tocmo0nlord/axolotl
1
0
Fork 0
Code Issues Pull Requests Actions 2 Packages Projects Releases Wiki Activity

attempt at getting around bf16 error

Browse Source
This commit is contained in:
Sunny Liu
2025-02-04 21:57:21 -05:00
parent 3f6be519d5
commit d0e739da24
1 changed files with 1 additions and 1 deletions
Download Patch File Download Diff File Expand all files Collapse all files

2
src/axolotl/utils/models.py
View File

@@ -1104,7 +1104,7 @@ class ModelLoader:
should_convert = ( should_convert = (
# LlamaRMSNorm layers are in fp32 after kbit_training or full finetune, so we need to # LlamaRMSNorm layers are in fp32 after kbit_training or full finetune, so we need to
# convert them back to fp16/bf16 for flash-attn compatibility. # convert them back to fp16/bf16 for flash-attn compatibility.
((needs_fa2_dtype or self.cfg.flash_attention) and not qlora_fsdp) ((needs_fa2_dtype or self.cfg.flash_attention or self.cfg.flex_attention) and not qlora_fsdp)
or self.cfg.cut_cross_entropy # Cut cross entropy requires embedding layers to be in fp16/bf16 for backward pass or self.cfg.cut_cross_entropy # Cut cross entropy requires embedding layers to be in fp16/bf16 for backward pass
) )