deepseek-r1-1.5b-packed/config.json

{
  "hidden_size": 1536,
  "intermediate_size": 8960,
  "num_attention_heads": 12,
  "num_key_value_heads": 2,
  "num_hidden_layers": 28,
  "vocab_size": 151936,
  "head_dim": 128,
  "rope_theta": 1000000.0,
  "rms_norm_eps": 1e-06,
  "n_levels": 7,
  "quant_type": "packed_unary"
}