使用同样的模型，cpp_infer推理比python PaddleOCR推理慢非常多，且结果不一致

[Phymin]

环境：win10 pro，未使用gpu， cpp_infer: v2.10, paddle_inference: CPU-vs2019-mkl
模型：检测：ch_PP-OCRv4_det_server_infer， 识别：ch_PP-OCRv4_rec_server_infer
在使用python脚本推理的时候：检测+识别耗时：93s左右
在使用cpp_infer推理的时候：检测+识别耗时：156s左右
python 推理脚本：

from paddleocr import PaddleOCR, draw_ocr
import time

ocr = PaddleOCR(lang='ch', det_model_dir='D:/Vision_Tech/ppocr/models/ch_PP-OCRv4_det_server_infer',  rec_model_dir='D:/Vision_Tech/ppocr/models/ch_PP-OCRv4_rec_server_infer') 
# need to run only once to download and load model into memory
img_path = 'D:/ocr_images/template.png'

st = time.time()
result = ocr.ocr(img_path, cls=False)
for idx in range(len(result)):
    res = result[idx]
    for line in res:
        print(line)
cost = time.time() - st

print("ocr cost time: ", cost)

# draw result
from PIL import Image
result = result[0]
image = Image.open(img_path).convert('RGB')
boxes = [line[0] for line in result]
txts = [line[1][0] for line in result]
scores = [line[1][1] for line in result]
im_show = draw_ocr(image, boxes, txts, scores, font_path='/path/to/PaddleOCR/doc/fonts/simfang.ttf')
im_show = Image.fromarray(im_show)
im_show.save('result.jpg')

cpp_infer推理 c++代码：

   FLAGS_use_gpu = false;
    FLAGS_gpu_id = 0;
    FLAGS_gpu_mem = 4000;
    FLAGS_cpu_threads = 10;

    FLAGS_enable_mkldnn = true;
    FLAGS_det_model_dir = "D:/Vision_Tech/ppocr/models/ch_PP-OCRv4_det_server_infer";
    FLAGS_rec_model_dir = "D:/Vision_Tech/ppocr/models/ch_PP-OCRv4_rec_server_infer";
    FLAGS_rec_char_dict_path = "D:/Vision_Tech/ppocr/ppocr_keys_v1.txt";

    PaddleOCR::PPOCR ocrImpl;
    std::vector<PaddleOCR::OCRPredictResult> ocr_results =
        ocrImpl.ocr(srcImg, FLAGS_det, FLAGS_rec, FLAGS_cls);

python 推理结果：

[[[30.0, 25.0], [146.0, 25.0], [146.0, 47.0], [30.0, 47.0]], ('(P)HUAWEI', 0.9205098748207092)]
[[[155.0, 24.0], [207.0, 24.0], [207.0, 47.0], [155.0, 47.0]], ('P/N:', 0.9215103387832642)]
[[[155.0, 24.0], [207.0, 24.0], [207.0, 47.0], [155.0, 47.0]], ('P/N:', 0.9215103387832642)]
[[[273.0, 24.0], [378.0, 24.0], [378.0, 46.0], [273.0, 46.0]], ('08070612', 0.9954185485839844)]
[[[33.0, 93.0], [129.0, 93.0], [129.0, 111.0], [33.0, 111.0]], ('（1P）Mfr', 0.8607510328292847)]
[[[273.0, 24.0], [378.0, 24.0], [378.0, 46.0], [273.0, 46.0]], ('08070612', 0.9954185485839844)]
[[[33.0, 93.0], [129.0, 93.0], [129.0, 111.0], [33.0, 111.0]], ('（1P）Mfr', 0.8607510328292847)]
[[[127.0, 90.0], [181.0, 90.0], [181.0, 112.0], [127.0, 112.0]], ('P/N:', 0.8674935698509216)]
[[[192.0, 87.0], [420.0, 91.0], [420.0, 115.0], [192.0, 112.0]], ('CL05A105KP5NNXC', 0.9789352416992188)]
[[[30.0, 160.0], [220.0, 160.0], [220.0, 180.0], [30.0, 180.0]], ('(10D）DateCode', 0.9151744842529297)]
[[[263.0, 159.0], [337.0, 156.0], [338.0, 179.0], [263.0, 181.0]], ('191101', 0.9944730401039124)]
[[[31.0, 227.0], [167.0, 227.0], [167.0, 244.0], [31.0, 244.0]], ('（1T）Lot#', 0.8181165456771851)]
[[[164.0, 226.0], [262.0, 226.0], [262.0, 244.0], [164.0, 244.0]], ('CF9HT5P', 0.9880648851394653)]
[[[30.0, 291.0], [199.0, 291.0], [199.0, 311.0], [30.0, 311.0]], ('(Q)QTY:10.000', 0.9011315703392029)]
[[[529.0, 300.0], [583.0, 300.0], [583.0, 317.0], [529.0, 317.0]], ('pass', 0.9919123649597168)]
[[[480.0, 318.0], [582.0, 320.0], [581.0, 342.0], [480.0, 339.0]], ('Rohs', 0.9899167418479919)]
[[[528.0, 343.0], [583.0, 342.0], [584.0, 360.0], [528.0, 361.0]], ('pass', 0.9922510385513306)]
[[[168.0, 359.0], [279.0, 359.0], [279.0, 376.0], [168.0, 376.0]], ('(V）Vendor', 0.902154803276062)]
[[[292.0, 356.0], [450.0, 354.0], [451.0, 375.0], [292.0, 377.0]], ('Code：Z00MCD', 0.8026291131973267)]
[[[28.0, 375.0], [131.0, 373.0], [131.0, 394.0], [29.0, 397.0]], ('SAMSUNG', 0.9068806767463684)]

cpp_infer推理结果：

0	det boxes: [[155,24],[207,24],[207,46],[155,46]] rec text: P/N rec score: 0.985921 
1	det boxes: [[273,24],[377,24],[377,45],[273,45]] rec text: 08070612 rec score: 0.988242 
2	det boxes: [[191,87],[420,90],[420,114],[191,111]] rec text: CCL05A105KP5NNXC rec score: 0.964699 
3	det boxes: [[30,25],[146,25],[146,46],[30,46]] rec text: (P)HUAWEI rec score: 0.885605 
4	det boxes: [[33,92],[128,92],[128,110],[33,110]] rec text: 锛P)Mfr rec score: 0.784595 
5	det boxes: [[126,90],[180,90],[180,111],[126,111]] rec text: P/N rec score: 0.953819 
6	det boxes: [[30,159],[220,159],[220,179],[30,179]] rec text: (10D锛塂ate Code rec score: 0.908462 
7	det boxes: [[262,158],[337,156],[337,178],[262,180]] rec text: 191101 rec score: 0.947901 
8	det boxes: [[31,226],[167,226],[167,244],[31,244]] rec text: 锛T锛塋ot#: rec score: 0.808263 
9	det boxes: [[164,225],[261,225],[261,244],[164,244]] rec text: CF9HT5P rec score: 0.984709 
10	det boxes: [[30,290],[198,290],[198,311],[30,311]] rec text: (Q)QTY:10.000 rec score: 0.888378 
11	det boxes: [[528,299],[582,299],[582,317],[528,317]] rec text: pass rec score: 0.972611 
12	det boxes: [[480,318],[581,320],[581,341],[480,339]] rec text: Rohs rec score: 0.974954 
13	det boxes: [[527,342],[583,341],[583,359],[527,360]] rec text: pass rec score: 0.981898 
14	det boxes: [[168,358],[279,358],[279,376],[168,376]] rec text: (V)Vendor rec score: 0.879852 
15	det boxes: [[292,355],[450,353],[450,375],[292,377]] rec text: Code:Z0eMCD rec score: 0.787138 
16	det boxes: [[28,375],[130,373],[130,394],[28,396]] rec text: SAMSUNG rec score: 0.928704 

可以看到第15个字符识别的结果差异很大

识别的图片请见附件

可以解释一下造成这种差异可能的原因吗？谢谢

[GreatV]

根据你提供的信息，可以归纳出以下两个主要问题：

一、CPP 推理比 Python 慢，差异很大（156s vs 93s）
二、推理结果存在明显差异（如 Code:Z00MCD 被识别为 Code:Z0eMCD）

▍一、CPP 推理比 Python 慢很多的可能原因

1. CPU线程设置是否生效
   你设置了 FLAGS_cpu_threads=10，但 Paddle Inference 的线程设置不一定默认生效。请确认在初始化 Place 和 Config 时线程数确实被正确设置。在 C++ 使用中需要调用以下方式设置：

config.SetCpuMathLibraryNumThreads(10);

此外，建议查看是否在 config 中启用了 config.EnableMKLDNN()，确保 MKL-DNN 被合理使用。

2. MKLDNN cache 和其他优化策略
   Python PaddleOCR 自带预热和动态图缓存，可以缓存部分计算图。而 C++ 端如果没有手动设置 enable_ir_optim 和开启缓存，性能会受影响：

config.SwitchIrOptim(true);
config.EnableMKLDNN();
config.SetMkldnnCacheCapacity(CAPACITY); // 例如设置为10或更多

3. 输入图像解码和预处理耗时差异
   Python PaddleOCR 会使用 Pillow 和 numpy 进行 SIMD 优化解析图片，而 C++ 端未必（如使用 opencv imread）。建议你在推理前分别打印模型输入大小及耗时，明确瓶颈在哪。

4. 文件I/O重复读取
   检查你的 C++ 实现是否每次都重新加载模型。Python 中 PaddleOCR 会在初始化时加载模型并常驻，而 C++ 示例中 PPOCR ocrImpl 可能每次图片推理都重新初始化模型。

5. 编译器优化参数
   检查你用 VS2019 编译 C++ 时是否启用 O2 或 Ox 优化，以及是否用 Release 模式。如果使用 Debug 模式，性能通常会慢上很多。

▍二、识别结果不一致的可能原因分析

部分字符差异较大，如 Code:Z00MCD 识别为 Code:Z0eMCD，可分为以下几类原因：

1. 图像预处理和切割存在差异
   Python 与 C++ 中图像旋转、缩放或仿射处理可能略微不同，特别是文字 box 的仿射裁剪是否精确、padding 是否一致不明确。

例如，PaddleOCR Python 中采用了 align_img() 和 padding64 方法确保图像切片为宽高的倍数，这部分有条件剪裁：

https://github.com/PaddlePaddle/PaddleOCR/blob/release/2.6/tools/infer/utility.py#L278

C++ 实现中如果未严格对齐，则释出图片质量或清晰度可能下降，影响识别精度。

2. 字典文件可能不同
   虽然你在 cpp 端设置了 FLAGS_rec_char_dict_path，但需要在 rec_model_dir 中核实字符集及顺序与你 Python 使用的一致。如有不同，例如缺失'：'等字符，也会影响模型最终输出。

3. 推理参数不一致
   Python 端识别模块中通常包括 Image Resize to 32xWidth，长宽比约束，mean/std 归一化等；C++ 是否执行完全相同的 InferTransform 流程需确认，一旦 resize 或归一方式不同，识别结果差异非常正常。

建议：可以将 Python 中识别前的输入图像保存下来（paddleocr/tools/infer/predict_rec.py 中），与 C++ 中送入 rec 的图片逐像素对比来看是否一致。

4. 模型推理顺序和多线程差异
   Python PaddleOCR 中输入是排好序的，按检测框从上到下，从左到右送入识别模型；C++ 源码中 OCRPredictor 可能顺序不同，加上多线程时，某些 batch 的结果顺序也可能略有不同，导致输出内容顺序错乱。

可参考 PaddleOCR Issue #14027：https://github.com/PaddlePaddle/PaddleOCR/issues/14027，类似的问题显示了“不同推理系统抽取的切图看似相同，但识别结果不同”的现象，跟裁剪精度、预处理差异高度相关。

▍建议解决方案总结：

• ✅ 在 C++ 中确保模型仅加载一次，避免重复初始化。
• ✅ 强烈建议你保存 C++ OCR 中 "送入rec模型前" 的裁剪图片，与 Python 对比分析像素值差异。
• ✅ C++ 中完整实现 rec 模型 preprocess，具体可对照 Python 的 InferTransform，包括：
  • 灰度转 RGB（如 OpenCV 默认是 BGR）
  • resize、padding 到固定高度（通常32）
  • mean/std 归一化处理
• ✅ 检查是否启用 MKL-DNN 和 IR 优化。
• ✅ 确保 TrueType font 字符集文件完全一致。
• ✅ 确保 Release 编译并启用多线程优化。

如果上述方法仍无法解决识别差异较大的问题，建议使用 Python 模型导出中间输入、输出数据，再在 C++ 中用相同数据跑通，以实现 byte-level 对齐验证。

如有进一步追溯需要，建议参考官方文档的 C++ inference API 文档及 Demo：

https://github.com/PaddlePaddle/PaddleOCR/tree/release/2.6/deploy/cpp_infer

希望这些信息能帮助你定位问题。如仍未解决可考虑提供 rec 输入图像数据对照分析。

---

Response generated by 🤖 feifei-bot | chatgpt-4o-latest

⚠️ This is an AI-generated response. Please verify any critical information.

[Phymin]

请查看python和cpp_infer的完整输出日志：

python

[2025/04/16 14:01:35] ppocr DEBUG: Namespace(help='==SUPPRESS==', use_gpu=False, use_xpu=False, use_npu=False, use_mlu=False, use_gcu=False, ir_optim=True, use_tensorrt=False, min_subgraph_size=15, precision='fp32', gpu_mem=500, gpu_id=0, image_dir=None, page_num=0, det_algorithm='DB', det_model_dir='D:/Vision_Tech/ppocr/models/ch_PP-OCRv4_det_server_infer', det_limit_side_len=960, det_limit_type='max', det_box_type='quad', det_db_thresh=0.3, det_db_box_thresh=0.6, det_db_unclip_ratio=1.5, max_batch_size=10, use_dilation=False, det_db_score_mode='fast', det_east_score_thresh=0.8, det_east_cover_thresh=0.1, det_east_nms_thresh=0.2, det_sast_score_thresh=0.5, det_sast_nms_thresh=0.2, det_pse_thresh=0, det_pse_box_thresh=0.85, det_pse_min_area=16, det_pse_scale=1, scales=[8, 16, 32], alpha=1.0, beta=1.0, fourier_degree=5, rec_algorithm='SVTR_LCNet', rec_model_dir='D:/Vision_Tech/ppocr/models/ch_PP-OCRv4_rec_server_infer', rec_image_inverse=True, rec_image_shape='3, 48, 320', rec_batch_num=6, max_text_length=25, rec_char_dict_path='C:\\Users\\phymin\\.conda\\envs\\label_dl\\lib\\site-packages\\paddleocr\\ppocr\\utils\\ppocr_keys_v1.txt', use_space_char=True, vis_font_path='./doc/fonts/simfang.ttf', drop_score=0.5, e2e_algorithm='PGNet', e2e_model_dir=None, e2e_limit_side_len=768, e2e_limit_type='max', e2e_pgnet_score_thresh=0.5, e2e_char_dict_path='./ppocr/utils/ic15_dict.txt', e2e_pgnet_valid_set='totaltext', e2e_pgnet_mode='fast', use_angle_cls=False, cls_model_dir='D:\\Vision_Tech\\models\\whl\\cls\\ch_ppocr_mobile_v2.0_cls_infer', cls_image_shape='3, 48, 192', label_list=['0', '180'], cls_batch_num=6, cls_thresh=0.9, enable_mkldnn=False, cpu_threads=10, use_pdserving=False, warmup=False, sr_model_dir=None, sr_image_shape='3, 32, 128', sr_batch_num=1, draw_img_save_dir='./inference_results', save_crop_res=False, crop_res_save_dir='./output', use_mp=False, total_process_num=1, process_id=0, benchmark=False, save_log_path='./log_output/', show_log=True, use_onnx=False, onnx_providers=False, onnx_sess_options=False, return_word_box=False, output='./output', table_max_len=488, table_algorithm='TableAttn', table_model_dir=None, merge_no_span_structure=True, table_char_dict_path=None, formula_algorithm='LaTeXOCR', formula_model_dir=None, formula_char_dict_path=None, formula_batch_num=1, layout_model_dir=None, layout_dict_path=None, layout_score_threshold=0.5, layout_nms_threshold=0.5, kie_algorithm='LayoutXLM', ser_model_dir=None, re_model_dir=None, use_visual_backbone=True, ser_dict_path='../train_data/XFUND/class_list_xfun.txt', ocr_order_method=None, mode='structure', image_orientation=False, layout=True, table=True, formula=False, ocr=True, recovery=False, recovery_to_markdown=False, use_pdf2docx_api=False, invert=False, binarize=False, alphacolor=(255, 255, 255), lang='ch', det=True, rec=True, type='ocr', savefile=False, ocr_version='PP-OCRv4', structure_version='PP-StructureV2')
[2025/04/16 14:01:43] ppocr DEBUG: dt_boxes num : 17, elapsed : 5.925155401229858
[2025/04/16 14:03:14] ppocr DEBUG: rec_res num  : 17, elapsed : 90.64945650100708
[[[30.0, 25.0], [146.0, 25.0], [146.0, 47.0], [30.0, 47.0]], ('(P)HUAWEI', 0.9205098748207092)]
[[[155.0, 24.0], [207.0, 24.0], [207.0, 47.0], [155.0, 47.0]], ('P/N:', 0.9215103387832642)]
[[[273.0, 24.0], [378.0, 24.0], [378.0, 46.0], [273.0, 46.0]], ('08070612', 0.9954185485839844)]
[[[33.0, 93.0], [129.0, 93.0], [129.0, 111.0], [33.0, 111.0]], ('（1P）Mfr', 0.8607510328292847)]        
[[[127.0, 90.0], [181.0, 90.0], [181.0, 112.0], [127.0, 112.0]], ('P/N:', 0.8674935698509216)]
[[[192.0, 87.0], [420.0, 91.0], [420.0, 115.0], [192.0, 112.0]], ('CL05A105KP5NNXC', 0.9789352416992188)]
[[[30.0, 160.0], [220.0, 160.0], [220.0, 180.0], [30.0, 180.0]], ('(10D）DateCode', 0.9151744842529297)] 
[[[263.0, 159.0], [337.0, 156.0], [338.0, 179.0], [263.0, 181.0]], ('191101', 0.9944730401039124)]       
[[[31.0, 227.0], [167.0, 227.0], [167.0, 244.0], [31.0, 244.0]], ('（1T）Lot#', 0.8181165456771851)]     
[[[164.0, 226.0], [262.0, 226.0], [262.0, 244.0], [164.0, 244.0]], ('CF9HT5P', 0.9880648851394653)]      
[[[30.0, 291.0], [199.0, 291.0], [199.0, 311.0], [30.0, 311.0]], ('(Q)QTY:10.000', 0.9011315703392029)]  
[[[529.0, 300.0], [583.0, 300.0], [583.0, 317.0], [529.0, 317.0]], ('pass', 0.9919123649597168)]
[[[480.0, 318.0], [582.0, 320.0], [581.0, 342.0], [480.0, 339.0]], ('Rohs', 0.9899167418479919)]
[[[528.0, 343.0], [583.0, 342.0], [584.0, 360.0], [528.0, 361.0]], ('pass', 0.9922510385513306)]
[[[168.0, 359.0], [279.0, 359.0], [279.0, 376.0], [168.0, 376.0]], ('(V）Vendor', 0.902154803276062)]    
[[[292.0, 356.0], [450.0, 354.0], [451.0, 375.0], [292.0, 377.0]], ('Code：Z00MCD', 0.8026291131973267)]
[[[28.0, 375.0], [131.0, 373.0], [131.0, 394.0], [29.0, 397.0]], ('SAMSUNG', 0.9068806767463684)]
ocr cost time:  96.5886881351471

cpp_infer:

In PP-OCRv3, rec_image_shape parameter defaults to '3, 48, 320',if you are using recognition model with PP-OCRv2 or an older version, please set --rec_image_shape='3,32,320
total images num: 1
WARNING: Logging before InitGoogleLogging() is written to STDERR
I0416 13:39:26.008740 37008 analysis_predictor.cc:1942] MKLDNN is enabled
e[1me[35m--- Running analysis [ir_graph_build_pass]e[0m
I0416 13:39:26.012741 37008 executor.cc:183] Old Executor is Running.
e[1me[35m--- Running analysis [ir_analysis_pass]e[0m
e[32m--- Running IR pass [onednn_placement_pass]e[0m
e[32m--- Running IR pass [simplify_with_basic_ops_pass]e[0m
e[32m--- Running IR pass [layer_norm_fuse_pass]e[0m
e[32m--- Running IR pass [attention_lstm_fuse_pass]e[0m
e[32m--- Running IR pass [seqconv_eltadd_relu_fuse_pass]e[0m
e[32m--- Running IR pass [seqpool_cvm_concat_fuse_pass]e[0m
e[32m--- Running IR pass [mul_lstm_fuse_pass]e[0m
e[32m--- Running IR pass [fc_gru_fuse_pass]e[0m
e[32m--- Running IR pass [mul_gru_fuse_pass]e[0m
e[32m--- Running IR pass [seq_concat_fc_fuse_pass]e[0m
e[32m--- Running IR pass [gpu_cpu_squeeze2_matmul_fuse_pass]e[0m
e[32m--- Running IR pass [gpu_cpu_reshape2_matmul_fuse_pass]e[0m
e[32m--- Running IR pass [gpu_cpu_flatten2_matmul_fuse_pass]e[0m
e[32m--- Running IR pass [matmul_v2_scale_fuse_pass]e[0m
e[32m--- Running IR pass [gpu_cpu_map_matmul_v2_to_mul_pass]e[0m
e[32m--- Running IR pass [gpu_cpu_map_matmul_v2_to_matmul_pass]e[0m
e[32m--- Running IR pass [matmul_scale_fuse_pass]e[0m
e[32m--- Running IR pass [gpu_cpu_map_matmul_to_mul_pass]e[0m
e[32m--- Running IR pass [fc_fuse_pass]e[0m
e[32m--- Running IR pass [repeated_fc_relu_fuse_pass]e[0m
e[32m--- Running IR pass [squared_mat_sub_fuse_pass]e[0m
e[32m--- Running IR pass [conv_bn_fuse_pass]e[0m
I0416 13:39:26.375403 37008 fuse_pass_base.cc:55] ---  detected 40 subgraphs
e[32m--- Running IR pass [conv_eltwiseadd_bn_fuse_pass]e[0m
e[32m--- Running IR pass [conv_transpose_bn_fuse_pass]e[0m
e[32m--- Running IR pass [conv_transpose_eltwiseadd_bn_fuse_pass]e[0m
e[32m--- Running IR pass [is_test_pass]e[0m
e[32m--- Running IR pass [constant_folding_pass]e[0m
I0416 13:39:26.426416 37008 fuse_pass_base.cc:55] ---  detected 52 subgraphs
e[32m--- Running IR pass [squeeze2_transpose2_onednn_fuse_pass]e[0m
e[37m--- fused 0 squeeze2 with transpose2e[0m
e[32m--- Running IR pass [depthwise_conv_onednn_pass]e[0m
I0416 13:39:26.427417 37008 fuse_pass_base.cc:55] ---  detected 3 subgraphs
e[32m--- Running IR pass [conv_bn_fuse_pass]e[0m
I0416 13:39:26.436416 37008 fuse_pass_base.cc:55] ---  detected 3 subgraphs
e[32m--- Running IR pass [conv_eltwiseadd_bn_fuse_pass]e[0m
W0416 13:39:26.457417 37008 op_compat_sensible_pass.cc:230]  Check the Attr(axis) of Op(elementwise_add) in pass(conv_eltwiseadd_bn_fuse_pass) failed!
W0416 13:39:26.457417 37008 conv_bn_fuse_pass.cc:644] Pass in op compat failed.
W0416 13:39:26.457417 37008 op_compat_sensible_pass.cc:230]  Check the Attr(axis) of Op(elementwise_add) in pass(conv_eltwiseadd_bn_fuse_pass) failed!
W0416 13:39:26.457417 37008 conv_bn_fuse_pass.cc:644] Pass in op compat failed.
W0416 13:39:26.457417 37008 op_compat_sensible_pass.cc:230]  Check the Attr(axis) of Op(elementwise_add) in pass(conv_eltwiseadd_bn_fuse_pass) failed!
W0416 13:39:26.457417 37008 conv_bn_fuse_pass.cc:644] Pass in op compat failed.
W0416 13:39:26.457417 37008 op_compat_sensible_pass.cc:230]  Check the Attr(axis) of Op(elementwise_add) in pass(conv_eltwiseadd_bn_fuse_pass) failed!
W0416 13:39:26.457417 37008 conv_bn_fuse_pass.cc:644] Pass in op compat failed.
e[32m--- Running IR pass [conv_affine_channel_onednn_fuse_pass]e[0m
e[32m--- Running IR pass [conv_transpose_bn_fuse_pass]e[0m
e[32m--- Running IR pass [conv_transpose_eltwiseadd_bn_fuse_pass]e[0m
W0416 13:39:26.473416 37008 op_compat_sensible_pass.cc:230]  Check the Attr(axis) of Op(elementwise_add) in pass(conv_transpose_eltwiseadd_bn_fuse_pass) failed!
W0416 13:39:26.473416 37008 conv_bn_fuse_pass.cc:644] Pass in op compat failed.
e[32m--- Running IR pass [conv_bias_onednn_fuse_pass]e[0m
W0416 13:39:26.489416 37008 op_compat_sensible_pass.cc:230]  Check the Attr(axis) of Op(elementwise_add) in pass(conv_bias_onednn_fuse_pass) failed!
W0416 13:39:26.489416 37008 op_compat_sensible_pass.cc:230]  Check the Attr(axis) of Op(elementwise_add) in pass(conv_bias_onednn_fuse_pass) failed!
W0416 13:39:26.489416 37008 op_compat_sensible_pass.cc:230]  Check the Attr(axis) of Op(elementwise_add) in pass(conv_bias_onednn_fuse_pass) failed!
W0416 13:39:26.489416 37008 op_compat_sensible_pass.cc:230]  Check the Attr(axis) of Op(elementwise_add) in pass(conv_bias_onednn_fuse_pass) failed!
W0416 13:39:26.490417 37008 op_compat_sensible_pass.cc:230]  Check the Attr(axis) of Op(elementwise_add) in pass(conv_bias_onednn_fuse_pass) failed!
W0416 13:39:26.490417 37008 op_compat_sensible_pass.cc:230]  Check the Attr(axis) of Op(elementwise_add) in pass(conv_bias_onednn_fuse_pass) failed!
W0416 13:39:26.490417 37008 op_compat_sensible_pass.cc:230]  Check the Attr(axis) of Op(elementwise_add) in pass(conv_bias_onednn_fuse_pass) failed!
W0416 13:39:26.490417 37008 op_compat_sensible_pass.cc:230]  Check the Attr(axis) of Op(elementwise_add) in pass(conv_bias_onednn_fuse_pass) failed!
W0416 13:39:26.490417 37008 op_compat_sensible_pass.cc:230]  Check the Attr(axis) of Op(elementwise_add) in pass(conv_bias_onednn_fuse_pass) failed!
W0416 13:39:26.490417 37008 op_compat_sensible_pass.cc:230]  Check the Attr(axis) of Op(elementwise_add) in pass(conv_bias_onednn_fuse_pass) failed!
W0416 13:39:26.491416 37008 op_compat_sensible_pass.cc:230]  Check the Attr(axis) of Op(elementwise_add) in pass(conv_bias_onednn_fuse_pass) failed!
W0416 13:39:26.491416 37008 op_compat_sensible_pass.cc:230]  Check the Attr(axis) of Op(elementwise_add) in pass(conv_bias_onednn_fuse_pass) failed!
W0416 13:39:26.491416 37008 op_compat_sensible_pass.cc:230]  Check the Attr(axis) of Op(elementwise_add) in pass(conv_bias_onednn_fuse_pass) failed!
W0416 13:39:26.491416 37008 op_compat_sensible_pass.cc:230]  Check the Attr(axis) of Op(elementwise_add) in pass(conv_bias_onednn_fuse_pass) failed!
W0416 13:39:26.491416 37008 op_compat_sensible_pass.cc:230]  Check the Attr(axis) of Op(elementwise_add) in pass(conv_bias_onednn_fuse_pass) failed!
W0416 13:39:26.491416 37008 op_compat_sensible_pass.cc:230]  Check the Attr(axis) of Op(elementwise_add) in pass(conv_bias_onednn_fuse_pass) failed!
W0416 13:39:26.491416 37008 op_compat_sensible_pass.cc:230]  Check the Attr(axis) of Op(elementwise_add) in pass(conv_bias_onednn_fuse_pass) failed!
W0416 13:39:26.491416 37008 op_compat_sensible_pass.cc:230]  Check the Attr(axis) of Op(elementwise_add) in pass(conv_bias_onednn_fuse_pass) failed!
W0416 13:39:26.492416 37008 op_compat_sensible_pass.cc:230]  Check the Attr(axis) of Op(elementwise_add) in pass(conv_bias_onednn_fuse_pass) failed!
W0416 13:39:26.492416 37008 op_compat_sensible_pass.cc:230]  Check the Attr(axis) of Op(elementwise_add) in pass(conv_bias_onednn_fuse_pass) failed!
W0416 13:39:26.492416 37008 op_compat_sensible_pass.cc:230]  Check the Attr(axis) of Op(elementwise_add) in pass(conv_bias_onednn_fuse_pass) failed!
W0416 13:39:26.492416 37008 op_compat_sensible_pass.cc:230]  Check the Attr(axis) of Op(elementwise_add) in pass(conv_bias_onednn_fuse_pass) failed!
W0416 13:39:26.492416 37008 op_compat_sensible_pass.cc:230]  Check the Attr(axis) of Op(elementwise_add) in pass(conv_bias_onednn_fuse_pass) failed!
W0416 13:39:26.493417 37008 op_compat_sensible_pass.cc:230]  Check the Attr(axis) of Op(elementwise_add) in pass(conv_bias_onednn_fuse_pass) failed!
W0416 13:39:26.493417 37008 op_compat_sensible_pass.cc:230]  Check the Attr(axis) of Op(elementwise_add) in pass(conv_bias_onednn_fuse_pass) failed!
W0416 13:39:26.493417 37008 op_compat_sensible_pass.cc:230]  Check the Attr(axis) of Op(elementwise_add) in pass(conv_bias_onednn_fuse_pass) failed!
W0416 13:39:26.493417 37008 op_compat_sensible_pass.cc:230]  Check the Attr(axis) of Op(elementwise_add) in pass(conv_bias_onednn_fuse_pass) failed!
W0416 13:39:26.494416 37008 op_compat_sensible_pass.cc:230]  Check the Attr(axis) of Op(elementwise_add) in pass(conv_bias_onednn_fuse_pass) failed!
W0416 13:39:26.494416 37008 op_compat_sensible_pass.cc:230]  Check the Attr(axis) of Op(elementwise_add) in pass(conv_bias_onednn_fuse_pass) failed!
W0416 13:39:26.494416 37008 op_compat_sensible_pass.cc:230]  Check the Attr(axis) of Op(elementwise_add) in pass(conv_bias_onednn_fuse_pass) failed!
W0416 13:39:26.494416 37008 op_compat_sensible_pass.cc:230]  Check the Attr(axis) of Op(elementwise_add) in pass(conv_bias_onednn_fuse_pass) failed!
W0416 13:39:26.494416 37008 op_compat_sensible_pass.cc:230]  Check the Attr(axis) of Op(elementwise_add) in pass(conv_bias_onednn_fuse_pass) failed!
W0416 13:39:26.495417 37008 op_compat_sensible_pass.cc:230]  Check the Attr(axis) of Op(elementwise_add) in pass(conv_bias_onednn_fuse_pass) failed!
W0416 13:39:26.495417 37008 op_compat_sensible_pass.cc:230]  Check the Attr(axis) of Op(elementwise_add) in pass(conv_bias_onednn_fuse_pass) failed!
W0416 13:39:26.495417 37008 op_compat_sensible_pass.cc:230]  Check the Attr(axis) of Op(elementwise_add) in pass(conv_bias_onednn_fuse_pass) failed!
W0416 13:39:26.495417 37008 op_compat_sensible_pass.cc:230]  Check the Attr(axis) of Op(elementwise_add) in pass(conv_bias_onednn_fuse_pass) failed!
W0416 13:39:26.495417 37008 op_compat_sensible_pass.cc:230]  Check the Attr(axis) of Op(elementwise_add) in pass(conv_bias_onednn_fuse_pass) failed!
W0416 13:39:26.495417 37008 op_compat_sensible_pass.cc:230]  Check the Attr(axis) of Op(elementwise_add) in pass(conv_bias_onednn_fuse_pass) failed!
W0416 13:39:26.495417 37008 op_compat_sensible_pass.cc:230]  Check the Attr(axis) of Op(elementwise_add) in pass(conv_bias_onednn_fuse_pass) failed!
W0416 13:39:26.495417 37008 op_compat_sensible_pass.cc:230]  Check the Attr(axis) of Op(elementwise_add) in pass(conv_bias_onednn_fuse_pass) failed!
W0416 13:39:26.495417 37008 op_compat_sensible_pass.cc:230]  Check the Attr(axis) of Op(elementwise_add) in pass(conv_bias_onednn_fuse_pass) failed!
W0416 13:39:26.496417 37008 op_compat_sensible_pass.cc:230]  Check the Attr(axis) of Op(elementwise_add) in pass(conv_bias_onednn_fuse_pass) failed!
W0416 13:39:26.496417 37008 op_compat_sensible_pass.cc:230]  Check the Attr(axis) of Op(elementwise_add) in pass(conv_bias_onednn_fuse_pass) failed!
W0416 13:39:26.496417 37008 op_compat_sensible_pass.cc:230]  Check the Attr(axis) of Op(elementwise_add) in pass(conv_bias_onednn_fuse_pass) failed!
W0416 13:39:26.496417 37008 op_compat_sensible_pass.cc:230]  Check the Attr(axis) of Op(elementwise_add) in pass(conv_bias_onednn_fuse_pass) failed!
W0416 13:39:26.496417 37008 op_compat_sensible_pass.cc:230]  Check the Attr(axis) of Op(elementwise_add) in pass(conv_bias_onednn_fuse_pass) failed!
W0416 13:39:26.497417 37008 op_compat_sensible_pass.cc:230]  Check the Attr(axis) of Op(elementwise_add) in pass(conv_bias_onednn_fuse_pass) failed!
W0416 13:39:26.497417 37008 op_compat_sensible_pass.cc:230]  Check the Attr(axis) of Op(elementwise_add) in pass(conv_bias_onednn_fuse_pass) failed!
W0416 13:39:26.497417 37008 op_compat_sensible_pass.cc:230]  Check the Attr(axis) of Op(elementwise_add) in pass(conv_bias_onednn_fuse_pass) failed!
W0416 13:39:26.497417 37008 op_compat_sensible_pass.cc:230]  Check the Attr(axis) of Op(elementwise_add) in pass(conv_bias_onednn_fuse_pass) failed!
e[32m--- Running IR pass [conv_transpose_bias_onednn_fuse_pass]e[0m
W0416 13:39:26.503417 37008 op_compat_sensible_pass.cc:230]  Check the Attr(axis) of Op(elementwise_add) in pass(conv_transpose_bias_onednn_fuse_pass) failed!
W0416 13:39:26.503417 37008 op_compat_sensible_pass.cc:230]  Check the Attr(axis) of Op(elementwise_add) in pass(conv_transpose_bias_onednn_fuse_pass) failed!
e[32m--- Running IR pass [conv_elementwise_add_onednn_fuse_pass]e[0m
e[37m---    Fused 3 projection conv (as y) + elementwise_add patternse[0m
I0416 13:39:26.737922 37008 fuse_pass_base.cc:55] ---  detected 3 subgraphs
e[32m--- Running IR pass [conv_activation_onednn_fuse_pass]e[0m
I0416 13:39:26.751922 37008 fuse_pass_base.cc:55] ---  detected 40 subgraphs
e[37m---    fused 40 fused_conv2d with relu activatione[0m
e[32m--- Running IR pass [scale_matmul_fuse_pass]e[0m
e[32m--- Running IR pass [reshape_transpose_matmul_onednn_fuse_pass]e[0m
e[32m--- Running IR pass [matmul_transpose_reshape_onednn_fuse_pass]e[0m
e[32m--- Running IR pass [matmul_elementwise_add_onednn_fuse_pass]e[0m
e[32m--- Running IR pass [matmul_activation_onednn_fuse_pass]e[0m
e[32m--- Running IR pass [fc_onednn_pass]e[0m
e[32m--- Running IR pass [fc_act_onednn_fuse_pass]e[0m
e[32m--- Running IR pass [self_attention_fuse_pass]e[0m
W0416 13:39:26.785923 37008 self_attention_fuse_pass.cc:52] No-avx512 or MKL supported!
e[32m--- Running IR pass [batch_norm_act_fuse_pass]e[0m
I0416 13:39:26.787922 37008 fuse_pass_base.cc:55] ---  detected 5 subgraphs
e[37m---    fused 5 batch norm with relu activatione[0m
e[32m--- Running IR pass [softplus_activation_onednn_fuse_pass]e[0m
e[32m--- Running IR pass [shuffle_channel_onednn_detect_pass]e[0m
e[32m--- Running IR pass [elementwise_act_onednn_fuse_pass]e[0m
I0416 13:39:26.798923 37008 fuse_pass_base.cc:55] ---  detected 7 subgraphs
e[37m---    fused 7 elementwise_add with sigmoid activatione[0m
e[32m--- Running IR pass [operator_scale_onednn_fuse_pass]e[0m
I0416 13:39:26.807922 37008 fuse_pass_base.cc:55] ---  detected 1 subgraphs
e[37m---    fused 1 elementwise_add with scalee[0m
e[32m--- Running IR pass [operator_unsqueeze2_onednn_fuse_pass]e[0m
e[32m--- Running IR pass [operator_reshape2_onednn_fuse_pass]e[0m
e[1me[35m--- Running analysis [ir_params_sync_among_devices_pass]e[0m
e[1me[35m--- Running analysis [adjust_cudnn_workspace_size_pass]e[0m
e[1me[35m--- Running analysis [inference_op_replace_pass]e[0m
e[1me[35m--- Running analysis [save_optimized_model_pass]e[0m
e[1me[35m--- Running analysis [memory_optimize_pass]e[0m
I0416 13:39:26.812922 37008 memory_optimize_pass.cc:114] The persistable params in main graph are : 107.918MB
I0416 13:39:26.815922 37008 memory_optimize_pass.cc:250] Cluster name : elementwise_mul_1  size: 2048
I0416 13:39:26.815922 37008 memory_optimize_pass.cc:250] Cluster name : relu_37.tmp_0  size: 4096
I0416 13:39:26.815922 37008 memory_optimize_pass.cc:250] Cluster name : concat_4.tmp_0  size: 8448
I0416 13:39:26.815922 37008 memory_optimize_pass.cc:250] Cluster name : pool2d_5.tmp_0  size: 4096
I0416 13:39:26.815922 37008 memory_optimize_pass.cc:250] Cluster name : elementwise_mul_0  size: 1024
I0416 13:39:26.815922 37008 memory_optimize_pass.cc:250] Cluster name : relu_30.tmp_0  size: 3072
I0416 13:39:26.815922 37008 memory_optimize_pass.cc:250] Cluster name : relu_32.tmp_0  size: 896
I0416 13:39:26.815922 37008 memory_optimize_pass.cc:250] Cluster name : conv2d_153.tmp_0  size: 1024
I0416 13:39:26.815922 37008 memory_optimize_pass.cc:250] Cluster name : relu_35.tmp_0  size: 896
I0416 13:39:26.815922 37008 memory_optimize_pass.cc:250] Cluster name : relu_33.tmp_0  size: 896
I0416 13:39:26.815922 37008 memory_optimize_pass.cc:250] Cluster name : relu_36.tmp_0  size: 896
I0416 13:39:26.815922 37008 memory_optimize_pass.cc:250] Cluster name : x  size: 12
e[1me[35m--- Running analysis [ir_graph_to_program_pass]e[0m
I0416 13:39:26.838922 37008 analysis_predictor.cc:2146] ======= ir optimization completed =======
I0416 13:39:26.838922 37008 naive_executor.cc:211] ---  skip [feed], feed -> x
I0416 13:39:26.839922 37008 naive_executor.cc:211] ---  skip [sigmoid_11.tmp_0], fetch -> fetch
I0416 13:39:26.840922 37008 naive_executor.cc:211] ---  skip [tmp_36], fetch -> fetch
In PP-OCRv3, default rec_img_h is 48,if you use other model, you should set the param rec_img_h=32
I0416 13:39:26.857923 37008 analysis_predictor.cc:1942] MKLDNN is enabled
e[1me[35m--- Running analysis [ir_graph_build_pass]e[0m
e[1me[35m--- Running analysis [ir_analysis_pass]e[0m
e[32m--- Running IR pass [onednn_placement_pass]e[0m
e[32m--- Running IR pass [simplify_with_basic_ops_pass]e[0m
e[32m--- Running IR pass [layer_norm_fuse_pass]e[0m
e[32m--- Running IR pass [attention_lstm_fuse_pass]e[0m
e[32m--- Running IR pass [seqconv_eltadd_relu_fuse_pass]e[0m
e[32m--- Running IR pass [seqpool_cvm_concat_fuse_pass]e[0m
e[32m--- Running IR pass [mul_lstm_fuse_pass]e[0m
e[32m--- Running IR pass [fc_gru_fuse_pass]e[0m
e[32m--- Running IR pass [mul_gru_fuse_pass]e[0m
e[32m--- Running IR pass [seq_concat_fc_fuse_pass]e[0m
e[32m--- Running IR pass [gpu_cpu_squeeze2_matmul_fuse_pass]e[0m
e[32m--- Running IR pass [gpu_cpu_reshape2_matmul_fuse_pass]e[0m
e[32m--- Running IR pass [gpu_cpu_flatten2_matmul_fuse_pass]e[0m
e[32m--- Running IR pass [matmul_v2_scale_fuse_pass]e[0m
e[32m--- Running IR pass [gpu_cpu_map_matmul_v2_to_mul_pass]e[0m
I0416 13:39:26.966922 37008 fuse_pass_base.cc:55] ---  detected 9 subgraphs
e[32m--- Running IR pass [gpu_cpu_map_matmul_v2_to_matmul_pass]e[0m
I0416 13:39:26.967922 37008 fuse_pass_base.cc:55] ---  detected 4 subgraphs
e[32m--- Running IR pass [matmul_scale_fuse_pass]e[0m
e[32m--- Running IR pass [gpu_cpu_map_matmul_to_mul_pass]e[0m
e[32m--- Running IR pass [fc_fuse_pass]e[0m
I0416 13:39:26.977922 37008 fuse_pass_base.cc:55] ---  detected 9 subgraphs
e[32m--- Running IR pass [repeated_fc_relu_fuse_pass]e[0m
e[32m--- Running IR pass [squared_mat_sub_fuse_pass]e[0m
e[32m--- Running IR pass [conv_bn_fuse_pass]e[0m
I0416 13:39:27.435386 37008 fuse_pass_base.cc:55] ---  detected 43 subgraphs
e[32m--- Running IR pass [conv_eltwiseadd_bn_fuse_pass]e[0m
e[32m--- Running IR pass [conv_transpose_bn_fuse_pass]e[0m
e[32m--- Running IR pass [conv_transpose_eltwiseadd_bn_fuse_pass]e[0m
e[32m--- Running IR pass [is_test_pass]e[0m
e[32m--- Running IR pass [constant_folding_pass]e[0m
e[32m--- Running IR pass [squeeze2_transpose2_onednn_fuse_pass]e[0m
I0416 13:39:27.450387 37008 fuse_pass_base.cc:55] ---  detected 1 subgraphs
e[37m--- fused 1 squeeze2 with transpose2e[0m
e[32m--- Running IR pass [depthwise_conv_onednn_pass]e[0m
I0416 13:39:27.451386 37008 fuse_pass_base.cc:55] ---  detected 4 subgraphs
e[32m--- Running IR pass [conv_bn_fuse_pass]e[0m
I0416 13:39:27.456386 37008 fuse_pass_base.cc:55] ---  detected 4 subgraphs
e[32m--- Running IR pass [conv_eltwiseadd_bn_fuse_pass]e[0m
e[32m--- Running IR pass [conv_affine_channel_onednn_fuse_pass]e[0m
e[32m--- Running IR pass [conv_transpose_bn_fuse_pass]e[0m
e[32m--- Running IR pass [conv_transpose_eltwiseadd_bn_fuse_pass]e[0m
e[32m--- Running IR pass [conv_bias_onednn_fuse_pass]e[0m
I0416 13:39:27.469386 37008 fuse_pass_base.cc:55] ---  detected 5 subgraphs
e[37m---    fused 5 conv2d with elementwise_add as biase[0m
e[32m--- Running IR pass [conv_transpose_bias_onednn_fuse_pass]e[0m
e[32m--- Running IR pass [conv_elementwise_add_onednn_fuse_pass]e[0m
e[32m--- Running IR pass [conv_activation_onednn_fuse_pass]e[0m
I0416 13:39:27.507386 37008 fuse_pass_base.cc:55] ---  detected 38 subgraphs
e[37m---    fused 38 fused_conv2d with relu activatione[0m
I0416 13:39:27.509387 37008 fuse_pass_base.cc:55] ---  detected 5 subgraphs
e[37m---    fused 5 fused_conv2d with sigmoid activatione[0m
W0416 13:39:27.511392 37008 op_compat_sensible_pass.cc:218]  Attribute(beta) of Op(swish) is not defined in opProto or is in extra set!The compatible check for this attribute is not use. Please remove it from the precondition of pass: conv_activation_onednn_fuse_pass
I0416 13:39:27.512387 37008 fuse_pass_base.cc:55] ---  detected 5 subgraphs
e[37m---    fused 5 fused_conv2d with swish activatione[0m
e[32m--- Running IR pass [scale_matmul_fuse_pass]e[0m
I0416 13:39:27.514386 37008 fuse_pass_base.cc:55] ---  detected 2 subgraphs
e[37m---    fused 2 scale with matmule[0m
e[32m--- Running IR pass [reshape_transpose_matmul_onednn_fuse_pass]e[0m
e[32m--- Running IR pass [matmul_transpose_reshape_onednn_fuse_pass]e[0m
I0416 13:39:27.531075 37008 fuse_pass_base.cc:55] ---  detected 2 subgraphs
e[37m---    fused 2 matmul + transpose + reshape patternse[0m
e[32m--- Running IR pass [matmul_elementwise_add_onednn_fuse_pass]e[0m
e[32m--- Running IR pass [matmul_activation_onednn_fuse_pass]e[0m
e[32m--- Running IR pass [fc_onednn_pass]e[0m
I0416 13:39:27.539062 37008 fuse_pass_base.cc:55] ---  detected 9 subgraphs
e[37m---    enabled FC MKL-DNN for 9 fc ops e[0m
e[32m--- Running IR pass [fc_act_onednn_fuse_pass]e[0m
I0416 13:39:27.542061 37008 fuse_pass_base.cc:55] ---  detected 2 subgraphs
e[37m---    fused 2 fc with swish activatione[0m
e[32m--- Running IR pass [self_attention_fuse_pass]e[0m
W0416 13:39:27.543061 37008 self_attention_fuse_pass.cc:52] No-avx512 or MKL supported!
e[32m--- Running IR pass [batch_norm_act_fuse_pass]e[0m
e[32m--- Running IR pass [softplus_activation_onednn_fuse_pass]e[0m
e[32m--- Running IR pass [shuffle_channel_onednn_detect_pass]e[0m
e[32m--- Running IR pass [elementwise_act_onednn_fuse_pass]e[0m
e[32m--- Running IR pass [operator_scale_onednn_fuse_pass]e[0m
e[32m--- Running IR pass [operator_unsqueeze2_onednn_fuse_pass]e[0m
e[32m--- Running IR pass [operator_reshape2_onednn_fuse_pass]e[0m
e[1me[35m--- Running analysis [ir_params_sync_among_devices_pass]e[0m
e[1me[35m--- Running analysis [adjust_cudnn_workspace_size_pass]e[0m
e[1me[35m--- Running analysis [inference_op_replace_pass]e[0m
e[1me[35m--- Running analysis [save_optimized_model_pass]e[0m
e[1me[35m--- Running analysis [memory_optimize_pass]e[0m
I0416 13:39:27.557061 37008 memory_optimize_pass.cc:114] The persistable params in main graph are : 86.1175MB
I0416 13:39:27.559062 37008 memory_optimize_pass.cc:250] Cluster name : relu_10.tmp_0  size: 7680
I0416 13:39:27.559062 37008 memory_optimize_pass.cc:250] Cluster name : concat_1.tmp_0  size: 58368
I0416 13:39:27.559062 37008 memory_optimize_pass.cc:250] Cluster name : elementwise_mul_2  size: 18432
I0416 13:39:27.559062 37008 memory_optimize_pass.cc:250] Cluster name : relu_16.tmp_0  size: 24576
I0416 13:39:27.559062 37008 memory_optimize_pass.cc:250] Cluster name : relu_30.tmp_0  size: 18432
I0416 13:39:27.559062 37008 memory_optimize_pass.cc:250] Cluster name : relu_13.tmp_0  size: 7680
I0416 13:39:27.559062 37008 memory_optimize_pass.cc:250] Cluster name : relu_11.tmp_0  size: 7680
I0416 13:39:27.559062 37008 memory_optimize_pass.cc:250] Cluster name : relu_14.tmp_0  size: 7680
I0416 13:39:27.559062 37008 memory_optimize_pass.cc:250] Cluster name : x  size: 576
e[1me[35m--- Running analysis [ir_graph_to_program_pass]e[0m
I0416 13:39:27.610102 37008 analysis_predictor.cc:2146] ======= ir optimization completed =======
I0416 13:39:27.611102 37008 naive_executor.cc:211] ---  skip [feed], feed -> x
I0416 13:39:27.612102 37008 naive_executor.cc:211] ---  skip [softmax_2.tmp_0], fetch -> fetch
I0416 13:39:27.626101 37008 onednn_context.cc:81] oneDNN v3.5.3
ocr costs: 207263
predict img: .\template.png
0       det boxes: [[155,24],[207,24],[207,46],[155,46]] rec text: P/N: rec score: 0.833894
1       det boxes: [[273,24],[377,24],[377,45],[273,45]] rec text: 08070612 rec score: 0.992806
2       det boxes: [[191,87],[420,90],[420,114],[191,111]] rec text: CL05A105KP5NNXC rec score: 0.964751
3       det boxes: [[30,25],[146,25],[146,46],[30,46]] rec text: (P)HUAWEI rec score: 0.9007
4       det boxes: [[33,92],[128,92],[128,110],[33,110]] rec text: （1P）Mfr rec score: 0.848296
5       det boxes: [[126,90],[180,90],[180,111],[126,111]] rec text: P/N rec score: 0.963511
6       det boxes: [[30,159],[220,159],[220,179],[30,179]] rec text: (10D）Date Code rec score: 0.886672
7       det boxes: [[262,158],[337,156],[337,178],[262,180]] rec text: 191101 rec score: 0.989412
8       det boxes: [[31,226],[167,226],[167,244],[31,244]] rec text: （1T）Lot#： rec score: 0.818124
9       det boxes: [[164,225],[261,225],[261,244],[164,244]] rec text: CF9HT5P rec score: 0.987019
10      det boxes: [[30,290],[198,290],[198,311],[30,311]] rec text: (Q)QTY:10.000 rec score: 0.856563
11      det boxes: [[528,299],[582,299],[582,317],[528,317]] rec text: pass rec score: 0.989197
12      det boxes: [[480,318],[581,320],[581,341],[480,339]] rec text: Rohs rec score: 0.989685
13      det boxes: [[527,342],[583,341],[583,359],[527,360]] rec text: pass rec score: 0.98579
14      det boxes: [[168,358],[279,358],[279,376],[168,376]] rec text: (V)Vendor rec score: 0.894614
15      det boxes: [[292,355],[450,353],[450,375],[292,377]] rec text: Code：ZOOMCD rec score: 0.802987
16      det boxes: [[28,375],[130,373],[130,394],[28,396]] rec text: SAMSUNG rec score: 0.935365
The detection visualized image saved in ./output//template.png
WARNING: Logging before InitGoogleLogging() is written to STDERR
I0416 13:42:54.999996 37008 autolog.h:51] ----------------------- Config info -----------------------
I0416 13:42:54.999996 37008 autolog.h:52] runtime_device: cpu
I0416 13:42:54.999996 37008 autolog.h:53] ir_optim: True
I0416 13:42:54.999996 37008 autolog.h:54] enable_memory_optim: True
I0416 13:42:54.999996 37008 autolog.h:55] enable_tensorrt: 0
I0416 13:42:55.000996 37008 autolog.h:56] enable_mkldnn: True
I0416 13:42:55.000996 37008 autolog.h:57] cpu_math_library_num_threads: 10
I0416 13:42:55.000996 37008 autolog.h:58] ----------------------- Data info -----------------------
I0416 13:42:55.000996 37008 autolog.h:59] batch_size: 1
I0416 13:42:55.000996 37008 autolog.h:60] input_shape: dynamic
I0416 13:42:55.000996 37008 autolog.h:61] data_num: 1
I0416 13:42:55.000996 37008 autolog.h:62] ----------------------- Model info -----------------------
I0416 13:42:55.000996 37008 autolog.h:63] model_name: ocr_det
I0416 13:42:55.000996 37008 autolog.h:64] precision: fp32
I0416 13:42:55.001995 37008 autolog.h:65] ----------------------- Perf info ------------------------
I0416 13:42:55.001995 37008 autolog.h:66] Total time spent(ms): 5971
I0416 13:42:55.001995 37008 autolog.h:68] preprocess_time(ms): 7.0819, inference_time(ms): 5958.27, postprocess_time(ms): 6.965
I0416 13:42:55.001995 37008 autolog.h:51] ----------------------- Config info -----------------------
I0416 13:42:55.001995 37008 autolog.h:52] runtime_device: cpu
I0416 13:42:55.001995 37008 autolog.h:53] ir_optim: True
I0416 13:42:55.001995 37008 autolog.h:54] enable_memory_optim: True
I0416 13:42:55.001995 37008 autolog.h:55] enable_tensorrt: 0
I0416 13:42:55.003001 37008 autolog.h:56] enable_mkldnn: True
I0416 13:42:55.003001 37008 autolog.h:57] cpu_math_library_num_threads: 10
I0416 13:42:55.003001 37008 autolog.h:58] ----------------------- Data info -----------------------
I0416 13:42:55.003001 37008 autolog.h:59] batch_size: 6
I0416 13:42:55.003001 37008 autolog.h:60] input_shape: dynamic
I0416 13:42:55.003001 37008 autolog.h:61] data_num: 1
I0416 13:42:55.003001 37008 autolog.h:62] ----------------------- Model info -----------------------
I0416 13:42:55.003001 37008 autolog.h:63] model_name: ocr_rec
I0416 13:42:55.003001 37008 autolog.h:64] precision: fp32
I0416 13:42:55.004000 37008 autolog.h:65] ----------------------- Perf info ------------------------
I0416 13:42:55.004000 37008 autolog.h:66] Total time spent(ms): 201279
I0416 13:42:55.004000 37008 autolog.h:68] preprocess_time(ms): 6.346, inference_time(ms): 201205, postprocess_time(ms): 69.9219

能否从这两份输出日志中得到你想要的信息？