-
Notifications
You must be signed in to change notification settings - Fork 10.5k
[AutoDiff] Crash when overriding _Differentiation
module
#59876
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
Comments
I found a workaround for the second crash. @differentiable(reverse)
func Tensor_stack(_ lhs: Tensor) -> Tensor {
var stack = Array<Tensor>()
stack.append(lhs)
return Tensor_init(stacking: stack)
} This uses |
* Initial support for release toolchains * Fix typo * Remove fatal error * Comment out x10_training_loop for Colab * Make `Mergeable.stack` differentiable, working around swiftlang/swift#59876 * Remove #available(macOS 9999, *) restriction * Enable RNN tests * Re-enable tensorflow#1162 * Allow Array replaceSubrange * Fit within 100 spaces * Type fix "partecipating" -> "participating" * Remove redundant same-type constraint * Remove deprecated Dataset API and reposition Zip2TensorGroup * Refactor tests * Refactor X10 tests * Fix failing tests on macOS + arm64 * Remove workarounds for swiftlang/swift#55703 (SR-13263) * Enable ops_test.swift * Optimize ops_test.swift * Optimize more tests * Rename gradient(at:in) to gradient(at:of:) * Rename argument label in withoutDerivative * Enable tests blocked by reflection crash * Workaround for swiftlang/swift#59135 on dev toolchains * Fix tests for Colab * Attempt to speed up tests on Colab * Revert changes to tests * Fix tests on Colab NVIDIA GPU * Revert specializations for NVIDIA GPUs
@philipturner Why did you close your issues? |
@philipturner I’m sorry to bother you, but it wasn’t just this one; there are at least 10 more. Was it an accident? Do you want me to reopen them? |
I felt like they should be closed since I am no longer involved with the Swift compiler. I have no intention to solve them, and there seems to be a low chance most will ever be touched again. Except for one thread with Anton, where I glanced over the fact that it was still actively investigated in January 2023. I was thinking in the mindset of issues in other repositories, where you generally close issues you no longer see as worth solving. Hence "closed as not planned". I forgot that for Swift, issues are an official bug tracker (like JIRA). Therefore closing them might make people mistake them as already being solved. If someone deems a specific bug worth investigating in the future, I can re-open them on a case by case basis. However, I do not wish to provide assistance to anyone investigating. |
I'll be glad to review these and see if they are still reproducible in top-of-tree Swift. If so, I can reopen them so that they can be tracked to be fixed as we work through a priority list of differentiable Swift issues. |
I don't know if we can fix this bug, but it should still be reported. In the current directory, make the following files:
Differentiable.swift
ArrayDifferentiation.swift
Run the following from the command line. Change
FRONTEND
to match your toolchain or apple/swift build products.Crash
There is another crash that's related, and must be worked around in S4TF by making
Tensor(stacking:alongAxis:)
not differentiable on release toolchains. Replace the contents of "ArrayDifferentiation.swift" with:Crash
Additional Context
This was originally mistaken for SR-14008.
The text was updated successfully, but these errors were encountered: