Update app.py

app.py

@@ -129,7 +129,7 @@ We currently only support one image at the start of the context! Please start a
         pixel_values = None 
         
     
-    generation_config = dict(max_new_tokens= 512, do_sample=False, num_beams = 3, repetition_penalty=2.0)
+    generation_config = dict(max_new_tokens= 512, do_sample=False, num_beams = 3, repetition_penalty=1.5)
     
     if len(history) == 0:
         if pixel_values is not None:
@@ -226,8 +226,8 @@ button.svelte-1lcyrx4[aria-label="user's message: a file of type image/jpeg, "]
 
 demo = gr.ChatInterface(
     fn=chat,
-    description="""Try [Vintern-3B-beta](https://huggingface.co/5CD-AI/Vintern-3B-beta) in this demo. Vintern-3B-beta consists of [InternViT-300M-448px](https://huggingface.co/OpenGVLab/InternViT-300M-448px), an MLP projector, and [Qwen2.5-3B-Instruct](https://huggingface.co/Qwen/Qwen2.5-3B-Instruct).
-Bias, Risks, and Limitations
+    description="""We introduce Vintern-1B-v3.5, the latest version in the Vintern series, offering significant improvements over v2 across all evaluation benchmarks. This model has been fine-tuned from InternVL-1B-2.5, which already good in Vietnamese 🇻🇳 tasks because it used Viet-ShareGPT-4o-Text-VQA data during its fine-tuning process by the InternVL 2.5 [1] team.
+To further enhance its performance in Vietnamese while maintaining good capabilities on existing English datasets, Vintern-1B-v3.5 has been fine-tuned using a vast amount of Vietnamese-specific data. This results in a model that is exceptionally powerful in text recognition, OCR, and understanding Vietnam-specific documents.
 The model might have biases because it learned from data that could be biased.
 Users should be cautious of these possible biases when using the model.""",
     examples=[{"text": "Mô tả hình ảnh.", "files":["./demo_3.jpg"]},