PhysiQuanty's picture
Update README.md
9e793ab verified
|
Raw
History Blame Contribute Delete
287 Bytes
---
license: apache-2.0
---
# Train From Scratch avec Flash Attention
Précision Supporté = FP32 , FP16 , BF16
Attention Supporté : Torch-Attention , Sage-Attention , Flash-Attention-2 , Flash-Attention-3
Bon train ! Que la loss tombe bien !
README.md en cours d'ecriture ...