[PR #5685] [CLOSED] Disable mmap by default for Windows ROCm #11885

Closed
opened 2026-04-12 23:41:49 -05:00 by GiteaMirror · 0 comments
Owner

📋 Pull Request Information

Original PR: https://github.com/ollama/ollama/pull/5685
Author: @zsmooter
Created: 7/14/2024
Status: Closed

Base: mainHead: main


📝 Commits (1)

  • d93fbf6 Disable mmap by default for Windows ROCm

📊 Changes

1 file changed (+2 additions, -2 deletions)

View changed files

📝 llm/server.go (+2 -2)

📄 Description

The same as with CUDA, disabling mmap when using ROCm on windows seems to speed up model load times significantly. I get a >2x speedup in model load times on my 7900xtx when disabling mmap.


🔄 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/5685 **Author:** [@zsmooter](https://github.com/zsmooter) **Created:** 7/14/2024 **Status:** ❌ Closed **Base:** `main` ← **Head:** `main` --- ### 📝 Commits (1) - [`d93fbf6`](https://github.com/ollama/ollama/commit/d93fbf6a60b95234656b2c6b2f734cf4429a716d) Disable mmap by default for Windows ROCm ### 📊 Changes **1 file changed** (+2 additions, -2 deletions) <details> <summary>View changed files</summary> 📝 `llm/server.go` (+2 -2) </details> ### 📄 Description The same as with CUDA, disabling mmap when using ROCm on windows seems to speed up model load times significantly. I get a >2x speedup in model load times on my 7900xtx when disabling mmap. --- <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 2026-04-12 23:41:49 -05:00
Sign in to join this conversation.
1 Participants
Notifications
Due Date
No due date set.
Dependencies

No dependencies set.

Reference: github-starred/ollama#11885