Why are giant models downloaded during build?

[mattroos]

I'm building from source on Linux, using the build.sh:

 ./build.sh --config Release --build_shared_lib --parallel --build_wheel

After building and running some tests, a lot of giant deep learning models are downloaded. What is the purpose, where are those files saved, and when are they used?

.sStart downloading model bvlc_alexnet from https://s3.amazonaws.com/download.onnx/models/opset_9/bvlc_alexnet.tar.gz
Done
.sStart downloading model densenet121 from https://s3.amazonaws.com/download.onnx/models/opset_9/densenet121.tar.gz
Done
.sStart downloading model inception_v1 from https://s3.amazonaws.com/download.onnx/models/opset_9/inception_v1.tar.gz
Done
.sStart downloading model inception_v2 from https://s3.amazonaws.com/download.onnx/models/opset_9/inception_v2.tar.gz
Done
.sStart downloading model resnet50 from https://s3.amazonaws.com/download.onnx/models/opset_9/resnet50.tar.gz
Done
2021-05-06 09:19:52.837508132 [W:onnxruntime:, graph.cc:3106 CleanUnusedInitializers] Removing initializer 'gpu_0/imagenet1k_blobs_queue_f22e83c9-22cd-4a8b-a66d-113af6b832b4_0'. It is not used by any node and should be removed from the model.
.sStart downloading model shufflenet from https://s3.amazonaws.com/download.onnx/models/opset_9/shufflenet.tar.gz
Done
.sStart downloading model squeezenet from https://s3.amazonaws.com/download.onnx/models/opset_9/squeezenet.tar.gz
Done
.sStart downloading model vgg19 from https://s3.amazonaws.com/download.onnx/models/opset_9/vgg19.tar.gz
Done
[libprotobuf WARNING google/protobuf/io/coded_stream.cc:604] Reading dangerously large protocol message.  If the message turns out to be larger than 2147483647 bytes, parsing will be halted for security reasons.  To increase the limit (or to disable these warnings), see CodedInputStream::SetTotalBytesLimit() in google/protobuf/io/coded_stream.h.
[libprotobuf WARNING google/protobuf/io/coded_stream.cc:81] The total number of bytes read was 574674712
.sStart downloading model zfnet512 from https://s3.amazonaws.com/download.onnx/models/opset_9/zfnet512.tar.gz
Done

[pranav-prakash]

I think this is for using onnxruntime to run the standard onnx backend model test suite
https://github.com/microsoft/onnxruntime/blob/master/docs/Model_Test.md
https://github.com/onnx/onnx/blob/master/docs/OnnxBackendTest.md

[mattroos]

Thanks, @pranav-prakash. Any idea what the default path is for those models? I can't seem to find them anywhere.

[pranav-prakash]

I'm not too sure, try ~/.onnx?
https://github.com/onnx/onnx/blob/ea8e57a71e01d1e70b188465c028eb3c6c6cc118/onnx/backend/test/runner/__init__.py#L206

[mattroos]

Yes, @pranav-prakash, ~/.onnx was correct. Thanks!