social-post-explorers (Social Post Explorers)

posted an update 6 days ago

Post

978

Hey there, 𝗶𝗻𝗴𝗲𝘀𝘁-𝗮𝗻𝘆𝘁𝗵𝗶𝗻𝗴 𝘃𝟭.𝟬.𝟬 just dropped with major changes:

✅ Embeddings: now works with Sentence Transformers, Jina AI, Cohere, OpenAI, and Model2Vec
All powered via 𝗖𝗵𝗼𝗻𝗸𝗶𝗲’𝘀 𝗔𝘂𝘁𝗼𝗘𝗺𝗯𝗲𝗱𝗱𝗶𝗻𝗴𝘀.
No more local-only limitations 🙌
✅ Vector DBs: now supports 𝗮𝗹𝗹 𝗟𝗹𝗮𝗺𝗮𝗜𝗻𝗱𝗲𝘅-𝗰𝗼𝗺𝗽𝗮𝘁𝗶𝗯𝗹𝗲 𝗯𝗮𝗰𝗸𝗲𝗻𝗱𝘀
Think: Qdrant, Pinecone, Weaviate, Milvus, etc.
No more bottlenecks🔓
✅ File parsing: now plugs into any 𝗟𝗹𝗮𝗺𝗮𝗜𝗻𝗱𝗲𝘅-𝗰𝗼𝗺𝗽𝗮𝘁𝗶𝗯𝗹𝗲 𝗱𝗮𝘁𝗮 𝗹𝗼𝗮𝗱𝗲𝗿
Using LlamaParse, Docling or your own setup? You’re covered.
Curious of knowing more? Try it out! 👉 https://github.com/AstraBert/ingest-anything

as-cle-bert

posted an update 7 days ago

Post

1819

One of the biggest challenges I've been facing since I started developing [𝐏𝐝𝐟𝐈𝐭𝐃𝐨𝐰𝐧](https://github.com/AstraBert/PdfItDown) was handling correctly the conversion of files like Excel sheets and CSVs: table conversion was bad and messy, almost unusable for downstream tasks🫣

That's why today I'm excited to introduce 𝐫𝐞𝐚𝐝𝐞𝐫𝐬, the new feature of PdfItDown v1.4.0!🎉

With 𝘳𝘦𝘢𝘥𝘦𝘳𝘴, you can choose among three (for now👀) flavors of text extraction and conversion to PDF:

- 𝗗𝗼𝗰𝗹𝗶𝗻𝗴, which does a fantastic work with presentations, spreadsheets and word documents🦆

- 𝗟𝗹𝗮𝗺𝗮𝗣𝗮𝗿𝘀𝗲 by LlamaIndex, suitable for more complex and articulated documents, with mixture of texts, images and tables🦙

- 𝗠𝗮𝗿𝗸𝗜𝘁𝗗𝗼𝘄𝗻 by Microsoft, not the best at handling highly structured documents, by extremly flexible in terms of input file format (it can even convert XML, JSON and ZIP files!)✒️

You can use this new feature in your python scripts (check the attached code snippet!😉) and in the command line interface as well!🐍

Have fun and don't forget to star the repo on GitHub ➡️ https://github.com/AstraBert/PdfItDown

Xenova

posted an update 10 days ago

Post

5441

Introducing the ONNX model explorer: Browse, search, and visualize neural networks directly in your browser. 🤯 A great tool for anyone studying Machine Learning! We're also releasing the entire dataset of graphs so you can use them in your own projects! 🤗

Check it out! 👇
Demo: onnx-community/model-explorer
Dataset: onnx-community/model-explorer
Source code: https://github.com/xenova/model-explorer

as-cle-bert

posted an update 13 days ago

Post

2886

Ever dreamt of ingesting into a vector DB that pile of CSVs, Word documents and presentations laying in some remote folders on your PC?🗂️
What if I told you that you can do it within three to six lines of code?🤯
Well, with my latest open-source project, 𝐢𝐧𝐠𝐞𝐬𝐭-𝐚𝐧𝐲𝐭𝐡𝐢𝐧𝐠 (https://github.com/AstraBert/ingest-anything), you can take all your non-PDF files, convert them to PDF, extract their text, chunk, embed and load them into a vector database, all in one go!🚀
How? It's pretty simple!
📁 The input files are converted into PDF by PdfItDown (https://github.com/AstraBert/PdfItDown)
📑 The PDF text is extracted using LlamaIndex readers
🦛 The text is chunked exploiting Chonkie
🧮 The chunks are embedded thanks to Sentence Transformers models
🗄️ The embeddings are loaded into a Qdrant vector database

And you're done!✅
Curious of trying it? Install it by running:

𝘱𝘪𝘱 𝘪𝘯𝘴𝘵𝘢𝘭𝘭 𝘪𝘯𝘨𝘦𝘴𝘵-𝘢𝘯𝘺𝘵𝘩𝘪𝘯𝘨

And you can start using it in your python scripts!🐍
Don't forget to star it on GitHub and let me know if you have any feedback! ➡️ https://github.com/AstraBert/ingest-anything

5 replies

·

natolambert

authored a paper 16 days ago

Reinforcement Learning from Human Feedback

Paper • 2504.12501 • Published 22 days ago • 3

as-cle-bert

posted an update 18 days ago

Post

2974

Finding a job that matches with our resume shouldn't be difficult, especially now that we have AI... And still, we're drowning in unclear announcements, jobs whose skill requirements might not really fit us, and tons of material😵‍💫
That's why I decided to build 𝐑𝐞𝐬𝐮𝐦𝐞 𝐌𝐚𝐭𝐜𝐡𝐞𝐫 (https://github.com/AstraBert/resume-matcher), a fully open-source application that scans your resume and searches the web for jobs that match with it!🎉
The workflow is very simple:
🦙 A LlamaExtract agent parses the resume and extracts valuable data that represent your profile
🗄️The structured data are passed on to a Job Matching Agent (built with LlamaIndex😉) that uses them to build a web search query based on your resume
🌐 The web search is handled by Linkup, which finds the top matches and returns them to the Agent
🔎 The agent evaluates the match between your profile and the jobs, and then returns a final answer to you

So, are you ready to find a job suitable for you?💼 You can spin up the application completely locally and with Docker, starting from the GitHub repo ➡️ https://github.com/AstraBert/resume-matcher
Feel free to leave your feedback and let me know in the comments if you want an online version of Resume Matcher as well!✨

2 replies

·

wanghaofan

authored a paper 21 days ago

InstantCharacter: Personalize Any Characters with a Scalable Diffusion Transformer Framework

Paper • 2504.12395 • Published 22 days ago • 17

ChuckMcSneed

posted an update 21 days ago

Post

698

Okay, folks, I need some help with this darn internet thing! My son, Timmy, showed me this interesting… forum thingy. He called it "/lmg/" and said it was the place to talk about… well, let's just say important matters 😉.

Timmy says something happened, though! He keeps mumbling about "Soy Jacks," "4chan is dead" and "hacked servers."

So, is this "/lmg/" thing GONE forever? Or did it move somewhere else? Timmy isn't being very helpful, and I'm sure some of you bright young minds on here probably know! I want to learn more and I really liked it there!

Thanks in advance for any help!

---

God bless America 🇺🇸
#WWG1WGA

5 replies

·

Xenova

posted an update 22 days ago

Post

2549

Reasoning models like o3 and o4-mini are advancing faster than ever, but imagine what will be possible when they can run locally in your browser! 🤯

Well, with 🤗 Transformers.js, you can do just that! Here's Zyphra's new ZR1 model running at over 100 tokens/second on WebGPU! ⚡️

Giving models access to browser APIs (like File System, Screen Capture, and more) could unlock an entirely new class of web experiences that are personalized, interactive, and run locally in a secure, sandboxed environment.

For now, try out the demo! 👇
webml-community/Zyphra-ZR1-WebGPU

1 reply

·

soldni

authored a paper 29 days ago

OLMoTrace: Tracing Language Model Outputs Back to Trillions of Training Tokens

Paper • 2504.07096 • Published 29 days ago • 73

yuchenlin

authored a paper 29 days ago

CrossWordBench: Evaluating the Reasoning Capabilities of LLMs and LVLMs with Controllable Puzzle Generation

Paper • 2504.00043 • Published Mar 30 • 9

Xenova

authored a paper about 1 month ago

SmolVLM: Redefining small and efficient multimodal models

Paper • 2504.05299 • Published Apr 7 • 180

loubnabnl

authored a paper about 1 month ago

SmolVLM: Redefining small and efficient multimodal models

Paper • 2504.05299 • Published Apr 7 • 180

as-cle-bert

posted an update about 1 month ago

Post

2944

Llama-4 is out and I couldn't resist but to cook something with it... So I came up with 𝐋𝐥𝐚𝐦𝐚𝐑𝐞𝐬𝐞𝐚𝐫𝐜𝐡𝐞𝐫 (https://llamaresearcher.com), your deep-research AI companion!🔎

The workflow behind 𝗟𝗹𝗮𝗺𝗮𝗥𝗲𝘀𝗲𝗮𝗿𝗰𝗵𝗲𝗿 is simple:
💬 You submit a query
🛡️ Your query is evaluated by Llama 3 guard model, which deems it safe or unsafe
🧠 If your query is safe, it is routed to the Researcher Agent
⚙️ The Researcher Agent expands the query into three sub-queries, with which to search the web
🌐 The web is searched for each of the sub-queries
📊 The retrieved information is evaluated for relevancy against your original query
✍️ The Researcher Agent produces an essay based on the information it gathered, paying attention to referencing its sources

The agent itself is also built with easy-to-use and intuitive blocks:
🦙 LlamaIndex provides the agentic architecture and the integrations with the language models
⚡Groq makes Llama-4 available with its lightning-fast inference
🔎 Linkup allows the agent to deep-search the web and provides sourced answers
💪 FastAPI does the heavy loading with wrapping everything within an elegant API interface
⏱️ Redis is used for API rate limiting
🎨 Gradio creates a simple but powerful user interface

Special mention also to Lovable, which helped me build the first draft of the landing page for LlamaResearcher!💖

If you're curious and you want to try LlamaResearcher, you can - completely for free and without subscription - for 30 days from now ➡️ https://llamaresearcher.com
And if you're like me, and you like getting your hands in code and build stuff on your own machine, I have good news: this is all open-source, fully reproducible locally and Docker-ready🐋
Just go to the GitHub repo: https://github.com/AstraBert/llama-4-researcher and don't forget to star it, if you find it useful!⭐

As always, have fun and feel free to leave your feedback✨

2 replies

·

grimjim

posted an update about 1 month ago

Post

1673

I recently have been looking at a paper titled "Why Warmup the Learning Rate? Underlying Mechanisms and Improvements", by Dayal Singh Kalra and Maissam Barkeshli, and was struck by "warmup" being analogous to simulated annealing.
https://arxiv.org/abs/2406.09405
Taking the physical analogy further, the "warmup" is a stochastic process to knock the system out of current local minima, allowing easier transition toward newer minima. It works because it reduces "fit" and therefore "friction".

as-cle-bert

posted an update about 1 month ago

Post

742

I heard someone saying 𝘃𝗼𝗶𝗰𝗲 assistants are the future, and someone else that 𝗠𝗖𝗣 will rule the AI world... So I decided to combine both!🚀

Meet 𝐓𝐲𝐒𝐕𝐀 (𝗧𝘆pe𝗦cript 𝗩oice 𝗔ssistant, https://github.com/AstraBert/TySVA), your (speaking) AI companion for everyday TypeScript programming tasks!🎙️

TySVA is a skilled TypeScript expert and, to provide accurate and up-to-date responses, she leverages the following workflow:
🗣️ If you talk to her, she converts the audio into a textual prompt, and use it a starting point to answer your questions (if you send a message, she'll use directly that💬)
🧠 She can solve your questions by (deep)searching the web and/or by retrieving relevant information from a vector database containing TypeScript documentation. If the answer is simple, she can also reply directly (no tools needed!)
🛜 To ease her life, TySVA has all the tools she needs available through Model Context Protocol (MCP)
🔊 Once she's done, she returns her answer to you, along with a voice summary of what she did and what solution she found

But how does she do that? What are her components?🤨

📖 Qdrant + HuggingFace give her the documentation knowledge, providing the vector database and the embeddings
🌐 Linkup provides her with up-to-date, grounded answers, connecting her to the web
🦙 LlamaIndex makes up her brain, with the whole agentic architecture
🎤 ElevenLabs gives her ears and mouth, transcribing and producing voice inputs and outoputs
📜 Groq provides her with speech, being the LLM provider behind TySVA
🎨 Gradio+FastAPI make up her face and fibers, providing a seamless backend-to-frontend integration

If you're now curious of trying her, you can easily do that by spinning her up locally (and with Docker!🐋) from the GitHub repo ➡️ https://github.com/AstraBert/TySVA

And feel free to leave any feedback!✨

as-cle-bert

posted an update about 2 months ago

Post

633

Drowning in handouts, documents and presentations from your professors and not knowing where to start?🌊😵‍💫
Well, I might have a tool for you: 𝐏𝐝𝐟𝟐𝐍𝐨𝐭𝐞𝐬 (https://github.com/AstraBert/pdf2notes) is an 𝗔𝗜-𝗽𝗼𝘄𝗲𝗿𝗲𝗱, 𝗼𝗽𝗲𝗻-𝘀𝗼𝘂𝗿𝗰𝗲 solution that lets you turn your unstructured and chaotic PDFs into nice and well-ordered notes in a matter of seconds!📝

𝗛𝗼𝘄 𝗱𝗼𝗲𝘀 𝗶𝘁 𝘄𝗼𝗿𝗸?
📄 You first upload a document
⚙️ LlamaParse by LlamaIndex extracts the text from the document, using DeepMind's Gemini 2 Flash to perform multi-modal parsing
🧠 Llama-3.3-70B by Groq turns the extracted text into notes!

The notes are not perfect or you want more in-depth insights? No problem:
💬 Send a direct message to the chatbot
⚙️ The chatbot will retrieve the chat history from a Postgres database
🧠 Llama-3.3-70B will produce the answer you need

All of this is nicely wrapped within a seamless backend-to-frontend framework powered by Gradio and FastAPI🎨

And you can even spin it up easily and locally, using Docker🐋

So, what are you waiting for? Go turn your hundreds of pages of chaotic learning material into neat and elegant notes ➡️ https://github.com/AstraBert/pdf2notes

And, if you would like an online demo, feel free to drop a comment - we'll see what we can build🚀

lbourdois

posted an update about 2 months ago

Post

2635

We introduce FAT5 (Flash Attention T5) ⚡

An implementation of T5 in PyTorch with UL2 objective optimized for GPGPU for both training and inference thanks to 13 different optimizations.
The main one is that we have designed a CUDA kernel to expand the Flash Attention by @tridao with RPE biases and supports other PE such as RoPE, ALiBi or FIRE.
The result kernel is 2 times faster than a SPDA implementation.
We also use Triton kernels to optimize certain parts of the architecture, such as the cross-entropy and RMSNorm layer.

The various kernels have been carefully built to be compatible with BF16 and torch.compile to go even faster and achieve efficient pretraining.

All other optimizations are described in a 📝 subsequent blog post available on @huggingface 🤗: CATIE-AQ/FAT5-report.

This methodology enabled us to efficiently pretrain as a proof of concept a FAT5 with 147M parameters in French in a reasonable time (1,461H for 419B tokens), with limited resources (1 A100 i.e. a computational budget of ~ €1,900) and a low carbon footprint (13.5kg eq CO2).

The model's weights are also available on Hugging Face: CATIE-AQ/FAT5-small.
Not very useful in practice, it's a PoC and not an instructed model (it's planned for later).

All the code is available on GitHub if you want to pretrain your own model in your own language or for a specific domain: https://github.com/catie-aq/flashT5 ⭐

Ending by indicating that was a joint project with @BorisAlbar at hf.co/CATIE-AQ.

as-cle-bert

posted an update about 2 months ago

Post

1685

𝐑𝐀𝐆𝐜𝐨𝐨𝐧🦝 - 𝐀𝐠𝐞𝐧𝐭𝐢𝐜 𝐑𝐀𝐆 𝐭𝐨 𝐡𝐞𝐥𝐩 𝐲𝐨𝐮 𝐛𝐮𝐢𝐥𝐝 𝐲𝐨𝐮𝐫 𝐬𝐭𝐚𝐫𝐭𝐮𝐩

GitHub 👉 https://github.com/AstraBert/ragcoon

Are you building a startup and you're stuck in the process, trying to navigate hundreds of resources, suggestions and LinkedIn posts?😶‍🌫️
Well, fear no more, because 𝗥𝗔𝗚𝗰𝗼𝗼𝗻🦝 is here to do some of the job for you:

📃 It's built on free resources written by successful founders
⚙️ It performs complex retrieval operations, exploiting "vanilla" hybrid search, query expansion with an 𝗵𝘆𝗽𝗼𝘁𝗵𝗲𝘁𝗶𝗰𝗮𝗹 𝗱𝗼𝗰𝘂𝗺𝗲𝗻𝘁 approach and 𝗺𝘂𝗹𝘁𝗶-𝘀𝘁𝗲𝗽 𝗾𝘂𝗲𝗿𝘆 𝗱𝗲𝗰𝗼𝗺𝗽𝗼𝘀𝗶𝘁𝗶𝗼𝗻
📊 It evaluates the 𝗿𝗲𝗹𝗶𝗮𝗯𝗶𝗹𝗶𝘁𝘆 of the retrieved context, and the 𝗿𝗲𝗹𝗲𝘃𝗮𝗻𝗰𝘆 and 𝗳𝗮𝗶𝘁𝗵𝗳𝘂𝗹𝗻𝗲𝘀𝘀 of its own responses, in an auto-correction effort

RAGcoon🦝 is 𝗼𝗽𝗲𝗻-𝘀𝗼𝘂𝗿𝗰𝗲 and relies on easy-to-use components:

🔹LlamaIndex is at the core of the agent architecture, provisions the integrations with language models and vector database services, and performs evaluations
🔹 Qdrant is your go-to, versatile and scalable companion for vector database services
🔹Groq provides lightning-fast LLM inference to support the agent, giving it the full power of 𝗤𝘄𝗤-𝟯𝟮𝗕 by Qwen
🔹Hugging Face provides the embedding models used for dense and sparse retrieval
🔹FastAPI wraps the whole backend into an API interface
🔹𝗠𝗲𝘀𝗼𝗽 by Google is used to serve the application frontend

RAGcoon🦝 can be spinned up locally - it's 𝗗𝗼𝗰𝗸𝗲𝗿-𝗿𝗲𝗮𝗱𝘆🐋, and you can find the whole code to reproduce it on GitHub 👉 https://github.com/AstraBert/ragcoon

But there might be room for an online version of RAGcoon🦝: let me know if you would use it - we can connect and build it together!🚀

wanghaofan

authored a paper about 2 months ago

Score-CAM: Score-Weighted Visual Explanations for Convolutional Neural Networks

Paper • 1910.01279 • Published Oct 3, 2019

Social Post Explorers

AI & ML interests

Recent Activity

social-post-explorers's activity

Reinforcement Learning from Human Feedback

InstantCharacter: Personalize Any Characters with a Scalable Diffusion Transformer Framework

OLMoTrace: Tracing Language Model Outputs Back to Trillions of Training Tokens

CrossWordBench: Evaluating the Reasoning Capabilities of LLMs and LVLMs with Controllable Puzzle Generation

SmolVLM: Redefining small and efficient multimodal models

SmolVLM: Redefining small and efficient multimodal models

Score-CAM: Score-Weighted Visual Explanations for Convolutional Neural Networks

AI & ML interests

Recent Activity

Team members 861

social-post-explorers's activity