Upload app.py

app.py

@@ -0,0 +1,72 @@
+# -*- coding: utf-8 -*-
+"""gradio.ipynb
+
+Automatically generated by Colab.
+
+Original file is located at
+    https://colab.research.google.com/drive/1goHcmXF0Gc4_X9PN-zecV77j9KeI6Dmn
+"""
+
+# !pip install -q -U gradio
+# !pip install -q -U torch transformers accelerate einops
+# !pip install -q peft
+
+
+
+
+
+import gradio as gr
+
+import torch
+from transformers import (
+    AutoTokenizer,
+    AutoModelForCausalLM,
+    TextIteratorStreamer,
+    pipeline,
+)
+
+# The huggingface model id for Microsoft's phi-2 model
+# Download and load model and tokenizer
+tokenizer = AutoTokenizer.from_pretrained("mudogruer/mixtral-7x8b-SciQ", trust_remote_code=True)
+
+#Download safetensor of adapter of fine-tune Phi-2 model
+from peft import PeftModel, PeftConfig
+
+
+config = PeftConfig.from_pretrained("mudogruer/mixtral-7x8b-SciQ")
+base_model = AutoModelForCausalLM.from_pretrained("mistralai/Mixtral-8x7B-Instruct-v0.1")
+model = PeftModel.from_pretrained(base_model, "mudogruer/mixtral-7x8b-SciQ")
+
+# Text generation pipeline
+phi2 = pipeline(
+    "text-generation",
+    tokenizer=tokenizer,
+    model=model,
+    pad_token_id=tokenizer.eos_token_id,
+    eos_token_id=tokenizer.eos_token_id,
+    device_map="cpu",
+)
+
+examples = [["Which organelle carries out the synthesis and packaging of digestive enzymes?"],
+            ["What is the change in speed of a moving object per unit time?"] ,
+            ["What is the formula of carbon tetrafluoride?"]]
+
+def generate(message, max_new_tokens):
+    instruction = "You are a helpful assistant to 'User'. You do not respond as 'User' or pretend to be 'User'. You only respond once as 'Assistant'."
+    final_prompt = f"Instruction: {instruction}\nUser: {message}\nOutput:"
+
+    # Generate text synchronously
+    response = phi2(final_prompt, max_new_tokens=max_new_tokens)
+    generated_text = response[0]['generated_text']
+
+    # Process to extract the last assistant's response
+    # Assuming the last line after 'Output:' is the response
+    last_response = generated_text.split('Output:')[-1].strip()
+    return last_response
+
+# Update the Gradio interface setup
+with gr.Blocks() as demo:
+    gr.Markdown("""### Mixtral-7x8b Scientific Question Chatbot(Fine-tuned from SciQ dataset)""")
+    tokens_slider = gr.Slider(8, 128, value=21, label="Maximum new tokens")
+    chatbot = gr.Interface(fn=generate, inputs=["text", tokens_slider], outputs="text", examples = examples)
+    demo.launch(share=True)