title
stringlengths 12
112
| published
stringlengths 19
23
| url
stringlengths 28
28
| video_id
stringlengths 11
11
| channel_id
stringclasses 5
values | id
stringlengths 16
31
| text
stringlengths 0
596
| start
float64 0
37.8k
| end
float64 2.18
37.8k
|
|---|---|---|---|---|---|---|---|---|
Training and Testing an Italian BERT - Transformers From Scratch #4
|
2021-07-06 13:00:03 UTC
|
https://youtu.be/35Pdoyi6ZoQ
|
35Pdoyi6ZoQ
|
UCv83tO5cePwHMt1952IVVHw
|
35Pdoyi6ZoQ-t1747.96
|
Avesero.
| 1,747.96 | 1,751.56 |
Training and Testing an Italian BERT - Transformers From Scratch #4
|
2021-07-06 13:00:03 UTC
|
https://youtu.be/35Pdoyi6ZoQ
|
35Pdoyi6ZoQ
|
UCv83tO5cePwHMt1952IVVHw
|
35Pdoyi6ZoQ-t1748.76
|
So let's run it.
| 1,748.76 | 1,752.36 |
Training and Testing an Italian BERT - Transformers From Scratch #4
|
2021-07-06 13:00:03 UTC
|
https://youtu.be/35Pdoyi6ZoQ
|
35Pdoyi6ZoQ
|
UCv83tO5cePwHMt1952IVVHw
|
35Pdoyi6ZoQ-t1751.56
|
Avesero.
| 1,751.56 | 1,753.56 |
Training and Testing an Italian BERT - Transformers From Scratch #4
|
2021-07-06 13:00:03 UTC
|
https://youtu.be/35Pdoyi6ZoQ
|
35Pdoyi6ZoQ
|
UCv83tO5cePwHMt1952IVVHw
|
35Pdoyi6ZoQ-t1752.36
|
You see?
| 1,752.36 | 1,754.36 |
Training and Testing an Italian BERT - Transformers From Scratch #4
|
2021-07-06 13:00:03 UTC
|
https://youtu.be/35Pdoyi6ZoQ
|
35Pdoyi6ZoQ
|
UCv83tO5cePwHMt1952IVVHw
|
35Pdoyi6ZoQ-t1753.56
|
That's cool.
| 1,753.56 | 1,755.96 |
Training and Testing an Italian BERT - Transformers From Scratch #4
|
2021-07-06 13:00:03 UTC
|
https://youtu.be/35Pdoyi6ZoQ
|
35Pdoyi6ZoQ
|
UCv83tO5cePwHMt1952IVVHw
|
35Pdoyi6ZoQ-t1754.36
|
That's very good.
| 1,754.36 | 1,759.56 |
Training and Testing an Italian BERT - Transformers From Scratch #4
|
2021-07-06 13:00:03 UTC
|
https://youtu.be/35Pdoyi6ZoQ
|
35Pdoyi6ZoQ
|
UCv83tO5cePwHMt1952IVVHw
|
35Pdoyi6ZoQ-t1755.96
|
And then the other one, se loro anno, is right?
| 1,755.96 | 1,760.16 |
Training and Testing an Italian BERT - Transformers From Scratch #4
|
2021-07-06 13:00:03 UTC
|
https://youtu.be/35Pdoyi6ZoQ
|
35Pdoyi6ZoQ
|
UCv83tO5cePwHMt1952IVVHw
|
35Pdoyi6ZoQ-t1759.56
|
Ave anno?
| 1,759.56 | 1,761.76 |
Training and Testing an Italian BERT - Transformers From Scratch #4
|
2021-07-06 13:00:03 UTC
|
https://youtu.be/35Pdoyi6ZoQ
|
35Pdoyi6ZoQ
|
UCv83tO5cePwHMt1952IVVHw
|
35Pdoyi6ZoQ-t1760.16
|
I mean, I'm saying...
| 1,760.16 | 1,764.56 |
Training and Testing an Italian BERT - Transformers From Scratch #4
|
2021-07-06 13:00:03 UTC
|
https://youtu.be/35Pdoyi6ZoQ
|
35Pdoyi6ZoQ
|
UCv83tO5cePwHMt1952IVVHw
|
35Pdoyi6ZoQ-t1761.76
|
Well, the verb is incorrect, but...
| 1,761.76 | 1,768.16 |
Training and Testing an Italian BERT - Transformers From Scratch #4
|
2021-07-06 13:00:03 UTC
|
https://youtu.be/35Pdoyi6ZoQ
|
35Pdoyi6ZoQ
|
UCv83tO5cePwHMt1952IVVHw
|
35Pdoyi6ZoQ-t1764.56
|
Yeah, it's in the wrong place, but it's saying the right...
| 1,764.56 | 1,772.76 |
Training and Testing an Italian BERT - Transformers From Scratch #4
|
2021-07-06 13:00:03 UTC
|
https://youtu.be/35Pdoyi6ZoQ
|
35Pdoyi6ZoQ
|
UCv83tO5cePwHMt1952IVVHw
|
35Pdoyi6ZoQ-t1768.16
|
Like the meaning is correct, but the grammar, it's not correct.
| 1,768.16 | 1,773.36 |
Training and Testing an Italian BERT - Transformers From Scratch #4
|
2021-07-06 13:00:03 UTC
|
https://youtu.be/35Pdoyi6ZoQ
|
35Pdoyi6ZoQ
|
UCv83tO5cePwHMt1952IVVHw
|
35Pdoyi6ZoQ-t1772.76
|
Okay.
| 1,772.76 | 1,773.96 |
Training and Testing an Italian BERT - Transformers From Scratch #4
|
2021-07-06 13:00:03 UTC
|
https://youtu.be/35Pdoyi6ZoQ
|
35Pdoyi6ZoQ
|
UCv83tO5cePwHMt1952IVVHw
|
35Pdoyi6ZoQ-t1773.3600000000001
|
Right, okay.
| 1,773.36 | 1,776.36 |
Training and Testing an Italian BERT - Transformers From Scratch #4
|
2021-07-06 13:00:03 UTC
|
https://youtu.be/35Pdoyi6ZoQ
|
35Pdoyi6ZoQ
|
UCv83tO5cePwHMt1952IVVHw
|
35Pdoyi6ZoQ-t1773.96
|
Yeah.
| 1,773.96 | 1,777.76 |
Training and Testing an Italian BERT - Transformers From Scratch #4
|
2021-07-06 13:00:03 UTC
|
https://youtu.be/35Pdoyi6ZoQ
|
35Pdoyi6ZoQ
|
UCv83tO5cePwHMt1952IVVHw
|
35Pdoyi6ZoQ-t1776.3600000000001
|
It's cool.
| 1,776.36 | 1,778.36 |
Training and Testing an Italian BERT - Transformers From Scratch #4
|
2021-07-06 13:00:03 UTC
|
https://youtu.be/35Pdoyi6ZoQ
|
35Pdoyi6ZoQ
|
UCv83tO5cePwHMt1952IVVHw
|
35Pdoyi6ZoQ-t1777.76
|
You're welcome.
| 1,777.76 | 1,782.56 |
Training and Testing an Italian BERT - Transformers From Scratch #4
|
2021-07-06 13:00:03 UTC
|
https://youtu.be/35Pdoyi6ZoQ
|
35Pdoyi6ZoQ
|
UCv83tO5cePwHMt1952IVVHw
|
35Pdoyi6ZoQ-t1778.3600000000001
|
Not happy it actually worked, because I wasn't sure if I could just...
| 1,778.36 | 1,789.56 |
Training and Testing an Italian BERT - Transformers From Scratch #4
|
2021-07-06 13:00:03 UTC
|
https://youtu.be/35Pdoyi6ZoQ
|
35Pdoyi6ZoQ
|
UCv83tO5cePwHMt1952IVVHw
|
35Pdoyi6ZoQ-t1782.56
|
Well, it worked with ciao coming back, but that was all I tested it with, so I was a little bit worried that it might not do anything else.
| 1,782.56 | 1,790.56 |
Training and Testing an Italian BERT - Transformers From Scratch #4
|
2021-07-06 13:00:03 UTC
|
https://youtu.be/35Pdoyi6ZoQ
|
35Pdoyi6ZoQ
|
UCv83tO5cePwHMt1952IVVHw
|
35Pdoyi6ZoQ-t1789.56
|
But thank you.
| 1,789.56 | 1,792.76 |
Training and Testing an Italian BERT - Transformers From Scratch #4
|
2021-07-06 13:00:03 UTC
|
https://youtu.be/35Pdoyi6ZoQ
|
35Pdoyi6ZoQ
|
UCv83tO5cePwHMt1952IVVHw
|
35Pdoyi6ZoQ-t1790.56
|
You're welcome.
| 1,790.56 | 1,795.36 |
Training and Testing an Italian BERT - Transformers From Scratch #4
|
2021-07-06 13:00:03 UTC
|
https://youtu.be/35Pdoyi6ZoQ
|
35Pdoyi6ZoQ
|
UCv83tO5cePwHMt1952IVVHw
|
35Pdoyi6ZoQ-t1792.76
|
Bye.
| 1,792.76 | 1,800.56 |
Training and Testing an Italian BERT - Transformers From Scratch #4
|
2021-07-06 13:00:03 UTC
|
https://youtu.be/35Pdoyi6ZoQ
|
35Pdoyi6ZoQ
|
UCv83tO5cePwHMt1952IVVHw
|
35Pdoyi6ZoQ-t1795.36
|
Okay, so I think that's a pretty good result.
| 1,795.36 | 1,808.16 |
Training and Testing an Italian BERT - Transformers From Scratch #4
|
2021-07-06 13:00:03 UTC
|
https://youtu.be/35Pdoyi6ZoQ
|
35Pdoyi6ZoQ
|
UCv83tO5cePwHMt1952IVVHw
|
35Pdoyi6ZoQ-t1800.56
|
So, I mean, that's pretty much everything we needed for building our model, our transform model.
| 1,800.56 | 1,818.96 |
Training and Testing an Italian BERT - Transformers From Scratch #4
|
2021-07-06 13:00:03 UTC
|
https://youtu.be/35Pdoyi6ZoQ
|
35Pdoyi6ZoQ
|
UCv83tO5cePwHMt1952IVVHw
|
35Pdoyi6ZoQ-t1808.16
|
Although I do want to... so we're going to do one more video after this, where we're going to upload our model to the Hugging Face model hub.
| 1,808.16 | 1,828.56 |
Training and Testing an Italian BERT - Transformers From Scratch #4
|
2021-07-06 13:00:03 UTC
|
https://youtu.be/35Pdoyi6ZoQ
|
35Pdoyi6ZoQ
|
UCv83tO5cePwHMt1952IVVHw
|
35Pdoyi6ZoQ-t1818.96
|
And then what we'll be able to do is actually download it directly from Hugging Face, which I think will be super cool to do that and figure out how we actually put all that together.
| 1,818.96 | 1,831.56 |
Training and Testing an Italian BERT - Transformers From Scratch #4
|
2021-07-06 13:00:03 UTC
|
https://youtu.be/35Pdoyi6ZoQ
|
35Pdoyi6ZoQ
|
UCv83tO5cePwHMt1952IVVHw
|
35Pdoyi6ZoQ-t1828.5600000000002
|
So, yeah, I think good result.
| 1,828.56 | 1,833.36 |
Training and Testing an Italian BERT - Transformers From Scratch #4
|
2021-07-06 13:00:03 UTC
|
https://youtu.be/35Pdoyi6ZoQ
|
35Pdoyi6ZoQ
|
UCv83tO5cePwHMt1952IVVHw
|
35Pdoyi6ZoQ-t1831.5600000000002
|
I'm pretty happy with that.
| 1,831.56 | 1,835.96 |
Training and Testing an Italian BERT - Transformers From Scratch #4
|
2021-07-06 13:00:03 UTC
|
https://youtu.be/35Pdoyi6ZoQ
|
35Pdoyi6ZoQ
|
UCv83tO5cePwHMt1952IVVHw
|
35Pdoyi6ZoQ-t1833.3600000000001
|
And thank you for watching.
| 1,833.36 | 1,839.76 |
Training and Testing an Italian BERT - Transformers From Scratch #4
|
2021-07-06 13:00:03 UTC
|
https://youtu.be/35Pdoyi6ZoQ
|
35Pdoyi6ZoQ
|
UCv83tO5cePwHMt1952IVVHw
|
35Pdoyi6ZoQ-t1835.96
| 1,835.96 | 1,839.76 |
|
Choosing Indexes for Similarity Search (Faiss in Python)
|
2021-08-09 15:04:10 UTC
|
https://youtu.be/B7wmo_NImgM
|
B7wmo_NImgM
|
UCv83tO5cePwHMt1952IVVHw
|
B7wmo_NImgM-t0.0
|
Hi, welcome to the video.
| 0 | 6.32 |
Choosing Indexes for Similarity Search (Faiss in Python)
|
2021-08-09 15:04:10 UTC
|
https://youtu.be/B7wmo_NImgM
|
B7wmo_NImgM
|
UCv83tO5cePwHMt1952IVVHw
|
B7wmo_NImgM-t1.8
|
I'm going to take you through a few different indexes in FIAS today.
| 1.8 | 8.32 |
Choosing Indexes for Similarity Search (Faiss in Python)
|
2021-08-09 15:04:10 UTC
|
https://youtu.be/B7wmo_NImgM
|
B7wmo_NImgM
|
UCv83tO5cePwHMt1952IVVHw
|
B7wmo_NImgM-t6.32
|
So FIAS for similarity search.
| 6.32 | 14.36 |
Choosing Indexes for Similarity Search (Faiss in Python)
|
2021-08-09 15:04:10 UTC
|
https://youtu.be/B7wmo_NImgM
|
B7wmo_NImgM
|
UCv83tO5cePwHMt1952IVVHw
|
B7wmo_NImgM-t8.32
|
And we're going to learn how we can decide which index to use based on our data.
| 8.32 | 18.56 |
Choosing Indexes for Similarity Search (Faiss in Python)
|
2021-08-09 15:04:10 UTC
|
https://youtu.be/B7wmo_NImgM
|
B7wmo_NImgM
|
UCv83tO5cePwHMt1952IVVHw
|
B7wmo_NImgM-t14.36
|
Now, these indexes are reasonably complex,
| 14.36 | 23.56 |
Choosing Indexes for Similarity Search (Faiss in Python)
|
2021-08-09 15:04:10 UTC
|
https://youtu.be/B7wmo_NImgM
|
B7wmo_NImgM
|
UCv83tO5cePwHMt1952IVVHw
|
B7wmo_NImgM-t18.56
|
but we're going to just have a high level look at each one of them.
| 18.56 | 26.96 |
Choosing Indexes for Similarity Search (Faiss in Python)
|
2021-08-09 15:04:10 UTC
|
https://youtu.be/B7wmo_NImgM
|
B7wmo_NImgM
|
UCv83tO5cePwHMt1952IVVHw
|
B7wmo_NImgM-t23.56
|
At some point in the future, we'll go into more depth for sure.
| 23.56 | 29.16 |
Choosing Indexes for Similarity Search (Faiss in Python)
|
2021-08-09 15:04:10 UTC
|
https://youtu.be/B7wmo_NImgM
|
B7wmo_NImgM
|
UCv83tO5cePwHMt1952IVVHw
|
B7wmo_NImgM-t26.96
|
But for now, this is what we're going to do.
| 26.96 | 31.68 |
Choosing Indexes for Similarity Search (Faiss in Python)
|
2021-08-09 15:04:10 UTC
|
https://youtu.be/B7wmo_NImgM
|
B7wmo_NImgM
|
UCv83tO5cePwHMt1952IVVHw
|
B7wmo_NImgM-t29.16
|
So we're going to cover the indexes that you see on the screen at the moment.
| 29.16 | 35.72 |
Choosing Indexes for Similarity Search (Faiss in Python)
|
2021-08-09 15:04:10 UTC
|
https://youtu.be/B7wmo_NImgM
|
B7wmo_NImgM
|
UCv83tO5cePwHMt1952IVVHw
|
B7wmo_NImgM-t31.68
|
So we have the flat indexes, which are just the plain and simple,
| 31.68 | 37.52 |
Choosing Indexes for Similarity Search (Faiss in Python)
|
2021-08-09 15:04:10 UTC
|
https://youtu.be/B7wmo_NImgM
|
B7wmo_NImgM
|
UCv83tO5cePwHMt1952IVVHw
|
B7wmo_NImgM-t35.72
|
nothing special going on there.
| 35.72 | 41.8 |
Choosing Indexes for Similarity Search (Faiss in Python)
|
2021-08-09 15:04:10 UTC
|
https://youtu.be/B7wmo_NImgM
|
B7wmo_NImgM
|
UCv83tO5cePwHMt1952IVVHw
|
B7wmo_NImgM-t37.519999999999996
|
And then we're going to have a look at LSH or locality sensitive hashing,
| 37.52 | 47.64 |
Choosing Indexes for Similarity Search (Faiss in Python)
|
2021-08-09 15:04:10 UTC
|
https://youtu.be/B7wmo_NImgM
|
B7wmo_NImgM
|
UCv83tO5cePwHMt1952IVVHw
|
B7wmo_NImgM-t41.8
|
HNSW, which is hierarchical navigable small worlds.
| 41.8 | 52.12 |
Choosing Indexes for Similarity Search (Faiss in Python)
|
2021-08-09 15:04:10 UTC
|
https://youtu.be/B7wmo_NImgM
|
B7wmo_NImgM
|
UCv83tO5cePwHMt1952IVVHw
|
B7wmo_NImgM-t47.64
|
And then finally, we're going to have a look at an IVF index as well.
| 47.64 | 57.92 |
Choosing Indexes for Similarity Search (Faiss in Python)
|
2021-08-09 15:04:10 UTC
|
https://youtu.be/B7wmo_NImgM
|
B7wmo_NImgM
|
UCv83tO5cePwHMt1952IVVHw
|
B7wmo_NImgM-t52.120000000000005
|
So first thing I'm going to show you is how to get some data for following through this.
| 52.12 | 60.56 |
Choosing Indexes for Similarity Search (Faiss in Python)
|
2021-08-09 15:04:10 UTC
|
https://youtu.be/B7wmo_NImgM
|
B7wmo_NImgM
|
UCv83tO5cePwHMt1952IVVHw
|
B7wmo_NImgM-t57.92
|
So we're going to be using the SIFT 1M dataset,
| 57.92 | 67.32 |
Choosing Indexes for Similarity Search (Faiss in Python)
|
2021-08-09 15:04:10 UTC
|
https://youtu.be/B7wmo_NImgM
|
B7wmo_NImgM
|
UCv83tO5cePwHMt1952IVVHw
|
B7wmo_NImgM-t60.56
|
which is one million vectors that we can use for testing similarity.
| 60.56 | 69.96 |
Choosing Indexes for Similarity Search (Faiss in Python)
|
2021-08-09 15:04:10 UTC
|
https://youtu.be/B7wmo_NImgM
|
B7wmo_NImgM
|
UCv83tO5cePwHMt1952IVVHw
|
B7wmo_NImgM-t67.32000000000001
|
Now, there's a little bit of code, so I'm just going to show it to you.
| 67.32 | 74.88 |
Choosing Indexes for Similarity Search (Faiss in Python)
|
2021-08-09 15:04:10 UTC
|
https://youtu.be/B7wmo_NImgM
|
B7wmo_NImgM
|
UCv83tO5cePwHMt1952IVVHw
|
B7wmo_NImgM-t69.96000000000001
|
So we have here, we're just downloading the code.
| 69.96 | 77.88 |
Choosing Indexes for Similarity Search (Faiss in Python)
|
2021-08-09 15:04:10 UTC
|
https://youtu.be/B7wmo_NImgM
|
B7wmo_NImgM
|
UCv83tO5cePwHMt1952IVVHw
|
B7wmo_NImgM-t74.88
|
There'll be a notebook for this in the description as well.
| 74.88 | 82.6 |
Choosing Indexes for Similarity Search (Faiss in Python)
|
2021-08-09 15:04:10 UTC
|
https://youtu.be/B7wmo_NImgM
|
B7wmo_NImgM
|
UCv83tO5cePwHMt1952IVVHw
|
B7wmo_NImgM-t77.88
|
So you can just use that and copy things across.
| 77.88 | 86.84 |
Choosing Indexes for Similarity Search (Faiss in Python)
|
2021-08-09 15:04:10 UTC
|
https://youtu.be/B7wmo_NImgM
|
B7wmo_NImgM
|
UCv83tO5cePwHMt1952IVVHw
|
B7wmo_NImgM-t82.6
|
But we're downloading it from here and this will give us a tar file.
| 82.6 | 88.88 |
Choosing Indexes for Similarity Search (Faiss in Python)
|
2021-08-09 15:04:10 UTC
|
https://youtu.be/B7wmo_NImgM
|
B7wmo_NImgM
|
UCv83tO5cePwHMt1952IVVHw
|
B7wmo_NImgM-t86.84
|
So we download that.
| 86.84 | 96.36 |
Choosing Indexes for Similarity Search (Faiss in Python)
|
2021-08-09 15:04:10 UTC
|
https://youtu.be/B7wmo_NImgM
|
B7wmo_NImgM
|
UCv83tO5cePwHMt1952IVVHw
|
B7wmo_NImgM-t88.88000000000001
|
And then here, all we're doing is extracting all the files from inside that tar file.
| 88.88 | 100.8 |
Choosing Indexes for Similarity Search (Faiss in Python)
|
2021-08-09 15:04:10 UTC
|
https://youtu.be/B7wmo_NImgM
|
B7wmo_NImgM
|
UCv83tO5cePwHMt1952IVVHw
|
B7wmo_NImgM-t96.36
|
And then here, I'm reading everything into the notebook.
| 96.36 | 105.76 |
Choosing Indexes for Similarity Search (Faiss in Python)
|
2021-08-09 15:04:10 UTC
|
https://youtu.be/B7wmo_NImgM
|
B7wmo_NImgM
|
UCv83tO5cePwHMt1952IVVHw
|
B7wmo_NImgM-t100.80000000000001
|
So inside that tar file, we'll get these FVEX files
| 100.8 | 109.76 |
Choosing Indexes for Similarity Search (Faiss in Python)
|
2021-08-09 15:04:10 UTC
|
https://youtu.be/B7wmo_NImgM
|
B7wmo_NImgM
|
UCv83tO5cePwHMt1952IVVHw
|
B7wmo_NImgM-t105.76
|
and we have to open them in a certain way, which is what we're doing here.
| 105.76 | 114.08 |
Choosing Indexes for Similarity Search (Faiss in Python)
|
2021-08-09 15:04:10 UTC
|
https://youtu.be/B7wmo_NImgM
|
B7wmo_NImgM
|
UCv83tO5cePwHMt1952IVVHw
|
B7wmo_NImgM-t109.76
|
So we're setting up the function to read them, sorry, here.
| 109.76 | 116.08 |
Choosing Indexes for Similarity Search (Faiss in Python)
|
2021-08-09 15:04:10 UTC
|
https://youtu.be/B7wmo_NImgM
|
B7wmo_NImgM
|
UCv83tO5cePwHMt1952IVVHw
|
B7wmo_NImgM-t114.08000000000001
|
And then here, I'm reading in two files.
| 114.08 | 118.2 |
Choosing Indexes for Similarity Search (Faiss in Python)
|
2021-08-09 15:04:10 UTC
|
https://youtu.be/B7wmo_NImgM
|
B7wmo_NImgM
|
UCv83tO5cePwHMt1952IVVHw
|
B7wmo_NImgM-t116.08
|
So we get a few different files here.
| 116.08 | 125.12 |
Choosing Indexes for Similarity Search (Faiss in Python)
|
2021-08-09 15:04:10 UTC
|
https://youtu.be/B7wmo_NImgM
|
B7wmo_NImgM
|
UCv83tO5cePwHMt1952IVVHw
|
B7wmo_NImgM-t118.2
|
So I'm sorry, this should be SIFT.
| 118.2 | 131.24 |
Choosing Indexes for Similarity Search (Faiss in Python)
|
2021-08-09 15:04:10 UTC
|
https://youtu.be/B7wmo_NImgM
|
B7wmo_NImgM
|
UCv83tO5cePwHMt1952IVVHw
|
B7wmo_NImgM-t125.12
|
So we get the base data, which is going to be the data that we're going to search through.
| 125.12 | 133.12 |
Choosing Indexes for Similarity Search (Faiss in Python)
|
2021-08-09 15:04:10 UTC
|
https://youtu.be/B7wmo_NImgM
|
B7wmo_NImgM
|
UCv83tO5cePwHMt1952IVVHw
|
B7wmo_NImgM-t131.24
|
And then we also have query data here.
| 131.24 | 136.6 |
Choosing Indexes for Similarity Search (Faiss in Python)
|
2021-08-09 15:04:10 UTC
|
https://youtu.be/B7wmo_NImgM
|
B7wmo_NImgM
|
UCv83tO5cePwHMt1952IVVHw
|
B7wmo_NImgM-t133.12
|
And then what I'm doing here is just selecting a single query,
| 133.12 | 139.56 |
Choosing Indexes for Similarity Search (Faiss in Python)
|
2021-08-09 15:04:10 UTC
|
https://youtu.be/B7wmo_NImgM
|
B7wmo_NImgM
|
UCv83tO5cePwHMt1952IVVHw
|
B7wmo_NImgM-t136.6
|
a single vector to query with rather than all of them,
| 136.6 | 141.6 |
Choosing Indexes for Similarity Search (Faiss in Python)
|
2021-08-09 15:04:10 UTC
|
https://youtu.be/B7wmo_NImgM
|
B7wmo_NImgM
|
UCv83tO5cePwHMt1952IVVHw
|
B7wmo_NImgM-t139.56
|
because we get quite a few in there.
| 139.56 | 143 |
Choosing Indexes for Similarity Search (Faiss in Python)
|
2021-08-09 15:04:10 UTC
|
https://youtu.be/B7wmo_NImgM
|
B7wmo_NImgM
|
UCv83tO5cePwHMt1952IVVHw
|
B7wmo_NImgM-t141.6
|
And then here, we can just see.
| 141.6 | 145.12 |
Choosing Indexes for Similarity Search (Faiss in Python)
|
2021-08-09 15:04:10 UTC
|
https://youtu.be/B7wmo_NImgM
|
B7wmo_NImgM
|
UCv83tO5cePwHMt1952IVVHw
|
B7wmo_NImgM-t143.0
|
So this is our query vector that gets Q.
| 143 | 150.24 |
Choosing Indexes for Similarity Search (Faiss in Python)
|
2021-08-09 15:04:10 UTC
|
https://youtu.be/B7wmo_NImgM
|
B7wmo_NImgM
|
UCv83tO5cePwHMt1952IVVHw
|
B7wmo_NImgM-t145.12
|
And then we also have WB here, which is going to be the data that we'll index and search through.
| 145.12 | 153.16 |
Choosing Indexes for Similarity Search (Faiss in Python)
|
2021-08-09 15:04:10 UTC
|
https://youtu.be/B7wmo_NImgM
|
B7wmo_NImgM
|
UCv83tO5cePwHMt1952IVVHw
|
B7wmo_NImgM-t150.24
|
And we can see some of it there as well.
| 150.24 | 156.08 |
Choosing Indexes for Similarity Search (Faiss in Python)
|
2021-08-09 15:04:10 UTC
|
https://youtu.be/B7wmo_NImgM
|
B7wmo_NImgM
|
UCv83tO5cePwHMt1952IVVHw
|
B7wmo_NImgM-t153.16
|
So that's how we get data.
| 153.16 | 160.2 |
Choosing Indexes for Similarity Search (Faiss in Python)
|
2021-08-09 15:04:10 UTC
|
https://youtu.be/B7wmo_NImgM
|
B7wmo_NImgM
|
UCv83tO5cePwHMt1952IVVHw
|
B7wmo_NImgM-t156.08
|
Let's move on to some flat indexes.
| 156.08 | 169.96 |
Choosing Indexes for Similarity Search (Faiss in Python)
|
2021-08-09 15:04:10 UTC
|
https://youtu.be/B7wmo_NImgM
|
B7wmo_NImgM
|
UCv83tO5cePwHMt1952IVVHw
|
B7wmo_NImgM-t160.20000000000002
|
So what you can see at the moment is a sort of a visual representation of a flat L2 index.
| 160.2 | 173.64 |
Choosing Indexes for Similarity Search (Faiss in Python)
|
2021-08-09 15:04:10 UTC
|
https://youtu.be/B7wmo_NImgM
|
B7wmo_NImgM
|
UCv83tO5cePwHMt1952IVVHw
|
B7wmo_NImgM-t169.96
|
Now up here, this is what we're doing.
| 169.96 | 176.44 |
Choosing Indexes for Similarity Search (Faiss in Python)
|
2021-08-09 15:04:10 UTC
|
https://youtu.be/B7wmo_NImgM
|
B7wmo_NImgM
|
UCv83tO5cePwHMt1952IVVHw
|
B7wmo_NImgM-t173.64
|
So we're calculating, we have all these points.
| 173.64 | 179.84 |
Choosing Indexes for Similarity Search (Faiss in Python)
|
2021-08-09 15:04:10 UTC
|
https://youtu.be/B7wmo_NImgM
|
B7wmo_NImgM
|
UCv83tO5cePwHMt1952IVVHw
|
B7wmo_NImgM-t176.44
|
So these are all of the WB points that we saw before.
| 176.44 | 181.8 |
Choosing Indexes for Similarity Search (Faiss in Python)
|
2021-08-09 15:04:10 UTC
|
https://youtu.be/B7wmo_NImgM
|
B7wmo_NImgM
|
UCv83tO5cePwHMt1952IVVHw
|
B7wmo_NImgM-t179.83999999999997
|
And this is our query vector.
| 179.84 | 184.64 |
Choosing Indexes for Similarity Search (Faiss in Python)
|
2021-08-09 15:04:10 UTC
|
https://youtu.be/B7wmo_NImgM
|
B7wmo_NImgM
|
UCv83tO5cePwHMt1952IVVHw
|
B7wmo_NImgM-t181.79999999999998
|
And we just calculate the distance between all of those.
| 181.8 | 188.44 |
Choosing Indexes for Similarity Search (Faiss in Python)
|
2021-08-09 15:04:10 UTC
|
https://youtu.be/B7wmo_NImgM
|
B7wmo_NImgM
|
UCv83tO5cePwHMt1952IVVHw
|
B7wmo_NImgM-t184.64
|
And then what we do is just take the top three.
| 184.64 | 192.44 |
Choosing Indexes for Similarity Search (Faiss in Python)
|
2021-08-09 15:04:10 UTC
|
https://youtu.be/B7wmo_NImgM
|
B7wmo_NImgM
|
UCv83tO5cePwHMt1952IVVHw
|
B7wmo_NImgM-t188.44
|
So the top K in reality, but in this case, it's top three.
| 188.44 | 194.76 |
Choosing Indexes for Similarity Search (Faiss in Python)
|
2021-08-09 15:04:10 UTC
|
https://youtu.be/B7wmo_NImgM
|
B7wmo_NImgM
|
UCv83tO5cePwHMt1952IVVHw
|
B7wmo_NImgM-t192.44
|
Now, we also have IP.
| 192.44 | 201.52 |
Choosing Indexes for Similarity Search (Faiss in Python)
|
2021-08-09 15:04:10 UTC
|
https://youtu.be/B7wmo_NImgM
|
B7wmo_NImgM
|
UCv83tO5cePwHMt1952IVVHw
|
B7wmo_NImgM-t194.76
|
So we have both L2 distance and IP distance as well.
| 194.76 | 204.92 |
Choosing Indexes for Similarity Search (Faiss in Python)
|
2021-08-09 15:04:10 UTC
|
https://youtu.be/B7wmo_NImgM
|
B7wmo_NImgM
|
UCv83tO5cePwHMt1952IVVHw
|
B7wmo_NImgM-t201.52
|
IP works in a different way.
| 201.52 | 211.52 |
Choosing Indexes for Similarity Search (Faiss in Python)
|
2021-08-09 15:04:10 UTC
|
https://youtu.be/B7wmo_NImgM
|
B7wmo_NImgM
|
UCv83tO5cePwHMt1952IVVHw
|
B7wmo_NImgM-t204.92000000000002
|
So we're using a different format to actually calculate the distance or similarity there.
| 204.92 | 215.32 |
Choosing Indexes for Similarity Search (Faiss in Python)
|
2021-08-09 15:04:10 UTC
|
https://youtu.be/B7wmo_NImgM
|
B7wmo_NImgM
|
UCv83tO5cePwHMt1952IVVHw
|
B7wmo_NImgM-t211.52
|
So it's not exactly as you see it here.
| 211.52 | 222.56 |
Choosing Indexes for Similarity Search (Faiss in Python)
|
2021-08-09 15:04:10 UTC
|
https://youtu.be/B7wmo_NImgM
|
B7wmo_NImgM
|
UCv83tO5cePwHMt1952IVVHw
|
B7wmo_NImgM-t215.32000000000002
|
But before we write any code, just want to say with flat indexes, they are 100% quality.
| 215.32 | 230 |
Choosing Indexes for Similarity Search (Faiss in Python)
|
2021-08-09 15:04:10 UTC
|
https://youtu.be/B7wmo_NImgM
|
B7wmo_NImgM
|
UCv83tO5cePwHMt1952IVVHw
|
B7wmo_NImgM-t222.56
|
And typically what we want to do with FI's and similarity search indexes is balance the
| 222.56 | 232.72 |
Choosing Indexes for Similarity Search (Faiss in Python)
|
2021-08-09 15:04:10 UTC
|
https://youtu.be/B7wmo_NImgM
|
B7wmo_NImgM
|
UCv83tO5cePwHMt1952IVVHw
|
B7wmo_NImgM-t230.0
|
search quality versus the search speed.
| 230 | 236.42 |
Choosing Indexes for Similarity Search (Faiss in Python)
|
2021-08-09 15:04:10 UTC
|
https://youtu.be/B7wmo_NImgM
|
B7wmo_NImgM
|
UCv83tO5cePwHMt1952IVVHw
|
B7wmo_NImgM-t232.72
|
Higher search quality, usually slower search speed.
| 232.72 | 242.2 |
Choosing Indexes for Similarity Search (Faiss in Python)
|
2021-08-09 15:04:10 UTC
|
https://youtu.be/B7wmo_NImgM
|
B7wmo_NImgM
|
UCv83tO5cePwHMt1952IVVHw
|
B7wmo_NImgM-t236.42
|
And flat indexes are just pure search quality because they are an exhaustive search.
| 236.42 | 248.4 |
Choosing Indexes for Similarity Search (Faiss in Python)
|
2021-08-09 15:04:10 UTC
|
https://youtu.be/B7wmo_NImgM
|
B7wmo_NImgM
|
UCv83tO5cePwHMt1952IVVHw
|
B7wmo_NImgM-t242.2
|
They check the distance between your query vector and every other vector in the index,
| 242.2 | 253.8 |
Choosing Indexes for Similarity Search (Faiss in Python)
|
2021-08-09 15:04:10 UTC
|
https://youtu.be/B7wmo_NImgM
|
B7wmo_NImgM
|
UCv83tO5cePwHMt1952IVVHw
|
B7wmo_NImgM-t248.4
|
which is fine if you don't have a particularly big data set or you don't care about time.
| 248.4 | 259.2 |
Choosing Indexes for Similarity Search (Faiss in Python)
|
2021-08-09 15:04:10 UTC
|
https://youtu.be/B7wmo_NImgM
|
B7wmo_NImgM
|
UCv83tO5cePwHMt1952IVVHw
|
B7wmo_NImgM-t253.8
|
But if you do, then you probably don't want to use that because it can take an incredibly
| 253.8 | 260.2 |
Choosing Indexes for Similarity Search (Faiss in Python)
|
2021-08-09 15:04:10 UTC
|
https://youtu.be/B7wmo_NImgM
|
B7wmo_NImgM
|
UCv83tO5cePwHMt1952IVVHw
|
B7wmo_NImgM-t259.2
|
long time.
| 259.2 | 267.88 |
Choosing Indexes for Similarity Search (Faiss in Python)
|
2021-08-09 15:04:10 UTC
|
https://youtu.be/B7wmo_NImgM
|
B7wmo_NImgM
|
UCv83tO5cePwHMt1952IVVHw
|
B7wmo_NImgM-t260.2
|
If you have a billion vectors in your data set and you do 100 queries a minute, then
| 260.2 | 271.32 |
Choosing Indexes for Similarity Search (Faiss in Python)
|
2021-08-09 15:04:10 UTC
|
https://youtu.be/B7wmo_NImgM
|
B7wmo_NImgM
|
UCv83tO5cePwHMt1952IVVHw
|
B7wmo_NImgM-t267.88
|
as far as I know, it's impossible to run that.
| 267.88 | 274.4 |
Choosing Indexes for Similarity Search (Faiss in Python)
|
2021-08-09 15:04:10 UTC
|
https://youtu.be/B7wmo_NImgM
|
B7wmo_NImgM
|
UCv83tO5cePwHMt1952IVVHw
|
B7wmo_NImgM-t271.32
|
And if you were going to run that, you'd need some pretty insane hardware.
| 271.32 | 281.74 |
Choosing Indexes for Similarity Search (Faiss in Python)
|
2021-08-09 15:04:10 UTC
|
https://youtu.be/B7wmo_NImgM
|
B7wmo_NImgM
|
UCv83tO5cePwHMt1952IVVHw
|
B7wmo_NImgM-t274.4
|
So we can't use flat indexes and exhaustive search in most cases.
| 274.4 | 286.2 |
Choosing Indexes for Similarity Search (Faiss in Python)
|
2021-08-09 15:04:10 UTC
|
https://youtu.be/B7wmo_NImgM
|
B7wmo_NImgM
|
UCv83tO5cePwHMt1952IVVHw
|
B7wmo_NImgM-t281.74
|
But I will show you how to do it.
| 281.74 | 291.92 |
Choosing Indexes for Similarity Search (Faiss in Python)
|
2021-08-09 15:04:10 UTC
|
https://youtu.be/B7wmo_NImgM
|
B7wmo_NImgM
|
UCv83tO5cePwHMt1952IVVHw
|
B7wmo_NImgM-t286.2
|
So first, I'm just going to define dimensionality of our data, which is 128, which we can see
| 286.2 | 292.92 |
Subsets and Splits
No community queries yet
The top public SQL queries from the community will appear here once available.