Alvaro Bartolome

1 July

GLM 5.2, open frontier-scale intelligence on Microsoft Foundry with AMD MI300X. Running a Codex goal with an open model never felt this good!

1 July

GLM 5.2, open frontier-scale intelligence on Microsoft Foundry with AMD MI300X. Running a Codex goal with an open model never felt this good!

1 July

I wrote a post so you can try GLM 5.2 on Microsoft Foundry from Codex too. alvarobartt.com/goal-glm-5.2...

Alvaro Bartolome reposted

Tom Aarsen

@tomaarsen.com

9 Apr

🌐 I've just released Sentence Transformers v5.4: we're going fully multimodal for embeddings & reranking! Also featuring a modular CrossEncoder, and automatic Flash Attention 2 input flattening. Highlights in 🧵

More embedding models and an even more reliable inference engine is what you get with @hf.co Text Embeddings Inference v1.9.0 💥 More in the thread 🧵

- 💚 NVIDIA Blackwell support, ready for next-gen GPUs as B200, GB200, or RTX 50-series.

github.comRelease v1.9.0 · huggingface/text-embeddings-inferenceWhat's changed? 🚨 Breaking changes Default HiddenAct::Gelu to GeLU + tanh in favour of GeLU erf by @vrdn-23 in #753 Default GeLU implementation is now GeLU + tanh approximation instead of exact...

And much more additions, improvements and fixes, that couldn't have been possible without the community support and contributions 🙏🏻 github.com/huggingface/...

More embedding models and an even more reliable inference engine is what you get with @hf.co Text Embeddings Inference v1.9.0 💥 More in the thread 🧵

- 🦙 Add support for Meta Llama 2 and 3 architectures with Flash Attention support, enabling embedding models as NVIDIA Llama Embed Nemotron.

- 🔄 Add bidirectional attention support for 3, enabling newer embedding models as Voyage AI by MongoDB.

More embedding models and an even more reliable inference engine is what you get with @hf.co Text Embeddings Inference v1.9.0 💥 More in the thread 🧵

- 🎉 Add support for Microsoft Deberta V2 and V3, for both feature-extraction (and sentence-similarity) and text-classification, enabling models as Meta Llama Prompt Guard.

5 Jan

`hf-mem` is all you need to estimate the required VRAM for inference of any model on @huggingface based on Safetensors metadata. - Written in Python - Lightweight, only depends on `httpx` - Runs w/ `uvx` as `uvx hf-mem ...` - Works with any Safetensors repository - Output inspired by usgraphics.com

github.com/alvarobartt/...

5 Jan

13 Mar 2025

🧨 I built something with #Zig! `tokeni.zig` is a std-only implementation of the Byte Pair Encoding (BPE) algorithm in Zig for tokenizing sequences of text, used by OpenAI (among many others) to tokenize the text when pretraining their large language models! github.com/alvarobartt/...

10 Feb 2025

For anyone interested in Zig I wrote a small post titled "How to read and parse JSON with Zig 0.13" that explains how to read JSON from a file with keys with different value types and how to access those values.

https://alvarobartt.me/how-to-read-and-parse-json-with-zig-0-13

10 Feb 2025

alvarobartt.meHow to read and parse JSON with Zig 0.13How to read and parse JSON with Zig 0.13 by alvarobartt

thesquareplanet.comAbout 40 hoursI often hear, especially from folks working at younger companies and the tech giants, that the 40 hour workweek is a lie. That “the company” secretly (or not so secretly) always wants you to work more...

3 Feb 2025

love this quote "working smarter helps, but the real superpower is resting smarter" a highly recommended read!

Jon Gjengset@jonhoo.eu

Perhaps unsurprisingly, I have a bunch of Opinions™ on work-life balance and time management. Lately, I've been particularly bugged by the common belief that employer incentives don't align with employee well-being. So, I wrote a thing (with graphs!): thesquareplanet.com/blog/about-4...

31 Jan 2025

🤗 Here's a simple script that calculates the required VRAM for serving DeepSeek R1 from @huggingface Hub safetensor's metadata! P.S. The result of the script above is: "model_id='deepseek-ai/DeepSeek-R1' requires memory=756.716GB"

31 Jan 2025

hmm refactoring in zig is not as easy as it's in rust, even though seems fairly common too, right? or is it just me? 🤔

github.comGitHub - huggingface/open-r1: Fully open reproduction of DeepSeek-R1Fully open reproduction of DeepSeek-R1. Contribute to huggingface/open-r1 development by creating an account on GitHub.

29 Jan 2025

stuff that matters takes time

Alvaro Bartolome reposted

Quentin Gallouédec

@qgallouedec.hf.co

25 Jan 2025

Last moments of closed-source AI 🪦 : Hugging Face is openly reproducing the pipeline of 🐳 DeepSeek-R1. Open data, open training. open models, open collaboration. 🫵 Let's go! github.com/huggingface/...

23 Jan 2025

🐐 DeepSeek is not on the @hf.co Hub to take part, they are there to take over! Amazing stuff from the DeepSeek team, ICYMI they recently released some reasoning models (DeepSeek-R1 and DeepSeek-R1-Zero), fully open-source, their performance is on par with OpenAI-o1 and it's MIT licensed!

huggingface.coDeepSeek-R1 - a deepseek-ai CollectionWe’re on a journey to advance and democratize artificial intelligence through open source and open science.

23 Jan 2025

Check DeepSeek-R1 collection on the Hugging Face Hub, with not just DeepSeek-R1 and DeepSeek-R1-Zero, but also distilled their reasoning patterns to fine-tune smaller models! huggingface.co/collections/...

23 Jan 2025

you can find so much gold in github gists wow, i was not a big fan because the discoverability doesn't seem great, but been exploring gists lately and so much gold stuff in there!

bit.lyHugging FaceHugging Face Email Forms

22 Jan 2025

in case anyone missed it, we're running a certified course on ai agents at hugging face starting on feb 2nd; the course is on how to build you own ai agents for different cool use cases built on top of open source! 👇 you can sign up in the link below, don't miss it! bit.ly/hf-learn-age...

22 Jan 2025

ok, here we go again 😅

27 Nov 2024

Not quite sure yet about how's following me here, but I may consider not just x-posting but also eventually post more random thoughts + content in Spanish, is that something you'd be interested in?

20 Nov 2024

here we go again! i work at hugging face and here you can expect posts about machine learning (llms mainly), some rust, some nvim nerdy stuff and anything related to hugging face 🤗 posting is not easy for me, but i’ll try to do better from now on, support is highly appreciated!

💡 Did you know that you can use over 13700 public open models and adapters on the @huggingface Hub for FREE? You just need a free account on the Hugging Face Hub (you can also subscribe to PRO to increase the requests per hour) More details on the thread 🧵

🔥 Finally, if you are willing to get started quickly and experiment with LLMs feel free to give the recently released Inference Playground a try! https://huggingface.co/playground

Read more about the Serverless Inference API in the documentation! https://huggingface.co/docs/api-inference

🔎 Now let's explore some of the different alternatives to run inference via the Serverless API! The most straightforward one is via the Hugging Face Hub available on the model card of the Serverless API supported models!

👨‍💻 Alternatively, you can also use the Serverless Inference API programmatically via cURL, the huggingface_hub Python SDK, the openai SDK for chat completion, and much more! Find all the alternatives at https://huggingface.co/docs/api-inference

❄️ Additionally, there are a bunch of models (around 1000) that are "Cold", meaning that those are not loaded in the Serverless API, but can be loaded when sending a request to them, also meaning that the first request may take a while until the model is loaded!

bsky.socialHow to set your domain as your handle - BlueskyUsing a domain as your handle helps with account identity, verification, and portability. Here's how to set your domain as your handle.

🔒 Before going on, you will first need to generate a Hugging Face fine-grained token with access to the Serverless API, as the requests need to be authenticated so keep the token safe and avoid exposing it!

You may be wondering how do you know what models out of the over a million publicly available on the Hub can be used via the Serverless API? Well, we have the "Warm" tag that indicates that a model is loaded in the Serverless API and ready to be used 🔥

Alvaro Bartolome reposted

Daniël de Kok

@danieldk.eu

18 Nov 2024

Bluesky pro-tip: you can set your domain as a handle: bsky.social/about/blog/4... Bonus: you can keep your handle if you ever want to move to another server in the future.