For the longest time I haven't been able to find a tokenizer online similar to the one for GPT-3 for other models. So I made one and put it up online for free for anyone.
It's built on top of the tiktoken library and is basically just a lambda function in the backend.
You could just download the library directly but I hope this saves someone time if they need the token ids for logit bias.
Decided to write this after doing a bunch of research into recording a webRTC meeting with several participants. Hopefully it helps other who decide to go down this path if they want to incorporate recording into their video chat app. I used an ingenious open source library called puppeteer-stream to basically get a prototype ready in a couple of minutes.
Paul Graham and Balaji really don't like the New York Times. Can't blame them really. The NYT lets you subscribe online, but you can only unsubscribe by calling them or talking to them through chat.
I made a repo where users can talk to someone at support to cancel their NYT subscription. It uses twilio to place the call, and streams the audio through websockets to google's speech to text api. When it hears that you are next in line, it forwards you the call.
I made a random startup idea generator. The idea was to have ideas that are plausible but terrible. It's solely for the purpose of humor. Hope one of these makes you laugh
It's built on top of the tiktoken library and is basically just a lambda function in the backend.
You could just download the library directly but I hope this saves someone time if they need the token ids for logit bias.