decoder.sh (@Decoder_sh) header
decoder.sh (@Decoder_sh) profile picture

Machine learning and Large Language Models 🤖https://teapawt.net/t.co/MRxfN2rOAY | ex-Uber

Joined January 2024
Have you been thinking about getting started with @langchain but don't know where to begin? Join me as I teach you how to use Templates, models and Runnables to make your very first chain, feat. @ollama youtu.be/qYJSNCPmDIk
4
8
42
7,596
My results seem to match the paper so far
1
98
This is just measuring activation level by layer, by the next step will be using the extracted "persona" at each level multiplied by some coefficient, then measuring the extent to which the persona is expressed in the response. Higher coeff, higher expression if all goes well
60
brb extracting an evil llm
56
decoder.sh retweeted
It's ugly but it's mine!
1
1
2
128
decoder.sh retweeted
When I test what I learned from ChatGPT about circuit design on an actual breadboard
2
1
3
156
Introducing our first set of Llama 4 models! We’ve been hard at work doing a complete re-design of the Llama series. I’m so excited to share it with the world today and mark another major milestone for the Llama herd as we release the *first* open source models in the Llama 4 collection 🦙. Here are some highlights: 📌 The Llama series have been re-designed to use state of the art mixture-of-experts (MoE) architecture and natively trained with multimodality. We’re dropping Llama 4 Scout & Llama 4 Maverick, and previewing Llama 4 Behemoth. 📌 Llama 4 Scout is highest performing small model with 17B activated parameters with 16 experts. It’s crazy fast, natively multimodal, and very smart. It achieves an industry leading 10M+ token context window and can also run on a single GPU! 📌 Llama 4 Maverick is the best multimodal model in its class, beating GPT-4o and Gemini 2.0 Flash across a broad range of widely reported benchmarks, while achieving comparable results to the new DeepSeek v3 on reasoning and coding – at less than half the active parameters. It offers a best-in-class performance to cost ratio with an experimental chat version scoring ELO of 1417 on LMArena. It can also run on a single host! 📌 Previewing Llama 4 Behemoth, our most powerful model yet and among the world’s smartest LLMs. Llama 4 Behemoth outperforms GPT4.5, Claude Sonnet 3.7, and Gemini 2.0 Pro on several STEM benchmarks. Llama 4 Behemoth is still training, and we’re excited to share more details about it even while it’s still in flight. A big thanks to all of our launch partners (full list in blog) for helping us bring Llama 4 to developers everywhere including @huggingface, @togethercompute, @SnowflakeDB, @ollama, @databricks and many others👏 This is just the start, we have more models coming and the team is really cooking – look out for Llama 4 Reasoning 😉 A few weeks ago, we celebrated Llama being downloaded over 1 billion times. Llama 4 demonstrates our long-term commitment to open source AI, the entire open source AI community, and our unwavering belief that open systems will produce the best small, mid-size and soon frontier models. Llama would be nothing without the global open source AI community & we are so ready to begin this next chapter with you. 🦙 Read more about the release here: llama.com/, and try it in our products today.
308
861
428
5,541
1,141,168
decoder.sh retweeted
you know you can rely on me for python advice. i’ve sensed the shift and adopted marimo, polars, altair early. if you still use matplotlib, delete it from your computer. install altair. this charting library is declarative and it just outputs a vega-lite spec you can use elsewhere. if you still use pandas, delete it from your computer. install polars. this dataframe library doesn’t have the indexes bullshit and has a much more sane api. adapt your workflow to use parquet for further speedup. if you still use jupyter, delete it from your computer. install marimo. it’s truly a next gen notebook. reactive, easily versioned, importable and executable as an app. goes without saying now, but migrate to astral’s tools. use ruff for linting and formatting (replaces black and all linters) and uv (replaces pyenv, pip, poetry) for installing python versions and working on projects. remember uv run command and forget you ever had to manage a venv.
41
49
8
634
47,552
.@cursor_ai user? Try my speech mcp to notify you when your agents are done thinking. Props to @xenovacom npmjs.com/package/speech-mcp…
1
1
204
All future tutorials will use @astral_sh 's `uv` instead of venv etc. This should hopefully make setup simpler for new coders!
74
Hey anon, do you edit your own videos?
1
170
Had an interesting experience this weekend. GF and I were refinishing a table. We live in an apartment in Brooklyn so we were sanding outside on the first floor. SO many neighbors stopped by to chat and give advice that we wouldn't have gotten if we were working in private.
1
2
135
tldr: build in public. Will try harder to apply this to my channel soon
59
Trying something new
64
This is big news, it’s not everyday that a novel dataset gets open sourced
Did you notice how most video AIs are commercial, closed-source & secret? It sucks and a big reason for the lack of open-source & transparency is that there are very few high-quality open video datasets. The @huggingface team is changing this today by releasing FineVideo, a high-quality annotated video dataset with a creative commons license. Time to build!
93
decoder.sh retweeted
magnet:?xt=urn:btih:7278e625de2b1da598b23954c13933047126238a&dn=pixtral-12b-240910&tr=udp%3A%2F%2Ftracker.opentrackr.org%3A1337%2Fannounce&tr=udp%3A%2F%https://teapawt.net/t.co/2UepcMHjvL%3A1337%2Fannounce&tr=http%3A%2F%https://teapawt.net/t.co/NsTRgy7h8S%3A80%2Fannounce
301
569
442
5,000
1,140,902
What should I use it for?
74
ML folks (but MLX folks sorry), what is your OS? Windows plus WSL? Or just straight up Ubuntu / Deb?
1
70
I’ve previously used a Mac for inference, but I’m building a PC (SFF anyone) and would know what the standard is
45