Edit model card

sft-trl

This model is a fine-tuned version of EleutherAI/gpt-j-6b on the anli dataset. It achieves the following results on the evaluation set:

  • Loss: 0.6837

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: 5e-05
  • train_batch_size: 8
  • eval_batch_size: 8
  • seed: 42
  • optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
  • lr_scheduler_type: linear
  • lr_scheduler_warmup_steps: 100
  • num_epochs: 3.0

Training results

Training Loss Epoch Step Validation Loss
2.1859 0.2 1000 2.0019
1.1342 0.4 2000 1.5352
0.8871 0.6 3000 1.0923
0.591 0.8 4000 0.8373
0.3165 1.0 5000 0.7852
0.3205 1.2 6000 0.7531
0.2338 1.4 7000 0.7155
0.2922 1.6 8000 0.6837
0.2427 1.8 9000 0.6837
0.2036 2.0 10000 0.6837
0.1205 2.2 11000 0.6837
0.2275 2.4 12000 0.6837
0.1094 2.6 13000 0.6837
0.1507 2.8 14000 0.6837
0.1739 3.0 15000 0.6837

Framework versions

  • Transformers 4.30.2
  • Pytorch 2.0.1+cu117
  • Datasets 2.13.1
  • Tokenizers 0.13.3
Downloads last month
1

Dataset used to train Tverous/sft-trl