size mismatch for stem.conv.weight: copying a param of torch.Size([64, 3, 7, 7]) from checkpoint, where the shape is torch.Size([16, 3, 3, 3]) in current model. size mismatch for stem.bn.weight: copying a param of torch.Size([64]) from checkpoint, where the shape is torch.Size([16]) in current model. about pycls

[rajprateek]

Hi chenbolinstudent, thanks for reporting this.
Can you please provide the command you used to launch pycls when you ran into this error? It will help us reproduce it and double check what the issue is.

I train with my dataset
use python tools/train_net.py
--cfg configs/archive/imagenet/resnet/R-50-1x64d_step_1gpu.yaml
OUT_DIR /tmp

train
a
1.png
2.png
b
3.png
4.png
c
5.png
6.png

val
a
1.png
2.png
b
3.png
4.png
c
5.png
6.png

the train picture is 2242243
the test picture is 2242243

MODEL:
TYPE: resnet
DEPTH: 50
NUM_CLASSES: 10
RESNET:
TRANS_FUN: bottleneck_transform
NUM_GROUPS: 1
WIDTH_PER_GROUP: 64
STRIDE_1X1: True
BN:
ZERO_INIT_FINAL_GAMMA: False
OPTIM:
BASE_LR: 0.0125
LR_POLICY: cos
STEPS: [0, 30, 60, 90]
LR_MULT: 0.1
MAX_EPOCH: 100
MOMENTUM: 0.9
NESTEROV: True
WEIGHT_DECAY: 0.0001
TRAIN:
DATASET: imagenet
SPLIT: train
BATCH_SIZE: 32
IM_SIZE: 224
EVAL_PERIOD: 1
TEST:
DATASET: imagenet
SPLIT: val
BATCH_SIZE: 25
IM_SIZE: 256
NUM_GPUS: 1
DATA_LOADER:
NUM_WORKERS: 4
CUDNN:
BENCHMARK: True
OUT_DIR: .

[ir413]

Hi @chenbolinstudent, there is likely a mismatch between the model you constructed and the model you are trying to load from checkpoint which results in an error. Please double-check that the constructed and the checkpoint models match.