Sabitlenmiş Tweet
في الثريد ده حابب اشارك معاكم رود ماب للي عايز يدخل في مجال الAI مع مصادر ونصائح ازاي تطور من نفسك
إن شاء الله هتكون مفيدة جدا
Red@whoroou
هو لو عايزه ابدأ Ai بذاكر data science الاول ؟
العربية
Eyad Ayman 🦜🦙
745 posts
Eyad Ayman 🦜🦙
@eyad_aiman
Machine Learning | Gen AI
Machine Learning Katılım Kasım 2020
471 Takip Edilen645 Takipçiler
@JustOmar21 @MrXroboT والله افتكرت قصدك الYOLO algorithm
You Only Look Once😂😂
@MrXroboT في الكتب و المخطوطات القديمة كان هذا الإختصار يرمز إلي المقولة الفلسفية
You Only Live Once
أول مرة أعرف إنهم عملوها أحمر شفايف في مصر بس أنا جلوبال مش لوكل 😆😆😆😆
العربية
@itsmebandr @abofahadjr @M0ATH هو يهودي بس بيدعم فلسطين وبينتقض السياسة الصهيونية وبيدعم فلسطين من اكتوبر 2023 يعني من البداية خالص, اما موضوع بيدوفيلي ده كان اتهام من كيندريك لامار في خلافات معاه لكن مفيش اي دليل بيأكد ده او اي اتهام او اي ادانه او اي تحقيق اتفتح ف الموضوع ده مع دراك اصلا
العربية
@abofahadjr @M0ATH معلومة دريك يهودي وبيدوفيلي فكسمه حتى لو تكلم عن فلسطين
العربية
🔴: المغني الشهير "دريك" في احد اغاني البومه الجديد قام بانتقاد "دي جي خالد" الامريكي من اصول فلسطينية لعدم دعمه لفلسطين وقال…
"شعبك ما زال ينتظر فلسطين حرة، لكن يبدو أن الأمور ليست كلها أبيض وأسود ولا بالأحمر والأخضر"
العربية
Eyad Ayman 🦜🦙 retweetledi
We collaborated with NVIDIA to teach you how we made LLM training ~25% faster! 🚀
Learn how 3 optimizations help your home GPU train models faster:
1. Packed-sequence metadata caching
2. Double-buffered checkpoint reloads
3. Faster MoE routing
Guide: unsloth.ai/blog/nvidia-co…
English
Eyad Ayman 🦜🦙 retweetledi
🤗🤗🤗introducing Hugging Science -- the home of AI for science 🤗🤗🤗
open models and datasets are the powerhouse of science (see the PDB), but finding the models and data you actually need for your breakthrough is hard af
you shouldn't need to scrape arxiv, own your own wetlab, fight a custom HDF5 parser, build a fusion stellarator, and beg for compute before you've trained a single epoch
so we're changing that
we've put all the best science on @huggingface in one place:
- 78GB of genomics data
- 11TB of PDE simulations
- 100M cell profiles
- 9T DNA base pairs
- 13M molecular trajectories
- 400k medical QA pairs
and much more, all open, and all ready for training (+ you can also now filter and search by domain, task, and keyword)
we've put together all the biggest releases from our partners at NASA, Google, OpenAI, Meta FAIR, Arc Institute, Ginkgo, SandboxAQ, Proxima Fusion, NVIDIA, Ai2, OpenADMET, InstaDeep, Future House, Polymathic AI, LeMaterial, Earth Species Project, Merck, and Eve Bio
if you're not sure where you fit in -- work on open challenges for problems that matter: including fusion stellarator design, ADMET, antibody developability, multilingual medicine, catalysis and materials, and scientific reasoning.
we're already changing how science gets done:
a fusion startup needed a benchmark for stellarator plasma confinement that didn't exist. @proximafusion shipped ConStellaration on Hugging Science: a leaderboard, dataset, and eval metrics, all in one place.
a drug discovery team wanted to predict hPXR induction. OpenADMET put up a blind challenge: 11,000+ compounds assayed at Octant, 513 held out, two tracks (pEC50 + structure). Anyone in the world can train and submit.
an antibody team at @Ginkgo released GDPa1, a developability dataset for stability, manufacturability, and immunogenicity prediction, with a live leaderboard scoring every submission.
if you know a problem the ML community should be working on, let us know. make a challenge! this is about putting all the tools for solving science in one place. so we can hillclimb!
→ huggingscience.co
English
Eyad Ayman 🦜🦙 retweetledi
First open-weight model from @poolsideai! Apache license, and available on Ollama to try.
👇👇👇 model page
Poolside@poolsideai
Today we’re releasing Laguna XS.2, Poolside’s first open-weight model. It’s a 33B total / 3B active MoE model built for agentic coding and long-horizon tasks. Trained fully in-house on our own stack. Runs on a single GPU. Released under Apache 2.0. Links 👇 Weights: huggingface.co/poolside/Lagun… API: platform.poolside.ai Blog: poolside.ai/blog/laguna-a-…
English
Eyad Ayman 🦜🦙 retweetledi
🐳 is back!
DeepSeek@deepseek_ai
🚀 DeepSeek-V4 Preview is officially live & open-sourced! Welcome to the era of cost-effective 1M context length. 🔹 DeepSeek-V4-Pro: 1.6T total / 49B active params. Performance rivaling the world's top closed-source models. 🔹 DeepSeek-V4-Flash: 284B total / 13B active params. Your fast, efficient, and economical choice. Try it now at chat.deepseek.com via Expert Mode / Instant Mode. API is updated & available today! 📄 Tech Report: huggingface.co/deepseek-ai/De… 🤗 Open Weights: huggingface.co/collections/de… 1/n
English
Eyad Ayman 🦜🦙 retweetledi
2-bit Qwen3.6-27B GGUF made 26 tool calls, triaged 15 GitHub issues and fixed, tested + reproed our repo’s 3 latest issues. 🔥
Try this locally in Unsloth Studio with just 12GB RAM. Studio also has a new look!
GitHub: github.com/unslothai/unsl…
English
Eyad Ayman 🦜🦙 retweetledi
Eyad Ayman 🦜🦙 retweetledi
This is crazy. ml-intern just passed the @huggingface internship test in 15 minutes.
The task: replicate a research baseline from a DeepMind paper on test-time compute scaling.
Here's what the agent did:
- Read the DeepMind paper, dug into Appendix E, picked the right scoring strategy (last-step PRM prediction, not min/product — matters a lot)
- Ran 16 solutions per problem through a reward model, grouped by final answer, summed scores, picked the winner
- Went from 45% → 65% accuracy. +20pp over greedy. Beat majority vote AND standard Best-of-N
- Generated 4 plots, pushed a full results dataset to the Hub, deployed a Docker Space on T4 GPU
The funniest part: it cited @lewtun's code snippet as a reference. The intern is already citing your bosses work to look good.
It also wrote the README, documented every design decision with paper citations, and added a co-authorship note explaining exactly which parts it did.
here's everything it produced:
full docs: huggingface.co/blog/cmpatino/…
trained model: huggingface.co/cmpatino/math5…
dataset: huggingface.co/datasets/cmpat…
the take home test: github.com/huggingface/po…
Aksel@akseljoonas
Introducing ml-intern, the agent that just automated the post-training team @huggingface It's an open-source implementation of the real research loop that our ML researchers do every day. You give it a prompt, it researches papers, goes through citations, implements ideas in GPU sandboxes, iterates and builds deeply research-backed models for any use case. All built on the Hugging Face ecosystem. It can pull off crazy things: We made it train the best model for scientific reasoning. It went through citations from the official benchmark paper. Found OpenScience and NemoTron-CrossThink, added 7 difficulty-filtered dataset variants from ARC/SciQ/MMLU, and ran 12 SFT runs on Qwen3-1.7B. This pushed the score 10% → 32% on GPQA in under 10h. Claude Code's best: 22.99%. In healthcare settings it inspected available datasets, concluded they were too low quality, and wrote a script to generate 1100 synthetic data points from scratch for emergencies, hedging, multilingual etc. Then upsampled 50x for training. Beat Codex on HealthBench by 60%. For competitive mathematics, it wrote a full GRPO script, launched training with A100 GPUs on hf.co/spaces, watched rewards claim and then collapse, and ran ablations until it succeeded. All fully backed by papers, autonomously. How it works? ml-intern makes full use of the HF ecosystem: - finds papers on arxiv and hf.co/papers, reads them fully, walks citation graphs, pulls datasets referenced in methodology sections and on hf.co/datasets - browses the Hub, reads recent docs, inspects datasets and reformats them before training so it doesn't waste GPU hours on bad data - launches training jobs on HF Jobs if no local GPUs are available, monitors runs, reads its own eval outputs, diagnoses failures, retrains ml-intern deeply embodies how researchers work and think. It knows how data should look like and what good models feel like. Releasing it today as a CLI and a web app you can use from your phone/desktop. CLI: github.com/huggingface/ml… Web + mobile: huggingface.co/spaces/smolage… And the best part? We also provisioned 1k$ GPU resources and Anthropic credits for the quickest among you to use.
English
Eyad Ayman 🦜🦙 retweetledi
Unplugging completely! No WiFi and zero notifications. A great way to get deep focus on a project.
Here is a walkthrough showing how to run Gemma 4 (26B A4B) fully offline with LM Studio & OpenCode to parse PDFs, ask questions, and build sites 100% locally.
English
Eyad Ayman 🦜🦙 retweetledi
I tested @huggingface ml-intern, given the prompt
"Fine-tune a Segment Anything Model (SAM) on a useful medical dataset. Train the model, and provide a comprehensive tutorial in a Jupyter Notebook file. Additionally, create a Hugging Face article/blog post documenting everything you have done."
It did it all autonomously:
- Researched via hf_papers & searched GitHub/HF Hub
- Found an HF dataset & wrote the finetuning script
- Trained it using HF compute (took ~1 hour)
- Pushed the weights & wrote the article
Here are the model weights, code, and the blog it generated:
hf article
huggingface.co/Mayank022/blog…
model weights
huggingface.co/Mayank022/sam-…
Awesome stuff @akseljoonas , looking forward to use this. 🔥
Aksel@akseljoonas
Introducing ml-intern, the agent that just automated the post-training team @huggingface It's an open-source implementation of the real research loop that our ML researchers do every day. You give it a prompt, it researches papers, goes through citations, implements ideas in GPU sandboxes, iterates and builds deeply research-backed models for any use case. All built on the Hugging Face ecosystem. It can pull off crazy things: We made it train the best model for scientific reasoning. It went through citations from the official benchmark paper. Found OpenScience and NemoTron-CrossThink, added 7 difficulty-filtered dataset variants from ARC/SciQ/MMLU, and ran 12 SFT runs on Qwen3-1.7B. This pushed the score 10% → 32% on GPQA in under 10h. Claude Code's best: 22.99%. In healthcare settings it inspected available datasets, concluded they were too low quality, and wrote a script to generate 1100 synthetic data points from scratch for emergencies, hedging, multilingual etc. Then upsampled 50x for training. Beat Codex on HealthBench by 60%. For competitive mathematics, it wrote a full GRPO script, launched training with A100 GPUs on hf.co/spaces, watched rewards claim and then collapse, and ran ablations until it succeeded. All fully backed by papers, autonomously. How it works? ml-intern makes full use of the HF ecosystem: - finds papers on arxiv and hf.co/papers, reads them fully, walks citation graphs, pulls datasets referenced in methodology sections and on hf.co/datasets - browses the Hub, reads recent docs, inspects datasets and reformats them before training so it doesn't waste GPU hours on bad data - launches training jobs on HF Jobs if no local GPUs are available, monitors runs, reads its own eval outputs, diagnoses failures, retrains ml-intern deeply embodies how researchers work and think. It knows how data should look like and what good models feel like. Releasing it today as a CLI and a web app you can use from your phone/desktop. CLI: github.com/huggingface/ml… Web + mobile: huggingface.co/spaces/smolage… And the best part? We also provisioned 1k$ GPU resources and Anthropic credits for the quickest among you to use.
English
Eyad Ayman 🦜🦙 retweetledi
What does it take to run 3, 5, or even 10 concurrent instances of Gemma 4 locally?
We've open-sourced a demo letting you run multiple models side-by-side on your hardware.
Gemma 4 26B A4B easily runs 10+ concurrent requests on a MacBook Pro M4 Max at 18 tokens/sec per request.
English
@nurhan__ashraf كلامك صح والقرارات الشخصية متعلقة بظروف الشخص، بس أنا شايف ان انت ممكن تاخد الكلام اللي متناسب مع ظروفك شوية او الكلام اللي على هواك والنصائح في المطلق بتبقى مفيدة
بس هي فعلا كل واحد وظروفه والظروف هي اللي بتعكس القرار الشخصي ف كلامك صح
العربية
@eyad_aiman أنا شخص مش بيحب التبريرات الكتير وبذات في الحاجات المتعلقه بقرارك الشخصي مينفعش تشتغل في حته وتطلع تعملي فيديو تقولي ليه اشتغلت في المطلق !! تسافر في حته وتقولي وفيه فرق بالنسبالي بين التبرير وبين اني احكي تجربه شخصيه ، فكل الاحوال ده قرارات شخصيه وبس
العربية
بالنسبالي المشكله مش في أنت رأيك ايه يعني يعم عايز تتجوز جو اون مش عايز جو اون عايز تسافر جو اون مش عايز برضو جو اون اما ليه تاخد قرار وتطلع تبرر ليا انت عملت كدا ليه وليه عايز تقنعني بوجهه نظرك يعني الإشكاليه بالنسبالي مش فكره انك بتقولي وبس تؤ انت بتحاول تقنعني انك صح
العربية
Eyad Ayman 🦜🦙 retweetledi
Train Gemma 4 with reinforcement learning!
Unsloth just added GRPO support for Gemma 4. You can now RL fine-tune Google's latest model on a consumer GPU.
The example notebook teaches Gemma 4 to solve Sudoku puzzles autonomously. The model learns through trial and error with reward signals instead of memorizing solutions.
Here's what RL training does differently. Standard fine-tuning teaches models to predict the next word based on examples. RL teaches them to reason through problems by rewarding correct approaches and penalizing wrong ones.
When Gemma 4 places a Sudoku number correctly, it gets a reward. When it violates the rules, it gets penalized. Over many attempts, it learns the actual logic instead of pattern matching.
GRPO makes this efficient by removing two of the three models traditional PPO needed. It uses statistics from multiple generations instead.
What this means for Gemma 4: you only need 9GB VRAM to run RL training. That's a single consumer GPU, not a data center.
The workflow is simple. Define what counts as correct. For Sudoku, a valid solved puzzle. For code, passing tests. For math, the right answer. Gemma 4 generates attempts, gets scored, and learns from feedback.
You can use any verifiable task where you can programmatically check correctness.
Free Colab notebook available.
I've shared the link in the replies!
English
Eyad Ayman 🦜🦙 retweetledi
1/6 Introducing VimRAG: Our most capable multimodal RAG framework yet.
We designed this framework to tackle "state blindness" in multimodal RAG. By moving from linear histories to a Multimodal Memory Graph, we structure reasoning as a dynamic DAG to eliminate redundant searches and track exploration paths effectively.
To handle heavy visual data, we’ve integrated Graph-Modulated Visual Memory Encoding for adaptive token allocation, alongside GGPO for fine-grained Credit Assignment.
Our goal is to move beyond simple retrieval toward truly structured and reliable reasoning.
English
Eyad Ayman 🦜🦙 retweetledi
GLM-5.1 can now be run locally!🔥
GLM-5.1 is a new open model for SOTA agentic coding & chat.
We shrank the 744B model from 1.65TB to 220GB (-86%) via Dynamic 2-bit.
Runs on a 256GB Mac or RAM/VRAM setups.
Guide: unsloth.ai/docs/models/gl…
GGUF: huggingface.co/unsloth/GLM-5.…
Z.ai@Zai_org
Introducing GLM-5.1: The Next Level of Open Source - Top-Tier Performance: #1 in open source and #3 globally across SWE-Bench Pro, Terminal-Bench, and NL2Repo. - Built for Long-Horizon Tasks: Runs autonomously for 8 hours, refining strategies through thousands of iterations. Blog: z.ai/blog/glm-5.1 Weights: huggingface.co/zai-org/GLM-5.1 API: docs.z.ai/guides/llm/glm… Coding Plan: z.ai/subscribe Coming to chat.z.ai in the next few days.
English
Eyad Ayman 🦜🦙 retweetledi
Introducing GLM-5.1: The Next Level of Open Source
- Top-Tier Performance: #1 in open source and #3 globally across SWE-Bench Pro, Terminal-Bench, and NL2Repo.
- Built for Long-Horizon Tasks: Runs autonomously for 8 hours, refining strategies through thousands of iterations.
Blog: z.ai/blog/glm-5.1
Weights: huggingface.co/zai-org/GLM-5.1
API: docs.z.ai/guides/llm/glm…
Coding Plan: z.ai/subscribe
Coming to chat.z.ai in the next few days.
English
Eyad Ayman 🦜🦙 retweetledi
NEW paper on multi-agents from Stanford.
More agents, better results, right?
Not so fast.
This paper challenges a core assumption in the multi-agent hype by controlling for what most studies don't: total computation.
It compares single-agent and multi-agent LLM architectures on multi-hop reasoning under matched thinking-token budgets across different models.
The finding is clear:
Single-agent systems are more information-efficient when reasoning tokens are held constant. The authors also identify significant artifacts in API-based budget control that may artificially inflate multi-agent advantages.
Why does it matter?
Many reported multi-agent gains disappear once you account for unequal computation.
Before building a multi-agent system, check whether a single agent with the same token budget would do the job. This paper gives you the framework to make that call.
Paper: arxiv.org/abs/2604.02460
Learn to build effective AI agents in our academy: academy.dair.ai
English
Eyad Ayman 🦜🦙 retweetledi
There's no way 🤯
Zai has just released a new open source model which is competitive with Opus 4.6 and GPT-5.4...
And even better on some benchmarks!
- 5x cheaper than Opus 4.6
- 3x cheaper than GPT-5.4
You can even use it in Claude Code or OpenClaw.
Weights and more below
Z.ai@Zai_org
Introducing GLM-5.1: The Next Level of Open Source - Top-Tier Performance: #1 in open source and #3 globally across SWE-Bench Pro, Terminal-Bench, and NL2Repo. - Built for Long-Horizon Tasks: Runs autonomously for 8 hours, refining strategies through thousands of iterations. Blog: z.ai/blog/glm-5.1 Weights: huggingface.co/zai-org/GLM-5.1 API: docs.z.ai/guides/llm/glm… Coding Plan: z.ai/subscribe Coming to chat.z.ai in the next few days.
English