Dear folks,
I enjoyed these article & youtube very much.
If you have any examples of how netCDF and chatgpt
have improved your coding efficiency,
can you give us some specific examples?
Best,
Yuichiro
----- Begin forwarded message -----
Subject: [HDF Forum] [News and Announcements from The HDF Group] Gerd Heber on
Call the Doctor - Tuesday, July 25, 2023
Date: Wed, 02 Aug 2023 21:12:56 +0000
From: Lori Cooper via HDF Forum <noreply@xxxxxxxxxxxxxxxxxx>
Here's the recording from Gerd's session: Enhance your coding productivity and
efficiency with ChatGPT
https://youtu.be/ckYkV_uBGqw
Because why not, here's the session description generated by ChatGPT:
In the July 25th episode of "Call the Doctor" hosted by Gerd Heber, the focus
was on using AI tools, specifically Chat GPT, to perform practical tasks
related to HDF5 and image processing.
The episode started with a brief introduction, and then Gerd mentioned that he
received a directory filled with TIFF image files and wanted to explore ways to
combine them into a single HDF5 file. He emphasized that the goal was to let
Chat GPT write the code for this task, avoiding manual coding.
Gerd interacted with Chat GPT, presenting different prompts related to the
task. Initially, he asked for a Python program to create an HDF5 file, and Chat
GPT provided the code using the Python Imaging Library (PIL) module.
Next, Gerd requested the code to compress the images while writing them into
the HDF5 file. Chat GPT complied and provided code for compressing the images
using GZIP compression.
The third prompt asked for the images to be combined into a 3D dataset instead
of individual 2D datasets. Chat GPT produced code to create a 3D dataset and
efficiently stack the images into it.
In the subsequent prompt (version 2.1), Gerd suggested that Chat GPT should
infer the dimensions of the images from the provided TIFF files instead of
explicitly specifying them. Chat GPT accurately determined the image dimensions
and generated the code accordingly.
Finally, Gerd introduced version 2.2 and 2.2.1. He explained that when using
version 2.2, HDF5's default chunking strategy resulted in poor performance when
dealing with compressed 3D datasets. To fix this, he asked Chat GPT to modify
the code to manually set the chunk size for each image to achieve better
performance. Chat GPT provided the code accordingly.
Throughout the episode, Gerd showcased how Chat GPT effectively wrote Python
code to handle various aspects of the task, making the process efficient and
practical. He also mentioned that the integration of GitHub Copilot with Visual
Studio Code had been helpful in assisting with the code generation process.
In conclusion, the episode demonstrated the power of AI tools like Chat GPT and
showcased how they could assist in practical tasks related to HDF5 and image
processing, making the development process more efficient and less manual.
--
HAGIHARA Yuichiro // yu.hagihara@xxxxxxxxx
Remote Sensing Laboratory,
National Institute of Information and Communications Tech.(NICT)
