optimum

/

m2m100_418M

like 1

Hello!
Why doesn't the model work?
If you can, I would appreciate if you could update the inference code in the model card
Thank you

I have the same issue!
Moreover, whenever I run the example, I keep getting this error :

 [ONNXRuntimeError] : 6 : RUNTIME_EXCEPTION : Non-zero status code returned while running Add node. Name:'Add_749' Status Message: /onnxruntime_src/onnxruntime/core/providers/cpu/math/element_wise_ops.h:540 void onnxruntime::BroadcastIterator::Init(ptrdiff_t, ptrdiff_t) axis == 1 || axis == largest was false. Attempting to broadcast an axis by a dimension other than 1. 3 by 12

I'm running it on google colab. And so I'm installing the pkgs as follows:

! pip install --upgrade pip
! pip install sentencepiece
! pip install transformers
! pip install langchain
! pip install onnxruntime
! pip install onnx
! pip install optimum
! pip install onnxruntime-gpu

Hi @Centaur31 and @YJ4219 ,

Thanks for opening the issue. The checkpoint has been uploaded for a while and according to the recent refactoring of the optimum library, it might be outdated and incompatible with the lastest Optimum. I will update it today when I have the bandwidth.

@Centaur31 @YJ4219 , hey folks, the checkpoint is updated, could you give it a try and tell me if it works? (tested with optimum v1.14.1 should be good)

I also added a export guide on the model card, if you meet outdated ONNX checkpoints again, don't hesitate to export a new one yourself, as there has been some update in the optimum library (and even better, open a pull request in the hub and tag the Optimum team members for a review 🤗!).

@Jingya Thank you very much for your response!
I have already converted the model manually (in the same way you provided in the model card) and everything seems to work fine!
Thank you!