standy66

from agentnet.utils.persistence import save,load
def build_pretrained_spaceinvaders(observation_reshape):
    """ a smaller version of DQN pre-trained for ~2 hours on GPU """
    assert tuple(observation_reshape.output_shape) == (None, 3, 105, 80)
    
    #main neural network body
    conv0 = Conv2DLayer(observation_reshape,16,filter_size=(8,8),stride=(4,4),name='conv0')
    conv1 = Conv2DLayer(conv0,32,filter_size=(4,4),stride=(2,2),name='conv1')
    dense0 = DenseLayer(conv1,256,name='dense',nonlinearity=lasagne.nonlinearities.tanh)
    

standy66

with g.as_default():
    print([v.name for v in tf.all_variables()])
    W = [v for v in tf.all_variables() if v.name == "Conv/weights:0"][0]
    W_val = sess.run(W)
    fig, axes = plt.subplots(5, 5, figsize=(15, 15))
    for i in range(5):
        for j in range(5):
            axes[i, j].imshow(W_val[:, :, 0, i * 5 + j], interpolation='nearest', cmap='gray')
    plt.show()

standy66

#! /usr/bin/env python3

import sys

def read_matrix(path):
    A = []
    with open(path) as f:
        content = f.readlines()
        for line in content:
            idx1, idx2, other = line.split()

standy66

#! /usr/bin/env python3

import sys
import random

width = int(sys.argv[1])
height = int(sys.argv[2])

for i in range(0, width):
    for j in range(0, height):

standy66

channel = new FileInputStream(dbFilePath).getChannel();
            ByteBuffer buffer = channel.map(FileChannel.MapMode.READ_ONLY, 0, channel.size());
            try {
                while (buffer.remaining() > 0) {
                    int keySize = buffer.getInt();
                    byte[] key = new byte[keySize];
                    buffer.get(key);
                    int valueSize = buffer.getInt();
                    byte[] value = new byte[valueSize];
                    buffer.get(value);