-
Skeletonization algorithms often produce small spurious branches. skan should be able to prune those branches and produce a revised skeleton. Additionally, one might want to prune branches based on ot…
-
#### Description
There is an implementation of the CART algorithm where the leafs are a kind of model (linear regression, svr, neural net). And the split is done thanks to a criterion that depends …
-
Does it support Knowledge Distillation?
-
启发OpenAI去做GPT的单个[sentiment neuron](https://openai.com/index/unsupervised-sentiment-neuron/),和后面Issue一样,不仅有具体的location还能editing并且操纵模型。[用GPT-4和数据寻找&解释GPT-2的neuron](https://openai.com/index/language-mode…
-
A general structure has to be agreed for the genetic algorithms. This may vary a little based on the algorithm in question, but a general approach is likely to be generally useful. I propose the follo…
-
### 🚀 The feature, motivation and pitch
Today, this triggers a cuda synchronization:
```
import torch
torch.set_default_device('cuda')
def f(x, y):
return x[y]
inps = (torch.randn(5…
-
-
It would be nice if there was a way to skip a single expression. It gets a bit cumbersome to write `#! format: off` and `#! format: on` before and after everything I want skipped. It would be cool if …
-
This is for discussing the future API of Graphs.jl
Here is a first shot to open the discussions:
In the light of [Why I no longer recommend Julia](https://discourse.julialang.org/t/discussion-on-w…
-
Hi @clbarnes,
@schlegelp and @NikDrummond reminded me about this effort today. This is not really an issue, but I figure the discussion may as well live with the repo. I was curious about where you …