I have a tensor, like:
input:
a = torch.rand(2,3,5)
output:
tensor([[[0.2764, 0.2209, 0.8007, 0.1246, 0.4302],
[0.9716, 0.8063, 0.3904, 0.7574, 0.2392],
[0.3366, 0.4209, 0.0527, 0.1328, 0.0441]],
[[0.8166, 0.6519, 0.5450, 0.3072, 0.2716],
[0.0583, 0.0613, 0.8984, 0.0110, 0.4744],
[0.2269, 0.2693, 0.6447, 0.6078, 0.6148]]])
how can I get :
tensor([[[0.2764, 0.2209, 0.8007, 0.1246, 0.4302],
[0.9716, 0.8063, 0.3904, 0.7574, 0.2392]],
[[0.8166, 0.6519, 0.5450, 0.3072, 0.2716],
[0.2269, 0.2693, 0.6447, 0.6078, 0.6148]]])
I have tried gather(), but it does not work.
try this .This should work
a[:,:-1, :]
Related
I am trying to use the tf.map_fn() , where my elems should be pointing to the channel dimension of my inputs(shape = 100,24,24,6), so my elems should be a list/tuple of tensors, pointing or accessing the values of the channel dimension(6) of the inputs .I am trying to do it by making a for loop in such a way :
#tf.function
def call(self, inputs, training=True):
elems = []
for b in inputs:
for h in b:
for w in h:
for c in w:
elems.append(c)
changed_inputs = tf.map_fn(self.do_mapping, elems)
return changed_inputs
What i am trying to achieve in the self.do_mapping is that it is doing a dictionary look up for the values of a dictionary (vmap) using the keys and the return the values. the dictionary vmap is made by accessing the output of a layer and appending only the similar values of the channel dimension of the output of layer so the keys in dictionary are tuple of 6 (as the size of channel dimension) tf.tensorobjects and values of dictionary is the count which i keep. This is how the dictionary is made :
value = list(self.get_values())
vmap = {}
cnt = 0
for v0 in value:
for v1 in v0:
for v2 in v1:
for v3 in v2:
v = tuple(v3)
if v not in vmap:
vmap[v]=cnt
cnt+=1
the do_mapping function is :
#tf.function
def do_mapping(self,pixel):
if self._compression :
pixel = tuple(pixel)
enumerated_value=self._vmap.get(pixel)
print(enumerated_value)
print(tf.shape(pixel))
exit()
return enumerated_value
If i try to use the tf.map_fn now where i try to point the elems to the channel dimension then i get the following error :(ValueError: elements in elems must be 1+ dimensional Tensors, not scalars ). Please help me to understand how can i use the tf.map_fn for my case ? Thank you in advance
First, instead of doing a for loop (try to avoid for efficiency), you can just reshape that way:
elems = tf.reshape(inputs,-1)
Second, what do you want to do exactly? What do you mean by "it doesn't work"? What is the error message? What is self.do_mapping?
Best,
Keivan
I've been trying to convert a generator I built to a tf.data.dataset.
I've come far and now I have something simple like this
def parse_image(filename):
file = tf.io.read_file(filename) # this will work only with filename as tensor
image = tf.image.decode_image(file)
return image
def transform_img(img):
img = parse_image(img).numpy()
img = transforms_train(image = img)["image"]
return img
transform img works as expected when I call it on a filename itself. like:
plt.imshow(transform_img(array_of_filenames[0]))
but when I map it on a dataset
dataset = tf.data.Dataset.from_tensor_slices(array_of_filenames)
dataset = dataset.map(transform_img)
I get the error in the title.
I am doing something silly again aren't I?
Thanks for helping!
It is not possible to use numpy inside the map function of tensorflow dataset. Otherwise, you need to wrap the function in tf.py_function or tf.numpy_function. So it should look like the following:
dataset = dataset.map(lambda: item: tf.py_function(transform_img, [item], [tf.float32]))
The first argument of py_function is the preprocessing function you want, the second argument is the parameter to pass to the function. The final argument is the dtype of the return of preprocess function. (same applies to tf.numpy_function)
I don't remember reading this in documentation but in a tutorial, you can find it here.
How do I get the value out of a tensor in Tensorflow.js after specifying the index?
You can use datasync for this.
const newTensor = tf.tensor2d([[2,4],[5,6]]);
const tensorData = newTensor.dataSync();
console.log("data[0] is " + tensorData[0]);
console.log("data[3] is " + tensorData[3]);
https://codepen.io/anon/pen/NMKgeO?editors=1011
const newTensor = tf.tensor2d([[2,4], [5,6]])
newTensor.get([0]) ##returns 2
newTensor.get([3]) ##returns 6
Thankfully, all this returns a number and not a tensor.
You can use the following more powerful method
tensor.buffer().get(0, 0);
This will allow you index into logical coordinates of the tensor (the 2d coordinates as opposed to the flattened 1d coordinate). See the link
tf.Tensor.dataSync() does not retain the original shape. If you would like to preserve the shape, you can use tf.Tensor.arraySync().
Alternatively you can also use slice to get the value:
let value = tensor.slice([i,j], [1, 1]);
or
let value = tensor.slice([i,j], [1, 1]).arraySync()[0][0];
I currently follow the tutorial to retrain Inception for image classification:
https://cloud.google.com/blog/big-data/2016/12/how-to-train-and-classify-images-using-google-cloud-machine-learning-and-cloud-dataflow
However, when I make a prediction with the API I get only the index of my class as a label. However I would like that the API actually gives me a string back with the actual class name e.g instead of
predictions:
- key: '0'
prediction: 4
scores:
- 8.11998e-09
- 2.64907e-08
- 1.10307e-06
I would like to get:
predictions:
- key: '0'
prediction: ROSES
scores:
- 8.11998e-09
- 2.64907e-08
- 1.10307e-06
Looking at the reference for the Google API it should be possible:
https://cloud.google.com/ml-engine/reference/rest/v1/projects/predict
I already tried to change in the model.py the following to
outputs = {
'key': keys.name,
'prediction': tensors.predictions[0].name,
'scores': tensors.predictions[1].name
}
tf.add_to_collection('outputs', json.dumps(outputs))
to
if tensors.predictions[0].name == 0:
pred_name ='roses'
elif tensors.predictions[0].name == 1:
pred_name ='tulips'
outputs = {
'key': keys.name,
'prediction': pred_name,
'scores': tensors.predictions[1].name
}
tf.add_to_collection('outputs', json.dumps(outputs))
but this doesn't work.
My next idea was to change this part in the preprocess.py file. So instead getting the index I want to use the string label.
def process(self, row, all_labels):
try:
row = row.element
except AttributeError:
pass
if not self.label_to_id_map:
for i, label in enumerate(all_labels):
label = label.strip()
if label:
self.label_to_id_map[label] = label #i
and
label_ids = []
for label in row[1:]:
try:
label_ids.append(label.strip())
#label_ids.append(self.label_to_id_map[label.strip()])
except KeyError:
unknown_label.inc()
but this gives the error:
TypeError: 'roses' has type <type 'str'>, but expected one of: (<type 'int'>, <type 'long'>) [while running 'Embed and make TFExample']
hence I thought that I should change something here in preprocess.py, in order to allow strings:
example = tf.train.Example(features=tf.train.Features(feature={
'image_uri': _bytes_feature([uri]),
'embedding': _float_feature(embedding.ravel().tolist()),
}))
if label_ids:
label_ids.sort()
example.features.feature['label'].int64_list.value.extend(label_ids)
But I don't know how to change it appropriately as I could not find someting like str_list. Could anyone please help me out here?
Online prediction certainly allows this, the model itself needs to be updated to do the conversion from int to string.
Keep in mind that the Python code is just building a graph which describes what computation to do in your model -- you're not sending the Python code to online prediction, you're sending the graph you build.
That distinction is important because the changes you have made are in Python -- you don't yet have any inputs or predictions, so you won't be able to inspect their values. What you need to do instead is add the equivalent lookups to the graph that you're exporting.
You could modify the code like so:
labels = tf.constant(['cars', 'trucks', 'suvs'])
predicted_indices = tf.argmax(softmax, 1)
prediction = tf.gather(labels, predicted_indices)
And leave the inputs/outputs untouched from the original code
I want to get the extension of image files to invoke different image decoder, and I found there's a function called tf.string_split in tensorflow r0.11.
filename_queue = tf.train.string_input_producer(filenames, shuffle=shuffle)
reader = tf.WholeFileReader()
img_src, img_bytes = reader.read(filename_queue)
split_result = tf.string_split(img_src, '.')
But when I run it, I get this error：
ValueError: Shape must be rank 1 but is rank 0 for 'StringSplit' (op: 'StringSplit') with input shapes: [], [].
I think it may caused by the shape inference of img_src. I try to use img_src.set_shape([1,]) to fix it, but it seems not work, I get this error:
ValueError: Shapes () and (1,) are not compatible
Also, I can't get the shape of img_src using
tf.Print(split_result, [tf.shape(img_src)],'img_src shape=')
The result is img_src shape=[]. But if I use the following code:
tf.Print(split_result, [img_src],'img_src=')
The result is img_src=test_img/test1.png. Am I doing something wrong?
Just pack img_src into a tensor.
split_result = tf.string_split([img_src], '.')