<!-- This model card has been generated automatically according to the information the Trainer had access to. You should probably proofread and complete it, then remove this comment. -->
wav2vec2-xls-r-300m-arabic_speech_commands_10s_one_speaker
This model is a fine-tuned version of facebook/wav2vec2-xls-r-300m on the None dataset. It achieves the following results on the evaluation set:
- Loss: 0.0825
- Accuracy: 0.9833
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: 4
- seed: 42
- gradient_accumulation_steps: 4
- total_train_batch_size: 16
- optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
- lr_scheduler_type: linear
- lr_scheduler_warmup_ratio: 0.1
- num_epochs: 80
Training results
Training Loss | Epoch | Step | Validation Loss | Accuracy |
---|---|---|---|---|
No log | 0.92 | 3 | 1.6103 | 0.2 |
No log | 1.92 | 6 | 1.6087 | 0.2 |
No log | 2.92 | 9 | 1.6056 | 0.3033 |
1.7251 | 3.92 | 12 | 1.6021 | 0.2 |
1.7251 | 4.92 | 15 | 1.5831 | 0.2533 |
1.7251 | 5.92 | 18 | 1.5251 | 0.5367 |
1.6654 | 6.92 | 21 | 1.3294 | 0.74 |
1.6654 | 7.92 | 24 | 1.0468 | 0.7567 |
1.6654 | 8.92 | 27 | 0.9286 | 0.8533 |
1.2188 | 9.92 | 30 | 0.7191 | 0.83 |
1.2188 | 10.92 | 33 | 0.6482 | 0.87 |
1.2188 | 11.92 | 36 | 0.6057 | 0.8433 |
1.2188 | 12.92 | 39 | 0.3797 | 0.9467 |
0.6897 | 13.92 | 42 | 0.3407 | 0.9267 |
0.6897 | 14.92 | 45 | 0.2305 | 0.95 |
0.6897 | 15.92 | 48 | 0.2806 | 0.92 |
0.2498 | 16.92 | 51 | 0.2146 | 0.9467 |
0.2498 | 17.92 | 54 | 0.2850 | 0.9233 |
0.2498 | 18.92 | 57 | 0.1945 | 0.9467 |
0.0658 | 19.92 | 60 | 0.0825 | 0.9833 |
0.0658 | 20.92 | 63 | 0.1280 | 0.9767 |
0.0658 | 21.92 | 66 | 0.1024 | 0.98 |
0.0658 | 22.92 | 69 | 0.1102 | 0.9767 |
0.044 | 23.92 | 72 | 0.0780 | 0.98 |
0.044 | 24.92 | 75 | 0.2282 | 0.94 |
0.044 | 25.92 | 78 | 0.0923 | 0.9767 |
0.153 | 26.92 | 81 | 0.0950 | 0.9767 |
0.153 | 27.92 | 84 | 0.1508 | 0.9633 |
0.153 | 28.92 | 87 | 0.0844 | 0.9833 |
0.0301 | 29.92 | 90 | 0.1224 | 0.97 |
0.0301 | 30.92 | 93 | 0.1852 | 0.95 |
0.0301 | 31.92 | 96 | 0.2527 | 0.94 |
0.0301 | 32.92 | 99 | 0.3035 | 0.9367 |
0.0092 | 33.92 | 102 | 0.4948 | 0.8967 |
0.0092 | 34.92 | 105 | 0.3825 | 0.9167 |
0.0092 | 35.92 | 108 | 0.1381 | 0.9733 |
0.0437 | 36.92 | 111 | 0.1392 | 0.9667 |
0.0437 | 37.92 | 114 | 0.1889 | 0.95 |
0.0437 | 38.92 | 117 | 0.2354 | 0.94 |
0.006 | 39.92 | 120 | 0.2462 | 0.9433 |
0.006 | 40.92 | 123 | 0.1140 | 0.97 |
0.006 | 41.92 | 126 | 0.3629 | 0.94 |
0.006 | 42.92 | 129 | 0.1350 | 0.97 |
0.0458 | 43.92 | 132 | 0.0908 | 0.98 |
0.0458 | 44.92 | 135 | 0.1091 | 0.98 |
0.0458 | 45.92 | 138 | 0.1283 | 0.9733 |
0.0063 | 46.92 | 141 | 0.1448 | 0.9733 |
0.0063 | 47.92 | 144 | 0.1495 | 0.9633 |
0.0063 | 48.92 | 147 | 0.1628 | 0.9633 |
0.0161 | 49.92 | 150 | 0.1812 | 0.96 |
0.0161 | 50.92 | 153 | 0.1954 | 0.96 |
0.0161 | 51.92 | 156 | 0.2043 | 0.96 |
0.0161 | 52.92 | 159 | 0.1715 | 0.9567 |
0.0989 | 53.92 | 162 | 0.1482 | 0.9633 |
0.0989 | 54.92 | 165 | 0.1305 | 0.97 |
0.0989 | 55.92 | 168 | 0.1269 | 0.97 |
0.0084 | 56.92 | 171 | 0.1302 | 0.9667 |
0.0084 | 57.92 | 174 | 0.1351 | 0.9667 |
0.0084 | 58.92 | 177 | 0.1408 | 0.9667 |
0.0054 | 59.92 | 180 | 0.1194 | 0.9767 |
0.0054 | 60.92 | 183 | 0.1110 | 0.98 |
0.0054 | 61.92 | 186 | 0.1064 | 0.98 |
0.0054 | 62.92 | 189 | 0.1050 | 0.9767 |
0.0031 | 63.92 | 192 | 0.1051 | 0.9767 |
0.0031 | 64.92 | 195 | 0.1047 | 0.98 |
0.0031 | 65.92 | 198 | 0.1121 | 0.98 |
0.0061 | 66.92 | 201 | 0.1170 | 0.9767 |
0.0061 | 67.92 | 204 | 0.1183 | 0.9733 |
0.0061 | 68.92 | 207 | 0.1181 | 0.9733 |
0.0032 | 69.92 | 210 | 0.1179 | 0.9733 |
0.0032 | 70.92 | 213 | 0.1000 | 0.98 |
0.0032 | 71.92 | 216 | 0.0883 | 0.98 |
0.0032 | 72.92 | 219 | 0.0787 | 0.9833 |
0.0029 | 73.92 | 222 | 0.0715 | 0.98 |
0.0029 | 74.92 | 225 | 0.0668 | 0.98 |
0.0029 | 75.92 | 228 | 0.0639 | 0.98 |
0.0027 | 76.92 | 231 | 0.0621 | 0.98 |
0.0027 | 77.92 | 234 | 0.0611 | 0.98 |
0.0027 | 78.92 | 237 | 0.0607 | 0.9833 |
0.0022 | 79.92 | 240 | 0.0606 | 0.9833 |
Framework versions
- Transformers 4.21.1
- Pytorch 1.12.1
- Datasets 2.4.0
- Tokenizers 0.12.1