Update proompt_template

app.py

@@ -1,13 +1,13 @@
 """Run codes."""
 # pylint: disable=line-too-long, broad-exception-caught, invalid-name, missing-function-docstring, too-many-instance-attributes, missing-class-docstring
 # ruff: noqa: E501
+import gc
 import os
 import platform
 import random
 import time
 from dataclasses import asdict, dataclass, field
 from pathlib import Path
-from textwrap import dedent
 
 # from types import SimpleNamespace
 import gradio as gr
@@ -22,29 +22,12 @@ from examples_list import examples_list
 url = "https://huggingface.co/TheBloke/WizardLM-1.0-Uncensored-Llama2-13B-GGML/blob/main/wizardlm-1.0-uncensored-llama2-13b.ggmlv3.q4_K_S.bin"  # 7.37G, Max RAM required 9.87 GB
 
 LLM = None
-
-if "forindo" in platform.node():  # deploy 70b model locally
-    # url = "https://huggingface.co/TheBloke/llama-2-70b-Guanaco-QLoRA-GGML/blob/main/llama-2-70b-guanaco-qlora.ggmlv3.q3_K_S.bin"  # 29.7G
-    # model_loc = "/home/mu2018/github/langchain-llama-2-70b-guanaco-qlora-ggml/models/llama-2-70b-guanaco-qlora.ggmlv3.q3_K_S.bin"
-    _ = """
-    url = "https://huggingface.co/TheBloke/StableBeluga2-70B-GGML/blob/main/stablebeluga2-70b.ggmlv3.q3_K_S.bin"
-    try:
-        model_loc, file_size = dl_hf_model(url)
-        logger.info(f"done load llm {model_loc=} {file_size=}G")
-    except Exception as exc_:
-        logger.error(exc_)
-        raise SystemExit(1) from exc_
-    # """
-    model_loc = "models/stablebeluga2-70b.ggmlv3.q3_K_S.bin"
-    assert Path(model_loc).exists(), f"Make sure {model_loc=} exists."
-else:
-    try:
-        logger.debug(f" dl {url}")
-        model_loc, file_size = dl_hf_model(url)
-        logger.info(f"done load llm {model_loc=} {file_size=}G")
-    except Exception as exc_:
-        logger.error(exc_)
-        raise SystemExit(1) from exc_
+gc.collect()
+try:
+    model_loc, file_size = dl_hf_model(url)
+except Exception as exc_:
+    logger.error(exc_)
+    raise SystemExit(1) from exc_
 
 # raise SystemExit(0)
 
@@ -58,21 +41,14 @@ human_prefix = "### Human"
 ai_prefix = "### Assistant"
 stop_list = [f"{human_prefix}:"]
 
-if "beluga" in model_loc.lower():
-    prompt_template = dedent(
-        """
-    ### System:
-    You are Stable Beluga, an AI that follows instructions extremely well. Help as much as you can.
-    Let's think step by step.
-
-    ### User: {question}
-
-    ### Assistant:
-    """
-    ).lstrip()
-    human_prefix = "### User"
-    ai_prefix = "### Assistant"
-    stop_list = [f"{human_prefix}:"]
+# Prompt template: WizardLM-Vicuna
+prompt_template = """You are a helpful AI assistant.
+
+USER: {question}
+ASSISTANT: """
+human_prefix = "USER"
+ai_prefix = "ASSISTANT"
+stop_list = [f"{human_prefix}:"]
 
 _ = psutil.cpu_count(logical=False) - 1
 cpu_count: int = int(_) if _ else 1

examples_list.py

@@ -9,7 +9,7 @@ examples_list = [
         "If it takes 10 hours to dry 10 clothes,  assuming all the clothes are hung together at the same time for drying , then how long will it take to dry a cloth?"
     ],
     [
-        "If it takes 10 hours to dry 10 clothes,  assuming all the clothes are hung together at the same time for drying , then how long will it take to dry a cloth? Think step by step."
+        "If it takes 10 hours to dry 10 clothes, assuming all the clothes are hung together at the same time for drying and there is only space to hang 10 clothes, then how long will it take to dry 23 clothes? Think step by step."
     ],
     ["is infinity + 1 bigger than infinity?"],
     ["Explain the plot of Cinderella in a sentence."],