right train env
Opened this issue · 0 comments
vagetablechicken commented
I use python 3.8.19, I think 3.8.x are fine.
timm==0.9.16
, cuz ModelEmaV3 is from huggingface/pytorch-image-models@dd84ef2
I download the dataset from kaggle, ref https://www.kaggle.com/competitions/multi-ffdi/data. Only phase1.
The data needs cvt, remove header, and use whitespace to sep, not comma.
file_path = 'trainset_label.txt'
prefix = 'dataset/trainset/' # 要添加的前缀
output_file_path = 'train.txt'
# file_path = 'valset_label.txt'
# prefix = 'dataset/valset/' # 要添加的前缀
# output_file_path = 'val.txt'
# 读取文件,处理内容
with open(file_path, 'r', encoding='utf-8') as file:
lines = file.readlines()
# 去掉第一行并处理剩下的行
processed_lines = []
for line in lines[1:]: # 跳过第一行
new_line = prefix + line.replace(',', ' ') # 添加前缀并替换逗号为空格
processed_lines.append(new_line)
# 输出处理后的内容,或写回文件
with open(output_file_path, 'w', encoding='utf-8') as output_file:
output_file.writelines(processed_lines)
If no 8 gpus, change config in main.sh
, e.g. I have 4 gpus
CUDA_VISIBLE_DEVICES=0,1,2,3 python -m torch.distributed.launch --nproc_per_node=4 --use_env main_train.py
DEVICES and nproc need to be modified.