关于“TPU Speed Data Pipelines”的评价

Instructions again make absolutely no sense - "iterate the files" is not an instruction that is complete is any way. Even the question numbers are wrong just listing question 1 repeatedly.

Oliver H. · 评论over 1 year之前

aishwarya b. · 评论over 1 year之前

Joost B. · 评论over 1 year之前

Thorvald Sahl J. · 评论over 1 year之前

It's kind of messy. We're not using accelerators (GPUs) and we're not using parallelism at all. I guess it's nice to learn the data loading.

Loading A. · 评论over 1 year之前

Aditya S. · 评论over 1 year之前

Brody I. · 评论over 1 year之前

Decent but there's a lot missing. TFRecord is introduced and then there is no explanation of tf.io.*, so you'll end up searching through a bunch of docs until maybe you come across parse_single_example or until you look at the solution.

chris w. · 评论over 1 year之前

LUCAS S. · 评论over 1 year之前

Sudharshan N. · 评论over 1 year之前

Alexey T. · 评论over 1 year之前

Suchita B. · 评论over 1 year之前

DANIEL M. · 评论over 1 year之前

Jayesh Arun B. · 评论over 1 year之前

ROBERTO O. · 评论over 1 year之前

Rahul M. · 评论over 1 year之前

David M. · 评论over 1 year之前

Harold H. · 评论over 1 year之前

2. To use TFRecord format to load training data efficiently from GCS it is more efficient is it because that it doesn't show the pictures, but it shows 000100 010000 01010?

Wanda W. · 评论over 1 year之前

Ming H. · 评论over 1 year之前

LUIS F. · 评论over 1 year之前

Kidus B. · 评论over 1 year之前

Menegozzo G. · 评论over 1 year之前

Marián Ž. · 评论over 1 year之前

Aditya P. · 评论over 1 year之前