执行 paddle serving 报错

[zouxiaoshi]

问题：

Q1:

执行如下代码时报错：

export SERVING_BIN=/usr/local/serving_bin/serving
python -m paddle_serving_server.serve \
--model ./serving_server \
--thread 8 --port 10010 \
--gpu_ids 0 

错误信息：

Error Message Summary:
----------------------
NotFoundError: Cannot open file ./serving_server/__model__, please confirm whether the file is normal.
  [Hint: Expected static_cast<bool>(fin.is_open()) == true, but received static_cast<bool>(fin.is_open()):0 != true:1.] (at /paddle/paddle/fluid/inference/api/analysis_predictor.cc:1119)

后用通过如下代码 进行转换：

python -m paddle_serving_client.convert --dirname . \
                                         --model_filename model.pdmodel          \
                                         --params_filename model.pdiparams       \
                                         --serving_server ./serving_server/ \
                                         --serving_client ./serving_client/

得到 如下文件：

.
├── model.pdiparams
├── model.pdmodel
├── serving_server_conf.prototxt
└── serving_server_conf.stream.prototxt

Q2:

强制 对 model.pdmodel 重命名， mv model.pdmodel __model__
然后启动 paddle serving 服务，得到如下错误：

SOLOv2 模型

Error Message Summary:
----------------------
UnavailableError: Load operator fail to open file ./serving_server/sync_batch_norm_48.w_1, please check whether the model file is complete or damaged.
  [Hint: Expected static_cast<bool>(fin) == true, but received static_cast<bool>(fin):0 != true:1.] (at /paddle/paddle/fluid/operators/load_op.h:41)
  [operator < load > error]

Yolov3 模型

Error Message Summary:
----------------------
UnavailableError: Load operator fail to open file ./serving_server/batch_norm_41.b_0, please check whether the model file is complete or damaged.
  [Hint: Expected static_cast<bool>(fin) == true, but received static_cast<bool>(fin):0 != true:1.] (at /paddle/paddle/fluid/operators/load_op.h:41)
  [operator < load > error]

YOLO V3 模型在 以下环境下运行是可以的：

paddle-serving-app        0.6.1
paddle-serving-client     0.6.1
paddle-serving-server-gpu 0.6.1.post102
paddlepaddle-gpu          2.1.0

环境

paddle-serving-app        0.7.0
paddle-serving-client     0.7.0
paddle-serving-server-gpu 0.7.0.post102
paddlepaddle-gpu          2.2.0

cuda 10.2
Tesla V100
python 3.8

[bjjwwang]

这个SERVING_BIN是从哪里来的

[TeslaZhao]

fail to open file ./serving_server/batch_norm_41.b_0

从报错信息上看，你的模型是散列多文件的？

[zouxiaoshi]

这个SERVING_BIN是从哪里来的

是从

Serving/tools/dockerfiles/build_scripts/install_whl.sh

Line 46 in 74a0315

serving_bin="https://paddle-serving.bj.bcebos.com/test-dev/bin/serving-gpu-1028-$SERVING_VERSION.tar.gz"

得来的。 把其中的 serving_version 换成 0.7.0

[zouxiaoshi]

fail to open file ./serving_server/batch_norm_41.b_0
从报错信息上看，你的模型是散列多文件的？

您好。
用的paddle detection上Yolo模型的脚本，训练后，导出的信息如下：
.
├── infer_cfg.yml
├── model.pdiparams
├── model.pdiparams.info
└── model.pdmodel

再通过 如下命令 得到 serving_server 以及 serving_client

python -m paddle_serving_client.convert --dirname . \
                                         --model_filename model.pdmodel          \
                                         --params_filename model.pdiparams       \
                                         --serving_server ./serving_server/ \
                                         --serving_client ./serving_client/

seving_server 信息如下：

.
├── fluid_time_file
├── model.pdiparams
├── model.pdmodel
├── serving_server_conf.prototxt
└── serving_server_conf.stream.prototxt

发生以下报错后

Error Message Summary:
----------------------
NotFoundError: Cannot open file ./serving_server/__model__, please confirm whether the file is normal.
  [Hint: Expected static_cast<bool>(fin.is_open()) == true, but received static_cast<bool>(fin.is_open()):0 != true:1.] (at /paddle/paddle/fluid/inference/api/analysis_predictor.cc:1119)

改名model: model.pdmodel ==> __model__
再执行，得到如下报错：

Error Message Summary:
----------------------
UnavailableError: Load operator fail to open file ./serving_server/batch_norm_41.b_0, please check whether the model file is complete or damaged.
  [Hint: Expected static_cast<bool>(fin) == true, but received static_cast<bool>(fin):0 != true:1.] (at /paddle/paddle/fluid/operators/load_op.h:41)
  [operator < load > error]

[bjjwwang]

好的 了解 我这边复现一下 大概2小时后给个结论。

[bjjwwang]

抱歉 我这里没能复现。
我在思考是不是SERVING_BIN的版本问题。
可以unset SERVING_BIN再运行一下

python -m paddle_serving_server.serve \
--model ./serving_server \
--thread 8 --port 10010 \
--gpu_ids 0 

[wenjia322]

@zouxiaoshi @bjjwwang 抱歉打扰了，我遇到了同样的问题，请教一下当时是怎么解决的呢？

我的操作步骤是：

1. 按照 安装文档 的1.2及2.1安装好了环境，并且在3环境检查时都是成功的。
2. 按照 模型转换文档 下载了PaddleOCR的模型，但转换后的 ppocr_det_v3_serving 文件夹下并不是__model__ 和__params__文件名，而是 inference.pdmodel 和 inference.pdiparams ，这是为什么？
3. 在执行：

python3 -m paddle_serving_server.serve --model ppocr_det_v3_serving  --port 8181

命令时，报错如下：

----------------------
Error Message Summary:
----------------------
NotFoundError: Cannot open file ppocr_det_v3_serving/__model__, please confirm whether the file is normal.
  [Hint: Expected static_cast<bool>(fin.is_open()) == true, but received static_cast<bool>(fin.is_open()):0 != true:1.] (at /paddle/paddle/fluid/inference/api/analysis_predictor.cc:1452)

请问该如何解决？

相关环境

使用镜像
registry.baidubce.com/paddlepaddle/paddle:2.3.0

其它

Python 3.7.13
paddle-serving-app    0.9.0
paddle-serving-client 0.9.0
paddle-serving-server 0.9.0
paddlepaddle          2.3.0