Int8 model export fail with ImportError: cannot import name 'metanet_pb2' from partially initialized module 'caffe2.proto'

When I export the model to INT8, I get this error。I installed the latest code of d2go according to revision in #137,but still get this error

import os
import json
import numpy as np
import cv2
from detectron2.data.datasets import register_coco_instances
from detectron2.structures import BoxMode
from detectron2.utils.visualizer import Visualizer
from detectron2.data import MetadataCatalog, DatasetCatalog
from d2go.model_zoo import model_zoo
from d2go.utils.demo_predictor import DemoPredictor
import matplotlib.pyplot as plt
import random
from d2go.runner import GeneralizedRCNNRunner
import copy
from detectron2.data import build_detection_test_loader
from d2go.export.api import convert_and_export_predictor
from d2go.utils.testing.data_loader_helper import create_fake_detection_data_loader
from d2go.export.d2_meta_arch import patch_d2_meta_arch
import logging
from mobile_cv.predictor.api import create_predictor
from d2go.utils.demo_predictor import DemoPredictor

register_coco_instances("my_dataset_train", {}, "/home/liupeng/sdb1/Image/ScaleDetection/Train/humanDetec/Annotations/coco_info.json",
                        "/home/liupeng/sdb1/Image/ScaleDetection/Train/humanDetec/Images")

def prepare_for_launch():
    runner = GeneralizedRCNNRunner()
    cfg = runner.get_default_cfg()
    cfg.merge_from_file(model_zoo.get_config_file("faster_rcnn_fbnetv3a_C4.yaml"))
    cfg.MODEL_EMA.ENABLED = False
    cfg.DATASETS.TRAIN = ("my_dataset_train",)
    cfg.DATASETS.TEST = ("my_dataset_train",)
    cfg.DATALOADER.NUM_WORKERS = 2
    cfg.MODEL.WEIGHTS = model_zoo.get_checkpoint_url("faster_rcnn_fbnetv3a_C4.yaml")  # Let training initialize from model zoo
    cfg.SOLVER.IMS_PER_BATCH = 2
    cfg.SOLVER.BASE_LR = 0.00025  # pick a good LR
    cfg.SOLVER.MAX_ITER = 1    # 600 iterations seems good enough for this toy dataset; you will need to train longer for a practical dataset
    cfg.SOLVER.STEPS = []        # do not decay learning rate
    cfg.MODEL.ROI_HEADS.BATCH_SIZE_PER_IMAGE = 256   # faster, and good enough for this toy dataset (default: 512)
    cfg.MODEL.ROI_HEADS.NUM_CLASSES = 1  # only has one class (ballon). (see https://detectron2.readthedocs.io/tutorials/datasets.html#update-the-config-for-new-datasets)
    # NOTE: this config means the number of classes, but a few popular unofficial tutorials incorrect uses num_classes+1 here.
    os.makedirs(cfg.OUTPUT_DIR, exist_ok=True)
    return cfg, runner

cfg, runner = prepare_for_launch()
model = runner.build_model(cfg)
runner.do_train(cfg, model, resume=False)


# disable all the warnings
previous_level = logging.root.manager.disable
logging.disable(logging.INFO)

patch_d2_meta_arch()

cfg_name = 'faster_rcnn_fbnetv3a_dsmask_C4.yaml'
pytorch_model = model_zoo.get(cfg_name, trained=True)
pytorch_model.cpu()

with create_fake_detection_data_loader(224, 320, is_train=False) as data_loader:
    predictor_path = convert_and_export_predictor(
            model_zoo.get_config(cfg_name),
            copy.deepcopy(pytorch_model),
            "torchscript_int8@tracing",
            './',
            data_loader,
        )

# recover the logging level
logging.disable(previous_level)


model = create_predictor(predictor_path)

im = cv2.imread("/home/liupeng/sdb1/PythonProject/scaleDetection/detectron2go/datasets/humandetect/Images/train/20.png")

predictor = DemoPredictor(model)
outputs = predictor(im)

v = Visualizer(im[:, :, ::-1], MetadataCatalog.get("my_dataset_train"))
out = v.draw_instance_predictions(outputs["instances"].to("cpu"))
plt.imshow(out.get_image()[:, :, ::-1]

this are all pacakge i used:

Package Version

---

absl-py 1.0.0 aiohttp 3.8.1 aiosignal 1.2.0 antlr4-python3-runtime 4.8 appdirs 1.4.4 async-timeout 4.0.1 attrs 21.2.0 black 21.4b2 cachetools 4.2.4 certifi 2021.10.8 charset-normalizer 2.0.8 click 8.0.3 cloudpickle 2.0.0 cycler 0.11.0 Cython 0.29.24 d2go 0.0.1b20211201 detectron2 0.6 fonttools 4.28.2 frozenlist 1.2.0 fsspec 2021.11.1 future 0.18.2 fvcore 0.1.5.post20211023 google-auth 2.3.3 google-auth-oauthlib 0.4.6 grpcio 1.42.0 hydra-core 1.1.1 idna 3.3 importlib 1.0.4 importlib-metadata 4.8.2 importlib-resources 5.4.0 iopath 0.1.9 kiwisolver 1.3.2 Markdown 3.3.6 matplotlib 3.5.0 mkl-fft 1.3.1 mkl-random 1.2.2 mkl-service 2.4.0 mobile-cv 0.1 mock 4.0.3 multidict 5.2.0 mypy-extensions 0.4.3 numpy 1.21.2 oauthlib 3.1.1 olefile 0.46 omegaconf 2.1.1 opencv-python 4.5.4.60 packaging 21.3 parameterized 0.8.1 pathspec 0.9.0 Pillow 8.4.0 pip 21.2.4 portalocker 2.3.2 protobuf 3.19.1 pyasn1 0.4.8 pyasn1-modules 0.2.8 pycocotools 2.0.3 pyDeprecate 0.3.1 pydot 1.4.2 pyparsing 3.0.6 python-dateutil 2.8.2 pytorch-lightning 1.5.4 PyYAML 6.0 regex 2021.11.10 requests 2.26.0 requests-oauthlib 1.3.0 rsa 4.8 setuptools 58.0.4 setuptools-scm 6.3.2 six 1.16.0 tabulate 0.8.9 tensorboard 2.7.0 tensorboard-data-server 0.6.1 tensorboard-plugin-wit 1.8.0 termcolor 1.1.0 toml 0.10.2 tomli 1.2.2 torch 1.11.0.dev20211201+cu113 torchaudio 0.11.0.dev20211201+cu113 torchmetrics 0.6.0 torchvision 0.12.0.dev20211201+cu113 tqdm 4.62.3 typing-extensions 3.10.0.2 urllib3 1.26.7 Werkzeug 2.0.2 wheel 0.37.0 yacs 0.1.8 yarl 1.7.2 zipp 3.6.0

this is error log:

creating index… index created! detectron2.layers.Linear is in expected type (torch.nn.Linear),consider removing this code mock_quantization_type. /home/liupeng/anaconda3/envs/scaleDetection/lib/python3.8/site-packages/torch/ao/quantization/observer.py:177: UserWarning: Please use quant_min and quant_max to specify the range for observers. reduce_range will be deprecated in a future release of PyTorch. warnings.warn( /home/liupeng/anaconda3/envs/scaleDetection/lib/python3.8/site-packages/torch/ao/quantization/observer.py:901: UserWarning: floordiv is deprecated, and its behavior will change in a future version of pytorch. It currently rounds toward 0 (like the ‘trunc’ function NOT ‘floor’). This results in incorrect rounding for negative values. To keep the current behavior, use torch.div(a, b, rounding_mode=‘trunc’), or for actual floor division, use torch.div(a, b, rounding_mode=‘floor’). src_bin_begin // dst_bin_width, 0, self.dst_nbins - 1 /home/liupeng/anaconda3/envs/scaleDetection/lib/python3.8/site-packages/torch/ao/quantization/observer.py:906: UserWarning: floordiv is deprecated, and its behavior will change in a future version of pytorch. It currently rounds toward 0 (like the ‘trunc’ function NOT ‘floor’). This results in incorrect rounding for negative values. To keep the current behavior, use torch.div(a, b, rounding_mode=‘trunc’), or for actual floor division, use torch.div(a, b, rounding_mode=‘floor’). src_bin_end // dst_bin_width, 0, self.dst_nbins - 1 Traceback (most recent call last): File “train.py”, line 114, in <module> predictor_path = convert_and_export_predictor( File “/home/liupeng/anaconda3/envs/scaleDetection/lib/python3.8/site-packages/d2go/export/api.py”, line 151, in convert_and_export_predictor return export_predictor(cfg, pytorch_model, predictor_type, output_dir, data_loader) File “/home/liupeng/anaconda3/envs/scaleDetection/lib/python3.8/site-packages/d2go/export/api.py”, line 174, in export_predictor return default_export_predictor( File “/home/liupeng/anaconda3/envs/scaleDetection/lib/python3.8/site-packages/d2go/export/api.py”, line 222, in default_export_predictor export_config = pytorch_model.prepare_for_export(cfg, inputs, predictor_type) File “/home/liupeng/anaconda3/envs/scaleDetection/lib/python3.8/site-packages/d2go/modeling/meta_arch/rcnn.py”, line 45, in prepare_for_export return func(self, cfg, *args, **kwargs) File “/home/liupeng/anaconda3/envs/scaleDetection/lib/python3.8/site-packages/d2go/modeling/meta_arch/rcnn.py”, line 60, in default_rcnn_prepare_for_export from detectron2.export.caffe2_modeling import META_ARCH_CAFFE2_EXPORT_TYPE_MAP File “/home/liupeng/anaconda3/envs/scaleDetection/lib/python3.8/site-packages/detectron2/export/caffe2_modeling.py”, line 14, in <module> from .c10 import Caffe2Compatible File “/home/liupeng/anaconda3/envs/scaleDetection/lib/python3.8/site-packages/detectron2/export/c10.py”, line 14, in <module> from .shared import alias, to_device File “/home/liupeng/anaconda3/envs/scaleDetection/lib/python3.8/site-packages/detectron2/export/shared.py”, line 12, in <module> import caffe2.python.utils as putils File “/home/liupeng/anaconda3/envs/scaleDetection/lib/python3.8/site-packages/caffe2/python/init.py”, line 2, in <module> from caffe2.proto import caffe2_pb2 File “/home/liupeng/anaconda3/envs/scaleDetection/lib/python3.8/site-packages/caffe2/proto/init.py”, line 11, in <module> from caffe2.proto import caffe2_pb2, metanet_pb2, torch_pb2 ImportError: cannot import name ‘metanet_pb2’ from partially initialized module ‘caffe2.proto’ (most likely due to a circular import) (/home/liupeng/anaconda3/envs/scaleDetection/lib/python3.8/site-packages/caffe2/proto/init.py)

Expected behavior: Is the package version mismatched?

[wat3rBro]

Hi the issue should be gone after https://github.com/facebookresearch/d2go/pull/154

[LP308210365]

No，i didn’t

发自我的iPad

在 2021年12月29日，上午8:54，Daniel @.***> 写道：

@AaronReidCI @LP308210365 Hi! Could I know do you solve this problem ?

#137 (comment)

— Reply to this email directly, view it on GitHub, or unsubscribe. You are receiving this because you were mentioned.