github-starred/ollama
2
0
Fork 0
mirror of https://github.com/ollama/ollama.git synced 2026-05-08 17:49:24 -05:00
Code Issues 14.6k Packages Projects Releases 122 Wiki Activity

[GH-ISSUE #5547] Mixtral 8x22b inference output is empty or gibberish #49978

New Issue
Open
opened 2026-04-28 13:37:34 -05:00 by GiteaMirror · 2 comments
GiteaMirror commented 2026-04-28 13:37:34 -05:00
Owner
Copy Link

Originally created by @PLK2 on GitHub (Jul 8, 2024).
Original GitHub issue: https://github.com/ollama/ollama/issues/5547

Originally assigned to: @dhiltgen on GitHub.

What is the issue?

Mixtral 8x22b instruct outputs are either empty or gibberish.
I have tried various quantizations: q4, q4_k_m, q5, etc. All seem problematic.
Other models (e.g., llama3, command-r, Mistral, etc) work fine.

Running 2x Nvidia 3090 GPUs = 48gb vram, 4.9 GHz AMD Ryzen 9 5950X, 128gb ram.

OS

Linux

GPU

Nvidia

CPU

AMD

Ollama version

0.1.48

Originally created by @PLK2 on GitHub (Jul 8, 2024). Original GitHub issue: https://github.com/ollama/ollama/issues/5547 Originally assigned to: @dhiltgen on GitHub. ### What is the issue? Mixtral 8x22b instruct outputs are either empty or gibberish. I have tried various quantizations: q4, q4_k_m, q5, etc. All seem problematic. Other models (e.g., llama3, command-r, Mistral, etc) work fine. Running 2x Nvidia 3090 GPUs = 48gb vram, 4.9 GHz AMD Ryzen 9 5950X, 128gb ram. ### OS Linux ### GPU Nvidia ### CPU AMD ### Ollama version 0.1.48
GiteaMirror added the memorybugnvidia labels 2026-04-28 13:37:34 -05:00
GiteaMirror commented 2026-04-28 13:37:35 -05:00
Author
Owner
Copy Link

@PLK2 commented on GitHub (Jul 8, 2024):

Have also tested on LM Studio and it works fine.

<!-- gh-comment-id:2215050415 --> @PLK2 commented on GitHub (Jul 8, 2024): Have also tested on LM Studio and it works fine.
GiteaMirror commented 2026-04-28 13:37:36 -05:00
Author
Owner
Copy Link

@dhiltgen commented on GitHub (Jul 24, 2024):

I don't have an identical setup, but on a dual 3060 setup, the model loads and works, albeit slowly. (1.64 tps)

My suspicion is we're off slightly in our memory predictions and loading ~1 too many layers leading to some sort of corruption. Can you share your server log so we can see the memory predictions and layer counts? You can also experiment and try getting it to load with fewer layers and see if that gets it working properly.

% curl http://localhost:11434/api/generate -d '{
  "model": "mixtral:8x22b",
  "prompt": "hello",
  "stream": false, "options": {"num_gpu": 12 }
}'

Please make sure to upgrade to the latest version too.

<!-- gh-comment-id:2246634204 --> @dhiltgen commented on GitHub (Jul 24, 2024): I don't have an identical setup, but on a dual 3060 setup, the model loads and works, albeit slowly. (1.64 tps) My suspicion is we're off slightly in our memory predictions and loading ~1 too many layers leading to some sort of corruption. Can you share your server log so we can see the memory predictions and layer counts? You can also experiment and try getting it to load with fewer layers and see if that gets it working properly. ``` % curl http://localhost:11434/api/generate -d '{ "model": "mixtral:8x22b", "prompt": "hello", "stream": false, "options": {"num_gpu": 12 } }' ``` Please make sure to upgrade to the latest version too.
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 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#49978
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?