github-starred/ollama
2
0
Fork 0
mirror of https://github.com/ollama/ollama.git synced 2026-05-08 00:51:34 -05:00
Code Issues 14.5k Packages Projects Releases 120 Wiki Activity

[GH-ISSUE #3074] CUDA error: an illegal memory access was encountered #63929

New Issue
Closed
opened 2026-05-03 15:26:39 -05:00 by GiteaMirror · 6 comments
GiteaMirror commented 2026-05-03 15:26:39 -05:00
Owner
Copy Link

Originally created by @lizhichao999 on GitHub (Mar 12, 2024).
Original GitHub issue: https://github.com/ollama/ollama/issues/3074

Originally assigned to: @dhiltgen on GitHub.

ollama run llama2
When I executed the command: ollama run llama2, an error occurred related to the data information.
1710236043402

I only have a 128G GPU. Do I need more than 2 GPUs? I can run it on my local laptop with 2 GPUs without any errors.

Originally created by @lizhichao999 on GitHub (Mar 12, 2024). Original GitHub issue: https://github.com/ollama/ollama/issues/3074 Originally assigned to: @dhiltgen on GitHub. ollama run llama2 When I executed the command: ollama run llama2, an error occurred related to the data information. ![1710236043402](https://github.com/ollama/ollama/assets/34128722/098fb024-f6f6-419e-99a0-354d29eecd2d) I only have a 128G GPU. Do I need more than 2 GPUs? I can run it on my local laptop with 2 GPUs without any errors.
GiteaMirror added the bugnvidia labels 2026-05-03 15:26:40 -05:00
GiteaMirror commented 2026-05-03 15:26:42 -05:00
Author
Owner
Copy Link

@dhiltgen commented on GitHub (Mar 12, 2024):

Could you share some additional information about your system? What GPU do you have? Windows version, nvidia driver version.

This might be a software defect, but I've also seen this behavior from a system with a hardware fault. Have you run other applications on your GPU without problem?

<!-- gh-comment-id:1991909696 --> @dhiltgen commented on GitHub (Mar 12, 2024): Could you share some additional information about your system? What GPU do you have? Windows version, nvidia driver version. This might be a software defect, but I've also seen this behavior from a system with a hardware fault. Have you run other applications on your GPU without problem?
GiteaMirror commented 2026-05-03 15:26:42 -05:00
Author
Owner
Copy Link

@lizhichao999 commented on GitHub (Mar 13, 2024):

system：Windows Server 2022
GPU： NVIDIA RTX A6000
nvidia driver version：31.0.15.5123

Could you share some additional information about your system? What GPU do you have? Windows version, nvidia driver version.

This might be a software defect, but I've also seen this behavior from a system with a hardware fault. Have you run other applications on your GPU without problem?

system：Windows Server 2022
GPU： NVIDIA RTX A6000
nvidia driver version：31.0.15.5123

<!-- gh-comment-id:1992777945 --> @lizhichao999 commented on GitHub (Mar 13, 2024): system：Windows Server 2022 GPU： NVIDIA RTX A6000 nvidia driver version：31.0.15.5123 > Could you share some additional information about your system? What GPU do you have? Windows version, nvidia driver version. > > This might be a software defect, but I've also seen this behavior from a system with a hardware fault. Have you run other applications on your GPU without problem? system：Windows Server 2022 GPU： NVIDIA RTX A6000 nvidia driver version：31.0.15.5123
GiteaMirror commented 2026-05-03 15:26:43 -05:00
Author
Owner
Copy Link

@lizhichao999 commented on GitHub (Mar 13, 2024):

1710288670426
Other programs don't have this problem.

<!-- gh-comment-id:1992789080 --> @lizhichao999 commented on GitHub (Mar 13, 2024): ![1710288670426](https://github.com/ollama/ollama/assets/34128722/a3b2f1e6-fa69-4c60-8860-f083bd48ccb0) Other programs don't have this problem.
GiteaMirror commented 2026-05-03 15:26:45 -05:00
Author
Owner
Copy Link

@dhiltgen commented on GitHub (Mar 13, 2024):

Doing a quick code inspection between 0.1.27 (the version you're running based on error line numbers) and 0.1.29, I see quite a bit of changes have been made upstream in llama.cpp in ggml-cuda.cu (where this crash is taking place) I haven't root caused your crash yet, but you might want to try out 0.1.29 and see if it fixes the problem. If not, we'll keep digging to understand where the defect is.

<!-- gh-comment-id:1994694357 --> @dhiltgen commented on GitHub (Mar 13, 2024): Doing a quick code inspection between 0.1.27 (the version you're running based on error line numbers) and 0.1.29, I see quite a bit of changes have been made upstream in llama.cpp in ggml-cuda.cu (where this crash is taking place) I haven't root caused your crash yet, but you might want to try out [0.1.29 ](https://github.com/ollama/ollama/releases/tag/v0.1.29)and see if it fixes the problem. If not, we'll keep digging to understand where the defect is.
GiteaMirror commented 2026-05-03 15:26:45 -05:00
Author
Owner
Copy Link

@lizhichao999 commented on GitHub (Mar 14, 2024):

Doing a quick code inspection between 0.1.27 (the version you're running based on error line numbers) and 0.1.29, I see quite a bit of changes have been made upstream in llama.cpp in ggml-cuda.cu (where this crash is taking place) I haven't root caused your crash yet, but you might want to try out 0.1.29 and see if it fixes the problem. If not, we'll keep digging to understand where the defect is.

I have updated the software to version 0.1.29, but the issue still persists.
1710376844249

<!-- gh-comment-id:1996191436 --> @lizhichao999 commented on GitHub (Mar 14, 2024): > Doing a quick code inspection between 0.1.27 (the version you're running based on error line numbers) and 0.1.29, I see quite a bit of changes have been made upstream in llama.cpp in ggml-cuda.cu (where this crash is taking place) I haven't root caused your crash yet, but you might want to try out [0.1.29 ](https://github.com/ollama/ollama/releases/tag/v0.1.29)and see if it fixes the problem. If not, we'll keep digging to understand where the defect is. I have updated the software to version 0.1.29, but the issue still persists. ![1710376844249](https://github.com/ollama/ollama/assets/34128722/221f2ec6-f452-40b7-847e-99fdd3ecd730)
GiteaMirror commented 2026-05-03 15:26:46 -05:00
Author
Owner
Copy Link

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

Is this still an issue with the latest version? If so, can you share a server log with OLLAMA_DEBUG=1 set?

<!-- gh-comment-id:2207247369 --> @dhiltgen commented on GitHub (Jul 3, 2024): Is this still an issue with the latest version? If so, can you share a server log with OLLAMA_DEBUG=1 set?
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 nvidia
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#63929
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?