whisper-ai-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: 84.4743

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
0.8208	0.8163	100	0.1289	13.0807
0.2627	1.6327	200	0.3265	94.8655
1.5786	2.4490	300	3.1093	100.0
0.4107	3.2653	400	0.2735	93.5208
0.2687	4.0816	500	0.2438	92.6650
0.2194	4.8980	600	0.2357	94.0098
0.1686	5.7143	700	0.2356	723.7164
0.1302	6.5306	800	0.0559	194.6210
0.0847	7.3469	900	0.0397	34.9633
0.0595	8.1633	1000	0.0123	17.8484
0.0622	8.9796	1100	0.2014	54.1565
0.0518	9.7959	1200	0.0192	27.9951
0.0285	10.6122	1300	0.0271	24.5721
0.0307	11.4286	1400	0.0565	243.0318
0.0207	12.2449	1500	0.0021	20.0489
0.0156	13.0612	1600	0.0091	21.1491
0.0157	13.8776	1700	0.0147	71.0269
0.0127	14.6939	1800	0.0019	20.2934
0.0084	15.5102	1900	0.0007	14.0587
0.008	16.3265	2000	0.0052	19.9267
0.0071	17.1429	2100	0.0009	98.5330
0.0023	17.9592	2200	0.0001	58.8020
0.002	18.7755	2300	0.0012	33.4963
0.002	19.5918	2400	0.0018	176.6504
0.0033	20.4082	2500	0.0001	148.5330
0.0017	21.2245	2600	0.0000	122.8606
0.0004	22.0408	2700	0.0000	125.3056
0.0001	22.8571	2800	0.0000	110.8802
0.0	23.6735	2900	0.0000	110.8802
0.0	24.4898	3000	0.0000	97.9218
0.0	25.3061	3100	0.0000	97.9218
0.0	26.1224	3200	0.0000	84.5966
0.0	26.9388	3300	0.0000	84.4743
0.0	27.7551	3400	0.0000	84.4743
0.0	28.5714	3500	0.0000	84.4743
0.0	29.3878	3600	0.0000	84.4743

Framework versions

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