-
Came here from https://youtu.be/EJ8okcxL2Iw?t=426
The talk says there are **token vectors** and **path vectors**.
I know what an AST is, but I am not that proficient in AI to answer what is the …
-
### 🐛 Describe the bug
Grouped convolutions should reduce the flops by dividing by the number of groups => ought to lead to massive speedup in training. This speedup however fails to materialize.
…
jxtps updated
2 years ago
-
Luckily, adjoint methods for (most) eigenvalue problems tend to be rather trivial _if you are optimizing the eigenvalue directly_. So, for example, if you want to match the effective index of two mode…
-
In the section named "What Do All the Colors Mean?", there's this paragraph:
"In the hidden layers, the lines are colored by the weights of the connections between neurons. Blue shows a positive weig…
-
Backpropagatable Linear Algebra methods are a vital part of Machine Learning including but not limited to Neural Nets. Here and there, requests for different LA methods pop up. More and more duplicate…
-
Karpathy, Andrej. 2015. [“The Unreasonable Effectiveness of Recurrent Neural Networks”](http://karpathy.github.io/2015/05/21/rnn-effectiveness/). Blogpost.
-
**Describe the bug**
When I was trying something like exponentially weighted moving average, I saw the gradients may be incorrect.
**To Reproduce**
```py
ti.init(arch=ti.cuda)
row_num, co…
-
## Keyword: sgd
### Doubly Stochastic Models: Learning with Unbiased Label Noises and Inference Stability
- **Authors:** Authors: Haoyi Xiong, Xuhong Li, Boyang Yu, Zhanxing Zhu, Dongrui Wu, Dejin…
-
I've noticed that my training take a long to train, probably because when using higher one has to loop through each batch individually as in
https://github.com/tristandeleu/pytorch-meta/blob/389e3…
-
# **Download Data**
If the Google drive links are dead, you can download data from [kaggle](https://www.kaggle.com/c/ml2021spring-hw1/data), and upload data manually to the workspace.
```py
t…