deepseek code v2 inference downgrade after a few inference

### What is the issue? I am using ollama for deepseed coder model make fill-in-middel inference, and it was working fine initially. However, after more than 50 consecutive inferences, the model output suddenly became abnormal, producing nonsensical content. abnormal output like this. #generategenerategenerategenerategenerategenerategenerategenerategenerategenerategenerategenerategenerategenerategenerategenerategenerategenerategenerategenerategenerategenerategenerategenerategenerategenerategenerategenerategenerategenerategenerate In these calls, the call parameters remain consistent. ``` data = { "model": "deepseek-coder-v2:latest", "prompt": prompt, "raw": True, "stream": False, "keep_alive": -1, "seed": 40, "options": { "num_predict": num_p, "top_k": 1, "top_p": 0.0, "temperature": 0.0, "repeat_penalty": 1.1, } } ``` ### OS Linux ### GPU Nvidia ### CPU Intel ### Ollama version 0.1.48