github-starred/ollama
2
0
Fork 0
mirror of https://github.com/ollama/ollama.git synced 2026-05-06 16:11:34 -05:00
Code Issues 14.5k Packages Projects Releases 120 Wiki Activity

[GH-ISSUE #5589] OOM crash loading codegeex4 on 4G GPU #65528

New Issue
Closed
opened 2026-05-03 21:35:35 -05:00 by GiteaMirror · 5 comments
GiteaMirror commented 2026-05-03 21:35:35 -05:00
Owner
Copy Link

Originally created by @chandan0000 on GitHub (Jul 10, 2024).
Original GitHub issue: https://github.com/ollama/ollama/issues/5589

Originally assigned to: @dhiltgen on GitHub.

What is the issue?

image
other work fine ollama3, gemma2 but i am running codegeex4 model throw error
how i can resolve this issue
my system config
image

OS

Linux

GPU

Nvidia

CPU

Intel

Ollama version

ollama version is 0.2.1

Originally created by @chandan0000 on GitHub (Jul 10, 2024). Original GitHub issue: https://github.com/ollama/ollama/issues/5589 Originally assigned to: @dhiltgen on GitHub. ### What is the issue? ![image](https://github.com/ollama/ollama/assets/60910265/dc193bf7-9418-469c-ab16-d342cf792c67) other work fine ollama3, gemma2 but i am running codegeex4 model throw error how i can resolve this issue my system config ![image](https://github.com/ollama/ollama/assets/60910265/edaf8a76-0274-4d0d-a7dc-7f5e59519ed7) ### OS Linux ### GPU Nvidia ### CPU Intel ### Ollama version ollama version is 0.2.1
GiteaMirror added the memorybug labels 2026-05-03 21:35:36 -05:00
GiteaMirror commented 2026-05-03 21:35:41 -05:00
Author
Owner
Copy Link

@darwinvelez58 commented on GitHub (Jul 10, 2024):

I have this issue Issue!

imagen

running with x3 7900 XTX, doesn't have any sense!

imagen

<!-- gh-comment-id:2221595402 --> @darwinvelez58 commented on GitHub (Jul 10, 2024): I have this issue Issue! ![imagen](https://github.com/ollama/ollama/assets/118543481/81932c5d-2316-4943-a755-d26a77e8341d) running with x3 7900 XTX, doesn't have any sense! ![imagen](https://github.com/ollama/ollama/assets/118543481/7d83d54c-cc4a-4cb2-95b9-52beae56309e)
GiteaMirror commented 2026-05-03 21:35:43 -05:00
Author
Owner
Copy Link

@darwinvelez58 commented on GitHub (Jul 10, 2024):

This is happening after 1.44rocm release!

<!-- gh-comment-id:2221596008 --> @darwinvelez58 commented on GitHub (Jul 10, 2024): This is happening after 1.44rocm release!
GiteaMirror commented 2026-05-03 21:35:46 -05:00
Author
Owner
Copy Link

@dhiltgen commented on GitHub (Jul 23, 2024):

@chandan0000 I believe your GPU is a 4G card. I have been able to reproduce the crash on another 4G card, and it looks like we're over-allocating 1 too many layers. In my setup, we tried to load 28/41 layers and that crashed. If I specify 27 layers it successfully loads the model.

% curl http://localhost:11434/api/generate -d '{
  "model": "codegeex4",
  "prompt": "hello?",
  "stream": false, "options": {"num_gpu": 27 }
}'
{"model":"codegeex4","created_at":"2024-07-23T23:33:52.341542262Z","response":"Hello! How can I assist you today?","done":true,"done_reason":"stop","context":[151331,151333,151336,198,14978,30,151337,198,9703,0,2585,646,358,7789,498,3351,30],"total_duration":13563765176,"load_duration":7871846153,"prompt_eval_count":8,"prompt_eval_duration":2509817000,"eval_count":10,"eval_duration":3180908000}%

% nvidia-smi
Tue Jul 23 23:36:20 2024
+-----------------------------------------------------------------------------------------+
| NVIDIA-SMI 555.42.02              Driver Version: 555.42.02      CUDA Version: 12.5     |
|-----------------------------------------+------------------------+----------------------+
| GPU  Name                 Persistence-M | Bus-Id          Disp.A | Volatile Uncorr. ECC |
| Fan  Temp   Perf          Pwr:Usage/Cap |           Memory-Usage | GPU-Util  Compute M. |
|                                         |                        |               MIG M. |
|=========================================+========================+======================|
|   0  NVIDIA GeForce GT 1030         Off |   00000000:01:00.0 Off |                  N/A |
| 35%   34C    P8             N/A /   19W |    3948MiB /   4096MiB |      0%      Default |
|                                         |                        |                  N/A |
+-----------------------------------------+------------------------+----------------------+

+-----------------------------------------------------------------------------------------+
| Processes:                                                                              |
|  GPU   GI   CI        PID   Type   Process name                              GPU Memory |
|        ID   ID                                                               Usage      |
|=========================================================================================|
|    0   N/A  N/A    202781      C   ...unners/cuda_v11/ollama_llama_server       3944MiB |
+-----------------------------------------------------------------------------------------+
<!-- gh-comment-id:2246516080 --> @dhiltgen commented on GitHub (Jul 23, 2024): @chandan0000 I believe your GPU is a 4G card. I have been able to reproduce the crash on another 4G card, and it looks like we're over-allocating 1 too many layers. In my setup, we tried to load 28/41 layers and that crashed. If I specify 27 layers it successfully loads the model. ``` % curl http://localhost:11434/api/generate -d '{ "model": "codegeex4", "prompt": "hello?", "stream": false, "options": {"num_gpu": 27 } }' {"model":"codegeex4","created_at":"2024-07-23T23:33:52.341542262Z","response":"Hello! How can I assist you today?","done":true,"done_reason":"stop","context":[151331,151333,151336,198,14978,30,151337,198,9703,0,2585,646,358,7789,498,3351,30],"total_duration":13563765176,"load_duration":7871846153,"prompt_eval_count":8,"prompt_eval_duration":2509817000,"eval_count":10,"eval_duration":3180908000}% ``` ``` % nvidia-smi Tue Jul 23 23:36:20 2024 +-----------------------------------------------------------------------------------------+ | NVIDIA-SMI 555.42.02 Driver Version: 555.42.02 CUDA Version: 12.5 | |-----------------------------------------+------------------------+----------------------+ | GPU Name Persistence-M | Bus-Id Disp.A | Volatile Uncorr. ECC | | Fan Temp Perf Pwr:Usage/Cap | Memory-Usage | GPU-Util Compute M. | | | | MIG M. | |=========================================+========================+======================| | 0 NVIDIA GeForce GT 1030 Off | 00000000:01:00.0 Off | N/A | | 35% 34C P8 N/A / 19W | 3948MiB / 4096MiB | 0% Default | | | | N/A | +-----------------------------------------+------------------------+----------------------+ +-----------------------------------------------------------------------------------------+ | Processes: | | GPU GI CI PID Type Process name GPU Memory | | ID ID Usage | |=========================================================================================| | 0 N/A N/A 202781 C ...unners/cuda_v11/ollama_llama_server 3944MiB | +-----------------------------------------------------------------------------------------+ ```
GiteaMirror commented 2026-05-03 21:35:49 -05:00
Author
Owner
Copy Link

@dhiltgen commented on GitHub (Jul 23, 2024):

@darwinvelez58 your issue is unrelated to the OOM crash of codegeex4 - we'll track the 3x ROCm bug via #5629

<!-- gh-comment-id:2246520768 --> @dhiltgen commented on GitHub (Jul 23, 2024): @darwinvelez58 your issue is unrelated to the OOM crash of codegeex4 - we'll track the 3x ROCm bug via #5629
GiteaMirror commented 2026-05-03 21:35:52 -05:00
Author
Owner
Copy Link

@dhiltgen commented on GitHub (Jul 23, 2024):

This was fixed in v0.2.2. Confirmed on 0.2.8 it loads 23 layers and doesn't crash on the 4G GPU.

<!-- gh-comment-id:2246545860 --> @dhiltgen commented on GitHub (Jul 23, 2024): This was fixed in v0.2.2. Confirmed on 0.2.8 it loads 23 layers and doesn't crash on the 4G GPU.
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 memory
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#65528
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?