Datasets:
jamescalam
/
youtube-transcriptions

Dataset card Data Studio Files
xet
Community
2
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-t3311.32
for training and we will output the results of that training batch to the
3,311.32
3,329.48
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-t3318.56
outputs variable. Our model, the IDs, we need the attention mask.
3,318.56
3,341.28
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-t3329.48
We also want our start positions and end positions.
3,329.48
3,361.8
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-t3341.28
Now from our training batch we want to extract the loss.
3,341.28
3,383.52
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-t3372.2400000000002
And then we want to calculate loss for every parameter and this is for our
3,372.24
3,391.48
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-t3383.52
gradient update and then we use the step method here to actually update those
3,383.52
3,398.72
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-t3391.48
gradients. And then this final little bit here is purely for us to see this is our
3,391.48
3,405.64
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-t3398.72
progress bar. So we call the loop we set the description which is going to be our
3,398.72
3,420.04
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-t3405.64
epoch. And then it would probably be quite useful to also see the loss in there as
3,405.64
3,430.56
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-t3420.04
well. We will set that as a post fix so it will appear after the progress bar.
3,420.04
3,446.6
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-t3438.84
Okay and that should be everything. Okay so that looks pretty good. We have our
3,438.84
3,453.64
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-t3446.6
model training and as I said this will take a little bit of time. So I will let
3,446.6
3,466.72
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-t3453.64
that run. Okay so we have this non type error here and this is because within
3,453.64
3,471.24
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-t3466.72
our end positions we normally expect integers but we're also getting some
3,466.72
3,476.12
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-t3471.24
non values because the code that we used earlier where we're checking if end
3,471.24
3,482.16
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-t3476.12
position is non essentially wasn't good enough. So as a fix for that we'll just
3,476.12
3,487.56
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-t3482.16
go back and we'll add like a while loop which will keep checking if it's non and
3,482.16
3,497.6
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-t3487.56
every time it is non reduce the value that we are seeing by one. So go back up
3,487.56
3,502.48
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-t3497.6
here and this is where the problem is coming from. So we're just going to
3,497.6
3,514.92
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-t3502.48
change this to be a while loop and just initialize a essentially a counter here.
3,502.48
3,525.04
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-t3514.92
And we'll use this as our go back value and every time the end position is still
3,514.92
3,533.72
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-t3525.04
non we'll just add one to go back and this should work.
3,525.04
3,560.28
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-t3533.72
Just need to remember to rerun anything we need to rerun. Yeah. Okay and that
3,533.72
3,566.64
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-t3560.28
looks like it solved the issue. So great we can just leave that training for a
3,560.28
3,572.04
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-t3566.6400000000003
little while and I will see you when it's done.
3,566.64
3,580.52
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-t3572.6000000000004
Okay so the model is finished and we'll go ahead and just save it. So obviously
3,572.6
3,589.88
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-t3580.52
we'll need to do that whenever actually doing this on any other projects. So I'm
3,580.52
3,597.08
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-t3589.88
just going to call it the Silbert custom and it's super easy to save we just do
3,589.88
3,608.76
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-t3597.08
save pre-trained and the model path. Now as well as this we might also want to
3,597.08
3,614.76
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-t3608.76
save the tokenizer so we have everything in one place. So to do that we also just
3,608.76
3,627.12
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-t3614.76
use tokenizer and save pre-trained again. Okay so if we go back into our folder
3,614.76
3,633.16
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-t3627.1200000000003
here, receive models and we have this Silbert custom and then in here we have
3,627.12
3,639.32
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-t3633.1600000000003
all of the files we need to build our PyTorch model. It's a little bit different
3,633.16
3,644.6
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-t3639.32
if we're using TensorFlow but the actual saving process is practically the same.
3,639.32
3,651.84
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-t3644.6
So now we've finished training we want to switch it out of the training mode so
3,644.6
3,659.4
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-t3651.8399999999997
we use a model eval. I'm just get all this information about our model as well
3,651.84
3,664.68
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-t3659.4
we don't actually need any of that and just like before we want to create a
3,659.4
3,672.6
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-t3664.68
data loader. So for that I'm gonna call it val loader and it's exactly the same
3,664.68
3,676.6
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-t3672.6
code as before in fact it's probably better if we just copy and paste some of
3,672.6
3,687.24
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-t3676.6
this. At least the loop. So what we're gonna do here is take the same loop and
3,676.6
3,693.96
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-t3687.24
apply it as a validation run with our validation data. Just paste that there
3,687.24
3,701.16
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-t3693.96
we'll initialize this data loader this time of course with the validation set
3,693.96
3,712.8
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-t3701.16
we'll stick with the same batch size. Now this time we do want to keep a log of
3,701.16
3,719.04
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-t3712.7999999999997
accuracy so we will keep that there and we also don't need to run most for epochs
3,712.8
3,723.04
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-t3719.04
because we're not training this time we're just running through all the
3,719.04
3,730.44
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-t3723.04
batches within our loop of validation data. So this is now a validation loader and
3,723.04
3,734.72
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-t3730.44
we just loop through each of those batches so we don't need to do anything
3,730.44
3,741.08
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-t3734.7200000000003
with the gradients here and because we're not doing anything to gradients we
3,734.72
3,750.08
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-t3741.08
actually add this in to stop PyTorch from calculating any gradients. This
3,741.08
3,756.36
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-t3750.08
will obviously save us a bit of time when we're processing all of this. And we
3,750.08
3,763.68
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-t3756.36
put those in there. The outputs we do so want this but of course we don't need to
3,756.36
3,768.84
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-t3763.6800000000003
be putting in the start and end positions so we can remove those and
3,763.68
3,777.04
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-t3768.84
this time we want to pull out the start prediction and end prediction. So if we
3,768.84
3,784.24
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-t3777.04
have a look at what our outputs look like before you see we have this model
3,777.04
3,790.52
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-t3784.24
output and within here we're a few different tensors which each have a
3,784.24
3,801.4
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-t3790.52
accessible name. So the ones that we care about are startLogits and that
3,790.52
3,809.92
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-t3801.3999999999996
will give us the logits for our start position which is essentially like a set
3,801.4
3,816
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-t3809.92
of predictions where the highest value within that vector represents the token
3,809.92
3,827.92
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-t3816.0
ID. So we can do that for both. You'll see we get these tensors. Now we only want
3,816
3,833.44
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-t3827.92
the largest value in each one of these vectors here because that will give us
3,827.92
3,846.68
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-t3833.44
the input ID. So to get that we use the argmax function and if we just use it by
3,833.44
3,852.58
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-t3846.68
itself that will give us the maximum index within the whole thing. But we
3,846.68
3,857.8
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-t3852.58
don't want that we want one for every single vector or every row and to do
3,852.58
3,864.32
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-t3857.8
that we just set dim equal to 1 and there you go we get a full batch of
3,857.8
3,870.76
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-t3864.32
outputs. So these are our starting positions and then we also want to do
3,864.32
3,880.76
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-t3870.76
the same for our ending positions. So we just change start to end. So it's pretty
3,870.76
3,888.24
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-t3880.76
easy. Now obviously we want to be doing this within our loop because this is
3,880.76
3,896.72
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-t3888.2400000000002
only doing one batch and we need to do this for every single batch. So we're
3,888.24
3,912.8
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-t3896.72
just going to assign them to a variable and there we have our predictions and
3,896.72
3,919.12
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-t3912.7999999999997
all we need to do now is check for an exact match. So what I mean by exact
3,912.8
3,925.64
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-t3919.12
match is we want to see whether the start positions here which we can rename
3,919.12
3,933.68
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-t3925.64
to the true values whether these are equal to the predicted values down here
3,925.64
3,945.76
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-t3933.68
and to calculate that so let me just run this so we have one batch. That
3,933.68
3,950.88
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-t3945.7599999999998
shouldn't take too long to process and we can just write the code out. So to
3,945.76
3,960.68
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-t3950.88
check this we just use the double equal syntax here and this will just check for
3,950.88
3,968.2
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-t3960.6800000000003
matches between two arrays. So we have the start predictions and the start true
3,960.68
3,984.24
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-t3968.2
values so we'll check for those. Okay so if we just have a look at what we have
3,968.2
3,990.36
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-t3984.24
here we get this array of true or false. So these ones don't look particularly
3,984.24
3,997.16
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-t3990.3599999999997
good but that's fine we just want to calculate the accuracy here. So take the
3,990.36
4,014.08
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-t3997.16
sum and we also want to divide that by the length. Okay so that will give us our
3,997.16
4,020.64
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-t4014.08
accuracy within the tensor and we just take it out using the item method but we
4,014.08
4,024.16
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-t4020.64
also just need to include brackets around this because at the moment we're
4,020.64
4,031.04
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-t4024.16
trying to take item of the length value. Okay and that gives us our very poor
4,024.16
4,037.36
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-t4031.04
accuracy on this final batch. So we can take that and within here we want to
4,031.04
4,047
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-t4037.3599999999997
append that to our accuracy list and then we also want to do that for the end
4,037.36
4,067.64
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-t4047.0
the prediction as well. And we'll just let that run through and then we can
4,047
4,073.56
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-t4067.64
calculate our accuracy from the end of that. And then we can have a quick look
4,067.64
4,081.32
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-t4073.56
at our accuracy here. We can see fortunately it's not as bad as it first
4,073.56
4,094.72
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-t4081.32
seemed. So we're getting a lot of 93%, 100%, 81% that's generally pretty good. So
4,081.32
4,101.8
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-t4094.72
of course if we want to get the overall accuracy all we do is sum that and
4,094.72
4,115.08
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-t4101.8
divide by the length. And we get 63.6% for an exact match accuracy. So what I
4,101.8
4,120.64
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-t4115.08
mean by exact match is say if we take a look at a few of these that do not match
4,115.08
4,129.16
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-t4120.64
so we have a 75% match on the fourth batch although that won't be particularly
4,120.64
4,135.6
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-t4129.16
useful because we can't see that batch right now. So let's just take the last
4,129.16
4,144.84
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-t4135.599999999999
batch because we have these values here. Now if we look at what start true is we
4,135.6
4,155.28
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-t4144.84
get these values. Then if we look at start pred we get this. So none of these
4,144.84
4,162.28
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-t4155.28
match but a couple of them do get pretty close. So these final four all of these
4,155.28
4,168.16
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-t4162.28
count as 0% on the exact match. But in reality if you look at what we
4,162.28
4,173.24
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-t4168.16
predicted for every single one of them it's predicting just one token
4,168.16
4,176.72
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-t4173.24
before so it's getting quite close but it's not an exact match so it scores
4,173.24
4,186.04