|Ray||Thespian Actor Library|
|7 days ago||6 months ago|
|Apache License 2.0||MIT License|
Stars - the number of stars that a project has on GitHub. Growth - month over month growth in stars.
Activity is a relative number indicating how actively a project is being developed. Recent commits have higher weight than older ones.
For example, an activity of 9.0 indicates that a project is amongst the top 10% of the most actively developed projects that we are tracking.
JORLDY: OpenSource Reinforcement Learning Framework
2 projects | reddit.com/r/reinforcementlearning | 8 Nov 2021
Distributed RL algorithms are provided using ray
Python stands to lose its GIL, and gain a lot of speed
5 projects | reddit.com/r/programming | 20 Oct 2021
I had a similar use case and ended up using ray. https://github.com/ray-project/ray
How to deploy a rllib-trained model?
3 projects | reddit.com/r/reinforcementlearning | 16 Oct 2021
Currently, rllib's "--export-formats" does nothing; I have folders of checkpoints, but no models. Looks like currently the internal export_model function isn't implemented: https://github.com/ray-project/ray/issues/190213 projects | reddit.com/r/reinforcementlearning | 16 Oct 2021
[HELP] Converting many individual workstations into a HPC cluster
1 project | reddit.com/r/HPC | 11 Oct 2021
Unless you have infiniband, you might want to build it as a kubernetes cluster and look at something like (ray-project)[https://github.com/ray-project/ray] it has a ton of distributed plugin packages that are Ethernet based.
Show HN: SpotML – Managed ML Training on Cheap AWS/GCP Spot Instances
6 projects | news.ycombinator.com | 3 Oct 2021
Neat. Congratulations on the launch!
Apart from the fact that it could deploy to both GCP and AWS, what does it do differently than AWS Batch ?
When we had a similar problem, we ran jobs on spots with AWS Batch and it worked nicely enough.
Some suggestions (for a later date):
1. Add built-in support for Ray  (you'd essentially be then competing with Anyscale, which is a VC funded startup, just to contrast it with another comment on this thread) and dbt .
2. Support deploying coin miners (might be good to widen the product's reach; and stand it up against the likes of consensys).
3. Get in front of many cost optimisation consultants out there, like the Duckbill Group.
If I may, where are you building this product from? And how many are on the team?
Writing your First Distributed Python Application with Ray (without multiprocessing)
4 projects | reddit.com/r/Python | 23 Aug 2021
Here is an older discussion on dask vs ray from the creators of both projects: https://github.com/ray-project/ray/issues/642
[D] Kubeflow vs. Argo for ML Pipeline Tool
2 projects | reddit.com/r/MachineLearning | 17 Aug 2021
Here is link number 1 - Previous text "Ray"2 projects | reddit.com/r/MachineLearning | 17 Aug 2021
If you are looking for a developer-friendly tool, I'd ditch the task/workflow orchestration paradigm altogether and use something like Ray. It's made by and for ML practitioners, it's much more versatile, has no unwarranted DSLs (pure python), and you can test locally before deploying with pretty much the same code.
1 project | news.ycombinator.com | 8 Jun 2021
Thespian Actor Library
Ask HN: How to 100% enable remote infrastructure?
1 project | news.ycombinator.com | 29 Apr 2021
"Even things that can't go wrong, do." - Troubleshooting Analog Circuits, Robert "Bob" Pease.
I had an application on a Raspberry PI that paired with a Bluetooth Low Energy device to fetch and send its data through a 3G dongle, on the premises of non-technical people who cannot troubleshoot, in different countries and time zones. There were a lot of things that could go wrong, and I wrote code to mitigate and recover, including pulling new code.
Part of it was using the Actor Model. I wrote actors to connect to the device, pulling data, sending data, computing what was sent, etc. The actor system handled the actors, when one died, it would recreate another one when an unhandled exception was met, for example.
If you're doing Scala, take a look at Akka. The new Scala has native support for this, if I remember correctly, without Akka.
What are some alternatives?
Faust - Python Stream Processing
stable-baselines3 - PyTorch version of Stable Baselines, reliable implementations of reinforcement learning algorithms.
gevent - Coroutine-based concurrency library for Python
SCOOP (Scalable COncurrent Operations in Python) - SCOOP (Scalable COncurrent Operations in Python)
stable-baselines - A fork of OpenAI Baselines, implementations of reinforcement learning algorithms
optuna - A hyperparameter optimization framework
Tomorrow - Magic decorator syntax for asynchronous code in Python
Wallaroo - Distributed Stream Processing
pipelines - An experimental programming language for data flow