Update app.py

app.py

@@ -1,10 +1,40 @@
 import gradio as gr
-from huggingface_hub import InferenceClient
+from transformers import AutoModelForCausalLM, AutoTokenizer, pipeline
+from peft import PeftModel, PeftConfig
+import torch
 
-"""
-For more information on `huggingface_hub` Inference API support, please check the docs: https://huggingface.co/docs/huggingface_hub/v0.22.2/en/guides/inference
-"""
-client = InferenceClient("HuggingFaceH4/zephyr-7b-beta")
+base_model = "meta-llama/Llama-3.2-3B-Instruct"
+adapter_model = "nickjelicic/Llama-3.2-3B-Instruct-Collectivist"
+
+SYSTEM_PROMPT = """
+You are playing the role of a university student, who is conducting an assignment together with another student. The assignment should be conducted together, as a collaboration. Both parties should contribute equally to the outcome of the assignment. In your communication, you are playing the role of a person with collectivist cultural background. Collectivism should be incorporated in the communication style, as well as the content of the assignment. The definitions of collectivism are provided below:
+
+Definitions
+
+* Collectivism is a philosophy that emphasizes the connection between the individual and the community. Its overriding philosophy is based on the belief that a person's social identity and personality are largely molded by community relationships. People in collectivistic cultures are closely connected and promote harmony and unity within society as a whole. (Source: Wikipedia)
+* Collectivism focuses on increasing overall group functioning as a way to create more overall happiness. (Source: Trompenaars)
+* In collectivist cultures, the needs, values, and goals of the in-group take precedence over the needs, values, and goals of the individual. (Triandis, 1995).
+* People in collectivist cultures tend to apply different value standards to ingroups and outgroups. (Gudykunst, 1997).
+* Collectivist communication is characterized by more frequent use of ‘we’ form.
+* Decisions are often referred back for verification. People ideally achieve in groups which assume joint responsibility. Vacations in organized groups or with extended family.
+* When doing business in a collectivist organization, negotiators can only agree tentatively and may withdraw an undertaking after consulting with superiors. The aim is to build lasting relationships.
+* Managers in collectivist organizations seek to integrate personality with authority within the group. Give attention to esprit de corps, morale and cohesiveness.
+* Have low job turnover and mobility. Extol the whole group and avoid showing favoritism. 
+""".strip()
+
+
+model = AutoModelForCausalLM.from_pretrained(
+    base_model, 
+    device_map='auto',
+    torch_dtype=torch.bfloat16)
+
+model = PeftModel.from_pretrained(model, adapter_model)
+tokenizer = AutoTokenizer.from_pretrained(base_model)
+
+pipe = pipeline(
+    'text-generation',
+    model=model,
+    tokenizer=tokenizer)
 
 
 def respond(
@@ -12,53 +42,25 @@ def respond(
     history: list[tuple[str, str]],
     system_message,
     max_tokens,
-    temperature,
-    top_p,
+    temperature
 ):
-    messages = [{"role": "system", "content": system_message}]
-
-    for val in history:
-        if val[0]:
-            messages.append({"role": "user", "content": val[0]})
-        if val[1]:
-            messages.append({"role": "assistant", "content": val[1]})
-
+    messages = [{"role": "system", "content": SYSTEM_PROMPT}]
     messages.append({"role": "user", "content": message})
+    
+    generation_config = {
+        'max_new_tokens':2048,
+        'temperature':0.7
+    }
+    res = pipe(messages,**generation_config)
+    return res[0]['generated_text'][-1]['content']
 
-    response = ""
-
-    for message in client.chat_completion(
-        messages,
-        max_tokens=max_tokens,
-        stream=True,
-        temperature=temperature,
-        top_p=top_p,
-    ):
-        token = message.choices[0].delta.content
 
-        response += token
-        yield response
-
-
-"""
-For information on how to customize the ChatInterface, peruse the gradio docs: https://www.gradio.app/docs/chatinterface
-"""
+chatbot = gr.Chatbot(show_copy_all_button=True)
 demo = gr.ChatInterface(
     respond,
-    additional_inputs=[
-        gr.Textbox(value="You are a friendly Chatbot.", label="System message"),
-        gr.Slider(minimum=1, maximum=2048, value=512, step=1, label="Max new tokens"),
-        gr.Slider(minimum=0.1, maximum=4.0, value=0.7, step=0.1, label="Temperature"),
-        gr.Slider(
-            minimum=0.1,
-            maximum=1.0,
-            value=0.95,
-            step=0.05,
-            label="Top-p (nucleus sampling)",
-        ),
-    ],
+    type='messages',
+    chatbot=chatbot
 )
 
-
 if __name__ == "__main__":
     demo.launch()