Skip to content

Error message: "Failed: Cannnot set version_counter for inference tensor" #115

Description

@atmh2

Error message occurs for depth rendering. I tried a couple different videos as well as a picture. Model used was Video Depth Anything Small. I tried a couple different batch values and overlap values. I tried different resolutions. I tried a couple different output encoders.

As it happens I have an old depth map for a video from a previous version of VD3D, so I was able to verify the 3D rendering works fine. Error is only with depth mapping.

Just downloaded the latest version of VD3D. I used previous versions (before the dark theme) successfully on my linux/AMD GPU system. Running VD3D on a windows 10 VM with gpu passthrough. GPU is a Radeon AI PRO R9700. I installed the DirectML build. I originally encounter and ffmpeg error as it appears VD3D was expecting ffmpeg to be included, but it was easy enough to install since the path it was looking for was provided in the error message.

Huge fan of the work you're doing here. I plan to buy the pro version once the depth engine error issue is resolved and I can do more with it. Right now I can only do the one single video I already have the depth map for :)

Metadata

Metadata

Assignees

No one assigned

    Labels

    No labels
    No labels

    Projects

    No projects

    Milestone

    No milestone

    Relationships

    None yet

    Development

    No branches or pull requests

    Issue actions