whisper-a-nomimo-18

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: 0.0290
• Wer: 143.6728

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: 18
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Wer
1.0616	1.0	104	0.2173	30.0926
0.14	2.0	208	0.0770	18.5957
0.0714	3.0	312	0.0723	19.6759
0.0513	4.0	416	0.0671	20.5247
0.0381	5.0	520	0.0415	17.9012
0.0468	6.0	624	0.0463	22.3765
0.0352	7.0	728	0.1139	42.5154
0.0195	8.0	832	0.0457	148.7654
0.0185	9.0	936	0.0430	172.6852
0.0129	10.0	1040	0.0312	154.8611
0.0106	11.0	1144	0.0405	145.5247
0.0084	12.0	1248	0.0325	154.9383
0.0058	13.0	1352	0.0320	152.0062
0.0039	14.0	1456	0.0263	144.5216
0.0044	15.0	1560	0.0270	148.3796
0.0028	16.0	1664	0.0285	148.6883
0.0015	17.0	1768	0.0285	147.3765
0.002	17.8309	1854	0.0290	143.6728

Framework versions

• Transformers 4.48.0.dev0
• Pytorch 2.4.0
• Datasets 3.1.0
• Tokenizers 0.21.0