tinyllama_sft_full

This model is a fine-tuned version of TinyLlama/TinyLlama-1.1B-step-50K-105b on the jailbreak_attack_sft_data_12197 dataset. It achieves the following results on the evaluation set:

  • Loss: 0.0074

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: 0.0001
  • train_batch_size: 14
  • eval_batch_size: 10
  • seed: 42
  • distributed_type: multi-GPU
  • num_devices: 4
  • total_train_batch_size: 56
  • total_eval_batch_size: 40
  • optimizer: Use adamw_torch with betas=(0.9,0.999) and epsilon=1e-08 and optimizer_args=No additional optimizer arguments
  • lr_scheduler_type: cosine
  • num_epochs: 8.0
  • mixed_precision_training: Native AMP

Training results

Training Loss Epoch Step Validation Loss
0.0109 1.8692 400 0.0107
0.0068 3.7383 800 0.0079
0.0059 5.6075 1200 0.0075
0.0053 7.4766 1600 0.0074

Framework versions

  • Transformers 4.47.0
  • Pytorch 2.3.1+cu121
  • Datasets 2.20.0
  • Tokenizers 0.21.0
Downloads last month
73
Inference Examples
This model does not have enough activity to be deployed to Inference API (serverless) yet. Increase its social visibility and check back later, or deploy to Inference Endpoints (dedicated) instead.

Model tree for xdrshjr/TinyLlama-1b-Rewrite-Jailbreak-Prompt

Finetuned
(4)
this model