github-starred/ollama
2
0
Fork 0
mirror of https://github.com/ollama/ollama.git synced 2026-05-07 16:40:08 -05:00
Code Issues 14.5k Packages Projects Releases 120 Wiki Activity

[PR #5527] Add Environment Variable For Row Split #43066

New Issue
Open
opened 2026-04-24 22:45:56 -05:00 by GiteaMirror · 0 comments
GiteaMirror commented 2026-04-24 22:45:56 -05:00
Owner
Copy Link

📋 Pull Request Information

Original PR: https://github.com/ollama/ollama/pull/5527
Author: @datacrystals
Created: 7/7/2024
Status: 🔄 Open

Base: main ← Head: main

📝 Commits (1)

  • e25faca Add Environment Variable For Row Split

📊 Changes

2 files changed (+16 additions, -0 deletions)

View changed files

📝 envconfig/config.go (+12 -0)
📝 llm/server.go (+4 -0)

📄 Description

As discussed in #5458, there does not appear to be a way to enable row-split rather than layer splitting, which on older multi-gpu setups seems to result in a 40-70% performance improvement.

I tested this on 3xP40 24GB GPUs running Ubuntu Server 22.04, and observed about a 70% improvement in throughput, but it does seem to vary model to model.

I've gone ahead and added it with a really simple change (added an environment variable, which when set to 1 enables row split rather than layer splitting) - please let me know if any edits are needed, I'd be happy to go and make any changes needed.

🔄 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/5527 **Author:** [@datacrystals](https://github.com/datacrystals) **Created:** 7/7/2024 **Status:** 🔄 Open **Base:** `main` ← **Head:** `main` --- ### 📝 Commits (1) - [`e25faca`](https://github.com/ollama/ollama/commit/e25faca8797828d5203f4e35df545a154e649eab) Add Environment Variable For Row Split ### 📊 Changes **2 files changed** (+16 additions, -0 deletions) <details> <summary>View changed files</summary> 📝 `envconfig/config.go` (+12 -0) 📝 `llm/server.go` (+4 -0) </details> ### 📄 Description As discussed in #5458, there does not appear to be a way to enable row-split rather than layer splitting, which on older multi-gpu setups seems to result in a 40-70% performance improvement. I tested this on 3xP40 24GB GPUs running Ubuntu Server 22.04, and observed about a 70% improvement in throughput, but it does seem to vary model to model. I've gone ahead and added it with a really simple change (added an environment variable, which when set to 1 enables row split rather than layer splitting) - please let me know if any edits are needed, I'd be happy to go and make any changes needed. --- <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 2026-04-24 22:45:56 -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#43066
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?