zhouzp610 opened a new issue, #14877: URL: https://github.com/apache/tvm/issues/14877
Each item in the features array in the auto_scheduler consists of some number of arrays of length 164. This number is not fixed and can be from 1 to 8. For each of this items (consisting of multiple features vectors) there is one target value.  But for the loss function it uses the sum of (f(x[i][1])+f(x[i][2])+…+f(x[i][len_i])-y[i])^2 for all i. Here f is our model, x[i][1], x[i][2], …, x[i][len_i] are vectors of features of object i, y[i] its target value of object i. So we sum the predictions for all feature vectors of the object and we want it to be close to the target value.  Why we have the sum of the predictions in the loss function, but not mean value of them? ### Triage Please refer to the list of label tags [here](https://github.com/apache/tvm/wiki/Issue-Triage-Labels) to find the relevant tags and add them below in a bullet format (example below). * needs-triage * tune:auto_scheduler -- 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]
