How to infer Janus on multi-images?

[2018211801]

Thanks for your nice model, but I cannot infer on multi-images. Counld you provide the code? Or the model cannot infer on multi-images.

[xwgnick]

Thanks for your nice model, but I cannot infer on multi-images. Counld you provide the code? Or the model cannot infer on multi-images.

Use multiple <image_placeholder>，like：
conversation = [ { "role": "<|User|>", "content": f"<image_placeholder>\n<image_placeholder>\n{question}", "images": ["./test_images/test_1.jpg", "./test_images/test_2.jpg"], }, {"role": "<|Assistant|>", "content": ""}, ]