rtformer训练和评估都可以正常运行，模型导出出错。

[loxoo6]

问题确认 Search before asking

• 我已经查询历史issue(包括open与closed)，没有发现相似的bug。I have searched the open and closed issues and found no similar bug report.

Bug描述 Describe the Bug

2023-07-27 15:42:29 [INFO] Loaded trained params of model successfully.
/opt/conda/envs/python35-paddle120-env/lib/python3.7/site-packages/paddle/fluid/layers/math_op_patch.py:341: UserWarning: /tmp/tmpybaytmvi.py:22
The behavior of expression A + B has been unified with elementwise_add(X, Y, axis=-1) from Paddle 2.0. If your code works well in the older versions but crashes in this version, try to use elementwise_add(X, Y, axis=0) instead of A + B. This transitional warning will be dropped in the future.
op_type, op_type, EXPRESSION_MAP[method_name]))
/opt/conda/envs/python35-paddle120-env/lib/python3.7/site-packages/paddle/fluid/layers/math_op_patch.py:341: UserWarning: /tmp/tmp_6zue3ji.py:22
The behavior of expression A + B has been unified with elementwise_add(X, Y, axis=-1) from Paddle 2.0. If your code works well in the older versions but crashes in this version, try to use elementwise_add(X, Y, axis=0) instead of A + B. This transitional warning will be dropped in the future.
op_type, op_type, EXPRESSION_MAP[method_name]))
/opt/conda/envs/python35-paddle120-env/lib/python3.7/site-packages/paddle/fluid/layers/math_op_patch.py:341: UserWarning: /tmp/tmpiv_0rys8.py:22
The behavior of expression A + B has been unified with elementwise_add(X, Y, axis=-1) from Paddle 2.0. If your code works well in the older versions but crashes in this version, try to use elementwise_add(X, Y, axis=0) instead of A + B. This transitional warning will be dropped in the future.
op_type, op_type, EXPRESSION_MAP[method_name]))
/opt/conda/envs/python35-paddle120-env/lib/python3.7/site-packages/paddle/fluid/layers/math_op_patch.py:341: UserWarning: /tmp/tmpc6phgklz.py:22
The behavior of expression A + B has been unified with elementwise_add(X, Y, axis=-1) from Paddle 2.0. If your code works well in the older versions but crashes in this version, try to use elementwise_add(X, Y, axis=0) instead of A + B. This transitional warning will be dropped in the future.
op_type, op_type, EXPRESSION_MAP[method_name]))
/opt/conda/envs/python35-paddle120-env/lib/python3.7/site-packages/paddle/fluid/layers/math_op_patch.py:341: UserWarning: /tmp/tmp95zzc269.py:22
The behavior of expression A + B has been unified with elementwise_add(X, Y, axis=-1) from Paddle 2.0. If your code works well in the older versions but crashes in this version, try to use elementwise_add(X, Y, axis=0) instead of A + B. This transitional warning will be dropped in the future.
op_type, op_type, EXPRESSION_MAP[method_name]))
/opt/conda/envs/python35-paddle120-env/lib/python3.7/site-packages/paddle/fluid/layers/math_op_patch.py:341: UserWarning: /tmp/tmp72zst8tw.py:22
The behavior of expression A + B has been unified with elementwise_add(X, Y, axis=-1) from Paddle 2.0. If your code works well in the older versions but crashes in this version, try to use elementwise_add(X, Y, axis=0) instead of A + B. This transitional warning will be dropped in the future.
op_type, op_type, EXPRESSION_MAP[method_name]))
/opt/conda/envs/python35-paddle120-env/lib/python3.7/site-packages/paddle/fluid/layers/math_op_patch.py:341: UserWarning: /tmp/tmpiyz4_3sv.py:15
The behavior of expression A + B has been unified with elementwise_add(X, Y, axis=-1) from Paddle 2.0. If your code works well in the older versions but crashes in this version, try to use elementwise_add(X, Y, axis=0) instead of A + B. This transitional warning will be dropped in the future.
op_type, op_type, EXPRESSION_MAP[method_name]))
/opt/conda/envs/python35-paddle120-env/lib/python3.7/site-packages/paddle/fluid/layers/math_op_patch.py:341: UserWarning: /tmp/tmprosubqf3.py:22
The behavior of expression A + B has been unified with elementwise_add(X, Y, axis=-1) from Paddle 2.0. If your code works well in the older versions but crashes in this version, try to use elementwise_add(X, Y, axis=0) instead of A + B. This transitional warning will be dropped in the future.
op_type, op_type, EXPRESSION_MAP[method_name]))
/opt/conda/envs/python35-paddle120-env/lib/python3.7/site-packages/paddle/fluid/layers/math_op_patch.py:341: UserWarning: /tmp/tmprvfnl6ko.py:13
The behavior of expression A / B has been unified with elementwise_div(X, Y, axis=-1) from Paddle 2.0. If your code works well in the older versions but crashes in this version, try to use elementwise_div(X, Y, axis=0) instead of A / B. This transitional warning will be dropped in the future.
op_type, op_type, EXPRESSION_MAP[method_name]))
/opt/conda/envs/python35-paddle120-env/lib/python3.7/site-packages/paddle/fluid/layers/math_op_patch.py:341: UserWarning: /tmp/tmpijsiv8ky.py:12
The behavior of expression A + B has been unified with elementwise_add(X, Y, axis=-1) from Paddle 2.0. If your code works well in the older versions but crashes in this version, try to use elementwise_add(X, Y, axis=0) instead of A + B. This transitional warning will be dropped in the future.
op_type, op_type, EXPRESSION_MAP[method_name]))
/opt/conda/envs/python35-paddle120-env/lib/python3.7/site-packages/paddle/fluid/layers/math_op_patch.py:341: UserWarning: /tmp/tmpijsiv8ky.py:14
The behavior of expression A + B has been unified with elementwise_add(X, Y, axis=-1) from Paddle 2.0. If your code works well in the older versions but crashes in this version, try to use elementwise_add(X, Y, axis=0) instead of A + B. This transitional warning will be dropped in the future.
op_type, op_type, EXPRESSION_MAP[method_name]))
/opt/conda/envs/python35-paddle120-env/lib/python3.7/site-packages/paddle/fluid/layers/math_op_patch.py:341: UserWarning: /tmp/tmpijsiv8ky.py:18
The behavior of expression A + B has been unified with elementwise_add(X, Y, axis=-1) from Paddle 2.0. If your code works well in the older versions but crashes in this version, try to use elementwise_add(X, Y, axis=0) instead of A + B. This transitional warning will be dropped in the future.
op_type, op_type, EXPRESSION_MAP[method_name]))
Traceback (most recent call last):
File "tools/export.py", line 140, in
main(args)
File "tools/export.py", line 115, in main
paddle.jit.save(new_net, save_path)
File "/opt/conda/envs/python35-paddle120-env/lib/python3.7/site-packages/paddle/fluid/dygraph/jit.py", line 631, in wrapper
func(layer, path, input_spec, **configs)
File "", line 2, in save
File "/opt/conda/envs/python35-paddle120-env/lib/python3.7/site-packages/paddle/fluid/wrapped_decorator.py", line 25, in impl
return wrapped_func(*args, **kwargs)
File "/opt/conda/envs/python35-paddle120-env/lib/python3.7/site-packages/paddle/fluid/dygraph/base.py", line 51, in impl
return func(*args, **kwargs)
File "/opt/conda/envs/python35-paddle120-env/lib/python3.7/site-packages/paddle/fluid/dygraph/jit.py", line 861, in save
inner_input_spec, with_hook=with_hook)
File "/opt/conda/envs/python35-paddle120-env/lib/python3.7/site-packages/paddle/fluid/dygraph/dygraph_to_static/program_translator.py", line 528, in concrete_program_specify_input_spec
*desired_input_spec, with_hook=with_hook)
File "/opt/conda/envs/python35-paddle120-env/lib/python3.7/site-packages/paddle/fluid/dygraph/dygraph_to_static/program_translator.py", line 436, in get_concrete_program
concrete_program, partial_program_layer = self._program_cache[cache_key]
File "/opt/conda/envs/python35-paddle120-env/lib/python3.7/site-packages/paddle/fluid/dygraph/dygraph_to_static/program_translator.py", line 801, in getitem
self._caches[item_id] = self._build_once(item)
File "/opt/conda/envs/python35-paddle120-env/lib/python3.7/site-packages/paddle/fluid/dygraph/dygraph_to_static/program_translator.py", line 790, in build_once
**cache_key.kwargs)
File "", line 2, in from_func_spec
File "/opt/conda/envs/python35-paddle120-env/lib/python3.7/site-packages/paddle/fluid/wrapped_decorator.py", line 25, in impl
return wrapped_func(*args, **kwargs)
File "/opt/conda/envs/python35-paddle120-env/lib/python3.7/site-packages/paddle/fluid/dygraph/base.py", line 51, in impl
return func(*args, **kwargs)
File "/opt/conda/envs/python35-paddle120-env/lib/python3.7/site-packages/paddle/fluid/dygraph/dygraph_to_static/program_translator.py", line 740, in from_func_spec
error_data.raise_new_exception()
File "/opt/conda/envs/python35-paddle120-env/lib/python3.7/site-packages/paddle/fluid/dygraph/dygraph_to_static/error.py", line 336, in raise_new_exception
six.exec("raise new_exception from None")
File "", line 1, in
AssertionError: In transformed code:

File "tools/export.py", line 68, in forward
outs = self.net(x)
File "/opt/conda/envs/python35-paddle120-env/lib/python3.7/site-packages/paddleseg/models/rtformer.py", line 189, in forward
[self.relu(x3_), self.relu(x3)])  # 2c, 1/8; 8c, 1/16
File "/opt/conda/envs/python35-paddle120-env/lib/python3.7/site-packages/paddleseg/models/rtformer.py", line 573, in forward
             1])  # n*out_channels_h,144,1,1
        x_h = x_h + self.drop_path(self.attn_h(x_h, cross_k,
                                               cross_v))  # n,out_chs_h,h,w
                                               ~~~~~~~~~~~~~~~~~~~~~~~~~~~~ <--- HERE

    x_h = x_h + self.drop_path(self.mlp_h(x_h))

File "/opt/conda/envs/python35-paddle120-env/lib/python3.7/site-packages/paddle/fluid/dygraph/layers.py", line 930, in __call__
return self._dygraph_call_func(*inputs, **kwargs)
File "/opt/conda/envs/python35-paddle120-env/lib/python3.7/site-packages/paddle/fluid/dygraph/layers.py", line 915, in _dygraph_call_func
outputs = self.forward(*inputs, **kwargs)
File "/tmp/tmp7vk1jylp.py", line 48, in forward
self, x), (cross_k, cross_v, self, x), (x,))
File "/opt/conda/envs/python35-paddle120-env/lib/python3.7/site-packages/paddle/fluid/dygraph/dygraph_to_static/convert_operators.py", line 211, in convert_ifelse
out = _run_py_ifelse(pred, true_fn, false_fn, true_args, false_args)
File "/opt/conda/envs/python35-paddle120-env/lib/python3.7/site-packages/paddle/fluid/dygraph/dygraph_to_static/convert_operators.py", line 257, in _run_py_ifelse
return true_fn(*true_args) if pred else false_fn(*false_args)
File "/tmp/tmp7vk1jylp.py", line 36, in false_fn_6
.format(B))
File "/opt/conda/envs/python35-paddle120-env/lib/python3.7/site-packages/paddle/fluid/dygraph/dygraph_to_static/convert_operators.py", line 478, in convert_assert
assert cond, message

AssertionError: The first dim of x (-1) should be greater than 0, please set input_shape for export.py

是不是配置文件里出错了呢？export:这里需要写什么信息嘛？
base: '../base/cityscapes.yml'

batch_size: 3 # total batch size: 4 * 3
iters: 120000

train_dataset:
transforms:
- type: ResizeStepScaling
min_scale_factor: 0.5
max_scale_factor: 2.0
scale_step_size: 0.25
- type: RandomPaddingCrop
crop_size: [1024, 512]
- type: RandomHorizontalFlip
- type: RandomDistort
brightness_range: 0.4
contrast_range: 0.4
saturation_range: 0.4
- type: Normalize
mean: [0.485, 0.456, 0.406]
std: [0.229, 0.224, 0.225]

val_dataset:
transforms:
- type: Resize
target_size: [2048, 1024]
keep_ratio: True
- type: Normalize
mean: [0.485, 0.456, 0.406]
std: [0.229, 0.224, 0.225]

export:
transforms:
- type: Resize
target_size: [2048, 512]
keep_ratio: True
- type: Normalize
mean: [0.485, 0.456, 0.406]
std: [0.229, 0.224, 0.225]

optimizer:
inherited: False
type: AdamW
beta1: 0.9
beta2: 0.999
weight_decay: 0.0125

lr_scheduler:
inherited: False
type: PolynomialDecay
learning_rate: 4.0e-4
power: 1.
end_lr: 1.0e-6
warmup_iters: 1500
warmup_start_lr: 1.0e-6

loss:
types:
- type: CrossEntropyLoss
coef: [1, 0.4]

model:
type: RTFormer
base_channels: 64
head_channels: 128
use_injection: [True, False]
pretrained: https://paddleseg.bj.bcebos.com/dygraph/backbone/rtformer_base_backbone_imagenet_pretrained.zip

复现环境 Environment

平台：aistudio
paddlepaddle 2.3.2
paddleseg 2.7

Bug描述确认 Bug description confirmation

• 我确认已经提供了Bug复现步骤、代码改动说明、以及环境信息，确认问题是可以复现的。I confirm that the bug replication steps, code change instructions, and environment information have been provided, and the problem can be reproduced.

是否愿意提交PR？ Are you willing to submit a PR?

• 我愿意提交PR！I'd like to help by submitting a PR!

[Asthestarsfalll]

导出命令是什么，指定一下 input_shape试试