whisper-ai-nomo

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: 12.2249

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
0.8516	0.8584	100	0.0650	14.5477
0.1705	1.7167	200	0.0484	18.7042
0.0597	2.5751	300	0.0123	14.1809
0.0515	3.4335	400	0.1206	59.9022
0.0337	4.2918	500	0.0030	6.3570
0.0232	5.1502	600	0.0023	6.3570
0.0356	6.0086	700	0.0155	12.9584
0.0124	6.8670	800	0.0014	4.7677
0.0152	7.7253	900	0.0030	168.7042
0.0089	8.5837	1000	0.0130	8.1907
0.0098	9.4421	1100	0.0023	12.7139
0.0095	10.3004	1200	0.0094	13.3252
0.0121	11.1588	1300	0.0008	6.9682
0.0082	12.0172	1400	0.0032	11.6137
0.0042	12.8755	1500	0.0126	12.8362
0.0091	13.7339	1600	0.0041	13.0807
0.0047	14.5923	1700	0.0005	9.5355
0.0082	15.4506	1800	0.0000	8.8020
0.003	16.3090	1900	0.0000	9.2910
0.0064	17.1674	2000	0.0002	56.9682
0.0017	18.0258	2100	0.0000	13.4474
0.0004	18.8841	2200	0.0000	16.2592
0.0014	19.7425	2300	0.0003	12.1027
0.0009	20.6009	2400	0.0002	12.1027
0.0001	21.4592	2500	0.0000	12.8362
0.0	22.3176	2600	0.0000	12.4694
0.0	23.1760	2700	0.0000	12.3472
0.0	24.0343	2800	0.0000	12.2249
0.0	24.8927	2900	0.0000	12.2249
0.0	25.7511	3000	0.0000	12.2249
0.0	26.6094	3100	0.0000	12.2249
0.0	27.4678	3200	0.0000	12.2249
0.0	28.3262	3300	0.0000	12.2249
0.0	29.1845	3400	0.0000	12.2249

Framework versions

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