首页
学习
活动
专区
圈层
工具
发布
    • 综合排序
    • 最热优先
    • 最新优先
    时间不限
  • 来自专栏人工智能

    大规模语言模型知识编辑:定位-编辑-再训练的一致性误差分析

    retraining_data = self. _generate_retraining_data(subject, relation, object_new) retraining_result = self.retraining_analyzer.analyze_local_retraining_impact ( edit_location, retraining_data, num_epochs=2 ) audit_report['retraining_analysis '] = { 'impact_analysis': retraining_result, 'risk_factors': self. _identify_retraining_risks(retraining_result) } # 阶段4: 一致性验证 print("[阶段4

    23810编辑于 2025-12-05
  • 来自专栏机器之心

    教程 | 从零开始:TensorFlow机器学习模型快速部署指南

    例如,https://www.tensorflow.org/tutorials/image_retraining 页面上有如何微调 ImageNet 模型对花样本数据集(3647 张图像,5 个类别)进行分类的教程 :retrain \ tensorflow/examples/image_retraining:label_image \ && \ bazel-bin/tensorflow/examples /image_retraining/retrain \ --image_dir "$HOME"/flower_photos \ --how_many_training_steps=200 && \ bazel-bin/tensorflow/examples/image_retraining/label_image \ --graph=/tmp/output_graph.pb \ --labels /image_retraining/label_image.py。

    1.1K50发布于 2018-05-10
  • 来自专栏大数据文摘

    手把手:我的深度学习模型训练好了,然后要做啥?

    比如,https://www.tensorflow.org/tutorials/image_retraining有一个关于如何微调Imagenet模型(在1.2M图像上训练1000个类别)以对花进行分类的样本数据集 :retrain \ tensorflow/examples/image_retraining:label_image \ && \ bazel-bin/tensorflow/examples /image_retraining/retrain \ --image_dir "$HOME"/flower_photos \ --how_many_training_steps=200 && 对我而言,这个脚本的位置在: in bazel-bin/tensorflow/examples/image_retraining/label_image.runfiles/org_tensorflow/ tensorflow/examples/image_retraining/label_image.py.

    2K20发布于 2018-05-24
  • 来自专栏大数据风控

    评分卡模型开发-主标尺设计及模型验证

    purpose"])=="business") { data_tmp[i,"purpose"]<-as.character("others/repairs/business") } #合并retraining 、education if(as.character(data_tmp[i,"purpose"])=="retraining") { data_tmp[i,"purpose"]<-as.character ("retraining/education") } if(as.character(data_tmp[i,"purpose"])=="education") { data_tmp[ i,"purpose"]<-as.character("retraining/education") } } ##purpose变量降维结束## ###用R代码实现打分卡模型### data1<-as.data.frame purpose"]=="car(new/used)") { score_purpose<--1 }else if(lst[,"purpose"]=="<em>retraining</em>

    2.1K100发布于 2018-01-09
  • 来自专栏人工智能

    从零开始:TensorFlow机器学习模型快速部署指南

    例如,https://www.tensorflow.org/tutorials/image_retraining 页面上有如何微调 ImageNet 模型对花样本数据集(3647 张图像,5 个类别)进行分类的教程 二者作为命令行参数被输入至 label_image.py (https://github.com/tensorflow/tensorflow/blob/r1.4/tensorflow/examples/image_retraining 它的地址为 bazel-bin/tensorflow/examples/image_retraining/label_image.runfiles/org_tensorflow/tensorflow/examples /image_retraining/label_image.py。

    1.7K70发布于 2018-02-02
  • 来自专栏大数据风控

    评分法模型开发-WOE值计算

    d[i,"purpose"]<-as.character("others/repairs/business") } if(as.character(d[i,"purpose"])=="<em>retraining</em> ") { d[i,"purpose"]<-as.character("retraining/education") } if(as.character(d[i,"purpose"]) =="education") { d[i,"purpose"]<-as.character("retraining/education") } } new_data<-cbind(discrete_data

    1.7K60发布于 2018-01-09
  • 来自专栏大数据风控

    评分卡模型开发-基于逻辑回归的标准评分卡实现

    i,"purpose"])=="business") { d[i,"purpose"]<-as.character("others/repairs/business") } #合并retraining 、education if(as.character(d[i,"purpose"])=="retraining") { d[i,"purpose"]<-as.character("retraining education") } if(as.character(d[i,"purpose"])=="education") { d[i,"purpose"]<-as.character("retraining =="radio/television/furniture/equipment") { purpose_WoE[i]<--0.23035114 } if(purpose[i]=="<em>retraining</em>

    5.2K81发布于 2018-01-09
  • 来自专栏机器之心

    AAAI 2019 提前看:融合质量不理想数据

    Partial Label Learning with Self-Guided Retraining 链接:http://www.ntu.edu.sg/home/boan/papers/AAAI19_Retraining.pdf 所介绍的算法叫做 SURE(Self-gUided REtraining)。 基于此思想,我们的介绍从对问题建模开始。 Partial Label Learning with Self-Guided Retraining. AAAI.] Partial Label Learning with Self-Guided Retraining. AAAI.] [Image: image.png] ? Partial Label Learning with Self-Guided Retraining. AAAI.] [Image: image.png] ?

    51810发布于 2019-04-29
  • 来自专栏vanguard

    Yolov5 Retrain

    训练模型 Retraining python train.py --img 640 --batch 16 --epochs 100 --data ./data/person.yaml --cfg .

    84330发布于 2021-07-21
  • 来自专栏智能生信

    [bioRxiv | 论文简读]

    简读分享 | 赵晏浠 编辑 | 龙文韬 论文题目 OpenFold: Retraining AlphaFold2 yields new insights into its learning mechanisms

    31510编辑于 2022-12-29
  • 来自专栏AutoML(自动机器学习)

    ICCV 2021 | BN-NAS: 只训练BN层来自动搜索模型

    Methodology 如下图示,整个算法包含三个步骤: Supernet training subnet searching subnet retraining [b4gha64e23.png? 1631640310&q-header-list=&q-url-param-list=&q-signature=9872b44d4b0c8888be2a679f3407e21b27275124] 2.3 Subnet Retraining

    51700发布于 2021-09-14
  • 来自专栏人工智能前沿讲习

    【AIDL专栏】Intel研究总监陈玉荣:如何高效的设计深度学习算法?(附PPT下载)

    这方面的工作主要是15年和16年提出的Pruning with retraining(NIPS’15, ICLR’16, ISCA’16),它的好处是保证accuracy 不loss、Model-free 该算法采用比Pruning with retraining更短的重训练时间,取得了比Pruning with retraining更高的压缩率,可以轻易地将LeNet和AlexNet这两个经典网络的参数总量分别压缩

    62220发布于 2020-05-14
  • 来自专栏全栈程序员必看

    信用标准评分卡模型开发及实现方案_信用评分卡模型的建立

    、education if(as.character(d[i,"purpose"])=="retraining") { d[i,"purpose"]<-as.character("retraining/ education") } if(as.character(d[i,"purpose"])=="education") { d[i,"purpose"]<-as.character("retraining 、education if(as.character(data_tmp[i,"purpose"])=="retraining") { data_tmp[i,"purpose"]<-as.character ("retraining/education") } if(as.character(data_tmp[i,"purpose"])=="education") { data_tmp[i,"purpose "]<-as.character("retraining/education") } } ##purpose变量降维结束## ###用R代码实现打分卡模型### data1<-as.data.frame

    1.7K20编辑于 2022-11-19
  • 来自专栏大数据风控

    信用标准评分卡模型开发及实现

    i,"purpose"])=="business") { d[i,"purpose"]<-as.character("others/repairs/business") } #合并retraining 、education if(as.character(d[i,"purpose"])=="retraining") { d[i,"purpose"]<-as.character("retraining =="radio/television/furniture/equipment") { purpose_WoE[i]<--0.23035114 } if(purpose[i]=="<em>retraining</em> 、education if(as.character(data_tmp[i,"purpose"])=="<em>retraining</em>") { data_tmp[i,"purpose"]<-as.character ("retraining/education") } if(as.character(data_tmp[i,"purpose"])=="education") { data_tmp[

    2.9K31发布于 2019-07-02
  • 来自专栏AI科技评论

    ICCV 2021放榜!发现一篇宝藏论文——如何一次性训练100,000+个Vision Transformers?

    不仅仅使得其在搜索后不再需要重新训练(retraining)结构,节约了搜索时间,也使得其能在各种不同的计算资源限制下快速搜索最优结构。 ? 在这种机制下,搜索空间的网络结构几乎都能被充分训练,省去了结构搜索后重新训练(Retraining)的时间。大量实验表明所提出的算法可以提高超网的排序能力并找到高性能的结构。

    88781发布于 2021-07-27
  • 来自专栏技术汇总专栏

    CI/CD与模型监控平台集成MLOps系统实现的全面路径

    airflow.operators.python import PythonOperator from datetime import datetime def retrain_model(): print("Retraining dag = DAG( 'model_retraining', description='Automated model retraining', schedule_interval

    53700编辑于 2025-07-28
  • AI体征营养指导系统:从数据到建议的技术闭环

    这些新的数据会作为强化信号,通过在线学习(Online Learning)或定期的模型再训练(Retraining),不断微调和优化推荐算法,让建议越来越精准、越来越贴合用户的实际生活。

    29810编辑于 2025-10-07
  • 来自专栏AI科技评论

    开发 | 机器学习零基础?手把手教你用TensorFlow搭建图像分类器

    第四步:下载图片 下面的步骤则基于TensorFlow的图形再训练案例(详情链接:https://www.tensorflow.org/versions/master/how_tos/image_retraining 如果你想知道后台具体是怎么运行的,可以点击https://www.tensorflow.org/versions/master/how_tos/image_retraining/#bottlenecks

    1.1K60发布于 2018-03-09
  • 从零到工业级落地的全栈实战指南

    最小可行产品(MVP)迭代策略: 第1周:用FastAPI搭建MNIST分类服务(准确率92%) 第3周:加入数据增强和模型蒸馏(准确率提升至96%) 第6周:实现模型监控和自动retraining pipeline

    44010编辑于 2025-03-19
  • 来自专栏AI 创作日记

    AI 创作日记 | 门店巡检的重构:DeepSeek多模态技术落地纪实

    () return f"Model upgraded to {self.model_version}" return "Data insufficient for retraining version_upgrade() return f"Model upgraded to {self.model_version}" return "Data insufficient for retraining

    57720编辑于 2025-03-24
领券