github-starred/ollama
2
0
Fork 0
mirror of https://github.com/ollama/ollama.git synced 2026-05-07 08:30:05 -05:00
Code Issues 14.5k Packages Projects Releases 120 Wiki Activity

[GH-ISSUE #15156] Ollama fails to use a coding model that doesn't fit into Vulkan memory #56215

New Issue
Closed
opened 2026-04-29 10:26:29 -05:00 by GiteaMirror · 4 comments
GiteaMirror commented 2026-04-29 10:26:29 -05:00
Owner
Copy Link

Originally created by @yurivict on GitHub (Mar 31, 2026).
Original GitHub issue: https://github.com/ollama/ollama/issues/15156

What is the issue?

The system has the Vulkan device NVIDIA GeForce RTX 2060 with 6 GB VRAM.

claude is configured to use the model gpt-oss:20b. 20b parameters requires 40+GB of memory.

Ollama serve log.

Ollama serv was loading the model very slowly for some reason.

Ollama should put some number of layers onto the Vulkan device and use CPU for the rest.
But this doesn't happen for some reason.

The log shows that after 15+ minutes computation doesn't start due to whatever issues with slow model loading and timeouts.

Relevant log output

See above.

OS

Linux

GPU

Nvidia

CPU

AMD

Ollama version

0.19.0

Originally created by @yurivict on GitHub (Mar 31, 2026). Original GitHub issue: https://github.com/ollama/ollama/issues/15156 ### What is the issue? The system has the Vulkan device NVIDIA GeForce RTX 2060 with 6 GB VRAM. claude is configured to use the model gpt-oss:20b. 20b parameters requires 40+GB of memory. Ollama serve [log](https://freebsd.org/~yuri/ollama-serve-failure-when-model-is-larger-than-vulkan-memory.log). Ollama serv was loading the model very slowly for some reason. Ollama should put some number of layers onto the Vulkan device and use CPU for the rest. But this doesn't happen for some reason. The log shows that after 15+ minutes computation doesn't start due to whatever issues with slow model loading and timeouts. ### Relevant log output ```shell See above. ``` ### OS Linux ### GPU Nvidia ### CPU AMD ### Ollama version 0.19.0
GiteaMirror added the bugneeds more info labels 2026-04-29 10:26:29 -05:00
GiteaMirror commented 2026-04-29 10:26:32 -05:00
Author
Owner
Copy Link

@rick-github commented on GitHub (Mar 31, 2026):

claude is configured to use the model gpt-oss:20b. 20b parameters requires 40+GB of memory.

NAME                    ID              SIZE      PROCESSOR    CONTEXT    UNTIL   
gpt-oss:20b             17052f91a42e    17 GB     100% GPU     131072     Forever

Ollama should put some number of layers onto the Vulkan device and use CPU for the rest.

time=2026-03-30T17:17:56.396-07:00 level=INFO source=ggml.go:494 msg="offloaded 9/25 layers to GPU"

The log shows that after 15+ minutes computation doesn't start due to whatever issues with slow model loading and timeouts.

time=2026-03-30T17:22:55.802-07:00 level=WARN source=server.go:1359 msg="client connection closed before server finished loading, aborting load"

The client closed the connection before the load was finished.

After the failed load, subsequent runners fail with a SIGSEGV in ggml_backend_dev_init so it looks like the device is wedged somehow. A reboot would probably fix it but that's suboptimal. Why are you using Vulkan with an RTX 2060 rather than CUDA?

<!-- gh-comment-id:4159121204 --> @rick-github commented on GitHub (Mar 31, 2026): > claude is configured to use the model gpt-oss:20b. 20b parameters requires 40+GB of memory. ``` NAME ID SIZE PROCESSOR CONTEXT UNTIL gpt-oss:20b 17052f91a42e 17 GB 100% GPU 131072 Forever ``` > Ollama should put some number of layers onto the Vulkan device and use CPU for the rest. ``` time=2026-03-30T17:17:56.396-07:00 level=INFO source=ggml.go:494 msg="offloaded 9/25 layers to GPU" ``` > The log shows that after 15+ minutes computation doesn't start due to whatever issues with slow model loading and timeouts. ``` time=2026-03-30T17:22:55.802-07:00 level=WARN source=server.go:1359 msg="client connection closed before server finished loading, aborting load" ``` The client closed the connection before the load was finished. After the failed load, subsequent runners fail with a SIGSEGV in `ggml_backend_dev_init` so it looks like the device is wedged somehow. A reboot would probably fix it but that's suboptimal. Why are you using Vulkan with an RTX 2060 rather than CUDA?
GiteaMirror commented 2026-04-29 10:26:34 -05:00
Author
Owner
Copy Link

@yurivict commented on GitHub (Mar 31, 2026):

Why are you using Vulkan with an RTX 2060 rather than CUDA?

CUDA isn't supported on FreeBSD.

Reboot didn't change anything.

<!-- gh-comment-id:4159155040 --> @yurivict commented on GitHub (Mar 31, 2026): > Why are you using Vulkan with an RTX 2060 rather than CUDA? CUDA isn't supported on FreeBSD. Reboot didn't change anything.
GiteaMirror commented 2026-04-29 10:26:34 -05:00
Author
Owner
Copy Link

@rick-github commented on GitHub (Mar 31, 2026):

Logs from after the reboot?

<!-- gh-comment-id:4159160646 --> @rick-github commented on GitHub (Mar 31, 2026): Logs from after the reboot?
GiteaMirror commented 2026-04-29 10:26:34 -05:00
Author
Owner
Copy Link

@yurivict commented on GitHub (Mar 31, 2026):

This is the log from after the reboot.

<!-- gh-comment-id:4159163395 --> @yurivict commented on GitHub (Mar 31, 2026): This is the log from after the reboot.
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 needs more info
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#56215
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?