FIM return escaping symbols

[Semihal]

Before submitting your bug report

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

Relevant environment info

- OS: MacOS 14.5
- Continue: v0.8.42
- IDE: VSCode 1.90.2
- Model: vLLM (CodeQwen1.5-7B) -> LiteLLM
- config.json:
  
{
  "tabAutocompleteModel": {
    "title": "Autocomplete model",
    "provider": "openai",
    "apiBase": "<endpoint>",
    "apiKey": "<token>",
    "model": "qwen",
    "useLegacyCompletionsEndpoint": true,
    "requestOptions": {
      "verifySsl": false
    }
  },
  "tabAutocompleteOptions":{
    "useSuffix": false,
    "useCache": false,
    "template": "<fim_prefix>{{ prefix }}<fim_suffix>{{ suffix }}<fim_middle> "
  },
  "allowAnonymousTelemetry": false
}

Description

I had broken autocomplete:

The same model with the CodeGPT plugin for JetBrainns is working properly :(

To reproduce

No response

Log output

No response