wav2vec2-E10_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: 4.3504
• Cer: 92.7732

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
57.6816	0.1289	200	6.9208	100.0
30.0241	0.2579	400	4.9760	99.2186
25.7956	0.3868	600	5.0354	93.9248
26.9553	0.5158	800	4.7842	94.0834
10.9188	0.6447	1000	5.0386	93.0141
4.8381	0.7737	1200	4.8832	93.1669
5.8977	0.9026	1400	4.7683	93.9953
5.9801	1.0316	1600	4.4577	94.1011
4.624	1.1605	1800	4.4622	93.7720
4.5279	1.2895	2000	4.4747	93.8132
4.5665	1.4184	2200	4.4467	93.8895
4.5274	1.5474	2400	4.4445	94.0541
4.5253	1.6763	2600	4.4448	93.8719
4.4933	1.8053	2800	4.4757	93.5723
4.4909	1.9342	3000	4.4318	93.5781
4.4942	2.0632	3200	4.4363	93.0905
4.3985	2.1921	3400	4.4396	92.9965
4.3714	2.3211	3600	4.3876	92.6263
4.3536	2.4500	3800	4.3960	92.9788
4.3598	2.5790	4000	4.3974	92.9553
4.3408	2.7079	4200	4.3550	92.7497
4.3141	2.8369	4400	4.3579	92.9377
4.2794	2.9658	4600	4.3504	92.7732

Framework versions

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