Quickstart
You can use the CLI or the SDK to work with our platform. A few concepts:
Models
- We support seven base models. You can run inference on these base models directly.
- We'll be supporting Code Llama and Starcoder soon
LLM Fine-Tuning
- If you want to fine-tune a model, you need to first create your own model instance, based off a base model you select. Check the CLI and SDK quickstarts for instructions.
- You can further fine-tune a model instance that you previously fine-tuned.
- If you have a large fine-tuning job, use the UI to upload your fine-tuning data.
- Fine-tuning limits via our public API are listed below. If you need more, email us at [email protected] for a custom quote.
Language Model | Fine-tuning Context Length |
---|---|
Bloom-560 | 500 |
Llama-2 | 2048 |
Nous-Hermes-Llama-2 | 4096 |
LLM Inference
- You can run inference on base models directly.
- You can also run inference on your fine-tuned model adapters. We only charge by the token—so you can run fine-tuned models on demand.
Updated 20 days ago