CTC-only training recipes for LibriSpeech (code from Samsung AI Cambridge) #2290

shucongzhang · 2023-12-05T14:15:52Z

What does this PR do?

This PR provides recipes for training CTC models from scratch (no W2V2, no Whisper). It yields to strong Branchformer-CTC and Conformer-CTC models.

TParcollet · 2023-12-05T14:20:14Z

speechbrain/lobes/models/transformer/TransformerASR.py

@@ -152,9 +152,12 @@ def __init__(
            ),
            torch.nn.Dropout(dropout),
        )
-        self.custom_tgt_module = ModuleList(


@asumagic I guess that this is the place of the problem?

Yes, this change does what I ultimately suggested to do in the discussion in my PR, with the same consequences:

Though, now, I'm thinking that we could just skip the headache and fix the problem by gating it behind num_decoder_layers > 0 and break models: As far as I can tell, actually, the only affected model in the SB repo seems to be the conformer-transducer model.
Then, just in case, we could provide a script that would remove the module from a checkpoint file.

TParcollet · 2023-12-05T14:21:14Z

Thanks @shucongzhang ! Could you start by fixing the tests? It can be done locally by installing pre-commit :-)

shucongzhang · 2023-12-05T17:06:13Z

Thanks @shucongzhang ! Could you start by fixing the tests? It can be done locally by installing pre-commit :-)

Definitely. I have fixed the tests :)

mravanelli · 2023-12-11T20:42:38Z

Thank you @shucongzhang! This is a great contribution. My main comment is that we need to update this PR to the new version of speechbrain that is currently available in the unstable branch (but soon will be merged in the development branch).
You can take a look at the recipes available in LibriSpeech/ASR/CTC for a reference. @Adel-Moumen can help with that conversion. With the new version of speechbrain, you can also improve the performance using CTC beamsearch with n-gram LM rescoring.

Adel-Moumen · 2024-01-16T11:38:39Z

Hello @shucongzhang,

Thanks for your great work!

We recently released in develop the new SpeechBrain 1.0 version (see: https://colab.research.google.com/drive/1IEPfKRuvJRSjoxu22GZhb3czfVHsAy0s?usp=sharing), and unfortunately, this PR needs to be sync with the latest commits in the develop branch so that we can move forward with this PR.

Best,
Adel

recipes/LibriSpeech/ASR/CTC/train_from_scratch.py

…into LS_CTC

Adel-Moumen · 2024-01-18T18:18:56Z

Hi everyone,

I have been modifying this PR so that it complains with SB 1.0. I also added support for beam search decoding (and latter I will run n-gram decoding as well).

I'm in the process of training the models. I found that one epoch is about ~40 minutes on an A100 80GB with fp16. However, in the yaml @shucongzhang specified number_of_epochs: 500. It seems a bit too much, isn't it ? 500 epochs means ~14 days of training ((500 * 40) / 60 / 24). I suppose we could lower this number right ? I can increase the number of GPUs, but it would have been great to have a competitive conformer/branchformer that works on 1 GPU at a time (<2-3 days).

Ping @TParcollet as well as I guess you were part of this PR.

Best,
Adel

speechbrain/nnet/schedulers.py

Adel-Moumen · 2024-01-25T11:29:25Z

Hello @shucongzhang small ping on my previous message about training time. Could you please confirm me that 500 epochs is reasonable ? Ty.

shucongzhang · 2024-01-25T15:27:41Z

Hello @shucongzhang small ping on my previous message about training time. Could you please confirm me that 500 epochs is reasonable ? Ty.

Hi @Adel-Moumen , Yes, the training time also makes me feel annoying. From my knowledge and my experiments, it seems for the vanilla Conformer/Branchformer a large batch size and a large number of epochs are necessary. Based on my Conformer training log,

epoch 250: 3.68 dev-clean WER 
epoch 300: 3.62 dev-clean WER
epoch 400: 3.50 dev-clean WER
epoch 500: 3.45 dev-clean WER

Thus, maybe it is Okay to do some trade off between the training time and WER? Thx!

Adel-Moumen · 2024-02-11T13:24:59Z

I'm wondering about something. Why are we using a sentencepiece 128 BPE vocab for our CTC branch/conformer ? If you are training with CTC then why not using label_encoder with chars ? Ping @TParcollet / @shucongzhang

shucongzhang · 2024-02-19T19:27:17Z

I'm wondering about something. Why are we using a sentencepiece 128 BPE vocab for our CTC branch/conformer ? If you are training with CTC then why not using label_encoder with chars ? Ping @TParcollet / @shucongzhang

@Adel-Moumen Hi Adel, I was in China for the Lunar New Year and just back to work. Happy Lunar New Year to you :) I used the 128 BPE to follow some previous works. I didn't test what will be the results if using using label_encoder with chars.

Adel-Moumen · 2024-02-20T10:29:37Z

I'm wondering about something. Why are we using a sentencepiece 128 BPE vocab for our CTC branch/conformer ? If you are training with CTC then why not using label_encoder with chars ? Ping @TParcollet / @shucongzhang

@Adel-Moumen Hi Adel, I was in China for the Lunar New Year and just back to work. Happy Lunar New Year to you :) I used the 128 BPE to follow some previous works. I didn't test what will be the results if using using label_encoder with chars.

@shucongzhang Welcome back, and Happy Lunar New Year to you too 🎉! Thanks for the clarification. It makes sense from your point of view. I will try with label_encoder since I think it could be useful here (and from a CTC point of view, I don't really understand the need of using BPE instead of phonemes/chars).

I let you know when I have some results to share. I might take a bit of time due to the next release of SB 1.0, but I'll let you know what happens :)

TParcollet · 2024-02-21T11:52:16Z

@Adel-Moumen we should avoid using label_encoder imho. SentencePiece can be degenerated to char only. I don't like our recipes that use label_encoder, we have developed this class because we did not want to use SentencePiece at first, but here we are. Maybe we should deprecate it.

…into LS_CTC

Adel-Moumen · 2024-04-18T10:11:04Z

Hello,

So I had the opportunity to train a bit (~200 epochs) with each model (Branchformer and Conformer), and everything ran smoothly :)

I will therefore merge this PR. Thanks a lot @shucongzhang for your very nice work. :)

Tests

python -c 'from tests.utils.recipe_tests import run_recipe_tests; print("TEST FAILED!") if not(run_recipe_tests(filters_fields=["Hparam_file"], filters=[["recipes/LibriSpeech/ASR/CTC/hparams/conformer_large.yaml", "recipes/LibriSpeech/ASR/CTC/hparams/branchformer_large.yaml"]], do_checks=False, run_opts="--device=cuda")) else print("TEST PASSED")'

(1/2) Running test for LibriSpeech_row_41...
        ... 11.01s
(2/2) Running test for LibriSpeech_row_42...
        ... 15.03s
TEST PASSED

Adel-Moumen

LGTM. Thanks again @shucongzhang :)

shucongzhang · 2024-04-22T14:29:26Z

LGTM. Thanks again @shucongzhang :)

Thank you so much @Adel-Moumen !

…AI Cambridge) (speechbrain#2290)" This reverts commit d086cde.

* Skip lazy imports when the caller is inspect.py This avoids having certain inspect functions import our lazy modules when we don't want them to. `getframeinfo` in particular appears to do it, and this gets called by PyTorch at some point. IPython might also be doing it but autocomplete still seems to work. This does not appear to break anything. Added test for hyperpyyaml to ensure we're not breaking that. * SSL_Semantic_Token _ new PR (speechbrain#2509) * remove unnecassry files and move to dasb * remove extra recepie from test * update ljspeech qunatization recepie * add discrete_ssl and remove extra files * fix precommit * update kmeans and add tokeizer for postprocessing * fix precommit * Update discrete_ssl.py * fix clone warning --------- Co-authored-by: Mirco Ravanelli <mirco.ravanelli@gmail.com> * _ensure_module Raises docstring * Expose `ensure_module` so that docs get generated for it This is already an internal class anyway, and this is safe to call. * Update actions/setup-python * Use `uv` in test CI + merge some dep installs The consequence is faster dependency installation. Merging some of the dependency installs helps avoid some packages being reinstalled from one line to the next. Additionally, CPU versions are specified when relevant, to avoid downloading CUDA stuff the CI can't use anyway. * Use `uv` in doc CI + merge some dep installs Similar rationale as for the test CI * Parallelize doc generation with Sphinx This does not affect the entire doc generation process but should allow some minor multithreading even with the 2-core CI workers. * Enable `uv` caching on the test CI * Enable `uv` caching on the docs CI * CTC-only training recipes for LibriSpeech (code from Samsung AI Cambridge) (speechbrain#2290) CTC-only pre-training of conformer and branchformer. --------- Co-authored-by: Shucong Zhang/Embedded AI /SRUK/Engineer/Samsung Electronics <s1.zhang@sruk-ccn4.eu.corp.samsungelectronics.net> Co-authored-by: Adel Moumen <adelmoumen.pro@gmail.com> Co-authored-by: Adel Moumen <88119391+Adel-Moumen@users.noreply.github.com> Co-authored-by: Parcollet Titouan <titouan.parcollet@univ-avignon.fr> * Update CommonVoice transformer recipes (code from Samsung AI Center Cambridge) (speechbrain#2465) * Update CV transformer recipes to match latest results with conformer. --------- Co-authored-by: Titouan Parcollet/Embedded AI /SRUK/Engineer/Samsung Electronics <t.parcollet@sruk-ccn4.eu.corp.samsungelectronics.net> Co-authored-by: Mirco Ravanelli <mirco.ravanelli@gmail.com> Co-authored-by: Adel Moumen <adelmoumen.pro@gmail.com> * Whisper improvements: flash attention, KV caching, lang_id, translation, training... (speechbrain#2450) Whisper improvements: - flash attention - kv caching - lang identifaction - translation - finetuning amelioration ... and more ... * Update README.md * precommit * update zed download link (speechbrain#2514) * `RelPosEncXL` refactor and precision fixes (speechbrain#2498) * Add `RelPosEncXL.make_pe`, rework precision handling * Rework RelPosEncXL output dtype selection * Fix in-place input normalization when using `sentence`/`speaker` norm (speechbrain#2504) * fix LOCAL_RANK to be RANK in if_main_process (speechbrain#2506) * Fix Separation and Enhancement recipes behavior when NaN encountered (speechbrain#2524) * Fix Separation and Enhancement recipes behavior when NaN encountered * Formatting using precommit hooks * Lock torch version in requirements.txt (speechbrain#2528) * Fix compatibility for torchaudio versions without `.io` (speechbrain#2532) This avoids having the Python interpreter attempt to resolve the type annotation directly. * fix docstrings * consistency tests - classification * consistency tests - classification * consistency tests - interpret * default to no wham * fix after tests pass * fix after tests pass * tests after that * fix consistency --------- Co-authored-by: asu <sdelang@sdelang.fr> Co-authored-by: Pooneh Mousavi <moosavi.pooneh@gmail.com> Co-authored-by: Mirco Ravanelli <mirco.ravanelli@gmail.com> Co-authored-by: shucongzhang <104781888+shucongzhang@users.noreply.github.com> Co-authored-by: Shucong Zhang/Embedded AI /SRUK/Engineer/Samsung Electronics <s1.zhang@sruk-ccn4.eu.corp.samsungelectronics.net> Co-authored-by: Adel Moumen <adelmoumen.pro@gmail.com> Co-authored-by: Adel Moumen <88119391+Adel-Moumen@users.noreply.github.com> Co-authored-by: Parcollet Titouan <titouan.parcollet@univ-avignon.fr> Co-authored-by: Parcollet Titouan <parcollet.titouan@gmail.com> Co-authored-by: Titouan Parcollet/Embedded AI /SRUK/Engineer/Samsung Electronics <t.parcollet@sruk-ccn4.eu.corp.samsungelectronics.net> Co-authored-by: Yingzhi WANG <41187612+BenoitWang@users.noreply.github.com> Co-authored-by: Peter Plantinga <plantinga.peter@protonmail.com> Co-authored-by: Séverin <123748182+SevKod@users.noreply.github.com>

…AI Cambridge) (speechbrain#2290)" This reverts commit d086cde.

… AI Cambridge) (speechbrain#2290)" This reverts commit 6d408c5.

Shucong Zhang/Embedded AI /SRUK/Engineer/Samsung Electronics and others added 3 commits December 5, 2023 12:10

librispeech train CTC from scratch scripts

5405fcf

Update README.md

ac0dfd5

Update README.md

47ac7bb

TParcollet requested review from Adel-Moumen and TParcollet December 5, 2023 14:19

TParcollet reviewed Dec 5, 2023

View reviewed changes

Shucong Zhang/Embedded AI /SRUK/Engineer/Samsung Electronics added 2 commits December 5, 2023 15:46

pre-commit tests

39b65f6

CI tests

5a40a97

mravanelli assigned shucongzhang Dec 11, 2023

Adel-Moumen reviewed Jan 18, 2024

View reviewed changes

recipes/LibriSpeech/ASR/CTC/train_from_scratch.py Outdated Show resolved Hide resolved

Adel-Moumen added 2 commits January 18, 2024 17:24

Merge branch 'develop' of https://github.com/speechbrain/speechbrain …

c613be2

…into LS_CTC

update code to comply with SB 1.0

9ab27e9

Adel-Moumen self-assigned this Jan 18, 2024

Adel-Moumen reviewed Jan 18, 2024

View reviewed changes

speechbrain/nnet/schedulers.py Show resolved Hide resolved

Adel-Moumen and others added 3 commits February 11, 2024 14:07

Merge branch 'develop' into LS_CTC

85c130c

update yamls with new data aug interface + dynamic batching

05f35de

dynamic batching

f2afd43

Adel-Moumen added 2 commits March 28, 2024 14:51

Merge branch 'develop' of https://github.com/speechbrain/speechbrain …

fb513a9

…into LS_CTC

update recipes

ba137be

Adel-Moumen and others added 10 commits March 28, 2024 20:49

update LibriSpeech test csv

4cbe99c

revert change

ceff583

revert change

15f7e3d

Title change.

7d10980

Merge remote-tracking branch 'speechbrain/develop' into LS_CTC

5dadc74

fix train.py val

edadc65

fix yamls val

275b134

fix naming

5171998

fix naming

6a7d00c

fix test

1170419

Adel-Moumen approved these changes Apr 18, 2024

View reviewed changes

Adel-Moumen merged commit d086cde into speechbrain:develop Apr 18, 2024
4 checks passed

asumagic added a commit to asumagic/speechbrain that referenced this pull request May 2, 2024

Revert "CTC-only training recipes for LibriSpeech (code from Samsung …

591b510

…AI Cambridge) (speechbrain#2290)" This reverts commit d086cde.

asumagic added a commit to asumagic/speechbrain that referenced this pull request May 6, 2024

Revert "CTC-only training recipes for LibriSpeech (code from Samsung …

6d408c5

…AI Cambridge) (speechbrain#2290)" This reverts commit d086cde.

asumagic added a commit to asumagic/speechbrain that referenced this pull request May 6, 2024

Reapply "CTC-only training recipes for LibriSpeech (code from Samsung…

cc556c6

… AI Cambridge) (speechbrain#2290)" This reverts commit 6d408c5.

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

CTC-only training recipes for LibriSpeech (code from Samsung AI Cambridge) #2290

CTC-only training recipes for LibriSpeech (code from Samsung AI Cambridge) #2290

shucongzhang commented Dec 5, 2023

TParcollet Dec 5, 2023

asumagic Dec 5, 2023

TParcollet commented Dec 5, 2023

shucongzhang commented Dec 5, 2023

mravanelli commented Dec 11, 2023

Adel-Moumen commented Jan 16, 2024

Adel-Moumen commented Jan 18, 2024

Adel-Moumen commented Jan 25, 2024

shucongzhang commented Jan 25, 2024 •

edited

Adel-Moumen commented Feb 11, 2024

shucongzhang commented Feb 19, 2024

Adel-Moumen commented Feb 20, 2024

TParcollet commented Feb 21, 2024

Adel-Moumen commented Apr 18, 2024

Adel-Moumen left a comment

shucongzhang commented Apr 22, 2024

CTC-only training recipes for LibriSpeech (code from Samsung AI Cambridge) #2290

CTC-only training recipes for LibriSpeech (code from Samsung AI Cambridge) #2290

Conversation

shucongzhang commented Dec 5, 2023

What does this PR do?

TParcollet Dec 5, 2023

Choose a reason for hiding this comment

asumagic Dec 5, 2023

Choose a reason for hiding this comment

TParcollet commented Dec 5, 2023

shucongzhang commented Dec 5, 2023

mravanelli commented Dec 11, 2023

Adel-Moumen commented Jan 16, 2024

Adel-Moumen commented Jan 18, 2024

Adel-Moumen commented Jan 25, 2024

shucongzhang commented Jan 25, 2024 • edited

Adel-Moumen commented Feb 11, 2024

shucongzhang commented Feb 19, 2024

Adel-Moumen commented Feb 20, 2024

TParcollet commented Feb 21, 2024

Adel-Moumen commented Apr 18, 2024

Tests

Adel-Moumen left a comment

Choose a reason for hiding this comment

shucongzhang commented Apr 22, 2024

shucongzhang commented Jan 25, 2024 •

edited