chore(deps): lock file maintenance

library/src/getitune/backend/lightning/models/detection/heads/base_head.py

@@ -315,6 +315,8 @@ def _predict_by_feat_single(
         results.scores = torch.cat(mlvl_scores)
         results.labels = torch.cat(mlvl_labels)
         if with_score_factors:
+            # pyrefly: ignore[no-matching-overload]
+            # If with_score_factors is True, mlvl_score_factors is always a list of Tensor
             results.score_factors = torch.cat(mlvl_score_factors)
 
         return self._bbox_post_process(results=results, cfg=cfg, rescale=rescale, with_nms=with_nms, img_meta=img_meta)
@@ -544,11 +546,13 @@ def export_by_feat(
 
         batch_mlvl_bboxes_pred = torch.cat(mlvl_valid_bboxes, dim=1)
         batch_scores = torch.cat(mlvl_valid_scores, dim=1)
-        batch_priors = torch.cat(mlvl_valid_priors, dim=1)
+        batch_priors = torch.cat(mlvl_valid_priors, dim=1)  # pyrefly: ignore[no-matching-overload]
 
         batch_bboxes = self.bbox_coder.decode_export(batch_priors, batch_mlvl_bboxes_pred, max_shape=img_shape)
 
         if with_score_factors:
+            # pyrefly: ignore[no-matching-overload]
+            # If with_score_factors is True, mlvl_score_factors is always a list of Tensor
             batch_score_factors = torch.cat(mlvl_score_factors, dim=1)
 
         if not self.use_sigmoid_cls:

library/src/getitune/backend/lightning/models/detection/heads/deim_decoder.py

@@ -794,7 +794,8 @@ def _get_decoder_input(
 
         if denoising_bbox_unact is not None:
             enc_topk_bbox_unact = torch.concat([denoising_bbox_unact, enc_topk_bbox_unact], dim=1)
-            content = torch.concat([denoising_logits, content], dim=1)
+            if denoising_logits is not None:
+                content = torch.concat([denoising_logits, content], dim=1)
 
         return content, enc_topk_bbox_unact, enc_topk_bboxes_list, enc_topk_logits_list, enc_outputs_logits
 

library/src/getitune/backend/lightning/models/detection/heads/dfine_decoder.py

@@ -625,7 +625,8 @@ def _get_decoder_input(
 
         if denoising_bbox_unact is not None:
             enc_topk_bbox_unact = torch.concat([denoising_bbox_unact, enc_topk_bbox_unact], dim=1)
-            content = torch.concat([denoising_logits, content], dim=1)
+            if denoising_logits is not None:
+                content = torch.concat([denoising_logits, content], dim=1)
 
         return content, enc_topk_bbox_unact, enc_topk_bboxes_list, enc_topk_logits_list, enc_outputs_logits
 

library/src/getitune/backend/lightning/models/detection/layers/common_layer.py

@@ -6,6 +6,8 @@
 Reference : https://github.com/WongKinYiu/YOLO
 """
 
+from __future__ import annotations
+
 import torch
 from torch import Tensor, nn
 
@@ -24,7 +26,7 @@ def __init__(self, dim: int = 1) -> None:
         super().__init__()
         self.dim = dim
 
-    def forward(self, x: Tensor) -> Tensor:
+    def forward(self, x: list[Tensor]) -> Tensor:
         """Forward function."""
         return torch.cat(x, self.dim)
 

library/src/getitune/backend/lightning/models/instance_segmentation/heads/rpn_head.py

@@ -434,7 +434,7 @@ def export_by_feat(
         # loop over features, decode boxes
         mlvl_valid_bboxes = []
         mlvl_scores = []
-        mlvl_valid_anchors = []
+        mlvl_valid_anchors: list[Tensor] = []
         for cls_score, bbox_pred, anchors in zip(
             mlvl_cls_scores,
             mlvl_bbox_preds,

library/src/getitune/backend/lightning/models/segmentation/heads/fcn_head.py

@@ -161,7 +161,7 @@ def _forward_feature(self, inputs: Tensor) -> Tensor:
         x = self._transform_inputs(inputs)
         feats = self.convs(x)
         if self.concat_input:
-            feats = self.conv_cat(torch.cat([x, feats], dim=1))
+            feats = self.conv_cat(torch.cat([x, feats], dim=1))  # pyrefly: ignore[no-matching-overload]
         return feats
 
     def forward(self, inputs: Tensor) -> Tensor:

library/src/getitune/data/utils/structures/mask/mask_util.py

@@ -7,12 +7,12 @@
 
 from typing import TYPE_CHECKING
 
-import numpy as np
 import pycocotools.mask as mask_utils
 import torch
 from torchvision.ops import roi_align
 
 if TYPE_CHECKING:
+    import numpy as np
     from torchvision import tv_tensors
 
 
@@ -106,19 +106,15 @@ def crop_and_resize_masks(
     if len(annos) == 0:
         return torch.empty((0, *out_shape), dtype=torch.float, device=device)
 
-    # convert bboxes to tensor
-    if isinstance(bboxes, np.ndarray):
-        bboxes = torch.from_numpy(bboxes).to(device=device)
-    if isinstance(inds, np.ndarray):
-        inds = torch.from_numpy(inds).to(device=device)
+    bboxes_tensor = torch.from_numpy(bboxes).to(device=device)
+    inds_tensor = torch.from_numpy(inds).to(device=device)
 
-    num_bbox = bboxes.shape[0]
-    # pyrefly: ignore[no-matching-overload]
-    fake_inds = torch.arange(num_bbox, device=device).to(dtype=bboxes.dtype)[:, None]
-    rois = torch.cat([fake_inds, bboxes], dim=1)  # Nx5
+    num_bbox = bboxes_tensor.shape[0]
+    fake_inds = torch.arange(num_bbox, device=device).to(dtype=bboxes_tensor.dtype)[:, None]
+    rois = torch.cat([fake_inds, bboxes_tensor], dim=1)  # Nx5
     rois = rois.to(device=device)
     if num_bbox > 0:
-        gt_masks_th = annos.index_select(0, inds).to(dtype=rois.dtype)
+        gt_masks_th = annos.index_select(0, inds_tensor).to(dtype=rois.dtype)
         targets = roi_align(gt_masks_th[:, None, :, :], rois, out_shape, 1.0, 0, True).squeeze(1)
         resized_masks = targets >= 0.5
     else: