-
A usecase: storing a full backtracking pointer matrix can be okay for needleman/ctc alignment (4x memory saving compared to uint8 representation), if 2bit data type is used. Currently it's possible to…
-
### Feature request
Need to DeformableDetrModel using fp16.
~~~
from transformers import AutoImageProcessor
from trf.models import DeformableDetrModel
from PIL import Image
import torch
impor…
-
Add DataFrame.to_coco_dataset() -> Path, which returns the path to the COCO json file.
For a description of the format: https://cocodataset.org/#format-data
The json file will only set annotatio…
-
环境是用dockerfile 构建的,应该没问题。
我要训练faster-rcnn模型,这是我的数据目录,
```
my_faster_rcnn_data/
├── train
……
│ ├── aafa46eb-99b9-499f-ad82-6c23270cc923.jpg
│ ├── aafa46eb-99b9-499f-ad82-6c23270cc923.tx…
-
**Checklist**
1. I have searched related issues but cannot get the expected help.
2. The issue has not been fixed in the latest version.
**Describe the issue**
I'm trying to run the deformab…
-
### Motivation
I have benchmarked the performance of [llava-v1.6-vicuna-13b](https://huggingface.co/liuhaotian/llava-v1.6-vicuna-13b) with api server on 2*A30. The detailed data is as follows.
Whe…
-
## 🐛 Bug
`pip install -U -r requirements.txt` bugs out with `ModuleNotFoundError: No module named 'numpy'` error.
## To Reproduce (REQUIRED)
Clone the repository and run `pip install -U -r requir…
-
Just want to share how to run on the latest version of Detectron2 (v0.6):
## 1. Environment
CUDA 11.1
Torch>=1.9.0 ```pip install torch==1.9.0+cu111 torchvision==0.10.0+cu111 torchaudio==0.9.0 …
-
I have trained a model.my datasets is DIOR(tend to cocodatasets),but the output file scalars.json only have the map of every class,no map50,how can I get the every class's map50 instead of map?Should …
1Iris updated
8 months ago
-
Hello respected author. I trained hmr2 on my own server. However, when I wanted to perform an evaluation, I found that I need the following data: "download the Human3.6M, 3DPW, LSP-Extended, COCO, and…