failed with 'CUDA_ERROR_INVALID_HANDLE'

[captainsalih]

Dear All,

I want to install this latest version of deepmd. I'm use Ubuntu, and NVIDIA 5090 graphics card. So here are what I had done

1. Create new env and install tensorflow and cuda. Here is the detail

+-----------------------------------------------------------------------------------------+ | NVIDIA-SMI 570.148.08 Driver Version: 570.148.08 CUDA Version: 12.8 |
|-----------------------------------------+------------------------+----------------------+

---

TF: 2.20.0
CUDA: 12.5.1
cuDNN: 9

conda list | egrep 'cudatoolkit|cudnn|cuda'
nvidia-cuda-cupti-cu12 12.9.79 pypi_0 pypi
nvidia-cuda-nvcc-cu12 12.9.86 pypi_0 pypi
nvidia-cuda-nvrtc-cu12 12.9.86 pypi_0 pypi
nvidia-cuda-runtime-cu12 12.9.79 pypi_0 pypi
nvidia-cudnn-cu12 9.12.0.46 pypi_0 pypi

2. Download 0,1,2 sh file and combine
3. run the sh file. up to this everything went normal

but when I try to train here is the error

2025-08-19 19:20:22.120030: W tensorflow/compiler/mlir/tools/kernel_gen/tf_gpu_runtime_wrappers.cc:40] 'cuModuleGetFunction(&function, module, kernel_name)' failed with 'CUDA_ERROR_INVALID_HANDLE'

2025-08-19 19:20:22.120039: W tensorflow/core/framework/op_kernel.cc:1829] INTERNAL: 'cuLaunchKernel(function, gridX, gridY, gridZ, blockX, blockY, blockZ, 0, reinterpret_cast(stream), params, nullptr)' failed with 'CUDA_ERROR_INVALID_HANDLE'
2025-08-19 19:20:22.122646: W tensorflow/compiler/mlir/tools/kernel_gen/tf_gpu_runtime_wrappers.cc:40] 'cuModuleLoadData(&module, data)' failed with 'CUDA_ERROR_INVALID_PTX'

2025-08-19 19:20:22.122659: W tensorflow/compiler/mlir/tools/kernel_gen/tf_gpu_runtime_wrappers.cc:40] 'cuModuleGetFunction(&function, module, kernel_name)' failed with 'CUDA_ERROR_INVALID_HANDLE'

2025-08-19 19:20:22.122665: W tensorflow/core/framework/op_kernel.cc:1829] INTERNAL: 'cuLaunchKernel(function, gridX, gridY, gridZ, blockX, blockY, blockZ, 0, reinterpret_cast(stream), params, nullptr)' failed with 'CUDA_ERROR_INVALID_HANDLE'
2025-08-19 19:20:22.125076: W tensorflow/compiler/mlir/tools/kernel_gen/tf_gpu_runtime_wrappers.cc:40] 'cuModuleLoadData(&module, data)' failed with 'CUDA_ERROR_INVALID_PTX'

2025-08-19 19:20:22.125088: W tensorflow/compiler/mlir/tools/kernel_gen/tf_gpu_runtime_wrappers.cc:40] 'cuModuleGetFunction(&function, module, kernel_name)' failed with 'CUDA_ERROR_INVALID_HANDLE'

thank you very much for the help

[njzjz]

Related to tensorflow/tensorflow#89272 and conda-forge/tensorflow-feedstock#436 - the compute_120 architecture was not built.

[brunobruzzo]

Is there a way to build dp with tf-nightly[and-cuda]?

Trying:

pip install tf-nightly[and-cuda]
pip install deepmd-kit[gpu] --no-binary deepmd-kit

Fails with:

$ pip install deepmd-kit[gpu] --no-binary deepmd-kit
Collecting deepmd-kit[gpu]
  Using cached deepmd_kit-3.1.2.tar.gz (1.3 MB)
  Installing build dependencies ... done
  Getting requirements to build wheel ... error
  error: subprocess-exited-with-error
  
  × Getting requirements to build wheel did not run successfully.
  │ exit code: 1
  ╰─> [32 lines of output]
      Traceback (most recent call last):
        File "/home/bruno/micromamba/envs/dp_env/lib/python3.11/site-packages/pip/_vendor/pyproject_hooks/_in_process/_in_process.py", line 389, in <module>
          main()
        File "/home/bruno/micromamba/envs/dp_env/lib/python3.11/site-packages/pip/_vendor/pyproject_hooks/_in_process/_in_process.py", line 373, in main
          json_out["return_val"] = hook(**hook_input["kwargs"])
                                   ^^^^^^^^^^^^^^^^^^^^^^^^^^^^
        File "/home/bruno/micromamba/envs/dp_env/lib/python3.11/site-packages/pip/_vendor/pyproject_hooks/_in_process/_in_process.py", line 137, in get_requires_for_build_wheel
          backend = _build_backend()
                    ^^^^^^^^^^^^^^^^
        File "/home/bruno/micromamba/envs/dp_env/lib/python3.11/site-packages/pip/_vendor/pyproject_hooks/_in_process/_in_process.py", line 70, in _build_backend
          obj = import_module(mod_path)
                ^^^^^^^^^^^^^^^^^^^^^^^
        File "/home/bruno/micromamba/envs/dp_env/lib/python3.11/importlib/__init__.py", line 126, in import_module
          return _bootstrap._gcd_import(name[level:], package, level)
                 ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
        File "<frozen importlib._bootstrap>", line 1204, in _gcd_import
        File "<frozen importlib._bootstrap>", line 1176, in _find_and_load
        File "<frozen importlib._bootstrap>", line 1147, in _find_and_load_unlocked
        File "<frozen importlib._bootstrap>", line 690, in _load_unlocked
        File "<frozen importlib._bootstrap_external>", line 940, in exec_module
        File "<frozen importlib._bootstrap>", line 241, in _call_with_frames_removed
        File "/tmp/pip-install-4x5x41as/deepmd-kit_30584a365c514aec8684ace739bd5f65/backend/dp_backend.py", line 31, in <module>
          set_scikit_build_env()
        File "/tmp/pip-install-4x5x41as/deepmd-kit_30584a365c514aec8684ace739bd5f65/backend/read_env.py", line 137, in set_scikit_build_env
          cmake_minimum_required_version, cmake_args, _, _, _, _ = get_argument_from_env()
                                                                   ^^^^^^^^^^^^^^^^^^^^^^^
        File "/tmp/pip-install-4x5x41as/deepmd-kit_30584a365c514aec8684ace739bd5f65/backend/read_env.py", line 90, in get_argument_from_env
          tf_version = get_tf_version(tf_install_dir)
                       ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
        File "/tmp/pip-install-4x5x41as/deepmd-kit_30584a365c514aec8684ace739bd5f65/backend/find_tensorflow.py", line 257, in get_tf_version
          raise RuntimeError("Failed to read TF version")
      RuntimeError: Failed to read TF version
      [end of output]
  
  note: This error originates from a subprocess, and is likely not a problem with pip.
ERROR: Failed to build 'deepmd-kit' when getting requirements to build wheel