Revert renaming in finetune_trainer (#9262)

examples/seq2seq/finetune_trainer.py

@@ -93,19 +93,27 @@ class DataTrainingArguments:
             "than this will be truncated, sequences shorter will be padded."
         },
     )
-    max_length: Optional[int] = field(
+    max_target_length: Optional[int] = field(
         default=128,
         metadata={
             "help": "The maximum total sequence length for target text after tokenization. Sequences longer "
             "than this will be truncated, sequences shorter will be padded."
         },
     )
-    eval_max_length: Optional[int] = field(
+    val_max_target_length: Optional[int] = field(
         default=142,
         metadata={
             "help": "The maximum total sequence length for validation target text after tokenization. Sequences longer "
+            "than this will be truncated, sequences shorter will be padded. "
+            "This argument is also used to override the ``max_length`` param of ``model.generate``, which is used "
+            "during ``evaluate`` and ``predict``."
+        },
+    )
+    test_max_target_length: Optional[int] = field(
+        default=142,
+        metadata={
+            "help": "The maximum total sequence length for test target text after tokenization. Sequences longer "
             "than this will be truncated, sequences shorter will be padded."
-            " This argument is also used to override the ``max_length`` param of ``model.generate``, which is used during ``evaluate`` and ``predict``"
         },
     )
     n_train: Optional[int] = field(default=-1, metadata={"help": "# training examples. -1 means use all."})
@@ -233,7 +241,7 @@ def main():
             type_path="train",
             data_dir=data_args.data_dir,
             n_obs=data_args.n_train,
-            max_target_length=data_args.max_length,
+            max_target_length=data_args.max_target_length,
             max_source_length=data_args.max_source_length,
             prefix=model.config.prefix or "",
         )
@@ -246,7 +254,7 @@ def main():
             type_path="val",
             data_dir=data_args.data_dir,
             n_obs=data_args.n_val,
-            max_target_length=data_args.eval_max_length,
+            max_target_length=data_args.val_max_target_length,
             max_source_length=data_args.max_source_length,
             prefix=model.config.prefix or "",
         )
@@ -259,7 +267,7 @@ def main():
             type_path="test",
             data_dir=data_args.data_dir,
             n_obs=data_args.n_test,
-            max_target_length=data_args.eval_max_length,
+            max_target_length=data_args.test_max_target_length,
             max_source_length=data_args.max_source_length,
             prefix=model.config.prefix or "",
         )
@@ -310,7 +318,7 @@ def main():
         logger.info("*** Evaluate ***")
 
         metrics = trainer.evaluate(
-            metric_key_prefix="val", max_length=data_args.eval_max_length, num_beams=data_args.eval_beams
+            metric_key_prefix="val", max_length=data_args.val_max_target_length, num_beams=data_args.eval_beams
         )
         metrics["val_n_objs"] = data_args.n_val
         metrics["val_loss"] = round(metrics["val_loss"], 4)
@@ -326,7 +334,7 @@ def main():
         test_output = trainer.predict(
             test_dataset=test_dataset,
             metric_key_prefix="test",
-            max_length=data_args.eval_max_length,
+            max_length=data_args.val_max_target_length,
             num_beams=data_args.eval_beams,
         )
         metrics = test_output.metrics

examples/seq2seq/test_finetune_trainer.py

@@ -137,8 +137,8 @@ class TestFinetuneTrainer(TestCasePlus):
             --n_train 8
             --n_val 8
             --max_source_length {max_len}
-            --max_length {max_len}
+            --max_target_length {max_len}
-            --eval_max_length {max_len}
+            --val_max_target_length {max_len}
             --do_train
             --do_eval
             --do_predict

examples/seq2seq/train_distil_marian_enro.sh

@@ -29,7 +29,8 @@ python finetune_trainer.py \
     --freeze_encoder --freeze_embeds \
     --num_train_epochs=6 \
     --save_steps 3000 --eval_steps 3000 \
-    --max_source_length $MAX_LEN --max_length $MAX_LEN --eval_max_length $MAX_LEN \
+    --max_source_length $MAX_LEN --max_target_length $MAX_LEN \
+    --val_max_target_length $MAX_TGT_LEN --test_max_target_length $MAX_TGT_LEN \
     --do_train --do_eval --do_predict \
     --evaluation_strategy steps \
     --predict_with_generate --logging_first_step \

examples/seq2seq/train_distil_marian_enro_tpu.sh

@@ -30,7 +30,8 @@ python xla_spawn.py --num_cores $TPU_NUM_CORES \
     --num_train_epochs=6 \
     --save_steps 500 --eval_steps 500 \
     --logging_first_step --logging_steps 200 \
-    --max_source_length $MAX_LEN --max_length $MAX_LEN --eval_max_length $MAX_LEN \
+    --max_source_length $MAX_LEN --max_target_length $MAX_LEN \
+    --val_max_target_length $MAX_TGT_LEN --test_max_target_length $MAX_TGT_LEN \
     --do_train --do_eval \
     --evaluation_strategy steps \
     --prediction_loss_only \

examples/seq2seq/train_distilbart_cnn.sh

@@ -32,7 +32,7 @@ python finetune_trainer.py \
     --num_train_epochs=2 \
     --save_steps 3000 --eval_steps 3000 \
     --logging_first_step \
-    --max_length 56 --eval_max_length $MAX_TGT_LEN \
+    --max_target_length 56 --val_max_target_length $MAX_TGT_LEN --test_max_target_length $MAX_TGT_LEN\
     --do_train --do_eval --do_predict \
     --evaluation_strategy steps \
     --predict_with_generate --sortish_sampler \

examples/seq2seq/train_mbart_cc25_enro.sh

@@ -24,7 +24,7 @@ python finetune_trainer.py \
     --src_lang en_XX --tgt_lang ro_RO \
     --freeze_embeds \
     --per_device_train_batch_size=4 --per_device_eval_batch_size=4 \
-    --max_source_length 128 --max_length 128 --eval_max_length 128 \
+    --max_source_length 128 --max_target_length 128 --val_max_target_length 128 --test_max_target_length 128\
     --sortish_sampler \
     --num_train_epochs 6 \
     --save_steps 25000 --eval_steps 25000 --logging_steps 1000 \

examples/seq2seq/utils.py

@@ -330,7 +330,7 @@ class Seq2SeqDataCollator:
             [x["src_texts"] for x in batch],
             tgt_texts=[x["tgt_texts"] for x in batch],
             max_length=self.data_args.max_source_length,
-            max_target_length=self.data_args.max_length,
+            max_target_length=self.data_args.max_target_length,
             padding="max_length" if self.tpu_num_cores is not None else "longest",  # TPU hack
             return_tensors="pt",
             **self.dataset_kwargs,