SUMIN/HuggingFace_transformer
1
0
Fork 0
Code Issues Pull Requests Actions 7 Packages Projects Releases Wiki Activity

Fix the bug in constructing the all_hidden_states of DeBERTa v2 (#10466)

Browse Source 
* fix all_hidden_states

* use output_states instead of next_kv
This commit is contained in:
felixgwu
2021-03-03 12:05:21 -05:00
committed by GitHub
parent 188574ac50
commit d064fb5647
1 changed files with 2 additions and 1 deletions
Download Patch File Download Diff File Expand all files Collapse all files

3
src/transformers/models/deberta_v2/modeling_deberta_v2.py
View File

@@ -450,10 +450,11 @@ class DebertaV2Encoder(nn.Module):
else: else:
next_kv = hidden_states next_kv = hidden_states
rel_embeddings = self.get_rel_embedding() rel_embeddings = self.get_rel_embedding()
output_states = next_kv
for i, layer_module in enumerate(self.layer): for i, layer_module in enumerate(self.layer):
if output_hidden_states: if output_hidden_states:
all_hidden_states = all_hidden_states + (hidden_states,) all_hidden_states = all_hidden_states + (output_states,)
output_states = layer_module( output_states = layer_module(
next_kv, next_kv,