search

microsoft / tensorflow-directml

Fork of TensorFlow accelerated by DirectML
Apache License 2.0
454 stars 32 forks source link

No device assignments were active #365

Open xiaomengxin123 opened 2 years ago

xiaomengxin123 commented 2 years ago

When I use on windows10, just run a complex model ,it failed, the error like this No device assignments were active during op 'Embedding-Token/embedding_lookup/Identity' creation., others are used correctly.

PatriceVignola commented 2 years ago

Hi @xiaomengxin123, we'll need more information to diagnose the issue. What kind of model are you trying to run? Can you repro the error while using the vanilla TensorFlow 1.15 repo with a GPU device? It would help if you could share a script to repro the issue.

xiaomengxin123 commented 2 years ago

Thanks and I see,the detail is ReadVariableOp (ReadVariableOp) /device:DML:0 framework assigned device=/job:localhost/replica:0/task:0/device:DML:0 VarIsInitializedOp (VarIsInitializedOp) /device:DML:0 framework assigned device=/job:localhost/replica:0/task:0/device:DML:0 training/Adam/gradients/Embedding-Token/embedding_lookup_grad/Const (Const) /device:DML:0 training/Adam/gradients/Embedding-Token/embedding_lookup_grad/Size (Size) /device:DML:0 training/Adam/gradients/Embedding-Token/embedding_lookup_grad/ExpandDims/dim (Const) /device:DML:0 training/Adam/gradients/Embedding-Token/embedding_lookup_grad/ExpandDims (ExpandDims) /device:DML:0 training/Adam/gradients/Embedding-Token/embedding_lookup_grad/strided_slice/stack (Const) /device:DML:0 training/Adam/gradients/Embedding-Token/embedding_lookup_grad/strided_slice/stack_1 (Const) /device:DML:0 training/Adam/gradients/Embedding-Token/embedding_lookup_grad/strided_slice/stack_2 (Const) /device:DML:0 training/Adam/gradients/Embedding-Token/embedding_lookup_grad/strided_slice (StridedSlice) /device:DML:0 training/Adam/gradients/Embedding-Token/embedding_lookup_grad/concat/axis (Const) /device:DML:0 training/Adam/gradients/Embedding-Token/embedding_lookup_grad/concat (ConcatV2) /device:DML:0 training/Adam/gradients/Embedding-Token/embedding_lookup_grad/Reshape (Reshape) /device:DML:0 training/Adam/gradients/Embedding-Token/embedding_lookup_grad/Reshape_1 (Reshape) /device:DML:0 training/Adam/gradients/AddN_296/strided_slice/stack (Const) /device:DML:0 training/Adam/gradients/AddN_296/strided_slice/stack_1 (Const) /device:DML:0 training/Adam/gradients/AddN_296/strided_slice/stack_2 (Const) /device:DML:0 training/Adam/gradients/AddN_296/strided_slice (StridedSlice) /device:DML:0 training/Adam/gradients/AddN_296/inputs_1 (UnsortedSegmentSum) /device:DML:0 training/Adam/gradients/AddN_296 (AddN) /device:DML:0 training/Adam/ReadVariableOp_6 (ReadVariableOp) /device:DML:0 training/Adam/AssignVariableOp_2 (AssignVariableOp) /device:DML:0 training/Adam/ReadVariableOp_9 (ReadVariableOp) /device:DML:0 [[node Embedding-Token/embedding_lookup/Identity (defined at \ProgramData\Anaconda3\envs\py37\lib\site-packages\tensorflow_core\python\framework\ops.py:1762) ]]Additional information about colocations:Node-device colocations active during op 'Embedding -Token/embedding_lookup/Identity' creation: with tf.colocate_with(Embedding-Token/embeddings): <G:\ProgramData\Anaconda3\envs\py37\lib\site-packages\tensorflow_core\python\ops\embedding_ops.py:133> No device assignments were active during op 'Embedding-Token/embedding_lookup/Identity' creation. , Maybe it was caused by tensorflow_core and not model .