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 #14657] bge-m3 only returns NaN on bitcoin whitepaper, other docs #71551

New Issue
Open
opened 2026-05-05 02:06:25 -05:00 by GiteaMirror · 5 comments
GiteaMirror commented 2026-05-05 02:06:25 -05:00
Owner
Copy Link

Originally created by @higgsaxhh on GitHub (Mar 6, 2026).
Original GitHub issue: https://github.com/ollama/ollama/issues/14657

What is the issue?

The bge-m3 model returns NaN values through the OpenAI-compatible embeddings API (/v1/embeddings) when processing certain text content, particularly technical documents.

This causes a 500 error with the message: failed to encode response: json: unsupported value: NaN

Environment

  • Ollama Version: 0.17.6
  • OS: Windows 10 / MSYS_NT-10.0-26100
  • Model: bge-m3:latest (ID: 790764642607, Size: 1.2 GB)
  • GPU: NVIDIA GeFORCE RTX 2080 Ti

Steps to Reproduce

  1. Pull the bge-m3 model
ollama pull bge-m3
  1. Test with simple text
  curl -X POST http://localhost:11434/v1/embeddings \
    -H "Content-Type: application/json" \
    -d '{
      "model": "bge-m3",
      "input": "This is a test sentence."
    }'

✅ Returns valid 1024-dimensional embedding

  1. Test with technical document content
  curl -X POST http://localhost:11434/v1/embeddings \
    -H "Content-Type: application/json" \
    -d '{
      "model": "bge-m3",
      "input": "Bitcoin: A Peer-to-Peer Electronic Cash System. Abstract. A purely peer-to-peer version of electronic cash
  would allow online payments to be sent directly from one party to another without going through a financial institution.
  Digital signatures provide part of the solution, but the main benefits are lost if a trusted third party is still required to
   prevent double-spending. We propose a solution to the double-spending problem using a peer-to-peer network. The network
  timestamps transactions by hashing them into an ongoing chain of hash-based proof-of-work, forming a record that cannot be
  changed without redoing the proof-of-work."
    }'

Expected Behavior

Should return a valid 1024-dimensional embedding array, similar to the simple text case.

Actual Behavior

  {
    "error": {
      "message": "failed to encode response: json: unsupported value: NaN",
      "type": "api_error",
      "param": null,
      "code": null
    }
  }

Additional Context

  • The same technical text works perfectly with nomic-embed-text:latest, which returns valid 768-dimensional embeddings without any NaN values
  • This issue occurs consistently with content from technical PDFs (e.g., Bitcoin whitepaper, research papers)
  • The issue appears to be specific to bge-m3 - other embedding models handle the same content without issues

Relevant log output

openai.InternalServerError: Error code: 500 - {'error': {'message': 'failed to encode response: json: unsupported value: NaN', 'type': 'api_error', 'param': None, 'code': None}}

OS

Windows

GPU

Nvidia

CPU

Intel

Ollama version

0.17.6

Originally created by @higgsaxhh on GitHub (Mar 6, 2026). Original GitHub issue: https://github.com/ollama/ollama/issues/14657 ### What is the issue? The bge-m3 model returns NaN values through the OpenAI-compatible embeddings API (/v1/embeddings) when processing certain text content, particularly technical documents. This causes a 500 error with the message: `failed to encode response: json: unsupported value: NaN` Environment - Ollama Version: 0.17.6 - OS: Windows 10 / MSYS_NT-10.0-26100 - Model: bge-m3:latest (ID: 790764642607, Size: 1.2 GB) - GPU: NVIDIA GeFORCE RTX 2080 Ti Steps to Reproduce 1. Pull the bge-m3 model ``` ollama pull bge-m3 ``` 2. Test with simple text ``` curl -X POST http://localhost:11434/v1/embeddings \ -H "Content-Type: application/json" \ -d '{ "model": "bge-m3", "input": "This is a test sentence." }' ``` ✅ Returns valid 1024-dimensional embedding 3. Test with technical document content ``` curl -X POST http://localhost:11434/v1/embeddings \ -H "Content-Type: application/json" \ -d '{ "model": "bge-m3", "input": "Bitcoin: A Peer-to-Peer Electronic Cash System. Abstract. A purely peer-to-peer version of electronic cash would allow online payments to be sent directly from one party to another without going through a financial institution. Digital signatures provide part of the solution, but the main benefits are lost if a trusted third party is still required to prevent double-spending. We propose a solution to the double-spending problem using a peer-to-peer network. The network timestamps transactions by hashing them into an ongoing chain of hash-based proof-of-work, forming a record that cannot be changed without redoing the proof-of-work." }' ``` **Expected Behavior** Should return a valid 1024-dimensional embedding array, similar to the simple text case. **Actual Behavior** ``` { "error": { "message": "failed to encode response: json: unsupported value: NaN", "type": "api_error", "param": null, "code": null } } ``` Additional Context - The same technical text works perfectly with `nomic-embed-text:latest`, which returns valid 768-dimensional embeddings without any NaN values - This issue occurs consistently with content from technical PDFs (e.g., Bitcoin whitepaper, research papers) - The issue appears to be specific to bge-m3 - other embedding models handle the same content without issues ### Relevant log output ```shell openai.InternalServerError: Error code: 500 - {'error': {'message': 'failed to encode response: json: unsupported value: NaN', 'type': 'api_error', 'param': None, 'code': None}} ``` ### OS Windows ### GPU Nvidia ### CPU Intel ### Ollama version 0.17.6
GiteaMirror added the bug label 2026-05-05 02:06:25 -05:00
GiteaMirror commented 2026-05-05 02:06:27 -05:00
Author
Owner
Copy Link

@telunyang commented on GitHub (Mar 9, 2026):

Have you tried the solution from https://github.com/ollama/ollama/issues/13572?

Setting OLLAMA_FLASH_ATTENTION=false solves the problem.

<!-- gh-comment-id:4023570559 --> @telunyang commented on GitHub (Mar 9, 2026): Have you tried the solution from [https://github.com/ollama/ollama/issues/13572](https://github.com/ollama/ollama/issues/13572)? Setting `OLLAMA_FLASH_ATTENTION=false` solves the problem.
GiteaMirror commented 2026-05-05 02:06:29 -05:00
Author
Owner
Copy Link

@mvanhorn commented on GitHub (Mar 9, 2026):

I've submitted a fix in #14739. Added NaN/Inf detection in the embedding response path (both runner-level handlers and the deprecated EmbeddingsHandler) to prevent the JSON serialization crash. The fix returns a clear error message instead of the cryptic json: unsupported value: NaN.

<!-- gh-comment-id:4024443460 --> @mvanhorn commented on GitHub (Mar 9, 2026): I've submitted a fix in #14739. Added NaN/Inf detection in the embedding response path (both runner-level handlers and the deprecated EmbeddingsHandler) to prevent the JSON serialization crash. The fix returns a clear error message instead of the cryptic `json: unsupported value: NaN`.
GiteaMirror commented 2026-05-05 02:06:31 -05:00
Author
Owner
Copy Link

@higgsaxhh commented on GitHub (Mar 9, 2026):

I've submitted a fix in #14739. Added NaN/Inf detection in the embedding response path (both runner-level handlers and the deprecated EmbeddingsHandler) to prevent the JSON serialization crash. The fix returns a clear error message instead of the cryptic json: unsupported value: NaN.

Thanks @mvanhorn. Does this solve the embedding problem or just more gracefully handle the error?

<!-- gh-comment-id:4026862589 --> @higgsaxhh commented on GitHub (Mar 9, 2026): > I've submitted a fix in [#14739](https://github.com/ollama/ollama/pull/14739). Added NaN/Inf detection in the embedding response path (both runner-level handlers and the deprecated EmbeddingsHandler) to prevent the JSON serialization crash. The fix returns a clear error message instead of the cryptic `json: unsupported value: NaN`. Thanks @mvanhorn. Does this solve the embedding problem or just more gracefully handle the error?
GiteaMirror commented 2026-05-05 02:06:32 -05:00
Author
Owner
Copy Link

@mvanhorn commented on GitHub (Mar 9, 2026):

Good question - to be clear, #14739 handles the error gracefully rather than fixing the root cause of why bge-m3 produces NaN values. The underlying NaN generation is a model/compute issue (likely related to flash attention as @telunyang mentioned - disabling it with OLLAMA_FLASH_ATTENTION=false is the current workaround).

What #14739 does is prevent the server from crashing with a cryptic json: unsupported value: NaN 500 error, and instead returns a clear error message explaining what happened. That way you get actionable feedback instead of a stack trace.

<!-- gh-comment-id:4027575663 --> @mvanhorn commented on GitHub (Mar 9, 2026): Good question - to be clear, #14739 handles the error gracefully rather than fixing the root cause of why bge-m3 produces NaN values. The underlying NaN generation is a model/compute issue (likely related to flash attention as @telunyang mentioned - disabling it with `OLLAMA_FLASH_ATTENTION=false` is the current workaround). What #14739 does is prevent the server from crashing with a cryptic `json: unsupported value: NaN` 500 error, and instead returns a clear error message explaining what happened. That way you get actionable feedback instead of a stack trace.
GiteaMirror commented 2026-05-05 02:06:34 -05:00
Author
Owner
Copy Link

@neuwcodebox commented on GitHub (Apr 7, 2026):

This issue still occurs in the latest ollama version (v0.20.1) and was resolved by setting the OLLAMA_FLASH_ATTENTION environment variable to false.

<!-- gh-comment-id:4196028561 --> @neuwcodebox commented on GitHub (Apr 7, 2026): This issue still occurs in the latest ollama version (v0.20.1) and was resolved by setting the OLLAMA_FLASH_ATTENTION environment variable to false.
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#71551
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?