Failed to export onnx model from flow

[zhangyike]

env:
torch.version = 2.0.1+cu118
onnx.version = '1.16.0'
command:
python cosyvoice/bin/export_onnx.py --model_dir $dir
error logs:
/root/miniconda3/envs/cosyvoice/lib/python3.8/site-packages/diffusers/models/attention_processor.py:645: TracerWarning: Converting a tensor to a Python boolean might cause the trace to be incorrect. We can't record the data flow of Python values, so this value will be treated as a constant in the future. This means that the trace might not generalize to other inputs!
if current_length != target_length:
/root/miniconda3/envs/cosyvoice/lib/python3.8/site-packages/diffusers/models/attention_processor.py:660: TracerWarning: Converting a tensor to a Python boolean might cause the trace to be incorrect. We can't record the data flow of Python values, so this value will be treated as a constant in the future. This means that the trace might not generalize to other inputs!
if attention_mask.shape[0] < batch_size * head_size:
/apdcephfs_cq8/share_784792/users/yikezhang/public_repo/CosyVoice/examples/libritts/cosyvoice/cosyvoice/bin/../../third_party/Matcha-TTS/matcha/models/components/decoder.py:149: TracerWarning: Converting a tensor to a Python boolean might cause the trace to be incorrect. We can't record the data flow of Python values, so this value will be treated as a constant in the future. This means that the trace might not generalize to other inputs!
assert inputs.shape[1] == self.channels
============= Diagnostic Run torch.onnx.export version 2.0.1+cu118 =============
verbose: False, log level: Level.ERROR
======================= 0 NONE 0 NOTE 0 WARNING 1 ERROR ========================
ERROR: missing-standard-symbolic-function

Exporting the operator 'aten::scaled_dot_product_attention' to ONNX opset version 18 is not supported. Please feel free to request support or submit a pull request on PyTorch GitHub: https://github.com/pytorch/pytorch/issues.
None

Traceback (most recent call last):
File "cosyvoice/bin/export_onnx.py", line 112, in
main()
File "cosyvoice/bin/export_onnx.py", line 68, in main
torch.onnx.export(
File "/root/miniconda3/envs/cosyvoice/lib/python3.8/site-packages/torch/onnx/utils.py", line 506, in export
_export(
File "/root/miniconda3/envs/cosyvoice/lib/python3.8/site-packages/torch/onnx/utils.py", line 1548, in _export
graph, params_dict, torch_out = _model_to_graph(
File "/root/miniconda3/envs/cosyvoice/lib/python3.8/site-packages/torch/onnx/utils.py", line 1117, in _model_to_graph
graph = _optimize_graph(
File "/root/miniconda3/envs/cosyvoice/lib/python3.8/site-packages/torch/onnx/utils.py", line 665, in _optimize_graph
graph = _C._jit_pass_onnx(graph, operator_export_type)
File "/root/miniconda3/envs/cosyvoice/lib/python3.8/site-packages/torch/onnx/utils.py", line 1901, in _run_symbolic_function
raise errors.UnsupportedOperatorError(
torch.onnx.errors.UnsupportedOperatorError: Exporting the operator 'aten::scaled_dot_product_attention' to ONNX opset version 18 is not supported. Please feel free to request support or submit a pull request on PyTorch GitHub: https://github.com/pytorch/pytorch/issues.

[aluminumbox]

use torch==2.2+, but onnx model rtf is not very stable