|
| 1 | +#!/usr/bin/env python |
| 2 | + |
| 3 | +# |
| 4 | +# The purpose of the script is to parse the DeepLearning.java file and emit |
| 5 | +# code related to parameters. |
| 6 | +# |
| 7 | +# Currently pieces of R code get emitted and need to be pasted in manually to the R file. |
| 8 | +# |
| 9 | + |
| 10 | +import sys |
| 11 | +import os |
| 12 | +import shutil |
| 13 | +import signal |
| 14 | +import time |
| 15 | +import random |
| 16 | +import getpass |
| 17 | +import re |
| 18 | +import subprocess |
| 19 | + |
| 20 | +def read_deeplearning_file(deeplearning_file): |
| 21 | + """ |
| 22 | + Read deep learning file and generate R parameter stub stuff. |
| 23 | + |
| 24 | + @param deeplearning_file: Java source code file |
| 25 | + @return: none |
| 26 | + """ |
| 27 | + try: |
| 28 | + nlist = [] |
| 29 | + in_api = False |
| 30 | + |
| 31 | + f = open(deeplearning_file, "r") |
| 32 | + s = f.readline() |
| 33 | + lineno = 0 |
| 34 | + while (len(s) != 0): |
| 35 | + lineno = lineno + 1 |
| 36 | + stripped = s.strip() |
| 37 | + if (len(stripped) == 0): |
| 38 | + s = f.readline() |
| 39 | + continue |
| 40 | + if (stripped.startswith("@API")): |
| 41 | + # print("") |
| 42 | + if (in_api): |
| 43 | + assert(False) |
| 44 | + in_api = True |
| 45 | + |
| 46 | + # match_groups = re.search("help\s*=\s*\"([^\"]*)\"", stripped) |
| 47 | + # if (match_groups == None): |
| 48 | + # print("Missing help") |
| 49 | + # sys.exit(1) |
| 50 | + # help = match_groups.group(1) |
| 51 | + # print(help) |
| 52 | + s = f.readline() |
| 53 | + continue |
| 54 | + if (in_api): |
| 55 | + skip = False |
| 56 | + if "checkpoint" in stripped: |
| 57 | + skip = True |
| 58 | + if "expert_mode" in stripped: |
| 59 | + skip = True |
| 60 | + # if "activation" in stripped: |
| 61 | + # skip = True |
| 62 | + # if "initial_weight_distribution" in stripped: |
| 63 | + # skip = True |
| 64 | + # if "loss" in stripped: |
| 65 | + # skip = True |
| 66 | + # if "score_validation_sampling" in stripped: |
| 67 | + # skip = True |
| 68 | + |
| 69 | + if (skip): |
| 70 | + in_api = False |
| 71 | + s = f.readline() |
| 72 | + continue |
| 73 | + |
| 74 | + match_groups = re.search("public boolean (\S+) = (\S+);", s) |
| 75 | + if (match_groups is not None): |
| 76 | + t = "boolean" |
| 77 | + n = match_groups.group(1) |
| 78 | + v = match_groups.group(2) |
| 79 | + print(" parms = .addBooleanParm(parms, k=\"{}\", v={})".format(n,n)) |
| 80 | + nlist.append(n) |
| 81 | + # print(t, n, v) |
| 82 | + in_api = False |
| 83 | + s = f.readline() |
| 84 | + continue |
| 85 | + |
| 86 | + match_groups = re.search("public Activation (\S+) = (\S+);", s) |
| 87 | + if (match_groups is not None): |
| 88 | + t = "string" |
| 89 | + n = match_groups.group(1) |
| 90 | + v = match_groups.group(2) |
| 91 | + print(" parms = .addStringParm(parms, k=\"{}\", v={})".format(n,n)) |
| 92 | + nlist.append(n) |
| 93 | + # print(t, n, v) |
| 94 | + in_api = False |
| 95 | + s = f.readline() |
| 96 | + continue |
| 97 | + |
| 98 | + match_groups = re.search("public int\[\] (\S+) = .*;", s) |
| 99 | + if (match_groups is not None): |
| 100 | + t = "int array" |
| 101 | + n = match_groups.group(1) |
| 102 | + print(" parms = .addIntArrayParm(parms, k=\"{}\", v={})".format(n,n)) |
| 103 | + nlist.append(n) |
| 104 | + # print(t, n) |
| 105 | + in_api = False |
| 106 | + s = f.readline() |
| 107 | + continue |
| 108 | + |
| 109 | + match_groups = re.search("public int (\S+) = .*;", s) |
| 110 | + if (match_groups is not None): |
| 111 | + t = "int" |
| 112 | + n = match_groups.group(1) |
| 113 | + print(" parms = .addIntParm(parms, k=\"{}\", v={})".format(n,n)) |
| 114 | + nlist.append(n) |
| 115 | + # print(t, n) |
| 116 | + in_api = False |
| 117 | + s = f.readline() |
| 118 | + continue |
| 119 | + |
| 120 | + match_groups = re.search("public double (\S+) = (\S+);", s) |
| 121 | + if (match_groups is not None): |
| 122 | + t = "double" |
| 123 | + n = match_groups.group(1) |
| 124 | + v = match_groups.group(2) |
| 125 | + print(" parms = .addDoubleParm(parms, k=\"{}\", v={})".format(n,n)) |
| 126 | + nlist.append(n) |
| 127 | + # print(t, n, v) |
| 128 | + in_api = False |
| 129 | + s = f.readline() |
| 130 | + continue |
| 131 | + |
| 132 | + match_groups = re.search("public float (\S+) = (\S+);", s) |
| 133 | + if (match_groups is not None): |
| 134 | + t = "float" |
| 135 | + n = match_groups.group(1) |
| 136 | + v = match_groups.group(2) |
| 137 | + print(" parms = .addFloatParm(parms, k=\"{}\", v={})".format(n,n)) |
| 138 | + nlist.append(n) |
| 139 | + # print(t, n, v) |
| 140 | + in_api = False |
| 141 | + s = f.readline() |
| 142 | + continue |
| 143 | + |
| 144 | + match_groups = re.search("public double\[\] (\S+);", s) |
| 145 | + if (match_groups is not None): |
| 146 | + t = "double array" |
| 147 | + n = match_groups.group(1) |
| 148 | + print(" parms = .addDoubleArrayParm(parms, k=\"{}\", v={})".format(n,n)) |
| 149 | + nlist.append(n) |
| 150 | + # print(t, n) |
| 151 | + in_api = False |
| 152 | + s = f.readline() |
| 153 | + continue |
| 154 | + |
| 155 | + match_groups = re.search("public long (\S+) = new Random.*;", s) |
| 156 | + if (match_groups is not None): |
| 157 | + t = "long" |
| 158 | + n = match_groups.group(1) |
| 159 | + v = -1 |
| 160 | + print(" parms = .addLongParm(parms, k=\"{}\", v={})".format(n,n)) |
| 161 | + nlist.append(n) |
| 162 | + # print(t, n, v) |
| 163 | + in_api = False |
| 164 | + s = f.readline() |
| 165 | + continue |
| 166 | + |
| 167 | + match_groups = re.search("public long (\S+) = (\S+);", s) |
| 168 | + if (match_groups is not None): |
| 169 | + t = "long" |
| 170 | + n = match_groups.group(1) |
| 171 | + v = match_groups.group(2) |
| 172 | + print(" parms = .addLongParm(parms, k=\"{}\", v={})".format(n,n)) |
| 173 | + nlist.append(n) |
| 174 | + # print(t, n, v) |
| 175 | + in_api = False |
| 176 | + s = f.readline() |
| 177 | + continue |
| 178 | + |
| 179 | + if (stripped == "public InitialWeightDistribution initial_weight_distribution = InitialWeightDistribution.UniformAdaptive;"): |
| 180 | + t = "string" |
| 181 | + n = "initial_weight_distribution" |
| 182 | + print(" parms = .addStringParm(parms, k=\"{}\", v={})".format(n,n)) |
| 183 | + nlist.append(n) |
| 184 | + # print(t, "initial_weight_distribution", "UniformAdaptive") |
| 185 | + in_api = False |
| 186 | + s = f.readline() |
| 187 | + continue |
| 188 | + |
| 189 | + if (stripped == "public Loss loss = Loss.CrossEntropy;"): |
| 190 | + t = "string" |
| 191 | + n = "loss" |
| 192 | + print(" parms = .addStringParm(parms, k=\"{}\", v={})".format(n,n)) |
| 193 | + nlist.append(n) |
| 194 | + # print(t, "loss", "CrossEntropy") |
| 195 | + in_api = False |
| 196 | + s = f.readline() |
| 197 | + continue |
| 198 | + |
| 199 | + if (stripped == "public ClassSamplingMethod score_validation_sampling = ClassSamplingMethod.Uniform;"): |
| 200 | + t = "string" |
| 201 | + n = "score_validation_sampling" |
| 202 | + print(" parms = .addStringParm(parms, k=\"{}\", v={})".format(n,n)) |
| 203 | + nlist.append(n) |
| 204 | + # print(t, "score_validation_sampling", "Uniform") |
| 205 | + in_api = False |
| 206 | + s = f.readline() |
| 207 | + continue |
| 208 | + |
| 209 | + print("ERROR: No match group found on line ", lineno) |
| 210 | + sys.exit(1) |
| 211 | + |
| 212 | + s = f.readline() |
| 213 | + f.close() |
| 214 | + |
| 215 | + for n in nlist: |
| 216 | + print(" {},".format(n)) |
| 217 | + |
| 218 | + except IOError as e: |
| 219 | + print("") |
| 220 | + print("ERROR: Failure reading test list: " + deeplearning_file) |
| 221 | + print(" (errno {0}): {1}".format(e.errno, e.strerror)) |
| 222 | + print("") |
| 223 | + sys.exit(1) |
| 224 | + |
| 225 | +def main(argv): |
| 226 | + read_deeplearning_file("./src/main/java/hex/deeplearning/DeepLearning.java") |
| 227 | + |
| 228 | +if __name__ == "__main__": |
| 229 | + main(sys.argv) |
0 commit comments