sharonzhou

6 ms·

1.
by sharonzhou 3y ago
ROCm is great. We were able to get run and finetune LLMs on AMD Instincts with parity to NVIDIA A100s - and built an SDK that’s as easy to use as HuggingFace or easier (Lamini). Or at the very least, our designer is able to finetune/tr
2.
by sharonzhou 3y ago
Can do training here (for free on small LLMs). With a simple interface. https://colab.research.google.com/drive/1QMeGzR9FnhNJJFmcHtm... https://lamini-ai.github.io/ from llama import QuestionAnswerMode
3.
Free and Fast LLM Finetuning(github.com)
7 points by sharonzhou 3y ago|4 comments
4.
by sharonzhou 3y ago
Official repo for the finetuning pipeline, so you can train custom models on your data. It's free, on small LLMs It's fast, taking 10-15 minutes It's like working with an unlimited prompt size, with 1000x+ more space than the
5.
Launch Lamini: The LLM Engine for Rapidly Customizing Models as Good as ChatGPT(lamini.ai)
123 points by sharonzhou 3y ago|70 comments
6.
by sharonzhou 3y ago
Hi HN! I’m super excited to announce Lamini, the LLM engine that gives every developer the superpowers that took the world from GPT-3 to ChatGPT! I’ve seen a lot of developers get stuck after prompt-tuning for a couple days or after fine-tu
7.
by sharonzhou 8y ago
I was a couple months ago, then I started my own thing and I feel great again. It’s called Sidepact and it’s actually really relevant to this problem! It’s a weekend program for employed engineers to start companies—to get people out of the
8.
by sharonzhou 8y ago
I think it's less about complexity and more about precision. A simple law can be extremely difficult to encode if it lacks specification.
9.
by sharonzhou 8y ago
I'm surprised there was no octopus to represent the bandits. In all seriousness, this post makes sense to me, as someone who does RL research. However, the intuition behind the concepts could be communicated more clearly. I would reaso
10.
by sharonzhou 8y ago
Sounds like she knows her priorities and adamantly avoids wasting money and chasing things for reputation's sake. I aspire to be like that.
11.
by sharonzhou 8y ago
This looks really cool and relevant. Following.