RTX3080 / CUDA 11.0 support

[mavanmanen]

Really wanted to try this out but sadly won't be able to because only CUDA 10.x is supported. Has anyone found a way to get this working with CUDA 11?

[simon-rob]

@mavanmanen I got it working on Ubuntu with a RTX 3090 CUDA 11.1 via the following

Assuming you have successfully compiled TensorFlow 2.x locally

change all: (in .py files)

import tensorflow as tf

To:

import tensorflow.compat.v1 as tf
tf.disable_v2_behavior()

To convert pkl files to tensorflow 2.x run from the StyeFlow folder:

sed -i -e 's/import tensorflow as tf/import tensorflow.compat.v1 as tf\ntf.disable_v2_behavior()/g' .stylegan2-cache/62233eb618af2672cd321cfb46c72c6a_http___d36zk2xti64re0.cloudfront.net_stylegan2_networks_stylegan2-ffhq-config-f.pkl

sed -i -e 's/Network architectures used in the StyleGAN2 paper./tyleGAN2 paper./g' .stylegan2-cache/62233eb618af2672cd321cfb46c72c6a_http___d36zk2xti64re0.cloudfront.net_stylegan2_networks_stylegan2-ffhq-config-f.pkl

Change: line 133 of nnlib/tflib/custom_ops.py

compile_opts += ' --compiler-options \'-fPIC -D_GLIBCXX_USE_CXX11_ABI=0

To:

compile_opts += ' --compiler-options \'-fPIC -D_GLIBCXX_USE_CXX11_ABI=1

comment out the following in: dnnlib/tflib/tfutil.py

#import tensorflow.contrib   # requires TensorFlow 1.x!
#tf.contrib = tensorflow.contrib

`

[mavanmanen]

@simon-rob That's great, any idea if this will work on windows too?

[justinjohn0306]

Yes it does

On Sat, Jan 16, 2021, 7:48 PM Mitchell van Manen notifications@github.com wrote:

@simon-rob https://github.com/simon-rob That's great, any idea if this will work on windows too?

— You are receiving this because you are subscribed to this thread. Reply to this email directly, view it on GitHub https://github.com/RameenAbdal/StyleFlow/issues/34#issuecomment-761570892, or unsubscribe https://github.com/notifications/unsubscribe-auth/AIDVKQ4RIXPBZG4XDKIE7Q3S2GN23ANCNFSM4WA4TTWA .

[mavanmanen]

Sadly I am now stuck with the following error:

C:/Users/mavan/Downloads/StyleFlow-Windows-10-master/dnnlib/tflib/ops/fused_bias_act.cu(174): error: expected an expression

C:/Users/mavan/Downloads/StyleFlow-Windows-10-master/dnnlib/tflib/ops/fused_bias_act.cu(174): error: no instance of constructor "tensorflow::register_op::OpDefBuilderWrapper::OpDefBuilderWrapper" matches the argument list
            argument types are: (const char [13], __nv_bool)

C:/Users/mavan/Downloads/StyleFlow-Windows-10-master/dnnlib/tflib/ops/fused_bias_act.cu(175): error: expected an expression

C:/Users/mavan/Downloads/StyleFlow-Windows-10-master/dnnlib/tflib/ops/fused_bias_act.cu(175): error: expected an expression

C:/Users/mavan/Downloads/StyleFlow-Windows-10-master/dnnlib/tflib/ops/fused_bias_act.cu(175): error: expected a type specifier

C:/Users/mavan/Downloads/StyleFlow-Windows-10-master/dnnlib/tflib/ops/fused_bias_act.cu(175): error: expected an expression

C:/Users/mavan/Downloads/StyleFlow-Windows-10-master/dnnlib/tflib/ops/fused_bias_act.cu(176): error: expected an expression

C:/Users/mavan/Downloads/StyleFlow-Windows-10-master/dnnlib/tflib/ops/fused_bias_act.cu(176): error: expected an expression

C:/Users/mavan/Downloads/StyleFlow-Windows-10-master/dnnlib/tflib/ops/fused_bias_act.cu(176): error: expected a type specifier

C:/Users/mavan/Downloads/StyleFlow-Windows-10-master/dnnlib/tflib/ops/fused_bias_act.cu(176): error: expected an expression

10 errors detected in the compilation of "C:/Users/mavan/Downloads/StyleFlow-Windows-10-master/dnnlib/tflib/ops/fused_bias_act.cu".
_pywrap_tensorflow_internal.lib
fused_bias_act.cu

[JakeHeintz]

Having the same error as @mavanmanen

[dmalyavin]

@simon-rob would you or anyone else who managed to get it to work on a 3090 be able to put the sequence in which you setup your environment? Thank you!

[frosty3907]

Also interested in this.

[fungtion]

@dmalyavin you can try to replace tf stylegan2 with pytorch stylegan2 .

[Minabsapi]

@mavanmanen I got it working on Ubuntu with a RTX 3090 CUDA 11.1 via the following

Assuming you have successfully compiled TensorFlow 2.x locally

change all: (in .py files)

import tensorflow as tf

To:

import tensorflow.compat.v1 as tf
tf.disable_v2_behavior()

To convert pkl files to tensorflow 2.x run from the StyeFlow folder:

sed -i -e 's/import tensorflow as tf/import tensorflow.compat.v1 as tf\ntf.disable_v2_behavior()/g' .stylegan2-cache/62233eb618af2672cd321cfb46c72c6a_http___d36zk2xti64re0.cloudfront.net_stylegan2_networks_stylegan2-ffhq-config-f.pkl

sed -i -e 's/Network architectures used in the StyleGAN2 paper./tyleGAN2 paper./g' .stylegan2-cache/62233eb618af2672cd321cfb46c72c6a_http___d36zk2xti64re0.cloudfront.net_stylegan2_networks_stylegan2-ffhq-config-f.pkl

Change: line 133 of nnlib/tflib/custom_ops.py

compile_opts += ' --compiler-options \'-fPIC -D_GLIBCXX_USE_CXX11_ABI=0

To:

compile_opts += ' --compiler-options \'-fPIC -D_GLIBCXX_USE_CXX11_ABI=1

comment out the following in: dnnlib/tflib/tfutil.py

#import tensorflow.contrib   # requires TensorFlow 1.x!
#tf.contrib = tensorflow.contrib

`

This trick didn't work for me on CUDA 11.4

2021-08-07 23:42:29.450490: I tensorflow/stream_executor/platform/default/dso_loader.cc:53] Successfully opened dynamic library libcudart.so.11.0
WARNING:tensorflow:From /home/minabsapi/anaconda3/envs/StyleFlow/lib/python3.7/site-packages/tensorflow/python/compat/v2_compat.py:96: disable_resource_variables (from tensorflow.python.ops.variable_scope) is deprecated and will be removed in a future version.
Instructions for updating:
non-resource variables are not supported in the long term
----------------- Options ---------------
                batchSize: 1                             
          checkpoints_dir: ./checkpoints                 
                 dataroot: ./data/datasetX               
                  gpu_ids: 0                             
     max_result_snapshots: 30                            
                    model: xxxx                          
                     name: XXXX                          
              network_pkl: gdrive:networks/stylegan2-ffhq-config-f.pkl
            only_for_test: ...                           
                    phase: test                          
----------------- End -------------------
Loading networks from "gdrive:networks/stylegan2-ffhq-config-f.pkl"...
Downloading http://d36zk2xti64re0.cloudfront.net/stylegan2/networks/stylegan2-ffhq-config-f.pkl ... done
2021-08-07 23:46:33.377141: I tensorflow/stream_executor/platform/default/dso_loader.cc:53] Successfully opened dynamic library libcuda.so.1
2021-08-07 23:46:33.473857: I tensorflow/stream_executor/cuda/cuda_gpu_executor.cc:937] successful NUMA node read from SysFS had negative value (-1), but there must be at least one NUMA node, so returning NUMA node zero
2021-08-07 23:46:33.474461: I tensorflow/core/common_runtime/gpu/gpu_device.cc:1733] Found device 0 with properties: 
pciBusID: 0000:01:00.0 name: NVIDIA GeForce GT 730 computeCapability: 3.5
coreClock: 0.9015GHz coreCount: 2 deviceMemorySize: 1,95GiB deviceMemoryBandwidth: 13,41GiB/s
2021-08-07 23:46:33.474609: I tensorflow/stream_executor/platform/default/dso_loader.cc:53] Successfully opened dynamic library libcudart.so.11.0
2021-08-07 23:46:34.353215: I tensorflow/stream_executor/platform/default/dso_loader.cc:53] Successfully opened dynamic library libcublas.so.11
2021-08-07 23:46:34.353418: I tensorflow/stream_executor/platform/default/dso_loader.cc:53] Successfully opened dynamic library libcublasLt.so.11
2021-08-07 23:46:34.529012: I tensorflow/stream_executor/platform/default/dso_loader.cc:53] Successfully opened dynamic library libcufft.so.10
2021-08-07 23:46:34.625099: I tensorflow/stream_executor/platform/default/dso_loader.cc:53] Successfully opened dynamic library libcurand.so.10
2021-08-07 23:46:34.784685: I tensorflow/stream_executor/platform/default/dso_loader.cc:53] Successfully opened dynamic library libcusolver.so.11
2021-08-07 23:46:34.878989: I tensorflow/stream_executor/platform/default/dso_loader.cc:53] Successfully opened dynamic library libcusparse.so.11
2021-08-07 23:46:34.953630: I tensorflow/stream_executor/platform/default/dso_loader.cc:53] Successfully opened dynamic library libcudnn.so.8
2021-08-07 23:46:34.953909: I tensorflow/stream_executor/cuda/cuda_gpu_executor.cc:937] successful NUMA node read from SysFS had negative value (-1), but there must be at least one NUMA node, so returning NUMA node zero
2021-08-07 23:46:34.954475: I tensorflow/stream_executor/cuda/cuda_gpu_executor.cc:937] successful NUMA node read from SysFS had negative value (-1), but there must be at least one NUMA node, so returning NUMA node zero
2021-08-07 23:46:34.966209: I tensorflow/core/common_runtime/gpu/gpu_device.cc:1871] Adding visible gpu devices: 0
2021-08-07 23:46:34.981201: I tensorflow/stream_executor/platform/default/dso_loader.cc:53] Successfully opened dynamic library libcudart.so.11.0
2021-08-07 23:46:41.748710: I tensorflow/core/common_runtime/gpu/gpu_device.cc:1258] Device interconnect StreamExecutor with strength 1 edge matrix:
2021-08-07 23:46:41.748773: I tensorflow/core/common_runtime/gpu/gpu_device.cc:1264]      0 
2021-08-07 23:46:41.748796: I tensorflow/core/common_runtime/gpu/gpu_device.cc:1277] 0:   N 
2021-08-07 23:46:41.799672: I tensorflow/stream_executor/cuda/cuda_gpu_executor.cc:937] successful NUMA node read from SysFS had negative value (-1), but there must be at least one NUMA node, so returning NUMA node zero
2021-08-07 23:46:41.800308: I tensorflow/stream_executor/cuda/cuda_gpu_executor.cc:937] successful NUMA node read from SysFS had negative value (-1), but there must be at least one NUMA node, so returning NUMA node zero
2021-08-07 23:46:41.800839: I tensorflow/stream_executor/cuda/cuda_gpu_executor.cc:937] successful NUMA node read from SysFS had negative value (-1), but there must be at least one NUMA node, so returning NUMA node zero
2021-08-07 23:46:41.801291: I tensorflow/core/common_runtime/gpu/gpu_device.cc:1418] Created TensorFlow device (/job:localhost/replica:0/task:0/device:GPU:0 with 801 MB memory) -> physical GPU (device: 0, name: NVIDIA GeForce GT 730, pci bus id: 0000:01:00.0, compute capability: 3.5)
Traceback (most recent call last):
  File "main.py", line 367, in <module>
    ex = ExWindow(opt)
  File "main.py", line 42, in __init__
    self.EX = Ex(opt)
  File "main.py", line 72, in __init__
    self.init_deep_model(opt)
  File "main.py", line 107, in init_deep_model
    self.model = Build_model(self.opt)
  File "/home/minabsapi/Téléchargements/StyleFlow/utils.py", line 202, in __init__
    _G, _D, Gs = pretrained_networks.load_networks(network_pkl)
  File "/home/minabsapi/Téléchargements/StyleFlow/pretrained_networks.py", line 76, in load_networks
    G, D, Gs = pickle.load(stream, encoding='latin1')
  File "/home/minabsapi/Téléchargements/StyleFlow/dnnlib/tflib/network.py", line 299, in __setstate__
    self._init_graph()
  File "/home/minabsapi/Téléchargements/StyleFlow/dnnlib/tflib/network.py", line 156, in _init_graph
    out_expr = self._build_func(*self.input_templates, **build_kwargs)
  File "<string>", line 451, in G_synthesis_stylegan2
AttributeError: module 'tensorflow' has no attribute 'get_variable'

GNU/Linux Ubuntu 20.04 tensorflow 2.5.0

[alicedingyueming]

tectur

I have the same problem when using stylegan2 in TF.2