Skip to content


A base class for all of Deci's model classes. A model stores data in constant fields, and let us manipulate the data in a more readable way.


Name Type Description Notes
update_time datetime The last time when the model data was changed. [optional]
creation_time datetime The time when the model was added to model repository. [optional]
id str [optional]
deleted bool Is this model had been deleted? [optional] [default to False]
model_id str The unique indicator of this model [optional]
name str The name of this model.
owner str What is the id of the owner of this model? [optional]
version str The version of the model, 1.0 for baseline 1.X for optimized [optional] [default to '1.0']
model_size float The size of the model file in mb. [optional] [default to 0]
source ModelSource [optional]
quantization_level QuantizationLevel [optional]
framework FrameworkType The deep learning framework of the model - which library is it based on?
platform_version str The platform version which the model was created by. Used in order to distinguish which packages version are required. [optional]
dl_task str
input_dimensions list[object] A list of numbers, describing the vector (dimensions) of the input model. A tuple - \"(1,2,3)\" or a list of integers as input. the integers have separated with commas \",\". [optional]
channel_first bool This field indicate if the channel of the model is the first input dim (True) or the last one (False). [optional] [default to True]
fetched_model_input_dimensions list[object] The input dimensions of the model as fetched from the model inferencer engine. Will be None if the engine doesn't support getting the input vector shape. [optional]
dataset_name DatasetName The dataset that the model is based on. [optional]
architecture str The architecture of the model, must be supported by deci. [optional]
primary_hardware HardwareReturnSchema
primary_batch_size int The batch size to present on default when viewing this model. [optional] [default to 1]
benchmark_state ModelBenchmarkState What is the state of the model benchmark? [optional]
benchmark_start_date datetime When the benchmark process was started? [optional]
benchmark_end_date datetime When the benchmark process was ended? [optional]
benchmark dict(str, list[ModelBenchmarkResultMetadata]) Benchmark result of the model per batch size per hardware. [optional]
optimization_state ModelOptimizationState Which optimizations ran on this model. [optional]
optimization_start_date datetime
optimization_end_date datetime
gru_state ModelGruState What is the state of gru on this model? [optional]
gru_start_date datetime When was gru started on this model? [optional]
gru_end_date datetime When was gru ended on this model? [optional]
input_tensor_name str The name of the input layer that will be used as an input layer for TensorFlow, Exactly as it is used inside of TensorFlow. [optional]
output_tensor_name str The name of the output layer that will be used as an input layer for TensorFlow, Exactly as it is used inside of TensorFlow. [optional]
description str Description explains this model purpose. [optional] [default to '']
tags list[str] List of tags used to query this model quickly. [optional] [default to []]
kpis list[KPI] Model goals for primary hardware and batch size. [optional] [default to []]
accuracy_metrics list[AccuracyMetric] List of all accuracy metrics on this model performance. [optional] [default to []]
hyper_parameters list[HyperParameter] List of all user defined hyper parameters. [optional] [default to []]
raw_format bool Whether we want to download files of the model in a raw (true) or regular (false) format [optional] [default to False]
company_name str The name of the company the model belongs to. [optional]
company_id str The unique id of the company the model belongs to. [optional]
workspace_id str [optional]
baseline_model_id str
error ModelErrorRecord Current errors for this model. [optional]
custom_hardware str A custom hardware defined by the user [optional]
colab_link str A link to a Colab with code to download this specific model. [optional]
is_int8_calibrated bool Indicates whether a model is INT8 calibrated [optional] [default to False]
dl_task_label DeepLearningTaskLabel

[Back to Model list] [Back to API list] [Back to README]