github-starred/ollama
2
0
Fork 0
mirror of https://github.com/ollama/ollama.git synced 2026-05-07 00:22:43 -05:00
Code Issues 14.5k Packages Projects Releases 120 Wiki Activity

[GH-ISSUE #10912] deepseek-r1-8b-qwen3 support tool call #7177

New Issue
Closed
opened 2026-04-12 19:10:29 -05:00 by GiteaMirror · 6 comments
GiteaMirror commented 2026-04-12 19:10:29 -05:00
Owner
Copy Link

Originally created by @K-O-T-W on GitHub (May 30, 2025).
Original GitHub issue: https://github.com/ollama/ollama/issues/10912

What is the issue?

Image
Both the teacher model and the base model of this architecture support tool calls, and Hugging Face's chat_template documentation also mentions handling for tool calls. However, this specific template does not support it—I get an error when passing the tool parameter:
{"error":"registry.ollama.ai/library/deepseek-r1:8b-0528-qwen3-q8_0 does not support tools"}

Relevant log output

OS

No response

GPU

No response

CPU

No response

Ollama version

No response

Originally created by @K-O-T-W on GitHub (May 30, 2025). Original GitHub issue: https://github.com/ollama/ollama/issues/10912 ### What is the issue? ![Image](https://github.com/user-attachments/assets/8652d2c5-b34a-499b-8f82-f9b9c4fd2c5e) Both the teacher model and the base model of this architecture support tool calls, and Hugging Face's chat_template documentation also mentions handling for tool calls. However, this specific template does not support it—I get an error when passing the tool parameter: {"error":"registry.ollama.ai/library/deepseek-r1:8b-0528-qwen3-q8_0 does not support tools"} ### 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 19:10:29 -05:00
GiteaMirror commented 2026-04-12 19:10:30 -05:00
Author
Owner
Copy Link

@Chanantho22 commented on GitHub (May 30, 2025):

What is the issue?

Image Both the teacher model and the base model of this architecture support tool calls, and Hugging Face's chat_template documentation also mentions handling for tool calls. However, this specific template does not support it—I get an error when passing the tool parameter: {"error":"registry.ollama.ai/library/deepseek-r1:8b-0528-qwen3-q8_0 does not support tools"}

Relevant log output

OS

No response

GPU

No response

CPU

No response

Ollama version

No response

<!-- gh-comment-id:2921742802 --> @Chanantho22 commented on GitHub (May 30, 2025): > ### What is the issue? > ![Image](https://github.com/user-attachments/assets/8652d2c5-b34a-499b-8f82-f9b9c4fd2c5e) Both the teacher model and the base model of this architecture support tool calls, and Hugging Face's chat_template documentation also mentions handling for tool calls. However, this specific template does not support it—I get an error when passing the tool parameter: {"error":"registry.ollama.ai/library/deepseek-r1:8b-0528-qwen3-q8_0 does not support tools"} > > ### Relevant log output > ### OS > _No response_ > > ### GPU > _No response_ > > ### CPU > _No response_ > > ### Ollama version > _No response_
GiteaMirror commented 2026-04-12 19:10:30 -05:00
Author
Owner
Copy Link

@Master-Pr0grammer commented on GitHub (Jun 1, 2025):

came here to say this, looks like it might just be a simple fix to the chat template. hopefully this gets updated soon.

<!-- gh-comment-id:2927847105 --> @Master-Pr0grammer commented on GitHub (Jun 1, 2025): came here to say this, looks like it might just be a simple fix to the chat template. hopefully this gets updated soon.
GiteaMirror commented 2026-04-12 19:10:31 -05:00
Author
Owner
Copy Link

@benhaotang commented on GitHub (Jun 1, 2025):

I also want to point out that although a template like this:https://ollama.com/okamototk/deepseek-r1:8b/blobs/e94a8ecb9327 do work, it also makes the model hallucinate a lot more and also much worse at instruction following.

<!-- gh-comment-id:2927853932 --> @benhaotang commented on GitHub (Jun 1, 2025): I also want to point out that although a template like this:https://ollama.com/okamototk/deepseek-r1:8b/blobs/e94a8ecb9327 do work, it also makes the model hallucinate a lot more and also much worse at instruction following.
GiteaMirror commented 2026-04-12 19:10:31 -05:00
Author
Owner
Copy Link

@tko commented on GitHub (Jul 1, 2025):

I suspect the problem with okamototks model is that the tokens/conventions used in the template (<|im_start|>assistant) don't match up with the params (<｜Assistant｜>)

I'd be inclined to try translate huggingface jinja2 template into ollama gotmpl format, though I've no idea which parts/tokens are significant and what's arbitrary.

<!-- gh-comment-id:3025246025 --> @tko commented on GitHub (Jul 1, 2025): I suspect the problem with `okamototk`s model is that the tokens/conventions used in the template (`<|im_start|>assistant`) don't match up with the params (`<｜Assistant｜>`) I'd be inclined to try translate [huggingface jinja2 template](https://huggingface.co/deepseek-ai/DeepSeek-R1-Distill-Qwen-32B?chat_template=default&format=true) into ollama gotmpl format, though I've no idea which parts/tokens are significant and what's arbitrary.
GiteaMirror commented 2026-04-12 19:10:31 -05:00
Author
Owner
Copy Link

@ParthSareen commented on GitHub (Jul 6, 2025):

I've spent a good amount of time trying to get this working. Template wise it's super simple to add but the model randomly misses tokens which are supposed to be there when doing tool calling. The distilled model's tokenizer and the parent model's tokenizers are different which could possibly cause the divergence. Found inadequate tool calling in other implementations as well – don't want to enable this if it's going to be a poor experience. Sorry for not communicating!

<!-- gh-comment-id:3040479108 --> @ParthSareen commented on GitHub (Jul 6, 2025): I've spent a good amount of time trying to get this working. Template wise it's super simple to add but the model randomly misses tokens which are supposed to be there when doing tool calling. The distilled model's tokenizer and the parent model's tokenizers are different which could possibly cause the divergence. Found inadequate tool calling in other implementations as well – don't want to enable this if it's going to be a poor experience. Sorry for not communicating!
GiteaMirror commented 2026-04-12 19:10:32 -05:00
Author
Owner
Copy Link

@tko commented on GitHub (Jul 6, 2025):

Oh, I see some of the deepseek-r1 model templates do include tools, for example deepseek-r1:671b

The web page is very unclear that the tools tag doesn't apply to all models. There's really no way of knowing from the list, you'd need to click through each model and look at its template. Adding a label on the list would help avoid the confusion, similar to how the Context and Input columns indicate what to expect.

<!-- gh-comment-id:3041134657 --> @tko commented on GitHub (Jul 6, 2025): Oh, I see _some_ of the deepseek-r1 model templates do include tools, for example [deepseek-r1:671b](https://ollama.com/library/deepseek-r1:671b) The web page is very unclear that the `tools` tag doesn't apply to all models. There's really no way of knowing from the list, you'd need to click through each model and look at its template. Adding a label on the list would help avoid the confusion, similar to how the Context and Input columns indicate what to expect.
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#7177
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?