github-starred/ollama
2
0
Fork 0
mirror of https://github.com/ollama/ollama.git synced 2026-05-21 13:11:39 -05:00
Code Issues 16.9k Packages Projects Releases 122 Wiki Activity

[GH-ISSUE #4742] VRAM allocation error when loading different models with different OLLAMA_VRAM_MAX configurations #80666

New Issue
Closed
opened 2026-05-09 09:17:22 -05:00 by GiteaMirror · 3 comments
GiteaMirror commented 2026-05-09 09:17:22 -05:00
Owner
Copy Link

Originally created by @hamkido on GitHub (May 31, 2024).
Original GitHub issue: https://github.com/ollama/ollama/issues/4742

What is the issue?

I have two amd 7900xtx 24g gpu. When using ollama, I encounter different memory allocation errors and exit errors.

  1. No OLLAMA_VRAM_MAX configuration
    The large model deepseek-llm:67b-chat can be loaded correctly
    But if you call something bigger, such as qwen:72b and command-r-plus, the display memory allocation will report an error and exit.
  2. Use OLLAMA_VRAM_MAX configuration
    Models larger than 67b-q4 load correctly, such as qwen:72b and command-r-plus.
    However, smaller models cannot be loaded correctly, and memory allocation errors will be reported and exited, such as deepseek-llm:67b-chat.

There might be some errors in the vram configuration.

OS

Linux

GPU

AMD

CPU

AMD

Ollama version

0.1.39

Originally created by @hamkido on GitHub (May 31, 2024). Original GitHub issue: https://github.com/ollama/ollama/issues/4742 ### What is the issue? I have two amd 7900xtx 24g gpu. When using ollama, I encounter different memory allocation errors and exit errors. 1. No OLLAMA_VRAM_MAX configuration The large model deepseek-llm:67b-chat can be loaded correctly But if you call something bigger, such as qwen:72b and command-r-plus, the display memory allocation will report an error and exit. 2. Use OLLAMA_VRAM_MAX configuration Models larger than 67b-q4 load correctly, such as qwen:72b and command-r-plus. However, smaller models cannot be loaded correctly, and memory allocation errors will be reported and exited, such as deepseek-llm:67b-chat. There might be some errors in the vram configuration. ### OS Linux ### GPU AMD ### CPU AMD ### Ollama version 0.1.39
GiteaMirror added the bug label 2026-05-09 09:17:22 -05:00
GiteaMirror commented 2026-05-09 09:17:24 -05:00
Author
Owner
Copy Link

@hamkido commented on GitHub (May 31, 2024):

This is my ollama configuration

cat /etc/ollama.conf | grep -v '#'
OLLAMA_NUM_PARALLEL=2
OLLAMA_MAX_LOADED_MODELS=2
HSA_OVERRIDE_GFX_VERSION=11.0.0
ROCR_VISIBLE_DEVICES=0,1
HIP_VISIBLE_DEVICES=0,1
HOME=/var/lib/ollama
GIN_MODE=debug
OLLAMA_DEBUG=1
AMD_SERIALIZE_KERNEL=3
OLLAMA_LLM_LIBRARY=rocm_v60002
OLLAMA_ORIGINS=*
OLLAMA_HOST=0.0.0.0

Another configuration has one more line
OLLAMA_MAX_VRAM=49392123904

If you need more information and tests, like systemd log or tests, I will upload them.

<!-- gh-comment-id:2141270001 --> @hamkido commented on GitHub (May 31, 2024): This is my ollama configuration ``` cat /etc/ollama.conf | grep -v '#' OLLAMA_NUM_PARALLEL=2 OLLAMA_MAX_LOADED_MODELS=2 HSA_OVERRIDE_GFX_VERSION=11.0.0 ROCR_VISIBLE_DEVICES=0,1 HIP_VISIBLE_DEVICES=0,1 HOME=/var/lib/ollama GIN_MODE=debug OLLAMA_DEBUG=1 AMD_SERIALIZE_KERNEL=3 OLLAMA_LLM_LIBRARY=rocm_v60002 OLLAMA_ORIGINS=* OLLAMA_HOST=0.0.0.0 ``` Another configuration has one more line ```OLLAMA_MAX_VRAM=49392123904``` If you need more information and tests, like systemd log or tests, I will upload them.
GiteaMirror commented 2026-05-09 09:17:24 -05:00
Author
Owner
Copy Link

@hamkido commented on GitHub (May 31, 2024):

seems caused by other reason.

<!-- gh-comment-id:2141319384 --> @hamkido commented on GitHub (May 31, 2024): seems caused by other reason.
GiteaMirror commented 2026-05-09 09:17:24 -05:00
Author
Owner
Copy Link

@hamkido commented on GitHub (Jun 5, 2024):

Just as a reference for anyone who runs into this error, I quickly fixed this by customizing the totalMemory var in gpu/amd_linux.go file and rebuilding the package. I was short on time and this wasn't a very elegant approach. But it just works.
Also /etc/ollama.conf OLLAMA_MAX_VRAM do not need config any more.

<!-- gh-comment-id:2148986748 --> @hamkido commented on GitHub (Jun 5, 2024): Just as a reference for anyone who runs into this error, I quickly fixed this by customizing the totalMemory var in gpu/amd_linux.go file and rebuilding the package. I was short on time and this wasn't a very elegant approach. But it just works. Also /etc/ollama.conf OLLAMA_MAX_VRAM do not need config any more.
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#80666
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?