solver-pytorch

Notes

  • Installation
    • Requirements
    • Installation via Pip
    • Installation from Source
    • Documentation Dependencies
  • Getting Started
    • Minimal Solver Skeleton
    • Required Hooks
    • Run Modes
    • What the Base Class Handles
  • Configuration
    • How Configuration Is Loaded
    • Top-Level Sections
    • Important SOLVER Fields
    • Dataset Configuration
    • Example Override
  • Extending thsolver
    • Model and Dataset Registries
    • Custom Result Handling
    • Custom Evaluation Jobs
    • Replacing the Dataset Helper
  • Distributed Training
    • Spawn Mode
    • torchrun Mode
    • Master Process Responsibilities
    • Reproducibility

Package Reference

  • thsolver.config
    • get_config()
    • parse_args()
  • thsolver.solver
    • Solver
  • thsolver.dataset
    • read_file()
    • Dataset
  • thsolver.registry
    • register_model()
    • model_entrypoints()
    • is_model()
    • list_models()
    • build_model()
    • register_dataset()
    • dataset_entrypoints()
    • is_dataset()
    • list_datasets()
    • build_dataset()
  • thsolver.sampler
    • InfSampler
    • DistributedInfSampler
  • thsolver.lr_scheduler
    • multi_step()
    • cos()
    • poly()
    • constant()
    • cos_warmup()
    • poly_warmup()
    • step_warmup()
    • get_lr_scheduler()
  • thsolver.tracker
    • AverageTracker
solver-pytorch
  • Search

© Copyright 2026, Peng-Shuai Wang.

Built with Sphinx using a theme provided by Read the Docs.