Hashes don't match?

#1
by RazielAU - opened

On https://huggingface.co/PygmalionAI/pygmalion-7b there is a list of file hashes for the Pygmalion 7B model as follows:

$ rhash -M *
4608facb4910118f8dfa80f090cbc4dc  config.json
2917a1cafb895cf57e746cfd7696bfe5  generation_config.json
98764eb949eea16f8e2e1c2d3dea0066  pytorch_model-00001-of-00002.bin
be9ba2f37228a0a9ea0eaf6530aba4de  pytorch_model-00002-of-00002.bin
81648ef3915ed2e83d49fed93122d53e  pytorch_model.bin.index.json
6b2e0a735969660e720c27061ef3f3d3  special_tokens_map.json
fdb311c39b8659a5d5c1991339bafc09  tokenizer.json
eeec4125e9c7560836b4873b6f8e3025  tokenizer.model
f0b65b44265ba51881b1e1881102504f  tokenizer_config.json

When I check the hashes of this repo using WSL2, I get the following:

$ rhash -M *
26f2f1a3b8db70fe9285ee19c3b2617b  README.md
a05498f1b7ad076ff7739f3b3b407451  config.json
148bfd184af630a7633b4de2f41bfc49  generation_config.json
0217bf6d894bc82c01575ead0f138184  huggingface-metadata.txt
557ba779b2895e7a7947e7cdc0197041  model-00001-of-00002.safetensors
b950b92be31b71040503bface1f972e2  model-00002-of-00002.safetensors
406284c6a66bc16d2a81cafb7d4dfdb5  model.safetensors.index.json
81648ef3915ed2e83d49fed93122d53e  pytorch_model.bin.index.json
6b2e0a735969660e720c27061ef3f3d3  special_tokens_map.json
fdb311c39b8659a5d5c1991339bafc09  tokenizer.json
eeec4125e9c7560836b4873b6f8e3025  tokenizer.model
9b3cf7b8c0e4783dbc1419b4cafe8e1e  tokenizer_config.json

Most of the files don't seem to match the hashes the Pygmalion team provided. Am I missing something?

EDIT: Okay, one obvious difference I've spotted is that I have the safetensors downloaded... but eve n the config.json doesn't match.

If few .json got updated when I uploaded safetensors take a look at the safetensors upload. Download the original bins and run md5sum and it'll match

Awesome, that makes sense, thanks!

RazielAU changed discussion status to closed

Sign up or log in to comment