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

[PR #10637] [MERGED] chore: update mllama to use ollama engine #60002

New Issue
Closed
opened 2026-04-29 14:55:27 -05:00 by GiteaMirror · 0 comments
GiteaMirror commented 2026-04-29 14:55:27 -05:00
Owner
Copy Link

📋 Pull Request Information

Original PR: https://github.com/ollama/ollama/pull/10637
Author: @mxyng
Created: 5/9/2025
Status: ✅ Merged
Merged: 5/14/2025
Merged by: @mxyng

Base: main ← Head: mxyng/mllama

📝 Commits (10+)

📊 Changes

67 files changed (+785 additions, -4354 deletions)

View changed files

📝 Makefile.sync (+10 -7)
📝 convert/convert.go (+14 -14)
➕ convert/convert_mllama.go (+160 -0)
📝 convert/reader.go (+4 -1)
📝 fs/ggml/ggml.go (+1 -0)
📝 llama/llama.cpp/include/llama.h (+0 -6)
📝 llama/llama.cpp/src/llama-arch.cpp (+0 -44)
📝 llama/llama.cpp/src/llama-arch.h (+0 -10)
📝 llama/llama.cpp/src/llama-batch.cpp (+0 -3)
📝 llama/llama.cpp/src/llama-context.cpp (+8 -15)
📝 llama/llama.cpp/src/llama-context.h (+0 -1)
📝 llama/llama.cpp/src/llama-cparams.h (+0 -1)
📝 llama/llama.cpp/src/llama-graph.cpp (+0 -25)
📝 llama/llama.cpp/src/llama-graph.h (+0 -12)
📝 llama/llama.cpp/src/llama-hparams.cpp (+0 -4)
📝 llama/llama.cpp/src/llama-hparams.h (+0 -7)
📝 llama/llama.cpp/src/llama-kv-cache.cpp (+3 -11)
📝 llama/llama.cpp/src/llama-model-loader.cpp (+0 -2)
📝 llama/llama.cpp/src/llama-model.cpp (+2 -307)
📝 llama/llama.cpp/src/llama-model.h (+0 -12)

...and 47 more files

📄 Description

this change drops the mllama patch to llama.cpp and routes those model requests to the ollama engine.

this also fixes an issue with the implementation where the processed image is contains less than 4 tiles. since the model is trained with 4 tiles, e.g. it's precomputed tile positional embeddings, this must be padded. previously this would've been implicit but the new implementation requires explicit padding

🔄 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/10637 **Author:** [@mxyng](https://github.com/mxyng) **Created:** 5/9/2025 **Status:** ✅ Merged **Merged:** 5/14/2025 **Merged by:** [@mxyng](https://github.com/mxyng) **Base:** `main` ← **Head:** `mxyng/mllama` --- ### 📝 Commits (10+) - [`0628b7a`](https://github.com/ollama/ollama/commit/0628b7a8b75111084ed2216b150c0006429a49d9) remove mllama patch - [`2760ec3`](https://github.com/ollama/ollama/commit/2760ec37b50d4f7d75bdb281331af37b36b3168f) remove mllama integration, use ollama engine - [`3394ae1`](https://github.com/ollama/ollama/commit/3394ae13b6b370629aa01e2d427cdd60d729eca9) pad negative - [`ae393c7`](https://github.com/ollama/ollama/commit/ae393c7569e41295e6be4170ab1a5e8db8f0a5a1) fix: mllama image processing - [`c1d5f16`](https://github.com/ollama/ollama/commit/c1d5f16593c359b29831517268d064554c799587) fix: build - [`73c0fd9`](https://github.com/ollama/ollama/commit/73c0fd9218cbb68f6815b21de3b2f2a2fe845bb9) error on legacy llama3.2-vision - [`4df6652`](https://github.com/ollama/ollama/commit/4df66527b817cb272d8aecee4db10255ee04f6bb) comments - [`750138e`](https://github.com/ollama/ollama/commit/750138e812ffc9c29326d365a105f2632acec56f) update mllama - [`75eaccf`](https://github.com/ollama/ollama/commit/75eaccf1d2fb9eeb51d8713009e532230c28b8cf) convert: mllama - [`dfa439d`](https://github.com/ollama/ollama/commit/dfa439d3bed70e10ea7ad05ee953bef5e8f59112) remove mask ### 📊 Changes **67 files changed** (+785 additions, -4354 deletions) <details> <summary>View changed files</summary> 📝 `Makefile.sync` (+10 -7) 📝 `convert/convert.go` (+14 -14) ➕ `convert/convert_mllama.go` (+160 -0) 📝 `convert/reader.go` (+4 -1) 📝 `fs/ggml/ggml.go` (+1 -0) 📝 `llama/llama.cpp/include/llama.h` (+0 -6) 📝 `llama/llama.cpp/src/llama-arch.cpp` (+0 -44) 📝 `llama/llama.cpp/src/llama-arch.h` (+0 -10) 📝 `llama/llama.cpp/src/llama-batch.cpp` (+0 -3) 📝 `llama/llama.cpp/src/llama-context.cpp` (+8 -15) 📝 `llama/llama.cpp/src/llama-context.h` (+0 -1) 📝 `llama/llama.cpp/src/llama-cparams.h` (+0 -1) 📝 `llama/llama.cpp/src/llama-graph.cpp` (+0 -25) 📝 `llama/llama.cpp/src/llama-graph.h` (+0 -12) 📝 `llama/llama.cpp/src/llama-hparams.cpp` (+0 -4) 📝 `llama/llama.cpp/src/llama-hparams.h` (+0 -7) 📝 `llama/llama.cpp/src/llama-kv-cache.cpp` (+3 -11) 📝 `llama/llama.cpp/src/llama-model-loader.cpp` (+0 -2) 📝 `llama/llama.cpp/src/llama-model.cpp` (+2 -307) 📝 `llama/llama.cpp/src/llama-model.h` (+0 -12) _...and 47 more files_ </details> ### 📄 Description this change drops the mllama patch to llama.cpp and routes those model requests to the ollama engine. this also fixes an issue with the implementation where the processed image is contains less than 4 tiles. since the model is trained with 4 tiles, e.g. it's precomputed tile positional embeddings, this must be padded. previously this would've been implicit but the new implementation requires explicit padding --- <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 2026-04-29 14:55:27 -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#60002
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?