wav2vec2-E10_speed_pause2

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: 1.2505
• Cer: 29.8296

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: 0.0001
• train_batch_size: 8
• eval_batch_size: 8
• seed: 42
• optimizer: Use adamw_torch with betas=(0.9,0.999) and epsilon=1e-08 and optimizer_args=No additional optimizer arguments
• lr_scheduler_type: linear
• lr_scheduler_warmup_steps: 50
• num_epochs: 3
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Cer
34.0244	0.1289	200	5.3369	100.0
5.0173	0.2579	400	4.7266	100.0
4.8452	0.3868	600	4.6853	100.0
4.7774	0.5158	800	4.5774	97.9377
4.6964	0.6447	1000	4.5999	97.9495
4.6269	0.7737	1200	4.4835	96.6569
4.4084	0.9026	1400	4.5428	96.0811
3.6551	1.0316	1600	3.4889	67.0094
2.8629	1.1605	1800	2.8535	54.0541
2.444	1.2895	2000	2.3817	48.1610
2.1451	1.4184	2200	2.1738	44.4418
1.8677	1.5474	2400	2.0255	42.4794
1.7019	1.6763	2600	1.8023	38.5546
1.5671	1.8053	2800	1.6997	36.3749
1.4107	1.9342	3000	1.6382	38.1316
1.2989	2.0632	3200	1.5183	33.8837
1.1951	2.1921	3400	1.4678	33.1610
1.0867	2.3211	3600	1.4365	33.0552
1.083	2.4500	3800	1.2930	30.1234
1.0106	2.5790	4000	1.2864	30.2585
0.9832	2.7079	4200	1.2784	30.0411
0.9353	2.8369	4400	1.2633	29.9471
0.9108	2.9658	4600	1.2505	29.8296

Framework versions

• Transformers 4.46.2
• Pytorch 2.5.1+cu121
• Datasets 3.1.0
• Tokenizers 0.20.3