Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Jump to bottom

Add Whisper and FLEURS #10

Merged
merged 53 commits into from
Aug 29, 2023
Merged

Add Whisper and FLEURS #10

merged 53 commits into from
Aug 29, 2023

Conversation

saattrupdan
Copy link
Collaborator

@saattrupdan saattrupdan commented Jul 31, 2023 •
edited
Loading

This PR:

  1. Adds support to finetune Whisper models
  2. Adds the Danish, Norwegian and Swedish FLEURS datasets

Closes #17 and #19.

@saattrupdan
refactor: Move load_model_setup to separate model_setup module
08fad7c
@saattrupdan
style: Use getattr for Processor.tokenizer
5859c15
@saattrupdan
style: Remove comment
ecd9f8e
@saattrupdan
style: Remove tokenizer attribute from Processor protocol
cf07771
@saattrupdan
chore: Rename script names to avoid name conflict
dfc5575
@saattrupdan
chore: Do not import finetune function in __init__
d22ae95
@saattrupdan
feat: Enable save_safetensors
f6b2333
@saattrupdan
chore: Update lock file
7deef7d
@saattrupdan
feat: Set up whisper finetuning
6be03a3
@saattrupdan
chore: More whisper configs, split up whisper/wav2vec2 specific options
50cf83b
@saattrupdan
chore: Only use sampling_rate in model config
92cccf8
@saattrupdan
feat: Add Nota config
d923798
@saattrupdan
chore: Change config
a54715c
@saattrupdan
feat: Set up more dataset configs, and set seed in config
582a54c
@saattrupdan
chore: Ignore wandb folder
9e032a8
@saattrupdan
docs: Update code coverage
748c787
@saattrupdan
chore: Update configs
8152ca4
@saattrupdan
tests: Add finetune test
d871012
@saattrupdan
tests: Fix logging test
a199df8
@saattrupdan
fix: Wandb import
62a73ef
@saattrupdan
fix: In preprocessing, allow both input_values and input_features
d28000c
@saattrupdan
fix: Do not apply processor in wav2vec2 data collator, as it has been…
8ab3d62 
… done in advance
@saattrupdan
docs: Update coverage badge
7871e38
@saattrupdan
chore: Config renaming
3058598
@saattrupdan
tests: Test whisper models too
4b47821
@saattrupdan
docs: Update coverage badge
071b0f3
@saattrupdan saattrupdan self-assigned this Jul 31, 2023
@saattrupdan saattrupdan requested review from emilhovad and removed request for rlrs August 7, 2023 11:01
@saattrupdan saattrupdan changed the title Add Whisper, Nota and FLEURS Add Whisper and FLEURS Aug 12, 2023
@saattrupdan saattrupdan removed a link to an issue Aug 12, 2023
Enable finetuning on Nota #18
Closed
@saattrupdan saattrupdan mentioned this pull request Aug 12, 2023
Merged
AJDERS
AJDERS requested changes Aug 15, 2023
View reviewed changes
src/coral_models/compute_metrics.py Show resolved Hide resolved
src/coral_models/compute_metrics.py Show resolved Hide resolved
src/coral_models/data.py Show resolved Hide resolved
sorenmulli
sorenmulli approved these changes Aug 24, 2023
View reviewed changes
Copy link
Collaborator

@sorenmulli sorenmulli left a comment

Choose a reason for hiding this comment

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

LGTM; seems like beautiful implementation of straight forward features

config/model/wav2vec2.yaml Outdated Show resolved Hide resolved
src/coral_models/compute_metrics.py Show resolved Hide resolved
src/coral_models/data.py Show resolved Hide resolved
src/coral_models/data.py Show resolved Hide resolved
src/coral_models/wav2vec2.py Show resolved Hide resolved
@saattrupdan saattrupdan merged commit 118552a into main Aug 29, 2023
7 checks passed
@saattrupdan saattrupdan deleted the feat/add-whisper branch August 29, 2023 08:27
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Reviewers

@AJDERS AJDERS AJDERS approved these changes

@sorenmulli sorenmulli sorenmulli approved these changes

@emilhovad emilhovad Awaiting requested review from emilhovad

@zer0x42 zer0x42 Awaiting requested review from zer0x42

Assignees

@saattrupdan saattrupdan

Labels
None yet
Projects
None yet
Milestone
No milestone
Development

Successfully merging this pull request may close these issues.

Enable finetuning on FLEURS Add whisper finetuning support
3 participants
@saattrupdan @AJDERS @sorenmulli