Skip to content

Fix num_gpus when TP > 1 #1851

Closed
Closed
@WoosukKwon

Description

@WoosukKwon

For now we should do:

num_gpus=self.cache_config.gpu_memory_utilization if self.parallel_config.tensor_parallel_size < 2 else 1

Originally posted by @Yard1 in #1821 (comment)

Metadata

Metadata

Assignees

No one assigned

    Labels

    No labels
    No labels

    Type

    No type

    Projects

    No projects

    Milestone

    No milestone

    Relationships

    None yet

    Development

    No branches or pull requests

    Issue actions