search

UbiquitousLearning / mllm

Fast Multimodal LLM on Mobile Devices
https://ubiquitouslearning.github.io/mllm_website
MIT License
504 stars 57 forks source link

Feat: Optimize the operation process #69

Closed yirongjie closed 7 months ago

yirongjie commented 7 months ago

Optimize the operation process:

  1. The process of running Op is: setup+execute;
  2. The initial state of the activation Tensor is TENSOR_STATIC_INIT, and after Op's setup(), it changes to TENSOR_STATIC_READY. Afterwards, Op's execute() can be executed