diff --git a/docs/source/api_ref_modules.rst b/docs/source/api_ref_modules.rst
index 5eb8fff358..979e57347f 100644
--- a/docs/source/api_ref_modules.rst
+++ b/docs/source/api_ref_modules.rst
@@ -48,10 +48,10 @@ model specific tokenizers.
     :toctree: generated/
     :nosignatures:
 
-    tokenizers.SentencePieceBaseTokenizer
-    tokenizers.TikTokenBaseTokenizer
-    tokenizers.ModelTokenizer
-    tokenizers.BaseTokenizer
+    transforms.tokenizers.SentencePieceBaseTokenizer
+    transforms.tokenizers.TikTokenBaseTokenizer
+    transforms.tokenizers.ModelTokenizer
+    transforms.tokenizers.BaseTokenizer
 
 Tokenizer Utilities
 -------------------
@@ -61,8 +61,8 @@ These are helper methods that can be used by any tokenizer.
     :toctree: generated/
     :nosignatures:
 
-    tokenizers.tokenize_messages_no_special_tokens
-    tokenizers.parse_hf_tokenizer_json
+    transforms.tokenizers.tokenize_messages_no_special_tokens
+    transforms.tokenizers.parse_hf_tokenizer_json
 
 
 PEFT Components
diff --git a/docs/source/basics/custom_components.rst b/docs/source/basics/custom_components.rst
index f252cb197e..0f742644dc 100644
--- a/docs/source/basics/custom_components.rst
+++ b/docs/source/basics/custom_components.rst
@@ -117,7 +117,7 @@ our models in torchtune - see :func:`~torchtune.models.llama3_2_vision.llama3_2_
     #
     from torchtune.datasets import SFTDataset, PackedDataset
     from torchtune.data import InputOutputToMessages
-    from torchtune.modules.tokenizers import ModelTokenizer
+    from torchtune.modules.transforms.tokenizers import ModelTokenizer
 
     # Example builder function for a custom code instruct dataset not in torchtune, but using
     # different dataset building blocks from torchtune
diff --git a/docs/source/basics/model_transforms.rst b/docs/source/basics/model_transforms.rst
index c10cb1abd8..71e7e08bd5 100644
--- a/docs/source/basics/model_transforms.rst
+++ b/docs/source/basics/model_transforms.rst
@@ -101,7 +101,7 @@ The following methods are required on the model transform:
 
 .. code-block:: python
 
-    from torchtune.modules.tokenizers import ModelTokenizer
+    from torchtune.modules.transforms.tokenizers import ModelTokenizer
     from torchtune.modules.transforms import Transform
 
     class MyMultimodalTransform(ModelTokenizer, Transform):
diff --git a/docs/source/basics/tokenizers.rst b/docs/source/basics/tokenizers.rst
index d637961c54..47be88fe0c 100644
--- a/docs/source/basics/tokenizers.rst
+++ b/docs/source/basics/tokenizers.rst
@@ -168,7 +168,7 @@ For example, here we change the ``"<|begin_of_text|>"`` and ``"<|end_of_text|>"`
 Base tokenizers
 ---------------
 
-:class:`~torchtune.modules.tokenizers.BaseTokenizer` are the underlying byte-pair encoding modules that perform the actual raw string to token ID conversion and back.
+:class:`~torchtune.modules.transforms.tokenizers.BaseTokenizer` are the underlying byte-pair encoding modules that perform the actual raw string to token ID conversion and back.
 In torchtune, they are required to implement ``encode`` and ``decode`` methods, which are called by the :ref:`model_tokenizers` to convert
 between raw text and token IDs.
 
@@ -202,13 +202,13 @@ between raw text and token IDs.
             """
             pass
 
-If you load any :ref:`model_tokenizers`, you can see that it calls its underlying :class:`~torchtune.modules.tokenizers.BaseTokenizer`
+If you load any :ref:`model_tokenizers`, you can see that it calls its underlying :class:`~torchtune.modules.transforms.tokenizers.BaseTokenizer`
 to do the actual encoding and decoding.
 
 .. code-block:: python
 
     from torchtune.models.mistral import mistral_tokenizer
-    from torchtune.modules.tokenizers import SentencePieceBaseTokenizer
+    from torchtune.modules.transforms.tokenizers import SentencePieceBaseTokenizer
 
     m_tokenizer = mistral_tokenizer("/tmp/Mistral-7B-v0.1/tokenizer.model")
     # Mistral uses SentencePiece for its underlying BPE
@@ -227,7 +227,7 @@ to do the actual encoding and decoding.
 Model tokenizers
 ----------------
 
-:class:`~torchtune.modules.tokenizers.ModelTokenizer` are specific to a particular model. They are required to implement the ``tokenize_messages`` method,
+:class:`~torchtune.modules.transforms.tokenizers.ModelTokenizer` are specific to a particular model. They are required to implement the ``tokenize_messages`` method,
 which converts a list of Messages into a list of token IDs.
 
 .. code-block:: python
@@ -259,7 +259,7 @@ is because they add all the necessary special tokens or prompt templates require
 .. code-block:: python
 
     from torchtune.models.mistral import mistral_tokenizer
-    from torchtune.modules.tokenizers import SentencePieceBaseTokenizer
+    from torchtune.modules.transforms.tokenizers import SentencePieceBaseTokenizer
     from torchtune.data import Message
 
     m_tokenizer = mistral_tokenizer("/tmp/Mistral-7B-v0.1/tokenizer.model")
diff --git a/recipes/eleuther_eval.py b/recipes/eleuther_eval.py
index 97744fb9b8..fd1ac8f6e1 100644
--- a/recipes/eleuther_eval.py
+++ b/recipes/eleuther_eval.py
@@ -31,8 +31,8 @@
 from torchtune.modules import TransformerDecoder
 from torchtune.modules.common_utils import local_kv_cache
 from torchtune.modules.model_fusion import DeepFusionModel
-from torchtune.modules.tokenizers import ModelTokenizer
 from torchtune.modules.transforms import Transform
+from torchtune.modules.transforms.tokenizers import ModelTokenizer
 from torchtune.recipe_interfaces import EvalRecipeInterface
 from torchtune.training import FullModelTorchTuneCheckpointer
 
diff --git a/tests/test_utils.py b/tests/test_utils.py
index 6497539869..ca28029710 100644
--- a/tests/test_utils.py
+++ b/tests/test_utils.py
@@ -20,8 +20,8 @@
 import torch
 from torch import nn
 from torchtune.data import Message, PromptTemplate, truncate
-from torchtune.modules.tokenizers import ModelTokenizer
 from torchtune.modules.transforms import Transform
+from torchtune.modules.transforms.tokenizers import ModelTokenizer
 
 skip_if_cuda_not_available = unittest.skipIf(
     not torch.cuda.is_available(), "CUDA is not available"
diff --git a/tests/torchtune/modules/tokenizers/test_sentencepiece.py b/tests/torchtune/modules/transforms/tokenizers/test_sentencepiece.py
similarity index 97%
rename from tests/torchtune/modules/tokenizers/test_sentencepiece.py
rename to tests/torchtune/modules/transforms/tokenizers/test_sentencepiece.py
index d11c1b9c52..217f0bf2d8 100644
--- a/tests/torchtune/modules/tokenizers/test_sentencepiece.py
+++ b/tests/torchtune/modules/transforms/tokenizers/test_sentencepiece.py
@@ -7,7 +7,7 @@
 import pytest
 
 from tests.common import ASSETS
-from torchtune.modules.tokenizers import SentencePieceBaseTokenizer
+from torchtune.modules.transforms.tokenizers import SentencePieceBaseTokenizer
 
 
 class TestSentencePieceBaseTokenizer:
diff --git a/tests/torchtune/modules/tokenizers/test_tiktoken.py b/tests/torchtune/modules/transforms/tokenizers/test_tiktoken.py
similarity index 98%
rename from tests/torchtune/modules/tokenizers/test_tiktoken.py
rename to tests/torchtune/modules/transforms/tokenizers/test_tiktoken.py
index e7e69f62d3..5d3608d4bd 100644
--- a/tests/torchtune/modules/tokenizers/test_tiktoken.py
+++ b/tests/torchtune/modules/transforms/tokenizers/test_tiktoken.py
@@ -8,7 +8,7 @@
 
 from tests.common import ASSETS
 from torchtune.models.llama3._tokenizer import CL100K_PATTERN
-from torchtune.modules.tokenizers import TikTokenBaseTokenizer
+from torchtune.modules.transforms.tokenizers import TikTokenBaseTokenizer
 
 
 class TestTikTokenBaseTokenizer:
diff --git a/tests/torchtune/modules/tokenizers/test_utils.py b/tests/torchtune/modules/transforms/tokenizers/test_utils.py
similarity index 94%
rename from tests/torchtune/modules/tokenizers/test_utils.py
rename to tests/torchtune/modules/transforms/tokenizers/test_utils.py
index 2c49d82a5a..e3a11e6f36 100644
--- a/tests/torchtune/modules/tokenizers/test_utils.py
+++ b/tests/torchtune/modules/transforms/tokenizers/test_utils.py
@@ -9,7 +9,7 @@
 from tests.test_utils import DummyTokenizer
 from torchtune.data import Message
 
-from torchtune.modules.tokenizers import tokenize_messages_no_special_tokens
+from torchtune.modules.transforms.tokenizers import tokenize_messages_no_special_tokens
 
 
 class TestTokenizerUtils:
diff --git a/torchtune/data/_messages.py b/torchtune/data/_messages.py
index a4e00834c2..170970e5c5 100644
--- a/torchtune/data/_messages.py
+++ b/torchtune/data/_messages.py
@@ -22,9 +22,10 @@
 class Message:
     """
     This class represents individual messages in a fine-tuning dataset. It supports
-    text-only content, text with interleaved images, and tool calls. The :class:`~torchtune.modules.tokenizers.ModelTokenizer`
-    will tokenize the content of the message using ``tokenize_messages`` and attach
-    the appropriate special tokens based on the flags set in this class.
+    text-only content, text with interleaved images, and tool calls. The
+    :class:`~torchtune.modules.transforms.tokenizers.ModelTokenizer` will tokenize
+    the content of the message using ``tokenize_messages`` and attach the appropriate
+    special tokens based on the flags set in this class.
 
     Args:
         role (Role): role of the message writer. Can be "system" for system prompts,
diff --git a/torchtune/datasets/_alpaca.py b/torchtune/datasets/_alpaca.py
index a881c149b0..c7795c8f28 100644
--- a/torchtune/datasets/_alpaca.py
+++ b/torchtune/datasets/_alpaca.py
@@ -12,7 +12,7 @@
 
 from torchtune.datasets._packed import PackedDataset
 from torchtune.datasets._sft import SFTDataset
-from torchtune.modules.tokenizers import ModelTokenizer
+from torchtune.modules.transforms.tokenizers import ModelTokenizer
 
 
 def alpaca_dataset(
diff --git a/torchtune/datasets/_chat.py b/torchtune/datasets/_chat.py
index f126fb3979..1e3962e14b 100644
--- a/torchtune/datasets/_chat.py
+++ b/torchtune/datasets/_chat.py
@@ -9,7 +9,7 @@
 from torchtune.data._messages import OpenAIToMessages, ShareGPTToMessages
 from torchtune.datasets._packed import PackedDataset
 from torchtune.datasets._sft import SFTDataset
-from torchtune.modules.tokenizers import ModelTokenizer
+from torchtune.modules.transforms.tokenizers import ModelTokenizer
 
 
 def chat_dataset(
diff --git a/torchtune/datasets/_cnn_dailymail.py b/torchtune/datasets/_cnn_dailymail.py
index d3c3af1f93..3995d46b22 100644
--- a/torchtune/datasets/_cnn_dailymail.py
+++ b/torchtune/datasets/_cnn_dailymail.py
@@ -8,7 +8,7 @@
 
 from torchtune.datasets._text_completion import TextCompletionDataset
 
-from torchtune.modules.tokenizers import ModelTokenizer
+from torchtune.modules.transforms.tokenizers import ModelTokenizer
 
 
 def cnn_dailymail_articles_dataset(
diff --git a/torchtune/datasets/_grammar.py b/torchtune/datasets/_grammar.py
index 9e9d700ea6..02970cedef 100644
--- a/torchtune/datasets/_grammar.py
+++ b/torchtune/datasets/_grammar.py
@@ -10,7 +10,7 @@
 from torchtune.data import InputOutputToMessages
 from torchtune.datasets._packed import PackedDataset
 from torchtune.datasets._sft import SFTDataset
-from torchtune.modules.tokenizers import ModelTokenizer
+from torchtune.modules.transforms.tokenizers import ModelTokenizer
 
 
 def grammar_dataset(
diff --git a/torchtune/datasets/_hh_rlhf_helpful.py b/torchtune/datasets/_hh_rlhf_helpful.py
index e466a8a4fd..8eea7e1a46 100644
--- a/torchtune/datasets/_hh_rlhf_helpful.py
+++ b/torchtune/datasets/_hh_rlhf_helpful.py
@@ -8,7 +8,7 @@
 
 from torchtune.data import ChosenRejectedToMessages
 from torchtune.datasets._preference import PreferenceDataset
-from torchtune.modules.tokenizers import ModelTokenizer
+from torchtune.modules.transforms.tokenizers import ModelTokenizer
 
 
 def hh_rlhf_helpful_dataset(
diff --git a/torchtune/datasets/_instruct.py b/torchtune/datasets/_instruct.py
index 0dfa46146d..20168aac1d 100644
--- a/torchtune/datasets/_instruct.py
+++ b/torchtune/datasets/_instruct.py
@@ -9,7 +9,7 @@
 from torchtune.data import InputOutputToMessages
 from torchtune.datasets._packed import PackedDataset
 from torchtune.datasets._sft import SFTDataset
-from torchtune.modules.tokenizers import ModelTokenizer
+from torchtune.modules.transforms.tokenizers import ModelTokenizer
 
 
 def instruct_dataset(
diff --git a/torchtune/datasets/_preference.py b/torchtune/datasets/_preference.py
index dea4eec852..c9615fe93c 100644
--- a/torchtune/datasets/_preference.py
+++ b/torchtune/datasets/_preference.py
@@ -11,10 +11,10 @@
 from torch.utils.data import Dataset
 
 from torchtune.data import ChosenRejectedToMessages, CROSS_ENTROPY_IGNORE_IDX
-
-from torchtune.modules.tokenizers import ModelTokenizer
 from torchtune.modules.transforms import Transform
 
+from torchtune.modules.transforms.tokenizers import ModelTokenizer
+
 
 class PreferenceDataset(Dataset):
     """
@@ -84,7 +84,7 @@ class requires the dataset to have "chosen" and "rejected" model responses. Thes
             of messages are stored in the ``"chosen"`` and ``"rejected"`` keys.
         tokenizer (ModelTokenizer): Tokenizer used by the model that implements the ``tokenize_messages`` method.
             Since PreferenceDataset only supports text data, it requires a
-            :class:`~torchtune.modules.tokenizers.ModelTokenizer` instead of the ``model_transform`` in
+            :class:`~torchtune.modules.transforms.tokenizers.ModelTokenizer` instead of the ``model_transform`` in
             :class:`~torchtune.datasets.SFTDataset`.
         filter_fn (Optional[Callable]): callable used to filter the dataset prior to any pre-processing. See
             the Hugging Face `docs <https://huggingface.co/docs/datasets/v2.20.0/process#select-and-filter>`_ for more
diff --git a/torchtune/datasets/_samsum.py b/torchtune/datasets/_samsum.py
index 905911d736..bd7f7dd8eb 100644
--- a/torchtune/datasets/_samsum.py
+++ b/torchtune/datasets/_samsum.py
@@ -10,7 +10,7 @@
 from torchtune.data import InputOutputToMessages
 from torchtune.datasets._packed import PackedDataset
 from torchtune.datasets._sft import SFTDataset
-from torchtune.modules.tokenizers import ModelTokenizer
+from torchtune.modules.transforms.tokenizers import ModelTokenizer
 
 
 def samsum_dataset(
diff --git a/torchtune/datasets/_sft.py b/torchtune/datasets/_sft.py
index 9ee11244b6..0d1461dd0d 100644
--- a/torchtune/datasets/_sft.py
+++ b/torchtune/datasets/_sft.py
@@ -69,11 +69,13 @@ class SFTDataset(Dataset):
     multimodal datasets requires processing the images in a way specific to the vision
     encoder being used by the model and is agnostic to the specific dataset.
 
-    Tokenization is handled by the ``model_transform``. All :class:`~torchtune.modules.tokenizers.ModelTokenizer`
-    can be treated as a ``model_transform`` since it uses the model-specific tokenizer to
-    transform the list of messages outputted from the ``message_transform`` into tokens
-    used by the model for training. Text-only datasets will simply pass the :class:`~torchtune.modules.tokenizers.ModelTokenizer`
-    into ``model_transform``. Tokenizers handle prompt templating, if configured.
+    Tokenization is handled by the ``model_transform``. All
+    :class:`~torchtune.modules.transforms.tokenizers.ModelTokenizer` can be treated as
+    a ``model_transform`` since it uses the model-specific tokenizer to transform the
+    list of messages outputted from the ``message_transform`` into tokens used by the
+    model for training. Text-only datasets will simply pass the
+    :class:`~torchtune.modules.transforms.tokenizers.ModelTokenizer` into ``model_transform``.
+    Tokenizers handle prompt templating, if configured.
 
     Args:
         source (str): path to dataset repository on Hugging Face. For local datasets,
diff --git a/torchtune/datasets/_slimorca.py b/torchtune/datasets/_slimorca.py
index 126b6b92e4..2701b2d717 100644
--- a/torchtune/datasets/_slimorca.py
+++ b/torchtune/datasets/_slimorca.py
@@ -10,7 +10,7 @@
 from torchtune.datasets._packed import PackedDataset
 
 from torchtune.datasets._sft import SFTDataset
-from torchtune.modules.tokenizers import ModelTokenizer
+from torchtune.modules.transforms.tokenizers import ModelTokenizer
 
 
 def slimorca_dataset(
diff --git a/torchtune/datasets/_stack_exchange_paired.py b/torchtune/datasets/_stack_exchange_paired.py
index 09eda929fe..a111d415d2 100644
--- a/torchtune/datasets/_stack_exchange_paired.py
+++ b/torchtune/datasets/_stack_exchange_paired.py
@@ -8,8 +8,8 @@
 
 from torchtune.data import Message
 from torchtune.datasets._preference import PreferenceDataset
-from torchtune.modules.tokenizers import ModelTokenizer
 from torchtune.modules.transforms import Transform
+from torchtune.modules.transforms.tokenizers import ModelTokenizer
 
 
 class StackExchangePairedToMessages(Transform):
diff --git a/torchtune/datasets/_text_completion.py b/torchtune/datasets/_text_completion.py
index 5b5cc94299..342c6aa816 100644
--- a/torchtune/datasets/_text_completion.py
+++ b/torchtune/datasets/_text_completion.py
@@ -10,7 +10,7 @@
 from torch.utils.data import Dataset
 from torchtune.data._utils import truncate
 from torchtune.datasets._packed import PackedDataset
-from torchtune.modules.tokenizers import ModelTokenizer
+from torchtune.modules.transforms.tokenizers import ModelTokenizer
 
 
 class TextCompletionDataset(Dataset):
diff --git a/torchtune/datasets/_wikitext.py b/torchtune/datasets/_wikitext.py
index 01111a25c6..4f9ada6741 100644
--- a/torchtune/datasets/_wikitext.py
+++ b/torchtune/datasets/_wikitext.py
@@ -13,7 +13,7 @@
     TextCompletionDataset,
 )
 
-from torchtune.modules.tokenizers import ModelTokenizer
+from torchtune.modules.transforms.tokenizers import ModelTokenizer
 
 
 def wikitext_dataset(
diff --git a/torchtune/models/clip/_tokenizer.py b/torchtune/models/clip/_tokenizer.py
index 69fed32c72..cdab2c9c05 100644
--- a/torchtune/models/clip/_tokenizer.py
+++ b/torchtune/models/clip/_tokenizer.py
@@ -7,7 +7,7 @@
 
 import regex as re
 
-from torchtune.modules.tokenizers._utils import BaseTokenizer
+from torchtune.modules.transforms.tokenizers._utils import BaseTokenizer
 
 WORD_BOUNDARY = "</w>"
 
diff --git a/torchtune/models/gemma/_tokenizer.py b/torchtune/models/gemma/_tokenizer.py
index e5eb89e230..dc5d2eadf8 100644
--- a/torchtune/models/gemma/_tokenizer.py
+++ b/torchtune/models/gemma/_tokenizer.py
@@ -7,12 +7,12 @@
 from typing import Any, List, Mapping, Optional, Tuple
 
 from torchtune.data import Message, PromptTemplate
-from torchtune.modules.tokenizers import (
+from torchtune.modules.transforms import Transform
+from torchtune.modules.transforms.tokenizers import (
     ModelTokenizer,
     SentencePieceBaseTokenizer,
     tokenize_messages_no_special_tokens,
 )
-from torchtune.modules.transforms import Transform
 
 WHITESPACE_CHARS = [" ", "\n", "\t", "\r", "\v"]
 
diff --git a/torchtune/models/llama2/_tokenizer.py b/torchtune/models/llama2/_tokenizer.py
index 078494c531..4e2ab6a40c 100644
--- a/torchtune/models/llama2/_tokenizer.py
+++ b/torchtune/models/llama2/_tokenizer.py
@@ -8,12 +8,12 @@
 
 from torchtune.data import Message, PromptTemplate
 from torchtune.models.llama2._prompt_template import Llama2ChatTemplate
-from torchtune.modules.tokenizers import (
+from torchtune.modules.transforms import Transform
+from torchtune.modules.transforms.tokenizers import (
     ModelTokenizer,
     SentencePieceBaseTokenizer,
     tokenize_messages_no_special_tokens,
 )
-from torchtune.modules.transforms import Transform
 
 WHITESPACE_CHARS = [" ", "\n", "\t", "\r", "\v"]
 
diff --git a/torchtune/models/llama3/_model_builders.py b/torchtune/models/llama3/_model_builders.py
index 0ddca90189..6c13e37cff 100644
--- a/torchtune/models/llama3/_model_builders.py
+++ b/torchtune/models/llama3/_model_builders.py
@@ -13,7 +13,7 @@
 
 from torchtune.modules import TransformerDecoder
 from torchtune.modules.peft import LORA_ATTN_MODULES
-from torchtune.modules.tokenizers import parse_hf_tokenizer_json
+from torchtune.modules.transforms.tokenizers import parse_hf_tokenizer_json
 
 
 """
diff --git a/torchtune/models/llama3/_tokenizer.py b/torchtune/models/llama3/_tokenizer.py
index 50ea0a7581..012aa9f584 100644
--- a/torchtune/models/llama3/_tokenizer.py
+++ b/torchtune/models/llama3/_tokenizer.py
@@ -8,8 +8,11 @@
 from typing import Any, Dict, List, Mapping, Optional, Tuple
 
 from torchtune.data import Message, PromptTemplate, truncate
-from torchtune.modules.tokenizers import ModelTokenizer, TikTokenBaseTokenizer
 from torchtune.modules.transforms import Transform
+from torchtune.modules.transforms.tokenizers import (
+    ModelTokenizer,
+    TikTokenBaseTokenizer,
+)
 
 
 CL100K_PATTERN = r"""(?i:'s|'t|'re|'ve|'m|'ll|'d)|[^\r\n\p{L}\p{N}]?\p{L}+|\p{N}{1,3}| ?[^\s\p{L}\p{N}]+[\r\n]*|\s*[\r\n]+|\s+(?!\S)|\s+"""  # noqa
diff --git a/torchtune/models/llama3_2_vision/_transform.py b/torchtune/models/llama3_2_vision/_transform.py
index eaf627d027..534ed4ab1c 100644
--- a/torchtune/models/llama3_2_vision/_transform.py
+++ b/torchtune/models/llama3_2_vision/_transform.py
@@ -10,8 +10,8 @@
 
 from torchtune.models.clip import CLIPImageTransform
 from torchtune.models.llama3 import llama3_tokenizer
-from torchtune.modules.tokenizers import ModelTokenizer
 from torchtune.modules.transforms import Transform, VisionCrossAttentionMask
+from torchtune.modules.transforms.tokenizers import ModelTokenizer
 
 
 class Llama3VisionTransform(ModelTokenizer, Transform):
diff --git a/torchtune/models/mistral/_tokenizer.py b/torchtune/models/mistral/_tokenizer.py
index c3bbc8a4a7..49617220c3 100644
--- a/torchtune/models/mistral/_tokenizer.py
+++ b/torchtune/models/mistral/_tokenizer.py
@@ -8,12 +8,12 @@
 
 from torchtune.data import Message, PromptTemplate
 from torchtune.models.mistral._prompt_template import MistralChatTemplate
-from torchtune.modules.tokenizers import (
+from torchtune.modules.transforms import Transform
+from torchtune.modules.transforms.tokenizers import (
     ModelTokenizer,
     SentencePieceBaseTokenizer,
     tokenize_messages_no_special_tokens,
 )
-from torchtune.modules.transforms import Transform
 
 WHITESPACE_CHARS = [" ", "\n", "\t", "\r", "\v"]
 
diff --git a/torchtune/models/phi3/_model_builders.py b/torchtune/models/phi3/_model_builders.py
index 91d42623d7..e1275df783 100644
--- a/torchtune/models/phi3/_model_builders.py
+++ b/torchtune/models/phi3/_model_builders.py
@@ -6,7 +6,7 @@
 from torchtune.modules import TransformerDecoder
 from torchtune.modules.peft import LORA_ATTN_MODULES
 from functools import partial
-from torchtune.modules.tokenizers import parse_hf_tokenizer_json
+from torchtune.modules.transforms.tokenizers import parse_hf_tokenizer_json
 from torchtune.data._prompt_templates import _TemplateType
 from torchtune.data._prompt_templates import _get_prompt_template
 
diff --git a/torchtune/models/phi3/_tokenizer.py b/torchtune/models/phi3/_tokenizer.py
index 38707bf26e..44f66b5934 100644
--- a/torchtune/models/phi3/_tokenizer.py
+++ b/torchtune/models/phi3/_tokenizer.py
@@ -9,8 +9,11 @@
 from torchtune.data._messages import Message
 from torchtune.data._prompt_templates import PromptTemplate
 from torchtune.data._utils import truncate
-from torchtune.modules.tokenizers import ModelTokenizer, SentencePieceBaseTokenizer
 from torchtune.modules.transforms import Transform
+from torchtune.modules.transforms.tokenizers import (
+    ModelTokenizer,
+    SentencePieceBaseTokenizer,
+)
 
 PHI3_SPECIAL_TOKENS = {
     "<|endoftext|>": 32000,
diff --git a/torchtune/models/qwen2/_model_builders.py b/torchtune/models/qwen2/_model_builders.py
index 2a0ee06f83..f1ca5b8506 100644
--- a/torchtune/models/qwen2/_model_builders.py
+++ b/torchtune/models/qwen2/_model_builders.py
@@ -11,7 +11,7 @@
 from torchtune.models.qwen2._tokenizer import QWEN2_SPECIAL_TOKENS, Qwen2Tokenizer
 from torchtune.modules import TransformerDecoder
 from torchtune.modules.peft import LORA_ATTN_MODULES
-from torchtune.modules.tokenizers import parse_hf_tokenizer_json
+from torchtune.modules.transforms.tokenizers import parse_hf_tokenizer_json
 
 """
 Model builders build specific instantiations using component builders. For example
diff --git a/torchtune/models/qwen2/_tokenizer.py b/torchtune/models/qwen2/_tokenizer.py
index 0e4ee6bd35..dd6d038003 100644
--- a/torchtune/models/qwen2/_tokenizer.py
+++ b/torchtune/models/qwen2/_tokenizer.py
@@ -11,7 +11,7 @@
 import regex as re
 
 from torchtune.data import ChatMLTemplate, Message, PromptTemplate, truncate
-from torchtune.modules.tokenizers import ModelTokenizer
+from torchtune.modules.transforms.tokenizers import ModelTokenizer
 
 PRETOKENIZE_REGEX = (
     r"(?i:'s|'t|'re|'ve|'m|'ll|'d)|"
diff --git a/torchtune/models/qwen2_5/_model_builders.py b/torchtune/models/qwen2_5/_model_builders.py
index 7d39802375..716ae48329 100644
--- a/torchtune/models/qwen2_5/_model_builders.py
+++ b/torchtune/models/qwen2_5/_model_builders.py
@@ -11,7 +11,7 @@
 from torchtune.models.qwen2_5._tokenizer import QWEN2_5_SPECIAL_TOKENS, Qwen2_5Tokenizer
 from torchtune.modules import TransformerDecoder
 from torchtune.modules.peft import LORA_ATTN_MODULES
-from torchtune.modules.tokenizers import parse_hf_tokenizer_json
+from torchtune.modules.transforms.tokenizers import parse_hf_tokenizer_json
 
 """
 Model builders build specific instantiations using component builders. For example
diff --git a/torchtune/models/t5/_tokenizer.py b/torchtune/models/t5/_tokenizer.py
index f89dff00f6..e4fa9c539e 100644
--- a/torchtune/models/t5/_tokenizer.py
+++ b/torchtune/models/t5/_tokenizer.py
@@ -5,7 +5,9 @@
 # LICENSE file in the root directory of this source tree.
 from typing import Any, Dict, List
 
-from torchtune.modules.tokenizers._sentencepiece import SentencePieceBaseTokenizer
+from torchtune.modules.transforms.tokenizers._sentencepiece import (
+    SentencePieceBaseTokenizer,
+)
 
 
 class T5Tokenizer(SentencePieceBaseTokenizer):
diff --git a/torchtune/modules/tokenizers/__init__.py b/torchtune/modules/tokenizers/__init__.py
index 2fecc279ee..f10a9b3dd6 100644
--- a/torchtune/modules/tokenizers/__init__.py
+++ b/torchtune/modules/tokenizers/__init__.py
@@ -4,20 +4,28 @@
 # This source code is licensed under the BSD-style license found in the
 # LICENSE file in the root directory of this source tree.
 
-from ._sentencepiece import SentencePieceBaseTokenizer
-from ._tiktoken import TikTokenBaseTokenizer
-from ._utils import (
+# flake8: noqa: F401
+
+# NOTE: This file is maintained for backward compatibility purposes.
+# The imports below point to the new location in `torchtune.modules.transforms.tokenizers`.
+# The import paths will be removed in v0.7. Please update your code to use the new path
+# (torchtune.modules.transforms.tokenizers) to avoid breaking changes in future releases.
+
+
+import warnings
+
+from torchtune.modules.transforms.tokenizers import (
     BaseTokenizer,
     ModelTokenizer,
     parse_hf_tokenizer_json,
+    SentencePieceBaseTokenizer,
+    TikTokenBaseTokenizer,
     tokenize_messages_no_special_tokens,
 )
 
-__all__ = [
-    "SentencePieceBaseTokenizer",
-    "TikTokenBaseTokenizer",
-    "ModelTokenizer",
-    "BaseTokenizer",
-    "tokenize_messages_no_special_tokens",
-    "parse_hf_tokenizer_json",
-]
+warnings.warn(
+    "The import path 'torchtune.modules.tokenizers' is deprecated and will be removed in v0.7. "
+    "Please update your imports to 'torchtune.modules.transforms.tokenizers'.",
+    DeprecationWarning,
+    stacklevel=2,
+)
diff --git a/torchtune/modules/transforms/tokenizers/__init__.py b/torchtune/modules/transforms/tokenizers/__init__.py
new file mode 100644
index 0000000000..2fecc279ee
--- /dev/null
+++ b/torchtune/modules/transforms/tokenizers/__init__.py
@@ -0,0 +1,23 @@
+# Copyright (c) Meta Platforms, Inc. and affiliates.
+# All rights reserved.
+#
+# This source code is licensed under the BSD-style license found in the
+# LICENSE file in the root directory of this source tree.
+
+from ._sentencepiece import SentencePieceBaseTokenizer
+from ._tiktoken import TikTokenBaseTokenizer
+from ._utils import (
+    BaseTokenizer,
+    ModelTokenizer,
+    parse_hf_tokenizer_json,
+    tokenize_messages_no_special_tokens,
+)
+
+__all__ = [
+    "SentencePieceBaseTokenizer",
+    "TikTokenBaseTokenizer",
+    "ModelTokenizer",
+    "BaseTokenizer",
+    "tokenize_messages_no_special_tokens",
+    "parse_hf_tokenizer_json",
+]
diff --git a/torchtune/modules/tokenizers/_sentencepiece.py b/torchtune/modules/transforms/tokenizers/_sentencepiece.py
similarity index 98%
rename from torchtune/modules/tokenizers/_sentencepiece.py
rename to torchtune/modules/transforms/tokenizers/_sentencepiece.py
index 0b22b63ee3..8d98617378 100644
--- a/torchtune/modules/tokenizers/_sentencepiece.py
+++ b/torchtune/modules/transforms/tokenizers/_sentencepiece.py
@@ -7,8 +7,7 @@
 from typing import List, Optional
 
 from sentencepiece import SentencePieceProcessor
-
-from torchtune.modules.tokenizers._utils import BaseTokenizer
+from torchtune.modules.transforms.tokenizers._utils import BaseTokenizer
 
 WHITESPACE_CHARS = [" ", "\n", "\t", "\r", "\v"]
 
diff --git a/torchtune/modules/tokenizers/_tiktoken.py b/torchtune/modules/transforms/tokenizers/_tiktoken.py
similarity index 98%
rename from torchtune/modules/tokenizers/_tiktoken.py
rename to torchtune/modules/transforms/tokenizers/_tiktoken.py
index 077b22b0cd..64733b4634 100644
--- a/torchtune/modules/tokenizers/_tiktoken.py
+++ b/torchtune/modules/transforms/tokenizers/_tiktoken.py
@@ -8,7 +8,7 @@
 
 from tiktoken import Encoding
 from tiktoken.load import load_tiktoken_bpe
-from torchtune.modules.tokenizers._utils import BaseTokenizer
+from torchtune.modules.transforms.tokenizers._utils import BaseTokenizer
 
 # Constants controlling encode logic
 MAX_ENCODE_CHARS = 400_000
diff --git a/torchtune/modules/tokenizers/_utils.py b/torchtune/modules/transforms/tokenizers/_utils.py
similarity index 97%
rename from torchtune/modules/tokenizers/_utils.py
rename to torchtune/modules/transforms/tokenizers/_utils.py
index b580eda1c0..ff374738c7 100644
--- a/torchtune/modules/tokenizers/_utils.py
+++ b/torchtune/modules/transforms/tokenizers/_utils.py
@@ -14,8 +14,8 @@
 class BaseTokenizer(Protocol):
     """
     Abstract token encoding model that implements ``encode`` and ``decode`` methods.
-    See :class:`~torchtune.modules.tokenizers.SentencePieceBaseTokenizer` and
-    :class:`~torchtune.modules.tokenizers.TikTokenBaseTokenizer` for example implementations of this protocol.
+    See :class:`~torchtune.modules.transforms.tokenizers.SentencePieceBaseTokenizer` and
+    :class:`~torchtune.modules.transforms.tokenizers.TikTokenBaseTokenizer` for example implementations of this protocol.
     """
 
     def encode(self, text: str, **kwargs: Dict[str, Any]) -> List[int]: