From 55e6acf8347873f41e0ca90401170592a3ef2e57 Mon Sep 17 00:00:00 2001 From: Kurkur <102506892+litmei@users.noreply.github.com> Date: Wed, 11 Mar 2026 15:03:08 +0800 Subject: [PATCH] [NPU][QwenVL] Support qwen image preprocess on npu (#20189) --- .../npu/modules/qwen_vl_processor.py | 153 ++++++++++++++++++ .../multimodal/processors/base_processor.py | 10 +- 2 files changed, 159 insertions(+), 4 deletions(-) create mode 100644 python/sglang/srt/hardware_backend/npu/modules/qwen_vl_processor.py diff --git a/python/sglang/srt/hardware_backend/npu/modules/qwen_vl_processor.py b/python/sglang/srt/hardware_backend/npu/modules/qwen_vl_processor.py new file mode 100644 index 000000000..1b3cf8b5a --- /dev/null +++ b/python/sglang/srt/hardware_backend/npu/modules/qwen_vl_processor.py @@ -0,0 +1,153 @@ +from typing import Optional + +import torch +import torchvision.transforms.v2.functional as tvF +from transformers.image_processing_utils import BatchFeature +from transformers.image_processing_utils_fast import ( + group_images_by_shape, + reorder_images, +) +from transformers.image_utils import SizeDict +from transformers.models.qwen2_vl.image_processing_qwen2_vl import smart_resize +from transformers.utils import TensorType + +from sglang.srt.utils import apply_module_patch + + +# Func refers to transformers.models.qwen2_vl.image_processing_qwen2_vl_fast.py +# Qwen2VLImageProcessorFast._preprocess +def npu_wrapper_preprocess(func): + + def _preprocess( + self, + images: list["torch.Tensor"], + do_resize: bool, + size: SizeDict, + interpolation: Optional["tvF.InterpolationMode"], + do_rescale: bool, + rescale_factor: float, + do_normalize: bool, + image_mean: float | list[float] | None, + image_std: float | list[float] | None, + patch_size: int, + temporal_patch_size: int, + merge_size: int, + disable_grouping: bool | None, + return_tensors: str | TensorType | None, + **kwargs, + ): + # Group images by size for batched resizing + grouped_images, grouped_images_index = group_images_by_shape( + images, disable_grouping=disable_grouping + ) + resized_images_grouped = {} + for shape, stacked_images in grouped_images.items(): + height, width = stacked_images.shape[-2:] + if do_resize: + resized_height, resized_width = smart_resize( + height, + width, + factor=patch_size * merge_size, + min_pixels=size["shortest_edge"], + max_pixels=size["longest_edge"], + ) + stacked_images = self.resize( + image=stacked_images, + size=SizeDict(height=resized_height, width=resized_width), + interpolation=interpolation, + ) + resized_images_grouped[shape] = stacked_images + resized_images = reorder_images(resized_images_grouped, grouped_images_index) + + # Group images by size for further processing + # Needed in case do_resize is False, or resize returns images with different sizes + grouped_images, grouped_images_index = group_images_by_shape( + resized_images, disable_grouping=disable_grouping + ) + processed_images_grouped = {} + processed_grids = {} + for shape, stacked_images in grouped_images.items(): + resized_height, resized_width = stacked_images.shape[-2:] + # Fused rescale and normalize + patches = self.rescale_and_normalize( + stacked_images, + do_rescale, + rescale_factor, + do_normalize, + image_mean, + image_std, + ) + if patches.ndim == 4: + # add a temporal dimension if we have images + patches = patches.unsqueeze(1) + if patches.shape[1] % temporal_patch_size != 0: + repeats = patches[:, -1:].repeat(1, temporal_patch_size - 1, 1, 1, 1) + patches = torch.cat([patches, repeats], dim=1) + batch_size, grid_t, channel = patches.shape[:3] + grid_t = grid_t // temporal_patch_size + grid_h, grid_w = resized_height // patch_size, resized_width // patch_size + + ###################################### + # Start of modifications for sglang # + ###################################### + patches = patches.view( + batch_size * grid_t, + temporal_patch_size * channel, + grid_h // merge_size, + merge_size, + patch_size, + grid_w // merge_size, + merge_size, + patch_size, + ) + patches = patches.permute(0, 1, 2, 5, 3, 6, 4, 7) + patches = patches.reshape( + batch_size, + grid_t, + temporal_patch_size, + channel, + grid_h * grid_w, + patch_size, + patch_size, + ) + patches = patches.permute(0, 1, 4, 3, 2, 5, 6) + flatten_patches = patches.reshape( + batch_size, + grid_t * grid_h * grid_w, + -1, + ) + ###################################### + # End of modifications for sglang # + ###################################### + + processed_images_grouped[shape] = flatten_patches + processed_grids[shape] = [[grid_t, grid_h, grid_w]] * batch_size + + processed_images = reorder_images( + processed_images_grouped, grouped_images_index + ) + processed_grids = reorder_images(processed_grids, grouped_images_index) + pixel_values = torch.cat(processed_images, dim=0) + image_grid_thw = torch.tensor(processed_grids) + + return BatchFeature( + data={"pixel_values": pixel_values, "image_grid_thw": image_grid_thw}, + tensor_type=return_tensors, + ) + + return _preprocess + + +_npu_preprocess_patched = False + + +def npu_apply_qwen_image_preprocess_patch(): + global _npu_preprocess_patched + if _npu_preprocess_patched: + return + apply_module_patch( + "transformers.models.qwen2_vl.image_processing_qwen2_vl_fast.Qwen2VLImageProcessorFast", + "_preprocess", + [npu_wrapper_preprocess], + ) + _npu_preprocess_patched = True diff --git a/python/sglang/srt/multimodal/processors/base_processor.py b/python/sglang/srt/multimodal/processors/base_processor.py index 36861d3a3..d21cca83e 100644 --- a/python/sglang/srt/multimodal/processors/base_processor.py +++ b/python/sglang/srt/multimodal/processors/base_processor.py @@ -346,11 +346,13 @@ class BaseMultimodalProcessor(ABC): kwargs["device"] = "xpu" elif not _is_npu: kwargs["device"] = "cuda" - elif processor.__class__.__name__ not in { - "Qwen2_5_VLProcessor", - "Qwen3VLProcessor", - }: + else: # Note: for qwen-vl, processor has some reshape issue because of dims restriction on Ascend. + from sglang.srt.hardware_backend.npu.modules.qwen_vl_processor import ( + npu_apply_qwen_image_preprocess_patch, + ) + + npu_apply_qwen_image_preprocess_patch() kwargs["device"] = "npu" result = processor.__call__(