wav2vec2-E50_feq_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: 4.2476
• Cer: 85.2526

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
48.4957	0.1289	200	6.8888	100.0
20.9588	0.2579	400	4.8707	97.5029
13.3764	0.3868	600	4.5300	91.7685
15.9565	0.5158	800	4.5414	92.9553
11.0676	0.6447	1000	4.6375	93.8367
5.7987	0.7737	1200	4.4724	93.2902
4.9382	0.9026	1400	4.4337	93.8543
7.107	1.0316	1600	4.4191	93.6428
5.0628	1.1605	1800	4.4137	94.3184
6.2941	1.2895	2000	4.4307	93.0317
5.6561	1.4184	2200	4.3684	93.0905
5.6863	1.5474	2400	4.3545	91.8743
5.9443	1.6763	2600	4.3462	91.3925
5.3199	1.8053	2800	4.3394	90.0823
5.9598	1.9342	3000	4.3221	90.0353
5.5016	2.0632	3200	4.3175	90.3055
5.7831	2.1921	3400	4.3053	89.9236
4.7223	2.3211	3600	4.2815	89.1657
5.0472	2.4500	3800	4.2367	88.4548
4.835	2.5790	4000	4.2107	85.7168
4.6838	2.7079	4200	4.2758	89.1833
4.7761	2.8369	4400	4.2710	85.8813
4.8559	2.9658	4600	4.2476	85.2526

Framework versions

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