Instructions to use zai-org/GLM-5.3-Flash with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Libraries

How to use zai-org/GLM-5.3-Flash with Transformers:

# Use a pipeline as a high-level helper
from transformers import pipeline

pipe = pipeline("text-generation", model="zai-org/GLM-5.3-Flash")
messages = [
    {
        "role": "user",
        "content": [
            {"type": "image", "url": "https://huggingface.co/datasets/huggingface/documentation-images/resolve/main/p-blog/candy.JPG"},
            {"type": "text", "text": "What animal is on the candy?"}
        ]
    },
]
pipe(text=messages)

# Load model directly
from transformers import AutoProcessor, AutoModelForMultimodalLM

processor = AutoProcessor.from_pretrained("zai-org/GLM-5.3-Flash")
model = AutoModelForMultimodalLM.from_pretrained("zai-org/GLM-5.3-Flash", device_map="auto")
messages = [
    {
        "role": "user",
        "content": [
            {"type": "image", "url": "https://huggingface.co/datasets/huggingface/documentation-images/resolve/main/p-blog/candy.JPG"},
            {"type": "text", "text": "What animal is on the candy?"}
        ]
    },
]
inputs = processor.apply_chat_template(
	messages,
	add_generation_prompt=True,
	tokenize=True,
	return_dict=True,
	return_tensors="pt",
).to(model.device)

outputs = model.generate(**inputs, max_new_tokens=40)
print(processor.decode(outputs[0][inputs["input_ids"].shape[-1]:]))

Inference
HuggingChat
Notebooks
Google Colab
Kaggle
Local Apps Settings

vLLM

How to use zai-org/GLM-5.3-Flash with vLLM:

Install from pip and serve model

# Install vLLM from pip:
pip install vllm
# Start the vLLM server:
vllm serve "zai-org/GLM-5.3-Flash"
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:8000/v1/chat/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "zai-org/GLM-5.3-Flash",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Use Docker

docker model run hf.co/zai-org/GLM-5.3-Flash

SGLang

How to use zai-org/GLM-5.3-Flash with SGLang:

Install from pip and serve model

# Install SGLang from pip:
pip install sglang
# Start the SGLang server:
python3 -m sglang.launch_server \
    --model-path "zai-org/GLM-5.3-Flash" \
    --host 0.0.0.0 \
    --port 30000
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:30000/v1/chat/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "zai-org/GLM-5.3-Flash",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Use Docker images

docker run --gpus all \
    --shm-size 32g \
    -p 30000:30000 \
    -v ~/.cache/huggingface:/root/.cache/huggingface \
    --env "HF_TOKEN=<secret>" \
    --ipc=host \
    lmsysorg/sglang:latest \
    python3 -m sglang.launch_server \
        --model-path "zai-org/GLM-5.3-Flash" \
        --host 0.0.0.0 \
        --port 30000
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:30000/v1/chat/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "zai-org/GLM-5.3-Flash",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Docker Model Runner
How to use zai-org/GLM-5.3-Flash with Docker Model Runner:
```
docker model run hf.co/zai-org/GLM-5.3-Flash
```

vLLM crash: "pe_dim must be 64 for fp8_ds_mla" on RTX PRO 6000 Blackwell.

#19

by NferKarn - opened 1 day ago

Discussion

NferKarn

1 day ago

•

edited about 23 hours ago

Full Error

Worker_TP0_EP0 pid=1125) INFO 08-26 17:54:39 [cuda.py:532] Using FLASHINFER_MLA_SPARSE_SM120 attention backend out of potential backends: ['FLASHINFER_MLA_SPARSE_SM120'].
(Worker_TP0_EP0 pid=1125) INFO 08-26 17:54:39 [mla_attention.py:475] Using fp8_ds_mla KV cache format for FLASHINFER_MLA_SPARSE_SM120 backend.
...
...
(EngineCore pid=951) ERROR 08-26 16:57:28 [core.py:1355] RuntimeError: Worker failed with error 'concat_and_cache_mla, /workspace/csrc/libtorch_stable/cache_kernels.cu:866, pe_dim must be 64 for fp8_ds_mla', please check the stack trace above for the root cause
2026-08-26 22:27:28.883	
(EngineCore pid=951) ERROR 08-26 16:57:28 [core.py:1355]     raise RuntimeError(
    2026-08-26 22:27:28.883	
(EngineCore pid=951) ERROR 08-26 16:57:28 [core.py:1355]   File "/usr/local/lib/python3.12/dist-packages/vllm/v1/executor/multiproc_executor.py", line 420, in get_response
    2026-08-26 22:27:28.883	
(EngineCore pid=951) ERROR 08-26 16:57:28 [core.py:1355]                               ^^^^^^^^^^^^^^^^^^^
    2026-08-26 22:27:28.883	
(EngineCore pid=951) ERROR 08-26 16:57:28 [core.py:1355]     response = self.aggregate(self.get_response())
    2026-08-26 22:27:28.883	
(EngineCore pid=951) ERROR 08-26 16:57:28 [core.py:1355]   File "/usr/local/lib/python3.12/dist-packages/vllm/v1/executor/multiproc_executor.py", line 102, in _wait_for_response
    2026-08-26 22:27:28.883	
(EngineCore pid=951) ERROR 08-26 16:57:28 [core.py:1355]     raise self._exception
    2026-08-26 22:27:28.883	
(EngineCore pid=951) ERROR 08-26 16:57:28 [core.py:1355]   File "/usr/lib/python3.12/concurrent/futures/_base.py", line 401, in __get_result
    2026-08-26 22:27:28.883	
(EngineCore pid=951) ERROR 08-26 16:57:28 [core.py:1355]            ^^^^^^^^^^^^^^^^^^^
    2026-08-26 22:27:28.883	
(EngineCore pid=951) ERROR 08-26 16:57:28 [core.py:1355]     return self.__get_result()
    2026-08-26 22:27:28.883	
(EngineCore pid=951) ERROR 08-26 16:57:28 [core.py:1355]   File "/usr/lib/python3.12/concurrent/futures/_base.py", line 449, in result
    2026-08-26 22:27:28.883	
(EngineCore pid=951) ERROR 08-26 16:57:28 [core.py:1355]            ^^^^^^^^^^^^^^^^
    2026-08-26 22:27:28.883	
(EngineCore pid=951) ERROR 08-26 16:57:28 [core.py:1355]     return super().result()
    2026-08-26 22:27:28.883	
(EngineCore pid=951) ERROR 08-26 16:57:28 [core.py:1355]   File "/usr/local/lib/python3.12/dist-packages/vllm/v1/executor/multiproc_executor.py", line 98, in result
    2026-08-26 22:27:28.883	
(EngineCore pid=951) ERROR 08-26 16:57:28 [core.py:1355]                                     ^^^^^^^^^^^^^^^
    2026-08-26 22:27:28.883	
(EngineCore pid=951) ERROR 08-26 16:57:28 [core.py:1355]     return future if non_block else future.result()
    2026-08-26 22:27:28.883	
(EngineCore pid=951) ERROR 08-26 16:57:28 [core.py:1355]   File "/usr/local/lib/python3.12/dist-packages/vllm/v1/executor/multiproc_executor.py", line 431, in collective_rpc
    2026-08-26 22:27:28.883	
(EngineCore pid=951) ERROR 08-26 16:57:28 [core.py:1355]            ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
    2026-08-26 22:27:28.883	
(EngineCore pid=951) ERROR 08-26 16:57:28 [core.py:1355]     return self.collective_rpc("determine_available_memory")
    2026-08-26 22:27:28.883	
(EngineCore pid=951) ERROR 08-26 16:57:28 [core.py:1355]   File "/usr/local/lib/python3.12/dist-packages/vllm/v1/executor/abstract.py", line 147, in determine_available_memory
    2026-08-26 22:27:28.883	
(EngineCore pid=951) ERROR 08-26 16:57:28 [core.py:1355]                            ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
    2026-08-26 22:27:28.883	
(EngineCore pid=951) ERROR 08-26 16:57:28 [core.py:1355]     available_gpu_memory = self.model_executor.determine_available_memory()
    2026-08-26 22:27:28.883	
(EngineCore pid=951) ERROR 08-26 16:57:28 [core.py:1355]   File "/usr/local/lib/python3.12/dist-packages/vllm/v1/engine/core.py", line 294, in _initialize_kv_caches
    2026-08-26 22:27:28.883	
(EngineCore pid=951) ERROR 08-26 16:57:28 [core.py:1355]            ^^^^^^^^^^^^^^^^^^^^^
    2026-08-26 22:27:28.883	
(EngineCore pid=951) ERROR 08-26 16:57:28 [core.py:1355]     return func(*args, **kwargs)
    2026-08-26 22:27:28.883	
(EngineCore pid=951) ERROR 08-26 16:57:28 [core.py:1355]   File "/usr/local/lib/python3.12/dist-packages/vllm/tracing/otel.py", line 178, in sync_wrapper
    2026-08-26 22:27:28.883	
(EngineCore pid=951) ERROR 08-26 16:57:28 [core.py:1355]                       ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
    2026-08-26 22:27:28.883	
(EngineCore pid=951) ERROR 08-26 16:57:28 [core.py:1355]     kv_cache_config = self._initialize_kv_caches(vllm_config)
    2026-08-26 22:27:28.883	
(EngineCore pid=951) ERROR 08-26 16:57:28 [core.py:1355]   File "/usr/local/lib/python3.12/dist-packages/vllm/v1/engine/core.py", line 144, in __init__
    2026-08-26 22:27:28.883	
(EngineCore pid=951) ERROR 08-26 16:57:28 [core.py:1355]     super().__init__(
    2026-08-26 22:27:28.883	
(EngineCore pid=951) ERROR 08-26 16:57:28 [core.py:1355]   File "/usr/local/lib/python3.12/dist-packages/vllm/v1/engine/core.py", line 1082, in __init__
    2026-08-26 22:27:28.883	
(EngineCore pid=951) ERROR 08-26 16:57:28 [core.py:1355]            ^^^^^^^^^^^^^^^^^^^^^
    2026-08-26 22:27:28.883	
(EngineCore pid=951) ERROR 08-26 16:57:28 [core.py:1355]     return func(*args, **kwargs)
    2026-08-26 22:27:28.883	
(EngineCore pid=951) ERROR 08-26 16:57:28 [core.py:1355]   File "/usr/local/lib/python3.12/dist-packages/vllm/tracing/otel.py", line 178, in sync_wrapper
    2026-08-26 22:27:28.883	
(EngineCore pid=951) ERROR 08-26 16:57:28 [core.py:1355]                   ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
    2026-08-26 22:27:28.883	
(EngineCore pid=951) ERROR 08-26 16:57:28 [core.py:1355]     engine_core = EngineCoreProc(*args, engine_index=dp_rank, **kwargs)
    2026-08-26 22:27:28.883	
(EngineCore pid=951) ERROR 08-26 16:57:28 [core.py:1355]   File "/usr/local/lib/python3.12/dist-packages/vllm/v1/engine/core.py", line 1324, in run_engine_core
    2026-08-26 22:27:28.883	
(EngineCore pid=951) ERROR 08-26 16:57:28 [core.py:1355] Traceback (most recent call last):
    2026-08-26 22:27:28.883	
(EngineCore pid=951) ERROR 08-26 16:57:28 [core.py:1355] EngineCore failed to start.
    2026-08-26 22:27:28.881	
(Worker_TP0_EP0 pid=1329) ERROR 08-26 16:57:28 [multiproc_executor.py:1033]
    2026-08-26 22:27:28.881	
(Worker_TP0_EP0 pid=1329) ERROR 08-26 16:57:28 [multiproc_executor.py:1033] RuntimeError: concat_and_cache_mla, /workspace/csrc/libtorch_stable/cache_kernels.cu:866, pe_dim must be 64 for fp8_ds_mla
    2026-08-26 22:27:28.881	
(Worker_TP0_EP0 pid=1329) ERROR 08-26 16:57:28 [multiproc_executor.py:1033]            ^^^^^^^^^^^^^^^^^^^^^^^^^
    2026-08-26 22:27:28.881	
(Worker_TP0_EP0 pid=1329) ERROR 08-26 16:57:28 [multiproc_executor.py:1033]     return self._op(*args, **kwargs)
    2026-08-26 22:27:28.881	
(Worker_TP0_EP0 pid=1329) ERROR 08-26 16:57:28 [multiproc_executor.py:1033]   File "/usr/local/lib/python3.12/dist-packages/torch/_ops.py", line 1279, in __call__
    2026-08-26 22:27:28.881	
(Worker_TP0_EP0 pid=1329) ERROR 08-26 16:57:28 [multiproc_executor.py:1033]     torch.ops._C_cache_ops.concat_and_cache_mla(
    2026-08-26 22:27:28.881	
(Worker_TP0_EP0 pid=1329) ERROR 08-26 16:57:28 [multiproc_executor.py:1033]   File "/usr/local/lib/python3.12/dist-packages/vllm/_custom_ops.py", line 2788, in concat_and_cache_mla
    2026-08-26 22:27:28.881	
(Worker_TP0_EP0 pid=1329) ERROR 08-26 16:57:28 [multiproc_executor.py:1033]     ops.concat_and_cache_mla(
    2026-08-26 22:27:28.881	
(Worker_TP0_EP0 pid=1329) ERROR 08-26 16:57:28 [multiproc_executor.py:1033]   File "/usr/local/lib/python3.12/dist-packages/vllm/v1/attention/backend.py", line 1131, in do_kv_cache_update
    2026-08-26 22:27:28.881	
(Worker_TP0_EP0 pid=1329) ERROR 08-26 16:57:28 [multiproc_executor.py:1033]     attn_layer.impl.do_kv_cache_update(  # type: ignore[attr-defined]
    2026-08-26 22:27:28.881	
(Worker_TP0_EP0 pid=1329) ERROR 08-26 16:57:28 [multiproc_executor.py:1033]   File "/usr/local/lib/python3.12/dist-packages/vllm/model_executor/layers/attention/mla_attention.py", line 1218, in unified_mla_kv_cache_update
    2026-08-26 22:27:28.881	
(Worker_TP0_EP0 pid=1329) ERROR 08-26 16:57:28 [multiproc_executor.py:1033]            ^^^^^^^^^^^^^^^^^^^^^^^^^
    2026-08-26 22:27:28.881	
(Worker_TP0_EP0 pid=1329) ERROR 08-26 16:57:28 [multiproc_executor.py:1033]     return self._op(*args, **kwargs)
    2026-08-26 22:27:28.881	
(Worker_TP0_EP0 pid=1329) ERROR 08-26 16:57:28 [multiproc_executor.py:1033]   File "/usr/local/lib/python3.12/dist-packages/torch/_ops.py", line 1279, in __call__
    2026-08-26 22:27:28.881	
(Worker_TP0_EP0 pid=1329) ERROR 08-26 16:57:28 [multiproc_executor.py:1033]                          ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
    2026-08-26 22:27:28.881	
(Worker_TP0_EP0 pid=1329) ERROR 08-26 16:57:28 [multiproc_executor.py:1033]     kv_cache_dummy_dep = torch.ops.vllm.unified_mla_kv_cache_update(
    2026-08-26 22:27:28.881	
(Worker_TP0_EP0 pid=1329) ERROR 08-26 16:57:28 [multiproc_executor.py:1033]   File "/usr/local/lib/python3.12/dist-packages/vllm/model_executor/layers/attention/mla_attention.py", line 701, in forward
    2026-08-26 22:27:28.881	
(Worker_TP0_EP0 pid=1329) ERROR 08-26 16:57:28 [multiproc_executor.py:1033]            ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
    2026-08-26 22:27:28.881	
(Worker_TP0_EP0 pid=1329) ERROR 08-26 16:57:28 [multiproc_executor.py:1033]     return forward_call(*args, **kwargs)
    2026-08-26 22:27:28.881	
(Worker_TP0_EP0 pid=1329) ERROR 08-26 16:57:28 [multiproc_executor.py:1033]   File "/usr/local/lib/python3.12/dist-packages/torch/nn/modules/module.py", line 1789, in _call_impl
    2026-08-26 22:27:28.881	
(Worker_TP0_EP0 pid=1329) ERROR 08-26 16:57:28 [multiproc_executor.py:1033]            ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
    2026-08-26 22:27:28.881	
(Worker_TP0_EP0 pid=1329) ERROR 08-26 16:57:28 [multiproc_executor.py:1033]     return self._call_impl(*args, **kwargs)
    2026-08-26 22:27:28.881	
(Worker_TP0_EP0 pid=1329) ERROR 08-26 16:57:28 [multiproc_executor.py:1033]   File "/usr/local/lib/python3.12/dist-packages/torch/nn/modules/module.py", line 1778, in _wrapped_call_impl
    2026-08-26 22:27:28.881	
(Worker_TP0_EP0 pid=1329) ERROR 08-26 16:57:28 [multiproc_executor.py:1033]                ^^^^^^^^^^^^^^
    2026-08-26 22:27:28.881	
(Worker_TP0_EP0 pid=1329) ERROR 08-26 16:57:28 [multiproc_executor.py:1033]     attn_out = self.mla_attn(
    2026-08-26 22:27:28.881	
(Worker_TP0_EP0 pid=1329) ERROR 08-26 16:57:28 [multiproc_executor.py:1033]   File "/usr/local/lib/python3.12/dist-packages/vllm/model_executor/layers/mla.py", line 234, in forward
    2026-08-26 22:27:28.881	
(Worker_TP0_EP0 pid=1329) ERROR 08-26 16:57:28 [multiproc_executor.py:1033]            ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
    2026-08-26 22:27:28.881	
(Worker_TP0_EP0 pid=1329) ERROR 08-26 16:57:28 [multiproc_executor.py:1033]     return forward_call(*args, **kwargs)
    2026-08-26 22:27:28.881	
(Worker_TP0_EP0 pid=1329) ERROR 08-26 16:57:28 [multiproc_executor.py:1033]   File "/usr/local/lib/python3.12/dist-packages/torch/nn/modules/module.py", line 1789, in _call_impl
    2026-08-26 22:27:28.881	
(Worker_TP0_EP0 pid=1329) ERROR 08-26 16:57:28 [multiproc_executor.py:1033]            ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
    2026-08-26 22:27:28.881	
(Worker_TP0_EP0 pid=1329) ERROR 08-26 16:57:28 [multiproc_executor.py:1033]     return self._call_impl(*args, **kwargs)
    2026-08-26 22:27:28.881	
(Worker_TP0_EP0 pid=1329) ERROR 08-26 16:57:28 [multiproc_executor.py:1033]   File "/usr/local/lib/python3.12/dist-packages/torch/nn/modules/module.py", line 1778, in _wrapped_call_impl
    2026-08-26 22:27:28.881	
(Worker_TP0_EP0 pid=1329) ERROR 08-26 16:57:28 [multiproc_executor.py:1033]            ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
    2026-08-26 22:27:28.881	
(Worker_TP0_EP0 pid=1329) ERROR 08-26 16:57:28 [multiproc_executor.py:1033]     return self.mla_attn(positions, hidden_states)
    2026-08-26 22:27:28.881	
(Worker_TP0_EP0 pid=1329) ERROR 08-26 16:57:28 [multiproc_executor.py:1033]   File "/usr/local/lib/python3.12/dist-packages/vllm/models/glm5next/nvidia/attention.py", line 624, in forward
    2026-08-26 22:27:28.881	
(Worker_TP0_EP0 pid=1329) ERROR 08-26 16:57:28 [multiproc_executor.py:1033]            ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
    2026-08-26 22:27:28.881	
(Worker_TP0_EP0 pid=1329) ERROR 08-26 16:57:28 [multiproc_executor.py:1033]     return forward_call(*args, **kwargs)
    2026-08-26 22:27:28.881	
(Worker_TP0_EP0 pid=1329) ERROR 08-26 16:57:28 [multiproc_executor.py:1033]   File "/usr/local/lib/python3.12/dist-packages/torch/nn/modules/module.py", line 1789, in _call_impl
    2026-08-26 22:27:28.881	
(Worker_TP0_EP0 pid=1329) ERROR 08-26 16:57:28 [multiproc_executor.py:1033]            ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
    2026-08-26 22:27:28.881	
(Worker_TP0_EP0 pid=1329) ERROR 08-26 16:57:28 [multiproc_executor.py:1033]     return self._call_impl(*args, **kwargs)
    2026-08-26 22:27:28.881	
(Worker_TP0_EP0 pid=1329) ERROR 08-26 16:57:28 [multiproc_executor.py:1033]   File "/usr/local/lib/python3.12/dist-packages/torch/nn/modules/module.py", line 1778, in _wrapped_call_impl
    2026-08-26 22:27:28.881	
(Worker_TP0_EP0 pid=1329) ERROR 08-26 16:57:28 [multiproc_executor.py:1033]         ^^^^^^^^^^^^^^^
    2026-08-26 22:27:28.881	
(Worker_TP0_EP0 pid=1329) ERROR 08-26 16:57:28 [multiproc_executor.py:1033]     x = self.self_attn(
    2026-08-26 22:27:28.881	
(Worker_TP0_EP0 pid=1329) ERROR 08-26 16:57:28 [multiproc_executor.py:1033]   File "/usr/local/lib/python3.12/dist-packages/vllm/models/glm5next/nvidia/model.py", line 476, in forward
    2026-08-26 22:27:28.881	
(Worker_TP0_EP0 pid=1329) ERROR 08-26 16:57:28 [multiproc_executor.py:1033]            ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
    2026-08-26 22:27:28.881	
(Worker_TP0_EP0 pid=1329) ERROR 08-26 16:57:28 [multiproc_executor.py:1033]     return forward_call(*args, **kwargs)
    2026-08-26 22:27:28.881	
(Worker_TP0_EP0 pid=1329) ERROR 08-26 16:57:28 [multiproc_executor.py:1033]   File "/usr/local/lib/python3.12/dist-packages/torch/nn/modules/module.py", line 1789, in _call_impl
    2026-08-26 22:27:28.881	
(Worker_TP0_EP0 pid=1329) ERROR 08-26 16:57:28 [multiproc_executor.py:1033]            ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
    2026-08-26 22:27:28.881	
(Worker_TP0_EP0 pid=1329) ERROR 08-26 16:57:28 [multiproc_executor.py:1033]     return self._call_impl(*args, **kwargs)
    2026-08-26 22:27:28.881	
(Worker_TP0_EP0 pid=1329) ERROR 08-26 16:57:28 [multiproc_executor.py:1033]   File "/usr/local/lib/python3.12/dist-packages/torch/nn/modules/module.py", line 1778, in _wrapped_call_impl
    2026-08-26 22:27:28.881	
(Worker_TP0_EP0 pid=1329) ERROR 08-26 16:57:28 [multiproc_executor.py:1033]                                           ^^^^^^
    2026-08-26 22:27:28.881	
(Worker_TP0_EP0 pid=1329) ERROR 08-26 16:57:28 [multiproc_executor.py:1033]     hidden_states, residual, post, comb = layer(
    2026-08-26 22:27:28.881	
(Worker_TP0_EP0 pid=1329) ERROR 08-26 16:57:28 [multiproc_executor.py:1033]   File "/usr/local/lib/python3.12/dist-packages/vllm/models/glm5next/nvidia/model.py", line 707, in forward
    2026-08-26 22:27:28.881	
(Worker_TP0_EP0 pid=1329) ERROR 08-26 16:57:28 [multiproc_executor.py:1033]            ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
    2026-08-26 22:27:28.881	
(Worker_TP0_EP0 pid=1329) ERROR 08-26 16:57:28 [multiproc_executor.py:1033]     return forward_call(*args, **kwargs)
    2026-08-26 22:27:28.881	
(Worker_TP0_EP0 pid=1329) ERROR 08-26 16:57:28 [multiproc_executor.py:1033]   File "/usr/local/lib/python3.12/dist-packages/torch/nn/modules/module.py", line 1789, in _call_impl
    2026-08-26 22:27:28.881	
(Worker_TP0_EP0 pid=1329) ERROR 08-26 16:57:28 [multiproc_executor.py:1033]            ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
    2026-08-26 22:27:28.881	
(Worker_TP0_EP0 pid=1329) ERROR 08-26 16:57:28 [multiproc_executor.py:1033]     return self._call_impl(*args, **kwargs)
    2026-08-26 22:27:28.881	
(Worker_TP0_EP0 pid=1329) ERROR 08-26 16:57:28 [multiproc_executor.py:1033]   File "/usr/local/lib/python3.12/dist-packages/torch/nn/modules/module.py", line 1778, in _wrapped_call_impl
    2026-08-26 22:27:28.881	
(Worker_TP0_EP0 pid=1329) ERROR 08-26 16:57:28 [multiproc_executor.py:1033]                     ^^^^^^^^^^^^^^^^^^^^^^^^^^
    2026-08-26 22:27:28.881	
(Worker_TP0_EP0 pid=1329) ERROR 08-26 16:57:28 [multiproc_executor.py:1033]     hidden_states = self.language_model.model(
    2026-08-26 22:27:28.881	
(Worker_TP0_EP0 pid=1329) ERROR 08-26 16:57:28 [multiproc_executor.py:1033]   File "/usr/local/lib/python3.12/dist-packages/vllm/model_executor/models/glm4_1v.py", line 2327, in forward
    2026-08-26 22:27:28.881	
(Worker_TP0_EP0 pid=1329) ERROR 08-26 16:57:28 [multiproc_executor.py:1033]            ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
    2026-08-26 22:27:28.881	
(Worker_TP0_EP0 pid=1329) ERROR 08-26 16:57:28 [multiproc_executor.py:1033]     return forward_call(*args, **kwargs)
    2026-08-26 22:27:28.881	
(Worker_TP0_EP0 pid=1329) ERROR 08-26 16:57:28 [multiproc_executor.py:1033]   File "/usr/local/lib/python3.12/dist-packages/torch/nn/modules/module.py", line 1789, in _call_impl
    2026-08-26 22:27:28.881	
(Worker_TP0_EP0 pid=1329) ERROR 08-26 16:57:28 [multiproc_executor.py:1033]            ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
    2026-08-26 22:27:28.881	
(Worker_TP0_EP0 pid=1329) ERROR 08-26 16:57:28 [multiproc_executor.py:1033]     return self._call_impl(*args, **kwargs)
    2026-08-26 22:27:28.881	
(Worker_TP0_EP0 pid=1329) ERROR 08-26 16:57:28 [multiproc_executor.py:1033]   File "/usr/local/lib/python3.12/dist-packages/torch/nn/modules/module.py", line 1778, in _wrapped_call_impl
    2026-08-26 22:27:28.881	
(Worker_TP0_EP0 pid=1329) ERROR 08-26 16:57:28 [multiproc_executor.py:1033]                    ^^^^^^^^^^^^^^^^^^^^^
    2026-08-26 22:27:28.881	
(Worker_TP0_EP0 pid=1329) ERROR 08-26 16:57:28 [multiproc_executor.py:1033]     model_output = model(**model_inputs)
    2026-08-26 22:27:28.881	
(Worker_TP0_EP0 pid=1329) ERROR 08-26 16:57:28 [multiproc_executor.py:1033]   File "/usr/local/lib/python3.12/dist-packages/vllm/v1/worker/gpu/cudagraph_utils.py", line 690, in forward_fn
    2026-08-26 22:27:28.881	
(Worker_TP0_EP0 pid=1329) ERROR 08-26 16:57:28 [multiproc_executor.py:1033]     forward_fn(CUDAGraphMode.NONE)
    2026-08-26 22:27:28.881	
(Worker_TP0_EP0 pid=1329) ERROR 08-26 16:57:28 [multiproc_executor.py:1033]   File "/usr/local/lib/python3.12/dist-packages/vllm/v1/worker/gpu/cudagraph_utils.py", line 339, in capture
    2026-08-26 22:27:28.881	
(Worker_TP0_EP0 pid=1329) ERROR 08-26 16:57:28 [multiproc_executor.py:1033]            ^^^^^^^^^^^^^^^^^^^^^
    2026-08-26 22:27:28.881	
(Worker_TP0_EP0 pid=1329) ERROR 08-26 16:57:28 [multiproc_executor.py:1033]     return func(*args, **kwargs)
    2026-08-26 22:27:28.881	
(Worker_TP0_EP0 pid=1329) ERROR 08-26 16:57:28 [multiproc_executor.py:1033]   File "/usr/local/lib/python3.12/dist-packages/torch/utils/_contextlib.py", line 124, in decorate_context
    2026-08-26 22:27:28.881	
(Worker_TP0_EP0 pid=1329) ERROR 08-26 16:57:28 [multiproc_executor.py:1033]     CudaGraphManager.capture(
    2026-08-26 22:27:28.881	
(Worker_TP0_EP0 pid=1329) ERROR 08-26 16:57:28 [multiproc_executor.py:1033]   File "/usr/local/lib/python3.12/dist-packages/vllm/v1/worker/gpu/cudagraph_utils.py", line 429, in profile_memory
    2026-08-26 22:27:28.881	
(Worker_TP0_EP0 pid=1329) ERROR 08-26 16:57:28 [multiproc_executor.py:1033]            ^^^^^^^^^^^^^^^^^^^^^
    2026-08-26 22:27:28.881	
(Worker_TP0_EP0 pid=1329) ERROR 08-26 16:57:28 [multiproc_executor.py:1033]     return func(*args, **kwargs)
    2026-08-26 22:27:28.881	
(Worker_TP0_EP0 pid=1329) ERROR 08-26 16:57:28 [multiproc_executor.py:1033]   File "/usr/local/lib/python3.12/dist-packages/torch/utils/_contextlib.py", line 124, in decorate_context
    2026-08-26 22:27:28.881	
(Worker_TP0_EP0 pid=1329) ERROR 08-26 16:57:28 [multiproc_executor.py:1033]            ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
    2026-08-26 22:27:28.881	
(Worker_TP0_EP0 pid=1329) ERROR 08-26 16:57:28 [multiproc_executor.py:1033]     return super().profile_memory(create_forward_fn)
    2026-08-26 22:27:28.881	
(Worker_TP0_EP0 pid=1329) ERROR 08-26 16:57:28 [multiproc_executor.py:1033]   File "/usr/local/lib/python3.12/dist-packages/vllm/v1/worker/gpu/cudagraph_utils.py", line 608, in profile_memory
    2026-08-26 22:27:28.881	
(Worker_TP0_EP0 pid=1329) ERROR 08-26 16:57:28 [multiproc_executor.py:1033]                 ^^^^^^^^^^^^^^^^^^^^^^^
    2026-08-26 22:27:28.881	
(Worker_TP0_EP0 pid=1329) ERROR 08-26 16:57:28 [multiproc_executor.py:1033]     estimates = manager.profile_memory(
    2026-08-26 22:27:28.881	
(Worker_TP0_EP0 pid=1329) ERROR 08-26 16:57:28 [multiproc_executor.py:1033]   File "/usr/local/lib/python3.12/dist-packages/vllm/v1/worker/gpu/model_runner.py", line 929, in profile_cudagraph_memory
    2026-08-26 22:27:28.881	
(Worker_TP0_EP0 pid=1329) ERROR 08-26 16:57:28 [multiproc_executor.py:1033]            ^^^^^^^^^^^^^^^^^^^^^
    2026-08-26 22:27:28.881	
(Worker_TP0_EP0 pid=1329) ERROR 08-26 16:57:28 [multiproc_executor.py:1033]     return func(*args, **kwargs)
    2026-08-26 22:27:28.881	
(Worker_TP0_EP0 pid=1329) ERROR 08-26 16:57:28 [multiproc_executor.py:1033]   File "/usr/local/lib/python3.12/dist-packages/torch/utils/_contextlib.py", line 124, in decorate_context
    2026-08-26 22:27:28.881	
(Worker_TP0_EP0 pid=1329) ERROR 08-26 16:57:28 [multiproc_executor.py:1033]                                 ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
    2026-08-26 22:27:28.881	
(Worker_TP0_EP0 pid=1329) ERROR 08-26 16:57:28 [multiproc_executor.py:1033]     cudagraph_memory_estimate = self.model_runner.profile_cudagraph_memory()
    2026-08-26 22:27:28.881	
(Worker_TP0_EP0 pid=1329) ERROR 08-26 16:57:28 [multiproc_executor.py:1033]   File "/usr/local/lib/python3.12/dist-packages/vllm/v1/worker/gpu_worker.py", line 518, in determine_available_memory
    2026-08-26 22:27:28.881	
(Worker_TP0_EP0 pid=1329) ERROR 08-26 16:57:28 [multiproc_executor.py:1033]            ^^^^^^^^^^^^^^^^^^^^^
    2026-08-26 22:27:28.881	
(Worker_TP0_EP0 pid=1329) ERROR 08-26 16:57:28 [multiproc_executor.py:1033]     return func(*args, **kwargs)
    2026-08-26 22:27:28.881	
(Worker_TP0_EP0 pid=1329) ERROR 08-26 16:57:28 [multiproc_executor.py:1033]   File "/usr/local/lib/python3.12/dist-packages/torch/utils/_contextlib.py", line 124, in decorate_context
    2026-08-26 22:27:28.881	
(Worker_TP0_EP0 pid=1329) ERROR 08-26 16:57:28 [multiproc_executor.py:1033]              ^^^^^^^^^^^^^^^^^^^^^
    2026-08-26 22:27:28.881	
(Worker_TP0_EP0 pid=1329) ERROR 08-26 16:57:28 [multiproc_executor.py:1033]     output = func(*args, **kwargs)
    2026-08-26 22:27:28.881	
(Worker_TP0_EP0 pid=1329) ERROR 08-26 16:57:28 [multiproc_executor.py:1033]   File "/usr/local/lib/python3.12/dist-packages/vllm/v1/executor/multiproc_executor.py", line 1025, in worker_busy_loop
    2026-08-26 22:27:28.881	
(Worker_TP0_EP0 pid=1329) ERROR 08-26 16:57:28 [multiproc_executor.py:1033] Traceback (most recent call last):
    2026-08-26 22:27:28.881	
(Worker_TP0_EP0 pid=1329) ERROR 08-26 16:57:28 [multiproc_executor.py:1033] RuntimeError: concat_and_cache_mla, /workspace/csrc/libtorch_stable/cache_kernels.cu:866, pe_dim must be 64 for fp8_ds_mla
    2026-08-26 22:27:28.881	
(Worker_TP0_EP0 pid=1329) ERROR 08-26 16:57:28 [multiproc_executor.py:1033]            ^^^^^^^^^^^^^^^^^^^^^^^^^
    2026-08-26 22:27:28.881	
(Worker_TP0_EP0 pid=1329) ERROR 08-26 16:57:28 [multiproc_executor.py:1033]     return self._op(*args, **kwargs)
    2026-08-26 22:27:28.881	
(Worker_TP0_EP0 pid=1329) ERROR 08-26 16:57:28 [multiproc_executor.py:1033]   File "/usr/local/lib/python3.12/dist-packages/torch/_ops.py", line 1279, in __call__
    2026-08-26 22:27:28.881	
(Worker_TP0_EP0 pid=1329) ERROR 08-26 16:57:28 [multiproc_executor.py:1033]     torch.ops._C_cache_ops.concat_and_cache_mla(
    2026-08-26 22:27:28.881	
(Worker_TP0_EP0 pid=1329) ERROR 08-26 16:57:28 [multiproc_executor.py:1033]   File "/usr/local/lib/python3.12/dist-packages/vllm/_custom_ops.py", line 2788, in concat_and_cache_mla
    2026-08-26 22:27:28.881	
(Worker_TP0_EP0 pid=1329) ERROR 08-26 16:57:28 [multiproc_executor.py:1033]     ops.concat_and_cache_mla(
    2026-08-26 22:27:28.881	
(Worker_TP0_EP0 pid=1329) ERROR 08-26 16:57:28 [multiproc_executor.py:1033]   File "/usr/local/lib/python3.12/dist-packages/vllm/v1/attention/backend.py", line 1131, in do_kv_cache_update
    2026-08-26 22:27:28.881	
(Worker_TP0_EP0 pid=1329) ERROR 08-26 16:57:28 [multiproc_executor.py:1033]     attn_layer.impl.do_kv_cache_update(  # type: ignore[attr-defined]
    2026-08-26 22:27:28.881	
(Worker_TP0_EP0 pid=1329) ERROR 08-26 16:57:28 [multiproc_executor.py:1033]   File "/usr/local/lib/python3.12/dist-packages/vllm/model_executor/layers/attention/mla_attention.py", line 1218, in unified_mla_kv_cache_update
    2026-08-26 22:27:28.881	
(Worker_TP0_EP0 pid=1329) ERROR 08-26 16:57:28 [multiproc_executor.py:1033]            ^^^^^^^^^^^^^^^^^^^^^^^^^
    2026-08-26 22:27:28.881	
(Worker_TP0_EP0 pid=1329) ERROR 08-26 16:57:28 [multiproc_executor.py:1033]     return self._op(*args, **kwargs)
    2026-08-26 22:27:28.881	
(Worker_TP0_EP0 pid=1329) ERROR 08-26 16:57:28 [multiproc_executor.py:1033]   File "/usr/local/lib/python3.12/dist-packages/torch/_ops.py", line 1279, in __call__
    2026-08-26 22:27:28.881	
(Worker_TP0_EP0 pid=1329) ERROR 08-26 16:57:28 [multiproc_executor.py:1033]                          ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
    2026-08-26 22:27:28.881	
(Worker_TP0_EP0 pid=1329) ERROR 08-26 16:57:28 [multiproc_executor.py:1033]     kv_cache_dummy_dep = torch.ops.vllm.unified_mla_kv_cache_update(
    2026-08-26 22:27:28.881	
(Worker_TP0_EP0 pid=1329) ERROR 08-26 16:57:28 [multiproc_executor.py:1033]   File "/usr/local/lib/python3.12/dist-packages/vllm/model_executor/layers/attention/mla_attention.py", line 701, in forward
    2026-08-26 22:27:28.881	
(Worker_TP0_EP0 pid=1329) ERROR 08-26 16:57:28 [multiproc_executor.py:1033]            ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
    2026-08-26 22:27:28.881	
(Worker_TP0_EP0 pid=1329) ERROR 08-26 16:57:28 [multiproc_executor.py:1033]     return forward_call(*args, **kwargs)
    2026-08-26 22:27:28.881	
(Worker_TP0_EP0 pid=1329) ERROR 08-26 16:57:28 [multiproc_executor.py:1033]   File "/usr/local/lib/python3.12/dist-packages/torch/nn/modules/module.py", line 1789, in _call_impl
    2026-08-26 22:27:28.881	
(Worker_TP0_EP0 pid=1329) ERROR 08-26 16:57:28 [multiproc_executor.py:1033]            ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
    2026-08-26 22:27:28.881	
(Worker_TP0_EP0 pid=1329) ERROR 08-26 16:57:28 [multiproc_executor.py:1033]     return self._call_impl(*args, **kwargs)
    2026-08-26 22:27:28.881	
(Worker_TP0_EP0 pid=1329) ERROR 08-26 16:57:28 [multiproc_executor.py:1033]   File "/usr/local/lib/python3.12/dist-packages/torch/nn/modules/module.py", line 1778, in _wrapped_call_impl
    2026-08-26 22:27:28.881	
(Worker_TP0_EP0 pid=1329) ERROR 08-26 16:57:28 [multiproc_executor.py:1033]                ^^^^^^^^^^^^^^
    2026-08-26 22:27:28.881	
(Worker_TP0_EP0 pid=1329) ERROR 08-26 16:57:28 [multiproc_executor.py:1033]     attn_out = self.mla_attn(
    2026-08-26 22:27:28.881	
(Worker_TP0_EP0 pid=1329) ERROR 08-26 16:57:28 [multiproc_executor.py:1033]   File "/usr/local/lib/python3.12/dist-packages/vllm/model_executor/layers/mla.py", line 234, in forward
    2026-08-26 22:27:28.881	
(Worker_TP0_EP0 pid=1329) ERROR 08-26 16:57:28 [multiproc_executor.py:1033]            ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
    2026-08-26 22:27:28.881	
(Worker_TP0_EP0 pid=1329) ERROR 08-26 16:57:28 [multiproc_executor.py:1033]     return forward_call(*args, **kwargs)
    2026-08-26 22:27:28.881	
(Worker_TP0_EP0 pid=1329) ERROR 08-26 16:57:28 [multiproc_executor.py:1033]   File "/usr/local/lib/python3.12/dist-packages/torch/nn/modules/module.py", line 1789, in _call_impl
    2026-08-26 22:27:28.881	
(Worker_TP0_EP0 pid=1329) ERROR 08-26 16:57:28 [multiproc_executor.py:1033]            ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
    2026-08-26 22:27:28.881	
(Worker_TP0_EP0 pid=1329) ERROR 08-26 16:57:28 [multiproc_executor.py:1033]     return self._call_impl(*args, **kwargs)
    2026-08-26 22:27:28.881	
(Worker_TP0_EP0 pid=1329) ERROR 08-26 16:57:28 [multiproc_executor.py:1033]   File "/usr/local/lib/python3.12/dist-packages/torch/nn/modules/module.py", line 1778, in _wrapped_call_impl
    2026-08-26 22:27:28.881	
(Worker_TP0_EP0 pid=1329) ERROR 08-26 16:57:28 [multiproc_executor.py:1033]            ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
    2026-08-26 22:27:28.881	
(Worker_TP0_EP0 pid=1329) ERROR 08-26 16:57:28 [multiproc_executor.py:1033]     return self.mla_attn(positions, hidden_states)
    2026-08-26 22:27:28.881	
(Worker_TP0_EP0 pid=1329) ERROR 08-26 16:57:28 [multiproc_executor.py:1033]   File "/usr/local/lib/python3.12/dist-packages/vllm/models/glm5next/nvidia/attention.py", line 624, in forward
    2026-08-26 22:27:28.881	
(Worker_TP0_EP0 pid=1329) ERROR 08-26 16:57:28 [multiproc_executor.py:1033]            ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
    2026-08-26 22:27:28.881	
(Worker_TP0_EP0 pid=1329) ERROR 08-26 16:57:28 [multiproc_executor.py:1033]     return forward_call(*args, **kwargs)
    2026-08-26 22:27:28.881	
(Worker_TP0_EP0 pid=1329) ERROR 08-26 16:57:28 [multiproc_executor.py:1033]   File "/usr/local/lib/python3.12/dist-packages/torch/nn/modules/module.py", line 1789, in _call_impl
    2026-08-26 22:27:28.881	
(Worker_TP0_EP0 pid=1329) ERROR 08-26 16:57:28 [multiproc_executor.py:1033]            ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
    2026-08-26 22:27:28.881	
(Worker_TP0_EP0 pid=1329) ERROR 08-26 16:57:28 [multiproc_executor.py:1033]     return self._call_impl(*args, **kwargs)
    2026-08-26 22:27:28.881	
(Worker_TP0_EP0 pid=1329) ERROR 08-26 16:57:28 [multiproc_executor.py:1033]   File "/usr/local/lib/python3.12/dist-packages/torch/nn/modules/module.py", line 1778, in _wrapped_call_impl
    2026-08-26 22:27:28.881	
(Worker_TP0_EP0 pid=1329) ERROR 08-26 16:57:28 [multiproc_executor.py:1033]         ^^^^^^^^^^^^^^^
    2026-08-26 22:27:28.881	
(Worker_TP0_EP0 pid=1329) ERROR 08-26 16:57:28 [multiproc_executor.py:1033]     x = self.self_attn(
    2026-08-26 22:27:28.881	
(Worker_TP0_EP0 pid=1329) ERROR 08-26 16:57:28 [multiproc_executor.py:1033]   File "/usr/local/lib/python3.12/dist-packages/vllm/models/glm5next/nvidia/model.py", line 476, in forward
    2026-08-26 22:27:28.881	
(Worker_TP0_EP0 pid=1329) ERROR 08-26 16:57:28 [multiproc_executor.py:1033]            ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
    2026-08-26 22:27:28.881	
(Worker_TP0_EP0 pid=1329) ERROR 08-26 16:57:28 [multiproc_executor.py:1033]     return forward_call(*args, **kwargs)
    2026-08-26 22:27:28.881	
(Worker_TP0_EP0 pid=1329) ERROR 08-26 16:57:28 [multiproc_executor.py:1033]   File "/usr/local/lib/python3.12/dist-packages/torch/nn/modules/module.py", line 1789, in _call_impl
    2026-08-26 22:27:28.881	
(Worker_TP0_EP0 pid=1329) ERROR 08-26 16:57:28 [multiproc_executor.py:1033]            ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
    2026-08-26 22:27:28.881	
(Worker_TP0_EP0 pid=1329) ERROR 08-26 16:57:28 [multiproc_executor.py:1033]     return self._call_impl(*args, **kwargs)
    2026-08-26 22:27:28.881	
(Worker_TP0_EP0 pid=1329) ERROR 08-26 16:57:28 [multiproc_executor.py:1033]   File "/usr/local/lib/python3.12/dist-packages/torch/nn/modules/module.py", line 1778, in _wrapped_call_impl
    2026-08-26 22:27:28.881	
(Worker_TP0_EP0 pid=1329) ERROR 08-26 16:57:28 [multiproc_executor.py:1033]                                           ^^^^^^
    2026-08-26 22:27:28.881	
(Worker_TP0_EP0 pid=1329) ERROR 08-26 16:57:28 [multiproc_executor.py:1033]     hidden_states, residual, post, comb = layer(
    2026-08-26 22:27:28.881	
(Worker_TP0_EP0 pid=1329) ERROR 08-26 16:57:28 [multiproc_executor.py:1033]   File "/usr/local/lib/python3.12/dist-packages/vllm/models/glm5next/nvidia/model.py", line 707, in forward
    2026-08-26 22:27:28.881	
(Worker_TP0_EP0 pid=1329) ERROR 08-26 16:57:28 [multiproc_executor.py:1033]            ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
    2026-08-26 22:27:28.881	
(Worker_TP0_EP0 pid=1329) ERROR 08-26 16:57:28 [multiproc_executor.py:1033]     return forward_call(*args, **kwargs)
    2026-08-26 22:27:28.881	
(Worker_TP0_EP0 pid=1329) ERROR 08-26 16:57:28 [multiproc_executor.py:1033]   File "/usr/local/lib/python3.12/dist-packages/torch/nn/modules/module.py", line 1789, in _call_impl
    2026-08-26 22:27:28.881	
(Worker_TP0_EP0 pid=1329) ERROR 08-26 16:57:28 [multiproc_executor.py:1033]            ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
    2026-08-26 22:27:28.881	
(Worker_TP0_EP0 pid=1329) ERROR 08-26 16:57:28 [multiproc_executor.py:1033]     return self._call_impl(*args, **kwargs)
    2026-08-26 22:27:28.881	
(Worker_TP0_EP0 pid=1329) ERROR 08-26 16:57:28 [multiproc_executor.py:1033]   File "/usr/local/lib/python3.12/dist-packages/torch/nn/modules/module.py", line 1778, in _wrapped_call_impl
    2026-08-26 22:27:28.881	
(Worker_TP0_EP0 pid=1329) ERROR 08-26 16:57:28 [multiproc_executor.py:1033]                     ^^^^^^^^^^^^^^^^^^^^^^^^^^
    2026-08-26 22:27:28.881	
(Worker_TP0_EP0 pid=1329) ERROR 08-26 16:57:28 [multiproc_executor.py:1033]     hidden_states = self.language_model.model(
    2026-08-26 22:27:28.881	
(Worker_TP0_EP0 pid=1329) ERROR 08-26 16:57:28 [multiproc_executor.py:1033]   File "/usr/local/lib/python3.12/dist-packages/vllm/model_executor/models/glm4_1v.py", line 2327, in forward
    2026-08-26 22:27:28.881	
(Worker_TP0_EP0 pid=1329) ERROR 08-26 16:57:28 [multiproc_executor.py:1033]            ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
    2026-08-26 22:27:28.881	
(Worker_TP0_EP0 pid=1329) ERROR 08-26 16:57:28 [multiproc_executor.py:1033]     return forward_call(*args, **kwargs)
    2026-08-26 22:27:28.881	
(Worker_TP0_EP0 pid=1329) ERROR 08-26 16:57:28 [multiproc_executor.py:1033]   File "/usr/local/lib/python3.12/dist-packages/torch/nn/modules/module.py", line 1789, in _call_impl
    2026-08-26 22:27:28.881	
(Worker_TP0_EP0 pid=1329) ERROR 08-26 16:57:28 [multiproc_executor.py:1033]            ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
    2026-08-26 22:27:28.881	
(Worker_TP0_EP0 pid=1329) ERROR 08-26 16:57:28 [multiproc_executor.py:1033]     return self._call_impl(*args, **kwargs)
    2026-08-26 22:27:28.881	
(Worker_TP0_EP0 pid=1329) ERROR 08-26 16:57:28 [multiproc_executor.py:1033]   File "/usr/local/lib/python3.12/dist-packages/torch/nn/modules/module.py", line 1778, in _wrapped_call_impl
    2026-08-26 22:27:28.881	
(Worker_TP0_EP0 pid=1329) ERROR 08-26 16:57:28 [multiproc_executor.py:1033]                    ^^^^^^^^^^^^^^^^^^^^^
    2026-08-26 22:27:28.881	
(Worker_TP0_EP0 pid=1329) ERROR 08-26 16:57:28 [multiproc_executor.py:1033]     model_output = model(**model_inputs)
    2026-08-26 22:27:28.881	
(Worker_TP0_EP0 pid=1329) ERROR 08-26 16:57:28 [multiproc_executor.py:1033]   File "/usr/local/lib/python3.12/dist-packages/vllm/v1/worker/gpu/cudagraph_utils.py", line 690, in forward_fn
    2026-08-26 22:27:28.881	
(Worker_TP0_EP0 pid=1329) ERROR 08-26 16:57:28 [multiproc_executor.py:1033]     forward_fn(CUDAGraphMode.NONE)
    2026-08-26 22:27:28.881	
(Worker_TP0_EP0 pid=1329) ERROR 08-26 16:57:28 [multiproc_executor.py:1033]   File "/usr/local/lib/python3.12/dist-packages/vllm/v1/worker/gpu/cudagraph_utils.py", line 339, in capture
    2026-08-26 22:27:28.881	
(Worker_TP0_EP0 pid=1329) ERROR 08-26 16:57:28 [multiproc_executor.py:1033]            ^^^^^^^^^^^^^^^^^^^^^
    2026-08-26 22:27:28.881	
(Worker_TP0_EP0 pid=1329) ERROR 08-26 16:57:28 [multiproc_executor.py:1033]     return func(*args, **kwargs)
    2026-08-26 22:27:28.881	
(Worker_TP0_EP0 pid=1329) ERROR 08-26 16:57:28 [multiproc_executor.py:1033]   File "/usr/local/lib/python3.12/dist-packages/torch/utils/_contextlib.py", line 124, in decorate_context
    2026-08-26 22:27:28.881	
(Worker_TP0_EP0 pid=1329) ERROR 08-26 16:57:28 [multiproc_executor.py:1033]     CudaGraphManager.capture(
    2026-08-26 22:27:28.881	
(Worker_TP0_EP0 pid=1329) ERROR 08-26 16:57:28 [multiproc_executor.py:1033]   File "/usr/local/lib/python3.12/dist-packages/vllm/v1/worker/gpu/cudagraph_utils.py", line 429, in profile_memory
    2026-08-26 22:27:28.881	
(Worker_TP0_EP0 pid=1329) ERROR 08-26 16:57:28 [multiproc_executor.py:1033]            ^^^^^^^^^^^^^^^^^^^^^
    2026-08-26 22:27:28.881	
(Worker_TP0_EP0 pid=1329) ERROR 08-26 16:57:28 [multiproc_executor.py:1033]     return func(*args, **kwargs)
    2026-08-26 22:27:28.881	
(Worker_TP0_EP0 pid=1329) ERROR 08-26 16:57:28 [multiproc_executor.py:1033]   File "/usr/local/lib/python3.12/dist-packages/torch/utils/_contextlib.py", line 124, in decorate_context
    2026-08-26 22:27:28.881	
(Worker_TP0_EP0 pid=1329) ERROR 08-26 16:57:28 [multiproc_executor.py:1033]            ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
    2026-08-26 22:27:28.881	
(Worker_TP0_EP0 pid=1329) ERROR 08-26 16:57:28 [multiproc_executor.py:1033]     return super().profile_memory(create_forward_fn)
    2026-08-26 22:27:28.881	
(Worker_TP0_EP0 pid=1329) ERROR 08-26 16:57:28 [multiproc_executor.py:1033]   File "/usr/local/lib/python3.12/dist-packages/vllm/v1/worker/gpu/cudagraph_utils.py", line 608, in profile_memory
    2026-08-26 22:27:28.881	
(Worker_TP0_EP0 pid=1329) ERROR 08-26 16:57:28 [multiproc_executor.py:1033]                 ^^^^^^^^^^^^^^^^^^^^^^^
    2026-08-26 22:27:28.881	
(Worker_TP0_EP0 pid=1329) ERROR 08-26 16:57:28 [multiproc_executor.py:1033]     estimates = manager.profile_memory(
    2026-08-26 22:27:28.881	
(Worker_TP0_EP0 pid=1329) ERROR 08-26 16:57:28 [multiproc_executor.py:1033]   File "/usr/local/lib/python3.12/dist-packages/vllm/v1/worker/gpu/model_runner.py", line 929, in profile_cudagraph_memory
    2026-08-26 22:27:28.881	
(Worker_TP0_EP0 pid=1329) ERROR 08-26 16:57:28 [multiproc_executor.py:1033]            ^^^^^^^^^^^^^^^^^^^^^
    2026-08-26 22:27:28.881	
(Worker_TP0_EP0 pid=1329) ERROR 08-26 16:57:28 [multiproc_executor.py:1033]     return func(*args, **kwargs)
    2026-08-26 22:27:28.881	
(Worker_TP0_EP0 pid=1329) ERROR 08-26 16:57:28 [multiproc_executor.py:1033]   File "/usr/local/lib/python3.12/dist-packages/torch/utils/_contextlib.py", line 124, in decorate_context
    2026-08-26 22:27:28.881	
(Worker_TP0_EP0 pid=1329) ERROR 08-26 16:57:28 [multiproc_executor.py:1033]                                 ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
    2026-08-26 22:27:28.881	
(Worker_TP0_EP0 pid=1329) ERROR 08-26 16:57:28 [multiproc_executor.py:1033]     cudagraph_memory_estimate = self.model_runner.profile_cudagraph_memory()
    2026-08-26 22:27:28.881	
(Worker_TP0_EP0 pid=1329) ERROR 08-26 16:57:28 [multiproc_executor.py:1033]   File "/usr/local/lib/python3.12/dist-packages/vllm/v1/worker/gpu_worker.py", line 518, in determine_available_memory
    2026-08-26 22:27:28.881	
(Worker_TP0_EP0 pid=1329) ERROR 08-26 16:57:28 [multiproc_executor.py:1033]            ^^^^^^^^^^^^^^^^^^^^^
    2026-08-26 22:27:28.881	
(Worker_TP0_EP0 pid=1329) ERROR 08-26 16:57:28 [multiproc_executor.py:1033]     return func(*args, **kwargs)
    2026-08-26 22:27:28.881	
(Worker_TP0_EP0 pid=1329) ERROR 08-26 16:57:28 [multiproc_executor.py:1033]   File "/usr/local/lib/python3.12/dist-packages/torch/utils/_contextlib.py", line 124, in decorate_context
    2026-08-26 22:27:28.881	
(Worker_TP0_EP0 pid=1329) ERROR 08-26 16:57:28 [multiproc_executor.py:1033]              ^^^^^^^^^^^^^^^^^^^^^
    2026-08-26 22:27:28.881	
(Worker_TP0_EP0 pid=1329) ERROR 08-26 16:57:28 [multiproc_executor.py:1033]     output = func(*args, **kwargs)
    2026-08-26 22:27:28.881	
(Worker_TP0_EP0 pid=1329) ERROR 08-26 16:57:28 [multiproc_executor.py:1033]   File "/usr/local/lib/python3.12/dist-packages/vllm/v1/executor/multiproc_executor.py", line 1025, in worker_busy_loop
    2026-08-26 22:27:28.881	
(Worker_TP0_EP0 pid=1329) ERROR 08-26 16:57:28 [multiproc_executor.py:1033] Traceback (most recent call last):
    2026-08-26 22:27:28.881	
(Worker_TP0_EP0 pid=1329) ERROR 08-26 16:57:28 [multiproc_executor.py:1033] WorkerProc hit an exception.

Hardware / environment

GPUs: 4× NVIDIA RTX 6000 Pro Blackwell (Server Edition), SM120
Model: zai-org/GLM-5.3-Flash (default native FP8 checkpoint)
vLLM image: vllm/vllm-openai:glm53-flash

What I tried

Removed --kv-cache-dtype fp8 (and set it to auto) → same error.
Forcing bfloat16 throws

Worker_TP3_EP3 pid=1759) ERROR 08-26 17:59:09 [multiproc_executor.py:927] ValueError: No valid attention backend found for cuda with AttentionSelectorConfig(head_size=512, dtype=torch.bfloat16, kv_cache_dtype=bfloat16, block_size=None, use_mla=True, has_sink=False, use_sparse=True, use_mm_prefix=False, use_per_head_quant_scales=False, attn_type=AttentionType.DECODER, has_sliding_window=False, use_non_causal=False, use_batch_invariant=False, use_kv_connector=True, use_adaptive_verification=False, use_pcp=False). Reasons: {TRITON_MLA: [sparse not supported], FLASHINFER_MLA_SPARSE_SM120: [kv_cache_dtype not supported, kv_cache_dtype not supported]}

Args

--served-model-name GLM-5.3-Flash
--max-num-batched-tokens 16384
--max-num-seqs 256
--max-model-len 262144
--enable-expert-parallel
--tensor-parallel-size 4
--enable-auto-tool-choice
--tool-call-parser glm47
--reasoning-parser glm45
--enable-cumem-allocator
--kv-transfer-config '{"kv_connector":"SimpleCPUOffloadConnector","kv_role":"kv_both","kv_connector_extra_config":{"cpu_bytes_to_use_per_rank":137438953472,"lazy_offload":false}}'
--api-server-count 1
--async-scheduling
--scheduling-policy priority