ihanif/xls-r-1b-pashto

facebook/wav2vec2-xls-r-1b

This model is a fine-tuned version of facebook/wav2vec2-xls-r-1b on the GOOGLE/FLEURS - PS_AF dataset. It achieves the following results on the evaluation set:

• Loss: 4.1921
• Wer: 0.9295
• Cer: 0.9608

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: 7.5e-05
• train_batch_size: 8
• eval_batch_size: 8
• seed: 42
• gradient_accumulation_steps: 4
• 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: 1000
• training_steps: 1000
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Cer	Validation Loss	Wer
19.9558	1.27	100	3.2660	20.9197	1.0
19.7186	2.53	200	1.1692	19.2447	1.0
15.203	3.8	300	0.9687	15.0053	0.9998
6.4303	5.06	400	0.9911	6.5437	0.9632
4.5712	6.33	500	0.9546	4.9040	0.9323
3.3986	12.66	1000	4.1921	0.9295	0.9608

Framework versions

• Transformers 4.26.0.dev0
• Pytorch 1.13.1+cu117
• Datasets 2.7.1.dev0
• Tokenizers 0.13.2