- 综合排序
- 最热优先
- 最新优先
pytorch computation graph
什么是 PyTorch 计算图（Computation Graph）在深度学习中，计算图是一个有向图，用于表示 张量操作的依赖关系。 特点：动态计算图（Dynamic Computation Graph）：图在前向计算时即时构建，每次 forward 都是新的图。节点（Node）：表示张量操作（例如加法、乘法、卷积）。 总结计算图（Computation Graph） 是 PyTorch 自动求导的核心机制。PyTorch 使用 动态计算图，灵活、可调试、支持任意控制流。
27910编辑于 2025-08-14- 来自专栏产品经理的人工智能学习库
算力 – computation
在普通电脑中，CPU就提供了算力帮助电脑快速运行。玩游戏的时候需要显卡提供算力，帮助电脑快速处理图形。而在 人工智能中，需要有类似CPU和GPU的硬件来提供算力，帮助算法快速运算出结果。
2.6K30发布于 2019-12-18 - 来自专栏大数据和云计算技术
Automatic Management of Data and Computation in Datacenters
image.png 最近在研究数据中心的数据管理和性能优化，看了一篇2010的论文Nectar:Automatic Management of Data and Computation in Datacenters
902100发布于 2018-03-08 - 来自专栏计算机视觉理论及其实现
torch(五)、Locally disabling gradient computation
torch.enable_grad(), and torch.set_grad_enabled() are helpful for locally disabling and enabling gradient computation See Locally disabling gradient computation for more details on their usage.
50130编辑于 2022-09-03 详解RuntimeError: one of the variables needed for gradient computation has been mo
在深度学习中，经常会使用自动微分技术（Automatic Differentiation）来计算模型参数的梯度，以进行模型的优化训练。然而，有时我们可能会遇到一个异常：RuntimeError: 一个用于梯度计算的变量已被就地操作修改。本文将详细解释这个异常的原因及解决方法。
3.7K10编辑于 2024-01-25yolov10训练报错GET was unable to find an engine to execute this computation
C++ engine to run the backward pass RuntimeError: GET was unable to find an engine to execute this computation
24400编辑于 2025-07-17- 来自专栏一个会写诗的程序员的博客
Actor 分布式并行计算模型: The Actor Model for Concurrent Computation
分布式计算技术之Actor计算模式 The Actor Model for Concurrent Computation 背景介绍 计算机CPU的计算速度提高（频率的提高）是有限度的，我们能做的是放入多个计算核心 The model The actor model is a conceptual model to deal with concurrent computation. Actors An actor is the primitive unit of computation. It’s the thing that receives a message and do some kind of computation based on it.
2.6K20编辑于 2021-12-20 - 来自专栏机器人课程与技术
ROS2机器人编程简述humble-第三章-COMPUTATION GRAPH .2
控制逻辑解释：输入的感知信息并产生控制命令（输出）。这个逻辑就是要用FSM实现的。逻辑控制将以20 Hz反复运行。执行频率取决于发布控制命令。
90020编辑于 2023-02-10 - 来自专栏函数式编程语言及工具
Akka-Cluster（5）- load-balancing with backoff-supervised stateless computation - 无状态任务集群节点均衡分配
分布式程序运算是一种水平扩展（scale-out）运算模式，其核心思想是能够充分利用服务器集群中每个服务器节点的计算资源，包括：CPU、内存、硬盘、IO总线等。首先对计算任务进行分割，然后把细分的任务分派给各节点去运算。细分的任务相互之间可以有关联或者各自为独立运算，使用akka-cluster可以把任务按照各节点运算资源的负载情况进行均匀的分配，从而达到资源的合理充分利用以实现运算效率最大化的目的。如果一项工作可以被分割成多个独立的运算任务，那么我们只需要关注如何合理地对细分任务进行分配以实现集群节点的负载均衡，这实际上是一种对无需维护内部状态的运算任务的分配方式：fire and forget。由于承担运算任务的目标actor具体的部署位置是由算法决定的，所以我们一般不需要控制指定的actor或者读取它的内部状态。当然，如果需要的话我们还是可以通过嵌入消息的方式来实现这样的功能。
1.8K20发布于 2019-01-07 - 来自专栏机器人课程与技术
ROS2机器人编程简述humble-第四章-COMPUTATION GRAPH .2
chatgpt： ROS2机器人的COMPUTATION GRAPH概念是指，通过构建一个图形结构，将机器人的计算任务分解成一系列的可执行步骤。
98810编辑于 2023-02-26 - 来自专栏机器学习原理
pytorch基础笔记pytorch的windoows下安装pytorch的使用Dynamic Computation Graph用pytorch实现线性回归
Tensor与Numpy的最大不同：Tensor可以在GPU上运算 Dynamic Computation Graph 它可以让我们的计算模型更灵活、复杂 它可以让反向传播算法随时进行
98250发布于 2018-06-19 - 来自专栏素质云笔记
python︱写markdown一样写网页，代码快速生成web工具：streamlit 缓存（五）
any function used inside the cached function cache在后台操作的步骤为： For example, when the function expensive_computation input key is not found in the cache, then: - Executes the cached function (i.e. output = expensive_computation (a, b): # Added this st.write("Cache miss: expensive_computation(", a, ",", b, ") ran") @st.cache(suppress_st_warning=True) def expensive_computation(a, b): st.write("Cache miss: expensive_computation @st.cache(suppress_st_warning=True) def expensive_computation(a, b): st.write("Cache miss: expensive_computation
85720编辑于 2021-12-07 - 来自专栏小徐学爬虫
Django代码中的TypeError 'float' object is not callable
例如：discount = computation.final_credit_rate( computation.pmt_loaner_final() + computation.pmt_broker_final (), computation.total_compounded_loaner() + computation.total_compounded_broker())通过使用括号来调用property 例如：discount = computation.final_credit_rate( int(computation.pmt_loaner_final()) + int(computation.pmt_broker_final ()), int(computation.total_compounded_loaner()) + int(computation.total_compounded_broker()))通过将a+ 例如：final_credit_rate = computation.get_final_credit_rate( int(computation.final_pmt_without_withdrawal_fees
71210编辑于 2024-01-23 - 来自专栏韩曙亮的移动开发专栏
【Flutter】FutureBuilder 异步编程 ( FutureBuilder 构造方法 | AsyncSnapshot 异步计算 )
= null); /// Current state of connection to the asynchronous computation. has completed /// successfully, if the computation did not return a non-null value. none, /// Connected to an asynchronous computation and awaiting interaction. active, /// Connected to a terminated asynchronous computation. done, } final T? has completed /// successfully, if the computation did not return a non-null value.
1.4K20编辑于 2023-03-29 - 来自专栏人工智能与演化计算成长与进阶
【征稿】2023 IEEE 进化计算国际会议专题：进化计算机视觉和图像处理
Session on Evolutionary Computer Vision and Image Processing (ECVIP) IEEE Congress on Evolutionary Computation In turn, evolutionary computation methods have been shown to be more efficient than classical optimization As a result of the convergence of the computer vision and evolutionary computation research fields, many Evolutionary computation paradigms include genetic algorithms, genetic programming, evolutionary strategies Her research focuses mainly on computer vision, image analysis, machine learning, evolutionary computation
1.7K40编辑于 2023-02-23 - 来自专栏人工智能与演化计算成长与进阶
征稿|CEC 2023 Special Session on "EC in Healthcare Industry"
can be formulated as scheduling, planning, predicting, and optimization problems, where evolutionary computation diagnosis and predicting for disease prevention have not been properly formulated for evolutionary computation techniques, and many evolutionary computation techniques are widely adopted in the healthcare community Application of evolutionary computation for job scheduling, such as ambulance scheduling, nurse scheduling Please select the main research topic as the Special Session on “Evolutionary Computation in Healthcare
56930编辑于 2023-02-23 - 来自专栏Rust语言学习交流
【Rust日报】2023-09-13 RustRover – JetBrains 推出的独立 Rust IDE
. fn expensive_computation<'a>(msg: OwnedMessage) -> String { info! ("Starting expensive computation on message {}", msg.offset()); thread::sleep(Duration::from_millis let computation_result = tokio::task::spawn_blocking(|| expensive_computation(owned_message )) .await .expect("failed to wait for expensive computation ("CARGO_PKG_VERSION").unwrap_or("")) .about("Asynchronous computation example") .arg(
66520编辑于 2023-09-26 - 来自专栏深度强化学习实验室
【Richard S. Sutton】谈 The Bitter Lesson(AI 研究中痛苦的教训)
Most AI research has been conducted as if the computation available to the agent were constant (in which Learning by self play, and learning in general, is like search in that it enables massive computation On the other side were newer methods that were more statistical in nature and did much more computation Deep learning methods rely even less on human knowledge, and use even more computation, together with even as the available computation becomes very great.
3.1K22编辑于 2022-09-23 - 来自专栏进步集
MIT6.006农民工学算法导论
文章目录 GOAL of algorithm Algorithms and computation Model of computation 学习顺序 GOAL of algorithm goal：—- communication solve compucating problems prove correctness argue efficiency Algorithms and computation Effeciency -asymptotic analysis渐进分析 O（）———- upper bound Omega（）——lower bounds Theta()——-both Model of computation
37240编辑于 2022-11-13 - 来自专栏阿策小和尚
【Flutter 专题】90 图解 Dart 单线程实现异步处理之 Future (一)
(FutureOr<T> computation()) { _Future<T> result = new _Future<T>(); Timer.run(() { try (() => print('Future(FutureOr<T> computation()) 01')); Future(() { print('Future(FutureOr<T> computation ()) 02'); return 'Future(FutureOr<T> computation()) 02 result!' ; }).then((val) => print(val)); Future(() { print('Future(FutureOr<T> computation()) 03'); return 'Future(FutureOr<T> computation()) 03 result!'
1.1K41发布于 2020-06-09
腾讯云开发者
Copyright © 2013 - 2026 Tencent Cloud. All Rights Reserved. 腾讯云 版权所有
深圳市腾讯计算机系统有限公司 ICP备案/许可证号：粤B2-20090059 
粤公网安备44030502008569号
腾讯云计算（北京）有限责任公司 京ICP证150476号 | 京ICP备11018762号