Add EmbeddingInfo NamedTuple (#392)

Co-authored-by: Ben Shapira <[email protected]>

fuse/data/tokenizers/modular_tokenizer/inject_utils.py

@@ -1,6 +1,6 @@
 import re
 from collections import defaultdict
-from typing import Dict, List, Optional, Tuple, Union
+from typing import Dict, List, NamedTuple, Optional, Tuple, Union
 from warnings import warn
 
 import torch
@@ -13,6 +13,11 @@
 from fuse.utils import NDict
 
 
+class EmbeddingInfo(NamedTuple):
+    location: int
+    embedding_input: str
+
+
 class InjectorToModularTokenizerLib:
     """
     InjectorTokenizer builds on top of ModularTokenizer.
@@ -227,7 +232,10 @@ def build_scalars_and_embeddings(
 
                     embedding_input = sample_dict[embeddings_key]
                     external_embeddings_info[embedding_model_name].append(
-                        (num_tokens_token_so_far, embedding_input)
+                        EmbeddingInfo(
+                            location=num_tokens_token_so_far,
+                            embedding_input=embedding_input,
+                        )
                     )
                 elif tokenizer_name.startswith("VECTORS_"):
                     raise NotImplementedError