-
Hello,
In the paper it is stated that
> ... given a random list of n documents [d1, d2, ..., dn], we extract randomly from each a pair of spans, [s11, s12, ..., sn1, sn2].
I was wondering how t…
-
Good day,
Please, I'm having a problem setting up the caffe, because I get an error from the constrative_loss file, is like their is something wrong with file and how do I fix. thank you and I wait …
-
### Background
Currently, the progressive learning network transformer is learned as a byproduct of optimizing the softmax objective loss for classification accuracy.
Contrastive loss ([referenc…
-
Hi, this is more like a question than an issue.
How could one do a pretraining run, without a downstream task?
What I want to achieve is to pretrain a model (using maybe both contrastive losses) a…
-
Hey guys, thanks for this implementation. But I am not really sure, how this is related to the [TopNet paper](https://arxiv.org/pdf/2304.03372). Could you please elaborate?
-
Hey Nils thank you for your create work!
I want to fine tune the pooling layer of all-mpnet-base-v2 with domain specific samples.
The problem is that the amount of available negatives is low.
Cr…
ghost updated
2 years ago
-
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 …
-
Hi.
There's a problem when resume training.
I tried to restart training DASR using this :
```
python main.py --dir_data='my/path' \
--model='blindsr' \
--scal…
sujyQ updated
2 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 …
-
Dear authors,
I was trying to reproduce repllama/repmistral and to understand the logic of the code.
# 1. Why use cross-entropy loss in contrastive learning?
in this file `/usa/dayu/Tab…