--- license: apache-2.0 language: - en pipeline_tag: text-generation datasets: - appvoid/no-prompt-15k --- ![palmer](https://huggingface.co/appvoid/palmer-003-turbo/resolve/main/image.png) # palmer ### a better base model palmer is a series of ~1b parameters language models fine-tuned to be used as base models instead of using custom prompts for tasks. This means that it can be further fine-tuned on more data with custom prompts as usual or be used for downstream tasks as any base model you can get. The model has the best of both worlds: some "bias" to act as an assistant, but also the abillity to predict the next-word from its internet knowledge base. It's a 600m llama 2 model (since is 4bit quantized) so you can use it with your favorite tools/frameworks. ### evaluation 🧪 note that this is a zero-shot setting as opposite to open llm leaderboard's few-shot evals ``` Model ARC_C HellaSwag PIQA Winogrande Average palmer-001 | 0.2807 | 0.5524 | 0.7106 | 0.5896 | 0.5333 | palmer-003-turbo | 0.3106 | 0.5806 | 0.7247 | 0.5951 | 0.5527 | palmer-002 | 0.3242 | 0.5956 | 0.7345 | 0.5888 | 0.5607 | ``` This model is as good as tinyllama base while being half the size. ### training 🦾 Training took 1.5 rtx 2060 gpu hours. It was trained on 15,000 gpt-4 shuffled samples. palmer was fine-tuned using lower learning rates ensuring it keeps as much general knowledge as possible. ### prompt 📝 ``` no prompt 🚀 ``` ### Note As of today 1/4/2024 is still not possible to convert to gguf, [see more here](https://github.com/ggerganov/llama.cpp/issues/4199#issuecomment-1825833475).