{
  "architecture": "solar_open2",
  "buffer_values": 15360,
  "config_sha256": "fb6428ba165af1ace1d98f9170f6bafce061347593a94bd16b4b8aa3d6fe09f9",
  "family_tensor_counts": {
    "gqa_projection": 60,
    "kda_matrix_projection": 324,
    "kda_small_control_norm_conv": 216,
    "lm_head": 1,
    "norm": 97,
    "routed_expert_down": 15360,
    "routed_expert_gate": 15360,
    "routed_expert_up": 15360,
    "router_control": 96,
    "shared_expert": 144,
    "token_embedding": 1
  },
  "gqa": {
    "head_dim": 128,
    "kv_heads": 8,
    "output_gate": true,
    "query_heads": 64,
    "use_rope": false
  },
  "gqa_layers": [
    0,
    4,
    8,
    12,
    16,
    20,
    24,
    28,
    32,
    36,
    40,
    44
  ],
  "kda": {
    "allow_negative_eigenvalue": true,
    "gate_equation": "g=-exp(A_log)*softplus(g_raw+dt_bias), clamped at -5",
    "head_dim": 128,
    "heads": 64,
    "short_conv_kernel": 4,
    "source": "upstageAI/transformers modular_solar_open2.py",
    "state_equation": "S_t=exp(g_t)*S_(t-1)+k_t*(beta_t*(v_t-S_decay^T*k_t))^T",
    "use_full_proj": false
  },
  "kda_layers": [
    1,
    2,
    3,
    5,
    6,
    7,
    9,
    10,
    11,
    13,
    14,
    15,
    17,
    18,
    19,
    21,
    22,
    23,
    25,
    26,
    27,
    29,
    30,
    31,
    33,
    34,
    35,
    37,
    38,
    39,
    41,
    42,
    43,
    45,
    46,
    47
  ],
  "layer_schedule": [
    "GQA",
    "KDA",
    "KDA",
    "KDA",
    "GQA",
    "KDA",
    "KDA",
    "KDA",
    "GQA",
    "KDA",
    "KDA",
    "KDA",
    "GQA",
    "KDA",
    "KDA",
    "KDA",
    "GQA",
    "KDA",
    "KDA",
    "KDA",
    "GQA",
    "KDA",
    "KDA",
    "KDA",
    "GQA",
    "KDA",
    "KDA",
    "KDA",
    "GQA",
    "KDA",
    "KDA",
    "KDA",
    "GQA",
    "KDA",
    "KDA",
    "KDA",
    "GQA",
    "KDA",
    "KDA",
    "KDA",
    "GQA",
    "KDA",
    "KDA",
    "KDA",
    "GQA",
    "KDA",
    "KDA",
    "KDA"
  ],
  "layers": 48,
  "moe": {
    "normalize_topk_weights": true,
    "routed_experts": 320,
    "routed_scaling_factor": 1.0,
    "routed_width": 1280,
    "router_activation": "sigmoid",
    "shared_experts": 1,
    "top_k": 8
  },
  "normalized_tensor_patterns": {
    "lm_head.weight": 1,
    "model.embed_tokens.weight": 1,
    "model.layers.{layer}.input_layernorm.weight": 48,
    "model.layers.{layer}.mlp.experts.{expert}.down_proj.weight": 15360,
    "model.layers.{layer}.mlp.experts.{expert}.gate_proj.weight": 15360,
    "model.layers.{layer}.mlp.experts.{expert}.up_proj.weight": 15360,
    "model.layers.{layer}.mlp.gate.e_score_correction_bias": 48,
    "model.layers.{layer}.mlp.gate.weight": 48,
    "model.layers.{layer}.mlp.shared_experts.down_proj.weight": 48,
    "model.layers.{layer}.mlp.shared_experts.gate_proj.weight": 48,
    "model.layers.{layer}.mlp.shared_experts.up_proj.weight": 48,
    "model.layers.{layer}.post_attention_layernorm.weight": 48,
    "model.layers.{layer}.self_attn.A_log": 36,
    "model.layers.{layer}.self_attn.b_proj.weight": 36,
    "model.layers.{layer}.self_attn.dt_bias": 36,
    "model.layers.{layer}.self_attn.f_a_proj.weight": 36,
    "model.layers.{layer}.self_attn.f_b_proj.weight": 36,
    "model.layers.{layer}.self_attn.g_a_proj.weight": 36,
    "model.layers.{layer}.self_attn.g_b_proj.weight": 36,
    "model.layers.{layer}.self_attn.g_proj.weight": 12,
    "model.layers.{layer}.self_attn.k_conv1d.weight": 36,
    "model.layers.{layer}.self_attn.k_proj.weight": 48,
    "model.layers.{layer}.self_attn.o_norm.weight": 36,
    "model.layers.{layer}.self_attn.o_proj.weight": 48,
    "model.layers.{layer}.self_attn.q_conv1d.weight": 36,
    "model.layers.{layer}.self_attn.q_proj.weight": 48,
    "model.layers.{layer}.self_attn.v_conv1d.weight": 36,
    "model.layers.{layer}.self_attn.v_proj.weight": 48,
    "model.norm.weight": 1
  },
  "schema_version": 1,
  "source": {
    "repo": "upstage/Solar-Open2-250B",
    "revision": "9190fbe63a2ad8e17fc766ccceb36de7c66f004b"
  },
  "source_dtype_counts": {
    "BF16": 46971,
    "F32": 48
  },
  "stored_values": 250287810304,
  "tensor_count": 47019,
  "tensor_header_sha256": "97281a0595ceeeec8c7579b8a384b9d5c35a4e5956e16b1b96cb690bd63c7f58",
  "trainable_parameters": 250287794944,
  "validation": {
    "errors": [],
    "ok": true,
    "warnings": [
      "pilot projection 97.76 GiB exceeds the 92 GiB artifact gate; pilot is an inventory/loadability artifact and must not be published as the capacity candidate"
    ]
  }
}