github-starred/ollama
2
0
Fork 0
mirror of https://github.com/ollama/ollama.git synced 2026-05-06 08:02:14 -05:00
Code Issues 14.5k Packages Projects Releases 120 Wiki Activity

[GH-ISSUE #6591] Ollama failing with CUDA error: PTX JIT compiler library not found #66186

New Issue
Closed
opened 2026-05-04 00:30:38 -05:00 by GiteaMirror · 9 comments
GiteaMirror commented 2026-05-04 00:30:38 -05:00
Owner
Copy Link

Originally created by @leobenkel on GitHub (Sep 2, 2024).
Original GitHub issue: https://github.com/ollama/ollama/issues/6591

What is the issue?

The GPU seems to be detected:

CUDA driver version: 11.4
time=2024-08-30T16:57:23.032Z level=DEBUG source=gpu.go:123 msg="detected GPUs" count=1 library=/usr/lib/aarch64-linux-gnu/tegra/libcuda.so.1.1
[GPU-d90c0d9d-5e59-56b1-b519-6439b1d74328] CUDA totalMem 30990 mb
[GPU-d90c0d9d-5e59-56b1-b519-6439b1d74328] CUDA freeMem 23431 mb
[GPU-d90c0d9d-5e59-56b1-b519-6439b1d74328] Compute Capability 7.2
...
time=2024-08-30T16:57:23.201Z level=INFO source=types.go:105 msg="inference compute" id=GPU-d90c0d9d-5e59-56b1-b519-6439b1d74328 library=cuda compute=7.2 driver=11.4 name=Xavier total="30.3 GiB" available="22.9 GiB"
...
time=2024-08-30T16:58:09.212Z level=DEBUG source=gpu.go:410 msg="updating cuda memory data" gpu=GPU-d90c0d9d-5e59-56b1-b519-6439b1d74328 name=Xavier overhead="0 B" before.total="30.3 GiB" before.free="22.9 GiB" now.total="30.3 GiB" now.free="23.0 GiB" now.used="7.3 GiB"

But then I get this error:

CUDA error: PTX JIT compiler library not found
  current device: 0, in function ggml_cuda_compute_forward at /go/src/github.com/ollama/ollama/llm/llama.cpp/ggml/src/ggml-cuda.cu:2313
/go/src/github.com/ollama/ollama/llm/llama.cpp/ggml/src/ggml-cuda.cu:101: CUDA error

Here are the installed version:

ii  nvidia-cuda                                5.1.3-b29                             arm64        NVIDIA CUDA Meta Package
ii  nvidia-cuda-dev                            5.1.3-b29                             arm64        NVIDIA CUDA dev Meta Package
ii  nvidia-jetpack                             5.1.3-b29                             arm64        NVIDIA Jetpack Meta Package
ii  nvidia-jetpack-dev                         5.1.3-b29                             arm64        NVIDIA Jetpack dev Meta Package
ii  nvidia-jetpack-runtime                     5.1.3-b29                             arm64        NVIDIA Jetpack runtime Meta Package
ii  nvidia-l4t-cuda                            35.5.0-20240219203809                 arm64        NVIDIA CUDA Package
ii  cuda                                       11.4.19-1                             arm64        CUDA meta-package

Ollama is used with the docker container et inside a kubernetes cluster.

OS

Docker

GPU

Nvidia

CPU

ARM

Ollama version

0.3.8

Originally created by @leobenkel on GitHub (Sep 2, 2024). Original GitHub issue: https://github.com/ollama/ollama/issues/6591 ### What is the issue? The GPU seems to be detected: ``` CUDA driver version: 11.4 time=2024-08-30T16:57:23.032Z level=DEBUG source=gpu.go:123 msg="detected GPUs" count=1 library=/usr/lib/aarch64-linux-gnu/tegra/libcuda.so.1.1 [GPU-d90c0d9d-5e59-56b1-b519-6439b1d74328] CUDA totalMem 30990 mb [GPU-d90c0d9d-5e59-56b1-b519-6439b1d74328] CUDA freeMem 23431 mb [GPU-d90c0d9d-5e59-56b1-b519-6439b1d74328] Compute Capability 7.2 ... time=2024-08-30T16:57:23.201Z level=INFO source=types.go:105 msg="inference compute" id=GPU-d90c0d9d-5e59-56b1-b519-6439b1d74328 library=cuda compute=7.2 driver=11.4 name=Xavier total="30.3 GiB" available="22.9 GiB" ... time=2024-08-30T16:58:09.212Z level=DEBUG source=gpu.go:410 msg="updating cuda memory data" gpu=GPU-d90c0d9d-5e59-56b1-b519-6439b1d74328 name=Xavier overhead="0 B" before.total="30.3 GiB" before.free="22.9 GiB" now.total="30.3 GiB" now.free="23.0 GiB" now.used="7.3 GiB" ``` But then I get this error: ``` CUDA error: PTX JIT compiler library not found current device: 0, in function ggml_cuda_compute_forward at /go/src/github.com/ollama/ollama/llm/llama.cpp/ggml/src/ggml-cuda.cu:2313 /go/src/github.com/ollama/ollama/llm/llama.cpp/ggml/src/ggml-cuda.cu:101: CUDA error ``` Here are the installed version: ``` ii nvidia-cuda 5.1.3-b29 arm64 NVIDIA CUDA Meta Package ii nvidia-cuda-dev 5.1.3-b29 arm64 NVIDIA CUDA dev Meta Package ii nvidia-jetpack 5.1.3-b29 arm64 NVIDIA Jetpack Meta Package ii nvidia-jetpack-dev 5.1.3-b29 arm64 NVIDIA Jetpack dev Meta Package ii nvidia-jetpack-runtime 5.1.3-b29 arm64 NVIDIA Jetpack runtime Meta Package ii nvidia-l4t-cuda 35.5.0-20240219203809 arm64 NVIDIA CUDA Package ii cuda 11.4.19-1 arm64 CUDA meta-package ``` Ollama is used with the docker container et inside a kubernetes cluster. ### OS Docker ### GPU Nvidia ### CPU ARM ### Ollama version 0.3.8
GiteaMirror added the bug label 2026-05-04 00:30:39 -05:00
GiteaMirror commented 2026-05-04 00:30:43 -05:00
Author
Owner
Copy Link

@dhiltgen commented on GitHub (Sep 3, 2024):

The Jetson ARM systems wont be working properly until we merge #6400

Support for Jetpack 5 is tracked via #4693

<!-- gh-comment-id:2327448413 --> @dhiltgen commented on GitHub (Sep 3, 2024): The Jetson ARM systems wont be working properly until we merge #6400 Support for Jetpack 5 is tracked via #4693
GiteaMirror commented 2026-05-04 00:30:45 -05:00
Author
Owner
Copy Link

@leobenkel commented on GitHub (Sep 4, 2024):

thank you for your response. Is there a work around in the mid-time ?

<!-- gh-comment-id:2328218280 --> @leobenkel commented on GitHub (Sep 4, 2024): thank you for your response. Is there a work around in the mid-time ?
GiteaMirror commented 2026-05-04 00:30:46 -05:00
Author
Owner
Copy Link

@dhiltgen commented on GitHub (Sep 4, 2024):

Building from source locally with the cuda toolkit installed should work.

<!-- gh-comment-id:2329496553 --> @dhiltgen commented on GitHub (Sep 4, 2024): Building from source locally with the cuda toolkit installed should work.
GiteaMirror commented 2026-05-04 00:30:47 -05:00
Author
Owner
Copy Link

@leobenkel commented on GitHub (Sep 4, 2024):

ok, i am going to try this then. thank you

<!-- gh-comment-id:2329502612 --> @leobenkel commented on GitHub (Sep 4, 2024): ok, i am going to try this then. thank you
GiteaMirror commented 2026-05-04 00:30:49 -05:00
Author
Owner
Copy Link

@leobenkel commented on GitHub (Sep 7, 2024):

@dhiltgen , just a question: should i pull and build on the physical node of the cluster or should i pull and build from inside the docker image ?

<!-- gh-comment-id:2336427620 --> @leobenkel commented on GitHub (Sep 7, 2024): @dhiltgen , just a question: should i pull and build on the physical node of the cluster or should i pull and build from inside the docker image ?
GiteaMirror commented 2026-05-04 00:30:51 -05:00
Author
Owner
Copy Link

@dhiltgen commented on GitHub (Sep 25, 2024):

@leobenkel it doesn't really matter, however make sure you have the right jetpack cuda version to match your target system.

<!-- gh-comment-id:2375210100 --> @dhiltgen commented on GitHub (Sep 25, 2024): @leobenkel it doesn't really matter, however make sure you have the right jetpack cuda version to match your target system.
GiteaMirror commented 2026-05-04 00:30:53 -05:00
Author
Owner
Copy Link

@leobenkel commented on GitHub (Sep 26, 2024):

@leobenkel it doesn't really matter, however make sure you have the right jetpack cuda version to match your target system.

thank you for your response.

Right now the system i am working with has jetpack 5, it should work right ?
which script should i run to build ? i found several ways in the documentation. is it ./build_linux.sh ?

<!-- gh-comment-id:2376538245 --> @leobenkel commented on GitHub (Sep 26, 2024): > @leobenkel it doesn't really matter, however make sure you have the right jetpack cuda version to match your target system. thank you for your response. Right now the system i am working with has jetpack 5, it should work right ? which script should i run to build ? i found several ways in the documentation. is it `./build_linux.sh` ?
GiteaMirror commented 2026-05-04 00:30:54 -05:00
Author
Owner
Copy Link

@dhiltgen commented on GitHub (Sep 26, 2024):

Linux dev guide: https://github.com/ollama/ollama/blob/main/docs/development.md#linux

<!-- gh-comment-id:2377671287 --> @dhiltgen commented on GitHub (Sep 26, 2024): Linux dev guide: https://github.com/ollama/ollama/blob/main/docs/development.md#linux
GiteaMirror commented 2026-05-04 00:30:55 -05:00
Author
Owner
Copy Link

@leobenkel commented on GitHub (Sep 26, 2024):

Linux dev guide: https://github.com/ollama/ollama/blob/main/docs/development.md#linux

ok thats what i followed but without success

<!-- gh-comment-id:2377816534 --> @leobenkel commented on GitHub (Sep 26, 2024): > Linux dev guide: https://github.com/ollama/ollama/blob/main/docs/development.md#linux ok thats what i followed but without success
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 bug
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#66186
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?