GLM-OCR-ONNX-q4f16

GLM-OCR ONNX (q4f16, browser-optimized)

Quantized q4f16 ONNX weights for GLM-OCR, packaged for Transformers.js WebGPU inference.

This repo contains only the q4f16 variant (~635 MB) instead of all precision variants in onnx-community/GLM-OCR-ONNX (~5+ GB).

Usage

import { AutoModelForImageTextToText, AutoProcessor } from '@huggingface/transformers';

const MODEL_ID = 'wolfofbackstreet/GLM-OCR-ONNX-q4f16';

const processor = await AutoProcessor.from_pretrained(MODEL_ID);
const model = await AutoModelForImageTextToText.from_pretrained(MODEL_ID, {
  device: 'webgpu',
  dtype: 'q4f16',
});

Source

Converted and quantized weights from onnx-community/GLM-OCR-ONNX.