Packages
CAMEL is a multi-agent framework for building, simulating, and researching systems of communicative agents that can collaborate on tasks, generate synthetic data, and study emergent behaviors at scale.
Stores and retrieves LangChain documents using a hybrid graph approach that combines vector similarity search with edge-based traversal between document chunks.
Owlready2 is a Python module for ontology-oriented programming that loads, manipulates, and reasons over OWL 2.0 ontologies and RDF knowledge graphs, with an embedded optimized quadstore and support for HermiT/Pellet reasoners.
Jax-jumpy provides a unified interface that runs NumPy code as-is when JAX is unavailable, and automatically switches to JAX's accelerated backend when JAX is installed and conditions are met (JIT compilation or JAX inputs).
Evaluates machine translation quality using neural metrics, scoring translations against references or without references, and explaining error types according to MQM typology.
NeuralProphet is a PyTorch-based framework for interpretable time series forecasting that combines neural networks with traditional algorithms, designed for iterative model building and human-in-the-loop refinement.
However, be aware that the last release was over 784 days ago and the project is in beta—if you require active bug fixes or compatibility guarantees with the latest…
LLMLingua compresses prompts by identifying and removing non-essential tokens, reducing input length to LLMs by up to 20x while preserving task performance and lowering API costs.
Connects LangChain applications to Together AI's language models and embeddings via a unified integration layer.
Cleanlab automatically detects and helps fix data quality issues—mislabeled examples, outliers, duplicates, and annotation errors—in datasets for machine learning, working with any model and any data modality.
PyTDC provides unified access to multimodal biomedical datasets, machine learning benchmarks, and state-of-the-art model weights for drug discovery and therapeutic AI research, with standardized training, evaluation, and inference endpoints.
GPT Researcher is an autonomous agent that conducts comprehensive online research on a given query and generates detailed, factual reports with citations by orchestrating planner and execution agents to gather and synthesize information from multiple sources.
torchtitan is a PyTorch-native platform for training large generative AI models at scale, providing distributed training infrastructure with support for multiple parallelism techniques, checkpointing, and quantization.
Accelerates Transformer model training and inference on NVIDIA GPUs using low-precision formats (FP8, MXFP8, NVFP4) with optimized kernels and mixed-precision APIs.
However, verify license terms before commercial use and ensure your system meets the strict hardware and software prerequisites (CUDA 12.1+, cuDNN 9.3+, C++17 compiler).
Extracts learned audio and text embeddings using contrastive language-audio pretraining, enabling cross-modal retrieval and downstream audio-text tasks.
However, verify that the HuggingFace checkpoints you need are still available, and be aware that the last release was 467 days ago—if you encounter bugs or need…
LitServe lets you build custom inference servers in Python by defining how requests are handled, models are loaded, and outputs are produced, with built-in support for batching, streaming, and multi-GPU scaling.
Loads and runs computer vision models from Roboflow or local sources using PyTorch, ONNX, TensorRT, or Hugging Face backends, with a unified API across different model architectures.
Identifies the language of text using a neural network model, returning language codes with confidence scores and reliability indicators.
Integrates Gladia's speech-to-text API with LiveKit Agents, enabling streaming transcription with multi-language support, code-switching, interim results, and optional real-time translation.
Install it if you are building a LiveKit Agents application and need Gladia's speech-to-text capabilities; it is the intended integration point for that use case.
Extends Model Explorer with visualization support for PyTorch Executorch PTE files targeting Arm Ethos-U NPU, Arm VGF, and XNNPACK backends.
Provides quantization, pruning, and clustering techniques to reduce model size and improve inference performance for TensorFlow and Keras models.
InstantTensor provides a high-throughput Safetensors loader optimized for moving model weights from disk to GPU memory, with support for distributed loading via torch.distributed.
Not recommended if you load small models infrequently or have ample host memory for caching—standard Safetensors loading will suffice.
slangtorch lets you write PyTorch kernels in the Slang shading language, compiling them to run on CUDA-capable GPUs.
Unitxt provides a unified framework for evaluating AI model performance across text, tables, vision, speech, and code using a modular catalog of benchmarks and datasets.
Install it if you need to benchmark models against established datasets or build reproducible evaluation workflows.
Provides runtime support for Databricks AutoML, enabling automated machine learning model training and evaluation workflows within the Databricks platform.
Provides CUDA kernels optimized for CPM, enabling GPU-accelerated computation on NVIDIA hardware.
Install only if you are maintaining legacy CPM code that explicitly requires this exact version and your environment is locked to compatible CUDA/Python versions.
Integrates Sarvam.ai's Indian-language voice AI services (speech-to-text, text-to-speech, and LLM chat) into LiveKit Agents for real-time conversational applications.
A deprecated wrapper package that redirects users to azureml-train-core for Azure Machine Learning training functionality.
Jina is a framework for building and deploying AI services that communicate via gRPC, HTTP, and WebSockets, with built-in support for scaling, containerization, and cloud deployment.
Not recommended if you prefer minimal dependencies or need cutting-edge feature velocity.
Python SDK for interacting with LM Studio, a local LLM application, enabling text completion and chat interactions with loaded language models via WebSocket.
However, the 357-day gap since the last release and unclear license terms warrant caution: verify license compliance before commercial use, and expect slower…
Provides CUDA kernels for equivariant neural network operations, loaded as a shared library when imported to enable GPU-accelerated equivariant primitives.
However, accept the proprietary license terms carefully: the SDK is restricted to NVIDIA GPU systems, prohibits reverse engineering, and requires your application to…
Polyglot is a multilingual natural language processing pipeline that performs tokenization, language detection, named entity recognition, part-of-speech tagging, sentiment analysis, word embeddings, morphological analysis, and transliteration across many languages.
No—not for new projects.
TorchIO reads, preprocesses, augments, and samples 3D medical images for deep learning with PyTorch, offering both standard computer vision transforms and domain-specific medical imaging operations like MRI artifact simulation.
Install it if you're building medical imaging models with PyTorch.
Marvin is a Python framework for building AI workflows that produce structured outputs and delegate tasks to language models. It provides utilities for extracting, casting, classifying, and generating data, plus a task-and-agent system for orchestrating multi-step agentic processes.
RecBole provides a unified framework for implementing, benchmarking, and comparing recommendation algorithms across general, sequential, context-aware, and knowledge-based categories.
However, it is research-focused and not designed for production systems; the aging maintenance status and large dependency footprint (16 runtime packages) mean you…
Detects nudity in images using lightweight ONNX models, returning bounding boxes and confidence scores for exposed or covered body parts, with optional image censoring.
Integrates NextPlaid, a ColBERT-style multi-vector search engine, with LangChain as a VectorStore, enabling late-interaction retrieval with metadata filtering.
However, be aware that this is a 0.1.0 release with limited track record; production use should be preceded by thorough testing and verification of performance…
PyGAD is a Python library for building and running genetic algorithms to optimize single-objective and multi-objective problems, with built-in support for training Keras and PyTorch models.
Install it if you need genetic algorithm optimization; skip it if you only do gradient-based machine learning.
TensorFlow Data Validation (TFDV) computes summary statistics, detects anomalies, and generates data schemas for machine learning datasets at scale using Apache Beam and TensorFlow.
Wraps OpenAI's Assistants API as a drop-in replacement, adding support for AstraDB vector stores and enabling credential pass-through to alternative LLM providers via litellm.
However, the aging maintenance status (482 days since last release) and unclear license terms are cautions—verify license compliance and test compatibility with your…
Enforces content moderation on LLM prompts and responses using configurable guard rules, blocking or modifying text according to a YAML-defined policy before and after LLM inference.
However, the proprietary license and unclear license treatment require legal review before use outside DataRobot deployments.