JunWorks/whisper-small-LIDmetric-4060-2lang

Whisper Small 2 languages normal

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

• Loss: 0.3118
• Cer: 90.4765
• Wer: 86.5335
• Both Er: 88.7786
• Lid: 44.2884
• Total: 144.4902

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: 1
• eval_batch_size: 1
• seed: 42
• gradient_accumulation_steps: 16
• 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
• training_steps: 20000
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Cer	Wer	Both Er	Lid	Total
0.2413	0.8017	1000	0.2265	81.5308	90.5108	85.3977	16.8515	168.5462
0.1096	1.6034	2000	0.2138	77.0089	115.7964	93.7113	11.5530	182.1583
0.0452	2.4052	3000	0.2195	67.2631	106.0961	83.9850	14.4088	169.5763
0.0192	3.2069	4000	0.2289	79.4602	114.9595	94.7466	5.9004	188.8462
0.0178	4.0086	5000	0.2350	66.7805	98.9821	80.6469	10.7741	169.8728
0.0091	4.8103	6000	0.2502	121.1795	172.3431	143.2111	15.3292	227.8819
0.0049	5.6121	7000	0.2636	45.7609	74.6202	58.1880	14.7746	143.4134
0.0032	6.4138	8000	0.2696	59.2028	88.8860	71.9847	18.1732	153.8115
0.0021	7.2155	9000	0.2700	88.3501	115.7851	100.1639	19.4831	180.6808
0.0019	8.0172	10000	0.2724	90.1750	122.9142	104.2729	19.4949	184.7780
0.001	8.8190	11000	0.2873	66.9644	87.4194	75.7726	23.4246	152.3480
0.0011	9.6207	12000	0.2893	148.6111	150.7107	149.5152	32.5112	217.0040
0.0008	10.4224	13000	0.2963	94.3029	117.2139	104.1687	28.0269	176.1418
0.0006	11.2241	14000	0.2984	53.1570	76.5127	63.2142	29.2896	133.9246
0.0004	12.0259	15000	0.2985	75.7925	97.0631	84.9519	35.8390	149.1129
0.0002	12.8276	16000	0.3013	88.8394	94.8011	91.4066	36.5943	154.8123
0.0002	13.6293	17000	0.3062	102.0049	102.1225	102.0555	41.2202	160.8353
0.0001	14.4310	18000	0.3097	95.7946	88.7842	92.7758	43.1555	149.6203
0.0002	15.2328	19000	0.3105	91.7297	87.2234	89.7892	43.7928	145.9964
0.0001	16.0345	20000	0.3118	90.4765	86.5335	88.7786	44.2884	144.4902

Framework versions

• Transformers 4.40.2
• Pytorch 2.3.0+cu121
• Datasets 2.20.0
• Tokenizers 0.19.1