testing question

[wyl12]

/home/lsw/miniconda3/envs/SOTS/bin/python /home/lsw/SOT/SOTS/tracking/test_sot.py
model backbone: ResNet50Dilated
model neck: ShrinkChannelS3S4
model head: Learn2Match
model build done!
SiamInference(
(backbone): ResNet50Dilated(
(conv1): Conv2d(3, 64, kernel_size=(7, 7), stride=(2, 2), bias=False)
(bn1): BatchNorm2d(64, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
(relu): ReLU(inplace=True)
(maxpool): MaxPool2d(kernel_size=3, stride=2, padding=1, dilation=1, ceil_mode=False)
(layer1): Sequential(
(0): Bottleneck(
(conv1): Conv2d(64, 64, kernel_size=(1, 1), stride=(1, 1), bias=False)
(bn1): BatchNorm2d(64, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
(conv2): Conv2d(64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False)
(bn2): BatchNorm2d(64, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
(conv3): Conv2d(64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False)
(bn3): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
(relu): ReLU(inplace=True)
(downsample): Sequential(
(0): Conv2d(64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False)
(1): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
)
)
(1): Bottleneck(
(conv1): Conv2d(256, 64, kernel_size=(1, 1), stride=(1, 1), bias=False)
(bn1): BatchNorm2d(64, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
(conv2): Conv2d(64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False)
(bn2): BatchNorm2d(64, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
(conv3): Conv2d(64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False)
(bn3): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
(relu): ReLU(inplace=True)
)
(2): Bottleneck(
(conv1): Conv2d(256, 64, kernel_size=(1, 1), stride=(1, 1), bias=False)
(bn1): BatchNorm2d(64, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
(conv2): Conv2d(64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False)
(bn2): BatchNorm2d(64, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
(conv3): Conv2d(64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False)
(bn3): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
(relu): ReLU(inplace=True)
)
)
(layer2): Sequential(
(0): Bottleneck(
(conv1): Conv2d(256, 128, kernel_size=(1, 1), stride=(1, 1), bias=False)
(bn1): BatchNorm2d(128, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
(conv2): Conv2d(128, 128, kernel_size=(3, 3), stride=(2, 2), bias=False)
(bn2): BatchNorm2d(128, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
(conv3): Conv2d(128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False)
(bn3): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
(relu): ReLU(inplace=True)
(downsample): Sequential(
(0): Conv2d(256, 512, kernel_size=(3, 3), stride=(2, 2), bias=False)
(1): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
)
)
(1): Bottleneck(
(conv1): Conv2d(512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False)
(bn1): BatchNorm2d(128, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
(conv2): Conv2d(128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False)
(bn2): BatchNorm2d(128, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
(conv3): Conv2d(128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False)
(bn3): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
(relu): ReLU(inplace=True)
)
(2): Bottleneck(
(conv1): Conv2d(512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False)
(bn1): BatchNorm2d(128, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
(conv2): Conv2d(128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False)
(bn2): BatchNorm2d(128, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
(conv3): Conv2d(128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False)
(bn3): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
(relu): ReLU(inplace=True)
)
(3): Bottleneck(
(conv1): Conv2d(512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False)
(bn1): BatchNorm2d(128, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
(conv2): Conv2d(128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False)
(bn2): BatchNorm2d(128, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
(conv3): Conv2d(128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False)
(bn3): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
(relu): ReLU(inplace=True)
)
)
(layer3): Sequential(
(0): Bottleneck(
(conv1): Conv2d(512, 256, kernel_size=(1, 1), stride=(1, 1), bias=False)
(bn1): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
(conv2): Conv2d(256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False)
(bn2): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
(conv3): Conv2d(256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False)
(bn3): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
(relu): ReLU(inplace=True)
(downsample): Sequential(
(0): Conv2d(512, 1024, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False)
(1): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
)
)
(1): Bottleneck(
(conv1): Conv2d(1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False)
(bn1): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
(conv2): Conv2d(256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False)
(bn2): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
(conv3): Conv2d(256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False)
(bn3): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
(relu): ReLU(inplace=True)
)
(2): Bottleneck(
(conv1): Conv2d(1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False)
(bn1): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
(conv2): Conv2d(256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False)
(bn2): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
(conv3): Conv2d(256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False)
(bn3): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
(relu): ReLU(inplace=True)
)
(3): Bottleneck(
(conv1): Conv2d(1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False)
(bn1): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
(conv2): Conv2d(256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False)
(bn2): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
(conv3): Conv2d(256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False)
(bn3): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
(relu): ReLU(inplace=True)
)
(4): Bottleneck(
(conv1): Conv2d(1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False)
(bn1): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
(conv2): Conv2d(256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False)
(bn2): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
(conv3): Conv2d(256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False)
(bn3): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
(relu): ReLU(inplace=True)
)
(5): Bottleneck(
(conv1): Conv2d(1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False)
(bn1): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
(conv2): Conv2d(256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False)
(bn2): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
(conv3): Conv2d(256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False)
(bn3): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
(relu): ReLU(inplace=True)
)
)
)
(neck): ShrinkChannelS3S4(
(downsample): Sequential(
(0): Conv2d(1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False)
(1): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
(2): ReLU(inplace=True)
)
(downsample_s3): Sequential(
(0): Conv2d(512, 256, kernel_size=(1, 1), stride=(1, 1), bias=False)
(1): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
(2): ReLU(inplace=True)
)
)
(head): Learn2Match(
(regression): L2Mregression(
(reg_encode): SimpleMatrix(
(matrix11_k): Sequential(
(0): Conv2d(256, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False)
)
(matrix11_s): Sequential(
(0): Conv2d(256, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False)
)
)
(roi_te): roi_template(
(fea_encoder): Sequential(
(0): Conv2d(256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
(1): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
(2): LeakyReLU(negative_slope=0.1)
(3): Conv2d(256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
(4): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
)
(fea_encoder_s3): Sequential(
(0): Conv2d(256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
(1): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
(2): LeakyReLU(negative_slope=0.1)
(3): Conv2d(256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
(4): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
)
(spatial_conv): Sequential(
(0): Conv2d(256, 256, kernel_size=(3, 3), stride=(1, 1))
(1): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
(2): LeakyReLU(negative_slope=0.1)
)
(spatial_conv_s3): Sequential(
(0): Conv2d(256, 256, kernel_size=(3, 3), stride=(1, 1))
(1): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
(2): LeakyReLU(negative_slope=0.1)
)
)
(LTM): LTM(
(FiLM): FiLM(
(s_embed): Conv2d(256, 256, kernel_size=(1, 1), stride=(1, 1))
(conv_g): Sequential(
(0): Conv2d(256, 256, kernel_size=(1, 1), stride=(1, 1))
(1): LeakyReLU(negative_slope=0.1)
)
(conv_b): Sequential(
(0): Conv2d(256, 256, kernel_size=(1, 1), stride=(1, 1))
(1): LeakyReLU(negative_slope=0.1)
)
)
(PC): PairRelation(
(s_embed): Conv2d(256, 64, kernel_size=(1, 1), stride=(1, 1))
(t_embed): Conv2d(256, 64, kernel_size=(1, 1), stride=(1, 1))
(down): Conv2d(225, 256, kernel_size=(1, 1), stride=(1, 1))
)
(embed2): Sequential(
(0): Conv2d(512, 256, kernel_size=(1, 1), stride=(1, 1))
(1): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
(2): LeakyReLU(negative_slope=0.1)
)
)
(bbox_tower): Sequential(
(0): Conv2d(256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
(1): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
(2): ReLU()
(3): Conv2d(256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
(4): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
(5): ReLU()
(6): Conv2d(256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
(7): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
(8): ReLU()
)
(bbox_pred): Conv2d(256, 4, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
)
(classification): L2Mclassification(
(LTM): LTM(
(Transformer): SimpleSelfAtt(
(s_embed): Conv2d(256, 256, kernel_size=(1, 1), stride=(1, 1))
(t_embed_v): Conv2d(256, 256, kernel_size=(1, 1), stride=(1, 1))
(t_embed): Conv2d(256, 256, kernel_size=(1, 1), stride=(1, 1))
(trans): MultiheadAttention(
(out_proj): NonDynamicallyQuantizableLinear(in_features=256, out_features=256, bias=True)
)
)
(FiLM): FiLM(
(s_embed): Conv2d(256, 256, kernel_size=(1, 1), stride=(1, 1))
(conv_g): Sequential(
(0): Conv2d(256, 256, kernel_size=(1, 1), stride=(1, 1))
(1): LeakyReLU(negative_slope=0.1)
)
(conv_b): Sequential(
(0): Conv2d(256, 256, kernel_size=(1, 1), stride=(1, 1))
(1): LeakyReLU(negative_slope=0.1)
)
)
(embed2): Sequential(
(0): Conv2d(512, 256, kernel_size=(1, 1), stride=(1, 1))
(1): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
(2): LeakyReLU(negative_slope=0.1)
)
)
(roi_cls): roi_classification(
(fea_encoder): Sequential(
(0): Conv2d(256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
(1): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
(2): LeakyReLU(negative_slope=0.1)
(3): Conv2d(256, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
(4): BatchNorm2d(128, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
(5): LeakyReLU(negative_slope=0.1)
)
(fea_encoder_s3): Sequential(
(0): Conv2d(256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
(1): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
(2): LeakyReLU(negative_slope=0.1)
(3): Conv2d(256, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
(4): BatchNorm2d(128, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
(5): LeakyReLU(negative_slope=0.1)
)
(down_spatial_conv): Sequential(
(0): Conv2d(128, 128, kernel_size=(3, 3), stride=(1, 1))
(1): BatchNorm2d(128, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
(2): LeakyReLU(negative_slope=0.1)
)
(down_spatial_linear): Sequential(
(0): Linear(in_features=128, out_features=128, bias=True)
(1): LeakyReLU(negative_slope=0.1)
)
(down_spatial_conv_s3): Sequential(
(0): Conv2d(128, 128, kernel_size=(3, 3), stride=(1, 1))
(1): BatchNorm2d(128, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
(2): LeakyReLU(negative_slope=0.1)
)
(down_target_s3): Sequential(
(0): Conv2d(256, 128, kernel_size=(1, 1), stride=(1, 1))
(1): BatchNorm2d(128, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
(2): LeakyReLU(negative_slope=0.1)
)
(down_target_s4): Sequential(
(0): Conv2d(256, 128, kernel_size=(1, 1), stride=(1, 1))
(1): BatchNorm2d(128, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
(2): LeakyReLU(negative_slope=0.1)
)
(down_spatial_linear_s3): Sequential(
(0): Linear(in_features=128, out_features=128, bias=True)
(1): LeakyReLU(negative_slope=0.1)
)
(merge_s3s4_s2): Sequential(
(0): Linear(in_features=256, out_features=256, bias=True)
(1): LeakyReLU(negative_slope=0.1)
(2): Linear(in_features=256, out_features=256, bias=True)
(3): LeakyReLU(negative_slope=0.1)
)
(merge_s3s4_s1): Sequential(
(0): Conv2d(256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
(1): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
(2): LeakyReLU(negative_slope=0.1)
(3): Conv2d(256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
(4): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
(5): LeakyReLU(negative_slope=0.1)
)
(pred_s1): Conv2d(256, 1, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
(pred_s2): Linear(in_features=256, out_features=1, bias=True)
)
)
)
)
===> init Siamese <====
load pretrained model from ../snapshot/AutoMatch.pth
remove prefix 'module.'
change features.features to backbone
2022-12-02 22:47:17.534 | INFO | lib.utils.model_helper:check_keys:178 - missing keys:[]
Traceback (most recent call last):
File "/home/lsw/SOT/SOTS/tracking/test_sot.py", line 150, in
main()
File "/home/lsw/SOT/SOTS/tracking/test_sot.py", line 135, in main
siam_net = loader.load_pretrain(siam_net, resume, addhead=True, print_unuse=False)
File "/home/lsw/SOT/SOTS/tracking/../lib/utils/model_helper.py", line 209, in load_pretrain
model.load_state_dict(pretrained_dict, strict=True)
File "/home/lsw/miniconda3/envs/Unicorn/lib/python3.7/site-packages/torch/nn/modules/module.py", line 1483, in load_state_dict
self.class.name, "\n\t".join(error_msgs)))
RuntimeError: Error(s) in loading state_dict for SiamInference:
Unexpected key(s) in state_dict: "head.regression.LTM.GuidedSP.s_embed.weight", "head.regression.LTM.GuidedSP.s_embed.bias", "head.regression.LTM.GuidedSP.t_embed.weight", "head.regression.LTM.GuidedSP.t_embed.bias", "head.regression.LTM.PointDW.s_embed.weight", "head.regression.LTM.PointDW.s_embed.bias", "head.regression.LTM.PointDW.t_embed.weight", "head.regression.LTM.PointDW.t_embed.bias", "head.regression.LTM.PointAdd.s_embed.weight", "head.regression.LTM.PointAdd.s_embed.bias", "head.regression.LTM.PointAdd.t_embed.weight", "head.regression.LTM.PointAdd.t_embed.bias", "head.regression.LTM.Transformer.s_embed.weight", "head.regression.LTM.Transformer.s_embed.bias", "head.regression.LTM.Transformer.t_embed_v.weight", "head.regression.LTM.Transformer.t_embed_v.bias", "head.regression.LTM.Transformer.t_embed.weight", "head.regression.LTM.Transformer.t_embed.bias", "head.regression.LTM.Transformer.trans.in_proj_weight", "head.regression.LTM.Transformer.trans.in_proj_bias", "head.regression.LTM.Transformer.trans.out_proj.weight", "head.regression.LTM.Transformer.trans.out_proj.bias", "head.classification.LTM.GuidedSP.s_embed.weight", "head.classification.LTM.GuidedSP.s_embed.bias", "head.classification.LTM.GuidedSP.t_embed.weight", "head.classification.LTM.GuidedSP.t_embed.bias", "head.classification.LTM.PointDW.s_embed.weight", "head.classification.LTM.PointDW.s_embed.bias", "head.classification.LTM.PointDW.t_embed.weight", "head.classification.LTM.PointDW.t_embed.bias", "head.classification.LTM.PointAdd.s_embed.weight", "head.classification.LTM.PointAdd.s_embed.bias", "head.classification.LTM.PointAdd.t_embed.weight", "head.classification.LTM.PointAdd.t_embed.bias", "head.classification.LTM.PC.s_embed.weight", "head.classification.LTM.PC.s_embed.bias", "head.classification.LTM.PC.t_embed.weight", "head.classification.LTM.PC.t_embed.bias", "head.classification.LTM.PC.down.weight", "head.classification.LTM.PC.down.bias".

Process finished with exit code 1

Hello,i met this problem when i try to testing ,but i can not solve it.Could you help me to solve it if you have some time?

[faicaiwawa]

I also met this problem,it seem model weight is not match with code .have you solve this problem?

[YOLOCy]

I also meet this question. May I ask if you have resolved it? If so, could you please let me know the method

[Air1000thsummer]

same issue . seams wrong pretrain.pth