wav2vec2-E30_speed_pause

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: 2.6351
• Cer: 61.7540

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: Adam with betas=(0.9,0.999) and epsilon=1e-08
• 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
32.976	0.1289	200	5.1291	100.0
4.9784	0.2579	400	4.7411	100.0
4.8371	0.3868	600	4.7057	100.0
4.7941	0.5158	800	4.7901	100.0
4.7063	0.6447	1000	4.7698	100.0
4.6822	0.7737	1200	4.6661	100.0
4.654	0.9026	1400	4.5412	100.0
4.5661	1.0316	1600	4.4702	99.9765
4.4285	1.1605	1800	4.3777	98.7253
4.1372	1.2895	2000	4.1695	97.1452
3.8557	1.4184	2200	3.8556	73.7664
3.5494	1.5474	2400	3.6468	70.5298
3.2487	1.6763	2600	3.3393	68.9086
3.0583	1.8053	2800	3.1815	67.4636
2.9404	1.9342	3000	3.1664	67.4283
2.8	2.0632	3200	2.9573	64.5266
2.6974	2.1921	3400	2.8951	63.7747
2.6229	2.3211	3600	2.8758	64.7909
2.5551	2.4500	3800	2.8143	63.4633
2.4771	2.5790	4000	2.7618	63.8628
2.4412	2.7079	4200	2.7518	64.2094
2.375	2.8369	4400	2.6586	62.5235
2.3835	2.9658	4600	2.6351	61.7540

Framework versions

• Transformers 4.44.2
• Pytorch 2.4.1+cu121
• Datasets 3.0.1
• Tokenizers 0.19.1