github-starred/ollama
2
0
Fork 0
mirror of https://github.com/ollama/ollama.git synced 2026-05-23 22:21:55 -05:00
Code Issues 16.9k Packages Projects Releases 122 Wiki Activity

[PR #9776] [MERGED] fix: gemma3 quantization #38930

New Issue
Closed
opened 2026-04-22 23:35:02 -05:00 by GiteaMirror · 0 comments
GiteaMirror commented 2026-04-22 23:35:02 -05:00
Owner
Copy Link

📋 Pull Request Information

Original PR: https://github.com/ollama/ollama/pull/9776
Author: @pdevine
Created: 3/14/2025
Status: ✅ Merged
Merged: 3/15/2025
Merged by: @pdevine

Base: main ← Head: pdevine/gemma3-quant

📝 Commits (1)

📊 Changes

5 files changed (+149 additions, -0 deletions)

View changed files

📝 llama/llama.cpp/src/llama-arch.cpp (+19 -0)
📝 llama/llama.cpp/src/llama-arch.h (+1 -0)
📝 llama/llama.cpp/src/llama-model.cpp (+7 -0)
📝 llama/llama.cpp/src/llama-quant.cpp (+9 -0)
➕ llama/patches/0021-gemma3-quantization.patch (+113 -0)

📄 Description

This change allows users to use ollama create --quantize <level> to quantize gemma3 models.

We still use llama.cpp in order to do quantization with ollama create and do not make calls directly to ggml (as with ollama run) so we need to define the architecture so that llama.cpp can correctly do the quantization. In the future we can make a call to quantize each tensor directly.

Note that the vision part of the model is now included in the GGUF file and does not get quantized along with the text part of the model.

🔄 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/9776 **Author:** [@pdevine](https://github.com/pdevine) **Created:** 3/14/2025 **Status:** ✅ Merged **Merged:** 3/15/2025 **Merged by:** [@pdevine](https://github.com/pdevine) **Base:** `main` ← **Head:** `pdevine/gemma3-quant` --- ### 📝 Commits (1) - [`a4203fa`](https://github.com/ollama/ollama/commit/a4203fabc4ca811104474d32c8819e013df89196) gemma3 quantization ### 📊 Changes **5 files changed** (+149 additions, -0 deletions) <details> <summary>View changed files</summary> 📝 `llama/llama.cpp/src/llama-arch.cpp` (+19 -0) 📝 `llama/llama.cpp/src/llama-arch.h` (+1 -0) 📝 `llama/llama.cpp/src/llama-model.cpp` (+7 -0) 📝 `llama/llama.cpp/src/llama-quant.cpp` (+9 -0) ➕ `llama/patches/0021-gemma3-quantization.patch` (+113 -0) </details> ### 📄 Description This change allows users to use `ollama create --quantize <level>` to quantize gemma3 models. We still use llama.cpp in order to do quantization with `ollama create` and do not make calls directly to ggml (as with `ollama run`) so we need to define the architecture so that llama.cpp can correctly do the quantization. In the future we can make a call to quantize each tensor directly. Note that the vision part of the model is now included in the GGUF file and does *not* get quantized along with the text part of the model. --- <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 2026-04-22 23:35:02 -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 pull-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#38930
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?