Open bkari2 opened 1 month ago
in flashdiffusion.src.flash.data.datasets.dataset.py
self.setup()
If not done then
data_module.train_pipeline = DataPipeline( config=data_module.train_config, filters_mappers=data_module.train_filters_mappers, ) data_module.train_dataloader()
returns
Cell In[43], [line 145](vscode-notebook-cell:?execution_count=43&line=145) [142](vscode-notebook-cell:?execution_count=43&line=142) def dataloader(self): [143](vscode-notebook-cell:?execution_count=43&line=143) # return the loader [144](vscode-notebook-cell:?execution_count=43&line=144) return wds.WebLoader( --> [145](vscode-notebook-cell:?execution_count=43&line=145) self.pipeline, [146](vscode-notebook-cell:?execution_count=43&line=146) batch_size=None, [147](vscode-notebook-cell:?execution_count=43&line=147) num_workers=self.config.num_workers, [148](vscode-notebook-cell:?execution_count=43&line=148) ) AttributeError: 'DataPipeline' object has no attribute 'pipeline'
from webdataset import DataPipeline
Please let me know if you have a better option. Otherwise, I am happy to do the PR.
in flashdiffusion.src.flash.data.datasets.dataset.py
self.setup()
in order to setup properly.If not done then
returns
from webdataset import DataPipeline
is unnecessary.Please let me know if you have a better option. Otherwise, I am happy to do the PR.