Erro!!!

[little-black-f]

Why can't I find the canonical_data file in MNLI, and when running the example provided in the webpage, I got the following error:

AssertionError Traceback (most recent call last) /data-tmp/TM-DNN/MT-DNN-master/mtdnn/data_builder_mtdnn.py in load_and_build_data(self, dump_rows) 131 task_load_func = self.supported_tasks_loader_map[name] --> 132 data = task_load_func(in_file_path, data_opts) 133 processed_rows = process_data_and_dump_rows(

/data-tmp/TM-DNN/MT-DNN-master/mtdnn/tasks/utils.py in load_mnli(file_path, kwargs) 125 blocks = line.strip().split("\t") --> 126 assert len(blocks) > 9 127 if blocks[-1] == "-":

AssertionError:

During handling of the above exception, another exception occurred:

OSError Traceback (most recent call last) in 5 data_dir=DATA_SOURCE_DIR_MNLI, 6 canonical_data_suffix="canonical_data", ----> 7 dump_rows=True, 8 ) 9

/data-tmp/TM-DNN/MT-DNN-master/mtdnn/data_builder_mtdnn.py in init(self, tokenizer, task_defs, do_lower_case, data_dir, canonical_data_suffix, dump_rows) 196 ) 197 self.processed_tasks_data = self.task_data_loader.load_and_build_data( --> 198 self.save_to_file 199 ) 200

/data-tmp/TM-DNN/MT-DNN-master/mtdnn/data_builder_mtdnn.py in load_and_build_data(self, dump_rows) 145 ) 146 except Exception as ex: --> 147 raise IOError(ex) 148 return processed_data 149

OSError:

which happened in : Data Preprocessing Create the Data Builder Object

Is this related to my pytorch using version 1.5.0？

[matteoghera]

Your MNLI data is corrupt. You can try to change assert with this if: if len(blocks) > 9: continue This omits the corrupt line.