ihanif/whisper-large-pashto

Whisper Large Pashto

This model is a fine-tuned version of openai/whisper-large-v2 on the google/fleurs ps_af dataset. It achieves the following results on the evaluation set:

• Loss: 0.8623
• Wer: 54.0685

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: 3e-07
• train_batch_size: 8
• eval_batch_size: 4
• seed: 42
• gradient_accumulation_steps: 8
• 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: 50
• training_steps: 700
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Wer
1.2281	16.59	100	1.0951	69.3118
0.7529	33.3	200	0.8693	57.5635
0.5372	49.89	300	0.8399	54.7350
0.4398	66.59	400	0.8623	54.0685
0.3244	83.3	500	0.9098	54.7505
0.238	99.89	600	0.9607	55.3782
0.2014	116.59	700	1.0077	55.9206

Framework versions

• Transformers 4.26.0.dev0
• Pytorch 1.13.0+cu117
• Datasets 2.7.1.dev0
• Tokenizers 0.13.2