Model Config¶
-
class
ModelConfig
(model_json, base_dir=None, labels=None, colors=None)¶ The model configuration parameters.
- Parameters
model_json (
dict
) – The parsed alwaysai.model.json.labels (
Optional
[List
[str
]]) – The label list for the model.colors (
Optional
[ndarray
]) – The color list for the model.
-
classmethod
from_model_id
(model_id)¶ - Return type
ModelConfig
-
property
config
¶ The config loaded from the model JSON file
- Return type
dict
-
property
model_parameters
¶ The model parameters in the config
- Return type
dict
-
property
id
¶ The model ID
- Return type
str
-
property
label_file
¶ Path to the label file
- Return type
Optional
[str
]
-
property
colors_file
¶ Path to the colors file
- Return type
Optional
[str
]
-
property
model_file
¶ Path to the model weights file
- Return type
str
-
property
config_file
¶ Relative path to the model framework config file
- Return type
Optional
[str
]
-
property
mean
¶ The RGB/BGR mean values for the model
- Return type
Tuple
[float
,float
,float
]
-
property
scalefactor
¶ The scale factor for the model input
- Return type
float
-
property
size
¶ The input image size of the model
- Return type
Tuple
[int
,int
]
-
property
purpose
¶ The purpose of the model
- Return type
SupportedPurposes
-
property
framework_type
¶ The framework type of the model
- Return type
str
-
property
crop
¶ Whether or not to crop the image prior to inferencing
- Return type
bool
-
property
colors_dtype
¶ The data type of the color values
- Return type
str
-
property
labels
¶ The labels of the model
- Return type
Optional
[List
[str
]]
-
property
colors
¶ The colors for each label of the model.
Each array element is a 3 dimensional array of 8 bit integers representing red, green, and blue
- Return type
Optional
[ndarray
]
-
property
swaprb
¶ Whether to swap the red and blue channels of the image prior to inference
- Return type
bool
-
property
architecture
¶ The architecture of the model
- Return type
Optional
[str
]
-
property
softmax
¶ Whether to perform softmax after the inference
- Return type
bool
-
property
device
¶ The device the model was built for
- Return type
Optional
[SupportedDevices
]
-
property
output_layer_names
¶ The output layer names of the model
- Return type
Optional
[List
[str
]]
-
property
hailo_quantize_input
¶ Whether to quantize input of Hailo model
- Return type
Optional
[bool
]
-
property
hailo_quantize_output
¶ Whether to quantize output of Hailo model
- Return type
Optional
[bool
]
-
property
hailo_input_format
¶ Input format for Hailo model
- Return type
Optional
[str
]
-
property
hailo_output_format
¶ Output format of Hailo model
- Return type
Optional
[str
]
-
property
dnn_support
¶ Whether DNN Engine supports the model
- Return type
bool
-
property
dnn_cuda_support
¶ Whether DNN CUDA Engine supports the model
- Return type
bool
-
property
tensor_rt_support
¶ Whether TensorRT Engine supports the model
- Return type
bool
-
property
hailo_support
¶ Whether Hailo RT Engine supports the model
- Return type
bool
-
property
qaic_support
¶ Whether QAIC RT Engine supports the model
- Return type
bool
-
property
onnx_rt_support
¶ Whether ONNX RT Engine supports the model
- Return type
bool
-
property
pytorch_support
¶ Whether PYTORCH Engine supports the model
- Return type
bool
-
property
batch_size
¶ Inference batch size of model
- Return type
int
-
property
hub_repo
¶ Torch hub repo
- Return type
Optional
[str
]
-
property
hub_model
¶ Torch hub model
- Return type
Optional
[str
]
-
property
hub_pretrained
¶ Torch pretrained model
- Return type
Optional
[bool
]
-
property
hub_force_reload
¶ Torch force reload the model
- Return type
Optional
[bool
]