ResNet - Live Coding in PyTorch - Wed Sep 15, 9pm IST

Community Events Paper Reading Group
1

In this session, we will live code ResNets from scratch using PyTorch.

Live stream: Live Coding Session on ResNet - YouTube

13 Likes
2

Hello Aman! Eagerly waiting for the session

2 Likes
3

Could we please make sure to carry on our conversation here?

5 Likes
4

can you bump up your font size a touch…

2 Likes
5

good morning :slight_smile: increase font please

1 Like
6

Is it fine now? I am at 125%

7

thanks. it is good now.

1 Like
8

Thank you for taking this session
Really needed it

9

thanks for this…i needed this session for resnet.

10

yes. 3am … I am dedicated.

5 Likes
11

Will you also take up coding from scratch and setting up the ecc2 instance
jeremy did discuss it but very briefly

1 Like
12

It would be really helpful if you could help us in setting up ec2 instance in a session.

2 Likes
13

I have tried to keep up with the PyTorch group, but the coding part takes time
How should one use the notebooks?
Should one try to replicate them?

1 Like
14

(with respect to pytorch and fast.ai)

15

sharing the repo referred to in the beginning

how to know when to have the fully connected layer, pooling layer and convolution layer?

3 Likes
16

Shouldn’t downsample be a conv2d instance?

17

Do we not need to add Padding to the Conv2d operations?
Also do we want to set self.downsample = 1 (Identify) if inchannels and out_channels are same?

18

I think you need need super.init() ??

19

Doesn’t conv1d introduce new parameters while resnet is supposed to be f(x) + x ? sorry for a stupid question

20

why is it needed ? never got this