We are trying to have caffe2 compiled on AIX. We were able to successfully compile it and run a few caffe2 sample code. But,a RNN sample script fails with an exception as shown below. Any pointers on the cause of this will be helpful.
python char_rnn.py --train_data input.txt
WARNING:root:This caffe2 python run does not have GPU support. Will run in CPU only mode.
WARNING:root:Debug message: No module named caffe2_pybind11_state_hip
Input has 65 characters. Total input size: 1115394
DEBUG:char_rnn:Start training
Entering interactive debugger. Type “bt” to print the full stacktrace. Type “help” to see command listing.
Exception when creating gradient for [LSTMUnit]:[enforce fail at pybind_state.cc:523] caffe2::GradientRegistry()->Has(def.type()). .
Op:
input: “LSTM/hidden_t_prev”
input: “LSTM/cell_t_prev”
input: “LSTM/gates_t”
input: “seq_lengths”
input: “timestep”
output: “LSTM/hidden_state”
output: “LSTM/cell_state”
name: “”
type: “LSTMUnit”
arg {
name: “sequence_lengths”
i: 1
}
arg {
name: “drop_states”
i: 0
}
arg {
name: “forget_bias”
f: 0.0
}
device_option {
device_type: 0
cuda_gpu_id: 0
}
/usr/local/lib/python2.7/site-packages/caffe2/python/core.py(1108)GetGradientForOp()
→ format(op.type, e, str(op))
(Pdb)