Closed ErfolgreichCharismatisch closed 6 years ago
yunfanye
commented
7 years ago One thing that could go wrong is that the image did not fit in the memory. GTX960 has a 2GB memory, which is relatively small compared with your image size. Try to reduce the size of the image and see if it drastically improves.
anishathalye
commented
7 years ago Are you using TensorFlow GPU? Does it print messages to stderr saying that it's running on the GPU?
On Jul 7, 2017, at 10:24 AM, yunfanye notifications@github.com wrote:
One thing that could go wrong is that the image did not fit in the memory. GTX960 has a 2GB memory, which is relatively small compared with your image size. Try to reduce the size of the image and see if it drastically improves.
— You are receiving this because you are subscribed to this thread. Reply to this email directly, view it on GitHub, or mute the thread.
ErfolgreichCharismatisch
commented
7 years ago This is the output(I changed the output to german in case you wonder, python = python 3.5):
E:\Python\Projekte\neural-style>python.exe E:\Python\Projekte\neural-style\neural_style.py --style-layer-weight-exp 0.3 --iterations 1000 --content "Nacht - 0002.jpg" --styles image001.jpg --output "RA.jpg"
2017-07-07 19:59:56.645752: W c:\tf_jenkins\home\workspace\release-win\m\windows\py\35\tensorflow\core\platform\cpu_feature_guard.cc:45] The TensorFlow library wasn't compiled to use SSE instructions, but these are available on your machine and could speed up CPU computations.
2017-07-07 19:59:56.645858: W c:\tf_jenkins\home\workspace\release-win\m\windows\py\35\tensorflow\core\platform\cpu_feature_guard.cc:45] The TensorFlow library wasn't compiled to use SSE2 instructions, but these are available on your machine and could speed up CPU computations.
2017-07-07 19:59:56.646243: W c:\tf_jenkins\home\workspace\release-win\m\windows\py\35\tensorflow\core\platform\cpu_feature_guard.cc:45] The TensorFlow library wasn't compiled to use SSE3 instructions, but these are available on your machine and could speed up CPU computations.
2017-07-07 19:59:56.646302: W c:\tf_jenkins\home\workspace\release-win\m\windows\py\35\tensorflow\core\platform\cpu_feature_guard.cc:45] The TensorFlow library wasn't compiled to use SSE4.1 instructions, but these are available on your machine and could speed up CPU computations.
2017-07-07 19:59:56.646364: W c:\tf_jenkins\home\workspace\release-win\m\windows\py\35\tensorflow\core\platform\cpu_feature_guard.cc:45] The TensorFlow library wasn't compiled to use SSE4.2 instructions, but these are available on your machine and could speed up CPU computations.
2017-07-07 19:59:56.646425: W c:\tf_jenkins\home\workspace\release-win\m\windows\py\35\tensorflow\core\platform\cpu_feature_guard.cc:45] The TensorFlow library wasn't compiled to use AVX instructions, but these are available on your machine and could speed up CPU computations.
Die Optimierung hat begonnen...
Durchlauf 1/1000
Durchlauf 2/1000
...I did some installing back and forth, now I get
2017-07-07 21:05:45.751030: W c:\tf_jenkins\home\workspace\release-win\m\windows-gpu\py\35\tensorflow\core\platform\cpu_feature_guard.cc:45] The TensorFlow library wasn't compiled to use SSE instructions, but these are available on your machine and could speed up CPU computations.
2017-07-07 21:05:45.751180: W c:\tf_jenkins\home\workspace\release-win\m\windows-gpu\py\35\tensorflow\core\platform\cpu_feature_guard.cc:45] The TensorFlow library wasn't compiled to use SSE2 instructions, but these are available on your machine and could speed up CPU computations.
2017-07-07 21:05:45.751423: W c:\tf_jenkins\home\workspace\release-win\m\windows-gpu\py\35\tensorflow\core\platform\cpu_feature_guard.cc:45] The TensorFlow library wasn't compiled to use SSE3 instructions, but these are available on your machine and could speed up CPU computations.
2017-07-07 21:05:45.751525: W c:\tf_jenkins\home\workspace\release-win\m\windows-gpu\py\35\tensorflow\core\platform\cpu_feature_guard.cc:45] The TensorFlow library wasn't compiled to use SSE4.1 instructions, but these are available on your machine and could speed up CPU computations.
2017-07-07 21:05:45.751642: W c:\tf_jenkins\home\workspace\release-win\m\windows-gpu\py\35\tensorflow\core\platform\cpu_feature_guard.cc:45] The TensorFlow library wasn't compiled to use SSE4.2 instructions, but these are available on your machine and could speed up CPU computations.
2017-07-07 21:05:45.751769: W c:\tf_jenkins\home\workspace\release-win\m\windows-gpu\py\35\tensorflow\core\platform\cpu_feature_guard.cc:45] The TensorFlow library wasn't compiled to use AVX instructions, but these are available on your machine and could speed up CPU computations.
2017-07-07 21:05:46.063231: I c:\tf_jenkins\home\workspace\release-win\m\windows-gpu\py\35\tensorflow\core\common_runtime\gpu\gpu_device.cc:940] Found device 0 with properties:
name: GeForce GTX 960
major: 5 minor: 2 memoryClockRate (GHz) 1.1775
pciBusID 0000:01:00.0
Total memory: 2.00GiB
Free memory: 1.64GiB
2017-07-07 21:05:46.063336: I c:\tf_jenkins\home\workspace\release-win\m\windows-gpu\py\35\tensorflow\core\common_runtime\gpu\gpu_device.cc:961] DMA: 0
2017-07-07 21:05:46.063695: I c:\tf_jenkins\home\workspace\release-win\m\windows-gpu\py\35\tensorflow\core\common_runtime\gpu\gpu_device.cc:971] 0: Y
2017-07-07 21:05:46.064298: I c:\tf_jenkins\home\workspace\release-win\m\windows-gpu\py\35\tensorflow\core\common_runtime\gpu\gpu_device.cc:1030] Creating TensorFlow device (/gpu:0) -> (device: 0, name: GeForce GTX 960, pci bus id: 0000:01:00.0)
2017-07-07 21:05:51.416742: I c:\tf_jenkins\home\workspace\release-win\m\windows-gpu\py\35\tensorflow\core\common_runtime\gpu\gpu_device.cc:1030] Creating TensorFlow device (/gpu:0) -> (device: 0, name: GeForce GTX 960, pci bus id: 0000:01:00.0)
2017-07-07 21:05:55.509367: I c:\tf_jenkins\home\workspace\release-win\m\windows-gpu\py\35\tensorflow\core\common_runtime\gpu\gpu_device.cc:1030] Creating TensorFlow device (/gpu:0) -> (device: 0, name: GeForce GTX 960, pci bus id: 0000:01:00.0)
Die Optimierung hat begonnen...
Durchlauf 1/1000
2017-07-07 21:05:58.595881: W c:\tf_jenkins\home\workspace\release-win\m\windows-gpu\py\35\tensorflow\core\common_runtime\bfc_allocator.cc:217] Allocator (GPU_0_bfc) ran out of memory trying to allocate 1.08GiB. The caller indicates that this is not a failure, but may mean that there could be performance gains if more memory is available.
2017-07-07 21:05:58.645705: W c:\tf_jenkins\home\workspace\release-win\m\windows-gpu\py\35\tensorflow\core\common_runtime\bfc_allocator.cc:217] Allocator (GPU_0_bfc) ran out of memory trying to allocate 1.08GiB. The caller indicates that this is not a failure, but may mean that there could be performance gains if more memory is available.
2017-07-07 21:05:58.663314: W c:\tf_jenkins\home\workspace\release-win\m\windows-gpu\py\35\tensorflow\core\common_runtime\bfc_allocator.cc:217] Allocator (GPU_0_bfc) ran out of memory trying to allocate 1.08GiB. The caller indicates that this is not a failure, but may mean that there could be performance gains if more memory is available.
2017-07-07 21:05:58.751849: W c:\tf_jenkins\home\workspace\release-win\m\windows-gpu\py\35\tensorflow\core\common_runtime\bfc_allocator.cc:217] Allocator (GPU_0_bfc) ran out of memory trying to allocate 1.08GiB. The caller indicates that this is not a failure, but may mean that there could be performance gains if more memory is available.
Durchlauf 2/1000And it is much faster. Yet my whole system seems to become a little bit unresponsive. Can I increase speed more by building from source and including those optimizations?
anishathalye
commented
7 years ago Not sure. You'd probably just benefit most by getting a GPU with more memory.
On Jul 7, 2017, at 12:10 PM, Erfolgreich charismatisch notifications@github.com wrote:
I did some installing back and forth, now I get
2017-07-07 21:05:45.751030: W c:\tf_jenkins\home\workspace\release-win\m\windows-gpu\py\35\tensorflow\core\platform\cpu_feature_guard.cc:45] The TensorFlow library wasn't compiled to use SSE instructions, but these are available on your machine and could speed up CPU computations. 2017-07-07 21:05:45.751180: W c:\tf_jenkins\home\workspace\release-win\m\windows-gpu\py\35\tensorflow\core\platform\cpu_feature_guard.cc:45] The TensorFlow library wasn't compiled to use SSE2 instructions, but these are available on your machine and could speed up CPU computations. 2017-07-07 21:05:45.751423: W c:\tf_jenkins\home\workspace\release-win\m\windows-gpu\py\35\tensorflow\core\platform\cpu_feature_guard.cc:45] The TensorFlow library wasn't compiled to use SSE3 instructions, but these are available on your machine and could speed up CPU computations. 2017-07-07 21:05:45.751525: W c:\tf_jenkins\home\workspace\release-win\m\windows-gpu\py\35\tensorflow\core\platform\cpu_feature_guard.cc:45] The TensorFlow library wasn't compiled to use SSE4.1 instructions, but these are available on your machine and could speed up CPU computations. 2017-07-07 21:05:45.751642: W c:\tf_jenkins\home\workspace\release-win\m\windows-gpu\py\35\tensorflow\core\platform\cpu_feature_guard.cc:45] The TensorFlow library wasn't compiled to use SSE4.2 instructions, but these are available on your machine and could speed up CPU computations. 2017-07-07 21:05:45.751769: W c:\tf_jenkins\home\workspace\release-win\m\windows-gpu\py\35\tensorflow\core\platform\cpu_feature_guard.cc:45] The TensorFlow library wasn't compiled to use AVX instructions, but these are available on your machine and could speed up CPU computations. 2017-07-07 21:05:46.063231: I c:\tf_jenkins\home\workspace\release-win\m\windows-gpu\py\35\tensorflow\core\common_runtime\gpu\gpu_device.cc:940] Found device 0 with properties: name: GeForce GTX 960 major: 5 minor: 2 memoryClockRate (GHz) 1.1775 pciBusID 0000:01:00.0 Total memory: 2.00GiB Free memory: 1.64GiB 2017-07-07 21:05:46.063336: I c:\tf_jenkins\home\workspace\release-win\m\windows-gpu\py\35\tensorflow\core\common_runtime\gpu\gpu_device.cc:961] DMA: 0 2017-07-07 21:05:46.063695: I c:\tf_jenkins\home\workspace\release-win\m\windows-gpu\py\35\tensorflow\core\common_runtime\gpu\gpu_device.cc:971] 0: Y 2017-07-07 21:05:46.064298: I c:\tf_jenkins\home\workspace\release-win\m\windows-gpu\py\35\tensorflow\core\common_runtime\gpu\gpu_device.cc:1030] Creating TensorFlow device (/gpu:0) -> (device: 0, name: GeForce GTX 960, pci bus id: 0000:01:00.0) 2017-07-07 21:05:51.416742: I c:\tf_jenkins\home\workspace\release-win\m\windows-gpu\py\35\tensorflow\core\common_runtime\gpu\gpu_device.cc:1030] Creating TensorFlow device (/gpu:0) -> (device: 0, name: GeForce GTX 960, pci bus id: 0000:01:00.0) 2017-07-07 21:05:55.509367: I c:\tf_jenkins\home\workspace\release-win\m\windows-gpu\py\35\tensorflow\core\common_runtime\gpu\gpu_device.cc:1030] Creating TensorFlow device (/gpu:0) -> (device: 0, name: GeForce GTX 960, pci bus id: 0000:01:00.0) Die Optimierung hat begonnen... Durchlauf 1/1000 2017-07-07 21:05:58.595881: W c:\tf_jenkins\home\workspace\release-win\m\windows-gpu\py\35\tensorflow\core\common_runtime\bfc_allocator.cc:217] Allocator (GPU_0_bfc) ran out of memory trying to allocate 1.08GiB. The caller indicates that this is not a failure, but may mean that there could be performance gains if more memory is available. 2017-07-07 21:05:58.645705: W c:\tf_jenkins\home\workspace\release-win\m\windows-gpu\py\35\tensorflow\core\common_runtime\bfc_allocator.cc:217] Allocator (GPU_0_bfc) ran out of memory trying to allocate 1.08GiB. The caller indicates that this is not a failure, but may mean that there could be performance gains if more memory is available. 2017-07-07 21:05:58.663314: W c:\tf_jenkins\home\workspace\release-win\m\windows-gpu\py\35\tensorflow\core\common_runtime\bfc_allocator.cc:217] Allocator (GPU_0_bfc) ran out of memory trying to allocate 1.08GiB. The caller indicates that this is not a failure, but may mean that there could be performance gains if more memory is available. 2017-07-07 21:05:58.751849: W c:\tf_jenkins\home\workspace\release-win\m\windows-gpu\py\35\tensorflow\core\common_runtime\bfc_allocator.cc:217] Allocator (GPU_0_bfc) ran out of memory trying to allocate 1.08GiB. The caller indicates that this is not a failure, but may mean that there could be performance gains if more memory is available. Durchlauf 2/1000 And it is much faster. Yet my whole system seems to become a little bit unresponsive. Can I increase speed more by building from source and including those optimizations?
— You are receiving this because you commented. Reply to this email directly, view it on GitHub, or mute the thread.
anishathalye
commented
6 years ago Also, for reference, it takes something like 60 seconds on my Maxwell Titan X.
Going to close this issue for now, feel free to reopen if anything new comes up.
I am using a 448 x 335 pixel image for the style image. I am using a 295 x 525 pixel image for the content image.
I have a Intel i7-2600 @ 3.40Ghz, 4 cores I have a GeForce GTX 960
It takes about two hours to finish with 1000 iterations.
Is that normal? Is it even using cuda? I get messages about tensorflow not being optimized for MMX etc, but then I thought that this is for cpu rendering, yet I am GPU rendering, so it doesn't matter or does it?