yolov5量化注意事项

news2025/1/20 1:56:28

（1）使用onnxsim ==0.4.1版本、ort版本1.13.1

 否则使用--dynamic True时所生成的onnx就会乱掉。

（2）利用trtexec生成engine

所使用的命令如下：

trtexec.exe --onnx=yolov5n_ptq_detect_dynamic.onnx --saveEngine=ptq_int8_fp16.trtmodel --int8 --fp16
 === Trace details ===
[11/28/2022-01:21:54] [I] Trace averages of 10 runs:
[11/28/2022-01:21:54] [I] Average on 10 runs - GPU latency: 10.0542 ms - Host latency: 11.208 ms (end to end 35 ms, enqueue 1.81131 ms)
[11/28/2022-01:21:54] [I] Average on 10 runs - GPU latency: 9.81727 ms - Host latency: 10.5855 ms (end to end 19.5386 ms, enqueue 1.85647 ms)
[11/28/2022-01:21:54] [I] Average on 10 runs - GPU latency: 9.69315 ms - Host latency: 10.4373 ms (end to end 19.2861 ms, enqueue 1.95906 ms)
[11/28/2022-01:21:54] [I] Average on 10 runs - GPU latency: 9.70017 ms - Host latency: 10.3944 ms (end to end 19.3087 ms, enqueue 1.81692 ms)
[11/28/2022-01:21:54] [I] Average on 10 runs - GPU latency: 9.65495 ms - Host latency: 10.3942 ms (end to end 19.2211 ms, enqueue 1.72121 ms)
[11/28/2022-01:21:54] [I] Average on 10 runs - GPU latency: 9.65385 ms - Host latency: 10.3949 ms (end to end 19.2132 ms, enqueue 1.82804 ms)
[11/28/2022-01:21:54] [I] Average on 10 runs - GPU latency: 9.68204 ms - Host latency: 10.3647 ms (end to end 19.2656 ms, enqueue 1.70457 ms)
[11/28/2022-01:21:54] [I] Average on 10 runs - GPU latency: 9.67172 ms - Host latency: 10.3892 ms (end to end 19.2918 ms, enqueue 1.89681 ms)
[11/28/2022-01:21:54] [I] Average on 10 runs - GPU latency: 9.68694 ms - Host latency: 10.3927 ms (end to end 19.2497 ms, enqueue 1.7677 ms)
[11/28/2022-01:21:54] [I] Average on 10 runs - GPU latency: 9.69882 ms - Host latency: 10.4191 ms (end to end 19.3633 ms, enqueue 1.74553 ms)
[11/28/2022-01:21:54] [I] Average on 10 runs - GPU latency: 9.7099 ms - Host latency: 10.4437 ms (end to end 19.2994 ms, enqueue 1.73564 ms)
[11/28/2022-01:21:54] [I] Average on 10 runs - GPU latency: 9.66166 ms - Host latency: 10.3627 ms (end to end 19.2514 ms, enqueue 1.70513 ms)
[11/28/2022-01:21:54] [I] Average on 10 runs - GPU latency: 9.6746 ms - Host latency: 10.4315 ms (end to end 19.2513 ms, enqueue 1.89272 ms)
[11/28/2022-01:21:54] [I] Average on 10 runs - GPU latency: 9.68583 ms - Host latency: 10.4422 ms (end to end 19.3063 ms, enqueue 1.74606 ms)
[11/28/2022-01:21:54] [I] Average on 10 runs - GPU latency: 9.67321 ms - Host latency: 10.4389 ms (end to end 19.2466 ms, enqueue 1.88253 ms)
[11/28/2022-01:21:54] [I] Average on 10 runs - GPU latency: 9.53226 ms - Host latency: 10.2362 ms (end to end 18.9538 ms, enqueue 1.80453 ms)
[11/28/2022-01:21:54] [I] Average on 10 runs - GPU latency: 9.54962 ms - Host latency: 10.2289 ms (end to end 18.9619 ms, enqueue 1.70771 ms)
[11/28/2022-01:21:54] [I] Average on 10 runs - GPU latency: 9.53646 ms - Host latency: 10.2664 ms (end to end 18.9249 ms, enqueue 1.71501 ms)
[11/28/2022-01:21:54] [I] Average on 10 runs - GPU latency: 9.52363 ms - Host latency: 10.2317 ms (end to end 18.9418 ms, enqueue 1.7092 ms)
[11/28/2022-01:21:54] [I] Average on 10 runs - GPU latency: 9.53206 ms - Host latency: 10.3252 ms (end to end 18.9318 ms, enqueue 1.81809 ms)
[11/28/2022-01:21:54] [I] Average on 10 runs - GPU latency: 9.54126 ms - Host latency: 10.3066 ms (end to end 18.955 ms, enqueue 1.74968 ms)
[11/28/2022-01:21:54] [I] Average on 10 runs - GPU latency: 9.53682 ms - Host latency: 10.2862 ms (end to end 18.9652 ms, enqueue 1.90947 ms)
[11/28/2022-01:21:54] [I] Average on 10 runs - GPU latency: 9.54568 ms - Host latency: 10.2775 ms (end to end 18.9686 ms, enqueue 1.80266 ms)
[11/28/2022-01:21:54] [I] Average on 10 runs - GPU latency: 9.5314 ms - Host latency: 10.2373 ms (end to end 18.9275 ms, enqueue 1.78276 ms)
[11/28/2022-01:21:54] [I] Average on 10 runs - GPU latency: 9.52163 ms - Host latency: 10.1887 ms (end to end 18.9425 ms, enqueue 1.57297 ms)
[11/28/2022-01:21:54] [I] Average on 10 runs - GPU latency: 9.52212 ms - Host latency: 10.3199 ms (end to end 18.9362 ms, enqueue 1.8717 ms)
[11/28/2022-01:21:54] [I] Average on 10 runs - GPU latency: 9.5271 ms - Host latency: 10.2847 ms (end to end 18.9282 ms, enqueue 1.7677 ms)
[11/28/2022-01:21:54] [I] Average on 10 runs - GPU latency: 9.52407 ms - Host latency: 10.2297 ms (end to end 18.9578 ms, enqueue 1.63486 ms)
[11/28/2022-01:21:54] [I] Average on 10 runs - GPU latency: 9.52361 ms - Host latency: 10.3314 ms (end to end 18.9119 ms, enqueue 1.82637 ms)
[11/28/2022-01:21:54] [I]
[11/28/2022-01:21:54] [I] === Performance summary ===
[11/28/2022-01:21:54] [I] Throughput: 98.1751 qps
[11/28/2022-01:21:54] [I] Latency: min = 10.1248 ms, max = 15.7787 ms, mean = 10.3718 ms, median = 10.312 ms, percentile(99%) = 10.9131 ms
[11/28/2022-01:21:54] [I] End-to-End Host Latency: min = 18.8291 ms, max = 170.617 ms, mean = 19.6443 ms, median = 19.0472 ms, percentile(99%) = 20.5318 ms
[11/28/2022-01:21:54] [I] Enqueue Time: min = 1.34326 ms, max = 2.96362 ms, mean = 1.78289 ms, median = 1.66577 ms, percentile(99%) = 2.81873 ms
[11/28/2022-01:21:54] [I] H2D Latency: min = 0.53125 ms, max = 4.94424 ms, mean = 0.646486 ms, median = 0.560547 ms, percentile(99%) = 1.05811 ms
[11/28/2022-01:21:54] [I] GPU Compute Time: min = 9.46582 ms, max = 10.7348 ms, mean = 9.62432 ms, median = 9.57275 ms, percentile(99%) = 10.1919 ms
[11/28/2022-01:21:54] [I] D2H Latency: min = 0.0830078 ms, max = 0.165771 ms, mean = 0.100955 ms, median = 0.0949707 ms, percentile(99%) = 0.150879 ms
[11/28/2022-01:21:54] [I] Total Host Walltime: 3.04558 s
[11/28/2022-01:21:54] [I] Total GPU Compute Time: 2.87767 s
[11/28/2022-01:21:54] [I] Explanations of the performance metrics are printed in the verbose logs.

注意：Throughput：越大代表吞吐量越大，模型推理速度越快

查看end-end数值也是可以的。

我对比只用--int8参数的：

trtexec.exe --onnx=yolov5n_ptq_detect_dynamic.onnx --saveEngine=ptq_int8_fp16.trtmodel --int8
=== Trace details ===
[11/28/2022-01:26:45] [I] Trace averages of 10 runs:
[11/28/2022-01:26:45] [I] Average on 10 runs - GPU latency: 11.9224 ms - Host latency: 12.9673 ms (end to end 26.5488 ms, enqueue 1.62947 ms)
[11/28/2022-01:26:45] [I] Average on 10 runs - GPU latency: 11.5994 ms - Host latency: 12.2361 ms (end to end 23.1969 ms, enqueue 1.63635 ms)
[11/28/2022-01:26:45] [I] Average on 10 runs - GPU latency: 11.5977 ms - Host latency: 12.2426 ms (end to end 23.1557 ms, enqueue 1.73348 ms)
[11/28/2022-01:26:45] [I] Average on 10 runs - GPU latency: 11.5168 ms - Host latency: 12.1538 ms (end to end 22.9547 ms, enqueue 1.61493 ms)
[11/28/2022-01:26:45] [I] Average on 10 runs - GPU latency: 11.4802 ms - Host latency: 12.1185 ms (end to end 22.8907 ms, enqueue 1.57255 ms)
[11/28/2022-01:26:45] [I] Average on 10 runs - GPU latency: 11.4798 ms - Host latency: 12.1063 ms (end to end 22.914 ms, enqueue 1.53058 ms)
[11/28/2022-01:26:45] [I] Average on 10 runs - GPU latency: 11.4603 ms - Host latency: 12.0941 ms (end to end 22.844 ms, enqueue 1.54984 ms)
[11/28/2022-01:26:45] [I] Average on 10 runs - GPU latency: 11.4718 ms - Host latency: 12.1083 ms (end to end 22.8685 ms, enqueue 1.54115 ms)
[11/28/2022-01:26:45] [I] Average on 10 runs - GPU latency: 11.4479 ms - Host latency: 12.0782 ms (end to end 22.8271 ms, enqueue 1.56331 ms)
[11/28/2022-01:26:45] [I] Average on 10 runs - GPU latency: 11.5097 ms - Host latency: 12.1481 ms (end to end 22.9363 ms, enqueue 1.56982 ms)
[11/28/2022-01:26:45] [I] Average on 10 runs - GPU latency: 11.4952 ms - Host latency: 12.1253 ms (end to end 22.9465 ms, enqueue 1.55604 ms)
[11/28/2022-01:26:45] [I] Average on 10 runs - GPU latency: 11.4569 ms - Host latency: 12.0909 ms (end to end 22.8553 ms, enqueue 1.53179 ms)
[11/28/2022-01:26:45] [I] Average on 10 runs - GPU latency: 11.4589 ms - Host latency: 12.1097 ms (end to end 22.8795 ms, enqueue 1.60232 ms)
[11/28/2022-01:26:45] [I] Average on 10 runs - GPU latency: 11.4927 ms - Host latency: 12.1319 ms (end to end 22.9371 ms, enqueue 1.61584 ms)
[11/28/2022-01:26:45] [I] Average on 10 runs - GPU latency: 11.4897 ms - Host latency: 12.1287 ms (end to end 22.937 ms, enqueue 1.60978 ms)
[11/28/2022-01:26:45] [I] Average on 10 runs - GPU latency: 11.4534 ms - Host latency: 12.0859 ms (end to end 22.8126 ms, enqueue 1.55414 ms)
[11/28/2022-01:26:45] [I] Average on 10 runs - GPU latency: 11.4507 ms - Host latency: 12.0853 ms (end to end 22.8174 ms, enqueue 1.50864 ms)
[11/28/2022-01:26:45] [I] Average on 10 runs - GPU latency: 11.4556 ms - Host latency: 12.0889 ms (end to end 22.8299 ms, enqueue 1.54136 ms)
[11/28/2022-01:26:45] [I] Average on 10 runs - GPU latency: 11.441 ms - Host latency: 12.0763 ms (end to end 22.8039 ms, enqueue 1.48589 ms)
[11/28/2022-01:26:45] [I] Average on 10 runs - GPU latency: 11.4657 ms - Host latency: 12.1066 ms (end to end 22.8564 ms, enqueue 1.55193 ms)
[11/28/2022-01:26:45] [I] Average on 10 runs - GPU latency: 11.467 ms - Host latency: 12.1061 ms (end to end 22.816 ms, enqueue 1.56992 ms)
[11/28/2022-01:26:45] [I] Average on 10 runs - GPU latency: 11.4546 ms - Host latency: 12.091 ms (end to end 22.8033 ms, enqueue 1.5084 ms)
[11/28/2022-01:26:45] [I] Average on 10 runs - GPU latency: 11.4535 ms - Host latency: 12.0993 ms (end to end 22.8176 ms, enqueue 1.61626 ms)
[11/28/2022-01:26:45] [I] Average on 10 runs - GPU latency: 11.4425 ms - Host latency: 12.0855 ms (end to end 22.7886 ms, enqueue 1.52632 ms)
[11/28/2022-01:26:45] [I] Average on 10 runs - GPU latency: 11.4571 ms - Host latency: 12.0936 ms (end to end 22.8192 ms, enqueue 1.61838 ms)
[11/28/2022-01:26:45] [I] Average on 10 runs - GPU latency: 11.4402 ms - Host latency: 12.0712 ms (end to end 22.821 ms, enqueue 1.50779 ms)
[11/28/2022-01:26:45] [I]
[11/28/2022-01:26:45] [I] === Performance summary ===
[11/28/2022-01:26:45] [I] Throughput: 85.541 qps
[11/28/2022-01:26:45] [I] Latency: min = 12.0249 ms, max = 17.3826 ms, mean = 12.1473 ms, median = 12.0969 ms, percentile(99%) = 12.6057 ms
[11/28/2022-01:26:45] [I] End-to-End Host Latency: min = 22.6846 ms, max = 52.0965 ms, mean = 23.0261 ms, median = 22.8641 ms, percentile(99%) = 24.234 ms
[11/28/2022-01:26:45] [I] Enqueue Time: min = 1.37061 ms, max = 2.40161 ms, mean = 1.57101 ms, median = 1.54297 ms, percentile(99%) = 1.99091 ms
[11/28/2022-01:26:45] [I] H2D Latency: min = 0.52832 ms, max = 4.58092 ms, mean = 0.559679 ms, median = 0.539001 ms, percentile(99%) = 0.637695 ms
[11/28/2022-01:26:45] [I] GPU Compute Time: min = 11.4016 ms, max = 12.7121 ms, mean = 11.4946 ms, median = 11.459 ms, percentile(99%) = 11.9795 ms
[11/28/2022-01:26:45] [I] D2H Latency: min = 0.0827026 ms, max = 0.121033 ms, mean = 0.0929625 ms, median = 0.0922852 ms, percentile(99%) = 0.116089 ms
[11/28/2022-01:26:45] [I] Total Host Walltime: 3.03948 s
[11/28/2022-01:26:45] [I] Total GPU Compute Time: 2.98861 s
[11/28/2022-01:26:45] [I] Explanations of the performance metrics are printed in the verbose logs.

通过以上Throughput或者end-end数值对比，可以看到使用--int8 --fp16的速度明显快于

--int8（实际是--int8 + fp32）。这样也就符合我们的预期了！

（3）为engine画图

这些要安装在conda虚拟环境下面

 另外注意trex安装（要在虚拟环境下面）：

git clone https://github.com/NVIDIA/TensorRT.git
cd TensorRT/tools/experimental/trt-engine-explorer
python -m pip install -e .

运行命令：

(yolo) wt-yjy@vt82-NF5468M6:~/TensorRT-8.5.1/tools/experimental/trt-engine-explorer/utils$ python process_engine.py yolov5n_ptq_detect_dynamic.onnx yolov5n_ptq_int8 int8 fp16

 

同时使用了int8和fp16来生成engine

 

 使用浏览器打开svg如下：

 

 

只展示如上的示意图。可以看到这个engine中包含了int8+fp16+少量的fp32

，如果我们只是用int8生成的engine呢？我们可以看下：

 

 明显可以看到，我们int8的engine中显示图中只有fp32和int8的算子。

注意：上面的可以使用process_engine.py来画，也可以使用draw_engine.py来画

（4）画出各个engine的性能图

利用这两个jupyter文件

首先看compare_engine：

import IPython
from ipywidgets import widgets
from trex import *


# Configure a wider output (for the wide graphs)
set_wide_display()

engine_name_1 = "./yolov5n_ptq_int8/yolov5n_ptq_detect_dynamic.onnx.engine"
engine_name_2 = "./yolov5n_ptq_int8_fp16/yolov5n_ptq_detect_dynamic.onnx.engine"

plan1 = EnginePlan(f'{engine_name_1}.graph.json', f'{engine_name_1}.profile.json', f"{engine_name_1}.profile.metadata.json")
plan2 = EnginePlan(f'{engine_name_2}.graph.json', f'{engine_name_2}.profile.json', f"{engine_name_2}.profile.metadata.json")
plans = (plan1, plan2)

compare_engines_summaries_tbl(plans, orientation='vertical')
compare_engines_overview(plans)

compare_engines_layer_latencies(
    plan1, plan2,
    # Allow for 3% error grace threshold when color highlighting performance differences
    threshold=0.03,
    # Inexact matching uses only the layer's first input and output to match to other layers.
    exact_matching=True)

compare_engines_layer_details(plans[0], plans[1])
report_card_perf_overview_widget(plan1);
print(plan1.name)



report_card_perf_overview_widget(plan2);
print(plan2.name)


report_card_table_view(plan1);
print(plan1.name)


report_card_table_view(plan2);
print(plan2.name)



for plan in plans:
    graph = to_dot(plan, layer_type_formatter, display_regions=True, expand_layer_details=True)
    render_dot(graph, plan.name, 'svg')

 

随便展示图如下：

 

 

engine_report_card：

import IPython
from ipywidgets import widgets
from trex import *

# Choose an engine file to load.
engine_name = "./yolov5n_ptq_int8/yolov5n_ptq_detect_dynamic.onnx.engine"
engine_name = "./yolov5n_ptq_int8_fp16/yolov5n_ptq_detect_dynamic.onnx.engine"

set_wide_display()

plan = EnginePlan(f"{engine_name}.graph.json", f"{engine_name}.profile.json", f"{engine_name}.profile.metadata.json")

report_card_draw_plan_graph_extended(plan, engine_name)


# "Average time" refers to the sum of the layer latencies, when profiling layers separately
# "Latency" refers to the [min, max, mean, median, 99% percentile] of the engine latency measurements, when timing the engine w/o profiling layers.
plan.summary()

layer_latency_sunburst(plan.df, "Layers Latencies (%)")

report_card_table_view(plan)

plot_engine_timings(timing_json_file= f"{engine_name}.timing.json")

report_card_perf_overview_widget(plan);

report_card_memory_footprint_widget(plan);

convs = plan.get_layers_by_type('Convolution')
report_card_convolutions_overview_widget(convs)

latency_vs_prec_per_conv = partial(
    plotly_bar2,
    convs,
    values_col='latency.pct_time',
    names_col='Name',
    color='tactic')

latency_vs_prec_per_conv("Latency per Layer (color=Tactics)")

tactic_cnt = group_count(plan.df, 'tactic')
display_df(tactic_cnt)


report_card_gemm_MNK(plan);
report_card_gemm_MNK_scatter(plan);
report_card_efficiency_vs_latency_3d(plan);
report_card_perf_scatter(plan);
display_df(ConvLinter(plan).lint())
report_card_reformat_overview(plan)
display_df(ReformatLinter(plan).lint())
display_df(SliceLinter(plan).lint())
display_df(QDQLinter(plan).lint())
report_card_pointwise_lint(plan)

summary = ExcelSummary(plan, path="default_summary.xlsx")
summary.generate_default_summary() # 'generate_default_summary' automatically saves the file


summary = ExcelSummary(plan, path="customized_summary.xlsx")
summary.add_dataframes({"df": plan.df})
summary.add_images({"trex_logo": "../images/trex_logo.png"})
summary.add_dataframes({"clean_df": clean_for_display(plan.df)})
summary.save() # 'save' needs to be called to propagate the changes to the file

# The context manager automatically saves the file
with ExcelSummary(plan, path="customized_summary_with_manager.xlsx") as summary:
    summary.add_dataframes({"df": plan.df})
    summary.add_images({"trex_logo": "../images/trex_logo.png"})
    summary.add_dataframes({"clean_df": clean_for_display(plan.df)})

展示部分图如下：

 

可以看到整个跟我们前面的也是符合一致的，前面使用了int8和fp16转换后，仍然存在少量部分fp32的算子操作。

注意：以上是PTQ的量化结果。

本文来自互联网用户投稿，该文观点仅代表作者本人，不代表本站立场。本站仅提供信息存储空间服务，不拥有所有权，不承担相关法律责任。如若转载，请注明出处：http://www.coloradmin.cn/o/44121.html

如若内容造成侵权/违法违规/事实不符，请联系多彩编程网进行投诉反馈，一经查实，立即删除！

相关文章

小白学流程引擎-FLowable(二) — 从零搭建自己的FLowable服务 — 搭建流程服务-FLowable的新手指南

小白学流程引擎-FLowable(二) — 从零搭建自己的FLowable服务 — 搭建流程服务-FLowable的新手指南

一、介绍 纵览Gitee搜索Flowable开源项目，大多都是已开发好的项目，而笔者从零开始搭建属于自己的Flowable引擎，并且是可以拿到生产上使用的。 二、软件架构 Springboot Flowable modeler idm Mysql SrpingBoot version：2.7…
阅读更多...
牛客网verilog刷题知识点盘点(75道题的版本)

牛客网verilog刷题知识点盘点(75道题的版本)

牛客网verilog刷题知识点盘点(75道题的版本) 还有几个坑没填 任务和函数 1.任务和函数必须在模块内定义，其作用范围仅适用于该模块，可以在模块内多次调用。 2.任务和函数中可以声明局部变量，如寄存器，时间，整数&…
阅读更多...
企业云工如何高效居家协同办公？试试这个方法

企业云工如何高效居家协同办公？试试这个方法

近日的疫情反扑让全国各地的防疫压力增加，并且甚至很多地方不知道的啥时候就要被居家隔离，所以在新的防疫背景下，居家就顺理成章地成为一种常态化的学习和工作方式。 现在越来越多公司也将日常业务搬到了线上，以保证疫情期间公司的…
阅读更多...
基于ssm的旅游网站的设计与实现

基于ssm的旅游网站的设计与实现

项目描述 临近学期结束，还是毕业设计，你还在做java程序网络编程，期末作业，老师的作业要求觉得大了吗?不知道毕业设计该怎么办?网页功能的数量是否太多?没有合适的类型或系统?等等。这里根据疫情当下，你想解决的问…
阅读更多...
vue——路由

vue——路由

目录 一、介绍路由 1、路由是什么 Vue Router 是 Vue.js 的官方路由。它与 Vue.js 核心深度集成，让用 Vue.js 构建单页应用变得轻而易举。 2、为什么要使用路由 3、主要用途 二、安装路由 1、安装命令 2、配置文件 3、在main.js 中进行挂载 三、案例展示 …
阅读更多...
Linux服务器配置与管理（基于Centos7.2）任务目标（五）

Linux服务器配置与管理（基于Centos7.2）任务目标（五）

文章目录一、知识目标二、能力目标二、任务实施任务一：安装Apache任务二：配置简单Web站点任务三：配置基于主机名的虚抛主机任务四：安装vsftpd任务五：匿名用户访问FTP服务器任务六：本地用户配置一、知识目标…
阅读更多...
HummerRisk 使用场景-混合云安全治理（2）--阿里云安全最佳实践

HummerRisk 使用场景-混合云安全治理（2）--阿里云安全最佳实践

背景： 阿里云安全最佳实践，是基于众多客户上云的成功案例萃取而成的最优化企业上云指导。每个最佳实践包括使用场景、多产品部署架构及部署手册。 最佳实践目前覆盖23种常用场景，目前有200篇最佳实践，涉及100款以上阿里云产品的…
阅读更多...
20221128-1Spring_day02(资料来自黑马程序)

20221128-1Spring_day02(资料来自黑马程序)

Spring_day02 今日目标 掌握IOC/DI配置管理第三方bean掌握IOC/DI的注解开发掌握IOC/DI注解管理第三方bean完成Spring与Mybatis及Junit的整合开发 1，IOC/DI配置管理第三方bean 前面所讲的知识点都是基于我们自己写的类，现在如果有需求让我们去管理第三方…
阅读更多...
安卓讲课笔记5.11 菜单

安卓讲课笔记5.11 菜单

文章目录零、本讲学习目标一、导入新课二、新课讲解（一）菜单概述1、选项菜单2、上下文菜单3、子菜单（二）选项菜单案例演示1、创建安卓应用2、准备图片素材3、字符串资源文件4 、主布局资源文件5、主界面类实现功能6、启动应用&…
阅读更多...
2. Vue3 Composition API

2. Vue3 Composition API

Composition API 1.Composition API 接下来我们来介绍一下Vue3中新增的Composition API如何使用。注意Composition API仅仅是Vue3中新增的API，我们依然可以使用Options API。先来实现一下之前演示的获取鼠标位置的案例。做这个案例之前，需要先介绍一下…
阅读更多...
MybatisPlus简单使用与自定义sql以及通过自定义sql实现多表联查的分页查询

MybatisPlus简单使用与自定义sql以及通过自定义sql实现多表联查的分页查询

MybatisPlus简单使用与自定义sql以及通过自定义sql实现多表联查的分页查询前言1. mybatis的简单使用2. MybatisPlus 的简单使用2.1 入门2.1.1 简单配置2.1.2 入门例子2.1.3 测试2.2 MybatisPlus自带封装的增删改查2.2.1 傻瓜式使用2.2.2 批量添加数据3 MybatisPlus 动态查询sql…
阅读更多...
计算机网络---数据链路层扩展的以太网

计算机网络---数据链路层扩展的以太网

（一）在物理层扩展以太网 使用光纤扩展：主机使用光纤和一对光纤调制解调器连接到集线器 使用集线器扩展：主机使用光纤和一对光纤调制解调器连接到集线器 使用集线器扩展的优缺点 优点 ：使原来属于不同碰撞域&#xff0…
阅读更多...
Nodejs -- Express中间件的概念及基本使用

Nodejs -- Express中间件的概念及基本使用

文章目录1 中间件的概念1.1 什么是中间件1.2 现实生活中的例子1.3 Express中间件的调用流程1.4 Express的中间件的格式1.5 next函数的作用2 Express中间件初体验2.1 定义中间件函数2.2 全局生效的中间件2.3 定义全局中间件的简化形式2.4 中间件的作用2.5 定义多个全局中间件2.6…
阅读更多...
机器学习-（手推）线性回归3-正则化-岭回归（Ridge）-频率角度贝叶斯角度

机器学习-（手推）线性回归3-正则化-岭回归（Ridge）-频率角度贝叶斯角度

一、正则化-岭回归-频率角度 回顾： Loss Function： 过拟合的解决方法： ①最直接：加数据 ②降维（特征选择/特征提取（PCA）） ③正则化（对参数空间，例如w的约…
阅读更多...
预处理，编译，汇编，链接，全过程。

预处理，编译，汇编，链接，全过程。

编译，链接，全过程。背景知识预处理：1.宏定义指令，如#define MAX 1；2.条件编译指令，如#ifdef、 #ifndef、#else、#elif、#endif等。3.头文件包含指令，如#include等。4.特殊符号，预编译…
阅读更多...
SQL explain解析器

SQL explain解析器

EXPLAIN 参数前言字段参数id 查询编号select_type 关联类型SIMPLEPRIMARYUNION & UNION RESULTDERIVEDSUBQUERYDEPENDENTUNCACHEABLEMATERIALIZEDtable 表名partitions 数据的分区信息type 关联类型system & consteq_refreffulltextref_or_nullindex_mergeunique_subqu…
阅读更多...
必备表格软件-FineReport正则表达式简介

必备表格软件-FineReport正则表达式简介

1. 概述 1.1 应用场景 有时候我们需要用到正则表达式进行信息的校验。 例如有一张使用了「文本控件」的查询报表，输入「销售员」姓名后可查询销售员的销售情况，此时希望设置销售员文本控件的填入信息校验内容为：若填入内容不是中文或中文的…
阅读更多...
【无标题】接口测试用例设计（精华）

【无标题】接口测试用例设计（精华）

接口测试 请求头 请求头中的Content-Type有哪几种： 1.application/x-www-form-urlencoded 最常见的 POST 提交数据的方式，原生Form表单，如果不设置 enctype 属性，默认为application/x-www-form-urlencoded 方式提交数据。 2.appli…
阅读更多...
Node.js 入门教程 19 package-lock.json 文件

Node.js 入门教程 19 package-lock.json 文件

Node.js 入门教程 Node.js官方入门教程 Node.js中文网 本文仅用于学习记录，不存在任何商业用途，如侵删 文章目录Node.js 入门教程19 package-lock.json 文件19.1 示例19 package-lock.json 文件 在版本 5 中，npm 引入了 package-lock.json 文…
阅读更多...
BUUCTF Reverse/[2019红帽杯]xx

BUUCTF Reverse/[2019红帽杯]xx

BUUCTF Reverse/[2019红帽杯]xx 先看下文件信息：没有加壳、64位程序 看别人wp时候发现个好东东,就是这个findcrypt插件，可以看加密算法的，具体安装可以看这个IDA7.5安装findcrypt3插件 可以看到这是tea加密 先一点点分析代码，输入…
阅读更多...
推荐文章
最新文章

Copyright @ 2022~2023