WEBVTT

1
00:00:00.000 --> 00:00:02.621
Part 7 of ML Libraries with a Bot.

2
00:00:02.621 --> 00:00:06.707
TensorFlow and Keras Install Requirements,
Explained.

3
00:00:06.707 --> 00:00:12.643
TensorFlow with its high-level Keras API
is the other big deep learning stack

4
00:00:12.643 --> 00:00:14.031
alongside PyTorch.

5
00:00:14.031 --> 00:00:20.430
Its install rules depend on your operating
system, CPU architecture, Python version

6
00:00:20.430 --> 00:00:22.974
and, for GPUs, the NVIDIA driver.

7
00:00:22.974 --> 00:00:29.295
This explainer walks through what
TensorFlow's own pip page says, points out
where

8
00:00:29.295 --> 00:00:34.460
the page disagrees with itself, and shows
a real CPU install check.

9
00:00:35.292 --> 00:00:37.978
What the bot can and cannot do.

10
00:00:37.978 --> 00:00:41.878
Summarize the install page for your
platform.

11
00:00:41.878 --> 00:00:49.072
Explain the difference between tensorflow,
tensorflow-cpu and tensorflow[and-cuda].

12
00:00:49.072 --> 00:00:54.445
Read an import error and suggest which
requirement is missing.

13
00:00:54.445 --> 00:00:58.172
Know the current supported Python
versions.

14
00:00:58.172 --> 00:01:00.425
They change with releases.

15
00:01:00.425 --> 00:01:04.585
Resolve a contradiction in the docs by
guessing.

16
00:01:04.585 --> 00:01:07.272
Only a real install settles it.

17
00:01:07.272 --> 00:01:09.525
Set up a GPU it can't see.

18
00:01:10.125 --> 00:01:14.986
Read the install page for your exact
platform, note the versions, then install
in a

19
00:01:14.986 --> 00:01:16.508
fresh virtual environment.

20
00:01:17.167 --> 00:01:23.275
When we checked TensorFlow's "Install
TensorFlow with pip" page on Nov 3, 2026:

21
00:01:23.875 --> 00:01:28.887
The Software requirements section listed
Python 3.9 to 3.12.

22
00:01:28.887 --> 00:01:35.485
Further down, a warning said that as of
TensorFlow 2.21 Python 3.9 is no longer

23
00:01:35.485 --> 00:01:42.334
supported, suggested 3.10 to 3.13, and the
wheel table included Python 3.13 wheels

24
00:01:42.334 --> 00:01:43.253
for 2.21.0.

25
00:01:43.253 --> 00:01:50.270
We tested the newer claim: That came from
tensorflow-cpu 2.21.0 on Python 3.13.5, so

26
00:01:50.270 --> 00:01:52.358
the 3.13 wheels are real.

27
00:01:53.208 --> 00:02:00.497
tf.config.list_physical_devices("GPU")
returned an empty list, as expected on a
CPU-

28
00:02:00.497 --> 00:02:01.278
only box.

29
00:02:01.278 --> 00:02:04.923
On a working GPU setup it lists your GPUs.

30
00:02:04.923 --> 00:02:12.125
The one-layer Keras model learned y = 2x +
1 with weight 2.0 and bias 1.0, the same

31
00:02:12.125 --> 00:02:19.154
check as the PyTorch part.
keras.utils.set_random_seed(0) seeds
Python, NumPy and

32
00:02:19.154 --> 00:02:20.976
the backend together.

33
00:02:21.750 --> 00:02:24.785
Our install came with Keras 3.15.1.

34
00:02:24.785 --> 00:02:28.861
TensorFlow's Keras guide is the place to
start.

35
00:02:28.861 --> 00:02:35.799
Keras 3 can also run on JAX or PyTorch
backends, so if a bot shows you code from

36
00:02:35.799 --> 00:02:42.563
older tf.keras tutorials, check it against
current docs before you rely on it.

37
00:02:43.250 --> 00:02:49.659
For a Linux GPU, the pip page's steps are:
install the NVIDIA driver and check it

38
00:02:49.659 --> 00:02:56.147
with nvidia-smi, create a venv, pip
install tensorflow[and-cuda], then run the
GPU

39
00:02:56.147 --> 00:02:56.622
check.

40
00:02:56.622 --> 00:03:02.635
If no GPU appears, it describes linking
NVIDIA libraries in the environment.

41
00:03:02.635 --> 00:03:08.965
For older TensorFlow releases, the "Build
from source" page has tables of tested

42
00:03:08.965 --> 00:03:12.288
Python, compiler, CUDA and cuDNN versions.

43
00:03:13.125 --> 00:03:19.184
We won't rank them; both are widely used
and both have thorough official docs.

44
00:03:19.184 --> 00:03:25.320
Practical questions are more useful: which
library does the tutorial, course or

45
00:03:25.320 --> 00:03:31.767
model you want to use rely on, which one
does your team already know, and which one

46
00:03:31.767 --> 00:03:33.553
supports your platform?

47
00:03:33.553 --> 00:03:40.078
On native Windows with a GPU, for example,
TensorFlow's own page points you to WSL2.

48
00:03:40.833 --> 00:03:42.963
Tested vs documented.

49
00:03:42.963 --> 00:03:51.483
Tested by us: tensorflow-cpu 2.21.0 with
Keras 3.15.1 on Python 3.13.5, the GPU
list

50
00:03:51.483 --> 00:03:54.323
(empty), and the tiny model.

51
00:03:54.323 --> 00:04:02.132
Documented, not tested by us: all GPU,
Windows, macOS and Arm64 details, from

52
00:04:02.132 --> 00:04:04.364
TensorFlow's pip page.

53
00:04:05.167 --> 00:04:08.108
The next part today tries JAX on the CPU.

54
00:04:08.708 --> 00:04:14.752
Before you run anything: check every bot
answer against the official docs, work in
a

55
00:04:14.752 --> 00:04:20.652
virtual environment, load model files only
from sources you trust, respect dataset

56
00:04:20.652 --> 00:04:25.904
and model licenses, and keep personal data
out of bots and training sets.

57
00:04:25.904 --> 00:04:28.423
Every source is linked on the page.