Category: Misc

Hello!

I am doing image classification with TensorFlow for learning purposes. I am splitting the data into 5 folds. I would like to get useful summary statistics on these validation sets. What could be useful other than the shape of the validation sets?

Hello,

I understand that Tensorflow is geared towards proprietary NVIDIA Cuda, but is there a workaround for AMD Radeon GPU? I’m on a Macbook Pro with an AMD Radeon 580 external GPU card.

I am trying to go through the DCGAN example on the tensorflow website https://www.tensorflow.org/tutorials/generative/dcgan. it seems to run fine up until the step where it uses the generator generated_image = generator(noise, training=False). At that point it exits with error code Process finished with exit code -1073740791 (0xC0000409).

I am running on Windows 10 using pycharm. I have tried messing with the batch size in case this is a memory issue, but even setting it to 1 gives the same results. I have also tried running pycharm as administrator.

I am running Tensorflow Lite on my Raspberry Pi 3b+ with a custom object detection mode. I have tested it on a Google COCO dataset and it works wonderfully but when I test it on my custom trained model it does not work despite the model passing TfLite Model Maker evaluation. When I run it the only error I get in my message is “Segmentation fault”. How can I fix this?

I am not able to upload my model to Stackoverflow but just some info about it. It is only detecting one object, It is Not quantized, it is trained based off the efficientdet_lite1 mode, and I trained it using the official Tensorflow Lite Model Maker Google Colab.

Here is the code used to interpret the model on my Pi.

https://pastebin.com/1at3ZAJd

I added a few print statements aswell to troubleshoot and it stops executing at around line 115.

Does anyone know how to fix this?

With the robotics set up we use opencv for the images. However in the cnn I use tf.io.decode_jpg to open the images. These two methods slighty alter the image and makes it such that the same image, but opened in the different methods can’t be classified on the cnn.

I found the differences in this blog: https://towardsdatascience.com/image-read-and-resize-with-opencv-tensorflow-and-pil-3e0f29b992be

which states that two things needs to be changed to ensure that the two files are the same

1. dct_method=’INTEGER_ACCURATE’ needs to be added to the decode
2. half_pixel_centers=True to the resize method and also force it to be bilinear.

However the half_pixel_centers keyword is not found

https://stackoverflow.com/questions/50591669/tf-image-resize-bilinear-vs-cv2-resize

This stackoverflow states that it is added in tf2.0 with a link to their github showing it has indeed been added: https://github.com/tensorflow/tensorflow/commit/3ae2c6691b7c6e0986d97b150c9283e5cc52c15f

About my code I map the dataset to a function that reads the file path

img = tf.io.read_file(file_path)

img = tf.io.decode_jpeg(img, channels=3, dct_method=’INTEGER_ACCURATE’)

resized img = tf.image.resize(img, (28,28), method=tf.image.ResizeMethod.BILINEAR, preserve_aspect_ratio=False, antialias=False, name=None, half_pixel_center=True)

I also tried it on an other machine with tf2.7 and it gives the same error.Could someone point out what I am doing wrong or perhaps there is a better way in general?

I’ve followed the same steps provided by Apple (https://developer.apple.com/metal/tensorflow-plugin/) to install the deep learning development environment on TensorFlow (based on tensorflow-metal plugin) on my MacBook Pro. My model employs the VGG19 through transfer-learning as its summary can be seen below. While I train this model on 1,859 75×75 RGB images, getting the error tensorflow.python.framework.errors_impl.InternalError: Failed copying input tensor from /job:localhost/replica:0/task:0/device:CPU:0 to /job:localhost/replica:0/task:0/device:GPU:0 in order to run _EagerConst: Dst tensor is not initialized. Isn’t this task an easy one for such a powerful SoC like M1 Pro 10-core CPU, 16-core GPU 16-core Neural Engine with 16 GB RAM? What is the issue here? Is this a bug or do I need to do some configuration to overcome this situation?

​

Here is the stack trace:

Metal device set to: Apple M1 Pro

​

systemMemory: 16.00 GB

maxCacheSize: 5.33 GB

​

Model: “sequential”

_________________________________________________________________

Layer (type) Output Shape Param #

=================================================================

vgg19 (Functional) (None, 512) 20024384

_________________________________________________________________

flatten (Flatten) (None, 512) 0

_________________________________________________________________

dense (Dense) (None, 1859) 953667

=================================================================

Total params: 20,978,051

Trainable params: 953,667

Non-trainable params: 20,024,384

_________________________________________________________________

Traceback (most recent call last):

File “/Users/talhakabakus/PycharmProjects/keras-matlab-comp-metal/run.py”, line 528, in <module>

run_all_stanford_dogs()

File “/Users/talhakabakus/PycharmProjects/keras-matlab-comp-metal/run.py”, line 481, in run_all_stanford_dogs

H = model.fit(X_train, y_train_cat, validation_split=n_val_split, epochs=n_epochs,

File “/Users/talhakabakus/miniforge3/envs/keras-matlab-comp-metal/lib/python3.9/site-packages/keras/engine/training.py”, line 1134, in fit

data_handler = data_adapter.get_data_handler(

File “/Users/talhakabakus/miniforge3/envs/keras-matlab-comp-metal/lib/python3.9/site-packages/keras/engine/data_adapter.py”, line 1383, in get_data_handler

return DataHandler(*args, **kwargs)

File “/Users/talhakabakus/miniforge3/envs/keras-matlab-comp-metal/lib/python3.9/site-packages/keras/engine/data_adapter.py”, line 1138, in __init__

self._adapter = adapter_cls(

File “/Users/talhakabakus/miniforge3/envs/keras-matlab-comp-metal/lib/python3.9/site-packages/keras/engine/data_adapter.py”, line 230, in __init__

x, y, sample_weights = _process_tensorlike((x, y, sample_weights))

File “/Users/talhakabakus/miniforge3/envs/keras-matlab-comp-metal/lib/python3.9/site-packages/keras/engine/data_adapter.py”, line 1031, in _process_tensorlike

inputs = tf.nest.map_structure(_convert_numpy_and_scipy, inputs)

File “/Users/talhakabakus/miniforge3/envs/keras-matlab-comp-metal/lib/python3.9/site-packages/tensorflow/python/util/nest.py”, line 869, in map_structure

structure[0], [func(*x) for x in entries],

File “/Users/talhakabakus/miniforge3/envs/keras-matlab-comp-metal/lib/python3.9/site-packages/tensorflow/python/util/nest.py”, line 869, in <listcomp>

structure[0], [func(*x) for x in entries],

File “/Users/talhakabakus/miniforge3/envs/keras-matlab-comp-metal/lib/python3.9/site-packages/keras/engine/data_adapter.py”, line 1026, in _convert_numpy_and_scipy

return tf.convert_to_tensor(x, dtype=dtype)

File “/Users/talhakabakus/miniforge3/envs/keras-matlab-comp-metal/lib/python3.9/site-packages/tensorflow/python/util/dispatch.py”, line 206, in wrapper

return target(*args, **kwargs)

File “/Users/talhakabakus/miniforge3/envs/keras-matlab-comp-metal/lib/python3.9/site-packages/tensorflow/python/framework/ops.py”, line 1430, in convert_to_tensor_v2_with_dispatch

return convert_to_tensor_v2(

File “/Users/talhakabakus/miniforge3/envs/keras-matlab-comp-metal/lib/python3.9/site-packages/tensorflow/python/framework/ops.py”, line 1436, in convert_to_tensor_v2

return convert_to_tensor(

File “/Users/talhakabakus/miniforge3/envs/keras-matlab-comp-metal/lib/python3.9/site-packages/tensorflow/python/profiler/trace.py”, line 163, in wrapped

return func(*args, **kwargs)

File “/Users/talhakabakus/miniforge3/envs/keras-matlab-comp-metal/lib/python3.9/site-packages/tensorflow/python/framework/ops.py”, line 1566, in convert_to_tensor

ret = conversion_func(value, dtype=dtype, name=name, as_ref=as_ref)

File “/Users/talhakabakus/miniforge3/envs/keras-matlab-comp-metal/lib/python3.9/site-packages/tensorflow/python/framework/tensor_conversion_registry.py”, line 52, in _default_conversion_function

return constant_op.constant(value, dtype, name=name)

File “/Users/talhakabakus/miniforge3/envs/keras-matlab-comp-metal/lib/python3.9/site-packages/tensorflow/python/framework/constant_op.py”, line 271, in constant

return _constant_impl(value, dtype, shape, name, verify_shape=False,

File “/Users/talhakabakus/miniforge3/envs/keras-matlab-comp-metal/lib/python3.9/site-packages/tensorflow/python/framework/constant_op.py”, line 283, in _constant_impl

return _constant_eager_impl(ctx, value, dtype, shape, verify_shape)

File “/Users/talhakabakus/miniforge3/envs/keras-matlab-comp-metal/lib/python3.9/site-packages/tensorflow/python/framework/constant_op.py”, line 308, in _constant_eager_impl

t = convert_to_eager_tensor(value, ctx, dtype)

File “/Users/talhakabakus/miniforge3/envs/keras-matlab-comp-metal/lib/python3.9/site-packages/tensorflow/python/framework/constant_op.py”, line 106, in convert_to_eager_tensor

return ops.EagerTensor(value, ctx.device_name, dtype)

tensorflow.python.framework.errors_impl.InternalError: Failed copying input tensor from /job:localhost/replica:0/task:0/device:CPU:0 to /job:localhost/replica:0/task:0/device:GPU:0 in order to run _EagerConst: Dst tensor is not initialized.

​

Process finished with exit code 1