5090 cuda Compatibility issues #205
Description
python -c "import torch; print(torch.version); print(torch.cuda.is_available()); print(torch.cuda.get_device_capability())"
2.9.0+cu128
True
(12, 0)
(ptxmm) adsb@MSI:/mnt/d/conda/protenix$ nvcc --version
nvcc: NVIDIA (R) Cuda compiler driver
Copyright (c) 2005-2025 NVIDIA Corporation
Built on Wed_Aug_20_01:58:59_PM_PDT_2025
Cuda compilation tools, release 13.0, V13.0.88
Build cuda_13.0.r13.0/compiler.36424714_0
(ptxmm) adsb@MSI:/mnt/d/conda/protenix$ export TORCH_CUDA_ARCH_LIST="8.9;12.0"
(ptxmm) adsb@MSI:/mnt/d/conda/protenix$ bash inference_demo.sh
Try to find the ccd cache data in the code directory for inference.
2025-10-06 16:13:55,702 [/mnt/d/conda/protenix/runner/inference.py:159] INFO main: Distributed environment: world size: 1, global rank: 0, local rank: 0
2025-10-06 16:13:55,702 [/mnt/d/conda/protenix/runner/inference.py:65] INFO root: LOCAL_RANK: 0 - CUDA_VISIBLE_DEVICES: [0]
2025-10-06 16:13:55,812 [/mnt/d/conda/protenix/runner/inference.py:159] INFO main: env: /mnt/d/conda/protenix/cutlass-3.5.1
2025-10-06 16:13:55,812 [/mnt/d/conda/protenix/runner/inference.py:80] INFO root: The kernels will be compiled when DS4Sci_EvoformerAttention is called for the first time.
2025-10-06 16:13:55,812 [/mnt/d/conda/protenix/runner/inference.py:85] INFO root: The kernels will be compiled when fast_layernorm is called for the first time.
2025-10-06 16:13:55,812 [/mnt/d/conda/protenix/runner/inference.py:89] INFO root: Finished init ENV.
train scheduler 16.0
inference scheduler 16.0
Diffusion Module has 16.0
2025-10-06 16:13:58,242 [/mnt/d/conda/protenix/runner/inference.py:159] INFO main: Loading from /home/adsb/conda/envs/ptxmm/lib/python3.10/site-packages/./release_data/checkpoint/model_v0.2.0.pt, strict: True
2025-10-06 16:14:03,742 [/mnt/d/conda/protenix/runner/inference.py:159] INFO main: Sampled key: module.input_embedder.atom_attention_encoder.linear_no_bias_f.weight
2025-10-06 16:14:04,038 [/mnt/d/conda/protenix/runner/inference.py:159] INFO main: Finish loading checkpoint.
2025-10-06 16:14:04,042 [/mnt/d/conda/protenix/runner/inference.py:221] INFO main: Loading data from
./examples/example.json
2025-10-06 16:14:04,226 [/home/adsb/conda/envs/ptxmm/lib/python3.10/site-packages/protenix/data/infer_data_pipeline.py:209] INFO protenix.data.infer_data_pipeline: Featurizing 7r6r...
2025-10-06 16:14:04,226 [/home/adsb/conda/envs/ptxmm/lib/python3.10/site-packages/protenix/data/infer_data_pipeline.py:209] INFO protenix.data.infer_data_pipeline: Featurizing 7wux...
2025-10-06 16:14:04,227 [/home/adsb/conda/envs/ptxmm/lib/python3.10/site-packages/protenix/data/infer_data_pipeline.py:209] INFO protenix.data.infer_data_pipeline: Featurizing 7pzb...
2025-10-06 16:14:10,904 [/mnt/d/conda/protenix/runner/inference.py:245] INFO main: [Rank 0 (1/3)] 7r6r: N_asym 3, N_token 245, N_atom 2529, N_msa 363
/home/adsb/conda/envs/ptxmm/lib/python3.10/site-packages/protenix/model/layer_norm/layer_norm.py:50: FutureWarning: torch.cuda.amp.autocast(args...) is deprecated. Please use torch.amp.autocast('cuda', args...) instead.
with torch.cuda.amp.autocast(enabled=False):
2025-10-06 16:14:11,382 [/mnt/d/conda/protenix/runner/inference.py:271] INFO main: [Rank 0]7r6r CUDA error: no kernel image is available for execution on the device
Search for cudaErrorNoKernelImageForDevice' in https://docs.nvidia.com/cuda/cuda-runtime-api/group__CUDART__TYPES.html for more information. Compile with TORCH_USE_CUDA_DSAto enable device-side assertions. : Traceback (most recent call last): File "/mnt/d/conda/protenix/runner/inference.py", line 254, in infer_predict prediction = runner.predict(data) File "/home/adsb/conda/envs/ptxmm/lib/python3.10/site-packages/torch/utils/_contextlib.py", line 120, in decorate_context return func(*args, **kwargs) File "/mnt/d/conda/protenix/runner/inference.py", line 148, in predict prediction, _, _ = self.model( File "/home/adsb/conda/envs/ptxmm/lib/python3.10/site-packages/torch/nn/modules/module.py", line 1775, in _wrapped_call_impl return self._call_impl(*args, **kwargs) File "/home/adsb/conda/envs/ptxmm/lib/python3.10/site-packages/torch/nn/modules/module.py", line 1786, in _call_impl return forward_call(*args, **kwargs) File "/home/adsb/conda/envs/ptxmm/lib/python3.10/site-packages/protenix/model/protenix.py", line 692, in forward pred_dict, log_dict, time_tracker = self.main_inference_loop( File "/home/adsb/conda/envs/ptxmm/lib/python3.10/site-packages/protenix/model/protenix.py", line 321, in main_inference_loop pred_dict, log_dict, time_tracker = self._main_inference_loop( File "/home/adsb/conda/envs/ptxmm/lib/python3.10/site-packages/protenix/model/protenix.py", line 382, in _main_inference_loop s_inputs, s, z = self.get_pairformer_output( File "/home/adsb/conda/envs/ptxmm/lib/python3.10/site-packages/protenix/model/protenix.py", line 148, in get_pairformer_output s_inputs = self.input_embedder( File "/home/adsb/conda/envs/ptxmm/lib/python3.10/site-packages/torch/nn/modules/module.py", line 1775, in _wrapped_call_impl return self._call_impl(*args, **kwargs) File "/home/adsb/conda/envs/ptxmm/lib/python3.10/site-packages/torch/nn/modules/module.py", line 1786, in _call_impl return forward_call(*args, **kwargs) File "/home/adsb/conda/envs/ptxmm/lib/python3.10/site-packages/protenix/model/modules/embedders.py", line 72, in forward a, _, _, _ = self.atom_attention_encoder( File "/home/adsb/conda/envs/ptxmm/lib/python3.10/site-packages/torch/nn/modules/module.py", line 1775, in _wrapped_call_impl return self._call_impl(*args, **kwargs) File "/home/adsb/conda/envs/ptxmm/lib/python3.10/site-packages/torch/nn/modules/module.py", line 1786, in _call_impl return forward_call(*args, **kwargs) File "/home/adsb/conda/envs/ptxmm/lib/python3.10/site-packages/protenix/model/modules/transformer.py", line 837, in forward q_l = self.atom_transformer( File "/home/adsb/conda/envs/ptxmm/lib/python3.10/site-packages/torch/nn/modules/module.py", line 1775, in _wrapped_call_impl return self._call_impl(*args, **kwargs) File "/home/adsb/conda/envs/ptxmm/lib/python3.10/site-packages/torch/nn/modules/module.py", line 1786, in _call_impl return forward_call(*args, **kwargs) File "/home/adsb/conda/envs/ptxmm/lib/python3.10/site-packages/protenix/model/modules/transformer.py", line 482, in forward return self.diffusion_transformer( File "/home/adsb/conda/envs/ptxmm/lib/python3.10/site-packages/torch/nn/modules/module.py", line 1775, in _wrapped_call_impl return self._call_impl(*args, **kwargs) File "/home/adsb/conda/envs/ptxmm/lib/python3.10/site-packages/torch/nn/modules/module.py", line 1786, in _call_impl return forward_call(*args, **kwargs) File "/home/adsb/conda/envs/ptxmm/lib/python3.10/site-packages/protenix/model/modules/transformer.py", line 405, in forward a, s, z = checkpoint_blocks( File "/home/adsb/conda/envs/ptxmm/lib/python3.10/site-packages/protenix/openfold_local/utils/checkpointing.py", line 85, in checkpoint_blocks return exec(blocks, args) File "/home/adsb/conda/envs/ptxmm/lib/python3.10/site-packages/protenix/openfold_local/utils/checkpointing.py", line 72, in exec a = wrap(block(*a)) File "/home/adsb/conda/envs/ptxmm/lib/python3.10/site-packages/torch/nn/modules/module.py", line 1775, in _wrapped_call_impl return self._call_impl(*args, **kwargs) File "/home/adsb/conda/envs/ptxmm/lib/python3.10/site-packages/torch/nn/modules/module.py", line 1786, in _call_impl return forward_call(*args, **kwargs) File "/home/adsb/conda/envs/ptxmm/lib/python3.10/site-packages/protenix/model/modules/transformer.py", line 282, in forward attn_out = self.attention_pair_bias( File "/home/adsb/conda/envs/ptxmm/lib/python3.10/site-packages/torch/nn/modules/module.py", line 1775, in _wrapped_call_impl return self._call_impl(*args, **kwargs) File "/home/adsb/conda/envs/ptxmm/lib/python3.10/site-packages/torch/nn/modules/module.py", line 1786, in _call_impl return forward_call(*args, **kwargs) File "/home/adsb/conda/envs/ptxmm/lib/python3.10/site-packages/protenix/model/modules/transformer.py", line 201, in forward a = self.local_multihead_attention( File "/home/adsb/conda/envs/ptxmm/lib/python3.10/site-packages/protenix/model/modules/transformer.py", line 131, in local_multihead_attention bias = self.linear_nobias_z( File "/home/adsb/conda/envs/ptxmm/lib/python3.10/site-packages/torch/nn/modules/module.py", line 1775, in _wrapped_call_impl return self._call_impl(*args, **kwargs) File "/home/adsb/conda/envs/ptxmm/lib/python3.10/site-packages/torch/nn/modules/module.py", line 1786, in _call_impl return forward_call(*args, **kwargs) File "/home/adsb/conda/envs/ptxmm/lib/python3.10/site-packages/torch/nn/modules/linear.py", line 134, in forward return F.linear(input, self.weight, self.bias) torch.AcceleratorError: CUDA error: no kernel image is available for execution on the device Search forcudaErrorNoKernelImageForDevice' in https://docs.nvidia.com/cuda/cuda-runtime-api/group__CUDART__TYPES.html for more information.
Compile with TORCH_USE_CUDA_DSA to enable device-side assertions.
2025-10-06 16:14:13,832 [/mnt/d/conda/protenix/runner/inference.py:245] INFO main: [Rank 0 (2/3)] 7wux: N_asym 10, N_token 1218, N_atom 9142, N_msa 1286
2025-10-06 16:14:14,173 [/mnt/d/conda/protenix/runner/inference.py:271] INFO main: [Rank 0]7wux CUDA error: no kernel image is available for execution on the device
Search for cudaErrorNoKernelImageForDevice' in https://docs.nvidia.com/cuda/cuda-runtime-api/group__CUDART__TYPES.html for more information. Compile with TORCH_USE_CUDA_DSAto enable device-side assertions. : Traceback (most recent call last): File "/mnt/d/conda/protenix/runner/inference.py", line 254, in infer_predict prediction = runner.predict(data) File "/home/adsb/conda/envs/ptxmm/lib/python3.10/site-packages/torch/utils/_contextlib.py", line 120, in decorate_context return func(*args, **kwargs) File "/mnt/d/conda/protenix/runner/inference.py", line 148, in predict prediction, _, _ = self.model( File "/home/adsb/conda/envs/ptxmm/lib/python3.10/site-packages/torch/nn/modules/module.py", line 1775, in _wrapped_call_impl return self._call_impl(*args, **kwargs) File "/home/adsb/conda/envs/ptxmm/lib/python3.10/site-packages/torch/nn/modules/module.py", line 1786, in _call_impl return forward_call(*args, **kwargs) File "/home/adsb/conda/envs/ptxmm/lib/python3.10/site-packages/protenix/model/protenix.py", line 692, in forward pred_dict, log_dict, time_tracker = self.main_inference_loop( File "/home/adsb/conda/envs/ptxmm/lib/python3.10/site-packages/protenix/model/protenix.py", line 321, in main_inference_loop pred_dict, log_dict, time_tracker = self._main_inference_loop( File "/home/adsb/conda/envs/ptxmm/lib/python3.10/site-packages/protenix/model/protenix.py", line 382, in _main_inference_loop s_inputs, s, z = self.get_pairformer_output( File "/home/adsb/conda/envs/ptxmm/lib/python3.10/site-packages/protenix/model/protenix.py", line 148, in get_pairformer_output s_inputs = self.input_embedder( File "/home/adsb/conda/envs/ptxmm/lib/python3.10/site-packages/torch/nn/modules/module.py", line 1775, in _wrapped_call_impl return self._call_impl(*args, **kwargs) File "/home/adsb/conda/envs/ptxmm/lib/python3.10/site-packages/torch/nn/modules/module.py", line 1786, in _call_impl return forward_call(*args, **kwargs) File "/home/adsb/conda/envs/ptxmm/lib/python3.10/site-packages/protenix/model/modules/embedders.py", line 72, in forward a, _, _, _ = self.atom_attention_encoder( File "/home/adsb/conda/envs/ptxmm/lib/python3.10/site-packages/torch/nn/modules/module.py", line 1775, in _wrapped_call_impl return self._call_impl(*args, **kwargs) File "/home/adsb/conda/envs/ptxmm/lib/python3.10/site-packages/torch/nn/modules/module.py", line 1786, in _call_impl return forward_call(*args, **kwargs) File "/home/adsb/conda/envs/ptxmm/lib/python3.10/site-packages/protenix/model/modules/transformer.py", line 837, in forward q_l = self.atom_transformer( File "/home/adsb/conda/envs/ptxmm/lib/python3.10/site-packages/torch/nn/modules/module.py", line 1775, in _wrapped_call_impl return self._call_impl(*args, **kwargs) File "/home/adsb/conda/envs/ptxmm/lib/python3.10/site-packages/torch/nn/modules/module.py", line 1786, in _call_impl return forward_call(*args, **kwargs) File "/home/adsb/conda/envs/ptxmm/lib/python3.10/site-packages/protenix/model/modules/transformer.py", line 482, in forward return self.diffusion_transformer( File "/home/adsb/conda/envs/ptxmm/lib/python3.10/site-packages/torch/nn/modules/module.py", line 1775, in _wrapped_call_impl return self._call_impl(*args, **kwargs) File "/home/adsb/conda/envs/ptxmm/lib/python3.10/site-packages/torch/nn/modules/module.py", line 1786, in _call_impl return forward_call(*args, **kwargs) File "/home/adsb/conda/envs/ptxmm/lib/python3.10/site-packages/protenix/model/modules/transformer.py", line 405, in forward a, s, z = checkpoint_blocks( File "/home/adsb/conda/envs/ptxmm/lib/python3.10/site-packages/protenix/openfold_local/utils/checkpointing.py", line 85, in checkpoint_blocks return exec(blocks, args) File "/home/adsb/conda/envs/ptxmm/lib/python3.10/site-packages/protenix/openfold_local/utils/checkpointing.py", line 72, in exec a = wrap(block(*a)) File "/home/adsb/conda/envs/ptxmm/lib/python3.10/site-packages/torch/nn/modules/module.py", line 1775, in _wrapped_call_impl return self._call_impl(*args, **kwargs) File "/home/adsb/conda/envs/ptxmm/lib/python3.10/site-packages/torch/nn/modules/module.py", line 1786, in _call_impl return forward_call(*args, **kwargs) File "/home/adsb/conda/envs/ptxmm/lib/python3.10/site-packages/protenix/model/modules/transformer.py", line 282, in forward attn_out = self.attention_pair_bias( File "/home/adsb/conda/envs/ptxmm/lib/python3.10/site-packages/torch/nn/modules/module.py", line 1775, in _wrapped_call_impl return self._call_impl(*args, **kwargs) File "/home/adsb/conda/envs/ptxmm/lib/python3.10/site-packages/torch/nn/modules/module.py", line 1786, in _call_impl return forward_call(*args, **kwargs) File "/home/adsb/conda/envs/ptxmm/lib/python3.10/site-packages/protenix/model/modules/transformer.py", line 201, in forward a = self.local_multihead_attention( File "/home/adsb/conda/envs/ptxmm/lib/python3.10/site-packages/protenix/model/modules/transformer.py", line 131, in local_multihead_attention bias = self.linear_nobias_z( File "/home/adsb/conda/envs/ptxmm/lib/python3.10/site-packages/torch/nn/modules/module.py", line 1775, in _wrapped_call_impl return self._call_impl(*args, **kwargs) File "/home/adsb/conda/envs/ptxmm/lib/python3.10/site-packages/torch/nn/modules/module.py", line 1786, in _call_impl return forward_call(*args, **kwargs) File "/home/adsb/conda/envs/ptxmm/lib/python3.10/site-packages/torch/nn/modules/linear.py", line 134, in forward return F.linear(input, self.weight, self.bias) torch.AcceleratorError: CUDA error: no kernel image is available for execution on the device Search forcudaErrorNoKernelImageForDevice' in https://docs.nvidia.com/cuda/cuda-runtime-api/group__CUDART__TYPES.html for more information.
Compile with TORCH_USE_CUDA_DSA to enable device-side assertions.
2025-10-06 16:14:14,178 [/mnt/d/conda/protenix/runner/inference.py:245] INFO main: [Rank 0 (3/3)] 7pzb: N_asym 8, N_token 600, N_atom 5222, N_msa 6502
2025-10-06 16:14:14,243 [/mnt/d/conda/protenix/runner/inference.py:271] INFO main: [Rank 0]7pzb CUDA error: no kernel image is available for execution on the device
Search for cudaErrorNoKernelImageForDevice' in https://docs.nvidia.com/cuda/cuda-runtime-api/group__CUDART__TYPES.html for more information. Compile with TORCH_USE_CUDA_DSAto enable device-side assertions. : Traceback (most recent call last): File "/mnt/d/conda/protenix/runner/inference.py", line 254, in infer_predict prediction = runner.predict(data) File "/home/adsb/conda/envs/ptxmm/lib/python3.10/site-packages/torch/utils/_contextlib.py", line 120, in decorate_context return func(*args, **kwargs) File "/mnt/d/conda/protenix/runner/inference.py", line 148, in predict prediction, _, _ = self.model( File "/home/adsb/conda/envs/ptxmm/lib/python3.10/site-packages/torch/nn/modules/module.py", line 1775, in _wrapped_call_impl return self._call_impl(*args, **kwargs) File "/home/adsb/conda/envs/ptxmm/lib/python3.10/site-packages/torch/nn/modules/module.py", line 1786, in _call_impl return forward_call(*args, **kwargs) File "/home/adsb/conda/envs/ptxmm/lib/python3.10/site-packages/protenix/model/protenix.py", line 692, in forward pred_dict, log_dict, time_tracker = self.main_inference_loop( File "/home/adsb/conda/envs/ptxmm/lib/python3.10/site-packages/protenix/model/protenix.py", line 321, in main_inference_loop pred_dict, log_dict, time_tracker = self._main_inference_loop( File "/home/adsb/conda/envs/ptxmm/lib/python3.10/site-packages/protenix/model/protenix.py", line 382, in _main_inference_loop s_inputs, s, z = self.get_pairformer_output( File "/home/adsb/conda/envs/ptxmm/lib/python3.10/site-packages/protenix/model/protenix.py", line 148, in get_pairformer_output s_inputs = self.input_embedder( File "/home/adsb/conda/envs/ptxmm/lib/python3.10/site-packages/torch/nn/modules/module.py", line 1775, in _wrapped_call_impl return self._call_impl(*args, **kwargs) File "/home/adsb/conda/envs/ptxmm/lib/python3.10/site-packages/torch/nn/modules/module.py", line 1786, in _call_impl return forward_call(*args, **kwargs) File "/home/adsb/conda/envs/ptxmm/lib/python3.10/site-packages/protenix/model/modules/embedders.py", line 72, in forward a, _, _, _ = self.atom_attention_encoder( File "/home/adsb/conda/envs/ptxmm/lib/python3.10/site-packages/torch/nn/modules/module.py", line 1775, in _wrapped_call_impl return self._call_impl(*args, **kwargs) File "/home/adsb/conda/envs/ptxmm/lib/python3.10/site-packages/torch/nn/modules/module.py", line 1786, in _call_impl return forward_call(*args, **kwargs) File "/home/adsb/conda/envs/ptxmm/lib/python3.10/site-packages/protenix/model/modules/transformer.py", line 837, in forward q_l = self.atom_transformer( File "/home/adsb/conda/envs/ptxmm/lib/python3.10/site-packages/torch/nn/modules/module.py", line 1775, in _wrapped_call_impl return self._call_impl(*args, **kwargs) File "/home/adsb/conda/envs/ptxmm/lib/python3.10/site-packages/torch/nn/modules/module.py", line 1786, in _call_impl return forward_call(*args, **kwargs) File "/home/adsb/conda/envs/ptxmm/lib/python3.10/site-packages/protenix/model/modules/transformer.py", line 482, in forward return self.diffusion_transformer( File "/home/adsb/conda/envs/ptxmm/lib/python3.10/site-packages/torch/nn/modules/module.py", line 1775, in _wrapped_call_impl return self._call_impl(*args, **kwargs) File "/home/adsb/conda/envs/ptxmm/lib/python3.10/site-packages/torch/nn/modules/module.py", line 1786, in _call_impl return forward_call(*args, **kwargs) File "/home/adsb/conda/envs/ptxmm/lib/python3.10/site-packages/protenix/model/modules/transformer.py", line 405, in forward a, s, z = checkpoint_blocks( File "/home/adsb/conda/envs/ptxmm/lib/python3.10/site-packages/protenix/openfold_local/utils/checkpointing.py", line 85, in checkpoint_blocks return exec(blocks, args) File "/home/adsb/conda/envs/ptxmm/lib/python3.10/site-packages/protenix/openfold_local/utils/checkpointing.py", line 72, in exec a = wrap(block(*a)) File "/home/adsb/conda/envs/ptxmm/lib/python3.10/site-packages/torch/nn/modules/module.py", line 1775, in _wrapped_call_impl return self._call_impl(*args, **kwargs) File "/home/adsb/conda/envs/ptxmm/lib/python3.10/site-packages/torch/nn/modules/module.py", line 1786, in _call_impl return forward_call(*args, **kwargs) File "/home/adsb/conda/envs/ptxmm/lib/python3.10/site-packages/protenix/model/modules/transformer.py", line 282, in forward attn_out = self.attention_pair_bias( File "/home/adsb/conda/envs/ptxmm/lib/python3.10/site-packages/torch/nn/modules/module.py", line 1775, in _wrapped_call_impl return self._call_impl(*args, **kwargs) File "/home/adsb/conda/envs/ptxmm/lib/python3.10/site-packages/torch/nn/modules/module.py", line 1786, in _call_impl return forward_call(*args, **kwargs) File "/home/adsb/conda/envs/ptxmm/lib/python3.10/site-packages/protenix/model/modules/transformer.py", line 201, in forward a = self.local_multihead_attention( File "/home/adsb/conda/envs/ptxmm/lib/python3.10/site-packages/protenix/model/modules/transformer.py", line 131, in local_multihead_attention bias = self.linear_nobias_z( File "/home/adsb/conda/envs/ptxmm/lib/python3.10/site-packages/torch/nn/modules/module.py", line 1775, in _wrapped_call_impl return self._call_impl(*args, **kwargs) File "/home/adsb/conda/envs/ptxmm/lib/python3.10/site-packages/torch/nn/modules/module.py", line 1786, in _call_impl return forward_call(*args, **kwargs) File "/home/adsb/conda/envs/ptxmm/lib/python3.10/site-packages/torch/nn/modules/linear.py", line 134, in forward return F.linear(input, self.weight, self.bias) torch.AcceleratorError: CUDA error: no kernel image is available for execution on the device Search forcudaErrorNoKernelImageForDevice' in https://docs.nvidia.com/cuda/cuda-runtime-api/group__CUDART__TYPES.html for more information.
Compile with TORCH_USE_CUDA_DSA to enable device-side assertions.
(ptxmm) adsb@MSI:/mnt/d/conda/protenix$ python -c "import torch; x = torch.randn(1, 128, 256, device='cuda'); y = torch.nn.Linear(256, 256).cuda()(x); print(y)"
tensor([[[-0.5351, -0.3450, -0.0861, ..., -0.3420, 0.0343, -0.1446],
[ 0.5010, -0.1754, -1.2315, ..., -0.9490, 0.2061, 0.0160],
[ 0.0726, 0.2123, -0.0957, ..., 0.3553, 0.0089, 0.3244],
...,
[-0.0861, -0.5129, -0.2459, ..., -0.6243, 0.5106, -0.7058],
[-0.1798, -0.2492, -0.5570, ..., 0.7262, -0.0278, 0.1920],
[-0.2529, 0.1932, -0.6778, ..., -0.0812, 0.4013, 0.6989]]],
device='cuda:0', grad_fn=)
老师您好，似乎5090 sm_12.0 有兼容问题，不知道能否根据上述信息，排查一下，谢谢您！