|
--- |
|
datasets: |
|
- seamew/ChnSentiCorp |
|
metrics: |
|
- accuracy |
|
- precision |
|
- f1 |
|
- recall |
|
model-index: |
|
- name: gpt2-imdb-sentiment-classifier |
|
results: |
|
- task: |
|
name: Text Classification |
|
type: text-classification |
|
dataset: |
|
name: imdb |
|
type: imdb |
|
args: plain_text |
|
metrics: |
|
- name: Accuracy |
|
type: accuracy |
|
value: 0.9394 |
|
language: |
|
- zh |
|
pipeline_tag: text-classification |
|
--- |
|
|
|
<!-- 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. --> |
|
|
|
# gpt2-imdb-sentiment-classifier |
|
|
|
This model is a fine-tuned version of [hfl/rbt6](https://huggingface.co/hfl/rbt6) on the ChnSentiCorp dataset. |
|
It achieves the following results on the evaluation set: |
|
- Loss: 0.294600 |
|
- Accuracy: 0.933884 |
|
|
|
## Intended uses & limitations |
|
|
|
This is comparable to [distilbert-imdb](https://huggingface.co/lvwerra/distilbert-imdb) and trained with exactly the same [script](https://huggingface.co/lvwerra/distilbert-imdb/blob/main/distilbert-imdb-training.ipynb) |
|
|
|
It achieves slightly lower loss (0.1703 vs 0.1903) and slightly higher accuracy (0.9394 vs 0.928) |
|
|
|
## 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: 4 |
|
- eval_batch_size: 4 |
|
- weight_decay=1e-2 |
|
- num_train_epochs=3 |
|
### Training results |
|
|
|
Epoch Training Loss Validation Loss Accuracy F1 Precision Recall |
|
1 0.359700 0.306089 0.924242 0.926230 0.918699 0.933884 |
|
2 0.200600 0.295512 0.942761 0.943615 0.946755 0.940496 |
|
3 0.105600 0.294600 0.941919 0.942452 0.951178 0.933884 |
|
|
|
### Framework versions |
|
|
|
- Pytorch 2.0.0 |
|
- Python 3.9.12 |
