This a ~90m assistant model for cameloid models like LLama/Alpaca/Vicuna/Guanaco that use the llama tokenizer, allowing for speedups up to 3x with greed sampling. Its trained on 5.5 billion tokens of refinedweb and uses the GPTBigcode architecture and has a context window: 1024. To use please see this article on assisted generation https://huggingface.co/blog/assisted-generation.
- Downloads last month
- 4
This model does not have enough activity to be deployed to Inference API (serverless) yet. Increase its social
visibility and check back later, or deploy to Inference Endpoints (dedicated)
instead.