wav2vec

This model is a fine-tuned version of vitouphy/wav2vec2-xls-r-300m-english on an unknown dataset. It achieves the following results on the evaluation set:

• Loss: 417.9874
• Pcc Accuracy: 0.2482
• Pcc Fluency: 0.2791
• Pcc Total Score: 0.3110
• Pcc Content: 0.3780

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: 4
• eval_batch_size: 6
• seed: 42
• gradient_accumulation_steps: 2
• total_train_batch_size: 8
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: cosine
• lr_scheduler_warmup_ratio: 0.4
• num_epochs: 25
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Pcc Accuracy	Pcc Fluency	Pcc Total Score	Pcc Content
3176.8777	1.01	100	2967.8770	0.2384	0.1513	-0.2037	-0.0386
3000.9279	2.02	200	2907.2825	0.2949	0.1813	-0.1250	0.0827
2716.2498	3.03	300	2805.8979	0.3034	0.2123	0.0290	0.2344
2600.8768	4.04	400	2666.0171	0.2859	0.2345	0.1642	0.3183
2222.3631	5.05	500	2490.9263	0.2698	0.2475	0.2305	0.3488
1940.7414	6.06	600	2284.4573	0.2573	0.2552	0.2588	0.3591
1962.4018	7.07	700	2051.6846	0.2504	0.2603	0.2738	0.3635
1506.0297	8.08	800	1798.2383	0.2444	0.2633	0.2813	0.3653
1448.3059	9.09	900	1534.5461	0.2396	0.2662	0.2845	0.3650
1202.578	10.1	1000	1265.2390	0.2376	0.2678	0.2873	0.3647
917.5093	11.11	1100	1021.2091	0.2356	0.2697	0.2896	0.3651
781.4407	12.12	1200	825.2852	0.2340	0.2710	0.2901	0.3647
633.8744	13.13	1300	674.1681	0.2337	0.2724	0.2918	0.3652
554.5075	14.14	1400	573.6318	0.2354	0.2737	0.2954	0.3677
500.6607	15.15	1500	510.6489	0.2378	0.2740	0.2978	0.3700
472.1874	16.16	1600	468.3256	0.2394	0.2751	0.3012	0.3720
406.9743	17.17	1700	444.8770	0.2421	0.2763	0.3041	0.3739
373.2401	18.18	1800	432.6308	0.2438	0.2771	0.3068	0.3751
447.599	19.19	1900	425.9487	0.2457	0.2778	0.3081	0.3762
360.8572	20.2	2000	421.8146	0.2466	0.2786	0.3093	0.3772
409.8801	21.21	2100	420.0713	0.2473	0.2786	0.3100	0.3777
419.8665	22.22	2200	418.7286	0.2478	0.2791	0.3107	0.3778
369.3772	23.23	2300	418.1939	0.2477	0.2791	0.3105	0.3776
449.1843	24.24	2400	417.9874	0.2482	0.2791	0.3110	0.3780

Framework versions

• Transformers 4.37.0
• Pytorch 2.1.2
• Datasets 2.17.0
• Tokenizers 0.15.1