github-starred/ollama
2
0
Fork 0
mirror of https://github.com/ollama/ollama.git synced 2026-05-23 06:01:37 -05:00
Code Issues 16.9k Packages Projects Releases 122 Wiki Activity

[GH-ISSUE #9511] snowflake-arctic-embed2 causes GGML_ASSERT(ctx->kv[key_id].get_type() != GGUF_TYPE_STRING) failed #6199

New Issue
Closed
opened 2026-04-12 17:35:06 -05:00 by GiteaMirror · 9 comments
GiteaMirror commented 2026-04-12 17:35:06 -05:00
Owner
Copy Link

Originally created by @jmorganca on GitHub (Mar 5, 2025).
Original GitHub issue: https://github.com/ollama/ollama/issues/9511

Originally assigned to: @jmorganca on GitHub.

What is the issue?

GGUF parsing code is broken as of the last llama.cpp vendor code commit for models such as snowflake-arctic-embed2

Logs:

2025-03-05 09:00:48 time=2025-03-05T07:00:48.651Z level=INFO source=runner.go:934 msg=system info="CPU : LLAMAFILE = 1 | CUDA : ARCHS = 500,600,610,700,750,800,860,870,890,900,1200 | USE_GRAPHS = 1 | PEER_MAX_BATCH_SIZE = 128 | CPU : SSE3 = 1 | SSSE3 = 1 | AVX = 1 | AVX_VNNI = 1 | AVX2 = 1 | F16C = 1 | FMA = 1 | LLAMAFILE = 1 | cgo(gcc)" threads=10
2025-03-05 09:00:48 time=2025-03-05T07:00:48.672Z level=INFO source=runner.go:992 msg="Server listening on 127.0.0.1:37753"
2025-03-05 09:00:48 time=2025-03-05T07:00:48.743Z level=INFO source=server.go:591 msg="waiting for server to become available" status="llm server loading model"
2025-03-05 09:00:48 llama_model_load_from_file_impl: using device CUDA0 (NVIDIA GeForce RTX 4080 Laptop GPU) - 11047 MiB free
2025-03-05 09:00:48 llama_model_loader: loaded meta data with 34 key-value pairs and 389 tensors from /root/.ollama/models/blobs/sha256-8c625c9569c3c799f5f9595b5a141f91d224233055608189d66746347c14e613 (version GGUF V3 (latest))
2025-03-05 09:00:48 llama_model_loader: Dumping metadata keys/values. Note: KV overrides do not apply in this output.
2025-03-05 09:00:48 llama_model_loader: - kv   0:                       general.architecture str              = bert
2025-03-05 09:00:48 llama_model_loader: - kv   1:                               general.type str              = model
2025-03-05 09:00:48 llama_model_loader: - kv   2:                         general.size_label str              = 567M
2025-03-05 09:00:48 llama_model_loader: - kv   3:                            general.license str              = apache-2.0
2025-03-05 09:00:48 llama_model_loader: - kv   4:                               general.tags arr[str,8]       = ["sentence-transformers", "feature-ex...
2025-03-05 09:00:48 llama_model_loader: - kv   5:                          general.languages arr[str,74]      = ["af", "ar", "az", "be", "bg", "bn", ...
2025-03-05 09:00:48 llama_model_loader: - kv   6:                           bert.block_count u32              = 24
2025-03-05 09:00:48 llama_model_loader: - kv   7:                        bert.context_length u32              = 8192
2025-03-05 09:00:48 llama_model_loader: - kv   8:                      bert.embedding_length u32              = 1024
2025-03-05 09:00:48 llama_model_loader: - kv   9:                   bert.feed_forward_length u32              = 4096
2025-03-05 09:00:48 llama_model_loader: - kv  10:                  bert.attention.head_count u32              = 16
2025-03-05 09:00:48 llama_model_loader: - kv  11:          bert.attention.layer_norm_epsilon f32              = 0.000010
2025-03-05 09:00:48 llama_model_loader: - kv  12:                          general.file_type u32              = 1
2025-03-05 09:00:48 llama_model_loader: - kv  13:                      bert.attention.causal bool             = false
2025-03-05 09:00:48 llama_model_loader: - kv  14:                          bert.pooling_type u32              = 2
2025-03-05 09:00:48 llama_model_loader: - kv  15:                       tokenizer.ggml.model str              = t5
2025-03-05 09:00:48 llama_model_loader: - kv  16:                         tokenizer.ggml.pre str              = default
2025-03-05 09:00:48 llama_model_loader: - kv  17:                      tokenizer.ggml.tokens arr[str,250002]  = ["<s>", "<pad>", "</s>", "<unk>", ","...
2025-03-05 09:00:48 llama_model_loader: - kv  18:                  tokenizer.ggml.token_type arr[i32,250002]  = [3, 3, 3, 3, 1, 1, 1, 1, 1, 1, 1, 1, ...
2025-03-05 09:00:48 llama_model_loader: - kv  19:                      tokenizer.ggml.scores arr[f32,250002]  = [-10000.000000, -10000.000000, -10000...
2025-03-05 09:00:48 llama_model_loader: - kv  20:               tokenizer.ggml.add_bos_token bool             = true
2025-03-05 09:00:48 llama_model_loader: - kv  21:               tokenizer.ggml.add_eos_token bool             = true
2025-03-05 09:00:48 llama_model_loader: - kv  22:            tokenizer.ggml.token_type_count u32              = 1
2025-03-05 09:00:48 llama_model_loader: - kv  23:                tokenizer.ggml.bos_token_id u32              = 0
2025-03-05 09:00:48 llama_model_loader: - kv  24:                tokenizer.ggml.eos_token_id u32              = 2
2025-03-05 09:00:48 llama_model_loader: - kv  25:            tokenizer.ggml.unknown_token_id u32              = 3
2025-03-05 09:00:48 llama_model_loader: - kv  26:          tokenizer.ggml.seperator_token_id u32              = 2
2025-03-05 09:00:48 llama_model_loader: - kv  27:            tokenizer.ggml.padding_token_id u32              = 1
2025-03-05 09:00:48 llama_model_loader: - kv  28:                tokenizer.ggml.cls_token_id u32              = 0
2025-03-05 09:00:48 llama_model_loader: - kv  29:               tokenizer.ggml.mask_token_id u32              = 250001
2025-03-05 09:00:48 llama_model_loader: - kv  30:        tokenizer.ggml.precompiled_charsmap arr[str,316720]  = ["A", "L", "Q", "C", "A", "A", "C", "...
2025-03-05 09:00:48 llama_model_loader: - kv  31:    tokenizer.ggml.remove_extra_whitespaces bool             = true
2025-03-05 09:00:48 llama_model_loader: - kv  32:            tokenizer.ggml.add_space_prefix bool             = true
2025-03-05 09:00:48 llama_model_loader: - kv  33:               general.quantization_version u32              = 2
2025-03-05 09:00:48 llama_model_loader: - type  f32:  244 tensors
2025-03-05 09:00:48 llama_model_loader: - type  f16:  145 tensors
2025-03-05 09:00:48 print_info: file format = GGUF V3 (latest)
2025-03-05 09:00:48 print_info: file type   = F16
2025-03-05 09:00:48 print_info: file size   = 1.07 GiB (16.25 BPW) 
2025-03-05 09:00:48 gguf.cpp:780: GGML_ASSERT(ctx->kv[key_id].get_type() != GGUF_TYPE_STRING) failed

Relevant log output

OS

No response

GPU

No response

CPU

No response

Ollama version

No response

Originally created by @jmorganca on GitHub (Mar 5, 2025). Original GitHub issue: https://github.com/ollama/ollama/issues/9511 Originally assigned to: @jmorganca on GitHub. ### What is the issue? GGUF parsing code is broken as of the last llama.cpp vendor code commit for models such as `snowflake-arctic-embed2` Logs: ``` 2025-03-05 09:00:48 time=2025-03-05T07:00:48.651Z level=INFO source=runner.go:934 msg=system info="CPU : LLAMAFILE = 1 | CUDA : ARCHS = 500,600,610,700,750,800,860,870,890,900,1200 | USE_GRAPHS = 1 | PEER_MAX_BATCH_SIZE = 128 | CPU : SSE3 = 1 | SSSE3 = 1 | AVX = 1 | AVX_VNNI = 1 | AVX2 = 1 | F16C = 1 | FMA = 1 | LLAMAFILE = 1 | cgo(gcc)" threads=10 2025-03-05 09:00:48 time=2025-03-05T07:00:48.672Z level=INFO source=runner.go:992 msg="Server listening on 127.0.0.1:37753" 2025-03-05 09:00:48 time=2025-03-05T07:00:48.743Z level=INFO source=server.go:591 msg="waiting for server to become available" status="llm server loading model" 2025-03-05 09:00:48 llama_model_load_from_file_impl: using device CUDA0 (NVIDIA GeForce RTX 4080 Laptop GPU) - 11047 MiB free 2025-03-05 09:00:48 llama_model_loader: loaded meta data with 34 key-value pairs and 389 tensors from /root/.ollama/models/blobs/sha256-8c625c9569c3c799f5f9595b5a141f91d224233055608189d66746347c14e613 (version GGUF V3 (latest)) 2025-03-05 09:00:48 llama_model_loader: Dumping metadata keys/values. Note: KV overrides do not apply in this output. 2025-03-05 09:00:48 llama_model_loader: - kv 0: general.architecture str = bert 2025-03-05 09:00:48 llama_model_loader: - kv 1: general.type str = model 2025-03-05 09:00:48 llama_model_loader: - kv 2: general.size_label str = 567M 2025-03-05 09:00:48 llama_model_loader: - kv 3: general.license str = apache-2.0 2025-03-05 09:00:48 llama_model_loader: - kv 4: general.tags arr[str,8] = ["sentence-transformers", "feature-ex... 2025-03-05 09:00:48 llama_model_loader: - kv 5: general.languages arr[str,74] = ["af", "ar", "az", "be", "bg", "bn", ... 2025-03-05 09:00:48 llama_model_loader: - kv 6: bert.block_count u32 = 24 2025-03-05 09:00:48 llama_model_loader: - kv 7: bert.context_length u32 = 8192 2025-03-05 09:00:48 llama_model_loader: - kv 8: bert.embedding_length u32 = 1024 2025-03-05 09:00:48 llama_model_loader: - kv 9: bert.feed_forward_length u32 = 4096 2025-03-05 09:00:48 llama_model_loader: - kv 10: bert.attention.head_count u32 = 16 2025-03-05 09:00:48 llama_model_loader: - kv 11: bert.attention.layer_norm_epsilon f32 = 0.000010 2025-03-05 09:00:48 llama_model_loader: - kv 12: general.file_type u32 = 1 2025-03-05 09:00:48 llama_model_loader: - kv 13: bert.attention.causal bool = false 2025-03-05 09:00:48 llama_model_loader: - kv 14: bert.pooling_type u32 = 2 2025-03-05 09:00:48 llama_model_loader: - kv 15: tokenizer.ggml.model str = t5 2025-03-05 09:00:48 llama_model_loader: - kv 16: tokenizer.ggml.pre str = default 2025-03-05 09:00:48 llama_model_loader: - kv 17: tokenizer.ggml.tokens arr[str,250002] = ["<s>", "<pad>", "</s>", "<unk>", ","... 2025-03-05 09:00:48 llama_model_loader: - kv 18: tokenizer.ggml.token_type arr[i32,250002] = [3, 3, 3, 3, 1, 1, 1, 1, 1, 1, 1, 1, ... 2025-03-05 09:00:48 llama_model_loader: - kv 19: tokenizer.ggml.scores arr[f32,250002] = [-10000.000000, -10000.000000, -10000... 2025-03-05 09:00:48 llama_model_loader: - kv 20: tokenizer.ggml.add_bos_token bool = true 2025-03-05 09:00:48 llama_model_loader: - kv 21: tokenizer.ggml.add_eos_token bool = true 2025-03-05 09:00:48 llama_model_loader: - kv 22: tokenizer.ggml.token_type_count u32 = 1 2025-03-05 09:00:48 llama_model_loader: - kv 23: tokenizer.ggml.bos_token_id u32 = 0 2025-03-05 09:00:48 llama_model_loader: - kv 24: tokenizer.ggml.eos_token_id u32 = 2 2025-03-05 09:00:48 llama_model_loader: - kv 25: tokenizer.ggml.unknown_token_id u32 = 3 2025-03-05 09:00:48 llama_model_loader: - kv 26: tokenizer.ggml.seperator_token_id u32 = 2 2025-03-05 09:00:48 llama_model_loader: - kv 27: tokenizer.ggml.padding_token_id u32 = 1 2025-03-05 09:00:48 llama_model_loader: - kv 28: tokenizer.ggml.cls_token_id u32 = 0 2025-03-05 09:00:48 llama_model_loader: - kv 29: tokenizer.ggml.mask_token_id u32 = 250001 2025-03-05 09:00:48 llama_model_loader: - kv 30: tokenizer.ggml.precompiled_charsmap arr[str,316720] = ["A", "L", "Q", "C", "A", "A", "C", "... 2025-03-05 09:00:48 llama_model_loader: - kv 31: tokenizer.ggml.remove_extra_whitespaces bool = true 2025-03-05 09:00:48 llama_model_loader: - kv 32: tokenizer.ggml.add_space_prefix bool = true 2025-03-05 09:00:48 llama_model_loader: - kv 33: general.quantization_version u32 = 2 2025-03-05 09:00:48 llama_model_loader: - type f32: 244 tensors 2025-03-05 09:00:48 llama_model_loader: - type f16: 145 tensors 2025-03-05 09:00:48 print_info: file format = GGUF V3 (latest) 2025-03-05 09:00:48 print_info: file type = F16 2025-03-05 09:00:48 print_info: file size = 1.07 GiB (16.25 BPW) 2025-03-05 09:00:48 gguf.cpp:780: GGML_ASSERT(ctx->kv[key_id].get_type() != GGUF_TYPE_STRING) failed ``` ### Relevant log output ```shell ``` ### OS _No response_ ### GPU _No response_ ### CPU _No response_ ### Ollama version _No response_
GiteaMirror added the bug label 2026-04-12 17:35:06 -05:00
GiteaMirror commented 2026-04-12 17:35:08 -05:00
Author
Owner
Copy Link

@Fade78 commented on GitHub (Mar 6, 2025):

got this too.

<!-- gh-comment-id:2703140052 --> @Fade78 commented on GitHub (Mar 6, 2025): got this too.
GiteaMirror commented 2026-04-12 17:35:08 -05:00
Author
Owner
Copy Link

@Fade78 commented on GitHub (Mar 6, 2025):

Fixed in 0.6.0.

old instructions:

How to rollback on docker

Waiting the fix you can replace with the image sha256:fe6cfe49855d66ff55c88022c7e0c92071a00803aff4e6f9aef884829174477f aka fe6cfe49855d. You can run a container simply by using its id instead of the image name. You can verify that you have this image with docker image ls --no-trunc ollama/ollama or docker image ls ollama/ollama.

<!-- gh-comment-id:2703174242 --> @Fade78 commented on GitHub (Mar 6, 2025): # Fixed in 0.6.0. old instructions: # How to rollback on docker Waiting the fix you can replace with the image `sha256:fe6cfe49855d66ff55c88022c7e0c92071a00803aff4e6f9aef884829174477f` aka `fe6cfe49855d`. You can `run` a container simply by using its id instead of the image name. You can verify that you have this image with `docker image ls --no-trunc ollama/ollama` or `docker image ls ollama/ollama`.
GiteaMirror commented 2026-04-12 17:35:09 -05:00
Author
Owner
Copy Link

@Carbaz commented on GitHub (Mar 6, 2025):

ggml-cpu.c:8624: GGML_ASSERT(i01 >= 0 && i01 < ne01) failed

When tried using snowflake-arctic-embed instead of snowflake-arctic-embed2

<!-- gh-comment-id:2703886595 --> @Carbaz commented on GitHub (Mar 6, 2025): `ggml-cpu.c:8624: GGML_ASSERT(i01 >= 0 && i01 < ne01) failed` When tried using `snowflake-arctic-embed` instead of `snowflake-arctic-embed2`
GiteaMirror commented 2026-04-12 17:35:09 -05:00
Author
Owner
Copy Link

@DangerousBerries commented on GitHub (Mar 8, 2025):

This still isn't fixed for me even with the latest version.

<!-- gh-comment-id:2708143843 --> @DangerousBerries commented on GitHub (Mar 8, 2025): This still isn't fixed for me even with the latest version.
GiteaMirror commented 2026-04-12 17:35:09 -05:00
Author
Owner
Copy Link

@thiswillbeyourgithub commented on GitHub (Mar 8, 2025):

As of today it is still present in the "latest" docker image (this one I believe) and disappears when I switch back to 0.5.12 via image: ollama/ollama:0.5.12. Using the 0.5.13 client with the 0.5.12 docker container so far causes no issue.

<!-- gh-comment-id:2708352188 --> @thiswillbeyourgithub commented on GitHub (Mar 8, 2025): As of today it is still present in the "latest" docker image ([this one I believe](https://hub.docker.com/layers/ollama/ollama/latest/images/sha256-2853c8e31d8cf3f9c6a1e15ffd3d1085226026030378415907f91ea16a6c3c86)) and disappears when I switch back to 0.5.12 via `image: ollama/ollama:0.5.12`. Using the 0.5.13 client with the 0.5.12 docker container so far causes no issue.
GiteaMirror commented 2026-04-12 17:35:10 -05:00
Author
Owner
Copy Link

@ChellShort commented on GitHub (Mar 11, 2025):

Downgrading to ollama 0.5.12 resolved the issue for me

<!-- gh-comment-id:2714478738 --> @ChellShort commented on GitHub (Mar 11, 2025): Downgrading to ollama 0.5.12 resolved the issue for me
GiteaMirror commented 2026-04-12 17:35:11 -05:00
Author
Owner
Copy Link

@DangerousBerries commented on GitHub (Mar 12, 2025):

@jmorganca this should probably be reopened considering that people continue coming here with the same problem.

<!-- gh-comment-id:2716703410 --> @DangerousBerries commented on GitHub (Mar 12, 2025): @jmorganca this should probably be reopened considering that people continue coming here with the same problem.
GiteaMirror commented 2026-04-12 17:35:11 -05:00
Author
Owner
Copy Link

@juangon commented on GitHub (Mar 12, 2025):

@DangerousBerries there was a new release few hours ago fixing it

<!-- gh-comment-id:2716787130 --> @juangon commented on GitHub (Mar 12, 2025): @DangerousBerries there was a new release few hours ago fixing it
GiteaMirror commented 2026-04-12 17:35:11 -05:00
Author
Owner
Copy Link

@DangerousBerries commented on GitHub (Mar 13, 2025):

@DangerousBerries there was a new release few hours ago fixing it

You're right, thank you for letting me know.

<!-- gh-comment-id:2720350942 --> @DangerousBerries commented on GitHub (Mar 13, 2025): > [@DangerousBerries](https://github.com/DangerousBerries) there was a new release few hours ago fixing it You're right, thank you for letting me know.
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#6199
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?