work on llama.cpp huge loading time

src/summarization.py

@@ -3,6 +3,11 @@ from llama_cpp import Llama
 from utils import available_gguf_llms, s2tw_converter
 import time
 from functools import lru_cache
+import multiprocessing
+
+# Detect logical cores (vCPUs available to the container)
+num_vcpus = multiprocessing.cpu_count()
+print(f"Detected vCPUs: {num_vcpus}")
 
 @lru_cache(maxsize=1)
 def get_model(gguf_repo_id, gguf_filename):
@@ -11,13 +16,15 @@ def get_model(gguf_repo_id, gguf_filename):
         filename=gguf_filename,
         verbose=False,
         n_ctx=32768,
-        n_threads=4,
+        n_threads=num_vcpus,
         repeat_penalty=1.2,
     )
 
 def summarize_transcript(transcript, selected_gguf_model, prompt_input):
     repo_id, filename = available_gguf_llms[selected_gguf_model]
+    t0 = time.time()
     llm = get_model(repo_id, filename)
+    print(f"Model loading time: {time.time() - t0:.1f} sec")
     full_summary = []
     is_1st_token = True
     t1 = time.time()