Merge pull request #714 from wangzhen38/autofis_fix

autofis fix

Author: frankwhzhang

datasets/criteo_autofis/Criteo.py

@@ -0,0 +1,70 @@
+# Copyright (c) 2020 PaddlePaddle Authors. All Rights Reserved.
+#
+# Licensed under the Apache License, Version 2.0 (the "License");
+# you may not use this file except in compliance with the License.
+# You may obtain a copy of the License at
+#
+#     http://www.apache.org/licenses/LICENSE-2.0
+#
+# Unless required by applicable law or agreed to in writing, software
+# distributed under the License is distributed on an "AS IS" BASIS,
+# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+# See the License for the specific language governing permissions and
+# limitations under the License.
+
+import numpy as np
+import os
+from Criteo import Criteo
+from tqdm import tqdm
+
+batch_size = 1024
+
+train_data_param = {
+    'gen_type': 'train',
+    'random_sample': True,
+    'batch_size': batch_size,
+    'split_fields': False,
+    'on_disk': True,
+    'squeeze_output': True,
+}
+test_data_param = {
+    'gen_type': 'test',
+    'random_sample': False,
+    'batch_size': batch_size,
+    'split_fields': False,
+    'on_disk': True,
+    'squeeze_output': True,
+}
+
+dataset = Criteo(initialized=True)
+train_gen = dataset.batch_generator(train_data_param)
+test_gen = dataset.batch_generator(test_data_param)
+
+output_dir = 'data/whole_data'
+xs = []
+ys = []
+for x, y in tqdm(train_gen):
+    xs.append(x)
+    ys.append(y)
+
+x = np.concatenate(xs, 0)
+y = np.concatenate(ys, 0)
+print(x.shape)
+#np.save(os.path.join(output_dir, 'train', 'train_x.npy'), x)
+#np.save(os.path.join(output_dir, 'train', 'train_y.npy'), y)
+np.savetxt(os.path.join(output_dir, 'train', 'train_x.txt'), x, fmt='%d')
+np.savetxt(os.path.join(output_dir, 'train', 'train_y.txt'), y, fmt='%d')
+
+xs = []
+ys = []
+for x, y in tqdm(test_gen):
+    xs.append(x)
+    ys.append(y)
+
+x = np.concatenate(xs, 0)
+y = np.concatenate(ys, 0)
+print(x.shape)
+#np.save(os.path.join(output_dir, 'test', 'test_x.npy'), x)
+#np.save(os.path.join(output_dir, 'test', 'test_y.npy'), y)
+np.savetxt(os.path.join(output_dir, 'test', 'test_x.txt'), x, fmt='%d')
+np.savetxt(os.path.join(output_dir, 'test', 'text_y.txt'), y, fmt='%d')

datasets/criteo_autofis/get_data.py

@@ -12,41 +12,39 @@
 # See the License for the specific language governing permissions and
 # limitations under the License.
 
+# global settings
 
 runner:
-  train_data_dir: "data/sample_data/train"
+  train_data_dir: "data/sample_data"
   train_reader_path: "criteo_reader" # importlib format
   use_gpu: False
-  use_auc: True
+  use_auc: False
   train_batch_size: 2
-  epochs: 3
-  print_interval: 2
-  # model_init_path: "output_model_deepfm/2" # init model
-  model_save_path: "output_model_deepfm"
-  test_data_dir: "data/sample_data/train"
+  epochs: 1
+  print_interval: 1
+  #model_init_path: "output_model/0" # init model
+  model_save_path: "output_model_autofis"
+  test_data_dir: "data/sample_data"
   infer_reader_path: "criteo_reader" # importlib format
-  infer_batch_size: 5
-  infer_load_path: "output_model_deepfm"
+  infer_batch_size: 2
+  infer_load_path: "output_model_autofis"
   infer_start_epoch: 0
-  infer_end_epoch: 3
-  #use inference save model
-  use_inference: False
-  save_inference_feed_varnames: ["C1","C2","C3","C4","C5","C6","C7","C8","C9","C10","C11","C12","C13","C14","C15","C16","C17","C18","C19","C20","C21","C22","C23","C24","C25","C26","dense_input"]
-  save_inference_fetch_varnames: ["sigmoid_0.tmp_0"]
-  # use fleet
-  use_fleet: False
-  
+  infer_end_epoch: 1
+
 # hyper parameters of user-defined network
 hyper_parameters:
   # optimizer config
   optimizer:
     class: Adam
     learning_rate: 0.001
-    strategy: async
+    gamma: 0.7
   # user-defined <key, value> pairs
-  sparse_inputs_slots: 27
-  sparse_feature_number: 1000001
-  sparse_feature_dim: 9
-  dense_input_dim: 13
-  fc_sizes: [512, 256, 128, 32]
-  distributed_embedding: 0
+  num_inputs: 39
+  input_size: 1178909
+  embedding_size: 40
+  width: 700
+  depth: 5
+  n_col: 741
+  grad_c: 0.0005
+  grad_mu: 0.8
+  pairs: 741

models/rank/autofis/config.yaml

@@ -23,11 +23,11 @@ runner:
   epochs: 1
   print_interval: 50
   #model_init_path: "output_model/0" # init model
-  model_save_path: "output_model_all_autodeepfm"
+  model_save_path: "output_model_autofis_all"
   test_data_dir: "../../../datasets/criteo_autofis/data/whole_data/test"
   infer_reader_path: "criteo_reader" # importlib format
-  infer_batch_size: 2000
-  infer_load_path: "output_model_all_autodeepfm"
+  infer_batch_size: 200
+  infer_load_path: "output_model_autofis_all"
   infer_start_epoch: 0
   infer_end_epoch: 1
 

models/rank/autofis/config_bigdata.yaml

@@ -22,10 +22,10 @@ class RecDataset(Dataset):
     def __init__(self, file_list, config):
         super(RecDataset, self).__init__()
         for file in file_list:
-            if file.endswith('x.npy'):
-                self.x = np.load(file)
-            elif file.endswith('y.npy'):
-                self.y = np.load(file)
+            if file.endswith('x.txt'):
+                self.x = np.loadtxt(file, dtype=np.int64)
+            elif file.endswith('y.txt'):
+                self.y = np.loadtxt(file, dtype=np.int64)
 
     def __getitem__(self, item):
         return self.x[item], self.y[item]