update gnn

models/recall/gnn/config.yaml

@@ -49,31 +49,31 @@ runner:
 - name: train_runner
   class: train
   # num of epochs
-  epochs: 2
+  epochs: 5
   # device to run training or infer
   device: cpu
   save_checkpoint_interval: 1 # save model interval of epochs
   save_inference_interval: 1 # save inference
-  save_checkpoint_path: "increment" # save checkpoint path
-  save_inference_path: "inference" # save inference path
+  save_checkpoint_path: "increment_gnn" # save checkpoint path
+  save_inference_path: "inference_gnn" # save inference path
   save_inference_feed_varnames: [] # feed vars of save inference
   save_inference_fetch_varnames: [] # fetch vars of save inference
   init_model_path: "" # load model path
-  print_interval: 1
+  print_interval: 10
 - name: infer_runner
   class: infer
   # device to run training or infer
   device: cpu
   print_interval: 1
-  init_model_path: "increment/0" # load model path
+  init_model_path: "increment_gnn" # load model path
 
 # runner will run all the phase in each epoch
 phase:
-- name: phase1
+- name: phase_train
   model: "{workspace}/model.py" # user-defined model
   dataset_name: dataset_train # select dataset by name
   thread_num: 1
-# - name: phase2
+# - name: phase_infer
 #   model: "{workspace}/model.py" # user-defined model
 #   dataset_name: dataset_infer # select dataset by name
 #   thread_num: 1

models/recall/gnn/data/download.py

@@ -57,5 +57,10 @@ def _download_file(url, savepath, print_progress):
             progress("[%-50s] %.2f%%" % ('=' * 50, 100), end=True)
 
 
-_download_file("https://sr-gnn.bj.bcebos.com/train-item-views.csv",
+if sys.argv[1] == "diginetica":
+    _download_file("https://sr-gnn.bj.bcebos.com/train-item-views.csv",
                    "./train-item-views.csv", True)
+elif sys.argv[1] == "yoochoose":
+    _download_file(
+        "https://paddlerec.bj.bcebos.com/gnn%2Fyoochoose-clicks.dat",
+        "./yoochoose-clicks.dat", True)

models/recall/gnn/data/preprocess.py

@@ -41,39 +41,29 @@ with open(dataset, "r") as f:
     curdate = None
     for data in reader:
         sessid = data['session_id']
-        if curdate and not curid == sessid:
         date = ''
-            if opt.dataset == 'yoochoose':
-                date = time.mktime(
-                    time.strptime(curdate[:19], '%Y-%m-%dT%H:%M:%S'))
-            else:
-                date = time.mktime(time.strptime(curdate, '%Y-%m-%d'))
-            sess_date[curid] = date
-        curid = sessid
         if opt.dataset == 'yoochoose':
             item = data['item_id']
+            date = time.mktime(
+                time.strptime(data['timestamp'][:19], '%Y-%m-%dT%H:%M:%S'))
         else:
             item = data['item_id'], int(data['timeframe'])
-        curdate = ''
-        if opt.dataset == 'yoochoose':
-            curdate = data['timestamp']
-        else:
-            curdate = data['eventdate']
+            date = time.mktime(time.strptime(data['eventdate'], '%Y-%m-%d'))
+
+        if sessid not in sess_date:
+            sess_date[sessid] = date
+        elif date > sess_date[sessid]:
+            sess_date[sessid] = date
 
         if sessid in sess_clicks:
             sess_clicks[sessid] += [item]
         else:
             sess_clicks[sessid] = [item]
         ctr += 1
-    date = ''
-    if opt.dataset == 'yoochoose':
-        date = time.mktime(time.strptime(curdate[:19], '%Y-%m-%dT%H:%M:%S'))
-    else:
-        date = time.mktime(time.strptime(curdate, '%Y-%m-%d'))
+    if opt.dataset != 'yoochoose':
         for i in list(sess_clicks):
             sorted_clicks = sorted(sess_clicks[i], key=operator.itemgetter(1))
             sess_clicks[i] = [c[0] for c in sorted_clicks]
-    sess_date[curid] = date
 print("-- Reading data @ %ss" % datetime.datetime.now())
 
 # Filter out length 1 sessions
@@ -160,7 +150,7 @@ def obtian_tra():
         train_dates += [date]
         train_seqs += [outseq]
     print(item_ctr)  # 43098, 37484
-    with open("./diginetica/config.txt", "w") as fout:
+    with open("./config.txt", "w") as fout:
         fout.write(str(item_ctr) + "\n")
     return train_ids, train_dates, train_seqs
 

models/recall/gnn/data_prepare.sh

@@ -15,21 +15,31 @@
 # limitations under the License.
 
 set -e
-echo "begin to download data"
 
-cd data && python download.py
-mkdir diginetica
-python preprocess.py --dataset diginetica
+dataset=$1
+src=$1
+
+if [[ $src == "yoochoose1_4" || $src == "yoochoose1_64" ]];then
+    src="yoochoose"
+elif [[ $src == "diginetica" ]];then
+    src="diginetica"
+else
+    echo "Usage: sh data_prepare.sh [diginetica|yoochoose1_4|yoochoose1_64]"
+    exit 1
+fi
+
+echo "begin to download data"
+cd data && python download.py $src
+mkdir $dataset
+python preprocess.py --dataset $src
 
 echo "begin to convert data (binary -> txt)"
-python convert_data.py --data_dir diginetica
+python convert_data.py --data_dir $dataset
 
-cat diginetica/train.txt | wc -l >> diginetica/config.txt
+cat ${dataset}/train.txt | wc -l >> config.txt
 
 rm -rf train && mkdir train
-mv diginetica/train.txt train
+mv ${dataset}/train.txt train
 
 rm -rf test && mkdir test
-mv diginetica/test.txt test
-
-mv diginetica/config.txt ./config.txt
+mv ${dataset}/test.txt test

models/recall/gnn/model.py

@@ -20,6 +20,7 @@ import paddle.fluid.layers as layers
 
 from paddlerec.core.utils import envs
 from paddlerec.core.model import ModelBase
+from paddlerec.core.metrics import Precision
 
 
 class Model(ModelBase):
@@ -235,16 +236,16 @@ class Model(ModelBase):
         softmax = layers.softmax_with_cross_entropy(
             logits=logits, label=inputs[6])  # [batch_size, 1]
         self.loss = layers.reduce_mean(softmax)  # [1]
-        self.acc = layers.accuracy(input=logits, label=inputs[6], k=20)
-
+        acc = Precision(input=logits, label=inputs[6], k=20)
         self._cost = self.loss
+
         if is_infer:
-            self._infer_results['acc'] = self.acc
-            self._infer_results['loss'] = self.loss
+            self._infer_results['P@20'] = acc
+            self._infer_results['LOSS'] = self.loss
             return
 
         self._metrics["LOSS"] = self.loss
-        self._metrics["train_acc"] = self.acc
+        self._metrics["Train_P@20"] = acc
 
     def optimizer(self):
         step_per_epoch = self.corpus_size // self.train_batch_size