fix: minor fix

app.py

@@ -112,6 +112,11 @@ def generate_prompt_with_history(
             else current_prompt
         )
 
+        if current_prompt.count("<image_placeholder>") > 2:
+            for _ in range(len(conversation_copy.messages) - 2):
+                conversation_copy.messages.pop(0)
+            return conversation_copy
+        
         if torch.tensor(tokenizer.encode(current_prompt)).size(-1) <= max_length:
             return conversation_copy
 
@@ -389,7 +394,7 @@ def build_demo(MODELS):
                     )
                     max_length_tokens = gr.Slider(
                         minimum=0,
-                        maximum=4096,
+                        maximum=2048,
                         value=2048,
                         step=8,
                         interactive=True,
@@ -397,8 +402,8 @@ def build_demo(MODELS):
                     )
                     max_context_length_tokens = gr.Slider(
                         minimum=0,
-                        maximum=4096,
-                        value=4096,
+                        maximum=2048,
+                        value=2048,
                         step=128,
                         interactive=True,
                         label="Max History Tokens",

app_modules/presets.py

@@ -20,7 +20,17 @@
 # -*- coding:utf-8 -*-
 import gradio as gr
 
-title = """<h1 align="left" style="min-width:200px; margin-top:0;">Chat with DeepSeek-VL </h1>"""
+title = """
+<h1 align="left" style="min-width:200px; margin-top:0;">Chat with DeepSeek-VL</h1>
+<p>
+    <a href="https://arxiv.org/abs/2403.05525" target="_blank">Paper</a> |
+    <a href="https://github.com/deepseek-ai/DeepSeek-VL" target="_blank">GitHub</a>
+</p>
+<p style="color: red;">
+    Note: Due to the limited memory of the A10 GPU, the demo supports a maximum of two images and 2048 tokens.
+</p>
+"""
+
 description_top = """"""
 description = """"""
 CONCURRENT_COUNT = 10