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
|
|---|---|---|---|---|---|---|---|---|
How to Build Custom Q&A Transformer Models in Python
|
2021-02-12 13:30:03 UTC
|
https://youtu.be/ZIRmXkHp0-c
|
ZIRmXkHp0-c
|
UCv83tO5cePwHMt1952IVVHw
|
ZIRmXkHp0-c-t1022.4
|
And this is where we get our question.
| 1,022.4 | 1,044.16 |
How to Build Custom Q&A Transformer Models in Python
|
2021-02-12 13:30:03 UTC
|
https://youtu.be/ZIRmXkHp0-c
|
ZIRmXkHp0-c
|
UCv83tO5cePwHMt1952IVVHw
|
ZIRmXkHp0-c-t1039.52
|
Then once we're here, we need to do something slightly different,
| 1,039.52 | 1,047.92 |
How to Build Custom Q&A Transformer Models in Python
|
2021-02-12 13:30:03 UTC
|
https://youtu.be/ZIRmXkHp0-c
|
ZIRmXkHp0-c
|
UCv83tO5cePwHMt1952IVVHw
|
ZIRmXkHp0-c-t1044.72
|
which is this plausible answers.
| 1,044.72 | 1,048.72 |
How to Build Custom Q&A Transformer Models in Python
|
2021-02-12 13:30:03 UTC
|
https://youtu.be/ZIRmXkHp0-c
|
ZIRmXkHp0-c
|
UCv83tO5cePwHMt1952IVVHw
|
ZIRmXkHp0-c-t1047.92
|
Okay.
| 1,047.92 | 1,054.72 |
How to Build Custom Q&A Transformer Models in Python
|
2021-02-12 13:30:03 UTC
|
https://youtu.be/ZIRmXkHp0-c
|
ZIRmXkHp0-c
|
UCv83tO5cePwHMt1952IVVHw
|
ZIRmXkHp0-c-t1048.72
|
And then we use this access variable in order to define what we're going to loop through next.
| 1,048.72 | 1,059.52 |
How to Build Custom Q&A Transformer Models in Python
|
2021-02-12 13:30:03 UTC
|
https://youtu.be/ZIRmXkHp0-c
|
ZIRmXkHp0-c
|
UCv83tO5cePwHMt1952IVVHw
|
ZIRmXkHp0-c-t1057.28
|
So here we go for answers.
| 1,057.28 | 1,067.92 |
How to Build Custom Q&A Transformer Models in Python
|
2021-02-12 13:30:03 UTC
|
https://youtu.be/ZIRmXkHp0-c
|
ZIRmXkHp0-c
|
UCv83tO5cePwHMt1952IVVHw
|
ZIRmXkHp0-c-t1060.88
|
Answer, sorry, in QA access, because this will switch to implausible answers or answers.
| 1,060.88 | 1,076.4 |
How to Build Custom Q&A Transformer Models in Python
|
2021-02-12 13:30:03 UTC
|
https://youtu.be/ZIRmXkHp0-c
|
ZIRmXkHp0-c
|
UCv83tO5cePwHMt1952IVVHw
|
ZIRmXkHp0-c-t1067.92
|
And then within this for loop, this is where we can begin adding this context, question, and answer
| 1,067.92 | 1,083.12 |
How to Build Custom Q&A Transformer Models in Python
|
2021-02-12 13:30:03 UTC
|
https://youtu.be/ZIRmXkHp0-c
|
ZIRmXkHp0-c
|
UCv83tO5cePwHMt1952IVVHw
|
ZIRmXkHp0-c-t1076.4
|
to a list of questions, contexts, and answers that we still need to define up here.
| 1,076.4 | 1,089.04 |
How to Build Custom Q&A Transformer Models in Python
|
2021-02-12 13:30:03 UTC
|
https://youtu.be/ZIRmXkHp0-c
|
ZIRmXkHp0-c
|
UCv83tO5cePwHMt1952IVVHw
|
ZIRmXkHp0-c-t1085.52
|
So each one of these is just going to be an empty list.
| 1,085.52 | 1,094.64 |
How to Build Custom Q&A Transformer Models in Python
|
2021-02-12 13:30:03 UTC
|
https://youtu.be/ZIRmXkHp0-c
|
ZIRmXkHp0-c
|
UCv83tO5cePwHMt1952IVVHw
|
ZIRmXkHp0-c-t1089.04
|
And then all we do is copy this across,
| 1,089.04 | 1,101.44 |
How to Build Custom Q&A Transformer Models in Python
|
2021-02-12 13:30:03 UTC
|
https://youtu.be/ZIRmXkHp0-c
|
ZIRmXkHp0-c
|
UCv83tO5cePwHMt1952IVVHw
|
ZIRmXkHp0-c-t1097.04
|
and we just append everything that we've extracted in this loop.
| 1,097.04 | 1,106.32 |
How to Build Custom Q&A Transformer Models in Python
|
2021-02-12 13:30:03 UTC
|
https://youtu.be/ZIRmXkHp0-c
|
ZIRmXkHp0-c
|
UCv83tO5cePwHMt1952IVVHw
|
ZIRmXkHp0-c-t1105.44
|
And the context.
| 1,105.44 | 1,111.36 |
How to Build Custom Q&A Transformer Models in Python
|
2021-02-12 13:30:03 UTC
|
https://youtu.be/ZIRmXkHp0-c
|
ZIRmXkHp0-c
|
UCv83tO5cePwHMt1952IVVHw
|
ZIRmXkHp0-c-t1109.84
|
And then we just add the context.
| 1,109.84 | 1,114.96 |
How to Build Custom Q&A Transformer Models in Python
|
2021-02-12 13:30:03 UTC
|
https://youtu.be/ZIRmXkHp0-c
|
ZIRmXkHp0-c
|
UCv83tO5cePwHMt1952IVVHw
|
ZIRmXkHp0-c-t1113.36
|
And then we just add the context.
| 1,113.36 | 1,125.92 |
How to Build Custom Q&A Transformer Models in Python
|
2021-02-12 13:30:03 UTC
|
https://youtu.be/ZIRmXkHp0-c
|
ZIRmXkHp0-c
|
UCv83tO5cePwHMt1952IVVHw
|
ZIRmXkHp0-c-t1114.96
|
And the context, question, and answer.
| 1,114.96 | 1,131.68 |
How to Build Custom Q&A Transformer Models in Python
|
2021-02-12 13:30:03 UTC
|
https://youtu.be/ZIRmXkHp0-c
|
ZIRmXkHp0-c
|
UCv83tO5cePwHMt1952IVVHw
|
ZIRmXkHp0-c-t1130.72
|
And that should work.
| 1,130.72 | 1,141.52 |
How to Build Custom Q&A Transformer Models in Python
|
2021-02-12 13:30:03 UTC
|
https://youtu.be/ZIRmXkHp0-c
|
ZIRmXkHp0-c
|
UCv83tO5cePwHMt1952IVVHw
|
ZIRmXkHp0-c-t1136.88
|
So now let's take a look at a few of our contexts.
| 1,136.88 | 1,146.32 |
How to Build Custom Q&A Transformer Models in Python
|
2021-02-12 13:30:03 UTC
|
https://youtu.be/ZIRmXkHp0-c
|
ZIRmXkHp0-c
|
UCv83tO5cePwHMt1952IVVHw
|
ZIRmXkHp0-c-t1141.52
|
Okay, and we can see we have this and because we have multiple question and answers for each context,
| 1,141.52 | 1,148.88 |
How to Build Custom Q&A Transformer Models in Python
|
2021-02-12 13:30:03 UTC
|
https://youtu.be/ZIRmXkHp0-c
|
ZIRmXkHp0-c
|
UCv83tO5cePwHMt1952IVVHw
|
ZIRmXkHp0-c-t1146.32
|
the context just repeat over and over again.
| 1,146.32 | 1,155.84 |
How to Build Custom Q&A Transformer Models in Python
|
2021-02-12 13:30:03 UTC
|
https://youtu.be/ZIRmXkHp0-c
|
ZIRmXkHp0-c
|
UCv83tO5cePwHMt1952IVVHw
|
ZIRmXkHp0-c-t1152.4
|
But then we should see something slightly different when we go with answers.
| 1,152.4 | 1,160.08 |
How to Build Custom Q&A Transformer Models in Python
|
2021-02-12 13:30:03 UTC
|
https://youtu.be/ZIRmXkHp0-c
|
ZIRmXkHp0-c
|
UCv83tO5cePwHMt1952IVVHw
|
ZIRmXkHp0-c-t1158.16
|
And questions.
| 1,158.16 | 1,163.28 |
How to Build Custom Q&A Transformer Models in Python
|
2021-02-12 13:30:03 UTC
|
https://youtu.be/ZIRmXkHp0-c
|
ZIRmXkHp0-c
|
UCv83tO5cePwHMt1952IVVHw
|
ZIRmXkHp0-c-t1162.0
|
Okay, so that's great.
| 1,162 | 1,166.56 |
How to Build Custom Q&A Transformer Models in Python
|
2021-02-12 13:30:03 UTC
|
https://youtu.be/ZIRmXkHp0-c
|
ZIRmXkHp0-c
|
UCv83tO5cePwHMt1952IVVHw
|
ZIRmXkHp0-c-t1163.28
|
We have our data in a reasonable format now.
| 1,163.28 | 1,171.12 |
How to Build Custom Q&A Transformer Models in Python
|
2021-02-12 13:30:03 UTC
|
https://youtu.be/ZIRmXkHp0-c
|
ZIRmXkHp0-c
|
UCv83tO5cePwHMt1952IVVHw
|
ZIRmXkHp0-c-t1166.56
|
But we want to do this for both the training set and the validation set.
| 1,166.56 | 1,179.84 |
How to Build Custom Q&A Transformer Models in Python
|
2021-02-12 13:30:03 UTC
|
https://youtu.be/ZIRmXkHp0-c
|
ZIRmXkHp0-c
|
UCv83tO5cePwHMt1952IVVHw
|
ZIRmXkHp0-c-t1172.08
|
So what we're going to do is just going to put this into a function like we were going to do before.
| 1,172.08 | 1,184.32 |
How to Build Custom Q&A Transformer Models in Python
|
2021-02-12 13:30:03 UTC
|
https://youtu.be/ZIRmXkHp0-c
|
ZIRmXkHp0-c
|
UCv83tO5cePwHMt1952IVVHw
|
ZIRmXkHp0-c-t1182.8799999999999
|
Just read squad.
| 1,182.88 | 1,191.04 |
How to Build Custom Q&A Transformer Models in Python
|
2021-02-12 13:30:03 UTC
|
https://youtu.be/ZIRmXkHp0-c
|
ZIRmXkHp0-c
|
UCv83tO5cePwHMt1952IVVHw
|
ZIRmXkHp0-c-t1184.32
|
So here we're going to read in our data, and then we run through it and transform it into our three lists.
| 1,184.32 | 1,194.08 |
How to Build Custom Q&A Transformer Models in Python
|
2021-02-12 13:30:03 UTC
|
https://youtu.be/ZIRmXkHp0-c
|
ZIRmXkHp0-c
|
UCv83tO5cePwHMt1952IVVHw
|
ZIRmXkHp0-c-t1191.04
|
All we need to do now is actually return those three lists.
| 1,191.04 | 1,203.84 |
How to Build Custom Q&A Transformer Models in Python
|
2021-02-12 13:30:03 UTC
|
https://youtu.be/ZIRmXkHp0-c
|
ZIRmXkHp0-c
|
UCv83tO5cePwHMt1952IVVHw
|
ZIRmXkHp0-c-t1202.8799999999999
|
And answers.
| 1,202.88 | 1,206.88 |
How to Build Custom Q&A Transformer Models in Python
|
2021-02-12 13:30:03 UTC
|
https://youtu.be/ZIRmXkHp0-c
|
ZIRmXkHp0-c
|
UCv83tO5cePwHMt1952IVVHw
|
ZIRmXkHp0-c-t1203.84
|
So now what we're going to do is just go back to our training set.
| 1,203.84 | 1,211.6 |
How to Build Custom Q&A Transformer Models in Python
|
2021-02-12 13:30:03 UTC
|
https://youtu.be/ZIRmXkHp0-c
|
ZIRmXkHp0-c
|
UCv83tO5cePwHMt1952IVVHw
|
ZIRmXkHp0-c-t1206.8799999999999
|
And we're going to do this for both the training set and the validation set.
| 1,206.88 | 1,220.48 |
How to Build Custom Q&A Transformer Models in Python
|
2021-02-12 13:30:03 UTC
|
https://youtu.be/ZIRmXkHp0-c
|
ZIRmXkHp0-c
|
UCv83tO5cePwHMt1952IVVHw
|
ZIRmXkHp0-c-t1211.6
|
And answers. So now what we can do is execute this function for both our training and validation sets.
| 1,211.6 | 1,240 |
How to Build Custom Q&A Transformer Models in Python
|
2021-02-12 13:30:03 UTC
|
https://youtu.be/ZIRmXkHp0-c
|
ZIRmXkHp0-c
|
UCv83tO5cePwHMt1952IVVHw
|
ZIRmXkHp0-c-t1220.48
|
So we're going to train context, questions and answers.
| 1,220.48 | 1,252 |
How to Build Custom Q&A Transformer Models in Python
|
2021-02-12 13:30:03 UTC
|
https://youtu.be/ZIRmXkHp0-c
|
ZIRmXkHp0-c
|
UCv83tO5cePwHMt1952IVVHw
|
ZIRmXkHp0-c-t1250.56
|
Okay, so that is one of them.
| 1,250.56 | 1,254 |
How to Build Custom Q&A Transformer Models in Python
|
2021-02-12 13:30:03 UTC
|
https://youtu.be/ZIRmXkHp0-c
|
ZIRmXkHp0-c
|
UCv83tO5cePwHMt1952IVVHw
|
ZIRmXkHp0-c-t1252.0
|
And we can just copy that.
| 1,252 | 1,270 |
How to Build Custom Q&A Transformer Models in Python
|
2021-02-12 13:30:03 UTC
|
https://youtu.be/ZIRmXkHp0-c
|
ZIRmXkHp0-c
|
UCv83tO5cePwHMt1952IVVHw
|
ZIRmXkHp0-c-t1258.24
|
And we just want this to be our validation set.
| 1,258.24 | 1,274.24 |
How to Build Custom Q&A Transformer Models in Python
|
2021-02-12 13:30:03 UTC
|
https://youtu.be/ZIRmXkHp0-c
|
ZIRmXkHp0-c
|
UCv83tO5cePwHMt1952IVVHw
|
ZIRmXkHp0-c-t1273.68
|
Like so.
| 1,273.68 | 1,279.52 |
How to Build Custom Q&A Transformer Models in Python
|
2021-02-12 13:30:03 UTC
|
https://youtu.be/ZIRmXkHp0-c
|
ZIRmXkHp0-c
|
UCv83tO5cePwHMt1952IVVHw
|
ZIRmXkHp0-c-t1278.56
|
Okay, so that's great.
| 1,278.56 | 1,288.48 |
How to Build Custom Q&A Transformer Models in Python
|
2021-02-12 13:30:03 UTC
|
https://youtu.be/ZIRmXkHp0-c
|
ZIRmXkHp0-c
|
UCv83tO5cePwHMt1952IVVHw
|
ZIRmXkHp0-c-t1279.52
|
We now have the training context and the validation context, which we can see right here.
| 1,279.52 | 1,297.6 |
How to Build Custom Q&A Transformer Models in Python
|
2021-02-12 13:30:03 UTC
|
https://youtu.be/ZIRmXkHp0-c
|
ZIRmXkHp0-c
|
UCv83tO5cePwHMt1952IVVHw
|
ZIRmXkHp0-c-t1291.28
|
So here let's hope that there is a slight difference in what we see between both.
| 1,291.28 | 1,305.84 |
How to Build Custom Q&A Transformer Models in Python
|
2021-02-12 13:30:03 UTC
|
https://youtu.be/ZIRmXkHp0-c
|
ZIRmXkHp0-c
|
UCv83tO5cePwHMt1952IVVHw
|
ZIRmXkHp0-c-t1304.96
|
Okay, great.
| 1,304.96 | 1,307.12 |
How to Build Custom Q&A Transformer Models in Python
|
2021-02-12 13:30:03 UTC
|
https://youtu.be/ZIRmXkHp0-c
|
ZIRmXkHp0-c
|
UCv83tO5cePwHMt1952IVVHw
|
ZIRmXkHp0-c-t1305.84
|
That's what we would expect.
| 1,305.84 | 1,311.76 |
How to Build Custom Q&A Transformer Models in Python
|
2021-02-12 13:30:03 UTC
|
https://youtu.be/ZIRmXkHp0-c
|
ZIRmXkHp0-c
|
UCv83tO5cePwHMt1952IVVHw
|
ZIRmXkHp0-c-t1307.12
|
Okay, so now we have our data almost in the right format.
| 1,307.12 | 1,315.44 |
How to Build Custom Q&A Transformer Models in Python
|
2021-02-12 13:30:03 UTC
|
https://youtu.be/ZIRmXkHp0-c
|
ZIRmXkHp0-c
|
UCv83tO5cePwHMt1952IVVHw
|
ZIRmXkHp0-c-t1311.76
|
We just need to add the ending position.
| 1,311.76 | 1,321.44 |
How to Build Custom Q&A Transformer Models in Python
|
2021-02-12 13:30:03 UTC
|
https://youtu.be/ZIRmXkHp0-c
|
ZIRmXkHp0-c
|
UCv83tO5cePwHMt1952IVVHw
|
ZIRmXkHp0-c-t1316.8799999999999
|
So we already have the start position if we take a look in our train answers.
| 1,316.88 | 1,327.36 |
How to Build Custom Q&A Transformer Models in Python
|
2021-02-12 13:30:03 UTC
|
https://youtu.be/ZIRmXkHp0-c
|
ZIRmXkHp0-c
|
UCv83tO5cePwHMt1952IVVHw
|
ZIRmXkHp0-c-t1323.4399999999998
|
Okay, we have the answer start, but we also need the answer end.
| 1,323.44 | 1,331.36 |
How to Build Custom Q&A Transformer Models in Python
|
2021-02-12 13:30:03 UTC
|
https://youtu.be/ZIRmXkHp0-c
|
ZIRmXkHp0-c
|
UCv83tO5cePwHMt1952IVVHw
|
ZIRmXkHp0-c-t1328.7199999999998
|
And that's not included within the data.
| 1,328.72 | 1,338.08 |
How to Build Custom Q&A Transformer Models in Python
|
2021-02-12 13:30:03 UTC
|
https://youtu.be/ZIRmXkHp0-c
|
ZIRmXkHp0-c
|
UCv83tO5cePwHMt1952IVVHw
|
ZIRmXkHp0-c-t1331.36
|
So what we need to do here is actually define a function that will go through each one of our
| 1,331.36 | 1,344.4 |
How to Build Custom Q&A Transformer Models in Python
|
2021-02-12 13:30:03 UTC
|
https://youtu.be/ZIRmXkHp0-c
|
ZIRmXkHp0-c
|
UCv83tO5cePwHMt1952IVVHw
|
ZIRmXkHp0-c-t1338.8799999999999
|
answers and context and figure out where that ending character actually is.
| 1,338.88 | 1,350.24 |
How to Build Custom Q&A Transformer Models in Python
|
2021-02-12 13:30:03 UTC
|
https://youtu.be/ZIRmXkHp0-c
|
ZIRmXkHp0-c
|
UCv83tO5cePwHMt1952IVVHw
|
ZIRmXkHp0-c-t1345.9199999999998
|
And of course, we could just say, okay, it's the length of the text.
| 1,345.92 | 1,353.76 |
How to Build Custom Q&A Transformer Models in Python
|
2021-02-12 13:30:03 UTC
|
https://youtu.be/ZIRmXkHp0-c
|
ZIRmXkHp0-c
|
UCv83tO5cePwHMt1952IVVHw
|
ZIRmXkHp0-c-t1350.24
|
We add that onto the answer start and we have our answer end.
| 1,350.24 | 1,359.84 |
How to Build Custom Q&A Transformer Models in Python
|
2021-02-12 13:30:03 UTC
|
https://youtu.be/ZIRmXkHp0-c
|
ZIRmXkHp0-c
|
UCv83tO5cePwHMt1952IVVHw
|
ZIRmXkHp0-c-t1354.56
|
However, that unfortunately won't work because some of the
| 1,354.56 | 1,365.6 |
How to Build Custom Q&A Transformer Models in Python
|
2021-02-12 13:30:03 UTC
|
https://youtu.be/ZIRmXkHp0-c
|
ZIRmXkHp0-c
|
UCv83tO5cePwHMt1952IVVHw
|
ZIRmXkHp0-c-t1359.84
|
answer starts are actually incorrect and they're usually off by one or two characters.
| 1,359.84 | 1,373.2 |
How to Build Custom Q&A Transformer Models in Python
|
2021-02-12 13:30:03 UTC
|
https://youtu.be/ZIRmXkHp0-c
|
ZIRmXkHp0-c
|
UCv83tO5cePwHMt1952IVVHw
|
ZIRmXkHp0-c-t1366.3999999999999
|
So we actually need to go through and one, fix that and two, add our end indices.
| 1,366.4 | 1,378 |
How to Build Custom Q&A Transformer Models in Python
|
2021-02-12 13:30:03 UTC
|
https://youtu.be/ZIRmXkHp0-c
|
ZIRmXkHp0-c
|
UCv83tO5cePwHMt1952IVVHw
|
ZIRmXkHp0-c-t1375.1999999999998
|
So to do that, we're just going to define a new function.
| 1,375.2 | 1,384.96 |
How to Build Custom Q&A Transformer Models in Python
|
2021-02-12 13:30:03 UTC
|
https://youtu.be/ZIRmXkHp0-c
|
ZIRmXkHp0-c
|
UCv83tO5cePwHMt1952IVVHw
|
ZIRmXkHp0-c-t1380.9599999999998
|
It's going to be add end index.
| 1,380.96 | 1,389.52 |
How to Build Custom Q&A Transformer Models in Python
|
2021-02-12 13:30:03 UTC
|
https://youtu.be/ZIRmXkHp0-c
|
ZIRmXkHp0-c
|
UCv83tO5cePwHMt1952IVVHw
|
ZIRmXkHp0-c-t1384.96
|
And here we will have our answers and the context.
| 1,384.96 | 1,392.16 |
How to Build Custom Q&A Transformer Models in Python
|
2021-02-12 13:30:03 UTC
|
https://youtu.be/ZIRmXkHp0-c
|
ZIRmXkHp0-c
|
UCv83tO5cePwHMt1952IVVHw
|
ZIRmXkHp0-c-t1390.08
|
And then we're going to just feed these in.
| 1,390.08 | 1,414.88 |
How to Build Custom Q&A Transformer Models in Python
|
2021-02-12 13:30:03 UTC
|
https://youtu.be/ZIRmXkHp0-c
|
ZIRmXkHp0-c
|
UCv83tO5cePwHMt1952IVVHw
|
ZIRmXkHp0-c-t1392.16
|
So first thing we do is loop through each answer and context pair.
| 1,392.16 | 1,421.04 |
How to Build Custom Q&A Transformer Models in Python
|
2021-02-12 13:30:03 UTC
|
https://youtu.be/ZIRmXkHp0-c
|
ZIRmXkHp0-c
|
UCv83tO5cePwHMt1952IVVHw
|
ZIRmXkHp0-c-t1415.68
|
And then we extract something which is called the gold text, which is essentially the
| 1,415.68 | 1,423.68 |
How to Build Custom Q&A Transformer Models in Python
|
2021-02-12 13:30:03 UTC
|
https://youtu.be/ZIRmXkHp0-c
|
ZIRmXkHp0-c
|
UCv83tO5cePwHMt1952IVVHw
|
ZIRmXkHp0-c-t1421.04
|
answer that we are looking for.
| 1,421.04 | 1,426.32 |
How to Build Custom Q&A Transformer Models in Python
|
2021-02-12 13:30:03 UTC
|
https://youtu.be/ZIRmXkHp0-c
|
ZIRmXkHp0-c
|
UCv83tO5cePwHMt1952IVVHw
|
ZIRmXkHp0-c-t1423.68
|
It's called the golden text or gold text.
| 1,423.68 | 1,435.44 |
How to Build Custom Q&A Transformer Models in Python
|
2021-02-12 13:30:03 UTC
|
https://youtu.be/ZIRmXkHp0-c
|
ZIRmXkHp0-c
|
UCv83tO5cePwHMt1952IVVHw
|
ZIRmXkHp0-c-t1427.92
|
So simply our answer and within that the text.
| 1,427.92 | 1,438.24 |
How to Build Custom Q&A Transformer Models in Python
|
2021-02-12 13:30:03 UTC
|
https://youtu.be/ZIRmXkHp0-c
|
ZIRmXkHp0-c
|
UCv83tO5cePwHMt1952IVVHw
|
ZIRmXkHp0-c-t1436.48
|
So we are pulling this out here.
| 1,436.48 | 1,442.32 |
How to Build Custom Q&A Transformer Models in Python
|
2021-02-12 13:30:03 UTC
|
https://youtu.be/ZIRmXkHp0-c
|
ZIRmXkHp0-c
|
UCv83tO5cePwHMt1952IVVHw
|
ZIRmXkHp0-c-t1439.44
|
So we should already know the starting index.
| 1,439.44 | 1,451.12 |
How to Build Custom Q&A Transformer Models in Python
|
2021-02-12 13:30:03 UTC
|
https://youtu.be/ZIRmXkHp0-c
|
ZIRmXkHp0-c
|
UCv83tO5cePwHMt1952IVVHw
|
ZIRmXkHp0-c-t1442.32
|
So what we do here is simply pull that out as well.
| 1,442.32 | 1,466.16 |
How to Build Custom Q&A Transformer Models in Python
|
2021-02-12 13:30:03 UTC
|
https://youtu.be/ZIRmXkHp0-c
|
ZIRmXkHp0-c
|
UCv83tO5cePwHMt1952IVVHw
|
ZIRmXkHp0-c-t1454.32
|
And then the end index ideally will be the start plus the length of the gold text.
| 1,454.32 | 1,471.68 |
How to Build Custom Q&A Transformer Models in Python
|
2021-02-12 13:30:03 UTC
|
https://youtu.be/ZIRmXkHp0-c
|
ZIRmXkHp0-c
|
UCv83tO5cePwHMt1952IVVHw
|
ZIRmXkHp0-c-t1466.16
|
However, that's not always the case because like I said before, they can be off by one or two characters.
| 1,466.16 | 1,476.72 |
How to Build Custom Q&A Transformer Models in Python
|
2021-02-12 13:30:03 UTC
|
https://youtu.be/ZIRmXkHp0-c
|
ZIRmXkHp0-c
|
UCv83tO5cePwHMt1952IVVHw
|
ZIRmXkHp0-c-t1473.68
|
So we need to add in some logic just to deal with that.
| 1,473.68 | 1,482.48 |
How to Build Custom Q&A Transformer Models in Python
|
2021-02-12 13:30:03 UTC
|
https://youtu.be/ZIRmXkHp0-c
|
ZIRmXkHp0-c
|
UCv83tO5cePwHMt1952IVVHw
|
ZIRmXkHp0-c-t1477.76
|
So in our first case, let's assume that the characters are not off.
| 1,477.76 | 1,491.36 |
How to Build Custom Q&A Transformer Models in Python
|
2021-02-12 13:30:03 UTC
|
https://youtu.be/ZIRmXkHp0-c
|
ZIRmXkHp0-c
|
UCv83tO5cePwHMt1952IVVHw
|
ZIRmXkHp0-c-t1483.76
|
So if context start to end.
| 1,483.76 | 1,503.76 |
How to Build Custom Q&A Transformer Models in Python
|
2021-02-12 13:30:03 UTC
|
https://youtu.be/ZIRmXkHp0-c
|
ZIRmXkHp0-c
|
UCv83tO5cePwHMt1952IVVHw
|
ZIRmXkHp0-c-t1491.36
|
Start to end equals the gold text.
| 1,491.36 | 1,509.6 |
How to Build Custom Q&A Transformer Models in Python
|
2021-02-12 13:30:03 UTC
|
https://youtu.be/ZIRmXkHp0-c
|
ZIRmXkHp0-c
|
UCv83tO5cePwHMt1952IVVHw
|
ZIRmXkHp0-c-t1505.12
|
This means everything is good and we don't need to worry about it.
| 1,505.12 | 1,517.68 |
How to Build Custom Q&A Transformer Models in Python
|
2021-02-12 13:30:03 UTC
|
https://youtu.be/ZIRmXkHp0-c
|
ZIRmXkHp0-c
|
UCv83tO5cePwHMt1952IVVHw
|
ZIRmXkHp0-c-t1509.6
|
So we can modify the original dictionary and we can add answer end into there.
| 1,509.6 | 1,522.8 |
How to Build Custom Q&A Transformer Models in Python
|
2021-02-12 13:30:03 UTC
|
https://youtu.be/ZIRmXkHp0-c
|
ZIRmXkHp0-c
|
UCv83tO5cePwHMt1952IVVHw
|
ZIRmXkHp0-c-t1517.68
|
And we made that equal to our end index.
| 1,517.68 | 1,528.24 |
How to Build Custom Q&A Transformer Models in Python
|
2021-02-12 13:30:03 UTC
|
https://youtu.be/ZIRmXkHp0-c
|
ZIRmXkHp0-c
|
UCv83tO5cePwHMt1952IVVHw
|
ZIRmXkHp0-c-t1522.8
|
However, if that's not the case, that means we have a problem.
| 1,522.8 | 1,530.88 |
How to Build Custom Q&A Transformer Models in Python
|
2021-02-12 13:30:03 UTC
|
https://youtu.be/ZIRmXkHp0-c
|
ZIRmXkHp0-c
|
UCv83tO5cePwHMt1952IVVHw
|
ZIRmXkHp0-c-t1528.24
|
It's one of those dodgy question answer pairs.
| 1,528.24 | 1,539.04 |
How to Build Custom Q&A Transformer Models in Python
|
2021-02-12 13:30:03 UTC
|
https://youtu.be/ZIRmXkHp0-c
|
ZIRmXkHp0-c
|
UCv83tO5cePwHMt1952IVVHw
|
ZIRmXkHp0-c-t1532.16
|
And so this time what we can do is we'll add a else statement.
| 1,532.16 | 1,545.04 |
How to Build Custom Q&A Transformer Models in Python
|
2021-02-12 13:30:03 UTC
|
https://youtu.be/ZIRmXkHp0-c
|
ZIRmXkHp0-c
|
UCv83tO5cePwHMt1952IVVHw
|
ZIRmXkHp0-c-t1540.0800000000002
|
So we're just going to go through when the position is off by one or two characters because
| 1,540.08 | 1,548.48 |
How to Build Custom Q&A Transformer Models in Python
|
2021-02-12 13:30:03 UTC
|
https://youtu.be/ZIRmXkHp0-c
|
ZIRmXkHp0-c
|
UCv83tO5cePwHMt1952IVVHw
|
ZIRmXkHp0-c-t1545.04
|
it is not off by any more than that in the squad data set.
| 1,545.04 | 1,554.4 |
How to Build Custom Q&A Transformer Models in Python
|
2021-02-12 13:30:03 UTC
|
https://youtu.be/ZIRmXkHp0-c
|
ZIRmXkHp0-c
|
UCv83tO5cePwHMt1952IVVHw
|
ZIRmXkHp0-c-t1550.8
|
Loop through each of those and we'll say, OK, if the context.
| 1,550.8 | 1,561.6 |
How to Build Custom Q&A Transformer Models in Python
|
2021-02-12 13:30:03 UTC
|
https://youtu.be/ZIRmXkHp0-c
|
ZIRmXkHp0-c
|
UCv83tO5cePwHMt1952IVVHw
|
ZIRmXkHp0-c-t1556.32
|
And then in here, we need to add the start index and this again.
| 1,556.32 | 1,563.28 |
How to Build Custom Q&A Transformer Models in Python
|
2021-02-12 13:30:03 UTC
|
https://youtu.be/ZIRmXkHp0-c
|
ZIRmXkHp0-c
|
UCv83tO5cePwHMt1952IVVHw
|
ZIRmXkHp0-c-t1561.6
|
So let's just copy and paste that across.
| 1,561.6 | 1,563.84 |
How to Build Custom Q&A Transformer Models in Python
|
2021-02-12 13:30:03 UTC
|
https://youtu.be/ZIRmXkHp0-c
|
ZIRmXkHp0-c
|
UCv83tO5cePwHMt1952IVVHw
|
ZIRmXkHp0-c-t1563.28
|
Be easier.
| 1,563.28 | 1,570.08 |
How to Build Custom Q&A Transformer Models in Python
|
2021-02-12 13:30:03 UTC
|
https://youtu.be/ZIRmXkHp0-c
|
ZIRmXkHp0-c
|
UCv83tO5cePwHMt1952IVVHw
|
ZIRmXkHp0-c-t1565.76
|
But this time we're checking to see if it is off by one or two characters.
| 1,565.76 | 1,571.92 |
How to Build Custom Q&A Transformer Models in Python
|
2021-02-12 13:30:03 UTC
|
https://youtu.be/ZIRmXkHp0-c
|
ZIRmXkHp0-c
|
UCv83tO5cePwHMt1952IVVHw
|
ZIRmXkHp0-c-t1570.08
|
So just minus N.
| 1,570.08 | 1,575.68 |
How to Build Custom Q&A Transformer Models in Python
|
2021-02-12 13:30:03 UTC
|
https://youtu.be/ZIRmXkHp0-c
|
ZIRmXkHp0-c
|
UCv83tO5cePwHMt1952IVVHw
|
ZIRmXkHp0-c-t1571.92
|
And it's always minus and it isn't shifted.
| 1,571.92 | 1,578.72 |
How to Build Custom Q&A Transformer Models in Python
|
2021-02-12 13:30:03 UTC
|
https://youtu.be/ZIRmXkHp0-c
|
ZIRmXkHp0-c
|
UCv83tO5cePwHMt1952IVVHw
|
ZIRmXkHp0-c-t1575.68
|
It's always shifted to the left rather than shifted to the right.
| 1,575.68 | 1,580.32 |
How to Build Custom Q&A Transformer Models in Python
|
2021-02-12 13:30:03 UTC
|
https://youtu.be/ZIRmXkHp0-c
|
ZIRmXkHp0-c
|
UCv83tO5cePwHMt1952IVVHw
|
ZIRmXkHp0-c-t1578.72
|
So that's this is fine.
| 1,578.72 | 1,587.44 |
How to Build Custom Q&A Transformer Models in Python
|
2021-02-12 13:30:03 UTC
|
https://youtu.be/ZIRmXkHp0-c
|
ZIRmXkHp0-c
|
UCv83tO5cePwHMt1952IVVHw
|
ZIRmXkHp0-c-t1582.72
|
So in this case, the answer is off by end tokens.
| 1,582.72 | 1,601.04 |
How to Build Custom Q&A Transformer Models in Python
|
2021-02-12 13:30:03 UTC
|
https://youtu.be/ZIRmXkHp0-c
|
ZIRmXkHp0-c
|
UCv83tO5cePwHMt1952IVVHw
|
ZIRmXkHp0-c-t1587.44
|
And so we need to update our answer start value and also add our answer end value.
| 1,587.44 | 1,618.64 |
How to Build Custom Q&A Transformer Models in Python
|
2021-02-12 13:30:03 UTC
|
https://youtu.be/ZIRmXkHp0-c
|
ZIRmXkHp0-c
|
UCv83tO5cePwHMt1952IVVHw
|
ZIRmXkHp0-c-t1601.04
|
So start index minus N and we also have the end.
| 1,601.04 | 1,624 |
How to Build Custom Q&A Transformer Models in Python
|
2021-02-12 13:30:03 UTC
|
https://youtu.be/ZIRmXkHp0-c
|
ZIRmXkHp0-c
|
UCv83tO5cePwHMt1952IVVHw
|
ZIRmXkHp0-c-t1622.6399999999999
|
So that's great.
| 1,622.64 | 1,628.88 |
How to Build Custom Q&A Transformer Models in Python
|
2021-02-12 13:30:03 UTC
|
https://youtu.be/ZIRmXkHp0-c
|
ZIRmXkHp0-c
|
UCv83tO5cePwHMt1952IVVHw
|
ZIRmXkHp0-c-t1624.0
|
We can take that and we can apply it to our train and validation sets.
| 1,624 | 1,635.76 |
How to Build Custom Q&A Transformer Models in Python
|
2021-02-12 13:30:03 UTC
|
https://youtu.be/ZIRmXkHp0-c
|
ZIRmXkHp0-c
|
UCv83tO5cePwHMt1952IVVHw
|
ZIRmXkHp0-c-t1628.88
|
So all we do here is call the function.
| 1,628.88 | 1,642.08 |
How to Build Custom Q&A Transformer Models in Python
|
2021-02-12 13:30:03 UTC
|
https://youtu.be/ZIRmXkHp0-c
|
ZIRmXkHp0-c
|
UCv83tO5cePwHMt1952IVVHw
|
ZIRmXkHp0-c-t1635.7600000000002
|
And we just see train answers and train context.
| 1,635.76 | 1,659.04 |
How to Build Custom Q&A Transformer Models in Python
|
2021-02-12 13:30:03 UTC
|
https://youtu.be/ZIRmXkHp0-c
|
ZIRmXkHp0-c
|
UCv83tO5cePwHMt1952IVVHw
|
ZIRmXkHp0-c-t1642.08
|
And of course, we can just copy this and do the same for our validation set.
| 1,642.08 | 1,660.24 |
How to Build Custom Q&A Transformer Models in Python
|
2021-02-12 13:30:03 UTC
|
https://youtu.be/ZIRmXkHp0-c
|
ZIRmXkHp0-c
|
UCv83tO5cePwHMt1952IVVHw
|
ZIRmXkHp0-c-t1659.04
|
OK, perfect.
| 1,659.04 | 1,665.36 |
How to Build Custom Q&A Transformer Models in Python
|
2021-02-12 13:30:03 UTC
|
https://youtu.be/ZIRmXkHp0-c
|
ZIRmXkHp0-c
|
UCv83tO5cePwHMt1952IVVHw
|
ZIRmXkHp0-c-t1660.24
|
So now if we have a quick look, we should be able to see that we have
| 1,660.24 | 1,671.6 |
Subsets and Splits
No community queries yet
The top public SQL queries from the community will appear here once available.