-
I have tried a two-tower model (user and query) in a real industrial scenario using contrastive learning. The samples are all actual click samples, and the loss function is InfoNCE. I have a few quest…
-
Hi, Cheng Hao,
Congratulations! This paper made a huge contribution to the continuous learning of stereo matching and inspired me a lot.
Recently, I've been trying to reproduce your work, and I've r…
-
```
class Classifier(pl.LightningModule):
def __init__(self):
super().__init__()
self.MFB = MFB(512,768,True,256,64,0.1)
self.fin_y_shape = torch.nn.Linear(768,512)
self.fi…
-
Hi there,
I was searching how people implemented CLIP and found this repo. Problems/differences with the loss function based on the [CLIP Paper](https://arxiv.org/pdf/2103.00020.pdf):
1) If you …
-
I have a question regarding the weights used in CAV-MAE. It seems like the $\lambda_c$ could play an important role in the optimization. I understand it is due to the gradient scale but It is surprisi…
-
Hello,
This code is consuming tens of gigabytes of RAM and VRAM (more than 40GB of RAM and 35GB of VRAM). Any reason why? I used the "lowest" model and then it works fine... but using 35GB of VRAM …
-
Não encontrei código associado, todavia parece simples de reproduzir.
-
### What version of `mystmd` are you using?
v1.0.5
### How did you install myst?
npm
### What operating system are you using?
Mac
### Which area is this feature request for?
Export: LaTeX or PD…
-
- https://arxiv.org/abs/2103.16748
- 2021
Generative Adversarial Networks (GAN)は、大規模な画像データセットを用いて無条件に画像を生成する際に素晴らしい結果を出す。
しかし、生成された画像は、特に分散性の高いデータセット(寝室や教会など)では、まだ見分けがつきにくい。
本論文では、画像生成の限界をさらに押し広…
e4exp updated
3 years ago
-
[paper](https://arxiv.org/pdf/2205.01917.pdf)
## TL;DR
**problem :** 좋은 vision backbone 만들기. 분류 레이블에 대한 이미지 프리트레이닝, 이미지-텍스트 pair를 받아 contrastive loss로 학습되는 dual-encoder model, image 인코더가 있고 …