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 14.5k Packages Projects Releases 122 Wiki Activity

[GH-ISSUE #12074] Seed-OSS-36B-Instruct-GGUF:Q4_K_M #54534

New Issue
Open
opened 2026-04-29 06:16:38 -05:00 by GiteaMirror · 14 comments
GiteaMirror commented 2026-04-29 06:16:38 -05:00
Owner
Copy Link

Originally created by @Eb7CAPJi on GitHub (Aug 25, 2025).
Original GitHub issue: https://github.com/ollama/ollama/issues/12074

Error: 500 Internal Server Error: unable to load model: e:.ollama\models\blobs\sha256-2d9f3c32a61f2668e012708be7ec04e753e35e366f87b9e4bce89f4d073cfc16

Originally created by @Eb7CAPJi on GitHub (Aug 25, 2025). Original GitHub issue: https://github.com/ollama/ollama/issues/12074 Error: 500 Internal Server Error: unable to load model: e:\.ollama\models\blobs\sha256-2d9f3c32a61f2668e012708be7ec04e753e35e366f87b9e4bce89f4d073cfc16
GiteaMirror added the model label 2026-04-29 06:16:38 -05:00
GiteaMirror commented 2026-04-29 06:16:47 -05:00
Author
Owner
Copy Link

@Eb7CAPJi commented on GitHub (Aug 25, 2025):

Attempting to use the Seed-OSS-36B-Instruct model results in an error, while Seed-Coder-8B-Instruct works without issues.

<!-- gh-comment-id:3220888604 --> @Eb7CAPJi commented on GitHub (Aug 25, 2025): Attempting to use the Seed-OSS-36B-Instruct model results in an error, while Seed-Coder-8B-Instruct works without issues.
GiteaMirror commented 2026-04-29 06:16:48 -05:00
Author
Owner
Copy Link

@rick-github commented on GitHub (Aug 25, 2025):

https://github.com/ggml-org/llama.cpp/pull/15490

<!-- gh-comment-id:3220918501 --> @rick-github commented on GitHub (Aug 25, 2025): https://github.com/ggml-org/llama.cpp/pull/15490
GiteaMirror commented 2026-04-29 06:16:48 -05:00
Author
Owner
Copy Link

@courtesysoft commented on GitHub (Aug 26, 2025):

Would love it if this was supported

<!-- gh-comment-id:3222479241 --> @courtesysoft commented on GitHub (Aug 26, 2025): Would love it if this was supported
GiteaMirror commented 2026-04-29 06:16:49 -05:00
Author
Owner
Copy Link

@AuditAIH commented on GitHub (Aug 27, 2025):

Would love it if this was supported

+1

<!-- gh-comment-id:3226603805 --> @AuditAIH commented on GitHub (Aug 27, 2025): > Would love it if this was supported +1
GiteaMirror commented 2026-04-29 06:16:49 -05:00
Author
Owner
Copy Link

@dojoca commented on GitHub (Aug 29, 2025):

+1

<!-- gh-comment-id:3237285397 --> @dojoca commented on GitHub (Aug 29, 2025): +1
GiteaMirror commented 2026-04-29 06:16:49 -05:00
Author
Owner
Copy Link

@courtesysoft commented on GitHub (Aug 29, 2025):

I am running this on LMStudio and it is slow, but on another level of intelligence than any other OSS model i evaluated.

Within 3 revisions, using VS Code + Cline, i can send it a whopper of a coding prompt and get output that's 95% as good as Deepseek R1.

For us it's like an 'AGI achieved internally moment'.. it's an important model and ollama should absolutely implement it

<!-- gh-comment-id:3237403308 --> @courtesysoft commented on GitHub (Aug 29, 2025): I am running this on LMStudio and it is slow, but on another level of intelligence than any other OSS model i evaluated. Within 3 revisions, using VS Code + Cline, i can send it a whopper of a coding prompt and get output that's 95% as good as Deepseek R1. For us it's like an 'AGI achieved internally moment'.. it's an important model and ollama should absolutely implement it
GiteaMirror commented 2026-04-29 06:16:50 -05:00
Author
Owner
Copy Link

@fengwang commented on GitHub (Aug 31, 2025):

I second this model request.

<!-- gh-comment-id:3240230630 --> @fengwang commented on GitHub (Aug 31, 2025): I second this model request.
GiteaMirror commented 2026-04-29 06:16:50 -05:00
Author
Owner
Copy Link

@d-bohn commented on GitHub (Sep 3, 2025):

+1

<!-- gh-comment-id:3247658796 --> @d-bohn commented on GitHub (Sep 3, 2025): +1
GiteaMirror commented 2026-04-29 06:16:50 -05:00
Author
Owner
Copy Link

@ulrichenslin commented on GitHub (Sep 3, 2025):

+1

<!-- gh-comment-id:3249862116 --> @ulrichenslin commented on GitHub (Sep 3, 2025): +1
GiteaMirror commented 2026-04-29 06:16:51 -05:00
Author
Owner
Copy Link

@islobodan commented on GitHub (Sep 16, 2025):

+1

<!-- gh-comment-id:3299478919 --> @islobodan commented on GitHub (Sep 16, 2025): +1
GiteaMirror commented 2026-04-29 06:16:51 -05:00
Author
Owner
Copy Link

@rick-github commented on GitHub (Sep 17, 2025):

ollama will support it when https://github.com/ollama/ollama/pull/12245 is merged.

<!-- gh-comment-id:3301879219 --> @rick-github commented on GitHub (Sep 17, 2025): ollama will support it when https://github.com/ollama/ollama/pull/12245 is merged.
GiteaMirror commented 2026-04-29 06:16:52 -05:00
Author
Owner
Copy Link

@AuditAIH commented on GitHub (Oct 23, 2025):

ollama will support it when #12245 is merged.

@rick-github now is support ? how to run it thanks

<!-- gh-comment-id:3435351145 --> @AuditAIH commented on GitHub (Oct 23, 2025): > ollama will support it when [#12245](https://github.com/ollama/ollama/pull/12245) is merged. @rick-github now is support ? how to run it thanks
GiteaMirror commented 2026-04-29 06:16:53 -05:00
Author
Owner
Copy Link

@Nantris commented on GitHub (Oct 30, 2025):

Is it possible to use Seed-OSS yet? With Codex landing some brutal usage caps today, I've heard several people say Seed-OSS gives the best local agentic coding results.

<!-- gh-comment-id:3470382857 --> @Nantris commented on GitHub (Oct 30, 2025): Is it possible to use Seed-OSS yet? With Codex landing some brutal usage caps today, I've heard several people say Seed-OSS gives the best local agentic coding results.
GiteaMirror commented 2026-04-29 06:16:55 -05:00
Author
Owner
Copy Link

@rick-github commented on GitHub (Oct 31, 2025):

Here's a work-in-progress Modelfile:

FROM hf.co/unsloth/Seed-OSS-36B-Instruct-GGUF:Q4_K_M

TEMPLATE """system
{{- $thinking_budget := -1 }}{{ $interval := 512 }}{{ $use_json_tooldef := false }}
{{- if .System }}
{{ .System }}
{{- else }}
You are Doubao, a helpful AI assistant.
{{- end }}
{{- if .Tools -}}
  You may call one or more functions to assist with the user query.
{{- if $use_json_tooldef }}

Tool List:
You are authorized to use the following tools (described in JSON Schema format). Before performing any task, you must decide how to call them based on the descriptions and parameters of these tools.
{{ .Tools | json }}
{{ else }}
{{ range .Tools }}

Function
def {{ .Function.Name }}(
  {{- $comma := false }}
  {{- range $name, $prop := .Function.Parameters.Properties }}
    {{- if $comma }},{{ end }}{{ $comma = true }}
    {{- $name }}: {{ $prop | toTypeScriptType }}
  {{- end }}):
    {{ `"""` }}
    {{ .Function.Description }}

    {{ if .Function.Parameters.Properties }}
    Args:
    {{- range $name, $prop := .Function.Parameters.Properties }}
    -  {{ $name }} ({{ $prop | toTypeScriptType }}): {{ $prop.Description }}
    {{- end }}
    {{- end }}
    {{ `"""` }}

{{ end }}
{{ end }}
工具调用请遵循如下格式:
<seed:tool_call>
<function=example_function_name>
<parameter=example_parameter_1>value_1</parameter>
<parameter=example_parameter_2>This is the value for the second parameter
that can span
multiple lines</parameter>
</function>
</seed:tool_call>
<seed:eos>
{{ end }}

{{- if .IsThinkSet }}
  {{- if .Think }}
    {{- $thinking_budget = 16384 }}
    {{- $interval = 1024 }}
  {{- else }}
    {{- $thinking_budget = 0 }}
  {{- end }}
  {{- if eq .ThinkLevel "low" }}
    {{- $thinking_budget = 0 }}
  {{- else if eq .ThinkLevel "high" }}
    {{- $thinking_budget = 16384 }}
    {{- $interval = 1024 }}
  {{- else if eq .ThinkLevel "medium" }}
    {{- $thinking_budget = 2048 }}
    {{- $interval = 512 }}
  {{- end }}
{{- end }}

{{- if ne $thinking_budget -1 }}
<seed:bos>system
{{- if eq $thinking_budget 0 }}
You are an intelligent assistant that can answer questions in one step without the need for reasoning and thinking, that is, your thinking budget is 0. Next, please skip the thinking process and directly start answering the user's questions.
{{ else }}
You are an intelligent assistant with reflective ability. In the process of thinking and reasoning, you need to strictly follow the thinking budget, which is {{ $thinking_budget }}. That is, you need to complete your thinking within {{ $thinking_budget }} tokens and start answering the user's questions. You will reflect on your thinking process every {{ $interval }} tokens, stating how many tokens have been used and how many are left.
{{ end }}
<seed:eos>
{{ end }}

{{- range $i, $_ := .Messages }}
{{- $last := eq (len (slice $.Messages $i)) 1}}

{{- if eq .Role "assistant" }}
<seed:bos>{{ .Role }}
{{- if .Thinking }}
<seed:think>{{ .Thinking }}</seed:think>
{{- end }}
{{ .Content }}
{{- if .ToolCalls }}
<seed:tool_call>
{{- range $i, $_ := .ToolCalls }}{"name": "{{ .Function.Name }}", "arguments": {{ .Function.Arguments }}{{ end -}}
</seed:tool_call>
{{- end }}{{/* .ToolCalls */}}
<seed:eos>
{{- end }}{{/* assistant */}}

{{- if or (eq .Role "user") (eq .Role "tool") }}
<seed:bos>{{ .Role }}
{{ .Content }}
<seed:eos>
{{- end }}{{/* user */}}

{{- if and $last (ne .Role "assistant") }}
<seed:bos>assistant
{{- if eq $thinking_budget 0 }}
<seed:think><seed:cot_budget_reflect>The current thinking budget is 0, so I will directly start answering the question.</seed:cot_budget_reflect>
</seed:think>
### Answer:
{{ end }}
{{- end }}
{{- end }}{{/* range */}}
"""

Tool calling doesn't work yet - as soon as the model finishes thinking about which tool to use, it ends the inference without emitting a tool call.

The model supports levels of thinking and a thinking budget, but ollama currently only allows gpt-oss models to set a level. So this template only supports setting think to true or false.

<!-- gh-comment-id:3470882387 --> @rick-github commented on GitHub (Oct 31, 2025): Here's a work-in-progress Modelfile: ```dockerfile FROM hf.co/unsloth/Seed-OSS-36B-Instruct-GGUF:Q4_K_M TEMPLATE """system {{- $thinking_budget := -1 }}{{ $interval := 512 }}{{ $use_json_tooldef := false }} {{- if .System }} {{ .System }} {{- else }} You are Doubao, a helpful AI assistant. {{- end }} {{- if .Tools -}} You may call one or more functions to assist with the user query. {{- if $use_json_tooldef }} Tool List: You are authorized to use the following tools (described in JSON Schema format). Before performing any task, you must decide how to call them based on the descriptions and parameters of these tools. {{ .Tools | json }} {{ else }} {{ range .Tools }} Function def {{ .Function.Name }}( {{- $comma := false }} {{- range $name, $prop := .Function.Parameters.Properties }} {{- if $comma }},{{ end }}{{ $comma = true }} {{- $name }}: {{ $prop | toTypeScriptType }} {{- end }}): {{ `"""` }} {{ .Function.Description }} {{ if .Function.Parameters.Properties }} Args: {{- range $name, $prop := .Function.Parameters.Properties }} - {{ $name }} ({{ $prop | toTypeScriptType }}): {{ $prop.Description }} {{- end }} {{- end }} {{ `"""` }} {{ end }} {{ end }} 工具调用请遵循如下格式: <seed:tool_call> <function=example_function_name> <parameter=example_parameter_1>value_1</parameter> <parameter=example_parameter_2>This is the value for the second parameter that can span multiple lines</parameter> </function> </seed:tool_call> <seed:eos> {{ end }} {{- if .IsThinkSet }} {{- if .Think }} {{- $thinking_budget = 16384 }} {{- $interval = 1024 }} {{- else }} {{- $thinking_budget = 0 }} {{- end }} {{- if eq .ThinkLevel "low" }} {{- $thinking_budget = 0 }} {{- else if eq .ThinkLevel "high" }} {{- $thinking_budget = 16384 }} {{- $interval = 1024 }} {{- else if eq .ThinkLevel "medium" }} {{- $thinking_budget = 2048 }} {{- $interval = 512 }} {{- end }} {{- end }} {{- if ne $thinking_budget -1 }} <seed:bos>system {{- if eq $thinking_budget 0 }} You are an intelligent assistant that can answer questions in one step without the need for reasoning and thinking, that is, your thinking budget is 0. Next, please skip the thinking process and directly start answering the user's questions. {{ else }} You are an intelligent assistant with reflective ability. In the process of thinking and reasoning, you need to strictly follow the thinking budget, which is {{ $thinking_budget }}. That is, you need to complete your thinking within {{ $thinking_budget }} tokens and start answering the user's questions. You will reflect on your thinking process every {{ $interval }} tokens, stating how many tokens have been used and how many are left. {{ end }} <seed:eos> {{ end }} {{- range $i, $_ := .Messages }} {{- $last := eq (len (slice $.Messages $i)) 1}} {{- if eq .Role "assistant" }} <seed:bos>{{ .Role }} {{- if .Thinking }} <seed:think>{{ .Thinking }}</seed:think> {{- end }} {{ .Content }} {{- if .ToolCalls }} <seed:tool_call> {{- range $i, $_ := .ToolCalls }}{"name": "{{ .Function.Name }}", "arguments": {{ .Function.Arguments }}{{ end -}} </seed:tool_call> {{- end }}{{/* .ToolCalls */}} <seed:eos> {{- end }}{{/* assistant */}} {{- if or (eq .Role "user") (eq .Role "tool") }} <seed:bos>{{ .Role }} {{ .Content }} <seed:eos> {{- end }}{{/* user */}} {{- if and $last (ne .Role "assistant") }} <seed:bos>assistant {{- if eq $thinking_budget 0 }} <seed:think><seed:cot_budget_reflect>The current thinking budget is 0, so I will directly start answering the question.</seed:cot_budget_reflect> </seed:think> ### Answer: {{ end }} {{- end }} {{- end }}{{/* range */}} """ ``` Tool calling doesn't work yet - as soon as the model finishes thinking about which tool to use, it ends the inference without emitting a tool call. The model supports levels of thinking and a thinking budget, but ollama currently only allows gpt-oss models to set a level. So this template only supports setting `think` to `true` or `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 model
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#54534
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?