I hereby claim:
- I am noorvir on github.
- I am noorvir (https://keybase.io/noorvir) on keybase.
- I have a public key ASBq_QwtJuQyvEFwjizF-Ty9lAm-6bycSjNG7ARRoxXGqAo
To claim this, I am signing this object:
Noorvir Aulakh noorvir
noorvir
/ scrollable-panel.tsx
Last active
February 18, 2024 09:03
Bare minimum hierarchy to enable scrollable component that stretches to fill all available space.
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
| <div style={{ display: 'flex', height: '100vh', flexGrow: 1, flexDirection: 'column' }}> | |
| <div | |
| style={{ | |
| display: 'flex', | |
| height: '100%', | |
| flexDirection: 'column', | |
| backgroundColor: 'red', | |
| }} | |
| > | |
| <div style={{ backgroundColor: 'yellow' }}>Element above</div> |
noorvir
/ keybase.md
Created
December 27, 2020 22:51
noorvir
/ map_and_assign_checkpoint.py
Last active
July 18, 2022 11:17
Load pre-trained weights in TensorFlow by mapping variable names in the new model to those in the saved checkpoint.
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
| """ | |
| You sometimes need to load weights from a checkpoint file to a new/different TensorFlow architecture where the | |
| variable names for the layers are different (see example below). For instance, you might need to load the first | |
| few layers of AlexNet pre-trained on ImageNet to your custom model which for some reason has a different naming | |
| convention. | |
| I couldn't find any convinient piece of code to do this; this gist is aimed at addressing this. | |
| Example mapping: |
noorvir
/ neural_net_training_blueprint.py
Created
June 12, 2018 10:41
A blueprint class to bunch together most of the boilerplate required to train neural-networks with TensorFlow. It is meant to be general enough to work with any neural net architecture/training regime.
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
| """ | |
| Inspired by: | |
| https://blog.metaflow.fr/tensorflow-a-proposal-of-good-practices-for-files-folders-and-models-architecture-f23171501ae3 | |
| Recommended guidelines for child classes: | |
| 1) Abstract away common/repeated functionality that remains the same for multiple models whose use is obvious to the | |
| reader. | |
| 2) Do not abstract essential parts of the model (such as loss terms, optimisers, layer names). Name these variables |