Narco pelicula, Robert Risko, Linear atrophoderma of Moulin

This commit is contained in:
nik gaffney 2021-02-09 18:19:14 +01:00
parent b976f2526b
commit e151fffd17
5 changed files with 121 additions and 9 deletions

View file

@ -7,3 +7,65 @@ various loose ends of the generative card making approach
setup script for linode GPU instances
- from debian 10 starting point ⟶ various GAN stuff
* lightweight & styleGAN
backups are on posthoc
* * styleGAN2-ADA
ongoing experiments with styleGAN2
current staging on Linode GPU instances (aka grapheme )
#+BEGIN_SRC shell :dir /ssh:grapheme:. :results raw :wrap SRC text :async
ps ax | grep train.py
#+end_src
#+RESULTS:
** generate and mix
#+BEGIN_SRC shell :dir /ssh:grapheme:/root/stylegan2-ada-pytorch/ :results raw replace :wrap SRC text :async
python3 style_mixing.py \
--outdir=/root/results/flickr6/generated/001600 \
--network=/root/results/flickr6/00003-512x512-auto1-resumecustom/network-snapshot-001600.pkl \
--rows=6000-6020 \
--cols=600-620
#+end_src
#+RESULTS:
#+BEGIN_SRC shell :dir /ssh:grapheme:/root/stylegan2-ada-pytorch/ :results raw replace :wrap SRC text :async
python3 generate.py \
--outdir=/root/results/flickr6/generated/001600 \
--network=/root/results/flickr6/00003-512x512-auto1-resumecustom/network-snapshot-001600.pkl \
--seeds=3300-3900
#+end_src
#+RESULTS:
** projection
image projection to explore latent space
--target=https://live.staticflickr.com/65535/50698595796_8accf83473_k_d.jpg \
#+BEGIN_SRC shell :dir /ssh:grapheme:/root/stylegan2-ada-pytorch/ :results raw replace :wrap SRC text :async
python3 projector.py --outdir=/root/results/flickr6/generated/101405 \
--target=/root/data/_transfer_test/50698595796_8accf83473_k.jpg \
--network=/root/results/flickr6/00003-512x512-auto1-resumecustom/network-snapshot-001440.pkl \
--num-steps=5000
#+END_SRC
#+RESULTS:
: 5efe249760c6a25a5188c67816b13ed4
#+BEGIN_SRC shell :results raw replace :wrap SRC text :async
rsync -av grapheme:results/ \
/Users/zzk/code/foam-repo/superject/cards/neural/styleGAN2-ADA/results
#+END_SRC

View file

@ -8,6 +8,7 @@
# - copy ssh keys via -> ssh-copy-id -i ~/.ssh/id_rsa root@grapheme
# - copy scripts -> scp *.sh grapheme:
# - sshfs grapheme:/tmp/neural /Volumes/
# -
# on grapheme
@ -32,7 +33,7 @@ _EOL_
apt-get update
apt-get upgrade -y
apt-get install nvidia-driver firmware-misc-nonfree python3-pip emacs-nox nvidia-cuda-dev nvtop screen rsync -y &> /dev/null | echo 'Installing packaged gunk...'
apt-get install git nvidia-driver firmware-misc-nonfree python3-pip emacs-nox nvidia-cuda-dev nvidia-cuda-toolkit nvtop screen rsync -y &> /dev/null | echo 'Installing packaged gunk...'
apt purge linux-image-4.19.0-13-amd64
@ -47,13 +48,17 @@ echo options nouveau modeset=0 >> /etc/modprobe.d/blacklist-nvidia-nouveau.conf
echo "partial attention..."
pip3 install deep-daze &> /dev/null | echo 'Installing deep-daze...'
git clone https://github.com/NVlabs/stylegan2-ada-pytorch
pip3 install big-sleep &> /dev/null | echo 'Installing big-sleep...'
pip3 install torch==1.7.1 pycuda psutil Pillow scipy click requests tqdm pyspng ninja tensorboard imageio imageio-ffmpeg==0.4.3
#pip3 install stylegan2_pytorch &> /dev/null | echo 'Installing styleGAN2...'
#pip3 install lightweight-gan &> /dev/null | echo 'Installing lightweight GAN...'
#pip3 install deep-daze &> /dev/null | echo 'Installing deep-daze...'
#pip3 install big-sleep &> /dev/null | echo 'Installing big-sleep...'
mkdir -p /tmp/neural
echo "you are now asleep."
# make new users if reqd?

View file

@ -0,0 +1,31 @@
#!/bin/bash
# a simple script run stylegan2_pytorch on a linode GPU instance
# PLACE=/tmp/neural/
# mkdir -p $PLACE
# cd $PLACE
echo "using stlyeGAN2..."
phrases=("An energetic nomad on a long hike with wooden staff a green leaf and a small dog." "At the beginning of a magical quest for wisdom an animus of a conman hide beneath the table." "A charismatic female figure meditates on the book of flesh in isolation, oval and white." "Fertility of nature in small green leaves, adam's apple, a pyramid and a baby eagle's wing." "The alchemical gold adorns the throne in a powerful house rooted in matter.")
echo "cosmic entanglement..."
cd /root
# transfer learning using ffhq model for sytleGAN2
stylegan2_pytorch --name flickr3 \
--data data/1024x1024 \
--models_dir models/ffhq-foam \
--results_dir results/flickr3 \
--multi-gpus \
--batch-size 16 --network-capacity 64 \
--aug-prob 0.25 --attn-layers 1
# note in big-sleep.py -> torch,cuda.set_device(1)
# and/or CUDA_VISIBLE_DEVICES=1,2 python myscript.py
# ffmpeg -r 24 -f image2 -s 1920x1080 -i The_Grand_Evocateur_of_intensity.%d.png -vcodec libx264 -crf 25 -pix_fmt yuv420p The_Grand_Evocateur_of_intensity_m3.mp4

View file

@ -2,8 +2,9 @@
# via https://github.com/huggingface/pytorch-pretrained-BigGAN
import torch
from pytorch_pretrained_biggan import (BigGAN, one_hot_from_names, truncated_noise_sample,
save_as_images, display_in_terminal)
from pytorch_pretrained_biggan import (BigGAN, one_hot_from_names, one_hot_from_int, truncated_noise_sample, save_as_images, display_in_terminal)
from PIL import Image
# OPTIONAL: if you want to have more information on what's happening, activate the logger as follows
import logging
@ -13,11 +14,15 @@ logging.basicConfig(level=logging.INFO)
model = BigGAN.from_pretrained('biggan-deep-512')
# Prepare a input
truncation = 0.4
class_vector = one_hot_from_names(['soap bubble', 'coffee', 'mushroom'], batch_size=3)
noise_vector = truncated_noise_sample(truncation=truncation, batch_size=3)
truncation = 0.001
class_vector = one_hot_from_int([22, 65, 555, 333], batch_size=4)
#class_vector = one_hot_from_names(['soap bubble', 'coffee', 'mushroom', 'daisy'], batch_size=4)
noise_vector = truncated_noise_sample(truncation=truncation, batch_size=4)
# All in tensors
# noise_vector = torch.from_numpy(noise_vector)
# class_vector = torch.from_numpy(class_vector)
noise_vector = torch.from_numpy(noise_vector)
class_vector = torch.from_numpy(class_vector)

View file

@ -741,6 +741,15 @@ improve the instuctions
* and 2021-02-09
via adam
»
Superject is an experiment. It is an experiment that is bold and humble at the same time. It mixes philosophy and design but is neither one of them. Philosophers would certainly object that Superject is not philosophy, while designers would likely balk at its many metaphysical conceits. It weaves philosophy and non-philosophy, or perhaps better: design and non-design; but then again, this experiment is not particularly interested in knowledge hierarchies. Alternatively, we could say that mixing philosophy and design creates an emergent species, something altogether different and unrecognizable, but then what of its components? Maybe its better not to worry about the dialectical essence of Superject and simply focus on its appearance: its a game.
* other games, gameplay and/or narrative tangents
Looking at various cards, decks and games in an attempt understand our possibility space, focus and intent, through the lens of Whitehead's propositions....