model_routines
This modules defines the interface which must be implemented in order to utilize a pretrained model and its weights
PretrainedModelRoutines
Bases: ABC
Set of routines that are required in order to use a pretrained model for nst.
Source code in src/artificial_artwork/pretrained_model/model_routines.py
10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 |
|
get_id(layer)
abstractmethod
Get the id of a model's network layer.
The pretrained model being a neural network has a specific architecture and each layer should a unique string id that one can reference it.
Parameters:
Name | Type | Description | Default |
---|---|---|---|
layer |
NDArray
|
the layer of a pretrained neural network model |
required |
Returns:
Name | Type | Description |
---|---|---|
str |
str
|
the layer id |
Source code in src/artificial_artwork/pretrained_model/model_routines.py
28 29 30 31 32 33 34 35 36 37 38 39 40 41 |
|
get_layers_dict(layers)
abstractmethod
Get a dict mapping strings to pretrained model layers.
Parameters:
Name | Type | Description | Default |
---|---|---|---|
layers |
NDArray
|
the pretrained model layers |
required |
Returns:
Type | Description |
---|---|
Dict[str, NDArray]
|
Dict[str, NDArray]: the dictionary mapping strings to layers |
Source code in src/artificial_artwork/pretrained_model/model_routines.py
43 44 45 46 47 48 49 50 51 52 53 |
|
get_weights(layer)
abstractmethod
Get the values of the weights of a given network layer.
Each pretrained model network layer has "learned" certain parameters in the form of "weights" (ie weight matrices A and b in equation Ax + b).
Call this method to get a tuple of the A and b mathematical matrices.
Parameters:
Name | Type | Description | Default |
---|---|---|---|
layer |
NDArray
|
the layer of a pretrained neural network model |
required |
Returns:
Type | Description |
---|---|
Tuple[NDArray, NDArray]
|
Tuple[NDArray, NDArray]: the weights in matrix A and b |
Source code in src/artificial_artwork/pretrained_model/model_routines.py
55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 |
|
load_layers(file_path)
abstractmethod
Load a pretrained model from disk.
Loads the model parameters, given the path to a file in the disk, that indicated where the pretrained model is.
Parameters:
Name | Type | Description | Default |
---|---|---|---|
file_path |
str
|
the path corresponding to a file in the disk |
required |
Returns:
Name | Type | Description |
---|---|---|
NDArray |
NDArray
|
the model parameters as a numpy array |
Source code in src/artificial_artwork/pretrained_model/model_routines.py
13 14 15 16 17 18 19 20 21 22 23 24 25 26 |
|