README.md

metadata

language:
  - ml
license: apache-2.0
tags:
  - whisper-event
  - generated_from_trainer
datasets:
  - thennal/imasc
metrics:
  - wer
model-index:
  - name: Whisper Small Ml - IMaSC
    results:
      - task:
          name: Automatic Speech Recognition
          type: automatic-speech-recognition
        dataset:
          name: ICFOSS Malayalam Speech Corpus
          type: thennal/imasc
          config: ml
          split: test
          args: ml
        metrics:
          - name: Wer
            type: wer
            value: 75.40229885057471

Whisper Small Ml - IMaSC

This model is a fine-tuned version of openai/whisper-small on the ICFOSS Malayalam Speech Corpus dataset. It achieves the following results on the evaluation set:

• Loss: 0.2750
• Wer: 75.4023
• Cer: 20.0050

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: 1e-05
• train_batch_size: 64
• eval_batch_size: 32
• seed: 42
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: linear
• lr_scheduler_warmup_steps: 500
• training_steps: 2000
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Wer	Cer
0.0678	0.93	500	0.2756	86.6667	31.1467
0.0342	1.86	1000	0.2424	73.7931	20.3305
0.0192	2.78	1500	0.2615	74.7126	19.8297
0.0107	3.71	2000	0.2750	75.4023	20.0050

Framework versions

• Transformers 4.26.0.dev0
• Pytorch 1.13.0+cu117
• Datasets 2.7.1.dev0
• Tokenizers 0.13.2