ceos/unsloth
1
0
Fork 0
Code Issues Pull requests Projects Releases Packages Wiki Activity Actions

Update llama.py

Browse source
This commit is contained in:
Daniel Han-Chen 2024-03-18 03:22:49 +11:00
commit 9032334c30
1 changed files with 4 additions and 2 deletions
Download patch file Download diff file Expand all files Collapse all files

6
unsloth/models/llama.py
View file

@ -1336,7 +1336,6 @@ class FastLlamaModel:
"We shall do it for you!"
)
train_lm_head = True
model.model.embed_tokens.to(torch.float32, non_blocking = True)
elif module == "embed_tokens":
logger.warning_once(
@ -1344,7 +1343,6 @@ class FastLlamaModel:
"We shall do it for you!"
)
train_embed_tokens = True
model.lm_head.to(torch.float32, non_blocking = True)
else:
assert(module in accepted_modules)
@ -1386,8 +1384,12 @@ class FastLlamaModel:
# Now patch lm_head and embed_tokens
if train_embed_tokens:
model.model.model.embed_tokens.to(torch.float32)
model.model.model.embed_tokens.requires_grad_(True)
pass
if train_lm_head:
model.model.lm_head.to(torch.float32)
model.model.lm_head.requires_grad_(True)
pass