Treat Visual Tokens as Text? But Your MLLM Only Needs Fewer Efforts to See https://
arxiv.org/abs/2410.06169 https://
github.com/ZhangAIPI/YOPO
_MLLM_Pruning/tree/main?tab=readme-ov-file
…
OPEN SOURCE
-
YOPO Pruning: Efficient Visual Token Processing for MLLMs
By
–
-
Celebrating Developer Creativity with MCP Server Contributions
By
–
Love the amount of creativity I'm seeing from devs using MCP!
— Alex Albert (@alexalbert__) 27 novembre 2024
If you make a cool MCP server, definitely create a PR to add it to our servers repo: https://t.co/nNZJmNlIh8 https://t.co/KcGq5kSguOLove the amount of creativity I'm seeing from devs using MCP! If you make a cool MCP server, definitely create a PR to add it to our servers repo: https://
github.com/modelcontextpr
otocol/servers
… -
Open Source Model Challenges OpenAI o1 Moat
By
–
That’s an Apache 2.0 licensed model competing with OpenAI o1 preview – the moat never existed!
-
QwQ-32B Model Now Available on Hugging Face Hub
By
–
Model on the hub, try it out: https://
huggingface.co/Qwen/QwQ-32B-P
review
… -
MCP Integration in Production: Multiple Clients Beyond Claude Desktop
By
–
Yep we do have a few clients beside Claude desktop that have already integrated MCP in production (Zed, Sourcegraph, etc). We have more details on how to build a client here: https://
modelcontextprotocol.io/clients That page is a little sparse right now in terms of quickly getting started -
AI4Bharat Releases Largest Indian Language Speech Translation Dataset
By
–
AI4Bharat introduces the largest speech translation dataset for Indian languages, featuring 44,400 hours of audio across 13 languages. Learn more: https://
indiaai.gov.in/article/ai4bha
rat-unveils-bhasaanuvaad-speech-translation-dataset-in-13-languages
… #MultilingualAI #BharatAI #IndiaAI #AI4Bharat #BhashaAnuvaad -
Llama 3.1 70B on TT-QuietBox with vLLM Framework
By
–
A Pseudolab developer event in Korea showcasing Llama 3.1 70B running on TT-QuietBox with the vLLM framework. Buy TT-QuietBox today –> http://
tenstorrent.com/hardware/tt-qu
ietbox
… -
Original MNIST dataset updated on Hugging Face
By
–
MNIST original dataset updated by the himself on Hugging Face!
-
Model weights and inference code now available
By
–
check out the model weights and inference code here:
-
NVIDIA releases Hymba-1.5B open-source language model weights
By
–
yo! @NVIDIAAIDev finally released the weights for Hymba-1.5B – outperforms Llama, Qwen, and SmolLM2 with 6-12x less training trained ONLY on 1.5T tokens > massive reductions in KV cache size and improved throughput
> combines Mamba and Attention in a hybrid parallel