title
stringlengths
12
112
published
stringlengths
19
23
url
stringlengths
28
28
video_id
stringlengths
11
11
channel_id
stringclasses
5 values
id
stringlengths
16
31
text
stringlengths
0
596
start
float64
0
37.8k
end
float64
2.18
37.8k
Training and Testing an Italian BERT - Transformers From Scratch #4
2021-07-06 13:00:03 UTC
https://youtu.be/35Pdoyi6ZoQ
35Pdoyi6ZoQ
UCv83tO5cePwHMt1952IVVHw
35Pdoyi6ZoQ-t0.0
Hi, welcome to the video.
0
9.36
Training and Testing an Italian BERT - Transformers From Scratch #4
2021-07-06 13:00:03 UTC
https://youtu.be/35Pdoyi6ZoQ
35Pdoyi6ZoQ
UCv83tO5cePwHMt1952IVVHw
35Pdoyi6ZoQ-t3.0
So this is the fourth video in a Transformers
3
11.56
Training and Testing an Italian BERT - Transformers From Scratch #4
2021-07-06 13:00:03 UTC
https://youtu.be/35Pdoyi6ZoQ
35Pdoyi6ZoQ
UCv83tO5cePwHMt1952IVVHw
35Pdoyi6ZoQ-t9.36
from Scratch mini series.
9.36
15.84
Training and Testing an Italian BERT - Transformers From Scratch #4
2021-07-06 13:00:03 UTC
https://youtu.be/35Pdoyi6ZoQ
35Pdoyi6ZoQ
UCv83tO5cePwHMt1952IVVHw
35Pdoyi6ZoQ-t11.56
So if you haven't been following along,
11.56
18.48
Training and Testing an Italian BERT - Transformers From Scratch #4
2021-07-06 13:00:03 UTC
https://youtu.be/35Pdoyi6ZoQ
35Pdoyi6ZoQ
UCv83tO5cePwHMt1952IVVHw
35Pdoyi6ZoQ-t15.84
we've essentially covered what you can see on the screen.
15.84
20.6
Training and Testing an Italian BERT - Transformers From Scratch #4
2021-07-06 13:00:03 UTC
https://youtu.be/35Pdoyi6ZoQ
35Pdoyi6ZoQ
UCv83tO5cePwHMt1952IVVHw
35Pdoyi6ZoQ-t18.48
So we got some data.
18.48
23.72
Training and Testing an Italian BERT - Transformers From Scratch #4
2021-07-06 13:00:03 UTC
https://youtu.be/35Pdoyi6ZoQ
35Pdoyi6ZoQ
UCv83tO5cePwHMt1952IVVHw
35Pdoyi6ZoQ-t20.6
We built a tokenizer with it.
20.6
25.76
Training and Testing an Italian BERT - Transformers From Scratch #4
2021-07-06 13:00:03 UTC
https://youtu.be/35Pdoyi6ZoQ
35Pdoyi6ZoQ
UCv83tO5cePwHMt1952IVVHw
35Pdoyi6ZoQ-t23.72
And then we've set up our input pipeline
23.72
28.48
Training and Testing an Italian BERT - Transformers From Scratch #4
2021-07-06 13:00:03 UTC
https://youtu.be/35Pdoyi6ZoQ
35Pdoyi6ZoQ
UCv83tO5cePwHMt1952IVVHw
35Pdoyi6ZoQ-t25.76
ready to begin actually training our model, which
25.76
32.36
Training and Testing an Italian BERT - Transformers From Scratch #4
2021-07-06 13:00:03 UTC
https://youtu.be/35Pdoyi6ZoQ
35Pdoyi6ZoQ
UCv83tO5cePwHMt1952IVVHw
35Pdoyi6ZoQ-t28.48
is what we're going to cover in this video.
28.48
35.96
Training and Testing an Italian BERT - Transformers From Scratch #4
2021-07-06 13:00:03 UTC
https://youtu.be/35Pdoyi6ZoQ
35Pdoyi6ZoQ
UCv83tO5cePwHMt1952IVVHw
35Pdoyi6ZoQ-t32.36
So let's move over to the code.
32.36
39.56
Training and Testing an Italian BERT - Transformers From Scratch #4
2021-07-06 13:00:03 UTC
https://youtu.be/35Pdoyi6ZoQ
35Pdoyi6ZoQ
UCv83tO5cePwHMt1952IVVHw
35Pdoyi6ZoQ-t35.96
And we see here that we have essentially everything
35.96
40.48
Training and Testing an Italian BERT - Transformers From Scratch #4
2021-07-06 13:00:03 UTC
https://youtu.be/35Pdoyi6ZoQ
35Pdoyi6ZoQ
UCv83tO5cePwHMt1952IVVHw
35Pdoyi6ZoQ-t39.56
we've done so far.
39.56
48.8
Training and Testing an Italian BERT - Transformers From Scratch #4
2021-07-06 13:00:03 UTC
https://youtu.be/35Pdoyi6ZoQ
35Pdoyi6ZoQ
UCv83tO5cePwHMt1952IVVHw
35Pdoyi6ZoQ-t40.480000000000004
So we've built our input data, our input pipeline.
40.48
51.52
Training and Testing an Italian BERT - Transformers From Scratch #4
2021-07-06 13:00:03 UTC
https://youtu.be/35Pdoyi6ZoQ
35Pdoyi6ZoQ
UCv83tO5cePwHMt1952IVVHw
35Pdoyi6ZoQ-t48.8
And we're now at a point where we have a data loader,
48.8
54.04
Training and Testing an Italian BERT - Transformers From Scratch #4
2021-07-06 13:00:03 UTC
https://youtu.be/35Pdoyi6ZoQ
35Pdoyi6ZoQ
UCv83tO5cePwHMt1952IVVHw
35Pdoyi6ZoQ-t51.519999999999996
PyTorch data loader, ready.
51.52
56.4
Training and Testing an Italian BERT - Transformers From Scratch #4
2021-07-06 13:00:03 UTC
https://youtu.be/35Pdoyi6ZoQ
35Pdoyi6ZoQ
UCv83tO5cePwHMt1952IVVHw
35Pdoyi6ZoQ-t54.040000000000006
And we can begin training a model with it.
54.04
61.84
Training and Testing an Italian BERT - Transformers From Scratch #4
2021-07-06 13:00:03 UTC
https://youtu.be/35Pdoyi6ZoQ
35Pdoyi6ZoQ
UCv83tO5cePwHMt1952IVVHw
35Pdoyi6ZoQ-t56.4
So there are a few things to be aware of.
56.4
64.88
Training and Testing an Italian BERT - Transformers From Scratch #4
2021-07-06 13:00:03 UTC
https://youtu.be/35Pdoyi6ZoQ
35Pdoyi6ZoQ
UCv83tO5cePwHMt1952IVVHw
35Pdoyi6ZoQ-t61.839999999999996
So I mean, first, let's just have a quick look
61.84
67.28
Training and Testing an Italian BERT - Transformers From Scratch #4
2021-07-06 13:00:03 UTC
https://youtu.be/35Pdoyi6ZoQ
35Pdoyi6ZoQ
UCv83tO5cePwHMt1952IVVHw
35Pdoyi6ZoQ-t64.88
at the structure of our data.
64.88
72.36
Training and Testing an Italian BERT - Transformers From Scratch #4
2021-07-06 13:00:03 UTC
https://youtu.be/35Pdoyi6ZoQ
35Pdoyi6ZoQ
UCv83tO5cePwHMt1952IVVHw
35Pdoyi6ZoQ-t67.28
So when we're training a model for mass language modeling,
67.28
74.12
Training and Testing an Italian BERT - Transformers From Scratch #4
2021-07-06 13:00:03 UTC
https://youtu.be/35Pdoyi6ZoQ
35Pdoyi6ZoQ
UCv83tO5cePwHMt1952IVVHw
35Pdoyi6ZoQ-t72.36
we need a few tensors.
72.36
76.04
Training and Testing an Italian BERT - Transformers From Scratch #4
2021-07-06 13:00:03 UTC
https://youtu.be/35Pdoyi6ZoQ
35Pdoyi6ZoQ
UCv83tO5cePwHMt1952IVVHw
35Pdoyi6ZoQ-t74.12
We need three tensors.
74.12
80.32
Training and Testing an Italian BERT - Transformers From Scratch #4
2021-07-06 13:00:03 UTC
https://youtu.be/35Pdoyi6ZoQ
35Pdoyi6ZoQ
UCv83tO5cePwHMt1952IVVHw
35Pdoyi6ZoQ-t76.03999999999999
And this is for training Roberta, by the way, as well.
76.04
83.24
Training and Testing an Italian BERT - Transformers From Scratch #4
2021-07-06 13:00:03 UTC
https://youtu.be/35Pdoyi6ZoQ
35Pdoyi6ZoQ
UCv83tO5cePwHMt1952IVVHw
35Pdoyi6ZoQ-t80.32
Same thing with Bert as well.
80.32
88.76
Training and Testing an Italian BERT - Transformers From Scratch #4
2021-07-06 13:00:03 UTC
https://youtu.be/35Pdoyi6ZoQ
35Pdoyi6ZoQ
UCv83tO5cePwHMt1952IVVHw
35Pdoyi6ZoQ-t83.24
We have our input IDs, attention mask, and our labels.
83.24
94.2
Training and Testing an Italian BERT - Transformers From Scratch #4
2021-07-06 13:00:03 UTC
https://youtu.be/35Pdoyi6ZoQ
35Pdoyi6ZoQ
UCv83tO5cePwHMt1952IVVHw
35Pdoyi6ZoQ-t88.75999999999999
Our input IDs have roughly 15% of their values masked.
88.76
96.64
Training and Testing an Italian BERT - Transformers From Scratch #4
2021-07-06 13:00:03 UTC
https://youtu.be/35Pdoyi6ZoQ
35Pdoyi6ZoQ
UCv83tO5cePwHMt1952IVVHw
35Pdoyi6ZoQ-t94.19999999999999
So we can see that here we have these two tensors.
94.2
98.04
Training and Testing an Italian BERT - Transformers From Scratch #4
2021-07-06 13:00:03 UTC
https://youtu.be/35Pdoyi6ZoQ
35Pdoyi6ZoQ
UCv83tO5cePwHMt1952IVVHw
35Pdoyi6ZoQ-t96.64
These are the labels.
96.64
102.56
Training and Testing an Italian BERT - Transformers From Scratch #4
2021-07-06 13:00:03 UTC
https://youtu.be/35Pdoyi6ZoQ
35Pdoyi6ZoQ
UCv83tO5cePwHMt1952IVVHw
35Pdoyi6ZoQ-t98.03999999999999
And we have the real tokens in here, the token IDs.
98.04
105.2
Training and Testing an Italian BERT - Transformers From Scratch #4
2021-07-06 13:00:03 UTC
https://youtu.be/35Pdoyi6ZoQ
35Pdoyi6ZoQ
UCv83tO5cePwHMt1952IVVHw
35Pdoyi6ZoQ-t102.56
And then in our input IDs tensor,
102.56
108.68
Training and Testing an Italian BERT - Transformers From Scratch #4
2021-07-06 13:00:03 UTC
https://youtu.be/35Pdoyi6ZoQ
35Pdoyi6ZoQ
UCv83tO5cePwHMt1952IVVHw
35Pdoyi6ZoQ-t105.19999999999999
we have these being replaced with mask tokens,
105.2
110.52
Training and Testing an Italian BERT - Transformers From Scratch #4
2021-07-06 13:00:03 UTC
https://youtu.be/35Pdoyi6ZoQ
35Pdoyi6ZoQ
UCv83tO5cePwHMt1952IVVHw
35Pdoyi6ZoQ-t108.67999999999999
the number fours.
108.68
114.44
Training and Testing an Italian BERT - Transformers From Scratch #4
2021-07-06 13:00:03 UTC
https://youtu.be/35Pdoyi6ZoQ
35Pdoyi6ZoQ
UCv83tO5cePwHMt1952IVVHw
35Pdoyi6ZoQ-t110.52
So that's the structure of our input data.
110.52
119
Training and Testing an Italian BERT - Transformers From Scratch #4
2021-07-06 13:00:03 UTC
https://youtu.be/35Pdoyi6ZoQ
35Pdoyi6ZoQ
UCv83tO5cePwHMt1952IVVHw
35Pdoyi6ZoQ-t114.44
We've created a Torch data set from it
114.44
122.6
Training and Testing an Italian BERT - Transformers From Scratch #4
2021-07-06 13:00:03 UTC
https://youtu.be/35Pdoyi6ZoQ
35Pdoyi6ZoQ
UCv83tO5cePwHMt1952IVVHw
35Pdoyi6ZoQ-t119.0
and use that to create a Torch data loader.
119
125.48
Training and Testing an Italian BERT - Transformers From Scratch #4
2021-07-06 13:00:03 UTC
https://youtu.be/35Pdoyi6ZoQ
35Pdoyi6ZoQ
UCv83tO5cePwHMt1952IVVHw
35Pdoyi6ZoQ-t122.6
And with that, we can actually begin
122.6
127.76
Training and Testing an Italian BERT - Transformers From Scratch #4
2021-07-06 13:00:03 UTC
https://youtu.be/35Pdoyi6ZoQ
35Pdoyi6ZoQ
UCv83tO5cePwHMt1952IVVHw
35Pdoyi6ZoQ-t125.47999999999999
setting up our model for training.
125.48
131.72
Training and Testing an Italian BERT - Transformers From Scratch #4
2021-07-06 13:00:03 UTC
https://youtu.be/35Pdoyi6ZoQ
35Pdoyi6ZoQ
UCv83tO5cePwHMt1952IVVHw
35Pdoyi6ZoQ-t127.75999999999999
So there are a few things to that.
127.76
134.36
Training and Testing an Italian BERT - Transformers From Scratch #4
2021-07-06 13:00:03 UTC
https://youtu.be/35Pdoyi6ZoQ
35Pdoyi6ZoQ
UCv83tO5cePwHMt1952IVVHw
35Pdoyi6ZoQ-t131.72
We can't just begin training straight away.
131.72
136.32
Training and Testing an Italian BERT - Transformers From Scratch #4
2021-07-06 13:00:03 UTC
https://youtu.be/35Pdoyi6ZoQ
35Pdoyi6ZoQ
UCv83tO5cePwHMt1952IVVHw
35Pdoyi6ZoQ-t134.35999999999999
So the first thing that we need to do
134.36
140.56
Training and Testing an Italian BERT - Transformers From Scratch #4
2021-07-06 13:00:03 UTC
https://youtu.be/35Pdoyi6ZoQ
35Pdoyi6ZoQ
UCv83tO5cePwHMt1952IVVHw
35Pdoyi6ZoQ-t136.32
is create a Roberta config object.
136.32
144
Training and Testing an Italian BERT - Transformers From Scratch #4
2021-07-06 13:00:03 UTC
https://youtu.be/35Pdoyi6ZoQ
35Pdoyi6ZoQ
UCv83tO5cePwHMt1952IVVHw
35Pdoyi6ZoQ-t140.56
And the config object is something
140.56
146.76
Training and Testing an Italian BERT - Transformers From Scratch #4
2021-07-06 13:00:03 UTC
https://youtu.be/35Pdoyi6ZoQ
35Pdoyi6ZoQ
UCv83tO5cePwHMt1952IVVHw
35Pdoyi6ZoQ-t144.0
that we use when we're initializing a transformer
144
149.16
Training and Testing an Italian BERT - Transformers From Scratch #4
2021-07-06 13:00:03 UTC
https://youtu.be/35Pdoyi6ZoQ
35Pdoyi6ZoQ
UCv83tO5cePwHMt1952IVVHw
35Pdoyi6ZoQ-t146.76
from scratch in order to initialize it
146.76
152.32
Training and Testing an Italian BERT - Transformers From Scratch #4
2021-07-06 13:00:03 UTC
https://youtu.be/35Pdoyi6ZoQ
35Pdoyi6ZoQ
UCv83tO5cePwHMt1952IVVHw
35Pdoyi6ZoQ-t149.16
with a certain set of parameters.
149.16
153.76
Training and Testing an Italian BERT - Transformers From Scratch #4
2021-07-06 13:00:03 UTC
https://youtu.be/35Pdoyi6ZoQ
35Pdoyi6ZoQ
UCv83tO5cePwHMt1952IVVHw
35Pdoyi6ZoQ-t152.32
So we'll do that first.
152.32
159.96
Training and Testing an Italian BERT - Transformers From Scratch #4
2021-07-06 13:00:03 UTC
https://youtu.be/35Pdoyi6ZoQ
35Pdoyi6ZoQ
UCv83tO5cePwHMt1952IVVHw
35Pdoyi6ZoQ-t153.76
So we want from transformers import Roberta config.
153.76
163.08
Training and Testing an Italian BERT - Transformers From Scratch #4
2021-07-06 13:00:03 UTC
https://youtu.be/35Pdoyi6ZoQ
35Pdoyi6ZoQ
UCv83tO5cePwHMt1952IVVHw
35Pdoyi6ZoQ-t159.96
And to create that config object, we do this.
159.96
167.92
Training and Testing an Italian BERT - Transformers From Scratch #4
2021-07-06 13:00:03 UTC
https://youtu.be/35Pdoyi6ZoQ
35Pdoyi6ZoQ
UCv83tO5cePwHMt1952IVVHw
35Pdoyi6ZoQ-t163.08
So we do Roberta config.
163.08
172.56
Training and Testing an Italian BERT - Transformers From Scratch #4
2021-07-06 13:00:03 UTC
https://youtu.be/35Pdoyi6ZoQ
35Pdoyi6ZoQ
UCv83tO5cePwHMt1952IVVHw
35Pdoyi6ZoQ-t167.92000000000002
And then in here, we need to specify different parameters.
167.92
177.12
Training and Testing an Italian BERT - Transformers From Scratch #4
2021-07-06 13:00:03 UTC
https://youtu.be/35Pdoyi6ZoQ
35Pdoyi6ZoQ
UCv83tO5cePwHMt1952IVVHw
35Pdoyi6ZoQ-t172.56
Now, one of the main ones is the voc up size.
172.56
180.48
Training and Testing an Italian BERT - Transformers From Scratch #4
2021-07-06 13:00:03 UTC
https://youtu.be/35Pdoyi6ZoQ
35Pdoyi6ZoQ
UCv83tO5cePwHMt1952IVVHw
35Pdoyi6ZoQ-t177.12
Now, this needs to match to whichever voc up size
177.12
186.08
Training and Testing an Italian BERT - Transformers From Scratch #4
2021-07-06 13:00:03 UTC
https://youtu.be/35Pdoyi6ZoQ
35Pdoyi6ZoQ
UCv83tO5cePwHMt1952IVVHw
35Pdoyi6ZoQ-t180.48000000000002
we have already created in our tokenizer
180.48
187.88
Training and Testing an Italian BERT - Transformers From Scratch #4
2021-07-06 13:00:03 UTC
https://youtu.be/35Pdoyi6ZoQ
35Pdoyi6ZoQ
UCv83tO5cePwHMt1952IVVHw
35Pdoyi6ZoQ-t186.08
when we're initializing it.
186.08
191.48
Training and Testing an Italian BERT - Transformers From Scratch #4
2021-07-06 13:00:03 UTC
https://youtu.be/35Pdoyi6ZoQ
35Pdoyi6ZoQ
UCv83tO5cePwHMt1952IVVHw
35Pdoyi6ZoQ-t187.88
In our tokenizer, when building our tokenizer.
187.88
201.28
Training and Testing an Italian BERT - Transformers From Scratch #4
2021-07-06 13:00:03 UTC
https://youtu.be/35Pdoyi6ZoQ
35Pdoyi6ZoQ
UCv83tO5cePwHMt1952IVVHw
35Pdoyi6ZoQ-t191.48
So I mean, for me, if I go all the way up here to here,
191.48
203.92
Training and Testing an Italian BERT - Transformers From Scratch #4
2021-07-06 13:00:03 UTC
https://youtu.be/35Pdoyi6ZoQ
35Pdoyi6ZoQ
UCv83tO5cePwHMt1952IVVHw
35Pdoyi6ZoQ-t201.28
this is where I created the tokenizer.
201.28
206.36
Training and Testing an Italian BERT - Transformers From Scratch #4
2021-07-06 13:00:03 UTC
https://youtu.be/35Pdoyi6ZoQ
35Pdoyi6ZoQ
UCv83tO5cePwHMt1952IVVHw
35Pdoyi6ZoQ-t203.92
I can see, OK, it's this number here.
203.92
209.24
Training and Testing an Italian BERT - Transformers From Scratch #4
2021-07-06 13:00:03 UTC
https://youtu.be/35Pdoyi6ZoQ
35Pdoyi6ZoQ
UCv83tO5cePwHMt1952IVVHw
35Pdoyi6ZoQ-t206.35999999999999
So 30,522.
206.36
213.2
Training and Testing an Italian BERT - Transformers From Scratch #4
2021-07-06 13:00:03 UTC
https://youtu.be/35Pdoyi6ZoQ
35Pdoyi6ZoQ
UCv83tO5cePwHMt1952IVVHw
35Pdoyi6ZoQ-t209.24
So I'm going to set that.
209.24
218.64
Training and Testing an Italian BERT - Transformers From Scratch #4
2021-07-06 13:00:03 UTC
https://youtu.be/35Pdoyi6ZoQ
35Pdoyi6ZoQ
UCv83tO5cePwHMt1952IVVHw
35Pdoyi6ZoQ-t213.2
But if you don't have that, you can just write tokenizer voc
213.2
219.92
Training and Testing an Italian BERT - Transformers From Scratch #4
2021-07-06 13:00:03 UTC
https://youtu.be/35Pdoyi6ZoQ
35Pdoyi6ZoQ
UCv83tO5cePwHMt1952IVVHw
35Pdoyi6ZoQ-t218.64
up size.
218.64
222.2
Training and Testing an Italian BERT - Transformers From Scratch #4
2021-07-06 13:00:03 UTC
https://youtu.be/35Pdoyi6ZoQ
35Pdoyi6ZoQ
UCv83tO5cePwHMt1952IVVHw
35Pdoyi6ZoQ-t219.92
So here.
219.92
224.24
Training and Testing an Italian BERT - Transformers From Scratch #4
2021-07-06 13:00:03 UTC
https://youtu.be/35Pdoyi6ZoQ
35Pdoyi6ZoQ
UCv83tO5cePwHMt1952IVVHw
35Pdoyi6ZoQ-t222.2
And that will return your voc up size.
222.2
226.64
Training and Testing an Italian BERT - Transformers From Scratch #4
2021-07-06 13:00:03 UTC
https://youtu.be/35Pdoyi6ZoQ
35Pdoyi6ZoQ
UCv83tO5cePwHMt1952IVVHw
35Pdoyi6ZoQ-t224.23999999999998
So I mean, let's replace that.
224.24
228.88
Training and Testing an Italian BERT - Transformers From Scratch #4
2021-07-06 13:00:03 UTC
https://youtu.be/35Pdoyi6ZoQ
35Pdoyi6ZoQ
UCv83tO5cePwHMt1952IVVHw
35Pdoyi6ZoQ-t226.64
We'll do this.
226.64
235.64
Training and Testing an Italian BERT - Transformers From Scratch #4
2021-07-06 13:00:03 UTC
https://youtu.be/35Pdoyi6ZoQ
35Pdoyi6ZoQ
UCv83tO5cePwHMt1952IVVHw
35Pdoyi6ZoQ-t228.88
Now, as well as that, we want to also set this.
228.88
239.72
Training and Testing an Italian BERT - Transformers From Scratch #4
2021-07-06 13:00:03 UTC
https://youtu.be/35Pdoyi6ZoQ
35Pdoyi6ZoQ
UCv83tO5cePwHMt1952IVVHw
35Pdoyi6ZoQ-t235.64
So max position embedding.
235.64
246.68
Training and Testing an Italian BERT - Transformers From Scratch #4
2021-07-06 13:00:03 UTC
https://youtu.be/35Pdoyi6ZoQ
35Pdoyi6ZoQ
UCv83tO5cePwHMt1952IVVHw
35Pdoyi6ZoQ-t239.72
And this needs to be set to your max length plus two
239.72
247.28
Training and Testing an Italian BERT - Transformers From Scratch #4
2021-07-06 13:00:03 UTC
https://youtu.be/35Pdoyi6ZoQ
35Pdoyi6ZoQ
UCv83tO5cePwHMt1952IVVHw
35Pdoyi6ZoQ-t246.68
in this case.
246.68
251.12
Training and Testing an Italian BERT - Transformers From Scratch #4
2021-07-06 13:00:03 UTC
https://youtu.be/35Pdoyi6ZoQ
35Pdoyi6ZoQ
UCv83tO5cePwHMt1952IVVHw
35Pdoyi6ZoQ-t247.28
So max length is set up here.
247.28
253.24
Training and Testing an Italian BERT - Transformers From Scratch #4
2021-07-06 13:00:03 UTC
https://youtu.be/35Pdoyi6ZoQ
35Pdoyi6ZoQ
UCv83tO5cePwHMt1952IVVHw
35Pdoyi6ZoQ-t251.12
So where is it?
251.12
256
Training and Testing an Italian BERT - Transformers From Scratch #4
2021-07-06 13:00:03 UTC
https://youtu.be/35Pdoyi6ZoQ
35Pdoyi6ZoQ
UCv83tO5cePwHMt1952IVVHw
35Pdoyi6ZoQ-t253.24
Max length here, 512.
253.24
259.96
Training and Testing an Italian BERT - Transformers From Scratch #4
2021-07-06 13:00:03 UTC
https://youtu.be/35Pdoyi6ZoQ
35Pdoyi6ZoQ
UCv83tO5cePwHMt1952IVVHw
35Pdoyi6ZoQ-t256.0
Plus two because we have these added special tokens.
256
263.2
Training and Testing an Italian BERT - Transformers From Scratch #4
2021-07-06 13:00:03 UTC
https://youtu.be/35Pdoyi6ZoQ
35Pdoyi6ZoQ
UCv83tO5cePwHMt1952IVVHw
35Pdoyi6ZoQ-t259.96
If we don't do that, we'll end up with a index error
259.96
268.56
Training and Testing an Italian BERT - Transformers From Scratch #4
2021-07-06 13:00:03 UTC
https://youtu.be/35Pdoyi6ZoQ
35Pdoyi6ZoQ
UCv83tO5cePwHMt1952IVVHw
35Pdoyi6ZoQ-t263.2
because we're going beyond the embedding limits.
263.2
270.24
Training and Testing an Italian BERT - Transformers From Scratch #4
2021-07-06 13:00:03 UTC
https://youtu.be/35Pdoyi6ZoQ
35Pdoyi6ZoQ
UCv83tO5cePwHMt1952IVVHw
35Pdoyi6ZoQ-t268.56
Now we want our hidden size.
268.56
274.72
Training and Testing an Italian BERT - Transformers From Scratch #4
2021-07-06 13:00:03 UTC
https://youtu.be/35Pdoyi6ZoQ
35Pdoyi6ZoQ
UCv83tO5cePwHMt1952IVVHw
35Pdoyi6ZoQ-t270.24
So this is the size of the vectors
270.24
277.84
Training and Testing an Italian BERT - Transformers From Scratch #4
2021-07-06 13:00:03 UTC
https://youtu.be/35Pdoyi6ZoQ
35Pdoyi6ZoQ
UCv83tO5cePwHMt1952IVVHw
35Pdoyi6ZoQ-t274.72
that our embedding layers within Roberta will create.
274.72
284.16
Training and Testing an Italian BERT - Transformers From Scratch #4
2021-07-06 13:00:03 UTC
https://youtu.be/35Pdoyi6ZoQ
35Pdoyi6ZoQ
UCv83tO5cePwHMt1952IVVHw
35Pdoyi6ZoQ-t277.84
So each token, so we have 514 or 12 tokens.
277.84
289.24
Training and Testing an Italian BERT - Transformers From Scratch #4
2021-07-06 13:00:03 UTC
https://youtu.be/35Pdoyi6ZoQ
35Pdoyi6ZoQ
UCv83tO5cePwHMt1952IVVHw
35Pdoyi6ZoQ-t284.16
And each one of those will be signed a vector of size 768.
284.16
290.76
Training and Testing an Italian BERT - Transformers From Scratch #4
2021-07-06 13:00:03 UTC
https://youtu.be/35Pdoyi6ZoQ
35Pdoyi6ZoQ
UCv83tO5cePwHMt1952IVVHw
35Pdoyi6ZoQ-t289.24
This is the typical number.
289.24
296.08
Training and Testing an Italian BERT - Transformers From Scratch #4
2021-07-06 13:00:03 UTC
https://youtu.be/35Pdoyi6ZoQ
35Pdoyi6ZoQ
UCv83tO5cePwHMt1952IVVHw
35Pdoyi6ZoQ-t290.76
So that's the originally came from the BERT based model.
290.76
301.76
Training and Testing an Italian BERT - Transformers From Scratch #4
2021-07-06 13:00:03 UTC
https://youtu.be/35Pdoyi6ZoQ
35Pdoyi6ZoQ
UCv83tO5cePwHMt1952IVVHw
35Pdoyi6ZoQ-t296.08
Then we set up the architecture of the internals of the model.
296.08
304.88
Training and Testing an Italian BERT - Transformers From Scratch #4
2021-07-06 13:00:03 UTC
https://youtu.be/35Pdoyi6ZoQ
35Pdoyi6ZoQ
UCv83tO5cePwHMt1952IVVHw
35Pdoyi6ZoQ-t301.76
So we want the number of attention heads,
301.76
307.08
Training and Testing an Italian BERT - Transformers From Scratch #4
2021-07-06 13:00:03 UTC
https://youtu.be/35Pdoyi6ZoQ
35Pdoyi6ZoQ
UCv83tO5cePwHMt1952IVVHw
35Pdoyi6ZoQ-t304.88
which I'm going to set to 12.
304.88
314.4
Training and Testing an Italian BERT - Transformers From Scratch #4
2021-07-06 13:00:03 UTC
https://youtu.be/35Pdoyi6ZoQ
35Pdoyi6ZoQ
UCv83tO5cePwHMt1952IVVHw
35Pdoyi6ZoQ-t307.08
And also the number of hidden layers, which I...
307.08
318.72
Training and Testing an Italian BERT - Transformers From Scratch #4
2021-07-06 13:00:03 UTC
https://youtu.be/35Pdoyi6ZoQ
35Pdoyi6ZoQ
UCv83tO5cePwHMt1952IVVHw
35Pdoyi6ZoQ-t314.4
So the default for this is for Roberta, 12.
314.4
324.12
Training and Testing an Italian BERT - Transformers From Scratch #4
2021-07-06 13:00:03 UTC
https://youtu.be/35Pdoyi6ZoQ
35Pdoyi6ZoQ
UCv83tO5cePwHMt1952IVVHw
35Pdoyi6ZoQ-t318.71999999999997
But I'm going to go with six for the sake of keeping train
318.72
326.96
Training and Testing an Italian BERT - Transformers From Scratch #4
2021-07-06 13:00:03 UTC
https://youtu.be/35Pdoyi6ZoQ
35Pdoyi6ZoQ
UCv83tO5cePwHMt1952IVVHw
35Pdoyi6ZoQ-t324.12
times a little shorter.
324.12
334.28
Training and Testing an Italian BERT - Transformers From Scratch #4
2021-07-06 13:00:03 UTC
https://youtu.be/35Pdoyi6ZoQ
35Pdoyi6ZoQ
UCv83tO5cePwHMt1952IVVHw
35Pdoyi6ZoQ-t326.96
Now we also need to add type, vocab, size, which is just one.
326.96
337.4
Training and Testing an Italian BERT - Transformers From Scratch #4
2021-07-06 13:00:03 UTC
https://youtu.be/35Pdoyi6ZoQ
35Pdoyi6ZoQ
UCv83tO5cePwHMt1952IVVHw
35Pdoyi6ZoQ-t334.28000000000003
So that's the different token types that we have.
334.28
338.36
Training and Testing an Italian BERT - Transformers From Scratch #4
2021-07-06 13:00:03 UTC
https://youtu.be/35Pdoyi6ZoQ
35Pdoyi6ZoQ
UCv83tO5cePwHMt1952IVVHw
35Pdoyi6ZoQ-t337.4
We just have one.
337.4
341.76
Training and Testing an Italian BERT - Transformers From Scratch #4
2021-07-06 13:00:03 UTC
https://youtu.be/35Pdoyi6ZoQ
35Pdoyi6ZoQ
UCv83tO5cePwHMt1952IVVHw
35Pdoyi6ZoQ-t338.36
Don't need to worry about that.
338.36
347.92
Training and Testing an Italian BERT - Transformers From Scratch #4
2021-07-06 13:00:03 UTC
https://youtu.be/35Pdoyi6ZoQ
35Pdoyi6ZoQ
UCv83tO5cePwHMt1952IVVHw
35Pdoyi6ZoQ-t341.76
OK, so that's our configuration object ready.
341.76
352.04
Training and Testing an Italian BERT - Transformers From Scratch #4
2021-07-06 13:00:03 UTC
https://youtu.be/35Pdoyi6ZoQ
35Pdoyi6ZoQ
UCv83tO5cePwHMt1952IVVHw
35Pdoyi6ZoQ-t347.92
And we can import and initialize a Roberta model with that.
347.92
353.72
Training and Testing an Italian BERT - Transformers From Scratch #4
2021-07-06 13:00:03 UTC
https://youtu.be/35Pdoyi6ZoQ
35Pdoyi6ZoQ
UCv83tO5cePwHMt1952IVVHw
35Pdoyi6ZoQ-t352.04
So we went from transformers.
352.04
356.88
Training and Testing an Italian BERT - Transformers From Scratch #4
2021-07-06 13:00:03 UTC
https://youtu.be/35Pdoyi6ZoQ
35Pdoyi6ZoQ
UCv83tO5cePwHMt1952IVVHw
35Pdoyi6ZoQ-t353.72
This is kind of similar to what we usually do.
353.72
359.28
Training and Testing an Italian BERT - Transformers From Scratch #4
2021-07-06 13:00:03 UTC
https://youtu.be/35Pdoyi6ZoQ
35Pdoyi6ZoQ
UCv83tO5cePwHMt1952IVVHw
35Pdoyi6ZoQ-t356.88000000000005
Import Roberta.
356.88
361.72
YAML Metadata Warning: The task_categories "conversational" is not in the official list: text-classification, token-classification, table-question-answering, question-answering, zero-shot-classification, translation, summarization, feature-extraction, text-generation, text2text-generation, fill-mask, sentence-similarity, text-to-speech, text-to-audio, automatic-speech-recognition, audio-to-audio, audio-classification, voice-activity-detection, depth-estimation, image-classification, object-detection, image-segmentation, text-to-image, image-to-text, image-to-image, image-to-video, unconditional-image-generation, video-classification, reinforcement-learning, robotics, tabular-classification, tabular-regression, tabular-to-text, table-to-text, multiple-choice, text-retrieval, time-series-forecasting, text-to-video, image-text-to-text, visual-question-answering, document-question-answering, zero-shot-image-classification, graph-ml, mask-generation, zero-shot-object-detection, text-to-3d, image-to-3d, image-feature-extraction, other

The YouTube transcriptions dataset contains technical tutorials (currently from James Briggs, Daniel Bourke, and AI Coffee Break) transcribed using OpenAI's Whisper (large). Each row represents roughly a sentence-length chunk of text alongside the video URL and timestamp.

Note that each item in the dataset contains just a short chunk of text. For most use cases you will likely need to merge multiple rows to create more substantial chunks of text, if you need to do that, this code snippet will help:

from datasets import load_dataset

# first download the dataset
data = load_dataset(
    'jamescalam/youtube-transcriptions',
    split='train'
)

new_data = []  # this will store adjusted data

window = 6  # number of sentences to combine
stride = 3  # number of sentences to 'stride' over, used to create overlap

for i in range(0, len(data), stride):
    i_end = min(len(data)-1, i+window)
    if data[i]['title'] != data[i_end]['title']:
        # in this case we skip this entry as we have start/end of two videos
        continue
    # create larger text chunk
    text = ' '.join(data[i:i_end]['text'])
    # add to adjusted data list
    new_data.append({
        'start': data[i]['start'],
        'end': data[i_end]['end'],
        'title': data[i]['title'],
        'text': text,
        'id': data[i]['id'],
        'url': data[i]['url'],
        'published': data[i]['published']
    })
Downloads last month
247