Edit model card

whisper-tiny-en-US

This model is a fine-tuned version of openai/whisper-tiny on the PolyAI/minds14 dataset. It achieves the following results on the evaluation set:

  • Loss: 0.5874
  • Wer Ortho: 0.1978
  • Wer: 0.1939

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: 16
  • eval_batch_size: 16
  • seed: 42
  • optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
  • lr_scheduler_type: constant_with_warmup
  • lr_scheduler_warmup_steps: 5
  • training_steps: 400

Training results

Training Loss Epoch Step Validation Loss Wer Ortho Wer
No log 0.36 10 0.5445 0.1970 0.1930
No log 0.71 20 0.5571 0.1973 0.1939
0.0009 1.07 30 0.5571 0.1981 0.1936
0.0009 1.43 40 0.5535 0.1958 0.1920
0.0024 1.79 50 0.5568 0.1967 0.1930
0.0024 2.14 60 0.5679 0.2024 0.1996
0.0024 2.5 70 0.5574 0.1970 0.1944
0.0019 2.86 80 0.5454 0.1956 0.1936
0.0019 3.21 90 0.5541 0.2044 0.1982
0.0043 3.57 100 0.5330 0.1998 0.1936
0.0043 3.93 110 0.5524 0.1981 0.1958
0.0043 4.29 120 0.5482 0.1958 0.1933
0.0043 4.64 130 0.5554 0.1984 0.1952
0.0043 5.0 140 0.5634 0.1998 0.1949
0.001 5.36 150 0.5526 0.1990 0.1930
0.001 5.71 160 0.5511 0.1973 0.1920
0.001 6.07 170 0.5548 0.1953 0.1917
0.0006 6.43 180 0.5589 0.2047 0.1998
0.0006 6.79 190 0.5657 0.2027 0.1982
0.0004 7.14 200 0.5686 0.1961 0.1933
0.0004 7.5 210 0.5714 0.1944 0.1911
0.0004 7.86 220 0.5710 0.1976 0.1941
0.0002 8.21 230 0.5656 0.1973 0.1925
0.0002 8.57 240 0.5661 0.1981 0.1928
0.0001 8.93 250 0.5688 0.1984 0.1930
0.0001 9.29 260 0.5720 0.1967 0.1920
0.0001 9.64 270 0.5745 0.1970 0.1930
0.0001 10.0 280 0.5758 0.1964 0.1925
0.0001 10.36 290 0.5767 0.1970 0.1930
0.0001 10.71 300 0.5780 0.1978 0.1941
0.0001 11.07 310 0.5793 0.1978 0.1941
0.0001 11.43 320 0.5803 0.1976 0.1939
0.0001 11.79 330 0.5815 0.1973 0.1933
0.0001 12.14 340 0.5824 0.1973 0.1933
0.0001 12.5 350 0.5833 0.1973 0.1933
0.0001 12.86 360 0.5843 0.1970 0.1933
0.0001 13.21 370 0.5850 0.1976 0.1936
0.0001 13.57 380 0.5856 0.1978 0.1939
0.0001 13.93 390 0.5865 0.1978 0.1939
0.0001 14.29 400 0.5874 0.1978 0.1939

Framework versions

  • Transformers 4.31.0.dev0
  • Pytorch 1.12.1+cu116
  • Datasets 2.4.0
  • Tokenizers 0.12.1
Downloads last month
27
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 ptah23/whisper-tiny-en-US

Finetuned
(1218)
this model

Dataset used to train ptah23/whisper-tiny-en-US

Evaluation results