mike272/whisper-large-v3-polish-impaired-v3

Whisper Large v3 - impaired polish speech

This model is a fine-tuned version of openai/whisper-large-v3 on the None dataset. It achieves the following results on the evaluation set:

• Loss: 0.6737
• Wer: 94.6970

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: 5e-05
• train_batch_size: 16
• eval_batch_size: 8
• seed: 42
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: linear
• lr_scheduler_warmup_steps: 3
• training_steps: 50
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Wer
0.8089	0.05	1	0.8794	99.6212
0.7774	0.1	2	0.8156	100.0
1.0544	0.15	3	0.8751	100.0
1.0424	0.2	4	0.9852	100.0
0.9079	0.25	5	0.9499	100.0
0.892	0.3	6	0.8850	71.9697
0.7047	0.35	7	0.8983	83.3333
0.7929	0.4	8	0.8541	64.0152
0.5985	0.45	9	0.8023	65.5303
0.918	0.5	10	0.7707	64.7727
0.441	0.55	11	0.7689	72.3485
0.712	0.6	12	0.7856	89.0152
0.615	0.65	13	0.7907	98.4848
0.8559	0.7	14	0.7920	97.3485
0.6121	0.75	15	0.8105	96.9697
0.6539	0.8	16	0.8331	97.3485
0.6593	0.85	17	0.8068	99.6212
0.7253	0.9	18	0.7870	100.0
0.6551	0.95	19	0.7596	100.0
0.5775	1.0	20	0.7319	100.0
0.2112	1.05	21	0.7120	100.0
0.1157	1.1	22	0.6858	99.6212
0.1208	1.15	23	0.6795	97.3485
0.1542	1.2	24	0.6846	92.4242
0.1902	1.25	25	0.6980	79.5455
0.3083	1.3	26	0.7203	78.7879
0.2311	1.35	27	0.7332	71.2121
0.1591	1.4	28	0.7313	67.8030
0.1156	1.45	29	0.7307	66.6667
0.2858	1.5	30	0.7188	64.3939
0.1179	1.55	31	0.7063	62.5
0.1541	1.6	32	0.6875	62.1212
0.1659	1.65	33	0.6741	62.5
0.1902	1.7	34	0.6713	77.2727
0.1	1.75	35	0.6729	93.1818
0.1201	1.8	36	0.6764	95.8333
0.1049	1.85	37	0.6784	98.4848
0.2873	1.9	38	0.6796	98.8636
0.1357	1.95	39	0.6800	98.8636
0.2411	2.0	40	0.6798	98.8636
0.0791	2.05	41	0.6829	98.8636
0.069	2.1	42	0.6811	98.4848
0.0996	2.15	43	0.6775	98.4848
0.0409	2.2	44	0.6762	97.3485
0.0355	2.25	45	0.6737	96.2121
0.0336	2.3	46	0.6732	95.8333
0.0183	2.35	47	0.6734	95.8333
0.1063	2.4	48	0.6739	95.0758
0.0326	2.45	49	0.6738	95.0758
0.0546	2.5	50	0.6737	94.6970

Framework versions

• Transformers 4.37.2
• Pytorch 2.2.1+cu121
• Datasets 3.2.0
• Tokenizers 0.15.2