Implementation:Speechbrain Speechbrain Hparams AISHELL1 Transformer

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

Overview

Hyperparameter configuration for Transformer ASR training on the AISHELL-1 dataset.

Description

HyperPyYAML configuration file that defines the model architecture, training schedule, and data processing pipeline for end-to-end ASR with a Transformer encoder-decoder on AISHELL-1 Mandarin Chinese data. The model uses CTC + KLdiv (label smoothing) losses with BPE unigram tokenization. It supports waveform and noise augmentation, dynamic batching, and configurable precision (fp16/bf16/fp32).

Usage

Pass this YAML file as the first argument to the corresponding training script.

Code Reference

Source Location

• Repository: SpeechBrain
• File: recipes/AISHELL-1/ASR/transformer/hparams/train_ASR_transformer.yaml

Key Parameters

seed: 8886
number_of_epochs: 50
batch_size: 8
ctc_weight: 0.3
grad_accumulation_factor: 4
loss_reduction: 'batchmean'
sorting: random
avg_checkpoints: 10
precision: fp32

# Feature parameters
sample_rate: 16000
n_fft: 400
n_mels: 80

# Stages related parameters
stage_one_epochs: 40
lr_adam: 1.0
lr_sgd: 0.000025

I/O Contract

Inputs

Name	Type	Required	Description
--data_folder	str	Yes	Path to AISHELL-1 dataset

Outputs

Name	Type	Description
Instantiated objects	Python objects	Model, optimizer, scheduler, etc.

Usage Examples

python train.py hparams/train_ASR_transformer.yaml --data_folder /path/to/aishell

Related Pages

• Principle:Speechbrain_Speechbrain_HyperPyYAML_Configuration