VentureBeat presents: AI Unleashed – An exclusive executive event for enterprise data leaders. Network and learn with industry peers. Learn More

A new platform that allows users to fine-tune open-source large language models (LLMs) without writing any code has recently launched new features that make the process even easier and faster.

The platform, called MonsterAPI, was created by a team of researchers and developers who wanted to make LLMs more accessible and affordable for everyone. LLMs are powerful artificial intelligence systems that can generate natural language texts for various tasks, such as writing, summarizing, translating, answering questions, and more.

However, LLMs are not perfect. They often have very general knowledge but struggle to solve specific problems. To make them more accurate and relevant, they need to be “fine-tuned,” which means teaching them how to perform a particular task using a custom dataset.

Fine-tuning LLMs is not a simple task. It requires a lot of time, effort, and GPU computing power. It also involves finding the optimal hyperparameters and dealing with underfitting and overfitting issues. Moreover, it is hard to find experienced people who know how to do it.


AI Unleashed

An exclusive invite-only evening of insights and networking, designed for senior enterprise executives overseeing data stacks and strategies.


Learn More

MonsterAPI aims to solve these problems by offering a no-code solution for fine-tuning LLMs. Users can choose from a variety of open-source models, such as Llama and Llama 2 7B, 13B and 70B; Falcon 7B and 40B; Open Llama; OPT; GPT J; and Mistral 7B. They can also upload their own datasets or use pre-made ones from the platform’s library. Then, they can fine-tune the models using a simple interface that guides them through the process.

The platform also uses a decentralized GPU platform that reduces the cost and increases the speed of fine-tuning. Users can pay as they go or subscribe to a plan that suits their needs. The platform also offers free credits for new users who sign up with a code.

The team behind MonsterAPI has recently announced new features that make the platform even better. These include:

  • QLora with 4-bit quantization and nf4: This feature reduces the size of the models by compressing them using quantization techniques. This allows users to fine-tune larger models using less memory and bandwidth.
  • Flash Attention 2: This feature improves the speed and efficiency of training by using a novel attention mechanism that reduces the computational complexity of the models.
  • Data and model parallelism on multiple GPUs: This feature enables users to train bigger models using larger context lengths by distributing the data and the model across multiple GPUs.

MonsterAPI has mostly received positive feedback from its users, who have used it for various purposes, such as creating content, generating summaries, building chatbots, and more. The platform also has an active community on Discord, where users can share their results, ask questions, get support, and receive updates and offers from the team.

MonsterAPI is one of the first platforms that offers no-code fine-tuning of open-source LLMs. It aims to democratize access to LLMs and make them more useful and affordable for everyone. To learn more about MonsterAPI or to sign up for free credits, visit http://monsterapi.ai or join their Discord server.

VentureBeat’s mission is to be a digital town square for technical decision-makers to gain knowledge about transformative enterprise technology and transact. Discover our Briefings.


Source link