| { |
| "step": 1060, |
| "best_loss": 0.3585319519042969, |
| "total_params": 601134721, |
| "total_params_M": 601.134721, |
| "config": { |
| "hidden_dim": 1024, |
| "num_layers": 24, |
| "num_heads": 32, |
| "head_dim": 32, |
| "ffn_dim": 4096, |
| "max_seq_len": 1024, |
| "vocab_size": 32000, |
| "hot_memory_slots": 1024, |
| "warm_rank": 64, |
| "time_dim": 128, |
| "time_num_frequencies": 16, |
| "time_max_events": 64, |
| "cognition_state_dim": 256, |
| "cognition_modes": 4, |
| "cognition_prompt_types": 5, |
| "register_dim": 4, |
| "intent_dim": 4, |
| "max_tools": 64, |
| "num_specialists": 5, |
| "router_top_k": 2, |
| "attention_sparse_budget": 128, |
| "batch_size": 4, |
| "grad_accumulation": 4, |
| "total_steps": 15000, |
| "warmup_steps": 200, |
| "max_grad_norm": 0.5, |
| "weight_decay": 0.01, |
| "seed": 42, |
| "lr_backbone": 3e-05, |
| "lr_specialists": 0.0003, |
| "lr_router": 0.001, |
| "lr_tools": 0.0005, |
| "lr_depth": 0.0003, |
| "lr_embeddings": 0.0003, |
| "balance_loss_weight": 0.05, |
| "cognition_loss_weight": 0.1, |
| "aux_loss_weight": 0.3, |
| "phase1_end": 5000, |
| "phase2_end": 10000, |
| "phase3_end": 13000, |
| "foundation_depth": [ |
| 0.4, |
| 0.4, |
| 0.2, |
| 0.0, |
| 0.0 |
| ], |
| "reasoning_depth": [ |
| 0.15, |
| 0.15, |
| 0.25, |
| 0.25, |
| 0.2 |
| ], |
| "depth_depth": [ |
| 0.05, |
| 0.05, |
| 0.2, |
| 0.2, |
| 0.5 |
| ], |
| "omega_depth": [ |
| 0.2, |
| 0.2, |
| 0.2, |
| 0.2, |
| 0.2 |
| ], |
| "bonus_ratio": 0.15, |
| "language_ratio": 0.4, |
| "science_ratio": 0.05, |
| "gaussian_ratio": 0.02, |
| "checkpoint_every": 250, |
| "log_every": 10, |
| "eval_every": 250, |
| "max_nan_retries": 5, |
| "max_total_nans": 50, |
| "vram_warning_gb": 13.0, |
| "min_disk_gb": 3.0 |
| }, |
| "components": { |
| "infrastructure": { |
| "params": 1049600, |
| "params_M": 1.0496, |
| "pct": 0.17, |
| "tensors": 2, |
| "avg_std": 0.251326, |
| "avg_sparsity": 0.249756, |
| "avg_norm": 378.3161, |
| "mean_range": [ |
| 0.500488, |
| 1.006288 |
| ], |
| "std_range": [ |
| 0.002653, |
| 0.5 |
| ] |
| }, |
| "embeddings": { |
| "params": 131930112, |
| "params_M": 131.9301, |
| "pct": 21.95, |
| "tensors": 9, |
| "avg_std": 0.032689, |
| "avg_sparsity": 0.001297, |
| "avg_norm": 102.0778, |
| "mean_range": [ |
| -0.002077, |
| 5e-05 |
| ], |
| "std_range": [ |
| 0.019981, |
| 0.044033 |
| ] |
| }, |
| "memory": { |
| "params": 2491908, |
| "params_M": 2.4919, |
| "pct": 0.41, |
| "tensors": 12, |
| "avg_std": 0.014853, |
| "avg_sparsity": 0.0, |
| "avg_norm": 21.0367, |
| "mean_range": [ |
| -4.3e-05, |
| 4.143555 |
| ], |
| "std_range": [ |
| 0.006705, |
| 0.035267 |
| ] |
| }, |
| "time_engine": { |
| "params": 137458, |
| "params_M": 0.1375, |
| "pct": 0.02, |
| "tensors": 11, |
| "avg_std": 0.075931, |
| "avg_sparsity": 0.0, |
| "avg_norm": 6.0218, |
| "mean_range": [ |
| -0.002298, |
| 0.001424 |
| ], |
| "std_range": [ |
| 0.050549, |
| 0.10451 |
| ] |
| }, |
| "cognition": { |
| "params": 329220, |
| "params_M": 0.3292, |
| "pct": 0.05, |
| "tensors": 6, |
| "avg_std": 0.029082, |
| "avg_sparsity": 0.0, |
| "avg_norm": 4.103, |
| "mean_range": [ |
| -0.004059, |
| 0.000486 |
| ], |
| "std_range": [ |
| 0.017853, |
| 0.038105 |
| ] |
| }, |
| "attention": { |
| "params": 108952705, |
| "params_M": 108.9527, |
| "pct": 18.12, |
| "tensors": 248, |
| "avg_std": 0.177187, |
| "avg_sparsity": 0.000118, |
| "avg_norm": 45.714, |
| "mean_range": [ |
| -0.001203, |
| 1.00175 |
| ], |
| "std_range": [ |
| 0.000543, |
| 0.72641 |
| ] |
| }, |
| "modulator": { |
| "params": 1702930, |
| "params_M": 1.7029, |
| "pct": 0.28, |
| "tensors": 30, |
| "avg_std": 0.027381, |
| "avg_sparsity": 0.136364, |
| "avg_norm": 7.5217, |
| "mean_range": [ |
| -0.002171, |
| 0.000924 |
| ], |
| "std_range": [ |
| 0.0, |
| 0.204096 |
| ] |
| }, |
| "tool_cortex": { |
| "params": 2098241, |
| "params_M": 2.0982, |
| "pct": 0.35, |
| "tensors": 6, |
| "avg_std": 0.017572, |
| "avg_sparsity": 0.0, |
| "avg_norm": 8.1068, |
| "mean_range": [ |
| -9.8e-05, |
| 3e-06 |
| ], |
| "std_range": [ |
| 0.012762, |
| 0.0264 |
| ] |
| }, |
| "router": { |
| "params": 5120, |
| "params_M": 0.0051, |
| "pct": 0.0, |
| "tensors": 1, |
| "avg_std": 0.017936, |
| "avg_sparsity": 0.0, |
| "avg_norm": 1.2833, |
| "mean_range": [ |
| -7.5e-05, |
| -7.5e-05 |
| ], |
| "std_range": [ |
| 0.017936, |
| 0.017936 |
| ] |
| }, |
| "feedforward": { |
| "params": 302014464, |
| "params_M": 302.0145, |
| "pct": 50.24, |
| "tensors": 96, |
| "avg_std": 0.015305, |
| "avg_sparsity": 0.0, |
| "avg_norm": 38.8077, |
| "mean_range": [ |
| -2.2e-05, |
| 1.003368 |
| ], |
| "std_range": [ |
| 0.000904, |
| 0.020069 |
| ] |
| }, |
| "gaussian_head": { |
| "params": 14497971, |
| "params_M": 14.498, |
| "pct": 2.41, |
| "tensors": 28, |
| "avg_std": 0.02671, |
| "avg_sparsity": 0.151429, |
| "avg_norm": 11.9429, |
| "mean_range": [ |
| -0.006763, |
| 1.0 |
| ], |
| "std_range": [ |
| 0.0, |
| 0.160107 |
| ] |
| }, |
| "plug_membranes": { |
| "params": 3154944, |
| "params_M": 3.1549, |
| "pct": 0.52, |
| "tensors": 12, |
| "avg_std": 0.011197, |
| "avg_sparsity": 0.166667, |
| "avg_norm": 16.4031, |
| "mean_range": [ |
| -6.1e-05, |
| 1.0 |
| ], |
| "std_range": [ |
| 0.0, |
| 0.033837 |
| ] |
| }, |
| "aux_head": { |
| "params": 32770048, |
| "params_M": 32.77, |
| "pct": 5.45, |
| "tensors": 3, |
| "avg_std": 0.021155, |
| "avg_sparsity": 0.0, |
| "avg_norm": 54.8849, |
| "mean_range": [ |
| -0.00062, |
| 1.069667 |
| ], |
| "std_range": [ |
| 0.001715, |
| 0.039174 |
| ] |
| } |
| }, |
| "per_layer": { |
| "0": { |
| "params_M": 16.88, |
| "avg_weight_norm": 44.2836, |
| "avg_weight_std": 0.127194 |
| }, |
| "1": { |
| "params_M": 16.88, |
| "avg_weight_norm": 44.3485, |
| "avg_weight_std": 0.127518 |
| }, |
| "2": { |
| "params_M": 16.88, |
| "avg_weight_norm": 44.3608, |
| "avg_weight_std": 0.127563 |
| }, |
| "3": { |
| "params_M": 16.88, |
| "avg_weight_norm": 44.368, |
| "avg_weight_std": 0.127538 |
| }, |
| "4": { |
| "params_M": 16.88, |
| "avg_weight_norm": 44.3372, |
| "avg_weight_std": 0.127405 |
| }, |
| "5": { |
| "params_M": 16.88, |
| "avg_weight_norm": 44.3438, |
| "avg_weight_std": 0.127414 |
| }, |
| "6": { |
| "params_M": 16.88, |
| "avg_weight_norm": 44.3081, |
| "avg_weight_std": 0.127205 |
| }, |
| "7": { |
| "params_M": 16.88, |
| "avg_weight_norm": 44.3078, |
| "avg_weight_std": 0.127193 |
| }, |
| "8": { |
| "params_M": 16.88, |
| "avg_weight_norm": 44.2412, |
| "avg_weight_std": 0.126937 |
| }, |
| "9": { |
| "params_M": 16.88, |
| "avg_weight_norm": 44.2339, |
| "avg_weight_std": 0.126905 |
| }, |
| "10": { |
| "params_M": 16.88, |
| "avg_weight_norm": 44.2545, |
| "avg_weight_std": 0.126958 |
| }, |
| "11": { |
| "params_M": 16.88, |
| "avg_weight_norm": 44.2508, |
| "avg_weight_std": 0.126945 |
| }, |
| "12": { |
| "params_M": 16.88, |
| "avg_weight_norm": 44.257, |
| "avg_weight_std": 0.126973 |
| }, |
| "13": { |
| "params_M": 16.88, |
| "avg_weight_norm": 44.2314, |
| "avg_weight_std": 0.126884 |
| }, |
| "14": { |
| "params_M": 16.88, |
| "avg_weight_norm": 44.2501, |
| "avg_weight_std": 0.126928 |
| }, |
| "15": { |
| "params_M": 16.88, |
| "avg_weight_norm": 44.2423, |
| "avg_weight_std": 0.126894 |
| }, |
| "16": { |
| "params_M": 16.88, |
| "avg_weight_norm": 44.2393, |
| "avg_weight_std": 0.126899 |
| }, |
| "17": { |
| "params_M": 16.88, |
| "avg_weight_norm": 44.2408, |
| "avg_weight_std": 0.126913 |
| }, |
| "18": { |
| "params_M": 16.88, |
| "avg_weight_norm": 44.2545, |
| "avg_weight_std": 0.126951 |
| }, |
| "19": { |
| "params_M": 16.88, |
| "avg_weight_norm": 44.2493, |
| "avg_weight_std": 0.126935 |
| }, |
| "20": { |
| "params_M": 16.88, |
| "avg_weight_norm": 44.2524, |
| "avg_weight_std": 0.126946 |
| }, |
| "21": { |
| "params_M": 16.88, |
| "avg_weight_norm": 44.2534, |
| "avg_weight_std": 0.12696 |
| }, |
| "22": { |
| "params_M": 16.88, |
| "avg_weight_norm": 44.2421, |
| "avg_weight_std": 0.126907 |
| }, |
| "23": { |
| "params_M": 16.88, |
| "avg_weight_norm": 44.2696, |
| "avg_weight_std": 0.127022 |
| } |
| }, |
| "optimizer": { |
| "avg_momentum": 4e-06, |
| "max_momentum": 0.000396, |
| "avg_variance": 4e-08, |
| "max_variance": 1.017e-05, |
| "top_active": [ |
| [ |
| "layers.14.attn_norm.weight", |
| 0.000396 |
| ], |
| [ |
| "layers.10.attention.rope.sin_cached", |
| 6.3e-05 |
| ], |
| [ |
| "layers.14.ffn.up_proj.weight", |
| 4.4e-05 |
| ], |
| [ |
| "layers.14.attention.rope.sin_cached", |
| 4.2e-05 |
| ], |
| [ |
| "layers.14.attention.alpha_router.0.weight", |
| 3.9e-05 |
| ], |
| [ |
| "layers.14.attention.o_proj.weight", |
| 2.8e-05 |
| ], |
| [ |
| "layers.14.ffn.gate_proj.weight", |
| 2.4e-05 |
| ], |
| [ |
| "layers.14.attention.rope.cos_cached", |
| 1.5e-05 |
| ], |
| [ |
| "layers.15.ffn.down_proj.weight", |
| 8e-06 |
| ], |
| [ |
| "memory.hot.write_gate.2.weight", |
| 8e-06 |
| ] |
| ], |
| "least_active": [ |
| [ |
| "layers.4.attention.v_proj.weight", |
| 1.4e-07 |
| ], |
| [ |
| "layers.13.attention.o_proj.weight", |
| 1.4e-07 |
| ], |
| [ |
| "embeddings.output_head.weight", |
| 4e-08 |
| ], |
| [ |
| "layers.10.ffn.gate_proj.weight", |
| 1e-08 |
| ], |
| [ |
| "embeddings.token_embed.weight", |
| 0.0 |
| ] |
| ] |
| }, |
| "curriculum": {}, |
| "health": { |
| "goods": [ |
| "Layer norms perfectly uniform (ratio 1.003088)", |
| "infrastructure: healthy (std=0.2513, sparsity=0.2498)", |
| "embeddings: healthy (std=0.0327, sparsity=0.0013)", |
| "memory: healthy (std=0.0149, sparsity=0.0000)", |
| "time_engine: healthy (std=0.0759, sparsity=0.0000)", |
| "cognition: healthy (std=0.0291, sparsity=0.0000)", |
| "attention: healthy (std=0.1772, sparsity=0.0001)", |
| "modulator: healthy (std=0.0274, sparsity=0.1364)", |
| "tool_cortex: healthy (std=0.0176, sparsity=0.0000)", |
| "router: healthy (std=0.0179, sparsity=0.0000)", |
| "feedforward: healthy (std=0.0153, sparsity=0.0000)", |
| "gaussian_head: healthy (std=0.0267, sparsity=0.1514)", |
| "plug_membranes: healthy (std=0.0112, sparsity=0.1667)", |
| "aux_head: healthy (std=0.0212, sparsity=0.0000)" |
| ], |
| "issues": [] |
| }, |
| "gpu_eval": true |
| } |
|
|
