GPU使用报错 #74192
Description
请提出你的问题 Please ask your question
在运行10分钟快速上手的例程的时候已经报了检测到GPU，W0723 15:29:00.590332 3208 gpu_resources.cc:114] Please NOTE: device: 0, GPU Compute Capability: 6.1, Driver API Version: 12.9, Runtime API Version: 12.6。但是在训练的时候报错
D:\paddle\venv\Lib\site-packages\paddle\hapi\model.py:1245: UserWarning: To copy construct from a tensor, it is recommended to use sourceTensor.clone().detach(), rather than paddle.to_tensor(sourceTensor).
labels = [paddle.to_tensor(l) for l in to_list(labels)]
Traceback (most recent call last):
File "D:\paddle\test.py", line 24, in
model.fit(train_dataset, epochs=5, batch_size=64, verbose=1)
File "D:\paddle\venv\Lib\site-packages\paddle\hapi\model.py", line 2429, in fit
logs = self._run_one_epoch(train_loader, cbks, 'train')
File "D:\paddle\venv\Lib\site-packages\paddle\hapi\model.py", line 2811, in _run_one_epoch
outs = getattr(self, mode + '_batch')(_inputs)
File "D:\paddle\venv\Lib\site-packages\paddle\hapi\model.py", line 1680, in train_batch
loss = self._adapter.train_batch(inputs, labels, update)
File "D:\paddle\venv\Lib\site-packages\paddle\hapi\model.py", line 1281, in train_batch
metric_outs = metric.compute((to_list(outputs) + labels))
File "D:\paddle\venv\Lib\site-packages\paddle\metric\metrics.py", line 282, in compute
pred = paddle.argsort(pred, descending=True)
File "D:\paddle\venv\Lib\site-packages\paddle\tensor\search.py", line 140, in argsort
_, ids = _C_ops.argsort(x, axis, descending, stable)
OSError: (External) CUDA error(101), invalid device ordinal.
[Hint: 'cudaErrorInvalidDevice'. This indicates that the device ordinal supplied by the user does not correspond to a valid CUDA device or that the actionrequested is invalid for the specified device.] (at C:\home\workspace\Paddle\paddle\phi\kernels\gpu\argsort_kernel.cu:271)
请问这个是什么问题啊？
附：系统信息
Win10 21H1
GPU:1080Ti
GPU驱动版本：576.88
CUDA版本：12.6
CuDNN版本：V9.6
TensorTR未安装