NER-Advance
This model is a fine-tuned version of xlm-roberta-base on an XTREME dataset. It achieves the following results on the evaluation set:
- Loss: 0.1363
- F1: 0.8658
Training and evaluation data
I used subset of the Cross-lingual TRansfer Evaluation of Multilingual Encoders (XTREME) benchmark called WikiANN or PAN-X.2 This dataset consists of Wikipedia articles in many languages, including the four most commonly spoken languages in Switzerland: German (62.9%), French (22.9%), Ital‐ ian (8.4%), and English (5.9%). Each article is annotated with LOC (location), PER (person), and ORG (organization) tags in the “inside-outside-beginning” (IOB2) for‐ mat. In this format, a B- prefix indicates the beginning of an entity, and consecutive tokens belonging to the same entity are given an I- prefix.
Training hyperparameters
The following hyperparameters were used during training:
- learning_rate: 5e-05
- train_batch_size: 24
- eval_batch_size: 24
- seed: 42
- optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
- lr_scheduler_type: linear
- num_epochs: 3
Training results
Training Loss | Epoch | Step | Validation Loss | F1 |
---|---|---|---|---|
0.2539 | 1.0 | 525 | 0.1505 | 0.8246 |
0.1268 | 2.0 | 1050 | 0.1380 | 0.8503 |
0.0794 | 3.0 | 1575 | 0.1363 | 0.8658 |
Framework versions
- Transformers 4.38.2
- Pytorch 2.2.1+cu121
- Datasets 2.18.0
- Tokenizers 0.15.2
- Downloads last month
- 24