I want to know how can I combine two layers with different spatial space in Tensorflow.
for example::
batch_size = 3
input1 = tf.ones([batch_size, 32, 32, 3], tf.float32)
input2 = tf.ones([batch_size, 16, 16, 3], tf.float32)
filt1 = tf.constant(0.1, shape = [3,3,3,64])
filt1_1 = tf.constant(0.1, shape = [1,1,64,64])
filt2 = tf.constant(0.1, shape = [3,3,3,128])
filt2_2 = tf.constant(0.1, shape = [1,1,128,128])
#first layer
conv1 = tf.nn.conv2d(input1, filt1, [1,2,2,1], "SAME")
pool1 = tf.nn.max_pool(conv1, [1,2,2,1],[1,2,2,1], "SAME")
conv1_1 = tf.nn.conv2d(pool1, filt1_1, [1,2,2,1], "SAME")
deconv1 = tf.nn.conv2d_transpose(conv1_1, filt1_1, pool1.get_shape().as_list(), [1,2,2,1], "SAME")
#seconda Layer
conv2 = tf.nn.conv2d(input2, filt2, [1,2,2,1], "SAME")
pool2 = tf.nn.max_pool(conv2, [1,2,2,1],[1,2,2,1], "SAME")
conv2_2 = tf.nn.conv2d(pool2, filt2_2, [1,2,2,1], "SAME")
deconv2 = tf.nn.conv2d_transpose(conv2_2, filt2_2, pool2.get_shape().as_list(), [1,2,2,1], "SAME")
The deconv1 shape is [3, 8, 8, 64] and the deconv2 shape is [3, 4, 4, 128]. Here I cannot use the tf.concat to combine the deconv1 and deconv2. So how can I do this???
Edit
This is image for the architecture that I tried to implement:: it is releated to this paper::
vii. He, W., Zhang, X. Y., Yin, F., & Liu, C. L. (2017). Deep Direct
Regression for Multi-Oriented Scene Text Detection. arXiv preprint
arXiv:1703.08289
I checked the paper you point and there is it, consider the input image to this network has size H x W (height and width), I write the size of the output image on the side of each layer. Now look at the most bottom layer which I circle the input arrows to that layer, let's check it. This layer has two input, the first from the previous layer which has shape H/2 x W/2 and the second from the first pooling layer which also has size H/2 x W/2. These two inputs are merged together (not concatenation, but added together based on paper) and goes into the last Upsample layer, which output image of size H x W.
The other Upsample layers also have the same inputs. As you can see all merging operations have the match shapes. Also, the filter number for all merging layers is 128 which has consistency with others.
You can also use concat instead of merging, but it results in a larger filter number, be careful about that. i.e. merging two matrices with shapes H/2 x W/2 x 128 results in the same shape H/2 x W/2 x 128, but concat two matrices on the last axis, with shapes H/2 x W/2 x 128 results in H/2 x W/2 x 256.
I tried to guide you as much as possible, hope that was useful.
Related
I am trying to implement a silly learn to rank example. Essentially, I have 2 descriptions of a location, size and number of bathrooms. I want to "combine" them to create a score. Then I wish to compare the scores for the "best". I will always be comparing 3 locations at a time.
The neuralnetwork I expect to do this:
# 3 locations with 2 descriptions.
rinputs = Input(shape=(3, 2), name ='inputlayer')
# take my 3 expected inputs, split them
split = Lambda( lambda x: tf.split(x,num_or_size_splits=3,axis=1))(rinputs)
input_one_tensor = split[0]
input_two_tensor = split[1]
input_three_tensor = split[2]
# combine each set of location elements into 1 "score"
layer2 = Dense(1, name = 'Layer2', use_bias = True, activation = 'sigmoid') # 60 was better than 100
layer2a = layer2(input_one_tensor)
layer2b = layer2(input_two_tensor)
layer2c = layer2(input_three_tensor)
concatLayer = Concatenate(name = 'ConcatLayer2')([layer2a,layer2b, layer2c])
# softmax my score to get "best selection"
softmaxLayer = Dense(3, activation='softmax', name = 'softmax', use_bias = False)
softmaxLayer = softmaxLayer(concatLayer)
model = Model(inputs=rinputs, outputs=softmaxLayer)
model.compile(loss=keras.losses.categorical_crossentropy, optimizer=keras.optimizers.Adam(),metrics=['accuracy'])
I now create my test data:
loc1 = [1, 5]
loc2 = [4, 1]
loc3 = [6, 7]
# create two entries for my trial run
inputs = np.asarray([[loc1, loc2, loc3], [loc3,loc3,loc1]]).reshape(2,3,2)
ytrue = np.asarray([[1, 0, 0], [0, 0, 1]]).reshape(2,3)
model.fit(inputs, ytrue,verbose=True,)
But then I get the following error about my outputs. That I am not understanding.
File "/.virtualenvs/python310/lib/python3.10/site-packages/keras/losses.py", line 1990, in categorical_crossentropy
return backend.categorical_crossentropy(
File "/.virtualenvs/python310/lib/python3.10/site-packages/keras/backend.py", line 5529, in categorical_crossentropy
target.shape.assert_is_compatible_with(output.shape)
ValueError: Shapes (None, 3) and (None, 1, 3) are incompatible
I'm not entirely understanding why the shapes don't match. I expect my softmax layer to output 3 numbers that sum to 1 and can be compared to my ytrue.
any insights appreciated
Just from the model architecture itself, it seems like you just need a two-dimensional data to be fed into Layer2:
One may use a Reshape/Flatten layer to fix it.
By reshaping the output of Lambda layer from (None, 1, 2) to (None, 2), the final output's shape should become compatible too (None, 3).
Additional notes:
As an example borrowed (with some modifications) from the TensorFlow website, let's assume we want to split an input tensor of the shape of (3, 2) into 3 smaller tensors along the axis=1:
x = tf.Variable(tf.random.uniform([3, 2], -1, 1))
s0, s1, s2 = tf.split(x, num_or_size_splits=3, axis=1)
Output:
Here are the smaller tensor splits:
Now, we can see the shape is (1, 2), i.e. a 2D tensor consistent with the tensor it is derived from, and not a vector of the shape of (2,). In the context of your problem, for a batch, that would be (None, 1, 2).
I am working with sequence modelling in pytorch and trying to determine if the order of the pooling and linear decoding layer matters. Given that I have a sequence with the shape (Batch, Seqlen, dim_model) and I want to transform it into (Batch, dim_output) I will need a pooling layer for reducing the second dimension (SeqLen) and an affine transformation that maps dim_model to dim_output. Assume Batch = 16, SeqLen = 6000, dim_model = 32, dim_output = 5, we have the following input:
import torch
pooler = lambda x: x.mean(dim=1)
decoder = torch.nn.Linear(32, 5)
x = torch.randn(16, 6000, 32)
Would this:
y = decoder(pooler(x))
Be the same as:
y = pooler(decoder(x))
The normalized difference between both outputs suggest that they are close:
torch.norm(decoder(pooler(x)) - pooler(decoder(x)))
output:
tensor(6.5412e-08, grad_fn=<CopyBackwards>)
But can one say they are equivalent? Are the gradients computed in the same way?
I am interesting the case of using arbitrary pooling layer, this includes for instance the "last" pooler:
pooler = lambda x: x[:,-1]
torch.norm(decoder(pooler(x)) - pooler(decoder(x)))
output:
tensor(0., grad_fn=<CopyBackwards>)
A linear layer does x -> Ax+b for some matrix A and vector b.
If you have a bunch of x (x1, x2, x3, ..., xn) then A[(x1+...+xn)/n] = (Ax1 +... +Axn)/n, so for mean pooling, applying pooling first and then doing the linear layer results (up to floating point errors) in the same value as applying the linear layer first and then doing the pooling.
For "last pooling", the result is the same because it doesn't matter whether you apply A to every element and then afterwards only pick the final one, or if you pick the final one, and apply A to it.
However, for plenty of other operations, the result would not be the same. E.g. for max pooling, the result would in general not be the same.
e.g. if x1 = (1, 0, 0), x2 = (0, 1, 0), x3 = (0, 0, 1), and A = ((1, 1, 1)) then Ax1 = Ax2 =Ax3 = (1), so applying max pooling after the linear layer just gives you (1),
but max pooling applied to x1, x2, x3 gives you (1, 1, 1) and A(1, 1, 1) = 3.
So I want to understand exactly how the outputs and hidden state of a GRU cell are calculated.
I obtained the pre-trained model from here and the GRU layer has been defined as nn.GRU(96, 96, bias=True).
I looked at the the PyTorch Documentation and confirmed the dimensions of the weights and bias as:
weight_ih_l0: (288, 96)
weight_hh_l0: (288, 96)
bias_ih_l0: (288)
bias_hh_l0: (288)
My input size and output size are (1000, 8, 96). I understand that there are 1000 tensors, each of size (8, 96). The hidden state is (1, 8, 96), which is one tensor of size (8, 96).
I have also printed the variable batch_first and found it to be False. This means that:
Sequence length: L=1000
Batch size: B=8
Input size: Hin=96
Now going by the equations from the documentation, for the reset gate, I need to multiply the weight by the input x. But my weights are 2-dimensions and my input has three dimensions.
Here is what I've tried, I took the first (8, 96) matrix from my input and multiplied it with the transpose of my weight matrix:
Input (8, 96) x Weight (96, 288) = (8, 288)
Then I add the bias by replicating the (288) eight times to give (8, 288). This would give the size of r(t) as (8, 288). Similarly, z(t) would also be (8, 288).
This r(t) is used in n(t), since Hadamard product is used, both the matrices being multiplied have to be the same size that is (8, 288). This implies that n(t) is also (8, 288).
Finally, h(t) is the Hadamard produce and matrix addition, which would give the size of h(t) as (8, 288) which is wrong.
Where am I going wrong in this process?
TLDR; This confusion comes from the fact that the weights of the layer are the concatenation of input_hidden and hidden-hidden respectively.
- nn.GRU layer weight/bias layout
You can take a closer look at what's inside the GRU layer implementation torch.nn.GRU by peaking through the weights and biases.
>>> gru = nn.GRU(input_size=96, hidden_size=96, num_layers=1)
First the parameters of the GRU layer:
>>> gru._all_weights
[['weight_ih_l0', 'weight_hh_l0', 'bias_ih_l0', 'bias_hh_l0']]
You can look at gru.state_dict() to get the dictionary of weights of the layer.
We have two weights and two biases, _ih stands for 'input-hidden' and _hh stands for 'hidden-hidden'.
For more efficient computation the parameters have been concatenated together, as the documentation page clearly explains (| means concatenation). In this particular example num_layers=1 and k=0:
~GRU.weight_ih_l[k] – the learnable input-hidden weights of the layer (W_ir | W_iz | W_in), of shape (3*hidden_size, input_size).
~GRU.weight_hh_l[k] – the learnable hidden-hidden weights of the layer (W_hr | W_hz | W_hn), of shape (3*hidden_size, hidden_size).
~GRU.bias_ih_l[k] – the learnable input-hidden bias of the layer (b_ir | b_iz | b_in), of shape (3*hidden_size).
~GRU.bias_hh_l[k] – the learnable hidden-hidden bias of the (b_hr | b_hz | b_hn).
For further inspection we can get those split up with the following code:
>>> W_ih, W_hh, b_ih, b_hh = gru._flat_weights
>>> W_ir, W_iz, W_in = W_ih.split(H_in)
>>> W_hr, W_hz, W_hn = W_hh.split(H_in)
>>> b_ir, b_iz, b_in = b_ih.split(H_in)
>>> b_hr, b_hz, b_hn = b_hh.split(H_in)
Now we have the 12 tensor parameters sorted out.
- Expressions
The four expressions for a GRU layer: r_t, z_t, n_t, and h_t, are computed at each timestep.
The first operation is r_t = σ(W_ir#x_t + b_ir + W_hr#h + b_hr). I used the # sign to designate the matrix multiplication operator (__matmul__). Remember W_ir is shaped (H_in=input_size, hidden_size) while x_t contains the element at step t from the x sequence. Tensor x_t = x[t] is shaped as (N=batch_size, H_in=input_size). At this point, it's simply a matrix multiplication between the input x[t] and the weight matrix. The resulting tensor r is shaped (N, hidden_size=H_in):
>>> (x[t]#W_ir.T).shape
(8, 96)
The same is true for all other weight multiplication operations performed. As a result, you end up with an output tensor shaped (N, H_out=hidden_size).
In the following expressions h is the tensor containing the hidden state of the previous step for each element in the batch, i.e. shaped (N, hidden_size=H_out), since num_layers=1, i.e. there's a single hidden layer.
>>> r_t = torch.sigmoid(x[t]#W_ir.T + b_ir + h#W_hr.T + b_hr)
>>> r_t.shape
(8, 96)
>>> z_t = torch.sigmoid(x[t]#W_iz.T + b_iz + h#W_hz.T + b_hz)
>>> z_t.shape
(8, 96)
The output of the layer is the concatenation of the computed h tensors at
consecutive timesteps t (between 0 and L-1).
- Demonstration
Here is a minimal example of an nn.GRU inference manually computed:
Parameters
Description
Values
H_in
feature size
3
H_out
hidden size
2
L
sequence length
3
N
batch size
1
k
number of layers
1
Setup:
gru = nn.GRU(input_size=H_in, hidden_size=H_out, num_layers=k)
W_ih, W_hh, b_ih, b_hh = gru._flat_weights
W_ir, W_iz, W_in = W_ih.split(H_out)
W_hr, W_hz, W_hn = W_hh.split(H_out)
b_ir, b_iz, b_in = b_ih.split(H_out)
b_hr, b_hz, b_hn = b_hh.split(H_out)
Random input:
x = torch.rand(L, N, H_in)
Inference loop:
output = []
h = torch.zeros(1, N, H_out)
for t in range(L):
r = torch.sigmoid(x[t]#W_ir.T + b_ir + h#W_hr.T + b_hr)
z = torch.sigmoid(x[t]#W_iz.T + b_iz + h#W_hz.T + b_hz)
n = torch.tanh(x[t]#W_in.T + b_in + r*(h#W_hn.T + b_hn))
h = (1-z)*n + z*h
output.append(h)
The final output is given by the stacking the tensors h at consecutive timesteps:
>>> torch.vstack(output)
tensor([[[0.1086, 0.0362]],
[[0.2150, 0.0108]],
[[0.3020, 0.0352]]], grad_fn=<CatBackward>)
In this case the output shape is (L, N, H_out), i.e. (3, 1, 2).
Which you can compare with output, _ = gru(x).
I started working with Pytorch recently so my understanding of it isn't quite strong. I previously had a 1 layer CNN but wanted to extend it to 2 layers, but the input and output channels have been throwing errors I can seem to decipher. Why does it expect 192 channels? Can someone give me a pointer to help me understand this better? I have seen several related problems on here, but I don't understand those solutions either.
import numpy as np
import pandas as pd
import torch
import torch.nn as nn
from transformers import BertConfig, BertModel, BertTokenizer
import math
from transformers import AdamW, get_linear_schedule_with_warmup
def pad_sents(sents, pad_token): # Pad list of sentences according to the longest sentence in the batch.
sents_padded = []
max_len = max(len(s) for s in sents)
for s in sents:
padded = [pad_token] * max_len
padded[:len(s)] = s
sents_padded.append(padded)
return sents_padded
def sents_to_tensor(tokenizer, sents, device):
tokens_list = [tokenizer.tokenize(str(sent)) for sent in sents]
sents_lengths = [len(tokens) for tokens in tokens_list]
tokens_list_padded = pad_sents(tokens_list, '[PAD]')
sents_lengths = torch.tensor(sents_lengths, device=device)
masks = []
for tokens in tokens_list_padded:
mask = [0 if token == '[PAD]' else 1 for token in tokens]
masks.append(mask)
masks_tensor = torch.tensor(masks, dtype=torch.long, device=device)
tokens_id_list = [tokenizer.convert_tokens_to_ids(tokens) for tokens in tokens_list_padded]
sents_tensor = torch.tensor(tokens_id_list, dtype=torch.long, device=device)
return sents_tensor, masks_tensor, sents_lengths
class ConvModel(nn.Module):
def __init__(self, device, dropout_rate, n_class, out_channel=16):
super(ConvModel, self).__init__()
self.bert_config = BertConfig.from_pretrained('bert-base-uncased', output_hidden_states=True)
self.dropout_rate = dropout_rate
self.n_class = n_class
self.out_channel = out_channel
self.bert = BertModel.from_pretrained('bert-base-uncased', config=self.bert_config)
self.out_channels = self.bert.config.num_hidden_layers * self.out_channel
self.tokenizer = BertTokenizer.from_pretrained('bert-base-uncased', config=self.bert_config)
self.conv = nn.Conv2d(in_channels=self.bert.config.num_hidden_layers,
out_channels=self.out_channels,
kernel_size=(3, self.bert.config.hidden_size),
groups=self.bert.config.num_hidden_layers)
self.conv1 = nn.Conv2d(in_channels=self.out_channels,
out_channels=48,
kernel_size=(3, self.bert.config.hidden_size),
groups=self.bert.config.num_hidden_layers)
self.hidden_to_softmax = nn.Linear(self.out_channels, self.n_class, bias=True)
self.dropout = nn.Dropout(p=self.dropout_rate)
self.device = device
def forward(self, sents):
sents_tensor, masks_tensor, sents_lengths = sents_to_tensor(self.tokenizer, sents, self.device)
encoded_layers = self.bert(input_ids=sents_tensor, attention_mask=masks_tensor)
hidden_encoded_layer = encoded_layers[2]
hidden_encoded_layer = hidden_encoded_layer[0]
hidden_encoded_layer = torch.unsqueeze(hidden_encoded_layer, dim=1)
hidden_encoded_layer = hidden_encoded_layer.repeat(1, 12, 1, 1)
conv_out = self.conv(hidden_encoded_layer) # (batch_size, channel_out, some_length, 1)
conv_out = self.conv1(conv_out)
conv_out = torch.squeeze(conv_out, dim=3) # (batch_size, channel_out, some_length)
conv_out, _ = torch.max(conv_out, dim=2) # (batch_size, channel_out)
pre_softmax = self.hidden_to_softmax(conv_out)
return pre_softmax
def batch_iter(data, batch_size, shuffle=False, bert=None):
batch_num = math.ceil(data.shape[0] / batch_size)
index_array = list(range(data.shape[0]))
if shuffle:
data = data.sample(frac=1)
for i in range(batch_num):
indices = index_array[i * batch_size: (i + 1) * batch_size]
examples = data.iloc[indices]
sents = list(examples.train_BERT_tweet)
targets = list(examples.train_label.values)
yield sents, targets # list[list[str]] if not bert else list[str], list[int]
def train():
label_name = ['Yes', 'Maybe', 'No']
device = torch.device("cpu")
df_train = pd.read_csv('trainn.csv') # , index_col=0)
train_label = dict(df_train.train_label.value_counts())
label_max = float(max(train_label.values()))
train_label_weight = torch.tensor([label_max / train_label[i] for i in range(len(train_label))], device=device)
model = ConvModel(device=device, dropout_rate=0.2, n_class=len(label_name))
optimizer = AdamW(model.parameters(), lr=1e-3, correct_bias=False)
scheduler = get_linear_schedule_with_warmup(optimizer, num_warmup_steps=100, num_training_steps=1000) # changed the last 2 arguments to old ones
model = model.to(device)
model.train()
cn_loss = torch.nn.CrossEntropyLoss(weight=train_label_weight, reduction='mean')
train_batch_size = 16
for epoch in range(1):
for sents, targets in batch_iter(df_train, batch_size=train_batch_size, shuffle=True): # for each epoch
optimizer.zero_grad()
pre_softmax = model(sents)
loss = cn_loss(pre_softmax, torch.tensor(targets, dtype=torch.long, device=device))
loss.backward()
optimizer.step()
scheduler.step()
TrainingModel = train()
Here's a snippet of data https://github.com/Kosisochi/DataSnippet
It seems that the original version of the code you had in this question behaved differently. The final version of the code you have here gives me a different error from what you posted, more specifically - this:
RuntimeError: Calculated padded input size per channel: (20 x 1). Kernel size: (3 x 768). Kernel size can't be greater than actual input size
I apologize if I misunderstood the situation, but it seems to me that your understanding of what exactly nn.Conv2d layer does is not 100% clear and that is the main source of your struggle. I interpret the part "detailed explanation on 2 layer CNN in Pytorch" you requested as an ask to explain in detail on how that layer works and I hope that after this is done there will be no problem applying it 1 time, 2 times or more.
You can find all the documentation about the layer here, but let me give you a recap which hopefully will help to understand more the errors you're getting.
First of all nn.Conv2d inputs are 4-d tensors of the shape (BatchSize, ChannelsIn, Height, Width) and outputs are 4-d tensors of the shape (BatchSize, ChannelsOut, HeightOut, WidthOut). The simplest way to think about nn.Conv2d is of something applied to 2d images with pixel grid of size Height x Width and having ChannelsIn different colors or features per pixel. Even if your inputs have nothing to do with actual images the behavior of the layer is still the same. Simplest situation is when the nn.Conv2d is not using padding (as in your code). In that case the kernel_size=(kernel_height, kernel_width) argument specifies the rectangle which you can imagine sweeping through Height x Width rectangle of your inputs and producing one pixel for each valid position. Without padding the coordinate of the rectangle's point can be any pair of indicies (x, y) with x between 0 and Height - kernel_height and y between 0 and Width - kernel_width. Thus the output will look like a 2d image of size (Height - kernel_height + 1) x (Width - kernel_width + 1) and will have as many output channels as specified to nn.Conv2d constructor, so the output tensor will be of shape (BatchSize, ChannelsOut, Height - kernel_height + 1, Width - kernel_width + 1).
The parameter groups is not affecting how shapes are changed by the layer - it is only controlling which input channels are used as inputs for the output channels (groups=1 means that every input channel is used as input for every output channel, otherwise input and output channels are divided into corresponding number of groups and only input channels from group i are used as inputs for the output channels from group i).
Now in your current version of the code you have BatchSize = 16 and the output of pre-trained model is (BatchSize, DynamicSize, 768) with DynamicSize depending on the input, e.g. 22. You then introduce additional dimension as axis 1 with unsqueeze and repeat the values along that dimension transforming the tensor of shape (16, 22, 768) into (16, 12, 22, 768). Effectively you are using the output of the pre-trained model as 12-channel (with each channel having same values as others) 2-d images here of size (22, 768), where 22 is not fixed (depends on the batch). Then you apply a nn.Conv2d with kernel size (3, 768) - which means that there is no "wiggle room" for width and output 2-d images will be of size (20, 1) and since your layer has 192 channels final size of the output of first convolution layer has shape (16, 192, 20, 1). Then you try to apply second layer of convolution on top of that with kernel size (3, 768) again, but since your 2-d "image" is now just (20 x 1) there is no valid position to fit (3, 768) kernel rectangle inside a rectangle (20 x 1) which leads to the error message Kernel size can't be greater than actual input size.
Hope this explanation helps. Now to the choices you have to avoid the issue:
(a) is to add padding in such a way that the size of the output is not changing comparing to input (I won't go into details here,
because I don't think this is what you need)
(b) Use smaller kernel on both first and/or second convolutions (e.g. if you don't change first convolution the only valid width for
the second kernel would be 1).
(c) Looking at what you're trying to do my guess is that you actually don't want to use 2d convolution, you want 1d convolution (on the sequence) with every position described by 768 values. When you're using one convolution layer with 768 width kernel (and same 768 width input) you're effectively doing exactly same thing as 1d convolution with 768 input channels, but then if you try to apply second one you have a problem. You can specify kernel width as 1 for the next layer(s) and that will work for you, but a more correct way would be to transpose pre-trained model's output tensor by switching the last dimensions - getting shape (16, 768, DynamicSize) from (16, DynamicSize, 768) and then apply nn.Conv1d layer with 768 input channels and arbitrary ChannelsOut as output channels and 1d kernel_size=3 (meaning you look at 3 consecutive elements of the sequence for convolution). If you do that than without padding input shape of (16, 768, DynamicSize) will become (16, ChannelsOut, DynamicSize-2), and after you apply second Conv1d with e.g. the same settings as first one you'll get a tensor of shape (16, ChannelsOut, DynamicSize-4), etc. (each time the 1d length will shrink by kernel_size-1). You can always change number of channels/kernel_size for each subsequent convolution layer too.
I'm building my first RNN in tensorflow. After understanding all the concepts regarding the 3D input shape, I came across with this issue.
In my numpy version (1.15.4), the shape representation of 3D arrays is the following: (panel, row, column). I will make each dimension different so that it is clearer:
In [1]: import numpy as np
In [2]: arr = np.arange(30).reshape((2,3,5))
In [3]: arr
Out[3]:
array([[[ 0, 1, 2, 3, 4],
[ 5, 6, 7, 8, 9],
[10, 11, 12, 13, 14]],
[[15, 16, 17, 18, 19],
[20, 21, 22, 23, 24],
[25, 26, 27, 28, 29]]])
In [4]: arr.shape
Out[4]: (2, 3, 5)
In [5]: np.__version__
Out[5]: '1.15.4'
Here my understanding is: I have two timesteps with each timestep having 3 observations with 5 features in each observation.
However, in tensorflow "theory" (which I believe it is strongly based in numpy) RNN cells expect tensors (i.e. just n-dimensional matrices) of shape [batch_size, timesteps, features], which could be translated to: (row, panel, column) in the numpy "jargon".
As can be seen, the representation doesn't match, leading to errors when feeding numpy data into a placeholder, which in most of the examples and theory is defined like:
x = tf.placeholder(tf.float32, shape=[None, N_TIMESTEPS_X, N_FEATURES], name='XPlaceholder')
np.reshape() doesn't solve the issue because it just rearranges the dimensions, but messes up with the data.
I'm using for the first time the Dataset API, but I encounter the problems once into the session, not in the Dataset API ops.
I'm using the static_rnn method, and everything works well until I have to feed the data into the placeholder, which obviously results in a shape error.
I have tried to change the placeholder shape to shape=[N_TIMESTEPS_X, None, N_FEATURES]. HOWEVER, I'm using the dataset API, and I get errors when making the initializer if I change the Xplaceholder to the shape=[N_TIMESTEPS_X, None, N_FEATURES].
So, to summarize:
First problem: Shape errors with different shape representations.
Second problem: Dataset error when equating the shape representations (I think that either static_rnn or dynamic_rnn would function if this is resolved).
My question is:
¿Is there anything I'm missing in regard to this different representation logic which makes the practice confusing?
¿Could the solution be attained to switching to dynamic_rnn? (although the problems about the shape I encounter are related to the dataset API initializer being fed with shape [N_TIMESTEPS_X, None, N_FEATURES], not with the RNN cell itself.
Thank you very much for your time.
Full code:
'''The idea is to create xt, yt, xval and yval. My numpy arrays to
be fed are of the following shapes:
The 3D xt array has a shape of: (11, 69579, 74)
The 3D xval array has a shape of: (11, 7732, 74)
The yt array has a shape of: (69579, 3)
The yval array has a shape of: (7732, 3)
'''
N_TIMESTEPS_X = xt.shape[0] ## The stack number
BATCH_SIZE = 256
#N_OBSERVATIONS = xt.shape[1]
N_FEATURES = xt.shape[2]
N_OUTPUTS = yt.shape[1]
N_NEURONS_LSTM = 128 ## Number of units in the LSTMCell
N_NEURONS_DENSE = 64 ## Number of units in the Dense layer
N_EPOCHS = 600
LEARNING_RATE = 0.1
### Define the placeholders anda gather the data.
train_data = (xt, yt)
validation_data = (xval, yval)
## We define the placeholders as a trick so that we do not break into memory problems, associated with feeding the data directly.
'''As an alternative, you can define the Dataset in terms of tf.placeholder() tensors, and feed the NumPy arrays when you initialize an Iterator over the dataset.'''
batch_size = tf.placeholder(tf.int64)
x = tf.placeholder(tf.float32, shape=[None, N_TIMESTEPS_X, N_FEATURES], name='XPlaceholder')
y = tf.placeholder(tf.float32, shape=[None, N_OUTPUTS], name='YPlaceholder')
# Creating the two different dataset objects.
train_dataset = tf.data.Dataset.from_tensor_slices((x,y)).batch(BATCH_SIZE).repeat()
val_dataset = tf.data.Dataset.from_tensor_slices((x,y)).batch(BATCH_SIZE)
# Creating the Iterator type that permits to switch between datasets.
itr = tf.data.Iterator.from_structure(train_dataset.output_types, train_dataset.output_shapes)
train_init_op = itr.make_initializer(train_dataset)
validation_init_op = itr.make_initializer(val_dataset)
next_features, next_labels = itr.get_next()
### Create the graph
cellType = tf.nn.rnn_cell.LSTMCell(num_units=N_NEURONS_LSTM, name='LSTMCell')
inputs = tf.unstack(next_features, N_TIMESTEPS_X, axis=0)
'''inputs: A length T list of inputs, each a Tensor of shape [batch_size, input_size]'''
RNNOutputs, _ = tf.nn.static_rnn(cell=cellType, inputs=inputs, dtype=tf.float32)
predictionsLayer = tf.layers.dense(inputs=tf.layers.batch_normalization(RNNOutputs[-1]), units=N_NEURONS_DENSE, activation=None, name='Dense_Layer')
### Define the cost function, that will be optimized by the optimizer.
cost = tf.reduce_mean(tf.nn.softmax_cross_entropy_with_logits_v2(logits=predictionsLayer, labels=next_labels, name='Softmax_plus_Cross_Entropy'))
optimizer_type = tf.train.AdamOptimizer(learning_rate=LEARNING_RATE, name='AdamOptimizer')
optimizer = optimizer_type.minimize(cost)
### Model evaluation
correctPrediction = tf.equal(tf.argmax(predictionsLayer,1), tf.argmax(y,1))
accuracy = tf.reduce_mean(tf.cast(correctPrediction,tf.float32))
#confusionMatrix = tf.confusion_matrix(next_labels, predictionsLayer, num_classes=3, name='ConfMatrix')
N_BATCHES = train_data[0].shape[0] // BATCH_SIZE
## Saving variables so that we can restore them afterwards.
saver = tf.train.Saver()
save_dir = '/home/zmlaptop/Desktop/tfModels/{}_{}'.format(cellType.__class__.__name__, datetime.now().strftime("%Y%m%d%H%M%S"))
os.mkdir(save_dir)
varDict = {'nTimeSteps':N_TIMESTEPS_X, 'BatchSize': BATCH_SIZE, 'nFeatures':N_FEATURES,
'nNeuronsLSTM':N_NEURONS_LSTM, 'nNeuronsDense':N_NEURONS_DENSE, 'nEpochs':N_EPOCHS,
'learningRate':LEARNING_RATE, 'optimizerType': optimizer_type.__class__.__name__}
varDicSavingTxt = save_dir + '/varDict.txt'
modelFilesDir = save_dir + '/modelFiles'
os.mkdir(modelFilesDir)
logDir = save_dir + '/TBoardLogs'
os.mkdir(logDir)
acc_summary = tf.summary.scalar('Accuracy', accuracy)
loss_summary = tf.summary.scalar('Cost_CrossEntropy', cost)
summary_merged = tf.summary.merge_all()
with open(varDicSavingTxt, 'w') as outfile:
outfile.write(repr(varDict))
with tf.Session() as sess:
tf.set_random_seed(2)
sess.run(tf.global_variables_initializer())
train_writer = tf.summary.FileWriter(logDir + '/train', sess.graph)
validation_writer = tf.summary.FileWriter(logDir + '/validation')
# initialise iterator with train data
sess.run(train_init_op, feed_dict = {x : train_data[0], y: train_data[1], batch_size: BATCH_SIZE})
print('¡Training starts!')
for epoch in range(N_EPOCHS):
batchAccList = []
tot_loss = 0
for batch in range(N_BATCHES):
optimizer_output, loss_value, summary = sess.run([optimizer, cost, summary_merged])
accBatch = sess.run(accuracy)
tot_loss += loss_value
batchAccList.append(accBatch)
if batch % 10 == 0:
train_writer.add_summary(summary, batch)
epochAcc = tf.reduce_mean(batchAccList)
if epoch%10 == 0:
print("Epoch: {}, Loss: {:.4f}, Accuracy: {}".format(epoch, tot_loss / N_BATCHES, epochAcc))
#confM = sess.run(confusionMatrix)
#confDic = {'confMatrix': confM}
#confTxt = save_dir + '/confMDict.txt'
#with open(confTxt, 'w') as outfile:
# outfile.write(repr(confDic))
#print(confM)
# initialise iterator with validation data
sess.run(validation_init_op, feed_dict = {x : validation_data[0], y: validation_data[1], batch_size:len(validation_data[0])})
print('Validation Loss: {:4f}, Validation Accuracy: {}'.format(sess.run(cost), sess.run(accuracy)))
summary_val = sess.run(summary_merged)
validation_writer.add_summary(summary_val)
saver.save(sess, modelFilesDir)
Is there anything I'm missing in regard to this different
representation logic which makes the practice confusing?
In fact, you made a mistake about the input shapes of static_rnn and dynamic_rnn. The input shape of static_rnn is [timesteps,batch_size, features](link),which is a list of 2D tensors of shape [batch_size, features]. But The input shape of dynamic_rnn is either [timesteps,batch_size, features] or [batch_size,timesteps, features] depending on time_major is True or False(link).
Could the solution be attained to switching to dynamic_rnn?
The key is not that you use static_rnn or dynamic_rnn, but that your data shape matches the required shape. The general format of placeholder is like your code is [None, N_TIMESTEPS_X, N_FEATURES]. It's also convenient for you to use dataset API.
You can use transpose()(link) instead of reshape().transpose() will permute the dimensions of an array and won't messes up with the data.
So your code needs to be modified.
# permute the dimensions
xt = xt.transpose([1,0,2])
xval = xval.transpose([1,0,2])
# adjust shape,axis=1 represents timesteps
inputs = tf.unstack(next_features, axis=1)
Other errors should have nothing to do with rnn shape.