mirror of
https://github.com/deepseek-ai/DeepSeek-R1.git
synced 2026-11-05 00:47:42 +00:00
[GH-ISSUE #36] Overthinking on FSM question #31
Labels
No milestone
No project
No assignees
1 participant
Notifications
Due date
No due date set.
Dependencies
No dependencies set.
Reference
deepseek-ai/DeepSeek-R1#31
Loading…
Add table
Add a link
Reference in a new issue
No description provided.
Delete branch "%!s()"
Deleting a branch is permanent. Although the deleted branch may continue to exist for a short time before it actually gets removed, it CANNOT be undone in most cases. Continue?
Originally created by @Northeus on GitHub (Jan 21, 2025).
Original GitHub issue: https://github.com/deepseek-ai/DeepSeek-R1/issues/36
Hi, when prompting 70b model, it seems that it overthinks too much on a problem concerning sequence detection using FSM. The model generated over 400 hundred lines long response, which seems to contains too much overthinking and revisiting similar steps again and again, so I had to stop the execution).
The prompt was:
create FSM detecting sequence 01001Are there any tips / guidlines, how can I mitigate such overthinking of the model?
@Northeus commented on GitHub (Jan 21, 2025):
When using online version with
DeepThinkthan there is a significantly less "thinking", is this problem relevant only to the smaller models? Could perhaps someone check, what is the output on the larger models on Ollama?@qryptiq-ai commented on GitHub (Jan 22, 2025):
Relatively new to a lot of this but pretty much everything I try results in it looping for 10+ minutes and the only way to stop it is to Terminate the Terminal window on my M1.
llama-cli
--hf-repo "unsloth/DeepSeek-R1-Distill-Llama-8B-GGUF"
--hf-file DeepSeek-R1-Distill-Llama-8B-F16.gguf
-p "In less than 5000 words, please explain in high level terms what the xxx function in the source code at the url yyy is doing. I will now provide the source code for that function. "
--conversation
It mostly keeps finding things to talk about, query and then provide detailed responses for.
@qryptiq-ai commented on GitHub (Jan 27, 2025):
Switched to ollama and deepseek-r1:14b and it's working fine on an M1 now.
@github-actions[bot] commented on GitHub (Feb 27, 2025):
This issue has been automatically marked as stale because it has not had recent activity. It will be closed if no further activity occurs. If you believe this issue is still relevant, please leave a comment to keep it open. Thank you for your contributions!
@github-actions[bot] commented on GitHub (Mar 17, 2025):
false