[bart-tiny-random] Put a 5MB model on S3 to allow faster exampl… (#3488)

examples/summarization/bart/evaluate_cnn.py

@@ -16,15 +16,17 @@ def chunks(lst, n):
         yield lst[i : i + n]
 
 
-def generate_summaries(lns, out_file, batch_size=8, device=DEFAULT_DEVICE):
+def generate_summaries(
+    examples: list, out_file: str, model_name: str, batch_size: int = 8, device: str = DEFAULT_DEVICE
+):
     fout = Path(out_file).open("w")
-    model = BartForConditionalGeneration.from_pretrained("bart-large-cnn", output_past=True,).to(device)
+    model = BartForConditionalGeneration.from_pretrained(model_name, output_past=True,).to(device)
     tokenizer = BartTokenizer.from_pretrained("bart-large")
 
     max_length = 140
     min_length = 55
 
-    for batch in tqdm(list(chunks(lns, batch_size))):
+    for batch in tqdm(list(chunks(examples, batch_size))):
         dct = tokenizer.batch_encode_plus(batch, max_length=1024, return_tensors="pt", pad_to_max_length=True)
         summaries = model.generate(
             input_ids=dct["input_ids"].to(device),
@@ -51,6 +53,9 @@ def _run_generate():
     parser.add_argument(
         "output_path", type=str, help="where to save summaries",
     )
+    parser.add_argument(
+        "model_name", type=str, default="bart-large-cnn", help="like bart-large-cnn",
+    )
     parser.add_argument(
         "--device", type=str, required=False, default=DEFAULT_DEVICE, help="cuda, cuda:1, cpu etc.",
     )
@@ -58,8 +63,8 @@ def _run_generate():
         "--bs", type=int, default=8, required=False, help="batch size: how many to summarize at a time",
     )
     args = parser.parse_args()
-    lns = [" " + x.rstrip() for x in open(args.source_path).readlines()]
+    examples = [" " + x.rstrip() for x in open(args.source_path).readlines()]
-    generate_summaries(lns, args.output_path, batch_size=args.bs, device=args.device)
+    generate_summaries(examples, args.output_path, args.model_name, batch_size=args.bs, device=args.device)
 
 
 if __name__ == "__main__":

examples/summarization/bart/test_bart_examples.py

@@ -25,7 +25,8 @@ class TestBartExamples(unittest.TestCase):
         tmp = Path(tempfile.gettempdir()) / "utest_generations_bart_sum.hypo"
         with tmp.open("w") as f:
             f.write("\n".join(articles))
-        testargs = ["evaluate_cnn.py", str(tmp), output_file_name]
+
+        testargs = ["evaluate_cnn.py", str(tmp), output_file_name, "sshleifer/bart-tiny-random"]
         with patch.object(sys, "argv", testargs):
             _run_generate()
             self.assertTrue(Path(output_file_name).exists())

tests/test_modeling_bart.py

@@ -27,7 +27,9 @@ from .utils import CACHE_DIR, require_torch, slow, torch_device
 if is_torch_available():
     import torch
     from transformers import (
+        AutoModel,
         AutoModelForSequenceClassification,
+        AutoTokenizer,
         BartModel,
         BartForConditionalGeneration,
         BartForSequenceClassification,
@@ -183,6 +185,15 @@ class BARTModelTest(ModelTesterMixin, unittest.TestCase):
     def test_inputs_embeds(self):
         pass
 
+    def test_tiny_model(self):
+        model_name = "sshleifer/bart-tiny-random"
+        tiny = AutoModel.from_pretrained(model_name)  # same vocab size
+        tok = AutoTokenizer.from_pretrained(model_name)  # same tokenizer
+        inputs_dict = tok.batch_encode_plus(["Hello my friends"], return_tensors="pt")
+
+        with torch.no_grad():
+            tiny(**inputs_dict)
+
 
 @require_torch
 class BartHeadTests(unittest.TestCase):