github-starred/ollama
2
0
Fork 0
mirror of https://github.com/ollama/ollama.git synced 2026-11-12 03:46:14 -06:00
Code Issues 16.9k Packages Projects Releases 122 Wiki Activity

[PR #15610] [CLOSED] Handle large Gemma4 thinking prompt defaults #41102

New Issue
Closed
opened 2026-04-23 01:50:00 -05:00 by GiteaMirror · 0 comments
GiteaMirror commented 2026-04-23 01:50:00 -05:00
Owner
Copy Link
Copy Source

📋 Pull Request Information

Original PR: https://github.com/ollama/ollama/pull/15610
Author: @dhiltgen
Created: 4/15/2026
Status: ❌ Closed

Base: main ← Head: gemma4-thinking

📝 Commits (1)

  • d28bfe4 Handle large Gemma4 thinking prompt defaults

📊 Changes

9 files changed (+171 additions, -27 deletions)

View changed files

📝 model/renderers/gemma4.go (+5 -1)
📝 model/renderers/gemma4_reference_test.go (+17 -0)
📝 model/renderers/renderer.go (+11 -0)
📝 server/create.go (+2 -0)
📝 server/prompt.go (+15 -1)
📝 server/prompt_test.go (+59 -0)
📝 types/model/config.go (+10 -8)
📝 x/create/client/create.go (+34 -17)
📝 x/create/client/create_test.go (+18 -0)

📄 Description

The ConfigV2 allow runtime selection of the renderers. If we want to bake different renderer names into the created models we can avoid that change.

For large Gemma4 models, when the request does not explicitly ask for thinking, the renderer appends this after the generation prompt:

<|channel>thought
<channel|>

That primes an empty thought channel so the model transitions into the response channel instead of emitting raw control fragments like thought, <channel|>, etc. into user-visible text.

🔄 This issue represents a GitHub Pull Request. It cannot be merged through Gitea due to API limitations.

## 📋 Pull Request Information **Original PR:** https://github.com/ollama/ollama/pull/15610 **Author:** [@dhiltgen](https://github.com/dhiltgen) **Created:** 4/15/2026 **Status:** ❌ Closed **Base:** `main` ← **Head:** `gemma4-thinking` --- ### 📝 Commits (1) - [`d28bfe4`](https://github.com/ollama/ollama/commit/d28bfe46b06522bc9a25679f0f30ce73d31e0908) Handle large Gemma4 thinking prompt defaults ### 📊 Changes **9 files changed** (+171 additions, -27 deletions) <details> <summary>View changed files</summary> 📝 `model/renderers/gemma4.go` (+5 -1) 📝 `model/renderers/gemma4_reference_test.go` (+17 -0) 📝 `model/renderers/renderer.go` (+11 -0) 📝 `server/create.go` (+2 -0) 📝 `server/prompt.go` (+15 -1) 📝 `server/prompt_test.go` (+59 -0) 📝 `types/model/config.go` (+10 -8) 📝 `x/create/client/create.go` (+34 -17) 📝 `x/create/client/create_test.go` (+18 -0) </details> ### 📄 Description The ConfigV2 allow runtime selection of the renderers. If we want to bake different renderer names into the created models we can avoid that change. For large Gemma4 models, when the request does not explicitly ask for thinking, the renderer appends this after the generation prompt: ``` <|channel>thought <channel|> ``` That primes an empty thought channel so the model transitions into the response channel instead of emitting raw control fragments like thought, <channel|>, etc. into user-visible text. --- <sub>🔄 This issue represents a GitHub Pull Request. It cannot be merged through Gitea due to API limitations.</sub>
GiteaMirror added the pull-request label 2026-04-23 01:50:00 -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 labels pull-request
Milestone
No items
No Milestone
Projects
Clear projects
No projects
Assignees
Clear assignees
GiteaMirror ninjasurge (Kaehlen Rog)
No Assignees
1 Participants
Notifications
Due Date
No due date set.
Dependencies

No dependencies set.

Reference: github-starred/ollama#41102