Skip to content

adds padding, output padding, and depthwise cpu support with tests#3

Open
mmwillet wants to merge 1 commit intomasterfrom
conv-transposition-1d-cpu-patch
Open

adds padding, output padding, and depthwise cpu support with tests#3
mmwillet wants to merge 1 commit intomasterfrom
conv-transposition-1d-cpu-patch

Conversation

@mmwillet
Copy link
Copy Markdown
Owner

Per mmwillet/TTS.cpp#66, this PR cleans up and adds the TTS.cpp patches to 1d convolutional transposition in order to be upstreamed to GGML main with the underlying goal of aligning modern GGML with TTS.cpp requirements.

Comment on lines +4024 to +4026
// This does not represent a real limitation in convolutional transposition. Rather, for simplicity the
// implementation currently assumes that padding is never greater than stride.
GGML_ASSERT(p0 < s0);
Copy link
Copy Markdown
Owner Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

I probably should work to completely remove this constraint.

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

1 participant