Joshua Batson (@thebasepoint)

작성자가 맥 노트북에서 빠르게 동작하는 오픈 가중치 모델용 해석(interpretability) 라이브러리를 만들었다고 알립니다. SSH나 Colab이 필요 없고, git clone 후 gemma-2에서 logit lens를 실행하는 데 90초가 걸린다고 하며 개발 루프에서 Claude 코드를 활용했다고 소개합니다.

https://x.com/thebasepoint/status/2006440158112133362

#interpretability #logitlens #gemma2 #claude #mac

Joshua Batson (@thebasepoint) on X

Because I somehow can't stop, I made a library for doing interpretability on open weights models that works -- fast -- on your mac laptop. No ssh, no colab. 90 seconds from git clone to logit lens on gemma-2. It's a really fun devloop with claude code.

X (formerly Twitter)

T5Gemma 2 từ Google: mô hình AI đa ngôn ngữ (140+ ngôn ngữ) và đa phương tiện, xử lý văn bản-hình ảnh, hỗ trợ 128K token. 3 cỡ mô hình 270M-4B. #AI #Gemma2 #DeepLearning #Google #Tech #ML #HệThốngAI

https://huggingface.co/collections/google/t5gemma-2

T5Gemma 2 - a google Collection

We’re on a journey to advance and democratize artificial intelligence through open source and open science.

Gemma-2-MoE: Công cụ nhẹ giúp biến các checkpoint Gemma 2 thành mô hình Mixture of Experts kiểu Frankenstein. Dễ dàng thử nghiệm các biến thể MoE của Gemma 2. #Gemma2 #MoE #AI #MachineLearning #TríTuệNhânTạo #HọcMáy

https://www.reddit.com/r/LocalLLaMA/comments/1ped5x6/gemma2moe_frankenstein_moe_builder_for_gemma_2/

"Cập nhật: Người dùng dùng LM Studio 0.3.31 thử kích hoạt suy tư Gemma 2.2b nhưng gặp phản hồi trống. Có thể do kết hợp với Qwen3-4b. Dùng thử để Vitória. #Gemma2 #LMStudio #AI #Hỗ trợ #Tiến Ẩn"

https://www.reddit.com/r/LocalLLaMA/comments/1ofom66/enable_gemma_2_2b_thinking_in_lm_studio/

Wow! I didn't really like Gemma 2, but Gemma 3, released today, is awesome. It comes in four sizes, 1b, 4b, 12b and 27b. It's super fast and except for the 1b version it can even handle images.

The 27B version apparently outperforms both DeepSeek v3 and LLaMA3-405 on the ChatbotArena benchmark.

It's also the first small model I've tested that's good at German.

#gemma3 #gemma #gemma2 #google #ai #programming #google #model #local #gemini #multimodal #vision #wow #chatbotarena #german

Top 15 Small Language #Models ( #SML ) for 2025 | #DataCamp

https://www.datacamp.com/blog/top-small-language-models

> Learn about the top 15 small language models of 2025, including #Llama 3.1 8B, #Gemma2 , #Qwen 2, #Mistral # Nemo, #Phi-3.5 , and more.

Just asked my little chat bot a question I didn't understand, got a great answer back.

You: What do the colours 'CMYKOR' stand for?
Gemma2: CMYKOR stands for:

* **C**yan
* **M**agenta
* **Y**ellow
* **K**ey (Black)
* **O**range
* **R**ed

This color model is used in some specialized printing processes.
You: Why is K black?
Gemma2: In CMYK printing, "K" stands for Key because black ink is used as the key plate to which the other colors (Cyan, Magenta, Yellow) are aligned.

Black ink is also crucial for achieving deep, rich blacks and shadows that mixing the other three inks cannot fully replicate.
#Ollama #Gemma2 #AI

🎯 #OpenSource Language Model Platform Launch

🔧 Leverages #vLLM technology with custom #GPU scheduler for running various #LLM models
🤖 Supports major models: #Llama3 (405B/70B/8B), #Qwen2 72B, #Mixtral, #Gemma2, #Jamba15, #Phi3

https://glhf.chat/

good luck have fun

Chat with open-source models

For those who want to test some #AI module on their #PC and want to do it without complications, there is a #GUI that makes things simple for you and allows you to use different types of AI.
Try #LMStudio. #qwen25 #mistral #phi #llama #gemma2 #deepseek2 #coding
https://lmstudio.ai/
LM Studio - Local AI on your computer

Run local AI models like gpt-oss, Llama, Gemma, Qwen, and DeepSeek privately on your computer.

LM Studio