Add Gradio training monitor and fix subprocess python path

- app.py: Gradio Space that streams GRPO training log live
- requirements.txt: training deps (unsloth, trl, wandb, gradio, etc.)
- README.md: sdk switched to gradio for training Space
- calibrate.py: fix python → sys.executable for portability
- train_grpo.py: fix python subprocess + add shutil.which fallback

Made-with: Cursor

requirements.txt

@@ -1,11 +1,11 @@
-fastapi==0.110.0
-uvicorn==0.29.0
-pydantic==2.6.4
-openai==2.7.2
-requests==2.31.0
-python-dotenv==1.0.1
-pytest==8.1.0
-httpx==0.27.0
-RestrictedPython==7.0
-openenv-core>=0.2.0
-gradio>=4.0
+--extra-index-url https://download.pytorch.org/whl/cu121
+gradio[oauth,mcp]==6.13.0
+pydantic==2.12.5
+wandb==0.18.7
+datasets==3.0.2
+transformers==4.46.3
+accelerate==1.0.1
+trl==0.14.0
+torch==2.5.1+cu121
+bitsandbytes==0.43.3
+peft==0.13.2