|
--- |
|
language: |
|
- en |
|
license: mit |
|
tags: |
|
- generated_from_trainer |
|
- nlu |
|
- slot-tagging |
|
datasets: |
|
- AmazonScience/massive |
|
metrics: |
|
- precision |
|
- recall |
|
- f1 |
|
- accuracy |
|
base_model: xlm-roberta-base |
|
model-index: |
|
- name: xlm-r-base-amazon-massive-slot |
|
results: |
|
- task: |
|
type: slot-filling |
|
name: slot-filling |
|
dataset: |
|
name: MASSIVE |
|
type: AmazonScience/massive |
|
split: test |
|
metrics: |
|
- type: f1 |
|
value: 0.8405 |
|
name: F1 |
|
--- |
|
|
|
<!-- 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. --> |
|
|
|
# xlm-r-base-amazon-massive-slot |
|
|
|
This model is a fine-tuned version of [xlm-roberta-base](https://huggingface.co/xlm-roberta-base) on the [MASSIVE1.1](https://huggingface.co/datasets/AmazonScience/massive) dataset. |
|
It achieves the following results on the evaluation set: |
|
- Loss: 0.5006 |
|
- Precision: 0.8144 |
|
- Recall: 0.8683 |
|
- F1: 0.8405 |
|
- Accuracy: 0.9333 |
|
|
|
## 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: 2e-05 |
|
- train_batch_size: 16 |
|
- eval_batch_size: 16 |
|
- seed: 42 |
|
- optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08 |
|
- lr_scheduler_type: linear |
|
- num_epochs: 20 |
|
|
|
### Training results |
|
|
|
| Training Loss | Epoch | Step | Validation Loss | Precision | Recall | F1 | Accuracy | |
|
|:-------------:|:-----:|:-----:|:---------------:|:---------:|:------:|:------:|:--------:| |
|
| 1.1445 | 1.0 | 720 | 0.5446 | 0.6681 | 0.6770 | 0.6725 | 0.8842 | |
|
| 0.5908 | 2.0 | 1440 | 0.3869 | 0.7331 | 0.7706 | 0.7514 | 0.9083 | |
|
| 0.3228 | 3.0 | 2160 | 0.3285 | 0.7658 | 0.8288 | 0.7961 | 0.9219 | |
|
| 0.2561 | 4.0 | 2880 | 0.3063 | 0.7819 | 0.8402 | 0.8100 | 0.9257 | |
|
| 0.1808 | 5.0 | 3600 | 0.3000 | 0.8011 | 0.8429 | 0.8214 | 0.9305 | |
|
| 0.1487 | 6.0 | 4320 | 0.2982 | 0.8201 | 0.8492 | 0.8344 | 0.9361 | |
|
| 0.1156 | 7.0 | 5040 | 0.3252 | 0.8009 | 0.8569 | 0.8280 | 0.9313 | |
|
| 0.094 | 8.0 | 5760 | 0.3481 | 0.8127 | 0.8502 | 0.8310 | 0.9333 | |
|
| 0.0843 | 9.0 | 6480 | 0.3764 | 0.7990 | 0.8613 | 0.8290 | 0.9304 | |
|
| 0.0641 | 10.0 | 7200 | 0.3822 | 0.7930 | 0.8609 | 0.8256 | 0.9280 | |
|
| 0.0547 | 11.0 | 7920 | 0.3889 | 0.8223 | 0.8649 | 0.8431 | 0.9354 | |
|
| 0.04 | 12.0 | 8640 | 0.4416 | 0.8019 | 0.8633 | 0.8314 | 0.9288 | |
|
| 0.0368 | 13.0 | 9360 | 0.4339 | 0.8117 | 0.8606 | 0.8354 | 0.9328 | |
|
| 0.0297 | 14.0 | 10080 | 0.4698 | 0.8062 | 0.8623 | 0.8333 | 0.9314 | |
|
| 0.0227 | 15.0 | 10800 | 0.4763 | 0.8058 | 0.8656 | 0.8346 | 0.9327 | |
|
| 0.0185 | 16.0 | 11520 | 0.4793 | 0.8124 | 0.8613 | 0.8361 | 0.9326 | |
|
| 0.0182 | 17.0 | 12240 | 0.4835 | 0.8191 | 0.8629 | 0.8404 | 0.9341 | |
|
| 0.0147 | 18.0 | 12960 | 0.4981 | 0.8140 | 0.8693 | 0.8407 | 0.9336 | |
|
| 0.0111 | 19.0 | 13680 | 0.5002 | 0.8099 | 0.8719 | 0.8398 | 0.9340 | |
|
| 0.0128 | 20.0 | 14400 | 0.5006 | 0.8144 | 0.8683 | 0.8405 | 0.9333 | |
|
|
|
|
|
### Framework versions |
|
|
|
- Transformers 4.22.2 |
|
- Pytorch 1.12.1+cu113 |
|
- Datasets 2.5.1 |
|
- Tokenizers 0.12.1 |
|
|
|
## Citation |
|
```bibtex |
|
@article{kubis2023back, |
|
title={Back Transcription as a Method for Evaluating Robustness of Natural Language Understanding Models to Speech Recognition Errors}, |
|
author={Kubis, Marek and Sk{\'o}rzewski, Pawe{\l} and Sowa{\'n}ski, Marcin and Zi{\k{e}}tkiewicz, Tomasz}, |
|
journal={arXiv preprint arXiv:2310.16609}, |
|
year={2023} |
|
eprint={2310.16609}, |
|
archivePrefix={arXiv}, |
|
} |
|
``` |
