Skip to content

Experiment_Tracking.md

Latest commit

 

History

History
59 lines (33 loc) · 2.57 KB

Experiment_Tracking.md

File metadata and controls

59 lines (33 loc) · 2.57 KB
graph LR
    PBNeuronLayerTracker["PBNeuronLayerTracker"]
    PB_Layer["PB_Layer"]
    PB_Utils["PB_Utils"]
    MNIST_PerforatedAI["MNIST_PerforatedAI"]
    PBNeuronLayerTracker -- "manages data from" --> PB_Layer
    PB_Layer -- "provides data to" --> PBNeuronLayerTracker
    PBNeuronLayerTracker -- "utilizes" --> PB_Utils
    PB_Utils -- "provides utilities to" --> PBNeuronLayerTracker
    MNIST_PerforatedAI -- "utilizes for logging" --> PBNeuronLayerTracker
    PBNeuronLayerTracker -- "provides tracking services to" --> MNIST_PerforatedAI
Loading

CodeBoardingDemoContact

Details

One paragraph explaining the functionality which is represented by this graph. What the main flow is and what is its purpose.

PBNeuronLayerTracker

The core component of the Experiment Tracking subsystem, responsible for initializing, managing, and logging training metrics and internal states of neuron layers, optimizers, and schedulers. It controls and tracks Perforated Backpropagation training modes for data collection and analysis.

Related Classes/Methods:

PB_Layer

Represents internal model state, with which PBNeuronLayerTracker interacts to collect state information and metrics from individual neuron layers.

Related Classes/Methods:

  • perforatedai.pb_layer

PB_Utils

Provides helper functions and data structures for tasks such as data aggregation, formatting, or specific calculations related to tracking, utilized by PBNeuronLayerTracker.

Related Classes/Methods:

  • perforatedai.pb_utils

MNIST_PerforatedAI

An example or application module that utilizes the Experiment Tracking component (PBNeuronLayerTracker) to log and monitor its training progress and model behavior.

Related Classes/Methods:

FAQ