mixtral-8x7b doesn't understand toolExecutionResponse and loops through tools

[FrancescoDiCat]

LocalAI version:
2.5.1-cublas-cuda12

Environment, CPU architecture, OS, and Version:
Ubuntu 22.04 with 2 RTX A5000 24Gb GPUs

Describe my bug
My problem is that this model mixtral-8x7b-instruct-v0.1.Q8_0.gguf seems to be not able to understand when to stop after tools response, so it loops through them without giving a final response. Even after correct tool response when this response is given to the model again it response continuosly with function_call.

Image Example of requests and model responses

[satyajitghana]

same here

[maxi1134]

Same here with "laser-dolphin-mixtral-2x7b-dpo.q6_k". Anyone has a solution?

[github-actions]

This issue is stale because it has been open 90 days with no activity. Remove stale label or comment or this will be closed in 5 days.