facebook/esm2_t33_650M_UR50D
🧠 AI Modelfacebook
650M parameter protein language model for masked sequence modeling.
ESM-2 is a family of protein language models based on the transformer architecture. This model (esm2_t33_650M_UR50D) is the 650 million parameter version with 33 layers, trained on the UniRef50 database (clustered at 50% sequence identity). It uses a masked language modeling objective to learn protein sequence representations. The model can be used for tasks like structure prediction (e.g., ESMFold), function annotation, and embedding extraction for downstream tasks. It supports both PyTorch and TensorFlow, is compatible with HuggingFace endpoints, and is released under the MIT license. The model has accumulated 2,668,729 downloads and 81 likes on HuggingFace, indicating strong community adoption.
💡Highlights
- ├─33 layers, 650M params
- ├─Trained on UniRef50
- └─Masked language modeling
🎯For
- ├─Bioinformaticians
- ├─Protein engineers
- └─AI researchers