github-starred/ollama
2
0
Fork 0
mirror of https://github.com/ollama/ollama.git synced 2026-05-08 08:59:13 -05:00
Code Issues 14.5k Packages Projects Releases 120 Wiki Activity

[PR #11635] cuda: leverage JIT for smaller footprint #13597

New Issue
Closed
opened 2026-04-13 00:30:54 -05:00 by GiteaMirror · 0 comments
GiteaMirror commented 2026-04-13 00:30:54 -05:00
Owner
Copy Link

Original Pull Request: https://github.com/ollama/ollama/pull/11635

State: closed
Merged: Yes

Prior to this change our official binaries contained both JIT PTX code and the cubin binary code for our chosen compute capabilities. This change switches to only compile the PTX code and rely on JIT at runtime for generating the cubin specific to the users GPU. The cubins are cached on the users system, so they should only see a small lag on the very first model load for a given Ollama release. This also adds the first generation of Blackwell GPUs so they aren't reliant on the Hopper PTX.

This change reduces the ggml-cuda.dll from 1.2G to 460M

I also removed CC 8.7 as that appears to be only a Jetson CC and unused on x86.

Testing on a dual 4060 Windows system, loading gpt-oss:20b takes 6.01s before this change. With this change, on the very first load, it takes 6.26s, then all subsequent loads take ~6.0s. Token rate is unaffected.

**Original Pull Request:** https://github.com/ollama/ollama/pull/11635 **State:** closed **Merged:** Yes --- Prior to this change our official binaries contained both JIT PTX code and the cubin binary code for our chosen compute capabilities. This change switches to only compile the PTX code and rely on JIT at runtime for generating the cubin specific to the users GPU. The cubins are cached on the users system, so they should only see a small lag on the very first model load for a given Ollama release. This also adds the first generation of Blackwell GPUs so they aren't reliant on the Hopper PTX. This change reduces the ggml-cuda.dll from 1.2G to 460M I also removed CC 8.7 as that appears to be only a Jetson CC and unused on x86. Testing on a dual 4060 Windows system, loading gpt-oss:20b takes 6.01s before this change. With this change, on the very first load, it takes 6.26s, then all subsequent loads take ~6.0s. Token rate is unaffected.
GiteaMirror added the pull-request label 2026-04-13 00:30:55 -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#13597
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?