Pharmaceutical domain instruction-tuning dataset in Alpaca format
(instruction, input, output fields).
Used to fine-tune TinyLlama/TinyLlama-1.1B-intermediate-step-1431k-3T
as Stage 2 (instruction tuning) on top of the domain-adapted Stage 1 model.
Format
Each JSONL record contains:
{
"instruction": "Explain the primary mechanism of action of metformin.",
"input": "",
"output": "Metformin works primarily by ..."
}
