metadata
license: apache-2.0
base_model: Qwen/Qwen2.5-Coder-1.5B-Instruct
tags:
- optimization
- code-generation
- retail
- pyomo
- fine-tuned
RetailOpt Codegen Coder
Fine-tuned Qwen2.5-Coder for natural-language retail supply-chain scenarios → executable Pyomo code.
Training
| Property | Value |
|---|---|
| Base model | Qwen/Qwen2.5-Coder-1.5B-Instruct |
| Dataset | retailopt-10k-verified |
| Method | SFT (TRL) |
| Task | NL + JSON → Pyomo + HiGHS |
| Reference | SIRL-Gurobi |
Training Command
hf jobs uv run scripts/run_finetune_job.py \
--flavor a10g-large --timeout 3h --secrets HF_TOKEN=$HF_TOKEN --detach
Intended Use
Generate Pyomo optimization models for retail inventory, allocation, procurement, distribution, fulfillment, and pricing scenarios from structured prompts.
Limitations
- Trained on simplified Pyomo templates; complex shelf-life aging may require repair loop
- HiGHS solver required at inference time
- Not a substitute for human OR validation on production planning systems