search

svip-lab / impersonator

PyTorch implementation of our ICCV 2019 paper: Liquid Warping GAN: A Unified Framework for Human Motion Imitation, Appearance Transfer and Novel View Synthesis
https://svip-lab.github.io/project/impersonator
Other
1.72k stars 317 forks source link

Sorry to bother you. Can this project run on cuda9.0? thx! #35

Closed OldChi closed 3 years ago

OldChi commented 4 years ago

i run "python demo_imitator.py --gpu_ids 1" on cuda9.0, torch1.1, torchvision0.3 and encountered an error:

import neural_renderer.cuda.load_textures as load_textures_cuda ImportError: /home/huangxinkai/env/impersonator/lib/python3.6/site-packages/neural_renderer-1.1.3-py3.6-linux-x86_64.egg/neural_renderer/cuda/load_textures.cpython-36m-x86_64-linux-gnu.so: undefined symbol: _ZN6caffe26detail37_typeMetaDataInstance_preallocated_29E

could you please give me a hand and solve this?thx!

taotaoyuhust commented 4 years ago

i run it on cuda9.2 ,windows 10, torch1.2, torchvision 0.4.1

piaozhx commented 4 years ago

@OldChi I think cuda9.0 is ok for this project. You should make sure your pytorch cuda version(usually is selected on the pytorch website when you download pytorch) and local cuda version(could be found in /usr/local/cuda/version.txt) are same.

OldChi commented 4 years ago

@OldChi I think cuda9.0 is ok for this project. You should make sure your pytorch cuda version(usually is selected on the pytorch website when you download pytorch) and local cuda version(could be found in /usr/local/cuda/version.txt) are same.

thank you! my cuda version is 9.0.176 i downloaded the whl file of torchvision0.3 and torch1.1(cuda 9.0 build) on the pytorch website and installed it, but the same error appeared

piaozhx commented 4 years ago

remove all files in /home/huangxinkai/env/impersonator/lib/python3.6/site-packages/neural_renderer-1.1.3-py3.6-linux-x86_64.egg and other compiled files in thirdparty/neural_renderer (recommend replacing it with a clean version), then install neural_render again.

OldChi commented 4 years ago

remove all files in /home/huangxinkai/env/impersonator/lib/python3.6/site-packages/neural_renderer-1.1.3-py3.6-linux-x86_64.egg and other compiled files in thirdparty/neural_renderer (recommend replacing it with a clean version), then install neural_render again.

thank you very much! now i'm running the imitator successfully. you guys have really done a cool job. bravo!!

here's a little question: i want to run some customs examples using imitator. and i wonder if there are some constraints on my reference pictures, for example, the image size? the character's position in the image? the character's background? how could i modify the reference pictures or the codes so that i could get good results as you guys did?

thanks again

OldChi commented 4 years ago

remove all files in /home/huangxinkai/env/impersonator/lib/python3.6/site-packages/neural_renderer-1.1.3-py3.6-linux-x86_64.egg and other compiled files in thirdparty/neural_renderer (recommend replacing it with a clean version), then install neural_render again.

excuse me. i've converted my own video file into 5445443 images and i used run_imitator.py to swap the man inside my video. but an error occured. what should i do to run it successfully.

Traceback (most recent call last): File "run_imitator.py", line 225, in adaptive_personalize(test_opt, imitator, visualizer) File "run_imitator.py", line 203, in adaptive_personalize imitator.personalize(opt.src_path, visualizer=None) File "/home/huangxinkai/env/impersonator/lib/python3.6/site-packages/torch/autograd/grad_mode.py", line 43, in decorate_no_grad return func(*args, *kwargs) File "/home/huangxinkai/project/impersonator/models/imitator.py", line 125, in personalize src_info['bg'] = self.bgnet(img, masks=body_mask, only_x=True) File "/home/huangxinkai/env/impersonator/lib/python3.6/site-packages/torch/nn/modules/module.py", line 493, in call result = self.forward(input, *kwargs) File "/home/huangxinkai/project/impersonator/networks/inpaintor.py", line 180, in forward masked_imgs = imgs (1 - masks) + masks RuntimeError: The size of tensor a (256) must match the size of tensor b (257) at non-singleton dimension 3

If you suspect this is an IPython bug, please report it at: https://github.com/ipython/ipython/issues or send an email to the mailing list at ipython-dev@python.org

You can print a more detailed traceback right now with "%tb", or use "%debug" to interactively debug it.

Extra-detailed tracebacks for bug-reporting purposes can be enabled via: %config Application.verbose_crash=True

AKclown commented 3 years ago

thirdparty/neural_renderer

I did not find the thirdparty/neural_renderer directory, pip uninstall neural_renderer_pytorch and then re-execute pip install neural_renderer_pytorch, sh scripts/run_car.sh and still report errors. What should I do?