Sakuna/LLaMaCoder

1---2language:3- en4library_name: peft5tags:6- llama27- peft8- bitsandbytes9datasets:10- HuggingFaceH4/CodeAlpaca_20K11pipeline_tag: text2text-generation12base_model: meta-llama/Llama-2-7b-hf13---14## Training procedure15 16 17The following `bitsandbytes` quantization config was used during training:18- load_in_8bit: False19- load_in_4bit: True20- llm_int8_threshold: 6.021- llm_int8_skip_modules: None22- llm_int8_enable_fp32_cpu_offload: False23- llm_int8_has_fp16_weight: False24- bnb_4bit_quant_type: nf425- bnb_4bit_use_double_quant: False26- bnb_4bit_compute_dtype: float1627### Framework versions28 29 30- PEFT 0.4.0.dev0