Implementation:Speechbrain Speechbrain Train PeoplesSpeech

Knowledge Sources	SpeechBrain
Domains	ASR, Training
Last Updated	2026-02-09 00:00 GMT

Overview

Concrete tool for training a Conformer ASR model on the People's Speech dataset provided by the SpeechBrain library.

Description

This recipe defines the ASR class (subclass of sb.core.Brain) for Conformer-based speech recognition on the People's Speech dataset (28,000 hours). The architecture uses a CNN frontend and a Conformer encoder with a Transformer decoder. Joint CTC/attention training with label smoothing is used. Feature augmentation is supported during training. Beam search decoding is applied at validation and test stages. BPE tokenization via SentencePiece is used for subword units.

Usage

Use this recipe to train a Conformer ASR model with CTC/attention joint decoding on the People's Speech dataset. Requires the corresponding hyperparameter YAML file and data preparation script.

Code Reference

Source Location

• Repository: SpeechBrain
• File: recipes/PeoplesSpeech/ASR/transformer/train.py

Signature

class ASR(sb.core.Brain):
    def compute_forward(self, batch, stage):
        """Forward computations from the waveform batches to the output probabilities."""
        ...
    def compute_objectives(self, predictions, batch, stage):
        """Computes the loss (CTC+NLL) given predictions and targets."""
        ...

Import

# Run as recipe script
python recipes/PeoplesSpeech/ASR/transformer/train.py hparams/conformer_large.yaml --data_folder /path/to/PeoplesSpeech

I/O Contract

Inputs

Name	Type	Required	Description
batch.sig	torch.Tensor	Yes	Input waveform signal
batch.tokens_bos	torch.Tensor	Yes	Target token sequence with BOS prefix
batch.tokens_eos	torch.Tensor	Yes	Target token sequence with EOS suffix
batch.tokens	torch.Tensor	Yes	Target token sequence (for CTC)

Outputs

Name	Type	Description
p_ctc	torch.Tensor	CTC log-probabilities from encoder
p_seq	torch.Tensor	Seq2seq log-probabilities from Transformer decoder
wav_lens	torch.Tensor	Relative waveform lengths
hyps	list	Beam search hypotheses (at validation/test)

Usage Examples

python train.py hparams/conformer_large.yaml --data_folder /path/to/PeoplesSpeech

Related Pages

• Principle:Speechbrain_Speechbrain_Transformer_ASR_Training