Alipay

This specification provides video memory management methods for large language model inference, devices, media, and products, which are applied to a service device deployed with a large language model. The method includes: allocating physical video memory resources on the service device, to separately map the physical video memory resources to a first video memory resource pool in which a cache object is a key-value cache and a second video memory resource pool in which a cache object is an intermediate activation value; and for an inference task submitted to the large language model, upon determining that an idle video memory resource in any video memory resource pool is insufficient to cache a corresponding cache object for the inference task, temporarily transferring at least a portion of idle video memory resources in another video memory resource pool to the any video memory resource pool.

This specification provides interaction methods for image processing and image processing methods corresponding to the interaction methods. First, a stroke that is used to represent a local area that needs to be changed and that is added by a user to a displayed first image by using an operable control is received, and then a prompt that is used to represent a modification target of the local area and that corresponds to the stroke is displayed. Then, a second image generated based on the stroke, the prompt, and the first image is displayed, where another part of the second image different from an area covered by the stroke is similar to the first image, and content represented by the modification target is correspondingly generated at a position at which the stroke is located in the second image. The user can draw, by adding a stroke to an image, content that the user wants to perform creative modification on the first image, and a device can identify the stroke of the user to generate the content needed by the user. As such, the user can perform efficient creative processing on an image by using simple operations even if the user is not familiar with image processing software.

Embodiments of the present specification disclose a lock control method and apparatus, a medium, and a device. This solution can be applied to a trusted execution environment for execution. A first wireless communication apparatus is disposed in an adjacent area of a lock, a user carries a second communication apparatus, and control permission for the lock is bound to the second communication module. Once the user enters a signal sensing range of the first wireless communication apparatus, the first wireless communication module is triggered to establish a communication connection with the second wireless communication module. Signal strength of the communication connection changes with movement of the second wireless communication apparatus carried by the user. Therefore, a distance change trend of the second wireless communication apparatus relative to the first wireless communication apparatus is determined based on a monitored signal strength change trend. If it is determined that the user is moving close to the lock, the lock can be controlled to be in an unlocked state. If it is determined that the user is moving away from the lock, the lock can be controlled to be in a locked state.

One or more embodiments of this specification provide a model training and checkpoint file storage system and method. The system includes a model training module and a checkpoint file processing module. The model training module is configured to execute a training task of an artificial intelligence model, where computation in the training task is performed by GPU chips; and during the execution of the training task, if a first checkpoint file is generated, the training task is suspended, and a request is made to the checkpoint file processing module to write the first checkpoint file. The checkpoint file processing module is configured to locally cache the first checkpoint file based on a received write request for the first checkpoint file, and then concurrently perform a notification operation and a storage operation, where the notification operation is used to return a write success notification for the first checkpoint file to the model training module, to indicate the model training module to resume the training task, and the storage operation is used to persist the first checkpoint file.

A file access method is applied to a computing node for executing a computing task. A user-space file system that runs in user space is mounted on a system kernel of the computing node. A first process adaptation program is implemented in a first user-space process corresponding to the computing task. A second process adaptation program is implemented in a second user-space process corresponding to the user-space file system. The method includes: The first process adaptation program intercepts an access request of the first user-space process for the user-space file system; the first process adaptation program transmits the access request to the second process adaptation program in the user space, and the second process adaptation program submits the access request to the second user-space process; and the second process adaptation program intercepts an access result generated by performing access processing on the user-space file system by the second user-space process, and transmits the access result to the first process adaptation program in the user space, and the first process adaptation program submits the access result to the first user-space process.

One or more embodiments of this specification provide a system for training an artificial intelligence model and a checkpoint file storage method. The system includes a model training module and a first cache module. The model training module is configured to: read a dataset needed for training from the first cache module, to execute a training task of an artificial intelligence model, where computation of the training task is performed by a GPU chip, and in a process of executing the training task, generate a checkpoint file and send the checkpoint file to the first cache module. The first cache module is configured to: identify a type of obtained to-be-stored data; and if the type of the to-be-stored data is a dataset, first write the to-be-stored data into a local buffer, and then store the to-be-stored data in a local hard disk from the local buffer; or if the type of the to-be-stored data is a checkpoint file, directly store the to-be-stored data in the local hard disk.

Zusammenfassung wird geladen …

Zusammenfassung wird geladen …

Zusammenfassung wird geladen …

Zusammenfassung wird geladen …