This repository is a knowledge base and documentation collection rather than a software application or data pipeline. It contains 117+ markdown files with interview questions and answers focused on Large Language Models (LLMs), Transformers, RAG, and Prompt Engineering. The content is curated by a single author and serves as a study resource for AI/ML engineers preparing for technical interviews. There is no executable code, data processing logic, or model training scripts present. The repository structure is flat and consistent, grouping questions into logical batches (e.g., QA_1-3.md, QA_4-6.md). While it lacks engineering artifacts like tests or CI/CD pipelines, the documentation quality is high, with clear formatting, consistent answer structures, and cross-references to related repositories. It is best suited for learners seeking structured knowledge rather than engineers looking for reusable code components. The content is static and does not require environment setup or data ingestion, making it highly reproducible in terms of accessibility. However, as a documentation repo, it does not offer the architectural depth or component reusability typical of software engineering projects. The 'active' maintenance status is derived from the provided metadata timestamp, though the content itself is static text.
This is the largest structured Chinese classical poetry dataset on GitHub, containing ~55,000 Tang poems, ~260,000 Song poems, ~21,000 Song ci (lyric poetry), and additional classical texts (Four Books and Five Classics, Book of Songs, Chu Ci, Yuan opera, etc.). The data is distributed as JSON files organized by dynasty and genre, with each poem containing author, title, and paragraph fields. The repository serves as a foundational dataset for Chinese NLP, digital humanities, and cultural heritage projects. The codebase is minimal — essentially a data repository with a JSON validation test (test_poetry.py), a simple data loader (loader/data_loader.py), and a web scraper for Song ci (宋词/UpdateCi.py). The data itself is the product of web scraping from sources like ctext.org and other online repositories, with manual curation and error correction over time. The repository includes an 'error' directory containing records with data quality issues (corrupted characters, missing content, etc.), which demonstrates transparency about data limitations. The data loader provides a basic abstraction for extracting poem bodies across different datasets, but the overall engineering is straightforward. The value lies entirely in the data breadth and accessibility — no complex transformations, feature engineering, or analytical pipelines are present. The repository is widely used (52K+ stars) as a reference dataset for Chinese poetry generation, analysis, and educational applications.
This repository provides access to Microsoft's global building footprint dataset (1.4B buildings) derived from Bing Maps satellite imagery. The core deliverable is a massive, publicly hosted dataset of building polygons in GeoJSONL format, with periodic updates. The repository itself contains only documentation, a Jupyter notebook demonstrating data download and spatial filtering, and a few utility scripts for converting compressed GeoJSONL files to GIS-friendly formats. There is no transformation pipeline or model training code present; the actual building extraction is done externally. The code is simple, well-documented, and parameterized for reuse, but lacks dependency management and environment definitions. The primary value is the dataset itself, with the scripts serving as examples for accessing and processing the data.
This repository is a curated, static collection of real-world legal case studies documenting the consequences of web scraping, data theft, and automated tool misuse in China. It contains no executable code, data pipelines, or analytical notebooks; instead, it organizes news articles, court documents, and lawyer analyses into categorized 'forbidden zones' (禁区) covering illegal crawler services, personal data trading, copyright infringement, and service disruption. The sole entry point is the README, which lists 16+ cases with links to external sources. While not a software project, it serves as a valuable educational reference for developers and legal practitioners navigating Chinese data compliance law. The repository is fully reproducible as it consists of static markdown files with no dependencies or runtime requirements.
This repository is a collection of case studies and presentations on how various companies operate Apache Kafka in production. It provides insights into the challenges and solutions encountered by big companies when running Kafka in production.
Microsoft Magma is a research codebase for a single large multimodal vision-language-action (VLA) foundation model, as opposed to a general-purpose application. The core is an 8B-parameter model built on LLaMA-3-8B-Instruct with a ConvNeXt vision encoder, capable of understanding images/video and producing both textual responses and discretized robot/UI actions. The repository contains the full training stack: model definitions, data preprocessing, dataset-specific constructors, mixed RLDS/PyTorch data loaders, action tokenization, DeepSpeed training configs, and evaluation hooks for robotics benchmarks like LIBERO and SimplerEnv. It also ships FastAPI servers and demo agents for UI automation and robot trajectory planning, but these are secondary to the model training and data pipeline core, making this a hybrid repo with a dominant data-research character. The data lineage is central to the project. Raw heterogeneous datasets—Open X-Embodiment, Ego4D, EpicKitchen, SeeClick, LLaVA, and the in-house Magma-820K—are normalized into a unified multimodal conversation format via per-dataset Constructor subclasses. Visual point tracks from CoTracker2 are clustered, quantized, and rendered as 'Trace-of-Mark' overlays; bounding boxes and labels are rendered as 'Set-of-Mark' overlays. Continuous robot actions are discretized into token IDs aligned with the LLM vocabulary, allowing a standard causal LM objective to train action prediction alongside visual understanding. The resulting model can be deployed via the included FastAPI server for interactive inference. Reproducing the full pipeline requires access to gated or proprietary datasets, CoTracker2 checkpoints, a custom transformers fork, and significant GPU resources, so end-to-end reproduction is not turnkey, though inference on the released HuggingFace checkpoint is more accessible.
UniTEX is a research-focused repository implementing a two-stage framework for high-fidelity 3D texture generation from arbitrary 3D shapes. Stage 1 uses LoRA-adapted Diffusion Transformers for multi-view texture synthesis, while Stage 2 employs a transformer-based Large Texturing Model (LTM) to predict Texture Functions in unified 3D space, bypassing UV mapping limitations. The repository integrates super-resolution (TSD_SR) for detail enhancement and provides tools for mesh processing, rendering, and texture reprojection. It includes training scripts, model architectures, and utility libraries, indicating its primary purpose is to support experimentation and reproduction of the CVPR 2026 paper's methodology rather than serving as a standalone end-user application.
This repository implements Dinomaly, a reconstruction-based unsupervised anomaly detection method for industrial inspection datasets. It uses vision transformers (primarily DINOv2) as feature extractors with a noisy bottleneck and linear attention mechanism. The code supports both class-separated and unified training across MVTec-AD, VisA, and Real-IAD datasets. Key components include dataset loaders, custom optimizers, training utilities, and evaluation metrics. While the research code achieves state-of-the-art results, it contains hardcoded paths, version-specific dependencies, and dataset-specific logic that limit immediate reproducibility.
This repository, branded as "The Malware Bible," is a curated educational knowledge base for cybersecurity, malware analysis, reverse engineering, and exploit development. It is not a conventional software application or an operational data pipeline; under the available classifications it is closest to data/research because its core assets are reference documentation, course-style Markdown modules, and research-oriented binaries rather than runnable application code. The content is organized into four learning paths: The Beginning (C, Python, x86 assembly, shellcode, reverse engineering basics), The Journey (PE/ELF structures, Android/iOS package formats), The Rituals (obfuscation, evasion, and writing malware), and The Scriptures (tool cheatsheets for radare2, IDA Pro, WinDbg, YARA, WinAPI, and evasion tactics). In practice, the repository serves as a self-study curriculum and analyst reference, not as a deployable system. It includes a large Windows DLL export catalog in Markdown, example ELF binaries demonstrating heap overflow, use-after-free, ROP chains and format string bugs, and numerous screenshots/diagrams. There is no CI/CD, dependency management, or automated testing; several files are marked "Coming soon" or "WORK IN PROGRESS," and external links or sponsor images may rot over time. The Windows API and DLL reference content is broad but may drift as Windows evolves. The last commit is recent, indicating active maintenance, and the repository benefits from a structured table of contents and consistent formatting across chapters.
This repository is a collection of blog posts, presentations, Jupyter notebooks, and scripts by John Lambert (@JohnLaTwC), a renowned security engineer from Microsoft. It serves as a knowledge base for security defenders, covering topics such as attack graph theory, the MS08-067 vulnerability story, and the 'Githubification of InfoSec' concept, which advocates for open, community-driven security practices using MITRE ATT&CK, Sigma rules, and Jupyter notebooks. The notebooks focus on malware analysis, shellcode reverse engineering, and threat hunting, often leveraging VirusTotal APIs and custom disassembly tools. The Yara rules provide detection for document-based attacks. The PowerShell scripts are utilities to restore Windows Defender to a secure configuration. While the content is highly educational and demonstrates deep security expertise, the repository does not constitute a structured application or a reusable data pipeline; it is a curated set of resources for learning and reference.
This repository implements Geometry Forcing, a research framework for video generation with geometric consistency. The core contribution marries video diffusion models (DFoT - Diffusion Forcing Transformer) with 3D representation learning (VGGT) to produce temporally consistent videos conditioned on camera poses. It includes full training pipelines (Hydra + PyTorch Lightning), a suite of video metrics (FVD, FID, LPIPS, VBench), and evaluation scripts for reprojection/revisit error using DROID-SLAM. The codebase is structured as a research template with reusable base classes for algorithms, data modules, experiments, and metric registries. Raw datasets (RealEstate10K, Kinetics-600, Minecraft) are downloaded and preprocessed into video clips with associated camera parameters. Transformations include VAE encoding for latent diffusion, augmentation (flip, reverse), and sliding-window clip extraction. The primary research question is whether video diffusion can be constrained by explicit 3D geometry to reduce inconsistencies in generated videos, measured by geometry-aware metrics. The repo also demonstrates composable sampling via History Guidance and an alignment loss that projects diffusion latents into VGGT feature space. While the code is well-engineered for research (modular design, config-driven, metric registry), it carries typical research code baggage: hardcoded paths, vendored third-party code (DROID-SLAM, VGGT), potential dependency pinning conflicts, and some dead/commented code. Reproducibility is moderate: dependencies are pinned but no environment definitions beyond requirements.txt, and dataset downloads rely on fragile YouTube scraping.
This repository provides ECMAScript compatibility tables for JavaScript engines, browsers, and compilers. It ingests test definitions (embedded as commented JavaScript code) and expected results per engine version, then builds static HTML pages via Node.js scripts. The data covers ES5 through ES2025+ proposals, Intl, and non-standard features, with thousands of subtests mapped to hundreds of engine versions. The codebase is primarily a data pipeline with a build system (using cheerio for HTML generation) and optional test runners that execute tests against real engines (e.g., Duktape, GraalVM) to validate or update results. The output is the well-known kangax.github.io/compat-table static site. While the data is meticulously maintained, the repository is not a reusable library or application; it is a specialized data project with moderate reproducibility due to reliance on external engine binaries and outdated dependencies like jQuery 1.12.
This repository is a lightweight automation pipeline that mirrors MaxMind's GeoLite2 IP geolocation databases (Country, City, ASN) via a GitHub Actions workflow. It does not contain application code or analytical scripts; its sole component is a CI/CD workflow that downloads the latest databases every three days from MaxMind, extracts the .mmdb files, and pushes them to both a dedicated 'download' branch and formal GitHub Releases, retaining only the two most recent releases. The repository serves as a dependable, proxy-free distribution point for these databases, eliminating the need for users to obtain their own MaxMind license key. While the repo has no traditional data transformations or feature engineering, the workflow demonstrates a clean pattern for automated data mirroring that can be adapted to other periodically updated datasets. The overall codebase is extremely minimal—just the workflow YAML, a README, and a LICENSE—making it functional but not rich in reusable data science assets. Reproducibility is limited by the requirement of a MaxMind license key stored as a GitHub secret, and the workflow is tightly coupled to the GitHub Actions environment.
This repository is a curated collection of robot description files (URDF, xacro, STL/DAE meshes) for multiple mobile manipulator platforms (Unitree B1+Z1, Aliengo+Z1, Go2+ARX, PR2, Ridgeback+UR5, Mabi Mobile) combined with an extensive awesome-list survey of loco-manipulation research papers. The robot models are well-structured with parameterized xacro macros, detailed inertial parameters, and simulation-ready Gazebo plugins. The paper list is meticulously organized by methodology (model-based control, learning-based, foundation models) with verified links. While the repository contains no application code or data pipelines, it serves as a valuable reference for researchers and engineers needing ready-to-use robot models for simulation or a comprehensive literature overview. The primary value lies in the reusable xacro patterns and the curated paper index.
This repository is a curated collection of baseline solutions for data science competitions, primarily from Chinese platforms (科大讯飞 AI开发者大赛, Tianchi, DataFountain, AIWIN, etc.) and some international ones (Kaggle). It contains hundreds of Jupyter notebooks and Python scripts covering a wide range of tasks: structured data classification/regression (LightGBM, XGBoost), computer vision (ResNet, EfficientNet, YOLO, CRNN), natural language processing (BERT, TF-IDF, Qwen), time series forecasting, and multimodal challenges. Each baseline is self-contained, providing a quick-start pipeline from data loading to submission generation. The repository serves as a practical learning resource for competition participants, demonstrating common feature engineering techniques, model training loops, and ensemble strategies. However, the code quality is uneven: many notebooks are exploratory with hardcoded paths, outdated library versions, and minimal error handling. The repo lacks centralized dependency management, environment definitions, or modular reusable libraries, making reproducibility low without manual adaptation. Despite these limitations, it offers a valuable catalog of patterns and templates that can be adapted for new competitions.
The repository ruanyf/weekly hosts the weekly tech newsletter '科技爱好者周刊' (Tech Lover's Weekly) curated by 阮一峰. Each issue is a markdown file in the `docs/` directory, following a standardized template that includes sections such as cover image, main topic, tech dynamics (short news), articles (annotated links), tools (software recommendations), resources (reference links), images, excerpts, quotes, past issues, and acknowledgements. The newsletter has been published every Friday since 2018, with issues numbered sequentially. The repository serves as an open-source, community-curated archive of technology news, tools, and resources. It accepts contributions via GitHub issues for article submissions and job postings ('谁在招人'). While the content is rich and well-organized, the repository contains no data processing code or build system; it relies entirely on manual curation and external links. This makes the long-term reproducibility dependent on the availability of external resources (images via CDN.beekka.com and linked articles). Despite this, the newsletter remains a valuable resource for tracking technology trends over time.
This repository is a massive collection of Jupyter notebooks, course materials, and documentation for the Hugging Face ecosystem (Transformers, Diffusers, Datasets, PEFT, smolagents, etc.). It serves as an educational resource, containing the official Hugging Face course in multiple languages, auto-generated documentation notebooks, and community examples covering NLP, computer vision, audio, and generative AI. It also includes SageMaker integration examples for training and deployment on AWS. The repository is primarily a documentation and learning resource, not a production application. It exhibits hybrid nature as it includes both educational content and reusable utility scripts and training patterns. The code quality is typical for tutorials: functional but not production-grade, with some deprecated API usage (e.g., `load_metric`), translation drift across languages, and auto-generated notebooks that may lag behind library updates. However, the core educational content is well-structured and actively maintained by Hugging Face.
This repository is a curated collection of study materials for the Chinese System Architect Designer exam (a software architect certification). It provides links to video tutorials, teaching materials, past papers, and exam resources, along with historical passing scores from study groups. The project is actively maintained, with regular updates to reflect the latest exam schedules, syllabus changes, and resource links. The repository does not contain any source code or data processing pipelines. Instead, it functions as a static data repository, hosting markdown files that record passing scores from various study groups (2019-2021) and an update log detailing the addition of new resources. The core value lies in the organized aggregation of external study materials and the community-driven collection of exam results. While the project lacks reusable data engineering components (such as ETL scripts or data validation routines), it demonstrates consistent maintenance and responsiveness to changes in the exam landscape. The data stored within the repository (primarily the group pass records and update log) is version-controlled and accessible, though users must rely on external links for the majority of study materials, which are kept up to date by the maintainer.
This repository implements a hierarchical controller for quadruped locomotion that combines a high-level reinforcement learning policy with a low-level model predictive controller (MPC). The RL policy dynamically predicts weight parameters for the MPC, which computes optimal ground reaction forces and joint torques. The codebase includes Isaac Gym training environments for the Unitree A1, Go1, and Aliengo robots, along with a custom C++ MPC solver (OSQP-based) and a finite state machine for robust transient handling. While primarily a research project that bridges RL and MPC for legged robots, it also provides reusable robotics components such as state estimators, leg controllers, gait pattern generators, and foot swing trajectory planners. The work is based on the MIT Cheetah software and includes detailed development logs and configuration files for reproducible experiments.
This repository is a collection of domain blocklists designed for use with Pi-hole and similar DNS-based ad-blocking services. It provides multiple specialized text files containing domain names to block, categorized by purpose (e.g., ads and trackers, telemetry, specific applications like TikTok and WhatsApp, malware, ransomware, and spam). The README explains how to integrate these lists into Pi-hole's adblocking configuration. While the repository lacks visible data processing code (showing only the final curated lists), the topical organization and maintenance updates suggest an ongoing curation effort aimed at removing false positives and optimizing the lists for specific use cases.