Other
Transformers
Safetensors
mint_stability
feature-extraction
immunology
affinity
pMHC
epitope
neoantigen
personalized_immunotherapy
custom_code
Instructions to use dkarthikeyan1/mint-stage1-affinity with libraries, inference providers, notebooks, and local apps. Follow these links to get started.
- Libraries
- Transformers
How to use dkarthikeyan1/mint-stage1-affinity with Transformers:
# Load model directly from transformers import AutoModel model = AutoModel.from_pretrained("dkarthikeyan1/mint-stage1-affinity", trust_remote_code=True, device_map="auto") - Notebooks
- Google Colab
- Kaggle
| """MINT Stability model configuration (S1/S2).""" | |
| from transformers import PretrainedConfig | |
| __all__ = ["MintStabilityConfig"] | |
| class MintStabilityConfig(PretrainedConfig): | |
| """Configuration for MINT pMHC-I stability prediction model. | |
| This model uses a custom ESM2-650M backbone (with optional cross-chain | |
| multimer attention from MINT pretraining) and a projection head for | |
| regression on complex half-life. | |
| """ | |
| model_type = "mint_stability" | |
| def __init__( | |
| self, | |
| # ESM2 backbone | |
| num_layers: int = 33, | |
| embed_dim: int = 1280, | |
| attention_heads: int = 20, | |
| token_dropout: bool = True, | |
| use_multimer: bool = True, | |
| # Projection head | |
| hidden_dim: int = 512, | |
| dropout: float = 0.2, | |
| output_size: int = 1, | |
| sigmoid_output: bool = False, | |
| **kwargs, | |
| ): | |
| self.num_layers = num_layers | |
| self.embed_dim = embed_dim | |
| self.attention_heads = attention_heads | |
| self.token_dropout = token_dropout | |
| self.use_multimer = use_multimer | |
| self.hidden_dim = hidden_dim | |
| self.dropout = dropout | |
| self.output_size = output_size | |
| self.sigmoid_output = sigmoid_output | |
| super().__init__(**kwargs) |