ihanif/pashto-asr-base

Whisper Base Pashto - Augmented

This model is a fine-tuned version of openai/whisper-base on the google/fleurs dataset. It achieves the following results on the evaluation set:

• Loss: 0.7901
• Wer: 59.6482
• Cer: 27.0947

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: 8
• seed: 42
• gradient_accumulation_steps: 4
• total_train_batch_size: 64
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: linear
• lr_scheduler_warmup_steps: 30
• training_steps: 600
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Wer	Cer
1.1215	2.38	100	0.9444	68.3354	30.2694
0.8268	4.75	200	0.8267	63.2440	28.2636
0.6912	7.14	300	0.7959	62.2443	28.2123
0.5725	9.52	400	0.7896	60.5859	27.6920
0.5231	11.89	500	0.7884	59.8574	27.1273
0.4752	14.28	600	0.7901	59.6482	27.0947

Framework versions

• Transformers 4.26.0.dev0
• Pytorch 1.13.1+cu116
• Datasets 2.8.1.dev0
• Tokenizers 0.13.2