wav2vec2-E50_speed2

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: 1.0444
• Cer: 21.8625

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
37.5369	0.1289	200	5.2051	100.0
4.8991	0.2579	400	4.6157	100.0
4.7556	0.3868	600	4.6579	100.0
4.7401	0.5158	800	4.7246	100.0
4.6787	0.6447	1000	4.5941	100.0
4.6531	0.7737	1200	4.5238	100.0
4.5631	0.9026	1400	4.3757	100.0
3.8885	1.0316	1600	3.1641	57.0682
2.9165	1.1605	1800	2.6892	47.8202
2.4332	1.2895	2000	2.2163	40.8226
2.1146	1.4184	2200	1.9082	35.0705
1.9011	1.5474	2400	1.8199	35.3702
1.7622	1.6763	2600	1.6833	32.0564
1.6191	1.8053	2800	1.5709	30.7109
1.4782	1.9342	3000	1.4401	28.7427
1.3359	2.0632	3200	1.3604	27.1093
1.2158	2.1921	3400	1.2156	24.7650
1.1436	2.3211	3600	1.1927	24.3948
1.1092	2.4500	3800	1.1650	24.4712
1.0546	2.5790	4000	1.0929	22.8966
1.0207	2.7079	4200	1.0990	22.8790
0.9799	2.8369	4400	1.0580	22.2679
0.9561	2.9658	4600	1.0444	21.8625

Framework versions

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