Help! CUDA 7.5 or 8 -- deviceQuery failing -- not working on MacbookPro(2013) OS X 10.11 / GT750M

My setup: MacbookPro late 2013; OS X 10.11.6; 16 GM RAM; NVIDIA GPU GeForce GT 750M with 2GB VRAM
Xcode: Version 7.3 (7D175)

I’m at my wits end trying to figure out how to get CUDA + CUDNN installation working with my TensorFlow. (https://www.tensorflow.org/get_started/os_setup#optional_setup_gpu_for_mac)

I followed the installation instructions here: (https://developer.nvidia.com/cuda-downloads)

With both CUDA 7.5 or CUDA 8.0, the deviceQuery is failing.

$ ./deviceQuery Starting...

 CUDA Device Query (Runtime API) version (CUDART static linking)

cudaGetDeviceCount returned 38
-> no CUDA-capable device is detected
Result = FAIL

#1 Results With CUDA 7.5: (a couple of months ago)

  • installed CUDA 7.5 and CUDNN 5.x
  • deviceQuery did not work
  • uninstalled CUDA 7.5 using uninstall perl script

#2 Results with CUDA 8.: (one month ago)

  • installed CUDA 8 (From Nvidia site, with local dmg installer)
  • installed Cudnn libs as required
  • deviceQuery did not work !!
  • uninstalled CUDA using the uninstall perl script

#3 Re-installed CUDA 7.5: (last week)

  • Using dmg: cuda_7.5.27_mac.dmg (local, not network)
  • again deviceQuery failed as before

I’ve now spent roughly two months (off and on) trying to figure out how to get deviceQuery to report correct status – but no dice.
[My understanding is that once deviceQuery reports correct GPU info… TensorFlow will be able to recognize and use the GPU.
I’ll of course use the GPU capable TensorFlow library]

Can anybody help?? Anybody seen this problem? I’m not sure what I’m missing.
Thanks!!
AtulA

---------- Driver ----------

CUDA Driver Version: 7.5.27
GPU Driver Version: 10.10.14 310.42.25f02


Model Name: MacBook Pro
Model Identifier: MacBookPro11,3

Processor Name: Intel Core i7
Processor Speed: 2.3 GHz
Number of Processors: 1
Total Number of Cores: 4
L2 Cache (per Core): 256 KB
L3 Cache: 6 MB
Memory: 16 GB
Boot ROM Version: MBP112.0138.B17
SMC Version (system): 2.19f12

NVIDIA GeForce GT 750M:
Chipset Model: NVIDIA GeForce GT 750M

Type: GPU
Bus: PCIe
PCIe Lane Width: x8
VRAM (Total): 2048 MB
Vendor: NVIDIA (0x10de)
Device ID: 0x0fe9
Revision ID: 0x00a2
ROM Revision: 3776
gMux Version: 4.0.8 [3.2.8]

Hmm – so I tried – one more time –
– Uninstalled existing CUDA
– Reinstalled CUDA 8 and cuDNN 5.1 – for OS X 10.11

One very basic question:
How would one verify if CUDA / cuDNN is installed correctly??

When I run kextstat, I get this:

kextstat | grep -i nvidia
  120    2 0xffffff7f80f1b000 0x279000   0x279000   com.apple.nvidia.driver.NVDAResman (10.1.0) 7C8DF8B7-32F9-3EFF-9B5A-E8EAD8EB3A42 <100 94 83 12 5 4 3 1>
  121    0 0xffffff7f8119f000 0x1af000   0x1af000   com.apple.nvidia.driver.NVDAGK100Hal (10.1.0) EB8A5980-AB59-368A-8244-60A00C7A933C <120 12 4 3>
  138    0 0xffffff7f8269c000 0x2000     0x2000     com.nvidia.CUDA (1.1.0) DD792765-CA28-395A-8593-D6837F05C4FF <4 1>
 ~ ▶ Developer ❯ … ❯ x86_64 ❯ darwin ❯ release ▶ $ ▶

Does this mean CUDA driver is installed correctly, or not?

The driver: (from Apple > System Preferences > CUDA):
CUDA Driver Version: 8.0.57
GPU Driver Version: 10.10.14 310.42.25f02

my results:

./deviceQuery 
./deviceQuery Starting...

 CUDA Device Query (Runtime API) version (CUDART static linking)

cudaGetDeviceCount returned 38
-> no CUDA-capable device is detected
Result = FAIL

I’ve detailed my CUDA / GPU problem here, on the TensorFlow github repo – in case anybody as similar experiences:

See here: (by @laventura) https://github.com/tensorflow/tensorflow/issues/2882

and here: (also by @laventura): https://github.com/tensorflow/tensorflow/issues/4858

IN brief – because deviceQuery does not recognize the GPU, TensorFlow gives an error.

For the life of me, I can’t find where this error is coming from:
"libcuda reported version is: 310.42.25; kernel reported version is: Invalid argument: expected %d.%d or %d.%d.%d form for driver version; got “”

What does this mean (from Apple > System Preference > CUDA)

<b>GPU Driver Version: 10.10.14 310.42.25f02</b>
CUDA Driver Version: 8.0.57

Is there a way to upgrade this NVIDIA Display driver from 310.42.2f02 to something new?
How? and where would I get the update from?
(This is for Mac OS X 10.11.6]

Thanks!

I am having the same problem , but with a gtx 770 . Very frustrating. I hope someone (Apple, nvidia) can shed some light on this problem. thanks for the post. fti here are my details below.

104 2 0xffffff7f81195000 0x279000 0x279000 com.apple.nvidia.driver.NVDAResman (10.1.0) 7C8DF8B7-32F9-3EFF-9B5A-E8EAD8EB3A42 <103 100 89 12 5 4 3 1>
105 0 0xffffff7f81419000 0x1af000 0x1af000 com.apple.nvidia.driver.NVDAGK100Hal (10.1.0) EB8A5980-AB59-368A-8244-60A00C7A933C <104 12 4 3>
145 0 0xffffff7f82e23000 0x2000 0x2000 com.nvidia.CUDA (1.1.0) DD792765-CA28-395A-8593-D6837F05C4FF <4 1>

CUDA:

Version: 8.0.90
Obtained from: Identified Developer
Last Modified: 2/14/18, 12:46 PM
Kind: Intel
64-Bit (Intel): Yes
Signed by: Developer ID Application: NVIDIA Corporation (6KR3T733EC), Developer ID Certification Authority, Apple Root CA
Location: /Library/Frameworks/CUDA.framework
Private: No

CUDA Driver Version: 8.0.90
GPU Driver Version: 10.10.14 310.42.25f02