HF-Party (Hugging Face Party @ PyTorch Conference)

clem

posted an update about 16 hours ago

Post

1142

Cool to see @ylecun joining the top 10 of most followed on HF!

(and leaderboard by @mvaloatto is here: mvaloatto/TCTF)

1 reply

·

csabakecskemeti

posted an update about 24 hours ago

Post

378

Just wondering why the number of parameters changed in the model attributes/Model size from 685B to 684B after converting deepseek-ai/DeepSeek-V3-Base from FP8 to BF16:
DevQuasar/deepseek-ai.DeepSeek-V3-Base-bf16
and not just for me:
opensourcerelease/DeepSeek-V3-Base-bf16

??

csabakecskemeti

posted an update 3 days ago

Post

1432

Happy New Year, Huggingface community!
In 2025, I'll continue my quantization (and some fine-tuning) efforts to support the open-source AI and Make knowledge free for everyone.

https://huggingface.co/DevQuasar
https://devquasar.com/

1 reply

·

csabakecskemeti

posted an update 4 days ago

Post

2023

The deepseek-ai/DeepSeek-V3-Base
model has featured today on CNBC tech news. The whale made a splash by using FP8 and shrink the cost of training significantly!

https://youtu.be/NJljq429cGk?si=kgk-ogPTMfJKsaA2

3 replies

·

csabakecskemeti

posted an update 6 days ago

Post

1433

I've built a small utility to split safetensors file by file.
The issue/need came up when I've tried to convert the new Deepseek V3 model from FP8 to BF16.
The only Ada architecture GPU I have is an RTX 4080 and the 16GB vram was just wasn't enough for the conversion.

BTW: I'll upload the bf16 version here:
DevQuasar/deepseek-ai.DeepSeek-V3-Base-bf16
(it will take a while - days with my upload speed)
If anyone has access the resources to test it I'd appreciate a feedback if it's working or not.

The tool, is available from here:
https://github.com/csabakecskemeti/ai_utils/blob/main/safetensor_splitter.py
It's splitting every file to n pieces by the layers if possible, and create a new "model.safetensors.index.json" file.
I've tested it with Llama 3.1 8B and multiple split sizes, and validated by using inference pipeline.
use --help for usage
Please note current version expects the model is already multiple file and have a "model.safetensors.index.json" layer-safetensor mapping file.

mbrack

authored a paper 12 days ago

LLMs Lost in Translation: M-ALERT uncovers Cross-Linguistic Safety Gaps

Paper • 2412.15035 • Published 15 days ago • 4

m-ric

posted an update 15 days ago

Post

1898

After 6 years, BERT, the workhorse of encoder models, finally gets a replacement: 𝗪𝗲𝗹𝗰𝗼𝗺𝗲 𝗠𝗼𝗱𝗲𝗿𝗻𝗕𝗘𝗥𝗧! 🤗

We talk a lot about ✨Generative AI✨, meaning "Decoder version of the Transformers architecture", but this is only one of the ways to build LLMs: encoder models, that turn a sentence in a vector, are maybe even more widely used in industry than generative models.

The workhorse for this category has been BERT since its release in 2018 (that's prehistory for LLMs).

It's not a fancy 100B parameters supermodel (just a few hundred millions), but it's an excellent workhorse, kind of a Honda Civic for LLMs.

Many applications use BERT-family models - the top models in this category cumulate millions of downloads on the Hub.

➡️ Now a collaboration between Answer.AI and LightOn just introduced BERT's replacement: ModernBERT.

𝗧𝗟;𝗗𝗥:
🏛️ Architecture changes:
⇒ First, standard modernizations:
- Rotary positional embeddings (RoPE)
- Replace GeLU with GeGLU,
- Use Flash Attention 2
✨ The team also introduced innovative techniques like alternating attention instead of full attention, and sequence packing to get rid of padding overhead.

🥇 As a result, the model tops the game of encoder models:
It beats previous standard DeBERTaV3 for 1/5th the memory footprint, and runs 4x faster!

Read the blog post 👉 https://huggingface.co/blog/modernbert

1 reply

·

m-ric

posted an update 16 days ago

Post

2282

𝐇𝐮𝐠𝐠𝐢𝐧𝐠 𝐅𝐚𝐜𝐞 𝐫𝐞𝐥𝐞𝐚𝐬𝐞𝐬 𝐏𝐢𝐜𝐨𝐭𝐫𝐨𝐧, 𝐚 𝐦𝐢𝐜𝐫𝐨𝐬𝐜𝐨𝐩𝐢𝐜 𝐥𝐢𝐛 𝐭𝐡𝐚𝐭 𝐬𝐨𝐥𝐯𝐞𝐬 𝐋𝐋𝐌 𝐭𝐫𝐚𝐢𝐧𝐢𝐧𝐠 𝟒𝐃 𝐩𝐚𝐫𝐚𝐥𝐥𝐞𝐥𝐢𝐳𝐚𝐭𝐢𝐨𝐧 🥳

🕰️ Llama-3.1-405B took 39 million GPU-hours to train, i.e. about 4.5 thousand years.

👴🏻 If they had needed all this time, we would have GPU stories from the time of Pharaoh 𓂀: "Alas, Lord of Two Lands, the shipment of counting-stones arriving from Cathay was lost to pirates, this shall delay the building of your computing temple by many moons "

🛠️ But instead, they just parallelized the training on 24k H100s, which made it take just a few months.
This required parallelizing across 4 dimensions: data, tensor, context, pipeline.
And it is infamously hard to do, making for bloated code repos that hold together only by magic.

🤏 𝗕𝘂𝘁 𝗻𝗼𝘄 𝘄𝗲 𝗱𝗼𝗻'𝘁 𝗻𝗲𝗲𝗱 𝗵𝘂𝗴𝗲 𝗿𝗲𝗽𝗼𝘀 𝗮𝗻𝘆𝗺𝗼𝗿𝗲! Instead of building mega-training codes, Hugging Face colleagues cooked in the other direction, towards tiny 4D parallelism libs. A team has built Nanotron, already widely used in industry.
And now a team releases Picotron, a radical approach to code 4D Parallelism in just a few hundred lines of code, a real engineering prowess, making it much easier to understand what's actually happening!

⚡ 𝗜𝘁'𝘀 𝘁𝗶𝗻𝘆, 𝘆𝗲𝘁 𝗽𝗼𝘄𝗲𝗿𝗳𝘂𝗹:
Counting in MFU (Model FLOPs Utilization, how much the model actually uses all the compute potential), this lib reaches ~50% on SmolLM-1.7B model with 8 H100 GPUs, which is really close to what huge libs would reach. (Caution: the team is leading further benchmarks to verify this)

Go take a look 👉 https://github.com/huggingface/picotron/tree/main/picotron

1 reply

·

csabakecskemeti

posted an update 17 days ago

Post

1222

tiiuae Falcon3 10B Q8 playground:
https://huggingface.co/spaces/DevQuasar/Mi50

Also find my tiiuae Falcon3 Quant collection here:
https://huggingface.co/collections/DevQuasar/tiiuae-falcon3-676236626f3c57d1a19c6c1d

Enjoy!

clem

posted an update 17 days ago

Post

1666

Coming back to Paris Friday to open our new Hugging Face office!

We're at capacity for the party but add your name in the waiting list as we're trying to privatize the passage du Caire for extra space for robots 🤖🦾🦿

https://t.co/enkFXjWndJ

1 reply

·

csabakecskemeti

posted an update 20 days ago

Post

4488

The AMD Instinct MI50 (~$110) is surprisingly fast for inference Quantized models.

This runs a Llama 3.1 8B Q8 with Llama.cpp
https://huggingface.co/spaces/DevQuasar/Mi50

A little blogpost about the HW
http://devquasar.com/uncategorized/amd-radeon-instinct-mi50-cheap-inference/

m-ric

posted an update 21 days ago

Post

2174

𝗣𝗼𝘁𝗲𝗻𝘁𝗶𝗮𝗹 𝗽𝗮𝗿𝗮𝗱𝗶𝗴𝗺 𝘀𝗵𝗶𝗳𝘁 𝗶𝗻 𝗟𝗟𝗠𝘀: 𝗻𝗲𝘄 𝗽𝗮𝗽𝗲𝗿 𝗯𝘆 𝗠𝗲𝘁𝗮 𝗰𝗹𝗮𝗶𝗺𝘀 𝘁𝗵𝗮𝘁 𝘄𝗲 𝗰𝗮𝗻 𝗴𝗲𝘁 𝗿𝗶𝗱 𝗼𝗳 𝘁𝗼𝗸𝗲𝗻𝗶𝘇𝗲𝗿𝘀! 🥳

Current LLMs process text by first splitting it into tokens. They use a module named "tokenizer", that -spl-it-s- th-e- te-xt- in-to- arbitrary tokens depending on a fixed dictionnary.
On the Hub you can find this dictionary in a model's files under tokenizer.json.

➡️ This process is called BPE tokenization. It is suboptimal, everyone says it. It breaks text into predefined chunks that often fail to capture the nuance of language. But it has been a necessary evil in language models since their inception.

💥 In Byte Latent Transformer (BLT), Meta researchers propose an elegant solution by eliminating tokenization entirely, working directly with raw bytes while maintaining efficiency through dynamic "patches."

This had been tried before with different byte-level tokenizations, but it's the first time that an architecture of this type scales as well as BPE tokenization. And it could mean a real paradigm shift! 👏👏

🏗️ 𝗔𝗿𝗰𝗵𝗶𝘁𝗲𝗰𝘁𝘂𝗿𝗲:
Instead of a lightweight tokenizer, BLT has a lightweight encoder that process raw bytes into patches. Then the patches are processed by the main heavy-duty transformers as we do normally (but for patches of bytes instead of tokens), before converting back to bytes.

🧩 𝗗𝘆𝗻𝗮𝗺𝗶𝗰 𝗣𝗮𝘁𝗰𝗵𝗶𝗻𝗴:
Instead of fixed tokens, BLT groups bytes based on their predictability (measured by entropy) - using more compute for complex sequences and efficiently handling simple ones. This allows efficient processing while maintaining byte-level understanding.

I hope this breakthrough is confirmed and we can get rid of all the tokenizer stuff, it will make model handling easier!

Read their paper here 👉 https://dl.fbaipublicfiles.com/blt/BLT__Patches_Scale_Better_Than_Tokens.pdf

2 replies

·

m-ric

posted an update 23 days ago

Post

2488

💥 𝗚𝗼𝗼𝗴𝗹𝗲 𝗿𝗲𝗹𝗲𝗮𝘀𝗲𝘀 𝗚𝗲𝗺𝗶𝗻𝗶 𝟮.𝟬, 𝘀𝘁𝗮𝗿𝘁𝗶𝗻𝗴 𝘄𝗶𝘁𝗵 𝗮 𝗙𝗹𝗮𝘀𝗵 𝗺𝗼𝗱𝗲𝗹 𝘁𝗵𝗮𝘁 𝘀𝘁𝗲𝗮𝗺𝗿𝗼𝗹𝗹𝘀 𝗚𝗣𝗧-𝟰𝗼 𝗮𝗻𝗱 𝗖𝗹𝗮𝘂𝗱𝗲-𝟯.𝟲 𝗦𝗼𝗻𝗻𝗲𝘁! And they start a huge effort on agentic capabilities.

🚀 The performance improvements are crazy for such a fast model:
‣ Gemini 2.0 Flash outperforms the previous 1.5 Pro model at twice the speed
‣ Now supports both input AND output of images, video, audio and text
‣ Can natively use tools like Google Search and execute code

➡️ If the price is on par with previous Flash iteration ($0.30 / M tokens, to compare with GPT-4o's $1.25) the competition will have a big problem with this 4x cheaper model that gets better benchmarks 🤯

🤖 What about the agentic capabilities?

‣ Project Astra: A universal AI assistant that can use Google Search, Lens and Maps
‣ Project Mariner: A Chrome extension that can complete complex web tasks (83.5% success rate on WebVoyager benchmark, this is really impressive!)
‣ Jules: An AI coding agent that integrates with GitHub workflows

I'll be eagerly awaiting further news from Google!

Read their blogpost here 👉 https://blog.google/technology/google-deepmind/google-gemini-ai-update-december-2024/

m-ric

posted an update 24 days ago

Post

1799

𝐒𝐜𝐚𝐥𝐢𝐧𝐠 𝐥𝐚𝐰𝐬 𝐚𝐫𝐞 𝐧𝐨𝐭 𝐝𝐞𝐚𝐝 𝐲𝐞𝐭! New blog post suggests Anthropic might have an extremely strong Opus-3.5 already available, but is not releasing it to keep their edge over the competition. 🧐

❓Since the release of Opus-3.5 has been delayed indefinitely, there have been lots of rumors and articles about LLMs plateauing. Scaling laws, the main powering factor of the LLM competence increase, could have stopped, according to these rumors, being the cause of this stalling of progress.

These rumors were quickly denied by many people at the leading LLM labs, including OpenAI and Anthropic. But these people would be expected to hype the future of LLMs even if scaling laws really plateaued, so the jury is still out.

🗞️ This new article by Semianalysis (generally a good source, specifically on hardware) provides a counter-rumor that I find more convincing:

➡️ Maybe scaling laws still work, Opus-3.5 is ready and as good as planned, but they just don't release it because the synthetic data it helps provide can bring cheaper/smaller models Claude and Haiku up in performance, without risking to leak this precious high-quality synthetic data to competitors.

Time will tell! I feel like we'll know more soon.

Read the article: https://semianalysis.com/2024/12/11/scaling-laws-o1-pro-architecture-reasoning-infrastructure-orion-and-claude-3-5-opus-failures/

1 reply

·

julien-c

posted an update 24 days ago

Post

7923

After some heated discussion 🔥, we clarify our intent re. storage limits on the Hub

TL;DR:
- public storage is free, and (unless blatant abuse) unlimited. We do ask that you consider upgrading to PRO and/or Enterprise Hub if possible
- private storage is paid above a significant free tier (1TB if you have a paid account, 100GB otherwise)

docs: https://huggingface.co/docs/hub/storage-limits

We optimize our infrastructure continuously to scale our storage for the coming years of growth in Machine learning, to the benefit of the community 🔥

cc: @reach-vb @pierric @victor and the HF team

28 replies

·

m-ric

posted an update 25 days ago

Post

2226

Last week was crazy in OS AI, with important models and datasets releases every day.

Here are the most important ones I've pinned:

🌎 Cohere relased GLobal-MMLU, a multilingual version of MMLU, to evaluate AI models' world knowledge in many languages!

🦙 Meta released Llama-3.3-70B-Instruct, a 70B model that's on par with Llama-3.1-405B-Instruct, GPT-4o and Claude. Probably my new go-to for agentic workflows.

🔉 FishAudio released fish-speech-1.5, multilingual text to speech model

🎨 Microsoft Research released TRELLIS, an extremely impressive image-to-3D model, which you can try here: JeffreyXiang/TRELLIS

📚 Yesterday, Hugging Face release FineWeb 2, a new version that extends the previous FineWeb to over 1000 languages, including extended coverage in Russina, Mandarin, German, Japanese, Spanish, French, so a huge, high-quality dataset of > 3 trillion words! HuggingFaceFW/fineweb-2

Now let's go build to make this week as productive as last one!

danielhanchen

posted an update 28 days ago

Post

1454

I uploaded GGUFs, 4bit bitsandbytes and full 16bit precision weights for Llama 3.3 70B Instruct are here: unsloth/llama-33-all-versions-67535d7d994794b9d7cf5e9f

You can also finetune Llama 3.3 70B in under 48GB of VRAM with Unsloth!
GGUFs: unsloth/Llama-3.3-70B-Instruct-GGUF
BnB 4bit: unsloth/Llama-3.3-70B-Instruct-bnb-4bit
16bit: unsloth/Llama-3.3-70B-Instruct

1 reply

·

m-ric

posted an update about 1 month ago

Post

1474

𝗦𝗵𝗼𝘄𝗨𝗜: 𝗮 𝘀𝗺𝗮𝗹𝗹 𝗲𝗻𝗱-𝘁𝗼-𝗲𝗻𝗱 𝗮𝗴𝗲𝗻𝘁 𝘁𝗵𝗮𝘁 𝗰𝗮𝗻 𝗻𝗮𝘃𝗶𝗴𝗮𝘁𝗲 𝗮𝗻𝘆 𝗨𝗜 𝗮𝗻𝗱 𝗼𝘂𝘁𝗽𝗲𝗿𝗳𝗼𝗿𝗺𝘀 𝗺𝘂𝗰𝗵 𝗯𝗶𝗴𝗴𝗲𝗿 𝘀𝘆𝘀𝘁𝗲𝗺𝘀! 📲

A team from NUS and Microsoft just released an agent that can act on any UI (Desktop, Android, Web) without needing additional text information. It works extremely well : they applied their method on a tiny Qwen2-VL-2B, and they managed to beat methods that use either much more powerful vision models (like GPT-4V) without using any additional info (e.g. leveraging the DOM of a webpage) like previous methods did ! 👏👏

They started from the idea that most existing methods rely heavily on text, which makes them less generalizable, while letting aside rich UI structure that user actually rely on when navigating this interfaces.

⚙️ They put several good ideas to work:

💡 Simplify screenshots to the max:
They prune a lot the heavy visual content of UI screenshots, by removing cloned image patches (like any vast patch of the same color will be reduced to a small patch, while maintaining positional embeddings), then group patches from the same GUI elements together to simplify even further

💡 Build a truly generalist dataset:
To train a general UI agent, you need trajectories from each possible UI, and express them in a common language. Authors merge datasets like OmniAct for Desktop, Mind2Web for websites, AMEX for Android trajectories to create a high-quality and diverse dataset.

➡️ Nice results ensued:
They fine-tune a tiny Qwen-2-VL-2B on their method, and it reaches SOTA on several task (element identification, web navigation), even beating methods that either use additional info from the DOM or use much bigger VLMS like GPT-4v! 🏆

And performance could certainly jump with a slightly bigger vision model. Let's hope the community builds this soon! 🚀

Paper added to my "Agents" collection 👉 m-ric/agents-65ba776fbd9e29f771c07d4e

m-ric

posted an update about 1 month ago

Post

1211

Need a measurement for traction of a GitHub repo, a more reliable one than Github star history? (which is a bit too hype-driven) 📈

➡️ I've made a Space to visualize PyPI downloads.

Try it here 👉 m-ric/package-download-history

1 reply

·

m-ric

posted an update about 1 month ago

Post

1276

🤖 𝗔𝗱𝗼𝗯𝗲'𝘀 𝗰𝗼𝗱𝗲-𝗴𝗲𝗻𝗲𝗿𝗮𝘁𝗶𝗻𝗴 𝗮𝗴𝗲𝗻𝘁 𝗿𝗲𝗮𝗰𝗵𝗲𝘀 𝘁𝗵𝗲 𝘁𝗼𝗽 𝗼𝗳 𝗚𝗔𝗜𝗔 𝗹𝗲𝗮𝗱𝗲𝗿𝗯𝗼𝗮𝗿𝗱 - and their paper cites my work!

💡 Reminder: In short, Agentic systems are a vehicle in which you put your LLM to allow it access to the outside world.

➡️ The team of researchers at Adobe started from the idea that current agentic systems lack the ability to define their own tools. So they decided to make an agent that writes actions as code, thus allowing it to write python functions that can be re-used later as tools!

Here's what the LLM generations can look like with the proper prompt:

Thought: I need to access the excel file using a different method.
Action:

def access_excel_file(file_path)
	... # rest of the code (the agent does writes it, but I don't have room in this post)
	return rows

Then your system executes this and appends the observation to the agent's memory.

Why is this code formulation better than classical tool use formulation as JSON? The paper explains:

"Most existing work uses text or JSON as the representation of actions, which significantly lacks the two criteria mentioned earlier: generality and composability. In contrast, DynaSaur can utilize available actions or create new ones if necessary, using code as a unified representation. In principle, acting with code enables agents to solve any Turing-complete problem."

The idea of using code is not new: in fact, we do it in transformers.agents (thus the citation that I got). They implementation adds further refinements, like using RAG to retrieve relevant functions before generating an action, which increases performance further.

And they observe that code agents perform much better, reaching the top of GAIA leaderboard! 🥇

Go take a look, it's really clear and informative!

Paper added to my agents collection 👉 m-ric/agents-65ba776fbd9e29f771c07d4e

Hugging Face Party @ PyTorch Conference

AI & ML interests

Recent Activity

HF-Party's activity

LLMs Lost in Translation: M-ALERT uncovers Cross-Linguistic Safety Gaps

AI & ML interests

Recent Activity

Team members 181

HF-Party's activity