translation

This is an AI translated post.

RevFactory

Building an AI Full Stack with Open Source

Select Language

  • English
  • 汉语
  • Español
  • Bahasa Indonesia
  • Português
  • Русский
  • 日本語
  • 한국어
  • Deutsch
  • Français
  • Italiano
  • Türkçe
  • Tiếng Việt
  • ไทย
  • Polski
  • Nederlands
  • हिन्दी
  • Magyar

Summarized by durumis AI

  • The open source ecosystem is experiencing an AI open source renaissance, with many models being released by the Open LLM camp.
  • Various inference and serving tools and LLM monitoring and management tools are emerging for LLM utilization.
  • Various frameworks for developing LLM-based applications are being introduced.

With the surge of numerous open-source projects related to AI, the open-source ecosystem is experiencing a true renaissance in AI open-source. Following the success of LangChain, many open-source projects have emerged, rapidly filling the gaps in AI industry systems.


Open LLM

LLM (Large Language Model), the core of generative AI, is divided into two axes: Closed LLM led by GPT and Open LLM in the Llama camp. Mistral team has released its model under an open-source license, drawing attention from many with its outstanding performance. Open LLM is primarily managed and provided through Hugging Face.


Mixtral-8x7B-Instruct-v0.1(Apache-2.0)


https://huggingface.co/mistralai/Mixtral-8x7B-Instruct-v0\.1


- Expert Mixture of Experts (SMoE) model was adopted.


- It shows performance that surpasses the Llama 2 70B model and even GPT-3.5 175B.


- It ranks third in Chatbot Arena, a blind chatbot test, following GPT-4 and Claude-2.


Llama-2-7b-chat(Llama 2 Community)


https://huggingface.co/meta-llama/Llama-2-7b-chat


- It is a license that allows commercial use for services with less than 700 million monthly active users.


- Numerous derivative models based on Llama-2 have emerged.


phi-2(MIT)


https://huggingface.co/microsoft/phi-2


- It is a lightweight model with 2.7B parameters released by MS.


- Tests on common sense, language understanding, and logical reasoning have shown that its performance is better than the 13B model.


LLM Inference and Serving

To effectively utilize well-trained LLMs, tools are needed that can provide fast inference and efficient management of computing resources.


Ollama(MIT)


https://github.com/jmorganca/ollama


- It allows you to run LLMs up to 7B in size directly on local environments like Mac, Linux, and Windows.


- You can download and run models with simple commands.


- Models can be managed through the CLI, and simple chatting is possible.


- Various applications are possible through the provided API.


vLLM(Apache-2.0)


https://github.com/vllm-project/vllm


- It is a fast and easy-to-use library for LLM inference and serving.


- Supports models provided by Hugging Face.


- Provides distributed processing, parallel processing, streaming output, and OpenAI compatible API.


- Supports Nvidia and AMD GPUs.


KServe(Apache-2.0)


https://github.com/kserve/kserve- It is a platform for ML model inference that can be built in a Kubernetes environment. - Provides an abstraction interface for scaling, networking, and monitoring.


LLM Proxying

LiteLLM(MIT)


https://github.com/BerriAI/litellm


- Integrates various LLM APIs and provides proxying.


- Follows the API format of OpenAI.


- Provides user-specific API authentication management.


One API(MIT)


https://github.com/songquanpeng/one-api


- It enables instant access to all large models through the standard OpenAI API format.


- Supports various LLMs and also provides proxy services.


- Load balancing and multi-deployment are possible, and it provides user management and group features.


AI Gateway(MIT)


https://github.com/Portkey-AI/gateway


- Provides connections to over 100 LLMs with a single, fast, and familiar API.


- Ensures fast access with a small installation size.


LLM Monitoring Great Expectations(Apache-2.0)


https://github.com/great-expectations/great_expectations


- Helps data teams build a shared understanding of their data through quality testing, documentation, and profiling.


- It can be integrated with CI/CD pipelines to add data quality precisely where needed.


LangFuse(MIT)


https://github.com/langfuse/langfuse


- Provides open-source LLM visibility, analysis, rapid management, evaluation, testing, monitoring, logging, and tracing.


- You can explore and debug complex logs and traces from a visual UI.


- Enterprise features are planned for the future.


Giskard(Apache-2.0, Dual License)


https://github.com/Giskard-AI/giskard


- It can automatically detect vulnerabilities in AI models, from tabular models to LLMs, such as bias, data leakage, spurious correlations, hallucinations, toxicity, and security issues.


- It scans for vulnerabilities in AI models and automatically generates test suites to support quality assurance processes for ML models and LLMs.


- Provides a SaaS platform for detecting AI safety risks in deployed LLM applications. (Premium)


LLM Framework

LangChain (MIT)


https://github.com/langchain-ai/langchain


- It is a framework for developing applications powered by language models.


- It is available in Python and Javascript, providing an abstraction layer that integrates numerous libraries.


- You can also deploy the built LangChain as an API.


LlamaIndex(MIT)


https://github.com/run-llama/llama_index


- It is a data-centric framework for LLM applications.


- Provides data connectors for collecting existing data sources and data formats (API, PDF, documents, SQL, etc.).


- Provides methods for structuring data (indexes, graphs) to make it easily usable by LLMs.


Haystack(Apache-2.0)


https://github.com/deepset-ai/haystack


- It is an LLM framework for easy construction of search augmented generation (RAG), document search, question answering, and answer generation.


- It is built based on the pipeline concept.


Flowise(Apache-2.0)


https://github.com/FlowiseAI/Flowise


- You can build custom LLM flows by dragging and dropping the UI.


LangFlow(MIT)


https://github.com/logspace-ai/langflow


- It makes it easy to experiment and prototype LangChain pipelines.


- It is executed using the CLI, and it also supports deploying Langflow to Google Cloud Platform (GCP).


Spring AI(Apache-2.0)


https://github.com/spring-projects/spring-ai


- AI framework provided by Spring Framework (still in snapshot stage)


- Supports API integration based on OpenAI and MS Azure, providing an abstraction layer


- The goal is to make it easier and more scalable to implement AI features using AI Templates



Data Juicer(Apache-2.0)


https://github.com/alibaba/data-juicer


- It is an open-source project released by Alibaba, providing a one-stop data processing system for LLMs.


- It provides a systematic library comprised of over 20 reusable configuration recipes, over 50 core OPs, and a rich functional dedicated toolkit.


- With automated report generation capabilities, it allows for in-depth data analysis to gain a deeper understanding of data sets.


revfactory
RevFactory
RevFactory
revfactory
Snowflake Launches 'Arctic,' an Enterprise-Grade LLM with Industry-Leading Openness Snowflake has launched 'Arctic,' an open-source large language model. It's free for commercial use under the Apache 2.0 license and boasts industry-leading performance and efficiency. It offers exceptional token efficiency and top-notch quality, supportin
스타트업 커뮤니티 씬디스 (SeenThis.kr)
스타트업 커뮤니티 씬디스 (SeenThis.kr)
스타트업 커뮤니티 씬디스 (SeenThis.kr)
스타트업 커뮤니티 씬디스 (SeenThis.kr)

April 25, 2024

Mr. Know-All – 2023.7 The first issue of "Mr. Know-All," a monthly AI magazine in July 2023, introduces the latest AI technologies and trends, including Claude 2, Azure OpenAI, LangChain, and LlamaIndex. In particular, it provides a detailed explanation of LlamaIndex, which em
Pilot AISmrteasy
Pilot AISmrteasy
Pilot AISmrteasy
Pilot AISmrteasy

March 21, 2024

What is LLM (Large Language Model)? Large Language Models (LLMs) are a core technology in artificial intelligence, learning from vast amounts of text data to acquire language processing abilities similar to humans. They can be used in various fields such as chatbots, translation, and text g
세상 모든 정보
세상 모든 정보
세상 모든 정보
세상 모든 정보

April 1, 2024

The Paradox of Leading AI Models, Transparency The lack of transparency in cutting-edge AI systems has emerged as a serious problem. Researchers at Stanford University analyzed 10 AI systems, including GPT-4, and concluded that there are no models that transparently disclose information such as data s
Byungchae Ryan Son
Byungchae Ryan Son
Byungchae Ryan Son
Byungchae Ryan Son
Byungchae Ryan Son

May 14, 2024

Apple's OpenELM / MS's Phi-3 / Meta's Llama 3 Released Major tech companies such as Apple, Microsoft, and Meta are injecting new energy into the AI industry by recently releasing their own large language models. These newly released models are evolving in various directions, including size reduction, data opt
해리슨 블로그
해리슨 블로그
해리슨 블로그
해리슨 블로그

April 27, 2024

SK C&C Unveils 'Soluer LLMOps,' a Platform Supporting Customized sLLM Implementation for Clients SK C&C has launched 'Soluer LLMOps,' a platform for building customized small-scale large language models (sLLMs) for enterprises. The platform supports easy creation of sLLMs using drag-and-drop functionality, leveraging various foundation models such as
스타트업 커뮤니티 씬디스 (SeenThis.kr)
스타트업 커뮤니티 씬디스 (SeenThis.kr)
스타트업 커뮤니티 씬디스 (SeenThis.kr)
스타트업 커뮤니티 씬디스 (SeenThis.kr)

May 20, 2024