-
Notifications
You must be signed in to change notification settings - Fork 48
New issue
Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.
By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.
Already on GitHub? Sign in to your account
Clarify intermediate values (activations) vs MLActivation vs operations #335
Comments
Having worked on several algorithms and having studied the implementation, I don't find the current usage of MLActivation sufficiently clear. I suggest we revert to the In addition, we need constructors for MLOperator and MLOperand (I have written them, but didn't make a PR). This would help make the spec more clear and also follow at least the CPU impl more closely. |
Thanks @huningxin. |
This is an old problem/discussion that has been addressed from multiple sides during the past year. Closing this for now. |
With consecutive recent changes I feel we should check and sort out the use of the following terms in the spec:
The latter are supposed to have functional semantics, so don't have state and don't change states of other objects. Therefore they are fine being represented by ops functions in MLGraphBuilder.
However,
(Possibly a V2 discussion, but would be nice to sort out in V1. The text definitely needs checking in V1.)
If there is a clear explanation, we just need to update the spec text, explainer etc.
The text was updated successfully, but these errors were encountered: