xlsr-nm-nose

This model is a fine-tuned version of facebook/wav2vec2-large-xlsr-53 on the None dataset. It achieves the following results on the evaluation set:

• Loss: 2.1813
• Wer: 0.5491

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: 0.0004
• train_batch_size: 8
• eval_batch_size: 8
• seed: 42
• gradient_accumulation_steps: 2
• total_train_batch_size: 16
• 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: 132
• num_epochs: 100
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Wer
4.9501	6.6667	200	3.0654	1.0
2.9958	13.3333	400	2.8175	1.0
2.5806	20.0	600	2.2179	1.0
1.4475	26.6667	800	1.4198	0.7254
0.6646	33.3333	1000	1.5241	0.6384
0.3545	40.0	1200	1.5196	0.5915
0.2218	46.6667	1400	1.6719	0.5737
0.1683	53.3333	1600	1.8144	0.5737
0.1259	60.0	1800	1.8164	0.5603
0.09	66.6667	2000	2.0002	0.5424
0.0888	73.3333	2200	2.0032	0.5536
0.0589	80.0	2400	2.0328	0.5469
0.0471	86.6667	2600	2.0949	0.5603
0.0392	93.3333	2800	2.2458	0.5469
0.0317	100.0	3000	2.1813	0.5491

Framework versions

• Transformers 4.47.0.dev0
• Pytorch 2.4.0
• Datasets 3.0.1
• Tokenizers 0.20.0