【无标题】parseq

一堆乱七八糟

conda create -n parseq python=3.9 -y
conda activate parseq
# CUDA 11.3
conda install pytorch==1.10.1 torchvision==0.11.2 torchaudio==0.10.1 cudatoolkit=11.3 -c pytorch -c conda-forge

# CUDA 10.2
pip install torch==1.10.1+cu102 torchvision==0.11.2+cu102 torchaudio==0.10.1 -f https://download.pytorch.org/whl/cu102/torch_stable.html

# Use specific platform build. Other PyTorch 1.13 options: cu116, cu117, rocm5.2
platform=CUDA 10.2
# Generate requirements files for specified PyTorch platform
make torch-CUDA 10.2

Install the project and core + train + test dependencies. Subsets: [train,test,bench,tune]

pip install -r requirements/core.cpu.txt -e .[train,test]

pip install -r requirements.txt 太难了就放弃吧，缺少module再注意pip就好，如果安装版本报错，可以参考一下 requirements.txt 指明的版本
一般都会有个 pip install -e .安装项目本身

./test.py outputs///checkpoints/last.ckpt # or use the released weights: ./test.py pretrained=parseq

python test.py work_dirs/parseq-bb5792a6.pt

python read.py work_dirs/parseq-bb5792a6.pt refine_iters:int=2 decode_ar:bool=false --images demo_images/*

做不下去了真的TTTTTTTT哭死
train 和test都运行不了
存档

TypeError: init() missing 21 required positional arguments: [When running test.py and read.py]
环境按照这个装的，只教了怎么用自己自制的数据集，类型是lmdb，tools里面有转化器