Video-LLaMA
Video-LLaMA copied to clipboard
Repeating itself
Hi,
Thanks for the code! For multiple videos we found that the model is quite often just repeating itself when answering. For instance, it might make descriptions like
"In this video, we see a black and white can of coffee on a wooden table. The can is black and white, and there is a black and white can of coffee next to it. The table is wooden, and there is a black and white can of coffee on it. The can is black and white, and there is a black and white can of coffee on the table. The can is black and white, and there is a black and white can of coffee on the table. The can is black and white, and there is a black and white can of coffee on the table. The can is black and white, and there is a black and white can of coffee on the table. The can is black and white, and there is a black and white can of coffee on the table. The can is black and white, and there is a black and white can of coffee on the table. The can is black and white, and there is a black and white can of coffee on the table. The can is black and white, and there is a black and white can of coffee on the table. The can is black and white, and there is a black and white can of coffee on the table. The can is black and white, and there is a black and white can of coffee on the table. The can is black and white, and there is a black and white can of coffee on the table. The can is black and white, and there is a black".
Has anyone experienced this before?