train: shrink to weekly horizon + bounded steps

train: default HF Job flavor l4x1 -> l40sx1 (48GB VRAM)

train: per-step credit + drop replies + larger batches

fix(notebook): py3.11 f-string backslash error in format_obs

Track PNGs with LFS

update

fix(notebook): pin typing_extensions>=4.13.0 to fix pydantic Sentinel ImportError

fix: restore parse_model_output exception parity with original bare except

chore: align train_grpo.ipynb with smoke/syntax patterns for Colab

add training/syntax_only.ipynb — kernel + Python syntax only (no project logic)

add train_grpo_smoke notebook; quote pip versions in train_grpo

fix: notebook loads Qwen without bitsandbytes on Mac; optional training deps

fix: robust notebook setup (no magic shell) + local CWD auto-detect

update

fix: rewrite training notebook for real LoRA fine-tuning on Colab

la la la --123

firstiteration