PyPi中的tf-nightly和tensorflow有什么区别?

时间:2019-12-27 11:42:42

标签: python tensorflow pip tensorflow2.0 pypi

PyPi中的Location.watchPositionAsyncimport React, { Component } from "react"; import { StyleSheet, View } from "react-native"; import MapView from "react-native-maps"; import * as Location from "expo-location"; import * as Permissions from "expo-permissions"; import * as TaskManager from "expo-task-manager"; const LOCATION_TASK_NAME = "background-location-task"; export default class App extends Component { constructor(props) { super(props); this.state = { region: null, error: '', }; } _getLocationAsync = async () => { console.log(`entered _getLocationAsync`) await Location.startLocationUpdatesAsync(LOCATION_TASK_NAME, { enableHighAccuracy: true, distanceInterval: 1, timeInterval: 5000 }); console.log(`start watchPositionAsync`) // watchPositionAsync Return Lat & Long on Position Change this.location = await Location.watchPositionAsync( { enableHighAccuracy: true, distanceInterval: 1, timeInterval: 10000 }, newLocation => { let { coords } = newLocation; // console.log(coords); let region = { latitude: coords.latitude, longitude: coords.longitude, latitudeDelta: 0.045, longitudeDelta: 0.045 }; console.log(`watchPositionAsync ${coords.latitude}, ${coords.longitude}`) this.setState({ region: region }); }, error => console.log(error) ); return this.location; }; async componentWillMount() { // Asking for device location permission const { status } = await Permissions.askAsync(Permissions.LOCATION); if (status === "granted") { this._getLocationAsync(); } else { this.setState({ error: "Locations services needed" }); } } render() { return ( <View style={styles.container}> <MapView initialRegion={this.state.region} showsCompass={true} showsUserLocation={true} rotateEnabled={true} ref={map => { this.map = map; }} style={{ flex: 1 }} /> </View> ); } } TaskManager.defineTask(LOCATION_TASK_NAME, async ({ data, error }) => { if (error) { console.log(error); return; } if (data) { const { locations } = data; let lat = locations[0].coords.latitude; let long = locations[0].coords.longitude; console.log(`TaskManager:- ${lat}, ${long}`) } }); const styles = StyleSheet.create({ container: { flex: 1, backgroundColor: "#fff" } }); 有什么区别?

tf-nightly

哪个可靠?

https://pypi.org/project/tf-nightly/

https://pypi.org/project/tensorflow/

2 个答案:

答案 0 :(得分:1)

Nightly用于获取较早的最新tensorflow开发思路,该版本每天更新。与浏览器(see here)相同。

  

哪个可靠?

经典的tensorflow pip install tensorflow是两者中最可靠的。此版本在发布之前已经过很多人的测试。

答案 1 :(得分:1)

只需添加到BSO编写的内容中即可:

  1. 顾名思义,tf-nightly pip软件包每晚都会生成并发布到PyPI(除非有任何生成失败的情况,这种情况很少发生)。结果,您可以看到一个almost once-per-day version update history。它具有接近github.com/tensorflow master分支的HEAD的最新功能。因此,如果您需要最新的功能,改进和错误修复,例如在上一个稳定的tensorflow版本之后提交的功能(请参见下文),则应使用pip install tf-nightly。但是不利的是,由于tf-nightly版本不受tensorflow相同的严格版本测试的限制,因此偶尔会包含一些错误,这些错误将在以后修复。另外,由于它是从HEAD构建的,因此将反映中间开发状态,例如功能不完整。

  2. tensorflow pip包由基于语义版本的计划发布。新版本大约2-6个月推出一次。由于具有全面的发布测试作业集,因此质量要高于tf-nightlytensorflow pip软件包中每个次要版本变更都会对https://www.tensorflow.org/api_docs/python/上的文档进行一次更新。