Update llama.py

This commit is contained in:
Daniel Han-Chen 2024-02-03 19:19:13 +11:00
commit 9225dd6708

View file

@ -209,7 +209,7 @@ def fast_mlp_inference(self, X):
gate = torch.nn.functional.silu(gate, inplace = True)
gate *= up
X = self.down_proj(gate)
down = self.down_proj(gate)
# down = fast_linear_forward(self.down_proj, gate)#, out = up[:,:,:hd])
return down
pass