Update deprecated type hinting in vllm/transformers_utils (#18058)

Signed-off-by: Harry Mellor <19981378+hmellor@users.noreply.github.com>
This commit is contained in:
Harry Mellor
2025-05-13 04:34:37 -07:00
committed by GitHub
parent ff334ca1cd
commit 8c946cecca
17 changed files with 98 additions and 102 deletions
@@ -24,7 +24,6 @@
# CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
import math
from typing import List, Tuple
import torch
import torchvision.transforms as T
@@ -36,8 +35,8 @@ from transformers.processing_utils import ProcessorMixin
class ImageTransform:
def __init__(self,
mean: Tuple[float, float, float] = (0.5, 0.5, 0.5),
std: Tuple[float, float, float] = (0.5, 0.5, 0.5),
mean: tuple[float, float, float] = (0.5, 0.5, 0.5),
std: tuple[float, float, float] = (0.5, 0.5, 0.5),
normalize: bool = True):
self.mean = mean
self.std = std
@@ -62,11 +61,11 @@ class DeepseekVLV2Processor(ProcessorMixin):
def __init__(
self,
tokenizer: LlamaTokenizerFast,
candidate_resolutions: Tuple[Tuple[int, int]],
candidate_resolutions: tuple[tuple[int, int]],
patch_size: int,
downsample_ratio: int,
image_mean: Tuple[float, float, float] = (0.5, 0.5, 0.5),
image_std: Tuple[float, float, float] = (0.5, 0.5, 0.5),
image_mean: tuple[float, float, float] = (0.5, 0.5, 0.5),
image_std: tuple[float, float, float] = (0.5, 0.5, 0.5),
normalize: bool = True,
image_token: str = "<image>",
pad_token: str = "<|▁pad▁|>",
@@ -170,13 +169,13 @@ class DeepseekVLV2Processor(ProcessorMixin):
return t
def decode(self, t: List[int], **kwargs) -> str:
def decode(self, t: list[int], **kwargs) -> str:
return self.tokenizer.decode(t, **kwargs)
def process_one(
self,
prompt: str,
images: List[Image.Image],
images: list[Image.Image],
inference_mode: bool = True,
**kwargs,
):
@@ -184,8 +183,8 @@ class DeepseekVLV2Processor(ProcessorMixin):
Args:
prompt (str): the formatted prompt;
conversations (List[Dict]): conversations with a list of messages;
images (List[ImageType]): the list of images;
conversations (list[dict]): conversations with a list of messages;
images (list[ImageType]): the list of images;
inference_mode (bool): if True, then remove the last eos token;
system_prompt (str): the system prompt;
**kwargs:
@@ -196,7 +195,7 @@ class DeepseekVLV2Processor(ProcessorMixin):
- target_ids (torch.LongTensor): [N + image tokens]
- pixel_values (torch.FloatTensor): [n_patches, 3, H, W]
- image_id (int): the id of the image token
- num_image_tokens (List[int]): the number of image tokens
- num_image_tokens (list[int]): the number of image tokens
"""
assert (prompt is not None and images is not None
@@ -257,7 +256,7 @@ class DeepseekVLV2Processor(ProcessorMixin):
self,
*,
prompt: str,
images: List[Image.Image],
images: list[Image.Image],
inference_mode: bool = True,
**kwargs,
):
@@ -265,7 +264,7 @@ class DeepseekVLV2Processor(ProcessorMixin):
Args:
prompt (str): the formatted prompt;
images (List[ImageType]): the list of images;
images (list[ImageType]): the list of images;
inference_mode (bool): if True, then remove the last eos token;
**kwargs:
@@ -274,7 +273,7 @@ class DeepseekVLV2Processor(ProcessorMixin):
- input_ids (torch.LongTensor): [N + image tokens]
- images (torch.FloatTensor): [n_images, 3, H, W]
- image_id (int): the id of the image token
- num_image_tokens (List[int]): the number of image tokens
- num_image_tokens (list[int]): the number of image tokens
"""
prepare = self.process_one(
@@ -288,7 +287,7 @@ class DeepseekVLV2Processor(ProcessorMixin):
def tokenize_with_images(
self,
conversation: str,
images: List[Image.Image],
images: list[Image.Image],
bos: bool = True,
eos: bool = True,
cropping: bool = True,
+6 -6
View File
@@ -23,7 +23,7 @@
# See the License for the specific language governing permissions and
# limitations under the License.
from functools import cached_property
from typing import List, Union
from typing import Union
import PIL
import torch
@@ -102,7 +102,7 @@ class OvisProcessor(ProcessorMixin):
def __call__(
self,
images: ImageInput = None,
text: Union[TextInput, PreTokenizedInput, List[TextInput], List[PreTokenizedInput]] = None,
text: Union[TextInput, PreTokenizedInput, list[TextInput], list[PreTokenizedInput]] = None,
**kwargs: Unpack[OvisProcessorKwargs],
) -> BatchFeature:
"""
@@ -111,14 +111,14 @@ class OvisProcessor(ProcessorMixin):
the text. To prepare the vision inputs, this method forwards the `vision_infos` and `kwrags` arguments to
Qwen2VLImageProcessor's [`~Qwen2VLImageProcessor.__call__`] if `vision_infos` is not `None`.
Args:
images (`PIL.Image.Image`, `np.ndarray`, `torch.Tensor`, `List[PIL.Image.Image]`, `List[np.ndarray]`, `List[torch.Tensor]`):
images (`PIL.Image.Image`, `np.ndarray`, `torch.Tensor`, `list[PIL.Image.Image]`, `list[np.ndarray]`, `list[torch.Tensor]`):
The image or batch of images to be prepared. Each image can be a PIL image, NumPy array or PyTorch
tensor. Both channels-first and channels-last formats are supported.
text (`str`, `List[str]`, `List[List[str]]`):
text (`str`, `list[str]`, `list[list[str]]`):
The sequence or batch of sequences to be encoded. Each sequence can be a string or a list of strings
(pretokenized string). If the sequences are provided as list of strings (pretokenized), you must set
`is_split_into_words=True` (to lift the ambiguity with a batch of sequences).
videos (`np.ndarray`, `torch.Tensor`, `List[np.ndarray]`, `List[torch.Tensor]`):
videos (`np.ndarray`, `torch.Tensor`, `list[np.ndarray]`, `list[torch.Tensor]`):
The image or batch of videos to be prepared. Each video can be a 4D NumPy array or PyTorch
tensor, or a nested list of 3D frames. Both channels-first and channels-last formats are supported.
return_tensors (`str` or [`~utils.TensorType`], *optional*):
@@ -400,7 +400,7 @@ class OvisProcessor(ProcessorMixin):
The output of the model `generate` function. The output is expected to be a tensor of shape `(batch_size, sequence_length)`
or `(sequence_length,)`.
Returns:
`List[str]`: The decoded text.
`list[str]`: The decoded text.
"""
return self.tokenizer.batch_decode(
generated_outputs, skip_special_tokens=True, clean_up_tokenization_spaces=False