ciflow/inductor/195373: Support complex dropout by allocating a real-valued mask
- PyTorch: 280 events in the last 90 days
- PyTorch: 277th Release in the last 90 days
- Previous: earlier the same day · ciflow/torchtitan/195383: [INITIAL] Update
What happened
nn.functional.dropout raised NotImplementedError for complex inputs on CPU (and CUDA), while silently "working" on MPS. Root cause: _dropout_impl allocated the Bernoulli mask ( noise ) in the input's dtype and called bernoulli_ on it. For complex inputs that requires a complex bernoulli_ , which CPU and CUDA do not implement (both dispatch via AT_DISPATCH_ALL_TYPES_AND3 , which does not list complex). MPS only accep…
Summary assembled by rule from the sources below