shawwn

a conversation about the einops lib and why it's hard to optimize ML:

<PapuaHardyNet> the solution is obviously jax acc to shawn

which is why i'm thinking to also build a jax equivalent and compare all three: pytorch, pytorch + einops, jax. Let's see if I have the bandwidth to do so

<nshepperd2> einops is really cool imo

FeepingCreature

diff --git a/src/generate_unconditional_samples.py b/src/generate_unconditional_samples.py
index d9e2131..366c411 100755
--- a/src/generate_unconditional_samples.py
+++ b/src/generate_unconditional_samples.py
@@ -9,7 +9,7 @@ import tensorflow as tf
 import model, sample, encoder
 
 def sample_model(
-    model_name='117M',
+    model_name='1558N',

Miladiouss

import torchvision
import torchvision.transforms as transforms
from torchvision.datasets import CIFAR10
from torch.utils.data import Dataset, DataLoader
import numpy as np

# Transformations
RC = transforms.RandomCrop(32, padding=4)
RHF = transforms.RandomHorizontalFlip()
RVF = transforms.RandomVerticalFlip()

fasiha

import numpy as np
import numpy.fft as fft


def stft(x, Nwin, Nfft=None):
    """
    Short-time Fourier transform: convert a 1D vector to a 2D array

    The short-time Fourier transform (STFT) breaks a long vector into disjoint
    chunks (no overlap) and runs an FFT (Fast Fourier Transform) on each chunk.

kylemcdonald

import PIL.Image
from cStringIO import StringIO
import IPython.display
import numpy as np
def showarray(a, fmt='png'):
    a = np.uint8(a)
    f = StringIO()
    PIL.Image.fromarray(a).save(f, fmt)
    IPython.display.display(IPython.display.Image(data=f.getvalue()))

kamiller

## {{{ http://code.activestate.com/recipes/577518/ (r4)
#!/usr/bin/env python
# -*- coding: utf-8 -*-
"""
This is a pure Python implementation of the [rsync algorithm](TM96).

[TM96] Andrew Tridgell and Paul Mackerras. The rsync algorithm.
Technical Report TR-CS-96-05, Canberra 0200 ACT, Australia, 1996.
http://samba.anu.edu.au/rsync/.