fix: fix lr

ppcls/arch/backbone/__init__.py

@@ -60,10 +60,10 @@ from ppcls.arch.backbone.model_zoo.rednet import RedNet26, RedNet38, RedNet50, R
 from ppcls.arch.backbone.model_zoo.tnt import TNT_small
 from ppcls.arch.backbone.model_zoo.hardnet import HarDNet68, HarDNet85, HarDNet39_ds, HarDNet68_ds
 from ppcls.arch.backbone.model_zoo.cspnet import CSPDarkNet53
+from ppcls.arch.backbone.model_zoo.pvt_v2 import PVT_V2_B0, PVT_V2_B1, PVT_V2_B2_Linear, PVT_V2_B2, PVT_V2_B3, PVT_V2_B4, PVT_V2_B5
 from ppcls.arch.backbone.variant_models.resnet_variant import ResNet50_last_stage_stride1
 from ppcls.arch.backbone.variant_models.vgg_variant import VGG19Sigmoid
 from ppcls.arch.backbone.variant_models.pp_lcnet_variant import PPLCNet_x2_5_Tanh
-from ppcls.arch.backbone.model_zoo.pvt_v2 import PVT_V2_B0, PVT_V2_B1, PVT_V2_B2_Linear, PVT_V2_B2, PVT_V2_B3, PVT_V2_B4, PVT_V2_B5
 
 
 # help whl get all the models' api (class type) and components' api (func type)

ppcls/arch/backbone/model_zoo/pvt_v2.py

@@ -95,7 +95,7 @@ class Attention(nn.Layer):
                  sr_ratio=1,
                  linear=False):
         super().__init__()
-        assert dim % num_heads == 0, f"dim {dim} should be divided by num_heads {num_heads}."
+        assert dim % num_heads == 0
 
         self.dim = dim
         self.num_heads = num_heads
@@ -195,7 +195,6 @@ class Block(nn.Layer):
             proj_drop=drop,
             sr_ratio=sr_ratio,
             linear=linear)
-        # NOTE: drop path for stochastic depth, we shall see if this is better than dropout here
         self.drop_path = DropPath(drop_path) if drop_path > 0. else Identity()
         self.norm2 = norm_layer(dim)
         mlp_hidden_dim = int(dim * mlp_ratio)

ppcls/configs/ImageNet/PVTV2/PVT_V2_B0.yaml

@@ -44,7 +44,7 @@ Optimizer:
   one_dim_param_no_weight_decay: True
   lr:
     name: Cosine
-    learning_rate: 1e-3
+    learning_rate: 5e-4
     eta_min: 1e-5
     warmup_epoch: 20
     warmup_start_lr: 1e-6

ppcls/configs/ImageNet/PVTV2/PVT_V2_B1.yaml

@@ -44,7 +44,7 @@ Optimizer:
   one_dim_param_no_weight_decay: True
   lr:
     name: Cosine
-    learning_rate: 1e-3
+    learning_rate: 5e-4
     eta_min: 1e-5
     warmup_epoch: 20
     warmup_start_lr: 1e-6

ppcls/configs/ImageNet/PVTV2/PVT_V2_B2.yaml

@@ -44,7 +44,7 @@ Optimizer:
   one_dim_param_no_weight_decay: True
   lr:
     name: Cosine
-    learning_rate: 1e-3
+    learning_rate: 5e-4
     eta_min: 1e-5
     warmup_epoch: 20
     warmup_start_lr: 1e-6

ppcls/configs/ImageNet/PVTV2/PVT_V2_B2_Linear.yaml

@@ -44,7 +44,7 @@ Optimizer:
   one_dim_param_no_weight_decay: True
   lr:
     name: Cosine
-    learning_rate: 1e-3
+    learning_rate: 5e-4
     eta_min: 1e-5
     warmup_epoch: 20
     warmup_start_lr: 1e-6

ppcls/configs/ImageNet/PVTV2/PVT_V2_B3.yaml

@@ -45,7 +45,7 @@ Optimizer:
   one_dim_param_no_weight_decay: True
   lr:
     name: Cosine
-    learning_rate: 1e-3
+    learning_rate: 5e-4
     eta_min: 1e-5
     warmup_epoch: 20
     warmup_start_lr: 1e-6

ppcls/configs/ImageNet/PVTV2/PVT_V2_B4.yaml

@@ -45,7 +45,7 @@ Optimizer:
   one_dim_param_no_weight_decay: True
   lr:
     name: Cosine
-    learning_rate: 1e-3
+    learning_rate: 5e-4
     eta_min: 1e-5
     warmup_epoch: 20
     warmup_start_lr: 1e-6

ppcls/configs/ImageNet/PVTV2/PVT_V2_B5.yaml

@@ -45,7 +45,7 @@ Optimizer:
   one_dim_param_no_weight_decay: True
   lr:
     name: Cosine
-    learning_rate: 1e-3
+    learning_rate: 5e-4
     eta_min: 1e-5
     warmup_epoch: 20
     warmup_start_lr: 1e-6

test_tipc/config/PVTV2/PVT_V2_B0.txt

@@ -37,7 +37,7 @@ pretrained_model_url:https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/P
 infer_model:../inference/
 infer_export:True
 infer_quant:Fasle
-inference:python/predict_cls.py -c configs/inference_cls.yaml -o PreProcess.transform_ops.0.ResizeImage.resize_short=384 -o PreProcess.transform_ops.1.CropImage.size=384
+inference:python/predict_cls.py -c configs/inference_cls.yaml -o PreProcess.transform_ops.0.ResizeImage.resize_short=256 -o PreProcess.transform_ops.1.CropImage.size=224
 -o Global.use_gpu:True|False
 -o Global.enable_mkldnn:True|False
 -o Global.cpu_num_threads:1|6

test_tipc/config/PVTV2/PVT_V2_B1.txt

@@ -37,7 +37,7 @@ pretrained_model_url:https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/P
 infer_model:../inference/
 infer_export:True
 infer_quant:Fasle
-inference:python/predict_cls.py -c configs/inference_cls.yaml -o PreProcess.transform_ops.0.ResizeImage.resize_short=384 -o PreProcess.transform_ops.1.CropImage.size=384
+inference:python/predict_cls.py -c configs/inference_cls.yaml -o PreProcess.transform_ops.0.ResizeImage.resize_short=256 -o PreProcess.transform_ops.1.CropImage.size=224
 -o Global.use_gpu:True|False
 -o Global.enable_mkldnn:True|False
 -o Global.cpu_num_threads:1|6

test_tipc/config/PVTV2/PVT_V2_B2.txt

@@ -37,7 +37,7 @@ pretrained_model_url:https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/P
 infer_model:../inference/
 infer_export:True
 infer_quant:Fasle
-inference:python/predict_cls.py -c configs/inference_cls.yaml -o PreProcess.transform_ops.0.ResizeImage.resize_short=384 -o PreProcess.transform_ops.1.CropImage.size=384
+inference:python/predict_cls.py -c configs/inference_cls.yaml -o PreProcess.transform_ops.0.ResizeImage.resize_short=256 -o PreProcess.transform_ops.1.CropImage.size=224
 -o Global.use_gpu:True|False
 -o Global.enable_mkldnn:True|False
 -o Global.cpu_num_threads:1|6

test_tipc/config/PVTV2/PVT_V2_B2_Linear.txt

@@ -37,7 +37,7 @@ pretrained_model_url:https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/P
 infer_model:../inference/
 infer_export:True
 infer_quant:Fasle
-inference:python/predict_cls.py -c configs/inference_cls.yaml -o PreProcess.transform_ops.0.ResizeImage.resize_short=384 -o PreProcess.transform_ops.1.CropImage.size=384
+inference:python/predict_cls.py -c configs/inference_cls.yaml -o PreProcess.transform_ops.0.ResizeImage.resize_short=256 -o PreProcess.transform_ops.1.CropImage.size=224
 -o Global.use_gpu:True|False
 -o Global.enable_mkldnn:True|False
 -o Global.cpu_num_threads:1|6

test_tipc/config/PVTV2/PVT_V2_B3.txt

@@ -37,7 +37,7 @@ pretrained_model_url:https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/P
 infer_model:../inference/
 infer_export:True
 infer_quant:Fasle
-inference:python/predict_cls.py -c configs/inference_cls.yaml -o PreProcess.transform_ops.0.ResizeImage.resize_short=384 -o PreProcess.transform_ops.1.CropImage.size=384
+inference:python/predict_cls.py -c configs/inference_cls.yaml -o PreProcess.transform_ops.0.ResizeImage.resize_short=256 -o PreProcess.transform_ops.1.CropImage.size=224
 -o Global.use_gpu:True|False
 -o Global.enable_mkldnn:True|False
 -o Global.cpu_num_threads:1|6

test_tipc/config/PVTV2/PVT_V2_B4.txt

@@ -37,7 +37,7 @@ pretrained_model_url:https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/P
 infer_model:../inference/
 infer_export:True
 infer_quant:Fasle
-inference:python/predict_cls.py -c configs/inference_cls.yaml -o PreProcess.transform_ops.0.ResizeImage.resize_short=384 -o PreProcess.transform_ops.1.CropImage.size=384
+inference:python/predict_cls.py -c configs/inference_cls.yaml -o PreProcess.transform_ops.0.ResizeImage.resize_short=256 -o PreProcess.transform_ops.1.CropImage.size=224
 -o Global.use_gpu:True|False
 -o Global.enable_mkldnn:True|False
 -o Global.cpu_num_threads:1|6

test_tipc/config/PVTV2/PVT_V2_B5.txt

@@ -37,7 +37,7 @@ pretrained_model_url:https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/P
 infer_model:../inference/
 infer_export:True
 infer_quant:Fasle
-inference:python/predict_cls.py -c configs/inference_cls.yaml -o PreProcess.transform_ops.0.ResizeImage.resize_short=384 -o PreProcess.transform_ops.1.CropImage.size=384
+inference:python/predict_cls.py -c configs/inference_cls.yaml -o PreProcess.transform_ops.0.ResizeImage.resize_short=256 -o PreProcess.transform_ops.1.CropImage.size=224
 -o Global.use_gpu:True|False
 -o Global.enable_mkldnn:True|False
 -o Global.cpu_num_threads:1|6