shivkumarganesh/whisper-small-vi

Whisper Small Vi - Shiv Kumar Ganesh

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: 0.7220
• Wer: 46.6769

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: 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_steps: 500
• training_steps: 1200
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Wer
1.7433	1.02	100	1.6824	155.0559
0.5929	2.04	200	0.8475	55.5824
0.1188	3.05	300	0.6646	47.2801
0.0672	5.0	400	0.7099	61.3292
0.0317	6.02	500	0.6951	49.9013
0.0169	7.04	600	0.7658	62.8866
0.0089	8.06	700	0.6681	34.2509
0.004	10.01	800	0.6875	43.8364
0.0015	11.03	900	0.7129	46.8195
0.0011	12.04	1000	0.7194	47.4775
0.0011	13.06	1100	0.7217	46.1505
0.001	15.01	1200	0.7220	46.6769

Framework versions

• Transformers 4.26.0.dev0
• Pytorch 1.13.0+cu116
• Datasets 2.8.1.dev0
• Tokenizers 0.13.2