marekgalovic

import os
from argparse import ArgumentParser
from PIL import Image
import json

import numpy as np
import tensorflow as tf
import tensorflow_hub as hub

from annoy import AnnoyIndex

marekgalovic

import os
import json
from PIL import Image
from io import BytesIO

from flask import Flask, jsonify, request, render_template, send_file

import numpy as np
import tensorflow as tf
import tensorflow_hub as hub

marekgalovic

import math
import random
import numpy as np
from multiprocessing import Pool, cpu_count

def prepend_start(route):
    return [0] + list(route)


def create_population(num_points, size):

marekgalovic

import random
import gym
import numpy as np
import tensorflow as tf

EPOCHS = 250
NUM_GAMES = 100
BATCH_SIZE = 128
MAX_UPDATES = 10
GAMMA = 0.99

marekgalovic

package main

import (
    "os";
    "os/signal";
    "syscall";
    "time";
    "context";
    "math";
    "bytes";

marekgalovic

# Q-network implementation using linear function aproximator.
# Estimates action-value given state and weights matrix Q(a|s,W)

import gym
import numpy as np
import matplotlib.pyplot as plt
import time

env = gym.make('FrozenLake-v0')

marekgalovic

import tensorflow as tf
import numpy as np

class ConvolutionalAttentionNLI(object):
    
    def __init__(self, embeddings_shape, target_classes=2, conv_filter_size=3, conv_projection_size=300, attention_output_size=200, comparison_output_size=100, learning_rate=0.05):
        self._embeddings_shape = embeddings_shape
        self._target_classes = target_classes
        self._conv_filter_size = conv_filter_size
        self._conv_projection_size = conv_projection_size

marekgalovic

import tensorflow as tf

class BiMPM(object):
    '''
        Bilateral Multi-Perspective Matching Model
        https://arxiv.org/pdf/1702.03814.pdf
    '''
    
    def __init__(self, state_size, l_perspectives, embeddings_shape):
        self._state_size = state_size

marekgalovic

# Decomposable attention model for NLI
# https://arxiv.org/pdf/1606.01933v1.pdf

dnli_graph = tf.Graph()
with dnli_graph.as_default():
    embedding_matrix = tf.Variable(tf.zeros([DICTIONARY_SIZE, EMBEDDING_SIZE]), name='word_embeddings', trainable=False)
    embedding_placeholder = tf.placeholder(tf.float32, [DICTIONARY_SIZE, EMBEDDING_SIZE])
    embedding_init_op = embedding_matrix.assign(embedding_placeholder)
    
    X_Q1 = tf.placeholder(tf.int32, [None, None])