Merge branch 'develop' of https://github.com/PaddlePaddle/PaddleClas into whole_chain

deploy/slim/quant_post_static.py

@@ -41,11 +41,11 @@ def main():
                      'inference.pdmodel')) and os.path.exists(
                          os.path.join(config["Global"]["save_inference_dir"],
                                       'inference.pdiparams'))
-    config["DataLoader"]["Train"]["sampler"]["batch_size"] = 1
-    config["DataLoader"]["Train"]["loader"]["num_workers"] = 0
+    config["DataLoader"]["Eval"]["sampler"]["batch_size"] = 1
+    config["DataLoader"]["Eval"]["loader"]["num_workers"] = 0
     init_logger()
     device = paddle.set_device("cpu")
-    train_dataloader = build_dataloader(config["DataLoader"], "Train", device,
+    train_dataloader = build_dataloader(config["DataLoader"], "Eval", device,
                                         False)
 
     def sample_generator(loader):

ppcls/configs/Vehicle/PPLCNet_2.5x_ReID.yaml

+# global configs
+# pretrained_model: https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/rec/models/pretrain/vehicle_reid_PPLCNet2.5x_VERIWild_v1.0_pretrained.pdparams
+# VeriWild v1 small: recall1: 0.93736, recall5: 0.98427, mAP: 0.82125
+Global:
+  checkpoints: null
+  pretrained_model: null
+  output_dir: "./output_reid/"
+  device: "gpu"
+  save_interval: 1
+  eval_during_train: True
+  eval_interval: 1
+  epochs: 160
+  print_batch_step: 10
+  use_visualdl: False
+  # used for static mode and model export
+  image_shape: [3, 224, 224]
+  save_inference_dir: "./inference"
+  eval_mode: "retrieval"
+
+# model architecture
+Arch:
+  name: "RecModel"
+  infer_output_key: "features"
+  infer_add_softmax: False
+  Backbone: 
+    name: "PPLCNet_x2_5"
+    pretrained: True
+    use_ssld: True
+  BackboneStopLayer:
+    name: "flatten_0"
+  Neck:
+    name: "FC"
+    embedding_size: 1280
+    class_num: 512
+  Head:
+    name: "ArcMargin"  
+    embedding_size: 512
+    class_num: 30671
+    margin: 0.15
+    scale: 32
+ 
+# loss function config for traing/eval process
+Loss:
+  Train:
+    - CELoss:
+        weight: 1.0
+    - SupConLoss:
+        weight: 1.0
+        views: 2
+  Eval:
+    - CELoss:
+        weight: 1.0
+
+Optimizer:
+  name: Momentum
+  momentum: 0.9
+  lr:
+    name: Cosine
+    learning_rate: 0.04
+  regularizer:
+    name: 'L2'
+    coeff: 0.0005
+
+# data loader for train and eval
+DataLoader:
+  Train:
+    dataset:
+        name: "VeriWild"
+        image_root: "./dataset/VeRI-Wild/images/"
+        cls_label_path: "./dataset/VeRI-Wild/train_test_split/train_list_start0.txt"
+        transform_ops:
+          - DecodeImage:
+              to_rgb: True
+              channel_first: False
+          - ResizeImage:
+              size: 224
+          - RandFlipImage:
+              flip_code: 1
+          - AugMix:
+              prob: 0.5
+          - NormalizeImage:
+              scale: 0.00392157
+              mean: [0.485, 0.456, 0.406]
+              std: [0.229, 0.224, 0.225]
+              order: ''
+          - RandomErasing:
+              EPSILON: 0.5
+              sl: 0.02
+              sh: 0.4
+              r1: 0.3
+              mean: [0., 0., 0.]
+
+    sampler:
+        name: PKSampler
+        batch_size: 128
+        sample_per_id: 2
+        drop_last: True
+        shuffle: True
+    loader:
+        num_workers: 6
+        use_shared_memory: True
+  Eval:
+    Query:
+      dataset: 
+        name: "VeriWild"
+        image_root: "./dataset/VeRI-Wild/images"
+        cls_label_path: "./dataset/VeRI-Wild/train_test_split/test_3000_id_query.txt"
+        transform_ops:
+          - DecodeImage:
+              to_rgb: True
+              channel_first: False
+          - ResizeImage:
+              size: 224
+          - NormalizeImage:
+              scale: 0.00392157
+              mean: [0.485, 0.456, 0.406]
+              std: [0.229, 0.224, 0.225]
+              order: ''
+      sampler:
+        name: DistributedBatchSampler
+        batch_size: 64
+        drop_last: False
+        shuffle: False
+      loader:
+        num_workers: 6
+        use_shared_memory: True
+
+    Gallery:
+      dataset: 
+        name: "VeriWild"
+        image_root: "./dataset/VeRI-Wild/images"
+        cls_label_path: "./dataset/VeRI-Wild/train_test_split/test_3000_id.txt"
+        transform_ops:
+          - DecodeImage:
+              to_rgb: True
+              channel_first: False
+          - ResizeImage:
+              size: 224
+          - NormalizeImage:
+              scale: 0.00392157
+              mean: [0.485, 0.456, 0.406]
+              std: [0.229, 0.224, 0.225]
+              order: ''
+      sampler:
+        name: DistributedBatchSampler
+        batch_size: 64
+        drop_last: False
+        shuffle: False
+      loader:
+        num_workers: 6
+        use_shared_memory: True
+
+Metric:
+  Eval:
+    - Recallk:
+        topk: [1, 5]
+    - mAP: {}
+

ppcls/loss/__init__.py

@@ -44,12 +44,18 @@ class CombinedLoss(nn.Layer):
 
     def __call__(self, input, batch):
         loss_dict = {}
-        for idx, loss_func in enumerate(self.loss_func):
-            loss = loss_func(input, batch)
-            weight = self.loss_weight[idx]
-            loss = {key: loss[key] * weight for key in loss}
+        # just for accelerate classification traing speed
+        if len(self.loss_func) == 1:
+            loss = self.loss_func[0](input, batch)
             loss_dict.update(loss)
-        loss_dict["loss"] = paddle.add_n(list(loss_dict.values()))
+            loss_dict["loss"] = list(loss.values())[0]
+        else:
+            for idx, loss_func in enumerate(self.loss_func):
+                loss = loss_func(input, batch)
+                weight = self.loss_weight[idx]
+                loss = {key: loss[key] * weight for key in loss}
+                loss_dict.update(loss)
+            loss_dict["loss"] = paddle.add_n(list(loss_dict.values()))
         return loss_dict
 
 

test_tipc/prepare.sh

@@ -51,7 +51,7 @@ if [ ${MODE} = "lite_train_lite_infer" ] || [ ${MODE} = "lite_train_whole_infer"
 	cp -r train/* val/
     fi
     cd ../../
-elif [ ${MODE} = "whole_infer" ] || [ ${MODE} = "cpp_infer" ];then
+elif [ ${MODE} = "whole_infer" ] || [ ${MODE} = "klquant_whole_infer" ];then
     # download data
     cd dataset
     rm -rf ILSVRC2012

test_tipc/test_train_inference_python.sh

@@ -156,7 +156,7 @@ function func_inference(){
     done
 }
 
-if [ ${MODE} = "whole_infer" ] || [${MODE} = "klquant_whole_infer" ]; then
+if [ ${MODE} = "whole_infer" ] || [ ${MODE} = "klquant_whole_infer" ]; then
    IFS="|"
    infer_export_flag=(${infer_export_flag})
    if [ ${infer_export_flag} != "null" ]  && [ ${infer_export_flag} != "False" ]; then
@@ -199,7 +199,7 @@ elif [ ${MODE} = "klquant_whole_infer" ]; then
 	ln -s __params__ inference.pdiparams
 	cd ../../deploy
 	is_quant=True
-        func_inference "${python}" "${inference_py}" "${infer_model}/quant_post_static_model" "../${LOG_PATH}" "${infer_img_dir}" ${is_quant}
+        func_inference "${python}" "${inference_py}" "${infer_model_dir_list}/quant_post_static_model" "../${LOG_PATH}" "${infer_img_dir}" ${is_quant}
 	cd ..
     fi
 else