Edit model card

Meta-Llama-3-8B-Instruct_derta_100step

This model is a fine-tuned version of meta-llama/Meta-Llama-3-8B-Instruct on the Evol-Instruct and BeaverTails dataset. The model is continued to train 100 steps with DeRTa on LLaMA3-8B-Instruct.

Model description

Please refer to the paper Refuse Whenever You Feel Unsafe: Improving Safety in LLMs via Decoupled Refusal Training and GitHub DeRTa.

Input format:

[INST] Your Instruction [\INST]

Intended uses & limitations

The model is trained with DeRTa, showing a high safety performance.

Training and evaluation data

More information needed

Training procedure

Training hyperparameters

The following hyperparameters were used during training:

  • learning_rate: 2e-05
  • train_batch_size: 16
  • weight_decay: 2e-5
  • eval_batch_size: 1
  • seed: 1
  • distributed_type: multi-GPU
  • num_devices: 8
  • total_train_batch_size: 128
  • total_eval_batch_size: 8
  • optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
  • lr_scheduler_type: cosine
  • lr_scheduler_warmup_ratio: 0.03
  • num_epochs: 2.0

Training results

Framework versions

  • Transformers 4.40.0
  • Pytorch 2.2.0+cu118
  • Datasets 2.10.0
  • Tokenizers 0.19.1
Downloads last month
7
Safetensors
Model size
8.03B params
Tensor type
BF16
·
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 Youliang/llama3-8b-instruct-derta-100step

Finetuned
(441)
this model