MPP-LLaVA

Multimodal Pipeline Parallel based on Qwen-LM for training large language models with support for video and image inputs.

GitHub

GraphCanon updated Sep 20, 2026 · GitHub synced Sep 20, 2026

40views this month

690 stars33 forksLast push Mar 10, 2025 Jupyter Notebook

Decision brief

MPP-LLaVA enables efficient fine-tuning of Qwen-based multimodal language models on consumer-grade GPUs for video, image, or multiple images inputs.

Good fit when

You are working with a limited GPU budget but need to fine-tune large MLLMs like Qwen14B using pipeline parallelism.
Handling video or image data is essential for your project but high-end hardware is not available.

Avoid when

High-performance and high-capacity GPUs are readily accessible, allowing other tools to leverage more comprehensive parallelisms beyond consumer-grade GPUs limitations.
The project does not require the handling of video or image data as inputs for MLLM fine-tuning.

Observed Jul 12, 2026 · Source: enrich:decision_facts

MPP-LLaVA alternatives Trust report

Verify the decision

Maintenance and security

Full trust report

Maintenance: Dormant (558d since push); As of Sep 19, 2026
Provenance: Not a fork · Personal account; As of Sep 19, 2026
Security (OSV): No lockfile; As of Jul 11, 2026

Public GitHub metadata and optional OSV scans. Signals, not a guarantee. Trust methodology.

Install

git clone https://github.com/Coobiw/MPP-LLaVA

Similar tools

Same-category neighbours. No typed graph edges are catalogued for this tool yet.

ollama

Get up and running with Kimi, GLM, MiniMax, DeepSeek, gpt-oss, Qwen, Gemma and other models.

181k★

open-webui

User-friendly AI Interface

152k★

awesome-gpt-image-2

World's largest GPT Image 2 prompt library, updated daily

9.9k★

Evidence and technical details

Sourced facts, taxonomy, compatibility claims, README excerpt, and machine-readable endpoints.

Overview

A project supporting the fine-tuning of multimodal large language models (MLLMs) like Qwen14B using pipeline parallelism to handle video/image/multi-image data. It allows training on consumer-grade GPUs such as RTX3090/4090 with 24GB VRAM.

Capability facts

Languages: jupyter notebook, python
Source: github.language+pyproject.toml · Sep 19, 2026

Compatibility

Sourced claims from the README excerpt - not unsourced marketing copy.

Python runtimePython

Source: README excerpt (regex_v1, Sep 19, 2026)

conda create -n minigpt4qwen python=3.8 && conda activate minigpt4qwen

Source link

README

Installation License 本仓库的许多代码是基于Lavis 的，其采用 BSD 3 Clause License. 本仓库采用Qwen 7B Chat，支持商用和科研、开发用途，其License为LICENSE

For agents

This page has a .md twin and JSON over the API.

/tools/coobiw-mpp-llava.md GET /api/graphcanon/tools/coobiw-mpp-llava GET /api/graphcanon/graph?tool=coobiw-mpp-llava