zotanika opened a new issue #6576: URL: https://github.com/apache/incubator-tvm/issues/6576
Dear TVM team, I'm struggling to figure out the reason why the models hosted at [https://www.tensorflow.org/lite/guide/hosted_models](url) experience huge accuracy drop after converting with relay.frontend.from_tflite(). For instance, mobilenet_v1_1.0_224 claimed to hit top1 72%, top5 91% accuracy becomes top1 61%, top5 87%. First, I would like to reproduce the result accuracy given at [https://tvm.apache.org/docs/dev/benchmark.html](url) for resnet18 and reuse these benchmark infrastructures to debug the problem efficiently, but I cannot find the source codes matching to schema v0.1 log format described in the page. Do I miss something, or is there any other reference to do this job? ---------------------------------------------------------------- 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. For queries about this service, please contact Infrastructure at: [email protected]
