Spaces:
Running
on
CPU Upgrade
Model request failed (fp32)
Hello @clefourrier ! My model failed according to the request log here. Can you kindly check it out?
Additional info, the model is in fp32. I set the eval request at fp16. Do I have to do anything, or does your script handle the conversion of the precision from 32 to 16?
Thank you!
Hi
@jaspercatapang
!
Thank you for this super clear issue!
Our script handles fp16 conversion, no worries.
Evaluation failed because we couldn't manage to download your model, model loading failed on file pytorch_model-00028-of-00029.bin.
Could you convert your model to safetensors format?
Hi
@clefourrier
, I uploaded the safetensors version of my model. You may check it here for reference. Thanks.
Hi
@jaspercatapang
, thnak you :)
I relaunched your model manually, we'll see how it goes this time.
Hello. Just commenting since I migrated accounts. I am the OP of this thread.
Hi @clefourrier , just wanted to ask why the same models have been stuck in the pending queue for days, while newer submissions go straight away to the running queue. Or I may be mistaken...? Thanks!
Hi! A number of request files were changed from .json to not having an extension during relaunches (which made them appear in the queue, but were not interpreted as valid request files by the backend).
It's been fixed!