Instructions to use Coralfil/Atlantis-Pyramid-v2 with libraries, inference providers, notebooks, and local apps. Follow these links to get started.
- Libraries
- PEFT
How to use Coralfil/Atlantis-Pyramid-v2 with PEFT:
from peft import PeftModel from transformers import AutoModelForCausalLM base_model = AutoModelForCausalLM.from_pretrained("Qwen/Qwen2.5-32B-Instruct") model = PeftModel.from_pretrained(base_model, "Coralfil/Atlantis-Pyramid-v2") - Notebooks
- Google Colab
- Kaggle
Official Release: Atlantis-Pyramid v2 (32.5B Specialized Foundation Model)
We are excited to announce the official open-weights release of Atlantis-Pyramid v2 (32.5B), Coralfil's flagship specialized foundation model for physical oceanography, benthic restoration, and Canadian maritime statutory compliance.\n\nKey Highlights:\n- 32.5B dense parameters combining domain-adapted Qwen2.5 and reasoning-distilled DeepSeek-R1.\n- 97.1% accuracy on the AboveBoard 52 benchmark and 100% Canadian marine statutory compliance.\n- Deep Dive Hexactine Graph Memory (degree = 6 orthogonal lattice).\n- 18.4 GB serving VRAM footprint (fits single RTX 3090/4090 or datacenter A10/A100).\n\nDocumentation and interactive playground: https://atlantis-llm.io\nPacific Ocean Monitor: https://coralfil.com/monitor