Qwen2.5-Coder-1.5B text-to-SQL (SQLite) LoRA
QLoRA adapter (r=16, alpha=32, all linear layers) trained for 1 epoch on parthasaradhi789/text2sql-sqlite-clean.
Quick check on 300 validation questions from unseen domains (execution accuracy, notebook
batched greedy decoding): base 57.0% -> fine-tuned 69.3%.
Final numbers come from the evaluation harness in the code repository (https://github.com/wittyparth/text2sql-specialist).
Git commit used for training: 1d71449. Training run: https://wandb.ai/parthasaradhimunakala-parthasaradhi-site/text2sql-specialist/runs/379av0n3
- Downloads last month
- 39
Inference Providers NEW
This model isn't deployed by any Inference Provider. 🙋 Ask for provider support
Model tree for parthasaradhi789/text2sql-qwen2.5-coder-1.5b-lora
Base model
Qwen/Qwen2.5-1.5B Finetuned
Qwen/Qwen2.5-Coder-1.5B Finetuned
Qwen/Qwen2.5-Coder-1.5B-Instruct