Instructions to use open-athena/marinfold-exp75 with libraries, inference providers, notebooks, and local apps. Follow these links to get started.
- Libraries
- Transformers
How to use open-athena/marinfold-exp75 with Transformers:
# Load model directly from transformers import AutoModel model = AutoModel.from_pretrained("open-athena/marinfold-exp75", device_map="auto") - Notebooks
- Google Colab
- Kaggle
metadata
library_name: transformers
tags:
- biology
- protein-language-model
- marinfold
MarinFold exp75 checkpoints
Final checkpoints from the MarinFold contacts-v1 1.5B tuning sweep tracked in Open-Athena/MarinFold#75.
Both checkpoint formats are retained:
hf/contains aQwen3ForCausalLMHugging Face safetensors export with the contacts-v1 tokenizer colocated with the weights.checkpoints/contains the original Levanter OCDBT checkpoint for training restart or re-export.
Checkpoint inventory
| Run ID | Final step | Format | Repository path | GCS source |
|---|---|---|---|---|
prot-exp75-cv1-1_5b-e8-lr1e-3-wd0p2-v1-bc3084 |
35679 | Hugging Face | hf/step-35679 |
gs://marin-us-east5/checkpoints/prot-exp75-cv1-1_5b-e8-lr1e-3-wd0p2-v1-bc3084/hf/step-35679/ |
prot-exp75-cv1-1_5b-e8-lr1e-3-wd0p2-v1-bc3084 |
35679 | Levanter | checkpoints/step-35679 |
gs://marin-us-east5/checkpoints/prot-exp75-cv1-1_5b-e8-lr1e-3-wd0p2-v1-bc3084/checkpoints/step-35679/ |
W&B recorded this run as finished at step 35,679. The GCS run contains exactly one permanent Levanter checkpoint, also at step 35,679.