csalt-voice-noLID

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

• Loss: 0.4769
• Wer: 17.0668

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: 2
• eval_batch_size: 8
• seed: 42
• gradient_accumulation_steps: 8
• 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: 500
• num_epochs: 100
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Wer
2.4828	0.9811	26	1.3369	45.4250
1.1813	2.0	53	0.8971	46.8527
0.8593	2.9811	79	0.7027	63.2057
0.618	4.0	106	0.5701	37.6379
0.4389	4.9811	132	0.4288	21.4796
0.1532	6.0	159	0.3010	20.8955
0.0838	6.9811	185	0.3005	19.0136
0.0489	8.0	212	0.3027	22.6476
0.0329	8.9811	238	0.3128	21.6742
0.0238	10.0	265	0.3228	18.6243
0.016	10.9811	291	0.3235	18.5594
0.0133	12.0	318	0.3145	18.2349
0.0116	12.9811	344	0.3394	16.7424
0.0117	14.0	371	0.3416	19.1434
0.011	14.9811	397	0.3728	18.8838
0.0089	16.0	424	0.3508	18.2349
0.0103	16.9811	450	0.3698	20.3115
0.0122	18.0	477	0.3686	20.1168
0.0146	18.9811	503	0.3735	19.1434
0.0154	20.0	530	0.3830	19.5328
0.0115	20.9811	556	0.3809	20.3764
0.0082	22.0	583	0.3982	19.9870
0.0066	22.9811	609	0.3936	19.0785
0.0048	24.0	636	0.4018	19.8572
0.0055	24.9811	662	0.3829	18.1051
0.005	26.0	689	0.3721	17.4562
0.0042	26.9811	715	0.3759	17.9104
0.0035	28.0	742	0.3930	17.7807
0.0024	28.9811	768	0.3987	18.2349
0.0024	30.0	795	0.4157	17.2615
0.0014	30.9811	821	0.4114	17.0019
0.0012	32.0	848	0.4123	16.8722
0.0009	32.9811	874	0.4210	17.5211
0.0009	34.0	901	0.4182	17.3264
0.0008	34.9811	927	0.4176	17.3913
0.0008	36.0	954	0.4168	17.4562
0.0004	36.9811	980	0.4222	17.3264
0.0004	38.0	1007	0.4252	17.5860
0.0003	38.9811	1033	0.4276	17.2615
0.0003	40.0	1060	0.4291	17.5211
0.0003	40.9811	1086	0.4298	17.3913
0.0003	42.0	1113	0.4308	17.3913
0.0003	42.9811	1139	0.4325	17.0668
0.0003	44.0	1166	0.4337	17.0668
0.0003	44.9811	1192	0.4348	17.0668
0.0002	46.0	1219	0.4358	17.0668
0.0002	46.9811	1245	0.4364	17.0668
0.0002	48.0	1272	0.4378	17.0668
0.0002	48.9811	1298	0.4388	17.0668
0.0002	50.0	1325	0.4400	17.0019
0.0002	50.9811	1351	0.4411	17.0019
0.0002	52.0	1378	0.4421	17.0019
0.0002	52.9811	1404	0.4425	17.0019
0.0002	54.0	1431	0.4438	17.0668
0.0002	54.9811	1457	0.4446	17.0668
0.0002	56.0	1484	0.4461	17.0668
0.0002	56.9811	1510	0.4467	17.1317
0.0002	58.0	1537	0.4479	17.1317
0.0002	58.9811	1563	0.4488	17.1317
0.0002	60.0	1590	0.4497	17.1317
0.0002	60.9811	1616	0.4502	17.0019
0.0002	62.0	1643	0.4512	16.8722
0.0002	62.9811	1669	0.4520	17.0019
0.0002	64.0	1696	0.4528	16.8722
0.0002	64.9811	1722	0.4541	16.8722
0.0002	66.0	1749	0.4548	17.0668
0.0002	66.9811	1775	0.4553	17.0668
0.0002	68.0	1802	0.4560	17.1317
0.0002	68.9811	1828	0.4566	17.2615
0.0002	70.0	1855	0.4579	17.3913
0.0002	70.9811	1881	0.4582	17.3913
0.0002	72.0	1908	0.4590	17.3913
0.0002	72.9811	1934	0.4599	17.3913
0.0002	74.0	1961	0.4605	17.3264
0.0002	74.9811	1987	0.4612	17.3264
0.0002	76.0	2014	0.4620	17.3264
0.0001	76.9811	2040	0.4684	17.2615
0.0001	78.0	2067	0.4715	17.2615
0.0001	78.9811	2093	0.4726	17.0668
0.0001	80.0	2120	0.4731	17.0668
0.0001	80.9811	2146	0.4733	17.0668
0.0001	82.0	2173	0.4738	17.0668
0.0001	82.9811	2199	0.4741	17.0668
0.0001	84.0	2226	0.4744	17.0668
0.0001	84.9811	2252	0.4748	17.0668
0.0001	86.0	2279	0.4751	17.0668
0.0001	86.9811	2305	0.4754	17.0668
0.0001	88.0	2332	0.4756	17.0668
0.0001	88.9811	2358	0.4759	17.0668
0.0001	90.0	2385	0.4762	17.0668
0.0001	90.9811	2411	0.4762	17.0668
0.0001	92.0	2438	0.4765	17.0668
0.0001	92.9811	2464	0.4767	17.1317
0.0001	94.0	2491	0.4767	17.0668
0.0001	94.9811	2517	0.4769	17.0668
0.0001	96.0	2544	0.4769	17.1317
0.0001	96.9811	2570	0.4769	17.1317
0.0001	98.0	2597	0.4769	17.1317
0.0001	98.1132	2600	0.4769	17.0668

Framework versions

• Transformers 4.44.2
• Pytorch 2.4.1+cu118
• Datasets 3.1.0
• Tokenizers 0.19.1