unparallelled commited on
Commit
5362d57
·
verified ·
1 Parent(s): 7ce9fb6

Upload folder using huggingface_hub

Browse files
Files changed (1) hide show
  1. logs/probepos_logs/master.log +86 -0
logs/probepos_logs/master.log ADDED
@@ -0,0 +1,86 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ [11:12:14] probe-pos sweep starting on NVIDIA H100 80GB HBM3
2
+ [11:12:14] PROBE_POS=8192 model=base (variants: p0.5001 p1 null absent)
3
+ [2026-08-31 11:12:15] needle sweep: model=1b_run8 run=0 factors=[4] caches=[yarn] blocks=3
4
+ [2026-08-31 11:12:15] GPUS=0 (n=1) kv=4 head_dim=128 theta=500000
5
+ [2026-08-31 11:12:15] == s=4 T=16384 CP=1 (single-GPU) N=8 slots=1 ==
6
+ [2026-08-31 11:19:43] scoring s=4 cache=yarn -> probe_pos8192_base/reach.csv
7
+ variant nll se score sig? notes
8
+ p0.5001 3.1080 0.1754 0.8767 YES
9
+ p1 1.5855 0.0505 1.0000 YES
10
+
11
+ floor(p1)=1.5855 absent=13.9323 null=12.3941 denom=12.3468 nats
12
+
13
+ wrote probe_pos8192_base/reach.csv
14
+ [2026-08-31 11:19:43] SWEEP DONE -> probe_pos8192_base/reach.csv (CSVs in probe_pos8192_base/csv, markers in probe_pos8192_base/state)
15
+ [11:19:43] -> probe_pos8192_base/reach.csv
16
+ [11:19:43] PROBE_POS=8192 model=armB (variants: p0.5001 p1 null absent)
17
+ [2026-08-31 11:19:44] needle sweep: model=1b_run8 run=0 factors=[4] caches=[yarn] blocks=3
18
+ [2026-08-31 11:19:44] GPUS=0 (n=1) kv=4 head_dim=128 theta=500000
19
+ [2026-08-31 11:19:44] == s=4 T=16384 CP=1 (single-GPU) N=8 slots=1 ==
20
+ [2026-08-31 11:27:22] scoring s=4 cache=yarn -> probe_pos8192_armB/reach.csv
21
+ variant nll se score sig? notes
22
+ p0.5001 4.8358 0.4371 0.7152 YES
23
+ p1 1.2433 0.0127 1.0000 YES
24
+
25
+ floor(p1)=1.2433 absent=13.8585 null=13.7556 denom=12.6152 nats
26
+
27
+ wrote probe_pos8192_armB/reach.csv
28
+ [2026-08-31 11:27:23] SWEEP DONE -> probe_pos8192_armB/reach.csv (CSVs in probe_pos8192_armB/csv, markers in probe_pos8192_armB/state)
29
+ [11:27:23] -> probe_pos8192_armB/reach.csv
30
+ [11:27:23] PROBE_POS=8192 model=armC (variants: p0.5001 p1 null absent)
31
+ [2026-08-31 11:27:23] needle sweep: model=1b_run8 run=0 factors=[4] caches=[yarn] blocks=3
32
+ [2026-08-31 11:27:23] GPUS=0 (n=1) kv=4 head_dim=128 theta=500000
33
+ [2026-08-31 11:27:23] == s=4 T=16384 CP=1 (single-GPU) N=8 slots=1 ==
34
+ [2026-08-31 11:35:01] scoring s=4 cache=yarn -> probe_pos8192_armC/reach.csv
35
+ variant nll se score sig? notes
36
+ p0.5001 2.7550 0.6873 0.8825 YES
37
+ p1 1.2675 0.0864 1.0000 YES
38
+
39
+ floor(p1)=1.2675 absent=13.9226 null=13.1675 denom=12.6551 nats
40
+
41
+ wrote probe_pos8192_armC/reach.csv
42
+ [2026-08-31 11:35:01] SWEEP DONE -> probe_pos8192_armC/reach.csv (CSVs in probe_pos8192_armC/csv, markers in probe_pos8192_armC/state)
43
+ [11:35:01] -> probe_pos8192_armC/reach.csv
44
+ [11:35:01] PROBE_POS=12288 model=base (variants: p0.6670 p1 null absent)
45
+ [2026-08-31 11:35:02] needle sweep: model=1b_run8 run=0 factors=[4] caches=[yarn] blocks=3
46
+ [2026-08-31 11:35:02] GPUS=0 (n=1) kv=4 head_dim=128 theta=500000
47
+ [2026-08-31 11:35:02] == s=4 T=16384 CP=1 (single-GPU) N=8 slots=1 ==
48
+ [2026-08-31 11:42:40] scoring s=4 cache=yarn -> probe_pos12288_base/reach.csv
49
+ variant nll se score sig? notes
50
+ p0.6670 3.1090 0.3586 0.8619 YES
51
+ p1 1.3972 0.1081 1.0000 YES
52
+
53
+ floor(p1)=1.3972 absent=13.7901 null=13.4591 denom=12.3929 nats
54
+
55
+ wrote probe_pos12288_base/reach.csv
56
+ [2026-08-31 11:42:40] SWEEP DONE -> probe_pos12288_base/reach.csv (CSVs in probe_pos12288_base/csv, markers in probe_pos12288_base/state)
57
+ [11:42:40] -> probe_pos12288_base/reach.csv
58
+ [11:42:40] PROBE_POS=12288 model=armB (variants: p0.6670 p1 null absent)
59
+ [2026-08-31 11:42:41] needle sweep: model=1b_run8 run=0 factors=[4] caches=[yarn] blocks=3
60
+ [2026-08-31 11:42:41] GPUS=0 (n=1) kv=4 head_dim=128 theta=500000
61
+ [2026-08-31 11:42:41] == s=4 T=16384 CP=1 (single-GPU) N=8 slots=1 ==
62
+ [2026-08-31 11:50:19] scoring s=4 cache=yarn -> probe_pos12288_armB/reach.csv
63
+ variant nll se score sig? notes
64
+ p0.6670 3.8071 0.5264 0.7995 YES
65
+ p1 1.2862 0.1035 1.0000 YES
66
+
67
+ floor(p1)=1.2862 absent=13.8572 null=13.7898 denom=12.5709 nats
68
+
69
+ wrote probe_pos12288_armB/reach.csv
70
+ [2026-08-31 11:50:20] SWEEP DONE -> probe_pos12288_armB/reach.csv (CSVs in probe_pos12288_armB/csv, markers in probe_pos12288_armB/state)
71
+ [11:50:20] -> probe_pos12288_armB/reach.csv
72
+ [11:50:20] PROBE_POS=12288 model=armC (variants: p0.6670 p1 null absent)
73
+ [2026-08-31 11:50:21] needle sweep: model=1b_run8 run=0 factors=[4] caches=[yarn] blocks=3
74
+ [2026-08-31 11:50:21] GPUS=0 (n=1) kv=4 head_dim=128 theta=500000
75
+ [2026-08-31 11:50:21] == s=4 T=16384 CP=1 (single-GPU) N=8 slots=1 ==
76
+ [2026-08-31 11:57:56] scoring s=4 cache=yarn -> probe_pos12288_armC/reach.csv
77
+ variant nll se score sig? notes
78
+ p0.6670 3.4169 0.5006 0.8317 YES
79
+ p1 1.2925 0.0859 1.0000 YES
80
+
81
+ floor(p1)=1.2925 absent=13.9188 null=13.4665 denom=12.6262 nats
82
+
83
+ wrote probe_pos12288_armC/reach.csv
84
+ [2026-08-31 11:57:56] SWEEP DONE -> probe_pos12288_armC/reach.csv (CSVs in probe_pos12288_armC/csv, markers in probe_pos12288_armC/state)
85
+ [11:57:56] -> probe_pos12288_armC/reach.csv
86
+ [11:57:56] ALL DONE