DrishtiSharma/whisper-medium-assamese

Whisper Small Assamese - Drishti Sharma

This model is a fine-tuned version of openai/whisper-small on the Common Voice 11.0 dataset. It achieves the following results on the evaluation set:

• Loss: 0.5164
• Wer: 30.6502

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: 16
• eval_batch_size: 8
• seed: 42
• gradient_accumulation_steps: 4
• total_train_batch_size: 64
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: linear
• lr_scheduler_warmup_steps: 150
• training_steps: 500
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Wer
1.1539	3.02	50	0.7835	89.0863
0.2089	7.0	100	0.3041	37.7378
0.0428	10.02	150	0.3760	33.9118
0.0141	14.01	200	0.4400	31.6538
0.0059	17.02	250	0.4472	31.2774
0.0022	21.01	300	0.4696	31.0475
0.0005	24.03	350	0.5032	31.2983
0.0003	28.02	400	0.5051	30.7129
0.0003	32.0	450	0.5137	30.7338
0.0003	35.02	500	0.5164	30.6502

Framework versions

• Transformers 4.26.0.dev0
• Pytorch 1.13.0+cu116
• Datasets 2.7.1.dev0
• Tokenizers 0.13.2