You need to agree to share your Hugging Face information to access this model

This repository is publicly accessible, but you have to accept the conditions to access its files and content.

MVDiT-VideoUFO

The target model of Membership Inference for Generative Video Models by Raina Panda and Maty Bohacek. Code: github.com/maty-bohacek/t2v-mia. Data: matybohacek/t2v-mia.

This is the MVDiT-XL/2 of OpenVid-1M (Nan et al., 2025), trained by Wang & Yang on the entirety of VideoUFO for 20 epochs (72,500 steps). The checkpoint holds the EMA weights of the final step. It has 28 transformer blocks of width 1,152 (1.07B parameters), generates 16 frames at 256×256, and uses the stabilityai/sd-vae-ft-ema VAE and the DeepFloyd/t5-v1_1-xxl text encoder.

Since its complete training set is known, the model allows membership inference to be studied on a modern T2V architecture. Access is reviewed by the research team.

huggingface-cli download matybohacek/mvdit-videoufo mvdit_videoufo_16x256_ema_epoch20_step72500.pt --local-dir ckpts

Downloads last month: -; Downloads are not tracked for this model. How to track