{"id":13812885,"url":"https://github.com/gplhegde/caffepresso","last_synced_at":"2025-05-14T22:31:23.934Z","repository":{"id":86447522,"uuid":"69979906","full_name":"gplhegde/caffepresso","owner":"gplhegde","description":"CaffePresso: An Optimized Library for Deep Learning on Embedded Accelerator-based platforms","archived":false,"fork":false,"pushed_at":"2024-10-16T07:23:33.000Z","size":15090,"stargazers_count":88,"open_issues_count":0,"forks_count":24,"subscribers_count":16,"default_branch":"master","last_synced_at":"2024-11-19T07:39:49.117Z","etag":null,"topics":[],"latest_commit_sha":null,"homepage":"","language":"C","has_issues":true,"has_wiki":null,"has_pages":null,"mirror_url":null,"source_name":null,"license":"mit","status":null,"scm":"git","pull_requests_enabled":true,"icon_url":"https://github.com/gplhegde.png","metadata":{"files":{"readme":"README.md","changelog":null,"contributing":null,"funding":null,"license":"LICENSE.txt","code_of_conduct":null,"threat_model":null,"audit":null,"citation":null,"codeowners":null,"security":null,"support":null,"governance":null,"roadmap":null,"authors":null,"dei":null,"publiccode":null,"codemeta":null}},"created_at":"2016-10-04T15:41:03.000Z","updated_at":"2024-10-16T07:23:37.000Z","dependencies_parsed_at":null,"dependency_job_id":"ccd41292-2dde-4c07-8b48-dec3427010e2","html_url":"https://github.com/gplhegde/caffepresso","commit_stats":null,"previous_names":[],"tags_count":0,"template":false,"template_full_name":null,"repository_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories/gplhegde%2Fcaffepresso","tags_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories/gplhegde%2Fcaffepresso/tags","releases_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories/gplhegde%2Fcaffepresso/releases","manifests_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories/gplhegde%2Fcaffepresso/manifests","owner_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/owners/gplhegde","download_url":"https://codeload.github.com/gplhegde/caffepresso/tar.gz/refs/heads/master","host":{"name":"GitHub","url":"https://github.com","kind":"github","repositories_count":254239559,"owners_count":22037726,"icon_url":"https://github.com/github.png","version":null,"created_at":"2022-05-30T11:31:42.601Z","updated_at":"2022-07-04T15:15:14.044Z","host_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub","repositories_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories","repository_names_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repository_names","owners_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/owners"}},"keywords":[],"created_at":"2024-08-04T04:00:57.313Z","updated_at":"2025-05-14T22:31:18.884Z","avatar_url":"https://github.com/gplhegde.png","language":"C","funding_links":[],"categories":["Data processing"],"sub_categories":["AI ML"],"readme":"# CaffePresso #\n\nThis git repository supports the CASES 2016 paper **\"CaffePresso: An Optimized Library for Deep Learning on Embedded Accelerator-based platforms\"**. It contains ConvNet implementations for DSP, FPGA, and NoC-based embedded accelerators. We provide a code generator that translates Caffe prototxt into low-level specifications for the various backends.\n\nThe framework is modularized so support for new hardware platforms is simple. It is also possible to change the ConvNet specifications and regenerate the implementations.\n\n### What are the pre-requisites to use this repo? ###\n\n**Hardware** (USD prices from May 2016)\n\n- TI Keystone II DSP (66AK2H12) -- $997\n- Adapteva Parallella/Epiphany-III SoC board -- $126\n- NVIDIA Jetson TX1 GPU platform -- $599\n- Xilinx ZC706 FPGA -- $2275\n- Also need USB cables (TI XDS100 programming cable) and SD cards as required.\n\n**Software**\n\n- TI Code Composer Studio v6, IMGLIB and DSPLIB\n- Epiphany SDK (https://github.com/adapteva/epiphany-sdk)\n- CUDA + cuDNNv4 libraries\n- Vectorblox bitstreams + Xilinx Vivado (https://github.com/VectorBlox/mxp/tree/master/examples/boards/zc706_arm_viv)\n- OS images for Parallella (https://www.parallella.org/create-sdcard/)\n- Caffe from caffe.berkeleyvision.org. Make sure to build it with CUDNN:=1 flag.\n\n### Building and running code ###\n\nThe ConvNet configurations used in the paper are stored in **nets** folder. The code-generation scripts are in **tools/caffe-proto** folders. \n\n**gpu** -- This contains the scripts to run various ConvNets via Caffe + cuDNN on the Jetson TX1. It is a seamless experience and it will get your started right away.\n\n**mxp** -- You can either do a simulation (via license from Vectorblox) or directly execute on the FPGA board with the dowloaded 64-lane ZC706 bitstream (freely available on Vectorblox github). The header and top-level files are assembled from the code generator block.\n\n**dsp** and **noc** -- These contain sub-folders for the various datasets. There is some manual assembly required to run this through the respective build systems. The Caffe code-generator frontend supplies the ConvNet-specific header configuration for execution.\n\n### How to cite this paper? ###\n\nIf you use this tool in your work and find it useful, please cite:\n\n```\n@article{caffepresso_cases2016,\n  title={CaffePresso: An Optimized Library for Deep Learning on Embedded Accelerator-based platforms},\n  author={Hegde, Gopalakrishna and Siddhartha and Ramasamy, Nachiappan and Kapre, Nachiket},\n  booktitle = {Proceedings of the 2016 International Conference on Compilers, Architecture and Synthesis for Embedded Systems},\n  year={2016}\n}\n```\n","project_url":"https://awesome.ecosyste.ms/api/v1/projects/github.com%2Fgplhegde%2Fcaffepresso","html_url":"https://awesome.ecosyste.ms/projects/github.com%2Fgplhegde%2Fcaffepresso","lists_url":"https://awesome.ecosyste.ms/api/v1/projects/github.com%2Fgplhegde%2Fcaffepresso/lists"}