Spaces:
Running
Running
| { | |
| "schema": "vr.kimi.output_formation_ledger@0.1", | |
| "model_id": "moonshotai/Kimi-K2.5", | |
| "prompt_tokens": [ | |
| 163584 | |
| ], | |
| "generated_tokens": [ | |
| { | |
| "step_index": 0, | |
| "input_token_ids": [ | |
| 163584 | |
| ], | |
| "output_token_id": 950, | |
| "output_logit": 27.711794, | |
| "logits_digest": "d24500ce107223c9a859c723211b59d43d96491a77459d3b97403d036f6f1341", | |
| "hidden_digest": "b3751954b18beb183a9d5f7b0601a31b28d74e36b573e8104c4007c3b970367a", | |
| "top_k_tokens": [ | |
| [ | |
| 950, | |
| 27.711794 | |
| ], | |
| [ | |
| 841, | |
| 20.556957 | |
| ], | |
| [ | |
| 4665, | |
| 15.4198 | |
| ], | |
| [ | |
| 32360, | |
| 14.036875 | |
| ], | |
| [ | |
| 6118, | |
| 13.822012 | |
| ], | |
| [ | |
| 91, | |
| 13.568241 | |
| ], | |
| [ | |
| 1121, | |
| 13.205778 | |
| ], | |
| [ | |
| 3100, | |
| 12.484051 | |
| ], | |
| [ | |
| 24576, | |
| 11.536135 | |
| ], | |
| [ | |
| 635, | |
| 11.474879 | |
| ], | |
| [ | |
| 38594, | |
| 11.33157 | |
| ], | |
| [ | |
| 2229, | |
| 11.053585 | |
| ], | |
| [ | |
| 11308, | |
| 10.809999 | |
| ], | |
| [ | |
| 146544, | |
| 9.996357 | |
| ], | |
| [ | |
| 5, | |
| 9.981656 | |
| ], | |
| [ | |
| 43627, | |
| 9.766911 | |
| ] | |
| ], | |
| "entropy": 0.0064897696, | |
| "margin": 7.1548367, | |
| "bytes_streamed": 35335326976, | |
| "layers": [ | |
| { | |
| "layer_idx": 0, | |
| "layer_kind": "dense", | |
| "residual_norm_in": 1.1708896, | |
| "residual_norm_out": 1.6389806, | |
| "expert_indices": null, | |
| "routing_weights": null, | |
| "bytes_streamed": 995000320 | |
| }, | |
| { | |
| "layer_idx": 1, | |
| "layer_kind": "moe", | |
| "residual_norm_in": 1.6389806, | |
| "residual_norm_out": 2.9620016, | |
| "expert_indices": [ | |
| 5, | |
| 33, | |
| 72, | |
| 105, | |
| 116, | |
| 154, | |
| 182, | |
| 272 | |
| ], | |
| "routing_weights": [ | |
| 0.08708401, | |
| 0.20157117, | |
| 0.14540458, | |
| 0.08845076, | |
| 0.12865292, | |
| 0.108699255, | |
| 0.12787405, | |
| 0.11226332 | |
| ], | |
| "bytes_streamed": 494044864 | |
| }, | |
| { | |
| "layer_idx": 2, | |
| "layer_kind": "moe", | |
| "residual_norm_in": 2.9620016, | |
| "residual_norm_out": 4.2335696, | |
| "expert_indices": [ | |
| 7, | |
| 25, | |
| 92, | |
| 179, | |
| 190, | |
| 240, | |
| 340, | |
| 343 | |
| ], | |
| "routing_weights": [ | |
| 0.1776416, | |
| 0.06865491, | |
| 0.15823834, | |
| 0.06387749, | |
| 0.12181032, | |
| 0.1846475, | |
| 0.060134605, | |
| 0.16499531 | |
| ], | |
| "bytes_streamed": 494044864 | |
| }, | |
| { | |
| "layer_idx": 3, | |
| "layer_kind": "moe", | |
| "residual_norm_in": 4.2335696, | |
| "residual_norm_out": 9.665392, | |
| "expert_indices": [ | |
| 15, | |
| 189, | |
| 190, | |
| 217, | |
| 253, | |
| 271, | |
| 337, | |
| 353 | |
| ], | |
| "routing_weights": [ | |
| 0.056847326, | |
| 0.062119152, | |
| 0.052167043, | |
| 0.35753915, | |
| 0.06827297, | |
| 0.049369168, | |
| 0.046695393, | |
| 0.30698994 | |
| ], | |
| "bytes_streamed": 494044864 | |
| }, | |
| { | |
| "layer_idx": 4, | |
| "layer_kind": "moe", | |
| "residual_norm_in": 9.665392, | |
| "residual_norm_out": 78.33159, | |
| "expert_indices": [ | |
| 52, | |
| 98, | |
| 106, | |
| 109, | |
| 120, | |
| 189, | |
| 191, | |
| 324 | |
| ], | |
| "routing_weights": [ | |
| 0.07070846, | |
| 0.08047036, | |
| 0.07718346, | |
| 0.08113837, | |
| 0.07101389, | |
| 0.47025698, | |
| 0.0818232, | |
| 0.067405306 | |
| ], | |
| "bytes_streamed": 494044864 | |
| }, | |
| { | |
| "layer_idx": 5, | |
| "layer_kind": "moe", | |
| "residual_norm_in": 78.33159, | |
| "residual_norm_out": 84.419685, | |
| "expert_indices": [ | |
| 12, | |
| 17, | |
| 121, | |
| 124, | |
| 160, | |
| 186, | |
| 195, | |
| 237 | |
| ], | |
| "routing_weights": [ | |
| 0.18882453, | |
| 0.100765295, | |
| 0.11140824, | |
| 0.17085916, | |
| 0.10019476, | |
| 0.103869334, | |
| 0.0954801, | |
| 0.12859854 | |
| ], | |
| "bytes_streamed": 494044864 | |
| }, | |
| { | |
| "layer_idx": 6, | |
| "layer_kind": "moe", | |
| "residual_norm_in": 84.419685, | |
| "residual_norm_out": 390.0009, | |
| "expert_indices": [ | |
| 18, | |
| 45, | |
| 88, | |
| 239, | |
| 243, | |
| 269, | |
| 334, | |
| 346 | |
| ], | |
| "routing_weights": [ | |
| 0.04630539, | |
| 0.02936879, | |
| 0.03381209, | |
| 0.03124016, | |
| 0.52461165, | |
| 0.059491146, | |
| 0.037579484, | |
| 0.23759125 | |
| ], | |
| "bytes_streamed": 494044864 | |
| }, | |
| { | |
| "layer_idx": 7, | |
| "layer_kind": "moe", | |
| "residual_norm_in": 390.0009, | |
| "residual_norm_out": 1106.5599, | |
| "expert_indices": [ | |
| 54, | |
| 57, | |
| 155, | |
| 183, | |
| 185, | |
| 192, | |
| 302, | |
| 361 | |
| ], | |
| "routing_weights": [ | |
| 0.025255663, | |
| 0.0061685992, | |
| 0.0072444, | |
| 0.92817444, | |
| 0.0096815275, | |
| 0.007881453, | |
| 0.00678511, | |
| 0.008808856 | |
| ], | |
| "bytes_streamed": 494044864 | |
| }, | |
| { | |
| "layer_idx": 8, | |
| "layer_kind": "moe", | |
| "residual_norm_in": 1106.5599, | |
| "residual_norm_out": 1492.6199, | |
| "expert_indices": [ | |
| 86, | |
| 128, | |
| 175, | |
| 287, | |
| 311, | |
| 336, | |
| 337, | |
| 351 | |
| ], | |
| "routing_weights": [ | |
| 0.06048175, | |
| 0.030799594, | |
| 0.030457731, | |
| 0.03846677, | |
| 0.029505622, | |
| 0.03274197, | |
| 0.034052663, | |
| 0.743494 | |
| ], | |
| "bytes_streamed": 494044864 | |
| }, | |
| { | |
| "layer_idx": 9, | |
| "layer_kind": "moe", | |
| "residual_norm_in": 1492.6199, | |
| "residual_norm_out": 1690.9316, | |
| "expert_indices": [ | |
| 30, | |
| 131, | |
| 182, | |
| 183, | |
| 198, | |
| 238, | |
| 365, | |
| 377 | |
| ], | |
| "routing_weights": [ | |
| 0.056319214, | |
| 0.047798157, | |
| 0.056981634, | |
| 0.051912706, | |
| 0.04586144, | |
| 0.05751384, | |
| 0.04544697, | |
| 0.63816607 | |
| ], | |
| "bytes_streamed": 494044864 | |
| }, | |
| { | |
| "layer_idx": 10, | |
| "layer_kind": "moe", | |
| "residual_norm_in": 1690.9316, | |
| "residual_norm_out": 1954.9259, | |
| "expert_indices": [ | |
| 5, | |
| 34, | |
| 46, | |
| 118, | |
| 234, | |
| 240, | |
| 292, | |
| 296 | |
| ], | |
| "routing_weights": [ | |
| 0.033065848, | |
| 0.015240267, | |
| 0.015210066, | |
| 0.015278822, | |
| 0.014619392, | |
| 0.013330271, | |
| 0.017498888, | |
| 0.87575644 | |
| ], | |
| "bytes_streamed": 494044864 | |
| }, | |
| { | |
| "layer_idx": 11, | |
| "layer_kind": "moe", | |
| "residual_norm_in": 1954.9259, | |
| "residual_norm_out": 1956.7909, | |
| "expert_indices": [ | |
| 31, | |
| 124, | |
| 141, | |
| 180, | |
| 224, | |
| 237, | |
| 252, | |
| 253 | |
| ], | |
| "routing_weights": [ | |
| 0.13398923, | |
| 0.12286287, | |
| 0.11951655, | |
| 0.12470391, | |
| 0.13138647, | |
| 0.11974149, | |
| 0.1242872, | |
| 0.12351226 | |
| ], | |
| "bytes_streamed": 494044864 | |
| }, | |
| { | |
| "layer_idx": 12, | |
| "layer_kind": "moe", | |
| "residual_norm_in": 1956.7909, | |
| "residual_norm_out": 1980.5905, | |
| "expert_indices": [ | |
| 25, | |
| 51, | |
| 92, | |
| 95, | |
| 139, | |
| 189, | |
| 270, | |
| 371 | |
| ], | |
| "routing_weights": [ | |
| 0.11804447, | |
| 0.12368669, | |
| 0.1154379, | |
| 0.12428873, | |
| 0.1271027, | |
| 0.13917553, | |
| 0.1261879, | |
| 0.1260761 | |
| ], | |
| "bytes_streamed": 494044864 | |
| }, | |
| { | |
| "layer_idx": 13, | |
| "layer_kind": "moe", | |
| "residual_norm_in": 1980.5905, | |
| "residual_norm_out": 1982.6885, | |
| "expert_indices": [ | |
| 65, | |
| 68, | |
| 103, | |
| 194, | |
| 203, | |
| 228, | |
| 312, | |
| 348 | |
| ], | |
| "routing_weights": [ | |
| 0.15230045, | |
| 0.12651002, | |
| 0.12470712, | |
| 0.11999359, | |
| 0.12181892, | |
| 0.11676216, | |
| 0.12590373, | |
| 0.112004004 | |
| ], | |
| "bytes_streamed": 494044864 | |
| }, | |
| { | |
| "layer_idx": 14, | |
| "layer_kind": "moe", | |
| "residual_norm_in": 1982.6885, | |
| "residual_norm_out": 1985.324, | |
| "expert_indices": [ | |
| 166, | |
| 182, | |
| 204, | |
| 207, | |
| 262, | |
| 269, | |
| 303, | |
| 377 | |
| ], | |
| "routing_weights": [ | |
| 0.12826355, | |
| 0.1121166, | |
| 0.1671414, | |
| 0.11411119, | |
| 0.12515007, | |
| 0.110577546, | |
| 0.11894186, | |
| 0.12369785 | |
| ], | |
| "bytes_streamed": 494044864 | |
| }, | |
| { | |
| "layer_idx": 15, | |
| "layer_kind": "moe", | |
| "residual_norm_in": 1985.324, | |
| "residual_norm_out": 2258.2163, | |
| "expert_indices": [ | |
| 11, | |
| 109, | |
| 136, | |
| 150, | |
| 194, | |
| 212, | |
| 224, | |
| 279 | |
| ], | |
| "routing_weights": [ | |
| 0.07922825, | |
| 0.082601346, | |
| 0.08270255, | |
| 0.078739665, | |
| 0.080214955, | |
| 0.08108263, | |
| 0.07728815, | |
| 0.4381424 | |
| ], | |
| "bytes_streamed": 494044864 | |
| }, | |
| { | |
| "layer_idx": 16, | |
| "layer_kind": "moe", | |
| "residual_norm_in": 2258.2163, | |
| "residual_norm_out": 2261.7844, | |
| "expert_indices": [ | |
| 19, | |
| 80, | |
| 89, | |
| 91, | |
| 204, | |
| 225, | |
| 249, | |
| 339 | |
| ], | |
| "routing_weights": [ | |
| 0.111320116, | |
| 0.12770423, | |
| 0.12904347, | |
| 0.12393298, | |
| 0.13755625, | |
| 0.11228151, | |
| 0.14229251, | |
| 0.11586892 | |
| ], | |
| "bytes_streamed": 494044864 | |
| }, | |
| { | |
| "layer_idx": 17, | |
| "layer_kind": "moe", | |
| "residual_norm_in": 2261.7844, | |
| "residual_norm_out": 2275.0195, | |
| "expert_indices": [ | |
| 81, | |
| 113, | |
| 128, | |
| 161, | |
| 282, | |
| 303, | |
| 304, | |
| 364 | |
| ], | |
| "routing_weights": [ | |
| 0.13470754, | |
| 0.12498503, | |
| 0.1351004, | |
| 0.13626957, | |
| 0.10787495, | |
| 0.1059576, | |
| 0.11651543, | |
| 0.13858952 | |
| ], | |
| "bytes_streamed": 494044864 | |
| }, | |
| { | |
| "layer_idx": 18, | |
| "layer_kind": "moe", | |
| "residual_norm_in": 2275.0195, | |
| "residual_norm_out": 2843.4258, | |
| "expert_indices": [ | |
| 30, | |
| 72, | |
| 95, | |
| 143, | |
| 230, | |
| 309, | |
| 329, | |
| 378 | |
| ], | |
| "routing_weights": [ | |
| 0.061736643, | |
| 0.06080599, | |
| 0.059534334, | |
| 0.5461069, | |
| 0.059427556, | |
| 0.06282737, | |
| 0.059586346, | |
| 0.08997482 | |
| ], | |
| "bytes_streamed": 494044864 | |
| }, | |
| { | |
| "layer_idx": 19, | |
| "layer_kind": "moe", | |
| "residual_norm_in": 2843.4258, | |
| "residual_norm_out": 2845.8062, | |
| "expert_indices": [ | |
| 28, | |
| 62, | |
| 84, | |
| 210, | |
| 258, | |
| 264, | |
| 278, | |
| 318 | |
| ], | |
| "routing_weights": [ | |
| 0.12718764, | |
| 0.11587796, | |
| 0.11643613, | |
| 0.11349039, | |
| 0.1294555, | |
| 0.121852785, | |
| 0.15666126, | |
| 0.11903835 | |
| ], | |
| "bytes_streamed": 494044864 | |
| }, | |
| { | |
| "layer_idx": 20, | |
| "layer_kind": "moe", | |
| "residual_norm_in": 2845.8062, | |
| "residual_norm_out": 2961.956, | |
| "expert_indices": [ | |
| 7, | |
| 108, | |
| 146, | |
| 227, | |
| 260, | |
| 273, | |
| 329, | |
| 351 | |
| ], | |
| "routing_weights": [ | |
| 0.12041012, | |
| 0.120457426, | |
| 0.11941098, | |
| 0.1164421, | |
| 0.15163466, | |
| 0.122804046, | |
| 0.13331617, | |
| 0.11552452 | |
| ], | |
| "bytes_streamed": 494044864 | |
| }, | |
| { | |
| "layer_idx": 21, | |
| "layer_kind": "moe", | |
| "residual_norm_in": 2961.956, | |
| "residual_norm_out": 2995.6318, | |
| "expert_indices": [ | |
| 8, | |
| 24, | |
| 140, | |
| 151, | |
| 250, | |
| 264, | |
| 313, | |
| 342 | |
| ], | |
| "routing_weights": [ | |
| 0.12219975, | |
| 0.11388809, | |
| 0.11746957, | |
| 0.11987376, | |
| 0.12484143, | |
| 0.122165635, | |
| 0.13231997, | |
| 0.14724179 | |
| ], | |
| "bytes_streamed": 494044864 | |
| }, | |
| { | |
| "layer_idx": 22, | |
| "layer_kind": "moe", | |
| "residual_norm_in": 2995.6318, | |
| "residual_norm_out": 2998.1895, | |
| "expert_indices": [ | |
| 3, | |
| 11, | |
| 78, | |
| 119, | |
| 206, | |
| 236, | |
| 249, | |
| 334 | |
| ], | |
| "routing_weights": [ | |
| 0.11585079, | |
| 0.12601943, | |
| 0.12537284, | |
| 0.12319894, | |
| 0.13266197, | |
| 0.116407886, | |
| 0.12136969, | |
| 0.13911854 | |
| ], | |
| "bytes_streamed": 494044864 | |
| }, | |
| { | |
| "layer_idx": 23, | |
| "layer_kind": "moe", | |
| "residual_norm_in": 2998.1895, | |
| "residual_norm_out": 3121.4033, | |
| "expert_indices": [ | |
| 8, | |
| 44, | |
| 125, | |
| 207, | |
| 209, | |
| 226, | |
| 283, | |
| 325 | |
| ], | |
| "routing_weights": [ | |
| 0.12773134, | |
| 0.11835282, | |
| 0.13288984, | |
| 0.11907172, | |
| 0.117436275, | |
| 0.12930648, | |
| 0.11873839, | |
| 0.13647315 | |
| ], | |
| "bytes_streamed": 494044864 | |
| }, | |
| { | |
| "layer_idx": 24, | |
| "layer_kind": "moe", | |
| "residual_norm_in": 3121.4033, | |
| "residual_norm_out": 3261.4487, | |
| "expert_indices": [ | |
| 11, | |
| 46, | |
| 58, | |
| 59, | |
| 89, | |
| 223, | |
| 294, | |
| 376 | |
| ], | |
| "routing_weights": [ | |
| 0.13790713, | |
| 0.12226996, | |
| 0.12022834, | |
| 0.1308028, | |
| 0.11101906, | |
| 0.12276518, | |
| 0.13959049, | |
| 0.11541706 | |
| ], | |
| "bytes_streamed": 494044864 | |
| }, | |
| { | |
| "layer_idx": 25, | |
| "layer_kind": "moe", | |
| "residual_norm_in": 3261.4487, | |
| "residual_norm_out": 3266.5002, | |
| "expert_indices": [ | |
| 32, | |
| 52, | |
| 186, | |
| 275, | |
| 292, | |
| 317, | |
| 325, | |
| 355 | |
| ], | |
| "routing_weights": [ | |
| 0.1295867, | |
| 0.13489805, | |
| 0.117950104, | |
| 0.12863116, | |
| 0.1278383, | |
| 0.120717674, | |
| 0.11842304, | |
| 0.12195493 | |
| ], | |
| "bytes_streamed": 494044864 | |
| }, | |
| { | |
| "layer_idx": 26, | |
| "layer_kind": "moe", | |
| "residual_norm_in": 3266.5002, | |
| "residual_norm_out": 3269.6042, | |
| "expert_indices": [ | |
| 28, | |
| 54, | |
| 109, | |
| 143, | |
| 164, | |
| 187, | |
| 234, | |
| 326 | |
| ], | |
| "routing_weights": [ | |
| 0.12403764, | |
| 0.124741524, | |
| 0.123365216, | |
| 0.12436722, | |
| 0.12376982, | |
| 0.13259424, | |
| 0.12488024, | |
| 0.122244105 | |
| ], | |
| "bytes_streamed": 494044864 | |
| }, | |
| { | |
| "layer_idx": 27, | |
| "layer_kind": "moe", | |
| "residual_norm_in": 3269.6042, | |
| "residual_norm_out": 3274.1047, | |
| "expert_indices": [ | |
| 5, | |
| 22, | |
| 23, | |
| 48, | |
| 115, | |
| 160, | |
| 188, | |
| 218 | |
| ], | |
| "routing_weights": [ | |
| 0.1290535, | |
| 0.11797063, | |
| 0.12373113, | |
| 0.119161785, | |
| 0.119591445, | |
| 0.122973636, | |
| 0.1374787, | |
| 0.13003916 | |
| ], | |
| "bytes_streamed": 494044864 | |
| }, | |
| { | |
| "layer_idx": 28, | |
| "layer_kind": "moe", | |
| "residual_norm_in": 3274.1047, | |
| "residual_norm_out": 3275.8123, | |
| "expert_indices": [ | |
| 83, | |
| 117, | |
| 178, | |
| 218, | |
| 233, | |
| 260, | |
| 305, | |
| 342 | |
| ], | |
| "routing_weights": [ | |
| 0.12889512, | |
| 0.12080206, | |
| 0.12741528, | |
| 0.119030036, | |
| 0.12386303, | |
| 0.12633918, | |
| 0.13027582, | |
| 0.12337943 | |
| ], | |
| "bytes_streamed": 494044864 | |
| }, | |
| { | |
| "layer_idx": 29, | |
| "layer_kind": "moe", | |
| "residual_norm_in": 3275.8123, | |
| "residual_norm_out": 3280.281, | |
| "expert_indices": [ | |
| 19, | |
| 132, | |
| 145, | |
| 213, | |
| 264, | |
| 275, | |
| 309, | |
| 374 | |
| ], | |
| "routing_weights": [ | |
| 0.1201872, | |
| 0.11707843, | |
| 0.13667186, | |
| 0.11928836, | |
| 0.11894966, | |
| 0.14129789, | |
| 0.13095954, | |
| 0.115567006 | |
| ], | |
| "bytes_streamed": 494044864 | |
| }, | |
| { | |
| "layer_idx": 30, | |
| "layer_kind": "moe", | |
| "residual_norm_in": 3280.281, | |
| "residual_norm_out": 3280.3618, | |
| "expert_indices": [ | |
| 55, | |
| 109, | |
| 165, | |
| 246, | |
| 254, | |
| 264, | |
| 275, | |
| 282 | |
| ], | |
| "routing_weights": [ | |
| 0.12114449, | |
| 0.1265742, | |
| 0.12592877, | |
| 0.13100544, | |
| 0.12474784, | |
| 0.12301072, | |
| 0.12573223, | |
| 0.121856295 | |
| ], | |
| "bytes_streamed": 494044864 | |
| }, | |
| { | |
| "layer_idx": 31, | |
| "layer_kind": "moe", | |
| "residual_norm_in": 3280.3618, | |
| "residual_norm_out": 3283.6667, | |
| "expert_indices": [ | |
| 45, | |
| 150, | |
| 196, | |
| 198, | |
| 243, | |
| 323, | |
| 333, | |
| 359 | |
| ], | |
| "routing_weights": [ | |
| 0.12434853, | |
| 0.12571236, | |
| 0.12164109, | |
| 0.12634422, | |
| 0.12098487, | |
| 0.12598804, | |
| 0.12444442, | |
| 0.1305364 | |
| ], | |
| "bytes_streamed": 494044864 | |
| }, | |
| { | |
| "layer_idx": 32, | |
| "layer_kind": "moe", | |
| "residual_norm_in": 3283.6667, | |
| "residual_norm_out": 3285.134, | |
| "expert_indices": [ | |
| 21, | |
| 76, | |
| 182, | |
| 270, | |
| 276, | |
| 286, | |
| 336, | |
| 345 | |
| ], | |
| "routing_weights": [ | |
| 0.1239256, | |
| 0.12669827, | |
| 0.12782235, | |
| 0.12475172, | |
| 0.12783591, | |
| 0.1255458, | |
| 0.12194816, | |
| 0.12147223 | |
| ], | |
| "bytes_streamed": 494044864 | |
| }, | |
| { | |
| "layer_idx": 33, | |
| "layer_kind": "moe", | |
| "residual_norm_in": 3285.134, | |
| "residual_norm_out": 3284.3105, | |
| "expert_indices": [ | |
| 9, | |
| 53, | |
| 54, | |
| 57, | |
| 267, | |
| 281, | |
| 303, | |
| 309 | |
| ], | |
| "routing_weights": [ | |
| 0.12560827, | |
| 0.12389937, | |
| 0.12423009, | |
| 0.12410614, | |
| 0.12650697, | |
| 0.12478608, | |
| 0.12493784, | |
| 0.12592526 | |
| ], | |
| "bytes_streamed": 494044864 | |
| }, | |
| { | |
| "layer_idx": 34, | |
| "layer_kind": "moe", | |
| "residual_norm_in": 3284.3105, | |
| "residual_norm_out": 3284.2253, | |
| "expert_indices": [ | |
| 54, | |
| 96, | |
| 152, | |
| 206, | |
| 250, | |
| 295, | |
| 325, | |
| 339 | |
| ], | |
| "routing_weights": [ | |
| 0.12830862, | |
| 0.12542318, | |
| 0.122119926, | |
| 0.12366778, | |
| 0.12476752, | |
| 0.12551583, | |
| 0.12344778, | |
| 0.12674934 | |
| ], | |
| "bytes_streamed": 494044864 | |
| }, | |
| { | |
| "layer_idx": 35, | |
| "layer_kind": "moe", | |
| "residual_norm_in": 3284.2253, | |
| "residual_norm_out": 3284.423, | |
| "expert_indices": [ | |
| 9, | |
| 68, | |
| 86, | |
| 142, | |
| 162, | |
| 176, | |
| 257, | |
| 326 | |
| ], | |
| "routing_weights": [ | |
| 0.12454218, | |
| 0.12767828, | |
| 0.12446839, | |
| 0.12491049, | |
| 0.12380854, | |
| 0.1271105, | |
| 0.12260751, | |
| 0.1248741 | |
| ], | |
| "bytes_streamed": 494044864 | |
| }, | |
| { | |
| "layer_idx": 36, | |
| "layer_kind": "moe", | |
| "residual_norm_in": 3284.423, | |
| "residual_norm_out": 3282.616, | |
| "expert_indices": [ | |
| 51, | |
| 76, | |
| 132, | |
| 200, | |
| 236, | |
| 276, | |
| 317, | |
| 383 | |
| ], | |
| "routing_weights": [ | |
| 0.12545687, | |
| 0.12348957, | |
| 0.12646745, | |
| 0.123585485, | |
| 0.12351756, | |
| 0.123819746, | |
| 0.1276341, | |
| 0.1260292 | |
| ], | |
| "bytes_streamed": 494044864 | |
| }, | |
| { | |
| "layer_idx": 37, | |
| "layer_kind": "moe", | |
| "residual_norm_in": 3282.616, | |
| "residual_norm_out": 3278.0051, | |
| "expert_indices": [ | |
| 22, | |
| 75, | |
| 156, | |
| 182, | |
| 187, | |
| 189, | |
| 230, | |
| 321 | |
| ], | |
| "routing_weights": [ | |
| 0.12714444, | |
| 0.12565665, | |
| 0.12507115, | |
| 0.12419245, | |
| 0.12688795, | |
| 0.12317565, | |
| 0.12531842, | |
| 0.12255332 | |
| ], | |
| "bytes_streamed": 494044864 | |
| }, | |
| { | |
| "layer_idx": 38, | |
| "layer_kind": "moe", | |
| "residual_norm_in": 3278.0051, | |
| "residual_norm_out": 3277.3687, | |
| "expert_indices": [ | |
| 116, | |
| 118, | |
| 141, | |
| 142, | |
| 165, | |
| 246, | |
| 296, | |
| 305 | |
| ], | |
| "routing_weights": [ | |
| 0.12589957, | |
| 0.12569398, | |
| 0.124183916, | |
| 0.12590075, | |
| 0.12354493, | |
| 0.1264788, | |
| 0.12358198, | |
| 0.124716096 | |
| ], | |
| "bytes_streamed": 494044864 | |
| }, | |
| { | |
| "layer_idx": 39, | |
| "layer_kind": "moe", | |
| "residual_norm_in": 3277.3687, | |
| "residual_norm_out": 3276.2964, | |
| "expert_indices": [ | |
| 1, | |
| 21, | |
| 53, | |
| 74, | |
| 153, | |
| 232, | |
| 278, | |
| 381 | |
| ], | |
| "routing_weights": [ | |
| 0.12392103, | |
| 0.12356896, | |
| 0.12509237, | |
| 0.12476749, | |
| 0.1261542, | |
| 0.1256581, | |
| 0.12628376, | |
| 0.124554105 | |
| ], | |
| "bytes_streamed": 494044864 | |
| }, | |
| { | |
| "layer_idx": 40, | |
| "layer_kind": "moe", | |
| "residual_norm_in": 3276.2964, | |
| "residual_norm_out": 3271.916, | |
| "expert_indices": [ | |
| 14, | |
| 54, | |
| 80, | |
| 205, | |
| 243, | |
| 350, | |
| 360, | |
| 378 | |
| ], | |
| "routing_weights": [ | |
| 0.123624906, | |
| 0.12560081, | |
| 0.12601347, | |
| 0.12464201, | |
| 0.124437995, | |
| 0.124908455, | |
| 0.1261274, | |
| 0.124645 | |
| ], | |
| "bytes_streamed": 494044864 | |
| }, | |
| { | |
| "layer_idx": 41, | |
| "layer_kind": "moe", | |
| "residual_norm_in": 3271.916, | |
| "residual_norm_out": 3268.541, | |
| "expert_indices": [ | |
| 0, | |
| 12, | |
| 102, | |
| 118, | |
| 143, | |
| 323, | |
| 346, | |
| 361 | |
| ], | |
| "routing_weights": [ | |
| 0.12718983, | |
| 0.12352078, | |
| 0.124371946, | |
| 0.1250337, | |
| 0.12582456, | |
| 0.12378337, | |
| 0.1258188, | |
| 0.12445698 | |
| ], | |
| "bytes_streamed": 494044864 | |
| }, | |
| { | |
| "layer_idx": 42, | |
| "layer_kind": "moe", | |
| "residual_norm_in": 3268.541, | |
| "residual_norm_out": 3262.465, | |
| "expert_indices": [ | |
| 11, | |
| 63, | |
| 66, | |
| 121, | |
| 233, | |
| 282, | |
| 286, | |
| 322 | |
| ], | |
| "routing_weights": [ | |
| 0.12231816, | |
| 0.1257218, | |
| 0.12679411, | |
| 0.12391464, | |
| 0.12539703, | |
| 0.12225602, | |
| 0.12437119, | |
| 0.129227 | |
| ], | |
| "bytes_streamed": 494044864 | |
| }, | |
| { | |
| "layer_idx": 43, | |
| "layer_kind": "moe", | |
| "residual_norm_in": 3262.465, | |
| "residual_norm_out": 3258.3584, | |
| "expert_indices": [ | |
| 48, | |
| 97, | |
| 111, | |
| 145, | |
| 166, | |
| 258, | |
| 275, | |
| 310 | |
| ], | |
| "routing_weights": [ | |
| 0.12841281, | |
| 0.12592939, | |
| 0.123062134, | |
| 0.11106127, | |
| 0.13067317, | |
| 0.12302472, | |
| 0.12871376, | |
| 0.1291227 | |
| ], | |
| "bytes_streamed": 494044864 | |
| }, | |
| { | |
| "layer_idx": 44, | |
| "layer_kind": "moe", | |
| "residual_norm_in": 3258.3584, | |
| "residual_norm_out": 3249.8809, | |
| "expert_indices": [ | |
| 52, | |
| 80, | |
| 135, | |
| 143, | |
| 145, | |
| 204, | |
| 209, | |
| 313 | |
| ], | |
| "routing_weights": [ | |
| 0.15059605, | |
| 0.1502398, | |
| 0.11775559, | |
| 0.08767917, | |
| 0.12829646, | |
| 0.09496362, | |
| 0.119430825, | |
| 0.15103848 | |
| ], | |
| "bytes_streamed": 494044864 | |
| }, | |
| { | |
| "layer_idx": 45, | |
| "layer_kind": "moe", | |
| "residual_norm_in": 3249.8809, | |
| "residual_norm_out": 3247.127, | |
| "expert_indices": [ | |
| 20, | |
| 45, | |
| 57, | |
| 59, | |
| 105, | |
| 176, | |
| 271, | |
| 320 | |
| ], | |
| "routing_weights": [ | |
| 0.12892541, | |
| 0.11697536, | |
| 0.1406315, | |
| 0.12651278, | |
| 0.13517754, | |
| 0.09021727, | |
| 0.13699074, | |
| 0.124569386 | |
| ], | |
| "bytes_streamed": 494044864 | |
| }, | |
| { | |
| "layer_idx": 46, | |
| "layer_kind": "moe", | |
| "residual_norm_in": 3247.127, | |
| "residual_norm_out": 3241.98, | |
| "expert_indices": [ | |
| 7, | |
| 13, | |
| 31, | |
| 58, | |
| 198, | |
| 206, | |
| 222, | |
| 361 | |
| ], | |
| "routing_weights": [ | |
| 0.13090932, | |
| 0.13061666, | |
| 0.11974917, | |
| 0.13284738, | |
| 0.13148265, | |
| 0.12531953, | |
| 0.10203685, | |
| 0.1270384 | |
| ], | |
| "bytes_streamed": 494044864 | |
| }, | |
| { | |
| "layer_idx": 47, | |
| "layer_kind": "moe", | |
| "residual_norm_in": 3241.98, | |
| "residual_norm_out": 3239.7107, | |
| "expert_indices": [ | |
| 50, | |
| 91, | |
| 248, | |
| 268, | |
| 274, | |
| 305, | |
| 359, | |
| 377 | |
| ], | |
| "routing_weights": [ | |
| 0.090512514, | |
| 0.12572674, | |
| 0.12560987, | |
| 0.13680649, | |
| 0.13461363, | |
| 0.13834189, | |
| 0.11906947, | |
| 0.12931943 | |
| ], | |
| "bytes_streamed": 494044864 | |
| }, | |
| { | |
| "layer_idx": 48, | |
| "layer_kind": "moe", | |
| "residual_norm_in": 3239.7107, | |
| "residual_norm_out": 3232.344, | |
| "expert_indices": [ | |
| 21, | |
| 140, | |
| 180, | |
| 211, | |
| 216, | |
| 265, | |
| 275, | |
| 356 | |
| ], | |
| "routing_weights": [ | |
| 0.08928093, | |
| 0.119472556, | |
| 0.14138734, | |
| 0.13919228, | |
| 0.1308221, | |
| 0.12825514, | |
| 0.12513568, | |
| 0.12645397 | |
| ], | |
| "bytes_streamed": 494044864 | |
| }, | |
| { | |
| "layer_idx": 49, | |
| "layer_kind": "moe", | |
| "residual_norm_in": 3232.344, | |
| "residual_norm_out": 3220.2454, | |
| "expert_indices": [ | |
| 65, | |
| 132, | |
| 136, | |
| 141, | |
| 257, | |
| 330, | |
| 331, | |
| 381 | |
| ], | |
| "routing_weights": [ | |
| 0.15157294, | |
| 0.14870884, | |
| 0.12575085, | |
| 0.14817211, | |
| 0.12935571, | |
| 0.12963834, | |
| 0.084066756, | |
| 0.08273444 | |
| ], | |
| "bytes_streamed": 494044864 | |
| }, | |
| { | |
| "layer_idx": 50, | |
| "layer_kind": "moe", | |
| "residual_norm_in": 3220.2454, | |
| "residual_norm_out": 3208.0757, | |
| "expert_indices": [ | |
| 2, | |
| 46, | |
| 77, | |
| 114, | |
| 160, | |
| 162, | |
| 175, | |
| 324 | |
| ], | |
| "routing_weights": [ | |
| 0.071680024, | |
| 0.1342091, | |
| 0.14428033, | |
| 0.13734862, | |
| 0.1444514, | |
| 0.13382247, | |
| 0.11261745, | |
| 0.12159063 | |
| ], | |
| "bytes_streamed": 494044864 | |
| }, | |
| { | |
| "layer_idx": 51, | |
| "layer_kind": "moe", | |
| "residual_norm_in": 3208.0757, | |
| "residual_norm_out": 3198.2842, | |
| "expert_indices": [ | |
| 36, | |
| 130, | |
| 189, | |
| 218, | |
| 263, | |
| 271, | |
| 317, | |
| 362 | |
| ], | |
| "routing_weights": [ | |
| 0.07808465, | |
| 0.14101124, | |
| 0.13042152, | |
| 0.115859196, | |
| 0.13718922, | |
| 0.12932166, | |
| 0.12141399, | |
| 0.14669852 | |
| ], | |
| "bytes_streamed": 494044864 | |
| }, | |
| { | |
| "layer_idx": 52, | |
| "layer_kind": "moe", | |
| "residual_norm_in": 3198.2842, | |
| "residual_norm_out": 3180.221, | |
| "expert_indices": [ | |
| 10, | |
| 31, | |
| 40, | |
| 49, | |
| 52, | |
| 96, | |
| 259, | |
| 342 | |
| ], | |
| "routing_weights": [ | |
| 0.13195449, | |
| 0.12658952, | |
| 0.14112595, | |
| 0.11027293, | |
| 0.16430713, | |
| 0.07610069, | |
| 0.12522084, | |
| 0.12442842 | |
| ], | |
| "bytes_streamed": 494044864 | |
| }, | |
| { | |
| "layer_idx": 53, | |
| "layer_kind": "moe", | |
| "residual_norm_in": 3180.221, | |
| "residual_norm_out": 3144.0103, | |
| "expert_indices": [ | |
| 42, | |
| 101, | |
| 131, | |
| 247, | |
| 272, | |
| 341, | |
| 371, | |
| 383 | |
| ], | |
| "routing_weights": [ | |
| 0.061393943, | |
| 0.17122178, | |
| 0.14294015, | |
| 0.10139631, | |
| 0.05643083, | |
| 0.16060388, | |
| 0.14770873, | |
| 0.15830438 | |
| ], | |
| "bytes_streamed": 494044864 | |
| }, | |
| { | |
| "layer_idx": 54, | |
| "layer_kind": "moe", | |
| "residual_norm_in": 3144.0103, | |
| "residual_norm_out": 3032.5757, | |
| "expert_indices": [ | |
| 68, | |
| 156, | |
| 169, | |
| 189, | |
| 192, | |
| 218, | |
| 240, | |
| 319 | |
| ], | |
| "routing_weights": [ | |
| 0.053958643, | |
| 0.05449723, | |
| 0.14695822, | |
| 0.14635251, | |
| 0.15215138, | |
| 0.1491312, | |
| 0.14903401, | |
| 0.14791682 | |
| ], | |
| "bytes_streamed": 494044864 | |
| }, | |
| { | |
| "layer_idx": 55, | |
| "layer_kind": "moe", | |
| "residual_norm_in": 3032.5757, | |
| "residual_norm_out": 2966.738, | |
| "expert_indices": [ | |
| 42, | |
| 87, | |
| 173, | |
| 246, | |
| 308, | |
| 309, | |
| 329, | |
| 348 | |
| ], | |
| "routing_weights": [ | |
| 0.051514734, | |
| 0.14723632, | |
| 0.1606687, | |
| 0.16340463, | |
| 0.15019183, | |
| 0.15263279, | |
| 0.11705787, | |
| 0.05729307 | |
| ], | |
| "bytes_streamed": 494044864 | |
| }, | |
| { | |
| "layer_idx": 56, | |
| "layer_kind": "moe", | |
| "residual_norm_in": 2966.738, | |
| "residual_norm_out": 2920.522, | |
| "expert_indices": [ | |
| 26, | |
| 67, | |
| 74, | |
| 84, | |
| 253, | |
| 316, | |
| 319, | |
| 363 | |
| ], | |
| "routing_weights": [ | |
| 0.16366942, | |
| 0.16105412, | |
| 0.16047256, | |
| 0.0766057, | |
| 0.07692882, | |
| 0.16656853, | |
| 0.123046815, | |
| 0.071654126 | |
| ], | |
| "bytes_streamed": 494044864 | |
| }, | |
| { | |
| "layer_idx": 57, | |
| "layer_kind": "moe", | |
| "residual_norm_in": 2920.522, | |
| "residual_norm_out": 2818.567, | |
| "expert_indices": [ | |
| 0, | |
| 30, | |
| 80, | |
| 84, | |
| 107, | |
| 222, | |
| 253, | |
| 288 | |
| ], | |
| "routing_weights": [ | |
| 0.18186912, | |
| 0.05262169, | |
| 0.12891345, | |
| 0.054279856, | |
| 0.17406718, | |
| 0.18127596, | |
| 0.17727545, | |
| 0.049697243 | |
| ], | |
| "bytes_streamed": 494044864 | |
| }, | |
| { | |
| "layer_idx": 58, | |
| "layer_kind": "moe", | |
| "residual_norm_in": 2818.567, | |
| "residual_norm_out": 2593.77, | |
| "expert_indices": [ | |
| 5, | |
| 234, | |
| 259, | |
| 279, | |
| 329, | |
| 344, | |
| 350, | |
| 359 | |
| ], | |
| "routing_weights": [ | |
| 0.12612459, | |
| 0.1269469, | |
| 0.12518734, | |
| 0.12676597, | |
| 0.118010186, | |
| 0.12671871, | |
| 0.12414087, | |
| 0.12610547 | |
| ], | |
| "bytes_streamed": 494044864 | |
| }, | |
| { | |
| "layer_idx": 59, | |
| "layer_kind": "moe", | |
| "residual_norm_in": 2593.77, | |
| "residual_norm_out": 3548.4114, | |
| "expert_indices": [ | |
| 47, | |
| 62, | |
| 83, | |
| 156, | |
| 161, | |
| 176, | |
| 372, | |
| 374 | |
| ], | |
| "routing_weights": [ | |
| 0.09576418, | |
| 0.07162922, | |
| 0.06112332, | |
| 0.07196138, | |
| 0.062052805, | |
| 0.48133284, | |
| 0.08541494, | |
| 0.070721366 | |
| ], | |
| "bytes_streamed": 494044864 | |
| }, | |
| { | |
| "layer_idx": 60, | |
| "layer_kind": "moe", | |
| "residual_norm_in": 3548.4114, | |
| "residual_norm_out": 9128.322, | |
| "expert_indices": [ | |
| 3, | |
| 55, | |
| 68, | |
| 73, | |
| 217, | |
| 234, | |
| 242, | |
| 378 | |
| ], | |
| "routing_weights": [ | |
| 0.03411231, | |
| 0.039301656, | |
| 0.19902284, | |
| 0.18033594, | |
| 0.074356996, | |
| 0.23592533, | |
| 0.19573753, | |
| 0.04120746 | |
| ], | |
| "bytes_streamed": 494044864 | |
| } | |
| ] | |
| }, | |
| { | |
| "step_index": 1, | |
| "input_token_ids": [ | |
| 163584, | |
| 950 | |
| ], | |
| "output_token_id": 1268, | |
| "output_logit": 12.380399, | |
| "logits_digest": "60b7ec8a387ef2236293f735372638dd15b95b2a6d9785e9fc40199e786cb7ed", | |
| "hidden_digest": "b4e64fc50682575dadbace19bdbbe95a9840a9e014fa3e08c3a625141195056a", | |
| "top_k_tokens": [ | |
| [ | |
| 1268, | |
| 12.380399 | |
| ], | |
| [ | |
| 950, | |
| 12.275541 | |
| ], | |
| [ | |
| 220, | |
| 11.824542 | |
| ], | |
| [ | |
| 17159, | |
| 11.68036 | |
| ], | |
| [ | |
| 646, | |
| 11.485584 | |
| ], | |
| [ | |
| 27, | |
| 11.00096 | |
| ], | |
| [ | |
| 13890, | |
| 10.856464 | |
| ], | |
| [ | |
| 61858, | |
| 10.649042 | |
| ], | |
| [ | |
| 401, | |
| 10.223328 | |
| ], | |
| [ | |
| 4665, | |
| 10.183086 | |
| ], | |
| [ | |
| 30996, | |
| 10.116564 | |
| ], | |
| [ | |
| 347, | |
| 10.069124 | |
| ], | |
| [ | |
| 32360, | |
| 10.050613 | |
| ], | |
| [ | |
| 994, | |
| 9.81243 | |
| ], | |
| [ | |
| 414, | |
| 9.763189 | |
| ], | |
| [ | |
| 4698, | |
| 9.717009 | |
| ] | |
| ], | |
| "entropy": 7.6537275, | |
| "margin": 0.104857445, | |
| "bytes_streamed": 46780282960, | |
| "layers": [ | |
| { | |
| "layer_idx": 0, | |
| "layer_kind": "dense", | |
| "residual_norm_in": 1.254553, | |
| "residual_norm_out": 1.7512261, | |
| "expert_indices": null, | |
| "routing_weights": null, | |
| "bytes_streamed": 995000320 | |
| }, | |
| { | |
| "layer_idx": 1, | |
| "layer_kind": "moe", | |
| "residual_norm_in": 1.7512261, | |
| "residual_norm_out": 3.1337905, | |
| "expert_indices": [ | |
| 5, | |
| 33, | |
| 72, | |
| 105, | |
| 116, | |
| 154, | |
| 182, | |
| 272 | |
| ], | |
| "routing_weights": [ | |
| 0.087083995, | |
| 0.20157112, | |
| 0.14540455, | |
| 0.08845074, | |
| 0.1286529, | |
| 0.10869927, | |
| 0.12787405, | |
| 0.112263314 | |
| ], | |
| "bytes_streamed": 667453288 | |
| }, | |
| { | |
| "layer_idx": 2, | |
| "layer_kind": "moe", | |
| "residual_norm_in": 3.1337905, | |
| "residual_norm_out": 4.4556847, | |
| "expert_indices": [ | |
| 7, | |
| 25, | |
| 92, | |
| 179, | |
| 190, | |
| 240, | |
| 340, | |
| 343 | |
| ], | |
| "routing_weights": [ | |
| 0.1776416, | |
| 0.068654895, | |
| 0.1582383, | |
| 0.0638775, | |
| 0.12181032, | |
| 0.18464753, | |
| 0.060134597, | |
| 0.16499528 | |
| ], | |
| "bytes_streamed": 667453288 | |
| }, | |
| { | |
| "layer_idx": 3, | |
| "layer_kind": "moe", | |
| "residual_norm_in": 4.4556847, | |
| "residual_norm_out": 9.8798485, | |
| "expert_indices": [ | |
| 15, | |
| 189, | |
| 190, | |
| 217, | |
| 253, | |
| 271, | |
| 337, | |
| 353 | |
| ], | |
| "routing_weights": [ | |
| 0.05684734, | |
| 0.062119197, | |
| 0.052167017, | |
| 0.35753912, | |
| 0.06827293, | |
| 0.049369186, | |
| 0.04669538, | |
| 0.30698988 | |
| ], | |
| "bytes_streamed": 667453288 | |
| }, | |
| { | |
| "layer_idx": 4, | |
| "layer_kind": "moe", | |
| "residual_norm_in": 9.8798485, | |
| "residual_norm_out": 78.37475, | |
| "expert_indices": [ | |
| 52, | |
| 98, | |
| 106, | |
| 109, | |
| 120, | |
| 189, | |
| 191, | |
| 324 | |
| ], | |
| "routing_weights": [ | |
| 0.07070838, | |
| 0.08047035, | |
| 0.077183425, | |
| 0.08113832, | |
| 0.071013846, | |
| 0.47025713, | |
| 0.08182316, | |
| 0.067405246 | |
| ], | |
| "bytes_streamed": 692225920 | |
| }, | |
| { | |
| "layer_idx": 5, | |
| "layer_kind": "moe", | |
| "residual_norm_in": 78.37475, | |
| "residual_norm_out": 84.48327, | |
| "expert_indices": [ | |
| 12, | |
| 17, | |
| 121, | |
| 124, | |
| 160, | |
| 186, | |
| 195, | |
| 237 | |
| ], | |
| "routing_weights": [ | |
| 0.1888245, | |
| 0.10076529, | |
| 0.11140824, | |
| 0.17085913, | |
| 0.10019477, | |
| 0.10386934, | |
| 0.095480084, | |
| 0.12859854 | |
| ], | |
| "bytes_streamed": 692225920 | |
| }, | |
| { | |
| "layer_idx": 6, | |
| "layer_kind": "moe", | |
| "residual_norm_in": 84.48327, | |
| "residual_norm_out": 390.02982, | |
| "expert_indices": [ | |
| 18, | |
| 45, | |
| 88, | |
| 239, | |
| 243, | |
| 269, | |
| 334, | |
| 346 | |
| ], | |
| "routing_weights": [ | |
| 0.046305433, | |
| 0.029368749, | |
| 0.033812124, | |
| 0.031240184, | |
| 0.5246116, | |
| 0.0594912, | |
| 0.037579488, | |
| 0.23759124 | |
| ], | |
| "bytes_streamed": 667453288 | |
| }, | |
| { | |
| "layer_idx": 7, | |
| "layer_kind": "moe", | |
| "residual_norm_in": 390.02982, | |
| "residual_norm_out": 1106.5735, | |
| "expert_indices": [ | |
| 54, | |
| 57, | |
| 155, | |
| 183, | |
| 185, | |
| 192, | |
| 302, | |
| 361 | |
| ], | |
| "routing_weights": [ | |
| 0.025255695, | |
| 0.0061686267, | |
| 0.007244412, | |
| 0.92817426, | |
| 0.009681543, | |
| 0.0078814775, | |
| 0.0067851017, | |
| 0.008808859 | |
| ], | |
| "bytes_streamed": 692225920 | |
| }, | |
| { | |
| "layer_idx": 8, | |
| "layer_kind": "moe", | |
| "residual_norm_in": 1106.5735, | |
| "residual_norm_out": 1492.6306, | |
| "expert_indices": [ | |
| 86, | |
| 128, | |
| 175, | |
| 287, | |
| 311, | |
| 336, | |
| 337, | |
| 351 | |
| ], | |
| "routing_weights": [ | |
| 0.06048171, | |
| 0.030799579, | |
| 0.030457757, | |
| 0.038466815, | |
| 0.029505648, | |
| 0.032741938, | |
| 0.034052644, | |
| 0.743494 | |
| ], | |
| "bytes_streamed": 692225920 | |
| }, | |
| { | |
| "layer_idx": 9, | |
| "layer_kind": "moe", | |
| "residual_norm_in": 1492.6306, | |
| "residual_norm_out": 1690.9564, | |
| "expert_indices": [ | |
| 30, | |
| 131, | |
| 182, | |
| 183, | |
| 198, | |
| 238, | |
| 365, | |
| 377 | |
| ], | |
| "routing_weights": [ | |
| 0.056319162, | |
| 0.047798146, | |
| 0.056981582, | |
| 0.05191267, | |
| 0.04586145, | |
| 0.057513874, | |
| 0.045446966, | |
| 0.63816607 | |
| ], | |
| "bytes_streamed": 692225920 | |
| }, | |
| { | |
| "layer_idx": 10, | |
| "layer_kind": "moe", | |
| "residual_norm_in": 1690.9564, | |
| "residual_norm_out": 1954.9684, | |
| "expert_indices": [ | |
| 5, | |
| 34, | |
| 46, | |
| 118, | |
| 234, | |
| 240, | |
| 292, | |
| 296 | |
| ], | |
| "routing_weights": [ | |
| 0.033065867, | |
| 0.015240303, | |
| 0.01521005, | |
| 0.015278836, | |
| 0.014619413, | |
| 0.013330309, | |
| 0.017498873, | |
| 0.87575644 | |
| ], | |
| "bytes_streamed": 692225920 | |
| }, | |
| { | |
| "layer_idx": 11, | |
| "layer_kind": "moe", | |
| "residual_norm_in": 1954.9684, | |
| "residual_norm_out": 1956.8363, | |
| "expert_indices": [ | |
| 31, | |
| 124, | |
| 141, | |
| 180, | |
| 224, | |
| 237, | |
| 252, | |
| 253 | |
| ], | |
| "routing_weights": [ | |
| 0.13398919, | |
| 0.12286291, | |
| 0.119516574, | |
| 0.12470392, | |
| 0.13138652, | |
| 0.11974144, | |
| 0.12428722, | |
| 0.12351225 | |
| ], | |
| "bytes_streamed": 692225920 | |
| }, | |
| { | |
| "layer_idx": 12, | |
| "layer_kind": "moe", | |
| "residual_norm_in": 1956.8363, | |
| "residual_norm_out": 1980.637, | |
| "expert_indices": [ | |
| 25, | |
| 51, | |
| 92, | |
| 95, | |
| 139, | |
| 189, | |
| 270, | |
| 371 | |
| ], | |
| "routing_weights": [ | |
| 0.11804446, | |
| 0.12368671, | |
| 0.115437895, | |
| 0.12428875, | |
| 0.12710269, | |
| 0.13917556, | |
| 0.12618789, | |
| 0.12607609 | |
| ], | |
| "bytes_streamed": 692225920 | |
| }, | |
| { | |
| "layer_idx": 13, | |
| "layer_kind": "moe", | |
| "residual_norm_in": 1980.637, | |
| "residual_norm_out": 1982.7433, | |
| "expert_indices": [ | |
| 65, | |
| 68, | |
| 103, | |
| 194, | |
| 203, | |
| 228, | |
| 312, | |
| 348 | |
| ], | |
| "routing_weights": [ | |
| 0.15230049, | |
| 0.12651007, | |
| 0.12470709, | |
| 0.11999359, | |
| 0.12181889, | |
| 0.116762176, | |
| 0.12590373, | |
| 0.112003945 | |
| ], | |
| "bytes_streamed": 667453288 | |
| }, | |
| { | |
| "layer_idx": 14, | |
| "layer_kind": "moe", | |
| "residual_norm_in": 1982.7433, | |
| "residual_norm_out": 1985.3844, | |
| "expert_indices": [ | |
| 166, | |
| 182, | |
| 204, | |
| 207, | |
| 262, | |
| 269, | |
| 303, | |
| 377 | |
| ], | |
| "routing_weights": [ | |
| 0.12826353, | |
| 0.11211658, | |
| 0.16714138, | |
| 0.11411119, | |
| 0.12515004, | |
| 0.110577546, | |
| 0.118941896, | |
| 0.12369785 | |
| ], | |
| "bytes_streamed": 692225920 | |
| }, | |
| { | |
| "layer_idx": 15, | |
| "layer_kind": "moe", | |
| "residual_norm_in": 1985.3844, | |
| "residual_norm_out": 2258.2778, | |
| "expert_indices": [ | |
| 11, | |
| 109, | |
| 136, | |
| 150, | |
| 194, | |
| 212, | |
| 224, | |
| 279 | |
| ], | |
| "routing_weights": [ | |
| 0.07922835, | |
| 0.08260134, | |
| 0.08270256, | |
| 0.07873966, | |
| 0.08021491, | |
| 0.08108263, | |
| 0.07728814, | |
| 0.43814245 | |
| ], | |
| "bytes_streamed": 667453288 | |
| }, | |
| { | |
| "layer_idx": 16, | |
| "layer_kind": "moe", | |
| "residual_norm_in": 2258.2778, | |
| "residual_norm_out": 2261.854, | |
| "expert_indices": [ | |
| 19, | |
| 80, | |
| 89, | |
| 91, | |
| 204, | |
| 225, | |
| 249, | |
| 339 | |
| ], | |
| "routing_weights": [ | |
| 0.11132005, | |
| 0.12770416, | |
| 0.12904347, | |
| 0.123933, | |
| 0.13755625, | |
| 0.11228153, | |
| 0.14229253, | |
| 0.115868956 | |
| ], | |
| "bytes_streamed": 692225920 | |
| }, | |
| { | |
| "layer_idx": 17, | |
| "layer_kind": "moe", | |
| "residual_norm_in": 2261.854, | |
| "residual_norm_out": 2275.1008, | |
| "expert_indices": [ | |
| 81, | |
| 113, | |
| 128, | |
| 161, | |
| 282, | |
| 303, | |
| 304, | |
| 364 | |
| ], | |
| "routing_weights": [ | |
| 0.13470753, | |
| 0.12498505, | |
| 0.13510035, | |
| 0.13626955, | |
| 0.10787492, | |
| 0.10595762, | |
| 0.1165155, | |
| 0.13858946 | |
| ], | |
| "bytes_streamed": 692225920 | |
| }, | |
| { | |
| "layer_idx": 18, | |
| "layer_kind": "moe", | |
| "residual_norm_in": 2275.1008, | |
| "residual_norm_out": 2846.2073, | |
| "expert_indices": [ | |
| 30, | |
| 72, | |
| 95, | |
| 143, | |
| 230, | |
| 309, | |
| 329, | |
| 378 | |
| ], | |
| "routing_weights": [ | |
| 0.06173668, | |
| 0.060805954, | |
| 0.059534308, | |
| 0.546107, | |
| 0.05942752, | |
| 0.06282742, | |
| 0.059586395, | |
| 0.08997479 | |
| ], | |
| "bytes_streamed": 667453288 | |
| }, | |
| { | |
| "layer_idx": 19, | |
| "layer_kind": "moe", | |
| "residual_norm_in": 2846.2073, | |
| "residual_norm_out": 2848.6094, | |
| "expert_indices": [ | |
| 28, | |
| 62, | |
| 84, | |
| 210, | |
| 258, | |
| 264, | |
| 278, | |
| 318 | |
| ], | |
| "routing_weights": [ | |
| 0.12718764, | |
| 0.11587793, | |
| 0.116436146, | |
| 0.11349043, | |
| 0.12945548, | |
| 0.12185279, | |
| 0.15666124, | |
| 0.1190383 | |
| ], | |
| "bytes_streamed": 692225920 | |
| }, | |
| { | |
| "layer_idx": 20, | |
| "layer_kind": "moe", | |
| "residual_norm_in": 2848.6094, | |
| "residual_norm_out": 2964.6714, | |
| "expert_indices": [ | |
| 7, | |
| 108, | |
| 146, | |
| 227, | |
| 260, | |
| 273, | |
| 329, | |
| 351 | |
| ], | |
| "routing_weights": [ | |
| 0.120410115, | |
| 0.12045736, | |
| 0.11941096, | |
| 0.116442, | |
| 0.15163471, | |
| 0.122804046, | |
| 0.13331619, | |
| 0.11552454 | |
| ], | |
| "bytes_streamed": 692225920 | |
| }, | |
| { | |
| "layer_idx": 21, | |
| "layer_kind": "moe", | |
| "residual_norm_in": 2964.6714, | |
| "residual_norm_out": 2998.3845, | |
| "expert_indices": [ | |
| 8, | |
| 24, | |
| 140, | |
| 151, | |
| 250, | |
| 264, | |
| 313, | |
| 342 | |
| ], | |
| "routing_weights": [ | |
| 0.12219975, | |
| 0.11388808, | |
| 0.11746955, | |
| 0.11987377, | |
| 0.12484145, | |
| 0.122165635, | |
| 0.13231997, | |
| 0.14724179 | |
| ], | |
| "bytes_streamed": 692225920 | |
| }, | |
| { | |
| "layer_idx": 22, | |
| "layer_kind": "moe", | |
| "residual_norm_in": 2998.3845, | |
| "residual_norm_out": 3000.9204, | |
| "expert_indices": [ | |
| 3, | |
| 11, | |
| 78, | |
| 119, | |
| 206, | |
| 236, | |
| 249, | |
| 334 | |
| ], | |
| "routing_weights": [ | |
| 0.11585083, | |
| 0.12601939, | |
| 0.12537284, | |
| 0.12319891, | |
| 0.13266194, | |
| 0.11640788, | |
| 0.12136969, | |
| 0.13911851 | |
| ], | |
| "bytes_streamed": 692225920 | |
| }, | |
| { | |
| "layer_idx": 23, | |
| "layer_kind": "moe", | |
| "residual_norm_in": 3000.9204, | |
| "residual_norm_out": 3124.0317, | |
| "expert_indices": [ | |
| 8, | |
| 44, | |
| 125, | |
| 207, | |
| 209, | |
| 226, | |
| 283, | |
| 325 | |
| ], | |
| "routing_weights": [ | |
| 0.1277313, | |
| 0.11835284, | |
| 0.13288988, | |
| 0.119071715, | |
| 0.117436275, | |
| 0.12930648, | |
| 0.11873839, | |
| 0.13647313 | |
| ], | |
| "bytes_streamed": 692225920 | |
| }, | |
| { | |
| "layer_idx": 24, | |
| "layer_kind": "moe", | |
| "residual_norm_in": 3124.0317, | |
| "residual_norm_out": 3263.8574, | |
| "expert_indices": [ | |
| 11, | |
| 46, | |
| 58, | |
| 59, | |
| 89, | |
| 223, | |
| 294, | |
| 376 | |
| ], | |
| "routing_weights": [ | |
| 0.13790713, | |
| 0.12226996, | |
| 0.12022834, | |
| 0.13080284, | |
| 0.11101902, | |
| 0.122765206, | |
| 0.1395905, | |
| 0.11541706 | |
| ], | |
| "bytes_streamed": 692225920 | |
| }, | |
| { | |
| "layer_idx": 25, | |
| "layer_kind": "moe", | |
| "residual_norm_in": 3263.8574, | |
| "residual_norm_out": 3268.826, | |
| "expert_indices": [ | |
| 32, | |
| 52, | |
| 186, | |
| 275, | |
| 292, | |
| 317, | |
| 325, | |
| 355 | |
| ], | |
| "routing_weights": [ | |
| 0.1295867, | |
| 0.1348981, | |
| 0.11795012, | |
| 0.12863119, | |
| 0.12783831, | |
| 0.12071772, | |
| 0.11842303, | |
| 0.12195492 | |
| ], | |
| "bytes_streamed": 692225920 | |
| }, | |
| { | |
| "layer_idx": 26, | |
| "layer_kind": "moe", | |
| "residual_norm_in": 3268.826, | |
| "residual_norm_out": 3271.8638, | |
| "expert_indices": [ | |
| 28, | |
| 54, | |
| 109, | |
| 143, | |
| 164, | |
| 187, | |
| 234, | |
| 326 | |
| ], | |
| "routing_weights": [ | |
| 0.12403761, | |
| 0.124741495, | |
| 0.1233652, | |
| 0.12436722, | |
| 0.12376981, | |
| 0.13259424, | |
| 0.12488026, | |
| 0.12224412 | |
| ], | |
| "bytes_streamed": 667453288 | |
| }, | |
| { | |
| "layer_idx": 27, | |
| "layer_kind": "moe", | |
| "residual_norm_in": 3271.8638, | |
| "residual_norm_out": 3274.3186, | |
| "expert_indices": [ | |
| 5, | |
| 22, | |
| 23, | |
| 48, | |
| 115, | |
| 160, | |
| 188, | |
| 218 | |
| ], | |
| "routing_weights": [ | |
| 0.12905355, | |
| 0.11797063, | |
| 0.123731114, | |
| 0.11916181, | |
| 0.119591415, | |
| 0.12297361, | |
| 0.13747871, | |
| 0.13003916 | |
| ], | |
| "bytes_streamed": 667453288 | |
| }, | |
| { | |
| "layer_idx": 28, | |
| "layer_kind": "moe", | |
| "residual_norm_in": 3274.3186, | |
| "residual_norm_out": 3275.987, | |
| "expert_indices": [ | |
| 83, | |
| 117, | |
| 178, | |
| 218, | |
| 233, | |
| 260, | |
| 305, | |
| 342 | |
| ], | |
| "routing_weights": [ | |
| 0.12889512, | |
| 0.12080206, | |
| 0.12741528, | |
| 0.119030036, | |
| 0.12386302, | |
| 0.12633918, | |
| 0.13027583, | |
| 0.12337941 | |
| ], | |
| "bytes_streamed": 692225920 | |
| }, | |
| { | |
| "layer_idx": 29, | |
| "layer_kind": "moe", | |
| "residual_norm_in": 3275.987, | |
| "residual_norm_out": 3280.4622, | |
| "expert_indices": [ | |
| 19, | |
| 132, | |
| 145, | |
| 213, | |
| 264, | |
| 275, | |
| 309, | |
| 374 | |
| ], | |
| "routing_weights": [ | |
| 0.1201872, | |
| 0.11707843, | |
| 0.13667186, | |
| 0.11928836, | |
| 0.11894964, | |
| 0.14129789, | |
| 0.13095954, | |
| 0.115567006 | |
| ], | |
| "bytes_streamed": 667453288 | |
| }, | |
| { | |
| "layer_idx": 30, | |
| "layer_kind": "moe", | |
| "residual_norm_in": 3280.4622, | |
| "residual_norm_out": 3280.6106, | |
| "expert_indices": [ | |
| 55, | |
| 109, | |
| 165, | |
| 246, | |
| 254, | |
| 264, | |
| 275, | |
| 282 | |
| ], | |
| "routing_weights": [ | |
| 0.12114447, | |
| 0.1265742, | |
| 0.12592877, | |
| 0.13100544, | |
| 0.12474787, | |
| 0.1230107, | |
| 0.12573223, | |
| 0.12185628 | |
| ], | |
| "bytes_streamed": 692225920 | |
| }, | |
| { | |
| "layer_idx": 31, | |
| "layer_kind": "moe", | |
| "residual_norm_in": 3280.6106, | |
| "residual_norm_out": 3283.9314, | |
| "expert_indices": [ | |
| 45, | |
| 150, | |
| 196, | |
| 198, | |
| 243, | |
| 323, | |
| 333, | |
| 359 | |
| ], | |
| "routing_weights": [ | |
| 0.124348536, | |
| 0.12571238, | |
| 0.121641114, | |
| 0.12634422, | |
| 0.12098488, | |
| 0.12598802, | |
| 0.12444443, | |
| 0.13053642 | |
| ], | |
| "bytes_streamed": 667453288 | |
| }, | |
| { | |
| "layer_idx": 32, | |
| "layer_kind": "moe", | |
| "residual_norm_in": 3283.9314, | |
| "residual_norm_out": 3285.4775, | |
| "expert_indices": [ | |
| 21, | |
| 76, | |
| 182, | |
| 270, | |
| 276, | |
| 286, | |
| 336, | |
| 345 | |
| ], | |
| "routing_weights": [ | |
| 0.123925604, | |
| 0.12669829, | |
| 0.12782235, | |
| 0.12475172, | |
| 0.12783591, | |
| 0.12554578, | |
| 0.12194816, | |
| 0.121472225 | |
| ], | |
| "bytes_streamed": 667453288 | |
| }, | |
| { | |
| "layer_idx": 33, | |
| "layer_kind": "moe", | |
| "residual_norm_in": 3285.4775, | |
| "residual_norm_out": 3284.7793, | |
| "expert_indices": [ | |
| 9, | |
| 53, | |
| 54, | |
| 57, | |
| 267, | |
| 281, | |
| 303, | |
| 309 | |
| ], | |
| "routing_weights": [ | |
| 0.12560827, | |
| 0.12389935, | |
| 0.12423008, | |
| 0.12410613, | |
| 0.12650695, | |
| 0.12478609, | |
| 0.12493785, | |
| 0.12592524 | |
| ], | |
| "bytes_streamed": 692225920 | |
| }, | |
| { | |
| "layer_idx": 34, | |
| "layer_kind": "moe", | |
| "residual_norm_in": 3284.7793, | |
| "residual_norm_out": 3284.764, | |
| "expert_indices": [ | |
| 54, | |
| 96, | |
| 152, | |
| 206, | |
| 250, | |
| 295, | |
| 325, | |
| 339 | |
| ], | |
| "routing_weights": [ | |
| 0.12830864, | |
| 0.12542318, | |
| 0.12211993, | |
| 0.123667784, | |
| 0.12476751, | |
| 0.12551583, | |
| 0.123447806, | |
| 0.12674934 | |
| ], | |
| "bytes_streamed": 692225920 | |
| }, | |
| { | |
| "layer_idx": 35, | |
| "layer_kind": "moe", | |
| "residual_norm_in": 3284.764, | |
| "residual_norm_out": 3285.0469, | |
| "expert_indices": [ | |
| 9, | |
| 68, | |
| 86, | |
| 142, | |
| 162, | |
| 176, | |
| 257, | |
| 326 | |
| ], | |
| "routing_weights": [ | |
| 0.12454217, | |
| 0.12767826, | |
| 0.124468386, | |
| 0.12491048, | |
| 0.12380852, | |
| 0.12711051, | |
| 0.122607514, | |
| 0.12487409 | |
| ], | |
| "bytes_streamed": 692225920 | |
| }, | |
| { | |
| "layer_idx": 36, | |
| "layer_kind": "moe", | |
| "residual_norm_in": 3285.0469, | |
| "residual_norm_out": 3283.272, | |
| "expert_indices": [ | |
| 51, | |
| 76, | |
| 132, | |
| 200, | |
| 236, | |
| 276, | |
| 317, | |
| 383 | |
| ], | |
| "routing_weights": [ | |
| 0.12545687, | |
| 0.12348959, | |
| 0.12646747, | |
| 0.12358548, | |
| 0.12351758, | |
| 0.123819746, | |
| 0.1276341, | |
| 0.12602918 | |
| ], | |
| "bytes_streamed": 667453288 | |
| }, | |
| { | |
| "layer_idx": 37, | |
| "layer_kind": "moe", | |
| "residual_norm_in": 3283.272, | |
| "residual_norm_out": 3278.7158, | |
| "expert_indices": [ | |
| 22, | |
| 75, | |
| 156, | |
| 182, | |
| 187, | |
| 189, | |
| 230, | |
| 321 | |
| ], | |
| "routing_weights": [ | |
| 0.12714443, | |
| 0.12565663, | |
| 0.12507114, | |
| 0.12419243, | |
| 0.12688793, | |
| 0.123175636, | |
| 0.12531842, | |
| 0.122553304 | |
| ], | |
| "bytes_streamed": 692225920 | |
| }, | |
| { | |
| "layer_idx": 38, | |
| "layer_kind": "moe", | |
| "residual_norm_in": 3278.7158, | |
| "residual_norm_out": 3278.0444, | |
| "expert_indices": [ | |
| 116, | |
| 118, | |
| 141, | |
| 142, | |
| 165, | |
| 246, | |
| 296, | |
| 305 | |
| ], | |
| "routing_weights": [ | |
| 0.12589957, | |
| 0.12569398, | |
| 0.12418391, | |
| 0.12590075, | |
| 0.12354491, | |
| 0.12647879, | |
| 0.123581976, | |
| 0.1247161 | |
| ], | |
| "bytes_streamed": 692225920 | |
| }, | |
| { | |
| "layer_idx": 39, | |
| "layer_kind": "moe", | |
| "residual_norm_in": 3278.0444, | |
| "residual_norm_out": 3277.1926, | |
| "expert_indices": [ | |
| 1, | |
| 21, | |
| 53, | |
| 74, | |
| 153, | |
| 232, | |
| 278, | |
| 381 | |
| ], | |
| "routing_weights": [ | |
| 0.12392103, | |
| 0.12356896, | |
| 0.12509237, | |
| 0.12476749, | |
| 0.1261542, | |
| 0.1256581, | |
| 0.12628376, | |
| 0.124554105 | |
| ], | |
| "bytes_streamed": 692225920 | |
| }, | |
| { | |
| "layer_idx": 40, | |
| "layer_kind": "moe", | |
| "residual_norm_in": 3277.1926, | |
| "residual_norm_out": 3272.9072, | |
| "expert_indices": [ | |
| 14, | |
| 54, | |
| 80, | |
| 205, | |
| 243, | |
| 350, | |
| 360, | |
| 378 | |
| ], | |
| "routing_weights": [ | |
| 0.123624906, | |
| 0.12560081, | |
| 0.12601347, | |
| 0.12464201, | |
| 0.124437995, | |
| 0.124908455, | |
| 0.1261274, | |
| 0.124645 | |
| ], | |
| "bytes_streamed": 667453288 | |
| }, | |
| { | |
| "layer_idx": 41, | |
| "layer_kind": "moe", | |
| "residual_norm_in": 3272.9072, | |
| "residual_norm_out": 3269.7615, | |
| "expert_indices": [ | |
| 0, | |
| 12, | |
| 102, | |
| 118, | |
| 143, | |
| 323, | |
| 346, | |
| 361 | |
| ], | |
| "routing_weights": [ | |
| 0.12718984, | |
| 0.12352078, | |
| 0.124371946, | |
| 0.1250337, | |
| 0.12582457, | |
| 0.12378337, | |
| 0.1258188, | |
| 0.12445698 | |
| ], | |
| "bytes_streamed": 692225920 | |
| }, | |
| { | |
| "layer_idx": 42, | |
| "layer_kind": "moe", | |
| "residual_norm_in": 3269.7615, | |
| "residual_norm_out": 3263.612, | |
| "expert_indices": [ | |
| 11, | |
| 63, | |
| 66, | |
| 121, | |
| 233, | |
| 282, | |
| 286, | |
| 322 | |
| ], | |
| "routing_weights": [ | |
| 0.12231816, | |
| 0.12572178, | |
| 0.12679413, | |
| 0.12391465, | |
| 0.12539703, | |
| 0.122256, | |
| 0.12437119, | |
| 0.12922701 | |
| ], | |
| "bytes_streamed": 692225920 | |
| }, | |
| { | |
| "layer_idx": 43, | |
| "layer_kind": "moe", | |
| "residual_norm_in": 3263.612, | |
| "residual_norm_out": 3259.6252, | |
| "expert_indices": [ | |
| 48, | |
| 97, | |
| 111, | |
| 145, | |
| 166, | |
| 258, | |
| 275, | |
| 310 | |
| ], | |
| "routing_weights": [ | |
| 0.12841284, | |
| 0.12592939, | |
| 0.1230621, | |
| 0.11106126, | |
| 0.13067316, | |
| 0.12302471, | |
| 0.12871376, | |
| 0.1291227 | |
| ], | |
| "bytes_streamed": 692225920 | |
| }, | |
| { | |
| "layer_idx": 44, | |
| "layer_kind": "moe", | |
| "residual_norm_in": 3259.6252, | |
| "residual_norm_out": 3251.1978, | |
| "expert_indices": [ | |
| 52, | |
| 80, | |
| 135, | |
| 143, | |
| 145, | |
| 204, | |
| 209, | |
| 313 | |
| ], | |
| "routing_weights": [ | |
| 0.1505961, | |
| 0.15023981, | |
| 0.11775557, | |
| 0.087679155, | |
| 0.12829639, | |
| 0.094963625, | |
| 0.11943085, | |
| 0.15103848 | |
| ], | |
| "bytes_streamed": 692225920 | |
| }, | |
| { | |
| "layer_idx": 45, | |
| "layer_kind": "moe", | |
| "residual_norm_in": 3251.1978, | |
| "residual_norm_out": 3248.4263, | |
| "expert_indices": [ | |
| 20, | |
| 45, | |
| 57, | |
| 59, | |
| 105, | |
| 176, | |
| 271, | |
| 320 | |
| ], | |
| "routing_weights": [ | |
| 0.12892541, | |
| 0.11697536, | |
| 0.14063151, | |
| 0.12651275, | |
| 0.13517754, | |
| 0.09021728, | |
| 0.13699076, | |
| 0.12456939 | |
| ], | |
| "bytes_streamed": 692225920 | |
| }, | |
| { | |
| "layer_idx": 46, | |
| "layer_kind": "moe", | |
| "residual_norm_in": 3248.4263, | |
| "residual_norm_out": 3243.3252, | |
| "expert_indices": [ | |
| 7, | |
| 13, | |
| 31, | |
| 58, | |
| 198, | |
| 206, | |
| 222, | |
| 361 | |
| ], | |
| "routing_weights": [ | |
| 0.13090932, | |
| 0.13061666, | |
| 0.119749196, | |
| 0.13284738, | |
| 0.13148266, | |
| 0.12531951, | |
| 0.102036834, | |
| 0.1270384 | |
| ], | |
| "bytes_streamed": 692225920 | |
| }, | |
| { | |
| "layer_idx": 47, | |
| "layer_kind": "moe", | |
| "residual_norm_in": 3243.3252, | |
| "residual_norm_out": 3241.226, | |
| "expert_indices": [ | |
| 50, | |
| 91, | |
| 248, | |
| 268, | |
| 274, | |
| 305, | |
| 359, | |
| 377 | |
| ], | |
| "routing_weights": [ | |
| 0.090512484, | |
| 0.12572679, | |
| 0.12560984, | |
| 0.13680649, | |
| 0.13461365, | |
| 0.1383419, | |
| 0.11906946, | |
| 0.12931941 | |
| ], | |
| "bytes_streamed": 692225920 | |
| }, | |
| { | |
| "layer_idx": 48, | |
| "layer_kind": "moe", | |
| "residual_norm_in": 3241.226, | |
| "residual_norm_out": 3233.8818, | |
| "expert_indices": [ | |
| 21, | |
| 140, | |
| 180, | |
| 211, | |
| 216, | |
| 265, | |
| 275, | |
| 356 | |
| ], | |
| "routing_weights": [ | |
| 0.089280955, | |
| 0.11947252, | |
| 0.14138731, | |
| 0.13919227, | |
| 0.13082214, | |
| 0.12825514, | |
| 0.12513566, | |
| 0.1264539 | |
| ], | |
| "bytes_streamed": 692225920 | |
| }, | |
| { | |
| "layer_idx": 49, | |
| "layer_kind": "moe", | |
| "residual_norm_in": 3233.8818, | |
| "residual_norm_out": 3221.8186, | |
| "expert_indices": [ | |
| 65, | |
| 132, | |
| 136, | |
| 141, | |
| 257, | |
| 330, | |
| 331, | |
| 381 | |
| ], | |
| "routing_weights": [ | |
| 0.15157294, | |
| 0.14870882, | |
| 0.12575087, | |
| 0.14817211, | |
| 0.12935574, | |
| 0.12963831, | |
| 0.084066786, | |
| 0.082734436 | |
| ], | |
| "bytes_streamed": 692225920 | |
| }, | |
| { | |
| "layer_idx": 50, | |
| "layer_kind": "moe", | |
| "residual_norm_in": 3221.8186, | |
| "residual_norm_out": 3209.8303, | |
| "expert_indices": [ | |
| 2, | |
| 46, | |
| 77, | |
| 114, | |
| 160, | |
| 162, | |
| 175, | |
| 324 | |
| ], | |
| "routing_weights": [ | |
| 0.07168002, | |
| 0.13420905, | |
| 0.14428033, | |
| 0.13734862, | |
| 0.14445138, | |
| 0.13382252, | |
| 0.11261744, | |
| 0.121590614 | |
| ], | |
| "bytes_streamed": 667453288 | |
| }, | |
| { | |
| "layer_idx": 51, | |
| "layer_kind": "moe", | |
| "residual_norm_in": 3209.8303, | |
| "residual_norm_out": 3200.2834, | |
| "expert_indices": [ | |
| 36, | |
| 130, | |
| 189, | |
| 218, | |
| 263, | |
| 271, | |
| 317, | |
| 362 | |
| ], | |
| "routing_weights": [ | |
| 0.07808468, | |
| 0.14101124, | |
| 0.13042147, | |
| 0.11585922, | |
| 0.13718921, | |
| 0.12932168, | |
| 0.12141402, | |
| 0.1466985 | |
| ], | |
| "bytes_streamed": 667453288 | |
| }, | |
| { | |
| "layer_idx": 52, | |
| "layer_kind": "moe", | |
| "residual_norm_in": 3200.2834, | |
| "residual_norm_out": 3182.2012, | |
| "expert_indices": [ | |
| 10, | |
| 31, | |
| 40, | |
| 49, | |
| 52, | |
| 96, | |
| 259, | |
| 342 | |
| ], | |
| "routing_weights": [ | |
| 0.13195448, | |
| 0.12658952, | |
| 0.14112595, | |
| 0.11027296, | |
| 0.16430713, | |
| 0.07610071, | |
| 0.12522088, | |
| 0.12442839 | |
| ], | |
| "bytes_streamed": 667453288 | |
| }, | |
| { | |
| "layer_idx": 53, | |
| "layer_kind": "moe", | |
| "residual_norm_in": 3182.2012, | |
| "residual_norm_out": 3145.9927, | |
| "expert_indices": [ | |
| 42, | |
| 101, | |
| 131, | |
| 247, | |
| 272, | |
| 341, | |
| 371, | |
| 383 | |
| ], | |
| "routing_weights": [ | |
| 0.06139395, | |
| 0.17122178, | |
| 0.14294012, | |
| 0.10139634, | |
| 0.056430835, | |
| 0.16060388, | |
| 0.14770874, | |
| 0.15830438 | |
| ], | |
| "bytes_streamed": 692225920 | |
| }, | |
| { | |
| "layer_idx": 54, | |
| "layer_kind": "moe", | |
| "residual_norm_in": 3145.9927, | |
| "residual_norm_out": 3034.8306, | |
| "expert_indices": [ | |
| 68, | |
| 156, | |
| 169, | |
| 189, | |
| 192, | |
| 218, | |
| 240, | |
| 319 | |
| ], | |
| "routing_weights": [ | |
| 0.053958632, | |
| 0.05449725, | |
| 0.14695823, | |
| 0.14635251, | |
| 0.15215138, | |
| 0.1491312, | |
| 0.14903401, | |
| 0.14791681 | |
| ], | |
| "bytes_streamed": 692225920 | |
| }, | |
| { | |
| "layer_idx": 55, | |
| "layer_kind": "moe", | |
| "residual_norm_in": 3034.8306, | |
| "residual_norm_out": 2969.1238, | |
| "expert_indices": [ | |
| 42, | |
| 87, | |
| 173, | |
| 246, | |
| 308, | |
| 309, | |
| 329, | |
| 348 | |
| ], | |
| "routing_weights": [ | |
| 0.05151474, | |
| 0.1472363, | |
| 0.1606687, | |
| 0.16340461, | |
| 0.15019181, | |
| 0.1526328, | |
| 0.11705792, | |
| 0.05729307 | |
| ], | |
| "bytes_streamed": 692225920 | |
| }, | |
| { | |
| "layer_idx": 56, | |
| "layer_kind": "moe", | |
| "residual_norm_in": 2969.1238, | |
| "residual_norm_out": 2923.2224, | |
| "expert_indices": [ | |
| 26, | |
| 67, | |
| 74, | |
| 84, | |
| 253, | |
| 316, | |
| 319, | |
| 363 | |
| ], | |
| "routing_weights": [ | |
| 0.16366942, | |
| 0.16105407, | |
| 0.16047254, | |
| 0.07660575, | |
| 0.07692882, | |
| 0.16656852, | |
| 0.123046815, | |
| 0.07165412 | |
| ], | |
| "bytes_streamed": 692225920 | |
| }, | |
| { | |
| "layer_idx": 57, | |
| "layer_kind": "moe", | |
| "residual_norm_in": 2923.2224, | |
| "residual_norm_out": 2821.4446, | |
| "expert_indices": [ | |
| 0, | |
| 30, | |
| 80, | |
| 84, | |
| 107, | |
| 222, | |
| 253, | |
| 288 | |
| ], | |
| "routing_weights": [ | |
| 0.18186912, | |
| 0.052621696, | |
| 0.12891348, | |
| 0.05427989, | |
| 0.1740672, | |
| 0.18127596, | |
| 0.17727546, | |
| 0.04969726 | |
| ], | |
| "bytes_streamed": 692225920 | |
| }, | |
| { | |
| "layer_idx": 58, | |
| "layer_kind": "moe", | |
| "residual_norm_in": 2821.4446, | |
| "residual_norm_out": 2597.2996, | |
| "expert_indices": [ | |
| 5, | |
| 234, | |
| 259, | |
| 279, | |
| 329, | |
| 344, | |
| 350, | |
| 359 | |
| ], | |
| "routing_weights": [ | |
| 0.12612459, | |
| 0.1269469, | |
| 0.12518732, | |
| 0.12676597, | |
| 0.1180102, | |
| 0.12671871, | |
| 0.12414087, | |
| 0.12610547 | |
| ], | |
| "bytes_streamed": 692225920 | |
| }, | |
| { | |
| "layer_idx": 59, | |
| "layer_kind": "moe", | |
| "residual_norm_in": 2597.2996, | |
| "residual_norm_out": 3552.2322, | |
| "expert_indices": [ | |
| 47, | |
| 62, | |
| 83, | |
| 156, | |
| 161, | |
| 176, | |
| 372, | |
| 374 | |
| ], | |
| "routing_weights": [ | |
| 0.09576432, | |
| 0.07162925, | |
| 0.061123345, | |
| 0.07196141, | |
| 0.062052693, | |
| 0.48133266, | |
| 0.085414946, | |
| 0.07072137 | |
| ], | |
| "bytes_streamed": 667453288 | |
| }, | |
| { | |
| "layer_idx": 60, | |
| "layer_kind": "moe", | |
| "residual_norm_in": 3552.2322, | |
| "residual_norm_out": 9132.588, | |
| "expert_indices": [ | |
| 3, | |
| 55, | |
| 68, | |
| 73, | |
| 217, | |
| 234, | |
| 242, | |
| 378 | |
| ], | |
| "routing_weights": [ | |
| 0.03411237, | |
| 0.039301682, | |
| 0.1990229, | |
| 0.18033567, | |
| 0.074357, | |
| 0.2359254, | |
| 0.19573748, | |
| 0.041207485 | |
| ], | |
| "bytes_streamed": 692225920 | |
| } | |
| ] | |
| }, | |
| { | |
| "step_index": 2, | |
| "input_token_ids": [ | |
| 163584, | |
| 950, | |
| 1268 | |
| ], | |
| "output_token_id": 426, | |
| "output_logit": 17.376055, | |
| "logits_digest": "cb476312b2c1c8177acca248bd9f169f122ab9542c4441b947b5c6ccf79d3b7c", | |
| "hidden_digest": "15f36305064a88a485385bde0b3c24ce48079c4ad1b00cc138aac8d05a9937dc", | |
| "top_k_tokens": [ | |
| [ | |
| 426, | |
| 17.376055 | |
| ], | |
| [ | |
| 1952, | |
| 16.618616 | |
| ], | |
| [ | |
| 8224, | |
| 13.956974 | |
| ], | |
| [ | |
| 29255, | |
| 13.74489 | |
| ], | |
| [ | |
| 100948, | |
| 13.390081 | |
| ], | |
| [ | |
| 950, | |
| 13.121993 | |
| ], | |
| [ | |
| 3146, | |
| 13.002652 | |
| ], | |
| [ | |
| 40591, | |
| 12.934351 | |
| ], | |
| [ | |
| 686, | |
| 12.907539 | |
| ], | |
| [ | |
| 16, | |
| 12.812391 | |
| ], | |
| [ | |
| 67, | |
| 12.530942 | |
| ], | |
| [ | |
| 271, | |
| 12.454943 | |
| ], | |
| [ | |
| 30162, | |
| 12.349958 | |
| ], | |
| [ | |
| 68, | |
| 12.280102 | |
| ], | |
| [ | |
| 7649, | |
| 12.252055 | |
| ], | |
| [ | |
| 382, | |
| 12.226152 | |
| ] | |
| ], | |
| "entropy": 3.3380015, | |
| "margin": 0.75743866, | |
| "bytes_streamed": 55376386264, | |
| "layers": [ | |
| { | |
| "layer_idx": 0, | |
| "layer_kind": "dense", | |
| "residual_norm_in": 1.3763882, | |
| "residual_norm_out": 1.8894544, | |
| "expert_indices": null, | |
| "routing_weights": null, | |
| "bytes_streamed": 995000320 | |
| }, | |
| { | |
| "layer_idx": 1, | |
| "layer_kind": "moe", | |
| "residual_norm_in": 1.8894544, | |
| "residual_norm_out": 3.2712896, | |
| "expert_indices": [ | |
| 5, | |
| 33, | |
| 72, | |
| 105, | |
| 116, | |
| 154, | |
| 182, | |
| 272 | |
| ], | |
| "routing_weights": [ | |
| 0.087083995, | |
| 0.20157112, | |
| 0.14540455, | |
| 0.08845074, | |
| 0.1286529, | |
| 0.10869927, | |
| 0.12787405, | |
| 0.112263314 | |
| ], | |
| "bytes_streamed": 766543816 | |
| }, | |
| { | |
| "layer_idx": 2, | |
| "layer_kind": "moe", | |
| "residual_norm_in": 3.2712896, | |
| "residual_norm_out": 4.6317177, | |
| "expert_indices": [ | |
| 7, | |
| 25, | |
| 92, | |
| 179, | |
| 190, | |
| 240, | |
| 340, | |
| 343 | |
| ], | |
| "routing_weights": [ | |
| 0.1776416, | |
| 0.068654895, | |
| 0.1582383, | |
| 0.0638775, | |
| 0.12181032, | |
| 0.18464753, | |
| 0.060134597, | |
| 0.16499528 | |
| ], | |
| "bytes_streamed": 766543816 | |
| }, | |
| { | |
| "layer_idx": 3, | |
| "layer_kind": "moe", | |
| "residual_norm_in": 4.6317177, | |
| "residual_norm_out": 10.012449, | |
| "expert_indices": [ | |
| 15, | |
| 189, | |
| 190, | |
| 217, | |
| 253, | |
| 271, | |
| 337, | |
| 353 | |
| ], | |
| "routing_weights": [ | |
| 0.05684734, | |
| 0.062119197, | |
| 0.052167017, | |
| 0.35753912, | |
| 0.06827293, | |
| 0.049369186, | |
| 0.04669538, | |
| 0.30698988 | |
| ], | |
| "bytes_streamed": 816089080 | |
| }, | |
| { | |
| "layer_idx": 4, | |
| "layer_kind": "moe", | |
| "residual_norm_in": 10.012449, | |
| "residual_norm_out": 78.407, | |
| "expert_indices": [ | |
| 52, | |
| 98, | |
| 106, | |
| 109, | |
| 120, | |
| 189, | |
| 191, | |
| 324 | |
| ], | |
| "routing_weights": [ | |
| 0.07070838, | |
| 0.08047035, | |
| 0.077183425, | |
| 0.08113832, | |
| 0.071013846, | |
| 0.47025713, | |
| 0.08182316, | |
| 0.067405246 | |
| ], | |
| "bytes_streamed": 840861712 | |
| }, | |
| { | |
| "layer_idx": 5, | |
| "layer_kind": "moe", | |
| "residual_norm_in": 78.407, | |
| "residual_norm_out": 84.53675, | |
| "expert_indices": [ | |
| 12, | |
| 17, | |
| 121, | |
| 124, | |
| 160, | |
| 186, | |
| 195, | |
| 237 | |
| ], | |
| "routing_weights": [ | |
| 0.1888245, | |
| 0.10076529, | |
| 0.11140824, | |
| 0.17085913, | |
| 0.10019477, | |
| 0.10386934, | |
| 0.095480084, | |
| 0.12859854 | |
| ], | |
| "bytes_streamed": 791316448 | |
| }, | |
| { | |
| "layer_idx": 6, | |
| "layer_kind": "moe", | |
| "residual_norm_in": 84.53675, | |
| "residual_norm_out": 390.0494, | |
| "expert_indices": [ | |
| 18, | |
| 45, | |
| 88, | |
| 239, | |
| 243, | |
| 269, | |
| 334, | |
| 346 | |
| ], | |
| "routing_weights": [ | |
| 0.046305433, | |
| 0.029368749, | |
| 0.033812124, | |
| 0.031240184, | |
| 0.5246116, | |
| 0.0594912, | |
| 0.037579488, | |
| 0.23759124 | |
| ], | |
| "bytes_streamed": 840861712 | |
| }, | |
| { | |
| "layer_idx": 7, | |
| "layer_kind": "moe", | |
| "residual_norm_in": 390.0494, | |
| "residual_norm_out": 1106.5815, | |
| "expert_indices": [ | |
| 54, | |
| 57, | |
| 155, | |
| 183, | |
| 185, | |
| 192, | |
| 302, | |
| 361 | |
| ], | |
| "routing_weights": [ | |
| 0.025255695, | |
| 0.0061686267, | |
| 0.007244412, | |
| 0.92817426, | |
| 0.009681543, | |
| 0.0078814775, | |
| 0.0067851017, | |
| 0.008808859 | |
| ], | |
| "bytes_streamed": 791316448 | |
| }, | |
| { | |
| "layer_idx": 8, | |
| "layer_kind": "moe", | |
| "residual_norm_in": 1106.5815, | |
| "residual_norm_out": 1492.6401, | |
| "expert_indices": [ | |
| 86, | |
| 128, | |
| 175, | |
| 287, | |
| 311, | |
| 336, | |
| 337, | |
| 351 | |
| ], | |
| "routing_weights": [ | |
| 0.06048171, | |
| 0.030799579, | |
| 0.030457757, | |
| 0.038466815, | |
| 0.029505648, | |
| 0.032741938, | |
| 0.034052644, | |
| 0.743494 | |
| ], | |
| "bytes_streamed": 816089080 | |
| }, | |
| { | |
| "layer_idx": 9, | |
| "layer_kind": "moe", | |
| "residual_norm_in": 1492.6401, | |
| "residual_norm_out": 1690.9658, | |
| "expert_indices": [ | |
| 30, | |
| 131, | |
| 182, | |
| 183, | |
| 198, | |
| 238, | |
| 365, | |
| 377 | |
| ], | |
| "routing_weights": [ | |
| 0.056319162, | |
| 0.047798146, | |
| 0.056981582, | |
| 0.05191267, | |
| 0.04586145, | |
| 0.057513874, | |
| 0.045446966, | |
| 0.63816607 | |
| ], | |
| "bytes_streamed": 890406976 | |
| }, | |
| { | |
| "layer_idx": 10, | |
| "layer_kind": "moe", | |
| "residual_norm_in": 1690.9658, | |
| "residual_norm_out": 1954.9796, | |
| "expert_indices": [ | |
| 5, | |
| 34, | |
| 46, | |
| 118, | |
| 234, | |
| 240, | |
| 292, | |
| 296 | |
| ], | |
| "routing_weights": [ | |
| 0.033065867, | |
| 0.015240303, | |
| 0.01521005, | |
| 0.015278836, | |
| 0.014619413, | |
| 0.013330309, | |
| 0.017498873, | |
| 0.87575644 | |
| ], | |
| "bytes_streamed": 816089080 | |
| }, | |
| { | |
| "layer_idx": 11, | |
| "layer_kind": "moe", | |
| "residual_norm_in": 1954.9796, | |
| "residual_norm_out": 1956.8497, | |
| "expert_indices": [ | |
| 31, | |
| 124, | |
| 141, | |
| 180, | |
| 224, | |
| 237, | |
| 252, | |
| 253 | |
| ], | |
| "routing_weights": [ | |
| 0.13398919, | |
| 0.12286291, | |
| 0.119516574, | |
| 0.12470392, | |
| 0.13138652, | |
| 0.11974144, | |
| 0.12428722, | |
| 0.12351225 | |
| ], | |
| "bytes_streamed": 816089080 | |
| }, | |
| { | |
| "layer_idx": 12, | |
| "layer_kind": "moe", | |
| "residual_norm_in": 1956.8497, | |
| "residual_norm_out": 1980.657, | |
| "expert_indices": [ | |
| 25, | |
| 51, | |
| 92, | |
| 95, | |
| 139, | |
| 189, | |
| 270, | |
| 371 | |
| ], | |
| "routing_weights": [ | |
| 0.11804446, | |
| 0.12368671, | |
| 0.115437895, | |
| 0.12428875, | |
| 0.12710269, | |
| 0.13917556, | |
| 0.12618789, | |
| 0.12607609 | |
| ], | |
| "bytes_streamed": 865634344 | |
| }, | |
| { | |
| "layer_idx": 13, | |
| "layer_kind": "moe", | |
| "residual_norm_in": 1980.657, | |
| "residual_norm_out": 1982.7671, | |
| "expert_indices": [ | |
| 65, | |
| 68, | |
| 103, | |
| 194, | |
| 203, | |
| 228, | |
| 312, | |
| 348 | |
| ], | |
| "routing_weights": [ | |
| 0.15230049, | |
| 0.12651007, | |
| 0.12470709, | |
| 0.11999359, | |
| 0.12181889, | |
| 0.116762176, | |
| 0.12590373, | |
| 0.112003945 | |
| ], | |
| "bytes_streamed": 816089080 | |
| }, | |
| { | |
| "layer_idx": 14, | |
| "layer_kind": "moe", | |
| "residual_norm_in": 1982.7671, | |
| "residual_norm_out": 1985.4127, | |
| "expert_indices": [ | |
| 166, | |
| 182, | |
| 204, | |
| 207, | |
| 262, | |
| 269, | |
| 303, | |
| 377 | |
| ], | |
| "routing_weights": [ | |
| 0.12826353, | |
| 0.11211658, | |
| 0.16714138, | |
| 0.11411119, | |
| 0.12515004, | |
| 0.110577546, | |
| 0.118941896, | |
| 0.12369785 | |
| ], | |
| "bytes_streamed": 840861712 | |
| }, | |
| { | |
| "layer_idx": 15, | |
| "layer_kind": "moe", | |
| "residual_norm_in": 1985.4127, | |
| "residual_norm_out": 2258.3035, | |
| "expert_indices": [ | |
| 11, | |
| 109, | |
| 136, | |
| 150, | |
| 194, | |
| 212, | |
| 224, | |
| 279 | |
| ], | |
| "routing_weights": [ | |
| 0.07922835, | |
| 0.08260134, | |
| 0.08270256, | |
| 0.07873966, | |
| 0.08021491, | |
| 0.08108263, | |
| 0.07728814, | |
| 0.43814245 | |
| ], | |
| "bytes_streamed": 840861712 | |
| }, | |
| { | |
| "layer_idx": 16, | |
| "layer_kind": "moe", | |
| "residual_norm_in": 2258.3035, | |
| "residual_norm_out": 2261.8838, | |
| "expert_indices": [ | |
| 19, | |
| 80, | |
| 89, | |
| 91, | |
| 204, | |
| 225, | |
| 249, | |
| 339 | |
| ], | |
| "routing_weights": [ | |
| 0.11132005, | |
| 0.12770416, | |
| 0.12904347, | |
| 0.123933, | |
| 0.13755625, | |
| 0.11228153, | |
| 0.14229253, | |
| 0.115868956 | |
| ], | |
| "bytes_streamed": 816089080 | |
| }, | |
| { | |
| "layer_idx": 17, | |
| "layer_kind": "moe", | |
| "residual_norm_in": 2261.8838, | |
| "residual_norm_out": 2275.1423, | |
| "expert_indices": [ | |
| 81, | |
| 113, | |
| 128, | |
| 161, | |
| 282, | |
| 303, | |
| 304, | |
| 364 | |
| ], | |
| "routing_weights": [ | |
| 0.13470753, | |
| 0.12498505, | |
| 0.13510035, | |
| 0.13626955, | |
| 0.10787492, | |
| 0.10595762, | |
| 0.1165155, | |
| 0.13858946 | |
| ], | |
| "bytes_streamed": 840861712 | |
| }, | |
| { | |
| "layer_idx": 18, | |
| "layer_kind": "moe", | |
| "residual_norm_in": 2275.1423, | |
| "residual_norm_out": 2846.2473, | |
| "expert_indices": [ | |
| 30, | |
| 72, | |
| 95, | |
| 143, | |
| 230, | |
| 309, | |
| 329, | |
| 378 | |
| ], | |
| "routing_weights": [ | |
| 0.06173668, | |
| 0.060805954, | |
| 0.059534308, | |
| 0.546107, | |
| 0.05942752, | |
| 0.06282742, | |
| 0.059586395, | |
| 0.08997479 | |
| ], | |
| "bytes_streamed": 791316448 | |
| }, | |
| { | |
| "layer_idx": 19, | |
| "layer_kind": "moe", | |
| "residual_norm_in": 2846.2473, | |
| "residual_norm_out": 2848.6553, | |
| "expert_indices": [ | |
| 28, | |
| 62, | |
| 84, | |
| 210, | |
| 258, | |
| 264, | |
| 278, | |
| 318 | |
| ], | |
| "routing_weights": [ | |
| 0.12718764, | |
| 0.11587793, | |
| 0.116436146, | |
| 0.11349043, | |
| 0.12945548, | |
| 0.12185279, | |
| 0.15666124, | |
| 0.1190383 | |
| ], | |
| "bytes_streamed": 890406976 | |
| }, | |
| { | |
| "layer_idx": 20, | |
| "layer_kind": "moe", | |
| "residual_norm_in": 2848.6553, | |
| "residual_norm_out": 2964.716, | |
| "expert_indices": [ | |
| 7, | |
| 108, | |
| 146, | |
| 227, | |
| 260, | |
| 273, | |
| 329, | |
| 351 | |
| ], | |
| "routing_weights": [ | |
| 0.120410115, | |
| 0.12045736, | |
| 0.11941096, | |
| 0.116442, | |
| 0.15163471, | |
| 0.122804046, | |
| 0.13331619, | |
| 0.11552454 | |
| ], | |
| "bytes_streamed": 840861712 | |
| }, | |
| { | |
| "layer_idx": 21, | |
| "layer_kind": "moe", | |
| "residual_norm_in": 2964.716, | |
| "residual_norm_out": 2998.4307, | |
| "expert_indices": [ | |
| 8, | |
| 24, | |
| 140, | |
| 151, | |
| 250, | |
| 264, | |
| 313, | |
| 342 | |
| ], | |
| "routing_weights": [ | |
| 0.12219975, | |
| 0.11388808, | |
| 0.11746955, | |
| 0.11987377, | |
| 0.12484145, | |
| 0.122165635, | |
| 0.13231997, | |
| 0.14724179 | |
| ], | |
| "bytes_streamed": 816089080 | |
| }, | |
| { | |
| "layer_idx": 22, | |
| "layer_kind": "moe", | |
| "residual_norm_in": 2998.4307, | |
| "residual_norm_out": 3000.9702, | |
| "expert_indices": [ | |
| 3, | |
| 11, | |
| 78, | |
| 119, | |
| 206, | |
| 236, | |
| 249, | |
| 334 | |
| ], | |
| "routing_weights": [ | |
| 0.11585083, | |
| 0.12601939, | |
| 0.12537284, | |
| 0.12319891, | |
| 0.13266194, | |
| 0.11640788, | |
| 0.12136969, | |
| 0.13911851 | |
| ], | |
| "bytes_streamed": 816089080 | |
| }, | |
| { | |
| "layer_idx": 23, | |
| "layer_kind": "moe", | |
| "residual_norm_in": 3000.9702, | |
| "residual_norm_out": 3124.0881, | |
| "expert_indices": [ | |
| 8, | |
| 44, | |
| 125, | |
| 207, | |
| 209, | |
| 226, | |
| 283, | |
| 325 | |
| ], | |
| "routing_weights": [ | |
| 0.1277313, | |
| 0.11835284, | |
| 0.13288988, | |
| 0.119071715, | |
| 0.117436275, | |
| 0.12930648, | |
| 0.11873839, | |
| 0.13647313 | |
| ], | |
| "bytes_streamed": 890406976 | |
| }, | |
| { | |
| "layer_idx": 24, | |
| "layer_kind": "moe", | |
| "residual_norm_in": 3124.0881, | |
| "residual_norm_out": 3263.917, | |
| "expert_indices": [ | |
| 11, | |
| 46, | |
| 58, | |
| 59, | |
| 89, | |
| 223, | |
| 294, | |
| 376 | |
| ], | |
| "routing_weights": [ | |
| 0.13790713, | |
| 0.12226996, | |
| 0.12022834, | |
| 0.13080284, | |
| 0.11101902, | |
| 0.122765206, | |
| 0.1395905, | |
| 0.11541706 | |
| ], | |
| "bytes_streamed": 890406976 | |
| }, | |
| { | |
| "layer_idx": 25, | |
| "layer_kind": "moe", | |
| "residual_norm_in": 3263.917, | |
| "residual_norm_out": 3268.9028, | |
| "expert_indices": [ | |
| 32, | |
| 52, | |
| 186, | |
| 275, | |
| 292, | |
| 317, | |
| 325, | |
| 355 | |
| ], | |
| "routing_weights": [ | |
| 0.1295867, | |
| 0.1348981, | |
| 0.11795012, | |
| 0.12863119, | |
| 0.12783831, | |
| 0.12071772, | |
| 0.11842303, | |
| 0.12195492 | |
| ], | |
| "bytes_streamed": 816089080 | |
| }, | |
| { | |
| "layer_idx": 26, | |
| "layer_kind": "moe", | |
| "residual_norm_in": 3268.9028, | |
| "residual_norm_out": 3271.9517, | |
| "expert_indices": [ | |
| 28, | |
| 54, | |
| 109, | |
| 143, | |
| 164, | |
| 187, | |
| 234, | |
| 326 | |
| ], | |
| "routing_weights": [ | |
| 0.12403761, | |
| 0.124741495, | |
| 0.1233652, | |
| 0.12436722, | |
| 0.12376981, | |
| 0.13259424, | |
| 0.12488026, | |
| 0.12224412 | |
| ], | |
| "bytes_streamed": 816089080 | |
| }, | |
| { | |
| "layer_idx": 27, | |
| "layer_kind": "moe", | |
| "residual_norm_in": 3271.9517, | |
| "residual_norm_out": 3274.4165, | |
| "expert_indices": [ | |
| 5, | |
| 22, | |
| 23, | |
| 48, | |
| 115, | |
| 160, | |
| 188, | |
| 218 | |
| ], | |
| "routing_weights": [ | |
| 0.12905355, | |
| 0.11797063, | |
| 0.123731114, | |
| 0.11916181, | |
| 0.119591415, | |
| 0.12297361, | |
| 0.13747871, | |
| 0.13003916 | |
| ], | |
| "bytes_streamed": 816089080 | |
| }, | |
| { | |
| "layer_idx": 28, | |
| "layer_kind": "moe", | |
| "residual_norm_in": 3274.4165, | |
| "residual_norm_out": 3276.1294, | |
| "expert_indices": [ | |
| 83, | |
| 117, | |
| 178, | |
| 218, | |
| 233, | |
| 260, | |
| 305, | |
| 342 | |
| ], | |
| "routing_weights": [ | |
| 0.12889512, | |
| 0.12080206, | |
| 0.12741528, | |
| 0.119030036, | |
| 0.12386302, | |
| 0.12633918, | |
| 0.13027583, | |
| 0.12337941 | |
| ], | |
| "bytes_streamed": 816089080 | |
| }, | |
| { | |
| "layer_idx": 29, | |
| "layer_kind": "moe", | |
| "residual_norm_in": 3276.1294, | |
| "residual_norm_out": 3280.6016, | |
| "expert_indices": [ | |
| 19, | |
| 132, | |
| 145, | |
| 213, | |
| 264, | |
| 275, | |
| 309, | |
| 374 | |
| ], | |
| "routing_weights": [ | |
| 0.1201872, | |
| 0.11707843, | |
| 0.13667186, | |
| 0.11928836, | |
| 0.11894964, | |
| 0.14129789, | |
| 0.13095954, | |
| 0.115567006 | |
| ], | |
| "bytes_streamed": 791316448 | |
| }, | |
| { | |
| "layer_idx": 30, | |
| "layer_kind": "moe", | |
| "residual_norm_in": 3280.6016, | |
| "residual_norm_out": 3280.7778, | |
| "expert_indices": [ | |
| 55, | |
| 109, | |
| 165, | |
| 246, | |
| 254, | |
| 264, | |
| 275, | |
| 282 | |
| ], | |
| "routing_weights": [ | |
| 0.12114447, | |
| 0.1265742, | |
| 0.12592877, | |
| 0.13100544, | |
| 0.12474787, | |
| 0.1230107, | |
| 0.12573223, | |
| 0.12185628 | |
| ], | |
| "bytes_streamed": 840861712 | |
| }, | |
| { | |
| "layer_idx": 31, | |
| "layer_kind": "moe", | |
| "residual_norm_in": 3280.7778, | |
| "residual_norm_out": 3284.137, | |
| "expert_indices": [ | |
| 45, | |
| 150, | |
| 196, | |
| 198, | |
| 243, | |
| 323, | |
| 333, | |
| 359 | |
| ], | |
| "routing_weights": [ | |
| 0.124348536, | |
| 0.12571238, | |
| 0.121641114, | |
| 0.12634422, | |
| 0.12098488, | |
| 0.12598802, | |
| 0.12444443, | |
| 0.13053642 | |
| ], | |
| "bytes_streamed": 816089080 | |
| }, | |
| { | |
| "layer_idx": 32, | |
| "layer_kind": "moe", | |
| "residual_norm_in": 3284.137, | |
| "residual_norm_out": 3285.814, | |
| "expert_indices": [ | |
| 21, | |
| 76, | |
| 182, | |
| 270, | |
| 276, | |
| 286, | |
| 336, | |
| 345 | |
| ], | |
| "routing_weights": [ | |
| 0.123925604, | |
| 0.12669829, | |
| 0.12782235, | |
| 0.12475172, | |
| 0.12783591, | |
| 0.12554578, | |
| 0.12194816, | |
| 0.121472225 | |
| ], | |
| "bytes_streamed": 816089080 | |
| }, | |
| { | |
| "layer_idx": 33, | |
| "layer_kind": "moe", | |
| "residual_norm_in": 3285.814, | |
| "residual_norm_out": 3285.1543, | |
| "expert_indices": [ | |
| 9, | |
| 53, | |
| 54, | |
| 57, | |
| 267, | |
| 281, | |
| 303, | |
| 309 | |
| ], | |
| "routing_weights": [ | |
| 0.12560827, | |
| 0.12389935, | |
| 0.12423008, | |
| 0.12410613, | |
| 0.12650695, | |
| 0.12478609, | |
| 0.12493785, | |
| 0.12592524 | |
| ], | |
| "bytes_streamed": 816089080 | |
| }, | |
| { | |
| "layer_idx": 34, | |
| "layer_kind": "moe", | |
| "residual_norm_in": 3285.1543, | |
| "residual_norm_out": 3285.1926, | |
| "expert_indices": [ | |
| 54, | |
| 96, | |
| 152, | |
| 206, | |
| 250, | |
| 295, | |
| 325, | |
| 339 | |
| ], | |
| "routing_weights": [ | |
| 0.12830864, | |
| 0.12542318, | |
| 0.12211993, | |
| 0.123667784, | |
| 0.12476751, | |
| 0.12551583, | |
| 0.123447806, | |
| 0.12674934 | |
| ], | |
| "bytes_streamed": 816089080 | |
| }, | |
| { | |
| "layer_idx": 35, | |
| "layer_kind": "moe", | |
| "residual_norm_in": 3285.1926, | |
| "residual_norm_out": 3285.509, | |
| "expert_indices": [ | |
| 9, | |
| 68, | |
| 86, | |
| 142, | |
| 162, | |
| 176, | |
| 257, | |
| 326 | |
| ], | |
| "routing_weights": [ | |
| 0.12454217, | |
| 0.12767826, | |
| 0.124468386, | |
| 0.12491048, | |
| 0.12380852, | |
| 0.12711051, | |
| 0.122607514, | |
| 0.12487409 | |
| ], | |
| "bytes_streamed": 840861712 | |
| }, | |
| { | |
| "layer_idx": 36, | |
| "layer_kind": "moe", | |
| "residual_norm_in": 3285.509, | |
| "residual_norm_out": 3283.8137, | |
| "expert_indices": [ | |
| 51, | |
| 76, | |
| 132, | |
| 200, | |
| 236, | |
| 276, | |
| 317, | |
| 383 | |
| ], | |
| "routing_weights": [ | |
| 0.12545687, | |
| 0.12348959, | |
| 0.12646747, | |
| 0.12358548, | |
| 0.12351758, | |
| 0.123819746, | |
| 0.1276341, | |
| 0.12602918 | |
| ], | |
| "bytes_streamed": 865634344 | |
| }, | |
| { | |
| "layer_idx": 37, | |
| "layer_kind": "moe", | |
| "residual_norm_in": 3283.8137, | |
| "residual_norm_out": 3279.3472, | |
| "expert_indices": [ | |
| 22, | |
| 75, | |
| 156, | |
| 182, | |
| 187, | |
| 189, | |
| 230, | |
| 321 | |
| ], | |
| "routing_weights": [ | |
| 0.12714443, | |
| 0.12565663, | |
| 0.12507114, | |
| 0.12419243, | |
| 0.12688793, | |
| 0.123175636, | |
| 0.12531842, | |
| 0.122553304 | |
| ], | |
| "bytes_streamed": 816089080 | |
| }, | |
| { | |
| "layer_idx": 38, | |
| "layer_kind": "moe", | |
| "residual_norm_in": 3279.3472, | |
| "residual_norm_out": 3278.7837, | |
| "expert_indices": [ | |
| 116, | |
| 118, | |
| 141, | |
| 142, | |
| 165, | |
| 246, | |
| 296, | |
| 305 | |
| ], | |
| "routing_weights": [ | |
| 0.12589957, | |
| 0.12569398, | |
| 0.12418391, | |
| 0.12590075, | |
| 0.12354491, | |
| 0.12647879, | |
| 0.123581976, | |
| 0.1247161 | |
| ], | |
| "bytes_streamed": 840861712 | |
| }, | |
| { | |
| "layer_idx": 39, | |
| "layer_kind": "moe", | |
| "residual_norm_in": 3278.7837, | |
| "residual_norm_out": 3278.0305, | |
| "expert_indices": [ | |
| 1, | |
| 21, | |
| 53, | |
| 74, | |
| 153, | |
| 232, | |
| 278, | |
| 381 | |
| ], | |
| "routing_weights": [ | |
| 0.12392103, | |
| 0.12356896, | |
| 0.12509237, | |
| 0.12476749, | |
| 0.1261542, | |
| 0.1256581, | |
| 0.12628376, | |
| 0.124554105 | |
| ], | |
| "bytes_streamed": 865634344 | |
| }, | |
| { | |
| "layer_idx": 40, | |
| "layer_kind": "moe", | |
| "residual_norm_in": 3278.0305, | |
| "residual_norm_out": 3273.8188, | |
| "expert_indices": [ | |
| 14, | |
| 54, | |
| 80, | |
| 205, | |
| 243, | |
| 350, | |
| 360, | |
| 378 | |
| ], | |
| "routing_weights": [ | |
| 0.123624906, | |
| 0.12560081, | |
| 0.12601347, | |
| 0.12464201, | |
| 0.124437995, | |
| 0.124908455, | |
| 0.1261274, | |
| 0.124645 | |
| ], | |
| "bytes_streamed": 816089080 | |
| }, | |
| { | |
| "layer_idx": 41, | |
| "layer_kind": "moe", | |
| "residual_norm_in": 3273.8188, | |
| "residual_norm_out": 3270.8535, | |
| "expert_indices": [ | |
| 0, | |
| 12, | |
| 102, | |
| 118, | |
| 143, | |
| 323, | |
| 346, | |
| 361 | |
| ], | |
| "routing_weights": [ | |
| 0.12718984, | |
| 0.12352078, | |
| 0.124371946, | |
| 0.1250337, | |
| 0.12582457, | |
| 0.12378337, | |
| 0.1258188, | |
| 0.12445698 | |
| ], | |
| "bytes_streamed": 791316448 | |
| }, | |
| { | |
| "layer_idx": 42, | |
| "layer_kind": "moe", | |
| "residual_norm_in": 3270.8535, | |
| "residual_norm_out": 3264.688, | |
| "expert_indices": [ | |
| 11, | |
| 63, | |
| 66, | |
| 121, | |
| 233, | |
| 282, | |
| 286, | |
| 322 | |
| ], | |
| "routing_weights": [ | |
| 0.12231816, | |
| 0.12572178, | |
| 0.12679413, | |
| 0.12391465, | |
| 0.12539703, | |
| 0.122256, | |
| 0.12437119, | |
| 0.12922701 | |
| ], | |
| "bytes_streamed": 840861712 | |
| }, | |
| { | |
| "layer_idx": 43, | |
| "layer_kind": "moe", | |
| "residual_norm_in": 3264.688, | |
| "residual_norm_out": 3260.6675, | |
| "expert_indices": [ | |
| 48, | |
| 97, | |
| 111, | |
| 145, | |
| 166, | |
| 258, | |
| 275, | |
| 310 | |
| ], | |
| "routing_weights": [ | |
| 0.12841284, | |
| 0.12592939, | |
| 0.1230621, | |
| 0.11106126, | |
| 0.13067316, | |
| 0.12302471, | |
| 0.12871376, | |
| 0.1291227 | |
| ], | |
| "bytes_streamed": 865634344 | |
| }, | |
| { | |
| "layer_idx": 44, | |
| "layer_kind": "moe", | |
| "residual_norm_in": 3260.6675, | |
| "residual_norm_out": 3252.343, | |
| "expert_indices": [ | |
| 52, | |
| 80, | |
| 135, | |
| 143, | |
| 145, | |
| 204, | |
| 209, | |
| 313 | |
| ], | |
| "routing_weights": [ | |
| 0.1505961, | |
| 0.15023981, | |
| 0.11775557, | |
| 0.087679155, | |
| 0.12829639, | |
| 0.094963625, | |
| 0.11943085, | |
| 0.15103848 | |
| ], | |
| "bytes_streamed": 816089080 | |
| }, | |
| { | |
| "layer_idx": 45, | |
| "layer_kind": "moe", | |
| "residual_norm_in": 3252.343, | |
| "residual_norm_out": 3249.5884, | |
| "expert_indices": [ | |
| 20, | |
| 45, | |
| 57, | |
| 59, | |
| 105, | |
| 176, | |
| 271, | |
| 320 | |
| ], | |
| "routing_weights": [ | |
| 0.12892541, | |
| 0.11697536, | |
| 0.14063151, | |
| 0.12651275, | |
| 0.13517754, | |
| 0.09021728, | |
| 0.13699076, | |
| 0.12456939 | |
| ], | |
| "bytes_streamed": 865634344 | |
| }, | |
| { | |
| "layer_idx": 46, | |
| "layer_kind": "moe", | |
| "residual_norm_in": 3249.5884, | |
| "residual_norm_out": 3244.682, | |
| "expert_indices": [ | |
| 7, | |
| 13, | |
| 31, | |
| 58, | |
| 198, | |
| 206, | |
| 222, | |
| 361 | |
| ], | |
| "routing_weights": [ | |
| 0.13090932, | |
| 0.13061666, | |
| 0.119749196, | |
| 0.13284738, | |
| 0.13148266, | |
| 0.12531951, | |
| 0.102036834, | |
| 0.1270384 | |
| ], | |
| "bytes_streamed": 840861712 | |
| }, | |
| { | |
| "layer_idx": 47, | |
| "layer_kind": "moe", | |
| "residual_norm_in": 3244.682, | |
| "residual_norm_out": 3242.7827, | |
| "expert_indices": [ | |
| 50, | |
| 91, | |
| 248, | |
| 268, | |
| 274, | |
| 305, | |
| 359, | |
| 377 | |
| ], | |
| "routing_weights": [ | |
| 0.090512484, | |
| 0.12572679, | |
| 0.12560984, | |
| 0.13680649, | |
| 0.13461365, | |
| 0.1383419, | |
| 0.11906946, | |
| 0.12931941 | |
| ], | |
| "bytes_streamed": 865634344 | |
| }, | |
| { | |
| "layer_idx": 48, | |
| "layer_kind": "moe", | |
| "residual_norm_in": 3242.7827, | |
| "residual_norm_out": 3235.4805, | |
| "expert_indices": [ | |
| 21, | |
| 140, | |
| 180, | |
| 211, | |
| 216, | |
| 265, | |
| 275, | |
| 356 | |
| ], | |
| "routing_weights": [ | |
| 0.089280955, | |
| 0.11947252, | |
| 0.14138731, | |
| 0.13919227, | |
| 0.13082214, | |
| 0.12825514, | |
| 0.12513566, | |
| 0.1264539 | |
| ], | |
| "bytes_streamed": 816089080 | |
| }, | |
| { | |
| "layer_idx": 49, | |
| "layer_kind": "moe", | |
| "residual_norm_in": 3235.4805, | |
| "residual_norm_out": 3223.483, | |
| "expert_indices": [ | |
| 65, | |
| 132, | |
| 136, | |
| 141, | |
| 257, | |
| 330, | |
| 331, | |
| 381 | |
| ], | |
| "routing_weights": [ | |
| 0.15157294, | |
| 0.14870882, | |
| 0.12575087, | |
| 0.14817211, | |
| 0.12935574, | |
| 0.12963831, | |
| 0.084066786, | |
| 0.082734436 | |
| ], | |
| "bytes_streamed": 840861712 | |
| }, | |
| { | |
| "layer_idx": 50, | |
| "layer_kind": "moe", | |
| "residual_norm_in": 3223.483, | |
| "residual_norm_out": 3211.675, | |
| "expert_indices": [ | |
| 2, | |
| 46, | |
| 77, | |
| 114, | |
| 160, | |
| 162, | |
| 175, | |
| 324 | |
| ], | |
| "routing_weights": [ | |
| 0.07168002, | |
| 0.13420905, | |
| 0.14428033, | |
| 0.13734862, | |
| 0.14445138, | |
| 0.13382252, | |
| 0.11261744, | |
| 0.121590614 | |
| ], | |
| "bytes_streamed": 840861712 | |
| }, | |
| { | |
| "layer_idx": 51, | |
| "layer_kind": "moe", | |
| "residual_norm_in": 3211.675, | |
| "residual_norm_out": 3202.074, | |
| "expert_indices": [ | |
| 36, | |
| 130, | |
| 189, | |
| 218, | |
| 263, | |
| 271, | |
| 317, | |
| 362 | |
| ], | |
| "routing_weights": [ | |
| 0.07808468, | |
| 0.14101124, | |
| 0.13042147, | |
| 0.11585922, | |
| 0.13718921, | |
| 0.12932168, | |
| 0.12141402, | |
| 0.1466985 | |
| ], | |
| "bytes_streamed": 791316448 | |
| }, | |
| { | |
| "layer_idx": 52, | |
| "layer_kind": "moe", | |
| "residual_norm_in": 3202.074, | |
| "residual_norm_out": 3184.0989, | |
| "expert_indices": [ | |
| 10, | |
| 31, | |
| 40, | |
| 49, | |
| 52, | |
| 96, | |
| 259, | |
| 342 | |
| ], | |
| "routing_weights": [ | |
| 0.13195448, | |
| 0.12658952, | |
| 0.14112595, | |
| 0.11027296, | |
| 0.16430713, | |
| 0.07610071, | |
| 0.12522088, | |
| 0.12442839 | |
| ], | |
| "bytes_streamed": 791316448 | |
| }, | |
| { | |
| "layer_idx": 53, | |
| "layer_kind": "moe", | |
| "residual_norm_in": 3184.0989, | |
| "residual_norm_out": 3148.3574, | |
| "expert_indices": [ | |
| 42, | |
| 101, | |
| 131, | |
| 247, | |
| 272, | |
| 341, | |
| 371, | |
| 383 | |
| ], | |
| "routing_weights": [ | |
| 0.06139395, | |
| 0.17122178, | |
| 0.14294012, | |
| 0.10139634, | |
| 0.056430835, | |
| 0.16060388, | |
| 0.14770874, | |
| 0.15830438 | |
| ], | |
| "bytes_streamed": 840861712 | |
| }, | |
| { | |
| "layer_idx": 54, | |
| "layer_kind": "moe", | |
| "residual_norm_in": 3148.3574, | |
| "residual_norm_out": 3037.2, | |
| "expert_indices": [ | |
| 68, | |
| 156, | |
| 169, | |
| 189, | |
| 192, | |
| 218, | |
| 240, | |
| 319 | |
| ], | |
| "routing_weights": [ | |
| 0.053958632, | |
| 0.05449725, | |
| 0.14695823, | |
| 0.14635251, | |
| 0.15215138, | |
| 0.1491312, | |
| 0.14903401, | |
| 0.14791681 | |
| ], | |
| "bytes_streamed": 840861712 | |
| }, | |
| { | |
| "layer_idx": 55, | |
| "layer_kind": "moe", | |
| "residual_norm_in": 3037.2, | |
| "residual_norm_out": 2971.9001, | |
| "expert_indices": [ | |
| 42, | |
| 87, | |
| 173, | |
| 246, | |
| 308, | |
| 309, | |
| 329, | |
| 348 | |
| ], | |
| "routing_weights": [ | |
| 0.05151474, | |
| 0.1472363, | |
| 0.1606687, | |
| 0.16340461, | |
| 0.15019181, | |
| 0.1526328, | |
| 0.11705792, | |
| 0.05729307 | |
| ], | |
| "bytes_streamed": 865634344 | |
| }, | |
| { | |
| "layer_idx": 56, | |
| "layer_kind": "moe", | |
| "residual_norm_in": 2971.9001, | |
| "residual_norm_out": 2925.79, | |
| "expert_indices": [ | |
| 26, | |
| 67, | |
| 74, | |
| 84, | |
| 253, | |
| 316, | |
| 319, | |
| 363 | |
| ], | |
| "routing_weights": [ | |
| 0.16366942, | |
| 0.16105407, | |
| 0.16047254, | |
| 0.07660575, | |
| 0.07692882, | |
| 0.16656852, | |
| 0.123046815, | |
| 0.07165412 | |
| ], | |
| "bytes_streamed": 840861712 | |
| }, | |
| { | |
| "layer_idx": 57, | |
| "layer_kind": "moe", | |
| "residual_norm_in": 2925.79, | |
| "residual_norm_out": 2824.1897, | |
| "expert_indices": [ | |
| 0, | |
| 30, | |
| 80, | |
| 84, | |
| 107, | |
| 222, | |
| 253, | |
| 288 | |
| ], | |
| "routing_weights": [ | |
| 0.18186912, | |
| 0.052621696, | |
| 0.12891348, | |
| 0.05427989, | |
| 0.1740672, | |
| 0.18127596, | |
| 0.17727546, | |
| 0.04969726 | |
| ], | |
| "bytes_streamed": 791316448 | |
| }, | |
| { | |
| "layer_idx": 58, | |
| "layer_kind": "moe", | |
| "residual_norm_in": 2824.1897, | |
| "residual_norm_out": 2600.4956, | |
| "expert_indices": [ | |
| 5, | |
| 234, | |
| 259, | |
| 279, | |
| 329, | |
| 344, | |
| 350, | |
| 359 | |
| ], | |
| "routing_weights": [ | |
| 0.12612459, | |
| 0.1269469, | |
| 0.12518732, | |
| 0.12676597, | |
| 0.1180102, | |
| 0.12671871, | |
| 0.12414087, | |
| 0.12610547 | |
| ], | |
| "bytes_streamed": 816089080 | |
| }, | |
| { | |
| "layer_idx": 59, | |
| "layer_kind": "moe", | |
| "residual_norm_in": 2600.4956, | |
| "residual_norm_out": 3555.355, | |
| "expert_indices": [ | |
| 47, | |
| 62, | |
| 83, | |
| 156, | |
| 161, | |
| 176, | |
| 372, | |
| 374 | |
| ], | |
| "routing_weights": [ | |
| 0.09576432, | |
| 0.07162925, | |
| 0.061123345, | |
| 0.07196141, | |
| 0.062052693, | |
| 0.48133266, | |
| 0.085414946, | |
| 0.07072137 | |
| ], | |
| "bytes_streamed": 791316448 | |
| }, | |
| { | |
| "layer_idx": 60, | |
| "layer_kind": "moe", | |
| "residual_norm_in": 3555.355, | |
| "residual_norm_out": 9134.746, | |
| "expert_indices": [ | |
| 3, | |
| 55, | |
| 68, | |
| 73, | |
| 217, | |
| 234, | |
| 242, | |
| 378 | |
| ], | |
| "routing_weights": [ | |
| 0.03411237, | |
| 0.039301682, | |
| 0.1990229, | |
| 0.18033567, | |
| 0.074357, | |
| 0.2359254, | |
| 0.19573748, | |
| 0.041207485 | |
| ], | |
| "bytes_streamed": 816089080 | |
| } | |
| ] | |
| }, | |
| { | |
| "step_index": 3, | |
| "input_token_ids": [ | |
| 163584, | |
| 950, | |
| 1268, | |
| 426 | |
| ], | |
| "output_token_id": 120287, | |
| "output_logit": 15.419181, | |
| "logits_digest": "e6f09139fbbc6f3ac49559a36875bb9065d30b9258b06145c8339c8235d0407d", | |
| "hidden_digest": "1c28ee7fac7c442fd659f21b5cadc65bcfb613ef6a5d4f8efd37baecfd7de51a", | |
| "top_k_tokens": [ | |
| [ | |
| 120287, | |
| 15.419181 | |
| ], | |
| [ | |
| 63, | |
| 14.588471 | |
| ], | |
| [ | |
| 56841, | |
| 14.501873 | |
| ], | |
| [ | |
| 2350, | |
| 14.499147 | |
| ], | |
| [ | |
| 16, | |
| 14.292819 | |
| ], | |
| [ | |
| 15052, | |
| 14.150986 | |
| ], | |
| [ | |
| 9197, | |
| 14.074427 | |
| ], | |
| [ | |
| 366, | |
| 14.06571 | |
| ], | |
| [ | |
| 12585, | |
| 13.425251 | |
| ], | |
| [ | |
| 4352, | |
| 13.358892 | |
| ], | |
| [ | |
| 17, | |
| 13.109247 | |
| ], | |
| [ | |
| 17958, | |
| 12.965369 | |
| ], | |
| [ | |
| 15, | |
| 12.964136 | |
| ], | |
| [ | |
| 1265, | |
| 12.955637 | |
| ], | |
| [ | |
| 141396, | |
| 12.939412 | |
| ], | |
| [ | |
| 6386, | |
| 12.891417 | |
| ] | |
| ], | |
| "entropy": 5.213575, | |
| "margin": 0.83070946, | |
| "bytes_streamed": 62882493760, | |
| "layers": [ | |
| { | |
| "layer_idx": 0, | |
| "layer_kind": "dense", | |
| "residual_norm_in": 1.4958955, | |
| "residual_norm_out": 2.049361, | |
| "expert_indices": null, | |
| "routing_weights": null, | |
| "bytes_streamed": 995000320 | |
| }, | |
| { | |
| "layer_idx": 1, | |
| "layer_kind": "moe", | |
| "residual_norm_in": 2.049361, | |
| "residual_norm_out": 3.407005, | |
| "expert_indices": [ | |
| 5, | |
| 33, | |
| 72, | |
| 105, | |
| 116, | |
| 154, | |
| 182, | |
| 272 | |
| ], | |
| "routing_weights": [ | |
| 0.087083995, | |
| 0.20157112, | |
| 0.14540455, | |
| 0.08845074, | |
| 0.1286529, | |
| 0.10869927, | |
| 0.12787405, | |
| 0.112263314 | |
| ], | |
| "bytes_streamed": 890406976 | |
| }, | |
| { | |
| "layer_idx": 2, | |
| "layer_kind": "moe", | |
| "residual_norm_in": 3.407005, | |
| "residual_norm_out": 4.7781343, | |
| "expert_indices": [ | |
| 7, | |
| 25, | |
| 92, | |
| 179, | |
| 190, | |
| 240, | |
| 340, | |
| 343 | |
| ], | |
| "routing_weights": [ | |
| 0.1776416, | |
| 0.068654895, | |
| 0.1582383, | |
| 0.0638775, | |
| 0.12181032, | |
| 0.18464753, | |
| 0.060134597, | |
| 0.16499528 | |
| ], | |
| "bytes_streamed": 865634344 | |
| }, | |
| { | |
| "layer_idx": 3, | |
| "layer_kind": "moe", | |
| "residual_norm_in": 4.7781343, | |
| "residual_norm_out": 10.123125, | |
| "expert_indices": [ | |
| 15, | |
| 189, | |
| 190, | |
| 217, | |
| 253, | |
| 271, | |
| 337, | |
| 353 | |
| ], | |
| "routing_weights": [ | |
| 0.05684734, | |
| 0.062119197, | |
| 0.052167017, | |
| 0.35753912, | |
| 0.06827293, | |
| 0.049369186, | |
| 0.04669538, | |
| 0.30698988 | |
| ], | |
| "bytes_streamed": 915179608 | |
| }, | |
| { | |
| "layer_idx": 4, | |
| "layer_kind": "moe", | |
| "residual_norm_in": 10.123125, | |
| "residual_norm_out": 78.43251, | |
| "expert_indices": [ | |
| 52, | |
| 98, | |
| 106, | |
| 109, | |
| 120, | |
| 189, | |
| 191, | |
| 324 | |
| ], | |
| "routing_weights": [ | |
| 0.07070838, | |
| 0.08047035, | |
| 0.077183425, | |
| 0.08113832, | |
| 0.071013846, | |
| 0.47025713, | |
| 0.08182316, | |
| 0.067405246 | |
| ], | |
| "bytes_streamed": 1014270136 | |
| }, | |
| { | |
| "layer_idx": 5, | |
| "layer_kind": "moe", | |
| "residual_norm_in": 78.43251, | |
| "residual_norm_out": 84.577156, | |
| "expert_indices": [ | |
| 12, | |
| 17, | |
| 121, | |
| 124, | |
| 160, | |
| 186, | |
| 195, | |
| 237 | |
| ], | |
| "routing_weights": [ | |
| 0.1888245, | |
| 0.10076529, | |
| 0.11140824, | |
| 0.17085913, | |
| 0.10019477, | |
| 0.10386934, | |
| 0.095480084, | |
| 0.12859854 | |
| ], | |
| "bytes_streamed": 890406976 | |
| }, | |
| { | |
| "layer_idx": 6, | |
| "layer_kind": "moe", | |
| "residual_norm_in": 84.577156, | |
| "residual_norm_out": 390.06165, | |
| "expert_indices": [ | |
| 18, | |
| 45, | |
| 88, | |
| 239, | |
| 243, | |
| 269, | |
| 334, | |
| 346 | |
| ], | |
| "routing_weights": [ | |
| 0.046305433, | |
| 0.029368749, | |
| 0.033812124, | |
| 0.031240184, | |
| 0.5246116, | |
| 0.0594912, | |
| 0.037579488, | |
| 0.23759124 | |
| ], | |
| "bytes_streamed": 964724872 | |
| }, | |
| { | |
| "layer_idx": 7, | |
| "layer_kind": "moe", | |
| "residual_norm_in": 390.06165, | |
| "residual_norm_out": 1106.5867, | |
| "expert_indices": [ | |
| 54, | |
| 57, | |
| 155, | |
| 183, | |
| 185, | |
| 192, | |
| 302, | |
| 361 | |
| ], | |
| "routing_weights": [ | |
| 0.025255695, | |
| 0.0061686267, | |
| 0.007244412, | |
| 0.92817426, | |
| 0.009681543, | |
| 0.0078814775, | |
| 0.0067851017, | |
| 0.008808859 | |
| ], | |
| "bytes_streamed": 890406976 | |
| }, | |
| { | |
| "layer_idx": 8, | |
| "layer_kind": "moe", | |
| "residual_norm_in": 1106.5867, | |
| "residual_norm_out": 1492.6448, | |
| "expert_indices": [ | |
| 86, | |
| 128, | |
| 175, | |
| 287, | |
| 311, | |
| 336, | |
| 337, | |
| 351 | |
| ], | |
| "routing_weights": [ | |
| 0.06048171, | |
| 0.030799579, | |
| 0.030457757, | |
| 0.038466815, | |
| 0.029505648, | |
| 0.032741938, | |
| 0.034052644, | |
| 0.743494 | |
| ], | |
| "bytes_streamed": 939952240 | |
| }, | |
| { | |
| "layer_idx": 9, | |
| "layer_kind": "moe", | |
| "residual_norm_in": 1492.6448, | |
| "residual_norm_out": 1690.9707, | |
| "expert_indices": [ | |
| 30, | |
| 131, | |
| 182, | |
| 183, | |
| 198, | |
| 238, | |
| 365, | |
| 377 | |
| ], | |
| "routing_weights": [ | |
| 0.056319162, | |
| 0.047798146, | |
| 0.056981582, | |
| 0.05191267, | |
| 0.04586145, | |
| 0.057513874, | |
| 0.045446966, | |
| 0.63816607 | |
| ], | |
| "bytes_streamed": 1039042768 | |
| }, | |
| { | |
| "layer_idx": 10, | |
| "layer_kind": "moe", | |
| "residual_norm_in": 1690.9707, | |
| "residual_norm_out": 1954.9858, | |
| "expert_indices": [ | |
| 5, | |
| 34, | |
| 46, | |
| 118, | |
| 234, | |
| 240, | |
| 292, | |
| 296 | |
| ], | |
| "routing_weights": [ | |
| 0.033065867, | |
| 0.015240303, | |
| 0.01521005, | |
| 0.015278836, | |
| 0.014619413, | |
| 0.013330309, | |
| 0.017498873, | |
| 0.87575644 | |
| ], | |
| "bytes_streamed": 915179608 | |
| }, | |
| { | |
| "layer_idx": 11, | |
| "layer_kind": "moe", | |
| "residual_norm_in": 1954.9858, | |
| "residual_norm_out": 1956.8572, | |
| "expert_indices": [ | |
| 31, | |
| 124, | |
| 141, | |
| 180, | |
| 224, | |
| 237, | |
| 252, | |
| 253 | |
| ], | |
| "routing_weights": [ | |
| 0.13398919, | |
| 0.12286291, | |
| 0.119516574, | |
| 0.12470392, | |
| 0.13138652, | |
| 0.11974144, | |
| 0.12428722, | |
| 0.12351225 | |
| ], | |
| "bytes_streamed": 964724872 | |
| }, | |
| { | |
| "layer_idx": 12, | |
| "layer_kind": "moe", | |
| "residual_norm_in": 1956.8572, | |
| "residual_norm_out": 1980.6659, | |
| "expert_indices": [ | |
| 25, | |
| 51, | |
| 92, | |
| 95, | |
| 139, | |
| 189, | |
| 270, | |
| 371 | |
| ], | |
| "routing_weights": [ | |
| 0.11804446, | |
| 0.12368671, | |
| 0.115437895, | |
| 0.12428875, | |
| 0.12710269, | |
| 0.13917556, | |
| 0.12618789, | |
| 0.12607609 | |
| ], | |
| "bytes_streamed": 964724872 | |
| }, | |
| { | |
| "layer_idx": 13, | |
| "layer_kind": "moe", | |
| "residual_norm_in": 1980.6659, | |
| "residual_norm_out": 1982.7783, | |
| "expert_indices": [ | |
| 65, | |
| 68, | |
| 103, | |
| 194, | |
| 203, | |
| 228, | |
| 312, | |
| 348 | |
| ], | |
| "routing_weights": [ | |
| 0.15230049, | |
| 0.12651007, | |
| 0.12470709, | |
| 0.11999359, | |
| 0.12181889, | |
| 0.116762176, | |
| 0.12590373, | |
| 0.112003945 | |
| ], | |
| "bytes_streamed": 915179608 | |
| }, | |
| { | |
| "layer_idx": 14, | |
| "layer_kind": "moe", | |
| "residual_norm_in": 1982.7783, | |
| "residual_norm_out": 1985.4268, | |
| "expert_indices": [ | |
| 166, | |
| 182, | |
| 204, | |
| 207, | |
| 262, | |
| 269, | |
| 303, | |
| 377 | |
| ], | |
| "routing_weights": [ | |
| 0.12826353, | |
| 0.11211658, | |
| 0.16714138, | |
| 0.11411119, | |
| 0.12515004, | |
| 0.110577546, | |
| 0.118941896, | |
| 0.12369785 | |
| ], | |
| "bytes_streamed": 915179608 | |
| }, | |
| { | |
| "layer_idx": 15, | |
| "layer_kind": "moe", | |
| "residual_norm_in": 1985.4268, | |
| "residual_norm_out": 2258.3145, | |
| "expert_indices": [ | |
| 11, | |
| 109, | |
| 136, | |
| 150, | |
| 194, | |
| 212, | |
| 224, | |
| 279 | |
| ], | |
| "routing_weights": [ | |
| 0.07922835, | |
| 0.08260134, | |
| 0.08270256, | |
| 0.07873966, | |
| 0.08021491, | |
| 0.08108263, | |
| 0.07728814, | |
| 0.43814245 | |
| ], | |
| "bytes_streamed": 964724872 | |
| }, | |
| { | |
| "layer_idx": 16, | |
| "layer_kind": "moe", | |
| "residual_norm_in": 2258.3145, | |
| "residual_norm_out": 2261.8975, | |
| "expert_indices": [ | |
| 19, | |
| 80, | |
| 89, | |
| 91, | |
| 204, | |
| 225, | |
| 249, | |
| 339 | |
| ], | |
| "routing_weights": [ | |
| 0.11132005, | |
| 0.12770416, | |
| 0.12904347, | |
| 0.123933, | |
| 0.13755625, | |
| 0.11228153, | |
| 0.14229253, | |
| 0.115868956 | |
| ], | |
| "bytes_streamed": 890406976 | |
| }, | |
| { | |
| "layer_idx": 17, | |
| "layer_kind": "moe", | |
| "residual_norm_in": 2261.8975, | |
| "residual_norm_out": 2275.1616, | |
| "expert_indices": [ | |
| 81, | |
| 113, | |
| 128, | |
| 161, | |
| 282, | |
| 303, | |
| 304, | |
| 364 | |
| ], | |
| "routing_weights": [ | |
| 0.13470753, | |
| 0.12498505, | |
| 0.13510035, | |
| 0.13626955, | |
| 0.10787492, | |
| 0.10595762, | |
| 0.1165155, | |
| 0.13858946 | |
| ], | |
| "bytes_streamed": 964724872 | |
| }, | |
| { | |
| "layer_idx": 18, | |
| "layer_kind": "moe", | |
| "residual_norm_in": 2275.1616, | |
| "residual_norm_out": 2846.267, | |
| "expert_indices": [ | |
| 30, | |
| 72, | |
| 95, | |
| 143, | |
| 230, | |
| 309, | |
| 329, | |
| 378 | |
| ], | |
| "routing_weights": [ | |
| 0.06173668, | |
| 0.060805954, | |
| 0.059534308, | |
| 0.546107, | |
| 0.05942752, | |
| 0.06282742, | |
| 0.059586395, | |
| 0.08997479 | |
| ], | |
| "bytes_streamed": 865634344 | |
| }, | |
| { | |
| "layer_idx": 19, | |
| "layer_kind": "moe", | |
| "residual_norm_in": 2846.267, | |
| "residual_norm_out": 2848.678, | |
| "expert_indices": [ | |
| 28, | |
| 62, | |
| 84, | |
| 210, | |
| 258, | |
| 264, | |
| 278, | |
| 318 | |
| ], | |
| "routing_weights": [ | |
| 0.12718764, | |
| 0.11587793, | |
| 0.116436146, | |
| 0.11349043, | |
| 0.12945548, | |
| 0.12185279, | |
| 0.15666124, | |
| 0.1190383 | |
| ], | |
| "bytes_streamed": 1014270136 | |
| }, | |
| { | |
| "layer_idx": 20, | |
| "layer_kind": "moe", | |
| "residual_norm_in": 2848.678, | |
| "residual_norm_out": 2964.7385, | |
| "expert_indices": [ | |
| 7, | |
| 108, | |
| 146, | |
| 227, | |
| 260, | |
| 273, | |
| 329, | |
| 351 | |
| ], | |
| "routing_weights": [ | |
| 0.120410115, | |
| 0.12045736, | |
| 0.11941096, | |
| 0.116442, | |
| 0.15163471, | |
| 0.122804046, | |
| 0.13331619, | |
| 0.11552454 | |
| ], | |
| "bytes_streamed": 915179608 | |
| }, | |
| { | |
| "layer_idx": 21, | |
| "layer_kind": "moe", | |
| "residual_norm_in": 2964.7385, | |
| "residual_norm_out": 2998.454, | |
| "expert_indices": [ | |
| 8, | |
| 24, | |
| 140, | |
| 151, | |
| 250, | |
| 264, | |
| 313, | |
| 342 | |
| ], | |
| "routing_weights": [ | |
| 0.12219975, | |
| 0.11388808, | |
| 0.11746955, | |
| 0.11987377, | |
| 0.12484145, | |
| 0.122165635, | |
| 0.13231997, | |
| 0.14724179 | |
| ], | |
| "bytes_streamed": 989497504 | |
| }, | |
| { | |
| "layer_idx": 22, | |
| "layer_kind": "moe", | |
| "residual_norm_in": 2998.454, | |
| "residual_norm_out": 3000.9976, | |
| "expert_indices": [ | |
| 3, | |
| 11, | |
| 78, | |
| 119, | |
| 206, | |
| 236, | |
| 249, | |
| 334 | |
| ], | |
| "routing_weights": [ | |
| 0.11585083, | |
| 0.12601939, | |
| 0.12537284, | |
| 0.12319891, | |
| 0.13266194, | |
| 0.11640788, | |
| 0.12136969, | |
| 0.13911851 | |
| ], | |
| "bytes_streamed": 865634344 | |
| }, | |
| { | |
| "layer_idx": 23, | |
| "layer_kind": "moe", | |
| "residual_norm_in": 3000.9976, | |
| "residual_norm_out": 3124.115, | |
| "expert_indices": [ | |
| 8, | |
| 44, | |
| 125, | |
| 207, | |
| 209, | |
| 226, | |
| 283, | |
| 325 | |
| ], | |
| "routing_weights": [ | |
| 0.1277313, | |
| 0.11835284, | |
| 0.13288988, | |
| 0.119071715, | |
| 0.117436275, | |
| 0.12930648, | |
| 0.11873839, | |
| 0.13647313 | |
| ], | |
| "bytes_streamed": 989497504 | |
| }, | |
| { | |
| "layer_idx": 24, | |
| "layer_kind": "moe", | |
| "residual_norm_in": 3124.115, | |
| "residual_norm_out": 3263.947, | |
| "expert_indices": [ | |
| 11, | |
| 46, | |
| 58, | |
| 59, | |
| 89, | |
| 223, | |
| 294, | |
| 376 | |
| ], | |
| "routing_weights": [ | |
| 0.13790713, | |
| 0.12226996, | |
| 0.12022834, | |
| 0.13080284, | |
| 0.11101902, | |
| 0.122765206, | |
| 0.1395905, | |
| 0.11541706 | |
| ], | |
| "bytes_streamed": 1014270136 | |
| }, | |
| { | |
| "layer_idx": 25, | |
| "layer_kind": "moe", | |
| "residual_norm_in": 3263.947, | |
| "residual_norm_out": 3268.942, | |
| "expert_indices": [ | |
| 32, | |
| 52, | |
| 186, | |
| 275, | |
| 292, | |
| 317, | |
| 325, | |
| 355 | |
| ], | |
| "routing_weights": [ | |
| 0.1295867, | |
| 0.1348981, | |
| 0.11795012, | |
| 0.12863119, | |
| 0.12783831, | |
| 0.12071772, | |
| 0.11842303, | |
| 0.12195492 | |
| ], | |
| "bytes_streamed": 939952240 | |
| }, | |
| { | |
| "layer_idx": 26, | |
| "layer_kind": "moe", | |
| "residual_norm_in": 3268.942, | |
| "residual_norm_out": 3271.9993, | |
| "expert_indices": [ | |
| 28, | |
| 54, | |
| 109, | |
| 143, | |
| 164, | |
| 187, | |
| 234, | |
| 326 | |
| ], | |
| "routing_weights": [ | |
| 0.12403761, | |
| 0.124741495, | |
| 0.1233652, | |
| 0.12436722, | |
| 0.12376981, | |
| 0.13259424, | |
| 0.12488026, | |
| 0.12224412 | |
| ], | |
| "bytes_streamed": 939952240 | |
| }, | |
| { | |
| "layer_idx": 27, | |
| "layer_kind": "moe", | |
| "residual_norm_in": 3271.9993, | |
| "residual_norm_out": 3274.4753, | |
| "expert_indices": [ | |
| 5, | |
| 22, | |
| 23, | |
| 48, | |
| 115, | |
| 160, | |
| 188, | |
| 218 | |
| ], | |
| "routing_weights": [ | |
| 0.12905355, | |
| 0.11797063, | |
| 0.123731114, | |
| 0.11916181, | |
| 0.119591415, | |
| 0.12297361, | |
| 0.13747871, | |
| 0.13003916 | |
| ], | |
| "bytes_streamed": 939952240 | |
| }, | |
| { | |
| "layer_idx": 28, | |
| "layer_kind": "moe", | |
| "residual_norm_in": 3274.4753, | |
| "residual_norm_out": 3276.2065, | |
| "expert_indices": [ | |
| 83, | |
| 117, | |
| 178, | |
| 218, | |
| 233, | |
| 260, | |
| 305, | |
| 342 | |
| ], | |
| "routing_weights": [ | |
| 0.12889512, | |
| 0.12080206, | |
| 0.12741528, | |
| 0.119030036, | |
| 0.12386302, | |
| 0.12633918, | |
| 0.13027583, | |
| 0.12337941 | |
| ], | |
| "bytes_streamed": 939952240 | |
| }, | |
| { | |
| "layer_idx": 29, | |
| "layer_kind": "moe", | |
| "residual_norm_in": 3276.2065, | |
| "residual_norm_out": 3280.7148, | |
| "expert_indices": [ | |
| 19, | |
| 132, | |
| 145, | |
| 213, | |
| 264, | |
| 275, | |
| 309, | |
| 374 | |
| ], | |
| "routing_weights": [ | |
| 0.1201872, | |
| 0.11707843, | |
| 0.13667186, | |
| 0.11928836, | |
| 0.11894964, | |
| 0.14129789, | |
| 0.13095954, | |
| 0.115567006 | |
| ], | |
| "bytes_streamed": 939952240 | |
| }, | |
| { | |
| "layer_idx": 30, | |
| "layer_kind": "moe", | |
| "residual_norm_in": 3280.7148, | |
| "residual_norm_out": 3280.9146, | |
| "expert_indices": [ | |
| 55, | |
| 109, | |
| 165, | |
| 246, | |
| 254, | |
| 264, | |
| 275, | |
| 282 | |
| ], | |
| "routing_weights": [ | |
| 0.12114447, | |
| 0.1265742, | |
| 0.12592877, | |
| 0.13100544, | |
| 0.12474787, | |
| 0.1230107, | |
| 0.12573223, | |
| 0.12185628 | |
| ], | |
| "bytes_streamed": 1014270136 | |
| }, | |
| { | |
| "layer_idx": 31, | |
| "layer_kind": "moe", | |
| "residual_norm_in": 3280.9146, | |
| "residual_norm_out": 3284.3037, | |
| "expert_indices": [ | |
| 45, | |
| 150, | |
| 196, | |
| 198, | |
| 243, | |
| 323, | |
| 333, | |
| 359 | |
| ], | |
| "routing_weights": [ | |
| 0.124348536, | |
| 0.12571238, | |
| 0.121641114, | |
| 0.12634422, | |
| 0.12098488, | |
| 0.12598802, | |
| 0.12444443, | |
| 0.13053642 | |
| ], | |
| "bytes_streamed": 939952240 | |
| }, | |
| { | |
| "layer_idx": 32, | |
| "layer_kind": "moe", | |
| "residual_norm_in": 3284.3037, | |
| "residual_norm_out": 3285.9998, | |
| "expert_indices": [ | |
| 21, | |
| 76, | |
| 182, | |
| 270, | |
| 276, | |
| 286, | |
| 336, | |
| 345 | |
| ], | |
| "routing_weights": [ | |
| 0.123925604, | |
| 0.12669829, | |
| 0.12782235, | |
| 0.12475172, | |
| 0.12783591, | |
| 0.12554578, | |
| 0.12194816, | |
| 0.121472225 | |
| ], | |
| "bytes_streamed": 939952240 | |
| }, | |
| { | |
| "layer_idx": 33, | |
| "layer_kind": "moe", | |
| "residual_norm_in": 3285.9998, | |
| "residual_norm_out": 3285.389, | |
| "expert_indices": [ | |
| 9, | |
| 53, | |
| 54, | |
| 57, | |
| 267, | |
| 281, | |
| 303, | |
| 309 | |
| ], | |
| "routing_weights": [ | |
| 0.12560827, | |
| 0.12389935, | |
| 0.12423008, | |
| 0.12410613, | |
| 0.12650695, | |
| 0.12478609, | |
| 0.12493785, | |
| 0.12592524 | |
| ], | |
| "bytes_streamed": 915179608 | |
| }, | |
| { | |
| "layer_idx": 34, | |
| "layer_kind": "moe", | |
| "residual_norm_in": 3285.389, | |
| "residual_norm_out": 3285.5088, | |
| "expert_indices": [ | |
| 54, | |
| 96, | |
| 152, | |
| 206, | |
| 250, | |
| 295, | |
| 325, | |
| 339 | |
| ], | |
| "routing_weights": [ | |
| 0.12830864, | |
| 0.12542318, | |
| 0.12211993, | |
| 0.123667784, | |
| 0.12476751, | |
| 0.12551583, | |
| 0.123447806, | |
| 0.12674934 | |
| ], | |
| "bytes_streamed": 989497504 | |
| }, | |
| { | |
| "layer_idx": 35, | |
| "layer_kind": "moe", | |
| "residual_norm_in": 3285.5088, | |
| "residual_norm_out": 3285.8599, | |
| "expert_indices": [ | |
| 9, | |
| 68, | |
| 86, | |
| 142, | |
| 162, | |
| 176, | |
| 257, | |
| 326 | |
| ], | |
| "routing_weights": [ | |
| 0.12454217, | |
| 0.12767826, | |
| 0.124468386, | |
| 0.12491048, | |
| 0.12380852, | |
| 0.12711051, | |
| 0.122607514, | |
| 0.12487409 | |
| ], | |
| "bytes_streamed": 964724872 | |
| }, | |
| { | |
| "layer_idx": 36, | |
| "layer_kind": "moe", | |
| "residual_norm_in": 3285.8599, | |
| "residual_norm_out": 3284.272, | |
| "expert_indices": [ | |
| 51, | |
| 76, | |
| 132, | |
| 200, | |
| 236, | |
| 276, | |
| 317, | |
| 383 | |
| ], | |
| "routing_weights": [ | |
| 0.12545687, | |
| 0.12348959, | |
| 0.12646747, | |
| 0.12358548, | |
| 0.12351758, | |
| 0.123819746, | |
| 0.1276341, | |
| 0.12602918 | |
| ], | |
| "bytes_streamed": 964724872 | |
| }, | |
| { | |
| "layer_idx": 37, | |
| "layer_kind": "moe", | |
| "residual_norm_in": 3284.272, | |
| "residual_norm_out": 3279.9285, | |
| "expert_indices": [ | |
| 22, | |
| 75, | |
| 156, | |
| 182, | |
| 187, | |
| 189, | |
| 230, | |
| 321 | |
| ], | |
| "routing_weights": [ | |
| 0.12714443, | |
| 0.12565663, | |
| 0.12507114, | |
| 0.12419243, | |
| 0.12688793, | |
| 0.123175636, | |
| 0.12531842, | |
| 0.122553304 | |
| ], | |
| "bytes_streamed": 915179608 | |
| }, | |
| { | |
| "layer_idx": 38, | |
| "layer_kind": "moe", | |
| "residual_norm_in": 3279.9285, | |
| "residual_norm_out": 3279.4634, | |
| "expert_indices": [ | |
| 116, | |
| 118, | |
| 141, | |
| 142, | |
| 165, | |
| 246, | |
| 296, | |
| 305 | |
| ], | |
| "routing_weights": [ | |
| 0.12589957, | |
| 0.12569398, | |
| 0.12418391, | |
| 0.12590075, | |
| 0.12354491, | |
| 0.12647879, | |
| 0.123581976, | |
| 0.1247161 | |
| ], | |
| "bytes_streamed": 1014270136 | |
| }, | |
| { | |
| "layer_idx": 39, | |
| "layer_kind": "moe", | |
| "residual_norm_in": 3279.4634, | |
| "residual_norm_out": 3278.8074, | |
| "expert_indices": [ | |
| 1, | |
| 21, | |
| 53, | |
| 74, | |
| 153, | |
| 232, | |
| 278, | |
| 381 | |
| ], | |
| "routing_weights": [ | |
| 0.12392103, | |
| 0.12356896, | |
| 0.12509237, | |
| 0.12476749, | |
| 0.1261542, | |
| 0.1256581, | |
| 0.12628376, | |
| 0.124554105 | |
| ], | |
| "bytes_streamed": 1039042768 | |
| }, | |
| { | |
| "layer_idx": 40, | |
| "layer_kind": "moe", | |
| "residual_norm_in": 3278.8074, | |
| "residual_norm_out": 3274.6726, | |
| "expert_indices": [ | |
| 14, | |
| 54, | |
| 80, | |
| 205, | |
| 243, | |
| 350, | |
| 360, | |
| 378 | |
| ], | |
| "routing_weights": [ | |
| 0.123624906, | |
| 0.12560081, | |
| 0.12601347, | |
| 0.12464201, | |
| 0.124437995, | |
| 0.124908455, | |
| 0.1261274, | |
| 0.124645 | |
| ], | |
| "bytes_streamed": 939952240 | |
| }, | |
| { | |
| "layer_idx": 41, | |
| "layer_kind": "moe", | |
| "residual_norm_in": 3274.6726, | |
| "residual_norm_out": 3271.9517, | |
| "expert_indices": [ | |
| 0, | |
| 12, | |
| 102, | |
| 118, | |
| 143, | |
| 323, | |
| 346, | |
| 361 | |
| ], | |
| "routing_weights": [ | |
| 0.12718984, | |
| 0.12352078, | |
| 0.124371946, | |
| 0.1250337, | |
| 0.12582457, | |
| 0.12378337, | |
| 0.1258188, | |
| 0.12445698 | |
| ], | |
| "bytes_streamed": 915179608 | |
| }, | |
| { | |
| "layer_idx": 42, | |
| "layer_kind": "moe", | |
| "residual_norm_in": 3271.9517, | |
| "residual_norm_out": 3265.8472, | |
| "expert_indices": [ | |
| 11, | |
| 63, | |
| 66, | |
| 121, | |
| 233, | |
| 282, | |
| 286, | |
| 322 | |
| ], | |
| "routing_weights": [ | |
| 0.12231816, | |
| 0.12572178, | |
| 0.12679413, | |
| 0.12391465, | |
| 0.12539703, | |
| 0.122256, | |
| 0.12437119, | |
| 0.12922701 | |
| ], | |
| "bytes_streamed": 1014270136 | |
| }, | |
| { | |
| "layer_idx": 43, | |
| "layer_kind": "moe", | |
| "residual_norm_in": 3265.8472, | |
| "residual_norm_out": 3261.9087, | |
| "expert_indices": [ | |
| 48, | |
| 97, | |
| 111, | |
| 145, | |
| 166, | |
| 258, | |
| 275, | |
| 310 | |
| ], | |
| "routing_weights": [ | |
| 0.12841284, | |
| 0.12592939, | |
| 0.1230621, | |
| 0.11106126, | |
| 0.13067316, | |
| 0.12302471, | |
| 0.12871376, | |
| 0.1291227 | |
| ], | |
| "bytes_streamed": 1039042768 | |
| }, | |
| { | |
| "layer_idx": 44, | |
| "layer_kind": "moe", | |
| "residual_norm_in": 3261.9087, | |
| "residual_norm_out": 3253.6006, | |
| "expert_indices": [ | |
| 52, | |
| 80, | |
| 135, | |
| 143, | |
| 145, | |
| 204, | |
| 209, | |
| 313 | |
| ], | |
| "routing_weights": [ | |
| 0.1505961, | |
| 0.15023981, | |
| 0.11775557, | |
| 0.087679155, | |
| 0.12829639, | |
| 0.094963625, | |
| 0.11943085, | |
| 0.15103848 | |
| ], | |
| "bytes_streamed": 915179608 | |
| }, | |
| { | |
| "layer_idx": 45, | |
| "layer_kind": "moe", | |
| "residual_norm_in": 3253.6006, | |
| "residual_norm_out": 3250.8918, | |
| "expert_indices": [ | |
| 20, | |
| 45, | |
| 57, | |
| 59, | |
| 105, | |
| 176, | |
| 271, | |
| 320 | |
| ], | |
| "routing_weights": [ | |
| 0.12892541, | |
| 0.11697536, | |
| 0.14063151, | |
| 0.12651275, | |
| 0.13517754, | |
| 0.09021728, | |
| 0.13699076, | |
| 0.12456939 | |
| ], | |
| "bytes_streamed": 964724872 | |
| }, | |
| { | |
| "layer_idx": 46, | |
| "layer_kind": "moe", | |
| "residual_norm_in": 3250.8918, | |
| "residual_norm_out": 3246.15, | |
| "expert_indices": [ | |
| 7, | |
| 13, | |
| 31, | |
| 58, | |
| 198, | |
| 206, | |
| 222, | |
| 361 | |
| ], | |
| "routing_weights": [ | |
| 0.13090932, | |
| 0.13061666, | |
| 0.119749196, | |
| 0.13284738, | |
| 0.13148266, | |
| 0.12531951, | |
| 0.102036834, | |
| 0.1270384 | |
| ], | |
| "bytes_streamed": 989497504 | |
| }, | |
| { | |
| "layer_idx": 47, | |
| "layer_kind": "moe", | |
| "residual_norm_in": 3246.15, | |
| "residual_norm_out": 3244.434, | |
| "expert_indices": [ | |
| 50, | |
| 91, | |
| 248, | |
| 268, | |
| 274, | |
| 305, | |
| 359, | |
| 377 | |
| ], | |
| "routing_weights": [ | |
| 0.090512484, | |
| 0.12572679, | |
| 0.12560984, | |
| 0.13680649, | |
| 0.13461365, | |
| 0.1383419, | |
| 0.11906946, | |
| 0.12931941 | |
| ], | |
| "bytes_streamed": 989497504 | |
| }, | |
| { | |
| "layer_idx": 48, | |
| "layer_kind": "moe", | |
| "residual_norm_in": 3244.434, | |
| "residual_norm_out": 3237.2173, | |
| "expert_indices": [ | |
| 21, | |
| 140, | |
| 180, | |
| 211, | |
| 216, | |
| 265, | |
| 275, | |
| 356 | |
| ], | |
| "routing_weights": [ | |
| 0.089280955, | |
| 0.11947252, | |
| 0.14138731, | |
| 0.13919227, | |
| 0.13082214, | |
| 0.12825514, | |
| 0.12513566, | |
| 0.1264539 | |
| ], | |
| "bytes_streamed": 964724872 | |
| }, | |
| { | |
| "layer_idx": 49, | |
| "layer_kind": "moe", | |
| "residual_norm_in": 3237.2173, | |
| "residual_norm_out": 3225.3926, | |
| "expert_indices": [ | |
| 65, | |
| 132, | |
| 136, | |
| 141, | |
| 257, | |
| 330, | |
| 331, | |
| 381 | |
| ], | |
| "routing_weights": [ | |
| 0.15157294, | |
| 0.14870882, | |
| 0.12575087, | |
| 0.14817211, | |
| 0.12935574, | |
| 0.12963831, | |
| 0.084066786, | |
| 0.082734436 | |
| ], | |
| "bytes_streamed": 964724872 | |
| }, | |
| { | |
| "layer_idx": 50, | |
| "layer_kind": "moe", | |
| "residual_norm_in": 3225.3926, | |
| "residual_norm_out": 3213.6804, | |
| "expert_indices": [ | |
| 2, | |
| 46, | |
| 77, | |
| 114, | |
| 160, | |
| 162, | |
| 175, | |
| 324 | |
| ], | |
| "routing_weights": [ | |
| 0.07168002, | |
| 0.13420905, | |
| 0.14428033, | |
| 0.13734862, | |
| 0.14445138, | |
| 0.13382252, | |
| 0.11261744, | |
| 0.121590614 | |
| ], | |
| "bytes_streamed": 964724872 | |
| }, | |
| { | |
| "layer_idx": 51, | |
| "layer_kind": "moe", | |
| "residual_norm_in": 3213.6804, | |
| "residual_norm_out": 3204.1482, | |
| "expert_indices": [ | |
| 36, | |
| 130, | |
| 189, | |
| 218, | |
| 263, | |
| 271, | |
| 317, | |
| 362 | |
| ], | |
| "routing_weights": [ | |
| 0.07808468, | |
| 0.14101124, | |
| 0.13042147, | |
| 0.11585922, | |
| 0.13718921, | |
| 0.12932168, | |
| 0.12141402, | |
| 0.1466985 | |
| ], | |
| "bytes_streamed": 964724872 | |
| }, | |
| { | |
| "layer_idx": 52, | |
| "layer_kind": "moe", | |
| "residual_norm_in": 3204.1482, | |
| "residual_norm_out": 3186.4143, | |
| "expert_indices": [ | |
| 10, | |
| 31, | |
| 40, | |
| 49, | |
| 52, | |
| 96, | |
| 259, | |
| 342 | |
| ], | |
| "routing_weights": [ | |
| 0.13195448, | |
| 0.12658952, | |
| 0.14112595, | |
| 0.11027296, | |
| 0.16430713, | |
| 0.07610071, | |
| 0.12522088, | |
| 0.12442839 | |
| ], | |
| "bytes_streamed": 915179608 | |
| }, | |
| { | |
| "layer_idx": 53, | |
| "layer_kind": "moe", | |
| "residual_norm_in": 3186.4143, | |
| "residual_norm_out": 3151.149, | |
| "expert_indices": [ | |
| 42, | |
| 101, | |
| 131, | |
| 247, | |
| 272, | |
| 341, | |
| 371, | |
| 383 | |
| ], | |
| "routing_weights": [ | |
| 0.06139395, | |
| 0.17122178, | |
| 0.14294012, | |
| 0.10139634, | |
| 0.056430835, | |
| 0.16060388, | |
| 0.14770874, | |
| 0.15830438 | |
| ], | |
| "bytes_streamed": 989497504 | |
| }, | |
| { | |
| "layer_idx": 54, | |
| "layer_kind": "moe", | |
| "residual_norm_in": 3151.149, | |
| "residual_norm_out": 3040.3433, | |
| "expert_indices": [ | |
| 68, | |
| 156, | |
| 169, | |
| 189, | |
| 192, | |
| 218, | |
| 240, | |
| 319 | |
| ], | |
| "routing_weights": [ | |
| 0.053958632, | |
| 0.05449725, | |
| 0.14695823, | |
| 0.14635251, | |
| 0.15215138, | |
| 0.1491312, | |
| 0.14903401, | |
| 0.14791681 | |
| ], | |
| "bytes_streamed": 989497504 | |
| }, | |
| { | |
| "layer_idx": 55, | |
| "layer_kind": "moe", | |
| "residual_norm_in": 3040.3433, | |
| "residual_norm_out": 2974.9587, | |
| "expert_indices": [ | |
| 42, | |
| 87, | |
| 173, | |
| 246, | |
| 308, | |
| 309, | |
| 329, | |
| 348 | |
| ], | |
| "routing_weights": [ | |
| 0.05151474, | |
| 0.1472363, | |
| 0.1606687, | |
| 0.16340461, | |
| 0.15019181, | |
| 0.1526328, | |
| 0.11705792, | |
| 0.05729307 | |
| ], | |
| "bytes_streamed": 989497504 | |
| }, | |
| { | |
| "layer_idx": 56, | |
| "layer_kind": "moe", | |
| "residual_norm_in": 2974.9587, | |
| "residual_norm_out": 2928.6697, | |
| "expert_indices": [ | |
| 26, | |
| 67, | |
| 74, | |
| 84, | |
| 253, | |
| 316, | |
| 319, | |
| 363 | |
| ], | |
| "routing_weights": [ | |
| 0.16366942, | |
| 0.16105407, | |
| 0.16047254, | |
| 0.07660575, | |
| 0.07692882, | |
| 0.16656852, | |
| 0.123046815, | |
| 0.07165412 | |
| ], | |
| "bytes_streamed": 964724872 | |
| }, | |
| { | |
| "layer_idx": 57, | |
| "layer_kind": "moe", | |
| "residual_norm_in": 2928.6697, | |
| "residual_norm_out": 2827.0847, | |
| "expert_indices": [ | |
| 0, | |
| 30, | |
| 80, | |
| 84, | |
| 107, | |
| 222, | |
| 253, | |
| 288 | |
| ], | |
| "routing_weights": [ | |
| 0.18186912, | |
| 0.052621696, | |
| 0.12891348, | |
| 0.05427989, | |
| 0.1740672, | |
| 0.18127596, | |
| 0.17727546, | |
| 0.04969726 | |
| ], | |
| "bytes_streamed": 890406976 | |
| }, | |
| { | |
| "layer_idx": 58, | |
| "layer_kind": "moe", | |
| "residual_norm_in": 2827.0847, | |
| "residual_norm_out": 2604.6702, | |
| "expert_indices": [ | |
| 5, | |
| 234, | |
| 259, | |
| 279, | |
| 329, | |
| 344, | |
| 350, | |
| 359 | |
| ], | |
| "routing_weights": [ | |
| 0.12612459, | |
| 0.1269469, | |
| 0.12518732, | |
| 0.12676597, | |
| 0.1180102, | |
| 0.12671871, | |
| 0.12414087, | |
| 0.12610547 | |
| ], | |
| "bytes_streamed": 964724872 | |
| }, | |
| { | |
| "layer_idx": 59, | |
| "layer_kind": "moe", | |
| "residual_norm_in": 2604.6702, | |
| "residual_norm_out": 3559.2024, | |
| "expert_indices": [ | |
| 47, | |
| 62, | |
| 83, | |
| 156, | |
| 161, | |
| 176, | |
| 372, | |
| 374 | |
| ], | |
| "routing_weights": [ | |
| 0.09576432, | |
| 0.07162925, | |
| 0.061123345, | |
| 0.07196141, | |
| 0.062052693, | |
| 0.48133266, | |
| 0.085414946, | |
| 0.07072137 | |
| ], | |
| "bytes_streamed": 964724872 | |
| }, | |
| { | |
| "layer_idx": 60, | |
| "layer_kind": "moe", | |
| "residual_norm_in": 3559.2024, | |
| "residual_norm_out": 9136.94, | |
| "expert_indices": [ | |
| 3, | |
| 55, | |
| 68, | |
| 73, | |
| 217, | |
| 234, | |
| 242, | |
| 378 | |
| ], | |
| "routing_weights": [ | |
| 0.03411237, | |
| 0.039301682, | |
| 0.1990229, | |
| 0.18033567, | |
| 0.074357, | |
| 0.2359254, | |
| 0.19573748, | |
| 0.041207485 | |
| ], | |
| "bytes_streamed": 939952240 | |
| } | |
| ] | |
| }, | |
| { | |
| "step_index": 4, | |
| "input_token_ids": [ | |
| 163584, | |
| 950, | |
| 1268, | |
| 426, | |
| 120287 | |
| ], | |
| "output_token_id": 63, | |
| "output_logit": 18.10574, | |
| "logits_digest": "aabece41b00114d4b0e831f17565a9686fc43c3742440f4f5b65fec22546ee1e", | |
| "hidden_digest": "a8ae2899ead71428e29defa6eed423e74766f7b0f0dcb1f14305bee6b9e38818", | |
| "top_k_tokens": [ | |
| [ | |
| 63, | |
| 18.10574 | |
| ], | |
| [ | |
| 12, | |
| 15.880532 | |
| ], | |
| [ | |
| 2343, | |
| 15.595577 | |
| ], | |
| [ | |
| 9971, | |
| 15.309732 | |
| ], | |
| [ | |
| 2390, | |
| 15.273062 | |
| ], | |
| [ | |
| 3868, | |
| 15.156176 | |
| ], | |
| [ | |
| 220, | |
| 14.71898 | |
| ], | |
| [ | |
| 1993, | |
| 14.635267 | |
| ], | |
| [ | |
| 382, | |
| 14.507246 | |
| ], | |
| [ | |
| 7355, | |
| 14.249199 | |
| ], | |
| [ | |
| 7690, | |
| 14.193375 | |
| ], | |
| [ | |
| 76485, | |
| 14.176044 | |
| ], | |
| [ | |
| 6820, | |
| 14.170868 | |
| ], | |
| [ | |
| 6, | |
| 14.105271 | |
| ], | |
| [ | |
| 1268, | |
| 14.091308 | |
| ], | |
| [ | |
| 1517, | |
| 14.066706 | |
| ] | |
| ], | |
| "entropy": 4.3132534, | |
| "margin": 2.2252073, | |
| "bytes_streamed": 69199514920, | |
| "layers": [ | |
| { | |
| "layer_idx": 0, | |
| "layer_kind": "dense", | |
| "residual_norm_in": 1.6775731, | |
| "residual_norm_out": 2.2587411, | |
| "expert_indices": null, | |
| "routing_weights": null, | |
| "bytes_streamed": 995000320 | |
| }, | |
| { | |
| "layer_idx": 1, | |
| "layer_kind": "moe", | |
| "residual_norm_in": 2.2587411, | |
| "residual_norm_out": 3.55967, | |
| "expert_indices": [ | |
| 5, | |
| 33, | |
| 72, | |
| 105, | |
| 116, | |
| 154, | |
| 182, | |
| 272 | |
| ], | |
| "routing_weights": [ | |
| 0.087083995, | |
| 0.20157112, | |
| 0.14540455, | |
| 0.08845074, | |
| 0.1286529, | |
| 0.10869927, | |
| 0.12787405, | |
| 0.112263314 | |
| ], | |
| "bytes_streamed": 1039042768 | |
| }, | |
| { | |
| "layer_idx": 2, | |
| "layer_kind": "moe", | |
| "residual_norm_in": 3.55967, | |
| "residual_norm_out": 4.9289865, | |
| "expert_indices": [ | |
| 7, | |
| 25, | |
| 92, | |
| 179, | |
| 190, | |
| 240, | |
| 340, | |
| 343 | |
| ], | |
| "routing_weights": [ | |
| 0.1776416, | |
| 0.068654895, | |
| 0.1582383, | |
| 0.0638775, | |
| 0.12181032, | |
| 0.18464753, | |
| 0.060134597, | |
| 0.16499528 | |
| ], | |
| "bytes_streamed": 890406976 | |
| }, | |
| { | |
| "layer_idx": 3, | |
| "layer_kind": "moe", | |
| "residual_norm_in": 4.9289865, | |
| "residual_norm_out": 10.357709, | |
| "expert_indices": [ | |
| 15, | |
| 189, | |
| 190, | |
| 217, | |
| 253, | |
| 271, | |
| 337, | |
| 353 | |
| ], | |
| "routing_weights": [ | |
| 0.05684734, | |
| 0.062119197, | |
| 0.052167017, | |
| 0.35753912, | |
| 0.06827293, | |
| 0.049369186, | |
| 0.04669538, | |
| 0.30698988 | |
| ], | |
| "bytes_streamed": 1039042768 | |
| }, | |
| { | |
| "layer_idx": 4, | |
| "layer_kind": "moe", | |
| "residual_norm_in": 10.357709, | |
| "residual_norm_out": 78.47345, | |
| "expert_indices": [ | |
| 52, | |
| 98, | |
| 106, | |
| 109, | |
| 120, | |
| 189, | |
| 191, | |
| 324 | |
| ], | |
| "routing_weights": [ | |
| 0.07070838, | |
| 0.08047035, | |
| 0.077183425, | |
| 0.08113832, | |
| 0.071013846, | |
| 0.47025713, | |
| 0.08182316, | |
| 0.067405246 | |
| ], | |
| "bytes_streamed": 1113360664 | |
| }, | |
| { | |
| "layer_idx": 5, | |
| "layer_kind": "moe", | |
| "residual_norm_in": 78.47345, | |
| "residual_norm_out": 84.63074, | |
| "expert_indices": [ | |
| 12, | |
| 17, | |
| 121, | |
| 124, | |
| 160, | |
| 186, | |
| 195, | |
| 237 | |
| ], | |
| "routing_weights": [ | |
| 0.1888245, | |
| 0.10076529, | |
| 0.11140824, | |
| 0.17085913, | |
| 0.10019477, | |
| 0.10386934, | |
| 0.095480084, | |
| 0.12859854 | |
| ], | |
| "bytes_streamed": 989497504 | |
| }, | |
| { | |
| "layer_idx": 6, | |
| "layer_kind": "moe", | |
| "residual_norm_in": 84.63074, | |
| "residual_norm_out": 390.0783, | |
| "expert_indices": [ | |
| 18, | |
| 45, | |
| 88, | |
| 239, | |
| 243, | |
| 269, | |
| 334, | |
| 346 | |
| ], | |
| "routing_weights": [ | |
| 0.046305433, | |
| 0.029368749, | |
| 0.033812124, | |
| 0.031240184, | |
| 0.5246116, | |
| 0.0594912, | |
| 0.037579488, | |
| 0.23759124 | |
| ], | |
| "bytes_streamed": 1063815400 | |
| }, | |
| { | |
| "layer_idx": 7, | |
| "layer_kind": "moe", | |
| "residual_norm_in": 390.0783, | |
| "residual_norm_out": 1106.5925, | |
| "expert_indices": [ | |
| 54, | |
| 57, | |
| 155, | |
| 183, | |
| 185, | |
| 192, | |
| 302, | |
| 361 | |
| ], | |
| "routing_weights": [ | |
| 0.025255695, | |
| 0.0061686267, | |
| 0.007244412, | |
| 0.92817426, | |
| 0.009681543, | |
| 0.0078814775, | |
| 0.0067851017, | |
| 0.008808859 | |
| ], | |
| "bytes_streamed": 1039042768 | |
| }, | |
| { | |
| "layer_idx": 8, | |
| "layer_kind": "moe", | |
| "residual_norm_in": 1106.5925, | |
| "residual_norm_out": 1492.6508, | |
| "expert_indices": [ | |
| 86, | |
| 128, | |
| 175, | |
| 287, | |
| 311, | |
| 336, | |
| 337, | |
| 351 | |
| ], | |
| "routing_weights": [ | |
| 0.06048171, | |
| 0.030799579, | |
| 0.030457757, | |
| 0.038466815, | |
| 0.029505648, | |
| 0.032741938, | |
| 0.034052644, | |
| 0.743494 | |
| ], | |
| "bytes_streamed": 1063815400 | |
| }, | |
| { | |
| "layer_idx": 9, | |
| "layer_kind": "moe", | |
| "residual_norm_in": 1492.6508, | |
| "residual_norm_out": 1690.9762, | |
| "expert_indices": [ | |
| 30, | |
| 131, | |
| 182, | |
| 183, | |
| 198, | |
| 238, | |
| 365, | |
| 377 | |
| ], | |
| "routing_weights": [ | |
| 0.056319162, | |
| 0.047798146, | |
| 0.056981582, | |
| 0.05191267, | |
| 0.04586145, | |
| 0.057513874, | |
| 0.045446966, | |
| 0.63816607 | |
| ], | |
| "bytes_streamed": 1113360664 | |
| }, | |
| { | |
| "layer_idx": 10, | |
| "layer_kind": "moe", | |
| "residual_norm_in": 1690.9762, | |
| "residual_norm_out": 1954.9926, | |
| "expert_indices": [ | |
| 5, | |
| 34, | |
| 46, | |
| 118, | |
| 234, | |
| 240, | |
| 292, | |
| 296 | |
| ], | |
| "routing_weights": [ | |
| 0.033065867, | |
| 0.015240303, | |
| 0.01521005, | |
| 0.015278836, | |
| 0.014619413, | |
| 0.013330309, | |
| 0.017498873, | |
| 0.87575644 | |
| ], | |
| "bytes_streamed": 1014270136 | |
| }, | |
| { | |
| "layer_idx": 11, | |
| "layer_kind": "moe", | |
| "residual_norm_in": 1954.9926, | |
| "residual_norm_out": 1956.8656, | |
| "expert_indices": [ | |
| 31, | |
| 124, | |
| 141, | |
| 180, | |
| 224, | |
| 237, | |
| 252, | |
| 253 | |
| ], | |
| "routing_weights": [ | |
| 0.13398919, | |
| 0.12286291, | |
| 0.119516574, | |
| 0.12470392, | |
| 0.13138652, | |
| 0.11974144, | |
| 0.12428722, | |
| 0.12351225 | |
| ], | |
| "bytes_streamed": 1113360664 | |
| }, | |
| { | |
| "layer_idx": 12, | |
| "layer_kind": "moe", | |
| "residual_norm_in": 1956.8656, | |
| "residual_norm_out": 1980.6772, | |
| "expert_indices": [ | |
| 25, | |
| 51, | |
| 92, | |
| 95, | |
| 139, | |
| 189, | |
| 270, | |
| 371 | |
| ], | |
| "routing_weights": [ | |
| 0.11804446, | |
| 0.12368671, | |
| 0.115437895, | |
| 0.12428875, | |
| 0.12710269, | |
| 0.13917556, | |
| 0.12618789, | |
| 0.12607609 | |
| ], | |
| "bytes_streamed": 1088588032 | |
| }, | |
| { | |
| "layer_idx": 13, | |
| "layer_kind": "moe", | |
| "residual_norm_in": 1980.6772, | |
| "residual_norm_out": 1982.7928, | |
| "expert_indices": [ | |
| 65, | |
| 68, | |
| 103, | |
| 194, | |
| 203, | |
| 228, | |
| 312, | |
| 348 | |
| ], | |
| "routing_weights": [ | |
| 0.15230049, | |
| 0.12651007, | |
| 0.12470709, | |
| 0.11999359, | |
| 0.12181889, | |
| 0.116762176, | |
| 0.12590373, | |
| 0.112003945 | |
| ], | |
| "bytes_streamed": 989497504 | |
| }, | |
| { | |
| "layer_idx": 14, | |
| "layer_kind": "moe", | |
| "residual_norm_in": 1982.7928, | |
| "residual_norm_out": 1985.4446, | |
| "expert_indices": [ | |
| 166, | |
| 182, | |
| 204, | |
| 207, | |
| 262, | |
| 269, | |
| 303, | |
| 377 | |
| ], | |
| "routing_weights": [ | |
| 0.12826353, | |
| 0.11211658, | |
| 0.16714138, | |
| 0.11411119, | |
| 0.12515004, | |
| 0.110577546, | |
| 0.118941896, | |
| 0.12369785 | |
| ], | |
| "bytes_streamed": 989497504 | |
| }, | |
| { | |
| "layer_idx": 15, | |
| "layer_kind": "moe", | |
| "residual_norm_in": 1985.4446, | |
| "residual_norm_out": 2258.33, | |
| "expert_indices": [ | |
| 11, | |
| 109, | |
| 136, | |
| 150, | |
| 194, | |
| 212, | |
| 224, | |
| 279 | |
| ], | |
| "routing_weights": [ | |
| 0.07922835, | |
| 0.08260134, | |
| 0.08270256, | |
| 0.07873966, | |
| 0.08021491, | |
| 0.08108263, | |
| 0.07728814, | |
| 0.43814245 | |
| ], | |
| "bytes_streamed": 1039042768 | |
| }, | |
| { | |
| "layer_idx": 16, | |
| "layer_kind": "moe", | |
| "residual_norm_in": 2258.33, | |
| "residual_norm_out": 2261.918, | |
| "expert_indices": [ | |
| 19, | |
| 80, | |
| 89, | |
| 91, | |
| 204, | |
| 225, | |
| 249, | |
| 339 | |
| ], | |
| "routing_weights": [ | |
| 0.11132005, | |
| 0.12770416, | |
| 0.12904347, | |
| 0.123933, | |
| 0.13755625, | |
| 0.11228153, | |
| 0.14229253, | |
| 0.115868956 | |
| ], | |
| "bytes_streamed": 989497504 | |
| }, | |
| { | |
| "layer_idx": 17, | |
| "layer_kind": "moe", | |
| "residual_norm_in": 2261.918, | |
| "residual_norm_out": 2275.189, | |
| "expert_indices": [ | |
| 81, | |
| 113, | |
| 128, | |
| 161, | |
| 282, | |
| 303, | |
| 304, | |
| 364 | |
| ], | |
| "routing_weights": [ | |
| 0.13470753, | |
| 0.12498505, | |
| 0.13510035, | |
| 0.13626955, | |
| 0.10787492, | |
| 0.10595762, | |
| 0.1165155, | |
| 0.13858946 | |
| ], | |
| "bytes_streamed": 1113360664 | |
| }, | |
| { | |
| "layer_idx": 18, | |
| "layer_kind": "moe", | |
| "residual_norm_in": 2275.189, | |
| "residual_norm_out": 2846.2942, | |
| "expert_indices": [ | |
| 30, | |
| 72, | |
| 95, | |
| 143, | |
| 230, | |
| 309, | |
| 329, | |
| 378 | |
| ], | |
| "routing_weights": [ | |
| 0.06173668, | |
| 0.060805954, | |
| 0.059534308, | |
| 0.546107, | |
| 0.05942752, | |
| 0.06282742, | |
| 0.059586395, | |
| 0.08997479 | |
| ], | |
| "bytes_streamed": 939952240 | |
| }, | |
| { | |
| "layer_idx": 19, | |
| "layer_kind": "moe", | |
| "residual_norm_in": 2846.2942, | |
| "residual_norm_out": 2848.7095, | |
| "expert_indices": [ | |
| 28, | |
| 62, | |
| 84, | |
| 210, | |
| 258, | |
| 264, | |
| 278, | |
| 318 | |
| ], | |
| "routing_weights": [ | |
| 0.12718764, | |
| 0.11587793, | |
| 0.116436146, | |
| 0.11349043, | |
| 0.12945548, | |
| 0.12185279, | |
| 0.15666124, | |
| 0.1190383 | |
| ], | |
| "bytes_streamed": 1088588032 | |
| }, | |
| { | |
| "layer_idx": 20, | |
| "layer_kind": "moe", | |
| "residual_norm_in": 2848.7095, | |
| "residual_norm_out": 2964.7712, | |
| "expert_indices": [ | |
| 7, | |
| 108, | |
| 146, | |
| 227, | |
| 260, | |
| 273, | |
| 329, | |
| 351 | |
| ], | |
| "routing_weights": [ | |
| 0.120410115, | |
| 0.12045736, | |
| 0.11941096, | |
| 0.116442, | |
| 0.15163471, | |
| 0.122804046, | |
| 0.13331619, | |
| 0.11552454 | |
| ], | |
| "bytes_streamed": 1063815400 | |
| }, | |
| { | |
| "layer_idx": 21, | |
| "layer_kind": "moe", | |
| "residual_norm_in": 2964.7712, | |
| "residual_norm_out": 2998.4875, | |
| "expert_indices": [ | |
| 8, | |
| 24, | |
| 140, | |
| 151, | |
| 250, | |
| 264, | |
| 313, | |
| 342 | |
| ], | |
| "routing_weights": [ | |
| 0.12219975, | |
| 0.11388808, | |
| 0.11746955, | |
| 0.11987377, | |
| 0.12484145, | |
| 0.122165635, | |
| 0.13231997, | |
| 0.14724179 | |
| ], | |
| "bytes_streamed": 1138133296 | |
| }, | |
| { | |
| "layer_idx": 22, | |
| "layer_kind": "moe", | |
| "residual_norm_in": 2998.4875, | |
| "residual_norm_out": 3001.0393, | |
| "expert_indices": [ | |
| 3, | |
| 11, | |
| 78, | |
| 119, | |
| 206, | |
| 236, | |
| 249, | |
| 334 | |
| ], | |
| "routing_weights": [ | |
| 0.11585083, | |
| 0.12601939, | |
| 0.12537284, | |
| 0.12319891, | |
| 0.13266194, | |
| 0.11640788, | |
| 0.12136969, | |
| 0.13911851 | |
| ], | |
| "bytes_streamed": 964724872 | |
| }, | |
| { | |
| "layer_idx": 23, | |
| "layer_kind": "moe", | |
| "residual_norm_in": 3001.0393, | |
| "residual_norm_out": 3124.1562, | |
| "expert_indices": [ | |
| 8, | |
| 44, | |
| 125, | |
| 207, | |
| 209, | |
| 226, | |
| 283, | |
| 325 | |
| ], | |
| "routing_weights": [ | |
| 0.1277313, | |
| 0.11835284, | |
| 0.13288988, | |
| 0.119071715, | |
| 0.117436275, | |
| 0.12930648, | |
| 0.11873839, | |
| 0.13647313 | |
| ], | |
| "bytes_streamed": 1063815400 | |
| }, | |
| { | |
| "layer_idx": 24, | |
| "layer_kind": "moe", | |
| "residual_norm_in": 3124.1562, | |
| "residual_norm_out": 3263.9944, | |
| "expert_indices": [ | |
| 11, | |
| 46, | |
| 58, | |
| 59, | |
| 89, | |
| 223, | |
| 294, | |
| 376 | |
| ], | |
| "routing_weights": [ | |
| 0.13790713, | |
| 0.12226996, | |
| 0.12022834, | |
| 0.13080284, | |
| 0.11101902, | |
| 0.122765206, | |
| 0.1395905, | |
| 0.11541706 | |
| ], | |
| "bytes_streamed": 1088588032 | |
| }, | |
| { | |
| "layer_idx": 25, | |
| "layer_kind": "moe", | |
| "residual_norm_in": 3263.9944, | |
| "residual_norm_out": 3268.9998, | |
| "expert_indices": [ | |
| 32, | |
| 52, | |
| 186, | |
| 275, | |
| 292, | |
| 317, | |
| 325, | |
| 355 | |
| ], | |
| "routing_weights": [ | |
| 0.1295867, | |
| 0.1348981, | |
| 0.11795012, | |
| 0.12863119, | |
| 0.12783831, | |
| 0.12071772, | |
| 0.11842303, | |
| 0.12195492 | |
| ], | |
| "bytes_streamed": 1063815400 | |
| }, | |
| { | |
| "layer_idx": 26, | |
| "layer_kind": "moe", | |
| "residual_norm_in": 3268.9998, | |
| "residual_norm_out": 3272.0737, | |
| "expert_indices": [ | |
| 28, | |
| 54, | |
| 109, | |
| 143, | |
| 164, | |
| 187, | |
| 234, | |
| 326 | |
| ], | |
| "routing_weights": [ | |
| 0.12403761, | |
| 0.124741495, | |
| 0.1233652, | |
| 0.12436722, | |
| 0.12376981, | |
| 0.13259424, | |
| 0.12488026, | |
| 0.12224412 | |
| ], | |
| "bytes_streamed": 1063815400 | |
| }, | |
| { | |
| "layer_idx": 27, | |
| "layer_kind": "moe", | |
| "residual_norm_in": 3272.0737, | |
| "residual_norm_out": 3274.5671, | |
| "expert_indices": [ | |
| 5, | |
| 22, | |
| 23, | |
| 48, | |
| 115, | |
| 160, | |
| 188, | |
| 218 | |
| ], | |
| "routing_weights": [ | |
| 0.12905355, | |
| 0.11797063, | |
| 0.123731114, | |
| 0.11916181, | |
| 0.119591415, | |
| 0.12297361, | |
| 0.13747871, | |
| 0.13003916 | |
| ], | |
| "bytes_streamed": 1063815400 | |
| }, | |
| { | |
| "layer_idx": 28, | |
| "layer_kind": "moe", | |
| "residual_norm_in": 3274.5671, | |
| "residual_norm_out": 3276.3123, | |
| "expert_indices": [ | |
| 83, | |
| 117, | |
| 178, | |
| 218, | |
| 233, | |
| 260, | |
| 305, | |
| 342 | |
| ], | |
| "routing_weights": [ | |
| 0.12889512, | |
| 0.12080206, | |
| 0.12741528, | |
| 0.119030036, | |
| 0.12386302, | |
| 0.12633918, | |
| 0.13027583, | |
| 0.12337941 | |
| ], | |
| "bytes_streamed": 1014270136 | |
| }, | |
| { | |
| "layer_idx": 29, | |
| "layer_kind": "moe", | |
| "residual_norm_in": 3276.3123, | |
| "residual_norm_out": 3280.83, | |
| "expert_indices": [ | |
| 19, | |
| 132, | |
| 145, | |
| 213, | |
| 264, | |
| 275, | |
| 309, | |
| 374 | |
| ], | |
| "routing_weights": [ | |
| 0.1201872, | |
| 0.11707843, | |
| 0.13667186, | |
| 0.11928836, | |
| 0.11894964, | |
| 0.14129789, | |
| 0.13095954, | |
| 0.115567006 | |
| ], | |
| "bytes_streamed": 1063815400 | |
| }, | |
| { | |
| "layer_idx": 30, | |
| "layer_kind": "moe", | |
| "residual_norm_in": 3280.83, | |
| "residual_norm_out": 3281.1006, | |
| "expert_indices": [ | |
| 55, | |
| 109, | |
| 165, | |
| 246, | |
| 254, | |
| 264, | |
| 275, | |
| 282 | |
| ], | |
| "routing_weights": [ | |
| 0.12114447, | |
| 0.1265742, | |
| 0.12592877, | |
| 0.13100544, | |
| 0.12474787, | |
| 0.1230107, | |
| 0.12573223, | |
| 0.12185628 | |
| ], | |
| "bytes_streamed": 1088588032 | |
| }, | |
| { | |
| "layer_idx": 31, | |
| "layer_kind": "moe", | |
| "residual_norm_in": 3281.1006, | |
| "residual_norm_out": 3284.5334, | |
| "expert_indices": [ | |
| 45, | |
| 150, | |
| 196, | |
| 198, | |
| 243, | |
| 323, | |
| 333, | |
| 359 | |
| ], | |
| "routing_weights": [ | |
| 0.124348536, | |
| 0.12571238, | |
| 0.121641114, | |
| 0.12634422, | |
| 0.12098488, | |
| 0.12598802, | |
| 0.12444443, | |
| 0.13053642 | |
| ], | |
| "bytes_streamed": 1063815400 | |
| }, | |
| { | |
| "layer_idx": 32, | |
| "layer_kind": "moe", | |
| "residual_norm_in": 3284.5334, | |
| "residual_norm_out": 3286.2473, | |
| "expert_indices": [ | |
| 21, | |
| 76, | |
| 182, | |
| 270, | |
| 276, | |
| 286, | |
| 336, | |
| 345 | |
| ], | |
| "routing_weights": [ | |
| 0.123925604, | |
| 0.12669829, | |
| 0.12782235, | |
| 0.12475172, | |
| 0.12783591, | |
| 0.12554578, | |
| 0.12194816, | |
| 0.121472225 | |
| ], | |
| "bytes_streamed": 1039042768 | |
| }, | |
| { | |
| "layer_idx": 33, | |
| "layer_kind": "moe", | |
| "residual_norm_in": 3286.2473, | |
| "residual_norm_out": 3285.6956, | |
| "expert_indices": [ | |
| 9, | |
| 53, | |
| 54, | |
| 57, | |
| 267, | |
| 281, | |
| 303, | |
| 309 | |
| ], | |
| "routing_weights": [ | |
| 0.12560827, | |
| 0.12389935, | |
| 0.12423008, | |
| 0.12410613, | |
| 0.12650695, | |
| 0.12478609, | |
| 0.12493785, | |
| 0.12592524 | |
| ], | |
| "bytes_streamed": 1039042768 | |
| }, | |
| { | |
| "layer_idx": 34, | |
| "layer_kind": "moe", | |
| "residual_norm_in": 3285.6956, | |
| "residual_norm_out": 3286.0151, | |
| "expert_indices": [ | |
| 54, | |
| 96, | |
| 152, | |
| 206, | |
| 250, | |
| 295, | |
| 325, | |
| 339 | |
| ], | |
| "routing_weights": [ | |
| 0.12830864, | |
| 0.12542318, | |
| 0.12211993, | |
| 0.123667784, | |
| 0.12476751, | |
| 0.12551583, | |
| 0.123447806, | |
| 0.12674934 | |
| ], | |
| "bytes_streamed": 1138133296 | |
| }, | |
| { | |
| "layer_idx": 35, | |
| "layer_kind": "moe", | |
| "residual_norm_in": 3286.0151, | |
| "residual_norm_out": 3286.385, | |
| "expert_indices": [ | |
| 9, | |
| 68, | |
| 86, | |
| 142, | |
| 162, | |
| 176, | |
| 257, | |
| 326 | |
| ], | |
| "routing_weights": [ | |
| 0.12454217, | |
| 0.12767826, | |
| 0.124468386, | |
| 0.12491048, | |
| 0.12380852, | |
| 0.12711051, | |
| 0.122607514, | |
| 0.12487409 | |
| ], | |
| "bytes_streamed": 1063815400 | |
| }, | |
| { | |
| "layer_idx": 36, | |
| "layer_kind": "moe", | |
| "residual_norm_in": 3286.385, | |
| "residual_norm_out": 3284.8843, | |
| "expert_indices": [ | |
| 51, | |
| 76, | |
| 132, | |
| 200, | |
| 236, | |
| 276, | |
| 317, | |
| 383 | |
| ], | |
| "routing_weights": [ | |
| 0.12545687, | |
| 0.12348959, | |
| 0.12646747, | |
| 0.12358548, | |
| 0.12351758, | |
| 0.123819746, | |
| 0.1276341, | |
| 0.12602918 | |
| ], | |
| "bytes_streamed": 1039042768 | |
| }, | |
| { | |
| "layer_idx": 37, | |
| "layer_kind": "moe", | |
| "residual_norm_in": 3284.8843, | |
| "residual_norm_out": 3280.6199, | |
| "expert_indices": [ | |
| 22, | |
| 75, | |
| 156, | |
| 182, | |
| 187, | |
| 189, | |
| 230, | |
| 321 | |
| ], | |
| "routing_weights": [ | |
| 0.12714443, | |
| 0.12565663, | |
| 0.12507114, | |
| 0.12419243, | |
| 0.12688793, | |
| 0.123175636, | |
| 0.12531842, | |
| 0.122553304 | |
| ], | |
| "bytes_streamed": 1014270136 | |
| }, | |
| { | |
| "layer_idx": 38, | |
| "layer_kind": "moe", | |
| "residual_norm_in": 3280.6199, | |
| "residual_norm_out": 3280.2507, | |
| "expert_indices": [ | |
| 116, | |
| 118, | |
| 141, | |
| 142, | |
| 165, | |
| 246, | |
| 296, | |
| 305 | |
| ], | |
| "routing_weights": [ | |
| 0.12589957, | |
| 0.12569398, | |
| 0.12418391, | |
| 0.12590075, | |
| 0.12354491, | |
| 0.12647879, | |
| 0.123581976, | |
| 0.1247161 | |
| ], | |
| "bytes_streamed": 1162905928 | |
| }, | |
| { | |
| "layer_idx": 39, | |
| "layer_kind": "moe", | |
| "residual_norm_in": 3280.2507, | |
| "residual_norm_out": 3279.603, | |
| "expert_indices": [ | |
| 1, | |
| 21, | |
| 53, | |
| 74, | |
| 153, | |
| 232, | |
| 278, | |
| 381 | |
| ], | |
| "routing_weights": [ | |
| 0.12392103, | |
| 0.12356896, | |
| 0.12509237, | |
| 0.12476749, | |
| 0.1261542, | |
| 0.1256581, | |
| 0.12628376, | |
| 0.124554105 | |
| ], | |
| "bytes_streamed": 1162905928 | |
| }, | |
| { | |
| "layer_idx": 40, | |
| "layer_kind": "moe", | |
| "residual_norm_in": 3279.603, | |
| "residual_norm_out": 3275.6978, | |
| "expert_indices": [ | |
| 14, | |
| 54, | |
| 80, | |
| 205, | |
| 243, | |
| 350, | |
| 360, | |
| 378 | |
| ], | |
| "routing_weights": [ | |
| 0.123624906, | |
| 0.12560081, | |
| 0.12601347, | |
| 0.12464201, | |
| 0.124437995, | |
| 0.124908455, | |
| 0.1261274, | |
| 0.124645 | |
| ], | |
| "bytes_streamed": 989497504 | |
| }, | |
| { | |
| "layer_idx": 41, | |
| "layer_kind": "moe", | |
| "residual_norm_in": 3275.6978, | |
| "residual_norm_out": 3273.1753, | |
| "expert_indices": [ | |
| 0, | |
| 12, | |
| 102, | |
| 118, | |
| 143, | |
| 323, | |
| 346, | |
| 361 | |
| ], | |
| "routing_weights": [ | |
| 0.12718984, | |
| 0.12352078, | |
| 0.124371946, | |
| 0.1250337, | |
| 0.12582457, | |
| 0.12378337, | |
| 0.1258188, | |
| 0.12445698 | |
| ], | |
| "bytes_streamed": 1039042768 | |
| }, | |
| { | |
| "layer_idx": 42, | |
| "layer_kind": "moe", | |
| "residual_norm_in": 3273.1753, | |
| "residual_norm_out": 3267.101, | |
| "expert_indices": [ | |
| 11, | |
| 63, | |
| 66, | |
| 121, | |
| 233, | |
| 282, | |
| 286, | |
| 322 | |
| ], | |
| "routing_weights": [ | |
| 0.12231816, | |
| 0.12572178, | |
| 0.12679413, | |
| 0.12391465, | |
| 0.12539703, | |
| 0.122256, | |
| 0.12437119, | |
| 0.12922701 | |
| ], | |
| "bytes_streamed": 1138133296 | |
| }, | |
| { | |
| "layer_idx": 43, | |
| "layer_kind": "moe", | |
| "residual_norm_in": 3267.101, | |
| "residual_norm_out": 3263.3801, | |
| "expert_indices": [ | |
| 48, | |
| 97, | |
| 111, | |
| 145, | |
| 166, | |
| 258, | |
| 275, | |
| 310 | |
| ], | |
| "routing_weights": [ | |
| 0.12841284, | |
| 0.12592939, | |
| 0.1230621, | |
| 0.11106126, | |
| 0.13067316, | |
| 0.12302471, | |
| 0.12871376, | |
| 0.1291227 | |
| ], | |
| "bytes_streamed": 1138133296 | |
| }, | |
| { | |
| "layer_idx": 44, | |
| "layer_kind": "moe", | |
| "residual_norm_in": 3263.3801, | |
| "residual_norm_out": 3254.9749, | |
| "expert_indices": [ | |
| 52, | |
| 80, | |
| 135, | |
| 143, | |
| 145, | |
| 204, | |
| 209, | |
| 313 | |
| ], | |
| "routing_weights": [ | |
| 0.1505961, | |
| 0.15023981, | |
| 0.11775557, | |
| 0.087679155, | |
| 0.12829639, | |
| 0.094963625, | |
| 0.11943085, | |
| 0.15103848 | |
| ], | |
| "bytes_streamed": 1088588032 | |
| }, | |
| { | |
| "layer_idx": 45, | |
| "layer_kind": "moe", | |
| "residual_norm_in": 3254.9749, | |
| "residual_norm_out": 3252.2402, | |
| "expert_indices": [ | |
| 20, | |
| 45, | |
| 57, | |
| 59, | |
| 105, | |
| 176, | |
| 271, | |
| 320 | |
| ], | |
| "routing_weights": [ | |
| 0.12892541, | |
| 0.11697536, | |
| 0.14063151, | |
| 0.12651275, | |
| 0.13517754, | |
| 0.09021728, | |
| 0.13699076, | |
| 0.12456939 | |
| ], | |
| "bytes_streamed": 1039042768 | |
| }, | |
| { | |
| "layer_idx": 46, | |
| "layer_kind": "moe", | |
| "residual_norm_in": 3252.2402, | |
| "residual_norm_out": 3247.5928, | |
| "expert_indices": [ | |
| 7, | |
| 13, | |
| 31, | |
| 58, | |
| 198, | |
| 206, | |
| 222, | |
| 361 | |
| ], | |
| "routing_weights": [ | |
| 0.13090932, | |
| 0.13061666, | |
| 0.119749196, | |
| 0.13284738, | |
| 0.13148266, | |
| 0.12531951, | |
| 0.102036834, | |
| 0.1270384 | |
| ], | |
| "bytes_streamed": 1063815400 | |
| }, | |
| { | |
| "layer_idx": 47, | |
| "layer_kind": "moe", | |
| "residual_norm_in": 3247.5928, | |
| "residual_norm_out": 3246.0535, | |
| "expert_indices": [ | |
| 50, | |
| 91, | |
| 248, | |
| 268, | |
| 274, | |
| 305, | |
| 359, | |
| 377 | |
| ], | |
| "routing_weights": [ | |
| 0.090512484, | |
| 0.12572679, | |
| 0.12560984, | |
| 0.13680649, | |
| 0.13461365, | |
| 0.1383419, | |
| 0.11906946, | |
| 0.12931941 | |
| ], | |
| "bytes_streamed": 1088588032 | |
| }, | |
| { | |
| "layer_idx": 48, | |
| "layer_kind": "moe", | |
| "residual_norm_in": 3246.0535, | |
| "residual_norm_out": 3238.9443, | |
| "expert_indices": [ | |
| 21, | |
| 140, | |
| 180, | |
| 211, | |
| 216, | |
| 265, | |
| 275, | |
| 356 | |
| ], | |
| "routing_weights": [ | |
| 0.089280955, | |
| 0.11947252, | |
| 0.14138731, | |
| 0.13919227, | |
| 0.13082214, | |
| 0.12825514, | |
| 0.12513566, | |
| 0.1264539 | |
| ], | |
| "bytes_streamed": 1113360664 | |
| }, | |
| { | |
| "layer_idx": 49, | |
| "layer_kind": "moe", | |
| "residual_norm_in": 3238.9443, | |
| "residual_norm_out": 3227.1724, | |
| "expert_indices": [ | |
| 65, | |
| 132, | |
| 136, | |
| 141, | |
| 257, | |
| 330, | |
| 331, | |
| 381 | |
| ], | |
| "routing_weights": [ | |
| 0.15157294, | |
| 0.14870882, | |
| 0.12575087, | |
| 0.14817211, | |
| 0.12935574, | |
| 0.12963831, | |
| 0.084066786, | |
| 0.082734436 | |
| ], | |
| "bytes_streamed": 1039042768 | |
| }, | |
| { | |
| "layer_idx": 50, | |
| "layer_kind": "moe", | |
| "residual_norm_in": 3227.1724, | |
| "residual_norm_out": 3215.628, | |
| "expert_indices": [ | |
| 2, | |
| 46, | |
| 77, | |
| 114, | |
| 160, | |
| 162, | |
| 175, | |
| 324 | |
| ], | |
| "routing_weights": [ | |
| 0.07168002, | |
| 0.13420905, | |
| 0.14428033, | |
| 0.13734862, | |
| 0.14445138, | |
| 0.13382252, | |
| 0.11261744, | |
| 0.121590614 | |
| ], | |
| "bytes_streamed": 1063815400 | |
| }, | |
| { | |
| "layer_idx": 51, | |
| "layer_kind": "moe", | |
| "residual_norm_in": 3215.628, | |
| "residual_norm_out": 3206.309, | |
| "expert_indices": [ | |
| 36, | |
| 130, | |
| 189, | |
| 218, | |
| 263, | |
| 271, | |
| 317, | |
| 362 | |
| ], | |
| "routing_weights": [ | |
| 0.07808468, | |
| 0.14101124, | |
| 0.13042147, | |
| 0.11585922, | |
| 0.13718921, | |
| 0.12932168, | |
| 0.12141402, | |
| 0.1466985 | |
| ], | |
| "bytes_streamed": 1088588032 | |
| }, | |
| { | |
| "layer_idx": 52, | |
| "layer_kind": "moe", | |
| "residual_norm_in": 3206.309, | |
| "residual_norm_out": 3188.682, | |
| "expert_indices": [ | |
| 10, | |
| 31, | |
| 40, | |
| 49, | |
| 52, | |
| 96, | |
| 259, | |
| 342 | |
| ], | |
| "routing_weights": [ | |
| 0.13195448, | |
| 0.12658952, | |
| 0.14112595, | |
| 0.11027296, | |
| 0.16430713, | |
| 0.07610071, | |
| 0.12522088, | |
| 0.12442839 | |
| ], | |
| "bytes_streamed": 1014270136 | |
| }, | |
| { | |
| "layer_idx": 53, | |
| "layer_kind": "moe", | |
| "residual_norm_in": 3188.682, | |
| "residual_norm_out": 3153.9102, | |
| "expert_indices": [ | |
| 42, | |
| 101, | |
| 131, | |
| 247, | |
| 272, | |
| 341, | |
| 371, | |
| 383 | |
| ], | |
| "routing_weights": [ | |
| 0.06139395, | |
| 0.17122178, | |
| 0.14294012, | |
| 0.10139634, | |
| 0.056430835, | |
| 0.16060388, | |
| 0.14770874, | |
| 0.15830438 | |
| ], | |
| "bytes_streamed": 1088588032 | |
| }, | |
| { | |
| "layer_idx": 54, | |
| "layer_kind": "moe", | |
| "residual_norm_in": 3153.9102, | |
| "residual_norm_out": 3043.4058, | |
| "expert_indices": [ | |
| 68, | |
| 156, | |
| 169, | |
| 189, | |
| 192, | |
| 218, | |
| 240, | |
| 319 | |
| ], | |
| "routing_weights": [ | |
| 0.053958632, | |
| 0.05449725, | |
| 0.14695823, | |
| 0.14635251, | |
| 0.15215138, | |
| 0.1491312, | |
| 0.14903401, | |
| 0.14791681 | |
| ], | |
| "bytes_streamed": 1014270136 | |
| }, | |
| { | |
| "layer_idx": 55, | |
| "layer_kind": "moe", | |
| "residual_norm_in": 3043.4058, | |
| "residual_norm_out": 2978.261, | |
| "expert_indices": [ | |
| 42, | |
| 87, | |
| 173, | |
| 246, | |
| 308, | |
| 309, | |
| 329, | |
| 348 | |
| ], | |
| "routing_weights": [ | |
| 0.05151474, | |
| 0.1472363, | |
| 0.1606687, | |
| 0.16340461, | |
| 0.15019181, | |
| 0.1526328, | |
| 0.11705792, | |
| 0.05729307 | |
| ], | |
| "bytes_streamed": 1113360664 | |
| }, | |
| { | |
| "layer_idx": 56, | |
| "layer_kind": "moe", | |
| "residual_norm_in": 2978.261, | |
| "residual_norm_out": 2931.9001, | |
| "expert_indices": [ | |
| 26, | |
| 67, | |
| 74, | |
| 84, | |
| 253, | |
| 316, | |
| 319, | |
| 363 | |
| ], | |
| "routing_weights": [ | |
| 0.16366942, | |
| 0.16105407, | |
| 0.16047254, | |
| 0.07660575, | |
| 0.07692882, | |
| 0.16656852, | |
| 0.123046815, | |
| 0.07165412 | |
| ], | |
| "bytes_streamed": 1063815400 | |
| }, | |
| { | |
| "layer_idx": 57, | |
| "layer_kind": "moe", | |
| "residual_norm_in": 2931.9001, | |
| "residual_norm_out": 2830.67, | |
| "expert_indices": [ | |
| 0, | |
| 30, | |
| 80, | |
| 84, | |
| 107, | |
| 222, | |
| 253, | |
| 288 | |
| ], | |
| "routing_weights": [ | |
| 0.18186912, | |
| 0.052621696, | |
| 0.12891348, | |
| 0.05427989, | |
| 0.1740672, | |
| 0.18127596, | |
| 0.17727546, | |
| 0.04969726 | |
| ], | |
| "bytes_streamed": 964724872 | |
| }, | |
| { | |
| "layer_idx": 58, | |
| "layer_kind": "moe", | |
| "residual_norm_in": 2830.67, | |
| "residual_norm_out": 2608.8313, | |
| "expert_indices": [ | |
| 5, | |
| 234, | |
| 259, | |
| 279, | |
| 329, | |
| 344, | |
| 350, | |
| 359 | |
| ], | |
| "routing_weights": [ | |
| 0.12612459, | |
| 0.1269469, | |
| 0.12518732, | |
| 0.12676597, | |
| 0.1180102, | |
| 0.12671871, | |
| 0.12414087, | |
| 0.12610547 | |
| ], | |
| "bytes_streamed": 1063815400 | |
| }, | |
| { | |
| "layer_idx": 59, | |
| "layer_kind": "moe", | |
| "residual_norm_in": 2608.8313, | |
| "residual_norm_out": 3563.1978, | |
| "expert_indices": [ | |
| 47, | |
| 62, | |
| 83, | |
| 156, | |
| 161, | |
| 176, | |
| 372, | |
| 374 | |
| ], | |
| "routing_weights": [ | |
| 0.09576432, | |
| 0.07162925, | |
| 0.061123345, | |
| 0.07196141, | |
| 0.062052693, | |
| 0.48133266, | |
| 0.085414946, | |
| 0.07072137 | |
| ], | |
| "bytes_streamed": 1113360664 | |
| }, | |
| { | |
| "layer_idx": 60, | |
| "layer_kind": "moe", | |
| "residual_norm_in": 3563.1978, | |
| "residual_norm_out": 9140.078, | |
| "expert_indices": [ | |
| 3, | |
| 55, | |
| 68, | |
| 73, | |
| 217, | |
| 234, | |
| 242, | |
| 378 | |
| ], | |
| "routing_weights": [ | |
| 0.03411237, | |
| 0.039301682, | |
| 0.1990229, | |
| 0.18033567, | |
| 0.074357, | |
| 0.2359254, | |
| 0.19573748, | |
| 0.041207485 | |
| ], | |
| "bytes_streamed": 1063815400 | |
| } | |
| ] | |
| } | |
| ], | |
| "total_bytes_streamed": 269574004880, | |
| "signpost": { | |
| "surface": "Output formation ledger", | |
| "provenance": [ | |
| "Measured", | |
| "Derived" | |
| ], | |
| "governance": "Lab", | |
| "admission": "Ratified", | |
| "contracts": [ | |
| "LocalCompute", | |
| "NasBacked" | |
| ], | |
| "scope": "5 tok / 61 layers / no-cache full pass", | |
| "synthetic_audit": "Clean", | |
| "perturbation": "NonPerturbing", | |
| "comparison": "Matched" | |
| } | |
| } |