Whether the pre- and post-processing operations of batch processing are parallel
pengxin233 opened this issue ยท 1 comments
pengxin233 commented
๐ The doc issue
During batch processing, torchserve accumulates the number of corresponding batches. When preprocessing, are there parameters that can control torchserve to perform preprocessing in parallel, and then perform inference together? Or do I need to implement parallel logic in the handle myself?
Suggest a potential alternative/fix
No response
agunapal commented
@pengxin233 you can use microbatching in TorchServe to do this
https://github.com/pytorch/serve/tree/master/examples/micro_batching