Edit model card

whisper

This model is a fine-tuned version of bofenghuang/whisper-large-v3-french-distil-dec16 on an unknown dataset. It achieves the following results on the evaluation set:

  • Loss: 0.1122
  • Wer: 5.3589

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: 5e-06
  • train_batch_size: 8
  • eval_batch_size: 4
  • seed: 42
  • gradient_accumulation_steps: 2
  • total_train_batch_size: 16
  • optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
  • lr_scheduler_type: linear
  • lr_scheduler_warmup_steps: 20
  • training_steps: 200

Training results

Training Loss Epoch Step Validation Loss Wer
0.4042 0.38 20 0.2881 4.5501
0.3463 0.77 40 0.2060 4.3478
0.125 1.15 60 0.1498 4.7523
0.0606 1.54 80 0.1154 4.3478
0.0884 1.92 100 0.1026 4.8534
0.0189 2.31 120 0.0995 4.8534
0.0235 2.69 140 0.1085 4.6512
0.0126 3.08 160 0.1100 4.6512
0.0096 3.46 180 0.1114 5.2578
0.0214 3.85 200 0.1122 5.3589

Framework versions

  • Transformers 4.37.2
  • Pytorch 2.1.0+cu121
  • Datasets 2.17.0
  • Tokenizers 0.15.1
Downloads last month
2
Safetensors
Model size
1.12B params
Tensor type
F32
·

Finetuned from