fix loss reduce from dict to list (#679)

* fix loss reduce from dict to list

* remove note

tools/program.py

@@ -143,8 +143,8 @@ def create_metric(out,
             out = out[1]
     softmax_out = F.softmax(out)
 
-    fetchs = OrderedDict()
-    metric_names = set()
+    fetch_list = []
+    metric_names = []
     if not multilabel:
         softmax_out = F.softmax(out)
 
@@ -154,12 +154,11 @@ def create_metric(out,
         k = min(topk, classes_num)
         topk = paddle.metric.accuracy(softmax_out, label=label, k=k)
 
-        metric_names.add("top1")
-        metric_names.add("top{}".format(k))
+        metric_names.append("top1")
+        metric_names.append("top{}".format(k))
 
-        fetchs['top1'] = top1
-        topk_name = "top{}".format(k)
-        fetchs[topk_name] = topk
+        fetch_list.append(top1)
+        fetch_list.append(topk)
     else:
         out = F.sigmoid(out)
         preds = multi_hot_encode(out.numpy())
@@ -169,19 +168,22 @@ def create_metric(out,
 
         ham_dist_name = "hamming_distance"
         accuracy_name = "multilabel_accuracy"
-        metric_names.add(ham_dist_name)
-        metric_names.add(accuracy_name)
+        metric_names.append(ham_dist_name)
+        metric_names.append(accuracy_name)
 
-        fetchs[accuracy_name] = accuracy
-        fetchs[ham_dist_name] = ham_dist
+        fetch_list.append(accuracy)
+        fetch_list.append(ham_dist)
 
     # multi cards' eval
     if mode != "train" and paddle.distributed.get_world_size() > 1:
-        for metric_name in metric_names:
-            fetchs[metric_name] = paddle.distributed.all_reduce(
-                fetchs[metric_name], op=paddle.distributed.ReduceOp.
+        for idx, fetch in enumerate(fetch_list):
+            fetch_list[idx] = paddle.distributed.all_reduce(
+                fetch, op=paddle.distributed.ReduceOp.
                 SUM) / paddle.distributed.get_world_size()
 
+    fetchs = OrderedDict()
+    for idx, name in enumerate(metric_names):
+        fetchs[name] = fetch_list[idx]
     return fetchs
 
 
@@ -282,7 +284,8 @@ def create_feeds(batch, use_mix, num_classes, multilabel=False):
         if not multilabel:
             label = to_tensor(batch[1].numpy().astype("int64").reshape(-1, 1))
         else:
-            label = to_tensor(batch[1].numpy().astype('float32').reshape(-1, num_classes))
+            label = to_tensor(batch[1].numpy().astype('float32').reshape(
+                -1, num_classes))
         feeds = {"image": image, "label": label}
     return feeds
 
@@ -336,9 +339,11 @@ def run(dataloader,
                 0, ("top1", AverageMeter(
                     "top1", '.5f', postfix=",")))
         else:
-            metric_list.insert(0, ("multilabel_accuracy", AverageMeter(
+            metric_list.insert(
+                0, ("multilabel_accuracy", AverageMeter(
                     "multilabel_accuracy", '.5f', postfix=",")))
-            metric_list.insert(0, ("hamming_distance", AverageMeter(
+            metric_list.insert(
+                0, ("hamming_distance", AverageMeter(
                     "hamming_distance", '.5f', postfix=",")))
 
     metric_list = OrderedDict(metric_list)