Onnxruntime: InvalidArgument: [ONNXRuntimeError] : 2 : INVALID_ARGUMENT : Unexpected input data type. Actual: (N11onnxruntime17PrimitiveDataTypeINS_9MLFloat16EEE) , expected: (N11onnxruntime17PrimitiveDataTypeIfEE)
Describe the bug
Dear Helper,
I have a tensorflow model
Model: "model"
__________________________________________________________________________________________________
Layer (type) Output Shape Param # Connected to
==================================================================================================
em_barrel_Lr1 (InputLayer) [(None, 56, 11)] 0
__________________________________________________________________________________________________
em_barrel_Lr0 (InputLayer) [(None, 7, 11)] 0
__________________________________________________________________________________________________
em_barrel_Lr2 (InputLayer) [(None, 7, 11)] 0
__________________________________________________________________________________________________
em_barrel_Lr3 (InputLayer) [(None, 7, 11)] 0
__________________________________________________________________________________________________
tile_barrel_Lr1 (InputLayer) [(None, 7, 11)] 0
__________________________________________________________________________________________________
tile_barrel_Lr2 (InputLayer) [(None, 7, 11)] 0
__________________________________________________________________________________________________
tile_barrel_Lr3 (InputLayer) [(None, 7, 11)] 0
__________________________________________________________________________________________________
reshape_7 (Reshape) (None, 56, 11, 1) 0 em_barrel_Lr1[0][0]
__________________________________________________________________________________________________
tracks_image (InputLayer) [(None, 4, 13)] 0
__________________________________________________________________________________________________
reshape_1 (Reshape) (None, 7, 11, 1) 0 em_barrel_Lr0[0][0]
__________________________________________________________________________________________________
reshape_2 (Reshape) (None, 7, 11, 1) 0 em_barrel_Lr2[0][0]
__________________________________________________________________________________________________
reshape_3 (Reshape) (None, 7, 11, 1) 0 em_barrel_Lr3[0][0]
__________________________________________________________________________________________________
reshape_4 (Reshape) (None, 7, 11, 1) 0 tile_barrel_Lr1[0][0]
__________________________________________________________________________________________________
reshape_5 (Reshape) (None, 7, 11, 1) 0 tile_barrel_Lr2[0][0]
__________________________________________________________________________________________________
reshape_6 (Reshape) (None, 7, 11, 1) 0 tile_barrel_Lr3[0][0]
__________________________________________________________________________________________________
conv2d_4 (Conv2D) (None, 54, 9, 200) 2000 reshape_7[0][0]
__________________________________________________________________________________________________
reshape (Reshape) (None, 4, 13, 1) 0 tracks_image[0][0]
__________________________________________________________________________________________________
concatenate (Concatenate) (None, 7, 11, 6) 0 reshape_1[0][0]
reshape_2[0][0]
reshape_3[0][0]
reshape_4[0][0]
reshape_5[0][0]
reshape_6[0][0]
__________________________________________________________________________________________________
leaky_re_lu_4 (LeakyReLU) (None, 54, 9, 200) 0 conv2d_4[0][0]
__________________________________________________________________________________________________
conv2d (Conv2D) (None, 4, 13, 200) 400 reshape[0][0]
__________________________________________________________________________________________________
conv2d_2 (Conv2D) (None, 6, 9, 200) 7400 concatenate[0][0]
__________________________________________________________________________________________________
max_pooling2d (MaxPooling2D) (None, 27, 4, 200) 0 leaky_re_lu_4[0][0]
__________________________________________________________________________________________________
leaky_re_lu (LeakyReLU) (None, 4, 13, 200) 0 conv2d[0][0]
__________________________________________________________________________________________________
leaky_re_lu_2 (LeakyReLU) (None, 6, 9, 200) 0 conv2d_2[0][0]
__________________________________________________________________________________________________
dropout_4 (Dropout) (None, 27, 4, 200) 0 max_pooling2d[0][0]
__________________________________________________________________________________________________
dropout (Dropout) (None, 4, 13, 200) 0 leaky_re_lu[0][0]
__________________________________________________________________________________________________
dropout_2 (Dropout) (None, 6, 9, 200) 0 leaky_re_lu_2[0][0]
__________________________________________________________________________________________________
conv2d_5 (Conv2D) (None, 25, 2, 200) 360200 dropout_4[0][0]
__________________________________________________________________________________________________
conv2d_1 (Conv2D) (None, 4, 13, 200) 40200 dropout[0][0]
__________________________________________________________________________________________________
conv2d_3 (Conv2D) (None, 5, 7, 200) 240200 dropout_2[0][0]
__________________________________________________________________________________________________
leaky_re_lu_5 (LeakyReLU) (None, 25, 2, 200) 0 conv2d_5[0][0]
__________________________________________________________________________________________________
leaky_re_lu_1 (LeakyReLU) (None, 4, 13, 200) 0 conv2d_1[0][0]
__________________________________________________________________________________________________
leaky_re_lu_3 (LeakyReLU) (None, 5, 7, 200) 0 conv2d_3[0][0]
__________________________________________________________________________________________________
max_pooling2d_1 (MaxPooling2D) (None, 12, 1, 200) 0 leaky_re_lu_5[0][0]
__________________________________________________________________________________________________
dropout_1 (Dropout) (None, 4, 13, 200) 0 leaky_re_lu_1[0][0]
__________________________________________________________________________________________________
dropout_3 (Dropout) (None, 5, 7, 200) 0 leaky_re_lu_3[0][0]
__________________________________________________________________________________________________
dropout_5 (Dropout) (None, 12, 1, 200) 0 max_pooling2d_1[0][0]
__________________________________________________________________________________________________
p_Eratio (InputLayer) [(None,)] 0
__________________________________________________________________________________________________
p_Reta (InputLayer) [(None,)] 0
__________________________________________________________________________________________________
p_Rhad (InputLayer) [(None,)] 0
__________________________________________________________________________________________________
p_Rphi (InputLayer) [(None,)] 0
__________________________________________________________________________________________________
p_TRTPID (InputLayer) [(None,)] 0
__________________________________________________________________________________________________
p_numberOfSCTHits (InputLayer) [(None,)] 0
__________________________________________________________________________________________________
p_ndof (InputLayer) [(None,)] 0
__________________________________________________________________________________________________
p_dPOverP (InputLayer) [(None,)] 0
__________________________________________________________________________________________________
p_deltaEta1 (InputLayer) [(None,)] 0
__________________________________________________________________________________________________
p_f1 (InputLayer) [(None,)] 0
__________________________________________________________________________________________________
p_f3 (InputLayer) [(None,)] 0
__________________________________________________________________________________________________
p_deltaPhiRescaled2 (InputLayer [(None,)] 0
__________________________________________________________________________________________________
p_weta2 (InputLayer) [(None,)] 0
__________________________________________________________________________________________________
p_d0 (InputLayer) [(None,)] 0
__________________________________________________________________________________________________
p_d0Sig (InputLayer) [(None,)] 0
__________________________________________________________________________________________________
p_qd0Sig (InputLayer) [(None,)] 0
__________________________________________________________________________________________________
p_nTracks (InputLayer) [(None,)] 0
__________________________________________________________________________________________________
p_sct_weight_charge (InputLayer [(None,)] 0
__________________________________________________________________________________________________
flatten (Flatten) (None, 10400) 0 dropout_1[0][0]
__________________________________________________________________________________________________
flatten_1 (Flatten) (None, 7000) 0 dropout_3[0][0]
__________________________________________________________________________________________________
flatten_2 (Flatten) (None, 2400) 0 dropout_5[0][0]
__________________________________________________________________________________________________
flatten_3 (Flatten) (None, 1) 0 p_Eratio[0][0]
__________________________________________________________________________________________________
flatten_4 (Flatten) (None, 1) 0 p_Reta[0][0]
__________________________________________________________________________________________________
flatten_5 (Flatten) (None, 1) 0 p_Rhad[0][0]
__________________________________________________________________________________________________
flatten_6 (Flatten) (None, 1) 0 p_Rphi[0][0]
__________________________________________________________________________________________________
flatten_7 (Flatten) (None, 1) 0 p_TRTPID[0][0]
__________________________________________________________________________________________________
flatten_8 (Flatten) (None, 1) 0 p_numberOfSCTHits[0][0]
__________________________________________________________________________________________________
flatten_9 (Flatten) (None, 1) 0 p_ndof[0][0]
__________________________________________________________________________________________________
flatten_10 (Flatten) (None, 1) 0 p_dPOverP[0][0]
__________________________________________________________________________________________________
flatten_11 (Flatten) (None, 1) 0 p_deltaEta1[0][0]
__________________________________________________________________________________________________
flatten_12 (Flatten) (None, 1) 0 p_f1[0][0]
__________________________________________________________________________________________________
flatten_13 (Flatten) (None, 1) 0 p_f3[0][0]
__________________________________________________________________________________________________
flatten_14 (Flatten) (None, 1) 0 p_deltaPhiRescaled2[0][0]
__________________________________________________________________________________________________
flatten_15 (Flatten) (None, 1) 0 p_weta2[0][0]
__________________________________________________________________________________________________
flatten_16 (Flatten) (None, 1) 0 p_d0[0][0]
__________________________________________________________________________________________________
flatten_17 (Flatten) (None, 1) 0 p_d0Sig[0][0]
__________________________________________________________________________________________________
flatten_18 (Flatten) (None, 1) 0 p_qd0Sig[0][0]
__________________________________________________________________________________________________
flatten_19 (Flatten) (None, 1) 0 p_nTracks[0][0]
__________________________________________________________________________________________________
flatten_20 (Flatten) (None, 1) 0 p_sct_weight_charge[0][0]
__________________________________________________________________________________________________
concatenate_1 (Concatenate) (None, 19818) 0 flatten[0][0]
flatten_1[0][0]
flatten_2[0][0]
flatten_3[0][0]
flatten_4[0][0]
flatten_5[0][0]
flatten_6[0][0]
flatten_7[0][0]
flatten_8[0][0]
flatten_9[0][0]
flatten_10[0][0]
flatten_11[0][0]
flatten_12[0][0]
flatten_13[0][0]
flatten_14[0][0]
flatten_15[0][0]
flatten_16[0][0]
flatten_17[0][0]
flatten_18[0][0]
flatten_19[0][0]
flatten_20[0][0]
__________________________________________________________________________________________________
dense (Dense) (None, 200) 3963800 concatenate_1[0][0]
__________________________________________________________________________________________________
leaky_re_lu_6 (LeakyReLU) (None, 200) 0 dense[0][0]
__________________________________________________________________________________________________
dropout_6 (Dropout) (None, 200) 0 leaky_re_lu_6[0][0]
__________________________________________________________________________________________________
dense_1 (Dense) (None, 200) 40200 dropout_6[0][0]
__________________________________________________________________________________________________
leaky_re_lu_7 (LeakyReLU) (None, 200) 0 dense_1[0][0]
__________________________________________________________________________________________________
dropout_7 (Dropout) (None, 200) 0 leaky_re_lu_7[0][0]
__________________________________________________________________________________________________
dense_2 (Dense) (None, 2) 402 dropout_7[0][0]
==================================================================================================
Total params: 4,654,802
Trainable params: 4,654,802
Non-trainable params: 0
__________________________________________________________________________________________________
which converts to .onnx model, and I have a sample to input in the model as dictionary with following keys.
'em_barrel_Lr0', 'em_barrel_Lr1', 'em_barrel_Lr2', 'em_barrel_Lr3', 'tile_barrel_Lr1',
'tile_barrel_Lr2', 'tile_barrel_Lr3', 'tracks_image','p_Eratio', 'p_Reta', 'p_Rhad',
'p_Rphi', 'p_TRTPID', 'p_numberOfSCTHits', 'p_ndof', 'p_dPOverP',
'p_deltaEta1', 'p_f1', 'p_f3', 'p_deltaPhiRescaled2', 'p_weta2',
'p_d0', 'p_d0Sig', 'p_qd0Sig', 'p_nTracks', 'p_sct_weight_charge'
I tried to feed the sample to the .onnx model by
input_feeds = {}
key = ['em_barrel_Lr0', 'em_barrel_Lr1', 'em_barrel_Lr2', 'em_barrel_Lr3', 'tile_barrel_Lr1',
'tile_barrel_Lr2', 'tile_barrel_Lr3', 'tracks_image','p_Eratio', 'p_Reta', 'p_Rhad',
'p_Rphi', 'p_TRTPID', 'p_numberOfSCTHits', 'p_ndof', 'p_dPOverP',
'p_deltaEta1', 'p_f1', 'p_f3', 'p_deltaPhiRescaled2', 'p_weta2',
'p_d0', 'p_d0Sig', 'p_qd0Sig', 'p_nTracks', 'p_sct_weight_charge']
for i in range(26):
input_feeds[sess_ort.get_inputs()[i].name] = final_sample1[key[i]]
and then
output_name = sess_ort.get_outputs()[0].name
res = sess_ort.run([output_name], input_feeds)
and I get following error
InvalidArgument Traceback (most recent call last)
<ipython-input-27-830965c44719> in <module>
----> 1 res = sess_ort.run([output_name], input_feeds)
/export/home/zp/gupta/envTf2/lib/python3.7/site-packages/onnxruntime/capi/session.py in run(self, output_names, input_feed, run_options)
140 output_names = [output.name for output in self._outputs_meta]
141 try:
--> 142 return self._sess.run(output_names, input_feed, run_options)
143 except C.EPFail as err:
144 if self._enable_fallback:
InvalidArgument: [ONNXRuntimeError] : 2 : INVALID_ARGUMENT : Unexpected input data type. Actual: (N11onnxruntime17PrimitiveDataTypeINS_9MLFloat16EEE) , expected: (N11onnxruntime17PrimitiveDataTypeIfEE)
need help,
Thanks,
Debo.
Urgency
none
System information
- OS Platform and Distribution (linux centos7):
- ONNX Runtime installed from binary:
- ONNX Runtime version:Latest
- Python version: 3.7
Debottam
All 6 comments
Is it possible to attach the onnx model here? Or you can just email me?
pranavsharma
on 11 May 2020
Hi @pranavsharma ,
Couldn't attach it here I emailed you the .onnx model.
Thanks,
Debo
Debottam
on 11 May 2020
I couldn't repro it with onnxruntime 1.2. Which version are you using? Here's the script I used https://gist.github.com/pranavsharma/f3f9b14a1ccc1e5567995a39568db9b7.
pranavsharma
on 12 May 2020
👍1
Hi @pranavsharma,
Thanks a lot for investigating. I am using onnxruntime 1.1.1 since our c++ api is still 1.1.1 please see the following
ort.__version__
'1.1.1'
and the sample I am using quite big maybe try to put everything in github but for now please see the following, the sample I am using has the following dimention
em_barrel_Lr0 : (1, 7, 11)
em_barrel_Lr1 : (1, 56, 11)
em_barrel_Lr2 : (1, 7, 11)
em_barrel_Lr3 : (1, 7, 11)
tile_barrel_Lr1 : (1, 7, 11)
tile_barrel_Lr2 : (1, 7, 11)
tile_barrel_Lr3 : (1, 7, 11)
tracks_image : (1, 4, 13)
p_Eratio : (1,)
p_Reta : (1,)
p_Rhad : (1,)
p_Rphi : (1,)
p_TRTPID : (1,)
p_numberOfSCTHits : (1,)
p_ndof : (1,)
p_dPOverP : (1,)
p_deltaEta1 : (1,)
p_f1 : (1,)
p_f3 : (1,)
p_deltaPhiRescaled2 : (1,)
p_weta2 : (1,)
p_d0 : (1,)
p_d0Sig : (1,)
p_qd0Sig : (1,)
p_nTracks : (1,)
p_sct_weight_charge : (1,)
and the model I have
em_barrel_Lr0 : ['N', 7, 11]
em_barrel_Lr1 : ['N', 56, 11]
em_barrel_Lr2 : ['N', 7, 11]
em_barrel_Lr3 : ['N', 7, 11]
tile_barrel_Lr1 : ['N', 7, 11]
tile_barrel_Lr2 : ['N', 7, 11]
tile_barrel_Lr3 : ['N', 7, 11]
tracks_image : ['N', 4, 13]
p_Eratio : ['N']
p_Reta : ['N']
p_Rhad : ['N']
p_Rphi : ['N']
p_TRTPID : ['N']
p_numberOfSCTHits : ['N']
p_ndof : ['N']
p_dPOverP : ['N']
p_deltaEta1 : ['N']
p_f1 : ['N']
p_f3 : ['N']
p_deltaPhiRescaled2 : ['N']
p_weta2 : ['N']
p_d0 : ['N']
p_d0Sig : ['N']
p_qd0Sig : ['N']
p_nTracks : ['N']
p_sct_weight_charge : ['N']
The same sample works for the tensorflow model from which the .onnx model developed.
Thanks,
Debo.
Debottam
on 12 May 2020
Hi @pranavsharma ,
I figured out, your example really helped; especially the part you did
d = create_random_data(shape_arr, np.float32, 0, 1, None)
The dataset I was feeding was of type np.float16, e.g.
'p_Eratio': array([0.40392396], dtype=float16),
'p_Reta': array([-0.64157045], dtype=float16),
I just change them to float32 by final_sample1[key].astype(np.float32) and it works with my code.
Thanks a lot,
Debo.
Debottam
on 12 May 2020
❤1
👍1
Glad it worked. Thanks!
pranavsharma
on 12 May 2020
Was this page helpful?
0 / 5 - 0 ratings
Related issues
daquexian
路
5Comments
pranav-prakash
路
4Comments
cbecker
路
4Comments
addisonklinke
路
3Comments
diwakar-ravichandran
路
5Comments
Most helpful comment
Hi @pranavsharma ,
I figured out, your example really helped; especially the part you did
The dataset I was feeding was of type np.float16, e.g.
I just change them to float32 by
final_sample1[key].astype(np.float32)and it works with my code.Thanks a lot,
Debo.