Add Vista example

.gitattributes

@@ -0,0 +1,2 @@
+* text=auto eol=lf
+Cargo.lock linguist-generated=false

.github/pull_request_template.md

@@ -0,0 +1,13 @@
+<!--
+* Keep your PR:s small and focused.
+* The PR title is what ends up in the changelog, so make it descriptive!
+* If applicable, add a screenshot or gif.
+* Do NOT open PR:s from your `main` branch, as that makes it hard for maintainers to test and add commits to your PR.
+* Remember to run `cargo fmt` and `cargo clippy`.
+* Open the PR as a draft until you have self-reviewed it and it passes CI.
+* When you have addressed a PR comment, mark it as resolved.
+
+Please be patient!
+-->
+
+* Closes #ISSUE_NUMBER

.github/workflows/labels.yml

@@ -0,0 +1,27 @@
+# Copied from https://github.com/rerun-io/rerun_template
+
+# https://github.com/marketplace/actions/require-labels
+# Check for existence of labels
+# See all our labels at https://github.com/rerun-io/rerun/issues/labels
+
+name: PR Labels
+
+on:
+  pull_request:
+    types:
+      - opened
+      - synchronize
+      - reopened
+      - labeled
+      - unlabeled
+
+jobs:
+  label:
+    runs-on: ubuntu-latest
+    steps:
+      - name: Check for a "do-not-merge" label
+        uses: mheap/github-action-required-labels@v3
+        with:
+          mode: exactly
+          count: 0
+          labels: "do-not-merge"

.github/workflows/links.yml

@@ -0,0 +1,29 @@
+# Copied from https://github.com/rerun-io/rerun_template
+on: [push, pull_request]
+
+name: Link checker
+
+jobs:
+  link-checker:
+    name: Check links
+    runs-on: ubuntu-latest
+    steps:
+      - uses: actions/checkout@v4
+
+      - name: Restore link checker cache
+        uses: actions/cache@v3
+        with:
+          path: .lycheecache
+          key: cache-lychee-${{ github.sha }}
+          restore-keys: cache-lychee-
+
+      # Check https://github.com/lycheeverse/lychee on how to run locally.
+      - name: Link Checker
+        id: lychee
+        uses: lycheeverse/lychee-action@v1.9.0
+        with:
+          fail: true
+          lycheeVersion: "0.14.3"
+          # When given a directory, lychee checks only markdown, html and text files, everything else we have to glob in manually.
+          args: |
+            --base . --cache --max-cache-age 1d . "**/*.rs" "**/*.toml" "**/*.hpp" "**/*.cpp" "**/CMakeLists.txt" "**/*.py" "**/*.yml"

.github/workflows/python.yml

@@ -0,0 +1,21 @@
+# Copied from https://github.com/rerun-io/rerun_template
+# Disabled since this contains a lot of non-conforming code from the original repository
+on: []
+
+name: C++
+
+jobs:
+  python-check:
+    name: Python
+    runs-on: ubuntu-latest
+    steps:
+      - uses: actions/checkout@v4
+
+      - uses: prefix-dev/setup-pixi@v0.5.2
+        with:
+          pixi-version: v0.19.0
+          cache: true
+
+      - run: pixi run py-fmt-check
+
+      - run: pixi run py-lint

.github/workflows/typos.yml

@@ -0,0 +1,19 @@
+# Copied from https://github.com/rerun-io/rerun_template
+
+# https://github.com/crate-ci/typos
+# Add exceptions to `.typos.toml`
+# install and run locally: cargo install typos-cli && typos
+
+name: Spell Check
+on: [pull_request]
+
+jobs:
+  run:
+    name: Spell Check
+    runs-on: ubuntu-latest
+    steps:
+      - name: Checkout Actions Repository
+        uses: actions/checkout@v4
+
+      - name: Check spelling of entire workspace
+        uses: crate-ci/typos@master

.gitignore

@@ -0,0 +1,23 @@
+# Mac stuff:
+.DS_Store
+
+# C++ build directory
+build
+
+# Rust compile target directories:
+target
+target_ra
+target_wasm
+
+# https://github.com/lycheeverse/lychee
+.lycheecache
+
+# Pixi environment
+.pixi
+
+# Python stuff:
+__pycache__
+.mypy_cache
+.ruff_cache
+venv
+.python-version

.mypy.ini

@@ -0,0 +1,11 @@
+[mypy]
+files = .
+exclude = build
+namespace_packages = True
+show_error_codes = True
+strict = True
+enable_error_code = redundant-expr, truthy-bool, ignore-without-code
+; plugins = numpy.typing.mypy_plugin
+ignore_missing_imports = True
+no_implicit_reexport = False
+disallow_untyped_calls = False

.typos.toml

@@ -0,0 +1,6 @@
+# https://github.com/crate-ci/typos
+# install:  cargo install typos-cli
+# run:      typos
+
+[default.extend-words]
+teh = "teh"     # part of @teh-cmc

.vscode/extensions.json

@@ -0,0 +1,26 @@
+{
+  // See https://go.microsoft.com/fwlink/?LinkId=827846
+  // for the documentation about the extensions.json format
+  "recommendations": [
+    "charliermarsh.ruff",
+    "gaborv.flatbuffers",
+    "github.vscode-github-actions",
+    "josetr.cmake-language-support-vscode",
+    "ms-python.mypy-type-checker",
+    "ms-python.python",
+    "ms-vscode.cmake-tools",
+    "ms-vscode.cpptools-extension-pack",
+    "ms-vsliveshare.vsliveshare",
+    "polymeilex.wgsl",
+    "rust-lang.rust-analyzer",
+    "serayuzgur.crates",
+    "streetsidesoftware.code-spell-checker",
+    "tamasfe.even-better-toml",
+    "vadimcn.vscode-lldb",
+    "wayou.vscode-todo-highlight",
+    "webfreak.debug",
+    "xaver.clang-format", // C++ formatter
+    "zxh404.vscode-proto3",
+    "esbenp.prettier-vscode"
+  ]
+}

.vscode/launch.json

@@ -0,0 +1,56 @@
+{
+    // Use IntelliSense to learn about possible attributes.
+    // Hover to view descriptions of existing attributes.
+    // For more information, visit: https://go.microsoft.com/fwlink/?linkid=830387
+    "version": "0.2.0",
+    "configurations": [
+        // Python
+        {
+            "name": "Python Debugger: Current File",
+            "type": "debugpy",
+            "request": "launch",
+            "program": "${file}",
+            "console": "integratedTerminal"
+        },
+        // Rust:
+        {
+            "name": "Debug 'PROJ_NAME'",
+            "type": "lldb",
+            "request": "launch",
+            "cargo": {
+                "args": [
+                    "build"
+                ],
+                "filter": {
+                    "name": "PROJ_NAME",
+                    "kind": "bin"
+                }
+            },
+            "args": [],
+            "cwd": "${workspaceFolder}",
+            "env": {
+                "RUST_LOG": "debug"
+            }
+        },
+        {
+            "name": "Launch Rust tests",
+            "type": "lldb",
+            "request": "launch",
+            "cargo": {
+                "args": [
+                    "test",
+                    "--no-run",
+                    "--lib",
+                    "--all-features"
+                ],
+                "filter": {
+                    "kind": "lib"
+                }
+            },
+            "cwd": "${workspaceFolder}",
+            "env": {
+                "RUST_LOG": "debug"
+            }
+        },
+    ]
+}

.vscode/settings.json

@@ -0,0 +1,52 @@
+{
+    "editor.formatOnSave": true,
+    "editor.semanticTokenColorCustomizations": {
+        "rules": {
+            "*.unsafe:rust": "#eb5046"
+        }
+    },
+    "files.autoGuessEncoding": true,
+    "files.insertFinalNewline": true,
+    "files.trimTrailingWhitespace": true,
+    // don't share a cargo lock with rust-analyzer.
+    // see https://github.com/rerun-io/rerun/pull/519 for rationale
+    "rust-analyzer.check.overrideCommand": [
+        "cargo",
+        "clippy",
+        "--target-dir=target_ra",
+        "--workspace",
+        "--message-format=json",
+        "--all-targets",
+        "--all-features"
+    ],
+    "rust-analyzer.cargo.buildScripts.overrideCommand": [
+        "cargo",
+        "check",
+        "--quiet",
+        "--target-dir=target_ra",
+        "--workspace",
+        "--message-format=json",
+        "--all-targets",
+        "--all-features",
+    ],
+    // Our build scripts are generating code.
+    // Having Rust Analyzer do this while doing other builds can lead to catastrophic failures.
+    // INCLUDING attempts to publish a new release!
+    "rust-analyzer.cargo.buildScripts.enable": false,
+    "C_Cpp.default.configurationProvider": "ms-vscode.cmake-tools", // Use cmake-tools to grab configs.
+    "C_Cpp.autoAddFileAssociations": false,
+    "cmake.buildDirectory": "${workspaceRoot}/build/debug",
+    "cmake.generator": "Ninja", // Use Ninja, just like we do in our just/pixi command.
+    "rust-analyzer.showUnlinkedFileNotification": false,
+    "ruff.format.args": [
+        "--config=pyproject.toml"
+    ],
+    "ruff.lint.args": [
+        "--config=pyproject.toml"
+    ],
+    "prettier.requireConfig": true,
+    "prettier.configPath": ".prettierrc.toml",
+    "[python]": {
+        "editor.defaultFormatter": "charliermarsh.ruff"
+    },
+}

CODE_OF_CONDUCT.md

@@ -0,0 +1,132 @@
+# Contributor Covenant Code of Conduct
+
+## Our Pledge
+
+We as members, contributors, and leaders pledge to make participation in our
+community a harassment-free experience for everyone, regardless of age, body
+size, visible or invisible disability, ethnicity, sex characteristics, gender
+identity and expression, level of experience, education, socio-economic status,
+nationality, personal appearance, race, caste, color, religion, or sexual
+identity and orientation.
+
+We pledge to act and interact in ways that contribute to an open, welcoming,
+diverse, inclusive, and healthy community.
+
+## Our Standards
+
+Examples of behavior that contributes to a positive environment for our
+community include:
+
+* Demonstrating empathy and kindness toward other people
+* Being respectful of differing opinions, viewpoints, and experiences
+* Giving and gracefully accepting constructive feedback
+* Accepting responsibility and apologizing to those affected by our mistakes,
+  and learning from the experience
+* Focusing on what is best not just for us as individuals, but for the overall
+  community
+
+Examples of unacceptable behavior include:
+
+* The use of sexualized language or imagery, and sexual attention or advances of
+  any kind
+* Trolling, insulting or derogatory comments, and personal or political attacks
+* Public or private harassment
+* Publishing others' private information, such as a physical or email address,
+  without their explicit permission
+* Other conduct which could reasonably be considered inappropriate in a
+  professional setting
+
+## Enforcement Responsibilities
+
+Community leaders are responsible for clarifying and enforcing our standards of
+acceptable behavior and will take appropriate and fair corrective action in
+response to any behavior that they deem inappropriate, threatening, offensive,
+or harmful.
+
+Community leaders have the right and responsibility to remove, edit, or reject
+comments, commits, code, wiki edits, issues, and other contributions that are
+not aligned to this Code of Conduct, and will communicate reasons for moderation
+decisions when appropriate.
+
+## Scope
+
+This Code of Conduct applies within all community spaces, and also applies when
+an individual is officially representing the community in public spaces.
+Examples of representing our community include using an official e-mail address,
+posting via an official social media account, or acting as an appointed
+representative at an online or offline event.
+
+## Enforcement
+
+Instances of abusive, harassing, or otherwise unacceptable behavior may be
+reported to the community leaders responsible for enforcement at
+opensource@rerun.io.
+All complaints will be reviewed and investigated promptly and fairly.
+
+All community leaders are obligated to respect the privacy and security of the
+reporter of any incident.
+
+## Enforcement Guidelines
+
+Community leaders will follow these Community Impact Guidelines in determining
+the consequences for any action they deem in violation of this Code of Conduct:
+
+### 1. Correction
+
+**Community Impact**: Use of inappropriate language or other behavior deemed
+unprofessional or unwelcome in the community.
+
+**Consequence**: A private, written warning from community leaders, providing
+clarity around the nature of the violation and an explanation of why the
+behavior was inappropriate. A public apology may be requested.
+
+### 2. Warning
+
+**Community Impact**: A violation through a single incident or series of
+actions.
+
+**Consequence**: A warning with consequences for continued behavior. No
+interaction with the people involved, including unsolicited interaction with
+those enforcing the Code of Conduct, for a specified period of time. This
+includes avoiding interactions in community spaces as well as external channels
+like social media. Violating these terms may lead to a temporary or permanent
+ban.
+
+### 3. Temporary Ban
+
+**Community Impact**: A serious violation of community standards, including
+sustained inappropriate behavior.
+
+**Consequence**: A temporary ban from any sort of interaction or public
+communication with the community for a specified period of time. No public or
+private interaction with the people involved, including unsolicited interaction
+with those enforcing the Code of Conduct, is allowed during this period.
+Violating these terms may lead to a permanent ban.
+
+### 4. Permanent Ban
+
+**Community Impact**: Demonstrating a pattern of violation of community
+standards, including sustained inappropriate behavior, harassment of an
+individual, or aggression toward or disparagement of classes of individuals.
+
+**Consequence**: A permanent ban from any sort of public interaction within the
+community.
+
+## Attribution
+
+This Code of Conduct is adapted from the [Contributor Covenant][homepage],
+version 2.1, available at
+[https://www.contributor-covenant.org/version/2/1/code_of_conduct.html][v2.1].
+
+Community Impact Guidelines were inspired by
+[Mozilla's code of conduct enforcement ladder][Mozilla CoC].
+
+For answers to common questions about this code of conduct, see the FAQ at
+[https://www.contributor-covenant.org/faq][FAQ]. Translations are available at
+[https://www.contributor-covenant.org/translations][translations].
+
+[homepage]: https://www.contributor-covenant.org
+[v2.1]: https://www.contributor-covenant.org/version/2/1/code_of_conduct.html
+[Mozilla CoC]: https://github.com/mozilla/diversity
+[FAQ]: https://www.contributor-covenant.org/faq
+[translations]: https://www.contributor-covenant.org/translations

LICENSE-APACHE

@@ -0,0 +1,201 @@
+                              Apache License
+                        Version 2.0, January 2004
+                     http://www.apache.org/licenses/
+
+TERMS AND CONDITIONS FOR USE, REPRODUCTION, AND DISTRIBUTION
+
+1. Definitions.
+
+   "License" shall mean the terms and conditions for use, reproduction,
+   and distribution as defined by Sections 1 through 9 of this document.
+
+   "Licensor" shall mean the copyright owner or entity authorized by
+   the copyright owner that is granting the License.
+
+   "Legal Entity" shall mean the union of the acting entity and all
+   other entities that control, are controlled by, or are under common
+   control with that entity. For the purposes of this definition,
+   "control" means (i) the power, direct or indirect, to cause the
+   direction or management of such entity, whether by contract or
+   otherwise, or (ii) ownership of fifty percent (50%) or more of the
+   outstanding shares, or (iii) beneficial ownership of such entity.
+
+   "You" (or "Your") shall mean an individual or Legal Entity
+   exercising permissions granted by this License.
+
+   "Source" form shall mean the preferred form for making modifications,
+   including but not limited to software source code, documentation
+   source, and configuration files.
+
+   "Object" form shall mean any form resulting from mechanical
+   transformation or translation of a Source form, including but
+   not limited to compiled object code, generated documentation,
+   and conversions to other media types.
+
+   "Work" shall mean the work of authorship, whether in Source or
+   Object form, made available under the License, as indicated by a
+   copyright notice that is included in or attached to the work
+   (an example is provided in the Appendix below).
+
+   "Derivative Works" shall mean any work, whether in Source or Object
+   form, that is based on (or derived from) the Work and for which the
+   editorial revisions, annotations, elaborations, or other modifications
+   represent, as a whole, an original work of authorship. For the purposes
+   of this License, Derivative Works shall not include works that remain
+   separable from, or merely link (or bind by name) to the interfaces of,
+   the Work and Derivative Works thereof.
+
+   "Contribution" shall mean any work of authorship, including
+   the original version of the Work and any modifications or additions
+   to that Work or Derivative Works thereof, that is intentionally
+   submitted to Licensor for inclusion in the Work by the copyright owner
+   or by an individual or Legal Entity authorized to submit on behalf of
+   the copyright owner. For the purposes of this definition, "submitted"
+   means any form of electronic, verbal, or written communication sent
+   to the Licensor or its representatives, including but not limited to
+   communication on electronic mailing lists, source code control systems,
+   and issue tracking systems that are managed by, or on behalf of, the
+   Licensor for the purpose of discussing and improving the Work, but
+   excluding communication that is conspicuously marked or otherwise
+   designated in writing by the copyright owner as "Not a Contribution."
+
+   "Contributor" shall mean Licensor and any individual or Legal Entity
+   on behalf of whom a Contribution has been received by Licensor and
+   subsequently incorporated within the Work.
+
+2. Grant of Copyright License. Subject to the terms and conditions of
+   this License, each Contributor hereby grants to You a perpetual,
+   worldwide, non-exclusive, no-charge, royalty-free, irrevocable
+   copyright license to reproduce, prepare Derivative Works of,
+   publicly display, publicly perform, sublicense, and distribute the
+   Work and such Derivative Works in Source or Object form.
+
+3. Grant of Patent License. Subject to the terms and conditions of
+   this License, each Contributor hereby grants to You a perpetual,
+   worldwide, non-exclusive, no-charge, royalty-free, irrevocable
+   (except as stated in this section) patent license to make, have made,
+   use, offer to sell, sell, import, and otherwise transfer the Work,
+   where such license applies only to those patent claims licensable
+   by such Contributor that are necessarily infringed by their
+   Contribution(s) alone or by combination of their Contribution(s)
+   with the Work to which such Contribution(s) was submitted. If You
+   institute patent litigation against any entity (including a
+   cross-claim or counterclaim in a lawsuit) alleging that the Work
+   or a Contribution incorporated within the Work constitutes direct
+   or contributory patent infringement, then any patent licenses
+   granted to You under this License for that Work shall terminate
+   as of the date such litigation is filed.
+
+4. Redistribution. You may reproduce and distribute copies of the
+   Work or Derivative Works thereof in any medium, with or without
+   modifications, and in Source or Object form, provided that You
+   meet the following conditions:
+
+   (a) You must give any other recipients of the Work or
+       Derivative Works a copy of this License; and
+
+   (b) You must cause any modified files to carry prominent notices
+       stating that You changed the files; and
+
+   (c) You must retain, in the Source form of any Derivative Works
+       that You distribute, all copyright, patent, trademark, and
+       attribution notices from the Source form of the Work,
+       excluding those notices that do not pertain to any part of
+       the Derivative Works; and
+
+   (d) If the Work includes a "NOTICE" text file as part of its
+       distribution, then any Derivative Works that You distribute must
+       include a readable copy of the attribution notices contained
+       within such NOTICE file, excluding those notices that do not
+       pertain to any part of the Derivative Works, in at least one
+       of the following places: within a NOTICE text file distributed
+       as part of the Derivative Works; within the Source form or
+       documentation, if provided along with the Derivative Works; or,
+       within a display generated by the Derivative Works, if and
+       wherever such third-party notices normally appear. The contents
+       of the NOTICE file are for informational purposes only and
+       do not modify the License. You may add Your own attribution
+       notices within Derivative Works that You distribute, alongside
+       or as an addendum to the NOTICE text from the Work, provided
+       that such additional attribution notices cannot be construed
+       as modifying the License.
+
+   You may add Your own copyright statement to Your modifications and
+   may provide additional or different license terms and conditions
+   for use, reproduction, or distribution of Your modifications, or
+   for any such Derivative Works as a whole, provided Your use,
+   reproduction, and distribution of the Work otherwise complies with
+   the conditions stated in this License.
+
+5. Submission of Contributions. Unless You explicitly state otherwise,
+   any Contribution intentionally submitted for inclusion in the Work
+   by You to the Licensor shall be under the terms and conditions of
+   this License, without any additional terms or conditions.
+   Notwithstanding the above, nothing herein shall supersede or modify
+   the terms of any separate license agreement you may have executed
+   with Licensor regarding such Contributions.
+
+6. Trademarks. This License does not grant permission to use the trade
+   names, trademarks, service marks, or product names of the Licensor,
+   except as required for reasonable and customary use in describing the
+   origin of the Work and reproducing the content of the NOTICE file.
+
+7. Disclaimer of Warranty. Unless required by applicable law or
+   agreed to in writing, Licensor provides the Work (and each
+   Contributor provides its Contributions) on an "AS IS" BASIS,
+   WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or
+   implied, including, without limitation, any warranties or conditions
+   of TITLE, NON-INFRINGEMENT, MERCHANTABILITY, or FITNESS FOR A
+   PARTICULAR PURPOSE. You are solely responsible for determining the
+   appropriateness of using or redistributing the Work and assume any
+   risks associated with Your exercise of permissions under this License.
+
+8. Limitation of Liability. In no event and under no legal theory,
+   whether in tort (including negligence), contract, or otherwise,
+   unless required by applicable law (such as deliberate and grossly
+   negligent acts) or agreed to in writing, shall any Contributor be
+   liable to You for damages, including any direct, indirect, special,
+   incidental, or consequential damages of any character arising as a
+   result of this License or out of the use or inability to use the
+   Work (including but not limited to damages for loss of goodwill,
+   work stoppage, computer failure or malfunction, or any and all
+   other commercial damages or losses), even if such Contributor
+   has been advised of the possibility of such damages.
+
+9. Accepting Warranty or Additional Liability. While redistributing
+   the Work or Derivative Works thereof, You may choose to offer,
+   and charge a fee for, acceptance of support, warranty, indemnity,
+   or other liability obligations and/or rights consistent with this
+   License. However, in accepting such obligations, You may act only
+   on Your own behalf and on Your sole responsibility, not on behalf
+   of any other Contributor, and only if You agree to indemnify,
+   defend, and hold each Contributor harmless for any liability
+   incurred by, or claims asserted against, such Contributor by reason
+   of your accepting any such warranty or additional liability.
+
+END OF TERMS AND CONDITIONS
+
+APPENDIX: How to apply the Apache License to your work.
+
+   To apply the Apache License to your work, attach the following
+   boilerplate notice, with the fields enclosed by brackets "[]"
+   replaced with your own identifying information. (Don't include
+   the brackets!)  The text should be enclosed in the appropriate
+   comment syntax for the file format. We also recommend that a
+   file or class name and description of purpose be included on the
+   same "printed page" as the copyright notice for easier
+   identification within third-party archives.
+
+Copyright [yyyy] [name of copyright owner]
+
+Licensed under the Apache License, Version 2.0 (the "License");
+you may not use this file except in compliance with the License.
+You may obtain a copy of the License at
+
+    http://www.apache.org/licenses/LICENSE-2.0
+
+Unless required by applicable law or agreed to in writing, software
+distributed under the License is distributed on an "AS IS" BASIS,
+WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+See the License for the specific language governing permissions and
+limitations under the License.

LICENSE-MIT

@@ -0,0 +1,25 @@
+Copyright (c) 2024 Rerun Technologies AB <opensource@rerun.io>
+
+Permission is hereby granted, free of charge, to any
+person obtaining a copy of this software and associated
+documentation files (the "Software"), to deal in the
+Software without restriction, including without
+limitation the rights to use, copy, modify, merge,
+publish, distribute, sublicense, and/or sell copies of
+the Software, and to permit persons to whom the Software
+is furnished to do so, subject to the following
+conditions:
+
+The above copyright notice and this permission notice
+shall be included in all copies or substantial portions
+of the Software.
+
+THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF
+ANY KIND, EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED
+TO THE WARRANTIES OF MERCHANTABILITY, FITNESS FOR A
+PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT
+SHALL THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY
+CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION
+OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR
+IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER
+DEALINGS IN THE SOFTWARE.

README.md

@@ -0,0 +1,41 @@
+---
+title: Vista
+emoji: 🚗
+colorFrom: blue
+colorTo: green
+sdk: gradio
+sdk_version: 4.36.1
+app_file: app.py
+pinned: false
+license: apache-2.0
+---
+
+# Vista: A Generalizable Driving World Model with High Fidelity and Versatile Controllability
+
+https://github.com/rerun-io/hf-example-vista/assets/9785832/0b9a01ca-90a2-4b36-98fc-a7a7b378fd54
+
+[Shenyuan Gao](https://github.com/Little-Podi), [Jiazhi Yang](https://scholar.google.com/citations?user=Ju7nGX8AAAAJ&hl=en), [Li Chen](https://scholar.google.com/citations?user=ulZxvY0AAAAJ&hl=en), [Kashyap Chitta](https://kashyap7x.github.io/), [Yihang Qiu](https://scholar.google.com/citations?user=qgRUOdIAAAAJ&hl=en), [Andreas Geiger](https://www.cvlibs.net/), [Jun Zhang](https://eejzhang.people.ust.hk/), [Hongyang Li](https://lihongyang.info/)
+
+This is a demo of the [Vista model](https://github.com/OpenDriveLab/Vista), a driving world model that can be used to simulate a variety of driving scenarios. This demo uses [Rerun](https://rerun.io/)'s custom [gradio component](https://www.gradio.app/custom-components/gallery?id=radames%2Fgradio_rerun) to livestream the model's output and show intermediate results.
+
+[📜technical report](https://arxiv.org/abs/2405.17398),  [🎬video demos](https://vista-demo.github.io/),  [🤗model weights](https://huggingface.co/OpenDriveLab/Vista)
+
+Please refer to the [original repository](https://github.com/OpenDriveLab/Vista) for the original code base and README.
+
+You can try the example on Rerun's HuggingFace space [here](https://huggingface.co/spaces/rerun/Vista).
+
+## Run the example locally
+To run this example locally use the following command (you need a GPU with at least 20GB of memory, tested with an RTX 4090):
+```bash
+pixi run example
+```
+
+You can specify the first image, the number of generated segments, and the number of diffusion steps per segment:
+```bash
+pixi run example --img-path "example_images/streetview.png" --num-segments 10 --num-steps 100
+```
+
+To see other all options, use the following:
+```bash
+pixi run example --help
+```

app.py

@@ -0,0 +1,127 @@
+"""Gradio interface for Vista model."""
+from __future__ import annotations
+
+import glob
+import os
+import queue
+import threading
+
+import gradio as gr
+import gradio_rerun
+import rerun as rr
+import spaces
+
+import vista
+
+
+@spaces.GPU(duration=400)
+@rr.thread_local_stream("Vista")
+def generate_gradio(
+    first_frame_file_name: str,
+    n_rounds: float=3,
+    n_steps: float=10,
+    height=576,
+    width=1024,
+    n_frames=25,
+    cfg_scale=2.5,
+    cond_aug=0.0,
+):
+    global model
+
+    n_rounds = int(n_rounds)
+    n_steps = int(n_steps)
+
+    # Use a queue to log immediately from internals
+    log_queue = queue.SimpleQueue()
+
+    stream = rr.binary_stream()
+
+    blueprint = vista.generate_blueprint(n_rounds)
+    rr.send_blueprint(blueprint)
+    yield stream.read()
+
+    handle = threading.Thread(
+        target=vista.run_sampling,
+        args=[
+            log_queue,
+            first_frame_file_name,
+            height,
+            width,
+            n_rounds,
+            n_frames,
+            n_steps,
+            cfg_scale,
+            cond_aug,
+            model,
+        ],
+    )
+    handle.start()
+    while True:
+        msg = log_queue.get()
+        if msg == "done":
+            break
+        else:
+            entity_path, entity, times = msg
+            rr.reset_time()
+            for timeline, time in times:
+                if isinstance(time, int):
+                    rr.set_time_sequence(timeline, time)
+                else:
+                    rr.set_time_seconds(timeline, time)
+            rr.log(entity_path, entity)
+            yield stream.read()
+    handle.join()
+
+
+model = vista.create_model()
+
+with gr.Blocks(css="style.css") as demo:
+    gr.Markdown(
+        """
+        # Vista: A Generalizable Driving World Model with High Fidelity and Versatile Controllability
+
+        [Shenyuan Gao](https://github.com/Little-Podi), [Jiazhi Yang](https://scholar.google.com/citations?user=Ju7nGX8AAAAJ&hl=en), [Li Chen](https://scholar.google.com/citations?user=ulZxvY0AAAAJ&hl=en), [Kashyap Chitta](https://kashyap7x.github.io/), [Yihang Qiu](https://scholar.google.com/citations?user=qgRUOdIAAAAJ&hl=en), [Andreas Geiger](https://www.cvlibs.net/), [Jun Zhang](https://eejzhang.people.ust.hk/), [Hongyang Li](https://lihongyang.info/)
+
+        This is a demo of the [Vista model](https://github.com/OpenDriveLab/Vista), a driving world model that can be used to simulate a variety of driving scenarios. This demo uses [Rerun](https://rerun.io/)'s custom [gradio component](https://www.gradio.app/custom-components/gallery?id=radames%2Fgradio_rerun) to livestream the model's output and show intermediate results.
+
+         [📜technical report](https://arxiv.org/abs/2405.17398),  [🎬video demos](https://vista-demo.github.io/),  [🤗model weights](https://huggingface.co/OpenDriveLab/Vista)
+
+         Note that the GPU time is limited to 400 seconds per run. If you need more time, you can run the model locally or on your own server.
+        """
+    )
+    first_frame = gr.Image(sources="upload", type="filepath")
+    example_dir_path = os.path.join(os.path.dirname(__file__), "example_images")
+    example_file_paths = sorted(glob.glob(os.path.join(example_dir_path, "*.*")))
+    example_gallery = gr.Examples(
+        examples=example_file_paths,
+        inputs=first_frame,
+        cache_examples=False,
+    )
+
+    btn = gr.Button("Generate video")
+    num_rounds = gr.Slider(
+        label="Segments",
+        info="Number of 25 frame segments to generate. Higher values lead to longer videos. Try to keep the product of segments and steps below 30 to avoid running out of time.",
+        minimum=1,
+        maximum=5,
+        value=2,
+        step=1
+    )
+    num_steps = gr.Slider(
+        label="Diffusion Steps",
+        info="Number of diffusion steps per segment. Higher values lead to more detailed videos. Try to keep the product of segments and steps below 30 to avoid running out of time.",
+        minimum=1,
+        maximum=50,
+        value=15,
+        step=1
+    )
+
+    with gr.Row():
+        viewer = gradio_rerun.Rerun(streaming=True)
+    btn.click(
+        generate_gradio,
+        inputs=[first_frame, num_rounds, num_steps],
+        outputs=[viewer],
+    )
+
+demo.launch()

lychee.toml

@@ -0,0 +1,82 @@
+# Copied from https://github.com/rerun-io/rerun_template
+
+################################################################################
+# Config for the link checker lychee.
+#
+# Download & learn more at:
+# https://github.com/lycheeverse/lychee
+#
+# Example config:
+# https://github.com/lycheeverse/lychee/blob/master/lychee.example.toml
+#
+# Run `lychee . --dump` to list all found links that are being checked.
+#
+# Note that by default lychee will only check markdown and html files,
+# to check any other files you have to point to them explicitly, e.g.:
+# `lychee **/*.rs`
+# To make things worse, `exclude_path` is ignored for these globs,
+# so local runs with lots of gitignored files will be slow.
+# (https://github.com/lycheeverse/lychee/issues/1405)
+#
+# This unfortunately doesn't list anything for non-glob checks.
+################################################################################
+
+# Maximum number of concurrent link checks.
+# Workaround for "too many open files" error on MacOS, see https://github.com/lycheeverse/lychee/issues/1248
+max_concurrency = 32
+
+# Check links inside `<code>` and `<pre>` blocks as well as Markdown code blocks.
+include_verbatim = true
+
+# Proceed for server connections considered insecure (invalid TLS).
+insecure = true
+
+# Exclude these filesystem paths from getting checked.
+exclude_path = [
+  # Unfortunately lychee doesn't yet read .gitignore https://github.com/lycheeverse/lychee/issues/1331
+  # The following entries are there because of that:
+  ".git",
+  "__pycache__",
+  "_deps/",
+  ".pixi",
+  "build",
+  "target_ra",
+  "target_wasm",
+  "target",
+  "venv",
+]
+
+# Exclude URLs and mail addresses from checking (supports regex).
+exclude = [
+  # Skip speculative links
+  '.*?speculative-link',
+
+  # Strings with replacements.
+  '/__VIEWER_VERSION__/', # Replacement variable __VIEWER_VERSION__.
+  '/\$',                  # Replacement variable $.
+  '/GIT_HASH/',           # Replacement variable GIT_HASH.
+  '\{\}',                 # Ignore links with string interpolation.
+  '\$relpath\^',          # Relative paths as used by rerun_cpp's doc header.
+  '%7B.+%7D',             # Ignore strings that look like ready to use links but contain a replacement strings. The URL escaping is for '{.+}' (this seems to be needed for html embedded urls since lychee assumes they use this encoding).
+  '%7B%7D',               # Ignore links with string interpolation, escaped variant.
+
+  # Local links that require further setup.
+  'http://127.0.0.1',
+  'http://localhost',
+  'recording:/',      # rrd recording link.
+  'ws:/',
+  're_viewer.js',     # Build artifact that html is linking to.
+
+  # Api endpoints.
+  'https://fonts.googleapis.com/', # Font API entrypoint, not a link.
+  'https://fonts.gstatic.com/',    # Font API entrypoint, not a link.
+  'https://tel.rerun.io/',         # Analytics endpoint.
+
+  # Avoid rate limiting.
+  'https://crates.io/crates/.*',                  # Avoid crates.io rate-limiting
+  'https://github.com/rerun-io/rerun/commit/\.*', # Ignore links to our own commits (typically in changelog).
+  'https://github.com/rerun-io/rerun/pull/\.*',   # Ignore links to our own pull requests (typically in changelog).
+
+  # Used in rerun_template repo until the user search-replaces `new_repo_name`
+  'https://github.com/rerun-io/new_repo_name',
+]

main.py

@@ -0,0 +1,87 @@
+"""Command line interface for generating videos from the model."""
+from __future__ import annotations
+
+import argparse
+import queue
+import threading
+
+import rerun as rr
+
+import vista
+
+
+def generate_local(
+    first_frame_file_name: str,
+    height=576,
+    width=1024,
+    n_rounds=4,
+    n_frames=25,
+    n_steps=10,
+    cfg_scale=2.5,
+    cond_aug=0.0,
+):
+    # Use a queue to log immediately from internals
+    log_queue = queue.SimpleQueue()
+
+    handle = threading.Thread(
+        target=vista.run_sampling,
+        args=[
+            log_queue,
+            first_frame_file_name,
+            height,
+            width,
+            n_rounds,
+            n_frames,
+            n_steps,
+            cfg_scale,
+            cond_aug,
+        ],
+    )
+    handle.start()
+    while True:
+        msg = log_queue.get()
+        if msg == "done":
+            break
+        else:
+            entity_path, entity, times = msg
+            rr.reset_time()
+            for timeline, time in times:
+                if isinstance(time, int):
+                    rr.set_time_sequence(timeline, time)
+                else:
+                    rr.set_time_seconds(timeline, time)
+            rr.log(entity_path, entity)
+    handle.join()
+
+
+if __name__ == "__main__":
+    parser = argparse.ArgumentParser(
+        description="Generate video conditioned on a single image using the Vista model."
+    )
+    parser.add_argument(
+        "--img-path",
+        type=str,
+        help="Path to image used as input for Canny edge detector.",
+        default="./example_images/nus-0.jpg",
+    )
+    parser.add_argument(
+        "--num-steps",
+        type=int,
+        help="Number of diffusion steps per image. Recommended range: 10-50. Higher values result in more detailed images and less blurry results.",
+        default=20,
+    )
+    parser.add_argument(
+        "--num-segments",
+        type=int,
+        help="Number of segments to generate. Each segment consists of 25 frames.",
+        default=3,
+    )
+    rr.script_add_args(parser)
+    args = parser.parse_args()
+    rr.script_setup(
+        args,
+        "rerun_example_vista",
+        default_blueprint=vista.generate_blueprint(args.num_segments),
+    )
+
+    generate_local(args.img_path, n_steps=args.num_steps, n_rounds=args.num_segments)

pixi.lock

@@ -0,0 +1,2130 @@
+version: 5
+environments:
+  default:
+    channels:
+    - url: https://conda.anaconda.org/conda-forge/
+    packages:
+      linux-64:
+      - conda: https://conda.anaconda.org/conda-forge/linux-64/_libgcc_mutex-0.1-conda_forge.tar.bz2
+      - conda: https://conda.anaconda.org/conda-forge/linux-64/_openmp_mutex-4.5-2_gnu.tar.bz2
+      - conda: https://conda.anaconda.org/conda-forge/linux-64/brotli-python-1.1.0-py310hc6cd4ac_1.conda
+      - conda: https://conda.anaconda.org/conda-forge/linux-64/bzip2-1.0.8-hd590300_5.conda
+      - conda: https://conda.anaconda.org/conda-forge/linux-64/ca-certificates-2024.2.2-hbcca054_0.conda
+      - conda: https://conda.anaconda.org/conda-forge/linux-64/ld_impl_linux-64-2.40-h41732ed_0.conda
+      - conda: https://conda.anaconda.org/conda-forge/linux-64/libffi-3.4.2-h7f98852_5.tar.bz2
+      - conda: https://conda.anaconda.org/conda-forge/linux-64/libgcc-ng-13.2.0-h807b86a_5.conda
+      - conda: https://conda.anaconda.org/conda-forge/linux-64/libgomp-13.2.0-h807b86a_5.conda
+      - conda: https://conda.anaconda.org/conda-forge/linux-64/libnsl-2.0.1-hd590300_0.conda
+      - conda: https://conda.anaconda.org/conda-forge/linux-64/libsqlite-3.45.2-h2797004_0.conda
+      - conda: https://conda.anaconda.org/conda-forge/linux-64/libstdcxx-ng-13.2.0-h7e041cc_5.conda
+      - conda: https://conda.anaconda.org/conda-forge/linux-64/libuuid-2.38.1-h0b41bf4_0.conda
+      - conda: https://conda.anaconda.org/conda-forge/linux-64/libxcrypt-4.4.36-hd590300_1.conda
+      - conda: https://conda.anaconda.org/conda-forge/linux-64/libzlib-1.2.13-hd590300_5.conda
+      - conda: https://conda.anaconda.org/conda-forge/linux-64/mypy-1.8.0-py310h2372a71_0.conda
+      - conda: https://conda.anaconda.org/conda-forge/noarch/mypy_extensions-1.0.0-pyha770c72_0.conda
+      - conda: https://conda.anaconda.org/conda-forge/linux-64/ncurses-6.4.20240210-h59595ed_0.conda
+      - conda: https://conda.anaconda.org/conda-forge/linux-64/openssl-3.2.1-hd590300_1.conda
+      - conda: https://conda.anaconda.org/conda-forge/noarch/pip-24.0-pyhd8ed1ab_0.conda
+      - conda: https://conda.anaconda.org/conda-forge/linux-64/psutil-6.0.0-py310hc51659f_0.conda
+      - conda: https://conda.anaconda.org/conda-forge/noarch/pysocks-1.7.1-pyha2e5f31_6.tar.bz2
+      - conda: https://conda.anaconda.org/conda-forge/linux-64/python-3.10.14-hd12c33a_0_cpython.conda
+      - conda: https://conda.anaconda.org/conda-forge/linux-64/python_abi-3.10-4_cp310.conda
+      - conda: https://conda.anaconda.org/conda-forge/linux-64/readline-8.2-h8228510_1.conda
+      - conda: https://conda.anaconda.org/conda-forge/linux-64/ruff-0.3.7-py310h3d77a66_0.conda
+      - conda: https://conda.anaconda.org/conda-forge/noarch/setuptools-70.1.1-pyhd8ed1ab_0.conda
+      - conda: https://conda.anaconda.org/conda-forge/linux-64/tk-8.6.13-noxft_h4845f30_101.conda
+      - conda: https://conda.anaconda.org/conda-forge/noarch/tomli-2.0.1-pyhd8ed1ab_0.tar.bz2
+      - conda: https://conda.anaconda.org/conda-forge/noarch/types-requests-2.31.0.20240406-pyhd8ed1ab_0.conda
+      - conda: https://conda.anaconda.org/conda-forge/noarch/typing_extensions-4.11.0-pyha770c72_0.conda
+      - conda: https://conda.anaconda.org/conda-forge/linux-64/typos-1.20.9-he8a937b_0.conda
+      - conda: https://conda.anaconda.org/conda-forge/noarch/tzdata-2024a-h0c530f3_0.conda
+      - conda: https://conda.anaconda.org/conda-forge/noarch/urllib3-2.2.1-pyhd8ed1ab_0.conda
+      - conda: https://conda.anaconda.org/conda-forge/noarch/wheel-0.43.0-pyhd8ed1ab_1.conda
+      - conda: https://conda.anaconda.org/conda-forge/linux-64/xz-5.2.6-h166bdaf_0.tar.bz2
+      linux-aarch64:
+      - conda: https://conda.anaconda.org/conda-forge/linux-aarch64/_openmp_mutex-4.5-2_gnu.tar.bz2
+      - conda: https://conda.anaconda.org/conda-forge/linux-aarch64/brotli-python-1.1.0-py310hbb3657e_1.conda
+      - conda: https://conda.anaconda.org/conda-forge/linux-aarch64/bzip2-1.0.8-h31becfc_5.conda
+      - conda: https://conda.anaconda.org/conda-forge/linux-aarch64/ca-certificates-2024.2.2-hcefe29a_0.conda
+      - conda: https://conda.anaconda.org/conda-forge/linux-aarch64/ld_impl_linux-aarch64-2.40-h2d8c526_0.conda
+      - conda: https://conda.anaconda.org/conda-forge/linux-aarch64/libffi-3.4.2-h3557bc0_5.tar.bz2
+      - conda: https://conda.anaconda.org/conda-forge/linux-aarch64/libgcc-ng-13.2.0-hf8544c7_5.conda
+      - conda: https://conda.anaconda.org/conda-forge/linux-aarch64/libgomp-13.2.0-hf8544c7_5.conda
+      - conda: https://conda.anaconda.org/conda-forge/linux-aarch64/libnsl-2.0.1-h31becfc_0.conda
+      - conda: https://conda.anaconda.org/conda-forge/linux-aarch64/libsqlite-3.45.2-h194ca79_0.conda
+      - conda: https://conda.anaconda.org/conda-forge/linux-aarch64/libstdcxx-ng-13.2.0-h9a76618_5.conda
+      - conda: https://conda.anaconda.org/conda-forge/linux-aarch64/libuuid-2.38.1-hb4cce97_0.conda
+      - conda: https://conda.anaconda.org/conda-forge/linux-aarch64/libxcrypt-4.4.36-h31becfc_1.conda
+      - conda: https://conda.anaconda.org/conda-forge/linux-aarch64/libzlib-1.2.13-h31becfc_5.conda
+      - conda: https://conda.anaconda.org/conda-forge/linux-aarch64/mypy-1.8.0-py310hb299538_0.conda
+      - conda: https://conda.anaconda.org/conda-forge/noarch/mypy_extensions-1.0.0-pyha770c72_0.conda
+      - conda: https://conda.anaconda.org/conda-forge/linux-aarch64/ncurses-6.4.20240210-h0425590_0.conda
+      - conda: https://conda.anaconda.org/conda-forge/linux-aarch64/openssl-3.2.1-h31becfc_1.conda
+      - conda: https://conda.anaconda.org/conda-forge/noarch/pip-24.0-pyhd8ed1ab_0.conda
+      - conda: https://conda.anaconda.org/conda-forge/linux-aarch64/psutil-6.0.0-py310hb52b2da_0.conda
+      - conda: https://conda.anaconda.org/conda-forge/noarch/pysocks-1.7.1-pyha2e5f31_6.tar.bz2
+      - conda: https://conda.anaconda.org/conda-forge/linux-aarch64/python-3.10.14-hbbe8eec_0_cpython.conda
+      - conda: https://conda.anaconda.org/conda-forge/linux-aarch64/python_abi-3.10-4_cp310.conda
+      - conda: https://conda.anaconda.org/conda-forge/linux-aarch64/readline-8.2-h8fc344f_1.conda
+      - conda: https://conda.anaconda.org/conda-forge/linux-aarch64/ruff-0.3.7-py310hf6424b7_0.conda
+      - conda: https://conda.anaconda.org/conda-forge/noarch/setuptools-70.1.1-pyhd8ed1ab_0.conda
+      - conda: https://conda.anaconda.org/conda-forge/linux-aarch64/tk-8.6.13-h194ca79_0.conda
+      - conda: https://conda.anaconda.org/conda-forge/noarch/tomli-2.0.1-pyhd8ed1ab_0.tar.bz2
+      - conda: https://conda.anaconda.org/conda-forge/noarch/types-requests-2.31.0.20240406-pyhd8ed1ab_0.conda
+      - conda: https://conda.anaconda.org/conda-forge/noarch/typing_extensions-4.11.0-pyha770c72_0.conda
+      - conda: https://conda.anaconda.org/conda-forge/linux-aarch64/typos-1.20.9-h1d8f897_0.conda
+      - conda: https://conda.anaconda.org/conda-forge/noarch/tzdata-2024a-h0c530f3_0.conda
+      - conda: https://conda.anaconda.org/conda-forge/noarch/urllib3-2.2.1-pyhd8ed1ab_0.conda
+      - conda: https://conda.anaconda.org/conda-forge/noarch/wheel-0.43.0-pyhd8ed1ab_1.conda
+      - conda: https://conda.anaconda.org/conda-forge/linux-aarch64/xz-5.2.6-h9cdd2b7_0.tar.bz2
+      osx-64:
+      - conda: https://conda.anaconda.org/conda-forge/osx-64/brotli-python-1.1.0-py310h9e9d8ca_1.conda
+      - conda: https://conda.anaconda.org/conda-forge/osx-64/bzip2-1.0.8-h10d778d_5.conda
+      - conda: https://conda.anaconda.org/conda-forge/osx-64/ca-certificates-2024.2.2-h8857fd0_0.conda
+      - conda: https://conda.anaconda.org/conda-forge/osx-64/libcxx-16.0.6-hd57cbcb_0.conda
+      - conda: https://conda.anaconda.org/conda-forge/osx-64/libffi-3.4.2-h0d85af4_5.tar.bz2
+      - conda: https://conda.anaconda.org/conda-forge/osx-64/libsqlite-3.45.2-h92b6c6a_0.conda
+      - conda: https://conda.anaconda.org/conda-forge/osx-64/libzlib-1.2.13-h8a1eda9_5.conda
+      - conda: https://conda.anaconda.org/conda-forge/osx-64/mypy-1.8.0-py310hb372a2b_0.conda
+      - conda: https://conda.anaconda.org/conda-forge/noarch/mypy_extensions-1.0.0-pyha770c72_0.conda
+      - conda: https://conda.anaconda.org/conda-forge/osx-64/ncurses-6.4.20240210-h73e2aa4_0.conda
+      - conda: https://conda.anaconda.org/conda-forge/osx-64/openssl-3.2.1-hd75f5a5_1.conda
+      - conda: https://conda.anaconda.org/conda-forge/noarch/pip-24.0-pyhd8ed1ab_0.conda
+      - conda: https://conda.anaconda.org/conda-forge/osx-64/psutil-6.0.0-py310h936d840_0.conda
+      - conda: https://conda.anaconda.org/conda-forge/noarch/pysocks-1.7.1-pyha2e5f31_6.tar.bz2
+      - conda: https://conda.anaconda.org/conda-forge/osx-64/python-3.10.14-h00d2728_0_cpython.conda
+      - conda: https://conda.anaconda.org/conda-forge/osx-64/python_abi-3.10-4_cp310.conda
+      - conda: https://conda.anaconda.org/conda-forge/osx-64/readline-8.2-h9e318b2_1.conda
+      - conda: https://conda.anaconda.org/conda-forge/osx-64/ruff-0.3.7-py310hdac29b7_0.conda
+      - conda: https://conda.anaconda.org/conda-forge/noarch/setuptools-70.1.1-pyhd8ed1ab_0.conda
+      - conda: https://conda.anaconda.org/conda-forge/osx-64/tk-8.6.13-h1abcd95_1.conda
+      - conda: https://conda.anaconda.org/conda-forge/noarch/tomli-2.0.1-pyhd8ed1ab_0.tar.bz2
+      - conda: https://conda.anaconda.org/conda-forge/noarch/types-requests-2.31.0.20240406-pyhd8ed1ab_0.conda
+      - conda: https://conda.anaconda.org/conda-forge/noarch/typing_extensions-4.11.0-pyha770c72_0.conda
+      - conda: https://conda.anaconda.org/conda-forge/osx-64/typos-1.20.9-h11a7dfb_0.conda
+      - conda: https://conda.anaconda.org/conda-forge/noarch/tzdata-2024a-h0c530f3_0.conda
+      - conda: https://conda.anaconda.org/conda-forge/noarch/urllib3-2.2.1-pyhd8ed1ab_0.conda
+      - conda: https://conda.anaconda.org/conda-forge/noarch/wheel-0.43.0-pyhd8ed1ab_1.conda
+      - conda: https://conda.anaconda.org/conda-forge/osx-64/xz-5.2.6-h775f41a_0.tar.bz2
+      osx-arm64:
+      - conda: https://conda.anaconda.org/conda-forge/osx-arm64/brotli-python-1.1.0-py310h1253130_1.conda
+      - conda: https://conda.anaconda.org/conda-forge/osx-arm64/bzip2-1.0.8-h93a5062_5.conda
+      - conda: https://conda.anaconda.org/conda-forge/osx-arm64/ca-certificates-2024.2.2-hf0a4a13_0.conda
+      - conda: https://conda.anaconda.org/conda-forge/osx-arm64/libcxx-16.0.6-h4653b0c_0.conda
+      - conda: https://conda.anaconda.org/conda-forge/osx-arm64/libffi-3.4.2-h3422bc3_5.tar.bz2
+      - conda: https://conda.anaconda.org/conda-forge/osx-arm64/libsqlite-3.45.2-h091b4b1_0.conda
+      - conda: https://conda.anaconda.org/conda-forge/osx-arm64/libzlib-1.2.13-h53f4e23_5.conda
+      - conda: https://conda.anaconda.org/conda-forge/osx-arm64/mypy-1.8.0-py310hd125d64_0.conda
+      - conda: https://conda.anaconda.org/conda-forge/noarch/mypy_extensions-1.0.0-pyha770c72_0.conda
+      - conda: https://conda.anaconda.org/conda-forge/osx-arm64/ncurses-6.4.20240210-h078ce10_0.conda
+      - conda: https://conda.anaconda.org/conda-forge/osx-arm64/openssl-3.2.1-h0d3ecfb_1.conda
+      - conda: https://conda.anaconda.org/conda-forge/noarch/pip-24.0-pyhd8ed1ab_0.conda
+      - conda: https://conda.anaconda.org/conda-forge/osx-arm64/psutil-6.0.0-py310ha6dd24b_0.conda
+      - conda: https://conda.anaconda.org/conda-forge/noarch/pysocks-1.7.1-pyha2e5f31_6.tar.bz2
+      - conda: https://conda.anaconda.org/conda-forge/osx-arm64/python-3.10.14-h2469fbe_0_cpython.conda
+      - conda: https://conda.anaconda.org/conda-forge/osx-arm64/python_abi-3.10-4_cp310.conda
+      - conda: https://conda.anaconda.org/conda-forge/osx-arm64/readline-8.2-h92ec313_1.conda
+      - conda: https://conda.anaconda.org/conda-forge/osx-arm64/ruff-0.3.7-py310h81561d7_0.conda
+      - conda: https://conda.anaconda.org/conda-forge/noarch/setuptools-70.1.1-pyhd8ed1ab_0.conda
+      - conda: https://conda.anaconda.org/conda-forge/osx-arm64/tk-8.6.13-h5083fa2_1.conda
+      - conda: https://conda.anaconda.org/conda-forge/noarch/tomli-2.0.1-pyhd8ed1ab_0.tar.bz2
+      - conda: https://conda.anaconda.org/conda-forge/noarch/types-requests-2.31.0.20240406-pyhd8ed1ab_0.conda
+      - conda: https://conda.anaconda.org/conda-forge/noarch/typing_extensions-4.11.0-pyha770c72_0.conda
+      - conda: https://conda.anaconda.org/conda-forge/osx-arm64/typos-1.20.9-h5ef7bb8_0.conda
+      - conda: https://conda.anaconda.org/conda-forge/noarch/tzdata-2024a-h0c530f3_0.conda
+      - conda: https://conda.anaconda.org/conda-forge/noarch/urllib3-2.2.1-pyhd8ed1ab_0.conda
+      - conda: https://conda.anaconda.org/conda-forge/noarch/wheel-0.43.0-pyhd8ed1ab_1.conda
+      - conda: https://conda.anaconda.org/conda-forge/osx-arm64/xz-5.2.6-h57fd34a_0.tar.bz2
+      win-64:
+      - conda: https://conda.anaconda.org/conda-forge/win-64/brotli-python-1.1.0-py310h00ffb61_1.conda
+      - conda: https://conda.anaconda.org/conda-forge/win-64/bzip2-1.0.8-hcfcfb64_5.conda
+      - conda: https://conda.anaconda.org/conda-forge/win-64/ca-certificates-2024.2.2-h56e8100_0.conda
+      - conda: https://conda.anaconda.org/conda-forge/win-64/libffi-3.4.2-h8ffe710_5.tar.bz2
+      - conda: https://conda.anaconda.org/conda-forge/win-64/libsqlite-3.45.2-hcfcfb64_0.conda
+      - conda: https://conda.anaconda.org/conda-forge/win-64/libzlib-1.2.13-hcfcfb64_5.conda
+      - conda: https://conda.anaconda.org/conda-forge/win-64/m2w64-gcc-libgfortran-5.3.0-6.tar.bz2
+      - conda: https://conda.anaconda.org/conda-forge/win-64/m2w64-gcc-libs-5.3.0-7.tar.bz2
+      - conda: https://conda.anaconda.org/conda-forge/win-64/m2w64-gcc-libs-core-5.3.0-7.tar.bz2
+      - conda: https://conda.anaconda.org/conda-forge/win-64/m2w64-gmp-6.1.0-2.tar.bz2
+      - conda: https://conda.anaconda.org/conda-forge/win-64/m2w64-libwinpthread-git-5.0.0.4634.697f757-2.tar.bz2
+      - conda: https://conda.anaconda.org/conda-forge/win-64/msys2-conda-epoch-20160418-1.tar.bz2
+      - conda: https://conda.anaconda.org/conda-forge/win-64/mypy-1.8.0-py310h8d17308_0.conda
+      - conda: https://conda.anaconda.org/conda-forge/noarch/mypy_extensions-1.0.0-pyha770c72_0.conda
+      - conda: https://conda.anaconda.org/conda-forge/win-64/openssl-3.2.1-hcfcfb64_1.conda
+      - conda: https://conda.anaconda.org/conda-forge/noarch/pip-24.0-pyhd8ed1ab_0.conda
+      - conda: https://conda.anaconda.org/conda-forge/win-64/psutil-6.0.0-py310ha8f682b_0.conda
+      - conda: https://conda.anaconda.org/conda-forge/noarch/pysocks-1.7.1-pyh0701188_6.tar.bz2
+      - conda: https://conda.anaconda.org/conda-forge/win-64/python-3.10.14-h4de0772_0_cpython.conda
+      - conda: https://conda.anaconda.org/conda-forge/win-64/python_abi-3.10-4_cp310.conda
+      - conda: https://conda.anaconda.org/conda-forge/win-64/ruff-0.3.7-py310h298983d_0.conda
+      - conda: https://conda.anaconda.org/conda-forge/noarch/setuptools-70.1.1-pyhd8ed1ab_0.conda
+      - conda: https://conda.anaconda.org/conda-forge/win-64/tk-8.6.13-h5226925_1.conda
+      - conda: https://conda.anaconda.org/conda-forge/noarch/tomli-2.0.1-pyhd8ed1ab_0.tar.bz2
+      - conda: https://conda.anaconda.org/conda-forge/noarch/types-requests-2.31.0.20240406-pyhd8ed1ab_0.conda
+      - conda: https://conda.anaconda.org/conda-forge/noarch/typing_extensions-4.11.0-pyha770c72_0.conda
+      - conda: https://conda.anaconda.org/conda-forge/win-64/typos-1.20.9-h7f3b576_0.conda
+      - conda: https://conda.anaconda.org/conda-forge/noarch/tzdata-2024a-h0c530f3_0.conda
+      - conda: https://conda.anaconda.org/conda-forge/win-64/ucrt-10.0.22621.0-h57928b3_0.tar.bz2
+      - conda: https://conda.anaconda.org/conda-forge/noarch/urllib3-2.2.1-pyhd8ed1ab_0.conda
+      - conda: https://conda.anaconda.org/conda-forge/win-64/vc-14.3-hcf57466_18.conda
+      - conda: https://conda.anaconda.org/conda-forge/win-64/vc14_runtime-14.38.33130-h82b7239_18.conda
+      - conda: https://conda.anaconda.org/conda-forge/win-64/vs2015_runtime-14.38.33130-hcb4865c_18.conda
+      - conda: https://conda.anaconda.org/conda-forge/noarch/wheel-0.43.0-pyhd8ed1ab_1.conda
+      - conda: https://conda.anaconda.org/conda-forge/noarch/win_inet_pton-1.1.0-pyhd8ed1ab_6.tar.bz2
+      - conda: https://conda.anaconda.org/conda-forge/win-64/xz-5.2.6-h8d14728_0.tar.bz2
+packages:
+- kind: conda
+  name: _libgcc_mutex
+  version: '0.1'
+  build: conda_forge
+  subdir: linux-64
+  url: https://conda.anaconda.org/conda-forge/linux-64/_libgcc_mutex-0.1-conda_forge.tar.bz2
+  sha256: fe51de6107f9edc7aa4f786a70f4a883943bc9d39b3bb7307c04c41410990726
+  md5: d7c89558ba9fa0495403155b64376d81
+  license: None
+  size: 2562
+  timestamp: 1578324546067
+- kind: conda
+  name: _openmp_mutex
+  version: '4.5'
+  build: 2_gnu
+  build_number: 16
+  subdir: linux-64
+  url: https://conda.anaconda.org/conda-forge/linux-64/_openmp_mutex-4.5-2_gnu.tar.bz2
+  sha256: fbe2c5e56a653bebb982eda4876a9178aedfc2b545f25d0ce9c4c0b508253d22
+  md5: 73aaf86a425cc6e73fcf236a5a46396d
+  depends:
+  - _libgcc_mutex 0.1 conda_forge
+  - libgomp >=7.5.0
+  constrains:
+  - openmp_impl 9999
+  license: BSD-3-Clause
+  license_family: BSD
+  size: 23621
+  timestamp: 1650670423406
+- kind: conda
+  name: _openmp_mutex
+  version: '4.5'
+  build: 2_gnu
+  build_number: 16
+  subdir: linux-aarch64
+  url: https://conda.anaconda.org/conda-forge/linux-aarch64/_openmp_mutex-4.5-2_gnu.tar.bz2
+  sha256: 3702bef2f0a4d38bd8288bbe54aace623602a1343c2cfbefd3fa188e015bebf0
+  md5: 6168d71addc746e8f2b8d57dfd2edcea
+  depends:
+  - libgomp >=7.5.0
+  constrains:
+  - openmp_impl 9999
+  license: BSD-3-Clause
+  license_family: BSD
+  size: 23712
+  timestamp: 1650670790230
+- kind: conda
+  name: brotli-python
+  version: 1.1.0
+  build: py310h00ffb61_1
+  build_number: 1
+  subdir: win-64
+  url: https://conda.anaconda.org/conda-forge/win-64/brotli-python-1.1.0-py310h00ffb61_1.conda
+  sha256: 8de77cf62a653dd6ffe19927b92c421f5fa73c078d7799181f5211a1bac2883b
+  md5: 42bfbc1d41cbe2696a3c9d8b0342324f
+  depends:
+  - python >=3.10,<3.11.0a0
+  - python_abi 3.10.* *_cp310
+  - ucrt >=10.0.20348.0
+  - vc >=14.2,<15
+  - vc14_runtime >=14.29.30139
+  constrains:
+  - libbrotlicommon 1.1.0 hcfcfb64_1
+  license: MIT
+  license_family: MIT
+  size: 321672
+  timestamp: 1695990897641
+- kind: conda
+  name: brotli-python
+  version: 1.1.0
+  build: py310h1253130_1
+  build_number: 1
+  subdir: osx-arm64
+  url: https://conda.anaconda.org/conda-forge/osx-arm64/brotli-python-1.1.0-py310h1253130_1.conda
+  sha256: dab21e18c0275bfd93a09b751096998485677ed17c2e2d08298bc5b43c10bee1
+  md5: 26fab7f65a80fff9f402ec3b7860b88a
+  depends:
+  - libcxx >=15.0.7
+  - python >=3.10,<3.11.0a0
+  - python >=3.10,<3.11.0a0 *_cpython
+  - python_abi 3.10.* *_cp310
+  constrains:
+  - libbrotlicommon 1.1.0 hb547adb_1
+  license: MIT
+  license_family: MIT
+  size: 344275
+  timestamp: 1695990848681
+- kind: conda
+  name: brotli-python
+  version: 1.1.0
+  build: py310h9e9d8ca_1
+  build_number: 1
+  subdir: osx-64
+  url: https://conda.anaconda.org/conda-forge/osx-64/brotli-python-1.1.0-py310h9e9d8ca_1.conda
+  sha256: 57d66ca3e072b889c94cfaf56eb7e1794d3b1b3179bd475a4edef50a03359354
+  md5: 2362e323293e7699cf1e621d502f86d6
+  depends:
+  - libcxx >=15.0.7
+  - python >=3.10,<3.11.0a0
+  - python_abi 3.10.* *_cp310
+  constrains:
+  - libbrotlicommon 1.1.0 h0dc2134_1
+  license: MIT
+  license_family: MIT
+  size: 367037
+  timestamp: 1695990378635
+- kind: conda
+  name: brotli-python
+  version: 1.1.0
+  build: py310hbb3657e_1
+  build_number: 1
+  subdir: linux-aarch64
+  url: https://conda.anaconda.org/conda-forge/linux-aarch64/brotli-python-1.1.0-py310hbb3657e_1.conda
+  sha256: 192f2537ca30953653375abd851f1ccf8b849988e3195e0189aa47384a3a88d9
+  md5: 5ed52d1d3c480022fe67ae00d1cab792
+  depends:
+  - libgcc-ng >=12
+  - libstdcxx-ng >=12
+  - python >=3.10,<3.11.0a0
+  - python >=3.10,<3.11.0a0 *_cpython
+  - python_abi 3.10.* *_cp310
+  constrains:
+  - libbrotlicommon 1.1.0 h31becfc_1
+  license: MIT
+  license_family: MIT
+  size: 355646
+  timestamp: 1695990521531
+- kind: conda
+  name: brotli-python
+  version: 1.1.0
+  build: py310hc6cd4ac_1
+  build_number: 1
+  subdir: linux-64
+  url: https://conda.anaconda.org/conda-forge/linux-64/brotli-python-1.1.0-py310hc6cd4ac_1.conda
+  sha256: e22268d81905338570786921b3def88e55f9ed6d0ccdd17d9fbae31a02fbef69
+  md5: 1f95722c94f00b69af69a066c7433714
+  depends:
+  - libgcc-ng >=12
+  - libstdcxx-ng >=12
+  - python >=3.10,<3.11.0a0
+  - python_abi 3.10.* *_cp310
+  constrains:
+  - libbrotlicommon 1.1.0 hd590300_1
+  license: MIT
+  license_family: MIT
+  size: 349397
+  timestamp: 1695990295884
+- kind: conda
+  name: bzip2
+  version: 1.0.8
+  build: h10d778d_5
+  build_number: 5
+  subdir: osx-64
+  url: https://conda.anaconda.org/conda-forge/osx-64/bzip2-1.0.8-h10d778d_5.conda
+  sha256: 61fb2b488928a54d9472113e1280b468a309561caa54f33825a3593da390b242
+  md5: 6097a6ca9ada32699b5fc4312dd6ef18
+  license: bzip2-1.0.6
+  license_family: BSD
+  size: 127885
+  timestamp: 1699280178474
+- kind: conda
+  name: bzip2
+  version: 1.0.8
+  build: h31becfc_5
+  build_number: 5
+  subdir: linux-aarch64
+  url: https://conda.anaconda.org/conda-forge/linux-aarch64/bzip2-1.0.8-h31becfc_5.conda
+  sha256: b9f170990625cb1eeefaca02e091dc009a64264b077166d8ed7aeb7a09e923b0
+  md5: a64e35f01e0b7a2a152eca87d33b9c87
+  depends:
+  - libgcc-ng >=12
+  license: bzip2-1.0.6
+  license_family: BSD
+  size: 189668
+  timestamp: 1699280060686
+- kind: conda
+  name: bzip2
+  version: 1.0.8
+  build: h93a5062_5
+  build_number: 5
+  subdir: osx-arm64
+  url: https://conda.anaconda.org/conda-forge/osx-arm64/bzip2-1.0.8-h93a5062_5.conda
+  sha256: bfa84296a638bea78a8bb29abc493ee95f2a0218775642474a840411b950fe5f
+  md5: 1bbc659ca658bfd49a481b5ef7a0f40f
+  license: bzip2-1.0.6
+  license_family: BSD
+  size: 122325
+  timestamp: 1699280294368
+- kind: conda
+  name: bzip2
+  version: 1.0.8
+  build: hcfcfb64_5
+  build_number: 5
+  subdir: win-64
+  url: https://conda.anaconda.org/conda-forge/win-64/bzip2-1.0.8-hcfcfb64_5.conda
+  sha256: ae5f47a5c86fd6db822931255dcf017eb12f60c77f07dc782ccb477f7808aab2
+  md5: 26eb8ca6ea332b675e11704cce84a3be
+  depends:
+  - ucrt >=10.0.20348.0
+  - vc >=14.2,<15
+  - vc14_runtime >=14.29.30139
+  license: bzip2-1.0.6
+  license_family: BSD
+  size: 124580
+  timestamp: 1699280668742
+- kind: conda
+  name: bzip2
+  version: 1.0.8
+  build: hd590300_5
+  build_number: 5
+  subdir: linux-64
+  url: https://conda.anaconda.org/conda-forge/linux-64/bzip2-1.0.8-hd590300_5.conda
+  sha256: 242c0c324507ee172c0e0dd2045814e746bb303d1eb78870d182ceb0abc726a8
+  md5: 69b8b6202a07720f448be700e300ccf4
+  depends:
+  - libgcc-ng >=12
+  license: bzip2-1.0.6
+  license_family: BSD
+  size: 254228
+  timestamp: 1699279927352
+- kind: conda
+  name: ca-certificates
+  version: 2024.2.2
+  build: h56e8100_0
+  subdir: win-64
+  url: https://conda.anaconda.org/conda-forge/win-64/ca-certificates-2024.2.2-h56e8100_0.conda
+  sha256: 4d587088ecccd393fec3420b64f1af4ee1a0e6897a45cfd5ef38055322cea5d0
+  md5: 63da060240ab8087b60d1357051ea7d6
+  license: ISC
+  size: 155886
+  timestamp: 1706843918052
+- kind: conda
+  name: ca-certificates
+  version: 2024.2.2
+  build: h8857fd0_0
+  subdir: osx-64
+  url: https://conda.anaconda.org/conda-forge/osx-64/ca-certificates-2024.2.2-h8857fd0_0.conda
+  sha256: 54a794aedbb4796afeabdf54287b06b1d27f7b13b3814520925f4c2c80f58ca9
+  md5: f2eacee8c33c43692f1ccfd33d0f50b1
+  license: ISC
+  size: 155665
+  timestamp: 1706843838227
+- kind: conda
+  name: ca-certificates
+  version: 2024.2.2
+  build: hbcca054_0
+  subdir: linux-64
+  url: https://conda.anaconda.org/conda-forge/linux-64/ca-certificates-2024.2.2-hbcca054_0.conda
+  sha256: 91d81bfecdbb142c15066df70cc952590ae8991670198f92c66b62019b251aeb
+  md5: 2f4327a1cbe7f022401b236e915a5fef
+  license: ISC
+  size: 155432
+  timestamp: 1706843687645
+- kind: conda
+  name: ca-certificates
+  version: 2024.2.2
+  build: hcefe29a_0
+  subdir: linux-aarch64
+  url: https://conda.anaconda.org/conda-forge/linux-aarch64/ca-certificates-2024.2.2-hcefe29a_0.conda
+  sha256: 0f6b34d835e26e5fa97cca4985dc46f0aba551a3a23f07c6f13cca2542b8c642
+  md5: 57c226edb90c4e973b9b7503537dd339
+  license: ISC
+  size: 155738
+  timestamp: 1706845723412
+- kind: conda
+  name: ca-certificates
+  version: 2024.2.2
+  build: hf0a4a13_0
+  subdir: osx-arm64
+  url: https://conda.anaconda.org/conda-forge/osx-arm64/ca-certificates-2024.2.2-hf0a4a13_0.conda
+  sha256: 49bc3439816ac72d0c0e0f144b8cc870fdcc4adec2e861407ec818d8116b2204
+  md5: fb416a1795f18dcc5a038bc2dc54edf9
+  license: ISC
+  size: 155725
+  timestamp: 1706844034242
+- kind: conda
+  name: ld_impl_linux-64
+  version: '2.40'
+  build: h41732ed_0
+  subdir: linux-64
+  url: https://conda.anaconda.org/conda-forge/linux-64/ld_impl_linux-64-2.40-h41732ed_0.conda
+  sha256: f6cc89d887555912d6c61b295d398cff9ec982a3417d38025c45d5dd9b9e79cd
+  md5: 7aca3059a1729aa76c597603f10b0dd3
+  constrains:
+  - binutils_impl_linux-64 2.40
+  license: GPL-3.0-only
+  license_family: GPL
+  size: 704696
+  timestamp: 1674833944779
+- kind: conda
+  name: ld_impl_linux-aarch64
+  version: '2.40'
+  build: h2d8c526_0
+  subdir: linux-aarch64
+  url: https://conda.anaconda.org/conda-forge/linux-aarch64/ld_impl_linux-aarch64-2.40-h2d8c526_0.conda
+  sha256: 1ba06e8645094b340b4aee23603a6abb1b0383788180e65f3de34e655c5f577c
+  md5: 16246d69e945d0b1969a6099e7c5d457
+  constrains:
+  - binutils_impl_linux-aarch64 2.40
+  license: GPL-3.0-only
+  license_family: GPL
+  size: 738776
+  timestamp: 1674833843183
+- kind: conda
+  name: libcxx
+  version: 16.0.6
+  build: h4653b0c_0
+  subdir: osx-arm64
+  url: https://conda.anaconda.org/conda-forge/osx-arm64/libcxx-16.0.6-h4653b0c_0.conda
+  sha256: 11d3fb51c14832d9e4f6d84080a375dec21ea8a3a381a1910e67ff9cedc20355
+  md5: 9d7d724faf0413bf1dbc5a85935700c8
+  license: Apache-2.0 WITH LLVM-exception
+  license_family: Apache
+  size: 1160232
+  timestamp: 1686896993785
+- kind: conda
+  name: libcxx
+  version: 16.0.6
+  build: hd57cbcb_0
+  subdir: osx-64
+  url: https://conda.anaconda.org/conda-forge/osx-64/libcxx-16.0.6-hd57cbcb_0.conda
+  sha256: 9063271847cf05f3a6cc6cae3e7f0ced032ab5f3a3c9d3f943f876f39c5c2549
+  md5: 7d6972792161077908b62971802f289a
+  license: Apache-2.0 WITH LLVM-exception
+  license_family: Apache
+  size: 1142172
+  timestamp: 1686896907750
+- kind: conda
+  name: libffi
+  version: 3.4.2
+  build: h0d85af4_5
+  build_number: 5
+  subdir: osx-64
+  url: https://conda.anaconda.org/conda-forge/osx-64/libffi-3.4.2-h0d85af4_5.tar.bz2
+  sha256: 7a2d27a936ceee6942ea4d397f9c7d136f12549d86f7617e8b6bad51e01a941f
+  md5: ccb34fb14960ad8b125962d3d79b31a9
+  license: MIT
+  license_family: MIT
+  size: 51348
+  timestamp: 1636488394370
+- kind: conda
+  name: libffi
+  version: 3.4.2
+  build: h3422bc3_5
+  build_number: 5
+  subdir: osx-arm64
+  url: https://conda.anaconda.org/conda-forge/osx-arm64/libffi-3.4.2-h3422bc3_5.tar.bz2
+  sha256: 41b3d13efb775e340e4dba549ab5c029611ea6918703096b2eaa9c015c0750ca
+  md5: 086914b672be056eb70fd4285b6783b6
+  license: MIT
+  license_family: MIT
+  size: 39020
+  timestamp: 1636488587153
+- kind: conda
+  name: libffi
+  version: 3.4.2
+  build: h3557bc0_5
+  build_number: 5
+  subdir: linux-aarch64
+  url: https://conda.anaconda.org/conda-forge/linux-aarch64/libffi-3.4.2-h3557bc0_5.tar.bz2
+  sha256: 7e9258a102480757fe3faeb225a3ca04dffd10fecd2a958c65cdb4cdf75f2c3c
+  md5: dddd85f4d52121fab0a8b099c5e06501
+  depends:
+  - libgcc-ng >=9.4.0
+  license: MIT
+  license_family: MIT
+  size: 59450
+  timestamp: 1636488255090
+- kind: conda
+  name: libffi
+  version: 3.4.2
+  build: h7f98852_5
+  build_number: 5
+  subdir: linux-64
+  url: https://conda.anaconda.org/conda-forge/linux-64/libffi-3.4.2-h7f98852_5.tar.bz2
+  sha256: ab6e9856c21709b7b517e940ae7028ae0737546122f83c2aa5d692860c3b149e
+  md5: d645c6d2ac96843a2bfaccd2d62b3ac3
+  depends:
+  - libgcc-ng >=9.4.0
+  license: MIT
+  license_family: MIT
+  size: 58292
+  timestamp: 1636488182923
+- kind: conda
+  name: libffi
+  version: 3.4.2
+  build: h8ffe710_5
+  build_number: 5
+  subdir: win-64
+  url: https://conda.anaconda.org/conda-forge/win-64/libffi-3.4.2-h8ffe710_5.tar.bz2
+  sha256: 1951ab740f80660e9bc07d2ed3aefb874d78c107264fd810f24a1a6211d4b1a5
+  md5: 2c96d1b6915b408893f9472569dee135
+  depends:
+  - vc >=14.1,<15.0a0
+  - vs2015_runtime >=14.16.27012
+  license: MIT
+  license_family: MIT
+  size: 42063
+  timestamp: 1636489106777
+- kind: conda
+  name: libgcc-ng
+  version: 13.2.0
+  build: h807b86a_5
+  build_number: 5
+  subdir: linux-64
+  url: https://conda.anaconda.org/conda-forge/linux-64/libgcc-ng-13.2.0-h807b86a_5.conda
+  sha256: d32f78bfaac282cfe5205f46d558704ad737b8dbf71f9227788a5ca80facaba4
+  md5: d4ff227c46917d3b4565302a2bbb276b
+  depends:
+  - _libgcc_mutex 0.1 conda_forge
+  - _openmp_mutex >=4.5
+  constrains:
+  - libgomp 13.2.0 h807b86a_5
+  license: GPL-3.0-only WITH GCC-exception-3.1
+  license_family: GPL
+  size: 770506
+  timestamp: 1706819192021
+- kind: conda
+  name: libgcc-ng
+  version: 13.2.0
+  build: hf8544c7_5
+  build_number: 5
+  subdir: linux-aarch64
+  url: https://conda.anaconda.org/conda-forge/linux-aarch64/libgcc-ng-13.2.0-hf8544c7_5.conda
+  sha256: 869e44e1cf329198f5bea56c146207ed639b24b6281187159435b9499ecb3959
+  md5: dee934e640275d9e74e7bbd455f25162
+  depends:
+  - _openmp_mutex >=4.5
+  constrains:
+  - libgomp 13.2.0 hf8544c7_5
+  license: GPL-3.0-only WITH GCC-exception-3.1
+  license_family: GPL
+  size: 456795
+  timestamp: 1706820691781
+- kind: conda
+  name: libgomp
+  version: 13.2.0
+  build: h807b86a_5
+  build_number: 5
+  subdir: linux-64
+  url: https://conda.anaconda.org/conda-forge/linux-64/libgomp-13.2.0-h807b86a_5.conda
+  sha256: 0d3d4b1b0134283ea02d58e8eb5accf3655464cf7159abf098cc694002f8d34e
+  md5: d211c42b9ce49aee3734fdc828731689
+  depends:
+  - _libgcc_mutex 0.1 conda_forge
+  license: GPL-3.0-only WITH GCC-exception-3.1
+  license_family: GPL
+  size: 419751
+  timestamp: 1706819107383
+- kind: conda
+  name: libgomp
+  version: 13.2.0
+  build: hf8544c7_5
+  build_number: 5
+  subdir: linux-aarch64
+  url: https://conda.anaconda.org/conda-forge/linux-aarch64/libgomp-13.2.0-hf8544c7_5.conda
+  sha256: a98d4f242a351feb7983a28e7d6a0ca51da764c6233ea3dfc776975a3aba8a01
+  md5: 379be2f115ffb73860e4e260dd2170b7
+  license: GPL-3.0-only WITH GCC-exception-3.1
+  license_family: GPL
+  size: 423091
+  timestamp: 1706820564165
+- kind: conda
+  name: libnsl
+  version: 2.0.1
+  build: h31becfc_0
+  subdir: linux-aarch64
+  url: https://conda.anaconda.org/conda-forge/linux-aarch64/libnsl-2.0.1-h31becfc_0.conda
+  sha256: fd18c2b75d7411096428d36a70b36b1a17e31f7b8956b6905d145792d49e97f8
+  md5: c14f32510f694e3185704d89967ec422
+  depends:
+  - libgcc-ng >=12
+  license: LGPL-2.1-only
+  license_family: GPL
+  size: 34501
+  timestamp: 1697358973269
+- kind: conda
+  name: libnsl
+  version: 2.0.1
+  build: hd590300_0
+  subdir: linux-64
+  url: https://conda.anaconda.org/conda-forge/linux-64/libnsl-2.0.1-hd590300_0.conda
+  sha256: 26d77a3bb4dceeedc2a41bd688564fe71bf2d149fdcf117049970bc02ff1add6
+  md5: 30fd6e37fe21f86f4bd26d6ee73eeec7
+  depends:
+  - libgcc-ng >=12
+  license: LGPL-2.1-only
+  license_family: GPL
+  size: 33408
+  timestamp: 1697359010159
+- kind: conda
+  name: libsqlite
+  version: 3.45.2
+  build: h091b4b1_0
+  subdir: osx-arm64
+  url: https://conda.anaconda.org/conda-forge/osx-arm64/libsqlite-3.45.2-h091b4b1_0.conda
+  sha256: 7c234320a1a2132b9cc972aaa06bb215bb220a5b1addb0bed7a5a321c805920e
+  md5: 9d07427ee5bd9afd1e11ce14368a48d6
+  depends:
+  - libzlib >=1.2.13,<1.3.0a0
+  license: Unlicense
+  size: 825300
+  timestamp: 1710255078823
+- kind: conda
+  name: libsqlite
+  version: 3.45.2
+  build: h194ca79_0
+  subdir: linux-aarch64
+  url: https://conda.anaconda.org/conda-forge/linux-aarch64/libsqlite-3.45.2-h194ca79_0.conda
+  sha256: 0ce6de6369c04386cfc8696b1f795f425843789609ae2e04e7a1eb7deae62a8b
+  md5: bf4c96a21fbfc6a6ef6a7781a534a4e0
+  depends:
+  - libgcc-ng >=12
+  - libzlib >=1.2.13,<1.3.0a0
+  license: Unlicense
+  size: 1038462
+  timestamp: 1710253998432
+- kind: conda
+  name: libsqlite
+  version: 3.45.2
+  build: h2797004_0
+  subdir: linux-64
+  url: https://conda.anaconda.org/conda-forge/linux-64/libsqlite-3.45.2-h2797004_0.conda
+  sha256: 8cdbeb7902729e319510a82d7c642402981818702b58812af265ef55d1315473
+  md5: 866983a220e27a80cb75e85cb30466a1
+  depends:
+  - libgcc-ng >=12
+  - libzlib >=1.2.13,<1.3.0a0
+  license: Unlicense
+  size: 857489
+  timestamp: 1710254744982
+- kind: conda
+  name: libsqlite
+  version: 3.45.2
+  build: h92b6c6a_0
+  subdir: osx-64
+  url: https://conda.anaconda.org/conda-forge/osx-64/libsqlite-3.45.2-h92b6c6a_0.conda
+  sha256: 320ec73a4e3dd377757a2595770b8137ec4583df4d7782472d76377cdbdc4543
+  md5: 086f56e13a96a6cfb1bf640505ae6b70
+  depends:
+  - libzlib >=1.2.13,<1.3.0a0
+  license: Unlicense
+  size: 902355
+  timestamp: 1710254991672
+- kind: conda
+  name: libsqlite
+  version: 3.45.2
+  build: hcfcfb64_0
+  subdir: win-64
+  url: https://conda.anaconda.org/conda-forge/win-64/libsqlite-3.45.2-hcfcfb64_0.conda
+  sha256: 4bb24b986550275a6d02835150d943c4c675808d05c0efc5c2a22154d007a69f
+  md5: f95359f8dc5abf7da7776ece9ef10bc5
+  depends:
+  - ucrt >=10.0.20348.0
+  - vc >=14.2,<15
+  - vc14_runtime >=14.29.30139
+  license: Unlicense
+  size: 869606
+  timestamp: 1710255095740
+- kind: conda
+  name: libstdcxx-ng
+  version: 13.2.0
+  build: h7e041cc_5
+  build_number: 5
+  subdir: linux-64
+  url: https://conda.anaconda.org/conda-forge/linux-64/libstdcxx-ng-13.2.0-h7e041cc_5.conda
+  sha256: a56c5b11f1e73a86e120e6141a42d9e935a99a2098491ac9e15347a1476ce777
+  md5: f6f6600d18a4047b54f803cf708b868a
+  license: GPL-3.0-only WITH GCC-exception-3.1
+  license_family: GPL
+  size: 3834139
+  timestamp: 1706819252496
+- kind: conda
+  name: libstdcxx-ng
+  version: 13.2.0
+  build: h9a76618_5
+  build_number: 5
+  subdir: linux-aarch64
+  url: https://conda.anaconda.org/conda-forge/linux-aarch64/libstdcxx-ng-13.2.0-h9a76618_5.conda
+  sha256: c209f23a8a497fc87107a68b6bbc8d2089cf15fd4015b558dfdce63544379b05
+  md5: 1b79d37dce0fad96bdf3de03925f43b4
+  license: GPL-3.0-only WITH GCC-exception-3.1
+  license_family: GPL
+  size: 3752658
+  timestamp: 1706820778418
+- kind: conda
+  name: libuuid
+  version: 2.38.1
+  build: h0b41bf4_0
+  subdir: linux-64
+  url: https://conda.anaconda.org/conda-forge/linux-64/libuuid-2.38.1-h0b41bf4_0.conda
+  sha256: 787eb542f055a2b3de553614b25f09eefb0a0931b0c87dbcce6efdfd92f04f18
+  md5: 40b61aab5c7ba9ff276c41cfffe6b80b
+  depends:
+  - libgcc-ng >=12
+  license: BSD-3-Clause
+  license_family: BSD
+  size: 33601
+  timestamp: 1680112270483
+- kind: conda
+  name: libuuid
+  version: 2.38.1
+  build: hb4cce97_0
+  subdir: linux-aarch64
+  url: https://conda.anaconda.org/conda-forge/linux-aarch64/libuuid-2.38.1-hb4cce97_0.conda
+  sha256: 616277b0c5f7616c2cdf36f6c316ea3f9aa5bb35f2d4476a349ab58b9b91675f
+  md5: 000e30b09db0b7c775b21695dff30969
+  depends:
+  - libgcc-ng >=12
+  license: BSD-3-Clause
+  license_family: BSD
+  size: 35720
+  timestamp: 1680113474501
+- kind: conda
+  name: libxcrypt
+  version: 4.4.36
+  build: h31becfc_1
+  build_number: 1
+  subdir: linux-aarch64
+  url: https://conda.anaconda.org/conda-forge/linux-aarch64/libxcrypt-4.4.36-h31becfc_1.conda
+  sha256: 6b46c397644091b8a26a3048636d10b989b1bf266d4be5e9474bf763f828f41f
+  md5: b4df5d7d4b63579d081fd3a4cf99740e
+  depends:
+  - libgcc-ng >=12
+  license: LGPL-2.1-or-later
+  size: 114269
+  timestamp: 1702724369203
+- kind: conda
+  name: libxcrypt
+  version: 4.4.36
+  build: hd590300_1
+  build_number: 1
+  subdir: linux-64
+  url: https://conda.anaconda.org/conda-forge/linux-64/libxcrypt-4.4.36-hd590300_1.conda
+  sha256: 6ae68e0b86423ef188196fff6207ed0c8195dd84273cb5623b85aa08033a410c
+  md5: 5aa797f8787fe7a17d1b0821485b5adc
+  depends:
+  - libgcc-ng >=12
+  license: LGPL-2.1-or-later
+  size: 100393
+  timestamp: 1702724383534
+- kind: conda
+  name: libzlib
+  version: 1.2.13
+  build: h31becfc_5
+  build_number: 5
+  subdir: linux-aarch64
+  url: https://conda.anaconda.org/conda-forge/linux-aarch64/libzlib-1.2.13-h31becfc_5.conda
+  sha256: aeeefbb61e5e8227e53566d5e42dbb49e120eb99109996bf0dbfde8f180747a7
+  md5: b213aa87eea9491ef7b129179322e955
+  depends:
+  - libgcc-ng >=12
+  constrains:
+  - zlib 1.2.13 *_5
+  license: Zlib
+  license_family: Other
+  size: 67036
+  timestamp: 1686575148440
+- kind: conda
+  name: libzlib
+  version: 1.2.13
+  build: h53f4e23_5
+  build_number: 5
+  subdir: osx-arm64
+  url: https://conda.anaconda.org/conda-forge/osx-arm64/libzlib-1.2.13-h53f4e23_5.conda
+  sha256: ab1c8aefa2d54322a63aaeeefe9cf877411851738616c4068e0dccc66b9c758a
+  md5: 1a47f5236db2e06a320ffa0392f81bd8
+  constrains:
+  - zlib 1.2.13 *_5
+  license: Zlib
+  license_family: Other
+  size: 48102
+  timestamp: 1686575426584
+- kind: conda
+  name: libzlib
+  version: 1.2.13
+  build: h8a1eda9_5
+  build_number: 5
+  subdir: osx-64
+  url: https://conda.anaconda.org/conda-forge/osx-64/libzlib-1.2.13-h8a1eda9_5.conda
+  sha256: fc58ad7f47ffea10df1f2165369978fba0a1cc32594aad778f5eec725f334867
+  md5: 4a3ad23f6e16f99c04e166767193d700
+  constrains:
+  - zlib 1.2.13 *_5
+  license: Zlib
+  license_family: Other
+  size: 59404
+  timestamp: 1686575566695
+- kind: conda
+  name: libzlib
+  version: 1.2.13
+  build: hcfcfb64_5
+  build_number: 5
+  subdir: win-64
+  url: https://conda.anaconda.org/conda-forge/win-64/libzlib-1.2.13-hcfcfb64_5.conda
+  sha256: c161822ee8130b71e08b6d282b9919c1de2c5274b29921a867bca0f7d30cad26
+  md5: 5fdb9c6a113b6b6cb5e517fd972d5f41
+  depends:
+  - ucrt >=10.0.20348.0
+  - vc >=14.2,<15
+  - vc14_runtime >=14.29.30139
+  constrains:
+  - zlib 1.2.13 *_5
+  license: Zlib
+  license_family: Other
+  size: 55800
+  timestamp: 1686575452215
+- kind: conda
+  name: libzlib
+  version: 1.2.13
+  build: hd590300_5
+  build_number: 5
+  subdir: linux-64
+  url: https://conda.anaconda.org/conda-forge/linux-64/libzlib-1.2.13-hd590300_5.conda
+  sha256: 370c7c5893b737596fd6ca0d9190c9715d89d888b8c88537ae1ef168c25e82e4
+  md5: f36c115f1ee199da648e0597ec2047ad
+  depends:
+  - libgcc-ng >=12
+  constrains:
+  - zlib 1.2.13 *_5
+  license: Zlib
+  license_family: Other
+  size: 61588
+  timestamp: 1686575217516
+- kind: conda
+  name: m2w64-gcc-libgfortran
+  version: 5.3.0
+  build: '6'
+  build_number: 6
+  subdir: win-64
+  url: https://conda.anaconda.org/conda-forge/win-64/m2w64-gcc-libgfortran-5.3.0-6.tar.bz2
+  sha256: 9de95a7996d5366ae0808eef2acbc63f9b11b874aa42375f55379e6715845dc6
+  md5: 066552ac6b907ec6d72c0ddab29050dc
+  depends:
+  - m2w64-gcc-libs-core
+  - msys2-conda-epoch ==20160418
+  license: GPL, LGPL, FDL, custom
+  size: 350687
+  timestamp: 1608163451316
+- kind: conda
+  name: m2w64-gcc-libs
+  version: 5.3.0
+  build: '7'
+  build_number: 7
+  subdir: win-64
+  url: https://conda.anaconda.org/conda-forge/win-64/m2w64-gcc-libs-5.3.0-7.tar.bz2
+  sha256: 3bd1ab02b7c89a5b153a17be03b36d833f1517ff2a6a77ead7c4a808b88196aa
+  md5: fe759119b8b3bfa720b8762c6fdc35de
+  depends:
+  - m2w64-gcc-libgfortran
+  - m2w64-gcc-libs-core
+  - m2w64-gmp
+  - m2w64-libwinpthread-git
+  - msys2-conda-epoch ==20160418
+  license: GPL3+, partial:GCCRLE, partial:LGPL2+
+  size: 532390
+  timestamp: 1608163512830
+- kind: conda
+  name: m2w64-gcc-libs-core
+  version: 5.3.0
+  build: '7'
+  build_number: 7
+  subdir: win-64
+  url: https://conda.anaconda.org/conda-forge/win-64/m2w64-gcc-libs-core-5.3.0-7.tar.bz2
+  sha256: 58afdfe859ed2e9a9b1cc06bc408720cb2c3a6a132e59d4805b090d7574f4ee0
+  md5: 4289d80fb4d272f1f3b56cfe87ac90bd
+  depends:
+  - m2w64-gmp
+  - m2w64-libwinpthread-git
+  - msys2-conda-epoch ==20160418
+  license: GPL3+, partial:GCCRLE, partial:LGPL2+
+  size: 219240
+  timestamp: 1608163481341
+- kind: conda
+  name: m2w64-gmp
+  version: 6.1.0
+  build: '2'
+  build_number: 2
+  subdir: win-64
+  url: https://conda.anaconda.org/conda-forge/win-64/m2w64-gmp-6.1.0-2.tar.bz2
+  sha256: 7e3cd95f554660de45f8323fca359e904e8d203efaf07a4d311e46d611481ed1
+  md5: 53a1c73e1e3d185516d7e3af177596d9
+  depends:
+  - msys2-conda-epoch ==20160418
+  license: LGPL3
+  size: 743501
+  timestamp: 1608163782057
+- kind: conda
+  name: m2w64-libwinpthread-git
+  version: 5.0.0.4634.697f757
+  build: '2'
+  build_number: 2
+  subdir: win-64
+  url: https://conda.anaconda.org/conda-forge/win-64/m2w64-libwinpthread-git-5.0.0.4634.697f757-2.tar.bz2
+  sha256: f63a09b2cae7defae0480f1740015d6235f1861afa6fe2e2d3e10bd0d1314ee0
+  md5: 774130a326dee16f1ceb05cc687ee4f0
+  depends:
+  - msys2-conda-epoch ==20160418
+  license: MIT, BSD
+  size: 31928
+  timestamp: 1608166099896
+- kind: conda
+  name: msys2-conda-epoch
+  version: '20160418'
+  build: '1'
+  build_number: 1
+  subdir: win-64
+  url: https://conda.anaconda.org/conda-forge/win-64/msys2-conda-epoch-20160418-1.tar.bz2
+  sha256: 99358d58d778abee4dca82ad29fb58058571f19b0f86138363c260049d4ac7f1
+  md5: b0309b72560df66f71a9d5e34a5efdfa
+  size: 3227
+  timestamp: 1608166968312
+- kind: conda
+  name: mypy
+  version: 1.8.0
+  build: py310h2372a71_0
+  subdir: linux-64
+  url: https://conda.anaconda.org/conda-forge/linux-64/mypy-1.8.0-py310h2372a71_0.conda
+  sha256: 6c01268327db83c70c38cfc87fc13a71d09cda123ae06cd6edbbe620c2b20f33
+  md5: 3320dc32fc6bd29ab4a16cf22bc35fc2
+  depends:
+  - libgcc-ng >=12
+  - mypy_extensions >=1.0.0
+  - psutil >=4.0
+  - python >=3.10,<3.11.0a0
+  - python_abi 3.10.* *_cp310
+  - tomli >=1.1.0
+  - typing_extensions >=4.1.0
+  license: MIT
+  license_family: MIT
+  size: 17160046
+  timestamp: 1703185152663
+- kind: conda
+  name: mypy
+  version: 1.8.0
+  build: py310h8d17308_0
+  subdir: win-64
+  url: https://conda.anaconda.org/conda-forge/win-64/mypy-1.8.0-py310h8d17308_0.conda
+  sha256: 8ca9e638a538225b6a5a935573964fa4d743456ece6171988d4116d57a635069
+  md5: 42c9adc3e138cd581a869d46dfdb3fcd
+  depends:
+  - mypy_extensions >=1.0.0
+  - psutil >=4.0
+  - python >=3.10,<3.11.0a0
+  - python_abi 3.10.* *_cp310
+  - tomli >=1.1.0
+  - typing_extensions >=4.1.0
+  - ucrt >=10.0.20348.0
+  - vc >=14.2,<15
+  - vc14_runtime >=14.29.30139
+  license: MIT
+  license_family: MIT
+  size: 9332107
+  timestamp: 1703185142866
+- kind: conda
+  name: mypy
+  version: 1.8.0
+  build: py310hb299538_0
+  subdir: linux-aarch64
+  url: https://conda.anaconda.org/conda-forge/linux-aarch64/mypy-1.8.0-py310hb299538_0.conda
+  sha256: a970f3c8d22641c7202a6b40c106a55e8b77b0fe4cca715c95ca551262cdebea
+  md5: bf64efdd960ab39f9cd99f562f7ad27f
+  depends:
+  - libgcc-ng >=12
+  - mypy_extensions >=1.0.0
+  - psutil >=4.0
+  - python >=3.10,<3.11.0a0
+  - python >=3.10,<3.11.0a0 *_cpython
+  - python_abi 3.10.* *_cp310
+  - tomli >=1.1.0
+  - typing_extensions >=4.1.0
+  license: MIT
+  license_family: MIT
+  size: 14537904
+  timestamp: 1703185330563
+- kind: conda
+  name: mypy
+  version: 1.8.0
+  build: py310hb372a2b_0
+  subdir: osx-64
+  url: https://conda.anaconda.org/conda-forge/osx-64/mypy-1.8.0-py310hb372a2b_0.conda
+  sha256: 3af15d65a207840b15b3298398f70e48263a20706d2bc48bede09f9077597759
+  md5: 5e8c2a9af839d3d23be1cf7e2c955c5c
+  depends:
+  - mypy_extensions >=1.0.0
+  - psutil >=4.0
+  - python >=3.10,<3.11.0a0
+  - python_abi 3.10.* *_cp310
+  - tomli >=1.1.0
+  - typing_extensions >=4.1.0
+  license: MIT
+  license_family: MIT
+  size: 11303456
+  timestamp: 1703184930605
+- kind: conda
+  name: mypy
+  version: 1.8.0
+  build: py310hd125d64_0
+  subdir: osx-arm64
+  url: https://conda.anaconda.org/conda-forge/osx-arm64/mypy-1.8.0-py310hd125d64_0.conda
+  sha256: 630f3bccefb8b7139dcee0941fb241664130953f319885134f8cd813d4e300ce
+  md5: 0884b51650eb1e4f77022ffd68819ce6
+  depends:
+  - mypy_extensions >=1.0.0
+  - psutil >=4.0
+  - python >=3.10,<3.11.0a0
+  - python >=3.10,<3.11.0a0 *_cpython
+  - python_abi 3.10.* *_cp310
+  - tomli >=1.1.0
+  - typing_extensions >=4.1.0
+  license: MIT
+  license_family: MIT
+  size: 8941826
+  timestamp: 1703185223331
+- kind: conda
+  name: mypy_extensions
+  version: 1.0.0
+  build: pyha770c72_0
+  subdir: noarch
+  noarch: python
+  url: https://conda.anaconda.org/conda-forge/noarch/mypy_extensions-1.0.0-pyha770c72_0.conda
+  sha256: f240217476e148e825420c6bc3a0c0efb08c0718b7042fae960400c02af858a3
+  md5: 4eccaeba205f0aed9ac3a9ea58568ca3
+  depends:
+  - python >=3.5
+  license: MIT
+  license_family: MIT
+  size: 10492
+  timestamp: 1675543414256
+- kind: conda
+  name: ncurses
+  version: 6.4.20240210
+  build: h0425590_0
+  subdir: linux-aarch64
+  url: https://conda.anaconda.org/conda-forge/linux-aarch64/ncurses-6.4.20240210-h0425590_0.conda
+  sha256: 4223dc34e2bddd37bf995158ae481e00be375b287d539bc7a0532634c0fc63b7
+  md5: c1a1612ddaee95c83abfa0b2ec858626
+  depends:
+  - libgcc-ng >=12
+  license: X11 AND BSD-3-Clause
+  size: 926594
+  timestamp: 1710866633409
+- kind: conda
+  name: ncurses
+  version: 6.4.20240210
+  build: h078ce10_0
+  subdir: osx-arm64
+  url: https://conda.anaconda.org/conda-forge/osx-arm64/ncurses-6.4.20240210-h078ce10_0.conda
+  sha256: 06f0905791575e2cd3aa961493c56e490b3d82ad9eb49f1c332bd338b0216911
+  md5: 616ae8691e6608527d0071e6766dcb81
+  license: X11 AND BSD-3-Clause
+  size: 820249
+  timestamp: 1710866874348
+- kind: conda
+  name: ncurses
+  version: 6.4.20240210
+  build: h59595ed_0
+  subdir: linux-64
+  url: https://conda.anaconda.org/conda-forge/linux-64/ncurses-6.4.20240210-h59595ed_0.conda
+  sha256: aa0f005b6727aac6507317ed490f0904430584fa8ca722657e7f0fb94741de81
+  md5: 97da8860a0da5413c7c98a3b3838a645
+  depends:
+  - libgcc-ng >=12
+  license: X11 AND BSD-3-Clause
+  size: 895669
+  timestamp: 1710866638986
+- kind: conda
+  name: ncurses
+  version: 6.4.20240210
+  build: h73e2aa4_0
+  subdir: osx-64
+  url: https://conda.anaconda.org/conda-forge/osx-64/ncurses-6.4.20240210-h73e2aa4_0.conda
+  sha256: 50b72acf08acbc4e5332807653e2ca6b26d4326e8af16fad1fd3f2ce9ea55503
+  md5: 50f28c512e9ad78589e3eab34833f762
+  license: X11 AND BSD-3-Clause
+  size: 823010
+  timestamp: 1710866856626
+- kind: conda
+  name: openssl
+  version: 3.2.1
+  build: h0d3ecfb_1
+  build_number: 1
+  subdir: osx-arm64
+  url: https://conda.anaconda.org/conda-forge/osx-arm64/openssl-3.2.1-h0d3ecfb_1.conda
+  sha256: 519dc941d7ab0ebf31a2878d85c2f444450e7c5f6f41c4d07252c6bb3417b78b
+  md5: eb580fb888d93d5d550c557323ac5cee
+  depends:
+  - ca-certificates
+  constrains:
+  - pyopenssl >=22.1
+  license: Apache-2.0
+  license_family: Apache
+  size: 2855250
+  timestamp: 1710793435903
+- kind: conda
+  name: openssl
+  version: 3.2.1
+  build: h31becfc_1
+  build_number: 1
+  subdir: linux-aarch64
+  url: https://conda.anaconda.org/conda-forge/linux-aarch64/openssl-3.2.1-h31becfc_1.conda
+  sha256: 055a26e99ebc12ae0cf23266a0e62e71b59b8ce8cafb1ebb87e375ef9c758d7b
+  md5: e95eb18d256edc72058e0dc9be5338a0
+  depends:
+  - ca-certificates
+  - libgcc-ng >=12
+  constrains:
+  - pyopenssl >=22.1
+  license: Apache-2.0
+  license_family: Apache
+  size: 3380844
+  timestamp: 1710793424665
+- kind: conda
+  name: openssl
+  version: 3.2.1
+  build: hcfcfb64_1
+  build_number: 1
+  subdir: win-64
+  url: https://conda.anaconda.org/conda-forge/win-64/openssl-3.2.1-hcfcfb64_1.conda
+  sha256: 61ce4e11c3c26ed4e4d9b7e7e2483121a1741ad0f9c8db0a91a28b6e05182ce6
+  md5: 958e0418e93e50c575bff70fbcaa12d8
+  depends:
+  - ca-certificates
+  - ucrt >=10.0.20348.0
+  - vc >=14.2,<15
+  - vc14_runtime >=14.29.30139
+  constrains:
+  - pyopenssl >=22.1
+  license: Apache-2.0
+  license_family: Apache
+  size: 8230112
+  timestamp: 1710796158475
+- kind: conda
+  name: openssl
+  version: 3.2.1
+  build: hd590300_1
+  build_number: 1
+  subdir: linux-64
+  url: https://conda.anaconda.org/conda-forge/linux-64/openssl-3.2.1-hd590300_1.conda
+  sha256: 2c689444ed19a603be457284cf2115ee728a3fafb7527326e96054dee7cdc1a7
+  md5: 9d731343cff6ee2e5a25c4a091bf8e2a
+  depends:
+  - ca-certificates
+  - libgcc-ng >=12
+  constrains:
+  - pyopenssl >=22.1
+  license: Apache-2.0
+  license_family: Apache
+  size: 2865379
+  timestamp: 1710793235846
+- kind: conda
+  name: openssl
+  version: 3.2.1
+  build: hd75f5a5_1
+  build_number: 1
+  subdir: osx-64
+  url: https://conda.anaconda.org/conda-forge/osx-64/openssl-3.2.1-hd75f5a5_1.conda
+  sha256: 7ae0ac6a1673584a8a380c2ff3d46eca48ed53bc7174c0d4eaa0dd2f247a0984
+  md5: 570a6f04802df580be529f3a72d2bbf7
+  depends:
+  - ca-certificates
+  constrains:
+  - pyopenssl >=22.1
+  license: Apache-2.0
+  license_family: Apache
+  size: 2506344
+  timestamp: 1710793930515
+- kind: conda
+  name: pip
+  version: '24.0'
+  build: pyhd8ed1ab_0
+  subdir: noarch
+  noarch: python
+  url: https://conda.anaconda.org/conda-forge/noarch/pip-24.0-pyhd8ed1ab_0.conda
+  sha256: b7c1c5d8f13e8cb491c4bd1d0d1896a4cf80fc47de01059ad77509112b664a4a
+  md5: f586ac1e56c8638b64f9c8122a7b8a67
+  depends:
+  - python >=3.7
+  - setuptools
+  - wheel
+  license: MIT
+  license_family: MIT
+  size: 1398245
+  timestamp: 1706960660581
+- kind: conda
+  name: psutil
+  version: 6.0.0
+  build: py310h936d840_0
+  subdir: osx-64
+  url: https://conda.anaconda.org/conda-forge/osx-64/psutil-6.0.0-py310h936d840_0.conda
+  sha256: c976819733772f63a1c8e704cb96bf4287c0eb477b10ba467be3adbe5974bf3a
+  md5: 2f5a3bd97ce3176794b59c160ed51fba
+  depends:
+  - __osx >=10.13
+  - python >=3.10,<3.11.0a0
+  - python_abi 3.10.* *_cp310
+  license: BSD-3-Clause
+  license_family: BSD
+  size: 378205
+  timestamp: 1719274714245
+- kind: conda
+  name: psutil
+  version: 6.0.0
+  build: py310ha6dd24b_0
+  subdir: osx-arm64
+  url: https://conda.anaconda.org/conda-forge/osx-arm64/psutil-6.0.0-py310ha6dd24b_0.conda
+  sha256: 7477bd84734668992cda9076147c5d07ce92f59c70441757a5b289401bd8ed85
+  md5: 0e0df689b8c6ea6676b786bd78a575d1
+  depends:
+  - __osx >=11.0
+  - python >=3.10,<3.11.0a0
+  - python >=3.10,<3.11.0a0 *_cpython
+  - python_abi 3.10.* *_cp310
+  license: BSD-3-Clause
+  license_family: BSD
+  size: 379588
+  timestamp: 1719274858964
+- kind: conda
+  name: psutil
+  version: 6.0.0
+  build: py310ha8f682b_0
+  subdir: win-64
+  url: https://conda.anaconda.org/conda-forge/win-64/psutil-6.0.0-py310ha8f682b_0.conda
+  sha256: 9801a18aa6fadd3a6286fd89e83fe6affbcb3ca275bb2a00ab0da299d32e92ad
+  md5: 32f5673b7aa2309dda74ccd01822caca
+  depends:
+  - python >=3.10,<3.11.0a0
+  - python_abi 3.10.* *_cp310
+  - ucrt >=10.0.20348.0
+  - vc >=14.2,<15
+  - vc14_runtime >=14.29.30139
+  license: BSD-3-Clause
+  license_family: BSD
+  size: 388249
+  timestamp: 1719275165312
+- kind: conda
+  name: psutil
+  version: 6.0.0
+  build: py310hb52b2da_0
+  subdir: linux-aarch64
+  url: https://conda.anaconda.org/conda-forge/linux-aarch64/psutil-6.0.0-py310hb52b2da_0.conda
+  sha256: 8ba0f87bbcbbd9f73c9695522ffe2082f1562a356f59ead8b3d02ab17151fc52
+  md5: a75da26171bc7400ed580382b5c18196
+  depends:
+  - libgcc-ng >=12
+  - python >=3.10,<3.11.0a0
+  - python >=3.10,<3.11.0a0 *_cpython
+  - python_abi 3.10.* *_cp310
+  license: BSD-3-Clause
+  license_family: BSD
+  size: 372806
+  timestamp: 1719274753799
+- kind: conda
+  name: psutil
+  version: 6.0.0
+  build: py310hc51659f_0
+  subdir: linux-64
+  url: https://conda.anaconda.org/conda-forge/linux-64/psutil-6.0.0-py310hc51659f_0.conda
+  sha256: d23e0a2bf49a752fcc8267484c5eff3e5b267703853c11cc7b4f762412d0f7ef
+  md5: b04405826f96f4eb2f502e642d121bb5
+  depends:
+  - libgcc-ng >=12
+  - python >=3.10,<3.11.0a0
+  - python_abi 3.10.* *_cp310
+  license: BSD-3-Clause
+  license_family: BSD
+  size: 371633
+  timestamp: 1719274668659
+- kind: conda
+  name: pysocks
+  version: 1.7.1
+  build: pyh0701188_6
+  build_number: 6
+  subdir: noarch
+  noarch: python
+  url: https://conda.anaconda.org/conda-forge/noarch/pysocks-1.7.1-pyh0701188_6.tar.bz2
+  sha256: b3a612bc887f3dd0fb7c4199ad8e342bd148cf69a9b74fd9468a18cf2bef07b7
+  md5: 56cd9fe388baac0e90c7149cfac95b60
+  depends:
+  - __win
+  - python >=3.8
+  - win_inet_pton
+  license: BSD-3-Clause
+  license_family: BSD
+  size: 19348
+  timestamp: 1661605138291
+- kind: conda
+  name: pysocks
+  version: 1.7.1
+  build: pyha2e5f31_6
+  build_number: 6
+  subdir: noarch
+  noarch: python
+  url: https://conda.anaconda.org/conda-forge/noarch/pysocks-1.7.1-pyha2e5f31_6.tar.bz2
+  sha256: a42f826e958a8d22e65b3394f437af7332610e43ee313393d1cf143f0a2d274b
+  md5: 2a7de29fb590ca14b5243c4c812c8025
+  depends:
+  - __unix
+  - python >=3.8
+  license: BSD-3-Clause
+  license_family: BSD
+  size: 18981
+  timestamp: 1661604969727
+- kind: conda
+  name: python
+  version: 3.10.14
+  build: h00d2728_0_cpython
+  subdir: osx-64
+  url: https://conda.anaconda.org/conda-forge/osx-64/python-3.10.14-h00d2728_0_cpython.conda
+  sha256: 00c1de2d46ede26609ef4e84a44b83be7876ba6a0215b7c83bff41a0656bf694
+  md5: 0a1cddc4382c5c171e791c70740546dd
+  depends:
+  - bzip2 >=1.0.8,<2.0a0
+  - libffi >=3.4,<4.0a0
+  - libsqlite >=3.45.2,<4.0a0
+  - libzlib >=1.2.13,<2.0.0a0
+  - ncurses >=6.4.20240210,<7.0a0
+  - openssl >=3.2.1,<4.0a0
+  - readline >=8.2,<9.0a0
+  - tk >=8.6.13,<8.7.0a0
+  - tzdata
+  - xz >=5.2.6,<6.0a0
+  constrains:
+  - python_abi 3.10.* *_cp310
+  license: Python-2.0
+  size: 11890228
+  timestamp: 1710940046031
+- kind: conda
+  name: python
+  version: 3.10.14
+  build: h2469fbe_0_cpython
+  subdir: osx-arm64
+  url: https://conda.anaconda.org/conda-forge/osx-arm64/python-3.10.14-h2469fbe_0_cpython.conda
+  sha256: 454d609fe25daedce9e886efcbfcadad103ed0362e7cb6d2bcddec90b1ecd3ee
+  md5: 4ae999c8227c6d8c7623d32d51d25ea9
+  depends:
+  - bzip2 >=1.0.8,<2.0a0
+  - libffi >=3.4,<4.0a0
+  - libsqlite >=3.45.2,<4.0a0
+  - libzlib >=1.2.13,<2.0.0a0
+  - ncurses >=6.4.20240210,<7.0a0
+  - openssl >=3.2.1,<4.0a0
+  - readline >=8.2,<9.0a0
+  - tk >=8.6.13,<8.7.0a0
+  - tzdata
+  - xz >=5.2.6,<6.0a0
+  constrains:
+  - python_abi 3.10.* *_cp310
+  license: Python-2.0
+  size: 12336005
+  timestamp: 1710939659384
+- kind: conda
+  name: python
+  version: 3.10.14
+  build: h4de0772_0_cpython
+  subdir: win-64
+  url: https://conda.anaconda.org/conda-forge/win-64/python-3.10.14-h4de0772_0_cpython.conda
+  sha256: 332f97d9927b65857d6d2d4d50d66dce9b37da81edb67833ae6b88ad52acbd0c
+  md5: 4a00e84f29d1eb418d84970598c444e1
+  depends:
+  - bzip2 >=1.0.8,<2.0a0
+  - libffi >=3.4,<4.0a0
+  - libsqlite >=3.45.2,<4.0a0
+  - libzlib >=1.2.13,<2.0.0a0
+  - openssl >=3.2.1,<4.0a0
+  - tk >=8.6.13,<8.7.0a0
+  - tzdata
+  - vc >=14.1,<15
+  - vc14_runtime >=14.16.27033
+  - xz >=5.2.6,<6.0a0
+  constrains:
+  - python_abi 3.10.* *_cp310
+  license: Python-2.0
+  size: 15864027
+  timestamp: 1710938888352
+- kind: conda
+  name: python
+  version: 3.10.14
+  build: hbbe8eec_0_cpython
+  subdir: linux-aarch64
+  url: https://conda.anaconda.org/conda-forge/linux-aarch64/python-3.10.14-hbbe8eec_0_cpython.conda
+  sha256: 992583064b95d256e1b1f03581a51e225a425894d865e35ea2bf3017444c3e84
+  md5: 8a8ee3a8c62032c554debc785a3b5aba
+  depends:
+  - bzip2 >=1.0.8,<2.0a0
+  - ld_impl_linux-aarch64 >=2.36.1
+  - libffi >=3.4,<4.0a0
+  - libgcc-ng >=12
+  - libnsl >=2.0.1,<2.1.0a0
+  - libsqlite >=3.45.2,<4.0a0
+  - libuuid >=2.38.1,<3.0a0
+  - libxcrypt >=4.4.36
+  - libzlib >=1.2.13,<2.0.0a0
+  - ncurses >=6.4.20240210,<7.0a0
+  - openssl >=3.2.1,<4.0a0
+  - readline >=8.2,<9.0a0
+  - tk >=8.6.13,<8.7.0a0
+  - tzdata
+  - xz >=5.2.6,<6.0a0
+  constrains:
+  - python_abi 3.10.* *_cp310
+  license: Python-2.0
+  size: 13116477
+  timestamp: 1710971217224
+- kind: conda
+  name: python
+  version: 3.10.14
+  build: hd12c33a_0_cpython
+  subdir: linux-64
+  url: https://conda.anaconda.org/conda-forge/linux-64/python-3.10.14-hd12c33a_0_cpython.conda
+  sha256: 76a5d12e73542678b70a94570f7b0f7763f9a938f77f0e75d9ea615ef22aa84c
+  md5: 2b4ba962994e8bd4be9ff5b64b75aff2
+  depends:
+  - bzip2 >=1.0.8,<2.0a0
+  - ld_impl_linux-64 >=2.36.1
+  - libffi >=3.4,<4.0a0
+  - libgcc-ng >=12
+  - libnsl >=2.0.1,<2.1.0a0
+  - libsqlite >=3.45.2,<4.0a0
+  - libuuid >=2.38.1,<3.0a0
+  - libxcrypt >=4.4.36
+  - libzlib >=1.2.13,<2.0.0a0
+  - ncurses >=6.4.20240210,<7.0a0
+  - openssl >=3.2.1,<4.0a0
+  - readline >=8.2,<9.0a0
+  - tk >=8.6.13,<8.7.0a0
+  - tzdata
+  - xz >=5.2.6,<6.0a0
+  constrains:
+  - python_abi 3.10.* *_cp310
+  license: Python-2.0
+  size: 25517742
+  timestamp: 1710939725109
+- kind: conda
+  name: python_abi
+  version: '3.10'
+  build: 4_cp310
+  build_number: 4
+  subdir: linux-64
+  url: https://conda.anaconda.org/conda-forge/linux-64/python_abi-3.10-4_cp310.conda
+  sha256: 456bec815bfc2b364763084d08b412fdc4c17eb9ccc66a36cb775fa7ac3cbaec
+  md5: 26322ec5d7712c3ded99dd656142b8ce
+  constrains:
+  - python 3.10.* *_cpython
+  license: BSD-3-Clause
+  license_family: BSD
+  size: 6398
+  timestamp: 1695147363189
+- kind: conda
+  name: python_abi
+  version: '3.10'
+  build: 4_cp310
+  build_number: 4
+  subdir: linux-aarch64
+  url: https://conda.anaconda.org/conda-forge/linux-aarch64/python_abi-3.10-4_cp310.conda
+  sha256: 9191cc3ddf380b655c08b3436a8174ce0cc798a6dfcfa8ee80fa793d0b7165de
+  md5: b0ff2ed109650f9e90d627d3119eb442
+  constrains:
+  - python 3.10.* *_cpython
+  license: BSD-3-Clause
+  license_family: BSD
+  size: 6436
+  timestamp: 1695147402616
+- kind: conda
+  name: python_abi
+  version: '3.10'
+  build: 4_cp310
+  build_number: 4
+  subdir: osx-64
+  url: https://conda.anaconda.org/conda-forge/osx-64/python_abi-3.10-4_cp310.conda
+  sha256: abc26b3b5a62f9c8112a2303d24b0c590d5f7fc9470521f5a520472d59c2223e
+  md5: b15c816c5a86abcc4d1458dd63aa4c65
+  constrains:
+  - python 3.10.* *_cpython
+  license: BSD-3-Clause
+  license_family: BSD
+  size: 6484
+  timestamp: 1695147705581
+- kind: conda
+  name: python_abi
+  version: '3.10'
+  build: 4_cp310
+  build_number: 4
+  subdir: osx-arm64
+  url: https://conda.anaconda.org/conda-forge/osx-arm64/python_abi-3.10-4_cp310.conda
+  sha256: f69bac2f28082a275ef67313968b2c366d8236c3a6869b9cdf5cdb97a5821812
+  md5: 1a3d9c6bb5f0b1b22d9e9296c127e8c7
+  constrains:
+  - python 3.10.* *_cpython
+  license: BSD-3-Clause
+  license_family: BSD
+  size: 6490
+  timestamp: 1695147522999
+- kind: conda
+  name: python_abi
+  version: '3.10'
+  build: 4_cp310
+  build_number: 4
+  subdir: win-64
+  url: https://conda.anaconda.org/conda-forge/win-64/python_abi-3.10-4_cp310.conda
+  sha256: 19066c462fd0e32c64503c688f77cb603beb4019b812caf855d03f2a5447960b
+  md5: b41195997c14fb7473d26637ea4c3946
+  constrains:
+  - python 3.10.* *_cpython
+  license: BSD-3-Clause
+  license_family: BSD
+  size: 6773
+  timestamp: 1695147715814
+- kind: conda
+  name: readline
+  version: '8.2'
+  build: h8228510_1
+  build_number: 1
+  subdir: linux-64
+  url: https://conda.anaconda.org/conda-forge/linux-64/readline-8.2-h8228510_1.conda
+  sha256: 5435cf39d039387fbdc977b0a762357ea909a7694d9528ab40f005e9208744d7
+  md5: 47d31b792659ce70f470b5c82fdfb7a4
+  depends:
+  - libgcc-ng >=12
+  - ncurses >=6.3,<7.0a0
+  license: GPL-3.0-only
+  license_family: GPL
+  size: 281456
+  timestamp: 1679532220005
+- kind: conda
+  name: readline
+  version: '8.2'
+  build: h8fc344f_1
+  build_number: 1
+  subdir: linux-aarch64
+  url: https://conda.anaconda.org/conda-forge/linux-aarch64/readline-8.2-h8fc344f_1.conda
+  sha256: 4c99f7417419734e3797d45bc355e61c26520e111893b0d7087a01a7fbfbe3dd
+  md5: 105eb1e16bf83bfb2eb380a48032b655
+  depends:
+  - libgcc-ng >=12
+  - ncurses >=6.3,<7.0a0
+  license: GPL-3.0-only
+  license_family: GPL
+  size: 294092
+  timestamp: 1679532238805
+- kind: conda
+  name: readline
+  version: '8.2'
+  build: h92ec313_1
+  build_number: 1
+  subdir: osx-arm64
+  url: https://conda.anaconda.org/conda-forge/osx-arm64/readline-8.2-h92ec313_1.conda
+  sha256: a1dfa679ac3f6007362386576a704ad2d0d7a02e98f5d0b115f207a2da63e884
+  md5: 8cbb776a2f641b943d413b3e19df71f4
+  depends:
+  - ncurses >=6.3,<7.0a0
+  license: GPL-3.0-only
+  license_family: GPL
+  size: 250351
+  timestamp: 1679532511311
+- kind: conda
+  name: readline
+  version: '8.2'
+  build: h9e318b2_1
+  build_number: 1
+  subdir: osx-64
+  url: https://conda.anaconda.org/conda-forge/osx-64/readline-8.2-h9e318b2_1.conda
+  sha256: 41e7d30a097d9b060037f0c6a2b1d4c4ae7e942c06c943d23f9d481548478568
+  md5: f17f77f2acf4d344734bda76829ce14e
+  depends:
+  - ncurses >=6.3,<7.0a0
+  license: GPL-3.0-only
+  license_family: GPL
+  size: 255870
+  timestamp: 1679532707590
+- kind: conda
+  name: ruff
+  version: 0.3.7
+  build: py310h298983d_0
+  subdir: win-64
+  url: https://conda.anaconda.org/conda-forge/win-64/ruff-0.3.7-py310h298983d_0.conda
+  sha256: fd0766092fb869c3c591da324cb8e675e9cd01e220dc72b9f2b857e92337c01d
+  md5: 9a5e1425ea8eac4f79a275c20d859cac
+  depends:
+  - python >=3.10,<3.11.0a0
+  - python_abi 3.10.* *_cp310
+  - ucrt >=10.0.20348.0
+  - vc >=14.2,<15
+  - vc14_runtime >=14.29.30139
+  license: MIT
+  license_family: MIT
+  size: 6353965
+  timestamp: 1712963138812
+- kind: conda
+  name: ruff
+  version: 0.3.7
+  build: py310h3d77a66_0
+  subdir: linux-64
+  url: https://conda.anaconda.org/conda-forge/linux-64/ruff-0.3.7-py310h3d77a66_0.conda
+  sha256: fc66dae77831b8110cd20ee257b462b4471204d310fc438d9760b769799969d9
+  md5: 55b40e33fae0b983b48d2f7aff8a8978
+  depends:
+  - libgcc-ng >=12
+  - libstdcxx-ng >=12
+  - python >=3.10,<3.11.0a0
+  - python_abi 3.10.* *_cp310
+  license: MIT
+  license_family: MIT
+  size: 6402959
+  timestamp: 1712962093948
+- kind: conda
+  name: ruff
+  version: 0.3.7
+  build: py310h81561d7_0
+  subdir: osx-arm64
+  url: https://conda.anaconda.org/conda-forge/osx-arm64/ruff-0.3.7-py310h81561d7_0.conda
+  sha256: b7c88aae99ef73ffac7886e4dbb02428230b66284fa360134985bb1741f8d569
+  md5: 4d922de656cdef20fb9fd47f0529bb95
+  depends:
+  - libcxx >=16
+  - python >=3.10,<3.11.0a0
+  - python >=3.10,<3.11.0a0 *_cpython
+  - python_abi 3.10.* *_cp310
+  constrains:
+  - __osx >=11.0
+  license: MIT
+  license_family: MIT
+  size: 5882969
+  timestamp: 1712963596623
+- kind: conda
+  name: ruff
+  version: 0.3.7
+  build: py310hdac29b7_0
+  subdir: osx-64
+  url: https://conda.anaconda.org/conda-forge/osx-64/ruff-0.3.7-py310hdac29b7_0.conda
+  sha256: 1d279579eb3282973102f3c903ef21b3b033a8635d00d08ff40959871baf91dc
+  md5: ab5f005a86062035d3eaf1adb081cc26
+  depends:
+  - libcxx >=16
+  - python >=3.10,<3.11.0a0
+  - python_abi 3.10.* *_cp310
+  constrains:
+  - __osx >=10.12
+  license: MIT
+  license_family: MIT
+  size: 6192290
+  timestamp: 1712963577697
+- kind: conda
+  name: ruff
+  version: 0.3.7
+  build: py310hf6424b7_0
+  subdir: linux-aarch64
+  url: https://conda.anaconda.org/conda-forge/linux-aarch64/ruff-0.3.7-py310hf6424b7_0.conda
+  sha256: 45cf4cfcaf97598959077d93e698db15aa17a7e215c5b050738367def0227e02
+  md5: c6dcc788fee5fefe50165e9c0353d065
+  depends:
+  - libgcc-ng >=12
+  - libstdcxx-ng >=12
+  - python >=3.10,<3.11.0a0
+  - python >=3.10,<3.11.0a0 *_cpython
+  - python_abi 3.10.* *_cp310
+  license: MIT
+  license_family: MIT
+  size: 6047252
+  timestamp: 1712962189494
+- kind: conda
+  name: setuptools
+  version: 70.1.1
+  build: pyhd8ed1ab_0
+  subdir: noarch
+  noarch: python
+  url: https://conda.anaconda.org/conda-forge/noarch/setuptools-70.1.1-pyhd8ed1ab_0.conda
+  sha256: 34ecbc63df6052a320838335a0e594b60050c92de79254045e52095bc27dde03
+  md5: 985e9e86e1b0fc75a74a9bfab9309ef7
+  depends:
+  - python >=3.8
+  license: MIT
+  license_family: MIT
+  size: 496940
+  timestamp: 1719325175003
+- kind: conda
+  name: tk
+  version: 8.6.13
+  build: h194ca79_0
+  subdir: linux-aarch64
+  url: https://conda.anaconda.org/conda-forge/linux-aarch64/tk-8.6.13-h194ca79_0.conda
+  sha256: 7fa27cc512d3a783f38bd16bbbffc008807372499d5b65d089a8e43bde9db267
+  md5: f75105e0585851f818e0009dd1dde4dc
+  depends:
+  - libgcc-ng >=12
+  - libzlib >=1.2.13,<1.3.0a0
+  license: TCL
+  license_family: BSD
+  size: 3351802
+  timestamp: 1695506242997
+- kind: conda
+  name: tk
+  version: 8.6.13
+  build: h1abcd95_1
+  build_number: 1
+  subdir: osx-64
+  url: https://conda.anaconda.org/conda-forge/osx-64/tk-8.6.13-h1abcd95_1.conda
+  sha256: 30412b2e9de4ff82d8c2a7e5d06a15f4f4fef1809a72138b6ccb53a33b26faf5
+  md5: bf830ba5afc507c6232d4ef0fb1a882d
+  depends:
+  - libzlib >=1.2.13,<1.3.0a0
+  license: TCL
+  license_family: BSD
+  size: 3270220
+  timestamp: 1699202389792
+- kind: conda
+  name: tk
+  version: 8.6.13
+  build: h5083fa2_1
+  build_number: 1
+  subdir: osx-arm64
+  url: https://conda.anaconda.org/conda-forge/osx-arm64/tk-8.6.13-h5083fa2_1.conda
+  sha256: 72457ad031b4c048e5891f3f6cb27a53cb479db68a52d965f796910e71a403a8
+  md5: b50a57ba89c32b62428b71a875291c9b
+  depends:
+  - libzlib >=1.2.13,<1.3.0a0
+  license: TCL
+  license_family: BSD
+  size: 3145523
+  timestamp: 1699202432999
+- kind: conda
+  name: tk
+  version: 8.6.13
+  build: h5226925_1
+  build_number: 1
+  subdir: win-64
+  url: https://conda.anaconda.org/conda-forge/win-64/tk-8.6.13-h5226925_1.conda
+  sha256: 2c4e914f521ccb2718946645108c9bd3fc3216ba69aea20c2c3cedbd8db32bb1
+  md5: fc048363eb8f03cd1737600a5d08aafe
+  depends:
+  - ucrt >=10.0.20348.0
+  - vc >=14.2,<15
+  - vc14_runtime >=14.29.30139
+  license: TCL
+  license_family: BSD
+  size: 3503410
+  timestamp: 1699202577803
+- kind: conda
+  name: tk
+  version: 8.6.13
+  build: noxft_h4845f30_101
+  build_number: 101
+  subdir: linux-64
+  url: https://conda.anaconda.org/conda-forge/linux-64/tk-8.6.13-noxft_h4845f30_101.conda
+  sha256: e0569c9caa68bf476bead1bed3d79650bb080b532c64a4af7d8ca286c08dea4e
+  md5: d453b98d9c83e71da0741bb0ff4d76bc
+  depends:
+  - libgcc-ng >=12
+  - libzlib >=1.2.13,<1.3.0a0
+  license: TCL
+  license_family: BSD
+  size: 3318875
+  timestamp: 1699202167581
+- kind: conda
+  name: tomli
+  version: 2.0.1
+  build: pyhd8ed1ab_0
+  subdir: noarch
+  noarch: python
+  url: https://conda.anaconda.org/conda-forge/noarch/tomli-2.0.1-pyhd8ed1ab_0.tar.bz2
+  sha256: 4cd48aba7cd026d17e86886af48d0d2ebc67ed36f87f6534f4b67138f5a5a58f
+  md5: 5844808ffab9ebdb694585b50ba02a96
+  depends:
+  - python >=3.7
+  license: MIT
+  license_family: MIT
+  size: 15940
+  timestamp: 1644342331069
+- kind: conda
+  name: types-requests
+  version: 2.31.0.20240406
+  build: pyhd8ed1ab_0
+  subdir: noarch
+  noarch: python
+  url: https://conda.anaconda.org/conda-forge/noarch/types-requests-2.31.0.20240406-pyhd8ed1ab_0.conda
+  sha256: de93470fe64b2baa5f8ef16a6edf849bb93542f301ed343d0ab4d6fd6116d742
+  md5: b4bc9b6dbc54191100b518a18be6045e
+  depends:
+  - python >=3.6
+  - urllib3 >=2
+  license: Apache-2.0 AND MIT
+  size: 26072
+  timestamp: 1712378106245
+- kind: conda
+  name: typing_extensions
+  version: 4.11.0
+  build: pyha770c72_0
+  subdir: noarch
+  noarch: python
+  url: https://conda.anaconda.org/conda-forge/noarch/typing_extensions-4.11.0-pyha770c72_0.conda
+  sha256: a7e8714d14f854058e971a6ed44f18cc37cc685f98ddefb2e6b7899a0cc4d1a2
+  md5: 6ef2fc37559256cf682d8b3375e89b80
+  depends:
+  - python >=3.8
+  license: PSF-2.0
+  license_family: PSF
+  size: 37583
+  timestamp: 1712330089194
+- kind: conda
+  name: typos
+  version: 1.20.9
+  build: h11a7dfb_0
+  subdir: osx-64
+  url: https://conda.anaconda.org/conda-forge/osx-64/typos-1.20.9-h11a7dfb_0.conda
+  sha256: c6345dad41706c3f925a3c76308e2af390ed3ce8eb92b84ebb82a37f2f599d1e
+  md5: 9e8f7d03be6ae3eb50f14fddffc7fe51
+  constrains:
+  - __osx >=10.12
+  license: MIT
+  license_family: MIT
+  size: 3347431
+  timestamp: 1713321633495
+- kind: conda
+  name: typos
+  version: 1.20.9
+  build: h1d8f897_0
+  subdir: linux-aarch64
+  url: https://conda.anaconda.org/conda-forge/linux-aarch64/typos-1.20.9-h1d8f897_0.conda
+  sha256: d98e8e5fb65b4d09cff752534418acf3e835f702003636ef2f2165039ddbb521
+  md5: 519bac2b977d194cc65f32c72bea5d12
+  depends:
+  - libgcc-ng >=12
+  license: MIT
+  license_family: MIT
+  size: 3561452
+  timestamp: 1713321791172
+- kind: conda
+  name: typos
+  version: 1.20.9
+  build: h5ef7bb8_0
+  subdir: osx-arm64
+  url: https://conda.anaconda.org/conda-forge/osx-arm64/typos-1.20.9-h5ef7bb8_0.conda
+  sha256: 42e47f16457a6193658e3052c2dbabbdcb5759f79c8f83eb06eb3af91287c18e
+  md5: f22e8b9559961907af67f4ec6d612456
+  constrains:
+  - __osx >=11.0
+  license: MIT
+  license_family: MIT
+  size: 3331993
+  timestamp: 1713321682918
+- kind: conda
+  name: typos
+  version: 1.20.9
+  build: h7f3b576_0
+  subdir: win-64
+  url: https://conda.anaconda.org/conda-forge/win-64/typos-1.20.9-h7f3b576_0.conda
+  sha256: ff9db9be0d78163b9b26ab6169e84d6026c05552a51bb335bbe62e0f4680cfd9
+  md5: 9e6937e2784adfe9088d8ee3ba944a07
+  depends:
+  - m2w64-gcc-libs
+  - m2w64-gcc-libs-core
+  license: MIT
+  license_family: MIT
+  size: 2596487
+  timestamp: 1713321776083
+- kind: conda
+  name: typos
+  version: 1.20.9
+  build: he8a937b_0
+  subdir: linux-64
+  url: https://conda.anaconda.org/conda-forge/linux-64/typos-1.20.9-he8a937b_0.conda
+  sha256: 2fc3888119f17b675edca08e1927f7f3445ec8c5ceb7841fc5dd6f20d8373fdd
+  md5: fc85f19eb1191bb7e6bf34d4b2ec39e6
+  depends:
+  - libgcc-ng >=12
+  license: MIT
+  license_family: MIT
+  size: 3728370
+  timestamp: 1713320476655
+- kind: conda
+  name: tzdata
+  version: 2024a
+  build: h0c530f3_0
+  subdir: noarch
+  noarch: generic
+  url: https://conda.anaconda.org/conda-forge/noarch/tzdata-2024a-h0c530f3_0.conda
+  sha256: 7b2b69c54ec62a243eb6fba2391b5e443421608c3ae5dbff938ad33ca8db5122
+  md5: 161081fc7cec0bfda0d86d7cb595f8d8
+  license: LicenseRef-Public-Domain
+  size: 119815
+  timestamp: 1706886945727
+- kind: conda
+  name: ucrt
+  version: 10.0.22621.0
+  build: h57928b3_0
+  subdir: win-64
+  url: https://conda.anaconda.org/conda-forge/win-64/ucrt-10.0.22621.0-h57928b3_0.tar.bz2
+  sha256: f29cdaf8712008f6b419b8b1a403923b00ab2504bfe0fb2ba8eb60e72d4f14c6
+  md5: 72608f6cd3e5898229c3ea16deb1ac43
+  constrains:
+  - vs2015_runtime >=14.29.30037
+  license: LicenseRef-Proprietary
+  license_family: PROPRIETARY
+  size: 1283972
+  timestamp: 1666630199266
+- kind: conda
+  name: urllib3
+  version: 2.2.1
+  build: pyhd8ed1ab_0
+  subdir: noarch
+  noarch: python
+  url: https://conda.anaconda.org/conda-forge/noarch/urllib3-2.2.1-pyhd8ed1ab_0.conda
+  sha256: d4009dcc9327684d6409706ce17656afbeae690d8522d3c9bc4df57649a352cd
+  md5: 08807a87fa7af10754d46f63b368e016
+  depends:
+  - brotli-python >=1.0.9
+  - pysocks >=1.5.6,<2.0,!=1.5.7
+  - python >=3.7
+  license: MIT
+  license_family: MIT
+  size: 94669
+  timestamp: 1708239595549
+- kind: conda
+  name: vc
+  version: '14.3'
+  build: hcf57466_18
+  build_number: 18
+  subdir: win-64
+  url: https://conda.anaconda.org/conda-forge/win-64/vc-14.3-hcf57466_18.conda
+  sha256: 447a8d8292a7b2107dcc18afb67f046824711a652725fc0f522c368e7a7b8318
+  md5: 20e1e652a4c740fa719002a8449994a2
+  depends:
+  - vc14_runtime >=14.38.33130
+  track_features:
+  - vc14
+  license: BSD-3-Clause
+  license_family: BSD
+  size: 16977
+  timestamp: 1702511255313
+- kind: conda
+  name: vc14_runtime
+  version: 14.38.33130
+  build: h82b7239_18
+  build_number: 18
+  subdir: win-64
+  url: https://conda.anaconda.org/conda-forge/win-64/vc14_runtime-14.38.33130-h82b7239_18.conda
+  sha256: bf94c9af4b2e9cba88207001197e695934eadc96a5c5e4cd7597e950aae3d8ff
+  md5: 8be79fdd2725ddf7bbf8a27a4c1f79ba
+  depends:
+  - ucrt >=10.0.20348.0
+  constrains:
+  - vs2015_runtime 14.38.33130.* *_18
+  license: LicenseRef-ProprietaryMicrosoft
+  license_family: Proprietary
+  size: 749868
+  timestamp: 1702511239004
+- kind: conda
+  name: vs2015_runtime
+  version: 14.38.33130
+  build: hcb4865c_18
+  build_number: 18
+  subdir: win-64
+  url: https://conda.anaconda.org/conda-forge/win-64/vs2015_runtime-14.38.33130-hcb4865c_18.conda
+  sha256: a2fec221f361d6263c117f4ea6d772b21c90a2f8edc6f3eb0eadec6bfe8843db
+  md5: 10d42885e3ed84e575b454db30f1aa93
+  depends:
+  - vc14_runtime >=14.38.33130
+  license: BSD-3-Clause
+  license_family: BSD
+  size: 16988
+  timestamp: 1702511261442
+- kind: conda
+  name: wheel
+  version: 0.43.0
+  build: pyhd8ed1ab_1
+  build_number: 1
+  subdir: noarch
+  noarch: python
+  url: https://conda.anaconda.org/conda-forge/noarch/wheel-0.43.0-pyhd8ed1ab_1.conda
+  sha256: cb318f066afd6fd64619f14c030569faf3f53e6f50abf743b4c865e7d95b96bc
+  md5: 0b5293a157c2b5cd513dd1b03d8d3aae
+  depends:
+  - python >=3.8
+  license: MIT
+  license_family: MIT
+  size: 57963
+  timestamp: 1711546009410
+- kind: conda
+  name: win_inet_pton
+  version: 1.1.0
+  build: pyhd8ed1ab_6
+  build_number: 6
+  subdir: noarch
+  noarch: python
+  url: https://conda.anaconda.org/conda-forge/noarch/win_inet_pton-1.1.0-pyhd8ed1ab_6.tar.bz2
+  sha256: a11ae693a0645bf6c7b8a47bac030be9c0967d0b1924537b9ff7458e832c0511
+  md5: 30878ecc4bd36e8deeea1e3c151b2e0b
+  depends:
+  - __win
+  - python >=3.6
+  license: PUBLIC-DOMAIN
+  size: 8191
+  timestamp: 1667051294134
+- kind: conda
+  name: xz
+  version: 5.2.6
+  build: h166bdaf_0
+  subdir: linux-64
+  url: https://conda.anaconda.org/conda-forge/linux-64/xz-5.2.6-h166bdaf_0.tar.bz2
+  sha256: 03a6d28ded42af8a347345f82f3eebdd6807a08526d47899a42d62d319609162
+  md5: 2161070d867d1b1204ea749c8eec4ef0
+  depends:
+  - libgcc-ng >=12
+  license: LGPL-2.1 and GPL-2.0
+  size: 418368
+  timestamp: 1660346797927
+- kind: conda
+  name: xz
+  version: 5.2.6
+  build: h57fd34a_0
+  subdir: osx-arm64
+  url: https://conda.anaconda.org/conda-forge/osx-arm64/xz-5.2.6-h57fd34a_0.tar.bz2
+  sha256: 59d78af0c3e071021cfe82dc40134c19dab8cdf804324b62940f5c8cd71803ec
+  md5: 39c6b54e94014701dd157f4f576ed211
+  license: LGPL-2.1 and GPL-2.0
+  size: 235693
+  timestamp: 1660346961024
+- kind: conda
+  name: xz
+  version: 5.2.6
+  build: h775f41a_0
+  subdir: osx-64
+  url: https://conda.anaconda.org/conda-forge/osx-64/xz-5.2.6-h775f41a_0.tar.bz2
+  sha256: eb09823f34cc2dd663c0ec4ab13f246f45dcd52e5b8c47b9864361de5204a1c8
+  md5: a72f9d4ea13d55d745ff1ed594747f10
+  license: LGPL-2.1 and GPL-2.0
+  size: 238119
+  timestamp: 1660346964847
+- kind: conda
+  name: xz
+  version: 5.2.6
+  build: h8d14728_0
+  subdir: win-64
+  url: https://conda.anaconda.org/conda-forge/win-64/xz-5.2.6-h8d14728_0.tar.bz2
+  sha256: 54d9778f75a02723784dc63aff4126ff6e6749ba21d11a6d03c1f4775f269fe0
+  md5: 515d77642eaa3639413c6b1bc3f94219
+  depends:
+  - vc >=14.1,<15
+  - vs2015_runtime >=14.16.27033
+  license: LGPL-2.1 and GPL-2.0
+  size: 217804
+  timestamp: 1660346976440
+- kind: conda
+  name: xz
+  version: 5.2.6
+  build: h9cdd2b7_0
+  subdir: linux-aarch64
+  url: https://conda.anaconda.org/conda-forge/linux-aarch64/xz-5.2.6-h9cdd2b7_0.tar.bz2
+  sha256: 93f58a7b393adf41fa007ac8c55978765e957e90cd31877ece1e5a343cb98220
+  md5: 83baad393a31d59c20b63ba4da6592df
+  depends:
+  - libgcc-ng >=12
+  license: LGPL-2.1 and GPL-2.0
+  size: 440555
+  timestamp: 1660348056328

pixi.toml

@@ -0,0 +1,57 @@
+# Pixi is a package management tool for developers.
+# Before running a task, pixi ensures that all listed dependencies are installed first.echop
+#
+# Pixi is not required for rerun, but it is a convenient way to install the
+# dependencies required for this example.
+#
+# https://prefix.dev/docs/pixi/overview
+#
+# Use `pixi task list` to list the available tasks,
+# and `pixi run TASK` to run it (e.g. `pixi run example`).
+
+[project]
+name = "rerun_vista_example"
+authors = ["rerun.io <opensource@rerun.io>"]
+channels = ["conda-forge"]
+description = "Visualizing the Vista model with Rerun."
+homepage = "https://rerun.io"
+license = "MIT OR Apache-2.0"
+
+platforms = ["linux-64", "linux-aarch64", "osx-arm64", "osx-64", "win-64"]
+readme = "README.md"
+repository = "https://github.com/rerun-io/hf-example-vista"
+version = "0.1.0"
+
+
+[tasks]
+# ------------------------------------------------------------------------------------------
+# Python stuff:
+
+# Run first ruff fix, then ruff format, order is important see also https://twitter.com/charliermarsh/status/1717229721954799727
+py-fmt = "ruff check --fix --config pyproject.toml . && ruff format --config pyproject.toml ."
+py-fmt-check = "ruff check --config pyproject.toml . && ruff format --check --config pyproject.toml"
+py-lint = "mypy --install-types --non-interactive --no-warn-unused-ignore"
+
+# ------------------------------------------------------------------------------------------
+# General stuff:
+lint-typos = "typos"
+
+# ------------------------------------------------------------------------------------------
+install-dependencies = "pip install -r requirements.txt"
+
+[tasks.example]
+cmd = "python main.py"
+depends_on = ["install-dependencies"]
+
+
+[dependencies]
+# Python stuff:
+mypy = "1.8.0"
+ruff = "0.3.7"
+python = "3.10.*"
+pip = ">=24.0,<25" # to install dependencies from requirements.txt
+
+types-requests = ">=2.31,<3" # mypy type hint stubs for generate_changelog.py
+
+# General stuff:
+typos = ">=1.16.20"

pyproject.toml

@@ -0,0 +1,72 @@
+# Copied from https://github.com/rerun-io/rerun_template
+
+[tool.ruff]
+# https://beta.ruff.rs/docs/configuration/
+
+target-version = "py38"
+
+# Enable unsafe fixes to allow ruff to apply fixes that may change the behavior of the code.
+# This is needed because otherwise ruff will not be able to trim whitespaces in (codegened) docstrings.
+unsafe-fixes = true
+
+# Allow preview lints to be enabled (like `PLW1514` to force `encoding` on open).
+preview = true
+# But we only want to opt-in to certain preview rules!
+lint.explicit-preview-rules = true
+
+extend-exclude = [
+    # Automatically generated test artifacts
+    "venv/",
+    "target/",
+]
+
+lint.ignore = [
+    # These makes sense to ignore in example code, but for a proper library we should not ignore these.
+    "D100", # Missing docstring in public module
+    "D101", # Missing docstring in public class
+    "D103", # Missing docstring in public function
+
+    # No blank lines allowed after function docstring.
+    "D202",
+
+    # npydocstyle: http://www.pydocstyle.org/en/stable/error_codes.html
+    # numpy convention with a few additional lints
+    "D107",
+    "D203",
+    "D212",
+    "D401",
+    "D402",
+    "D415",
+    "D416",
+
+    # Ruff can't fix this error on its own (yet)
+    # Having ruff check this causes errors that prevent the code-formatting process from completing.
+    "E501",
+
+    # allow relative imports
+    "TID252",
+
+    "UP007", # We need this, or `ruff format` will convert `Union[X, Y]` to `X | Y` which break on Python 3.8
+]
+
+line-length = 120
+lint.select = [
+    "D",       # pydocstyle codes https://www.pydocstyle.org/en/latest/error_codes.html
+    "E",       # pycodestyle error codes: https://pycodestyle.pycqa.org/en/latest/intro.html#error-codes
+    "F",       # Flake8 error codes https://flake8.pycqa.org/en/latest/user/error-codes.html
+    "I",       # Isort
+    "TID",     # flake8-tidy-imports
+    "W",       # pycodestyle warning codes: https://pycodestyle.pycqa.org/en/latest/intro.html#error-codes
+    "UP",      # pyupgrade (ensures idomatic code for supported python version)
+    "PLW1514", # Force setting `encoding` for open calls. This is in order to prevent issues when opening utf8 files on windows where the default encoding may depend on the active locale. https://docs.astral.sh/ruff/rules/unspecified-encoding/
+]
+
+lint.unfixable = [
+    "PLW1514", # Automatic fix for `encoding` doesn't do what we want - it queries the locale for the preferred encoding which is exactly what we want to avoid.
+]
+
+[tool.ruff.lint.per-file-ignores]
+"__init__.py" = ["F401", "F403"]
+
+[tool.ruff.lint.isort]
+required-imports = ["from __future__ import annotations"]

requirements.txt

@@ -0,0 +1,168 @@
+aiofiles==23.2.1
+aiohttp==3.9.5
+aiosignal==1.3.1
+altair==5.3.0
+annotated-types==0.7.0
+antlr4-python3-runtime==4.9.3
+anyio==4.4.0
+async-timeout==4.0.3
+attrs==23.2.0
+black==23.7.0
+blinker==1.8.2
+braceexpand==0.1.7
+cachetools==5.3.3
+certifi==2024.6.2
+chardet==5.1.0
+charset-normalizer==3.3.2
+click==8.1.7
+clip @ git+https://github.com/openai/CLIP.git
+cmake==3.29.5.1
+contourpy==1.2.1
+cycler==0.12.1
+deepspeed
+dnspython==2.6.1
+docker-pycreds==0.4.0
+einops==0.8.0
+email_validator==2.1.1
+exceptiongroup==1.2.1
+fairscale==0.4.13
+fastapi==0.111.0
+fastapi-cli==0.0.4
+ffmpy==0.3.2
+filelock==3.15.1
+fire==0.6.0
+fonttools==4.53.0
+frozenlist==1.4.1
+fsspec==2024.6.0
+ftfy==6.2.0
+gitdb==4.0.11
+GitPython==3.1.43
+gradio==4.26.0
+gradio_client==0.15.1
+gradio_rerun==0.0.3
+h11==0.14.0
+hjson==3.1.0
+httpcore==1.0.5
+httptools==0.6.1
+httpx==0.27.0
+huggingface-hub==0.23.3
+idna==3.7
+imageio==2.31.1
+imageio-ffmpeg==0.4.8
+importlib_resources==6.4.0
+invisible-watermark==0.2.0
+jedi==0.19.1
+Jinja2==3.1.4
+jsonschema==4.22.0
+jsonschema-specifications==2023.12.1
+kiwisolver==1.4.5
+kornia==0.7.2
+kornia_rs==0.1.3
+lightning-utilities==0.11.2
+lit==18.1.7
+markdown-it-py==3.0.0
+MarkupSafe==2.1.5
+matplotlib==3.9.0
+mdurl==0.1.2
+mpmath==1.3.0
+multidict==6.0.5
+mypy-extensions==1.0.0
+natsort==8.4.0
+networkx==3.3
+ninja==1.11.1.1
+numpy==1.26.4
+nvidia-cublas-cu11==11.10.3.66
+nvidia-cuda-cupti-cu11==11.7.101
+nvidia-cuda-nvrtc-cu11==11.7.99
+nvidia-cuda-runtime-cu11==11.7.99
+nvidia-cudnn-cu11==8.5.0.96
+nvidia-cufft-cu11==10.9.0.58
+nvidia-curand-cu11==10.2.10.91
+nvidia-cusolver-cu11==11.4.0.1
+nvidia-cusparse-cu11==11.7.4.91
+nvidia-ml-py==12.555.43
+nvidia-nccl-cu11==2.14.3
+nvidia-nvtx-cu11==11.7.91
+omegaconf==2.3.0
+open-clip-torch==2.24.0
+opencv-python==4.6.0.66
+orjson==3.10.4
+packaging==24.1
+pandas==2.2.2
+parso==0.8.4
+pathspec==0.12.1
+pillow==10.3.0
+platformdirs==4.2.2
+protobuf==3.20.3
+psutil==5.9.8
+pudb==2024.1
+py-cpuinfo==9.0.0
+pyarrow==16.1.0
+pydantic==2.7.4
+pydantic_core==2.18.4
+pydeck==0.9.1
+pydub==0.25.1
+Pygments==2.18.0
+pyparsing==3.1.2
+python-dateutil==2.9.0.post0
+python-dotenv==1.0.1
+python-multipart==0.0.9
+pytorch-lightning==2.0.1
+pytz==2024.1
+PyWavelets==1.6.0
+PyYAML==6.0.1
+referencing==0.35.1
+regex==2024.5.15
+requests==2.32.3
+rerun-sdk==0.16.1
+rich==13.7.1
+rpds-py==0.18.1
+ruff==0.4.8
+safetensors==0.4.3
+scipy==1.13.1
+semantic-version==2.10.0
+sentencepiece==0.2.0
+sentry-sdk==2.5.1
+setproctitle==1.3.3
+shellingham==1.5.4
+six==1.16.0
+smmap==5.0.1
+sniffio==1.3.1
+spaces==0.28.3
+starlette==0.37.2
+streamlit==1.35.0
+sympy==1.12.1
+tensorboardX==2.6
+termcolor==2.4.0
+timm==1.0.3
+tokenizers==0.12.1
+toml==0.10.2
+tomli==2.0.1
+tomlkit==0.12.0
+toolz==0.12.1
+torch==2.0.1
+torchaudio==2.0.2
+torchdata==0.6.1
+torchmetrics==1.4.0.post0
+torchvision==0.15.2
+tornado==6.4.1
+tqdm==4.66.4
+transformers==4.19.1
+triton==2.0.0
+typer==0.12.3
+typing_extensions==4.12.2
+tzdata==2024.1
+ujson==5.10.0
+urllib3==1.26.18
+urwid==2.6.14
+urwid_readline==0.14
+uvicorn==0.30.1
+uvloop==0.19.0
+wandb==0.17.1
+watchdog==4.0.1
+watchfiles==0.22.0
+wcwidth==0.2.13
+webdataset==0.2.86
+websockets==11.0.3
+xformers==0.0.22
+yarl==1.9.4

scripts/template_update.py

@@ -0,0 +1,191 @@
+#!/usr/bin/env python3
+# Copied from https://github.com/rerun-io/rerun_template
+
+"""
+The script has two purposes.
+
+After using `rerun_template` as a template, run this to clean out things you don't need.
+Use `scripts/template_update.py init --languages cpp,rust,python` for this.
+
+Update an existing repository with the latest changes from the template.
+Use `scripts/template_update.py update --languages cpp,rust,python` for this.
+
+In either case, make sure the list of languages matches the languages you want to support.
+You can also use `--dry-run` to see what would happen without actually changing anything.
+"""
+
+from __future__ import annotations
+
+import argparse
+import os
+import shutil
+import tempfile
+
+from git import Repo  # pip install GitPython
+
+OWNER = "rerun-io"
+
+# Don't overwrite these when updating existing repository from the template
+DO_NOT_OVERWRITE = {
+    "Cargo.lock",
+    "CHANGELOG.md",
+    "main.py",
+    "pixi.lock",
+    "README.md",
+    "requirements.txt",
+}
+
+# Files required by C++, but not by _both_ Python and Rust
+CPP_FILES = {
+    ".clang-format",
+    ".github/workflows/cpp.yml",
+    "CMakeLists.txt",
+    "pixi.lock",  # Pixi is only C++ & Python - For Rust we only use cargo
+    "pixi.toml",  # Pixi is only C++ & Python - For Rust we only use cargo
+    "src/",
+    "src/main.cpp",
+}
+
+# Files required by Python, but not by _both_ C++ and Rust
+PYTHON_FILES = {
+    ".github/workflows/python.yml",
+    ".mypy.ini",
+    "main.py",
+    "pixi.lock",  # Pixi is only C++ & Python - For Rust we only use cargo
+    "pixi.toml",  # Pixi is only C++ & Python - For Rust we only use cargo
+    "pyproject.toml",
+    "requirements.txt",
+}
+
+# Files required by Rust, but not by _both_ C++ and Python
+RUST_FILES = {
+    ".github/workflows/rust.yml",
+    "bacon.toml",
+    "Cargo.lock",
+    "Cargo.toml",
+    "CHANGELOG.md",  # We only keep a changelog for Rust crates at the moment
+    "clippy.toml",
+    "Cranky.toml",
+    "deny.toml",
+    "rust-toolchain",
+    "scripts/clippy_wasm/",
+    "scripts/clippy_wasm/clippy.toml",
+    "scripts/generate_changelog.py",  # We only keep a changelog for Rust crates at the moment
+    "src/",
+    "src/lib.rs",
+    "src/main.rs",
+}
+
+# Files we used to have, but have been removed in never version of rerun_template
+DEAD_FILES = ["bacon.toml", "Cranky.toml"]
+
+
+def parse_languages(lang_str: str) -> set[str]:
+    languages = lang_str.split(",") if lang_str else []
+    for lang in languages:
+        assert lang in ["cpp", "python", "rust"], f"Unsupported language: {lang}"
+    return set(languages)
+
+
+def calc_deny_set(languages: set[str]) -> set[str]:
+    """The set of files to delete/ignore."""
+    files_to_delete = CPP_FILES | PYTHON_FILES | RUST_FILES
+    if "cpp" in languages:
+        files_to_delete -= CPP_FILES
+    if "python" in languages:
+        files_to_delete -= PYTHON_FILES
+    if "rust" in languages:
+        files_to_delete -= RUST_FILES
+    return files_to_delete
+
+
+def init(languages: set[str], dry_run: bool) -> None:
+    print("Removing all language-specific files not needed for languages {languages}.")
+    files_to_delete = calc_deny_set(languages)
+    delete_files_and_folder(files_to_delete, dry_run)
+
+
+def remove_file(filepath: str):
+    try:
+        os.remove(filepath)
+    except FileNotFoundError:
+        pass
+
+
+def delete_files_and_folder(paths: set[str], dry_run: bool) -> None:
+    repo_path = os.path.dirname(os.path.dirname(os.path.realpath(__file__)))
+    for path in paths:
+        full_path = os.path.join(repo_path, path)
+        if os.path.exists(full_path):
+            if os.path.isfile(full_path):
+                print(f"Removing file {full_path}…")
+                if not dry_run:
+                    remove_file(full_path)
+            elif os.path.isdir(full_path):
+                print(f"Removing folder {full_path}…")
+                if not dry_run:
+                    shutil.rmtree(full_path)
+
+
+def update(languages: set[str], dry_run: bool) -> None:
+    for file in DEAD_FILES:
+        print(f"Removing dead file {file}…")
+        if not dry_run:
+            remove_file(file)
+
+    files_to_ignore = calc_deny_set(languages) | DO_NOT_OVERWRITE
+    repo_path = os.path.dirname(os.path.dirname(os.path.realpath(__file__)))
+
+    with tempfile.TemporaryDirectory() as temp_dir:
+        Repo.clone_from("https://github.com/rerun-io/rerun_template.git", temp_dir)
+        for root, dirs, files in os.walk(temp_dir):
+            for file in files:
+                src_path = os.path.join(root, file)
+                rel_path = os.path.relpath(src_path, temp_dir)
+
+                if rel_path.startswith(".git/"):
+                    continue
+                if rel_path.startswith("src/"):
+                    continue
+                if rel_path in files_to_ignore:
+                    continue
+
+                dest_path = os.path.join(repo_path, rel_path)
+
+                print(f"Updating {rel_path}…")
+                if not dry_run:
+                    os.makedirs(os.path.dirname(dest_path), exist_ok=True)
+                    shutil.copy2(src_path, dest_path)
+
+
+def main() -> None:
+    parser = argparse.ArgumentParser(description="Handle the Rerun template.")
+    subparsers = parser.add_subparsers(dest="command")
+
+    init_parser = subparsers.add_parser("init", help="Initialize a new checkout of the template.")
+    init_parser.add_argument(
+        "--languages", default="", nargs="?", const="", help="The languages to support (e.g. `cpp,python,rust`)."
+    )
+    init_parser.add_argument("--dry-run", action="store_true", help="Don't actually delete any files.")
+
+    update_parser = subparsers.add_parser(
+        "update", help="Update all existing Rerun repositories with the latest changes from the template"
+    )
+    update_parser.add_argument(
+        "--languages", default="", nargs="?", const="", help="The languages to support (e.g. `cpp,python,rust`)."
+    )
+    update_parser.add_argument("--dry-run", action="store_true", help="Don't actually delete any files.")
+
+    args = parser.parse_args()
+
+    if args.command == "init":
+        init(parse_languages(args.languages), args.dry_run)
+    elif args.command == "update":
+        update(parse_languages(args.languages), args.dry_run)
+    else:
+        parser.print_help()
+        exit(1)
+
+
+if __name__ == "__main__":
+    main()

style.css

@@ -0,0 +1,4 @@
+gradio-app {
+  max-width: 900px;
+  margin: auto;
+}

vista/.gitignore

@@ -0,0 +1,168 @@
+# Byte-compiled / optimized / DLL files
+__pycache__/
+*.py[cod]
+*$py.class
+
+# C extensions
+*.so
+
+# Distribution / packaging
+.Python
+build/
+develop-eggs/
+dist/
+downloads/
+eggs/
+.eggs/
+lib/
+lib64/
+parts/
+sdist/
+var/
+wheels/
+share/python-wheels/
+*.egg-info/
+.installed.cfg
+*.egg
+MANIFEST
+
+# PyInstaller
+#  Usually these files are written by a python script from a template
+#  before PyInstaller builds the exe, so as to inject date/other infos into it.
+*.manifest
+*.spec
+
+# Installer logs
+pip-log.txt
+pip-delete-this-directory.txt
+
+# Unit test / coverage reports
+htmlcov/
+.tox/
+.nox/
+.coverage
+.coverage.*
+.cache
+nosetests.xml
+coverage.xml
+*.cover
+*.py,cover
+.hypothesis/
+.pytest_cache/
+cover/
+
+# Translations
+*.mo
+*.pot
+
+# Django stuff:
+*.log
+local_settings.py
+db.sqlite3
+db.sqlite3-journal
+
+# Flask stuff:
+instance/
+.webassets-cache
+
+# Scrapy stuff:
+.scrapy
+
+# Sphinx documentation
+docs/_build/
+
+# PyBuilder
+.pybuilder/
+target/
+
+# Jupyter Notebook
+.ipynb_checkpoints
+
+# IPython
+profile_default/
+ipython_config.py
+
+# pyenv
+#   For a library or package, you might want to ignore these files since the code is
+#   intended to run in multiple environments; otherwise, check them in:
+# .python-version
+
+# pipenv
+#   According to pypa/pipenv#598, it is recommended to include Pipfile.lock in version control.
+#   However, in case of collaboration, if having platform-specific dependencies or dependencies
+#   having no cross-platform support, pipenv may install dependencies that don't work, or not
+#   install all needed dependencies.
+#Pipfile.lock
+
+# poetry
+#   Similar to Pipfile.lock, it is generally recommended to include poetry.lock in version control.
+#   This is especially recommended for binary packages to ensure reproducibility, and is more
+#   commonly ignored for libraries.
+#   https://python-poetry.org/docs/basic-usage/#commit-your-poetrylock-file-to-version-control
+#poetry.lock
+
+# pdm
+#   Similar to Pipfile.lock, it is generally recommended to include pdm.lock in version control.
+#pdm.lock
+#   pdm stores project-wide configurations in .pdm.toml, but it is recommended to not include it
+#   in version control.
+#   https://pdm.fming.dev/latest/usage/project/#working-with-version-control
+.pdm.toml
+.pdm-python
+.pdm-build/
+
+# PEP 582; used by e.g. github.com/David-OConnor/pyflow and github.com/pdm-project/pdm
+__pypackages__/
+
+# Celery stuff
+celerybeat-schedule
+celerybeat.pid
+
+# SageMath parsed files
+*.sage.py
+
+# Environments
+.env
+.venv
+env/
+venv/
+ENV/
+env.bak/
+venv.bak/
+
+# Spyder project settings
+.spyderproject
+.spyproject
+
+# Rope project settings
+.ropeproject
+
+# mkdocs documentation
+/site
+
+# mypy
+.mypy_cache/
+.dmypy.json
+dmypy.json
+
+# Pyre type checker
+.pyre/
+
+# pytype static type analyzer
+.pytype/
+
+# Cython debug symbols
+cython_debug/
+
+# PyCharm
+#  JetBrains specific template is maintained in a separate JetBrains.gitignore that can
+#  be found at https://github.com/github/gitignore/blob/main/Global/JetBrains.gitignore
+#  and can be added to the global gitignore or merged into this file.  For a more nuclear
+#  option (not recommended) you can uncomment the following to ignore the entire idea folder.
+.idea/
+
+outputs/
+logs/
+
+.DS_Store
+*/.DS_Store

vista/LICENSE

@@ -0,0 +1,201 @@
+                                 Apache License
+                           Version 2.0, January 2004
+                        http://www.apache.org/licenses/
+
+   TERMS AND CONDITIONS FOR USE, REPRODUCTION, AND DISTRIBUTION
+
+   1. Definitions.
+
+      "License" shall mean the terms and conditions for use, reproduction,
+      and distribution as defined by Sections 1 through 9 of this document.
+
+      "Licensor" shall mean the copyright owner or entity authorized by
+      the copyright owner that is granting the License.
+
+      "Legal Entity" shall mean the union of the acting entity and all
+      other entities that control, are controlled by, or are under common
+      control with that entity. For the purposes of this definition,
+      "control" means (i) the power, direct or indirect, to cause the
+      direction or management of such entity, whether by contract or
+      otherwise, or (ii) ownership of fifty percent (50%) or more of the
+      outstanding shares, or (iii) beneficial ownership of such entity.
+
+      "You" (or "Your") shall mean an individual or Legal Entity
+      exercising permissions granted by this License.
+
+      "Source" form shall mean the preferred form for making modifications,
+      including but not limited to software source code, documentation
+      source, and configuration files.
+
+      "Object" form shall mean any form resulting from mechanical
+      transformation or translation of a Source form, including but
+      not limited to compiled object code, generated documentation,
+      and conversions to other media types.
+
+      "Work" shall mean the work of authorship, whether in Source or
+      Object form, made available under the License, as indicated by a
+      copyright notice that is included in or attached to the work
+      (an example is provided in the Appendix below).
+
+      "Derivative Works" shall mean any work, whether in Source or Object
+      form, that is based on (or derived from) the Work and for which the
+      editorial revisions, annotations, elaborations, or other modifications
+      represent, as a whole, an original work of authorship. For the purposes
+      of this License, Derivative Works shall not include works that remain
+      separable from, or merely link (or bind by name) to the interfaces of,
+      the Work and Derivative Works thereof.
+
+      "Contribution" shall mean any work of authorship, including
+      the original version of the Work and any modifications or additions
+      to that Work or Derivative Works thereof, that is intentionally
+      submitted to Licensor for inclusion in the Work by the copyright owner
+      or by an individual or Legal Entity authorized to submit on behalf of
+      the copyright owner. For the purposes of this definition, "submitted"
+      means any form of electronic, verbal, or written communication sent
+      to the Licensor or its representatives, including but not limited to
+      communication on electronic mailing lists, source code control systems,
+      and issue tracking systems that are managed by, or on behalf of, the
+      Licensor for the purpose of discussing and improving the Work, but
+      excluding communication that is conspicuously marked or otherwise
+      designated in writing by the copyright owner as "Not a Contribution."
+
+      "Contributor" shall mean Licensor and any individual or Legal Entity
+      on behalf of whom a Contribution has been received by Licensor and
+      subsequently incorporated within the Work.
+
+   2. Grant of Copyright License. Subject to the terms and conditions of
+      this License, each Contributor hereby grants to You a perpetual,
+      worldwide, non-exclusive, no-charge, royalty-free, irrevocable
+      copyright license to reproduce, prepare Derivative Works of,
+      publicly display, publicly perform, sublicense, and distribute the
+      Work and such Derivative Works in Source or Object form.
+
+   3. Grant of Patent License. Subject to the terms and conditions of
+      this License, each Contributor hereby grants to You a perpetual,
+      worldwide, non-exclusive, no-charge, royalty-free, irrevocable
+      (except as stated in this section) patent license to make, have made,
+      use, offer to sell, sell, import, and otherwise transfer the Work,
+      where such license applies only to those patent claims licensable
+      by such Contributor that are necessarily infringed by their
+      Contribution(s) alone or by combination of their Contribution(s)
+      with the Work to which such Contribution(s) was submitted. If You
+      institute patent litigation against any entity (including a
+      cross-claim or counterclaim in a lawsuit) alleging that the Work
+      or a Contribution incorporated within the Work constitutes direct
+      or contributory patent infringement, then any patent licenses
+      granted to You under this License for that Work shall terminate
+      as of the date such litigation is filed.
+
+   4. Redistribution. You may reproduce and distribute copies of the
+      Work or Derivative Works thereof in any medium, with or without
+      modifications, and in Source or Object form, provided that You
+      meet the following conditions:
+
+      (a) You must give any other recipients of the Work or
+          Derivative Works a copy of this License; and
+
+      (b) You must cause any modified files to carry prominent notices
+          stating that You changed the files; and
+
+      (c) You must retain, in the Source form of any Derivative Works
+          that You distribute, all copyright, patent, trademark, and
+          attribution notices from the Source form of the Work,
+          excluding those notices that do not pertain to any part of
+          the Derivative Works; and
+
+      (d) If the Work includes a "NOTICE" text file as part of its
+          distribution, then any Derivative Works that You distribute must
+          include a readable copy of the attribution notices contained
+          within such NOTICE file, excluding those notices that do not
+          pertain to any part of the Derivative Works, in at least one
+          of the following places: within a NOTICE text file distributed
+          as part of the Derivative Works; within the Source form or
+          documentation, if provided along with the Derivative Works; or,
+          within a display generated by the Derivative Works, if and
+          wherever such third-party notices normally appear. The contents
+          of the NOTICE file are for informational purposes only and
+          do not modify the License. You may add Your own attribution
+          notices within Derivative Works that You distribute, alongside
+          or as an addendum to the NOTICE text from the Work, provided
+          that such additional attribution notices cannot be construed
+          as modifying the License.
+
+      You may add Your own copyright statement to Your modifications and
+      may provide additional or different license terms and conditions
+      for use, reproduction, or distribution of Your modifications, or
+      for any such Derivative Works as a whole, provided Your use,
+      reproduction, and distribution of the Work otherwise complies with
+      the conditions stated in this License.
+
+   5. Submission of Contributions. Unless You explicitly state otherwise,
+      any Contribution intentionally submitted for inclusion in the Work
+      by You to the Licensor shall be under the terms and conditions of
+      this License, without any additional terms or conditions.
+      Notwithstanding the above, nothing herein shall supersede or modify
+      the terms of any separate license agreement you may have executed
+      with Licensor regarding such Contributions.
+
+   6. Trademarks. This License does not grant permission to use the trade
+      names, trademarks, service marks, or product names of the Licensor,
+      except as required for reasonable and customary use in describing the
+      origin of the Work and reproducing the content of the NOTICE file.
+
+   7. Disclaimer of Warranty. Unless required by applicable law or
+      agreed to in writing, Licensor provides the Work (and each
+      Contributor provides its Contributions) on an "AS IS" BASIS,
+      WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or
+      implied, including, without limitation, any warranties or conditions
+      of TITLE, NON-INFRINGEMENT, MERCHANTABILITY, or FITNESS FOR A
+      PARTICULAR PURPOSE. You are solely responsible for determining the
+      appropriateness of using or redistributing the Work and assume any
+      risks associated with Your exercise of permissions under this License.
+
+   8. Limitation of Liability. In no event and under no legal theory,
+      whether in tort (including negligence), contract, or otherwise,
+      unless required by applicable law (such as deliberate and grossly
+      negligent acts) or agreed to in writing, shall any Contributor be
+      liable to You for damages, including any direct, indirect, special,
+      incidental, or consequential damages of any character arising as a
+      result of this License or out of the use or inability to use the
+      Work (including but not limited to damages for loss of goodwill,
+      work stoppage, computer failure or malfunction, or any and all
+      other commercial damages or losses), even if such Contributor
+      has been advised of the possibility of such damages.
+
+   9. Accepting Warranty or Additional Liability. While redistributing
+      the Work or Derivative Works thereof, You may choose to offer,
+      and charge a fee for, acceptance of support, warranty, indemnity,
+      or other liability obligations and/or rights consistent with this
+      License. However, in accepting such obligations, You may act only
+      on Your own behalf and on Your sole responsibility, not on behalf
+      of any other Contributor, and only if You agree to indemnify,
+      defend, and hold each Contributor harmless for any liability
+      incurred by, or claims asserted against, such Contributor by reason
+      of your accepting any such warranty or additional liability.
+
+   END OF TERMS AND CONDITIONS
+
+   APPENDIX: How to apply the Apache License to your work.
+
+      To apply the Apache License to your work, attach the following
+      boilerplate notice, with the fields enclosed by brackets "[]"
+      replaced with your own identifying information. (Don't include
+      the brackets!)  The text should be enclosed in the appropriate
+      comment syntax for the file format. We also recommend that a
+      file or class name and description of purpose be included on the
+      same "printed page" as the copyright notice for easier
+      identification within third-party archives.
+
+   Copyright [yyyy] [name of copyright owner]
+
+   Licensed under the Apache License, Version 2.0 (the "License");
+   you may not use this file except in compliance with the License.
+   You may obtain a copy of the License at
+
+       http://www.apache.org/licenses/LICENSE-2.0
+
+   Unless required by applicable law or agreed to in writing, software
+   distributed under the License is distributed on an "AS IS" BASIS,
+   WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+   See the License for the specific language governing permissions and
+   limitations under the License.

vista/__init__.py

@@ -0,0 +1,96 @@
+from __future__ import annotations
+
+import rerun.blueprint as rrb
+import torch
+from transformers.utils import hub
+
+from . import sample, sample_utils
+
+
+def create_model():
+    return sample_utils.init_model(
+        {
+            "config": "./vista/configs/inference/vista.yaml",
+            "ckpt": hub.get_file_from_repo("OpenDriveLab/Vista", "vista.safetensors"),
+        }
+    )
+
+
+def generate_blueprint(n_rounds: int) -> rrb.Blueprint:
+    row1 = rrb.Horizontal(
+        *[
+            rrb.TensorView(origin=f"diffusion_{i}", name=f"Latents Segment {i+1}")
+            for i in range(n_rounds)
+        ],
+    )
+    row2 = rrb.Spatial2DView(origin="generated_image", name="Generated Video")
+
+    return rrb.Blueprint(rrb.Vertical(row1, row2), collapse_panels=True)
+
+
+def run_sampling(
+    log_queue,
+    first_frame_file_name,
+    height,
+    width,
+    n_rounds,
+    n_frames,
+    n_steps,
+    cfg_scale,
+    cond_aug,
+    model=None,
+) -> None:
+    if model is None:
+        model = create_model()
+
+    unique_keys = set([x.input_key for x in model.conditioner.embedders])
+    value_dict = sample_utils.init_embedder_options(unique_keys)
+
+    action_dict = None
+
+    first_frame = sample.load_img(first_frame_file_name, height, width, "cuda")[None]
+    repeated_frame = first_frame.expand(n_frames, -1, -1, -1)
+
+    value_dict = sample_utils.init_embedder_options(unique_keys)
+    cond_img = first_frame
+    value_dict["cond_frames_without_noise"] = cond_img
+    value_dict["cond_aug"] = cond_aug
+    value_dict["cond_frames"] = cond_img + cond_aug * torch.randn_like(cond_img)
+    if action_dict is not None:
+        for key, value in action_dict.items():
+            value_dict[key] = value
+
+    if n_rounds > 1:
+        guider = "TrianglePredictionGuider"
+    else:
+        guider = "VanillaCFG"
+    sampler = sample_utils.init_sampling(
+        guider=guider,
+        steps=n_steps,
+        cfg_scale=cfg_scale,
+        num_frames=n_frames,
+    )
+
+    uc_keys = [
+        "cond_frames",
+        "cond_frames_without_noise",
+        "command",
+        "trajectory",
+        "speed",
+        "angle",
+        "goal",
+    ]
+
+    _generated_images, _samples_z, _inputs = sample_utils.do_sample(
+        repeated_frame,
+        model,
+        sampler,
+        value_dict,
+        num_rounds=n_rounds,
+        num_frames=n_frames,
+        force_uc_zero_embeddings=uc_keys,
+        initial_cond_indices=[0],
+        log_queue=log_queue,
+    )
+
+    log_queue.put("done")

vista/bin_to_st.py

@@ -0,0 +1,56 @@
+from __future__ import annotations
+
+import os
+
+import torch
+from safetensors.torch import save_file
+
+ckpt = "path_to/pytorch_model.bin"
+
+vista_bin = torch.load(ckpt, map_location="cpu")  # only contains model weights
+
+for k in list(vista_bin.keys()):  # merge LoRA weights (if exist) for inference
+    if "adapter_down" in k:
+        if "q_adapter_down" in k:
+            up_k = k.replace("q_adapter_down", "q_adapter_up")
+            pretrain_k = k.replace("q_adapter_down", "to_q")
+        elif "k_adapter_down" in k:
+            up_k = k.replace("k_adapter_down", "k_adapter_up")
+            pretrain_k = k.replace("k_adapter_down", "to_k")
+        elif "v_adapter_down" in k:
+            up_k = k.replace("v_adapter_down", "v_adapter_up")
+            pretrain_k = k.replace("v_adapter_down", "to_v")
+        else:
+            up_k = k.replace("out_adapter_down", "out_adapter_up")
+            if "model_ema" in k:
+                pretrain_k = k.replace("out_adapter_down", "to_out0")
+            else:
+                pretrain_k = k.replace("out_adapter_down", "to_out.0")
+
+        lora_weights = vista_bin[up_k] @ vista_bin[k]
+        del vista_bin[k]
+        del vista_bin[up_k]
+        vista_bin[pretrain_k] = vista_bin[pretrain_k] + lora_weights
+
+for k in list(vista_bin.keys()):  # remove the prefix
+    if "_forward_module" in k and "decay" not in k and "num_updates" not in k:
+        vista_bin[k.replace("_forward_module.", "")] = vista_bin[k]
+    del vista_bin[k]
+
+for k in list(vista_bin.keys()):  # combine EMA weights
+    if "model_ema" in k:
+        orig_k = None
+        for kk in list(vista_bin.keys()):
+            if "model_ema" not in kk and k[10:] == kk[6:].replace(".", ""):
+                orig_k = kk
+        assert orig_k is not None
+        vista_bin[orig_k] = vista_bin[k]
+        del vista_bin[k]
+        print("Replace", orig_k, "with", k)
+
+vista_st = dict()
+for k in list(vista_bin.keys()):
+    vista_st[k] = vista_bin[k]
+
+os.makedirs("ckpts", exist_ok=True)
+save_file(vista_st, "ckpts/vista.safetensors")

vista/configs/example/nusc_train.yaml

@@ -0,0 +1,292 @@
+model:
+  base_learning_rate: 5.e-5
+  target: vista.vwm.models.diffusion.DiffusionEngine
+  params:
+    use_ema: True
+    input_key: img_seq
+    scale_factor: 0.18215
+    disable_first_stage_autocast: True
+    en_and_decode_n_samples_a_time: 1
+    num_frames: &num_frames 25
+    slow_spatial_layers: True
+    train_peft_adapters: False
+    replace_cond_frames: &replace_cond_frames True
+    fixed_cond_frames: # only used for logging images
+      - [ 0, 1, 2 ]
+
+    denoiser_config:
+      target: vista.vwm.modules.diffusionmodules.denoiser.Denoiser
+      params:
+        num_frames: *num_frames
+
+        scaling_config:
+          target: vista.vwm.modules.diffusionmodules.denoiser_scaling.VScalingWithEDMcNoise
+
+    network_config:
+      target: vista.vwm.modules.diffusionmodules.video_model.VideoUNet
+      params:
+        adm_in_channels: 768
+        num_classes: sequential
+        use_checkpoint: True
+        in_channels: 8
+        out_channels: 4
+        model_channels: 320
+        attention_resolutions: [ 4, 2, 1 ]
+        num_res_blocks: 2
+        channel_mult: [ 1, 2, 4, 4 ]
+        num_head_channels: 64
+        use_linear_in_transformer: True
+        transformer_depth: 1
+        context_dim: 1024
+        spatial_transformer_attn_type: softmax-xformers
+        extra_ff_mix_layer: True
+        use_spatial_context: True
+        merge_strategy: learned_with_images
+        video_kernel_size: [ 3, 1, 1 ]
+        add_lora: False
+        action_control: False
+
+    conditioner_config:
+      target: vista.vwm.modules.GeneralConditioner
+      params:
+        emb_models:
+          - input_key: cond_frames_without_noise
+            is_trainable: False
+            ucg_rate: 0.15
+            target: vista.vwm.modules.encoders.modules.FrozenOpenCLIPImagePredictionEmbedder
+            params:
+              n_cond_frames: 1
+              n_copies: 1
+              open_clip_embedding_config:
+                target: vista.vwm.modules.encoders.modules.FrozenOpenCLIPImageEmbedder
+                params:
+                  freeze: True
+
+          - input_key: fps_id
+            is_trainable: False
+            ucg_rate: 0.0
+            target: vista.vwm.modules.encoders.modules.ConcatTimestepEmbedderND
+            params:
+              outdim: 256
+
+          - input_key: motion_bucket_id
+            is_trainable: False
+            ucg_rate: 0.0
+            target: vista.vwm.modules.encoders.modules.ConcatTimestepEmbedderND
+            params:
+              outdim: 256
+
+          - input_key: cond_frames
+            is_trainable: False
+            ucg_rate: 0.15
+            target: vista.vwm.modules.encoders.modules.VideoPredictionEmbedderWithEncoder
+            params:
+              disable_encoder_autocast: True
+              n_cond_frames: 1
+              n_copies: 1
+              is_ae: True
+
+              encoder_config:
+                target: vista.vwm.models.autoencoder.AutoencoderKLModeOnly
+                params:
+                  embed_dim: 4
+                  monitor: val/rec_loss
+
+                  ddconfig:
+                    attn_type: vanilla-xformers
+                    double_z: True
+                    z_channels: 4
+                    resolution: 256
+                    in_channels: 3
+                    out_ch: 3
+                    ch: 128
+                    ch_mult: [ 1, 2, 4, 4 ]
+                    num_res_blocks: 2
+                    attn_resolutions: [ ]
+                    dropout: 0.0
+
+                  loss_config:
+                    target: torch.nn.Identity
+
+          - input_key: cond_aug
+            is_trainable: False
+            ucg_rate: 0.0
+            target: vista.vwm.modules.encoders.modules.ConcatTimestepEmbedderND
+            params:
+              outdim: 256
+
+          - input_key: command
+            is_trainable: False
+            ucg_rate: 0.15
+            target: vista.vwm.modules.encoders.modules.ConcatTimestepEmbedderND
+            params:
+              outdim: &action_emb_dim 128
+              num_features: 1
+              add_sequence_dim: True
+
+          - input_key: trajectory
+            is_trainable: False
+            ucg_rate: 0.15
+            target: vista.vwm.modules.encoders.modules.ConcatTimestepEmbedderND
+            params:
+              outdim: *action_emb_dim
+              num_features: 8
+              add_sequence_dim: True
+
+          - input_key: speed
+            is_trainable: False
+            ucg_rate: 0.15
+            target: vista.vwm.modules.encoders.modules.ConcatTimestepEmbedderND
+            params:
+              outdim: *action_emb_dim
+              num_features: 4
+              add_sequence_dim: True
+
+          - input_key: angle
+            is_trainable: False
+            ucg_rate: 0.15
+            target: vista.vwm.modules.encoders.modules.ConcatTimestepEmbedderND
+            params:
+              outdim: *action_emb_dim
+              num_features: 4
+              add_sequence_dim: True
+
+          - input_key: goal
+            is_trainable: False
+            ucg_rate: 0.15
+            target: vista.vwm.modules.encoders.modules.ConcatTimestepEmbedderND
+            params:
+              outdim: *action_emb_dim
+              num_features: 2
+              add_sequence_dim: True
+
+    first_stage_config:
+      target: vista.vwm.models.autoencoder.AutoencodingEngine
+      params:
+        loss_config:
+          target: torch.nn.Identity
+
+        regularizer_config:
+          target: vista.vwm.modules.autoencoding.regularizers.DiagonalGaussianRegularizer
+
+        encoder_config:
+          target: vista.vwm.modules.diffusionmodules.model.Encoder
+          params:
+            attn_type: vanilla
+            double_z: True
+            z_channels: 4
+            resolution: 256
+            in_channels: 3
+            out_ch: 3
+            ch: 128
+            ch_mult: [ 1, 2, 4, 4 ]
+            num_res_blocks: 2
+            attn_resolutions: [ ]
+            dropout: 0.0
+
+        decoder_config:
+          target: vista.vwm.modules.autoencoding.temporal_ae.VideoDecoder
+          params:
+            attn_type: vanilla
+            double_z: True
+            z_channels: 4
+            resolution: 256
+            in_channels: 3
+            out_ch: 3
+            ch: 128
+            ch_mult: [ 1, 2, 4, 4 ]
+            num_res_blocks: 2
+            attn_resolutions: [ ]
+            dropout: 0.0
+            video_kernel_size: [ 3, 1, 1 ]
+
+    scheduler_config:
+      target: vista.vwm.lr_scheduler.LambdaLinearScheduler
+      params:
+        warm_up_steps: [ 1000 ]
+        cycle_lengths: [ 10000000000000 ]
+        f_start: [ 1.e-6 ]
+        f_max: [ 1. ]
+        f_min: [ 1. ]
+
+    loss_fn_config:
+      target: vista.vwm.modules.diffusionmodules.loss.StandardDiffusionLoss
+      params:
+        use_additional_loss: True
+        offset_noise_level: 0.02
+        additional_loss_weight: 0.1
+        num_frames: *num_frames
+        replace_cond_frames: *replace_cond_frames
+        cond_frames_choices:
+          - [ ]
+          - [ 0 ]
+          - [ 0, 1 ]
+          - [ 0, 1, 2 ]
+
+        sigma_sampler_config:
+          target: vista.vwm.modules.diffusionmodules.sigma_sampling.EDMSampling
+          params:
+            p_mean: 1.0
+            p_std: 1.6
+            num_frames: *num_frames
+
+        loss_weighting_config:
+          target: vista.vwm.modules.diffusionmodules.loss_weighting.VWeighting
+
+    sampler_config:
+      target: vista.vwm.modules.diffusionmodules.sampling.EulerEDMSampler
+      params:
+        num_steps: 15
+
+        discretization_config:
+          target: vista.vwm.modules.diffusionmodules.discretizer.EDMDiscretization
+          params:
+            sigma_max: 700.0
+
+        guider_config:
+          target: vista.vwm.modules.diffusionmodules.guiders.LinearPredictionGuider
+          params:
+            num_frames: *num_frames
+            max_scale: 3.0
+            min_scale: 1.5
+
+data:
+  target: vista.vwm.data.dataset.Sampler
+  params:
+    batch_size: 1
+    num_workers: 16
+    subsets:
+      - NuScenes
+    probs:
+      - 1
+    samples_per_epoch: 16000
+    target_height: 576
+    target_width: 1024
+    num_frames: *num_frames
+
+lightning:
+  callbacks:
+    image_logger:
+      target: train.ImageLogger
+      params:
+        num_frames: *num_frames
+        disabled: False
+        enable_autocast: False
+        batch_frequency: 100
+        increase_log_steps: True
+        log_first_step: False
+        log_images_kwargs:
+          N: *num_frames
+
+  modelcheckpoint:
+    params:
+      every_n_epochs: 1  # every_n_train_steps: 5000, set the same as image_logger batch_frequency
+
+  trainer:
+    devices: 0,1
+    benchmark: True
+    num_sanity_val_steps: 0
+    accumulate_grad_batches: 1
+    max_epochs: 100
+    strategy: deepspeed_stage_2
+    gradient_clip_val: 0.3

vista/configs/inference/vista.yaml

@@ -0,0 +1,184 @@
+model:
+  target: vista.vwm.models.diffusion.DiffusionEngine
+  params:
+    input_key: img_seq
+    scale_factor: 0.18215
+    disable_first_stage_autocast: True
+    en_and_decode_n_samples_a_time: 1
+    num_frames: &num_frames 25
+
+    denoiser_config:
+      target: vista.vwm.modules.diffusionmodules.denoiser.Denoiser
+      params:
+        num_frames: *num_frames
+
+        scaling_config:
+          target: vista.vwm.modules.diffusionmodules.denoiser_scaling.VScalingWithEDMcNoise
+
+    network_config:
+      target: vista.vwm.modules.diffusionmodules.video_model.VideoUNet
+      params:
+        adm_in_channels: 768
+        num_classes: sequential
+        use_checkpoint: False
+        in_channels: 8
+        out_channels: 4
+        model_channels: 320
+        attention_resolutions: [ 4, 2, 1 ]
+        num_res_blocks: 2
+        channel_mult: [ 1, 2, 4, 4 ]
+        num_head_channels: 64
+        use_linear_in_transformer: True
+        transformer_depth: 1
+        context_dim: 1024
+        spatial_transformer_attn_type: softmax-xformers
+        extra_ff_mix_layer: True
+        use_spatial_context: True
+        merge_strategy: learned_with_images
+        video_kernel_size: [ 3, 1, 1 ]
+        add_lora: False
+        action_control: True
+
+    conditioner_config:
+      target: vista.vwm.modules.GeneralConditioner
+      params:
+        emb_models:
+          - input_key: cond_frames_without_noise
+            is_trainable: False
+            target: vista.vwm.modules.encoders.modules.FrozenOpenCLIPImagePredictionEmbedder
+            params:
+              n_cond_frames: 1
+              n_copies: 1
+              open_clip_embedding_config:
+                target: vista.vwm.modules.encoders.modules.FrozenOpenCLIPImageEmbedder
+                params:
+                  freeze: True
+
+          - input_key: fps_id
+            is_trainable: False
+            target: vista.vwm.modules.encoders.modules.ConcatTimestepEmbedderND
+            params:
+              outdim: 256
+
+          - input_key: motion_bucket_id
+            is_trainable: False
+            target: vista.vwm.modules.encoders.modules.ConcatTimestepEmbedderND
+            params:
+              outdim: 256
+
+          - input_key: cond_frames
+            is_trainable: False
+            target: vista.vwm.modules.encoders.modules.VideoPredictionEmbedderWithEncoder
+            params:
+              disable_encoder_autocast: True
+              n_cond_frames: 1
+              n_copies: 1
+              is_ae: True
+
+              encoder_config:
+                target: vista.vwm.models.autoencoder.AutoencoderKLModeOnly
+                params:
+                  embed_dim: 4
+                  monitor: val/rec_loss
+
+                  ddconfig:
+                    attn_type: vanilla-xformers
+                    double_z: True
+                    z_channels: 4
+                    resolution: 256
+                    in_channels: 3
+                    out_ch: 3
+                    ch: 128
+                    ch_mult: [ 1, 2, 4, 4 ]
+                    num_res_blocks: 2
+                    attn_resolutions: [ ]
+                    dropout: 0.0
+
+                  loss_config:
+                    target: torch.nn.Identity
+
+          - input_key: cond_aug
+            is_trainable: False
+            target: vista.vwm.modules.encoders.modules.ConcatTimestepEmbedderND
+            params:
+              outdim: 256
+
+          - input_key: command
+            is_trainable: False
+            target: vista.vwm.modules.encoders.modules.ConcatTimestepEmbedderND
+            params:
+              outdim: &action_emb_dim 128
+              num_features: 1
+              add_sequence_dim: True
+
+          - input_key: trajectory
+            is_trainable: False
+            target: vista.vwm.modules.encoders.modules.ConcatTimestepEmbedderND
+            params:
+              outdim: *action_emb_dim
+              num_features: 8
+              add_sequence_dim: True
+
+          - input_key: speed
+            is_trainable: False
+            target: vista.vwm.modules.encoders.modules.ConcatTimestepEmbedderND
+            params:
+              outdim: *action_emb_dim
+              num_features: 4
+              add_sequence_dim: True
+
+          - input_key: angle
+            is_trainable: False
+            target: vista.vwm.modules.encoders.modules.ConcatTimestepEmbedderND
+            params:
+              outdim: *action_emb_dim
+              num_features: 4
+              add_sequence_dim: True
+
+          - input_key: goal
+            is_trainable: False
+            target: vista.vwm.modules.encoders.modules.ConcatTimestepEmbedderND
+            params:
+              outdim: *action_emb_dim
+              num_features: 2
+              add_sequence_dim: True
+
+    first_stage_config:
+      target: vista.vwm.models.autoencoder.AutoencodingEngine
+      params:
+        loss_config:
+          target: torch.nn.Identity
+
+        regularizer_config:
+          target: vista.vwm.modules.autoencoding.regularizers.DiagonalGaussianRegularizer
+
+        encoder_config:
+          target: vista.vwm.modules.diffusionmodules.model.Encoder
+          params:
+            attn_type: vanilla
+            double_z: True
+            z_channels: 4
+            resolution: 256
+            in_channels: 3
+            out_ch: 3
+            ch: 128
+            ch_mult: [ 1, 2, 4, 4 ]
+            num_res_blocks: 2
+            attn_resolutions: [ ]
+            dropout: 0.0
+
+        decoder_config:
+          target: vista.vwm.modules.autoencoding.temporal_ae.VideoDecoder
+          params:
+            attn_type: vanilla
+            double_z: True
+            z_channels: 4
+            resolution: 256
+            in_channels: 3
+            out_ch: 3
+            ch: 128
+            ch_mult: [ 1, 2, 4, 4 ]
+            num_res_blocks: 2
+            attn_resolutions: [ ]
+            dropout: 0.0
+            video_kernel_size: [ 3, 1, 1 ]

vista/configs/training/vista_phase1.yaml

@@ -0,0 +1,247 @@
+model:
+  base_learning_rate: 1.e-5
+  target: vista.vwm.models.diffusion.DiffusionEngine
+  params:
+    use_ema: True
+    input_key: img_seq
+    scale_factor: 0.18215
+    disable_first_stage_autocast: True
+    en_and_decode_n_samples_a_time: 1
+    num_frames: &num_frames 25
+    slow_spatial_layers: True
+    train_peft_adapters: False
+    replace_cond_frames: &replace_cond_frames True
+    fixed_cond_frames: # only used for logging images
+      - [ 0, 1, 2 ]
+
+    denoiser_config:
+      target: vista.vwm.modules.diffusionmodules.denoiser.Denoiser
+      params:
+        num_frames: *num_frames
+
+        scaling_config:
+          target: vista.vwm.modules.diffusionmodules.denoiser_scaling.VScalingWithEDMcNoise
+
+    network_config:
+      target: vista.vwm.modules.diffusionmodules.video_model.VideoUNet
+      params:
+        adm_in_channels: 768
+        num_classes: sequential
+        use_checkpoint: True
+        in_channels: 8
+        out_channels: 4
+        model_channels: 320
+        attention_resolutions: [ 4, 2, 1 ]
+        num_res_blocks: 2
+        channel_mult: [ 1, 2, 4, 4 ]
+        num_head_channels: 64
+        use_linear_in_transformer: True
+        transformer_depth: 1
+        context_dim: 1024
+        spatial_transformer_attn_type: softmax-xformers
+        extra_ff_mix_layer: True
+        use_spatial_context: True
+        merge_strategy: learned_with_images
+        video_kernel_size: [ 3, 1, 1 ]
+        add_lora: False
+        action_control: False
+
+    conditioner_config:
+      target: vista.vwm.modules.GeneralConditioner
+      params:
+        emb_models:
+          - input_key: cond_frames_without_noise
+            is_trainable: False
+            ucg_rate: 0.15
+            target: vista.vwm.modules.encoders.modules.FrozenOpenCLIPImagePredictionEmbedder
+            params:
+              n_cond_frames: 1
+              n_copies: 1
+              open_clip_embedding_config:
+                target: vista.vwm.modules.encoders.modules.FrozenOpenCLIPImageEmbedder
+                params:
+                  freeze: True
+
+          - input_key: fps_id
+            is_trainable: False
+            ucg_rate: 0.0
+            target: vista.vwm.modules.encoders.modules.ConcatTimestepEmbedderND
+            params:
+              outdim: 256
+
+          - input_key: motion_bucket_id
+            is_trainable: False
+            ucg_rate: 0.0
+            target: vista.vwm.modules.encoders.modules.ConcatTimestepEmbedderND
+            params:
+              outdim: 256
+
+          - input_key: cond_frames
+            is_trainable: False
+            ucg_rate: 0.15
+            target: vista.vwm.modules.encoders.modules.VideoPredictionEmbedderWithEncoder
+            params:
+              disable_encoder_autocast: True
+              n_cond_frames: 1
+              n_copies: 1
+              is_ae: True
+
+              encoder_config:
+                target: vista.vwm.models.autoencoder.AutoencoderKLModeOnly
+                params:
+                  embed_dim: 4
+                  monitor: val/rec_loss
+
+                  ddconfig:
+                    attn_type: vanilla-xformers
+                    double_z: True
+                    z_channels: 4
+                    resolution: 256
+                    in_channels: 3
+                    out_ch: 3
+                    ch: 128
+                    ch_mult: [ 1, 2, 4, 4 ]
+                    num_res_blocks: 2
+                    attn_resolutions: [ ]
+                    dropout: 0.0
+
+                  loss_config:
+                    target: torch.nn.Identity
+
+          - input_key: cond_aug
+            is_trainable: False
+            ucg_rate: 0.0
+            target: vista.vwm.modules.encoders.modules.ConcatTimestepEmbedderND
+            params:
+              outdim: 256
+
+    first_stage_config:
+      target: vista.vwm.models.autoencoder.AutoencodingEngine
+      params:
+        loss_config:
+          target: torch.nn.Identity
+
+        regularizer_config:
+          target: vista.vwm.modules.autoencoding.regularizers.DiagonalGaussianRegularizer
+
+        encoder_config:
+          target: vista.vwm.modules.diffusionmodules.model.Encoder
+          params:
+            attn_type: vanilla
+            double_z: True
+            z_channels: 4
+            resolution: 256
+            in_channels: 3
+            out_ch: 3
+            ch: 128
+            ch_mult: [ 1, 2, 4, 4 ]
+            num_res_blocks: 2
+            attn_resolutions: [ ]
+            dropout: 0.0
+
+        decoder_config:
+          target: vista.vwm.modules.autoencoding.temporal_ae.VideoDecoder
+          params:
+            attn_type: vanilla
+            double_z: True
+            z_channels: 4
+            resolution: 256
+            in_channels: 3
+            out_ch: 3
+            ch: 128
+            ch_mult: [ 1, 2, 4, 4 ]
+            num_res_blocks: 2
+            attn_resolutions: [ ]
+            dropout: 0.0
+            video_kernel_size: [ 3, 1, 1 ]
+
+    scheduler_config:
+      target: vista.vwm.lr_scheduler.LambdaLinearScheduler
+      params:
+        warm_up_steps: [ 1000 ]
+        cycle_lengths: [ 10000000000000 ]
+        f_start: [ 1.e-6 ]
+        f_max: [ 1. ]
+        f_min: [ 1. ]
+
+    loss_fn_config:
+      target: vista.vwm.modules.diffusionmodules.loss.StandardDiffusionLoss
+      params:
+        use_additional_loss: True
+        offset_noise_level: 0.02
+        additional_loss_weight: 0.1
+        num_frames: *num_frames
+        replace_cond_frames: *replace_cond_frames
+        cond_frames_choices:
+          - [ ]
+          - [ 0 ]
+          - [ 0, 1 ]
+          - [ 0, 1, 2 ]
+
+        sigma_sampler_config:
+          target: vista.vwm.modules.diffusionmodules.sigma_sampling.EDMSampling
+          params:
+            p_mean: 1.0
+            p_std: 1.6
+            num_frames: *num_frames
+
+        loss_weighting_config:
+          target: vista.vwm.modules.diffusionmodules.loss_weighting.VWeighting
+
+    sampler_config:
+      target: vista.vwm.modules.diffusionmodules.sampling.EulerEDMSampler
+      params:
+        num_steps: 15
+
+        discretization_config:
+          target: vista.vwm.modules.diffusionmodules.discretizer.EDMDiscretization
+          params:
+            sigma_max: 700.0
+
+        guider_config:
+          target: vista.vwm.modules.diffusionmodules.guiders.LinearPredictionGuider
+          params:
+            num_frames: *num_frames
+            max_scale: 3.0
+            min_scale: 1.5
+
+data:
+  target: vista.vwm.data.dataset.Sampler
+  params:
+    batch_size: 1
+    num_workers: 16
+    subsets:
+      - YouTube
+    probs:
+      - 1
+    samples_per_epoch: 256000
+    target_height: 576
+    target_width: 1024
+    num_frames: *num_frames
+
+lightning:
+  callbacks:
+    image_logger:
+      target: train.ImageLogger
+      params:
+        num_frames: *num_frames
+        disabled: False
+        enable_autocast: False
+        batch_frequency: 1000
+        increase_log_steps: True
+        log_first_step: False
+        log_images_kwargs:
+          N: *num_frames
+
+  modelcheckpoint:
+    params:
+      every_n_epochs: 1  # every_n_train_steps: 5000, set the same as image_logger batch_frequency
+
+  trainer:
+    devices: 0,1
+    benchmark: True
+    num_sanity_val_steps: 0
+    accumulate_grad_batches: 2
+    max_epochs: 100
+    strategy: deepspeed_stage_2
+    gradient_clip_val: 0.3

vista/configs/training/vista_phase2_stage1.yaml

@@ -0,0 +1,294 @@
+model:
+  base_learning_rate: 5.e-5
+  target: vista.vwm.models.diffusion.DiffusionEngine
+  params:
+    use_ema: True
+    input_key: img_seq
+    scale_factor: 0.18215
+    disable_first_stage_autocast: True
+    en_and_decode_n_samples_a_time: 1
+    num_frames: &num_frames 25
+    slow_spatial_layers: False
+    train_peft_adapters: True
+    replace_cond_frames: &replace_cond_frames True
+    fixed_cond_frames: # only used for logging images
+      - [ 0 ]
+
+    denoiser_config:
+      target: vista.vwm.modules.diffusionmodules.denoiser.Denoiser
+      params:
+        num_frames: *num_frames
+
+        scaling_config:
+          target: vista.vwm.modules.diffusionmodules.denoiser_scaling.VScalingWithEDMcNoise
+
+    network_config:
+      target: vista.vwm.modules.diffusionmodules.video_model.VideoUNet
+      params:
+        adm_in_channels: 768
+        num_classes: sequential
+        use_checkpoint: True
+        in_channels: 8
+        out_channels: 4
+        model_channels: 320
+        attention_resolutions: [ 4, 2, 1 ]
+        num_res_blocks: 2
+        channel_mult: [ 1, 2, 4, 4 ]
+        num_head_channels: 64
+        use_linear_in_transformer: True
+        transformer_depth: 1
+        context_dim: 1024
+        spatial_transformer_attn_type: softmax-xformers
+        extra_ff_mix_layer: True
+        use_spatial_context: True
+        merge_strategy: learned_with_images
+        video_kernel_size: [ 3, 1, 1 ]
+        add_lora: True
+        action_control: True
+
+    conditioner_config:
+      target: vista.vwm.modules.GeneralConditioner
+      params:
+        emb_models:
+          - input_key: cond_frames_without_noise
+            is_trainable: False
+            ucg_rate: 0.15
+            target: vista.vwm.modules.encoders.modules.FrozenOpenCLIPImagePredictionEmbedder
+            params:
+              n_cond_frames: 1
+              n_copies: 1
+              open_clip_embedding_config:
+                target: vista.vwm.modules.encoders.modules.FrozenOpenCLIPImageEmbedder
+                params:
+                  freeze: True
+
+          - input_key: fps_id
+            is_trainable: False
+            ucg_rate: 0.0
+            target: vista.vwm.modules.encoders.modules.ConcatTimestepEmbedderND
+            params:
+              outdim: 256
+
+          - input_key: motion_bucket_id
+            is_trainable: False
+            ucg_rate: 0.0
+            target: vista.vwm.modules.encoders.modules.ConcatTimestepEmbedderND
+            params:
+              outdim: 256
+
+          - input_key: cond_frames
+            is_trainable: False
+            ucg_rate: 0.15
+            target: vista.vwm.modules.encoders.modules.VideoPredictionEmbedderWithEncoder
+            params:
+              disable_encoder_autocast: True
+              n_cond_frames: 1
+              n_copies: 1
+              is_ae: True
+
+              encoder_config:
+                target: vista.vwm.models.autoencoder.AutoencoderKLModeOnly
+                params:
+                  embed_dim: 4
+                  monitor: val/rec_loss
+
+                  ddconfig:
+                    attn_type: vanilla-xformers
+                    double_z: True
+                    z_channels: 4
+                    resolution: 256
+                    in_channels: 3
+                    out_ch: 3
+                    ch: 128
+                    ch_mult: [ 1, 2, 4, 4 ]
+                    num_res_blocks: 2
+                    attn_resolutions: [ ]
+                    dropout: 0.0
+
+                  loss_config:
+                    target: torch.nn.Identity
+
+          - input_key: cond_aug
+            is_trainable: False
+            ucg_rate: 0.0
+            target: vista.vwm.modules.encoders.modules.ConcatTimestepEmbedderND
+            params:
+              outdim: 256
+
+          - input_key: command
+            is_trainable: False
+            ucg_rate: 0.15
+            target: vista.vwm.modules.encoders.modules.ConcatTimestepEmbedderND
+            params:
+              outdim: &action_emb_dim 128
+              num_features: 1
+              add_sequence_dim: True
+
+          - input_key: trajectory
+            is_trainable: False
+            ucg_rate: 0.15
+            target: vista.vwm.modules.encoders.modules.ConcatTimestepEmbedderND
+            params:
+              outdim: *action_emb_dim
+              num_features: 8
+              add_sequence_dim: True
+
+          - input_key: speed
+            is_trainable: False
+            ucg_rate: 0.15
+            target: vista.vwm.modules.encoders.modules.ConcatTimestepEmbedderND
+            params:
+              outdim: *action_emb_dim
+              num_features: 4
+              add_sequence_dim: True
+
+          - input_key: angle
+            is_trainable: False
+            ucg_rate: 0.15
+            target: vista.vwm.modules.encoders.modules.ConcatTimestepEmbedderND
+            params:
+              outdim: *action_emb_dim
+              num_features: 4
+              add_sequence_dim: True
+
+          - input_key: goal
+            is_trainable: False
+            ucg_rate: 0.15
+            target: vista.vwm.modules.encoders.modules.ConcatTimestepEmbedderND
+            params:
+              outdim: *action_emb_dim
+              num_features: 2
+              add_sequence_dim: True
+
+    first_stage_config:
+      target: vista.vwm.models.autoencoder.AutoencodingEngine
+      params:
+        loss_config:
+          target: torch.nn.Identity
+
+        regularizer_config:
+          target: vista.vwm.modules.autoencoding.regularizers.DiagonalGaussianRegularizer
+
+        encoder_config:
+          target: vista.vwm.modules.diffusionmodules.model.Encoder
+          params:
+            attn_type: vanilla
+            double_z: True
+            z_channels: 4
+            resolution: 256
+            in_channels: 3
+            out_ch: 3
+            ch: 128
+            ch_mult: [ 1, 2, 4, 4 ]
+            num_res_blocks: 2
+            attn_resolutions: [ ]
+            dropout: 0.0
+
+        decoder_config:
+          target: vista.vwm.modules.autoencoding.temporal_ae.VideoDecoder
+          params:
+            attn_type: vanilla
+            double_z: True
+            z_channels: 4
+            resolution: 256
+            in_channels: 3
+            out_ch: 3
+            ch: 128
+            ch_mult: [ 1, 2, 4, 4 ]
+            num_res_blocks: 2
+            attn_resolutions: [ ]
+            dropout: 0.0
+            video_kernel_size: [ 3, 1, 1 ]
+
+    scheduler_config:
+      target: vista.vwm.lr_scheduler.LambdaLinearScheduler
+      params:
+        warm_up_steps: [ 1000 ]
+        cycle_lengths: [ 10000000000000 ]
+        f_start: [ 1.e-6 ]
+        f_max: [ 1. ]
+        f_min: [ 1. ]
+
+    loss_fn_config:
+      target: vista.vwm.modules.diffusionmodules.loss.StandardDiffusionLoss
+      params:
+        use_additional_loss: True
+        offset_noise_level: 0.02
+        additional_loss_weight: 0.1
+        num_frames: *num_frames
+        replace_cond_frames: *replace_cond_frames
+        cond_frames_choices:
+          - [ ]
+          - [ 0 ]
+          - [ 0, 1 ]
+          - [ 0, 1, 2 ]
+
+        sigma_sampler_config:
+          target: vista.vwm.modules.diffusionmodules.sigma_sampling.EDMSampling
+          params:
+            p_mean: 1.0
+            p_std: 1.6
+            num_frames: *num_frames
+
+        loss_weighting_config:
+          target: vista.vwm.modules.diffusionmodules.loss_weighting.VWeighting
+
+    sampler_config:
+      target: vista.vwm.modules.diffusionmodules.sampling.EulerEDMSampler
+      params:
+        num_steps: 15
+
+        discretization_config:
+          target: vista.vwm.modules.diffusionmodules.discretizer.EDMDiscretization
+          params:
+            sigma_max: 700.0
+
+        guider_config:
+          target: vista.vwm.modules.diffusionmodules.guiders.LinearPredictionGuider
+          params:
+            num_frames: *num_frames
+            max_scale: 3.0
+            min_scale: 1.5
+
+data:
+  target: vista.vwm.data.dataset.Sampler
+  params:
+    batch_size: 1
+    num_workers: 16
+    subsets:
+      - YouTube
+      - NuScenes
+    probs:
+      - 1
+      - 1
+    samples_per_epoch: 256000
+    target_height: 320
+    target_width: 576
+    num_frames: *num_frames
+
+lightning:
+  callbacks:
+    image_logger:
+      target: train.ImageLogger
+      params:
+        num_frames: *num_frames
+        disabled: False
+        enable_autocast: False
+        batch_frequency: 1000
+        increase_log_steps: True
+        log_first_step: False
+        log_images_kwargs:
+          N: *num_frames
+
+  modelcheckpoint:
+    params:
+      every_n_epochs: 1  # every_n_train_steps: 5000, set the same as image_logger batch_frequency
+
+  trainer:
+    devices: 0,1
+    benchmark: True
+    num_sanity_val_steps: 0
+    accumulate_grad_batches: 1
+    max_epochs: 100
+    strategy: deepspeed_stage_2
+    gradient_clip_val: 0.3

vista/configs/training/vista_phase2_stage2.yaml

@@ -0,0 +1,294 @@
+model:
+  base_learning_rate: 5.e-5
+  target: vista.vwm.models.diffusion.DiffusionEngine
+  params:
+    use_ema: True
+    input_key: img_seq
+    scale_factor: 0.18215
+    disable_first_stage_autocast: True
+    en_and_decode_n_samples_a_time: 1
+    num_frames: &num_frames 25
+    slow_spatial_layers: False
+    train_peft_adapters: True
+    replace_cond_frames: &replace_cond_frames True
+    fixed_cond_frames: # only used for logging images
+      - [ 0 ]
+
+    denoiser_config:
+      target: vista.vwm.modules.diffusionmodules.denoiser.Denoiser
+      params:
+        num_frames: *num_frames
+
+        scaling_config:
+          target: vista.vwm.modules.diffusionmodules.denoiser_scaling.VScalingWithEDMcNoise
+
+    network_config:
+      target: vista.vwm.modules.diffusionmodules.video_model.VideoUNet
+      params:
+        adm_in_channels: 768
+        num_classes: sequential
+        use_checkpoint: True
+        in_channels: 8
+        out_channels: 4
+        model_channels: 320
+        attention_resolutions: [ 4, 2, 1 ]
+        num_res_blocks: 2
+        channel_mult: [ 1, 2, 4, 4 ]
+        num_head_channels: 64
+        use_linear_in_transformer: True
+        transformer_depth: 1
+        context_dim: 1024
+        spatial_transformer_attn_type: softmax-xformers
+        extra_ff_mix_layer: True
+        use_spatial_context: True
+        merge_strategy: learned_with_images
+        video_kernel_size: [ 3, 1, 1 ]
+        add_lora: True
+        action_control: True
+
+    conditioner_config:
+      target: vista.vwm.modules.GeneralConditioner
+      params:
+        emb_models:
+          - input_key: cond_frames_without_noise
+            is_trainable: False
+            ucg_rate: 0.15
+            target: vista.vwm.modules.encoders.modules.FrozenOpenCLIPImagePredictionEmbedder
+            params:
+              n_cond_frames: 1
+              n_copies: 1
+              open_clip_embedding_config:
+                target: vista.vwm.modules.encoders.modules.FrozenOpenCLIPImageEmbedder
+                params:
+                  freeze: True
+
+          - input_key: fps_id
+            is_trainable: False
+            ucg_rate: 0.0
+            target: vista.vwm.modules.encoders.modules.ConcatTimestepEmbedderND
+            params:
+              outdim: 256
+
+          - input_key: motion_bucket_id
+            is_trainable: False
+            ucg_rate: 0.0
+            target: vista.vwm.modules.encoders.modules.ConcatTimestepEmbedderND
+            params:
+              outdim: 256
+
+          - input_key: cond_frames
+            is_trainable: False
+            ucg_rate: 0.15
+            target: vista.vwm.modules.encoders.modules.VideoPredictionEmbedderWithEncoder
+            params:
+              disable_encoder_autocast: True
+              n_cond_frames: 1
+              n_copies: 1
+              is_ae: True
+
+              encoder_config:
+                target: vista.vwm.models.autoencoder.AutoencoderKLModeOnly
+                params:
+                  embed_dim: 4
+                  monitor: val/rec_loss
+
+                  ddconfig:
+                    attn_type: vanilla-xformers
+                    double_z: True
+                    z_channels: 4
+                    resolution: 256
+                    in_channels: 3
+                    out_ch: 3
+                    ch: 128
+                    ch_mult: [ 1, 2, 4, 4 ]
+                    num_res_blocks: 2
+                    attn_resolutions: [ ]
+                    dropout: 0.0
+
+                  loss_config:
+                    target: torch.nn.Identity
+
+          - input_key: cond_aug
+            is_trainable: False
+            ucg_rate: 0.0
+            target: vista.vwm.modules.encoders.modules.ConcatTimestepEmbedderND
+            params:
+              outdim: 256
+
+          - input_key: command
+            is_trainable: False
+            ucg_rate: 0.15
+            target: vista.vwm.modules.encoders.modules.ConcatTimestepEmbedderND
+            params:
+              outdim: &action_emb_dim 128
+              num_features: 1
+              add_sequence_dim: True
+
+          - input_key: trajectory
+            is_trainable: False
+            ucg_rate: 0.15
+            target: vista.vwm.modules.encoders.modules.ConcatTimestepEmbedderND
+            params:
+              outdim: *action_emb_dim
+              num_features: 8
+              add_sequence_dim: True
+
+          - input_key: speed
+            is_trainable: False
+            ucg_rate: 0.15
+            target: vista.vwm.modules.encoders.modules.ConcatTimestepEmbedderND
+            params:
+              outdim: *action_emb_dim
+              num_features: 4
+              add_sequence_dim: True
+
+          - input_key: angle
+            is_trainable: False
+            ucg_rate: 0.15
+            target: vista.vwm.modules.encoders.modules.ConcatTimestepEmbedderND
+            params:
+              outdim: *action_emb_dim
+              num_features: 4
+              add_sequence_dim: True
+
+          - input_key: goal
+            is_trainable: False
+            ucg_rate: 0.15
+            target: vista.vwm.modules.encoders.modules.ConcatTimestepEmbedderND
+            params:
+              outdim: *action_emb_dim
+              num_features: 2
+              add_sequence_dim: True
+
+    first_stage_config:
+      target: vista.vwm.models.autoencoder.AutoencodingEngine
+      params:
+        loss_config:
+          target: torch.nn.Identity
+
+        regularizer_config:
+          target: vista.vwm.modules.autoencoding.regularizers.DiagonalGaussianRegularizer
+
+        encoder_config:
+          target: vista.vwm.modules.diffusionmodules.model.Encoder
+          params:
+            attn_type: vanilla
+            double_z: True
+            z_channels: 4
+            resolution: 256
+            in_channels: 3
+            out_ch: 3
+            ch: 128
+            ch_mult: [ 1, 2, 4, 4 ]
+            num_res_blocks: 2
+            attn_resolutions: [ ]
+            dropout: 0.0
+
+        decoder_config:
+          target: vista.vwm.modules.autoencoding.temporal_ae.VideoDecoder
+          params:
+            attn_type: vanilla
+            double_z: True
+            z_channels: 4
+            resolution: 256
+            in_channels: 3
+            out_ch: 3
+            ch: 128
+            ch_mult: [ 1, 2, 4, 4 ]
+            num_res_blocks: 2
+            attn_resolutions: [ ]
+            dropout: 0.0
+            video_kernel_size: [ 3, 1, 1 ]
+
+    scheduler_config:
+      target: vista.vwm.lr_scheduler.LambdaLinearScheduler
+      params:
+        warm_up_steps: [ 1000 ]
+        cycle_lengths: [ 10000000000000 ]
+        f_start: [ 1.e-6 ]
+        f_max: [ 1. ]
+        f_min: [ 1. ]
+
+    loss_fn_config:
+      target: vista.vwm.modules.diffusionmodules.loss.StandardDiffusionLoss
+      params:
+        use_additional_loss: True
+        offset_noise_level: 0.02
+        additional_loss_weight: 0.1
+        num_frames: *num_frames
+        replace_cond_frames: *replace_cond_frames
+        cond_frames_choices:
+          - [ ]
+          - [ 0 ]
+          - [ 0, 1 ]
+          - [ 0, 1, 2 ]
+
+        sigma_sampler_config:
+          target: vista.vwm.modules.diffusionmodules.sigma_sampling.EDMSampling
+          params:
+            p_mean: 1.0
+            p_std: 1.6
+            num_frames: *num_frames
+
+        loss_weighting_config:
+          target: vista.vwm.modules.diffusionmodules.loss_weighting.VWeighting
+
+    sampler_config:
+      target: vista.vwm.modules.diffusionmodules.sampling.EulerEDMSampler
+      params:
+        num_steps: 15
+
+        discretization_config:
+          target: vista.vwm.modules.diffusionmodules.discretizer.EDMDiscretization
+          params:
+            sigma_max: 700.0
+
+        guider_config:
+          target: vista.vwm.modules.diffusionmodules.guiders.LinearPredictionGuider
+          params:
+            num_frames: *num_frames
+            max_scale: 3.0
+            min_scale: 1.5
+
+data:
+  target: vista.vwm.data.dataset.Sampler
+  params:
+    batch_size: 1
+    num_workers: 16
+    subsets:
+      - YouTube
+      - NuScenes
+    probs:
+      - 1
+      - 1
+    samples_per_epoch: 256000
+    target_height: 576
+    target_width: 1024
+    num_frames: *num_frames
+
+lightning:
+  callbacks:
+    image_logger:
+      target: train.ImageLogger
+      params:
+        num_frames: *num_frames
+        disabled: False
+        enable_autocast: False
+        batch_frequency: 1000
+        increase_log_steps: True
+        log_first_step: False
+        log_images_kwargs:
+          N: *num_frames
+
+  modelcheckpoint:
+    params:
+      every_n_epochs: 1  # every_n_train_steps: 5000, set the same as image_logger batch_frequency
+
+  trainer:
+    devices: 0,1
+    benchmark: True
+    num_sanity_val_steps: 0
+    accumulate_grad_batches: 1
+    max_epochs: 100
+    strategy: deepspeed_stage_2
+    gradient_clip_val: 0.3

vista/docs/INSTALL.md

@@ -0,0 +1,48 @@
+## Installation
+
+- ### Requirement
+
+  Our experiments are conducted with **PyTorch 2.0.1**, **CUDA 11.7**, **Ubuntu 22.04**, and **NVIDIA Tesla A100** (80 GB). For other requirements, please check [TRAINING.md](https://github.com/OpenDriveLab/Vista/blob/main/docs/TRAINING.md) and [SAMPLING.md](https://github.com/OpenDriveLab/Vista/blob/main/docs/SAMPLING.md).
+
+- ### Preparation
+
+  Clone the repository to your local directory.
+
+  ```shell
+  git clone https://github.com/OpenDriveLab/Vista.git
+  ```
+
+  We provide an example on nuScenes dataset for training and sampling. Before you start, make sure you have:
+  
+  - Downloaded the translated action annotations from [here](https://drive.google.com/drive/folders/1JpZObdR0OXagCbnPZfMSI8vhGLom5pht?usp=sharing) and put the JSON files into `annos`.
+    
+  - Downloaded all splits of **Trainval** in **Full dataset (v1.0)** to your device following [official instructions](https://www.nuscenes.org/download). After downloading, it should contain:
+
+    ```
+    $<your-nusc-data-root>
+    ├── samples
+    ├── sweeps
+    ├── ...
+    └── v1.0-trainval
+    ```
+
+- ### Installation
+
+  - We use conda to manage the environment.
+
+    ```shell
+    conda create -n vista python=3.9 -y
+    conda activate vista
+    ```
+  
+  - Install dependencies.
+  
+    ```shell
+    conda install -y pytorch==2.0.1 torchvision==0.15.2 torchaudio==2.0.2 pytorch-cuda=11.7 -c pytorch -c nvidia
+    pip3 install -r requirements.txt
+    pip3 install -e git+https://github.com/Stability-AI/datapipelines.git@main#egg=sdata
+    ```
+
+---
+
+=> Next: [[Training](https://github.com/OpenDriveLab/Vista/blob/main/docs/TRAINING.md)]

vista/docs/ISSUES.md

@@ -0,0 +1,36 @@
+## Trouble Shooting
+
+1. #### Out of memory during sampling.
+
+   - Possible reason:
+     - Too many high-resolution frames for parallel decoding. The default setting will request ca. 66 GB peak VARM.
+
+   - Try this:
+     - Reduce the number of jointly decoded frames *en_and_decode_n_samples_a_time* in `inference/vista.yaml`.
+
+2. #### Get stuck at loading FrozenCLIPEmbedder or FrozenOpenCLIPImageEmbedder.
+
+   - Possible reason:
+     - A network failure.
+   
+   - Try this:
+     1. Download [openai/clip-vit-large-patch14](https://huggingface.co/openai/clip-vit-large-patch14/tree/main) and [laion/CLIP-ViT-H-14-laion2B-s32B-b79K](https://huggingface.co/laion/CLIP-ViT-H-14-laion2B-s32B-b79K/tree/main) in advance.
+     2. Set *version* of FrozenCLIPEmbedder and FrozenOpenCLIPImageEmbedder in `vwm/modules/encoders/modules.py` to the new paths of `pytorch_model.bin`.
+
+3. #### Datasets not yet available during training.
+
+   - Possible reason:
+
+     - The installed [sdata](https://github.com/Stability-AI/datapipelines) is not detected.
+
+   - Try this:
+
+     - Reinstall in the current project directory.
+
+       ````shell
+       pip3 install -e git+https://github.com/Stability-AI/datapipelines.git@main#egg=sdata
+       ````
+
+---
+
+<= Previous: [[Sampling](https://github.com/OpenDriveLab/Vista/blob/main/docs/SAMPLING.md)]

vista/docs/SAMPLING.md

@@ -0,0 +1,60 @@
+## Sampling
+
+- ### Requirement
+
+  Currently, we suggest using Nvidia GPUs with a minimum of **32 GB** VRAM for sampling. Check [ISSUES.md](https://github.com/OpenDriveLab/Vista/blob/main/docs/ISSUES.md) if you do not have enough memory.
+
+- ### Preparation
+
+  Make sure you have downloaded `vista.safetensors` from [Hugging Face](https://huggingface.co/OpenDriveLab/Vista/blob/main/vista.safetensors) or [Google Drive](https://drive.google.com/file/d/1bCM7XLDquRqnnpauQAK5j1jP-n0y1ama/view). Move (or link) the checkpoint into `ckpts`.
+
+- ### Future Prediction
+
+  - We provide a sampling example for nuScenes. Make sure to prepare the dataset as [INSTALL.md](https://github.com/OpenDriveLab/Vista/blob/main/docs/INSTALL.md) and replace the correct *data_root* in `sample.py`.
+
+    - Short-term action-free prediction.
+
+      ```shell
+      python sample.py
+      ```
+
+    - Long-term rollout.
+    
+      ```shell
+      python sample.py --n_rounds 6
+      ```
+    
+    - Action-conditioned simulation (take trajectory as an example).
+    
+      ```shell
+      python sample.py --action traj
+      ```
+      
+    > Make sure the loaded checkpoint strictly match all parameters. Otherwise, you may get a sequence of blur.
+
+  - Important arguments:
+
+    - `--dataset`: You can also customize the scenes by providing other driving views within a folder of images. They will serve as the initial frames for prediction when you set `--dataset` to "IMG".
+    - `--action`: The mode of control inputs. By default, we perform action-free prediction. You can try different actions using "traj", "cmd", "steer", or "goal". It will import ground truth actions (if available), but you can enforce any actions by making slight modifications.
+    - `--n_rounds`: The number of sampling rounds, which determines the duration to predict. You can increase it to perform long-horizon rollout. Each additional round extends the prediction by 2.3 seconds.
+    - `--n_steps`: The number of DDIM sampling steps, which can be reduced for efficiency.
+    - `--rand_gen`: Whether to generate samples randomly selected from the whole dataset or go through all samples one by one. 
+    - `--low_vram`: Enable the low VRAM mode if you are using a GPU with less than 80 GB VRAM.
+
+- ### Reward Estimation
+
+  - We provide a simplified example to estimate the rewards on nuScenes. Make sure to replace the correct *data_root* in `reward.py`.
+
+    ```shell
+    python reward.py
+    ```
+
+  - Important arguments:
+  
+    - `--ens_size`: The number of samples to generate per case (initial frame and action condition).
+
+---
+
+<= Previous: [[Training](https://github.com/OpenDriveLab/Vista/blob/main/docs/TRAINING.md)]
+
+=> Next: [[Trouble Shooting](https://github.com/OpenDriveLab/Vista/blob/main/docs/ISSUES.md)]

vista/docs/TRAINING.md

@@ -0,0 +1,113 @@
+## Training
+
+- ### Requirement
+
+  Nvidia GPUs with **80 GB** VRAM are required for training, but you can train low-resolution variants on smaller GPUs.
+
+- ### Preparation
+
+  Download the pretrained `svd_xt.safetensors` from [Hugging Face](https://huggingface.co/stabilityai/stable-video-diffusion-img2vid-xt/blob/main/svd_xt.safetensors) and place the checkpoint into `ckpts`.
+
+- ### Training (example)
+
+  - We take **nuScenes** dataset as an example for training. After finishing the setups in [INSTALL.md](https://github.com/OpenDriveLab/Vista/blob/main/docs/INSTALL.md), remember to edit *data_root* in `vwm/data/subsets/nuscenes.py` to the proper path of nuScenes.
+
+  - We use DeepSpeed ZeRO stage 2 to improve data parallelism and lower memory footprint during training. The training can be launched as: 
+
+    - Distributed training (suppose you train with 2 nodes, and each node has 8 GPUs).
+
+      ```shell
+      torchrun \
+          --nnodes=2 \
+          --nproc_per_node=8 \
+          train.py \
+          --base configs/example/nusc_train.yaml \
+          --num_nodes 2 \
+          --n_devices 8
+      ```
+
+    - Single GPU debugging (too slow, not recommended for training).
+
+      ```shell
+      python train.py --num_nodes 1 --n_devices 1
+      ```
+
+    > The training logs, including visualization samples and model checkpoints, will be saved in the project directory by default. Given that the size of checkpoints could be very large, you can set another directory to save these logs by providing an available path to `--logdir`.
+    >
+    > You can disable `--no_test` to test a bunch of samples for every checkpoint, but we recommend evaluating them offline for flexible comparison and uninterrupted training.
+
+  - After training, switch to the log directory with the model checkpoint. You should find a Python script named `zero_to_fp32.py` and a `checkpoint` folder that contains all partitioned checkpoints. The final checkpoint can be obtained by:
+
+    1. [*if you only want to resume training*] Merge the partitioned checkpoints as `pytorch_model.bin` using `zero_to_fp32.py`.
+    
+       ```shell
+       python zero_to_fp32.py . pytorch_model.bin
+       ```
+    
+    2. [*if you also want to do inference*] Navigate into the project root, and use `bin_to_st.py` to convert the resulting `path_to/pytorch_model.bin` to `ckpts/vista.safetensors`.
+
+- ### Training of Vista
+
+  - Download **OpenDV-YouTube** dataset (or a part of it) from [DriveAGI](https://github.com/OpenDriveLab/DriveAGI#genad-dataset-opendv-youtube). You can refer to the structure in `vwm/data/subsets/youtube.py` to organize the dataset.
+  
+  - #### Phase 1: learning high-fidelity future prediction
+  
+    - This phase uses unlabeled OpenDV-YouTube for training.
+  
+    - The model is trained at a resolution of 576x1024 on 128 GPUs for 20K iterations with gradient accumulation.
+  
+      ```shell
+      torchrun \
+          --nnodes=16 \
+          --nproc_per_node=8 \
+          train.py \
+          --base configs/training/vista_phase1.yaml \
+          --num_nodes 16 \
+          --n_devices 8
+      ```
+  
+    - We pause the training after the effect of dynamics priors can be witnessed. The last checkpoint is merged for the training of next phase.
+  
+  - #### Phase 2:  learning versatile action controllability
+  
+    - This phase uses OpenDV-YouTube and nuScenes for collaborative training.
+  
+    - ##### Stage 1: low-resolution training
+  
+      - The model is finetuned at a resolution of 320x576 on 8 GPUs for 120K iterations.
+  
+        ```shell
+        torchrun \
+            --nnodes=1 \
+            --nproc_per_node=8 \
+            train.py \
+            --base configs/training/vista_phase2_stage1.yaml \
+            --finetune ${PATH_TO_PHASE1_CKPT}/pytorch_model.bin \
+            --num_nodes 1 \
+            --n_devices 8
+        ```
+  
+      - We pause the training after the controllability can be clearly witnessed. The last checkpoint is merged for the training of next stage.
+  
+    - ##### Stage 2: high-resolution training
+  
+      - The model is finetuned at a resolution of 576x1024 on 8 GPUs for another 10K iterations.
+  
+        ```shell
+        torchrun \
+            --nnodes=1 \
+            --nproc_per_node=8 \
+            train.py \
+            --base configs/training/vista_phase2_stage2.yaml \
+            --finetune ${PATH_TO_STAGE1_CKPT}/pytorch_model.bin \
+            --num_nodes 1 \
+            --n_devices 8
+        ```
+  
+      - We pause the training after the model adapt to the desired resolution. The last checkpoint is merged for application.
+
+---
+
+<= Previous: [[Installation](https://github.com/OpenDriveLab/Vista/blob/main/docs/INSTALL.md)]
+
+=> Next: [[Sampling](https://github.com/OpenDriveLab/Vista/blob/main/docs/SAMPLING.md)]

vista/reward.py

@@ -0,0 +1,266 @@
+from __future__ import annotations
+
+import argparse
+import json
+import random
+
+from pytorch_lightning import seed_everything
+from reward_utils import *
+from torchvision import transforms
+
+VERSION2SPECS = {
+    "vwm": {
+        "config": "configs/inference/vista.yaml",
+        "ckpt": "ckpts/vista.safetensors"
+    }
+}
+
+DATASET2SOURCES = {
+    "NUSCENES": {
+        "data_root": "data/nuscenes",
+        "anno_file": "annos/nuScenes_val.json"
+    },
+    "IMG": {
+        "data_root": "image_folder"
+    }
+}
+
+
+def parse_args(**parser_kwargs):
+    parser = argparse.ArgumentParser(**parser_kwargs)
+    parser.add_argument(
+        "--version",
+        type=str,
+        default="vwm",
+        help="model version"
+    )
+    parser.add_argument(
+        "--dataset",
+        type=str,
+        default="NUSCENES",
+        help="dataset name"
+    )
+    parser.add_argument(
+        "--save",
+        type=str,
+        default="outputs",
+        help="directory to save samples"
+    )
+    parser.add_argument(
+        "--action",
+        type=str,
+        default="traj",
+        help="action mode for control, such as traj, cmd, steer, goal"
+    )
+    parser.add_argument(
+        "--n_frames",
+        type=int,
+        default=25,
+        help="number of frames for each round"
+    )
+    parser.add_argument(
+        "--n_conds",
+        type=int,
+        default=1,
+        help="number of initial condition frames for the first round"
+    )
+    parser.add_argument(
+        "--ens_size",
+        type=int,
+        default=5,
+        help="number of samples per case"
+    )
+    parser.add_argument(
+        "--seed",
+        type=int,
+        default=23,
+        help="random seed for seed_everything"
+    )
+    parser.add_argument(
+        "--height",
+        type=int,
+        default=576,
+        help="target height of the generated video"
+    )
+    parser.add_argument(
+        "--width",
+        type=int,
+        default=1024,
+        help="target width of the generated video"
+    )
+    parser.add_argument(
+        "--cfg_scale",
+        type=float,
+        default=2.5,
+        help="scale of the classifier-free guidance"
+    )
+    parser.add_argument(
+        "--cond_aug",
+        type=float,
+        default=0.0,
+        help="strength of the noise augmentation"
+    )
+    parser.add_argument(
+        "--n_steps",
+        type=int,
+        default=10,
+        help="number of sampling steps"
+    )
+    parser.add_argument(
+        "--rand_gen",
+        action="store_false",
+        help="whether to generate samples randomly or sequentially"
+    )
+    parser.add_argument(
+        "--low_vram",
+        action="store_true",
+        help="whether to save memory or not"
+    )
+    return parser
+
+
+def get_sample(selected_index=0, dataset_name="NUSCENES", num_frames=25, action_mode="free"):
+    dataset_dict = DATASET2SOURCES[dataset_name]
+    action_dict = None
+    if dataset_name == "IMG":
+        image_list = os.listdir(dataset_dict["data_root"])
+        total_length = len(image_list)
+        while selected_index >= total_length:
+            selected_index -= total_length
+        image_file = image_list[selected_index]
+
+        path_list = [os.path.join(dataset_dict["data_root"], image_file)] * num_frames
+    else:
+        with open(dataset_dict["anno_file"]) as anno_json:
+            all_samples = json.load(anno_json)
+        total_length = len(all_samples)
+        while selected_index >= total_length:
+            selected_index -= total_length
+        sample_dict = all_samples[selected_index]
+
+        path_list = list()
+        if dataset_name == "NUSCENES":
+            for index in range(num_frames):
+                image_path = os.path.join(dataset_dict["data_root"], sample_dict["frames"][index])
+                assert os.path.exists(image_path), image_path
+                path_list.append(image_path)
+            action_dict = dict()
+            if action_mode == "traj" or action_mode == "trajectory":
+                action_dict["trajectory"] = torch.tensor(sample_dict["traj"][2:])
+            elif action_mode == "cmd" or action_mode == "command":
+                action_dict["command"] = torch.tensor(sample_dict["cmd"])
+            elif action_mode == "steer":
+                # scene might be empty
+                if sample_dict["speed"]:
+                    action_dict["speed"] = torch.tensor(sample_dict["speed"][1:])
+                # scene might be empty
+                if sample_dict["angle"]:
+                    action_dict["angle"] = torch.tensor(sample_dict["angle"][1:]) / 780
+            elif action_mode == "goal":
+                # point might be invalid
+                if sample_dict["z"] > 0 and 0 < sample_dict["goal"][0] < 1600 and 0 < sample_dict["goal"][1] < 900:
+                    action_dict["goal"] = torch.tensor([
+                        sample_dict["goal"][0] / 1600,
+                        sample_dict["goal"][1] / 900
+                    ])
+            else:
+                raise ValueError(f"Unsupported action mode {action_mode}")
+        else:
+            raise ValueError(f"Invalid dataset {dataset_name}")
+    return path_list, selected_index, total_length, action_dict
+
+
+def load_img(file_name, target_height=320, target_width=576, device="cuda"):
+    if file_name is not None:
+        image = Image.open(file_name)
+        if not image.mode == "RGB":
+            image = image.convert("RGB")
+    else:
+        raise ValueError(f"Invalid image file {file_name}")
+    ori_w, ori_h = image.size
+    # print(f"Loaded input image of size ({ori_w}, {ori_h})")
+
+    if ori_w / ori_h > target_width / target_height:
+        tmp_w = int(target_width / target_height * ori_h)
+        left = (ori_w - tmp_w) // 2
+        right = (ori_w + tmp_w) // 2
+        image = image.crop((left, 0, right, ori_h))
+    elif ori_w / ori_h < target_width / target_height:
+        tmp_h = int(target_height / target_width * ori_w)
+        top = (ori_h - tmp_h) // 2
+        bottom = (ori_h + tmp_h) // 2
+        image = image.crop((0, top, ori_w, bottom))
+    image = image.resize((target_width, target_height), resample=Image.LANCZOS)
+    if not image.mode == "RGB":
+        image = image.convert("RGB")
+    image = transforms.Compose([
+        transforms.ToTensor(),
+        transforms.Lambda(lambda x: x * 2.0 - 1.0)
+    ])(image)
+    return image.to(device)
+
+
+if __name__ == "__main__":
+    parser = parse_args()
+    opt, unknown = parser.parse_known_args()
+
+    set_lowvram_mode(opt.low_vram)
+    version_dict = VERSION2SPECS[opt.version]
+    model = init_model(version_dict)
+    unique_keys = set([x.input_key for x in model.conditioner.embedders])
+
+    sample_index = 0
+    while sample_index >= 0:
+        seed_everything(opt.seed)
+
+        frame_list, sample_index, dataset_length, action_dict = get_sample(sample_index,
+                                                                           opt.dataset,
+                                                                           opt.n_frames,
+                                                                           opt.action)
+
+        img_seq = list()
+        for each_path in frame_list:
+            img = load_img(each_path, opt.height, opt.width)
+            img_seq.append(img)
+        images = torch.stack(img_seq)
+
+        value_dict = init_embedder_options(unique_keys)
+        cond_img = img_seq[0][None]
+        value_dict["cond_frames_without_noise"] = cond_img
+        value_dict["cond_aug"] = opt.cond_aug
+        value_dict["cond_frames"] = cond_img + opt.cond_aug * torch.randn_like(cond_img)
+        if action_dict is not None:
+            for key, value in action_dict.items():
+                value_dict[key] = value
+
+        guider = "VanillaCFG"
+        sampler = init_sampling(guider=guider, steps=opt.n_steps, cfg_scale=opt.cfg_scale, num_frames=opt.n_frames)
+
+        uc_keys = ["cond_frames", "cond_frames_without_noise", "command", "trajectory", "speed", "angle", "goal"]
+
+        out = do_sample(
+            images,
+            model,
+            sampler,
+            value_dict,
+            num_frames=opt.n_frames,
+            ensemble_size=opt.ens_size,
+            force_uc_zero_embeddings=uc_keys,
+            initial_cond_indices=[index for index in range(opt.n_conds)]
+        )
+
+        if isinstance(out, (tuple, list)):
+            inputs, reward = out
+            real_path = os.path.join(opt.save, "real")
+            perform_save_locally(real_path, inputs, "videos", opt.dataset, sample_index)
+            perform_save_locally(real_path, inputs, "grids", opt.dataset, sample_index)
+            perform_save_locally(real_path, inputs, "images", opt.dataset, sample_index)
+        else:
+            raise TypeError
+
+        if opt.rand_gen:
+            sample_index += random.randint(1, dataset_length - 1)
+        else:
+            sample_index += 1
+            if dataset_length <= sample_index:
+                sample_index = -1

vista/reward_utils.py

@@ -0,0 +1,342 @@
+from __future__ import annotations
+
+import math
+import os
+from typing import Optional, Union
+
+import numpy as np
+import torch
+import torchvision
+from einops import rearrange, repeat
+from omegaconf import ListConfig, OmegaConf
+from PIL import Image
+from safetensors.torch import load_file as load_safetensors
+from torch import autocast
+from train import save_img_seq_to_video
+from vwm.modules.diffusionmodules.sampling import EulerEDMSampler
+from vwm.util import default, instantiate_from_config
+
+
+def init_model(version_dict, load_ckpt=True):
+    config = OmegaConf.load(version_dict["config"])
+    model = load_model_from_config(config, version_dict["ckpt"] if load_ckpt else None)
+    return model
+
+
+lowvram_mode = True
+
+
+def set_lowvram_mode(mode):
+    global lowvram_mode
+    lowvram_mode = mode
+
+
+def initial_model_load(model):
+    global lowvram_mode
+    if lowvram_mode:
+        model.model.half()
+    else:
+        model.cuda()
+    return model
+
+
+def load_model(model):
+    model.cuda()
+
+
+def unload_model(model):
+    global lowvram_mode
+    if lowvram_mode:
+        model.cpu()
+        torch.cuda.empty_cache()
+
+
+def load_model_from_config(config, ckpt=None):
+    model = instantiate_from_config(config.model)
+
+    if ckpt is not None:
+        print(f"Loading model from {ckpt}")
+        if ckpt.endswith("ckpt"):
+            pl_svd = torch.load(ckpt, map_location="cpu")
+            # dict contains:
+            # "epoch", "global_step", "pytorch-lightning_version",
+            # "state_dict", "loops", "callbacks", "optimizer_states", "lr_schedulers"
+            if "global_step" in pl_svd:
+                print(f"Global step: {pl_svd['global_step']}")
+            svd = pl_svd["state_dict"]
+        elif ckpt.endswith("safetensors"):
+            svd = load_safetensors(ckpt)
+        else:
+            raise NotImplementedError("Please convert the checkpoint to safetensors first")
+
+        missing, unexpected = model.load_state_dict(svd, strict=False)
+        if len(missing) > 0:
+            print(f"Missing keys: {missing}")
+        if len(unexpected) > 0:
+            print(f"Unexpected keys: {unexpected}")
+
+    model = initial_model_load(model)
+    model.eval()
+    return model
+
+
+def init_embedder_options(keys):
+    # hardcoded demo settings, might undergo some changes in the future
+    value_dict = dict()
+    for key in keys:
+        if key in ["fps_id", "fps"]:
+            fps = 10
+            value_dict["fps"] = fps
+            value_dict["fps_id"] = fps - 1
+        elif key == "motion_bucket_id":
+            value_dict["motion_bucket_id"] = 127  # [0, 511]
+    return value_dict
+
+
+def perform_save_locally(save_path, samples, mode, dataset_name, sample_index):
+    assert mode in ["images", "grids", "videos"]
+    merged_path = os.path.join(save_path, mode)
+    os.makedirs(merged_path, exist_ok=True)
+    samples = samples.cpu()
+
+    if mode == "images":
+        frame_count = 0
+        for sample in samples:
+            sample = rearrange(sample.numpy(), "c h w -> h w c")
+            if "real" in save_path:
+                sample = 255.0 * (sample + 1.0) / 2.0
+            else:
+                sample = 255.0 * sample
+            image_save_path = os.path.join(merged_path, f"{dataset_name}_{sample_index:06}_{frame_count:04}.png")
+            # if os.path.exists(image_save_path):
+            #     return
+            Image.fromarray(sample.astype(np.uint8)).save(image_save_path)
+            frame_count += 1
+    elif mode == "grids":
+        grid = torchvision.utils.make_grid(samples, nrow=int(samples.shape[0] ** 0.5))
+        grid = grid.transpose(0, 1).transpose(1, 2).squeeze(-1).numpy()
+        if "real" in save_path:
+            grid = 255.0 * (grid + 1.0) / 2.0
+        else:
+            grid = 255.0 * grid
+        grid_save_path = os.path.join(merged_path, f"{dataset_name}_{sample_index:06}.png")
+        # if os.path.exists(grid_save_path):
+        #     return
+        Image.fromarray(grid.astype(np.uint8)).save(grid_save_path)
+    elif mode == "videos":
+        img_seq = rearrange(samples.numpy(), "t c h w -> t h w c")
+        if "real" in save_path:
+            img_seq = 255.0 * (img_seq + 1.0) / 2.0
+        else:
+            img_seq = 255.0 * img_seq
+        video_save_path = os.path.join(merged_path, f"{dataset_name}_{sample_index:06}.mp4")
+        # if os.path.exists(video_save_path):
+        #     return
+        save_img_seq_to_video(video_save_path, img_seq.astype(np.uint8), 10)
+    else:
+        raise NotImplementedError
+
+
+def init_sampling(sampler="EulerEDMSampler", guider="VanillaCFG", discretization="EDMDiscretization",
+                  steps=50, cfg_scale=2.5, num_frames=25):
+    discretization_config = get_discretization(discretization)
+    guider_config = get_guider(guider, cfg_scale, num_frames)
+    sampler = get_sampler(sampler, steps, discretization_config, guider_config)
+    return sampler
+
+
+def get_discretization(discretization):
+    if discretization == "LegacyDDPMDiscretization":
+        discretization_config = {
+            "target": "vista.vwm.modules.diffusionmodules.discretizer.LegacyDDPMDiscretization"
+        }
+    elif discretization == "EDMDiscretization":
+        discretization_config = {
+            "target": "vista.vwm.modules.diffusionmodules.discretizer.EDMDiscretization",
+            "params": {
+                "sigma_min": 0.002,
+                "sigma_max": 700.0,
+                "rho": 7.0
+            }
+        }
+    else:
+        raise NotImplementedError
+    return discretization_config
+
+
+def get_guider(guider="LinearPredictionGuider", cfg_scale=2.5, num_frames=25):
+    if guider == "IdentityGuider":
+        guider_config = {
+            "target": "vista.vwm.modules.diffusionmodules.guiders.IdentityGuider"
+        }
+    elif guider == "VanillaCFG":
+        scale = cfg_scale
+
+        guider_config = {
+            "target": "vista.vwm.modules.diffusionmodules.guiders.VanillaCFG",
+            "params": {
+                "scale": scale
+            }
+        }
+    elif guider == "LinearPredictionGuider":
+        max_scale = cfg_scale
+        min_scale = 1.0
+
+        guider_config = {
+            "target": "vista.vwm.modules.diffusionmodules.guiders.LinearPredictionGuider",
+            "params": {
+                "max_scale": max_scale,
+                "min_scale": min_scale,
+                "num_frames": num_frames
+            }
+        }
+    elif guider == "TrianglePredictionGuider":
+        max_scale = cfg_scale
+        min_scale = 1.0
+
+        guider_config = {
+            "target": "vista.vwm.modules.diffusionmodules.guiders.TrianglePredictionGuider",
+            "params": {
+                "max_scale": max_scale,
+                "min_scale": min_scale,
+                "num_frames": num_frames
+            }
+        }
+    else:
+        raise NotImplementedError
+    return guider_config
+
+
+def get_sampler(sampler, steps, discretization_config, guider_config):
+    if sampler == "EulerEDMSampler":
+        s_churn = 0.0
+        s_tmin = 0.0
+        s_tmax = 999.0
+        s_noise = 1.0
+
+        sampler = EulerEDMSampler(
+            num_steps=steps,
+            discretization_config=discretization_config,
+            guider_config=guider_config,
+            s_churn=s_churn,
+            s_tmin=s_tmin,
+            s_tmax=s_tmax,
+            s_noise=s_noise,
+            verbose=False
+        )
+    else:
+        raise ValueError(f"Unknown sampler {sampler}")
+    return sampler
+
+
+def get_batch(keys, value_dict, N: Union[list, ListConfig], device="cuda"):
+    # hardcoded demo setups, might undergo some changes in the future
+    batch = dict()
+    batch_uc = dict()
+
+    for key in keys:
+        if key in value_dict:
+            if key in ["fps", "fps_id", "motion_bucket_id", "cond_aug"]:
+                batch[key] = repeat(torch.tensor([value_dict[key]]).to(device), "1 -> b", b=math.prod(N))
+            elif key in ["command", "trajectory", "speed", "angle", "goal"]:
+                batch[key] = repeat(value_dict[key][None].to(device), "1 ... -> b ...", b=N[0])
+            elif key in ["cond_frames", "cond_frames_without_noise"]:
+                batch[key] = repeat(value_dict[key], "1 ... -> b ...", b=N[0])
+            else:
+                # batch[key] = value_dict[key]
+                raise NotImplementedError
+
+    for key in batch.keys():
+        if key not in batch_uc and isinstance(batch[key], torch.Tensor):
+            batch_uc[key] = torch.clone(batch[key])
+    return batch, batch_uc
+
+
+def get_condition(model, value_dict, num_samples, force_uc_zero_embeddings, device):
+    load_model(model.conditioner)
+    batch, batch_uc = get_batch(
+        list(set([x.input_key for x in model.conditioner.embedders])),
+        value_dict,
+        [num_samples]
+    )
+    c, uc = model.conditioner.get_unconditional_conditioning(
+        batch,
+        batch_uc=batch_uc,
+        force_uc_zero_embeddings=force_uc_zero_embeddings
+    )
+    unload_model(model.conditioner)
+
+    for k in c:
+        if isinstance(c[k], torch.Tensor):
+            c[k], uc[k] = map(lambda y: y[k][:num_samples].to(device), (c, uc))
+            if c[k].shape[0] < num_samples:
+                c[k] = c[k][[0]]
+            if uc[k].shape[0] < num_samples:
+                uc[k] = uc[k][[0]]
+    return c, uc
+
+
+def fill_latent(cond, length, cond_indices, device):
+    latent = torch.zeros(length, *cond.shape[1:]).to(device)
+    latent[cond_indices] = cond
+    return latent
+
+
+@torch.no_grad()
+def do_sample(
+        images,
+        model,
+        sampler,
+        value_dict,
+        num_frames,
+        ensemble_size: int = 5,
+        force_uc_zero_embeddings: Optional[list] = None,
+        initial_cond_indices: Optional[list] = None,
+        device="cuda"
+):
+    if initial_cond_indices is None:
+        initial_cond_indices = [0]
+
+    force_uc_zero_embeddings = default(force_uc_zero_embeddings, list())
+    precision_scope = autocast
+
+    with torch.no_grad(), precision_scope(device), model.ema_scope("Sampling"):
+        load_model(model.first_stage_model)
+        z = model.encode_first_stage(images)
+        unload_model(model.first_stage_model)
+
+        def denoiser(x, sigma, cond, cond_mask):
+            return model.denoiser(model.model, x, sigma, cond, cond_mask)
+
+        load_model(model.denoiser)
+        load_model(model.model)
+
+        initial_cond_mask = torch.zeros(num_frames).to(device)
+        initial_cond_mask[initial_cond_indices] = 1
+
+        c, uc = get_condition(model, value_dict, num_frames, force_uc_zero_embeddings, device)
+
+        sample_ensemble = list()
+        for _ in range(ensemble_size):
+            noise = torch.randn_like(z)
+            sample = sampler(
+                denoiser,
+                noise,
+                cond=c,
+                uc=uc,
+                cond_frame=z,  # cond_frame will be rescaled when calling the sampler
+                cond_mask=initial_cond_mask
+            )
+            sample[0] = z[0]
+            sample_ensemble.append(sample)
+
+        u = torch.mean(torch.stack(sample_ensemble), 0)
+        diff = torch.zeros_like(sample)
+        for each_sample in sample_ensemble:
+            diff.add_((each_sample - u) ** 2)
+        variance = diff / (ensemble_size - 1)
+        reward = torch.exp(-variance.mean()).cpu()
+
+        unload_model(model.model)
+        unload_model(model.denoiser)
+        return images, reward

vista/sample.py

@@ -0,0 +1,269 @@
+from __future__ import annotations
+
+import argparse
+import json
+import os
+import random
+
+import PIL
+import torch
+from pytorch_lightning import seed_everything
+from torchvision import transforms
+
+from . import sample_utils
+
+VERSION2SPECS = {
+    "vwm": {"config": "configs/inference/vista.yaml", "ckpt": "ckpts/vista.safetensors"}
+}
+
+DATASET2SOURCES = {
+    "NUSCENES": {"data_root": "data/nuscenes", "anno_file": "annos/nuScenes_val.json"},
+    "IMG": {"data_root": "image_folder"},
+}
+
+
+def parse_args(**parser_kwargs):
+    parser = argparse.ArgumentParser(**parser_kwargs)
+    parser.add_argument("--version", type=str, default="vwm", help="model version")
+    parser.add_argument("--dataset", type=str, default="NUSCENES", help="dataset name")
+    parser.add_argument(
+        "--save", type=str, default="outputs", help="directory to save samples"
+    )
+    parser.add_argument(
+        "--action",
+        type=str,
+        default="free",
+        help="action mode for control, such as traj, cmd, steer, goal",
+    )
+    parser.add_argument(
+        "--n_rounds", type=int, default=1, help="number of sampling rounds"
+    )
+    parser.add_argument(
+        "--n_frames", type=int, default=25, help="number of frames for each round"
+    )
+    parser.add_argument(
+        "--n_conds",
+        type=int,
+        default=1,
+        help="number of initial condition frames for the first round",
+    )
+    parser.add_argument(
+        "--seed", type=int, default=23, help="random seed for seed_everything"
+    )
+    parser.add_argument(
+        "--height", type=int, default=576, help="target height of the generated video"
+    )
+    parser.add_argument(
+        "--width", type=int, default=1024, help="target width of the generated video"
+    )
+    parser.add_argument(
+        "--cfg_scale",
+        type=float,
+        default=2.5,
+        help="scale of the classifier-free guidance",
+    )
+    parser.add_argument(
+        "--cond_aug", type=float, default=0.0, help="strength of the noise augmentation"
+    )
+    parser.add_argument(
+        "--n_steps", type=int, default=50, help="number of sampling steps"
+    )
+    parser.add_argument(
+        "--rand_gen",
+        action="store_false",
+        help="whether to generate samples randomly or sequentially",
+    )
+    parser.add_argument(
+        "--low_vram", action="store_true", help="whether to save memory or not"
+    )
+    return parser
+
+
+def get_sample(
+    selected_index=0, dataset_name="NUSCENES", num_frames=25, action_mode="free"
+):
+    dataset_dict = DATASET2SOURCES[dataset_name]
+    action_dict = None
+    if dataset_name == "IMG":
+        image_list = os.listdir(dataset_dict["data_root"])
+        total_length = len(image_list)
+        while selected_index >= total_length:
+            selected_index -= total_length
+        image_file = image_list[selected_index]
+
+        path_list = [os.path.join(dataset_dict["data_root"], image_file)] * num_frames
+    else:
+        with open(dataset_dict["anno_file"]) as anno_json:
+            all_samples = json.load(anno_json)
+        total_length = len(all_samples)
+        while selected_index >= total_length:
+            selected_index -= total_length
+        sample_dict = all_samples[selected_index]
+
+        path_list = list()
+        if dataset_name == "NUSCENES":
+            for index in range(num_frames):
+                image_path = os.path.join(
+                    dataset_dict["data_root"], sample_dict["frames"][index]
+                )
+                assert os.path.exists(image_path), image_path
+                path_list.append(image_path)
+            if action_mode != "free":
+                action_dict = dict()
+                if action_mode == "traj" or action_mode == "trajectory":
+                    action_dict["trajectory"] = torch.tensor(sample_dict["traj"][2:])
+                elif action_mode == "cmd" or action_mode == "command":
+                    action_dict["command"] = torch.tensor(sample_dict["cmd"])
+                elif action_mode == "steer":
+                    # scene might be empty
+                    if sample_dict["speed"]:
+                        action_dict["speed"] = torch.tensor(sample_dict["speed"][1:])
+                    # scene might be empty
+                    if sample_dict["angle"]:
+                        action_dict["angle"] = (
+                            torch.tensor(sample_dict["angle"][1:]) / 780
+                        )
+                elif action_mode == "goal":
+                    # point might be invalid
+                    if (
+                        sample_dict["z"] > 0
+                        and 0 < sample_dict["goal"][0] < 1600
+                        and 0 < sample_dict["goal"][1] < 900
+                    ):
+                        action_dict["goal"] = torch.tensor(
+                            [
+                                sample_dict["goal"][0] / 1600,
+                                sample_dict["goal"][1] / 900,
+                            ]
+                        )
+                else:
+                    raise ValueError(f"Unsupported action mode {action_mode}")
+        else:
+            raise ValueError(f"Invalid dataset {dataset_name}")
+    return path_list, selected_index, total_length, action_dict
+
+
+def load_img(file_name, target_height=320, target_width=576, device="cuda"):
+    if file_name is not None:
+        image = PIL.Image.open(file_name)
+        if not image.mode == "RGB":
+            image = image.convert("RGB")
+    else:
+        raise ValueError(f"Invalid image file {file_name}")
+    ori_w, ori_h = image.size
+    # print(f"Loaded input image of size ({ori_w}, {ori_h})")
+
+    if ori_w / ori_h > target_width / target_height:
+        tmp_w = int(target_width / target_height * ori_h)
+        left = (ori_w - tmp_w) // 2
+        right = (ori_w + tmp_w) // 2
+        image = image.crop((left, 0, right, ori_h))
+    elif ori_w / ori_h < target_width / target_height:
+        tmp_h = int(target_height / target_width * ori_w)
+        top = (ori_h - tmp_h) // 2
+        bottom = (ori_h + tmp_h) // 2
+        image = image.crop((0, top, ori_w, bottom))
+    image = image.resize((target_width, target_height), resample=PIL.Image.LANCZOS)
+    if not image.mode == "RGB":
+        image = image.convert("RGB")
+    image = transforms.Compose(
+        [transforms.ToTensor(), transforms.Lambda(lambda x: x * 2.0 - 1.0)]
+    )(image)
+    return image.to(device)
+
+
+if __name__ == "__main__":
+    parser = parse_args()
+    opt, unknown = parser.parse_known_args()
+
+    sample_utils.set_lowvram_mode(opt.low_vram)
+    version_dict = VERSION2SPECS[opt.version]
+    model = sample_utils.init_model(version_dict)
+    unique_keys = set([x.input_key for x in model.conditioner.embedders])
+
+    sample_index = 0
+    while sample_index >= 0:
+        seed_everything(opt.seed)
+
+        frame_list, sample_index, dataset_length, action_dict = get_sample(
+            sample_index, opt.dataset, opt.n_frames, opt.action
+        )
+
+        img_seq = list()
+        for each_path in frame_list:
+            img = load_img(each_path, opt.height, opt.width)
+            img_seq.append(img)
+        images = torch.stack(img_seq)
+
+        value_dict = sample_utils.init_embedder_options(unique_keys)
+        cond_img = img_seq[0][None]
+        value_dict["cond_frames_without_noise"] = cond_img
+        value_dict["cond_aug"] = opt.cond_aug
+        value_dict["cond_frames"] = cond_img + opt.cond_aug * torch.randn_like(cond_img)
+        if action_dict is not None:
+            for key, value in action_dict.items():
+                value_dict[key] = value
+
+        if opt.n_rounds > 1:
+            guider = "TrianglePredictionGuider"
+        else:
+            guider = "VanillaCFG"
+        sampler = sample_utils.init_sampling(
+            guider=guider,
+            steps=opt.n_steps,
+            cfg_scale=opt.cfg_scale,
+            num_frames=opt.n_frames,
+        )
+
+        uc_keys = [
+            "cond_frames",
+            "cond_frames_without_noise",
+            "command",
+            "trajectory",
+            "speed",
+            "angle",
+            "goal",
+        ]
+
+        out = sample_utils.do_sample(
+            images,
+            model,
+            sampler,
+            value_dict,
+            num_rounds=opt.n_rounds,
+            num_frames=opt.n_frames,
+            force_uc_zero_embeddings=uc_keys,
+            initial_cond_indices=[index for index in range(opt.n_conds)],
+        )
+
+        if isinstance(out, (tuple, list)):
+            samples, samples_z, inputs = out
+            virtual_path = os.path.join(opt.save, "virtual")
+            real_path = os.path.join(opt.save, "real")
+            sample_utils.perform_save_locally(
+                virtual_path, samples, "videos", opt.dataset, sample_index
+            )
+            sample_utils.perform_save_locally(
+                virtual_path, samples, "grids", opt.dataset, sample_index
+            )
+            sample_utils.perform_save_locally(
+                virtual_path, samples, "images", opt.dataset, sample_index
+            )
+            sample_utils.perform_save_locally(
+                real_path, inputs, "videos", opt.dataset, sample_index
+            )
+            sample_utils.perform_save_locally(
+                real_path, inputs, "grids", opt.dataset, sample_index
+            )
+            sample_utils.perform_save_locally(
+                real_path, inputs, "images", opt.dataset, sample_index
+            )
+        else:
+            raise TypeError
+
+        if opt.rand_gen:
+            sample_index += random.randint(1, dataset_length - 1)
+        else:
+            sample_index += 1
+            if dataset_length <= sample_index:
+                sample_index = -1

vista/sample_utils.py

@@ -0,0 +1,442 @@
+from __future__ import annotations
+
+import math
+import os
+import queue
+from typing import Optional, Union
+
+import numpy as np
+import rerun as rr
+import torch
+import torchvision
+from einops import rearrange, repeat
+from omegaconf import ListConfig, OmegaConf
+from PIL import Image
+from safetensors.torch import load_file as load_safetensors
+from torch import autocast
+from tqdm import tqdm
+
+from .vwm.modules.diffusionmodules.sampling import EulerEDMSampler
+from .vwm.util import default, instantiate_from_config
+
+
+def init_model(version_dict, load_ckpt=True):
+    config = OmegaConf.load(version_dict["config"])
+    model = load_model_from_config(config, version_dict["ckpt"] if load_ckpt else None)
+    return model
+
+
+lowvram_mode = True
+
+
+def set_lowvram_mode(mode):
+    global lowvram_mode
+    lowvram_mode = mode
+
+
+def initial_model_load(model):
+    global lowvram_mode
+    if lowvram_mode:
+        model.model.half()
+    else:
+        model.cuda()
+    return model
+
+
+def load_model(model):
+    model.cuda()
+
+
+def unload_model(model):
+    global lowvram_mode
+    print(lowvram_mode)
+    if lowvram_mode:
+        model.cpu()
+        torch.cuda.empty_cache()
+        torch.cuda.synchronize()
+
+
+def load_model_from_config(config, ckpt=None):
+    model = instantiate_from_config(config.model)
+    print(ckpt)
+
+    if ckpt is not None:
+        print(f"Loading model from {ckpt}")
+        if ckpt.endswith("ckpt"):
+            pl_svd = torch.load(ckpt, map_location="cpu")
+            # dict contains:
+            # "epoch", "global_step", "pytorch-lightning_version",
+            # "state_dict", "loops", "callbacks", "optimizer_states", "lr_schedulers"
+            if "global_step" in pl_svd:
+                print(f"Global step: {pl_svd['global_step']}")
+            svd = pl_svd["state_dict"]
+        else:
+            svd = load_safetensors(ckpt)
+
+        missing, unexpected = model.load_state_dict(svd, strict=False)
+        if len(missing) > 0:
+            print(f"Missing keys: {missing}")
+        if len(unexpected) > 0:
+            print(f"Unexpected keys: {unexpected}")
+
+    model = initial_model_load(model)
+    model.eval()
+    return model
+
+
+def init_embedder_options(keys):
+    # hardcoded demo settings, might undergo some changes in the future
+    value_dict = dict()
+    for key in keys:
+        if key in ["fps_id", "fps"]:
+            fps = 10
+            value_dict["fps"] = fps
+            value_dict["fps_id"] = fps - 1
+        elif key == "motion_bucket_id":
+            value_dict["motion_bucket_id"] = 127  # [0, 511]
+    return value_dict
+
+
+def perform_save_locally(save_path, samples, mode, dataset_name, sample_index):
+    assert mode in ["images", "grids", "videos"]
+    merged_path = os.path.join(save_path, mode)
+    os.makedirs(merged_path, exist_ok=True)
+    samples = samples.cpu()
+
+    if mode == "images":
+        frame_count = 0
+        for sample in samples:
+            sample = rearrange(sample.numpy(), "c h w -> h w c")
+            if "real" in save_path:
+                sample = 255.0 * (sample + 1.0) / 2.0
+            else:
+                sample = 255.0 * sample
+            image_save_path = os.path.join(
+                merged_path, f"{dataset_name}_{sample_index:06}_{frame_count:04}.png"
+            )
+            # if os.path.exists(image_save_path):
+            #     return
+            Image.fromarray(sample.astype(np.uint8)).save(image_save_path)
+            frame_count += 1
+    elif mode == "grids":
+        grid = torchvision.utils.make_grid(samples, nrow=int(samples.shape[0] ** 0.5))
+        grid = grid.transpose(0, 1).transpose(1, 2).squeeze(-1).numpy()
+        if "real" in save_path:
+            grid = 255.0 * (grid + 1.0) / 2.0
+        else:
+            grid = 255.0 * grid
+        grid_save_path = os.path.join(
+            merged_path, f"{dataset_name}_{sample_index:06}.png"
+        )
+        # if os.path.exists(grid_save_path):
+        #     return
+        Image.fromarray(grid.astype(np.uint8)).save(grid_save_path)
+    elif mode == "videos":
+        img_seq = rearrange(samples.numpy(), "t c h w -> t h w c")
+        if "real" in save_path:
+            img_seq = 255.0 * (img_seq + 1.0) / 2.0
+        else:
+            img_seq = 255.0 * img_seq
+        video_save_path = os.path.join(
+            merged_path, f"{dataset_name}_{sample_index:06}.mp4"
+        )
+        # if os.path.exists(video_save_path):
+        #     return
+        save_img_seq_to_video(video_save_path, img_seq.astype(np.uint8), 10)
+    else:
+        raise NotImplementedError
+
+
+def init_sampling(
+    sampler="EulerEDMSampler",
+    guider="VanillaCFG",
+    discretization="EDMDiscretization",
+    steps=50,
+    cfg_scale=2.5,
+    num_frames=25,
+):
+    discretization_config = get_discretization(discretization)
+    guider_config = get_guider(guider, cfg_scale, num_frames)
+    sampler = get_sampler(sampler, steps, discretization_config, guider_config)
+    return sampler
+
+
+def get_discretization(discretization):
+    if discretization == "LegacyDDPMDiscretization":
+        discretization_config = {
+            "target": "vista.vwm.modules.diffusionmodules.discretizer.LegacyDDPMDiscretization"
+        }
+    elif discretization == "EDMDiscretization":
+        discretization_config = {
+            "target": "vista.vwm.modules.diffusionmodules.discretizer.EDMDiscretization",
+            "params": {"sigma_min": 0.002, "sigma_max": 700.0, "rho": 7.0},
+        }
+    else:
+        raise NotImplementedError
+    return discretization_config
+
+
+def get_guider(guider="LinearPredictionGuider", cfg_scale=2.5, num_frames=25):
+    if guider == "IdentityGuider":
+        guider_config = {
+            "target": "vista.vwm.modules.diffusionmodules.guiders.IdentityGuider"
+        }
+    elif guider == "VanillaCFG":
+        scale = cfg_scale
+
+        guider_config = {
+            "target": "vista.vwm.modules.diffusionmodules.guiders.VanillaCFG",
+            "params": {"scale": scale},
+        }
+    elif guider == "LinearPredictionGuider":
+        max_scale = cfg_scale
+        min_scale = 1.0
+
+        guider_config = {
+            "target": "vista.vwm.modules.diffusionmodules.guiders.LinearPredictionGuider",
+            "params": {
+                "max_scale": max_scale,
+                "min_scale": min_scale,
+                "num_frames": num_frames,
+            },
+        }
+    elif guider == "TrianglePredictionGuider":
+        max_scale = cfg_scale
+        min_scale = 1.0
+
+        guider_config = {
+            "target": "vista.vwm.modules.diffusionmodules.guiders.TrianglePredictionGuider",
+            "params": {
+                "max_scale": max_scale,
+                "min_scale": min_scale,
+                "num_frames": num_frames,
+            },
+        }
+    else:
+        raise NotImplementedError
+    return guider_config
+
+
+def get_sampler(sampler, steps, discretization_config, guider_config):
+    if sampler == "EulerEDMSampler":
+        s_churn = 0.0
+        s_tmin = 0.0
+        s_tmax = 999.0
+        s_noise = 1.0
+
+        sampler = EulerEDMSampler(
+            num_steps=steps,
+            discretization_config=discretization_config,
+            guider_config=guider_config,
+            s_churn=s_churn,
+            s_tmin=s_tmin,
+            s_tmax=s_tmax,
+            s_noise=s_noise,
+            verbose=False,
+        )
+    else:
+        raise ValueError(f"Unknown sampler {sampler}")
+    return sampler
+
+
+def get_batch(keys, value_dict, N: Union[list, ListConfig], device="cuda"):
+    # hardcoded demo setups, might undergo some changes in the future
+    batch = dict()
+    batch_uc = dict()
+
+    for key in keys:
+        if key in value_dict:
+            if key in ["fps", "fps_id", "motion_bucket_id", "cond_aug"]:
+                batch[key] = repeat(
+                    torch.tensor([value_dict[key]]).to(device), "1 -> b", b=math.prod(N)
+                )
+            elif key in ["command", "trajectory", "speed", "angle", "goal"]:
+                batch[key] = repeat(
+                    value_dict[key][None].to(device), "1 ... -> b ...", b=N[0]
+                )
+            elif key in ["cond_frames", "cond_frames_without_noise"]:
+                batch[key] = repeat(value_dict[key], "1 ... -> b ...", b=N[0])
+            else:
+                # batch[key] = value_dict[key]
+                raise NotImplementedError
+
+    for key in batch.keys():
+        if key not in batch_uc and isinstance(batch[key], torch.Tensor):
+            batch_uc[key] = torch.clone(batch[key])
+    return batch, batch_uc
+
+
+def get_condition(model, value_dict, num_samples, force_uc_zero_embeddings, device):
+    load_model(model.conditioner)
+    batch, batch_uc = get_batch(
+        list(set([x.input_key for x in model.conditioner.embedders])),
+        value_dict,
+        [num_samples],
+    )
+    c, uc = model.conditioner.get_unconditional_conditioning(
+        batch, batch_uc=batch_uc, force_uc_zero_embeddings=force_uc_zero_embeddings
+    )
+    unload_model(model.conditioner)
+
+    for k in c:
+        if isinstance(c[k], torch.Tensor):
+            c[k], uc[k] = map(lambda y: y[k][:num_samples].to(device), (c, uc))
+            if c[k].shape[0] < num_samples:
+                c[k] = c[k][[0]]
+            if uc[k].shape[0] < num_samples:
+                uc[k] = uc[k][[0]]
+    return c, uc
+
+
+def fill_latent(cond, length, cond_indices, device):
+    latent = torch.zeros(length, *cond.shape[1:]).to(device)
+    latent[cond_indices] = cond
+    return latent
+
+
+@torch.no_grad()
+def do_sample(
+    images,
+    model,
+    sampler,
+    value_dict,
+    num_rounds,
+    num_frames,
+    force_uc_zero_embeddings: Optional[list] = None,
+    initial_cond_indices: Optional[list] = None,
+    device="cuda",
+    log_queue: queue.SimpleQueue = None,
+):
+    if initial_cond_indices is None:
+        initial_cond_indices = [0]
+
+    force_uc_zero_embeddings = default(force_uc_zero_embeddings, list())
+    precision_scope = autocast
+
+    with torch.no_grad(), precision_scope(device), model.ema_scope("Sampling"):
+        c, uc = get_condition(
+            model, value_dict, num_frames, force_uc_zero_embeddings, device
+        )
+
+        load_model(model.first_stage_model)
+        z = model.encode_first_stage(images)
+        unload_model(model.first_stage_model)
+
+        samples_z = torch.zeros((num_rounds * (num_frames - 3) + 3, *z.shape[1:])).to(
+            device
+        )
+
+        sampling_progress = tqdm(total=num_rounds, desc="Compute sequences")
+
+        def denoiser(x, sigma, cond, cond_mask):
+            return model.denoiser(model.model, x, sigma, cond, cond_mask)
+
+        load_model(model.denoiser)
+        load_model(model.model)
+
+        initial_cond_mask = torch.zeros(num_frames).to(device)
+        prediction_cond_mask = torch.zeros(num_frames).to(device)
+        initial_cond_mask[initial_cond_indices] = 1
+        prediction_cond_mask[[0, 1, 2]] = 1
+
+        generated_images = []
+
+        noise = torch.randn_like(z)
+        sample = sampler(
+            denoiser,
+            noise,
+            cond=c,
+            uc=uc,
+            cond_frame=z,  # cond_frame will be rescaled when calling the sampler
+            cond_mask=initial_cond_mask,
+            num_sequence=0,
+            log_queue=log_queue,
+        )
+        sampling_progress.update(1)
+        sample[0] = z[0]
+        samples_z[:num_frames] = sample
+
+        generated_images.append(decode_samples(sample[:num_frames], model))
+
+        for i, generated_image in enumerate(generated_images[-1]):
+            log_queue.put(
+                (
+                    "generated_image",
+                    rr.Image(generated_image.cpu().permute(1, 2, 0)),
+                    [
+                        ("frame_id", i),
+                        ("diffusion", 0),
+                        (
+                            "combined",
+                            1 + 2 * 0 + (i * 1.0 / len(generated_images[-1])),
+                        ),
+                    ],
+                )
+            )
+
+        for n in range(num_rounds - 1):
+            load_model(model.first_stage_model)
+            samples_x_for_guidance = model.decode_first_stage(sample[-14:])
+            unload_model(model.first_stage_model)
+            value_dict["cond_frames_without_noise"] = samples_x_for_guidance[[-3]]
+            value_dict["cond_frames"] = sample[[-3]] / model.scale_factor
+
+            for embedder in model.conditioner.embedders:
+                if hasattr(embedder, "skip_encode"):
+                    embedder.skip_encode = True
+            c, uc = get_condition(
+                model, value_dict, num_frames, force_uc_zero_embeddings, device
+            )
+            for embedder in model.conditioner.embedders:
+                if hasattr(embedder, "skip_encode"):
+                    embedder.skip_encode = False
+
+            filled_latent = fill_latent(sample[-3:], num_frames, [0, 1, 2], device)
+
+            noise = torch.randn_like(filled_latent)
+            sample = sampler(
+                denoiser,
+                noise,
+                cond=c,
+                uc=uc,
+                cond_frame=filled_latent,  # cond_frame will be rescaled when calling the sampler
+                cond_mask=prediction_cond_mask,
+                num_sequence=n + 1,
+                log_queue=log_queue,
+            )
+            sampling_progress.update(1)
+            first_frame_id = (n + 1) * (num_frames - 3) + 3
+            last_frame_id = (n + 1) * (num_frames - 3) + num_frames
+            samples_z[first_frame_id:last_frame_id] = sample[3:]
+
+            generated_images.append(decode_samples(sample[3:], model))
+
+            for i, generated_image in enumerate(generated_images[-1]):
+                log_queue.put(
+                    (
+                        "generated_image",
+                        rr.Image(generated_image.cpu().permute(1, 2, 0)),
+                        [
+                            ("frame_id", first_frame_id + i),
+                            ("diffusion", 0),
+                            (
+                                "combined",
+                                1 + 2 * (n + 1) + (i * 1.0 / len(generated_images[-1])),
+                            ),
+                        ],
+                    )
+                )
+
+        unload_model(model.model)
+        unload_model(model.denoiser)
+
+        generated_images = torch.concat(generated_images, dim=0)
+        return generated_images, samples_z, images
+
+
+def decode_samples(samples, model):
+    load_model(model.first_stage_model)
+    samples_x = model.decode_first_stage(samples)
+    unload_model(model.first_stage_model)
+    samples = torch.clamp((samples_x + 1.0) / 2.0, min=0.0, max=1.0)
+    return samples

vista/train.py

@@ -0,0 +1,924 @@
+from __future__ import annotations
+
+import argparse
+import datetime
+import glob
+import inspect
+import os
+import sys
+from inspect import Parameter
+
+import imageio
+import numpy as np
+import pytorch_lightning as pl
+import torch
+import torchvision
+from einops import rearrange
+from matplotlib import pyplot as plt
+from natsort import natsorted
+from omegaconf import OmegaConf
+from packaging import version
+from PIL import Image
+from pytorch_lightning import seed_everything
+from pytorch_lightning.callbacks import Callback
+from pytorch_lightning.trainer import Trainer
+from pytorch_lightning.utilities import rank_zero_only
+from safetensors.torch import load_file as load_safetensors
+
+from .vwm.util import instantiate_from_config, isheatmap
+
+MULTINODE_HACKS = True
+
+
+def default_trainer_args():
+    argspec = dict(inspect.signature(Trainer.__init__).parameters)
+    argspec.pop("self")
+    default_args = {
+        param: argspec[param].default
+        for param in argspec
+        if argspec[param] != Parameter.empty
+    }
+    return default_args
+
+
+def get_parser(**parser_kwargs):
+    def str2bool(v):
+        if isinstance(v, bool):
+            return v
+        if v.lower() in ("yes", "true", "t", "y", "1"):
+            return True
+        elif v.lower() in ("no", "false", "f", "n", "0"):
+            return False
+        else:
+            raise argparse.ArgumentTypeError("Boolean value expected")
+
+    parser = argparse.ArgumentParser(**parser_kwargs)
+    parser.add_argument(
+        "-n",
+        "--name",
+        type=str,
+        const=True,
+        default="",
+        nargs="?",
+        help="postfix for logdir"
+    )
+    parser.add_argument(
+        "--no_date",
+        type=str2bool,
+        nargs="?",
+        const=True,
+        default=False,
+        help="if True, skip date generation for logdir and only use naming via opt.base or opt.name (+ opt.postfix, optionally)"
+    )
+    parser.add_argument(
+        "-r",
+        "--resume",
+        type=str,
+        const=True,
+        default="",
+        nargs="?",
+        help="resume from logdir or checkpoint in logdir"
+    )
+    parser.add_argument(
+        "-b",
+        "--base",
+        nargs="*",
+        metavar="base_config.yaml",
+        help="paths to base configs. "
+             "Loaded from left-to-right. "
+             "Parameters can be overwritten or added with command-line options of the form `--key value`",
+        default=list()
+    )
+    parser.add_argument(
+        "-t",
+        "--train",
+        type=str2bool,
+        const=True,
+        default=True,
+        nargs="?",
+        help="train"
+    )
+    parser.add_argument(
+        "--no_test",
+        type=str2bool,
+        const=True,
+        default=True,
+        nargs="?",
+        help="disable test"
+    )
+    parser.add_argument(
+        "-p",
+        "--project",
+        help="name of new or path to existing project"
+    )
+    parser.add_argument(
+        "-d",
+        "--debug",
+        type=str2bool,
+        nargs="?",
+        const=True,
+        default=False,
+        help="enable post-mortem debugging"
+    )
+    parser.add_argument(
+        "-s",
+        "--seed",
+        type=int,
+        default=23,
+        help="seed for seed_everything"
+    )
+    parser.add_argument(
+        "-f",
+        "--postfix",
+        type=str,
+        default="",
+        help="post-postfix for default name"
+    )
+    parser.add_argument(
+        "-l",
+        "--logdir",
+        type=str,
+        default="logs",
+        help="directory for logging data"
+    )
+    parser.add_argument(
+        "--scale_lr",
+        type=str2bool,
+        nargs="?",
+        const=True,
+        default=False,
+        help="scale base-lr by ngpu * batch_size * n_accumulate"
+    )
+    parser.add_argument(
+        "--legacy_naming",
+        type=str2bool,
+        nargs="?",
+        const=True,
+        default=False,
+        help="name run based on config file name if true, else by whole path"
+    )
+    parser.add_argument(
+        "--enable_tf32",
+        type=str2bool,
+        nargs="?",
+        const=True,
+        default=False,
+        help="enables the TensorFloat32 format both for matmuls and cuDNN for pytorch 1.12"
+    )
+    parser.add_argument(
+        "--no_base_name",
+        type=str2bool,
+        nargs="?",
+        const=True,
+        default=False,
+        help="no config name"
+    )
+    if version.parse(pl.__version__) >= version.parse("2.0.0"):
+        parser.add_argument(
+            "--resume_from_checkpoint",
+            type=str,
+            default=None,
+            help="single checkpoint file to resume from"
+        )
+    parser.add_argument(
+        "--n_devices",
+        type=int,
+        default=8,
+        help="number of gpus in training"
+    )
+    parser.add_argument(
+        "--finetune",
+        type=str,
+        default="ckpts/pytorch_model.bin",
+        help="path to checkpoint to finetune from"
+    )
+    default_args = default_trainer_args()
+    for key in default_args:
+        parser.add_argument("--" + key, default=default_args[key])
+    return parser
+
+
+def get_checkpoint_name(logdir):
+    ckpt = os.path.join(logdir, "checkpoints", "last**.ckpt")
+    ckpt = natsorted(glob.glob(ckpt))
+    print("Available last checkpoints:", ckpt)
+    if len(ckpt) > 1:
+        print("Got most recent checkpoint")
+        ckpt = sorted(ckpt, key=lambda x: os.path.getmtime(x))[-1]
+        print(f"Most recent ckpt is {ckpt}")
+        with open(os.path.join(logdir, "most_recent_ckpt.txt"), "w") as f:
+            f.write(ckpt + "\n")
+        try:
+            version = int(ckpt.split("/")[-1].split("-v")[-1].split(".")[0])
+        except Exception as e:
+            # version confusion but not bad
+            print(e)
+            version = 1
+        # version = last_version + 1
+    else:
+        # in this case, we only have one "last.ckpt"
+        ckpt = ckpt[0]
+        version = 1
+    melk_ckpt_name = f"last-v{version}.ckpt"
+    print(f"Current melk ckpt name: {melk_ckpt_name}")
+    return ckpt, melk_ckpt_name
+
+
+def save_img_seq_to_video(out_path, img_seq, fps):
+    # img_seq: np array
+    writer = imageio.get_writer(out_path, fps=fps)
+    for img in img_seq:
+        writer.append_data(img)
+    writer.close()
+
+
+class SetupCallback(Callback):
+    def __init__(
+            self,
+            resume,
+            now,
+            logdir,
+            ckptdir,
+            cfgdir,
+            config,
+            lightning_config,
+            debug,
+            ckpt_name=None
+    ):
+        super().__init__()
+        self.resume = resume
+        self.now = now
+        self.logdir = logdir
+        self.ckptdir = ckptdir
+        self.cfgdir = cfgdir
+        self.config = config
+        self.lightning_config = lightning_config
+        self.debug = debug
+        self.ckpt_name = ckpt_name
+
+    def on_exception(self, trainer: pl.Trainer, pl_module, exception):
+        if not self.debug and trainer.global_rank == 0:
+            # print("Summoning checkpoint")
+            # if self.ckpt_name is None:
+            #     ckpt_path = os.path.join(self.ckptdir, "last.ckpt")
+            # else:
+            #     ckpt_path = os.path.join(self.ckptdir, self.ckpt_name)
+            # trainer.save_checkpoint(ckpt_path)
+            print("Exiting")
+
+    def on_fit_start(self, trainer, pl_module):
+        if trainer.global_rank == 0:
+            # create logdirs and save configs
+            os.makedirs(self.logdir, exist_ok=True)
+            os.makedirs(self.ckptdir, exist_ok=True)
+            os.makedirs(self.cfgdir, exist_ok=True)
+
+            if "callbacks" in self.lightning_config:
+                if "metrics_over_trainsteps_checkpoint" in self.lightning_config["callbacks"]:
+                    os.makedirs(
+                        os.path.join(self.ckptdir, "trainstep_checkpoints"),
+                        exist_ok=True
+                    )
+            print("Project config")
+            print(OmegaConf.to_yaml(self.config))
+            if MULTINODE_HACKS:
+                import time
+
+                time.sleep(5)
+            OmegaConf.save(
+                self.config,
+                os.path.join(self.cfgdir, f"{self.now}-project.yaml")
+            )
+
+            print("Lightning config")
+            print(OmegaConf.to_yaml(self.lightning_config))
+            OmegaConf.save(
+                OmegaConf.create({"lightning": self.lightning_config}),
+                os.path.join(self.cfgdir, f"{self.now}-lightning.yaml")
+            )
+        else:
+            # ModelCheckpoint callback created log directory, remove it
+            if not MULTINODE_HACKS and not self.resume and os.path.exists(self.logdir):
+                dst, name = os.path.split(self.logdir)
+                dst = os.path.join(dst, "child_runs", name)
+                os.makedirs(os.path.split(dst)[0], exist_ok=True)
+                try:
+                    os.rename(self.logdir, dst)
+                except FileNotFoundError:
+                    pass
+
+
+class ImageLogger(Callback):
+    def __init__(
+            self,
+            batch_frequency,
+            clamp=True,
+            increase_log_steps=True,
+            rescale=True,
+            disabled=False,
+            log_on_batch_idx=False,
+            log_first_step=False,
+            log_images_kwargs=None,
+            log_before_first_step=False,
+            enable_autocast=True,
+            num_frames=25
+    ):
+        super().__init__()
+        self.enable_autocast = enable_autocast
+        self.rescale = rescale
+        self.batch_freq = batch_frequency
+        self.log_steps = [2 ** n for n in range(int(np.log2(self.batch_freq)) + 1)]
+        if not increase_log_steps:
+            self.log_steps = [self.batch_freq]
+        self.clamp = clamp
+        self.disabled = disabled
+        self.log_on_batch_idx = log_on_batch_idx
+        self.log_images_kwargs = log_images_kwargs if log_images_kwargs else dict()
+        self.log_first_step = log_first_step
+        self.log_before_first_step = log_before_first_step
+        self.num_frames = num_frames
+
+    @rank_zero_only
+    def log_local(
+            self,
+            save_dir,
+            split,
+            images,
+            global_step,
+            current_epoch,
+            batch_idx
+    ):
+        root = os.path.join(save_dir, "images", split)
+        for log_type in images:
+            if isheatmap(images[log_type]):
+                _fig, ax = plt.subplots()
+                ax = ax.matshow(
+                    images[log_type].cpu().numpy(), cmap="hot", interpolation="lanczos"
+                )
+                plt.colorbar(ax)
+                plt.axis("off")
+
+                filename = f"{log_type}_epoch{current_epoch:03}_batch{batch_idx:06}_step{global_step:06}.png"
+                os.makedirs(root, exist_ok=True)
+                path = os.path.join(root, log_type, filename)
+                plt.savefig(path)
+                plt.close()
+            elif "mp4" in log_type:
+                dir_path = os.path.join(root, log_type)
+                os.makedirs(dir_path, exist_ok=True)
+                img_seq = images[log_type]
+                if self.rescale:
+                    img_seq = (img_seq + 1.0) / 2.0
+                img_seq = rearrange(img_seq, "(b t) c h w -> b t h w c", t=self.num_frames)
+                B, _T = img_seq.shape[:2]
+                for b_i in range(B):
+                    cur_img_seq = img_seq[b_i].numpy()  # [t h w c]
+                    cur_img_seq = (cur_img_seq * 255).astype(np.uint8)  # [t h w c]
+                    filename = f"{log_type}_epoch{current_epoch:02}_batch{batch_idx:04}_step{global_step:06}.mp4"
+                    save_img_seq_to_video(os.path.join(root, log_type, filename), cur_img_seq, fps=10)
+            else:
+                grid = torchvision.utils.make_grid(images[log_type], nrow=int(images[log_type].shape[0] ** 0.5))
+                if self.rescale:
+                    grid = (grid + 1.0) / 2.0  # -1,1 -> 0,1; c,h,w
+                grid = grid.transpose(0, 1).transpose(1, 2).squeeze(-1)
+                grid = grid.numpy()
+                grid = (grid * 255).astype(np.uint8)
+                filename = f"{log_type}_epoch{current_epoch:02}_batch{batch_idx:04}_step{global_step:06}.png"
+                dir_path = os.path.join(root, log_type)
+                os.makedirs(dir_path, exist_ok=True)
+                path = os.path.join(dir_path, filename)
+                img = Image.fromarray(grid)
+                img.save(path)
+
+    @rank_zero_only
+    def log_img(self, pl_module, batch, batch_idx, split="train"):
+        check_idx = batch_idx if self.log_on_batch_idx else pl_module.global_step
+        if (
+                self.check_frequency(check_idx)
+                and hasattr(pl_module, "log_images")  # batch_idx % self.batch_freq == 0
+                and callable(pl_module.log_images)
+        ) or split == "test":
+            is_train = pl_module.training
+            if is_train:
+                pl_module.eval()
+
+            gpu_autocast_kwargs = {
+                "enabled": self.enable_autocast,  # torch.is_autocast_enabled(),
+                "dtype": torch.get_autocast_gpu_dtype(),
+                "cache_enabled": torch.is_autocast_cache_enabled()
+            }
+
+            with torch.no_grad(), torch.cuda.amp.autocast(**gpu_autocast_kwargs):
+                images = pl_module.log_images(batch, split=split, **self.log_images_kwargs)
+
+            for log_type in images:
+                if isinstance(images[log_type], torch.Tensor):
+                    images[log_type] = images[log_type].detach().float().cpu()
+                    if self.clamp and not isheatmap(images[log_type]):
+                        images[log_type] = torch.clamp(images[log_type], -1.0, 1.0)
+
+            self.log_local(
+                pl_module.logger.save_dir,
+                split,
+                images,
+                pl_module.global_step,
+                pl_module.current_epoch,
+                batch_idx
+            )
+
+            if is_train:
+                pl_module.train()
+
+    def check_frequency(self, check_idx):
+        if (check_idx % self.batch_freq == 0 or check_idx in self.log_steps) and (check_idx > 0 or self.log_first_step):
+            try:
+                self.log_steps.pop(0)
+            except IndexError as e:
+                print(e)
+                pass
+            return True
+        else:
+            return False
+
+    @rank_zero_only
+    def on_train_batch_end(self, trainer, pl_module, outputs, batch, batch_idx):
+        if not self.disabled and (pl_module.global_step > 0 or self.log_first_step):
+            self.log_img(pl_module, batch, batch_idx, split="train")
+
+    @rank_zero_only
+    def on_train_batch_start(self, trainer, pl_module, batch, batch_idx):
+        if self.log_before_first_step and pl_module.global_step == 0:
+            print(f"{self.__class__.__name__}: logging before training")
+            self.log_img(pl_module, batch, batch_idx, split="train")
+
+    @rank_zero_only
+    def on_validation_batch_end(self, trainer, pl_module, outputs, batch, batch_idx, *args, **kwargs):
+        if not self.disabled and pl_module.global_step > 0:
+            self.log_img(pl_module, batch, batch_idx, split="val")
+
+    @rank_zero_only
+    def on_test_batch_end(self, trainer, pl_module, outputs, batch, batch_idx):
+        self.log_img(pl_module, batch, batch_idx, split="test")
+
+
+if __name__ == "__main__":
+    # custom parser to specify config files, train, test and debug mode, postfix, resume
+    # `--key value` arguments are interpreted as arguments to the trainer
+    # `nested.key=value` arguments are interpreted as config parameters
+    # configs are merged from left-to-right followed by command line parameters
+
+    # model:
+    #   base_learning_rate: float
+    #   target: path to lightning module
+    #   params:
+    #       key: value
+    # data:
+    #   target: train.DataModuleFromConfig
+    #   params:
+    #      batch_size: int
+    #      wrap: bool
+    #      train:
+    #          target: path to train dataset
+    #          params:
+    #              key: value
+    #      validation:
+    #          target: path to validation dataset
+    #          params:
+    #              key: value
+    #      test:
+    #          target: path to test dataset
+    #          params:
+    #              key: value
+    # lightning: (optional, has sane defaults and can be specified on cmd line)
+    #   trainer:
+    #       additional arguments to trainer
+    #   logger:
+    #       logger to instantiate
+    #   modelcheckpoint:
+    #       modelcheckpoint to instantiate
+    #   callbacks:
+    #       callback1:
+    #           target: importpath
+    #           params:
+    #               key: value
+
+    now = datetime.datetime.now().strftime("%Y-%m-%dT%H-%M-%S")
+
+    # add cwd for convenience and to make classes in this file available when
+    # running as `python train.py`
+    # (in particular `train.DataModuleFromConfig`)
+    sys.path.append(os.getcwd())
+
+    parser = get_parser()
+    opt, unknown = parser.parse_known_args()
+
+    if opt.name and opt.resume:
+        raise ValueError(
+            "-n/--name and -r/--resume cannot be specified both. "
+            "If you want to resume training in a new log folder, "
+            "use -n/--name in combination with --resume_from_checkpoint"
+        )
+    melk_ckpt_name = None
+    name = None
+    if opt.resume:
+        if not os.path.exists(opt.resume):
+            raise ValueError(f"Cannot find {opt.resume}")
+        if os.path.isfile(opt.resume):
+            paths = opt.resume.split("/")
+            # idx = len(paths)-paths[::-1].index("logs")+1
+            # logdir = "/".join(paths[:idx])
+            logdir = "/".join(paths[:-2])
+            ckpt = opt.resume
+            _, melk_ckpt_name = get_checkpoint_name(logdir)
+        else:
+            assert os.path.isdir(opt.resume), opt.resume
+            logdir = opt.resume.rstrip("/")
+            ckpt, melk_ckpt_name = get_checkpoint_name(logdir)
+
+        print("#" * 100)
+        print(f"Resuming from checkpoint `{ckpt}`")
+        print("#" * 100)
+
+        opt.resume_from_checkpoint = ckpt
+        base_configs = sorted(glob.glob(os.path.join(logdir, "configs/*.yaml")))
+        opt.base = base_configs + opt.base
+        _tmp = logdir.split("/")
+        nowname = _tmp[-1]
+    else:
+        if opt.name:
+            name = "_" + opt.name
+        elif opt.base:
+            if opt.no_base_name:
+                name = ""
+            else:
+                if opt.legacy_naming:
+                    cfg_fname = os.path.split(opt.base[0])[-1]
+                    cfg_name = os.path.splitext(cfg_fname)[0]
+                else:
+                    assert "configs" in os.path.split(opt.base[0])[0], os.path.split(
+                        opt.base[0]
+                    )[0]
+                    cfg_path = os.path.split(opt.base[0])[0].split(os.sep)[
+                               os.path.split(opt.base[0])[0].split(os.sep).index("configs")
+                               + 1:
+                               ]  # cut away the first one (we assert all configs are in "configs")
+                    cfg_name = os.path.splitext(os.path.split(opt.base[0])[-1])[0]
+                    cfg_name = "-".join(cfg_path) + f"-{cfg_name}"
+                name = "_" + cfg_name
+        else:
+            name = ""
+        if opt.no_date:
+            nowname = name + opt.postfix
+            if nowname.startswith("_"):
+                nowname = nowname[1:]
+        else:
+            nowname = now + name + opt.postfix
+        logdir = os.path.join(opt.logdir, nowname)
+
+    ckptdir = os.path.join(logdir, "checkpoints")
+    cfgdir = os.path.join(logdir, "configs")
+    seed_everything(opt.seed, workers=True)
+
+    # move before model init, in case a torch.compile(...) is called somewhere
+    if opt.enable_tf32:
+        # pt_version = version.parse(torch.__version__)
+        torch.backends.cuda.matmul.allow_tf32 = True
+        torch.backends.cudnn.allow_tf32 = True
+        print(f"Enabling TF32 for PyTorch {torch.__version__}")
+    else:
+        print(f"Using default TF32 settings for PyTorch {torch.__version__}:")
+        print(f"torch.backends.cuda.matmul.allow_tf32={torch.backends.cuda.matmul.allow_tf32}")
+        print(f"torch.backends.cudnn.allow_tf32={torch.backends.cudnn.allow_tf32}")
+
+    try:
+        # init and save configs
+        configs = [OmegaConf.load(cfg) for cfg in opt.base]
+        cli = OmegaConf.from_dotlist(unknown)
+        config = OmegaConf.merge(*configs, cli)
+        lightning_config = config.pop("lightning", OmegaConf.create())
+        # merge trainer cli with config
+        trainer_config = lightning_config.get("trainer", OmegaConf.create())
+
+        # default to gpu
+        trainer_config["accelerator"] = "gpu"
+
+        standard_args = default_trainer_args()
+        for k in standard_args:
+            if getattr(opt, k) != standard_args[k]:
+                trainer_config[k] = getattr(opt, k)
+
+        n_devices = getattr(opt, "n_devices", None)
+        if n_devices is not None:
+            assert isinstance(n_devices, int) and n_devices > 0
+            devices = [str(i) for i in range(n_devices)]
+            trainer_config["devices"] = ",".join(devices) + ","
+        else:
+            assert "devices" in trainer_config, "Must specify either n_devices or devices"
+
+        ckpt_resume_path = opt.resume_from_checkpoint
+
+        if "devices" not in trainer_config and trainer_config["accelerator"] != "gpu":
+            del trainer_config["accelerator"]
+            cpu = True
+        else:
+            gpuinfo = trainer_config["devices"]
+            print(f"Running on GPUs {gpuinfo}")
+            cpu = False
+        trainer_opt = argparse.Namespace(**trainer_config)
+        lightning_config.trainer = trainer_config
+
+        # model
+        model = instantiate_from_config(config.model)
+
+        # use pretrained model
+        if not opt.resume or opt.finetune:
+            if not opt.finetune or not os.path.exists(opt.finetune):
+                default_ckpt = "ckpts/svd_xt.safetensors"
+                print(f"Loading pretrained model from {default_ckpt}")
+                svd = load_safetensors(default_ckpt)
+                for k in list(svd.keys()):
+                    if "time_embed" in k:  # duplicate a new timestep embedding from the pretrained weights
+                        svd[k.replace("time_embed", "cond_time_stack_embed")] = svd[k]
+            else:
+                ckpt_path = opt.finetune
+                print(f"Loading pretrained model from {ckpt_path}")
+                if ckpt_path.endswith("ckpt"):
+                    svd = torch.load(ckpt_path, map_location="cpu")["state_dict"]
+                elif ckpt_path.endswith("bin"):  # for deepspeed merged checkpoints
+                    svd = torch.load(ckpt_path, map_location="cpu")
+                    for k in list(svd.keys()):  # remove the prefix
+                        if "_forward_module" in k:
+                            svd[k.replace("_forward_module.", "")] = svd[k]
+                        del svd[k]
+                elif ckpt_path.endswith("safetensors"):
+                    svd = load_safetensors(ckpt_path)
+                else:
+                    raise NotImplementedError
+            missing, unexpected = model.load_state_dict(svd, strict=False)
+
+            # avoid empty weights when resuming from EMA weights
+            for miss_k in missing:
+                ema_name = miss_k.replace(".", "").replace("modeldiffusion_model", "model_ema.diffusion_model")
+                svd[miss_k] = svd[ema_name]
+                print("Fill", miss_k, "with", ema_name)
+            missing, unexpected = model.load_state_dict(svd, strict=False)
+
+            if len(missing) > 0:
+                if not opt.finetune or not os.path.exists(opt.finetune):
+                    model.reinit_ema()
+                missing = [model_key for model_key in missing if "model_ema" not in model_key]
+                # print(f"Missing keys: {missing}")
+            print(f"Missing keys: {missing}")
+            # if len(unexpected) > 0:
+            #     print(f"Unexpected keys: {unexpected}")
+            print(f"Unexpected keys: {unexpected}")
+
+        # trainer and callbacks
+        trainer_kwargs = dict()
+
+        # default logger configs
+        default_logger_cfgs = {
+            "csv": {
+                "target": "pytorch_lightning.loggers.CSVLogger",
+                "params": {
+                    "name": "testtube",  # hack for sbord fanatics
+                    "save_dir": logdir
+                }
+            }
+        }
+        default_logger_cfg = default_logger_cfgs["csv"]
+        if "logger" in lightning_config:
+            logger_cfg = lightning_config.logger
+        else:
+            logger_cfg = OmegaConf.create()
+        logger_cfg = OmegaConf.merge(default_logger_cfg, logger_cfg)
+        trainer_kwargs["logger"] = instantiate_from_config(logger_cfg)
+
+        # use TrainResult/EvalResult(checkpoint_on=metric) to specify which metric is used to determine best models
+        default_modelckpt_cfg = {
+            "target": "pytorch_lightning.callbacks.ModelCheckpoint",
+            "params": {
+                "dirpath": ckptdir,
+                "filename": "{epoch:02}",
+                "verbose": True,
+                "save_last": True,
+                "save_top_k": -1
+            }
+        }
+        # if hasattr(model, "monitor"):
+        #     print(f"Monitoring {model.monitor} as checkpoint metric")
+        #     default_modelckpt_cfg["params"]["monitor"] = model.monitor
+        #     default_modelckpt_cfg["params"]["save_top_k"] = 3
+
+        if "modelcheckpoint" in lightning_config:
+            modelckpt_cfg = lightning_config.modelcheckpoint
+        else:
+            modelckpt_cfg = OmegaConf.create()
+        modelckpt_cfg = OmegaConf.merge(default_modelckpt_cfg, modelckpt_cfg)
+        print(f"Merged modelckpt-cfg: \n{modelckpt_cfg}")
+
+        # default to ddp if not further specified
+        default_strategy_config = {"target": "pytorch_lightning.strategies.DDPStrategy"}
+
+        if "strategy" in lightning_config:
+            strategy_cfg = lightning_config.strategy
+        else:
+            strategy_cfg = OmegaConf.create()
+            default_strategy_config["params"] = {
+                "find_unused_parameters": True
+            }
+        strategy_cfg = OmegaConf.merge(default_strategy_config, strategy_cfg)
+        print(
+            f"strategy config: \n ++++++++++++++ \n {strategy_cfg} \n ++++++++++++++ "
+        )
+        trainer_kwargs["strategy"] = instantiate_from_config(strategy_cfg)
+
+        # add callback which sets up log directory
+        default_callbacks_cfg = {
+            "setup_callback": {
+                "target": "train.SetupCallback",
+                "params": {
+                    "resume": opt.resume,
+                    "now": now,
+                    "logdir": logdir,
+                    "ckptdir": ckptdir,
+                    "cfgdir": cfgdir,
+                    "config": config,
+                    "lightning_config": lightning_config,
+                    "debug": opt.debug,
+                    "ckpt_name": melk_ckpt_name
+                }
+            },
+            "image_logger": {
+                "target": "train.ImageLogger",
+                "params": {
+                    "batch_frequency": 1000,
+                    "clamp": True
+                }
+            },
+            "learning_rate_logger": {
+                "target": "pytorch_lightning.callbacks.LearningRateMonitor",
+                "params": {
+                    "logging_interval": "step"
+                }
+            }
+        }
+        if version.parse(pl.__version__) >= version.parse("1.4.0"):
+            default_callbacks_cfg.update({"checkpoint_callback": modelckpt_cfg})
+
+        if "callbacks" in lightning_config:
+            callbacks_cfg = lightning_config.callbacks
+        else:
+            callbacks_cfg = OmegaConf.create()
+
+        # if "metrics_over_trainsteps_checkpoint" in callbacks_cfg:
+        #     print(
+        #         "WARNING: saving checkpoints every n train steps without deleting, this might require some free space"
+        #     )
+        #     default_metrics_over_trainsteps_ckpt_dict = {
+        #         "metrics_over_trainsteps_checkpoint": {
+        #             "target": "pytorch_lightning.callbacks.ModelCheckpoint",
+        #             "params": {
+        #                 "dirpath": os.path.join(ckptdir, "trainstep_checkpoints"),
+        #                 "filename": "{epoch:06}-{step:09}",
+        #                 "verbose": True,
+        #                 "save_top_k": -1,
+        #                 "every_n_train_steps": 10000,
+        #                 "save_weights_only": True
+        #             }
+        #         }
+        #     }
+        #     default_callbacks_cfg.update(default_metrics_over_trainsteps_ckpt_dict)
+
+        callbacks_cfg = OmegaConf.merge(default_callbacks_cfg, callbacks_cfg)
+        if "ignore_keys_callback" in callbacks_cfg and ckpt_resume_path is not None:
+            callbacks_cfg.ignore_keys_callback.params["ckpt_path"] = ckpt_resume_path
+        elif "ignore_keys_callback" in callbacks_cfg:
+            del callbacks_cfg["ignore_keys_callback"]
+
+        trainer_kwargs["callbacks"] = [
+            instantiate_from_config(callbacks_cfg[k]) for k in callbacks_cfg
+        ]
+        if "plugins" not in trainer_kwargs:
+            trainer_kwargs["plugins"] = list()
+
+        # cmd line trainer args (which are in trainer_opt) have always priority over
+        # config-trainer-args (which are in trainer_kwargs)
+        trainer_opt = vars(trainer_opt)
+        trainer_kwargs = {
+            key: val for key, val in trainer_kwargs.items() if key not in trainer_opt
+        }
+        trainer = Trainer(**trainer_opt, **trainer_kwargs)
+
+        trainer.logdir = logdir
+
+        # data
+        data = instantiate_from_config(config.data)
+        # calling these ourselves should not be necessary, but it is
+        # lightning still takes care of proper multiprocessing though
+        data.prepare_data()
+        # data.setup()
+        print("#### Data #####")
+        try:
+            for k in data.datasets:
+                print(
+                    f"{k}, {data.datasets[k].__class__.__name__}, {len(data.datasets[k])}"
+                )
+        except:
+            print("Datasets not yet initialized")
+
+        # configure learning rate
+        if "batch_size" in config.data.params:
+            bs, base_lr = config.data.params.batch_size, config.model.base_learning_rate
+        else:
+            bs, base_lr = (
+                config.data.params.train.loader.batch_size,
+                config.model.base_learning_rate
+            )
+        if cpu:
+            ngpu = 1
+        else:
+            ngpu = len(lightning_config.trainer.devices.strip(",").split(","))
+        if "accumulate_grad_batches" in lightning_config.trainer:
+            accumulate_grad_batches = lightning_config.trainer.accumulate_grad_batches
+        else:
+            accumulate_grad_batches = 1
+        print(f"accumulate_grad_batches = {accumulate_grad_batches}")
+        lightning_config.trainer.accumulate_grad_batches = accumulate_grad_batches
+        if opt.scale_lr:
+            model.learning_rate = accumulate_grad_batches * ngpu * bs * base_lr
+            print(
+                "Setting learning rate to "
+                f"{model.learning_rate:.2e} = {accumulate_grad_batches} (accumulate_grad_batches) * {ngpu} (num_gpus) * {bs} (batch_size) * {base_lr:.2e} (base_lr)"
+            )
+        else:
+            model.learning_rate = base_lr
+            print("++++ NOT USING LR SCALING ++++")
+            print(f"Setting learning rate to {model.learning_rate:.2e}")
+
+
+        # allow checkpointing via USR1
+        def melk(*args, **kwargs):
+            # run all checkpoint hooks
+            if trainer.global_rank == 0:
+                # print("Summoning checkpoint")
+                # if melk_ckpt_name is None:
+                #     ckpt_path = os.path.join(ckptdir, "last.ckpt")
+                # else:
+                #     ckpt_path = os.path.join(ckptdir, melk_ckpt_name)
+                # trainer.save_checkpoint(ckpt_path)
+                print("Exiting")
+
+
+        def divein(*args, **kwargs):
+            if trainer.global_rank == 0:
+                import pudb
+                pudb.set_trace()
+
+
+        import signal
+
+        signal.signal(signal.SIGUSR1, melk)
+        signal.signal(signal.SIGUSR2, divein)
+
+        # run
+        if opt.train:
+            trainer.fit(model, data, ckpt_path=ckpt_resume_path)
+        if not opt.no_test and not trainer.interrupted:
+            trainer.test(model, data)
+    except RuntimeError as error:
+        # if MULTINODE_HACKS:
+        #     import datetime
+        #     import os
+        #     import socket
+        #
+        #     import requests
+        #
+        #     device = os.environ.get("CUDA_VISIBLE_DEVICES", "?")
+        #     hostname = socket.gethostname()
+        #     ts = datetime.datetime.utcnow().strftime("%Y-%m-%d %H:%M:%S")
+        #     resp = requests.get("http://169.254.169.254/latest/meta-data/instance-id")
+        #     print(
+        #         f"ERROR at {ts} "
+        #         f"on {hostname}/{resp.text} (CUDA_VISIBLE_DEVICES={device}): {type(err).__name__}: {err}",
+        #         flush=True
+        #     )
+        raise error
+    except Exception:
+        if opt.debug and trainer.global_rank == 0:
+            try:
+                import pudb as debugger
+            except ImportError:
+                import pdb as debugger
+            debugger.post_mortem()
+        raise
+    finally:
+        # move newly created debug project to debug_runs
+        if opt.debug and not opt.resume and trainer.global_rank == 0:
+            dst, name = os.path.split(logdir)
+            dst = os.path.join(dst, "debug_runs", name)
+            os.makedirs(os.path.split(dst)[0], exist_ok=True)
+            os.rename(logdir, dst)
+
+        # if trainer.global_rank == 0:
+        #    print(trainer.profiler.summary())

vista/vwm/__init__.py

@@ -0,0 +1,6 @@
+from __future__ import annotations
+
+from .models import AutoencodingEngine, DiffusionEngine
+from .util import get_configs_path, instantiate_from_config
+
+__version__ = "0.1.0"