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
|
|---|---|---|---|---|---|---|---|---|
Sentence Similarity With Transformers and PyTorch (Python)
|
2021-05-05 15:00:20 UTC
|
https://youtu.be/jVPd7lEvjtg
|
jVPd7lEvjtg
|
UCv83tO5cePwHMt1952IVVHw
|
jVPd7lEvjtg-t308.0
|
So tokenizer encode plus.
| 308 | 315 |
Sentence Similarity With Transformers and PyTorch (Python)
|
2021-05-05 15:00:20 UTC
|
https://youtu.be/jVPd7lEvjtg
|
jVPd7lEvjtg
|
UCv83tO5cePwHMt1952IVVHw
|
jVPd7lEvjtg-t312.0
|
And then in here, we need to pass our sentence.
| 312 | 319 |
Sentence Similarity With Transformers and PyTorch (Python)
|
2021-05-05 15:00:20 UTC
|
https://youtu.be/jVPd7lEvjtg
|
jVPd7lEvjtg
|
UCv83tO5cePwHMt1952IVVHw
|
jVPd7lEvjtg-t315.0
|
We need to pass the maximum length of our sequence.
| 315 | 323 |
Sentence Similarity With Transformers and PyTorch (Python)
|
2021-05-05 15:00:20 UTC
|
https://youtu.be/jVPd7lEvjtg
|
jVPd7lEvjtg
|
UCv83tO5cePwHMt1952IVVHw
|
jVPd7lEvjtg-t319.0
|
So with BERT, usually we would set this to 512.
| 319 | 327 |
Sentence Similarity With Transformers and PyTorch (Python)
|
2021-05-05 15:00:20 UTC
|
https://youtu.be/jVPd7lEvjtg
|
jVPd7lEvjtg
|
UCv83tO5cePwHMt1952IVVHw
|
jVPd7lEvjtg-t323.0
|
But because we're using this BERT based NLIME tokens model,
| 323 | 330 |
Sentence Similarity With Transformers and PyTorch (Python)
|
2021-05-05 15:00:20 UTC
|
https://youtu.be/jVPd7lEvjtg
|
jVPd7lEvjtg
|
UCv83tO5cePwHMt1952IVVHw
|
jVPd7lEvjtg-t327.0
|
this should actually be set to 128.
| 327 | 335 |
Sentence Similarity With Transformers and PyTorch (Python)
|
2021-05-05 15:00:20 UTC
|
https://youtu.be/jVPd7lEvjtg
|
jVPd7lEvjtg
|
UCv83tO5cePwHMt1952IVVHw
|
jVPd7lEvjtg-t330.0
|
So we set max length to 128.
| 330 | 338 |
Sentence Similarity With Transformers and PyTorch (Python)
|
2021-05-05 15:00:20 UTC
|
https://youtu.be/jVPd7lEvjtg
|
jVPd7lEvjtg
|
UCv83tO5cePwHMt1952IVVHw
|
jVPd7lEvjtg-t335.0
|
And anything longer than this, we want to truncate.
| 335 | 342 |
Sentence Similarity With Transformers and PyTorch (Python)
|
2021-05-05 15:00:20 UTC
|
https://youtu.be/jVPd7lEvjtg
|
jVPd7lEvjtg
|
UCv83tO5cePwHMt1952IVVHw
|
jVPd7lEvjtg-t338.0
|
So we set truncation equal to true.
| 338 | 346 |
Sentence Similarity With Transformers and PyTorch (Python)
|
2021-05-05 15:00:20 UTC
|
https://youtu.be/jVPd7lEvjtg
|
jVPd7lEvjtg
|
UCv83tO5cePwHMt1952IVVHw
|
jVPd7lEvjtg-t342.0
|
And anything shorter than this, which they all will be in our case,
| 342 | 349 |
Sentence Similarity With Transformers and PyTorch (Python)
|
2021-05-05 15:00:20 UTC
|
https://youtu.be/jVPd7lEvjtg
|
jVPd7lEvjtg
|
UCv83tO5cePwHMt1952IVVHw
|
jVPd7lEvjtg-t346.0
|
we set padding equal to the max length.
| 346 | 352 |
Sentence Similarity With Transformers and PyTorch (Python)
|
2021-05-05 15:00:20 UTC
|
https://youtu.be/jVPd7lEvjtg
|
jVPd7lEvjtg
|
UCv83tO5cePwHMt1952IVVHw
|
jVPd7lEvjtg-t349.0
|
To pad it up to that max length.
| 349 | 356 |
Sentence Similarity With Transformers and PyTorch (Python)
|
2021-05-05 15:00:20 UTC
|
https://youtu.be/jVPd7lEvjtg
|
jVPd7lEvjtg
|
UCv83tO5cePwHMt1952IVVHw
|
jVPd7lEvjtg-t352.0
|
And then here, we want to say return tensors.
| 352 | 362 |
Sentence Similarity With Transformers and PyTorch (Python)
|
2021-05-05 15:00:20 UTC
|
https://youtu.be/jVPd7lEvjtg
|
jVPd7lEvjtg
|
UCv83tO5cePwHMt1952IVVHw
|
jVPd7lEvjtg-t356.0
|
And we set this equal to PT, because we're using PyTorch.
| 356 | 366 |
Sentence Similarity With Transformers and PyTorch (Python)
|
2021-05-05 15:00:20 UTC
|
https://youtu.be/jVPd7lEvjtg
|
jVPd7lEvjtg
|
UCv83tO5cePwHMt1952IVVHw
|
jVPd7lEvjtg-t362.0
|
Now this will return a dictionary containing input IDs
| 362 | 369 |
Sentence Similarity With Transformers and PyTorch (Python)
|
2021-05-05 15:00:20 UTC
|
https://youtu.be/jVPd7lEvjtg
|
jVPd7lEvjtg
|
UCv83tO5cePwHMt1952IVVHw
|
jVPd7lEvjtg-t366.0
|
and attention masks for a single sentence.
| 366 | 377 |
Sentence Similarity With Transformers and PyTorch (Python)
|
2021-05-05 15:00:20 UTC
|
https://youtu.be/jVPd7lEvjtg
|
jVPd7lEvjtg
|
UCv83tO5cePwHMt1952IVVHw
|
jVPd7lEvjtg-t369.0
|
So we'll take the new tokens, assign it to that variable.
| 369 | 382 |
Sentence Similarity With Transformers and PyTorch (Python)
|
2021-05-05 15:00:20 UTC
|
https://youtu.be/jVPd7lEvjtg
|
jVPd7lEvjtg
|
UCv83tO5cePwHMt1952IVVHw
|
jVPd7lEvjtg-t377.0
|
And then what we're going to do is access our tokens dictionary.
| 377 | 385 |
Sentence Similarity With Transformers and PyTorch (Python)
|
2021-05-05 15:00:20 UTC
|
https://youtu.be/jVPd7lEvjtg
|
jVPd7lEvjtg
|
UCv83tO5cePwHMt1952IVVHw
|
jVPd7lEvjtg-t382.0
|
Which inputs IDs first.
| 382 | 390 |
Sentence Similarity With Transformers and PyTorch (Python)
|
2021-05-05 15:00:20 UTC
|
https://youtu.be/jVPd7lEvjtg
|
jVPd7lEvjtg
|
UCv83tO5cePwHMt1952IVVHw
|
jVPd7lEvjtg-t385.0
|
And append the input IDs for the single sentence
| 385 | 393 |
Sentence Similarity With Transformers and PyTorch (Python)
|
2021-05-05 15:00:20 UTC
|
https://youtu.be/jVPd7lEvjtg
|
jVPd7lEvjtg
|
UCv83tO5cePwHMt1952IVVHw
|
jVPd7lEvjtg-t390.0
|
from the new tokens variable.
| 390 | 396 |
Sentence Similarity With Transformers and PyTorch (Python)
|
2021-05-05 15:00:20 UTC
|
https://youtu.be/jVPd7lEvjtg
|
jVPd7lEvjtg
|
UCv83tO5cePwHMt1952IVVHw
|
jVPd7lEvjtg-t393.0
|
So input IDs.
| 393 | 400 |
Sentence Similarity With Transformers and PyTorch (Python)
|
2021-05-05 15:00:20 UTC
|
https://youtu.be/jVPd7lEvjtg
|
jVPd7lEvjtg
|
UCv83tO5cePwHMt1952IVVHw
|
jVPd7lEvjtg-t396.0
|
And then we do the same for our attention mask.
| 396 | 406 |
Sentence Similarity With Transformers and PyTorch (Python)
|
2021-05-05 15:00:20 UTC
|
https://youtu.be/jVPd7lEvjtg
|
jVPd7lEvjtg
|
UCv83tO5cePwHMt1952IVVHw
|
jVPd7lEvjtg-t404.0
|
Okay.
| 404 | 411 |
Sentence Similarity With Transformers and PyTorch (Python)
|
2021-05-05 15:00:20 UTC
|
https://youtu.be/jVPd7lEvjtg
|
jVPd7lEvjtg
|
UCv83tO5cePwHMt1952IVVHw
|
jVPd7lEvjtg-t406.0
|
So that gives us those.
| 406 | 413 |
Sentence Similarity With Transformers and PyTorch (Python)
|
2021-05-05 15:00:20 UTC
|
https://youtu.be/jVPd7lEvjtg
|
jVPd7lEvjtg
|
UCv83tO5cePwHMt1952IVVHw
|
jVPd7lEvjtg-t411.0
|
There's another thing as well.
| 411 | 416 |
Sentence Similarity With Transformers and PyTorch (Python)
|
2021-05-05 15:00:20 UTC
|
https://youtu.be/jVPd7lEvjtg
|
jVPd7lEvjtg
|
UCv83tO5cePwHMt1952IVVHw
|
jVPd7lEvjtg-t413.0
|
These are wrapped as vectors.
| 413 | 420 |
Sentence Similarity With Transformers and PyTorch (Python)
|
2021-05-05 15:00:20 UTC
|
https://youtu.be/jVPd7lEvjtg
|
jVPd7lEvjtg
|
UCv83tO5cePwHMt1952IVVHw
|
jVPd7lEvjtg-t416.0
|
So we also want to just extract the first element there.
| 416 | 427 |
Sentence Similarity With Transformers and PyTorch (Python)
|
2021-05-05 15:00:20 UTC
|
https://youtu.be/jVPd7lEvjtg
|
jVPd7lEvjtg
|
UCv83tO5cePwHMt1952IVVHw
|
jVPd7lEvjtg-t420.0
|
Because they're almost like lists within a list, but in tensor format.
| 420 | 430 |
Sentence Similarity With Transformers and PyTorch (Python)
|
2021-05-05 15:00:20 UTC
|
https://youtu.be/jVPd7lEvjtg
|
jVPd7lEvjtg
|
UCv83tO5cePwHMt1952IVVHw
|
jVPd7lEvjtg-t427.0
|
And we want to extract the list.
| 427 | 432 |
Sentence Similarity With Transformers and PyTorch (Python)
|
2021-05-05 15:00:20 UTC
|
https://youtu.be/jVPd7lEvjtg
|
jVPd7lEvjtg
|
UCv83tO5cePwHMt1952IVVHw
|
jVPd7lEvjtg-t430.0
|
Now that's good.
| 430 | 434 |
Sentence Similarity With Transformers and PyTorch (Python)
|
2021-05-05 15:00:20 UTC
|
https://youtu.be/jVPd7lEvjtg
|
jVPd7lEvjtg
|
UCv83tO5cePwHMt1952IVVHw
|
jVPd7lEvjtg-t432.0
|
But obviously we're using PyTorch here.
| 432 | 438 |
Sentence Similarity With Transformers and PyTorch (Python)
|
2021-05-05 15:00:20 UTC
|
https://youtu.be/jVPd7lEvjtg
|
jVPd7lEvjtg
|
UCv83tO5cePwHMt1952IVVHw
|
jVPd7lEvjtg-t434.0
|
We want PyTorch tensors, not lists.
| 434 | 442 |
Sentence Similarity With Transformers and PyTorch (Python)
|
2021-05-05 15:00:20 UTC
|
https://youtu.be/jVPd7lEvjtg
|
jVPd7lEvjtg
|
UCv83tO5cePwHMt1952IVVHw
|
jVPd7lEvjtg-t438.0
|
So within these lists, we do have PyTorch tensors.
| 438 | 447 |
Sentence Similarity With Transformers and PyTorch (Python)
|
2021-05-05 15:00:20 UTC
|
https://youtu.be/jVPd7lEvjtg
|
jVPd7lEvjtg
|
UCv83tO5cePwHMt1952IVVHw
|
jVPd7lEvjtg-t442.0
|
So in fact, let me just show you.
| 442 | 451 |
Sentence Similarity With Transformers and PyTorch (Python)
|
2021-05-05 15:00:20 UTC
|
https://youtu.be/jVPd7lEvjtg
|
jVPd7lEvjtg
|
UCv83tO5cePwHMt1952IVVHw
|
jVPd7lEvjtg-t447.0
|
So if we have a look in here.
| 447 | 456 |
Sentence Similarity With Transformers and PyTorch (Python)
|
2021-05-05 15:00:20 UTC
|
https://youtu.be/jVPd7lEvjtg
|
jVPd7lEvjtg
|
UCv83tO5cePwHMt1952IVVHw
|
jVPd7lEvjtg-t451.0
|
We'll see that we have our PyTorch tensors.
| 451 | 460 |
Sentence Similarity With Transformers and PyTorch (Python)
|
2021-05-05 15:00:20 UTC
|
https://youtu.be/jVPd7lEvjtg
|
jVPd7lEvjtg
|
UCv83tO5cePwHMt1952IVVHw
|
jVPd7lEvjtg-t456.0
|
But they're contained within a normal Python list.
| 456 | 463 |
Sentence Similarity With Transformers and PyTorch (Python)
|
2021-05-05 15:00:20 UTC
|
https://youtu.be/jVPd7lEvjtg
|
jVPd7lEvjtg
|
UCv83tO5cePwHMt1952IVVHw
|
jVPd7lEvjtg-t460.0
|
So we can even check that.
| 460 | 465 |
Sentence Similarity With Transformers and PyTorch (Python)
|
2021-05-05 15:00:20 UTC
|
https://youtu.be/jVPd7lEvjtg
|
jVPd7lEvjtg
|
UCv83tO5cePwHMt1952IVVHw
|
jVPd7lEvjtg-t463.0
|
We do type.
| 463 | 466 |
Sentence Similarity With Transformers and PyTorch (Python)
|
2021-05-05 15:00:20 UTC
|
https://youtu.be/jVPd7lEvjtg
|
jVPd7lEvjtg
|
UCv83tO5cePwHMt1952IVVHw
|
jVPd7lEvjtg-t465.0
|
We see that we get lists.
| 465 | 469 |
Sentence Similarity With Transformers and PyTorch (Python)
|
2021-05-05 15:00:20 UTC
|
https://youtu.be/jVPd7lEvjtg
|
jVPd7lEvjtg
|
UCv83tO5cePwHMt1952IVVHw
|
jVPd7lEvjtg-t466.0
|
And inside there, we have the Torch tensor.
| 466 | 471 |
Sentence Similarity With Transformers and PyTorch (Python)
|
2021-05-05 15:00:20 UTC
|
https://youtu.be/jVPd7lEvjtg
|
jVPd7lEvjtg
|
UCv83tO5cePwHMt1952IVVHw
|
jVPd7lEvjtg-t469.0
|
Which is what we want for all of them.
| 469 | 478 |
Sentence Similarity With Transformers and PyTorch (Python)
|
2021-05-05 15:00:20 UTC
|
https://youtu.be/jVPd7lEvjtg
|
jVPd7lEvjtg
|
UCv83tO5cePwHMt1952IVVHw
|
jVPd7lEvjtg-t471.0
|
So to convert this list of PyTorch tensors into a single PyTorch tensor.
| 471 | 483 |
Sentence Similarity With Transformers and PyTorch (Python)
|
2021-05-05 15:00:20 UTC
|
https://youtu.be/jVPd7lEvjtg
|
jVPd7lEvjtg
|
UCv83tO5cePwHMt1952IVVHw
|
jVPd7lEvjtg-t478.0
|
What we do is we take this Torch.
| 478 | 489 |
Sentence Similarity With Transformers and PyTorch (Python)
|
2021-05-05 15:00:20 UTC
|
https://youtu.be/jVPd7lEvjtg
|
jVPd7lEvjtg
|
UCv83tO5cePwHMt1952IVVHw
|
jVPd7lEvjtg-t483.0
|
And we use the stack method.
| 483 | 494 |
Sentence Similarity With Transformers and PyTorch (Python)
|
2021-05-05 15:00:20 UTC
|
https://youtu.be/jVPd7lEvjtg
|
jVPd7lEvjtg
|
UCv83tO5cePwHMt1952IVVHw
|
jVPd7lEvjtg-t489.0
|
And what the stack method does is takes a list.
| 489 | 497 |
Sentence Similarity With Transformers and PyTorch (Python)
|
2021-05-05 15:00:20 UTC
|
https://youtu.be/jVPd7lEvjtg
|
jVPd7lEvjtg
|
UCv83tO5cePwHMt1952IVVHw
|
jVPd7lEvjtg-t494.0
|
And within that list we'll expect PyTorch tensors.
| 494 | 500 |
Sentence Similarity With Transformers and PyTorch (Python)
|
2021-05-05 15:00:20 UTC
|
https://youtu.be/jVPd7lEvjtg
|
jVPd7lEvjtg
|
UCv83tO5cePwHMt1952IVVHw
|
jVPd7lEvjtg-t497.0
|
And it will stack all of those on top of each other.
| 497 | 502 |
Sentence Similarity With Transformers and PyTorch (Python)
|
2021-05-05 15:00:20 UTC
|
https://youtu.be/jVPd7lEvjtg
|
jVPd7lEvjtg
|
UCv83tO5cePwHMt1952IVVHw
|
jVPd7lEvjtg-t500.0
|
Essentially adding another dimension.
| 500 | 505 |
Sentence Similarity With Transformers and PyTorch (Python)
|
2021-05-05 15:00:20 UTC
|
https://youtu.be/jVPd7lEvjtg
|
jVPd7lEvjtg
|
UCv83tO5cePwHMt1952IVVHw
|
jVPd7lEvjtg-t502.0
|
And stacking them all on top of each other.
| 502 | 508 |
Sentence Similarity With Transformers and PyTorch (Python)
|
2021-05-05 15:00:20 UTC
|
https://youtu.be/jVPd7lEvjtg
|
jVPd7lEvjtg
|
UCv83tO5cePwHMt1952IVVHw
|
jVPd7lEvjtg-t505.0
|
Hence the name.
| 505 | 509 |
Sentence Similarity With Transformers and PyTorch (Python)
|
2021-05-05 15:00:20 UTC
|
https://youtu.be/jVPd7lEvjtg
|
jVPd7lEvjtg
|
UCv83tO5cePwHMt1952IVVHw
|
jVPd7lEvjtg-t508.0
|
So take that.
| 508 | 516 |
Sentence Similarity With Transformers and PyTorch (Python)
|
2021-05-05 15:00:20 UTC
|
https://youtu.be/jVPd7lEvjtg
|
jVPd7lEvjtg
|
UCv83tO5cePwHMt1952IVVHw
|
jVPd7lEvjtg-t509.0
|
And we want to do it for both input IDs and attention mask.
| 509 | 517 |
Sentence Similarity With Transformers and PyTorch (Python)
|
2021-05-05 15:00:20 UTC
|
https://youtu.be/jVPd7lEvjtg
|
jVPd7lEvjtg
|
UCv83tO5cePwHMt1952IVVHw
|
jVPd7lEvjtg-t516.0
|
And then let's have a look at what we have.
| 516 | 522 |
Sentence Similarity With Transformers and PyTorch (Python)
|
2021-05-05 15:00:20 UTC
|
https://youtu.be/jVPd7lEvjtg
|
jVPd7lEvjtg
|
UCv83tO5cePwHMt1952IVVHw
|
jVPd7lEvjtg-t517.0
|
So let's go attention or input IDs.
| 517 | 525 |
Sentence Similarity With Transformers and PyTorch (Python)
|
2021-05-05 15:00:20 UTC
|
https://youtu.be/jVPd7lEvjtg
|
jVPd7lEvjtg
|
UCv83tO5cePwHMt1952IVVHw
|
jVPd7lEvjtg-t522.0
|
And now we just have a single tensor.
| 522 | 532 |
Sentence Similarity With Transformers and PyTorch (Python)
|
2021-05-05 15:00:20 UTC
|
https://youtu.be/jVPd7lEvjtg
|
jVPd7lEvjtg
|
UCv83tO5cePwHMt1952IVVHw
|
jVPd7lEvjtg-t525.0
|
Okay, so we do type.
| 525 | 534 |
Sentence Similarity With Transformers and PyTorch (Python)
|
2021-05-05 15:00:20 UTC
|
https://youtu.be/jVPd7lEvjtg
|
jVPd7lEvjtg
|
UCv83tO5cePwHMt1952IVVHw
|
jVPd7lEvjtg-t532.0
|
And now we just have a tensor.
| 532 | 540 |
Sentence Similarity With Transformers and PyTorch (Python)
|
2021-05-05 15:00:20 UTC
|
https://youtu.be/jVPd7lEvjtg
|
jVPd7lEvjtg
|
UCv83tO5cePwHMt1952IVVHw
|
jVPd7lEvjtg-t534.0
|
Now, that's great.
| 534 | 542 |
Sentence Similarity With Transformers and PyTorch (Python)
|
2021-05-05 15:00:20 UTC
|
https://youtu.be/jVPd7lEvjtg
|
jVPd7lEvjtg
|
UCv83tO5cePwHMt1952IVVHw
|
jVPd7lEvjtg-t540.0
|
Check its size.
| 540 | 549 |
Sentence Similarity With Transformers and PyTorch (Python)
|
2021-05-05 15:00:20 UTC
|
https://youtu.be/jVPd7lEvjtg
|
jVPd7lEvjtg
|
UCv83tO5cePwHMt1952IVVHw
|
jVPd7lEvjtg-t542.0
|
So we have six sentences that have all been encoded into the 128 tokens.
| 542 | 552 |
Sentence Similarity With Transformers and PyTorch (Python)
|
2021-05-05 15:00:20 UTC
|
https://youtu.be/jVPd7lEvjtg
|
jVPd7lEvjtg
|
UCv83tO5cePwHMt1952IVVHw
|
jVPd7lEvjtg-t549.0
|
Ready to go into our model.
| 549 | 556 |
Sentence Similarity With Transformers and PyTorch (Python)
|
2021-05-05 15:00:20 UTC
|
https://youtu.be/jVPd7lEvjtg
|
jVPd7lEvjtg
|
UCv83tO5cePwHMt1952IVVHw
|
jVPd7lEvjtg-t552.0
|
So to process these through our model.
| 552 | 562 |
Sentence Similarity With Transformers and PyTorch (Python)
|
2021-05-05 15:00:20 UTC
|
https://youtu.be/jVPd7lEvjtg
|
jVPd7lEvjtg
|
UCv83tO5cePwHMt1952IVVHw
|
jVPd7lEvjtg-t556.0
|
We'll output the outputs to this outputs variable.
| 556 | 564 |
Sentence Similarity With Transformers and PyTorch (Python)
|
2021-05-05 15:00:20 UTC
|
https://youtu.be/jVPd7lEvjtg
|
jVPd7lEvjtg
|
UCv83tO5cePwHMt1952IVVHw
|
jVPd7lEvjtg-t562.0
|
And we take our model.
| 562 | 573 |
Sentence Similarity With Transformers and PyTorch (Python)
|
2021-05-05 15:00:20 UTC
|
https://youtu.be/jVPd7lEvjtg
|
jVPd7lEvjtg
|
UCv83tO5cePwHMt1952IVVHw
|
jVPd7lEvjtg-t564.0
|
And we pass our tokens as keyword arguments into the model input there.
| 564 | 577 |
Sentence Similarity With Transformers and PyTorch (Python)
|
2021-05-05 15:00:20 UTC
|
https://youtu.be/jVPd7lEvjtg
|
jVPd7lEvjtg
|
UCv83tO5cePwHMt1952IVVHw
|
jVPd7lEvjtg-t573.0
|
So we process that.
| 573 | 582 |
Sentence Similarity With Transformers and PyTorch (Python)
|
2021-05-05 15:00:20 UTC
|
https://youtu.be/jVPd7lEvjtg
|
jVPd7lEvjtg
|
UCv83tO5cePwHMt1952IVVHw
|
jVPd7lEvjtg-t577.0
|
And that will give us this output object.
| 577 | 589 |
Sentence Similarity With Transformers and PyTorch (Python)
|
2021-05-05 15:00:20 UTC
|
https://youtu.be/jVPd7lEvjtg
|
jVPd7lEvjtg
|
UCv83tO5cePwHMt1952IVVHw
|
jVPd7lEvjtg-t582.0
|
And inside this output object, we have the last hidden state tensor here.
| 582 | 593 |
Sentence Similarity With Transformers and PyTorch (Python)
|
2021-05-05 15:00:20 UTC
|
https://youtu.be/jVPd7lEvjtg
|
jVPd7lEvjtg
|
UCv83tO5cePwHMt1952IVVHw
|
jVPd7lEvjtg-t589.0
|
And we can also see that if we print out keys.
| 589 | 595 |
Sentence Similarity With Transformers and PyTorch (Python)
|
2021-05-05 15:00:20 UTC
|
https://youtu.be/jVPd7lEvjtg
|
jVPd7lEvjtg
|
UCv83tO5cePwHMt1952IVVHw
|
jVPd7lEvjtg-t593.0
|
You see that we have the last hidden state.
| 593 | 597 |
Sentence Similarity With Transformers and PyTorch (Python)
|
2021-05-05 15:00:20 UTC
|
https://youtu.be/jVPd7lEvjtg
|
jVPd7lEvjtg
|
UCv83tO5cePwHMt1952IVVHw
|
jVPd7lEvjtg-t595.0
|
And we also have this pooler output.
| 595 | 604 |
Sentence Similarity With Transformers and PyTorch (Python)
|
2021-05-05 15:00:20 UTC
|
https://youtu.be/jVPd7lEvjtg
|
jVPd7lEvjtg
|
UCv83tO5cePwHMt1952IVVHw
|
jVPd7lEvjtg-t597.0
|
Now, we want to take our last hidden state tensor.
| 597 | 612 |
Sentence Similarity With Transformers and PyTorch (Python)
|
2021-05-05 15:00:20 UTC
|
https://youtu.be/jVPd7lEvjtg
|
jVPd7lEvjtg
|
UCv83tO5cePwHMt1952IVVHw
|
jVPd7lEvjtg-t604.0
|
And then perform the mean pooling operation to convert it into a sentence vector.
| 604 | 621 |
Sentence Similarity With Transformers and PyTorch (Python)
|
2021-05-05 15:00:20 UTC
|
https://youtu.be/jVPd7lEvjtg
|
jVPd7lEvjtg
|
UCv83tO5cePwHMt1952IVVHw
|
jVPd7lEvjtg-t612.0
|
So to get that last hidden state, we will assign it to this embeddings variable.
| 612 | 630 |
Sentence Similarity With Transformers and PyTorch (Python)
|
2021-05-05 15:00:20 UTC
|
https://youtu.be/jVPd7lEvjtg
|
jVPd7lEvjtg
|
UCv83tO5cePwHMt1952IVVHw
|
jVPd7lEvjtg-t621.0
|
And we extract it using hidden or last hidden state.
| 621 | 631 |
Sentence Similarity With Transformers and PyTorch (Python)
|
2021-05-05 15:00:20 UTC
|
https://youtu.be/jVPd7lEvjtg
|
jVPd7lEvjtg
|
UCv83tO5cePwHMt1952IVVHw
|
jVPd7lEvjtg-t630.0
|
Like that.
| 630 | 634 |
Sentence Similarity With Transformers and PyTorch (Python)
|
2021-05-05 15:00:20 UTC
|
https://youtu.be/jVPd7lEvjtg
|
jVPd7lEvjtg
|
UCv83tO5cePwHMt1952IVVHw
|
jVPd7lEvjtg-t631.0
|
And let's just check what we have here.
| 631 | 635 |
Sentence Similarity With Transformers and PyTorch (Python)
|
2021-05-05 15:00:20 UTC
|
https://youtu.be/jVPd7lEvjtg
|
jVPd7lEvjtg
|
UCv83tO5cePwHMt1952IVVHw
|
jVPd7lEvjtg-t634.0
|
So we'll just hold it at shape.
| 634 | 639 |
Sentence Similarity With Transformers and PyTorch (Python)
|
2021-05-05 15:00:20 UTC
|
https://youtu.be/jVPd7lEvjtg
|
jVPd7lEvjtg
|
UCv83tO5cePwHMt1952IVVHw
|
jVPd7lEvjtg-t635.0
|
And you see now we have the six sentences.
| 635 | 642 |
Sentence Similarity With Transformers and PyTorch (Python)
|
2021-05-05 15:00:20 UTC
|
https://youtu.be/jVPd7lEvjtg
|
jVPd7lEvjtg
|
UCv83tO5cePwHMt1952IVVHw
|
jVPd7lEvjtg-t639.0
|
We have the 128 tokens.
| 639 | 646 |
Sentence Similarity With Transformers and PyTorch (Python)
|
2021-05-05 15:00:20 UTC
|
https://youtu.be/jVPd7lEvjtg
|
jVPd7lEvjtg
|
UCv83tO5cePwHMt1952IVVHw
|
jVPd7lEvjtg-t642.0
|
And then we have the 768 dimension size.
| 642 | 651 |
Sentence Similarity With Transformers and PyTorch (Python)
|
2021-05-05 15:00:20 UTC
|
https://youtu.be/jVPd7lEvjtg
|
jVPd7lEvjtg
|
UCv83tO5cePwHMt1952IVVHw
|
jVPd7lEvjtg-t646.0
|
Which is just the hidden state dimensions within BERT.
| 646 | 658 |
Sentence Similarity With Transformers and PyTorch (Python)
|
2021-05-05 15:00:20 UTC
|
https://youtu.be/jVPd7lEvjtg
|
jVPd7lEvjtg
|
UCv83tO5cePwHMt1952IVVHw
|
jVPd7lEvjtg-t651.0
|
So what we have at the moment is this last hidden state tensor.
| 651 | 665 |
Sentence Similarity With Transformers and PyTorch (Python)
|
2021-05-05 15:00:20 UTC
|
https://youtu.be/jVPd7lEvjtg
|
jVPd7lEvjtg
|
UCv83tO5cePwHMt1952IVVHw
|
jVPd7lEvjtg-t658.0
|
And what we're going to do is now convert it into this using a mean pooling operation.
| 658 | 679 |
Sentence Similarity With Transformers and PyTorch (Python)
|
2021-05-05 15:00:20 UTC
|
https://youtu.be/jVPd7lEvjtg
|
jVPd7lEvjtg
|
UCv83tO5cePwHMt1952IVVHw
|
jVPd7lEvjtg-t665.0
|
So the first thing we need to do is multiply every value within this last hidden state tensor by zero.
| 665 | 683 |
Sentence Similarity With Transformers and PyTorch (Python)
|
2021-05-05 15:00:20 UTC
|
https://youtu.be/jVPd7lEvjtg
|
jVPd7lEvjtg
|
UCv83tO5cePwHMt1952IVVHw
|
jVPd7lEvjtg-t679.0
|
Where we shouldn't have a real token.
| 679 | 686 |
Sentence Similarity With Transformers and PyTorch (Python)
|
2021-05-05 15:00:20 UTC
|
https://youtu.be/jVPd7lEvjtg
|
jVPd7lEvjtg
|
UCv83tO5cePwHMt1952IVVHw
|
jVPd7lEvjtg-t683.0
|
So if we look up here, we've padded all of these.
| 683 | 693 |
Sentence Similarity With Transformers and PyTorch (Python)
|
2021-05-05 15:00:20 UTC
|
https://youtu.be/jVPd7lEvjtg
|
jVPd7lEvjtg
|
UCv83tO5cePwHMt1952IVVHw
|
jVPd7lEvjtg-t686.0
|
And obviously there's more padding tokens in this sentence than there are in this sentence.
| 686 | 698 |
Sentence Similarity With Transformers and PyTorch (Python)
|
2021-05-05 15:00:20 UTC
|
https://youtu.be/jVPd7lEvjtg
|
jVPd7lEvjtg
|
UCv83tO5cePwHMt1952IVVHw
|
jVPd7lEvjtg-t693.0
|
So we need to take each of those attention mass tensors that we took here.
| 693 | 700 |
Sentence Similarity With Transformers and PyTorch (Python)
|
2021-05-05 15:00:20 UTC
|
https://youtu.be/jVPd7lEvjtg
|
jVPd7lEvjtg
|
UCv83tO5cePwHMt1952IVVHw
|
jVPd7lEvjtg-t698.0
|
Which just contain ones and zeros.
| 698 | 702 |
Sentence Similarity With Transformers and PyTorch (Python)
|
2021-05-05 15:00:20 UTC
|
https://youtu.be/jVPd7lEvjtg
|
jVPd7lEvjtg
|
UCv83tO5cePwHMt1952IVVHw
|
jVPd7lEvjtg-t700.0
|
Ones where there's real tokens.
| 700 | 704 |
Sentence Similarity With Transformers and PyTorch (Python)
|
2021-05-05 15:00:20 UTC
|
https://youtu.be/jVPd7lEvjtg
|
jVPd7lEvjtg
|
UCv83tO5cePwHMt1952IVVHw
|
jVPd7lEvjtg-t702.0
|
Zeros where there are padding tokens.
| 702 | 711 |
Sentence Similarity With Transformers and PyTorch (Python)
|
2021-05-05 15:00:20 UTC
|
https://youtu.be/jVPd7lEvjtg
|
jVPd7lEvjtg
|
UCv83tO5cePwHMt1952IVVHw
|
jVPd7lEvjtg-t704.0
|
And multiply that out to remove any activations where there should just be padding tokens.
| 704 | 712 |
Sentence Similarity With Transformers and PyTorch (Python)
|
2021-05-05 15:00:20 UTC
|
https://youtu.be/jVPd7lEvjtg
|
jVPd7lEvjtg
|
UCv83tO5cePwHMt1952IVVHw
|
jVPd7lEvjtg-t711.0
|
E.g. zeros.
| 711 | 719 |
Sentence Similarity With Transformers and PyTorch (Python)
|
2021-05-05 15:00:20 UTC
|
https://youtu.be/jVPd7lEvjtg
|
jVPd7lEvjtg
|
UCv83tO5cePwHMt1952IVVHw
|
jVPd7lEvjtg-t712.0
|
Now the only problem is that if we have a look at our attention mass.
| 712 | 726 |
Sentence Similarity With Transformers and PyTorch (Python)
|
2021-05-05 15:00:20 UTC
|
https://youtu.be/jVPd7lEvjtg
|
jVPd7lEvjtg
|
UCv83tO5cePwHMt1952IVVHw
|
jVPd7lEvjtg-t719.0
|
So tokens attention mass.
| 719 | 730 |
Sentence Similarity With Transformers and PyTorch (Python)
|
2021-05-05 15:00:20 UTC
|
https://youtu.be/jVPd7lEvjtg
|
jVPd7lEvjtg
|
UCv83tO5cePwHMt1952IVVHw
|
jVPd7lEvjtg-t726.0
|
If we have a look at the size.
| 726 | 732 |
Sentence Similarity With Transformers and PyTorch (Python)
|
2021-05-05 15:00:20 UTC
|
https://youtu.be/jVPd7lEvjtg
|
jVPd7lEvjtg
|
UCv83tO5cePwHMt1952IVVHw
|
jVPd7lEvjtg-t730.0
|
We get a six by 128.
| 730 | 738 |
Subsets and Splits
No community queries yet
The top public SQL queries from the community will appear here once available.