Warm tip: This article is reproduced from serverfault.com, please click
keras python tensorflow reshape

python-临时合并Keras中的批次尺寸

(python - Temporarily merge the batch dimension in Keras)

发布于 2020-11-28 13:43:05

我有一个Keras模型，输入形状为[None, 500, 500, 3]，输出形状为[None, 1]。现在，我想制作一个包装模型，其输入形状为[None, 48, 500, 500, 3]，输出形状为[None, 48]。

为此，幼稚的方法是在第二个轴上迭代48次并应用第一个模型，然后使用KerasConcatenate层获得所需的形状。

model_outputs = []
for i in range(inputs.shape[1]):
    im_block = inputs[:, i]
    model_outputs += [self.model(im_block)]
return Concatenate()(model_outputs)

但是，这使图形变得相当复杂。因此，我想改为执行以下操作：

        [None, 48, 500, 500, 3]
     -> [None*48,  500, 500, 3]
           (apply the model)
     -> [None*48,  1]
     -> [None, 48, 1]

我的尝试是

outputs = tf.reshape(inputs, (inputs[0] * inputs[1], *inputs[2:]))
outputs = self.model(outputs)
outputs = tf.reshape(outputs, (inputs[0], inputs[1]))
return outputs

但这给了我

TypeError: Cannot iterate over a tensor with unknown first dimension.

有没有办法做到这一点 ？

Questioner
Samuel Prevost
Viewed
11
分享
Andrey 2020-11-29 01:29:39

这应该工作：

inp = tf.reshape(inp, (-1, 500, 500, 3))
res = model(inp)
res = tf.reshape(res, (-1, 48))
Samuel Prevost 2020-11-28 17:56:23

您是对的，我只是忘记了，.shape并且使用的是张量的值，该值不存在。

热门github

1
Implementation of paper - YOLOv9: Learning What You Want to Learn Using Programmable Gradient Information (翻译：论文实现——YOLOv9: Learning What You Want to Learn using Programmable Gradient Information（利用可编程梯度信息学习您想学的内容）)
2
A Windows and Office activator using HWID / Ohook / KMS38 / Online KMS activation methods, with a focus on open-source code and fewer antivirus detections. (翻译：使用 HWID / KMS38 / Online KMS 激活方法激活 Microsoft 产品的脚本集合，重点是开源代码、较少的防病毒检测和用户友好性。)
3
Get up and running with Llama 2, Mistral, Gemma, and other large language models. (翻译：启动并运行 Llama 2、Mistral、Gemma 和其他大型语言模型。)
4
该项目可以让你通过订阅的方式使用Cloudflare WARP+，自动获取流量。This project enables you to use Cloudflare WARP+ through subscription, automatically acquiring traffic.
5
Multi functional app to find duplicates, empty folders, similar images etc. (翻译：多功能应用程序可查找重复项、空文件夹、相似图像等。)
6
Xray panel supporting multi-protocol multi-user expire day & traffic & ip limit (Vmess & Vless & Trojan & ShadowSocks & Wireguard) (翻译：Xray面板支持多协议多用户到期日&流量&IP限制（Vmess & Vless & Trojan & ShadowSocks & Wireguard)
7
The Free Software Media System (翻译：Jellyfin 是一个免费软件媒体系统，可让您控制管理和流式传输您的媒体。它是专有 Emby 和 Plex 的替代方案，通过多个应用程序从专用服务器向最终用户设备提供媒体。)
8
lightweight, standalone C++ inference engine for Google's Gemma models. (翻译：适用于 Google Gemma 模型的轻量级独立 C++ 推理引擎。)
9
📚 Freely available programming books (翻译：📚 免费提供的编程书籍)
10
A collective list of free APIs (翻译：免费 API 的集合列表)
11
1️⃣🐝🏎️ The One Billion Row Challenge -- A fun exploration of how quickly 1B rows from a text file can be aggregated with Java (翻译：十亿行挑战 —— 使用 Java 对文本文件中的 10 亿行数据进行聚合的有趣探索)
12
🎓 Path to a free self-taught education in Computer Science! (翻译：🎓计算机科学免费自学教程！)
13
Curso para aprender el lenguaje de programación Python desde cero y para principiantes. 75 clases, 37 horas en vídeo, código, proyectos y grupo de chat. Fundamentos, frontend, backend, testing, IA... (翻译：从零开始学习 Python 编程语言的课程，适合初学者)
14
This repository contains System Design resources which are useful while preparing for interviews and learning Distributed Systems (翻译：该存储库包含系统设计资源，在准备面试和学习分布式系统时非常有用)
15
Mamba is a new state space model architecture showing promising performance on information-dense data such as language modeling, where previous subquadratic models fall short of Transformers. It is based on the line of progress on structured state space models, with an efficient hardware-aware design and implementation in the spirit of FlashAttention. (翻译：Mamba 是一种新的状态空间模型架构，在信息密集型数据（例如语言建模）上显示出良好的性能，而之前的二次模型在 Transformers 方面存在不足。它基于结构化状态空间模型的进展，并本着FlashAttention的精神进行高效的硬件感知设计和实现。)