fix: key in config to match

fix fast tokenizer for transformers >= "from_slow": false,

chore: readme update

Update README.md

chore: update readme

chore: add checkpoint import

chore: upload pytorch standalone checkpoint

Update README.md

chore: add info on dtypes

chore: kernel disclaimer

chore: model name

chore: update logo size

chore: update readme

chore: update readme

chore: more info in the readme

chore: small fix in dec

Update tokenizer class in tokenizer_config,json to llama tokenizer.

Add model_type to config.json.

chore: update gradient checkpointing

chore: update config

fix: force correct mixed dtype after HF load

fix: force correct dtype in HF load

chore: remove generate

chore: remove flashfft dep

chore: info

chore: info

chore: sync readme

chore: add more information

chore: add some model card info

feat: initial commit

initial commit