Provided model name is ignored when using TGI

[HenFo]

Before submitting your bug report

• I believe this is a bug. I'll try to join the Continue Discord for questions
• I'm not able to find an open issue that reports the same bug
• I've seen the troubleshooting guide on the Continue Docs

Relevant environment info

- OS: Windows_NT x64 10.0.26100
- Continue version: 1.1.19
- IDE version: vscode 1.98.2
- Model: qwen2.5 coder 3b
- config:
  
name: Local Assistant
version: 1.0.0
schema: v1
models: 
  - name: Qwen2.5-coder 3b
    provider: huggingface-tgi
    model: qwen2.5-coder:3b
    apiBase: <url>
    roles:
      - autocomplete
context:
  - provider: code
  - provider: docs
  - provider: diff
  - provider: terminal
  - provider: problems
  - provider: folder
  - provider: codebase

Description

I host the model using huggingface tgi on a server under a specific name which did not include the "coder" in qwen.
I noticed that the wrong fim template was used, although I specified the model name in my config. While debugging into the extension, I noticed that the model name was overwritten with the abstract name I used to host the model.
Further, I noticed that this does not happen in the sandbox environment, but only when I open a new folder in the extension development host vscode window. My user config is as the same as the debug config.

I guess this could be called a feature, but from my perspective it is unexpected behavior.
Maybe a flag like "infereModelName" could be added, but by default the provided model name should be used.

To reproduce

1. host a local model using huggingface-tgi with a custom name, i.e. modelId.
2. setup the config as provided using the url to your tgi instance.
3. launch continue in debug mode and log the lowerCaseModel in getTemplateForModel in AutocompleteTemplate.ts.
4. Trigger a suggestion in the sandbox example and the provided model name will be logged.
5. Change the folder to a different project or whatever and trigger a new suggestion. The modelId as provided by tgi should be logged.

Log output