Spaces:
Runtime error
Runtime error
File size: 2,106 Bytes
1896ffa 355a772 1896ffa 355a772 1896ffa 355a772 |
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 |
import gradio as gr
import os
import keras_nlp
# Set Kaggle API credentials (if needed)
# Set Kaggle API credentials
os.environ["KAGGLE_USERNAME"] = "rogerkorantenng"
os.environ["KAGGLE_KEY"] = "9a33b6e88bcb6058b1281d777fa6808d"
# Set the JAX backend and configure memory
os.environ["KERAS_BACKEND"] = "tensorflow"
os.environ["XLA_PYTHON_CLIENT_MEM_FRACTION"] = "1.00"
# Load the GemmaCausalLM model
gemma_lm = keras_nlp.models.GemmaCausalLM.from_preset("gemma_2b_en")
# Load LoRA weights if you have them
LoRA_weights_path = "model.lora.h5"
gemma_lm.backbone.enable_lora(rank=4) # Enable LoRA with rank 4
gemma_lm.preprocessor.sequence_length = 512 # Limit sequence length
gemma_lm.backbone.load_lora_weights(LoRA_weights_path) # Load LoRA weights
# Define the response generation function
def generate_response(message, history):
# Create a prompt template
template = "Instruction:\n{instruction}\n\nResponse:\n{response}"
# Format the history and the current message into the prompt
formatted_history = ""
for user_msg, bot_msg in history:
formatted_history += template.format(instruction=user_msg, response=bot_msg)
# Add the latest message from the user
prompt = template.format(instruction=message, response="")
# Combine history with the latest prompt
final_prompt = formatted_history + prompt
# Generate response from the model
response = gemma_lm.generate(final_prompt, max_length=256)
# Extract and return the generated response text
return response[0] # Adjust this if your model's output structure differs
# Create the Gradio chat interface
interface = gr.ChatInterface(
fn=generate_response, # Function that generates responses
chatbot=gr.Chatbot(height=300), # Chatbot UI component
textbox=gr.Textbox(placeholder="You can ask me anything", container=False, scale=7),
title="Local Model Chat Bot",
retry_btn=None, # Disable retry button
undo_btn="Delete Previous", # Enable undo button
clear_btn="Clear" # Enable clear button
)
# Launch the Gradio app
interface.launch(share=True)
|