trunk/9c77ff51b6b2a30dc0964289e54c55b3f1bcdf13: Fix functional accumulate_grad state modeling (#184082)
🔒
https://github.com
«Make Inductor's accumulate_grad op take the current grad explicitly and return the updated grad, so functionalization and compiled autograd no longer rely on hidden Tensor.grad reads.
Remove the custom op output alias an...»
Automatische Weiterleitung...
1.5s