From 9a710b20d7c26363bddae12ca806f60d8692482f Mon Sep 17 00:00:00 2001
From: wukesong <wukesong1@huawei.com>
Date: Sat, 30 May 2020 21:33:53 +0800
Subject: [PATCH] modify wide&deep

---
 model_zoo/wide_and_deep/run_multinpu_train.sh   | 17 +++++++++++++++++
 model_zoo/wide_and_deep/train_and_test.py       |  2 +-
 .../wide_and_deep/train_and_test_multinpu.py    | 14 ++++++++++----
 3 files changed, 28 insertions(+), 5 deletions(-)
 create mode 100644 model_zoo/wide_and_deep/run_multinpu_train.sh

diff --git a/model_zoo/wide_and_deep/run_multinpu_train.sh b/model_zoo/wide_and_deep/run_multinpu_train.sh
new file mode 100644
index 0000000000..db7823eed7
--- /dev/null
+++ b/model_zoo/wide_and_deep/run_multinpu_train.sh
@@ -0,0 +1,17 @@
+#!/bin/bash
+# bash run_multinpu_train.sh
+execute_path=$(pwd)
+
+export RANK_TABLE_FILE=${execute_path}/rank_table_8p.json
+export RANK_SIZE=8
+export MINDSPORE_HCCL_CONFIG_PATH=${execute_path}/rank_table_8p.json
+
+for((i=0;i<=7;i++));
+do
+  rm -rf ${execute_path}/device_$i/
+  mkdir ${execute_path}/device_$i/
+  cd ${execute_path}/device_$i/ || exit
+  export RANK_ID=$i
+  export DEVICE_ID=$i
+  pytest -s ${execute_path}/train_and_test_multinpu.py >train_deep$i.log 2>&1 &
+done
diff --git a/model_zoo/wide_and_deep/train_and_test.py b/model_zoo/wide_and_deep/train_and_test.py
index 5d9832b84b..0b37b67a11 100644
--- a/model_zoo/wide_and_deep/train_and_test.py
+++ b/model_zoo/wide_and_deep/train_and_test.py
@@ -82,7 +82,7 @@ def test_train_eval(config):
     eval_callback = EvalCallBack(model, ds_eval, auc_metric, config)
 
     callback = LossCallBack(config=config)
-    ckptconfig = CheckpointConfig(save_checkpoint_steps=1, keep_checkpoint_max=5)
+    ckptconfig = CheckpointConfig(save_checkpoint_steps=ds_train.get_dataset_size(), keep_checkpoint_max=5)
     ckpoint_cb = ModelCheckpoint(prefix='widedeep_train', directory=config.ckpt_path, config=ckptconfig)
 
     out = model.eval(ds_eval)
diff --git a/model_zoo/wide_and_deep/train_and_test_multinpu.py b/model_zoo/wide_and_deep/train_and_test_multinpu.py
index 9e4b99f546..9a4e154326 100644
--- a/model_zoo/wide_and_deep/train_and_test_multinpu.py
+++ b/model_zoo/wide_and_deep/train_and_test_multinpu.py
@@ -30,7 +30,7 @@ from src.metrics import AUCMetric
 from src.config import WideDeepConfig
 
 sys.path.append(os.path.dirname(os.path.dirname(os.path.abspath(__file__))))
-context.set_context(mode=GRAPH_MODE, device_target="Davinci", save_graph=True)
+context.set_context(mode=context.GRAPH_MODE, device_target="Ascend", save_graphs=True)
 context.set_auto_parallel_context(parallel_mode=ParallelMode.DATA_PARALLEL, mirror_mean=True)
 init()
 
@@ -71,8 +71,8 @@ def test_train_eval():
     test_train_eval
     """
     np.random.seed(1000)
-    config = WideDeepConfig
-    data_path = Config.data_path
+    config = WideDeepConfig()
+    data_path = config.data_path
     batch_size = config.batch_size
     epochs = config.epochs
     print("epochs is {}".format(epochs))
@@ -94,8 +94,14 @@ def test_train_eval():
     eval_callback = EvalCallBack(model, ds_eval, auc_metric, config)
 
     callback = LossCallBack(config=config)
-    ckptconfig = CheckpointConfig(save_checkpoint_steps=1, keep_checkpoint_max=5)
+    ckptconfig = CheckpointConfig(save_checkpoint_steps=ds_train.get_dataset_size(), keep_checkpoint_max=5)
     ckpoint_cb = ModelCheckpoint(prefix='widedeep_train',
                                  directory=config.ckpt_path, config=ckptconfig)
+    out = model.eval(ds_eval)
+    print("=====" * 5 + "model.eval() initialized: {}".format(out))
     model.train(epochs, ds_train,
                 callbacks=[TimeMonitor(ds_train.get_dataset_size()), eval_callback, callback, ckpoint_cb])
+
+
+if __name__ == "__main__":
+    test_train_eval()