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 #581] How to use num_predict? #261

New Issue
Closed
opened 2026-04-12 09:47:22 -05:00 by GiteaMirror · 8 comments
GiteaMirror commented 2026-04-12 09:47:22 -05:00
Owner
Copy Link

Originally created by @jamesbraza on GitHub (Sep 23, 2023).
Original GitHub issue: https://github.com/ollama/ollama/issues/581

From https://github.com/jmorganca/ollama/issues/318#issuecomment-1710181439, I see num_predict exists, and am trying to figure out how to use it.

Where are the docs on parameters like this?

More specifically, I am trying to figure out how to specify num_predict (and similar parameters) to the Ollama server process and/or /generate API calls.

Originally created by @jamesbraza on GitHub (Sep 23, 2023). Original GitHub issue: https://github.com/ollama/ollama/issues/581 From https://github.com/jmorganca/ollama/issues/318#issuecomment-1710181439, I see `num_predict` exists, and am trying to figure out how to use it. Where are the docs on parameters like this? More specifically, I am trying to figure out how to specify `num_predict` (and similar parameters) to the Ollama server process and/or `/generate` API calls.
GiteaMirror commented 2026-04-12 09:47:23 -05:00
Author
Owner
Copy Link

@willowell commented on GitHub (Sep 25, 2023):

Hello @jamesbraza!

Had a look around - looks like the docs for these parameters are in llama.cpp. For instance, here's the doc for n_predict: https://github.com/ggerganov/llama.cpp/tree/master/examples/main#number-of-tokens-to-predict.

I found these types for the parameters https://github.com/jmorganca/ollama/blob/main/llm/llama.go#L378 - it looks like these are out-of-sync with the ones in https://github.com/jmorganca/ollama/blob/main/api/types.go#L161?

I don't see num_predict used outside types.go.

Does this help?

<!-- gh-comment-id:1733774990 --> @willowell commented on GitHub (Sep 25, 2023): Hello @jamesbraza! Had a look around - looks like the docs for these parameters are in llama.cpp. For instance, here's the doc for `n_predict`: https://github.com/ggerganov/llama.cpp/tree/master/examples/main#number-of-tokens-to-predict. I found these types for the parameters https://github.com/jmorganca/ollama/blob/main/llm/llama.go#L378 - it looks like these are out-of-sync with the ones in https://github.com/jmorganca/ollama/blob/main/api/types.go#L161? I don't see `num_predict` used outside `types.go`. Does this help?
GiteaMirror commented 2026-04-12 09:47:24 -05:00
Author
Owner
Copy Link

@BruceMacD commented on GitHub (Sep 25, 2023):

This parameter tells the LLM the maximum number of tokens it is allowed to generate.

It's not exposed in the CLI at the moment, but you can define it directly in the body of requests make to the API at the /generate endpoint. Here is an example of that, setting the num_predict to 1.

Request

curl --request POST \
     --url http://localhost:11434/api/generate \
     --header "Content-Type: application/json" \
     --data '{
         "prompt": "hi",
         "model": "llama2",
         "options": {
             "num_predict": 1
         }
     }'

Response Stream

{
    "model": "llama2",
    "created_at": "2023-09-25T14:32:13.093801Z",
    "response": " Hello",
    "done": false
}
{
    "model": "llama2",
    "created_at": "2023-09-25T14:32:13.095352Z",
    "done": true,
    "context": [29961,25580,29962,7251,518,29914,25580,29962,29871,15043],
    "total_duration": 383724416,
    "load_duration": 1833458,
    "prompt_eval_count": 5,
    "prompt_eval_duration": 373471000,
    "eval_count": 1
}
<!-- gh-comment-id:1733841969 --> @BruceMacD commented on GitHub (Sep 25, 2023): This parameter tells the LLM the maximum number of tokens it is allowed to generate. It's not exposed in the CLI at the moment, but you can define it directly in the body of requests make to the API at the `/generate` endpoint. Here is an example of that, setting the `num_predict` to 1. ## Request ``` curl --request POST \ --url http://localhost:11434/api/generate \ --header "Content-Type: application/json" \ --data '{ "prompt": "hi", "model": "llama2", "options": { "num_predict": 1 } }' ``` ## Response Stream ``` { "model": "llama2", "created_at": "2023-09-25T14:32:13.093801Z", "response": " Hello", "done": false } { "model": "llama2", "created_at": "2023-09-25T14:32:13.095352Z", "done": true, "context": [29961,25580,29962,7251,518,29914,25580,29962,29871,15043], "total_duration": 383724416, "load_duration": 1833458, "prompt_eval_count": 5, "prompt_eval_duration": 373471000, "eval_count": 1 } ```
GiteaMirror commented 2026-04-12 09:47:25 -05:00
Author
Owner
Copy Link

@jamesbraza commented on GitHub (Sep 25, 2023):

Thank you both @willowell and @BruceMacD! Is there a relevant portion of the Ollama docs for this somewhere? Otherwise I would gladly add this somewhere, if you could point me to the right place

<!-- gh-comment-id:1734379394 --> @jamesbraza commented on GitHub (Sep 25, 2023): Thank you both @willowell and @BruceMacD! Is there a relevant portion of the Ollama docs for this somewhere? Otherwise I would gladly add this somewhere, if you could point me to the right place
GiteaMirror commented 2026-04-12 09:47:27 -05:00
Author
Owner
Copy Link

@BruceMacD commented on GitHub (Sep 26, 2023):

The closest documentation would be this table of parameter options in the modelfile docs. It looks like it is missing a few options now though, num_predict isn't there:
https://github.com/jmorganca/ollama/blob/main/docs/modelfile.md#valid-parameters-and-values

<!-- gh-comment-id:1735643153 --> @BruceMacD commented on GitHub (Sep 26, 2023): The closest documentation would be this table of parameter options in the modelfile docs. It looks like it is missing a few options now though, num_predict isn't there: https://github.com/jmorganca/ollama/blob/main/docs/modelfile.md#valid-parameters-and-values
GiteaMirror commented 2026-04-12 09:47:27 -05:00
Author
Owner
Copy Link

@jamesbraza commented on GitHub (Sep 26, 2023):

As long as a search for num_predict has some matches in docs/, I will call it a win.

Looks like that table comes from Options, I will add a few entries to it shortly

<!-- gh-comment-id:1735831779 --> @jamesbraza commented on GitHub (Sep 26, 2023): As long as a search for `num_predict` has some matches in `docs/`, I will call it a win. Looks like that table comes from [`Options`](https://github.com/jmorganca/ollama/blob/v0.0.21/api/types.go#L161), I will add a few entries to it shortly
GiteaMirror commented 2026-04-12 09:47:28 -05:00
Author
Owner
Copy Link

@jamesbraza commented on GitHub (Sep 27, 2023):

PR opened, thanks all!

<!-- gh-comment-id:1736544212 --> @jamesbraza commented on GitHub (Sep 27, 2023): [PR opened](https://github.com/jmorganca/ollama/pull/614), thanks all!
GiteaMirror commented 2026-04-12 09:47:29 -05:00
Author
Owner
Copy Link

@vusiernestmthuk commented on GitHub (Nov 5, 2025):

Ind lunchtime and teatime for today

<!-- gh-comment-id:3489578619 --> @vusiernestmthuk commented on GitHub (Nov 5, 2025): Ind lunchtime and teatime for today
GiteaMirror commented 2026-04-12 09:47:30 -05:00
Author
Owner
Copy Link

@vusiernestmthuk commented on GitHub (Nov 5, 2025):

Lunchtime and teatime

<!-- gh-comment-id:3489579874 --> @vusiernestmthuk commented on GitHub (Nov 5, 2025): Lunchtime and teatime
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#261
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?