[GH-ISSUE #161] Question about the training strategy #126

Open
opened 2026-10-26 21:12:36 +00:00 by bsakaguchi · 0 comments
Owner

Originally created by @Hertz19 on GitHub (Feb 10, 2025).
Original GitHub issue: https://github.com/deepseek-ai/Janus/issues/161

In the technical report of Janus-Pro, you mentioned 'even with the LLM parameters fixed, the model could effectively model pixel dependence and generate reasonable images based on category names'. I tried to reproduce with other LLMs but failed. Could please you show some case of the generated images after Stage I training?

Originally created by @Hertz19 on GitHub (Feb 10, 2025). Original GitHub issue: https://github.com/deepseek-ai/Janus/issues/161 In the technical report of Janus-Pro, you mentioned 'even with the LLM parameters fixed, the model could effectively model pixel dependence and generate reasonable images based on category names'. I tried to reproduce with other LLMs but failed. Could please you show some case of the generated images after Stage I training?
Sign in to join this conversation.
No labels
pull-request
No milestone
No project
No assignees
1 participant
Notifications
Due date
The due date is invalid or out of range. Please use the format "yyyy-mm-dd".

No due date set.

Dependencies

No dependencies set.

Reference
deepseek-ai/Janus#126
No description provided.