Nebium

Causal Transformer for self-supervised next-move prediction on chess games.

Architecture

  • d_model: 64
  • n_heads: 4
  • n_layers: 1
  • vocab_size: 136
  • max_seq_len: 128
  • positional_encoding: rope
  • activation: swiglu
  • norm: rmsnorm

Last validation metrics

  • val/loss: 8.162410736083984
  • val/accuracy: 0.10000000149011612
  • val/top5_accuracy: 0.15000000596046448
  • val/perplexity: 3506.629981726336
Downloads last month

-

Downloads are not tracked for this model. How to track
Inference Providers NEW
This model isn't deployed by any Inference Provider. 🙋 Ask for provider support