Deploy CoRGI demo - 2025-10-29 14:27:36

Features:
- Structured reasoning with CoRGI protocol
- ROI extraction using Qwen3-VL grounding
- Visual evidence synthesis
- Gradio UI with per-step visualization

Model: Qwen/Qwen3-VL-8B-Thinking

README.md

@@ -12,7 +12,7 @@ license: apache-2.0
 
 # CoRGI Qwen3-VL Demo
 
-This Space showcases the CoRGI reasoning pipeline powered entirely by **Qwen/Qwen3-VL-4B-Instruct**.  
+This Space showcases the CoRGI reasoning pipeline powered entirely by **Qwen/Qwen3-VL-2B-Instruct**.  
 Upload an image, ask a visual question, and the app will:
 
 1. Generate structured reasoning steps with visual-verification flags.  
@@ -24,7 +24,7 @@ Upload an image, ask a visual question, and the app will:
 ```bash
 pip install -r requirements.txt
 python examples/demo_qwen_corgi.py \
-  --model-id Qwen/Qwen3-VL-4B-Instruct \
+  --model-id Qwen/Qwen3-VL-2B-Instruct \
   --max-steps 3 \
   --max-regions 3
 ```
@@ -35,9 +35,17 @@ To launch the Gradio demo locally:
 python app.py
 ```
 
+## 📚 Full Documentation
+
+See **[docs/](docs/)** folder for complete documentation:
+- 🚀 **[Quick Start](docs/START_HERE.md)** - Begin here!
+- 📖 **[Usage Guide](docs/USAGE_GUIDE.md)** - How to use
+- 🔧 **[Deployment](docs/DEPLOY_NOW.md)** - Deploy to HF Spaces
+- 📊 **[Summary Report](docs/SUMMARY_REPORT.md)** - Full overview
+
 ## Configuration Notes
 
-- **Model**: Uses `Qwen/Qwen3-VL-4B-Instruct` (4B parameters, ~8GB VRAM)
+- **Model**: Uses `Qwen/Qwen3-VL-2B-Instruct` (2B parameters, ~5GB VRAM)
 - **Single GPU**: Model loads on single GPU (cuda:0) to avoid memory fragmentation
 - **Hardware**: The Space runs on `cpu-basic` tier by default
 - **Customization**: Set `CORGI_QWEN_MODEL` environment variable to use a different checkpoint

corgi/cli.py

@@ -12,7 +12,7 @@ from .pipeline import CoRGIPipeline
 from .qwen_client import Qwen3VLClient, QwenGenerationConfig
 from .types import GroundedEvidence, ReasoningStep
 
-DEFAULT_MODEL_ID = "Qwen/Qwen3-VL-4B-Instruct"
+DEFAULT_MODEL_ID = "Qwen/Qwen3-VL-2B-Instruct"
 
 
 def build_parser() -> argparse.ArgumentParser:

corgi/qwen_client.py

@@ -128,7 +128,7 @@ def _load_backend(model_id: str) -> tuple[AutoModelForImageTextToText, AutoProce
 
 @dataclass
 class QwenGenerationConfig:
-    model_id: str = "Qwen/Qwen3-VL-4B-Instruct"
+    model_id: str = "Qwen/Qwen3-VL-2B-Instruct"
     max_new_tokens: int = 512
     temperature: float | None = None
     do_sample: bool = False