Introduction

Vicuna-style model based on LoRA tuning

• Dataset: https://huggingface.co/datasets/Aeala/ShareGPT_Vicuna_unfiltered
• Base model: https://huggingface.co/huggyllama/llama-7b

Training procedure

The following bitsandbytes quantization config was used during training:

• load_in_8bit: True
• load_in_4bit: False
• llm_int8_threshold: 6.0
• llm_int8_skip_modules: None
• llm_int8_enable_fp32_cpu_offload: False
• llm_int8_has_fp16_weight: False
• bnb_4bit_quant_type: fp4
• bnb_4bit_use_double_quant: False
• bnb_4bit_compute_dtype: float32

Framework versions

• PEFT 0.4.0