wav2vec2-XLS-R-Fleurs-demo-google-colab-Ezra_William_Prod9

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: 1.6655
• Wer: 0.8832

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: 50
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Wer
9.2247	1.0	39	3.1705	1.0
2.9516	2.0	78	2.8732	1.0
2.8953	3.0	117	2.8663	1.0
2.8878	4.0	156	2.8693	1.0
2.8818	5.0	195	2.8656	1.0
2.8814	6.0	234	2.8478	1.0
2.8817	7.0	273	2.8603	1.0
2.8771	8.0	312	2.8647	1.0
2.8811	9.0	351	2.8630	1.0
2.8703	10.0	390	2.8667	1.0
2.8608	11.0	429	2.8429	1.0
2.8578	12.0	468	2.8399	1.0
2.856	13.0	507	2.8531	1.0
2.8438	14.0	546	2.7872	1.0
2.7833	15.0	585	2.7015	1.0
2.7126	16.0	624	2.5606	1.0
2.4797	17.0	663	2.2529	1.0
2.2495	18.0	702	2.1600	1.0
1.8737	19.0	741	1.6194	1.0
1.69	20.0	780	1.4995	0.9999
1.5018	21.0	819	1.3398	0.9834
1.3264	22.0	858	1.2688	0.9692
1.1944	23.0	897	1.2211	0.9585
1.1186	24.0	936	1.1754	0.9517
1.0038	25.0	975	1.2082	0.9758
0.9096	26.0	1014	1.1463	0.9210
0.7954	27.0	1053	1.1530	0.9184
0.7337	28.0	1092	1.1948	0.9208
0.6438	29.0	1131	1.1907	0.9021
0.5933	30.0	1170	1.1994	0.9032
0.5646	31.0	1209	1.2765	0.9019
0.5314	32.0	1248	1.3331	0.9387
0.4208	33.0	1287	1.4003	0.9271
0.3769	34.0	1326	1.4226	0.9635
0.425	35.0	1365	1.3948	0.8890
0.3446	36.0	1404	1.4492	0.8901
0.3411	37.0	1443	1.5271	0.9136
0.3147	38.0	1482	1.4801	0.9139
0.2843	39.0	1521	1.5223	0.9011
0.2908	40.0	1560	1.6087	0.8871
0.2816	41.0	1599	1.5167	0.9097
0.2586	42.0	1638	1.5968	0.9129
0.2428	43.0	1677	1.6335	0.9100
0.2569	44.0	1716	1.5888	0.8967
0.2119	45.0	1755	1.6366	0.8910
0.2496	46.0	1794	1.6392	0.8807
0.2246	47.0	1833	1.6780	0.9197
0.2231	48.0	1872	1.7074	0.8969
0.2083	49.0	1911	1.6566	0.8811
0.2091	50.0	1950	1.6655	0.8832

Framework versions

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