While referring to "Installing TensorFlow on Windows" on the official website, record the contents of the installation in your environment.
-OS: Windows 7 Professional Service Pack 1 64bit -GPU: NVIDIA GeForce GTX 650 → Since the compute capability is "3.0", there are various dangers.
Actually, this was already installed, so I didn't bother to install it for this reason. A memo of the installed module for the time being. ・ CUDA 8.0 ・ CuDNN 5.1
Python 3.6 (32bit version) was already included, but it must be Python 3.5 (64bit version), so here I installed it from /). (It has been confirmed that an error occurs in 3.5 (32bit version) and 3.6 (32bit version, 64bit version))
After that, when I tried to install TensorFlow, it said "Pip is old", so I updated it.
Python -m pip install --upgrade pip
It's finally TensorFlow installation. Use Pip.
pip3 install --upgrade tensorflow-gpu
I installed the GPU version.
-The environment variable "PATH" was too long to add the CUDA / cuDNN path. → An error is displayed when importing TensorFlow
-The execution log is displayed
I c:\tf_jenkins\home\workspace\release-win\device\gpu\os\windows\tensorflow\core\common_runtime\gpu\gpu_device.cc:885] Found device 0 with properties:
name: GeForce GTX 650
major: 3 minor: 0 memoryClockRate (GHz) 1.0585
pciBusID 0000:01:00.0
Total memory: 1.00GiB
Free memory: 190.88MiB
→ Unsolved → Is that something like that?
-An error is displayed (executed)
E c:\tf_jenkins\home\workspace\release-win\device\gpu\os\windows\tensorflow\core\framework\op_kernel.cc:943] OpKernel ('op: "BestSplits" device_type: "CPU"') for unknown op: BestSplits
E c:\tf_jenkins\home\workspace\release-win\device\gpu\os\windows\tensorflow\core\framework\op_kernel.cc:943] OpKernel ('op: "CountExtremelyRandomStats" device_type: "CPU"') for unknown op: CountExtremelyRandomStats
E c:\tf_jenkins\home\workspace\release-win\device\gpu\os\windows\tensorflow\core\framework\op_kernel.cc:943] OpKernel ('op: "FinishedNodes" device_type: "CPU"') for unknown op: FinishedNodes
E c:\tf_jenkins\home\workspace\release-win\device\gpu\os\windows\tensorflow\core\framework\op_kernel.cc:943] OpKernel ('op: "GrowTree" device_type: "CPU"') for unknown op: GrowTree
...
→ Unresolved (here is being confirmed) → The CPU version has been fixed, but is the GPU version still?
★ After that, it turned out that TensorFlow brought by pip is not the latest ★ Bring the latest whl from here and pip the file to eliminate the above error etc.
pip3 install tensorflow-1.0.0-cp35-cp35m-win_amd64.whl
You'll want a tutorial or something, so it's a good idea to bring it from GitHub source set.
Recommended Posts