whisper-small-nomi

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: 0.0000
• Wer: 0.6383

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
1.4305	1.9417	100	0.1954	36.8085
0.3179	3.8835	200	0.1760	71.7021
0.1876	5.8252	300	0.2092	24.4681
0.1377	7.7670	400	0.1063	23.1915
0.0933	9.7087	500	0.0556	7.8723
0.0713	11.6505	600	0.0556	9.5745
0.0548	13.5922	700	0.0289	5.1064
0.0436	15.5340	800	0.0330	3.6170
0.0258	17.4757	900	0.0041	0.8511
0.0094	19.4175	1000	0.0072	2.1277
0.0047	21.3592	1100	0.0001	1.7021
0.001	23.3010	1200	0.0001	0.8511
0.0	25.2427	1300	0.0000	0.6383
0.0	27.1845	1400	0.0000	0.6383
0.0	29.1262	1500	0.0000	0.6383

Framework versions

• Transformers 4.45.0.dev0
• Pytorch 2.4.0
• Datasets 2.21.0
• Tokenizers 0.19.1