Upload README.md

Upload evaluate_lcb.py

Upload merge_and_test.py

Upload README.md

Upload Dockerfile

Upload requirements.txt

Upload README.md

Upload train_ssd_sft.py

Update README with training results

Training in progress, step 300

Update README: fix hardware recommendations, add implementation notes from sandbox testing

Fix train_ssd.py: use correct model class (AutoModelForImageTextToText), target language_model only for LoRA, fix deprecated torch_dtype, remove OOM-causing prepare_model_for_kbit_training

Add comprehensive README

Add full SSD training script (non-QLoRA, for A100+)

Add SSD training script for Gemma-4-E2B-it

initial commit