hivemind

Tesseract

Distributed training of large neural networks across volunteer computers.

[WIP] - this branch is in progress of updating. If you're interested in supplementary code for Learning@home paper, you can find it at https://github.com/mryab/learning-at-home .

What do I need to run it?

• One or several computers, each equipped with at least one GPU
• Each computer should have at least two open ports (if not, consider ssh port forwarding)
• Some popular Linux x64 distribution
  • Tested on Ubuntu16.04, should work fine on any popular linux64 and even MacOS;
  • Running on Windows natively is not supported, please use vm or docker;

How do I run it?

Currently, there isn't any way to do it easily. There are some tests (you can look into CI logs and/or config) and we want to expand them, but if you want to do something complex with it, you're on your own.

tesseract quick tour

Trainer process:

• RemoteExpert(tesseract/client/remote_expert.py) behaves like a pytorch module with autograd support but actually sends request to a remote runtime.
• GatingFunction(tesseract/client/gating_function.py) finds best experts for a given input and either returns them as RemoteExpert or applies them right away.

Runtime process:

• TesseractRuntime (tesseract/runtime/__init__.py) aggregates batches and performs inference/training of experts according to their priority.
• TesseractServer (tesseract/server/__init__.py) wraps runtime and periodically uploads experts into TesseractNetwork.

DHT:

• TesseractNetwork(tesseract/network/__init__.py) is a node of Kademlia-based DHT that stores metadata used by trainer and runtime.

Limitations

DHT:

• DHT functionality is severely limited by its inability to traverse NAT.
• Because of this all the features that require DHT are in deep pre-alpha state and cannot be used without special setup.