Video summary
François Chollet recounts the origins of Keras, which he began developing in February 2015 amidst a deep learning landscape dominated by C++ libraries like Caffe and Torch7. At that time, the field was relatively small with fewer than 10,000 practitioners, and computer vision using convolutional neural networks (CNNs) was the primary focus while natural language processing remained niche. Chollet identified a significant gap in reusable open-source implementations for NLP architectures like LSTMs and decided to build Keras as a Python-based solution that defined models directly through code rather than static configuration files, a design choice that contrasted with the prevailing industry standard at the time but aligned more closely with usability-focused tools like PyTorch. The project gained traction because it offered an easy-to-use interface for neural networks exactly when interest in NLP was surging, leading Chollet to join Google later in 2015 where he encountered early internal versions of TensorFlow. Although his initial role involved computer vision research rather than Keras development, the release of TensorFlow 0.7 served as a catalyst for him to port Keras onto the new framework by December 2015. This transition was not merely a simple port but an abstraction effort that allowed the same codebase to run on multiple backends like Theano and CNTK before eventually settling with TensorFlow, which quickly overtook other options in speed and adoption due to its architectural advantages for distributed training. In October 2016, Chollet moved from his independent research work at Google to a dedicated role within the TensorFlow team following an invitation from Rajat Monga. This shift allowed him to integrate the Keras API more tightly into TensorFlow's core infrastructure, evolving it from a separate contrib library into the standard high-level interface for the framework. The integration process transformed what was essentially a temporary tent-like version of Keras into a permanent component that defined how users interacted with deep learning models within Google's ecosystem, marking a pivotal moment where Chollet transitioned from building an external tool to shaping the internal architecture of one of AI's most influential platforms. Chollet expresses significant excitement about TensorFlow 2.0 and its implementation of eager execution, which fundamentally changes the framework's usability by making it more flexible without sacrificing high-level convenience. Previously, users faced a trade-off between using Keras for ease of use or writing raw code for flexibility; however, the new version eliminates this dichotomy by allowing developers to seamlessly mix low-level control with higher-level abstractions through subclassing and custom training loops. This spectrum of workflows enables researchers, data scientists, and machine learning engineers to operate at their preferred level of abstraction while maintaining access to powerful features like automatic differentiation and optimized execution graphs that were previously difficult to debug or customize in eager mode environments.
Read the full video transcript
let's go from the philosophical to the
practical I can give me a history of
Karis and all the major deep learning
frameworks that you kind of remember in
relation to chaos and in general
tensorflow siano the old days you give a
brief overview Wikipedia style history
and your role in it before return to AGI
discussions yeah that's a broad topic so
I started working on chaos to the name
chaos at the time I actually picked the
name like just today I was gonna release
it so I started working on it in
February 2015 and so at the time there
weren't too many people working on deep
learning maybe like fewer than 10,000
the software tuning was not really
developed
so the
deepening library was cafe which was
mostly C++ why do I say cafe was the
main one cafe was vastly more popular
than ya know in in late 2014 early 2015
cafe was the one library that everyone
was using for computer vision and
computer vision was the most popular
problem absolutely company like
covenants was like the subfield of
deplaning it everyone was working on so
myself suing in in late 2014 I was
actually interested in islands in Rico
neural networks which was a very niche
topic at the time right III a tree to
catherine around 2016 and so I was
looking for good tools and I had used
torch 7 News Channel you stay on a lot
in Carroll competitions mmm I just cafe
and there was no like good solution for
Ireland's at the time like there was no
reusable open-source implementation of
in lsdm for instance so I decided to
build my own and that first the pitch
for that was it was going to be mostly
around LSTA memory on your networks it
was going to be in Python an important
decision at the time that was Canon are
obvious is that the models would be
defined yeah a Python code which was
kind of like going against the
mainstream at the time because cafe
thailand who wants on like all the big
libraries were actually going with you
approach sharing static configuration
files in yemen to define models so some
libraries were using code to define
models like torch 7 obviously that was
not
python Lezyne was like a piano based
very early library that was I think
developed I don't remember exactly
probably late 2014 Python as well
it's Python as well it was it was like
on top of Tiano and so I started working
on something
and in the value proposition at the time
was that not only that the what I think
was the first reducible open-source
implementation affair astrium
you could combine Islands and covenants
with the same library which is not
really possible before like a he was on
into incontinence and it was kind of
easy to use because so before I was
using the N I was actually doing cycling
and I loved psychically for its
usability so I drew a lot of inspiration
from cycling when I meant Cara's it's
almost like psychically and for neural
networks yeah the fit function exactly
the v function like reducing a complex
training loop to a single function core
right and of course you know some people
will say this is hiding a lot of details
but that's exactly the point
all right the magic is the point right
so it's magical but in a good way it's
magical in the sense that it's
delightful yeah right yeah I'm actually
quite surprised I didn't know that it
was born out of desire to implement our
hands in lc/ms it was that's fascinating
so you were actually one of the first
people to really try to attempt to get
the major architectures together and
it's also interesting you made me
realize that that was a design decision
at all is defining the modeling code
just I'm putting myself in your shoes
whether the yamo especially if cafe was
the most popular it was the most spoken
I might fall if I was I'm if I were yeah
I don't it I didn't like the nominal
thing but it makes more sense that you
will put in a configuration file the
definition of a model that's an
interesting gutsy move just stick with
defining it in code just if you look
back other libraries where we're doing
it this way but it was definitely the
more niche option yeah okay Cara's and
then your ass sorry discus in March 2015
and it got you just pretty much from the
start so the deepening community was
very small at the time
lots of people were starting to be
interested in the rest um so it was
gonna release it at the right time
because it was offering and easy to use
it as implementation exactly at the time
where lots of you started to be
intrigued by the capabilities of O&N on
and so NLP so it it grew from there
then I joined Google
months later and that was actually
completely unrelated to took care of
actually joined a research team working
on image classification mostly like
computer vision so I was doing computer
vision research at Google initially and
immediately when I joined Google I was
exposed to the early internal version of
tensorflow
and the way to peel to me at the time
and it was definitely wait West at the
time is that this was an improved
version of Tiano
so I immediately knew I had to Port
Charles to this new tensorflow thing and
I was actually very busy as as as a
noogler as new Googler so I had not time
to work on that but then in November I
think twist November 2015
tensile flu got released and it was kind
of like my my wake-up call at hey to
actually you know go on make it happen
so in December I I putted cars to run on
two of tensorflow but it was not exactly
port it was more accurate factoring
where I was abstracting away all the
backend functionality into one module
then the same codebase could run on top
of multiple backends right so on top of
danceBlue orthia no and for the next
year yeah no you know state as the
default option it was you know it was
easier to use somewhat let's begin it
was much faster especially when he came
to ordinance but eventually you know a
tensorflow overtook it right and tester
for the early test for a similar
architectural decisions this piano yeah
so what is there was a natural as a
natural transition yeah absolutely so
what I mean that still carries is the
side almost fun project right yeah so it
it was not my job assignment it's not I
was doing it on the side so I'm and even
though it grew to have you know a lot of
uses for a deepening library at the time
like Stroud 2016 but I wasn't doing it
as my main job so things started
changing
in I think it's mustard maybe October
2016 so one year later so Rashad who has
the lead intensive law basically showed
up one day in our building I was doing
like so I was doing research and things
like so I added of computer vision
research also collaborations with
Christians are getting and deplaning for
theory improving this is a really
interesting research topic and so Rajat
was saying hey we so Kara's we liked it
we saw that you had Google why don't you
come over for like a quarter and and and
work with us I was like yeah that sounds
like a great opportunity let's do it and
so I started working on integrating the
chaos API into tends to flow more
tightly so what fold up is sort of like
temporary tents of lonely version of
chaos that was in tents for that contrib
for a while and finally moved to dance
to the core and you know I've never
actually gotten back to my old sim doing
research well it's kind of funny that
somebody like you who dreams of or at
least sees the power of AI systems the
reason and theorem proving we'll talk
about has also created a system and
makes the the most basic kind of Lego
building that is deep learning super
accessible super easy so beautifully so
that's the funny irony that your book is
just both you're responsible for both
things but so Tessa flow 2.0 it's kind
of there's a sprint I don't know how
long I'll take but there's a sprint
towards the finish what do you look what
are you working on these days what are
you excited about what are you excited
about in 2.0 I mean eager execution
there's so many things that just make it
a lot easier
ya know to work what are you excited
about and what's also really hard what
are the problems you have to kind of
solve so I've spent the past yeah aha
working on 1002 it's been a long journey
I'm actually extremely excited about it
I think it's a great product it's a
delightful product competitors for one
we met huge progress so on the carrot
side what I'm really excited about is
that so you know previously Kara's has
been this very easy-to-use high level
interface to do deep learning but if you
wanted to
you know if you wanted a lot of
flexibility the chaos framework you know
was probably not the optimal way to do
things compared to just writing
everything from scratch so in some way
the framework was getting in the way and
in terms of you - you don't have this at
all actually you have the usability of
the high level interface but you have
the flexibility of this lower level
interface and you have this spectrum of
workflows where you can get more or less
usability and flexibility the trade-offs
depending on your needs right you can
write everything from scratch and you
get a lot of help doing so by you know
sub-classing models and writing some
train loops using ego execution it's
very flexible is very easy to debug is
very powerful but all of these
integrates seamlessly with higher level
features up to you know the classic era
square fruits which which are very
psychically unlike and and you know
ideal for a data scientist machining
engineer type of profile so now you can
have the same framework offering the
same set of api's that enable a spectrum
of workflows that are more or less
Louisville more less high level that are
suitable for you know profiles ranging
from researchers to data scientists and
everything in between
you