Update README.md

Upload v4_results.json with huggingface_hub

Upload sft_model_v4.pt with huggingface_hub

Upload model_arch.py with huggingface_hub

Upload exp_c_tokenizer_ablation.json with huggingface_hub

Upload exp_ab_results.json with huggingface_hub

Upload config.json with huggingface_hub

Upload base_model.pt with huggingface_hub

Upload README.md with huggingface_hub

Replace model_arch.py with correct architecture (train_sft_3b.py)

Fix architecture params in model card

Fix architecture params: DIM=3072, DEPTH=26, VOCAB=32000, N_HEADS=24

Update README.md

Add model card, config, tokenizer, and architecture code

initial commit