Skip to content

Imatrix input data should not be unescaped - #9543

Merged
ggerganov merged 1 commit into
ggml-org:masterfrom
CISC:imatrix-no-escape
Sep 19, 2024
Merged

Imatrix input data should not be unescaped#9543
ggerganov merged 1 commit into
ggml-org:masterfrom
CISC:imatrix-no-escape

Conversation

@CISC

@CISC CISC commented Sep 18, 2024

Copy link
Copy Markdown
Member

At some point prompt input unescaping was turned on by default, which is fine in itself, but imatrix is usually generated from prompt_file, which is fed into params.prompt and then unescaped, generating bad input data.

This will (and has) go(ne) unnoticed until you have sequences like \xdef in your input data, which will throw an "invalid character" error (only with BPE tokenizer).

@ggerganov
ggerganov merged commit eca0fab into ggml-org:master Sep 19, 2024
@slaren

slaren commented Sep 19, 2024

Copy link
Copy Markdown
Member

perplexity will probably need a similar fix.

@CISC
CISC deleted the imatrix-no-escape branch September 19, 2024 08:46
@CISC

CISC commented Sep 19, 2024

Copy link
Copy Markdown
Member Author

@slaren I'll make another PR for perplexity.

dsx1986 pushed a commit to dsx1986/llama.cpp that referenced this pull request Oct 29, 2024
arthw pushed a commit to arthw/llama.cpp that referenced this pull request Nov 15, 2024
arthw pushed a commit to arthw/llama.cpp that referenced this pull request Nov 18, 2024
Seunghhon pushed a commit to Seunghhon/llama.cpp that referenced this pull request Apr 26, 2026
ljubomirj pushed a commit to ljubomirj/llama.cpp that referenced this pull request May 6, 2026
my-other-github-account pushed a commit to my-other-github-account/llama.cpp that referenced this pull request May 15, 2026
my-other-github-account pushed a commit to my-other-github-account/llama.cpp that referenced this pull request May 15, 2026
phibya pushed a commit to ziee-ai/llama.cpp that referenced this pull request May 29, 2026
AlexiAlp pushed a commit to minghaop/llama.cpp that referenced this pull request Jun 2, 2026
AlexiAlp pushed a commit to minghaop/llama.cpp that referenced this pull request Jun 2, 2026
fukuro-kun pushed a commit to fukuro-kun/fukuro-llama-cpp-turboquant that referenced this pull request Jul 5, 2026
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

Projects

None yet

Development

Successfully merging this pull request may close these issues.

3 participants