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 2024-02-01 18:34:43 +11:00
commit a5ee70b63f
1 changed files with 7 additions and 3 deletions
Download patch file Download diff file Expand all files Collapse all files

10
unsloth/models/llama.py
View file

@ -284,13 +284,17 @@ pass
def fast_mlp_inference(self, X):
# gate = self.gate_proj(X)
# up = self.up_proj(X)
gate = fast_linear_forward(self.gate_proj, X)
up = fast_linear_forward(self. up_proj, X)
bsz, _, hd = X.shape
mlp_size = model.config.intermediate_size
# temp = torch.empty((2, bsz, 1, mlp_size), dtype = X.dtype, device = "cuda")
gate = fast_linear_forward(self.gate_proj, X)#, out = temp[0])
up = fast_linear_forward(self. up_proj, X)#, out = temp[1])
gate = torch.nn.functional.silu(gate, inplace = True)
gate *= up
# X = self.down_proj(gate)
down = fast_linear_forward(self.down_proj, gate, out = up[:,:,:X.shape[2]])
down = fast_linear_forward(self.down_proj, gate)#, out = up[:,:,:hd])
return down
pass