proger

# prompt: https://twitter.com/francoisfleuret/status/1783479122418716805

import os
os.environ['TORCH_LOGS'] = 'output_code' # shows all the bmms
import torch
torch.set_float32_matmul_precision('high')

N, T, D, U, C = 3, 128, 5, 32, 32 # batch, time, heads, head_dim, dim
S = T
A = torch.randn(N, T, D, U) / U**0.5

francois-rozet

#!/usr/bin/env python

import math
import matplotlib.pyplot as plt
import torch
import torch.nn as nn

from sklearn.datasets import make_moons
from torch import Tensor
from tqdm import tqdm

crowsonkb

import geoopt


def spherical_avg(p, w=None, tol=1e-6):
    sphere = geoopt.Sphere()
    if w is None:
        w = p.new_ones([p.shape[0]])
    assert p.ndim == 2 and w.ndim == 1 and len(p) == len(w)
    w = w / w.sum()
    p = sphere.projx(p)

crowsonkb

import abc

import torch
from torch import nn, optim


class Constraint(nn.Module, metaclass=abc.ABCMeta):
    def __init__(self, fn, damping):
        super().__init__()
        self.fn = fn

rolux

# git clone https://github.com/NVlabs/stylegan2
import os
import numpy as np
from scipy.interpolate import interp1d
from scipy.io import wavfile
import matplotlib.pyplot as plt
import PIL.Image
import moviepy.editor

import dnnlib

hiromu

#!/usr/bin/env python
# -*- coding: utf-8 -*-

import glob
import librosa
import numpy as np
import os
import sklearn.mixture
import sys

lpillmann

#!/usr/bin/env python
'''Crop an image to just the portions containing text.

Usage:

    ./crop_morphology.py path/to/image.jpg

This will place the cropped image in path/to/image.crop.png.

For details on the methodology, see

victor-shepardson

from contextlib import contextmanager
import numpy as np
import torch
from torch import Tensor, ByteTensor
import torch.nn.functional as F
from torch.autograd import Variable
import pycuda.driver
from pycuda.gl import graphics_map_flags
from glumpy import app, gloo, gl