Varun Khare

Local synchronization and communication

4

Really love Blaze for the time I have used. Why not share the files over local connections if available like Bluetooth, wifi, wifi-p2p. This will remove the necessity to have...

remove job arguments from authenticate

1

## Feature Description Authenticate requires a job name and mode version as parameters to identify whether a speed test is needed or not. This should however be done in a...

I have added a [pytorch](https://github.com/varunkhare1234/Generative-and-Discriminative-Voxel-Modeling.git) implementation of the voxception-resnet. It might be useful for others.

🚀 Feature: Supporting multiple federated learning datasets in the library

2

### 🔖 Feature description While at one end we are implementing several state-of-the-art papers in federated learning, we need to support those implementations with readily available datasets. There have been...

### 🔖 Feature description Support [Deep speed Zero Infinity](https://www.deepspeed.ai/tutorials/zero/#offloading-to-cpu-and-nvme-with-zero-infinity) in the multiprocessing job executor ### 🎤 Pitch The actor system can support large-scale worker modeling. However, the ML Job Executor...

Implement non iid dataset for recommendation

1

- [ ] Implement Kaggle Dataset NimbleEdge/fedDLRM#1 - [ ] Implement paper recommendation dataset NimbleEdge/fedDLRM#2 - [ ] Add arxiv scraper for recommendation NimbleEdge/FederatedArxivSanity#15

### 🔖 Feature description We need to support the paper [Personalized Federated Learning for Heterogeneous Clients with Clustered Knowledge Transfer](https://arxiv.org/abs/2109.08119) by Yae Jee Cho, Jianyu Wang, Tarun Chiruvolu, Gauri Joshi....

## Where? All the javadoc comments in the library ## What? Dokka to javadoc convertor doesn't support @see annotation. ## Additional Context Add any other context about the problem here.

## Description PyGrid currently does not support multiple instances of jobs per worker mostly because it would need to have multiple socket connections per worker (one for each job). This...

For basic int/string constants we would be happier if we used typedef instead of enum class to reduce the memory footprint.