Upload folder using huggingface_hub
Browse files- logs/in2/arm_a.trained +1 -0
- logs/in2/arm_a_train.log +0 -0
- logs/in2/arm_b.trained +1 -0
- logs/in2/arm_b_train.log +693 -0
- logs/in2/arm_c.trained +1 -0
- logs/in2/arm_c_train.log +297 -0
- logs/in2/driver_20260829_1535.log +114 -0
- logs/in2/driver_20260829_1542.log +92 -0
- logs/in2/driver_20260829_1545.log +94 -0
- logs/in2/driver_20260829_1548.log +100 -0
logs/in2/arm_a.trained
ADDED
|
@@ -0,0 +1 @@
|
|
|
|
|
|
|
| 1 |
+
12
|
logs/in2/arm_a_train.log
ADDED
|
The diff for this file is too large to render.
See raw diff
|
|
|
logs/in2/arm_b.trained
ADDED
|
@@ -0,0 +1 @@
|
|
|
|
|
|
|
| 1 |
+
13
|
logs/in2/arm_b_train.log
ADDED
|
@@ -0,0 +1,693 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
=== Llama Context Parallel Training Script (bluscriptCP) ===
|
| 2 |
+
[CP attn fusion] split (Stage-2 rope-once + Stage-3 Hopper causal-attn kernels)
|
| 3 |
+
[ZeRO-1] CP_ZERO_SYMMETRIC=1: NCCL symmetric memory mode (NVLS)
|
| 4 |
+
[watchdog] collective no-progress deadline 600 s, poll 2 s (BLU_NCCL_WATCHDOG=0 to disable)
|
| 5 |
+
[watchdog] collective no-progress deadline 600 s, poll 2 s (BLU_NCCL_WATCHDOG=0 to disable)
|
| 6 |
+
================ Configuration (bluscriptCP) ================
|
| 7 |
+
d_model/layers : 1536 / 36
|
| 8 |
+
q/kv heads : 12 / 4 (G=3, hd=128)
|
| 9 |
+
ffn_hidden : 4096 (SwiGLU)
|
| 10 |
+
context_length : 16384
|
| 11 |
+
vocab_size : 50304 tying=0
|
| 12 |
+
B x T : 4 x 16384
|
| 13 |
+
world_size : 4 (dp=1 x cp=4)
|
| 14 |
+
CP_ATTN_MODE : ring (load_balance=HeadTail)
|
| 15 |
+
global_batch : 524288 (grad_accum=8)
|
| 16 |
+
optimizer : Muon (2-D body) + AdamW (wte + 1-D) [ZeRO-1] muon_lr=0.001000
|
| 17 |
+
precision : bf16 Linears + fp32 norms/residual/logits (fp32 master weights in AdamW) [mask=15 QKV WO MLP HEAD]
|
| 18 |
+
max_lr/min_lr : 0.001 / 0.0001
|
| 19 |
+
warmup/steps : 1 / 400 (cosine, no stable phase)
|
| 20 |
+
checkpointing : true (ckpt_freq=100, dir=checkpoints_bluscriptcp)
|
| 21 |
+
=============================================================
|
| 22 |
+
Initializing model on CUDA device 0...
|
| 23 |
+
[YaRN][warn] seq_len(16384) != scale(4.000)*orig_maxpos(1024.0)=4096.0 -- cache scaled for a different context than it is sized for
|
| 24 |
+
[YaRN] build_rope_cache: seq_len=16384 hd=128 scale=4.000 low=7 high=25 m=1.1386
|
| 25 |
+
[InitFrom] weights loaded from /root/Adhi/BluTrain/dist/Context_Parallelism/checkpoints_bluscriptcp/rank_0/blumodelcp_run7_step_750.ckpt (base step 750, loss 1.28324) -- optimizer state and RNG intentionally NOT restored; pair with CP_REWARMUP
|
| 26 |
+
Parameters: 1060624896
|
| 27 |
+
Parameters per GPU: 1060624896
|
| 28 |
+
max_steps: 400
|
| 29 |
+
warmup_steps: 1
|
| 30 |
+
[Muon] adjust_lr=match_rms_adamw (default) ns_steps=5 momentum=0.95
|
| 31 |
+
[ZeRO-1/Muon] reduce-scatter/all-gather layout: 6 bucket(s), 12 collectives/step, shard 226492416 elems/rank, padding 0.000000% (flat buffer 1728.000000 MB)
|
| 32 |
+
[ZeRO-1/Muon] adjust_lr=match_rms_adamw | params=905969664 (216 orthogonalized / 0 AdamW) muon_elems=905969664 | W=4 owned=226492416 (42 tensors) | master+momentum: 1728.000000 MB/GPU
|
| 33 |
+
[ZeRO-1 v2] build=v2.4 supports: rs+ag | can-symmetric-window | can-fp32-reduce | can-bf16-reduce | can-reshard-v1-ckpt
|
| 34 |
+
[ZeRO-1 v2] ACTIVE: reduce_dtype=bfloat16 (= param dtype) symmetric=ON overlap=-/-
|
| 35 |
+
[ZeRO-1 v2] zero1-v2 reduce-scatter/all-gather (symmetric/NVLS)
|
| 36 |
+
[ZeRO-1 v2] params=77266944 padded=77266944 W=4 shard=19316736 | 3 buckets of 50.000000 MB (6553600 elems/rank/bucket) | optimizer state/GPU: 221.062500 MB (vs 884.250000 MB unsharded)
|
| 37 |
+
[ZeRO-1 v2] build=v2.4 supports: rs+ag | can-symmetric-window | can-fp32-reduce | can-bf16-reduce | can-reshard-v1-ckpt
|
| 38 |
+
[ZeRO-1 v2] ACTIVE: reduce_dtype=float32 (= param dtype) symmetric=ON overlap=-/-
|
| 39 |
+
[ZeRO-1 v2] zero1-v2 reduce-scatter/all-gather (symmetric/NVLS)
|
| 40 |
+
[ZeRO-1 v2] params=77388288 padded=77388288 W=4 shard=19347072 | 6 buckets of 50.000000 MB (3276800 elems/rank/bucket) | optimizer state/GPU: 221.409668 MB (vs 885.638672 MB unsharded)
|
| 41 |
+
[ZeRO-1] variant=v2 reduce-scatter/all-gather groups=3 [muon:bfloat16/905M] [adamw:bfloat16/77266944] [adamw:float32/77388288]
|
| 42 |
+
[ZeRO-1] DataParallel is DISABLED (both would double-reduce)
|
| 43 |
+
[precision] bf16 params=217, all with fp32 master weights
|
| 44 |
+
found 3 shards for split train
|
| 45 |
+
found 1 shards for split val
|
| 46 |
+
Saving logs to: CP_BluScreipt_Training_logs/CP_Training_log13.csv (run 13)
|
| 47 |
+
[rope_theta] 500000 compiled default (no sidecar -- legacy run, assuming it was trained at this base)
|
| 48 |
+
Checkpointing: ON -> checkpoints_bluscriptcp/blumodelcp_run13_step_<N>.ckpt (every 100 steps, keep 2)
|
| 49 |
+
[re-warmup] anchored at step 0 for 100 steps -> peak 0.000300 (0.300000 x max_lr), then cosine-decay to 0.000100 | resuming at step 0 (mid-ramp)
|
| 50 |
+
|
| 51 |
+
Starting training...
|
| 52 |
+
|
| 53 |
+
[CP ring P2P] OVERLAP: dedicated non-blocking stream
|
| 54 |
+
Current shard: /root/Adhi/BluTrain/dist/Context_Parallelism/in2_data/arm_b/in2b_val_00000.bin
|
| 55 |
+
100000000
|
| 56 |
+
validation loss: 1.6834
|
| 57 |
+
[CP bwd] overlap=ON
|
| 58 |
+
step 0 | loss: 1.719764 | lr 3.0000e-06 | norm: 0.9177 | dt: 15904.57ms | tok/sec: 32964.6 | mem: 107213MB | Time Left: 01 hrs : 46 mins
|
| 59 |
+
[TIMING] data: 3.0ms | fwd: 4237.0ms | loss: 6.5ms | bwd: 4571.0ms | clip: 0.0ms | optim: 62.6ms
|
| 60 |
+
step 1 | loss: 1.689209 | lr 6.0000e-06 | norm: 0.9809 | dt: 8602.98ms | tok/sec: 60942.6 | mem: 108909MB | Time Left: 00 hrs : 57 mins
|
| 61 |
+
[TIMING] data: 3.5ms | fwd: 4004.7ms | loss: 6.4ms | bwd: 4530.9ms | clip: 0.0ms | optim: 49.6ms
|
| 62 |
+
step 2 | loss: 1.710424 | lr 9.0000e-06 | norm: 1.2316 | dt: 8313.79ms | tok/sec: 63062.5 | mem: 108909MB | Time Left: 00 hrs : 55 mins
|
| 63 |
+
[TIMING] data: 4.5ms | fwd: 4090.5ms | loss: 6.3ms | bwd: 4154.7ms | clip: 0.0ms | optim: 49.5ms
|
| 64 |
+
step 3 | loss: 1.731818 | lr 1.2000e-05 | norm: 0.9002 | dt: 6208.15ms | tok/sec: 84451.6 | mem: 108909MB | Time Left: 00 hrs : 41 mins
|
| 65 |
+
[TIMING] data: 3.5ms | fwd: 1990.5ms | loss: 6.4ms | bwd: 4150.5ms | clip: 0.0ms | optim: 49.4ms
|
| 66 |
+
step 4 | loss: 1.702109 | lr 1.5000e-05 | norm: 0.9435 | dt: 6212.17ms | tok/sec: 84396.9 | mem: 108909MB | Time Left: 00 hrs : 41 mins
|
| 67 |
+
[TIMING] data: 3.4ms | fwd: 1991.1ms | loss: 6.4ms | bwd: 4153.8ms | clip: 0.0ms | optim: 49.5ms
|
| 68 |
+
step 5 | loss: 1.766153 | lr 1.8000e-05 | norm: 0.8298 | dt: 6250.52ms | tok/sec: 83879.1 | mem: 108909MB | Time Left: 00 hrs : 41 mins
|
| 69 |
+
[TIMING] data: 3.4ms | fwd: 2008.1ms | loss: 6.4ms | bwd: 4175.3ms | clip: 0.0ms | optim: 49.4ms
|
| 70 |
+
step 6 | loss: 1.703925 | lr 2.1000e-05 | norm: 0.9083 | dt: 6372.13ms | tok/sec: 82278.3 | mem: 108909MB | Time Left: 00 hrs : 41 mins
|
| 71 |
+
[TIMING] data: 3.3ms | fwd: 2084.4ms | loss: 6.4ms | bwd: 4220.8ms | clip: 0.0ms | optim: 49.3ms
|
| 72 |
+
step 7 | loss: 1.660973 | lr 2.4000e-05 | norm: 0.7858 | dt: 6383.07ms | tok/sec: 82137.3 | mem: 108909MB | Time Left: 00 hrs : 41 mins
|
| 73 |
+
[TIMING] data: 3.3ms | fwd: 2029.9ms | loss: 6.4ms | bwd: 4286.0ms | clip: 0.0ms | optim: 49.6ms
|
| 74 |
+
step 8 | loss: 1.628360 | lr 2.7000e-05 | norm: 0.7381 | dt: 6420.10ms | tok/sec: 81663.6 | mem: 108909MB | Time Left: 00 hrs : 41 mins
|
| 75 |
+
[TIMING] data: 3.4ms | fwd: 2036.6ms | loss: 6.4ms | bwd: 4316.3ms | clip: 0.0ms | optim: 49.5ms
|
| 76 |
+
step 9 | loss: 1.665800 | lr 3.0000e-05 | norm: 0.7373 | dt: 6401.36ms | tok/sec: 81902.6 | mem: 108909MB | Time Left: 00 hrs : 41 mins
|
| 77 |
+
[TIMING] data: 3.5ms | fwd: 2035.0ms | loss: 6.4ms | bwd: 4299.1ms | clip: 0.0ms | optim: 49.5ms
|
| 78 |
+
step 10 | loss: 1.610076 | lr 3.3000e-05 | norm: 0.8668 | dt: 6412.79ms | tok/sec: 81756.6 | mem: 108909MB | Time Left: 00 hrs : 41 mins
|
| 79 |
+
[TIMING] data: 3.4ms | fwd: 2034.6ms | loss: 6.4ms | bwd: 4310.9ms | clip: 0.0ms | optim: 49.5ms
|
| 80 |
+
step 11 | loss: 1.618258 | lr 3.6000e-05 | norm: 0.7161 | dt: 6374.74ms | tok/sec: 82244.7 | mem: 108909MB | Time Left: 00 hrs : 41 mins
|
| 81 |
+
[TIMING] data: 3.6ms | fwd: 2030.3ms | loss: 6.4ms | bwd: 4277.1ms | clip: 0.0ms | optim: 49.4ms
|
| 82 |
+
step 12 | loss: 1.605687 | lr 3.9000e-05 | norm: 0.7537 | dt: 6461.94ms | tok/sec: 81134.8 | mem: 108909MB | Time Left: 00 hrs : 41 mins
|
| 83 |
+
[TIMING] data: 3.5ms | fwd: 2101.0ms | loss: 6.4ms | bwd: 4293.6ms | clip: 0.0ms | optim: 49.5ms
|
| 84 |
+
step 13 | loss: 1.612419 | lr 4.2000e-05 | norm: 0.8240 | dt: 6356.31ms | tok/sec: 82483.1 | mem: 108909MB | Time Left: 00 hrs : 40 mins
|
| 85 |
+
[TIMING] data: 3.5ms | fwd: 2021.3ms | loss: 6.4ms | bwd: 4267.4ms | clip: 0.0ms | optim: 49.7ms
|
| 86 |
+
step 14 | loss: 1.638093 | lr 4.5000e-05 | norm: 0.8258 | dt: 6346.77ms | tok/sec: 82607.0 | mem: 108909MB | Time Left: 00 hrs : 40 mins
|
| 87 |
+
[TIMING] data: 3.5ms | fwd: 2020.4ms | loss: 6.4ms | bwd: 4258.5ms | clip: 0.0ms | optim: 50.0ms
|
| 88 |
+
step 15 | loss: 1.627281 | lr 4.8000e-05 | norm: 0.8539 | dt: 6335.85ms | tok/sec: 82749.5 | mem: 108909MB | Time Left: 00 hrs : 40 mins
|
| 89 |
+
[TIMING] data: 3.5ms | fwd: 2022.1ms | loss: 6.4ms | bwd: 4246.4ms | clip: 0.0ms | optim: 49.6ms
|
| 90 |
+
step 16 | loss: 1.602286 | lr 5.1000e-05 | norm: 0.8507 | dt: 6329.67ms | tok/sec: 82830.3 | mem: 108909MB | Time Left: 00 hrs : 40 mins
|
| 91 |
+
[TIMING] data: 3.5ms | fwd: 2024.3ms | loss: 6.4ms | bwd: 4238.2ms | clip: 0.0ms | optim: 49.3ms
|
| 92 |
+
step 17 | loss: 1.543830 | lr 5.4000e-05 | norm: 0.7251 | dt: 6331.24ms | tok/sec: 82809.7 | mem: 108909MB | Time Left: 00 hrs : 40 mins
|
| 93 |
+
[TIMING] data: 3.5ms | fwd: 2029.2ms | loss: 6.4ms | bwd: 4234.7ms | clip: 0.0ms | optim: 49.4ms
|
| 94 |
+
step 18 | loss: 1.574815 | lr 5.7000e-05 | norm: 0.6813 | dt: 6321.41ms | tok/sec: 82938.5 | mem: 108909MB | Time Left: 00 hrs : 40 mins
|
| 95 |
+
[TIMING] data: 3.5ms | fwd: 2008.9ms | loss: 6.4ms | bwd: 4245.3ms | clip: 0.0ms | optim: 49.4ms
|
| 96 |
+
step 19 | loss: 1.587047 | lr 6.0000e-05 | norm: 0.7060 | dt: 6342.47ms | tok/sec: 82663.1 | mem: 108909MB | Time Left: 00 hrs : 40 mins
|
| 97 |
+
[TIMING] data: 3.4ms | fwd: 2014.7ms | loss: 6.4ms | bwd: 4260.5ms | clip: 0.0ms | optim: 49.4ms
|
| 98 |
+
step 20 | loss: 1.558500 | lr 6.3000e-05 | norm: 0.8647 | dt: 6334.12ms | tok/sec: 82772.0 | mem: 108909MB | Time Left: 00 hrs : 40 mins
|
| 99 |
+
[TIMING] data: 3.4ms | fwd: 2016.2ms | loss: 6.4ms | bwd: 4250.7ms | clip: 0.0ms | optim: 49.4ms
|
| 100 |
+
step 21 | loss: 1.628016 | lr 6.6000e-05 | norm: 0.8759 | dt: 6310.75ms | tok/sec: 83078.6 | mem: 108909MB | Time Left: 00 hrs : 39 mins
|
| 101 |
+
[TIMING] data: 3.4ms | fwd: 2009.1ms | loss: 6.4ms | bwd: 4234.4ms | clip: 0.0ms | optim: 49.4ms
|
| 102 |
+
step 22 | loss: 1.556988 | lr 6.9000e-05 | norm: 0.7970 | dt: 6326.02ms | tok/sec: 82878.0 | mem: 108909MB | Time Left: 00 hrs : 39 mins
|
| 103 |
+
[TIMING] data: 3.5ms | fwd: 2009.3ms | loss: 6.4ms | bwd: 4249.5ms | clip: 0.0ms | optim: 49.5ms
|
| 104 |
+
step 23 | loss: 1.555099 | lr 7.2000e-05 | norm: 0.8540 | dt: 6324.02ms | tok/sec: 82904.2 | mem: 108909MB | Time Left: 00 hrs : 39 mins
|
| 105 |
+
[TIMING] data: 3.5ms | fwd: 2013.5ms | loss: 6.4ms | bwd: 4243.2ms | clip: 0.0ms | optim: 49.4ms
|
| 106 |
+
step 24 | loss: 1.524715 | lr 7.5000e-05 | norm: 0.6101 | dt: 6328.00ms | tok/sec: 82852.1 | mem: 108909MB | Time Left: 00 hrs : 39 mins
|
| 107 |
+
[TIMING] data: 3.4ms | fwd: 2036.3ms | loss: 6.4ms | bwd: 4224.5ms | clip: 0.0ms | optim: 49.4ms
|
| 108 |
+
step 25 | loss: 1.595337 | lr 7.8000e-05 | norm: 0.6755 | dt: 6358.19ms | tok/sec: 82458.7 | mem: 108909MB | Time Left: 00 hrs : 39 mins
|
| 109 |
+
[TIMING] data: 3.5ms | fwd: 2036.2ms | loss: 6.4ms | bwd: 4254.8ms | clip: 0.0ms | optim: 49.4ms
|
| 110 |
+
step 26 | loss: 1.504139 | lr 8.1000e-05 | norm: 0.9090 | dt: 6344.17ms | tok/sec: 82641.0 | mem: 108909MB | Time Left: 00 hrs : 39 mins
|
| 111 |
+
[TIMING] data: 3.5ms | fwd: 2056.7ms | loss: 6.4ms | bwd: 4220.2ms | clip: 0.0ms | optim: 49.4ms
|
| 112 |
+
step 27 | loss: 1.509489 | lr 8.4000e-05 | norm: 0.6161 | dt: 6334.27ms | tok/sec: 82770.1 | mem: 108909MB | Time Left: 00 hrs : 39 mins
|
| 113 |
+
[TIMING] data: 3.5ms | fwd: 2021.7ms | loss: 6.4ms | bwd: 4244.9ms | clip: 0.0ms | optim: 49.8ms
|
| 114 |
+
step 28 | loss: 1.541055 | lr 8.7000e-05 | norm: 0.7380 | dt: 6335.22ms | tok/sec: 82757.6 | mem: 108909MB | Time Left: 00 hrs : 39 mins
|
| 115 |
+
[TIMING] data: 3.4ms | fwd: 2021.2ms | loss: 6.4ms | bwd: 4246.9ms | clip: 0.0ms | optim: 49.4ms
|
| 116 |
+
step 29 | loss: 1.520158 | lr 9.0000e-05 | norm: 0.6387 | dt: 6310.05ms | tok/sec: 83087.7 | mem: 108909MB | Time Left: 00 hrs : 39 mins
|
| 117 |
+
[TIMING] data: 3.5ms | fwd: 2014.5ms | loss: 6.4ms | bwd: 4228.3ms | clip: 0.0ms | optim: 49.5ms
|
| 118 |
+
step 30 | loss: 1.497158 | lr 9.3000e-05 | norm: 0.6212 | dt: 6331.83ms | tok/sec: 82801.9 | mem: 108909MB | Time Left: 00 hrs : 39 mins
|
| 119 |
+
[TIMING] data: 3.5ms | fwd: 2019.8ms | loss: 6.4ms | bwd: 4244.8ms | clip: 0.0ms | optim: 49.4ms
|
| 120 |
+
step 31 | loss: 1.558612 | lr 9.6000e-05 | norm: 0.6279 | dt: 6387.33ms | tok/sec: 82082.4 | mem: 108909MB | Time Left: 00 hrs : 39 mins
|
| 121 |
+
[TIMING] data: 3.5ms | fwd: 2064.0ms | loss: 6.4ms | bwd: 4256.1ms | clip: 0.0ms | optim: 49.4ms
|
| 122 |
+
step 32 | loss: 1.469230 | lr 9.9000e-05 | norm: 0.6770 | dt: 6318.88ms | tok/sec: 82971.6 | mem: 108909MB | Time Left: 00 hrs : 38 mins
|
| 123 |
+
[TIMING] data: 3.5ms | fwd: 2014.8ms | loss: 6.4ms | bwd: 4236.7ms | clip: 0.0ms | optim: 49.6ms
|
| 124 |
+
step 33 | loss: 1.531537 | lr 1.0200e-04 | norm: 0.6956 | dt: 6318.93ms | tok/sec: 82971.0 | mem: 108909MB | Time Left: 00 hrs : 38 mins
|
| 125 |
+
[TIMING] data: 3.4ms | fwd: 2017.4ms | loss: 6.4ms | bwd: 4234.6ms | clip: 0.0ms | optim: 49.3ms
|
| 126 |
+
step 34 | loss: 1.509960 | lr 1.0500e-04 | norm: 1.2464 | dt: 6330.87ms | tok/sec: 82814.5 | mem: 108909MB | Time Left: 00 hrs : 38 mins
|
| 127 |
+
[TIMING] data: 3.4ms | fwd: 2013.2ms | loss: 6.4ms | bwd: 4250.5ms | clip: 0.0ms | optim: 49.4ms
|
| 128 |
+
step 35 | loss: 1.521332 | lr 1.0800e-04 | norm: 0.6600 | dt: 6320.12ms | tok/sec: 82955.4 | mem: 108909MB | Time Left: 00 hrs : 38 mins
|
| 129 |
+
[TIMING] data: 3.4ms | fwd: 2014.6ms | loss: 6.4ms | bwd: 4238.4ms | clip: 0.0ms | optim: 49.4ms
|
| 130 |
+
step 36 | loss: 1.499588 | lr 1.1100e-04 | norm: 0.7906 | dt: 6328.81ms | tok/sec: 82841.5 | mem: 108909MB | Time Left: 00 hrs : 38 mins
|
| 131 |
+
[TIMING] data: 3.4ms | fwd: 2018.6ms | loss: 6.4ms | bwd: 4243.0ms | clip: 0.0ms | optim: 49.4ms
|
| 132 |
+
step 37 | loss: 1.499973 | lr 1.1400e-04 | norm: 0.6590 | dt: 6330.89ms | tok/sec: 82814.3 | mem: 108909MB | Time Left: 00 hrs : 38 mins
|
| 133 |
+
[TIMING] data: 3.4ms | fwd: 2018.6ms | loss: 6.4ms | bwd: 4245.1ms | clip: 0.0ms | optim: 49.4ms
|
| 134 |
+
step 38 | loss: 1.491481 | lr 1.1700e-04 | norm: 0.8726 | dt: 6324.11ms | tok/sec: 82903.0 | mem: 108909MB | Time Left: 00 hrs : 38 mins
|
| 135 |
+
[TIMING] data: 3.5ms | fwd: 2010.4ms | loss: 6.4ms | bwd: 4246.4ms | clip: 0.0ms | optim: 49.5ms
|
| 136 |
+
step 39 | loss: 1.515249 | lr 1.2000e-04 | norm: 0.6338 | dt: 6316.15ms | tok/sec: 83007.5 | mem: 108909MB | Time Left: 00 hrs : 38 mins
|
| 137 |
+
[TIMING] data: 3.5ms | fwd: 2014.1ms | loss: 6.4ms | bwd: 4234.7ms | clip: 0.0ms | optim: 49.4ms
|
| 138 |
+
step 40 | loss: 1.475689 | lr 1.2300e-04 | norm: 0.7981 | dt: 6356.20ms | tok/sec: 82484.5 | mem: 108909MB | Time Left: 00 hrs : 38 mins
|
| 139 |
+
[TIMING] data: 3.5ms | fwd: 2020.7ms | loss: 6.7ms | bwd: 4268.0ms | clip: 0.0ms | optim: 49.4ms
|
| 140 |
+
step 41 | loss: 1.453508 | lr 1.2600e-04 | norm: 0.6444 | dt: 6302.20ms | tok/sec: 83191.3 | mem: 108909MB | Time Left: 00 hrs : 37 mins
|
| 141 |
+
[TIMING] data: 3.4ms | fwd: 2008.4ms | loss: 6.4ms | bwd: 4226.6ms | clip: 0.0ms | optim: 49.4ms
|
| 142 |
+
step 42 | loss: 1.502223 | lr 1.2900e-04 | norm: 0.6451 | dt: 6360.25ms | tok/sec: 82431.9 | mem: 108909MB | Time Left: 00 hrs : 37 mins
|
| 143 |
+
[TIMING] data: 3.4ms | fwd: 2054.1ms | loss: 6.4ms | bwd: 4239.1ms | clip: 0.0ms | optim: 49.3ms
|
| 144 |
+
step 43 | loss: 1.504963 | lr 1.3200e-04 | norm: 0.6668 | dt: 6317.30ms | tok/sec: 82992.4 | mem: 108909MB | Time Left: 00 hrs : 37 mins
|
| 145 |
+
[TIMING] data: 3.5ms | fwd: 2015.3ms | loss: 6.4ms | bwd: 4234.8ms | clip: 0.0ms | optim: 49.4ms
|
| 146 |
+
step 44 | loss: 1.425290 | lr 1.3500e-04 | norm: 1.2208 | dt: 6359.73ms | tok/sec: 82438.7 | mem: 108909MB | Time Left: 00 hrs : 37 mins
|
| 147 |
+
[TIMING] data: 3.5ms | fwd: 2050.7ms | loss: 6.4ms | bwd: 4241.5ms | clip: 0.0ms | optim: 49.6ms
|
| 148 |
+
step 45 | loss: 1.504238 | lr 1.3800e-04 | norm: 0.6194 | dt: 6322.99ms | tok/sec: 82917.7 | mem: 108909MB | Time Left: 00 hrs : 37 mins
|
| 149 |
+
[TIMING] data: 3.5ms | fwd: 2016.5ms | loss: 6.4ms | bwd: 4238.9ms | clip: 0.0ms | optim: 49.8ms
|
| 150 |
+
step 46 | loss: 1.491822 | lr 1.4100e-04 | norm: 0.6576 | dt: 6328.94ms | tok/sec: 82839.7 | mem: 108909MB | Time Left: 00 hrs : 37 mins
|
| 151 |
+
[TIMING] data: 3.4ms | fwd: 2018.7ms | loss: 6.4ms | bwd: 4243.1ms | clip: 0.0ms | optim: 49.4ms
|
| 152 |
+
step 47 | loss: 1.515073 | lr 1.4400e-04 | norm: 0.7951 | dt: 6324.23ms | tok/sec: 82901.5 | mem: 108909MB | Time Left: 00 hrs : 37 mins
|
| 153 |
+
[TIMING] data: 3.5ms | fwd: 2010.3ms | loss: 6.4ms | bwd: 4246.7ms | clip: 0.0ms | optim: 49.3ms
|
| 154 |
+
step 48 | loss: 1.509569 | lr 1.4700e-04 | norm: 0.9722 | dt: 6319.16ms | tok/sec: 82968.0 | mem: 108909MB | Time Left: 00 hrs : 37 mins
|
| 155 |
+
[TIMING] data: 3.5ms | fwd: 2021.6ms | loss: 6.4ms | bwd: 4230.4ms | clip: 0.0ms | optim: 49.4ms
|
| 156 |
+
step 49 | loss: 1.487431 | lr 1.5000e-04 | norm: 0.6924 | dt: 6332.06ms | tok/sec: 82798.9 | mem: 108909MB | Time Left: 00 hrs : 37 mins
|
| 157 |
+
[TIMING] data: 3.5ms | fwd: 2020.5ms | loss: 6.7ms | bwd: 4243.8ms | clip: 0.0ms | optim: 49.6ms
|
| 158 |
+
step 50 | loss: 1.519441 | lr 1.5300e-04 | norm: 0.6099 | dt: 6325.66ms | tok/sec: 82882.8 | mem: 108909MB | Time Left: 00 hrs : 36 mins
|
| 159 |
+
[TIMING] data: 3.6ms | fwd: 2024.9ms | loss: 6.4ms | bwd: 4233.1ms | clip: 0.0ms | optim: 49.7ms
|
| 160 |
+
step 51 | loss: 1.539274 | lr 1.5600e-04 | norm: 0.6793 | dt: 6339.00ms | tok/sec: 82708.4 | mem: 108909MB | Time Left: 00 hrs : 36 mins
|
| 161 |
+
[TIMING] data: 3.6ms | fwd: 2011.3ms | loss: 6.4ms | bwd: 4260.0ms | clip: 0.0ms | optim: 49.9ms
|
| 162 |
+
step 52 | loss: 1.548381 | lr 1.5900e-04 | norm: 0.5928 | dt: 6296.16ms | tok/sec: 83271.0 | mem: 108909MB | Time Left: 00 hrs : 36 mins
|
| 163 |
+
[TIMING] data: 3.5ms | fwd: 2011.4ms | loss: 6.4ms | bwd: 4217.7ms | clip: 0.0ms | optim: 49.3ms
|
| 164 |
+
step 53 | loss: 1.472854 | lr 1.6200e-04 | norm: 0.8747 | dt: 6324.35ms | tok/sec: 82899.9 | mem: 108909MB | Time Left: 00 hrs : 36 mins
|
| 165 |
+
[TIMING] data: 3.5ms | fwd: 2009.7ms | loss: 6.6ms | bwd: 4247.1ms | clip: 0.0ms | optim: 49.4ms
|
| 166 |
+
step 54 | loss: 1.479821 | lr 1.6500e-04 | norm: 0.7724 | dt: 6332.94ms | tok/sec: 82787.5 | mem: 108909MB | Time Left: 00 hrs : 36 mins
|
| 167 |
+
[TIMING] data: 3.5ms | fwd: 2018.3ms | loss: 6.4ms | bwd: 4247.4ms | clip: 0.0ms | optim: 49.5ms
|
| 168 |
+
step 55 | loss: 1.559876 | lr 1.6800e-04 | norm: 0.6723 | dt: 6319.43ms | tok/sec: 82964.4 | mem: 108909MB | Time Left: 00 hrs : 36 mins
|
| 169 |
+
[TIMING] data: 3.4ms | fwd: 2010.3ms | loss: 6.4ms | bwd: 4242.0ms | clip: 0.0ms | optim: 49.4ms
|
| 170 |
+
step 56 | loss: 1.525480 | lr 1.7100e-04 | norm: 0.6874 | dt: 6329.41ms | tok/sec: 82833.6 | mem: 108909MB | Time Left: 00 hrs : 36 mins
|
| 171 |
+
[TIMING] data: 3.6ms | fwd: 2019.1ms | loss: 6.4ms | bwd: 4241.9ms | clip: 0.0ms | optim: 50.6ms
|
| 172 |
+
step 57 | loss: 1.459871 | lr 1.7400e-04 | norm: 0.6680 | dt: 6321.78ms | tok/sec: 82933.6 | mem: 108909MB | Time Left: 00 hrs : 36 mins
|
| 173 |
+
[TIMING] data: 3.5ms | fwd: 2011.0ms | loss: 6.4ms | bwd: 4243.3ms | clip: 0.0ms | optim: 49.6ms
|
| 174 |
+
step 58 | loss: 1.471684 | lr 1.7700e-04 | norm: 0.7933 | dt: 6347.73ms | tok/sec: 82594.6 | mem: 108909MB | Time Left: 00 hrs : 36 mins
|
| 175 |
+
[TIMING] data: 3.4ms | fwd: 2032.0ms | loss: 6.7ms | bwd: 4248.2ms | clip: 0.0ms | optim: 49.4ms
|
| 176 |
+
step 59 | loss: 1.480517 | lr 1.8000e-04 | norm: 0.7333 | dt: 6323.14ms | tok/sec: 82915.8 | mem: 108909MB | Time Left: 00 hrs : 35 mins
|
| 177 |
+
[TIMING] data: 3.5ms | fwd: 2015.6ms | loss: 6.4ms | bwd: 4240.3ms | clip: 0.0ms | optim: 49.4ms
|
| 178 |
+
step 60 | loss: 1.519410 | lr 1.8300e-04 | norm: 0.7315 | dt: 6331.64ms | tok/sec: 82804.5 | mem: 108909MB | Time Left: 00 hrs : 35 mins
|
| 179 |
+
[TIMING] data: 3.5ms | fwd: 2021.2ms | loss: 6.4ms | bwd: 4243.3ms | clip: 0.0ms | optim: 49.3ms
|
| 180 |
+
step 61 | loss: 1.455516 | lr 1.8600e-04 | norm: 0.6936 | dt: 6323.94ms | tok/sec: 82905.3 | mem: 108909MB | Time Left: 00 hrs : 35 mins
|
| 181 |
+
[TIMING] data: 3.5ms | fwd: 2014.4ms | loss: 6.4ms | bwd: 4242.4ms | clip: 0.0ms | optim: 49.3ms
|
| 182 |
+
step 62 | loss: 1.486469 | lr 1.8900e-04 | norm: 0.7827 | dt: 6324.21ms | tok/sec: 82901.7 | mem: 108909MB | Time Left: 00 hrs : 35 mins
|
| 183 |
+
[TIMING] data: 3.5ms | fwd: 2013.1ms | loss: 6.4ms | bwd: 4244.0ms | clip: 0.0ms | optim: 49.4ms
|
| 184 |
+
step 63 | loss: 1.483447 | lr 1.9200e-04 | norm: 0.8920 | dt: 6335.92ms | tok/sec: 82748.5 | mem: 108909MB | Time Left: 00 hrs : 35 mins
|
| 185 |
+
[TIMING] data: 3.5ms | fwd: 2021.0ms | loss: 6.4ms | bwd: 4247.6ms | clip: 0.0ms | optim: 49.6ms
|
| 186 |
+
step 64 | loss: 1.452058 | lr 1.9500e-04 | norm: 0.8363 | dt: 6364.76ms | tok/sec: 82373.6 | mem: 108909MB | Time Left: 00 hrs : 35 mins
|
| 187 |
+
[TIMING] data: 3.5ms | fwd: 2062.0ms | loss: 6.4ms | bwd: 4235.4ms | clip: 0.0ms | optim: 49.4ms
|
| 188 |
+
step 65 | loss: 1.500889 | lr 1.9800e-04 | norm: 0.8404 | dt: 6336.29ms | tok/sec: 82743.6 | mem: 108909MB | Time Left: 00 hrs : 35 mins
|
| 189 |
+
[TIMING] data: 3.4ms | fwd: 2023.0ms | loss: 6.5ms | bwd: 4245.9ms | clip: 0.0ms | optim: 49.4ms
|
| 190 |
+
step 66 | loss: 1.483080 | lr 2.0100e-04 | norm: 0.7789 | dt: 6334.96ms | tok/sec: 82761.1 | mem: 108909MB | Time Left: 00 hrs : 35 mins
|
| 191 |
+
[TIMING] data: 3.5ms | fwd: 2017.6ms | loss: 6.4ms | bwd: 4250.1ms | clip: 0.0ms | optim: 49.4ms
|
| 192 |
+
step 67 | loss: 1.524449 | lr 2.0400e-04 | norm: 0.8876 | dt: 6310.82ms | tok/sec: 83077.7 | mem: 108909MB | Time Left: 00 hrs : 35 mins
|
| 193 |
+
[TIMING] data: 3.5ms | fwd: 2011.6ms | loss: 6.4ms | bwd: 4232.0ms | clip: 0.0ms | optim: 49.4ms
|
| 194 |
+
step 68 | loss: 1.493887 | lr 2.0700e-04 | norm: 0.6793 | dt: 6330.85ms | tok/sec: 82814.8 | mem: 108909MB | Time Left: 00 hrs : 35 mins
|
| 195 |
+
[TIMING] data: 3.5ms | fwd: 2018.1ms | loss: 6.4ms | bwd: 4244.6ms | clip: 0.0ms | optim: 50.3ms
|
| 196 |
+
step 69 | loss: 1.532737 | lr 2.1000e-04 | norm: 0.7137 | dt: 6315.90ms | tok/sec: 83010.9 | mem: 108909MB | Time Left: 00 hrs : 34 mins
|
| 197 |
+
[TIMING] data: 3.5ms | fwd: 2019.8ms | loss: 6.4ms | bwd: 4228.6ms | clip: 0.0ms | optim: 49.7ms
|
| 198 |
+
step 70 | loss: 1.498729 | lr 2.1300e-04 | norm: 0.8502 | dt: 6321.53ms | tok/sec: 82936.9 | mem: 108909MB | Time Left: 00 hrs : 34 mins
|
| 199 |
+
[TIMING] data: 3.5ms | fwd: 2015.6ms | loss: 6.4ms | bwd: 4238.6ms | clip: 0.0ms | optim: 49.5ms
|
| 200 |
+
step 71 | loss: 1.501005 | lr 2.1600e-04 | norm: 0.6561 | dt: 6370.35ms | tok/sec: 82301.2 | mem: 108909MB | Time Left: 00 hrs : 34 mins
|
| 201 |
+
[TIMING] data: 3.5ms | fwd: 2075.7ms | loss: 6.4ms | bwd: 4227.3ms | clip: 0.0ms | optim: 49.5ms
|
| 202 |
+
step 72 | loss: 1.435198 | lr 2.1900e-04 | norm: 0.8811 | dt: 6322.63ms | tok/sec: 82922.5 | mem: 108909MB | Time Left: 00 hrs : 34 mins
|
| 203 |
+
[TIMING] data: 3.5ms | fwd: 2016.4ms | loss: 6.4ms | bwd: 4239.0ms | clip: 0.0ms | optim: 49.5ms
|
| 204 |
+
step 73 | loss: 1.507871 | lr 2.2200e-04 | norm: 0.8746 | dt: 6320.86ms | tok/sec: 82945.7 | mem: 108909MB | Time Left: 00 hrs : 34 mins
|
| 205 |
+
[TIMING] data: 3.5ms | fwd: 2010.0ms | loss: 6.4ms | bwd: 4243.6ms | clip: 0.0ms | optim: 49.4ms
|
| 206 |
+
step 74 | loss: 1.489377 | lr 2.2500e-04 | norm: 0.8673 | dt: 6323.05ms | tok/sec: 82916.9 | mem: 108909MB | Time Left: 00 hrs : 34 mins
|
| 207 |
+
[TIMING] data: 3.5ms | fwd: 2014.3ms | loss: 6.4ms | bwd: 4241.4ms | clip: 0.0ms | optim: 49.5ms
|
| 208 |
+
step 75 | loss: 1.502547 | lr 2.2800e-04 | norm: 0.7430 | dt: 6332.58ms | tok/sec: 82792.2 | mem: 108909MB | Time Left: 00 hrs : 34 mins
|
| 209 |
+
[TIMING] data: 3.5ms | fwd: 2015.2ms | loss: 6.4ms | bwd: 4250.1ms | clip: 0.0ms | optim: 49.4ms
|
| 210 |
+
step 76 | loss: 1.455447 | lr 2.3100e-04 | norm: 0.7772 | dt: 6310.79ms | tok/sec: 83078.1 | mem: 108909MB | Time Left: 00 hrs : 34 mins
|
| 211 |
+
[TIMING] data: 3.5ms | fwd: 2011.1ms | loss: 6.4ms | bwd: 4232.4ms | clip: 0.0ms | optim: 49.4ms
|
| 212 |
+
step 77 | loss: 1.513057 | lr 2.3400e-04 | norm: 0.6982 | dt: 6349.46ms | tok/sec: 82572.1 | mem: 108909MB | Time Left: 00 hrs : 34 mins
|
| 213 |
+
[TIMING] data: 3.6ms | fwd: 2030.2ms | loss: 6.4ms | bwd: 4252.0ms | clip: 0.0ms | optim: 49.3ms
|
| 214 |
+
step 78 | loss: 1.487326 | lr 2.3700e-04 | norm: 0.7463 | dt: 6344.28ms | tok/sec: 82639.5 | mem: 108909MB | Time Left: 00 hrs : 34 mins
|
| 215 |
+
[TIMING] data: 3.6ms | fwd: 2046.4ms | loss: 6.4ms | bwd: 4230.1ms | clip: 0.0ms | optim: 49.8ms
|
| 216 |
+
step 79 | loss: 1.497366 | lr 2.4000e-04 | norm: 0.7971 | dt: 6336.67ms | tok/sec: 82738.7 | mem: 108909MB | Time Left: 00 hrs : 33 mins
|
| 217 |
+
[TIMING] data: 3.5ms | fwd: 2010.9ms | loss: 6.6ms | bwd: 4258.2ms | clip: 0.0ms | optim: 49.5ms
|
| 218 |
+
step 80 | loss: 1.496414 | lr 2.4300e-04 | norm: 0.6843 | dt: 6305.02ms | tok/sec: 83154.0 | mem: 108909MB | Time Left: 00 hrs : 33 mins
|
| 219 |
+
[TIMING] data: 3.4ms | fwd: 2012.4ms | loss: 6.4ms | bwd: 4225.0ms | clip: 0.0ms | optim: 49.9ms
|
| 220 |
+
step 81 | loss: 1.544732 | lr 2.4600e-04 | norm: 0.7336 | dt: 6309.07ms | tok/sec: 83100.7 | mem: 108909MB | Time Left: 00 hrs : 33 mins
|
| 221 |
+
[TIMING] data: 3.4ms | fwd: 2023.1ms | loss: 6.4ms | bwd: 4218.9ms | clip: 0.0ms | optim: 49.3ms
|
| 222 |
+
step 82 | loss: 1.565590 | lr 2.4900e-04 | norm: 0.7821 | dt: 6314.54ms | tok/sec: 83028.7 | mem: 108909MB | Time Left: 00 hrs : 33 mins
|
| 223 |
+
[TIMING] data: 3.4ms | fwd: 2020.2ms | loss: 6.4ms | bwd: 4227.3ms | clip: 0.0ms | optim: 49.3ms
|
| 224 |
+
step 83 | loss: 1.488052 | lr 2.5200e-04 | norm: 0.7503 | dt: 6320.02ms | tok/sec: 82956.7 | mem: 108909MB | Time Left: 00 hrs : 33 mins
|
| 225 |
+
[TIMING] data: 3.6ms | fwd: 2012.4ms | loss: 6.4ms | bwd: 4240.3ms | clip: 0.0ms | optim: 49.4ms
|
| 226 |
+
step 84 | loss: 1.506120 | lr 2.5500e-04 | norm: 0.8259 | dt: 6357.75ms | tok/sec: 82464.4 | mem: 108909MB | Time Left: 00 hrs : 33 mins
|
| 227 |
+
[TIMING] data: 3.5ms | fwd: 2063.1ms | loss: 6.4ms | bwd: 4227.5ms | clip: 0.0ms | optim: 49.3ms
|
| 228 |
+
step 85 | loss: 1.508562 | lr 2.5800e-04 | norm: 0.8080 | dt: 6323.59ms | tok/sec: 82909.9 | mem: 108909MB | Time Left: 00 hrs : 33 mins
|
| 229 |
+
[TIMING] data: 3.6ms | fwd: 2016.4ms | loss: 6.4ms | bwd: 4240.0ms | clip: 0.0ms | optim: 49.3ms
|
| 230 |
+
step 86 | loss: 1.506531 | lr 2.6100e-04 | norm: 0.8455 | dt: 6321.49ms | tok/sec: 82937.4 | mem: 108909MB | Time Left: 00 hrs : 33 mins
|
| 231 |
+
[TIMING] data: 3.4ms | fwd: 2012.8ms | loss: 6.4ms | bwd: 4241.1ms | clip: 0.0ms | optim: 49.9ms
|
| 232 |
+
step 87 | loss: 1.568609 | lr 2.6400e-04 | norm: 0.8385 | dt: 6346.16ms | tok/sec: 82615.0 | mem: 108909MB | Time Left: 00 hrs : 33 mins
|
| 233 |
+
[TIMING] data: 3.4ms | fwd: 2055.7ms | loss: 6.4ms | bwd: 4223.2ms | clip: 0.0ms | optim: 49.6ms
|
| 234 |
+
step 88 | loss: 1.578616 | lr 2.6700e-04 | norm: 0.8432 | dt: 6316.88ms | tok/sec: 82997.9 | mem: 108909MB | Time Left: 00 hrs : 32 mins
|
| 235 |
+
[TIMING] data: 3.4ms | fwd: 2016.8ms | loss: 6.4ms | bwd: 4232.9ms | clip: 0.0ms | optim: 49.4ms
|
| 236 |
+
step 89 | loss: 1.520792 | lr 2.7000e-04 | norm: 0.8000 | dt: 6328.62ms | tok/sec: 82844.0 | mem: 108909MB | Time Left: 00 hrs : 32 mins
|
| 237 |
+
[TIMING] data: 3.7ms | fwd: 1999.6ms | loss: 6.4ms | bwd: 4261.6ms | clip: 0.0ms | optim: 49.4ms
|
| 238 |
+
step 90 | loss: 1.589205 | lr 2.7300e-04 | norm: 0.7820 | dt: 6301.03ms | tok/sec: 83206.7 | mem: 108909MB | Time Left: 00 hrs : 32 mins
|
| 239 |
+
[TIMING] data: 3.5ms | fwd: 2015.3ms | loss: 6.4ms | bwd: 4218.2ms | clip: 0.0ms | optim: 49.7ms
|
| 240 |
+
step 91 | loss: 1.535868 | lr 2.7600e-04 | norm: 0.8531 | dt: 6363.63ms | tok/sec: 82388.1 | mem: 108909MB | Time Left: 00 hrs : 32 mins
|
| 241 |
+
[TIMING] data: 3.5ms | fwd: 2066.5ms | loss: 6.4ms | bwd: 4229.5ms | clip: 0.0ms | optim: 49.7ms
|
| 242 |
+
step 92 | loss: 1.544926 | lr 2.7900e-04 | norm: 0.7781 | dt: 6364.68ms | tok/sec: 82374.6 | mem: 108909MB | Time Left: 00 hrs : 32 mins
|
| 243 |
+
[TIMING] data: 3.5ms | fwd: 2043.3ms | loss: 6.4ms | bwd: 4254.1ms | clip: 0.0ms | optim: 49.5ms
|
| 244 |
+
step 93 | loss: 1.568713 | lr 2.8200e-04 | norm: 0.7326 | dt: 6315.09ms | tok/sec: 83021.5 | mem: 108909MB | Time Left: 00 hrs : 32 mins
|
| 245 |
+
[TIMING] data: 3.4ms | fwd: 2012.7ms | loss: 6.4ms | bwd: 4235.1ms | clip: 0.0ms | optim: 49.5ms
|
| 246 |
+
step 94 | loss: 1.447728 | lr 2.8500e-04 | norm: 0.8374 | dt: 6328.15ms | tok/sec: 82850.1 | mem: 108909MB | Time Left: 00 hrs : 32 mins
|
| 247 |
+
[TIMING] data: 3.5ms | fwd: 2018.8ms | loss: 6.4ms | bwd: 4241.1ms | clip: 0.0ms | optim: 50.0ms
|
| 248 |
+
step 95 | loss: 1.528693 | lr 2.8800e-04 | norm: 0.7675 | dt: 6318.46ms | tok/sec: 82977.2 | mem: 108909MB | Time Left: 00 hrs : 32 mins
|
| 249 |
+
[TIMING] data: 3.4ms | fwd: 2017.1ms | loss: 6.4ms | bwd: 4234.2ms | clip: 0.0ms | optim: 49.4ms
|
| 250 |
+
step 96 | loss: 1.526013 | lr 2.9100e-04 | norm: 0.7406 | dt: 6334.85ms | tok/sec: 82762.5 | mem: 108909MB | Time Left: 00 hrs : 32 mins
|
| 251 |
+
[TIMING] data: 3.4ms | fwd: 2018.0ms | loss: 6.4ms | bwd: 4249.8ms | clip: 0.0ms | optim: 49.3ms
|
| 252 |
+
step 97 | loss: 1.522310 | lr 2.9400e-04 | norm: 0.7262 | dt: 6366.42ms | tok/sec: 82352.1 | mem: 108909MB | Time Left: 00 hrs : 32 mins
|
| 253 |
+
[TIMING] data: 3.5ms | fwd: 2054.2ms | loss: 6.4ms | bwd: 4244.9ms | clip: 0.0ms | optim: 49.4ms
|
| 254 |
+
step 98 | loss: 1.517379 | lr 2.9700e-04 | norm: 0.7505 | dt: 6335.56ms | tok/sec: 82753.2 | mem: 108909MB | Time Left: 00 hrs : 31 mins
|
| 255 |
+
[TIMING] data: 3.5ms | fwd: 2027.7ms | loss: 6.4ms | bwd: 4240.7ms | clip: 0.0ms | optim: 49.3ms
|
| 256 |
+
step 99 | loss: 1.542074 | lr 3.0000e-04 | norm: 0.9468 | dt: 6316.63ms | tok/sec: 83001.2 | mem: 108909MB | Time Left: 00 hrs : 31 mins
|
| 257 |
+
[TIMING] data: 3.4ms | fwd: 2015.2ms | loss: 6.4ms | bwd: 4234.2ms | clip: 0.0ms | optim: 49.4ms
|
| 258 |
+
[CheckpointManager] Saved: checkpoints_bluscriptcp/rank_0/blumodelcp_run13_step_100.ckpt (safely)
|
| 259 |
+
step 100 | loss: 1.560971 | lr 3.0000e-04 | norm: 0.8281 | dt: 6193.54ms | tok/sec: 84650.8 | mem: 108909MB | Time Left: 00 hrs : 30 mins
|
| 260 |
+
[TIMING] data: 3.4ms | fwd: 1987.0ms | loss: 6.4ms | bwd: 4139.4ms | clip: 0.0ms | optim: 49.4ms
|
| 261 |
+
step 101 | loss: 1.590868 | lr 2.9999e-04 | norm: 0.7858 | dt: 6223.60ms | tok/sec: 84241.9 | mem: 108909MB | Time Left: 00 hrs : 31 mins
|
| 262 |
+
[TIMING] data: 3.3ms | fwd: 1992.2ms | loss: 6.4ms | bwd: 4164.1ms | clip: 0.0ms | optim: 49.6ms
|
| 263 |
+
step 102 | loss: 1.500099 | lr 2.9998e-04 | norm: 0.7740 | dt: 6261.98ms | tok/sec: 83725.6 | mem: 108909MB | Time Left: 00 hrs : 31 mins
|
| 264 |
+
[TIMING] data: 3.4ms | fwd: 1995.6ms | loss: 6.4ms | bwd: 4199.4ms | clip: 0.0ms | optim: 49.3ms
|
| 265 |
+
step 103 | loss: 1.553445 | lr 2.9995e-04 | norm: 0.7512 | dt: 6281.82ms | tok/sec: 83461.2 | mem: 108909MB | Time Left: 00 hrs : 31 mins
|
| 266 |
+
[TIMING] data: 3.3ms | fwd: 2017.0ms | loss: 6.4ms | bwd: 4197.8ms | clip: 0.0ms | optim: 49.4ms
|
| 267 |
+
step 104 | loss: 1.622695 | lr 2.9991e-04 | norm: 0.9523 | dt: 6280.21ms | tok/sec: 83482.6 | mem: 108909MB | Time Left: 00 hrs : 30 mins
|
| 268 |
+
[TIMING] data: 3.4ms | fwd: 2006.3ms | loss: 6.4ms | bwd: 4206.8ms | clip: 0.0ms | optim: 49.3ms
|
| 269 |
+
step 105 | loss: 1.534858 | lr 2.9986e-04 | norm: 0.8496 | dt: 6321.54ms | tok/sec: 82936.8 | mem: 108909MB | Time Left: 00 hrs : 31 mins
|
| 270 |
+
[TIMING] data: 3.4ms | fwd: 2015.3ms | loss: 6.4ms | bwd: 4239.0ms | clip: 0.0ms | optim: 49.4ms
|
| 271 |
+
step 106 | loss: 1.540255 | lr 2.9980e-04 | norm: 0.9536 | dt: 6364.30ms | tok/sec: 82379.5 | mem: 108909MB | Time Left: 00 hrs : 31 mins
|
| 272 |
+
[TIMING] data: 3.3ms | fwd: 2052.6ms | loss: 6.4ms | bwd: 4244.7ms | clip: 0.0ms | optim: 49.4ms
|
| 273 |
+
step 107 | loss: 1.606415 | lr 2.9973e-04 | norm: 0.7491 | dt: 6333.17ms | tok/sec: 82784.5 | mem: 108909MB | Time Left: 00 hrs : 30 mins
|
| 274 |
+
[TIMING] data: 3.5ms | fwd: 2016.5ms | loss: 6.4ms | bwd: 4249.4ms | clip: 0.0ms | optim: 49.5ms
|
| 275 |
+
step 108 | loss: 1.496417 | lr 2.9965e-04 | norm: 0.9238 | dt: 6313.81ms | tok/sec: 83038.3 | mem: 108909MB | Time Left: 00 hrs : 30 mins
|
| 276 |
+
[TIMING] data: 3.3ms | fwd: 2012.1ms | loss: 6.4ms | bwd: 4234.7ms | clip: 0.0ms | optim: 49.3ms
|
| 277 |
+
step 109 | loss: 1.536228 | lr 2.9956e-04 | norm: 0.7126 | dt: 6318.94ms | tok/sec: 82970.9 | mem: 108909MB | Time Left: 00 hrs : 30 mins
|
| 278 |
+
[TIMING] data: 3.4ms | fwd: 2017.6ms | loss: 6.4ms | bwd: 4233.7ms | clip: 0.0ms | optim: 49.9ms
|
| 279 |
+
step 110 | loss: 1.559572 | lr 2.9945e-04 | norm: 0.8443 | dt: 6326.96ms | tok/sec: 82865.7 | mem: 108909MB | Time Left: 00 hrs : 30 mins
|
| 280 |
+
[TIMING] data: 3.3ms | fwd: 2019.8ms | loss: 6.4ms | bwd: 4239.7ms | clip: 0.0ms | optim: 49.8ms
|
| 281 |
+
step 111 | loss: 1.541899 | lr 2.9934e-04 | norm: 0.7483 | dt: 6321.80ms | tok/sec: 82933.4 | mem: 108909MB | Time Left: 00 hrs : 30 mins
|
| 282 |
+
[TIMING] data: 3.4ms | fwd: 2009.3ms | loss: 6.6ms | bwd: 4245.2ms | clip: 0.0ms | optim: 49.2ms
|
| 283 |
+
step 112 | loss: 1.518180 | lr 2.9921e-04 | norm: 0.9077 | dt: 6326.17ms | tok/sec: 82876.0 | mem: 108909MB | Time Left: 00 hrs : 30 mins
|
| 284 |
+
[TIMING] data: 3.4ms | fwd: 2020.6ms | loss: 6.4ms | bwd: 4238.0ms | clip: 0.0ms | optim: 49.7ms
|
| 285 |
+
step 113 | loss: 1.504054 | lr 2.9907e-04 | norm: 0.8560 | dt: 6305.27ms | tok/sec: 83150.7 | mem: 108909MB | Time Left: 00 hrs : 30 mins
|
| 286 |
+
[TIMING] data: 3.4ms | fwd: 2015.3ms | loss: 6.4ms | bwd: 4223.0ms | clip: 0.0ms | optim: 49.3ms
|
| 287 |
+
step 114 | loss: 1.536195 | lr 2.9893e-04 | norm: 0.6512 | dt: 6374.01ms | tok/sec: 82254.0 | mem: 108909MB | Time Left: 00 hrs : 30 mins
|
| 288 |
+
[TIMING] data: 3.5ms | fwd: 2062.1ms | loss: 6.4ms | bwd: 4244.8ms | clip: 0.0ms | optim: 49.3ms
|
| 289 |
+
step 115 | loss: 1.523702 | lr 2.9877e-04 | norm: 0.7223 | dt: 6380.53ms | tok/sec: 82170.0 | mem: 108909MB | Time Left: 00 hrs : 30 mins
|
| 290 |
+
[TIMING] data: 3.3ms | fwd: 2062.4ms | loss: 6.4ms | bwd: 4250.8ms | clip: 0.0ms | optim: 49.6ms
|
| 291 |
+
step 116 | loss: 1.471415 | lr 2.9860e-04 | norm: 0.9105 | dt: 6321.89ms | tok/sec: 82932.1 | mem: 108909MB | Time Left: 00 hrs : 29 mins
|
| 292 |
+
[TIMING] data: 3.4ms | fwd: 2014.8ms | loss: 6.4ms | bwd: 4239.0ms | clip: 0.0ms | optim: 50.4ms
|
| 293 |
+
step 117 | loss: 1.486171 | lr 2.9842e-04 | norm: 0.8120 | dt: 6331.32ms | tok/sec: 82808.6 | mem: 108909MB | Time Left: 00 hrs : 29 mins
|
| 294 |
+
[TIMING] data: 3.4ms | fwd: 2022.3ms | loss: 6.4ms | bwd: 4241.8ms | clip: 0.0ms | optim: 49.5ms
|
| 295 |
+
step 118 | loss: 1.551453 | lr 2.9823e-04 | norm: 0.7651 | dt: 6317.55ms | tok/sec: 82989.2 | mem: 108909MB | Time Left: 00 hrs : 29 mins
|
| 296 |
+
[TIMING] data: 3.4ms | fwd: 2016.7ms | loss: 6.5ms | bwd: 4233.6ms | clip: 0.0ms | optim: 49.4ms
|
| 297 |
+
step 119 | loss: 1.563141 | lr 2.9803e-04 | norm: 0.7465 | dt: 6339.21ms | tok/sec: 82705.6 | mem: 108909MB | Time Left: 00 hrs : 29 mins
|
| 298 |
+
[TIMING] data: 3.3ms | fwd: 2017.5ms | loss: 6.4ms | bwd: 4254.6ms | clip: 0.0ms | optim: 49.4ms
|
| 299 |
+
step 120 | loss: 1.506313 | lr 2.9781e-04 | norm: 0.6294 | dt: 6324.94ms | tok/sec: 82892.2 | mem: 108909MB | Time Left: 00 hrs : 29 mins
|
| 300 |
+
[TIMING] data: 3.3ms | fwd: 2018.4ms | loss: 6.4ms | bwd: 4239.5ms | clip: 0.0ms | optim: 49.4ms
|
| 301 |
+
step 121 | loss: 1.531366 | lr 2.9759e-04 | norm: 0.7520 | dt: 6304.79ms | tok/sec: 83157.2 | mem: 108909MB | Time Left: 00 hrs : 29 mins
|
| 302 |
+
[TIMING] data: 3.4ms | fwd: 2019.4ms | loss: 6.4ms | bwd: 4218.3ms | clip: 0.0ms | optim: 49.4ms
|
| 303 |
+
step 122 | loss: 1.583121 | lr 2.9736e-04 | norm: 0.7222 | dt: 6373.51ms | tok/sec: 82260.5 | mem: 108909MB | Time Left: 00 hrs : 29 mins
|
| 304 |
+
[TIMING] data: 3.3ms | fwd: 2064.9ms | loss: 6.4ms | bwd: 4241.5ms | clip: 0.0ms | optim: 49.5ms
|
| 305 |
+
step 123 | loss: 1.549728 | lr 2.9711e-04 | norm: 0.6932 | dt: 6385.67ms | tok/sec: 82103.8 | mem: 108909MB | Time Left: 00 hrs : 29 mins
|
| 306 |
+
[TIMING] data: 3.4ms | fwd: 2075.5ms | loss: 6.4ms | bwd: 4243.0ms | clip: 0.0ms | optim: 49.4ms
|
| 307 |
+
step 124 | loss: 1.571770 | lr 2.9686e-04 | norm: 0.7142 | dt: 6307.75ms | tok/sec: 83118.0 | mem: 108909MB | Time Left: 00 hrs : 29 mins
|
| 308 |
+
[TIMING] data: 3.4ms | fwd: 2008.9ms | loss: 6.4ms | bwd: 4231.8ms | clip: 0.0ms | optim: 49.3ms
|
| 309 |
+
step 125 | loss: 1.530097 | lr 2.9659e-04 | norm: 0.5959 | dt: 6302.30ms | tok/sec: 83190.0 | mem: 108909MB | Time Left: 00 hrs : 28 mins
|
| 310 |
+
[TIMING] data: 3.4ms | fwd: 2014.8ms | loss: 6.4ms | bwd: 4220.4ms | clip: 0.0ms | optim: 49.4ms
|
| 311 |
+
step 126 | loss: 1.548526 | lr 2.9632e-04 | norm: 0.6799 | dt: 6342.53ms | tok/sec: 82662.3 | mem: 108909MB | Time Left: 00 hrs : 28 mins
|
| 312 |
+
[TIMING] data: 3.4ms | fwd: 2019.1ms | loss: 6.4ms | bwd: 4256.5ms | clip: 0.0ms | optim: 49.3ms
|
| 313 |
+
step 127 | loss: 1.587535 | lr 2.9603e-04 | norm: 0.5902 | dt: 6309.10ms | tok/sec: 83100.2 | mem: 108909MB | Time Left: 00 hrs : 28 mins
|
| 314 |
+
[TIMING] data: 3.4ms | fwd: 2015.9ms | loss: 6.4ms | bwd: 4225.9ms | clip: 0.0ms | optim: 49.6ms
|
| 315 |
+
step 128 | loss: 1.606323 | lr 2.9573e-04 | norm: 0.6112 | dt: 6359.27ms | tok/sec: 82444.7 | mem: 108909MB | Time Left: 00 hrs : 28 mins
|
| 316 |
+
[TIMING] data: 3.4ms | fwd: 2067.1ms | loss: 6.4ms | bwd: 4225.1ms | clip: 0.0ms | optim: 49.3ms
|
| 317 |
+
step 129 | loss: 1.556684 | lr 2.9542e-04 | norm: 0.6113 | dt: 6341.04ms | tok/sec: 82681.8 | mem: 108909MB | Time Left: 00 hrs : 28 mins
|
| 318 |
+
[TIMING] data: 3.4ms | fwd: 2057.6ms | loss: 6.4ms | bwd: 4216.4ms | clip: 0.0ms | optim: 49.3ms
|
| 319 |
+
step 130 | loss: 1.676277 | lr 2.9511e-04 | norm: 0.7016 | dt: 6325.64ms | tok/sec: 82883.0 | mem: 108909MB | Time Left: 00 hrs : 28 mins
|
| 320 |
+
[TIMING] data: 3.4ms | fwd: 2015.0ms | loss: 6.4ms | bwd: 4242.7ms | clip: 0.0ms | optim: 49.4ms
|
| 321 |
+
step 131 | loss: 1.559720 | lr 2.9478e-04 | norm: 0.6344 | dt: 6334.89ms | tok/sec: 82762.0 | mem: 108909MB | Time Left: 00 hrs : 28 mins
|
| 322 |
+
[TIMING] data: 3.3ms | fwd: 2019.9ms | loss: 6.4ms | bwd: 4247.9ms | clip: 0.0ms | optim: 49.4ms
|
| 323 |
+
Current shard: /root/Adhi/BluTrain/dist/Context_Parallelism/in2_data/arm_b/in2b_train_00001.bin
|
| 324 |
+
100000000
|
| 325 |
+
step 132 | loss: 1.561504 | lr 2.9444e-04 | norm: 0.6851 | dt: 6315.48ms | tok/sec: 83016.3 | mem: 108909MB | Time Left: 00 hrs : 28 mins
|
| 326 |
+
[TIMING] data: 15.4ms | fwd: 2015.8ms | loss: 6.4ms | bwd: 4220.3ms | clip: 0.0ms | optim: 49.7ms
|
| 327 |
+
step 133 | loss: 1.544661 | lr 2.9409e-04 | norm: 0.5797 | dt: 6326.40ms | tok/sec: 82873.1 | mem: 108909MB | Time Left: 00 hrs : 28 mins
|
| 328 |
+
[TIMING] data: 3.3ms | fwd: 2020.0ms | loss: 6.4ms | bwd: 4239.4ms | clip: 0.0ms | optim: 49.4ms
|
| 329 |
+
step 134 | loss: 1.552186 | lr 2.9373e-04 | norm: 0.6372 | dt: 6314.07ms | tok/sec: 83034.9 | mem: 108909MB | Time Left: 00 hrs : 27 mins
|
| 330 |
+
[TIMING] data: 3.4ms | fwd: 2019.4ms | loss: 6.4ms | bwd: 4227.6ms | clip: 0.0ms | optim: 49.3ms
|
| 331 |
+
step 135 | loss: 1.576610 | lr 2.9336e-04 | norm: 0.5871 | dt: 6309.44ms | tok/sec: 83095.8 | mem: 108909MB | Time Left: 00 hrs : 27 mins
|
| 332 |
+
[TIMING] data: 3.4ms | fwd: 2016.4ms | loss: 6.4ms | bwd: 4225.9ms | clip: 0.0ms | optim: 49.4ms
|
| 333 |
+
step 136 | loss: 1.529440 | lr 2.9298e-04 | norm: 0.6040 | dt: 6360.19ms | tok/sec: 82432.7 | mem: 108909MB | Time Left: 00 hrs : 27 mins
|
| 334 |
+
[TIMING] data: 3.3ms | fwd: 2091.1ms | loss: 6.4ms | bwd: 4202.1ms | clip: 0.0ms | optim: 49.4ms
|
| 335 |
+
step 137 | loss: 1.524042 | lr 2.9259e-04 | norm: 0.5933 | dt: 6319.37ms | tok/sec: 82965.3 | mem: 108909MB | Time Left: 00 hrs : 27 mins
|
| 336 |
+
[TIMING] data: 3.3ms | fwd: 2019.4ms | loss: 6.4ms | bwd: 4231.8ms | clip: 0.0ms | optim: 49.4ms
|
| 337 |
+
step 138 | loss: 1.582924 | lr 2.9219e-04 | norm: 0.5469 | dt: 6309.43ms | tok/sec: 83095.9 | mem: 108909MB | Time Left: 00 hrs : 27 mins
|
| 338 |
+
[TIMING] data: 3.4ms | fwd: 2018.0ms | loss: 6.4ms | bwd: 4224.4ms | clip: 0.0ms | optim: 49.4ms
|
| 339 |
+
step 139 | loss: 1.578228 | lr 2.9178e-04 | norm: 0.7709 | dt: 6318.26ms | tok/sec: 82979.8 | mem: 108909MB | Time Left: 00 hrs : 27 mins
|
| 340 |
+
[TIMING] data: 3.4ms | fwd: 2007.4ms | loss: 6.6ms | bwd: 4243.6ms | clip: 0.0ms | optim: 49.4ms
|
| 341 |
+
step 140 | loss: 1.595851 | lr 2.9135e-04 | norm: 0.5805 | dt: 6296.67ms | tok/sec: 83264.3 | mem: 108909MB | Time Left: 00 hrs : 27 mins
|
| 342 |
+
[TIMING] data: 3.5ms | fwd: 2014.1ms | loss: 6.4ms | bwd: 4214.2ms | clip: 0.0ms | optim: 50.4ms
|
| 343 |
+
step 141 | loss: 1.587405 | lr 2.9092e-04 | norm: 0.6079 | dt: 6355.14ms | tok/sec: 82498.3 | mem: 108909MB | Time Left: 00 hrs : 27 mins
|
| 344 |
+
[TIMING] data: 3.4ms | fwd: 2052.6ms | loss: 6.4ms | bwd: 4235.1ms | clip: 0.0ms | optim: 49.7ms
|
| 345 |
+
step 142 | loss: 1.570241 | lr 2.9048e-04 | norm: 0.5607 | dt: 6316.69ms | tok/sec: 83000.5 | mem: 108909MB | Time Left: 00 hrs : 27 mins
|
| 346 |
+
[TIMING] data: 3.3ms | fwd: 2013.2ms | loss: 6.4ms | bwd: 4236.3ms | clip: 0.0ms | optim: 49.5ms
|
| 347 |
+
step 143 | loss: 1.514309 | lr 2.9003e-04 | norm: 1.8534 | dt: 6304.19ms | tok/sec: 83165.0 | mem: 108909MB | Time Left: 00 hrs : 27 mins
|
| 348 |
+
[TIMING] data: 3.3ms | fwd: 2014.8ms | loss: 6.4ms | bwd: 4222.2ms | clip: 0.0ms | optim: 49.5ms
|
| 349 |
+
step 144 | loss: 1.589352 | lr 2.8957e-04 | norm: 0.5719 | dt: 6310.08ms | tok/sec: 83087.4 | mem: 108909MB | Time Left: 00 hrs : 26 mins
|
| 350 |
+
[TIMING] data: 3.4ms | fwd: 2014.5ms | loss: 6.6ms | bwd: 4227.9ms | clip: 0.0ms | optim: 49.7ms
|
| 351 |
+
step 145 | loss: 1.502340 | lr 2.8910e-04 | norm: 0.5646 | dt: 6302.00ms | tok/sec: 83194.0 | mem: 108909MB | Time Left: 00 hrs : 26 mins
|
| 352 |
+
[TIMING] data: 3.3ms | fwd: 2008.6ms | loss: 6.4ms | bwd: 4226.4ms | clip: 0.0ms | optim: 49.3ms
|
| 353 |
+
step 146 | loss: 1.517694 | lr 2.8862e-04 | norm: 0.5770 | dt: 6306.29ms | tok/sec: 83137.3 | mem: 108909MB | Time Left: 00 hrs : 26 mins
|
| 354 |
+
[TIMING] data: 3.4ms | fwd: 2009.6ms | loss: 6.4ms | bwd: 4229.5ms | clip: 0.0ms | optim: 49.5ms
|
| 355 |
+
step 147 | loss: 1.520943 | lr 2.8813e-04 | norm: 0.5190 | dt: 6292.18ms | tok/sec: 83323.8 | mem: 108909MB | Time Left: 00 hrs : 26 mins
|
| 356 |
+
[TIMING] data: 3.3ms | fwd: 2014.0ms | loss: 6.4ms | bwd: 4211.2ms | clip: 0.0ms | optim: 49.3ms
|
| 357 |
+
step 148 | loss: 1.562016 | lr 2.8763e-04 | norm: 0.6044 | dt: 6279.47ms | tok/sec: 83492.4 | mem: 108909MB | Time Left: 00 hrs : 26 mins
|
| 358 |
+
[TIMING] data: 3.4ms | fwd: 2011.2ms | loss: 6.4ms | bwd: 4201.2ms | clip: 0.0ms | optim: 49.4ms
|
| 359 |
+
step 149 | loss: 1.563453 | lr 2.8712e-04 | norm: 0.6146 | dt: 6333.33ms | tok/sec: 82782.4 | mem: 108909MB | Time Left: 00 hrs : 26 mins
|
| 360 |
+
[TIMING] data: 3.4ms | fwd: 2044.6ms | loss: 6.4ms | bwd: 4221.5ms | clip: 0.0ms | optim: 49.5ms
|
| 361 |
+
step 150 | loss: 1.517942 | lr 2.8660e-04 | norm: 0.6800 | dt: 6318.64ms | tok/sec: 82974.8 | mem: 108909MB | Time Left: 00 hrs : 26 mins
|
| 362 |
+
[TIMING] data: 3.4ms | fwd: 2032.5ms | loss: 6.4ms | bwd: 4219.1ms | clip: 0.0ms | optim: 49.3ms
|
| 363 |
+
step 151 | loss: 1.538941 | lr 2.8607e-04 | norm: 0.5141 | dt: 6293.21ms | tok/sec: 83310.2 | mem: 108909MB | Time Left: 00 hrs : 26 mins
|
| 364 |
+
[TIMING] data: 3.4ms | fwd: 2011.1ms | loss: 6.4ms | bwd: 4214.7ms | clip: 0.0ms | optim: 49.6ms
|
| 365 |
+
step 152 | loss: 1.583701 | lr 2.8554e-04 | norm: 0.4730 | dt: 6299.01ms | tok/sec: 83233.4 | mem: 108909MB | Time Left: 00 hrs : 26 mins
|
| 366 |
+
[TIMING] data: 3.4ms | fwd: 2007.5ms | loss: 6.4ms | bwd: 4224.3ms | clip: 0.0ms | optim: 49.5ms
|
| 367 |
+
step 153 | loss: 1.539420 | lr 2.8499e-04 | norm: 0.6077 | dt: 6288.10ms | tok/sec: 83377.8 | mem: 108909MB | Time Left: 00 hrs : 25 mins
|
| 368 |
+
[TIMING] data: 3.4ms | fwd: 2009.9ms | loss: 6.4ms | bwd: 4211.2ms | clip: 0.0ms | optim: 49.4ms
|
| 369 |
+
step 154 | loss: 1.568313 | lr 2.8443e-04 | norm: 0.7170 | dt: 6303.74ms | tok/sec: 83170.9 | mem: 108909MB | Time Left: 00 hrs : 25 mins
|
| 370 |
+
[TIMING] data: 3.3ms | fwd: 2015.0ms | loss: 6.4ms | bwd: 4221.6ms | clip: 0.0ms | optim: 49.4ms
|
| 371 |
+
step 155 | loss: 1.547035 | lr 2.8387e-04 | norm: 0.4806 | dt: 6288.74ms | tok/sec: 83369.3 | mem: 108909MB | Time Left: 00 hrs : 25 mins
|
| 372 |
+
[TIMING] data: 3.3ms | fwd: 2009.2ms | loss: 6.4ms | bwd: 4212.5ms | clip: 0.0ms | optim: 49.4ms
|
| 373 |
+
step 156 | loss: 1.542251 | lr 2.8329e-04 | norm: 0.6079 | dt: 6310.97ms | tok/sec: 83075.7 | mem: 108909MB | Time Left: 00 hrs : 25 mins
|
| 374 |
+
[TIMING] data: 3.4ms | fwd: 2015.4ms | loss: 6.4ms | bwd: 4228.5ms | clip: 0.0ms | optim: 49.4ms
|
| 375 |
+
step 157 | loss: 1.593042 | lr 2.8271e-04 | norm: 0.5018 | dt: 6307.18ms | tok/sec: 83125.6 | mem: 108909MB | Time Left: 00 hrs : 25 mins
|
| 376 |
+
[TIMING] data: 3.5ms | fwd: 2012.0ms | loss: 6.4ms | bwd: 4227.8ms | clip: 0.0ms | optim: 49.7ms
|
| 377 |
+
step 158 | loss: 1.604503 | lr 2.8211e-04 | norm: 0.6080 | dt: 6336.21ms | tok/sec: 82744.7 | mem: 108909MB | Time Left: 00 hrs : 25 mins
|
| 378 |
+
[TIMING] data: 3.4ms | fwd: 2065.8ms | loss: 6.4ms | bwd: 4203.2ms | clip: 0.0ms | optim: 49.5ms
|
| 379 |
+
step 159 | loss: 1.554489 | lr 2.8151e-04 | norm: 0.5733 | dt: 6309.68ms | tok/sec: 83092.6 | mem: 108909MB | Time Left: 00 hrs : 25 mins
|
| 380 |
+
[TIMING] data: 3.4ms | fwd: 2021.8ms | loss: 6.4ms | bwd: 4220.8ms | clip: 0.0ms | optim: 49.4ms
|
| 381 |
+
step 160 | loss: 1.589306 | lr 2.8090e-04 | norm: 0.4747 | dt: 6356.37ms | tok/sec: 82482.4 | mem: 108909MB | Time Left: 00 hrs : 25 mins
|
| 382 |
+
[TIMING] data: 3.5ms | fwd: 2075.3ms | loss: 6.4ms | bwd: 4213.9ms | clip: 0.0ms | optim: 49.5ms
|
| 383 |
+
step 161 | loss: 1.558533 | lr 2.8028e-04 | norm: 0.4875 | dt: 6343.32ms | tok/sec: 82651.9 | mem: 108909MB | Time Left: 00 hrs : 25 mins
|
| 384 |
+
[TIMING] data: 3.3ms | fwd: 2045.2ms | loss: 6.4ms | bwd: 4230.9ms | clip: 0.0ms | optim: 49.5ms
|
| 385 |
+
step 162 | loss: 1.552307 | lr 2.7965e-04 | norm: 0.4857 | dt: 6305.58ms | tok/sec: 83146.6 | mem: 108909MB | Time Left: 00 hrs : 25 mins
|
| 386 |
+
[TIMING] data: 3.4ms | fwd: 2016.4ms | loss: 6.4ms | bwd: 4222.1ms | clip: 0.0ms | optim: 49.4ms
|
| 387 |
+
step 163 | loss: 1.561074 | lr 2.7902e-04 | norm: 0.4315 | dt: 6313.94ms | tok/sec: 83036.6 | mem: 108909MB | Time Left: 00 hrs : 24 mins
|
| 388 |
+
[TIMING] data: 3.2ms | fwd: 2015.4ms | loss: 6.4ms | bwd: 4231.3ms | clip: 0.0ms | optim: 49.4ms
|
| 389 |
+
step 164 | loss: 1.494718 | lr 2.7837e-04 | norm: 0.4518 | dt: 6313.21ms | tok/sec: 83046.2 | mem: 108909MB | Time Left: 00 hrs : 24 mins
|
| 390 |
+
[TIMING] data: 3.4ms | fwd: 2010.0ms | loss: 6.4ms | bwd: 4236.0ms | clip: 0.0ms | optim: 49.5ms
|
| 391 |
+
step 165 | loss: 1.550173 | lr 2.7771e-04 | norm: 0.5249 | dt: 6329.39ms | tok/sec: 82833.9 | mem: 108909MB | Time Left: 00 hrs : 24 mins
|
| 392 |
+
[TIMING] data: 3.3ms | fwd: 2015.0ms | loss: 6.4ms | bwd: 4247.3ms | clip: 0.0ms | optim: 49.5ms
|
| 393 |
+
step 166 | loss: 1.549638 | lr 2.7705e-04 | norm: 0.5183 | dt: 6318.25ms | tok/sec: 82979.9 | mem: 108909MB | Time Left: 00 hrs : 24 mins
|
| 394 |
+
[TIMING] data: 3.4ms | fwd: 2019.4ms | loss: 6.4ms | bwd: 4231.6ms | clip: 0.0ms | optim: 49.5ms
|
| 395 |
+
step 167 | loss: 1.567677 | lr 2.7638e-04 | norm: 0.5612 | dt: 6319.49ms | tok/sec: 82963.6 | mem: 108909MB | Time Left: 00 hrs : 24 mins
|
| 396 |
+
[TIMING] data: 3.3ms | fwd: 2016.6ms | loss: 6.4ms | bwd: 4235.6ms | clip: 0.0ms | optim: 49.7ms
|
| 397 |
+
step 168 | loss: 1.598396 | lr 2.7570e-04 | norm: 0.4648 | dt: 6299.20ms | tok/sec: 83230.9 | mem: 108909MB | Time Left: 00 hrs : 24 mins
|
| 398 |
+
[TIMING] data: 3.4ms | fwd: 2015.9ms | loss: 6.4ms | bwd: 4216.2ms | clip: 0.0ms | optim: 49.4ms
|
| 399 |
+
step 169 | loss: 1.517734 | lr 2.7501e-04 | norm: 0.5128 | dt: 6302.75ms | tok/sec: 83184.0 | mem: 108909MB | Time Left: 00 hrs : 24 mins
|
| 400 |
+
[TIMING] data: 3.4ms | fwd: 2011.6ms | loss: 6.4ms | bwd: 4223.9ms | clip: 0.0ms | optim: 49.4ms
|
| 401 |
+
step 170 | loss: 1.526498 | lr 2.7431e-04 | norm: 0.4985 | dt: 6296.95ms | tok/sec: 83260.6 | mem: 108909MB | Time Left: 00 hrs : 24 mins
|
| 402 |
+
[TIMING] data: 3.4ms | fwd: 2008.0ms | loss: 6.4ms | bwd: 4221.9ms | clip: 0.0ms | optim: 49.3ms
|
| 403 |
+
step 171 | loss: 1.589458 | lr 2.7361e-04 | norm: 0.5621 | dt: 6331.36ms | tok/sec: 82808.1 | mem: 108909MB | Time Left: 00 hrs : 24 mins
|
| 404 |
+
[TIMING] data: 3.3ms | fwd: 2049.7ms | loss: 6.4ms | bwd: 4214.5ms | clip: 0.0ms | optim: 49.5ms
|
| 405 |
+
step 172 | loss: 1.551229 | lr 2.7290e-04 | norm: 0.4868 | dt: 6348.52ms | tok/sec: 82584.2 | mem: 108909MB | Time Left: 00 hrs : 24 mins
|
| 406 |
+
[TIMING] data: 3.4ms | fwd: 2061.1ms | loss: 6.4ms | bwd: 4219.9ms | clip: 0.0ms | optim: 49.8ms
|
| 407 |
+
step 173 | loss: 1.536721 | lr 2.7218e-04 | norm: 0.4725 | dt: 6330.28ms | tok/sec: 82822.2 | mem: 108909MB | Time Left: 00 hrs : 23 mins
|
| 408 |
+
[TIMING] data: 3.4ms | fwd: 2028.6ms | loss: 6.4ms | bwd: 4234.3ms | clip: 0.0ms | optim: 49.5ms
|
| 409 |
+
step 174 | loss: 1.561139 | lr 2.7145e-04 | norm: 0.4796 | dt: 6308.29ms | tok/sec: 83111.0 | mem: 108909MB | Time Left: 00 hrs : 23 mins
|
| 410 |
+
[TIMING] data: 3.4ms | fwd: 2020.0ms | loss: 6.4ms | bwd: 4221.1ms | clip: 0.0ms | optim: 49.4ms
|
| 411 |
+
step 175 | loss: 1.546170 | lr 2.7071e-04 | norm: 0.5087 | dt: 6284.76ms | tok/sec: 83422.1 | mem: 108909MB | Time Left: 00 hrs : 23 mins
|
| 412 |
+
[TIMING] data: 3.4ms | fwd: 2010.3ms | loss: 6.4ms | bwd: 4207.3ms | clip: 0.0ms | optim: 49.4ms
|
| 413 |
+
step 176 | loss: 1.510554 | lr 2.6997e-04 | norm: 0.4955 | dt: 6326.66ms | tok/sec: 82869.6 | mem: 108909MB | Time Left: 00 hrs : 23 mins
|
| 414 |
+
[TIMING] data: 3.4ms | fwd: 2015.5ms | loss: 6.4ms | bwd: 4244.2ms | clip: 0.0ms | optim: 49.3ms
|
| 415 |
+
step 177 | loss: 1.594345 | lr 2.6921e-04 | norm: 0.4702 | dt: 6303.45ms | tok/sec: 83174.7 | mem: 108909MB | Time Left: 00 hrs : 23 mins
|
| 416 |
+
[TIMING] data: 3.4ms | fwd: 2019.3ms | loss: 6.4ms | bwd: 4217.1ms | clip: 0.0ms | optim: 49.4ms
|
| 417 |
+
step 178 | loss: 1.565544 | lr 2.6845e-04 | norm: 0.6543 | dt: 6356.01ms | tok/sec: 82487.0 | mem: 108909MB | Time Left: 00 hrs : 23 mins
|
| 418 |
+
[TIMING] data: 3.5ms | fwd: 2067.6ms | loss: 6.4ms | bwd: 4220.6ms | clip: 0.0ms | optim: 49.8ms
|
| 419 |
+
step 179 | loss: 1.570838 | lr 2.6769e-04 | norm: 0.4854 | dt: 6338.31ms | tok/sec: 82717.4 | mem: 108909MB | Time Left: 00 hrs : 23 mins
|
| 420 |
+
[TIMING] data: 3.4ms | fwd: 2060.6ms | loss: 6.4ms | bwd: 4210.6ms | clip: 0.0ms | optim: 49.4ms
|
| 421 |
+
step 180 | loss: 1.608467 | lr 2.6691e-04 | norm: 0.4543 | dt: 6348.96ms | tok/sec: 82578.6 | mem: 108909MB | Time Left: 00 hrs : 23 mins
|
| 422 |
+
[TIMING] data: 3.3ms | fwd: 2078.2ms | loss: 6.4ms | bwd: 4203.8ms | clip: 0.0ms | optim: 49.4ms
|
| 423 |
+
step 181 | loss: 1.518504 | lr 2.6613e-04 | norm: 0.4037 | dt: 6300.05ms | tok/sec: 83219.7 | mem: 108909MB | Time Left: 00 hrs : 22 mins
|
| 424 |
+
[TIMING] data: 3.3ms | fwd: 2012.5ms | loss: 6.4ms | bwd: 4220.4ms | clip: 0.0ms | optim: 49.4ms
|
| 425 |
+
step 182 | loss: 1.519081 | lr 2.6534e-04 | norm: 0.3706 | dt: 6300.37ms | tok/sec: 83215.4 | mem: 108909MB | Time Left: 00 hrs : 22 mins
|
| 426 |
+
[TIMING] data: 3.5ms | fwd: 2013.4ms | loss: 6.4ms | bwd: 4219.8ms | clip: 0.0ms | optim: 49.5ms
|
| 427 |
+
step 183 | loss: 1.545064 | lr 2.6455e-04 | norm: 0.4007 | dt: 6298.80ms | tok/sec: 83236.2 | mem: 108909MB | Time Left: 00 hrs : 22 mins
|
| 428 |
+
[TIMING] data: 3.3ms | fwd: 2010.6ms | loss: 6.4ms | bwd: 4221.2ms | clip: 0.0ms | optim: 49.4ms
|
| 429 |
+
step 184 | loss: 1.546265 | lr 2.6374e-04 | norm: 0.4499 | dt: 6286.92ms | tok/sec: 83393.4 | mem: 108909MB | Time Left: 00 hrs : 22 mins
|
| 430 |
+
[TIMING] data: 3.3ms | fwd: 2014.6ms | loss: 6.4ms | bwd: 4205.4ms | clip: 0.0ms | optim: 49.2ms
|
| 431 |
+
step 185 | loss: 1.517903 | lr 2.6293e-04 | norm: 0.3830 | dt: 6304.95ms | tok/sec: 83155.0 | mem: 108909MB | Time Left: 00 hrs : 22 mins
|
| 432 |
+
[TIMING] data: 3.3ms | fwd: 2015.7ms | loss: 6.4ms | bwd: 4221.2ms | clip: 0.0ms | optim: 50.4ms
|
| 433 |
+
step 186 | loss: 1.585939 | lr 2.6211e-04 | norm: 0.4245 | dt: 6312.37ms | tok/sec: 83057.2 | mem: 108909MB | Time Left: 00 hrs : 22 mins
|
| 434 |
+
[TIMING] data: 3.4ms | fwd: 2015.8ms | loss: 6.4ms | bwd: 4229.2ms | clip: 0.0ms | optim: 49.6ms
|
| 435 |
+
step 187 | loss: 1.557112 | lr 2.6129e-04 | norm: 0.4315 | dt: 6293.00ms | tok/sec: 83312.9 | mem: 108909MB | Time Left: 00 hrs : 22 mins
|
| 436 |
+
[TIMING] data: 3.3ms | fwd: 2002.2ms | loss: 6.4ms | bwd: 4223.7ms | clip: 0.0ms | optim: 49.4ms
|
| 437 |
+
step 188 | loss: 1.538441 | lr 2.6046e-04 | norm: 0.4569 | dt: 6300.70ms | tok/sec: 83211.0 | mem: 108909MB | Time Left: 00 hrs : 22 mins
|
| 438 |
+
[TIMING] data: 3.4ms | fwd: 2009.8ms | loss: 6.4ms | bwd: 4223.4ms | clip: 0.0ms | optim: 49.6ms
|
| 439 |
+
step 189 | loss: 1.548894 | lr 2.5962e-04 | norm: 0.3525 | dt: 6327.37ms | tok/sec: 82860.4 | mem: 108909MB | Time Left: 00 hrs : 22 mins
|
| 440 |
+
[TIMING] data: 3.2ms | fwd: 2030.5ms | loss: 6.4ms | bwd: 4229.9ms | clip: 0.0ms | optim: 49.4ms
|
| 441 |
+
step 190 | loss: 1.522655 | lr 2.5878e-04 | norm: 0.4214 | dt: 6354.27ms | tok/sec: 82509.6 | mem: 108909MB | Time Left: 00 hrs : 22 mins
|
| 442 |
+
[TIMING] data: 3.3ms | fwd: 2063.5ms | loss: 6.4ms | bwd: 4223.4ms | clip: 0.0ms | optim: 49.4ms
|
| 443 |
+
step 191 | loss: 1.544763 | lr 2.5793e-04 | norm: 0.3714 | dt: 6323.34ms | tok/sec: 82913.1 | mem: 108909MB | Time Left: 00 hrs : 22 mins
|
| 444 |
+
[TIMING] data: 3.4ms | fwd: 2012.5ms | loss: 6.4ms | bwd: 4243.6ms | clip: 0.0ms | optim: 49.5ms
|
| 445 |
+
step 192 | loss: 1.531644 | lr 2.5707e-04 | norm: 0.5406 | dt: 6287.51ms | tok/sec: 83385.6 | mem: 108909MB | Time Left: 00 hrs : 21 mins
|
| 446 |
+
[TIMING] data: 3.5ms | fwd: 2009.8ms | loss: 6.4ms | bwd: 4210.6ms | clip: 0.0ms | optim: 49.4ms
|
| 447 |
+
step 193 | loss: 1.527502 | lr 2.5621e-04 | norm: 0.4157 | dt: 6304.97ms | tok/sec: 83154.7 | mem: 108909MB | Time Left: 00 hrs : 21 mins
|
| 448 |
+
[TIMING] data: 3.3ms | fwd: 2018.3ms | loss: 6.4ms | bwd: 4218.9ms | clip: 0.0ms | optim: 50.0ms
|
| 449 |
+
step 194 | loss: 1.537738 | lr 2.5534e-04 | norm: 0.4412 | dt: 6309.68ms | tok/sec: 83092.6 | mem: 108909MB | Time Left: 00 hrs : 21 mins
|
| 450 |
+
[TIMING] data: 3.5ms | fwd: 2010.9ms | loss: 6.4ms | bwd: 4231.6ms | clip: 0.0ms | optim: 49.4ms
|
| 451 |
+
step 195 | loss: 1.519402 | lr 2.5446e-04 | norm: 0.4267 | dt: 6281.65ms | tok/sec: 83463.5 | mem: 108909MB | Time Left: 00 hrs : 21 mins
|
| 452 |
+
[TIMING] data: 3.3ms | fwd: 2014.2ms | loss: 6.4ms | bwd: 4200.4ms | clip: 0.0ms | optim: 49.5ms
|
| 453 |
+
step 196 | loss: 1.552346 | lr 2.5358e-04 | norm: 0.9164 | dt: 6314.61ms | tok/sec: 83027.8 | mem: 108909MB | Time Left: 00 hrs : 21 mins
|
| 454 |
+
[TIMING] data: 3.3ms | fwd: 2010.0ms | loss: 6.4ms | bwd: 4237.6ms | clip: 0.0ms | optim: 49.3ms
|
| 455 |
+
step 197 | loss: 1.558954 | lr 2.5270e-04 | norm: 0.4911 | dt: 6335.28ms | tok/sec: 82756.9 | mem: 108909MB | Time Left: 00 hrs : 21 mins
|
| 456 |
+
[TIMING] data: 3.4ms | fwd: 2077.9ms | loss: 6.4ms | bwd: 4190.0ms | clip: 0.0ms | optim: 49.7ms
|
| 457 |
+
step 198 | loss: 1.508947 | lr 2.5180e-04 | norm: 0.3714 | dt: 6359.11ms | tok/sec: 82446.8 | mem: 108909MB | Time Left: 00 hrs : 21 mins
|
| 458 |
+
[TIMING] data: 3.5ms | fwd: 2068.8ms | loss: 6.4ms | bwd: 4222.9ms | clip: 0.0ms | optim: 49.6ms
|
| 459 |
+
step 199 | loss: 1.520034 | lr 2.5090e-04 | norm: 0.3761 | dt: 6309.29ms | tok/sec: 83097.8 | mem: 108909MB | Time Left: 00 hrs : 21 mins
|
| 460 |
+
[TIMING] data: 3.3ms | fwd: 2007.9ms | loss: 6.4ms | bwd: 4234.2ms | clip: 0.0ms | optim: 49.5ms
|
| 461 |
+
[CheckpointManager] Saved: checkpoints_bluscriptcp/rank_0/blumodelcp_run13_step_200.ckpt (safely)
|
| 462 |
+
step 200 | loss: 1.515468 | lr 2.5000e-04 | norm: 0.3363 | dt: 6274.53ms | tok/sec: 83558.1 | mem: 108909MB | Time Left: 00 hrs : 20 mins
|
| 463 |
+
[TIMING] data: 3.3ms | fwd: 2091.9ms | loss: 6.4ms | bwd: 4115.5ms | clip: 0.0ms | optim: 49.4ms
|
| 464 |
+
step 201 | loss: 1.543592 | lr 2.4909e-04 | norm: 0.3853 | dt: 6197.18ms | tok/sec: 84601.1 | mem: 108909MB | Time Left: 00 hrs : 20 mins
|
| 465 |
+
[TIMING] data: 3.3ms | fwd: 1988.0ms | loss: 6.4ms | bwd: 4142.1ms | clip: 0.0ms | optim: 49.5ms
|
| 466 |
+
step 202 | loss: 1.518495 | lr 2.4818e-04 | norm: 0.3703 | dt: 6216.37ms | tok/sec: 84339.9 | mem: 108909MB | Time Left: 00 hrs : 20 mins
|
| 467 |
+
[TIMING] data: 3.5ms | fwd: 1992.1ms | loss: 6.3ms | bwd: 4157.0ms | clip: 0.0ms | optim: 49.5ms
|
| 468 |
+
step 203 | loss: 1.594687 | lr 2.4726e-04 | norm: 0.3279 | dt: 6224.03ms | tok/sec: 84236.1 | mem: 108909MB | Time Left: 00 hrs : 20 mins
|
| 469 |
+
[TIMING] data: 3.4ms | fwd: 1991.0ms | loss: 6.4ms | bwd: 4165.4ms | clip: 0.0ms | optim: 49.8ms
|
| 470 |
+
step 204 | loss: 1.508122 | lr 2.4633e-04 | norm: 0.3350 | dt: 6244.08ms | tok/sec: 83965.6 | mem: 108909MB | Time Left: 00 hrs : 20 mins
|
| 471 |
+
[TIMING] data: 3.2ms | fwd: 1998.7ms | loss: 6.3ms | bwd: 4178.5ms | clip: 0.0ms | optim: 49.3ms
|
| 472 |
+
step 205 | loss: 1.481425 | lr 2.4540e-04 | norm: 0.4867 | dt: 6300.96ms | tok/sec: 83207.6 | mem: 108909MB | Time Left: 00 hrs : 20 mins
|
| 473 |
+
[TIMING] data: 3.4ms | fwd: 2004.3ms | loss: 6.4ms | bwd: 4229.4ms | clip: 0.0ms | optim: 49.5ms
|
| 474 |
+
step 206 | loss: 1.561807 | lr 2.4446e-04 | norm: 0.4345 | dt: 6308.09ms | tok/sec: 83113.5 | mem: 108909MB | Time Left: 00 hrs : 20 mins
|
| 475 |
+
[TIMING] data: 3.5ms | fwd: 2014.5ms | loss: 6.4ms | bwd: 4226.3ms | clip: 0.0ms | optim: 49.3ms
|
| 476 |
+
step 207 | loss: 1.582822 | lr 2.4352e-04 | norm: 0.3482 | dt: 6330.54ms | tok/sec: 82818.9 | mem: 108909MB | Time Left: 00 hrs : 20 mins
|
| 477 |
+
[TIMING] data: 3.4ms | fwd: 2021.2ms | loss: 6.4ms | bwd: 4241.9ms | clip: 0.0ms | optim: 49.6ms
|
| 478 |
+
step 208 | loss: 1.465932 | lr 2.4258e-04 | norm: 0.3645 | dt: 6351.70ms | tok/sec: 82543.0 | mem: 108909MB | Time Left: 00 hrs : 20 mins
|
| 479 |
+
[TIMING] data: 3.4ms | fwd: 2027.3ms | loss: 6.4ms | bwd: 4257.3ms | clip: 0.0ms | optim: 49.3ms
|
| 480 |
+
step 209 | loss: 1.552704 | lr 2.4163e-04 | norm: 0.5003 | dt: 6382.09ms | tok/sec: 82149.9 | mem: 108909MB | Time Left: 00 hrs : 20 mins
|
| 481 |
+
[TIMING] data: 3.5ms | fwd: 2067.9ms | loss: 6.4ms | bwd: 4246.9ms | clip: 0.0ms | optim: 49.5ms
|
| 482 |
+
step 210 | loss: 1.560924 | lr 2.4067e-04 | norm: 0.4204 | dt: 6343.38ms | tok/sec: 82651.2 | mem: 108909MB | Time Left: 00 hrs : 20 mins
|
| 483 |
+
[TIMING] data: 3.4ms | fwd: 2021.8ms | loss: 6.4ms | bwd: 4254.5ms | clip: 0.0ms | optim: 49.3ms
|
| 484 |
+
step 211 | loss: 1.577441 | lr 2.3971e-04 | norm: 0.2780 | dt: 6369.29ms | tok/sec: 82315.0 | mem: 108909MB | Time Left: 00 hrs : 20 mins
|
| 485 |
+
[TIMING] data: 3.3ms | fwd: 2023.9ms | loss: 6.4ms | bwd: 4278.3ms | clip: 0.0ms | optim: 49.4ms
|
| 486 |
+
step 212 | loss: 1.523178 | lr 2.3875e-04 | norm: 0.3415 | dt: 6376.56ms | tok/sec: 82221.1 | mem: 108909MB | Time Left: 00 hrs : 19 mins
|
| 487 |
+
[TIMING] data: 3.4ms | fwd: 2029.4ms | loss: 6.5ms | bwd: 4279.9ms | clip: 0.0ms | optim: 49.4ms
|
| 488 |
+
step 213 | loss: 1.532332 | lr 2.3778e-04 | norm: 0.3428 | dt: 6336.42ms | tok/sec: 82742.0 | mem: 108909MB | Time Left: 00 hrs : 19 mins
|
| 489 |
+
[TIMING] data: 3.4ms | fwd: 2022.6ms | loss: 6.4ms | bwd: 4246.6ms | clip: 0.0ms | optim: 49.3ms
|
| 490 |
+
step 214 | loss: 1.526189 | lr 2.3681e-04 | norm: 0.4383 | dt: 6379.65ms | tok/sec: 82181.3 | mem: 108909MB | Time Left: 00 hrs : 19 mins
|
| 491 |
+
[TIMING] data: 3.3ms | fwd: 2044.2ms | loss: 6.4ms | bwd: 4268.4ms | clip: 0.0ms | optim: 49.4ms
|
| 492 |
+
step 215 | loss: 1.536538 | lr 2.3584e-04 | norm: 0.4139 | dt: 6378.23ms | tok/sec: 82199.7 | mem: 108909MB | Time Left: 00 hrs : 19 mins
|
| 493 |
+
[TIMING] data: 3.4ms | fwd: 2030.4ms | loss: 6.4ms | bwd: 4280.6ms | clip: 0.0ms | optim: 49.5ms
|
| 494 |
+
step 216 | loss: 1.505207 | lr 2.3486e-04 | norm: 0.3868 | dt: 6372.58ms | tok/sec: 82272.5 | mem: 108909MB | Time Left: 00 hrs : 19 mins
|
| 495 |
+
[TIMING] data: 3.4ms | fwd: 2027.4ms | loss: 6.4ms | bwd: 4278.1ms | clip: 0.0ms | optim: 49.4ms
|
| 496 |
+
step 217 | loss: 1.519254 | lr 2.3387e-04 | norm: 0.2924 | dt: 6370.76ms | tok/sec: 82296.0 | mem: 108909MB | Time Left: 00 hrs : 19 mins
|
| 497 |
+
[TIMING] data: 3.3ms | fwd: 2027.3ms | loss: 6.4ms | bwd: 4276.3ms | clip: 0.0ms | optim: 49.5ms
|
| 498 |
+
step 218 | loss: 1.547643 | lr 2.3289e-04 | norm: 0.3269 | dt: 6415.39ms | tok/sec: 81723.5 | mem: 108909MB | Time Left: 00 hrs : 19 mins
|
| 499 |
+
[TIMING] data: 3.4ms | fwd: 2060.3ms | loss: 6.4ms | bwd: 4287.9ms | clip: 0.0ms | optim: 49.4ms
|
| 500 |
+
step 219 | loss: 1.571966 | lr 2.3190e-04 | norm: 0.3471 | dt: 6380.81ms | tok/sec: 82166.4 | mem: 108909MB | Time Left: 00 hrs : 19 mins
|
| 501 |
+
[TIMING] data: 3.3ms | fwd: 2030.7ms | loss: 6.4ms | bwd: 4283.0ms | clip: 0.0ms | optim: 49.3ms
|
| 502 |
+
step 220 | loss: 1.509900 | lr 2.3090e-04 | norm: 0.2972 | dt: 6356.96ms | tok/sec: 82474.6 | mem: 108909MB | Time Left: 00 hrs : 19 mins
|
| 503 |
+
[TIMING] data: 3.3ms | fwd: 2023.5ms | loss: 6.4ms | bwd: 4266.3ms | clip: 0.0ms | optim: 49.4ms
|
| 504 |
+
step 221 | loss: 1.497495 | lr 2.2990e-04 | norm: 0.3361 | dt: 6365.62ms | tok/sec: 82362.4 | mem: 108909MB | Time Left: 00 hrs : 18 mins
|
| 505 |
+
[TIMING] data: 3.4ms | fwd: 2052.2ms | loss: 6.4ms | bwd: 4246.2ms | clip: 0.0ms | optim: 49.5ms
|
| 506 |
+
step 222 | loss: 1.492427 | lr 2.2890e-04 | norm: 0.2918 | dt: 6375.24ms | tok/sec: 82238.2 | mem: 108909MB | Time Left: 00 hrs : 18 mins
|
| 507 |
+
[TIMING] data: 3.4ms | fwd: 2051.9ms | loss: 6.7ms | bwd: 4255.9ms | clip: 0.0ms | optim: 49.4ms
|
| 508 |
+
step 223 | loss: 1.519729 | lr 2.2790e-04 | norm: 0.3266 | dt: 6357.23ms | tok/sec: 82471.1 | mem: 108909MB | Time Left: 00 hrs : 18 mins
|
| 509 |
+
[TIMING] data: 3.4ms | fwd: 2018.6ms | loss: 6.4ms | bwd: 4271.3ms | clip: 0.0ms | optim: 49.5ms
|
| 510 |
+
step 224 | loss: 1.466115 | lr 2.2689e-04 | norm: 0.3746 | dt: 6343.84ms | tok/sec: 82645.3 | mem: 108909MB | Time Left: 00 hrs : 18 mins
|
| 511 |
+
[TIMING] data: 3.4ms | fwd: 2023.9ms | loss: 6.4ms | bwd: 4252.8ms | clip: 0.0ms | optim: 49.4ms
|
| 512 |
+
step 225 | loss: 1.554846 | lr 2.2588e-04 | norm: 0.3662 | dt: 6337.59ms | tok/sec: 82726.7 | mem: 108909MB | Time Left: 00 hrs : 18 mins
|
| 513 |
+
[TIMING] data: 3.3ms | fwd: 2021.0ms | loss: 6.4ms | bwd: 4249.4ms | clip: 0.0ms | optim: 49.4ms
|
| 514 |
+
step 226 | loss: 1.487506 | lr 2.2487e-04 | norm: 0.4012 | dt: 6335.98ms | tok/sec: 82747.8 | mem: 108909MB | Time Left: 00 hrs : 18 mins
|
| 515 |
+
[TIMING] data: 3.4ms | fwd: 2021.6ms | loss: 6.4ms | bwd: 4247.2ms | clip: 0.0ms | optim: 49.4ms
|
| 516 |
+
step 227 | loss: 1.479402 | lr 2.2385e-04 | norm: 0.2841 | dt: 6320.69ms | tok/sec: 82947.9 | mem: 108909MB | Time Left: 00 hrs : 18 mins
|
| 517 |
+
[TIMING] data: 3.4ms | fwd: 2020.3ms | loss: 6.4ms | bwd: 4233.3ms | clip: 0.0ms | optim: 49.4ms
|
| 518 |
+
step 228 | loss: 1.533440 | lr 2.2284e-04 | norm: 0.5841 | dt: 6387.96ms | tok/sec: 82074.4 | mem: 108909MB | Time Left: 00 hrs : 18 mins
|
| 519 |
+
[TIMING] data: 3.4ms | fwd: 2070.9ms | loss: 6.4ms | bwd: 4250.1ms | clip: 0.0ms | optim: 49.3ms
|
| 520 |
+
step 229 | loss: 1.506774 | lr 2.2181e-04 | norm: 0.2893 | dt: 6334.24ms | tok/sec: 82770.5 | mem: 108909MB | Time Left: 00 hrs : 18 mins
|
| 521 |
+
[TIMING] data: 3.4ms | fwd: 2020.0ms | loss: 6.4ms | bwd: 4247.2ms | clip: 0.0ms | optim: 49.4ms
|
| 522 |
+
step 230 | loss: 1.499669 | lr 2.2079e-04 | norm: 0.3589 | dt: 6331.55ms | tok/sec: 82805.6 | mem: 108909MB | Time Left: 00 hrs : 17 mins
|
| 523 |
+
[TIMING] data: 3.5ms | fwd: 2021.4ms | loss: 6.4ms | bwd: 4243.0ms | clip: 0.0ms | optim: 49.4ms
|
| 524 |
+
step 231 | loss: 1.486993 | lr 2.1977e-04 | norm: 0.3249 | dt: 6306.00ms | tok/sec: 83141.2 | mem: 108909MB | Time Left: 00 hrs : 17 mins
|
| 525 |
+
[TIMING] data: 3.5ms | fwd: 2017.3ms | loss: 6.4ms | bwd: 4221.5ms | clip: 0.0ms | optim: 49.5ms
|
| 526 |
+
step 232 | loss: 1.574881 | lr 2.1874e-04 | norm: 0.3193 | dt: 6340.37ms | tok/sec: 82690.4 | mem: 108909MB | Time Left: 00 hrs : 17 mins
|
| 527 |
+
[TIMING] data: 3.5ms | fwd: 2021.6ms | loss: 6.4ms | bwd: 4251.7ms | clip: 0.0ms | optim: 49.4ms
|
| 528 |
+
step 233 | loss: 1.508013 | lr 2.1771e-04 | norm: 0.2468 | dt: 6327.01ms | tok/sec: 82865.0 | mem: 108909MB | Time Left: 00 hrs : 17 mins
|
| 529 |
+
[TIMING] data: 3.3ms | fwd: 2025.7ms | loss: 6.4ms | bwd: 4234.3ms | clip: 0.0ms | optim: 49.3ms
|
| 530 |
+
step 234 | loss: 1.524792 | lr 2.1668e-04 | norm: 0.2990 | dt: 6327.45ms | tok/sec: 82859.3 | mem: 108909MB | Time Left: 00 hrs : 17 mins
|
| 531 |
+
[TIMING] data: 3.3ms | fwd: 2011.9ms | loss: 6.4ms | bwd: 4248.5ms | clip: 0.0ms | optim: 49.4ms
|
| 532 |
+
step 235 | loss: 1.561522 | lr 2.1564e-04 | norm: 0.2423 | dt: 6412.86ms | tok/sec: 81755.8 | mem: 108909MB | Time Left: 00 hrs : 17 mins
|
| 533 |
+
[TIMING] data: 3.5ms | fwd: 2092.1ms | loss: 6.4ms | bwd: 4253.4ms | clip: 0.0ms | optim: 49.5ms
|
| 534 |
+
step 236 | loss: 1.532568 | lr 2.1461e-04 | norm: 0.2922 | dt: 6317.35ms | tok/sec: 82991.7 | mem: 108909MB | Time Left: 00 hrs : 17 mins
|
| 535 |
+
[TIMING] data: 3.3ms | fwd: 2024.8ms | loss: 6.4ms | bwd: 4225.2ms | clip: 0.0ms | optim: 49.7ms
|
| 536 |
+
step 237 | loss: 1.511153 | lr 2.1357e-04 | norm: 0.3742 | dt: 6312.04ms | tok/sec: 83061.6 | mem: 108909MB | Time Left: 00 hrs : 17 mins
|
| 537 |
+
[TIMING] data: 3.4ms | fwd: 2012.8ms | loss: 6.4ms | bwd: 4232.1ms | clip: 0.0ms | optim: 49.4ms
|
| 538 |
+
step 238 | loss: 1.461659 | lr 2.1253e-04 | norm: 0.3239 | dt: 6330.51ms | tok/sec: 82819.3 | mem: 108909MB | Time Left: 00 hrs : 17 mins
|
| 539 |
+
[TIMING] data: 3.4ms | fwd: 2024.6ms | loss: 6.4ms | bwd: 4238.8ms | clip: 0.0ms | optim: 49.3ms
|
| 540 |
+
step 239 | loss: 1.480994 | lr 2.1149e-04 | norm: 0.2434 | dt: 6309.65ms | tok/sec: 83093.0 | mem: 108909MB | Time Left: 00 hrs : 16 mins
|
| 541 |
+
[TIMING] data: 3.4ms | fwd: 2017.9ms | loss: 6.4ms | bwd: 4224.6ms | clip: 0.0ms | optim: 49.4ms
|
| 542 |
+
step 240 | loss: 1.504617 | lr 2.1045e-04 | norm: 0.3472 | dt: 6399.22ms | tok/sec: 81930.0 | mem: 108909MB | Time Left: 00 hrs : 17 mins
|
| 543 |
+
[TIMING] data: 3.4ms | fwd: 2084.7ms | loss: 6.4ms | bwd: 4247.0ms | clip: 0.0ms | optim: 49.7ms
|
| 544 |
+
step 241 | loss: 1.484900 | lr 2.0941e-04 | norm: 0.2350 | dt: 6329.71ms | tok/sec: 82829.8 | mem: 108909MB | Time Left: 00 hrs : 16 mins
|
| 545 |
+
[TIMING] data: 3.3ms | fwd: 2021.4ms | loss: 6.4ms | bwd: 4241.2ms | clip: 0.0ms | optim: 49.4ms
|
| 546 |
+
step 242 | loss: 1.526330 | lr 2.0837e-04 | norm: 0.3010 | dt: 6328.15ms | tok/sec: 82850.1 | mem: 108909MB | Time Left: 00 hrs : 16 mins
|
| 547 |
+
[TIMING] data: 3.4ms | fwd: 2017.8ms | loss: 6.4ms | bwd: 4243.3ms | clip: 0.0ms | optim: 49.3ms
|
| 548 |
+
step 243 | loss: 1.494656 | lr 2.0732e-04 | norm: 0.2456 | dt: 6314.70ms | tok/sec: 83026.5 | mem: 108909MB | Time Left: 00 hrs : 16 mins
|
| 549 |
+
[TIMING] data: 3.4ms | fwd: 2018.8ms | loss: 6.4ms | bwd: 4228.7ms | clip: 0.0ms | optim: 49.4ms
|
| 550 |
+
step 244 | loss: 1.518198 | lr 2.0628e-04 | norm: 0.2675 | dt: 6348.66ms | tok/sec: 82582.5 | mem: 108909MB | Time Left: 00 hrs : 16 mins
|
| 551 |
+
[TIMING] data: 3.4ms | fwd: 2023.7ms | loss: 6.4ms | bwd: 4257.0ms | clip: 0.0ms | optim: 50.2ms
|
| 552 |
+
step 245 | loss: 1.521052 | lr 2.0523e-04 | norm: 0.2581 | dt: 6308.14ms | tok/sec: 83112.9 | mem: 108909MB | Time Left: 00 hrs : 16 mins
|
| 553 |
+
[TIMING] data: 3.3ms | fwd: 2015.5ms | loss: 6.4ms | bwd: 4225.3ms | clip: 0.0ms | optim: 49.7ms
|
| 554 |
+
step 246 | loss: 1.502181 | lr 2.0419e-04 | norm: 0.1972 | dt: 6346.20ms | tok/sec: 82614.5 | mem: 108909MB | Time Left: 00 hrs : 16 mins
|
| 555 |
+
[TIMING] data: 3.4ms | fwd: 2057.8ms | loss: 6.4ms | bwd: 4221.3ms | clip: 0.0ms | optim: 49.3ms
|
| 556 |
+
step 247 | loss: 1.523221 | lr 2.0314e-04 | norm: 0.2856 | dt: 6349.26ms | tok/sec: 82574.7 | mem: 108909MB | Time Left: 00 hrs : 16 mins
|
| 557 |
+
[TIMING] data: 3.3ms | fwd: 2065.1ms | loss: 6.5ms | bwd: 4216.8ms | clip: 0.0ms | optim: 49.6ms
|
| 558 |
+
step 248 | loss: 1.499708 | lr 2.0209e-04 | norm: 0.2951 | dt: 6318.24ms | tok/sec: 82980.1 | mem: 108909MB | Time Left: 00 hrs : 16 mins
|
| 559 |
+
[TIMING] data: 3.3ms | fwd: 2012.7ms | loss: 6.4ms | bwd: 4238.3ms | clip: 0.0ms | optim: 49.5ms
|
| 560 |
+
step 249 | loss: 1.498004 | lr 2.0105e-04 | norm: 0.2532 | dt: 6318.23ms | tok/sec: 82980.2 | mem: 108909MB | Time Left: 00 hrs : 15 mins
|
| 561 |
+
[TIMING] data: 3.3ms | fwd: 2011.2ms | loss: 6.4ms | bwd: 4238.5ms | clip: 0.0ms | optim: 50.7ms
|
| 562 |
+
Current shard: /root/Adhi/BluTrain/dist/Context_Parallelism/in2_data/arm_b/in2b_val_00000.bin
|
| 563 |
+
100000000
|
| 564 |
+
validation loss: 1.5051
|
| 565 |
+
step 250 | loss: 1.479036 | lr 2.0000e-04 | norm: 0.2610 | dt: 11387.78ms | tok/sec: 46039.5 | mem: 108909MB | Time Left: 00 hrs : 28 mins
|
| 566 |
+
[TIMING] data: 3.3ms | fwd: 2013.6ms | loss: 6.4ms | bwd: 4229.9ms | clip: 0.0ms | optim: 49.6ms
|
| 567 |
+
step 251 | loss: 1.524658 | lr 1.9895e-04 | norm: 0.2094 | dt: 6312.76ms | tok/sec: 83052.2 | mem: 108909MB | Time Left: 00 hrs : 15 mins
|
| 568 |
+
[TIMING] data: 3.3ms | fwd: 2007.3ms | loss: 6.4ms | bwd: 4238.3ms | clip: 0.0ms | optim: 49.5ms
|
| 569 |
+
step 252 | loss: 1.530877 | lr 1.9791e-04 | norm: 0.3166 | dt: 6324.16ms | tok/sec: 82902.5 | mem: 108909MB | Time Left: 00 hrs : 15 mins
|
| 570 |
+
[TIMING] data: 3.3ms | fwd: 2023.0ms | loss: 6.4ms | bwd: 4234.2ms | clip: 0.0ms | optim: 49.3ms
|
| 571 |
+
step 253 | loss: 1.498913 | lr 1.9686e-04 | norm: 0.2670 | dt: 6348.04ms | tok/sec: 82590.5 | mem: 108909MB | Time Left: 00 hrs : 15 mins
|
| 572 |
+
[TIMING] data: 3.3ms | fwd: 2048.2ms | loss: 6.4ms | bwd: 4232.8ms | clip: 0.0ms | optim: 49.5ms
|
| 573 |
+
step 254 | loss: 1.514113 | lr 1.9581e-04 | norm: 0.2580 | dt: 6325.70ms | tok/sec: 82882.2 | mem: 108909MB | Time Left: 00 hrs : 15 mins
|
| 574 |
+
[TIMING] data: 3.3ms | fwd: 2033.2ms | loss: 6.4ms | bwd: 4225.4ms | clip: 0.0ms | optim: 49.4ms
|
| 575 |
+
step 255 | loss: 1.534029 | lr 1.9477e-04 | norm: 0.1956 | dt: 6346.03ms | tok/sec: 82616.8 | mem: 108909MB | Time Left: 00 hrs : 15 mins
|
| 576 |
+
[TIMING] data: 3.3ms | fwd: 2024.9ms | loss: 6.4ms | bwd: 4254.1ms | clip: 0.0ms | optim: 49.4ms
|
| 577 |
+
step 256 | loss: 1.479002 | lr 1.9372e-04 | norm: 0.2145 | dt: 6320.09ms | tok/sec: 82955.8 | mem: 108909MB | Time Left: 00 hrs : 15 mins
|
| 578 |
+
[TIMING] data: 3.3ms | fwd: 2017.1ms | loss: 6.4ms | bwd: 4236.2ms | clip: 0.0ms | optim: 49.2ms
|
| 579 |
+
step 257 | loss: 1.474892 | lr 1.9268e-04 | norm: 0.3064 | dt: 6323.17ms | tok/sec: 82915.4 | mem: 108909MB | Time Left: 00 hrs : 15 mins
|
| 580 |
+
[TIMING] data: 3.3ms | fwd: 2013.7ms | loss: 6.4ms | bwd: 4242.2ms | clip: 0.0ms | optim: 49.6ms
|
| 581 |
+
step 258 | loss: 1.479650 | lr 1.9163e-04 | norm: 0.2714 | dt: 6328.00ms | tok/sec: 82852.1 | mem: 108909MB | Time Left: 00 hrs : 14 mins
|
| 582 |
+
[TIMING] data: 3.4ms | fwd: 2018.5ms | loss: 6.4ms | bwd: 4242.2ms | clip: 0.0ms | optim: 49.5ms
|
| 583 |
+
step 259 | loss: 1.469126 | lr 1.9059e-04 | norm: 0.2147 | dt: 6316.69ms | tok/sec: 83000.4 | mem: 108909MB | Time Left: 00 hrs : 14 mins
|
| 584 |
+
[TIMING] data: 3.4ms | fwd: 2010.3ms | loss: 6.4ms | bwd: 4239.3ms | clip: 0.0ms | optim: 49.4ms
|
| 585 |
+
step 260 | loss: 1.482932 | lr 1.8955e-04 | norm: 0.3710 | dt: 6349.33ms | tok/sec: 82573.7 | mem: 108909MB | Time Left: 00 hrs : 14 mins
|
| 586 |
+
[TIMING] data: 3.5ms | fwd: 2025.4ms | loss: 6.4ms | bwd: 4256.7ms | clip: 0.0ms | optim: 49.3ms
|
| 587 |
+
step 261 | loss: 1.472604 | lr 1.8851e-04 | norm: 0.3087 | dt: 6299.28ms | tok/sec: 83229.9 | mem: 108909MB | Time Left: 00 hrs : 14 mins
|
| 588 |
+
[TIMING] data: 3.4ms | fwd: 2014.0ms | loss: 6.4ms | bwd: 4218.1ms | clip: 0.0ms | optim: 49.4ms
|
| 589 |
+
step 262 | loss: 1.510692 | lr 1.8747e-04 | norm: 0.2209 | dt: 6306.56ms | tok/sec: 83133.7 | mem: 108909MB | Time Left: 00 hrs : 14 mins
|
| 590 |
+
[TIMING] data: 3.5ms | fwd: 2010.7ms | loss: 6.4ms | bwd: 4228.6ms | clip: 0.0ms | optim: 49.4ms
|
| 591 |
+
step 263 | loss: 1.521401 | lr 1.8643e-04 | norm: 0.2542 | dt: 6335.75ms | tok/sec: 82750.8 | mem: 108909MB | Time Left: 00 hrs : 14 mins
|
| 592 |
+
[TIMING] data: 3.5ms | fwd: 2042.3ms | loss: 6.4ms | bwd: 4226.2ms | clip: 0.0ms | optim: 49.4ms
|
| 593 |
+
step 264 | loss: 1.518334 | lr 1.8539e-04 | norm: 0.2676 | dt: 6346.58ms | tok/sec: 82609.5 | mem: 108909MB | Time Left: 00 hrs : 14 mins
|
| 594 |
+
[TIMING] data: 3.3ms | fwd: 2058.0ms | loss: 6.4ms | bwd: 4221.4ms | clip: 0.0ms | optim: 49.4ms
|
| 595 |
+
step 265 | loss: 1.467476 | lr 1.8436e-04 | norm: 0.3925 | dt: 6373.01ms | tok/sec: 82267.0 | mem: 108909MB | Time Left: 00 hrs : 14 mins
|
| 596 |
+
[TIMING] data: 3.3ms | fwd: 2068.0ms | loss: 6.4ms | bwd: 4237.8ms | clip: 0.0ms | optim: 49.5ms
|
| 597 |
+
Current shard: /root/Adhi/BluTrain/dist/Context_Parallelism/in2_data/arm_b/in2b_train_00002.bin
|
| 598 |
+
100000000
|
| 599 |
+
step 266 | loss: 1.439376 | lr 1.8332e-04 | norm: 0.2033 | dt: 6320.79ms | tok/sec: 82946.6 | mem: 108909MB | Time Left: 00 hrs : 14 mins
|
| 600 |
+
[TIMING] data: 14.4ms | fwd: 2033.8ms | loss: 6.4ms | bwd: 4209.0ms | clip: 0.0ms | optim: 49.3ms
|
| 601 |
+
step 267 | loss: 1.543246 | lr 1.8229e-04 | norm: 0.2407 | dt: 6323.34ms | tok/sec: 82913.1 | mem: 108909MB | Time Left: 00 hrs : 14 mins
|
| 602 |
+
[TIMING] data: 3.3ms | fwd: 2019.1ms | loss: 6.4ms | bwd: 4237.3ms | clip: 0.0ms | optim: 49.3ms
|
| 603 |
+
step 268 | loss: 1.523543 | lr 1.8126e-04 | norm: 0.2025 | dt: 6298.32ms | tok/sec: 83242.5 | mem: 108909MB | Time Left: 00 hrs : 13 mins
|
| 604 |
+
[TIMING] data: 3.5ms | fwd: 2010.2ms | loss: 6.4ms | bwd: 4221.0ms | clip: 0.0ms | optim: 49.3ms
|
| 605 |
+
step 269 | loss: 1.496513 | lr 1.8023e-04 | norm: 0.1905 | dt: 6305.69ms | tok/sec: 83145.3 | mem: 108909MB | Time Left: 00 hrs : 13 mins
|
| 606 |
+
[TIMING] data: 3.4ms | fwd: 2014.6ms | loss: 6.4ms | bwd: 4223.9ms | clip: 0.0ms | optim: 49.4ms
|
| 607 |
+
step 270 | loss: 1.457562 | lr 1.7921e-04 | norm: 0.2139 | dt: 6299.40ms | tok/sec: 83228.3 | mem: 108909MB | Time Left: 00 hrs : 13 mins
|
| 608 |
+
[TIMING] data: 3.3ms | fwd: 2011.4ms | loss: 6.4ms | bwd: 4221.0ms | clip: 0.0ms | optim: 49.3ms
|
| 609 |
+
step 271 | loss: 1.535891 | lr 1.7819e-04 | norm: 0.2134 | dt: 6311.65ms | tok/sec: 83066.8 | mem: 108909MB | Time Left: 00 hrs : 13 mins
|
| 610 |
+
[TIMING] data: 3.4ms | fwd: 2015.1ms | loss: 6.4ms | bwd: 4229.5ms | clip: 0.0ms | optim: 49.4ms
|
| 611 |
+
step 272 | loss: 1.501040 | lr 1.7716e-04 | norm: 0.2187 | dt: 6319.93ms | tok/sec: 82957.9 | mem: 108909MB | Time Left: 00 hrs : 13 mins
|
| 612 |
+
[TIMING] data: 3.3ms | fwd: 2014.7ms | loss: 6.4ms | bwd: 4238.3ms | clip: 0.0ms | optim: 49.3ms
|
| 613 |
+
step 273 | loss: 1.566223 | lr 1.7615e-04 | norm: 0.2625 | dt: 6308.45ms | tok/sec: 83108.8 | mem: 108909MB | Time Left: 00 hrs : 13 mins
|
| 614 |
+
[TIMING] data: 3.3ms | fwd: 2019.4ms | loss: 6.4ms | bwd: 4222.0ms | clip: 0.0ms | optim: 49.4ms
|
| 615 |
+
step 274 | loss: 1.495419 | lr 1.7513e-04 | norm: 0.2264 | dt: 6317.88ms | tok/sec: 82984.8 | mem: 108909MB | Time Left: 00 hrs : 13 mins
|
| 616 |
+
[TIMING] data: 3.4ms | fwd: 2022.2ms | loss: 6.4ms | bwd: 4228.5ms | clip: 0.0ms | optim: 49.5ms
|
| 617 |
+
step 275 | loss: 1.529794 | lr 1.7412e-04 | norm: 0.2556 | dt: 6314.58ms | tok/sec: 83028.1 | mem: 108909MB | Time Left: 00 hrs : 13 mins
|
| 618 |
+
[TIMING] data: 3.4ms | fwd: 2021.7ms | loss: 6.4ms | bwd: 4225.6ms | clip: 0.0ms | optim: 49.5ms
|
| 619 |
+
step 276 | loss: 1.466739 | lr 1.7311e-04 | norm: 0.1947 | dt: 6323.10ms | tok/sec: 82916.3 | mem: 108909MB | Time Left: 00 hrs : 13 mins
|
| 620 |
+
[TIMING] data: 3.4ms | fwd: 2010.3ms | loss: 6.4ms | bwd: 4245.6ms | clip: 0.0ms | optim: 49.5ms
|
| 621 |
+
step 277 | loss: 1.534216 | lr 1.7210e-04 | norm: 0.2265 | dt: 6318.67ms | tok/sec: 82974.4 | mem: 108909MB | Time Left: 00 hrs : 12 mins
|
| 622 |
+
[TIMING] data: 3.3ms | fwd: 2007.6ms | loss: 6.6ms | bwd: 4243.7ms | clip: 0.0ms | optim: 49.4ms
|
| 623 |
+
step 278 | loss: 1.468664 | lr 1.7110e-04 | norm: 0.2138 | dt: 6335.29ms | tok/sec: 82756.8 | mem: 108909MB | Time Left: 00 hrs : 12 mins
|
| 624 |
+
[TIMING] data: 3.4ms | fwd: 2018.6ms | loss: 6.4ms | bwd: 4249.3ms | clip: 0.0ms | optim: 49.7ms
|
| 625 |
+
step 279 | loss: 1.551666 | lr 1.7010e-04 | norm: 0.2405 | dt: 6311.24ms | tok/sec: 83072.1 | mem: 108909MB | Time Left: 00 hrs : 12 mins
|
| 626 |
+
[TIMING] data: 3.4ms | fwd: 2018.4ms | loss: 6.4ms | bwd: 4225.7ms | clip: 0.0ms | optim: 49.4ms
|
| 627 |
+
step 280 | loss: 1.486062 | lr 1.6910e-04 | norm: 0.3126 | dt: 6307.03ms | tok/sec: 83127.6 | mem: 108909MB | Time Left: 00 hrs : 12 mins
|
| 628 |
+
[TIMING] data: 3.4ms | fwd: 2014.8ms | loss: 6.4ms | bwd: 4225.2ms | clip: 0.0ms | optim: 49.3ms
|
| 629 |
+
step 281 | loss: 1.491243 | lr 1.6810e-04 | norm: 0.2130 | dt: 6330.03ms | tok/sec: 82825.5 | mem: 108909MB | Time Left: 00 hrs : 12 mins
|
| 630 |
+
[TIMING] data: 3.4ms | fwd: 2013.2ms | loss: 6.4ms | bwd: 4249.7ms | clip: 0.0ms | optim: 49.4ms
|
| 631 |
+
step 282 | loss: 1.489311 | lr 1.6711e-04 | norm: 0.2014 | dt: 6292.83ms | tok/sec: 83315.1 | mem: 108909MB | Time Left: 00 hrs : 12 mins
|
| 632 |
+
[TIMING] data: 3.4ms | fwd: 2013.9ms | loss: 6.4ms | bwd: 4211.8ms | clip: 0.0ms | optim: 49.4ms
|
| 633 |
+
step 283 | loss: 1.416811 | lr 1.6613e-04 | norm: 0.1893 | dt: 6325.16ms | tok/sec: 82889.3 | mem: 108909MB | Time Left: 00 hrs : 12 mins
|
| 634 |
+
[TIMING] data: 3.4ms | fwd: 2031.2ms | loss: 6.4ms | bwd: 4226.7ms | clip: 0.0ms | optim: 49.4ms
|
| 635 |
+
step 284 | loss: 1.525040 | lr 1.6514e-04 | norm: 0.1914 | dt: 6333.96ms | tok/sec: 82774.1 | mem: 108909MB | Time Left: 00 hrs : 12 mins
|
| 636 |
+
[TIMING] data: 3.4ms | fwd: 2041.1ms | loss: 6.4ms | bwd: 4225.4ms | clip: 0.0ms | optim: 49.5ms
|
| 637 |
+
step 285 | loss: 1.497472 | lr 1.6416e-04 | norm: 0.2121 | dt: 6326.71ms | tok/sec: 82869.0 | mem: 108909MB | Time Left: 00 hrs : 12 mins
|
| 638 |
+
[TIMING] data: 3.4ms | fwd: 2041.4ms | loss: 6.4ms | bwd: 4218.3ms | clip: 0.0ms | optim: 49.3ms
|
| 639 |
+
step 286 | loss: 1.476106 | lr 1.6319e-04 | norm: 0.2120 | dt: 6310.10ms | tok/sec: 83087.1 | mem: 108909MB | Time Left: 00 hrs : 11 mins
|
| 640 |
+
[TIMING] data: 3.3ms | fwd: 2012.5ms | loss: 6.4ms | bwd: 4230.5ms | clip: 0.0ms | optim: 49.4ms
|
| 641 |
+
step 287 | loss: 1.430440 | lr 1.6222e-04 | norm: 0.1705 | dt: 6310.17ms | tok/sec: 83086.2 | mem: 108909MB | Time Left: 00 hrs : 11 mins
|
| 642 |
+
[TIMING] data: 3.3ms | fwd: 2015.5ms | loss: 6.4ms | bwd: 4227.6ms | clip: 0.0ms | optim: 49.4ms
|
| 643 |
+
step 288 | loss: 1.457449 | lr 1.6125e-04 | norm: 0.2106 | dt: 6313.65ms | tok/sec: 83040.4 | mem: 108909MB | Time Left: 00 hrs : 11 mins
|
| 644 |
+
[TIMING] data: 3.4ms | fwd: 2015.3ms | loss: 6.4ms | bwd: 4231.1ms | clip: 0.0ms | optim: 49.5ms
|
| 645 |
+
step 289 | loss: 1.432535 | lr 1.6029e-04 | norm: 0.1601 | dt: 6317.72ms | tok/sec: 82986.9 | mem: 108909MB | Time Left: 00 hrs : 11 mins
|
| 646 |
+
[TIMING] data: 3.3ms | fwd: 2011.7ms | loss: 6.4ms | bwd: 4238.6ms | clip: 0.0ms | optim: 49.7ms
|
| 647 |
+
step 290 | loss: 1.415617 | lr 1.5933e-04 | norm: 0.3453 | dt: 6311.40ms | tok/sec: 83070.0 | mem: 108909MB | Time Left: 00 hrs : 11 mins
|
| 648 |
+
[TIMING] data: 3.5ms | fwd: 2016.9ms | loss: 6.4ms | bwd: 4227.4ms | clip: 0.0ms | optim: 49.3ms
|
| 649 |
+
step 291 | loss: 1.482946 | lr 1.5837e-04 | norm: 0.1834 | dt: 6309.17ms | tok/sec: 83099.4 | mem: 108909MB | Time Left: 00 hrs : 11 mins
|
| 650 |
+
[TIMING] data: 3.3ms | fwd: 2009.8ms | loss: 6.4ms | bwd: 4232.3ms | clip: 0.0ms | optim: 49.4ms
|
| 651 |
+
step 292 | loss: 1.491100 | lr 1.5742e-04 | norm: 0.2317 | dt: 6353.94ms | tok/sec: 82513.8 | mem: 108909MB | Time Left: 00 hrs : 11 mins
|
| 652 |
+
[TIMING] data: 3.4ms | fwd: 2082.3ms | loss: 6.4ms | bwd: 4204.5ms | clip: 0.0ms | optim: 49.5ms
|
| 653 |
+
step 293 | loss: 1.501492 | lr 1.5648e-04 | norm: 0.1793 | dt: 6313.35ms | tok/sec: 83044.3 | mem: 108909MB | Time Left: 00 hrs : 11 mins
|
| 654 |
+
[TIMING] data: 3.4ms | fwd: 2013.2ms | loss: 6.4ms | bwd: 4233.0ms | clip: 0.0ms | optim: 49.4ms
|
| 655 |
+
step 294 | loss: 1.435696 | lr 1.5554e-04 | norm: 0.1851 | dt: 6319.12ms | tok/sec: 82968.6 | mem: 108909MB | Time Left: 00 hrs : 11 mins
|
| 656 |
+
[TIMING] data: 3.4ms | fwd: 2018.0ms | loss: 6.4ms | bwd: 4233.6ms | clip: 0.0ms | optim: 49.8ms
|
| 657 |
+
step 295 | loss: 1.491356 | lr 1.5460e-04 | norm: 0.1827 | dt: 6321.44ms | tok/sec: 82938.0 | mem: 108909MB | Time Left: 00 hrs : 11 mins
|
| 658 |
+
[TIMING] data: 3.4ms | fwd: 2015.2ms | loss: 6.4ms | bwd: 4239.2ms | clip: 0.0ms | optim: 49.2ms
|
| 659 |
+
step 296 | loss: 1.504897 | lr 1.5367e-04 | norm: 0.2337 | dt: 6323.81ms | tok/sec: 82907.0 | mem: 108909MB | Time Left: 00 hrs : 10 mins
|
| 660 |
+
[TIMING] data: 3.3ms | fwd: 2018.9ms | loss: 6.4ms | bwd: 4236.5ms | clip: 0.0ms | optim: 50.7ms
|
| 661 |
+
step 297 | loss: 1.437167 | lr 1.5274e-04 | norm: 0.1649 | dt: 6315.66ms | tok/sec: 83014.0 | mem: 108909MB | Time Left: 00 hrs : 10 mins
|
| 662 |
+
[TIMING] data: 3.5ms | fwd: 2015.1ms | loss: 6.4ms | bwd: 4230.9ms | clip: 0.0ms | optim: 51.8ms
|
| 663 |
+
step 298 | loss: 1.450542 | lr 1.5182e-04 | norm: 0.2649 | dt: 6318.76ms | tok/sec: 82973.3 | mem: 108909MB | Time Left: 00 hrs : 10 mins
|
| 664 |
+
[TIMING] data: 3.4ms | fwd: 2012.5ms | loss: 6.4ms | bwd: 4237.0ms | clip: 0.0ms | optim: 51.4ms
|
| 665 |
+
step 299 | loss: 1.449647 | lr 1.5091e-04 | norm: 0.1781 | dt: 6340.15ms | tok/sec: 82693.3 | mem: 108909MB | Time Left: 00 hrs : 10 mins
|
| 666 |
+
[TIMING] data: 3.4ms | fwd: 2048.9ms | loss: 6.4ms | bwd: 4222.6ms | clip: 0.0ms | optim: 50.9ms
|
| 667 |
+
[WARN] checkpoint save at step 300 failed: Failed to flush checkpoint to: checkpoints_bluscriptcp/rank_3/blumodelcp_run13_step_300.tmp
|
| 668 |
+
[WARN] checkpoint save at step 300 failed: Failed to flush checkpoint to: checkpoints_bluscriptcp/rank_2/blumodelcp_run13_step_300.tmp
|
| 669 |
+
[WARN] checkpoint save at step 300 failed: Failed to flush checkpoint to: checkpoints_bluscriptcp/rank_1/blumodelcp_run13_step_300.tmp
|
| 670 |
+
[WARN] checkpoint save at step 300 failed: Failed to flush checkpoint to: checkpoints_bluscriptcp/rank_0/blumodelcp_run13_step_300.tmp
|
| 671 |
+
step 300 | loss: 1.499225 | lr 1.5000e-04 | norm: 0.2531 | dt: 6180.78ms | tok/sec: 84825.5 | mem: 108909MB | Time Left: 00 hrs : 10 mins
|
| 672 |
+
[TIMING] data: 3.3ms | fwd: 1987.5ms | loss: 6.4ms | bwd: 4126.2ms | clip: 0.0ms | optim: 49.4ms
|
| 673 |
+
step 301 | loss: 1.472106 | lr 1.4910e-04 | norm: 0.1649 | dt: 6223.69ms | tok/sec: 84240.7 | mem: 108909MB | Time Left: 00 hrs : 10 mins
|
| 674 |
+
[TIMING] data: 3.3ms | fwd: 1990.1ms | loss: 6.4ms | bwd: 4166.3ms | clip: 0.0ms | optim: 49.8ms
|
| 675 |
+
step 302 | loss: 1.473293 | lr 1.4820e-04 | norm: 0.1976 | dt: 6230.85ms | tok/sec: 84144.0 | mem: 108909MB | Time Left: 00 hrs : 10 mins
|
| 676 |
+
[TIMING] data: 3.2ms | fwd: 1996.4ms | loss: 6.4ms | bwd: 4167.6ms | clip: 0.0ms | optim: 49.4ms
|
| 677 |
+
step 303 | loss: 1.397302 | lr 1.4730e-04 | norm: 0.2037 | dt: 6252.91ms | tok/sec: 83847.1 | mem: 108909MB | Time Left: 00 hrs : 10 mins
|
| 678 |
+
[TIMING] data: 3.2ms | fwd: 2001.3ms | loss: 6.4ms | bwd: 4184.7ms | clip: 0.0ms | optim: 49.4ms
|
| 679 |
+
step 304 | loss: 1.4step 396 | loss: 1.463978 | lr 1.0009e-04 | norm: 0.1220 | dt: 6371.53ms | tok/sec: 82286.0 | mem: 108909MB | Time Left: 00 hrs : 00 mins
|
| 680 |
+
[TIMING] data: 3.3ms | fwd: 2075.1ms | loss: 6.4ms | bwd: 4229.4ms | clip: 0.0ms | optim: 49.4ms
|
| 681 |
+
step 397 | loss: 1.432412 | lr 1.0005e-04 | norm: 0.1516 | dt: 6305.59ms | tok/sec: 83146.6 | mem: 108909MB | Time Left: 00 hrs : 00 mins
|
| 682 |
+
[TIMING] data: 3.3ms | fwd: 2008.0ms | loss: 6.4ms | bwd: 4230.5ms | clip: 0.0ms | optim: 49.5ms
|
| 683 |
+
step 398 | loss: 1.444381 | lr 1.0002e-04 | norm: 0.1591 | dt: 6290.80ms | tok/sec: 83342.1 | mem: 108909MB | Time Left: 00 hrs : 00 mins
|
| 684 |
+
[TIMING] data: 3.3ms | fwd: 2007.2ms | loss: 6.4ms | bwd: 4216.6ms | clip: 0.0ms | optim: 49.5ms
|
| 685 |
+
Current shard: /root/Adhi/BluTrain/dist/Context_Parallelism/in2_data/arm_b/in2b_val_00000.bin
|
| 686 |
+
100000000
|
| 687 |
+
validation loss: 1.4434
|
| 688 |
+
Current shard: /root/Adhi/BluTrain/dist/Context_Parallelism/in2_data/arm_b/in2b_train_00000.bin
|
| 689 |
+
100000000
|
| 690 |
+
step 399 | loss: 1.451133 | lr 1.0001e-04 | norm: 0.1280 | dt: 11381.27ms | tok/sec: 46065.9 | mem: 108909MB | Time Left: 00 hrs : 00 mins
|
| 691 |
+
[TIMING] data: 15.9ms | fwd: 2017.6ms | loss: 6.4ms | bwd: 4236.8ms | clip: 0.0ms | optim: 49.4ms
|
| 692 |
+
|
| 693 |
+
=== Training complete ===
|
logs/in2/arm_c.trained
ADDED
|
@@ -0,0 +1 @@
|
|
|
|
|
|
|
| 1 |
+
14
|
logs/in2/arm_c_train.log
ADDED
|
@@ -0,0 +1,297 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
=== Llama Context Parallel Training Script (bluscriptCP) ===
|
| 2 |
+
[CP attn fusion] split (Stage-2 rope-once + Stage-3 Hopper causal-attn kernels)
|
| 3 |
+
[ZeRO-1] CP_ZERO_SYMMETRIC=1: NCCL symmetric memory mode (NVLS)
|
| 4 |
+
[watchdog] collective no-progress deadline 600 s, poll 2 s (BLU_NCCL_WATCHDOG=0 to disable)
|
| 5 |
+
[watchdog] collective no-progress deadline 600 s, poll 2 s (BLU_NCCL_WATCHDOG=0 to disable)
|
| 6 |
+
================ Configuration (bluscriptCP) ================
|
| 7 |
+
d_model/layers : 1536 / 36
|
| 8 |
+
q/kv heads : 12 / 4 (G=3, hd=128)
|
| 9 |
+
ffn_hidden : 4096 (SwiGLU)
|
| 10 |
+
context_length : 16384
|
| 11 |
+
vocab_size : 50304 tying=0
|
| 12 |
+
B x T : 4 x 16384
|
| 13 |
+
world_size : 4 (dp=1 x cp=4)
|
| 14 |
+
CP_ATTN_MODE : ring (load_balance=HeadTail)
|
| 15 |
+
global_batch : 524288 (grad_accum=8)
|
| 16 |
+
optimizer : Muon (2-D body) + AdamW (wte + 1-D) [ZeRO-1] muon_lr=0.001000
|
| 17 |
+
precision : bf16 Linears + fp32 norms/residual/logits (fp32 master weights in AdamW) [mask=15 QKV WO MLP HEAD]
|
| 18 |
+
max_lr/min_lr : 0.001 / 0.0001
|
| 19 |
+
warmup/steps : 1 / 400 (cosine, no stable phase)
|
| 20 |
+
checkpointing : true (ckpt_freq=100, dir=checkpoints_bluscriptcp)
|
| 21 |
+
=============================================================
|
| 22 |
+
Initializing model on CUDA device 0...
|
| 23 |
+
[YaRN][warn] seq_len(16384) != scale(4.000)*orig_maxpos(1024.0)=4096.0 -- cache scaled for a different context than it is sized for
|
| 24 |
+
[YaRN] build_rope_cache: seq_len=16384 hd=128 scale=4.000 low=7 high=25 m=1.1386
|
| 25 |
+
[InitFrom] weights loaded from /root/Adhi/BluTrain/dist/Context_Parallelism/checkpoints_bluscriptcp/rank_0/blumodelcp_run7_step_750.ckpt (base step 750, loss 1.28324) -- optimizer state and RNG intentionally NOT restored; pair with CP_REWARMUP
|
| 26 |
+
Parameters: 1060624896
|
| 27 |
+
Parameters per GPU: 1060624896
|
| 28 |
+
max_steps: 400
|
| 29 |
+
warmup_steps: 1
|
| 30 |
+
[Muon] adjust_lr=match_rms_adamw (default) ns_steps=5 momentum=0.95
|
| 31 |
+
[ZeRO-1/Muon] reduce-scatter/all-gather layout: 6 bucket(s), 12 collectives/step, shard 226492416 elems/rank, padding 0.000000% (flat buffer 1728.000000 MB)
|
| 32 |
+
[ZeRO-1/Muon] adjust_lr=match_rms_adamw | params=905969664 (216 orthogonalized / 0 AdamW) muon_elems=905969664 | W=4 owned=226492416 (42 tensors) | master+momentum: 1728.000000 MB/GPU
|
| 33 |
+
[ZeRO-1 v2] build=v2.4 supports: rs+ag | can-symmetric-window | can-fp32-reduce | can-bf16-reduce | can-reshard-v1-ckpt
|
| 34 |
+
[ZeRO-1 v2] ACTIVE: reduce_dtype=bfloat16 (= param dtype) symmetric=ON overlap=-/-
|
| 35 |
+
[ZeRO-1 v2] zero1-v2 reduce-scatter/all-gather (symmetric/NVLS)
|
| 36 |
+
[ZeRO-1 v2] params=77266944 padded=77266944 W=4 shard=19316736 | 3 buckets of 50.000000 MB (6553600 elems/rank/bucket) | optimizer state/GPU: 221.062500 MB (vs 884.250000 MB unsharded)
|
| 37 |
+
[ZeRO-1 v2] build=v2.4 supports: rs+ag | can-symmetric-window | can-fp32-reduce | can-bf16-reduce | can-reshard-v1-ckpt
|
| 38 |
+
[ZeRO-1 v2] ACTIVE: reduce_dtype=float32 (= param dtype) symmetric=ON overlap=-/-
|
| 39 |
+
[ZeRO-1 v2] zero1-v2 reduce-scatter/all-gather (symmetric/NVLS)
|
| 40 |
+
[ZeRO-1 v2] params=77388288 padded=77388288 W=4 shard=19347072 | 6 buckets of 50.000000 MB (3276800 elems/rank/bucket) | optimizer state/GPU: 221.409668 MB (vs 885.638672 MB unsharded)
|
| 41 |
+
[ZeRO-1] variant=v2 reduce-scatter/all-gather groups=3 [muon:bfloat16/905M] [adamw:bfloat16/77266944] [adamw:float32/77388288]
|
| 42 |
+
[ZeRO-1] DataParallel is DISABLED (both would double-reduce)
|
| 43 |
+
[precision] bf16 params=217, all with fp32 master weights
|
| 44 |
+
found 3 shards for split train
|
| 45 |
+
found 1 shards for split val
|
| 46 |
+
Saving logs to: CP_BluScreipt_Training_logs/CP_Training_log14.csv (run 14)
|
| 47 |
+
[rope_theta] 500000 compiled default (no sidecar -- legacy run, assuming it was trained at this base)
|
| 48 |
+
Checkpointing: ON -> checkpoints_bluscriptcp/blumodelcp_run14_step_<N>.ckpt (every 100 steps, keep 2)
|
| 49 |
+
[re-warmup] anchored at step 0 for 100 steps -> peak 0.000300 (0.300000 x max_lr), then cosine-decay to 0.000100 | resuming at step 0 (mid-ramp)
|
| 50 |
+
|
| 51 |
+
Starting training...
|
| 52 |
+
|
| 53 |
+
[CP ring P2P] OVERLAP: dedicated non-blocking stream
|
| 54 |
+
Current shard: /root/Adhi/BluTrain/dist/Context_Parallelism/in2_data/arm_c/in2c_val_00000.bin
|
| 55 |
+
100000000
|
| 56 |
+
validation loss: 1.8887
|
| 57 |
+
[CP bwd] overlap=ON
|
| 58 |
+
step 0 | loss: 1.920389 | lr 3.0000e-06 | norm: 1.1394 | dt: 16001.44ms | tok/sec: 32765.1 | mem: 107213MB | Time Left: 01 hrs : 46 mins
|
| 59 |
+
[TIMING] data: 2.9ms | fwd: 4222.9ms | loss: 6.4ms | bwd: 4663.3ms | clip: 0.0ms | optim: 62.6ms
|
| 60 |
+
step 1 | loss: 1.898834 | lr 6.0000e-06 | norm: 1.1567 | dt: 8280.12ms | tok/sec: 63318.9 | mem: 108909MB | Time Left: 00 hrs : 55 mins
|
| 61 |
+
[TIMING] data: 4.0ms | fwd: 3667.5ms | loss: 6.4ms | bwd: 4544.2ms | clip: 0.0ms | optim: 49.8ms
|
| 62 |
+
step 2 | loss: 1.910765 | lr 9.0000e-06 | norm: 1.2803 | dt: 9915.14ms | tok/sec: 52877.5 | mem: 108909MB | Time Left: 01 hrs : 05 mins
|
| 63 |
+
[TIMING] data: 2.6ms | fwd: 5223.5ms | loss: 6.4ms | bwd: 4625.0ms | clip: 0.0ms | optim: 49.6ms
|
| 64 |
+
step 3 | loss: 1.936233 | lr 1.2000e-05 | norm: 1.2666 | dt: 6216.73ms | tok/sec: 84335.0 | mem: 108909MB | Time Left: 00 hrs : 41 mins
|
| 65 |
+
[TIMING] data: 2.6ms | fwd: 1991.8ms | loss: 6.4ms | bwd: 4158.1ms | clip: 0.0ms | optim: 49.9ms
|
| 66 |
+
step 4 | loss: 1.905092 | lr 1.5000e-05 | norm: 1.1602 | dt: 6238.63ms | tok/sec: 84039.0 | mem: 108909MB | Time Left: 00 hrs : 41 mins
|
| 67 |
+
[TIMING] data: 3.1ms | fwd: 1995.3ms | loss: 6.4ms | bwd: 4176.5ms | clip: 0.0ms | optim: 49.3ms
|
| 68 |
+
step 5 | loss: 1.966748 | lr 1.8000e-05 | norm: 1.1494 | dt: 6276.42ms | tok/sec: 83532.9 | mem: 108909MB | Time Left: 00 hrs : 41 mins
|
| 69 |
+
[TIMING] data: 3.1ms | fwd: 2005.5ms | loss: 6.4ms | bwd: 4204.3ms | clip: 0.0ms | optim: 49.3ms
|
| 70 |
+
step 6 | loss: 1.904285 | lr 2.1000e-05 | norm: 1.0437 | dt: 6345.53ms | tok/sec: 82623.2 | mem: 108909MB | Time Left: 00 hrs : 41 mins
|
| 71 |
+
[TIMING] data: 3.2ms | fwd: 2013.2ms | loss: 6.4ms | bwd: 4265.3ms | clip: 0.0ms | optim: 49.4ms
|
| 72 |
+
step 7 | loss: 1.854531 | lr 2.4000e-05 | norm: 1.0216 | dt: 6402.65ms | tok/sec: 81886.1 | mem: 108909MB | Time Left: 00 hrs : 41 mins
|
| 73 |
+
[TIMING] data: 3.1ms | fwd: 2033.1ms | loss: 6.4ms | bwd: 4302.6ms | clip: 0.0ms | optim: 49.6ms
|
| 74 |
+
step 8 | loss: 1.830776 | lr 2.7000e-05 | norm: 0.9876 | dt: 6419.80ms | tok/sec: 81667.3 | mem: 108909MB | Time Left: 00 hrs : 41 mins
|
| 75 |
+
[TIMING] data: 3.0ms | fwd: 2037.5ms | loss: 6.4ms | bwd: 4315.3ms | clip: 0.0ms | optim: 49.6ms
|
| 76 |
+
step 9 | loss: 1.852669 | lr 3.0000e-05 | norm: 0.8843 | dt: 6426.62ms | tok/sec: 81580.7 | mem: 108909MB | Time Left: 00 hrs : 41 mins
|
| 77 |
+
[TIMING] data: 3.1ms | fwd: 2048.3ms | loss: 6.4ms | bwd: 4311.6ms | clip: 0.0ms | optim: 49.4ms
|
| 78 |
+
step 10 | loss: 1.806110 | lr 3.3000e-05 | norm: 0.8296 | dt: 6450.76ms | tok/sec: 81275.3 | mem: 108909MB | Time Left: 00 hrs : 41 mins
|
| 79 |
+
[TIMING] data: 3.1ms | fwd: 2058.2ms | loss: 6.7ms | bwd: 4325.5ms | clip: 0.0ms | optim: 49.4ms
|
| 80 |
+
step 11 | loss: 1.792623 | lr 3.6000e-05 | norm: 1.0501 | dt: 6435.64ms | tok/sec: 81466.3 | mem: 108909MB | Time Left: 00 hrs : 41 mins
|
| 81 |
+
[TIMING] data: 3.1ms | fwd: 2058.2ms | loss: 6.6ms | bwd: 4310.1ms | clip: 0.0ms | optim: 49.8ms
|
| 82 |
+
step 12 | loss: 1.784030 | lr 3.9000e-05 | norm: 1.0424 | dt: 6422.32ms | tok/sec: 81635.3 | mem: 108909MB | Time Left: 00 hrs : 41 mins
|
| 83 |
+
[TIMING] data: 3.1ms | fwd: 2025.7ms | loss: 6.4ms | bwd: 4329.7ms | clip: 0.0ms | optim: 49.5ms
|
| 84 |
+
step 13 | loss: 1.786141 | lr 4.2000e-05 | norm: 1.3447 | dt: 6375.98ms | tok/sec: 82228.6 | mem: 108909MB | Time Left: 00 hrs : 41 mins
|
| 85 |
+
[TIMING] data: 3.2ms | fwd: 2025.6ms | loss: 6.4ms | bwd: 4283.1ms | clip: 0.0ms | optim: 49.7ms
|
| 86 |
+
step 14 | loss: 1.809938 | lr 4.5000e-05 | norm: 1.2942 | dt: 6358.03ms | tok/sec: 82460.8 | mem: 108909MB | Time Left: 00 hrs : 40 mins
|
| 87 |
+
[TIMING] data: 3.1ms | fwd: 2027.8ms | loss: 6.4ms | bwd: 4263.5ms | clip: 0.0ms | optim: 49.4ms
|
| 88 |
+
step 15 | loss: 1.793668 | lr 4.8000e-05 | norm: 1.2557 | dt: 6368.77ms | tok/sec: 82321.7 | mem: 108909MB | Time Left: 00 hrs : 40 mins
|
| 89 |
+
[TIMING] data: 3.2ms | fwd: 2042.4ms | loss: 6.4ms | bwd: 4259.3ms | clip: 0.0ms | optim: 49.7ms
|
| 90 |
+
step 16 | loss: 1.765961 | lr 5.1000e-05 | norm: 1.0888 | dt: 6343.39ms | tok/sec: 82651.1 | mem: 108909MB | Time Left: 00 hrs : 40 mins
|
| 91 |
+
[TIMING] data: 3.2ms | fwd: 2021.3ms | loss: 6.4ms | bwd: 4254.8ms | clip: 0.0ms | optim: 49.9ms
|
| 92 |
+
step 17 | loss: 1.712484 | lr 5.4000e-05 | norm: 0.7903 | dt: 6323.92ms | tok/sec: 82905.5 | mem: 108909MB | Time Left: 00 hrs : 40 mins
|
| 93 |
+
[TIMING] data: 3.2ms | fwd: 2022.9ms | loss: 6.4ms | bwd: 4234.1ms | clip: 0.0ms | optim: 49.5ms
|
| 94 |
+
step 18 | loss: 1.731035 | lr 5.7000e-05 | norm: 0.8584 | dt: 6329.27ms | tok/sec: 82835.4 | mem: 108909MB | Time Left: 00 hrs : 40 mins
|
| 95 |
+
[TIMING] data: 3.0ms | fwd: 2017.8ms | loss: 6.6ms | bwd: 4244.6ms | clip: 0.0ms | optim: 49.3ms
|
| 96 |
+
step 19 | loss: 1.743202 | lr 6.0000e-05 | norm: 0.7385 | dt: 6334.25ms | tok/sec: 82770.3 | mem: 108909MB | Time Left: 00 hrs : 40 mins
|
| 97 |
+
[TIMING] data: 3.2ms | fwd: 2031.9ms | loss: 6.4ms | bwd: 4235.3ms | clip: 0.0ms | optim: 49.7ms
|
| 98 |
+
step 20 | loss: 1.718590 | lr 6.3000e-05 | norm: 1.1694 | dt: 6330.50ms | tok/sec: 82819.3 | mem: 108909MB | Time Left: 00 hrs : 40 mins
|
| 99 |
+
[TIMING] data: 3.1ms | fwd: 2022.3ms | loss: 6.4ms | bwd: 4241.4ms | clip: 0.0ms | optim: 49.6ms
|
| 100 |
+
step 21 | loss: 1.767774 | lr 6.6000e-05 | norm: 0.9925 | dt: 6338.12ms | tok/sec: 82719.8 | mem: 108909MB | Time Left: 00 hrs : 40 mins
|
| 101 |
+
[TIMING] data: 3.1ms | fwd: 2018.0ms | loss: 6.4ms | bwd: 4253.2ms | clip: 0.0ms | optim: 49.6ms
|
| 102 |
+
step 22 | loss: 1.697396 | lr 6.9000e-05 | norm: 0.6909 | dt: 6326.69ms | tok/sec: 82869.2 | mem: 108909MB | Time Left: 00 hrs : 39 mins
|
| 103 |
+
[TIMING] data: 3.2ms | fwd: 2019.1ms | loss: 6.4ms | bwd: 4240.6ms | clip: 0.0ms | optim: 49.6ms
|
| 104 |
+
step 23 | loss: 1.707077 | lr 7.2000e-05 | norm: 0.7638 | dt: 6371.41ms | tok/sec: 82287.6 | mem: 108909MB | Time Left: 00 hrs : 40 mins
|
| 105 |
+
[TIMING] data: 3.2ms | fwd: 2050.0ms | loss: 6.4ms | bwd: 4254.6ms | clip: 0.0ms | optim: 49.3ms
|
| 106 |
+
step 24 | loss: 1.671977 | lr 7.5000e-05 | norm: 0.7776 | dt: 6375.66ms | tok/sec: 82232.8 | mem: 108909MB | Time Left: 00 hrs : 39 mins
|
| 107 |
+
[TIMING] data: 3.3ms | fwd: 2069.2ms | loss: 6.4ms | bwd: 4239.3ms | clip: 0.0ms | optim: 49.6ms
|
| 108 |
+
step 25 | loss: 1.739844 | lr 7.8000e-05 | norm: 0.8230 | dt: 6332.83ms | tok/sec: 82788.9 | mem: 108909MB | Time Left: 00 hrs : 39 mins
|
| 109 |
+
[TIMING] data: 3.2ms | fwd: 2011.0ms | loss: 6.4ms | bwd: 4254.9ms | clip: 0.0ms | optim: 49.6ms
|
| 110 |
+
step 26 | loss: 1.652054 | lr 8.1000e-05 | norm: 0.9077 | dt: 6340.05ms | tok/sec: 82694.6 | mem: 108909MB | Time Left: 00 hrs : 39 mins
|
| 111 |
+
[TIMING] data: 3.1ms | fwd: 2025.9ms | loss: 6.4ms | bwd: 4247.1ms | clip: 0.0ms | optim: 49.7ms
|
| 112 |
+
step 27 | loss: 1.654510 | lr 8.4000e-05 | norm: 0.8483 | dt: 6341.17ms | tok/sec: 82680.0 | mem: 108909MB | Time Left: 00 hrs : 39 mins
|
| 113 |
+
[TIMING] data: 3.1ms | fwd: 2023.2ms | loss: 6.4ms | bwd: 4250.9ms | clip: 0.0ms | optim: 49.7ms
|
| 114 |
+
step 28 | loss: 1.691144 | lr 8.7000e-05 | norm: 0.7625 | dt: 6357.39ms | tok/sec: 82469.1 | mem: 108909MB | Time Left: 00 hrs : 39 mins
|
| 115 |
+
[TIMING] data: 3.2ms | fwd: 2045.9ms | loss: 6.4ms | bwd: 4244.7ms | clip: 0.0ms | optim: 49.4ms
|
| 116 |
+
step 29 | loss: 1.663364 | lr 9.0000e-05 | norm: 0.9021 | dt: 6359.41ms | tok/sec: 82442.9 | mem: 108909MB | Time Left: 00 hrs : 39 mins
|
| 117 |
+
[TIMING] data: 3.1ms | fwd: 2024.0ms | loss: 6.4ms | bwd: 4268.5ms | clip: 0.0ms | optim: 49.4ms
|
| 118 |
+
step 30 | loss: 1.633048 | lr 9.3000e-05 | norm: 0.7084 | dt: 6332.02ms | tok/sec: 82799.4 | mem: 108909MB | Time Left: 00 hrs : 39 mins
|
| 119 |
+
[TIMING] data: 3.1ms | fwd: 2012.0ms | loss: 6.4ms | bwd: 4253.2ms | clip: 0.0ms | optim: 49.5ms
|
| 120 |
+
step 31 | loss: 1.700897 | lr 9.6000e-05 | norm: 0.5941 | dt: 6339.41ms | tok/sec: 82702.9 | mem: 108909MB | Time Left: 00 hrs : 38 mins
|
| 121 |
+
[TIMING] data: 3.1ms | fwd: 2024.3ms | loss: 6.4ms | bwd: 4248.3ms | clip: 0.0ms | optim: 49.5ms
|
| 122 |
+
step 32 | loss: 1.608716 | lr 9.9000e-05 | norm: 0.6328 | dt: 6395.62ms | tok/sec: 81976.1 | mem: 108909MB | Time Left: 00 hrs : 39 mins
|
| 123 |
+
[TIMING] data: 3.2ms | fwd: 2066.5ms | loss: 6.4ms | bwd: 4262.0ms | clip: 0.0ms | optim: 49.6ms
|
| 124 |
+
step 33 | loss: 1.657396 | lr 1.0200e-04 | norm: 0.6388 | dt: 6355.23ms | tok/sec: 82497.0 | mem: 108909MB | Time Left: 00 hrs : 38 mins
|
| 125 |
+
[TIMING] data: 3.1ms | fwd: 2024.5ms | loss: 6.4ms | bwd: 4263.7ms | clip: 0.0ms | optim: 49.7ms
|
| 126 |
+
step 34 | loss: 1.648264 | lr 1.0500e-04 | norm: 0.6340 | dt: 6350.56ms | tok/sec: 82557.8 | mem: 108909MB | Time Left: 00 hrs : 38 mins
|
| 127 |
+
[TIMING] data: 3.1ms | fwd: 2026.4ms | loss: 6.4ms | bwd: 4257.2ms | clip: 0.0ms | optim: 49.6ms
|
| 128 |
+
step 35 | loss: 1.660285 | lr 1.0800e-04 | norm: 0.7759 | dt: 6348.23ms | tok/sec: 82588.0 | mem: 108909MB | Time Left: 00 hrs : 38 mins
|
| 129 |
+
[TIMING] data: 3.1ms | fwd: 2080.1ms | loss: 6.4ms | bwd: 4201.3ms | clip: 0.0ms | optim: 49.5ms
|
| 130 |
+
step 36 | loss: 1.631099 | lr 1.1100e-04 | norm: 0.5843 | dt: 6381.27ms | tok/sec: 82160.5 | mem: 108909MB | Time Left: 00 hrs : 38 mins
|
| 131 |
+
[TIMING] data: 3.1ms | fwd: 2063.8ms | loss: 6.4ms | bwd: 4250.2ms | clip: 0.0ms | optim: 50.0ms
|
| 132 |
+
step 37 | loss: 1.626856 | lr 1.1400e-04 | norm: 0.6801 | dt: 6330.01ms | tok/sec: 82825.7 | mem: 108909MB | Time Left: 00 hrs : 38 mins
|
| 133 |
+
[TIMING] data: 3.2ms | fwd: 2013.3ms | loss: 6.4ms | bwd: 4249.9ms | clip: 0.0ms | optim: 49.4ms
|
| 134 |
+
step 38 | loss: 1.619444 | lr 1.1700e-04 | norm: 0.6525 | dt: 6337.69ms | tok/sec: 82725.4 | mem: 108909MB | Time Left: 00 hrs : 38 mins
|
| 135 |
+
[TIMING] data: 3.2ms | fwd: 2020.1ms | loss: 6.4ms | bwd: 4250.6ms | clip: 0.0ms | optim: 49.7ms
|
| 136 |
+
step 39 | loss: 1.642475 | lr 1.2000e-04 | norm: 0.7724 | dt: 6377.82ms | tok/sec: 82204.8 | mem: 108909MB | Time Left: 00 hrs : 38 mins
|
| 137 |
+
[TIMING] data: 3.2ms | fwd: 2075.6ms | loss: 6.4ms | bwd: 4235.5ms | clip: 0.0ms | optim: 49.3ms
|
| 138 |
+
step 40 | loss: 1.600698 | lr 1.2300e-04 | norm: 0.8544 | dt: 6376.74ms | tok/sec: 82218.8 | mem: 108909MB | Time Left: 00 hrs : 38 mins
|
| 139 |
+
[TIMING] data: 3.2ms | fwd: 2071.6ms | loss: 6.4ms | bwd: 4238.0ms | clip: 0.0ms | optim: 49.7ms
|
| 140 |
+
step 41 | loss: 1.580887 | lr 1.2600e-04 | norm: 0.9816 | dt: 6324.28ms | tok/sec: 82900.8 | mem: 108909MB | Time Left: 00 hrs : 37 mins
|
| 141 |
+
[TIMING] data: 3.1ms | fwd: 2022.0ms | loss: 6.4ms | bwd: 4235.4ms | clip: 0.0ms | optim: 49.6ms
|
| 142 |
+
step 42 | loss: 1.624956 | lr 1.2900e-04 | norm: 0.8724 | dt: 6332.33ms | tok/sec: 82795.5 | mem: 108909MB | Time Left: 00 hrs : 37 mins
|
| 143 |
+
[TIMING] data: 3.1ms | fwd: 2018.1ms | loss: 6.7ms | bwd: 4246.9ms | clip: 0.0ms | optim: 49.7ms
|
| 144 |
+
step 43 | loss: 1.623077 | lr 1.3200e-04 | norm: 0.6374 | dt: 6346.73ms | tok/sec: 82607.6 | mem: 108909MB | Time Left: 00 hrs : 37 mins
|
| 145 |
+
[TIMING] data: 3.3ms | fwd: 2024.3ms | loss: 6.4ms | bwd: 4255.4ms | clip: 0.0ms | optim: 49.5ms
|
| 146 |
+
step 44 | loss: 1.554411 | lr 1.3500e-04 | norm: 0.8312 | dt: 6327.95ms | tok/sec: 82852.7 | mem: 108909MB | Time Left: 00 hrs : 37 mins
|
| 147 |
+
[TIMING] data: 3.3ms | fwd: 2012.0ms | loss: 6.4ms | bwd: 4248.7ms | clip: 0.0ms | optim: 49.7ms
|
| 148 |
+
step 45 | loss: 1.633371 | lr 1.3800e-04 | norm: 0.7739 | dt: 6346.30ms | tok/sec: 82613.2 | mem: 108909MB | Time Left: 00 hrs : 37 mins
|
| 149 |
+
[TIMING] data: 3.1ms | fwd: 2037.5ms | loss: 6.4ms | bwd: 4241.6ms | clip: 0.0ms | optim: 49.8ms
|
| 150 |
+
step 46 | loss: 1.607605 | lr 1.4100e-04 | norm: 0.7427 | dt: 6313.64ms | tok/sec: 83040.5 | mem: 108909MB | Time Left: 00 hrs : 37 mins
|
| 151 |
+
[TIMING] data: 3.1ms | fwd: 2019.0ms | loss: 6.4ms | bwd: 4227.8ms | clip: 0.0ms | optim: 49.5ms
|
| 152 |
+
step 47 | loss: 1.635939 | lr 1.4400e-04 | norm: 0.6916 | dt: 6347.58ms | tok/sec: 82596.6 | mem: 108909MB | Time Left: 00 hrs : 37 mins
|
| 153 |
+
[TIMING] data: 3.1ms | fwd: 2024.1ms | loss: 6.4ms | bwd: 4256.4ms | clip: 0.0ms | optim: 49.7ms
|
| 154 |
+
step 48 | loss: 1.633764 | lr 1.4700e-04 | norm: 0.8044 | dt: 6336.69ms | tok/sec: 82738.5 | mem: 108909MB | Time Left: 00 hrs : 37 mins
|
| 155 |
+
[TIMING] data: 3.3ms | fwd: 2021.2ms | loss: 6.4ms | bwd: 4248.1ms | clip: 0.0ms | optim: 49.6ms
|
| 156 |
+
step 49 | loss: 1.607873 | lr 1.5000e-04 | norm: 0.6211 | dt: 6305.45ms | tok/sec: 83148.4 | mem: 108909MB | Time Left: 00 hrs : 36 mins
|
| 157 |
+
[TIMING] data: 3.2ms | fwd: 2019.0ms | loss: 6.4ms | bwd: 4219.5ms | clip: 0.0ms | optim: 49.6ms
|
| 158 |
+
step 50 | loss: 1.639636 | lr 1.5300e-04 | norm: 0.7115 | dt: 6370.11ms | tok/sec: 82304.3 | mem: 108909MB | Time Left: 00 hrs : 37 mins
|
| 159 |
+
[TIMING] data: 3.2ms | fwd: 2078.6ms | loss: 6.4ms | bwd: 4224.5ms | clip: 0.0ms | optim: 49.5ms
|
| 160 |
+
step 51 | loss: 1.656035 | lr 1.5600e-04 | norm: 0.7963 | dt: 6345.47ms | tok/sec: 82624.0 | mem: 108909MB | Time Left: 00 hrs : 36 mins
|
| 161 |
+
[TIMING] data: 3.2ms | fwd: 2031.6ms | loss: 6.6ms | bwd: 4246.8ms | clip: 0.0ms | optim: 49.5ms
|
| 162 |
+
step 52 | loss: 1.662412 | lr 1.5900e-04 | norm: 0.7057 | dt: 6316.86ms | tok/sec: 82998.2 | mem: 108909MB | Time Left: 00 hrs : 36 mins
|
| 163 |
+
[TIMING] data: 3.1ms | fwd: 2016.9ms | loss: 6.4ms | bwd: 4232.1ms | clip: 0.0ms | optim: 49.5ms
|
| 164 |
+
step 53 | loss: 1.581693 | lr 1.6200e-04 | norm: 0.6668 | dt: 6342.80ms | tok/sec: 82658.8 | mem: 108909MB | Time Left: 00 hrs : 36 mins
|
| 165 |
+
[TIMING] data: 3.2ms | fwd: 2028.6ms | loss: 6.4ms | bwd: 4247.1ms | clip: 0.0ms | optim: 49.7ms
|
| 166 |
+
step 54 | loss: 1.594439 | lr 1.6500e-04 | norm: 0.8471 | dt: 6319.42ms | tok/sec: 82964.6 | mem: 108909MB | Time Left: 00 hrs : 36 mins
|
| 167 |
+
[TIMING] data: 3.2ms | fwd: 2024.6ms | loss: 6.4ms | bwd: 4227.7ms | clip: 0.0ms | optim: 49.6ms
|
| 168 |
+
step 55 | loss: 1.670181 | lr 1.6800e-04 | norm: 0.8091 | dt: 6371.77ms | tok/sec: 82282.9 | mem: 108909MB | Time Left: 00 hrs : 36 mins
|
| 169 |
+
[TIMING] data: 3.2ms | fwd: 2057.9ms | loss: 6.4ms | bwd: 4247.0ms | clip: 0.0ms | optim: 49.4ms
|
| 170 |
+
step 56 | loss: 1.632861 | lr 1.7100e-04 | norm: 0.7737 | dt: 6320.74ms | tok/sec: 82947.3 | mem: 108909MB | Time Left: 00 hrs : 36 mins
|
| 171 |
+
[TIMING] data: 3.2ms | fwd: 2018.6ms | loss: 6.4ms | bwd: 4235.3ms | clip: 0.0ms | optim: 49.5ms
|
| 172 |
+
step 57 | loss: 1.572145 | lr 1.7400e-04 | norm: 0.8922 | dt: 6340.90ms | tok/sec: 82683.6 | mem: 108909MB | Time Left: 00 hrs : 36 mins
|
| 173 |
+
[TIMING] data: 3.2ms | fwd: 2020.7ms | loss: 6.4ms | bwd: 4253.2ms | clip: 0.0ms | optim: 49.5ms
|
| 174 |
+
step 58 | loss: 1.589554 | lr 1.7700e-04 | norm: 0.6774 | dt: 6324.35ms | tok/sec: 82899.9 | mem: 108909MB | Time Left: 00 hrs : 36 mins
|
| 175 |
+
[TIMING] data: 3.1ms | fwd: 2026.5ms | loss: 6.4ms | bwd: 4230.7ms | clip: 0.0ms | optim: 49.8ms
|
| 176 |
+
step 59 | loss: 1.586414 | lr 1.8000e-04 | norm: 0.7578 | dt: 6376.86ms | tok/sec: 82217.2 | mem: 108909MB | Time Left: 00 hrs : 36 mins
|
| 177 |
+
[TIMING] data: 3.2ms | fwd: 2054.2ms | loss: 6.5ms | bwd: 4255.8ms | clip: 0.0ms | optim: 49.3ms
|
| 178 |
+
step 60 | loss: 1.630998 | lr 1.8300e-04 | norm: 0.7535 | dt: 6323.77ms | tok/sec: 82907.5 | mem: 108909MB | Time Left: 00 hrs : 35 mins
|
| 179 |
+
[TIMING] data: 3.2ms | fwd: 2018.2ms | loss: 6.4ms | bwd: 4238.6ms | clip: 0.0ms | optim: 49.5ms
|
| 180 |
+
step 61 | loss: 1.556511 | lr 1.8600e-04 | norm: 0.8945 | dt: 6345.84ms | tok/sec: 82619.2 | mem: 108909MB | Time Left: 00 hrs : 35 mins
|
| 181 |
+
[TIMING] data: 3.2ms | fwd: 2025.4ms | loss: 6.4ms | bwd: 4252.0ms | clip: 0.0ms | optim: 50.9ms
|
| 182 |
+
step 62 | loss: 1.594290 | lr 1.8900e-04 | norm: 0.8499 | dt: 6348.36ms | tok/sec: 82586.3 | mem: 108909MB | Time Left: 00 hrs : 35 mins
|
| 183 |
+
[TIMING] data: 3.2ms | fwd: 2026.3ms | loss: 6.4ms | bwd: 4255.1ms | clip: 0.0ms | optim: 49.6ms
|
| 184 |
+
step 63 | loss: 1.591010 | lr 1.9200e-04 | norm: 0.7212 | dt: 6382.64ms | tok/sec: 82142.8 | mem: 108909MB | Time Left: 00 hrs : 35 mins
|
| 185 |
+
[TIMING] data: 3.1ms | fwd: 2080.1ms | loss: 6.4ms | bwd: 4235.5ms | clip: 0.0ms | optim: 49.6ms
|
| 186 |
+
step 64 | loss: 1.558378 | lr 1.9500e-04 | norm: 0.8342 | dt: 6370.17ms | tok/sec: 82303.6 | mem: 108909MB | Time Left: 00 hrs : 35 mins
|
| 187 |
+
[TIMING] data: 3.3ms | fwd: 2050.7ms | loss: 6.6ms | bwd: 4252.4ms | clip: 0.0ms | optim: 49.4ms
|
| 188 |
+
step 65 | loss: 1.605214 | lr 1.9800e-04 | norm: 0.6799 | dt: 6333.11ms | tok/sec: 82785.2 | mem: 108909MB | Time Left: 00 hrs : 35 mins
|
| 189 |
+
[TIMING] data: 3.1ms | fwd: 2021.7ms | loss: 6.4ms | bwd: 4244.5ms | clip: 0.0ms | optim: 49.5ms
|
| 190 |
+
step 66 | loss: 1.590822 | lr 2.0100e-04 | norm: 0.9325 | dt: 6346.55ms | tok/sec: 82609.9 | mem: 108909MB | Time Left: 00 hrs : 35 mins
|
| 191 |
+
[TIMING] data: 3.2ms | fwd: 2013.1ms | loss: 6.4ms | bwd: 4266.6ms | clip: 0.0ms | optim: 49.6ms
|
| 192 |
+
step 67 | loss: 1.622504 | lr 2.0400e-04 | norm: 0.8867 | dt: 6339.60ms | tok/sec: 82700.5 | mem: 108909MB | Time Left: 00 hrs : 35 mins
|
| 193 |
+
[TIMING] data: 3.2ms | fwd: 2021.6ms | loss: 6.4ms | bwd: 4250.8ms | clip: 0.0ms | optim: 49.8ms
|
| 194 |
+
step 68 | loss: 1.598213 | lr 2.0700e-04 | norm: 0.9454 | dt: 6362.03ms | tok/sec: 82408.9 | mem: 108909MB | Time Left: 00 hrs : 35 mins
|
| 195 |
+
[TIMING] data: 3.0ms | fwd: 2052.2ms | loss: 6.4ms | bwd: 4242.9ms | clip: 0.0ms | optim: 49.7ms
|
| 196 |
+
step 69 | loss: 1.635720 | lr 2.1000e-04 | norm: 0.8285 | dt: 6375.84ms | tok/sec: 82230.5 | mem: 108909MB | Time Left: 00 hrs : 35 mins
|
| 197 |
+
[TIMING] data: 3.1ms | fwd: 2071.8ms | loss: 6.4ms | bwd: 4237.3ms | clip: 0.0ms | optim: 49.5ms
|
| 198 |
+
step 70 | loss: 1.608484 | lr 2.1300e-04 | norm: 0.7056 | dt: 6333.45ms | tok/sec: 82780.8 | mem: 108909MB | Time Left: 00 hrs : 34 mins
|
| 199 |
+
[TIMING] data: 3.1ms | fwd: 2030.7ms | loss: 6.4ms | bwd: 4235.7ms | clip: 0.0ms | optim: 49.8ms
|
| 200 |
+
step 71 | loss: 1.609575 | lr 2.1600e-04 | norm: 0.7693 | dt: 6333.45ms | tok/sec: 82780.8 | mem: 108909MB | Time Left: 00 hrs : 34 mins
|
| 201 |
+
[TIMING] data: 3.2ms | fwd: 2022.2ms | loss: 6.4ms | bwd: 4244.5ms | clip: 0.0ms | optim: 49.4ms
|
| 202 |
+
step 72 | loss: 1.544902 | lr 2.1900e-04 | norm: 0.7656 | dt: 6330.44ms | tok/sec: 82820.1 | mem: 108909MB | Time Left: 00 hrs : 34 mins
|
| 203 |
+
[TIMING] data: 3.0ms | fwd: 2021.9ms | loss: 6.4ms | bwd: 4241.3ms | clip: 0.0ms | optim: 50.0ms
|
| 204 |
+
step 73 | loss: 1.616233 | lr 2.2200e-04 | norm: 0.9089 | dt: 6347.53ms | tok/sec: 82597.2 | mem: 108909MB | Time Left: 00 hrs : 34 mins
|
| 205 |
+
[TIMING] data: 3.1ms | fwd: 2046.7ms | loss: 6.4ms | bwd: 4234.0ms | clip: 0.0ms | optim: 49.5ms
|
| 206 |
+
step 74 | loss: 1.594353 | lr 2.2500e-04 | norm: 0.8706 | dt: 6332.66ms | tok/sec: 82791.1 | mem: 108909MB | Time Left: 00 hrs : 34 mins
|
| 207 |
+
[TIMING] data: 3.0ms | fwd: 2023.6ms | loss: 6.4ms | bwd: 4242.1ms | clip: 0.0ms | optim: 49.7ms
|
| 208 |
+
step 75 | loss: 1.609479 | lr 2.2800e-04 | norm: 0.9014 | dt: 6339.55ms | tok/sec: 82701.2 | mem: 108909MB | Time Left: 00 hrs : 34 mins
|
| 209 |
+
[TIMING] data: 3.1ms | fwd: 2023.0ms | loss: 6.6ms | bwd: 4249.3ms | clip: 0.0ms | optim: 49.7ms
|
| 210 |
+
step 76 | loss: 1.552467 | lr 2.3100e-04 | norm: 0.7200 | dt: 6335.94ms | tok/sec: 82748.2 | mem: 108909MB | Time Left: 00 hrs : 34 mins
|
| 211 |
+
[TIMING] data: 3.1ms | fwd: 2018.6ms | loss: 6.4ms | bwd: 4250.5ms | clip: 0.0ms | optim: 49.5ms
|
| 212 |
+
step 77 | loss: 1.623230 | lr 2.3400e-04 | norm: 0.8687 | dt: 6327.53ms | tok/sec: 82858.2 | mem: 108909MB | Time Left: 00 hrs : 34 mins
|
| 213 |
+
[TIMING] data: 3.2ms | fwd: 2033.3ms | loss: 6.4ms | bwd: 4227.3ms | clip: 0.0ms | optim: 49.6ms
|
| 214 |
+
step 78 | loss: 1.589706 | lr 2.3700e-04 | norm: 0.7531 | dt: 6334.82ms | tok/sec: 82762.9 | mem: 108909MB | Time Left: 00 hrs : 33 mins
|
| 215 |
+
[TIMING] data: 3.1ms | fwd: 2022.4ms | loss: 6.4ms | bwd: 4245.0ms | clip: 0.0ms | optim: 50.0ms
|
| 216 |
+
step 79 | loss: 1.593231 | lr 2.4000e-04 | norm: 0.9309 | dt: 6320.51ms | tok/sec: 82950.2 | mem: 108909MB | Time Left: 00 hrs : 33 mins
|
| 217 |
+
[TIMING] data: 3.2ms | fwd: 2013.5ms | loss: 6.4ms | bwd: 4239.9ms | clip: 0.0ms | optim: 49.6ms
|
| 218 |
+
step 80 | loss: 1.596107 | lr 2.4300e-04 | norm: 1.6061 | dt: 6329.07ms | tok/sec: 82838.0 | mem: 108909MB | Time Left: 00 hrs : 33 mins
|
| 219 |
+
[TIMING] data: 3.1ms | fwd: 2014.9ms | loss: 6.4ms | bwd: 4247.6ms | clip: 0.0ms | optim: 49.2ms
|
| 220 |
+
step 81 | loss: 1.639003 | lr 2.4600e-04 | norm: 0.8177 | dt: 6320.26ms | tok/sec: 82953.6 | mem: 108909MB | Time Left: 00 hrs : 33 mins
|
| 221 |
+
[TIMING] data: 3.1ms | fwd: 2019.2ms | loss: 6.4ms | bwd: 4234.4ms | clip: 0.0ms | optim: 49.4ms
|
| 222 |
+
step 82 | loss: 1.667713 | lr 2.4900e-04 | norm: 0.8361 | dt: 6397.35ms | tok/sec: 81954.0 | mem: 108909MB | Time Left: 00 hrs : 33 mins
|
| 223 |
+
[TIMING] data: 3.1ms | fwd: 2098.5ms | loss: 6.4ms | bwd: 4232.2ms | clip: 0.0ms | optim: 49.3ms
|
| 224 |
+
step 83 | loss: 1.597706 | lr 2.5200e-04 | norm: 0.8508 | dt: 6376.57ms | tok/sec: 82221.0 | mem: 108909MB | Time Left: 00 hrs : 33 mins
|
| 225 |
+
[TIMING] data: 3.1ms | fwd: 2045.3ms | loss: 6.4ms | bwd: 4264.5ms | clip: 0.0ms | optim: 49.5ms
|
| 226 |
+
step 84 | loss: 1.606992 | lr 2.5500e-04 | norm: 0.8824 | dt: 6307.10ms | tok/sec: 83126.6 | mem: 108909MB | Time Left: 00 hrs : 33 mins
|
| 227 |
+
[TIMING] data: 3.2ms | fwd: 2014.7ms | loss: 6.4ms | bwd: 4225.3ms | clip: 0.0ms | optim: 49.5ms
|
| 228 |
+
step 85 | loss: 1.606081 | lr 2.5800e-04 | norm: 0.8665 | dt: 6330.49ms | tok/sec: 82819.5 | mem: 108909MB | Time Left: 00 hrs : 33 mins
|
| 229 |
+
[TIMING] data: 3.2ms | fwd: 2021.7ms | loss: 6.4ms | bwd: 4241.2ms | clip: 0.0ms | optim: 50.0ms
|
| 230 |
+
step 86 | loss: 1.608906 | lr 2.6100e-04 | norm: 0.9115 | dt: 6355.76ms | tok/sec: 82490.2 | mem: 108909MB | Time Left: 00 hrs : 33 mins
|
| 231 |
+
[TIMING] data: 3.1ms | fwd: 2027.0ms | loss: 6.4ms | bwd: 4261.7ms | clip: 0.0ms | optim: 49.7ms
|
| 232 |
+
step 87 | loss: 1.671125 | lr 2.6400e-04 | norm: 0.8144 | dt: 6334.82ms | tok/sec: 82762.8 | mem: 108909MB | Time Left: 00 hrs : 33 mins
|
| 233 |
+
[TIMING] data: 3.1ms | fwd: 2028.3ms | loss: 6.4ms | bwd: 4239.7ms | clip: 0.0ms | optim: 49.4ms
|
| 234 |
+
step 88 | loss: 1.673897 | lr 2.6700e-04 | norm: 0.8308 | dt: 6355.62ms | tok/sec: 82492.0 | mem: 108909MB | Time Left: 00 hrs : 33 mins
|
| 235 |
+
[TIMING] data: 3.1ms | fwd: 2057.7ms | loss: 6.4ms | bwd: 4230.9ms | clip: 0.0ms | optim: 49.7ms
|
| 236 |
+
step 89 | loss: 1.632511 | lr 2.7000e-04 | norm: 0.7274 | dt: 6347.37ms | tok/sec: 82599.2 | mem: 108909MB | Time Left: 00 hrs : 32 mins
|
| 237 |
+
[TIMING] data: 3.3ms | fwd: 2019.5ms | loss: 6.4ms | bwd: 4260.6ms | clip: 0.0ms | optim: 49.8ms
|
| 238 |
+
step 90 | loss: 1.688837 | lr 2.7300e-04 | norm: 0.8296 | dt: 6319.86ms | tok/sec: 82958.8 | mem: 108909MB | Time Left: 00 hrs : 32 mins
|
| 239 |
+
[TIMING] data: 3.2ms | fwd: 2016.4ms | loss: 6.4ms | bwd: 4236.4ms | clip: 0.0ms | optim: 49.5ms
|
| 240 |
+
step 91 | loss: 1.636074 | lr 2.7600e-04 | norm: 0.8177 | dt: 6328.08ms | tok/sec: 82851.0 | mem: 108909MB | Time Left: 00 hrs : 32 mins
|
| 241 |
+
[TIMING] data: 3.4ms | fwd: 2024.8ms | loss: 6.5ms | bwd: 4236.1ms | clip: 0.0ms | optim: 49.4ms
|
| 242 |
+
step 92 | loss: 1.643463 | lr 2.7900e-04 | norm: 0.8352 | dt: 6385.58ms | tok/sec: 82105.0 | mem: 108909MB | Time Left: 00 hrs : 32 mins
|
| 243 |
+
[TIMING] data: 3.1ms | fwd: 2085.3ms | loss: 6.4ms | bwd: 4233.4ms | clip: 0.0ms | optim: 49.5ms
|
| 244 |
+
step 93 | loss: 1.674679 | lr 2.8200e-04 | norm: 0.8445 | dt: 6345.84ms | tok/sec: 82619.2 | mem: 108909MB | Time Left: 00 hrs : 32 mins
|
| 245 |
+
[TIMING] data: 3.2ms | fwd: 2050.6ms | loss: 6.4ms | bwd: 4228.3ms | clip: 0.0ms | optim: 49.5ms
|
| 246 |
+
step 94 | loss: 1.552707 | lr 2.8500e-04 | norm: 0.8634 | dt: 6333.64ms | tok/sec: 82778.4 | mem: 108909MB | Time Left: 00 hrs : 32 mins
|
| 247 |
+
[TIMING] data: 3.1ms | fwd: 2024.6ms | loss: 6.4ms | bwd: 4242.3ms | clip: 0.0ms | optim: 49.4ms
|
| 248 |
+
step 95 | loss: 1.630905 | lr 2.8800e-04 | norm: 0.8320 | dt: 6353.83ms | tok/sec: 82515.3 | mem: 108909MB | Time Left: 00 hrs : 32 mins
|
| 249 |
+
[TIMING] data: 3.3ms | fwd: 2028.2ms | loss: 6.4ms | bwd: 4258.5ms | clip: 0.0ms | optim: 49.6ms
|
| 250 |
+
step 96 | loss: 1.621827 | lr 2.9100e-04 | norm: 0.9184 | dt: 6323.02ms | tok/sec: 82917.4 | mem: 108909MB | Time Left: 00 hrs : 32 mins
|
| 251 |
+
[TIMING] data: 3.2ms | fwd: 2012.9ms | loss: 6.5ms | bwd: 4242.1ms | clip: 0.0ms | optim: 50.3ms
|
| 252 |
+
step 97 | loss: 1.617399 | lr 2.9400e-04 | norm: 0.7699 | dt: 6362.22ms | tok/sec: 82406.5 | mem: 108909MB | Time Left: 00 hrs : 32 mins
|
| 253 |
+
[TIMING] data: 3.1ms | fwd: 2055.8ms | loss: 6.4ms | bwd: 4239.4ms | clip: 0.0ms | optim: 49.7ms
|
| 254 |
+
step 98 | loss: 1.612508 | lr 2.9700e-04 | norm: 0.7359 | dt: 6338.88ms | tok/sec: 82709.9 | mem: 108909MB | Time Left: 00 hrs : 31 mins
|
| 255 |
+
[TIMING] data: 3.3ms | fwd: 2024.5ms | loss: 6.4ms | bwd: 4246.3ms | clip: 0.0ms | optim: 50.0ms
|
| 256 |
+
step 99 | loss: 1.637773 | lr 3.0000e-04 | norm: 0.8468 | dt: 6349.17ms | tok/sec: 82575.8 | mem: 108909MB | Time Left: 00 hrs : 31 mins
|
| 257 |
+
[TIMING] data: 3.3ms | fwd: 2017.7ms | loss: 6.5ms | bwd: 4264.2ms | clip: 0.0ms | optim: 49.6ms
|
| 258 |
+
[WARN] checkpoint save at step 100 failed: Failed to flush checkpoint to: checkpoints_bluscriptcp/rank_2/blumodelcp_run14_step_100.tmp
|
| 259 |
+
[WARN] checkpoint save at step 100 failed: Failed to flush checkpoint to: checkpoints_bluscriptcp/rank_3/blumodelcp_run14_step_100.tmp
|
| 260 |
+
[WARN] checkpoint save at step 100 failed: Failed to flush checkpoint to: checkpoints_bluscriptcp/rank_1/blumodelcp_run14_step_100.tmp
|
| 261 |
+
[WARN] checkpoint save at step 100 failed: Failed to flush checkpoint to: checkpoints_bluscriptcp/rank_0/blumodelcp_run14_step_100.tmp
|
| 262 |
+
step 100 | loss: 1.653108 | lr 3.0000e-04 | norm: 0.7348 | dt: 6202.84ms | tok/sec: 84523.8 | mem: 108909MB | Time Left: 00 hrs : 31 mins
|
| 263 |
+
[TIMING] data: 3.2ms | step 173 | loss: 1.634869 | lr 2.7218e-04 | norm: 0.4948 | dt: 6352.32ms | tok/sec: 82534.8 | mem: 108909MB | Time Left: 00 hrs : 24 mins
|
| 264 |
+
[TIMING] data: 3.4ms | fwd: 2025.4ms | loss: 6.5ms | bwd: 4259.7ms | clip: 0.0ms | optim: 49.4ms
|
| 265 |
+
step 174 | loss: 1.655519 | lr 2.7145e-04 | norm: 0.5231 | dt: 6317.55ms | tok/sec: 82989.1 | mem: 108909MB | Time Left: 00 hrs : 23 mins
|
| 266 |
+
[TIMING] data: 3.5ms | fwd: 2024.2ms | loss: 6.5ms | bwd: 4225.9ms | clip: 0.0ms | optim: 49.5ms
|
| 267 |
+
step 175 | loss: 1.636195 | lr 2.7071e-04 | norm: 0.4414 | dt: 6314.49ms | tok/sec: 83029.3 | mem: 108909MB | Time Left: 00 hrs : 23 mins
|
| 268 |
+
[TIMING] data: 3.5ms | fwd: 2019.0ms | loss: 6.5ms | bwd: 4228.3ms | clip: 0.0ms | optim: 49.3ms
|
| 269 |
+
step 176 | loss: 1.596313 | lr 2.6997e-04 | norm: 0.4120 | dt: 6341.42ms | tok/sec: 82676.8 | mem: 108909MB | Time Left: 00 hrs : 23 mins
|
| 270 |
+
[TIMING] data: 3.6ms | fwd: 2056.1ms | loss: 6.5ms | bwd: 4217.3ms | clip: 0.0ms | optim: 50.0ms
|
| 271 |
+
step 177 | loss: 1.690520 | lr 2.6921e-04 | norm: 0.5356 | dt: 6335.65ms | tok/sec: 82752.1 | mem: 108909MB | Time Left: 00 hrs : 23 mins
|
| 272 |
+
[TIMING] data: 3.6ms | fwd: 2028.3ms | loss: 6.5ms | bwd: 4239.8ms | clip: 0.0ms | optim: 49.5ms
|
| 273 |
+
step 178 | loss: 1.654294 | lr 2.6845e-04 | norm: 0.5014 | dt: 6333.05ms | tok/sec: 82786.0 | mem: 108909MB | Time Left: 00 hrs : 23 mins
|
| 274 |
+
[TIMING] data: 3.7ms | fwd: 2020.2ms | loss: 6.5ms | bwd: 4245.2ms | clip: 0.0ms | optim: 49.5ms
|
| 275 |
+
step 179 | loss: 1.655236 | lr 2.6769e-04 | norm: 0.5110 | dt: 6326.79ms | tok/sec: 82868.0 | mem: 108909MB | Time Left: 00 hrs : 23 mins
|
| 276 |
+
[TIMING] data: 3.4ms | fwd: 2023.6ms | loss: 6.5ms | bwd: 4235.8ms | clip: 0.0ms | optim: 49.5ms
|
| 277 |
+
step 180 | loss: 1.699248 | lr 2.6691e-04 | norm: 0.4708 | dt: 6315.29ms | tok/sec: 83018.9 | mem: 108909MB | Time Left: 00 hrs : 23 mins
|
| 278 |
+
[TIMING] data: 3.4ms | fwd: 2018.1ms | loss: 6.5ms | bwd: 4229.6ms | clip: 0.0ms | optim: 49.7ms
|
| 279 |
+
step 181 | loss: 1.610737 | lr 2.6613e-04 | norm: 0.5130 | dt: 6332.34ms | tok/sec: 82795.2 | mem: 108909MB | Time Left: 00 hrs : 23 mins
|
| 280 |
+
[TIMING] data: 3.4ms | fwd: 2021.4ms | loss: 6.5ms | bwd: 4243.0ms | clip: 0.0ms | optim: 50.1ms
|
| 281 |
+
step 182 | loss: 1.598042 | lr 2.6534e-04 | norm: 0.3652 | dt: 6330.79ms | tok/sec: 82815.6 | mem: 108909MB | Time Left: 00 hrs : 23 mins
|
| 282 |
+
[TIMING] data: 3.6ms | fwd: 2023.0ms | loss: 6.5ms | bwd: 4240.2ms | clip: 0.0ms | optim: 49.5ms
|
| 283 |
+
step 183 | loss: 1.636767 | lr 2.6455e-04 | norm: 0.3973 | dt: 6319.70ms | tok/sec: 82961.0 | mem: 108909MB | Time Left: 00 hrs : 22 mins
|
| 284 |
+
[TIMING] data: 3.5ms | fwd: 2023.7ms | loss: 6.5ms | bwd: 4228.4ms | clip: 0.0ms | optim: 49.6ms
|
| 285 |
+
step 184 | loss: 1.634775 | lr 2.6374e-04 | norm: 0.4649 | dt: 6357.98ms | tok/sec: 82461.5 | mem: 108909MB | Time Left: 00 hrs : 22 mins
|
| 286 |
+
[TIMING] data: 3.6ms | fwd: 2051.7ms | loss: 6.5ms | bwd: 4238.5ms | clip: 0.0ms | optim: 49.7ms
|
| 287 |
+
step 185 | loss: 1.609302 | lr 2.6293e-04 | norm: 0.4189 | dt: 6341.67ms | tok/sec: 82673.4 | mem: 108909MB | Time Left: 00 hrs : 22 mins
|
| 288 |
+
[TIMING] data: 3.5ms | fwd: 2027.3ms | loss: 6.5ms | bwd: 4246.8ms | clip: 0.0ms | optim: 49.5ms
|
| 289 |
+
step 186 | loss: 1.671908 | lr 2.6211e-04 | norm: 0.5121 | dt: 6334.59ms | tok/sec: 82765.8 | mem: 108909MB | Time Left: 00 hrs : 22 mins
|
| 290 |
+
[TIMING] data: 3.6ms | fwd: 2050.0ms | loss: 6.5ms | bwd: 4216.8ms | clip: 0.0ms | optim: 49.6ms
|
| 291 |
+
step 187 | loss: 1.645242 | lr 2.6129e-04 | norm: 0.4458 | dt: 6326.63ms | tok/sec: 82870.1 | mem: 108909MB | Time Left: 00 hrs : 22 mins
|
| 292 |
+
[TIMING] data: 3.6ms | fwd: 2029.0ms | loss: 6.8ms | bwd: 4229.5ms | clip: 0.0ms | optim: 49.8ms
|
| 293 |
+
step 188 | loss: 1.626171 | lr 2.6046e-04 | norm: 0.4034 | dt: 6339.96ms | tok/sec: 82695.8 | mem: 108909MB | Time Left: 00 hrs : 22 mins
|
| 294 |
+
[TIMING] data: 3.5ms | fwd: 2025.2ms | loss: 6.5ms | bwd: 4247.0ms | clip: 0.0ms | optim: 49.6ms
|
| 295 |
+
step 189 | loss: 1.634669 | lr 2.5962e-04 | norm: 0.3758 | dt: 6325.95ms | tok/sec: 82878.9 | mem: 108909MB | Time Left: 00 hrs : 22 mins
|
| 296 |
+
[TIMING] data: 3.5ms | fwd: 2019.9ms | loss: 6.5ms | bwd: 4238.5ms | clip: 0.0ms | optim: 49.5ms
|
| 297 |
+
step 190 | loss: 1.610499 | lr
|
logs/in2/driver_20260829_1535.log
ADDED
|
@@ -0,0 +1,114 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
[15:35:38] IN2 driver starting. base=/root/Adhi/BluTrain/dist/Context_Parallelism/checkpoints_bluscriptcp/rank_0/blumodelcp_run7_step_750.ckpt d_min=9835 R=9835
|
| 2 |
+
[15:35:38] generating arm C (train) at --d-min 9835
|
| 3 |
+
[15:35:45] IN2 driver starting. base=/root/Adhi/BluTrain/dist/Context_Parallelism/checkpoints_bluscriptcp/rank_0/blumodelcp_run7_step_750.ckpt d_min=9835 R=9835
|
| 4 |
+
[15:35:45] generating arm C (train) at --d-min 9835
|
| 5 |
+
wrote 3 shard(s) in in2_data/arm_c 209977344 tokens total
|
| 6 |
+
answer tokens 6561792 (3.12%) synthetic 22146048 (10.55%)
|
| 7 |
+
distance: min 9847 median 10879 max 16348 n=820224
|
| 8 |
+
[15:36:06] generating arm C (val)
|
| 9 |
+
wrote 1 shard(s) in in2_data/arm_c 1048576 tokens total
|
| 10 |
+
answer tokens 32768 (3.12%) synthetic 110592 (10.55%)
|
| 11 |
+
distance: min 9847 median 10857 max 16157 n=4096
|
| 12 |
+
[15:36:06] arm C distance: min 9847 median 10879 max 16348 n=820224
|
| 13 |
+
[15:36:06] arm C distance: min 9847 median 10857 max 16157 n=4096
|
| 14 |
+
[15:36:06] gate (train) on all three arms
|
| 15 |
+
wrote 3 shard(s) in in2_data/arm_c 209977344 tokens total
|
| 16 |
+
answer tokens 6561792 (3.12%) synthetic 22146048 (10.55%)
|
| 17 |
+
distance: min 9847 median 10879 max 16348 n=820224
|
| 18 |
+
[15:36:15] generating arm C (val)
|
| 19 |
+
wrote 1 shard(s) in in2_data/arm_c 1048576 tokens total
|
| 20 |
+
answer tokens 32768 (3.12%) synthetic 110592 (10.55%)
|
| 21 |
+
distance: min 9847 median 10857 max 16157 n=4096
|
| 22 |
+
[15:36:15] arm C distance: min 9847 median 10879 max 16348 n=820224
|
| 23 |
+
[15:36:15] arm C distance: min 9847 median 10857 max 16157 n=4096
|
| 24 |
+
[15:36:15] gate (train) on all three arms
|
| 25 |
+
|
| 26 |
+
=== arm a T=16384 examples=12816 K=64 ===
|
| 27 |
+
ok 1 total size across 3 shard(s) == examples*T*2 (419954688 bytes, got 419954688)
|
| 28 |
+
ok 2 tokens in [0,50304) (min 0, max 50256)
|
| 29 |
+
ok 3 no answer within the last 4 offsets (0 bad)
|
| 30 |
+
ok 4 planted spans never overlap (0 bad)
|
| 31 |
+
ok 5 RETRIEVAL POSSIBLE: probe==key and answer==value (0 bad)
|
| 32 |
+
ok 7 arm-a distance invariant (0 bad)
|
| 33 |
+
ok 6a batch stride 65536 is a multiple of T=16384
|
| 34 |
+
ok 6b total 209977344 tokens is a whole number of batches (stride 65536)
|
| 35 |
+
ok 6c row r starts at token r*T for all B=4 rows
|
| 36 |
+
ok 6d y[T-1] targets the next example's first token (why PAD_TAIL exists)
|
| 37 |
+
|
| 38 |
+
=== arm b T=16384 examples=12816 K=64 ===
|
| 39 |
+
ok 1 total size across 3 shard(s) == examples*T*2 (419954688 bytes, got 419954688)
|
| 40 |
+
ok 2 tokens in [0,50304) (min 0, max 50256)
|
| 41 |
+
ok 3 no answer within the last 4 offsets (0 bad)
|
| 42 |
+
ok 4 planted spans never overlap (0 bad)
|
| 43 |
+
ok 5 RETRIEVAL POSSIBLE: probe==key and answer==value (0 bad)
|
| 44 |
+
ok 7 arm-b distance invariant (0 bad)
|
| 45 |
+
ok 6a batch stride 65536 is a multiple of T=16384
|
| 46 |
+
ok 6b total 209977344 tokens is a whole number of batches (stride 65536)
|
| 47 |
+
ok 6c row r starts at token r*T for all B=4 rows
|
| 48 |
+
ok 6d y[T-1] targets the next example's first token (why PAD_TAIL exists)
|
| 49 |
+
|
| 50 |
+
=== arm c T=16384 examples=12816 K=64 ===
|
| 51 |
+
ok 1 total size across 3 shard(s) == examples*T*2 (419954688 bytes, got 419954688)
|
| 52 |
+
ok 2 tokens in [0,50304) (min 0, max 50256)
|
| 53 |
+
ok 3 no answer within the last 4 offsets (0 bad)
|
| 54 |
+
ok 4 planted spans never overlap (0 bad)
|
| 55 |
+
ok 5 RETRIEVAL POSSIBLE: probe==key and answer==value (0 bad)
|
| 56 |
+
ok 7 arm-c distance invariant (0 bad)
|
| 57 |
+
ok 6a batch stride 65536 is a multiple of T=16384
|
| 58 |
+
ok 6b total 209977344 tokens is a whole number of batches (stride 65536)
|
| 59 |
+
ok 6c row r starts at token r*T for all B=4 rows
|
| 60 |
+
ok 6d y[T-1] targets the next example's first token (why PAD_TAIL exists)
|
| 61 |
+
|
| 62 |
+
=== cross-arm ===
|
| 63 |
+
FAIL 8 arms B and C have BIT-IDENTICAL answer_start histograms (16293 vs 6522 distinct offsets)
|
| 64 |
+
ok 8b same seed
|
| 65 |
+
ok 8c same answer-token count
|
| 66 |
+
|
| 67 |
+
1 FAILURE(S)
|
| 68 |
+
[15:36:32] FATAL: GATE FAILED (train)
|
| 69 |
+
[15:36:32] aborting -- see in2_logs/driver_20260829_1535.log
|
| 70 |
+
|
| 71 |
+
=== arm a T=16384 examples=12816 K=64 ===
|
| 72 |
+
ok 1 total size across 3 shard(s) == examples*T*2 (419954688 bytes, got 419954688)
|
| 73 |
+
ok 2 tokens in [0,50304) (min 0, max 50256)
|
| 74 |
+
ok 3 no answer within the last 4 offsets (0 bad)
|
| 75 |
+
ok 4 planted spans never overlap (0 bad)
|
| 76 |
+
ok 5 RETRIEVAL POSSIBLE: probe==key and answer==value (0 bad)
|
| 77 |
+
ok 7 arm-a distance invariant (0 bad)
|
| 78 |
+
ok 6a batch stride 65536 is a multiple of T=16384
|
| 79 |
+
ok 6b total 209977344 tokens is a whole number of batches (stride 65536)
|
| 80 |
+
ok 6c row r starts at token r*T for all B=4 rows
|
| 81 |
+
ok 6d y[T-1] targets the next example's first token (why PAD_TAIL exists)
|
| 82 |
+
|
| 83 |
+
=== arm b T=16384 examples=12816 K=64 ===
|
| 84 |
+
ok 1 total size across 3 shard(s) == examples*T*2 (419954688 bytes, got 419954688)
|
| 85 |
+
ok 2 tokens in [0,50304) (min 0, max 50256)
|
| 86 |
+
ok 3 no answer within the last 4 offsets (0 bad)
|
| 87 |
+
ok 4 planted spans never overlap (0 bad)
|
| 88 |
+
ok 5 RETRIEVAL POSSIBLE: probe==key and answer==value (0 bad)
|
| 89 |
+
ok 7 arm-b distance invariant (0 bad)
|
| 90 |
+
ok 6a batch stride 65536 is a multiple of T=16384
|
| 91 |
+
ok 6b total 209977344 tokens is a whole number of batches (stride 65536)
|
| 92 |
+
ok 6c row r starts at token r*T for all B=4 rows
|
| 93 |
+
ok 6d y[T-1] targets the next example's first token (why PAD_TAIL exists)
|
| 94 |
+
|
| 95 |
+
=== arm c T=16384 examples=12816 K=64 ===
|
| 96 |
+
ok 1 total size across 3 shard(s) == examples*T*2 (419954688 bytes, got 419954688)
|
| 97 |
+
ok 2 tokens in [0,50304) (min 0, max 50256)
|
| 98 |
+
ok 3 no answer within the last 4 offsets (0 bad)
|
| 99 |
+
ok 4 planted spans never overlap (0 bad)
|
| 100 |
+
ok 5 RETRIEVAL POSSIBLE: probe==key and answer==value (0 bad)
|
| 101 |
+
ok 7 arm-c distance invariant (0 bad)
|
| 102 |
+
ok 6a batch stride 65536 is a multiple of T=16384
|
| 103 |
+
ok 6b total 209977344 tokens is a whole number of batches (stride 65536)
|
| 104 |
+
ok 6c row r starts at token r*T for all B=4 rows
|
| 105 |
+
ok 6d y[T-1] targets the next example's first token (why PAD_TAIL exists)
|
| 106 |
+
|
| 107 |
+
=== cross-arm ===
|
| 108 |
+
FAIL 8 arms B and C have BIT-IDENTICAL answer_start histograms (16293 vs 6522 distinct offsets)
|
| 109 |
+
ok 8b same seed
|
| 110 |
+
ok 8c same answer-token count
|
| 111 |
+
|
| 112 |
+
1 FAILURE(S)
|
| 113 |
+
[15:36:41] FATAL: GATE FAILED (train)
|
| 114 |
+
[15:36:41] aborting -- see in2_logs/driver_20260829_1535.log
|
logs/in2/driver_20260829_1542.log
ADDED
|
@@ -0,0 +1,92 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
[15:42:31] IN2 driver starting. base=/root/Adhi/BluTrain/dist/Context_Parallelism/checkpoints_bluscriptcp/rank_0/blumodelcp_run7_step_750.ckpt d_min=64 R=9835
|
| 2 |
+
[15:42:31] gate (train) on all three arms
|
| 3 |
+
|
| 4 |
+
=== arm a T=16384 examples=12816 K=64 ===
|
| 5 |
+
ok 1 total size across 3 shard(s) == examples*T*2 (419954688 bytes, got 419954688)
|
| 6 |
+
ok 2 tokens in [0,50304) (min 0, max 50256)
|
| 7 |
+
ok 3 no answer within the last 4 offsets (0 bad)
|
| 8 |
+
ok 4 planted spans never overlap (0 bad)
|
| 9 |
+
ok 5 RETRIEVAL POSSIBLE: probe==key and answer==value (0 bad)
|
| 10 |
+
ok 7 arm-a distance invariant (0 bad)
|
| 11 |
+
ok 6a batch stride 65536 is a multiple of T=16384
|
| 12 |
+
ok 6b total 209977344 tokens is a whole number of batches (stride 65536)
|
| 13 |
+
ok 6c row r starts at token r*T for all B=4 rows
|
| 14 |
+
ok 6d y[T-1] targets the next example's first token (why PAD_TAIL exists)
|
| 15 |
+
|
| 16 |
+
=== arm b T=16384 examples=12816 K=64 ===
|
| 17 |
+
ok 1 total size across 3 shard(s) == examples*T*2 (419954688 bytes, got 419954688)
|
| 18 |
+
ok 2 tokens in [0,50304) (min 0, max 50256)
|
| 19 |
+
ok 3 no answer within the last 4 offsets (0 bad)
|
| 20 |
+
ok 4 planted spans never overlap (0 bad)
|
| 21 |
+
ok 5 RETRIEVAL POSSIBLE: probe==key and answer==value (0 bad)
|
| 22 |
+
ok 7 arm-b distance invariant (0 bad)
|
| 23 |
+
ok 6a batch stride 65536 is a multiple of T=16384
|
| 24 |
+
ok 6b total 209977344 tokens is a whole number of batches (stride 65536)
|
| 25 |
+
ok 6c row r starts at token r*T for all B=4 rows
|
| 26 |
+
ok 6d y[T-1] targets the next example's first token (why PAD_TAIL exists)
|
| 27 |
+
|
| 28 |
+
=== arm c T=16384 examples=12816 K=64 ===
|
| 29 |
+
ok 1 total size across 3 shard(s) == examples*T*2 (419954688 bytes, got 419954688)
|
| 30 |
+
ok 2 tokens in [0,50304) (min 0, max 50256)
|
| 31 |
+
ok 3 no answer within the last 4 offsets (0 bad)
|
| 32 |
+
ok 4 planted spans never overlap (0 bad)
|
| 33 |
+
ok 5 RETRIEVAL POSSIBLE: probe==key and answer==value (0 bad)
|
| 34 |
+
ok 7 arm-c distance invariant (0 bad)
|
| 35 |
+
ok 6a batch stride 65536 is a multiple of T=16384
|
| 36 |
+
ok 6b total 209977344 tokens is a whole number of batches (stride 65536)
|
| 37 |
+
ok 6c row r starts at token r*T for all B=4 rows
|
| 38 |
+
ok 6d y[T-1] targets the next example's first token (why PAD_TAIL exists)
|
| 39 |
+
|
| 40 |
+
=== cross-arm ===
|
| 41 |
+
ok 8 arms B and C have BIT-IDENTICAL answer_start histograms (16293 vs 16293 distinct offsets)
|
| 42 |
+
ok 8b same seed
|
| 43 |
+
ok 8c same answer-token count
|
| 44 |
+
|
| 45 |
+
ALL PASS
|
| 46 |
+
[15:42:57] gate train: ALL PASS
|
| 47 |
+
[15:42:57] gate (val) on all three arms
|
| 48 |
+
|
| 49 |
+
=== arm a T=16384 examples=64 K=64 ===
|
| 50 |
+
ok 1 total size across 1 shard(s) == examples*T*2 (2097152 bytes, got 2097152)
|
| 51 |
+
ok 2 tokens in [0,50304) (min 0, max 50256)
|
| 52 |
+
ok 3 no answer within the last 4 offsets (0 bad)
|
| 53 |
+
ok 4 planted spans never overlap (0 bad)
|
| 54 |
+
ok 5 RETRIEVAL POSSIBLE: probe==key and answer==value (0 bad)
|
| 55 |
+
ok 7 arm-a distance invariant (0 bad)
|
| 56 |
+
ok 6a batch stride 65536 is a multiple of T=16384
|
| 57 |
+
ok 6b total 1048576 tokens is a whole number of batches (stride 65536)
|
| 58 |
+
ok 6c row r starts at token r*T for all B=4 rows
|
| 59 |
+
ok 6d y[T-1] targets the next example's first token (why PAD_TAIL exists)
|
| 60 |
+
|
| 61 |
+
=== arm b T=16384 examples=64 K=64 ===
|
| 62 |
+
ok 1 total size across 1 shard(s) == examples*T*2 (2097152 bytes, got 2097152)
|
| 63 |
+
ok 2 tokens in [0,50304) (min 0, max 50256)
|
| 64 |
+
ok 3 no answer within the last 4 offsets (0 bad)
|
| 65 |
+
ok 4 planted spans never overlap (0 bad)
|
| 66 |
+
ok 5 RETRIEVAL POSSIBLE: probe==key and answer==value (0 bad)
|
| 67 |
+
ok 7 arm-b distance invariant (0 bad)
|
| 68 |
+
ok 6a batch stride 65536 is a multiple of T=16384
|
| 69 |
+
ok 6b total 1048576 tokens is a whole number of batches (stride 65536)
|
| 70 |
+
ok 6c row r starts at token r*T for all B=4 rows
|
| 71 |
+
ok 6d y[T-1] targets the next example's first token (why PAD_TAIL exists)
|
| 72 |
+
|
| 73 |
+
=== arm c T=16384 examples=64 K=64 ===
|
| 74 |
+
ok 1 total size across 1 shard(s) == examples*T*2 (2097152 bytes, got 2097152)
|
| 75 |
+
ok 2 tokens in [0,50304) (min 0, max 50256)
|
| 76 |
+
ok 3 no answer within the last 4 offsets (0 bad)
|
| 77 |
+
ok 4 planted spans never overlap (0 bad)
|
| 78 |
+
ok 5 RETRIEVAL POSSIBLE: probe==key and answer==value (0 bad)
|
| 79 |
+
ok 7 arm-c distance invariant (0 bad)
|
| 80 |
+
ok 6a batch stride 65536 is a multiple of T=16384
|
| 81 |
+
ok 6b total 1048576 tokens is a whole number of batches (stride 65536)
|
| 82 |
+
ok 6c row r starts at token r*T for all B=4 rows
|
| 83 |
+
ok 6d y[T-1] targets the next example's first token (why PAD_TAIL exists)
|
| 84 |
+
|
| 85 |
+
=== cross-arm ===
|
| 86 |
+
ok 8 arms B and C have BIT-IDENTICAL answer_start histograms (3593 vs 3593 distinct offsets)
|
| 87 |
+
ok 8b same seed
|
| 88 |
+
ok 8c same answer-token count
|
| 89 |
+
|
| 90 |
+
ALL PASS
|
| 91 |
+
[15:42:57] gate val: ALL PASS
|
| 92 |
+
[15:42:57] training arm a (400 steps x 524288 tokens, T=16384, cp=4, YaRN=4)
|
logs/in2/driver_20260829_1545.log
ADDED
|
@@ -0,0 +1,94 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
[15:45:12] IN2 driver starting. base=/root/Adhi/BluTrain/dist/Context_Parallelism/checkpoints_bluscriptcp/rank_0/blumodelcp_run7_step_750.ckpt d_min=64 R=9835
|
| 2 |
+
[15:45:12] gate (train) on all three arms
|
| 3 |
+
|
| 4 |
+
=== arm a T=16384 examples=12816 K=64 ===
|
| 5 |
+
ok 1 total size across 3 shard(s) == examples*T*2 (419954688 bytes, got 419954688)
|
| 6 |
+
ok 2 tokens in [0,50304) (min 0, max 50256)
|
| 7 |
+
ok 3 no answer within the last 4 offsets (0 bad)
|
| 8 |
+
ok 4 planted spans never overlap (0 bad)
|
| 9 |
+
ok 5 RETRIEVAL POSSIBLE: probe==key and answer==value (0 bad)
|
| 10 |
+
ok 7 arm-a distance invariant (0 bad)
|
| 11 |
+
ok 6a batch stride 65536 is a multiple of T=16384
|
| 12 |
+
ok 6b total 209977344 tokens is a whole number of batches (stride 65536)
|
| 13 |
+
ok 6c row r starts at token r*T for all B=4 rows
|
| 14 |
+
ok 6d y[T-1] targets the next example's first token (why PAD_TAIL exists)
|
| 15 |
+
|
| 16 |
+
=== arm b T=16384 examples=12816 K=64 ===
|
| 17 |
+
ok 1 total size across 3 shard(s) == examples*T*2 (419954688 bytes, got 419954688)
|
| 18 |
+
ok 2 tokens in [0,50304) (min 0, max 50256)
|
| 19 |
+
ok 3 no answer within the last 4 offsets (0 bad)
|
| 20 |
+
ok 4 planted spans never overlap (0 bad)
|
| 21 |
+
ok 5 RETRIEVAL POSSIBLE: probe==key and answer==value (0 bad)
|
| 22 |
+
ok 7 arm-b distance invariant (0 bad)
|
| 23 |
+
ok 6a batch stride 65536 is a multiple of T=16384
|
| 24 |
+
ok 6b total 209977344 tokens is a whole number of batches (stride 65536)
|
| 25 |
+
ok 6c row r starts at token r*T for all B=4 rows
|
| 26 |
+
ok 6d y[T-1] targets the next example's first token (why PAD_TAIL exists)
|
| 27 |
+
|
| 28 |
+
=== arm c T=16384 examples=12816 K=64 ===
|
| 29 |
+
ok 1 total size across 3 shard(s) == examples*T*2 (419954688 bytes, got 419954688)
|
| 30 |
+
ok 2 tokens in [0,50304) (min 0, max 50256)
|
| 31 |
+
ok 3 no answer within the last 4 offsets (0 bad)
|
| 32 |
+
ok 4 planted spans never overlap (0 bad)
|
| 33 |
+
ok 5 RETRIEVAL POSSIBLE: probe==key and answer==value (0 bad)
|
| 34 |
+
ok 7 arm-c distance invariant (0 bad)
|
| 35 |
+
ok 6a batch stride 65536 is a multiple of T=16384
|
| 36 |
+
ok 6b total 209977344 tokens is a whole number of batches (stride 65536)
|
| 37 |
+
ok 6c row r starts at token r*T for all B=4 rows
|
| 38 |
+
ok 6d y[T-1] targets the next example's first token (why PAD_TAIL exists)
|
| 39 |
+
|
| 40 |
+
=== cross-arm ===
|
| 41 |
+
ok 8 arms B and C have BIT-IDENTICAL answer_start histograms (16293 vs 16293 distinct offsets)
|
| 42 |
+
ok 8b same seed
|
| 43 |
+
ok 8c same answer-token count
|
| 44 |
+
|
| 45 |
+
ALL PASS
|
| 46 |
+
[15:45:38] gate train: ALL PASS
|
| 47 |
+
[15:45:38] gate (val) on all three arms
|
| 48 |
+
|
| 49 |
+
=== arm a T=16384 examples=64 K=64 ===
|
| 50 |
+
ok 1 total size across 1 shard(s) == examples*T*2 (2097152 bytes, got 2097152)
|
| 51 |
+
ok 2 tokens in [0,50304) (min 0, max 50256)
|
| 52 |
+
ok 3 no answer within the last 4 offsets (0 bad)
|
| 53 |
+
ok 4 planted spans never overlap (0 bad)
|
| 54 |
+
ok 5 RETRIEVAL POSSIBLE: probe==key and answer==value (0 bad)
|
| 55 |
+
ok 7 arm-a distance invariant (0 bad)
|
| 56 |
+
ok 6a batch stride 65536 is a multiple of T=16384
|
| 57 |
+
ok 6b total 1048576 tokens is a whole number of batches (stride 65536)
|
| 58 |
+
ok 6c row r starts at token r*T for all B=4 rows
|
| 59 |
+
ok 6d y[T-1] targets the next example's first token (why PAD_TAIL exists)
|
| 60 |
+
|
| 61 |
+
=== arm b T=16384 examples=64 K=64 ===
|
| 62 |
+
ok 1 total size across 1 shard(s) == examples*T*2 (2097152 bytes, got 2097152)
|
| 63 |
+
ok 2 tokens in [0,50304) (min 0, max 50256)
|
| 64 |
+
ok 3 no answer within the last 4 offsets (0 bad)
|
| 65 |
+
ok 4 planted spans never overlap (0 bad)
|
| 66 |
+
ok 5 RETRIEVAL POSSIBLE: probe==key and answer==value (0 bad)
|
| 67 |
+
ok 7 arm-b distance invariant (0 bad)
|
| 68 |
+
ok 6a batch stride 65536 is a multiple of T=16384
|
| 69 |
+
ok 6b total 1048576 tokens is a whole number of batches (stride 65536)
|
| 70 |
+
ok 6c row r starts at token r*T for all B=4 rows
|
| 71 |
+
ok 6d y[T-1] targets the next example's first token (why PAD_TAIL exists)
|
| 72 |
+
|
| 73 |
+
=== arm c T=16384 examples=64 K=64 ===
|
| 74 |
+
ok 1 total size across 1 shard(s) == examples*T*2 (2097152 bytes, got 2097152)
|
| 75 |
+
ok 2 tokens in [0,50304) (min 0, max 50256)
|
| 76 |
+
ok 3 no answer within the last 4 offsets (0 bad)
|
| 77 |
+
ok 4 planted spans never overlap (0 bad)
|
| 78 |
+
ok 5 RETRIEVAL POSSIBLE: probe==key and answer==value (0 bad)
|
| 79 |
+
ok 7 arm-c distance invariant (0 bad)
|
| 80 |
+
ok 6a batch stride 65536 is a multiple of T=16384
|
| 81 |
+
ok 6b total 1048576 tokens is a whole number of batches (stride 65536)
|
| 82 |
+
ok 6c row r starts at token r*T for all B=4 rows
|
| 83 |
+
ok 6d y[T-1] targets the next example's first token (why PAD_TAIL exists)
|
| 84 |
+
|
| 85 |
+
=== cross-arm ===
|
| 86 |
+
ok 8 arms B and C have BIT-IDENTICAL answer_start histograms (3593 vs 3593 distinct offsets)
|
| 87 |
+
ok 8b same seed
|
| 88 |
+
ok 8c same answer-token count
|
| 89 |
+
|
| 90 |
+
ALL PASS
|
| 91 |
+
[15:45:38] gate val: ALL PASS
|
| 92 |
+
[15:45:38] training arm a (400 steps x 524288 tokens, T=16384, cp=4, YaRN=4)
|
| 93 |
+
[15:46:34] FATAL: arm a training failed -- see in2_logs/arm_a_train.log
|
| 94 |
+
[15:46:34] aborting -- see in2_logs/driver_20260829_1545.log
|
logs/in2/driver_20260829_1548.log
ADDED
|
@@ -0,0 +1,100 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
[15:48:58] IN2 driver starting. base=/root/Adhi/BluTrain/dist/Context_Parallelism/checkpoints_bluscriptcp/rank_0/blumodelcp_run7_step_750.ckpt d_min=64 R=9835
|
| 2 |
+
[15:48:58] gate (train) on all three arms
|
| 3 |
+
|
| 4 |
+
=== arm a T=16384 examples=12816 K=64 ===
|
| 5 |
+
ok 1 total size across 3 shard(s) == examples*T*2 (419954688 bytes, got 419954688)
|
| 6 |
+
ok 2 tokens in [0,50304) (min 0, max 50256)
|
| 7 |
+
ok 3 no answer within the last 4 offsets (0 bad)
|
| 8 |
+
ok 4 planted spans never overlap (0 bad)
|
| 9 |
+
ok 5 RETRIEVAL POSSIBLE: probe==key and answer==value (0 bad)
|
| 10 |
+
ok 7 arm-a distance invariant (0 bad)
|
| 11 |
+
ok 6a batch stride 65536 is a multiple of T=16384
|
| 12 |
+
ok 6b total 209977344 tokens is a whole number of batches (stride 65536)
|
| 13 |
+
ok 6c row r starts at token r*T for all B=4 rows
|
| 14 |
+
ok 6d y[T-1] targets the next example's first token (why PAD_TAIL exists)
|
| 15 |
+
|
| 16 |
+
=== arm b T=16384 examples=12816 K=64 ===
|
| 17 |
+
ok 1 total size across 3 shard(s) == examples*T*2 (419954688 bytes, got 419954688)
|
| 18 |
+
ok 2 tokens in [0,50304) (min 0, max 50256)
|
| 19 |
+
ok 3 no answer within the last 4 offsets (0 bad)
|
| 20 |
+
ok 4 planted spans never overlap (0 bad)
|
| 21 |
+
ok 5 RETRIEVAL POSSIBLE: probe==key and answer==value (0 bad)
|
| 22 |
+
ok 7 arm-b distance invariant (0 bad)
|
| 23 |
+
ok 6a batch stride 65536 is a multiple of T=16384
|
| 24 |
+
ok 6b total 209977344 tokens is a whole number of batches (stride 65536)
|
| 25 |
+
ok 6c row r starts at token r*T for all B=4 rows
|
| 26 |
+
ok 6d y[T-1] targets the next example's first token (why PAD_TAIL exists)
|
| 27 |
+
|
| 28 |
+
=== arm c T=16384 examples=12816 K=64 ===
|
| 29 |
+
ok 1 total size across 3 shard(s) == examples*T*2 (419954688 bytes, got 419954688)
|
| 30 |
+
ok 2 tokens in [0,50304) (min 0, max 50256)
|
| 31 |
+
ok 3 no answer within the last 4 offsets (0 bad)
|
| 32 |
+
ok 4 planted spans never overlap (0 bad)
|
| 33 |
+
ok 5 RETRIEVAL POSSIBLE: probe==key and answer==value (0 bad)
|
| 34 |
+
ok 7 arm-c distance invariant (0 bad)
|
| 35 |
+
ok 6a batch stride 65536 is a multiple of T=16384
|
| 36 |
+
ok 6b total 209977344 tokens is a whole number of batches (stride 65536)
|
| 37 |
+
ok 6c row r starts at token r*T for all B=4 rows
|
| 38 |
+
ok 6d y[T-1] targets the next example's first token (why PAD_TAIL exists)
|
| 39 |
+
|
| 40 |
+
=== cross-arm ===
|
| 41 |
+
ok 8 arms B and C have BIT-IDENTICAL answer_start histograms (16293 vs 16293 distinct offsets)
|
| 42 |
+
ok 8b same seed
|
| 43 |
+
ok 8c same answer-token count
|
| 44 |
+
|
| 45 |
+
ALL PASS
|
| 46 |
+
[15:49:23] gate train: ALL PASS
|
| 47 |
+
[15:49:23] gate (val) on all three arms
|
| 48 |
+
|
| 49 |
+
=== arm a T=16384 examples=64 K=64 ===
|
| 50 |
+
ok 1 total size across 1 shard(s) == examples*T*2 (2097152 bytes, got 2097152)
|
| 51 |
+
ok 2 tokens in [0,50304) (min 0, max 50256)
|
| 52 |
+
ok 3 no answer within the last 4 offsets (0 bad)
|
| 53 |
+
ok 4 planted spans never overlap (0 bad)
|
| 54 |
+
ok 5 RETRIEVAL POSSIBLE: probe==key and answer==value (0 bad)
|
| 55 |
+
ok 7 arm-a distance invariant (0 bad)
|
| 56 |
+
ok 6a batch stride 65536 is a multiple of T=16384
|
| 57 |
+
ok 6b total 1048576 tokens is a whole number of batches (stride 65536)
|
| 58 |
+
ok 6c row r starts at token r*T for all B=4 rows
|
| 59 |
+
ok 6d y[T-1] targets the next example's first token (why PAD_TAIL exists)
|
| 60 |
+
|
| 61 |
+
=== arm b T=16384 examples=64 K=64 ===
|
| 62 |
+
ok 1 total size across 1 shard(s) == examples*T*2 (2097152 bytes, got 2097152)
|
| 63 |
+
ok 2 tokens in [0,50304) (min 0, max 50256)
|
| 64 |
+
ok 3 no answer within the last 4 offsets (0 bad)
|
| 65 |
+
ok 4 planted spans never overlap (0 bad)
|
| 66 |
+
ok 5 RETRIEVAL POSSIBLE: probe==key and answer==value (0 bad)
|
| 67 |
+
ok 7 arm-b distance invariant (0 bad)
|
| 68 |
+
ok 6a batch stride 65536 is a multiple of T=16384
|
| 69 |
+
ok 6b total 1048576 tokens is a whole number of batches (stride 65536)
|
| 70 |
+
ok 6c row r starts at token r*T for all B=4 rows
|
| 71 |
+
ok 6d y[T-1] targets the next example's first token (why PAD_TAIL exists)
|
| 72 |
+
|
| 73 |
+
=== arm c T=16384 examples=64 K=64 ===
|
| 74 |
+
ok 1 total size across 1 shard(s) == examples*T*2 (2097152 bytes, got 2097152)
|
| 75 |
+
ok 2 tokens in [0,50304) (min 0, max 50256)
|
| 76 |
+
ok 3 no answer within the last 4 offsets (0 bad)
|
| 77 |
+
ok 4 planted spans never overlap (0 bad)
|
| 78 |
+
ok 5 RETRIEVAL POSSIBLE: probe==key and answer==value (0 bad)
|
| 79 |
+
ok 7 arm-c distance invariant (0 bad)
|
| 80 |
+
ok 6a batch stride 65536 is a multiple of T=16384
|
| 81 |
+
ok 6b total 1048576 tokens is a whole number of batches (stride 65536)
|
| 82 |
+
ok 6c row r starts at token r*T for all B=4 rows
|
| 83 |
+
ok 6d y[T-1] targets the next example's first token (why PAD_TAIL exists)
|
| 84 |
+
|
| 85 |
+
=== cross-arm ===
|
| 86 |
+
ok 8 arms B and C have BIT-IDENTICAL answer_start histograms (3593 vs 3593 distinct offsets)
|
| 87 |
+
ok 8b same seed
|
| 88 |
+
ok 8c same answer-token count
|
| 89 |
+
|
| 90 |
+
ALL PASS
|
| 91 |
+
[15:49:23] gate val: ALL PASS
|
| 92 |
+
[15:49:23] training arm a (400 steps x 524288 tokens, T=16384, cp=4, YaRN=4)
|
| 93 |
+
[16:33:29] arm a done -> run 12, final loss loss: 1.349215
|
| 94 |
+
[16:33:29] training arm b (400 steps x 524288 tokens, T=16384, cp=4, YaRN=4)
|
| 95 |
+
[17:17:52] arm b done -> run 13, final loss loss: 1.451133
|
| 96 |
+
[17:17:52] training arm c (400 steps x 524288 tokens, T=16384, cp=4, YaRN=4)
|
| 97 |
+
[18:01:55] arm c done -> run 14, final loss loss: 1.610499
|
| 98 |
+
[18:01:55] arm C val_loss: first=1.888691 last=1.888691
|
| 99 |
+
[18:01:55] FATAL: LEARNABILITY GATE FAILED: arm C val_loss did not fall (1.888691 -> 1.888691). The run says nothing about IN2.
|
| 100 |
+
[18:01:55] aborting -- see in2_logs/driver_20260829_1548.log
|