Merge branch 'update_codegen_doc' of https://github.com/gongel/PaddleNLP into update_codegen_doc

Author: gongel

applications/text_classification/hierarchical/train.py

@@ -182,9 +182,6 @@ def train():
             logits = model(**batch)
             loss = criterion(logits, labels)
 
-            probs = F.sigmoid(logits)
-            metric.update(probs, labels)
-
             loss.backward()
             optimizer.step()
             if args.warmup:
@@ -193,11 +190,10 @@ def train():
 
             global_step += 1
             if global_step % args.logging_steps == 0 and rank == 0:
-                micro_f1_score, macro_f1_score = metric.accumulate()
                 logger.info(
-                    "global step %d, epoch: %d, batch: %d, loss: %.5f, micro f1 score: %.5f, macro f1 score: %.5f, speed: %.2f step/s"
-                    % (global_step, epoch, step, loss, micro_f1_score,
-                       macro_f1_score, 10 / (time.time() - tic_train)))
+                    "global step %d, epoch: %d, batch: %d, loss: %.5f, speed: %.2f step/s"
+                    % (global_step, epoch, step, loss, 10 /
+                       (time.time() - tic_train)))
                 tic_train = time.time()
 
         early_stop_count += 1

applications/text_classification/multi_class/train.py

@@ -180,10 +180,6 @@ def train():
             logits = model(**batch)
             loss = criterion(logits, labels)
 
-            probs = F.softmax(logits, axis=1)
-            correct = metric.compute(probs, labels)
-            metric.update(correct)
-
             loss.backward()
             optimizer.step()
             if args.warmup:
@@ -192,10 +188,9 @@ def train():
 
             global_step += 1
             if global_step % args.logging_steps == 0 and rank == 0:
-                acc = metric.accumulate()
                 logger.info(
-                    "global step %d, epoch: %d, batch: %d, loss: %.5f, acc: %.5f, speed: %.2f step/s"
-                    % (global_step, epoch, step, loss, acc, args.logging_steps /
+                    "global step %d, epoch: %d, batch: %d, loss: %.5f, speed: %.2f step/s"
+                    % (global_step, epoch, step, loss, args.logging_steps /
                        (time.time() - tic_train)))
                 tic_train = time.time()
 

applications/text_classification/multi_label/train.py

@@ -181,9 +181,6 @@ def train():
             logits = model(**batch)
             loss = criterion(logits, labels)
 
-            probs = F.sigmoid(logits)
-            metric.update(probs, labels)
-
             loss.backward()
             optimizer.step()
             if args.warmup:
@@ -192,11 +189,9 @@ def train():
 
             global_step += 1
             if global_step % args.logging_steps == 0 and rank == 0:
-                micro_f1_score, macro_f1_score = metric.accumulate()
                 logger.info(
-                    "global step %d, epoch: %d, batch: %d, loss: %.5f, micro f1 score: %.5f, macro f1 score: %.5f, speed: %.2f step/s"
-                    % (global_step, epoch, step, loss, micro_f1_score,
-                       macro_f1_score, args.logging_steps /
+                    "global step %d, epoch: %d, batch: %d, loss: %.5f, speed: %.2f step/s"
+                    % (global_step, epoch, step, loss, args.logging_steps /
                        (time.time() - tic_train)))
                 tic_train = time.time()