On April 22nd, Alibaba Cloud's Pailitao Large Model Service Platform announced the launch of limited-time free training, deployment, and inference services for the Llama 3 series in China.
This service means that enterprises and developers can now conveniently build their own large models based on the Llama 3 series on Alibaba Cloud without worrying about the cumbersome training and deployment processes.
Llama 3, as a new generation of open-source large language models introduced by Meta, has demonstrated powerful performance with its 8 billion and 70 billion parameter versions. However, for specific language or vertical domain requirements, further training or fine-tuning of these open-source models is often necessary. This process may have been complex in the past, but now, Alibaba Cloud's Pailitao platform offers a one-stop solution, making everything simple.
Alibaba Cloud's Hundred Refinements platform not only offers one-click training, deployment, and inference solutions for Llama 3 but also temporarily waives the computational costs for model development and invocation, significantly reducing the financial burden on enterprises and developers.
Developers simply need to apply to use Llama 3 on the Hundred Refinements Model Plaza. Once approved, they can quickly experience the performance of Llama 3 and intuitively compare it with other models.
For users who wish to directly utilize Llama 3 for inference applications, the Hundred Refinements platform provides features such as model debugging and prompt templates, enabling developers to swiftly build their AI assistants based on Llama 3.
For enterprises looking to undertake deep development, the Hundred Refinements platform supports integration with enterprise datasets, enabling one-click full-parameter training or efficient training. It also offers advanced features such as RAG retrieval enhancement and intelligent agent creation to meet diverse application requirements.
The robust services provided by Alibaba Cloud's Hundred Refinements platform are backed by its powerful AI infrastructure capabilities. The Lingjun Intelligent Computing Cluster, as Alibaba Cloud's infrastructure layer, boasts scalability of up to hundreds of thousands of GPUs, providing a solid hardware foundation for training large-scale models.
At the AI platform layer, the Platform of Artificial Intelligence (PAI) offers efficient and flexible scheduling capabilities, significantly improving the training performance of large models and ensuring stable task execution.
Additionally, at the model-as-a-service level, Alibaba Cloud has built the largest AI model community in China, MagicDart, and the model service platform Hundred Refinements, providing enterprises and developers with one-stop services.
From model selection, experience, retraining, deployment, application, to inference, Alibaba Cloud offers comprehensive support, allowing enterprises and developers to focus more on model innovation and application without worrying about the technical implementation details.
