Fine-tuned Falcon-7B Model for Medical Diagnosis

Model Details

Model Description:

This model is a fine-tuned version of the Falcon-7B model. This model was fine-tuned on Gretel.ai's "Symtoms to Diagnosis" dataset, found at the following link: https://huggingface.co/datasets/gretelai/symptom_to_diagnosis, in order to provide preliminary diagnoses based on the symptom descriptions it is prompted with.

Baseline Model:

For more details about the baseline Falcon-7B model, please see the following links:

1. https://huggingface.co/tiiuae/falcon-7b
2. https://huggingface.co/blog/falcon

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: bfloat16

Framework versions

• PEFT 0.5.0.dev0