whisper-nm-clp

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: 0.3038
• Wer: 37.9421

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: 0.0004
• train_batch_size: 8
• eval_batch_size: 8
• seed: 42
• gradient_accumulation_steps: 2
• total_train_batch_size: 16
• optimizer: Use adamw_torch with betas=(0.9,0.999) and epsilon=1e-08 and optimizer_args=No additional optimizer arguments
• lr_scheduler_type: linear
• lr_scheduler_warmup_steps: 132
• num_epochs: 50
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Wer
No log	2.4390	100	0.3751	228.2958
0.9619	4.8780	200	0.4282	51.4469
0.9619	7.3171	300	0.2953	46.3023
0.2051	9.7561	400	0.3202	49.3569
0.2051	12.1951	500	0.3298	45.9807
0.1113	14.6341	600	0.2517	44.6945
0.1113	17.0732	700	0.3117	47.4277
0.0846	19.5122	800	0.3111	113.5048
0.0846	21.9512	900	0.3486	47.9100
0.0584	24.3902	1000	0.3190	48.3923
0.0584	26.8293	1100	0.2865	37.2990
0.0264	29.2683	1200	0.4153	47.9100
0.0264	31.7073	1300	0.3360	43.8907
0.0122	34.1463	1400	0.2909	36.8167
0.0122	36.5854	1500	0.3393	38.4244
0.004	39.0244	1600	0.3064	37.9421
0.004	41.4634	1700	0.2980	34.7267
0.0006	43.9024	1800	0.3004	36.8167
0.0006	46.3415	1900	0.3034	37.9421
0.0002	48.7805	2000	0.3038	37.9421

Framework versions

• Transformers 4.47.0.dev0
• Pytorch 2.4.0
• Datasets 3.0.1
• Tokenizers 0.20.0