Skip to content

mkrum/latte

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 

History

94 Commits
 
 
 
 
 
 
 
 
 
 
 
 
 
 

Repository files navigation

latte build status

Developed by: Michael Krumdick, Tristan Mitchell, Matthew Lanus

A minimal dependency network builder styled after Caffe, built for customizability

  • Disclaimer: Program currently only works on Mac machines - linux machines will sometimes get errors

How to use the latte:

	To test latte -
1. Navigate to the src/ directory and run make
2. Go to the example directory and run $./latte sample.ptx final

	To use create own layers and use latte to compute results -
1. Create a .ptx file in similar format to sample.ptx with multiple layers and functions
	To find layers that one can use please navigate to src/layers/
2. Create an input file for the .ptx file
3. Run $./latte <ptx.ptx> <name of last layer of .ptx file>

Benchmarks:

Graph Traversals - 
	A took on average 0.000387 seconds
	B took on average 0.000372 seconds
	C took on average 0.000364 seconds
	D took on average 0.000385 seconds
	E took on average 0.000390 seconds
	F took on average 0.000386 seconds
	G took on average 0.000396 seconds
	H took on average 0.000379 seconds
	I took on average 0.000389 seconds
	J took on average 0.000397 seconds
	K took on average 0.000388 seconds

Total time to run:
	Run time of latte for 1 rows and cols is 0.00035 seconds
	Run time of latte for 10 rows and cols is 0.00027 seconds
	Run time of latte for 100 rows and cols is 0.00029 seconds
	Run time of latte for 1000 rows and cols is 0.00040 seconds
	Run time of latte for 10000 rows and cols is 0.00046 seconds

Individual Contributions:

Michael Krumdick - Created the Python-C++ interface and connected the layers of the graph 
together. Also helped debug errors until the graph worked as expected.

Matt Lanus - Worked on the graph class, helped with output formatting, ran the benchmarks, and 
helped with the README write-up.

Tristan Mitchell - Worked on the matrix class operations to organize input, helped debug graph 
errors, made test suite, and helped with README write-up.

About

Lightweight Caffe-Inspired network builder, focusing on customization

Resources

Stars

Watchers

Forks

Releases

No releases published

Packages

 
 
 

Contributors