Our great sponsors
-
gpt-neox
An implementation of model parallel autoregressive transformers on GPUs, based on the DeepSpeed library.
-
WorkOS
The modern identity platform for B2B SaaS. The APIs are flexible and easy-to-use, supporting authentication, user identity, and complex enterprise features like SSO and SCIM provisioning.
In terms of not having to worry about usage costs, I don't think you can get around this with GPT3. You could try working with GPT-NEOX, which is the currently largest available self hostable model (20B params, and they're planning on eventually matching GPT3's 175B params), and set up a remote machine via some cloud provider that you turn on/off with some command in your playground so it's only burning money when in use, though this may be a huge technical challenge, and may end up being very expensive as it requires multiple GPUs or a single very powerful GPU.