def test_eval_model_zip(bioimageio_model_bytes):
with ZipFile(bioimageio_model_bytes) as zf:
pipeline = eval_model_zip(zf, devices=["cpu"], preserve_batch_dim=True)
assert pipeline.input_axes == "bcyx"
assert pipeline.output_axes == "bcyx"
assert pipeline.input_shape == [("b", 1), ("c", 1), ("y", 512), ("x", 512)]
assert pipeline.halo == [("b", 0), ("c", 0), ("y", 32), ("x", 32)]
def test_eval_onnx_model_zip_predict(bioimageio_unet2d_onnx_bytes,
bioimageio_unet2d_onnx_test_data):
with ZipFile(bioimageio_unet2d_onnx_bytes) as zf:
adapter = eval_model_zip(zf, devices=["cpu"], preserve_batch_dim=True)
test_input = xarray.DataArray(np.load(
bioimageio_unet2d_onnx_test_data["test_input"]),
dims=("b", "c", "x", "y"))
# TODO: Figure out why test output doesn't match result
adapter.forward(test_input)
def test_eval_model_zip_metadata_with_batch_dim(
bioimageio_unet2d_torchscript_bytes,
bioimageio_unet2d_torchscript_test_data):
with ZipFile(bioimageio_unet2d_torchscript_bytes) as zf:
pipeline = eval_model_zip(zf, devices=["cpu"], preserve_batch_dim=True)
assert pipeline.input_axes == "bcyx"
assert pipeline.output_axes == "bcyx"
assert pipeline.input_shape == [("b", 1), ("c", 1), ("y", 512),
("x", 512)]
assert pipeline.halo == [("b", 0), ("c", 0), ("y", 32), ("x", 32)]
def __init__(self, model_zip: bytes, devices: List[str]) -> None:
cache_path = os.getenv("PYBIO_CACHE_PATH", None)
if cache_path is not None:
cache_path = Path(cache_path)
with zipfile.ZipFile(io.BytesIO(model_zip)) as model_file:
self._model = eval_model_zip(model_file, devices, cache_path=cache_path)
self._datasets = {}
self._worker = base.SessionBackend(self._model)
def test_eval_tensorflow_model_zip(pybio_dummy_tensorflow_model_bytes, cache_path):
with ZipFile(pybio_dummy_tensorflow_model_bytes) as zf:
pipeline = eval_model_zip(zf, devices=["cpu"], cache_path=cache_path)
assert isinstance(pipeline, PredictionPipeline)
test_input = xr.DataArray(np.zeros(shape=(1, 128, 128)), dims=("c", "y", "x"))
out_arr = np.ones(shape=(1, 128, 128))
out_arr.fill(42)
test_output = xr.DataArray(out_arr, dims=("c", "y", "x"))
result = pipeline.forward(test_input)
xr.testing.assert_equal(result, test_output)
def test_eval_model_zip_metadata_no_batch_dim(
pybio_unet2d_torchscript_bytes, pybio_unet2d_torchscript_test_data,
cache_path):
with ZipFile(pybio_unet2d_torchscript_bytes) as zf:
pipeline = eval_model_zip(zf,
devices=["cpu"],
cache_path=cache_path,
preserve_batch_dim=False)
assert pipeline.name == "UNet 2D Nuclei Broad"
assert pipeline.input_axes == "cyx"
assert pipeline.output_axes == "cyx"
assert pipeline.input_shape == [("c", 1), ("y", 512), ("x", 512)]
assert pipeline.halo == [("c", 0), ("y", 32), ("x", 32)]
def test_eval_torchscript_model_zip_predict(
bioimageio_unet2d_torchscript_bytes,
bioimageio_unet2d_torchscript_test_data):
with ZipFile(bioimageio_unet2d_torchscript_bytes) as zf:
pipeline = eval_model_zip(zf, devices=["cpu"], preserve_batch_dim=True)
test_input = xarray.DataArray(np.load(
bioimageio_unet2d_torchscript_test_data["test_input"]).astype(
np.float32),
dims=("b", "c", "x", "y"))
test_output = np.load(
bioimageio_unet2d_torchscript_test_data["test_output"])
result = pipeline.forward(test_input)
assert_array_almost_equal(result.data, test_output, decimal=4)
def test_eval_onnx_model_zip(pybio_unet2d_onnx_bytes, cache_path):
with ZipFile(pybio_unet2d_onnx_bytes) as zf:
adapter = eval_model_zip(zf, devices=["cpu"], cache_path=cache_path)
assert isinstance(adapter, PredictionPipeline)
def test_eval_model_zip(pybio_model_bytes, cache_path):
with ZipFile(pybio_model_bytes) as zf:
exemplum = eval_model_zip(zf, devices=["cpu"], cache_path=cache_path)
assert isinstance(exemplum, PredictionPipeline)
def test_eval_tensorflow_model_zip(pybio_dummy_tensorflow_model_bytes,
cache_path):
with ZipFile(pybio_dummy_tensorflow_model_bytes) as zf:
exemplum = eval_model_zip(zf, devices=["cpu"], cache_path=cache_path)
assert isinstance(exemplum, ModelAdapter)
def __init__(self, model_zip: bytes, devices: List[str]) -> None:
with zipfile.ZipFile(io.BytesIO(model_zip)) as model_file:
self._model = eval_model_zip(model_file, devices)
self._datasets = {}
self._worker = base.SessionBackend(self._model)
def test_eval_onnx_model_zip(bioimageio_unet2d_onnx_bytes):
with ZipFile(bioimageio_unet2d_onnx_bytes) as zf:
adapter = eval_model_zip(zf, devices=["cpu"])
assert isinstance(adapter, PredictionPipeline)
def test_eval_model_zip(bioimageio_model_bytes):
with ZipFile(bioimageio_model_bytes) as zf:
exemplum = eval_model_zip(zf, devices=["cpu"])
assert isinstance(exemplum, PredictionPipeline)