Open meharbhatia opened 8 months ago
@meharbhatia were you able to find a fix for this? I am also unable to download the model checkpoints as per the instructions in the https://github.com/microsoft/Oscar/blob/master/VinVL_DOWNLOAD.md file. Getting the same error.
yes , my work also report AuthenticationFailed, how to fix it?
I also meet this problem, do you know how to solve it?
Okay I had the same error. After a few attempts I noticed that I miss "?" character in my script after the name of a blob and before the SAS token. I had this in my input parameters: ".vhdse=2024" It should be this: ".vhd?se=2024"
After I fixed that it was working again.
Currently only the VinVL vision encoder model can be downloaded from an updated google link. We do not provide extracted features any more. If you need image features, you need to run the VinVL vision encoder model to extract features by yourselves. Sorry for the inconvenience!
I am using the given SAS token to extract this YAML file
./azcopy copy "https://biglmdiag.blob.core.windows.net/vinvl/pretrain_corpus/coco_flickr30k_googlecc_gqa_sbu_oi_x152c4big2exp168.yaml?sp=r&st=2023-08-28T01:12:41Z&se=3023-08-28T09:12:41Z&sv=2022-11-02&sr=c&sig=6R1YmWluiXmPLsdVn1rDUpeBp2SYBMxDjc6KoKNlY8Q%3D" data/ --recursive
but face the issue.
Can anyone help me resolve this? @xjli @ChunyuanLI @xiyinmsu @pzzhang. My work needs access to the VinVL files and I am stuck currently.