whisper-large-v2-ft-tms-good-and-bad-60-250504-v2

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: 1.5119

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
12.6652	1.0	1	12.6099
12.7284	2.0	2	12.6099
12.6979	3.0	3	12.6099
12.8604	4.0	4	12.6099
12.6964	5.0	5	12.6099
12.6169	6.0	6	12.6099
12.5936	7.0	7	12.6099
12.6436	8.0	8	12.6099
12.6892	9.0	9	12.5756
12.6292	10.0	10	12.4211
12.3777	11.0	11	12.1817
12.3014	12.0	12	11.8436
11.9713	13.0	13	11.8436
11.916	14.0	14	11.4236
11.3473	15.0	15	10.9360
10.7909	16.0	16	10.3283
10.3267	17.0	17	9.6044
9.4613	18.0	18	8.7731
8.7764	19.0	19	7.9344
7.6595	20.0	20	7.0902
6.8138	21.0	21	6.1688
5.8664	22.0	22	5.3031
5.4038	23.0	23	4.9312
5.1601	24.0	24	4.7820
4.9727	25.0	25	4.6886
4.954	26.0	26	4.6085
4.8044	27.0	27	4.5212
4.6758	28.0	28	4.4243
4.6124	29.0	29	4.3207
4.5222	30.0	30	4.2157
4.3924	31.0	31	4.1097
4.2287	32.0	32	4.0007
4.1772	33.0	33	3.8919
4.0695	34.0	34	3.7831
3.9619	35.0	35	3.6808
3.734	36.0	36	3.5853
3.6489	37.0	37	3.4984
3.6132	38.0	38	3.4186
3.472	39.0	39	3.3445
3.3173	40.0	40	3.2727
3.2766	41.0	41	3.2021
3.1815	42.0	42	3.1324
3.1	43.0	43	3.0623
3.0455	44.0	44	2.9923
2.9112	45.0	45	2.9198
2.8782	46.0	46	2.8427
2.8031	47.0	47	2.7588
2.6295	48.0	48	2.6644
2.5573	49.0	49	2.5570
2.3889	50.0	50	2.4584
2.316	51.0	51	2.3695
2.1967	52.0	52	2.2957
2.1272	53.0	53	2.2318
2.0393	54.0	54	2.1753
1.9701	55.0	55	2.1288
1.9604	56.0	56	2.0918
1.9301	57.0	57	2.0618
1.8725	58.0	58	2.0354
1.8635	59.0	59	2.0108
1.8153	60.0	60	1.9883
1.7738	61.0	61	1.9668
1.7451	62.0	62	1.9461
1.7126	63.0	63	1.9264
1.6683	64.0	64	1.9074
1.645	65.0	65	1.8894
1.6364	66.0	66	1.8724
1.6139	67.0	67	1.8558
1.5709	68.0	68	1.8405
1.5621	69.0	69	1.8259
1.5538	70.0	70	1.8118
1.535	71.0	71	1.7986
1.5007	72.0	72	1.7854
1.4835	73.0	73	1.7729
1.4732	74.0	74	1.7607
1.4428	75.0	75	1.7488
1.4459	76.0	76	1.7368
1.4287	77.0	77	1.7247
1.4191	78.0	78	1.7124
1.401	79.0	79	1.7001
1.3784	80.0	80	1.6870
1.3731	81.0	81	1.6744
1.3668	82.0	82	1.6618
1.3465	83.0	83	1.6490
1.3381	84.0	84	1.6363
1.329	85.0	85	1.6228
1.3128	86.0	86	1.6104
1.3025	87.0	87	1.5992
1.2873	88.0	88	1.5880
1.2777	89.0	89	1.5778
1.2738	90.0	90	1.5686
1.2671	91.0	91	1.5597
1.2636	92.0	92	1.5518
1.2564	93.0	93	1.5451
1.2429	94.0	94	1.5387
1.2498	95.0	95	1.5330
1.2365	96.0	96	1.5277
1.2289	97.0	97	1.5232
1.2182	98.0	98	1.5189
1.2133	99.0	99	1.5151
1.2252	100.0	100	1.5119

Framework versions

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