VladS159/whisper-medium-8000-steps_40_percent_8_speakers_synthetic_data_12_05_2026

Whisper Medium Ro - PEFT

This model is a fine-tuned version of openai/whisper-medium on the Common Voice 17.0 40% Synthetic 8 Speakers dataset. It achieves the following results on the evaluation set:

• Loss: 0.1109
• Wer: 5.8547

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: 8
• eval_batch_size: 8
• seed: 42
• gradient_accumulation_steps: 2
• total_train_batch_size: 16
• optimizer: Use adamw_8bit 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: 800
• training_steps: 8000
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Wer
0.1734	0.7253	1600	0.1488	7.6679
0.112	1.4506	3200	0.1242	6.5781
0.0774	2.1759	4800	0.1130	6.0714
0.0641	2.9012	6400	0.1095	5.8333
0.0525	3.6265	8000	0.1109	5.8547

Framework versions

• PEFT 0.18.1.dev0
• Transformers 4.57.1
• Pytorch 2.9.1+rocm6.4
• Datasets 3.6.0
• Tokenizers 0.22.1