metadata
license: mit
base_model: gpt2
tags:
- generated_from_trainer
model-index:
- name: xsum-gpt2-long
results: []
xsum-gpt2-long
This model is a fine-tuned version of gpt2 on an unknown dataset. It achieves the following results on the evaluation set:
- Loss: 3.0751
- Ppl: 22.3764
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: 0.0005
- train_batch_size: 2
- eval_batch_size: 2
- seed: 22554
- distributed_type: multi-GPU
- num_devices: 4
- gradient_accumulation_steps: 4
- total_train_batch_size: 32
- total_eval_batch_size: 8
- optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
- lr_scheduler_type: cosine
- lr_scheduler_warmup_steps: 2000
- num_epochs: 15
- mixed_precision_training: Native AMP
Training results
| Training Loss | Epoch | Step | Validation Loss | Ppl |
|---|---|---|---|---|
| 3.662 | 2.5 | 4000 | 3.5005 | 34.2753 |
| 3.3365 | 4.99 | 8000 | 3.2677 | 27.1321 |
| 3.2005 | 7.49 | 12000 | 3.1715 | 24.6352 |
| 3.1102 | 9.98 | 16000 | 3.1144 | 23.2719 |
| 3.0517 | 12.48 | 20000 | 3.0830 | 22.5529 |
| 3.0267 | 14.97 | 24000 | 3.0751 | 22.3764 |
Framework versions
- Transformers 4.35.0.dev0
- Pytorch 1.13.1+cu117
- Datasets 2.14.6
- Tokenizers 0.14.1