This model is fine-tuned to generate SAS code based on healthcare data specifications, specifically using the 2015 MEPS dataset. It was trained with specifications and corresponding SAS code pairs.
Model Details
Base Model: Llama-3.2-1B-bnb-4bit
Fine-Tuning Dataset: MEPS dataset specifications and SAS code pairs.
Objective: Generate executable SAS code based on input specifications.
Training Framework: Hugging Face transformers with LoRA (PEFT).
Hardware: Google Colab Pro (A100 GPU).
Example Usage
Input:
As a healthcare data analyst, generate SAS code to identify individuals with mental disorders (CCS Code = 650-670) using the 2015 MEPS data.
Output:
sas
1PROC MEPSDATA;/* Example SAS code */2RUN;
Intended Use
This model is intended for healthcare analysts working with SAS and MEPS data. It is designed to streamline SAS code generation based on structured specifications.
Limitations
May not generalize well to non-healthcare datasets or specifications.
Requires familiarity with SAS to validate generated code.
