bhuvanesh25/whis-tam-small

Whisper Small Ta - Bharat Ramanathan (Kudos to him for developing it)

This is a copy of his model for academic purpose.

This model is a fine-tuned version of openai/whisper-small on the None dataset. It achieves the following results on the evaluation set:

• Loss: 0.1803
• Wer: 17.1456

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: 32
• eval_batch_size: 16
• seed: 42
• gradient_accumulation_steps: 2
• 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: 500
• training_steps: 5000
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Wer
0.3374	0.1	500	0.2579	23.3804
0.29	0.2	1000	0.2260	20.9937
0.2522	0.3	1500	0.2139	20.0682
0.2338	0.4	2000	0.2025	19.6785
0.223	0.5	2500	0.1979	18.3147
0.211	0.6	3000	0.1927	17.8276
0.2032	0.7	3500	0.1865	17.3892
0.1978	0.8	4000	0.1839	17.5353
0.1972	0.9	4500	0.1812	17.0969
0.1894	1.0	5000	0.1803	17.1456

Framework versions

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