---
license: gpl-3.0
language:
- ca
tags:
- audio
- automatic-speech-recognition
- catalan
- mamba-models
- catalonia
- barcelona-supercomputing-center
model-index:
- name: ConMamba-small-ca
  results:
  - task:
      type: automatic-speech-recognition
      name: Automatic Speech Recognition
    dataset:
      name: 3cat-parla-asr
      type: "projecte-aina/3catparla_asr"
      config: ca
      split: test
    metrics:
    - type: wer
      value: 2.81
      name: Test WER
  - task:
      type: automatic-speech-recognition
      name: Automatic Speech Recognition
    dataset:
      name: parla_clean
      type: "projecte-aina/parlament_parla_v3"
      config: ca
      split: test
    metrics:
    - type: wer
      value: 4.81
      name: Test WER
  - task:
      type: automatic-speech-recognition
      name: Automatic Speech Recognition
    dataset:
      name: parla_other
      type: "projecte-aina/parlament_parla_v3"
      config: ca
      split: test
    metrics:
    - type: wer
      value: 10.91
      name: Test WER
  - task:
      type: automatic-speech-recognition
      name: Automatic Speech Recognition
    dataset:
      name: corts_clean_anon
      type: "projecte-aina/corts_valencianes_asr_a"
      config: ca
      split: test
    metrics:
    - type: wer
      value: 7.60
      name: Test WER
  - task:
      type: automatic-speech-recognition
      name: Automatic Speech Recognition
    dataset:
      name: corts_other_anon
      type: "projecte-aina/corts_valencianes_asr_a"
      config: ca
      split: test
    metrics:
    - type: wer
      value: 18.01
      name: Test WER
  - task:
      type: automatic-speech-recognition
      name: Automatic Speech Recognition
    dataset:
      name: central_male
      type: "projecte-aina/annotated_catalan_common_voice_v17"
      config: ca
      split: test
    metrics:
    - type: wer
      value: 11.63
      name: Test WER
  - task:
      type: automatic-speech-recognition
      name: Automatic Speech Recognition
    dataset:
      name: central_female
      type: "projecte-aina/annotated_catalan_common_voice_v17"
      config: ca
      split: test
    metrics:
    - type: wer
      value: 10.06
      name: Test WER
  - task:
      type: automatic-speech-recognition
      name: Automatic Speech Recognition
    dataset:
      name: valencia_male
      type: "projecte-aina/annotated_catalan_common_voice_v17"
      config: ca
      split: test
    metrics:
    - type: wer
      value: 11.24
      name: Test WER
  - task:
      type: automatic-speech-recognition
      name: Automatic Speech Recognition
    dataset:
      name: valencia_female
      type: "projecte-aina/annotated_catalan_common_voice_v17"
      config: ca
      split: test
    metrics:
    - type: wer
      value: 10.29
      name: Test WER
  - task:
      type: automatic-speech-recognition
      name: Automatic Speech Recognition
    dataset:
      name: balearic_male
      type: "projecte-aina/annotated_catalan_common_voice_v17"
      config: ca
      split: test
    metrics:
    - type: wer
      value: 13.58
      name: Test WER
  - task:
      type: automatic-speech-recognition
      name: Automatic Speech Recognition
    dataset:
      name: balearic_female
      type: "projecte-aina/annotated_catalan_common_voice_v17"
      config: ca
      split: test
    metrics:
    - type: wer
      value: 12.39
      name: Test WER
  - task:
      type: automatic-speech-recognition
      name: Automatic Speech Recognition
    dataset:
      name: northern_male
      type: "projecte-aina/annotated_catalan_common_voice_v17"
      config: ca
      split: test
    metrics:
    - type: wer
      value: 11.58
      name: Test WER
  - task:
      type: automatic-speech-recognition
      name: Automatic Speech Recognition
    dataset:
      name: northern_female
      type: "projecte-aina/annotated_catalan_common_voice_v17"
      config: ca
      split: test
    metrics:
    - type: wer
      value: 9.65
      name: Test WER
  - task:
      type: automatic-speech-recognition
      name: Automatic Speech Recognition
    dataset:
      name: northwestern_male
      type: "projecte-aina/annotated_catalan_common_voice_v17"
      config: ca
      split: test
    metrics:
    - type: wer
      value: 11.55
      name: Test WER
  - task:
      type: automatic-speech-recognition
      name: Automatic Speech Recognition
    dataset:
      name: northwestern_female
      type: "projecte-aina/annotated_catalan_common_voice_v17"
      config: ca
      split: test
    metrics:
    - type: wer
      value: 8.64
      name: Test WER
---
# ConMamba-small-ca-

## Table of Contents
<details>
<summary>Click to expand</summary>

- [Model Description](#model-description)
- [Intended Uses and Limitations](#intended-uses-and-limitations)
- [How to Get Started with the Model](#how-to-get-started-with-the-model)
- [Conversion Details](#conversion-details)
- [Citation](#citation)
- [Additional information](#additional-information)

</details>

## Summary

The **ConMamba-small-ca** is an acoustic model for Automatic Speech Recognition (ASR) in Catalan. It is based on the **ConMamba** architecture, which uses a **Mamba** (State Space Model) encoder augmented with convolutions for efficient sequence processing.

## Model Description

The **ConMamba-small-ca** model implements the **Convolution-augmented Mamba (ConMamba)** architecture, an adaptation of State Space Models (SSMs) designed to improve performance and efficiency in speech recognition tasks by integrating convolutional layers.

This model has been specifically trained for the **Catalan** language. The corpus used for training has 4929 hours.

## Intended Uses and Limitations

This model can be used for Automatic Speech Recognition (ASR) in Catalan. The model is intended to transcribe audio files in Catalan to plain text without punctuation.

## How to Get Started with the Model


### Installation

The implementation of the **ConMamba-small-ca** architecture often depends on specific libraries such as `mamba-ssm` and `causal-conv1d`. It is recommended to follow the installation steps from the original Mamba ASR repository:

1.  **Create a virtual environment (mamba_asr, for example):**
    ```bash
    conda create --name mamba_asr python=3.9
    conda activate mamba_asr
    ```
2.  **Install dependencies:**
    ```bash
    clone github https://github.com/langtech-bsc/ConMamba_ASR
    cd ConMamba_ASR
    pip install -r requirements.txt
    # Make sure that the versions of torch, torchaudio, causal-conv1d, and mamba-ssm are compatible with your hardware.
    ```

### For Inference

Inference is performed using the dedicated `run_inference.py` script provided within the repository.

1.  **Define Paths:** Set the paths for the repository, the input audio, and the specific configuration file for inference.
2.  **Execute Inference:** Run the script using the defined paths.

```bash
# Define your paths
REPO_PATH="/path/to/ConMamba-ASR" 
AUDIO="/path/to/your/audio.wav"
HPARAMS="conmambamamba_debug_catalan_small_1k_unigram_inference.yaml" # Use your specific inference YAML

# Execute inference script
python  $REPO_PATH/run_inference.py \
  --hparams $HPARAMS \
  --audio $AUDIO
```

**Dev Result - WER**: 8.6


## Training Details

### Training data

The model was trained for a total of 4929 hours. Including:

- [Parlament-Parla-v3](https://huggingface.co/datasets/projecte-aina/parlament_parla_v3) (Only the anonymized version of the dataset is public. We trained the model with the non-anonymized version.)
- [Corts Valencianes](https://huggingface.co/datasets/projecte-aina/corts_valencianes_asr_a) (Only the anonymized version of the dataset is public. We trained the model with the non-anonymized version.)
- [3cat](https://www.isca-archive.org/iberspeech_2024/hernandezmena24_iberspeech.pdf) 
- [IB3](https://huggingface.co/datasets/projecte-aina/ib3_ca_asr) (The datasets will be made accessible shortly.)
- [Common Voice ca 17 Benchmark](https://huggingface.co/datasets/projecte-aina/commonvoice_benchmark_catalan_accents)


### Training Hyperparameters

* Training hours: 4929
* language: catalan
* number_of_epochs: 110
* batch_size: 30
* ctc_weight: 0.6
* grad_accumulation_factor: 1
* max_grad_norm: 5.0
* loss_reduction: 'batchmean'
* sorting: random
* num_workers: 8
* precision: bf16
* avg_checkpoints: 10 
* lr_adam: 0.001 



## Citation
If this model contributes to your research, please cite the work:
```bibtex
@inproceedings{zevallos2025conmambasmallca,
  title={Evaluating High-Performance and Lightweight ASR Systems for Catalan},
  author={Zevallos, Rodolfo}
  organization={Barcelona Supercomputing Center},
  year={2025}
}
```

## Additional Information

### Author

The model was trained during September (2025) in the [Language Technologies Laboratory](https://huggingface.co/BSC-LT) of the [Barcelona Supercomputing Center](https://www.bsc.es/) by [Rodolfo Zevallos](https://huggingface.co/rjzevallos).

### Contact
For further information, please send an email to <bsc-lt@bsc.es>.

### Copyright
Copyright(c) 2025 by Language Technologies Laboratory, Barcelona Supercomputing Center.

### License
[GPL-3.0](https://www.gnu.org/licenses/gpl-3.0.html)

### Funding
This work has been promoted and financed by the Generalitat de Catalunya through the [Aina project](https://projecteaina.cat/).

The conversion of the model was possible thanks to the computing time provided by [Barcelona Supercomputing Center](https://www.bsc.es/) through MareNostrum 5.