On November 21st, the "Building Foundation and Empowering Intelligent Future" Jiuzhang Yunji DataCanvas large model series results conference (hereinafter referred to as the "press conference") was held in Beijing. This results release was held on June 30 this year. The Alaya Jiuzhang Yuanshi large model was released just over 4 months ago, which is a testament to the innovative capabilities of Jiuzhang Yunji DataCanvas' large model technology application.
At the press conference, a series of large model innovations focusing on the DataCanvas Alaya Jiuzhangyuanshi large model, including the open source underlying LLMOps large model tool chain, the open source Alaya-7B large model series, and the TableAgent data analysis agent, were released. Industry-leading model performance, zero threshold, and high-value data analysis applications once again refresh the cognitive ceiling of the large model application industry!
During the same period of the conference, Guangdong Private Investment Co., Ltd. (hereinafter referred to as "Guangdong Min Investment"), Tibet Saifu Heyin Investment Co., Ltd. (hereinafter referred to as "Tibet Saifu"), Guangzhou Saifu Heyin Asset Management Co., Ltd. (hereinafter referred to as "Saif") Asset Management"), and officially signed the "Strategic Cooperation to Jointly Promote the Popularization and Update of Artificial Intelligence in the Industrial Field" and the "Strategic Cooperation to Expand the Ecosystem of Intelligent Computing Center" with Beijing Jiuzhang Yunji Technology Co., Ltd. (hereinafter referred to as "Jiuzhang Yunji DataCanvas") 》. In addition, Jiuzhang Yunji and Hanbo Semiconductor signed the "Strategic Cooperation to Jointly Build an Intelligent Computing Cluster Ecosystem" agreement, which will promote the integration and innovation of artificial intelligence software + hardware, the large-scale application of artificial intelligence technology in the industrial field, and create a nationally leading and internationally recognized company. Influential artificial intelligence innovative application demonstration area.
Alaya-7B general knowledge & dialogue large model + LLMOps tool chain, the large model "Family Bucket" is open source
At the press conference, the DataCanvas Alaya Jiuzhang Yuanshi large model matrix independently developed by Jiuzhang Yunji DataCanvas Company officially open sourced a series of new results, including the Alaya-7B Foundation Model general knowledge large model in the "Alaya-7B large model series" There are two major models, the Alaya-7B Chat Model dialogue large model, and the two major tools of the "LLMOps large model tool chain", the LMS model running tool and the LMPM prompt word manager.
Jiuzhang Yunji DataCanvas Alaya-7B series large models are officially open source
[Alaya-7B with lower technical threshold and less computing power consumption]
Dr. Yu Jiangang, Vice President of Jiuzhang Yunji DataCanvas Company, introduced that the Alaya-7B large model series is one of the members of the DataCanvas Alaya Jiuzhang Yuanshi large model matrix. It is based on the Alaya general knowledge large model and is self-collected and carefully screened. It is pre-trained from 0 on the trillion token data set (including Chinese and English articles, news, encyclopedia and other data sources on the Internet). In previous participation in authoritative large model evaluation lists such as C-Eval, CMMLU, AGIEval, MMLU, and BBH, Alaya-7B has shown industry-leading standards.
Alaya-7B Chat Model is a conversational version of the Alaya-7B Foundation Model. It is fine-tuned on carefully selected fine-tuned data sets and detoxifies data based on drug-related, pornographic-related and adverse biases to generate data that is aligned with human values. Conversational large model. Alaya-7B Chat Model has the capabilities of multiple rounds of dialogue, self-awareness and biased refusal to answer, and can complete multiple language tasks such as knowledge Q&A, code writing, information extraction, reading comprehension, and creative writing.
Dr. Yu Jiangang said that while ensuring model performance, the Alaya-7B large model series has lower hardware requirements for users to install and use, lower application technical difficulty, and consumes less computing resources required for training, which helps Accelerate the practical application of large models in various industry scenarios.
Jiuzhang Yunji DataCanvas LLMOps large model tool chain is officially open source
【Full life cycle large model tool chain】
The LLMOps large model tool chain covers the entire life cycle of large models from training, fine-tuning, compression, deployment, inference to monitoring. The two major tools open sourced this time - LMS (Large Model Serving) model running tool and LMPM (Large Model Prompt Manager) prompt word manager - both play a key role in the process of large model business empowerment.
-
List itemLMS model running tool can improve the delivery speed and quality of large models, reduce the maintenance and operating costs of large models, and complete the needs of large model production and service operations.
-
The List item LMPM prompt word manager guides large models to generate more accurate, reliable, and expected output content by helping users design better prompt words. This tool can not only provide development toolkit development mode for technical personnel, but also provide human-computer interaction operation mode for non-technical personnel, meeting the needs of more people to use large models.
Jiuzhang Yunji DataCanvas is constantly trying to integrate innovations in tool chains, large models and industry applications. Previously, one of the tool chains, DingoDB multi-modal vector database, and DataCanvas Alaya Jiuzhang Yuanshi large model jointly created an enterprise knowledge steward solution, empowering enterprises to build highly automated and intelligent enterprise knowledge bases, and accelerate multi-modal large models Floor application.
From DingoDB multi-modal vector database, LMS model running tool, LMPM prompt word manager, to Alaya-7B general knowledge + dialogue large model series, Jiuzhang Yunji DataCanvas has provided users with a one-stop solution from data management to large model application. formula, zero threshold, full chain open source tool combination.
Open source address:
Alaya-7B大model：https://github.com/DataCanvasIO/Alaya
DingoDB multi-mode vector database:https://github.com/DingoDB
LMS model running tool:https://github.com/DataCanvasIO/LMS
LMPM prompt word manager:https://github.com/DataCanvasIO/LMPM
Hit the most commercially valuable application scenarios! TableAgent makes everyone a data analyst
This conference officially released the public beta of the TableAgent data analysis agent, which is also an intelligent agent innovation based on the DataCanvas Alaya nine chapters of the Yuanshi large model and the LLMOps large model tool chain.
TableAgent data analysis agent officially released
Yang Jian, chief architect of Jiuzhang Yunji DataCanvas Company, said that TableAgent evolved on the basis of Alaya Yuanshi. It is a breakthrough in interactive structured data analysis from 0 to 1. It is a new way of enterprise data analysis, allowing " "Everyone is a data analyst" can move from dream to reality.
TableAgent is an enterprise-level data analysis agent that can be deployed privately. It has very powerful intent understanding capabilities, analysis modeling capabilities and insights. After fully understanding user intentions, TableAgent can autonomously use advanced modeling technologies such as statistical science, machine learning, and causal inference to mine value from data, and then provide insights to analyze opinions and guide actions. This heuristic and guided analysis capability can continuously mine the information and value in the data from the depth and breadth, helping users complete high-quality analysis work.
TableAgent data analysis agent features and advantages
At the same time, thanks to its own large model and self-developed T+ underlying system, TableAgent can be applied to various industries and professions to achieve professional fine-tuning of personalized data analysis scenarios in any specific field.
Yang Jian pointed out that the current market presents a rich and diverse form of generative AI. Putting aside a number of superficial forms and scenarios, the TableAgent team found that "data analysis" is a deeper step in the integration of large models and specific businesses, and is the best way to The core areas where users generate direct commercial value will also be the areas with the most business value that companies really need to accumulate. Consistent with Jiuzhang Yunji DataCanvas’ AI basic software research and development goals of “all for application”, TableAgent focuses on data analysis. As a product of Data+AI, it will transform huge business value for enterprises in the future AI era dominated by large models, welcoming There are immeasurable blue ocean opportunities.
TableAgent public location: https://tableagent.DataCanvas.com
The "algorithm + computing power" integrated service is advancing at full speed, and powerful forces are joining forces to build a strong artificial intelligence ecosystem!
At the press conference, Jiuzhang Yunji DataCanvas Company formally signed three major strategic cooperation with its upstream and downstream ecological partners. The cooperation areas cover independent innovation of AI software and hardware, large-scale application of AI technology in the industrial field and the construction of localized intelligent computing power, realizing the company's The "algorithm + computing power" integrated service ecological strategy has advanced!
Based on the industry's leading AI technology accumulation and productization capabilities, Jiuzhang Yunji DataCanvas Company and Hanbo Semiconductor signed a "Strategic Cooperation Agreement to Jointly Build an Intelligent Computing Cluster Ecosystem". Both parties will make full use of their respective advantages and adopt excellent domestic software on a large scale. Hardware products create a full-featured new intelligent computing cluster that covers hardware and software and can provide full-stack capabilities from bottom-level computing power to top-level application enablement.
Jiuzhang Yunji DataCanvas Company and Tibet Saifu signed a "Strategic Cooperation to Jointly Promote the Popularization and Update of Artificial Intelligence in the Industrial Field"
Relying on its leading advantages and rich industry experience in AIFS artificial intelligence basic software products including large model series tools, Tibet Saifu officially signed the "Strategic Cooperation to Jointly Promote the Popularization and Update of Artificial Intelligence in the Industrial Field" with Jiuzhang Yunji DataCanvas. The aim is to jointly actively promote the latter's artificial intelligence system solutions and industrial applications in advanced manufacturing and hydropower, thermal power, wind power, energy storage, mining, petrochemical and other industrial fields, and promote the application and popularization of artificial intelligence technology in domestic industrial fields. renew.
Jiuzhang Yunji DataCanvas Company signed the "Strategic Cooperation for Ecological Development of Intelligent Computing Center" with Guangdong Mint Investment and SAIF Asset Management
During the same period, Guangdong Mint Investment, SAIF Asset Management and Jiuzhang Yunji DataCanvas Company jointly signed the "Strategic Cooperation for Ecological Expansion of Intelligent Computing Center", which added strong impetus to the company's "algorithm + computing power" integrated service strategic upgrade. According to the strategic cooperation agreement, all parties will cooperate to jointly promote the creation of a new domestic intelligent computing center that is leading in scale, independent and controllable, and promote industrial integration innovation and cooperative ecological construction plans, using the advanced software and hardware technologies and technologies of each party in the field of artificial intelligence. Partner resources empower the real industry, promote my country's innovative application of artificial intelligence industries such as general large models and vertical large models, promote industrialization transformation, and create benchmarking intelligent computing center demonstration projects.
Dr. Fang Lei, Chairman of Jiuzhang Yunji DataCanvas, delivered a speech
Dr. Fang Lei, chairman of Jiuzhang Yunji DataCanvas Company, gave a keynote speech on "The Development Road of Domestic AI Computing Power in the New Era" at the press conference. He pointed out that in the first year of the big model, the company grasped the times, persisted in leading, and cooperated with the big models. Models coevolve. The development of large models not only provides fertile ground for the deeper and wider application of artificial intelligence technology in various industries, but also brings unprecedented technological boost and market opportunities to the innovative integration of software and hardware based on AI technology. Among them, the domestic replacement of AI computing power is a new field in which my country's artificial intelligence software and hardware empower each other. It has great potential. Working hand in hand with high-quality ecological partners will accelerate the realization of a win-win situation for the times and us.
This large model series results conference further established Jiuzhang Yunji DataCanvas as the leader in artificial intelligence basic software. In the future, the company will anchor AI applications and deepen the Ecological cooperation in the AI industry chain accelerates the continuous infusion of independent innovative AI basic software source power into AI Foundation Service artificial intelligence basic services!