need help with phi3-mini-lora

[reidisaki]

Has anyone gotten this to work? Also after running the command to make the .pte file it came out to 15.29 gigs.. I thought this would be much smaller? can you explain to me like i'm a dog. I'm an android developer with no python or llm experience but learning a lot!!

I keep getting this error when I try to run python export_model.py

python export_model.py
W0904 13:50:55.270000 21535 torch/distributed/elastic/multiprocessing/redirects.py:29] NOTE: Redirects are currently not supported in Windows or MacOs.
/Users/risaki/git/executorch/pyenv/lib/python3.12/site-packages/executorch/exir/dialects/edge/_ops.py:9: UserWarning: pkg_resources is deprecated as an API. See https://setuptools.pypa.io/en/latest/pkg_resources.html. The pkg_resources package is slated for removal as early as 2025-11-30. Refrain from using this package or pin to Setuptools<81.
import pkg_resources
Main
Exporting to aten dialect
Lowering to edge dialect
Exporting to executorch
Saving to phi3_mini_lora.pte
Done.
Exporting phi3-mini with LoRA for training
Exporting to aten dialect
Creating a joint forward-backwards graph for training
Lowering to edge dialect
Traceback (most recent call last):
File "/Users/risaki/git/executorch/examples/models/phi-3-mini-lora/export_model.py", line 144, in
main()
File "/Users/risaki/git/executorch/examples/models/phi-3-mini-lora/export_model.py", line 140, in main
export_phi3_mini_lora_training(lora_training_model)
File "/Users/risaki/git/executorch/examples/models/phi-3-mini-lora/export_model.py", line 102, in export_phi3_mini_lora_training
edge_program = to_edge(joint_graph)
^^^^^^^^^^^^^^^^^^^^
File "/Users/risaki/git/executorch/pyenv/lib/python3.12/site-packages/executorch/exir/program/_program.py", line 114, in wrapper
return func(self, *args, **kwargs)
^^^^^^^^^^^^^^^^^^^^^^^^^^^
File "/Users/risaki/git/executorch/pyenv/lib/python3.12/site-packages/executorch/exir/program/_program.py", line 1373, in to_edge
edge_programs[name] = _generate_edge_program(name, config, program)
^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
File "/Users/risaki/git/executorch/pyenv/lib/python3.12/site-packages/executorch/exir/program/_program.py", line 814, in _generate_edge_program
EXIRATenDialectVerifier(
File "/Users/risaki/git/executorch/pyenv/lib/python3.12/site-packages/executorch/exir/verification/verifier.py", line 74, in call
return self._check_graph_module(*args, **kwargs)
^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
File "/Users/risaki/git/executorch/pyenv/lib/python3.12/site-packages/torch/_export/verifier.py", line 263, in _check_graph_module
_check_valid_op(node.target)
File "/Users/risaki/git/executorch/pyenv/lib/python3.12/site-packages/torch/_export/verifier.py", line 244, in _check_valid_op
self.check_valid_op(op)
File "/Users/risaki/git/executorch/pyenv/lib/python3.12/site-packages/executorch/exir/verification/verifier.py", line 129, in check_valid_op
raise SpecViolationError(
torch._export.verifier.SpecViolationError:
Operator torch._ops.aten.empty_permuted.default is not in Core ATen opset (https://pytorch.org/docs/stable/torch.compiler_ir.html#core-aten-ir)."
There are a few things to try:

1. You can proceed with to_edge(compile_config=EdgeCompileConfig(_core_aten_ops_exception_list=[torch.ops.aten.empty_permuted.default])).
   Please make sure that the backend(s) you are planning to lower to is able to handle aten.empty_permuted.default, or you have a corresponding kernel linked to your runtime.

2. Sometimes inference and training gives slightly different op set. Try adding with torch.no_grad(): context manager if you are export for inference only.

3. If the error persists after 2, this is likely caused by torch.export() + core ATen decomposition producing unexpected operators for your model.
   If you believe this operator should be included into core ATen opset, please create an issue in https://github.com/pytorch/pytorch/issues and add module: core aten tag.

When I tried to replace the to_edge function call with :
to_edge(compile_config=EdgeCompileConfig(_core_aten_ops_exception_list=[torch.ops.aten.empty_permuted.default]))
the compiler couldn't find EdgeCompileConfig.. what library do I need or is this not the right path? and it should just work like the readme shows.?