Delete run_data

run_data/qwen_full.log

@@ -1,196 +0,0 @@
-NCCL version 2.21.5+cuda12.4
-`torch_dtype` is deprecated! Use `dtype` instead!
-`torch_dtype` is deprecated! Use `dtype` instead!
-`torch_dtype` is deprecated! Use `dtype` instead!
-`torch_dtype` is deprecated! Use `dtype` instead!
-`torch_dtype` is deprecated! Use `dtype` instead!
-`torch_dtype` is deprecated! Use `dtype` instead!
-`torch_dtype` is deprecated! Use `dtype` instead!
-`torch_dtype` is deprecated! Use `dtype` instead!
-
-
-
-
-
-
-
-
-Too many dataloader workers: 32 (max is dataset.num_shards=7). Stopping 25 dataloader workers.
-Too many dataloader workers: 32 (max is dataset.num_shards=7). Stopping 25 dataloader workers.
-Too many dataloader workers: 32 (max is dataset.num_shards=6). Stopping 26 dataloader workers.
-Too many dataloader workers: 32 (max is dataset.num_shards=6). Stopping 26 dataloader workers.
-Too many dataloader workers: 32 (max is dataset.num_shards=6). Stopping 26 dataloader workers.
-Too many dataloader workers: 32 (max is dataset.num_shards=6). Stopping 26 dataloader workers.
-[rank1]:W0307 03:26:40.690000 29391 torch/_logging/_internal.py:1089] [0/0] Profiler function <class 'torch.autograd.profiler.record_function'> will be ignored
-Too many dataloader workers: 32 (max is dataset.num_shards=6). Stopping 26 dataloader workers.
-Too many dataloader workers: 32 (max is dataset.num_shards=6). Stopping 26 dataloader workers.
-[rank7]:W0307 03:26:43.017000 29397 torch/_logging/_internal.py:1089] [0/0] Profiler function <class 'torch.autograd.profiler.record_function'> will be ignored
-[rank0]:W0307 03:26:43.105000 29390 torch/_logging/_internal.py:1089] [0/0] Profiler function <class 'torch.autograd.profiler.record_function'> will be ignored
-[rank3]:W0307 03:26:43.220000 29393 torch/_logging/_internal.py:1089] [0/0] Profiler function <class 'torch.autograd.profiler.record_function'> will be ignored
-[rank5]:W0307 03:26:44.141000 29395 torch/_logging/_internal.py:1089] [0/0] Profiler function <class 'torch.autograd.profiler.record_function'> will be ignored
-[rank4]:W0307 03:26:44.231000 29394 torch/_logging/_internal.py:1089] [0/0] Profiler function <class 'torch.autograd.profiler.record_function'> will be ignored
-[rank6]:W0307 03:26:46.461000 29396 torch/_logging/_internal.py:1089] [0/0] Profiler function <class 'torch.autograd.profiler.record_function'> will be ignored
-[rank2]:W0307 03:26:49.647000 29392 torch/_logging/_internal.py:1089] [0/0] Profiler function <class 'torch.autograd.profiler.record_function'> will be ignored
-/root/PyC/.venv/lib/python3.10/site-packages/torch/_inductor/compile_fx.py:194: UserWarning: TensorFloat32 tensor cores for float32 matrix multiplication available but not enabled. Consider setting `torch.set_float32_matmul_precision('high')` for better performance.
-  warnings.warn(
-/root/PyC/.venv/lib/python3.10/site-packages/torch/_inductor/compile_fx.py:194: UserWarning: TensorFloat32 tensor cores for float32 matrix multiplication available but not enabled. Consider setting `torch.set_float32_matmul_precision('high')` for better performance.
-  warnings.warn(
-/root/PyC/.venv/lib/python3.10/site-packages/torch/_inductor/compile_fx.py:194: UserWarning: TensorFloat32 tensor cores for float32 matrix multiplication available but not enabled. Consider setting `torch.set_float32_matmul_precision('high')` for better performance.
-  warnings.warn(
-/root/PyC/.venv/lib/python3.10/site-packages/torch/_inductor/compile_fx.py:194: UserWarning: TensorFloat32 tensor cores for float32 matrix multiplication available but not enabled. Consider setting `torch.set_float32_matmul_precision('high')` for better performance.
-  warnings.warn(
-/root/PyC/.venv/lib/python3.10/site-packages/torch/_inductor/compile_fx.py:194: UserWarning: TensorFloat32 tensor cores for float32 matrix multiplication available but not enabled. Consider setting `torch.set_float32_matmul_precision('high')` for better performance.
-  warnings.warn(
-/root/PyC/.venv/lib/python3.10/site-packages/torch/_inductor/compile_fx.py:194: UserWarning: TensorFloat32 tensor cores for float32 matrix multiplication available but not enabled. Consider setting `torch.set_float32_matmul_precision('high')` for better performance.
-  warnings.warn(
-/root/PyC/.venv/lib/python3.10/site-packages/torch/_inductor/compile_fx.py:194: UserWarning: TensorFloat32 tensor cores for float32 matrix multiplication available but not enabled. Consider setting `torch.set_float32_matmul_precision('high')` for better performance.
-  warnings.warn(
-/root/PyC/.venv/lib/python3.10/site-packages/torch/_inductor/compile_fx.py:194: UserWarning: TensorFloat32 tensor cores for float32 matrix multiplication available but not enabled. Consider setting `torch.set_float32_matmul_precision('high')` for better performance.
-  warnings.warn(
-[train_sft] step=10/2000 loss=0.3106 tokens=66019 samples_per_sec=0.42 tokens_per_sec=172.88
-[train_sft] step=20/2000 loss=0.2507 tokens=130469 samples_per_sec=0.51 tokens_per_sec=209.38
-[train_sft] step=30/2000 loss=0.2291 tokens=196381 samples_per_sec=0.55 tokens_per_sec=226.84
-[train_sft] step=40/2000 loss=0.2799 tokens=260313 samples_per_sec=0.58 tokens_per_sec=234.92
-[train_sft] step=50/2000 loss=0.0738 tokens=326884 samples_per_sec=0.59 tokens_per_sec=241.97
-[train_sft] step=60/2000 loss=0.2171 tokens=392579 samples_per_sec=0.60 tokens_per_sec=246.20
-[train_sft] step=70/2000 loss=0.1810 tokens=458551 samples_per_sec=0.61 tokens_per_sec=249.57
-[train_sft] step=80/2000 loss=0.1329 tokens=527838 samples_per_sec=0.62 tokens_per_sec=253.71
-[train_sft] step=90/2000 loss=0.1080 tokens=598351 samples_per_sec=0.62 tokens_per_sec=257.55
-[train_sft] step=100/2000 loss=0.1271 tokens=665669 samples_per_sec=0.62 tokens_per_sec=259.31
-[train_sft] step=110/2000 loss=0.1596 tokens=727376 samples_per_sec=0.63 tokens_per_sec=258.71
-[train_sft] step=120/2000 loss=0.1730 tokens=789579 samples_per_sec=0.63 tokens_per_sec=258.39
-[train_sft] step=130/2000 loss=0.1666 tokens=852405 samples_per_sec=0.63 tokens_per_sec=258.28
-[train_sft] step=140/2000 loss=0.1658 tokens=922146 samples_per_sec=0.63 tokens_per_sec=260.12
-[train_sft] step=150/2000 loss=0.0985 tokens=986385 samples_per_sec=0.63 tokens_per_sec=260.33
-[train_sft] step=160/2000 loss=0.0701 tokens=1049899 samples_per_sec=0.63 tokens_per_sec=260.30
-[train_sft] step=170/2000 loss=0.0789 tokens=1114566 samples_per_sec=0.64 tokens_per_sec=260.56
-[train_sft] step=180/2000 loss=0.1549 tokens=1184173 samples_per_sec=0.64 tokens_per_sec=261.97
-[train_sft] step=190/2000 loss=0.1046 tokens=1248076 samples_per_sec=0.64 tokens_per_sec=261.99
-[train_sft] step=200/2000 loss=0.2644 tokens=1312805 samples_per_sec=0.64 tokens_per_sec=262.13
-[train_sft] eval step=200 loss=n/a
-[train_sft] step=210/2000 loss=0.1345 tokens=1378046 samples_per_sec=0.64 tokens_per_sec=262.34
-[train_sft] step=220/2000 loss=0.1363 tokens=1440118 samples_per_sec=0.64 tokens_per_sec=261.97
-[train_sft] step=230/2000 loss=0.1488 tokens=1506320 samples_per_sec=0.64 tokens_per_sec=262.35
-[train_sft] step=240/2000 loss=0.1060 tokens=1571030 samples_per_sec=0.64 tokens_per_sec=262.45
-[train_sft] step=250/2000 loss=0.1710 tokens=1636503 samples_per_sec=0.64 tokens_per_sec=262.64
-[train_sft] step=260/2000 loss=0.1923 tokens=1702374 samples_per_sec=0.64 tokens_per_sec=262.90
-[train_sft] step=270/2000 loss=0.1278 tokens=1765267 samples_per_sec=0.64 tokens_per_sec=262.69
-[train_sft] step=280/2000 loss=0.1669 tokens=1834329 samples_per_sec=0.64 tokens_per_sec=263.40
-[train_sft] step=290/2000 loss=0.1825 tokens=1901061 samples_per_sec=0.64 tokens_per_sec=263.73
-[train_sft] step=300/2000 loss=0.1071 tokens=1965869 samples_per_sec=0.64 tokens_per_sec=263.76
-[train_sft] step=310/2000 loss=0.1100 tokens=2026328 samples_per_sec=0.64 tokens_per_sec=263.24
-[train_sft] step=320/2000 loss=0.1569 tokens=2091438 samples_per_sec=0.64 tokens_per_sec=263.35
-[train_sft] step=330/2000 loss=0.0981 tokens=2159388 samples_per_sec=0.64 tokens_per_sec=263.79
-[train_sft] step=340/2000 loss=0.1418 tokens=2224937 samples_per_sec=0.65 tokens_per_sec=263.91
-[train_sft] step=350/2000 loss=0.1125 tokens=2290801 samples_per_sec=0.65 tokens_per_sec=264.05
-[train_sft] step=360/2000 loss=0.0595 tokens=2356921 samples_per_sec=0.65 tokens_per_sec=264.21
-[train_sft] step=370/2000 loss=0.0877 tokens=2419567 samples_per_sec=0.65 tokens_per_sec=264.01
-[train_sft] step=380/2000 loss=0.2055 tokens=2484959 samples_per_sec=0.65 tokens_per_sec=264.08
-[train_sft] step=390/2000 loss=0.2391 tokens=2548421 samples_per_sec=0.65 tokens_per_sec=263.97
-[train_sft] step=400/2000 loss=0.1708 tokens=2611454 samples_per_sec=0.65 tokens_per_sec=263.81
-[train_sft] eval step=400 loss=n/a
-[train_sft] step=410/2000 loss=0.1243 tokens=2673912 samples_per_sec=0.65 tokens_per_sec=263.61
-[train_sft] step=420/2000 loss=0.1155 tokens=2741532 samples_per_sec=0.65 tokens_per_sec=263.91
-[train_sft] step=430/2000 loss=0.1285 tokens=2803400 samples_per_sec=0.65 tokens_per_sec=263.66
-[train_sft] step=440/2000 loss=0.0763 tokens=2870087 samples_per_sec=0.65 tokens_per_sec=263.86
-[train_sft] step=450/2000 loss=0.1212 tokens=2938147 samples_per_sec=0.65 tokens_per_sec=264.18
-[train_sft] step=460/2000 loss=0.1697 tokens=3006919 samples_per_sec=0.65 tokens_per_sec=264.55
-[train_sft] step=470/2000 loss=0.0634 tokens=3074418 samples_per_sec=0.65 tokens_per_sec=264.79
-[train_sft] step=480/2000 loss=0.1071 tokens=3140612 samples_per_sec=0.65 tokens_per_sec=264.92
-[train_sft] step=490/2000 loss=0.0840 tokens=3207555 samples_per_sec=0.65 tokens_per_sec=265.09
-[train_sft] step=500/2000 loss=0.1685 tokens=3272736 samples_per_sec=0.65 tokens_per_sec=265.12
-
-[train_sft] step=510/2000 loss=11.9358 tokens=3333255 samples_per_sec=0.64 tokens_per_sec=261.82
-[train_sft] step=520/2000 loss=11.9305 tokens=3400813 samples_per_sec=0.64 tokens_per_sec=260.02
-[train_sft] step=530/2000 loss=11.9318 tokens=3468065 samples_per_sec=0.63 tokens_per_sec=258.30
-[train_sft] step=540/2000 loss=11.9336 tokens=3535022 samples_per_sec=0.63 tokens_per_sec=256.63
-[train_sft] step=550/2000 loss=11.9334 tokens=3600853 samples_per_sec=0.62 tokens_per_sec=254.97
-[train_sft] step=560/2000 loss=11.9341 tokens=3662071 samples_per_sec=0.62 tokens_per_sec=253.07
-[train_sft] step=570/2000 loss=11.9347 tokens=3730017 samples_per_sec=0.62 tokens_per_sec=251.71
-[train_sft] step=580/2000 loss=11.9316 tokens=3796693 samples_per_sec=0.61 tokens_per_sec=250.33
-[train_sft] step=590/2000 loss=11.9386 tokens=3858933 samples_per_sec=0.61 tokens_per_sec=248.73
-[train_sft] step=600/2000 loss=11.9369 tokens=3920622 samples_per_sec=0.61 tokens_per_sec=247.17
-[train_sft] eval step=600 loss=n/a
-[train_sft] step=610/2000 loss=11.9311 tokens=3984325 samples_per_sec=0.60 tokens_per_sec=245.79
-[train_sft] step=620/2000 loss=11.9305 tokens=4051226 samples_per_sec=0.60 tokens_per_sec=244.67
-[train_sft] step=630/2000 loss=11.9329 tokens=4117533 samples_per_sec=0.60 tokens_per_sec=243.55
-[train_sft] step=640/2000 loss=11.9343 tokens=4183625 samples_per_sec=0.59 tokens_per_sec=242.47
-[train_sft] step=650/2000 loss=11.9327 tokens=4252154 samples_per_sec=0.59 tokens_per_sec=241.57
-[train_sft] step=660/2000 loss=11.9336 tokens=4313556 samples_per_sec=0.59 tokens_per_sec=240.31
-[train_sft] step=670/2000 loss=11.9328 tokens=4376047 samples_per_sec=0.59 tokens_per_sec=239.16
-[train_sft] step=680/2000 loss=11.9349 tokens=4443065 samples_per_sec=0.58 tokens_per_sec=238.29
-[train_sft] step=690/2000 loss=11.9312 tokens=4508609 samples_per_sec=0.58 tokens_per_sec=237.37
-[train_sft] step=700/2000 loss=11.9365 tokens=4571767 samples_per_sec=0.58 tokens_per_sec=236.37
-[train_sft] step=710/2000 loss=11.9332 tokens=4635156 samples_per_sec=0.58 tokens_per_sec=235.42
-[train_sft] step=720/2000 loss=11.9340 tokens=4700197 samples_per_sec=0.57 tokens_per_sec=234.57
-[train_sft] step=730/2000 loss=11.9302 tokens=4764290 samples_per_sec=0.57 tokens_per_sec=233.71
-[train_sft] step=740/2000 loss=11.9358 tokens=4831552 samples_per_sec=0.57 tokens_per_sec=233.03
-[train_sft] step=750/2000 loss=11.9338 tokens=4896926 samples_per_sec=0.57 tokens_per_sec=232.29
-[train_sft] step=760/2000 loss=11.9299 tokens=4963790 samples_per_sec=0.57 tokens_per_sec=231.64
-[train_sft] step=770/2000 loss=11.9375 tokens=5030112 samples_per_sec=0.57 tokens_per_sec=230.98
-[train_sft] step=780/2000 loss=11.9328 tokens=5093917 samples_per_sec=0.56 tokens_per_sec=230.23
-[train_sft] step=790/2000 loss=11.9376 tokens=5160155 samples_per_sec=0.56 tokens_per_sec=229.61
-[train_sft] step=800/2000 loss=11.9353 tokens=5229774 samples_per_sec=0.56 tokens_per_sec=229.16
-[train_sft] eval step=800 loss=n/a
-[train_sft] step=810/2000 loss=11.9334 tokens=5295154 samples_per_sec=0.56 tokens_per_sec=228.54
-[train_sft] step=820/2000 loss=11.9345 tokens=5357311 samples_per_sec=0.56 tokens_per_sec=227.80
-[train_sft] step=830/2000 loss=11.9387 tokens=5422927 samples_per_sec=0.56 tokens_per_sec=227.23
-[train_sft] step=840/2000 loss=11.9358 tokens=5490488 samples_per_sec=0.56 tokens_per_sec=226.75
-[train_sft] step=850/2000 loss=11.9340 tokens=5551808 samples_per_sec=0.55 tokens_per_sec=226.04
-[train_sft] step=860/2000 loss=11.9372 tokens=5617359 samples_per_sec=0.55 tokens_per_sec=225.52
-[train_sft] step=870/2000 loss=11.9311 tokens=5685346 samples_per_sec=0.55 tokens_per_sec=225.10
-[train_sft] step=880/2000 loss=11.9340 tokens=5750741 samples_per_sec=0.55 tokens_per_sec=224.59
-[train_sft] step=890/2000 loss=11.9300 tokens=5816629 samples_per_sec=0.55 tokens_per_sec=224.12
-[train_sft] step=900/2000 loss=11.9343 tokens=5882089 samples_per_sec=0.55 tokens_per_sec=223.64
-[train_sft] step=910/2000 loss=11.9320 tokens=5948442 samples_per_sec=0.55 tokens_per_sec=223.21
-[train_sft] step=920/2000 loss=11.9363 tokens=6009051 samples_per_sec=0.55 tokens_per_sec=222.58
-[train_sft] step=930/2000 loss=11.9309 tokens=6078081 samples_per_sec=0.54 tokens_per_sec=222.27
-[train_sft] step=940/2000 loss=11.9353 tokens=6138758 samples_per_sec=0.54 tokens_per_sec=221.66
-[train_sft] step=950/2000 loss=11.9371 tokens=6204982 samples_per_sec=0.54 tokens_per_sec=221.27
-[train_sft] step=960/2000 loss=11.9341 tokens=6272247 samples_per_sec=0.54 tokens_per_sec=220.93
-[train_sft] step=970/2000 loss=11.9314 tokens=6337243 samples_per_sec=0.54 tokens_per_sec=220.51
-[train_sft] step=980/2000 loss=11.9361 tokens=6401913 samples_per_sec=0.54 tokens_per_sec=220.10
-[train_sft] step=990/2000 loss=11.9325 tokens=6465635 samples_per_sec=0.54 tokens_per_sec=219.66
-[train_sft] step=1000/2000 loss=11.9347 tokens=6532582 samples_per_sec=0.54 tokens_per_sec=219.34
-[train_sft] eval step=1000 loss=n/a
-
-[train_sft] step=1010/2000 loss=nan tokens=6598011 samples_per_sec=0.54 tokens_per_sec=219.25
-W0307 11:46:34.841000 29377 torch/distributed/elastic/agent/server/api.py:719] Received Signals.SIGINT death signal, shutting down workers
-W0307 11:46:34.843000 29377 torch/distributed/elastic/multiprocessing/api.py:897] Sending process 29390 closing signal SIGINT
-W0307 11:46:34.843000 29377 torch/distributed/elastic/multiprocessing/api.py:897] Sending process 29391 closing signal SIGINT
-W0307 11:46:34.844000 29377 torch/distributed/elastic/multiprocessing/api.py:897] Sending process 29392 closing signal SIGINT
-W0307 11:46:34.844000 29377 torch/distributed/elastic/multiprocessing/api.py:897] Sending process 29393 closing signal SIGINT
-W0307 11:46:34.844000 29377 torch/distributed/elastic/multiprocessing/api.py:897] Sending process 29394 closing signal SIGINT
-W0307 11:46:34.844000 29377 torch/distributed/elastic/multiprocessing/api.py:897] Sending process 29395 closing signal SIGINT
-W0307 11:46:34.845000 29377 torch/distributed/elastic/multiprocessing/api.py:897] Sending process 29396 closing signal SIGINT
-W0307 11:46:34.845000 29377 torch/distributed/elastic/multiprocessing/api.py:897] Sending process 29397 closing signal SIGINT
-W0307 11:47:04.846000 29377 torch/distributed/elastic/multiprocessing/api.py:916] Unable to shutdown process 29390 via Signals.SIGINT, forcefully exiting via Signals.SIGKILL
-W0307 11:47:05.104000 29377 torch/distributed/elastic/multiprocessing/api.py:916] Unable to shutdown process 29391 via Signals.SIGINT, forcefully exiting via Signals.SIGKILL
-W0307 11:47:05.500000 29377 torch/distributed/elastic/multiprocessing/api.py:916] Unable to shutdown process 29392 via Signals.SIGINT, forcefully exiting via Signals.SIGKILL
-W0307 11:47:05.889000 29377 torch/distributed/elastic/multiprocessing/api.py:916] Unable to shutdown process 29393 via Signals.SIGINT, forcefully exiting via Signals.SIGKILL
-W0307 11:47:06.314000 29377 torch/distributed/elastic/multiprocessing/api.py:916] Unable to shutdown process 29394 via Signals.SIGINT, forcefully exiting via Signals.SIGKILL
-W0307 11:47:06.691000 29377 torch/distributed/elastic/multiprocessing/api.py:916] Unable to shutdown process 29395 via Signals.SIGINT, forcefully exiting via Signals.SIGKILL
-W0307 11:47:07.074000 29377 torch/distributed/elastic/multiprocessing/api.py:916] Unable to shutdown process 29396 via Signals.SIGINT, forcefully exiting via Signals.SIGKILL
-W0307 11:47:07.463000 29377 torch/distributed/elastic/multiprocessing/api.py:916] Unable to shutdown process 29397 via Signals.SIGINT, forcefully exiting via Signals.SIGKILL
-Traceback (most recent call last):
-  File "/root/PyC/.venv/bin/torchrun", line 6, in <module>
-    sys.exit(main())
-  File "/root/PyC/.venv/lib/python3.10/site-packages/torch/distributed/elastic/multiprocessing/errors/__init__.py", line 355, in wrapper
-    return f(*args, **kwargs)
-  File "/root/PyC/.venv/lib/python3.10/site-packages/torch/distributed/run.py", line 918, in main
-    run(args)
-  File "/root/PyC/.venv/lib/python3.10/site-packages/torch/distributed/run.py", line 909, in run
-    elastic_launch(
-  File "/root/PyC/.venv/lib/python3.10/site-packages/torch/distributed/launcher/api.py", line 138, in __call__
-    return launch_agent(self._config, self._entrypoint, list(args))
-  File "/root/PyC/.venv/lib/python3.10/site-packages/torch/distributed/launcher/api.py", line 260, in launch_agent
-    result = agent.run()
-  File "/root/PyC/.venv/lib/python3.10/site-packages/torch/distributed/elastic/metrics/api.py", line 137, in wrapper
-    result = f(*args, **kwargs)
-  File "/root/PyC/.venv/lib/python3.10/site-packages/torch/distributed/elastic/agent/server/api.py", line 711, in run
-    result = self._invoke_run(role)
-  File "/root/PyC/.venv/lib/python3.10/site-packages/torch/distributed/elastic/agent/server/api.py", line 870, in _invoke_run
-    time.sleep(monitor_interval)
-  File "/root/PyC/.venv/lib/python3.10/site-packages/torch/distributed/elastic/multiprocessing/api.py", line 84, in _terminate_process_handler
-    raise SignalException(f"Process {os.getpid()} got signal: {sigval}", sigval=sigval)
-torch.distributed.elastic.multiprocessing.api.SignalException: Process 29377 got signal: 2

run_data/train_metrics.json

@@ -1,286 +0,0 @@
-{
-  "run_id": "qwen14b_codeinstruct",
-  "timestamp_utc": "2026-03-07T03:24:36Z",
-  "host": "5cea9450de15",
-  "train_runtime_sec": 30093.7884,
-  "samples_per_sec": 0.537,
-  "steps_per_sec": 0.0336,
-  "tokens_per_sec": 219.2483,
-  "loss_final": NaN,
-  "loss_curve": [
-    11.933758,
-    11.935425,
-    11.936735,
-    11.934477,
-    11.936081,
-    11.934277,
-    11.933105,
-    11.934149,
-    11.931114,
-    11.931803,
-    11.931644,
-    11.933023,
-    11.935471,
-    11.934968,
-    11.935336,
-    11.936069,
-    11.931381,
-    11.929764,
-    11.933843,
-    11.932031,
-    11.934188,
-    11.933489,
-    11.932564,
-    11.933738,
-    11.933002,
-    11.935679,
-    11.930973,
-    11.933707,
-    11.935345,
-    11.932213,
-    11.931738,
-    11.938071,
-    11.934389,
-    11.935558,
-    11.933566,
-    11.931173,
-    11.933773,
-    11.930953,
-    11.933154,
-    11.935999,
-    11.934997,
-    11.934099,
-    11.93278,
-    11.938663,
-    11.933143,
-    11.932686,
-    11.934217,
-    11.934953,
-    11.936878,
-    11.933104,
-    11.930088,
-    11.929935,
-    11.932352,
-    11.932141,
-    11.934212,
-    11.937968,
-    11.93185,
-    11.935164,
-    11.933124,
-    11.934265,
-    11.931581,
-    11.93028,
-    11.935074,
-    11.930362,
-    11.93231,
-    11.934386,
-    11.931125,
-    11.933617,
-    11.934947,
-    11.935481,
-    11.931095,
-    11.936635,
-    11.934564,
-    11.934128,
-    11.933428,
-    11.933599,
-    11.93398,
-    11.930263,
-    11.934587,
-    11.933071,
-    11.932242,
-    11.933808,
-    11.931849,
-    11.936655,
-    11.932829,
-    11.931339,
-    11.936092,
-    11.935149,
-    11.934838,
-    11.932211,
-    11.939731,
-    11.935974,
-    11.93283,
-    11.933028,
-    11.938888,
-    11.932542,
-    11.935485,
-    11.93673,
-    11.930267,
-    11.933331,
-    11.934375,
-    11.932673,
-    11.93618,
-    11.933468,
-    11.932549,
-    11.936689,
-    11.932197,
-    11.934677,
-    11.934959,
-    11.932065,
-    11.934155,
-    11.9322,
-    11.935388,
-    11.935803,
-    11.936669,
-    11.929192,
-    11.937821,
-    11.933041,
-    11.933342,
-    11.935324,
-    11.934742,
-    11.934669,
-    11.935793,
-    11.937335,
-    11.931247,
-    11.932714,
-    11.933646,
-    11.935673,
-    11.932477,
-    11.931051,
-    11.935923,
-    11.935831,
-    11.935593,
-    11.937632,
-    11.933506,
-    11.933863,
-    11.936568,
-    11.93511,
-    11.929846,
-    11.928335,
-    11.935637,
-    11.931632,
-    11.933343,
-    11.932386,
-    11.935142,
-    11.934553,
-    11.937015,
-    11.930168,
-    11.937183,
-    11.934591,
-    11.937156,
-    11.934317,
-    11.933546,
-    11.937324,
-    11.92873,
-    11.934661,
-    11.935086,
-    11.934525,
-    11.934164,
-    11.933876,
-    11.932452,
-    11.932648,
-    11.934206,
-    11.932563,
-    11.934017,
-    11.934988,
-    11.933892,
-    11.933537,
-    11.929354,
-    11.935569,
-    11.93031,
-    11.934333,
-    11.934294,
-    11.931088,
-    11.936719,
-    11.934738,
-    11.932435,
-    11.934471,
-    11.934482,
-    11.93504,
-    11.934249,
-    11.931784,
-    11.932494,
-    11.932414,
-    11.933865,
-    11.931005,
-    11.93541,
-    11.932836,
-    11.935291,
-    11.933075,
-    11.931273,
-    11.931405,
-    11.93127,
-    11.931889,
-    11.935592,
-    11.930029,
-    11.935498,
-    11.934025,
-    11.935754,
-    11.932719,
-    11.933537,
-    11.935834,
-    11.933773,
-    11.931642,
-    11.931832,
-    11.934774,
-    11.932487,
-    11.932528,
-    11.932735,
-    11.934239,
-    11.93345,
-    11.933942,
-    11.931937,
-    11.937059,
-    11.938506,
-    11.932685,
-    11.932717,
-    11.936692,
-    11.9363,
-    11.936151,
-    11.931432,
-    11.933336,
-    11.932334,
-    11.932056,
-    11.934895,
-    11.933721,
-    11.934547,
-    11.933809,
-    11.933655,
-    11.933565,
-    11.933448,
-    11.932833,
-    11.936954,
-    11.935034,
-    11.931088,
-    11.935302,
-    11.933675,
-    11.935065,
-    11.935748,
-    11.932056,
-    11.938281,
-    11.934908,
-    11.932071,
-    11.933029,
-    11.934909,
-    11.933907,
-    11.933957,
-    NaN,
-    NaN,
-    NaN,
-    NaN,
-    NaN,
-    NaN,
-    NaN,
-    NaN,
-    NaN
-  ],
-  "eval_loss": null,
-  "gpu_util_mean": 99.2073,
-  "gpu_util_p95": 100.0,
-  "h2d_time_ms_mean": 0.0726,
-  "compute_time_ms_mean": 3697.3475,
-  "comm_time_ms_mean": 6.5765,
-  "idle_gap_ms_mean": 11.6392,
-  "idle_gap_ms_p95": 0.3205,
-  "mode": "sft",
-  "dist": "fsdp",
-  "backend": "nccl",
-  "world_size": 8,
-  "model_name": "Qwen/Qwen2.5-14B",
-  "dataset_name": "nvidia/OpenCodeInstruct",
-  "dataset_config": "",
-  "precision": "bf16",
-  "seq_length": 4096,
-  "gradient_accumulation_steps": 8
-}