metadata

library_name: transformers
license: cc-by-nc-4.0
base_model: facebook/mms-1b-all
tags:
  - generated_from_trainer
datasets:
  - common_voice_17_0
metrics:
  - wer
  - bleu
model-index:
  - name: wav2vec2-mms-1b-CV17.0-training_set_variations
    results:
      - task:
          name: Automatic Speech Recognition
          type: automatic-speech-recognition
        dataset:
          name: common_voice_17_0
          type: common_voice_17_0
          config: ta
          split: >-
            validation[:5%]+validation[20%:25%]+validation[60%:65%]+validation[90%:]
          args: ta
        metrics:
          - name: Wer
            type: wer
            value: 0.5119016249451032
          - name: Bleu
            type: bleu
            value: 0.24178033350654143

wav2vec2-mms-1b-CV17.0-training_set_variations

This model is a fine-tuned version of facebook/mms-1b-all on the common_voice_17_0 dataset. It achieves the following results on the evaluation set:

Loss: 0.4958
Wer: 0.5119
Cer: 0.0973
Bleu: 0.2418

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: 0.001
train_batch_size: 16
eval_batch_size: 8
seed: 42
gradient_accumulation_steps: 2
total_train_batch_size: 32
optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
lr_scheduler_type: linear
lr_scheduler_warmup_ratio: 0.15
training_steps: 2000
mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Wer	Cer	Bleu
6.9079	100.0	100	0.3463	0.4393	0.0786	0.3177
0.0658	200.0	200	0.3912	0.4538	0.0839	0.3158
0.0346	300.0	300	0.4707	0.5046	0.0947	0.2477
0.0265	400.0	400	0.4906	0.5137	0.0967	0.2393
0.0184	500.0	500	0.5407	0.5240	0.1005	0.2301
0.0158	600.0	600	0.4958	0.5119	0.0973	0.2418

Framework versions

Transformers 4.44.2
Pytorch 2.4.1+cu121
Datasets 3.0.0
Tokenizers 0.19.1