calista-whisper-base.en-drop-002-production

This model is a fine-tuned version of openai/whisper-base.en on the audiofolder dataset. It achieves the following results on the evaluation set:

Model description

More information needed

More information needed

More information needed

The following hyperparameters were used during training:

learning_rate: 1e-05
train_batch_size: 16
eval_batch_size: 16
seed: 42
optimizer: Use OptimizerNames.ADAMW_TORCH_FUSED with betas=(0.9,0.999) and epsilon=1e-08 and optimizer_args=No additional optimizer arguments
lr_scheduler_type: constant_with_warmup
lr_scheduler_warmup_steps: 50
training_steps: 500
mixed_precision_training: Native AMP

Training Loss	Epoch	Step	Validation Loss	Wer Ortho	Wer
No log	0.5556	20	1.7441	93.6508	57.1429
2.9422	1.1111	40	0.0001	0.0	0.0
0.0495	1.6667	60	0.0000	0.0	0.0
0.0000	2.2222	80	0.0000	0.0	0.0
0.0000	2.7778	100	0.0000	0.0	0.0
0.0000	3.3333	120	0.0000	0.0	0.0
0.0000	3.8889	140	0.0000	0.0	0.0
0.0000	4.4444	160	0.0000	0.0	0.0
0.0000	5.0	180	0.0000	0.0	0.0
0.0000	5.5556	200	0.0000	0.0	0.0
0.0000	6.1111	220	0.0000	0.0	0.0
0.0000	6.6667	240	0.0000	0.0	0.0
0.0000	7.2222	260	0.0000	0.0	0.0
0.0000	7.7778	280	0.0000	0.0	0.0
0.0000	8.3333	300	0.0000	0.0	0.0
0.0000	8.8889	320	0.0000	0.0	0.0
0.0000	9.4444	340	0.0000	0.0	0.0
0.0000	10.0	360	0.0000	0.0	0.0
0.0000	10.5556	380	0.0000	0.0	0.0
0.0000	11.1111	400	0.0000	0.0	0.0
0.0000	11.6667	420	0.0000	0.0	0.0
0.0000	12.2222	440	0.0000	0.0	0.0
0.0000	12.7778	460	0.0000	0.0	0.0
0.0000	13.3333	480	0.0000	0.0	0.0
0.0000	13.8889	500	0.0000	0.0	0.0

Loading…