Document created on September 7 2023

Endo_Llama: Thyroid Edition (v1)

Model Description

What is it?

Endo_Llama is a large language model (LLama 2) finetuned to answer questions related to endocrinology, with a focus on thyroid-related issues in its first version. This model aims to assist healthcare professionals and researchers in obtaining better information about thyroid conditions, treatments, and best practices.

Input Models input text only.

Output Models generate text only.

Model Architecture

Model Type: Llama 2 Parameters: ~7B Training Data: Curated dataset of endocrinology questions and answers, with a focus on thyroid issues for Version 1.

Intended Use

Clinical decision support Medical research

Limitations

Not a substitute for professional medical advice May not cover all nuances of a complex medical condition Version 1 is primarily focused on thyroid-related issues

Training Data

The model is trained on a curated dataset of endocrinology questions and answers, with an emphasis on thyroid conditions.

Usage

coming soon!

library_name: peft

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: float16

Framework versions

• PEFT 0.5.0

• PEFT 0.5.0