Update app.py

app.py

@@ -47,44 +47,11 @@ from tensorflow import keras
 from PIL import Image
 
 # Cell 1: Image Classification Model
-model1 = from_pretrained_keras("rocioadlc/EfficientNetV2L")
-
-# Define class labels
-class_labels = ['battery',
- 'biological',
- 'brown-glass',
- 'cardboard',
- 'clothes',
- 'green-glass',
- 'metal',
- 'paper',
- 'plastic',
- 'shoes',
- 'trash',
- 'white-glass']
-
-# Function to predict image label and score
-def predict_image(input):
-    # Resize the image to the size expected by the model
-    image = input.resize((244, 224))
-    # Convert the image to a NumPy array
-    image_array = tf.keras.preprocessing.image.img_to_array(image)
-    # Normalize the image
-    image_array /= 255.0
-    # Expand the dimensions to create a batch
-    image_array = tf.expand_dims(image_array, 0)
-    # Predict using the model
-    predictions = model1.predict(image_array)
-    
-    # Get the predicted class label
-    predicted_class_index = tf.argmax(predictions, axis=1).numpy()[0]
-    predicted_class_label = class_labels[predicted_class_index]
-    
-    # Get the confidence score of the predicted class
-    confidence_score = predictions[0][predicted_class_index]
-    
-    # Return predicted class label and confidence score
-    return {predicted_class_label: confidence_score}
+pipeline = pipeline(task="image-classification", model="julien-c/hotdog-not-hotdog")
+
+def predict_image(image):
+  predictions = pipeline(image)
+  return {p["label"]: p["score"] for p in predictions}
 
 
 image_gradio_app = gr.Interface(
@@ -95,100 +62,12 @@ image_gradio_app = gr.Interface(
     theme=theme
 )
 
-loader = WebBaseLoader(["https://www.epa.gov/recycle/frequent-questions-recycling", "https://www.whitehorsedc.gov.uk/vale-of-white-horse-district-council/recycling-rubbish-and-waste/lets-get-real-about-recycling/", "https://www.teimas.com/blog/13-preguntas-y-respuestas-sobre-la-ley-de-residuos-07-2022", "https://www.molok.com/es/blog/gestion-de-residuos-solidos-urbanos-rsu-10-dudas-comunes"])
-data=loader.load()
-# split documents
-text_splitter = RecursiveCharacterTextSplitter(
-    chunk_size=1024,
-    chunk_overlap=150,
-    length_function=len
-)
-docs = text_splitter.split_documents(data)
-# define embedding
-embeddings = HuggingFaceEmbeddings(model_name='thenlper/gte-small')
-# create vector database from data
-persist_directory = 'docs/chroma/'
-
-# Remove old database files if any
-shutil.rmtree(persist_directory, ignore_errors=True)
-vectordb = Chroma.from_documents(
-    documents=docs,
-    embedding=embeddings,
-    persist_directory=persist_directory
-)
-# define retriever
-retriever = vectordb.as_retriever(search_kwargs={"k": 2}, search_type="mmr")
-
-class FinalAnswer(BaseModel):
-    question: str = Field(description="the original question")
-    answer: str = Field(description="the extracted answer")
-
-# Assuming you have a parser for the FinalAnswer class
-parser = PydanticOutputParser(pydantic_object=FinalAnswer)
-
-template = """
-Your name is Greta and you are a recycling chatbot with the objective to anwer questions from user in English or Spanish /
-Use the following pieces of context to answer the question /
-If the question is English answer in English /
-If the question is Spanish answer in Spanish /
-Do not mention the word context when you answer a question /
-Answer the question fully and provide as much relevant detail as possible. Do not cut your response short /
-Context: {context}
-User: {question}
-{format_instructions}
-"""
-
-# Create the chat prompt templates
-sys_prompt = SystemMessagePromptTemplate.from_template(template)
-qa_prompt = ChatPromptTemplate(
-    messages=[
-        sys_prompt,
-        HumanMessagePromptTemplate.from_template("{question}")],
-    partial_variables={"format_instructions": parser.get_format_instructions()}
-)
-llm = HuggingFaceHub(
-    repo_id="mistralai/Mixtral-8x7B-Instruct-v0.1",
-    task="text-generation",
-    model_kwargs={
-        "max_new_tokens": 2000,
-        "top_k": 30,
-        "temperature": 0.1,
-        "repetition_penalty": 1.03
-    },
-)
-
-qa_chain = ConversationalRetrievalChain.from_llm(
-    llm = llm,
-    memory = ConversationBufferMemory(llm=llm, memory_key="chat_history", input_key='question', output_key='output'),
-    retriever = retriever,
-    verbose = True,
-    combine_docs_chain_kwargs={'prompt': qa_prompt},
-    get_chat_history = lambda h : h,
-    rephrase_question = False,
-    output_key = 'output',
-)
-
-def chat_interface(question,history):
-    result = qa_chain.invoke({'question': question})
-    output_string = result['output']
-
-    # Find the index of the last occurrence of "answer": in the string
-    answer_index = output_string.rfind('"answer":')
-
-    # Extract the substring starting from the "answer": index
-    answer_part = output_string[answer_index + len('"answer":'):].strip()
-
-    # Find the next occurrence of a double quote to get the start of the answer value
-    quote_index = answer_part.find('"')
-
-    # Extract the answer value between double quotes
-    answer_value = answer_part[quote_index + 1:answer_part.find('"', quote_index + 1)]
-    
-    return answer_value
+def echo(message, history):
+    return message
 
 
 chatbot_gradio_app = gr.ChatInterface(
-    fn=chat_interface,
+    fn=echo,
     title=custom_title
 )