Resource allocation method and apparatus and artificial intelligence training system

An artificial intelligence training system, comprising at least one node, a node manager and a client plugin library, a plurality of development environments being created in each node;

• each of the plurality of development environments is provided with a corresponding graphics processing unit quota, the graphics processing unit quota is a maximum available graphics processing unit resource of the development environment input by a user for a corresponding development environment when each of the plurality of development environments is created, wherein
• the client plugin library is configured for, when detecting that a target training task in a target development environment is started, generating a target graphics processing unit request after redirecting a loading process of a target deep learning framework for the target training task, and sending the target graphics processing unit request to the node manager;
• the node manager is configured for determining a target graphics processing unit quota pre-configured for the target development environment, and according to the target graphics processing unit quota and the target graphics processing unit request, allocating the graphics processing unit resources for the target training task;
• and
• the target deep learning framework performs artificial intelligence training using the graphics processing unit resources allocated by the node manager;
• wherein the target graphics processing unit quota further comprises a target graphics processing unit memory quota capacity;
• the node manager is further configured for, according to the target graphics processing unit memory quota capacity, allocating a graphics processing unit memory capacity for the target training task.

An electronic device, comprising a processor, a memory, and a computer program stored on the memory and capable of running on the processor, wherein the computer program, when executed by the processor, implements resource operations applied to an artificial intelligence training system, the artificial intelligence training system comprises a client plugin library and at least one node, a plurality of development environments are created in each node, each of the plurality of development environments is provided with a corresponding graphics processing unit quota, the graphics processing unit quota is a maximum available graphics processing unit resource of the development environment input by a user for a corresponding development environment when each of the plurality of development environments is created, and the operations comprises:

• when it is detected that a target training task in a target development environment is started, acquiring a target graphics processing unit request from the client plugin library, wherein the target graphics processing unit request is generated after redirecting a loading process of a target deep learning framework for the target training task when the client plugin library detects that the target training task in the target development environment is started;
• determining a target graphics processing unit quota pre-configured for the target development environment;
• and
• according to the target graphics processing unit quota and the target graphics processing unit request, allocating the graphics processing unit resources for the target training task;
• wherein the target graphics processing unit quota further comprises a target graphics processing unit memory quota capacity, and the operation of, according to the target graphics processing unit quota and the target graphics processing unit request, allocating the graphics processing unit resources for the target training task further comprises:
• according to the target graphics processing unit memory quota capacity, allocating a graphics processing unit memory capacity for the target training task.