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: 0.2668
- Model Preparation Time: 0.0161
- Wer: 0.3832
- Cer: 0.0671
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 |
---|---|---|---|---|---|---|
75.5144 | 1.0 | 139 | 0.6716 | 0.0161 | 0.6021 | 0.1091 |
2.2356 | 2.0 | 278 | 0.1570 | 0.0161 | 0.2587 | 0.0378 |
1.6751 | 3.0 | 417 | 0.1414 | 0.0161 | 0.2355 | 0.0342 |
1.5484 | 4.0 | 556 | 0.1314 | 0.0161 | 0.2148 | 0.0308 |
1.4406 | 5.0 | 695 | 0.1275 | 0.0161 | 0.2100 | 0.0301 |
1.3547 | 6.0 | 834 | 0.1247 | 0.0161 | 0.1967 | 0.0288 |
1.3095 | 7.0 | 973 | 0.1191 | 0.0161 | 0.1909 | 0.0277 |
1.2359 | 8.0 | 1112 | 0.1176 | 0.0161 | 0.1865 | 0.0271 |
1.1879 | 9.0 | 1251 | 0.1173 | 0.0161 | 0.1818 | 0.0269 |
1.1513 | 10.0 | 1390 | 0.1181 | 0.0161 | 0.1818 | 0.0267 |
1.1037 | 11.0 | 1529 | 0.1145 | 0.0161 | 0.1797 | 0.0258 |
1.0629 | 12.0 | 1668 | 0.1090 | 0.0161 | 0.1726 | 0.0253 |
1.0302 | 13.0 | 1807 | 0.1087 | 0.0161 | 0.1705 | 0.0249 |
0.9804 | 14.0 | 1946 | 0.1059 | 0.0161 | 0.1654 | 0.0243 |
0.9649 | 15.0 | 2085 | 0.1098 | 0.0161 | 0.1644 | 0.0242 |
0.9074 | 16.0 | 2224 | 0.1052 | 0.0161 | 0.1624 | 0.0239 |
0.8947 | 17.0 | 2363 | 0.1045 | 0.0161 | 0.1579 | 0.0236 |
0.8442 | 18.0 | 2502 | 0.1063 | 0.0161 | 0.1572 | 0.0233 |
0.8151 | 19.0 | 2641 | 0.1036 | 0.0161 | 0.1566 | 0.0230 |
0.809 | 20.0 | 2780 | 0.1046 | 0.0161 | 0.1583 | 0.0234 |
0.8049 | 21.0 | 2919 | 0.1055 | 0.0161 | 0.1596 | 0.0229 |
0.7888 | 22.0 | 3058 | 0.1037 | 0.0161 | 0.1521 | 0.0221 |
0.7426 | 23.0 | 3197 | 0.1052 | 0.0161 | 0.1521 | 0.0223 |
0.7328 | 24.0 | 3336 | 0.1033 | 0.0161 | 0.1477 | 0.0221 |
0.7215 | 25.0 | 3475 | 0.1046 | 0.0161 | 0.1521 | 0.0224 |
0.6755 | 26.0 | 3614 | 0.1053 | 0.0161 | 0.1515 | 0.0225 |
0.6844 | 27.0 | 3753 | 0.1016 | 0.0161 | 0.1464 | 0.0217 |
0.6419 | 28.0 | 3892 | 0.1042 | 0.0161 | 0.1453 | 0.0214 |
0.6459 | 29.0 | 4031 | 0.1029 | 0.0161 | 0.1436 | 0.0209 |
0.6528 | 30.0 | 4170 | 0.1020 | 0.0161 | 0.1423 | 0.0214 |
0.6177 | 31.0 | 4309 | 0.1037 | 0.0161 | 0.1430 | 0.0213 |
0.5899 | 32.0 | 4448 | 0.1042 | 0.0161 | 0.1487 | 0.0225 |
0.6216 | 33.0 | 4587 | 0.1064 | 0.0161 | 0.1416 | 0.0212 |
0.5864 | 34.0 | 4726 | 0.1071 | 0.0161 | 0.1443 | 0.0217 |
0.5797 | 35.0 | 4865 | 0.1032 | 0.0161 | 0.1467 | 0.0215 |
0.5562 | 36.0 | 5004 | 0.1059 | 0.0161 | 0.1433 | 0.0213 |
0.5575 | 37.0 | 5143 | 0.1039 | 0.0161 | 0.1440 | 0.0214 |
0.5236 | 38.0 | 5282 | 0.1036 | 0.0161 | 0.1382 | 0.0207 |
0.5128 | 39.0 | 5421 | 0.1021 | 0.0161 | 0.1385 | 0.0203 |
0.4987 | 40.0 | 5560 | 0.1032 | 0.0161 | 0.1430 | 0.0208 |
0.5254 | 41.0 | 5699 | 0.1056 | 0.0161 | 0.1355 | 0.0204 |
0.5086 | 42.0 | 5838 | 0.1038 | 0.0161 | 0.1406 | 0.0206 |
0.4772 | 43.0 | 5977 | 0.1075 | 0.0161 | 0.1402 | 0.0205 |
0.4541 | 44.0 | 6116 | 0.1050 | 0.0161 | 0.1355 | 0.0200 |
0.4791 | 45.0 | 6255 | 0.1066 | 0.0161 | 0.1324 | 0.0198 |
0.4764 | 46.0 | 6394 | 0.1038 | 0.0161 | 0.1358 | 0.0198 |
0.4556 | 47.0 | 6533 | 0.1043 | 0.0161 | 0.1334 | 0.0197 |
0.4485 | 48.0 | 6672 | 0.1041 | 0.0161 | 0.1321 | 0.0195 |
0.44 | 49.0 | 6811 | 0.1071 | 0.0161 | 0.1385 | 0.0204 |
0.4265 | 50.0 | 6950 | 0.1086 | 0.0161 | 0.1361 | 0.0201 |
0.4388 | 51.0 | 7089 | 0.1066 | 0.0161 | 0.1304 | 0.0195 |
0.4326 | 52.0 | 7228 | 0.1087 | 0.0161 | 0.1331 | 0.0201 |
0.3916 | 53.0 | 7367 | 0.1091 | 0.0161 | 0.1368 | 0.0203 |
0.396 | 54.0 | 7506 | 0.1125 | 0.0161 | 0.1324 | 0.0198 |
0.4223 | 55.0 | 7645 | 0.1104 | 0.0161 | 0.1310 | 0.0195 |
0.3937 | 56.0 | 7784 | 0.1093 | 0.0161 | 0.1293 | 0.0193 |
0.4038 | 57.0 | 7923 | 0.1145 | 0.0161 | 0.1317 | 0.0197 |
0.402 | 58.0 | 8062 | 0.1112 | 0.0161 | 0.1331 | 0.0199 |
0.3912 | 59.0 | 8201 | 0.1109 | 0.0161 | 0.1270 | 0.0190 |
0.3888 | 60.0 | 8340 | 0.1103 | 0.0161 | 0.1249 | 0.0188 |
0.3797 | 61.0 | 8479 | 0.1074 | 0.0161 | 0.1276 | 0.0191 |
0.3682 | 62.0 | 8618 | 0.1111 | 0.0161 | 0.1293 | 0.0194 |
0.3678 | 63.0 | 8757 | 0.1117 | 0.0161 | 0.1304 | 0.0191 |
0.3611 | 64.0 | 8896 | 0.1136 | 0.0161 | 0.1253 | 0.0191 |
0.3406 | 65.0 | 9035 | 0.1114 | 0.0161 | 0.1232 | 0.0186 |
0.328 | 66.0 | 9174 | 0.1101 | 0.0161 | 0.1246 | 0.0187 |
0.3482 | 67.0 | 9313 | 0.1103 | 0.0161 | 0.1205 | 0.0182 |
0.344 | 68.0 | 9452 | 0.1130 | 0.0161 | 0.1215 | 0.0183 |
0.3465 | 69.0 | 9591 | 0.1115 | 0.0161 | 0.1290 | 0.0191 |
0.3411 | 70.0 | 9730 | 0.1124 | 0.0161 | 0.1249 | 0.0186 |
0.3182 | 71.0 | 9869 | 0.1142 | 0.0161 | 0.1225 | 0.0182 |
0.3303 | 72.0 | 10008 | 0.1116 | 0.0161 | 0.1232 | 0.0183 |
0.3632 | 73.0 | 10147 | 0.1099 | 0.0161 | 0.1229 | 0.0184 |
0.3183 | 74.0 | 10286 | 0.1128 | 0.0161 | 0.1266 | 0.0190 |
0.3215 | 75.0 | 10425 | 0.1119 | 0.0161 | 0.1229 | 0.0184 |
0.3019 | 76.0 | 10564 | 0.1132 | 0.0161 | 0.1253 | 0.0184 |
0.3275 | 77.0 | 10703 | 0.1103 | 0.0161 | 0.1212 | 0.0182 |
Framework versions
- Transformers 4.47.1
- Pytorch 2.1.0+cu118
- Datasets 3.2.0
- Tokenizers 0.21.0
- Downloads last month
- 0
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 asr-africa/mms-1B_all_NCHLT_ZULU_5hr_v1
Base model
facebook/mms-1b-all