whisper-small-mn-6

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: 0.3296
• Wer: 35.8860
• Cer: 13.3108

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: 32
• 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: 15000
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Wer	Cer
0.3774	0.8	1000	0.4319	53.2773	19.6627
0.2926	1.61	2000	0.3493	40.4960	15.0214
0.2331	2.41	3000	0.3346	39.1741	14.7689
0.1636	3.22	4000	0.3287	36.9237	13.7943
0.1157	4.02	5000	0.3296	35.8860	13.3108
0.1271	4.82	6000	0.3422	36.0717	13.5702
0.0879	5.63	7000	0.3661	36.6943	13.7780
0.0574	6.43	8000	0.3884	36.4595	13.5015
0.036	7.23	9000	0.4128	37.1422	13.8424
0.0229	8.04	10000	0.4321	36.8582	13.8475
0.0241	8.84	11000	0.4530	37.1095	13.8673
0.0123	9.65	12000	0.4763	37.5956	13.9583
0.007	10.45	13000	0.4939	37.3116	13.9360
0.0047	11.25	14000	0.5054	37.1750	13.8106
0.0036	12.06	15000	0.5093	37.5082	13.8930

Framework versions

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