paddlepaddle / x2paddle

Deep learning model converter for PaddlePaddle. (『飞桨』深度学习模型转换工具)

X2Paddle

简介

X2Paddle是飞桨生态下的模型转换工具，致力于帮助其它深度学习框架用户快速迁移至飞桨框架。目前支持推理模型的框架转换与PyTorch训练代码迁移，我们还提供了详细的不同框架间API对比文档，降低开发者将模型迁移到飞桨的时间成本。

特性

• 支持主流深度学习框架

  • 目前已经支持Caffe/TensorFlow/ONNX/PyTorch四大框架的预测模型的转换，PyTorch训练项目的转换，涵盖了目前市面主流深度学习框架

• 支持的模型丰富

  • 在主流的CV和NLP模型上支持大部分模型转换，目前X2Paddle支持130+ PyTorch OP，90+ ONNX OP，90+ TensorFlow OP 以及 30+ Caffe OP，详见 支持列表

• 简洁易用

  • 一条命令行或者一个API即可完成模型转换

能力

• 预测模型转换

  • 支持Caffe/TensorFlow/ONNX/PyTorch的模型一键转为飞桨的预测模型，并使用PaddleInference/PaddleLite进行CPU/GPU/Arm等设备的部署

• PyTorch训练项目转换

  • 支持PyTorch项目Python代码（包括训练、预测）一键转为基于飞桨框架的项目代码，帮助开发者快速迁移项目，并可享受AIStudio平台对于飞桨框架提供的海量免费计算资源【新功能，试一下！】

• API对应文档

  • 详细的API文档对比分析，帮助开发者快速从PyTorch框架的使用迁移至飞桨框架的使用，大大降低学习成本 【新内容，了解一下！】

安装

环境依赖

• python >= 3.5
• paddlepaddle >= 2.2.2
• tensorflow == 1.14 (如需转换TensorFlow模型)
• onnx >= 1.6.0 (如需转换ONNX模型)
• torch >= 1.5.0 (如需转换PyTorch模型)
• paddlelite >= 2.9.0 (如需一键转换成Paddle-Lite支持格式,推荐最新版本)

pip安装(推荐）

如需使用稳定版本，可通过pip方式安装X2Paddle：

pip install x2paddle

源码安装

如需体验最新功能，可使用源码安装方式：

git clone https://github.com/PaddlePaddle/X2Paddle.git
cd X2Paddle
git checkout develop
python setup.py install

快速开始

功能一：推理模型转换

PyTorch模型转换

from x2paddle.convert import pytorch2paddle
pytorch2paddle(module=torch_module,
               save_dir="./pd_model",
               jit_type="trace",
               input_examples=[torch_input])
# module (torch.nn.Module): PyTorch的Module。
# save_dir (str): 转换后模型的保存路径。
# jit_type (str): 转换方式。默认为"trace"。
# input_examples (list[torch.tensor]): torch.nn.Module的输入示例，list的长度必须与输入的长度一致。默认为None。

script模式以及更多细节可参考PyTorch模型转换文档。

TensorFlow模型转换

x2paddle --framework=tensorflow --model=tf_model.pb --save_dir=pd_model

ONNX模型转换

x2paddle --framework=onnx --model=onnx_model.onnx --save_dir=pd_model

Caffe模型转换

x2paddle --framework=caffe --prototxt=deploy.prototxt --weight=deploy.caffemodel --save_dir=pd_model

转换参数说明

参数	作用
--framework	源模型类型 (tensorflow、caffe、onnx)
--prototxt	当framework为caffe时，该参数指定caffe模型的proto文件路径
--weight	当framework为caffe时，该参数指定caffe模型的参数文件路径
--save_dir	指定转换后的模型保存目录路径
--model	当framework为tensorflow/onnx时，该参数指定tensorflow的pb模型文件或onnx模型路径
--input_shape_dict	[可选] For ONNX, 定义ONNX模型输入大小
--caffe_proto	[可选] 由caffe.proto编译成caffe_pb2.py文件的存放路径，当存在自定义Layer时使用，默认为None
--define_input_shape	[可选] For TensorFlow, 当指定该参数时，强制用户输入每个Placeholder的shape，见文档Q2
--enable_code_optim	[可选] For PyTorch, 是否对生成代码进行优化，默认为False
--to_lite	[可选] 是否使用opt工具转成Paddle-Lite支持格式，默认为False
--lite_valid_places	[可选] 指定转换类型，可以同时指定多个backend(以逗号分隔)，opt将会自动选择最佳方式，默认为arm
--lite_model_type	[可选] 指定模型转化类型，目前支持两种类型：protobuf和naive_buffer，默认为naive_buffer
--disable_feedback	[可选] 是否关闭X2Paddle使用反馈；X2Paddle默认会统计用户在进行模型转换时的成功率，以及转换框架来源等信息，以便于帮忙X2Paddle根据用户需求进行迭代，不会上传用户的模型文件。如若不想参与反馈，可指定此参数为False即可

X2Paddle API

目前X2Paddle提供API方式转换模型，可参考X2PaddleAPI

一键转换Paddle-Lite支持格式

可参考使用X2paddle导出Padde-Lite支持格式

功能二：PyTorch模型训练迁移

项目转换包括3个步骤

1. 项目代码预处理
2. 代码/预训练模型一键转换
3. 转换后代码后处理

详见PyTorch训练项目转换文档。

使用VisualDL进行模型转换

飞桨可视化工具VisualDL已经将模型转换工具部署在官网提供服务，可以点击服务链接进行在线的ONNX2Paddle模型转换。

使用教程

1. TensorFlow预测模型转换教程
2. MMDetection模型转换指南
3. PyTorch预测模型转换教程
4. PyTorch训练项目转换教程

🤗贡献代码🤗

我们非常欢迎您为X2Paddle贡献代码或者提供使用建议。如果您可以修复某个issue或者增加一个新功能，欢迎给我们提交Pull Requests，如果有PyTorch训练项目转换需求欢迎随时提issue~

x2paddle's Issues

Can X2Paddle convert SSD_mobileNet_V1 model zoo pretrained in tensorflow?

I download SSD_Mobilenet Tensorflow detection model zoo from https://github.com/tensorflow/models/blob/master/research/object_detection/g3doc/detection_model_zoo.md and use X2Paddle convert to Paddle-lite model and use it run on an Neural Network Inference Engine supported by huawei chip. BUT none of them success. Could you please do a try and give me some suggest?

模型的双向GRU怎么转比较好

复现的pytorch代码：
import torch
import torch.nn as nn
rnn = nn.GRU(10,20, bidirectional=True)
input = torch.randn(5, 3, 10)
h0 = torch.randn(2, 3, 20)
output, hn = rnn(input,h0)
torch.onnx.export(rnn, input, 'test.onnx', verbose=True)

转onnx出的结果：
graph(%input : Float(5, 3, 10),
%weight_ih_l0 : Float(60, 10),
%weight_hh_l0 : Float(60, 20),
%bias_ih_l0 : Float(60),
%bias_hh_l0 : Float(60),
%weight_ih_l0_reverse : Float(60, 10),
%weight_hh_l0_reverse : Float(60, 20),
%bias_ih_l0_reverse : Float(60),
%bias_hh_l0_reverse : Float(60)):
%9 : Long() = onnx::Constantvalue={1}, scope: GRU
%10 : Tensor = onnx::Shape(%input), scope: GRU
%11 : Long() = onnx::Gather[axis=0](%10, %9), scope: GRU
%12 : Long() = onnx::Constantvalue={2}, scope: GRU
%13 : Long() = onnx::Constantvalue={20}, scope: GRU
%14 : Tensor = onnx::Unsqueezeaxes=[0]
%15 : Tensor = onnx::Unsqueezeaxes=[0]
%16 : Tensor = onnx::Unsqueezeaxes=[0]
%17 : Tensor = onnx::Concat[axis=0](%14, %15, %16)
%18 : Float(2, 3, 20) = onnx::ConstantOfShapevalue={0}, scope: GRU
%19 : Tensor? = prim::Constant(), scope: GRU
%20 : Tensor = onnx::Sliceaxes=[0], ends=[40], starts=[20], scope: GRU
%21 : Tensor = onnx::Sliceaxes=[0], ends=[20], starts=[0], scope: GRU
%22 : Tensor = onnx::Sliceaxes=[0], ends=[60], starts=[40], scope: GRU
%23 : Tensor = onnx::Concat[axis=0](%20, %21, %22), scope: GRU
%24 : Tensor = onnx::Sliceaxes=[0], ends=[40], starts=[20], scope: GRU
%25 : Tensor = onnx::Sliceaxes=[0], ends=[20], starts=[0], scope: GRU
%26 : Tensor = onnx::Sliceaxes=[0], ends=[60], starts=[40], scope: GRU
%27 : Tensor = onnx::Concat[axis=0](%24, %25, %26), scope: GRU
%28 : Tensor = onnx::Sliceaxes=[0], ends=[40], starts=[20], scope: GRU
%29 : Tensor = onnx::Sliceaxes=[0], ends=[20], starts=[0], scope: GRU
%30 : Tensor = onnx::Sliceaxes=[0], ends=[60], starts=[40], scope: GRU
%31 : Tensor = onnx::Concat[axis=0](%28, %29, %30), scope: GRU
%32 : Tensor = onnx::Sliceaxes=[0], ends=[40], starts=[20], scope: GRU
%33 : Tensor = onnx::Sliceaxes=[0], ends=[20], starts=[0], scope: GRU
%34 : Tensor = onnx::Sliceaxes=[0], ends=[60], starts=[40], scope: GRU
%35 : Tensor = onnx::Concat[axis=0](%32, %33, %34), scope: GRU
%36 : Tensor = onnx::Concat[axis=0](%31, %35), scope: GRU
%37 : Tensor = onnx::Unsqueezeaxes=[0], scope: GRU
%38 : Tensor = onnx::Unsqueezeaxes=[0], scope: GRU
%39 : Tensor = onnx::Unsqueezeaxes=[0], scope: GRU
%40 : Tensor = onnx::Sliceaxes=[0], ends=[40], starts=[20], scope: GRU
%41 : Tensor = onnx::Sliceaxes=[0], ends=[20], starts=[0], scope: GRU
%42 : Tensor = onnx::Sliceaxes=[0], ends=[60], starts=[40], scope: GRU
%43 : Tensor = onnx::Concat[axis=0](%40, %41, %42), scope: GRU
%44 : Tensor = onnx::Sliceaxes=[0], ends=[40], starts=[20], scope: GRU
%45 : Tensor = onnx::Sliceaxes=[0], ends=[20], starts=[0], scope: GRU
%46 : Tensor = onnx::Sliceaxes=[0], ends=[60], starts=[40], scope: GRU
%47 : Tensor = onnx::Concat[axis=0](%44, %45, %46), scope: GRU
%48 : Tensor = onnx::Sliceaxes=[0], ends=[40], starts=[20], scope: GRU
%49 : Tensor = onnx::Sliceaxes=[0], ends=[20], starts=[0], scope: GRU
%50 : Tensor = onnx::Sliceaxes=[0], ends=[60], starts=[40], scope: GRU
%51 : Tensor = onnx::Concat[axis=0](%48, %49, %50), scope: GRU
%52 : Tensor = onnx::Sliceaxes=[0], ends=[40], starts=[20], scope: GRU
%53 : Tensor = onnx::Sliceaxes=[0], ends=[20], starts=[0], scope: GRU
%54 : Tensor = onnx::Sliceaxes=[0], ends=[60], starts=[40], scope: GRU
%55 : Tensor = onnx::Concat[axis=0](%52, %53, %54), scope: GRU
%56 : Tensor = onnx::Concat[axis=0](%51, %55), scope: GRU
%57 : Tensor = onnx::Unsqueezeaxes=[0], scope: GRU
%58 : Tensor = onnx::Unsqueezeaxes=[0], scope: GRU
%59 : Tensor = onnx::Unsqueezeaxes=[0], scope: GRU
%60 : Tensor = onnx::Concat[axis=0](%37, %57), scope: GRU
%61 : Tensor = onnx::Concat[axis=0](%38, %58), scope: GRU
%62 : Tensor = onnx::Concat[axis=0](%39, %59), scope: GRU
%63 : Tensor, %64 : Float(2, 3, 20) = onnx::GRU[direction="bidirectional", hidden_size=20, linear_before_reset=1](%input, %60, %61, %62, %19, %18), scope: GRU
%65 : Tensor = onnx::Transposeperm=[0, 2, 1, 3], scope: GRU
%66 : Tensor = onnx::Constantvalue= 0 0 -1 [ Variable[CPUType]{3} ], scope: GRU
%67 : Float(5, 3, 40) = onnx::Reshape(%65, %66), scope: GRU
return (%67, %64)

x2paddle --framework=onnx --model=test.onnx --save_dir=./转paddle模型，会找不到%19这个节点，报错如下：
Traceback (most recent call last):
File "/home/zhangzexin/anaconda3/envs/paddle/bin/x2paddle", line 10, in
sys.exit(main())
File "/home/zhangzexin/anaconda3/envs/paddle/lib/python3.6/site-packages/x2paddle/convert.py", line 211, in main
onnx2paddle(args.model, args.save_dir)
File "/home/zhangzexin/anaconda3/envs/paddle/lib/python3.6/site-packages/x2paddle/convert.py", line 154, in onnx2paddle
model = ONNXDecoder(model_path)
File "/home/zhangzexin/anaconda3/envs/paddle/lib/python3.6/site-packages/x2paddle/decoder/onnx_decoder.py", line 334, in init
self.standardize_variable_name(model.graph)
File "/home/zhangzexin/anaconda3/envs/paddle/lib/python3.6/site-packages/x2paddle/decoder/onnx_decoder.py", line 501, in standardize_variable_name
node.input[i] = self.make_variable_name(node.input[i])
File "/home/zhangzexin/anaconda3/envs/paddle/lib/python3.6/site-packages/x2paddle/decoder/onnx_decoder.py", line 478, in make_variable_name
raise ValueError('name should not be empty')
ValueError: name should not be empty

已看github上一个gru的issue，看到推荐是使用nn.GRUCell的方式，但由于使用了nn.GRU封装的bidirection等属性，想知道（1）现在nn.GRU是一定没有办法转成功嘛，（2）有没有办法在不重新训练模型的情况下转成paddle呢~非常期待回复，感谢！

加载tf的meta文件，输出是个tensor的列表，转换失败

python tf2fluid/convert.py --meta_file=${meta_file}
5 --ckpt_dir=./textnet-part-vgg-synth-8gpu-tf-base/
6 --in_nodes=input_images,input_proposals,input_decoders,input_target_weights
7 --input_shape=None,512,512,3
8 --input_format=NHWC
9 --output_nodes=model_wo_buckets/embedding_attention_decoder/attention_decoder/AttnOutputProjection/AttnOutputProjection/BiasAdd,model_wo_buckets/embedd ing_attention_decoder/attention_decoder/AttnOutputProjection_1/AttnOutputProjection/BiasAdd,model_wo_buckets/embedding_attention_decoder/attention_decoder/ AttnOutputProjection_2/AttnOutputProjection/BiasAdd,model_wo_buckets/embedding_attention_decoder/attention_decoder/AttnOutputProjection_3/AttnOutputProject ion/BiasAdd,model_wo_buckets/embedding_attention_decoder/attention_decoder/AttnOutputProjection_4/AttnOutputProjection/BiasAdd,model_wo_buckets/embedding_a ttention_decoder/attention_decoder/AttnOutputProjection_5/AttnOutputProjection/BiasAdd,model_wo_buckets/embedding_attention_decoder/attention_decoder/AttnO utputProjection_6/AttnOutputProjection/BiasAdd,model_wo_buckets/embedding_attention_decoder/attention_decoder/AttnOutputProjection_7/AttnOutputProjection/B iasAdd,model_wo_buckets/embedding_attention_decoder/attention_decoder/AttnOutputProjection_8/AttnOutputProjection/BiasAdd,model_wo_buckets/embedding_attent ion_decoder/attention_decoder/AttnOutputProjection_9/AttnOutputProjection/BiasAdd,model_wo_buckets/embedding_attention_decoder/attention_decoder/AttnOutput Projection_10/AttnOutputProjection/BiasAdd,model_wo_buckets/embedding_attention_decoder/attention_decoder/AttnOutputProjection_11/AttnOutputProjection/Bias Add,model_wo_buckets/embedding_attention_decoder/attention_decoder/AttnOutputProjection_12/AttnOutputProjection/BiasAdd,model_wo_buckets/embedding_attentio n_decoder/attention_decoder/AttnOutputProjection_13/AttnOutputProjection/BiasAdd,model_wo_buckets/embedding_attention_decoder/attention_decoder/AttnOutputP rojection_14/AttnOutputProjection/BiasAdd,model_wo_buckets/embedding_attention_decoder/attention_decoder/AttnOutputProjection_15/AttnOutputProjection/BiasA dd,model_wo_buckets/embedding_attention_decoder/attention_decoder/AttnOutputProjection_16/AttnOutputProjection/BiasAdd,model_wo_buckets/embedding_attention decoder/attention_decoder/AttnOutputProjection_17/AttnOutputProjection/BiasAdd,model_wo_buckets/embedding_attention_decoder/attention_decoder/AttnOutputPr ojection_18/AttnOutputProjection/BiasAdd,model_wo_buckets/embedding_attention_decoder/attention_decoder/AttnOutputProjection_19/AttnOutputProjection/BiasAd d,model_wo_buckets/embedding_attention_decoder/attention_decoder/AttnOutputProjection_20/AttnOutputProjection/BiasAdd,model_wo_buckets/embedding_attention decoder/attention_decoder/AttnOutputProjection_21/AttnOutputProjection/BiasAdd,model_wo_buckets/embedding_attention_decoder/attention_decoder/AttnOutputPro jection_22/AttnOutputProjection/BiasAdd,model_wo_buckets/embedding_attention_decoder/attention_decoder/AttnOutputProjection_23/AttnOutputProjection/BiasAdd ,model_wo_buckets/embedding_attention_decoder/attention_decoder/AttnOutputProjection_24/AttnOutputProjection/BiasAdd
10 --use_cuda=True
11 --save_dir=./save_fluid_model/ \

TypeError: OpDesc() missing 1 required positional argument: 'attrs'

[ INFO]onnx2fluid::convert:0071: loading model: baidu_903_ep39.onnx ...
[ INFO]onnx2fluid::convert:0075: checking model ...
[ INFO]onnx2fluid::convert:0082: using opset version: 9
[ INFO]onnx2fluid::convert:0093: model has 588 ops
[ INFO]onnx2fluid::convert:0098: optimizing model ...
[ DEBUG]onnx2fluid.onnx_utils::polish_model:0342: builtin optimizations to perform in ONNX:
['eliminate_deadend', 'eliminate_identity', 'eliminate_nop_dropout', 'eliminate_nop_monotone_argmax', 'eliminate_nop_pad', 'eliminate_nop_transpose', 'eliminate_unused_initializer', 'extract_constant_to_initializer', 'fuse_add_bias_into_conv', 'fuse_bn_into_conv', 'fuse_consecutive_concats', 'fuse_consecutive_log_softmax', 'fuse_consecutive_reduce_unsqueeze', 'fuse_consecutive_squeezes', 'fuse_consecutive_transposes', 'fuse_matmul_add_bias_into_gemm', 'fuse_pad_into_conv', 'fuse_transpose_into_gemm', 'lift_lexical_references', 'nop']
[ INFO]onnx2fluid::convert:0104: folder baidu_903_ep39/ cleared
[ INFO]onnx2fluid::convert:0136: conversion started
[ DEBUG]onnx2fluid::convert:0146: translating op op_0(op_0) ai.onnx::Shape ...
[CRITICAL]onnx2fluid::convert:0163: conversion failed for:
['var_0'] -> ::Shape -> ['var_670']
Traceback (most recent call last):
File "/home/zhangzexin/anaconda3/envs/paddle/bin/onnx2fluid", line 11, in
load_entry_point('onnx2fluid==0.1.1', 'console_scripts', 'onnx2fluid')()
File "/home/zhangzexin/anaconda3/envs/paddle/lib/python3.6/site-packages/pkg_resources/init.py", line 489, in load_entry_point
return get_distribution(dist).load_entry_point(group, name)
File "/home/zhangzexin/anaconda3/envs/paddle/lib/python3.6/site-packages/pkg_resources/init.py", line 2843, in load_entry_point
return ep.load()
File "/home/zhangzexin/anaconda3/envs/paddle/lib/python3.6/site-packages/pkg_resources/init.py", line 2434, in load
return self.resolve()
File "/home/zhangzexin/anaconda3/envs/paddle/lib/python3.6/site-packages/pkg_resources/init.py", line 2440, in resolve
module = import(self.module_name, fromlist=['name'], level=0)
File "", line 971, in _find_and_load
File "", line 955, in _find_and_load_unlocked
File "", line 656, in _load_unlocked
File "", line 626, in _load_backward_compatible
File "/home/zhangzexin/anaconda3/envs/paddle/lib/python3.6/site-packages/onnx2fluid-0.1.1-py3.6.egg/onnx2fluid/main.py", line 121, in
File "/home/zhangzexin/anaconda3/envs/paddle/lib/python3.6/site-packages/onnx2fluid-0.1.1-py3.6.egg/onnx2fluid/cmdline.py", line 62, in main
File "/home/zhangzexin/anaconda3/envs/paddle/lib/python3.6/site-packages/onnx2fluid-0.1.1-py3.6.egg/onnx2fluid/conversion.py", line 164, in convert
File "/home/zhangzexin/anaconda3/envs/paddle/lib/python3.6/site-packages/onnx2fluid-0.1.1-py3.6.egg/onnx2fluid/conversion.py", line 159, in convert
File "/home/zhangzexin/anaconda3/envs/paddle/lib/python3.6/site-packages/onnx2fluid-0.1.1-py3.6.egg/onnx2fluid/writer.py", line 328, in emit_op
File "/home/zhangzexin/anaconda3/envs/paddle/lib/python3.6/site-packages/onnx2fluid-0.1.1-py3.6.egg/onnx2fluid/writer.py", line 232, in Op
File "/home/zhangzexin/anaconda3/envs/paddle/lib/python3.6/site-packages/onnx2fluid-0.1.1-py3.6.egg/onnx2fluid/symbolic.py", line 2037, in Shape
TypeError: OpDesc() missing 1 required positional argument: 'attrs'

想问一下这个可能是什么问题呢~

minor problems

1. function name:
   https://github.com/PaddlePaddle/X2Paddle/blob/develop/x2paddle/core/graph.py#L102
   Seems conflict with built-in 'print'.
2. function definition:
   https://github.com/PaddlePaddle/X2Paddle/blob/develop/x2paddle/op_mapper/onnx_op_mapper.py#L97
   should be

def directly_map(self, node, name='', *args,  **kwargs):

3. need to handle Unicode string
   https://github.com/PaddlePaddle/X2Paddle/blob/develop/x2paddle/core/fluid_code.py#L31
   https://github.com/PaddlePaddle/X2Paddle/blob/develop/x2paddle/core/fluid_code.py#L75

onnx转paddle失败

Now translating model from onnx to paddle.
model ir_version: 4, op version: 9
W1127 21:32:25.501857 54097 init.cc:125] Compiled with WITH_GPU, but no GPU found in runtime.
Traceback (most recent call last):
File "/home/yulu/anaconda3/envs/paddle/bin/onnx_infer", line 11, in
load_entry_point('x2paddle==0.6.0', 'console_scripts', 'onnx_infer')()
File "/home/yulu/anaconda3/envs/paddle/lib/python3.6/site-packages/x2paddle-0.6.0-py3.6.egg/x2paddle/onnx_infer.py", line 39, in main
sess = rt.InferenceSession(model_dir)
File "/home/yulu/anaconda3/envs/paddle/lib/python3.6/site-packages/onnxruntime/capi/session.py", line 29, in init
self._sess.load_model(path_or_bytes)
RuntimeError: [ONNXRuntimeError] : 1 : GENERAL ERROR : Load model from pd_model/tmp_data/onnx_model_infer.onnx failed:Type Error: Type parameter (T) bound to different types (tensor(float) and tensor(int64) in node (_1211).
Total nodes: 545
in (Constant -> _1146): attribute "shape" of _1146 not inferred, using value as 1-D tensor may lead to fails
in (Constant -> _1161): attribute "shape" of _1161 not inferred, using value as 1-D tensor may lead to fails
in (Constant -> _1162): attribute "shape" of _1162 not inferred, using value as 1-D tensor may lead to fails
in (Constant -> _1171): attribute "shape" of _1171 not inferred, using value as 1-D tensor may lead to fails
in (Constant -> _1174): attribute "shape" of _1174 not inferred, using value as 1-D tensor may lead to fails
in (Constant -> _1177): attribute "shape" of _1177 not inferred, using value as 1-D tensor may lead to fails
in (Constant -> _1181): attribute "shape" of _1181 not inferred, using value as 1-D tensor may lead to fails
in (Constant -> _1184): attribute "shape" of _1184 not inferred, using value as 1-D tensor may lead to fails
in (Constant -> _1187): attribute "shape" of _1187 not inferred, using value as 1-D tensor may lead to fails
in (Constant -> _1190): attribute "shape" of _1190 not inferred, using value as 1-D tensor may lead to fails
in (Constant -> _1193): attribute "shape" of _1193 not inferred, using value as 1-D tensor may lead to fails
in (Constant -> _1201): attribute "shape" of _1201 not inferred, using value as 1-D tensor may lead to fails
in (Constant -> _1224): attribute "shape" of _1224 not inferred, using value as 1-D tensor may lead to fails
in (Constant -> _1227): attribute "shape" of _1227 not inferred, using value as 1-D tensor may lead to fails
in (Constant -> _1230): attribute "shape" of _1230 not inferred, using value as 1-D tensor may lead to fails
in (Constant -> _1234): attribute "shape" of _1234 not inferred, using value as 1-D tensor may lead to fails
in (Constant -> _1242): attribute "shape" of _1242 not inferred, using value as 1-D tensor may lead to fails
in (Constant -> _1255): attribute "shape" of _1255 not inferred, using value as 1-D tensor may lead to fails
in (Constant -> _1262): attribute "shape" of _1262 not inferred, using value as 1-D tensor may lead to fails
in (Constant -> _1275): attribute "shape" of _1275 not inferred, using value as 1-D tensor may lead to fails
in (Constant -> _1285): attribute "shape" of _1285 not inferred, using value as 1-D tensor may lead to fails
in (Constant -> _1286): attribute "shape" of _1286 not inferred, using value as 1-D tensor may lead to fails
Traceback (most recent call last):
File "/home/yulu/anaconda3/envs/paddle/bin/x2paddle", line 11, in
load_entry_point('x2paddle==0.6.0', 'console_scripts', 'x2paddle')()
File "/home/yulu/anaconda3/envs/paddle/lib/python3.6/site-packages/x2paddle-0.6.0-py3.6.egg/x2paddle/convert.py", line 233, in main
onnx2paddle(args.model, args.save_dir, params_merge)
File "/home/yulu/anaconda3/envs/paddle/lib/python3.6/site-packages/x2paddle-0.6.0-py3.6.egg/x2paddle/convert.py", line 170, in onnx2paddle
mapper = ONNXOpMapper(model, save_dir)
File "/home/yulu/anaconda3/envs/paddle/lib/python3.6/site-packages/x2paddle-0.6.0-py3.6.egg/x2paddle/op_mapper/onnx_op_mapper.py", line 90, in init
func(node)
File "/home/yulu/anaconda3/envs/paddle/lib/python3.6/site-packages/x2paddle-0.6.0-py3.6.egg/x2paddle/op_mapper/onnx_op_mapper.py", line 513, in Unsqueeze
if len(val_x.out_shapes[0]) == 0:
TypeError: object of type 'NoneType' has no len()

Is caffe slice layer supported?

When trying to convert caffe to paddle, I got following error.

squeeze idx:1, with kind:Convolution,name:conv1 Traceback (most recent call last): File "/home/turing/anaconda3/bin/x2paddle", line 10, in <module> sys.exit(main()) File "/home/turing/anaconda3/lib/python3.6/site-packages/x2paddle/convert.py", line 163, in main args.caffe_proto) File "/home/turing/anaconda3/lib/python3.6/site-packages/x2paddle/convert.py", line 101, in caffe2paddle mapper = CaffeOpMapper(model) File "/home/turing/anaconda3/lib/python3.6/site-packages/x2paddle/op_mapper/caffe_op_mapper.py", line 44, in __init__ self.set_node_shape(node) File "/home/turing/anaconda3/lib/python3.6/site-packages/x2paddle/op_mapper/caffe_op_mapper.py", line 80, in set_node_shape input_shape.append(last_node.output_shape[idx]) IndexError: list index out of range

Hope to any suggestions.
Following is my prototxt.

name: "model"
layer {
name: "input"
type: "Input"
top: "data"
input_param {
shape {
dim: 1
dim: 1
dim: 128
dim: 128
}
}
}
layer {
name: "conv1"
type: "Convolution"
bottom: "data"
top: "conv1"
param {
lr_mult: 1
decay_mult: 1
}
param {
lr_mult: 2
decay_mult: 0
}
convolution_param {
num_output: 16
pad: 2
kernel_size: 5
stride: 1
weight_filler {
type: "xavier"
}
bias_filler {
type: "constant"
value: 0.1
}
}
}
layer {
name: "slice1"
type: "Slice"
bottom: "conv1"
top: "slice1_1"
top: "slice1_2"
slice_param {
slice_dim: 1
}
}
...

IndexError: list index out of range?

Hi,

x2paddle conversion gives this error. I used Ubuntu18.04 in VitrualBox.

Model: ssd_mobilenet_custom.zip

x2paddle --framework=caffe --prototxt=deploy.prototxt --weight=coco/ssd_mobilenet_custom.caffemodel --save_dir=pd_model
Now translating model from caffe to paddle.
Total nodes: 268
Traceback (most recent call last):
  File "/home/ghimire/.local/bin/x2paddle", line 11, in <module>
    sys.exit(main())
  File "/home/ghimire/.local/lib/python2.7/site-packages/x2paddle/convert.py", line 227, in main
    args.caffe_proto, params_merge)
  File "/home/ghimire/.local/lib/python2.7/site-packages/x2paddle/convert.py", line 146, in caffe2paddle
    mapper = CaffeOpMapper(model)
  File "/home/ghimire/.local/lib/python2.7/site-packages/x2paddle/op_mapper/caffe_op_mapper.py", line 51, in __init__
    self.deal_custom_layer(node)
  File "/home/ghimire/.local/lib/python2.7/site-packages/x2paddle/op_mapper/caffe_op_mapper.py", line 944, in deal_custom_layer
    input = self.graph.get_bottom_node(input, idx=0, copy=True)
  File "/home/ghimire/.local/lib/python2.7/site-packages/x2paddle/decoder/caffe_decoder.py", line 204, in get_bottom_node
    input_node_name = node.inputs[idx]
IndexError: list index out of range

Please help.

Best,
Deepak

error

(pytorch1_1_0) boyun@boyun:~/software/ubuntu-wine/BaiduYunDownload/X2Paddle-master/onnx2fluid$ onnx2fluid resnet18.onnx -t sample_1.npz
[ INFO]convert::convert:0051: loading model: resnet18.onnx ...
[ INFO]convert::convert:0054: checking model ...
[ INFO]convert::convert:0072: model has 296 ops
[ INFO]convert::convert:0073: optimizing model ...
[ INFO]convert::convert:0082: folder resnet18/ cleared
[ INFO]convert::convert:0113: conversion started
[CRITICAL]convert::convert:0136: conversion failed for:
['646', '653'] -> ::Reshape -> ['654']
Traceback (most recent call last):
File "/home/boyun/anaconda3/envs/pytorch1_1_0/bin/onnx2fluid", line 11, in
load_entry_point('onnx2fluid==0.1.1', 'console_scripts', 'onnx2fluid')()
File "/home/boyun/anaconda3/envs/pytorch1_1_0/lib/python3.6/site-packages/pkg_resources/init.py", line 489, in load_entry_point
return get_distribution(dist).load_entry_point(group, name)
File "/home/boyun/anaconda3/envs/pytorch1_1_0/lib/python3.6/site-packages/pkg_resources/init.py", line 2843, in load_entry_point
return ep.load()
File "/home/boyun/anaconda3/envs/pytorch1_1_0/lib/python3.6/site-packages/pkg_resources/init.py", line 2434, in load
return self.resolve()
File "/home/boyun/anaconda3/envs/pytorch1_1_0/lib/python3.6/site-packages/pkg_resources/init.py", line 2440, in resolve
module = import(self.module_name, fromlist=['name'], level=0)
File "", line 971, in _find_and_load
File "", line 955, in _find_and_load_unlocked
File "", line 656, in _load_unlocked
File "", line 626, in _load_backward_compatible
File "/home/boyun/anaconda3/envs/pytorch1_1_0/lib/python3.6/site-packages/onnx2fluid-0.1.1-py3.6.egg/onnx2fluid/main.py", line 106, in
File "/home/boyun/anaconda3/envs/pytorch1_1_0/lib/python3.6/site-packages/onnx2fluid-0.1.1-py3.6.egg/onnx2fluid/cmdline.py", line 60, in main
File "/home/boyun/anaconda3/envs/pytorch1_1_0/lib/python3.6/site-packages/onnx2fluid-0.1.1-py3.6.egg/onnx2fluid/conversion.py", line 137, in convert
File "/home/boyun/anaconda3/envs/pytorch1_1_0/lib/python3.6/site-packages/onnx2fluid-0.1.1-py3.6.egg/onnx2fluid/conversion.py", line 132, in convert
File "/home/boyun/anaconda3/envs/pytorch1_1_0/lib/python3.6/site-packages/onnx2fluid-0.1.1-py3.6.egg/onnx2fluid/writer.py", line 312, in emit_op
File "/home/boyun/anaconda3/envs/pytorch1_1_0/lib/python3.6/site-packages/onnx2fluid-0.1.1-py3.6.egg/onnx2fluid/writer.py", line 244, in Op
File "/home/boyun/anaconda3/envs/pytorch1_1_0/lib/python3.6/site-packages/onnx2fluid-0.1.1-py3.6.egg/onnx2fluid/symbolic.py", line 1559, in Reshape
File "/home/boyun/anaconda3/envs/pytorch1_1_0/lib/python3.6/site-packages/onnx2fluid-0.1.1-py3.6.egg/onnx2fluid/writer.py", line 200, in OpDesc
File "/home/boyun/anaconda3/envs/pytorch1_1_0/lib/python3.6/site-packages/onnx2fluid-0.1.1-py3.6.egg/onnx2fluid/writer.py", line 152, in OpDescAttrs
ValueError: unsupported attribute shape = []

为什么onnx转paddle需要是1.6版本的，有没有支持1.4版本的

版本太高，有些不支持

不支持instance normalization吗

The op DepthwiseConvolution in model is not supported yet

register layer[ROIPooling]
register layer[PriorBox]
register layer[Permute]
register layer[DetectionOutput]
register layer[Normalize]
register layer[Select]
register layer[ShuffleChannel]
register layer[ConvolutionDepthwise]
register layer[Axpy]
Now translating model from caffe to paddle.
...
Traceback (most recent call last):
  File "/Users//anaconda3/bin/x2paddle", line 10, in <module>
    sys.exit(main())
  File "/Users//anaconda3/lib/python3.5/site-packages/x2paddle/convert.py", line 201, in main
    args.caffe_proto)
  File "/Users//anaconda3/lib/python3.5/site-packages/x2paddle/convert.py", line 127, in caffe2paddle
    mapper = CaffeOpMapper(model)
  File "/Users//anaconda3/lib/python3.5/site-packages/x2paddle/op_mapper/caffe_op_mapper.py", line 55, in __init__
    "The op {} in model is not supported yet.".format(op))
Exception: The op DepthwiseConvolution in model is not supported yet.

看到注册的layer里有ConvolutionDepthwise，能直接映射过去吗？

使用x2paddle转换mobilenet(tensorflow),报错

转换命令：x2paddle -f tensorflow -m frozen_inference_graph.pb -s pd_model/ --without_data_format_optimization --define_input_shape
报错情况：
Now translating model from tensorflow to paddle.
Define shape[now is [-1L, -1L, -1L, 3L]] for input tensor[tensor name: "image_tensor']
Use your keyboard type the shape of input tensor below :)
Shape of Input(e.g. None,224,224,3): None,224,224,3
Traceback (most recent call last):
File "/usr/local/bin/x2paddle", line 11, in
load_entry_point('x2paddle==0.6.0', 'console_scripts', 'x2paddle')()
File "/usr/local/lib/python2.7/dist-packages/x2paddle-0.6.0-py2.7.egg/x2paddle/convert.py", line 219, in main
define_input_shape, params_merge)
File "/usr/local/lib/python2.7/dist-packages/x2paddle-0.6.0-py2.7.egg/x2paddle/convert.py", line 106, in tf2paddle
model = TFDecoder(model_path, define_input_shape=define_input_shape)
File "/usr/local/lib/python2.7/dist-packages/x2paddle-0.6.0-py2.7.egg/x2paddle/decoder/tf_decoder.py", line 263, in init
self.tf_graph.build()
File "/usr/local/lib/python2.7/dist-packages/x2paddle-0.6.0-py2.7.egg/x2paddle/decoder/tf_decoder.py", line 123, in build
format(in_node, layer_name))
Exception: input[^Preprocessor_map_while_Identity] of node[Preprocessor_map_while_add_y] does not exist in node_map

测试模型及环境配置说明见百度网盘：
链接：https://pan.baidu.com/s/1LTRQdLbGk35KZGL7Mloy_g
提取码：jm0u
复制这段内容后打开百度网盘手机App，操作更方便哦

Exception: input[_394] of node[_413] does not exist in node_map

I have successfully transferred the pytorch model to onnx and the report for onnx transfer paddle is as follows
Exception: input[_394] of node[_413] does not exist in node_map

GPT-2模型转换时出错。

在我使用onnx转paddle时候，报错：
There are 1 ops not supported yet, list as below
NonZero

希望能尽快支持

模型位置：https://github.com/onnx/models/tree/master/text/machine_comprehension/gpt-2/model

version must be larger than 3.6.0!

这代码是谁写的???

能不能把paddle的模型转换为tf呢

master caffe2fluid net_template.py有问题

应修改如下
diff --git a/caffe2fluid/kaffe/net_template.py b/caffe2fluid/kaffe/net_template.py
index f9387c9..c6810c6 100644
--- a/caffe2fluid/kaffe/net_template.py
+++ b/caffe2fluid/kaffe/net_template.py
@@ -100,7 +100,7 @@ def main():
npy_weight = args.npy_path
fluid_model = args.model_param_path
outputs = None

• if len(sys.argv) >= 6:

• if args.need_layers_name:
  outputs = args.need_layers_name.split(',')

  ret = MyNet.convert(npy_weight, fluid_model, outputs)

AssertionError: The version of protobuf must be larger than 3.6.0!

pip3 list显示python3.5对应的protobuf版本是3.6.0了已经，但是protoc --version的版本是3.4.0.所以那个assert过不去，但是不影响使用。
还有就是python2.7版本的啥时候放出来？

can not support get_dynamic_shape

i use dynamic shape in onnx model. and i have successfully convert to onnx model. when convert to paddle, the error is:

 File "/root/miniconda3/envs/torch120/bin/x2paddle", line 11, in <module>
    load_entry_point('x2paddle==0.5.0', 'console_scripts', 'x2paddle')()
  File "/root/miniconda3/envs/torch120/lib/python3.6/site-packages/x2paddle-0.5.0-py3.6.egg/x2paddle/convert.py", line 211, in main
    onnx2paddle(args.model, args.save_dir)
  File "/root/miniconda3/envs/torch120/lib/python3.6/site-packages/x2paddle-0.5.0-py3.6.egg/x2paddle/convert.py", line 157, in onnx2paddle
    mapper = ONNXOpMapper(model, save_dir)
  File "/root/miniconda3/envs/torch120/lib/python3.6/site-packages/x2paddle-0.5.0-py3.6.egg/x2paddle/op_mapper/onnx_op_mapper.py", line 65, in __init__
    self.get_output_shapes()
  File "/root/miniconda3/envs/torch120/lib/python3.6/site-packages/x2paddle-0.5.0-py3.6.egg/x2paddle/op_mapper/onnx_op_mapper.py", line 157, in get_output_shapes
    _, dtype, shape = self.get_dynamic_shape(opt)
  File "/root/miniconda3/envs/torch120/lib/python3.6/site-packages/x2paddle-0.5.0-py3.6.egg/x2paddle/op_mapper/onnx_op_mapper.py", line 134, in get_dynamic_shape
    output = np.load(os.path.join(self.tmp_data_dir, layer + '.npy'))

the model i used is:链接: https://pan.baidu.com/s/1WrTbTyVmBBJrT6ZVffx_3g 提取码: y9hb

Ops are not supported yet:Sqrt,MirrorPad,Square

(paddle) bash-3.2$ x2paddle --framework=tensorflow --model=/Users/dev01/下载目录/测试model.pb/Angel_a025_c256_s512_a0.25_b2_sn16-32000_frozen.pb --save_dir=/Users/dev01/下载目录/paddle
Now translating model from tensorflow to paddle.
Total nodes: 468
Converting node 468 ... ==========3 Ops are not supported yet======
========== Sqrt ==========
========== MirrorPad ==========
========== Square ==========

tf model convert error

tf model convert error: Exception: OP: [ResizeBilinear] not support yet

don't support torch.tensor.expand?

when i use pytorch torch.tensor.expand, the error

Traceback (most recent call last):
  File "/root/miniconda3/envs/torch120/bin/onnx_infer", line 11, in <module>
    load_entry_point('x2paddle==0.5.0', 'console_scripts', 'onnx_infer')()
  File "/root/miniconda3/envs/torch120/lib/python3.6/site-packages/x2paddle-0.5.0-py3.6.egg/x2paddle/onnx_infer.py", line 48, in main
    res = sess.run(None, input_feed=inputs_dict)
  File "/root/miniconda3/envs/torch120/lib/python3.6/site-packages/onnxruntime/capi/session.py", line 72, in run
    return self._sess.run(output_names, input_feed, run_options)
RuntimeError: Method run failed due to: [ONNXRuntimeError] : 1 : GENERAL ERROR : /onnxruntime_src/onnxruntime/core/providers/cpu/math/element_wise_ops.h:341 void onnxruntime::BroadcastIterator::Init(int64_t, int64_t) axis == 1 || axis == largest was false. Attempting to broadcast an axis by a dimension other than 1. -1 by 264
Stacktrace:

There are 1 ops not supported yet, list as below
Expand
Traceback (most recent call last):
  File "/root/miniconda3/envs/torch120/bin/x2paddle", line 11, in <module>
    load_entry_point('x2paddle==0.5.0', 'console_scripts', 'x2paddle')()
  File "/root/miniconda3/envs/torch120/lib/python3.6/site-packages/x2paddle-0.5.0-py3.6.egg/x2paddle/convert.py", line 211, in main
    onnx2paddle(args.model, args.save_dir)
  File "/root/miniconda3/envs/torch120/lib/python3.6/site-packages/x2paddle-0.5.0-py3.6.egg/x2paddle/convert.py", line 157, in onnx2paddle
    mapper = ONNXOpMapper(model, save_dir)
  File "/root/miniconda3/envs/torch120/lib/python3.6/site-packages/x2paddle-0.5.0-py3.6.egg/x2paddle/op_mapper/onnx_op_mapper.py", line 68, in __init__
    raise Exception("Model are not supported yet.")
Exception: Model are not supported yet.

it seems don't support expand?

加载转换好的模型时报错Enforce failed. Expected version == 0U, but received version:5 != 0U:0.

完整报错如下：

Traceback (most recent call last):
File "E:/CNNNET/paddle/tensorflow2fluid/tf2fluid/predict.py", line 6, in
model = ml.ModelLoader("translated_paddle_model", use_cuda=True)
File "E:\CNNNET\paddle\tensorflow2fluid\tf2fluid\model_loader.py", line 42, in init
fluid.io.load_vars(self.exe, model_dir, vars=var_list)
File "D:\Anaconda3\lib\site-packages\paddle\fluid\io.py", line 610, in load_vars
executor.run(load_prog)
File "D:\Anaconda3\lib\site-packages\paddle\fluid\executor.py", line 565, in run
use_program_cache=use_program_cache)
File "D:\Anaconda3\lib\site-packages\paddle\fluid\executor.py", line 642, in _run
exe.run(program.desc, scope, 0, True, True, fetch_var_name)
paddle.fluid.core.EnforceNotMet: Invoke operator load error.
Python Callstacks:
File "D:\Anaconda3\lib\site-packages\paddle\fluid\framework.py", line 1654, in append_op
attrs=kwargs.get("attrs", None))
File "D:\Anaconda3\lib\site-packages\paddle\fluid\io.py", line 596, in load_vars
attrs={'file_path': os.path.join(dirname, new_var.name)})
File "E:\CNNNET\paddle\tensorflow2fluid\tf2fluid\model_loader.py", line 42, in init
fluid.io.load_vars(self.exe, model_dir, vars=var_list)
File "E:/CNNNET/paddle/tensorflow2fluid/tf2fluid/predict.py", line 6, in
model = ml.ModelLoader("translated_paddle_model", use_cuda=True)
C++ Callstacks:
Enforce failed. Expected version == 0U, but received version:5 != 0U:0.
Only version 0 is supported at [D:/1.4.1/paddle/paddle/fluid/framework/.tensor_util.cu:453]
PaddlePaddle Call Stacks:
Windows not support stack backtrace yet.
W0508 10:14:41.673883 22764 device_context.cc:261] Please NOTE: device: 0, CUDA Capability: 61, Driver API Version: 9.2, Runtime API Version: 8.0
W0508 10:14:41.679867 22764 device_context.cc:269] device: 0, cuDNN Version: 7.0.

使用x2paddle转caffe模型后，paddle1.6.1的C++ inference速度慢了20%以上

caffe模型选用x2paddle_model_zoo
中的shufflenetv2

转模型后，使用paddle1.6.1的 cpu-mkl版本进行预测，比caffe的模型（cpu端）慢20%以上。

AttributeError: module 'model' has no attribute 'x2paddle_net'

lib/python3.6/site-packages/x2paddle-0.6.0-py3.6.egg/x2paddle/core/op_mapper.py", line 122, in save_inference_model
    inputs, outputs = model.x2paddle_net()
AttributeError: module 'model' has no attribute 'x2paddle_net'

During handling of the above exception, another exception occurred:

Traceback (most recent call last):
  File "bin/x2paddle", line 11, in <module>
    load_entry_point('x2paddle==0.6.0', 'console_scripts', 'x2paddle')()
  File "/lib/python3.6/site-packages/x2paddle-0.6.0-py3.6.egg/x2paddle/convert.py", line 233, in main
    onnx2paddle(args.model, args.save_dir, params_merge)
  File "/lib/python3.6/site-packages/x2paddle-0.6.0-py3.6.egg/x2paddle/convert.py", line 176, in onnx2paddle
    mapper.save_inference_model(save_dir, params_merge)
  File "lib/python3.6/site-packages/x2paddle-0.6.0-py3.6.egg/x2paddle/core/op_mapper.py", line 158, in save_inference_model
    .format(py_code_dir))
Exception: Paddle code was saved in ./model.py, but seems there's wrong exist, please check model.py manually.

tf.tensordot和tf.einsum有没有相应的paddle实现方案？

paddle中如何实现tensor乘？

使用paddle训练后的paddle模型可以转caffe模型吗？

使用paddle训练后的paddle模型可以转caffe模型吗？

ValueError: Object arrays cannot be loaded when allow_pickle=False

我是要做caffe2paddle，
使用convert.py转换那一步成功，想把npy参数和py网络转成paddle可读取的模型文件时出错，
我看问题是出在
data_dict = np.load(data_path).item()
因为这是生成的net.py里的，所以哪一行可能不确定，错误内容基本上是说np.load的时候设置了allow_pickle=False，这可能是默认参数。修改为
data_dict = np.load(data_path, allow_pickle=True).item()之后就可以成功导出模型了。
导出的只有两个文件，一个叫model，一个叫params。

Did you have support for Paddle2X ?

Did you have support for Paddle2X ?

don't support onnx::gather?

the onnx model used onnx::gather, detail:

%position_enc.weight : Float(264, 1024)):
  %2 : Float(1, 30, 1024) = onnx::Gather(%position_enc.weight, %x)

but it report an error:

Traceback (most recent call last):
  File "/root/miniconda3/envs/torch120/bin/x2paddle", line 11, in <module>
    load_entry_point('x2paddle==0.5.0', 'console_scripts', 'x2paddle')()
  File "/root/miniconda3/envs/torch120/lib/python3.6/site-packages/x2paddle-0.5.0-py3.6.egg/x2paddle/convert.py", line 211, in main
    onnx2paddle(args.model, args.save_dir)
  File "/root/miniconda3/envs/torch120/lib/python3.6/site-packages/x2paddle-0.5.0-py3.6.egg/x2paddle/convert.py", line 157, in onnx2paddle
    mapper = ONNXOpMapper(model, save_dir)
  File "/root/miniconda3/envs/torch120/lib/python3.6/site-packages/x2paddle-0.5.0-py3.6.egg/x2paddle/op_mapper/onnx_op_mapper.py", line 81, in __init__
    func(node)
  File "/root/miniconda3/envs/torch120/lib/python3.6/site-packages/x2paddle-0.5.0-py3.6.egg/x2paddle/op_mapper/onnx_op_mapper.py", line 500, in Gather
    indices_shape) <= 1, "Gather op don't support dim of indice >1 "
AssertionError: Gather op don't support dim of indice >1

the pytorch code is:

class test_model(nn.Module):
    def __init__(self, src_n_position=264, d_word_vec=1024):
        super().__init__()
        self.position_enc = nn.Embedding(src_n_position, d_word_vec)
        
    def forward(self, x):
        x = self.position_enc(x)
        
        return x
    
net2 = test_model()
x = torch.randint(30, (1, 30))
torch.onnx.export(net2, x, './Fonnx/tranformertwo.onnx', verbose=True, input_names=['x'])

pytorch GRU转paddle 报错

先用pytorch写了一个gru,  转成onnx, 再使用X2Paddle/onnx2fluid转的时候，报错onnx2fluid/writer.py", line 247, in Op ValueError: conversion for ::ConstantFill not supported

使用x2paddle转换之后的模型进行预测，是否必须要求预测库paddle>1.6.0

如题

使用x2paddle对caffe转模型后，利用paddle预测库进行预测，是否要求预测库paddle>1.6.0

x2paddle transfer model of tensorflow, report error.

(venv-python36) PS D:\tensorflow-learning\venv-python36\Scripts> x2paddle -m D:\work\AI\model\mobilenet_v1_1.0_224_frozen\mobilenet_v1_1.0_224_frozen.pb -s D:\work\AI\PaddlePaddleLite -f tensorflow -f tensorflow

paddlepaddle not installed, use "pip install paddlepaddle"

1.0.0<=tensorflow<2.0.0 is required, and v1.14.0 is recommended

but paddlepaddle installed . And use "pip list" can see : "paddlepaddle 1.6.1"

X2Paddle 支持 tensorflow 2.0 的模型转换吗？

'TENSOR_TYPE_TO_NP_TYPE'

x2paddle/op_mapper/onnx_op_mapper.py line511
NameError: name 'TENSOR_TYPE_TO_NP_TYPE' is not defined

tensorflow2fluid转出的模型Inference时间不正常

tensorflow2fluid将VGG16转成paddle格式，如实例代码。
转出的模型如下：

按照示例的Inference代码进行ml.ModelLoader后，分别进行100次Inference。
GPU：

CPU：

@jiangjiajun @SunAhong1993 你们看一下是不是有这个问题，这Inference时间不正常。

pb模型无法读取input tensor name

已训练好的pb模型可以在tensorflow的sess.run中正常调用，在feed_dict字典中的input为
self.graph.get_tensor_by_name('prefix/input_1:0')
在convert.py中设置in_nodes为"prefix/input_1:0"，报错如下

修改in_nodes为‘input_1’，报错相同。

onnx转paddle，Upsample动态shape

模型转换Pytorch->onnx->paddle。

1. Pytorch代码中对应部分代码：

def forward(self, x):
    n,c,h,w = x.shape
    up = nn.functional.interpolate(rb2, size=((h+1)//2, (w+1)//2), mode='bilinear', align_corners=False) 
    return up

2. 转onnx时候，Upsample对应层是：
   %827 : Float(1, 576, 4, 35) = onnx::Upsample[mode="linear"](%808, %826)，其中%826是计算动态shape

3. 使用develop分支的X2Paddle转，在model = ONNXDecoder(model_path)这步里有一个model = onnx.shape_inference.infer_shapes(model)，onnx貌似不支持动态的shape推断，打印出来Upsample的四个dim都是空

4. 由于3中所述，在onnx转paddle的时，Upsample转成paddle的out_shape=[0,0]，转出来的model.py对应层是_827 = fluid.layers.resize_bilinear(_808, scale=None, out_shape=[0, 0], name='_827')，这里手动把2中的mode=“linear”改成了paddle支持的“bilinear”。

我看到paddle的文档中写了“根据指定的out_shape执行双线性插值调整输入大小，输出形状按优先级由actual_shape、out_shape和scale指定。”其中acutal_shape是动态图时用到的。
我现在想法是在ONNXOpMapper这里改，Upsample层对应的是_interpolate函数，但不清楚用actual_shape参数要怎么改，或者有更方便的改法嘛~
其中_interpolate函数里层的现在是提供scale和out_shape的参数，如下：

        attr = {
            'scale': scale,
            'out_shape': out_shape,
            'name': string(node.layer_name)
        }
        node.fluid_code.add_layer(fluid_op,
                                  inputs=val_x,
                                  output=node,
                                  param_attr=attr)

加载转换好的模型时报错Enforce failed. Expected version == 0U, but received version:5 != 0U:0.

已有issue，#24 ，但未给出Windows下的解决方案，请教下。

google.protobuf.message.DecodeError: Error parsing message

Traceback (most recent call last):
File "/usr/local/bin/x2paddle", line 11, in
load_entry_point('x2paddle==0.6.0', 'console_scripts', 'x2paddle')()
File "/usr/local/lib/python3.5/dist-packages/x2paddle-0.6.0-py3.5.egg/x2paddle/convert.py", line 227, in main
args.caffe_proto, params_merge)
File "/usr/local/lib/python3.5/dist-packages/x2paddle-0.6.0-py3.5.egg/x2paddle/convert.py", line 145, in caffe2paddle
model = CaffeDecoder(proto, weight, caffe_proto)
File "/usr/local/lib/python3.5/dist-packages/x2paddle-0.6.0-py3.5.egg/x2paddle/decoder/caffe_decoder.py", line 228, in init
self.load_using_pb()
File "/usr/local/lib/python3.5/dist-packages/x2paddle-0.6.0-py3.5.egg/x2paddle/decoder/caffe_decoder.py", line 237, in load_using_pb
data.MergeFromString(open(self.model_path, 'rb').read())
google.protobuf.message.DecodeError: Error parsing message

This issue happen when convert a caffe model to paddle paddle model, any specified requirement on google protobuf? I am installing the python3 protobuf.

AssertionError: The parameter of my_classifier (type is InnerProduct) is not set. You need to use python package of caffe to set the default value.

转换resnet50模型(来自caffe-model-zoo)，报错

register layer[ROIPooling]
register layer[PriorBox]
register layer[Permute]
register layer[DetectionOutput]
register layer[Normalize]
register layer[Select]
register layer[ShuffleChannel]
register layer[ConvolutionDepthwise]
register layer[Axpy]
Now translating model from caffe to paddle.
cost: 1.406264305114746
Ignoring parameters for non-existent layer: fc1000
Total nodes: 229
squeeze idx:1, with kind:Convolution,name:conv1
Traceback (most recent call last):
  File "/usr/local/bin/x2paddle", line 9, in <module>
    load_entry_point('x2paddle==0.4.5', 'console_scripts', 'x2paddle')()
  File "/home/***/CNN/X2Paddle/x2paddle/convert.py", line 211, in main
    args.caffe_proto)
  File "/home/***/CNN/X2Paddle/x2paddle/convert.py", line 137, in caffe2paddle
    mapper = CaffeOpMapper(model)
  File "/home/***/CNN/X2Paddle/x2paddle/op_mapper/caffe_op_mapper.py", line 46, in __init__
    func(node)
  File "/home/***/CNN/X2Paddle/x2paddle/op_mapper/caffe_op_mapper.py", line 346, in InnerProduct
    node.layer_name, node.layer_type)
AssertionError: The parameter of my_classifier (type is InnerProduct) is not set. You need to use python package of caffe to set the default value.

转换用的命令为:

x2paddle --framework=caffe --prototxt=resnet50.prototxt --weight=resnet50.caffemodel --save_dir=pd_model  --caffe_proto=***/build/include/caffe/proto/caffe_pb2.py

0.3之前用的caffe2fluid还好好的呢，怎么升级了就不兼容了？

develop分支能否支持单向GRU呢

我看到paddle实现双向GRU的方式是：

    fc_1 = fluid.layers.fc(input=x,
                           size=rnn_hidden_size * 3,
                           param_attr=para_attr,
                           bias_attr=bias_attr_nobias)
    fc_2 = fluid.layers.fc(input=x,
                           size=rnn_hidden_size * 3,
                           param_attr=para_attr,
                           bias_attr=bias_attr_nobias)
    gru_forward = fluid.layers.dynamic_gru(
        input=fc_1,
        size=rnn_hidden_size,
        param_attr=para_attr,
        bias_attr=bias_attr,
        candidate_activation='relu')
    gru_backward = fluid.layers.dynamic_gru(
        input=fc_2,
        size=rnn_hidden_size,
        is_reverse=True,
        param_attr=para_attr,
        bias_attr=bias_attr,
        candidate_activation='relu')

那对于单向的Pytorch上的代码是：

model = nn.GRU(10, 20, 2)
input = torch.randn(5, 3, 10)
#h0 = torch.zeros(2, 3, 20)
yp = model(input)
torch.onnx.export(model, input, 'gru.onnx',verbose=True)

转后的onnx模型是
链接：https://pan.baidu.com/s/1NtYNN5DSALo-tLrGFDlGxA
提取码：897a

想问一下目前develop分支能否支持单向GRU

ConstantFill op的转换失败

使用X2Paddle/onnx2fluid转一个gru的时候，报错onnx2fluid/writer.py", line 247, in Op
ValueError: conversion for ::ConstantFill not supported。希望能够支持对ConstantFill op的转换。

Multiple top nodes are not supported.

I am converting caffe slice layer to paddle. Slice layer has Multiple top nodes but X2Paddle does not support it.

pytorch ssd模型导出onnx 转paddle出错

Now translating model from onnx to paddle.
model ir_version: 4, op version: 9
Total nodes: 98
Traceback (most recent call last):
File "/home/zt/androidprojects/X2Paddle-develop/x2paddle/core/op_mapper.py", line 122, in save_inference_model
inputs, outputs = model.x2paddle_net()
File "pd_model/model_with_code/model.py", line 7, in x2paddle_net
_107 = fluid.layers.create_parameter(dtype='float32', shape=[], name='_107', attr='_107', default_initializer=Constant(0.0))
File "/home/zt/tools/anaconda3/envs/pytorch1.3/lib/python3.6/site-packages/paddle/fluid/layers/tensor.py", line 103, in create_parameter
default_initializer)
File "/home/zt/tools/anaconda3/envs/pytorch1.3/lib/python3.6/site-packages/paddle/fluid/layer_helper_base.py", line 330, in create_parameter
**attr._to_kwargs(with_initializer=True))
File "/home/zt/tools/anaconda3/envs/pytorch1.3/lib/python3.6/site-packages/paddle/fluid/framework.py", line 2384, in create_parameter
param = Parameter(global_block, *args, **kwargs)
File "/home/zt/tools/anaconda3/envs/pytorch1.3/lib/python3.6/site-packages/paddle/fluid/framework.py", line 4471, in init
"The dimensions of shape for Parameter must be greater than 0")
ValueError: The dimensions of shape for Parameter must be greater than 0

During handling of the above exception, another exception occurred:

Traceback (most recent call last):
File "/home/zt/tools/anaconda3/envs/pytorch1.3/bin/x2paddle", line 11, in
load_entry_point('x2paddle==0.6.0', 'console_scripts', 'x2paddle')()
File "/home/zt/androidprojects/X2Paddle-develop/x2paddle/convert.py", line 233, in main
onnx2paddle(args.model, args.save_dir, params_merge)
File "/home/zt/androidprojects/X2Paddle-develop/x2paddle/convert.py", line 176, in onnx2paddle
mapper.save_inference_model(save_dir, params_merge)
File "/home/zt/androidprojects/X2Paddle-develop/x2paddle/core/op_mapper.py", line 158, in save_inference_model
.format(py_code_dir))
Exception: Paddle code was saved in pd_model/model_with_code/model.py, but seems there's wrong exist, please check model.py manually.

onnx文件先经过onnx-simplifier-master工具转化过,出错的地方为L2Norm部分,不知道是不是因为_107处常量没有scope.

%106 : Float(1, 1024, 19, 19) = onnx::Relu(%105), scope: SSD/VGG16[base]/ReLU # /home/zt/tools/anaconda3/envs/pytorch1.3/lib/python3.6/site-packages/torch/nn/functional.py:914:0
%107 : Float() = onnx::Constantvalue={2}
%108 : Float(1, 512, 38, 38) = onnx::Pow(%94, %107), scope: SSD/L2Norm[L2Norm] # /home/zt/PycharmProjects/fssd.pytorch-master/models/SSD_vgg.py:268:0

onnx模型转paddle，batch normalization为啥没了

转出来的模型，BN没了。。。。 而且使用X2Paddle的时候，报segment fault，卸载掉paddle就好了~~

caffe2fluid模型加载与预测代码问题

将Caffe模型转成Paddle模型，参考https://github.com/PaddlePaddle/X2Paddle/tree/master/caffe2fluid，执行完第一步和第二步后，生成的文件如下：

文档上写模型的加载及预测可参考PaddlePaddle官方文档： https://www.paddlepaddle.org.cn/documentation/docs/zh/1.3/api_guides/low_level/inference.html#id4
但你们参考的文档上模型目录下需要包含 ./infer_model/ __ model __ 文件，Caffe模型转Paddle模型的文件夹中并没有该文件。请问怎么解决，或者有这个Inference的代码参考一下吗？
@SunAhong1993 @jiangjiajun

在含有paddle的docker中安装tensor1.14.0后，使用x2paddle工具还是提示未安装tensorflow

don't support onnx::equal and onnx::where?

it seems don't support equal and where operater
the error:

There are 2 ops not supported yet, list as below
Where
Equal
Traceback (most recent call last):
  File "/root/miniconda3/envs/torch120/bin/x2paddle", line 11, in <module>
    load_entry_point('x2paddle==0.5.0', 'console_scripts', 'x2paddle')()
  File "/root/miniconda3/envs/torch120/lib/python3.6/site-packages/x2paddle-0.5.0-py3.6.egg/x2paddle/convert.py", line 215, in main
    onnx2paddle(args.model, args.save_dir)
  File "/root/miniconda3/envs/torch120/lib/python3.6/site-packages/x2paddle-0.5.0-py3.6.egg/x2paddle/convert.py", line 161, in onnx2paddle
    mapper = ONNXOpMapper(model, save_dir)
  File "/root/miniconda3/envs/torch120/lib/python3.6/site-packages/x2paddle-0.5.0-py3.6.egg/x2paddle/op_mapper/onnx_op_mapper.py", line 76, in __init__
    raise Exception("Model are not supported yet.")
Exception: Model are not supported yet.