Skip to content
Language models

Tehma/text2sql-qwen2.5-coder-3b-onnx

browser-based SQL generation model

2 min read

This is a fine-tuned version of Qwen2.5-Coder-3B-Instruct (a 3-billion-parameter code language model) trained with QLoRA (quantized low-rank adaptation) on the Spider dataset for text-to-SQL generation. The main highlight of this release is the conversion of weights into a 4-bit ONNX format (q4f16), allowing it to run directly in the browser using Transformers.js and WebGPU hardware acceleration. It is built for developers who want to add local text-to-SQL capabilities without sending user data to a server.

Key features

  • Exported to ONNX (a cross-platform model inference format) with 4-bit weight quantization (total file size is around 1.74 GB).
  • Runs in the browser via WebGPU at high generation speeds (around 45 tokens per second on an RTX 3060).
  • Tailored for generating SQLite queries from a given database schema and a natural language question.
  • Local execution eliminates the need to deploy heavy backend servers.

What you can do

  • Integrate the model into a Transformers.js web app to build interactive browser-based database interfaces.
  • Build a local analytics assistant that translates natural language user queries into SQL code.

Example

import { AutoTokenizer, AutoModelForCausalLM } from "@huggingface/transformers";

const id = "Tehma/text2sql-qwen2.5-coder-3b-onnx";
const tokenizer = await AutoTokenizer.from_pretrained(id);
const model = await AutoModelForCausalLM.from_pretrained(id, { dtype: "q4f16", device: "webgpu" });

const inputs = tokenizer.apply_chat_template(messages, { add_generation_prompt: true, return_dict: true });
const out = await model.generate({ ...inputs, max_new_tokens: 256, do_sample: false });
Read moreDetails
Benchmarks: On the Spider dataset (first 200 questions of the dev split), the q4f16 ONNX model achieves 64.0% execution accuracy.
Hardware requirements: A WebGPU-compatible device (for example, an RTX 3060 GPU delivers ~45 tokens/s).
Limitations: Trained on the Spider dataset (English only, SQLite dialect, academic schemas); not tested on other SQL dialects or real-world business databases.
License: Qwen Research License (non-commercial use only).
Discuss Telegram 0 3

Will it run?

Runs in BF16

Fits without quantisation.

  • BF16weights 0.0 GB · ready files0.6 GB
  • FP8 / INT8recommendedweights 0.0 GB0.6 GB
  • 4-bitweights 0.0 GB0.6 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 · BF16 — from RTX 2060

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

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.

    # Tehma/text2sql-qwen2.5-coder-3b-onnx — 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 Tehma/text2sql-qwen2.5-coder-3b-onnx --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. Qwen2.5-3Bfamily root
  2. Qwen2.5-Coder-3B
  3. Qwen2.5-Coder-3B-Instruct
  4. text2sql-qwen2.5-coder-3b-onnx

New posts about versions of Qwen2.5-3B (quants, finetunes, LoRAs) will show up under the bell.

Quantised versions 12All on Hugging Face
Fine-tunes 12All on Hugging Face
LoRAs and adapters 2All on Hugging Face
Merges 3All on Hugging Face

About the model

Tehma/text2sql-qwen2.5-coder-3b-onnx
Author
Tehma
Category
Language models
Released
Posted

On NeyroLearn

Hugging Face

30

Ask AI about this model

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

Comments