Fix precision errors from casting rotary parameters to FP16 with AMP #27700
Update modeling_llama.py
643098a9
Update modeling_open_llama.py
cd763d21
Update modeling_gpt_neox.py
659b8fc3
Update modeling_mistral.py
9dfac9e0
Update modeling_persimmon.py
d3aef861
Update modeling_phi.py
3f81d376
kevinhu
changed the title Fix precision errors from casting rotary parameters to FP16 with AMP in Llama Fix precision errors from casting rotary parameters to FP16 with AMP 2 years ago
Update modeling_falcon.py
db4c79fe
Update modeling_gpt_neox_japanese.py
d00f0868
Merge branch 'huggingface:main' into fix-einsum-amp
ecaf5121
Merge branch 'huggingface:main' into fix-einsum-amp
3f171917
Assignees
No one assigned
Login to write a write a comment.
Login via GitHub