wav2vec2-gpt2-enc-dec

This model is a fine-tuned version of on the common_voice_17_0 dataset. It achieves the following results on the evaluation set:

• Loss: 0.3276
• Wer: 0.8489

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: 1e-05
• train_batch_size: 16
• eval_batch_size: 16
• 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_ratio: 0.08
• num_epochs: 20
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Wer
1.9498	1.5625	50	0.6548	0.9324
0.4531	3.125	100	0.3959	0.9020
0.4087	4.6875	150	0.3735	0.8894
0.3992	6.25	200	0.3572	0.8747
0.3725	7.8125	250	0.3500	0.8763
0.3635	9.375	300	0.3419	0.8626
0.3647	10.9375	350	0.3381	0.8632
0.36	12.5	400	0.3340	0.8566
0.3588	14.0625	450	0.3316	0.8547
0.362	15.625	500	0.3299	0.8547
0.3613	17.1875	550	0.3280	0.8498
0.3505	18.75	600	0.3276	0.8489

Framework versions

• Transformers 4.48.3
• Pytorch 2.6.0+cu124
• Datasets 3.2.0
• Tokenizers 0.21.0