GeneralRecognitionV2_PPLCNetV2_base.yaml微调特征提取模型报错

[chinesejunzai12]

库对应的版本如下:
paddleclas==2.5.2
paddlepaddle==2.5.2
paddlepaddle-gpt==2.5.2

运行命令报错如下:
python3.7 tools/train.py \

-c ./ppcls/configs/GeneralRecognitionV2/CUB_GeneralRecognitionV2_PPLCNetV2_base.yaml
grep: warning: GREP_OPTIONS is deprecated; please use an alias or script
[2024/06/26 17:17:35] ppcls INFO:
===========================================================
== PaddleClas is powered by PaddlePaddle ! ==
===========================================================
== ==
== For more info please go to the following website. ==
== ==
== https://github.com/PaddlePaddle/PaddleClas ==
===========================================================

[2024/06/26 17:17:35] ppcls INFO: AMP :
[2024/06/26 17:17:35] ppcls INFO: level : O1
[2024/06/26 17:17:35] ppcls INFO: scale_loss : 65536
[2024/06/26 17:17:35] ppcls INFO: use_dynamic_loss_scaling : True
[2024/06/26 17:17:35] ppcls INFO: ------------------------------------------------------------
[2024/06/26 17:17:35] ppcls INFO: Arch :
[2024/06/26 17:17:35] ppcls INFO: Backbone :
[2024/06/26 17:17:35] ppcls INFO: class_expand : 512
[2024/06/26 17:17:35] ppcls INFO: name : PPLCNetV2_base_ShiTu
[2024/06/26 17:17:35] ppcls INFO: pretrained : True
[2024/06/26 17:17:35] ppcls INFO: use_ssld : True
[2024/06/26 17:17:35] ppcls INFO: BackboneStopLayer :
[2024/06/26 17:17:35] ppcls INFO: name : flatten
[2024/06/26 17:17:35] ppcls INFO: Head :
[2024/06/26 17:17:35] ppcls INFO: bias_attr : False
[2024/06/26 17:17:35] ppcls INFO: class_num : 101
[2024/06/26 17:17:35] ppcls INFO: embedding_size : 512
[2024/06/26 17:17:35] ppcls INFO: name : FC
[2024/06/26 17:17:35] ppcls INFO: weight_attr :
[2024/06/26 17:17:35] ppcls INFO: initializer :
[2024/06/26 17:17:35] ppcls INFO: name : Normal
[2024/06/26 17:17:35] ppcls INFO: std : 0.001
[2024/06/26 17:17:35] ppcls INFO: Neck :
[2024/06/26 17:17:35] ppcls INFO: bias_attr :
[2024/06/26 17:17:35] ppcls INFO: initializer :
[2024/06/26 17:17:35] ppcls INFO: name : Constant
[2024/06/26 17:17:35] ppcls INFO: value : 0.0
[2024/06/26 17:17:35] ppcls INFO: learning_rate : 1e-20
[2024/06/26 17:17:35] ppcls INFO: name : BNNeck
[2024/06/26 17:17:35] ppcls INFO: num_features : 512
[2024/06/26 17:17:35] ppcls INFO: weight_attr :
[2024/06/26 17:17:35] ppcls INFO: initializer :
[2024/06/26 17:17:35] ppcls INFO: name : Constant
[2024/06/26 17:17:35] ppcls INFO: value : 1.0
[2024/06/26 17:17:35] ppcls INFO: infer_add_softmax : False
[2024/06/26 17:17:35] ppcls INFO: infer_output_key : features
[2024/06/26 17:17:35] ppcls INFO: name : RecModel
[2024/06/26 17:17:35] ppcls INFO: DataLoader :
[2024/06/26 17:17:35] ppcls INFO: Eval :
[2024/06/26 17:17:35] ppcls INFO: Gallery :
[2024/06/26 17:17:35] ppcls INFO: dataset :
[2024/06/26 17:17:35] ppcls INFO: cls_label_path : ./dataset/CUB_200_2011/test_list.txt
[2024/06/26 17:17:35] ppcls INFO: image_root : ./dataset/CUB_200_2011/
[2024/06/26 17:17:35] ppcls INFO: name : VeriWild
[2024/06/26 17:17:35] ppcls INFO: transform_ops :
[2024/06/26 17:17:35] ppcls INFO: DecodeImage :
[2024/06/26 17:17:35] ppcls INFO: channel_first : False
[2024/06/26 17:17:35] ppcls INFO: to_rgb : True
[2024/06/26 17:17:35] ppcls INFO: ResizeImage :
[2024/06/26 17:17:35] ppcls INFO: backend : cv2
[2024/06/26 17:17:35] ppcls INFO: interpolation : bilinear
[2024/06/26 17:17:35] ppcls INFO: return_numpy : False
[2024/06/26 17:17:35] ppcls INFO: size : [224, 224]
[2024/06/26 17:17:35] ppcls INFO: NormalizeImage :
[2024/06/26 17:17:35] ppcls INFO: mean : [0.485, 0.456, 0.406]
[2024/06/26 17:17:35] ppcls INFO: order : hwc
[2024/06/26 17:17:35] ppcls INFO: scale : 1.0/255.0
[2024/06/26 17:17:35] ppcls INFO: std : [0.229, 0.224, 0.225]
[2024/06/26 17:17:35] ppcls INFO: loader :
[2024/06/26 17:17:35] ppcls INFO: num_workers : 4
[2024/06/26 17:17:35] ppcls INFO: use_shared_memory : True
[2024/06/26 17:17:35] ppcls INFO: sampler :
[2024/06/26 17:17:35] ppcls INFO: batch_size : 64
[2024/06/26 17:17:35] ppcls INFO: drop_last : False
[2024/06/26 17:17:35] ppcls INFO: name : DistributedBatchSampler
[2024/06/26 17:17:35] ppcls INFO: shuffle : False
[2024/06/26 17:17:35] ppcls INFO: Query :
[2024/06/26 17:17:35] ppcls INFO: dataset :
[2024/06/26 17:17:35] ppcls INFO: cls_label_path : ./dataset/CUB_200_2011/test_list.txt
[2024/06/26 17:17:35] ppcls INFO: image_root : ./dataset/CUB_200_2011/
[2024/06/26 17:17:35] ppcls INFO: name : VeriWild
[2024/06/26 17:17:35] ppcls INFO: transform_ops :
[2024/06/26 17:17:35] ppcls INFO: DecodeImage :
[2024/06/26 17:17:35] ppcls INFO: channel_first : False
[2024/06/26 17:17:35] ppcls INFO: to_rgb : True
[2024/06/26 17:17:35] ppcls INFO: ResizeImage :
[2024/06/26 17:17:35] ppcls INFO: backend : cv2
[2024/06/26 17:17:35] ppcls INFO: interpolation : bilinear
[2024/06/26 17:17:35] ppcls INFO: return_numpy : False
[2024/06/26 17:17:35] ppcls INFO: size : [224, 224]
[2024/06/26 17:17:35] ppcls INFO: NormalizeImage :
[2024/06/26 17:17:35] ppcls INFO: mean : [0.485, 0.456, 0.406]
[2024/06/26 17:17:35] ppcls INFO: order : hwc
[2024/06/26 17:17:35] ppcls INFO: scale : 1.0/255.0
[2024/06/26 17:17:35] ppcls INFO: std : [0.229, 0.224, 0.225]
[2024/06/26 17:17:35] ppcls INFO: loader :
[2024/06/26 17:17:35] ppcls INFO: num_workers : 4
[2024/06/26 17:17:35] ppcls INFO: use_shared_memory : True
[2024/06/26 17:17:35] ppcls INFO: sampler :
[2024/06/26 17:17:35] ppcls INFO: batch_size : 64
[2024/06/26 17:17:35] ppcls INFO: drop_last : False
[2024/06/26 17:17:35] ppcls INFO: name : DistributedBatchSampler
[2024/06/26 17:17:35] ppcls INFO: shuffle : False
[2024/06/26 17:17:35] ppcls INFO: Train :
[2024/06/26 17:17:35] ppcls INFO: dataset :
[2024/06/26 17:17:35] ppcls INFO: cls_label_path : ./dataset/CUB_200_2011/train_list.txt
[2024/06/26 17:17:35] ppcls INFO: image_root : ./dataset/CUB_200_2011/
[2024/06/26 17:17:35] ppcls INFO: name : VeriWild
[2024/06/26 17:17:35] ppcls INFO: relabel : True
[2024/06/26 17:17:35] ppcls INFO: transform_ops :
[2024/06/26 17:17:35] ppcls INFO: DecodeImage :
[2024/06/26 17:17:35] ppcls INFO: channel_first : False
[2024/06/26 17:17:35] ppcls INFO: to_rgb : True
[2024/06/26 17:17:35] ppcls INFO: ResizeImage :
[2024/06/26 17:17:35] ppcls INFO: backend : cv2
[2024/06/26 17:17:35] ppcls INFO: interpolation : bilinear
[2024/06/26 17:17:35] ppcls INFO: return_numpy : False
[2024/06/26 17:17:35] ppcls INFO: size : [224, 224]
[2024/06/26 17:17:35] ppcls INFO: RandFlipImage :
[2024/06/26 17:17:35] ppcls INFO: flip_code : 1
[2024/06/26 17:17:35] ppcls INFO: Pad :
[2024/06/26 17:17:35] ppcls INFO: backend : cv2
[2024/06/26 17:17:35] ppcls INFO: padding : 10
[2024/06/26 17:17:35] ppcls INFO: RandCropImageV2 :
[2024/06/26 17:17:35] ppcls INFO: size : [224, 224]
[2024/06/26 17:17:35] ppcls INFO: RandomRotation :
[2024/06/26 17:17:35] ppcls INFO: degrees : 90
[2024/06/26 17:17:35] ppcls INFO: interpolation : bilinear
[2024/06/26 17:17:35] ppcls INFO: prob : 0.5
[2024/06/26 17:17:35] ppcls INFO: ResizeImage :
[2024/06/26 17:17:35] ppcls INFO: backend : cv2
[2024/06/26 17:17:35] ppcls INFO: interpolation : bilinear
[2024/06/26 17:17:35] ppcls INFO: return_numpy : False
[2024/06/26 17:17:35] ppcls INFO: size : [224, 224]
[2024/06/26 17:17:35] ppcls INFO: NormalizeImage :
[2024/06/26 17:17:35] ppcls INFO: mean : [0.485, 0.456, 0.406]
[2024/06/26 17:17:35] ppcls INFO: order : hwc
[2024/06/26 17:17:35] ppcls INFO: scale : 1.0/255.0
[2024/06/26 17:17:35] ppcls INFO: std : [0.229, 0.224, 0.225]
[2024/06/26 17:17:35] ppcls INFO: loader :
[2024/06/26 17:17:35] ppcls INFO: num_workers : 4
[2024/06/26 17:17:35] ppcls INFO: use_shared_memory : True
[2024/06/26 17:17:35] ppcls INFO: sampler :
[2024/06/26 17:17:35] ppcls INFO: batch_size : 256
[2024/06/26 17:17:35] ppcls INFO: drop_last : False
[2024/06/26 17:17:35] ppcls INFO: id_list : [50030, 80700, 92019, 96015]
[2024/06/26 17:17:35] ppcls INFO: name : PKSampler
[2024/06/26 17:17:35] ppcls INFO: ratio : [4, 4]
[2024/06/26 17:17:35] ppcls INFO: sample_method : id_avg_prob
[2024/06/26 17:17:35] ppcls INFO: sample_per_id : 4
[2024/06/26 17:17:35] ppcls INFO: shuffle : True
[2024/06/26 17:17:35] ppcls INFO: Global :
[2024/06/26 17:17:35] ppcls INFO: checkpoints : None
[2024/06/26 17:17:35] ppcls INFO: device : gpu
[2024/06/26 17:17:35] ppcls INFO: epochs : 100
[2024/06/26 17:17:35] ppcls INFO: eval_during_train : True
[2024/06/26 17:17:35] ppcls INFO: eval_interval : 1
[2024/06/26 17:17:35] ppcls INFO: eval_mode : retrieval
[2024/06/26 17:17:35] ppcls INFO: image_shape : [3, 224, 224]
[2024/06/26 17:17:35] ppcls INFO: output_dir : ./output
[2024/06/26 17:17:35] ppcls INFO: pretrained_model : None
[2024/06/26 17:17:35] ppcls INFO: print_batch_step : 20
[2024/06/26 17:17:35] ppcls INFO: re_ranking : False
[2024/06/26 17:17:35] ppcls INFO: retrieval_feature_from : features
[2024/06/26 17:17:35] ppcls INFO: save_inference_dir : ./inference
[2024/06/26 17:17:35] ppcls INFO: save_interval : 10
[2024/06/26 17:17:35] ppcls INFO: use_dali : False
[2024/06/26 17:17:35] ppcls INFO: use_visualdl : False
[2024/06/26 17:17:35] ppcls INFO: Loss :
[2024/06/26 17:17:35] ppcls INFO: Eval :
[2024/06/26 17:17:35] ppcls INFO: CELoss :
[2024/06/26 17:17:35] ppcls INFO: weight : 1.0
[2024/06/26 17:17:35] ppcls INFO: Train :
[2024/06/26 17:17:35] ppcls INFO: CELoss :
[2024/06/26 17:17:35] ppcls INFO: epsilon : 0.1
[2024/06/26 17:17:35] ppcls INFO: weight : 1.0
[2024/06/26 17:17:35] ppcls INFO: TripletAngularMarginLoss :
[2024/06/26 17:17:35] ppcls INFO: absolute_loss_weight : 0.1
[2024/06/26 17:17:35] ppcls INFO: add_absolute : True
[2024/06/26 17:17:35] ppcls INFO: an_value : 0.4
[2024/06/26 17:17:35] ppcls INFO: ap_value : 0.8
[2024/06/26 17:17:35] ppcls INFO: feature_from : features
[2024/06/26 17:17:35] ppcls INFO: margin : 0.5
[2024/06/26 17:17:35] ppcls INFO: normalize_feature : True
[2024/06/26 17:17:35] ppcls INFO: reduction : mean
[2024/06/26 17:17:35] ppcls INFO: weight : 1.0
[2024/06/26 17:17:35] ppcls INFO: Metric :
[2024/06/26 17:17:35] ppcls INFO: Eval :
[2024/06/26 17:17:35] ppcls INFO: Recallk :
[2024/06/26 17:17:35] ppcls INFO: topk : [1, 5]
[2024/06/26 17:17:35] ppcls INFO: mAP :
[2024/06/26 17:17:35] ppcls INFO: Optimizer :
[2024/06/26 17:17:35] ppcls INFO: lr :
[2024/06/26 17:17:35] ppcls INFO: learning_rate : 0.06
[2024/06/26 17:17:35] ppcls INFO: name : Cosine
[2024/06/26 17:17:35] ppcls INFO: warmup_epoch : 5
[2024/06/26 17:17:35] ppcls INFO: momentum : 0.9
[2024/06/26 17:17:35] ppcls INFO: name : Momentum
[2024/06/26 17:17:35] ppcls INFO: regularizer :
[2024/06/26 17:17:35] ppcls INFO: coeff : 1e-05
[2024/06/26 17:17:35] ppcls INFO: name : L2
[2024/06/26 17:17:35] ppcls INFO: profiler_options : None
[2024/06/26 17:17:35] ppcls INFO: train with paddle 2.5.2 and device Place(gpu:0)
Traceback (most recent call last):
File "tools/train.py", line 32, in
engine = Engine(config, mode="train")
File "/mnt/ppcls/engine/engine.py", line 121, in init
self.config["DataLoader"], "Train", self.device, self.use_dali)
File "/mnt/ppcls/data/init.py", line 91, in build_dataloader
dataset = eval(dataset_name)(**config_dataset)
TypeError: init() got an unexpected keyword argument 'relabel'

[cuicheng01]

请问您有改过默认的config吗

[cuicheng01]

问题我们没有复现，您可以尝试用干净的代码再试试呢？

[chinesejunzai12]

这个是在2.3版本出现的, 升级到2.5后没有了, 因为当时看文档的时候指定的是2.3版本