Upload folder using huggingface_hub
Browse files
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
|