fix(Llama): increase window context to 1024

app.py

@@ -9,7 +9,7 @@ from llama_cpp import Llama
 from huggingface_hub import hf_hub_download
 
 hf_hub_download(repo_id="TheBloke/dolphin-2_6-phi-2-GGUF", filename="dolphin-2_6-phi-2.Q5_K_S.gguf", local_dir=".")
-llm = Llama(model_path="./dolphin-2_6-phi-2.Q5_K_S.gguf", n_ctx=512, last_n_tokens_size=256, n_threads=4, n_gpu_layers=0)
+llm = Llama(model_path="./dolphin-2_6-phi-2.Q5_K_S.gguf", n_ctx=1024)
 
 ins = '''<|im_start|>user
 {question}<|im_end|>