[Bug] have a problem when convert the pytorch model to onnx using tools/deploy.py #2846
Open
1 of 3 tasks
Comments
|
maybe you should add "from mmdeploy.apis.tensorrt import onnx2tensorrt" in tools/deploy.py |
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Checklist
Describe the bug
mmsegmentation-main/mmdeploy$ python tools/deploy.py \
11/22 15:16:18 - mmengine - INFO - Start pipeline mmdeploy.apis.pytorch2onnx.torch2onnx in subprocess
11/22 15:16:19 - mmengine - WARNING - Failed to search registry with scope "mmseg" in the "Codebases" registry tree. As a workaround, the current "Codebases" registry in "mmdeploy" is used to build instance. This may cause unexpected failure when running the built modules. Please check whether "mmseg" is a correct scope, or whether the registry is initialized.
11/22 15:16:19 - mmengine - WARNING - Failed to search registry with scope "mmseg" in the "mmseg_tasks" registry tree. As a workaround, the current "mmseg_tasks" registry in "mmdeploy" is used to build instance. This may cause unexpected failure when running the built modules. Please check whether "mmseg" is a correct scope, or whether the registry is initialized.
/home/itdtirtx/anaconda3/envs/tgp/lib/python3.8/site-packages/mmcv/cnn/bricks/hsigmoid.py:35: UserWarning: In MMCV v1.4.4, we modified the default value of args to align with PyTorch official. Previous Implementation: Hsigmoid(x) = min(max((x + 1) / 2, 0), 1). Current Implementation: Hsigmoid(x) = min(max((x + 3) / 6, 0), 1).
warnings.warn(
check in_channels (16, 24, 960)
/mmsegmentation-main/mmsegmentation/mmseg/models/decode_heads/decode_head.py:121: UserWarning: For binary segmentation, we suggest using
out_channels = 1to define the outputchannels of segmentor, and usethresholdto convertseg_logitsinto a predictionapplying a thresholdwarnings.warn('For binary segmentation, we suggest using'
/mmsegmentation-main/mmsegmentation/mmseg/models/losses/cross_entropy_loss.py:250: UserWarning: Default
avg_non_ignoreis False, if you would like to ignore the certain label and average loss over non-ignore labels, which is the same with PyTorch official cross_entropy, setavg_non_ignore=True.warnings.warn(
Loads checkpoint by local backend from path: ../work_dirs/mobilenet-v3-d8_lraspp_4xb4-320k_skinseg-512x512-1122/iter_80000.pth
11/22 15:16:20 - mmengine - WARNING - DeprecationWarning: get_onnx_config will be deprecated in the future.
11/22 15:16:20 - mmengine - INFO - Export PyTorch model to ONNX: mmdeploy_models/mmseg/ort-1122/end2end.onnx.
11/22 15:16:20 - mmengine - WARNING - Can not find torch.nn.functional.scaled_dot_product_attention, function rewrite will not be applied
/home/itdtirtx/anaconda3/envs/tgp/lib/python3.8/site-packages/mmdeploy/core/optimizers/function_marker.py:160: TracerWarning: Converting a tensor to a Python integer 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!
ys_shape = tuple(int(s) for s in ys.shape)
/home/itdtirtx/anaconda3/envs/tgp/lib/python3.8/site-packages/mmdeploy/codebase/mmseg/models/segmentors/base.py:47: TracerWarning: Converting a tensor to a Python integer 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!
img_shape = [int(val) for val in img_shape]
/home/itdtirtx/anaconda3/envs/tgp/lib/python3.8/site-packages/mmcv/cnn/bricks/conv2d_adaptive_padding.py:50: TracerWarning: Converting a tensor to a Python float 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!
output_h = math.ceil(img_h / stride_h)
/home/itdtirtx/anaconda3/envs/tgp/lib/python3.8/site-packages/mmcv/cnn/bricks/conv2d_adaptive_padding.py:51: TracerWarning: Converting a tensor to a Python float 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!
output_w = math.ceil(img_w / stride_w)
/home/itdtirtx/anaconda3/envs/tgp/lib/python3.8/site-packages/mmcv/cnn/bricks/conv2d_adaptive_padding.py:53: 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!
max((output_h - 1) * self.stride[0] +
/home/itdtirtx/anaconda3/envs/tgp/lib/python3.8/site-packages/mmcv/cnn/bricks/conv2d_adaptive_padding.py:56: 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!
max((output_w - 1) * self.stride[1] +
/home/itdtirtx/anaconda3/envs/tgp/lib/python3.8/site-packages/mmcv/cnn/bricks/conv2d_adaptive_padding.py:58: 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 pad_h > 0 or pad_w > 0:
/home/itdtirtx/anaconda3/envs/tgp/lib/python3.8/site-packages/mmdeploy/codebase/mmseg/models/segmentors/base.py:61: 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 seg_logit.shape[1] == 1:
/home/itdtirtx/anaconda3/envs/tgp/lib/python3.8/site-packages/torch/onnx/utils.py:687: UserWarning: The shape inference of prim::Constant type is missing, so it may result in wrong shape inference for the exported graph. Please consider adding it in symbolic function. (Triggered internally at ../torch/csrc/jit/passes/onnx/shape_type_inference.cpp:1884.)
_C._jit_pass_onnx_graph_shape_type_inference(
/home/itdtirtx/anaconda3/envs/tgp/lib/python3.8/site-packages/torch/onnx/utils.py:687: UserWarning: Constant folding - Only steps=1 can be constant folded for opset >= 10 onnx::Slice op. Constant folding not applied. (Triggered internally at ../torch/csrc/jit/passes/onnx/constant_fold.cpp:179.)
_C._jit_pass_onnx_graph_shape_type_inference(
/home/itdtirtx/anaconda3/envs/tgp/lib/python3.8/site-packages/torch/onnx/utils.py:1178: UserWarning: The shape inference of prim::Constant type is missing, so it may result in wrong shape inference for the exported graph. Please consider adding it in symbolic function. (Triggered internally at ../torch/csrc/jit/passes/onnx/shape_type_inference.cpp:1884.)
_C._jit_pass_onnx_graph_shape_type_inference(
/home/itdtirtx/anaconda3/envs/tgp/lib/python3.8/site-packages/torch/onnx/utils.py:1178: UserWarning: Constant folding - Only steps=1 can be constant folded for opset >= 10 onnx::Slice op. Constant folding not applied. (Triggered internally at ../torch/csrc/jit/passes/onnx/constant_fold.cpp:179.)
_C._jit_pass_onnx_graph_shape_type_inference(
11/22 15:16:25 - mmengine - INFO - Execute onnx optimize passes.
11/22 15:16:25 - mmengine - WARNING - Can not optimize model, please build torchscipt extension.
More details: mmdeploy/docs/en/experimental/onnx_optimizer.md at main · open-mmlab/mmdeploy
11/22 15:16:25 - mmengine - INFO - Finish pipeline mmdeploy.apis.pytorch2onnx.torch2onnx
11/22 15:16:26 - mmengine - INFO - Start pipeline mmdeploy.apis.utils.utils.to_backend in main process
11/22 15:16:26 - mmengine - INFO - Finish pipeline mmdeploy.apis.utils.utils.to_backend
11/22 15:16:26 - mmengine - INFO - visualize onnxruntime model start.
11/22 15:16:30 - mmengine - WARNING - Failed to search registry with scope "mmseg" in the "Codebases" registry tree. As a workaround, the current "Codebases" registry in "mmdeploy" is used to build instance. This may cause unexpected failure when running the built modules. Please check whether "mmseg" is a correct scope, or whether the registry is initialized.
11/22 15:16:30 - mmengine - WARNING - Failed to search registry with scope "mmseg" in the "mmseg_tasks" registry tree. As a workaround, the current "mmseg_tasks" registry in "mmdeploy" is used to build instance. This may cause unexpected failure when running the built modules. Please check whether "mmseg" is a correct scope, or whether the registry is initialized.
11/22 15:16:30 - mmengine - WARNING - Failed to search registry with scope "mmseg" in the "backend_segmentors" registry tree. As a workaround, the current "backend_segmentors" registry in "mmdeploy" is used to build instance. This may cause unexpected failure when running the built modules. Please check whether "mmseg" is a correct scope, or whether the registry is initialized.
11/22 15:16:30 - mmengine - INFO - Successfully loaded onnxruntime custom ops from /home/itdtirtx/anaconda3/envs/tgp/lib/python3.8/site-packages/mmdeploy/lib/libmmdeploy_onnxruntime_ops.so
11/22 15:17:05 - mmengine - INFO - visualize onnxruntime model success.
11/22 15:17:05 - mmengine - INFO - visualize pytorch model start.
11/22 15:17:08 - mmengine - WARNING - Failed to search registry with scope "mmseg" in the "Codebases" registry tree. As a workaround, the current "Codebases" registry in "mmdeploy" is used to build instance. This may cause unexpected failure when running the built modules. Please check whether "mmseg" is a correct scope, or whether the registry is initialized.
11/22 15:17:08 - mmengine - WARNING - Failed to search registry with scope "mmseg" in the "mmseg_tasks" registry tree. As a workaround, the current "mmseg_tasks" registry in "mmdeploy" is used to build instance. This may cause unexpected failure when running the built modules. Please check whether "mmseg" is a correct scope, or whether the registry is initialized.
/home/itdtirtx/anaconda3/envs/tgp/lib/python3.8/site-packages/mmcv/cnn/bricks/hsigmoid.py:35: UserWarning: In MMCV v1.4.4, we modified the default value of args to align with PyTorch official. Previous Implementation: Hsigmoid(x) = min(max((x + 1) / 2, 0), 1). Current Implementation: Hsigmoid(x) = min(max((x + 3) / 6, 0), 1).
warnings.warn(
check in_channels (16, 24, 960)
/mmsegmentation-main/mmsegmentation/mmseg/models/decode_heads/decode_head.py:121: UserWarning: For binary segmentation, we suggest using
out_channels = 1to define the outputchannels of segmentor, and usethresholdto convertseg_logitsinto a predictionapplying a thresholdwarnings.warn('For binary segmentation, we suggest using'
/mmsegmentation-main/mmsegmentation/mmseg/models/losses/cross_entropy_loss.py:250: UserWarning: Default
avg_non_ignoreis False, if you would like to ignore the certain label and average loss over non-ignore labels, which is the same with PyTorch official cross_entropy, setavg_non_ignore=True.warnings.warn(
Loads checkpoint by local backend from path: ../work_dirs/mobilenet-v3-d8_lraspp_4xb4-320k_skinseg-512x512-1122/iter_80000.pth
11/22 15:17:44 - mmengine - INFO - visualize pytorch model success.
11/22 15:17:44 - mmengine - INFO - All process success.
Reproduction
python tools/deploy.py
configs/mmseg/segmentation_onnxruntime_static-512x512.py
../configs/mobilenet_v3/mobilenet-v3-d8_lraspp_4xb4-320k_skinseg-512x512.py
../work_dirs/mobilenet-v3-d8_lraspp_4xb4-320k_skinseg-512x512-1122/iter_80000.pth
demo/15kdata_20240501_44.jpg
--work-dir mmdeploy_models/mmseg/ort-1122
Environment
Error traceback
No response
The text was updated successfully, but these errors were encountered: