Skip to content
GitLab
菜单
项目
群组
代码片段
帮助
帮助
支持
社区论坛
快捷键
?
提交反馈
登录/注册
切换导航
菜单
打开侧边栏
KMSCAKKSCFKA AKFACAMADCAS
tensorflow-federated
提交
1dcb490d
提交
1dcb490d
编辑于
12月 15, 2020
作者:
Krzysztof Ostrowski
提交者:
tensorflow-copybara
12月 15, 2020
浏览文件
Top-level introductory paragraph for multi-framework documentation in OSS.
PiperOrigin-RevId: 347653240
上级
91f0a85d
变更
3
Hide whitespace changes
Inline
Side-by-side
tensorflow_federated/experimental/BUILD
0 → 100644
浏览文件 @
1dcb490d
package
(
default_visibility
=
[
"//visibility:private"
])
licenses
([
"notice"
])
tensorflow_federated/experimental/README.md
0 → 100644
浏览文件 @
1dcb490d
This directory contains experimental code that is being incubated. Do not depend
on this code from other parts of TFF, as it can change at any time. All code
included here should retain private visibility while in the incubation stage.
tensorflow_federated/experimental/g3doc/multiframework.md
0 → 100644
浏览文件 @
1dcb490d
# Multi-Framework Support in TensorFlow Federated
TensorFlow Federated (TFF) has been designed to support a broad range of
federated computations, expressed through a combination of TFF's federated
operators that model distributed communication, and local processing logic.
Currently local processing logic can be expressed using TensorFlow APIs (via
`@tff.tf_computation`
) at the frontend, and is executed via the TensorFlow
runtime at the backend. However, we aim to support multiple other
(non-TensorFlow) frontend and backend frameworks for local computations,
including non-ML frameworks (e.g., for logic expressed in SQL or general-purpose
programming languages).
In this section, we'll include information on:
*
Mechanisms that TFF provides to support alternative frameworks, and how you
can add support for your preferred type of frontend or backend to TFF.
*
Experimental implementations of support for non-TensorFlow frameworks, with
examples.
*
Tentative future roadmap for graduating these capabilities beyond the
experimental status.
编辑
预览
Supports
Markdown
0%
请重试
或
添加新附件
.
添加附件
取消
You are about to add
0
people
to the discussion. Proceed with caution.
先完成此消息的编辑!
取消
想要评论请
注册
或
登录