haoshiming

Hi, I recently looked into your code (csrc/selective_scan/*), and I guess the largest part of the speedup comes from Ktraits::BlockScanT(smem_scan).InclusiveScan. But I see that besides that, you have also implemented...

GITHUB_TOKEN=xxxxx python3 create_text_dataset.py --dataset_name_or_path /data/to/my/swe_bench_orcale --output_dir ./base_datasets --prompt_style style-2 --file_source oracle --split train I get:  And I see that the repo indeed does not exist here:...