Question

以某种方式，当我使用bfloat16时，tensorflow无法检测到我的GPU。

是否可以在GPU上使用bfloat16？或者我只能在张量处理单元上使用bfloat16？

错误消息是否意味着张量流无法计算梯度？</ p>

tensorflow.python.framework.errors_impl.NotFoundError: No registered 'Reciprocal' OpKernel for CPU devices compatible with node {{node ConstantFolding/foo/gradients/foo/Mean_1_grad/truediv_recip}} = Reciprocal[T=DT_BFLOAT16, _device="/job:localhost/replica:0/task:0/device:CPU:0"](foo/gradients/foo/Mean_1_grad/Const_1)
     (OpKernel was found, but attributes didn't match)
    .  Registered:  device='XLA_GPU'; T in [DT_FLOAT, DT_DOUBLE, DT_INT32, DT_COMPLEX64, DT_INT64, DT_BFLOAT16, DT_HALF]
  device='XLA_CPU'; T in [DT_FLOAT, DT_DOUBLE, DT_INT32, DT_COMPLEX64, DT_INT64, DT_HALF]
  device='XLA_CPU_JIT'; T in [DT_FLOAT, DT_DOUBLE, DT_INT32, DT_COMPLEX64, DT_INT64, DT_HALF]
  device='XLA_GPU_JIT'; T in [DT_FLOAT, DT_DOUBLE, DT_INT32, DT_COMPLEX64, DT_INT64, DT_BFLOAT16, DT_HALF]
  device='GPU'; T in [DT_INT64]
  device='GPU'; T in [DT_DOUBLE]
  device='GPU'; T in [DT_HALF]
  device='GPU'; T in [DT_FLOAT]
  device='CPU'; T in [DT_COMPLEX128]
  device='CPU'; T in [DT_COMPLEX64]
  device='CPU'; T in [DT_DOUBLE]
  device='CPU'; T in [DT_HALF]
  device='CPU'; T in [DT_FLOAT]

     [[{{node ConstantFolding/foo/gradients/foo/Mean_1_grad/truediv_recip}} = Reciprocal[T=DT_BFLOAT16, _device="/job:localhost/replica:0/task:0/device:CPU:0"](foo/gradients/foo/Mean_1_grad/Const_1)]]

在GPU上使用bfloat16和tensorflow

0 个答案: