generation_config.json

{
  "architectures": [
    "Max2ForCausalLM"
  ],
  "model_type": "max2",
  "model_name": "max2-nano",
  "hidden_size": 768,
  "intermediate_size": 2048,
  "num_hidden_layers": 12,
  "num_attention_heads": 12,
  "num_key_value_heads": 3,
  "vocab_size": 32000,
  "max_position_embeddings": 4096,
  "rope_theta": 10000.0,
  "use_moe": true,
  "num_experts": 4,
  "num_experts_per_tok": 1,
  "expert_hidden_size": 512,
  "rms_norm_eps": 1e-06,
  "torch_dtype": "float16",
  "transformers_version": "4.40.0"
}