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 2024-01-21 19:21:56 +11:00
commit 2f80890578
1 changed files with 1 additions and 2 deletions
Download patch file Download diff file Expand all files Collapse all files

3
unsloth/models/llama.py
View file

@ -144,7 +144,7 @@ def LlamaAttention_fast_forward_inference(
A = torch.matmul(A, Vnn)
A = A.transpose(1, 2)
A = A.reshape(bsz, 1, self.hidden_size)
A = self.o_proj(self, A)
A = self.o_proj(A)
return A, (Kn, Vn)
pass
@ -304,7 +304,6 @@ def LlamaDecoderLayer_fast_forward(
past_key_value (`Tuple(torch.FloatTensor)`, *optional*): cached past key and value projection states
"""
bsz, q_len, hd = hidden_states.size()
print(hidden_states.size())
if (past_key_value is not None and q_len == 1):
# Self Attention