Skip to content

GitLab

Unverified Commit 7ff7c493 authored by Stas Bekman's avatar Stas Bekman Committed by GitHub
Browse files

fixing crash (#8057)

parent cbad90d8
Hide whitespace changes
Inline Side-by-side
Showing with 2 additions and 2 deletions
src/transformers/modeling_deberta.py
View file @ 7ff7c493
...@@ -571,8 +571,8 @@ class DisentangledSelfAttention(torch.nn.Module): ...@@ -571,8 +571,8 @@ class DisentangledSelfAttention(torch.nn.Module):
k, v = [linear(qkvw[i], qkvb[i], hidden_states) for i in range(1, 3)] k, v = [linear(qkvw[i], qkvb[i], hidden_states) for i in range(1, 3)]
query_layer, key_layer, value_layer = [self.transpose_for_scores(x) for x in [q, k, v]] query_layer, key_layer, value_layer = [self.transpose_for_scores(x) for x in [q, k, v]]
query_layer += self.transpose_for_scores(self.q_bias[None, None, :]) query_layer = query_layer + self.transpose_for_scores(self.q_bias[None, None, :])
value_layer += self.transpose_for_scores(self.v_bias[None, None, :]) value_layer = value_layer + self.transpose_for_scores(self.v_bias[None, None, :])
rel_att = None rel_att = None
# Take the dot product between "query" and "key" to get the raw attention scores. # Take the dot product between "query" and "key" to get the raw attention scores.
......
Markdown is supported
0% or .
You are about to add 0 people to the discussion. Proceed with caution.
Finish editing this message first!
Please register or to comment