tf.to_float(tf.convert_to_tensor(python_object))在Tensorflow对象检测api中多次使用grid_anchor_generator。我会使用tf.constant(python_object, dtype=tf.float32)。我想知道它们之间的区别。感谢
答案 0 :(得分：6)
对于tf.convert_to_tensor，输入值必须是静态非张量类型。例如一个numpy数组。
对于getNews(stocks: string, daysBack: number) {
this.newsItems = [];
const securityArray = stocks.split(',');
this.recordsProcessed = 0;
this.recordCount = securityArray.length;
this.securityService.progressStart.next(0);
this.securityService.progressFinish.next(this.recordCount);
this.showRecs = false;
for (let i = 0; i < securityArray.length; i++) {
const stk = securityArray[i];
this.securityService.getNews(securityArray[i]).takeUntil(this.ngUnsubscribe).subscribe(n => {
try {
for (let x = 0; x < n.rss.channel.item.length; x++) {
const iDate = new Date(new Date().setDate(new Date().getDate() - daysBack));
iDate.setHours(0, 0, 0, 0);
const pDate = new Date(n.rss.channel.item[x].pubDate);
if (pDate >= iDate) {
const newsItem = new NewsItem(
stk,
decodeURI(n.rss.channel.item[x].description),
n.rss.channel.item[x].guid,
n.rss.channel.item[x].link,
decodeURI(n.rss.channel.item[x].title),
pDate
);
this.newsItems.push(newsItem);
}
}
this.recordsProcessed++;
this.securityService.progressStart.next(this.recordsProcessed);
if (this.recordCount === this.recordsProcessed) {
this.showRecs = true;
}
} catch (e) {
console.log(e);
this.recordsProcessed++;
this.securityService.progressStart.next(this.recordsProcessed);
if (this.recordCount === this.recordsProcessed) {
this.showRecs = true;
}
}
}, error => {
console.log('Error', error);
this.recordsProcessed++;
this.securityService.progressStart.next(this.recordsProcessed);
});
}
}
，值＆＃34;一个对象，其类型具有注册的Tensor转换函数。＆＃34;这意味着输入类型如Tensors或tf.Variables也可以作为输入提供。例如，请在此处查看变量的张量转换函数：https://github.com/tensorflow/tensorflow/blob/r1.8/tensorflow/python/ops/variables.py#L762