3 ms·

I wish there was a light version of gpt3 with maybe like 100 api query limit per day so just amateurs like me got to play with this API too on weekends. I have

I wish there was a light version of gpt3 with maybe like 100 api query limit per day so just amateurs like me got to play with this API too on weekends. I have no agenda of changing the world and I have nothing substantial to fill in the gpt3 invite form but sometimes the best ideas come from tinkering and I just wish they gave some limited access to everyone.

ersiees 6y ago
You might like: https://transformer.huggingface.co https://transformer.huggingface.co The GPT-2 model is very similar to GPT-3 just smaller.
- superasn 6y ago
  Thanks for the link. So as someone who is completely new to this, if this guy used this model instead of gpt 3 would he get similar results?
  - sanxiyn 6y ago
    No, because the model is smaller.
  - dhairya 6y ago
    out of the box probably not. gpt-3 is pretrained on a much larger corpus and is a much higher capacity model, which is why its been performing very well in zero-shot learning examples like the one above. That being said you should try it. Note the huggingface library is a python library, so you'd technically be writing some code, though the library API is super user friendly and the community is very helpful.
  - calebkaiser 6y ago
    If you want to play with your own GPT-2 API, I put this together and shared it on HN yesterday: https://github.com/cortexlabs/cortex/tree/master/examples/pytorch/text-generator https://github.com/cortexlabs/cortex/tree/master/examples/py... You can run that and deploy your own GPT-2 API locally, or in the cloud. I'd also recommend checking out Max Woolf's gpt-2-simple, which is a library that makes it really easy to fine tune GPT-2 with your own text (if you've played the free version of AI Dungeon, this is how they trained their model): https://github.com/minimaxir/gpt-2-simple https://github.com/minimaxir/gpt-2-simple
dharma1 6y ago
I thought they have a free tier from October? https://twitter.com/hardmaru/status/1301362995356774401 https://twitter.com/hardmaru/status/1301362995356774401 Not sure what the 100k tokens translates to in terms of query limit
- dtech 6y ago
  That's 3 month max though
  - dharma1 6y ago
    would be nice if it was longer, but should have decent ideas in 3 months. And having another go with a throwaway email probably not that hard
- satorii 6y ago
  assume each prediction use 1000 tokens (including both input prompt and output text), then 100k tokens -> 100 predictions

ersiees 6y ago

You might like: https://transformer.huggingface.co https://transformer.huggingface.co The GPT-2 model is very similar to GPT-3 just smaller.

superasn 6y ago
Thanks for the link. So as someone who is completely new to this, if this guy used this model instead of gpt 3 would he get similar results?
- sanxiyn 6y ago
  No, because the model is smaller.
- dhairya 6y ago
  out of the box probably not. gpt-3 is pretrained on a much larger corpus and is a much higher capacity model, which is why its been performing very well in zero-shot learning examples like the one above. That being said you should try it. Note the huggingface library is a python library, so you'd technically be writing some code, though the library API is super user friendly and the community is very helpful.
- calebkaiser 6y ago
  If you want to play with your own GPT-2 API, I put this together and shared it on HN yesterday: https://github.com/cortexlabs/cortex/tree/master/examples/pytorch/text-generator https://github.com/cortexlabs/cortex/tree/master/examples/py... You can run that and deploy your own GPT-2 API locally, or in the cloud. I'd also recommend checking out Max Woolf's gpt-2-simple, which is a library that makes it really easy to fine tune GPT-2 with your own text (if you've played the free version of AI Dungeon, this is how they trained their model): https://github.com/minimaxir/gpt-2-simple https://github.com/minimaxir/gpt-2-simple

dharma1 6y ago

I thought they have a free tier from October? https://twitter.com/hardmaru/status/1301362995356774401 https://twitter.com/hardmaru/status/1301362995356774401 Not sure what the 100k tokens translates to in terms of query limit

dtech 6y ago
That's 3 month max though
- dharma1 6y ago
  would be nice if it was longer, but should have decent ideas in 3 months. And having another go with a throwaway email probably not that hard
satorii 6y ago
assume each prediction use 1000 tokens (including both input prompt and output text), then 100k tokens -> 100 predictions