Upload folder using huggingface_hub

app_new.py

@@ -64,31 +64,48 @@ MODEL_OPTIONS = {
 }
 
 MODEL = None
+CURRENT_MODEL_KEY = None
 
 def get_model_key(model_type, model_size):
     return f"{model_type} {model_size}"
 
-def initialize_model(model_type="Apollo2", model_size="7B"):
-    global MODEL
+def initialize_model(model_type, model_size):
+    """Load the selected model dynamically."""
+    global MODEL, CURRENT_MODEL_KEY
     model_key = get_model_key(model_type, model_size)
+    
+    # Only reload the model if it's not already loaded
+    if CURRENT_MODEL_KEY == model_key and MODEL is not None:
+        print(f"Model {model_key} is already loaded.")
+        return
+
+    print(f"Initializing model: {model_key}")
     try:
-        print(f"Initializing model: {model_key}")
         selected_model = MODEL_OPTIONS[model_key]
         MODEL = load_model(
             directory=selected_model["directory"],
             model_name=selected_model["model_name"],
             model_url=selected_model["model_url"]
         )
+        CURRENT_MODEL_KEY = model_key
         print(f"Model initialized: {model_key}")
     except Exception as e:
         print(f"Failed to initialize model {model_key}: {e}")
         MODEL = None
 
-def user(message, history):
+# Functions for chat interactions
+def user(message, history, model_type, model_size):
+    """Handle user input and dynamically initialize the selected model."""
+    global MODEL
+
+    # Dynamically initialize the selected model
+    initialize_model(model_type, model_size)
+
     new_history = history + [[message, None]]
     return "", new_history
 
 def bot(history, top_p, top_k, temp):
+    """Generate a response from the bot based on chat history."""
     global MODEL
     if MODEL is None:
         raise RuntimeError("Model has not been initialized. Please select a model to load.")
@@ -96,18 +113,13 @@ def bot(history, top_p, top_k, temp):
     tokens = get_system_tokens(model)[:]
 
     for user_message, bot_message in history[:-1]:
-        message_tokens = get_message_tokens(model=model, role="user", content=user_message)
-        tokens.extend(message_tokens)
+        tokens.extend(get_message_tokens(model=model, role="user", content=user_message))
         if bot_message:
-            message_tokens = get_message_tokens(model=model, role="bot", content=bot_message)
-            tokens.extend(message_tokens)
+            tokens.extend(get_message_tokens(model=model, role="bot", content=bot_message))
 
     last_user_message = history[-1][0]
-    message_tokens = get_message_tokens(model=model, role="user", content=last_user_message)
-    tokens.extend(message_tokens)
-
-    role_tokens = model.tokenize("bot\n".encode("utf-8"), special=True)
-    tokens.extend(role_tokens)
+    tokens.extend(get_message_tokens(model=model, role="user", content=last_user_message))
+    tokens.extend(model.tokenize("bot\n".encode("utf-8"), special=True))
 
     generator = model.generate(tokens, top_k=top_k, top_p=top_p, temp=temp)
 
@@ -120,16 +132,22 @@ def bot(history, top_p, top_k, temp):
         history[-1][1] = partial_text
         yield history
 
-with gr.Blocks(
-    theme=gr.themes.Monochrome(),
-    analytics_enabled=False,
-) as demo:
+def clear_chat():
+    """Clear the chat history."""
+    return []
+
+def stop_generation():
+    """Placeholder to stop generation."""
+    print("Generation stopped.")  # Implement stop logic if supported
+    return None
+
+# Gradio UI
+with gr.Blocks(theme=gr.themes.Monochrome(), analytics_enabled=False) as demo:
     favicon = '<img src="https://huggingface.co/FreedomIntelligence/Apollo2-7B/resolve/main/assets/apollo_medium_final.png" width="148px" style="display: inline">'
     gr.Markdown(
         f"""# {favicon} Apollo GGUF Playground
 
         This is a demo of multilingual medical model series **[Apollo](https://huggingface.co/FreedomIntelligence/Apollo-7B-GGUF)**, GGUF version. [Apollo1](https://arxiv.org/abs/2403.03640) covers 6 languages. [Apollo2](https://arxiv.org/abs/2410.10626) covers 50 languages.
-
         """
     )
     with gr.Row():
@@ -142,7 +160,6 @@ with gr.Blocks(
                 elem_id="send-message-box"
             )
         with gr.Column(scale=1):
-            # 将 model_type 和 model_size 包含在同一个 gr.Row 中
             with gr.Row(equal_height=False):
                 model_type = gr.Dropdown(
                     choices=["Apollo", "Apollo2"],
@@ -158,7 +175,6 @@ with gr.Blocks(
                     interactive=True,
                     elem_id="model-size-dropdown",
                 )
-            #gr.Markdown("### Generation Parameters")
             top_p = gr.Slider(
                 minimum=0.0,
                 maximum=1.0,
@@ -188,16 +204,22 @@ with gr.Blocks(
                 stop = gr.Button("Stop", elem_id="stop-btn")
                 clear = gr.Button("Clear", elem_id="clear-btn")
 
+    # Event bindings
+    submit_event = msg.submit(
+        fn=user,
+        inputs=[msg, chatbot, model_type, model_size],
+        outputs=[msg, chatbot],
+        queue=False,
+    ).success(
+        fn=bot,
+        inputs=[chatbot, top_p, top_k, temp],
+        outputs=chatbot,
+        queue=True,
+    )
 
-    def update_model(model_type, model_size):
-        initialize_model(model_type, model_size)
-
-    model_type.change(update_model, [model_type, model_size], None)
-    model_size.change(update_model, [model_type, model_size], None)
-
-    msg.submit(
+    submit_click_event = submit.click(
         fn=user,
-        inputs=[msg, chatbot],
+        inputs=[msg, chatbot, model_type, model_size],
         outputs=[msg, chatbot],
         queue=False,
     ).success(
@@ -207,28 +229,40 @@ with gr.Blocks(
         queue=True,
     )
 
+
+    stop.click(
+        fn=stop_generation,
+        inputs=None,
+        outputs=None,
+        cancels=[submit_event, submit_click_event],
+        queue=False,
+    )
+
+    clear.click(fn=clear_chat, inputs=None, outputs=chatbot, queue=False)
+
 demo.queue(max_size=128)
 demo.css = """
 footer {display: none !important;}
 #send-message-box {width: 100%;}
 #send-btn, #stop-btn, #clear-btn {
-    display: inline-block; /* 强制内联块 */
-    width: 30%; /* 设置按钮宽度为父容器的 30% */
-    margin-right: 2px; /* 按钮之间增加间距 */
-    text-align: center; /* 按钮内容居中 */
+    display: inline-block;
+    width: 30%;
+    margin-right: 2px;
+    text-align: center;
 }
-
 .gr-row {
-    display: flex !important; /* 强制使用 flex 布局 */
-    flex-direction: row !important; /* 水平排列 */
-    justify-content: space-between; /* 组件之间的间距调整 */
-    align-items: center; /* 垂直居中对齐 */
-    flex-wrap: nowrap; /* 禁止按钮换行 */
+    display: flex !important;
+    flex-direction: row !important;
+    justify-content: space-between;
+    align-items: center;
+    flex-wrap: nowrap;
 }
 """
 
+# Initialize
+
 
 # Initialize the default model at startup
-initialize_model("Apollo2", "7B")
+#initialize_model("Apollo2", "7B")
 
 demo.launch(show_error=True, share=True)