Update app.py

app.py

@@ -13,16 +13,23 @@ HF_TOKEN = os.environ.get("HF_TOKEN", None)
 MODEL_ID = os.environ.get("MODEL_ID")
 MODEL_NAME = MODEL_ID.split("/")[-1]
 
-TITLE = "<h1><center>VL-Chatbox</center></h1>"
+TITLE = "<h1>VL-Chatbox</h1>"
 
-DESCRIPTION = f'<h3><center>MODEL: <a href="https://hf.co/{MODEL_ID}">{MODEL_NAME}</a></center></h3>'
+DESCRIPTION = f'<p>A SPACE FOR VLM MODELS</p><br><h3><center>MODEL NOW: <a href="https://hf.co/{MODEL_ID}">{MODEL_NAME}</a></center></h3>'
 
 CSS = """
 .duplicate-button {
-  margin: auto !important;
-  color: white !important;
-  background: black !important;
-  border-radius: 100vh !important;
+    margin: auto !important;
+    color: white !important;
+    background: black !important;
+    border-radius: 100vh !important;
+}
+h1 {
+    text-align: center;
+    display: block;
+}
+p {
+    text-align: center;
 }
 """
 
@@ -37,7 +44,7 @@ model.eval()
 
 
 @spaces.GPU()
-def stream_chat(message, history: list, temperature: float, max_length: int):
+def stream_chat(message, history: list, temperature: float, max_length: int, top_p: float, top_k: int, penalty: float):
     print(f'message is - {message}')
     print(f'history is - {history}')
     conversation = []
@@ -46,8 +53,9 @@ def stream_chat(message, history: list, temperature: float, max_length: int):
         conversation.append({"role": "user", "image": image, "content": message['text']})
     else:
         if len(history) == 0:
-            raise gr.Error("Please upload an image first.")
+            #raise gr.Error("Please upload an image first.")
             image = None
+            conversation.append({"role": "user", "content": message['text']})
         else:
             image = Image.open(history[0][0][0])
             for prompt, answer in history:
@@ -65,9 +73,11 @@ def stream_chat(message, history: list, temperature: float, max_length: int):
         max_length=max_length,
         streamer=streamer,
         do_sample=True,
-        top_k=1,
+        top_p=top_p,
+        top_k=top_k,
         temperature=temperature,
-        repetition_penalty=1.2,
+        repetition_penalty=penalty,
+        eos_token_id=[151329, 151336, 151338],
     )
     gen_kwargs = {**input_ids, **generate_kwargs}
 
@@ -91,9 +101,9 @@ chat_input = gr.MultimodalTextbox(
 
 )
 EXAMPLES = [
-        [{"text": "Describe it in great detailed.", "files": ["./laptop.jpg"]}],
-        [{"text": "Describe it in great detailed.", "files": ["./hotel.jpg"]}],
-        [{"text": "Describe it in great detailed.", "files": ["./spacecat.png"]}]
+        [{"text": "Describe it in detailed", "files": ["./laptop.jpg"]}],
+        [{"text": "Where it is?", "files": ["./hotel.jpg"]}],
+        [{"text": "Is it real?", "files": ["./spacecat.png"]}]
 ]
 
 with gr.Blocks(css=CSS) as demo:
@@ -118,12 +128,37 @@ with gr.Blocks(css=CSS) as demo:
             ),
             gr.Slider(
                 minimum=128,
-                maximum=4096,
+                maximum=8192,
                 step=1,
                 value=1024,
                 label="Max Length",
                 render=False,
             ),
+            with gr.Row():
+                gr.Slider(
+                    minimum=0.0,
+                    maximum=1.0,
+                    step=0.1,
+                    value=1.0,
+                    label="top_p",
+                    render=False,
+                ),
+                gr.Slider(
+                    minimum=1,
+                    maximum=20,
+                    step=1,
+                    value=10,
+                    label="top_k",
+                    render=False,
+                ),
+            gr.Slider(
+                minimum=0.0,
+                maximum=2.0,
+                step=0.1,
+                value=1.0,
+                label="Repetition penalty",
+                render=False,
+            ),
         ],
     ),
     gr.Examples(EXAMPLES,[chat_input])