metadata
license: apache-2.0
tags:
- generated_from_trainer
model-index:
- name: distilgpt2-finetune-acl22
results: []
distilgpt2-finetune-acl22
This model is a fine-tuned version of distilgpt2 on the ACL-anthology-corpus dataset. It achieves the following results on the evaluation set:
- Loss: 3.4835
Model description
We finetune the gpt2 LLM on the full-text from ACL-anthology-corpus
Training hyperparameters
The following hyperparameters were used during training:
- learning_rate: 2e-05
- train_batch_size: 256
- eval_batch_size: 64
- 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 |
|---|---|---|---|
| 3.6676 | 1.0 | 9852 | 3.5623 |
| 3.5959 | 2.0 | 19704 | 3.4995 |
| 3.5719 | 3.0 | 29556 | 3.4835 |
Framework versions
- Transformers 4.21.2
- Pytorch 1.12.1
- Datasets 2.4.0
- Tokenizers 0.12.1