Hugging Face's logo

asr-africa
/
mms-1B_all_NCHLT_ZULU_20hr_v1

Automatic Speech Recognition
Transformers
Safetensors
Zulu
wav2vec2
Generated from Trainer
Eval Results
Inference Endpoints
Model card Files Community
mms-1B_all_NCHLT_ZULU_20hr_v1 / README.md
Beijuka's picture
Beijuka
End of training
4f46add verified
preview code
|
raw
history blame contribute delete
8.75 kB
metadata
library_name: transformers
language:
  - zul
license: cc-by-nc-4.0
base_model: facebook/mms-1b-all
tags:
  - generated_from_trainer
datasets:
  - NCHLT_speech_corpus
metrics:
  - wer
model-index:
  - name: facebook mms-1b-all zulu - Beijuka Bruno
    results:
      - task:
          name: Automatic Speech Recognition
          type: automatic-speech-recognition
        dataset:
          name: NCHLT_speech_corpus/Zulu
          type: NCHLT_speech_corpus
        metrics:
          - name: Wer
            type: wer
            value: 0.47644382219110953

facebook mms-1b-all zulu - Beijuka Bruno

This model is a fine-tuned version of facebook/mms-1b-all on the NCHLT_speech_corpus/Zulu dataset. It achieves the following results on the evaluation set:

  • Loss: 0.2982
  • Model Preparation Time: 0.0133
  • Wer: 0.4764
  • Cer: 0.0937

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: 0.0003
  • train_batch_size: 4
  • eval_batch_size: 8
  • seed: 42
  • gradient_accumulation_steps: 8
  • total_train_batch_size: 32
  • optimizer: Use adamw_torch with betas=(0.9,0.999) and epsilon=1e-08 and optimizer_args=No additional optimizer arguments
  • lr_scheduler_type: linear
  • lr_scheduler_warmup_steps: 100
  • num_epochs: 100

Training results

Training Loss Epoch Step Validation Loss Model Preparation Time Wer Cer
1.8637 0.9989 568 0.1607 0.0133 0.2395 0.0374
0.1869 1.9989 1136 0.1470 0.0133 0.2161 0.0338
0.1694 2.9989 1704 0.1361 0.0133 0.1968 0.0311
0.153 3.9989 2272 0.1263 0.0133 0.1840 0.0294
0.1413 4.9989 2840 0.1232 0.0133 0.1727 0.0278
0.1359 5.9989 3408 0.1170 0.0133 0.1675 0.0268
0.1271 6.9989 3976 0.1132 0.0133 0.1623 0.0264
0.1202 7.9989 4544 0.1102 0.0133 0.1570 0.0254
0.1161 8.9989 5112 0.1105 0.0133 0.1549 0.0251
0.111 9.9989 5680 0.1036 0.0133 0.1501 0.0244
0.1075 10.9989 6248 0.1022 0.0133 0.1429 0.0235
0.1036 11.9989 6816 0.1016 0.0133 0.1414 0.0233
0.0992 12.9989 7384 0.1009 0.0133 0.1387 0.0227
0.0951 13.9989 7952 0.0986 0.0133 0.1370 0.0224
0.0931 14.9989 8520 0.0964 0.0133 0.1331 0.0221
0.0901 15.9989 9088 0.0958 0.0133 0.1296 0.0215
0.0887 16.9989 9656 0.0942 0.0133 0.1304 0.0214
0.0845 17.9989 10224 0.0935 0.0133 0.1247 0.0208
0.082 18.9989 10792 0.0938 0.0133 0.1234 0.0206
0.079 19.9989 11360 0.0922 0.0133 0.1229 0.0206
0.0773 20.9989 11928 0.0900 0.0133 0.1191 0.0199
0.0749 21.9989 12496 0.0907 0.0133 0.1191 0.0199
0.0732 22.9989 13064 0.0911 0.0133 0.1188 0.0198
0.0714 23.9989 13632 0.0883 0.0133 0.1157 0.0194
0.0687 24.9989 14200 0.0909 0.0133 0.1180 0.0197
0.0676 25.9989 14768 0.0889 0.0133 0.1072 0.0182
0.0659 26.9989 15336 0.0875 0.0133 0.1109 0.0185
0.0632 27.9989 15904 0.0876 0.0133 0.1071 0.0183
0.0626 28.9989 16472 0.0894 0.0133 0.1099 0.0184
0.0606 29.9989 17040 0.0845 0.0133 0.1036 0.0176
0.0583 30.9989 17608 0.0901 0.0133 0.1073 0.0181
0.057 31.9989 18176 0.0879 0.0133 0.1021 0.0174
0.0562 32.9989 18744 0.0878 0.0133 0.1042 0.0179
0.0556 33.9989 19312 0.0878 0.0133 0.1023 0.0175
0.0543 34.9989 19880 0.0825 0.0133 0.0983 0.0167
0.0531 35.9989 20448 0.0846 0.0133 0.0982 0.0167
0.0513 36.9989 21016 0.0858 0.0133 0.1010 0.0171
0.0499 37.9989 21584 0.0878 0.0133 0.1002 0.0168
0.0492 38.9989 22152 0.0872 0.0133 0.0988 0.0166
0.0474 39.9989 22720 0.0863 0.0133 0.0964 0.0166
0.0467 40.9989 23288 0.0877 0.0133 0.0986 0.0167
0.0455 41.9989 23856 0.0897 0.0133 0.0969 0.0167
0.0454 42.9989 24424 0.0892 0.0133 0.0960 0.0164
0.0437 43.9989 24992 0.0897 0.0133 0.0979 0.0168
0.0423 44.9989 25560 0.0888 0.0133 0.0924 0.0159
0.0415 45.9989 26128 0.0899 0.0133 0.0919 0.0160
0.0412 46.9989 26696 0.0906 0.0133 0.0934 0.0161
0.0397 47.9989 27264 0.0854 0.0133 0.0889 0.0153
0.0388 48.9989 27832 0.0904 0.0133 0.0936 0.0160
0.0386 49.9989 28400 0.0903 0.0133 0.0892 0.0156
0.0381 50.9989 28968 0.0879 0.0133 0.0921 0.0160
0.0378 51.9989 29536 0.0876 0.0133 0.0893 0.0155
0.0378 52.9989 30104 0.0887 0.0133 0.0889 0.0155
0.0353 53.9989 30672 0.0902 0.0133 0.0887 0.0155
0.0351 54.9989 31240 0.0904 0.0133 0.0911 0.0158
0.0332 55.9989 31808 0.0881 0.0133 0.0854 0.0151
0.0349 56.9989 32376 0.0892 0.0133 0.0903 0.0155
0.0339 57.9989 32944 0.0919 0.0133 0.0876 0.0154
0.032 58.9989 33512 0.0892 0.0133 0.0862 0.0148
0.0316 59.9989 34080 0.0881 0.0133 0.0831 0.0149
0.0304 60.9989 34648 0.0892 0.0133 0.0830 0.0147
0.0314 61.9989 35216 0.0880 0.0133 0.0856 0.0151
0.0317 62.9989 35784 0.0891 0.0133 0.0870 0.0149
0.0304 63.9989 36352 0.0883 0.0133 0.0825 0.0148
0.0298 64.9989 36920 0.0923 0.0133 0.0839 0.0148
0.0303 65.9989 37488 0.0909 0.0133 0.0841 0.0147
0.0297 66.9989 38056 0.0906 0.0133 0.0845 0.0148
0.0293 67.9989 38624 0.0921 0.0133 0.0843 0.0147
0.0279 68.9989 39192 0.0934 0.0133 0.0823 0.0144
0.0268 69.9989 39760 0.0935 0.0133 0.0833 0.0144

Framework versions

  • Transformers 4.47.0
  • Pytorch 2.1.0+cu118
  • Datasets 3.2.0
  • Tokenizers 0.21.0