mirror of
https://github.com/clearml/clearml-docs
synced 2025-02-25 13:45:33 +00:00
56 lines
2.1 KiB
Markdown
56 lines
2.1 KiB
Markdown
---
|
|
title: MegEngine
|
|
---
|
|
|
|
The [megengine_mnist.py](https://github.com/clearml/clearml/blob/master/examples/frameworks/megengine/megengine_mnist.py)
|
|
example demonstrates the integration of ClearML into code that uses [MegEngine](https://github.com/MegEngine/MegEngine)
|
|
and [TensorBoardX](https://github.com/lanpa/tensorboardX). ClearML automatically captures models saved with `megengine`.
|
|
|
|
The example script does the following:
|
|
* Trains a simple deep neural network on MegEngine's built-in [MNIST](https://www.megengine.org.cn/doc/master/en/reference/api/megengine.data.dataset.MNIST.html)
|
|
dataset.
|
|
* Creates a TensorBoardX `SummaryWriter` object to log scalars during training.
|
|
* Creates a ClearML task named `megengine mnist train` in the `examples` project.
|
|
|
|
## Hyperparameters
|
|
|
|
ClearML automatically logs command line options defined with `argparse`. They appear in the task's **CONFIGURATION**
|
|
tab under **HYPERPARAMETERS** **>** **Args**.
|
|
|
|
|
|
|
|
## Scalars
|
|
|
|
The example script's `train` function calls TensorBoardX's `SummaryWriter.add_scalar` method to report `loss`.
|
|
ClearML automatically captures the data that is added to the `SummaryWriter` object.
|
|
|
|
These scalars can be visualized in plots, which appear in the ClearML [WebApp](../../../webapp/webapp_home.md), in the
|
|
task's **SCALARS** tab.
|
|
|
|
|
|
|
|
|
|
## Models
|
|
|
|
ClearML automatically captures the model logged using the `megengine.save` method, and saves it as an artifact.
|
|
|
|
View saved snapshots in the task's **ARTIFACTS** tab.
|
|
|
|
|
|
|
|
To view the model details, click the model name in the **ARTIFACTS** page, which will open the model's info tab. Alternatively, download the model.
|
|
|
|
The model info panel contains the model details, including:
|
|
* Model URL
|
|
* Framework
|
|
* Snapshot locations.
|
|
|
|
|
|
|
|
## Console
|
|
|
|
All console output during the script's execution appears in the task's **CONSOLE** page.
|
|
|
|
|
|
|