Liang-Su's picture
Upload 74 files
18187e5 verified
|
raw
history blame
508 Bytes
metadata
library_name: peft

Training procedure

The following bitsandbytes quantization config was used during training:

  • quant_method: bitsandbytes
  • _load_in_8bit: False
  • _load_in_4bit: True
  • 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: nf4
  • bnb_4bit_use_double_quant: True
  • bnb_4bit_compute_dtype: float16
  • load_in_4bit: True
  • load_in_8bit: False

Framework versions

  • PEFT 0.4.0