MichaelAI23/falcon-rw-1b_8bit_finetuned

Training procedure

This is a Falcon 1B model that was finetuned on the MichaelAI23/hotel_requests dataset. LoRA was used for the training in combination with 8-bit quantization.

The following bitsandbytes quantization config was used during training:

• quant_method: bitsandbytes
• 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 0.6.0.dev0