Deutsch
English Indonesia Türkçe Deutsch Español Português Pусский Français Italiano Nederlands Polski العربية ไทย Tiếng Việt 繁體中文 简体中文 日本語 한국어
javaeeeee1's profile picture. Software developer with physics background, teacher, entrepreneur

Dmitry Noranovich

@javaeeeee1

Software developer with physics background, teacher, entrepreneur

17Tsd.Posts 2Tsd.Follower 2Tsd.Folge ich
Was dir gefallen könnte
Dmitry Noranovich hat repostet
LightningAI's profile picture.
Lightning AI ⚡️
@LightningAI
03.11.

Inference engines turn trained models into a live, callable API. FastAPI works for simple models... until you need streaming, batching, or speculative decoding. And when your “model” is really multiple models stitched together, every API hop adds latency. That's why we built…

LightningAI's tweet image. Inference engines turn trained models into a live, callable API. FastAPI works for simple models... until you need streaming, batching, or speculative decoding. And when your “model” is really multiple models stitched together, every API hop adds latency. That's why we built…
1
1
9
6
1Tsd.
Dmitry Noranovich hat repostet
CShorten30's profile picture.
Connor Shorten
@CShorten30
07.10.

REFRAG from Meta Superintelligence Labs is a SUPER EXCITING breakthrough that may spark the second summer of Vector Databases! ☀️🏖️ REFRAG illustrates how Database Systems are becoming even more integral to LLM inference 🧬 By making clever use of how context vectors are…

CShorten30's tweet image. REFRAG from Meta Superintelligence Labs is a SUPER EXCITING breakthrough that may spark the second summer of Vector Databases! ☀️🏖️ REFRAG illustrates how Database Systems are becoming even more integral to LLM inference 🧬 By making clever use of how context vectors are…
10
39
185
154
28Tsd.
javaeeeee1's profile picture.
Dmitry Noranovich
@javaeeeee1
03.11.

Want to learn how to use precomputed, compressed vector representations of the data from a vector database instead of using the full text of retrieved documents? Listen to a podcast about REFRAG. REFRAG significantly reduces the amount of information the large language model…

CShorten30's profile picture. AI and Databases @weaviate_io
Connor Shorten
@CShorten30
03.11.

I am SUPER EXCITED to publish the 130th episode of the Weaviate Podcast featuring Xiaoqiang Lin (@xiaoqiang_98), the lead author of REFRAG from Meta Superintelligence Labs! 🎙️🎉 Traditional RAG systems use vectors to retrieve relevant context, but then throw away the vectors,…

CShorten30's tweet image. I am SUPER EXCITED to publish the 130th episode of the Weaviate Podcast featuring Xiaoqiang Lin (@xiaoqiang_98), the lead author of REFRAG from Meta Superintelligence Labs! 🎙️🎉 Traditional RAG systems use vectors to retrieve relevant context, but then throw away the vectors,…
5
14
33
11
8Tsd.
2
2
7
0
698
Dmitry Noranovich hat repostet
UnslothAI's profile picture.
Unsloth AI
@UnslothAI
31.10.

You can now run Qwen3-VL locally! 💜 Run the 235B variant for SOTA vision/OCR on 128GB unified memory (dynamic 4-bit). Includes our chat template fixes. Qwen3-VL-2B runs at ~40 t/s on 4GB RAM. Fine-tune & RL via Unsloth free notebooks & export to GGUF. docs.unsloth.ai/models/qwen3-vl

UnslothAI's tweet image. You can now run Qwen3-VL locally! 💜 Run the 235B variant for SOTA vision/OCR on 128GB unified memory (dynamic 4-bit). Includes our chat template fixes. Qwen3-VL-2B runs at ~40 t/s on 4GB RAM. Fine-tune & RL via Unsloth free notebooks & export to GGUF. docs.unsloth.ai/models/qwen3-vl
26
102
592
293
82Tsd.
Dmitry Noranovich hat repostet
Alibaba_Qwen's profile picture.
Qwen
@Alibaba_Qwen
02.11.

You can now run Qwen3-VL locally with Unsloth AI. 👇Fine-tune & RL via free notebooks.

UnslothAI's profile picture. Open source LLM fine-tuning & RL! 🦥 https://t.co/2kXqhhvLsb
Unsloth AI
@UnslothAI
31.10.

You can now run Qwen3-VL locally! 💜 Run the 235B variant for SOTA vision/OCR on 128GB unified memory (dynamic 4-bit). Includes our chat template fixes. Qwen3-VL-2B runs at ~40 t/s on 4GB RAM. Fine-tune & RL via Unsloth free notebooks & export to GGUF. docs.unsloth.ai/models/qwen3-vl

UnslothAI's tweet image. You can now run Qwen3-VL locally! 💜 Run the 235B variant for SOTA vision/OCR on 128GB unified memory (dynamic 4-bit). Includes our chat template fixes. Qwen3-VL-2B runs at ~40 t/s on 4GB RAM. Fine-tune & RL via Unsloth free notebooks & export to GGUF. docs.unsloth.ai/models/qwen3-vl
26
102
592
293
82Tsd.
21
71
580
228
57Tsd.
Dmitry Noranovich hat repostet
SergioPaniego's profile picture.
Sergio Paniego
@SergioPaniego
03.11.

Gave a smol 🤏 intro to Agents using smolagents last Monday! Sharing the slides in case you're curious. They serve as a gentle first step into the Agents Course we developed at @huggingface 🫶🫶

SergioPaniego's tweet image. Gave a smol 🤏 intro to Agents using smolagents last Monday! Sharing the slides in case you're curious. They serve as a gentle first step into the Agents Course we developed at @huggingface 🫶🫶
5
12
105
75
5Tsd.
Dmitry Noranovich hat repostet
rasbt's profile picture.
Sebastian Raschka
@rasbt
02.11.

Ah sure, that's magazine.sebastianraschka.com/p/the-big-llm-… I am also working on one where I am explaining DeltaNet in more detail.

rasbt's tweet card. From DeepSeek-V3 to Kimi K2: A Look At Modern LLM Architecture Design
The Big LLM Architecture Comparison
Quelle: magazine.sebastianraschka.com
0
9
42
21
3Tsd.
javaeeeee1's profile picture.
Dmitry Noranovich
@javaeeeee1
03.11.

NVIDIA GeForce RTX 5090 vs 4080 for AI (2025): VRAM, Bandwidth, Tensor Cores bestgpusforai.com/gpu-comparison…

javaeeeee1's tweet image. NVIDIA GeForce RTX 5090 vs 4080 for AI (2025): VRAM, Bandwidth, Tensor Cores bestgpusforai.com/gpu-comparison…
0
0
0
0
21
javaeeeee1's profile picture.
Dmitry Noranovich
@javaeeeee1
02.11.

NVIDIA GeForce RTX 5090 vs 4080 for AI (2025): VRAM, Bandwidth, Tensor Cores bestgpusforai.com/gpu-comparison…

javaeeeee1's tweet image. NVIDIA GeForce RTX 5090 vs 4080 for AI (2025): VRAM, Bandwidth, Tensor Cores bestgpusforai.com/gpu-comparison…
0
0
0
0
36

United States Trends

Was dir gefallen könnte

Something went wrong.


Something went wrong.