-1

I try to run tensorflow 1.13.1 inside a docker (the image with the wanted configuration is evariste/autodl:gpu-latest).

The docker has access to a RTX 2080 Ti GPU.

I get the following error:

2020-09-10 16:09:47.428460: F tensorflow/core/platform/cpu_feature_guard.cc:37] The TensorFlow library was compiled to use SSE4.1 instructions, but these aren't available on your machine.

Adrien Pavao
  • 452
  • 5
  • 11

1 Answers1

1

SSE4.1 is an instruction set supported by CPU, not GPU. Thus you need to check if your CPU supports it; more discussions about this topic can be found here.

Minh Nguyen
  • 755
  • 5
  • 11
  • You are right. My CPU (Intel Xeon Gold 6126) supports SSE4.2 instruction set. – Adrien Pavao Sep 17 '20 at 15:16
  • Was the above docker image built to support SSE4.1? If it's not (and generally Tensorflow is built without SSE4.1 support), then you can safely ignore that warning. – Minh Nguyen Sep 17 '20 at 15:27
  • I don't know. This is what the log suggests. Personally I aim at running the code on GPU. – Adrien Pavao Sep 17 '20 at 16:17
  • Nevermind, based on the log the image probably supports SSE4.1, but since you only care about GPU you can ignore the warning, it''s not an error anyway. – Minh Nguyen Sep 18 '20 at 00:29
  • It seems like a real error because Python crashes when I try to import tensorflow. When running with tensorflow 1.5.1 I have no issue (because this version does support SSE4.2). Now I'm stuck because I really need to run tf 1.13.1 – Adrien Pavao Sep 18 '20 at 11:07
  • Then I think it's probably best for you to try to contact people who made or work closely work with the above docker image. Since I'm not familiar with the above docker image, I'm sorry that I cannot help you much. – Minh Nguyen Sep 20 '20 at 07:09