-
히든 레이어 부분을 공부하다 의문이 든 점이 있습니다.
딥러닝에서 히든 레이어를 많이 쌓으면 더 정확해진다는 사실은 알고 있습니다. 이는 레이어의 층을 두껍게 쌓는다는 의미인 것 같습니다.
여기서 의문점이 든 것이, 퍼셉트론의 수를 늘려도 더 정확해질 거 같다고 생각을 했는데, 그 이유는 퍼셉트론의 수를 늘리면 그만큼 심화되게 데이터를 쌓을 수 …
-
Hello, the author! I followed all your feature extraction and charades training suggestions in your github homepage. But in my work environment, the reprocude results of LVTR-CLIP in 200-th epoch look…
-
## Description
Trying to get predictions for a simple logistic regression model I geet the error posted in the relative section. I am working with R on ArchLinux.
## Environment info (Required)
R…
-
#### Describe the bug
I encountered a memory issue when moving a model to CUDA using `model.to('cuda')`. Specifically, when using `torch.nn.ModuleList` or `torch.nn.Sequential` containing multiple …
-
Add a meaningful example case. Say a feedforward NN that approximates a linear function and link that to a Fortran or C program that uses this NN approx.
-
# ❓ Questions and Help
>pip install -v -U git+https://github.com/facebookresearch/xformers.git@main#egg=xformers
Using pip 23.2.1 from /data/anaconda3/envs/sd/lib/python3.10/site-packages/pip (pytho…
bzwqq updated
12 months ago
-
## ❓ How to do something using DETR
Hello, I want to run DETR on my customer datasets. when I use my COCO label instances_train2017.json, it shows the error.
Not using distributed mode
git:
…
-
请问如果想把swin-b改成swin-s的话,是不是只要
arg = dict(pretrain_img_size=384, window_size=12, embed_dim=128, out_indices=[2, 1],
depths=[2, 2, 18, 2], num_heads=[4, 8, 16, 32])
中dict这一块改成
dict(pre…
-
I read the paper, but I'm not quite sure what 'Output Head' means. For example, if I have a 3D molecular graph of my protein pocket, with each node embedded with features from the BLOSUM62 matrix, and…
-
I am currently working on a four wheel robot (One axis for steering and one axis for driving).
I have implemented the ackermann controller for the robot.
The ackermann controller does not give out…