Bram Wasti

@Jia-HongHenryLee You'll need to build for those architectures and place them inside the correct folders in jniLibs. See armeabi-v7a for an example.

You're going to want to compile with -O3, let me add that to the CMakeLists in caffe2 :) To do so yourself right now, edit [this line](https://github.com/caffe2/caffe2/blob/master/CMakeLists.txt#L93) to say set(CMAKE_CXX_FLAGS...

The Makefile should read ``` bindings: $(DEBUG_OBJ_DIR) $(DEBUG_OBJ_FILES) $(DEBUG_WEBCORE_OBJ_FILES) $(DEBUG_SQLITE_OBJ_FILES) $(DEBUG_NODE_OS_OBJ_FILES) $(DEBUG_BUILTINS_OBJ_FILES) $(DEBUG_IO_FILES) ``` (that first `$(DEBUG_OBJ_DIR)` is missing)

the other fix is to add DEBUG_IO_FILES to the headers

would you be able to post the entire output from a run of setup.py? I'm not seeing the root cause of the failure here ``` python setup.py install &> /tmp/setup_fail.log...

this seems similar to issue #77 - please try this fix: https://github.com/pytorch/tvm/issues/77#issuecomment-519614588

Have you built PyTorch from source?

> > Have you built PyTorch from source? > > I used pip to install pytorch. Is it necessary to build pytorch from souce? It shouldn't be necessary to build...

Hey, you've certainly caught a bug. Would you mind printing `trace_tvm.graph`? Generally, to use the `to_relay` API, you'll need the entire graph representable in relay (note the error message) --...

you'll need to change some more configuration params (e.g. n_local_heads should be 8) I'd copy them from here https://huggingface.co/docs/transformers/main/model_doc/mistral#transformers.MistralConfig