Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Integrate IPFS loading from ipfs-lite-cpp library to load File formats for MNN #14

Open
SuperGNUS opened this issue Apr 10, 2023 · 8 comments · Fixed by #23
Open

Integrate IPFS loading from ipfs-lite-cpp library to load File formats for MNN #14

SuperGNUS opened this issue Apr 10, 2023 · 8 comments · Fixed by #23
Assignees

Comments

@SuperGNUS
Copy link
Contributor

After integrating MNN into thirdparty as a submodule, remove Kompute from the build and then integrate the loading through ipfs-lite-cpp to load the file formats MNN supports. I think there is the TensorFlow file format and others. Should be a wrapper around their file system loading code.

@scorpioluck20 scorpioluck20 linked a pull request Apr 20, 2023 that will close this issue
@scorpioluck20 scorpioluck20 removed a link to a pull request Apr 20, 2023
@scorpioluck20
Copy link
Contributor

I'm working on this issue.
BRs,
Tan Nguyen

@scorpioluck20
Copy link
Contributor

scorpioluck20 commented May 7, 2023

Implement was done for Linux OS.
ipfs-lite-cpp: GeniusVentures/ipfs-lite-cpp#1
thirdparty: #23

@SuperGNUS
Copy link
Contributor Author

Please refactor this. There is no reason this needs to be embedded in ipfs-lite-cpp, it should be its own library and when added to SuperGenius, it should hook itself into ipfs-lite-cpp, not be embedded in the library

@scorpioluck20
Copy link
Contributor

scorpioluck20 commented May 15, 2023

Update design description:

Please don't hack in other features to libraries where they don't belong.  
The MNN parser has no business being a subdirectory of ipfs-lite-cpp.  
I just updated the feature request, so refactor it into it own repo called something like MNNParser.  
Then SuperGenius can call MNNParser, when it receives a .MNN file or whatever the suffix is.

If the MNNParser is also meant to be a streaming data handler for the ML/AI data, 
please make it separate and when constructing the main MNN class,
have it hook into the streaming part of ipfs-lite-cpp.

@scorpioluck20
Copy link
Contributor

https://github.com/GeniusVentures/MNNLoader|
MNN Loader library under development. Will be integrate it to ipfs-lite-cpp when it done.

@scorpioluck20 scorpioluck20 linked a pull request May 30, 2023 that will close this issue
@scorpioluck20
Copy link
Contributor

scorpioluck20 commented May 30, 2023

PR was create: #32
Build is working on Linux and Windows OS.
For other OS build got failed by another library.
For the SuperGenius can call MNNLoader library to loading or print-out information .MNN file.
Just follow-up with the MNNExample program include in MNNLoader.
We can closed this ticker after merge the PR #32
BRs,
TN

@itsafuu
Copy link
Contributor

itsafuu commented Apr 1, 2024

MNN expects TensorFlow, etc to be converted to mnn format using the converter util. We should probably just have this done before it is sent out to clients for processing. Otherwise, the ASyncIOmanager can load any type of file from IPFS as is, so I think this is fine.

@Super-Genius
Copy link
Contributor

Just a comment to see if github notifications go to element. Maybe move this do done @itsafuu

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment