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-t359.28000000000003
And we're doing this for massed LL.
359.28
363.36
Training and Testing an Italian BERT - Transformers From Scratch #4
2021-07-06 13:00:03 UTC
https://youtu.be/35Pdoyi6ZoQ
35Pdoyi6ZoQ
UCv83tO5cePwHMt1952IVVHw
35Pdoyi6ZoQ-t361.72
So MLM, right?
361.72
366.52
Training and Testing an Italian BERT - Transformers From Scratch #4
2021-07-06 13:00:03 UTC
https://youtu.be/35Pdoyi6ZoQ
35Pdoyi6ZoQ
UCv83tO5cePwHMt1952IVVHw
35Pdoyi6ZoQ-t363.36
So we're training using MLM.
363.36
369.56
Training and Testing an Italian BERT - Transformers From Scratch #4
2021-07-06 13:00:03 UTC
https://youtu.be/35Pdoyi6ZoQ
35Pdoyi6ZoQ
UCv83tO5cePwHMt1952IVVHw
35Pdoyi6ZoQ-t366.52000000000004
So we want Roberta for massed LL.
366.52
374.12
Training and Testing an Italian BERT - Transformers From Scratch #4
2021-07-06 13:00:03 UTC
https://youtu.be/35Pdoyi6ZoQ
35Pdoyi6ZoQ
UCv83tO5cePwHMt1952IVVHw
35Pdoyi6ZoQ-t369.56
And we initialize our model using that Roberta for massed LL
369.56
375.08
Training and Testing an Italian BERT - Transformers From Scratch #4
2021-07-06 13:00:03 UTC
https://youtu.be/35Pdoyi6ZoQ
35Pdoyi6ZoQ
UCv83tO5cePwHMt1952IVVHw
35Pdoyi6ZoQ-t374.12
object.
374.12
377.48
Training and Testing an Italian BERT - Transformers From Scratch #4
2021-07-06 13:00:03 UTC
https://youtu.be/35Pdoyi6ZoQ
35Pdoyi6ZoQ
UCv83tO5cePwHMt1952IVVHw
35Pdoyi6ZoQ-t375.08000000000004
And we just pass in our config.
375.08
379
Training and Testing an Italian BERT - Transformers From Scratch #4
2021-07-06 13:00:03 UTC
https://youtu.be/35Pdoyi6ZoQ
35Pdoyi6ZoQ
UCv83tO5cePwHMt1952IVVHw
35Pdoyi6ZoQ-t377.48
And this will...
377.48
382.92
Training and Testing an Italian BERT - Transformers From Scratch #4
2021-07-06 13:00:03 UTC
https://youtu.be/35Pdoyi6ZoQ
35Pdoyi6ZoQ
UCv83tO5cePwHMt1952IVVHw
35Pdoyi6ZoQ-t379.0
That's right there, initialize our Roberta model.
379
387.76
Training and Testing an Italian BERT - Transformers From Scratch #4
2021-07-06 13:00:03 UTC
https://youtu.be/35Pdoyi6ZoQ
35Pdoyi6ZoQ
UCv83tO5cePwHMt1952IVVHw
35Pdoyi6ZoQ-t382.92
So that's a plain Roberta model, randomly initialized weights,
382.92
389.6
Training and Testing an Italian BERT - Transformers From Scratch #4
2021-07-06 13:00:03 UTC
https://youtu.be/35Pdoyi6ZoQ
35Pdoyi6ZoQ
UCv83tO5cePwHMt1952IVVHw
35Pdoyi6ZoQ-t387.76
and so on.
387.76
393.52
Training and Testing an Italian BERT - Transformers From Scratch #4
2021-07-06 13:00:03 UTC
https://youtu.be/35Pdoyi6ZoQ
35Pdoyi6ZoQ
UCv83tO5cePwHMt1952IVVHw
35Pdoyi6ZoQ-t389.6
And now we can move on to setting up everything
389.6
395
Training and Testing an Italian BERT - Transformers From Scratch #4
2021-07-06 13:00:03 UTC
https://youtu.be/35Pdoyi6ZoQ
35Pdoyi6ZoQ
UCv83tO5cePwHMt1952IVVHw
35Pdoyi6ZoQ-t393.52000000000004
for training.
393.52
395.88
Training and Testing an Italian BERT - Transformers From Scratch #4
2021-07-06 13:00:03 UTC
https://youtu.be/35Pdoyi6ZoQ
35Pdoyi6ZoQ
UCv83tO5cePwHMt1952IVVHw
35Pdoyi6ZoQ-t395.0
So we have our model.
395
400.04
Training and Testing an Italian BERT - Transformers From Scratch #4
2021-07-06 13:00:03 UTC
https://youtu.be/35Pdoyi6ZoQ
35Pdoyi6ZoQ
UCv83tO5cePwHMt1952IVVHw
35Pdoyi6ZoQ-t395.88
Now we need to prepare a few things before we train it.
395.88
403.04
Training and Testing an Italian BERT - Transformers From Scratch #4
2021-07-06 13:00:03 UTC
https://youtu.be/35Pdoyi6ZoQ
35Pdoyi6ZoQ
UCv83tO5cePwHMt1952IVVHw
35Pdoyi6ZoQ-t400.04
First thing is we need to decide which device we're
400.04
403.92
Training and Testing an Italian BERT - Transformers From Scratch #4
2021-07-06 13:00:03 UTC
https://youtu.be/35Pdoyi6ZoQ
35Pdoyi6ZoQ
UCv83tO5cePwHMt1952IVVHw
35Pdoyi6ZoQ-t403.04
going to be training on.
403.04
407.44
Training and Testing an Italian BERT - Transformers From Scratch #4
2021-07-06 13:00:03 UTC
https://youtu.be/35Pdoyi6ZoQ
35Pdoyi6ZoQ
UCv83tO5cePwHMt1952IVVHw
35Pdoyi6ZoQ-t403.92
So whether that's CPU or a CUDA-enabled GPU.
403.92
411.24
Training and Testing an Italian BERT - Transformers From Scratch #4
2021-07-06 13:00:03 UTC
https://youtu.be/35Pdoyi6ZoQ
35Pdoyi6ZoQ
UCv83tO5cePwHMt1952IVVHw
35Pdoyi6ZoQ-t407.44
And to figure out if we have that, we write...
407.44
416.92
Training and Testing an Italian BERT - Transformers From Scratch #4
2021-07-06 13:00:03 UTC
https://youtu.be/35Pdoyi6ZoQ
35Pdoyi6ZoQ
UCv83tO5cePwHMt1952IVVHw
35Pdoyi6ZoQ-t411.24
Well, we can write torch CUDA is available.
411.24
417.76
Training and Testing an Italian BERT - Transformers From Scratch #4
2021-07-06 13:00:03 UTC
https://youtu.be/35Pdoyi6ZoQ
35Pdoyi6ZoQ
UCv83tO5cePwHMt1952IVVHw
35Pdoyi6ZoQ-t416.92
So write this.
416.92
419.84
Training and Testing an Italian BERT - Transformers From Scratch #4
2021-07-06 13:00:03 UTC
https://youtu.be/35Pdoyi6ZoQ
35Pdoyi6ZoQ
UCv83tO5cePwHMt1952IVVHw
35Pdoyi6ZoQ-t417.76
And for me, it is.
417.76
423.2
Training and Testing an Italian BERT - Transformers From Scratch #4
2021-07-06 13:00:03 UTC
https://youtu.be/35Pdoyi6ZoQ
35Pdoyi6ZoQ
UCv83tO5cePwHMt1952IVVHw
35Pdoyi6ZoQ-t419.84000000000003
So the typical way that you would decide whether you're
419.84
428.28
Training and Testing an Italian BERT - Transformers From Scratch #4
2021-07-06 13:00:03 UTC
https://youtu.be/35Pdoyi6ZoQ
35Pdoyi6ZoQ
UCv83tO5cePwHMt1952IVVHw
35Pdoyi6ZoQ-t423.2
using CUDA or CPU or the typical line of code that will decide it
423.2
434.52
Training and Testing an Italian BERT - Transformers From Scratch #4
2021-07-06 13:00:03 UTC
https://youtu.be/35Pdoyi6ZoQ
35Pdoyi6ZoQ
UCv83tO5cePwHMt1952IVVHw
35Pdoyi6ZoQ-t428.28000000000003
for you is you write device and you do torch.CUDA,
428.28
437.88
Training and Testing an Italian BERT - Transformers From Scratch #4
2021-07-06 13:00:03 UTC
https://youtu.be/35Pdoyi6ZoQ
35Pdoyi6ZoQ
UCv83tO5cePwHMt1952IVVHw
35Pdoyi6ZoQ-t434.52
or torch device, sorry.
434.52
440.56
Training and Testing an Italian BERT - Transformers From Scratch #4
2021-07-06 13:00:03 UTC
https://youtu.be/35Pdoyi6ZoQ
35Pdoyi6ZoQ
UCv83tO5cePwHMt1952IVVHw
35Pdoyi6ZoQ-t437.88
And then you write CUDA inside here.
437.88
449.12
Training and Testing an Italian BERT - Transformers From Scratch #4
2021-07-06 13:00:03 UTC
https://youtu.be/35Pdoyi6ZoQ
35Pdoyi6ZoQ
UCv83tO5cePwHMt1952IVVHw
35Pdoyi6ZoQ-t440.56
If it's available, otherwise we are going to use torch device CPU.
440.56
453.4
Training and Testing an Italian BERT - Transformers From Scratch #4
2021-07-06 13:00:03 UTC
https://youtu.be/35Pdoyi6ZoQ
35Pdoyi6ZoQ
UCv83tO5cePwHMt1952IVVHw
35Pdoyi6ZoQ-t449.12
Now, CPU takes...
449.12
456.76
Training and Testing an Italian BERT - Transformers From Scratch #4
2021-07-06 13:00:03 UTC
https://youtu.be/35Pdoyi6ZoQ
35Pdoyi6ZoQ
UCv83tO5cePwHMt1952IVVHw
35Pdoyi6ZoQ-t453.4
Yeah, it's just takes a really long time.
453.4
463.72
Training and Testing an Italian BERT - Transformers From Scratch #4
2021-07-06 13:00:03 UTC
https://youtu.be/35Pdoyi6ZoQ
35Pdoyi6ZoQ
UCv83tO5cePwHMt1952IVVHw
35Pdoyi6ZoQ-t456.76
So if you are using CPU, you have to leave it overnight,
456.76
464.72
Training and Testing an Italian BERT - Transformers From Scratch #4
2021-07-06 13:00:03 UTC
https://youtu.be/35Pdoyi6ZoQ
35Pdoyi6ZoQ
UCv83tO5cePwHMt1952IVVHw
35Pdoyi6ZoQ-t463.72
for sure.
463.72
466.32
Training and Testing an Italian BERT - Transformers From Scratch #4
2021-07-06 13:00:03 UTC
https://youtu.be/35Pdoyi6ZoQ
35Pdoyi6ZoQ
UCv83tO5cePwHMt1952IVVHw
35Pdoyi6ZoQ-t464.72
Maybe even longer.
464.72
471.12
Training and Testing an Italian BERT - Transformers From Scratch #4
2021-07-06 13:00:03 UTC
https://youtu.be/35Pdoyi6ZoQ
35Pdoyi6ZoQ
UCv83tO5cePwHMt1952IVVHw
35Pdoyi6ZoQ-t466.32
Even if it's just like a little bit of data, it takes so long.
466.32
474.28
Training and Testing an Italian BERT - Transformers From Scratch #4
2021-07-06 13:00:03 UTC
https://youtu.be/35Pdoyi6ZoQ
35Pdoyi6ZoQ
UCv83tO5cePwHMt1952IVVHw
35Pdoyi6ZoQ-t471.12
So but hopefully you have a GPU.
471.12
477.24
Training and Testing an Italian BERT - Transformers From Scratch #4
2021-07-06 13:00:03 UTC
https://youtu.be/35Pdoyi6ZoQ
35Pdoyi6ZoQ
UCv83tO5cePwHMt1952IVVHw
35Pdoyi6ZoQ-t474.28
If not, just you're going to have to be patient.
474.28
478.76
Training and Testing an Italian BERT - Transformers From Scratch #4
2021-07-06 13:00:03 UTC
https://youtu.be/35Pdoyi6ZoQ
35Pdoyi6ZoQ
UCv83tO5cePwHMt1952IVVHw
35Pdoyi6ZoQ-t477.24
That's all.
477.24
482.88
Training and Testing an Italian BERT - Transformers From Scratch #4
2021-07-06 13:00:03 UTC
https://youtu.be/35Pdoyi6ZoQ
35Pdoyi6ZoQ
UCv83tO5cePwHMt1952IVVHw
35Pdoyi6ZoQ-t478.76
Or if you could maybe try and use Google Colab,
478.76
485.92
Training and Testing an Italian BERT - Transformers From Scratch #4
2021-07-06 13:00:03 UTC
https://youtu.be/35Pdoyi6ZoQ
35Pdoyi6ZoQ
UCv83tO5cePwHMt1952IVVHw
35Pdoyi6ZoQ-t482.88
but you have to use a premium version because otherwise it's
482.88
489.52
Training and Testing an Italian BERT - Transformers From Scratch #4
2021-07-06 13:00:03 UTC
https://youtu.be/35Pdoyi6ZoQ
35Pdoyi6ZoQ
UCv83tO5cePwHMt1952IVVHw
35Pdoyi6ZoQ-t485.92
just going to shut off after like an hour or two.
485.92
490.04
Training and Testing an Italian BERT - Transformers From Scratch #4
2021-07-06 13:00:03 UTC
https://youtu.be/35Pdoyi6ZoQ
35Pdoyi6ZoQ
UCv83tO5cePwHMt1952IVVHw
35Pdoyi6ZoQ-t489.52
I don't know.
489.52
490.96
Training and Testing an Italian BERT - Transformers From Scratch #4
2021-07-06 13:00:03 UTC
https://youtu.be/35Pdoyi6ZoQ
35Pdoyi6ZoQ
UCv83tO5cePwHMt1952IVVHw
35Pdoyi6ZoQ-t490.03999999999996
I don't really use it.
490.04
496.04
Training and Testing an Italian BERT - Transformers From Scratch #4
2021-07-06 13:00:03 UTC
https://youtu.be/35Pdoyi6ZoQ
35Pdoyi6ZoQ
UCv83tO5cePwHMt1952IVVHw
35Pdoyi6ZoQ-t490.96
So I don't know how long it will train for before just deciding
490.96
496.88
Training and Testing an Italian BERT - Transformers From Scratch #4
2021-07-06 13:00:03 UTC
https://youtu.be/35Pdoyi6ZoQ
35Pdoyi6ZoQ
UCv83tO5cePwHMt1952IVVHw
35Pdoyi6ZoQ-t496.04
that it's done.
496.04
499.64
Training and Testing an Italian BERT - Transformers From Scratch #4
2021-07-06 13:00:03 UTC
https://youtu.be/35Pdoyi6ZoQ
35Pdoyi6ZoQ
UCv83tO5cePwHMt1952IVVHw
35Pdoyi6ZoQ-t496.88
And the GPU is also not that good anyway.
496.88
503.08
Training and Testing an Italian BERT - Transformers From Scratch #4
2021-07-06 13:00:03 UTC
https://youtu.be/35Pdoyi6ZoQ
35Pdoyi6ZoQ
UCv83tO5cePwHMt1952IVVHw
35Pdoyi6ZoQ-t499.64000000000004
So, yeah.
499.64
506.84
Training and Testing an Italian BERT - Transformers From Scratch #4
2021-07-06 13:00:03 UTC
https://youtu.be/35Pdoyi6ZoQ
35Pdoyi6ZoQ
UCv83tO5cePwHMt1952IVVHw
35Pdoyi6ZoQ-t503.08000000000004
However, you can do it.
503.08
511.96
Training and Testing an Italian BERT - Transformers From Scratch #4
2021-07-06 13:00:03 UTC
https://youtu.be/35Pdoyi6ZoQ
35Pdoyi6ZoQ
UCv83tO5cePwHMt1952IVVHw
35Pdoyi6ZoQ-t506.84000000000003
And then after that, we want to move our model to our device.
506.84
516.12
Training and Testing an Italian BERT - Transformers From Scratch #4
2021-07-06 13:00:03 UTC
https://youtu.be/35Pdoyi6ZoQ
35Pdoyi6ZoQ
UCv83tO5cePwHMt1952IVVHw
35Pdoyi6ZoQ-t511.96000000000004
So whether it's GPU or CPU, we move over there.
511.96
519.12
Training and Testing an Italian BERT - Transformers From Scratch #4
2021-07-06 13:00:03 UTC
https://youtu.be/35Pdoyi6ZoQ
35Pdoyi6ZoQ
UCv83tO5cePwHMt1952IVVHw
35Pdoyi6ZoQ-t516.12
We're going to get a really big output now.
516.12
521.16
Training and Testing an Italian BERT - Transformers From Scratch #4
2021-07-06 13:00:03 UTC
https://youtu.be/35Pdoyi6ZoQ
35Pdoyi6ZoQ
UCv83tO5cePwHMt1952IVVHw
35Pdoyi6ZoQ-t519.12
So it's just our model.
519.12
523.08
Training and Testing an Italian BERT - Transformers From Scratch #4
2021-07-06 13:00:03 UTC
https://youtu.be/35Pdoyi6ZoQ
35Pdoyi6ZoQ
UCv83tO5cePwHMt1952IVVHw
35Pdoyi6ZoQ-t521.16
So this is like the structure of our model.
521.16
524.84
Training and Testing an Italian BERT - Transformers From Scratch #4
2021-07-06 13:00:03 UTC
https://youtu.be/35Pdoyi6ZoQ
35Pdoyi6ZoQ
UCv83tO5cePwHMt1952IVVHw
35Pdoyi6ZoQ-t523.08
So we can see a few interesting things.
523.08
527.52
Training and Testing an Italian BERT - Transformers From Scratch #4
2021-07-06 13:00:03 UTC
https://youtu.be/35Pdoyi6ZoQ
35Pdoyi6ZoQ
UCv83tO5cePwHMt1952IVVHw
35Pdoyi6ZoQ-t524.84
We've got Roberta for MLM.
524.84
529.56
Training and Testing an Italian BERT - Transformers From Scratch #4
2021-07-06 13:00:03 UTC
https://youtu.be/35Pdoyi6ZoQ
35Pdoyi6ZoQ
UCv83tO5cePwHMt1952IVVHw
35Pdoyi6ZoQ-t527.52
We have the Roberta model.
527.52
531.52
Training and Testing an Italian BERT - Transformers From Scratch #4
2021-07-06 13:00:03 UTC
https://youtu.be/35Pdoyi6ZoQ
35Pdoyi6ZoQ
UCv83tO5cePwHMt1952IVVHw
35Pdoyi6ZoQ-t529.5600000000001
And then inside that we have our embeddings.
529.56
534.4
Training and Testing an Italian BERT - Transformers From Scratch #4
2021-07-06 13:00:03 UTC
https://youtu.be/35Pdoyi6ZoQ
35Pdoyi6ZoQ
UCv83tO5cePwHMt1952IVVHw
35Pdoyi6ZoQ-t531.52
And then we have our 12.
531.52
535.28
Training and Testing an Italian BERT - Transformers From Scratch #4
2021-07-06 13:00:03 UTC
https://youtu.be/35Pdoyi6ZoQ
35Pdoyi6ZoQ
UCv83tO5cePwHMt1952IVVHw
35Pdoyi6ZoQ-t534.4
Did I say 12?
534.4
537.16
Training and Testing an Italian BERT - Transformers From Scratch #4
2021-07-06 13:00:03 UTC
https://youtu.be/35Pdoyi6ZoQ
35Pdoyi6ZoQ
UCv83tO5cePwHMt1952IVVHw
35Pdoyi6ZoQ-t535.2800000000001
I think it was six.
535.28
538.96
Training and Testing an Italian BERT - Transformers From Scratch #4
2021-07-06 13:00:03 UTC
https://youtu.be/35Pdoyi6ZoQ
35Pdoyi6ZoQ
UCv83tO5cePwHMt1952IVVHw
35Pdoyi6ZoQ-t537.1600000000001
Six encoders should be.
537.16
539.44
Training and Testing an Italian BERT - Transformers From Scratch #4
2021-07-06 13:00:03 UTC
https://youtu.be/35Pdoyi6ZoQ
35Pdoyi6ZoQ
UCv83tO5cePwHMt1952IVVHw
35Pdoyi6ZoQ-t538.96
Yeah.
538.96
541.48
Training and Testing an Italian BERT - Transformers From Scratch #4
2021-07-06 13:00:03 UTC
https://youtu.be/35Pdoyi6ZoQ
35Pdoyi6ZoQ
UCv83tO5cePwHMt1952IVVHw
35Pdoyi6ZoQ-t539.44
So it goes from 0 to 5.
539.44
542.52
Training and Testing an Italian BERT - Transformers From Scratch #4
2021-07-06 13:00:03 UTC
https://youtu.be/35Pdoyi6ZoQ
35Pdoyi6ZoQ
UCv83tO5cePwHMt1952IVVHw
35Pdoyi6ZoQ-t541.48
It's our six.
541.48
544.8
Training and Testing an Italian BERT - Transformers From Scratch #4
2021-07-06 13:00:03 UTC
https://youtu.be/35Pdoyi6ZoQ
35Pdoyi6ZoQ
UCv83tO5cePwHMt1952IVVHw
35Pdoyi6ZoQ-t542.52
And then we have the outputs here.
542.52
547.92
Training and Testing an Italian BERT - Transformers From Scratch #4
2021-07-06 13:00:03 UTC
https://youtu.be/35Pdoyi6ZoQ
35Pdoyi6ZoQ
UCv83tO5cePwHMt1952IVVHw
35Pdoyi6ZoQ-t544.8000000000001
And then our final bit, which is the language modeling
544.8
549.6
Training and Testing an Italian BERT - Transformers From Scratch #4
2021-07-06 13:00:03 UTC
https://youtu.be/35Pdoyi6ZoQ
35Pdoyi6ZoQ
UCv83tO5cePwHMt1952IVVHw
35Pdoyi6ZoQ-t547.9200000000001
head, the MLM head.
547.92
551.16
Training and Testing an Italian BERT - Transformers From Scratch #4
2021-07-06 13:00:03 UTC
https://youtu.be/35Pdoyi6ZoQ
35Pdoyi6ZoQ
UCv83tO5cePwHMt1952IVVHw
35Pdoyi6ZoQ-t549.6
So that's cool.
549.6
554
Training and Testing an Italian BERT - Transformers From Scratch #4
2021-07-06 13:00:03 UTC
https://youtu.be/35Pdoyi6ZoQ
35Pdoyi6ZoQ
UCv83tO5cePwHMt1952IVVHw
35Pdoyi6ZoQ-t551.1600000000001
Now we need our optimizer.
551.16
559.88
Training and Testing an Italian BERT - Transformers From Scratch #4
2021-07-06 13:00:03 UTC
https://youtu.be/35Pdoyi6ZoQ
35Pdoyi6ZoQ
UCv83tO5cePwHMt1952IVVHw
35Pdoyi6ZoQ-t554.0
So from transformers, import AdamW,
554
562.8
Training and Testing an Italian BERT - Transformers From Scratch #4
2021-07-06 13:00:03 UTC
https://youtu.be/35Pdoyi6ZoQ
35Pdoyi6ZoQ
UCv83tO5cePwHMt1952IVVHw
35Pdoyi6ZoQ-t559.88
which is Adam with weight and decay.
559.88
566.2
Training and Testing an Italian BERT - Transformers From Scratch #4
2021-07-06 13:00:03 UTC
https://youtu.be/35Pdoyi6ZoQ
35Pdoyi6ZoQ
UCv83tO5cePwHMt1952IVVHw
35Pdoyi6ZoQ-t562.8
And what we're going to do is just
562.8
568.6
Training and Testing an Italian BERT - Transformers From Scratch #4
2021-07-06 13:00:03 UTC
https://youtu.be/35Pdoyi6ZoQ
35Pdoyi6ZoQ
UCv83tO5cePwHMt1952IVVHw
35Pdoyi6ZoQ-t566.2
going to activate the training mode of our model.
566.2
572.68
Training and Testing an Italian BERT - Transformers From Scratch #4
2021-07-06 13:00:03 UTC
https://youtu.be/35Pdoyi6ZoQ
35Pdoyi6ZoQ
UCv83tO5cePwHMt1952IVVHw
35Pdoyi6ZoQ-t568.6
It's going to give us loads of output again.
568.6
574.64
Training and Testing an Italian BERT - Transformers From Scratch #4
2021-07-06 13:00:03 UTC
https://youtu.be/35Pdoyi6ZoQ
35Pdoyi6ZoQ
UCv83tO5cePwHMt1952IVVHw
35Pdoyi6ZoQ-t572.68
So just, yeah.
572.68
575.24
Training and Testing an Italian BERT - Transformers From Scratch #4
2021-07-06 13:00:03 UTC
https://youtu.be/35Pdoyi6ZoQ
35Pdoyi6ZoQ
UCv83tO5cePwHMt1952IVVHw
35Pdoyi6ZoQ-t574.64
You know what?
574.64
578.64
Training and Testing an Italian BERT - Transformers From Scratch #4
2021-07-06 13:00:03 UTC
https://youtu.be/35Pdoyi6ZoQ
35Pdoyi6ZoQ
UCv83tO5cePwHMt1952IVVHw
35Pdoyi6ZoQ-t575.24
Maybe I can just, let's just remove that.
575.24
579.24
Training and Testing an Italian BERT - Transformers From Scratch #4
2021-07-06 13:00:03 UTC
https://youtu.be/35Pdoyi6ZoQ
35Pdoyi6ZoQ
UCv83tO5cePwHMt1952IVVHw
35Pdoyi6ZoQ-t578.64
There we go.
578.64
580.36
Training and Testing an Italian BERT - Transformers From Scratch #4
2021-07-06 13:00:03 UTC
https://youtu.be/35Pdoyi6ZoQ
35Pdoyi6ZoQ
UCv83tO5cePwHMt1952IVVHw
35Pdoyi6ZoQ-t579.24
Easier.
579.24
585.12
Training and Testing an Italian BERT - Transformers From Scratch #4
2021-07-06 13:00:03 UTC
https://youtu.be/35Pdoyi6ZoQ
35Pdoyi6ZoQ
UCv83tO5cePwHMt1952IVVHw
35Pdoyi6ZoQ-t580.36
And then our optimizer is going to be AdamW.
580.36
589.28
Training and Testing an Italian BERT - Transformers From Scratch #4
2021-07-06 13:00:03 UTC
https://youtu.be/35Pdoyi6ZoQ
35Pdoyi6ZoQ
UCv83tO5cePwHMt1952IVVHw
35Pdoyi6ZoQ-t585.12
We need to pass in our model parameters.
585.12
590.48
Training and Testing an Italian BERT - Transformers From Scratch #4
2021-07-06 13:00:03 UTC
https://youtu.be/35Pdoyi6ZoQ
35Pdoyi6ZoQ
UCv83tO5cePwHMt1952IVVHw
35Pdoyi6ZoQ-t589.28
And we need a learning rate.
589.28
595.08
Training and Testing an Italian BERT - Transformers From Scratch #4
2021-07-06 13:00:03 UTC
https://youtu.be/35Pdoyi6ZoQ
35Pdoyi6ZoQ
UCv83tO5cePwHMt1952IVVHw
35Pdoyi6ZoQ-t590.48
So from, I mean, I don't usually use Roberta.
590.48
601.48
Training and Testing an Italian BERT - Transformers From Scratch #4
2021-07-06 13:00:03 UTC
https://youtu.be/35Pdoyi6ZoQ
35Pdoyi6ZoQ
UCv83tO5cePwHMt1952IVVHw
35Pdoyi6ZoQ-t595.08
But looking online, this looks like a reasonable learning
595.08
601.96
Training and Testing an Italian BERT - Transformers From Scratch #4
2021-07-06 13:00:03 UTC
https://youtu.be/35Pdoyi6ZoQ
35Pdoyi6ZoQ
UCv83tO5cePwHMt1952IVVHw
35Pdoyi6ZoQ-t601.48
rate.
601.48
605.4
Training and Testing an Italian BERT - Transformers From Scratch #4
2021-07-06 13:00:03 UTC
https://youtu.be/35Pdoyi6ZoQ
35Pdoyi6ZoQ
UCv83tO5cePwHMt1952IVVHw
35Pdoyi6ZoQ-t601.96
I think you can go from sort of here to, I think,
601.96
608.92
Training and Testing an Italian BERT - Transformers From Scratch #4
2021-07-06 13:00:03 UTC
https://youtu.be/35Pdoyi6ZoQ
35Pdoyi6ZoQ
UCv83tO5cePwHMt1952IVVHw
35Pdoyi6ZoQ-t605.4
from what I remember down to like here.
605.4
610.28
Training and Testing an Italian BERT - Transformers From Scratch #4
2021-07-06 13:00:03 UTC
https://youtu.be/35Pdoyi6ZoQ
35Pdoyi6ZoQ
UCv83tO5cePwHMt1952IVVHw
35Pdoyi6ZoQ-t608.92
That's the sort of typical range.
608.92
612.64
Training and Testing an Italian BERT - Transformers From Scratch #4
2021-07-06 13:00:03 UTC
https://youtu.be/35Pdoyi6ZoQ
35Pdoyi6ZoQ
UCv83tO5cePwHMt1952IVVHw
35Pdoyi6ZoQ-t610.28
But obviously, it's going to depend on how much data
610.28
613.2
Training and Testing an Italian BERT - Transformers From Scratch #4
2021-07-06 13:00:03 UTC
https://youtu.be/35Pdoyi6ZoQ
35Pdoyi6ZoQ
UCv83tO5cePwHMt1952IVVHw
35Pdoyi6ZoQ-t612.64
you have.
612.64
614.84
Training and Testing an Italian BERT - Transformers From Scratch #4
2021-07-06 13:00:03 UTC
https://youtu.be/35Pdoyi6ZoQ
35Pdoyi6ZoQ
UCv83tO5cePwHMt1952IVVHw
35Pdoyi6ZoQ-t613.1999999999999
And we don't want to do that.
613.2
617.76
Training and Testing an Italian BERT - Transformers From Scratch #4
2021-07-06 13:00:03 UTC
https://youtu.be/35Pdoyi6ZoQ
35Pdoyi6ZoQ
UCv83tO5cePwHMt1952IVVHw
35Pdoyi6ZoQ-t614.8399999999999
How much data you have and loads of different things, right?
614.84
623.32
Training and Testing an Italian BERT - Transformers From Scratch #4
2021-07-06 13:00:03 UTC
https://youtu.be/35Pdoyi6ZoQ
35Pdoyi6ZoQ
UCv83tO5cePwHMt1952IVVHw
35Pdoyi6ZoQ-t617.76
So that's what I'm going to go with.
617.76
627.8
Training and Testing an Italian BERT - Transformers From Scratch #4
2021-07-06 13:00:03 UTC
https://youtu.be/35Pdoyi6ZoQ
35Pdoyi6ZoQ
UCv83tO5cePwHMt1952IVVHw
35Pdoyi6ZoQ-t623.3199999999999
And that should be pretty much it.
623.32
630.44
Training and Testing an Italian BERT - Transformers From Scratch #4
2021-07-06 13:00:03 UTC
https://youtu.be/35Pdoyi6ZoQ
35Pdoyi6ZoQ
UCv83tO5cePwHMt1952IVVHw
35Pdoyi6ZoQ-t627.8
So that's our setup.
627.8
633.08
Training and Testing an Italian BERT - Transformers From Scratch #4
2021-07-06 13:00:03 UTC
https://youtu.be/35Pdoyi6ZoQ
35Pdoyi6ZoQ
UCv83tO5cePwHMt1952IVVHw
35Pdoyi6ZoQ-t630.4399999999999
Now we're just going to create our training loop.
630.44
638.6
Training and Testing an Italian BERT - Transformers From Scratch #4
2021-07-06 13:00:03 UTC
https://youtu.be/35Pdoyi6ZoQ
35Pdoyi6ZoQ
UCv83tO5cePwHMt1952IVVHw
35Pdoyi6ZoQ-t633.0799999999999
Now for the training loop, we want to import TQDM.
633.08
641.76
Training and Testing an Italian BERT - Transformers From Scratch #4
2021-07-06 13:00:03 UTC
https://youtu.be/35Pdoyi6ZoQ
35Pdoyi6ZoQ
UCv83tO5cePwHMt1952IVVHw
35Pdoyi6ZoQ-t638.6
So we can see how far through we are.
638.6
647.6
Training and Testing an Italian BERT - Transformers From Scratch #4
2021-07-06 13:00:03 UTC
https://youtu.be/35Pdoyi6ZoQ
35Pdoyi6ZoQ
UCv83tO5cePwHMt1952IVVHw
35Pdoyi6ZoQ-t645.32
We're going to train for two epochs.
645.32
652.08
Training and Testing an Italian BERT - Transformers From Scratch #4
2021-07-06 13:00:03 UTC
https://youtu.be/35Pdoyi6ZoQ
35Pdoyi6ZoQ
UCv83tO5cePwHMt1952IVVHw
35Pdoyi6ZoQ-t647.6
And we're going to initialize our loop object using TQDM.
647.6
656.12
Training and Testing an Italian BERT - Transformers From Scratch #4
2021-07-06 13:00:03 UTC
https://youtu.be/35Pdoyi6ZoQ
35Pdoyi6ZoQ
UCv83tO5cePwHMt1952IVVHw
35Pdoyi6ZoQ-t652.08
So TQDM.
652.08
658.32