I'm writing a custom Tensorflow op using the tutorial and I'm having trouble understanding how to read and write to/from Tensors.
let's say I have a Tensor in my OpKernel that I get from
const Tensor& values_tensor = context->input(0); (where context = OpKernelConstruction*)
if that Tensor has shape, say, [2, 10, 20], how can I index into it (e.g. auto x = values_tensor[1, 4, 12], etc.)?
equivalently, if I have
Tensor *output_tensor = NULL;
OP_REQUIRES_OK(context, context->allocate_output(
0,
{batch_size, value_len - window_size, window_size},
&output_tensor
));
how can I assign to output_tensor, like output_tensor[1, 2, 3] = 11, etc.?
sorry for the dumb question, but the docs are really tripping me up here and the examples in the Tensorflow kernel code for built-in ops somehow obfuscate this to the point that I get very confused :)
thank you!
The easiest way to read from and write to tensorflow::Tensor objects is to convert them to an Eigen tensor, using the tensorflow::Tensor::tensor<T, NDIMS>() method. Note that you have to specify the (C++) type of elements in tensor as template parameter T.
For example, to read a particular value from a DT_FLOAT32 tensor:
const Tensor& values_tensor = context->input(0);
auto x = value_tensor.tensor<float, 3>()(1, 4, 12);
To write a particular value to a DT_FLOAT32 tensor:
Tensor* output_tensor = ...;
output_tensor->tensor<float, 3>()(1, 2, 3) = 11.0;
There are also convenience methods for accessing a scalar, vector, or matrix.
Related
TF version: 2.11
I try to train a simple 2input classifier with TFRecords tf.data pipeline
I do not manage to convert the tf.dense Tensor with containing only a scalar to a tf.onehot vector
# get all recorddatasets abspath
training_names= [record_path+'/'+rec for rec in os.listdir(record_path) if rec.startswith('train')]
# load in tf dataset
train_dataset = tf.data.TFRecordDataset(training_names[1])
train_dataset = train_dataset.map(return_xy)
mapping function:
def return_xy(example_proto):
#parse example
sample= parse_function(example_proto)
#decode image 1
encoded_image1 = sample['image/encoded_1']
decoded_image1 = decode_image(encoded_image1)
#decode image 2
encoded_image2 = sample['image/encoded_2']
decoded_image2 = decode_image(encoded_image2)
#decode label
print(f'image/object/class/'+level: {sample['image/object/class/'+level]}')
class_label = tf.sparse.to_dense(sample['image/object/class/'+level])
print(f'type of class label :{type(class_label)}')
print(class_label)
# conversion to onehot with depth 26 :: -> how can i extract only the value or convert directly to tf.onehot??
label_onehot=tf.one_hot(class_label,26)
#resizing image
input_left=tf.image.resize(decoded_image1,[416, 416])
input_right=tf.image.resize(decoded_image2,[416, 416])
return {'input_3res1':input_left, 'input_5res2':input_right} , label_onehot
output:
image/object/class/'+level: SparseTensor(indices=Tensor("ParseSingleExample/ParseExample/ParseExampleV2:14", shape=(None, 1), dtype=int64), values=Tensor("ParseSingleExample/ParseExample/ParseExampleV2:31", shape=(None,), dtype=int64), dense_shape=Tensor("ParseSingleExample/ParseExample/ParseExampleV2:48", shape=(1,), dtype=int64))
type of class label :<class 'tensorflow.python.framework.ops.Tensor'>
Tensor("SparseToDense:0", shape=(None,), dtype=int64)
However I am sure that the label is in this Tensor because when run it eagerly
raw_dataset = tf.data.TFRecordDataset([rec_file])
parsed_dataset = raw_dataset.map(parse_function) # only parsing
for sample in parsed_dataset:
class_label=tf.sparse.to_dense(sample['image/object/class/label_level3'])[0]
print(f'type of class label :{type(class_label)}')
print(f'labels from labelmap :{class_label}')
I get output:
type of class label :<class 'tensorflow.python.framework.ops.EagerTensor'>
labels from labelmap :7
If I just chose a random number for the label and pass it to tf_one_hot(randint, 26) then the model begins to train (obviously nonsensical).
So the question is how can i convert the:
Tensor("SparseToDense:0", shape=(None,), dtype=int64)
to a
Tensor("one_hot:0", shape=(26,), dtype=float32)
What I tried so far
in the call data.map(parse_xy)
i tried to just call .numpy() on the tf tensors but didnt work , this only works for eager tensors.
In my understanding i cannot use eager execution because everthing in the parse_xy function gets excecuted on the whole graph:
ive already tried to enable eager execution -> failed
https://www.tensorflow.org/api_docs/python/tf/config/run_functions_eagerly
Note: This flag has no effect on functions passed into tf.data transformations as arguments.
tf.data functions are never executed eagerly and are always executed as a compiled Tensorflow Graph.
ive also tried to use the tf_pyfunc but this only returns another tf.Tensor with an unknown shape
def get_onehot(tensor):
class_label=tensor[0]
return tf.one_hot(class_label,26)
and add the line in parse_xy:
label_onehot=tf.py_function(func=get_onehot, inp=[class_label], Tout=tf.int64)
but there i always get an unknown shape which a cannot just alter with .set_shape()
I was able to solve the issue by only using TensorFlow functions.
tf.gather allows to index a TensorFlow tensor:
class_label_gather = tf.sparse.to_dense(sample['image/object/class/'+level])
class_indices = tf.gather(tf.cast(class_label_gather,dtype=tf.int32),0)
label_onehot=tf.one_hot(class_indices,26)
simple question and im sure answer is straightforward but im really struggling to match model shape with tensor fitting into model.
this simple code
let tf = require('#tensorflow/tfjs-node');
let features = {
x: [1,2,3,4,5,6,7,8,9],
y: [1,2,3,4,5,6,7,8,9]
}
let tensorfeature = tf.tensor2d(Object.values(features))
console.log(tensorfeature.shape)
const model = tf.sequential();
model.add(tf.layers.dense(
{
inputShape: tensorfeature.shape,
units: 1
}
))
const optimizer = tf.train.sgd(0.005);
model.compile({optimizer: optimizer, loss: 'meanAbsoluteError'});
model.fit(tensorfeature,
{epochs: 5}
)
Results in Error: Error when checking input: expected dense_Dense1_input to have 3 dimension(s). but got array with shape 2,9
tried multiple things with reshape, slice, etc with no luck. Can someone point me what exactly is wrong?
model.fit takes at least two parameters x, y which are either tensors or array of tensors. The config object is the third parameter.
Also, the feature(tensorfeature) tensor passed as argument to model.fit should be one dimension higher than the inputShape of the model. Since tensorfeature.shape is used as the inputShape, if we want to traing the model with tensorfeature its dimension should be expanded. It can be done using reshape or expandDims.
model.fit(tensorfeature.expandDims(0))
// or possibly
model.fit(tensorfeature.reshape([1, ...tensorfeature.shape])
This shape mismatch between the model and the training data has been discussed here and there
For faster inference one model, I want to merge 'Conv-BN-Scale' into a single 'Conv' layer for my tensorflow model, but I can not find some useful complete example about how to do it?
Anyone can give some advises or complete code example?
Thanks!
To merge two layers, you will need to pass a Tensor and get a tensor back that is after both the layers are applied, suppose your input tensor is X.
def MlConvBnScale(X ,kernel,strides , padding = 'SAME' , scale = False, beta_initializer = 0.1, gamma_initializer = 0.1, moving_mean_initializer = 0.1, moving_variance_initializer = 0.1):
convLout = tf.nn.conv2d(X,
filter = Kernel,
strides = strides,
padding = padding)
return tf.nn.batch_normalization(convLout,
scale = scale,
beta_initializer = beta_initializer,
gamma_initializer = gamma_initializer,
moving_mean_initializer = moving_mean_intializer,
moving_variance_initializer = moving_variance_initializer )
And that will return a tensor after performing both the operations, I have taken default values of variables but you can modify them in your function call, and in case your input is not already a tensor but a numpy array you can use tf.convert_to_tensor() from this link https://www.tensorflow.org/api_docs/python/tf/convert_to_tensor, and in case you are struggling with kernel/filter and its application, check out this thread. What does tf.nn.conv2d do in tensorflow?
If you have any queries or run into trouble implementing it, comment down below and we will see.
How do I mutate value of a tensor in Tensorflow.js? For example if I have a tensor created like this:
const a = tf.tensor1d([1,2,3,4])
How do I change the value of the third element of the tensor? I know that tensors are immutable and variables are mutable.
Doing this: const a = tf.variable(tf.tensor1d([1,2,3,4])) doesn't seem to solve the problem. I cannot do:
const a = a[0].assign(5)
I am able to do this in python tensorflow like this:
a = tf.Variable([1,2,3,4])
a = a[0].assign(100)
with tf.Session() as sess:
sess.run(tf.global_variables_iniliazer())
print sess.run(a)
This outputs [100, 2,3,4]
Does tf.buffer work for you?
// Create a buffer and set values at particular indices.
const a = tf.tensor1d([1, 2, 3, 4]);
const buffer = tf.buffer(a.shape, a.dtype, a.dataSync());
buffer.set(5, 0);
const b = buffer.toTensor();
// Convert the buffer back to a tensor.
b.print();
Tensor
[5, 2, 3, 4]
I had to do this using mulStrict and addStrict which do element-wise multiplication and addition.
const a = tf.tensor1d([1,2,3,4]);
tf.mulStrict(a, tf.tensor1d([0,1,1,1]))
.addStrict(tf.tensor1d([100, 0, 0, 0]);
This was based on asnwer
here
I’m running into a number of issues relating to dynamic axes. I am trying to implement a convolutional rnn similar to the of the LSTM() function but handles sequential image input and outputs an image.
I’m able to build the network and pass dummy data through it to produce output, but when I try to compute the error with an input_variable label I consistently see the following error:
RuntimeError: Node '__v2libuid__Input471__v2libname__img_label' (InputValue operation): DataFor: FrameRange's dynamic axis is inconsistent with matrix: {numTimeSteps:1, numParallelSequences:2, sequences:[{seqId:0, s:0, begin:0, end:1}, {seqId:1, s:1, begin:0, end:1}]} vs. {numTimeSteps:2, numParallelSequences:1, sequences:[{seqId:0, s:0, begin:0, end:2}]}`
If I understand this error message correctly, it claims that the value I passed in as the label has inconsistent axes to what is expected with 2 time steps and 1 parallel sequence, when what is desired is 1 time-step and 2 sequences. This makes sense to me, but I’m not sure how the data I’m passing in is not conforming to this. Here are (roughly) the variable declarations and eval statements:
…
img_input = input_variable(shape=img_shape, dtype=np.float32, name="img_input")
convlstm = Recurrence(conv_lstm_cell, initial_state=initial_state)(img_input)
out = select_last(convlstm)
img_label = input_variable(shape=img_shape, dynamic_axes=out.dynamic_axes, dtype=np.float32, name="img_label”)
error = squared_error(out, img_label)
…
dummy_input = np.ones(shape=(2, 3, 3, 32, 32)) # (batch, seq_len, channels, height, width)
dummy_label = np.ones(shape=(2, 3, 32, 32)) # (batch, channels, height, width)
out = error.eval({img_input:dummy_input, img_label:dummy_label})
I believe part of the issue is with the dynamic_axes set when creating the img_label input_variable, I’ve also tried setting it to [Axis.default_batch_axis()] and not setting it at all and either squared error complains about inconsistent axes between out and img_label or I see the same error as above.
The only issue I see with the above setup is that your dummy label should have an explicit dynamic axis so it should be declared as
dummy_label = np.ones(shape=(2, 1, 3, 32, 32))
Assuming your convlstm works similar to an lstm, then the following works without issues for me and it evaluates the loss for two input/output pairs.
x = C.input_variable((3,32,32))
cx = convlstm(x)
lx = C.sequence.last(cx)
y = C.input_variable(lx.shape, dynamic_axes=lx.dynamic_axes)
loss = C.squared_error(y, lx)
x0 = np.arange(2*3*3*32*32,dtype=np.float32).reshape(2,3,3,32,32)
y0 = np.arange(2*1*3*32*32,dtype=np.float32).reshape(2,1,3,32,32)
loss.eval({x:x0, y:y0})