whisper-nm-nomimo

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: 0.2305
• Wer: 17.7326

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: 50
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Wer
No log	2.1739	100	0.4521	81.3953
0.8708	4.3478	200	1.0045	35.4651
0.8708	6.5217	300	0.3180	68.8953
0.1996	8.6957	400	0.2474	39.9709
0.1996	10.8696	500	0.1908	52.7616
0.1119	13.0435	600	0.1939	21.5116
0.1119	15.2174	700	0.2796	31.1047
0.0747	17.3913	800	0.2581	29.2151
0.0747	19.5652	900	0.2637	31.1047
0.0493	21.7391	1000	0.2337	25.1453
0.0493	23.9130	1100	0.2266	125.5814
0.0364	26.0870	1200	0.2047	18.6047
0.0364	28.2609	1300	0.2721	28.1977
0.0202	30.4348	1400	0.2931	24.8547
0.0202	32.6087	1500	0.2378	56.5407
0.0083	34.7826	1600	0.2336	19.6221
0.0083	36.9565	1700	0.2283	18.3140
0.0005	39.1304	1800	0.2326	18.6047
0.0005	41.3043	1900	0.2283	18.1686
0.0003	43.4783	2000	0.2308	17.7326
0.0003	45.6522	2100	0.2305	17.7326
0.0	47.8261	2200	0.2305	17.7326
0.0	50.0	2300	0.2305	17.7326

Framework versions

• Transformers 4.47.0.dev0
• Pytorch 2.4.0
• Datasets 3.0.1
• Tokenizers 0.20.0