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

adding jenkins test for speech_to_text_aed model #8368

Merged
merged 2 commits into from
Feb 9, 2024
+43 −1
Merged

adding jenkins test for speech_to_text_aed model #8368

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
2 commits
Select commit Hold shift + click to select a range
ee90b6d
bug fix in fast-conformer-aed.yaml and adding jenkins test for speech…
Feb 7, 2024
62ad4e0
Merge branch 'r1.23.0' into r123_canary_jenkins
titu1994 Feb 7, 2024
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
42 changes: 42 additions & 0 deletions Jenkinsfile
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -607,6 +607,48 @@ pipeline {

}

stage('L2: Speech to Text AED') {
when {
anyOf {
branch 'r1.23.0'
changeRequest target: 'r1.23.0'
}
}
steps {
sh 'python examples/asr/speech_multitask/speech_to_text_aed.py \
model.prompt_format=canary \
model.model_defaults.asr_enc_hidden=256 \
model.model_defaults.lm_dec_hidden=256 \
model.encoder.n_layers=12 \
model.transf_encoder.num_layers=0 \
model.transf_decoder.config_dict.num_layers=12 \
model.train_ds.manifest_filepath=/home/TestData/asr/manifests/canary/an4_canary_train.json \
++model.train_ds.is_tarred=false \
Copy link
Collaborator

Choose a reason for hiding this comment

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

Hmm is it necessary for Lhotse to use is_tarred true?

Copy link
Collaborator

@pzelasko pzelasko Feb 7, 2024 •
edited
Loading

Choose a reason for hiding this comment

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

No, it auto-infers the right type. This arg is ignored. I missed that this arg is here earlier actually...

Copy link
Collaborator

Choose a reason for hiding this comment

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

Can we remove it from config and code in main in another PR? After release that is

model.train_ds.batch_duration=60 \
+model.train_ds.text_field="answer" \
Copy link
Collaborator

Choose a reason for hiding this comment

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

In the manifest, why isnt text the key for the transcript - why is it now "answer"?

Copy link
Collaborator

Choose a reason for hiding this comment

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

AFAIK that was an already established convention in speech lm expts, in order to support these cases we allow users to choose whichever keys they want to read text and lang from.

Copy link
Collaborator

Choose a reason for hiding this comment

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

Ok - but why is the default not text as has been nemos convention for the past 4 years? If this is arbitrary, please make it revert to the nemo default of reading text as the text field, whatever it may be. If user wants to change it, that's their prerogative, but nemos defaults are set as done before.

+model.train_ds.lang_field="target_lang" \
model.validation_ds.manifest_filepath=/home/TestData/asr/manifests/canary/an4_canary_val.json \
+model.validation_ds.text_field="answer" \
+model.validation_ds.lang_field="target_lang" \
model.test_ds.manifest_filepath=/home/TestData/asr/manifests/canary/an4_canary_val.json \
+model.test_ds.text_field="answer" \
+model.test_ds.lang_field="target_lang" \
model.tokenizer.langs.spl_tokens.dir=/home/TestData/asr_tokenizers/canary/canary_spl_tokenizer_v32 \
model.tokenizer.langs.spl_tokens.type="bpe" \
model.tokenizer.langs.en.dir=/home/TestData/asr_tokenizers/canary/en/tokenizer_spe_bpe_v1024_max_4 \
model.tokenizer.langs.en.type=bpe \
++model.tokenizer.langs.es.dir=/home/TestData/asr_tokenizers/canary/es/tokenizer_spe_bpe_v1024_max_4 \
++model.tokenizer.langs.es.type=bpe \
trainer.devices=[0] \
trainer.accelerator="gpu" \
+trainer.use_distributed_sampler=false \
+trainer.fast_dev_run=True \
exp_manager.exp_dir=examples/asr/speech_to_text_aed_results'
sh 'rm -rf examples/asr/speech_to_text_results'
}

}

stage('L2: Speaker dev run') {
when {
anyOf {
Expand Down
2 changes: 1 addition & 1 deletion examples/asr/conf/speech_multitask/fast-conformer_aed.yaml
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -38,7 +38,7 @@ model:
# https://github.com/NVIDIA/NeMo/blob/main/docs/source/asr/datasets.rst#lhotse-dataloading
# You can also check the following configuration dataclass:
# https://github.com/NVIDIA/NeMo/blob/main/nemo/collections/common/data/lhotse/dataloader.py#L36
batch_size: None
batch_size: null
batch_duration: 360
quadratic_duration: 15
use_bucketing: True
Expand Down
Loading