4

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: 2.0909
• Wer: 65.6470

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: 5e-07
• train_batch_size: 8
• eval_batch_size: 8
• seed: 42
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: linear
• lr_scheduler_warmup_steps: 10
• num_epochs: 1
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Wer
4.2263	0.04	10	4.0304	15.6470
3.8803	0.08	20	3.6564	15.4712
3.7847	0.13	30	3.2816	14.6624
3.4068	0.17	40	3.0934	14.3108
3.0547	0.21	50	2.9635	14.0647
3.0245	0.25	60	2.8523	17.7918
3.3518	0.29	70	2.7638	18.8467
3.105	0.34	80	2.6917	20.8509
2.8325	0.38	90	2.6234	21.3783
2.8667	0.42	100	2.5582	21.4838
2.7528	0.46	110	2.4994	21.2377
2.8502	0.5	120	2.4430	28.0591
2.5623	0.55	130	2.3898	30.5556
2.877	0.59	140	2.3409	36.0056
2.6492	0.63	150	2.2965	45.1828
2.4119	0.67	160	2.2554	60.0563
2.6273	0.71	170	2.2175	73.8045
2.4091	0.76	180	2.1841	70.2883
2.2931	0.8	190	2.1558	74.8242
2.297	0.84	200	2.1324	71.0970
2.1893	0.88	210	2.1133	72.0113
2.3408	0.92	220	2.0994	70.8861
2.2892	0.97	230	2.0909	65.6470

Framework versions

• Transformers 4.38.0.dev0
• Pytorch 2.1.0+cu121
• Datasets 2.16.1
• Tokenizers 0.15.1