wav2vec2-XLS-R-Fleurs-demo-google-colab-Ezra_William_Prod10

This model is a fine-tuned version of facebook/wav2vec2-large-xlsr-53 on the xtreme_s dataset. It achieves the following results on the evaluation set:

• Loss: 2.5424
• Wer: 0.9609

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: 0.001
• train_batch_size: 16
• eval_batch_size: 8
• seed: 42
• 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
9.2693	1.0	39	3.1721	1.0
2.9485	2.0	78	2.8726	1.0
2.8943	3.0	117	2.8655	1.0
2.9036	4.0	156	2.8631	1.0
2.8869	5.0	195	2.8614	1.0
2.8802	6.0	234	2.8854	1.0
2.8802	7.0	273	2.8515	1.0
2.8706	8.0	312	2.8609	1.0
2.8712	9.0	351	2.8458	1.0
2.8583	10.0	390	2.8361	1.0
2.857	11.0	429	2.8355	1.0
2.8546	12.0	468	2.8439	1.0
2.8463	13.0	507	2.8348	1.0
2.8307	14.0	546	2.7596	1.0
2.7673	15.0	585	2.6289	1.0
2.5597	16.0	624	2.3411	1.0
2.224	17.0	663	2.0992	1.0
2.0145	18.0	702	1.7290	1.0
1.7274	19.0	741	1.5571	0.9954
1.6774	20.0	780	1.4439	0.9906
1.4585	21.0	819	1.3841	1.1238
1.342	22.0	858	1.2805	0.9662
1.215	23.0	897	1.2965	0.9628
1.188	24.0	936	1.2713	0.9659
1.1147	25.0	975	1.2936	1.0251
1.0374	26.0	1014	1.2900	0.9483
0.9352	27.0	1053	1.3671	0.9908
0.9249	28.0	1092	1.3018	0.9404
0.7973	29.0	1131	1.3253	0.9631
0.7451	30.0	1170	1.4314	1.0451
0.7391	31.0	1209	1.4553	0.9909
0.699	32.0	1248	1.5116	0.9487
0.5484	33.0	1287	1.5492	0.9829
0.5106	34.0	1326	1.6631	1.0674
0.5989	35.0	1365	1.6305	1.0150
0.464	36.0	1404	1.6285	0.9430
0.4925	37.0	1443	1.7208	1.0183
0.4206	38.0	1482	1.7476	1.0040
0.3848	39.0	1521	1.8125	1.0341
0.4057	40.0	1560	1.8245	0.9750
0.3978	41.0	1599	1.7153	0.9326
0.3806	42.0	1638	1.8650	1.0025
0.3376	43.0	1677	1.9067	0.9653
0.3798	44.0	1716	2.0028	0.9396
0.2902	45.0	1755	2.0901	0.9920
0.3324	46.0	1794	1.8935	0.9729
0.3241	47.0	1833	2.0133	1.0074
0.3055	48.0	1872	2.0352	0.9943
0.2927	49.0	1911	1.9539	1.0022
0.2729	50.0	1950	2.0982	0.9910
0.2569	51.0	1989	2.1607	0.9832
0.2683	52.0	2028	2.2544	0.9705
0.2685	53.0	2067	2.1528	0.9857
0.2757	54.0	2106	2.1648	0.9490
0.2379	55.0	2145	2.2498	0.9693
0.2501	56.0	2184	2.2509	1.0282
0.2347	57.0	2223	2.2095	0.9897
0.2697	58.0	2262	2.1933	0.9695
0.2255	59.0	2301	2.2140	0.9756
0.2071	60.0	2340	2.2364	0.9787
0.228	61.0	2379	2.3069	0.9551
0.2112	62.0	2418	2.3191	0.9769
0.2224	63.0	2457	2.2679	1.0025
0.2081	64.0	2496	2.2548	0.9660
0.2149	65.0	2535	2.1813	0.9720
0.2156	66.0	2574	2.0609	0.9633
0.1916	67.0	2613	2.4192	0.9594
0.2221	68.0	2652	2.3571	1.0186
0.1849	69.0	2691	2.3650	0.9705
0.1999	70.0	2730	2.3588	0.9700
0.2314	71.0	2769	2.5680	1.0693
0.1995	72.0	2808	2.3918	1.0490
0.1842	73.0	2847	2.3448	0.9706
0.1841	74.0	2886	2.3811	0.9945
0.1836	75.0	2925	2.4134	0.9659
0.1844	76.0	2964	2.3892	0.9657
0.1933	77.0	3003	2.3327	0.9606
0.1757	78.0	3042	2.4641	0.9702
0.1794	79.0	3081	2.4175	0.9535
0.1795	80.0	3120	2.3742	0.9503
0.1859	81.0	3159	2.5093	0.9508
0.1641	82.0	3198	2.4232	0.9647
0.195	83.0	3237	2.4070	0.9474
0.1712	84.0	3276	2.4726	0.9674
0.1882	85.0	3315	2.4682	0.9643
0.1746	86.0	3354	2.4826	0.9523
0.1655	87.0	3393	2.5652	0.9495
0.1895	88.0	3432	2.4967	0.9489
0.1659	89.0	3471	2.4620	0.9695
0.1618	90.0	3510	2.4974	0.9433
0.1559	91.0	3549	2.5137	0.9599
0.1646	92.0	3588	2.4645	0.9579
0.1599	93.0	3627	2.4751	0.9612
0.1735	94.0	3666	2.5473	0.9597
0.1571	95.0	3705	2.5158	0.9675
0.1606	96.0	3744	2.5234	0.9645
0.1499	97.0	3783	2.5328	0.9612
0.1571	98.0	3822	2.5535	0.9594
0.166	99.0	3861	2.5450	0.9592
0.1651	100.0	3900	2.5424	0.9609

Framework versions

• Transformers 4.39.2
• Pytorch 2.2.2+cu121
• Datasets 2.18.0
• Tokenizers 0.15.2