NVIDIA / tensorflow

An Open Source Machine Learning Framework for Everyone
https://developer.nvidia.com/deep-learning-frameworks
Apache License 2.0
962 stars 144 forks source link

Prebuilt docker images compute capability question #57

Open bmccord2 opened 2 years ago

bmccord2 commented 2 years ago

Sorry, I am not sure if this is the correct place to ask this question, I am using an image from the nvcr.io repository and noticed that I was not getting gpu support on the k80 I was using, it appears that the latest tensorflow image, 22.01-tf1-py3 has a minimum compute capability set to 5.2 and I require 3.7.

My question is--is there anyway to look at the history of when this changed, is their a separate repository for those docker builds? I see that the nvbuild.sh script seems to use the value "all" for the compute capability which ends up setting the minimum compute capability to 5.2, but there seems to be no straightforward way to track history, or see when that changed (assuming it has changed at any point), its not in any documentation that I have found.