Data Geek
@datageek-pl.bsky.social
AI codegens / agents + applied AI / CS / tech insights 🎯 target shooter, sci-fi geek 🚀 techno-optimist
Pinned
Data Geek
@datageek-pl.bsky.social
· Nov 28
My answer to AiSuite - focused even more on the GPU-poor.
It's still in early development, but you can take a look here:
github.com/mobarski/ai-...
All feedback is more than welcome (comments, likes, github stars, reposts).
It's still in early development, but you can take a look here:
github.com/mobarski/ai-...
All feedback is more than welcome (comments, likes, github stars, reposts).
Reposted by Data Geek
Concordia is a library for generative agent-based modeling that works like a table-top role-playing game.
It's open source and model agnostic.
Try it today!
github.com/google-deepm...
It's open source and model agnostic.
Try it today!
github.com/google-deepm...
GitHub - google-deepmind/concordia: A library for generative social simulation
A library for generative social simulation. Contribute to google-deepmind/concordia development by creating an account on GitHub.
github.com
November 16, 2024 at 11:49 PM
Concordia is a library for generative agent-based modeling that works like a table-top role-playing game.
It's open source and model agnostic.
Try it today!
github.com/google-deepm...
It's open source and model agnostic.
Try it today!
github.com/google-deepm...
Still cooking my aibricks project.
Simple example of what I mean by "configuration driven":
Simple example of what I mean by "configuration driven":
December 14, 2024 at 10:19 AM
Still cooking my aibricks project.
Simple example of what I mean by "configuration driven":
Simple example of what I mean by "configuration driven":
Reposted by Data Geek
From the same people (lmarena.ai) who brought you Chatbot Arena, they are introducing WebDev Arena.
Leaderboard: web.lmarena.ai/leaderboard
Leaderboard: web.lmarena.ai/leaderboard
December 13, 2024 at 9:14 PM
From the same people (lmarena.ai) who brought you Chatbot Arena, they are introducing WebDev Arena.
Leaderboard: web.lmarena.ai/leaderboard
Leaderboard: web.lmarena.ai/leaderboard
Reposted by Data Geek
LEXICO defining the Pareto frontier of KV cache compression
December 12, 2024 at 7:53 PM
LEXICO defining the Pareto frontier of KV cache compression
Reposted by Data Geek
Textual 1.0 has been released. 🥳
Three years in the making. A TUI framework that is bigger than the terminal.
To celebrate, I want to give away some trade secrets. Because I am appalling at keeping secrets.
Tell me what you think of the diagrams...
textual.textualize.io/blog/2024/12...
Three years in the making. A TUI framework that is bigger than the terminal.
To celebrate, I want to give away some trade secrets. Because I am appalling at keeping secrets.
Tell me what you think of the diagrams...
textual.textualize.io/blog/2024/12...
Textual - Algorithms for high performance terminal apps
Textual is a TUI framework for Python, inspired by modern web development.
textual.textualize.io
December 12, 2024 at 3:27 PM
Textual 1.0 has been released. 🥳
Three years in the making. A TUI framework that is bigger than the terminal.
To celebrate, I want to give away some trade secrets. Because I am appalling at keeping secrets.
Tell me what you think of the diagrams...
textual.textualize.io/blog/2024/12...
Three years in the making. A TUI framework that is bigger than the terminal.
To celebrate, I want to give away some trade secrets. Because I am appalling at keeping secrets.
Tell me what you think of the diagrams...
textual.textualize.io/blog/2024/12...
1000x inference cost reduction by converting qwen/llama to rwkv architecture without retraining from scratch.
Big if true/without-any-major-issues. Definitely worth checking.
huggingface.co/recursal/QRW...
Big if true/without-any-major-issues. Definitely worth checking.
huggingface.co/recursal/QRW...
December 12, 2024 at 8:04 AM
1000x inference cost reduction by converting qwen/llama to rwkv architecture without retraining from scratch.
Big if true/without-any-major-issues. Definitely worth checking.
huggingface.co/recursal/QRW...
Big if true/without-any-major-issues. Definitely worth checking.
huggingface.co/recursal/QRW...
Reposted by Data Geek
December 11, 2024 at 8:11 PM
Reposted by Data Geek
In case you didn't know, Bluesky has built-in RSS support 👀
openrss.org/blog/bluesky...
openrss.org/blog/bluesky...
Bluesky has launched RSS feeds
Now that Bluesky's got its own RSS feeds, here's what you should know
openrss.org
December 9, 2024 at 1:00 PM
In case you didn't know, Bluesky has built-in RSS support 👀
openrss.org/blog/bluesky...
openrss.org/blog/bluesky...
Reposted by Data Geek
The Semantic Hub Hypothesis: Language Models Share Semantic Representations Across Languages and Modalities
arxiv.org/abs/2411.04986
arxiv.org/abs/2411.04986
December 10, 2024 at 4:35 AM
The Semantic Hub Hypothesis: Language Models Share Semantic Representations Across Languages and Modalities
arxiv.org/abs/2411.04986
arxiv.org/abs/2411.04986
Reposted by Data Geek
entrapix — your LLM should raise a ConfusedAgentError when it doesn’t know
Why? Because as an application developer, I have a lot of things I can do to give the LLM more information
github.com/tkellogg/oll...
Why? Because as an application developer, I have a lot of things I can do to give the LLM more information
github.com/tkellogg/oll...
github.com
December 8, 2024 at 11:36 PM
entrapix — your LLM should raise a ConfusedAgentError when it doesn’t know
Why? Because as an application developer, I have a lot of things I can do to give the LLM more information
github.com/tkellogg/oll...
Why? Because as an application developer, I have a lot of things I can do to give the LLM more information
github.com/tkellogg/oll...
Leaked Windsurf #codegen prompt: www.reddit.com/r/LocalLLaMA...
From the LocalLLaMA community on Reddit
Explore this post and more from the LocalLLaMA community
www.reddit.com
December 7, 2024 at 6:12 PM
Leaked Windsurf #codegen prompt: www.reddit.com/r/LocalLLaMA...
Reposted by Data Geek
Paper📰: Evaluating Language Models as Synthetic Data Generators ( arxiv.org/abs/2412.03679 )
Code💻: github.com/neulab/data-...
Data: huggingface.co/Data-Agora
Code💻: github.com/neulab/data-...
Data: huggingface.co/Data-Agora
December 7, 2024 at 6:51 AM
Paper📰: Evaluating Language Models as Synthetic Data Generators ( arxiv.org/abs/2412.03679 )
Code💻: github.com/neulab/data-...
Data: huggingface.co/Data-Agora
Code💻: github.com/neulab/data-...
Data: huggingface.co/Data-Agora
Please don’t mind any unfollow from my side - I’m experimenting with the atproto api.
December 5, 2024 at 3:47 PM
Please don’t mind any unfollow from my side - I’m experimenting with the atproto api.
Reposted by Data Geek
www.bnnbloomberg.ca/business/tec...
One of the best pieces on the topic I've read this year.
Endless ruminations on this: @josiecox.bsky.social
One of the best pieces on the topic I've read this year.
Endless ruminations on this: @josiecox.bsky.social
Why Are Women Less Likely to Use AI?
AI’s early adopters are disproportionately men, a disconnect that could stands to exacerbate the gender pay gap
www.bnnbloomberg.ca
December 4, 2024 at 7:07 AM
www.bnnbloomberg.ca/business/tec...
One of the best pieces on the topic I've read this year.
Endless ruminations on this: @josiecox.bsky.social
One of the best pieces on the topic I've read this year.
Endless ruminations on this: @josiecox.bsky.social
This is how happiness looks like.
December 3, 2024 at 7:54 PM
This is how happiness looks like.
Now cooking the config layer and streaming.
Middleware for streaming might require some trial and error but hey, thats the fun part!
The results might be nice - imagine streaming but line by line instead by random number of tokens - the latency is still low and you can easily act on produced text!
Middleware for streaming might require some trial and error but hey, thats the fun part!
The results might be nice - imagine streaming but line by line instead by random number of tokens - the latency is still low and you can easily act on produced text!
My answer to AiSuite - focused even more on the GPU-poor.
It's still in early development, but you can take a look here:
github.com/mobarski/ai-...
All feedback is more than welcome (comments, likes, github stars, reposts).
It's still in early development, but you can take a look here:
github.com/mobarski/ai-...
All feedback is more than welcome (comments, likes, github stars, reposts).
December 3, 2024 at 4:08 PM
Now cooking the config layer and streaming.
Middleware for streaming might require some trial and error but hey, thats the fun part!
The results might be nice - imagine streaming but line by line instead by random number of tokens - the latency is still low and you can easily act on produced text!
Middleware for streaming might require some trial and error but hey, thats the fun part!
The results might be nice - imagine streaming but line by line instead by random number of tokens - the latency is still low and you can easily act on produced text!
This 1h free course is an excellent intro to memgpt like agent architecture. No frameworks - just doing fun things from scratch.
I highly recommend this one even to people not interested in AI powered games - agentic internals are very similar everywhere.
www.deeplearning.ai/short-course...
I highly recommend this one even to people not interested in AI powered games - agentic internals are very similar everywhere.
www.deeplearning.ai/short-course...
Building an AI-Powered Game
Learn to build with LLMs by creating a fun interactive game from scratch.
www.deeplearning.ai
December 1, 2024 at 5:36 PM
This 1h free course is an excellent intro to memgpt like agent architecture. No frameworks - just doing fun things from scratch.
I highly recommend this one even to people not interested in AI powered games - agentic internals are very similar everywhere.
www.deeplearning.ai/short-course...
I highly recommend this one even to people not interested in AI powered games - agentic internals are very similar everywhere.
www.deeplearning.ai/short-course...
Reposted by Data Geek
This is crazy, run all workflows for generative AI on local GPU from one simple to use interface, includes snap-ins for all the popular workflows. pinokio.computer
#ai #aiTools #aiArtist
#ai #aiTools #aiArtist
Pinokio
AI Browser
pinokio.computer
November 30, 2024 at 8:10 PM
This is crazy, run all workflows for generative AI on local GPU from one simple to use interface, includes snap-ins for all the popular workflows. pinokio.computer
#ai #aiTools #aiArtist
#ai #aiTools #aiArtist
Twitter's algo has made us forget the value of links in short messages from people who share our interests.
To say that I’m in tears would be an overstatement but I’m seriously touched.
To say that I’m in tears would be an overstatement but I’m seriously touched.
November 30, 2024 at 10:59 AM
Twitter's algo has made us forget the value of links in short messages from people who share our interests.
To say that I’m in tears would be an overstatement but I’m seriously touched.
To say that I’m in tears would be an overstatement but I’m seriously touched.
Reposted by Data Geek
New study shows LLMs outperform neuroscience experts at predicting experimental results in advance of experiments (86% vs 63% accuracy).
They use a fine-tuned Mistral 7B but other models worked too. Suggests LLMs can integrate scientific knowledge at scale to support research.
They use a fine-tuned Mistral 7B but other models worked too. Suggests LLMs can integrate scientific knowledge at scale to support research.
November 29, 2024 at 5:10 PM
New study shows LLMs outperform neuroscience experts at predicting experimental results in advance of experiments (86% vs 63% accuracy).
They use a fine-tuned Mistral 7B but other models worked too. Suggests LLMs can integrate scientific knowledge at scale to support research.
They use a fine-tuned Mistral 7B but other models worked too. Suggests LLMs can integrate scientific knowledge at scale to support research.
My answer to AiSuite - focused even more on the GPU-poor.
It's still in early development, but you can take a look here:
github.com/mobarski/ai-...
All feedback is more than welcome (comments, likes, github stars, reposts).
It's still in early development, but you can take a look here:
github.com/mobarski/ai-...
All feedback is more than welcome (comments, likes, github stars, reposts).
November 28, 2024 at 7:24 PM
My answer to AiSuite - focused even more on the GPU-poor.
It's still in early development, but you can take a look here:
github.com/mobarski/ai-...
All feedback is more than welcome (comments, likes, github stars, reposts).
It's still in early development, but you can take a look here:
github.com/mobarski/ai-...
All feedback is more than welcome (comments, likes, github stars, reposts).
Reposted by Data Geek
note to the gpu poor: you can still train
A new approach to training models in memory-constrained settings, LoQT allows for the pre-training of a 13B LLM on a 24GB GPU without model parallelism, checkpointing, or offloading strategies during training
Code: github.com/sebulo/LoQT
Code: github.com/sebulo/LoQT
GitHub - sebulo/LoQT
Contribute to sebulo/LoQT development by creating an account on GitHub.
github.com
November 17, 2024 at 11:49 PM
note to the gpu poor: you can still train