README.md

metadata

library_name: transformers
language:
  - sn
license: cc-by-nc-4.0
base_model: facebook/mms-1b-all
tags:
  - generated_from_trainer
datasets:
  - DigitalUmuganda_Afrivoice/Shona
metrics:
  - wer
model-index:
  - name: facebook/mms-1b-all
    results:
      - task:
          name: Automatic Speech Recognition
          type: automatic-speech-recognition
        dataset:
          name: DigitalUmuganda
          type: DigitalUmuganda_Afrivoice/Shona
        metrics:
          - name: Wer
            type: wer
            value: 0.2619660819281792

facebook/mms-1b-all

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

• Loss: 0.2146
• Model Preparation Time: 0.0203
• Wer: 0.2620
• Cer: 0.0482

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: 0.0003
• train_batch_size: 4
• eval_batch_size: 8
• seed: 42
• gradient_accumulation_steps: 4
• total_train_batch_size: 16
• optimizer: Use adamw_torch with betas=(0.9,0.999) and epsilon=1e-08 and optimizer_args=No additional optimizer arguments
• lr_scheduler_type: linear
• lr_scheduler_warmup_steps: 100
• num_epochs: 150
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Model Preparation Time	Wer	Cer
9.7791	1.0	546	0.2166	0.0203	0.3124	0.0531
0.9395	2.0	1092	0.2028	0.0203	0.2937	0.0494
0.8823	3.0	1638	0.1954	0.0203	0.2847	0.0479
0.8476	4.0	2184	0.1902	0.0203	0.2733	0.0460
0.8178	5.0	2730	0.1886	0.0203	0.2677	0.0451
0.7954	6.0	3276	0.1859	0.0203	0.2673	0.0443
0.7758	7.0	3822	0.1829	0.0203	0.2641	0.0440
0.7604	8.0	4368	0.1824	0.0203	0.2576	0.0430
0.746	9.0	4914	0.1821	0.0203	0.2678	0.0438
0.7375	10.0	5460	0.1802	0.0203	0.2609	0.0433
0.7223	11.0	6006	0.1800	0.0203	0.2572	0.0428
0.714	12.0	6552	0.1782	0.0203	0.2564	0.0426
0.7007	13.0	7098	0.1753	0.0203	0.2552	0.0421
0.6923	14.0	7644	0.1764	0.0203	0.2560	0.0421
0.6842	15.0	8190	0.1769	0.0203	0.2563	0.0422
0.6742	16.0	8736	0.1733	0.0203	0.2483	0.0410
0.6655	17.0	9282	0.1767	0.0203	0.2506	0.0415
0.6594	18.0	9828	0.1750	0.0203	0.2503	0.0412
0.651	19.0	10374	0.1733	0.0203	0.2513	0.0413
0.6435	20.0	10920	0.1767	0.0203	0.2458	0.0410
0.637	21.0	11466	0.1773	0.0203	0.2449	0.0407
0.6306	22.0	12012	0.1750	0.0203	0.2488	0.0411
0.6216	23.0	12558	0.1738	0.0203	0.2507	0.0412
0.6174	24.0	13104	0.1788	0.0203	0.2450	0.0407
0.6093	25.0	13650	0.1802	0.0203	0.2479	0.0408
0.6033	26.0	14196	0.1801	0.0203	0.2463	0.0407
0.5932	27.0	14742	0.1793	0.0203	0.2462	0.0407
0.5885	28.0	15288	0.1780	0.0203	0.2465	0.0407
0.5823	29.0	15834	0.1802	0.0203	0.2465	0.0404
0.5749	30.0	16380	0.1794	0.0203	0.2465	0.0407

Framework versions

• Transformers 4.47.0.dev0
• Pytorch 2.1.0+cu118
• Datasets 3.0.2
• Tokenizers 0.20.1