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 #14580] structured output addon #9453

New Issue
Open
opened 2026-04-12 22:22:53 -05:00 by GiteaMirror · 0 comments
GiteaMirror commented 2026-04-12 22:22:53 -05:00
Owner
Copy Link

Originally created by @Abdulrahman392011 on GitHub (Mar 3, 2026).
Original GitHub issue: https://github.com/ollama/ollama/issues/14580

i noticed that when using structured output, the model generate the name of the class which takes a lot of time. for example:

if you say the object in the class is called (( 'confidence percentage that the user will like this' : int )) the model will reply with manually generating each single token in 'confidence percentage that the user will like this' and then generate the actual token that matter which the confidence percentage itself.

you try to play around prompt engineering to reduce the amount of word used in the object name but ....

why not just copy and paste the object name and then carry the generation only for the value of the class.
sure there's a way to do that from low level prospective. it will help speed up the generation significantly.
keep in mind that structured output is usually used repeatedly and having a small change like this can make the workflow a lot faster, not to mention the energy consumption.

Originally created by @Abdulrahman392011 on GitHub (Mar 3, 2026). Original GitHub issue: https://github.com/ollama/ollama/issues/14580 i noticed that when using structured output, the model generate the name of the class which takes a lot of time. for example: if you say the object in the class is called (( 'confidence percentage that the user will like this' : int )) the model will reply with manually generating each single token in 'confidence percentage that the user will like this' and then generate the actual token that matter which the confidence percentage itself. you try to play around prompt engineering to reduce the amount of word used in the object name but ....<sad> why not just copy and paste the object name and then carry the generation only for the value of the class. sure there's a way to do that from low level prospective. it will help speed up the generation significantly. keep in mind that structured output is usually used repeatedly and having a small change like this can make the workflow a lot faster, not to mention the energy consumption.
GiteaMirror added the feature request label 2026-04-12 22:22:53 -05:00
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 feature request
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#9453
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?