mrq/torchscale
1
0
Fork 0
Code Issues Pull Requests Packages Projects Releases Wiki Activity

Merge pull request #8 from buaahsh/main

Browse Source 
don't need attn weight in decoder
This commit is contained in:
Shuming Ma 2022-12-06 19:06:26 +08:00 committed by GitHub
commit 6d62bbbf67
No known key found for this signature in database
GPG Key ID: 4AEE18F83AFDEB23
1 changed files with 1 additions and 1 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

2
torchscale/architecture/decoder.py
View File

@ -462,7 +462,7 @@ class Decoder(nn.Module):
return x, { return x, {
"inner_states": inner_states, "inner_states": inner_states,
"l_aux": l_aux, "l_aux": l_aux,
"attn": [layer_attn.mean(dim=0)], "attn": None,
} }
def output_layer(self, features): def output_layer(self, features):