# Falcon ## Overview Falcon is a state-of-the-art language model trained on the [RefinedWeb dataset](https://arxiv.org/abs/2306.01116). At the time of writing, it is the leading model on the [OpenLLM leaderboard](https://huggingface.co/spaces/HuggingFaceH4/open_llm_leaderboard). There is no paper associated with Falcon yet, but for citation information please see [the repository for Falcon-40B](https://huggingface.co/tiiuae/falcon-40b#citation), the highest-performance Falcon model. - The model and tokenizer can be loaded via: ```python from transformers import AutoModelForCausalLM, AutoTokenizer tokenizer = AutoTokenizer.from_pretrained("tiiuae/falcon-40b-instruct") model = AutoModelForCausalLM.from_pretrained("tiiuae/falcon-40b-instruct") inputs = tokenizer("What's the best way to divide a pizza between three people?", return_tensors="pt") outputs = model.generate(**inputs, max_length=50) ``` - The Falcon tokenizer is a BPE model. ## FalconConfig [[autodoc]] FalconConfig ## FalconModel [[autodoc]] FalconModel - forward ## FalconForCausalLM [[autodoc]] FalconForCausalLM - forward ## FalconForSequenceClassification [[autodoc]] FalconForSequenceClassification - forward ## FalconForTokenClassification [[autodoc]] FalconForTokenClassification - forward ## FalconForQuestionAnswering [[autodoc]] FalconForQuestionAnswering - forward