Hugging Face's logo

qmeeus
/
whisper-large-multilingual-spoken-ner-pipeline-ft

Transformers
Safetensors
whisper_for_slu
whisper-event
Generated from Trainer
Eval Results
Inference Endpoints
Model card Files Community
whisper-large-multilingual-spoken-ner-pipeline-ft / README.md
Quentin Meeus
trained model
6d4bc00
preview code
|
raw history blame contribute delete
No virus
3.95 kB
metadata
license: apache-2.0
base_model: openai/whisper-large-v2
tags:
  - whisper-event
  - generated_from_trainer
datasets:
  - facebook/voxpopuli
metrics:
  - wer
model-index:
  - name: WhisperForSpokenNER
    results:
      - task:
          name: Automatic Speech Recognition
          type: automatic-speech-recognition
        dataset:
          name: facebook/voxpopuli de+es+fr+nl
          type: facebook/voxpopuli
          config: de+es+fr+nl
          split: None
        metrics:
          - name: Wer
            type: wer
            value: 0.06196300023221612

WhisperForSpokenNER

This model is a fine-tuned version of openai/whisper-large-v2 on the facebook/voxpopuli de+es+fr+nl dataset. It achieves the following results on the evaluation set:

  • Loss: 0.2797
  • F1 Score: 0.7918
  • Label F1: 0.8933
  • Wer: 0.0620

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: 4
  • eval_batch_size: 8
  • seed: 42
  • gradient_accumulation_steps: 32
  • total_train_batch_size: 128
  • optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
  • lr_scheduler_type: cosine
  • lr_scheduler_warmup_steps: 500
  • training_steps: 5000
  • mixed_precision_training: Native AMP

Training results

Training Loss Epoch Step Validation Loss F1 Score Label F1 Wer
0.1748 0.36 200 0.1706 0.6541 0.8032 0.0656
0.1754 0.71 400 0.1769 0.7194 0.8502 0.0674
0.1606 1.07 600 0.1856 0.6991 0.8407 0.0708
0.1282 1.43 800 0.1835 0.7455 0.8724 0.0728
0.131 1.79 1000 0.1762 0.7331 0.8691 0.0713
0.0804 2.14 1200 0.1792 0.7544 0.8744 0.0685
0.0712 2.5 1400 0.1833 0.75 0.8846 0.0691
0.0746 2.86 1600 0.1800 0.7554 0.8732 0.0738
0.0331 3.22 1800 0.1992 0.7757 0.8804 0.0702
0.0363 3.57 2000 0.1938 0.7625 0.8805 0.0688
0.037 3.93 2200 0.1986 0.7771 0.8865 0.0677
0.0153 4.29 2400 0.2125 0.7765 0.8794 0.0666
0.0144 4.65 2600 0.2115 0.7763 0.8922 0.0681
0.0148 5.0 2800 0.2180 0.7781 0.8891 0.0647
0.0058 5.36 3000 0.2310 0.7918 0.8913 0.0629
0.0058 5.72 3200 0.2268 0.7828 0.8938 0.0627
0.0036 6.08 3400 0.2462 0.7911 0.8937 0.0621
0.0019 6.43 3600 0.2493 0.7948 0.8950 0.0629
0.0016 6.79 3800 0.2543 0.7917 0.8980 0.0631
0.0009 7.15 4000 0.2667 0.7880 0.8944 0.0619
0.0007 7.51 4200 0.2735 0.7909 0.8934 0.0624
0.0007 7.86 4400 0.2756 0.7901 0.8926 0.0621
0.0005 8.22 4600 0.2779 0.7913 0.8931 0.0624
0.0004 8.58 4800 0.2795 0.7909 0.8932 0.0620
0.0005 8.94 5000 0.2797 0.7918 0.8933 0.0620

Framework versions

  • Transformers 4.37.0.dev0
  • Pytorch 2.1.0
  • Datasets 2.14.6
  • Tokenizers 0.14.1