Scribd
Upload
70 views16 pages

TensorFlow Basics for Beginners

The document discusses TensorFlow tensors and their properties. It demonstrates how to create 0D, 1D, 2D, 3D and 4D tensors with TensorFlow and manipulate them. It also shows how to create identity, filled and ones tensors.

Uploaded by

Ganesh Nagal
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF, TXT or read online on Scribd
70 views16 pages

TensorFlow Basics for Beginners

The document discusses TensorFlow tensors and their properties. It demonstrates how to create 0D, 1D, 2D, 3D and 4D tensors with TensorFlow and manipulate them. It also shows how to create identity, filled and ones tensors.

Uploaded by

Ganesh Nagal
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF, TXT or read online on Scribd

8/5/23, 11:46 PM TensorFlow_ComputerVision_01.

ipynb - Colaboratory

TensorFlow from Scratch -- 01


Author: Ramdas Dharavath
Linkedin: [Link]
Source Code GitHub link: [Link]

Show code

Tensors are Multi Dimentional Arrays


An array is an ordered arrangement of numbers

Show code

Show code

[Link] 1/16
8/5/23, 11:46 PM TensorFlow_ComputerVision_01.ipynb - Colaboratory

8 --> oD

[0,4,6] --> 1D --> (3,)

2D [[1,2,3],
[4,5,6]] --> 2D --> (2,3)

3D --> [[[4,5,6],[7,2,1], [8,9,2]]] --> (1,3,3)

Import tensorflow as tf

import tensorflow as tf

0D Tensor

tensor_zero_D = [Link](4)
print(tensor_zero_D)

[Link](4, shape=(), dtype=int32)

1D Tensor

tensor_one_d = [Link]([3,4,5])
print(tensor_one_d)

[Link]([3 4 5], shape=(3,), dtype=int32)

tensor_one_d = [Link]([3,4,5],dtype=tf.float32)
print(tensor_one_d)

[Link]([3. 4. 5.], shape=(3,), dtype=float32)

tensor_one_d = [Link]([3,4,5],dtype=tf.float16)
print(tensor_one_d)

[Link]([3. 4. 5.], shape=(3,), dtype=float16)

[Link] 2/16
8/5/23, 11:46 PM TensorFlow_ComputerVision_01.ipynb - Colaboratory

tensor_one_d = [Link]([3,4.4,5,8.1],dtype=tf.int64)
print(tensor_one_d)

---------------------------------------------------------------------------
TypeError Traceback (most recent call last)
<ipython-input-30-e4c0cd578aa7> in <cell line: 1>()
----> 1 tensor_one_d = [Link]([3,4.4,5,8.1],dtype=tf.int64)
2 print(tensor_one_d)

3 frames
/usr/local/lib/python3.10/dist-packages/tensorflow/python/framework/constant_op.py i
convert_to_eager_tensor(value, ctx, dtype)
101 dtype = dtypes.as_dtype(dtype).as_datatype_enum
102 ctx.ensure_initialized()
--> 103 return [Link](value, ctx.device_name, dtype)
104
105

TypeError: Cannot convert [3, 4.4, 5, 8.1] to EagerTensor of dtype int64

SEARCH STACK OVERFLOW

solving above error using [Link]

tensor_one_d = [Link]([3,4.4,5,8.1],dtype=tf.float32)

casted_tensor_one_d =[Link](tensor_one_d,dtype=tf.int32)
print(tensor_one_d)
print(casted_tensor_one_d)

[Link]([3. 4.4 5. 8.1], shape=(4,), dtype=float32)


[Link]([3 4 5 8], shape=(4,), dtype=int32)

in bool if zero is false else: true

tensor_one_d = [Link]([3,4.4,0,5,8.1],dtype=tf.float32)

casted_tensor_one_d =[Link](tensor_one_d,dtype=[Link])
print(tensor_one_d)
print(casted_tensor_one_d)

[Link]([3. 4.4 0. 5. 8.1], shape=(5,), dtype=float32)


[Link]([ True True False True True], shape=(5,), dtype=bool)

tensor_bool_d = [Link]([True,False,True])
print(tensor_bool_d)

[Link]([ True False True], shape=(3,), dtype=bool)

[Link] 3/16
8/5/23, 11:46 PM TensorFlow_ComputerVision_01.ipynb - Colaboratory

tensor_string_d = [Link]("Computer Vision")
print(tensor_string_d)

[Link](b'Computer Vision', shape=(), dtype=string)

tensor_string_d = [Link](["Computer Vision"])
print(tensor_string_d)

[Link]([b'Computer Vision'], shape=(1,), dtype=string)

tensor_one_d = [Link]([3,4,5.6])
print(tensor_one_d)

[Link]([3. 4. 5.6], shape=(3,), dtype=float32)

tensor_one_d = [Link]([-3,4,5])
print(tensor_one_d)

[Link]([-3 4 5], shape=(3,), dtype=int32)

numpy array into tensor

import numpy as np

np_array = [Link]([2,3,4,5])
print(np_array)

[2 3 4 5]

converted_tensor = tf.convert_to_tensor(np_array)
print(converted_tensor)

[Link]([2 3 4 5], shape=(4,), dtype=int64)

2D Tensor

tensor_two_d = [Link]([
    [1,2,3],
    [4,5,6],
    [0,7,5]
])
print(tensor_two_d)
# stack of one d tensor form 2d tensor

[Link](
[[1 2 3]

[Link] 4/16
8/5/23, 11:46 PM TensorFlow_ComputerVision_01.ipynb - Colaboratory

[4 5 6]
[0 7 5]], shape=(3, 3), dtype=int32)

tensor_two_d.dtype

tf.int32

tensor_two_d.shape

TensorShape([3, 3])

3D Tensor

tensor_three_d = [Link]([
    [[1,2,3],
     [5,4,3]],
    [[1,6,9],
     [0,4,1]]
])

print(tensor_three_d)

[Link](
[[[1 2 3]
[5 4 3]]

[[1 6 9]
[0 4 1]]], shape=(2, 2, 3), dtype=int32)

tensor_three_d.shape

TensorShape([2, 2, 3])

tensor_three_d.ndim

tensor_three_d.dtype

tf.int32

4D Tensor

tensor_four_d = [Link]([
    [[[1,2,3],
      [90,50,44]]],

[Link] 5/16
8/5/23, 11:46 PM TensorFlow_ComputerVision_01.ipynb - Colaboratory

    [[[5,6,7],
      [90,30,22]]]
])
print(tensor_four_d)

[Link](
[[[[ 1 2 3]
[90 50 44]]]

[[[ 5 6 7]
[90 30 22]]]], shape=(2, 1, 2, 3), dtype=int32)

tensor_four_d.ndim

tensor_four_d.shape

TensorShape([2, 1, 2, 3])

tensor_four_d.dtype

tf.int32

Eye tensor

eye_tensor = [Link](
    num_rows=4,
    num_columns=None,
    batch_shape=None,
    dtype=[Link].float32,
    name=None
)
print(eye_tensor)
# construct identity matrix

[Link](
[[1. 0. 0. 0.]
[0. 1. 0. 0.]
[0. 0. 1. 0.]
[0. 0. 0. 1.]], shape=(4, 4), dtype=float32)

eye_tensor = [Link](
    num_rows=4,
    num_columns=None,
    batch_shape=None,
    dtype=[Link].float16,

[Link] 6/16
8/5/23, 11:46 PM TensorFlow_ComputerVision_01.ipynb - Colaboratory

    name=None
)
print(eye_tensor)
# construct identity matrix

[Link](
[[1. 0. 0. 0.]
[0. 1. 0. 0.]
[0. 0. 1. 0.]
[0. 0. 0. 1.]], shape=(4, 4), dtype=float16)

eye_tensor = [Link](
    num_rows=4,
    num_columns=None,
    batch_shape=None,
    dtype=[Link],
    name=None
)
print(eye_tensor)
# construct identity matrix

[Link](
[[ True False False False]
[False True False False]
[False False True False]
[False False False True]], shape=(4, 4), dtype=bool)

eye_tensor = [Link](
    num_rows=4,
    num_columns=3,
    batch_shape=None,
    dtype=[Link].float32,
    name=None
)
print(eye_tensor)
# construct identity matrix

[Link](
[[1. 0. 0.]
[0. 1. 0.]
[0. 0. 1.]
[0. 0. 0.]], shape=(4, 3), dtype=float32)

eye_tensor = [Link](
    num_rows=4,
    num_columns=None,
    batch_shape=[2,],
    dtype=[Link].float32,
    name=None
)
print(eye_tensor)
# construct identity matrix
# batch size 2 means i need two 4,4 matrix

[Link] 7/16
8/5/23, 11:46 PM TensorFlow_ComputerVision_01.ipynb - Colaboratory

[Link](
[[[1. 0. 0. 0.]
[0. 1. 0. 0.]
[0. 0. 1. 0.]
[0. 0. 0. 1.]]

[[1. 0. 0. 0.]
[0. 1. 0. 0.]
[0. 0. 1. 0.]
[0. 0. 0. 1.]]], shape=(2, 4, 4), dtype=float32)

fill tensor

fill_tensor = [Link](
    [3,4],5,name=None
)
print(fill_tensor)

[Link](
[[5 5 5 5]
[5 5 5 5]
[5 5 5 5]], shape=(3, 4), dtype=int32)

ones tensor

ones_tensor = [Link](
    [2,3],
    dtype=[Link].int32,
    name=None
)
print(ones_tensor)

[Link](
[[1 1 1]
[1 1 1]], shape=(2, 3), dtype=int32)

ones_tensor = [Link](
    [2,3,2],
    dtype=[Link].float32,
    name=None
)
print(ones_tensor)

[Link](
[[[1. 1.]
[1. 1.]
[1. 1.]]

[[1. 1.]

[Link] 8/16
8/5/23, 11:46 PM TensorFlow_ComputerVision_01.ipynb - Colaboratory

[1. 1.]
[1. 1.]]], shape=(2, 3, 2), dtype=float32)

fill2_tensor = [Link](
    [3,3],3,name=None
)
print(fill2_tensor)

[Link](
[[3 3 3]
[3 3 3]
[3 3 3]], shape=(3, 3), dtype=int32)

ones _like tensor

ones_like_tensor = tf.ones_like(
   fill2_tensor
)
print(ones_like_tensor)

[Link](
[[1 1 1]
[1 1 1]
[1 1 1]], shape=(3, 3), dtype=int32)

zeros tensor

zeros2_tensor = [Link](
    [3,2],
    dtype = [Link].float32,
    name=None
)
print(zeros2_tensor)

[Link](
[[0. 0.]
[0. 0.]
[0. 0.]], shape=(3, 2), dtype=float32)

ones_like2_tensor = tf.ones_like(zeros2_tensor)
print(ones_like2_tensor)

[Link](
[[1. 1.]
[1. 1.]
[1. 1.]], shape=(3, 2), dtype=float32)

print(tensor_three_d)

[Link] 9/16
8/5/23, 11:46 PM TensorFlow_ComputerVision_01.ipynb - Colaboratory

[Link](
[[[1 2 3]
[5 4 3]]

[[1 6 9]
[0 4 1]]], shape=(2, 2, 3), dtype=int32)

tensor_three_d.shape

TensorShape([2, 2, 3])

print([Link](tensor_three_d))

[Link]([2 2 3], shape=(3,), dtype=int32)

[Link] return rank of tensor

# [Link]   return rank of tensor

t = [Link]([[[1,2,3],[6,7,8]]])
print([Link](t))
[Link](t)

[Link](3, shape=(), dtype=int32)


<[Link]: shape=(), dtype=int32, numpy=3>

t1 = [Link]([[1,2,3],[5,6,7]])
print([Link](t1))
[Link](t1)

[Link](2, shape=(), dtype=int32)


<[Link]: shape=(), dtype=int32, numpy=2>

t1 = [Link]([1,2,3])
print([Link](t1))
[Link](t1)

[Link](1, shape=(), dtype=int32)


<[Link]: shape=(), dtype=int32, numpy=1>

# [Link]

[Link] 10/16
8/5/23, 11:46 PM TensorFlow_ComputerVision_01.ipynb - Colaboratory

t2 = [Link]([[1,2,3,4],[6,7,8,9]])
[Link](t2)

<[Link]: shape=(), dtype=int32, numpy=8>

t2 = [Link]([[1,2,3,4],[6,7,8,9]])
[Link](t2,out_type=tf.float32)

<[Link]: shape=(), dtype=float32, numpy=8.0>

# [Link]

random_tensor = [Link](
    [3,3],
    mean=0,
    stddev=1.0,
    dtype=[Link].float32,
    seed=None,
    name=None
)
print(random_tensor)

[Link](
[[-0.39316094 0.02094128 -0.6328461 ]
[ 2.299625 0.6251846 0.96148187]
[-0.80120754 -0.38471952 0.34140763]], shape=(3, 3), dtype=float32)

# [Link]
# range of values in between minval,maxvalue

random_tensor2 = [Link](
    [4,],
    minval=0,
    maxval=6,
    dtype = [Link].int32,
    seed = None,
    name=None

)
print(random_tensor2)

[Link]([0 5 2 2], shape=(4,), dtype=int32)

random_uniform_tensor = [Link](
    [3,4],
    minval = 1,
    maxval = 4,
    dtype = [Link].float64,
    seed = None,
    name = None

[Link] 11/16
8/5/23, 11:46 PM TensorFlow_ComputerVision_01.ipynb - Colaboratory

)
print(random_uniform_tensor)

[Link](
[[3.99208412 1.76304071 2.05448341 2.56923315]
[2.55513969 2.18310585 1.75114135 3.04348884]
[2.79654863 1.39844371 2.08378495 3.53174024]], shape=(3, 4), dtype=float64)

random_uniform_tensor = [Link](
    [3,3],
    minval = 1,
    maxval = None,
    dtype = [Link].float64,
    seed = None,
    name = None
)
print(random_uniform_tensor)

[Link](
[[1. 1. 1.]
[1. 1. 1.]
[1. 1. 1.]], shape=(3, 3), dtype=float64)

# seed argument
# tf.

[Link].set_seed(5)
print([Link](shape=[2,],maxval=4, dtype=tf.float32, seed=10))
print([Link](shape=[3,],maxval=3, dtype=tf.int32, seed=10))
print([Link](shape=[4,], maxval=6, dtype=tf.int64, seed=10))

[Link]([0.36659575 3.8268886 ], shape=(2,), dtype=float32)


[Link]([2 0 1], shape=(3,), dtype=int32)
[Link]([5 0 5 4], shape=(4,), dtype=int64)

[Link].set_seed(5)
print([Link](shape=[2,],maxval=4, dtype=tf.float32, seed=10))
print([Link](shape=[3,],maxval=3, dtype=tf.int32, seed=10))
print([Link](shape=[4,], maxval=6, dtype=tf.int64, seed=10))

[Link]([0.36659575 3.8268886 ], shape=(2,), dtype=float32)


[Link]([2 0 1], shape=(3,), dtype=int32)
[Link]([5 0 5 4], shape=(4,), dtype=int64)

[Link] 12/16
8/5/23, 11:46 PM TensorFlow_ComputerVision_01.ipynb - Colaboratory

random_uniform_tensor = [Link](
    [3,],
    minval = 1,
    maxval = 20,
    dtype = [Link].float64,
    seed = None,
    name = None
)
print(random_uniform_tensor)

[Link]([ 1.21176329 2.28550813 15.31137277], shape=(3,), dtype=float64)

random_uniform_tensor = [Link](
    [3,],
    minval = 1,
    maxval = 20,
    dtype = [Link].float64,
    seed = 10,
    name = None
)
print(random_uniform_tensor)

[Link]([14.93065576 5.51886788 18.68575432], shape=(3,), dtype=float64)

Tensor index

tensor_indexed = [Link]([1,2,3,4,5,6])
print(tensor_indexed)

[Link]([1 2 3 4 5 6], shape=(6,), dtype=int32)

print(tensor_indexed[0:5])

[Link]([1 2 3 4 5], shape=(5,), dtype=int32)

print(tensor_indexed[3:4])

[Link]([4], shape=(1,), dtype=int32)

print(tensor_indexed[1:4+1])

[Link]([2 3 4 5], shape=(4,), dtype=int32)

print(tensor_indexed[Link])

[Link]([1 3], shape=(2,), dtype=int32)

[Link] 13/16
8/5/23, 11:46 PM TensorFlow_ComputerVision_01.ipynb - Colaboratory

print(tensor_indexed[2:])

[Link]([3 4 5 6], shape=(4,), dtype=int32)

print(tensor_indexed[1:-2])

[Link]([2 3 4], shape=(3,), dtype=int32)

tensor_two_d2 = [Link]([[2,1,3,4],
                             [8,7,6,5]])
print(tensor_two_d2)

[Link](
[[2 1 3 4]
[8 7 6 5]], shape=(2, 4), dtype=int32)

print(tensor_two_d2[1:,1])

[Link]([7], shape=(1,), dtype=int32)

print(tensor_two_d2[0:,2])

[Link]([3 6], shape=(2,), dtype=int32)

print(tensor_two_d2)
print(tensor_two_d2[...,1])

[Link](
[[2 1 3 4]
[8 7 6 5]], shape=(2, 4), dtype=int32)
[Link]([1 7], shape=(2,), dtype=int32)

print(tensor_three_d)

[Link](
[[[1 2 3]
[5 4 3]]

[[1 6 9]
[0 4 1]]], shape=(2, 2, 3), dtype=int32)

print(tensor_three_d[0, 0, 1])

[Link](2, shape=(), dtype=int32)

print(tensor_three_d[1,0,2])

[Link](9, shape=(), dtype=int32)

print(tensor_three_d[0,:,1])

[Link] 14/16
8/5/23, 11:46 PM TensorFlow_ComputerVision_01.ipynb - Colaboratory

[Link]([2 4], shape=(2,), dtype=int32)

print(tensor_three_d[:,:,1])

[Link](
[[2 4]
[6 4]], shape=(2, 2), dtype=int32)

print(tensor_three_d[1,1,1])

[Link](4, shape=(), dtype=int32)

print(tensor_three_d)

[Link](
[[[1 2 3]
[5 4 3]]

[[1 6 9]
[0 4 1]]], shape=(2, 2, 3), dtype=int32)

print(tensor_three_d[0:2, : , -1])

[Link](
[[3 3]
[9 1]], shape=(2, 2), dtype=int32)

print(tensor_three_d[..., :, 1])

[Link](
[[2 4]
[6 4]], shape=(2, 2), dtype=int32)

# range

[Link](1,5)

<[Link]: shape=(4,), dtype=int32, numpy=array([1, 2, 3, 4], dtype=int32)>

# [Link]

x_Abs = [Link]([-1,3,4,5.5])
[Link](x_Abs)

[Link] 15/16
8/5/23, 11:46 PM TensorFlow_ComputerVision_01.ipynb - Colaboratory

<[Link]: shape=(4,), dtype=float32, numpy=array([1. , 3. , 4. , 5.5],


dtype=float32)>

check 0s completed at 11:33 PM

[Link] 16/16

You might also like