Whisper encoder + No 30 second padding #5
Merged
Conversation
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
farzadab
commented
Jun 4, 2024
juberti
approved these changes
Jun 4, 2024
akshat0311
pushed a commit
to jiviai/audio-llm
that referenced
this pull request
Jan 30, 2025
* enable whisper model (no need for max_padding) * bugfix: boolq_in didn't have audio_transcript -> use _get_transcribe_sample * bugfix: move model to device before merging lora weights * rename modified_whisper -> whisper_model_modified
zqhuang211
pushed a commit
that referenced
this pull request
Feb 12, 2025
* LSM training * use WavTokenizer directly instead of CustomWavTokenizer
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
This PR enables Whisper model for training and inference, without the need for full 30 second padding.
I verified that the inference works with:
$ python -m ultravox.tools.infer_tool -T 32 -n 10 -d boolq_in --asr \ -m wandb://fixie/ultravox/model-llama3_whisper_s__gs__cont_lora_gs_ai_constant__longest_pad:v1 --- Sample 0 --- Q: Transcribe <|audio|> ["do iran and afghanistan speak the same language?"] A: Do Iran and Afghanistan speak the same language? X: do iran and afghanistan speak the same language? [wer: 0.00, avg: 0.00] --- Sample 1 --- Q: Transcribe <|audio|> ["do good samaritan laws protect those who help at an accident?"] A: Do good Samaritan laws protect those who help at an accident? X: do good samaritan laws protect those who help at an accident? [wer: 0.00, avg: 0.00] --- Sample 2 --- Q: Transcribe <|audio|> ["is windows movie maker part of windows essentials?"] A: Is Windows Movie Maker part of windows essentials? X: is windows movie maker part of windows essentials? [wer: 0.00, avg: 0.00] --- Sample 3 --- Q: Transcribe <|audio|> ["is confectionary sugar the same as powdered sugar?"] A: Is confectionary sugar the same as powdered sugar? X: is confectionary sugar the same as powdered sugar? [wer: 0.00, avg: 0.00] --- Sample 4 --- Q: Transcribe <|audio|> ["is elder scrolls online the same as skyrim?"] A: Is elder scrolls Online the same as Skyrim? X: is elder scrolls online the same as skyrim? [wer: 0.00, avg: 0.00] --- Sample 5 --- Q: Transcribe <|audio|> ["can you use oyster card at epsom station?"] A: Can you use Oyster card at Upminster station? X: can you use oyster card at epsom station? [wer: 0.12, avg: 0.02] --- Sample 6 --- Q: Transcribe <|audio|> ["will there be a season 4 of da vinci's demons?"] A: Will there be a season four of Downton Abbey? X: will there be a season 4 of da vinci's demons? [wer: 0.45, avg: 0.08] ...