{
 "preset": "mini-beatrix-2s-control-attn-fp16-r10k",
 "model_config": {
  "name": "mini-beatrix-2s-control-attn-fp16-r10k",
  "d_model": 1024,
  "n_layers": 20,
  "n_heads": 16,
  "context": 4096,
  "vocab_size": 256,
  "tokenizer": "byte-trigram",
  "hub_layers": [],
  "hub_K": 64,
  "hub_D": 128,
  "tau": 0.1,
  "bank_experts": 3,
  "bank_ff": 1024,
  "head_K": 256,
  "head_D": 256,
  "gate_init": -3.0,
  "tie_embeddings": false,
  "hub_chunk": 256,
  "hub_const": 4,
  "hub_ckpt": 0,
  "fusion": null,
  "qk_norm": "",
  "attn_fp32": true,
  "attn_kernel": "fp16"
 },
 "created_utc": "2026-10-10T03:52:07Z",
 "updated_utc": "2026-10-11T05:08:16Z",
 "steps": 36000,
 "tokens_seen": 9437184000,
 "wall_hours": 22.372104821006456,
 "phases": [
  {
   "name": "warmup_wikitext",
   "dataset": "wikitext-103",
   "planned_tokens": 300000000,
   "status": "done",
   "tokens_done": 300154880
  },
  {
   "name": "fineweb_main",
   "dataset": "fineweb-edu",
   "planned_tokens": 5000000000,
   "status": "done",
   "tokens_done": 5000134656
  },
  {
   "name": "curriculum_s0",
   "dataset": "curriculum-s0",
   "planned_tokens": 700000000,
   "status": "done",
   "tokens_done": 700186624
  },
  {
   "name": "curriculum_s1",
   "dataset": "curriculum-s1",
   "planned_tokens": 700000000,
   "status": "done",
   "tokens_done": 700186624
  },
  {
   "name": "curriculum_s2",
   "dataset": "curriculum-s2",
   "planned_tokens": 1000000000,
   "status": "done",
   "tokens_done": 1000079360
  },
  {
   "name": "curriculum_s3",
   "dataset": "curriculum-s3",
   "planned_tokens": 1200000000,
   "status": "done",
   "tokens_done": 1200095232
  },
  {
   "name": "curriculum_s4",
   "dataset": "curriculum-s4",
   "planned_tokens": 1200000000,
   "status": "active",
   "tokens_done": 536346624
  },
  {
   "name": "curriculum_s5",
   "dataset": "curriculum-s5",
   "planned_tokens": 1000000000,
   "status": "planned",
   "tokens_done": 0
  },
  {
   "name": "curriculum_s6",
   "dataset": "curriculum-s6",
   "planned_tokens": 800000000,
   "status": "planned",
   "tokens_done": 0
  },
  {
   "name": "curriculum_s7",
   "dataset": "curriculum-s7",
   "planned_tokens": 1400000000,
   "status": "planned",
   "tokens_done": 0
  },
  {
   "name": "curriculum_s8",
   "dataset": "curriculum-s8",
   "planned_tokens": 800000000,
   "status": "planned",
   "tokens_done": 0
  },
  {
   "name": "anneal_nochat",
   "dataset": "anneal-nochat",
   "planned_tokens": 1000000000,
   "status": "planned",
   "tokens_done": 0
  },
  {
   "name": "anneal_mix",
   "dataset": "anneal-mix",
   "planned_tokens": 1000000000,
   "status": "planned",
   "tokens_done": 0
  }
 ],
 "checkpoints": [
  {
   "step": 12000,
   "tokens": 3145728000,
   "kind": "safetensors",
   "path": "checkpoints/step_00012000.safetensors",
   "val_bpb": 1.0852089576026334,
   "utc": "2026-10-10T05:35:32Z"
  },
  {
   "step": 12000,
   "tokens": 3145728000,
   "kind": "resume",
   "path": "resume/latest.pt",
   "val_bpb": null,
   "utc": "2026-10-10T05:35:32Z"
  },
  {
   "step": 13778,
   "tokens": 3611820032,
   "kind": "safetensors",
   "path": "checkpoints/step_00013778.safetensors",
   "val_bpb": 1.0852089576026334,
   "utc": "2026-10-10T07:07:27Z"
  },
  {
   "step": 13778,
   "tokens": 3611820032,
   "kind": "fp8",
   "path": "checkpoints/fp8/step_00013778.safetensors",
   "val_bpb": null,
   "utc": "2026-10-10T07:07:32Z"
  },
  {
   "step": 13778,
   "tokens": 3611820032,
   "kind": "resume",
   "path": "resume/latest.pt",
   "val_bpb": null,
   "utc": "2026-10-10T07:07:32Z"
  },
  {
   "step": 14000,
   "tokens": 3670016000,
   "kind": "safetensors",
   "path": "checkpoints/step_00014000.safetensors",
   "val_bpb": 1.0706639871092856,
   "utc": "2026-10-10T07:23:58Z"
  },
  {
   "step": 14000,
   "tokens": 3670016000,
   "kind": "resume",
   "path": "resume/latest.pt",
   "val_bpb": null,
   "utc": "2026-10-10T07:23:58Z"
  },
  {
   "step": 16000,
   "tokens": 4194304000,
   "kind": "safetensors",
   "path": "checkpoints/step_00016000.safetensors",
   "val_bpb": 1.0579480198611166,
   "utc": "2026-10-10T09:09:13Z"
  },
  {
   "step": 16000,
   "tokens": 4194304000,
   "kind": "resume",
   "path": "resume/latest.pt",
   "val_bpb": null,
   "utc": "2026-10-10T09:09:13Z"
  },
  {
   "step": 18000,
   "tokens": 4718592000,
   "kind": "safetensors",
   "path": "checkpoints/step_00018000.safetensors",
   "val_bpb": 1.0496593267516992,
   "utc": "2026-10-10T10:54:29Z"
  },
  {
   "step": 18000,
   "tokens": 4718592000,
   "kind": "resume",
   "path": "resume/latest.pt",
   "val_bpb": null,
   "utc": "2026-10-10T10:54:29Z"
  },
  {
   "step": 20000,
   "tokens": 5242880000,
   "kind": "safetensors",
   "path": "checkpoints/step_00020000.safetensors",
   "val_bpb": 1.0453013938679925,
   "utc": "2026-10-10T12:39:51Z"
  },
  {
   "step": 20000,
   "tokens": 5242880000,
   "kind": "resume",
   "path": "resume/latest.pt",
   "val_bpb": null,
   "utc": "2026-10-10T12:39:51Z"
  },
  {
   "step": 20219,
   "tokens": 5300289536,
   "kind": "safetensors",
   "path": "checkpoints/step_00020219.safetensors",
   "val_bpb": 1.0453013938679925,
   "utc": "2026-10-10T12:51:55Z"
  },
  {
   "step": 20219,
   "tokens": 5300289536,
   "kind": "fp8",
   "path": "checkpoints/fp8/step_00020219.safetensors",
   "val_bpb": null,
   "utc": "2026-10-10T12:52:01Z"
  },
  {
   "step": 20219,
   "tokens": 5300289536,
   "kind": "resume",
   "path": "resume/latest.pt",
   "val_bpb": null,
   "utc": "2026-10-10T12:52:01Z"
  },
  {
   "step": 20219,
   "tokens": 5300289536,
   "kind": "safetensors",
   "path": "checkpoints/step_00020219.safetensors",
   "val_bpb": 1.0453013938679925,
   "utc": "2026-10-10T12:52:35Z"
  },
  {
   "step": 20219,
   "tokens": 5300289536,
   "kind": "fp8",
   "path": "checkpoints/fp8/step_00020219.safetensors",
   "val_bpb": null,
   "utc": "2026-10-10T12:52:37Z"
  },
  {
   "step": 20219,
   "tokens": 5300289536,
   "kind": "resume",
   "path": "resume/latest.pt",
   "val_bpb": null,
   "utc": "2026-10-10T12:52:37Z"
  },
  {
   "step": 22000,
   "tokens": 5767168000,
   "kind": "safetensors",
   "path": "checkpoints/step_00022000.safetensors",
   "val_bpb": 1.1479324062335066,
   "utc": "2026-10-10T14:27:06Z"
  },
  {
   "step": 22000,
   "tokens": 5767168000,
   "kind": "resume",
   "path": "resume/latest.pt",
   "val_bpb": null,
   "utc": "2026-10-10T14:27:06Z"
  },
  {
   "step": 22890,
   "tokens": 6000476160,
   "kind": "safetensors",
   "path": "checkpoints/step_00022890.safetensors",
   "val_bpb": 1.1479324062335066,
   "utc": "2026-10-10T15:14:21Z"
  },
  {
   "step": 22890,
   "tokens": 6000476160,
   "kind": "resume",
   "path": "resume/latest.pt",
   "val_bpb": null,
   "utc": "2026-10-10T15:14:21Z"
  },
  {
   "step": 22890,
   "tokens": 6000476160,
   "kind": "safetensors",
   "path": "checkpoints/step_00022890.safetensors",
   "val_bpb": 1.1479324062335066,
   "utc": "2026-10-10T15:15:04Z"
  },
  {
   "step": 22890,
   "tokens": 6000476160,
   "kind": "fp8",
   "path": "checkpoints/fp8/step_00022890.safetensors",
   "val_bpb": null,
   "utc": "2026-10-10T15:15:09Z"
  },
  {
   "step": 22890,
   "tokens": 6000476160,
   "kind": "resume",
   "path": "resume/latest.pt",
   "val_bpb": null,
   "utc": "2026-10-10T15:15:09Z"
  },
  {
   "step": 24000,
   "tokens": 6291456000,
   "kind": "safetensors",
   "path": "checkpoints/step_00024000.safetensors",
   "val_bpb": 1.1565509513183632,
   "utc": "2026-10-10T16:14:45Z"
  },
  {
   "step": 24000,
   "tokens": 6291456000,
   "kind": "fp8",
   "path": "checkpoints/fp8/step_00024000.safetensors",
   "val_bpb": null,
   "utc": "2026-10-10T16:14:51Z"
  },
  {
   "step": 24000,
   "tokens": 6291456000,
   "kind": "resume",
   "path": "resume/latest.pt",
   "val_bpb": null,
   "utc": "2026-10-10T16:14:51Z"
  },
  {
   "step": 25561,
   "tokens": 6700662784,
   "kind": "safetensors",
   "path": "checkpoints/step_00025561.safetensors",
   "val_bpb": 1.1565509513183632,
   "utc": "2026-10-10T17:37:49Z"
  },
  {
   "step": 25561,
   "tokens": 6700662784,
   "kind": "resume",
   "path": "resume/latest.pt",
   "val_bpb": null,
   "utc": "2026-10-10T17:37:49Z"
  },
  {
   "step": 25561,
   "tokens": 6700662784,
   "kind": "safetensors",
   "path": "checkpoints/step_00025561.safetensors",
   "val_bpb": 1.1565509513183632,
   "utc": "2026-10-10T17:38:24Z"
  },
  {
   "step": 25561,
   "tokens": 6700662784,
   "kind": "fp8",
   "path": "checkpoints/fp8/step_00025561.safetensors",
   "val_bpb": null,
   "utc": "2026-10-10T17:38:29Z"
  },
  {
   "step": 25561,
   "tokens": 6700662784,
   "kind": "resume",
   "path": "resume/latest.pt",
   "val_bpb": null,
   "utc": "2026-10-10T17:38:29Z"
  },
  {
   "step": 26000,
   "tokens": 6815744000,
   "kind": "safetensors",
   "path": "checkpoints/step_00026000.safetensors",
   "val_bpb": 1.1457359083101808,
   "utc": "2026-10-10T18:02:56Z"
  },
  {
   "step": 26000,
   "tokens": 6815744000,
   "kind": "resume",
   "path": "resume/latest.pt",
   "val_bpb": null,
   "utc": "2026-10-10T18:02:56Z"
  },
  {
   "step": 28000,
   "tokens": 7340032000,
   "kind": "safetensors",
   "path": "checkpoints/step_00028000.safetensors",
   "val_bpb": 1.1394487922871677,
   "utc": "2026-10-10T19:49:59Z"
  },
  {
   "step": 28000,
   "tokens": 7340032000,
   "kind": "resume",
   "path": "resume/latest.pt",
   "val_bpb": null,
   "utc": "2026-10-10T19:49:59Z"
  },
  {
   "step": 29376,
   "tokens": 7700742144,
   "kind": "safetensors",
   "path": "checkpoints/step_00029376.safetensors",
   "val_bpb": 1.1394487922871677,
   "utc": "2026-10-10T21:03:56Z"
  },
  {
   "step": 29376,
   "tokens": 7700742144,
   "kind": "fp8",
   "path": "checkpoints/fp8/step_00029376.safetensors",
   "val_bpb": null,
   "utc": "2026-10-10T21:04:02Z"
  },
  {
   "step": 29376,
   "tokens": 7700742144,
   "kind": "resume",
   "path": "resume/latest.pt",
   "val_bpb": null,
   "utc": "2026-10-10T21:04:02Z"
  },
  {
   "step": 29376,
   "tokens": 7700742144,
   "kind": "safetensors",
   "path": "checkpoints/step_00029376.safetensors",
   "val_bpb": 1.1394487922871677,
   "utc": "2026-10-10T21:04:38Z"
  },
  {
   "step": 29376,
   "tokens": 7700742144,
   "kind": "fp8",
   "path": "checkpoints/fp8/step_00029376.safetensors",
   "val_bpb": null,
   "utc": "2026-10-10T21:04:39Z"
  },
  {
   "step": 29376,
   "tokens": 7700742144,
   "kind": "resume",
   "path": "resume/latest.pt",
   "val_bpb": null,
   "utc": "2026-10-10T21:04:39Z"
  },
  {
   "step": 30000,
   "tokens": 7864320000,
   "kind": "safetensors",
   "path": "checkpoints/step_00030000.safetensors",
   "val_bpb": 1.1527475764923636,
   "utc": "2026-10-10T21:38:37Z"
  },
  {
   "step": 30000,
   "tokens": 7864320000,
   "kind": "resume",
   "path": "resume/latest.pt",
   "val_bpb": null,
   "utc": "2026-10-10T21:38:37Z"
  },
  {
   "step": 30841,
   "tokens": 8084783104,
   "kind": "safetensors",
   "path": "checkpoints/step_00030841.safetensors",
   "val_bpb": 1.1527475764923636,
   "utc": "2026-10-10T22:23:36Z"
  },
  {
   "step": 30841,
   "tokens": 8084783104,
   "kind": "fp8",
   "path": "checkpoints/fp8/step_00030841.safetensors",
   "val_bpb": null,
   "utc": "2026-10-10T22:23:42Z"
  },
  {
   "step": 30841,
   "tokens": 8084783104,
   "kind": "resume",
   "path": "resume/latest.pt",
   "val_bpb": null,
   "utc": "2026-10-10T22:23:42Z"
  },
  {
   "step": 32000,
   "tokens": 8388608000,
   "kind": "safetensors",
   "path": "checkpoints/step_00032000.safetensors",
   "val_bpb": 1.1610089243602937,
   "utc": "2026-10-11T00:01:30Z"
  },
  {
   "step": 32000,
   "tokens": 8388608000,
   "kind": "resume",
   "path": "resume/latest.pt",
   "val_bpb": null,
   "utc": "2026-10-11T00:01:30Z"
  },
  {
   "step": 33954,
   "tokens": 8900837376,
   "kind": "safetensors",
   "path": "checkpoints/step_00033954.safetensors",
   "val_bpb": 1.1610089243602937,
   "utc": "2026-10-11T02:30:06Z"
  },
  {
   "step": 33954,
   "tokens": 8900837376,
   "kind": "resume",
   "path": "resume/latest.pt",
   "val_bpb": null,
   "utc": "2026-10-11T02:30:06Z"
  },
  {
   "step": 33954,
   "tokens": 8900837376,
   "kind": "safetensors",
   "path": "checkpoints/step_00033954.safetensors",
   "val_bpb": 1.1610089243602937,
   "utc": "2026-10-11T02:30:43Z"
  },
  {
   "step": 33954,
   "tokens": 8900837376,
   "kind": "fp8",
   "path": "checkpoints/fp8/step_00033954.safetensors",
   "val_bpb": null,
   "utc": "2026-10-11T02:30:50Z"
  },
  {
   "step": 33954,
   "tokens": 8900837376,
   "kind": "resume",
   "path": "resume/latest.pt",
   "val_bpb": null,
   "utc": "2026-10-11T02:30:50Z"
  },
  {
   "step": 34000,
   "tokens": 8912896000,
   "kind": "safetensors",
   "path": "checkpoints/step_00034000.safetensors",
   "val_bpb": 1.18245087210553,
   "utc": "2026-10-11T02:35:10Z"
  },
  {
   "step": 34000,
   "tokens": 8912896000,
   "kind": "resume",
   "path": "resume/latest.pt",
   "val_bpb": null,
   "utc": "2026-10-11T02:35:10Z"
  },
  {
   "step": 36000,
   "tokens": 9437184000,
   "kind": "safetensors",
   "path": "checkpoints/step_00036000.safetensors",
   "val_bpb": 1.1530911870798717,
   "utc": "2026-10-11T05:07:42Z"
  },
  {
   "step": 36000,
   "tokens": 9437184000,
   "kind": "fp8",
   "path": "checkpoints/fp8/step_00036000.safetensors",
   "val_bpb": null,
   "utc": "2026-10-11T05:07:48Z"
  },
  {
   "step": 36000,
   "tokens": 9437184000,
   "kind": "resume",
   "path": "resume/latest.pt",
   "val_bpb": null,
   "utc": "2026-10-11T05:07:48Z"
  }
 ],
 "data_state": {
  "dataset": "curriculum-s4",
  "rows_consumed": 4043546,
  "epoch": 0
 },
 "notes": [
  {
   "utc": "2026-10-10T03:52:07Z",
   "msg": "manifest created"
  },
  {
   "utc": "2026-10-10T03:52:07Z",
   "msg": "WEIGHTS-ONLY START at step 10,000 from AbstractPhil/alephllm-mini-beatrix-training/mini-beatrix-2s-control/checkpoints/step_00010000.safetensors: fresh optimizer states; the curriculum cursor set from the step (active phase 'fineweb_main', 2.621B tokens counted); its stream opens at seed offset 7919 (a data-order discontinuity by design: the source run consumed the shuffle head)"
  },
  {
   "utc": "2026-10-10T12:51:48Z",
   "msg": "phase 'fineweb_main' complete at 5,000,134,656 tokens"
  },
  {
   "utc": "2026-10-10T12:52:32Z",
   "msg": "fp16 kernel this process: 1,030,620 calls, 0 backward retries (the gradient scale lowered after an overflow), 0 fp32 fallbacks (a forward overflow)"
  },
  {
   "utc": "2026-10-10T15:14:14Z",
   "msg": "phase 'curriculum_s0' complete at 700,186,624 tokens"
  },
  {
   "utc": "2026-10-10T15:15:02Z",
   "msg": "fp16 kernel this process: 1,457,980 calls, 0 backward retries (the gradient scale lowered after an overflow), 0 fp32 fallbacks (a forward overflow)"
  },
  {
   "utc": "2026-10-10T17:37:39Z",
   "msg": "phase 'curriculum_s1' complete at 700,186,624 tokens"
  },
  {
   "utc": "2026-10-10T17:38:22Z",
   "msg": "fp16 kernel this process: 1,885,340 calls, 0 backward retries (the gradient scale lowered after an overflow), 0 fp32 fallbacks (a forward overflow)"
  },
  {
   "utc": "2026-10-10T21:03:48Z",
   "msg": "phase 'curriculum_s2' complete at 1,000,079,360 tokens"
  },
  {
   "utc": "2026-10-10T21:04:36Z",
   "msg": "fp16 kernel this process: 2,495,740 calls, 0 backward retries (the gradient scale lowered after an overflow), 0 fp32 fallbacks (a forward overflow)"
  },
  {
   "utc": "2026-10-10T22:23:29Z",
   "msg": "fp16 kernel this process: 2,730,260 calls, 0 backward retries (the gradient scale lowered after an overflow), 0 fp32 fallbacks (a forward overflow)"
  },
  {
   "utc": "2026-10-11T02:29:57Z",
   "msg": "phase 'curriculum_s3' complete at 1,200,095,232 tokens"
  },
  {
   "utc": "2026-10-11T02:30:41Z",
   "msg": "fp16 kernel this process: 498,140 calls, 0 backward retries (the gradient scale lowered after an overflow), 0 fp32 fallbacks (a forward overflow)"
  }
 ],
 "halt": null,
 "data_plane": {
  "data_scale": 1.0,
  "epoch_cap": 4.0,
  "rebalance_to": null,
  "recipe_hash": "1b0cebcb3490",
  "minted_lexicon": null
 },
 "init_from": {
  "source": "AbstractPhil/alephllm-mini-beatrix-training/mini-beatrix-2s-control/checkpoints/step_00010000.safetensors",
  "step": 10000,
  "phase": "fineweb_main",
  "seed_offset": 7919,
  "tokens_per_step": 262144,
  "live_tokens_per_step": 262144,
  "weights": "the shipped bf16 checkpoint, upcast to the fp32 masters",
  "optimizers": "fresh (Muon momentum and Adam moments start at zero)",
  "guards_added": [],
  "utc": "2026-10-10T03:52:07Z"
 }
}