ayo/ipex-llm
1
Fork 0
Code Issues Pull requests Projects Releases Packages Wiki Activity Actions 1

llm: quick fix of fast_rms_norm (#9480)

Browse source
This commit is contained in:
Ruonan Wang 2023-11-16 14:42:16 +08:00 committed by GitHub
parent d5263e6681
commit c0ef70df02
2 changed files with 10 additions and 10 deletions
Show stats Download patch file Download diff file Expand all files Collapse all files

10
python/llm/src/bigdl/llm/transformers/models/baichuan2.py
View file

@ -53,11 +53,11 @@ def baichuan_13b_rms_norm_forward(self, hidden_states):
[self.weight.size(0)],
self.weight)
else:
hidden_states, _ = torch.ops.torch_ipex.fast_rms_norm(hidden_states,
[self.weight.size(0)],
self.weight,
None,
self.epsilon)
hidden_states = torch.ops.torch_ipex.fast_rms_norm(hidden_states,
[self.weight.size(0)],
self.weight,
None,
self.epsilon)
else:
input_dtype = hidden_states.dtype
hidden_states = hidden_states.to(torch.float32)

10
python/llm/src/bigdl/llm/transformers/models/llama.py
View file

@ -79,11 +79,11 @@ def llama_rms_norm_forward(self, hidden_states):
hidden_states, _ = torch.ops.torch_ipex.rms_norm(hidden_states,
[self.weight.size(0)], self.weight)
else:
hidden_states, _ = torch.ops.torch_ipex.fast_rms_norm(hidden_states,
[self.weight.size(0)],
self.weight,
None,
self.variance_epsilon)
hidden_states = torch.ops.torch_ipex.fast_rms_norm(hidden_states,
[self.weight.size(0)],
self.weight,
None,
self.variance_epsilon)
else:
input_dtype = hidden_states.dtype
hidden_states = hidden_states.to(torch.float32)