Skip to content
experimaestroPublic

About

Experimaestro add-on for PyTorch-based experiments

Resources

Stars

3 stars

Watchers

1 watching

Forks

Repository files navigation

xpm-torch

PyPI version RTD

PyTorch training framework built on experimaestro and Lightning Fabric.

Documentation

The full documentation is at xpm-torch.readthedocs.io:

  • Overview — what's in the package and how it fits together
  • Module system — Module, ModuleLoader and safetensors serialization
  • Training — Learner, trainers, checkpointing and validation
  • Optimization — optimizers, schedulers and gradient hooks
  • HuggingFace Hub — export actions, pushing and loading models
  • Several CUDA versions — xpm-torch uv-cuda-sync, one environment per torch build, chosen on each SLURM node

Features

  • Module system: Module base class combining experimaestro Config with torch.nn.Module, with safetensors serialization
  • Training: Learner task with checkpointing, validation listeners, and TensorBoard logging
  • HuggingFace Hub: Upload/download models via ExperimaestroHFHub (from experimaestro)
  • Distributed training: Lightning Fabric integration for multi-GPU/node training
  • Several CUDA versions: xpm-torch uv-cuda-sync creates one uv environment per torch build (e.g. cu126 and cu130), and experimaestro runs each job with the one matching the GPU of its node

Installation

pip install xpm-torch

Quick Start

from xpm_torch.module import Module
from experimaestro import Param
import torch
import torch.nn as nn


class MyModel(Module):
    input_dim: Param[int]
    hidden_dim: Param[int]

    def __initialize__(self):
        self.fc1 = nn.Linear(self.input_dim, self.hidden_dim)

    def forward(self, x):
        return self.fc1(x)


# Create config, then instance
cfg = MyModel.C(input_dim=50, hidden_dim=100)
model = cfg.instance()
model.initialize()

# Save/load with safetensors
from pathlib import Path

model.save_model(Path("checkpoint/model"))
model.load_model(Path("checkpoint/model"))

License

GPL-3.0

About

Experimaestro add-on for PyTorch-based experiments

Resources

Stars

3 stars

Watchers

1 watching

Forks

Releases

Packages

Contributors

Languages