dltetris/load_functions.py at master · smlgorta/dltetris · GitHub

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
import numpy as np
import pandas as pd
import ast
import pickle

def from_np_array(array_string):
    array_string = ','.join(array_string.replace('[ ', '[').split())
    return np.array(ast.literal_eval(array_string))

def pickleSave(path, obj):
    fileObject = open(path, 'wb')
    pickle.dump(obj, fileObject)
    fileObject.close()
    return;

def pickleLoad(path):
    fileObject = open(path, 'rb')
    return pickle.load(fileObject)

def parseState(state):
    splitted_state = state.split("|")
    board = np.zeros((10, 10))
    for row in range(1, len(splitted_state)):
        splitted_row = splitted_state[row].split(":")
        for col in range(0, len(splitted_row)):
            if splitted_row[col] != "":
                board[row-1][int(splitted_row[col])] = 1
    return board

def loadAndParseData(input_path, output_path):
    print("loading csv")
    csv = pd.read_csv(input_path)
    def parseStates(states):
        parsed_states = []
        for i in range(states.shape[0]):
            parsed_states.append(parseState(states[i]).flatten())
        return parsed_states


    print("parsing states")
    csv["stateaction"] = parseStates(csv["stateaction"])
    csv = pd.DataFrame(csv)
    # print("writing to csv")
    # csv.to_csv(output_path, header=True, index=False)
    print("writing to pickle")
    fileObject = open(output_path, 'wb')
    pickle.dump(csv, fileObject)
    fileObject.close()

def loadParsedReshapeAndFlatten(input_path, input_outpath, output_outpath):
    # print("loading parsed states from data/parsed.csv")
    # csv = pd.read_csv(input_path, converters={'stateaction':from_np_array})

    print("reading from pickle")
    csv = pickleLoad(input_path)

    print("grouping by state")
    grouped_csv = csv.groupby("step", as_index=True)
    print("setting indices for actions")
    indices = []
    for i, row in grouped_csv.size().iteritems():
        indices.extend(grouped_csv.get_group(i).reset_index().index.values)
    print("attaching indices to main frame")
    csv['option'] = indices
    print("reshaping")
    pivoted_csv = csv.pivot(index="step", columns='option')

    def softmax(x):
        z = np.sum(np.exp(x))
        sm = np.exp(x) / z
        return np.nan_to_num(sm);

    input_ret = []
    output_ret = []
    for index, row in pivoted_csv.iterrows():
        input_row = []
        repeat = 0
        for i in range(34):
            if row.get_values()[i] is not None:
                input_row = np.append(input_row, row.get_values()[i])
            else:
                if row.get_values()[repeat] is None:
                    repeat = 0
                input_row = np.append(input_row, row.get_values()[repeat])
                repeat += 1

        # print("row: ", input_row)
        input_ret.append(input_row)
        output_row = []
        repeat = 0
        for i in range(34, 68):
            if row.get_values()[i] is not None:
                output_row = np.append(output_row, row.get_values()[i])
            else:
                if row.get_values()[34 + repeat] is None:
                    repeat = 0
                output_row = np.append(output_row, row.get_values()[34+repeat])
                repeat += 1
        output_ret.append(softmax(output_row.flatten()))
    output_ret = np.array(output_ret)
    input_ret = np.array(input_ret)
    print("writing to pickle")
    print(input_ret.shape)
    print(output_ret.shape)
    pickleSave(output_outpath, output_ret)
    pickleSave(input_outpath, input_ret)

def loadParsedAndMakePairwiseComparisons(input_path, output_path):
    l = pickleLoad(input_path)
    steps = l["step"]
    steps = steps.drop_duplicates()
    steps = steps.reset_index(drop = True)
    pw_dataset = []
    for i in range(len(steps)):
        step = l[l['step'] == steps[i]]
        for j in range(len(step)):
            for k in range(len(step)):
                pw = []
                if(j != k):
                    pw = [l['stateaction'][j],l['stateaction'][k],l['value'][j]-l['value'][k]]
                    if(len(pw_dataset) == 0):
                        pw_dataset = pw
                    else:
                        pw_dataset = np.vstack([pw_dataset, pw])
    pickleSave(output_path, pw_dataset)

def makePairwiseComparisons(placements, steps):
    pw_dataset = []
    for i in range(len(steps)):
        step = placements[placements['step'] == steps[i]]
        step = step.reset_index(drop=True)
        for j in range(len(step)):
            for k in range(len(step)):
                pw = []
                if (j != k):
                    pw = [step['stateaction'][j], step['stateaction'][k], step['value'][j] - step['value'][k]]
                    if (len(pw_dataset) == 0):
                        pw_dataset = pw
                    else:
                        pw_dataset = np.vstack([pw_dataset, pw])
    return pw_dataset

def makePairwiseComparisons(placements):
    pw_dataset = []
    for j in range(len(placements)):
            for k in range(len(placements)):
                pw = []
                if (j != k):
                    pw = [placements['stateaction'][j], placements['stateaction'][k], float(placements['value'][j]) - float(placements['value'][k])]
                    if (len(pw_dataset) == 0):
                        pw_dataset = pw
                    else:
                        pw_dataset = np.vstack([pw_dataset, pw])
    return pw_dataset