[misc] upgrade format to py39 (#7256)

src/llamafactory/data/processor/__init__.py

@@ -9,9 +9,9 @@ from .unsupervised import UnsupervisedDatasetProcessor
 __all__ = [
     "DatasetProcessor",
     "FeedbackDatasetProcessor",
+    "PackedSupervisedDatasetProcessor",
     "PairwiseDatasetProcessor",
     "PretrainDatasetProcessor",
-    "PackedSupervisedDatasetProcessor",
     "SupervisedDatasetProcessor",
     "UnsupervisedDatasetProcessor",
 ]

src/llamafactory/data/processor/feedback.py

@@ -13,7 +13,8 @@
 # limitations under the License.
 
 from collections import defaultdict
-from typing import TYPE_CHECKING, Any, Dict, List, Optional, Sequence, Tuple
+from collections.abc import Sequence
+from typing import TYPE_CHECKING, Any, Optional
 
 from ...extras import logging
 from ...extras.constants import IGNORE_INDEX
@@ -30,15 +31,15 @@ logger = logging.get_logger(__name__)
 class FeedbackDatasetProcessor(DatasetProcessor):
     def _encode_data_example(
         self,
-        prompt: Sequence[Dict[str, str]],
-        response: Sequence[Dict[str, str]],
-        kl_response: Sequence[Dict[str, str]],
+        prompt: Sequence[dict[str, str]],
+        response: Sequence[dict[str, str]],
+        kl_response: Sequence[dict[str, str]],
         system: Optional[str],
         tools: Optional[str],
         images: Sequence["ImageInput"],
         videos: Sequence["VideoInput"],
         audios: Sequence["AudioInput"],
-    ) -> Tuple[List[int], List[int], List[int], List[int], bool]:
+    ) -> tuple[list[int], list[int], list[int], list[int], bool]:
         if response[0]["content"]:  # desired example
             kto_tag = True
             messages = prompt + [response[0]]
@@ -82,7 +83,7 @@ class FeedbackDatasetProcessor(DatasetProcessor):
         kl_labels = [IGNORE_INDEX] * kl_source_len + kl_response_ids
         return input_ids, labels, kl_input_ids, kl_labels, kto_tag
 
-    def preprocess_dataset(self, examples: Dict[str, List[Any]]) -> Dict[str, List[Any]]:
+    def preprocess_dataset(self, examples: dict[str, list[Any]]) -> dict[str, list[Any]]:
         # create unrelated input-output pairs for estimating the KL term by flipping the matched pairs
         kl_response = examples["_response"][::-1]
         model_inputs = defaultdict(list)
@@ -121,7 +122,7 @@ class FeedbackDatasetProcessor(DatasetProcessor):
 
         return model_inputs
 
-    def print_data_example(self, example: Dict[str, List[int]]) -> None:
+    def print_data_example(self, example: dict[str, list[int]]) -> None:
         valid_labels = list(filter(lambda x: x != IGNORE_INDEX, example["labels"]))
         print("input_ids:\n{}".format(example["input_ids"]))
         print("inputs:\n{}".format(self.tokenizer.decode(example["input_ids"], skip_special_tokens=False)))

src/llamafactory/data/processor/pairwise.py

@@ -13,7 +13,8 @@
 # limitations under the License.
 
 from collections import defaultdict
-from typing import TYPE_CHECKING, Any, Dict, List, Optional, Sequence, Tuple
+from collections.abc import Sequence
+from typing import TYPE_CHECKING, Any, Optional
 
 from ...extras import logging
 from ...extras.constants import IGNORE_INDEX
@@ -30,14 +31,14 @@ logger = logging.get_logger(__name__)
 class PairwiseDatasetProcessor(DatasetProcessor):
     def _encode_data_example(
         self,
-        prompt: Sequence[Dict[str, str]],
-        response: Sequence[Dict[str, str]],
+        prompt: Sequence[dict[str, str]],
+        response: Sequence[dict[str, str]],
         system: Optional[str],
         tools: Optional[str],
         images: Sequence["ImageInput"],
         videos: Sequence["VideoInput"],
         audios: Sequence["AudioInput"],
-    ) -> Tuple[List[int], List[int], List[int], List[int]]:
+    ) -> tuple[list[int], list[int], list[int], list[int]]:
         chosen_messages = self.template.mm_plugin.process_messages(
             prompt + [response[0]], images, videos, audios, self.processor
         )
@@ -68,7 +69,7 @@ class PairwiseDatasetProcessor(DatasetProcessor):
         rejected_labels = [IGNORE_INDEX] * source_len + rejected_ids
         return chosen_input_ids, chosen_labels, rejected_input_ids, rejected_labels
 
-    def preprocess_dataset(self, examples: Dict[str, List[Any]]) -> Dict[str, List[Any]]:
+    def preprocess_dataset(self, examples: dict[str, list[Any]]) -> dict[str, list[Any]]:
         # build input pairs with format `<bos> X`, `Y1 <eos>` and `Y2 <eos>`
         model_inputs = defaultdict(list)
         for i in range(len(examples["_prompt"])):
@@ -99,7 +100,7 @@ class PairwiseDatasetProcessor(DatasetProcessor):
 
         return model_inputs
 
-    def print_data_example(self, example: Dict[str, List[int]]) -> None:
+    def print_data_example(self, example: dict[str, list[int]]) -> None:
         valid_chosen_labels = list(filter(lambda x: x != IGNORE_INDEX, example["chosen_labels"]))
         valid_rejected_labels = list(filter(lambda x: x != IGNORE_INDEX, example["rejected_labels"]))
         print("chosen_input_ids:\n{}".format(example["chosen_input_ids"]))

src/llamafactory/data/processor/pretrain.py

@@ -17,14 +17,14 @@
 
 from dataclasses import dataclass
 from itertools import chain
-from typing import Any, Dict, List
+from typing import Any
 
 from .processor_utils import DatasetProcessor
 
 
 @dataclass
 class PretrainDatasetProcessor(DatasetProcessor):
-    def preprocess_dataset(self, examples: Dict[str, List[Any]]) -> Dict[str, List[Any]]:
+    def preprocess_dataset(self, examples: dict[str, list[Any]]) -> dict[str, list[Any]]:
         # build grouped texts with format `X1 X2 X3 ...` if packing is enabled
         eos_token = "<|end_of_text|>" if self.data_args.template == "llama3" else self.tokenizer.eos_token
         text_examples = [messages[0]["content"] + eos_token for messages in examples["_prompt"]]
@@ -52,6 +52,6 @@ class PretrainDatasetProcessor(DatasetProcessor):
 
         return result
 
-    def print_data_example(self, example: Dict[str, List[int]]) -> None:
+    def print_data_example(self, example: dict[str, list[int]]) -> None:
         print("input_ids:\n{}".format(example["input_ids"]))
         print("inputs:\n{}".format(self.tokenizer.decode(example["input_ids"], skip_special_tokens=False)))

src/llamafactory/data/processor/processor_utils.py

@@ -14,8 +14,9 @@
 
 import bisect
 from abc import ABC, abstractmethod
+from collections.abc import Sequence
 from dataclasses import dataclass
-from typing import TYPE_CHECKING, Any, Dict, List, Optional, Sequence, Tuple
+from typing import TYPE_CHECKING, Any, Optional
 
 
 if TYPE_CHECKING:
@@ -27,9 +28,7 @@ if TYPE_CHECKING:
 
 @dataclass
 class DatasetProcessor(ABC):
-    r"""
-    A class for data processors.
-    """
+    r"""A class for data processors."""
 
     template: "Template"
     tokenizer: "PreTrainedTokenizer"
@@ -37,32 +36,24 @@ class DatasetProcessor(ABC):
     data_args: "DataArguments"
 
     @abstractmethod
-    def preprocess_dataset(self, examples: Dict[str, List[Any]]) -> Dict[str, List[Any]]:
-        r"""
-        Builds model inputs from the examples.
-        """
+    def preprocess_dataset(self, examples: dict[str, list[Any]]) -> dict[str, list[Any]]:
+        r"""Build model inputs from the examples."""
         ...
 
     @abstractmethod
-    def print_data_example(self, example: Dict[str, List[int]]) -> None:
-        r"""
-        Print a data example to stdout.
-        """
+    def print_data_example(self, example: dict[str, list[int]]) -> None:
+        r"""Print a data example to stdout."""
         ...
 
 
 def search_for_fit(numbers: Sequence[int], capacity: int) -> int:
-    r"""
-    Finds the index of largest number that fits into the knapsack with the given capacity.
-    """
+    r"""Find the index of largest number that fits into the knapsack with the given capacity."""
     index = bisect.bisect(numbers, capacity)
     return -1 if index == 0 else (index - 1)
 
 
-def greedy_knapsack(numbers: List[int], capacity: int) -> List[List[int]]:
-    r"""
-    An efficient greedy algorithm with binary search for the knapsack problem.
-    """
+def greedy_knapsack(numbers: list[int], capacity: int) -> list[list[int]]:
+    r"""Implement efficient greedy algorithm with binary search for the knapsack problem."""
     numbers.sort()  # sort numbers in ascending order for binary search
     knapsacks = []
 
@@ -83,10 +74,8 @@ def greedy_knapsack(numbers: List[int], capacity: int) -> List[List[int]]:
     return knapsacks
 
 
-def infer_seqlen(source_len: int, target_len: int, cutoff_len: int) -> Tuple[int, int]:
-    r"""
-    Computes the real sequence length after truncation by the cutoff_len.
-    """
+def infer_seqlen(source_len: int, target_len: int, cutoff_len: int) -> tuple[int, int]:
+    r"""Compute the real sequence length after truncation by the cutoff_len."""
     if target_len * 2 < cutoff_len:  # truncate source
         max_target_len = cutoff_len
     elif source_len * 2 < cutoff_len:  # truncate target

src/llamafactory/data/processor/supervised.py

@@ -13,8 +13,9 @@
 # limitations under the License.
 
 from collections import defaultdict
+from collections.abc import Sequence
 from dataclasses import dataclass
-from typing import TYPE_CHECKING, Any, Dict, List, Optional, Sequence, Tuple
+from typing import TYPE_CHECKING, Any, Optional
 
 from ...extras import logging
 from ...extras.constants import IGNORE_INDEX
@@ -32,14 +33,14 @@ logger = logging.get_logger(__name__)
 class SupervisedDatasetProcessor(DatasetProcessor):
     def _encode_data_example(
         self,
-        prompt: Sequence[Dict[str, str]],
-        response: Sequence[Dict[str, str]],
+        prompt: Sequence[dict[str, str]],
+        response: Sequence[dict[str, str]],
         system: Optional[str],
         tools: Optional[str],
         images: Sequence["ImageInput"],
         videos: Sequence["VideoInput"],
         audios: Sequence["AudioInput"],
-    ) -> Tuple[List[int], List[int]]:
+    ) -> tuple[list[int], list[int]]:
         messages = self.template.mm_plugin.process_messages(prompt + response, images, videos, audios, self.processor)
         input_ids, labels = self.template.mm_plugin.process_token_ids(
             [], [], images, videos, audios, self.tokenizer, self.processor
@@ -85,7 +86,7 @@ class SupervisedDatasetProcessor(DatasetProcessor):
 
         return input_ids, labels
 
-    def preprocess_dataset(self, examples: Dict[str, List[Any]]) -> Dict[str, List[Any]]:
+    def preprocess_dataset(self, examples: dict[str, list[Any]]) -> dict[str, list[Any]]:
         # build inputs with format `<bos> X Y <eos>` and labels with format `<ignore> ... <ignore> Y <eos>`
         # for multiturn examples, we only mask the prompt part in each prompt-response pair.
         model_inputs = defaultdict(list)
@@ -114,7 +115,7 @@ class SupervisedDatasetProcessor(DatasetProcessor):
 
         return model_inputs
 
-    def print_data_example(self, example: Dict[str, List[int]]) -> None:
+    def print_data_example(self, example: dict[str, list[int]]) -> None:
         valid_labels = list(filter(lambda x: x != IGNORE_INDEX, example["labels"]))
         print("input_ids:\n{}".format(example["input_ids"]))
         print("inputs:\n{}".format(self.tokenizer.decode(example["input_ids"], skip_special_tokens=False)))
@@ -124,7 +125,7 @@ class SupervisedDatasetProcessor(DatasetProcessor):
 
 @dataclass
 class PackedSupervisedDatasetProcessor(SupervisedDatasetProcessor):
-    def preprocess_dataset(self, examples: Dict[str, List[Any]]) -> Dict[str, List[Any]]:
+    def preprocess_dataset(self, examples: dict[str, list[Any]]) -> dict[str, list[Any]]:
         # TODO: use `position_ids` to achieve packing
         # build inputs with format `<bos> X1 Y1 <eos> <bos> X2 Y2 <eos>`
         # and labels with format `<ignore> ... <ignore> Y1 <eos> <ignore> ... <ignore> Y2 <eos>`

src/llamafactory/data/processor/unsupervised.py

@@ -13,7 +13,8 @@
 # limitations under the License.
 
 from collections import defaultdict
-from typing import TYPE_CHECKING, Any, Dict, List, Optional, Sequence, Tuple
+from collections.abc import Sequence
+from typing import TYPE_CHECKING, Any, Optional
 
 from ...extras import logging
 from ..data_utils import Role
@@ -30,14 +31,14 @@ logger = logging.get_logger(__name__)
 class UnsupervisedDatasetProcessor(DatasetProcessor):
     def _encode_data_example(
         self,
-        prompt: Sequence[Dict[str, str]],
-        response: Sequence[Dict[str, str]],
+        prompt: Sequence[dict[str, str]],
+        response: Sequence[dict[str, str]],
         system: Optional[str],
         tools: Optional[str],
         images: Sequence["ImageInput"],
         videos: Sequence["VideoInput"],
         audios: Sequence["AudioInput"],
-    ) -> Tuple[List[int], List[int]]:
+    ) -> tuple[list[int], list[int]]:
         if len(response) == 1:
             messages = prompt + response
         else:
@@ -56,7 +57,7 @@ class UnsupervisedDatasetProcessor(DatasetProcessor):
         labels = labels[:target_len]
         return input_ids, labels
 
-    def preprocess_dataset(self, examples: Dict[str, List[Any]]) -> Dict[str, List[Any]]:
+    def preprocess_dataset(self, examples: dict[str, list[Any]]) -> dict[str, list[Any]]:
         # build inputs with format `<bos> X` and labels with format `Y <eos>`
         model_inputs = defaultdict(list)
         for i in range(len(examples["_prompt"])):
@@ -84,7 +85,7 @@ class UnsupervisedDatasetProcessor(DatasetProcessor):
 
         return model_inputs
 
-    def print_data_example(self, example: Dict[str, List[int]]) -> None:
+    def print_data_example(self, example: dict[str, list[int]]) -> None:
         print("input_ids:\n{}".format(example["input_ids"]))
         print("inputs:\n{}".format(self.tokenizer.decode(example["input_ids"], skip_special_tokens=False)))
         print("label_ids:\n{}".format(example["labels"]))