| license: apache-2.0 | |
| datasets: | |
| - HuggingFaceH4/ultrachat_200k | |
| language: | |
| - en | |
| tags: | |
| - tiny | |
| - llm | |
| - lm | |
| - language | |
| - mobile | |
| - phone | |
| - '48' | |
| - m | |
| - million | |
| - parameter | |
| - instruct | |
| - chat | |
| - chatbot | |
| - bot | |
| pipeline_tag: text-generation | |
| # Typewriter 48M Instruct! | |
| Trained on an RTX 3090, purely on chat data, as a test to see how small you could make LLMs before they can't chat anymore. | |
| Not very useful, but good if you need some cheap chat embeddings. |