fix ckpt names

.gitignore

@@ -22,6 +22,15 @@ text/.DS_Store
 src/
 workdir/
 *save*/
+test_py.sh
+
+# paraphraser
+*/paraphrase/*.txt
+lightning_logs/
+pytorch_model.bin
+*/paraphrase/bart-*
+cnn*
+/tests/*/
 
 
 # C extensions

genienlp/paraphrase/finetune_bart.py

@@ -10,8 +10,6 @@ from torch.utils.data import DataLoader
 from genienlp.paraphrase.transformer_base import BaseTransformer, add_generic_args, generic_train, get_linear_schedule_with_warmup, MODEL_MODES
 from genienlp.paraphrase.utils import Seq2SeqDataset, sort_checkpoints
 
-from transformers import BartTokenizer, MBartTokenizer
-
 
 logger = logging.getLogger(__name__)
 

genienlp/paraphrase/transformer_base.py

@@ -251,7 +251,8 @@ def add_generic_args(parser, root_dir):
         help="For fp16: Apex AMP optimization level selected in ['O0', 'O1', 'O2', and 'O3']."
         "See details at https://nvidia.github.io/apex/amp.html",
     )
-
+    
+    parser.add_argument("--max_to_keep", type=int, default=2, help="Number of checkpoints to keep")
     parser.add_argument("--n_gpu", type=int, default=1)
     parser.add_argument("--n_tpu_cores", type=int, default=0)
     parser.add_argument("--max_grad_norm", default=1.0, type=float, help="Max gradient norm.")
@@ -289,8 +290,8 @@ def generic_train(model, args):
             raise ValueError("Output directory ({}) already exists and is not empty.".format(args.output_dir))
 
     checkpoint_callback = pl.callbacks.ModelCheckpoint(
-        filepath=args.output_dir, prefix="checkpoint", monitor="val_loss", mode="min", save_top_k=5
-    )
+        filepath=os.path.join(args.output_dir, 'mbart-{epoch:02d}'), monitor="val_loss", mode="min", save_top_k=args.max_to_keep
+)
 
     train_params = dict(
         accumulate_grad_batches=args.gradient_accumulation_steps,