Add the routed 8k learning-rate sweep: 18 runs with per-expert load

.gitattributes

@@ -283,3 +283,39 @@ lr-transfer-250M/250m-5tpp-bs128-lr2e-9-s1338/diagnostics.riglog filter=lfs diff
 lr-transfer-250M/250m-5tpp-bs128-lr2e-9-s1338/training.riglog filter=lfs diff=lfs merge=lfs -text
 lr-transfer-250M/250m-5tpp-bs128-lr2e-9-s1339/diagnostics.riglog filter=lfs diff=lfs merge=lfs -text
 lr-transfer-250M/250m-5tpp-bs128-lr2e-9-s1339/training.riglog filter=lfs diff=lfs merge=lfs -text
+moe-lr-sweep-8k/125m-5tpp-bs16-lr2e-7-s1337/diagnostics.riglog filter=lfs diff=lfs merge=lfs -text
+moe-lr-sweep-8k/125m-5tpp-bs16-lr2e-7-s1337/training.riglog filter=lfs diff=lfs merge=lfs -text
+moe-lr-sweep-8k/125m-5tpp-bs16-lr2e-8-s1337/diagnostics.riglog filter=lfs diff=lfs merge=lfs -text
+moe-lr-sweep-8k/125m-5tpp-bs16-lr2e-8-s1337/training.riglog filter=lfs diff=lfs merge=lfs -text
+moe-lr-sweep-8k/125m-5tpp-bs16-lr2e-9-s1337/diagnostics.riglog filter=lfs diff=lfs merge=lfs -text
+moe-lr-sweep-8k/125m-5tpp-bs16-lr2e-9-s1337/training.riglog filter=lfs diff=lfs merge=lfs -text
+moe-lr-sweep-8k/60m-5tpp-bs16-lr2e-10-s1337/diagnostics.riglog filter=lfs diff=lfs merge=lfs -text
+moe-lr-sweep-8k/60m-5tpp-bs16-lr2e-10-s1337/training.riglog filter=lfs diff=lfs merge=lfs -text
+moe-lr-sweep-8k/60m-5tpp-bs16-lr2e-10-s1338/diagnostics.riglog filter=lfs diff=lfs merge=lfs -text
+moe-lr-sweep-8k/60m-5tpp-bs16-lr2e-10-s1338/training.riglog filter=lfs diff=lfs merge=lfs -text
+moe-lr-sweep-8k/60m-5tpp-bs16-lr2e-10-s1339/diagnostics.riglog filter=lfs diff=lfs merge=lfs -text
+moe-lr-sweep-8k/60m-5tpp-bs16-lr2e-10-s1339/training.riglog filter=lfs diff=lfs merge=lfs -text
+moe-lr-sweep-8k/60m-5tpp-bs16-lr2e-6-s1337/diagnostics.riglog filter=lfs diff=lfs merge=lfs -text
+moe-lr-sweep-8k/60m-5tpp-bs16-lr2e-6-s1337/training.riglog filter=lfs diff=lfs merge=lfs -text
+moe-lr-sweep-8k/60m-5tpp-bs16-lr2e-6-s1338/diagnostics.riglog filter=lfs diff=lfs merge=lfs -text
+moe-lr-sweep-8k/60m-5tpp-bs16-lr2e-6-s1338/training.riglog filter=lfs diff=lfs merge=lfs -text
+moe-lr-sweep-8k/60m-5tpp-bs16-lr2e-6-s1339/diagnostics.riglog filter=lfs diff=lfs merge=lfs -text
+moe-lr-sweep-8k/60m-5tpp-bs16-lr2e-6-s1339/training.riglog filter=lfs diff=lfs merge=lfs -text
+moe-lr-sweep-8k/60m-5tpp-bs16-lr2e-7-s1337/diagnostics.riglog filter=lfs diff=lfs merge=lfs -text
+moe-lr-sweep-8k/60m-5tpp-bs16-lr2e-7-s1337/training.riglog filter=lfs diff=lfs merge=lfs -text
+moe-lr-sweep-8k/60m-5tpp-bs16-lr2e-7-s1338/diagnostics.riglog filter=lfs diff=lfs merge=lfs -text
+moe-lr-sweep-8k/60m-5tpp-bs16-lr2e-7-s1338/training.riglog filter=lfs diff=lfs merge=lfs -text
+moe-lr-sweep-8k/60m-5tpp-bs16-lr2e-7-s1339/diagnostics.riglog filter=lfs diff=lfs merge=lfs -text
+moe-lr-sweep-8k/60m-5tpp-bs16-lr2e-7-s1339/training.riglog filter=lfs diff=lfs merge=lfs -text
+moe-lr-sweep-8k/60m-5tpp-bs16-lr2e-8-s1337/diagnostics.riglog filter=lfs diff=lfs merge=lfs -text
+moe-lr-sweep-8k/60m-5tpp-bs16-lr2e-8-s1337/training.riglog filter=lfs diff=lfs merge=lfs -text
+moe-lr-sweep-8k/60m-5tpp-bs16-lr2e-8-s1338/diagnostics.riglog filter=lfs diff=lfs merge=lfs -text
+moe-lr-sweep-8k/60m-5tpp-bs16-lr2e-8-s1338/training.riglog filter=lfs diff=lfs merge=lfs -text
+moe-lr-sweep-8k/60m-5tpp-bs16-lr2e-8-s1339/diagnostics.riglog filter=lfs diff=lfs merge=lfs -text
+moe-lr-sweep-8k/60m-5tpp-bs16-lr2e-8-s1339/training.riglog filter=lfs diff=lfs merge=lfs -text
+moe-lr-sweep-8k/60m-5tpp-bs16-lr2e-9-s1337/diagnostics.riglog filter=lfs diff=lfs merge=lfs -text
+moe-lr-sweep-8k/60m-5tpp-bs16-lr2e-9-s1337/training.riglog filter=lfs diff=lfs merge=lfs -text
+moe-lr-sweep-8k/60m-5tpp-bs16-lr2e-9-s1338/diagnostics.riglog filter=lfs diff=lfs merge=lfs -text
+moe-lr-sweep-8k/60m-5tpp-bs16-lr2e-9-s1338/training.riglog filter=lfs diff=lfs merge=lfs -text
+moe-lr-sweep-8k/60m-5tpp-bs16-lr2e-9-s1339/diagnostics.riglog filter=lfs diff=lfs merge=lfs -text
+moe-lr-sweep-8k/60m-5tpp-bs16-lr2e-9-s1339/training.riglog filter=lfs diff=lfs merge=lfs -text

moe-lr-sweep-8k/125m-5tpp-bs16-lr2e-7-s1337/diagnostics.riglog

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:7362e85f565d8180db0234a5d771caea29b11a36c14efa33bd1c8582e58f4585
+size 552576

moe-lr-sweep-8k/125m-5tpp-bs16-lr2e-7-s1337/metrics.json

@@ -0,0 +1,243 @@
+{
+  "artifacts": {
+    "diagnostics": "diagnostics.riglog",
+    "training_curve": "training.riglog",
+    "validation_curve": "validation.csv"
+  },
+  "checkpoint": null,
+  "contract": {
+    "dataset_id": "fineweb-8b-gpt2",
+    "model": {
+      "d_model": 640,
+      "heads": 10,
+      "layers": 12,
+      "mlp_activation": "gelu",
+      "mlp_mult": 4,
+      "normalization": "rms_norm",
+      "parameterization": "complete_d_p",
+      "position_encoding": "rope_base_10000",
+      "semantic_vocab_size": 50304,
+      "tied_embeddings": false,
+      "tier": "125m",
+      "vocab_size": 50304
+    },
+    "model_id": "reference-gpt-v3-family",
+    "sequence_length": 8192,
+    "tokenizer_id": "gpt2"
+  },
+  "evaluations": {
+    "fineweb": {
+      "canonical": true,
+      "loss": 3.5722893476486206,
+      "perplexity": 35.59799612494644,
+      "scored_tokens": 1048576,
+      "seconds": 0.8216901089763269
+    }
+  },
+  "implementation": {
+    "attention_backend": "tpu_flash",
+    "attention_tuning": {
+      "key_digest": "bf93286e9d38983bc78d617a9a666f8b00490f5bfff49b38c7fff97fd570e3ea",
+      "resolution_source": "heuristic",
+      "tiles": {
+        "block_kv": 512,
+        "block_kv_compute": 256,
+        "block_kv_dkv": 512,
+        "block_kv_dkv_compute": 256,
+        "block_kv_dq": 512,
+        "block_kv_dq_compute": 256,
+        "block_q": 512,
+        "block_q_dkv": 512,
+        "block_q_dkv_compute": 256,
+        "block_q_dq": 256
+      },
+      "tune_seconds": 0.0
+    },
+    "configuration": {
+      "overrides": {
+        "tokens_per_parameter_micros": 5000000
+      },
+      "path": "config.yaml",
+      "profile": "dev",
+      "resolved": {
+        "evaluation": {
+          "eval_batches": 8,
+          "val_every": 0,
+          "val_probe_batches": 8
+        },
+        "kernels": {
+          "attention_backend": "tpu_flash",
+          "loss_backend": "tiled",
+          "vocab_tile_size": 2048
+        },
+        "logging": {
+          "diagnostics_every": 10,
+          "log_every": 10
+        },
+        "model": {
+          "d_model": 640,
+          "heads": 10,
+          "layers": 12,
+          "mlp_activation": "gelu",
+          "mlp_mult": 4,
+          "normalization": "rms_norm",
+          "parameterization": "complete_d_p",
+          "position_encoding": "rope_base_10000",
+          "semantic_vocab_size": 50304,
+          "tied_embeddings": false,
+          "tier": "125m",
+          "vocab_size": 50304
+        },
+        "optimizer": {
+          "adam_epsilon": 1e-08,
+          "beta1": 0.9,
+          "beta2": 0.95,
+          "effective": {
+            "adam_epsilon_horizon_multiplier": 1.4354161175193187,
+            "beta1": 0.9514661925028901,
+            "beta2": 0.975733096251445,
+            "global_peak_learning_rate": 0.0054426726192134,
+            "weight_decay_horizon_multiplier": 0.6966620952593152
+          },
+          "grad_clip": 0.0,
+          "learning_rate": 0.0078125,
+          "min_lr_ratio": 0.1,
+          "warmup_steps": 471,
+          "weight_decay": 0.1
+        },
+        "parameterization": {
+          "attention_scale": "inverse_head_dim",
+          "base_depth": 12,
+          "base_width": 384,
+          "batch_multiplier": 1.0,
+          "data_multiplier": 2.0604194304342345,
+          "depth_alpha": 1.0,
+          "depth_multiplier": 1.0,
+          "embeddings": "untied",
+          "init_std": 0.02,
+          "name": "complete_d_p",
+          "width_multiplier": 1.6666666666666667
+        },
+        "training": {
+          "batch_size": 16,
+          "dtype": "bfloat16",
+          "sampling": "shuffled_epochs",
+          "seq_len": 8192,
+          "steps": 4709,
+          "tokens_per_parameter": 4.999472268158278,
+          "train_tokens": 617218048
+        }
+      },
+      "schema_version": 2,
+      "sha256": "a4bad1be29b57d20ec7ae7b0d539f2e777d11f5539607655550b7afca185affe"
+    },
+    "loss_backend": "tiled",
+    "vocab_tile_size": 2048
+  },
+  "metrics": {
+    "achieved_tflops": 543.8900671494216,
+    "attention_tune_seconds": 0.0,
+    "base_learning_rate": 0.0078125,
+    "diagnostic_compile_seconds": 97.0277210950153,
+    "diagnostic_point_count": 472,
+    "diagnostics_every": 10,
+    "early_stopping_step": null,
+    "estimated_total_flops": 846442363009105920,
+    "eval_compile_seconds": 18.104762503993697,
+    "expert_top_k": 2,
+    "experts": 8,
+    "final_validation_seconds": 0.8216901089763269,
+    "flop_accounting": {
+      "by_site": {
+        "dot_general": 3116881674240,
+        "pallas_call": 1932735283200,
+        "tpu_flash_causal_attention_bwd_dkv": 2061584302080,
+        "tpu_flash_causal_attention_bwd_dq": 2061584302080,
+        "tpu_flash_causal_attention_fwd": 2061584302080
+      },
+      "elementwise_per_sequence": 13316361400,
+      "matmul_per_sequence": 11234369863680,
+      "method": "traced-jaxpr",
+      "warnings": []
+    },
+    "flops_per_token": 1371383040,
+    "mfu_estimate": 0.12361137889759582,
+    "model_tier": "125m",
+    "parameter_count": 123525760,
+    "parameters": 241628800,
+    "schedule_steps": 4709,
+    "tokens_per_parameter": 4.999472268158278,
+    "tokens_per_second": 396599.6744056435,
+    "tokens_processed": 617218048,
+    "total_compile_seconds": 181.27476295700762,
+    "total_parameter_count": 241628800,
+    "train_compile_seconds": 66.14227935799863,
+    "train_loss": 3.494847536087036,
+    "train_seconds": 1556.2747219220037,
+    "training_data_epochs": 0.07812906684585416,
+    "training_data_sharding": "rank_disjoint_shuffled_windows",
+    "training_sampling": "shuffled_epochs",
+    "training_steps": 4709,
+    "training_token_budget": 617218048,
+    "training_usable_tokens_per_epoch": 7899979776,
+    "validation_loss": 3.5722893476486206,
+    "validation_probe_count": 0,
+    "validation_probe_seconds": 0.0,
+    "validation_tokens": 1048576
+  },
+  "profile": "dev",
+  "schema_version": 1,
+  "seed": 1337,
+  "status": "ok",
+  "system": {
+    "controller_process_index": 1,
+    "device_count": 16,
+    "device_ids": [
+      0,
+      1,
+      2,
+      3,
+      4,
+      5,
+      6,
+      7,
+      8,
+      9,
+      10,
+      11,
+      12,
+      13,
+      14,
+      15
+    ],
+    "device_kinds": [
+      "TPU v4"
+    ],
+    "jax_version": "0.11.0",
+    "jaxlib_version": "0.11.0",
+    "libtpu_version": "0.0.44.1",
+    "local_device_count": 4,
+    "platform": "tpu",
+    "process_count": 4,
+    "process_indices": [
+      0,
+      0,
+      0,
+      0,
+      1,
+      1,
+      1,
+      1,
+      2,
+      2,
+      2,
+      2,
+      3,
+      3,
+      3,
+      3
+    ],
+    "python_version": "3.12.13"
+  },
+  "track": "open"
+}

moe-lr-sweep-8k/125m-5tpp-bs16-lr2e-7-s1337/result.json

@@ -0,0 +1,243 @@
+{
+  "artifacts": {
+    "diagnostics": "diagnostics.riglog",
+    "training_curve": "training.riglog",
+    "validation_curve": "validation.csv"
+  },
+  "checkpoint": null,
+  "contract": {
+    "dataset_id": "fineweb-8b-gpt2",
+    "model": {
+      "d_model": 640,
+      "heads": 10,
+      "layers": 12,
+      "mlp_activation": "gelu",
+      "mlp_mult": 4,
+      "normalization": "rms_norm",
+      "parameterization": "complete_d_p",
+      "position_encoding": "rope_base_10000",
+      "semantic_vocab_size": 50304,
+      "tied_embeddings": false,
+      "tier": "125m",
+      "vocab_size": 50304
+    },
+    "model_id": "reference-gpt-v3-family",
+    "sequence_length": 8192,
+    "tokenizer_id": "gpt2"
+  },
+  "evaluations": {
+    "fineweb": {
+      "canonical": true,
+      "loss": 3.5722893476486206,
+      "perplexity": 35.59799612494644,
+      "scored_tokens": 1048576,
+      "seconds": 0.8216901089763269
+    }
+  },
+  "implementation": {
+    "attention_backend": "tpu_flash",
+    "attention_tuning": {
+      "key_digest": "bf93286e9d38983bc78d617a9a666f8b00490f5bfff49b38c7fff97fd570e3ea",
+      "resolution_source": "heuristic",
+      "tiles": {
+        "block_kv": 512,
+        "block_kv_compute": 256,
+        "block_kv_dkv": 512,
+        "block_kv_dkv_compute": 256,
+        "block_kv_dq": 512,
+        "block_kv_dq_compute": 256,
+        "block_q": 512,
+        "block_q_dkv": 512,
+        "block_q_dkv_compute": 256,
+        "block_q_dq": 256
+      },
+      "tune_seconds": 0.0
+    },
+    "configuration": {
+      "overrides": {
+        "tokens_per_parameter_micros": 5000000
+      },
+      "path": "config.yaml",
+      "profile": "dev",
+      "resolved": {
+        "evaluation": {
+          "eval_batches": 8,
+          "val_every": 0,
+          "val_probe_batches": 8
+        },
+        "kernels": {
+          "attention_backend": "tpu_flash",
+          "loss_backend": "tiled",
+          "vocab_tile_size": 2048
+        },
+        "logging": {
+          "diagnostics_every": 10,
+          "log_every": 10
+        },
+        "model": {
+          "d_model": 640,
+          "heads": 10,
+          "layers": 12,
+          "mlp_activation": "gelu",
+          "mlp_mult": 4,
+          "normalization": "rms_norm",
+          "parameterization": "complete_d_p",
+          "position_encoding": "rope_base_10000",
+          "semantic_vocab_size": 50304,
+          "tied_embeddings": false,
+          "tier": "125m",
+          "vocab_size": 50304
+        },
+        "optimizer": {
+          "adam_epsilon": 1e-08,
+          "beta1": 0.9,
+          "beta2": 0.95,
+          "effective": {
+            "adam_epsilon_horizon_multiplier": 1.4354161175193187,
+            "beta1": 0.9514661925028901,
+            "beta2": 0.975733096251445,
+            "global_peak_learning_rate": 0.0054426726192134,
+            "weight_decay_horizon_multiplier": 0.6966620952593152
+          },
+          "grad_clip": 0.0,
+          "learning_rate": 0.0078125,
+          "min_lr_ratio": 0.1,
+          "warmup_steps": 471,
+          "weight_decay": 0.1
+        },
+        "parameterization": {
+          "attention_scale": "inverse_head_dim",
+          "base_depth": 12,
+          "base_width": 384,
+          "batch_multiplier": 1.0,
+          "data_multiplier": 2.0604194304342345,
+          "depth_alpha": 1.0,
+          "depth_multiplier": 1.0,
+          "embeddings": "untied",
+          "init_std": 0.02,
+          "name": "complete_d_p",
+          "width_multiplier": 1.6666666666666667
+        },
+        "training": {
+          "batch_size": 16,
+          "dtype": "bfloat16",
+          "sampling": "shuffled_epochs",
+          "seq_len": 8192,
+          "steps": 4709,
+          "tokens_per_parameter": 4.999472268158278,
+          "train_tokens": 617218048
+        }
+      },
+      "schema_version": 2,
+      "sha256": "a4bad1be29b57d20ec7ae7b0d539f2e777d11f5539607655550b7afca185affe"
+    },
+    "loss_backend": "tiled",
+    "vocab_tile_size": 2048
+  },
+  "metrics": {
+    "achieved_tflops": 543.8900671494216,
+    "attention_tune_seconds": 0.0,
+    "base_learning_rate": 0.0078125,
+    "diagnostic_compile_seconds": 97.0277210950153,
+    "diagnostic_point_count": 472,
+    "diagnostics_every": 10,
+    "early_stopping_step": null,
+    "estimated_total_flops": 846442363009105920,
+    "eval_compile_seconds": 18.104762503993697,
+    "expert_top_k": 2,
+    "experts": 8,
+    "final_validation_seconds": 0.8216901089763269,
+    "flop_accounting": {
+      "by_site": {
+        "dot_general": 3116881674240,
+        "pallas_call": 1932735283200,
+        "tpu_flash_causal_attention_bwd_dkv": 2061584302080,
+        "tpu_flash_causal_attention_bwd_dq": 2061584302080,
+        "tpu_flash_causal_attention_fwd": 2061584302080
+      },
+      "elementwise_per_sequence": 13316361400,
+      "matmul_per_sequence": 11234369863680,
+      "method": "traced-jaxpr",
+      "warnings": []
+    },
+    "flops_per_token": 1371383040,
+    "mfu_estimate": 0.12361137889759582,
+    "model_tier": "125m",
+    "parameter_count": 123525760,
+    "parameters": 241628800,
+    "schedule_steps": 4709,
+    "tokens_per_parameter": 4.999472268158278,
+    "tokens_per_second": 396599.6744056435,
+    "tokens_processed": 617218048,
+    "total_compile_seconds": 181.27476295700762,
+    "total_parameter_count": 241628800,
+    "train_compile_seconds": 66.14227935799863,
+    "train_loss": 3.494847536087036,
+    "train_seconds": 1556.2747219220037,
+    "training_data_epochs": 0.07812906684585416,
+    "training_data_sharding": "rank_disjoint_shuffled_windows",
+    "training_sampling": "shuffled_epochs",
+    "training_steps": 4709,
+    "training_token_budget": 617218048,
+    "training_usable_tokens_per_epoch": 7899979776,
+    "validation_loss": 3.5722893476486206,
+    "validation_probe_count": 0,
+    "validation_probe_seconds": 0.0,
+    "validation_tokens": 1048576
+  },
+  "profile": "dev",
+  "schema_version": 1,
+  "seed": 1337,
+  "status": "ok",
+  "system": {
+    "controller_process_index": 1,
+    "device_count": 16,
+    "device_ids": [
+      0,
+      1,
+      2,
+      3,
+      4,
+      5,
+      6,
+      7,
+      8,
+      9,
+      10,
+      11,
+      12,
+      13,
+      14,
+      15
+    ],
+    "device_kinds": [
+      "TPU v4"
+    ],
+    "jax_version": "0.11.0",
+    "jaxlib_version": "0.11.0",
+    "libtpu_version": "0.0.44.1",
+    "local_device_count": 4,
+    "platform": "tpu",
+    "process_count": 4,
+    "process_indices": [
+      0,
+      0,
+      0,
+      0,
+      1,
+      1,
+      1,
+      1,
+      2,
+      2,
+      2,
+      2,
+      3,
+      3,
+      3,
+      3
+    ],
+    "python_version": "3.12.13"
+  },
+  "track": "open"
+}

moe-lr-sweep-8k/125m-5tpp-bs16-lr2e-7-s1337/training.riglog

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:6bbfe067b48c85047631bc161a47e2b18c3c85a9c988ade69d27cd30e3a221bc
+size 2678128

moe-lr-sweep-8k/125m-5tpp-bs16-lr2e-7-s1337/validation.csv

@@ -0,0 +1,2 @@
+step,tokens_processed,kind,domain,validation_tokens,validation_loss,perplexity,validation_seconds,canonical
+4709,617218048,fineweb,fineweb,1048576,3.5722893476486206,35.59799612494644,0.8216901089763269,true

moe-lr-sweep-8k/125m-5tpp-bs16-lr2e-8-s1337/diagnostics.riglog

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:0ad5eb251c51535db6ef54d6c15cef717c1a5f9a09fb5622b310b7876c1325b0
+size 552576

moe-lr-sweep-8k/125m-5tpp-bs16-lr2e-8-s1337/metrics.json

@@ -0,0 +1,243 @@
+{
+  "artifacts": {
+    "diagnostics": "diagnostics.riglog",
+    "training_curve": "training.riglog",
+    "validation_curve": "validation.csv"
+  },
+  "checkpoint": null,
+  "contract": {
+    "dataset_id": "fineweb-8b-gpt2",
+    "model": {
+      "d_model": 640,
+      "heads": 10,
+      "layers": 12,
+      "mlp_activation": "gelu",
+      "mlp_mult": 4,
+      "normalization": "rms_norm",
+      "parameterization": "complete_d_p",
+      "position_encoding": "rope_base_10000",
+      "semantic_vocab_size": 50304,
+      "tied_embeddings": false,
+      "tier": "125m",
+      "vocab_size": 50304
+    },
+    "model_id": "reference-gpt-v3-family",
+    "sequence_length": 8192,
+    "tokenizer_id": "gpt2"
+  },
+  "evaluations": {
+    "fineweb": {
+      "canonical": true,
+      "loss": 3.5765029788017273,
+      "perplexity": 35.74830941038075,
+      "scored_tokens": 1048576,
+      "seconds": 0.8198627600213513
+    }
+  },
+  "implementation": {
+    "attention_backend": "tpu_flash",
+    "attention_tuning": {
+      "key_digest": "bf93286e9d38983bc78d617a9a666f8b00490f5bfff49b38c7fff97fd570e3ea",
+      "resolution_source": "heuristic",
+      "tiles": {
+        "block_kv": 512,
+        "block_kv_compute": 256,
+        "block_kv_dkv": 512,
+        "block_kv_dkv_compute": 256,
+        "block_kv_dq": 512,
+        "block_kv_dq_compute": 256,
+        "block_q": 512,
+        "block_q_dkv": 512,
+        "block_q_dkv_compute": 256,
+        "block_q_dq": 256
+      },
+      "tune_seconds": 0.0
+    },
+    "configuration": {
+      "overrides": {
+        "tokens_per_parameter_micros": 5000000
+      },
+      "path": "config.yaml",
+      "profile": "dev",
+      "resolved": {
+        "evaluation": {
+          "eval_batches": 8,
+          "val_every": 0,
+          "val_probe_batches": 8
+        },
+        "kernels": {
+          "attention_backend": "tpu_flash",
+          "loss_backend": "tiled",
+          "vocab_tile_size": 2048
+        },
+        "logging": {
+          "diagnostics_every": 10,
+          "log_every": 10
+        },
+        "model": {
+          "d_model": 640,
+          "heads": 10,
+          "layers": 12,
+          "mlp_activation": "gelu",
+          "mlp_mult": 4,
+          "normalization": "rms_norm",
+          "parameterization": "complete_d_p",
+          "position_encoding": "rope_base_10000",
+          "semantic_vocab_size": 50304,
+          "tied_embeddings": false,
+          "tier": "125m",
+          "vocab_size": 50304
+        },
+        "optimizer": {
+          "adam_epsilon": 1e-08,
+          "beta1": 0.9,
+          "beta2": 0.95,
+          "effective": {
+            "adam_epsilon_horizon_multiplier": 1.4354161175193187,
+            "beta1": 0.9514661925028901,
+            "beta2": 0.975733096251445,
+            "global_peak_learning_rate": 0.0027213363096067,
+            "weight_decay_horizon_multiplier": 0.6966620952593152
+          },
+          "grad_clip": 0.0,
+          "learning_rate": 0.00390625,
+          "min_lr_ratio": 0.1,
+          "warmup_steps": 471,
+          "weight_decay": 0.1
+        },
+        "parameterization": {
+          "attention_scale": "inverse_head_dim",
+          "base_depth": 12,
+          "base_width": 384,
+          "batch_multiplier": 1.0,
+          "data_multiplier": 2.0604194304342345,
+          "depth_alpha": 1.0,
+          "depth_multiplier": 1.0,
+          "embeddings": "untied",
+          "init_std": 0.02,
+          "name": "complete_d_p",
+          "width_multiplier": 1.6666666666666667
+        },
+        "training": {
+          "batch_size": 16,
+          "dtype": "bfloat16",
+          "sampling": "shuffled_epochs",
+          "seq_len": 8192,
+          "steps": 4709,
+          "tokens_per_parameter": 4.999472268158278,
+          "train_tokens": 617218048
+        }
+      },
+      "schema_version": 2,
+      "sha256": "a4bad1be29b57d20ec7ae7b0d539f2e777d11f5539607655550b7afca185affe"
+    },
+    "loss_backend": "tiled",
+    "vocab_tile_size": 2048
+  },
+  "metrics": {
+    "achieved_tflops": 543.7682606917205,
+    "attention_tune_seconds": 0.0,
+    "base_learning_rate": 0.00390625,
+    "diagnostic_compile_seconds": 99.17232654301915,
+    "diagnostic_point_count": 472,
+    "diagnostics_every": 10,
+    "early_stopping_step": null,
+    "estimated_total_flops": 846442363009105920,
+    "eval_compile_seconds": 18.211572511005215,
+    "expert_top_k": 2,
+    "experts": 8,
+    "final_validation_seconds": 0.8198627600213513,
+    "flop_accounting": {
+      "by_site": {
+        "dot_general": 3116881674240,
+        "pallas_call": 1932735283200,
+        "tpu_flash_causal_attention_bwd_dkv": 2061584302080,
+        "tpu_flash_causal_attention_bwd_dq": 2061584302080,
+        "tpu_flash_causal_attention_fwd": 2061584302080
+      },
+      "elementwise_per_sequence": 13316361400,
+      "matmul_per_sequence": 11234369863680,
+      "method": "traced-jaxpr",
+      "warnings": []
+    },
+    "flops_per_token": 1371383040,
+    "mfu_estimate": 0.12358369561175465,
+    "model_tier": "125m",
+    "parameter_count": 123525760,
+    "parameters": 241628800,
+    "schedule_steps": 4709,
+    "tokens_per_parameter": 4.999472268158278,
+    "tokens_per_second": 396510.8542480738,
+    "tokens_processed": 617218048,
+    "total_compile_seconds": 177.4247099260101,
+    "total_parameter_count": 241628800,
+    "train_compile_seconds": 60.04081087198574,
+    "train_loss": 3.5012834072113037,
+    "train_seconds": 1556.6233342349879,
+    "training_data_epochs": 0.07812906684585416,
+    "training_data_sharding": "rank_disjoint_shuffled_windows",
+    "training_sampling": "shuffled_epochs",
+    "training_steps": 4709,
+    "training_token_budget": 617218048,
+    "training_usable_tokens_per_epoch": 7899979776,
+    "validation_loss": 3.5765029788017273,
+    "validation_probe_count": 0,
+    "validation_probe_seconds": 0.0,
+    "validation_tokens": 1048576
+  },
+  "profile": "dev",
+  "schema_version": 1,
+  "seed": 1337,
+  "status": "ok",
+  "system": {
+    "controller_process_index": 1,
+    "device_count": 16,
+    "device_ids": [
+      0,
+      1,
+      2,
+      3,
+      4,
+      5,
+      6,
+      7,
+      8,
+      9,
+      10,
+      11,
+      12,
+      13,
+      14,
+      15
+    ],
+    "device_kinds": [
+      "TPU v4"
+    ],
+    "jax_version": "0.11.0",
+    "jaxlib_version": "0.11.0",
+    "libtpu_version": "0.0.44.1",
+    "local_device_count": 4,
+    "platform": "tpu",
+    "process_count": 4,
+    "process_indices": [
+      0,
+      0,
+      0,
+      0,
+      1,
+      1,
+      1,
+      1,
+      2,
+      2,
+      2,
+      2,
+      3,
+      3,
+      3,
+      3
+    ],
+    "python_version": "3.12.13"
+  },
+  "track": "open"
+}

moe-lr-sweep-8k/125m-5tpp-bs16-lr2e-8-s1337/result.json

@@ -0,0 +1,243 @@
+{
+  "artifacts": {
+    "diagnostics": "diagnostics.riglog",
+    "training_curve": "training.riglog",
+    "validation_curve": "validation.csv"
+  },
+  "checkpoint": null,
+  "contract": {
+    "dataset_id": "fineweb-8b-gpt2",
+    "model": {
+      "d_model": 640,
+      "heads": 10,
+      "layers": 12,
+      "mlp_activation": "gelu",
+      "mlp_mult": 4,
+      "normalization": "rms_norm",
+      "parameterization": "complete_d_p",
+      "position_encoding": "rope_base_10000",
+      "semantic_vocab_size": 50304,
+      "tied_embeddings": false,
+      "tier": "125m",
+      "vocab_size": 50304
+    },
+    "model_id": "reference-gpt-v3-family",
+    "sequence_length": 8192,
+    "tokenizer_id": "gpt2"
+  },
+  "evaluations": {
+    "fineweb": {
+      "canonical": true,
+      "loss": 3.5765029788017273,
+      "perplexity": 35.74830941038075,
+      "scored_tokens": 1048576,
+      "seconds": 0.8198627600213513
+    }
+  },
+  "implementation": {
+    "attention_backend": "tpu_flash",
+    "attention_tuning": {
+      "key_digest": "bf93286e9d38983bc78d617a9a666f8b00490f5bfff49b38c7fff97fd570e3ea",
+      "resolution_source": "heuristic",
+      "tiles": {
+        "block_kv": 512,
+        "block_kv_compute": 256,
+        "block_kv_dkv": 512,
+        "block_kv_dkv_compute": 256,
+        "block_kv_dq": 512,
+        "block_kv_dq_compute": 256,
+        "block_q": 512,
+        "block_q_dkv": 512,
+        "block_q_dkv_compute": 256,
+        "block_q_dq": 256
+      },
+      "tune_seconds": 0.0
+    },
+    "configuration": {
+      "overrides": {
+        "tokens_per_parameter_micros": 5000000
+      },
+      "path": "config.yaml",
+      "profile": "dev",
+      "resolved": {
+        "evaluation": {
+          "eval_batches": 8,
+          "val_every": 0,
+          "val_probe_batches": 8
+        },
+        "kernels": {
+          "attention_backend": "tpu_flash",
+          "loss_backend": "tiled",
+          "vocab_tile_size": 2048
+        },
+        "logging": {
+          "diagnostics_every": 10,
+          "log_every": 10
+        },
+        "model": {
+          "d_model": 640,
+          "heads": 10,
+          "layers": 12,
+          "mlp_activation": "gelu",
+          "mlp_mult": 4,
+          "normalization": "rms_norm",
+          "parameterization": "complete_d_p",
+          "position_encoding": "rope_base_10000",
+          "semantic_vocab_size": 50304,
+          "tied_embeddings": false,
+          "tier": "125m",
+          "vocab_size": 50304
+        },
+        "optimizer": {
+          "adam_epsilon": 1e-08,
+          "beta1": 0.9,
+          "beta2": 0.95,
+          "effective": {
+            "adam_epsilon_horizon_multiplier": 1.4354161175193187,
+            "beta1": 0.9514661925028901,
+            "beta2": 0.975733096251445,
+            "global_peak_learning_rate": 0.0027213363096067,
+            "weight_decay_horizon_multiplier": 0.6966620952593152
+          },
+          "grad_clip": 0.0,
+          "learning_rate": 0.00390625,
+          "min_lr_ratio": 0.1,
+          "warmup_steps": 471,
+          "weight_decay": 0.1
+        },
+        "parameterization": {
+          "attention_scale": "inverse_head_dim",
+          "base_depth": 12,
+          "base_width": 384,
+          "batch_multiplier": 1.0,
+          "data_multiplier": 2.0604194304342345,
+          "depth_alpha": 1.0,
+          "depth_multiplier": 1.0,
+          "embeddings": "untied",
+          "init_std": 0.02,
+          "name": "complete_d_p",
+          "width_multiplier": 1.6666666666666667
+        },
+        "training": {
+          "batch_size": 16,
+          "dtype": "bfloat16",
+          "sampling": "shuffled_epochs",
+          "seq_len": 8192,
+          "steps": 4709,
+          "tokens_per_parameter": 4.999472268158278,
+          "train_tokens": 617218048
+        }
+      },
+      "schema_version": 2,
+      "sha256": "a4bad1be29b57d20ec7ae7b0d539f2e777d11f5539607655550b7afca185affe"
+    },
+    "loss_backend": "tiled",
+    "vocab_tile_size": 2048
+  },
+  "metrics": {
+    "achieved_tflops": 543.7682606917205,
+    "attention_tune_seconds": 0.0,
+    "base_learning_rate": 0.00390625,
+    "diagnostic_compile_seconds": 99.17232654301915,
+    "diagnostic_point_count": 472,
+    "diagnostics_every": 10,
+    "early_stopping_step": null,
+    "estimated_total_flops": 846442363009105920,
+    "eval_compile_seconds": 18.211572511005215,
+    "expert_top_k": 2,
+    "experts": 8,
+    "final_validation_seconds": 0.8198627600213513,
+    "flop_accounting": {
+      "by_site": {
+        "dot_general": 3116881674240,
+        "pallas_call": 1932735283200,
+        "tpu_flash_causal_attention_bwd_dkv": 2061584302080,
+        "tpu_flash_causal_attention_bwd_dq": 2061584302080,
+        "tpu_flash_causal_attention_fwd": 2061584302080
+      },
+      "elementwise_per_sequence": 13316361400,
+      "matmul_per_sequence": 11234369863680,
+      "method": "traced-jaxpr",
+      "warnings": []
+    },
+    "flops_per_token": 1371383040,
+    "mfu_estimate": 0.12358369561175465,
+    "model_tier": "125m",
+    "parameter_count": 123525760,
+    "parameters": 241628800,
+    "schedule_steps": 4709,
+    "tokens_per_parameter": 4.999472268158278,
+    "tokens_per_second": 396510.8542480738,
+    "tokens_processed": 617218048,
+    "total_compile_seconds": 177.4247099260101,
+    "total_parameter_count": 241628800,
+    "train_compile_seconds": 60.04081087198574,
+    "train_loss": 3.5012834072113037,
+    "train_seconds": 1556.6233342349879,
+    "training_data_epochs": 0.07812906684585416,
+    "training_data_sharding": "rank_disjoint_shuffled_windows",
+    "training_sampling": "shuffled_epochs",
+    "training_steps": 4709,
+    "training_token_budget": 617218048,
+    "training_usable_tokens_per_epoch": 7899979776,
+    "validation_loss": 3.5765029788017273,
+    "validation_probe_count": 0,
+    "validation_probe_seconds": 0.0,
+    "validation_tokens": 1048576
+  },
+  "profile": "dev",
+  "schema_version": 1,
+  "seed": 1337,
+  "status": "ok",
+  "system": {
+    "controller_process_index": 1,
+    "device_count": 16,
+    "device_ids": [
+      0,
+      1,
+      2,
+      3,
+      4,
+      5,
+      6,
+      7,
+      8,
+      9,
+      10,
+      11,
+      12,
+      13,
+      14,
+      15
+    ],
+    "device_kinds": [
+      "TPU v4"
+    ],
+    "jax_version": "0.11.0",
+    "jaxlib_version": "0.11.0",
+    "libtpu_version": "0.0.44.1",
+    "local_device_count": 4,
+    "platform": "tpu",
+    "process_count": 4,
+    "process_indices": [
+      0,
+      0,
+      0,
+      0,
+      1,
+      1,
+      1,
+      1,
+      2,
+      2,
+      2,
+      2,
+      3,
+      3,
+      3,
+      3
+    ],
+    "python_version": "3.12.13"
+  },
+  "track": "open"
+}

moe-lr-sweep-8k/125m-5tpp-bs16-lr2e-8-s1337/training.riglog

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:07f92e2471ddb1479cfbdd43d256e1f3e1566d8efc1ad397bfc539cf0293651c
+size 2678128

moe-lr-sweep-8k/125m-5tpp-bs16-lr2e-8-s1337/validation.csv

@@ -0,0 +1,2 @@
+step,tokens_processed,kind,domain,validation_tokens,validation_loss,perplexity,validation_seconds,canonical
+4709,617218048,fineweb,fineweb,1048576,3.5765029788017273,35.74830941038075,0.8198627600213513,true

moe-lr-sweep-8k/125m-5tpp-bs16-lr2e-9-s1337/diagnostics.riglog

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:c08e4f290cd0276ae108de5818b301d9996f66bfab0008d3400d895fb0b25b30
+size 552576

moe-lr-sweep-8k/125m-5tpp-bs16-lr2e-9-s1337/metrics.json

@@ -0,0 +1,243 @@
+{
+  "artifacts": {
+    "diagnostics": "diagnostics.riglog",
+    "training_curve": "training.riglog",
+    "validation_curve": "validation.csv"
+  },
+  "checkpoint": null,
+  "contract": {
+    "dataset_id": "fineweb-8b-gpt2",
+    "model": {
+      "d_model": 640,
+      "heads": 10,
+      "layers": 12,
+      "mlp_activation": "gelu",
+      "mlp_mult": 4,
+      "normalization": "rms_norm",
+      "parameterization": "complete_d_p",
+      "position_encoding": "rope_base_10000",
+      "semantic_vocab_size": 50304,
+      "tied_embeddings": false,
+      "tier": "125m",
+      "vocab_size": 50304
+    },
+    "model_id": "reference-gpt-v3-family",
+    "sequence_length": 8192,
+    "tokenizer_id": "gpt2"
+  },
+  "evaluations": {
+    "fineweb": {
+      "canonical": true,
+      "loss": 3.636278748512268,
+      "perplexity": 37.95035083687206,
+      "scored_tokens": 1048576,
+      "seconds": 0.820926460030023
+    }
+  },
+  "implementation": {
+    "attention_backend": "tpu_flash",
+    "attention_tuning": {
+      "key_digest": "bf93286e9d38983bc78d617a9a666f8b00490f5bfff49b38c7fff97fd570e3ea",
+      "resolution_source": "heuristic",
+      "tiles": {
+        "block_kv": 512,
+        "block_kv_compute": 256,
+        "block_kv_dkv": 512,
+        "block_kv_dkv_compute": 256,
+        "block_kv_dq": 512,
+        "block_kv_dq_compute": 256,
+        "block_q": 512,
+        "block_q_dkv": 512,
+        "block_q_dkv_compute": 256,
+        "block_q_dq": 256
+      },
+      "tune_seconds": 0.0
+    },
+    "configuration": {
+      "overrides": {
+        "tokens_per_parameter_micros": 5000000
+      },
+      "path": "config.yaml",
+      "profile": "dev",
+      "resolved": {
+        "evaluation": {
+          "eval_batches": 8,
+          "val_every": 0,
+          "val_probe_batches": 8
+        },
+        "kernels": {
+          "attention_backend": "tpu_flash",
+          "loss_backend": "tiled",
+          "vocab_tile_size": 2048
+        },
+        "logging": {
+          "diagnostics_every": 10,
+          "log_every": 10
+        },
+        "model": {
+          "d_model": 640,
+          "heads": 10,
+          "layers": 12,
+          "mlp_activation": "gelu",
+          "mlp_mult": 4,
+          "normalization": "rms_norm",
+          "parameterization": "complete_d_p",
+          "position_encoding": "rope_base_10000",
+          "semantic_vocab_size": 50304,
+          "tied_embeddings": false,
+          "tier": "125m",
+          "vocab_size": 50304
+        },
+        "optimizer": {
+          "adam_epsilon": 1e-08,
+          "beta1": 0.9,
+          "beta2": 0.95,
+          "effective": {
+            "adam_epsilon_horizon_multiplier": 1.4354161175193187,
+            "beta1": 0.9514661925028901,
+            "beta2": 0.975733096251445,
+            "global_peak_learning_rate": 0.00136066815480335,
+            "weight_decay_horizon_multiplier": 0.6966620952593152
+          },
+          "grad_clip": 0.0,
+          "learning_rate": 0.001953125,
+          "min_lr_ratio": 0.1,
+          "warmup_steps": 471,
+          "weight_decay": 0.1
+        },
+        "parameterization": {
+          "attention_scale": "inverse_head_dim",
+          "base_depth": 12,
+          "base_width": 384,
+          "batch_multiplier": 1.0,
+          "data_multiplier": 2.0604194304342345,
+          "depth_alpha": 1.0,
+          "depth_multiplier": 1.0,
+          "embeddings": "untied",
+          "init_std": 0.02,
+          "name": "complete_d_p",
+          "width_multiplier": 1.6666666666666667
+        },
+        "training": {
+          "batch_size": 16,
+          "dtype": "bfloat16",
+          "sampling": "shuffled_epochs",
+          "seq_len": 8192,
+          "steps": 4709,
+          "tokens_per_parameter": 4.999472268158278,
+          "train_tokens": 617218048
+        }
+      },
+      "schema_version": 2,
+      "sha256": "a4bad1be29b57d20ec7ae7b0d539f2e777d11f5539607655550b7afca185affe"
+    },
+    "loss_backend": "tiled",
+    "vocab_tile_size": 2048
+  },
+  "metrics": {
+    "achieved_tflops": 543.9659870683994,
+    "attention_tune_seconds": 0.0,
+    "base_learning_rate": 0.001953125,
+    "diagnostic_compile_seconds": 86.72710289398674,
+    "diagnostic_point_count": 472,
+    "diagnostics_every": 10,
+    "early_stopping_step": null,
+    "estimated_total_flops": 846442363009105920,
+    "eval_compile_seconds": 17.856146146019455,
+    "expert_top_k": 2,
+    "experts": 8,
+    "final_validation_seconds": 0.820926460030023,
+    "flop_accounting": {
+      "by_site": {
+        "dot_general": 3116881674240,
+        "pallas_call": 1932735283200,
+        "tpu_flash_causal_attention_bwd_dkv": 2061584302080,
+        "tpu_flash_causal_attention_bwd_dq": 2061584302080,
+        "tpu_flash_causal_attention_fwd": 2061584302080
+      },
+      "elementwise_per_sequence": 13316361400,
+      "matmul_per_sequence": 11234369863680,
+      "method": "traced-jaxpr",
+      "warnings": []
+    },
+    "flops_per_token": 1371383040,
+    "mfu_estimate": 0.12362863342463623,
+    "model_tier": "125m",
+    "parameter_count": 123525760,
+    "parameters": 241628800,
+    "schedule_steps": 4709,
+    "tokens_per_parameter": 4.999472268158278,
+    "tokens_per_second": 396655.03451785387,
+    "tokens_processed": 617218048,
+    "total_compile_seconds": 154.1182081500301,
+    "total_parameter_count": 241628800,
+    "train_compile_seconds": 49.53495911002392,
+    "train_loss": 3.56113862991333,
+    "train_seconds": 1556.0575167040224,
+    "training_data_epochs": 0.07812906684585416,
+    "training_data_sharding": "rank_disjoint_shuffled_windows",
+    "training_sampling": "shuffled_epochs",
+    "training_steps": 4709,
+    "training_token_budget": 617218048,
+    "training_usable_tokens_per_epoch": 7899979776,
+    "validation_loss": 3.636278748512268,
+    "validation_probe_count": 0,
+    "validation_probe_seconds": 0.0,
+    "validation_tokens": 1048576
+  },
+  "profile": "dev",
+  "schema_version": 1,
+  "seed": 1337,
+  "status": "ok",
+  "system": {
+    "controller_process_index": 1,
+    "device_count": 16,
+    "device_ids": [
+      0,
+      1,
+      2,
+      3,
+      4,
+      5,
+      6,
+      7,
+      8,
+      9,
+      10,
+      11,
+      12,
+      13,
+      14,
+      15
+    ],
+    "device_kinds": [
+      "TPU v4"
+    ],
+    "jax_version": "0.11.0",
+    "jaxlib_version": "0.11.0",
+    "libtpu_version": "0.0.44.1",
+    "local_device_count": 4,
+    "platform": "tpu",
+    "process_count": 4,
+    "process_indices": [
+      0,
+      0,
+      0,
+      0,
+      1,
+      1,
+      1,
+      1,
+      2,
+      2,
+      2,
+      2,
+      3,
+      3,
+      3,
+      3
+    ],
+    "python_version": "3.12.13"
+  },
+  "track": "open"
+}

moe-lr-sweep-8k/125m-5tpp-bs16-lr2e-9-s1337/result.json

@@ -0,0 +1,243 @@
+{
+  "artifacts": {
+    "diagnostics": "diagnostics.riglog",
+    "training_curve": "training.riglog",
+    "validation_curve": "validation.csv"
+  },
+  "checkpoint": null,
+  "contract": {
+    "dataset_id": "fineweb-8b-gpt2",
+    "model": {
+      "d_model": 640,
+      "heads": 10,
+      "layers": 12,
+      "mlp_activation": "gelu",
+      "mlp_mult": 4,
+      "normalization": "rms_norm",
+      "parameterization": "complete_d_p",
+      "position_encoding": "rope_base_10000",
+      "semantic_vocab_size": 50304,
+      "tied_embeddings": false,
+      "tier": "125m",
+      "vocab_size": 50304
+    },
+    "model_id": "reference-gpt-v3-family",
+    "sequence_length": 8192,
+    "tokenizer_id": "gpt2"
+  },
+  "evaluations": {
+    "fineweb": {
+      "canonical": true,
+      "loss": 3.636278748512268,
+      "perplexity": 37.95035083687206,
+      "scored_tokens": 1048576,
+      "seconds": 0.820926460030023
+    }
+  },
+  "implementation": {
+    "attention_backend": "tpu_flash",
+    "attention_tuning": {
+      "key_digest": "bf93286e9d38983bc78d617a9a666f8b00490f5bfff49b38c7fff97fd570e3ea",
+      "resolution_source": "heuristic",
+      "tiles": {
+        "block_kv": 512,
+        "block_kv_compute": 256,
+        "block_kv_dkv": 512,
+        "block_kv_dkv_compute": 256,
+        "block_kv_dq": 512,
+        "block_kv_dq_compute": 256,
+        "block_q": 512,
+        "block_q_dkv": 512,
+        "block_q_dkv_compute": 256,
+        "block_q_dq": 256
+      },
+      "tune_seconds": 0.0
+    },
+    "configuration": {
+      "overrides": {
+        "tokens_per_parameter_micros": 5000000
+      },
+      "path": "config.yaml",
+      "profile": "dev",
+      "resolved": {
+        "evaluation": {
+          "eval_batches": 8,
+          "val_every": 0,
+          "val_probe_batches": 8
+        },
+        "kernels": {
+          "attention_backend": "tpu_flash",
+          "loss_backend": "tiled",
+          "vocab_tile_size": 2048
+        },
+        "logging": {
+          "diagnostics_every": 10,
+          "log_every": 10
+        },
+        "model": {
+          "d_model": 640,
+          "heads": 10,
+          "layers": 12,
+          "mlp_activation": "gelu",
+          "mlp_mult": 4,
+          "normalization": "rms_norm",
+          "parameterization": "complete_d_p",
+          "position_encoding": "rope_base_10000",
+          "semantic_vocab_size": 50304,
+          "tied_embeddings": false,
+          "tier": "125m",
+          "vocab_size": 50304
+        },
+        "optimizer": {
+          "adam_epsilon": 1e-08,
+          "beta1": 0.9,
+          "beta2": 0.95,
+          "effective": {
+            "adam_epsilon_horizon_multiplier": 1.4354161175193187,
+            "beta1": 0.9514661925028901,
+            "beta2": 0.975733096251445,
+            "global_peak_learning_rate": 0.00136066815480335,
+            "weight_decay_horizon_multiplier": 0.6966620952593152
+          },
+          "grad_clip": 0.0,
+          "learning_rate": 0.001953125,
+          "min_lr_ratio": 0.1,
+          "warmup_steps": 471,
+          "weight_decay": 0.1
+        },
+        "parameterization": {
+          "attention_scale": "inverse_head_dim",
+          "base_depth": 12,
+          "base_width": 384,
+          "batch_multiplier": 1.0,
+          "data_multiplier": 2.0604194304342345,
+          "depth_alpha": 1.0,
+          "depth_multiplier": 1.0,
+          "embeddings": "untied",
+          "init_std": 0.02,
+          "name": "complete_d_p",
+          "width_multiplier": 1.6666666666666667
+        },
+        "training": {
+          "batch_size": 16,
+          "dtype": "bfloat16",
+          "sampling": "shuffled_epochs",
+          "seq_len": 8192,
+          "steps": 4709,
+          "tokens_per_parameter": 4.999472268158278,
+          "train_tokens": 617218048
+        }
+      },
+      "schema_version": 2,
+      "sha256": "a4bad1be29b57d20ec7ae7b0d539f2e777d11f5539607655550b7afca185affe"
+    },
+    "loss_backend": "tiled",
+    "vocab_tile_size": 2048
+  },
+  "metrics": {
+    "achieved_tflops": 543.9659870683994,
+    "attention_tune_seconds": 0.0,
+    "base_learning_rate": 0.001953125,
+    "diagnostic_compile_seconds": 86.72710289398674,
+    "diagnostic_point_count": 472,
+    "diagnostics_every": 10,
+    "early_stopping_step": null,
+    "estimated_total_flops": 846442363009105920,
+    "eval_compile_seconds": 17.856146146019455,
+    "expert_top_k": 2,
+    "experts": 8,
+    "final_validation_seconds": 0.820926460030023,
+    "flop_accounting": {
+      "by_site": {
+        "dot_general": 3116881674240,
+        "pallas_call": 1932735283200,
+        "tpu_flash_causal_attention_bwd_dkv": 2061584302080,
+        "tpu_flash_causal_attention_bwd_dq": 2061584302080,
+        "tpu_flash_causal_attention_fwd": 2061584302080
+      },
+      "elementwise_per_sequence": 13316361400,
+      "matmul_per_sequence": 11234369863680,
+      "method": "traced-jaxpr",
+      "warnings": []
+    },
+    "flops_per_token": 1371383040,
+    "mfu_estimate": 0.12362863342463623,
+    "model_tier": "125m",
+    "parameter_count": 123525760,
+    "parameters": 241628800,
+    "schedule_steps": 4709,
+    "tokens_per_parameter": 4.999472268158278,
+    "tokens_per_second": 396655.03451785387,
+    "tokens_processed": 617218048,
+    "total_compile_seconds": 154.1182081500301,
+    "total_parameter_count": 241628800,
+    "train_compile_seconds": 49.53495911002392,
+    "train_loss": 3.56113862991333,
+    "train_seconds": 1556.0575167040224,
+    "training_data_epochs": 0.07812906684585416,
+    "training_data_sharding": "rank_disjoint_shuffled_windows",
+    "training_sampling": "shuffled_epochs",
+    "training_steps": 4709,
+    "training_token_budget": 617218048,
+    "training_usable_tokens_per_epoch": 7899979776,
+    "validation_loss": 3.636278748512268,
+    "validation_probe_count": 0,
+    "validation_probe_seconds": 0.0,
+    "validation_tokens": 1048576
+  },
+  "profile": "dev",
+  "schema_version": 1,
+  "seed": 1337,
+  "status": "ok",
+  "system": {
+    "controller_process_index": 1,
+    "device_count": 16,
+    "device_ids": [
+      0,
+      1,
+      2,
+      3,
+      4,
+      5,
+      6,
+      7,
+      8,
+      9,
+      10,
+      11,
+      12,
+      13,
+      14,
+      15
+    ],
+    "device_kinds": [
+      "TPU v4"
+    ],
+    "jax_version": "0.11.0",
+    "jaxlib_version": "0.11.0",
+    "libtpu_version": "0.0.44.1",
+    "local_device_count": 4,
+    "platform": "tpu",
+    "process_count": 4,
+    "process_indices": [
+      0,
+      0,
+      0,
+      0,
+      1,
+      1,
+      1,
+      1,
+      2,
+      2,
+      2,
+      2,
+      3,
+      3,
+      3,
+      3
+    ],
+    "python_version": "3.12.13"
+  },
+  "track": "open"
+}

moe-lr-sweep-8k/125m-5tpp-bs16-lr2e-9-s1337/training.riglog

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:f46d2f1d84742558f0e76c980a10c2a27351c0e0b141655706fe1eebc0a12631
+size 2678128

moe-lr-sweep-8k/125m-5tpp-bs16-lr2e-9-s1337/validation.csv

@@ -0,0 +1,2 @@
+step,tokens_processed,kind,domain,validation_tokens,validation_loss,perplexity,validation_seconds,canonical
+4709,617218048,fineweb,fineweb,1048576,3.636278748512268,37.95035083687206,0.820926460030023,true

moe-lr-sweep-8k/60m-5tpp-bs16-lr2e-10-s1337/diagnostics.riglog

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:c038a5d5f83af15b22d3179e36611ef098c008dfc10ed94f84517500dc4da943
+size 272824

moe-lr-sweep-8k/60m-5tpp-bs16-lr2e-10-s1337/metrics.json

@@ -0,0 +1,243 @@
+{
+  "artifacts": {
+    "diagnostics": "diagnostics.riglog",
+    "training_curve": "training.riglog",
+    "validation_curve": "validation.csv"
+  },
+  "checkpoint": null,
+  "contract": {
+    "dataset_id": "fineweb-8b-gpt2",
+    "model": {
+      "d_model": 384,
+      "heads": 6,
+      "layers": 12,
+      "mlp_activation": "gelu",
+      "mlp_mult": 4,
+      "normalization": "rms_norm",
+      "parameterization": "complete_d_p",
+      "position_encoding": "rope_base_10000",
+      "semantic_vocab_size": 50304,
+      "tied_embeddings": false,
+      "tier": "60m",
+      "vocab_size": 50304
+    },
+    "model_id": "reference-gpt-v3-family",
+    "sequence_length": 8192,
+    "tokenizer_id": "gpt2"
+  },
+  "evaluations": {
+    "fineweb": {
+      "canonical": true,
+      "loss": 4.140182375907898,
+      "perplexity": 62.81427621528696,
+      "scored_tokens": 1048576,
+      "seconds": 0.4522744969581254
+    }
+  },
+  "implementation": {
+    "attention_backend": "tpu_flash",
+    "attention_tuning": {
+      "key_digest": "ec38db3431308f3e9f99acf40c7e3bb85ca0695b0d97491e44a93bc7adfada00",
+      "resolution_source": "heuristic",
+      "tiles": {
+        "block_kv": 512,
+        "block_kv_compute": 256,
+        "block_kv_dkv": 512,
+        "block_kv_dkv_compute": 256,
+        "block_kv_dq": 512,
+        "block_kv_dq_compute": 256,
+        "block_q": 512,
+        "block_q_dkv": 512,
+        "block_q_dkv_compute": 256,
+        "block_q_dq": 256
+      },
+      "tune_seconds": 0.0
+    },
+    "configuration": {
+      "overrides": {
+        "tokens_per_parameter_micros": 5000000
+      },
+      "path": "config.yaml",
+      "profile": "dev",
+      "resolved": {
+        "evaluation": {
+          "eval_batches": 8,
+          "val_every": 0,
+          "val_probe_batches": 8
+        },
+        "kernels": {
+          "attention_backend": "tpu_flash",
+          "loss_backend": "tiled",
+          "vocab_tile_size": 2048
+        },
+        "logging": {
+          "diagnostics_every": 10,
+          "log_every": 10
+        },
+        "model": {
+          "d_model": 384,
+          "heads": 6,
+          "layers": 12,
+          "mlp_activation": "gelu",
+          "mlp_mult": 4,
+          "normalization": "rms_norm",
+          "parameterization": "complete_d_p",
+          "position_encoding": "rope_base_10000",
+          "semantic_vocab_size": 50304,
+          "tied_embeddings": false,
+          "tier": "60m",
+          "vocab_size": 50304
+        },
+        "optimizer": {
+          "adam_epsilon": 1e-08,
+          "beta1": 0.9,
+          "beta2": 0.95,
+          "effective": {
+            "adam_epsilon_horizon_multiplier": 1.0,
+            "beta1": 0.9,
+            "beta2": 0.95,
+            "global_peak_learning_rate": 0.0009765625,
+            "weight_decay_horizon_multiplier": 1.0
+          },
+          "grad_clip": 0.0,
+          "learning_rate": 0.0009765625,
+          "min_lr_ratio": 0.1,
+          "warmup_steps": 229,
+          "weight_decay": 0.1
+        },
+        "parameterization": {
+          "attention_scale": "inverse_head_dim",
+          "base_depth": 12,
+          "base_width": 384,
+          "batch_multiplier": 1.0,
+          "data_multiplier": 1.0,
+          "depth_alpha": 1.0,
+          "depth_multiplier": 1.0,
+          "embeddings": "untied",
+          "init_std": 0.02,
+          "name": "complete_d_p",
+          "width_multiplier": 1.0
+        },
+        "training": {
+          "batch_size": 16,
+          "dtype": "bfloat16",
+          "sampling": "shuffled_epochs",
+          "seq_len": 8192,
+          "steps": 2286,
+          "tokens_per_parameter": 5.000660099848113,
+          "train_tokens": 299630592
+        }
+      },
+      "schema_version": 2,
+      "sha256": "a4bad1be29b57d20ec7ae7b0d539f2e777d11f5539607655550b7afca185affe"
+    },
+    "loss_backend": "tiled",
+    "vocab_tile_size": 2048
+  },
+  "metrics": {
+    "achieved_tflops": 597.1994385567854,
+    "attention_tune_seconds": 0.0,
+    "base_learning_rate": 0.0009765625,
+    "diagnostic_compile_seconds": 83.89820041001076,
+    "diagnostic_point_count": 230,
+    "diagnostics_every": 10,
+    "early_stopping_step": null,
+    "estimated_total_flops": 221095966021779456,
+    "eval_compile_seconds": 14.67128990398487,
+    "expert_top_k": 2,
+    "experts": 8,
+    "final_validation_seconds": 0.4522744969581254,
+    "flop_accounting": {
+      "by_site": {
+        "dot_general": 1638200770560,
+        "pallas_call": 695784701952,
+        "tpu_flash_causal_attention_bwd_dkv": 1236950581248,
+        "tpu_flash_causal_attention_bwd_dq": 1236950581248,
+        "tpu_flash_causal_attention_fwd": 1236950581248
+      },
+      "elementwise_per_sequence": 9518381752,
+      "matmul_per_sequence": 6044837216256,
+      "method": "traced-jaxpr",
+      "warnings": []
+    },
+    "flops_per_token": 737895168,
+    "mfu_estimate": 0.13572714512654213,
+    "model_tier": "60m",
+    "parameter_count": 59959680,
+    "parameters": 102509952,
+    "schedule_steps": 2286,
+    "tokens_per_parameter": 5.000660099848113,
+    "tokens_per_second": 809328.2954751445,
+    "tokens_processed": 299630592,
+    "total_compile_seconds": 145.22406579001108,
+    "total_parameter_count": 102509952,
+    "train_compile_seconds": 46.65457547601545,
+    "train_loss": 4.212207794189453,
+    "train_seconds": 370.2213226390304,
+    "training_data_epochs": 0.037928020133706226,
+    "training_data_sharding": "rank_disjoint_shuffled_windows",
+    "training_sampling": "shuffled_epochs",
+    "training_steps": 2286,
+    "training_token_budget": 299630592,
+    "training_usable_tokens_per_epoch": 7899979776,
+    "validation_loss": 4.140182375907898,
+    "validation_probe_count": 0,
+    "validation_probe_seconds": 0.0,
+    "validation_tokens": 1048576
+  },
+  "profile": "dev",
+  "schema_version": 1,
+  "seed": 1337,
+  "status": "ok",
+  "system": {
+    "controller_process_index": 1,
+    "device_count": 16,
+    "device_ids": [
+      0,
+      1,
+      2,
+      3,
+      4,
+      5,
+      6,
+      7,
+      8,
+      9,
+      10,
+      11,
+      12,
+      13,
+      14,
+      15
+    ],
+    "device_kinds": [
+      "TPU v4"
+    ],
+    "jax_version": "0.11.0",
+    "jaxlib_version": "0.11.0",
+    "libtpu_version": "0.0.44.1",
+    "local_device_count": 4,
+    "platform": "tpu",
+    "process_count": 4,
+    "process_indices": [
+      0,
+      0,
+      0,
+      0,
+      1,
+      1,
+      1,
+      1,
+      2,
+      2,
+      2,
+      2,
+      3,
+      3,
+      3,
+      3
+    ],
+    "python_version": "3.12.13"
+  },
+  "track": "open"
+}

moe-lr-sweep-8k/60m-5tpp-bs16-lr2e-10-s1337/result.json

@@ -0,0 +1,243 @@
+{
+  "artifacts": {
+    "diagnostics": "diagnostics.riglog",
+    "training_curve": "training.riglog",
+    "validation_curve": "validation.csv"
+  },
+  "checkpoint": null,
+  "contract": {
+    "dataset_id": "fineweb-8b-gpt2",
+    "model": {
+      "d_model": 384,
+      "heads": 6,
+      "layers": 12,
+      "mlp_activation": "gelu",
+      "mlp_mult": 4,
+      "normalization": "rms_norm",
+      "parameterization": "complete_d_p",
+      "position_encoding": "rope_base_10000",
+      "semantic_vocab_size": 50304,
+      "tied_embeddings": false,
+      "tier": "60m",
+      "vocab_size": 50304
+    },
+    "model_id": "reference-gpt-v3-family",
+    "sequence_length": 8192,
+    "tokenizer_id": "gpt2"
+  },
+  "evaluations": {
+    "fineweb": {
+      "canonical": true,
+      "loss": 4.140182375907898,
+      "perplexity": 62.81427621528696,
+      "scored_tokens": 1048576,
+      "seconds": 0.4522744969581254
+    }
+  },
+  "implementation": {
+    "attention_backend": "tpu_flash",
+    "attention_tuning": {
+      "key_digest": "ec38db3431308f3e9f99acf40c7e3bb85ca0695b0d97491e44a93bc7adfada00",
+      "resolution_source": "heuristic",
+      "tiles": {
+        "block_kv": 512,
+        "block_kv_compute": 256,
+        "block_kv_dkv": 512,
+        "block_kv_dkv_compute": 256,
+        "block_kv_dq": 512,
+        "block_kv_dq_compute": 256,
+        "block_q": 512,
+        "block_q_dkv": 512,
+        "block_q_dkv_compute": 256,
+        "block_q_dq": 256
+      },
+      "tune_seconds": 0.0
+    },
+    "configuration": {
+      "overrides": {
+        "tokens_per_parameter_micros": 5000000
+      },
+      "path": "config.yaml",
+      "profile": "dev",
+      "resolved": {
+        "evaluation": {
+          "eval_batches": 8,
+          "val_every": 0,
+          "val_probe_batches": 8
+        },
+        "kernels": {
+          "attention_backend": "tpu_flash",
+          "loss_backend": "tiled",
+          "vocab_tile_size": 2048
+        },
+        "logging": {
+          "diagnostics_every": 10,
+          "log_every": 10
+        },
+        "model": {
+          "d_model": 384,
+          "heads": 6,
+          "layers": 12,
+          "mlp_activation": "gelu",
+          "mlp_mult": 4,
+          "normalization": "rms_norm",
+          "parameterization": "complete_d_p",
+          "position_encoding": "rope_base_10000",
+          "semantic_vocab_size": 50304,
+          "tied_embeddings": false,
+          "tier": "60m",
+          "vocab_size": 50304
+        },
+        "optimizer": {
+          "adam_epsilon": 1e-08,
+          "beta1": 0.9,
+          "beta2": 0.95,
+          "effective": {
+            "adam_epsilon_horizon_multiplier": 1.0,
+            "beta1": 0.9,
+            "beta2": 0.95,
+            "global_peak_learning_rate": 0.0009765625,
+            "weight_decay_horizon_multiplier": 1.0
+          },
+          "grad_clip": 0.0,
+          "learning_rate": 0.0009765625,
+          "min_lr_ratio": 0.1,
+          "warmup_steps": 229,
+          "weight_decay": 0.1
+        },
+        "parameterization": {
+          "attention_scale": "inverse_head_dim",
+          "base_depth": 12,
+          "base_width": 384,
+          "batch_multiplier": 1.0,
+          "data_multiplier": 1.0,
+          "depth_alpha": 1.0,
+          "depth_multiplier": 1.0,
+          "embeddings": "untied",
+          "init_std": 0.02,
+          "name": "complete_d_p",
+          "width_multiplier": 1.0
+        },
+        "training": {
+          "batch_size": 16,
+          "dtype": "bfloat16",
+          "sampling": "shuffled_epochs",
+          "seq_len": 8192,
+          "steps": 2286,
+          "tokens_per_parameter": 5.000660099848113,
+          "train_tokens": 299630592
+        }
+      },
+      "schema_version": 2,
+      "sha256": "a4bad1be29b57d20ec7ae7b0d539f2e777d11f5539607655550b7afca185affe"
+    },
+    "loss_backend": "tiled",
+    "vocab_tile_size": 2048
+  },
+  "metrics": {
+    "achieved_tflops": 597.1994385567854,
+    "attention_tune_seconds": 0.0,
+    "base_learning_rate": 0.0009765625,
+    "diagnostic_compile_seconds": 83.89820041001076,
+    "diagnostic_point_count": 230,
+    "diagnostics_every": 10,
+    "early_stopping_step": null,
+    "estimated_total_flops": 221095966021779456,
+    "eval_compile_seconds": 14.67128990398487,
+    "expert_top_k": 2,
+    "experts": 8,
+    "final_validation_seconds": 0.4522744969581254,
+    "flop_accounting": {
+      "by_site": {
+        "dot_general": 1638200770560,
+        "pallas_call": 695784701952,
+        "tpu_flash_causal_attention_bwd_dkv": 1236950581248,
+        "tpu_flash_causal_attention_bwd_dq": 1236950581248,
+        "tpu_flash_causal_attention_fwd": 1236950581248
+      },
+      "elementwise_per_sequence": 9518381752,
+      "matmul_per_sequence": 6044837216256,
+      "method": "traced-jaxpr",
+      "warnings": []
+    },
+    "flops_per_token": 737895168,
+    "mfu_estimate": 0.13572714512654213,
+    "model_tier": "60m",
+    "parameter_count": 59959680,
+    "parameters": 102509952,
+    "schedule_steps": 2286,
+    "tokens_per_parameter": 5.000660099848113,
+    "tokens_per_second": 809328.2954751445,
+    "tokens_processed": 299630592,
+    "total_compile_seconds": 145.22406579001108,
+    "total_parameter_count": 102509952,
+    "train_compile_seconds": 46.65457547601545,
+    "train_loss": 4.212207794189453,
+    "train_seconds": 370.2213226390304,
+    "training_data_epochs": 0.037928020133706226,
+    "training_data_sharding": "rank_disjoint_shuffled_windows",
+    "training_sampling": "shuffled_epochs",
+    "training_steps": 2286,
+    "training_token_budget": 299630592,
+    "training_usable_tokens_per_epoch": 7899979776,
+    "validation_loss": 4.140182375907898,
+    "validation_probe_count": 0,
+    "validation_probe_seconds": 0.0,
+    "validation_tokens": 1048576
+  },
+  "profile": "dev",
+  "schema_version": 1,
+  "seed": 1337,
+  "status": "ok",
+  "system": {
+    "controller_process_index": 1,
+    "device_count": 16,
+    "device_ids": [
+      0,
+      1,
+      2,
+      3,
+      4,
+      5,
+      6,
+      7,
+      8,
+      9,
+      10,
+      11,
+      12,
+      13,
+      14,
+      15
+    ],
+    "device_kinds": [
+      "TPU v4"
+    ],
+    "jax_version": "0.11.0",
+    "jaxlib_version": "0.11.0",
+    "libtpu_version": "0.0.44.1",
+    "local_device_count": 4,
+    "platform": "tpu",
+    "process_count": 4,
+    "process_indices": [
+      0,
+      0,
+      0,
+      0,
+      1,
+      1,
+      1,
+      1,
+      2,
+      2,
+      2,
+      2,
+      3,
+      3,
+      3,
+      3
+    ],
+    "python_version": "3.12.13"
+  },
+  "track": "open"
+}

moe-lr-sweep-8k/60m-5tpp-bs16-lr2e-10-s1337/training.riglog

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:8eababbce742ceb8f3d3ac1ee872460c3637a856180b34681d47fc22312081f3
+size 1301864

moe-lr-sweep-8k/60m-5tpp-bs16-lr2e-10-s1337/validation.csv

@@ -0,0 +1,2 @@
+step,tokens_processed,kind,domain,validation_tokens,validation_loss,perplexity,validation_seconds,canonical
+2286,299630592,fineweb,fineweb,1048576,4.140182375907898,62.81427621528696,0.4522744969581254,true

moe-lr-sweep-8k/60m-5tpp-bs16-lr2e-10-s1338/diagnostics.riglog

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:b283d39fcce7fa570c608c7ee0f63b60685531973ee6e8f86dfd13b5d3f361b3
+size 272824

moe-lr-sweep-8k/60m-5tpp-bs16-lr2e-10-s1338/metrics.json

@@ -0,0 +1,243 @@
+{
+  "artifacts": {
+    "diagnostics": "diagnostics.riglog",
+    "training_curve": "training.riglog",
+    "validation_curve": "validation.csv"
+  },
+  "checkpoint": null,
+  "contract": {
+    "dataset_id": "fineweb-8b-gpt2",
+    "model": {
+      "d_model": 384,
+      "heads": 6,
+      "layers": 12,
+      "mlp_activation": "gelu",
+      "mlp_mult": 4,
+      "normalization": "rms_norm",
+      "parameterization": "complete_d_p",
+      "position_encoding": "rope_base_10000",
+      "semantic_vocab_size": 50304,
+      "tied_embeddings": false,
+      "tier": "60m",
+      "vocab_size": 50304
+    },
+    "model_id": "reference-gpt-v3-family",
+    "sequence_length": 8192,
+    "tokenizer_id": "gpt2"
+  },
+  "evaluations": {
+    "fineweb": {
+      "canonical": true,
+      "loss": 4.1316089034080505,
+      "perplexity": 62.278041726011224,
+      "scored_tokens": 1048576,
+      "seconds": 0.4497800850076601
+    }
+  },
+  "implementation": {
+    "attention_backend": "tpu_flash",
+    "attention_tuning": {
+      "key_digest": "ec38db3431308f3e9f99acf40c7e3bb85ca0695b0d97491e44a93bc7adfada00",
+      "resolution_source": "heuristic",
+      "tiles": {
+        "block_kv": 512,
+        "block_kv_compute": 256,
+        "block_kv_dkv": 512,
+        "block_kv_dkv_compute": 256,
+        "block_kv_dq": 512,
+        "block_kv_dq_compute": 256,
+        "block_q": 512,
+        "block_q_dkv": 512,
+        "block_q_dkv_compute": 256,
+        "block_q_dq": 256
+      },
+      "tune_seconds": 0.0
+    },
+    "configuration": {
+      "overrides": {
+        "tokens_per_parameter_micros": 5000000
+      },
+      "path": "config.yaml",
+      "profile": "dev",
+      "resolved": {
+        "evaluation": {
+          "eval_batches": 8,
+          "val_every": 0,
+          "val_probe_batches": 8
+        },
+        "kernels": {
+          "attention_backend": "tpu_flash",
+          "loss_backend": "tiled",
+          "vocab_tile_size": 2048
+        },
+        "logging": {
+          "diagnostics_every": 10,
+          "log_every": 10
+        },
+        "model": {
+          "d_model": 384,
+          "heads": 6,
+          "layers": 12,
+          "mlp_activation": "gelu",
+          "mlp_mult": 4,
+          "normalization": "rms_norm",
+          "parameterization": "complete_d_p",
+          "position_encoding": "rope_base_10000",
+          "semantic_vocab_size": 50304,
+          "tied_embeddings": false,
+          "tier": "60m",
+          "vocab_size": 50304
+        },
+        "optimizer": {
+          "adam_epsilon": 1e-08,
+          "beta1": 0.9,
+          "beta2": 0.95,
+          "effective": {
+            "adam_epsilon_horizon_multiplier": 1.0,
+            "beta1": 0.9,
+            "beta2": 0.95,
+            "global_peak_learning_rate": 0.0009765625,
+            "weight_decay_horizon_multiplier": 1.0
+          },
+          "grad_clip": 0.0,
+          "learning_rate": 0.0009765625,
+          "min_lr_ratio": 0.1,
+          "warmup_steps": 229,
+          "weight_decay": 0.1
+        },
+        "parameterization": {
+          "attention_scale": "inverse_head_dim",
+          "base_depth": 12,
+          "base_width": 384,
+          "batch_multiplier": 1.0,
+          "data_multiplier": 1.0,
+          "depth_alpha": 1.0,
+          "depth_multiplier": 1.0,
+          "embeddings": "untied",
+          "init_std": 0.02,
+          "name": "complete_d_p",
+          "width_multiplier": 1.0
+        },
+        "training": {
+          "batch_size": 16,
+          "dtype": "bfloat16",
+          "sampling": "shuffled_epochs",
+          "seq_len": 8192,
+          "steps": 2286,
+          "tokens_per_parameter": 5.000660099848113,
+          "train_tokens": 299630592
+        }
+      },
+      "schema_version": 2,
+      "sha256": "a4bad1be29b57d20ec7ae7b0d539f2e777d11f5539607655550b7afca185affe"
+    },
+    "loss_backend": "tiled",
+    "vocab_tile_size": 2048
+  },
+  "metrics": {
+    "achieved_tflops": 596.5610758674637,
+    "attention_tune_seconds": 0.0,
+    "base_learning_rate": 0.0009765625,
+    "diagnostic_compile_seconds": 81.42221394897206,
+    "diagnostic_point_count": 230,
+    "diagnostics_every": 10,
+    "early_stopping_step": null,
+    "estimated_total_flops": 221095966021779456,
+    "eval_compile_seconds": 15.508095662982669,
+    "expert_top_k": 2,
+    "experts": 8,
+    "final_validation_seconds": 0.4497800850076601,
+    "flop_accounting": {
+      "by_site": {
+        "dot_general": 1638200770560,
+        "pallas_call": 695784701952,
+        "tpu_flash_causal_attention_bwd_dkv": 1236950581248,
+        "tpu_flash_causal_attention_bwd_dq": 1236950581248,
+        "tpu_flash_causal_attention_fwd": 1236950581248
+      },
+      "elementwise_per_sequence": 9518381752,
+      "matmul_per_sequence": 6044837216256,
+      "method": "traced-jaxpr",
+      "warnings": []
+    },
+    "flops_per_token": 737895168,
+    "mfu_estimate": 0.13558206269715084,
+    "model_tier": "60m",
+    "parameter_count": 59959680,
+    "parameters": 102509952,
+    "schedule_steps": 2286,
+    "tokens_per_parameter": 5.000660099848113,
+    "tokens_per_second": 808463.1824929686,
+    "tokens_processed": 299630592,
+    "total_compile_seconds": 144.46100250293966,
+    "total_parameter_count": 102509952,
+    "train_compile_seconds": 47.53069289098494,
+    "train_loss": 4.155775547027588,
+    "train_seconds": 370.61748572898796,
+    "training_data_epochs": 0.037928020133706226,
+    "training_data_sharding": "rank_disjoint_shuffled_windows",
+    "training_sampling": "shuffled_epochs",
+    "training_steps": 2286,
+    "training_token_budget": 299630592,
+    "training_usable_tokens_per_epoch": 7899979776,
+    "validation_loss": 4.1316089034080505,
+    "validation_probe_count": 0,
+    "validation_probe_seconds": 0.0,
+    "validation_tokens": 1048576
+  },
+  "profile": "dev",
+  "schema_version": 1,
+  "seed": 1338,
+  "status": "ok",
+  "system": {
+    "controller_process_index": 1,
+    "device_count": 16,
+    "device_ids": [
+      0,
+      1,
+      2,
+      3,
+      4,
+      5,
+      6,
+      7,
+      8,
+      9,
+      10,
+      11,
+      12,
+      13,
+      14,
+      15
+    ],
+    "device_kinds": [
+      "TPU v4"
+    ],
+    "jax_version": "0.11.0",
+    "jaxlib_version": "0.11.0",
+    "libtpu_version": "0.0.44.1",
+    "local_device_count": 4,
+    "platform": "tpu",
+    "process_count": 4,
+    "process_indices": [
+      0,
+      0,
+      0,
+      0,
+      1,
+      1,
+      1,
+      1,
+      2,
+      2,
+      2,
+      2,
+      3,
+      3,
+      3,
+      3
+    ],
+    "python_version": "3.12.13"
+  },
+  "track": "open"
+}

moe-lr-sweep-8k/60m-5tpp-bs16-lr2e-10-s1338/result.json

@@ -0,0 +1,243 @@
+{
+  "artifacts": {
+    "diagnostics": "diagnostics.riglog",
+    "training_curve": "training.riglog",
+    "validation_curve": "validation.csv"
+  },
+  "checkpoint": null,
+  "contract": {
+    "dataset_id": "fineweb-8b-gpt2",
+    "model": {
+      "d_model": 384,
+      "heads": 6,
+      "layers": 12,
+      "mlp_activation": "gelu",
+      "mlp_mult": 4,
+      "normalization": "rms_norm",
+      "parameterization": "complete_d_p",
+      "position_encoding": "rope_base_10000",
+      "semantic_vocab_size": 50304,
+      "tied_embeddings": false,
+      "tier": "60m",
+      "vocab_size": 50304
+    },
+    "model_id": "reference-gpt-v3-family",
+    "sequence_length": 8192,
+    "tokenizer_id": "gpt2"
+  },
+  "evaluations": {
+    "fineweb": {
+      "canonical": true,
+      "loss": 4.1316089034080505,
+      "perplexity": 62.278041726011224,
+      "scored_tokens": 1048576,
+      "seconds": 0.4497800850076601
+    }
+  },
+  "implementation": {
+    "attention_backend": "tpu_flash",
+    "attention_tuning": {
+      "key_digest": "ec38db3431308f3e9f99acf40c7e3bb85ca0695b0d97491e44a93bc7adfada00",
+      "resolution_source": "heuristic",
+      "tiles": {
+        "block_kv": 512,
+        "block_kv_compute": 256,
+        "block_kv_dkv": 512,
+        "block_kv_dkv_compute": 256,
+        "block_kv_dq": 512,
+        "block_kv_dq_compute": 256,
+        "block_q": 512,
+        "block_q_dkv": 512,
+        "block_q_dkv_compute": 256,
+        "block_q_dq": 256
+      },
+      "tune_seconds": 0.0
+    },
+    "configuration": {
+      "overrides": {
+        "tokens_per_parameter_micros": 5000000
+      },
+      "path": "config.yaml",
+      "profile": "dev",
+      "resolved": {
+        "evaluation": {
+          "eval_batches": 8,
+          "val_every": 0,
+          "val_probe_batches": 8
+        },
+        "kernels": {
+          "attention_backend": "tpu_flash",
+          "loss_backend": "tiled",
+          "vocab_tile_size": 2048
+        },
+        "logging": {
+          "diagnostics_every": 10,
+          "log_every": 10
+        },
+        "model": {
+          "d_model": 384,
+          "heads": 6,
+          "layers": 12,
+          "mlp_activation": "gelu",
+          "mlp_mult": 4,
+          "normalization": "rms_norm",
+          "parameterization": "complete_d_p",
+          "position_encoding": "rope_base_10000",
+          "semantic_vocab_size": 50304,
+          "tied_embeddings": false,
+          "tier": "60m",
+          "vocab_size": 50304
+        },
+        "optimizer": {
+          "adam_epsilon": 1e-08,
+          "beta1": 0.9,
+          "beta2": 0.95,
+          "effective": {
+            "adam_epsilon_horizon_multiplier": 1.0,
+            "beta1": 0.9,
+            "beta2": 0.95,
+            "global_peak_learning_rate": 0.0009765625,
+            "weight_decay_horizon_multiplier": 1.0
+          },
+          "grad_clip": 0.0,
+          "learning_rate": 0.0009765625,
+          "min_lr_ratio": 0.1,
+          "warmup_steps": 229,
+          "weight_decay": 0.1
+        },
+        "parameterization": {
+          "attention_scale": "inverse_head_dim",
+          "base_depth": 12,
+          "base_width": 384,
+          "batch_multiplier": 1.0,
+          "data_multiplier": 1.0,
+          "depth_alpha": 1.0,
+          "depth_multiplier": 1.0,
+          "embeddings": "untied",
+          "init_std": 0.02,
+          "name": "complete_d_p",
+          "width_multiplier": 1.0
+        },
+        "training": {
+          "batch_size": 16,
+          "dtype": "bfloat16",
+          "sampling": "shuffled_epochs",
+          "seq_len": 8192,
+          "steps": 2286,
+          "tokens_per_parameter": 5.000660099848113,
+          "train_tokens": 299630592
+        }
+      },
+      "schema_version": 2,
+      "sha256": "a4bad1be29b57d20ec7ae7b0d539f2e777d11f5539607655550b7afca185affe"
+    },
+    "loss_backend": "tiled",
+    "vocab_tile_size": 2048
+  },
+  "metrics": {
+    "achieved_tflops": 596.5610758674637,
+    "attention_tune_seconds": 0.0,
+    "base_learning_rate": 0.0009765625,
+    "diagnostic_compile_seconds": 81.42221394897206,
+    "diagnostic_point_count": 230,
+    "diagnostics_every": 10,
+    "early_stopping_step": null,
+    "estimated_total_flops": 221095966021779456,
+    "eval_compile_seconds": 15.508095662982669,
+    "expert_top_k": 2,
+    "experts": 8,
+    "final_validation_seconds": 0.4497800850076601,
+    "flop_accounting": {
+      "by_site": {
+        "dot_general": 1638200770560,
+        "pallas_call": 695784701952,
+        "tpu_flash_causal_attention_bwd_dkv": 1236950581248,
+        "tpu_flash_causal_attention_bwd_dq": 1236950581248,
+        "tpu_flash_causal_attention_fwd": 1236950581248
+      },
+      "elementwise_per_sequence": 9518381752,
+      "matmul_per_sequence": 6044837216256,
+      "method": "traced-jaxpr",
+      "warnings": []
+    },
+    "flops_per_token": 737895168,
+    "mfu_estimate": 0.13558206269715084,
+    "model_tier": "60m",
+    "parameter_count": 59959680,
+    "parameters": 102509952,
+    "schedule_steps": 2286,
+    "tokens_per_parameter": 5.000660099848113,
+    "tokens_per_second": 808463.1824929686,
+    "tokens_processed": 299630592,
+    "total_compile_seconds": 144.46100250293966,
+    "total_parameter_count": 102509952,
+    "train_compile_seconds": 47.53069289098494,
+    "train_loss": 4.155775547027588,
+    "train_seconds": 370.61748572898796,
+    "training_data_epochs": 0.037928020133706226,
+    "training_data_sharding": "rank_disjoint_shuffled_windows",
+    "training_sampling": "shuffled_epochs",
+    "training_steps": 2286,
+    "training_token_budget": 299630592,
+    "training_usable_tokens_per_epoch": 7899979776,
+    "validation_loss": 4.1316089034080505,
+    "validation_probe_count": 0,
+    "validation_probe_seconds": 0.0,
+    "validation_tokens": 1048576
+  },
+  "profile": "dev",
+  "schema_version": 1,
+  "seed": 1338,
+  "status": "ok",
+  "system": {
+    "controller_process_index": 1,
+    "device_count": 16,
+    "device_ids": [
+      0,
+      1,
+      2,
+      3,
+      4,
+      5,
+      6,
+      7,
+      8,
+      9,
+      10,
+      11,
+      12,
+      13,
+      14,
+      15
+    ],
+    "device_kinds": [
+      "TPU v4"
+    ],
+    "jax_version": "0.11.0",
+    "jaxlib_version": "0.11.0",
+    "libtpu_version": "0.0.44.1",
+    "local_device_count": 4,
+    "platform": "tpu",
+    "process_count": 4,
+    "process_indices": [
+      0,
+      0,
+      0,
+      0,
+      1,
+      1,
+      1,
+      1,
+      2,
+      2,
+      2,
+      2,
+      3,
+      3,
+      3,
+      3
+    ],
+    "python_version": "3.12.13"
+  },
+  "track": "open"
+}

moe-lr-sweep-8k/60m-5tpp-bs16-lr2e-10-s1338/training.riglog

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:5237c2cb19951ee6a2ae63c8a1775b3292dacb2a5451e107970b5d53d7c401ab
+size 1301864

moe-lr-sweep-8k/60m-5tpp-bs16-lr2e-10-s1338/validation.csv

@@ -0,0 +1,2 @@
+step,tokens_processed,kind,domain,validation_tokens,validation_loss,perplexity,validation_seconds,canonical
+2286,299630592,fineweb,fineweb,1048576,4.1316089034080505,62.278041726011224,0.4497800850076601,true

moe-lr-sweep-8k/60m-5tpp-bs16-lr2e-10-s1339/diagnostics.riglog

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:47c912f8aded72e15b289bb7113686557ab379f43d19c131282127f5de98ab72
+size 272824

moe-lr-sweep-8k/60m-5tpp-bs16-lr2e-10-s1339/metrics.json

@@ -0,0 +1,243 @@
+{
+  "artifacts": {
+    "diagnostics": "diagnostics.riglog",
+    "training_curve": "training.riglog",
+    "validation_curve": "validation.csv"
+  },
+  "checkpoint": null,
+  "contract": {
+    "dataset_id": "fineweb-8b-gpt2",
+    "model": {
+      "d_model": 384,
+      "heads": 6,
+      "layers": 12,
+      "mlp_activation": "gelu",
+      "mlp_mult": 4,
+      "normalization": "rms_norm",
+      "parameterization": "complete_d_p",
+      "position_encoding": "rope_base_10000",
+      "semantic_vocab_size": 50304,
+      "tied_embeddings": false,
+      "tier": "60m",
+      "vocab_size": 50304
+    },
+    "model_id": "reference-gpt-v3-family",
+    "sequence_length": 8192,
+    "tokenizer_id": "gpt2"
+  },
+  "evaluations": {
+    "fineweb": {
+      "canonical": true,
+      "loss": 4.118675529956818,
+      "perplexity": 61.477762862592776,
+      "scored_tokens": 1048576,
+      "seconds": 0.44936058699386194
+    }
+  },
+  "implementation": {
+    "attention_backend": "tpu_flash",
+    "attention_tuning": {
+      "key_digest": "ec38db3431308f3e9f99acf40c7e3bb85ca0695b0d97491e44a93bc7adfada00",
+      "resolution_source": "heuristic",
+      "tiles": {
+        "block_kv": 512,
+        "block_kv_compute": 256,
+        "block_kv_dkv": 512,
+        "block_kv_dkv_compute": 256,
+        "block_kv_dq": 512,
+        "block_kv_dq_compute": 256,
+        "block_q": 512,
+        "block_q_dkv": 512,
+        "block_q_dkv_compute": 256,
+        "block_q_dq": 256
+      },
+      "tune_seconds": 0.0
+    },
+    "configuration": {
+      "overrides": {
+        "tokens_per_parameter_micros": 5000000
+      },
+      "path": "config.yaml",
+      "profile": "dev",
+      "resolved": {
+        "evaluation": {
+          "eval_batches": 8,
+          "val_every": 0,
+          "val_probe_batches": 8
+        },
+        "kernels": {
+          "attention_backend": "tpu_flash",
+          "loss_backend": "tiled",
+          "vocab_tile_size": 2048
+        },
+        "logging": {
+          "diagnostics_every": 10,
+          "log_every": 10
+        },
+        "model": {
+          "d_model": 384,
+          "heads": 6,
+          "layers": 12,
+          "mlp_activation": "gelu",
+          "mlp_mult": 4,
+          "normalization": "rms_norm",
+          "parameterization": "complete_d_p",
+          "position_encoding": "rope_base_10000",
+          "semantic_vocab_size": 50304,
+          "tied_embeddings": false,
+          "tier": "60m",
+          "vocab_size": 50304
+        },
+        "optimizer": {
+          "adam_epsilon": 1e-08,
+          "beta1": 0.9,
+          "beta2": 0.95,
+          "effective": {
+            "adam_epsilon_horizon_multiplier": 1.0,
+            "beta1": 0.9,
+            "beta2": 0.95,
+            "global_peak_learning_rate": 0.0009765625,
+            "weight_decay_horizon_multiplier": 1.0
+          },
+          "grad_clip": 0.0,
+          "learning_rate": 0.0009765625,
+          "min_lr_ratio": 0.1,
+          "warmup_steps": 229,
+          "weight_decay": 0.1
+        },
+        "parameterization": {
+          "attention_scale": "inverse_head_dim",
+          "base_depth": 12,
+          "base_width": 384,
+          "batch_multiplier": 1.0,
+          "data_multiplier": 1.0,
+          "depth_alpha": 1.0,
+          "depth_multiplier": 1.0,
+          "embeddings": "untied",
+          "init_std": 0.02,
+          "name": "complete_d_p",
+          "width_multiplier": 1.0
+        },
+        "training": {
+          "batch_size": 16,
+          "dtype": "bfloat16",
+          "sampling": "shuffled_epochs",
+          "seq_len": 8192,
+          "steps": 2286,
+          "tokens_per_parameter": 5.000660099848113,
+          "train_tokens": 299630592
+        }
+      },
+      "schema_version": 2,
+      "sha256": "a4bad1be29b57d20ec7ae7b0d539f2e777d11f5539607655550b7afca185affe"
+    },
+    "loss_backend": "tiled",
+    "vocab_tile_size": 2048
+  },
+  "metrics": {
+    "achieved_tflops": 597.5103503897918,
+    "attention_tune_seconds": 0.0,
+    "base_learning_rate": 0.0009765625,
+    "diagnostic_compile_seconds": 89.28556085104356,
+    "diagnostic_point_count": 230,
+    "diagnostics_every": 10,
+    "early_stopping_step": null,
+    "estimated_total_flops": 221095966021779456,
+    "eval_compile_seconds": 15.467254528019112,
+    "expert_top_k": 2,
+    "experts": 8,
+    "final_validation_seconds": 0.44936058699386194,
+    "flop_accounting": {
+      "by_site": {
+        "dot_general": 1638200770560,
+        "pallas_call": 695784701952,
+        "tpu_flash_causal_attention_bwd_dkv": 1236950581248,
+        "tpu_flash_causal_attention_bwd_dq": 1236950581248,
+        "tpu_flash_causal_attention_fwd": 1236950581248
+      },
+      "elementwise_per_sequence": 9518381752,
+      "matmul_per_sequence": 6044837216256,
+      "method": "traced-jaxpr",
+      "warnings": []
+    },
+    "flops_per_token": 737895168,
+    "mfu_estimate": 0.13579780690677087,
+    "model_tier": "60m",
+    "parameter_count": 59959680,
+    "parameters": 102509952,
+    "schedule_steps": 2286,
+    "tokens_per_parameter": 5.000660099848113,
+    "tokens_per_second": 809749.6450739625,
+    "tokens_processed": 299630592,
+    "total_compile_seconds": 154.75221040507313,
+    "total_parameter_count": 102509952,
+    "train_compile_seconds": 49.99939502601046,
+    "train_loss": 3.9968578815460205,
+    "train_seconds": 370.0286796329892,
+    "training_data_epochs": 0.037928020133706226,
+    "training_data_sharding": "rank_disjoint_shuffled_windows",
+    "training_sampling": "shuffled_epochs",
+    "training_steps": 2286,
+    "training_token_budget": 299630592,
+    "training_usable_tokens_per_epoch": 7899979776,
+    "validation_loss": 4.118675529956818,
+    "validation_probe_count": 0,
+    "validation_probe_seconds": 0.0,
+    "validation_tokens": 1048576
+  },
+  "profile": "dev",
+  "schema_version": 1,
+  "seed": 1339,
+  "status": "ok",
+  "system": {
+    "controller_process_index": 1,
+    "device_count": 16,
+    "device_ids": [
+      0,
+      1,
+      2,
+      3,
+      4,
+      5,
+      6,
+      7,
+      8,
+      9,
+      10,
+      11,
+      12,
+      13,
+      14,
+      15
+    ],
+    "device_kinds": [
+      "TPU v4"
+    ],
+    "jax_version": "0.11.0",
+    "jaxlib_version": "0.11.0",
+    "libtpu_version": "0.0.44.1",
+    "local_device_count": 4,
+    "platform": "tpu",
+    "process_count": 4,
+    "process_indices": [
+      0,
+      0,
+      0,
+      0,
+      1,
+      1,
+      1,
+      1,
+      2,
+      2,
+      2,
+      2,
+      3,
+      3,
+      3,
+      3
+    ],
+    "python_version": "3.12.13"
+  },
+  "track": "open"
+}

moe-lr-sweep-8k/60m-5tpp-bs16-lr2e-10-s1339/result.json

@@ -0,0 +1,243 @@
+{
+  "artifacts": {
+    "diagnostics": "diagnostics.riglog",
+    "training_curve": "training.riglog",
+    "validation_curve": "validation.csv"
+  },
+  "checkpoint": null,
+  "contract": {
+    "dataset_id": "fineweb-8b-gpt2",
+    "model": {
+      "d_model": 384,
+      "heads": 6,
+      "layers": 12,
+      "mlp_activation": "gelu",
+      "mlp_mult": 4,
+      "normalization": "rms_norm",
+      "parameterization": "complete_d_p",
+      "position_encoding": "rope_base_10000",
+      "semantic_vocab_size": 50304,
+      "tied_embeddings": false,
+      "tier": "60m",
+      "vocab_size": 50304
+    },
+    "model_id": "reference-gpt-v3-family",
+    "sequence_length": 8192,
+    "tokenizer_id": "gpt2"
+  },
+  "evaluations": {
+    "fineweb": {
+      "canonical": true,
+      "loss": 4.118675529956818,
+      "perplexity": 61.477762862592776,
+      "scored_tokens": 1048576,
+      "seconds": 0.44936058699386194
+    }
+  },
+  "implementation": {
+    "attention_backend": "tpu_flash",
+    "attention_tuning": {
+      "key_digest": "ec38db3431308f3e9f99acf40c7e3bb85ca0695b0d97491e44a93bc7adfada00",
+      "resolution_source": "heuristic",
+      "tiles": {
+        "block_kv": 512,
+        "block_kv_compute": 256,
+        "block_kv_dkv": 512,
+        "block_kv_dkv_compute": 256,
+        "block_kv_dq": 512,
+        "block_kv_dq_compute": 256,
+        "block_q": 512,
+        "block_q_dkv": 512,
+        "block_q_dkv_compute": 256,
+        "block_q_dq": 256
+      },
+      "tune_seconds": 0.0
+    },
+    "configuration": {
+      "overrides": {
+        "tokens_per_parameter_micros": 5000000
+      },
+      "path": "config.yaml",
+      "profile": "dev",
+      "resolved": {
+        "evaluation": {
+          "eval_batches": 8,
+          "val_every": 0,
+          "val_probe_batches": 8
+        },
+        "kernels": {
+          "attention_backend": "tpu_flash",
+          "loss_backend": "tiled",
+          "vocab_tile_size": 2048
+        },
+        "logging": {
+          "diagnostics_every": 10,
+          "log_every": 10
+        },
+        "model": {
+          "d_model": 384,
+          "heads": 6,
+          "layers": 12,
+          "mlp_activation": "gelu",
+          "mlp_mult": 4,
+          "normalization": "rms_norm",
+          "parameterization": "complete_d_p",
+          "position_encoding": "rope_base_10000",
+          "semantic_vocab_size": 50304,
+          "tied_embeddings": false,
+          "tier": "60m",
+          "vocab_size": 50304
+        },
+        "optimizer": {
+          "adam_epsilon": 1e-08,
+          "beta1": 0.9,
+          "beta2": 0.95,
+          "effective": {
+            "adam_epsilon_horizon_multiplier": 1.0,
+            "beta1": 0.9,
+            "beta2": 0.95,
+            "global_peak_learning_rate": 0.0009765625,
+            "weight_decay_horizon_multiplier": 1.0
+          },
+          "grad_clip": 0.0,
+          "learning_rate": 0.0009765625,
+          "min_lr_ratio": 0.1,
+          "warmup_steps": 229,
+          "weight_decay": 0.1
+        },
+        "parameterization": {
+          "attention_scale": "inverse_head_dim",
+          "base_depth": 12,
+          "base_width": 384,
+          "batch_multiplier": 1.0,
+          "data_multiplier": 1.0,
+          "depth_alpha": 1.0,
+          "depth_multiplier": 1.0,
+          "embeddings": "untied",
+          "init_std": 0.02,
+          "name": "complete_d_p",
+          "width_multiplier": 1.0
+        },
+        "training": {
+          "batch_size": 16,
+          "dtype": "bfloat16",
+          "sampling": "shuffled_epochs",
+          "seq_len": 8192,
+          "steps": 2286,
+          "tokens_per_parameter": 5.000660099848113,
+          "train_tokens": 299630592
+        }
+      },
+      "schema_version": 2,
+      "sha256": "a4bad1be29b57d20ec7ae7b0d539f2e777d11f5539607655550b7afca185affe"
+    },
+    "loss_backend": "tiled",
+    "vocab_tile_size": 2048
+  },
+  "metrics": {
+    "achieved_tflops": 597.5103503897918,
+    "attention_tune_seconds": 0.0,
+    "base_learning_rate": 0.0009765625,
+    "diagnostic_compile_seconds": 89.28556085104356,
+    "diagnostic_point_count": 230,
+    "diagnostics_every": 10,
+    "early_stopping_step": null,
+    "estimated_total_flops": 221095966021779456,
+    "eval_compile_seconds": 15.467254528019112,
+    "expert_top_k": 2,
+    "experts": 8,
+    "final_validation_seconds": 0.44936058699386194,
+    "flop_accounting": {
+      "by_site": {
+        "dot_general": 1638200770560,
+        "pallas_call": 695784701952,
+        "tpu_flash_causal_attention_bwd_dkv": 1236950581248,
+        "tpu_flash_causal_attention_bwd_dq": 1236950581248,
+        "tpu_flash_causal_attention_fwd": 1236950581248
+      },
+      "elementwise_per_sequence": 9518381752,
+      "matmul_per_sequence": 6044837216256,
+      "method": "traced-jaxpr",
+      "warnings": []
+    },
+    "flops_per_token": 737895168,
+    "mfu_estimate": 0.13579780690677087,
+    "model_tier": "60m",
+    "parameter_count": 59959680,
+    "parameters": 102509952,
+    "schedule_steps": 2286,
+    "tokens_per_parameter": 5.000660099848113,
+    "tokens_per_second": 809749.6450739625,
+    "tokens_processed": 299630592,
+    "total_compile_seconds": 154.75221040507313,
+    "total_parameter_count": 102509952,
+    "train_compile_seconds": 49.99939502601046,
+    "train_loss": 3.9968578815460205,
+    "train_seconds": 370.0286796329892,
+    "training_data_epochs": 0.037928020133706226,
+    "training_data_sharding": "rank_disjoint_shuffled_windows",
+    "training_sampling": "shuffled_epochs",
+    "training_steps": 2286,
+    "training_token_budget": 299630592,
+    "training_usable_tokens_per_epoch": 7899979776,
+    "validation_loss": 4.118675529956818,
+    "validation_probe_count": 0,
+    "validation_probe_seconds": 0.0,
+    "validation_tokens": 1048576
+  },
+  "profile": "dev",
+  "schema_version": 1,
+  "seed": 1339,
+  "status": "ok",
+  "system": {
+    "controller_process_index": 1,
+    "device_count": 16,
+    "device_ids": [
+      0,
+      1,
+      2,
+      3,
+      4,
+      5,
+      6,
+      7,
+      8,
+      9,
+      10,
+      11,
+      12,
+      13,
+      14,
+      15
+    ],
+    "device_kinds": [
+      "TPU v4"
+    ],
+    "jax_version": "0.11.0",
+    "jaxlib_version": "0.11.0",
+    "libtpu_version": "0.0.44.1",
+    "local_device_count": 4,
+    "platform": "tpu",
+    "process_count": 4,
+    "process_indices": [
+      0,
+      0,
+      0,
+      0,
+      1,
+      1,
+      1,
+      1,
+      2,
+      2,
+      2,
+      2,
+      3,
+      3,
+      3,
+      3
+    ],
+    "python_version": "3.12.13"
+  },
+  "track": "open"
+}

moe-lr-sweep-8k/60m-5tpp-bs16-lr2e-10-s1339/training.riglog

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:1de69488eabf0579a0cecb5fa31d1402fd82eaa4e8ac0591c3b0e51538e121bb
+size 1301864

moe-lr-sweep-8k/60m-5tpp-bs16-lr2e-10-s1339/validation.csv

@@ -0,0 +1,2 @@
+step,tokens_processed,kind,domain,validation_tokens,validation_loss,perplexity,validation_seconds,canonical
+2286,299630592,fineweb,fineweb,1048576,4.118675529956818,61.477762862592776,0.44936058699386194,true

moe-lr-sweep-8k/60m-5tpp-bs16-lr2e-6-s1337/diagnostics.riglog

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:f91fb635351dbd43ed42b81ece5b89079252822186797b2a3ff0e4456ba3dcbf
+size 272824

moe-lr-sweep-8k/60m-5tpp-bs16-lr2e-6-s1337/metrics.json

@@ -0,0 +1,243 @@
+{
+  "artifacts": {
+    "diagnostics": "diagnostics.riglog",
+    "training_curve": "training.riglog",
+    "validation_curve": "validation.csv"
+  },
+  "checkpoint": null,
+  "contract": {
+    "dataset_id": "fineweb-8b-gpt2",
+    "model": {
+      "d_model": 384,
+      "heads": 6,
+      "layers": 12,
+      "mlp_activation": "gelu",
+      "mlp_mult": 4,
+      "normalization": "rms_norm",
+      "parameterization": "complete_d_p",
+      "position_encoding": "rope_base_10000",
+      "semantic_vocab_size": 50304,
+      "tied_embeddings": false,
+      "tier": "60m",
+      "vocab_size": 50304
+    },
+    "model_id": "reference-gpt-v3-family",
+    "sequence_length": 8192,
+    "tokenizer_id": "gpt2"
+  },
+  "evaluations": {
+    "fineweb": {
+      "canonical": true,
+      "loss": 4.004855066537857,
+      "perplexity": 54.86387221186876,
+      "scored_tokens": 1048576,
+      "seconds": 0.44860493595479056
+    }
+  },
+  "implementation": {
+    "attention_backend": "tpu_flash",
+    "attention_tuning": {
+      "key_digest": "ec38db3431308f3e9f99acf40c7e3bb85ca0695b0d97491e44a93bc7adfada00",
+      "resolution_source": "heuristic",
+      "tiles": {
+        "block_kv": 512,
+        "block_kv_compute": 256,
+        "block_kv_dkv": 512,
+        "block_kv_dkv_compute": 256,
+        "block_kv_dq": 512,
+        "block_kv_dq_compute": 256,
+        "block_q": 512,
+        "block_q_dkv": 512,
+        "block_q_dkv_compute": 256,
+        "block_q_dq": 256
+      },
+      "tune_seconds": 0.0
+    },
+    "configuration": {
+      "overrides": {
+        "tokens_per_parameter_micros": 5000000
+      },
+      "path": "config.yaml",
+      "profile": "dev",
+      "resolved": {
+        "evaluation": {
+          "eval_batches": 8,
+          "val_every": 0,
+          "val_probe_batches": 8
+        },
+        "kernels": {
+          "attention_backend": "tpu_flash",
+          "loss_backend": "tiled",
+          "vocab_tile_size": 2048
+        },
+        "logging": {
+          "diagnostics_every": 10,
+          "log_every": 10
+        },
+        "model": {
+          "d_model": 384,
+          "heads": 6,
+          "layers": 12,
+          "mlp_activation": "gelu",
+          "mlp_mult": 4,
+          "normalization": "rms_norm",
+          "parameterization": "complete_d_p",
+          "position_encoding": "rope_base_10000",
+          "semantic_vocab_size": 50304,
+          "tied_embeddings": false,
+          "tier": "60m",
+          "vocab_size": 50304
+        },
+        "optimizer": {
+          "adam_epsilon": 1e-08,
+          "beta1": 0.9,
+          "beta2": 0.95,
+          "effective": {
+            "adam_epsilon_horizon_multiplier": 1.0,
+            "beta1": 0.9,
+            "beta2": 0.95,
+            "global_peak_learning_rate": 0.015625,
+            "weight_decay_horizon_multiplier": 1.0
+          },
+          "grad_clip": 0.0,
+          "learning_rate": 0.015625,
+          "min_lr_ratio": 0.1,
+          "warmup_steps": 229,
+          "weight_decay": 0.1
+        },
+        "parameterization": {
+          "attention_scale": "inverse_head_dim",
+          "base_depth": 12,
+          "base_width": 384,
+          "batch_multiplier": 1.0,
+          "data_multiplier": 1.0,
+          "depth_alpha": 1.0,
+          "depth_multiplier": 1.0,
+          "embeddings": "untied",
+          "init_std": 0.02,
+          "name": "complete_d_p",
+          "width_multiplier": 1.0
+        },
+        "training": {
+          "batch_size": 16,
+          "dtype": "bfloat16",
+          "sampling": "shuffled_epochs",
+          "seq_len": 8192,
+          "steps": 2286,
+          "tokens_per_parameter": 5.000660099848113,
+          "train_tokens": 299630592
+        }
+      },
+      "schema_version": 2,
+      "sha256": "a4bad1be29b57d20ec7ae7b0d539f2e777d11f5539607655550b7afca185affe"
+    },
+    "loss_backend": "tiled",
+    "vocab_tile_size": 2048
+  },
+  "metrics": {
+    "achieved_tflops": 597.6152834934923,
+    "attention_tune_seconds": 0.0,
+    "base_learning_rate": 0.015625,
+    "diagnostic_compile_seconds": 88.96127280901419,
+    "diagnostic_point_count": 230,
+    "diagnostics_every": 10,
+    "early_stopping_step": null,
+    "estimated_total_flops": 221095966021779456,
+    "eval_compile_seconds": 15.446089457953349,
+    "expert_top_k": 2,
+    "experts": 8,
+    "final_validation_seconds": 0.44860493595479056,
+    "flop_accounting": {
+      "by_site": {
+        "dot_general": 1638200770560,
+        "pallas_call": 695784701952,
+        "tpu_flash_causal_attention_bwd_dkv": 1236950581248,
+        "tpu_flash_causal_attention_bwd_dq": 1236950581248,
+        "tpu_flash_causal_attention_fwd": 1236950581248
+      },
+      "elementwise_per_sequence": 9518381752,
+      "matmul_per_sequence": 6044837216256,
+      "method": "traced-jaxpr",
+      "warnings": []
+    },
+    "flops_per_token": 737895168,
+    "mfu_estimate": 0.13582165533943005,
+    "model_tier": "60m",
+    "parameter_count": 59959680,
+    "parameters": 102509952,
+    "schedule_steps": 2286,
+    "tokens_per_parameter": 5.000660099848113,
+    "tokens_per_second": 809891.85105152,
+    "tokens_processed": 299630592,
+    "total_compile_seconds": 151.08850227796938,
+    "total_parameter_count": 102509952,
+    "train_compile_seconds": 46.681140011001844,
+    "train_loss": 4.086471080780029,
+    "train_seconds": 369.9637076369836,
+    "training_data_epochs": 0.037928020133706226,
+    "training_data_sharding": "rank_disjoint_shuffled_windows",
+    "training_sampling": "shuffled_epochs",
+    "training_steps": 2286,
+    "training_token_budget": 299630592,
+    "training_usable_tokens_per_epoch": 7899979776,
+    "validation_loss": 4.004855066537857,
+    "validation_probe_count": 0,
+    "validation_probe_seconds": 0.0,
+    "validation_tokens": 1048576
+  },
+  "profile": "dev",
+  "schema_version": 1,
+  "seed": 1337,
+  "status": "ok",
+  "system": {
+    "controller_process_index": 1,
+    "device_count": 16,
+    "device_ids": [
+      0,
+      1,
+      2,
+      3,
+      4,
+      5,
+      6,
+      7,
+      8,
+      9,
+      10,
+      11,
+      12,
+      13,
+      14,
+      15
+    ],
+    "device_kinds": [
+      "TPU v4"
+    ],
+    "jax_version": "0.11.0",
+    "jaxlib_version": "0.11.0",
+    "libtpu_version": "0.0.44.1",
+    "local_device_count": 4,
+    "platform": "tpu",
+    "process_count": 4,
+    "process_indices": [
+      0,
+      0,
+      0,
+      0,
+      1,
+      1,
+      1,
+      1,
+      2,
+      2,
+      2,
+      2,
+      3,
+      3,
+      3,
+      3
+    ],
+    "python_version": "3.12.13"
+  },
+  "track": "open"
+}

moe-lr-sweep-8k/60m-5tpp-bs16-lr2e-6-s1337/result.json

@@ -0,0 +1,243 @@
+{
+  "artifacts": {
+    "diagnostics": "diagnostics.riglog",
+    "training_curve": "training.riglog",
+    "validation_curve": "validation.csv"
+  },
+  "checkpoint": null,
+  "contract": {
+    "dataset_id": "fineweb-8b-gpt2",
+    "model": {
+      "d_model": 384,
+      "heads": 6,
+      "layers": 12,
+      "mlp_activation": "gelu",
+      "mlp_mult": 4,
+      "normalization": "rms_norm",
+      "parameterization": "complete_d_p",
+      "position_encoding": "rope_base_10000",
+      "semantic_vocab_size": 50304,
+      "tied_embeddings": false,
+      "tier": "60m",
+      "vocab_size": 50304
+    },
+    "model_id": "reference-gpt-v3-family",
+    "sequence_length": 8192,
+    "tokenizer_id": "gpt2"
+  },
+  "evaluations": {
+    "fineweb": {
+      "canonical": true,
+      "loss": 4.004855066537857,
+      "perplexity": 54.86387221186876,
+      "scored_tokens": 1048576,
+      "seconds": 0.44860493595479056
+    }
+  },
+  "implementation": {
+    "attention_backend": "tpu_flash",
+    "attention_tuning": {
+      "key_digest": "ec38db3431308f3e9f99acf40c7e3bb85ca0695b0d97491e44a93bc7adfada00",
+      "resolution_source": "heuristic",
+      "tiles": {
+        "block_kv": 512,
+        "block_kv_compute": 256,
+        "block_kv_dkv": 512,
+        "block_kv_dkv_compute": 256,
+        "block_kv_dq": 512,
+        "block_kv_dq_compute": 256,
+        "block_q": 512,
+        "block_q_dkv": 512,
+        "block_q_dkv_compute": 256,
+        "block_q_dq": 256
+      },
+      "tune_seconds": 0.0
+    },
+    "configuration": {
+      "overrides": {
+        "tokens_per_parameter_micros": 5000000
+      },
+      "path": "config.yaml",
+      "profile": "dev",
+      "resolved": {
+        "evaluation": {
+          "eval_batches": 8,
+          "val_every": 0,
+          "val_probe_batches": 8
+        },
+        "kernels": {
+          "attention_backend": "tpu_flash",
+          "loss_backend": "tiled",
+          "vocab_tile_size": 2048
+        },
+        "logging": {
+          "diagnostics_every": 10,
+          "log_every": 10
+        },
+        "model": {
+          "d_model": 384,
+          "heads": 6,
+          "layers": 12,
+          "mlp_activation": "gelu",
+          "mlp_mult": 4,
+          "normalization": "rms_norm",
+          "parameterization": "complete_d_p",
+          "position_encoding": "rope_base_10000",
+          "semantic_vocab_size": 50304,
+          "tied_embeddings": false,
+          "tier": "60m",
+          "vocab_size": 50304
+        },
+        "optimizer": {
+          "adam_epsilon": 1e-08,
+          "beta1": 0.9,
+          "beta2": 0.95,
+          "effective": {
+            "adam_epsilon_horizon_multiplier": 1.0,
+            "beta1": 0.9,
+            "beta2": 0.95,
+            "global_peak_learning_rate": 0.015625,
+            "weight_decay_horizon_multiplier": 1.0
+          },
+          "grad_clip": 0.0,
+          "learning_rate": 0.015625,
+          "min_lr_ratio": 0.1,
+          "warmup_steps": 229,
+          "weight_decay": 0.1
+        },
+        "parameterization": {
+          "attention_scale": "inverse_head_dim",
+          "base_depth": 12,
+          "base_width": 384,
+          "batch_multiplier": 1.0,
+          "data_multiplier": 1.0,
+          "depth_alpha": 1.0,
+          "depth_multiplier": 1.0,
+          "embeddings": "untied",
+          "init_std": 0.02,
+          "name": "complete_d_p",
+          "width_multiplier": 1.0
+        },
+        "training": {
+          "batch_size": 16,
+          "dtype": "bfloat16",
+          "sampling": "shuffled_epochs",
+          "seq_len": 8192,
+          "steps": 2286,
+          "tokens_per_parameter": 5.000660099848113,
+          "train_tokens": 299630592
+        }
+      },
+      "schema_version": 2,
+      "sha256": "a4bad1be29b57d20ec7ae7b0d539f2e777d11f5539607655550b7afca185affe"
+    },
+    "loss_backend": "tiled",
+    "vocab_tile_size": 2048
+  },
+  "metrics": {
+    "achieved_tflops": 597.6152834934923,
+    "attention_tune_seconds": 0.0,
+    "base_learning_rate": 0.015625,
+    "diagnostic_compile_seconds": 88.96127280901419,
+    "diagnostic_point_count": 230,
+    "diagnostics_every": 10,
+    "early_stopping_step": null,
+    "estimated_total_flops": 221095966021779456,
+    "eval_compile_seconds": 15.446089457953349,
+    "expert_top_k": 2,
+    "experts": 8,
+    "final_validation_seconds": 0.44860493595479056,
+    "flop_accounting": {
+      "by_site": {
+        "dot_general": 1638200770560,
+        "pallas_call": 695784701952,
+        "tpu_flash_causal_attention_bwd_dkv": 1236950581248,
+        "tpu_flash_causal_attention_bwd_dq": 1236950581248,
+        "tpu_flash_causal_attention_fwd": 1236950581248
+      },
+      "elementwise_per_sequence": 9518381752,
+      "matmul_per_sequence": 6044837216256,
+      "method": "traced-jaxpr",
+      "warnings": []
+    },
+    "flops_per_token": 737895168,
+    "mfu_estimate": 0.13582165533943005,
+    "model_tier": "60m",
+    "parameter_count": 59959680,
+    "parameters": 102509952,
+    "schedule_steps": 2286,
+    "tokens_per_parameter": 5.000660099848113,
+    "tokens_per_second": 809891.85105152,
+    "tokens_processed": 299630592,
+    "total_compile_seconds": 151.08850227796938,
+    "total_parameter_count": 102509952,
+    "train_compile_seconds": 46.681140011001844,
+    "train_loss": 4.086471080780029,
+    "train_seconds": 369.9637076369836,
+    "training_data_epochs": 0.037928020133706226,
+    "training_data_sharding": "rank_disjoint_shuffled_windows",
+    "training_sampling": "shuffled_epochs",
+    "training_steps": 2286,
+    "training_token_budget": 299630592,
+    "training_usable_tokens_per_epoch": 7899979776,
+    "validation_loss": 4.004855066537857,
+    "validation_probe_count": 0,
+    "validation_probe_seconds": 0.0,
+    "validation_tokens": 1048576
+  },
+  "profile": "dev",
+  "schema_version": 1,
+  "seed": 1337,
+  "status": "ok",
+  "system": {
+    "controller_process_index": 1,
+    "device_count": 16,
+    "device_ids": [
+      0,
+      1,
+      2,
+      3,
+      4,
+      5,
+      6,
+      7,
+      8,
+      9,
+      10,
+      11,
+      12,
+      13,
+      14,
+      15
+    ],
+    "device_kinds": [
+      "TPU v4"
+    ],
+    "jax_version": "0.11.0",
+    "jaxlib_version": "0.11.0",
+    "libtpu_version": "0.0.44.1",
+    "local_device_count": 4,
+    "platform": "tpu",
+    "process_count": 4,
+    "process_indices": [
+      0,
+      0,
+      0,
+      0,
+      1,
+      1,
+      1,
+      1,
+      2,
+      2,
+      2,
+      2,
+      3,
+      3,
+      3,
+      3
+    ],
+    "python_version": "3.12.13"
+  },
+  "track": "open"
+}

moe-lr-sweep-8k/60m-5tpp-bs16-lr2e-6-s1337/training.riglog

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:046a6ec970a81a7dbb3676bcc4783bd2e5ddda48c4201823f2752add4e56c8bd
+size 1301864

moe-lr-sweep-8k/60m-5tpp-bs16-lr2e-6-s1337/validation.csv

@@ -0,0 +1,2 @@
+step,tokens_processed,kind,domain,validation_tokens,validation_loss,perplexity,validation_seconds,canonical
+2286,299630592,fineweb,fineweb,1048576,4.004855066537857,54.86387221186876,0.44860493595479056,true

moe-lr-sweep-8k/60m-5tpp-bs16-lr2e-6-s1338/diagnostics.riglog

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:b62c81ddcb5ee0ad2f099954624f5c86e1d6b4bab001342d1bfba0ca4a7ba0a7
+size 272824

moe-lr-sweep-8k/60m-5tpp-bs16-lr2e-6-s1338/metrics.json

@@ -0,0 +1,243 @@
+{
+  "artifacts": {
+    "diagnostics": "diagnostics.riglog",
+    "training_curve": "training.riglog",
+    "validation_curve": "validation.csv"
+  },
+  "checkpoint": null,
+  "contract": {
+    "dataset_id": "fineweb-8b-gpt2",
+    "model": {
+      "d_model": 384,
+      "heads": 6,
+      "layers": 12,
+      "mlp_activation": "gelu",
+      "mlp_mult": 4,
+      "normalization": "rms_norm",
+      "parameterization": "complete_d_p",
+      "position_encoding": "rope_base_10000",
+      "semantic_vocab_size": 50304,
+      "tied_embeddings": false,
+      "tier": "60m",
+      "vocab_size": 50304
+    },
+    "model_id": "reference-gpt-v3-family",
+    "sequence_length": 8192,
+    "tokenizer_id": "gpt2"
+  },
+  "evaluations": {
+    "fineweb": {
+      "canonical": true,
+      "loss": 3.9843336641788483,
+      "perplexity": 53.74946235063435,
+      "scored_tokens": 1048576,
+      "seconds": 0.4521297650062479
+    }
+  },
+  "implementation": {
+    "attention_backend": "tpu_flash",
+    "attention_tuning": {
+      "key_digest": "ec38db3431308f3e9f99acf40c7e3bb85ca0695b0d97491e44a93bc7adfada00",
+      "resolution_source": "heuristic",
+      "tiles": {
+        "block_kv": 512,
+        "block_kv_compute": 256,
+        "block_kv_dkv": 512,
+        "block_kv_dkv_compute": 256,
+        "block_kv_dq": 512,
+        "block_kv_dq_compute": 256,
+        "block_q": 512,
+        "block_q_dkv": 512,
+        "block_q_dkv_compute": 256,
+        "block_q_dq": 256
+      },
+      "tune_seconds": 0.0
+    },
+    "configuration": {
+      "overrides": {
+        "tokens_per_parameter_micros": 5000000
+      },
+      "path": "config.yaml",
+      "profile": "dev",
+      "resolved": {
+        "evaluation": {
+          "eval_batches": 8,
+          "val_every": 0,
+          "val_probe_batches": 8
+        },
+        "kernels": {
+          "attention_backend": "tpu_flash",
+          "loss_backend": "tiled",
+          "vocab_tile_size": 2048
+        },
+        "logging": {
+          "diagnostics_every": 10,
+          "log_every": 10
+        },
+        "model": {
+          "d_model": 384,
+          "heads": 6,
+          "layers": 12,
+          "mlp_activation": "gelu",
+          "mlp_mult": 4,
+          "normalization": "rms_norm",
+          "parameterization": "complete_d_p",
+          "position_encoding": "rope_base_10000",
+          "semantic_vocab_size": 50304,
+          "tied_embeddings": false,
+          "tier": "60m",
+          "vocab_size": 50304
+        },
+        "optimizer": {
+          "adam_epsilon": 1e-08,
+          "beta1": 0.9,
+          "beta2": 0.95,
+          "effective": {
+            "adam_epsilon_horizon_multiplier": 1.0,
+            "beta1": 0.9,
+            "beta2": 0.95,
+            "global_peak_learning_rate": 0.015625,
+            "weight_decay_horizon_multiplier": 1.0
+          },
+          "grad_clip": 0.0,
+          "learning_rate": 0.015625,
+          "min_lr_ratio": 0.1,
+          "warmup_steps": 229,
+          "weight_decay": 0.1
+        },
+        "parameterization": {
+          "attention_scale": "inverse_head_dim",
+          "base_depth": 12,
+          "base_width": 384,
+          "batch_multiplier": 1.0,
+          "data_multiplier": 1.0,
+          "depth_alpha": 1.0,
+          "depth_multiplier": 1.0,
+          "embeddings": "untied",
+          "init_std": 0.02,
+          "name": "complete_d_p",
+          "width_multiplier": 1.0
+        },
+        "training": {
+          "batch_size": 16,
+          "dtype": "bfloat16",
+          "sampling": "shuffled_epochs",
+          "seq_len": 8192,
+          "steps": 2286,
+          "tokens_per_parameter": 5.000660099848113,
+          "train_tokens": 299630592
+        }
+      },
+      "schema_version": 2,
+      "sha256": "a4bad1be29b57d20ec7ae7b0d539f2e777d11f5539607655550b7afca185affe"
+    },
+    "loss_backend": "tiled",
+    "vocab_tile_size": 2048
+  },
+  "metrics": {
+    "achieved_tflops": 597.4534119803859,
+    "attention_tune_seconds": 0.0,
+    "base_learning_rate": 0.015625,
+    "diagnostic_compile_seconds": 85.61127052997472,
+    "diagnostic_point_count": 230,
+    "diagnostics_every": 10,
+    "early_stopping_step": null,
+    "estimated_total_flops": 221095966021779456,
+    "eval_compile_seconds": 15.335515389975626,
+    "expert_top_k": 2,
+    "experts": 8,
+    "final_validation_seconds": 0.4521297650062479,
+    "flop_accounting": {
+      "by_site": {
+        "dot_general": 1638200770560,
+        "pallas_call": 695784701952,
+        "tpu_flash_causal_attention_bwd_dkv": 1236950581248,
+        "tpu_flash_causal_attention_bwd_dq": 1236950581248,
+        "tpu_flash_causal_attention_fwd": 1236950581248
+      },
+      "elementwise_per_sequence": 9518381752,
+      "matmul_per_sequence": 6044837216256,
+      "method": "traced-jaxpr",
+      "warnings": []
+    },
+    "flops_per_token": 737895168,
+    "mfu_estimate": 0.1357848663591786,
+    "model_tier": "60m",
+    "parameter_count": 59959680,
+    "parameters": 102509952,
+    "schedule_steps": 2286,
+    "tokens_per_parameter": 5.000660099848113,
+    "tokens_per_second": 809672.4817967447,
+    "tokens_processed": 299630592,
+    "total_compile_seconds": 146.9452016199357,
+    "total_parameter_count": 102509952,
+    "train_compile_seconds": 45.998415699985344,
+    "train_loss": 3.988403797149658,
+    "train_seconds": 370.06394404696766,
+    "training_data_epochs": 0.037928020133706226,
+    "training_data_sharding": "rank_disjoint_shuffled_windows",
+    "training_sampling": "shuffled_epochs",
+    "training_steps": 2286,
+    "training_token_budget": 299630592,
+    "training_usable_tokens_per_epoch": 7899979776,
+    "validation_loss": 3.9843336641788483,
+    "validation_probe_count": 0,
+    "validation_probe_seconds": 0.0,
+    "validation_tokens": 1048576
+  },
+  "profile": "dev",
+  "schema_version": 1,
+  "seed": 1338,
+  "status": "ok",
+  "system": {
+    "controller_process_index": 1,
+    "device_count": 16,
+    "device_ids": [
+      0,
+      1,
+      2,
+      3,
+      4,
+      5,
+      6,
+      7,
+      8,
+      9,
+      10,
+      11,
+      12,
+      13,
+      14,
+      15
+    ],
+    "device_kinds": [
+      "TPU v4"
+    ],
+    "jax_version": "0.11.0",
+    "jaxlib_version": "0.11.0",
+    "libtpu_version": "0.0.44.1",
+    "local_device_count": 4,
+    "platform": "tpu",
+    "process_count": 4,
+    "process_indices": [
+      0,
+      0,
+      0,
+      0,
+      1,
+      1,
+      1,
+      1,
+      2,
+      2,
+      2,
+      2,
+      3,
+      3,
+      3,
+      3
+    ],
+    "python_version": "3.12.13"
+  },
+  "track": "open"
+}

moe-lr-sweep-8k/60m-5tpp-bs16-lr2e-6-s1338/result.json

@@ -0,0 +1,243 @@
+{
+  "artifacts": {
+    "diagnostics": "diagnostics.riglog",
+    "training_curve": "training.riglog",
+    "validation_curve": "validation.csv"
+  },
+  "checkpoint": null,
+  "contract": {
+    "dataset_id": "fineweb-8b-gpt2",
+    "model": {
+      "d_model": 384,
+      "heads": 6,
+      "layers": 12,
+      "mlp_activation": "gelu",
+      "mlp_mult": 4,
+      "normalization": "rms_norm",
+      "parameterization": "complete_d_p",
+      "position_encoding": "rope_base_10000",
+      "semantic_vocab_size": 50304,
+      "tied_embeddings": false,
+      "tier": "60m",
+      "vocab_size": 50304
+    },
+    "model_id": "reference-gpt-v3-family",
+    "sequence_length": 8192,
+    "tokenizer_id": "gpt2"
+  },
+  "evaluations": {
+    "fineweb": {
+      "canonical": true,
+      "loss": 3.9843336641788483,
+      "perplexity": 53.74946235063435,
+      "scored_tokens": 1048576,
+      "seconds": 0.4521297650062479
+    }
+  },
+  "implementation": {
+    "attention_backend": "tpu_flash",
+    "attention_tuning": {
+      "key_digest": "ec38db3431308f3e9f99acf40c7e3bb85ca0695b0d97491e44a93bc7adfada00",
+      "resolution_source": "heuristic",
+      "tiles": {
+        "block_kv": 512,
+        "block_kv_compute": 256,
+        "block_kv_dkv": 512,
+        "block_kv_dkv_compute": 256,
+        "block_kv_dq": 512,
+        "block_kv_dq_compute": 256,
+        "block_q": 512,
+        "block_q_dkv": 512,
+        "block_q_dkv_compute": 256,
+        "block_q_dq": 256
+      },
+      "tune_seconds": 0.0
+    },
+    "configuration": {
+      "overrides": {
+        "tokens_per_parameter_micros": 5000000
+      },
+      "path": "config.yaml",
+      "profile": "dev",
+      "resolved": {
+        "evaluation": {
+          "eval_batches": 8,
+          "val_every": 0,
+          "val_probe_batches": 8
+        },
+        "kernels": {
+          "attention_backend": "tpu_flash",
+          "loss_backend": "tiled",
+          "vocab_tile_size": 2048
+        },
+        "logging": {
+          "diagnostics_every": 10,
+          "log_every": 10
+        },
+        "model": {
+          "d_model": 384,
+          "heads": 6,
+          "layers": 12,
+          "mlp_activation": "gelu",
+          "mlp_mult": 4,
+          "normalization": "rms_norm",
+          "parameterization": "complete_d_p",
+          "position_encoding": "rope_base_10000",
+          "semantic_vocab_size": 50304,
+          "tied_embeddings": false,
+          "tier": "60m",
+          "vocab_size": 50304
+        },
+        "optimizer": {
+          "adam_epsilon": 1e-08,
+          "beta1": 0.9,
+          "beta2": 0.95,
+          "effective": {
+            "adam_epsilon_horizon_multiplier": 1.0,
+            "beta1": 0.9,
+            "beta2": 0.95,
+            "global_peak_learning_rate": 0.015625,
+            "weight_decay_horizon_multiplier": 1.0
+          },
+          "grad_clip": 0.0,
+          "learning_rate": 0.015625,
+          "min_lr_ratio": 0.1,
+          "warmup_steps": 229,
+          "weight_decay": 0.1
+        },
+        "parameterization": {
+          "attention_scale": "inverse_head_dim",
+          "base_depth": 12,
+          "base_width": 384,
+          "batch_multiplier": 1.0,
+          "data_multiplier": 1.0,
+          "depth_alpha": 1.0,
+          "depth_multiplier": 1.0,
+          "embeddings": "untied",
+          "init_std": 0.02,
+          "name": "complete_d_p",
+          "width_multiplier": 1.0
+        },
+        "training": {
+          "batch_size": 16,
+          "dtype": "bfloat16",
+          "sampling": "shuffled_epochs",
+          "seq_len": 8192,
+          "steps": 2286,
+          "tokens_per_parameter": 5.000660099848113,
+          "train_tokens": 299630592
+        }
+      },
+      "schema_version": 2,
+      "sha256": "a4bad1be29b57d20ec7ae7b0d539f2e777d11f5539607655550b7afca185affe"
+    },
+    "loss_backend": "tiled",
+    "vocab_tile_size": 2048
+  },
+  "metrics": {
+    "achieved_tflops": 597.4534119803859,
+    "attention_tune_seconds": 0.0,
+    "base_learning_rate": 0.015625,
+    "diagnostic_compile_seconds": 85.61127052997472,
+    "diagnostic_point_count": 230,
+    "diagnostics_every": 10,
+    "early_stopping_step": null,
+    "estimated_total_flops": 221095966021779456,
+    "eval_compile_seconds": 15.335515389975626,
+    "expert_top_k": 2,
+    "experts": 8,
+    "final_validation_seconds": 0.4521297650062479,
+    "flop_accounting": {
+      "by_site": {
+        "dot_general": 1638200770560,
+        "pallas_call": 695784701952,
+        "tpu_flash_causal_attention_bwd_dkv": 1236950581248,
+        "tpu_flash_causal_attention_bwd_dq": 1236950581248,
+        "tpu_flash_causal_attention_fwd": 1236950581248
+      },
+      "elementwise_per_sequence": 9518381752,
+      "matmul_per_sequence": 6044837216256,
+      "method": "traced-jaxpr",
+      "warnings": []
+    },
+    "flops_per_token": 737895168,
+    "mfu_estimate": 0.1357848663591786,
+    "model_tier": "60m",
+    "parameter_count": 59959680,
+    "parameters": 102509952,
+    "schedule_steps": 2286,
+    "tokens_per_parameter": 5.000660099848113,
+    "tokens_per_second": 809672.4817967447,
+    "tokens_processed": 299630592,
+    "total_compile_seconds": 146.9452016199357,
+    "total_parameter_count": 102509952,
+    "train_compile_seconds": 45.998415699985344,
+    "train_loss": 3.988403797149658,
+    "train_seconds": 370.06394404696766,
+    "training_data_epochs": 0.037928020133706226,
+    "training_data_sharding": "rank_disjoint_shuffled_windows",
+    "training_sampling": "shuffled_epochs",
+    "training_steps": 2286,
+    "training_token_budget": 299630592,
+    "training_usable_tokens_per_epoch": 7899979776,
+    "validation_loss": 3.9843336641788483,
+    "validation_probe_count": 0,
+    "validation_probe_seconds": 0.0,
+    "validation_tokens": 1048576
+  },
+  "profile": "dev",
+  "schema_version": 1,
+  "seed": 1338,
+  "status": "ok",
+  "system": {
+    "controller_process_index": 1,
+    "device_count": 16,
+    "device_ids": [
+      0,
+      1,
+      2,
+      3,
+      4,
+      5,
+      6,
+      7,
+      8,
+      9,
+      10,
+      11,
+      12,
+      13,
+      14,
+      15
+    ],
+    "device_kinds": [
+      "TPU v4"
+    ],
+    "jax_version": "0.11.0",
+    "jaxlib_version": "0.11.0",
+    "libtpu_version": "0.0.44.1",
+    "local_device_count": 4,
+    "platform": "tpu",
+    "process_count": 4,
+    "process_indices": [
+      0,
+      0,
+      0,
+      0,
+      1,
+      1,
+      1,
+      1,
+      2,
+      2,
+      2,
+      2,
+      3,
+      3,
+      3,
+      3
+    ],
+    "python_version": "3.12.13"
+  },
+  "track": "open"
+}

moe-lr-sweep-8k/60m-5tpp-bs16-lr2e-6-s1338/training.riglog

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:bb07d16d4b22e4ce3bddb3a1c9c09712df7a181a9c184cc6a7faef3e2cbfd90c
+size 1301864

moe-lr-sweep-8k/60m-5tpp-bs16-lr2e-6-s1338/validation.csv

@@ -0,0 +1,2 @@
+step,tokens_processed,kind,domain,validation_tokens,validation_loss,perplexity,validation_seconds,canonical
+2286,299630592,fineweb,fineweb,1048576,3.9843336641788483,53.74946235063435,0.4521297650062479,true

moe-lr-sweep-8k/60m-5tpp-bs16-lr2e-6-s1339/diagnostics.riglog

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:19e82079ffba08762f687bbda71b6446fe5d47cfe60f925b87d1cfb16225bbe8
+size 272824

moe-lr-sweep-8k/60m-5tpp-bs16-lr2e-6-s1339/metrics.json

@@ -0,0 +1,243 @@
+{
+  "artifacts": {
+    "diagnostics": "diagnostics.riglog",
+    "training_curve": "training.riglog",
+    "validation_curve": "validation.csv"
+  },
+  "checkpoint": null,
+  "contract": {
+    "dataset_id": "fineweb-8b-gpt2",
+    "model": {
+      "d_model": 384,
+      "heads": 6,
+      "layers": 12,
+      "mlp_activation": "gelu",
+      "mlp_mult": 4,
+      "normalization": "rms_norm",
+      "parameterization": "complete_d_p",
+      "position_encoding": "rope_base_10000",
+      "semantic_vocab_size": 50304,
+      "tied_embeddings": false,
+      "tier": "60m",
+      "vocab_size": 50304
+    },
+    "model_id": "reference-gpt-v3-family",
+    "sequence_length": 8192,
+    "tokenizer_id": "gpt2"
+  },
+  "evaluations": {
+    "fineweb": {
+      "canonical": true,
+      "loss": 4.05812394618988,
+      "perplexity": 57.8656501105525,
+      "scored_tokens": 1048576,
+      "seconds": 0.45437015598872676
+    }
+  },
+  "implementation": {
+    "attention_backend": "tpu_flash",
+    "attention_tuning": {
+      "key_digest": "ec38db3431308f3e9f99acf40c7e3bb85ca0695b0d97491e44a93bc7adfada00",
+      "resolution_source": "heuristic",
+      "tiles": {
+        "block_kv": 512,
+        "block_kv_compute": 256,
+        "block_kv_dkv": 512,
+        "block_kv_dkv_compute": 256,
+        "block_kv_dq": 512,
+        "block_kv_dq_compute": 256,
+        "block_q": 512,
+        "block_q_dkv": 512,
+        "block_q_dkv_compute": 256,
+        "block_q_dq": 256
+      },
+      "tune_seconds": 0.0
+    },
+    "configuration": {
+      "overrides": {
+        "tokens_per_parameter_micros": 5000000
+      },
+      "path": "config.yaml",
+      "profile": "dev",
+      "resolved": {
+        "evaluation": {
+          "eval_batches": 8,
+          "val_every": 0,
+          "val_probe_batches": 8
+        },
+        "kernels": {
+          "attention_backend": "tpu_flash",
+          "loss_backend": "tiled",
+          "vocab_tile_size": 2048
+        },
+        "logging": {
+          "diagnostics_every": 10,
+          "log_every": 10
+        },
+        "model": {
+          "d_model": 384,
+          "heads": 6,
+          "layers": 12,
+          "mlp_activation": "gelu",
+          "mlp_mult": 4,
+          "normalization": "rms_norm",
+          "parameterization": "complete_d_p",
+          "position_encoding": "rope_base_10000",
+          "semantic_vocab_size": 50304,
+          "tied_embeddings": false,
+          "tier": "60m",
+          "vocab_size": 50304
+        },
+        "optimizer": {
+          "adam_epsilon": 1e-08,
+          "beta1": 0.9,
+          "beta2": 0.95,
+          "effective": {
+            "adam_epsilon_horizon_multiplier": 1.0,
+            "beta1": 0.9,
+            "beta2": 0.95,
+            "global_peak_learning_rate": 0.015625,
+            "weight_decay_horizon_multiplier": 1.0
+          },
+          "grad_clip": 0.0,
+          "learning_rate": 0.015625,
+          "min_lr_ratio": 0.1,
+          "warmup_steps": 229,
+          "weight_decay": 0.1
+        },
+        "parameterization": {
+          "attention_scale": "inverse_head_dim",
+          "base_depth": 12,
+          "base_width": 384,
+          "batch_multiplier": 1.0,
+          "data_multiplier": 1.0,
+          "depth_alpha": 1.0,
+          "depth_multiplier": 1.0,
+          "embeddings": "untied",
+          "init_std": 0.02,
+          "name": "complete_d_p",
+          "width_multiplier": 1.0
+        },
+        "training": {
+          "batch_size": 16,
+          "dtype": "bfloat16",
+          "sampling": "shuffled_epochs",
+          "seq_len": 8192,
+          "steps": 2286,
+          "tokens_per_parameter": 5.000660099848113,
+          "train_tokens": 299630592
+        }
+      },
+      "schema_version": 2,
+      "sha256": "a4bad1be29b57d20ec7ae7b0d539f2e777d11f5539607655550b7afca185affe"
+    },
+    "loss_backend": "tiled",
+    "vocab_tile_size": 2048
+  },
+  "metrics": {
+    "achieved_tflops": 597.3680857773434,
+    "attention_tune_seconds": 0.0,
+    "base_learning_rate": 0.015625,
+    "diagnostic_compile_seconds": 78.2013416410191,
+    "diagnostic_point_count": 230,
+    "diagnostics_every": 10,
+    "early_stopping_step": null,
+    "estimated_total_flops": 221095966021779456,
+    "eval_compile_seconds": 14.844683077011723,
+    "expert_top_k": 2,
+    "experts": 8,
+    "final_validation_seconds": 0.45437015598872676,
+    "flop_accounting": {
+      "by_site": {
+        "dot_general": 1638200770560,
+        "pallas_call": 695784701952,
+        "tpu_flash_causal_attention_bwd_dkv": 1236950581248,
+        "tpu_flash_causal_attention_bwd_dq": 1236950581248,
+        "tpu_flash_causal_attention_fwd": 1236950581248
+      },
+      "elementwise_per_sequence": 9518381752,
+      "matmul_per_sequence": 6044837216256,
+      "method": "traced-jaxpr",
+      "warnings": []
+    },
+    "flops_per_token": 737895168,
+    "mfu_estimate": 0.1357654740403053,
+    "model_tier": "60m",
+    "parameter_count": 59959680,
+    "parameters": 102509952,
+    "schedule_steps": 2286,
+    "tokens_per_parameter": 5.000660099848113,
+    "tokens_per_second": 809556.8472096884,
+    "tokens_processed": 299630592,
+    "total_compile_seconds": 138.09480160003295,
+    "total_parameter_count": 102509952,
+    "train_compile_seconds": 45.04877688200213,
+    "train_loss": 3.9488604068756104,
+    "train_seconds": 370.1168028319953,
+    "training_data_epochs": 0.037928020133706226,
+    "training_data_sharding": "rank_disjoint_shuffled_windows",
+    "training_sampling": "shuffled_epochs",
+    "training_steps": 2286,
+    "training_token_budget": 299630592,
+    "training_usable_tokens_per_epoch": 7899979776,
+    "validation_loss": 4.05812394618988,
+    "validation_probe_count": 0,
+    "validation_probe_seconds": 0.0,
+    "validation_tokens": 1048576
+  },
+  "profile": "dev",
+  "schema_version": 1,
+  "seed": 1339,
+  "status": "ok",
+  "system": {
+    "controller_process_index": 1,
+    "device_count": 16,
+    "device_ids": [
+      0,
+      1,
+      2,
+      3,
+      4,
+      5,
+      6,
+      7,
+      8,
+      9,
+      10,
+      11,
+      12,
+      13,
+      14,
+      15
+    ],
+    "device_kinds": [
+      "TPU v4"
+    ],
+    "jax_version": "0.11.0",
+    "jaxlib_version": "0.11.0",
+    "libtpu_version": "0.0.44.1",
+    "local_device_count": 4,
+    "platform": "tpu",
+    "process_count": 4,
+    "process_indices": [
+      0,
+      0,
+      0,
+      0,
+      1,
+      1,
+      1,
+      1,
+      2,
+      2,
+      2,
+      2,
+      3,
+      3,
+      3,
+      3
+    ],
+    "python_version": "3.12.13"
+  },
+  "track": "open"
+}

moe-lr-sweep-8k/60m-5tpp-bs16-lr2e-6-s1339/result.json

@@ -0,0 +1,243 @@
+{
+  "artifacts": {
+    "diagnostics": "diagnostics.riglog",
+    "training_curve": "training.riglog",
+    "validation_curve": "validation.csv"
+  },
+  "checkpoint": null,
+  "contract": {
+    "dataset_id": "fineweb-8b-gpt2",
+    "model": {
+      "d_model": 384,
+      "heads": 6,
+      "layers": 12,
+      "mlp_activation": "gelu",
+      "mlp_mult": 4,
+      "normalization": "rms_norm",
+      "parameterization": "complete_d_p",
+      "position_encoding": "rope_base_10000",
+      "semantic_vocab_size": 50304,
+      "tied_embeddings": false,
+      "tier": "60m",
+      "vocab_size": 50304
+    },
+    "model_id": "reference-gpt-v3-family",
+    "sequence_length": 8192,
+    "tokenizer_id": "gpt2"
+  },
+  "evaluations": {
+    "fineweb": {
+      "canonical": true,
+      "loss": 4.05812394618988,
+      "perplexity": 57.8656501105525,
+      "scored_tokens": 1048576,
+      "seconds": 0.45437015598872676
+    }
+  },
+  "implementation": {
+    "attention_backend": "tpu_flash",
+    "attention_tuning": {
+      "key_digest": "ec38db3431308f3e9f99acf40c7e3bb85ca0695b0d97491e44a93bc7adfada00",
+      "resolution_source": "heuristic",
+      "tiles": {
+        "block_kv": 512,
+        "block_kv_compute": 256,
+        "block_kv_dkv": 512,
+        "block_kv_dkv_compute": 256,
+        "block_kv_dq": 512,
+        "block_kv_dq_compute": 256,
+        "block_q": 512,
+        "block_q_dkv": 512,
+        "block_q_dkv_compute": 256,
+        "block_q_dq": 256
+      },
+      "tune_seconds": 0.0
+    },
+    "configuration": {
+      "overrides": {
+        "tokens_per_parameter_micros": 5000000
+      },
+      "path": "config.yaml",
+      "profile": "dev",
+      "resolved": {
+        "evaluation": {
+          "eval_batches": 8,
+          "val_every": 0,
+          "val_probe_batches": 8
+        },
+        "kernels": {
+          "attention_backend": "tpu_flash",
+          "loss_backend": "tiled",
+          "vocab_tile_size": 2048
+        },
+        "logging": {
+          "diagnostics_every": 10,
+          "log_every": 10
+        },
+        "model": {
+          "d_model": 384,
+          "heads": 6,
+          "layers": 12,
+          "mlp_activation": "gelu",
+          "mlp_mult": 4,
+          "normalization": "rms_norm",
+          "parameterization": "complete_d_p",
+          "position_encoding": "rope_base_10000",
+          "semantic_vocab_size": 50304,
+          "tied_embeddings": false,
+          "tier": "60m",
+          "vocab_size": 50304
+        },
+        "optimizer": {
+          "adam_epsilon": 1e-08,
+          "beta1": 0.9,
+          "beta2": 0.95,
+          "effective": {
+            "adam_epsilon_horizon_multiplier": 1.0,
+            "beta1": 0.9,
+            "beta2": 0.95,
+            "global_peak_learning_rate": 0.015625,
+            "weight_decay_horizon_multiplier": 1.0
+          },
+          "grad_clip": 0.0,
+          "learning_rate": 0.015625,
+          "min_lr_ratio": 0.1,
+          "warmup_steps": 229,
+          "weight_decay": 0.1
+        },
+        "parameterization": {
+          "attention_scale": "inverse_head_dim",
+          "base_depth": 12,
+          "base_width": 384,
+          "batch_multiplier": 1.0,
+          "data_multiplier": 1.0,
+          "depth_alpha": 1.0,
+          "depth_multiplier": 1.0,
+          "embeddings": "untied",
+          "init_std": 0.02,
+          "name": "complete_d_p",
+          "width_multiplier": 1.0
+        },
+        "training": {
+          "batch_size": 16,
+          "dtype": "bfloat16",
+          "sampling": "shuffled_epochs",
+          "seq_len": 8192,
+          "steps": 2286,
+          "tokens_per_parameter": 5.000660099848113,
+          "train_tokens": 299630592
+        }
+      },
+      "schema_version": 2,
+      "sha256": "a4bad1be29b57d20ec7ae7b0d539f2e777d11f5539607655550b7afca185affe"
+    },
+    "loss_backend": "tiled",
+    "vocab_tile_size": 2048
+  },
+  "metrics": {
+    "achieved_tflops": 597.3680857773434,
+    "attention_tune_seconds": 0.0,
+    "base_learning_rate": 0.015625,
+    "diagnostic_compile_seconds": 78.2013416410191,
+    "diagnostic_point_count": 230,
+    "diagnostics_every": 10,
+    "early_stopping_step": null,
+    "estimated_total_flops": 221095966021779456,
+    "eval_compile_seconds": 14.844683077011723,
+    "expert_top_k": 2,
+    "experts": 8,
+    "final_validation_seconds": 0.45437015598872676,
+    "flop_accounting": {
+      "by_site": {
+        "dot_general": 1638200770560,
+        "pallas_call": 695784701952,
+        "tpu_flash_causal_attention_bwd_dkv": 1236950581248,
+        "tpu_flash_causal_attention_bwd_dq": 1236950581248,
+        "tpu_flash_causal_attention_fwd": 1236950581248
+      },
+      "elementwise_per_sequence": 9518381752,
+      "matmul_per_sequence": 6044837216256,
+      "method": "traced-jaxpr",
+      "warnings": []
+    },
+    "flops_per_token": 737895168,
+    "mfu_estimate": 0.1357654740403053,
+    "model_tier": "60m",
+    "parameter_count": 59959680,
+    "parameters": 102509952,
+    "schedule_steps": 2286,
+    "tokens_per_parameter": 5.000660099848113,
+    "tokens_per_second": 809556.8472096884,
+    "tokens_processed": 299630592,
+    "total_compile_seconds": 138.09480160003295,
+    "total_parameter_count": 102509952,
+    "train_compile_seconds": 45.04877688200213,
+    "train_loss": 3.9488604068756104,
+    "train_seconds": 370.1168028319953,
+    "training_data_epochs": 0.037928020133706226,
+    "training_data_sharding": "rank_disjoint_shuffled_windows",
+    "training_sampling": "shuffled_epochs",
+    "training_steps": 2286,
+    "training_token_budget": 299630592,
+    "training_usable_tokens_per_epoch": 7899979776,
+    "validation_loss": 4.05812394618988,
+    "validation_probe_count": 0,
+    "validation_probe_seconds": 0.0,
+    "validation_tokens": 1048576
+  },
+  "profile": "dev",
+  "schema_version": 1,
+  "seed": 1339,
+  "status": "ok",
+  "system": {
+    "controller_process_index": 1,
+    "device_count": 16,
+    "device_ids": [
+      0,
+      1,
+      2,
+      3,
+      4,
+      5,
+      6,
+      7,
+      8,
+      9,
+      10,
+      11,
+      12,
+      13,
+      14,
+      15
+    ],
+    "device_kinds": [
+      "TPU v4"
+    ],
+    "jax_version": "0.11.0",
+    "jaxlib_version": "0.11.0",
+    "libtpu_version": "0.0.44.1",
+    "local_device_count": 4,
+    "platform": "tpu",
+    "process_count": 4,
+    "process_indices": [
+      0,
+      0,
+      0,
+      0,
+      1,
+      1,
+      1,
+      1,
+      2,
+      2,
+      2,
+      2,
+      3,
+      3,
+      3,
+      3
+    ],
+    "python_version": "3.12.13"
+  },
+  "track": "open"
+}

moe-lr-sweep-8k/60m-5tpp-bs16-lr2e-6-s1339/training.riglog

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:fc8405d8ccdca80dcd167a63ea3a40130931336f42d620213e485a20f26f61fb
+size 1301864

moe-lr-sweep-8k/60m-5tpp-bs16-lr2e-6-s1339/validation.csv

@@ -0,0 +1,2 @@
+step,tokens_processed,kind,domain,validation_tokens,validation_loss,perplexity,validation_seconds,canonical
+2286,299630592,fineweb,fineweb,1048576,4.05812394618988,57.8656501105525,0.45437015598872676,true

moe-lr-sweep-8k/60m-5tpp-bs16-lr2e-7-s1337/diagnostics.riglog

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:6cea041ff2911a0e2193389e333696be9fc3c4619b3a5ea3aceca5579781d083
+size 272824

moe-lr-sweep-8k/60m-5tpp-bs16-lr2e-7-s1337/metrics.json

@@ -0,0 +1,243 @@
+{
+  "artifacts": {
+    "diagnostics": "diagnostics.riglog",
+    "training_curve": "training.riglog",
+    "validation_curve": "validation.csv"
+  },
+  "checkpoint": null,
+  "contract": {
+    "dataset_id": "fineweb-8b-gpt2",
+    "model": {
+      "d_model": 384,
+      "heads": 6,
+      "layers": 12,
+      "mlp_activation": "gelu",
+      "mlp_mult": 4,
+      "normalization": "rms_norm",
+      "parameterization": "complete_d_p",
+      "position_encoding": "rope_base_10000",
+      "semantic_vocab_size": 50304,
+      "tied_embeddings": false,
+      "tier": "60m",
+      "vocab_size": 50304
+    },
+    "model_id": "reference-gpt-v3-family",
+    "sequence_length": 8192,
+    "tokenizer_id": "gpt2"
+  },
+  "evaluations": {
+    "fineweb": {
+      "canonical": true,
+      "loss": 3.962973475456238,
+      "perplexity": 52.61353865253143,
+      "scored_tokens": 1048576,
+      "seconds": 0.449693754024338
+    }
+  },
+  "implementation": {
+    "attention_backend": "tpu_flash",
+    "attention_tuning": {
+      "key_digest": "ec38db3431308f3e9f99acf40c7e3bb85ca0695b0d97491e44a93bc7adfada00",
+      "resolution_source": "heuristic",
+      "tiles": {
+        "block_kv": 512,
+        "block_kv_compute": 256,
+        "block_kv_dkv": 512,
+        "block_kv_dkv_compute": 256,
+        "block_kv_dq": 512,
+        "block_kv_dq_compute": 256,
+        "block_q": 512,
+        "block_q_dkv": 512,
+        "block_q_dkv_compute": 256,
+        "block_q_dq": 256
+      },
+      "tune_seconds": 0.0
+    },
+    "configuration": {
+      "overrides": {
+        "tokens_per_parameter_micros": 5000000
+      },
+      "path": "config.yaml",
+      "profile": "dev",
+      "resolved": {
+        "evaluation": {
+          "eval_batches": 8,
+          "val_every": 0,
+          "val_probe_batches": 8
+        },
+        "kernels": {
+          "attention_backend": "tpu_flash",
+          "loss_backend": "tiled",
+          "vocab_tile_size": 2048
+        },
+        "logging": {
+          "diagnostics_every": 10,
+          "log_every": 10
+        },
+        "model": {
+          "d_model": 384,
+          "heads": 6,
+          "layers": 12,
+          "mlp_activation": "gelu",
+          "mlp_mult": 4,
+          "normalization": "rms_norm",
+          "parameterization": "complete_d_p",
+          "position_encoding": "rope_base_10000",
+          "semantic_vocab_size": 50304,
+          "tied_embeddings": false,
+          "tier": "60m",
+          "vocab_size": 50304
+        },
+        "optimizer": {
+          "adam_epsilon": 1e-08,
+          "beta1": 0.9,
+          "beta2": 0.95,
+          "effective": {
+            "adam_epsilon_horizon_multiplier": 1.0,
+            "beta1": 0.9,
+            "beta2": 0.95,
+            "global_peak_learning_rate": 0.0078125,
+            "weight_decay_horizon_multiplier": 1.0
+          },
+          "grad_clip": 0.0,
+          "learning_rate": 0.0078125,
+          "min_lr_ratio": 0.1,
+          "warmup_steps": 229,
+          "weight_decay": 0.1
+        },
+        "parameterization": {
+          "attention_scale": "inverse_head_dim",
+          "base_depth": 12,
+          "base_width": 384,
+          "batch_multiplier": 1.0,
+          "data_multiplier": 1.0,
+          "depth_alpha": 1.0,
+          "depth_multiplier": 1.0,
+          "embeddings": "untied",
+          "init_std": 0.02,
+          "name": "complete_d_p",
+          "width_multiplier": 1.0
+        },
+        "training": {
+          "batch_size": 16,
+          "dtype": "bfloat16",
+          "sampling": "shuffled_epochs",
+          "seq_len": 8192,
+          "steps": 2286,
+          "tokens_per_parameter": 5.000660099848113,
+          "train_tokens": 299630592
+        }
+      },
+      "schema_version": 2,
+      "sha256": "a4bad1be29b57d20ec7ae7b0d539f2e777d11f5539607655550b7afca185affe"
+    },
+    "loss_backend": "tiled",
+    "vocab_tile_size": 2048
+  },
+  "metrics": {
+    "achieved_tflops": 597.4925117457461,
+    "attention_tune_seconds": 0.0,
+    "base_learning_rate": 0.0078125,
+    "diagnostic_compile_seconds": 81.99022210703697,
+    "diagnostic_point_count": 230,
+    "diagnostics_every": 10,
+    "early_stopping_step": null,
+    "estimated_total_flops": 221095966021779456,
+    "eval_compile_seconds": 15.240791740012355,
+    "expert_top_k": 2,
+    "experts": 8,
+    "final_validation_seconds": 0.449693754024338,
+    "flop_accounting": {
+      "by_site": {
+        "dot_general": 1638200770560,
+        "pallas_call": 695784701952,
+        "tpu_flash_causal_attention_bwd_dkv": 1236950581248,
+        "tpu_flash_causal_attention_bwd_dq": 1236950581248,
+        "tpu_flash_causal_attention_fwd": 1236950581248
+      },
+      "elementwise_per_sequence": 9518381752,
+      "matmul_per_sequence": 6044837216256,
+      "method": "traced-jaxpr",
+      "warnings": []
+    },
+    "flops_per_token": 737895168,
+    "mfu_estimate": 0.13579375266948776,
+    "model_tier": "60m",
+    "parameter_count": 59959680,
+    "parameters": 102509952,
+    "schedule_steps": 2286,
+    "tokens_per_parameter": 5.000660099848113,
+    "tokens_per_second": 809725.4700355296,
+    "tokens_processed": 299630592,
+    "total_compile_seconds": 140.92854372307193,
+    "total_parameter_count": 102509952,
+    "train_compile_seconds": 43.6975298760226,
+    "train_loss": 4.045482635498047,
+    "train_seconds": 370.0397271520342,
+    "training_data_epochs": 0.037928020133706226,
+    "training_data_sharding": "rank_disjoint_shuffled_windows",
+    "training_sampling": "shuffled_epochs",
+    "training_steps": 2286,
+    "training_token_budget": 299630592,
+    "training_usable_tokens_per_epoch": 7899979776,
+    "validation_loss": 3.962973475456238,
+    "validation_probe_count": 0,
+    "validation_probe_seconds": 0.0,
+    "validation_tokens": 1048576
+  },
+  "profile": "dev",
+  "schema_version": 1,
+  "seed": 1337,
+  "status": "ok",
+  "system": {
+    "controller_process_index": 1,
+    "device_count": 16,
+    "device_ids": [
+      0,
+      1,
+      2,
+      3,
+      4,
+      5,
+      6,
+      7,
+      8,
+      9,
+      10,
+      11,
+      12,
+      13,
+      14,
+      15
+    ],
+    "device_kinds": [
+      "TPU v4"
+    ],
+    "jax_version": "0.11.0",
+    "jaxlib_version": "0.11.0",
+    "libtpu_version": "0.0.44.1",
+    "local_device_count": 4,
+    "platform": "tpu",
+    "process_count": 4,
+    "process_indices": [
+      0,
+      0,
+      0,
+      0,
+      1,
+      1,
+      1,
+      1,
+      2,
+      2,
+      2,
+      2,
+      3,
+      3,
+      3,
+      3
+    ],
+    "python_version": "3.12.13"
+  },
+  "track": "open"
+}

moe-lr-sweep-8k/60m-5tpp-bs16-lr2e-7-s1337/result.json

@@ -0,0 +1,243 @@
+{
+  "artifacts": {
+    "diagnostics": "diagnostics.riglog",
+    "training_curve": "training.riglog",
+    "validation_curve": "validation.csv"
+  },
+  "checkpoint": null,
+  "contract": {
+    "dataset_id": "fineweb-8b-gpt2",
+    "model": {
+      "d_model": 384,
+      "heads": 6,
+      "layers": 12,
+      "mlp_activation": "gelu",
+      "mlp_mult": 4,
+      "normalization": "rms_norm",
+      "parameterization": "complete_d_p",
+      "position_encoding": "rope_base_10000",
+      "semantic_vocab_size": 50304,
+      "tied_embeddings": false,
+      "tier": "60m",
+      "vocab_size": 50304
+    },
+    "model_id": "reference-gpt-v3-family",
+    "sequence_length": 8192,
+    "tokenizer_id": "gpt2"
+  },
+  "evaluations": {
+    "fineweb": {
+      "canonical": true,
+      "loss": 3.962973475456238,
+      "perplexity": 52.61353865253143,
+      "scored_tokens": 1048576,
+      "seconds": 0.449693754024338
+    }
+  },
+  "implementation": {
+    "attention_backend": "tpu_flash",
+    "attention_tuning": {
+      "key_digest": "ec38db3431308f3e9f99acf40c7e3bb85ca0695b0d97491e44a93bc7adfada00",
+      "resolution_source": "heuristic",
+      "tiles": {
+        "block_kv": 512,
+        "block_kv_compute": 256,
+        "block_kv_dkv": 512,
+        "block_kv_dkv_compute": 256,
+        "block_kv_dq": 512,
+        "block_kv_dq_compute": 256,
+        "block_q": 512,
+        "block_q_dkv": 512,
+        "block_q_dkv_compute": 256,
+        "block_q_dq": 256
+      },
+      "tune_seconds": 0.0
+    },
+    "configuration": {
+      "overrides": {
+        "tokens_per_parameter_micros": 5000000
+      },
+      "path": "config.yaml",
+      "profile": "dev",
+      "resolved": {
+        "evaluation": {
+          "eval_batches": 8,
+          "val_every": 0,
+          "val_probe_batches": 8
+        },
+        "kernels": {
+          "attention_backend": "tpu_flash",
+          "loss_backend": "tiled",
+          "vocab_tile_size": 2048
+        },
+        "logging": {
+          "diagnostics_every": 10,
+          "log_every": 10
+        },
+        "model": {
+          "d_model": 384,
+          "heads": 6,
+          "layers": 12,
+          "mlp_activation": "gelu",
+          "mlp_mult": 4,
+          "normalization": "rms_norm",
+          "parameterization": "complete_d_p",
+          "position_encoding": "rope_base_10000",
+          "semantic_vocab_size": 50304,
+          "tied_embeddings": false,
+          "tier": "60m",
+          "vocab_size": 50304
+        },
+        "optimizer": {
+          "adam_epsilon": 1e-08,
+          "beta1": 0.9,
+          "beta2": 0.95,
+          "effective": {
+            "adam_epsilon_horizon_multiplier": 1.0,
+            "beta1": 0.9,
+            "beta2": 0.95,
+            "global_peak_learning_rate": 0.0078125,
+            "weight_decay_horizon_multiplier": 1.0
+          },
+          "grad_clip": 0.0,
+          "learning_rate": 0.0078125,
+          "min_lr_ratio": 0.1,
+          "warmup_steps": 229,
+          "weight_decay": 0.1
+        },
+        "parameterization": {
+          "attention_scale": "inverse_head_dim",
+          "base_depth": 12,
+          "base_width": 384,
+          "batch_multiplier": 1.0,
+          "data_multiplier": 1.0,
+          "depth_alpha": 1.0,
+          "depth_multiplier": 1.0,
+          "embeddings": "untied",
+          "init_std": 0.02,
+          "name": "complete_d_p",
+          "width_multiplier": 1.0
+        },
+        "training": {
+          "batch_size": 16,
+          "dtype": "bfloat16",
+          "sampling": "shuffled_epochs",
+          "seq_len": 8192,
+          "steps": 2286,
+          "tokens_per_parameter": 5.000660099848113,
+          "train_tokens": 299630592
+        }
+      },
+      "schema_version": 2,
+      "sha256": "a4bad1be29b57d20ec7ae7b0d539f2e777d11f5539607655550b7afca185affe"
+    },
+    "loss_backend": "tiled",
+    "vocab_tile_size": 2048
+  },
+  "metrics": {
+    "achieved_tflops": 597.4925117457461,
+    "attention_tune_seconds": 0.0,
+    "base_learning_rate": 0.0078125,
+    "diagnostic_compile_seconds": 81.99022210703697,
+    "diagnostic_point_count": 230,
+    "diagnostics_every": 10,
+    "early_stopping_step": null,
+    "estimated_total_flops": 221095966021779456,
+    "eval_compile_seconds": 15.240791740012355,
+    "expert_top_k": 2,
+    "experts": 8,
+    "final_validation_seconds": 0.449693754024338,
+    "flop_accounting": {
+      "by_site": {
+        "dot_general": 1638200770560,
+        "pallas_call": 695784701952,
+        "tpu_flash_causal_attention_bwd_dkv": 1236950581248,
+        "tpu_flash_causal_attention_bwd_dq": 1236950581248,
+        "tpu_flash_causal_attention_fwd": 1236950581248
+      },
+      "elementwise_per_sequence": 9518381752,
+      "matmul_per_sequence": 6044837216256,
+      "method": "traced-jaxpr",
+      "warnings": []
+    },
+    "flops_per_token": 737895168,
+    "mfu_estimate": 0.13579375266948776,
+    "model_tier": "60m",
+    "parameter_count": 59959680,
+    "parameters": 102509952,
+    "schedule_steps": 2286,
+    "tokens_per_parameter": 5.000660099848113,
+    "tokens_per_second": 809725.4700355296,
+    "tokens_processed": 299630592,
+    "total_compile_seconds": 140.92854372307193,
+    "total_parameter_count": 102509952,
+    "train_compile_seconds": 43.6975298760226,
+    "train_loss": 4.045482635498047,
+    "train_seconds": 370.0397271520342,
+    "training_data_epochs": 0.037928020133706226,
+    "training_data_sharding": "rank_disjoint_shuffled_windows",
+    "training_sampling": "shuffled_epochs",
+    "training_steps": 2286,
+    "training_token_budget": 299630592,
+    "training_usable_tokens_per_epoch": 7899979776,
+    "validation_loss": 3.962973475456238,
+    "validation_probe_count": 0,
+    "validation_probe_seconds": 0.0,
+    "validation_tokens": 1048576
+  },
+  "profile": "dev",
+  "schema_version": 1,
+  "seed": 1337,
+  "status": "ok",
+  "system": {
+    "controller_process_index": 1,
+    "device_count": 16,
+    "device_ids": [
+      0,
+      1,
+      2,
+      3,
+      4,
+      5,
+      6,
+      7,
+      8,
+      9,
+      10,
+      11,
+      12,
+      13,
+      14,
+      15
+    ],
+    "device_kinds": [
+      "TPU v4"
+    ],
+    "jax_version": "0.11.0",
+    "jaxlib_version": "0.11.0",
+    "libtpu_version": "0.0.44.1",
+    "local_device_count": 4,
+    "platform": "tpu",
+    "process_count": 4,
+    "process_indices": [
+      0,
+      0,
+      0,
+      0,
+      1,
+      1,
+      1,
+      1,
+      2,
+      2,
+      2,
+      2,
+      3,
+      3,
+      3,
+      3
+    ],
+    "python_version": "3.12.13"
+  },
+  "track": "open"
+}

moe-lr-sweep-8k/60m-5tpp-bs16-lr2e-7-s1337/training.riglog

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:b468ab569cc5ca9a1d1d2e9a39606dadca83d6bde95044f64907ddbdbedd2e3b
+size 1301864