1

I'm new in python, I have python 3.5, windows 10, AVX 5200. I want to install tensorflow on gpu. I tried vary blogs step by step, but I cannot get tensorflow on gpu, always get the following message:

    Your CPU supports instructions that this TensorFlow binary was not compiled to use: AVX

please can anyone help me how can I get tensorflow-gpu step by step? I will be thankfull to him.

alyaa mahdi
  • 29
  • 4
  • It is possible that you successfully install tensorflow-gpu, but your code gets to use tensorflow-cpu nonetheless, e.g., because of some environment setting. When you do `pip install tensorflow-gpu`, does the installation complete successfully? – Inon Peled Apr 02 '19 at 15:37
  • @InonPeled thank you for your reply. yes it says the installation complete successfully, but when I train the CNN model in python with pycharm, It give me this message. How can I enable my CNN to train and evaluate data using tensorflow-gpu, it is very slow and stuck with my tensorflow, can you please help me? – alyaa mahdi Apr 02 '19 at 17:14
  • Maybe start here: https://stackoverflow.com/questions/51306862/how-to-use-tensorflow-gpu – Inon Peled Apr 02 '19 at 17:17
  • Possible duplicate of [How to use TensorFlow GPU?](https://stackoverflow.com/questions/51306862/how-to-use-tensorflow-gpu) – Inon Peled Apr 02 '19 at 17:18
  • @InonPeled I'm followed what is mentioned, but I get: I C:\tf_jenkins\workspace\rel-win\M\windows-gpu\PY\35\tensorflow\core\platform\cpu_feature_guard.cc:137] Your CPU supports instructions that this TensorFlow binary was not compiled to use: AVX 2019-04-02 20:26:49.174284: I C:\tf_jenkins\workspace\rel-win\M\windows-gpu\PY\35\tensorflow\core\common_runtime\gpu\gpu_device.cc:1105] Found device 0 with properties: name: NVS 5200M major: 2 minor: 1 memoryClockRate(GHz): 1.344 pciBusID: 0000:01:00.0 totalMemory: 1.00GiB freeMemory: 823.00MiB – alyaa mahdi Apr 02 '19 at 17:34
  • 2019-04-02 20:26:49.174641: I C:\tf_jenkins\workspace\rel-win\M\windows-gpu\PY\35\tensorflow\core\common_runtime\gpu\gpu_device.cc:1168] Ignoring visible gpu device (device: 0, name: NVS 5200M, pci bus id: 0000:01:00.0, compute capability: 2.1) with Cuda compute capability 2.1. The minimum required Cuda capability is 3.0. [name: "/device:CPU:0" device_type: "CPU" memory_limit: 268435456 locality { } incarnation: 15798718000949594164 ] – alyaa mahdi Apr 02 '19 at 17:34
  • So it appears that you need to upgrade your CUDA libs. Again, refer to the instructions in the link that I pasted above. – Inon Peled Apr 02 '19 at 17:41

1 Answers1

0
  1. You have installed and get TF running.
  2. You are seeing a warning message, not an error at all. You have installed a precompiled package, which didn't enable AVX instruction set, AVX is SIMD instruction set of C.P.U.
  3. You may ensure tensorflow gpu is working by run:
sess = tf.Session(config=tf.ConfigProto(log_device_placement=True))
sess.close()
  1. Highly recommend using ananconda and install everything with conda.
  2. Find a book or just read the official guides.
pinxue
  • 1,736
  • 12
  • 17