AI stops generating text

[Milucide]

Description

I use llamasharp in kombination with KernelMemory.
When I send a question to the model it generates a response whithout any error messages but the response stops mid sentence.
I use
Embeddingmodel: "multilingual-e5-base-q8_0.gguf"
Model: "em_german_leo_mistral.Q4_K_M.gguf"
and the following code:
InferenceParams infParams = new()
{
//MaxTokens = setMaxTokens,
MaxTokens = -1,
TokensKeep = 100,
SamplingPipeline = new DefaultSamplingPipeline
{
Temperature = 0.7f,
TopP = 0.9f,
TopK = 50,
RepeatPenalty = 1.08f

                }, 
            };

            LLamaSharpConfig llamasharpconfig = new(modelPath) { 
                DefaultInferenceParams = infParams, 
                ContextSize = 4096,
            };

         
            memory = new KernelMemoryBuilder()
             .WithLLamaSharpTextEmbeddingGeneration(new LLamaSharpConfig(modelPathEmbedding) { 
                 DefaultInferenceParams = infParams, 
                 ContextSize = 4096,
             })
             .WithLLamaSharpTextGeneration(llamasharpconfig)
             .WithPostgresMemoryDb(postgresConfig)
             .WithSimpleFileStorage(@".\database\")
             .Build(new KernelMemoryBuilderBuildOptions { AllowMixingVolatileAndPersistentData = false });

What am I missing? I already tried adding other parameters and changing the parameters with no success.

[m0nsky]

Did you try setting MaxTokens and increasing the ContextSize? By the way, the model you're using (em_german_leo_mistral) is very old and uses a custom prompt format. If you need German text, I recommend using Gemma 2 9B or Gemma 3 12B.