XLS-R_Synthesis_GN_v1

This model is a fine-tuned version of facebook/wav2vec2-xls-r-300m on the None dataset. It achieves the following results on the evaluation set:

• Loss: 0.5071
• Wer: 0.3125

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: 18
• eval_batch_size: 9
• seed: 42
• distributed_type: multi-GPU
• num_devices: 2
• gradient_accumulation_steps: 2
• total_train_batch_size: 72
• total_eval_batch_size: 18
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: polynomial
• lr_scheduler_warmup_ratio: 0.1
• num_epochs: 100.0
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Wer
5.5691	1.0	1252	2.9844	1.0
1.7927	2.0	2505	0.5051	0.5816
0.5062	3.0	3757	0.3025	0.3928
0.3566	4.0	5010	0.2515	0.3208
0.2883	5.0	6262	0.2237	0.2811
0.2474	6.0	7515	0.2055	0.2538
0.223	7.0	8767	0.1929	0.2394
0.2068	8.0	10020	0.1957	0.2354
0.2043	9.0	11272	0.2222	0.2486
0.3093	10.0	12525	0.3168	0.3350
0.2501	11.0	13777	0.2099	0.2458
0.3961	12.0	15030	0.5071	0.3125

Framework versions

• Transformers 4.35.2
• Pytorch 2.1.1+cu121
• Datasets 2.15.0
• Tokenizers 0.15.0