hoseinshr1055/op_dir_full_1epoch

Whisper Large fa - Mobin Tadbir Sharif

This model is a fine-tuned version of openai/whisper-small on the Common Voice 17.0 dataset. It achieves the following results on the evaluation set:

• Loss: 0.8106
• Wer: 91.6584

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: 0.001
• train_batch_size: 1
• eval_batch_size: 1
• seed: 42
• gradient_accumulation_steps: 32
• total_train_batch_size: 32
• optimizer: Use OptimizerNames.ADAMW_BNB with betas=(0.9,0.999) and epsilon=1e-08 and optimizer_args=No additional optimizer arguments
• lr_scheduler_type: linear
• lr_scheduler_warmup_steps: 1000
• num_epochs: 1
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Wer
1.581	0.0869	1000	2.5843	104.8093
2.2554	0.1738	2000	2.3093	111.6202
2.1214	0.2607	3000	2.1839	105.5556
2.024	0.3477	4000	2.1036	113.6312
1.9005	0.4346	5000	2.0217	135.5618
1.7344	0.5215	6000	1.8019	107.0619
1.4862	0.6084	7000	1.5560	101.7136
1.253	0.6953	8000	1.3641	106.0004
1.0361	0.7822	9000	1.1275	98.4010
0.8509	0.8692	10000	0.9458	97.6069
0.7212	0.9561	11000	0.8106	91.6584

Framework versions

• Transformers 4.49.0
• Pytorch 2.6.0+cu124
• Datasets 3.3.2
• Tokenizers 0.21.0