Practical overview
When it fits
Use Xmdpt when you want code for a research experiment in pose-guided person-image synthesis and can work with the documented Python, CUDA and DeepFashion setup.
Who it is for
- Researchers exploring diffusion-transformer methods for person-image synthesis.
- Developers preparing DeepFashion experiments involving cross-view, pose-guided image generation.
Documented uses
- Prepare the documented Python environment and DeepFashion LMDB data for experiments.
- Train X-MDPT and save model checkpoints.
- Run the documented inference workflow on DeepFashion test data and save generated samples.
- Compare the repository's listed model variants using its reported parameter and inference-time figures.
Requirements & limits
The repository is public and lists an Apache-2.0 license.
- The README reports a PyTorch 1.12 and CUDA 11.6 setup using Docker, and gives Python 3.8 Conda setup instructions.
- Training and documented test-set inference depend on DeepFashion data in the described LMDB structure.
- The README reports training and testing its models on one A100 with 80 GB of memory.
- The arbitrary-image inference example is explicitly marked not implemented.
Check current requirements in the official documentation. Source-based guidance does not mean OpenViST has run or independently tested this project.
Set up Xmdpt for DeepFashion experiments
Prepare the documented Python environment and dataset structure for training or test-set inference.
Before you start
- Python 3.8 and Conda
- PyTorch 1.12 with CUDA 11.6 is the README's reported setup
- DeepFashion data processed into LMDB format, following the repository's described directory structure
Steps
Create and activate the Conda environment with Python 3.8.
Install the repository requirements with pip.
Prepare DeepFashion in LMDB format and arrange the files under the documented datasets/deepfashion structure.
Expected outcome
An environment and dataset layout prepared for the repository's documented training and test-set inference stages.
Guide limitations
- The README reports testing with PyTorch 1.12 and CUDA 11.6 in Docker; it does not guarantee other configurations.
- This setup recipe does not cover execution or confirm that the environment will work on a particular machine.
Run inference on the DeepFashion test set
Use the repository's documented inference script to save test-set image samples.
Before you start
- The prepared Xmdpt environment and DeepFashion test-set data
- Downloaded model checkpoints and VAE placed where infer_xmdpt.py expects them
- A CUDA-capable setup; the README reports using one A100 with 80 GB for its models
Steps
Download the model checkpoints and VAE and put them in the locations expected by the default infer_xmdpt.py configuration.
Run the documented CUDA_VISIBLE_DEVICES=0 python infer_xmdpt.py command.
Expected outcome
The README says the script saves output image samples in the repository's test_img area.
Guide limitations
- The arbitrary-image inference command is explicitly marked not implemented.
- The README's hardware and timing figures are reported results, not compatibility guarantees.
Sources & review
Source record details
- Source availability
- Active
- Sources checked
- 15 October 2026
- Original website title
- trungpx/xmdpt · Official homepage title · retained in its source language