With this PR, we'll still try to bind tensors to logical domains. However, tensor sizes are "unsharded" before binding. The core change isn't complicated, but I had to fix quite some dependencies, making this PR large.
Most tests are passing. There are a few failing corner cases that are arguably problems in the tests.
Fixes #3282
With this PR, we'll still try to bind tensors to logical domains. However, tensor sizes are "unsharded" before binding. The core change isn't complicated, but I had to fix quite some dependencies, making this PR large.
Most tests are passing. There are a few failing corner cases that are arguably problems in the tests.