Skip to content
This repository has been archived by the owner on Oct 13, 2022. It is now read-only.

Use a pre-trained bigram P for LF-MMI training #222

Merged
merged 4 commits into from
Jul 2, 2021
Merged

Use a pre-trained bigram P for LF-MMI training #222

Changes from 1 commit
Commits
Show all changes
4 commits
Select commit Hold shift + click to select a range
c2c215b
Use a pre-trained bigram P for LF-MMI training
csukuangfj Jun 29, 2021
b80a3e8
Fix errors for decoding.
csukuangfj Jun 30, 2021
6632629
Support loading models trained with P_scores.
csukuangfj Jul 1, 2021
42c3256
More fixes for checkpoints loading.
csukuangfj Jul 1, 2021
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
6 changes: 4 additions & 2 deletions snowfall/common.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -88,9 +88,11 @@ def load_checkpoint(
src_key = '{}.{}'.format('module', key)
dst_state_dict[key] = src_state_dict.pop(src_key)
assert len(src_state_dict) == 0
model.load_state_dict(dst_state_dict)
model.load_state_dict(dst_state_dict, strict=False)
Copy link
Collaborator Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

@danpovey
Adding strict=False should prevent PyTorch from complaining
about extra key P_scores in the checkpoints.

else:
model.load_state_dict(checkpoint['state_dict'])
model.load_state_dict(checkpoint['state_dict'], strict=False)
# Note we used strict=False above so that the current code
# can load models trained with P_scores.

model.num_features = checkpoint['num_features']
model.num_classes = checkpoint['num_classes']
Expand Down