Package io.github.kirstenali.deepj.training

Interface Trainable

All Known Subinterfaces:
CausalLM, Layer, NormLayer, Projection
All Known Implementing Classes:
DecoderOnlyModel, DeepSeekModel, DeepSeekTransformerBlock, Embedding, FNN, GPTModel, GPTTransformerBlock, LayerNorm1D, Linear, LlamaModel, LlamaTransformerBlock, MultiHeadLatentAttention, MultiHeadSelfAttention, PositionalEmbedding, RMSNorm1D, RoPEMultiHeadSelfAttention, SwiGLULayer, TransformerStack
public interface Trainable

  • Method Summary

    Modifier and Type
    Method
    Description
    List<Parameter>
    parameters()
     
    default void
    zeroGrad()
     

  • Method Details

    • parameters

      List<Parameter> parameters()

    • zeroGrad

      default void zeroGrad()