github-starred/ollama
2
0
Fork 0
mirror of https://github.com/ollama/ollama.git synced 2026-05-07 00:22:43 -05:00
Code Issues 14.5k Packages Projects Releases 120 Wiki Activity

[GH-ISSUE #12276] Ollama v0.11.11-rc1 - CUDA error: invalid argument #70220

New Issue
Closed
opened 2026-05-04 20:42:02 -05:00 by GiteaMirror · 6 comments
GiteaMirror commented 2026-05-04 20:42:02 -05:00
Owner
Copy Link

Originally created by @eXt73 on GitHub (Sep 12, 2025).
Original GitHub issue: https://github.com/ollama/ollama/issues/12276

What is the issue?

Ollama works under settings

Environment="OLLAMA_FLASH_ATTENTION=1"
Environment="OLLAMA_KV_CACHE_TYPE=q4_0"
Environment="OLLAMA_NEW_ESTIMATES=1"
Environment="OLLAMA_NEW_ENGINE=1"

The model is Mistral-3.2-nov-2506-eXt73-tuned-v1.5-GGUF:Q6_K with the vision module removed and the parameters modified (hf.co/unsloth/Mistral-Small-3.2-24B-Instruct-2506-GGUF:Q6_K).

NVIDIA-SMI 580.82.07 Driver Version: 580.82.07 CUDA Version: 13.0

It works flawlessly with ollama v0.11.10, but with v0.11.11-rc1 it receives the following errors:

ollama run Mistral-3.2-nov-2506-eXt73-tuned-v1.5-GGUF:Q6_K
Error: 500 Internal Server Error: llama runner process has terminated: CUDA error: invalid argument
current device: 0, in function ggml_backend_cuda_buffer_set_tensor at //ml/backend/ggml/ggml/src/ggml-cuda/ggml-cuda.cu:667
cudaMemcpyAsync((char *)tensor->data + offset, data, size, cudaMemcpyHostToDevice, ((cudaStream_t)0x2))
//ml/backend/ggml/ggml/src/ggml-cuda/ggml-cuda.cu:84: CUDA error
ext73@ext73-kernel:~$

Image

Relevant log output

OS

Kubuntu Linux

GPU

RTX 5090

CPU

AMD Ryzen 9950x3D

Ollama version

v0.11.11-rc1

Originally created by @eXt73 on GitHub (Sep 12, 2025). Original GitHub issue: https://github.com/ollama/ollama/issues/12276 ### What is the issue? Ollama works under settings Environment="OLLAMA_FLASH_ATTENTION=1" Environment="OLLAMA_KV_CACHE_TYPE=q4_0" Environment="OLLAMA_NEW_ESTIMATES=1" Environment="OLLAMA_NEW_ENGINE=1" The model is Mistral-3.2-nov-2506-eXt73-tuned-v1.5-GGUF:Q6_K with the vision module removed and the parameters modified (hf.co/unsloth/Mistral-Small-3.2-24B-Instruct-2506-GGUF:Q6_K). NVIDIA-SMI 580.82.07 Driver Version: 580.82.07 CUDA Version: 13.0 It works flawlessly with ollama v0.11.10, but with v0.11.11-rc1 it receives the following errors: ollama run Mistral-3.2-nov-2506-eXt73-tuned-v1.5-GGUF:Q6_K Error: 500 Internal Server Error: llama runner process has terminated: CUDA error: invalid argument current device: 0, in function ggml_backend_cuda_buffer_set_tensor at //ml/backend/ggml/ggml/src/ggml-cuda/ggml-cuda.cu:667 cudaMemcpyAsync((char *)tensor->data + offset, data, size, cudaMemcpyHostToDevice, ((cudaStream_t)0x2)) //ml/backend/ggml/ggml/src/ggml-cuda/ggml-cuda.cu:84: CUDA error ext73@ext73-kernel:~$ <img width="3840" height="2160" alt="Image" src="https://github.com/user-attachments/assets/388c28ad-400b-473b-aa10-7882c03fd81e" /> ### Relevant log output ```shell ``` ### OS Kubuntu Linux ### GPU RTX 5090 ### CPU AMD Ryzen 9950x3D ### Ollama version v0.11.11-rc1
GiteaMirror added the bug label 2026-05-04 20:42:02 -05:00
GiteaMirror commented 2026-05-04 20:42:07 -05:00
Author
Owner
Copy Link

@rick-github commented on GitHub (Sep 12, 2025):

A full text log will include a stack trace that may help with debugging. Details of the fine-tuning may also be useful, as would access to the model.

<!-- gh-comment-id:3286942677 --> @rick-github commented on GitHub (Sep 12, 2025): A full text [log](https://github.com/ollama/ollama/blob/main/docs/troubleshooting.md#how-to-troubleshoot-issues) will include a stack trace that may help with debugging. Details of the fine-tuning may also be useful, as would access to the model.
GiteaMirror commented 2026-05-04 20:42:07 -05:00
Author
Owner
Copy Link

@eXt73 commented on GitHub (Sep 13, 2025):

Interestingly, this happens on every model that works with the new engine - for example, when the standard Mistarl is started with the old engine, there is no problem... and what's strange is that I don't see anything in the logs > it doesn't generate them when listening > sudo journalctl -u ollama --follow

ext73@ext73-kernel:$ ollama ls
NAME ID SIZE MODIFIED
Mistral-3.2-nov-2506-eXt73-tuned-v1.5-GGUF:Q4 754d5d5cb6b8 14 GB 45 minutes ago
hf.co/unsloth/Mistral-Small-3.2-24B-Instruct-2506-GGUF:latest 39224c6559c9 15 GB 28 hours ago
hf.co/unsloth/Mistral-Small-3.2-24B-Instruct-2506-GGUF:Q6_K 4206c4a9ec35 20 GB 2 days ago
embeddinggemma:latest 85462619ee72 621 MB 7 days ago
hf.co/bartowski/Qwen_Qwen3-4B-Instruct-2507-GGUF:Q8_0 67fab01b4791 4.3 GB 5 weeks ago
ext73@ext73-kernel:$ ollama run hf.co/bartowski/Qwen_Qwen3-4B-Instruct-2507-GGUF:Q8_0
Error: 500 Internal Server Error: llama runner process has terminated: CUDA error: context is destroyed
current device: 0, in function ggml_backend_cuda_buffer_set_tensor at //ml/backend/ggml/ggml/src/ggml-cuda/ggml-cuda.cu:667
cudaMemcpyAsync((char *)tensor->data + offset, data, size, cudaMemcpyHostToDevice, ((cudaStream_t)0x2))
//ml/backend/ggml/ggml/src/ggml-cuda/ggml-cuda.cu:84: CUDA error
ext73@ext73-kernel:~$ ollama run hf.co/unsloth/Mistral-Small-3.2-24B-Instruct-2506-GGUF:latest

Image
<!-- gh-comment-id:3288958144 --> @eXt73 commented on GitHub (Sep 13, 2025): Interestingly, this happens on every model that works with the new engine - for example, when the standard Mistarl is started with the old engine, there is no problem... and what's strange is that I don't see anything in the logs > it doesn't generate them when listening > sudo journalctl -u ollama --follow ext73@ext73-kernel:~$ ollama ls NAME ID SIZE MODIFIED Mistral-3.2-nov-2506-eXt73-tuned-v1.5-GGUF:Q4 754d5d5cb6b8 14 GB 45 minutes ago hf.co/unsloth/Mistral-Small-3.2-24B-Instruct-2506-GGUF:latest 39224c6559c9 15 GB 28 hours ago hf.co/unsloth/Mistral-Small-3.2-24B-Instruct-2506-GGUF:Q6_K 4206c4a9ec35 20 GB 2 days ago embeddinggemma:latest 85462619ee72 621 MB 7 days ago hf.co/bartowski/Qwen_Qwen3-4B-Instruct-2507-GGUF:Q8_0 67fab01b4791 4.3 GB 5 weeks ago ext73@ext73-kernel:~$ ollama run hf.co/bartowski/Qwen_Qwen3-4B-Instruct-2507-GGUF:Q8_0 Error: 500 Internal Server Error: llama runner process has terminated: CUDA error: context is destroyed current device: 0, in function ggml_backend_cuda_buffer_set_tensor at //ml/backend/ggml/ggml/src/ggml-cuda/ggml-cuda.cu:667 cudaMemcpyAsync((char *)tensor->data + offset, data, size, cudaMemcpyHostToDevice, ((cudaStream_t)0x2)) //ml/backend/ggml/ggml/src/ggml-cuda/ggml-cuda.cu:84: CUDA error ext73@ext73-kernel:~$ ollama run hf.co/unsloth/Mistral-Small-3.2-24B-Instruct-2506-GGUF:latest <img width="1632" height="1287" alt="Image" src="https://github.com/user-attachments/assets/2953151f-d063-4bf0-8729-8b4cc8f24be0" />
GiteaMirror commented 2026-05-04 20:42:08 -05:00
Author
Owner
Copy Link

@alsavu commented on GitHub (Sep 16, 2025):

I am having the exact same issue but not with rc1 of 0.11.11, with the full final release, I updated today and I am having the same issue, any progress on this ?

<!-- gh-comment-id:3297417480 --> @alsavu commented on GitHub (Sep 16, 2025): I am having the exact same issue but not with rc1 of 0.11.11, with the full final release, I updated today and I am having the same issue, any progress on this ?
GiteaMirror commented 2026-05-04 20:42:09 -05:00
Author
Owner
Copy Link

@rick-github commented on GitHub (Sep 16, 2025):

https://github.com/ollama/ollama/issues/12276#issuecomment-3286942677

<!-- gh-comment-id:3297659113 --> @rick-github commented on GitHub (Sep 16, 2025): https://github.com/ollama/ollama/issues/12276#issuecomment-3286942677
GiteaMirror commented 2026-05-04 20:42:11 -05:00
Author
Owner
Copy Link

@eXt73 commented on GitHub (Sep 16, 2025):

I can confirm that this problem no longer occurs in my case under stable 0.11.11 - it beautifully sets the 'neutered' Mistral from the vision module on the new engine :)

Image
<!-- gh-comment-id:3300468209 --> @eXt73 commented on GitHub (Sep 16, 2025): I can confirm that this problem no longer occurs in my case under stable 0.11.11 - it beautifully sets the 'neutered' Mistral from the vision module on the new engine :) <img width="3840" height="2160" alt="Image" src="https://github.com/user-attachments/assets/1e77fb35-ddaa-40e9-b28a-1b14b88dc0cd" />
GiteaMirror commented 2026-05-04 20:42:12 -05:00
Author
Owner
Copy Link

@alsavu commented on GitHub (Sep 18, 2025):

I also can confirm this problem no longer occurs for me as well, as I reinstalled version 0.11.11 and it works very well, I tested more than 30 models and I had no issues, thank you

<!-- gh-comment-id:3306410700 --> @alsavu commented on GitHub (Sep 18, 2025): I also can confirm this problem no longer occurs for me as well, as I reinstalled version 0.11.11 and it works very well, I tested more than 30 models and I had no issues, thank you
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#70220
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?