Describe the bug
I am attempting to deploy Tabby in an air-gapped environment using Docker, following the tutorial. While the default models are working as expected, I encountered an error error loading model architecture: unknown model architecture: 'qwen3next' when trying to deploy a Qwen3-Coder-Next GGUF model from Hugging Face .
Docker image based on main-5731104
docker run cmd
docker run -it \
--gpus '"device=1"' -p 7777:7777 -v $HOME/.tabby:/data \
tabby-offline \
serve --model /data/models/TabbyML/Qwen3-Coder-Next --device cuda --port 7777
Qwen3-Coder-Next directory structure
L Qwen3-Coder-Next
L ggml
L model-00001-of-00004.gguf
L model-00002-of-00004.gguf
L model-00003-of-00004.gguf
L model-00004-of-00004.gguf
L tabby.json
Is the 'qwen3next' model architecture currently not supported?
Information about your version
tabby 0.33.0-dev.0
Information about your GPU
+-----------------------------------------------------------------------------------------+
| NVIDIA-SMI 580.126.09 Driver Version: 580.126.09 CUDA Version: 13.0 |
|-----------------------------------------+------------------------+----------------------+
| GPU Name Persistence-M | Bus-Id Disp.A | Volatile Uncorr. ECC |
| Fan Temp Perf Pwr:Usage/Cap | Memory-Usage | GPU-Util Compute M. |
| | | MIG M. |
|=========================================+========================+======================|
| 0 NVIDIA RTX PRO 6000 Blac... Off | 00000000:72:00.0 Off | 0 |
| N/A 39C P0 102W / 600W | 28740MiB / 97887MiB | 21% Default |
| | | Disabled |
+-----------------------------------------+------------------------+----------------------+
| 1 NVIDIA RTX PRO 6000 Blac... Off | 00000000:82:00.0 Off | 0 |
| N/A 39C P0 101W / 600W | 30970MiB / 97887MiB | 21% Default |
| | | Disabled |
+-----------------------------------------+------------------------+----------------------+
Additional context
Log
ggml_cuda_init: GGML_CUDA_FORCE_MMQ: no
ggml_cuda_init: GGML_CUDA_FORCE_CUBLAS: no
ggml_cuda_init: found 1 CUDA devices:
Device 0: NVIDIA RTX PRO 6000 Blackwell Server Edition, compute capability 12.0, VMM: yes
build: 1 (952a47f) with cc (Ubuntu 11.4.0-1ubuntu1~22.04) 11.4.0 for x86_64-linux-gnu
system_info: n_threads = 96, n_threads_batch = 96, total_threads = 192
system_info: n_threads = 96 (n_threads_batch = 96) / 192 | CUDA : ARCHS = 500,610,700,750,800,860,890 | USE_GRAPHS = 1 | PEER_MAX_BATCH_SIZE = 128 | CPU : SSE3 = 1 | SSSE3 = 1 | AVX = 1 | AVX2 = 1 | F16C = 1 | FMA = 1 | BMI2 = 1 | LLAMAFILE = 1 | OPENMP = 1 | REPACK = 1 |
main: binding port with default address family
main: HTTP server is listening, hostname: 127.0.0.1, port: 30888, http threads: 191
main: loading model
srv load_model: loading model '/data/models/TabbyML/Qwen3-Coder-Next/ggml/model-00001-of-00004.gguf'
llama_model_loader: loaded meta data with 44 key-value pairs and 807 tensors from /data/models/TabbyML/Qwen3-Coder-Next/ggml/model-00001-of-00004.gguf (version GGUF V3 (latest))
llama_model_loader: Dumping metadata keys/values. Note: KV overrides do not apply in this output.
llama_model_loader: - kv 0: general.architecture str = qwen3next
llama_model_loader: - kv 1: general.type str = model
llama_model_loader: - kv 2: general.sampling.top_k i32 = 40
llama_model_loader: - kv 3: general.sampling.top_p f32 = 0.950000
llama_model_loader: - kv 4: general.sampling.temp f32 = 1.000000
llama_model_loader: - kv 5: general.name str = Qwen3 Coder Next 0129
llama_model_loader: - kv 6: general.version str = 0129
llama_model_loader: - kv 7: general.basename str = Qwen3-Coder-Next
llama_model_loader: - kv 8: general.size_label str = 512x2.5B
llama_model_loader: - kv 9: qwen3next.block_count u32 = 48
llama_model_loader: - kv 10: qwen3next.context_length u32 = 262144
llama_model_loader: - kv 11: qwen3next.embedding_length u32 = 2048
llama_model_loader: - kv 12: qwen3next.feed_forward_length u32 = 5120
llama_model_loader: - kv 13: qwen3next.attention.head_count u32 = 16
llama_model_loader: - kv 14: qwen3next.attention.head_count_kv u32 = 2
llama_model_loader: - kv 15: qwen3next.rope.freq_base f32 = 5000000.000000
llama_model_loader: - kv 16: qwen3next.attention.layer_norm_rms_epsilon f32 = 0.000001
llama_model_loader: - kv 17: qwen3next.expert_used_count u32 = 10
llama_model_loader: - kv 18: qwen3next.attention.key_length u32 = 256
llama_model_loader: - kv 19: qwen3next.attention.value_length u32 = 256
llama_model_loader: - kv 20: general.file_type u32 = 1
llama_model_loader: - kv 21: qwen3next.expert_count u32 = 512
llama_model_loader: - kv 22: qwen3next.expert_feed_forward_length u32 = 512
llama_model_loader: - kv 23: qwen3next.expert_shared_feed_forward_length u32 = 512
llama_model_loader: - kv 24: qwen3next.ssm.conv_kernel u32 = 4
llama_model_loader: - kv 25: qwen3next.ssm.state_size u32 = 128
llama_model_loader: - kv 26: qwen3next.ssm.group_count u32 = 16
llama_model_loader: - kv 27: qwen3next.ssm.time_step_rank u32 = 32
llama_model_loader: - kv 28: qwen3next.ssm.inner_size u32 = 4096
llama_model_loader: - kv 29: qwen3next.rope.dimension_count u32 = 64
llama_model_loader: - kv 30: general.quantization_version u32 = 2
llama_model_loader: - kv 31: tokenizer.ggml.model str = gpt2
llama_model_loader: - kv 32: tokenizer.ggml.pre str = qwen2
llama_model_loader: - kv 33: tokenizer.ggml.tokens arr[str,151936] = ["!", "\"", "#", "$", "%", "&", "'", ...
llama_model_loader: - kv 34: tokenizer.ggml.token_type arr[i32,151936] = [1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, ...
llama_model_loader: - kv 35: tokenizer.ggml.merges arr[str,151387] = ["Ġ Ġ", "ĠĠ ĠĠ", "i n", "Ġ t", ...
llama_model_loader: - kv 36: tokenizer.ggml.eos_token_id u32 = 151645
llama_model_loader: - kv 37: tokenizer.ggml.padding_token_id u32 = 151643
llama_model_loader: - kv 38: tokenizer.ggml.bos_token_id u32 = 151643
llama_model_loader: - kv 39: tokenizer.ggml.add_bos_token bool = false
llama_model_loader: - kv 40: tokenizer.chat_template str = {% macro render_extra_keys(json_dict, ...
llama_model_loader: - kv 41: split.no u16 = 0
llama_model_loader: - kv 42: split.count u16 = 4
llama_model_loader: - kv 43: split.tensors.count i32 = 807
llama_model_loader: type f32: 313 tensors
llama_model_loader: type f16: 494 tensors
print_info: file format = GGUF V3 (latest)
print_info: file type = F16
print_info: file size = 148.50 GiB (16.01 BPW)
llama_model_load: error loading model: error loading model architecture: unknown model architecture: 'qwen3next'
llama_model_load_from_file_impl: failed to load model
common_init_from_params: failed to load model '/data/models/TabbyML/Qwen3-Coder-Next/ggml/model-00001-of-00004.gguf'
srv load_model: failed to load model, '/data/models/TabbyML/Qwen3-Coder-Next/ggml/model-00001-of-00004.gguf'
srv operator(): operator(): cleaning up before exit...
Describe the bug
I am attempting to deploy Tabby in an air-gapped environment using Docker, following the tutorial. While the default models are working as expected, I encountered an error
error loading model architecture: unknown model architecture: 'qwen3next'when trying to deploy a Qwen3-Coder-Next GGUF model from Hugging Face .Docker image based on
main-5731104docker run cmd
Qwen3-Coder-Next directory structure
Is the 'qwen3next' model architecture currently not supported?
Information about your version
tabby 0.33.0-dev.0Information about your GPU
Additional context
Log