feat: use latest modeling code

merge: upload `transformers` implementation

update(README): lift gate

fix(modeling): use correct `base_model_prefix` name

feat: add dropout support

fix: make `flash_attn` optional (`trust_remote_code` breaks dynamic module check)

feat: add flash attention v2 (#9)

fix: remove `StableLMEpochForCausalLM.transformer` refs

fix: rename incorrect access to model

fix: init proper term

Upload folder using huggingface_hub

initial commit