Add application file

app.py

@@ -3,6 +3,8 @@ from pydantic import BaseModel
 from llama_cpp import Llama
 from typing import Optional
 import uvicorn
+import os
+import huggingface_hub
 
 app = FastAPI(
 	title="OmniVLM API",
@@ -10,10 +12,17 @@ app = FastAPI(
 	version="1.0.0"
 )
 
-# Initialize the model
-llm = Llama.from_pretrained(
+# Download the model from Hugging Face Hub
+model_path = huggingface_hub.hf_hub_download(
 	repo_id="NexaAIDev/OmniVLM-968M",
-	filename="omnivision-text-optimized-llm-Q8_0.gguf",
+	filename="omnivision-text-optimized-llm-Q8_0.gguf"
+)
+
+# Initialize the model with the downloaded file
+llm = Llama(
+	model_path=model_path,
+	n_ctx=2048,  # Context window
+	n_threads=4   # Number of CPU threads to use
 )
 
 class GenerationRequest(BaseModel):

requirements.txt

@@ -1,4 +1,5 @@
 fastapi==0.104.1
 uvicorn==0.24.0
 pydantic==2.4.2
-llama-cpp-python==0.1.76
+llama-cpp-python==0.1.76
+huggingface-hub>=0.19.0