[GH-ISSUE #71] confusion about learning rate scaling law #50

Open
opened 2026-10-26 21:02:08 +00:00 by bsakaguchi · 0 comments
Owner

Originally created by @ds22058 on GitHub (Jul 8, 2025).
Original GitHub issue: https://github.com/deepseek-ai/DeepSeek-LLM/issues/71

Your study on learning rate is very helpful to me, but I still have some questions.

  1. In your learning rate scaling law experiment, are all trainings reduced to 31.6% of the initial learning rate at 80% and 10% of the initial learning rate at 90% as mentioned before?

  2. Does the learning rate used in the scaling law refer to the initial learning rate?

Looking forward to your reply.

Originally created by @ds22058 on GitHub (Jul 8, 2025). Original GitHub issue: https://github.com/deepseek-ai/DeepSeek-LLM/issues/71 Your study on learning rate is very helpful to me, but I still have some questions. 1. In your learning rate scaling law experiment, are all trainings reduced to 31.6% of the initial learning rate at 80% and 10% of the initial learning rate at 90% as mentioned before? 2. Does the learning rate used in the scaling law refer to the initial learning rate? Looking forward to your reply.
Sign in to join this conversation.
No labels
bug
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/DeepSeek-LLM#50
No description provided.