github-starred/ollama
2
0
Fork 0
mirror of https://github.com/ollama/ollama.git synced 2026-05-07 16:40:08 -05:00
Code Issues 14.5k Packages Projects Releases 120 Wiki Activity

[GH-ISSUE #10685] v0.7.0 crashes when using snowflake-arctic-embed2 #32783

New Issue
Closed
opened 2026-04-22 14:37:08 -05:00 by GiteaMirror · 6 comments
GiteaMirror commented 2026-04-22 14:37:08 -05:00
Owner
Copy Link

Originally created by @Blumlaut on GitHub (May 13, 2025).
Original GitHub issue: https://github.com/ollama/ollama/issues/10685

Originally assigned to: @jmorganca on GitHub.

What is the issue?

When using Ollama 0.7.0-rc0 with qwen3:8b and open-webui knowledge (using snowflake-arctic-embed2 as the embedding model), ollama crashes and restarts, works fine when not using knowledge.

0.6.9 works perfectly fine using the same parameters

Relevant log output

https://pastebin.com/ike4qdHF

OS

Docker (Debian Bullseye in a Proxmox KVM)

GPU

Nvidia RTX 3060 12GB LHR (Driver Version 535.216.03)

CPU

AMD Ryzen 5 5600G

Ollama version

v0.7.0-rc0

Originally created by @Blumlaut on GitHub (May 13, 2025). Original GitHub issue: https://github.com/ollama/ollama/issues/10685 Originally assigned to: @jmorganca on GitHub. ### What is the issue? When using Ollama 0.7.0-rc0 with `qwen3:8b` and open-webui knowledge (using `snowflake-arctic-embed2` as the embedding model), ollama crashes and restarts, works fine when not using knowledge. 0.6.9 works perfectly fine using the same parameters ### Relevant log output ```shell https://pastebin.com/ike4qdHF ``` ### OS Docker (Debian Bullseye in a Proxmox KVM) ### GPU Nvidia RTX 3060 12GB LHR (Driver Version 535.216.03) ### CPU AMD Ryzen 5 5600G ### Ollama version v0.7.0-rc0
GiteaMirror added the bug label 2026-04-22 14:37:08 -05:00
GiteaMirror commented 2026-04-22 14:37:10 -05:00
Author
Owner
Copy Link

@rick-github commented on GitHub (May 13, 2025):

Seems to be a problem with snowflake-arctic-embed2:latest. None of the other embedding models in my library are affected.

llama-vocab.cpp:1472: GGML_ASSERT(pc_type == GGUF_TYPE_INT8 || pc_type == GGUF_TYPE_UINT8) failed
/bin/ollama(+0x1057118)[0x5584c1e66118]
/bin/ollama(ggml_abort+0x136)[0x5584c1e66436]
/bin/ollama(_ZN11llama_vocab4impl4loadER18llama_model_loaderRK6LLM_KV+0x2cf3)[0x5584c1f73dd3]
/bin/ollama(_ZN11llama_model10load_vocabER18llama_model_loader+0x30)[0x5584c1f0b1c0]
/bin/ollama(+0x116c96a)[0x5584c1f7b96a]
/bin/ollama(llama_model_load_from_file+0xa2)[0x5584c1f7c152]
/bin/ollama(_cgo_e5c1b5ac37b2_Cfunc_llama_model_load_from_file+0x4b)[0x5584c1e4043b]
/bin/ollama(+0x3a2461)[0x5584c11b1461]
SIGABRT: abort
PC=0x7fe1de65d00b m=11 sigcode=18446744073709551610
signal arrived during cgo execution
 
goroutine 23 gp=0xc000103340 m=11 mp=0xc000680008 [syscall]:
runtime.cgocall(0x5584c1e403f0, 0xc000698a10)
	runtime/cgocall.go:167 +0x4b fp=0xc0006989e8 sp=0xc0006989b0 pc=0x5584c11a6dab
github.com/ollama/ollama/llama._Cfunc_llama_model_load_from_file(0x7fe1480c11f0, {0x0, 0x0, 0x0, 0x1, 0x0, 0x0, 0x0, 0x0, 0x0, ...})
<!-- gh-comment-id:2876097857 --> @rick-github commented on GitHub (May 13, 2025): Seems to be a problem with snowflake-arctic-embed2:latest. None of the other embedding models in my library are affected. ``` llama-vocab.cpp:1472: GGML_ASSERT(pc_type == GGUF_TYPE_INT8 || pc_type == GGUF_TYPE_UINT8) failed /bin/ollama(+0x1057118)[0x5584c1e66118] /bin/ollama(ggml_abort+0x136)[0x5584c1e66436] /bin/ollama(_ZN11llama_vocab4impl4loadER18llama_model_loaderRK6LLM_KV+0x2cf3)[0x5584c1f73dd3] /bin/ollama(_ZN11llama_model10load_vocabER18llama_model_loader+0x30)[0x5584c1f0b1c0] /bin/ollama(+0x116c96a)[0x5584c1f7b96a] /bin/ollama(llama_model_load_from_file+0xa2)[0x5584c1f7c152] /bin/ollama(_cgo_e5c1b5ac37b2_Cfunc_llama_model_load_from_file+0x4b)[0x5584c1e4043b] /bin/ollama(+0x3a2461)[0x5584c11b1461] SIGABRT: abort PC=0x7fe1de65d00b m=11 sigcode=18446744073709551610 signal arrived during cgo execution goroutine 23 gp=0xc000103340 m=11 mp=0xc000680008 [syscall]: runtime.cgocall(0x5584c1e403f0, 0xc000698a10) runtime/cgocall.go:167 +0x4b fp=0xc0006989e8 sp=0xc0006989b0 pc=0x5584c11a6dab github.com/ollama/ollama/llama._Cfunc_llama_model_load_from_file(0x7fe1480c11f0, {0x0, 0x0, 0x0, 0x1, 0x0, 0x0, 0x0, 0x0, 0x0, ...}) ```
GiteaMirror commented 2026-04-22 14:37:10 -05:00
Author
Owner
Copy Link

@Blumlaut commented on GitHub (May 13, 2025):

Seems to be a problem with snowflake-arctic-embed2:latest. None of the other embedding models in my library are affected.

good to know! I didn't have time to test a different model but it seems you're right, i updated the issue title accordingly.

<!-- gh-comment-id:2876183156 --> @Blumlaut commented on GitHub (May 13, 2025): > Seems to be a problem with snowflake-arctic-embed2:latest. None of the other embedding models in my library are affected. good to know! I didn't have time to test a different model but it seems you're right, i updated the issue title accordingly.
GiteaMirror commented 2026-04-22 14:37:10 -05:00
Author
Owner
Copy Link

@rick-github commented on GitHub (May 13, 2025):

Appears to be a result of 10d2af0eaa in llama.cpp, where an assert was added for the type of tokenizer.ggml.precompiled_charsmap:

c7f4ae7b9c/llama/llama.cpp/src/llama-vocab.cpp (L1469-L1472)

The problem seems to be that in snowflake-arctic-embed2, this is declared as [STRING]:

$ dumpgguf --no-tensors /root/.ollama/models/blobs/sha256-8c625c9569c3c799f5f9595b5a141f91d224233055608189d66746347c14e613 | grep tokenizer.ggml.precompiled_charsmap
     34: [STRING]   |   316720 | tokenizer.ggml.precompiled_charsmap = ['A', 'L', 'Q', 'C', 'A', 'A', ...]

whereas in other models it's [UINT8], eg granite-embedding:278m:

$ dumpgguf --no-tensors /root/.ollama/models/blobs/sha256-a658c903bf83a9d84fe1b9869874229ddfdfc1c88b04f958f17dd24e243d262c | grep tokenizer.ggml.precompiled_charsmap
     30: [UINT8]    |   237539 | tokenizer.ggml.precompiled_charsmap = [0, 180, 2, 0, 0, 132, ...]
<!-- gh-comment-id:2876368773 --> @rick-github commented on GitHub (May 13, 2025): Appears to be a result of https://github.com/ggml-org/llama.cpp/commit/10d2af0eaa0aafd7c6577b279dfa5221ff44a63f in llama.cpp, where an assert was added for the type of `tokenizer.ggml.precompiled_charsmap`: https://github.com/ollama/ollama/blob/c7f4ae7b9c8976b4d50c59eb87e9582ea9c5c82f/llama/llama.cpp/src/llama-vocab.cpp#L1469-L1472 The problem seems to be that in snowflake-arctic-embed2, this is declared as [STRING]: ```console $ dumpgguf --no-tensors /root/.ollama/models/blobs/sha256-8c625c9569c3c799f5f9595b5a141f91d224233055608189d66746347c14e613 | grep tokenizer.ggml.precompiled_charsmap 34: [STRING] | 316720 | tokenizer.ggml.precompiled_charsmap = ['A', 'L', 'Q', 'C', 'A', 'A', ...] ``` whereas in other models it's [UINT8], eg granite-embedding:278m: ```console $ dumpgguf --no-tensors /root/.ollama/models/blobs/sha256-a658c903bf83a9d84fe1b9869874229ddfdfc1c88b04f958f17dd24e243d262c | grep tokenizer.ggml.precompiled_charsmap 30: [UINT8] | 237539 | tokenizer.ggml.precompiled_charsmap = [0, 180, 2, 0, 0, 132, ...] ```
GiteaMirror commented 2026-04-22 14:37:10 -05:00
Author
Owner
Copy Link

@jmorganca commented on GitHub (May 13, 2025):

Closing for https://github.com/ollama/ollama/issues/10685

<!-- gh-comment-id:2877379352 --> @jmorganca commented on GitHub (May 13, 2025): Closing for https://github.com/ollama/ollama/issues/10685
GiteaMirror commented 2026-04-22 14:37:10 -05:00
Author
Owner
Copy Link

@ProjectMoon commented on GitHub (May 13, 2025):

Err, isn't this #10685? Lol.

<!-- gh-comment-id:2877439027 --> @ProjectMoon commented on GitHub (May 13, 2025): Err, isn't this #10685? Lol.
GiteaMirror commented 2026-04-22 14:37:11 -05:00
Author
Owner
Copy Link

@jmorganca commented on GitHub (May 13, 2025):

Oops!

<!-- gh-comment-id:2877675143 --> @jmorganca commented on GitHub (May 13, 2025): Oops!
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#32783
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?