Hannes Kuchelmeister
commited on
Commit
·
6693f22
1
Parent(s):
e447dcf
add hyperparameter search for convolutional model
Browse files
configs/hparams_search/focusConvMSE_150_hyperparameter_search.yaml
ADDED
|
@@ -0,0 +1,76 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
# @package _global_
|
| 2 |
+
|
| 3 |
+
# example hyperparameter optimization of some experiment with Optuna:
|
| 4 |
+
# python train.py -m hparams_search=mnist_optuna experiment=example
|
| 5 |
+
|
| 6 |
+
defaults:
|
| 7 |
+
- override /datamodule: focus150.yaml
|
| 8 |
+
- override /model: focusConv_150.yaml
|
| 9 |
+
- override /hydra/sweeper: optuna
|
| 10 |
+
|
| 11 |
+
# choose metric which will be optimized by Optuna
|
| 12 |
+
# make sure this is the correct name of some metric logged in lightning module!
|
| 13 |
+
optimized_metric: "val/mae_best"
|
| 14 |
+
|
| 15 |
+
name: "focusConvMSE_150_hyperparameter_search"
|
| 16 |
+
|
| 17 |
+
# here we define Optuna hyperparameter search
|
| 18 |
+
# it optimizes for value returned from function with @hydra.main decorator
|
| 19 |
+
# docs: https://hydra.cc/docs/next/plugins/optuna_sweeper
|
| 20 |
+
hydra:
|
| 21 |
+
sweeper:
|
| 22 |
+
_target_: hydra_plugins.hydra_optuna_sweeper.optuna_sweeper.OptunaSweeper
|
| 23 |
+
|
| 24 |
+
# storage URL to persist optimization results
|
| 25 |
+
# for example, you can use SQLite if you set 'sqlite:///example.db'
|
| 26 |
+
storage: null
|
| 27 |
+
|
| 28 |
+
# name of the study to persist optimization results
|
| 29 |
+
study_name: focusConvMSE_150_hyperparameter_search
|
| 30 |
+
|
| 31 |
+
# number of parallel workers
|
| 32 |
+
n_jobs: 1
|
| 33 |
+
|
| 34 |
+
# 'minimize' or 'maximize' the objective
|
| 35 |
+
direction: minimize
|
| 36 |
+
|
| 37 |
+
# total number of runs that will be executed
|
| 38 |
+
n_trials: 20
|
| 39 |
+
|
| 40 |
+
# choose Optuna hyperparameter sampler
|
| 41 |
+
# docs: https://optuna.readthedocs.io/en/stable/reference/samplers.html
|
| 42 |
+
sampler:
|
| 43 |
+
_target_: optuna.samplers.TPESampler
|
| 44 |
+
seed: 12345
|
| 45 |
+
n_startup_trials: 10 # number of random sampling runs before optimization starts
|
| 46 |
+
|
| 47 |
+
# define range of hyperparameters
|
| 48 |
+
search_space:
|
| 49 |
+
datamodule.batch_size:
|
| 50 |
+
type: categorical
|
| 51 |
+
choices: [64, 128]
|
| 52 |
+
model.lr:
|
| 53 |
+
type: float
|
| 54 |
+
low: 0.0001
|
| 55 |
+
high: 0.2
|
| 56 |
+
model.pool_size:
|
| 57 |
+
type: categorical
|
| 58 |
+
choices: [1, 2, 3]
|
| 59 |
+
model.conv1_size:
|
| 60 |
+
type: categorical
|
| 61 |
+
choices: [3, 5, 7, 9]
|
| 62 |
+
model.conv1_channels:
|
| 63 |
+
type: categorical
|
| 64 |
+
choices: [1, 3, 6, 9]
|
| 65 |
+
model.conv2_size:
|
| 66 |
+
type: categorical
|
| 67 |
+
choices: [3, 5, 7, 9]
|
| 68 |
+
model.conv2_channels:
|
| 69 |
+
type: categorical
|
| 70 |
+
choices: [1, 3, 6, 9]
|
| 71 |
+
model.lin1_size:
|
| 72 |
+
type: categorical
|
| 73 |
+
choices: [16, 32, 64, 96, 128]
|
| 74 |
+
model.lin2_size:
|
| 75 |
+
type: categorical
|
| 76 |
+
choices: [16, 32, 64, 96, 128]
|