move files

dataset/flowers102/generate_flowers102_list.py

@@ -1,38 +0,0 @@
-"""
-.mat files data format
-imagelabel.mat
-jpg_name   1     2     3 ... 
-label      32    12    66 ... 
-
-setid.mat
-jpg_name(10 records in a class)  24 6 100 65 32 ...
-label                            4 ...
-"""
-"""
-Usage: 
-    python generate_flower_list.py prefix_folder mode
-    python generate_flower_list.py jpg train > train_list.txt
-    python generate_flower_list.py jpg valid > val_list.txt
-"""
-
-import scipy.io
-import numpy as np
-import os
-import sys
-
-data_path = sys.argv[1]
-imagelabels_path = './imagelabels.mat'
-setid_path = './setid.mat'
-
-labels = scipy.io.loadmat(imagelabels_path)
-labels = np.array(labels['labels'][0])
-setid = scipy.io.loadmat(setid_path)
-
-d = {}
-d['train'] = np.array(setid['trnid'][0])
-d['valid'] = np.array(setid['valid'][0])
-d['test'] = np.array(setid['tstid'][0])
-
-for id in d[sys.argv[2]]:
-   message = str(data_path) + "/image_" + str(id).zfill(5) + ".jpg " + str(labels[id - 1] - 1)
-   print(message)

deploy/slim/quant/export_model.py

@@ -21,7 +21,7 @@ sys.path.append(os.path.abspath(os.path.join(__dir__, '..', '..', '..')))
 sys.path.append(
     os.path.abspath(os.path.join(__dir__, '..', '..', '..', 'tools')))
 
-from ppcls.modeling import architectures
+from ppcls.arch import backbone
 from ppcls.utils.save_load import load_dygraph_pretrain
 import paddle
 import paddle.nn.functional as F
@@ -63,7 +63,7 @@ class Net(paddle.nn.Layer):
 def main():
     args = parse_args()
 
-    net = architectures.__dict__[args.model]
+    net = backbone.__dict__[args.model]
     model = Net(net, args.class_dim, args.model)
 
     # get QAT model

docs/zh_CN/faq_series/faq_2020_s1.md

@@ -136,7 +136,7 @@ ResNet系列模型中，相比于其他模型，ResNet_vd模型在预测速度
 
 **A**：
 
-* 可以使用自动混合精度进行训练，这在精度几乎无损的情况下，可以有比较明显的速度收益，以ResNet50为例，PaddleClas中使用自动混合精度训练的配置文件可以参考：[ResNet50_fp16.yml](../../../configs/ResNet/ResNet50_fp16.yml)，主要就是需要在标准的配置文件中添加以下几行
+* 可以使用自动混合精度进行训练，这在精度几乎无损的情况下，可以有比较明显的速度收益，以ResNet50为例，PaddleClas中使用自动混合精度训练的配置文件可以参考：[ResNet50_fp16.yml](../../../ppcls/configs/ResNet/ResNet50_fp16.yml)，主要就是需要在标准的配置文件中添加以下几行
 
 ```
 use_fp16: True

docs/zh_CN/feature_visiualization/get_started.md

@@ -6,7 +6,7 @@
 
 ## 二、准备工作
 
-首先需要选定研究的模型，本文设定ResNet50作为研究模型，将resnet.py从[模型库](../../../ppcls/modeling/architecture/)拷贝到当前目录下，并下载预训练模型[预训练模型](../../zh_CN/models/models_intro), 复制resnet50的模型链接，使用下列命令下载并解压预训练模型。
+首先需要选定研究的模型，本文设定ResNet50作为研究模型，将resnet.py从[模型库](../../../ppcls/arch/architecture/)拷贝到当前目录下，并下载预训练模型[预训练模型](../../zh_CN/models/models_intro), 复制resnet50的模型链接，使用下列命令下载并解压预训练模型。
 
 ```bash
 wget The Link for Pretrained Model

tools/export_model.py

@@ -19,7 +19,7 @@ __dir__ = os.path.dirname(os.path.abspath(__file__))
 sys.path.append(__dir__)
 sys.path.append(os.path.abspath(os.path.join(__dir__, '..')))
 
-from ppcls.modeling import architectures
+from ppcls.arch import backbone
 from ppcls.utils.save_load import load_dygraph_pretrain
 import paddle
 import paddle.nn.functional as F
@@ -64,7 +64,7 @@ class Net(paddle.nn.Layer):
 def main():
     args = parse_args()
 
-    net = architectures.__dict__[args.model]
+    net = backbone.__dict__[args.model]
     model = Net(net, args.class_dim, args.model)
     load_dygraph_pretrain(
         model.pre_net,

tools/export_serving_model.py

@@ -14,7 +14,7 @@
 
 import argparse
 import os
-from ppcls.modeling import architectures
+from ppcls.arch import backbone
 
 import paddle.fluid as fluid
 import paddle_serving_client.io as serving_io
@@ -49,7 +49,7 @@ def create_model(args, model, input, class_dim=1000):
 def main():
     args = parse_args()
 
-    model = architectures.__dict__[args.model]()
+    model = backbone.__dict__[args.model]()
 
     place = fluid.CPUPlace()
     exe = fluid.Executor(place)

tools/infer/infer.py

@@ -26,7 +26,7 @@ sys.path.append(os.path.abspath(os.path.join(__dir__, '../..')))
 
 from ppcls.utils.save_load import load_dygraph_pretrain
 from ppcls.utils import logger
-from ppcls.modeling import architectures
+from ppcls.arch import backbone
 from utils import parse_args, get_image_list, preprocess, postprocess, save_prelabel_results
 
 
@@ -36,7 +36,7 @@ def main():
     place = paddle.set_device('gpu' if args.use_gpu else 'cpu')
     multilabel = True if args.multilabel else False
 
-    net = architectures.__dict__[args.model](class_dim=args.class_num)
+    net = backbone.__dict__[args.model](class_dim=args.class_num)
     load_dygraph_pretrain(net, args.pretrained_model, args.load_static_weights)
     image_list = get_image_list(args.image_file)
     batch_input_list = []

tools/program.py

@@ -16,24 +16,22 @@ from __future__ import absolute_import
 from __future__ import division
 from __future__ import print_function
 
-import os
 import time
 import datetime
 from collections import OrderedDict
 
 import paddle
 from paddle import to_tensor
-import paddle.nn as nn
 import paddle.nn.functional as F
 
 from ppcls.optimizer import LearningRateBuilder
 from ppcls.optimizer import OptimizerBuilder
-from ppcls.modeling import architectures
-from ppcls.modeling.loss import MultiLabelLoss
-from ppcls.modeling.loss import CELoss
-from ppcls.modeling.loss import MixCELoss
-from ppcls.modeling.loss import JSDivLoss
-from ppcls.modeling.loss import GoogLeNetLoss
+from ppcls.arch import backbone
+from ppcls.arch.loss import MultiLabelLoss
+from ppcls.arch.loss import CELoss
+from ppcls.arch.loss import MixCELoss
+from ppcls.arch.loss import JSDivLoss
+from ppcls.arch.loss import GoogLeNetLoss
 from ppcls.utils.misc import AverageMeter
 from ppcls.utils import logger
 from ppcls.utils import profiler
@@ -57,7 +55,7 @@ def create_model(architecture, classes_num):
     """
     name = architecture["name"]
     params = architecture.get("params", {})
-    return architectures.__dict__[name](class_dim=classes_num, **params)
+    return backbone.__dict__[name](class_dim=classes_num, **params)
 
 
 def create_loss(feeds,

tools/static/program.py

@@ -16,7 +16,6 @@ from __future__ import absolute_import
 from __future__ import division
 from __future__ import print_function
 
-import os
 import time
 import numpy as np
 
@@ -27,11 +26,11 @@ import paddle
 import paddle.nn.functional as F
 
 from ppcls.optimizer.learning_rate import LearningRateBuilder
-from ppcls.modeling import architectures
-from ppcls.modeling.loss import CELoss
-from ppcls.modeling.loss import MixCELoss
-from ppcls.modeling.loss import JSDivLoss
-from ppcls.modeling.loss import GoogLeNetLoss
+from ppcls.arch import backbone
+from ppcls.arch.loss import CELoss
+from ppcls.arch.loss import MixCELoss
+from ppcls.arch.loss import JSDivLoss
+from ppcls.arch.loss import GoogLeNetLoss
 from ppcls.utils.misc import AverageMeter
 from ppcls.utils import logger, profiler
 
@@ -95,7 +94,7 @@ def create_model(architecture, image, classes_num, config, is_train):
         params["input_image_channel"] = input_image_channel
     if "is_test" in params:
         params['is_test'] = not is_train
-    model = architectures.__dict__[name](class_dim=classes_num, **params)
+    model = backbone.__dict__[name](class_dim=classes_num, **params)
 
     out = model(image)
     return out