ywjhs commented 4 months ago

训练可以正常进行，但是不知道训练中途保存的参数会保存在哪里，同时发现inference.py需要的标签checkpoint的内容是空的，请问配置文件里的checkpoint应该怎么填，什么时候填，以及训练时的参数会保存在哪里

Quiyyy commented 3 months ago

你好，请问问题解决了吗，我尝试在jupyter中训练checkpoints文件夹打不开，但是使用colab训练checkpoints文件可以正常打开，并且模型会正常保存

Quiyyy commented 3 months ago

问题解决了，模型其实已经保存了，只是文件太大打不开，你可以使用这段代码将checkpoints文件夹压缩 `import shutil import os

def compress_folder(folder_name, output_path): """ Compresses the specified folder into a zip file.

Parameters:
folder_name (str): The name of the folder to compress.
output_path (str): The path of the output zip file.
"""
folder_path = os.path.abspath(folder_name)
if not os.path.exists(folder_path):
    print(f'Folder "{folder_path}" does not exist.')
    return

shutil.make_archive(output_path, 'zip', folder_path)
print(f'Folder "{folder_path}" has been compressed into "{output_path}.zip"')

示例用法，将当前工作目录下的 "checkpoints" 文件夹压缩为 "checkpoints.zip"

compress_folder('checkpoints', 'checkpoints')`

LBH1024 / CAN

复现时遇到的问题 #41

示例用法，将当前工作目录下的 "checkpoints" 文件夹压缩为 "checkpoints.zip"