卷积层通道剪裁报错

[zhuguiqian]

在使用slim.prune进行剪裁时会报错，模型可以正常推理。 errror信息如下： [12-03 16:06:43 MainThread @logger.py:224] Argv: paddle_pruning.py /home/zgq/anaconda3/envs/paddle/lib/python3.7/importlib/_bootstrap.py:219: RuntimeWarning: numpy.ufunc size changed, may indicate binary incompatibility. Expected 192 from C header, got 216 from PyObject return f(*args, **kwds) FLOPs: 37047296.0 2020-12-03 16:06:44,036-INFO: pruning: x2paddle_convolution_W Traceback (most recent call last): File "paddle_pruning.py", line 52, in place=fluid.CPUPlace()) File "/home/zgq/anaconda3/envs/paddle/lib/python3.7/site-packages/paddleslim/prune/pruner.py", line 165, in prune graph.infer_shape() File "/home/zgq/anaconda3/envs/paddle/lib/python3.7/site-packages/paddleslim/core/graph_wrapper.py", line 362, in infer_shape op._op.desc.infer_shape(op._op.block.desc) paddle.fluid.core_avx.EnforceNotMet:

---

C++ Call Stacks (More useful to developers):

0 std::string paddle::platform::GetTraceBackString<char const*>(char const*&&, char const*, int) 1 paddle::platform::EnforceNotMet::EnforceNotMet(std::__exception_ptr::exception_ptr, char const*, int) 2 paddle::framework::OpDesc::InferShape(paddle::framework::BlockDesc const&) const

---

Python Call Stacks (More useful to users):

File "/home/zgq/anaconda3/envs/paddle/lib/python3.7/site-packages/paddle/fluid/framework.py", line 2679, in _prepend_op attrs=kwargs.get("attrs", None)) File "/home/zgq/anaconda3/envs/paddle/lib/python3.7/site-packages/paddle/fluid/io.py", line 1048, in prepend_feed_ops attrs={'col': i}) File "/home/zgq/anaconda3/envs/paddle/lib/python3.7/site-packages/paddle/fluid/io.py", line 1247, in save_inference_model prepend_feed_ops(main_program, feeded_var_names) File "/home/zgq/anaconda3/envs/paddle/lib/python3.7/site-packages/x2paddle/core/op_mapper.py", line 193, in save_inference_model params_filename=None) File "/home/zgq/anaconda3/envs/paddle/lib/python3.7/site-packages/x2paddle/convert.py", line 184, in onnx2paddle mapper.save_inference_model(save_dir, params_merge) File "/home/zgq/anaconda3/envs/paddle/lib/python3.7/site-packages/x2paddle/convert.py", line 306, in main onnx2paddle(args.model, args.save_dir, params_merge) File "/home/zgq/anaconda3/envs/paddle/bin/x2paddle", line 8, in sys.exit(main())

---

Error Message Summary:

Error: feed's infer_shape has not been registered at (/paddle/paddle/fluid/framework/op_desc.cc:686) [operator < feed > error]

请问有什么建议吗

[qingqing01]

@zhuguiqian 您是x2paddle转换过来的模型，再使用PaddleSlim的裁剪吗？

[Water2style]

@zhuguiqian 您是x2paddle转换过来的模型，再使用PaddleSlim的裁剪吗？

我用 load_inference_model 了一个paddle模型 剪枝也会出这个错.