Andrew M. Sica

Sounds good! @aarnphm with 'use the docker container as an entrypoint for onnx-mlir' I assume this would be to take an ONNX model as input and handle compile as part...

If the community is willing to accept this capability, we will contribute the support and tests.

@aarnphm great - excited for v1! Is there anything in particular we can provide to help an evaluation?

Just a quick note: I was able to successfully load vgg16 using torchvision, e.g.,: `model = models.vgg16(pretrained=True)` I inspected the weights and compared them to my laptop, they matched my...

Sorry for leaving that out @ezyang ! The minimal example above is saved as follows: `torch.save(lstm.state_dict(), filepath)` We then load (after instantiating a model) with: `lstm.load_state_dict(torch.load(filepath),strict=True) ` @driazati I will...

specifying _use_new_zipfile_serialization=False on the model save fixed both my minimal example, as well as yolov5 model. I ended up loading the pre-trained yolov5 model on my x86 machine, saving it...

I've tested a set of additional models and see no other portability issues when specifying _use_new_zipfile_serialization=False on LE platform save and subsequent load/use on s390x. @driazati thank you, this is...