Class for the PXD background generator module. More...

Inheritance diagram for PXDBackgroundGenerator:

Public Member Functions
	__init__ (self, str model="convnet", Union[None, str, pathlib.Path] checkpoint=None, Union[None, int] seed=None, int nintra=1, int ninter=1, str globaltag="PXDBackgroundGenerator")
	Constructor for the PXD background generator module.

	initialize (self)
	Method called before event processing to initialize the module.

	event (self)
	Method called each time an event is processed.

Public Attributes
	model = _verify_model(model)
	Name of the generator model.

	checkpoint = _verify_checkpoint(checkpoint)
	Path to the checkpoint file with the pre-trained model weights.

	seed = _verify_seed(seed)
	Integer number in the interval $[-2^{63}, 2^{63} - 1]$ set as the initial seed.

	nintra = _verify_nintra(nintra)
	Number of intra-op threads utilized.

	ninter = _verify_ninter(ninter)
	Number of inter-op threads utilized.

	globaltag = _verify_globaltag(globaltag)
	Global tag of the conditions database providing the default checkpoints stored as payloads.

Protected Attributes
	_generate_func = _get_generate_func(self.model)
	Generation function applied on the model instance to return an output that is transcoded into digits.

	_generator = _get_model_cls(self.model)()
	Generator model instance.

	_vxdids = tuple(VxdID(*arg) for arg in VXDID_ARGS)
	Sequence of identifier objects for each PXD module.

	_pystorearray = PyStoreArray("PXDDigits", DataStore.c_DontWriteOut)
	Accessor for PXD background digits in the data store.

Detailed Description

Class for the PXD background generator module.

Generates PXD background samples for background overlay on the fly.

:param model: Name of the generator model to use - either "convnet" or "resnet",
    defaults to "convnet" (optional)
:type model: str

:param checkpoint: Path to the checkpoint file with weights for the selected model,
    defaults to None - use the default checkpoint from the conditions database (optional)
:type checkpoint: str

:param seed: Integer number in the interval :math:`[-2^{63}, 2^{63} - 1]`
    used internally as the initial seed,
    defaults to None - derive a deterministic seed from the
    value returned by :py:func:`basf2.get_random_seed` (optional)
:type seed: int

:param nintra: Number of intra-op threads to be utilized for the generation,
    defaults to 1 (optional)
:type nintra: int

:param ninter: Number of inter-op threads to be utilized for the generation,
    defaults to 1 (optional)
:type ninter: int

:param globaltag: Global tag of the conditions database
    providing the default checkpoints stored as payloads,
    defaults to "PXDBackgroundGenerator" (optional)
:type globaltag: str

Definition at line 138 of file __init__.py.

Constructor & Destructor Documentation

◆ init()

__init__	(		self,
		str	model = "convnet",
		Union[None, str, pathlib.Path]	checkpoint = None,
		Union[None, int]	seed = None,
		int	nintra = 1,
		int	ninter = 1,
		str	globaltag = "PXDBackgroundGenerator" )

Constructor for the PXD background generator module.

Parameters

model	Name of the generator model to use - either "convnet" or "resnet", defaults to "convnet"
checkpoint	Path to the checkpoint file with weights for the selected model, defaults to None - use the default checkpoint from the conditions database
seed	Integer number in the interval $[-2^{63}, 2^{63} - 1]$ used internally as the initial seed, defaults to None - derive a deterministic seed from the value returned by `basf2.get_random_seed()`
nintra	Number of intra-op threads to be utilized for the generation, defaults to 1
ninter	Number of inter-op threads to be utilized for the generation, defaults to 1
globaltag	Global tag of the conditions database providing the default checkpoints stored as payloads, defaults to "PXDBackgroundGenerator"

Definition at line 192 of file __init__.py.

    ):
        super().__init__()
        # process `model`
        self.model = _verify_model(model)
 
        # process `checkpoint`
        self.checkpoint = _verify_checkpoint(checkpoint)
 
        # process `seed`
        self.seed = _verify_seed(seed)
 
        # process `nintra`
        self.nintra = _verify_nintra(nintra)
 
        # process `ninter`
        self.ninter = _verify_ninter(ninter)
 
        # process `globaltag`
        self.globaltag = _verify_globaltag(globaltag)
 
        # enable the specified global tag
        basf2.conditions.append_globaltag(self.globaltag)
 

Member Function Documentation

◆ event()

event ( self )

Method called each time an event is processed.

Definition at line 288 of file __init__.py.

    def event(self):
        # get the low-level array accessor
        digit_array = self._pystorearray.getPtr()
 
        # clear digits stored by BGOverlayInput
        digit_array.Clear()
 
        # generate a batch of images - one image for each PXD module
        x = self._generate_func(self._generator)
 
        # locate indices of pixels with non-zero values - pixel hits
        nonzero = x.nonzero(as_tuple=True)
        args = [indices.tolist() for indices in nonzero]
        vals = x[nonzero].tolist()
 
        # store indices and pixel values into the data store as background digits
        for n, (idx, ucell, vcell, charge) in enumerate(zip(*args, vals)):
            # append a new default digit to expand the array
            self._pystorearray.appendNew()
            # modify the array to point to the correct digit
            digit_array[n] = PXDDigit(self._vxdids[idx], ucell, vcell, charge)
 
        # delete references to release memory
        del x, nonzero, args, vals
 

◆ initialize()

initialize ( self )

Method called before event processing to initialize the module.

Definition at line 225 of file __init__.py.

    def initialize(self):
        try:
            import torch
        except ImportError as exc:
            exc.msg = "please install PyTorch: `pip3 install torch==1.4.0`"
            raise
 
        # set the number of inter-op CPU threads
        torch.set_num_interop_threads(self.ninter)
 
        # set the number of intra-op CPU threads
        torch.set_num_threads(self.nintra)
 
        # get the generation function for the selected model
        self._generate_func = _get_generate_func(self.model)
 
        # instantiate the generator model
        self._generator = _get_model_cls(self.model)()
 
        # initialize the model weights
        checkpoint = self.checkpoint
        if self.checkpoint is None:
            # use the default checkpoint from the conditions database
            payload = f"PXDBackgroundGenerator_{self.model}"
            accessor = DBAccessorBase(DBStoreEntry.c_RawFile, payload, True)
            checkpoint = accessor.getFilename()
        self._generator.load_state_dict(torch.load(checkpoint, map_location="cpu"))
 
        # set mode of operation to inference
        self._generator.eval()
 
        # disable the computation of gradients
        for param in self._generator.parameters():
            param.requires_grad = False
 
        # initialize the seed
        seed = self.seed
        if seed is None:
            # derive from the basf2 seed value
            obj = basf2.get_random_seed()
            func = hashlib.sha512()
            func.update(bytes(str(obj), "utf-8"))
            digest = func.digest()[:8]
            seed = int.from_bytes(digest, "big", signed=True)
        basf2.B2INFO(f"PXD background generator seed initialized to {seed}")
        torch.manual_seed(seed)
 
        # instantiate objects for specifying distinct PXD modules
        self._vxdids = tuple(VxdID(*arg) for arg in VXDID_ARGS)
 
        # get the name of the extension used by BGOverlayInput for background collections
        bkginfo = PyStoreObj("BackgroundInfo", DataStore.c_Persistent)
        if not bkginfo.isValid():
            # no information about background overlay input available
            basf2.B2ERROR("path must contain the BGOverlayInput module")
        extension = bkginfo.getExtensionName()
 
        # register the PXD background digit collection - array - in the data store
        self._pystorearray = PyStoreArray("PXDDigits", DataStore.c_DontWriteOut)
        self._pystorearray.registerInDataStore(f"PXDDigits{extension}")
 

Member Data Documentation

◆ _generate_func

_generate_func = _get_generate_func(self.model)

protected

Generation function applied on the model instance to return an output that is transcoded into digits.

Definition at line 239 of file __init__.py.

◆ _generator

_generator = _get_model_cls(self.model)()

protected

Generator model instance.

Definition at line 242 of file __init__.py.

◆ _pystorearray

_pystorearray = PyStoreArray("PXDDigits", DataStore.c_DontWriteOut)

protected

Accessor for PXD background digits in the data store.

Definition at line 283 of file __init__.py.

◆ _vxdids

_vxdids = tuple(VxdID(*arg) for arg in VXDID_ARGS)

protected

Sequence of identifier objects for each PXD module.

Definition at line 273 of file __init__.py.

◆ checkpoint

checkpoint = _verify_checkpoint(checkpoint)

Path to the checkpoint file with the pre-trained model weights.

Definition at line 206 of file __init__.py.

◆ globaltag

globaltag = _verify_globaltag(globaltag)

Global tag of the conditions database providing the default checkpoints stored as payloads.

Definition at line 218 of file __init__.py.

◆ model

model = _verify_model(model)

Name of the generator model.

Definition at line 203 of file __init__.py.

◆ ninter

ninter = _verify_ninter(ninter)

Number of inter-op threads utilized.

Definition at line 215 of file __init__.py.

◆ nintra

nintra = _verify_nintra(nintra)

Number of intra-op threads utilized.

Definition at line 212 of file __init__.py.

◆ seed

seed = _verify_seed(seed)

Integer number in the interval $[-2^{63}, 2^{63} - 1]$ set as the initial seed.

Definition at line 209 of file __init__.py.

The documentation for this class was generated from the following file:

pxd/scripts/pxd/background_generator/__init__.py

Public Member Functions

Public Attributes

Protected Attributes

Detailed Description

Constructor & Destructor Documentation

◆ __init__()

Member Function Documentation

◆ event()

◆ initialize()

Member Data Documentation

◆ _generate_func

◆ _generator

◆ _pystorearray

◆ _vxdids

◆ checkpoint

◆ globaltag

◆ model

◆ ninter

◆ nintra

◆ seed

◆ init()