Onnx error when saving model.

[Haradai]

Describe the bug I was following the initial example tutorial and managed to train a model. When I stop the training it cannot save the model because of issues with onnx. It looks like torch is unable to find onnx? But I have it installed and which the correct version.

Console logs / stack traces ^C[INFO] Learning was interrupted. Please wait while the graph is generated. Traceback (most recent call last): File "/Users/haradai/miniconda3/envs/mlagents/lib/python3.10/site-packages/torch/onnx/_internal/onnx_proto_utils.py", line 219, in _add_onnxscript_fn import onnx File "/Users/haradai/miniconda3/envs/mlagents/lib/python3.10/site-packages/onnx/init.py", line 6, in from onnx.external_data_helper import load_external_data_for_model, write_external_data_tensors, convert_model_to_external_data File "/Users/haradai/miniconda3/envs/mlagents/lib/python3.10/site-packages/onnx/external_data_helper.py", line 9, in from .onnx_pb import TensorProto, ModelProto, AttributeProto, GraphProto File "/Users/haradai/miniconda3/envs/mlagents/lib/python3.10/site-packages/onnx/onnx_pb.py", line 4, in from .onnx_ml_pb2 import * # noqa File "/Users/haradai/miniconda3/envs/mlagents/lib/python3.10/site-packages/onnx/onnx_ml_pb2.py", line 5, in from google.protobuf.internal import builder as _builder ImportError: cannot import name 'builder' from 'google.protobuf.internal' (/Users/haradai/miniconda3/envs/mlagents/lib/python3.10/site-packages/google/protobuf/internal/init.py)

The above exception was the direct cause of the following exception:

Traceback (most recent call last): File "/Users/haradai/miniconda3/envs/mlagents/bin/mlagents-learn", line 8, in sys.exit(main()) File "/Users/haradai/miniconda3/envs/mlagents/lib/python3.10/site-packages/mlagents/trainers/learn.py", line 267, in main run_cli(parse_command_line()) File "/Users/haradai/miniconda3/envs/mlagents/lib/python3.10/site-packages/mlagents/trainers/learn.py", line 263, in run_cli run_training(run_seed, options, num_areas) File "/Users/haradai/miniconda3/envs/mlagents/lib/python3.10/site-packages/mlagents/trainers/learn.py", line 137, in run_training tc.start_learning(env_manager) File "/Users/haradai/miniconda3/envs/mlagents/lib/python3.10/site-packages/mlagents_envs/timers.py", line 305, in wrapped return func(*args, **kwargs) File "/Users/haradai/miniconda3/envs/mlagents/lib/python3.10/site-packages/mlagents/trainers/trainer_controller.py", line 200, in start_learning self._save_models() File "/Users/haradai/miniconda3/envs/mlagents/lib/python3.10/site-packages/mlagents_envs/timers.py", line 305, in wrapped return func(*args, **kwargs) File "/Users/haradai/miniconda3/envs/mlagents/lib/python3.10/site-packages/mlagents/trainers/trainer_controller.py", line 80, in _save_models self.trainers[brain_name].save_model() File "/Users/haradai/miniconda3/envs/mlagents/lib/python3.10/site-packages/mlagents/trainers/trainer/rl_trainer.py", line 172, in save_model model_checkpoint = self._checkpoint() File "/Users/haradai/miniconda3/envs/mlagents/lib/python3.10/site-packages/mlagents_envs/timers.py", line 305, in wrapped return func(*args, **kwargs) File "/Users/haradai/miniconda3/envs/mlagents/lib/python3.10/site-packages/mlagents/trainers/trainer/rl_trainer.py", line 144, in _checkpoint export_path, auxillary_paths = self.model_saver.save_checkpoint( File "/Users/haradai/miniconda3/envs/mlagents/lib/python3.10/site-packages/mlagents/trainers/model_saver/torch_model_saver.py", line 60, in save_checkpoint self.export(checkpoint_path, behavior_name) File "/Users/haradai/miniconda3/envs/mlagents/lib/python3.10/site-packages/mlagents/trainers/model_saver/torch_model_saver.py", line 65, in export self.exporter.export_policy_model(output_filepath) File "/Users/haradai/miniconda3/envs/mlagents/lib/python3.10/site-packages/mlagents/trainers/torch_entities/model_serialization.py", line 164, in export_policy_model torch.onnx.export( File "/Users/haradai/miniconda3/envs/mlagents/lib/python3.10/site-packages/torch/onnx/utils.py", line 516, in export _export( File "/Users/haradai/miniconda3/envs/mlagents/lib/python3.10/site-packages/torch/onnx/utils.py", line 1686, in _export proto = onnx_proto_utils._add_onnxscript_fn( File "/Users/haradai/miniconda3/envs/mlagents/lib/python3.10/site-packages/torch/onnx/_internal/onnx_proto_utils.py", line 221, in _add_onnxscript_fn raise errors.OnnxExporterError("Module onnx is not installed!") from e torch.onnx.errors.OnnxExporterError: Module onnx is not installed!

Environment (please complete the following information):

• Unity Version: [e.g. Unity 2022.3.29f1]
• OS + version: [Mac M1, Macos Sonoma 14.5]
• ML-Agents version: release 21
• Torch version: 2.3.0
• Protobuf version: 3.19.6
• Onnx version: 1.12.0

NOTE: I found this issue where seems that in newer versions of onnx the error has been fixed but I cannot upgrade as mlagents needs onnx==1.12.0.

[Astronog]

pip install onnx

[onurkrsrml]

I also trying to install on M1 Mac ml agents release 21. I have issue with onnx on the this step(python -m pip install ./ml-agents). can you help me to fix it.

[Haradai]

pip install onnx

Not recognizing the package. I even tried installing it through brew and no luck

[github-actions[bot]]

This issue is stale because it has been open for 30 days with no activity.

[github-actions[bot]]

This issue was closed because it has been inactive for 14 days since being marked as stale. Please open a new issue for related bugs.

[mucunwuxian]

My solotion was the following.

pip install --upgrade pip
pip uninstall -y onnx onnxruntime
pip install onnx onnxruntime

Maybe onnx was old compared to torch.