different MD5 but same SHA256

#1
by Hamlyn - opened

my downloaded model-0004 and model-0005 have different MD5 values with those shown on the homepage, but the same SHA256 values.
It seems that these files have some minor modifications.

Hamlyn changed discussion title from MD5值不同但是SHA256值相同 to different MD5 but same SHA256
Hamlyn changed discussion status to closed
Hamlyn changed discussion status to open

I got the following message when I the code.

OSError: Unable to load weights from pytorch checkpoint file for '/home/jovyan/.cache/huggingface/hub/models--codefuse-ai--CodeFuse-13B/snapshots/e00d68bfb3464c11c6929914a5d969d70b04c654/pytorch_model-00004-of-00006.bin' at '/home/jovyan/.cache/huggingface/hub/models--codefuse-ai--CodeFuse-13B/snapshots/e00d68bfb3464c11c6929914a5d969d70b04c654/pytorch_model-00004-of-00006.bin'. If you tried to load a PyTorch model from a TF 2.0 checkpoint, please set from_tf=True.

I'm experiencing this as well, with different md5 values but same sha256 values and the same error when loading the model

image.png

CodeFuse AI org

We are looking into it...

MD5 were calculated before upload, it seems that the two files were corrupted during uploading process. We will re-upload the corrupted files tonight.

Meanwhile, you can use ModelScope files instead.

curl -o pytorch_model-00004-of-00006.bin "https://modelscope.cn/api/v1/models/codefuse-ai/CodeFuse-13B/repo?Revision=master&FilePath=pytorch_model-00004-of-00006.bin"

curl -o pytorch_model-00005-of-00006.bin "https://modelscope.cn/api/v1/models/codefuse-ai/CodeFuse-13B/repo?Revision=master&FilePath=pytorch_model-00005-of-00006.bin"

Nice, I'am trying!

CodeFuse AI org

#4 and #5 model file on huggingface refreshed :)

codefuse-admin changed discussion status to closed

Sign up or log in to comment