github-starred/ollama
2
0
Fork 0
mirror of https://github.com/ollama/ollama.git synced 2026-05-07 00:22:43 -05:00
Code Issues 14.5k Packages Projects Releases 120 Wiki Activity

[GH-ISSUE #2792] Subsequent generation requests hang after successful generation request with num_predict: 0 #48198

New Issue
Closed
opened 2026-04-28 07:08:14 -05:00 by GiteaMirror · 1 comment
GiteaMirror commented 2026-04-28 07:08:14 -05:00
Owner
Copy Link

Originally created by @stanier on GitHub (Feb 27, 2024).
Original GitHub issue: https://github.com/ollama/ollama/issues/2792

If you submit a generation request with num_predict: 0, the request will be handled successfully but all subsequent generation requests will hang indefinitely regardless of their num_predict values.

Below is an example of how to reproduce this behavior:

curl http://localhost:11434/api/generate -d '{
    "model": "vicuna:13b-16k",
    "template":"Hello world!",
    "stream":false,
    "num_predict":0
}'

Any requests following this will not return until the daemon has been restarted.

Also to note is that the daemon might be in an unresponsive state following this request, I've had to kill it each time it's happened to me so far, but haven't tested this against typical behavior for signals sent mid-generation or otherwise.

I think it also affects raw mode, and neither num_keep or num_ctx seem to be related.

Originally created by @stanier on GitHub (Feb 27, 2024). Original GitHub issue: https://github.com/ollama/ollama/issues/2792 If you submit a generation request with `num_predict: 0`, the request will be handled successfully but all subsequent generation requests will hang indefinitely regardless of their `num_predict` values. Below is an example of how to reproduce this behavior: ``` curl http://localhost:11434/api/generate -d '{ "model": "vicuna:13b-16k", "template":"Hello world!", "stream":false, "num_predict":0 }' ``` Any requests following this will not return until the daemon has been restarted. Also to note is that the daemon might be in an unresponsive state following this request, I've had to `kill` it each time it's happened to me so far, but haven't tested this against typical behavior for signals sent mid-generation or otherwise. I think it also affects `raw` mode, and neither `num_keep` or `num_ctx` seem to be related.
GiteaMirror commented 2026-04-28 07:08:17 -05:00
Author
Owner
Copy Link

@jmorganca commented on GitHub (May 10, 2024):

This should be fixed now. However, setting template might mean the model isn't using it's correct prompt template which could lead to infinitely long generations (which will stop automatically after ~10k tokens currently, and less in the future)

<!-- gh-comment-id:2103676275 --> @jmorganca commented on GitHub (May 10, 2024): This should be fixed now. However, setting `template` might mean the model isn't using it's correct prompt template which could lead to infinitely long generations (which will stop automatically after ~10k tokens currently, and less in the future)
Sign in to join this conversation.
No Branch/Tag Specified
Branches Tags
Labels
Clear labels
amd
api
app
bug
build
cli
cloud
compatibility
context-length
create
docker
documentation
embeddings
feature request
feedback wanted
good first issue
gpt-oss
gpu
harmony
help wanted
image
install
intel
js
launch
linux
macos
memory
mlx
model
needs more info
networking
nvidia
ollama.com
performance
pull-request
Mirrored from GitHub Pull Request
 python
question
registry
rendering
thinking
tools
top
vulkan
windows
wsl
No Label
Milestone
No items
No Milestone
Projects
Clear projects
No project
Assignees
Clear assignees
GiteaMirror ninjasurge
No Assignees
1 Participants
Notifications
Due Date
No due date set.
Dependencies

No dependencies set.

Reference: github-starred/ollama#48198
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?