训练实例分割时发现一下错误(InvalidArgument) Broadcast dimension mismatch

[Du-J]

您好，我在训练实例分割时发现一下错误，排查数据正常用pytorch版训练也正常，麻烦您帮忙看一下问题是否有解决办法，谢谢。 Traceback (most recent call last): File "segment/train.py", line 663, in main(opt) File "segment/train.py", line 554, in main train(opt.hyp, opt, device, callbacks) File "segment/train.py", line 309, in train loss, loss_items = compute_loss(pred, targets, masks=masks.astype("float32")) File "/home/qwe/FRN/yolov5-Paddle-main/utils/segment/loss.py", line 47, in call tcls, tbox, indices, anchors, tidxs, xywhn = self.build_targets(p, targets) # targets File "/home/qwe/FRN/yolov5-Paddle-main/utils/segment/loss.py", line 159, in build_targets gxi = gain[[2, 4]] - gxy # inverse File "/home/qwe/.local/lib/python3.8/site-packages/paddle/fluid/dygraph/math_op_patch.py", line 304, in impl return math_op(self, other_var, -1) ValueError: (InvalidArgument) Broadcast dimension mismatch. Operands could not be broadcast together with the shape of X = [2] and the shape of Y = [0, 2, 8]. Received [2] in X is not equal to [8] in Y at i:2. [Hint: Expected x_dims_array[i] == y_dims_array[i] || x_dims_array[i] <= 1 || y_dims_array[i] <= 1 == true, but received x_dims_array[i] == y_dims_array[i] || x_dims_array[i] <= 1 || y_dims_array[i] <= 1:0 != true:1.] (at /paddle/paddle/phi/kernels/funcs/common_shape.h:84)

[GuoQuanhao]

能提供一部分数据集样例吗，你使用的什么paddle版本

[frn888]

paddle是2.3.2 CUDA11.2 CUDNN是8.5 使用--resume报以下错误Traceback (most recent call last): File "segment/train.py", line 664, in main(opt) File "segment/train.py", line 555, in main train(opt.hyp, opt, device, callbacks) File "segment/train.py", line 310, in train loss, loss_items = compute_loss(pred, targets, masks=masks.astype("float32")) File "/home/cathay/FRN/yolov5-Paddle/utils/segment/loss.py", line 84, in call for bi in b.unique(): File "/home/cathay/.conda/envs/paddle/lib/python3.6/site-packages/paddle/tensor/manipulation.py", line 1229, in unique return_counts, 'axis', axis, "is_sorted", True) SystemError: (Fatal) Operator unique raises an thrust::system::system_error exception. The exception content is :parallel_for failed: cudaErrorInvalidConfiguration: invalid configuration argument. (at /paddle/paddle/fluid/imperative/tracer.cc:307) terminate called without an active exception 训练开始时报以下错误

[frn888]

Validating runs/train-seg/exp2/weights/best.pdparams... Traceback (most recent call last): File "segment/train.py", line 664, in main(opt) File "segment/train.py", line 555, in main train(opt.hyp, opt, device, callbacks) File "segment/train.py", line 429, in train model=attempt_load(f, fp16=cuda, verbose=False), File "/home/cathay/FRN/yolov5-Paddle/models/experimental.py", line 102, in attempt_load anchors=ckpt_weights['hyp'].get('anchors') if hasattr(ckpt_weights, 'hyp') else None) File "/home/cathay/FRN/yolov5-Paddle/models/yolo.py", line 199, in init m.stride = paddle.to_tensor([s / x.shape[-2] for x in forward(paddle.zeros([1, ch, s, s], dtype=paddle.get_default_dtype()))]) # forward File "/home/cathay/FRN/yolov5-Paddle/models/yolo.py", line 198, in forward = lambda x: self.forward(x)[0] if isinstance(m, Segment) else self.forward(x) File "/home/cathay/FRN/yolov5-Paddle/models/yolo.py", line 214, in forward return self._forward_once(x, profile, visualize) # single-scale inference, train File "/home/cathay/FRN/yolov5-Paddle/models/yolo.py", line 123, in _forward_once x = m(x) # run File "/home/cathay/.conda/envs/paddle/lib/python3.6/site-packages/paddle/fluid/dygraph/layers.py", line 930, in call return self._dygraph_call_func(*inputs, *kwargs) File "/home/cathay/.conda/envs/paddle/lib/python3.6/site-packages/paddle/fluid/dygraph/layers.py", line 915, in _dygraph_call_func outputs = self.forward(inputs, *kwargs) File "/home/cathay/.conda/envs/paddle/lib/python3.6/site-packages/paddle/nn/layer/common.py", line 417, in forward name=self.name) File "/home/cathay/.conda/envs/paddle/lib/python3.6/site-packages/paddle/nn/functional/common.py", line 475, in interpolate out = _C_ops.nearest_interp_v2(x, dy_attr) RuntimeError: (NotFound) Operator nearest_interp_v2 does not have kernel for {data_type[::paddle::platform::float16]; data_layout[Undefined(AnyLayout)]; place[Place(gpu:0)]; library_type[PLAIN]}. [Hint: Expected kernel_iter != kernels.end(), but received kernel_iter == kernels.end().] (at /paddle/paddle/fluid/imperative/prepared_operator.cc:419) [operator < nearest_interp_v2 > error]

[frn888]

Traceback (most recent call last): File "detect.py", line 255, in main(opt) File "detect.py", line 250, in main run(*vars(opt)) File "", line 2, in run File "/home/cathay/.conda/envs/paddle/lib/python3.6/site-packages/paddle/fluid/dygraph/base.py", line 354, in _decorate_function return func(args, **kwargs) File "detect.py", line 149, in run det[:, :4] = scale_boxes(im.shape[2:], det[:, :4], im0.shape).round() File "/home/cathay/.conda/envs/paddle/lib/python3.6/site-packages/paddle/fluid/dygraph/varbase_patch_methods.py", line 740, in getitem return self._getitem_index_not_tensor(item) ValueError: (InvalidArgument) Too many indices (2) for tensor of dimension 1. [Hint: Expected valid_indexs <= rank == true, but received valid_indexs <= rank:0 != true:1.] (at /paddle/paddle/fluid/pybind/slice_utils.h:288)

[frn888]

您好 训练使用的动态度可以改成静态图训练吗？

[GuoQuanhao]

不能哦

[Du-J]

导出的模型可以是静态图模型吗？