stulcrad's picture
Update README.md
a54ffc6 verified
---
license: mit
base_model: FacebookAI/xlm-roberta-large
tags:
- generated_from_trainer
datasets:
- wikiann
metrics:
- precision
- recall
- f1
- accuracy
model-index:
- name: fine_tuned_XLMROBERTA_cs_wikann
results:
- task:
name: Token Classification
type: token-classification
dataset:
name: wikiann
type: wikiann
config: default
split: validation
args: default
metrics:
- name: Precision
type: precision
value: 0.920336
- name: Recall
type: recall
value: 0.934218
- name: F1
type: f1
value: 0.927225
- name: Accuracy
type: accuracy
value: 0.973202
---
<!-- 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. -->
# fine_tuned_XLMROBERTA_cs_wikann
This model is a fine-tuned version of [FacebookAI/xlm-roberta-large](https://huggingface.co/FacebookAI/xlm-roberta-large) on a czech wikiann dataset.
It achieves the following results on the evaluation set:
- Loss: 0.1543
- Precision: 0.9203
- Recall: 0.9342
- F1: 0.9272
- Accuracy: 0.9732
## 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: 5
### Training results
| Training Loss | Epoch | Step | Validation Loss | Precision | Recall | F1 | Accuracy |
|:-------------:|:-----:|:----:|:---------------:|:---------:|:------:|:------:|:--------:|
| 0.51 | 0.27 | 500 | 0.1995 | 0.7873 | 0.8274 | 0.8069 | 0.9435 |
| 0.2164 | 0.53 | 1000 | 0.2216 | 0.7743 | 0.8430 | 0.8072 | 0.9407 |
| 0.1963 | 0.8 | 1500 | 0.1673 | 0.8465 | 0.8849 | 0.8653 | 0.9534 |
| 0.1478 | 1.07 | 2000 | 0.1612 | 0.8850 | 0.9 | 0.8925 | 0.9629 |
| 0.1316 | 1.33 | 2500 | 0.1508 | 0.8765 | 0.9081 | 0.8920 | 0.9615 |
| 0.1156 | 1.6 | 3000 | 0.1561 | 0.9028 | 0.9081 | 0.9054 | 0.9656 |
| 0.1069 | 1.87 | 3500 | 0.1544 | 0.9009 | 0.9091 | 0.9050 | 0.9651 |
| 0.0925 | 2.13 | 4000 | 0.1724 | 0.9008 | 0.9216 | 0.9111 | 0.9662 |
| 0.0791 | 2.4 | 4500 | 0.1385 | 0.9096 | 0.9201 | 0.9148 | 0.9705 |
| 0.0739 | 2.67 | 5000 | 0.1309 | 0.9130 | 0.9254 | 0.9192 | 0.9701 |
| 0.0732 | 2.93 | 5500 | 0.1593 | 0.9035 | 0.9190 | 0.9112 | 0.9679 |
| 0.0538 | 3.2 | 6000 | 0.1550 | 0.9193 | 0.9309 | 0.9251 | 0.9722 |
| 0.0529 | 3.47 | 6500 | 0.1451 | 0.9112 | 0.9330 | 0.9220 | 0.9710 |
| 0.0521 | 3.73 | 7000 | 0.1510 | 0.9185 | 0.9323 | 0.9253 | 0.9721 |
| 0.0526 | 4.0 | 7500 | 0.1378 | 0.9173 | 0.9325 | 0.9249 | 0.9727 |
| 0.0377 | 4.27 | 8000 | 0.1501 | 0.9164 | 0.9344 | 0.9253 | 0.9728 |
| 0.0382 | 4.53 | 8500 | 0.1541 | 0.9213 | 0.9352 | 0.9282 | 0.9729 |
| 0.0358 | 4.8 | 9000 | 0.1543 | 0.9203 | 0.9342 | 0.9272 | 0.9732 |
### Framework versions
- Transformers 4.36.2
- Pytorch 2.1.2+cu121
- Datasets 2.16.1
- Tokenizers 0.15.0