Loading ...

explosion/curated-transformers

🤖 A PyTorch library of curated Transformer models and their composable components

895

12

17

Notes

Sign in to add notes for this repository

1460

Followers

0

Following

Software company specializing in developer tools and tailored solutions for AI and Natural Language Processing

Branches

wip/maybe-abandoned-l2-loss-approach

wip/distillation

Releases

v2.0.1: Fix Python 3.12.3 compatibility

v1.3.2: Fix Python 3.12.3 compatibility

v2.0.0 (Superposition)

Commits

Fix activation lookup with Python 3.12.3 (#375) We used the metaclass `EnumMeta`/`EnumType` to override reporting of missing enum values (to give the full set of supported activations). However, in

Finalize the API changes for 2.0 (#374) * `qkv_split` arguments for attention heads are now mandatory * Rename `FromHFHub` mixins to `FromHF` * Remove `FromHF.convert_hf_state_dict`

Set version to 2.0.0 (#373) Also update the `curated-tokenizers` dependency to 2.0.0.

Issues

Thoughts on jaxtyping

Output logits for generation

Add Low-Rank Adapters injection into base models