config.json

{
  "model_type": "kimi-k2",
  "vocab_size": 50304,
  "n_layer": 12,
  "n_head": 12,
  "n_embd": 768,
  "dropout": 0.0,
  "bias": true,
  "use_moe": true,
  "num_experts": 8,
  "expert_capacity": 32,
  "top_k_experts": 2,
  "use_latent_attention": true,
  "latent_dim": 64,
  "block_size": 256,
  "tokenizer": "tiktoken/gpt2",
  "architectures": [
    "KimiK2ForCausalLM"
  ],
  "torch_dtype": "float32",
  "auto_map": {
    "AutoConfig": "modeling_kimik2.KimiK2Config",
    "AutoModelForCausalLM": "modeling_kimik2.KimiK2ForCausalLM"
  }
}