Downloads · 30 days
6
1% of all-time downloads
IceBERT-finetuned-grouped
ueb1/IceBERT-finetuned-grouped
IceBERT-finetuned-grouped is a text classification model from ueb1. Use it when you need a label for a piece of text. It is set up for transformers. The card lists the license as gpl-3.0.
should probably proofread and complete it, then remove this comment. --
All-time downloads
657
Public
Repo size
1.5 GB
Likes
0
Public
Repo makeup
Click a slice to open those files.
.bin498 MB · 99%
At a glance
- Task
- Text Classification
- Library
- transformers
- License
- gpl-3.0
- Model type
- roberta
- Access
- Public
- Created
- Mar 2, 2022
- Updated
- Nov 24, 2021
- SHA
- 02e88a7e
Try a prompt
Model card
From the Hugging Face model README
<!-- 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. -->
IceBERT-finetuned-grouped
This model is a fine-tuned version of vesteinn/IceBERT on an unknown dataset. It achieves the following results on the evaluation set:
- Loss: 3.5660
- Accuracy: 0.2259
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 | Accuracy |
|---|---|---|---|---|
| No log | 1.0 | 269 | 4.1727 | 0.1172 |
| 4.3535 | 2.0 | 538 | 3.8406 | 0.1632 |
| 4.3535 | 3.0 | 807 | 3.6718 | 0.2113 |
| 3.6711 | 4.0 | 1076 | 3.5660 | 0.2259 |
| 3.6711 | 5.0 | 1345 | 3.5332 | 0.2176 |
Framework versions
- Transformers 4.12.5
- Pytorch 1.10.0+cu111
- Datasets 1.15.1
- Tokenizers 0.10.3
- Task
- Text Classification
- Library
- transformers
- Type
- roberta
- License
- gpl-3.0
- Created
- Mar 2, 2022
- Updated
- Nov 24, 2021