Tensorflow op [batch_join: QueueDequeueUpToV2] is not supported #1044
Description
Describe the bug
2020-08-07 15:13:14,603 - ERROR - Tensorflow op [batch_join: QueueDequeueUpToV2] is not supported
2020-08-07 15:13:19,636 - ERROR - Unsupported ops: Counter({'QueueDequeueUpToV2': 1})
I apologize in advance if there is a support op list that I should have cross referenced ahead of this. I did look, but I didn't find one. I figure that might be more trouble than its worth, but then again...
Urgency
Low - evaluation phase
This model was chosen due to is structure which is similar to one I will attempt to move from C code directly to the ONNX API.
My alternative is to pull the mobile model from the model zoo.
System information
- OS Platform and Distribution (e.g., Linux Ubuntu 16.04): Linux RHEL 8.1
- Tensorflow Version: tensorflow=1.15.3, onnx=1.7.0, tf2onnx=1.6.2/8d5253
- Python version: 3.3.7
To Reproduce
Download the model from: https://drive.google.com/open?id=1R77HmFADxe87GmoLwzfgMu_HY0IhcyBz
python -m tf2onnx.convert --opset 12 --graphdef 20180408-102900.pb --fold_const --inputs input:0 --outputs embeddings:0
2020-08-07 15:19:10,587 - WARNING - From /fomshare/home/robers/anaconda3/envs/tfonnx/lib/python3.7/site-packages/tf2onnx-1.6.2-py3.7.egg/tf2onnx/verbose_logging.py:76: The name tf.logging.set_verbosity is deprecated. Please use tf.compat.v1.logging.set_verbosity instead.
Output_names from frozen session:
embeddings:0
Input_names from frozen session:
input:0
2020-08-07 15:19:17,712 - INFO - Using tensorflow=1.15.3, onnx=1.7.0, tf2onnx=1.6.2/8d5253
2020-08-07 15:19:17,712 - INFO - Using opset <onnx, 12>
2020-08-07 15:19:17,712 - WARNING - Shape of placeholder batch_size is unknown, treated it as a scalar
2020-08-07 15:19:17,712 - WARNING - Shape of placeholder phase_train is unknown, treated it as a scalar
2020-08-07 15:19:22,703 - WARNING - Cannot infer shape for batch_join: batch_join:0,batch_join:1
2020-08-07 15:19:22,703 - WARNING - Cannot infer shape for image_batch: image_batch:0
2020-08-07 15:19:22,703 - WARNING - Cannot infer shape for input: input:0
2020-08-07 15:19:47,659 - ERROR - Tensorflow op [batch_join: QueueDequeueUpToV2] is not supported
2020-08-07 15:19:52,683 - ERROR - Unsupported ops: Counter({'QueueDequeueUpToV2': 1})
Traceback (most recent call last):
File "/fomshare/home/robers/anaconda3/envs/tfonnx/lib/python3.7/runpy.py", line 193, in _run_module_as_main
"main", mod_spec)
File "/fomshare/home/robers/anaconda3/envs/tfonnx/lib/python3.7/runpy.py", line 85, in _run_code
exec(code, run_globals)
File "/fomshare/home/robers/anaconda3/envs/tfonnx/lib/python3.7/site-packages/tf2onnx-1.6.2-py3.7.egg/tf2onnx/convert.py", line 169, in
File "/fomshare/home/robers/anaconda3/envs/tfonnx/lib/python3.7/site-packages/tf2onnx-1.6.2-py3.7.egg/tf2onnx/convert.py", line 153, in main
File "/fomshare/home/robers/anaconda3/envs/tfonnx/lib/python3.7/site-packages/tf2onnx-1.6.2-py3.7.egg/tf2onnx/tfonnx.py", line 487, in process_tf_graph
File "/fomshare/home/robers/anaconda3/envs/tfonnx/lib/python3.7/site-packages/tf2onnx-1.6.2-py3.7.egg/tf2onnx/tfonnx.py", line 292, in topological_sort
File "/fomshare/home/robers/anaconda3/envs/tfonnx/lib/python3.7/site-packages/tf2onnx-1.6.2-py3.7.egg/tf2onnx/graph.py", line 922, in topological_sort
File "/fomshare/home/robers/anaconda3/envs/tfonnx/lib/python3.7/site-packages/tf2onnx-1.6.2-py3.7.egg/tf2onnx/utils.py", line 193, in make_sure
ValueError: make_sure failure: Cannot find node with output batch_join/fifo_queue:0
Expected behavior
The error reporting is fine, but it would be nice to have the converter finish successfully.
Additional context
Note that I could not pull this up in Tensorboard, but through code review of https://github.com/davidsandberg/facenet I found the inputs and outputs.