Adding files

chat_logic.py

@@ -0,0 +1,33 @@
+from model import tokenizer, model
+
+HISTORY = []  # Initialize an empty history list
+
+def chat(message, history):
+    """
+    Handles user input, generates a response using the model, and updates the chat history.
+    """
+    # Combine history with the current message
+    conversation = "\n".join(history) + f"\nUser: {message}\nAssistant:"
+
+    # Tokenize and generate response
+    inputs = tokenizer(conversation, return_tensors="pt", truncation=True, max_length=1024).to("cuda")
+    outputs = model.generate(inputs.input_ids, max_length=1024, temperature=0.7, do_sample=True)
+    reply = tokenizer.decode(outputs[:, inputs.input_ids.shape[-1]:][0], skip_special_tokens=True)
+
+    # Update history
+    update_history(message, reply)
+    return reply
+
+def update_history(message, reply):
+    """
+    Update the global history with the latest message and reply.
+    """
+    global HISTORY
+    HISTORY.append(f"User: {message}")
+    HISTORY.append(f"Assistant: {reply}")
+
+def get_history():
+    """
+    Retrieve the chat history as a string.
+    """
+    return "\n".join(HISTORY)

main.py

@@ -0,0 +1,23 @@
+import gradio as gr
+from chat_logic import chat, get_history
+
+def launch_gradio_ui():
+    with gr.Blocks() as gui:
+        gr.Markdown("## Chat With Llama 3.1-8B")
+
+        with gr.Row():
+            with gr.Column(scale=3):
+                chat_interface = gr.ChatInterface(fn=chat)
+
+            with gr.Column(scale=1):
+                gr.Markdown("### Message History")
+                history_display = gr.Textbox(label="Chat History", lines=27, interactive=False)
+                refresh_button = gr.Button("Refresh History")
+
+                # Update history display when the button is clicked
+                refresh_button.click(get_history, [], history_display)
+
+    gui.launch(share=True)
+
+if __name__ == "__main__":
+    launch_gradio_ui()

model.py

@@ -0,0 +1,25 @@
+# from transformers import AutoTokenizer, AutoModelForCausalLM
+
+# def load_model(model_name="meta-llama/Llama-3.1-8B"):
+#     """
+#     Load the Hugging Face Llama model and tokenizer.
+#     """
+#     tokenizer = AutoTokenizer.from_pretrained(model_name)
+#     model = AutoModelForCausalLM.from_pretrained(model_name, device_map="auto", torch_dtype="auto")
+#     return tokenizer, model
+
+# # Initialize model and tokenizer
+# tokenizer, model = load_model()
+
+from transformers import pipeline
+
+# Replace with your Hugging Face API token
+api_token = "your_huggingface_api_token"
+
+# Load the model using the API
+generator = pipeline(
+    "text-generation",
+    model="meta-llama/Llama-3.1-8B",
+    use_auth_token=api_token
+)
+

utils.py

@@ -0,0 +1,6 @@
+def truncate_conversation(history, max_tokens=1024):
+    """
+    Truncate the conversation history to fit within the token limit.
+    """
+    truncated_history = history[-max_tokens:]
+    return truncated_history