query_available_dl_devices — Get list of deep-learning-capable hardware devices.
query_available_dl_devices( : : GenParamName, GenParamValue : DLDeviceHandles)
query_available_dl_devices returns a list of handles. Each
handle refers to a deep-learning-capable hardware device
(hereafter referred to as device) that can be
used for inference or training of a deep learning model. For each
returned device, every parameter mentioned in GenParamName
must be equal to at least one of its corresponding values that appear
in GenParamValue. A parameter can have more than one value
by duplicating its name in GenParamName and adding different
corresponding value in GenParamValue.
A deep-learning-capable device is either supported directly through HALCON or through an AI 2-interface.
The devices that are supported directly through HALCON are equivalent
to those that can be set to a deep learning model via
set_dl_model_param using 'runtime' = 'cpu' or
'runtime' = 'gpu'. HALCON provides an internal
implementation for the inference or training of a deep learning model
for those devices. See Deep Learning for more details.
Devices that are supported through the AI
2-interface
can also be set to a deep learning model using set_dl_model_param.
In this case the inference is not executed by HALCON but by the device
itself.
query_available_dl_devices returns a handle for each
deep-learning-capable device supported through HALCON and through an
inference engine.
If a device is supported through HALCON and one or several inference
engines, query_available_dl_devices returns a handle for HALCON
and for each inference engine.
GenParamName can be used to filter for the devices.
All GenParamName that are gettable by get_dl_device_param
and that do not return a handle-typed value for GenParamValue
are supported for filtering. See the operator reference of
get_dl_device_param for the list of gettable parameters.
In addition, the following values are supported:
The devices, which are directly supported by HALCON for this device type.
Possible values (standard): 'cpu', 'gpu'.
GenParamName can have multiple entries for the same value.
In this case filter combines the entries with a logical 'or'.
Please see the example code below for some examples how to use the
filter.
This operator returns a handle. Note that the state of an instance of this handle type may be changed by specific operators even though the handle is used as an input parameter by those operators.
GenParamName (input_control) attribute.name(-array) → (string)
Name of the generic parameter.
Default: []
List of values: 'ai_accelerator_interface', 'calibration_precisions', 'cast_precisions', 'conversion_supported', 'id', 'inference_only', 'name', 'optimize_for_inference_params', 'precisions', 'runtime', 'settable_device_params', 'type'
GenParamValue (input_control) attribute.value(-array) → (string / integer / real)
Value of the generic parameter.
Default: []
Suggested values:
DLDeviceHandles (output_control) dl_device(-array) → (handle)
Tuple of DLDevice handles
* Query all deep-learning-capable hardware devices
query_available_dl_devices ([], [], DLDeviceHandles)
* Query all GPUs with ID 0 or 2
query_available_dl_devices (['type', 'id', 'id'], ['gpu', 0, 2],\
DLDeviceHandles)
* Query the unique GPU with ID 1 supported by HALCON
query_available_dl_devices (['runtime', 'id'], ['gpu', 1], DLDeviceHandles)
If the parameters are valid, the operator
query_available_dl_devices returns the value 2 (
H_MSG_TRUE)
.
If necessary, an exception is raised.
Foundation