qbx2

def k_subset(s, k):
    if k == len(s):
        return (tuple([(x,) for x in s]),)
    k_subs = []
    for i in range(len(s)):
        partials = k_subset(s[:i] + s[i + 1:], k)
        for partial in partials:
            for p in range(len(partial)):
                k_subs.append(partial[:p] + (partial[p] + (s[i],),) + partial[p + 1:])
    return k_subs

qbx2

import gzip
import tensorflow as tf
import struct
import numpy as np
import random

def one_hot_encode(i):
	ret = [0] * 10
	ret[i] = 1
	return ret

qbx2

import gzip
import tensorflow as tf
import struct
import numpy as np
import random

def one_hot_encode(i):
	ret = [0] * 10
	ret[i] = 1
	return ret

qbx2

from PIL import Image
from struct import unpack
from sys import argv

def rgb565torgb888(c):
	return ((c&0xf800) >> 8, (c&0x07e0) >> 3, (c&0x001f) << 3)

if len(argv) < 2:
	print('Usage: {} <filename>'.format(argv[0]))
	exit()

qbx2

import tensorflow as tf
import random
import copy
import sys

MINIBATCH_SIZE = 1000

def gen_model():
    cell = tf.contrib.rnn.BasicRNNCell(128)

qbx2

import tensorflow as tf
import random


W = tf.get_variable('W', shape=[1, 4])
ph_y = tf.placeholder(tf.int32, [1])  # labels

prob = tf.nn.softmax(W)

loss = tf.reduce_mean(

qbx2

import tensorflow as tf
import random


W = tf.get_variable('W', shape=[1, 4])
ph_y = tf.placeholder(tf.int32, [1])  # labels

prob = tf.nn.softmax(W)

loss = tf.reduce_mean(

qbx2

-----BEGIN CERTIFICATE-----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qbx2

from collections import defaultdict
metrics = defaultdict(float)
num_metrics = 0

# training loop
for xs, ys in training_dataloader:
  batch_size = xs.size(0)
  loss = criterion(...)
  metrics['loss'] += float(loss) * batch_size
  num_metrics += batch_size

qbx2

import torch


def calculate_loss(x, y):
    clip_range = 100.
    clipped = y + (x - y).clamp(-clip_range, clip_range)
    l_vf = 0.5 * torch.max((clipped - y) ** 2, (x - y) ** 2).mean()
    return l_vf