Packages
tlparse parses structured PyTorch torch.compile trace logs and generates HTML analysis files from the output.
Automatically selects and tunes machine learning models given training and test data, handling model search and hyperparameter optimization as part of Azure ML's AutoML workflow.
nlpaug generates synthetic augmented text and audio data for machine learning projects by applying transformations like word substitution, insertion, deletion, and audio modifications to improve model robustness and performance.
Install it if you need to augment text or audio data for training; skip it if your training pipeline already has dedicated augmentation logic or if you work…
Integrates Google's Gemini models into LlamaIndex applications, providing a unified interface for text generation, chat, and streaming responses via the Google GenAI API.
Flower is a framework for building federated learning systems where machine learning models are trained collaboratively across distributed clients without centralizing raw data.
Install it if you need to build federated learning systems or conduct distributed model training while preserving data privacy.
Distrax provides JAX-native probability distributions and bijectors (invertible functions with Jacobian computation) as a lightweight reimplementation of a subset of TensorFlow Probability, designed for extensibility and readability.
Install it if you need probability distributions in JAX and value extensibility and mathematical clarity over a comprehensive feature set.
TensorDict is a batched, nested dictionary container for PyTorch tensors that behaves like a single tensor—enabling slicing, reshaping, device transfer, and arithmetic operations across all leaves simultaneously while maintaining a shared batch size.
ReMe is a local-first knowledge base that converts conversations and resources into searchable, interconnected Markdown files with wikilinks, designed to give AI agents and users a persistent, editable memory layer.
ETA is an extensible computer vision and machine learning analytics toolkit that provides core utilities for working with images, videos, embeddings, and ML inference pipelines, along with a CLI for building and running analytics workflows.
TensorFlow Transform preprocesses data with full-pass operations like normalization, vocabulary generation, and bucketing, exporting a reusable TensorFlow graph for consistent training and serving.
Combines vector similarity search with graph traversal over metadata to retrieve documents from existing vector stores, enabling both unstructured and structured retrieval in a single query.
However, the package is in Beta with limited recent activity (last release April 2025), so verify that your vector store is supported and test thoroughly before…
Retrieves documents from a graph structure using vector similarity search, combining graph traversal strategies with LangChain's retriever framework for efficient document discovery.
However, verify that the graph traversal strategies and vector store adapters you need are production-ready before deploying to critical systems.
PyTorch bindings for NVIDIA's CUDA-accelerated library for building equivariant geometric neural networks using segmented polynomials and optimized tensor operations.
However, it is in Beta state, so evaluate stability for your production timeline.
Adds type annotations for PyTorch tensor shape, dtype, layout, and dimension names with optional runtime checking via typeguard.
However, be aware that the package is aging and the author recommends jaxtyping for new projects due to better static type checker support.
FiftyOne Brain provides AI/ML capabilities for analyzing and manipulating datasets and models, including visual similarity search, text-based querying, sample uniqueness detection, and quality/annotation issue identification.
Install it if you work with computer vision datasets and need systematic quality and similarity analysis.
LiteRT is a runtime for running trained machine learning models on mobile and embedded devices with low latency and minimal binary footprint.
However, avoid it for production deployments requiring stability—use a stable release instead.
Provides PyTorch implementations of Vision Transformer (ViT) and related transformer-based vision models for image classification and self-supervised learning tasks.
Real-ESRGAN upscales images and videos using deep learning models trained on synthetic data, with specialized support for anime content and face enhancement through integrated GFPGAN.
Install it if you have a GPU or can tolerate CPU inference; expect to download model weights on first use.
CellPyLib simulates 1- and 2-dimensional cellular automata on regular grids with configurable neighbourhoods, colors, and evolution rules, producing visualizations of their behavior over time.
Coqui TTS synthesizes speech from text using deep learning models, supporting over 1100 languages with pretrained weights and tools for training and fine-tuning custom models.
Install it if you want pretrained models out-of-the-box or plan to fine-tune.
Trinity orchestrates multiple AI coding agents (Claude, Codex, Antigravity) in structured round-based deliberation, reaching consensus before distributing tasks to each agent's specialty.
However, alpha status means workflows and APIs may change; use it for exploration and prototyping rather than critical production pipelines.
Parse documents via a remote Compass Parser API and manage indexed documents through a Compass API client, supporting multiple file types and file systems.
GLiNER2 extracts entities, classifies text, parses structured data, and identifies relations from unstructured text using a single 205M parameter model that runs efficiently on CPU without external API dependencies.
Provides an HTTP server for serving Azure Machine Learning models with built-in support for Flask-based scoring scripts, OpenTelemetry observability, and Azure Monitor integration.
However, review the unclear proprietary license terms before use, and ensure your scoring scripts are compatible with Flask 2.x and Pydantic 2.0+.
Converts between SELFIES (Self-Referencing Embedded Strings) and SMILES molecular representations, ensuring syntactically and semantically valid molecular strings for machine learning applications.
Connects LangChain applications to Databricks services including vector search, chat models, and MLflow integration, but is now deprecated in favor of databricks-langchain.
Python SDK client for connecting to the Labelbox API, enabling programmatic access to data curation, annotation, and model evaluation workflows.
Install it if you need programmatic access to the Labelbox platform; the main requirement is a valid API key.
Provides Python protobuf message definitions and serialization support for Flyte's interface definition language (IDL), enabling communication between Flyte workflow components.
Triton Ascend is a compiler framework that enables Triton code to run efficiently on Ascend hardware, handling memory allocation, data movement, and kernel optimization automatically.
This package is a redirect or placeholder; the actual PyTorch library is distributed under the name 'torch', not 'pytorch'.
NGBoost implements Natural Gradient Boosting for probabilistic prediction, built on scikit-learn, enabling regression models that output full probability distributions rather than point estimates.
Install it if you need prediction intervals, distribution outputs, or uncertainty estimates alongside your regression predictions.
Python client library for calling the Aleph Alpha API, supporting both synchronous and asynchronous requests for text completion, semantic search, and embeddings.
However, the aging maintenance status (196 days since last release) means you should verify API compatibility before relying on it for production work, and monitor…
Dora is an experiment launching and management tool for machine learning projects that handles grid search scheduling, automatic deduplication of runs, and terminal-based monitoring of experiments.
However, the package is dormant (last release 2023-05-23, no recent commits), so expect no active maintenance or support for newer dependencies.
Integrates Rime speech recognition into LiveKit's agent framework for building real-time voice agents that can hear and respond to users.
Provides schedule-free optimizers for PyTorch that eliminate the need to specify stopping time in advance, replacing momentum with interpolation and averaging to achieve faster training without learning rate schedules.
A Python SDK for accessing JigsawStack's hosted AI services—translation, summarization, web scraping, speech-to-text, image recognition, and content generation—through a typed API with webhook and file-upload support.
Tunix is a JAX-based library for post-training large language models with support for supervised fine-tuning, reinforcement learning, and agentic RL workflows, optimized for TPU execution.
However, the 26 runtime dependencies and alpha status mean you should verify that optional dependencies match your use case and accept that the API may evolve.
Visualizes and debugs machine learning model graphs with hierarchical layer expansion, metadata overlay, and interactive exploration across TFLite, TensorFlow, TFJS, MLIR, and PyTorch formats.
Implements a hardware-aware Gram Newton-Schulz algorithm for polar decomposition that serves as a drop-in replacement for standard Newton-Schulz in PyTorch optimizers like Muon, achieving faster convergence through symmetric GEMM kernels on NVIDIA GPUs.
Integrates Speechmatics speech-to-text into LiveKit Agents, enabling real-time transcription with optional speaker diarization and configurable turn detection modes.
Install it if you are building a LiveKit agent and need Speechmatics transcription with speaker diarization or flexible turn detection.