qqplot23's picture
Model save
b2ecdb7
metadata
license: mit
base_model: gpt2
tags:
  - generated_from_trainer
model-index:
  - name: xsum-gpt2-long-pegasus
    results: []

xsum-gpt2-long-pegasus

This model is a fine-tuned version of gpt2 on an unknown dataset. It achieves the following results on the evaluation set:

  • Loss: 3.2524
  • Ppl: 26.6834

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: 2
  • gradient_accumulation_steps: 8
  • total_train_batch_size: 32
  • total_eval_batch_size: 4
  • 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
3.7921 2.67 4000 3.6382 39.1940
3.4486 5.34 8000 3.4164 31.3953
3.299 8.01 12000 3.3291 28.7823
3.2019 10.68 16000 3.2769 27.3369
3.1403 13.36 20000 3.2524 26.6834

Framework versions

  • Transformers 4.35.0.dev0
  • Pytorch 1.13.1+cu117
  • Datasets 2.14.6
  • Tokenizers 0.14.1