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 BERT #5 - Training With BertForPretraining
|
2021-06-15 15:00:19 UTC
|
https://youtu.be/IC9FaVPKlYc
|
IC9FaVPKlYc
|
UCv83tO5cePwHMt1952IVVHw
|
IC9FaVPKlYc-t374.4
|
sentence is not equal to empty, then once we're there, what we want to do is we want to get the
| 374.4 | 394 |
Training BERT #5 - Training With BertForPretraining
|
2021-06-15 15:00:19 UTC
|
https://youtu.be/IC9FaVPKlYc
|
IC9FaVPKlYc
|
UCv83tO5cePwHMt1952IVVHw
|
IC9FaVPKlYc-t385.76
|
number of sentences within each sentence or sentences variable. So just get length.
| 385.76 | 398.88 |
Training BERT #5 - Training With BertForPretraining
|
2021-06-15 15:00:19 UTC
|
https://youtu.be/IC9FaVPKlYc
|
IC9FaVPKlYc
|
UCv83tO5cePwHMt1952IVVHw
|
IC9FaVPKlYc-t394.0
|
And the reason we do that is because we want to check that a couple of times in the next few
| 394 | 405.28 |
Training BERT #5 - Training With BertForPretraining
|
2021-06-15 15:00:19 UTC
|
https://youtu.be/IC9FaVPKlYc
|
IC9FaVPKlYc
|
UCv83tO5cePwHMt1952IVVHw
|
IC9FaVPKlYc-t398.88
|
lines of code. And first time we check that is now. So we check that the number of sentences is
| 398.88 | 412.08 |
Training BERT #5 - Training With BertForPretraining
|
2021-06-15 15:00:19 UTC
|
https://youtu.be/IC9FaVPKlYc
|
IC9FaVPKlYc
|
UCv83tO5cePwHMt1952IVVHw
|
IC9FaVPKlYc-t405.28
|
greater than one. Now this because we're concatenating two sentences to create our
| 405.28 | 417.36 |
Training BERT #5 - Training With BertForPretraining
|
2021-06-15 15:00:19 UTC
|
https://youtu.be/IC9FaVPKlYc
|
IC9FaVPKlYc
|
UCv83tO5cePwHMt1952IVVHw
|
IC9FaVPKlYc-t412.08
|
training data, we don't want to get just one sentence. We need it where we have, for example,
| 412.08 | 422.4 |
Training BERT #5 - Training With BertForPretraining
|
2021-06-15 15:00:19 UTC
|
https://youtu.be/IC9FaVPKlYc
|
IC9FaVPKlYc
|
UCv83tO5cePwHMt1952IVVHw
|
IC9FaVPKlYc-t417.36
|
in this one, we have multiple sentences so that we can select like this sentence followed by this
| 417.36 | 427.76 |
Training BERT #5 - Training With BertForPretraining
|
2021-06-15 15:00:19 UTC
|
https://youtu.be/IC9FaVPKlYc
|
IC9FaVPKlYc
|
UCv83tO5cePwHMt1952IVVHw
|
IC9FaVPKlYc-t422.4
|
sentence. We can't do that with these because there's no guarantee that this paragraph here
| 422.4 | 432.48 |
Training BERT #5 - Training With BertForPretraining
|
2021-06-15 15:00:19 UTC
|
https://youtu.be/IC9FaVPKlYc
|
IC9FaVPKlYc
|
UCv83tO5cePwHMt1952IVVHw
|
IC9FaVPKlYc-t427.76
|
is going to be talking about the same topic as this paragraph here. So we just avoid that.
| 427.76 | 438.4 |
Training BERT #5 - Training With BertForPretraining
|
2021-06-15 15:00:19 UTC
|
https://youtu.be/IC9FaVPKlYc
|
IC9FaVPKlYc
|
UCv83tO5cePwHMt1952IVVHw
|
IC9FaVPKlYc-t432.47999999999996
|
And in here, first thing we want to do is set out start sentence. So this is where
| 432.48 | 442.64 |
Training BERT #5 - Training With BertForPretraining
|
2021-06-15 15:00:19 UTC
|
https://youtu.be/IC9FaVPKlYc
|
IC9FaVPKlYc
|
UCv83tO5cePwHMt1952IVVHw
|
IC9FaVPKlYc-t438.4
|
sentence A is going to come from. And we're going to randomly select,
| 438.4 | 450.8 |
Training BERT #5 - Training With BertForPretraining
|
2021-06-15 15:00:19 UTC
|
https://youtu.be/IC9FaVPKlYc
|
IC9FaVPKlYc
|
UCv83tO5cePwHMt1952IVVHw
|
IC9FaVPKlYc-t443.84
|
say for this example, we want to randomly select any of the first one, two, three sentences.
| 443.84 | 455.6 |
Training BERT #5 - Training With BertForPretraining
|
2021-06-15 15:00:19 UTC
|
https://youtu.be/IC9FaVPKlYc
|
IC9FaVPKlYc
|
UCv83tO5cePwHMt1952IVVHw
|
IC9FaVPKlYc-t450.8
|
Okay, we'd want to select any of these three, but not this one, because if this sentence A,
| 450.8 | 458.64 |
Training BERT #5 - Training With BertForPretraining
|
2021-06-15 15:00:19 UTC
|
https://youtu.be/IC9FaVPKlYc
|
IC9FaVPKlYc
|
UCv83tO5cePwHMt1952IVVHw
|
IC9FaVPKlYc-t455.6
|
we don't have a sentence B which follows it to extract.
| 455.6 | 474.48 |
Training BERT #5 - Training With BertForPretraining
|
2021-06-15 15:00:19 UTC
|
https://youtu.be/IC9FaVPKlYc
|
IC9FaVPKlYc
|
UCv83tO5cePwHMt1952IVVHw
|
IC9FaVPKlYc-t461.28000000000003
|
So we write random, randint 0 up to the length of num sentences minus two. Now we can now get
| 461.28 | 485.12 |
Training BERT #5 - Training With BertForPretraining
|
2021-06-15 15:00:19 UTC
|
https://youtu.be/IC9FaVPKlYc
|
IC9FaVPKlYc
|
UCv83tO5cePwHMt1952IVVHw
|
IC9FaVPKlYc-t474.48
|
our sentence A, which is append, and we just write sentences start. And then for our sentence B,
| 474.48 | 490.8 |
Training BERT #5 - Training With BertForPretraining
|
2021-06-15 15:00:19 UTC
|
https://youtu.be/IC9FaVPKlYc
|
IC9FaVPKlYc
|
UCv83tO5cePwHMt1952IVVHw
|
IC9FaVPKlYc-t485.12
|
50% we want to select random one from bag up here, 50% of time we want to select the genuine
| 485.12 | 498.48 |
Training BERT #5 - Training With BertForPretraining
|
2021-06-15 15:00:19 UTC
|
https://youtu.be/IC9FaVPKlYc
|
IC9FaVPKlYc
|
UCv83tO5cePwHMt1952IVVHw
|
IC9FaVPKlYc-t490.8
|
next sentence. So say if random.random, so this will select a random float between 0 and 1,
| 490.8 | 510.4 |
Training BERT #5 - Training With BertForPretraining
|
2021-06-15 15:00:19 UTC
|
https://youtu.be/IC9FaVPKlYc
|
IC9FaVPKlYc
|
UCv83tO5cePwHMt1952IVVHw
|
IC9FaVPKlYc-t498.48
|
it's greater than 0.5. And sentence B is going to be, we'll make this our coherent version. So
| 498.48 | 522.88 |
Training BERT #5 - Training With BertForPretraining
|
2021-06-15 15:00:19 UTC
|
https://youtu.be/IC9FaVPKlYc
|
IC9FaVPKlYc
|
UCv83tO5cePwHMt1952IVVHw
|
IC9FaVPKlYc-t510.40000000000003
|
sentences start plus one. And that means our label will have to be zero because that means
| 510.4 | 529.36 |
Training BERT #5 - Training With BertForPretraining
|
2021-06-15 15:00:19 UTC
|
https://youtu.be/IC9FaVPKlYc
|
IC9FaVPKlYc
|
UCv83tO5cePwHMt1952IVVHw
|
IC9FaVPKlYc-t522.88
|
that these two sentences are coherent. Sentence B does follow sentence A. Otherwise,
| 522.88 | 538.48 |
Training BERT #5 - Training With BertForPretraining
|
2021-06-15 15:00:19 UTC
|
https://youtu.be/IC9FaVPKlYc
|
IC9FaVPKlYc
|
UCv83tO5cePwHMt1952IVVHw
|
IC9FaVPKlYc-t530.8
|
we select a random sentence for sentence B. So do append, and here we would write bag,
| 530.8 | 544.8 |
Training BERT #5 - Training With BertForPretraining
|
2021-06-15 15:00:19 UTC
|
https://youtu.be/IC9FaVPKlYc
|
IC9FaVPKlYc
|
UCv83tO5cePwHMt1952IVVHw
|
IC9FaVPKlYc-t539.52
|
and we need to select a random one. So we do random, same as we did earlier on for the start,
| 539.52 | 556.16 |
Training BERT #5 - Training With BertForPretraining
|
2021-06-15 15:00:19 UTC
|
https://youtu.be/IC9FaVPKlYc
|
IC9FaVPKlYc
|
UCv83tO5cePwHMt1952IVVHw
|
IC9FaVPKlYc-t544.8
|
we do random, randint from zero to the length of the bag size minus one. So we also need to do the
| 544.8 | 563.28 |
Training BERT #5 - Training With BertForPretraining
|
2021-06-15 15:00:19 UTC
|
https://youtu.be/IC9FaVPKlYc
|
IC9FaVPKlYc
|
UCv83tO5cePwHMt1952IVVHw
|
IC9FaVPKlYc-t556.16
|
label, which is going to be one in this case. We can execute that. Now that will work. I go a
| 556.16 | 570.48 |
Training BERT #5 - Training With BertForPretraining
|
2021-06-15 15:00:19 UTC
|
https://youtu.be/IC9FaVPKlYc
|
IC9FaVPKlYc
|
UCv83tO5cePwHMt1952IVVHw
|
IC9FaVPKlYc-t563.28
|
little more into depth on this in the previous NSP video. So I'll leave a link to that in the
| 563.28 | 576.88 |
Training BERT #5 - Training With BertForPretraining
|
2021-06-15 15:00:19 UTC
|
https://youtu.be/IC9FaVPKlYc
|
IC9FaVPKlYc
|
UCv83tO5cePwHMt1952IVVHw
|
IC9FaVPKlYc-t570.48
|
description if you want to go through it. And now what we can do is tokenize our data. So to do that,
| 570.48 | 581.84 |
Training BERT #5 - Training With BertForPretraining
|
2021-06-15 15:00:19 UTC
|
https://youtu.be/IC9FaVPKlYc
|
IC9FaVPKlYc
|
UCv83tO5cePwHMt1952IVVHw
|
IC9FaVPKlYc-t576.88
|
we just write inputs and we use a tokenizer. So this is just normal, you know,
| 576.88 | 591.6 |
Training BERT #5 - Training With BertForPretraining
|
2021-06-15 15:00:19 UTC
|
https://youtu.be/IC9FaVPKlYc
|
IC9FaVPKlYc
|
UCv83tO5cePwHMt1952IVVHw
|
IC9FaVPKlYc-t581.84
|
hugging face transformers. And we just write sentence A and sentence B. So hugging face
| 581.84 | 595.84 |
Training BERT #5 - Training With BertForPretraining
|
2021-06-15 15:00:19 UTC
|
https://youtu.be/IC9FaVPKlYc
|
IC9FaVPKlYc
|
UCv83tO5cePwHMt1952IVVHw
|
IC9FaVPKlYc-t591.6
|
transformers will know what we want to do with that. It will deal with formatting for us,
| 591.6 | 605.04 |
Training BERT #5 - Training With BertForPretraining
|
2021-06-15 15:00:19 UTC
|
https://youtu.be/IC9FaVPKlYc
|
IC9FaVPKlYc
|
UCv83tO5cePwHMt1952IVVHw
|
IC9FaVPKlYc-t595.84
|
which is pretty useful. We want to return PyTorch tensors. So return tensors equals pt.
| 595.84 | 615.84 |
Training BERT #5 - Training With BertForPretraining
|
2021-06-15 15:00:19 UTC
|
https://youtu.be/IC9FaVPKlYc
|
IC9FaVPKlYc
|
UCv83tO5cePwHMt1952IVVHw
|
IC9FaVPKlYc-t607.2
|
And we need to set everything to a max length of 512 tokens. So max length equals 512.
| 607.2 | 623.84 |
Training BERT #5 - Training With BertForPretraining
|
2021-06-15 15:00:19 UTC
|
https://youtu.be/IC9FaVPKlYc
|
IC9FaVPKlYc
|
UCv83tO5cePwHMt1952IVVHw
|
IC9FaVPKlYc-t615.84
|
The truncation needs to be set to true. And we also need to set padding equal to max length.
| 615.84 | 633.52 |
Training BERT #5 - Training With BertForPretraining
|
2021-06-15 15:00:19 UTC
|
https://youtu.be/IC9FaVPKlYc
|
IC9FaVPKlYc
|
UCv83tO5cePwHMt1952IVVHw
|
IC9FaVPKlYc-t625.84
|
Okay. So that creates three different tensors for us.
| 625.84 | 642.8 |
Training BERT #5 - Training With BertForPretraining
|
2021-06-15 15:00:19 UTC
|
https://youtu.be/IC9FaVPKlYc
|
IC9FaVPKlYc
|
UCv83tO5cePwHMt1952IVVHw
|
IC9FaVPKlYc-t635.2
|
Impart IDs, token type IDs, and attention mask. Now for the pre-trained model, we need two more
| 635.2 | 649.12 |
Training BERT #5 - Training With BertForPretraining
|
2021-06-15 15:00:19 UTC
|
https://youtu.be/IC9FaVPKlYc
|
IC9FaVPKlYc
|
UCv83tO5cePwHMt1952IVVHw
|
IC9FaVPKlYc-t642.8
|
tensors. We need our next sentence label tensor. So to create that, we write inputs,
| 642.8 | 656.88 |
Training BERT #5 - Training With BertForPretraining
|
2021-06-15 15:00:19 UTC
|
https://youtu.be/IC9FaVPKlYc
|
IC9FaVPKlYc
|
UCv83tO5cePwHMt1952IVVHw
|
IC9FaVPKlYc-t650.0799999999999
|
next sentence label. And that needs to be a long tensor
| 650.08 | 667.2 |
Training BERT #5 - Training With BertForPretraining
|
2021-06-15 15:00:19 UTC
|
https://youtu.be/IC9FaVPKlYc
|
IC9FaVPKlYc
|
UCv83tO5cePwHMt1952IVVHw
|
IC9FaVPKlYc-t659.1999999999999
|
containing our labels, which we created before in the correct dimensionality. So that's why we're
| 659.2 | 674.24 |
Training BERT #5 - Training With BertForPretraining
|
2021-06-15 15:00:19 UTC
|
https://youtu.be/IC9FaVPKlYc
|
IC9FaVPKlYc
|
UCv83tO5cePwHMt1952IVVHw
|
IC9FaVPKlYc-t667.2
|
using the list here and the transpose. And we can have a look at what that creates as well.
| 667.2 | 683.84 |
Training BERT #5 - Training With BertForPretraining
|
2021-06-15 15:00:19 UTC
|
https://youtu.be/IC9FaVPKlYc
|
IC9FaVPKlYc
|
UCv83tO5cePwHMt1952IVVHw
|
IC9FaVPKlYc-t674.96
|
So look at the first 10. We get that. Okay. And now what we want to do is create our mask data. So
| 674.96 | 691.92 |
Training BERT #5 - Training With BertForPretraining
|
2021-06-15 15:00:19 UTC
|
https://youtu.be/IC9FaVPKlYc
|
IC9FaVPKlYc
|
UCv83tO5cePwHMt1952IVVHw
|
IC9FaVPKlYc-t684.5600000000001
|
we need the labels for our mask first. So when we do this, what we'll do is we're going to clone
| 684.56 | 698 |
Training BERT #5 - Training With BertForPretraining
|
2021-06-15 15:00:19 UTC
|
https://youtu.be/IC9FaVPKlYc
|
IC9FaVPKlYc
|
UCv83tO5cePwHMt1952IVVHw
|
IC9FaVPKlYc-t691.92
|
the input IDs tensor. We're going to use that clone for the labels tensor. And then we're going to go
| 691.92 | 704.48 |
Training BERT #5 - Training With BertForPretraining
|
2021-06-15 15:00:19 UTC
|
https://youtu.be/IC9FaVPKlYc
|
IC9FaVPKlYc
|
UCv83tO5cePwHMt1952IVVHw
|
IC9FaVPKlYc-t698.0
|
back to our input IDs and mask around 15% of the tokens in that tensor. So let's create that labels
| 698 | 719.76 |
Training BERT #5 - Training With BertForPretraining
|
2021-06-15 15:00:19 UTC
|
https://youtu.be/IC9FaVPKlYc
|
IC9FaVPKlYc
|
UCv83tO5cePwHMt1952IVVHw
|
IC9FaVPKlYc-t704.4799999999999
|
tensor. It's going to be equal to inputs, input IDs, detach, and clone. Okay. So now we've got
| 704.48 | 727.76 |
Training BERT #5 - Training With BertForPretraining
|
2021-06-15 15:00:19 UTC
|
https://youtu.be/IC9FaVPKlYc
|
IC9FaVPKlYc
|
UCv83tO5cePwHMt1952IVVHw
|
IC9FaVPKlYc-t719.76
|
our mask data. Okay. So now we'll see in here, we have all of the tensors we need, but we still need
| 719.76 | 734 |
Training BERT #5 - Training With BertForPretraining
|
2021-06-15 15:00:19 UTC
|
https://youtu.be/IC9FaVPKlYc
|
IC9FaVPKlYc
|
UCv83tO5cePwHMt1952IVVHw
|
IC9FaVPKlYc-t727.76
|
to mask around 15% of these before moving on to training our model. And to do that, we'll
| 727.76 | 740.88 |
Training BERT #5 - Training With BertForPretraining
|
2021-06-15 15:00:19 UTC
|
https://youtu.be/IC9FaVPKlYc
|
IC9FaVPKlYc
|
UCv83tO5cePwHMt1952IVVHw
|
IC9FaVPKlYc-t734.0
|
use, we'll create a random array using the torch rend. That needs to be in the same shape as our
| 734 | 750.56 |
Training BERT #5 - Training With BertForPretraining
|
2021-06-15 15:00:19 UTC
|
https://youtu.be/IC9FaVPKlYc
|
IC9FaVPKlYc
|
UCv83tO5cePwHMt1952IVVHw
|
IC9FaVPKlYc-t740.88
|
input IDs. And that will just create a big tensor between values of zero to one. And what we want
| 740.88 | 758.64 |
Training BERT #5 - Training With BertForPretraining
|
2021-06-15 15:00:19 UTC
|
https://youtu.be/IC9FaVPKlYc
|
IC9FaVPKlYc
|
UCv83tO5cePwHMt1952IVVHw
|
IC9FaVPKlYc-t750.56
|
to do is mask around 15% of those. So we write something like this. Okay. And that will give us
| 750.56 | 764.96 |
Training BERT #5 - Training With BertForPretraining
|
2021-06-15 15:00:19 UTC
|
https://youtu.be/IC9FaVPKlYc
|
IC9FaVPKlYc
|
UCv83tO5cePwHMt1952IVVHw
|
IC9FaVPKlYc-t759.28
|
our mask here, but we also don't want to mask special tokens, which we are doing here. We're
| 759.28 | 771.12 |
Training BERT #5 - Training With BertForPretraining
|
2021-06-15 15:00:19 UTC
|
https://youtu.be/IC9FaVPKlYc
|
IC9FaVPKlYc
|
UCv83tO5cePwHMt1952IVVHw
|
IC9FaVPKlYc-t764.96
|
masking classification tokens and also masking padding tokens up here. So we need to add a little
| 764.96 | 781.76 |
Training BERT #5 - Training With BertForPretraining
|
2021-06-15 15:00:19 UTC
|
https://youtu.be/IC9FaVPKlYc
|
IC9FaVPKlYc
|
UCv83tO5cePwHMt1952IVVHw
|
IC9FaVPKlYc-t771.12
|
bit more logic to that. So let me just add this to a variable. So we add that logic, which says,
| 771.12 | 794.88 |
Training BERT #5 - Training With BertForPretraining
|
2021-06-15 15:00:19 UTC
|
https://youtu.be/IC9FaVPKlYc
|
IC9FaVPKlYc
|
UCv83tO5cePwHMt1952IVVHw
|
IC9FaVPKlYc-t784.8000000000001
|
and input IDs is not equal to one zero one, which is our CLS token, which is what we get down here.
| 784.8 | 803.36 |
Training BERT #5 - Training With BertForPretraining
|
2021-06-15 15:00:19 UTC
|
https://youtu.be/IC9FaVPKlYc
|
IC9FaVPKlYc
|
UCv83tO5cePwHMt1952IVVHw
|
IC9FaVPKlYc-t794.88
|
See the impact. See we get faults now. And we also want to do the same for our separator tokens,
| 794.88 | 810.72 |
Training BERT #5 - Training With BertForPretraining
|
2021-06-15 15:00:19 UTC
|
https://youtu.be/IC9FaVPKlYc
|
IC9FaVPKlYc
|
UCv83tO5cePwHMt1952IVVHw
|
IC9FaVPKlYc-t805.2
|
which is one zero two. We can't see any of those. And our padding tokens, we use zero.
| 805.2 | 821.2 |
Training BERT #5 - Training With BertForPretraining
|
2021-06-15 15:00:19 UTC
|
https://youtu.be/IC9FaVPKlYc
|
IC9FaVPKlYc
|
UCv83tO5cePwHMt1952IVVHw
|
IC9FaVPKlYc-t812.16
|
So you see these are all that will go false now, like so. So that's our masking array.
| 812.16 | 830.8 |
Training BERT #5 - Training With BertForPretraining
|
2021-06-15 15:00:19 UTC
|
https://youtu.be/IC9FaVPKlYc
|
IC9FaVPKlYc
|
UCv83tO5cePwHMt1952IVVHw
|
IC9FaVPKlYc-t821.2
|
And now what we want to do is loop through all of these, extract the points at which they are not
| 821.2 | 838.72 |
Training BERT #5 - Training With BertForPretraining
|
2021-06-15 15:00:19 UTC
|
https://youtu.be/IC9FaVPKlYc
|
IC9FaVPKlYc
|
UCv83tO5cePwHMt1952IVVHw
|
IC9FaVPKlYc-t830.8000000000001
|
false. So where we have the mask and use those indices values to mask our actual input IDs up
| 830.8 | 851.92 |
Training BERT #5 - Training With BertForPretraining
|
2021-06-15 15:00:19 UTC
|
https://youtu.be/IC9FaVPKlYc
|
IC9FaVPKlYc
|
UCv83tO5cePwHMt1952IVVHw
|
IC9FaVPKlYc-t838.72
|
here. To do that, we go for i in range inputs, input IDs dot shape zero. This is like iterating
| 838.72 | 861.44 |
Training BERT #5 - Training With BertForPretraining
|
2021-06-15 15:00:19 UTC
|
https://youtu.be/IC9FaVPKlYc
|
IC9FaVPKlYc
|
UCv83tO5cePwHMt1952IVVHw
|
IC9FaVPKlYc-t851.9200000000001
|
through each row. And what we do here is we get selection. So these are the indices where we have
| 851.92 | 874.56 |
Training BERT #5 - Training With BertForPretraining
|
2021-06-15 15:00:19 UTC
|
https://youtu.be/IC9FaVPKlYc
|
IC9FaVPKlYc
|
UCv83tO5cePwHMt1952IVVHw
|
IC9FaVPKlYc-t861.44
|
true values and mask array. And we do that using torch flatten mask array at the given index,
| 861.44 | 885.6 |
Training BERT #5 - Training With BertForPretraining
|
2021-06-15 15:00:19 UTC
|
https://youtu.be/IC9FaVPKlYc
|
IC9FaVPKlYc
|
UCv83tO5cePwHMt1952IVVHw
|
IC9FaVPKlYc-t874.5600000000001
|
where they are non-zero. And we want to create a list from that. Okay. So we have that. Oh, and so
| 874.56 | 891.28 |
Training BERT #5 - Training With BertForPretraining
|
2021-06-15 15:00:19 UTC
|
https://youtu.be/IC9FaVPKlYc
|
IC9FaVPKlYc
|
UCv83tO5cePwHMt1952IVVHw
|
IC9FaVPKlYc-t885.6
|
I want to show you what the selection looks like quickly. So it's just a selection of indices to
| 885.6 | 904.72 |
Training BERT #5 - Training With BertForPretraining
|
2021-06-15 15:00:19 UTC
|
https://youtu.be/IC9FaVPKlYc
|
IC9FaVPKlYc
|
UCv83tO5cePwHMt1952IVVHw
|
IC9FaVPKlYc-t891.28
|
mask. And we want to apply that to our inputs, input IDs. So at the current index, and we
| 891.28 | 911.68 |
Training BERT #5 - Training With BertForPretraining
|
2021-06-15 15:00:19 UTC
|
https://youtu.be/IC9FaVPKlYc
|
IC9FaVPKlYc
|
UCv83tO5cePwHMt1952IVVHw
|
IC9FaVPKlYc-t905.76
|
select those specific items and we set them equal to one zero three, which is the masking token ID.
| 905.76 | 922 |
Training BERT #5 - Training With BertForPretraining
|
2021-06-15 15:00:19 UTC
|
https://youtu.be/IC9FaVPKlYc
|
IC9FaVPKlYc
|
UCv83tO5cePwHMt1952IVVHw
|
IC9FaVPKlYc-t911.68
|
Okay. So that's our masking. And now what we need to do is we need to take all of our data here and
| 911.68 | 928.48 |
Training BERT #5 - Training With BertForPretraining
|
2021-06-15 15:00:19 UTC
|
https://youtu.be/IC9FaVPKlYc
|
IC9FaVPKlYc
|
UCv83tO5cePwHMt1952IVVHw
|
IC9FaVPKlYc-t922.0
|
load it into a PyTorch data loader. And to do that, we need to reform our data into a PyTorch
| 922 | 937.28 |
Training BERT #5 - Training With BertForPretraining
|
2021-06-15 15:00:19 UTC
|
https://youtu.be/IC9FaVPKlYc
|
IC9FaVPKlYc
|
UCv83tO5cePwHMt1952IVVHw
|
IC9FaVPKlYc-t928.4799999999999
|
data set object. And we do that here. So main thing to note is we pass our data into this
| 928.48 | 945.28 |
Training BERT #5 - Training With BertForPretraining
|
2021-06-15 15:00:19 UTC
|
https://youtu.be/IC9FaVPKlYc
|
IC9FaVPKlYc
|
UCv83tO5cePwHMt1952IVVHw
|
IC9FaVPKlYc-t937.28
|
initialization that assigns them to this self encodings attribute. And then here we say, okay,
| 937.28 | 953.36 |
Training BERT #5 - Training With BertForPretraining
|
2021-06-15 15:00:19 UTC
|
https://youtu.be/IC9FaVPKlYc
|
IC9FaVPKlYc
|
UCv83tO5cePwHMt1952IVVHw
|
IC9FaVPKlYc-t945.28
|
given a certain index, we want to extract the tensors in a dictionary format for that index.
| 945.28 | 960.48 |
Training BERT #5 - Training With BertForPretraining
|
2021-06-15 15:00:19 UTC
|
https://youtu.be/IC9FaVPKlYc
|
IC9FaVPKlYc
|
UCv83tO5cePwHMt1952IVVHw
|
IC9FaVPKlYc-t954.16
|
And then here we're just passing the lengths of how many tensors or how many samples we have in
| 954.16 | 970.4 |
Training BERT #5 - Training With BertForPretraining
|
2021-06-15 15:00:19 UTC
|
https://youtu.be/IC9FaVPKlYc
|
IC9FaVPKlYc
|
UCv83tO5cePwHMt1952IVVHw
|
IC9FaVPKlYc-t960.48
|
the full data set. So run that. We initialize our data sets using that class. So right, data set
| 960.48 | 977.76 |
Training BERT #5 - Training With BertForPretraining
|
2021-06-15 15:00:19 UTC
|
https://youtu.be/IC9FaVPKlYc
|
IC9FaVPKlYc
|
UCv83tO5cePwHMt1952IVVHw
|
IC9FaVPKlYc-t970.4
|
equals meditations data set, pass our data in there, which is inputs. And then with that,
| 970.4 | 991.84 |
Training BERT #5 - Training With BertForPretraining
|
2021-06-15 15:00:19 UTC
|
https://youtu.be/IC9FaVPKlYc
|
IC9FaVPKlYc
|
UCv83tO5cePwHMt1952IVVHw
|
IC9FaVPKlYc-t977.76
|
we can create our data loader like this. So torch utils data data loader. And we have data set.
| 977.76 | 997.76 |
Training BERT #5 - Training With BertForPretraining
|
2021-06-15 15:00:19 UTC
|
https://youtu.be/IC9FaVPKlYc
|
IC9FaVPKlYc
|
UCv83tO5cePwHMt1952IVVHw
|
IC9FaVPKlYc-t991.84
|
Okay. So that's ready. Now we need to set up our training loop. So first thing we need to do is
| 991.84 | 1,004.56 |
Training BERT #5 - Training With BertForPretraining
|
2021-06-15 15:00:19 UTC
|
https://youtu.be/IC9FaVPKlYc
|
IC9FaVPKlYc
|
UCv83tO5cePwHMt1952IVVHw
|
IC9FaVPKlYc-t997.76
|
check if we are on GPU or not. If we are, we use it and we do that like so. So device equals torch
| 997.76 | 1,011.2 |
Training BERT #5 - Training With BertForPretraining
|
2021-06-15 15:00:19 UTC
|
https://youtu.be/IC9FaVPKlYc
|
IC9FaVPKlYc
|
UCv83tO5cePwHMt1952IVVHw
|
IC9FaVPKlYc-t1004.56
|
device cuda if torch cuda is available. Else torch device CPU. So that's saying use the GPU
| 1,004.56 | 1,019.28 |
Training BERT #5 - Training With BertForPretraining
|
2021-06-15 15:00:19 UTC
|
https://youtu.be/IC9FaVPKlYc
|
IC9FaVPKlYc
|
UCv83tO5cePwHMt1952IVVHw
|
IC9FaVPKlYc-t1011.1999999999999
|
if we have a cuda enabled GPU, otherwise use CPU. And then what we want to do is move our model
| 1,011.2 | 1,028.48 |
Training BERT #5 - Training With BertForPretraining
|
2021-06-15 15:00:19 UTC
|
https://youtu.be/IC9FaVPKlYc
|
IC9FaVPKlYc
|
UCv83tO5cePwHMt1952IVVHw
|
IC9FaVPKlYc-t1019.92
|
over to that device. And we also want to activate the training mode of our model.
| 1,019.92 | 1,037.12 |
Training BERT #5 - Training With BertForPretraining
|
2021-06-15 15:00:19 UTC
|
https://youtu.be/IC9FaVPKlYc
|
IC9FaVPKlYc
|
UCv83tO5cePwHMt1952IVVHw
|
IC9FaVPKlYc-t1028.48
|
And then we need to initialize our optimizer. I'm going to be using Adam with weighted decay.
| 1,028.48 | 1,049.84 |
Training BERT #5 - Training With BertForPretraining
|
2021-06-15 15:00:19 UTC
|
https://youtu.be/IC9FaVPKlYc
|
IC9FaVPKlYc
|
UCv83tO5cePwHMt1952IVVHw
|
IC9FaVPKlYc-t1038.64
|
So from transformers import Adam w. And initialize it like this. So optim equals Adam w.
| 1,038.64 | 1,058.08 |
Training BERT #5 - Training With BertForPretraining
|
2021-06-15 15:00:19 UTC
|
https://youtu.be/IC9FaVPKlYc
|
IC9FaVPKlYc
|
UCv83tO5cePwHMt1952IVVHw
|
IC9FaVPKlYc-t1051.6
|
We pass our model parameters to that. And we also pass a learning rate. So learning rate
| 1,051.6 | 1,065.84 |
Training BERT #5 - Training With BertForPretraining
|
2021-06-15 15:00:19 UTC
|
https://youtu.be/IC9FaVPKlYc
|
IC9FaVPKlYc
|
UCv83tO5cePwHMt1952IVVHw
|
IC9FaVPKlYc-t1058.08
|
is going to be 5e to the minus 5. Okay. And now we can create our training loop.
| 1,058.08 | 1,074.4 |
Training BERT #5 - Training With BertForPretraining
|
2021-06-15 15:00:19 UTC
|
https://youtu.be/IC9FaVPKlYc
|
IC9FaVPKlYc
|
UCv83tO5cePwHMt1952IVVHw
|
IC9FaVPKlYc-t1065.84
|
So you're going to use TQDM to create the progress bar. And we're going to go through
| 1,065.84 | 1,084.4 |
Training BERT #5 - Training With BertForPretraining
|
2021-06-15 15:00:19 UTC
|
https://youtu.be/IC9FaVPKlYc
|
IC9FaVPKlYc
|
UCv83tO5cePwHMt1952IVVHw
|
IC9FaVPKlYc-t1074.3999999999999
|
two epochs. So for epoch in range two, we initialize our loop by wrapping it within TQDM.
| 1,074.4 | 1,091.28 |
Training BERT #5 - Training With BertForPretraining
|
2021-06-15 15:00:19 UTC
|
https://youtu.be/IC9FaVPKlYc
|
IC9FaVPKlYc
|
UCv83tO5cePwHMt1952IVVHw
|
IC9FaVPKlYc-t1084.4
|
And in here we have our data loader. And we set leave equal to true so that we can see that progress
| 1,084.4 | 1,102.48 |
Training BERT #5 - Training With BertForPretraining
|
2021-06-15 15:00:19 UTC
|
https://youtu.be/IC9FaVPKlYc
|
IC9FaVPKlYc
|
UCv83tO5cePwHMt1952IVVHw
|
IC9FaVPKlYc-t1091.2800000000002
|
bar. And then we loop through each batch within that loop. Up here, so I didn't actually set the
| 1,091.28 | 1,108.08 |
Training BERT #5 - Training With BertForPretraining
|
2021-06-15 15:00:19 UTC
|
https://youtu.be/IC9FaVPKlYc
|
IC9FaVPKlYc
|
UCv83tO5cePwHMt1952IVVHw
|
IC9FaVPKlYc-t1102.48
|
batches. My mistake. So up here we want to set where we initialize the data loader. We want to
| 1,102.48 | 1,120.56 |
Training BERT #5 - Training With BertForPretraining
|
2021-06-15 15:00:19 UTC
|
https://youtu.be/IC9FaVPKlYc
|
IC9FaVPKlYc
|
UCv83tO5cePwHMt1952IVVHw
|
IC9FaVPKlYc-t1108.08
|
set batch size equal to 16. And also shuffle the data set as well. Okay. So for batch in loop,
| 1,108.08 | 1,129.52 |
Training BERT #5 - Training With BertForPretraining
|
2021-06-15 15:00:19 UTC
|
https://youtu.be/IC9FaVPKlYc
|
IC9FaVPKlYc
|
UCv83tO5cePwHMt1952IVVHw
|
IC9FaVPKlYc-t1121.52
|
here we want to initialize the gradient on our optimizer. And then we need to load in
| 1,121.52 | 1,139.44 |
Training BERT #5 - Training With BertForPretraining
|
2021-06-15 15:00:19 UTC
|
https://youtu.be/IC9FaVPKlYc
|
IC9FaVPKlYc
|
UCv83tO5cePwHMt1952IVVHw
|
IC9FaVPKlYc-t1129.52
|
each of our tensors, which there are quite a few of them. So we have inputs.keys. We need to load
| 1,129.52 | 1,148.56 |
Training BERT #5 - Training With BertForPretraining
|
2021-06-15 15:00:19 UTC
|
https://youtu.be/IC9FaVPKlYc
|
IC9FaVPKlYc
|
UCv83tO5cePwHMt1952IVVHw
|
IC9FaVPKlYc-t1139.44
|
in each one of these. So input IDs equals batch. We access this like a dictionary. So input IDs.
| 1,139.44 | 1,156.08 |
Training BERT #5 - Training With BertForPretraining
|
2021-06-15 15:00:19 UTC
|
https://youtu.be/IC9FaVPKlYc
|
IC9FaVPKlYc
|
UCv83tO5cePwHMt1952IVVHw
|
IC9FaVPKlYc-t1150.32
|
We also want to move each one of those tensors that we're using to our device.
| 1,150.32 | 1,166.88 |
Training BERT #5 - Training With BertForPretraining
|
2021-06-15 15:00:19 UTC
|
https://youtu.be/IC9FaVPKlYc
|
IC9FaVPKlYc
|
UCv83tO5cePwHMt1952IVVHw
|
IC9FaVPKlYc-t1156.08
|
So we do that for each one of those. And we have attention mask.
| 1,156.08 | 1,172.88 |
Training BERT #5 - Training With BertForPretraining
|
2021-06-15 15:00:19 UTC
|
https://youtu.be/IC9FaVPKlYc
|
IC9FaVPKlYc
|
UCv83tO5cePwHMt1952IVVHw
|
IC9FaVPKlYc-t1168.56
|
And next sentence labels and also labels.
| 1,168.56 | 1,187.36 |
Training BERT #5 - Training With BertForPretraining
|
2021-06-15 15:00:19 UTC
|
https://youtu.be/IC9FaVPKlYc
|
IC9FaVPKlYc
|
UCv83tO5cePwHMt1952IVVHw
|
IC9FaVPKlYc-t1172.88
|
Labels and also labels. Okay. And now we can actually process that through our model.
| 1,172.88 | 1,193.52 |
Training BERT #5 - Training With BertForPretraining
|
2021-06-15 15:00:19 UTC
|
https://youtu.be/IC9FaVPKlYc
|
IC9FaVPKlYc
|
UCv83tO5cePwHMt1952IVVHw
|
IC9FaVPKlYc-t1188.8000000000002
|
So in here, we just need to pass all of these tensors that we have. So input IDs.
| 1,188.8 | 1,200.8 |
Training BERT #5 - Training With BertForPretraining
|
2021-06-15 15:00:19 UTC
|
https://youtu.be/IC9FaVPKlYc
|
IC9FaVPKlYc
|
UCv83tO5cePwHMt1952IVVHw
|
IC9FaVPKlYc-t1193.52
|
And then we have token type IDs. Just copy this.
| 1,193.52 | 1,204.8 |
Training BERT #5 - Training With BertForPretraining
|
2021-06-15 15:00:19 UTC
|
https://youtu.be/IC9FaVPKlYc
|
IC9FaVPKlYc
|
UCv83tO5cePwHMt1952IVVHw
|
IC9FaVPKlYc-t1203.76
|
Attention mask.
| 1,203.76 | 1,209.84 |
Training BERT #5 - Training With BertForPretraining
|
2021-06-15 15:00:19 UTC
|
https://youtu.be/IC9FaVPKlYc
|
IC9FaVPKlYc
|
UCv83tO5cePwHMt1952IVVHw
|
IC9FaVPKlYc-t1207.84
|
Next sentence label.
| 1,207.84 | 1,214.8 |
Subsets and Splits
No community queries yet
The top public SQL queries from the community will appear here once available.