Downloads · 30 days
28
6% of all-time downloads
tetis-textmine-2024-camembert-large-based
rdecoupes/tetis-textmine-2024-camembert-large-based
tetis-textmine-2024-camembert-large-based is a token classification model from rdecoupes. Use it when you need labels on individual words, such as names. It is set up for transformers.
All-time downloads
502
Public
Parameters
336M
2.7 GB on disk
Likes
0
Public
Repo makeup
Click a slice to open those files.
.bin1.3 GB · 50%
Parameter types
How the weights are stored.
F32336M · 100%
Try a prompt
Model card
From the Hugging Face model README
license: cc-by-nc-4.0
TETIS @ Challenge TextMine 2024
This model is a NER based on Camembert-Large for the Kaggle Competition (in French): https://www.kaggle.com/competitions/defi-textmine-2024/
This model could be re-use with HuggingFace transormers pipeline. To use it, please refer to its Github
<img align="left" src="https://www.umr-tetis.fr/images/logo-header-tetis.png">| Participants |
|---|
| Rémy Decoupes |
| Roberto Interdonato |
| Rodrique Kafando |
| Mehtab Syed Alam |
| Maguelonne Teisseire |
| Mathieu Roche |
| Sarah Valentin |
<!-- 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. -->
tetis-textmine-2024-camembert-large-based
This model is a fine-tuned version of camembert/camembert-large on the None dataset. It achieves the following results on the evaluation set:
- Loss: 0.1106
- Precision: 0.9327
- Recall: 0.9471
- F1: 0.9398
- Accuracy: 0.9843
- Aucun F1: 0.9434
- Geogfeat F1: 0.9193
- Geogfeat geogname F1: 0.9554
- Geogname F1: 0.9133
- Name geogname F1: 0.9519
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: 8
- eval_batch_size: 8
- seed: 42
- optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
- lr_scheduler_type: linear
- num_epochs: 10
Training results
| Training Loss | Epoch | Step | Validation Loss | Precision | Recall | F1 | Accuracy | Aucun F1 | Geogfeat F1 | Geogfeat geogname F1 | Geogname F1 | Name geogname F1 |
|---|---|---|---|---|---|---|---|---|---|---|---|---|
| No log | 1.0 | 192 | 0.1045 | 0.9171 | 0.9369 | 0.9269 | 0.9828 | 0.9303 | 0.8943 | 0.9509 | 0.9174 | 0.9373 |
| No log | 2.0 | 384 | 0.1029 | 0.9223 | 0.9471 | 0.9345 | 0.9830 | 0.9339 | 0.9170 | 0.9522 | 0.9419 | 0.9377 |
| 0.0072 | 3.0 | 576 | 0.0952 | 0.9136 | 0.9466 | 0.9298 | 0.9840 | 0.9226 | 0.8993 | 0.9587 | 0.9440 | 0.9571 |
| 0.0072 | 4.0 | 768 | 0.1054 | 0.9347 | 0.9409 | 0.9378 | 0.9838 | 0.9380 | 0.9256 | 0.9603 | 0.9164 | 0.9433 |
| 0.0072 | 5.0 | 960 | 0.1165 | 0.9229 | 0.9347 | 0.9288 | 0.9814 | 0.9328 | 0.9013 | 0.9441 | 0.9060 | 0.9451 |
| 0.0071 | 6.0 | 1152 | 0.1070 | 0.9306 | 0.9462 | 0.9383 | 0.9840 | 0.9419 | 0.9144 | 0.9487 | 0.9213 | 0.9533 |
| 0.0071 | 7.0 | 1344 | 0.1037 | 0.9285 | 0.9453 | 0.9368 | 0.9844 | 0.9392 | 0.9100 | 0.9534 | 0.9271 | 0.9507 |
| 0.0013 | 8.0 | 1536 | 0.1127 | 0.9335 | 0.9475 | 0.9405 | 0.9841 | 0.9451 | 0.9175 | 0.9505 | 0.9222 | 0.9520 |
| 0.0013 | 9.0 | 1728 | 0.1110 | 0.9356 | 0.9488 | 0.9422 | 0.9849 | 0.9452 | 0.9195 | 0.9571 | 0.9186 | 0.9572 |
| 0.0013 | 10.0 | 1920 | 0.1106 | 0.9327 | 0.9471 | 0.9398 | 0.9843 | 0.9434 | 0.9193 | 0.9554 | 0.9133 | 0.9519 |
Framework versions
- Transformers 4.30.2
- Pytorch 2.0.1+cu117
- Datasets 2.13.0
- Tokenizers 0.13.3
- Task
- Token Classification
- Library
- transformers
- Type
- camembert
- Created
- Nov 29, 2023
- Updated
- Feb 13, 2025