Update Space

app.py

@@ -1,6 +1,7 @@
 import gradio as gr
 from transformers import AutoTokenizer, AutoModelForCausalLM
 import torch
+import markdown
 
 model_name = "deepseek-ai/deepseek-coder-1.3b-base"
 
@@ -8,8 +9,17 @@ tokenizer = AutoTokenizer.from_pretrained(model_name, trust_remote_code=True)
 model = AutoModelForCausalLM.from_pretrained(model_name, trust_remote_code=True)
 model.eval()
 
-def respond(prompt, max_tokens, temperature, top_p):
-    inputs = tokenizer(prompt, return_tensors="pt")
+def respond(message, chat_history, max_tokens, temperature, top_p):
+    chat_history = chat_history or []
+    chat_history.append(("User", message))
+    
+    full_prompt = ""
+    for speaker, text in chat_history:
+        prefix = "User: " if speaker == "User" else "Assistant: "
+        full_prompt += prefix + text + "\n"
+    full_prompt += "Assistant: "
+    
+    inputs = tokenizer(full_prompt, return_tensors="pt")
     outputs = model.generate(
         **inputs,
         max_new_tokens=max_tokens,
@@ -18,17 +28,35 @@ def respond(prompt, max_tokens, temperature, top_p):
         do_sample=True,
         pad_token_id=tokenizer.eos_token_id,
     )
-    return tokenizer.decode(outputs[0], skip_special_tokens=True)[len(prompt):].strip()
+    reply = tokenizer.decode(outputs[0], skip_special_tokens=True)[len(full_prompt):].strip()
+    chat_history.append(("Assistant", reply))
+    
+    formatted_history = []
+    for i in range(0, len(chat_history), 2):
+        user_msg = chat_history[i][1] if i < len(chat_history) else ""
+        bot_msg = chat_history[i+1][1] if i+1 < len(chat_history) else ""
+        # Render assistant message as markdown
+        formatted_history.append([user_msg, gr.Markdown(bot_msg)])
+    
+    return formatted_history, ""
 
 with gr.Blocks() as demo:
-    gr.Markdown("# DeepSeek Coder without Login")
-    prompt = gr.Textbox(label="Enter your prompt", lines=5)
-    max_tokens = gr.Slider(1, 1024, value=512, step=1, label="Max Tokens")
-    temperature = gr.Slider(0.1, 1.0, value=0.7, step=0.05, label="Temperature")
-    top_p = gr.Slider(0.1, 1.0, value=0.9, step=0.05, label="Top-p")
-    btn = gr.Button("Generate")
-    output = gr.Textbox(label="Output", lines=15)
-    btn.click(respond, inputs=[prompt, max_tokens, temperature, top_p], outputs=output)
+    gr.Markdown("# DeepSeek Coder Chatbot")
+    
+    chatbot = gr.Chatbot(markdown=True)
+    with gr.Row():
+        user_input = gr.Textbox(show_label=False, placeholder="Enter your prompt here and press Enter")
+    with gr.Row():
+        max_tokens = gr.Slider(1, 1024, value=512, step=1, label="Max Tokens")
+        temperature = gr.Slider(0.1, 1.0, value=0.7, step=0.05, label="Temperature")
+        top_p = gr.Slider(0.1, 1.0, value=0.9, step=0.05, label="Top-p")
+    
+    def user_submit(text, history, max_tokens, temperature, top_p):
+        if not text.strip():
+            return history, ""
+        return respond(text, history, max_tokens, temperature, top_p)
+    
+    user_input.submit(user_submit, inputs=[user_input, chatbot, max_tokens, temperature, top_p], outputs=[chatbot, user_input])
 
 if __name__ == "__main__":
     demo.launch()