aqader

3 ms·

1.
by aqader 3y ago
this is really cool, can’t wait to try it out for some ML pipeline development. kudos myles and akshay!
2.
by aqader 4y ago
Yes, we did! The dataset has since been cleaned even more so we're due to update the model.
3.
by aqader 4y ago
There are a couple open source implementations. I'll list a couple below: 7B: - https://huggingface.co/tloen/alpaca-lora-7b - https://huggingface.co/ozcur/alpaca-native-4bit 13B: - https:&#
4.
by aqader 4y ago
Depends on the model size. A model like GPT3 that has hundreds of billions of paramaters, you can do few-shot learning with. You'll still pay for the tokens processed and it'll at least linearly increase response times the larger
5.
by aqader 4y ago
Almost. If your dataset contains questions and answers about your own projects documentation, then yes. The UX around how to prompt a fine-tuned model depends on the format of the dataset it's trained on. One way you can do this is pas
6.
by aqader 4y ago
Yeah, this is running in 8bit mode. The 30b 8bit version we released seems to do a lot better but it requires significantly more compute. https://huggingface.co/baseten/alpaca-30b
7.
by aqader 4y ago
For this demo, we're using the 8bit version here: https://huggingface.co/tloen/alpaca-lora-7b We also fine-tuned and OSS'd a 30b version here that you can checkout (on the cleaned 52k Alpaca dataset) https:&
8.
Show HN: Fine-tune generative models in 1 line of code(blueprint.baseten.co)
16 points by aqader 4y ago|0 comments
9.
Show HN: I fine-tuned Flan-T5. Can it cook?(abuqader.substack.com)
55 points by aqader 4y ago|43 comments
10.
Show HN: Generate recipes with a fine-tuned language model(lechef.fyi)
7 points by aqader 4y ago|0 comments
11.
On Browser Tabs(abuqader.substack.com)
121 points by aqader 6y ago|186 comments