whisper-large-v2-ft-tms-good-and-bad-50-250505-v1

This model is a fine-tuned version of openai/whisper-large-v2 on the None dataset. It achieves the following results on the evaluation set:

• Loss: 1.8823

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: 8
• eval_batch_size: 8
• seed: 42
• distributed_type: multi-GPU
• num_devices: 8
• total_train_batch_size: 64
• total_eval_batch_size: 64
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: linear
• lr_scheduler_warmup_ratio: 0.2
• num_epochs: 100
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss
11.3816	1.0	1	12.5223
11.5454	2.0	2	12.5223
11.7174	3.0	3	12.5223
11.5318	4.0	4	12.5223
11.5732	5.0	5	12.5223
11.5473	6.0	6	12.5223
11.3888	7.0	7	12.5223
11.605	8.0	8	12.5223
11.6527	9.0	9	12.4712
11.2839	10.0	10	12.2859
11.0946	11.0	11	11.9612
11.1023	12.0	12	11.5144
10.7491	13.0	13	10.9702
10.254	14.0	14	10.3222
9.4715	15.0	15	9.5843
8.6174	16.0	16	8.7475
8.3552	17.0	17	7.8500
7.2577	18.0	18	7.0251
6.6791	19.0	19	6.2691
5.795	20.0	20	5.6403
5.468	21.0	21	5.3147
5.1026	22.0	22	5.1391
4.9813	23.0	23	5.0325
4.8977	24.0	24	4.9578
4.7086	25.0	25	4.8893
4.7115	26.0	26	4.8260
4.5435	27.0	27	4.7666
4.4317	28.0	28	4.7114
4.4249	29.0	29	4.6608
4.3736	30.0	30	4.6085
4.2158	31.0	31	4.5557
4.1822	32.0	32	4.4951
4.0871	33.0	33	4.4217
4.0832	34.0	34	4.3240
3.9507	35.0	35	4.1705
3.7172	36.0	36	3.9373
3.6009	37.0	37	3.6852
3.4037	38.0	38	3.5052
3.2777	39.0	39	3.3997
3.1126	40.0	40	3.3483
3.1557	41.0	41	3.3135
2.9871	42.0	42	3.2783
2.9043	43.0	43	3.2402
2.8365	44.0	44	3.1998
2.7988	45.0	45	3.1561
2.7496	46.0	46	3.1120
2.6076	47.0	47	3.0659
2.6008	48.0	48	3.0197
2.5564	49.0	49	2.9724
2.5158	50.0	50	2.9254
2.4585	51.0	51	2.8792
2.4219	52.0	52	2.8358
2.4166	53.0	53	2.7935
2.33	54.0	54	2.7552
2.2931	55.0	55	2.7183
2.273	56.0	56	2.6839
2.2016	57.0	57	2.6502
2.1833	58.0	58	2.6179
2.1684	59.0	59	2.5877
2.1622	60.0	60	2.5584
2.0812	61.0	61	2.5304
2.1008	62.0	62	2.5038
2.0411	63.0	63	2.4779
2.0118	64.0	64	2.4534
1.9913	65.0	65	2.4292
1.9612	66.0	66	2.4059
1.9046	67.0	67	2.3834
1.8865	68.0	68	2.3611
1.9055	69.0	69	2.3389
1.8638	70.0	70	2.3172
1.8163	71.0	71	2.2956
1.8067	72.0	72	2.2744
1.7946	73.0	73	2.2542
1.7988	74.0	74	2.2343
1.7813	75.0	75	2.2147
1.7345	76.0	76	2.1957
1.7475	77.0	77	2.1769
1.7323	78.0	78	2.1590
1.7226	79.0	79	2.1410
1.6695	80.0	80	2.1237
1.6686	81.0	81	2.1066
1.6696	82.0	82	2.0899
1.6251	83.0	83	2.0737
1.6318	84.0	84	2.0580
1.5919	85.0	85	2.0432
1.5714	86.0	86	2.0281
1.5807	87.0	87	2.0136
1.5506	88.0	88	2.0002
1.538	89.0	89	1.9870
1.5381	90.0	90	1.9737
1.5244	91.0	91	1.9619
1.5001	92.0	92	1.9505
1.4815	93.0	93	1.9393
1.4942	94.0	94	1.9293
1.5128	95.0	95	1.9193
1.4744	96.0	96	1.9102
1.464	97.0	97	1.9025
1.4661	98.0	98	1.8948
1.4467	99.0	99	1.8879
1.4349	100.0	100	1.8823

Framework versions

• PEFT 0.13.0
• Transformers 4.45.1
• Pytorch 2.5.0+cu124
• Datasets 2.21.0
• Tokenizers 0.20.0