Use model as ONNX type
Is there any change to use this model as ONNX? If yes - how it can be smoothly converted to it?
Sorry, I am not familiar with ONNX
Got you. Is there any way to get this model in PyTorch format (pt, pth)? Then i can convert it in needed format.
Thanks in advance.
Any solutions for the ONNX? I'll struggeling with this, and need it for inference speed in production env
@intfloat
,
Here is one brief article - https://medium.com/deci-ai/tutorial-converting-a-pytorch-model-to-onnx-format-f1bbce156d2a
I also want to host e5 model in production using ONNX runtime as it's faster and can be operated in JAVA.
It would be very helpful if you can provide the model in .pt or .pth format . Right now the pytorch_model.bin is not a model file but a state file.
Could you please help.
@intfloat
,
Here is one brief article - https://medium.com/deci-ai/tutorial-converting-a-pytorch-model-to-onnx-format-f1bbce156d2a
I also want to host e5 model in production using ONNX runtime as it's faster and can be operated in JAVA.
It would be very helpful if you can provide the model in .pt or .pth format . Right now the pytorch_model.bin is not a model file but a state file.
Could you please help.
@preetnitkkr I would love to help, but the ONNX stuff goes beyond my skill set. Maybe you can convert this model to ONNX type and then open a pull request.
Turns out HuggingFace has an amazing interface to automatically export models to ONNX format at https://huggingface.co/spaces/onnx/export
Also thanks to romaindeveaud for the pull request.