fix: correct batch tokenization when sequence exceeds encoder length (#…

…352)

Signed-off-by: George Armstrong <[email protected]>
Co-authored-by: pre-commit-ci[bot] <66853113+pre-commit-ci[bot]@users.noreply.github.com>

nemo_aligner/utils/text_generation_utils.py

@@ -92,10 +92,10 @@ def tokenize(sentence):
         return output
 
     context_tokens = list(map(tokenize, sentences))
+    context_tokens = [x[:max_len] for x in context_tokens]
     max_sequence_length = max(len(x) for x in context_tokens)
 
     context_tokens, context_lengths = pad_batch(context_tokens, tokenizer.eos_id, max_len - max_sequence_length)
-    context_tokens = [x[:max_len] for x in context_tokens]
     context_tokens_tensor = torch.cuda.LongTensor(context_tokens)
     context_length_tensor = torch.cuda.LongTensor(context_lengths)
     return context_tokens_tensor, context_length_tensor

tests/test_text_generation_utils.py

@@ -0,0 +1,69 @@
+# Copyright (c) 2023, NVIDIA CORPORATION.  All rights reserved.
+#
+# Licensed under the Apache License, Version 2.0 (the "License");
+# you may not use this file except in compliance with the License.
+# You may obtain a copy of the License at
+#
+#     http://www.apache.org/licenses/LICENSE-2.0
+#
+# Unless required by applicable law or agreed to in writing, software
+# distributed under the License is distributed on an "AS IS" BASIS,
+# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+# See the License for the specific language governing permissions and
+# limitations under the License.
+
+from nemo_aligner.utils.text_generation_utils import tokenize_batch
+
+
+class MockTokenizer:
+    def __init__(self):
+        self.vocab = dict()
+        self.bos_id = 0
+        self.eos_id = 1
+        self.vocab["<bos>"] = self.bos_id
+        self.vocab["<eos>"] = self.eos_id
+
+    def text_to_ids(self, text):
+        tokens = list(text)
+        ids = [self.vocab.get(token, len(self.vocab)) for token in tokens]
+        return ids
+
+
+def test_tokenize_batch():
+    sentences = ["I went to the store.", "I bought a zoo."]
+    tokenizer = MockTokenizer()
+    max_len = 30
+    context_tokens_tensor, context_length_tensor = tokenize_batch(
+        sentences, tokenizer, max_len, add_BOS=False, add_EOS=False
+    )
+    assert context_tokens_tensor.shape == (
+        2,
+        30,
+    ), f"expected context_tokens_tensor shape to be (2, 30) but got {context_tokens_tensor.shape}"
+    assert context_length_tensor.shape == (
+        2,
+    ), f"expected context_length_tensor shape to be (2,) but got {context_length_tensor.shape}"
+    assert context_length_tensor.tolist() == [
+        20,
+        15,
+    ], f"expected context_length_tensor to be [20, 15] but got {context_length_tensor.tolist()}"
+
+
+def test_tokenize_batch_with_sentence_longer_than_max_len():
+    sentences = ["I went to the store.", "I bought a zoo."]
+    tokenizer = MockTokenizer()
+    max_len = 10
+    context_tokens_tensor, context_length_tensor = tokenize_batch(
+        sentences, tokenizer, max_len, add_BOS=False, add_EOS=False
+    )
+    assert context_tokens_tensor.shape == (
+        2,
+        10,
+    ), f"expected context_tokens_tensor shape to be (2, 10) but got {context_tokens_tensor.shape}"
+    assert context_length_tensor.shape == (
+        2,
+    ), f"expected context_length_tensor shape to be (2,) but got {context_length_tensor.shape}"
+    assert context_length_tensor.tolist() == [
+        10,
+        10,
+    ], f"expected context_length_tensor to be [10, 10] but got {context_length_tensor.tolist()}"