Maggio33 commited on
Commit
e9f0f81
·
verified ·
1 Parent(s): 83aca85

Final checkpoint sums (to 2860); run record: metrics, logs, arguments, data manifests, 13-gram overlap

Browse files
README.md CHANGED
@@ -23,8 +23,8 @@ tokens per parameter of 1.00. Each file holds the model weights **and the optimi
23
  | Last checkpoint before the decay | **`ckpt_2250.pt`** — starting point for further continued training |
24
  | Training curve and evaluations during training | https://track.fabryka.ai/run/4934429a-d393-42e0-8aac-03eae6b71ff7 |
25
 
26
- Checkpoints are added every 150 steps while the run is in progress (final: step 2860). Model card with evaluation
27
- results: SlayerLab/Slayer1B-1.01 (after the run).
28
 
29
  ## File format
30
 
@@ -46,6 +46,10 @@ db254bfb77e302fc618f9cc7e3c4e23f1e3d6923386e54ad5fb930ff78a04bc4 ckpt_1650.pt
46
  b46f5e02f74dab2db1055fa009668c784411f403e01198dfff36deae9a0200f5 ckpt_2100.pt
47
  aeddfb023abb11c8654b47c728c9680c232d1b6568697e3a1eee8128efd37054 ckpt_2250.pt
48
  616df6070fbc6e338e106d4b535030d06767063168e80494b67f508005904b73 ckpt_2400.pt
 
 
 
 
49
  ```
50
 
51
  ## License
 
23
  | Last checkpoint before the decay | **`ckpt_2250.pt`** — starting point for further continued training |
24
  | Training curve and evaluations during training | https://track.fabryka.ai/run/4934429a-d393-42e0-8aac-03eae6b71ff7 |
25
 
26
+ Checkpoints every 150 steps from 1200 to 2850, plus the final step 2860. Released model (weights in safetensors) and
27
+ model card: [SlayerLab/Slayer1B-1.01](https://huggingface.co/SlayerLab/Slayer1B-1.01).
28
 
29
  ## File format
30
 
 
46
  b46f5e02f74dab2db1055fa009668c784411f403e01198dfff36deae9a0200f5 ckpt_2100.pt
47
  aeddfb023abb11c8654b47c728c9680c232d1b6568697e3a1eee8128efd37054 ckpt_2250.pt
48
  616df6070fbc6e338e106d4b535030d06767063168e80494b67f508005904b73 ckpt_2400.pt
49
+ 82fd3d07914308411a2ddec965ce6881f296b8bb6956369096534bdc2c45b0c8 ckpt_2550.pt
50
+ e2b5b29f05d3113c0bc95a5a4d74fe07fee057031068bc82844f129a341c09f0 ckpt_2700.pt
51
+ decd0f02487fa15c65aaaaf9e2c2320f196571efa0337c8c588705915f7fd315 ckpt_2850.pt
52
+ b0a288798ae614c38e37ff8176a293f6650cec2221d59e46d5116afac7c31eba ckpt_2860.pt
53
  ```
54
 
55
  ## License
run/13gram_p01.json ADDED
@@ -0,0 +1,130 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "train": "wsad_p01/train.bin",
3
+ "train_sha256": "0494df580a1bd75099fba85df5fcd1df248b02bd47228f99a0ea6088e635e906",
4
+ "n": 13,
5
+ "tokeny_train": 1880520445,
6
+ "val": {
7
+ "arc_challenge.jsonl": {
8
+ "sha256": "8e6aacea5f9a7f6c65fe41f6e11a9f5e9bb0a4ef959bf1e57d3a6ce92838c9d5",
9
+ "dokumenty": 1172,
10
+ "ngramy": 53484,
11
+ "ngramy_z_trafieniem": 5,
12
+ "dokumenty_z_trafieniem": 3,
13
+ "udzial_dokumentow": 0.002559726962457338,
14
+ "dokumenty_pokrycie_ge_50proc": 0,
15
+ "dokumenty_pokrycie_ge_90proc": 0,
16
+ "dokumenty_krotsze_niz_n": 0
17
+ },
18
+ "arc_easy.jsonl": {
19
+ "sha256": "bc50a1c6b6b672f08d3ea719ebfff2c1a64d8f89e24b99e9d3377716903bf8d8",
20
+ "dokumenty": 2376,
21
+ "ngramy": 84531,
22
+ "ngramy_z_trafieniem": 31,
23
+ "dokumenty_z_trafieniem": 10,
24
+ "udzial_dokumentow": 0.004208754208754209,
25
+ "dokumenty_pokrycie_ge_50proc": 0,
26
+ "dokumenty_pokrycie_ge_90proc": 0,
27
+ "dokumenty_krotsze_niz_n": 0
28
+ },
29
+ "arc_easy_pl_ddudek.jsonl": {
30
+ "sha256": "88dfce7e6bf40e16c0f79ecd0ba00dc062825f754e5ef2f3e48e467f78b0c1dd",
31
+ "dokumenty": 2376,
32
+ "ngramy": 86925,
33
+ "ngramy_z_trafieniem": 42,
34
+ "dokumenty_z_trafieniem": 13,
35
+ "udzial_dokumentow": 0.005471380471380472,
36
+ "dokumenty_pokrycie_ge_50proc": 0,
37
+ "dokumenty_pokrycie_ge_90proc": 0,
38
+ "dokumenty_krotsze_niz_n": 1
39
+ },
40
+ "arc_easy_pl_eu20.jsonl": {
41
+ "sha256": "a5ec96a974ed143651ec3353f2221047ef6d39d326ccc4050b790a70b0e454ef",
42
+ "dokumenty": 2376,
43
+ "ngramy": 85737,
44
+ "ngramy_z_trafieniem": 30,
45
+ "dokumenty_z_trafieniem": 8,
46
+ "udzial_dokumentow": 0.003367003367003367,
47
+ "dokumenty_pokrycie_ge_50proc": 0,
48
+ "dokumenty_pokrycie_ge_90proc": 0,
49
+ "dokumenty_krotsze_niz_n": 3
50
+ },
51
+ "belebele_pol_Latn.jsonl": {
52
+ "sha256": "4aef39d8d2fb716f70a06d049f524390571f24afaf49457f8e01b04ea69503b6",
53
+ "dokumenty": 900,
54
+ "ngramy": 152263,
55
+ "ngramy_z_trafieniem": 76,
56
+ "dokumenty_z_trafieniem": 29,
57
+ "udzial_dokumentow": 0.03222222222222222,
58
+ "dokumenty_pokrycie_ge_50proc": 0,
59
+ "dokumenty_pokrycie_ge_90proc": 0,
60
+ "dokumenty_krotsze_niz_n": 0
61
+ },
62
+ "blimp.jsonl": {
63
+ "sha256": "643b0a041fe99aedd27c014e8642409d70fb6611a9c626546fbb7bc0fce02fd6",
64
+ "dokumenty": 67000,
65
+ "ngramy": 738562,
66
+ "ngramy_z_trafieniem": 0,
67
+ "dokumenty_z_trafieniem": 0,
68
+ "udzial_dokumentow": 0.0,
69
+ "dokumenty_pokrycie_ge_50proc": 0,
70
+ "dokumenty_pokrycie_ge_90proc": 0,
71
+ "dokumenty_krotsze_niz_n": 4447
72
+ },
73
+ "hellaswag.jsonl": {
74
+ "sha256": "92a6bcfabd4d5b67726fb08f534baf2c97343c80c0a194ff9324a52b3438a0f2",
75
+ "dokumenty": 10042,
76
+ "ngramy": 1710914,
77
+ "ngramy_z_trafieniem": 253,
78
+ "dokumenty_z_trafieniem": 27,
79
+ "udzial_dokumentow": 0.002688707428799044,
80
+ "dokumenty_pokrycie_ge_50proc": 0,
81
+ "dokumenty_pokrycie_ge_90proc": 0,
82
+ "dokumenty_krotsze_niz_n": 0
83
+ },
84
+ "multiblimp_pol.jsonl": {
85
+ "sha256": "3ef818f28e4514c31db5e36443e22c170a56860bfce7e1efba25bbf79be212cf",
86
+ "dokumenty": 3272,
87
+ "ngramy": 115533,
88
+ "ngramy_z_trafieniem": 610,
89
+ "dokumenty_z_trafieniem": 40,
90
+ "udzial_dokumentow": 0.012224938875305624,
91
+ "dokumenty_pokrycie_ge_50proc": 10,
92
+ "dokumenty_pokrycie_ge_90proc": 0,
93
+ "dokumenty_krotsze_niz_n": 183
94
+ },
95
+ "piqa.jsonl": {
96
+ "sha256": "97c82004b5f6846173369ba05ff2019414f86f8ec61b739c7bd73d5f0c2aa2a5",
97
+ "dokumenty": 1838,
98
+ "ngramy": 96232,
99
+ "ngramy_z_trafieniem": 0,
100
+ "dokumenty_z_trafieniem": 0,
101
+ "udzial_dokumentow": 0.0,
102
+ "dokumenty_pokrycie_ge_50proc": 0,
103
+ "dokumenty_pokrycie_ge_90proc": 0,
104
+ "dokumenty_krotsze_niz_n": 3
105
+ },
106
+ "wikitext.jsonl": {
107
+ "sha256": "ece83c9ce264b93f9ef954f38e167cb65292675107cfc49dfd4feaf16b03a38d",
108
+ "dokumenty": 62,
109
+ "ngramy": 304597,
110
+ "ngramy_z_trafieniem": 229,
111
+ "dokumenty_z_trafieniem": 26,
112
+ "udzial_dokumentow": 0.41935483870967744,
113
+ "dokumenty_pokrycie_ge_50proc": 0,
114
+ "dokumenty_pokrycie_ge_90proc": 0,
115
+ "dokumenty_krotsze_niz_n": 0
116
+ },
117
+ "winogrande.jsonl": {
118
+ "sha256": "acd69a9ddbca0ae020b14329e29bf16dd28f7357626aabc78372be989fd88ad4",
119
+ "dokumenty": 1267,
120
+ "ngramy": 40992,
121
+ "ngramy_z_trafieniem": 0,
122
+ "dokumenty_z_trafieniem": 0,
123
+ "udzial_dokumentow": 0.0,
124
+ "dokumenty_pokrycie_ge_50proc": 0,
125
+ "dokumenty_pokrycie_ge_90proc": 0,
126
+ "dokumenty_krotsze_niz_n": 0
127
+ }
128
+ },
129
+ "skrypt_sha256": "bec97edd47da5911305a6f3d0c52c793bb30214fef4e14065f524ade1814948c"
130
+ }
run/args.txt ADDED
@@ -0,0 +1 @@
 
 
1
+ --data-dir /root/pilot1b/data --out-dir /root/pilot1b/runs/doucz --steps 2860 --ckpt-every 150 --log-every 10 --vocab 32768 --dtype uint16 --block 4096 --n-layer 20 --n-embd 2048 --n-head 16 --value-residual --batch 256 --grad-accum 8 --norm rmsnorm --pos rope --rope-theta 500000 --ffn swiglu --ffn-mult 2.667 --qk-norm --attn-bias 0 --optimizer muon --muon-lr 0.02 --lr 6e-4 --min-lr 6e-5 --wd 0.1 --schedule wsd --decay-frac 0.2 --warmup 150 --act-ckpt --loss-chunk 1024 --sampler epoch --seed 1 --data-seed 20261023 --compile --eval-every 1000000 --eval-iters 1 --data-dir-2 /root/pilot1b/data_p01 --switch-step 1801
run/checkpoints_SHA256SUMS ADDED
@@ -0,0 +1,13 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ 3637f1d90f60c2151724ad88b4b7c52925c87b7330a7680974b9093e8c1b4021 ckpt_1200.pt
2
+ d6d75584ffbd7261ed23f489ed4927d3e33fc3a6fef3bd9d3ad46b309b63daf4 ckpt_1350.pt
3
+ 3b2af30f8f4331f7bf549b1e7f0b2643ac2ec8ad3369a42b579d578930ba2934 ckpt_1500.pt
4
+ db254bfb77e302fc618f9cc7e3c4e23f1e3d6923386e54ad5fb930ff78a04bc4 ckpt_1650.pt
5
+ 9591dfa0d41f68b9e5cc40db2caf6f601b102c31b37c76d2ddeffa9f26cec49e ckpt_1800.pt
6
+ 6b8037e19b3fc42bb1c70599754749cdbb6a507a27fa4e2c3e316d676e969555 ckpt_1950.pt
7
+ b46f5e02f74dab2db1055fa009668c784411f403e01198dfff36deae9a0200f5 ckpt_2100.pt
8
+ aeddfb023abb11c8654b47c728c9680c232d1b6568697e3a1eee8128efd37054 ckpt_2250.pt
9
+ 616df6070fbc6e338e106d4b535030d06767063168e80494b67f508005904b73 ckpt_2400.pt
10
+ 82fd3d07914308411a2ddec965ce6881f296b8bb6956369096534bdc2c45b0c8 ckpt_2550.pt
11
+ e2b5b29f05d3113c0bc95a5a4d74fe07fee057031068bc82844f129a341c09f0 ckpt_2700.pt
12
+ decd0f02487fa15c65aaaaf9e2c2320f196571efa0337c8c588705915f7fd315 ckpt_2850.pt
13
+ b0a288798ae614c38e37ff8176a293f6650cec2221d59e46d5116afac7c31eba ckpt_2860.pt
run/metrics.jsonl ADDED
@@ -0,0 +1,181 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {"step": 1060, "loss": 2.8355888724327087, "lr": 0.0006, "tok_s": 49514.65569001781}
2
+ {"step": 1070, "loss": 2.868690076470375, "lr": 0.0006, "tok_s": 109239.62583205878}
3
+ {"step": 1080, "loss": 2.849772420525551, "lr": 0.0006, "tok_s": 108731.41270258823}
4
+ {"step": 1090, "loss": 2.8553335845470427, "lr": 0.0006, "tok_s": 108247.71876262185}
5
+ {"step": 1100, "loss": 2.8493208408355715, "lr": 0.0006, "tok_s": 107870.8176958336}
6
+ {"step": 1110, "loss": 2.8484700292348863, "lr": 0.0006, "tok_s": 107629.93356256538}
7
+ {"step": 1120, "loss": 2.8432138890028, "lr": 0.0006, "tok_s": 108071.58766880153}
8
+ {"step": 1130, "loss": 2.841386917233467, "lr": 0.0006, "tok_s": 107987.28124382511}
9
+ {"step": 1140, "loss": 2.8191855221986772, "lr": 0.0006, "tok_s": 107987.22264654815}
10
+ {"step": 1150, "loss": 2.836642736196518, "lr": 0.0006, "tok_s": 107796.65818903201}
11
+ {"step": 1160, "loss": 2.8427219957113268, "lr": 0.0006, "tok_s": 108155.50931191628}
12
+ {"step": 1170, "loss": 2.8140008479356764, "lr": 0.0006, "tok_s": 108311.0283709636}
13
+ {"step": 1180, "loss": 2.815743073821068, "lr": 0.0006, "tok_s": 108201.03404464756}
14
+ {"step": 1190, "loss": 2.807954716682434, "lr": 0.0006, "tok_s": 107801.6596697397}
15
+ {"step": 1200, "loss": 2.837801617383957, "lr": 0.0006, "tok_s": 107815.18731532386}
16
+ {"step": 1210, "loss": 2.8055685728788378, "lr": 0.0006, "tok_s": 98763.15615477115}
17
+ {"step": 1220, "loss": 2.814597424864769, "lr": 0.0006, "tok_s": 108307.3042974815}
18
+ {"step": 1230, "loss": 2.8082002848386765, "lr": 0.0006, "tok_s": 108132.19462823083}
19
+ {"step": 1240, "loss": 2.809553936123848, "lr": 0.0006, "tok_s": 107957.26840983769}
20
+ {"step": 1250, "loss": 2.8165749609470367, "lr": 0.0006, "tok_s": 108113.87849951097}
21
+ {"step": 1260, "loss": 2.812753301858902, "lr": 0.0006, "tok_s": 108061.12958622658}
22
+ {"step": 1270, "loss": 2.7992117911577226, "lr": 0.0006, "tok_s": 108210.42788418503}
23
+ {"step": 1280, "loss": 2.7989447355270385, "lr": 0.0006, "tok_s": 108128.41664798697}
24
+ {"step": 1290, "loss": 2.7926212579011915, "lr": 0.0006, "tok_s": 107743.6193828888}
25
+ {"step": 1300, "loss": 2.7574854761362078, "lr": 0.0006, "tok_s": 107732.50555301548}
26
+ {"step": 1310, "loss": 2.7675517439842223, "lr": 0.0006, "tok_s": 107801.73524113033}
27
+ {"step": 1320, "loss": 2.776053211092949, "lr": 0.0006, "tok_s": 108185.44433973345}
28
+ {"step": 1330, "loss": 2.793119618296623, "lr": 0.0006, "tok_s": 107999.25339132732}
29
+ {"step": 1340, "loss": 2.788493865728378, "lr": 0.0006, "tok_s": 107916.69591270621}
30
+ {"step": 1350, "loss": 2.7733452320098877, "lr": 0.0006, "tok_s": 108009.5093116523}
31
+ {"step": 1360, "loss": 2.7871426165103914, "lr": 0.0006, "tok_s": 98694.3397800095}
32
+ {"step": 1370, "loss": 2.7688917189836504, "lr": 0.0006, "tok_s": 108305.93284031458}
33
+ {"step": 1380, "loss": 2.7654865086078644, "lr": 0.0006, "tok_s": 107983.47361360268}
34
+ {"step": 1390, "loss": 2.7674186497926714, "lr": 0.0006, "tok_s": 107907.34611639766}
35
+ {"step": 1400, "loss": 2.7651878595352173, "lr": 0.0006, "tok_s": 107833.06712285145}
36
+ {"step": 1410, "loss": 2.76025028526783, "lr": 0.0006, "tok_s": 107875.23733077542}
37
+ {"step": 1420, "loss": 2.782536968588829, "lr": 0.0006, "tok_s": 107990.15523990303}
38
+ {"step": 1430, "loss": 2.7737979143857956, "lr": 0.0006, "tok_s": 108139.92740386253}
39
+ {"step": 1440, "loss": 2.7555275529623033, "lr": 0.0006, "tok_s": 108163.19754063788}
40
+ {"step": 1450, "loss": 2.761045199632645, "lr": 0.0006, "tok_s": 108188.44492825941}
41
+ {"step": 1460, "loss": 2.741342243552208, "lr": 0.0006, "tok_s": 108134.87826847275}
42
+ {"step": 1470, "loss": 2.754783868789673, "lr": 0.0006, "tok_s": 108067.43313671649}
43
+ {"step": 1480, "loss": 2.75789669752121, "lr": 0.0006, "tok_s": 108193.53606360417}
44
+ {"step": 1490, "loss": 2.747355082631111, "lr": 0.0006, "tok_s": 107963.11884104139}
45
+ {"step": 1500, "loss": 2.738798829913139, "lr": 0.0006, "tok_s": 107926.10109308893}
46
+ {"step": 1510, "loss": 2.7642705708742143, "lr": 0.0006, "tok_s": 98670.87299957886}
47
+ {"step": 1520, "loss": 2.7517167776823044, "lr": 0.0006, "tok_s": 108297.32961934055}
48
+ {"step": 1530, "loss": 2.741634425520897, "lr": 0.0006, "tok_s": 108167.13570742913}
49
+ {"step": 1540, "loss": 2.7395107358694077, "lr": 0.0006, "tok_s": 108347.34230448217}
50
+ {"step": 1550, "loss": 2.7479649424552917, "lr": 0.0006, "tok_s": 108358.47389903935}
51
+ {"step": 1560, "loss": 2.724382811784744, "lr": 0.0006, "tok_s": 108232.30577715716}
52
+ {"step": 1570, "loss": 2.738993287086487, "lr": 0.0006, "tok_s": 108028.18731416993}
53
+ {"step": 1580, "loss": 2.7299241334199906, "lr": 0.0006, "tok_s": 108055.70843067027}
54
+ {"step": 1590, "loss": 2.7522672712802887, "lr": 0.0006, "tok_s": 108020.51428617351}
55
+ {"step": 1600, "loss": 2.7404644578695296, "lr": 0.0006, "tok_s": 108153.73503545183}
56
+ {"step": 1610, "loss": 2.7167934119701385, "lr": 0.0006, "tok_s": 107893.32631204944}
57
+ {"step": 1620, "loss": 2.708075675368309, "lr": 0.0006, "tok_s": 107857.24408208238}
58
+ {"step": 1630, "loss": 2.7199666798114777, "lr": 0.0006, "tok_s": 107659.20198923118}
59
+ {"step": 1640, "loss": 2.713624891638756, "lr": 0.0006, "tok_s": 107823.27634913012}
60
+ {"step": 1650, "loss": 2.7406162858009337, "lr": 0.0006, "tok_s": 107778.54059229337}
61
+ {"step": 1660, "loss": 2.7050065279006956, "lr": 0.0006, "tok_s": 98087.3950005446}
62
+ {"step": 1670, "loss": 2.7108397483825684, "lr": 0.0006, "tok_s": 108101.71406147916}
63
+ {"step": 1680, "loss": 2.725912392139435, "lr": 0.0006, "tok_s": 107896.69771037067}
64
+ {"step": 1690, "loss": 2.701611289381981, "lr": 0.0006, "tok_s": 107780.49750863285}
65
+ {"step": 1700, "loss": 2.7218257039785385, "lr": 0.0006, "tok_s": 107921.41776246236}
66
+ {"step": 1710, "loss": 2.704691159725189, "lr": 0.0006, "tok_s": 107868.74717487369}
67
+ {"step": 1720, "loss": 2.7217802166938783, "lr": 0.0006, "tok_s": 107871.84584247085}
68
+ {"step": 1730, "loss": 2.710379382967949, "lr": 0.0006, "tok_s": 107896.8125909666}
69
+ {"step": 1740, "loss": 2.686915796995163, "lr": 0.0006, "tok_s": 107873.95643546285}
70
+ {"step": 1750, "loss": 2.6925191670656203, "lr": 0.0006, "tok_s": 107968.93783457343}
71
+ {"step": 1760, "loss": 2.685075503587723, "lr": 0.0006, "tok_s": 107962.53127815499}
72
+ {"step": 1770, "loss": 2.6855502486228944, "lr": 0.0006, "tok_s": 107892.29166826584}
73
+ {"step": 1780, "loss": 2.6988736003637315, "lr": 0.0006, "tok_s": 107713.04300499891}
74
+ {"step": 1790, "loss": 2.6702101826667786, "lr": 0.0006, "tok_s": 107620.20281422044}
75
+ {"step": 1800, "loss": 2.697207808494568, "lr": 0.0006, "tok_s": 107334.96294875555}
76
+ {"step": 1810, "loss": 2.6812927216291427, "lr": 0.0006, "tok_s": 97748.77790713932}
77
+ {"step": 1820, "loss": 2.6995500564575194, "lr": 0.0006, "tok_s": 108119.03066232133}
78
+ {"step": 1830, "loss": 2.7000478714704514, "lr": 0.0006, "tok_s": 107858.4648003429}
79
+ {"step": 1840, "loss": 2.692162424325943, "lr": 0.0006, "tok_s": 108091.78996543158}
80
+ {"step": 1850, "loss": 2.6932918190956117, "lr": 0.0006, "tok_s": 108196.15141512894}
81
+ {"step": 1860, "loss": 2.70576034784317, "lr": 0.0006, "tok_s": 108082.94661351903}
82
+ {"step": 1870, "loss": 2.6963472723960877, "lr": 0.0006, "tok_s": 108178.3441864483}
83
+ {"step": 1880, "loss": 2.6996336460113524, "lr": 0.0006, "tok_s": 108172.96899882071}
84
+ {"step": 1890, "loss": 2.693413272500038, "lr": 0.0006, "tok_s": 107940.68468094197}
85
+ {"step": 1900, "loss": 2.688009849190712, "lr": 0.0006, "tok_s": 108129.06795922168}
86
+ {"step": 1910, "loss": 2.692662012577057, "lr": 0.0006, "tok_s": 108020.95523222258}
87
+ {"step": 1920, "loss": 2.6745044946670533, "lr": 0.0006, "tok_s": 107919.62970815877}
88
+ {"step": 1930, "loss": 2.6717816829681396, "lr": 0.0006, "tok_s": 107949.21225162986}
89
+ {"step": 1940, "loss": 2.684726443886757, "lr": 0.0006, "tok_s": 108209.70583763365}
90
+ {"step": 1950, "loss": 2.6857749342918398, "lr": 0.0006, "tok_s": 107953.37757224798}
91
+ {"step": 1960, "loss": 2.6984307318925858, "lr": 0.0006, "tok_s": 97835.4561594414}
92
+ {"step": 1970, "loss": 2.668472445011139, "lr": 0.0006, "tok_s": 107868.04370264609}
93
+ {"step": 1980, "loss": 2.6823298960924147, "lr": 0.0006, "tok_s": 107690.36272470573}
94
+ {"step": 1990, "loss": 2.6722473174333574, "lr": 0.0006, "tok_s": 107690.80361639523}
95
+ {"step": 2000, "loss": 2.661471354961395, "lr": 0.0006, "tok_s": 107818.74784199921}
96
+ {"step": 2010, "loss": 2.6695849776268004, "lr": 0.0006, "tok_s": 108002.6465305499}
97
+ {"step": 2020, "loss": 2.6536724478006364, "lr": 0.0006, "tok_s": 107785.25261753154}
98
+ {"step": 2030, "loss": 2.667010545730591, "lr": 0.0006, "tok_s": 107548.99336353678}
99
+ {"step": 2040, "loss": 2.704330566525459, "lr": 0.0006, "tok_s": 107558.0675691955}
100
+ {"step": 2050, "loss": 2.6757187455892564, "lr": 0.0006, "tok_s": 107565.35381794942}
101
+ {"step": 2060, "loss": 2.673007407784462, "lr": 0.0006, "tok_s": 107804.65671149152}
102
+ {"step": 2070, "loss": 2.659758207201958, "lr": 0.0006, "tok_s": 107968.77429519371}
103
+ {"step": 2080, "loss": 2.670881485939026, "lr": 0.0006, "tok_s": 108008.00162123257}
104
+ {"step": 2090, "loss": 2.6693315744400024, "lr": 0.0006, "tok_s": 107843.55038987851}
105
+ {"step": 2100, "loss": 2.634330800175667, "lr": 0.0006, "tok_s": 107861.92448982407}
106
+ {"step": 2110, "loss": 2.664397010207176, "lr": 0.0006, "tok_s": 97559.2329144601}
107
+ {"step": 2120, "loss": 2.6637308210134507, "lr": 0.0006, "tok_s": 107790.94202451478}
108
+ {"step": 2130, "loss": 2.6516553431749346, "lr": 0.0006, "tok_s": 107568.98284224629}
109
+ {"step": 2140, "loss": 2.6471674919128416, "lr": 0.0006, "tok_s": 107636.00830411588}
110
+ {"step": 2150, "loss": 2.6531218707561495, "lr": 0.0006, "tok_s": 107872.56391636291}
111
+ {"step": 2160, "loss": 2.6534654557704926, "lr": 0.0006, "tok_s": 107674.79407404842}
112
+ {"step": 2170, "loss": 2.656611406803131, "lr": 0.0006, "tok_s": 107484.54929863823}
113
+ {"step": 2180, "loss": 2.668790024518967, "lr": 0.0006, "tok_s": 107483.29395039317}
114
+ {"step": 2190, "loss": 2.645035815238953, "lr": 0.0006, "tok_s": 107499.40372080148}
115
+ {"step": 2200, "loss": 2.649842268228531, "lr": 0.0006, "tok_s": 107382.78167397874}
116
+ {"step": 2210, "loss": 2.638586989045143, "lr": 0.0006, "tok_s": 107588.0030494198}
117
+ {"step": 2220, "loss": 2.6282501727342606, "lr": 0.0006, "tok_s": 107429.98750500844}
118
+ {"step": 2230, "loss": 2.643375760316849, "lr": 0.0006, "tok_s": 107231.01276669964}
119
+ {"step": 2240, "loss": 2.665416491031647, "lr": 0.0006, "tok_s": 107515.66629711786}
120
+ {"step": 2250, "loss": 2.631808540225029, "lr": 0.0006, "tok_s": 107529.30102989414}
121
+ {"step": 2260, "loss": 2.6378825336694716, "lr": 0.0006, "tok_s": 96309.13742732577}
122
+ {"step": 2270, "loss": 2.646059420704842, "lr": 0.0006, "tok_s": 107862.70459727371}
123
+ {"step": 2280, "loss": 2.6531503170728685, "lr": 0.0006, "tok_s": 107508.6995105937}
124
+ {"step": 2290, "loss": 2.64722101688385, "lr": 0.0005774214657298085, "tok_s": 107550.78046643571}
125
+ {"step": 2300, "loss": 2.6418594866991043, "lr": 0.000525115473509594, "tok_s": 107790.70029769574}
126
+ {"step": 2310, "loss": 2.631280091404915, "lr": 0.0004965321576256931, "tok_s": 107473.69369939031}
127
+ {"step": 2320, "loss": 2.620929941534996, "lr": 0.0004742880415254916, "tok_s": 107316.30364677172}
128
+ {"step": 2330, "loss": 2.6298107355833054, "lr": 0.0004554268399688289, "tok_s": 107306.17737206843}
129
+ {"step": 2340, "loss": 2.6195673823356627, "lr": 0.0004387570134880257, "tok_s": 107439.97600776868}
130
+ {"step": 2350, "loss": 2.6066989421844484, "lr": 0.00042365601003378984, "tok_s": 107240.89212964392}
131
+ {"step": 2360, "loss": 2.596417245268822, "lr": 0.0004097498885805716, "tok_s": 106925.43909553958}
132
+ {"step": 2370, "loss": 2.5959843754768372, "lr": 0.00039679319156827686, "tok_s": 107135.17575005986}
133
+ {"step": 2380, "loss": 2.5997823506593702, "lr": 0.00038461451049128426, "tok_s": 107354.079195453}
134
+ {"step": 2390, "loss": 2.589158457517624, "lr": 0.00037308853887736037, "tok_s": 107685.60438301729}
135
+ {"step": 2400, "loss": 2.600908464193344, "lr": 0.0003621203806823432, "tok_s": 107815.84305199448}
136
+ {"step": 2410, "loss": 2.6034529328346254, "lr": 0.0003516361230278939, "tok_s": 95862.69320362947}
137
+ {"step": 2420, "loss": 2.590833434462547, "lr": 0.0003415768634922223, "tok_s": 107941.37055650413}
138
+ {"step": 2430, "loss": 2.5812904596328736, "lr": 0.0003318947602518376, "tok_s": 107512.93287954979}
139
+ {"step": 2440, "loss": 2.577410587668419, "lr": 0.0003225503258637313, "tok_s": 107572.94284647373}
140
+ {"step": 2450, "loss": 2.5865168035030366, "lr": 0.0003135105171629789, "tok_s": 107692.00685771921}
141
+ {"step": 2460, "loss": 2.584952947497368, "lr": 0.0003047473524688967, "tok_s": 107579.01010373922}
142
+ {"step": 2470, "loss": 2.572823813557625, "lr": 0.0002962368883948744, "tok_s": 107413.26525153017}
143
+ {"step": 2480, "loss": 2.5803961187601088, "lr": 0.00028795844816766774, "tok_s": 107500.08110949333}
144
+ {"step": 2490, "loss": 2.586494725942612, "lr": 0.00027989402978414763, "tok_s": 107690.71580693712}
145
+ {"step": 2500, "loss": 2.5757115572690963, "lr": 0.0002720278452890637, "tok_s": 107345.64879438932}
146
+ {"step": 2510, "loss": 2.5683379590511324, "lr": 0.000264345957325647, "tok_s": 107447.27514024232}
147
+ {"step": 2520, "loss": 2.5658774793148043, "lr": 0.000256835988976789, "tok_s": 107255.80142772399}
148
+ {"step": 2530, "loss": 2.5492791712284086, "lr": 0.0002494868896040541, "tok_s": 107325.72493999987}
149
+ {"step": 2540, "loss": 2.552611404657364, "lr": 0.00024228874401643067, "tok_s": 107521.5217978572}
150
+ {"step": 2550, "loss": 2.55129331946373, "lr": 0.00023523261555445397, "tok_s": 107401.83076728905}
151
+ {"step": 2560, "loss": 2.537569320201874, "lr": 0.0002283104160012728, "tok_s": 95800.38994188794}
152
+ {"step": 2570, "loss": 2.5496610373258592, "lr": 0.0002215147969192865, "tok_s": 107578.76590565075}
153
+ {"step": 2580, "loss": 2.5446118742227553, "lr": 0.00021483905825103307, "tok_s": 107306.97093025854}
154
+ {"step": 2590, "loss": 2.550077572464943, "lr": 0.00020827707094573443, "tok_s": 107222.24068411377}
155
+ {"step": 2600, "loss": 2.5474867790937425, "lr": 0.0002018232110673008, "tok_s": 107512.63194928189}
156
+ {"step": 2610, "loss": 2.5570268005132677, "lr": 0.00019547230336767132, "tok_s": 107488.53356208059}
157
+ {"step": 2620, "loss": 2.537807011604309, "lr": 0.00018921957271486332, "tok_s": 107508.90213072642}
158
+ {"step": 2630, "loss": 2.5365156441926957, "lr": 0.00018306060207931707, "tok_s": 107669.88554366646}
159
+ {"step": 2640, "loss": 2.5356369584798815, "lr": 0.00017699129602765423, "tok_s": 107663.9245256355}
160
+ {"step": 2650, "loss": 2.556650850176811, "lr": 0.00017100784886636244, "tok_s": 107419.9919387986}
161
+ {"step": 2660, "loss": 2.5153169214725493, "lr": 0.00016510671673137113, "tok_s": 107310.84201830508}
162
+ {"step": 2670, "loss": 2.534062457084656, "lr": 0.00015928459304209713, "tok_s": 107136.08787689375}
163
+ {"step": 2680, "loss": 2.5275665760040282, "lr": 0.00015353838683715258, "tok_s": 107567.76735229077}
164
+ {"step": 2690, "loss": 2.5233274906873704, "lr": 0.00014786520358871499, "tok_s": 107605.58405771478}
165
+ {"step": 2700, "loss": 2.527819338440895, "lr": 0.00014226232815751742, "tok_s": 107162.97897660267}
166
+ {"step": 2710, "loss": 2.531130611896515, "lr": 0.00013672720960358748, "tok_s": 97103.1665525264}
167
+ {"step": 2720, "loss": 2.5199529200792314, "lr": 0.00013125744761161224, "tok_s": 107241.61934820058}
168
+ {"step": 2730, "loss": 2.5417379319667814, "lr": 0.00012585078032597953, "tok_s": 107247.85641721041}
169
+ {"step": 2740, "loss": 2.523855650424957, "lr": 0.00012050507342059957, "tok_s": 107313.99329280482}
170
+ {"step": 2750, "loss": 2.5106892466545103, "lr": 0.00011521831025369088, "tok_s": 107318.19981914318}
171
+ {"step": 2760, "loss": 2.4959987819194795, "lr": 0.00010998858297873422, "tok_s": 107011.92873306715}
172
+ {"step": 2770, "loss": 2.5276853412389757, "lr": 0.0001048140845004869, "tok_s": 106942.89184751152}
173
+ {"step": 2780, "loss": 2.524294966459274, "lr": 9.969310117989273e-05, "tok_s": 107074.4528784364}
174
+ {"step": 2790, "loss": 2.5118527859449387, "lr": 9.462400620439525e-05, "tok_s": 107164.10517203102}
175
+ {"step": 2800, "loss": 2.5065575748682023, "lr": 8.960525355094298e-05, "tok_s": 107364.47346421603}
176
+ {"step": 2810, "loss": 2.5089343070983885, "lr": 8.463537247818483e-05, "tok_s": 107308.7840361509}
177
+ {"step": 2820, "loss": 2.508688193559647, "lr": 7.971296249223989e-05, "tok_s": 107408.49516611568}
178
+ {"step": 2830, "loss": 2.49304376244545, "lr": 7.483668873720484e-05, "tok_s": 107398.15636394244}
179
+ {"step": 2840, "loss": 2.512237471342087, "lr": 7.000527776740496e-05, "tok_s": 107539.41713450302}
180
+ {"step": 2850, "loss": 2.4829426646232604, "lr": 6.521751366344496e-05, "tok_s": 107476.82381632294}
181
+ {"step": 2860, "loss": 2.5067392736673355, "lr": 6.0472234458494435e-05, "tok_s": 95658.72062795027}
run/pip_freeze.txt ADDED
@@ -0,0 +1,113 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ archspec @ file:///home/conda/feedstock_root/build_artifacts/archspec_1737352602016/work
2
+ asttokens @ file:///home/conda/feedstock_root/build_artifacts/asttokens_1733250440834/work
3
+ astunparse==1.6.3
4
+ attrs @ file:///home/conda/feedstock_root/build_artifacts/attrs_1741918516150/work
5
+ beautifulsoup4 @ file:///home/conda/feedstock_root/build_artifacts/beautifulsoup4_1744783198182/work
6
+ boltons @ file:///home/conda/feedstock_root/build_artifacts/boltons_1749686179973/work
7
+ Brotli @ file:///home/conda/feedstock_root/build_artifacts/brotli-split_1749229842835/work
8
+ certifi @ file:///home/conda/feedstock_root/build_artifacts/certifi_1754231422783/work/certifi
9
+ cffi @ file:///home/conda/feedstock_root/build_artifacts/cffi_1725560564262/work
10
+ chardet @ file:///home/conda/feedstock_root/build_artifacts/chardet_1741797914774/work
11
+ charset-normalizer @ file:///home/conda/feedstock_root/build_artifacts/charset-normalizer_1746214863626/work
12
+ click @ file:///home/conda/feedstock_root/build_artifacts/click_1747811314515/work
13
+ cmake==4.0.3
14
+ colorama @ file:///home/conda/feedstock_root/build_artifacts/colorama_1733218098505/work
15
+ conda @ file:///home/conda/feedstock_root/build_artifacts/conda_1754405241914/work/conda-src
16
+ conda-build @ file:///home/conda/feedstock_root/build_artifacts/conda-build_1754316273870/work
17
+ conda-libmamba-solver @ file:///home/conda/feedstock_root/build_artifacts/conda-libmamba-solver_1742219570693/work/src
18
+ conda-package-handling @ file:///home/conda/feedstock_root/build_artifacts/conda-package-handling_1736345463896/work
19
+ conda_index @ file:///home/conda/feedstock_root/build_artifacts/conda-index_1748375757308/work
20
+ conda_package_streaming @ file:///home/conda/feedstock_root/build_artifacts/conda-package-streaming_1751548120229/work
21
+ decorator @ file:///home/conda/feedstock_root/build_artifacts/decorator_1740384970518/work
22
+ distro @ file:///home/conda/feedstock_root/build_artifacts/distro_1734729835256/work
23
+ dnspython==2.7.0
24
+ evalidate @ file:///home/conda/feedstock_root/build_artifacts/bld/rattler-build_evalidate_1746793833/work
25
+ exceptiongroup @ file:///home/conda/feedstock_root/build_artifacts/exceptiongroup_1746947292760/work
26
+ executing @ file:///home/conda/feedstock_root/build_artifacts/executing_1745502089858/work
27
+ expecttest==0.3.0
28
+ filelock @ file:///home/conda/feedstock_root/build_artifacts/filelock_1741969488311/work
29
+ frozendict @ file:///home/conda/feedstock_root/build_artifacts/frozendict_1728841334936/work
30
+ fsspec==2025.7.0
31
+ h2 @ file:///home/conda/feedstock_root/build_artifacts/h2_1738578511449/work
32
+ hpack @ file:///home/conda/feedstock_root/build_artifacts/hpack_1737618293087/work
33
+ hyperframe @ file:///home/conda/feedstock_root/build_artifacts/hyperframe_1737618333194/work
34
+ hypothesis==6.137.1
35
+ idna @ file:///home/conda/feedstock_root/build_artifacts/idna_1733211830134/work
36
+ ipython @ file:///home/conda/feedstock_root/build_artifacts/bld/rattler-build_ipython_1751465044/work
37
+ ipython_pygments_lexers @ file:///home/conda/feedstock_root/build_artifacts/ipython_pygments_lexers_1737123620466/work
38
+ jedi @ file:///home/conda/feedstock_root/build_artifacts/jedi_1733300866624/work
39
+ Jinja2 @ file:///home/conda/feedstock_root/build_artifacts/jinja2_1741263328855/work
40
+ jsonpatch @ file:///home/conda/feedstock_root/build_artifacts/jsonpatch_1733814567314/work
41
+ jsonpointer @ file:///home/conda/feedstock_root/build_artifacts/jsonpointer_1725302941992/work
42
+ jsonschema @ file:///home/conda/feedstock_root/build_artifacts/bld/rattler-build_jsonschema_1752925388/work
43
+ jsonschema-specifications @ file:///tmp/tmpuvkyqc9y/src
44
+ libarchive-c @ file:///home/conda/feedstock_root/build_artifacts/bld/rattler-build_python-libarchive-c_1747927321/work
45
+ libmambapy @ file:///home/conda/feedstock_root/build_artifacts/mamba-split_1746515836725/work/libmambapy
46
+ lief @ file:///home/conda/feedstock_root/build_artifacts/lief_1750151383011/work/api/python
47
+ lintrunner==0.12.7
48
+ MarkupSafe @ file:///home/conda/feedstock_root/build_artifacts/markupsafe_1733219680183/work
49
+ matplotlib-inline @ file:///home/conda/feedstock_root/build_artifacts/matplotlib-inline_1733416936468/work
50
+ menuinst @ file:///home/conda/feedstock_root/build_artifacts/menuinst_1753546279984/work
51
+ mpmath==1.3.0
52
+ msgpack @ file:///home/conda/feedstock_root/build_artifacts/msgpack-python_1749813202382/work
53
+ networkx==3.5
54
+ ninja==1.11.1.4
55
+ numpy @ file:///home/conda/feedstock_root/build_artifacts/bld/rattler-build_numpy_1753401560/work/dist/numpy-2.3.2-cp311-cp311-linux_x86_64.whl#sha256=469440884ff65cdd5601d2ff9e1c5b9bcc2d8176c37e37ec78a0666703a54157
56
+ nvidia-cublas-cu12==12.8.4.1
57
+ nvidia-cuda-cupti-cu12==12.8.90
58
+ nvidia-cuda-nvrtc-cu12==12.8.93
59
+ nvidia-cuda-runtime-cu12==12.8.90
60
+ nvidia-cudnn-cu12==9.10.2.21
61
+ nvidia-cufft-cu12==11.3.3.83
62
+ nvidia-cufile-cu12==1.13.1.3
63
+ nvidia-curand-cu12==10.3.9.90
64
+ nvidia-cusolver-cu12==11.7.3.90
65
+ nvidia-cusparse-cu12==12.5.8.93
66
+ nvidia-cusparselt-cu12==0.7.1
67
+ nvidia-nccl-cu12==2.27.3
68
+ nvidia-nvjitlink-cu12==12.8.93
69
+ nvidia-nvtx-cu12==12.8.90
70
+ optree==0.17.0
71
+ packaging @ file:///home/conda/feedstock_root/build_artifacts/bld/rattler-build_packaging_1745345660/work
72
+ parso @ file:///home/conda/feedstock_root/build_artifacts/parso_1733271261340/work
73
+ pexpect @ file:///home/conda/feedstock_root/build_artifacts/pexpect_1733301927746/work
74
+ pickleshare @ file:///home/conda/feedstock_root/build_artifacts/pickleshare_1733327343728/work
75
+ pillow==11.0.0
76
+ pkginfo @ file:///home/conda/feedstock_root/build_artifacts/pkginfo_1739984581450/work
77
+ platformdirs @ file:///home/conda/feedstock_root/build_artifacts/bld/rattler-build_platformdirs_1746710438/work
78
+ pluggy @ file:///home/conda/feedstock_root/build_artifacts/pluggy_1747339660894/work
79
+ prompt_toolkit @ file:///home/conda/feedstock_root/build_artifacts/prompt-toolkit_1744724089886/work
80
+ psutil @ file:///home/conda/feedstock_root/build_artifacts/psutil_1740663149797/work
81
+ ptyprocess @ file:///home/conda/feedstock_root/build_artifacts/ptyprocess_1733302279685/work/dist/ptyprocess-0.7.0-py2.py3-none-any.whl#sha256=92c32ff62b5fd8cf325bec5ab90d7be3d2a8ca8c8a3813ff487a8d2002630d1f
82
+ pure_eval @ file:///home/conda/feedstock_root/build_artifacts/pure_eval_1733569405015/work
83
+ pycosat @ file:///home/conda/feedstock_root/build_artifacts/pycosat_1732588400443/work
84
+ pycparser @ file:///home/conda/feedstock_root/build_artifacts/bld/rattler-build_pycparser_1733195786/work
85
+ Pygments @ file:///home/conda/feedstock_root/build_artifacts/pygments_1750615794071/work
86
+ PySocks @ file:///home/conda/feedstock_root/build_artifacts/pysocks_1733217236728/work
87
+ python-etcd==0.4.5
88
+ pytz @ file:///home/conda/feedstock_root/build_artifacts/pytz_1742920838005/work
89
+ PyYAML @ file:///home/conda/feedstock_root/build_artifacts/pyyaml_1737454647378/work
90
+ referencing @ file:///home/conda/feedstock_root/build_artifacts/bld/rattler-build_referencing_1737836872/work
91
+ requests @ file:///home/conda/feedstock_root/build_artifacts/requests_1749498106507/work
92
+ rpds-py @ file:///home/conda/feedstock_root/build_artifacts/bld/rattler-build_rpds-py_1751468291/work
93
+ ruamel.yaml @ file:///home/conda/feedstock_root/build_artifacts/ruamel.yaml_1749479918291/work
94
+ ruamel.yaml.clib @ file:///home/conda/feedstock_root/build_artifacts/ruamel.yaml.clib_1728724459810/work
95
+ six==1.17.0
96
+ sortedcontainers==2.4.0
97
+ soupsieve @ file:///home/conda/feedstock_root/build_artifacts/soupsieve_1746563585861/work
98
+ stack_data @ file:///home/conda/feedstock_root/build_artifacts/stack_data_1733569443808/work
99
+ sympy==1.14.0
100
+ tokenizers==0.23.2
101
+ torch==2.8.0+cu128
102
+ torchaudio==2.8.0+cu128
103
+ torchelastic==0.2.2
104
+ torchvision==0.23.0+cu128
105
+ tqdm @ file:///home/conda/feedstock_root/build_artifacts/tqdm_1735661334605/work
106
+ traitlets @ file:///home/conda/feedstock_root/build_artifacts/traitlets_1733367359838/work
107
+ triton==3.4.0
108
+ truststore @ file:///home/conda/feedstock_root/build_artifacts/bld/rattler-build_truststore_1739009763/work
109
+ types-dataclasses==0.6.6
110
+ typing_extensions @ file:///home/conda/feedstock_root/build_artifacts/bld/rattler-build_typing_extensions_1751643513/work
111
+ urllib3 @ file:///home/conda/feedstock_root/build_artifacts/urllib3_1750271362675/work
112
+ wcwidth @ file:///home/conda/feedstock_root/build_artifacts/wcwidth_1733231326287/work
113
+ zstandard==0.23.0
run/train.log ADDED
@@ -0,0 +1,210 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ DOUCZ-SMOKE-OK 2026-10-23T13:55:52Z BATCH-SHA krok biegu 1051 = krok strumienia 1051: ee67ca8d5d4ca05831b84e5852c88c6f534f68c1d4c2f2d57305d1ee5aaabc5a BATCH-SHA krok biegu 1800 = krok strumienia 1800: f8efcf575e2700045a50a401ff3ea9a99636196ec1595030a93c34f71b66c801 BATCH-SHA krok biegu 1801 = krok strumienia 1801: f4e091e8441771d06191822b63d2ae18186b27d93829445d938ddb05226ff176
2
+ DOUCZ start 2026-10-23T13:55:52Z z /root/pilot1b/start/ckpt_1050.pt sha 236108e8f204995f4adcfea7d9a99437d230c1af12708c64c68c74afbc2910fd steps 2860 switch 1801
3
+ === doucz 2026-10-23T13:55:52Z steps 2860 ckpt_every 150
4
+ [13:55:57] device=cuda:0 bf16=True dev=NVIDIA GeForce RTX 5090 world=8
5
+ [13:55:57] dane: train=1,889,519,627B block=4096 batch=256 tok/step=1,048,576
6
+ [13:56:10] model GPT-ref: 1073.9M param (L20 d2048 h16)
7
+ [13:56:10] optimizer=muon muon_lr=0.02 (mult=33.3x)
8
+ [13:56:10] r11 pamiec: act_ckpt=True loss_chunk=1024
9
+ [13:56:10] r13: mixer_layout=AAAAAAAAAAAAAAAAAAAA gdn_head_dim=64 attn_bias=0 rope_style=interleaved
10
+ [13:56:13] torch.compile enabled
11
+ [13:56:21] RESUME @ 1050
12
+ [13:56:21] START gpt-ref: steps=2860 (od 1050) lr=0.0006->6e-05
13
+ [13:56:21] sampler epoch: 461,307 okien/epoke (offset0 2245), start od kroku 1050 = epoka 0.583
14
+ [13:56:21] data-dir-2: /root/pilot1b/data_p01 1,880,520,445 tok, 459,110 okien/epoke = 1793 krokow, od kroku strumienia 1801
15
+ [13:59:53] step 1060/2860 loss 2.8356 lr 6.00e-04 gnorm 0.10 49,515 tok/s peakVRAM 27.0GB
16
+ [14:01:29] step 1070/2860 loss 2.8687 lr 6.00e-04 gnorm 0.10 109,240 tok/s peakVRAM 27.0GB
17
+ [14:03:05] step 1080/2860 loss 2.8498 lr 6.00e-04 gnorm 0.10 108,731 tok/s peakVRAM 27.0GB
18
+ [14:04:42] step 1090/2860 loss 2.8553 lr 6.00e-04 gnorm 0.11 108,248 tok/s peakVRAM 27.0GB
19
+ [14:06:19] step 1100/2860 loss 2.8493 lr 6.00e-04 gnorm 0.09 107,871 tok/s peakVRAM 27.0GB
20
+ [14:07:57] step 1110/2860 loss 2.8485 lr 6.00e-04 gnorm 0.09 107,630 tok/s peakVRAM 27.0GB
21
+ [14:09:34] step 1120/2860 loss 2.8432 lr 6.00e-04 gnorm 0.09 108,072 tok/s peakVRAM 27.0GB
22
+ [14:11:11] step 1130/2860 loss 2.8414 lr 6.00e-04 gnorm 0.09 107,987 tok/s peakVRAM 27.0GB
23
+ [14:12:48] step 1140/2860 loss 2.8192 lr 6.00e-04 gnorm 0.10 107,987 tok/s peakVRAM 27.0GB
24
+ [14:14:25] step 1150/2860 loss 2.8366 lr 6.00e-04 gnorm 0.11 107,797 tok/s peakVRAM 27.0GB
25
+ [14:16:02] step 1160/2860 loss 2.8427 lr 6.00e-04 gnorm 0.11 108,156 tok/s peakVRAM 27.0GB
26
+ [14:17:39] step 1170/2860 loss 2.8140 lr 6.00e-04 gnorm 0.10 108,311 tok/s peakVRAM 27.0GB
27
+ [14:19:16] step 1180/2860 loss 2.8157 lr 6.00e-04 gnorm 0.10 108,201 tok/s peakVRAM 27.0GB
28
+ [14:20:53] step 1190/2860 loss 2.8080 lr 6.00e-04 gnorm 0.11 107,802 tok/s peakVRAM 27.0GB
29
+ [14:22:31] step 1200/2860 loss 2.8378 lr 6.00e-04 gnorm 0.11 107,815 tok/s peakVRAM 27.0GB
30
+ [14:22:40] ckpt @ 1200
31
+ [14:24:17] step 1210/2860 loss 2.8056 lr 6.00e-04 gnorm 0.10 98,763 tok/s peakVRAM 27.0GB
32
+ [14:25:54] step 1220/2860 loss 2.8146 lr 6.00e-04 gnorm 0.10 108,307 tok/s peakVRAM 27.0GB
33
+ [14:27:31] step 1230/2860 loss 2.8082 lr 6.00e-04 gnorm 0.09 108,132 tok/s peakVRAM 27.0GB
34
+ [14:29:08] step 1240/2860 loss 2.8096 lr 6.00e-04 gnorm 0.10 107,957 tok/s peakVRAM 27.0GB
35
+ [14:30:45] step 1250/2860 loss 2.8166 lr 6.00e-04 gnorm 0.11 108,114 tok/s peakVRAM 27.0GB
36
+ [14:32:22] step 1260/2860 loss 2.8128 lr 6.00e-04 gnorm 0.10 108,061 tok/s peakVRAM 27.0GB
37
+ [14:33:59] step 1270/2860 loss 2.7992 lr 6.00e-04 gnorm 0.09 108,210 tok/s peakVRAM 27.0GB
38
+ [14:35:36] step 1280/2860 loss 2.7989 lr 6.00e-04 gnorm 0.10 108,128 tok/s peakVRAM 27.0GB
39
+ [14:37:13] step 1290/2860 loss 2.7926 lr 6.00e-04 gnorm 0.12 107,744 tok/s peakVRAM 27.0GB
40
+ [14:38:50] step 1300/2860 loss 2.7575 lr 6.00e-04 gnorm 0.10 107,733 tok/s peakVRAM 27.0GB
41
+ [14:40:28] step 1310/2860 loss 2.7676 lr 6.00e-04 gnorm 0.09 107,802 tok/s peakVRAM 27.0GB
42
+ [14:42:04] step 1320/2860 loss 2.7761 lr 6.00e-04 gnorm 0.09 108,185 tok/s peakVRAM 27.0GB
43
+ [14:43:42] step 1330/2860 loss 2.7931 lr 6.00e-04 gnorm 0.09 107,999 tok/s peakVRAM 27.0GB
44
+ [14:45:19] step 1340/2860 loss 2.7885 lr 6.00e-04 gnorm 0.09 107,917 tok/s peakVRAM 27.0GB
45
+ [14:46:56] step 1350/2860 loss 2.7733 lr 6.00e-04 gnorm 0.08 108,010 tok/s peakVRAM 27.0GB
46
+ [14:47:05] ckpt @ 1350
47
+ [14:48:42] step 1360/2860 loss 2.7871 lr 6.00e-04 gnorm 0.10 98,694 tok/s peakVRAM 27.0GB
48
+ [14:50:19] step 1370/2860 loss 2.7689 lr 6.00e-04 gnorm 0.09 108,306 tok/s peakVRAM 27.0GB
49
+ [14:51:56] step 1380/2860 loss 2.7655 lr 6.00e-04 gnorm 0.09 107,983 tok/s peakVRAM 27.0GB
50
+ [14:53:33] step 1390/2860 loss 2.7674 lr 6.00e-04 gnorm 0.10 107,907 tok/s peakVRAM 27.0GB
51
+ [14:55:10] step 1400/2860 loss 2.7652 lr 6.00e-04 gnorm 0.09 107,833 tok/s peakVRAM 27.0GB
52
+ [14:56:48] step 1410/2860 loss 2.7603 lr 6.00e-04 gnorm 0.09 107,875 tok/s peakVRAM 27.0GB
53
+ [14:58:25] step 1420/2860 loss 2.7825 lr 6.00e-04 gnorm 0.10 107,990 tok/s peakVRAM 27.0GB
54
+ [15:00:02] step 1430/2860 loss 2.7738 lr 6.00e-04 gnorm 0.08 108,140 tok/s peakVRAM 27.0GB
55
+ [15:01:39] step 1440/2860 loss 2.7555 lr 6.00e-04 gnorm 0.09 108,163 tok/s peakVRAM 27.0GB
56
+ [15:03:15] step 1450/2860 loss 2.7610 lr 6.00e-04 gnorm 0.09 108,188 tok/s peakVRAM 27.0GB
57
+ [15:04:52] step 1460/2860 loss 2.7413 lr 6.00e-04 gnorm 0.09 108,135 tok/s peakVRAM 27.0GB
58
+ [15:06:29] step 1470/2860 loss 2.7548 lr 6.00e-04 gnorm 0.09 108,067 tok/s peakVRAM 27.0GB
59
+ [15:08:06] step 1480/2860 loss 2.7579 lr 6.00e-04 gnorm 0.10 108,194 tok/s peakVRAM 27.0GB
60
+ [15:09:44] step 1490/2860 loss 2.7474 lr 6.00e-04 gnorm 0.10 107,963 tok/s peakVRAM 27.0GB
61
+ [15:11:21] step 1500/2860 loss 2.7388 lr 6.00e-04 gnorm 0.10 107,926 tok/s peakVRAM 27.0GB
62
+ [15:11:30] ckpt @ 1500
63
+ [15:13:07] step 1510/2860 loss 2.7643 lr 6.00e-04 gnorm 0.10 98,671 tok/s peakVRAM 27.0GB
64
+ [15:14:44] step 1520/2860 loss 2.7517 lr 6.00e-04 gnorm 0.09 108,297 tok/s peakVRAM 27.0GB
65
+ [15:16:21] step 1530/2860 loss 2.7416 lr 6.00e-04 gnorm 0.08 108,167 tok/s peakVRAM 27.0GB
66
+ [15:17:58] step 1540/2860 loss 2.7395 lr 6.00e-04 gnorm 0.10 108,347 tok/s peakVRAM 27.0GB
67
+ [15:19:34] step 1550/2860 loss 2.7480 lr 6.00e-04 gnorm 0.09 108,358 tok/s peakVRAM 27.0GB
68
+ [15:21:11] step 1560/2860 loss 2.7244 lr 6.00e-04 gnorm 0.08 108,232 tok/s peakVRAM 27.0GB
69
+ [15:22:48] step 1570/2860 loss 2.7390 lr 6.00e-04 gnorm 0.09 108,028 tok/s peakVRAM 27.0GB
70
+ [15:24:25] step 1580/2860 loss 2.7299 lr 6.00e-04 gnorm 0.08 108,056 tok/s peakVRAM 27.0GB
71
+ [15:26:02] step 1590/2860 loss 2.7523 lr 6.00e-04 gnorm 0.10 108,021 tok/s peakVRAM 27.0GB
72
+ [15:27:39] step 1600/2860 loss 2.7405 lr 6.00e-04 gnorm 0.08 108,154 tok/s peakVRAM 27.0GB
73
+ [15:29:16] step 1610/2860 loss 2.7168 lr 6.00e-04 gnorm 0.08 107,893 tok/s peakVRAM 27.0GB
74
+ [15:30:54] step 1620/2860 loss 2.7081 lr 6.00e-04 gnorm 0.09 107,857 tok/s peakVRAM 27.0GB
75
+ [15:32:31] step 1630/2860 loss 2.7200 lr 6.00e-04 gnorm 0.09 107,659 tok/s peakVRAM 27.0GB
76
+ [15:34:08] step 1640/2860 loss 2.7136 lr 6.00e-04 gnorm 0.16 107,823 tok/s peakVRAM 27.0GB
77
+ [15:35:46] step 1650/2860 loss 2.7406 lr 6.00e-04 gnorm 0.09 107,779 tok/s peakVRAM 27.0GB
78
+ [15:35:56] ckpt @ 1650
79
+ [15:37:33] step 1660/2860 loss 2.7050 lr 6.00e-04 gnorm 0.09 98,087 tok/s peakVRAM 27.0GB
80
+ [15:39:10] step 1670/2860 loss 2.7108 lr 6.00e-04 gnorm 0.10 108,102 tok/s peakVRAM 27.0GB
81
+ [15:40:47] step 1680/2860 loss 2.7259 lr 6.00e-04 gnorm 0.10 107,897 tok/s peakVRAM 27.0GB
82
+ [15:42:24] step 1690/2860 loss 2.7016 lr 6.00e-04 gnorm 0.10 107,780 tok/s peakVRAM 27.0GB
83
+ [15:44:01] step 1700/2860 loss 2.7218 lr 6.00e-04 gnorm 0.09 107,921 tok/s peakVRAM 27.0GB
84
+ [15:45:38] step 1710/2860 loss 2.7047 lr 6.00e-04 gnorm 0.09 107,869 tok/s peakVRAM 27.0GB
85
+ [15:47:16] step 1720/2860 loss 2.7218 lr 6.00e-04 gnorm 0.10 107,872 tok/s peakVRAM 27.0GB
86
+ [15:48:53] step 1730/2860 loss 2.7104 lr 6.00e-04 gnorm 0.10 107,897 tok/s peakVRAM 27.0GB
87
+ [15:50:30] step 1740/2860 loss 2.6869 lr 6.00e-04 gnorm 0.08 107,874 tok/s peakVRAM 27.0GB
88
+ [15:52:07] step 1750/2860 loss 2.6925 lr 6.00e-04 gnorm 0.10 107,969 tok/s peakVRAM 27.0GB
89
+ [15:53:44] step 1760/2860 loss 2.6851 lr 6.00e-04 gnorm 0.10 107,963 tok/s peakVRAM 27.0GB
90
+ [15:55:21] step 1770/2860 loss 2.6856 lr 6.00e-04 gnorm 0.08 107,892 tok/s peakVRAM 27.0GB
91
+ [15:56:59] step 1780/2860 loss 2.6989 lr 6.00e-04 gnorm 0.10 107,713 tok/s peakVRAM 27.0GB
92
+ [15:58:36] step 1790/2860 loss 2.6702 lr 6.00e-04 gnorm 0.08 107,620 tok/s peakVRAM 27.0GB
93
+ [16:00:14] step 1800/2860 loss 2.6972 lr 6.00e-04 gnorm 0.09 107,335 tok/s peakVRAM 27.0GB
94
+ [16:00:24] ckpt @ 1800
95
+ [16:02:01] step 1810/2860 loss 2.6813 lr 6.00e-04 gnorm 0.09 97,749 tok/s peakVRAM 27.0GB
96
+ [16:03:38] step 1820/2860 loss 2.6996 lr 6.00e-04 gnorm 0.10 108,119 tok/s peakVRAM 27.0GB
97
+ [16:05:15] step 1830/2860 loss 2.7000 lr 6.00e-04 gnorm 0.08 107,858 tok/s peakVRAM 27.0GB
98
+ [16:06:52] step 1840/2860 loss 2.6922 lr 6.00e-04 gnorm 0.08 108,092 tok/s peakVRAM 27.0GB
99
+ [16:08:29] step 1850/2860 loss 2.6933 lr 6.00e-04 gnorm 0.08 108,196 tok/s peakVRAM 27.0GB
100
+ [16:10:06] step 1860/2860 loss 2.7058 lr 6.00e-04 gnorm 0.08 108,083 tok/s peakVRAM 27.0GB
101
+ [16:11:43] step 1870/2860 loss 2.6963 lr 6.00e-04 gnorm 0.08 108,178 tok/s peakVRAM 27.0GB
102
+ [16:13:20] step 1880/2860 loss 2.6996 lr 6.00e-04 gnorm 0.10 108,173 tok/s peakVRAM 27.0GB
103
+ [16:14:57] step 1890/2860 loss 2.6934 lr 6.00e-04 gnorm 0.09 107,941 tok/s peakVRAM 27.0GB
104
+ [16:16:34] step 1900/2860 loss 2.6880 lr 6.00e-04 gnorm 0.09 108,129 tok/s peakVRAM 27.0GB
105
+ [16:18:11] step 1910/2860 loss 2.6927 lr 6.00e-04 gnorm 0.09 108,021 tok/s peakVRAM 27.0GB
106
+ [16:19:49] step 1920/2860 loss 2.6745 lr 6.00e-04 gnorm 0.09 107,920 tok/s peakVRAM 27.0GB
107
+ [16:21:26] step 1930/2860 loss 2.6718 lr 6.00e-04 gnorm 0.09 107,949 tok/s peakVRAM 27.0GB
108
+ [16:23:03] step 1940/2860 loss 2.6847 lr 6.00e-04 gnorm 0.09 108,210 tok/s peakVRAM 27.0GB
109
+ [16:24:40] step 1950/2860 loss 2.6858 lr 6.00e-04 gnorm 0.08 107,953 tok/s peakVRAM 27.0GB
110
+ [16:24:50] ckpt @ 1950
111
+ [16:26:27] step 1960/2860 loss 2.6984 lr 6.00e-04 gnorm 0.09 97,835 tok/s peakVRAM 27.0GB
112
+ [16:28:04] step 1970/2860 loss 2.6685 lr 6.00e-04 gnorm 0.09 107,868 tok/s peakVRAM 27.0GB
113
+ [16:29:41] step 1980/2860 loss 2.6823 lr 6.00e-04 gnorm 0.09 107,690 tok/s peakVRAM 27.0GB
114
+ [16:31:19] step 1990/2860 loss 2.6722 lr 6.00e-04 gnorm 0.08 107,691 tok/s peakVRAM 27.0GB
115
+ [16:32:56] step 2000/2860 loss 2.6615 lr 6.00e-04 gnorm 0.09 107,819 tok/s peakVRAM 27.0GB
116
+ [16:34:33] step 2010/2860 loss 2.6696 lr 6.00e-04 gnorm 0.09 108,003 tok/s peakVRAM 27.0GB
117
+ [16:36:10] step 2020/2860 loss 2.6537 lr 6.00e-04 gnorm 0.08 107,785 tok/s peakVRAM 27.0GB
118
+ [16:37:48] step 2030/2860 loss 2.6670 lr 6.00e-04 gnorm 0.09 107,549 tok/s peakVRAM 27.0GB
119
+ [16:39:25] step 2040/2860 loss 2.7043 lr 6.00e-04 gnorm 0.09 107,558 tok/s peakVRAM 27.0GB
120
+ [16:41:03] step 2050/2860 loss 2.6757 lr 6.00e-04 gnorm 0.08 107,565 tok/s peakVRAM 27.0GB
121
+ [16:42:40] step 2060/2860 loss 2.6730 lr 6.00e-04 gnorm 0.09 107,805 tok/s peakVRAM 27.0GB
122
+ [16:44:17] step 2070/2860 loss 2.6598 lr 6.00e-04 gnorm 0.08 107,969 tok/s peakVRAM 27.0GB
123
+ [16:45:54] step 2080/2860 loss 2.6709 lr 6.00e-04 gnorm 0.10 108,008 tok/s peakVRAM 27.0GB
124
+ [16:47:32] step 2090/2860 loss 2.6693 lr 6.00e-04 gnorm 0.10 107,844 tok/s peakVRAM 27.0GB
125
+ [16:49:09] step 2100/2860 loss 2.6343 lr 6.00e-04 gnorm 0.08 107,862 tok/s peakVRAM 27.0GB
126
+ [16:49:20] ckpt @ 2100
127
+ [16:50:56] step 2110/2860 loss 2.6644 lr 6.00e-04 gnorm 0.10 97,559 tok/s peakVRAM 27.0GB
128
+ [16:52:34] step 2120/2860 loss 2.6637 lr 6.00e-04 gnorm 0.07 107,791 tok/s peakVRAM 27.0GB
129
+ [16:54:11] step 2130/2860 loss 2.6517 lr 6.00e-04 gnorm 0.09 107,569 tok/s peakVRAM 27.0GB
130
+ [16:55:48] step 2140/2860 loss 2.6472 lr 6.00e-04 gnorm 0.08 107,636 tok/s peakVRAM 27.0GB
131
+ [16:57:26] step 2150/2860 loss 2.6531 lr 6.00e-04 gnorm 0.08 107,873 tok/s peakVRAM 27.0GB
132
+ [16:59:03] step 2160/2860 loss 2.6535 lr 6.00e-04 gnorm 0.08 107,675 tok/s peakVRAM 27.0GB
133
+ [17:00:41] step 2170/2860 loss 2.6566 lr 6.00e-04 gnorm 0.08 107,485 tok/s peakVRAM 27.0GB
134
+ [17:02:18] step 2180/2860 loss 2.6688 lr 6.00e-04 gnorm 0.09 107,483 tok/s peakVRAM 27.0GB
135
+ [17:03:56] step 2190/2860 loss 2.6450 lr 6.00e-04 gnorm 0.08 107,499 tok/s peakVRAM 27.0GB
136
+ [17:05:33] step 2200/2860 loss 2.6498 lr 6.00e-04 gnorm 0.09 107,383 tok/s peakVRAM 27.0GB
137
+ [17:07:11] step 2210/2860 loss 2.6386 lr 6.00e-04 gnorm 0.09 107,588 tok/s peakVRAM 27.0GB
138
+ [17:08:48] step 2220/2860 loss 2.6283 lr 6.00e-04 gnorm 0.09 107,430 tok/s peakVRAM 27.0GB
139
+ [17:10:26] step 2230/2860 loss 2.6434 lr 6.00e-04 gnorm 0.07 107,231 tok/s peakVRAM 27.0GB
140
+ [17:12:04] step 2240/2860 loss 2.6654 lr 6.00e-04 gnorm 0.08 107,516 tok/s peakVRAM 27.0GB
141
+ [17:13:41] step 2250/2860 loss 2.6318 lr 6.00e-04 gnorm 0.08 107,529 tok/s peakVRAM 27.0GB
142
+ [17:13:53] ckpt @ 2250
143
+ [17:15:30] step 2260/2860 loss 2.6379 lr 6.00e-04 gnorm 0.10 96,309 tok/s peakVRAM 27.0GB
144
+ [17:17:07] step 2270/2860 loss 2.6461 lr 6.00e-04 gnorm 0.08 107,863 tok/s peakVRAM 27.0GB
145
+ [17:18:45] step 2280/2860 loss 2.6532 lr 6.00e-04 gnorm 0.08 107,509 tok/s peakVRAM 27.0GB
146
+ [17:20:22] step 2290/2860 loss 2.6472 lr 5.77e-04 gnorm 0.08 107,551 tok/s peakVRAM 27.0GB
147
+ [17:22:00] step 2300/2860 loss 2.6419 lr 5.25e-04 gnorm 0.07 107,791 tok/s peakVRAM 27.0GB
148
+ [17:23:37] step 2310/2860 loss 2.6313 lr 4.97e-04 gnorm 0.07 107,474 tok/s peakVRAM 27.0GB
149
+ [17:25:15] step 2320/2860 loss 2.6209 lr 4.74e-04 gnorm 0.06 107,316 tok/s peakVRAM 27.0GB
150
+ [17:26:53] step 2330/2860 loss 2.6298 lr 4.55e-04 gnorm 0.07 107,306 tok/s peakVRAM 27.0GB
151
+ [17:28:30] step 2340/2860 loss 2.6196 lr 4.39e-04 gnorm 0.07 107,440 tok/s peakVRAM 27.0GB
152
+ [17:30:08] step 2350/2860 loss 2.6067 lr 4.24e-04 gnorm 0.06 107,241 tok/s peakVRAM 27.0GB
153
+ [17:31:46] step 2360/2860 loss 2.5964 lr 4.10e-04 gnorm 0.07 106,925 tok/s peakVRAM 27.0GB
154
+ [17:33:24] step 2370/2860 loss 2.5960 lr 3.97e-04 gnorm 0.06 107,135 tok/s peakVRAM 27.0GB
155
+ [17:35:02] step 2380/2860 loss 2.5998 lr 3.85e-04 gnorm 0.06 107,354 tok/s peakVRAM 27.0GB
156
+ [17:36:39] step 2390/2860 loss 2.5892 lr 3.73e-04 gnorm 0.06 107,686 tok/s peakVRAM 27.0GB
157
+ [17:38:16] step 2400/2860 loss 2.6009 lr 3.62e-04 gnorm 0.05 107,816 tok/s peakVRAM 27.0GB
158
+ [17:38:29] ckpt @ 2400
159
+ [17:40:06] step 2410/2860 loss 2.6035 lr 3.52e-04 gnorm 0.06 95,863 tok/s peakVRAM 27.0GB
160
+ [17:41:43] step 2420/2860 loss 2.5908 lr 3.42e-04 gnorm 0.06 107,941 tok/s peakVRAM 27.0GB
161
+ [17:43:20] step 2430/2860 loss 2.5813 lr 3.32e-04 gnorm 0.05 107,513 tok/s peakVRAM 27.0GB
162
+ [17:44:58] step 2440/2860 loss 2.5774 lr 3.23e-04 gnorm 0.06 107,573 tok/s peakVRAM 27.0GB
163
+ [17:46:35] step 2450/2860 loss 2.5865 lr 3.14e-04 gnorm 0.06 107,692 tok/s peakVRAM 27.0GB
164
+ [17:48:13] step 2460/2860 loss 2.5850 lr 3.05e-04 gnorm 0.05 107,579 tok/s peakVRAM 27.0GB
165
+ [17:49:50] step 2470/2860 loss 2.5728 lr 2.96e-04 gnorm 0.05 107,413 tok/s peakVRAM 27.0GB
166
+ [17:51:28] step 2480/2860 loss 2.5804 lr 2.88e-04 gnorm 0.05 107,500 tok/s peakVRAM 27.0GB
167
+ [17:53:05] step 2490/2860 loss 2.5865 lr 2.80e-04 gnorm 0.06 107,691 tok/s peakVRAM 27.0GB
168
+ [17:54:43] step 2500/2860 loss 2.5757 lr 2.72e-04 gnorm 0.05 107,346 tok/s peakVRAM 27.0GB
169
+ [17:56:20] step 2510/2860 loss 2.5683 lr 2.64e-04 gnorm 0.05 107,447 tok/s peakVRAM 27.0GB
170
+ [17:57:58] step 2520/2860 loss 2.5659 lr 2.57e-04 gnorm 0.05 107,256 tok/s peakVRAM 27.0GB
171
+ [17:59:36] step 2530/2860 loss 2.5493 lr 2.49e-04 gnorm 0.05 107,326 tok/s peakVRAM 27.0GB
172
+ [18:01:13] step 2540/2860 loss 2.5526 lr 2.42e-04 gnorm 0.05 107,522 tok/s peakVRAM 27.0GB
173
+ [18:02:51] step 2550/2860 loss 2.5513 lr 2.35e-04 gnorm 0.05 107,402 tok/s peakVRAM 27.0GB
174
+ [18:03:04] ckpt @ 2550
175
+ [18:04:41] step 2560/2860 loss 2.5376 lr 2.28e-04 gnorm 0.05 95,800 tok/s peakVRAM 27.0GB
176
+ [18:06:18] step 2570/2860 loss 2.5497 lr 2.22e-04 gnorm 0.05 107,579 tok/s peakVRAM 27.0GB
177
+ [18:07:56] step 2580/2860 loss 2.5446 lr 2.15e-04 gnorm 0.05 107,307 tok/s peakVRAM 27.0GB
178
+ [18:09:34] step 2590/2860 loss 2.5501 lr 2.08e-04 gnorm 0.05 107,222 tok/s peakVRAM 27.0GB
179
+ [18:11:11] step 2600/2860 loss 2.5475 lr 2.02e-04 gnorm 0.05 107,513 tok/s peakVRAM 27.0GB
180
+ [18:12:49] step 2610/2860 loss 2.5570 lr 1.95e-04 gnorm 0.05 107,489 tok/s peakVRAM 27.0GB
181
+ [18:14:26] step 2620/2860 loss 2.5378 lr 1.89e-04 gnorm 0.05 107,509 tok/s peakVRAM 27.0GB
182
+ [18:16:04] step 2630/2860 loss 2.5365 lr 1.83e-04 gnorm 0.05 107,670 tok/s peakVRAM 27.0GB
183
+ [18:17:41] step 2640/2860 loss 2.5356 lr 1.77e-04 gnorm 0.05 107,664 tok/s peakVRAM 27.0GB
184
+ [18:19:19] step 2650/2860 loss 2.5567 lr 1.71e-04 gnorm 0.05 107,420 tok/s peakVRAM 27.0GB
185
+ [18:20:56] step 2660/2860 loss 2.5153 lr 1.65e-04 gnorm 0.05 107,311 tok/s peakVRAM 27.0GB
186
+ [18:22:34] step 2670/2860 loss 2.5341 lr 1.59e-04 gnorm 0.05 107,136 tok/s peakVRAM 27.0GB
187
+ [18:24:12] step 2680/2860 loss 2.5276 lr 1.54e-04 gnorm 0.05 107,568 tok/s peakVRAM 27.0GB
188
+ [18:25:49] step 2690/2860 loss 2.5233 lr 1.48e-04 gnorm 0.05 107,606 tok/s peakVRAM 27.0GB
189
+ [18:27:27] step 2700/2860 loss 2.5278 lr 1.42e-04 gnorm 0.05 107,163 tok/s peakVRAM 27.0GB
190
+ [18:27:37] ckpt @ 2700
191
+ [18:29:15] step 2710/2860 loss 2.5311 lr 1.37e-04 gnorm 0.05 97,103 tok/s peakVRAM 27.0GB
192
+ [18:30:53] step 2720/2860 loss 2.5200 lr 1.31e-04 gnorm 0.05 107,242 tok/s peakVRAM 27.0GB
193
+ [18:32:30] step 2730/2860 loss 2.5417 lr 1.26e-04 gnorm 0.05 107,248 tok/s peakVRAM 27.0GB
194
+ [18:34:08] step 2740/2860 loss 2.5239 lr 1.21e-04 gnorm 0.05 107,314 tok/s peakVRAM 27.0GB
195
+ [18:35:46] step 2750/2860 loss 2.5107 lr 1.15e-04 gnorm 0.05 107,318 tok/s peakVRAM 27.0GB
196
+ [18:37:24] step 2760/2860 loss 2.4960 lr 1.10e-04 gnorm 0.05 107,012 tok/s peakVRAM 27.0GB
197
+ [18:39:02] step 2770/2860 loss 2.5277 lr 1.05e-04 gnorm 0.05 106,943 tok/s peakVRAM 27.0GB
198
+ [18:40:40] step 2780/2860 loss 2.5243 lr 9.97e-05 gnorm 0.05 107,074 tok/s peakVRAM 27.0GB
199
+ [18:42:18] step 2790/2860 loss 2.5119 lr 9.46e-05 gnorm 0.05 107,164 tok/s peakVRAM 27.0GB
200
+ [18:43:55] step 2800/2860 loss 2.5066 lr 8.96e-05 gnorm 0.05 107,364 tok/s peakVRAM 27.0GB
201
+ [18:45:33] step 2810/2860 loss 2.5089 lr 8.46e-05 gnorm 0.05 107,309 tok/s peakVRAM 27.0GB
202
+ [18:47:11] step 2820/2860 loss 2.5087 lr 7.97e-05 gnorm 0.05 107,408 tok/s peakVRAM 27.0GB
203
+ [18:48:48] step 2830/2860 loss 2.4930 lr 7.48e-05 gnorm 0.05 107,398 tok/s peakVRAM 27.0GB
204
+ [18:50:26] step 2840/2860 loss 2.5122 lr 7.00e-05 gnorm 0.05 107,539 tok/s peakVRAM 27.0GB
205
+ [18:52:03] step 2850/2860 loss 2.4829 lr 6.52e-05 gnorm 0.05 107,477 tok/s peakVRAM 27.0GB
206
+ [18:52:16] ckpt @ 2850
207
+ [18:53:53] step 2860/2860 loss 2.5067 lr 6.05e-05 gnorm 0.05 95,659 tok/s peakVRAM 27.0GB
208
+ [18:54:06] ckpt @ 2860
209
+ [18:54:06] DONE
210
+ TRAIN-DONE 2026-10-23T18:54:52Z
run/train.out ADDED
@@ -0,0 +1,214 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ [13:55:57] device=cuda:0 bf16=True dev=NVIDIA GeForce RTX 5090 world=8
2
+ [13:55:57] dane: train=1,889,519,627B block=4096 batch=256 tok/step=1,048,576
3
+ [13:56:10] model GPT-ref: 1073.9M param (L20 d2048 h16)
4
+ [13:56:10] optimizer=muon muon_lr=0.02 (mult=33.3x)
5
+ [13:56:10] r11 pamiec: act_ckpt=True loss_chunk=1024
6
+ [13:56:10] r13: mixer_layout=AAAAAAAAAAAAAAAAAAAA gdn_head_dim=64 attn_bias=0 rope_style=interleaved
7
+ [13:56:13] torch.compile enabled
8
+ [rank6]:W1023 13:56:20.301000 6145 site-packages/torch/_logging/_internal.py:1154] [0/0] Profiler function <class 'torch.autograd.profiler.record_function'> will be ignored
9
+ [rank7]:W1023 13:56:20.302000 6146 site-packages/torch/_logging/_internal.py:1154] [0/0] Profiler function <class 'torch.autograd.profiler.record_function'> will be ignored
10
+ [rank4]:W1023 13:56:20.302000 6143 site-packages/torch/_logging/_internal.py:1154] [0/0] Profiler function <class 'torch.autograd.profiler.record_function'> will be ignored
11
+ [13:56:21] RESUME @ 1050
12
+ [13:56:21] START gpt-ref: steps=2860 (od 1050) lr=0.0006->6e-05
13
+ [13:56:21] sampler epoch: 461,307 okien/epoke (offset0 2245), start od kroku 1050 = epoka 0.583
14
+ [13:56:21] data-dir-2: /root/pilot1b/data_p01 1,880,520,445 tok, 459,110 okien/epoke = 1793 krokow, od kroku strumienia 1801
15
+ [rank5]:W1023 13:56:21.716000 6144 site-packages/torch/_logging/_internal.py:1154] [0/0] Profiler function <class 'torch.autograd.profiler.record_function'> will be ignored
16
+ [rank3]:W1023 13:56:21.716000 6142 site-packages/torch/_logging/_internal.py:1154] [0/0] Profiler function <class 'torch.autograd.profiler.record_function'> will be ignored
17
+ [rank0]:W1023 13:56:21.720000 6139 site-packages/torch/_logging/_internal.py:1154] [0/0] Profiler function <class 'torch.autograd.profiler.record_function'> will be ignored
18
+ [rank2]:W1023 13:56:21.720000 6141 site-packages/torch/_logging/_internal.py:1154] [0/0] Profiler function <class 'torch.autograd.profiler.record_function'> will be ignored
19
+ [rank1]:W1023 13:56:21.721000 6140 site-packages/torch/_logging/_internal.py:1154] [0/0] Profiler function <class 'torch.autograd.profiler.record_function'> will be ignored
20
+ [13:59:53] step 1060/2860 loss 2.8356 lr 6.00e-04 gnorm 0.10 49,515 tok/s peakVRAM 27.0GB
21
+ [14:01:29] step 1070/2860 loss 2.8687 lr 6.00e-04 gnorm 0.10 109,240 tok/s peakVRAM 27.0GB
22
+ [14:03:05] step 1080/2860 loss 2.8498 lr 6.00e-04 gnorm 0.10 108,731 tok/s peakVRAM 27.0GB
23
+ [14:04:42] step 1090/2860 loss 2.8553 lr 6.00e-04 gnorm 0.11 108,248 tok/s peakVRAM 27.0GB
24
+ [14:06:19] step 1100/2860 loss 2.8493 lr 6.00e-04 gnorm 0.09 107,871 tok/s peakVRAM 27.0GB
25
+ [14:07:57] step 1110/2860 loss 2.8485 lr 6.00e-04 gnorm 0.09 107,630 tok/s peakVRAM 27.0GB
26
+ [14:09:34] step 1120/2860 loss 2.8432 lr 6.00e-04 gnorm 0.09 108,072 tok/s peakVRAM 27.0GB
27
+ [14:11:11] step 1130/2860 loss 2.8414 lr 6.00e-04 gnorm 0.09 107,987 tok/s peakVRAM 27.0GB
28
+ [14:12:48] step 1140/2860 loss 2.8192 lr 6.00e-04 gnorm 0.10 107,987 tok/s peakVRAM 27.0GB
29
+ [14:14:25] step 1150/2860 loss 2.8366 lr 6.00e-04 gnorm 0.11 107,797 tok/s peakVRAM 27.0GB
30
+ [14:16:02] step 1160/2860 loss 2.8427 lr 6.00e-04 gnorm 0.11 108,156 tok/s peakVRAM 27.0GB
31
+ [14:17:39] step 1170/2860 loss 2.8140 lr 6.00e-04 gnorm 0.10 108,311 tok/s peakVRAM 27.0GB
32
+ [14:19:16] step 1180/2860 loss 2.8157 lr 6.00e-04 gnorm 0.10 108,201 tok/s peakVRAM 27.0GB
33
+ [14:20:53] step 1190/2860 loss 2.8080 lr 6.00e-04 gnorm 0.11 107,802 tok/s peakVRAM 27.0GB
34
+ [14:22:31] step 1200/2860 loss 2.8378 lr 6.00e-04 gnorm 0.11 107,815 tok/s peakVRAM 27.0GB
35
+ [14:22:40] ckpt @ 1200
36
+ [14:24:17] step 1210/2860 loss 2.8056 lr 6.00e-04 gnorm 0.10 98,763 tok/s peakVRAM 27.0GB
37
+ [14:25:54] step 1220/2860 loss 2.8146 lr 6.00e-04 gnorm 0.10 108,307 tok/s peakVRAM 27.0GB
38
+ [14:27:31] step 1230/2860 loss 2.8082 lr 6.00e-04 gnorm 0.09 108,132 tok/s peakVRAM 27.0GB
39
+ [14:29:08] step 1240/2860 loss 2.8096 lr 6.00e-04 gnorm 0.10 107,957 tok/s peakVRAM 27.0GB
40
+ [14:30:45] step 1250/2860 loss 2.8166 lr 6.00e-04 gnorm 0.11 108,114 tok/s peakVRAM 27.0GB
41
+ [14:32:22] step 1260/2860 loss 2.8128 lr 6.00e-04 gnorm 0.10 108,061 tok/s peakVRAM 27.0GB
42
+ [14:33:59] step 1270/2860 loss 2.7992 lr 6.00e-04 gnorm 0.09 108,210 tok/s peakVRAM 27.0GB
43
+ [14:35:36] step 1280/2860 loss 2.7989 lr 6.00e-04 gnorm 0.10 108,128 tok/s peakVRAM 27.0GB
44
+ [14:37:13] step 1290/2860 loss 2.7926 lr 6.00e-04 gnorm 0.12 107,744 tok/s peakVRAM 27.0GB
45
+ [14:38:50] step 1300/2860 loss 2.7575 lr 6.00e-04 gnorm 0.10 107,733 tok/s peakVRAM 27.0GB
46
+ [14:40:28] step 1310/2860 loss 2.7676 lr 6.00e-04 gnorm 0.09 107,802 tok/s peakVRAM 27.0GB
47
+ [14:42:04] step 1320/2860 loss 2.7761 lr 6.00e-04 gnorm 0.09 108,185 tok/s peakVRAM 27.0GB
48
+ [14:43:42] step 1330/2860 loss 2.7931 lr 6.00e-04 gnorm 0.09 107,999 tok/s peakVRAM 27.0GB
49
+ [14:45:19] step 1340/2860 loss 2.7885 lr 6.00e-04 gnorm 0.09 107,917 tok/s peakVRAM 27.0GB
50
+ [14:46:56] step 1350/2860 loss 2.7733 lr 6.00e-04 gnorm 0.08 108,010 tok/s peakVRAM 27.0GB
51
+ [14:47:05] ckpt @ 1350
52
+ [14:48:42] step 1360/2860 loss 2.7871 lr 6.00e-04 gnorm 0.10 98,694 tok/s peakVRAM 27.0GB
53
+ [14:50:19] step 1370/2860 loss 2.7689 lr 6.00e-04 gnorm 0.09 108,306 tok/s peakVRAM 27.0GB
54
+ [14:51:56] step 1380/2860 loss 2.7655 lr 6.00e-04 gnorm 0.09 107,983 tok/s peakVRAM 27.0GB
55
+ [14:53:33] step 1390/2860 loss 2.7674 lr 6.00e-04 gnorm 0.10 107,907 tok/s peakVRAM 27.0GB
56
+ [14:55:10] step 1400/2860 loss 2.7652 lr 6.00e-04 gnorm 0.09 107,833 tok/s peakVRAM 27.0GB
57
+ [14:56:48] step 1410/2860 loss 2.7603 lr 6.00e-04 gnorm 0.09 107,875 tok/s peakVRAM 27.0GB
58
+ [14:58:25] step 1420/2860 loss 2.7825 lr 6.00e-04 gnorm 0.10 107,990 tok/s peakVRAM 27.0GB
59
+ [15:00:02] step 1430/2860 loss 2.7738 lr 6.00e-04 gnorm 0.08 108,140 tok/s peakVRAM 27.0GB
60
+ [15:01:39] step 1440/2860 loss 2.7555 lr 6.00e-04 gnorm 0.09 108,163 tok/s peakVRAM 27.0GB
61
+ [15:03:15] step 1450/2860 loss 2.7610 lr 6.00e-04 gnorm 0.09 108,188 tok/s peakVRAM 27.0GB
62
+ [15:04:52] step 1460/2860 loss 2.7413 lr 6.00e-04 gnorm 0.09 108,135 tok/s peakVRAM 27.0GB
63
+ [15:06:29] step 1470/2860 loss 2.7548 lr 6.00e-04 gnorm 0.09 108,067 tok/s peakVRAM 27.0GB
64
+ [15:08:06] step 1480/2860 loss 2.7579 lr 6.00e-04 gnorm 0.10 108,194 tok/s peakVRAM 27.0GB
65
+ [15:09:44] step 1490/2860 loss 2.7474 lr 6.00e-04 gnorm 0.10 107,963 tok/s peakVRAM 27.0GB
66
+ [15:11:21] step 1500/2860 loss 2.7388 lr 6.00e-04 gnorm 0.10 107,926 tok/s peakVRAM 27.0GB
67
+ [15:11:30] ckpt @ 1500
68
+ [15:13:07] step 1510/2860 loss 2.7643 lr 6.00e-04 gnorm 0.10 98,671 tok/s peakVRAM 27.0GB
69
+ [15:14:44] step 1520/2860 loss 2.7517 lr 6.00e-04 gnorm 0.09 108,297 tok/s peakVRAM 27.0GB
70
+ [15:16:21] step 1530/2860 loss 2.7416 lr 6.00e-04 gnorm 0.08 108,167 tok/s peakVRAM 27.0GB
71
+ [15:17:58] step 1540/2860 loss 2.7395 lr 6.00e-04 gnorm 0.10 108,347 tok/s peakVRAM 27.0GB
72
+ [15:19:34] step 1550/2860 loss 2.7480 lr 6.00e-04 gnorm 0.09 108,358 tok/s peakVRAM 27.0GB
73
+ [15:21:11] step 1560/2860 loss 2.7244 lr 6.00e-04 gnorm 0.08 108,232 tok/s peakVRAM 27.0GB
74
+ [15:22:48] step 1570/2860 loss 2.7390 lr 6.00e-04 gnorm 0.09 108,028 tok/s peakVRAM 27.0GB
75
+ [15:24:25] step 1580/2860 loss 2.7299 lr 6.00e-04 gnorm 0.08 108,056 tok/s peakVRAM 27.0GB
76
+ [15:26:02] step 1590/2860 loss 2.7523 lr 6.00e-04 gnorm 0.10 108,021 tok/s peakVRAM 27.0GB
77
+ [15:27:39] step 1600/2860 loss 2.7405 lr 6.00e-04 gnorm 0.08 108,154 tok/s peakVRAM 27.0GB
78
+ [15:29:16] step 1610/2860 loss 2.7168 lr 6.00e-04 gnorm 0.08 107,893 tok/s peakVRAM 27.0GB
79
+ [15:30:54] step 1620/2860 loss 2.7081 lr 6.00e-04 gnorm 0.09 107,857 tok/s peakVRAM 27.0GB
80
+ [15:32:31] step 1630/2860 loss 2.7200 lr 6.00e-04 gnorm 0.09 107,659 tok/s peakVRAM 27.0GB
81
+ [15:34:08] step 1640/2860 loss 2.7136 lr 6.00e-04 gnorm 0.16 107,823 tok/s peakVRAM 27.0GB
82
+ [15:35:46] step 1650/2860 loss 2.7406 lr 6.00e-04 gnorm 0.09 107,779 tok/s peakVRAM 27.0GB
83
+ [15:35:56] ckpt @ 1650
84
+ [15:37:33] step 1660/2860 loss 2.7050 lr 6.00e-04 gnorm 0.09 98,087 tok/s peakVRAM 27.0GB
85
+ [15:39:10] step 1670/2860 loss 2.7108 lr 6.00e-04 gnorm 0.10 108,102 tok/s peakVRAM 27.0GB
86
+ [15:40:47] step 1680/2860 loss 2.7259 lr 6.00e-04 gnorm 0.10 107,897 tok/s peakVRAM 27.0GB
87
+ [15:42:24] step 1690/2860 loss 2.7016 lr 6.00e-04 gnorm 0.10 107,780 tok/s peakVRAM 27.0GB
88
+ [15:44:01] step 1700/2860 loss 2.7218 lr 6.00e-04 gnorm 0.09 107,921 tok/s peakVRAM 27.0GB
89
+ [15:45:38] step 1710/2860 loss 2.7047 lr 6.00e-04 gnorm 0.09 107,869 tok/s peakVRAM 27.0GB
90
+ [15:47:16] step 1720/2860 loss 2.7218 lr 6.00e-04 gnorm 0.10 107,872 tok/s peakVRAM 27.0GB
91
+ [15:48:53] step 1730/2860 loss 2.7104 lr 6.00e-04 gnorm 0.10 107,897 tok/s peakVRAM 27.0GB
92
+ [15:50:30] step 1740/2860 loss 2.6869 lr 6.00e-04 gnorm 0.08 107,874 tok/s peakVRAM 27.0GB
93
+ [15:52:07] step 1750/2860 loss 2.6925 lr 6.00e-04 gnorm 0.10 107,969 tok/s peakVRAM 27.0GB
94
+ [15:53:44] step 1760/2860 loss 2.6851 lr 6.00e-04 gnorm 0.10 107,963 tok/s peakVRAM 27.0GB
95
+ [15:55:21] step 1770/2860 loss 2.6856 lr 6.00e-04 gnorm 0.08 107,892 tok/s peakVRAM 27.0GB
96
+ [15:56:59] step 1780/2860 loss 2.6989 lr 6.00e-04 gnorm 0.10 107,713 tok/s peakVRAM 27.0GB
97
+ [15:58:36] step 1790/2860 loss 2.6702 lr 6.00e-04 gnorm 0.08 107,620 tok/s peakVRAM 27.0GB
98
+ [16:00:14] step 1800/2860 loss 2.6972 lr 6.00e-04 gnorm 0.09 107,335 tok/s peakVRAM 27.0GB
99
+ [16:00:24] ckpt @ 1800
100
+ [16:02:01] step 1810/2860 loss 2.6813 lr 6.00e-04 gnorm 0.09 97,749 tok/s peakVRAM 27.0GB
101
+ [16:03:38] step 1820/2860 loss 2.6996 lr 6.00e-04 gnorm 0.10 108,119 tok/s peakVRAM 27.0GB
102
+ [16:05:15] step 1830/2860 loss 2.7000 lr 6.00e-04 gnorm 0.08 107,858 tok/s peakVRAM 27.0GB
103
+ [16:06:52] step 1840/2860 loss 2.6922 lr 6.00e-04 gnorm 0.08 108,092 tok/s peakVRAM 27.0GB
104
+ [16:08:29] step 1850/2860 loss 2.6933 lr 6.00e-04 gnorm 0.08 108,196 tok/s peakVRAM 27.0GB
105
+ [16:10:06] step 1860/2860 loss 2.7058 lr 6.00e-04 gnorm 0.08 108,083 tok/s peakVRAM 27.0GB
106
+ [16:11:43] step 1870/2860 loss 2.6963 lr 6.00e-04 gnorm 0.08 108,178 tok/s peakVRAM 27.0GB
107
+ [16:13:20] step 1880/2860 loss 2.6996 lr 6.00e-04 gnorm 0.10 108,173 tok/s peakVRAM 27.0GB
108
+ [16:14:57] step 1890/2860 loss 2.6934 lr 6.00e-04 gnorm 0.09 107,941 tok/s peakVRAM 27.0GB
109
+ [16:16:34] step 1900/2860 loss 2.6880 lr 6.00e-04 gnorm 0.09 108,129 tok/s peakVRAM 27.0GB
110
+ [16:18:11] step 1910/2860 loss 2.6927 lr 6.00e-04 gnorm 0.09 108,021 tok/s peakVRAM 27.0GB
111
+ [16:19:49] step 1920/2860 loss 2.6745 lr 6.00e-04 gnorm 0.09 107,920 tok/s peakVRAM 27.0GB
112
+ [16:21:26] step 1930/2860 loss 2.6718 lr 6.00e-04 gnorm 0.09 107,949 tok/s peakVRAM 27.0GB
113
+ [16:23:03] step 1940/2860 loss 2.6847 lr 6.00e-04 gnorm 0.09 108,210 tok/s peakVRAM 27.0GB
114
+ [16:24:40] step 1950/2860 loss 2.6858 lr 6.00e-04 gnorm 0.08 107,953 tok/s peakVRAM 27.0GB
115
+ [16:24:50] ckpt @ 1950
116
+ [16:26:27] step 1960/2860 loss 2.6984 lr 6.00e-04 gnorm 0.09 97,835 tok/s peakVRAM 27.0GB
117
+ [16:28:04] step 1970/2860 loss 2.6685 lr 6.00e-04 gnorm 0.09 107,868 tok/s peakVRAM 27.0GB
118
+ [16:29:41] step 1980/2860 loss 2.6823 lr 6.00e-04 gnorm 0.09 107,690 tok/s peakVRAM 27.0GB
119
+ [16:31:19] step 1990/2860 loss 2.6722 lr 6.00e-04 gnorm 0.08 107,691 tok/s peakVRAM 27.0GB
120
+ [16:32:56] step 2000/2860 loss 2.6615 lr 6.00e-04 gnorm 0.09 107,819 tok/s peakVRAM 27.0GB
121
+ [16:34:33] step 2010/2860 loss 2.6696 lr 6.00e-04 gnorm 0.09 108,003 tok/s peakVRAM 27.0GB
122
+ [16:36:10] step 2020/2860 loss 2.6537 lr 6.00e-04 gnorm 0.08 107,785 tok/s peakVRAM 27.0GB
123
+ [16:37:48] step 2030/2860 loss 2.6670 lr 6.00e-04 gnorm 0.09 107,549 tok/s peakVRAM 27.0GB
124
+ [16:39:25] step 2040/2860 loss 2.7043 lr 6.00e-04 gnorm 0.09 107,558 tok/s peakVRAM 27.0GB
125
+ [16:41:03] step 2050/2860 loss 2.6757 lr 6.00e-04 gnorm 0.08 107,565 tok/s peakVRAM 27.0GB
126
+ [16:42:40] step 2060/2860 loss 2.6730 lr 6.00e-04 gnorm 0.09 107,805 tok/s peakVRAM 27.0GB
127
+ [16:44:17] step 2070/2860 loss 2.6598 lr 6.00e-04 gnorm 0.08 107,969 tok/s peakVRAM 27.0GB
128
+ [16:45:54] step 2080/2860 loss 2.6709 lr 6.00e-04 gnorm 0.10 108,008 tok/s peakVRAM 27.0GB
129
+ [16:47:32] step 2090/2860 loss 2.6693 lr 6.00e-04 gnorm 0.10 107,844 tok/s peakVRAM 27.0GB
130
+ [16:49:09] step 2100/2860 loss 2.6343 lr 6.00e-04 gnorm 0.08 107,862 tok/s peakVRAM 27.0GB
131
+ [16:49:20] ckpt @ 2100
132
+ [16:50:56] step 2110/2860 loss 2.6644 lr 6.00e-04 gnorm 0.10 97,559 tok/s peakVRAM 27.0GB
133
+ [16:52:34] step 2120/2860 loss 2.6637 lr 6.00e-04 gnorm 0.07 107,791 tok/s peakVRAM 27.0GB
134
+ [16:54:11] step 2130/2860 loss 2.6517 lr 6.00e-04 gnorm 0.09 107,569 tok/s peakVRAM 27.0GB
135
+ [16:55:48] step 2140/2860 loss 2.6472 lr 6.00e-04 gnorm 0.08 107,636 tok/s peakVRAM 27.0GB
136
+ [16:57:26] step 2150/2860 loss 2.6531 lr 6.00e-04 gnorm 0.08 107,873 tok/s peakVRAM 27.0GB
137
+ [16:59:03] step 2160/2860 loss 2.6535 lr 6.00e-04 gnorm 0.08 107,675 tok/s peakVRAM 27.0GB
138
+ [17:00:41] step 2170/2860 loss 2.6566 lr 6.00e-04 gnorm 0.08 107,485 tok/s peakVRAM 27.0GB
139
+ [17:02:18] step 2180/2860 loss 2.6688 lr 6.00e-04 gnorm 0.09 107,483 tok/s peakVRAM 27.0GB
140
+ [17:03:56] step 2190/2860 loss 2.6450 lr 6.00e-04 gnorm 0.08 107,499 tok/s peakVRAM 27.0GB
141
+ [17:05:33] step 2200/2860 loss 2.6498 lr 6.00e-04 gnorm 0.09 107,383 tok/s peakVRAM 27.0GB
142
+ [17:07:11] step 2210/2860 loss 2.6386 lr 6.00e-04 gnorm 0.09 107,588 tok/s peakVRAM 27.0GB
143
+ [17:08:48] step 2220/2860 loss 2.6283 lr 6.00e-04 gnorm 0.09 107,430 tok/s peakVRAM 27.0GB
144
+ [17:10:26] step 2230/2860 loss 2.6434 lr 6.00e-04 gnorm 0.07 107,231 tok/s peakVRAM 27.0GB
145
+ [17:12:04] step 2240/2860 loss 2.6654 lr 6.00e-04 gnorm 0.08 107,516 tok/s peakVRAM 27.0GB
146
+ [17:13:41] step 2250/2860 loss 2.6318 lr 6.00e-04 gnorm 0.08 107,529 tok/s peakVRAM 27.0GB
147
+ [17:13:53] ckpt @ 2250
148
+ [17:15:30] step 2260/2860 loss 2.6379 lr 6.00e-04 gnorm 0.10 96,309 tok/s peakVRAM 27.0GB
149
+ [17:17:07] step 2270/2860 loss 2.6461 lr 6.00e-04 gnorm 0.08 107,863 tok/s peakVRAM 27.0GB
150
+ [17:18:45] step 2280/2860 loss 2.6532 lr 6.00e-04 gnorm 0.08 107,509 tok/s peakVRAM 27.0GB
151
+ [17:20:22] step 2290/2860 loss 2.6472 lr 5.77e-04 gnorm 0.08 107,551 tok/s peakVRAM 27.0GB
152
+ [17:22:00] step 2300/2860 loss 2.6419 lr 5.25e-04 gnorm 0.07 107,791 tok/s peakVRAM 27.0GB
153
+ [17:23:37] step 2310/2860 loss 2.6313 lr 4.97e-04 gnorm 0.07 107,474 tok/s peakVRAM 27.0GB
154
+ [17:25:15] step 2320/2860 loss 2.6209 lr 4.74e-04 gnorm 0.06 107,316 tok/s peakVRAM 27.0GB
155
+ [17:26:53] step 2330/2860 loss 2.6298 lr 4.55e-04 gnorm 0.07 107,306 tok/s peakVRAM 27.0GB
156
+ [17:28:30] step 2340/2860 loss 2.6196 lr 4.39e-04 gnorm 0.07 107,440 tok/s peakVRAM 27.0GB
157
+ [17:30:08] step 2350/2860 loss 2.6067 lr 4.24e-04 gnorm 0.06 107,241 tok/s peakVRAM 27.0GB
158
+ [17:31:46] step 2360/2860 loss 2.5964 lr 4.10e-04 gnorm 0.07 106,925 tok/s peakVRAM 27.0GB
159
+ [17:33:24] step 2370/2860 loss 2.5960 lr 3.97e-04 gnorm 0.06 107,135 tok/s peakVRAM 27.0GB
160
+ [17:35:02] step 2380/2860 loss 2.5998 lr 3.85e-04 gnorm 0.06 107,354 tok/s peakVRAM 27.0GB
161
+ [17:36:39] step 2390/2860 loss 2.5892 lr 3.73e-04 gnorm 0.06 107,686 tok/s peakVRAM 27.0GB
162
+ [17:38:16] step 2400/2860 loss 2.6009 lr 3.62e-04 gnorm 0.05 107,816 tok/s peakVRAM 27.0GB
163
+ [17:38:29] ckpt @ 2400
164
+ [17:40:06] step 2410/2860 loss 2.6035 lr 3.52e-04 gnorm 0.06 95,863 tok/s peakVRAM 27.0GB
165
+ [17:41:43] step 2420/2860 loss 2.5908 lr 3.42e-04 gnorm 0.06 107,941 tok/s peakVRAM 27.0GB
166
+ [17:43:20] step 2430/2860 loss 2.5813 lr 3.32e-04 gnorm 0.05 107,513 tok/s peakVRAM 27.0GB
167
+ [17:44:58] step 2440/2860 loss 2.5774 lr 3.23e-04 gnorm 0.06 107,573 tok/s peakVRAM 27.0GB
168
+ [17:46:35] step 2450/2860 loss 2.5865 lr 3.14e-04 gnorm 0.06 107,692 tok/s peakVRAM 27.0GB
169
+ [17:48:13] step 2460/2860 loss 2.5850 lr 3.05e-04 gnorm 0.05 107,579 tok/s peakVRAM 27.0GB
170
+ [17:49:50] step 2470/2860 loss 2.5728 lr 2.96e-04 gnorm 0.05 107,413 tok/s peakVRAM 27.0GB
171
+ [17:51:28] step 2480/2860 loss 2.5804 lr 2.88e-04 gnorm 0.05 107,500 tok/s peakVRAM 27.0GB
172
+ [17:53:05] step 2490/2860 loss 2.5865 lr 2.80e-04 gnorm 0.06 107,691 tok/s peakVRAM 27.0GB
173
+ [17:54:43] step 2500/2860 loss 2.5757 lr 2.72e-04 gnorm 0.05 107,346 tok/s peakVRAM 27.0GB
174
+ [17:56:20] step 2510/2860 loss 2.5683 lr 2.64e-04 gnorm 0.05 107,447 tok/s peakVRAM 27.0GB
175
+ [17:57:58] step 2520/2860 loss 2.5659 lr 2.57e-04 gnorm 0.05 107,256 tok/s peakVRAM 27.0GB
176
+ [17:59:36] step 2530/2860 loss 2.5493 lr 2.49e-04 gnorm 0.05 107,326 tok/s peakVRAM 27.0GB
177
+ [18:01:13] step 2540/2860 loss 2.5526 lr 2.42e-04 gnorm 0.05 107,522 tok/s peakVRAM 27.0GB
178
+ [18:02:51] step 2550/2860 loss 2.5513 lr 2.35e-04 gnorm 0.05 107,402 tok/s peakVRAM 27.0GB
179
+ [18:03:04] ckpt @ 2550
180
+ [18:04:41] step 2560/2860 loss 2.5376 lr 2.28e-04 gnorm 0.05 95,800 tok/s peakVRAM 27.0GB
181
+ [18:06:18] step 2570/2860 loss 2.5497 lr 2.22e-04 gnorm 0.05 107,579 tok/s peakVRAM 27.0GB
182
+ [18:07:56] step 2580/2860 loss 2.5446 lr 2.15e-04 gnorm 0.05 107,307 tok/s peakVRAM 27.0GB
183
+ [18:09:34] step 2590/2860 loss 2.5501 lr 2.08e-04 gnorm 0.05 107,222 tok/s peakVRAM 27.0GB
184
+ [18:11:11] step 2600/2860 loss 2.5475 lr 2.02e-04 gnorm 0.05 107,513 tok/s peakVRAM 27.0GB
185
+ [18:12:49] step 2610/2860 loss 2.5570 lr 1.95e-04 gnorm 0.05 107,489 tok/s peakVRAM 27.0GB
186
+ [18:14:26] step 2620/2860 loss 2.5378 lr 1.89e-04 gnorm 0.05 107,509 tok/s peakVRAM 27.0GB
187
+ [18:16:04] step 2630/2860 loss 2.5365 lr 1.83e-04 gnorm 0.05 107,670 tok/s peakVRAM 27.0GB
188
+ [18:17:41] step 2640/2860 loss 2.5356 lr 1.77e-04 gnorm 0.05 107,664 tok/s peakVRAM 27.0GB
189
+ [18:19:19] step 2650/2860 loss 2.5567 lr 1.71e-04 gnorm 0.05 107,420 tok/s peakVRAM 27.0GB
190
+ [18:20:56] step 2660/2860 loss 2.5153 lr 1.65e-04 gnorm 0.05 107,311 tok/s peakVRAM 27.0GB
191
+ [18:22:34] step 2670/2860 loss 2.5341 lr 1.59e-04 gnorm 0.05 107,136 tok/s peakVRAM 27.0GB
192
+ [18:24:12] step 2680/2860 loss 2.5276 lr 1.54e-04 gnorm 0.05 107,568 tok/s peakVRAM 27.0GB
193
+ [18:25:49] step 2690/2860 loss 2.5233 lr 1.48e-04 gnorm 0.05 107,606 tok/s peakVRAM 27.0GB
194
+ [18:27:27] step 2700/2860 loss 2.5278 lr 1.42e-04 gnorm 0.05 107,163 tok/s peakVRAM 27.0GB
195
+ [18:27:37] ckpt @ 2700
196
+ [18:29:15] step 2710/2860 loss 2.5311 lr 1.37e-04 gnorm 0.05 97,103 tok/s peakVRAM 27.0GB
197
+ [18:30:53] step 2720/2860 loss 2.5200 lr 1.31e-04 gnorm 0.05 107,242 tok/s peakVRAM 27.0GB
198
+ [18:32:30] step 2730/2860 loss 2.5417 lr 1.26e-04 gnorm 0.05 107,248 tok/s peakVRAM 27.0GB
199
+ [18:34:08] step 2740/2860 loss 2.5239 lr 1.21e-04 gnorm 0.05 107,314 tok/s peakVRAM 27.0GB
200
+ [18:35:46] step 2750/2860 loss 2.5107 lr 1.15e-04 gnorm 0.05 107,318 tok/s peakVRAM 27.0GB
201
+ [18:37:24] step 2760/2860 loss 2.4960 lr 1.10e-04 gnorm 0.05 107,012 tok/s peakVRAM 27.0GB
202
+ [18:39:02] step 2770/2860 loss 2.5277 lr 1.05e-04 gnorm 0.05 106,943 tok/s peakVRAM 27.0GB
203
+ [18:40:40] step 2780/2860 loss 2.5243 lr 9.97e-05 gnorm 0.05 107,074 tok/s peakVRAM 27.0GB
204
+ [18:42:18] step 2790/2860 loss 2.5119 lr 9.46e-05 gnorm 0.05 107,164 tok/s peakVRAM 27.0GB
205
+ [18:43:55] step 2800/2860 loss 2.5066 lr 8.96e-05 gnorm 0.05 107,364 tok/s peakVRAM 27.0GB
206
+ [18:45:33] step 2810/2860 loss 2.5089 lr 8.46e-05 gnorm 0.05 107,309 tok/s peakVRAM 27.0GB
207
+ [18:47:11] step 2820/2860 loss 2.5087 lr 7.97e-05 gnorm 0.05 107,408 tok/s peakVRAM 27.0GB
208
+ [18:48:48] step 2830/2860 loss 2.4930 lr 7.48e-05 gnorm 0.05 107,398 tok/s peakVRAM 27.0GB
209
+ [18:50:26] step 2840/2860 loss 2.5122 lr 7.00e-05 gnorm 0.05 107,539 tok/s peakVRAM 27.0GB
210
+ [18:52:03] step 2850/2860 loss 2.4829 lr 6.52e-05 gnorm 0.05 107,477 tok/s peakVRAM 27.0GB
211
+ [18:52:16] ckpt @ 2850
212
+ [18:53:53] step 2860/2860 loss 2.5067 lr 6.05e-05 gnorm 0.05 95,659 tok/s peakVRAM 27.0GB
213
+ [18:54:06] ckpt @ 2860
214
+ [18:54:06] DONE
run/wsad_p01_SHA256SUMS ADDED
@@ -0,0 +1,5 @@
 
 
 
 
 
 
1
+ 0494df580a1bd75099fba85df5fcd1df248b02bd47228f99a0ea6088e635e906 train.bin
2
+ 49ec1403347aebd5fced178d5aeb986a336f6542d1da9631f78ed5bdc1159d68 val_pl.bin
3
+ 9ac7ac75f18e45f89e25ce01631e07e4928a0badb928398e0c69a756a98ce2f5 val_en.bin
4
+ ab669dabeba476fc22dbe9ad08e43358037eb99bd1260fba9d8db1520ea03663 val.bin
5
+ 0640d3bd3674d7a6e59c540d945a2ad275a12e1bcaeb6007c37a90751f9815dd tokenizer.json
run/wsad_p01_manifest.json ADDED
@@ -0,0 +1,201 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "zrodlo": "SlayerLab/slayer-pl-8x3b pack-01",
3
+ "rewizja": "d50df04a6ad46876fa559a87cca5d3e2a96e7098",
4
+ "rewizja_sprawdzona": "sha parquetów == SHA_REV",
5
+ "manifesty_wejscia_sha256": {
6
+ "pack-01/manifest.json": "132b941f3e3d3568e36b345a848d250ef802f5ae6da08048742030d188dd8a6f",
7
+ "shared/manifest.json": "78a53c5df6f48d29a4032c093f08ae2b98c551f541c290d335cf5e5f7ae7fcfc"
8
+ },
9
+ "pliki": [
10
+ {
11
+ "plik": "pack-01/web.parquet",
12
+ "komponent": "web",
13
+ "sha256": "802f4829f41ef83e23979d2460500907bcfd44278e300eccb5e39ac666a04c2c",
14
+ "tokeny_cl100k": 1889999832
15
+ },
16
+ {
17
+ "plik": "shared/core.parquet",
18
+ "komponent": "rdzen",
19
+ "sha256": "ea0b4b901ace3780501b02a3972f014f77ce722315476e6f93681616958e0a99",
20
+ "tokeny_cl100k": 162205785
21
+ },
22
+ {
23
+ "plik": "shared/core_sa.parquet",
24
+ "komponent": "rdzen",
25
+ "sha256": "910b18eeebbc26efbbd76554d7484d016c6df692a3b6a18492b5c9ceb32549cb",
26
+ "tokeny_cl100k": 3546606209
27
+ },
28
+ {
29
+ "plik": "shared/legal.parquet",
30
+ "komponent": "legal",
31
+ "sha256": "6d4cc2f604d76452cd7e7ea52d1a04826f5e44f09bb035ba0853232fe373e74a",
32
+ "tokeny_cl100k": 1920115782
33
+ },
34
+ {
35
+ "plik": "shared/legal_sa.parquet",
36
+ "komponent": "legal",
37
+ "sha256": "a1609cd31e12e0f6fdfdb030edb538bc230fc5f5ab376d23136180338e9dbd1e",
38
+ "tokeny_cl100k": 80666129
39
+ },
40
+ {
41
+ "plik": "shared/en.parquet",
42
+ "komponent": "en",
43
+ "sha256": "f3d451b7a202d7748a598c6c958c38cbd42bfc981c8d2fcbc3cdd721276b935a",
44
+ "tokeny_cl100k": 2401565319
45
+ }
46
+ ],
47
+ "wagi": {
48
+ "web": 0.63,
49
+ "rdzen": 0.19,
50
+ "legal": 0.08,
51
+ "en": 0.1
52
+ },
53
+ "budzet_cl100k": 3000000000.0,
54
+ "frakcje": {
55
+ "web": 1.0,
56
+ "rdzen": 0.1536880275738237,
57
+ "legal": 0.11995310367437644,
58
+ "en": 0.12491852610734674
59
+ },
60
+ "przedzialy": {
61
+ "web": [
62
+ 0.0,
63
+ 1.0
64
+ ],
65
+ "rdzen": [
66
+ 0.1536880275738237,
67
+ 0.1536880275738237
68
+ ],
69
+ "legal": [
70
+ 0.11995310367437644,
71
+ 0.11995310367437644
72
+ ],
73
+ "en": [
74
+ 0.12491852610734674,
75
+ 0.12491852610734674
76
+ ]
77
+ },
78
+ "seed": 20261023,
79
+ "poprzedni_manifest_sha256": "cc831cb8b91e5cd1714b32bb8b9fcd5d7be0720184b6b40835b9e6af221b4249",
80
+ "val_pl_frac": 0.001,
81
+ "val_en_frac": 0.005,
82
+ "tokenizer_sha256": "0640d3bd3674d7a6e59c540d945a2ad275a12e1bcaeb6007c37a90751f9815dd",
83
+ "eos": 0,
84
+ "dtype": "uint16",
85
+ "liczniki": {
86
+ "web": {
87
+ "wiersze": 1412646,
88
+ "garbage": 0,
89
+ "wybrane": 1412646,
90
+ "puste": 0,
91
+ "specjalne_pominiete": 0,
92
+ "email": 0,
93
+ "pesel": 57,
94
+ "nip": 264,
95
+ "regon": 252,
96
+ "dowod": 69,
97
+ "telefon": 3747,
98
+ "dok_train": 1411278,
99
+ "dok_val": 1368,
100
+ "tok_train": 980966067,
101
+ "tok_val": 905131
102
+ },
103
+ "rdzen": {
104
+ "wiersze": 8579260,
105
+ "garbage": 0,
106
+ "wybrane": 1318510,
107
+ "puste": 0,
108
+ "specjalne_pominiete": 0,
109
+ "email": 7466,
110
+ "pesel": 43,
111
+ "nip": 298,
112
+ "regon": 115,
113
+ "dowod": 58,
114
+ "telefon": 7648,
115
+ "dok_train": 1317212,
116
+ "dok_val": 1298,
117
+ "tok_train": 416418851,
118
+ "tok_val": 380690
119
+ },
120
+ "legal": {
121
+ "wiersze": 350213,
122
+ "garbage": 0,
123
+ "wybrane": 41884,
124
+ "puste": 0,
125
+ "specjalne_pominiete": 0,
126
+ "email": 2998,
127
+ "pesel": 23,
128
+ "nip": 298,
129
+ "regon": 578,
130
+ "dowod": 41,
131
+ "telefon": 5266,
132
+ "dok_train": 41842,
133
+ "dok_val": 42,
134
+ "tok_train": 154193068,
135
+ "tok_val": 133175
136
+ },
137
+ "en": {
138
+ "wiersze": 2384000,
139
+ "garbage": 0,
140
+ "wybrane": 296955,
141
+ "puste": 0,
142
+ "specjalne_pominiete": 0,
143
+ "email": 8871,
144
+ "pesel": 51,
145
+ "nip": 758,
146
+ "regon": 122,
147
+ "dowod": 43,
148
+ "telefon": 4920,
149
+ "dok_train": 295499,
150
+ "dok_val": 1456,
151
+ "tok_train": 328942459,
152
+ "tok_val": 1703971
153
+ }
154
+ },
155
+ "tokeny_v6_na_komponent": {
156
+ "web": 981871198,
157
+ "rdzen": 416799541,
158
+ "legal": 154326243,
159
+ "en": 330646430
160
+ },
161
+ "udzial_v6": {
162
+ "web": 0.5212617163869018,
163
+ "rdzen": 0.2212730596166574,
164
+ "legal": 0.08192964868872962,
165
+ "en": 0.17553557530771116
166
+ },
167
+ "tokeny": {
168
+ "train.bin": 1880520445,
169
+ "val_pl.bin": 1418996,
170
+ "val_en.bin": 1703971
171
+ },
172
+ "sha256": {
173
+ "train.bin": "0494df580a1bd75099fba85df5fcd1df248b02bd47228f99a0ea6088e635e906",
174
+ "val_pl.bin": "49ec1403347aebd5fced178d5aeb986a336f6542d1da9631f78ed5bdc1159d68",
175
+ "val_en.bin": "9ac7ac75f18e45f89e25ce01631e07e4928a0badb928398e0c69a756a98ce2f5",
176
+ "val.bin": "ab669dabeba476fc22dbe9ad08e43358037eb99bd1260fba9d8db1520ea03663",
177
+ "tokenizer.json": "0640d3bd3674d7a6e59c540d945a2ad275a12e1bcaeb6007c37a90751f9815dd"
178
+ },
179
+ "skrypt_sha256": "e1ab4e930d8f9918593cf560c826830b6925ca19ad5576727847ccfef6638406",
180
+ "srodowisko": {
181
+ "python": "3.11.13",
182
+ "pyarrow": "20.0.0",
183
+ "tokenizers": "0.23.2",
184
+ "numpy": "2.3.2"
185
+ },
186
+ "argv": [
187
+ "--root",
188
+ "/root/pilot1b/slayer-pl-8x3b",
189
+ "--tokenizer",
190
+ "tok_32768.json",
191
+ "--out",
192
+ "wsad_p01",
193
+ "--pack",
194
+ "01",
195
+ "--poprzedni-manifest",
196
+ "wsad_v2/manifest.json",
197
+ "--workers",
198
+ "240"
199
+ ],
200
+ "uwaga": "dane wewnętrzne pilota 1B; PII maskowane regułami (bez gwarancji kompletności); val.bin = val_pl + val_en tylko do trendu w pętli trenera, decyzje z val_bpb osobno na val_pl i val_en; liczbę kroków planować z 'tokeny', nie z budżetu cl100k; bez publikacji wag bez bramki PII i przeglądu licencji (_sa = CC-BY-SA)"
201
+ }
run/wsad_v2_SHA256SUMS ADDED
@@ -0,0 +1,5 @@
 
 
 
 
 
 
1
+ 760ef0cc43ffeeaadf079697ff9281052210464696465705904a887b746c9b84 train.bin
2
+ 28a02765817347beaac328ff33d628861114092b45f4ba6856d58419e8a51115 val_pl.bin
3
+ ea978c0e2b6557ba1dde374e73c151b4beeddb1bde1e4a5ff6c0a055721fc8ae val_en.bin
4
+ acbbfaceb16555b3650cb25f33e049de408ad17f201c6a67cdef0b4648ea8319 val.bin
5
+ 0640d3bd3674d7a6e59c540d945a2ad275a12e1bcaeb6007c37a90751f9815dd tokenizer.json
run/wsad_v2_manifest.json ADDED
@@ -0,0 +1,199 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "zrodlo": "SlayerLab/slayer-pl-8x3b pack-00",
3
+ "rewizja": "d50df04a6ad46876fa559a87cca5d3e2a96e7098",
4
+ "rewizja_sprawdzona": "sha parquetów == SHA_REV",
5
+ "manifesty_wejscia_sha256": {
6
+ "pack-00/manifest.json": "f811021870b9600af8684cdfb137b423cd43a172d7258cf9c8f68edcf9938b09",
7
+ "shared/manifest.json": "78a53c5df6f48d29a4032c093f08ae2b98c551f541c290d335cf5e5f7ae7fcfc"
8
+ },
9
+ "pliki": [
10
+ {
11
+ "plik": "pack-00/web.parquet",
12
+ "komponent": "web",
13
+ "sha256": "5cbb6dfa3a276a30db3c2034105ffffa75cc29b767456e7d75c305f9bd7ec562",
14
+ "tokeny_cl100k": 1890000496
15
+ },
16
+ {
17
+ "plik": "shared/core.parquet",
18
+ "komponent": "rdzen",
19
+ "sha256": "ea0b4b901ace3780501b02a3972f014f77ce722315476e6f93681616958e0a99",
20
+ "tokeny_cl100k": 162205785
21
+ },
22
+ {
23
+ "plik": "shared/core_sa.parquet",
24
+ "komponent": "rdzen",
25
+ "sha256": "910b18eeebbc26efbbd76554d7484d016c6df692a3b6a18492b5c9ceb32549cb",
26
+ "tokeny_cl100k": 3546606209
27
+ },
28
+ {
29
+ "plik": "shared/legal.parquet",
30
+ "komponent": "legal",
31
+ "sha256": "6d4cc2f604d76452cd7e7ea52d1a04826f5e44f09bb035ba0853232fe373e74a",
32
+ "tokeny_cl100k": 1920115782
33
+ },
34
+ {
35
+ "plik": "shared/legal_sa.parquet",
36
+ "komponent": "legal",
37
+ "sha256": "a1609cd31e12e0f6fdfdb030edb538bc230fc5f5ab376d23136180338e9dbd1e",
38
+ "tokeny_cl100k": 80666129
39
+ },
40
+ {
41
+ "plik": "shared/en.parquet",
42
+ "komponent": "en",
43
+ "sha256": "f3d451b7a202d7748a598c6c958c38cbd42bfc981c8d2fcbc3cdd721276b935a",
44
+ "tokeny_cl100k": 2401565319
45
+ }
46
+ ],
47
+ "wagi": {
48
+ "web": 0.63,
49
+ "rdzen": 0.19,
50
+ "legal": 0.08,
51
+ "en": 0.1
52
+ },
53
+ "budzet_cl100k": 3000000000.0,
54
+ "frakcje": {
55
+ "web": 0.9999997375662064,
56
+ "rdzen": 0.1536880275738237,
57
+ "legal": 0.11995310367437644,
58
+ "en": 0.12491852610734674
59
+ },
60
+ "przedzialy": {
61
+ "web": [
62
+ 0.0,
63
+ 0.9999997375662064
64
+ ],
65
+ "rdzen": [
66
+ 0.0,
67
+ 0.1536880275738237
68
+ ],
69
+ "legal": [
70
+ 0.0,
71
+ 0.11995310367437644
72
+ ],
73
+ "en": [
74
+ 0.0,
75
+ 0.12491852610734674
76
+ ]
77
+ },
78
+ "seed": 20261023,
79
+ "poprzedni_manifest_sha256": null,
80
+ "val_pl_frac": 0.001,
81
+ "val_en_frac": 0.005,
82
+ "tokenizer_sha256": "0640d3bd3674d7a6e59c540d945a2ad275a12e1bcaeb6007c37a90751f9815dd",
83
+ "eos": 0,
84
+ "dtype": "uint16",
85
+ "liczniki": {
86
+ "web": {
87
+ "wiersze": 1412141,
88
+ "garbage": 0,
89
+ "wybrane": 1412141,
90
+ "puste": 0,
91
+ "specjalne_pominiete": 0,
92
+ "email": 0,
93
+ "pesel": 48,
94
+ "nip": 272,
95
+ "regon": 240,
96
+ "dowod": 44,
97
+ "telefon": 4114,
98
+ "dok_train": 1410724,
99
+ "dok_val": 1417,
100
+ "tok_train": 978608734,
101
+ "tok_val": 991180
102
+ },
103
+ "rdzen": {
104
+ "wiersze": 8579260,
105
+ "garbage": 0,
106
+ "wybrane": 1317383,
107
+ "puste": 0,
108
+ "specjalne_pominiete": 0,
109
+ "email": 7285,
110
+ "pesel": 33,
111
+ "nip": 237,
112
+ "regon": 140,
113
+ "dowod": 35,
114
+ "telefon": 7130,
115
+ "dok_train": 1316041,
116
+ "dok_val": 1342,
117
+ "tok_train": 419344519,
118
+ "tok_val": 319737
119
+ },
120
+ "legal": {
121
+ "wiersze": 350213,
122
+ "garbage": 0,
123
+ "wybrane": 42030,
124
+ "puste": 0,
125
+ "specjalne_pominiete": 0,
126
+ "email": 2913,
127
+ "pesel": 10,
128
+ "nip": 319,
129
+ "regon": 859,
130
+ "dowod": 141,
131
+ "telefon": 5640,
132
+ "dok_train": 41976,
133
+ "dok_val": 54,
134
+ "tok_train": 159799980,
135
+ "tok_val": 251730
136
+ },
137
+ "en": {
138
+ "wiersze": 2384000,
139
+ "garbage": 0,
140
+ "wybrane": 298438,
141
+ "puste": 0,
142
+ "specjalne_pominiete": 0,
143
+ "email": 8882,
144
+ "pesel": 40,
145
+ "nip": 774,
146
+ "regon": 182,
147
+ "dowod": 39,
148
+ "telefon": 4722,
149
+ "dok_train": 297002,
150
+ "dok_val": 1436,
151
+ "tok_train": 331766394,
152
+ "tok_val": 1643997
153
+ }
154
+ },
155
+ "tokeny_v6_na_komponent": {
156
+ "web": 979599914,
157
+ "rdzen": 419664256,
158
+ "legal": 160051710,
159
+ "en": 333410391
160
+ },
161
+ "udzial_v6": {
162
+ "web": 0.5175602669066561,
163
+ "rdzen": 0.22172474827977912,
164
+ "legal": 0.08456146694441903,
165
+ "en": 0.1761535178691457
166
+ },
167
+ "tokeny": {
168
+ "train.bin": 1889519627,
169
+ "val_pl.bin": 1562647,
170
+ "val_en.bin": 1643997
171
+ },
172
+ "sha256": {
173
+ "train.bin": "760ef0cc43ffeeaadf079697ff9281052210464696465705904a887b746c9b84",
174
+ "val_pl.bin": "28a02765817347beaac328ff33d628861114092b45f4ba6856d58419e8a51115",
175
+ "val_en.bin": "ea978c0e2b6557ba1dde374e73c151b4beeddb1bde1e4a5ff6c0a055721fc8ae",
176
+ "val.bin": "acbbfaceb16555b3650cb25f33e049de408ad17f201c6a67cdef0b4648ea8319",
177
+ "tokenizer.json": "0640d3bd3674d7a6e59c540d945a2ad275a12e1bcaeb6007c37a90751f9815dd"
178
+ },
179
+ "skrypt_sha256": "e1ab4e930d8f9918593cf560c826830b6925ca19ad5576727847ccfef6638406",
180
+ "srodowisko": {
181
+ "python": "3.11.13",
182
+ "pyarrow": "20.0.0",
183
+ "tokenizers": "0.23.2",
184
+ "numpy": "2.3.2"
185
+ },
186
+ "argv": [
187
+ "--root",
188
+ "/root/pilot1b/slayer-pl-8x3b",
189
+ "--tokenizer",
190
+ "tok_32768.json",
191
+ "--out",
192
+ "wsad_v2",
193
+ "--pack",
194
+ "00",
195
+ "--workers",
196
+ "240"
197
+ ],
198
+ "uwaga": "dane wewnętrzne pilota 1B; PII maskowane regułami (bez gwarancji kompletności); val.bin = val_pl + val_en tylko do trendu w pętli trenera, decyzje z val_bpb osobno na val_pl i val_en; liczbę kroków planować z 'tokeny', nie z budżetu cl100k; bez publikacji wag bez bramki PII i przeglądu licencji (_sa = CC-BY-SA)"
199
+ }