parambharat/whisper-base-ml

Whisper Base ML - Bharat Ramanathan

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

• Loss: 0.2456
• Wer: 48.0535

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: 5000
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Wer
0.7249	4.02	500	0.3786	70.8029
0.3377	4.02	1000	0.2477	56.2044
0.25	9.01	1500	0.2241	49.5134
0.2009	14.01	2000	0.2158	46.9586
0.1674	19.0	2500	0.2188	49.3917
0.142	23.02	3000	0.2194	49.6350
0.123	28.01	3500	0.2280	49.7567
0.1103	33.01	4000	0.2424	51.4599
0.0999	38.0	4500	0.2435	50.6083
0.0951	42.02	5000	0.2456	48.0535

Framework versions

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