YaLM-100B VS boinc

Its probably just open source software/training sets repurposed... https://github.com/yandex/YaLM-100B

There is a 100b model published on pache 2.0 license. Though there is no information about finetuning it or using in 4-bit with smth like llama.cpp. Trying to figure out how to try it without renting extremely expensive gpu set. https://github.com/yandex/YaLM-100B

Hey everyone! I just discovered an open-source 100 billion parameter language model called YaLM, which is published under the Apache 2.0 license. The model is trained on more than 1 TB of Russian and English text. Here's the GitHub repo: https://github.com/yandex/YaLM-100B and an article explaining how it was trained: https://medium.com/yandex/yandex-publishes-yalm-100b-its-the-largest-gpt-like-neural-network-in-open-source-d1df53d0e9a6

Yandex has already released a LLM: https://github.com/yandex/YaLM-100B

So there is this open project YaLM 100B require 200 GB of disk space, it is trained on 1.7 TB of text

You'd think so, but to train a model like the one CAI uses, it would require truly jaw-breaking amount of funds. That's why CAI is so suspicious tbh. Just to give you an example, YaML (100 billion parameters which is probably less than CAI) took 65 days to train, and 800 A100 graphics cards. 175 billion parameters would not be 1.75 times higher because it's not a linear function. It would probably be 10x or even more. IIRC, "Open"Ai could only afford to train GPT-3 a single time...

I don't much follow AI news beyond what I randomly happen to see on HN, but this might still be the largest open source model: https://github.com/yandex/YaLM-100B . There's discussion of it here: https://old.reddit.com/r/MachineLearning/comments/vpn0r1/d_h... - at the bottom of that page is a comment from someone who actually ran it in the cloud.

Hilariously, Yandex Alisa runs circles around it, because it's not just a collection of gimmicks but has an actual 100B-class language model (YaLM, opensourced) as its core, plus lots of decent engineering. It's helpful, skillful and feels alive, almost like ChatGPT.

The only way I can foresee a cryptocoin actually holding value is if spending the coin meant spending processing cycles and RAM doing things like this:

https://en.wikipedia.org/wiki/List_of_volunteer_computing_pr...

But in more general sense, less like https://boinc.berkeley.edu/ and more like AWS...

It's the only way to have value, actually holding computing power in a distributed network.

Made me think of Gridcoin and BOINC https://boinc.berkeley.edu/

The BOINC Census is back for another year! BOINC is an open source software and network for volunteer computing. People can use it do donate their CPU/GPU power to various scientific research areas like cancer, drug discovery, mapping the galaxy, and more.

A few years back, I was in a similar situation and found BOINC(https://boinc.berkeley.edu/) to be a great way to contribute. It's a platform that lets you support various scientific research projects by sharing your computational power and bandwidth. However, it's worth noting that BOINC might tends to be more CPU/GPU intensive rather than bandwidth-heavy

Sign up or login to the Milky Way MayoCoin team (CPU only) and Einstein MayoCoin team (GPU and CPU) using a BOINC account. Use your Reddit or Discord username.

Berkeley Open Infrastructure for Network Computing (BOINC)

https://boinc.berkeley.edu/

https://boinc.berkeley.edu/projects.php

Has a unified management experience with the ability to subscribe to various projects, and set priorities/schedules for work units.

Probably BOINC (https://boinc.berkeley.edu/) could be a good solution for you. You can write me a DM, and I could help you to clarify is this something that could help you with your research. By default, to run your computations on BOINC you need to create a server, but we can deal with that and run your research on our own server first, so this could help you to start faster, and then later decide if you need a separate server. And yes - it's totally free.