fedn.network.storage.models package
The models package handles storing of model updates durign the federated training process.
The functionality is used by the combiner service during aggregation of model updates from clients.
By implementing the interface in the base class modelstorage.py, a developer may customize the
behaviour of the framework.
fedn.network.storage.models.memorymodelstorage module
-
class fedn.network.storage.models.memorymodelstorage.MemoryModelStorage[source]
Bases: ModelStorage
Class for in-memory storage of model artifacts.
Models are stored as BytesIO objects in a dictionary.
-
exist(model_id)[source]
Check if model exists in storage
- Parameters:
model_id (str) – The model id
- Returns:
True if model exists, False otherwise
- Return type:
bool
-
get(model_id)[source]
Get model from storage
- Parameters:
model_id (str) – The model id
- Returns:
The model
- Return type:
object
-
get_model_metadata(model_id)[source]
Get model metadata from storage
- Parameters:
model_id (str) – The model id
- Returns:
The model metadata
- Return type:
dict
-
get_ptr(model_id)[source]
- Parameters:
model_id
- Returns:
-
-
set_model_metadata(model_id, model_metadata)[source]
Set model metadata in storage
- Parameters:
-
- Returns:
True if successful, False otherwise
- Return type:
bool
fedn.network.storage.models.modelstorage module
-
class fedn.network.storage.models.modelstorage.ModelStorage[source]
Bases: ABC
-
abstract delete(model_id)[source]
Delete model from storage
- Parameters:
model_id (str) – The model id
- Returns:
True if successful, False otherwise
- Return type:
bool
-
abstract delete_all()[source]
Delete all models from storage
- Returns:
True if successful, False otherwise
- Return type:
bool
-
abstract exist(model_id)[source]
Check if model exists in storage
- Parameters:
model_id (str) – The model id
- Returns:
True if model exists, False otherwise
- Return type:
bool
-
abstract get(model_id)[source]
Get model from storage
- Parameters:
model_id (str) – The model id
- Returns:
The model
- Return type:
object
-
abstract get_model_metadata(model_id)[source]
Get model metadata from storage
- Parameters:
model_id (str) – The model id
- Returns:
The model metadata
- Return type:
dict
-
abstract set_model_metadata(model_id, model_metadata)[source]
Set model metadata in storage
- Parameters:
-
- Returns:
True if successful, False otherwise
- Return type:
bool
fedn.network.storage.models.tempmodelstorage module
-
class fedn.network.storage.models.tempmodelstorage.TempModelStorage[source]
Bases: ModelStorage
Class for managing local temporary models on file on combiners.
-
delete(model_id)[source]
Delete model from storage
- Parameters:
model_id (str) – The model id
- Returns:
True if successful, False otherwise
- Return type:
bool
-
delete_all()[source]
Delete all models from storage
- Returns:
True if successful, False otherwise
- Return type:
bool
-
exist(model_id)[source]
Check if model exists in storage
- Parameters:
model_id (str) – The model id
- Returns:
True if model exists, False otherwise
- Return type:
bool
-
get(model_id)[source]
Get model from storage
- Parameters:
model_id (str) – The model id
- Returns:
The model
- Return type:
object
-
get_model_metadata(model_id)[source]
Get model metadata from storage
- Parameters:
model_id (str) – The model id
- Returns:
The model metadata
- Return type:
dict
-
get_ptr(model_id)[source]
- Parameters:
model_id
- Returns:
-
-
set_model_metadata(model_id, model_metadata)[source]
Set model metadata in storage
- Parameters:
-
- Returns:
True if successful, False otherwise
- Return type:
bool