Skip to content

readme : clarify MODEL_ENDPOINT usage#20941

Merged
angt merged 1 commit intoggml-org:masterfrom
angt:readme-clarify-model_endpoint-usage
Mar 24, 2026
Merged

readme : clarify MODEL_ENDPOINT usage#20941
angt merged 1 commit intoggml-org:masterfrom
angt:readme-clarify-model_endpoint-usage

Conversation

@angt
Copy link
Copy Markdown
Member

@angt angt commented Mar 24, 2026

Overview

ModelScope is not an HF-compatible endpoint.

Additional information

A dedicated -ms option could be added in the future.

Requirements

Signed-off-by: Adrien Gallouët <angt@huggingface.co>
@angt angt requested a review from ggerganov as a code owner March 24, 2026 09:09
@angt angt merged commit c9dc433 into ggml-org:master Mar 24, 2026
2 checks passed
@yingdachen
Copy link
Copy Markdown

Not sure why this change is made to the readme without any code change? With no code hange, it would appear that the usage

MODEL_ENDPOINT=https://www.modelscope.cn/ ./llama-cli -hf ggml-org/gemma-3-1b-it-GGUF

would still be valid.

There was disccusion before the implementation was oringally merged and the MODEL_ENDPOINT approach was suggested, are the communities now favoring option like

./llama-cli -ms ggml-org/gemma-3-1b-it-GGUF`

instead?

@angt @ggerganov

@angt
Copy link
Copy Markdown
Member Author

angt commented Mar 27, 2026

Hi @yingdachen,

Indeed, we updated the -hf option recently to share the same cache as the hf tool and do the discovery of the model and the mmproj locally, not relying on the API. So it is not possible anymore to mix different source under the same -hf flag:

$ MODEL_ENDPOINT=https://www.modelscope.cn/ build/bin/llama-cli -hf ggml-org/gemma-3-1b-it-GGUF
get_repo_commit: JSON error: [json.exception.parse_error.101] parse error at line 2, column 1: syntax error while parsing value - invalid literal; last read: '<U+000A><'
error: failed to download model from Hugging Face

I did a PR especially to clarify that as I was not aware of this usage and so I proposed a -ms for ModelScope.

@yingdachen
Copy link
Copy Markdown

Hi @yingdachen,

Indeed, we updated the -hf option recently to share the same cache as the hf tool and do the discovery of the model and the mmproj locally, not relying on the API. So it is not possible anymore to mix different source under the same -hf flag:

$ MODEL_ENDPOINT=https://www.modelscope.cn/ build/bin/llama-cli -hf ggml-org/gemma-3-1b-it-GGUF
get_repo_commit: JSON error: [json.exception.parse_error.101] parse error at line 2, column 1: syntax error while parsing value - invalid literal; last read: '<U+000A><'
error: failed to download model from Hugging Face

I did a PR especially to clarify that as I was not aware of this usage and so I proposed a -ms for ModelScope.

thanks for the clarification @angt , we will take a look and submit a PR for the -ms option later.

Seunghhon pushed a commit to Seunghhon/llama.cpp that referenced this pull request Apr 26, 2026
Signed-off-by: Adrien Gallouët <angt@huggingface.co>
rsenthilkumar6 pushed a commit to rsenthilkumar6/llama.cpp that referenced this pull request May 1, 2026
Signed-off-by: Adrien Gallouët <angt@huggingface.co>
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.

3 participants