Get Error while optimizing SDXL of DirectML example

[klin2024]

Describe the bug
Get Error while optimizing SDXL

To Reproduce
Command: python stable_diffusion_xl.py --optimize

Expected behavior
Optimize successfully.

Olive logs

Optimizing unet
[2024-05-19 05:11:53,655] [INFO] [run.py:279:run] Loading Olive module configuration from: C:\Users\AMD\miniconda3\envs\sdxl_turbo\lib\site-packages\olive\olive_config.json
[2024-05-19 05:11:53,665] [INFO] [accelerator_creator.py:224:create_accelerators] Running workflow on accelerator specs: gpu-dml
[2024-05-19 05:11:53,666] [INFO] [engine.py:107:initialize] Using cache directory: cache
[2024-05-19 05:11:53,666] [INFO] [engine.py:263:run] Running Olive on accelerator: gpu-dml
[2024-05-19 05:11:53,666] [INFO] [engine.py:1075:_create_system] Creating target system ...
[2024-05-19 05:11:53,667] [INFO] [engine.py:1078:_create_system] Target system created in 0.000997 seconds
[2024-05-19 05:11:53,667] [INFO] [engine.py:1087:_create_system] Creating host system ...
[2024-05-19 05:11:53,667] [INFO] [engine.py:1090:_create_system] Host system created in 0.000000 seconds
[2024-05-19 05:11:53,692] [INFO] [engine.py:865:_run_pass] Running pass convert:OnnxConversion
[2024-05-19 05:18:58,157] [INFO] [engine.py:952:_run_pass] Pass convert:OnnxConversion finished in 424.462611 seconds
[2024-05-19 05:18:58,168] [INFO] [engine.py:865:_run_pass] Running pass optimize:OrtTransformersOptimization
fusion: 100%|█████████████████████████████████████████████████████████████████████████| 18/18 [07:59<00:00, 26.61s/it]
Traceback (most recent call last):
File "D:\olive_0516\Olive\examples\directml\stable_diffusion_xl\stable_diffusion_xl.py", line 637, in
main()
File "D:\olive_0516\Olive\examples\directml\stable_diffusion_xl\stable_diffusion_xl.py", line 603, in main
optimize(
File "D:\olive_0516\Olive\examples\directml\stable_diffusion_xl\stable_diffusion_xl.py", line 369, in optimize
olive_run(olive_config)
File "C:\Users\KT\miniconda3\envs\sdxl_turbo\lib\site-packages\olive\workflows\run\run.py", line 301, in run
return run_engine(package_config, run_config, data_root)
File "C:\Users\KT\miniconda3\envs\sdxl_turbo\lib\site-packages\olive\workflows\run\run.py", line 255, in run_engine
engine.run(
File "C:\Users\KT\miniconda3\envs\sdxl_turbo\lib\site-packages\olive\engine\engine.py", line 265, in run
run_result = self.run_accelerator(
File "C:\Users\KT\miniconda3\envs\sdxl_turbo\lib\site-packages\olive\engine\engine.py", line 337, in run_accelerator
output_footprint = self.run_no_search(
File "C:\Users\KT\miniconda3\envs\sdxl_turbo\lib\site-packages\olive\engine\engine.py", line 429, in run_no_search
should_prune, signal, model_ids = self._run_passes(
File "C:\Users\KT\miniconda3\envs\sdxl_turbo\lib\site-packages\olive\engine\engine.py", line 827, in _run_passes
model_config, model_id = self._run_pass(
File "C:\Users\KT\miniconda3\envs\sdxl_turbo\lib\site-packages\olive\engine\engine.py", line 935, in _run_pass
output_model_config = host.run_pass(p, input_model_config, data_root, output_model_path, pass_search_point)
File "C:\Users\KT\miniconda3\envs\sdxl_turbo\lib\site-packages\olive\systems\local.py", line 32, in run_pass
output_model = the_pass.run(model, data_root, output_model_path, point)
File "C:\Users\KT\miniconda3\envs\sdxl_turbo\lib\site-packages\olive\passes\olive_pass.py", line 223, in run
output_model = self._run_for_config(model, data_root, config, output_model_path)
File "C:\Users\KT\miniconda3\envs\sdxl_turbo\lib\site-packages\olive\passes\onnx\transformer_optimization.py", line 332, in run_for_config
return model_proto_to_olive_model(optimizer.model, output_model_path, config)
File "C:\Users\KT\miniconda3\envs\sdxl_turbo\lib\site-packages\olive\passes\onnx\common.py", line 170, in model_proto_to_olive_model
has_external_data = model_proto_to_file(
File "C:\Users\KT\miniconda3\envs\sdxl_turbo\lib\site-packages\olive\passes\onnx\common.py", line 114, in model_proto_to_file
onnx.save_model(model, str(output_path))
File "C:\Users\KT\miniconda3\envs\sdxl_turbo\lib\site-packages\onnx_init.py", line 327, in save_model
serialized = _get_serializer(format, model_filepath).serialize_proto(proto)
File "C:\Users\KT\miniconda3\envs\sdxl_turbo\lib\site-packages\onnx\serialization.py", line 100, in serialize_proto
result = proto.SerializeToString()
ValueError: Message onnx.ModelProto exceeds maximum protobuf size of 2GB: 5136056262

In model_proto_to_file, model_size was 841088966.

[Jay19751103]

Hi Olive team
It looks that file size over 2GB during optimization pass. convert pass have flag to prevent this.
But the flag is not set in Transformation pass. I tried to add it can pass transformation under tag v0.6.2

Could any one fix this?