Hello,
I am attempting to use TensorFlow with my project in AI Workbench. There is no image provided for TensorFlow in the default NGC Catalog base environment containers. After trying a few different approaches I’m out of luck.
Adding tensorflow[and-cuda] in the workbench packages menu, gets TensorFlow installed in the workbench container build, but its not compiled to work with newer cuda versions, meaning it doesn’t work for blackwell 100%.. (missing proper CUDA 12.0 PTX support)
I tried creating a custom container using the tensorflow image from NVCR.io as a base, which works in the terminal, but I am unable to figure out how to deploy it somewhere where a custom container can read from in AI workbench. I tried uploading a public image to hub.docker.com, and workbench said it isn’t supported. I tried uploading to ghcr.io and receive 404 errors in AI workbench when using the repo URL. NVCR.IO doesn’t seem to allow me to upload anything, I’m not sure I have access to that.
So I am left unable to use my 5090 with properly supported GPU acceleration for TensorFlow in AI Workbench. Any ideas welcome. Yes, there are docker images out there that will work on their own, but I was really hoping to use tensorflow with workbench somehow with my fancy card.
Best,
Jeremy K
If you are reporting a bug or error, consider submitting a Support Bundle to aiworkbench-ea@nvidia.com. This will help us solve your issue more quickly.
Please describe your issue or request: (tick the boxes after creating this topic):
Please tick the appropriate box to help us categorize your post
X Bug or Error
Feature Request
Documentation Issue
Other