refactor(TokenizerProcessor): improve dependency handling and observation management

- Updated TokenizerProcessor to conditionally import AutoTokenizer based on the availability of the transformers library, enhancing flexibility.
- Modified tokenizer attribute type to Any to accommodate scenarios where transformers may not be installed.
- Improved observation handling by using a more concise approach to manage the transition dictionary, ensuring compatibility with existing data structures.
- Added error handling for missing transformers library, providing clear guidance for users on installation requirements.

src/lerobot/processor/tokenizer_processor.py

@@ -5,14 +5,19 @@ Tokenizer processor for handling text tokenization in robot transitions.
 from __future__ import annotations
 
 from dataclasses import dataclass, field
-from typing import Any
+from typing import TYPE_CHECKING, Any
 
 import torch
-from transformers import AutoTokenizer
 
 from lerobot.configs.types import FeatureType, PolicyFeature
 from lerobot.constants import OBS_LANGUAGE
 from lerobot.processor.pipeline import EnvTransition, ProcessorStepRegistry, TransitionKey
+from lerobot.utils.import_utils import _transformers_available
+
+if TYPE_CHECKING or _transformers_available:
+    from transformers import AutoTokenizer
+else:
+    AutoTokenizer = None
 
 
 @dataclass
@@ -54,7 +59,7 @@ class TokenizerProcessor:
     """
 
     tokenizer_name: str | None = None
-    tokenizer: AutoTokenizer | None = None
+    tokenizer: Any | None = None  # Otherwise transformers is not available in the core dependencies
     max_length: int = 512
     task_key: str = "task"
     padding_side: str = "right"
@@ -66,10 +71,18 @@ class TokenizerProcessor:
 
     def __post_init__(self):
         """Initialize the tokenizer from the provided tokenizer or tokenizer name."""
+        if not _transformers_available:
+            raise ImportError(
+                "The 'transformers' library is not installed. "
+                "Please install it with `pip install 'lerobot[transformers-dep]'` to use TokenizerProcessor."
+            )
+
         if self.tokenizer is not None:
             # Use provided tokenizer object directly
             self._tokenizer = self.tokenizer
         elif self.tokenizer_name is not None:
+            if AutoTokenizer is None:
+                raise ImportError("AutoTokenizer is not available")
             self._tokenizer = AutoTokenizer.from_pretrained(self.tokenizer_name)
         else:
             raise ValueError(
@@ -125,9 +138,11 @@ class TokenizerProcessor:
         tokenized_prompt = self._tokenize_text(task)
 
         # Get or create observation dict
-        if TransitionKey.OBSERVATION not in transition or transition[TransitionKey.OBSERVATION] is None:
-            transition[TransitionKey.OBSERVATION] = {}
-        observation = transition[TransitionKey.OBSERVATION]
+        observation = transition.get(TransitionKey.OBSERVATION)
+        if observation is None:
+            observation = {}
+        else:
+            observation = dict(observation)  # Make a copy
 
         # Add tokenized data to observation
         observation[f"{OBS_LANGUAGE}.tokens"] = tokenized_prompt["input_ids"]
@@ -135,6 +150,7 @@ class TokenizerProcessor:
             dtype=torch.bool
         )
 
+        transition[TransitionKey.OBSERVATION.value] = observation  # type: ignore[misc]
         return transition
 
     def _tokenize_text(self, text: str | list[str]) -> dict[str, torch.Tensor]:

src/lerobot/utils/import_utils.py

@@ -58,6 +58,7 @@ def is_package_available(pkg_name: str, return_version: bool = False) -> tuple[b
 
 
 _torch_available, _torch_version = is_package_available("torch", return_version=True)
+_transformers_available = is_package_available("transformers")
 _gym_xarm_available = is_package_available("gym_xarm")
 _gym_aloha_available = is_package_available("gym_aloha")
 _gym_pusht_available = is_package_available("gym_pusht")