> ML_LIBRARIES_CATALOG_v1.0

Qualified ML Libraries

Every library is independently qualified with primary citations, supported version ranges, hardware accelerator separation, and real-world failure patterns.

Matrix View Navigator

Showing 18 of 161 Qualified Libraries & Tools (Page 3 of 9)Accelerator Mappings & Failure Modes

NLP & LLM Serving

4.3.3LGPL-2.1

Gensim

RaRe Technologies / Radim Řehůřek

Topic Modelling for Humans: Fast Word2Vec, LDA, and Document Similarity in Python.

Training:

CPU

Inference:

CPU

#text

Docs Deep Dive

NLP & LLM Serving

1.9.2Apache-2.0

Stanza

Stanford NLP Group

Official Stanford NLP Python library for deep linguistic analysis across 70+ languages.

Training:

CPUCUDA

Inference:

CPUCUDA

#text

Docs Deep Dive

NLP & LLM Serving

0.14.0MIT

Flair

Humboldt University of Berlin / Open Source

Very simple framework for state-of-the-art NLP, developed by Humboldt University.

Training:

CPUCUDAMPS

Inference:

CPUCUDAMPS

#text

Docs Deep Dive

NLP & LLM Serving

0.9.2MIT

fastText

Meta AI Research (FAIR)

Library for fast text representation and classification developed by Facebook AI Research.

Training:

CPU

Inference:

CPUWASM

#text

Docs Deep Dive

NLP & LLM Serving

0.16.4MIT

BERTopic

Maarten Grootendorst / Open Source

Leveraging transformers and c-TF-IDF to create easily interpretable topics.

Training:

CPUCUDAROCMMPS

Inference:

CPUCUDAROCMMPS

#text

Docs Deep Dive

NLP & LLM Serving

0.13.0Apache-2.0

PEFT

Hugging Face

State-of-the-art Parameter-Efficient Fine-Tuning methods for large pretrained models.

Training:

CPUCUDAROCMMPSXPU

Inference:

CPUCUDAROCMMPS

#text#image#multimodal

Docs Deep Dive

NLP & LLM Serving

0.10.1Apache-2.0

TRL

Hugging Face

Transformer Reinforcement Learning for post-training and alignment.

Training:

CPUCUDAROCMMPS

Inference:

CPUCUDAROCMMPS

#text#multimodal

Docs Deep Dive

privacy-security-optimization

0.43.3MIT

bitsandbytes

bitsandbytes Foundation / Tim Dettmers

Accessible large language models via 8-bit and 4-bit quantization.

Training:

CPUCUDAROCMXPU

Inference:

CPUCUDAROCM

#text#multimodal

Docs Deep Dive

NLP & LLM Serving

0.3.1MIT

LangChain

LangChain, Inc.

Framework for developing context-aware, reasoning applications powered by language models.

Training:

Not Supported (Inference Only)

Inference:

CPUCUDAROCMMPS

#text#multimodal

Docs Deep Dive

NLP & LLM Serving

0.11.13MIT

LlamaIndex

LlamaIndex (Jerry Liu)

The data framework for connecting enterprise data sources to large language models.

Training:

Not Supported (Inference Only)

Inference:

CPUCUDAROCMMPS

#text#multimodal

Docs Deep Dive

NLP & LLM Serving

2.5.0Apache-2.0

Haystack

deepset

An open-source NLP framework for building production-ready LLM pipelines.

Training:

Not Supported (Inference Only)

Inference:

CPUCUDAROCMMPS

#text

Docs Deep Dive

NLP & LLM Serving

2.5.0MIT

DSPy

Stanford NLP Group / Omar Khattab

Programming—not prompting—Foundation Models.

Training:

CPUCUDA

Inference:

CPUCUDA

#text

Docs Deep Dive

Inference Serving

0.6.2Apache-2.0

vLLM

vLLM Project / UC Berkeley

High-throughput and memory-efficient inference and serving engine for LLMs.

Training:

Not Supported (Inference Only)

Inference:

CPUCUDAROCM

#text#multimodal

Docs Deep Dive

Inference Serving

b3650MIT

llama.cpp

Georgi Gerganov / Open Source

Port of Facebook's LLaMA model in C/C++ for efficient local inference.

Training:

Not Supported (Inference Only)

Inference:

CPUCUDAROCMMPSWEBGPUWASM

#text#multimodal

Docs Deep Dive

Inference Serving

0.3.12MIT

Ollama

Ollama, Inc.

Get up and running with large language models locally.

Training:

Not Supported (Inference Only)

Inference:

CPUCUDAROCMMPS

#text#multimodal

Docs Deep Dive

Inference Serving

0.12.0Apache-2.0

TensorRT-LLM

NVIDIA

NVIDIA TensorRT-LLM provides users with an easy-to-use Python API to define and compile LLMs for extreme performance.

Training:

Not Supported (Inference Only)

Inference:

CUDA

#text#multimodal

Docs Deep Dive

Inference Serving

2.3.1HFOIL v1.0

Text Generation Inference

Hugging Face

A purpose-built solution for deploying and serving Large Language Models in production.

Training:

Not Supported (Inference Only)

Inference:

CUDAROCM

#text#multimodal

Docs Deep Dive

Inference Serving

0.3.1Apache-2.0

SGLang

LMSYS Org / UC Berkeley

Fast serving framework for large language models and complex multi-turn programs.

Training:

Not Supported (Inference Only)

Inference:

CUDAROCM

#text#multimodal

Docs Deep Dive