nv-dfm-core 3.1.29

Data Federation Mesh Core Framework

nv-dfm-core

Core framework for NVIDIA Data Federation Mesh (DFM).

What is DFM?

Data Federation Mesh (DFM) is a Python-based framework for creating and orchestrating complex workflows that process data from various distributed sources and stream results into applications. DFM determines where to run each operation of a data processing pipeline and handles data movement between sites automatically.

DFM is built on top of NVIDIA Flare, which provides distributed messaging, job management, security, deployment, and a simulation framework.

Overview

nv-dfm-core is the core package of the DFM framework. It provides:

Component	Description
Pipeline API	Declarative Python API for defining pipelines (Pipeline, Yield, PlaceParam, ForEach, etc.)
Compilation	IRGen and ModGen convert pipelines to executable Petri nets
Execution	NetRunner, JobController, and Router for distributed execution
CLI	dfm command for development, federation management, and deployment
Targets	Flare (distributed) and Local (multiprocessing) execution backends

Additional DFM packages provide domain-specific adapters:

Package	Description
nv-dfm-lib-common	Shared schemas and utilities used across adapter libraries
nv-dfm-lib-weather	Experimental weather and climate data adapters (GFS, ECMWF, HRRR, SFNO, cBottle)

Installation

pip install nv-dfm-core

Or from source with uv:

git clone https://github.com/NVIDIA/data-federation-mesh.git
cd data-federation-mesh
uv sync --package nv-dfm-core

Quick Start

from nv_dfm_core.api import Pipeline, Yield, PlaceParam
from nv_dfm_core.session import Session

# Define a pipeline
with Pipeline() as p:
    result = SomeOperation(...)
    Yield(value=result)

# Execute with a session (see tutorials for federation setup)
with Session(federation="my_federation") as session:
    job = session.execute(p)
    for result in job.results():
        print(result)

Documentation

• DFM Documentation
• Installation Guide
• CLI Reference
• Tutorials

License

Apache License 2.0. See the LICENSE file for details.