anuragshas/whisper-small-as

Whisper Small Assamese

This model is a fine-tuned version of openai/whisper-small on the mozilla-foundation/common_voice_11_0 as dataset. It achieves the following results on the evaluation set:

• Loss: 0.6033
• Wer: 35.4990

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: 40
• training_steps: 400
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Wer
1.0676	3.01	50	0.6487	62.5338
0.2252	6.03	100	0.3487	36.4916
0.0787	9.04	150	0.3934	35.6434
0.0178	13.01	200	0.5057	36.0043
0.0048	16.02	250	0.5589	35.8239
0.0022	19.04	300	0.5882	35.7336
0.0015	23.01	350	0.5985	35.5712
0.0013	26.02	400	0.6033	35.4990

Framework versions

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