README.md

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
          args: ta
        metrics:
          - name: Wer
            type: wer
            value: 1.0759957831049183
          - name: Bleu
            type: bleu
            value: 0

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: 6.3100
• Wer: 1.0760
• Cer: 0.7242
• Bleu: 0.0

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
13.5071	6.25	50	7.0048	1.0001	0.9823	0.0
5.6893	12.5	100	5.2167	1.0000	0.9089	0.0
4.3569	18.75	150	4.3516	1.0125	0.8661	0.0
3.5211	25.0	200	3.5506	1.0269	0.8318	0.0
3.2359	31.25	250	3.4822	1.0039	0.8458	0.0
3.1004	37.5	300	3.5209	1.0086	0.8185	0.0
2.9953	43.75	350	3.5498	1.0121	0.8104	0.0
2.779	50.0	400	3.6218	1.0170	0.7777	0.0
2.6019	56.25	450	4.0907	1.0134	0.7504	0.0
2.4589	62.5	500	3.8633	1.0287	0.7476	0.0
2.2318	68.75	550	3.7976	1.0367	0.7239	0.0
2.0073	75.0	600	4.0050	1.0234	0.7288	0.0
1.7416	81.25	650	4.2022	1.0126	0.7231	0.0
1.5467	87.5	700	4.4087	1.0469	0.7197	0.0
1.3716	93.75	750	4.5391	1.0471	0.7185	0.0
1.2237	100.0	800	4.8405	1.0398	0.7152	0.0
1.1216	106.25	850	5.0209	1.0421	0.7160	0.0
1.0274	112.5	900	5.0349	1.0669	0.7160	0.0
0.9325	118.75	950	5.2384	1.0500	0.7198	0.0
0.8486	125.0	1000	5.3843	1.0614	0.7131	0.0
0.7996	131.25	1050	5.4558	1.0622	0.7181	0.0
0.7479	137.5	1100	5.6746	1.0622	0.7205	0.0
0.6972	143.75	1150	5.7069	1.1182	0.7218	0.0
0.6596	150.0	1200	5.7678	1.0883	0.7221	0.0
0.6213	156.25	1250	5.9645	1.0700	0.7191	0.0
0.5905	162.5	1300	6.0098	1.0970	0.7210	0.0
0.5532	168.75	1350	6.0379	1.0981	0.7235	0.0
0.5336	175.0	1400	6.2079	1.0564	0.7210	0.0
0.4955	181.25	1450	6.1618	1.0717	0.7225	0.0
0.4851	187.5	1500	6.3100	1.0760	0.7242	0.0

Framework versions

• Transformers 4.44.2
• Pytorch 2.4.1+cu121
• Datasets 3.0.0
• Tokenizers 0.19.1