vllm-project/vllm-omni
[Bug] FP8 quantization and HSDP cannot be enabled simultaneously for Diffusion models (dynamic_scaled_fp8_quant called on CPU tensor)
Chiusa
#2159 aperta il 25 mar 2026
bughelp wantedmedium priority
Metriche repository
- Star
- (4990 stelle)
- Metriche merge PR
- (Metriche PR in attesa)
Descrizione
Your current environment
==============================
System Info
==============================
OS : Ubuntu 24.04.1 LTS (x86_64)
GCC version : (Ubuntu 13.3.0-6ubuntu2~24.04) 13.3.0
Clang version : Could not collect
CMake version : version 3.31.4
Libc version : glibc-2.39
==============================
PyTorch Info
==============================
PyTorch version : 2.10.0+cu129
Is debug build : False
CUDA used to build PyTorch : 12.9
ROCM used to build PyTorch : N/A
==============================
Python Environment
==============================
Python version : 3.12.3 (main, Nov 6 2024, 18:32:19) [GCC 13.2.0] (64-bit runtime)
Python platform : Linux-4.15.0-213-generic-x86_64-with-glibc2.39
==============================
CUDA / GPU Info
==============================
Is CUDA available : True
CUDA runtime version : 12.8.61
CUDA_MODULE_LOADING set to : LAZY
GPU models and configuration :
GPU 0: NVIDIA A100-SXM4-80GB
GPU 1: NVIDIA A100-SXM4-80GB
GPU 2: NVIDIA A100-SXM4-80GB
GPU 3: NVIDIA A100-SXM4-80GB
GPU 4: NVIDIA A100-SXM4-80GB
GPU 5: NVIDIA A100-SXM4-80GB
GPU 6: NVIDIA A100-SXM4-80GB
GPU 7: NVIDIA A100-SXM4-80GB
Nvidia driver version : 530.30.02
cuDNN version : Probably one of the following:
/usr/lib/x86_64-linux-gnu/libcudnn.so.9.7.0
/usr/lib/x86_64-linux-gnu/libcudnn_adv.so.9.7.0
/usr/lib/x86_64-linux-gnu/libcudnn_cnn.so.9.7.0
/usr/lib/x86_64-linux-gnu/libcudnn_engines_precompiled.so.9.7.0
/usr/lib/x86_64-linux-gnu/libcudnn_engines_runtime_compiled.so.9.7.0
/usr/lib/x86_64-linux-gnu/libcudnn_graph.so.9.7.0
/usr/lib/x86_64-linux-gnu/libcudnn_heuristic.so.9.7.0
/usr/lib/x86_64-linux-gnu/libcudnn_ops.so.9.7.0
HIP runtime version : N/A
MIOpen runtime version : N/A
Is XNNPACK available : True
==============================
CPU Info
==============================
Architecture: x86_64
CPU op-mode(s): 32-bit, 64-bit
Address sizes: 43 bits physical, 48 bits virtual
Byte Order: Little Endian
CPU(s): 192
On-line CPU(s) list: 0-191
Vendor ID: AuthenticAMD
BIOS Vendor ID: Advanced Micro Devices, Inc.
Model name: AMD EPYC 7642 48-Core Processor
BIOS Model name: AMD EPYC 7642 48-Core Processor Unknown CPU @ 2.3GHz
BIOS CPU family: 107
CPU family: 23
Model: 49
Thread(s) per core: 2
Core(s) per socket: 48
Socket(s): 2
Stepping: 0
Frequency boost: enabled
CPU(s) scaling MHz: 66%
CPU max MHz: 2300.0000
CPU min MHz: 1500.0000
BogoMIPS: 4600.06
Flags: fpu vme de pse tsc msr pae mce cx8 apic sep mtrr pge mca cmov pat pse36 clflush mmx fxsr sse sse2 ht syscall nx mmxext fxsr_opt pdpe1gb rdtscp lm constant_tsc rep_good nopl xtopology nonstop_tsc cpuid extd_apicid aperfmperf pni pclmulqdq monitor ssse3 fma cx16 sse4_1 sse4_2 movbe popcnt aes xsave avx f16c rdrand lahf_lm cmp_legacy svm extapic cr8_legacy abm sse4a misalignsse 3dnowprefetch osvw ibs skinit wdt tce topoext perfctr_core perfctr_nb bpext perfctr_llc mwaitx cpb cat_l3 cdp_l3 hw_pstate sme ssbd ibrs ibpb stibp vmmcall fsgsbase bmi1 avx2 smep bmi2 cqm rdt_a rdseed adx smap clflushopt clwb sha_ni xsaveopt xsavec xgetbv1 xsaves cqm_llc cqm_occup_llc cqm_mbm_total cqm_mbm_local clzero irperf xsaveerptr rdpru arat npt lbrv svm_lock nrip_save tsc_scale vmcb_clean flushbyasid decodeassists pausefilter pfthreshold avic v_vmsave_vmload vgif umip rdpid overflow_recov succor smca
Virtualization: AMD-V
L1d cache: 3 MiB (96 instances)
L1i cache: 3 MiB (96 instances)
L2 cache: 48 MiB (96 instances)
L3 cache: 512 MiB (32 instances)
NUMA node(s): 2
NUMA node0 CPU(s): 0-47,96-143
NUMA node1 CPU(s): 48-95,144-191
Vulnerability Itlb multihit: Not affected
Vulnerability L1tf: Not affected
Vulnerability Mds: Not affected
Vulnerability Meltdown: Not affected
Vulnerability Mmio stale data: Not affected
Vulnerability Retbleed: Vulnerable
Vulnerability Spec store bypass: Mitigation; Speculative Store Bypass disabled via prctl and seccomp
Vulnerability Spectre v1: Mitigation; usercopy/swapgs barriers and __user pointer sanitization
Vulnerability Spectre v2: Mitigation; Retpolines, IBPB conditional, IBRS_FW, STIBP conditional, RSB filling, PBRSB-eIBRS Not affected
Vulnerability Srbds: Not affected
Vulnerability Tsx async abort: Not affected
==============================
Versions of relevant libraries
==============================
[pip3] flashinfer-python==0.6.6
[pip3] mypy==1.11.1
[pip3] mypy_extensions==1.1.0
[pip3] numpy==2.2.6
[pip3] nvidia-cublas-cu12==12.9.1.4
[pip3] nvidia-cuda-cupti-cu12==12.9.79
[pip3] nvidia-cuda-nvrtc-cu12==12.9.86
[pip3] nvidia-cuda-runtime-cu12==12.9.79
[pip3] nvidia-cudnn-cu12==9.10.2.21
[pip3] nvidia-cudnn-frontend==1.18.0
[pip3] nvidia-cufft-cu12==11.4.1.4
[pip3] nvidia-cufile-cu12==1.14.1.1
[pip3] nvidia-curand-cu12==10.3.10.19
[pip3] nvidia-cusolver-cu12==11.7.5.82
[pip3] nvidia-cusparse-cu12==12.5.10.65
[pip3] nvidia-cusparselt-cu12==0.7.1
[pip3] nvidia-cutlass-dsl==4.4.1
[pip3] nvidia-cutlass-dsl-libs-base==4.4.1
[pip3] nvidia-ml-py==13.590.48
[pip3] nvidia-nccl-cu12==2.27.5
[pip3] nvidia-nvjitlink-cu12==12.9.86
[pip3] nvidia-nvshmem-cu12==3.4.5
[pip3] nvidia-nvtx-cu12==12.9.79
[pip3] onnxruntime==1.24.3
[pip3] pyzmq==27.1.0
[pip3] torch==2.10.0+cu129
[pip3] torch_c_dlpack_ext==0.1.5
[pip3] torchaudio==2.10.0+cu129
[pip3] torchsde==0.2.6
[pip3] torchvision==0.25.0+cu129
[pip3] transformers==4.57.6
[pip3] triton==3.6.0
[pip3] x-transformers==2.16.2
[conda] Could not collect
==============================
vLLM Info
==============================
ROCM Version : Could not collect
vLLM Version : 0.18.0
vLLM-Omni Version : 0.18.0rc2.dev71+g699ca2eb7 (git sha: 699ca2eb7)
vLLM Build Flags:
CUDA Archs: 7.5 8.0 8.6 9.0 10.0 12.0+PTX; ROCm: Disabled
GPU Topology:
GPU0 GPU1 GPU2 GPU3 GPU4 GPU5 GPU6 GPU7 NIC0 NIC1 NIC2 NIC3 NIC4 NIC5 NIC6 NIC7 NIC8 NIC9 NIC10 NIC11 NIC12 NIC13 NIC14 NIC15 NIC16NIC17 CPU Affinity NUMA Affinity
GPU0 X NV12 NV12 NV12 NV12 NV12 NV12 NV12 NODE NODE NODE NODE PXB PXB PXB PXB SYS SYS SYS SYS SYS SYS SYS SYS SYS SYS 0-47,96-143 0
GPU1 NV12 X NV12 NV12 NV12 NV12 NV12 NV12 NODE NODE NODE NODE PXB PXB PXB PXB SYS SYS SYS SYS SYS SYS SYS SYS SYS SYS 0-47,96-143 0
GPU2 NV12 NV12 X NV12 NV12 NV12 NV12 NV12 PXB PXB PXB PXB NODE NODE NODE NODE SYS SYS SYS SYS SYS SYS SYS SYS SYS SYS 0-47,96-143 0
GPU3 NV12 NV12 NV12 X NV12 NV12 NV12 NV12 PXB PXB PXB PXB NODE NODE NODE NODE SYS SYS SYS SYS SYS SYS SYS SYS SYS SYS 0-47,96-143 0
GPU4 NV12 NV12 NV12 NV12 X NV12 NV12 NV12 SYS SYS SYS SYS SYS SYS SYS SYS NODE NODE NODE NODE PXB PXB PXB PXB NODE NODE 48-95,144-191 1
GPU5 NV12 NV12 NV12 NV12 NV12 X NV12 NV12 SYS SYS SYS SYS SYS SYS SYS SYS NODE NODE NODE NODE PXB PXB PXB PXB NODE NODE 48-95,144-191 1
GPU6 NV12 NV12 NV12 NV12 NV12 NV12 X NV12 SYS SYS SYS SYS SYS SYS SYS SYS PXB PXB PXB PXB NODE NODE NODE NODE NODE NODE 48-95,144-191 1
GPU7 NV12 NV12 NV12 NV12 NV12 NV12 NV12 X SYS SYS SYS SYS SYS SYS SYS SYS PXB PXB PXB PXB NODE NODE NODE NODE NODE NODE 48-95,144-191 1
NIC0 NODE NODE PXB PXB SYS SYS SYS SYS X PIX PIX PIX NODE NODE NODE NODE SYS SYS SYS SYS SYS SYS SYS SYS SYS SYS
NIC1 NODE NODE PXB PXB SYS SYS SYS SYS PIX X PIX PIX NODE NODE NODE NODE SYS SYS SYS SYS SYS SYS SYS SYS SYS SYS
NIC2 NODE NODE PXB PXB SYS SYS SYS SYS PIX PIX X PIX NODE NODE NODE NODE SYS SYS SYS SYS SYS SYS SYS SYS SYS SYS
NIC3 NODE NODE PXB PXB SYS SYS SYS SYS PIX PIX PIX X NODE NODE NODE NODE SYS SYS SYS SYS SYS SYS SYS SYS SYS SYS
NIC4 PXB PXB NODE NODE SYS SYS SYS SYS NODE NODE NODE NODE X PIX PXB PXB SYS SYS SYS SYS SYS SYS SYS SYS SYS SYS
NIC5 PXB PXB NODE NODE SYS SYS SYS SYS NODE NODE NODE NODE PIX X PXB PXB SYS SYS SYS SYS SYS SYS SYS SYS SYS SYS
NIC6 PXB PXB NODE NODE SYS SYS SYS SYS NODE NODE NODE NODE PXB PXB X PIX SYS SYS SYS SYS SYS SYS SYS SYS SYS SYS
NIC7 PXB PXB NODE NODE SYS SYS SYS SYS NODE NODE NODE NODE PXB PXB PIX X SYS SYS SYS SYS SYS SYS SYS SYS SYS SYS
NIC8 SYS SYS SYS SYS NODE NODE PXB PXB SYS SYS SYS SYS SYS SYS SYS SYS X PIX PXB PXB NODE NODE NODE NODE NODE NODE
NIC9 SYS SYS SYS SYS NODE NODE PXB PXB SYS SYS SYS SYS SYS SYS SYS SYS PIX X PXB PXB NODE NODE NODE NODE NODE NODE
NIC10 SYS SYS SYS SYS NODE NODE PXB PXB SYS SYS SYS SYS SYS SYS SYS SYS PXB PXB X PIX NODE NODE NODE NODE NODE NODE
NIC11 SYS SYS SYS SYS NODE NODE PXB PXB SYS SYS SYS SYS SYS SYS SYS SYS PXB PXB PIX X NODE NODE NODE NODE NODE NODE
NIC12 SYS SYS SYS SYS PXB PXB NODE NODE SYS SYS SYS SYS SYS SYS SYS SYS NODE NODE NODE NODE X PIX PXB PXB NODE NODE
NIC13 SYS SYS SYS SYS PXB PXB NODE NODE SYS SYS SYS SYS SYS SYS SYS SYS NODE NODE NODE NODE PIX X PXB PXB NODE NODE
NIC14 SYS SYS SYS SYS PXB PXB NODE NODE SYS SYS SYS SYS SYS SYS SYS SYS NODE NODE NODE NODE PXB PXB X PIX NODE NODE
NIC15 SYS SYS SYS SYS PXB PXB NODE NODE SYS SYS SYS SYS SYS SYS SYS SYS NODE NODE NODE NODE PXB PXB PIX X NODE NODE
NIC16 SYS SYS SYS SYS NODE NODE NODE NODE SYS SYS SYS SYS SYS SYS SYS SYS NODE NODE NODE NODE NODE NODE NODE NODE X PIX
NIC17 SYS SYS SYS SYS NODE NODE NODE NODE SYS SYS SYS SYS SYS SYS SYS SYS NODE NODE NODE NODE NODE NODE NODE NODE PIX X
Legend:
X = Self
SYS = Connection traversing PCIe as well as the SMP interconnect between NUMA nodes (e.g., QPI/UPI)
NODE = Connection traversing PCIe as well as the interconnect between PCIe Host Bridges within a NUMA node
PHB = Connection traversing PCIe as well as a PCIe Host Bridge (typically the CPU)
PXB = Connection traversing multiple PCIe bridges (without traversing the PCIe Host Bridge)
PIX = Connection traversing at most a single PCIe bridge
NV# = Connection traversing a bonded set of # NVLinks
NIC Legend:
NIC0: mlx5_0
NIC1: mlx5_1
NIC2: mlx5_2
NIC3: mlx5_3
NIC4: mlx5_4
NIC5: mlx5_5
NIC6: mlx5_6
NIC7: mlx5_7
NIC8: mlx5_8
NIC9: mlx5_9
NIC10: mlx5_10
NIC11: mlx5_11
NIC12: mlx5_12
NIC13: mlx5_13
NIC14: mlx5_14
NIC15: mlx5_15
NIC16: mlx5_16
NIC17: mlx5_17
==============================
Environment Variables
==============================
NVIDIA_VISIBLE_DEVICES=all
CUBLAS_VERSION=12.8.3.14
NVIDIA_REQUIRE_CUDA=cuda>=9.0
CUDA_CACHE_DISABLE=1
TORCH_CUDA_ARCH_LIST=7.5 8.0 8.6 9.0 10.0 12.0+PTX
NCCL_VERSION=2.25.1
NCCL_NVLS_ENABLE=0
NVIDIA_DRIVER_CAPABILITIES=compute,utility,video
TORCH_NCCL_USE_COMM_NONBLOCKING=0
NVIDIA_PRODUCT_NAME=PyTorch
CUDA_VERSION=12.8.0.038
PYTORCH_VERSION=2.6.0a0+ecf3bae
PYTORCH_BUILD_NUMBER=0
CUDNN_FRONTEND_VERSION=1.9.0
CUDNN_VERSION=9.7.0.66
PYTORCH_HOME=/opt/pytorch/pytorch
LD_LIBRARY_PATH=/usr/local/lib/python3.12/dist-packages/torch/lib:/usr/local/lib/python3.12/dist-packages/torch_tensorrt/lib:/usr/local/cuda/compat/lib:/usr/local/nvidia/lib:/usr/local/nvidia/lib64
NVIDIA_BUILD_ID=134983853
CUDA_DRIVER_VERSION=570.86.10
PYTORCH_BUILD_VERSION=2.6.0a0+ecf3bae
CUDA_HOME=/usr/local/cuda
CUDA_HOME=/usr/local/cuda
CUDA_MODULE_LOADING=LAZY
NVIDIA_REQUIRE_JETPACK_HOST_MOUNTS=
NVIDIA_PYTORCH_VERSION=25.01
TORCH_ALLOW_TF32_CUBLAS_OVERRIDE=1
PYTORCH_NVML_BASED_CUDA_CHECK=1
TORCHINDUCTOR_COMPILE_THREADS=1
TORCHINDUCTOR_CACHE_DIR=/tmp/torchinductor_root
Your code version
🐛 Describe the bug
Description
When validating acceleration features for Diffusion models using the compatibility test script from PR #2083, enabling both FP8 quantization and HSDP (Hybrid Sharded Data Parallelism) causes the process to crash during model weight loading. Each feature works correctly when enabled alone, but the combination fails.
Steps to Reproduce
- Run the compatibility test script with the following command (as in PR #2083):
python run_compat_test.py \
--baseline-feature fp8 \
--addons hsdp \
--model Qwen/Qwen-Image \
--num-prompts 3 \
--steps 10
- Observe the error traceback.
Expected Behavior
FP8 quantization and HSDP can be used together; the model loads successfully and inference proceeds normally.
Actual Behavior
The worker process crashes during weight loading with a NotImplementedError, indicating that the dynamic_scaled_fp8_quant operation is being called on a CPU tensor while only CUDA and other backends are supported.
Error Log (Key Part)
File "/home/l30053556/.venv/lib/python3.12/site-packages/vllm/model_executor/layers/quantization/fp8.py", line 625, in process_weights_after_loading
qweight, weight_scale = ops.scaled_fp8_quant(layer.weight, scale=None)
^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
File "/home/l30053556/.venv/lib/python3.12/site-packages/vllm/_custom_ops.py", line 1941, in scaled_fp8_quant
torch.ops._C.dynamic_scaled_fp8_quant(output, input, scale)
File "/home/l30053556/.venv/lib/python3.12/site-packages/torch/_ops.py", line 1209, in __call__
return self._op(*args, **kwargs)
^^^^^^^^^^^^^^^^^^^^^^^^^
NotImplementedError: Could not run '_C::dynamic_scaled_fp8_quant' with arguments from the 'CPU' backend. ...
Full log attached below.
Multi-thread loading shards: 0% Completed | 0/9 [00:00<?, ?it/s]
Process DiffusionWorker-1:
Traceback (most recent call last):
File "/usr/lib/python3.12/multiprocessing/process.py", line 314, in _bootstrap
self.run()
File "/usr/lib/python3.12/multiprocessing/process.py", line 108, in run
self._target(*self._args, **self._kwargs)
File "/home/l30053556/vllm-omni/vllm_omni/diffusion/worker/diffusion_worker.py", line 527, in worker_main
worker_proc = WorkerProc(
^^^^^^^^^^^
File "/home/l30053556/vllm-omni/vllm_omni/diffusion/worker/diffusion_worker.py", line 401, in __init__
self.worker = self._create_worker(gpu_id, od_config, worker_extension_cls, custom_pipeline_args)
^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
File "/home/l30053556/vllm-omni/vllm_omni/diffusion/worker/diffusion_worker.py", line 412, in _create_worker
wrapper = WorkerWrapperBase(
^^^^^^^^^^^^^^^^^^
File "/home/l30053556/vllm-omni/vllm_omni/diffusion/worker/diffusion_worker.py", line 582, in __init__
self.worker = worker_class(
^^^^^^^^^^^^^
File "/home/l30053556/vllm-omni/vllm_omni/diffusion/worker/diffusion_worker.py", line 96, in __init__
self.load_model(load_format=self.od_config.diffusion_load_format)
File "/home/l30053556/vllm-omni/vllm_omni/diffusion/worker/diffusion_worker.py", line 152, in load_model
self.model_runner.load_model(
File "/home/l30053556/vllm-omni/vllm_omni/diffusion/worker/diffusion_model_runner.py", line 132, in load_model
self.pipeline = model_loader.load_model(
^^^^^^^^^^^^^^^^^^^^^^^^
File "/home/l30053556/vllm-omni/vllm_omni/diffusion/model_loader/diffusers_loader.py", line 270, in load_model
model = self._load_model_with_hsdp(
^^^^^^^^^^^^^^^^^^^^^^^^^^^
File "/home/l30053556/vllm-omni/vllm_omni/diffusion/model_loader/diffusers_loader.py", line 488, in _load_model_with_hsdp
self.load_weights(model)
File "/home/l30053556/vllm-omni/vllm_omni/diffusion/model_loader/diffusers_loader.py", line 318, in load_weights
loaded_weights = model.load_weights(self.get_all_weights(model))
^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
File "/home/l30053556/vllm-omni/vllm_omni/diffusion/models/qwen_image/pipeline_qwen_image.py", line 1003, in load_weights
return loader.load_weights(weights)
^^^^^^^^^^^^^^^^^^^^^^^^^^^^
File "/home/l30053556/.venv/lib/python3.12/site-packages/vllm/model_executor/model_loader/reload/torchao_decorator.py", line 50, in patched_model_load_weights
return original_load_weights(self, weights, *args, **kwargs)
^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
File "/home/l30053556/.venv/lib/python3.12/site-packages/vllm/model_executor/models/utils.py", line 348, in load_weights
autoloaded_weights = set(self._load_module("", self.module, weights))
^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
File "/home/l30053556/.venv/lib/python3.12/site-packages/vllm/model_executor/models/utils.py", line 295, in _load_module
yield from self._load_module(
File "/home/l30053556/.venv/lib/python3.12/site-packages/vllm/model_executor/models/utils.py", line 268, in _load_module
loaded_params = module_load_weights(weights)
^^^^^^^^^^^^^^^^^^^^^^^^^^^^
File "/home/l30053556/vllm-omni/vllm_omni/diffusion/models/qwen_image/qwen_image_transformer.py", line 1179, in load_weights
weight_loader(param, loaded_weight)
File "/home/l30053556/.venv/lib/python3.12/site-packages/vllm/model_executor/layers/quantization/fp8.py", line 575, in patched_weight_loader
self.process_weights_after_loading(layer)
File "/home/l30053556/.venv/lib/python3.12/site-packages/vllm/model_executor/layers/quantization/fp8.py", line 625, in process_weights_after_loading
qweight, weight_scale = ops.scaled_fp8_quant(layer.weight, scale=None)
^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
File "/home/l30053556/.venv/lib/python3.12/site-packages/vllm/_custom_ops.py", line 1941, in scaled_fp8_quant
torch.ops._C.dynamic_scaled_fp8_quant(output, input, scale)
File "/home/l30053556/.venv/lib/python3.12/site-packages/torch/_ops.py", line 1209, in __call__
return self._op(*args, **kwargs)
^^^^^^^^^^^^^^^^^^^^^^^^^
File "/home/l30053556/.venv/lib/python3.12/site-packages/vllm/model_executor/parameter.py", line 126, in __torch_function__
return super().__torch_function__(func, types, args, kwargs)
^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
File "/home/l30053556/.venv/lib/python3.12/site-packages/torch/_ops.py", line 1209, in __call__
return self._op(*args, **kwargs)
^^^^^^^^^^^^^^^^^^^^^^^^^
NotImplementedError: Could not run '_C::dynamic_scaled_fp8_quant' with arguments from the 'CPU' backend. This could be because the operator doesn't exist for this backend, or was omitted during the selective/custom build process (if using custom build). If you are a Facebook employee using PyTorch on mobile, please visit https://fburl.com/ptmfixes for possible resolutions. '_C::dynamic_scaled_fp8_quant' is only available for these backends: [CUDA, Meta, BackendSelect, Python, FuncTorchDynamicLayerBackMode, Functionalize, Named, Conjugate, Negative, ZeroTensor, ADInplaceOrView, AutogradOther, AutogradCPU, AutogradCUDA, AutogradXLA, AutogradMPS, AutogradXPU, AutogradHPU, AutogradLazy, AutogradMTIA, AutogradMAIA, AutogradPrivateUse1, AutogradMeta, Tracer, AutocastCPU, AutocastMTIA, AutocastMAIA, AutocastXPU, AutocastMPS, AutocastCUDA, FuncTorchBatched, BatchedNestedTensor, FuncTorchVmapMode, Batched, VmapMode, FuncTorchGradWrapper, PythonTLSSnapshot, FuncTorchDynamicLayerFrontMode, PreDispatch, PythonDispatcher].
Additional Context
-
The error occurs when loading weights in the HSDP path, and the FP8 quantization attempts to call a CUDA-only operator on a CPU tensor. It seems that under HSDP, the model parameters might be on CPU during the FP8 weight transformation stage, leading to the backend mismatch.
-
This was reproduced using the compatibility test script from PR #2083 with the command provided.
-
Tested with Qwen-Image-2512 ; likely affects other Diffusion models as well.
Before submitting a new issue...
- Make sure you already searched for relevant issues, and asked the chatbot living at the bottom right corner of the documentation page, which can answer lots of frequently asked questions.