Commit Graph

4 Commits

Author SHA1 Message Date
DocShotgun
39f7a2aabd Expose draft_rope_scale 2023-12-05 12:59:32 -08:00
DocShotgun
67507105d0 Update colab, expose additional args
* Exposed draft model args for speculative decoding
* Exposed int8 cache, dummy models, and no flash attention
* Resolved CUDA 11.8 dependency issue
2023-12-04 22:20:46 -08:00
veryamazinglystupid
ad1a12a0f2 make colab better, fix libcudart errors
:3
2023-12-03 14:07:52 +05:30
DocShotgun
2a9e4ca051 Add Colab example
*note: this uses wheels for python 3.10 and torch 2.1.0+cu118 which is the current default in colab
2023-12-03 02:21:51 -05:00