COPAS-withctrl-whisper-lg-3-Dec4

This model is a fine-tuned version of openai/whisper-large-v3 on an unknown dataset. It achieves the following results on the evaluation set:

• Loss: 0.0830
• Wer: 22.8522

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: 1e-05
• 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: 100
• num_epochs: 100
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Wer
0.7468	1.2048	100	0.3724	30.8849
0.1779	2.4096	200	0.1806	26.6323
0.0631	3.6145	300	0.1187	25.4296
0.031	4.8193	400	0.1113	25.2148
0.0165	6.0241	500	0.0997	25.0
0.0104	7.2289	600	0.1012	23.7328
0.0062	8.4337	700	0.1017	25.3651
0.0053	9.6386	800	0.0928	24.0979
0.0051	10.8434	900	0.0956	24.9356
0.004	12.0482	1000	0.0868	23.3462
0.0018	13.2530	1100	0.0907	23.9905
0.0022	14.4578	1200	0.0915	23.3247
0.0025	15.6627	1300	0.0871	23.1744
0.0029	16.8675	1400	0.0930	23.8832
0.0025	18.0723	1500	0.0909	23.8617
0.0029	19.2771	1600	0.0818	23.0455
0.0027	20.4819	1700	0.0984	24.1409
0.0015	21.6867	1800	0.0904	23.8832
0.0009	22.8916	1900	0.0933	23.5610
0.0013	24.0964	2000	0.0917	23.9261
0.0016	25.3012	2100	0.0881	23.7758
0.0037	26.5060	2200	0.0983	24.4416
0.0039	27.7108	2300	0.1022	25.0215
0.0027	28.9157	2400	0.0839	23.4536
0.0032	30.1205	2500	0.0768	23.8617
0.0012	31.3253	2600	0.0810	24.8497
0.0006	32.5301	2700	0.0809	24.8926
0.0012	33.7349	2800	0.0738	23.7543
0.0012	34.9398	2900	0.0730	23.8187
0.0015	36.1446	3000	0.0793	23.5180
0.0015	37.3494	3100	0.0813	23.9046
0.0008	38.5542	3200	0.0784	23.2603
0.0006	39.7590	3300	0.0844	22.9596
0.0006	40.9639	3400	0.0798	21.9502
0.0004	42.1687	3500	0.0785	22.4442
0.0001	43.3735	3600	0.0792	22.6589
0.0	44.5783	3700	0.0795	22.0361
0.0	45.7831	3800	0.0799	22.5515
0.0	46.9880	3900	0.0803	23.0026
0.0	48.1928	4000	0.0807	23.2388
0.0	49.3976	4100	0.0809	23.1100
0.0	50.6024	4200	0.0812	22.9381
0.0	51.8072	4300	0.0815	22.9381
0.0	53.0120	4400	0.0816	22.9381
0.0	54.2169	4500	0.0818	23.0241
0.0	55.4217	4600	0.0820	23.0241
0.0	56.6265	4700	0.0822	22.8308
0.0	57.8313	4800	0.0823	22.8952
0.0	59.0361	4900	0.0825	22.8737
0.0	60.2410	5000	0.0826	22.8952
0.0	61.4458	5100	0.0827	22.8522
0.0	62.6506	5200	0.0829	22.8522
0.0	63.8554	5300	0.0830	22.8522

Framework versions

• Transformers 4.43.4
• Pytorch 2.4.1
• Datasets 3.0.0
• Tokenizers 0.19.1