đ wav2vec2-base-intent-classification-ori-f1
This model is a fine - tuned version of facebook/wav2vec2-base on the None dataset. It can achieve good performance in intent classification tasks, with an F1 score of 0.875 on the evaluation set.
đ Quick Start
This model is a fine - tuned version of facebook/wav2vec2-base on the None dataset.
It achieves the following results on the evaluation set:
đ Documentation
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: 3e - 05
- train_batch_size: 1
- eval_batch_size: 1
- seed: 42
- gradient_accumulation_steps: 4
- total_train_batch_size: 4
- optimizer: Adam with betas=(0.9,0.999) and epsilon = 1e - 08
- lr_scheduler_type: linear
- lr_scheduler_warmup_ratio: 0.1
- num_epochs: 45
Training results
Training Loss |
Epoch |
Step |
Validation Loss |
F1 |
2.19 |
1.0 |
28 |
2.1733 |
0.2708 |
2.1205 |
2.0 |
56 |
2.1125 |
0.2708 |
2.0965 |
3.0 |
84 |
2.0543 |
0.2708 |
1.9694 |
4.0 |
112 |
1.9125 |
0.2917 |
1.9091 |
5.0 |
140 |
1.8455 |
0.3542 |
1.8399 |
6.0 |
168 |
1.7895 |
0.3958 |
1.8424 |
7.0 |
196 |
1.8828 |
0.3125 |
1.5475 |
8.0 |
224 |
1.4255 |
0.5208 |
1.2653 |
9.0 |
252 |
1.3953 |
0.5417 |
1.1465 |
10.0 |
280 |
1.3501 |
0.5417 |
1.281 |
11.0 |
308 |
1.2800 |
0.5417 |
1.0996 |
12.0 |
336 |
1.2797 |
0.6042 |
1.1288 |
13.0 |
364 |
1.1341 |
0.6667 |
0.8577 |
14.0 |
392 |
1.0104 |
0.7083 |
0.8047 |
15.0 |
420 |
1.0906 |
0.6667 |
0.7098 |
16.0 |
448 |
0.9710 |
0.7917 |
0.5407 |
17.0 |
476 |
0.9363 |
0.7708 |
0.4634 |
18.0 |
504 |
0.8283 |
0.75 |
0.4368 |
19.0 |
532 |
0.7587 |
0.7708 |
0.2818 |
20.0 |
560 |
0.6551 |
0.8333 |
0.1951 |
21.0 |
588 |
0.5865 |
0.8333 |
0.1456 |
22.0 |
616 |
0.7378 |
0.7917 |
0.1269 |
23.0 |
644 |
0.6327 |
0.8333 |
0.0801 |
24.0 |
672 |
0.6896 |
0.8333 |
0.0723 |
25.0 |
700 |
0.7179 |
0.8333 |
0.0626 |
26.0 |
728 |
1.0643 |
0.7708 |
0.0434 |
27.0 |
756 |
0.4353 |
0.875 |
0.0499 |
28.0 |
784 |
0.6656 |
0.8333 |
0.0396 |
29.0 |
812 |
0.6788 |
0.8333 |
0.0352 |
30.0 |
840 |
0.8139 |
0.8333 |
0.0348 |
31.0 |
868 |
0.8745 |
0.8125 |
0.0313 |
32.0 |
896 |
0.8693 |
0.8125 |
0.0269 |
33.0 |
924 |
0.9393 |
0.8125 |
0.0242 |
34.0 |
952 |
0.9351 |
0.8333 |
0.0217 |
35.0 |
980 |
0.9406 |
0.8333 |
0.0234 |
36.0 |
1008 |
0.9464 |
0.8333 |
0.0219 |
37.0 |
1036 |
0.9507 |
0.8333 |
0.0215 |
38.0 |
1064 |
0.9471 |
0.8333 |
0.0206 |
39.0 |
1092 |
0.9260 |
0.8333 |
0.0229 |
40.0 |
1120 |
0.9420 |
0.8333 |
0.0216 |
41.0 |
1148 |
0.9570 |
0.8333 |
0.0227 |
42.0 |
1176 |
0.9573 |
0.8333 |
0.0208 |
43.0 |
1204 |
0.9609 |
0.8333 |
0.0201 |
44.0 |
1232 |
0.9617 |
0.8333 |
0.0208 |
45.0 |
1260 |
0.9620 |
0.8333 |
Framework versions
- Transformers 4.20.1
- Pytorch 1.11.0
- Datasets 2.1.0
- Tokenizers 0.12.1
đ License
This project is licensed under the Apache - 2.0 license.