Feature/add -ms option for modelscope

[yrk111222]

Overview

This PR introduces ModelScope integration for model downloading and loading.
Related PR: #20941
#21985

Key changes:

Added new cli arguments: -ms (ModelScope repo ID)
Added corresponding environment variables support for seamless configuration.
Enabled community model downloading via MODEL_ENDPOINT (defaults to https://modelscope.cn/), allowing flexible usage with or without explicit endpoint specification.

Usage examples:

# 1. Download via ModelScope ID (uses default or MODEL_ENDPOINT)
./build/bin/llama-cli -ms Qwen/Qwen3-0.6B-GGUF:Q8_0 -p "hello"
or
MODEL_ENDPOINT=https://modelscope.cn/ ./build/bin/llama-cli -ms Qwen/Qwen3-0.6B-GGUF:Q8_0 -p "hello"

# 2. Specify file and token explicitly (or set 'MS_TOKEN' env)
./build/bin/llama-cli -ms <repo> -hff <file> -hft <token> -p "hello"

#3. For vision models, the mmproj usage with `-ms` is identical to `-hf`.

Additional information

Requirements

• I have read and agree with the contributing guidelines
• AI usage disclosure: NO

[ggml-gh-bot]

Hi @yrk111222, thanks for your contribution!

Per our contribution guidelines, the automated PR checker found the following issue(s) that need your attention:

• Multiple open PRs from a new contributor: We limit new contributors (those without a previously merged PR) to 1 open PR at a time. You currently have 2 open PRs.

---

Please note that maintainers reserve the right to make final decisions on PRs. If you believe there is a mistake, please comment below.

[yrk111222]

Hi @yrk111222, thanks for your contribution!

Per our contribution guidelines, the automated PR checker found the following issue(s) that need your attention:

• Multiple open PRs from a new contributor: We limit new contributors (those without a previously merged PR) to 1 open PR at a time. You currently have 2 open PRs.

Please note that maintainers reserve the right to make final decisions on PRs. If you believe there is a mistake, please comment below.

I have closed the previous PR #21985 to comply with the single-PR rule for new contributors.

[yrk111222]

Hi @ngxson @angt , here is a new PR adding ModelScope -ms support. Please take a look when time permits. Thanks!🙏

[angt]

Hi @yrk111222, thanks for the update.
However, the PR feels particularly wrong because it duplicates a lot of code that shouldn’t be 👀

[yrk111222]

Hi @angt — thanks for the earlier review. I reworked this PR to remove duplication by switching to common_download_file_single and keeping only the ModelScope-specific bits (dropped the custom progress bar/download loop, extracted shared helpers).
Could you please take another look when you have a moment?
If anything else comes up, feel free to ping me anytime — thanks!😊

[yrk111222]

Hi @angt @ngxson , hope you’re having a great day! 😊
Just a gentle follow-up — I’ve refactored the PR to remove the duplicated code as you suggested.
Whenever you have a moment, could you kindly take another look?
Please also let me know if there’s anything else that needs improvement or if you have any other concerns — I’d be happy to adjust further.
Thanks so much for your time and guidance!

[angt]

Hey @yrk111222 thanks for the update, there is still a duplication of the download code 👀

[Leonard-Li777]

Expect the feature to be merged and released immediately

[yrk111222]

Hi @angt, thanks for your feedback. Since your last review, I've further reduced duplication.
Could you take another look when you have time? Thanks!🙏

[yrk111222]

Hi @angt. Friendly ping — could you take a look when you get a chance? Thanks! 😊

[yingdachen]

this has been pending for a while, could you please follow up and take a look when you got a chance? @angt @Leonard-Li777

thanks a bunch!

[angt]

Hey!

We know this feature is requested by users, so we really appreciate the effort. But rn, we are rethinking our overall cache system and how users interact with it. Because of this upcoming redesign, several related PRs are currently on hold, and I apologize for the delay this causes.

At the same time, this PR currently mixes a few different responsibilities. While we acknowledge that the existing codebase is already a bit messy, we want to avoid adding more technical debt before the redesign. All the logic for downloading, quantization selection, and filtering should stay in download.cpp for now. To keep things future-proof, ms-cache.cpp needs to follow the pattern of hf-cache.cpp more closely.

[yingdachen]

Hey!

We know this feature is requested by users, so we really appreciate the effort. But rn, we are rethinking our overall cache system and how users interact with it. Because of this upcoming redesign, several related PRs are currently on hold, and I apologize for the delay this causes.

At the same time, this PR currently mixes a few different responsibilities. While we acknowledge that the existing codebase is already a bit messy, we want to avoid adding more technical debt before the redesign. All the logic for downloading, quantization selection, and filtering should stay in download.cpp for now. To keep things future-proof, ms-cache.cpp needs to follow the pattern of hf-cache.cpp more closely.

thanks @angt , we will be working on a revision according to your suggestions.

[yrk111222]

Hello @angt
Thanks for the your detailed feedback! Based on your guidance about following the hf-cache.cpp pattern more closely and keeping download/selection logic in download.cpp, I've restructured the implementation and opened a fresh PR #24716 to keep the history clean.

Closing this one in favor of #24716. Happy to address any further feedback there!