Teach LLMs how to follow schemas like JSON.
LLMs often make stupid (and sometimes naive) JSON mistakes, such as mistaking JSON for Javascript (doesn't require double quotes for key names), missing characters like brackets, or bad escape strings.
Below is a simple prompt that attempts to lower the probability of those happening, and I think it's pretty cool.
## Tips and tricksFernando Miguel
curl "https://static.wikia.nocookie.net/vsbattles/images/b/b0/Fernando_Miguel.png/revision/latest?cb=20240216175342" --output fernando.pngScreen
replace {{ url }} to your discussion url
| from typing import List, Tuple | |
| import requests | |
| from datasets import Dataset | |
| from selectolax.lexbor import LexborHTMLParser | |
| # News pages search total iterations | |
| N_NEWS_ITERS = 100 | |
| user_agent = ( |
If you're watching the video tutorial Let's build the GPT Tokenizer by Andrej Karpathy and you want to copy the whole article, just clone this gist.
Clone this:
In [25]: !git clone https://gist.github.com/f48f35dd9456af1f278a349501aafab9.git
Out[25]: # Cloning into 'f48f35dd9456af1f278a349501aafab9'...
# remote: Enumerating objects: 4, done.| from rapiz.app import API, APIContext | |
| import time | |
| api = API() | |
| @api('/api/user', methods=['GET', 'POST']) | |
| def user(res: APIContext): | |
| return { | |
| "name": "John Cena", |
| from linelib import Client # pip install linelib | |
| client = Client('channel secret', 'channel access token') | |
| @client.event('text') | |
| async def text_msg(ctx): | |
| await ctx.send('Hello, World!') | |
| client.run() |