mulanxiaodingdang commented on issue #18048: URL: https://github.com/apache/tvm/issues/18048#issuecomment-2962209343
Hello, thank you for your response. The model we utilized is derived from a pre-trained encoder available on Hugging Face. We believe its architecture is sufficiently complex for our purposes. The compiled version of the original model has been made available via the link below. For compiling this ONNX model, we employed both the Relay and Relax compilers. When compiling with Relay at the O3 optimization level, we observed significant operator fusion. In contrast, no operator fusion occurred at the O0 level. Notably, for the identical ONNX model, compiling with the Relax compiler at both the O0 and O3 optimization levels resulted in identical .so files.[https://huggingface.co/sentence-transformers/multi-qa-mpnet-base-dot-v1/tree/main/onnx](url) -- 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: [email protected] For queries about this service, please contact Infrastructure at: [email protected]
