def create_navigation(self, navigation_html, experiment_folder,
experiment_path, data):
if experiment_folder != "":
if os.path.exists(experiment_path + "/custom_flexp_chain.py"):
try:
custom_flexp_chain = import_by_filename(
'custom_flexp_chain',
experiment_path + "/custom_flexp_chain.py")
html_chain = custom_flexp_chain.get_chain()
html_chain = Chain(html_chain)
except:
html_chain = Chain([
StringToHtml(
"<h2>Error processing custom chain. {}</h2>".
format(traceback.format_exc().replace(
"\n", "</br>")),
title="Error in custom chain")
] + self.html_chain.modules)
finally:
if "custom_flexp_chain" in sys.modules:
del sys.modules["custom_flexp_chain"]
else:
html_chain = self.html_chain
html_chain.process(data)
html_chain.close()
navigation_html = html_anchor_navigation(
experiment_path, experiment_folder,
html_chain) + navigation_html
return navigation_html
def test_with(self):
c = Chain([
Add(13),
])
data = {"input": 10}
with c:
c.process(data)
assert data == {"input": 10, "output": 23}
def test_chain_from_fuction(self):
data = {"input": 10}
def add(x):
x["output"] = x["input"] + 13
c = Chain(add)
c.process(data)
self.assertEqual(data, {"input": 10, "output": 23})
assert str(c) == "Chain[add]"
def test_time(self):
data = {"input": 10}
c = Chain([
Add(13),
])
c.process(data)
c.close()
assert len(c.times) == 1
assert c.iterations == 1
c.process(data)
assert c.iterations == 2
def test_chain_inspect_deep(self):
data = {"input": {i: i for i in range(11)}}
with LogCapture() as l:
c = Chain([
inspector.inspect(DummyModule(), stream=True)])
c.process(data)
c.close()
l.check(
('flexp.flow.flow', 'DEBUG', 'DummyModule.process()'),
('flexp.flow.inspector', 'INFO', 'Data flow structure'),
('flexp.flow.inspector', 'INFO', "{\'input\': {\"<class \'int\'>#11 times (0)\": 0}}"),
('flexp.flow.inspector', 'INFO', 'End of data flow structure'),
('flexp.flow.flow', 'INFO', 'DummyModule average execution time 0.00 sec')
)
def test_chain_inspect(self):
data = {"input": 20}
with LogCapture() as l:
c = Chain([
inspector.inspect(Add(10), stream=True)])
c.process(data)
c.close()
l.check(
('flexp.flow.flow', 'DEBUG', 'Add.process()'),
('flexp.flow.inspector', 'INFO', 'Data flow structure'),
('flexp.flow.inspector', 'INFO', "{'input': 20, 'output': 30}"),
('flexp.flow.inspector', 'INFO', 'End of data flow structure'),
('flexp.flow.flow', 'INFO', 'Add average execution time 0.00 sec')
)
def test_chain(self):
data = {"input": 10}
c = Chain([
Add(13),
])
c.process(data)
c.close()
self.assertEqual(data, {"input": 10, "output": 23})
assert str(c) == "Chain[Add]"
c.add(Mult(2))
assert str(c) == "Chain[Add-Mult]"
c.process(data)
self.assertEqual(data, {"input": 10, "output": 20})
def test_chain_from_object(self):
data = {"input": 10}
c = Chain(Add(13))
c.process(data)
self.assertEqual(data, {"input": 10, "output": 23})
assert str(c) == "Chain[Add]"
flexp.describe("Query parameter prediction with TF-IDF and linear regression")
# Setup logging
log.debug("flow setup complete.")
# Create the chain to load the data, lowercase and tokenize it.
file_name = "example_queries.tsv"
data_chain = Chain([
LoadData(file_name),
Lowercase(),
TfIdf(),
])
# data["id"] should contain all info required to replicate the experiment.
data = {"id": file_name}
data_chain.process(data)
data_chain.close()
log.debug("Data chain complete.")
# Train and evaluate a classifier
train_chain = Chain([
# Create data["train"] and data["dev"]
TrainTestSplit(),
# Train our classifier on data["train"]
Train(),
# Evaluate the classifier and store the results in the experiment folder
# using flexp.get_file_path("results.csv")
Eval(),
])
train_chain.process(data)
train_chain.close()
def get(self):
experiment_folder = self.get_argument("experiment", default="")
experiment_path = path.join(self.experiments_folder, experiment_folder)
if not path.isdir(experiment_path):
experiment_folder = ""
navigation_html = html_navigation(self.experiments_folder, experiment_folder)
header_html = ""
scripts_html = ""
if experiment_folder != "":
data = {"experiment_path": experiment_path,
"experiment_folder": experiment_folder,
"html": [],
"header": dict(),
"scripts": dict()
}
# Use custom chain, if present
if os.path.exists(experiment_path + "/custom_flexp_chain.py"):
try:
custom_flexp_chain = import_by_filename('custom_flexp_chain',
experiment_path + "/custom_flexp_chain.py")
html_chain = custom_flexp_chain.get_chain()
html_chain = Chain(html_chain)
except:
html_chain = Chain([StringToHtml("<h2>Error processing custom chain. {}</h2>"
.format(traceback.format_exc().replace("\n", "</br>")),
title="Error in custom chain")] + self.html_chain.modules)
finally:
if "custom_flexp_chain" in sys.modules:
del sys.modules["custom_flexp_chain"]
else:
html_chain = self.html_chain
html_chain.process(data)
html_chain.close()
title_html = "<h1>{}</h1>".format(experiment_folder)
content_html = u"\n".join(data['html'])
navigation_html = html_anchor_navigation(
experiment_path, experiment_folder, html_chain) + navigation_html
header_html = u"\n".join(u"\n".join(html_lines)
for head_section, html_lines
in data["header"].items())
scripts_html = u"\n".join(u"\n".join(script_lines)
for script_section, script_lines
in data["scripts"].items())
else:
title_html = "<h1>Experiments</h1>"
content_html = html_table(self.experiments_folder, self.get_metrics_fcn, self.metrics_file)
html = self._template.format(
title=title_html,
navigation=navigation_html,
content=content_html,
header=header_html,
scripts=scripts_html)
self.write(html)