Follow along on Twitter for detailed updates and improvements made to Banana.
Playground and Templates
We're excited to introduce three new models to our playground this week:
- segmentation: This model can segment audio, a feature that is particularly useful for projects requiring differentiation between speakers in a conversation. Imagine inputting a phone call between you and a friend, and the model being able to detect different speakers on the call. Explore this model here.
- stable-diffusion-xl-base-1.0: With this model, you can generate high-quality images. Try it out here.
- CodeLlama-7B-Instruct-GPTQ: Following last week's announcement, the LLM community has been exploring Code Llama models. We're thrilled to bring one of them to our playground!
If you want any of the other Code Llama models deployed as an endpoint on Banana, they are all available as templates. Find them in this post 🦙.
We've made significant updates to our documentation to address common issues our users face:
- 1MB JSON payload limit
- Inference calls taking longer than 5 minutes
- Adding multiple models to your Potassium app
We've got you covered. We've extensively updated our documentation to address these issues, with a particular focus on how to handle Potassium. Check out the updated guide here 📄.