From e0c1fb3a82231a8e3e888973e0b3c00be56b344e Mon Sep 17 00:00:00 2001
From: xuezhong <xuezhongq@gmail.com>
Date: Fri, 2 Nov 2018 06:40:05 +0000
Subject: [PATCH] default setting

---
 .../machine_reading_comprehension/.run_ce.sh  | 45 ++++++++++++
 .../machine_reading_comprehension/_ce.py      | 68 +++++++++++++++++++
 2 files changed, 113 insertions(+)
 create mode 100644 fluid/PaddleNLP/machine_reading_comprehension/.run_ce.sh
 create mode 100644 fluid/PaddleNLP/machine_reading_comprehension/_ce.py

diff --git a/fluid/PaddleNLP/machine_reading_comprehension/.run_ce.sh b/fluid/PaddleNLP/machine_reading_comprehension/.run_ce.sh
new file mode 100644
index 00000000..715d5bb3
--- /dev/null
+++ b/fluid/PaddleNLP/machine_reading_comprehension/.run_ce.sh
@@ -0,0 +1,45 @@
+#!/bin/bash
+
+DATA_PATH=./data
+if [ ! -e $DATA_PATH/demo ] ; then
+    mkdir -p $DATA_PATH/demo
+    if [ ! -e $DATA_PATH/demo.tgz ] ; then
+        cd $DATA_PATH
+        wget -c --no-check-certificate http://dureader.gz.bcebos.com/demo.tgz
+        cd -
+    fi
+    tar -zxf $DATA_PATH/demo.tgz -C $DATA_PATH/demo
+fi
+
+train(){
+python  -u run.py   \
+	--trainset 'data/demo/search.train.json' \
+	--devset 'data/demo/search.dev.json' \
+	--testset 'data/demo/search.test.json' \
+	--vocab_dir 'data/demo/' \
+	--use_gpu true \
+	--save_dir ./models \
+	--pass_num 1 \
+	--learning_rate 0.001 \
+	--batch_size 32 \
+	--embed_size 300 \
+	--hidden_size 150 \
+	--max_p_num 5 \
+	--max_p_len 500 \
+	--max_q_len 60 \
+	--max_a_len 200 \
+	--drop_rate 0.2 \
+	--log_interval 1 \
+        --enable_ce \
+        --train 
+}
+
+cudaid=${transformer:=0} # use 0-th card as default
+export CUDA_VISIBLE_DEVICES=$cudaid
+
+train | python _ce.py
+
+cudaid=${transformer_m:=0,1} # use 0,1 card as default
+export CUDA_VISIBLE_DEVICES=$cudaid
+
+train | python _ce.py
diff --git a/fluid/PaddleNLP/machine_reading_comprehension/_ce.py b/fluid/PaddleNLP/machine_reading_comprehension/_ce.py
new file mode 100644
index 00000000..24243846
--- /dev/null
+++ b/fluid/PaddleNLP/machine_reading_comprehension/_ce.py
@@ -0,0 +1,68 @@
+####this file is only used for continuous evaluation test!
+
+import os
+import sys
+#sys.path.insert(0, os.environ['ceroot'])
+from kpi import CostKpi, DurationKpi, AccKpi
+
+#### NOTE kpi.py should shared in models in some way!!!!
+
+train_cost_card1_kpi = CostKpi('train_cost_card1', 0.02, 0, actived=True)
+test_cost_card1_kpi = CostKpi('test_cost_card1', 0.005, 0, actived=True)
+train_duration_card1_kpi = DurationKpi(
+    'train_duration_card1', 0.06, 0, actived=True)
+train_cost_card4_kpi = CostKpi('train_cost_card2', 0.01, 0, actived=True)
+test_cost_card4_kpi = CostKpi('test_cost_card2', 0.005, 0, actived=True)
+train_duration_card4_kpi = DurationKpi(
+    'train_duration_card2', 0.06, 0, actived=True)
+
+tracking_kpis = [
+    train_cost_card1_kpi,
+    test_cost_card1_kpi,
+    train_duration_card1_kpi,
+    train_cost_card4_kpi,
+    test_cost_card4_kpi,
+    train_duration_card4_kpi,
+]
+
+
+def parse_log(log):
+    '''
+    This method should be implemented by model developers.
+    The suggestion:
+    each line in the log should be key, value, for example:
+    "
+    train_cost\t1.0
+    test_cost\t1.0
+    train_cost\t1.0
+    train_cost\t1.0
+    train_acc\t1.2
+    "
+    '''
+    for line in log.split('\n'):
+        fs = line.strip().split('\t')
+        print(fs)
+        if len(fs) == 3 and fs[0] == 'kpis':
+            print("-----%s" % fs)
+            kpi_name = fs[1]
+            kpi_value = float(fs[2])
+            yield kpi_name, kpi_value
+
+
+def log_to_ce(log):
+    kpi_tracker = {}
+    for kpi in tracking_kpis:
+        kpi_tracker[kpi.name] = kpi
+
+    for (kpi_name, kpi_value) in parse_log(log):
+        print(kpi_name, kpi_value)
+        kpi_tracker[kpi_name].add_record(kpi_value)
+        kpi_tracker[kpi_name].persist()
+
+
+if __name__ == '__main__':
+    log = sys.stdin.read()
+    print("*****")
+    print(log)
+    print("****")
+    log_to_ce(log)
-- 
GitLab