OpenAssistant

/

oasst-rlhf-2-llama-30b-7k-steps-xor

like 25

Hello,

I am trying to apply the final step from the instructions provided on the "https://huggingface.co/OpenAssistant/oasst-rlhf-2-llama-30b-7k-steps-xor" page. Specifically, I am having issues while running the "xor_codec.py" script. The MD5 hashes of the output model files do not match the ones posted in the instructions (all previous steps are ok). Instead, these are the hashes I get:

20788f92b8e441ac1f4fc6f7f9ddd4af  ./pytorch_model-00005-of-00007.bin
ff416d3a305b4d74c0cad3323a7fa6eb  ./pytorch_model-00002-of-00007.bin
eeec4125e9c7560836b4873b6f8e3025  ./tokenizer.model
ed59bfee4e87b9193fea5897d610ab24  ./tokenizer_config.json
7e91abda6782ecbb58202e40cd618002  ./pytorch_model-00007-of-00007.bin
704373f0c0d62be75e5f7d41d39a7e57  ./special_tokens_map.json
9a4d2468ecf85bf07420b200faefb4af  ./config.json
148bfd184af630a7633b4de2f41bfc49  ./generation_config.json
f60488bcbf2cc29d591d442c7ff30af3  ./pytorch_model-00004-of-00007.bin
0c86e7d88aac7583d88953957dc543a6  ./pytorch_model-00003-of-00007.bin
492c80909300e3cb2e9dcd88f1cf15e3  ./pytorch_model-00006-of-00007.bin
deb33dd4ffc3d2baddcce275a00b7c1b  ./tokenizer.json
27b0dc092f99aa2efaf467b2d8026c3f  ./added_tokens.json
9723be7088896c9211b76be7aa88aafc  ./pytorch_model.bin.index.json
ec27a39bafd24e94c5a37b887d115561  ./pytorch_model-00001-of-00007.bin

I am completely puzzled and surprised that I can't get the correct hashes. I have exhausted all the troubleshooting options I could think of, and I would deeply appreciate any assistance or insights you might be able to provide.

Best regards,

Carmona