search

ArtemisDicoTiar / FastLLM

1 stars 0 forks source link

Target Model Distribution Dataset Generation #3

Closed ArtemisDicoTiar closed 7 months ago

ArtemisDicoTiar commented 7 months ago

Resources Available

3090: 24G A6000: 48G

Target Model 1

T5-Large this model requires 3G GPU mem with full precision

Target Model 2

LLama2-7B this model requires 28G GPU mem with full precision and 14G with half precision (float16).