[Bug]: The lstmsequence model cannot be compiled on npu device.

[luaperl]

OpenVINO Version

2024.1.0

Operating System

Windows System

Device used for inference

NPU

Framework

None

Model used

Custom

Issue description

I defined the model using openvino opset's lstmsequence and compiled it with an NPU device, but it failed.
It compiled normally on CPU and GPU.
NPU driver was installed using this link.

If it was my mistake, please let me know what I did wrong.

• Env
  • CPU: Intel(R) Core(TM) Ultra 8 155H
  • Window version: 23H2 22631.3737
  • NPU driver version: npu_win_32.0.100.2408

Step-by-step reproduction

import openvino as ov 
ops = ov.runtime.opset13

# param 
B = 1 
T = 300
D = 512
H = 256
DIR = 2

# set input for lstm
X                    = ops.parameter([B, T, D], ov.Type.f32)
initial_hidden_state = ops.parameter([B, DIR, H], ov.Type.f32)
initial_cell_state   = ops.parameter([B, DIR, H], ov.Type.f32)
sequence_lengths     = ops.parameter([B], ov.Type.i64)
W                    = ops.parameter([DIR, 4*H, D], ov.Type.f32)
R                    = ops.parameter([DIR, 4*H, H], ov.Type.f32)
B                    = ops.parameter([DIR, 4*H], ov.Type.f32)

# define and compile model 
lstm = ops.lstm_sequence(
    X=X, 
    initial_hidden_state=initial_hidden_state, 
    initial_cell_state=initial_cell_state, 
    sequence_lengths=sequence_lengths, 
    W=W, 
    R=R, 
    B=B, 
    hidden_size=H,
    direction="bidirectional",
)
model = ov.Model([lstm.output(0), lstm.output(1), lstm.output(2)], [X, initial_hidden_state, initial_cell_state, sequence_lengths, W, R, B], "Model")
model = ov.compile_model(model, "NPU")
print(model)

Relevant log output

Traceback (most recent call last):
  File "C:\Users\test\Desktop\noah\redle\test.py", line 33, in <module>
    model = ov.compile_model(model, "NPU")
            ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
  File "C:\Users\test\AppData\Local\Programs\Python\Python312\Lib\site-packages\openvino\runtime\ie_api.py", line 609, in compile_model
    return core.compile_model(model, device_name, {} if config is None else config)
           ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
  File "C:\Users\test\AppData\Local\Programs\Python\Python312\Lib\site-packages\openvino\runtime\ie_api.py", line 521, in compile_model
    super().compile_model(model, device_name, {} if config is None else config),
    ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
RuntimeError: Exception from src\inference\src\cpp\core.cpp:109:
Exception from src\inference\src\dev\plugin.cpp:54:
Exception from src\plugins\intel_npu\src\plugin\src\plugin.cpp:513:
Check 'result == ZE_RESULT_SUCCESS' failed at src\plugins\intel_npu\src\compiler\src\zero_compiler_in_driver.cpp:753:
Failed to compile network. L0 createGraph result: ZE_RESULT_ERROR_INVALID_ARGUMENT, code 0x78000004. nGraph TensorIterator node 'LSTMSequence_1568' has incompatible output number 6 with its body outputs descriptions concatOutput '1' + invariantOutput '2'
Failed to create executable

Issue submission checklist

• I'm reporting an issue. It's not a question.
• I checked the problem with the documentation, FAQ, open issues, Stack Overflow, etc., and have not found a solution.
• There is reproducer code and related data files such as images, videos, models, etc.