segmentation fault during training 3D match

[tangbohu]

hello! Thanks for your sharing.

The code seems raise an error during runing the following code when I attempt to train on 3D Match. train_data = train_data.map(map_func=map_func, num_parallel_calls=self.num_threads)

Looking forward to your kind reply!

[XuyangBai]

Hi, @tangbohu Could you paste the error information here?

[tangbohu]

Hi, @tangbohu Could you paste the error information here?

Hi, @XuyangBai Thanks for your kind reply!

The error information is as follow:

Dataset Preparation

---

Preparing ply files Num_train 3680

Preparing ply files Num_val 1238 Initiating input pipelines self.batch_limit: 32235.291046036124 WARNING: Logging before flag parsing goes to stderr. W0419 10:41:07.728522 139672721119040 deprecation.py:323] From /home/user/.local/lib/python3.6/site-packages/tensorflow/python/data/ops/dataset_ops.py:494: py_func (from tensorflow.python.ops.script_ops) is deprecated and will be removed in a future version. Instructions for updating: tf.py_func is deprecated in TF V2. Instead, there are two options available in V2.

• tf.py_function takes a python function which manipulates tf eager tensors instead of numpy arrays. It's easy to convert a tf eager tensor to an ndarray (just call tensor.numpy()) but having access to eager tensors means tf.py_functions can use accelerators such as GPUs as well as being differentiable using a gradient tape.
• tf.numpy_function maintains the semantics of the deprecated tf.py_func (it is not differentiable, and manipulates numpy arrays). It drops the stateful argument making all functions stateful.

W0419 10:41:07.835422 139672721119040 deprecation.py:506] From /media/user/400a44dc-53f0-42c3-b7af-9a9931e73ec0/OriginWorkSpace/D3Feat/datasets/common.py:1308: calling reduce_min_v1 (from tensorflow.python.ops.math_ops) with keep_dims is deprecated and will be removed in a future version. Instructions for updating: keep_dims is deprecated, use keepdims instead Segmentation fault (core dumped)

And I have debugged the code, and found that the code dumped at line 600 of the file dataset/common.py, namely the code "train_data = train_data.map(map_func=map_func, num_parallel_calls=self.num_threads)".

Thanks!

[XuyangBai]

Hi @tangbohu I think you are using tensorflow2.0 or newer version? But there might be segmentation fault when using cpp custom op in tf.data.Dataset.map in tensorflow2.0, see this issue. Maybe you can try tensorflow1.12.0 instead or try g++-4.8 to build the custom ops (I am not sure this can work or not). And my code is heavily built on KPConv, and the author of KPConv might not test their code on tf 2.0.

[tangbohu]

Hi @tangbohu I think you are using tensorflow2.0 or newer version? But there might be segmentation fault when using cpp custom op in tf.data.Dataset.map in tensorflow2.0, see this issue. Maybe you can try tensorflow1.12.0 instead or try g++-4.8 to build the custom ops (I am not sure this can work or not). And my code is heavily built on KPConv, and the author of KPConv might not test their code on tf 2.0.

Thanks! It is caused by my upgrading of tensorflow from 1.12 to 1.14. The problem has been solved.

[zhouhao957]

@tangbohu hello I also have an error in this code train_data = train_data.map(map_func=map_func, num_parallel_calls=self.num_threads)

error: Process finished with exit code 139 (interrupted by signal 11: SIGSEGV) Are you like that? How did you solve it