Edit model card

he-cantillation

This model is a fine-tuned version of openai/whisper-small on an unknown dataset. It achieves the following results on the evaluation set:

  • Loss: 0.2146
  • Wer: 12.3134
  • Avg Precision Exact: 0.8990
  • Avg Recall Exact: 0.9012
  • Avg F1 Exact: 0.8997
  • Avg Precision Letter Shift: 0.9196
  • Avg Recall Letter Shift: 0.9219
  • Avg F1 Letter Shift: 0.9204
  • Avg Precision Word Level: 0.9216
  • Avg Recall Word Level: 0.9240
  • Avg F1 Word Level: 0.9224
  • Avg Precision Word Shift: 0.9730
  • Avg Recall Word Shift: 0.9761
  • Avg F1 Word Shift: 0.9741
  • Precision Median Exact: 1.0
  • Recall Median Exact: 1.0
  • F1 Median Exact: 1.0
  • Precision Max Exact: 1.0
  • Recall Max Exact: 1.0
  • F1 Max Exact: 1.0
  • Precision Min Exact: 0.0
  • Recall Min Exact: 0.0
  • F1 Min Exact: 0.0
  • Precision Min Letter Shift: 0.0
  • Recall Min Letter Shift: 0.0
  • F1 Min Letter Shift: 0.0
  • Precision Min Word Level: 0.0
  • Recall Min Word Level: 0.0
  • F1 Min Word Level: 0.0
  • Precision Min Word Shift: 0.1429
  • Recall Min Word Shift: 0.1
  • F1 Min Word Shift: 0.1176

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: 8
  • eval_batch_size: 32
  • seed: 42
  • optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
  • lr_scheduler_type: linear
  • lr_scheduler_warmup_steps: 500
  • training_steps: 200000
  • mixed_precision_training: Native AMP

Training results

Training Loss Epoch Step Validation Loss Wer Avg Precision Exact Avg Recall Exact Avg F1 Exact Avg Precision Letter Shift Avg Recall Letter Shift Avg F1 Letter Shift Avg Precision Word Level Avg Recall Word Level Avg F1 Word Level Avg Precision Word Shift Avg Recall Word Shift Avg F1 Word Shift Precision Median Exact Recall Median Exact F1 Median Exact Precision Max Exact Recall Max Exact F1 Max Exact Precision Min Exact Recall Min Exact F1 Min Exact Precision Min Letter Shift Recall Min Letter Shift F1 Min Letter Shift Precision Min Word Level Recall Min Word Level F1 Min Word Level Precision Min Word Shift Recall Min Word Shift F1 Min Word Shift
No log 8e-05 1 7.1551 102.7088 0.0004 0.0034 0.0008 0.0169 0.0167 0.0165 0.0045 0.0485 0.0082 0.0865 0.0873 0.0863 0.0 0.0 0.0 0.1111 1.0 0.2000 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0
0.0271 0.8 10000 0.1443 15.7539 0.8698 0.8704 0.8696 0.8921 0.8928 0.8920 0.8951 0.8962 0.8951 0.9611 0.9641 0.9620 0.9333 0.9333 0.9474 1.0 1.0 1.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.1429 0.125 0.1333
0.0138 1.6 20000 0.1523 14.3570 0.8784 0.8793 0.8784 0.9010 0.9020 0.9010 0.9042 0.9053 0.9043 0.9672 0.9691 0.9676 0.9737 1.0 0.9565 1.0 1.0 1.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0769 0.0769 0.0769
0.0059 2.4 30000 0.1584 17.4242 0.8838 0.8843 0.8835 0.9054 0.9061 0.9051 0.9077 0.9090 0.9076 0.9658 0.9681 0.9661 1.0 1.0 0.9655 1.0 1.0 1.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0
0.0038 3.2 40000 0.1696 13.3925 0.8917 0.8933 0.8921 0.9129 0.9147 0.9134 0.9155 0.9171 0.9159 0.9724 0.9743 0.9729 1.0 1.0 0.9677 1.0 1.0 1.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.1429 0.1 0.1176
0.0039 4.0 50000 0.1695 13.2779 0.8819 0.8816 0.8814 0.9029 0.9027 0.9024 0.9052 0.9052 0.9048 0.9707 0.9714 0.9706 1.0 1.0 0.9697 1.0 1.0 1.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.1429 0.125 0.1333
0.0027 4.8 60000 0.1823 13.4294 0.8837 0.8852 0.8841 0.9047 0.9066 0.9052 0.9073 0.9094 0.9078 0.9699 0.9729 0.9708 1.0 1.0 1.0 1.0 1.0 1.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.1429 0.1 0.1176
0.0008 5.6 70000 0.1762 13.0340 0.8870 0.8902 0.8883 0.9071 0.9104 0.9084 0.9097 0.9131 0.9110 0.9710 0.9751 0.9726 1.0 1.0 1.0 1.0 1.0 1.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.1429 0.1111 0.125
0.0006 6.4 80000 0.1845 12.8012 0.8926 0.8943 0.8931 0.9131 0.9150 0.9137 0.9155 0.9174 0.9161 0.9729 0.9753 0.9737 1.0 1.0 1.0 1.0 1.0 1.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.1429 0.1 0.1176
0.0009 7.2 90000 0.1883 12.6497 0.8960 0.8943 0.8948 0.9173 0.9158 0.9162 0.9197 0.9184 0.9187 0.9745 0.9747 0.9742 1.0 1.0 1.0 1.0 1.0 1.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.1429 0.1 0.1176
0.0003 8.0 100000 0.1888 12.3688 0.8960 0.8977 0.8965 0.9169 0.9186 0.9174 0.9192 0.9211 0.9198 0.9745 0.9772 0.9755 1.0 1.0 1.0 1.0 1.0 1.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.1429 0.125 0.1333
0.0023 8.8 110000 0.1934 12.5240 0.8962 0.8948 0.8951 0.9162 0.9148 0.9151 0.9184 0.9172 0.9174 0.9737 0.9740 0.9734 1.0 1.0 1.0 1.0 1.0 1.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.1429 0.1 0.1176
0.0005 9.6 120000 0.2024 12.6534 0.8919 0.8935 0.8924 0.9122 0.9139 0.9127 0.9146 0.9163 0.9151 0.9720 0.9745 0.9728 1.0 1.0 1.0 1.0 1.0 1.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.1429 0.1111 0.125
0.0001 10.4 130000 0.1999 13.0229 0.8901 0.8925 0.8909 0.9109 0.9135 0.9118 0.9134 0.9160 0.9143 0.9714 0.9749 0.9727 1.0 1.0 1.0 1.0 1.0 1.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.1429 0.125 0.1333
0.0002 11.2 140000 0.2010 12.9675 0.8873 0.8920 0.8892 0.9085 0.9135 0.9106 0.9112 0.9163 0.9133 0.9698 0.9752 0.9720 1.0 1.0 1.0 1.0 1.0 1.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.1429 0.125 0.1333
0.0001 12.0 150000 0.2078 12.8455 0.8932 0.8961 0.8943 0.9149 0.9181 0.9161 0.9170 0.9202 0.9182 0.9724 0.9762 0.9739 1.0 1.0 1.0 1.0 1.0 1.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.1429 0.1111 0.125
0.0001 12.8 160000 0.2063 12.6718 0.8950 0.8962 0.8952 0.9161 0.9175 0.9164 0.9183 0.9198 0.9187 0.9731 0.9757 0.9740 1.0 1.0 1.0 1.0 1.0 1.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.1429 0.1111 0.125
0.0005 13.6 170000 0.2137 12.5203 0.8941 0.8960 0.8947 0.9143 0.9164 0.9150 0.9165 0.9186 0.9172 0.9712 0.9742 0.9723 1.0 1.0 1.0 1.0 1.0 1.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0909 0.1 0.1000
0.0002 14.4 180000 0.2121 12.5979 0.8945 0.8964 0.8951 0.9155 0.9176 0.9162 0.9178 0.9199 0.9185 0.9731 0.9760 0.9741 1.0 1.0 1.0 1.0 1.0 1.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.1429 0.1111 0.125
0.0 15.2 190000 0.2143 12.3282 0.8985 0.9008 0.8993 0.9193 0.9218 0.9202 0.9215 0.9240 0.9224 0.9734 0.9766 0.9746 1.0 1.0 1.0 1.0 1.0 1.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.1429 0.125 0.1333
0.0 16.0 200000 0.2146 12.3134 0.8990 0.9012 0.8997 0.9196 0.9219 0.9204 0.9216 0.9240 0.9224 0.9730 0.9761 0.9741 1.0 1.0 1.0 1.0 1.0 1.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.1429 0.1 0.1176

Framework versions

  • Transformers 4.41.2
  • Pytorch 2.2.1
  • Datasets 2.20.0
  • Tokenizers 0.19.1
Downloads last month
0
Safetensors
Model size
242M params
Tensor type
F32
·
Inference API
Unable to determine this model's library. Check the docs .

Model tree for cantillation/Teamim-small_Random_WeightDecay-0.05_Augmented_New-Data_date-02-08-2024

Finetuned
(1867)
this model