Why Chrome Automatically Redirects HTTP to HTTPS — and How to Bypass It Core Principles Understanding view(), reshape(), and transpose() in PyTorch Where Should optimizer.zero_grad() Go in Your PyTorch Training Loop?Agent AI, Machine Learning, and NLP: Origins, Relationships, and Taxonomy Proxy backend/open_webui/utils/middleware.py Deep Dive: RAGFlow’s Admin & Superuser Initialization System How RAGflow’s APIToken.query Works Under the Hood Understanding RAGflow’s api_token Table Deep Dive: RAGFlow’s Document Parsing Pipeline How RAGflow Uses operator.attrgetter for Lazy ORM Query Building Building Q&A with ragflow_sdk: Chat, Session, and ask()From Elasticsearch Queries to Kubernetes Cluster Deployment: A RAGFlow Journey How RAGFlow Assigns create_time in BaseModel How RAGFlow Creates a Dataset: A Code Walkthrough Where Does a Dataset Actually Live in RAGFlow?Upload Is Not Enough: Understanding RAGFlow’s Two-Step Document Processing Pipeline How RAGFlow Uses Elasticsearch to Filter by Dataset How RAGFlow Generates Knowledge Graphs and RAPTOR Hierarchies Important tables How RAGFlow Resolves index_name and knowledgebase_id During Chat Why Infinity Is a Strong Document Engine for RAGFlow How RAGflow Bootstraps a Knowledge Base: Inside init_kb and create_idx How RAGFlow Handles User Invitations and Multi-Tenant Access Control kb/create and /datasets How RAGFlow Creates and Persists a Knowledge Base: A Service-Layer Walkthrough How RAGFlow Handles Passwords: Transport Encryption + One-Way Hashing Deep Dive into RAGFlow’s PDF Parser Understanding the positions Parameter in RAGFlow Cause Studied code Why User ID Equals Tenant ID in RAGFlow: A Practical Multi-Tenant Design How RAGFlow Searches Across Multiple Datasets RAGFlow Search Flow: A Deep Dive into the Retrieval Pipeline Deep Dive into RAGFlow’s Search Architecture: Document Storage, Hybrid Retrieval, and Similarity Computation How RAGFlow Stores Files in MinIO Deep Dive into RAGFlow’s Task Executor: How Asynchronous Document Processing Works Inside RAGFlow User Management: A Practical Look at the user Table Understanding the Blackjack Value Function — Notes on Sutton & Barto Figure 5.1 Books Does DP Always Need Probabilities? A Look at DP vs. Monte Carlo in Reinforcement Learning Is the Convergence of Monte Carlo ES Still an Open Problem?Data Process Understanding the DDPM Forward Diffusion Process Embedding模型的内部机制详解长文本信息稀释原因及可能方向 Understanding Conv2d: How Multi-Channel Convolution Really Works Embedding Index in Vector Databases: Storage, Search, and Similarity Understanding Information Retrieval Evaluation Metrics: NDCG, MAP, and MRR Understanding Tensor Shape Changes: Conv2d, BatchNorm2d, and Flatten in PyTorch GAN vs Diffusion: A Deep Dive into Modern Generative AI ResNet: The Architecture That Won ImageNet Fixing PyTorch dtype Mismatch: Long vs Float in Matrix Multiplication PyTorch Gradient Pitfalls: Lessons from Building a Word2Vec Training Loop Understanding PyTorch transforms.Normalize and ToTensor: Common Pitfalls Understanding NDCG Evaluation with ranx torchtune vs HuggingFace Transformers: A Training Comparison The End of T.Scale: What Happened to torchvision.transforms.Scale and What to Use Instead From PyTorch Tensor to PIL Image: A Line-by-Line Breakdown Datasets Optimum Transformers Calling Claude on AWS Bedrock with OpenAI SDK LiteLLM: More Than Just a Router Tips MCP Authorization Quickstart (Linux)Model Context Protocol (MCP)MCP Memory Mixture of Experts (MoE): Scalable Machine Learning Architecture Monte Carlo Methods in Reinforcement Learning: A Deep Dive From MNIST to Full-Document OCR: Bridging the Gap Between CNN Classification and Real-World Text Recognition LiteLLM, Ollama, and vLLM: Understanding Model Switching at Every Layer Peoples Players Products Prompt RAG RL Watching a Character-Level RNN Learn: Training Observations and the Loss Plateau Problem From Fast Loss Drop to Weird Tokens: Reading Character-Level LTSM Training Signals Why Cannot backpropagate more than once Happens in RNN/LSTM Training Why LSTM Still Needs BPTT — and Why It Does Not Reason Like a Transformer Why Is My RNN Generating “wan wan wan…”? A One-Line Bug That Breaks Backpropagation Why Perplexity Rises at the Start of RNN Training Skill Usages Table Understanding with RAG Template Why np.array Breaks Autograd Backpropagation: A Subtle Python Reference Bug Machine Prices for AI Training: A Complete GPU Server Buyer’s & Rental Guide Installing PyTorch with the Correct CUDA Version PyTorch LayerNorm Warning: “Cannot Dispatch to Fused Implementation”Fine-Tuning LLMs with Torchtune: A Practical Guide (Qwen & Llama2)Running Inference on a Torchtune LoRA Fine-Tuned Model The Modern Landscape of LLM Training & Fine-tuning: Frameworks and Methods in 2026 Pytorch Utilities World Models, Language Models, and the Architecture Beneath Them Explicit vs. Implicit Features: What Every Software Architect Must Consider Fact-Finding, Fix Issue, and Develop Feature: Three Types of Engineering Tasks “扛责”Why sh -c "cd dir && pwd" Prints the Wrong Directory Exposing EMIB Inter-Die Signals Using FIB Pads: A Deep Dive into Advanced Packaging Failure Analysis DDR5 Memory Addressing: From Physical Address to Bits on the Wire From System Physical Address to DRAM Chip: What You Can (and Cannot) Locate Decoding Intel’s Xeon “Rapids” Family: What’s in a Name?Chromadb Python Package MinIO Mirror and Backup: A Practical Guide to mc mirror --watch MinIO: Why Object Storage Beats a Plain Directory Structure PostgreSQL vs SQLite ChromaDB Python Package File Management in Linux: file and ls Github Pages GitHub Copilot Customization and Resources Understanding nvidia-smi Output Company and Patent RMB Mainstream Linux Releases: Debian, Ubuntu, CentOS, Fedora, and SUSE Memory Management Linux Process Management: A Practical Guide to ps and Process Inspection Formula Tensor Azure OpenAI Documentation Azure Rich Progress: Beautiful Progress Bars in Python Pip Package Index Configuration Guide Create Package PyPI Server Deployment Guide Install Package Uploading the distribution archives Wireless Device Regulations: What Every Hardware Engineer Needs to Know Converting HTML to PDF in Python: A Complete Guide Scrapy: Why custom_settings on Your Spider Gets Ignored Scrapy: Spider.parse → ItemPipeline → New Request Flow Handling Pages That Require Login with Playwright Handling Large Tables When Converting Web Pages to PDF with Playwright Scrapy Log Saving Saving Scraped Data to a Database in Scrapy Post-Quantum Cryptography: Understanding XMSS, LMS, SPHINCS+, and Dilithium Secure Secret Input: How to Read Sensitive Data Without Exposing It useradd VS adduser Make Certificate SSH (Secure Shell)Docker Reference Guide Docker Compose Multiple Files Configuration Docker Environment File Q&A Guide Environment Variable Configuration gcr.io/distroless/base-debian12 Docker Restart Policies in Docker Compose Docker Compose Volumes Guide FastAPI 本地 Uvicorn + FastAPI 获取 IP NGINX Learn Artificial Intelligence Open-WebUI RAG Flow Building Better Neural Networks: ReLU and Dropout in a Three-Layer Architecture Langchain Inference Train Practices Algorithm 架构师 Bash Compiler CPU Architecture Notes Database NEO4j Disk Documentation Mermaid Editor VIM Visual Code Edk2 Flutter Github GPU Invest Linux commands and kernels Mathematics Microsoft Network numpy Python Command Line Interface Python Package Index UV - Python Package and Project Manager Research Robot Natural Science Scrapy Security Share Universities Video Virtualization & Containers: A Complete Technical Guide Container Visualization Weights & Biases VNC Web X Window

Practices

Practices

torchtune finetune for qwen model
torchtune generate for finetuned model
torch with correct version
torch without fused implementation
training frameworks

Learning

Learning

Record learning from practice