def test_package(self):
with mkvenv() as tempdir:
with chdir(tempdir):
subprocess.check_call(["sh", filepath("create.sh")])
subprocess.check_call(["sh", filepath("cd.sh")])
cd_path = pathlib.Path(filepath("cd.sh"))
pkg = cd_path.read_text().strip().split()[-1]
with chdir(pkg):
# Check that the imports header is correct
check_imports_path = pathlib.Path(
filepath("test_model_imports.txt")
)
check_imports = check_imports_path.read_text().strip()
real_imports_path = pathlib.Path("tests", "test_model.py")
real_imports = real_imports_path.read_text().strip()
self.assertTrue(real_imports.startswith(check_imports))
# Run the pip install
subprocess.check_call(["sh", filepath("pip_install.sh")])
# Run the tests
subprocess.check_call(["sh", filepath("unittest.sh")])
subprocess.check_call(
["sh", filepath("unittest_logging.sh")]
)
# Try training
subprocess.check_call(
[
"sh",
str(
pathlib.Path(__file__).parent.parent
/ "slr"
/ "train_data.sh"
),
]
)
subprocess.check_call(["sh", filepath("train.sh")])
def directory_with_csv_files():
with tempfile.TemporaryDirectory() as tempdir:
with chdir(tempdir):
subprocess.check_output(["sh", sh_filepath("train_data.sh")])
subprocess.check_output(["sh", sh_filepath("test_data.sh")])
subprocess.check_output(["sh", sh_filepath("predict_data.sh")])
yield tempdir
def directory_with_data_files():
with tempfile.TemporaryDirectory() as tempdir:
with chdir(tempdir):
# Replace Yann's website with GitHub mirror for stability
image_data_sh_path = pathlib.Path(sh_filepath("image_data.sh"))
image_data_sh_contents = image_data_sh_path.read_text().split("\n")
# First line is the curl command
curl = image_data_sh_contents[0].split()[:-1]
# Make 4 curl commands and then append the sha validation
image_data_sh_github_mirror_path = pathlib.Path(
tempdir, "image_data.sh")
image_data_sh_github_mirror_path.write_text("\n".join([
" ".join(curl + [url]) for url in [
"https://github.com/intel/dffml/files/4283897/train-labels-idx1-ubyte.gz",
"https://github.com/intel/dffml/files/4283898/train-images-idx3-ubyte.gz",
"https://github.com/intel/dffml/files/6138929/t10k-labels-idx1-ubyte.gz",
"https://github.com/intel/dffml/files/6138930/t10k-images-idx3-ubyte.gz",
]
] + image_data_sh_contents[1:], ))
subprocess.check_output(
["bash", str(image_data_sh_github_mirror_path)])
subprocess.check_output(["bash", sh_filepath("image_file.sh")])
for image in pathlib.Path(__file__).parent.glob("*.png"):
shutil.copy(str(image.absolute()), image.name)
yield tempdir
def directory_with_data_files():
with tempfile.TemporaryDirectory() as tempdir:
with chdir(tempdir):
subprocess.check_output(["bash", sh_filepath("image_data.sh")])
subprocess.check_output(["bash", sh_filepath("image_file.sh")])
for image in pathlib.Path(__file__).parent.glob("*.png"):
shutil.copy(str(image.absolute()), image.name)
yield tempdir
async def make_dataflow(ops, operations, seed):
# Create temp dir and write op to ops.py
with tempfile.TemporaryDirectory() as tmpdirname:
# Change directory into the tempdir
with chdir(tmpdirname):
# Write out op to op.py
pathlib.Path(tmpdirname, "ops.py").write_text(ops)
# Reload conents
sys.path.insert(0, tmpdirname)
module = importlib.import_module("ops")
importlib.reload(module)
sys.path.pop(0)
# $ dffml dataflow create $operations -seed $seed
with io.StringIO() as dataflow:
with contextlib.redirect_stdout(dataflow):
await CLI.cli("dataflow", "create", *operations,
"-seed", *seed)
yield DataFlow._fromdict(**json.loads(dataflow.getvalue()))
async def test_create_from_path(self):
# Create temp dir and write op to ops.py
with tempfile.TemporaryDirectory() as tmpdirname:
# Change directory into the tempdir
with chdir(tmpdirname):
# Write out op to op.py
operation_file_path = pathlib.Path(tmpdirname, "ops.py")
operation_file_path.write_text(OP_DEF_STRING)
# We make the name the path relative to our cwd
operation_qualname = "ops:echo_string"
dataflow_file_path = pathlib.Path(tmpdirname, "dataflow.json")
# $ dffml dataflow create \
# ops:echo_string get_single
with io.StringIO() as dataflow:
with contextlib.redirect_stdout(dataflow):
await CLI.cli(
"dataflow",
"create",
*[operation_qualname, "get_single"],
"-seed",
'["OutputString"]=get_single_spec',
)
test_dataflow = DataFlow._fromdict(
**json.loads(dataflow.getvalue()))
# Make sure the operation is in the dataflow
self.assertIn(operation_qualname, test_dataflow.operations)
# Run the dataflow
async for ctx_str, results in run(
test_dataflow,
[
Input(
value="Irregular at magic school",
definition=test_dataflow.operations[
operation_qualname].inputs["input_string"],
)
],
):
self.assertIn("OutputString", results)
self.assertEqual(
results["OutputString"],
"Irregular at magic school",
)
def directory_with_csv_files():
with tempfile.TemporaryDirectory() as tempdir:
with chdir(tempdir):
subprocess.check_output(["bash", sh_filepath("dataset.sh")])
yield tempdir
