Packages
Demucs separates music into individual stems—drums, bass, vocals, and accompaniment—using a hybrid transformer-based neural network trained on waveform and spectrogram domains.
Provides embedding and reranker model integrations for Milvus vector database, supporting OpenAI, Voyage AI, Cohere, SentenceTransformers, and Hugging Face models.
However, the aging maintenance status (501 days since last release) warrants checking that its dependencies remain compatible with your environment.
Counts tokens and estimates USD costs for LLM API calls across OpenAI, Anthropic, and other major providers before or after requests are sent.
However, be aware that maintenance is aging—pricing tables may lag behind provider updates, and support for newly released models is not guaranteed.
Converts TensorFlow, Keras, TensorFlow.js, and TFLite models to ONNX format via command line or Python API, enabling model portability across different inference runtimes.
MLX-Audio provides text-to-speech, speech-to-text, and speech-to-speech inference optimized for Apple Silicon, with support for multiple model architectures and multilingual capabilities.
Multi Model Server is a tool for serving deep learning models exported from MXNet or ONNX via HTTP endpoints for inference requests, using a CLI or Docker images to manage the service.
Install only if you are maintaining legacy code that already depends on it and cannot migrate.
Optimistix provides nonlinear solvers for root finding, minimization, fixed-point problems, and least-squares fitting, built on JAX with support for autodiff, GPU/TPU acceleration, and PyTree-based state.
Model Archiver creates `.mar` archive files from trained neural network models for use with MXNet-Model-Server inference.
PyGlove provides symbolic object-oriented programming for Python, enabling direct manipulation of Python objects and programs through a mutable symbolic model with built-in search and iteration primitives.
Install it if you need to search over program variants, explore hyperparameters systematically, or build flexible ML code that evolves.
Builds and runs AI agents powered by Google's Gemini, handling the agentic loop, tool wiring, and stateful conversation management so you focus on agent behavior rather than infrastructure.
However, it is in Alpha status, depends on 7 runtime libraries, requires Python 3.10+, and mandates installation from PyPI due to compiled binaries.
Faiss provides GPU-accelerated similarity search and clustering for dense vectors, supporting exact and approximate nearest-neighbor queries on datasets from memory to billions of vectors.
Connects Apache Spark DataFrames to Amazon SageMaker FeatureStore for ingesting feature data into online and offline stores, with automatic feature definition loading.
TorchEval provides a collection of PyTorch model metrics with both functional and class-based interfaces, supporting single-process, multi-GPU, and distributed training scenarios.
Integrates PyTorch profiling data with TensorBoard, providing GPU timeline tracing and performance diagnostics for ML workloads through a TensorBoard plugin interface.
However, the last release was in October 2023; verify compatibility with your current PyTorch version before relying on it for new projects.
Command-line interface for managing Bittensor platform operations including wallet creation, subnet registration, delegation, and governance voting.
Provides utilities for working with Hugging Face Spaces, including downloading files, uploading content, and managing repositories on the Hugging Face Hub platform.
Install it if you need to programmatically interact with the Hub.
Provides shared utilities and common infrastructure for AutoGluon's automated machine learning framework, supporting tabular, time series, multimodal, and image data tasks.
RSL-RL is a GPU-accelerated reinforcement learning library for robotics research, providing implementations of algorithms like PPO with multi-GPU training support and a minimal, extensible codebase.
Install it if you are working with Isaac Lab, Legged Gym, or similar robot learning environments; otherwise, verify that its algorithm implementations and API match…
Provides cloud-specific integrations and utilities for running Pathways workloads on Google Cloud, enabling a single JAX client to orchestrate machine learning computations across multiple TPU slices.
However, it is tightly coupled to Google Cloud infrastructure and requires explicit preview access, making it unsuitable for general-purpose use outside that ecosystem.
Integrates Perplexity AI models into LangChain applications, enabling you to use Perplexity's language models as a component in LangChain workflows.
Install it if you're already using LangChain and want to add Perplexity as a model provider—it's the standard way to do so.
Coordinates internal Azure Machine Learning SDK dependencies for dataset operations; not intended for direct use.
Diffrax provides numerical solvers for ordinary, stochastic, and controlled differential equations in JAX, with automatic differentiation and GPU acceleration built in.
Extends TensorFlow with support for file systems and data formats not built into TensorFlow, including HTTP/HTTPS access and automatic decompression for remote datasets.
Install only if you have a specific need for extended I/O capabilities; it is not required for standard TensorFlow workflows.
Provides a Python SDK to build, train, and deploy machine learning models on the Azure Machine Learning service, supporting workflows from any Python environment including Jupyter Notebooks and IDEs.
However, clarify the custom license terms first, and verify that the five runtime dependencies align with your use case—this is a heavyweight, cloud-coupled package,…
Inspect Scout analyzes and visualizes transcripts from AI agent runs, providing a web-based interface to explore execution traces, inspect intermediate states, and debug agent behavior.
cassIO integrates Apache Cassandra with machine learning and LLM workloads, providing a Python library to connect Cassandra databases to AI/genAI applications.
However, be aware of the aging maintenance cadence (680 days since last release) and the possibility of breaking changes in the 0.* series.
Integrates Elasticsearch with LangChain to provide vector storage, retrieval, embeddings, chat history, and LLM caching backed by Elasticsearch.
Install it if you are building a LangChain application that requires persistent vector storage, retrieval, or LLM/embedding caching with Elasticsearch.
AutoGluon TimeSeries automates machine learning for time series forecasting, training and deploying high-accuracy models with minimal code using deep learning and statistical approaches.
Modifies Hugging Face Transformers model.generate() to return a token-by-token generator instead of waiting for full inference completion, enabling real-time streaming output during text generation.
Programmatically create and manage Weights & Biases workspaces and reports, including building dashboards with charts and panels from Python code.
Fills missing values in time-series data while respecting irregular time intervals, using spline-based interpolation methods and providing gap analysis tools.
However, it is early-stage (Alpha, 2 stars, 167 days old), so expect the API or behavior to evolve; use it for production work only if you can tolerate potential…
Reasoning Gym generates procedurally verifiable reasoning datasets and RL training environments across algebra, arithmetic, logic, games, and other domains, with adjustable complexity and algorithmic answer verification.
PyTorch Forecasting provides neural network models and utilities for time series forecasting, including architectures like Temporal Fusion Transformers, N-BEATS, and N-HiTS, with training orchestrated through PyTorch Lightning.
Install it if you need to train neural forecasting models on time series data and want to avoid reimplementing standard architectures or training boilerplate.
Official Python SDK for ZhipuAI's large model APIs, providing type-safe client methods to call chat completions, multimodal models, video generation, and assistant endpoints.
However, the aging maintenance status (354 days since last release, no recent commits tracked) and unclear license are concerns—verify the license for your use case…
Evaluates generative AI application outputs using built-in and custom evaluators that measure quality, safety, and performance through mathematical metrics, NLP scores, and AI-assisted assessment.
A command-line tool for running computer vision inference locally via Docker or against Roboflow's hosted API, supporting object detection, classification, instance segmentation, and foundation models like CLIP and SAM.
However, verify that the GPL-3.0 and AGPL-3.0 licenses on bundled models (YOLOv5, YOLOv8) align with your project's licensing requirements before committing to…
Provides Python bindings to NVIDIA's fatbin compiler library for working with CUDA binary formats and GPU code compilation artifacts.
However, verify the license terms before use in proprietary projects, and confirm that your system meets any undocumented CUDA runtime dependencies.
python-doctr extracts and localizes text from documents and images using a two-stage deep learning pipeline: text detection (finding word locations) followed by text recognition (identifying characters), with support for PDFs, images, and rotated pages.
Install only if you have a genuine OCR task; the dependency footprint is substantial but justified for deep learning work.
BlenderMCP is a Model Context Protocol server that enables Claude AI to directly control and interact with Blender, allowing AI-assisted 3D modeling, scene creation, and object manipulation through natural language prompts.
Load, visualize, and evaluate solutions on the ARC-AGI and ARC-AGI-2 reasoning datasets through Python API and command-line tools, with built-in support for multiple dataset versions.