metadata

language:
  - rm-sursilv
license: apache-2.0
tags:
  - automatic-speech-recognition
  - mozilla-foundation/common_voice_7_0
  - generated_from_trainer
  - rm-sursilv
  - robust-speech-event
  - model_for_talk
  - hf-asr-leaderboard
datasets:
  - mozilla-foundation/common_voice_7_0
model-index:
  - name: XLS-R-300M - Romansh Sursilvan
    results:
      - task:
          name: Automatic Speech Recognition
          type: automatic-speech-recognition
        dataset:
          name: Common Voice 7
          type: mozilla-foundation/common_voice_7_0
          args: rm-sursilv
        metrics:
          - name: Test WER
            type: wer
            value: 19.816
          - name: Test CER
            type: cer
            value: 4.153

wav2vec2-large-xls-r-300m-romansh-sursilvan

This model is a fine-tuned version of facebook/wav2vec2-xls-r-300m on the MOZILLA-FOUNDATION/COMMON_VOICE_7_0 - RM-SURSILV dataset. It achieves the following results on the evaluation set:

Loss: 0.2163
Wer: 0.1981

Model description

More information needed

Intended uses & limitations

More information needed

Training and evaluation data

More information needed

Training procedure

Training hyperparameters

The following hyperparameters were used during training:

learning_rate: 7e-05
train_batch_size: 32
eval_batch_size: 1
seed: 42
optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
lr_scheduler_type: linear
lr_scheduler_warmup_steps: 2000
num_epochs: 120.0
mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Wer
1.1004	23.81	2000	0.3710	0.4191
0.7002	47.62	4000	0.2342	0.2562
0.5573	71.43	6000	0.2175	0.2177
0.4799	95.24	8000	0.2109	0.1987
0.4511	119.05	10000	0.2164	0.1975

Framework versions

Transformers 4.16.0.dev0
Pytorch 1.10.1+cu102
Datasets 1.17.1.dev0
Tokenizers 0.11.0