Whisper Large-V2 Basque

This model is a fine-tuned version of openai/whisper-large-v2 on the mozilla-foundation/common_voice_16_1 eu dataset. It achieves the following results on the evaluation set:

  • Loss: 0.4206
  • Wer: 7.7204

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: 32
  • eval_batch_size: 16
  • seed: 42
  • gradient_accumulation_steps: 8
  • total_train_batch_size: 256
  • optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
  • lr_scheduler_type: linear
  • lr_scheduler_warmup_steps: 500
  • training_steps: 40000
  • mixed_precision_training: Native AMP

Training results

Training Loss Epoch Step Validation Loss Wer
0.0112 10.04 1000 0.2182 10.1571
0.0052 20.08 2000 0.2372 9.6276
0.0017 30.11 3000 0.2417 9.0150
0.0022 40.15 4000 0.2341 8.8938
0.0023 50.19 5000 0.2451 8.9388
0.0006 60.23 6000 0.2517 8.4161
0.0006 70.26 7000 0.2499 8.0985
0.0008 80.3 8000 0.2548 8.3467
0.0004 90.34 9000 0.2498 7.9559
0.0003 100.38 10000 0.2489 7.6940
0.0 110.41 11000 0.2906 7.5455
0.0 120.45 12000 0.3027 7.4596
0.0 130.49 13000 0.3137 7.4517
0.0 140.53 14000 0.3243 7.4644
0.0 150.56 15000 0.3351 7.4762
0.0 160.6 16000 0.3459 7.4556
0.0 170.64 17000 0.3565 7.4605
0.0 180.68 18000 0.3689 7.4996
0.0 190.72 19000 0.3806 7.5934
0.0 200.75 20000 0.3912 7.6344
0.0 210.79 21000 0.4005 7.5485
0.0 220.83 22000 0.4102 7.6266
0.0079 230.87 23000 0.2467 9.1654
0.0 240.9 24000 0.3097 7.7615
0.0 250.94 25000 0.3311 7.7243
0.0 260.98 26000 0.3446 7.7028
0.0 271.02 27000 0.3551 7.7546
0.0 281.05 28000 0.3646 7.7986
0.0 291.09 29000 0.3729 7.7781
0.0 301.13 30000 0.3811 7.7634
0.0 311.17 31000 0.3878 7.7702
0.0 321.2 32000 0.3948 7.7722
0.0 331.24 33000 0.4003 7.7302
0.0 341.28 34000 0.4058 7.7312
0.0 351.32 35000 0.4108 7.7292
0.0 361.36 36000 0.4142 7.7321
0.0 371.39 37000 0.4170 7.7204
0.0 381.43 38000 0.4189 7.7253
0.0 391.47 39000 0.4202 7.7263
0.0 401.51 40000 0.4206 7.7204

Framework versions

  • Transformers 4.37.2
  • Pytorch 2.2.0+cu121
  • Datasets 2.16.1
  • Tokenizers 0.15.1
Downloads last month
30
Safetensors
Model size
1.54B params
Tensor type
F32
·
Inference Examples
This model does not have enough activity to be deployed to Inference API (serverless) yet. Increase its social visibility and check back later, or deploy to Inference Endpoints (dedicated) instead.

Model tree for zuazo/whisper-large-v2-eu-cv16_1

Finetuned
(191)
this model
Finetunes
1 model

Dataset used to train zuazo/whisper-large-v2-eu-cv16_1

Evaluation results