Skip to content

DDP recipe tests and fixes #2130

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

Merged
mravanelli merged 6 commits into from
Sep 21, 2023
Merged

DDP recipe tests and fixes #2130

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
6 commits
Select commit Hold shift + click to select a range
331acdb
add notes on tests with non-default gpu
mravanelli Aug 14, 2023
091b3ce
fixed hard-coded device
mravanelli Aug 14, 2023
cc72c9e
fixed hard-coded device
mravanelli Aug 14, 2023
c60e606
fixed hard-coded device
mravanelli Aug 14, 2023
3298a29
Merge branch 'develop' into fixDDP
mravanelli Aug 18, 2023
9856912
Merge branch 'develop' into fixDDP
mravanelli Sep 21, 2023
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
4 changes: 1 addition & 3 deletions recipes/SLURP/direct/hparams/train.yaml
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -60,9 +60,7 @@ epoch_counter: !new:speechbrain.utils.epoch_loop.EpochCounter
limit: !ref <number_of_epochs>

# Models
asr_model: !apply:speechbrain.pretrained.EncoderDecoderASR.from_hparams
source: speechbrain/asr-crdnn-rnnlm-librispeech
run_opts: {"device":"cuda:0"}
asr_model_source: speechbrain/asr-crdnn-rnnlm-librispeech

slu_enc: !new:speechbrain.nnet.containers.Sequential
input_shape: [null, null, !ref <ASR_encoder_dim>]
Expand Down
8 changes: 8 additions & 0 deletions recipes/SLURP/direct/train.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -340,6 +340,14 @@ def text_pipeline(semantics):
run_on_main(hparams["pretrainer"].collect_files)
hparams["pretrainer"].load_collected(device=run_opts["device"])

# Download pretrained ASR model
from speechbrain.pretrained import EncoderDecoderASR

hparams["asr_model"] = EncoderDecoderASR.from_hparams(
source=hparams["asr_model_source"],
run_opts={"device": run_opts["device"]},
)

# Brain class initialization
slu_brain = SLU(
modules=hparams["modules"],
Expand Down
4 changes: 1 addition & 3 deletions recipes/fluent-speech-commands/direct/hparams/train.yaml
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -56,9 +56,7 @@ epoch_counter: !new:speechbrain.utils.epoch_loop.EpochCounter
limit: !ref <number_of_epochs>

# Models
asr_model: !apply:speechbrain.pretrained.EncoderDecoderASR.from_hparams
source: speechbrain/asr-crdnn-rnnlm-librispeech
run_opts: {"device":"cuda:0"}
asr_model_source: speechbrain/asr-crdnn-rnnlm-librispeech

slu_enc: !new:speechbrain.nnet.containers.Sequential
input_shape: [null, null, !ref <ASR_encoder_dim>]
Expand Down
8 changes: 8 additions & 0 deletions recipes/fluent-speech-commands/direct/train.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -322,6 +322,14 @@ def text_pipeline(semantics):
run_on_main(hparams["pretrainer"].collect_files)
hparams["pretrainer"].load_collected(device=run_opts["device"])

# Download pretrained ASR model
from speechbrain.pretrained import EncoderDecoderASR

hparams["asr_model"] = EncoderDecoderASR.from_hparams(
source=hparams["asr_model_source"],
run_opts={"device": run_opts["device"]},
)

# Brain class initialization
slu_brain = SLU(
modules=hparams["modules"],
Expand Down
4 changes: 1 addition & 3 deletions recipes/timers-and-such/direct/hparams/train.yaml
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -66,9 +66,7 @@ epoch_counter: !new:speechbrain.utils.epoch_loop.EpochCounter
limit: !ref <number_of_epochs>

# Models
asr_model: !apply:speechbrain.pretrained.EncoderDecoderASR.from_hparams
source: speechbrain/asr-crdnn-rnnlm-librispeech
run_opts: {"device":"cuda:0"}
asr_model_source: speechbrain/asr-crdnn-rnnlm-librispeech

slu_enc: !new:speechbrain.nnet.containers.Sequential
input_shape: [null, null, !ref <ASR_encoder_dim>]
Expand Down
8 changes: 8 additions & 0 deletions recipes/timers-and-such/direct/train.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -344,6 +344,14 @@ def text_pipeline(semantics):
run_on_main(hparams["pretrainer"].collect_files)
hparams["pretrainer"].load_collected(device=run_opts["device"])

# Download pretrained ASR model
from speechbrain.pretrained import EncoderDecoderASR

hparams["asr_model"] = EncoderDecoderASR.from_hparams(
source=hparams["asr_model_source"],
run_opts={"device": run_opts["device"]},
)

# Brain class initialization
slu_brain = SLU(
modules=hparams["modules"],
Expand Down
12 changes: 12 additions & 0 deletions tests/recipes/README.md
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -16,6 +16,18 @@ If you like to test for all recipes belonging to one dataset:
python -c 'from tests.utils.recipe_tests import run_recipe_tests; print("TEST FAILED!") if not(run_recipe_tests(filters_fields=["Dataset"], filters=[["CommonLanguage", "LibriSpeech"]], do_checks=False, run_opts="--device=cuda")) else print("TEST PASSED")'
```

You can run the recipe on the CPU just by setting the run_opts properly:
```
python -c 'from tests.utils.recipe_tests import run_recipe_tests; print("TEST FAILED!") if not(run_recipe_tests(filters_fields=["Dataset"], filters=[["CommonLanguage", "LibriSpeech"]], do_checks=False, run_opts="--device=cpu")) else print("TEST PASSED")'
```

In some cases, you might want to test the recipe on a non-default GPU (e.g, cuda:1). Thia helps detecting issues in recipes where the device was hard-coded. You can do that simply with:

```
python -c 'from tests.utils.recipe_tests import run_recipe_tests; print("TEST FAILED!") if not(run_recipe_tests(filters_fields=["Dataset"], filters=[["CommonLanguage", "LibriSpeech"]], do_checks=False, run_opts="--device=cuda:0")) else print("TEST PASSED")'
```


To target a specific recipe (here by its hparam yaml):
```
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/TIMIT/ASR/transducer/hparams/train_wav2vec.yaml"]], do_checks=False, run_opts="--device=cuda")) else print("TEST PASSED")'
Expand Down