Metadata Class

class modelindex.models.Metadata.Metadata(flops: Union[str, int] = None, parameters: Union[str, int] = None, epochs: Union[str, int] = None, batch_size: Union[str, int] = None, training_data: [<class 'str'>, typing.List] = None, training_techniques: [<class 'str'>, typing.List] = None, training_resources: str = None, architecture: [<class 'str'>, typing.List] = None, _filepath: str = None, **kwargs)[source]

Metadata for a model.

Parameters

  • flops (str,int) – number of FLOPs

  • parameters (str,int) – total number of parameters for the model

  • epochs (str,int) – how many epochs the model was trained

  • batch_size (str,int) – batch size for the model

  • training_data (str,list) – one or a list of datasets used in training

  • training_techniques (str,list) – one or a list of training techniques

  • training_resources (str) – hardware used to train

  • architecture (str, List) – one or a list of architectures used in the model

  • _filepath (str) – path to the file where the data is coming from

  • **kwargs – any other custom metadata

__init__(flops: Union[str, int] = None, parameters: Union[str, int] = None, epochs: Union[str, int] = None, batch_size: Union[str, int] = None, training_data: [<class 'str'>, typing.List] = None, training_techniques: [<class 'str'>, typing.List] = None, training_resources: str = None, architecture: [<class 'str'>, typing.List] = None, _filepath: str = None, **kwargs)[source]
Parameters

  • flops (str,int) – number of FLOPs

  • parameters (str,int) – total number of parameters for the model

  • epochs (str,int) – how many epochs the model was trained

  • batch_size (str,int) – batch size for the model

  • training_data (str,list) – one or a list of datasets used in training

  • training_techniques (str,list) – one or a list of training techniques

  • training_resources (str) – hardware used to train

  • architecture (str, List) – one or a list of architectures used in the model

  • _filepath (str) – path to the file where the data is coming from

  • **kwargs – any other custom metadata

static from_dict(d: Dict, _filepath: Optional[str] = None)[source]

Construct Metadata from a dict.

Parameters

  • d (dict) – A dictionary of values

  • _filepath (str) – path to the file where the data is coming from

static from_file(filepath: Optional[str] = None, parent_filepath: Optional[str] = None)[source]

Load a Metadata from a file.

Parameters

  • filepath (str) – File from which to load the metadata

  • parent_filepath (str) – Parent filename (if file is imported from another file)

property flops

Get the FLOPs

property parameters

Get number of parameters

property epochs

Get epochs

property batch_size

Get batch size

property training_data

Get training data used

property training_techniques

Get techniques used

property training_resources

Get training resources used

property architecture

Get the architecture(s) used.