Skip to content

function47/CS7670MLTasks

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 

History

2 Commits
 
 

Repository files navigation

CS7670 MLTasks

Train a NN on pytorch on CIFAR-10.

Selection of NNs

A common selection is ResNet18 which is a built-in model of pytorch. However, an 18-layer model may be too large for this task. We could use ResNet-9 or even LeNet-5. However, since the dataloader is handling the data loads. I belive the different models will not affect the results.


Hyperparameters

The image loading in pytorch is handled by torch.dataloader. A few hyperparameters will affect the images loading processes.

Batch size

Batch size is the number of samples used for updating the internal model parameters.

Sample approach

1. Sequential vs shuffled sampler
2. memory pinning

Data parallelism

The dataloader allow user to define number of sub-processes as number of workers for handling dataset.

Prefech factors


References

Ofeidis, Iason, Diego Kiedanski, and Leandros Tassiulas. "An Overview of the Data-Loader Landscape: Comparative Performance Analysis." arXiv preprint arXiv:2209.13705 (2022).

About

No description, website, or topics provided.

Resources

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published