import arm qwen3/head_k16_solo_o0_lamA2: anchors (every close), stats, results, log (the second home until the publishing convention is settled; Phil 2026-10-08)

mini-beatrix-3/arm_refit/import/qwen3/head_k16_solo_o0_lamA2/closes/close1_heads.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:b9282214ccd39a3c4eec9dee0961472a3f1cbcf301e7eeab11577797e83c1669
+size 134380517

mini-beatrix-3/arm_refit/import/qwen3/head_k16_solo_o0_lamA2/closes/close2_heads.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:b5cfc6c6dbb53174eff4b2f39b60168cc4cecf54be1c6232ea6ff3b0b78be1b0
+size 134380517

mini-beatrix-3/arm_refit/import/qwen3/head_k16_solo_o0_lamA2/closes/qimp_head_k16_solo_o0_lamA2_close1_s15_qwen_import.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:ee54ed327f000687d66ad6a5a2f370b09cab89ce3de4e3cf8166e6c1849300c3
+size 54819536

mini-beatrix-3/arm_refit/import/qwen3/head_k16_solo_o0_lamA2/closes/qimp_head_k16_solo_o0_lamA2_close2_s15_qwen_import.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:40f4b0d12a252c0ae60c51ad22decbb5cac77fc707a462d64a05ef3b428348e1
+size 54819536

mini-beatrix-3/arm_refit/import/qwen3/head_k16_solo_o0_lamA2/frame.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:de185c8e673419c19e65aaf61768da566c05382bd4a0aefe1d2bf921522b38a3
+size 134371583

mini-beatrix-3/arm_refit/import/qwen3/head_k16_solo_o0_lamA2/manifest.json

@@ -0,0 +1,18 @@
+{
+ "tag": "qimp_head_k16_solo_o0_lamA2",
+ "run": "head_k16_solo_o0_lamA2",
+ "convention": "qwen3",
+ "member": "s15_qwen_import",
+ "final_anchor": "qimp_head_k16_solo_o0_lamA2_close2_s15_qwen_import.safetensors",
+ "closes": [
+  "qimp_head_k16_solo_o0_lamA2_close1_s15_qwen_import.safetensors",
+  "qimp_head_k16_solo_o0_lamA2_close2_s15_qwen_import.safetensors"
+ ],
+ "finished": true,
+ "form": "head",
+ "base": "solo",
+ "steps": 2000,
+ "trunk_step": 245674,
+ "shipped": "2026-10-08 19:58 local",
+ "note": "every close's anchor ships (mids included); the final close's anchor is the run's s15_qwen_import.safetensors"
+}

mini-beatrix-3/arm_refit/import/qwen3/head_k16_solo_o0_lamA2/s15_qwen_import.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:40f4b0d12a252c0ae60c51ad22decbb5cac77fc707a462d64a05ef3b428348e1
+size 54819536

mini-beatrix-3/arm_refit/import/qwen3/head_k16_solo_o0_lamA2/stats.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:33cd02438163cabe1e3c135f976e98fad21dab80fd9cea632a256ccfafb78a3b
+size 264373

mini-beatrix-3/arm_refit/import/qwen3/head_k16_solo_o0_lamA2/train.log

@@ -0,0 +1,96 @@
+[    0.5s] THE JOB: qimp_head_k16_solo_o0_lamA2: the Qwen import arm s15_qwen_import (head form; the teacher Qwen3 0.6B base at depth 16) alone on the bare trunk; blocks all; 2000 steps, a close every 1000; lambda 2.0 off-domain, 2.0 on the caption rows; pure Adam lr 0.001; the compiled chain with recompute; card 30.5 GB with 30.4 GB free; host RSS 0.6 GB
+Warning: You are sending unauthenticated requests to the HF Hub. Please set a HF_TOKEN to enable higher rate limits and faster downloads.
+[   19.0s] mini-beatrix-3 @ 245,674: trunk loaded from /workspace/qarm/dl/mini-beatrix-3/checkpoints/step_00245674.safetensors and frozen (376.0M); alephllm 0.10.7 at /workspace/qarm/alephllm/geolip/alephllm; amoe-lora 0.2.11; torch 2.8.0+cu128
+[   19.7s] tokenizer /workspace/qarm/dl/qwen3_tokenizer (gpt2 spelling; vocab 151669; sha256 be75606093db2094): 'a taco truck' -> ['a', 'Ġtaco', 'Ġtruck']; sites [(1, 6, 7)]
+[   29.2s] the teacher: Qwen3 0.6B base (/workspace/qarm/dl/qwen_3_06b_base.safetensors, sha256 cd2a512003e2f9f3; 28 layers, width 1024; config /workspace/qarm/dl/qwen3_06b); depth 16 = the input of layer 16
+[   29.2s] the read's captions: 512 of each draw of /workspace/qarm/dl/coco_caps_2x2048.json (4,090 captions of both draws kept out of training)
+[   42.1s] caption pack (/workspace/qarm/dl/mini-beatrix-3/caption_pack/pack_v2, train): dff 121,153 rows, scf 42,080 rows, danbooru 603,026 rows, cc12m 1,242,176 rows, coco 19,227 rows; host RSS 4.0 GB
+[   46.7s] read rows draw1: 512 captions (0 unspellable or too long), rows (512, 99), 4,920 sites; Qwen's states at ['16', '28post'] (fp32; final-state size 132.0)
+[   50.7s] read rows draw2: 512 captions (0 unspellable or too long), rows (512, 151), 5,086 sites; Qwen's states at ['16', '28post'] (fp32; final-state size 131.5)
+[   50.8s] streams: web text; partners none (two web rows a quiet chunk); held-out sets ['captions', 'web'] x 8 batches
+[arms] attached 's15_qwen_import' for phase 'qwen_import': 13.69M params x 32 blocks, lambda 2.0, spec {'n_slots': 16, 'K': 16, 'D': 8, 'hidden': 256}, recompute on, compile on
+[   68.9s] BIRTH: the member attached over the trunk: web rows read exactly as before; the hand-run trunk against the model's forward: armed 0.0e+00, masked 0.0e+00 (gates: 0.0)
+[   80.5s] standardization statistics from 13,911 reference sites on 8 batches (hers per block; Qwen's depth-16 states: size 61.7, sigma range 0.486..5.375); saved /workspace/qarm/out/anchors/qimp_head_k16_solo_o0_lamA2_stats.pt
+[  219.2s] THE HEADS (linear, her standardized states -> Qwen's standardized depth-16 states, born as fp64 ridges, 139 s): held-out R2 per block 8: 0.351, 12: 0.397, 16: 0.432, 20: 0.458, 24: 0.438, 28: 0.385, 31: 0.334; 33.6M head parameters; saved /workspace/qarm/out/anchors/qimp_head_k16_solo_o0_lamA2_frame.pt
+[  497.1s] CLOSE 0 (step 0): the import read on draw 1 (512 captions, 4,920 sites); per block: her reading against Qwen's final state, masked -> armed / against depth 16 masked -> armed; context (token means removed) masked -> armed
+[  497.1s]    block  8: final 0.426 -> 0.426 / depth 0.496 -> 0.496; context 0.285 -> 0.285; mse 0.866 cos 0.512
+[  497.1s]    block 12: final 0.470 -> 0.470 / depth 0.548 -> 0.548; context 0.331 -> 0.331; mse 0.827 cos 0.546
+[  497.1s]    block 16: final 0.508 -> 0.508 / depth 0.579 -> 0.579; context 0.372 -> 0.372; mse 0.785 cos 0.577
+[  497.1s]    block 18: final 0.514 -> 0.514 / depth 0.581 -> 0.581; context 0.386 -> 0.386; mse 0.762 cos 0.593
+[  497.1s]    block 20: final 0.526 -> 0.526 / depth 0.573 -> 0.573; context 0.399 -> 0.399; mse 0.759 cos 0.595
+[  497.1s]    block 24: final 0.562 -> 0.562 / depth 0.526 -> 0.526; context 0.466 -> 0.466; mse 0.792 cos 0.571
+[  497.1s]    block 28: final 0.540 -> 0.540 / depth 0.452 -> 0.452; context 0.448 -> 0.448; mse 0.852 cos 0.527
+[  497.1s]    block 31: final 0.475 -> 0.475 / depth 0.385 -> 0.385; context 0.385 -> 0.385; mse 0.900 cos 0.494
+[  497.1s]    draw 2 (the repeat), armed against the final state at 12/16/20/24/28: 0.471 / 0.508 / 0.529 / 0.567 / 0.546
+[  497.1s]    I1 the reach: the largest rise over her bare reading +0.000 at block 31 (draw 2 +0.000 at 31) against the bar +.10; I2 the context: +0.000 at block 31 against +.10; bits: captions +0.00000, web +0.00000 against +.012
+[  497.1s]    gate mean 0.047 / widest 0.047; compiled chain: 2 cache entries (limit 256), 0 limit warnings; docs drawn 256 (skipped {'test': 2, 'spelling': 0, 'empty': 0, 'short': 0}); card peak 6.9 GB; host RSS 9.2 GB
+[  497.1s] THE INSTRUMENT CHECK: her bare reading against Qwen's final state on draw 1 (this driver / e038's row): block 8 0.426 / 0.426, block 12 0.470 / 0.470, block 16 0.508 / 0.509, block 20 0.526 / 0.526, block 24 0.562 / 0.562, block 28 0.540 / 0.540; largest difference 0.001
+[stream] opening 'fineweb-edu' (HuggingFaceFW/fineweb-edu) …
+[  504.9s] step 1/2000: task 0.6398 (mse 0.640 cos 0.603); quiet A 0.00000 off 0.00000; 1,727 sites; 7.79 s/step; the job: 8.4 min spent, about 259.4 min of steps left; card peak 16.1 GB
+[  635.3s] step 50/2000: task 0.6273 (mse 0.627 cos 0.610); quiet A 0.00026 off 0.00021; 1,733 sites; 2.76 s/step; the job: 10.6 min spent, about 89.8 min of steps left; card peak 16.5 GB
+[  767.6s] step 100/2000: task 0.6159 (mse 0.616 cos 0.619); quiet A 0.00029 off 0.00024; 1,740 sites; 2.70 s/step; the job: 12.8 min spent, about 85.6 min of steps left; card peak 16.5 GB
+[  901.3s] step 150/2000: task 0.6095 (mse 0.609 cos 0.624); quiet A 0.00032 off 0.00025; 1,738 sites; 2.69 s/step; the job: 15.0 min spent, about 83.1 min of steps left; card peak 16.5 GB
+[ 1034.6s] step 200/2000: task 0.6110 (mse 0.611 cos 0.622); quiet A 0.00036 off 0.00025; 1,717 sites; 2.69 s/step; the job: 17.2 min spent, about 80.6 min of steps left; card peak 16.5 GB
+[ 1167.5s] step 250/2000: task 0.6040 (mse 0.604 cos 0.628); quiet A 0.00042 off 0.00025; 1,721 sites; 2.68 s/step; the job: 19.5 min spent, about 78.2 min of steps left; card peak 16.5 GB
+[ 1301.1s] step 300/2000: task 0.6176 (mse 0.618 cos 0.620); quiet A 0.00049 off 0.00025; 1,750 sites; 2.68 s/step; the job: 21.7 min spent, about 75.9 min of steps left; card peak 16.5 GB
+[ 1434.7s] step 350/2000: task 0.6135 (mse 0.614 cos 0.623); quiet A 0.00059 off 0.00025; 1,725 sites; 2.68 s/step; the job: 23.9 min spent, about 73.7 min of steps left; card peak 16.5 GB
+[ 1568.4s] step 400/2000: task 0.6014 (mse 0.601 cos 0.630); quiet A 0.00068 off 0.00024; 1,772 sites; 2.68 s/step; the job: 26.1 min spent, about 71.4 min of steps left; card peak 16.5 GB
+[ 1701.4s] step 450/2000: task 0.6069 (mse 0.607 cos 0.627); quiet A 0.00079 off 0.00025; 1,827 sites; 2.68 s/step; the job: 28.4 min spent, about 69.1 min of steps left; card peak 16.5 GB
+[ 1834.8s] step 500/2000: task 0.5995 (mse 0.600 cos 0.630); quiet A 0.00091 off 0.00025; 1,767 sites; 2.68 s/step; the job: 30.6 min spent, about 66.9 min of steps left; card peak 16.5 GB
+[ 1968.4s] step 550/2000: task 0.6955 (mse 0.695 cos 0.628); quiet A 0.00101 off 0.00026; 1,682 sites; 2.68 s/step; the job: 32.8 min spent, about 64.6 min of steps left; card peak 16.5 GB
+[ 2101.6s] step 600/2000: task 0.6458 (mse 0.646 cos 0.631); quiet A 0.00114 off 0.00028; 1,717 sites; 2.67 s/step; the job: 35.0 min spent, about 62.4 min of steps left; card peak 16.5 GB
+[ 2234.8s] step 650/2000: task 0.6084 (mse 0.608 cos 0.627); quiet A 0.00127 off 0.00026; 1,805 sites; 2.67 s/step; the job: 37.2 min spent, about 60.2 min of steps left; card peak 16.5 GB
+[ 2368.8s] step 700/2000: task 0.6400 (mse 0.640 cos 0.632); quiet A 0.00139 off 0.00027; 1,672 sites; 2.67 s/step; the job: 39.5 min spent, about 57.9 min of steps left; card peak 16.5 GB
+[ 2501.5s] step 750/2000: task 0.6858 (mse 0.686 cos 0.631); quiet A 0.00152 off 0.00028; 1,708 sites; 2.67 s/step; the job: 41.7 min spent, about 55.7 min of steps left; card peak 16.5 GB
+[ 2634.9s] step 800/2000: task 0.5959 (mse 0.596 cos 0.634); quiet A 0.00170 off 0.00028; 1,771 sites; 2.67 s/step; the job: 43.9 min spent, about 53.4 min of steps left; card peak 16.5 GB
+[ 2767.9s] step 850/2000: task 0.6023 (mse 0.602 cos 0.630); quiet A 0.00182 off 0.00027; 1,665 sites; 2.67 s/step; the job: 46.1 min spent, about 51.2 min of steps left; card peak 16.5 GB
+[ 2900.9s] step 900/2000: task 0.6052 (mse 0.605 cos 0.630); quiet A 0.00197 off 0.00029; 1,848 sites; 2.67 s/step; the job: 48.3 min spent, about 49.0 min of steps left; card peak 16.5 GB
+[ 3034.4s] step 950/2000: task 0.6582 (mse 0.658 cos 0.626); quiet A 0.00212 off 0.00028; 1,781 sites; 2.67 s/step; the job: 50.6 min spent, about 46.7 min of steps left; card peak 16.5 GB
+[ 3167.9s] step 1000/2000: task 0.6043 (mse 0.604 cos 0.631); quiet A 0.00228 off 0.00029; 1,745 sites; 2.67 s/step; the job: 52.8 min spent, about 44.5 min of steps left; card peak 16.5 GB
+[ 3168.4s] anchor saved: /workspace/qarm/out/anchors/qimp_head_k16_solo_o0_lamA2_close1_s15_qwen_import.safetensors
+[ 3310.4s] CLOSE 1 (step 1000): the import read on draw 1 (512 captions, 4,920 sites); per block: her reading against Qwen's final state, masked -> armed / against depth 16 masked -> armed; context (token means removed) masked -> armed
+[ 3310.4s]    block  8: final 0.426 -> 0.426 / depth 0.496 -> 0.498; context 0.285 -> 0.289; mse 0.833 cos 0.538
+[ 3310.4s]    block 12: final 0.470 -> 0.470 / depth 0.548 -> 0.552; context 0.331 -> 0.333; mse 0.784 cos 0.577
+[ 3310.4s]    block 16: final 0.508 -> 0.510 / depth 0.579 -> 0.586; context 0.372 -> 0.374; mse 0.739 cos 0.610
+[ 3310.4s]    block 18: final 0.514 -> 0.518 / depth 0.581 -> 0.591; context 0.386 -> 0.388; mse 0.714 cos 0.627
+[ 3310.4s]    block 20: final 0.526 -> 0.529 / depth 0.573 -> 0.584; context 0.399 -> 0.402; mse 0.710 cos 0.629
+[ 3310.4s]    block 24: final 0.562 -> 0.566 / depth 0.526 -> 0.541; context 0.466 -> 0.471; mse 0.734 cos 0.613
+[ 3310.4s]    block 28: final 0.540 -> 0.549 / depth 0.452 -> 0.470; context 0.448 -> 0.456; mse 0.788 cos 0.574
+[ 3310.4s]    block 31: final 0.475 -> 0.486 / depth 0.385 -> 0.402; context 0.385 -> 0.396; mse 0.836 cos 0.542
+[ 3310.4s]    draw 2 (the repeat), armed against the final state at 12/16/20/24/28: 0.472 / 0.510 / 0.532 / 0.572 / 0.556
+[ 3310.4s]    I1 the reach: the largest rise over her bare reading +0.011 at block 30 (draw 2 +0.012 at 31) against the bar +.10; I2 the context: +0.011 at block 30 against +.10; bits: captions -0.01469, web -0.00002 against +.012
+[ 3310.4s]    gate mean 0.192 / widest 0.260; compiled chain: 5 cache entries (limit 256), 0 limit warnings; docs drawn 32,523 (skipped {'test': 44, 'spelling': 0, 'empty': 0, 'short': 0}); card peak 16.5 GB; host RSS 11.3 GB
+[ 3443.8s] step 1050/2000: task 0.5904 (mse 0.590 cos 0.638); quiet A 0.00240 off 0.00030; 1,763 sites; 2.67 s/step; the job: 57.4 min spent, about 42.3 min of steps left; card peak 16.5 GB
+[ 3577.7s] step 1100/2000: task 0.5866 (mse 0.587 cos 0.641); quiet A 0.00253 off 0.00030; 1,731 sites; 2.67 s/step; the job: 59.6 min spent, about 40.1 min of steps left; card peak 16.5 GB
+[ 3710.5s] step 1150/2000: task 0.5838 (mse 0.584 cos 0.643); quiet A 0.00267 off 0.00030; 1,714 sites; 2.67 s/step; the job: 61.8 min spent, about 37.8 min of steps left; card peak 16.5 GB
+[ 3843.6s] step 1200/2000: task 0.5902 (mse 0.590 cos 0.640); quiet A 0.00280 off 0.00031; 1,724 sites; 2.67 s/step; the job: 64.1 min spent, about 35.6 min of steps left; card peak 16.5 GB
+[ 3977.3s] step 1250/2000: task 0.5930 (mse 0.593 cos 0.637); quiet A 0.00294 off 0.00032; 1,695 sites; 2.67 s/step; the job: 66.3 min spent, about 33.4 min of steps left; card peak 16.5 GB
+[ 4110.3s] step 1300/2000: task 0.5927 (mse 0.593 cos 0.637); quiet A 0.00302 off 0.00032; 1,701 sites; 2.67 s/step; the job: 68.5 min spent, about 31.1 min of steps left; card peak 16.5 GB
+[ 4244.3s] step 1350/2000: task 0.6386 (mse 0.639 cos 0.635); quiet A 0.00317 off 0.00033; 1,710 sites; 2.67 s/step; the job: 70.7 min spent, about 28.9 min of steps left; card peak 16.5 GB
+[ 4377.8s] step 1400/2000: task 0.5822 (mse 0.582 cos 0.644); quiet A 0.00332 off 0.00031; 1,767 sites; 2.67 s/step; the job: 73.0 min spent, about 26.7 min of steps left; card peak 16.5 GB
+[ 4511.5s] step 1450/2000: task 0.6378 (mse 0.638 cos 0.639); quiet A 0.00340 off 0.00032; 1,719 sites; 2.67 s/step; the job: 75.2 min spent, about 24.5 min of steps left; card peak 16.5 GB
+[ 4644.0s] step 1500/2000: task 0.5967 (mse 0.597 cos 0.636); quiet A 0.00352 off 0.00035; 1,670 sites; 2.67 s/step; the job: 77.4 min spent, about 22.2 min of steps left; card peak 16.5 GB
+[ 4777.2s] step 1550/2000: task 0.6301 (mse 0.630 cos 0.645); quiet A 0.00364 off 0.00034; 1,703 sites; 2.67 s/step; the job: 79.6 min spent, about 20.0 min of steps left; card peak 16.5 GB
+[ 4910.3s] step 1600/2000: task 0.6607 (mse 0.661 cos 0.645); quiet A 0.00377 off 0.00033; 1,705 sites; 2.67 s/step; the job: 81.8 min spent, about 17.8 min of steps left; card peak 16.5 GB
+[ 5043.4s] step 1650/2000: task 0.6226 (mse 0.623 cos 0.644); quiet A 0.00391 off 0.00034; 1,780 sites; 2.67 s/step; the job: 84.1 min spent, about 15.6 min of steps left; card peak 16.5 GB
+[ 5176.3s] step 1700/2000: task 0.5878 (mse 0.588 cos 0.642); quiet A 0.00397 off 0.00035; 1,733 sites; 2.67 s/step; the job: 86.3 min spent, about 13.3 min of steps left; card peak 16.5 GB
+[ 5309.0s] step 1750/2000: task 0.5856 (mse 0.586 cos 0.643); quiet A 0.00412 off 0.00035; 1,704 sites; 2.66 s/step; the job: 88.5 min spent, about 11.1 min of steps left; card peak 16.5 GB
+[ 5442.2s] step 1800/2000: task 0.5825 (mse 0.582 cos 0.647); quiet A 0.00424 off 0.00035; 1,777 sites; 2.66 s/step; the job: 90.7 min spent, about 8.9 min of steps left; card peak 16.5 GB
+[ 5575.6s] step 1850/2000: task 0.5788 (mse 0.579 cos 0.646); quiet A 0.00434 off 0.00036; 1,728 sites; 2.66 s/step; the job: 92.9 min spent, about 6.7 min of steps left; card peak 16.5 GB
+[ 5708.9s] step 1900/2000: task 0.6257 (mse 0.626 cos 0.645); quiet A 0.00443 off 0.00035; 1,836 sites; 2.67 s/step; the job: 95.1 min spent, about 4.4 min of steps left; card peak 16.5 GB
+[ 5842.3s] step 1950/2000: task 0.6617 (mse 0.662 cos 0.650); quiet A 0.00448 off 0.00036; 1,742 sites; 2.67 s/step; the job: 97.4 min spent, about 2.2 min of steps left; card peak 16.5 GB
+[ 5975.1s] step 2000/2000: task 0.5833 (mse 0.583 cos 0.645); quiet A 0.00459 off 0.00037; 1,747 sites; 2.66 s/step; the job: 99.6 min spent, about 0.0 min of steps left; card peak 16.5 GB
+[ 5975.4s] anchor saved: /workspace/qarm/out/anchors/qimp_head_k16_solo_o0_lamA2_close2_s15_qwen_import.safetensors
+[ 6117.4s] CLOSE 2 (step 2000): the import read on draw 1 (512 captions, 4,920 sites); per block: her reading against Qwen's final state, masked -> armed / against depth 16 masked -> armed; context (token means removed) masked -> armed
+[ 6117.4s]    block  8: final 0.426 -> 0.427 / depth 0.496 -> 0.501; context 0.285 -> 0.292; mse 0.824 cos 0.543
+[ 6117.4s]    block 12: final 0.470 -> 0.473 / depth 0.548 -> 0.554; context 0.331 -> 0.336; mse 0.775 cos 0.581
+[ 6117.4s]    block 16: final 0.508 -> 0.511 / depth 0.579 -> 0.592; context 0.372 -> 0.377; mse 0.728 cos 0.616
+[ 6117.4s]    block 18: final 0.514 -> 0.519 / depth 0.581 -> 0.595; context 0.386 -> 0.389; mse 0.707 cos 0.631
+[ 6117.4s]    block 20: final 0.526 -> 0.531 / depth 0.573 -> 0.591; context 0.399 -> 0.405; mse 0.700 cos 0.635
+[ 6117.4s]    block 24: final 0.562 -> 0.568 / depth 0.526 -> 0.552; context 0.466 -> 0.472; mse 0.721 cos 0.620
+[ 6117.4s]    block 28: final 0.540 -> 0.556 / depth 0.452 -> 0.485; context 0.448 -> 0.462; mse 0.770 cos 0.585
+[ 6117.4s]    block 31: final 0.475 -> 0.494 / depth 0.385 -> 0.416; context 0.385 -> 0.401; mse 0.816 cos 0.556
+[ 6117.4s]    draw 2 (the repeat), armed against the final state at 12/16/20/24/28: 0.474 / 0.511 / 0.533 / 0.574 / 0.564
+[ 6117.4s]    I1 the reach: the largest rise over her bare reading +0.021 at block 30 (draw 2 +0.020 at 30) against the bar +.10; I2 the context: +0.019 at block 30 against +.10; bits: captions -0.02484, web -0.00005 against +.012
+[ 6117.4s]    gate mean 0.455 / widest 0.633; compiled chain: 5 cache entries (limit 256), 0 limit warnings; docs drawn 64,564 (skipped {'test': 97, 'spelling': 1, 'empty': 0, 'short': 0}); card peak 16.5 GB; host RSS 11.3 GB
+[ 6117.4s] DONE qimp_head_k16_solo_o0_lamA2: 2000 steps in 88.9 min of steps (2.67 s/step); 2 closes; 6117 s in all; card peak 16.5 GB; host RSS 11.3 GB; compiled chain: 5 cache entries (limit 256), 0 limit warnings