yas1-wav2vec2-darija-marocain-test-2000

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

• Loss: 15.0797
• Wer: 1.0

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: 2
• seed: 42
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: linear
• lr_scheduler_warmup_steps: 1000
• num_epochs: 3
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Wer
39.0836	0.0538	50	368.6155	1.0
44.8514	0.1075	100	367.4959	1.0
51.6622	0.1613	150	365.2090	1.0
47.5838	0.2151	200	358.0549	1.0
51.5157	0.2688	250	345.3773	1.0
41.4047	0.3226	300	325.8074	1.0
50.1842	0.3763	350	291.5808	1.0
40.3827	0.4301	400	242.1378	1.0
30.337	0.4839	450	205.6870	1.0
29.3318	0.5376	500	182.6260	1.0
27.6937	0.5914	550	163.7182	1.0
29.1351	0.6452	600	151.1235	1.0
24.1047	0.6989	650	139.5537	1.0
24.3775	0.7527	700	130.2811	1.0
21.2306	0.8065	750	120.8115	1.0
16.4903	0.8602	800	112.0836	1.0
22.5541	0.9140	850	103.7457	1.0
16.4038	0.9677	900	95.6137	1.0
9.3657	1.0215	950	87.3968	1.0
8.9414	1.0753	1000	79.3383	1.0
9.2177	1.1290	1050	71.6232	1.0
9.2464	1.1828	1100	64.6608	1.0
6.8945	1.2366	1150	58.2576	1.0
6.5385	1.2903	1200	52.4811	1.0
7.0235	1.3441	1250	47.2149	1.0
5.8139	1.3978	1300	42.5255	1.0
5.4924	1.4516	1350	38.4367	1.0
6.3404	1.5054	1400	34.7247	1.0
5.5641	1.5591	1450	31.5253	1.0
4.6851	1.6129	1500	28.8176	1.0
6.0874	1.6667	1550	26.3767	1.0
4.8159	1.7204	1600	24.4155	1.0
4.3678	1.7742	1650	22.7697	1.0
4.2616	1.8280	1700	21.0956	1.0
4.2859	1.8817	1750	19.8534	1.0
4.2486	1.9355	1800	19.0108	1.0
3.9602	1.9892	1850	18.1411	1.0
4.3156	2.0430	1900	17.8388	1.0
3.7601	2.0968	1950	17.6259	1.0
3.9751	2.1505	2000	16.8560	1.0
4.5834	2.2043	2050	16.6510	1.0
3.7144	2.2581	2100	16.2614	1.0
3.7045	2.3118	2150	16.2588	1.0
3.6049	2.3656	2200	15.9820	1.0
3.75	2.4194	2250	15.6029	1.0
3.7796	2.4731	2300	15.8595	1.0
3.6727	2.5269	2350	15.6830	1.0
3.8072	2.5806	2400	15.3710	1.0
3.5989	2.6344	2450	15.2521	1.0
3.5626	2.6882	2500	15.3968	1.0
3.8358	2.7419	2550	15.3590	1.0
3.5086	2.7957	2600	15.1679	1.0
3.5551	2.8495	2650	15.0797	1.0
3.5962	2.9032	2700	15.0924	1.0
3.6453	2.9570	2750	15.1095	1.0

Framework versions

• Transformers 4.41.0
• Pytorch 2.4.1+cu121
• Datasets 2.18.0
• Tokenizers 0.19.1