/
SConstruct
289 lines (261 loc) · 18.3 KB
/
SConstruct
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
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
import os
import os.path
import logging
import random
import subprocess
import shlex
import gzip
import re
import functools
import time
import imp
import sys
import json
from hashlib import md5
#import starcoder
import steamroller
# workaround needed to fix bug with SCons and the pickle module
del sys.modules['pickle']
sys.modules['pickle'] = imp.load_module('pickle', *imp.find_module('pickle'))
import pickle
vars = Variables("custom.py")
vars.AddVariables(
("OUTPUT_WIDTH", "", 100),
("RANDOM_COMPONENTS", "", 1000),
("MINIMUM_CONSTANTS", "", 1),
("MAXIMUM_CONSTANTS", "", 4),
("TRAIN_PROPORTION", "", 0.9),
("DEV_PROPORTION", "", 0.1),
("MAX_CATEGORICAL", "", 50),
("MAX_SEQUENCE_LENGTH", "", 100),
("RANDOM_LINE_COUNT", "", 1000),
("MAX_COLLAPSE", "", 0),
("MAX_SEQUENCE_LENGTH", "", 32),
("LOG_LEVEL", "", "INFO"),
("LINE_COUNT", "", 1000),
("BATCH_SIZE", "", 128),
("MAX_EPOCHS", "", 10),
("LEARNING_RATE", "", 0.001),
("RANDOM_RESTARTS", "", 0),
("ELASTIC_HOST", "", "localhost"),
("ELASTIC_PORT", "", 5601),
("MOMENTUM", "", None),
("EARLY_STOP", "", 10),
("PATIENCE", "", 5),
("HIDDEN_SIZE", "", 32),
("EMBEDDING_SIZE", "", 32),
("AUTOENCODER_SHAPES", "", [32, 16]),
("CLUSTER_REDUCTION", "", 0.5),
BoolVariable("USE_GPU", "", False),
BoolVariable("USE_GRID", "", False),
("GPU_PREAMBLE", "", "module load cuda10.1/toolkit"),
("SPLIT_PROPORTIONS", "", [("train", 0.80), ("dev", 0.10), ("test", 0.10)]),
("DEPTH", "", 1),
("SPLITTER_CLASS", "", "sample_components"),
("BATCHIFIER_CLASS", "", "sample_components"),
("SLAVE_TRADE_PATH", "", "${DATA_PATH}/slavery"),
("MAROON_PATH", "", "${DATA_PATH}/maroon_ads"),
("FLUENCY_PATH", "", "${DATA_PATH}/russian_fluency"),
("SENTIMENT_PATH", "", "${DATA_PATH}/stanford_sentiment_treebank"),
("SMOKING_AND_VAPING_PATH", "", "${DATA_PATH}/smoking_and_vaping.txt.bz2"),
("ENTERTAINING_PATH", "", "${DATA_PATH}/entertaining_america"),
("DH_PATH", "", "${DATA_PATH}/documentary_hypothesis"),
("ME_PATH", "", "${DATA_PATH}/middle_english"),
("AFFICHES_PATH", "", "${DATA_PATH}/affiches_americaines"),
("ROYAL_INSCRIPTIONS_PATH", "", "${DATA_PATH}"),
("WALS_PATH", "", "${DATA_PATH}"),
("TWITTER_LID_PATH", "", "${DATA_PATH}/twitter_lid"),
("WOMEN_WRITERS_PATH", "", "${DATA_PATH}"),
("PARIS_TAX_ROLLS_PATH", "", "${DATA_PATH}"),
("LITBANK_PATH", "", "${DATA_PATH}/litbank"),
("MULTIMODAL_WIKIPEDIA_PATH", "", "${DATA_PATH}/multimodal_wikipedia"),
("DATA_PATH", "", ""),
("EXPERIMENTS", "", {}),
("ELASTIC_HOST", "", "localhost"),
("ELASTIC_PORT", "", 9200),
("ELASTIC_USER", "", "elastic"),
("ELASTIC_PASSWORD", "", ""),
("KIBANA_HOST", "", "localhost"),
("KIBANA_PORT", "", 5601),
("KIBANA_USER", "", "elastic"),
("KIBANA_PASSWORD", "", ""),
("DJANGO_USER", "", "admin"),
("DJANGO_PASSWORD", "", "admin"),
("DJANGO_EMAIL", "", "nothing@nothing.com"),
BoolVariable("ELASTIC_UPLOAD", "", False),
("SERVER_HOST", "", "localhost"),
("SERVER_PORT", "", 8080),
)
env = Environment(variables=vars, ENV=os.environ, TARFLAGS="-c -z", TARSUFFIX=".tgz",
tools=["default", steamroller.generate])
env.Append(BUILDERS={"PreprocessArithmetic" : env.Builder(**env.ActionMaker("python",
"scripts/preprocess_arithmetic.py",
"--output ${TARGETS[0]} --components ${RANDOM_COMPONENTS} --minimum_constants ${MINIMUM_CONSTANTS} --maximum_constants ${MAXIMUM_CONSTANTS}")),
"PreprocessTargetedSentiment" : env.Builder(**env.ActionMaker("python",
"scripts/preprocess_targeted_sentiment.py",
"${SOURCES} --output ${TARGETS[0]}")),
"PreprocessMaroonAds" : env.Builder(**env.ActionMaker("python",
"scripts/preprocess_maroon_ads.py",
"${SOURCES[0]} --output ${TARGETS[0]}")),
"PreprocessFluency" : env.Builder(**env.ActionMaker("python",
"scripts/preprocess_fluency.py",
"${SOURCES} --output ${TARGETS[0]}")),
"PreprocessSmokingAndVaping" : env.Builder(**env.ActionMaker("python",
"scripts/preprocess_reddit.py",
"${SOURCES} --output ${TARGETS[0]}")),
"PreprocessMiddleEnglish" : env.Builder(**env.ActionMaker("python",
"scripts/preprocess_middle_english.py",
"${SOURCES} --output ${TARGETS[0]}")),
"PreprocessDocumentaryHypothesis" : env.Builder(**env.ActionMaker("python",
"scripts/preprocess_documentary_hypothesis.py",
"${SOURCES} --output ${TARGETS[0]}")),
"PreprocessAffichesAmericaines" : env.Builder(**env.ActionMaker("python",
"scripts/preprocess_affiches_americaines.py",
"${SOURCES} --output ${TARGETS[0]}")),
"PreprocessParisTaxRolls" : env.Builder(**env.ActionMaker("python",
"scripts/preprocess_paris_tax_rolls.py",
"${SOURCES} --output ${TARGETS[0]}")),
"PreprocessEntertainingAmerica" : env.Builder(**env.ActionMaker("python",
"scripts/preprocess_entertaining_america.py",
"${SOURCES} --output ${TARGETS[0]}")),
"PreprocessWomenWriters" : env.Builder(**env.ActionMaker("python",
"scripts/preprocess_women_writers.py",
"${SOURCES} --output ${TARGETS[0]}")),
"PreprocessRoyalInscriptions" : env.Builder(**env.ActionMaker("python",
"scripts/preprocess_royal_inscriptions.py",
"${SOURCES} --output ${TARGETS[0]}")),
"PreprocessLinguisticLid" : env.Builder(**env.ActionMaker("python",
"scripts/preprocess_linguistic_lid.py",
"--output ${TARGETS[0]} ${SOURCES}")),
"PreprocessLitbank" : env.Builder(**env.ActionMaker("python",
"scripts/preprocess_litbank.py",
"--output ${TARGETS[0]} --input ${SOURCES[0]}")),
"PreprocessMultimodalWikipedia" : env.Builder(**env.ActionMaker("python",
"scripts/preprocess_multimodal_wikipedia.py",
"--output ${TARGETS[0]} --input ${SOURCES[0]}")),
"PreprocessPostAtlanticSlaveTrade" : env.Builder(**env.ActionMaker("python",
"scripts/preprocess_post_atlantic_slave_trade.py",
"--output ${TARGETS[0]} ${SOURCES}")),
"PrepareDataset" : env.Builder(**env.ActionMaker("python",
"scripts/prepare_dataset.py",
"--schema_output ${TARGETS[0]} --data_output ${TARGETS[1]} --data_input ${SOURCES[0]} --schema_input ${SOURCES[1]}",
other_deps=[])),
"SplitData" : env.Builder(**env.ActionMaker("python",
"scripts/split_data.py",
"--input ${SOURCES[0]} --proportions ${PROPORTIONS} --outputs ${TARGETS} --random_seed ${RANDOM_SEED} --splitter_class ${SPLITTER_CLASS} --shared_entity_types ${SHARED_ENTITY_TYPES}",
other_deps=[],
)),
"TrainModel" : env.Builder(**env.ActionMaker("python",
"scripts/train_model.py",
"--data ${SOURCES[0]} --train ${SOURCES[1]} --dev ${SOURCES[2]} --model_output ${TARGETS[0]} --trace_output ${TARGETS[1]} ${'--gpu' if USE_GPU else ''} ${'--autoencoder_shapes ' + ' '.join(map(str, AUTOENCODER_SHAPES)) if AUTOENCODER_SHAPES != None else ''} ${'--mask ' + ' '.join(MASK) if MASK else ''} --log_level ${LOG_LEVEL} ${'--autoencoder' if AUTOENCODER else ''} --random_restarts ${RANDOM_RESTARTS} ${' --subselect ' if SUBSELECT==True else ''} --batchifier_class ${BATCHIFIER_CLASS} --shared_entity_types ${SHARED_ENTITY_TYPES}",
other_args=["DEPTH", "MAX_EPOCHS", "LEARNING_RATE", "RANDOM_SEED", "PATIENCE", "MOMENTUM", "BATCH_SIZE",
"EMBEDDING_SIZE", "HIDDEN_SIZE", "FIELD_DROPOUT", "HIDDEN_DROPOUT", "EARLY_STOP"],
USE_GPU=env["USE_GPU"],
)),
"ApplyModel" : env.Builder(**env.ActionMaker("python",
"scripts/apply_model.py",
"--model ${SOURCES[0]} --dataset ${SOURCES[1]} ${'--split ' + SOURCES[2].rstr() if len(SOURCES) == 3 else ''} --output ${TARGETS[0]} ${'--gpu' if USE_GPU else ''}",
)),
"UploadResults" : env.Builder(**env.ActionMaker("python",
"scripts/upload_results.py",
"--results ${SOURCES[0]} --log ${TARGETS[0]} --elastic_host ${ELASTIC_HOST} --elastic_port ${ELASTIC_PORT} --create_indices --overwrite_existing --experiment_id ${APPLY_CONFIG_ID} --experiment_name ${EXPERIMENT_NAME} --elastic_user ${ELASTIC_USER} --elastic_password ${ELASTIC_PASSWORD} --kibana_host ${KIBANA_HOST} --kibana_port ${KIBANA_PORT} --kibana_user ${KIBANA_USER} --kibana_password ${KIBANA_PASSWORD} --model ${SOURCES[1]}",
)),
#"ClusterEntities" : env.Builder(**env.ActionMaker("python",
# "scripts/cluster_entities.py",
# "--input ${SOURCES[0]} --schema ${SOURCES[1]} --output ${TARGETS[0]} --reduction ${CLUSTER_REDUCTION}")),
#"InspectClusters" : env.Builder(**env.ActionMaker("python",
# "scripts/inspect_clusters.py",
# "--input ${SOURCES[0]} --schema ${SOURCES[1]} --output ${TARGETS[0]}")),
"PlotTrace" : env.Builder(**env.ActionMaker("python",
"scripts/plot_trace.py",
"--input ${SOURCES[0]} --output ${TARGETS[0]}")),
"MakeServerConfig" : env.Builder(**env.ActionMaker("python",
"scripts/make_server_config.py",
"--models ${SOURCES} --names ${NAMES} --elastic_host ${ELASTIC_HOST} --elastic_port ${ELASTIC_PORT} --elastic_user ${ELASTIC_USER} --elastic_password ${ELASTIC_PASSWORD} --host ${SERVER_HOST} --port ${SERVER_PORT} --django_user ${DJANGO_USER} --django_password ${DJANGO_PASSWORD} --django_email ${DJANGO_EMAIL} --output ${TARGETS[0]}")),
},
tools=["default"],
)
# function for width-aware printing of commands
def print_cmd_line(s, target, source, env):
if len(s) > int(env["OUTPUT_WIDTH"]):
print(s[:int(float(env["OUTPUT_WIDTH"]) / 2) - 2] + "..." + s[-int(float(env["OUTPUT_WIDTH"]) / 2) + 1:])
else:
print(s)
# and the command-printing function
env['PRINT_CMD_LINE_FUNC'] = print_cmd_line
# and how we decide if a dependency is out of date
env.Decider("timestamp-newer")
def run_experiment(env, experiment_config, **args):
data = sum([env.Glob(env.subst(p)) for p in experiment_config.get("DATA_FILES", [])], [])
schema = experiment_config.get("SCHEMA", None)
title = experiment_name.replace("_", " ").title().replace(" ", "")
data = getattr(env, "Preprocess{}".format(title))("work/${EXPERIMENT_NAME}/data.json.gz",
data, **args)
# prepare the final spec and dataset
observed_schema, dataset = env.PrepareDataset(["work/${EXPERIMENT_NAME}/schema.json.gz", "work/${EXPERIMENT_NAME}/dataset.pkl.gz"],
[data] + ([] if schema == None else [schema]),
**args)
split_names = [n for n, _ in experiment_config.get("SPLIT_PROPORTIONS", env["SPLIT_PROPORTIONS"])]
split_props = [p for _, p in experiment_config.get("SPLIT_PROPORTIONS", env["SPLIT_PROPORTIONS"])]
train, dev, test = env.SplitData(["work/${{EXPERIMENT_NAME}}/{0}.pkl.gz".format(n) for n in split_names],
dataset,
**experiment_config,
**args, RANDOM_SEED=0, PROPORTIONS=split_props)
env.Alias("splits", [train, dev, test])
# expand training configurations
train_configs = [[]]
for arg_name, values in experiment_config.get("TRAIN_CONFIG", {}).items():
train_configs = sum([[config + [(arg_name.upper(), v)] for config in train_configs] for v in values], [])
train_configs = [dict(config) for config in train_configs]
# expand apply configurations
apply_configs = [[]]
for arg_name, values in experiment_config.get("APPLY_CONFIG", {}).items():
apply_configs = sum([[config + [(arg_name.upper(), v)] for config in apply_configs] for v in values], [])
apply_configs = [dict(config) for config in apply_configs]
results = []
for config in train_configs:
args["TRAIN_CONFIG_ID"] = md5(str(sorted(list(config.items()))).encode()).hexdigest()
model, trace = env.TrainModel(["work/${EXPERIMENT_NAME}/model_${TRAIN_CONFIG_ID}.pkl.gz",
"work/${EXPERIMENT_NAME}/trace_${TRAIN_CONFIG_ID}.json.gz"],
[dataset, train, dev],
**args,
**experiment_config,
**config)
env.PlotTrace("work/${EXPERIMENT_NAME}/traceplot_${TRAIN_CONFIG_ID}.png",
trace,
**args
)
for apply_config in apply_configs:
config.update(apply_config)
args["APPLY_CONFIG_ID"] = md5(str(sorted(list(config.items()))).encode()).hexdigest()
output = env.ApplyModel("work/${EXPERIMENT_NAME}/${FOLD}/output_${APPLY_CONFIG_ID}.json.gz",
[model, dataset],
**args,
**experiment_config,
**config)
if env["ELASTIC_UPLOAD"] == True:
upload = env.UploadResults("work/${EXPERIMENT_NAME}/${FOLD}/upload_${APPLY_CONFIG_ID}.txt",
[output, model],
**args,
**experiment_config,
**config)
#continue
#clusters = env.ClusterEntities("work/${EXPERIMENT_NAME}/${FOLD}/clusters_${APPLY_CONFIG_ID}.json.gz",
# [output, schema],
# **args,
# **config)
#inspect_clusters = env.InspectClusters("work/${EXPERIMENT_NAME}/${FOLD}/inspect_clusters_${APPLY_CONFIG_ID}.txt",
# [clusters, observed_schema],
# **args,
# **config)
return model
env.AddMethod(run_experiment, "RunExperiment")
#
# Run all experiments
#
models = []
names = []
for experiment_name, experiment_config in env["EXPERIMENTS"].items():
names.append(experiment_name)
models.append(env.RunExperiment(experiment_config, EXPERIMENT_NAME=experiment_name))
env.MakeServerConfig("work/server_config.json", models, NAMES=names)