prova_codice_01

This model is a fine-tuned version of facebook/wav2vec2-xls-r-300m on the common_voice_11_0 dataset. It achieves the following results on the evaluation set:

• Loss: 0.3804
• Wer: 0.2560

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: 0.0003
• train_batch_size: 16
• eval_batch_size: 8
• seed: 42
• gradient_accumulation_steps: 2
• total_train_batch_size: 32
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: linear
• lr_scheduler_warmup_steps: 500
• num_epochs: 30
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Wer
0.6378	3.39	400	0.3791	0.3805
0.2294	6.78	800	0.3722	0.3490
0.1374	10.17	1200	0.3786	0.3267
0.0994	13.56	1600	0.3894	0.3170
0.0759	16.95	2000	0.3841	0.2858
0.0567	20.34	2400	0.3869	0.2764
0.0441	23.73	2800	0.3841	0.2602
0.0355	27.12	3200	0.3804	0.2560

Framework versions

• Transformers 4.28.1
• Pytorch 2.0.0+cu118
• Datasets 2.11.0
• Tokenizers 0.13.3