Name: glm 4vq
Author: nikravan

Overview

GLM-4VQ is a specialized model optimized for high-precision document question-answering (DocQA). Unlike general-purpose LLMs that may struggle with long-context retrieval or hallucinate details from dense texts, this model is engineered to extract and synthesize information directly from provided source documents. For developers, this makes it an ideal engine for building RAG (Retrieval-Augmented Generation) pipelines, automated technical support bots, or internal knowledge base interfaces. It balances accuracy with efficiency, offering a reliable alternative for production environments where groundedness is more critical than creative generation. Integration is straightforward via standard API endpoints, fitting easily into existing document processing workflows.

Highlights

Optimized for high-accuracy document question-answering tasks
Ideal for RAG pipelines and knowledge base integration
Reduces hallucinations by focusing on provided source text
Open-source flexibility under the Apache-2.0 license

Usage

Install

# Install Hugging Face transformers
pip install transformers torch

SDK Usage

# Load model with transformers
from transformers import AutoModel, AutoTokenizer

model = AutoModel.from_pretrained("nikravan/glm-4vq")
tokenizer = AutoTokenizer.from_pretrained("nikravan/glm-4vq")

Hugging Face Download

We recommend downloading the model via the Hugging Face CLI or Hub SDK.

Guidance：Before downloading, install huggingface_hub with:

Guidance

pip install -U huggingface_hub

CLI Download

Download the full repository

huggingface-cli download nikravan/glm-4vq

Download a single file to a local folder (e.g. config.json into ./dir)

huggingface-cli download nikravan/glm-4vq config.json --local-dir ./dir

See the official docs for more CLI options

SDK Download

# 模型下载
from huggingface_hub import snapshot_download
model_dir = snapshot_download('nikravan/glm-4vq')

Git Download

Make sure git-lfs is installed first

Git Download

git lfs install
git clone https://huggingface.co/nikravan/glm-4vq

To skip LFS large-file downloads, use:

Skip LFS

GIT_LFS_SKIP_SMUDGE=1 git clone https://huggingface.co/nikravan/glm-4vq

Model files are hosted on the Hugging Face Hub — download directly via HF CLI / SDK / Git, not through this site.

PyTorch / Transformers Usage

Install Transformers

pip install -U transformers torch

Load the model and run inference

from transformers import AutoModelForCausalLM, AutoTokenizer

model = AutoModelForCausalLM.from_pretrained('nikravan/glm-4vq')
tokenizer = AutoTokenizer.from_pretrained('nikravan/glm-4vq')

Full Documentation

来源: HuggingFace

---
language:

en

de

fr

fa

ar

tr

es

it

zh

ko

ja

hi

metrics:

accuracy

pipeline_tag: document-question-answering
tags:

text-generation-inference

---

MultiModal MultiLingual (3ML)

This model is 4bit quantized of glm-4v-9b Model (Less than 9G).

It excels in document, image, chart questioning answering and delivers superior performance over GPT-4-turbo-2024-04-09, Gemini 1.0 Pro, Qwen-VL-Max, and Claude 3 Opus.

Some part of the original Model changed and It can excute on free version of google colab.

Try it: ![Open In Colab](https://colab.research.google.com/drive/1aZGX9f5Yw1WbiOrS3TpvPk_UJUP_yYQU?usp=sharing)

[![Github Source]](https://github.com/nikravan1/3ML)

Note: For optimal performance with document and image understanding, please use English or Chinese. The model can still handle chat in any supported language.

About GLM-4V-9B

GLM-4V-9B is a multimodal language model with visual understanding capabilities. The evaluation results of its related classic tasks are as follows:

| | MMBench-EN-Test | MMBench-CN-Test | SEEDBench_IMG | MMStar | MMMU | MME | HallusionBench | AI2D | OCRBench |
|-------------------------|---------------------|---------------------|-------------------|------------|----------|---------|--------------------|----------|--------------|
| | 英文综合 | 中文综合 | 综合能力 | 综合能力 | 学科综合 | 感知推理 | 幻觉性 | 图表理解 | 文字识别 |
| GPT-4o, 20240513 | 83.4 | 82.1 | 77.1 | 63.9 | 69.2 | 2310.3 | 55 | 84.6 | 736 |
| GPT-4v, 20240409 | 81 | 80.2 | 73 | 56 | 61.7 | 2070.2 | 43.9 | 78.6 | 656 |
| GPT-4v, 20231106 | 77 | 74.4 | 72.3 | 49.7 | 53.8 | 1771.5 | 46.5 | 75.9 | 516 |
| InternVL-Chat-V1.5 | 82.3 | 80.7 | 75.2 | 57.1 | 46.8 | 2189.6 | 47.4 | 80.6 | 720 |
| LlaVA-Next-Yi-34B | 81.1 | 79 | 75.7 | 51.6 | 48.8 | 2050.2 | 34.8 | 78.9 | 574 |
| Step-1V | 80.7 | 79.9 | 70.3 | 50 | 49.9 | 2206.4 | 48.4 | 79.2 | 625 |
| MiniCPM-Llama3-V2.5 | 77.6 | 73.8 | 72.3 | 51.8 | 45.8 | 2024.6 | 42.4 | 78.4 | 725 |
| Qwen-VL-Max | 77.6 | 75.7 | 72.7 | 49.5 | 52 | 2281.7 | 41.2 | 75.7 | 684 |
| GeminiProVision | 73.6 | 74.3 | 70.7 | 38.6 | 49 | 2148.9 | 45.7 | 72.9 | 680 |
| Claude-3V Opus | 63.3 | 59.2 | 64 | 45.7 | 54.9 | 1586.8 | 37.8 | 70.6 | 694 |
| GLM-4v-9B | 81.1 | 79.4 | 76.8 | 58.7 | 47.2 | 2163.8 | 46.6 | 81.1 | 786 |
This repository is the model repository of 4bit quantized of GLM-4V-9B model, supporting 8K context length.

Quick Start

Use colab model or this python script.

python

import torch
from transformers import AutoModelForCausalLM, AutoTokenizer
from PIL import Image
device = "cuda"
modelPath="nikravan/glm-4vq"
tokenizer = AutoTokenizer.from_pretrained(modelPath, trust_remote_code=True)
model = AutoModelForCausalLM.from_pretrained(
    modelPath,
    torch_dtype=torch.bfloat16,
    low_cpu_mem_usage=True,
    trust_remote_code=True,
    device_map="auto"
)

query ='explain all the details in this picture'
image = Image.open("a3.png").convert('RGB')
#image=""
inputs = tokenizer.apply_chat_template([{"role": "user", "image": image, "content": query}],
                                       add_generation_prompt=True, tokenize=True, return_tensors="pt",
                                       return_dict=True)  # chat with image mode
inputs = inputs.to(device)
gen_kwargs = {"max_length": 2500, "do_sample": True, "top_k": 1}
with torch.no_grad():
    outputs = model.generate(inputs, gen_kwargs)
    outputs = outputs[:, inputs['input_ids'].shape[1]:]
    print(tokenizer.decode(outputs[0]))