How to install Tensorflow on Intel Xeon Processor based PC?
Question:
Not really a coding related question.
I am trying to install tensorflow on Intel Xeon(R) CPU E5620 which has Ubuntu 19.10. I can’t install it with pip( could someone please explain why). It does gets installed using pip but it just says,
illegal instruction (core dumped)
I have tried building it from the source, but that was a pretty painful process, takes way too much time.
My question is, is their any easier way to install tensorflow on my PC?
Answers:
Try using conda environment. I have no problem installing and using tensorflow on my machine. I am using tensorflow 1.15 btw and as far as I recall I installed it using pip install tensorflow-gpu==1.15
Cpu model: Intel(R) Xeon(R) CPU E5-2620 v4 @ 2.10GHz
OS: Ubuntu 18.04
conda environment python version: 3.5.6
- For Intel Z600
- Processor: Xeon(R) CPU E5620
- OS : Ubuntu 19.10
- GPU : GTX 1050
- CUDA : 10.0
- cuDNN : 7.4
- NCCL : 2.6
with the above configuration I successfully build tensorflow 1.14 from source, NO their is no other way to get tensorflow/tensorflow-gpu on this particular PC. I have experienced this that pip and conda don’t work when installing tensorflow and that building from source or using docker as mentioned on tf official documentation is the best possible way as for 15 May 2020.
I figured that someone in near future might want to build tensorflow from source on a xeon based CPU, so I wrote how I did it step by step here, https://github.com/kazzastic/Tensorflow-BuildFromSource
Not really a coding related question.
I am trying to install tensorflow on Intel Xeon(R) CPU E5620 which has Ubuntu 19.10. I can’t install it with pip( could someone please explain why). It does gets installed using pip but it just says,
illegal instruction (core dumped)
I have tried building it from the source, but that was a pretty painful process, takes way too much time.
My question is, is their any easier way to install tensorflow on my PC?
Try using conda environment. I have no problem installing and using tensorflow on my machine. I am using tensorflow 1.15 btw and as far as I recall I installed it using pip install tensorflow-gpu==1.15
Cpu model: Intel(R) Xeon(R) CPU E5-2620 v4 @ 2.10GHz
OS: Ubuntu 18.04
conda environment python version: 3.5.6
- For Intel Z600
- Processor: Xeon(R) CPU E5620
- OS : Ubuntu 19.10
- GPU : GTX 1050
- CUDA : 10.0
- cuDNN : 7.4
- NCCL : 2.6
with the above configuration I successfully build tensorflow 1.14 from source, NO their is no other way to get tensorflow/tensorflow-gpu on this particular PC. I have experienced this that pip and conda don’t work when installing tensorflow and that building from source or using docker as mentioned on tf official documentation is the best possible way as for 15 May 2020.
I figured that someone in near future might want to build tensorflow from source on a xeon based CPU, so I wrote how I did it step by step here, https://github.com/kazzastic/Tensorflow-BuildFromSource