Skip to content

P
PaddleDetection

PaddlePaddle / PaddleDetection
大约 1 年 前同步成功

通知 695
Star 11112
Fork 2696
P
PaddleDetection
未验证 提交 0077c107 编写于 作者: N niefeng 提交者： GitHub
浏览文件
操作

fix ppyoloe+ distill mask_positive (#8340)

上级 83e538fe
隐藏空白更改
内联 并排
Showing with 2 addition and 3 deletion
ppdet/slim/distill_model.py
浏览文件 @ 0077c107
...@@ -332,12 +332,11 @@ class PPYOLOEDistillModel(DistillModel): ...@@ -332,12 +332,11 @@ class PPYOLOEDistillModel(DistillModel):
with paddle.no_grad(): with paddle.no_grad():
teacher_loss = self.teacher_model(inputs) teacher_loss = self.teacher_model(inputs)
if hasattr(self.teacher_model.yolo_head, "assigned_labels"): if hasattr(self.teacher_model.yolo_head, "assigned_labels"):
self.student_model.yolo_head.assigned_labels, self.student_model.yolo_head.assigned_bboxes, self.student_model.yolo_head.assigned_scores, self.student_model.yolo_head.mask_positive = \ self.student_model.yolo_head.assigned_labels, self.student_model.yolo_head.assigned_bboxes, self.student_model.yolo_head.assigned_scores = \
self.teacher_model.yolo_head.assigned_labels, self.teacher_model.yolo_head.assigned_bboxes, self.teacher_model.yolo_head.assigned_scores, self.teacher_model.yolo_head.mask_positive self.teacher_model.yolo_head.assigned_labels, self.teacher_model.yolo_head.assigned_bboxes, self.teacher_model.yolo_head.assigned_scores
delattr(self.teacher_model.yolo_head, "assigned_labels") delattr(self.teacher_model.yolo_head, "assigned_labels")
delattr(self.teacher_model.yolo_head, "assigned_bboxes") delattr(self.teacher_model.yolo_head, "assigned_bboxes")
delattr(self.teacher_model.yolo_head, "assigned_scores") delattr(self.teacher_model.yolo_head, "assigned_scores")
delattr(self.teacher_model.yolo_head, "mask_positive")
student_loss = self.student_model(inputs) student_loss = self.student_model(inputs)
logits_loss, feat_loss = self.distill_loss(self.teacher_model, logits_loss, feat_loss = self.distill_loss(self.teacher_model,
......
Markdown is supported
0% 或 .
You are about to add 0 people to the discussion. Proceed with caution.
先完成此消息的编辑！
想要评论请 注册 或