Run r3: logs and results
Browse files- runs/r3/logs/pipeline.log +1 -0
- runs/r3/logs/pipeline.out +1 -0
- runs/r3/logs/stats.log +2 -0
- runs/r3/logs/sync.log +1 -0
- runs/r3/logs/train.log +39 -0
runs/r3/logs/pipeline.log
CHANGED
|
@@ -19,3 +19,4 @@
|
|
| 19 |
[2026-11-19 02:10:26] done extract (2492s)
|
| 20 |
[2026-11-19 02:10:27] start ablate
|
| 21 |
[2026-11-19 02:49:15] done ablate (2328s)
|
|
|
|
|
|
| 19 |
[2026-11-19 02:10:26] done extract (2492s)
|
| 20 |
[2026-11-19 02:10:27] start ablate
|
| 21 |
[2026-11-19 02:49:15] done ablate (2328s)
|
| 22 |
+
[2026-11-19 02:49:16] start train
|
runs/r3/logs/pipeline.out
CHANGED
|
@@ -19,3 +19,4 @@
|
|
| 19 |
[2026-11-19 02:10:26] done extract (2492s)
|
| 20 |
[2026-11-19 02:10:27] start ablate
|
| 21 |
[2026-11-19 02:49:15] done ablate (2328s)
|
|
|
|
|
|
| 19 |
[2026-11-19 02:10:26] done extract (2492s)
|
| 20 |
[2026-11-19 02:10:27] start ablate
|
| 21 |
[2026-11-19 02:49:15] done ablate (2328s)
|
| 22 |
+
[2026-11-19 02:49:16] start train
|
runs/r3/logs/stats.log
CHANGED
|
@@ -33,3 +33,5 @@
|
|
| 33 |
[2026-11-19 02:36:19] ram 74.8/172 GB used (cache 0.1, shmem 0.0) | cgroup 74.8/172 GB | disk 141 GB used | gpu 42.5/95.6 GiB, 100%, 457 W | load 0.0 | largest python 81.0 GB, python 0.1 GB, python 0.1 GB
|
| 34 |
[2026-11-19 02:46:21] ram 74.8/172 GB used (cache 0.1, shmem 0.0) | cgroup 74.8/172 GB | disk 141 GB used | gpu 35.6/95.6 GiB, 100%, 460 W | load 0.0 | largest python 80.9 GB, python 0.1 GB, python 0.1 GB
|
| 35 |
[2026-11-19 02:49:15] ram 62.8/172 GB used (cache 0.0, shmem 0.0) | cgroup 62.8/172 GB | disk 141 GB used | gpu 0.0/95.6 GiB, 0%, 111 W | load 0.0 | largest python 0.1 GB, python 0.1 GB, python 0.0 GB
|
|
|
|
|
|
|
|
|
| 33 |
[2026-11-19 02:36:19] ram 74.8/172 GB used (cache 0.1, shmem 0.0) | cgroup 74.8/172 GB | disk 141 GB used | gpu 42.5/95.6 GiB, 100%, 457 W | load 0.0 | largest python 81.0 GB, python 0.1 GB, python 0.1 GB
|
| 34 |
[2026-11-19 02:46:21] ram 74.8/172 GB used (cache 0.1, shmem 0.0) | cgroup 74.8/172 GB | disk 141 GB used | gpu 35.6/95.6 GiB, 100%, 460 W | load 0.0 | largest python 80.9 GB, python 0.1 GB, python 0.1 GB
|
| 35 |
[2026-11-19 02:49:15] ram 62.8/172 GB used (cache 0.0, shmem 0.0) | cgroup 62.8/172 GB | disk 141 GB used | gpu 0.0/95.6 GiB, 0%, 111 W | load 0.0 | largest python 0.1 GB, python 0.1 GB, python 0.0 GB
|
| 36 |
+
[2026-11-19 02:49:16] ram 41.1/172 GB used (cache 0.0, shmem 0.0) | cgroup 41.1/172 GB | disk 141 GB used | gpu 0.0/95.6 GiB, 0%, 85 W | load 0.0 | largest python 0.1 GB, python 0.1 GB, python 0.0 GB
|
| 37 |
+
[2026-11-19 02:56:23] ram 74.8/172 GB used (cache 0.1, shmem 0.0) | cgroup 74.8/172 GB | disk 141 GB used | gpu 35.6/95.6 GiB, 100%, 463 W | load 0.0 | largest python 80.9 GB, python 0.1 GB, python 0.1 GB
|
runs/r3/logs/sync.log
CHANGED
|
@@ -23,3 +23,4 @@ uploaded logs/results to mudassiruddin21/EMC:runs/r3
|
|
| 23 |
uploaded logs/results to mudassiruddin21/EMC:runs/r3
|
| 24 |
uploaded logs/results to mudassiruddin21/EMC:runs/r3
|
| 25 |
uploaded logs/results to mudassiruddin21/EMC:runs/r3
|
|
|
|
|
|
| 23 |
uploaded logs/results to mudassiruddin21/EMC:runs/r3
|
| 24 |
uploaded logs/results to mudassiruddin21/EMC:runs/r3
|
| 25 |
uploaded logs/results to mudassiruddin21/EMC:runs/r3
|
| 26 |
+
uploaded logs/results to mudassiruddin21/EMC:runs/r3
|
runs/r3/logs/train.log
ADDED
|
@@ -0,0 +1,39 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
training recipe: --step-w auto --topk 10 --floor 0.3 --rel-lr 1e-3 --lora 128
|
| 2 |
+
[02:49:57] data: 14284 train / 160 val sequences (held on cpu), generated tokens by domain {'chat': 1382965, 'code': 669126, 'math': 1197671, 'longform': 2828495}
|
| 3 |
+
[02:50:05] [init] val tau 2.8371 | dev tau 2.8809 {'code': 3.753, 'chat': 3.057, 'math': 3.678, 'longform': 2.479} steps [0.851, 0.791, 0.788]
|
| 4 |
+
[02:50:05] equalisation factors {'g': 0.9997948828943418, 'gamma': 0.9999797389881102, 's': 0.9997603272856522, 'beta': 1.0000041710353693, 'c': 0.9997597321195537, 'a_k': 0.9997507954939153, 'max_rel_drift': 5.650238108501071e-06} (must be function-preserving)
|
| 5 |
+
[02:50:12] [equalized] val tau 2.8381 | dev tau 2.8801 {'code': 3.754, 'chat': 3.052, 'math': 3.675, 'longform': 2.48} steps [0.85, 0.791, 0.788]
|
| 6 |
+
[02:50:22] [fp8_untrained] val tau 2.8376 | dev tau 2.8808 {'code': 3.747, 'chat': 3.058, 'math': 3.68, 'longform': 2.479} steps [0.85, 0.791, 0.788]
|
| 7 |
+
[02:50:22] LoRA rank 128: 15.9 M trainable parameters
|
| 8 |
+
[02:50:25] training: 1790 steps planned (895/epoch), qat=True, step weights auto, top-k 10, floor 0.3, rel-lr 0.001
|
| 9 |
+
[02:50:26] step 1/1790 ce 0.716/1.158/1.457 acc 0.780/0.677/0.631 batch-tau~2.641 lr x0.02 1.78s/step w [1.0, 0.63, 0.28]
|
| 10 |
+
[02:50:38] step 10/1790 ce 0.064/0.128/0.189 acc 0.979/0.959/0.939 batch-tau~3.799 lr x0.17 1.31s/step w [1.0, 0.63, 0.28]
|
| 11 |
+
[02:50:51] step 20/1790 ce 0.422/0.717/0.948 acc 0.874/0.804/0.774 batch-tau~3.122 lr x0.34 1.31s/step w [1.0, 0.63, 0.28]
|
| 12 |
+
[02:51:03] step 30/1790 ce 0.573/0.912/1.111 acc 0.829/0.766/0.747 batch-tau~2.938 lr x0.51 1.29s/step w [1.0, 0.63, 0.28]
|
| 13 |
+
[02:51:16] step 40/1790 ce 0.085/0.148/0.200 acc 0.975/0.963/0.952 batch-tau~3.807 lr x0.68 1.29s/step w [1.0, 0.62, 0.28]
|
| 14 |
+
[02:51:29] step 50/1790 ce 0.126/0.229/0.299 acc 0.958/0.932/0.916 batch-tau~3.668 lr x0.85 1.28s/step w [1.0, 0.63, 0.28]
|
| 15 |
+
[02:51:41] step 60/1790 ce 0.740/1.263/1.592 acc 0.780/0.668/0.604 batch-tau~2.616 lr x1.00 1.27s/step w [1.0, 0.63, 0.28]
|
| 16 |
+
[02:51:47] eval step 60: val tau 2.8410 | dev(1/4) tau 2.8985 (best 2.8376 @ 0)
|
| 17 |
+
[02:52:00] step 70/1790 ce 0.300/0.480/0.555 acc 0.906/0.868/0.862 batch-tau~3.370 lr x1.00 1.36s/step w [1.0, 0.63, 0.28]
|
| 18 |
+
[02:52:13] step 80/1790 ce 0.795/1.302/1.596 acc 0.767/0.658/0.614 batch-tau~2.582 lr x1.00 1.35s/step w [1.0, 0.63, 0.28]
|
| 19 |
+
[02:52:25] step 90/1790 ce 0.863/1.395/1.693 acc 0.746/0.635/0.593 batch-tau~2.500 lr x1.00 1.34s/step w [1.0, 0.63, 0.28]
|
| 20 |
+
[02:52:38] step 100/1790 ce 0.769/1.302/1.649 acc 0.761/0.647/0.592 batch-tau~2.545 lr x1.00 1.34s/step w [1.0, 0.63, 0.28]
|
| 21 |
+
[02:52:50] step 110/1790 ce 0.496/0.887/1.117 acc 0.845/0.769/0.738 batch-tau~2.975 lr x1.00 1.32s/step w [1.0, 0.63, 0.28]
|
| 22 |
+
[02:53:03] step 120/1790 ce 0.664/1.111/1.391 acc 0.794/0.708/0.662 batch-tau~2.729 lr x1.00 1.32s/step w [1.0, 0.63, 0.28]
|
| 23 |
+
[02:53:09] eval step 120: val tau 2.8411 | dev(1/4) tau 2.8933 (best 2.8410 @ 60)
|
| 24 |
+
[02:53:23] step 130/1790 ce 0.800/1.352/1.681 acc 0.758/0.644/0.593 batch-tau~2.535 lr x1.00 1.37s/step w [1.0, 0.63, 0.28]
|
| 25 |
+
[02:53:36] step 140/1790 ce 0.157/0.250/0.323 acc 0.948/0.925/0.909 batch-tau~3.623 lr x1.00 1.36s/step w [1.0, 0.63, 0.28]
|
| 26 |
+
[02:53:48] step 150/1790 ce 0.807/1.363/1.695 acc 0.761/0.656/0.601 batch-tau~2.559 lr x0.99 1.36s/step w [1.0, 0.63, 0.28]
|
| 27 |
+
[02:54:01] step 160/1790 ce 0.690/1.163/1.446 acc 0.788/0.683/0.638 batch-tau~2.670 lr x0.99 1.35s/step w [1.0, 0.62, 0.28]
|
| 28 |
+
[02:54:14] step 170/1790 ce 0.676/1.123/1.325 acc 0.798/0.710/0.699 batch-tau~2.760 lr x0.99 1.35s/step w [1.0, 0.62, 0.27]
|
| 29 |
+
[02:54:27] step 180/1790 ce 0.770/1.276/1.628 acc 0.773/0.663/0.608 batch-tau~2.598 lr x0.99 1.35s/step w [1.0, 0.63, 0.28]
|
| 30 |
+
[02:54:33] eval step 180: val tau 2.8412 | dev(1/4) tau 2.9070 (best 2.8411 @ 120)
|
| 31 |
+
[02:54:45] step 190/1790 ce 0.175/0.297/0.392 acc 0.945/0.915/0.899 batch-tau~3.586 lr x0.99 1.37s/step w [1.0, 0.63, 0.28]
|
| 32 |
+
[02:54:58] step 200/1790 ce 0.575/0.966/1.180 acc 0.825/0.750/0.719 batch-tau~2.889 lr x0.99 1.37s/step w [1.0, 0.63, 0.28]
|
| 33 |
+
[02:55:11] step 210/1790 ce 0.222/0.379/0.457 acc 0.932/0.896/0.885 batch-tau~3.506 lr x0.98 1.36s/step w [1.0, 0.63, 0.28]
|
| 34 |
+
[02:55:23] step 220/1790 ce 0.492/0.853/1.079 acc 0.853/0.779/0.742 batch-tau~3.011 lr x0.98 1.36s/step w [1.0, 0.63, 0.28]
|
| 35 |
+
[02:55:37] step 230/1790 ce 0.797/1.334/1.678 acc 0.762/0.655/0.597 batch-tau~2.559 lr x0.98 1.36s/step w [1.0, 0.63, 0.28]
|
| 36 |
+
[02:55:50] step 240/1790 ce 0.803/1.370/1.707 acc 0.756/0.639/0.591 batch-tau~2.525 lr x0.98 1.35s/step w [1.0, 0.63, 0.28]
|
| 37 |
+
[02:55:56] eval step 240: val tau 2.8427 | dev(1/4) tau 2.9140 (best 2.8412 @ 180)
|
| 38 |
+
[02:56:09] step 250/1790 ce 0.144/0.237/0.334 acc 0.954/0.933/0.908 batch-tau~3.653 lr x0.97 1.38s/step w [1.0, 0.63, 0.28]
|
| 39 |
+
[02:56:22] step 260/1790 ce 0.778/1.284/1.608 acc 0.762/0.651/0.597 batch-tau~2.553 lr x0.97 1.37s/step w [1.0, 0.63, 0.28]
|