|
--- |
|
tags: |
|
- generated_from_trainer |
|
model-index: |
|
- name: model_broadclass_onSet3 |
|
results: [] |
|
--- |
|
|
|
<!-- 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. --> |
|
|
|
# model_broadclass_onSet3 |
|
|
|
This model was trained from scratch on the None dataset. |
|
It achieves the following results on the evaluation set: |
|
- eval_loss: 0.1389 |
|
- eval_0_precision: 1.0 |
|
- eval_0_recall: 1.0 |
|
- eval_0_f1-score: 1.0 |
|
- eval_0_support: 23 |
|
- eval_1_precision: 0.9697 |
|
- eval_1_recall: 0.9697 |
|
- eval_1_f1-score: 0.9697 |
|
- eval_1_support: 33 |
|
- eval_2_precision: 1.0 |
|
- eval_2_recall: 1.0 |
|
- eval_2_f1-score: 1.0 |
|
- eval_2_support: 26 |
|
- eval_3_precision: 0.9333 |
|
- eval_3_recall: 0.9333 |
|
- eval_3_f1-score: 0.9333 |
|
- eval_3_support: 15 |
|
- eval_accuracy: 0.9794 |
|
- eval_macro avg_precision: 0.9758 |
|
- eval_macro avg_recall: 0.9758 |
|
- eval_macro avg_f1-score: 0.9758 |
|
- eval_macro avg_support: 97 |
|
- eval_weighted avg_precision: 0.9794 |
|
- eval_weighted avg_recall: 0.9794 |
|
- eval_weighted avg_f1-score: 0.9794 |
|
- eval_weighted avg_support: 97 |
|
- eval_wer: 0.1037 |
|
- eval_mtrix: [[0, 1, 2, 3], [0, 23, 0, 0, 0], [1, 0, 32, 0, 1], [2, 0, 0, 26, 0], [3, 0, 1, 0, 14]] |
|
- eval_runtime: 5.6481 |
|
- eval_samples_per_second: 17.174 |
|
- eval_steps_per_second: 2.302 |
|
- step: 0 |
|
|
|
## 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: 8 |
|
- eval_batch_size: 8 |
|
- seed: 42 |
|
- gradient_accumulation_steps: 2 |
|
- total_train_batch_size: 16 |
|
- optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08 |
|
- lr_scheduler_type: linear |
|
- lr_scheduler_warmup_steps: 200 |
|
- num_epochs: 80 |
|
- mixed_precision_training: Native AMP |
|
|
|
### Framework versions |
|
|
|
- Transformers 4.25.1 |
|
- Pytorch 1.13.0+cu116 |
|
- Datasets 2.8.0 |
|
- Tokenizers 0.13.2 |
|
|