hoops_ai.storage.MLStorage

class hoops_ai.storage.MLStorage

Bases: ABC

abstract append_edata(edge_features, feature_name, torch_type=torch.float)

Appends an edge feature to the graph.

Parameters:

  • edge_features (np.ndarray) – The edge feature data as a NumPy array.

  • feature_name (str) – The name to assign to the feature in the graph’s edge data.

  • torch_type (torch.dtype, optional) – The PyTorch data type for the tensor. Defaults to torch.float.

Raises:

  • ValueError – If the graph has not been initialized.

  • TypeError – If edge_features is not a numpy.ndarray.

Return type:

None

abstract append_extra_data(extra_data, feature_name, torch_type=torch.float)

Appends extra data to the graph’s internal dictionary.

Parameters:

  • extra_data (np.ndarray) – The extra feature data as a NumPy array.

  • feature_name (str) – The name to assign to the feature in the internal dictionary.

  • torch_type (torch.dtype, optional) – The PyTorch data type for the tensor. Defaults to torch.float.

Raises:

TypeError – If extra_data is not a numpy.ndarray.

Return type:

None

abstract append_ndata(node_features, feature_name, torch_type=torch.float)

Appends a node feature to the graph.

Parameters:

  • node_features (np.ndarray) – The node feature data as a NumPy array.

  • feature_name (str) – The name to assign to the feature in the graph’s node data.

  • torch_type (torch.dtype, optional) – The PyTorch data type for the tensor. Defaults to torch.float.

Raises:

  • ValueError – If the graph has not been initialized.

  • TypeError – If node_features is not a numpy.ndarray.

Return type:

None

abstract concatenate_data(target_dgl_tensor, source_dgl_tensor)

Concatenates the source_dgl_tensor to the target_dgl_tensor along the last dimension.

Parameters:

  • target_dgl_tensor (torch.Tensor) – The tensor to append to.

  • source_dgl_tensor (torch.Tensor) – The tensor to append from.

Returns:

The concatenated tensor.

Return type:

torch.Tensor

Raises:

TypeError – If either target_dgl_tensor or source_dgl_tensor is not a torch.Tensor.

format()

a string specifying the concrete format for this storage.

Return type:

str

abstract load_graph(filepath)

Loads a graph from the specified filepath.

Parameters:

filepath (str) – Path to load the graph from.

Returns:

The loaded graph object.

Return type:

Any

Raises:

ValueError – If the graph cannot be loaded.

abstract save_graph(filepath)

Saves the graph to the specified filepath.

Parameters:

filepath (str) – Path to save the graph.

Raises:

ValueError – If the graph has not been initialized.

Return type:

int

abstract setup_graph(source_nodes, target_nodes, num_nodes)

Creates a graph using separate lists of source and target nodes and the total number of nodes.

Parameters:

  • source_nodes (List[int]) – List of source node indices.

  • target_nodes (List[int]) – List of target node indices.

  • num_nodes (int) – Total number of nodes in the graph.

Raises:

ValueError – If source_nodes and target_nodes do not have the same length.

Return type:

None