[GH-ISSUE #159] A small question about multi images input for recognition #124

New issue

Open

opened 2026-10-26 21:12:35 +00:00 by bsakaguchi · 0 comments

bsakaguchi commented

2026-10-26 21:12:35 +00:00

Owner

Originally created by @sddthree on GitHub (Feb 9, 2025).
Original GitHub issue: https://github.com/deepseek-ai/Janus/issues/159

When I use JanusFlow for multi image recognition, I find that it cannot input multiple images at once. After inspection, it was found that the problem lies in Janus/janus/janusflow/models/processed-vlm.py file, the call function which is from lines 378 to 383, only processes one of the images when multiple images are input, and does not form multiple VLChatProcessorOutput for List [VLChatProcessorOutput] in the batchify function. The current solution is to generate multiple VLChatProcessorOutput using loops with multiple imgaes inputs and add them to the prepare list for batchify. We hope to optimize the code in the future to improve the functionality of multiple images and write it down.

Originally created by @sddthree on GitHub (Feb 9, 2025). Original GitHub issue: https://github.com/deepseek-ai/Janus/issues/159 When I use JanusFlow for multi image recognition, I find that it cannot input multiple images at once. After inspection, it was found that the problem lies in Janus/janus/janusflow/models/processed-vlm.py file, the call function which is from lines 378 to 383, only processes one of the images when multiple images are input, and does not form multiple VLChatProcessorOutput for List [VLChatProcessorOutput] in the batchify function. The current solution is to generate multiple VLChatProcessorOutput using loops with multiple imgaes inputs and add them to the prepare list for batchify. We hope to optimize the code in the future to improve the functionality of multiple images and write it down.

bsakaguchi referenced this issue

2026-10-26 21:14:14 +00:00

[PR #124] [CLOSED] streaming version of app_januspro #219