adi-iitd / ai-art Goto Github PK

Python noobie here. I am running Style_Transfer.py on a macbook with the following specs:

Is it normal for the script to run more than 5 minutes or am I doing something wrong? I shut the process off at some point because my laptop fans were going crazy.

I tried executing CycleGAN version for the style transfer using ! python CycleGAN-PL.py in gcolab.

But its showing following error

Traceback (most recent call last):
File "CycleGAN-PL.py", line 2, in
from Imports import *
File "/content/drive/Shareddrives/ML/AI-Art/src/CycleGAN/Imports.py", line 3, in
import os, wget, zipfile, shutil
ModuleNotFoundError: No module named 'wget'

In the neural style transfer models, as you increase the resolution of images that you are working with (e.g. the content, style and output images), what are the impacts on results and how does the computation scale?

Is it practical to use such models for HD image tasks at their current stage?

So how would I install this AI-art on my machine and run it? The src just has the python source code, but no requirements.txt, or even what commands to run.

It should be val dataloader at https://github.com/Adi-iitd/AI-Art/blob/master/CycleGAN.py#L788?

Getting this error... Maybe a problem introduced by a new torch version?

File "C:\Users\dgilk\anaconda3\envs\aiart\lib\site-packages\torch\distributed\rendezvous.py", line 190, in _env_rendezvous_handler
File "C:\Users\dgilk\anaconda3\envs\aiart\lib\site-packages\pytorch_lightning\accelerators\accelerator.py", line 83, in pre_dispatch
RuntimeError self.init_ddp_connection(self.global_rank, self.world_size): store = TCPStore(master_addr, master_port, world_size, start_daemon, timeout)
Only one usage of each socket address (protocol/network address/port) is normally permitted.
self.training_type_plugin.pre_dispatch()
File "C:\Users\dgilk\anaconda3\envs\aiart\lib\site-packages\pytorch_lightning\plugins\training_type\ddp.py", line 241, in init_ddp_connection
RuntimeError: Only one usage of each socket address (protocol/network address/port) is normally permitted.

File "C:\Users\dgilk\anaconda3\envs\aiart\lib\site-packages\pytorch_lightning\plugins\training_type\ddp.py", line 258, in pre_dispatch
torch_distrib.init_process_group(self.torch_distributed_backend, rank=global_rank, world_size=world_size)
File "C:\Users\dgilk\anaconda3\envs\aiart\lib\site-packages\torch\distributed\distributed_c10d.py", line 500, in init_process_group
self.init_ddp_connection(self.global_rank, self.world_size) store, rank, world_size = next(rendezvous_iterator)

File "C:\Users\dgilk\anaconda3\envs\aiart\lib\site-packages\torch\distributed\rendezvous.py", line 190, in _env_rendezvous_handler
File "C:\Users\dgilk\anaconda3\envs\aiart\lib\site-packages\pytorch_lightning\plugins\training_type\ddp.py", line 241, in init_ddp_connection
store = TCPStore(master_addr, master_port, world_size, start_daemon, timeout)torch_distrib.init_process_group(self.torch_distributed_backend, rank=global_rank, world_size=world_size)

RuntimeError: File "C:\Users\dgilk\anaconda3\envs\aiart\lib\site-packages\torch\distributed\distributed_c10d.py", line 500, in init_process_group
Only one usage of each socket address (protocol/network address/port) is normally permitted.
store, rank, world_size = next(rendezvous_iterator)
File "C:\Users\dgilk\anaconda3\envs\aiart\lib\site-packages\torch\distributed\rendezvous.py", line 190, in _env_rendezvous_handler
store = TCPStore(master_addr, master_port, world_size, start_daemon, timeout)
RuntimeError: Only one usage of each socket address (protocol/network address/port) is normally permitted.

Tried running in Google Colab and on GPUs on AWS

Same error both times:

9 frames

/usr/local/lib/python3.7/dist-packages/imageio/core/request.py in _parse_uri(self, uri)
271 # Reading: check that the file exists (but is allowed a dir)
272 if not os.path.exists(fn):
--> 273 raise FileNotFoundError("No such file: '%s'" % fn)
274 else:
275 # Writing: check that the directory to write to does exist

FileNotFoundError: No such file: '/content/AI-Art/Dataset/StyleTransfer/Content.jpg'

When you're training the gan, in the train function, it does all the updating and the loss goes down. Suppose I wanted to interrupt and resume training or suppose I wanted to generate test time images, how do I do that?