[s2s] allow task_specific_params=summarization_xsum (#6923)

examples/seq2seq/callbacks.py

@@ -75,7 +75,7 @@ class Seq2SeqLoggingCallback(pl.Callback):
         return self._write_logs(trainer, pl_module, "test")
 
 
-def get_checkpoint_callback(output_dir, metric):
+def get_checkpoint_callback(output_dir, metric, save_top_k=1):
     """Saves the best model by validation ROUGE2 score."""
     if metric == "rouge2":
         exp = "{val_avg_rouge2:.4f}-{step_count}"
@@ -90,7 +90,7 @@ def get_checkpoint_callback(output_dir, metric):
         filepath=os.path.join(output_dir, exp),
         monitor=f"val_{metric}",
         mode="max",
-        save_top_k=1,
+        save_top_k=save_top_k,
         period=0,  # maybe save a checkpoint every time val is run, not just end of epoch.
     )
     return checkpoint_callback

examples/seq2seq/finetune.py

@@ -306,6 +306,7 @@ class SummarizationModule(BaseTransformer):
         parser.add_argument("--tgt_lang", type=str, default="", required=False)
         parser.add_argument("--eval_beams", type=int, default=None, required=False)
         parser.add_argument("--val_metric", type=str, default=None, required=False)
+        parser.add_argument("--save_top_k", type=int, default=1, required=False, help="How many checkpoints to save")
         parser.add_argument(
             "--early_stopping_patience",
             type=int,
@@ -336,7 +337,7 @@ def main(args, model=None) -> SummarizationModule:
     if len(os.listdir(args.output_dir)) > 3 and args.do_train:
         raise ValueError("Output directory ({}) already exists and is not empty.".format(args.output_dir))
     if model is None:
-        if args.task == "summarization":
+        if "summarization" in args.task:
             model: SummarizationModule = SummarizationModule(args)
         else:
             model: SummarizationModule = TranslationModule(args)
@@ -368,7 +369,7 @@ def main(args, model=None) -> SummarizationModule:
         model,
         args,
         logging_callback=Seq2SeqLoggingCallback(),
-        checkpoint_callback=get_checkpoint_callback(args.output_dir, model.val_metric),
+        checkpoint_callback=get_checkpoint_callback(args.output_dir, model.val_metric, args.save_top_k),
         early_stopping_callback=es_callback,
         logger=logger,
         # TODO: early stopping callback seems messed up

examples/seq2seq/test_seq2seq_examples.py

@@ -34,6 +34,7 @@ CHEAP_ARGS = {
     "label_smoothing": 0.2,
     "eval_beams": 1,
     "val_metric": None,
+    "save_top_k": 1,
     "adafactor": True,
     "early_stopping_patience": 2,
     "logger_name": "default",