fine tuning a llm model, performance tests yet to perform.

Model: tiiuae/falcon-7b, Dataset: nampdn-ai/tiny-codes

before training output: Instruction:"Generate a python function to find number of CPU cores" Response: "def num_cpu_cores(): num_cores = (int)(os.cpu_count() * 2) return"

After training output: Instruction:"Generate a python function to find number of CPU cores" Response: "def num_cpu_cores(): num_cores = 0 for i in range(0, os.cpu_count()):""

Training procedure

The following bitsandbytes quantization config was used during training:

• 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: bfloat16

Framework versions

• PEFT 0.5.0.dev0