Скачать или смотреть Qwen 3 VL + LM Studio: Building a Market Profile AI with RAG & OCR Fixes

Chapter Time stamps
00:25 Broader plan
02:25 Lesson 1 : Qwen VL Broken Jinja Template
03:36 Lesson 2: Qwen3-VL-8B has issues with 12GB VRAM
03:52 Qwen3-VL-4B-Q4K_M half worked
04:46 Unsloth settings for Qwen3-VL
05:59 How parameters affect VLM/LLM outcome
07:54 How I tackled OCR issue in Qwen3-VL
10:03 Vision Language models are not an OCR tool: There is a fundamental difference
12:51 Intro to Market Profile Python Script,
14:56 Intro to helper text file for RAG
17:52 RAG explained in 1 minute
20:34 Embedding Model Leaderboard
20:54 Intro to AnythingLLM
21:19 LM Studio settings to run VLM as a server
21:57 Qwen3-VL30-a3b is a GOAT
23:13 AnythingLLM setup instructions
25:07 System Prompt for RAG
27:11 Market Profile Test using Qwen3-VL
29:13 mProof of how RAG makes LLM/VLM smarter
31:13 OCR fallback mechanism is working
34:57 Token/second speed with our setup

Link for the repo:
https://github.com/beinghorizontal/Qu...