How exactly does LSTMCell from TensorFlow operates?

i examined this link and your code is almost perfect but you forgot to add forget_bias value(default 1.0) in this line F = vsigmoid(g3) its actualy F = vsigmoid(g3+self.forget_bias) or in your case its 1 F = vsigmoid(g3+1)

here is my imp with numpy:

import numpy as np
import tensorflow as tf

num_units = 3
lstm = tf.nn.rnn_cell.LSTMCell(num_units = num_units)
batch=1
timesteps = 7
num_input = 4
X = tf.placeholder("float", [batch, timesteps, num_input])
x = tf.unstack(X, timesteps, 1)
outputs, states = tf.contrib.rnn.static_rnn(lstm, x, dtype=tf.float32)
sess = tf.Session()
init = tf.global_variables_initializer()
sess.run(init)
x_val = np.reshape(range(28),[batch, timesteps, num_input])
res = sess.run(outputs, feed_dict = {X:x_val})
for e in res:
    print(e)
print("\nmy imp\n")
#my impl
def sigmoid(x):
    return 1/(1+np.exp(-x))

kernel,bias=sess.run([lstm._kernel,lstm._bias])
f_b_=lstm._forget_bias
c,h=np.zeros([batch,num_input-1]),np.zeros([batch,num_input-1])
for step in range(timesteps):
    inpt=np.split(x_val,7,1)[step][0]
    lstm_mtrx=np.matmul(np.concatenate([inpt,h],1),kernel)+bias
    i,j,f,o=np.split(lstm_mtrx,4,1)
    c=sigmoid(f+f_b_)*c+sigmoid(i)*np.tanh(j)
    h=sigmoid(o)*np.tanh(c)
    print(h)

output:

[[ 0.06964055 -0.06541953 -0.00682676]]
[[ 0.005264   -0.03234607  0.00014838]]
[[ 1.617855e-04 -1.316892e-02  8.596722e-06]]
[[ 3.9425286e-06 -5.1347450e-03  7.5078127e-08]]
[[ 8.7508155e-08 -1.9560163e-03  6.3853928e-10]]
[[ 1.8867894e-09 -7.3784427e-04  5.8551406e-12]]
[[ 4.0385355e-11 -2.7728223e-04  5.3957669e-14]]

my imp

[[ 0.06964057 -0.06541953 -0.00682676]]
[[ 0.005264   -0.03234607  0.00014838]]
[[ 1.61785520e-04 -1.31689185e-02  8.59672610e-06]]
[[ 3.94252745e-06 -5.13474567e-03  7.50781122e-08]]
[[ 8.75080644e-08 -1.95601574e-03  6.38539112e-10]]
[[ 1.88678843e-09 -7.37844070e-04  5.85513438e-12]]
[[ 4.03853841e-11 -2.77282006e-04  5.39576024e-14]]

Tensorflow uses glorot_uniform() function to initialize the lstm kernel, which samples weights from a random uniform distribution. We need to fix a value for the kernel to get reproducible results:

import tensorflow as tf
import numpy as np

np.random.seed(0)
timesteps = 7
num_input = 4
x_val = np.random.normal(size = (1, timesteps, num_input))

num_units = 3

def glorot_uniform(shape):
    limit = np.sqrt(6.0 / (shape[0] + shape[1]))
    return np.random.uniform(low=-limit, high=limit, size=shape)

kernel_init = glorot_uniform((num_input + num_units, 4 * num_units))

My implementation of the LSTMCell (well, actually it's just slightly rewritten tensorflow's code):

def sigmoid(x):
    return 1. / (1 + np.exp(-x))

class LSTMCell():
    """Long short-term memory unit (LSTM) recurrent network cell.
    """
    def __init__(self, num_units, initializer=glorot_uniform,
               forget_bias=1.0, activation=np.tanh):
        """Initialize the parameters for an LSTM cell.
        Args:
          num_units: int, The number of units in the LSTM cell.
          initializer: The initializer to use for the kernel matrix. Default: glorot_uniform
          forget_bias: Biases of the forget gate are initialized by default to 1
            in order to reduce the scale of forgetting at the beginning of
            the training. 
          activation: Activation function of the inner states.  Default: np.tanh.
        """
        # Inputs must be 2-dimensional.
        self._num_units = num_units
        self._forget_bias = forget_bias
        self._activation = activation
        self._initializer = initializer

    def build(self, inputs_shape):
        input_depth = inputs_shape[-1]
        h_depth = self._num_units
        self._kernel = self._initializer(shape=(input_depth + h_depth, 4 * self._num_units))
        self._bias = np.zeros(shape=(4 * self._num_units))

    def call(self, inputs, state):
        """Run one step of LSTM.
        Args:
          inputs: input numpy array, must be 2-D, `[batch, input_size]`.
          state:  a tuple of numpy arrays, both `2-D`, with column sizes `c_state` and
            `m_state`.
        Returns:
          A tuple containing:
          - A `2-D, [batch, output_dim]`, numpy array representing the output of the
            LSTM after reading `inputs` when previous state was `state`.
            Here output_dim is equal to num_units.
          - Numpy array(s) representing the new state of LSTM after reading `inputs` when
            the previous state was `state`.  Same type and shape(s) as `state`.
        """
        num_proj = self._num_units
        (c_prev, m_prev) = state

        input_size = inputs.shape[-1]

        # i = input_gate, j = new_input, f = forget_gate, o = output_gate
        lstm_matrix = np.hstack([inputs, m_prev]).dot(self._kernel)
        lstm_matrix += self._bias

        i, j, f, o = np.split(lstm_matrix, indices_or_sections=4, axis=0)
        # Diagonal connections
        c = (sigmoid(f + self._forget_bias) * c_prev + sigmoid(i) *
               self._activation(j))

        m = sigmoid(o) * self._activation(c)

        new_state = (c, m)
        return m, new_state

X = x_val.reshape(x_val.shape[1:])

cell = LSTMCell(num_units, initializer=lambda shape: kernel_init)
cell.build(X.shape)

state = (np.zeros(num_units), np.zeros(num_units))
for i in range(timesteps):
    x = X[i,:]
    output, state = cell.call(x, state)
    print(output)

Produces output:

[-0.21386017 -0.08401277 -0.25431477]
[-0.22243588 -0.25817422 -0.1612211 ]
[-0.2282134  -0.14207162 -0.35017249]
[-0.23286737 -0.17129192 -0.2706512 ]
[-0.11768674 -0.20717363 -0.13339118]
[-0.0599215  -0.17756104 -0.2028935 ]
[ 0.11437953 -0.19484555  0.05371994]

While your Tensorflow code, if you replace the second line with

lstm = tf.nn.rnn_cell.LSTMCell(num_units = num_units, initializer = tf.constant_initializer(kernel_init))

returns:

[[-0.2138602  -0.08401276 -0.25431478]]
[[-0.22243595 -0.25817424 -0.16122109]]
[[-0.22821338 -0.1420716  -0.35017252]]
[[-0.23286738 -0.1712919  -0.27065122]]
[[-0.1176867  -0.2071736  -0.13339119]]
[[-0.05992149 -0.177561   -0.2028935 ]]
[[ 0.11437953 -0.19484554  0.05371996]]

Considering Linear Algebra, it's possible to exist a dimension mismatch in the matrix multiplication between I*N (red circle), affecting the output, given that n x m dot m x p will give you a n x p dimensional output.

LSTM


Here is a blog which will answer any conceptual questions related to LSTM's. Seems that there is a lot which goes into building an LSTM from scratch!

Of course, this answer doesn't solve your question but just giving a direction.