All Tools
P
OtherFreeOpen Source
PADDLEMIX
Multimodal understanding and generation toolkit on PaddlePaddle
Apache-2.0
ABOUT
Multimodal training usually means separate stacks for VLMs, diffusion, and video generation. PaddleMIX puts Qwen-VL, LLaVA, Stable Diffusion, and related models on PaddlePaddle with shared data tools so teams pretrain, fine-tune, and generate from one multimodal suite.
INTEGRATION GUIDE
1. Fine-tune Qwen-VL or LLaVA for image captioning and visual QA
2. Generate images or video with PPDiffusers and Stable Diffusion
3. Run document understanding with PP-DocBee on scanned pages
4. Prepare multimodal datasets with DataCopilot before training
TAGS
pythonmultimodalpaddlepaddlediffusionvision-languagellavaqwen-vl