Packages
PyGrinder introduces missing values into datasets using multiple patterns (MCAR, MAR, MNAR, RDO, sequential, block) to simulate incomplete data for model evaluation and testing.
Install it if you work with time-series data or incomplete observations in model development or research.
Provides type definitions and data structures for the Guardrails AI REST client, enabling type-safe interaction with the guardrails-api service for LLM output validation and guarded chat completions.
Provides CUDA-accelerated PyTorch kernels and operators for equivariant neural network primitives, including specialized Pairformer inference operations optimized for low-precision tensor computations.
BenchPOTS provides standardized preprocessing pipelines and evaluation tasks for benchmarking machine learning algorithms on partially-observed time series datasets.
Supports participation in the ECMWF AI Weather Quest competition by enabling forecast submission, evaluation of sub-seasonal forecasts, and training data download for AI-based weather prediction models.
Augmax is a JAX-based image data augmentation framework that chains transformations together and compiles them efficiently for GPU/TPU execution, minimizing redundant memory operations.
However, maintenance is aging (last update 455 days ago), so verify that the transformation set and API meet your needs before committing to it in production.
Implements a CUDA-optimized causal depthwise 1D convolution operation with PyTorch bindings, supporting fp32, fp16, and bf16 precision for kernel sizes 2, 3, and 4.
The high install friction (CUDA, build tools, ROCm patching for some users) is a real cost; install only if you have a GPU environment already set up and the…
TensorFlow Lite runtime enables on-device machine learning inference on mobile and embedded devices with low latency and minimal binary size.
However, the dormant maintenance status means no recent updates or security patches; verify compatibility with your target model format and hardware before committing…
DiffQ performs differentiable quantization of PyTorch models using pseudo quantization noise, automatically tuning the number of bits per weight to balance model size and accuracy.
AntiBERTy is a transformer language model pre-trained on antibody sequences that generates embeddings, classifies antibody species and chain types, predicts masked residues, and scores sequence likelihood.
Consider it a research artifact: useful for one-off analysis or as a feature extractor in a larger pipeline, but not for production systems requiring updates or legal…
k-diffusion is a PyTorch library implementing diffusion-based generative models from Karras et al. (2022), with improved sampling algorithms, transformer-based model support, and compatibility wrappers for other diffusion frameworks.
fev is a lightweight benchmarking library for time series forecasting models that provides standardized evaluation workflows, reproducible task definitions, and metric computation without heavy dependencies.
Implements parameter-efficient fine-tuning algorithms (LoRA, LoHa, LoKr, DyLoRA, and others) for Stable Diffusion models, enabling training and inference with reduced memory and storage overhead.
Official Python SDK for Z.ai's large model APIs, providing access to chat completions, embeddings, video generation, audio processing, and assistant features.
However, verify the license terms before use in proprietary projects, since the license status is currently unclear in the package metadata.
Provides CLI and MCP server access to Google Gemini Notebook (formerly NotebookLM), enabling programmatic notebook creation, source management, content generation, and AI agent integration.
A self-contained voice activity detector that processes 10ms chunks of 16-bit mono PCM audio at 16kHz and returns a speech probability score to classify audio as speech or silence.
Loads CUDA kernels for equivariant neural network operations on NVIDIA GPUs, providing compiled binary extensions without Python bindings.
Converts between bounding box formats (COCO, YOLO, VOC, FiftyOne, Albumentations) and computes geometric operations like IoU, area, and intersection.
TNT provides training utilities and tools for PyTorch models, including abstractions for common training workflows, distributed training support, and monitoring integrations.
However, the Pre-Alpha status and 814-day release gap mean the API may shift and updates are infrequent—verify that its abstractions match your workflow before…
SDV generates synthetic tabular data by learning patterns from real datasets using machine learning models, then creates new data that preserves statistical properties while anonymizing sensitive columns.
However, the BUSL-1.1 license treatment is unclear; verify the license terms match your use case (commercial, internal, or research) before committing to production…
Provides Python tools for building, inspecting, and working with LiteRT-LM file formats to run large language models on edge devices with hardware acceleration.
However, verify that you need the builder/inspection tools specifically—if you only need to run inference, you may need the full LiteRT-LM runtime separately.
Powershap performs feature selection by testing whether each feature's Shapley values are statistically significantly larger than a random feature's, using hypothesis testing and automatic power calculations to determine the required iterations.
However, verify that its power-calculation defaults and statistical assumptions fit your problem domain before relying on it for critical feature selection decisions.
AI4TS provides a framework for building machine learning models that work with time-series data, including incomplete, irregularly sampled, and partially observed sequences through imputation, interpolation, classification, clustering, and forecasting.
Install it if you work with incomplete or irregularly sampled time-series data and need a framework to handle imputation, classification, clustering, or forecasting.
Fits the smallest-area quadrilateral that contains all points from a noisy polygon or point cloud, useful for cleaning up irregular detections before perspective correction or pattern matching.
Install it if your workflow involves cleaning up shape detections; skip it if you need active support or ongoing feature development.
PyTextRank implements graph-based TextRank and related algorithms as a spaCy pipeline extension to extract key phrases and perform extractive summarization on text documents.
A terminal-based AI coding agent that works with any LLM supporting tool calling, providing an interactive interface for code generation, execution, and debugging without requiring code setup.
Performs fast hierarchical agglomerative clustering from distance matrices or vector data, replacing scipy.cluster.hierarchy functions with optimized C++ implementations.
Integrates Baseten's chat models and embeddings into LangChain applications, enabling use of Baseten-hosted models through LangChain's standard interfaces.
openvino-genai simplifies running inference on generative AI models by wrapping OpenVINO's inference engine with a high-level pipeline API that handles tokenization, model loading, and text generation.
The main gotcha is strict version pinning of openvino_tokenizers; verify compatibility before updating dependencies.
OGB provides standardized benchmark datasets, data loaders, and evaluators for graph machine learning tasks across node, link, and graph prediction problems.
However, maintenance is aging (last release 2023-04-07)—verify compatibility with your PyTorch and graph framework versions before relying on it for new research.
Cellpose segments cells and nuclei in microscopy images using deep learning, with support for 2D/3D data, custom model training, and image restoration.
Provides GPU-accelerated approximate nearest neighbor search and clustering algorithms optimized for NVIDIA GPUs.
Compresses and optimizes neural networks through quantization, pruning, and hardware-aware techniques, reducing model size and computational cost for edge deployment.
Install it if you need to compress trained models for deployment; the main constraint is the Python >=3.10 requirement and the need to provide a pre-trained model as…
GPU-accelerated vector search and clustering library providing approximate nearest neighbor algorithms for similarity search on NVIDIA hardware.
Tianshou is a PyTorch-based reinforcement learning library that implements a wide range of RL algorithms (DQN, PPO, SAC, and others) with both high-level and low-level APIs for training agents on Gymnasium environments.
Label Studio is a web-based data annotation platform that lets you label images, audio, text, video, and time-series data through a browser UI, then export annotations in formats ready for machine learning training.
tinygrad is a minimal deep learning framework providing a tensor library with autograd, an IR-based compiler, JIT execution, and neural network training utilities—positioned between PyTorch's ergonomics and lower-level compiler systems.
LFX is a lightweight command-line tool that executes Langflow AI flows from JSON files without requiring the full Langflow package, serving them as stateless FastAPI endpoints or running them locally.
However, verify the license terms first, and confirm that your flows do not depend on persistent state (saved flows, message history, user management) before…
A framework for building, prototyping, and deploying AI and ML workflows from notebooks to production, handling data management, compute scaling, and orchestration across local and cloud environments.
Measures Korean speech-to-text accuracy using character error rate (CER), word error rate (WER), character correct rate (CRR), and entity preservation metrics with reproducible evaluation reports.