Skip to content
Language models

Qwen2.5-Coder-1.5B-Text2SQL-LoRA

a compact model for generating SQLite queries

2 min readupdated

This model turns plain-English questions and table schemas into ready-to-run SQLite queries. Built on the compact Qwen2.5-Coder-1.5B-Instruct language model (1.54B parameters), it was fine-tuned using QLoRA (quantized low-rank adaptation, which cuts VRAM usage during training). Because the adapter weights are already merged into the base model in fp16, you can load it just like a standard LLM with no extra setup. Compared to the original base model, this release delivers higher execution accuracy on real databases, making it a practical fit for developers who want a fast, lightweight local assistant for SQLite.

Key features

  • Higher execution accuracy: scores jumped from 61.0% to 71.3% (+10.3 percentage points) on the Gretel test set, and from 46.1% to 55.2% (+9.1 percentage points) on unseen Spider dev schemas.
  • Ready-to-use merged weights in fp16 in the repository root, alongside a standalone adapter in the adapter/ folder.
  • Trained on 12,000 validated examples from the gretelai/synthetic_text_to_sql dataset with loss calculated strictly on response tokens.
  • Open Apache-2.0 license, fully permissive for commercial use.

What you can do

  • Embed a local natural-language interface into a lightweight app or Telegram bot to query SQLite databases without sending your schemas to cloud APIs.
  • Speed up database prototyping by generating draft SELECT queries for quick manual review.
  • Grab the standalone adapter from the adapter/ folder to experiment with merging or further fine-tuning on top of Qwen2.5-Coder-1.5B-Instruct.

Example

import torch
from transformers import AutoTokenizer, AutoModelForCausalLM

repo = "2003Rohansinha/Qwen2.5-Coder-1.5B-Text2SQL-LoRA"
tok = AutoTokenizer.from_pretrained(repo)
model = AutoModelForCausalLM.from_pretrained(repo, torch_dtype=torch.float16, device_map="auto")
msgs = [{"role": "system", "content": "You are an expert SQL assistant. Given a database schema and a question, write one SQLite query that answers the question. Return only the SQL query in a ```sql code block."}, {"role": "user", "content": "CREATE TABLE employees (id INTEGER, name TEXT, dept TEXT, salary REAL); What is the average salary in each department?"}]
ids = tok.apply_chat_template(msgs, add_generation_prompt=True, return_tensors="pt").to(model.device)
print(tok.decode(model.generate(ids, max_new_tokens=256, do_sample=False)[0][ids.shape[1]:], skip_special_tokens=True))
Read moreDetails
Benchmarks (execution accuracy, greedy decoding): Gretel test (300 questions) — 71.3% vs. 61.0% for the base model; Spider dev (1,034 questions) — 55.2% vs. 46.1% for the base model.
Training: QLoRA (4-bit NF4, r=16, alpha=32, dropout 0.05) on 2× T4 GPUs in Kaggle, lr 2e-4.
Limitations: strictly tailored to SQLite syntax (requires adjustments for Postgres or MySQL), may struggle with complex JOINs and nested subqueries, and generated SQL should only be executed over a read-only connection.
Discuss Telegram 0 55

Will it run?

Runs in F16

Fits without quantisation.

  • F16weights 2.9 GB · ready files3.7 GB
  • FP8 / INT8recommendedweights 1.4 GB2.3 GB
  • 4-bitweights 0.8 GB1.6 GB

Tip: go 8-bit.FP8 / INT8 is within about a percent of full precision in quality, takes half the memory and runs faster on RTX 40/50 and H100. Full precision only pays off for fine-tuning.

What to get for this model

Massed ComputeOpen Massed Compute

Your own GPU: FP8 / INT8 — from RTX 2060 · F16 — from RTX 2060

Massed Compute hourly prices, November 2026. The discounted price applies with the coupon at checkout.

Language models: with a 8192-token context.

Estimate: weights plus working memory. Long videos, high resolutions and long contexts need more.

Run it on Massed Compute

A ready script for the machine we advise above: paste it and the model downloads and starts.

  1. 1. Rent A30

    $0.26 an hour with the coupon. Pick the Ubuntu image when you create the machine.

    Open Massed Compute
  2. 2. Paste the script into the machine's terminal

    Connect over SSH (the Massed Compute panel shows the command) or open a terminal on the machine's desktop.

    HF_TOKEN / CIVITAI_TOKENThe script starts with token lines: paste yours between the quotes. The Hugging Face token is for gated models, the Civitai one for Civitai models.

    # 2003Rohansinha/Qwen2.5-Coder-1.5B-Text2SQL-LoRA — A30, FP8 / INT8
    # NVIDIA driver: CUDA 13 needs 580+. If older or missing, the current nvidia-open is installed (keep yours: KEEP_DRIVER=1)
    DRV=$(nvidia-smi --query-gpu=driver_version --format=csv,noheader 2>/dev/null | head -1 | cut -d. -f1)
    case "$DRV" in ""|*[!0-9]*) DRV=0 ;; esac
    if [ "$DRV" -lt 580 ] && [ -z "${KEEP_DRIVER:-}" ]; then
      . /etc/os-release && \
      curl -fsSLO "https://developer.download.nvidia.com/compute/cuda/repos/ubuntu${VERSION_ID//./}/x86_64/cuda-keyring_1.1-1_all.deb" && \
      sudo dpkg -i cuda-keyring_1.1-1_all.deb && rm -f cuda-keyring_1.1-1_all.deb && \
      sudo apt-get update && \
      sudo DEBIAN_FRONTEND=noninteractive apt-get install -y nvidia-open || \
      { echo "Could not install the NVIDIA driver: see the error above. To keep the current one, run with KEEP_DRIVER=1."; exit 1; }
      echo "NVIDIA driver updated. Reboot the machine (sudo reboot) and run this same script again."
      exit 0
    fi
    curl -LsSf https://astral.sh/uv/install.sh | sh
    source "$HOME/.local/bin/env"
    uv venv ~/vllm --python 3.12 && source ~/vllm/bin/activate
    uv pip install vllm
    vllm serve 2003Rohansinha/Qwen2.5-Coder-1.5B-Text2SQL-LoRA --max-model-len 8192 --quantization fp8
  3. 3. Use it

    The server listens on port 8000 and speaks the OpenAI API: http://localhost:8000/v1 on the machine. From your computer: ssh -L 8000:localhost:8000 user@machine-IP.

Stop the machine in the panel when you are done: it bills by the minute while it runs.

How to speed it up

Picked for this model type. Choose your hardware above to see what fits you exactly.

  • FlashAttention 44.0.0b40from November 18, 2026new release

    fastest exact attention on B200

    Kernels for Blackwell (beta). Exact attention without quantisation — fine for everything, LLMs included.

    How to enable: pip install flash-attn-4; vLLM and SGLang pick it up themselves.

    Earlier generations (2)
    • FlashAttention 3 — The Hopper (H100/H200) version with FP8.
    • FlashAttention 2 — The standard for RTX 30/40/50 and A100: exact attention without the memory cost of long contexts.
  • torch.compile2.15.0from October 28, 2026new release

    10–30% for free

    Compiles the model for your GPU. Slower first run, faster afterwards.

    How to enable: model = torch.compile(model); in ComfyUI the TorchCompileModel node.

  • MTP (multi-token prediction)0.34.0from November 16, 2026new release

    1.5–2× generation speed

    If the model has MTP layers (DeepSeek-V3/R1, Qwen3-Next, GLM, MiMo and others), it predicts several tokens per step itself.

    How to enable: vLLM / SGLang: --speculative-config '{"method": "mtp", "num_speculative_tokens": 1}'. Check the model card.

    Earlier generations (2)
    • EAGLE-3 — Speculative decoding with a small draft head. For models without their own MTP layers.
    • EAGLE-2 / Medusa — Earlier draft-head generations. Still work if that is all you find.
  • FP8 KV cache0.34.0from November 16, 2026new release

    2× more context in the same memory

    Keeps the attention cache in 8 bits, so long contexts fit where they did not.

    How to enable: vLLM: --kv-cache-dtype fp8; llama.cpp: -ctk q8_0 -ctv q8_0.

  • vLLM / SGLang0.34.0from November 16, 2026new release

    many times more tokens/s on a server

    Server engines: batching, FlashAttention, FP8 and speculative decoding out of the box.

    How to enable: vllm serve <model> — an OpenAI-compatible API.

    Earlier generations (1)
    • llama.cpp — GGUF models on any GPU, Mac or even CPU; some layers can stay in RAM.

Model family

  1. Qwen2.5-1.5Bfamily root
  2. Qwen2.5-Coder-1.5B
  3. Qwen2.5-Coder-1.5B-Instruct
  4. Qwen2.5-Coder-1.5B-Text2SQL-LoRA

New posts about versions of Qwen2.5-1.5B (quants, finetunes, LoRAs) will show up under the bell.

Quantised versions 12All on Hugging Face
Fine-tunes 12All on Hugging Face
LoRAs and adapters 5All on Hugging Face
Merges 2All on Hugging Face

About the model

2003Rohansinha/Qwen2.5-Coder-1.5B-Text2SQL-LoRA
Author
2003Rohansinha
Category
Language models
Released
Posted

On NeyroLearn

Hugging Face

550

Ask AI about this model

Answered by Gemini 4 · the chat is kept only in this tab

Comments