def testCorpusSizeSimple(self):
# There should be a difference between forceLast and not forceLast.
patternDir = os.path.join(self.testContext["MAT_PKG_HOME"], "sample",
"ne", "resources", "data", "json")
from MAT.Bootstrap import Bootstrapper, DocumentSet, TrainingRun, TestRun, CorpusSizeIterator
e = Bootstrapper(dir=self.expDir,
task=self.task,
corpora=[
DocumentSet("test",
partitions=[("train", 4),
("test", 1)],
filePats=["*.json"],
prefix=patternDir)
],
models=[
TrainingRun("test",
trainingCorpora=[("test", "train")],
iterators=[CorpusSizeIterator(2)])
])
e.run()
# OK, this works. There should be a subdirectory for each model.
allInstances = e.getModel("test").allInstances
self.assertEqual(len(allInstances), 4)
self.assertEqual([len(m.trainingSet.getFiles()) for m in allInstances],
[2, 4, 6, 8])
self.assertTrue(
set(allInstances[0].trainingSet.getFiles()) < set(
allInstances[1].trainingSet.getFiles()))
self.assertTrue(
set(allInstances[1].trainingSet.getFiles()) < set(
allInstances[2].trainingSet.getFiles()))
self.assertTrue(
set(allInstances[2].trainingSet.getFiles()) < set(
allInstances[3].trainingSet.getFiles()))
def testIncrementLastStep(self):
# There should be a difference between forceLast and not forceLast.
patternDir = os.path.join(self.testContext["MAT_PKG_HOME"], "sample",
"ne", "resources", "data", "json")
from MAT.Bootstrap import Bootstrapper, DocumentSet, TrainingRun, TestRun, IncrementIterator
e = Bootstrapper(
dir=self.expDir,
task=self.task,
corpora=[
DocumentSet("test",
partitions=[("train", 4), ("test", 1)],
filePats=["*.json"],
prefix=patternDir)
],
models=[
TrainingRun("test",
trainingCorpora=[("test", "train")],
iterators=[
IncrementIterator("engineSettings",
"max_iterations",
4,
7,
2,
forceLast=True)
])
])
e.run()
# OK, this works. There should be a subdirectory for each model.
self.assertTrue(len(e.getModel("test").allInstances) == 3)
def testDoubleModelIncrement(self):
# Forcing last here because of float rounding.
patternDir = os.path.join(self.testContext["MAT_PKG_HOME"], "sample",
"ne", "resources", "data", "json")
from MAT.Bootstrap import Bootstrapper, DocumentSet, TrainingRun, TestRun, IncrementIterator
e = Bootstrapper(
dir=self.expDir,
task=self.task,
corpora=[
DocumentSet("test",
partitions=[("train", 4), ("test", 1)],
filePats=["*.json"],
prefix=patternDir)
],
models=[
TrainingRun("test",
trainingCorpora=[("test", "train")],
engineSettings={"l1": True},
iterators=[
IncrementIterator("engineSettings",
"max_iterations", 4, 8, 2),
IncrementIterator("engineSettings",
"l1_c",
0.1,
0.3,
.1,
forceLast=True)
])
])
e.run()
self.assertTrue(len(e.getModel("test").allInstances) == 9)
self.assertEqual([
m.engineSettings["max_iterations"]
for m in e.getModel("test").allInstances
], [4, 4, 4, 6, 6, 6, 8, 8, 8])
self.assertEqual([
str(m.engineSettings["l1_c"])
for m in e.getModel("test").allInstances
], ['0.1', '0.2', '0.3', '0.1', '0.2', '0.3', '0.1', '0.2', '0.3'])
self.assertEqual(
[m.engineSettings["l1"] for m in e.getModel("test").allInstances],
[True] * 9)
self.assertEqual([
m.modelSubdir for m in e.getModel("test").allInstances
], [
"test_max_iterations_4_l1_c_0_1", "test_max_iterations_4_l1_c_0_2",
"test_max_iterations_4_l1_c_0_3", "test_max_iterations_6_l1_c_0_1",
"test_max_iterations_6_l1_c_0_2", "test_max_iterations_6_l1_c_0_3",
"test_max_iterations_8_l1_c_0_1", "test_max_iterations_8_l1_c_0_2",
"test_max_iterations_8_l1_c_0_3"
])
def testModelPlusRunIncrement(self):
patternDir = os.path.join(self.testContext["MAT_PKG_HOME"], "sample",
"ne", "resources", "data", "json")
from MAT.Bootstrap import Bootstrapper, DocumentSet, TrainingRun, TestRun, IncrementIterator
e = Bootstrapper(
dir=self.expDir,
task=self.task,
corpora=[
DocumentSet("test",
partitions=[("train", 4), ("test", 1)],
filePats=["*.json"],
prefix=patternDir)
],
models=[
TrainingRun("test",
trainingCorpora=[("test", "train")],
iterators=[
IncrementIterator("engineSettings",
"max_iterations", 4, 8, 2)
])
],
runs=[
TestRun("test",
model="test",
testCorpora=[("test", "test")],
engineOptions={
"steps": "zone,tokenize,tag",
"workflow": "Demo"
},
iterators=[
IncrementIterator("engineOptions", "prior_adjust",
-1, 1, 1)
])
])
e.run()
# OK, this works. There should be a subdirectory for each model.
self.assertTrue(len(e.getModel("test").allInstances) == 3)
self.assertEqual([
m.engineSettings["max_iterations"]
for m in e.getModel("test").allInstances
], [4, 6, 8])
self.assertEqual(
[m.modelSubdir for m in e.getModel("test").allInstances], [
"test_max_iterations_4", "test_max_iterations_6",
"test_max_iterations_8"
])
self.assertTrue(len(e.runTable["test"].allInstances) == 9)
# Interleaved, so model dominant.
self.assertEqual([(os.path.basename(os.path.dirname(
r.runDir)), os.path.basename(r.runDir))
for r in e.runTable["test"].allInstances],
[("test_prior_adjust__1", "test_max_iterations_4"),
("test_prior_adjust_0", "test_max_iterations_4"),
("test_prior_adjust_1", "test_max_iterations_4"),
("test_prior_adjust__1", "test_max_iterations_6"),
("test_prior_adjust_0", "test_max_iterations_6"),
("test_prior_adjust_1", "test_max_iterations_6"),
("test_prior_adjust__1", "test_max_iterations_8"),
("test_prior_adjust_0", "test_max_iterations_8"),
("test_prior_adjust_1", "test_max_iterations_8")])
def testSimple(self):
# I'm going to do a simple boostrap, constructed from objects.
patternDir = os.path.join(self.testContext["MAT_PKG_HOME"], "sample",
"ne", "resources", "data", "json")
from MAT.Bootstrap import Bootstrapper, DocumentSet, TrainingRun, TestRun
e = Bootstrapper(
dir=self.expDir,
task=self.task,
corpora=[
DocumentSet("test",
partitions=[
("train", 3),
("test", DocumentSet.FIXED_PARTITION_REMAINDER)
],
partitionIsFixed=True,
filePats=["*.json"],
prefix=patternDir)
],
models=[TrainingRun("test", trainingCorpora=[("test", "train")])],
runs=[
TestRun("test",
model="test",
testCorpora=[("test", "test")],
engineOptions={
"steps": "zone,tokenize,tag",
"workflow": "Demo"
})
])
e.run()
# OK, now we need to make sure that there's a model, and that in the
# run input and the hyp input, there are two files each, whose names are
# the prefix of the test slice of the test corpus.
self.assertTrue(len(e.getModel("test").allInstances) == 1)
m = e.getModel("test").allInstances[0]
self.assertTrue(os.path.exists(os.path.join(m.modelDir, "model")))
files = e.corporaTable["test"].getFiles(partition="test")
self.assertEqual(len(files), 7)
self.assertEqual(len(m.trainingSet.getFiles()), 3)
self.assertTrue(len(e.runTable["test"].allInstances) == 1)
r = e.runTable["test"].allInstances[0]
runDir = r.runDir
for file in files:
self.assertTrue(
os.path.exists(
os.path.join(runDir, "hyp", os.path.basename(file)) +
".prepped.tag.json"))
self.assertTrue(
os.path.exists(
os.path.join(runDir, "run_input", os.path.basename(file)) +
".prepped"))
def testFiveWay(self):
# I'm going to do a simple boostrap, constructed from objects.
patternDir = os.path.join(self.testContext["MAT_PKG_HOME"], "sample",
"ne", "resources", "data", "json")
from MAT.Bootstrap import Bootstrapper, DocumentSet, TrainingRun, TestRun
e = Bootstrapper(dir=self.expDir,
task=self.task,
corpora=[
DocumentSet("test",
partitions=[("s1", 1), ("s2", 1),
("s3", 1), ("s4", 1),
("s5", 1)],
filePats=["*.json"],
prefix=patternDir)
],
models=[
TrainingRun("s1234",
trainingCorpora=[("test", "s1"),
("test", "s2"),
("test", "s3"),
("test", "s4")]),
TrainingRun("s1235",
trainingCorpora=[("test", "s1"),
("test", "s2"),
("test", "s3"),
("test", "s5")]),
TrainingRun("s1245",
trainingCorpora=[("test", "s1"),
("test", "s2"),
("test", "s4"),
("test", "s5")]),
TrainingRun("s1345",
trainingCorpora=[("test", "s1"),
("test", "s3"),
("test", "s4"),
("test", "s5")]),
TrainingRun("s2345",
trainingCorpora=[("test", "s2"),
("test", "s3"),
("test", "s4"),
("test", "s5")])
],
runs=[
TestRun("s1",
model="s2345",
testCorpora=[("test", "s1")],
engineOptions={
"steps": "zone,tokenize,tag",
"workflow": "Demo"
}),
TestRun("s2",
model="s1345",
testCorpora=[("test", "s2")],
engineOptions={
"steps": "zone,tokenize,tag",
"workflow": "Demo"
}),
TestRun("s3",
model="s1245",
testCorpora=[("test", "s3")],
engineOptions={
"steps": "zone,tokenize,tag",
"workflow": "Demo"
}),
TestRun("s4",
model="s1235",
testCorpora=[("test", "s4")],
engineOptions={
"steps": "zone,tokenize,tag",
"workflow": "Demo"
}),
TestRun("s5",
model="s1234",
testCorpora=[("test", "s5")],
engineOptions={
"steps": "zone,tokenize,tag",
"workflow": "Demo"
})
])
e.run()
# OK, now we need to make sure that there's a model, and that in the
# run input and the hyp input, there are two files each, whose names are
# the prefix of the test slice of the test corpus.
for m, mTemplate in e.modelSetTable.items():
self.assertTrue(len(mTemplate.allInstances) == 1)
self.assertTrue(
os.path.exists(
os.path.join(e.getModelDir(mTemplate.allInstances[0]),
"model")))
self.assertEqual(
len(mTemplate.allInstances[0].trainingSet.getFiles()), 8)
for p in ["s1", "s2", "s3", "s4", "s5"]:
files = e.corporaTable["test"].getFiles(partition=p)
self.assertEqual(len(files), 2)
r = e.runTable[p]
self.assertTrue(len(r.allInstances) == 1)
runDir = r.allInstances[0].runDir
for file in files:
self.assertTrue(
os.path.exists(
os.path.join(runDir, "hyp", os.path.basename(file)) +
".prepped.tag.json"))
self.assertTrue(
os.path.exists(
os.path.join(runDir, "run_input", os.path.basename(
file)) + ".prepped"))
def testIncrement(self):
patternDir = os.path.join(self.testContext["MAT_PKG_HOME"], "sample",
"ne", "resources", "data", "json")
from MAT.Bootstrap import Bootstrapper, DocumentSet, TrainingRun, TestRun, IncrementIterator
e = Bootstrapper(dir=self.expDir,
task=self.task,
corpora=[
DocumentSet("test",
partitions=[("train", 4),
("test", 1)],
filePats=["*.json"],
prefix=patternDir)
],
models=[
TrainingRun("test",
trainingCorpora=[("test", "train")],
iterators=[
IncrementIterator(
"engineSettings",
"max_iterations", 4, 8, 2)
])
],
runs=[
TestRun("test",
model="test",
testCorpora=[("test", "test")],
engineOptions={
"steps": "zone,tokenize,tag",
"workflow": "Demo"
})
])
e.run()
# OK, this works. There should be a subdirectory for each model.
self.assertTrue(len(e.getModel("test").allInstances) == 3)
for m in e.getModel("test").allInstances:
self.assertTrue(os.path.exists(os.path.join(m.modelDir, "model")))
self.assertEqual(len(m.trainingSet.getFiles()), 8)
self.assertEqual([
m.engineSettings["max_iterations"]
for m in e.getModel("test").allInstances
], [4, 6, 8])
self.assertEqual(
[m.modelSubdir for m in e.getModel("test").allInstances], [
"test_max_iterations_4", "test_max_iterations_6",
"test_max_iterations_8"
])
files = e.corporaTable["test"].getFiles(partition="test")
self.assertEqual(len(files), 2)
self.assertTrue(len(e.runTable["test"].allInstances) == 3)
for r in e.runTable["test"].allInstances:
runDir = r.runDir
for file in files:
self.assertTrue(
os.path.exists(
os.path.join(runDir, "hyp", os.path.basename(file)) +
".prepped.tag.json"))
self.assertTrue(
os.path.exists(
os.path.join(runDir, "run_input", os.path.basename(
file)) + ".prepped"))
self.assertEqual([
os.path.basename(r.runDir) for r in e.runTable["test"].allInstances
], [
"test_max_iterations_4", "test_max_iterations_6",
"test_max_iterations_8"
])