Update app.py

app.py

@@ -75,14 +75,16 @@ def main():
                     # 'threads': int(os.cpu_count() / 2)
                     # }
                         
-                    # llm = CTransformers(
-                    # # model = "TheBloke/Mistral-7B-Instruct-v0.2-GGUF",
-                    # model= "TheBloke/phi-2-GGUF",
+                    llm = CTransformers(
+                    # model = "TheBloke/Mistral-7B-Instruct-v0.2-GGUF",
+                    model= "second-state/stablelm-2-zephyr-1.6b-GGU",
+                    model_file = "stablelm-2-zephyr-1_6b-Q4_0.gguf"
                     # model_type="mistral",
-                    # max_new_tokens = 1048,
-                    # temperature = 0.3
-                    # )
-                    llm = AutoModelForCausalLM.from_pretrained("second-state/stablelm-2-zephyr-1.6b-GGUF", model_type="stablelm-2-zephyr-1_6b-Q4_0.gguf")
+                    max_new_tokens = 1048,
+                    temperature = 0.3,
+                    lib="avx2", # for CPU
+                    )
+                    # llm = AutoModelForCausalLM.from_pretrained("second-state/stablelm-2-zephyr-1.6b-GGUF", model_type="stablelm-2-zephyr-1_6b-Q4_0.gguf")
                         
                     print("LLM Initialized...")