whisper-nmcpc-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: 0.2688
• Wer: 12.1277

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: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: linear
• lr_scheduler_warmup_steps: 132
• num_epochs: 30
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Wer
1.3625	1.5625	100	0.2926	234.4681
0.4183	3.125	200	0.2864	124.2553
0.25	4.6875	300	0.2957	39.5745
0.1874	6.25	400	0.2477	31.0638
0.1471	7.8125	500	0.2908	35.9574
0.1121	9.375	600	0.3177	35.9574
0.1028	10.9375	700	0.3177	30.4255
0.0843	12.5	800	0.1921	19.5745
0.0633	14.0625	900	0.2242	18.2979
0.0415	15.625	1000	0.2623	23.4043
0.0377	17.1875	1100	0.2313	20.4255
0.0266	18.75	1200	0.3112	33.4043
0.0177	20.3125	1300	0.2643	17.0213
0.0097	21.875	1400	0.2476	17.0213
0.0017	23.4375	1500	0.2722	11.4894
0.001	25.0	1600	0.2661	12.1277
0.0006	26.5625	1700	0.2696	12.1277
0.0003	28.125	1800	0.2681	12.1277
0.0001	29.6875	1900	0.2688	12.1277

Framework versions

• Transformers 4.45.0.dev0
• Pytorch 2.4.0
• Datasets 3.0.0
• Tokenizers 0.19.1