lmejias/w2v2-small-baseline

This model is a fine-tuned version of Jzuluaga/wav2vec2-large-960h-lv60-self-en-atc-uwb-atcc-and-atcosim on the ATC Jzuluaga on uwb_atcc dataset. It achieves the following results on the evaluation set:

• Loss: 0.0230
• Wer: 0.0

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: 32
• eval_batch_size: 8
• seed: 42
• optimizer: Use OptimizerNames.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: 10
• num_epochs: 30
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Wer
0.4021	5.0	5	0.2164	12.5
0.215	10.0	10	0.0293	0.0
0.1191	15.0	15	0.0243	0.0
0.0453	20.0	20	0.0166	0.0
0.0384	25.0	25	0.0190	0.0
0.0663	30.0	30	0.0230	0.0

Framework versions

• Transformers 4.51.3
• Pytorch 2.9.1+cu128
• Datasets 4.4.1
• Tokenizers 0.21.4