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.

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

Speechllm develop gen duplex2 clean #11450

Draft
wants to merge 53 commits into
base: speechllm-develop-gen
Choose a base branch
from

Conversation

zhehuaichen
Copy link
Collaborator

What does this PR do ?

Add a one line overview of what this PR aims to accomplish.

Collection: [Note which collection this PR will affect]

Changelog

  • Add specific line by line info of high level changes in this PR.

Usage

  • You can potentially add a usage example below
# Add a code snippet demonstrating how to use this 

GitHub Actions CI

The Jenkins CI system has been replaced by GitHub Actions self-hosted runners.

The GitHub Actions CI will run automatically when the "Run CICD" label is added to the PR.
To re-run CI remove and add the label again.
To run CI on an untrusted fork, a NeMo user with write access must first click "Approve and run".

Before your PR is "Ready for review"

Pre checks:

  • Make sure you read and followed Contributor guidelines
  • Did you write any new necessary tests?
  • Did you add or update any necessary documentation?
  • Does the PR affect components that are optional to install? (Ex: Numba, Pynini, Apex etc)
    • Reviewer: Does the PR have correct import guards for all optional libraries?

PR Type:

  • New Feature
  • Bugfix
  • Documentation

If you haven't finished some of the above items you can still open "Draft" PR.

Who can review?

Anyone in the community is free to review the PR once the checks have passed.
Contributor guidelines contains specific people who can review PRs to various areas.

Additional Information

  • Related to # (issue)

Subhankar Ghosh and others added 30 commits October 28, 2024 13:51
https://docs.google.com/document/d/1LRMwpUt8TH96oRVi0zz7OOTbejrMk9zs4VH1SEX59ZM/edit?tab=t.0
and
/lustre/fsw/portfolios/llmservice/users/kevinhu/works/mod_speech_llm/code/NeMo_s2s_align_debug/

Signed-off-by: zhehuaichen <[email protected]>
…eechllm-develop-gen_align_otf

Signed-off-by: zhehuaichen <[email protected]>
…arg' into speechllm-develop-gen_align_otf

Signed-off-by: zhehuaichen <[email protected]>
Signed-off-by: zhehuaichen <[email protected]>
[NeMo I 2024-11-14 08:26:21 modular_s2s_models:677] Label: `Relatively insensitive to differences in musical pitch.` text_answer_text: `⁇   Relatively ⁇  ⁇  ⁇  ⁇  ⁇  ⁇  insensitive ⁇
  ⁇  ⁇  ⁇  ⁇  ⁇  ⁇  ⁇  ⁇  ⁇  ⁇  ⁇  to ⁇  ⁇  ⁇  ⁇  ⁇  ⁇  ⁇  ⁇  ⁇  ⁇  differences ⁇  ⁇  ⁇  ⁇  ⁇  ⁇  ⁇  in ⁇  ⁇  ⁇  musical ⁇  ⁇  ⁇  ⁇  ⁇  ⁇  ⁇  ⁇  ⁇  ⁇  ⁇  ⁇  ⁇  ⁇  ⁇  ⁇  ⁇  ⁇  ⁇  ⁇  ⁇
  ⁇  ⁇  ⁇  ⁇  ⁇  ⁇  ⁇  ⁇  ⁇  ⁇  ⁇  ⁇  ⁇  ⁇  ⁇  ⁇  ⁇  ⁇  ⁇  ⁇  ⁇`
[NeMo I 2024-11-14 08:26:21 modular_s2s_models:678] Pred: `⁇   Relatively ⁇  ⁇  ⁇  ⁇  ⁇  ⁇  insensitive ⁇  ⁇  ⁇  ⁇  ⁇  ⁇  ⁇  ⁇  ⁇  ⁇  ⁇  ⁇  to ⁇  ⁇  ⁇  ⁇  ⁇  ⁇  ⁇  ⁇  ⁇  ⁇  differences ⁇  ⁇  ⁇  ⁇  ⁇  ⁇  ⁇  in ⁇  ⁇  ⁇  musical ⁇  ⁇  ⁇  ⁇  ⁇  ⁇  ⁇  ⁇  ⁇  ⁇  ⁇  ⁇  ⁇  ⁇  ⁇  ⁇  ⁇  ⁇  ⁇  ⁇  ⁇  ⁇  ⁇  ⁇  ⁇  ⁇  ⁇  ⁇  ⁇  ⁇  ⁇  ⁇  ⁇  ⁇  ⁇  ⁇  ⁇  ⁇  ⁇  ⁇  ⁇  ⁇  ⁇  ⁇  ⁇  ⁇
⁇  ⁇  ⁇  ⁇  ⁇  ⁇  ⁇  ⁇  ⁇  ⁇  ⁇  ⁇  ⁇  ⁇  ⁇  ⁇  ⁇  ⁇  ⁇  ⁇  ⁇  ⁇  ⁇  ⁇  ⁇  ⁇  ⁇  ⁇  ⁇  ⁇  ⁇  ⁇  ⁇  ⁇`
[NeMo I 2024-11-14 08:26:21 modular_s2s_models:679] speech_preds_transcribed: `relatively insensitive to differences in musical pitch`
[NeMo I 2024-11-14 08:26:21 modular_s2s_models:680] speech_answers_transcribed: ``
[NeMo I 2024-11-14 08:26:21 modular_s2s_models:681] Speech out len: pred (53, 8) label (53, 8)
[NeMo I 2024-11-14 08:26:21 modular_s2s_models:717] validation bleu: 0.04974522814154625
[NeMo I 2024-11-14 08:26:21 modular_s2s_models:717] validation asr-bleu: 53.29462814331055

Signed-off-by: zhehuaichen <[email protected]>
Signed-off-by: zhehuaichen <[email protected]>
Signed-off-by: zhehuaichen <[email protected]>
Signed-off-by: zhehuaichen <[email protected]>
Signed-off-by: zhehuaichen <[email protected]>
Signed-off-by: zhehuaichen <[email protected]>
also fix previous bug in instruction_lengths and features_lens from
codec_model_downsampling_factor

Signed-off-by: zhehuaichen <[email protected]>
Signed-off-by: zhehuaichen <[email protected]>
Signed-off-by: zhehuaichen <[email protected]>
Signed-off-by: zhehuaichen <[email protected]>
* Fix instruction_lengths and bos

* Update training config

* Add Causal Decoder

* Apply isort and black reformatting

Signed-off-by: Edresson <[email protected]>

* Apply isort and black reformatting

Signed-off-by: artbataev <[email protected]>

---------

Signed-off-by: Edresson <[email protected]>
Signed-off-by: artbataev <[email protected]>
Co-authored-by: kevinhu <[email protected]>
Co-authored-by: Edresson <[email protected]>
Co-authored-by: artbataev <[email protected]>
Signed-off-by: zhehuaichen <[email protected]>
add inference code
Signed-off-by: zhehuaichen <[email protected]>
Signed-off-by: zhehuaichen <[email protected]>
Signed-off-by: zhehuaichen <[email protected]>
…uplex shars

verify the overfit on 2 examples with 4 turns and 8 turns

[NeMo I 2024-11-26 12:56:49 modular_s2s_models:682] Label: `⁇  ⁇  ⁇  ⁇
⁇  ⁇  ⁇  ⁇  ⁇  ⁇  ⁇  ⁇  ⁇  ⁇  ⁇  ⁇  ⁇  ⁇  ⁇  ⁇  ⁇  ⁇  ⁇  ⁇  ⁇  ⁇  ⁇   I
could not find the answer ⁇  ⁇  ⁇  ⁇ ⁇  ⁇  ⁇  ⁇  ⁇  ⁇  ⁇  ⁇  ⁇  ⁇  ⁇  ⁇
⁇  ⁇  ⁇  ⁇  ⁇  ⁇  ⁇  ⁇  ⁇  ⁇  ⁇   The plural form of axis. ⁇  ⁇  ⁇  ⁇  ⁇
⁇  ⁇  ⁇  ⁇  ⁇  ⁇  ⁇  ⁇  ⁇  ⁇  ⁇  ⁇  ⁇  ⁇  ⁇  ⁇  ⁇  ⁇  ⁇   Relatively
insensitive to differences pitch. ⁇  ⁇  ⁇  ⁇  ⁇  ⁇  ⁇  ⁇  ⁇  ⁇  ⁇  ⁇  ⁇
⁇  ⁇  ⁇  ⁇  ⁇  ⁇  ⁇  ⁇  ⁇  ⁇  ⁇  ⁇  ⁇   The Eustachian tube`
text_answer_text: `⁇  ⁇  ⁇  ⁇  ⁇  ⁇  ⁇  ⁇  ⁇  ⁇  ⁇  ⁇  ⁇  ⁇  ⁇  ⁇  ⁇  ⁇
⁇  ⁇  ⁇  ⁇  ⁇  ⁇  ⁇  ⁇  ⁇   I could not find the answer ⁇  ⁇  ⁇  ⁇  ⁇  ⁇
⁇  ⁇  ⁇  ⁇  ⁇  ⁇  ⁇  ⁇  ⁇  ⁇  ⁇  ⁇  ⁇  ⁇  ⁇  ⁇  ⁇  ⁇  ⁇  ⁇  ⁇   The
plural form of axis. ⁇  ⁇  ⁇  ⁇  ⁇  ⁇  ⁇  ⁇  ⁇  ⁇  ⁇  ⁇  ⁇  ⁇  ⁇  ⁇  ⁇
⁇  ⁇  ⁇  ⁇  ⁇  ⁇  ⁇   Relatively insensitive to differences pitch. ⁇  ⁇
⁇  ⁇  ⁇  ⁇  ⁇  ⁇  ⁇  ⁇  ⁇  ⁇  ⁇  ⁇  ⁇  ⁇  ⁇  ⁇  ⁇  ⁇  ⁇  ⁇  ⁇  ⁇  ⁇  ⁇
The Eustachian tube`
[NeMo I 2024-11-26 12:56:49 modular_s2s_models:683] Pred: `⁇  ⁇  ⁇  ⁇  ⁇
⁇  ⁇  ⁇  ⁇  ⁇  ⁇  ⁇  ⁇  ⁇  ⁇  ⁇  ⁇  ⁇  ⁇  ⁇  ⁇  ⁇  ⁇  ⁇  ⁇   I could not
not the answer ⁇  ⁇  ⁇  ⁇  ⁇  ⁇  ⁇  ⁇  ⁇  ⁇  ⁇  ⁇  ⁇  ⁇  ⁇  ⁇  ⁇  ⁇  ⁇
⁇  ⁇  ⁇  ⁇  ⁇  ⁇  ⁇   The plural form of axis. ⁇  ⁇  ⁇  ⁇  ⁇  ⁇  ⁇  ⁇  ⁇
⁇  ⁇  ⁇  ⁇  ⁇  ⁇  ⁇  ⁇  ⁇  ⁇  ⁇  ⁇  ⁇  ⁇  ⁇   Relatively insensitive to
differences pitch.. ⁇  ⁇  ⁇  ⁇  ⁇  ⁇  ⁇  ⁇  ⁇  ⁇  ⁇  ⁇  ⁇  ⁇  ⁇  ⁇  ⁇  ⁇
⁇  ⁇  ⁇  ⁇  ⁇  ⁇  ⁇   The Eustachian tube pl pl of. axis. ⁇ . ⁇  ⁇  ⁇  ⁇
⁇ ach ⁇  ⁇  ⁇  ⁇  ⁇ .. axisach.itiveitive of differences... of the
differences ⁇  ⁇  ⁇ atively.. of ⁇  ⁇ .... differences differences
differences differences differences differences . differences      ⁇
differences  ⁇  ⁇  ⁇  ⁇  ⁇  ⁇  ⁇  ⁇  ins ⁇  ⁇  ⁇  ⁇  ⁇  ins ⁇ .
differences ian of ⁇  ⁇ itive differences ⁇  differences differences
differences axis differences differences axis`

[NeMo I 2024-11-26 12:56:49 modular_s2s_models:684]
speech_preds_transcribed: `i could not find the answer in the audio form
of axis is axes relatively insensitive to differences in musical pitch
the eustacian tube leads in the eard drum`

Signed-off-by: zhehuaichen <[email protected]>
Signed-off-by: zhehuaichen <[email protected]>
Signed-off-by: zhehuaichen <[email protected]>
Signed-off-by: zhehuaichen <[email protected]>
…nto speechllm-develop-gen_duplex2

Signed-off-by: zhehuaichen <[email protected]>
zhehuaichen and others added 14 commits December 2, 2024 14:20
…uaichen/NeMo into speechllm-develop-gen_duplex2_clean
Signed-off-by: zhehuaichen <[email protected]>
Signed-off-by: zhehuaichen <[email protected]>
Signed-off-by: zhehuaichen <[email protected]>
Signed-off-by: zhehuaichen <[email protected]>
Signed-off-by: zhehuaichen <[email protected]>
@github-actions github-actions bot added the ASR label Dec 10, 2024
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Projects
None yet
Development

Successfully merging this pull request may close these issues.

4 participants