


The effect can reach 96% of the equivalent model of OpanAI, and the domestic open source AI language model TigerBot is released
According to news on June 8, the domestic multi-modal large language model TigerBot was officially released recently, including two versions of 7 billion parameters and 180 billion parameters. It is now open source on GitHub.
▲ Picture source TigerBot’s GitHub page
It is reported that the innovation brought by TigerBot mainly lies in:
- Propose instructions and complete them Innovative algorithms for supervised fine-tuning improve model learnability.
- Use ensemble and probabilistic modeling methods to achieve controllable facts and creativity.
- Break through the memory and communication issues in mainstream frameworks such as deep-speed in parallel training.
In addition, this model also makes more suitable optimizations from the tokenizer to the training algorithm for the more irregular distribution of the Chinese language.
Researcher Chen Ye said on the official website of Hubo Technology: "This model can quickly understand what type of questions humans have asked using only a small number of parameters. According to the OpenAI InstructGPT paper on the public NLP data set According to the automatic evaluation, TigerBot-7B has reached 96% of the comprehensive performance of OpenAI models of the same size."
▲ Picture source TigerBot's GitHub page
According to According to the report, the performance of TigerBot-7B-base is "better than that of OpenAI's comparable models." The open source code includes basic training and inference code, and the quantification and inference code of the dual-card inference 180B model. The data includes 100G pre-training data and 1G or 1 million pieces of data for supervised fine-tuning.
IT House friends canfind GitHub’s open source projects here.
The above is the detailed content of The effect can reach 96% of the equivalent model of OpanAI, and the domestic open source AI language model TigerBot is released. For more information, please follow other related articles on the PHP Chinese website!

Hot AI Tools

Undresser.AI Undress
AI-powered app for creating realistic nude photos

AI Clothes Remover
Online AI tool for removing clothes from photos.

Undress AI Tool
Undress images for free

Clothoff.io
AI clothes remover

AI Hentai Generator
Generate AI Hentai for free.

Hot Article

Hot Tools

Notepad++7.3.1
Easy-to-use and free code editor

SublimeText3 Chinese version
Chinese version, very easy to use

Zend Studio 13.0.1
Powerful PHP integrated development environment

Dreamweaver CS6
Visual web development tools

SublimeText3 Mac version
God-level code editing software (SublimeText3)

Hot Topics

OpenAI recently announced the launch of their latest generation embedding model embeddingv3, which they claim is the most performant embedding model with higher multi-language performance. This batch of models is divided into two types: the smaller text-embeddings-3-small and the more powerful and larger text-embeddings-3-large. Little information is disclosed about how these models are designed and trained, and the models are only accessible through paid APIs. So there have been many open source embedding models. But how do these open source models compare with the OpenAI closed source model? This article will empirically compare the performance of these new models with open source models. We plan to create a data

In 2023, AI technology has become a hot topic and has a huge impact on various industries, especially in the programming field. People are increasingly aware of the importance of AI technology, and the Spring community is no exception. With the continuous advancement of GenAI (General Artificial Intelligence) technology, it has become crucial and urgent to simplify the creation of applications with AI functions. Against this background, "SpringAI" emerged, aiming to simplify the process of developing AI functional applications, making it simple and intuitive and avoiding unnecessary complexity. Through "SpringAI", developers can more easily build applications with AI functions, making them easier to use and operate.

Author丨Compiled by TimAnderson丨Produced by Noah|51CTO Technology Stack (WeChat ID: blog51cto) The Zed editor project is still in the pre-release stage and has been open sourced under AGPL, GPL and Apache licenses. The editor features high performance and multiple AI-assisted options, but is currently only available on the Mac platform. Nathan Sobo explained in a post that in the Zed project's code base on GitHub, the editor part is licensed under the GPL, the server-side components are licensed under the AGPL, and the GPUI (GPU Accelerated User) The interface) part adopts the Apache2.0 license. GPUI is a product developed by the Zed team

If the answer given by the AI model is incomprehensible at all, would you dare to use it? As machine learning systems are used in more important areas, it becomes increasingly important to demonstrate why we can trust their output, and when not to trust them. One possible way to gain trust in the output of a complex system is to require the system to produce an interpretation of its output that is readable to a human or another trusted system, that is, fully understandable to the point that any possible errors can be found. For example, to build trust in the judicial system, we require courts to provide clear and readable written opinions that explain and support their decisions. For large language models, we can also adopt a similar approach. However, when taking this approach, ensure that the language model generates

Not long ago, OpenAISora quickly became popular with its amazing video generation effects. It stood out among the crowd of literary video models and became the focus of global attention. Following the launch of the Sora training inference reproduction process with a 46% cost reduction 2 weeks ago, the Colossal-AI team has fully open sourced the world's first Sora-like architecture video generation model "Open-Sora1.0", covering the entire training process, including data processing, all training details and model weights, and join hands with global AI enthusiasts to promote a new era of video creation. For a sneak peek, let’s take a look at a video of a bustling city generated by the “Open-Sora1.0” model released by the Colossal-AI team. Open-Sora1.0

Ollama is a super practical tool that allows you to easily run open source models such as Llama2, Mistral, and Gemma locally. In this article, I will introduce how to use Ollama to vectorize text. If you have not installed Ollama locally, you can read this article. In this article we will use the nomic-embed-text[2] model. It is a text encoder that outperforms OpenAI text-embedding-ada-002 and text-embedding-3-small on short context and long context tasks. Start the nomic-embed-text service when you have successfully installed o

Microsoft and OpenAI were revealed to be investing large sums of money into a humanoid robot startup at the beginning of the year. Among them, Microsoft plans to invest US$95 million, and OpenAI will invest US$5 million. According to Bloomberg, the company is expected to raise a total of US$500 million in this round, and its pre-money valuation may reach US$1.9 billion. What attracts them? Let’s take a look at this company’s robotics achievements first. This robot is all silver and black, and its appearance resembles the image of a robot in a Hollywood science fiction blockbuster: Now, he is putting a coffee capsule into the coffee machine: If it is not placed correctly, it will adjust itself without any human remote control: However, After a while, a cup of coffee can be taken away and enjoyed: Do you have any family members who have recognized it? Yes, this robot was created some time ago.

Open AI’s ChatGPT Mac application is now available to everyone, having been limited to only those with a ChatGPT Plus subscription for the last few months. The app installs just like any other native Mac app, as long as you have an up to date Apple S
