Support loading model into pipeline from local filesystem

[Jeadie]

Motivation

• Support loading models from local filesystem to make it easier for productionising using mistral.rs.

Changes

• Change trait Loader. From load_model
  • load_model_from_hf
  • load_model_from_path
• Both of these can use the same logic once HF files have been downloaded locally.

[EricLBuehler]

Thank you for adding this! I think it's a great addition to the Rust API and just clears up the code a bit. I left a few minor comments regarding docs and one small question.

[EricLBuehler]

@Jeadie, I think there are some formatting/linting issues. After those are fixed, I think this should be ready to merge.

[github-actions]

Code Metrics Report

  ===============================================================================
 Language            Files        Lines         Code     Comments       Blanks
===============================================================================
 Dockerfile              1           34           25            0            9
 Happy                   1          442          369            0           73
 JSON                    5            9            9            0            0
 Python                 21          741          622           21           98
 TOML                   16          419          378            1           40
-------------------------------------------------------------------------------
 Jupyter Notebooks       1            0            0            0            0
 |- Markdown             1           60           30           22            8
 |- Python               1           96           87            1            8
 (Total)                            156          117           23           16
-------------------------------------------------------------------------------
 Markdown               16         1026            0          758          268
 |- BASH                 6          205          192            0           13
 |- Python               6          121          110            0           11
 |- Rust                 3          185          172            9            4
 (Total)                           1537          474          767          296
-------------------------------------------------------------------------------
 Rust                   81        26428        24327          334         1767
 |- Markdown            38          359            0          354            5
 (Total)                          26787        24327          688         1772
===============================================================================
 Total                 143        29099        25730         1114         2255
===============================================================================
  

[Jeadie]

Fixed and ran both cargo fmt --all --check and cargo clippy --workspace --tests --examples -- -D warnings locally.

[EricLBuehler]

Thank you!

[polarathene]

@Jeadie Is the local filesystem support only a few days old? Can you explain if HF API is mandatory?

Active issue: #326

• It's much simpler to run a local GGUF model via llama-cpp.
• The expected parameters (redundant?) can also be a tad confusing. This GGUF presumably needs the additional files sourced from here, but there is no tokenizer.json? From what I've read all the relevant metadata should already be available in the GGUF file itself, and llama-cpp happily runs with only the GGUF file provided.
• Without a HF token provided (--token-source none), the HF API is still hit but encounters an "Unauthorized" response (401), while only 404 responses have an exception which triggers a panic.

Perhaps there is some benefit with local filesystem going through HF API? Otherwise it'd probably be better to avoid that, or at least offer a way to opt-out if possible.