TensorFlow with CUDA: RTX 5xxx series isn't supported (CUDA_ERROR_INVALID_HANDLE)

### CUDA/cuDNN version
12.9

### GPU model and memory
5070TI(16G)

### Command
```
uv run playground/open_duck_mini_v2/runner.py
```

### Relevant log output

```bash
(mujoco) jay@USER-20250603TE:~/Open_Duck_Playground$ uv run playground/open_duck_mini_v2/runner.py
2025-06-17 09:03:38.354798: E external/local_xla/xla/stream_executor/cuda/cuda_fft.cc:467] Unable to register cuFFT factory: Attempting to register factory for plugin cuFFT when one has already been registered
WARNING: All log messages before absl::InitializeLog() is called are written to STDERR
E0000 00:00:1750122218.405515 2972 cuda_dnn.cc:8579] Unable to register cuDNN factory: Attempting to register factory for plugin cuDNN when one has already been registered
E0000 00:00:1750122218.420688 2972 cuda_blas.cc:1407] Unable to register cuBLAS factory: Attempting to register factory for plugin cuBLAS when one has already been registered
W0000 00:00:1750122218.514032 2972 computation_placer.cc:177] computation placer already registered. Please check linkage and avoid linking the same target more than once.
W0000 00:00:1750122218.514061 2972 computation_placer.cc:177] computation placer already registered. Please check linkage and avoid linking the same target more than once.
W0000 00:00:1750122218.514063 2972 computation_placer.cc:177] computation placer already registered. Please check linkage and avoid linking the same target more than once.
W0000 00:00:1750122218.514064 2972 computation_placer.cc:177] computation placer already registered. Please check linkage and avoid linking the same target more than once.
xml: /home/jay/Open_Duck_Playground/playground/open_duck_mini_v2/xmls/scene_flat_terrain.xml
actuators: ['left_hip_yaw', 'left_hip_roll', 'left_hip_pitch', 'left_knee', 'left_ankle', 'neck_pitch', 'head_pitch', 'head_yaw', 'head_roll', 'right_hip_yaw', 'right_hip_roll', 'right_hip_pitch', 'right_knee', 'right_ankle']
joints: ['floating_base', 'left_hip_yaw', 'left_hip_roll', 'left_hip_pitch', 'left_knee', 'left_ankle', 'neck_pitch', 'head_pitch', 'head_yaw', 'head_roll', 'right_hip_yaw', 'right_hip_roll', 'right_hip_pitch', 'right_knee', 'right_ankle']
backlash joints: []
actuator joints ids: [1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14]
actuator joints dict: {'left_hip_yaw': 1, 'left_hip_roll': 2, 'left_hip_pitch': 3, 'left_knee': 4, 'left_ankle': 5, 'neck_pitch': 6, 'head_pitch': 7, 'head_yaw': 8, 'head_roll': 9, 'right_hip_yaw': 10, 'right_hip_roll': 11, 'right_hip_pitch': 12, 'right_knee': 13, 'right_ankle': 14}
floating qpos addr: 0 qvel addr: 0
[Poly ref data] Processing ...
[Poly ref data] Done processing
xml: /home/jay/Open_Duck_Playground/playground/open_duck_mini_v2/xmls/scene_flat_terrain.xml
actuators: ['left_hip_yaw', 'left_hip_roll', 'left_hip_pitch', 'left_knee', 'left_ankle', 'neck_pitch', 'head_pitch', 'head_yaw', 'head_roll', 'right_hip_yaw', 'right_hip_roll', 'right_hip_pitch', 'right_knee', 'right_ankle']
joints: ['floating_base', 'left_hip_yaw', 'left_hip_roll', 'left_hip_pitch', 'left_knee', 'left_ankle', 'neck_pitch', 'head_pitch', 'head_yaw', 'head_roll', 'right_hip_yaw', 'right_hip_roll', 'right_hip_pitch', 'right_knee', 'right_ankle']
backlash joints: []
actuator joints ids: [1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14]
actuator joints dict: {'left_hip_yaw': 1, 'left_hip_roll': 2, 'left_hip_pitch': 3, 'left_knee': 4, 'left_ankle': 5, 'neck_pitch': 6, 'head_pitch': 7, 'head_yaw': 8, 'head_roll': 9, 'right_hip_yaw': 10, 'right_hip_roll': 11, 'right_hip_pitch': 12, 'right_knee': 13, 'right_ankle': 14}
floating qpos addr: 0 qvel addr: 0
[Poly ref data] Processing ...
[Poly ref data] Done processing
Observation size: 101
PPO params: {'action_repeat': 1, 'batch_size': 256, 'clipping_epsilon': 0.2, 'discounting': 0.97, 'entropy_cost': 0.005, 'episode_length': 1000, 'learning_rate': 0.0003, 'max_grad_norm': 1.0, 'normalize_observations': True, 'num_envs': 8192, 'num_evals': 15, 'num_minibatches': 32, 'num_resets_per_eval': 1, 'num_timesteps': 150000000, 'num_updates_per_batch': 4, 'reward_scaling': 1.0, 'unroll_length': 20}
/home/jay/Open_Duck_Playground/.venv/lib/python3.11/site-packages/jax/_src/interpreters/xla.py:112: RuntimeWarning: overflow encountered in cast
return np.asarray(x, dtypes.canonicalize_dtype(x.dtype))
-----------
STEP: 0 reward: 14.808151245117188 reward_std: 12.227242469787598
-----------
Saving checkpoint (step: 0): /home/jay/Open_Duck_Playground/checkpoints/2025_06_17_090411_0
=== EXPORT ONNX ===
W0000 00:00:1750122251.239234 2972 gpu_device.cc:2430] TensorFlow was not built with CUDA kernel binaries compatible with compute capability 12.0. CUDA kernels will be jit-compiled from PTX, which could take 30 minutes or longer.
W0000 00:00:1750122251.241327 2972 gpu_device.cc:2430] TensorFlow was not built with CUDA kernel binaries compatible with compute capability 12.0. CUDA kernels will be jit-compiled from PTX, which could take 30 minutes or longer.
I0000 00:00:1750122251.241767 2972 gpu_device.cc:2019] Created device /job:localhost/replica:0/task:0/device:GPU:0 with 708 MB memory: -> device: 0, name: NVIDIA GeForce RTX 5070 Ti, pci bus id: 0000:01:00.0, compute capability: 12.0
(101,) (101,)
2025-06-17 09:04:11.336528: W tensorflow/compiler/mlir/tools/kernel_gen/tf_gpu_runtime_wrappers.cc:40] 'cuModuleLoadData(&module, data)' failed with 'CUDA_ERROR_INVALID_PTX'

2025-06-17 09:04:11.336557: W tensorflow/compiler/mlir/tools/kernel_gen/tf_gpu_runtime_wrappers.cc:40] 'cuModuleGetFunction(&function, module, kernel_name)' failed with 'CUDA_ERROR_INVALID_HANDLE'

2025-06-17 09:04:11.336740: W tensorflow/core/framework/op_kernel.cc:1844] INTERNAL: 'cuLaunchKernel(function, gridX, gridY, gridZ, blockX, blockY, blockZ, 0, reinterpret_cast(stream), params, nullptr)' failed with 'CUDA_ERROR_INVALID_HANDLE'
(101,) (101,)
2025-06-17 09:04:11.340111: W tensorflow/compiler/mlir/tools/kernel_gen/tf_gpu_runtime_wrappers.cc:40] 'cuModuleLoadData(&module, data)' failed with 'CUDA_ERROR_INVALID_PTX'

2025-06-17 09:04:11.340135: W tensorflow/compiler/mlir/tools/kernel_gen/tf_gpu_runtime_wrappers.cc:40] 'cuModuleGetFunction(&function, module, kernel_name)' failed with 'CUDA_ERROR_INVALID_HANDLE'

2025-06-17 09:04:11.340151: W tensorflow/core/framework/op_kernel.cc:1844] INTERNAL: 'cuLaunchKernel(function, gridX, gridY, gridZ, blockX, blockY, blockZ, 0, reinterpret_cast(stream), params, nullptr)' failed with 'CUDA_ERROR_INVALID_HANDLE'
Traceback (most recent call last):
File "/home/jay/Open_Duck_Playground/playground/open_duck_mini_v2/runner.py", line 64, in
main()
File "/home/jay/Open_Duck_Playground/playground/open_duck_mini_v2/runner.py", line 60, in main
runner.train()
File "/home/jay/Open_Duck_Playground/playground/common/runner.py", line 114, in train
_, params, _ = train_fn(
^^^^^^^^^
File "/home/jay/Open_Duck_Playground/.venv/lib/python3.11/site-packages/brax/training/agents/ppo/train.py", line 692, in train
policy_params_fn(current_step, make_policy, params)
File "/home/jay/Open_Duck_Playground/playground/common/runner.py", line 78, in policy_params_fn
export_onnx(
File "/home/jay/Open_Duck_Playground/playground/common/export_onnx.py", line 105, in export_onnx
example_output = tf_policy_network(example_input)
^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
File "/home/jay/Open_Duck_Playground/.venv/lib/python3.11/site-packages/keras/src/utils/traceback_utils.py", line 122, in error_handler
raise e.with_traceback(filtered_tb) from None
File "/home/jay/Open_Duck_Playground/playground/common/export_onnx.py", line 69, in call
inputs = (inputs - self.mean) / self.std
~~~~~~~^~~~~~~~~~~
tensorflow.python.framework.errors_impl.InternalError: Exception encountered when calling MLP.call().

{{function_node __wrapped__Sub_device_/job:localhost/replica:0/task:0/device:GPU:0}} 'cuLaunchKernel(function, gridX, gridY, gridZ, blockX, blockY, blockZ, 0, reinterpret_cast(stream), params, nullptr)' failed with 'CUDA_ERROR_INVALID_HANDLE' [Op:Sub] name:

Arguments received by MLP.call():
• inputs=tf.Tensor(shape=(1, 101), dtype=float32)
```

anyone has any idea how to solve this problem?