jcf94 commented on a change in pull request #8808: URL: https://github.com/apache/tvm/pull/8808#discussion_r696225637
########## File path: src/runtime/contrib/tensorrt/tensorrt_runtime.cc ########## @@ -66,7 +78,16 @@ class TensorRTRuntime : public JSONRuntimeBase { use_implicit_batch_(true), max_workspace_size_(size_t(1) << 30), max_batch_size_(-1), - multi_engine_mode_(false) {} + multi_engine_mode_(false) { + const bool use_int8 = dmlc::GetEnv("TVM_TENSORRT_USE_INT8", false); Review comment: > Hi @FrozenGene > We already have environment variables for runtime options such as `TVM_TENSORRT_USE_FP16`. This calibration is purely done at runtime, so it is too restricting to require user to specify this during compilation. I think environment variable is the only clean way to add new runtime functionality. OK, I agree with the part that this's purely a runtime behavior. Though I still think using environment variable here is just like a "jump wire", I don't have better suggestions. -- This is an automated message from the Apache Git Service. To respond to the message, please log on to GitHub and use the URL above to go to the specific comment. To unsubscribe, e-mail: commits-unsubscr...@tvm.apache.org For queries about this service, please contact Infrastructure at: us...@infra.apache.org