Here you can several examples of how to use SimpleAI to expose models and integrate them. There is also a list of external projects using SimpleAI below.
Model | Description | Number of parameters | Chat | Chat (streaming) | Instruct | Completion | Completion (streaming) | Embedding |
---|---|---|---|---|---|---|---|---|
MPT-7B-Chat | MosaicML's MPT model fine-tuned for chat | 7B | ✔️ | ✔️ | ❌ | ❌ | ❌ | ❌ |
MPT-7B-Storywriter-65k+ | MosaicML's MPT model fine-tuned with a context length of 65k tokens on a filtered fiction subset of the books3 dataset | 7B | ❌ | ❌ | ❌ | ✔️ | ✔️ | ❌ |
StableLM - Open-Assistant | StabilityAI's StableLM model fine-tuned for chat by Open-Assistant | 7B | ✔️ | ✔️ | ❌ | ❌ | ❌ | ❌ |
Alpaca | Instruct model using LoRA to reproduce the Stanford Alpaca model | 7B | ❌ | ❌ | ✔️ | ❌ | ❌ | ❌ |
Sentence-Transformers | A basic example on how to use sentence-transformers to create embeddings from text |
N/A | ❌ | ❌ | ❌ | ❌ | ❌ | ✔️ |
Find it here.
This project uses the RWKV-LM model and turns it into an gRPC service that can be used through SimpleAI.
RWKV is an RNN with Transformer-level language model performance that can be trained like a GPT transformer and is 100% attention-free. It combines the best of RNN and transformer, providing great performance, fast inference, saves VRAM, fast training, "infinite" ctx_len, and free sentence embedding.
- Author : @Nintorac
- Implements for the same model and in a single gRPC server:
chat
(including streaming),completions
,- and
embeddings
.
Feel free to contribute with your own examples, either by submitting a PR or getting in touch to get your project referenced here.