'roi_heads.box_predictor.cls_score.weight' has shape (81, 1024) in the checkpoint but (157, 1024) in the model! Skipped.

Problem:

I have some difficulties understanding this output. I have almost the exact same setup working with another data set - the only difference in setup is that the working data set has 1 category while this data set has 156.

Model setup:

cfg = get_cfg()
cfg.merge_from_file("/detectron2_repo/configs/COCO-Detection/faster_rcnn_R_50_FPN_3x.yaml")
cfg.OUTPUT_DIR = '/mnt/s3bucket/All_Products_with_Category_Clean_Output'
cfg.DATASETS.TRAIN = ("train_labels",)
cfg.DATASETS.TEST = ()  # no metrics implemented for this dataset
cfg.DATALOADER.NUM_WORKERS = 0
cfg.MODEL.WEIGHTS = "detectron2://COCO-Detection/faster_rcnn_R_50_FPN_3x/137849458/model_final_280758.pkl"
cfg.SOLVER.IMS_PER_BATCH = 2
cfg.SOLVER.BASE_LR = 0.02
cfg.SOLVER.MAX_ITER = (200000)
cfg.MODEL.ROI_HEADS.BATCH_SIZE_PER_IMAGE = (128)
cfg.MODEL.ROI_HEADS.NUM_CLASSES = 156

Output:

[10/29 23:26:51 d2.data.detection_utils]: TransformGens used in training: [ResizeShortestEdge(short_edge_length=(640, 672, 704, 736, 768, 800), max_size=1333, sample_style='choice'), RandomFlip()]
[10/29 23:26:51 d2.data.build]: Using training sampler TrainingSampler
'roi_heads.box_predictor.cls_score.weight' has shape (81, 1024) in the checkpoint but (157, 1024) in the model! Skipped.
'roi_heads.box_predictor.cls_score.bias' has shape (81,) in the checkpoint but (157,) in the model! Skipped.
'roi_heads.box_predictor.bbox_pred.weight' has shape (320, 1024) in the checkpoint but (624, 1024) in the model! Skipped.
'roi_heads.box_predictor.bbox_pred.bias' has shape (320,) in the checkpoint but (624,) in the model! Skipped.
[10/29 23:26:53 d2.engine.train_loop]: Starting training from iteration 0

Environment:

Output of python3 -m detectron2.utils.collect_env:

------------------------  --------------------------------------------------
sys.platform              linux
Python                    3.7.4 (default, Aug 13 2019, 20:35:49) [GCC 7.3.0]
Numpy                     1.17.3
Detectron2 Compiler       GCC 7.4
Detectron2 CUDA Compiler  10.1
DETECTRON2_ENV_MODULE     <not set>
PyTorch                   1.3.0
PyTorch Debug Build       False
torchvision               0.4.1
CUDA available            True
GPU 0                     Tesla K80
CUDA_HOME                 /usr/local/cuda
NVCC                      Cuda compilation tools, release 10.1, V10.1.243
Pillow                    6.2.1
cv2                       4.1.1
------------------------  --------------------------------------------------
PyTorch built with:
  - GCC 7.3
  - Intel(R) Math Kernel Library Version 2019.0.4 Product Build 20190411 for Intel(R) 64 architecture applications
  - Intel(R) MKL-DNN v0.20.5 (Git Hash 0125f28c61c1f822fd48570b4c1066f96fcb9b2e)
  - OpenMP 201511 (a.k.a. OpenMP 4.5)
  - NNPACK is enabled
  - CUDA Runtime 10.1
  - NVCC architecture flags: -gencode;arch=compute_35,code=sm_35;-gencode;arch=compute_50,code=sm_50;-gencode;arch=compute_60,code=sm_60;-gencode;arch=compute_61,code=sm_61;-gencode;arch=compute_70,code=sm_70;-gencode;arch=compute_75,code=sm_75;-gencode;arch=compute_50,code=compute_50
  - CuDNN 7.6.3
  - Magma 2.5.1
  - Build settings: BLAS=MKL, BUILD_NAMEDTENSOR=OFF, BUILD_TYPE=Release, CXX_FLAGS= -Wno-deprecated -fvisibility-inlines-hidden -fopenmp -DUSE_FBGEMM -DUSE_QNNPACK -DUSE_PYTORCH_QNNPACK -O2 -fPIC -Wno-narrowing -Wall -Wextra -Wno-missing-field-initializers -Wno-type-limits -Wno-array-bounds -Wno-unknown-pragmas -Wno-sign-compare -Wno-unused-parameter -Wno-unused-variable -Wno-unused-function -Wno-unused-result -Wno-strict-overflow -Wno-strict-aliasing -Wno-error=deprecated-declarations -Wno-stringop-overflow -Wno-error=pedantic -Wno-error=redundant-decls -Wno-error=old-style-cast -fdiagnostics-color=always -faligned-new -Wno-unused-but-set-variable -Wno-maybe-uninitialized -fno-math-errno -fno-trapping-math -Wno-stringop-overflow, DISABLE_NUMA=1, PERF_WITH_AVX=1, PERF_WITH_AVX2=1, PERF_WITH_AVX512=1, USE_CUDA=True, USE_EXCEPTION_PTR=1, USE_GFLAGS=OFF, USE_GLOG=OFF, USE_MKL=ON, USE_MKLDNN=ON, USE_MPI=OFF, USE_NCCL=ON, USE_NNPACK=ON, USE_OPENMP=ON, USE_STATIC_DISPATCH=OFF, 

[ppwwyyxx]

The warning message is expected. Because your dataset has different number of classes from the pre-trained model, a few pre-trained weights have to be skipped.

[arniwesth]

@diego1109 The warning messages basically just tells you that the model setup is different from the pre-trained model (e.g. different number of classes). This does not affect the training - it’s expected behaviour.