github-starred/ollama
2
0
Fork 0
mirror of https://github.com/ollama/ollama.git synced 2026-07-16 05:05:37 -05:00
Code Issues 16.9k Packages Projects Releases 122 Wiki Activity

[GH-ISSUE #11361] please support model architecture: 'hunyuan-moe' #85181

New Issue
Closed
opened 2026-05-09 22:45:59 -05:00 by GiteaMirror · 5 comments
GiteaMirror commented 2026-05-09 22:45:59 -05:00
Owner
Copy Link

Originally created by @johanteekens on GitHub (Jul 10, 2025).
Original GitHub issue: https://github.com/ollama/ollama/issues/11361

Please add support for Hunyuan-A13B-Instruct
Model: https://huggingface.co/tencent/Hunyuan-A13B-Instruct

llama_model_loader: Dumping metadata keys/values. Note: KV overrides do not apply in this output.
llama_model_loader: - kv 0: general.architecture str = hunyuan-moe
llama_model_loader: - kv 1: general.type str = model
llama_model_loader: - kv 2: general.name str = Hunyuan-A13B-Instruct
llama_model_loader: - kv 3: general.finetune str = Instruct
llama_model_loader: - kv 4: general.basename str = Hunyuan-A13B-Instruct
llama_model_loader: - kv 5: general.quantized_by str = Unsloth
llama_model_loader: - kv 6: general.size_label str = A13B
llama_model_loader: - kv 7: general.license str = other
llama_model_loader: - kv 8: general.license.name str = tencent-hunyuan-a13b
llama_model_loader: - kv 9: general.license.link str = https://github.com/Tencent-Hunyuan/Hu...
llama_model_loader: - kv 10: general.repo_url str = https://huggingface.co/unsloth
llama_model_loader: - kv 11: general.base_model.count u32 = 1
llama_model_loader: - kv 12: general.base_model.0.name str = Hunyuan A13B Instruct
llama_model_loader: - kv 13: general.base_model.0.organization str = Tencent
llama_model_loader: - kv 14: general.base_model.0.repo_url str = https://huggingface.co/tencent/Hunyua...
llama_model_loader: - kv 15: general.tags arr[str,1] = ["unsloth"]
llama_model_loader: - kv 16: hunyuan-moe.block_count u32 = 32
llama_model_loader: - kv 17: hunyuan-moe.context_length u32 = 262144
llama_model_loader: - kv 18: hunyuan-moe.embedding_length u32 = 4096
llama_model_loader: - kv 19: hunyuan-moe.feed_forward_length u32 = 3072
llama_model_loader: - kv 20: hunyuan-moe.attention.head_count u32 = 32
llama_model_loader: - kv 21: hunyuan-moe.attention.head_count_kv u32 = 8
llama_model_loader: - kv 22: hunyuan-moe.rope.freq_base f32 = 11158840.000000
llama_model_loader: - kv 23: hunyuan-moe.attention.layer_norm_rms_epsilon f32 = 0.000010
llama_model_loader: - kv 24: hunyuan-moe.expert_count u32 = 64
llama_model_loader: - kv 25: hunyuan-moe.expert_shared_feed_forward_length u32 = 3072
llama_model_loader: - kv 26: hunyuan-moe.expert_feed_forward_length u32 = 3072
llama_model_loader: - kv 27: hunyuan-moe.expert_used_count u32 = 8
llama_model_loader: - kv 28: hunyuan-moe.expert_shared_count u32 = 1
llama_model_loader: - kv 29: hunyuan-moe.rope.scaling.type str = none
llama_model_loader: - kv 30: hunyuan-moe.rope.scaling.factor f32 = 1.000000
llama_model_loader: - kv 31: hunyuan-moe.rope.scaling.original_context_length u32 = 262144
llama_model_loader: - kv 32: tokenizer.ggml.model str = gpt2
llama_model_loader: - kv 33: tokenizer.ggml.pre str = hunyuan
llama_model_loader: - kv 34: tokenizer.ggml.tokens arr[str,128167] = ["!", """, "#", "$", "%", "&", "'", ...
llama_model_loader: - kv 35: tokenizer.ggml.token_type arr[i32,128167] = [1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, ...
llama_model_loader: - kv 36: tokenizer.ggml.merges arr[str,127698] = ["Ġ Ġ", "ĠĠ ĠĠ", "i n", "Ġ t",...
llama_model_loader: - kv 37: tokenizer.ggml.eos_token_id u32 = 127960
llama_model_loader: - kv 38: tokenizer.ggml.seperator_token_id u32 = 127962
llama_model_loader: - kv 39: tokenizer.ggml.padding_token_id u32 = 127961
llama_model_loader: - kv 40: tokenizer.chat_template str = {% set loop_messages = messages %}\n{%...
llama_model_loader: - kv 41: tokenizer.ggml.bos_token_id u32 = 127959
llama_model_loader: - kv 42: general.quantization_version u32 = 2
llama_model_loader: - kv 43: general.file_type u32 = 3
llama_model_loader: - kv 44: quantize.imatrix.file str = Hunyuan-A13B-Instruct-GGUF/imatrix_un...
llama_model_loader: - kv 45: quantize.imatrix.dataset str = unsloth_calibration_Hunyuan-A13B-Inst...
llama_model_loader: - kv 46: quantize.imatrix.entries_count u32 = 352
llama_model_loader: - kv 47: quantize.imatrix.chunks_count u32 = 695
llama_model_loader: - kv 48: split.no u16 = 0
llama_model_loader: - kv 49: split.tensors.count i32 = 482
llama_model_loader: - kv 50: split.count u16 = 0
llama_model_loader: - type f32: 161 tensors
llama_model_loader: - type q4_1: 320 tensors
llama_model_loader: - type q6_K: 1 tensors
print_info: file format = GGUF V3 (latest)
print_info: file type = Q4_1
print_info: file size = 46.92 GiB (5.01 BPW)
llama_model_load: error loading model: error loading model architecture: unknown model architecture: 'hunyuan-moe'
llama_model_load_from_file_impl: failed to load model

Originally created by @johanteekens on GitHub (Jul 10, 2025). Original GitHub issue: https://github.com/ollama/ollama/issues/11361 Please add support for Hunyuan-A13B-Instruct Model: https://huggingface.co/tencent/Hunyuan-A13B-Instruct llama_model_loader: Dumping metadata keys/values. Note: KV overrides do not apply in this output. llama_model_loader: - kv 0: general.architecture str = hunyuan-moe llama_model_loader: - kv 1: general.type str = model llama_model_loader: - kv 2: general.name str = Hunyuan-A13B-Instruct llama_model_loader: - kv 3: general.finetune str = Instruct llama_model_loader: - kv 4: general.basename str = Hunyuan-A13B-Instruct llama_model_loader: - kv 5: general.quantized_by str = Unsloth llama_model_loader: - kv 6: general.size_label str = A13B llama_model_loader: - kv 7: general.license str = other llama_model_loader: - kv 8: general.license.name str = tencent-hunyuan-a13b llama_model_loader: - kv 9: general.license.link str = https://github.com/Tencent-Hunyuan/Hu... llama_model_loader: - kv 10: general.repo_url str = https://huggingface.co/unsloth llama_model_loader: - kv 11: general.base_model.count u32 = 1 llama_model_loader: - kv 12: general.base_model.0.name str = Hunyuan A13B Instruct llama_model_loader: - kv 13: general.base_model.0.organization str = Tencent llama_model_loader: - kv 14: general.base_model.0.repo_url str = https://huggingface.co/tencent/Hunyua... llama_model_loader: - kv 15: general.tags arr[str,1] = ["unsloth"] llama_model_loader: - kv 16: hunyuan-moe.block_count u32 = 32 llama_model_loader: - kv 17: hunyuan-moe.context_length u32 = 262144 llama_model_loader: - kv 18: hunyuan-moe.embedding_length u32 = 4096 llama_model_loader: - kv 19: hunyuan-moe.feed_forward_length u32 = 3072 llama_model_loader: - kv 20: hunyuan-moe.attention.head_count u32 = 32 llama_model_loader: - kv 21: hunyuan-moe.attention.head_count_kv u32 = 8 llama_model_loader: - kv 22: hunyuan-moe.rope.freq_base f32 = 11158840.000000 llama_model_loader: - kv 23: hunyuan-moe.attention.layer_norm_rms_epsilon f32 = 0.000010 llama_model_loader: - kv 24: hunyuan-moe.expert_count u32 = 64 llama_model_loader: - kv 25: hunyuan-moe.expert_shared_feed_forward_length u32 = 3072 llama_model_loader: - kv 26: hunyuan-moe.expert_feed_forward_length u32 = 3072 llama_model_loader: - kv 27: hunyuan-moe.expert_used_count u32 = 8 llama_model_loader: - kv 28: hunyuan-moe.expert_shared_count u32 = 1 llama_model_loader: - kv 29: hunyuan-moe.rope.scaling.type str = none llama_model_loader: - kv 30: hunyuan-moe.rope.scaling.factor f32 = 1.000000 llama_model_loader: - kv 31: hunyuan-moe.rope.scaling.original_context_length u32 = 262144 llama_model_loader: - kv 32: tokenizer.ggml.model str = gpt2 llama_model_loader: - kv 33: tokenizer.ggml.pre str = hunyuan llama_model_loader: - kv 34: tokenizer.ggml.tokens arr[str,128167] = ["!", "\"", "#", "$", "%", "&", "'", ... llama_model_loader: - kv 35: tokenizer.ggml.token_type arr[i32,128167] = [1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, ... llama_model_loader: - kv 36: tokenizer.ggml.merges arr[str,127698] = ["Ġ Ġ", "ĠĠ ĠĠ", "i n", "Ġ t",... llama_model_loader: - kv 37: tokenizer.ggml.eos_token_id u32 = 127960 llama_model_loader: - kv 38: tokenizer.ggml.seperator_token_id u32 = 127962 llama_model_loader: - kv 39: tokenizer.ggml.padding_token_id u32 = 127961 llama_model_loader: - kv 40: tokenizer.chat_template str = {% set loop_messages = messages %}\n{%... llama_model_loader: - kv 41: tokenizer.ggml.bos_token_id u32 = 127959 llama_model_loader: - kv 42: general.quantization_version u32 = 2 llama_model_loader: - kv 43: general.file_type u32 = 3 llama_model_loader: - kv 44: quantize.imatrix.file str = Hunyuan-A13B-Instruct-GGUF/imatrix_un... llama_model_loader: - kv 45: quantize.imatrix.dataset str = unsloth_calibration_Hunyuan-A13B-Inst... llama_model_loader: - kv 46: quantize.imatrix.entries_count u32 = 352 llama_model_loader: - kv 47: quantize.imatrix.chunks_count u32 = 695 llama_model_loader: - kv 48: split.no u16 = 0 llama_model_loader: - kv 49: split.tensors.count i32 = 482 llama_model_loader: - kv 50: split.count u16 = 0 llama_model_loader: - type f32: 161 tensors llama_model_loader: - type q4_1: 320 tensors llama_model_loader: - type q6_K: 1 tensors print_info: file format = GGUF V3 (latest) print_info: file type = Q4_1 print_info: file size = 46.92 GiB (5.01 BPW) llama_model_load: error loading model: error loading model architecture: unknown model architecture: 'hunyuan-moe' llama_model_load_from_file_impl: failed to load model
GiteaMirror added the model label 2026-05-09 22:45:59 -05:00
GiteaMirror commented 2026-05-09 22:46:03 -05:00
Author
Owner
Copy Link

@rick-github commented on GitHub (Jul 10, 2025):

https://github.com/ggml-org/llama.cpp/pull/14425

Should work after the next ollama sync to the vendored llama.cpp code.

<!-- gh-comment-id:3058085416 --> @rick-github commented on GitHub (Jul 10, 2025): https://github.com/ggml-org/llama.cpp/pull/14425 Should work after the next ollama sync to the vendored llama.cpp code.
GiteaMirror commented 2026-05-09 22:46:05 -05:00
Author
Owner
Copy Link

@stevenkuang-tencent commented on GitHub (Jul 11, 2025):

hi @rick-github, when will sync llama.cpp approximately?

<!-- gh-comment-id:3061217619 --> @stevenkuang-tencent commented on GitHub (Jul 11, 2025): hi @rick-github, when will sync llama.cpp approximately?
GiteaMirror commented 2026-05-09 22:46:06 -05:00
Author
Owner
Copy Link

@rick-github commented on GitHub (Jul 11, 2025):

Some time in the future.

<!-- gh-comment-id:3061237071 --> @rick-github commented on GitHub (Jul 11, 2025): Some time in the future.
GiteaMirror commented 2026-05-09 22:46:07 -05:00
Author
Owner
Copy Link

@stevenkuang-tencent commented on GitHub (Jul 11, 2025):

Some time in the future.

ok, thanks~

<!-- gh-comment-id:3061252209 --> @stevenkuang-tencent commented on GitHub (Jul 11, 2025): > Some time in the future. ok, thanks~
GiteaMirror commented 2026-05-09 22:46:08 -05:00
Author
Owner
Copy Link

@rick-github commented on GitHub (Aug 18, 2025):

FROM hf.co/unsloth/Hunyuan-A13B-Instruct-GGUF:q4_k_m

TEMPLATE """
{{- $lastUserIdx := -1 }}
{{- range $i, $_ := .Messages }}
{{- if eq .Role "user" }}{{- $lastUserIdx = $i }}{{ end }}
{{- end -}}
{{- if .System }}<|startoftext|>
{{- .System }}<|extra_4|>
{{- end }}
{{- range $i, $_ := .Messages }}
{{- $last := eq (len (slice $.Messages $i)) 1 }}
{{- if eq .Role "user" }}<|startoftext|>
{{- .Content }}<|extra_0|>
{{- else if eq .Role "assistant" }}<|startoftext|>
{{- if (and $.IsThinkSet (and .Thinking (or $last (gt $i $lastUserIdx)))) -}}
<think>{{ .Thinking }}</think>
{{- end }}
{{- .Content }}<|eos|>
{{- end }}
{{- if and $.IsThinkSet (not $.Think) -}}
<think>

</think>
{{ end -}}
{{- end }}"""

PARAMETER stop <|startoftext|>
PARAMETER stop <|extra_4|>
PARAMETER stop <|extra_0|>
PARAMETER stop <|eos|>

$ ollama run hunyuan:80b-a13b-q4_K_M --think=false hello
<answer>
Hello there, how can I assist you today? 😊
</answer>

$ ollama run hunyuan:80b-a13b-q4_K_M --think=true hello
Thinking...
Okay, the user said "hello". I need to respond in a friendly and welcoming way. Let me think 
of a simple greeting.

"Hello there!" sounds good. Adding an emoji might make it friendlier. Maybe 😊. Then offer 
assistance. Something like "How can I help you today?" That's standard but effective. Keep it 
concise and open-ended so the user feels encouraged to ask anything they need.
...done thinking.

<answer>
Hello there 😊 How can I help you today?
</answer>
<!-- gh-comment-id:3198737802 --> @rick-github commented on GitHub (Aug 18, 2025): ```dockerfile FROM hf.co/unsloth/Hunyuan-A13B-Instruct-GGUF:q4_k_m TEMPLATE """ {{- $lastUserIdx := -1 }} {{- range $i, $_ := .Messages }} {{- if eq .Role "user" }}{{- $lastUserIdx = $i }}{{ end }} {{- end -}} {{- if .System }}<|startoftext|> {{- .System }}<|extra_4|> {{- end }} {{- range $i, $_ := .Messages }} {{- $last := eq (len (slice $.Messages $i)) 1 }} {{- if eq .Role "user" }}<|startoftext|> {{- .Content }}<|extra_0|> {{- else if eq .Role "assistant" }}<|startoftext|> {{- if (and $.IsThinkSet (and .Thinking (or $last (gt $i $lastUserIdx)))) -}} <think>{{ .Thinking }}</think> {{- end }} {{- .Content }}<|eos|> {{- end }} {{- if and $.IsThinkSet (not $.Think) -}} <think> </think> {{ end -}} {{- end }}""" PARAMETER stop <|startoftext|> PARAMETER stop <|extra_4|> PARAMETER stop <|extra_0|> PARAMETER stop <|eos|> ``` ```console $ ollama run hunyuan:80b-a13b-q4_K_M --think=false hello <answer> Hello there, how can I assist you today? 😊 </answer> ``` ```console $ ollama run hunyuan:80b-a13b-q4_K_M --think=true hello Thinking... Okay, the user said "hello". I need to respond in a friendly and welcoming way. Let me think of a simple greeting. "Hello there!" sounds good. Adding an emoji might make it friendlier. Maybe 😊. Then offer assistance. Something like "How can I help you today?" That's standard but effective. Keep it concise and open-ended so the user feels encouraged to ask anything they need. ...done thinking. <answer> Hello there 😊 How can I help you today? </answer> ```
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#85181
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?