[GH-ISSUE #41] What tasks in training for image-text interleaved data? #28

Closed
opened 2026-10-26 21:08:58 +00:00 by bsakaguchi · 1 comment
Owner

Originally created by @BlueBlueFF on GitHub (Mar 31, 2024).
Original GitHub issue: https://github.com/deepseek-ai/DeepSeek-VL/issues/41

Originally created by @BlueBlueFF on GitHub (Mar 31, 2024). Original GitHub issue: https://github.com/deepseek-ai/DeepSeek-VL/issues/41
Author
Owner

@StevenLiuWen commented on GitHub (Apr 1, 2024):

@BlueBlueFF, we use interleaved image-text data to improve the in-context learning and support the inputs of multiple images with interleaved image-text format for the VL models. Same with other image-text data formats like image caption data, we apply the next token prediction loss only on the text part.

<!-- gh-comment-id:2029097800 --> @StevenLiuWen commented on GitHub (Apr 1, 2024): @BlueBlueFF, we use interleaved image-text data to improve the in-context learning and support the inputs of multiple images with interleaved image-text format for the VL models. Same with other image-text data formats like image caption data, we apply the next token prediction loss only on the text part.
Sign in to join this conversation.
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/DeepSeek-VL#28
No description provided.