Add files using upload-large-folder tool

LTA_openwebtext_dualt/logs/lta_owt_gpt2cached_len1024_rollout1_p1_ddit768x12_gbs512_8gpu_1m_20260513_143213.log

@@ -0,0 +1,924 @@
+t-20260513223132-g9wrc-worker-0:10255:10255 [0] NCCL INFO NCCL_SOCKET_IFNAME set by environment to eth1
+t-20260513223132-g9wrc-worker-0:10255:10255 [0] NCCL INFO Bootstrap: Using eth1:10.82.24.54<0>
+t-20260513223132-g9wrc-worker-0:10255:10255 [0] NCCL INFO cudaDriverVersion 12080
+t-20260513223132-g9wrc-worker-0:10255:10255 [0] NCCL INFO NCCL version 2.25.1+cuda12.8
+t-20260513223132-g9wrc-worker-0:10255:10255 [0] NCCL INFO Comm config Blocking set to 1
+t-20260513223132-g9wrc-worker-0:10260:10260 [5] NCCL INFO cudaDriverVersion 12080
+t-20260513223132-g9wrc-worker-0:10260:10260 [5] NCCL INFO NCCL_SOCKET_IFNAME set by environment to eth1
+t-20260513223132-g9wrc-worker-0:10260:10260 [5] NCCL INFO Bootstrap: Using eth1:10.82.24.54<0>
+t-20260513223132-g9wrc-worker-0:10260:10260 [5] NCCL INFO NCCL version 2.25.1+cuda12.8
+t-20260513223132-g9wrc-worker-0:10260:10260 [5] NCCL INFO Comm config Blocking set to 1
+t-20260513223132-g9wrc-worker-0:10261:10261 [6] NCCL INFO cudaDriverVersion 12080
+t-20260513223132-g9wrc-worker-0:10261:10261 [6] NCCL INFO NCCL_SOCKET_IFNAME set by environment to eth1
+t-20260513223132-g9wrc-worker-0:10261:10261 [6] NCCL INFO Bootstrap: Using eth1:10.82.24.54<0>
+t-20260513223132-g9wrc-worker-0:10261:10261 [6] NCCL INFO NCCL version 2.25.1+cuda12.8
+t-20260513223132-g9wrc-worker-0:10261:10261 [6] NCCL INFO Comm config Blocking set to 1
+t-20260513223132-g9wrc-worker-0:10262:10262 [7] NCCL INFO cudaDriverVersion 12080
+t-20260513223132-g9wrc-worker-0:10262:10262 [7] NCCL INFO NCCL_SOCKET_IFNAME set by environment to eth1
+t-20260513223132-g9wrc-worker-0:10262:10262 [7] NCCL INFO Bootstrap: Using eth1:10.82.24.54<0>
+t-20260513223132-g9wrc-worker-0:10262:10262 [7] NCCL INFO NCCL version 2.25.1+cuda12.8
+t-20260513223132-g9wrc-worker-0:10262:10262 [7] NCCL INFO Comm config Blocking set to 1
+t-20260513223132-g9wrc-worker-0:10256:10256 [1] NCCL INFO cudaDriverVersion 12080
+t-20260513223132-g9wrc-worker-0:10256:10256 [1] NCCL INFO NCCL_SOCKET_IFNAME set by environment to eth1
+t-20260513223132-g9wrc-worker-0:10256:10256 [1] NCCL INFO Bootstrap: Using eth1:10.82.24.54<0>
+t-20260513223132-g9wrc-worker-0:10256:10256 [1] NCCL INFO NCCL version 2.25.1+cuda12.8
+t-20260513223132-g9wrc-worker-0:10256:10256 [1] NCCL INFO Comm config Blocking set to 1
+t-20260513223132-g9wrc-worker-0:10259:10259 [4] NCCL INFO cudaDriverVersion 12080
+t-20260513223132-g9wrc-worker-0:10259:10259 [4] NCCL INFO NCCL_SOCKET_IFNAME set by environment to eth1
+t-20260513223132-g9wrc-worker-0:10259:10259 [4] NCCL INFO Bootstrap: Using eth1:10.82.24.54<0>
+t-20260513223132-g9wrc-worker-0:10259:10259 [4] NCCL INFO NCCL version 2.25.1+cuda12.8
+t-20260513223132-g9wrc-worker-0:10258:10258 [3] NCCL INFO cudaDriverVersion 12080
+t-20260513223132-g9wrc-worker-0:10258:10258 [3] NCCL INFO NCCL_SOCKET_IFNAME set by environment to eth1
+t-20260513223132-g9wrc-worker-0:10257:10257 [2] NCCL INFO cudaDriverVersion 12080
+t-20260513223132-g9wrc-worker-0:10257:10257 [2] NCCL INFO NCCL_SOCKET_IFNAME set by environment to eth1
+t-20260513223132-g9wrc-worker-0:10258:10258 [3] NCCL INFO Bootstrap: Using eth1:10.82.24.54<0>
+t-20260513223132-g9wrc-worker-0:10258:10258 [3] NCCL INFO NCCL version 2.25.1+cuda12.8
+t-20260513223132-g9wrc-worker-0:10259:10259 [4] NCCL INFO Comm config Blocking set to 1
+t-20260513223132-g9wrc-worker-0:10257:10257 [2] NCCL INFO Bootstrap: Using eth1:10.82.24.54<0>
+t-20260513223132-g9wrc-worker-0:10257:10257 [2] NCCL INFO NCCL version 2.25.1+cuda12.8
+t-20260513223132-g9wrc-worker-0:10258:10258 [3] NCCL INFO Comm config Blocking set to 1
+t-20260513223132-g9wrc-worker-0:10257:10257 [2] NCCL INFO Comm config Blocking set to 1
+t-20260513223132-g9wrc-worker-0:10255:10327 [0] NCCL INFO NET/Plugin: Loaded net plugin NCCL RDMA Plugin v9 (v9)
+t-20260513223132-g9wrc-worker-0:10255:10327 [0] NCCL INFO NET/Plugin: Loaded collnet plugin SHARP (v9)
+t-20260513223132-g9wrc-worker-0:10255:10327 [0] NCCL INFO Plugin Path : /opt/hpcx/nccl_rdma_sharp_plugin/lib/libnccl-net.so
+t-20260513223132-g9wrc-worker-0:10255:10327 [0] NCCL INFO P2P plugin v9 IBext_v9
+t-20260513223132-g9wrc-worker-0:10255:10327 [0] NCCL INFO NCCL_SOCKET_IFNAME set by environment to eth1
+t-20260513223132-g9wrc-worker-0:10260:10328 [5] NCCL INFO NET/Plugin: Loaded net plugin NCCL RDMA Plugin v9 (v9)
+t-20260513223132-g9wrc-worker-0:10260:10328 [5] NCCL INFO NET/Plugin: Loaded collnet plugin SHARP (v9)
+t-20260513223132-g9wrc-worker-0:10260:10328 [5] NCCL INFO Plugin Path : /opt/hpcx/nccl_rdma_sharp_plugin/lib/libnccl-net.so
+t-20260513223132-g9wrc-worker-0:10260:10328 [5] NCCL INFO P2P plugin v9 IBext_v9
+t-20260513223132-g9wrc-worker-0:10260:10328 [5] NCCL INFO NCCL_SOCKET_IFNAME set by environment to eth1
+t-20260513223132-g9wrc-worker-0:10255:10327 [0] NCCL INFO NCCL_IB_PCI_RELAXED_ORDERING set by environment to 1.
+t-20260513223132-g9wrc-worker-0:10255:10327 [0] NCCL INFO NET/IB : Using [0]mlx5_1:1/RoCE [1]mlx5_4:1/RoCE [2]mlx5_5:1/RoCE [3]mlx5_6:1/RoCE [4]mlx5_7:1/RoCE [5]mlx5_8:1/RoCE [6]mlx5_9:1/RoCE [7]mlx5_10:1/RoCE [RO]; OOB eth1:10.82.24.54<0>
+t-20260513223132-g9wrc-worker-0:10255:10327 [0] NCCL INFO PROFILER/Plugin: Could not find: libnccl-profiler.so.
+t-20260513223132-g9wrc-worker-0:10255:10327 [0] NCCL INFO Using network IBext_v9
+t-20260513223132-g9wrc-worker-0:10261:10329 [6] NCCL INFO NET/Plugin: Loaded net plugin NCCL RDMA Plugin v9 (v9)
+t-20260513223132-g9wrc-worker-0:10261:10329 [6] NCCL INFO NET/Plugin: Loaded collnet plugin SHARP (v9)
+t-20260513223132-g9wrc-worker-0:10261:10329 [6] NCCL INFO Plugin Path : /opt/hpcx/nccl_rdma_sharp_plugin/lib/libnccl-net.so
+t-20260513223132-g9wrc-worker-0:10261:10329 [6] NCCL INFO P2P plugin v9 IBext_v9
+t-20260513223132-g9wrc-worker-0:10261:10329 [6] NCCL INFO NCCL_SOCKET_IFNAME set by environment to eth1
+t-20260513223132-g9wrc-worker-0:10262:10330 [7] NCCL INFO NET/Plugin: Loaded net plugin NCCL RDMA Plugin v9 (v9)
+t-20260513223132-g9wrc-worker-0:10262:10330 [7] NCCL INFO NET/Plugin: Loaded collnet plugin SHARP (v9)
+t-20260513223132-g9wrc-worker-0:10262:10330 [7] NCCL INFO Plugin Path : /opt/hpcx/nccl_rdma_sharp_plugin/lib/libnccl-net.so
+t-20260513223132-g9wrc-worker-0:10262:10330 [7] NCCL INFO P2P plugin v9 IBext_v9
+t-20260513223132-g9wrc-worker-0:10262:10330 [7] NCCL INFO NCCL_SOCKET_IFNAME set by environment to eth1
+t-20260513223132-g9wrc-worker-0:10256:10331 [1] NCCL INFO NET/Plugin: Loaded net plugin NCCL RDMA Plugin v9 (v9)
+t-20260513223132-g9wrc-worker-0:10256:10331 [1] NCCL INFO NET/Plugin: Loaded collnet plugin SHARP (v9)
+t-20260513223132-g9wrc-worker-0:10256:10331 [1] NCCL INFO Plugin Path : /opt/hpcx/nccl_rdma_sharp_plugin/lib/libnccl-net.so
+t-20260513223132-g9wrc-worker-0:10256:10331 [1] NCCL INFO P2P plugin v9 IBext_v9
+t-20260513223132-g9wrc-worker-0:10256:10331 [1] NCCL INFO NCCL_SOCKET_IFNAME set by environment to eth1
+t-20260513223132-g9wrc-worker-0:10260:10328 [5] NCCL INFO NCCL_IB_PCI_RELAXED_ORDERING set by environment to 1.
+t-20260513223132-g9wrc-worker-0:10260:10328 [5] NCCL INFO NET/IB : Using [0]mlx5_1:1/RoCE [1]mlx5_4:1/RoCE [2]mlx5_5:1/RoCE [3]mlx5_6:1/RoCE [4]mlx5_7:1/RoCE [5]mlx5_8:1/RoCE [6]mlx5_9:1/RoCE [7]mlx5_10:1/RoCE [RO]; OOB eth1:10.82.24.54<0>
+t-20260513223132-g9wrc-worker-0:10260:10328 [5] NCCL INFO PROFILER/Plugin: Could not find: libnccl-profiler.so.
+t-20260513223132-g9wrc-worker-0:10260:10328 [5] NCCL INFO Using network IBext_v9
+t-20260513223132-g9wrc-worker-0:10261:10329 [6] NCCL INFO NCCL_IB_PCI_RELAXED_ORDERING set by environment to 1.
+t-20260513223132-g9wrc-worker-0:10261:10329 [6] NCCL INFO NET/IB : Using [0]mlx5_1:1/RoCE [1]mlx5_4:1/RoCE [2]mlx5_5:1/RoCE [3]mlx5_6:1/RoCE [4]mlx5_7:1/RoCE [5]mlx5_8:1/RoCE [6]mlx5_9:1/RoCE [7]mlx5_10:1/RoCE [RO]; OOB eth1:10.82.24.54<0>
+t-20260513223132-g9wrc-worker-0:10261:10329 [6] NCCL INFO PROFILER/Plugin: Could not find: libnccl-profiler.so.
+t-20260513223132-g9wrc-worker-0:10261:10329 [6] NCCL INFO Using network IBext_v9
+t-20260513223132-g9wrc-worker-0:10262:10330 [7] NCCL INFO NCCL_IB_PCI_RELAXED_ORDERING set by environment to 1.
+t-20260513223132-g9wrc-worker-0:10262:10330 [7] NCCL INFO NET/IB : Using [0]mlx5_1:1/RoCE [1]mlx5_4:1/RoCE [2]mlx5_5:1/RoCE [3]mlx5_6:1/RoCE [4]mlx5_7:1/RoCE [5]mlx5_8:1/RoCE [6]mlx5_9:1/RoCE [7]mlx5_10:1/RoCE [RO]; OOB eth1:10.82.24.54<0>
+t-20260513223132-g9wrc-worker-0:10262:10330 [7] NCCL INFO PROFILER/Plugin: Could not find: libnccl-profiler.so.
+t-20260513223132-g9wrc-worker-0:10262:10330 [7] NCCL INFO Using network IBext_v9
+t-20260513223132-g9wrc-worker-0:10256:10331 [1] NCCL INFO NCCL_IB_PCI_RELAXED_ORDERING set by environment to 1.
+t-20260513223132-g9wrc-worker-0:10256:10331 [1] NCCL INFO NET/IB : Using [0]mlx5_1:1/RoCE [1]mlx5_4:1/RoCE [2]mlx5_5:1/RoCE [3]mlx5_6:1/RoCE [4]mlx5_7:1/RoCE [5]mlx5_8:1/RoCE [6]mlx5_9:1/RoCE [7]mlx5_10:1/RoCE [RO]; OOB eth1:10.82.24.54<0>
+t-20260513223132-g9wrc-worker-0:10256:10331 [1] NCCL INFO PROFILER/Plugin: Could not find: libnccl-profiler.so.
+t-20260513223132-g9wrc-worker-0:10256:10331 [1] NCCL INFO Using network IBext_v9
+t-20260513223132-g9wrc-worker-0:10258:10333 [3] NCCL INFO NET/Plugin: Loaded net plugin NCCL RDMA Plugin v9 (v9)
+t-20260513223132-g9wrc-worker-0:10258:10333 [3] NCCL INFO NET/Plugin: Loaded collnet plugin SHARP (v9)
+t-20260513223132-g9wrc-worker-0:10258:10333 [3] NCCL INFO Plugin Path : /opt/hpcx/nccl_rdma_sharp_plugin/lib/libnccl-net.so
+t-20260513223132-g9wrc-worker-0:10258:10333 [3] NCCL INFO P2P plugin v9 IBext_v9
+t-20260513223132-g9wrc-worker-0:10258:10333 [3] NCCL INFO NCCL_SOCKET_IFNAME set by environment to eth1
+t-20260513223132-g9wrc-worker-0:10258:10333 [3] NCCL INFO NCCL_IB_PCI_RELAXED_ORDERING set by environment to 1.
+t-20260513223132-g9wrc-worker-0:10258:10333 [3] NCCL INFO NET/IB : Using [0]mlx5_1:1/RoCE [1]mlx5_4:1/RoCE [2]mlx5_5:1/RoCE [3]mlx5_6:1/RoCE [4]mlx5_7:1/RoCE [5]mlx5_8:1/RoCE [6]mlx5_9:1/RoCE [7]mlx5_10:1/RoCE [RO]; OOB eth1:10.82.24.54<0>
+t-20260513223132-g9wrc-worker-0:10258:10333 [3] NCCL INFO PROFILER/Plugin: Could not find: libnccl-profiler.so.
+t-20260513223132-g9wrc-worker-0:10258:10333 [3] NCCL INFO Using network IBext_v9
+t-20260513223132-g9wrc-worker-0:10257:10334 [2] NCCL INFO NET/Plugin: Loaded net plugin NCCL RDMA Plugin v9 (v9)
+t-20260513223132-g9wrc-worker-0:10257:10334 [2] NCCL INFO NET/Plugin: Loaded collnet plugin SHARP (v9)
+t-20260513223132-g9wrc-worker-0:10257:10334 [2] NCCL INFO Plugin Path : /opt/hpcx/nccl_rdma_sharp_plugin/lib/libnccl-net.so
+t-20260513223132-g9wrc-worker-0:10257:10334 [2] NCCL INFO P2P plugin v9 IBext_v9
+t-20260513223132-g9wrc-worker-0:10257:10334 [2] NCCL INFO NCCL_SOCKET_IFNAME set by environment to eth1
+t-20260513223132-g9wrc-worker-0:10257:10334 [2] NCCL INFO NCCL_IB_PCI_RELAXED_ORDERING set by environment to 1.
+t-20260513223132-g9wrc-worker-0:10257:10334 [2] NCCL INFO NET/IB : Using [0]mlx5_1:1/RoCE [1]mlx5_4:1/RoCE [2]mlx5_5:1/RoCE [3]mlx5_6:1/RoCE [4]mlx5_7:1/RoCE [5]mlx5_8:1/RoCE [6]mlx5_9:1/RoCE [7]mlx5_10:1/RoCE [RO]; OOB eth1:10.82.24.54<0>
+t-20260513223132-g9wrc-worker-0:10257:10334 [2] NCCL INFO PROFILER/Plugin: Could not find: libnccl-profiler.so.
+t-20260513223132-g9wrc-worker-0:10257:10334 [2] NCCL INFO Using network IBext_v9
+t-20260513223132-g9wrc-worker-0:10259:10332 [4] NCCL INFO NET/Plugin: Loaded net plugin NCCL RDMA Plugin v9 (v9)
+t-20260513223132-g9wrc-worker-0:10259:10332 [4] NCCL INFO NET/Plugin: Loaded collnet plugin SHARP (v9)
+t-20260513223132-g9wrc-worker-0:10259:10332 [4] NCCL INFO Plugin Path : /opt/hpcx/nccl_rdma_sharp_plugin/lib/libnccl-net.so
+t-20260513223132-g9wrc-worker-0:10259:10332 [4] NCCL INFO P2P plugin v9 IBext_v9
+t-20260513223132-g9wrc-worker-0:10259:10332 [4] NCCL INFO NCCL_SOCKET_IFNAME set by environment to eth1
+t-20260513223132-g9wrc-worker-0:10259:10332 [4] NCCL INFO NCCL_IB_PCI_RELAXED_ORDERING set by environment to 1.
+t-20260513223132-g9wrc-worker-0:10259:10332 [4] NCCL INFO NET/IB : Using [0]mlx5_1:1/RoCE [1]mlx5_4:1/RoCE [2]mlx5_5:1/RoCE [3]mlx5_6:1/RoCE [4]mlx5_7:1/RoCE [5]mlx5_8:1/RoCE [6]mlx5_9:1/RoCE [7]mlx5_10:1/RoCE [RO]; OOB eth1:10.82.24.54<0>
+t-20260513223132-g9wrc-worker-0:10259:10332 [4] NCCL INFO PROFILER/Plugin: Could not find: libnccl-profiler.so.
+t-20260513223132-g9wrc-worker-0:10259:10332 [4] NCCL INFO Using network IBext_v9
+t-20260513223132-g9wrc-worker-0:10255:10327 [0] NCCL INFO ncclCommInitRankConfig comm 0xb5c2750 rank 0 nranks 8 cudaDev 0 nvmlDev 0 busId 65040 commId 0x93a3726e358b8ce4 - Init START
+t-20260513223132-g9wrc-worker-0:10260:10328 [5] NCCL INFO ncclCommInitRankConfig comm 0xa89c920 rank 5 nranks 8 cudaDev 5 nvmlDev 5 busId 71020 commId 0x93a3726e358b8ce4 - Init START
+t-20260513223132-g9wrc-worker-0:10261:10329 [6] NCCL INFO ncclCommInitRankConfig comm 0xb044c70 rank 6 nranks 8 cudaDev 6 nvmlDev 6 busId 73020 commId 0x93a3726e358b8ce4 - Init START
+t-20260513223132-g9wrc-worker-0:10262:10330 [7] NCCL INFO ncclCommInitRankConfig comm 0xab21280 rank 7 nranks 8 cudaDev 7 nvmlDev 7 busId 75020 commId 0x93a3726e358b8ce4 - Init START
+t-20260513223132-g9wrc-worker-0:10262:10330 [7] NCCL INFO RAS client listening socket at ::1<28028>
+t-20260513223132-g9wrc-worker-0:10261:10329 [6] NCCL INFO RAS client listening socket at ::1<28028>
+t-20260513223132-g9wrc-worker-0:10256:10331 [1] NCCL INFO ncclCommInitRankConfig comm 0x9e75e40 rank 1 nranks 8 cudaDev 1 nvmlDev 1 busId 67020 commId 0x93a3726e358b8ce4 - Init START
+t-20260513223132-g9wrc-worker-0:10255:10327 [0] NCCL INFO RAS client listening socket at ::1<28028>
+t-20260513223132-g9wrc-worker-0:10258:10333 [3] NCCL INFO ncclCommInitRankConfig comm 0xb5ef120 rank 3 nranks 8 cudaDev 3 nvmlDev 3 busId 6b020 commId 0x93a3726e358b8ce4 - Init START
+t-20260513223132-g9wrc-worker-0:10257:10334 [2] NCCL INFO ncclCommInitRankConfig comm 0xb0afb80 rank 2 nranks 8 cudaDev 2 nvmlDev 2 busId 69020 commId 0x93a3726e358b8ce4 - Init START
+t-20260513223132-g9wrc-worker-0:10257:10334 [2] NCCL INFO RAS client listening socket at ::1<28028>
+t-20260513223132-g9wrc-worker-0:10256:10331 [1] NCCL INFO RAS client listening socket at ::1<28028>
+t-20260513223132-g9wrc-worker-0:10259:10332 [4] NCCL INFO ncclCommInitRankConfig comm 0xb1d1a20 rank 4 nranks 8 cudaDev 4 nvmlDev 4 busId 6f020 commId 0x93a3726e358b8ce4 - Init START
+t-20260513223132-g9wrc-worker-0:10258:10333 [3] NCCL INFO RAS client listening socket at ::1<28028>
+t-20260513223132-g9wrc-worker-0:10259:10332 [4] NCCL INFO RAS client listening socket at ::1<28028>
+t-20260513223132-g9wrc-worker-0:10260:10328 [5] NCCL INFO RAS client listening socket at ::1<28028>
+t-20260513223132-g9wrc-worker-0:10258:10333 [3] NCCL INFO Bootstrap timings total 0.034379 (create 0.000020, send 0.000057, recv 0.033803, ring 0.000246, delay 0.000001)
+t-20260513223132-g9wrc-worker-0:10257:10334 [2] NCCL INFO Bootstrap timings total 0.011056 (create 0.000023, send 0.000066, recv 0.000108, ring 0.010248, delay 0.000001)
+t-20260513223132-g9wrc-worker-0:10259:10332 [4] NCCL INFO Bootstrap timings total 0.000652 (create 0.000021, send 0.000064, recv 0.000078, ring 0.000251, delay 0.000001)
+t-20260513223132-g9wrc-worker-0:10260:10328 [5] NCCL INFO Bootstrap timings total 0.138167 (create 0.000020, send 0.000063, recv 0.047278, ring 0.000119, delay 0.000001)
+t-20260513223132-g9wrc-worker-0:10262:10330 [7] NCCL INFO Bootstrap timings total 0.072953 (create 0.000026, send 0.000075, recv 0.000083, ring 0.072479, delay 0.000001)
+t-20260513223132-g9wrc-worker-0:10255:10327 [0] NCCL INFO Bootstrap timings total 0.347872 (create 0.000026, send 0.000083, recv 0.282423, ring 0.064983, delay 0.000001)
+t-20260513223132-g9wrc-worker-0:10256:10331 [1] NCCL INFO Bootstrap timings total 0.067223 (create 0.000024, send 0.000091, recv 0.056518, ring 0.010257, delay 0.000001)
+t-20260513223132-g9wrc-worker-0:10261:10329 [6] NCCL INFO Bootstrap timings total 0.091017 (create 0.000024, send 0.000084, recv 0.018098, ring 0.072477, delay 0.000001)
+t-20260513223132-g9wrc-worker-0:10259:10332 [4] NCCL INFO MNNVL busId 0x6f020 fabric UUID 0.0 cliqueId 0x0 state 3 healthMask 0x0
+t-20260513223132-g9wrc-worker-0:10262:10330 [7] NCCL INFO MNNVL busId 0x75020 fabric UUID 0.0 cliqueId 0x0 state 3 healthMask 0x0
+t-20260513223132-g9wrc-worker-0:10258:10333 [3] NCCL INFO MNNVL busId 0x6b020 fabric UUID 0.0 cliqueId 0x0 state 3 healthMask 0x0
+t-20260513223132-g9wrc-worker-0:10255:10327 [0] NCCL INFO MNNVL busId 0x65040 fabric UUID 0.0 cliqueId 0x0 state 3 healthMask 0x0
+t-20260513223132-g9wrc-worker-0:10257:10334 [2] NCCL INFO MNNVL busId 0x69020 fabric UUID 0.0 cliqueId 0x0 state 3 healthMask 0x0
+t-20260513223132-g9wrc-worker-0:10256:10331 [1] NCCL INFO MNNVL busId 0x67020 fabric UUID 0.0 cliqueId 0x0 state 3 healthMask 0x0
+t-20260513223132-g9wrc-worker-0:10261:10329 [6] NCCL INFO MNNVL busId 0x73020 fabric UUID 0.0 cliqueId 0x0 state 3 healthMask 0x0
+t-20260513223132-g9wrc-worker-0:10260:10328 [5] NCCL INFO MNNVL busId 0x71020 fabric UUID 0.0 cliqueId 0x0 state 3 healthMask 0x0
+t-20260513223132-g9wrc-worker-0:10259:10332 [4] NCCL INFO NCCL_TOPO_FILE set by environment to /var/run/nvidia-topologyd/virtualTopology.xml
+t-20260513223132-g9wrc-worker-0:10255:10327 [0] NCCL INFO NCCL_TOPO_FILE set by environment to /var/run/nvidia-topologyd/virtualTopology.xml
+t-20260513223132-g9wrc-worker-0:10258:10333 [3] NCCL INFO NCCL_TOPO_FILE set by environment to /var/run/nvidia-topologyd/virtualTopology.xml
+t-20260513223132-g9wrc-worker-0:10256:10331 [1] NCCL INFO NCCL_TOPO_FILE set by environment to /var/run/nvidia-topologyd/virtualTopology.xml
+t-20260513223132-g9wrc-worker-0:10262:10330 [7] NCCL INFO NCCL_TOPO_FILE set by environment to /var/run/nvidia-topologyd/virtualTopology.xml
+t-20260513223132-g9wrc-worker-0:10257:10334 [2] NCCL INFO NCCL_TOPO_FILE set by environment to /var/run/nvidia-topologyd/virtualTopology.xml
+t-20260513223132-g9wrc-worker-0:10261:10329 [6] NCCL INFO NCCL_TOPO_FILE set by environment to /var/run/nvidia-topologyd/virtualTopology.xml
+t-20260513223132-g9wrc-worker-0:10260:10328 [5] NCCL INFO NCCL_TOPO_FILE set by environment to /var/run/nvidia-topologyd/virtualTopology.xml
+t-20260513223132-g9wrc-worker-0:10261:10329 [6] NCCL INFO Setting affinity for GPU 6 to 0fffff,ffffffff,ffffffff,fc000000,00000000,00000000
+t-20260513223132-g9wrc-worker-0:10261:10329 [6] NCCL INFO NVLS multicast support is available on dev 6
+t-20260513223132-g9wrc-worker-0:10259:10332 [4] NCCL INFO Setting affinity for GPU 4 to 0fffff,ffffffff,ffffffff,fc000000,00000000,00000000
+t-20260513223132-g9wrc-worker-0:10259:10332 [4] NCCL INFO NVLS multicast support is available on dev 4
+t-20260513223132-g9wrc-worker-0:10260:10328 [5] NCCL INFO Setting affinity for GPU 5 to 0fffff,ffffffff,ffffffff,fc000000,00000000,00000000
+t-20260513223132-g9wrc-worker-0:10258:10333 [3] NCCL INFO Setting affinity for GPU 3 to 03ffffff,ffffffff,ffffffff
+t-20260513223132-g9wrc-worker-0:10258:10333 [3] NCCL INFO NVLS multicast support is available on dev 3
+t-20260513223132-g9wrc-worker-0:10257:10334 [2] NCCL INFO Setting affinity for GPU 2 to 03ffffff,ffffffff,ffffffff
+t-20260513223132-g9wrc-worker-0:10257:10334 [2] NCCL INFO NVLS multicast support is available on dev 2
+t-20260513223132-g9wrc-worker-0:10255:10327 [0] NCCL INFO Setting affinity for GPU 0 to 03ffffff,ffffffff,ffffffff
+t-20260513223132-g9wrc-worker-0:10255:10327 [0] NCCL INFO NVLS multicast support is available on dev 0
+t-20260513223132-g9wrc-worker-0:10260:10328 [5] NCCL INFO NVLS multicast support is available on dev 5
+t-20260513223132-g9wrc-worker-0:10256:10331 [1] NCCL INFO Setting affinity for GPU 1 to 03ffffff,ffffffff,ffffffff
+t-20260513223132-g9wrc-worker-0:10262:10330 [7] NCCL INFO Setting affinity for GPU 7 to 0fffff,ffffffff,ffffffff,fc000000,00000000,00000000
+t-20260513223132-g9wrc-worker-0:10262:10330 [7] NCCL INFO NVLS multicast support is available on dev 7
+t-20260513223132-g9wrc-worker-0:10256:10331 [1] NCCL INFO NVLS multicast support is available on dev 1
+t-20260513223132-g9wrc-worker-0:10258:10333 [3] NCCL INFO comm 0xb5ef120 rank 3 nRanks 8 nNodes 1 localRanks 8 localRank 3 MNNVL 0
+t-20260513223132-g9wrc-worker-0:10257:10334 [2] NCCL INFO comm 0xb0afb80 rank 2 nRanks 8 nNodes 1 localRanks 8 localRank 2 MNNVL 0
+t-20260513223132-g9wrc-worker-0:10262:10330 [7] NCCL INFO comm 0xab21280 rank 7 nRanks 8 nNodes 1 localRanks 8 localRank 7 MNNVL 0
+t-20260513223132-g9wrc-worker-0:10255:10327 [0] NCCL INFO comm 0xb5c2750 rank 0 nRanks 8 nNodes 1 localRanks 8 localRank 0 MNNVL 0
+t-20260513223132-g9wrc-worker-0:10256:10331 [1] NCCL INFO comm 0x9e75e40 rank 1 nRanks 8 nNodes 1 localRanks 8 localRank 1 MNNVL 0
+t-20260513223132-g9wrc-worker-0:10261:10329 [6] NCCL INFO comm 0xb044c70 rank 6 nRanks 8 nNodes 1 localRanks 8 localRank 6 MNNVL 0
+t-20260513223132-g9wrc-worker-0:10260:10328 [5] NCCL INFO comm 0xa89c920 rank 5 nRanks 8 nNodes 1 localRanks 8 localRank 5 MNNVL 0
+t-20260513223132-g9wrc-worker-0:10259:10332 [4] NCCL INFO comm 0xb1d1a20 rank 4 nRanks 8 nNodes 1 localRanks 8 localRank 4 MNNVL 0
+t-20260513223132-g9wrc-worker-0:10255:10327 [0] NCCL INFO Channel 00/24 : 0 1 2 3 4 5 6 7
+t-20260513223132-g9wrc-worker-0:10255:10327 [0] NCCL INFO Channel 01/24 : 0 1 2 3 4 5 6 7
+t-20260513223132-g9wrc-worker-0:10255:10327 [0] NCCL INFO Channel 02/24 : 0 1 2 3 4 5 6 7
+t-20260513223132-g9wrc-worker-0:10255:10327 [0] NCCL INFO Channel 03/24 : 0 1 2 3 4 5 6 7
+t-20260513223132-g9wrc-worker-0:10255:10327 [0] NCCL INFO Channel 04/24 : 0 1 2 3 4 5 6 7
+t-20260513223132-g9wrc-worker-0:10255:10327 [0] NCCL INFO Channel 05/24 : 0 1 2 3 4 5 6 7
+t-20260513223132-g9wrc-worker-0:10262:10330 [7] NCCL INFO Trees [0] -1/-1/-1->7->6 [1] -1/-1/-1->7->6 [2] -1/-1/-1->7->6 [3] -1/-1/-1->7->6 [4] -1/-1/-1->7->6 [5] -1/-1/-1->7->6 [6] -1/-1/-1->7->6 [7] -1/-1/-1->7->6 [8] -1/-1/-1->7->6 [9] -1/-1/-1->7->6 [10] -1/-1/-1->7->6 [11] -1/-1/-1->7->6 [12] -1/-1/-1->7->6 [13] -1/-1/-1->7->6 [14] -1/-1/-1->7->6 [15] -1/-1/-1->7->6 [16] -1/-1/-1->7->6 [17] -1/-1/-1->7->6 [18] -1/-1/-1->7->6 [19] -1/-1/-1->7->6 [20] -1/-1/-1->7->6 [21] -1/-1/-1->7->6 [22] -1/-1/-1->7->6 [23] -1/-1/-1->7->6
+t-20260513223132-g9wrc-worker-0:10259:10332 [4] NCCL INFO Trees [0] 5/-1/-1->4->3 [1] 5/-1/-1->4->3 [2] 5/-1/-1->4->3 [3] 5/-1/-1->4->3 [4] 5/-1/-1->4->3 [5] 5/-1/-1->4->3 [6] 5/-1/-1->4->3 [7] 5/-1/-1->4->3 [8] 5/-1/-1->4->3 [9] 5/-1/-1->4->3 [10] 5/-1/-1->4->3 [11] 5/-1/-1->4->3 [12] 5/-1/-1->4->3 [13] 5/-1/-1->4->3 [14] 5/-1/-1->4->3 [15] 5/-1/-1->4->3 [16] 5/-1/-1->4->3 [17] 5/-1/-1->4->3 [18] 5/-1/-1->4->3 [19] 5/-1/-1->4->3 [20] 5/-1/-1->4->3 [21] 5/-1/-1->4->3 [22] 5/-1/-1->4->3 [23] 5/-1/-1->4->3
+t-20260513223132-g9wrc-worker-0:10255:10327 [0] NCCL INFO Channel 06/24 : 0 1 2 3 4 5 6 7
+t-20260513223132-g9wrc-worker-0:10262:10330 [7] NCCL INFO P2P Chunksize set to 524288
+t-20260513223132-g9wrc-worker-0:10259:10332 [4] NCCL INFO P2P Chunksize set to 524288
+t-20260513223132-g9wrc-worker-0:10258:10333 [3] NCCL INFO Trees [0] 4/-1/-1->3->2 [1] 4/-1/-1->3->2 [2] 4/-1/-1->3->2 [3] 4/-1/-1->3->2 [4] 4/-1/-1->3->2 [5] 4/-1/-1->3->2 [6] 4/-1/-1->3->2 [7] 4/-1/-1->3->2 [8] 4/-1/-1->3->2 [9] 4/-1/-1->3->2 [10] 4/-1/-1->3->2 [11] 4/-1/-1->3->2 [12] 4/-1/-1->3->2 [13] 4/-1/-1->3->2 [14] 4/-1/-1->3->2 [15] 4/-1/-1->3->2 [16] 4/-1/-1->3->2 [17] 4/-1/-1->3->2 [18] 4/-1/-1->3->2 [19] 4/-1/-1->3->2 [20] 4/-1/-1->3->2 [21] 4/-1/-1->3->2 [22] 4/-1/-1->3->2 [23] 4/-1/-1->3->2
+t-20260513223132-g9wrc-worker-0:10255:10327 [0] NCCL INFO Channel 07/24 : 0 1 2 3 4 5 6 7
+t-20260513223132-g9wrc-worker-0:10255:10327 [0] NCCL INFO Channel 08/24 : 0 1 2 3 4 5 6 7
+t-20260513223132-g9wrc-worker-0:10258:10333 [3] NCCL INFO P2P Chunksize set to 524288
+t-20260513223132-g9wrc-worker-0:10257:10334 [2] NCCL INFO Trees [0] 3/-1/-1->2->1 [1] 3/-1/-1->2->1 [2] 3/-1/-1->2->1 [3] 3/-1/-1->2->1 [4] 3/-1/-1->2->1 [5] 3/-1/-1->2->1 [6] 3/-1/-1->2->1 [7] 3/-1/-1->2->1 [8] 3/-1/-1->2->1 [9] 3/-1/-1->2->1 [10] 3/-1/-1->2->1 [11] 3/-1/-1->2->1 [12] 3/-1/-1->2->1 [13] 3/-1/-1->2->1 [14] 3/-1/-1->2->1 [15] 3/-1/-1->2->1 [16] 3/-1/-1->2->1 [17] 3/-1/-1->2->1 [18] 3/-1/-1->2->1 [19] 3/-1/-1->2->1 [20] 3/-1/-1->2->1 [21] 3/-1/-1->2->1 [22] 3/-1/-1->2->1 [23] 3/-1/-1->2->1
+t-20260513223132-g9wrc-worker-0:10256:10331 [1] NCCL INFO Trees [0] 2/-1/-1->1->0 [1] 2/-1/-1->1->0 [2] 2/-1/-1->1->0 [3] 2/-1/-1->1->0 [4] 2/-1/-1->1->0 [5] 2/-1/-1->1->0 [6] 2/-1/-1->1->0 [7] 2/-1/-1->1->0 [8] 2/-1/-1->1->0 [9] 2/-1/-1->1->0 [10] 2/-1/-1->1->0 [11] 2/-1/-1->1->0 [12] 2/-1/-1->1->0 [13] 2/-1/-1->1->0 [14] 2/-1/-1->1->0 [15] 2/-1/-1->1->0 [16] 2/-1/-1->1->0 [17] 2/-1/-1->1->0 [18] 2/-1/-1->1->0 [19] 2/-1/-1->1->0 [20] 2/-1/-1->1->0 [21] 2/-1/-1->1->0 [22] 2/-1/-1->1->0 [23] 2/-1/-1->1->0
+t-20260513223132-g9wrc-worker-0:10261:10329 [6] NCCL INFO Trees [0] 7/-1/-1->6->5 [1] 7/-1/-1->6->5 [2] 7/-1/-1->6->5 [3] 7/-1/-1->6->5 [4] 7/-1/-1->6->5 [5] 7/-1/-1->6->5 [6] 7/-1/-1->6->5 [7] 7/-1/-1->6->5 [8] 7/-1/-1->6->5 [9] 7/-1/-1->6->5 [10] 7/-1/-1->6->5 [11] 7/-1/-1->6->5 [12] 7/-1/-1->6->5 [13] 7/-1/-1->6->5 [14] 7/-1/-1->6->5 [15] 7/-1/-1->6->5 [16] 7/-1/-1->6->5 [17] 7/-1/-1->6->5 [18] 7/-1/-1->6->5 [19] 7/-1/-1->6->5 [20] 7/-1/-1->6->5 [21] 7/-1/-1->6->5 [22] 7/-1/-1->6->5 [23] 7/-1/-1->6->5
+t-20260513223132-g9wrc-worker-0:10255:10327 [0] NCCL INFO Channel 09/24 : 0 1 2 3 4 5 6 7
+t-20260513223132-g9wrc-worker-0:10260:10328 [5] NCCL INFO Trees [0] 6/-1/-1->5->4 [1] 6/-1/-1->5->4 [2] 6/-1/-1->5->4 [3] 6/-1/-1->5->4 [4] 6/-1/-1->5->4 [5] 6/-1/-1->5->4 [6] 6/-1/-1->5->4 [7] 6/-1/-1->5->4 [8] 6/-1/-1->5->4 [9] 6/-1/-1->5->4 [10] 6/-1/-1->5->4 [11] 6/-1/-1->5->4 [12] 6/-1/-1->5->4 [13] 6/-1/-1->5->4 [14] 6/-1/-1->5->4 [15] 6/-1/-1->5->4 [16] 6/-1/-1->5->4 [17] 6/-1/-1->5->4 [18] 6/-1/-1->5->4 [19] 6/-1/-1->5->4 [20] 6/-1/-1->5->4 [21] 6/-1/-1->5->4 [22] 6/-1/-1->5->4 [23] 6/-1/-1->5->4
+t-20260513223132-g9wrc-worker-0:10256:10331 [1] NCCL INFO P2P Chunksize set to 524288
+t-20260513223132-g9wrc-worker-0:10257:10334 [2] NCCL INFO P2P Chunksize set to 524288
+t-20260513223132-g9wrc-worker-0:10255:10327 [0] NCCL INFO Channel 10/24 : 0 1 2 3 4 5 6 7
+t-20260513223132-g9wrc-worker-0:10261:10329 [6] NCCL INFO P2P Chunksize set to 524288
+t-20260513223132-g9wrc-worker-0:10255:10327 [0] NCCL INFO Channel 11/24 : 0 1 2 3 4 5 6 7
+t-20260513223132-g9wrc-worker-0:10260:10328 [5] NCCL INFO P2P Chunksize set to 524288
+t-20260513223132-g9wrc-worker-0:10255:10327 [0] NCCL INFO Channel 12/24 : 0 1 2 3 4 5 6 7
+t-20260513223132-g9wrc-worker-0:10255:10327 [0] NCCL INFO Channel 13/24 : 0 1 2 3 4 5 6 7
+t-20260513223132-g9wrc-worker-0:10255:10327 [0] NCCL INFO Channel 14/24 : 0 1 2 3 4 5 6 7
+t-20260513223132-g9wrc-worker-0:10255:10327 [0] NCCL INFO Channel 15/24 : 0 1 2 3 4 5 6 7
+t-20260513223132-g9wrc-worker-0:10255:10327 [0] NCCL INFO Channel 16/24 : 0 1 2 3 4 5 6 7
+t-20260513223132-g9wrc-worker-0:10255:10327 [0] NCCL INFO Channel 17/24 : 0 1 2 3 4 5 6 7
+t-20260513223132-g9wrc-worker-0:10255:10327 [0] NCCL INFO Channel 18/24 : 0 1 2 3 4 5 6 7
+t-20260513223132-g9wrc-worker-0:10255:10327 [0] NCCL INFO Channel 19/24 : 0 1 2 3 4 5 6 7
+t-20260513223132-g9wrc-worker-0:10255:10327 [0] NCCL INFO Channel 20/24 : 0 1 2 3 4 5 6 7
+t-20260513223132-g9wrc-worker-0:10255:10327 [0] NCCL INFO Channel 21/24 : 0 1 2 3 4 5 6 7
+t-20260513223132-g9wrc-worker-0:10255:10327 [0] NCCL INFO Channel 22/24 : 0 1 2 3 4 5 6 7
+t-20260513223132-g9wrc-worker-0:10255:10327 [0] NCCL INFO Channel 23/24 : 0 1 2 3 4 5 6 7
+t-20260513223132-g9wrc-worker-0:10255:10327 [0] NCCL INFO Trees [0] 1/-1/-1->0->-1 [1] 1/-1/-1->0->-1 [2] 1/-1/-1->0->-1 [3] 1/-1/-1->0->-1 [4] 1/-1/-1->0->-1 [5] 1/-1/-1->0->-1 [6] 1/-1/-1->0->-1 [7] 1/-1/-1->0->-1 [8] 1/-1/-1->0->-1 [9] 1/-1/-1->0->-1 [10] 1/-1/-1->0->-1 [11] 1/-1/-1->0->-1 [12] 1/-1/-1->0->-1 [13] 1/-1/-1->0->-1 [14] 1/-1/-1->0->-1 [15] 1/-1/-1->0->-1 [16] 1/-1/-1->0->-1 [17] 1/-1/-1->0->-1 [18] 1/-1/-1->0->-1 [19] 1/-1/-1->0->-1 [20] 1/-1/-1->0->-1 [21] 1/-1/-1->0->-1 [22] 1/-1/-1->0->-1 [23] 1/-1/-1->0->-1
+t-20260513223132-g9wrc-worker-0:10255:10327 [0] NCCL INFO P2P Chunksize set to 524288
+t-20260513223132-g9wrc-worker-0:10257:10408 [2] NCCL INFO [Proxy Service UDS] Device 2 CPU core 4
+t-20260513223132-g9wrc-worker-0:10257:10407 [2] NCCL INFO [Proxy Service] Device 2 CPU core 2
+t-20260513223132-g9wrc-worker-0:10256:10409 [1] NCCL INFO [Proxy Service] Device 1 CPU core 2
+t-20260513223132-g9wrc-worker-0:10256:10410 [1] NCCL INFO [Proxy Service UDS] Device 1 CPU core 4
+t-20260513223132-g9wrc-worker-0:10261:10411 [6] NCCL INFO [Proxy Service] Device 6 CPU core 92
+t-20260513223132-g9wrc-worker-0:10261:10412 [6] NCCL INFO [Proxy Service UDS] Device 6 CPU core 96
+t-20260513223132-g9wrc-worker-0:10260:10414 [5] NCCL INFO [Proxy Service] Device 5 CPU core 92
+t-20260513223132-g9wrc-worker-0:10259:10415 [4] NCCL INFO [Proxy Service] Device 4 CPU core 96
+t-20260513223132-g9wrc-worker-0:10262:10413 [7] NCCL INFO [Proxy Service] Device 7 CPU core 106
+t-20260513223132-g9wrc-worker-0:10262:10416 [7] NCCL INFO [Proxy Service UDS] Device 7 CPU core 108
+t-20260513223132-g9wrc-worker-0:10259:10417 [4] NCCL INFO [Proxy Service UDS] Device 4 CPU core 98
+t-20260513223132-g9wrc-worker-0:10260:10419 [5] NCCL INFO [Proxy Service UDS] Device 5 CPU core 100
+t-20260513223132-g9wrc-worker-0:10258:10418 [3] NCCL INFO [Proxy Service] Device 3 CPU core 55
+t-20260513223132-g9wrc-worker-0:10258:10420 [3] NCCL INFO [Proxy Service UDS] Device 3 CPU core 56
+t-20260513223132-g9wrc-worker-0:10255:10327 [0] NCCL INFO Check P2P Type intraNodeP2pSupport 1 directMode 0
+t-20260513223132-g9wrc-worker-0:10255:10421 [0] NCCL INFO [Proxy Service] Device 0 CPU core 86
+t-20260513223132-g9wrc-worker-0:10255:10422 [0] NCCL INFO [Proxy Service UDS] Device 0 CPU core 88
+t-20260513223132-g9wrc-worker-0:10257:10334 [2] NCCL INFO threadThresholds 8/8/64 | 64/8/64 | 512 | 512
+t-20260513223132-g9wrc-worker-0:10257:10334 [2] NCCL INFO 24 coll channels, 24 collnet channels, 16 nvls channels, 32 p2p channels, 32 p2p channels per peer
+t-20260513223132-g9wrc-worker-0:10258:10333 [3] NCCL INFO threadThresholds 8/8/64 | 64/8/64 | 512 | 512
+t-20260513223132-g9wrc-worker-0:10258:10333 [3] NCCL INFO 24 coll channels, 24 collnet channels, 16 nvls channels, 32 p2p channels, 32 p2p channels per peer
+t-20260513223132-g9wrc-worker-0:10259:10332 [4] NCCL INFO threadThresholds 8/8/64 | 64/8/64 | 512 | 512
+t-20260513223132-g9wrc-worker-0:10259:10332 [4] NCCL INFO 24 coll channels, 24 collnet channels, 16 nvls channels, 32 p2p channels, 32 p2p channels per peer
+t-20260513223132-g9wrc-worker-0:10260:10328 [5] NCCL INFO threadThresholds 8/8/64 | 64/8/64 | 512 | 512
+t-20260513223132-g9wrc-worker-0:10260:10328 [5] NCCL INFO 24 coll channels, 24 collnet channels, 16 nvls channels, 32 p2p channels, 32 p2p channels per peer
+t-20260513223132-g9wrc-worker-0:10261:10329 [6] NCCL INFO threadThresholds 8/8/64 | 64/8/64 | 512 | 512
+t-20260513223132-g9wrc-worker-0:10261:10329 [6] NCCL INFO 24 coll channels, 24 collnet channels, 16 nvls channels, 32 p2p channels, 32 p2p channels per peer
+t-20260513223132-g9wrc-worker-0:10262:10330 [7] NCCL INFO threadThresholds 8/8/64 | 64/8/64 | 512 | 512
+t-20260513223132-g9wrc-worker-0:10255:10327 [0] NCCL INFO threadThresholds 8/8/64 | 64/8/64 | 512 | 512
+t-20260513223132-g9wrc-worker-0:10255:10327 [0] NCCL INFO 24 coll channels, 24 collnet channels, 16 nvls channels, 32 p2p channels, 32 p2p channels per peer
+t-20260513223132-g9wrc-worker-0:10262:10330 [7] NCCL INFO 24 coll channels, 24 collnet channels, 16 nvls channels, 32 p2p channels, 32 p2p channels per peer
+t-20260513223132-g9wrc-worker-0:10256:10331 [1] NCCL INFO threadThresholds 8/8/64 | 64/8/64 | 512 | 512
+t-20260513223132-g9wrc-worker-0:10256:10331 [1] NCCL INFO 24 coll channels, 24 collnet channels, 16 nvls channels, 32 p2p channels, 32 p2p channels per peer
+t-20260513223132-g9wrc-worker-0:10255:10327 [0] NCCL INFO CC Off, workFifoBytes 1048576
+t-20260513223132-g9wrc-worker-0:10257:10334 [2] NCCL INFO TUNER/Plugin: Failed to find ncclTunerPlugin_v4 symbol.
+t-20260513223132-g9wrc-worker-0:10256:10331 [1] NCCL INFO TUNER/Plugin: Failed to find ncclTunerPlugin_v4 symbol.
+t-20260513223132-g9wrc-worker-0:10260:10328 [5] NCCL INFO TUNER/Plugin: Failed to find ncclTunerPlugin_v4 symbol.
+t-20260513223132-g9wrc-worker-0:10262:10330 [7] NCCL INFO TUNER/Plugin: Failed to find ncclTunerPlugin_v4 symbol.
+t-20260513223132-g9wrc-worker-0:10257:10334 [2] NCCL INFO TUNER/Plugin: Failed to find ncclTunerPlugin_v3 symbol.
+t-20260513223132-g9wrc-worker-0:10256:10331 [1] NCCL INFO TUNER/Plugin: Failed to find ncclTunerPlugin_v3 symbol.
+t-20260513223132-g9wrc-worker-0:10260:10328 [5] NCCL INFO TUNER/Plugin: Failed to find ncclTunerPlugin_v3 symbol.
+t-20260513223132-g9wrc-worker-0:10259:10332 [4] NCCL INFO TUNER/Plugin: Failed to find ncclTunerPlugin_v4 symbol.
+t-20260513223132-g9wrc-worker-0:10257:10334 [2] NCCL INFO TUNER/Plugin: Failed to find ncclTunerPlugin_v2 symbol, using internal tuner instead.
+t-20260513223132-g9wrc-worker-0:10255:10327 [0] NCCL INFO TUNER/Plugin: Failed to find ncclTunerPlugin_v4 symbol.
+t-20260513223132-g9wrc-worker-0:10258:10333 [3] NCCL INFO TUNER/Plugin: Failed to find ncclTunerPlugin_v4 symbol.
+t-20260513223132-g9wrc-worker-0:10259:10332 [4] NCCL INFO TUNER/Plugin: Failed to find ncclTunerPlugin_v3 symbol.
+t-20260513223132-g9wrc-worker-0:10262:10330 [7] NCCL INFO TUNER/Plugin: Failed to find ncclTunerPlugin_v3 symbol.
+t-20260513223132-g9wrc-worker-0:10255:10327 [0] NCCL INFO TUNER/Plugin: Failed to find ncclTunerPlugin_v3 symbol.
+t-20260513223132-g9wrc-worker-0:10260:10328 [5] NCCL INFO TUNER/Plugin: Failed to find ncclTunerPlugin_v2 symbol, using internal tuner instead.
+t-20260513223132-g9wrc-worker-0:10256:10331 [1] NCCL INFO TUNER/Plugin: Failed to find ncclTunerPlugin_v2 symbol, using internal tuner instead.
+t-20260513223132-g9wrc-worker-0:10262:10330 [7] NCCL INFO TUNER/Plugin: Failed to find ncclTunerPlugin_v2 symbol, using internal tuner instead.
+t-20260513223132-g9wrc-worker-0:10261:10329 [6] NCCL INFO TUNER/Plugin: Failed to find ncclTunerPlugin_v4 symbol.
+t-20260513223132-g9wrc-worker-0:10256:10331 [1] NCCL INFO ncclCommInitRankConfig comm 0x9e75e40 rank 1 nranks 8 cudaDev 1 nvmlDev 1 busId 67020 commId 0x93a3726e358b8ce4 - Init COMPLETE
+t-20260513223132-g9wrc-worker-0:10261:10329 [6] NCCL INFO TUNER/Plugin: Failed to find ncclTunerPlugin_v3 symbol.
+t-20260513223132-g9wrc-worker-0:10261:10329 [6] NCCL INFO TUNER/Plugin: Failed to find ncclTunerPlugin_v2 symbol, using internal tuner instead.
+t-20260513223132-g9wrc-worker-0:10257:10334 [2] NCCL INFO ncclCommInitRankConfig comm 0xb0afb80 rank 2 nranks 8 cudaDev 2 nvmlDev 2 busId 69020 commId 0x93a3726e358b8ce4 - Init COMPLETE
+t-20260513223132-g9wrc-worker-0:10259:10332 [4] NCCL INFO TUNER/Plugin: Failed to find ncclTunerPlugin_v2 symbol, using internal tuner instead.
+t-20260513223132-g9wrc-worker-0:10258:10333 [3] NCCL INFO TUNER/Plugin: Failed to find ncclTunerPlugin_v3 symbol.
+t-20260513223132-g9wrc-worker-0:10261:10329 [6] NCCL INFO ncclCommInitRankConfig comm 0xb044c70 rank 6 nranks 8 cudaDev 6 nvmlDev 6 busId 73020 commId 0x93a3726e358b8ce4 - Init COMPLETE
+t-20260513223132-g9wrc-worker-0:10255:10327 [0] NCCL INFO TUNER/Plugin: Failed to find ncclTunerPlugin_v2 symbol, using internal tuner instead.
+t-20260513223132-g9wrc-worker-0:10260:10328 [5] NCCL INFO ncclCommInitRankConfig comm 0xa89c920 rank 5 nranks 8 cudaDev 5 nvmlDev 5 busId 71020 commId 0x93a3726e358b8ce4 - Init COMPLETE
+t-20260513223132-g9wrc-worker-0:10262:10330 [7] NCCL INFO ncclCommInitRankConfig comm 0xab21280 rank 7 nranks 8 cudaDev 7 nvmlDev 7 busId 75020 commId 0x93a3726e358b8ce4 - Init COMPLETE
+t-20260513223132-g9wrc-worker-0:10256:10331 [1] NCCL INFO Init timings - ncclCommInitRankConfig: rank 1 nranks 8 total 2.03 (kernels 0.18, alloc 0.92, bootstrap 0.07, allgathers 0.00, topo 0.55, graphs 0.01, connections 0.28, rest 0.02)
+t-20260513223132-g9wrc-worker-0:10259:10332 [4] NCCL INFO ncclCommInitRankConfig comm 0xb1d1a20 rank 4 nranks 8 cudaDev 4 nvmlDev 4 busId 6f020 commId 0x93a3726e358b8ce4 - Init COMPLETE
+t-20260513223132-g9wrc-worker-0:10257:10334 [2] NCCL INFO Init timings - ncclCommInitRankConfig: rank 2 nranks 8 total 2.01 (kernels 0.30, alloc 0.84, bootstrap 0.01, allgathers 0.01, topo 0.55, graphs 0.01, connections 0.28, rest 0.02)
+t-20260513223132-g9wrc-worker-0:10258:10333 [3] NCCL INFO TUNER/Plugin: Failed to find ncclTunerPlugin_v2 symbol, using internal tuner instead.
+t-20260513223132-g9wrc-worker-0:10255:10327 [0] NCCL INFO ncclCommInitRankConfig comm 0xb5c2750 rank 0 nranks 8 cudaDev 0 nvmlDev 0 busId 65040 commId 0x93a3726e358b8ce4 - Init COMPLETE
+t-20260513223132-g9wrc-worker-0:10261:10329 [6] NCCL INFO Init timings - ncclCommInitRankConfig: rank 6 nranks 8 total 2.04 (kernels 0.18, alloc 0.91, bootstrap 0.09, allgathers 0.01, topo 0.55, graphs 0.01, connections 0.27, rest 0.02)
+t-20260513223132-g9wrc-worker-0:10260:10328 [5] NCCL INFO Init timings - ncclCommInitRankConfig: rank 5 nranks 8 total 2.06 (kernels 0.18, alloc 0.88, bootstrap 0.14, allgathers 0.00, topo 0.55, graphs 0.01, connections 0.26, rest 0.03)
+t-20260513223132-g9wrc-worker-0:10262:10330 [7] NCCL INFO Init timings - ncclCommInitRankConfig: rank 7 nranks 8 total 2.03 (kernels 0.18, alloc 0.92, bootstrap 0.07, allgathers 0.01, topo 0.55, graphs 0.01, connections 0.26, rest 0.03)
+t-20260513223132-g9wrc-worker-0:10258:10333 [3] NCCL INFO ncclCommInitRankConfig comm 0xb5ef120 rank 3 nranks 8 cudaDev 3 nvmlDev 3 busId 6b020 commId 0x93a3726e358b8ce4 - Init COMPLETE
+t-20260513223132-g9wrc-worker-0:10259:10332 [4] NCCL INFO Init timings - ncclCommInitRankConfig: rank 4 nranks 8 total 2.02 (kernels 0.39, alloc 0.77, bootstrap 0.00, allgathers 0.01, topo 0.55, graphs 0.01, connections 0.26, rest 0.03)
+t-20260513223132-g9wrc-worker-0:10255:10327 [0] NCCL INFO Init timings - ncclCommInitRankConfig: rank 0 nranks 8 total 2.09 (kernels 0.18, alloc 0.71, bootstrap 0.35, allgathers 0.01, topo 0.55, graphs 0.01, connections 0.26, rest 0.03)
+t-20260513223132-g9wrc-worker-0:10258:10333 [3] NCCL INFO Init timings - ncclCommInitRankConfig: rank 3 nranks 8 total 2.01 (kernels 0.23, alloc 0.90, bootstrap 0.03, allgathers 0.01, topo 0.55, graphs 0.01, connections 0.26, rest 0.03)
+t-20260513223132-g9wrc-worker-0:10261:10423 [6] NCCL INFO Channel 00/0 : 6[6] -> 7[7] via P2P/CUMEM
+t-20260513223132-g9wrc-worker-0:10259:10424 [4] NCCL INFO Channel 00/0 : 4[4] -> 5[5] via P2P/CUMEM
+t-20260513223132-g9wrc-worker-0:10261:10423 [6] NCCL INFO Channel 01/0 : 6[6] -> 7[7] via P2P/CUMEM
+t-20260513223132-g9wrc-worker-0:10259:10424 [4] NCCL INFO Channel 01/0 : 4[4] -> 5[5] via P2P/CUMEM
+t-20260513223132-g9wrc-worker-0:10261:10423 [6] NCCL INFO Channel 02/0 : 6[6] -> 7[7] via P2P/CUMEM
+t-20260513223132-g9wrc-worker-0:10258:10426 [3] NCCL INFO Channel 00/0 : 3[3] -> 4[4] via P2P/CUMEM
+t-20260513223132-g9wrc-worker-0:10259:10424 [4] NCCL INFO Channel 02/0 : 4[4] -> 5[5] via P2P/CUMEM
+t-20260513223132-g9wrc-worker-0:10261:10423 [6] NCCL INFO Channel 03/0 : 6[6] -> 7[7] via P2P/CUMEM
+t-20260513223132-g9wrc-worker-0:10255:10425 [0] NCCL INFO Channel 00/0 : 0[0] -> 1[1] via P2P/CUMEM
+t-20260513223132-g9wrc-worker-0:10258:10426 [3] NCCL INFO Channel 01/0 : 3[3] -> 4[4] via P2P/CUMEM
+t-20260513223132-g9wrc-worker-0:10259:10424 [4] NCCL INFO Channel 03/0 : 4[4] -> 5[5] via P2P/CUMEM
+t-20260513223132-g9wrc-worker-0:10261:10423 [6] NCCL INFO Channel 04/0 : 6[6] -> 7[7] via P2P/CUMEM
+t-20260513223132-g9wrc-worker-0:10255:10425 [0] NCCL INFO Channel 01/0 : 0[0] -> 1[1] via P2P/CUMEM
+t-20260513223132-g9wrc-worker-0:10258:10426 [3] NCCL INFO Channel 02/0 : 3[3] -> 4[4] via P2P/CUMEM
+t-20260513223132-g9wrc-worker-0:10259:10424 [4] NCCL INFO Channel 04/0 : 4[4] -> 5[5] via P2P/CUMEM
+t-20260513223132-g9wrc-worker-0:10261:10423 [6] NCCL INFO Channel 05/0 : 6[6] -> 7[7] via P2P/CUMEM
+t-20260513223132-g9wrc-worker-0:10255:10425 [0] NCCL INFO Channel 02/0 : 0[0] -> 1[1] via P2P/CUMEM
+t-20260513223132-g9wrc-worker-0:10258:10426 [3] NCCL INFO Channel 03/0 : 3[3] -> 4[4] via P2P/CUMEM
+t-20260513223132-g9wrc-worker-0:10259:10424 [4] NCCL INFO Channel 05/0 : 4[4] -> 5[5] via P2P/CUMEM
+t-20260513223132-g9wrc-worker-0:10257:10428 [2] NCCL INFO Channel 00/0 : 2[2] -> 3[3] via P2P/CUMEM
+t-20260513223132-g9wrc-worker-0:10261:10423 [6] NCCL INFO Channel 06/0 : 6[6] -> 7[7] via P2P/CUMEM
+t-20260513223132-g9wrc-worker-0:10255:10425 [0] NCCL INFO Channel 03/0 : 0[0] -> 1[1] via P2P/CUMEM
+t-20260513223132-g9wrc-worker-0:10258:10426 [3] NCCL INFO Channel 04/0 : 3[3] -> 4[4] via P2P/CUMEM
+t-20260513223132-g9wrc-worker-0:10259:10424 [4] NCCL INFO Channel 06/0 : 4[4] -> 5[5] via P2P/CUMEM
+t-20260513223132-g9wrc-worker-0:10257:10428 [2] NCCL INFO Channel 01/0 : 2[2] -> 3[3] via P2P/CUMEM
+t-20260513223132-g9wrc-worker-0:10261:10423 [6] NCCL INFO Channel 07/0 : 6[6] -> 7[7] via P2P/CUMEM
+t-20260513223132-g9wrc-worker-0:10255:10425 [0] NCCL INFO Channel 04/0 : 0[0] -> 1[1] via P2P/CUMEM
+t-20260513223132-g9wrc-worker-0:10260:10429 [5] NCCL INFO Channel 00/0 : 5[5] -> 6[6] via P2P/CUMEM
+t-20260513223132-g9wrc-worker-0:10258:10426 [3] NCCL INFO Channel 05/0 : 3[3] -> 4[4] via P2P/CUMEM
+t-20260513223132-g9wrc-worker-0:10256:10427 [1] NCCL INFO Channel 00/0 : 1[1] -> 2[2] via P2P/CUMEM
+t-20260513223132-g9wrc-worker-0:10259:10424 [4] NCCL INFO Channel 07/0 : 4[4] -> 5[5] via P2P/CUMEM
+t-20260513223132-g9wrc-worker-0:10257:10428 [2] NCCL INFO Channel 02/0 : 2[2] -> 3[3] via P2P/CUMEM
+t-20260513223132-g9wrc-worker-0:10261:10423 [6] NCCL INFO Channel 08/0 : 6[6] -> 7[7] via P2P/CUMEM
+t-20260513223132-g9wrc-worker-0:10255:10425 [0] NCCL INFO Channel 05/0 : 0[0] -> 1[1] via P2P/CUMEM
+t-20260513223132-g9wrc-worker-0:10260:10429 [5] NCCL INFO Channel 01/0 : 5[5] -> 6[6] via P2P/CUMEM
+t-20260513223132-g9wrc-worker-0:10258:10426 [3] NCCL INFO Channel 06/0 : 3[3] -> 4[4] via P2P/CUMEM
+t-20260513223132-g9wrc-worker-0:10256:10427 [1] NCCL INFO Channel 01/0 : 1[1] -> 2[2] via P2P/CUMEM
+t-20260513223132-g9wrc-worker-0:10262:10430 [7] NCCL INFO Channel 00/0 : 7[7] -> 0[0] via P2P/CUMEM
+t-20260513223132-g9wrc-worker-0:10259:10424 [4] NCCL INFO Channel 08/0 : 4[4] -> 5[5] via P2P/CUMEM
+t-20260513223132-g9wrc-worker-0:10257:10428 [2] NCCL INFO Channel 03/0 : 2[2] -> 3[3] via P2P/CUMEM
+t-20260513223132-g9wrc-worker-0:10261:10423 [6] NCCL INFO Channel 09/0 : 6[6] -> 7[7] via P2P/CUMEM
+t-20260513223132-g9wrc-worker-0:10255:10425 [0] NCCL INFO Channel 06/0 : 0[0] -> 1[1] via P2P/CUMEM
+t-20260513223132-g9wrc-worker-0:10260:10429 [5] NCCL INFO Channel 02/0 : 5[5] -> 6[6] via P2P/CUMEM
+t-20260513223132-g9wrc-worker-0:10258:10426 [3] NCCL INFO Channel 07/0 : 3[3] -> 4[4] via P2P/CUMEM
+t-20260513223132-g9wrc-worker-0:10256:10427 [1] NCCL INFO Channel 02/0 : 1[1] -> 2[2] via P2P/CUMEM
+t-20260513223132-g9wrc-worker-0:10262:10430 [7] NCCL INFO Channel 01/0 : 7[7] -> 0[0] via P2P/CUMEM
+t-20260513223132-g9wrc-worker-0:10259:10424 [4] NCCL INFO Channel 09/0 : 4[4] -> 5[5] via P2P/CUMEM
+t-20260513223132-g9wrc-worker-0:10257:10428 [2] NCCL INFO Channel 04/0 : 2[2] -> 3[3] via P2P/CUMEM
+t-20260513223132-g9wrc-worker-0:10261:10423 [6] NCCL INFO Channel 10/0 : 6[6] -> 7[7] via P2P/CUMEM
+t-20260513223132-g9wrc-worker-0:10255:10425 [0] NCCL INFO Channel 07/0 : 0[0] -> 1[1] via P2P/CUMEM
+t-20260513223132-g9wrc-worker-0:10260:10429 [5] NCCL INFO Channel 03/0 : 5[5] -> 6[6] via P2P/CUMEM
+t-20260513223132-g9wrc-worker-0:10258:10426 [3] NCCL INFO Channel 08/0 : 3[3] -> 4[4] via P2P/CUMEM
+t-20260513223132-g9wrc-worker-0:10256:10427 [1] NCCL INFO Channel 03/0 : 1[1] -> 2[2] via P2P/CUMEM
+t-20260513223132-g9wrc-worker-0:10262:10430 [7] NCCL INFO Channel 02/0 : 7[7] -> 0[0] via P2P/CUMEM
+t-20260513223132-g9wrc-worker-0:10259:10424 [4] NCCL INFO Channel 10/0 : 4[4] -> 5[5] via P2P/CUMEM
+t-20260513223132-g9wrc-worker-0:10257:10428 [2] NCCL INFO Channel 05/0 : 2[2] -> 3[3] via P2P/CUMEM
+t-20260513223132-g9wrc-worker-0:10261:10423 [6] NCCL INFO Channel 11/0 : 6[6] -> 7[7] via P2P/CUMEM
+t-20260513223132-g9wrc-worker-0:10255:10425 [0] NCCL INFO Channel 08/0 : 0[0] -> 1[1] via P2P/CUMEM
+t-20260513223132-g9wrc-worker-0:10260:10429 [5] NCCL INFO Channel 04/0 : 5[5] -> 6[6] via P2P/CUMEM
+t-20260513223132-g9wrc-worker-0:10258:10426 [3] NCCL INFO Channel 09/0 : 3[3] -> 4[4] via P2P/CUMEM
+t-20260513223132-g9wrc-worker-0:10256:10427 [1] NCCL INFO Channel 04/0 : 1[1] -> 2[2] via P2P/CUMEM
+t-20260513223132-g9wrc-worker-0:10262:10430 [7] NCCL INFO Channel 03/0 : 7[7] -> 0[0] via P2P/CUMEM
+t-20260513223132-g9wrc-worker-0:10259:10424 [4] NCCL INFO Channel 11/0 : 4[4] -> 5[5] via P2P/CUMEM
+t-20260513223132-g9wrc-worker-0:10257:10428 [2] NCCL INFO Channel 06/0 : 2[2] -> 3[3] via P2P/CUMEM
+t-20260513223132-g9wrc-worker-0:10261:10423 [6] NCCL INFO Channel 12/0 : 6[6] -> 7[7] via P2P/CUMEM
+t-20260513223132-g9wrc-worker-0:10255:10425 [0] NCCL INFO Channel 09/0 : 0[0] -> 1[1] via P2P/CUMEM
+t-20260513223132-g9wrc-worker-0:10260:10429 [5] NCCL INFO Channel 05/0 : 5[5] -> 6[6] via P2P/CUMEM
+t-20260513223132-g9wrc-worker-0:10258:10426 [3] NCCL INFO Channel 10/0 : 3[3] -> 4[4] via P2P/CUMEM
+t-20260513223132-g9wrc-worker-0:10256:10427 [1] NCCL INFO Channel 05/0 : 1[1] -> 2[2] via P2P/CUMEM
+t-20260513223132-g9wrc-worker-0:10262:10430 [7] NCCL INFO Channel 04/0 : 7[7] -> 0[0] via P2P/CUMEM
+t-20260513223132-g9wrc-worker-0:10259:10424 [4] NCCL INFO Channel 12/0 : 4[4] -> 5[5] via P2P/CUMEM
+t-20260513223132-g9wrc-worker-0:10257:10428 [2] NCCL INFO Channel 07/0 : 2[2] -> 3[3] via P2P/CUMEM
+t-20260513223132-g9wrc-worker-0:10261:10423 [6] NCCL INFO Channel 13/0 : 6[6] -> 7[7] via P2P/CUMEM
+t-20260513223132-g9wrc-worker-0:10255:10425 [0] NCCL INFO Channel 10/0 : 0[0] -> 1[1] via P2P/CUMEM
+t-20260513223132-g9wrc-worker-0:10260:10429 [5] NCCL INFO Channel 06/0 : 5[5] -> 6[6] via P2P/CUMEM
+t-20260513223132-g9wrc-worker-0:10258:10426 [3] NCCL INFO Channel 11/0 : 3[3] -> 4[4] via P2P/CUMEM
+t-20260513223132-g9wrc-worker-0:10256:10427 [1] NCCL INFO Channel 06/0 : 1[1] -> 2[2] via P2P/CUMEM
+t-20260513223132-g9wrc-worker-0:10262:10430 [7] NCCL INFO Channel 05/0 : 7[7] -> 0[0] via P2P/CUMEM
+t-20260513223132-g9wrc-worker-0:10259:10424 [4] NCCL INFO Channel 13/0 : 4[4] -> 5[5] via P2P/CUMEM
+t-20260513223132-g9wrc-worker-0:10257:10428 [2] NCCL INFO Channel 08/0 : 2[2] -> 3[3] via P2P/CUMEM
+t-20260513223132-g9wrc-worker-0:10261:10423 [6] NCCL INFO Channel 14/0 : 6[6] -> 7[7] via P2P/CUMEM
+t-20260513223132-g9wrc-worker-0:10255:10425 [0] NCCL INFO Channel 11/0 : 0[0] -> 1[1] via P2P/CUMEM
+t-20260513223132-g9wrc-worker-0:10260:10429 [5] NCCL INFO Channel 07/0 : 5[5] -> 6[6] via P2P/CUMEM
+t-20260513223132-g9wrc-worker-0:10258:10426 [3] NCCL INFO Channel 12/0 : 3[3] -> 4[4] via P2P/CUMEM
+t-20260513223132-g9wrc-worker-0:10256:10427 [1] NCCL INFO Channel 07/0 : 1[1] -> 2[2] via P2P/CUMEM
+t-20260513223132-g9wrc-worker-0:10262:10430 [7] NCCL INFO Channel 06/0 : 7[7] -> 0[0] via P2P/CUMEM
+t-20260513223132-g9wrc-worker-0:10259:10424 [4] NCCL INFO Channel 14/0 : 4[4] -> 5[5] via P2P/CUMEM
+t-20260513223132-g9wrc-worker-0:10257:10428 [2] NCCL INFO Channel 09/0 : 2[2] -> 3[3] via P2P/CUMEM
+t-20260513223132-g9wrc-worker-0:10261:10423 [6] NCCL INFO Channel 15/0 : 6[6] -> 7[7] via P2P/CUMEM
+t-20260513223132-g9wrc-worker-0:10255:10425 [0] NCCL INFO Channel 12/0 : 0[0] -> 1[1] via P2P/CUMEM
+t-20260513223132-g9wrc-worker-0:10260:10429 [5] NCCL INFO Channel 08/0 : 5[5] -> 6[6] via P2P/CUMEM
+t-20260513223132-g9wrc-worker-0:10258:10426 [3] NCCL INFO Channel 13/0 : 3[3] -> 4[4] via P2P/CUMEM
+t-20260513223132-g9wrc-worker-0:10256:10427 [1] NCCL INFO Channel 08/0 : 1[1] -> 2[2] via P2P/CUMEM
+t-20260513223132-g9wrc-worker-0:10262:10430 [7] NCCL INFO Channel 07/0 : 7[7] -> 0[0] via P2P/CUMEM
+t-20260513223132-g9wrc-worker-0:10259:10424 [4] NCCL INFO Channel 15/0 : 4[4] -> 5[5] via P2P/CUMEM
+t-20260513223132-g9wrc-worker-0:10257:10428 [2] NCCL INFO Channel 10/0 : 2[2] -> 3[3] via P2P/CUMEM
+t-20260513223132-g9wrc-worker-0:10261:10423 [6] NCCL INFO Channel 16/0 : 6[6] -> 7[7] via P2P/CUMEM
+t-20260513223132-g9wrc-worker-0:10255:10425 [0] NCCL INFO Channel 13/0 : 0[0] -> 1[1] via P2P/CUMEM
+t-20260513223132-g9wrc-worker-0:10260:10429 [5] NCCL INFO Channel 09/0 : 5[5] -> 6[6] via P2P/CUMEM
+t-20260513223132-g9wrc-worker-0:10258:10426 [3] NCCL INFO Channel 14/0 : 3[3] -> 4[4] via P2P/CUMEM
+t-20260513223132-g9wrc-worker-0:10256:10427 [1] NCCL INFO Channel 09/0 : 1[1] -> 2[2] via P2P/CUMEM
+t-20260513223132-g9wrc-worker-0:10262:10430 [7] NCCL INFO Channel 08/0 : 7[7] -> 0[0] via P2P/CUMEM
+t-20260513223132-g9wrc-worker-0:10259:10424 [4] NCCL INFO Channel 16/0 : 4[4] -> 5[5] via P2P/CUMEM
+t-20260513223132-g9wrc-worker-0:10257:10428 [2] NCCL INFO Channel 11/0 : 2[2] -> 3[3] via P2P/CUMEM
+t-20260513223132-g9wrc-worker-0:10261:10423 [6] NCCL INFO Channel 17/0 : 6[6] -> 7[7] via P2P/CUMEM
+t-20260513223132-g9wrc-worker-0:10255:10425 [0] NCCL INFO Channel 14/0 : 0[0] -> 1[1] via P2P/CUMEM
+t-20260513223132-g9wrc-worker-0:10260:10429 [5] NCCL INFO Channel 10/0 : 5[5] -> 6[6] via P2P/CUMEM
+t-20260513223132-g9wrc-worker-0:10258:10426 [3] NCCL INFO Channel 15/0 : 3[3] -> 4[4] via P2P/CUMEM
+t-20260513223132-g9wrc-worker-0:10256:10427 [1] NCCL INFO Channel 10/0 : 1[1] -> 2[2] via P2P/CUMEM
+t-20260513223132-g9wrc-worker-0:10262:10430 [7] NCCL INFO Channel 09/0 : 7[7] -> 0[0] via P2P/CUMEM
+t-20260513223132-g9wrc-worker-0:10259:10424 [4] NCCL INFO Channel 17/0 : 4[4] -> 5[5] via P2P/CUMEM
+t-20260513223132-g9wrc-worker-0:10257:10428 [2] NCCL INFO Channel 12/0 : 2[2] -> 3[3] via P2P/CUMEM
+t-20260513223132-g9wrc-worker-0:10261:10423 [6] NCCL INFO Channel 18/0 : 6[6] -> 7[7] via P2P/CUMEM
+t-20260513223132-g9wrc-worker-0:10255:10425 [0] NCCL INFO Channel 15/0 : 0[0] -> 1[1] via P2P/CUMEM
+t-20260513223132-g9wrc-worker-0:10260:10429 [5] NCCL INFO Channel 11/0 : 5[5] -> 6[6] via P2P/CUMEM
+t-20260513223132-g9wrc-worker-0:10258:10426 [3] NCCL INFO Channel 16/0 : 3[3] -> 4[4] via P2P/CUMEM
+t-20260513223132-g9wrc-worker-0:10256:10427 [1] NCCL INFO Channel 11/0 : 1[1] -> 2[2] via P2P/CUMEM
+t-20260513223132-g9wrc-worker-0:10262:10430 [7] NCCL INFO Channel 10/0 : 7[7] -> 0[0] via P2P/CUMEM
+t-20260513223132-g9wrc-worker-0:10259:10424 [4] NCCL INFO Channel 18/0 : 4[4] -> 5[5] via P2P/CUMEM
+t-20260513223132-g9wrc-worker-0:10257:10428 [2] NCCL INFO Channel 13/0 : 2[2] -> 3[3] via P2P/CUMEM
+t-20260513223132-g9wrc-worker-0:10261:10423 [6] NCCL INFO Channel 19/0 : 6[6] -> 7[7] via P2P/CUMEM
+t-20260513223132-g9wrc-worker-0:10255:10425 [0] NCCL INFO Channel 16/0 : 0[0] -> 1[1] via P2P/CUMEM
+t-20260513223132-g9wrc-worker-0:10260:10429 [5] NCCL INFO Channel 12/0 : 5[5] -> 6[6] via P2P/CUMEM
+t-20260513223132-g9wrc-worker-0:10258:10426 [3] NCCL INFO Channel 17/0 : 3[3] -> 4[4] via P2P/CUMEM
+t-20260513223132-g9wrc-worker-0:10256:10427 [1] NCCL INFO Channel 12/0 : 1[1] -> 2[2] via P2P/CUMEM
+t-20260513223132-g9wrc-worker-0:10262:10430 [7] NCCL INFO Channel 11/0 : 7[7] -> 0[0] via P2P/CUMEM
+t-20260513223132-g9wrc-worker-0:10259:10424 [4] NCCL INFO Channel 19/0 : 4[4] -> 5[5] via P2P/CUMEM
+t-20260513223132-g9wrc-worker-0:10257:10428 [2] NCCL INFO Channel 14/0 : 2[2] -> 3[3] via P2P/CUMEM
+t-20260513223132-g9wrc-worker-0:10261:10423 [6] NCCL INFO Channel 20/0 : 6[6] -> 7[7] via P2P/CUMEM
+t-20260513223132-g9wrc-worker-0:10255:10425 [0] NCCL INFO Channel 17/0 : 0[0] -> 1[1] via P2P/CUMEM
+t-20260513223132-g9wrc-worker-0:10260:10429 [5] NCCL INFO Channel 13/0 : 5[5] -> 6[6] via P2P/CUMEM
+t-20260513223132-g9wrc-worker-0:10258:10426 [3] NCCL INFO Channel 18/0 : 3[3] -> 4[4] via P2P/CUMEM
+t-20260513223132-g9wrc-worker-0:10256:10427 [1] NCCL INFO Channel 13/0 : 1[1] -> 2[2] via P2P/CUMEM
+t-20260513223132-g9wrc-worker-0:10262:10430 [7] NCCL INFO Channel 12/0 : 7[7] -> 0[0] via P2P/CUMEM
+t-20260513223132-g9wrc-worker-0:10259:10424 [4] NCCL INFO Channel 20/0 : 4[4] -> 5[5] via P2P/CUMEM
+t-20260513223132-g9wrc-worker-0:10257:10428 [2] NCCL INFO Channel 15/0 : 2[2] -> 3[3] via P2P/CUMEM
+t-20260513223132-g9wrc-worker-0:10261:10423 [6] NCCL INFO Channel 21/0 : 6[6] -> 7[7] via P2P/CUMEM
+t-20260513223132-g9wrc-worker-0:10255:10425 [0] NCCL INFO Channel 18/0 : 0[0] -> 1[1] via P2P/CUMEM
+t-20260513223132-g9wrc-worker-0:10260:10429 [5] NCCL INFO Channel 14/0 : 5[5] -> 6[6] via P2P/CUMEM
+t-20260513223132-g9wrc-worker-0:10258:10426 [3] NCCL INFO Channel 19/0 : 3[3] -> 4[4] via P2P/CUMEM
+t-20260513223132-g9wrc-worker-0:10256:10427 [1] NCCL INFO Channel 14/0 : 1[1] -> 2[2] via P2P/CUMEM
+t-20260513223132-g9wrc-worker-0:10262:10430 [7] NCCL INFO Channel 13/0 : 7[7] -> 0[0] via P2P/CUMEM
+t-20260513223132-g9wrc-worker-0:10259:10424 [4] NCCL INFO Channel 21/0 : 4[4] -> 5[5] via P2P/CUMEM
+t-20260513223132-g9wrc-worker-0:10257:10428 [2] NCCL INFO Channel 16/0 : 2[2] -> 3[3] via P2P/CUMEM
+t-20260513223132-g9wrc-worker-0:10261:10423 [6] NCCL INFO Channel 22/0 : 6[6] -> 7[7] via P2P/CUMEM
+t-20260513223132-g9wrc-worker-0:10255:10425 [0] NCCL INFO Channel 19/0 : 0[0] -> 1[1] via P2P/CUMEM
+t-20260513223132-g9wrc-worker-0:10260:10429 [5] NCCL INFO Channel 15/0 : 5[5] -> 6[6] via P2P/CUMEM
+t-20260513223132-g9wrc-worker-0:10258:10426 [3] NCCL INFO Channel 20/0 : 3[3] -> 4[4] via P2P/CUMEM
+t-20260513223132-g9wrc-worker-0:10256:10427 [1] NCCL INFO Channel 15/0 : 1[1] -> 2[2] via P2P/CUMEM
+t-20260513223132-g9wrc-worker-0:10262:10430 [7] NCCL INFO Channel 14/0 : 7[7] -> 0[0] via P2P/CUMEM
+t-20260513223132-g9wrc-worker-0:10259:10424 [4] NCCL INFO Channel 22/0 : 4[4] -> 5[5] via P2P/CUMEM
+t-20260513223132-g9wrc-worker-0:10257:10428 [2] NCCL INFO Channel 17/0 : 2[2] -> 3[3] via P2P/CUMEM
+t-20260513223132-g9wrc-worker-0:10261:10423 [6] NCCL INFO Channel 23/0 : 6[6] -> 7[7] via P2P/CUMEM
+t-20260513223132-g9wrc-worker-0:10255:10425 [0] NCCL INFO Channel 20/0 : 0[0] -> 1[1] via P2P/CUMEM
+t-20260513223132-g9wrc-worker-0:10260:10429 [5] NCCL INFO Channel 16/0 : 5[5] -> 6[6] via P2P/CUMEM
+t-20260513223132-g9wrc-worker-0:10258:10426 [3] NCCL INFO Channel 21/0 : 3[3] -> 4[4] via P2P/CUMEM
+t-20260513223132-g9wrc-worker-0:10256:10427 [1] NCCL INFO Channel 16/0 : 1[1] -> 2[2] via P2P/CUMEM
+t-20260513223132-g9wrc-worker-0:10262:10430 [7] NCCL INFO Channel 15/0 : 7[7] -> 0[0] via P2P/CUMEM
+t-20260513223132-g9wrc-worker-0:10259:10424 [4] NCCL INFO Channel 23/0 : 4[4] -> 5[5] via P2P/CUMEM
+t-20260513223132-g9wrc-worker-0:10257:10428 [2] NCCL INFO Channel 18/0 : 2[2] -> 3[3] via P2P/CUMEM
+t-20260513223132-g9wrc-worker-0:10255:10425 [0] NCCL INFO Channel 21/0 : 0[0] -> 1[1] via P2P/CUMEM
+t-20260513223132-g9wrc-worker-0:10260:10429 [5] NCCL INFO Channel 17/0 : 5[5] -> 6[6] via P2P/CUMEM
+t-20260513223132-g9wrc-worker-0:10258:10426 [3] NCCL INFO Channel 22/0 : 3[3] -> 4[4] via P2P/CUMEM
+t-20260513223132-g9wrc-worker-0:10256:10427 [1] NCCL INFO Channel 17/0 : 1[1] -> 2[2] via P2P/CUMEM
+t-20260513223132-g9wrc-worker-0:10262:10430 [7] NCCL INFO Channel 16/0 : 7[7] -> 0[0] via P2P/CUMEM
+t-20260513223132-g9wrc-worker-0:10257:10428 [2] NCCL INFO Channel 19/0 : 2[2] -> 3[3] via P2P/CUMEM
+t-20260513223132-g9wrc-worker-0:10255:10425 [0] NCCL INFO Channel 22/0 : 0[0] -> 1[1] via P2P/CUMEM
+t-20260513223132-g9wrc-worker-0:10260:10429 [5] NCCL INFO Channel 18/0 : 5[5] -> 6[6] via P2P/CUMEM
+t-20260513223132-g9wrc-worker-0:10258:10426 [3] NCCL INFO Channel 23/0 : 3[3] -> 4[4] via P2P/CUMEM
+t-20260513223132-g9wrc-worker-0:10256:10427 [1] NCCL INFO Channel 18/0 : 1[1] -> 2[2] via P2P/CUMEM
+t-20260513223132-g9wrc-worker-0:10262:10430 [7] NCCL INFO Channel 17/0 : 7[7] -> 0[0] via P2P/CUMEM
+t-20260513223132-g9wrc-worker-0:10257:10428 [2] NCCL INFO Channel 20/0 : 2[2] -> 3[3] via P2P/CUMEM
+t-20260513223132-g9wrc-worker-0:10255:10425 [0] NCCL INFO Channel 23/0 : 0[0] -> 1[1] via P2P/CUMEM
+t-20260513223132-g9wrc-worker-0:10260:10429 [5] NCCL INFO Channel 19/0 : 5[5] -> 6[6] via P2P/CUMEM
+t-20260513223132-g9wrc-worker-0:10256:10427 [1] NCCL INFO Channel 19/0 : 1[1] -> 2[2] via P2P/CUMEM
+t-20260513223132-g9wrc-worker-0:10262:10430 [7] NCCL INFO Channel 18/0 : 7[7] -> 0[0] via P2P/CUMEM
+t-20260513223132-g9wrc-worker-0:10257:10428 [2] NCCL INFO Channel 21/0 : 2[2] -> 3[3] via P2P/CUMEM
+t-20260513223132-g9wrc-worker-0:10260:10429 [5] NCCL INFO Channel 20/0 : 5[5] -> 6[6] via P2P/CUMEM
+t-20260513223132-g9wrc-worker-0:10256:10427 [1] NCCL INFO Channel 20/0 : 1[1] -> 2[2] via P2P/CUMEM
+t-20260513223132-g9wrc-worker-0:10262:10430 [7] NCCL INFO Channel 19/0 : 7[7] -> 0[0] via P2P/CUMEM
+t-20260513223132-g9wrc-worker-0:10257:10428 [2] NCCL INFO Channel 22/0 : 2[2] -> 3[3] via P2P/CUMEM
+t-20260513223132-g9wrc-worker-0:10260:10429 [5] NCCL INFO Channel 21/0 : 5[5] -> 6[6] via P2P/CUMEM
+t-20260513223132-g9wrc-worker-0:10256:10427 [1] NCCL INFO Channel 21/0 : 1[1] -> 2[2] via P2P/CUMEM
+t-20260513223132-g9wrc-worker-0:10262:10430 [7] NCCL INFO Channel 20/0 : 7[7] -> 0[0] via P2P/CUMEM
+t-20260513223132-g9wrc-worker-0:10257:10428 [2] NCCL INFO Channel 23/0 : 2[2] -> 3[3] via P2P/CUMEM
+t-20260513223132-g9wrc-worker-0:10260:10429 [5] NCCL INFO Channel 22/0 : 5[5] -> 6[6] via P2P/CUMEM
+t-20260513223132-g9wrc-worker-0:10256:10427 [1] NCCL INFO Channel 22/0 : 1[1] -> 2[2] via P2P/CUMEM
+t-20260513223132-g9wrc-worker-0:10262:10430 [7] NCCL INFO Channel 21/0 : 7[7] -> 0[0] via P2P/CUMEM
+t-20260513223132-g9wrc-worker-0:10260:10429 [5] NCCL INFO Channel 23/0 : 5[5] -> 6[6] via P2P/CUMEM
+t-20260513223132-g9wrc-worker-0:10256:10427 [1] NCCL INFO Channel 23/0 : 1[1] -> 2[2] via P2P/CUMEM
+t-20260513223132-g9wrc-worker-0:10262:10430 [7] NCCL INFO Channel 22/0 : 7[7] -> 0[0] via P2P/CUMEM
+t-20260513223132-g9wrc-worker-0:10262:10430 [7] NCCL INFO Channel 23/0 : 7[7] -> 0[0] via P2P/CUMEM
+t-20260513223132-g9wrc-worker-0:10258:10426 [3] NCCL INFO Connected all rings, use ring PXN 0 GDR 1
+t-20260513223132-g9wrc-worker-0:10257:10428 [2] NCCL INFO Connected all rings, use ring PXN 0 GDR 1
+t-20260513223132-g9wrc-worker-0:10256:10427 [1] NCCL INFO Connected all rings, use ring PXN 0 GDR 1
+t-20260513223132-g9wrc-worker-0:10261:10423 [6] NCCL INFO Connected all rings, use ring PXN 0 GDR 1
+t-20260513223132-g9wrc-worker-0:10262:10430 [7] NCCL INFO Connected all rings, use ring PXN 0 GDR 1
+t-20260513223132-g9wrc-worker-0:10260:10429 [5] NCCL INFO Connected all rings, use ring PXN 0 GDR 1
+t-20260513223132-g9wrc-worker-0:10259:10424 [4] NCCL INFO Connected all rings, use ring PXN 0 GDR 1
+t-20260513223132-g9wrc-worker-0:10255:10425 [0] NCCL INFO Connected all rings, use ring PXN 0 GDR 1
+{
+  "device": "cuda:0",
+  "rank": 0,
+  "world_size": 8,
+  "samples": "owt_cached_chunks:8734897",
+  "vocab_size": 50257,
+  "tokenizer_vocab_size": 50257,
+  "save_dir": "runs/lta_owt_gpt2cached_len1024_rollout1_p1_ddit768x12_gbs512_8gpu_1m_20260513_143213",
+  "batch_size": 32,
+  "grad_accum": 2,
+  "effective_batch_size": 512,
+  "global_batch_size": 512,
+  "lr_schedule": "cosine",
+  "optimizer": "adamw",
+  "warmup_steps": 2000,
+  "min_lr": 6e-05,
+  "weight_decay": 0.1,
+  "adamw_param_groups": "nanogpt",
+  "adam_beta1": 0.9,
+  "adam_beta2": 0.95,
+  "adam_eps": 1e-08,
+  "muon_momentum": 0.95,
+  "muon_ns_steps": 5,
+  "muon_update_scale": 1.0,
+  "ema_decay": 0.0,
+  "ema_start_step": 0,
+  "model_type": "ddit",
+  "dual_t": true,
+  "corrupt_t_mode": "same",
+  "corrupt_min_t": 0.0,
+  "corrupt_max_t": 1.0,
+  "prefix_block_prob": 0.0,
+  "prefix_block_len": 128,
+  "dirichlet_endpoint_mode": "categorical_dual_t",
+  "dirichlet_semantic_t_mode": "same",
+  "dirichlet_semantic_t_value": 0.0,
+  "categorical_wrong_from_full_vocab": true,
+  "categorical_wrong_from_batch_valid_tokens": false,
+  "mask_mixture_original_prob": 0.0,
+  "mask_mixture_lowk_prob": 0.0,
+  "mask_mixture_lowcorrupt_prob": 0.0,
+  "mask_mixture_block_prob": 0.0,
+  "mask_mixture_all_prob": 0.0,
+  "mask_mixture_lowk_clean_tokens": "1,2,4,8,16,32,64",
+  "mask_mixture_lowcorrupt_tokens": "1,2,4,8,16,32,64",
+  "mask_mixture_block_tokens": "64,128",
+  "simplex_bridge_sampler": "dirichlet",
+  "logistic_normal_sigma_min": 0.18,
+  "logistic_normal_sigma_max": 2.2,
+  "logistic_normal_tau_min": 0.65,
+  "logistic_normal_tau_max": 1.15,
+  "torch_compile": false,
+  "compile_mode": "max-autotune",
+  "state_format": "prob",
+  "target_loss": "hard_ce",
+  "meanflow_weight": 0.0,
+  "rollout_train_prob": 1.0,
+  "rollout_train_steps": 1,
+  "rollout_train_infer_steps": 64,
+  "rollout_train_temp": 1.45,
+  "rollout_train_max_gamma": 1.0,
+  "rollout_train_corrupt_only": true,
+  "bridge_noise_init": "logistic_normal",
+  "noise_sigma": -1.0,
+  "allow_tf32": true,
+  "activation_checkpointing": false,
+  "activation_checkpoint_interval": 1,
+  "ddp_static_graph": false,
+  "ddp_gradient_as_bucket_view": true,
+  "blocking_data_transfer": false,
+  "dataloader_prefetch_factor": 2,
+  "full_train_stats": false,
+  "record_pad_truncate": false,
+  "record_add_eos": false,
+  "record_add_special_tokens": false,
+  "record_pad_token": "pad",
+  "record_shuffle_buffer": 10000,
+  "wrap": true,
+  "wrap_mode": "stream",
+  "wrap_record_buffer_size": 200,
+  "owt_cached_chunks": true,
+  "owt_chunk_cache_dir": "/e2e-data/evad-tech-vla/wanghan58/data/small_benchmarks/langflow_2604_11748/openwebtext_lta_cached_chunks/gpt2_len1024_train_minus_100k",
+  "owt_chunk_cache_rebuild": false,
+  "owt_chunk_cache_write_batch": 4096,
+  "owt_exact_repeat_per_chunk": 0,
+  "online_chunk_shuffle": false,
+  "online_chunk_shuffle_buffer": 10000,
+  "openwebtext_split": "train_minus_100k",
+  "detokenizer": "auto",
+  "resolved_detokenizer": null,
+  "num_workers": 4,
+  "latest_every": 1000,
+  "resume_path": ""
+}
+[rank6]: Traceback (most recent call last):
+[rank6]:   File "/e2e-data/evad-tech-vla/wanghan58/workspace/LTA_openwebtext_dualt/train.py", line 1487, in <module>
+[rank6]:     main()
+[rank6]:   File "/e2e-data/evad-tech-vla/wanghan58/workspace/LTA_openwebtext_dualt/train.py", line 1372, in main
+[rank6]:     logits = trainable_model(loss_state, model_t, bridge.attn_mask)
+[rank6]:              ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
+[rank6]:   File "/usr/local/lib/python3.12/dist-packages/torch/nn/modules/module.py", line 1739, in _wrapped_call_impl
+[rank6]:     return self._call_impl(*args, **kwargs)
+[rank6]:            ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
+[rank6]:   File "/usr/local/lib/python3.12/dist-packages/torch/nn/modules/module.py", line 1750, in _call_impl
+[rank6]:     return forward_call(*args, **kwargs)
+[rank6]:            ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
+[rank6]:   File "/usr/local/lib/python3.12/dist-packages/torch/nn/parallel/distributed.py", line 1650, in forward
+[rank6]:     inputs, kwargs = self._pre_forward(*inputs, **kwargs)
+[rank6]:                      ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
+[rank6]:   File "/usr/local/lib/python3.12/dist-packages/torch/nn/parallel/distributed.py", line 1539, in _pre_forward
+[rank6]:     if torch.is_grad_enabled() and self.reducer._rebuild_buckets():
+[rank6]:                                    ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
+[rank6]: RuntimeError: Expected to have finished reduction in the prior iteration before starting a new one. This error indicates that your module has parameters that were not used in producing loss. You can enable unused parameter detection by passing the keyword argument `find_unused_parameters=True` to `torch.nn.parallel.DistributedDataParallel`, and by 
+[rank6]: making sure all `forward` function outputs participate in calculating loss. 
+[rank6]: If you already have done the above, then the distributed data parallel module wasn't able to locate the output tensors in the return value of your module's `forward` function. Please include the loss function and the structure of the return value of `forward` of your module when reporting this issue (e.g. list, dict, iterable).
+[rank6]: Parameter indices which did not receive grad for rank 6: 0 1 2 3 4 6 7 9 10 11 12 13 14 16 17 19 20 21 22 23 24 26 27 29 30 31 32 33 34 36 37 39 40 41 42 43 44 46 47 49 50 51 52 53 54 56 57 59 60 61 62 63 64 66 67 69 70 71 72 73 74 76 77 79 80 81 82 83 84 86 87 89 90 91 92 93 94 96 97 99 100 101 102 103 104 106 107 109 110 111 112 113 114 116 117 119 120 121 122 123 ...
+[rank6]:  In addition, you can set the environment variable TORCH_DISTRIBUTED_DEBUG to either INFO or DETAIL to print out information about which particular parameters did not receive gradient on this rank as part of this error
+[rank4]: Traceback (most recent call last):
+[rank4]:   File "/e2e-data/evad-tech-vla/wanghan58/workspace/LTA_openwebtext_dualt/train.py", line 1487, in <module>
+[rank4]:     main()
+[rank4]:   File "/e2e-data/evad-tech-vla/wanghan58/workspace/LTA_openwebtext_dualt/train.py", line 1372, in main
+[rank4]:     logits = trainable_model(loss_state, model_t, bridge.attn_mask)
+[rank4]:              ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
+[rank4]:   File "/usr/local/lib/python3.12/dist-packages/torch/nn/modules/module.py", line 1739, in _wrapped_call_impl
+[rank4]:     return self._call_impl(*args, **kwargs)
+[rank4]:            ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
+[rank4]:   File "/usr/local/lib/python3.12/dist-packages/torch/nn/modules/module.py", line 1750, in _call_impl
+[rank4]:     return forward_call(*args, **kwargs)
+[rank4]:            ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
+[rank4]:   File "/usr/local/lib/python3.12/dist-packages/torch/nn/parallel/distributed.py", line 1650, in forward
+[rank4]:     inputs, kwargs = self._pre_forward(*inputs, **kwargs)
+[rank4]:                      ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
+[rank4]:   File "/usr/local/lib/python3.12/dist-packages/torch/nn/parallel/distributed.py", line 1539, in _pre_forward
+[rank4]:     if torch.is_grad_enabled() and self.reducer._rebuild_buckets():
+[rank4]:                                    ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
+[rank4]: RuntimeError: Expected to have finished reduction in the prior iteration before starting a new one. This error indicates that your module has parameters that were not used in producing loss. You can enable unused parameter detection by passing the keyword argument `find_unused_parameters=True` to `torch.nn.parallel.DistributedDataParallel`, and by 
+[rank4]: making sure all `forward` function outputs participate in calculating loss. 
+[rank4]: If you already have done the above, then the distributed data parallel module wasn't able to locate the output tensors in the return value of your module's `forward` function. Please include the loss function and the structure of the return value of `forward` of your module when reporting this issue (e.g. list, dict, iterable).
+[rank4]: Parameter indices which did not receive grad for rank 4: 0 1 2 3 4 6 7 9 10 11 12 13 14 16 17 19 20 21 22 23 24 26 27 29 30 31 32 33 34 36 37 39 40 41 42 43 44 46 47 49 50 51 52 53 54 56 57 59 60 61 62 63 64 66 67 69 70 71 72 73 74 76 77 79 80 81 82 83 84 86 87 89 90 91 92 93 94 96 97 99 100 101 102 103 104 106 107 109 110 111 112 113 114 116 117 119 120 121 122 123 ...
+[rank4]:  In addition, you can set the environment variable TORCH_DISTRIBUTED_DEBUG to either INFO or DETAIL to print out information about which particular parameters did not receive gradient on this rank as part of this error
+[rank7]: Traceback (most recent call last):
+[rank7]:   File "/e2e-data/evad-tech-vla/wanghan58/workspace/LTA_openwebtext_dualt/train.py", line 1487, in <module>
+[rank7]:     main()
+[rank7]:   File "/e2e-data/evad-tech-vla/wanghan58/workspace/LTA_openwebtext_dualt/train.py", line 1372, in main
+[rank7]:     logits = trainable_model(loss_state, model_t, bridge.attn_mask)
+[rank7]:              ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
+[rank7]:   File "/usr/local/lib/python3.12/dist-packages/torch/nn/modules/module.py", line 1739, in _wrapped_call_impl
+[rank7]:     return self._call_impl(*args, **kwargs)
+[rank7]:            ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
+[rank7]:   File "/usr/local/lib/python3.12/dist-packages/torch/nn/modules/module.py", line 1750, in _call_impl
+[rank7]:     return forward_call(*args, **kwargs)
+[rank7]:            ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
+[rank7]:   File "/usr/local/lib/python3.12/dist-packages/torch/nn/parallel/distributed.py", line 1650, in forward
+[rank7]:     inputs, kwargs = self._pre_forward(*inputs, **kwargs)
+[rank7]:                      ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
+[rank7]:   File "/usr/local/lib/python3.12/dist-packages/torch/nn/parallel/distributed.py", line 1539, in _pre_forward
+[rank7]:     if torch.is_grad_enabled() and self.reducer._rebuild_buckets():
+[rank7]:                                    ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
+[rank7]: RuntimeError: Expected to have finished reduction in the prior iteration before starting a new one. This error indicates that your module has parameters that were not used in producing loss. You can enable unused parameter detection by passing the keyword argument `find_unused_parameters=True` to `torch.nn.parallel.DistributedDataParallel`, and by 
+[rank7]: making sure all `forward` function outputs participate in calculating loss. 
+[rank7]: If you already have done the above, then the distributed data parallel module wasn't able to locate the output tensors in the return value of your module's `forward` function. Please include the loss function and the structure of the return value of `forward` of your module when reporting this issue (e.g. list, dict, iterable).
+[rank7]: Parameter indices which did not receive grad for rank 7: 0 1 2 3 4 6 7 9 10 11 12 13 14 16 17 19 20 21 22 23 24 26 27 29 30 31 32 33 34 36 37 39 40 41 42 43 44 46 47 49 50 51 52 53 54 56 57 59 60 61 62 63 64 66 67 69 70 71 72 73 74 76 77 79 80 81 82 83 84 86 87 89 90 91 92 93 94 96 97 99 100 101 102 103 104 106 107 109 110 111 112 113 114 116 117 119 120 121 122 123 ...
+[rank7]:  In addition, you can set the environment variable TORCH_DISTRIBUTED_DEBUG to either INFO or DETAIL to print out information about which particular parameters did not receive gradient on this rank as part of this error
+[rank3]: Traceback (most recent call last):
+[rank3]:   File "/e2e-data/evad-tech-vla/wanghan58/workspace/LTA_openwebtext_dualt/train.py", line 1487, in <module>
+[rank3]:     main()
+[rank3]:   File "/e2e-data/evad-tech-vla/wanghan58/workspace/LTA_openwebtext_dualt/train.py", line 1372, in main
+[rank3]:     logits = trainable_model(loss_state, model_t, bridge.attn_mask)
+[rank3]:              ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
+[rank3]:   File "/usr/local/lib/python3.12/dist-packages/torch/nn/modules/module.py", line 1739, in _wrapped_call_impl
+[rank3]:     return self._call_impl(*args, **kwargs)
+[rank3]:            ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
+[rank3]:   File "/usr/local/lib/python3.12/dist-packages/torch/nn/modules/module.py", line 1750, in _call_impl
+[rank3]:     return forward_call(*args, **kwargs)
+[rank3]:            ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
+[rank3]:   File "/usr/local/lib/python3.12/dist-packages/torch/nn/parallel/distributed.py", line 1650, in forward
+[rank3]:     inputs, kwargs = self._pre_forward(*inputs, **kwargs)
+[rank3]:                      ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
+[rank3]:   File "/usr/local/lib/python3.12/dist-packages/torch/nn/parallel/distributed.py", line 1539, in _pre_forward
+[rank3]:     if torch.is_grad_enabled() and self.reducer._rebuild_buckets():
+[rank3]:                                    ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
+[rank3]: RuntimeError: Expected to have finished reduction in the prior iteration before starting a new one. This error indicates that your module has parameters that were not used in producing loss. You can enable unused parameter detection by passing the keyword argument `find_unused_parameters=True` to `torch.nn.parallel.DistributedDataParallel`, and by 
+[rank3]: making sure all `forward` function outputs participate in calculating loss. 
+[rank3]: If you already have done the above, then the distributed data parallel module wasn't able to locate the output tensors in the return value of your module's `forward` function. Please include the loss function and the structure of the return value of `forward` of your module when reporting this issue (e.g. list, dict, iterable).
+[rank3]: Parameter indices which did not receive grad for rank 3: 0 1 2 3 4 6 7 9 10 11 12 13 14 16 17 19 20 21 22 23 24 26 27 29 30 31 32 33 34 36 37 39 40 41 42 43 44 46 47 49 50 51 52 53 54 56 57 59 60 61 62 63 64 66 67 69 70 71 72 73 74 76 77 79 80 81 82 83 84 86 87 89 90 91 92 93 94 96 97 99 100 101 102 103 104 106 107 109 110 111 112 113 114 116 117 119 120 121 122 123 ...
+[rank3]:  In addition, you can set the environment variable TORCH_DISTRIBUTED_DEBUG to either INFO or DETAIL to print out information about which particular parameters did not receive gradient on this rank as part of this error
+[rank2]: Traceback (most recent call last):
+[rank2]:   File "/e2e-data/evad-tech-vla/wanghan58/workspace/LTA_openwebtext_dualt/train.py", line 1487, in <module>
+[rank2]:     main()
+[rank2]:   File "/e2e-data/evad-tech-vla/wanghan58/workspace/LTA_openwebtext_dualt/train.py", line 1372, in main
+[rank2]:     logits = trainable_model(loss_state, model_t, bridge.attn_mask)
+[rank2]:              ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
+[rank2]:   File "/usr/local/lib/python3.12/dist-packages/torch/nn/modules/module.py", line 1739, in _wrapped_call_impl
+[rank2]:     return self._call_impl(*args, **kwargs)
+[rank2]:            ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
+[rank2]:   File "/usr/local/lib/python3.12/dist-packages/torch/nn/modules/module.py", line 1750, in _call_impl
+[rank2]:     return forward_call(*args, **kwargs)
+[rank2]:            ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
+[rank2]:   File "/usr/local/lib/python3.12/dist-packages/torch/nn/parallel/distributed.py", line 1650, in forward
+[rank2]:     inputs, kwargs = self._pre_forward(*inputs, **kwargs)
+[rank2]:                      ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
+[rank2]:   File "/usr/local/lib/python3.12/dist-packages/torch/nn/parallel/distributed.py", line 1539, in _pre_forward
+[rank2]:     if torch.is_grad_enabled() and self.reducer._rebuild_buckets():
+[rank2]:                                    ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
+[rank2]: RuntimeError: Expected to have finished reduction in the prior iteration before starting a new one. This error indicates that your module has parameters that were not used in producing loss. You can enable unused parameter detection by passing the keyword argument `find_unused_parameters=True` to `torch.nn.parallel.DistributedDataParallel`, and by 
+[rank2]: making sure all `forward` function outputs participate in calculating loss. 
+[rank2]: If you already have done the above, then the distributed data parallel module wasn't able to locate the output tensors in the return value of your module's `forward` function. Please include the loss function and the structure of the return value of `forward` of your module when reporting this issue (e.g. list, dict, iterable).
+[rank2]: Parameter indices which did not receive grad for rank 2: 0 1 2 3 4 6 7 9 10 11 12 13 14 16 17 19 20 21 22 23 24 26 27 29 30 31 32 33 34 36 37 39 40 41 42 43 44 46 47 49 50 51 52 53 54 56 57 59 60 61 62 63 64 66 67 69 70 71 72 73 74 76 77 79 80 81 82 83 84 86 87 89 90 91 92 93 94 96 97 99 100 101 102 103 104 106 107 109 110 111 112 113 114 116 117 119 120 121 122 123 ...
+[rank2]:  In addition, you can set the environment variable TORCH_DISTRIBUTED_DEBUG to either INFO or DETAIL to print out information about which particular parameters did not receive gradient on this rank as part of this error
+[rank5]: Traceback (most recent call last):
+[rank5]:   File "/e2e-data/evad-tech-vla/wanghan58/workspace/LTA_openwebtext_dualt/train.py", line 1487, in <module>
+[rank5]:     main()
+[rank5]:   File "/e2e-data/evad-tech-vla/wanghan58/workspace/LTA_openwebtext_dualt/train.py", line 1372, in main
+[rank5]:     logits = trainable_model(loss_state, model_t, bridge.attn_mask)
+[rank5]:              ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
+[rank5]:   File "/usr/local/lib/python3.12/dist-packages/torch/nn/modules/module.py", line 1739, in _wrapped_call_impl
+[rank5]:     return self._call_impl(*args, **kwargs)
+[rank5]:            ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
+[rank5]:   File "/usr/local/lib/python3.12/dist-packages/torch/nn/modules/module.py", line 1750, in _call_impl
+[rank5]:     return forward_call(*args, **kwargs)
+[rank5]:            ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
+[rank5]:   File "/usr/local/lib/python3.12/dist-packages/torch/nn/parallel/distributed.py", line 1650, in forward
+[rank5]:     inputs, kwargs = self._pre_forward(*inputs, **kwargs)
+[rank5]:                      ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
+[rank5]:   File "/usr/local/lib/python3.12/dist-packages/torch/nn/parallel/distributed.py", line 1539, in _pre_forward
+[rank5]:     if torch.is_grad_enabled() and self.reducer._rebuild_buckets():
+[rank5]:                                    ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
+[rank5]: RuntimeError: Expected to have finished reduction in the prior iteration before starting a new one. This error indicates that your module has parameters that were not used in producing loss. You can enable unused parameter detection by passing the keyword argument `find_unused_parameters=True` to `torch.nn.parallel.DistributedDataParallel`, and by 
+[rank5]: making sure all `forward` function outputs participate in calculating loss. 
+[rank5]: If you already have done the above, then the distributed data parallel module wasn't able to locate the output tensors in the return value of your module's `forward` function. Please include the loss function and the structure of the return value of `forward` of your module when reporting this issue (e.g. list, dict, iterable).
+[rank5]: Parameter indices which did not receive grad for rank 5: 0 1 2 3 4 6 7 9 10 11 12 13 14 16 17 19 20 21 22 23 24 26 27 29 30 31 32 33 34 36 37 39 40 41 42 43 44 46 47 49 50 51 52 53 54 56 57 59 60 61 62 63 64 66 67 69 70 71 72 73 74 76 77 79 80 81 82 83 84 86 87 89 90 91 92 93 94 96 97 99 100 101 102 103 104 106 107 109 110 111 112 113 114 116 117 119 120 121 122 123 ...
+[rank5]:  In addition, you can set the environment variable TORCH_DISTRIBUTED_DEBUG to either INFO or DETAIL to print out information about which particular parameters did not receive gradient on this rank as part of this error
+[rank1]: Traceback (most recent call last):
+[rank1]:   File "/e2e-data/evad-tech-vla/wanghan58/workspace/LTA_openwebtext_dualt/train.py", line 1487, in <module>
+[rank1]:     main()
+[rank1]:   File "/e2e-data/evad-tech-vla/wanghan58/workspace/LTA_openwebtext_dualt/train.py", line 1372, in main
+[rank1]:     logits = trainable_model(loss_state, model_t, bridge.attn_mask)
+[rank1]:              ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
+[rank1]:   File "/usr/local/lib/python3.12/dist-packages/torch/nn/modules/module.py", line 1739, in _wrapped_call_impl
+[rank1]:     return self._call_impl(*args, **kwargs)
+[rank1]:            ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
+[rank1]:   File "/usr/local/lib/python3.12/dist-packages/torch/nn/modules/module.py", line 1750, in _call_impl
+[rank1]:     return forward_call(*args, **kwargs)
+[rank1]:            ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
+[rank1]:   File "/usr/local/lib/python3.12/dist-packages/torch/nn/parallel/distributed.py", line 1650, in forward
+[rank1]:     inputs, kwargs = self._pre_forward(*inputs, **kwargs)
+[rank1]:                      ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
+[rank1]:   File "/usr/local/lib/python3.12/dist-packages/torch/nn/parallel/distributed.py", line 1539, in _pre_forward
+[rank1]:     if torch.is_grad_enabled() and self.reducer._rebuild_buckets():
+[rank1]:                                    ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
+[rank1]: RuntimeError: Expected to have finished reduction in the prior iteration before starting a new one. This error indicates that your module has parameters that were not used in producing loss. You can enable unused parameter detection by passing the keyword argument `find_unused_parameters=True` to `torch.nn.parallel.DistributedDataParallel`, and by 
+[rank1]: making sure all `forward` function outputs participate in calculating loss. 
+[rank1]: If you already have done the above, then the distributed data parallel module wasn't able to locate the output tensors in the return value of your module's `forward` function. Please include the loss function and the structure of the return value of `forward` of your module when reporting this issue (e.g. list, dict, iterable).
+[rank1]: Parameter indices which did not receive grad for rank 1: 0 1 2 3 4 6 7 9 10 11 12 13 14 16 17 19 20 21 22 23 24 26 27 29 30 31 32 33 34 36 37 39 40 41 42 43 44 46 47 49 50 51 52 53 54 56 57 59 60 61 62 63 64 66 67 69 70 71 72 73 74 76 77 79 80 81 82 83 84 86 87 89 90 91 92 93 94 96 97 99 100 101 102 103 104 106 107 109 110 111 112 113 114 116 117 119 120 121 122 123 ...
+[rank1]:  In addition, you can set the environment variable TORCH_DISTRIBUTED_DEBUG to either INFO or DETAIL to print out information about which particular parameters did not receive gradient on this rank as part of this error
+[rank0]: Traceback (most recent call last):
+[rank0]:   File "/e2e-data/evad-tech-vla/wanghan58/workspace/LTA_openwebtext_dualt/train.py", line 1487, in <module>
+[rank0]:     main()
+[rank0]:   File "/e2e-data/evad-tech-vla/wanghan58/workspace/LTA_openwebtext_dualt/train.py", line 1372, in main
+[rank0]:     logits = trainable_model(loss_state, model_t, bridge.attn_mask)
+[rank0]:              ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
+[rank0]:   File "/usr/local/lib/python3.12/dist-packages/torch/nn/modules/module.py", line 1739, in _wrapped_call_impl
+[rank0]:     return self._call_impl(*args, **kwargs)
+[rank0]:            ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
+[rank0]:   File "/usr/local/lib/python3.12/dist-packages/torch/nn/modules/module.py", line 1750, in _call_impl
+[rank0]:     return forward_call(*args, **kwargs)
+[rank0]:            ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
+[rank0]:   File "/usr/local/lib/python3.12/dist-packages/torch/nn/parallel/distributed.py", line 1650, in forward
+[rank0]:     inputs, kwargs = self._pre_forward(*inputs, **kwargs)
+[rank0]:                      ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
+[rank0]:   File "/usr/local/lib/python3.12/dist-packages/torch/nn/parallel/distributed.py", line 1539, in _pre_forward
+[rank0]:     if torch.is_grad_enabled() and self.reducer._rebuild_buckets():
+[rank0]:                                    ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
+[rank0]: RuntimeError: Expected to have finished reduction in the prior iteration before starting a new one. This error indicates that your module has parameters that were not used in producing loss. You can enable unused parameter detection by passing the keyword argument `find_unused_parameters=True` to `torch.nn.parallel.DistributedDataParallel`, and by 
+[rank0]: making sure all `forward` function outputs participate in calculating loss. 
+[rank0]: If you already have done the above, then the distributed data parallel module wasn't able to locate the output tensors in the return value of your module's `forward` function. Please include the loss function and the structure of the return value of `forward` of your module when reporting this issue (e.g. list, dict, iterable).
+[rank0]: Parameter indices which did not receive grad for rank 0: 0 1 2 3 4 6 7 9 10 11 12 13 14 16 17 19 20 21 22 23 24 26 27 29 30 31 32 33 34 36 37 39 40 41 42 43 44 46 47 49 50 51 52 53 54 56 57 59 60 61 62 63 64 66 67 69 70 71 72 73 74 76 77 79 80 81 82 83 84 86 87 89 90 91 92 93 94 96 97 99 100 101 102 103 104 106 107 109 110 111 112 113 114 116 117 119 120 121 122 123 ...
+[rank0]:  In addition, you can set the environment variable TORCH_DISTRIBUTED_DEBUG to either INFO or DETAIL to print out information about which particular parameters did not receive gradient on this rank as part of this error
+t-20260513223132-g9wrc-worker-0:10256:10750 [1] NCCL INFO misc/socket.cc:64 -> 3
+t-20260513223132-g9wrc-worker-0:10256:10750 [1] NCCL INFO misc/socket.cc:80 -> 3
+t-20260513223132-g9wrc-worker-0:10256:10750 [1] NCCL INFO misc/socket.cc:828 -> 3
+t-20260513223132-g9wrc-worker-0:10256:10750 [1] NCCL INFO misc/socket.cc:64 -> 3
+t-20260513223132-g9wrc-worker-0:10256:10750 [1] NCCL INFO misc/socket.cc:80 -> 3
+t-20260513223132-g9wrc-worker-0:10256:10750 [1] NCCL INFO misc/socket.cc:828 -> 3
+t-20260513223132-g9wrc-worker-0:10256:10750 [1] NCCL INFO misc/socket.cc:64 -> 3
+t-20260513223132-g9wrc-worker-0:10256:10750 [1] NCCL INFO misc/socket.cc:80 -> 3
+t-20260513223132-g9wrc-worker-0:10256:10750 [1] NCCL INFO misc/socket.cc:828 -> 3
+t-20260513223132-g9wrc-worker-0:10256:10409 [1] NCCL INFO misc/socket.cc:880 -> 3
+t-20260513223132-g9wrc-worker-0:10261:10752 [6] NCCL INFO misc/socket.cc:64 -> 3
+t-20260513223132-g9wrc-worker-0:10261:10752 [6] NCCL INFO misc/socket.cc:80 -> 3
+t-20260513223132-g9wrc-worker-0:10261:10752 [6] NCCL INFO misc/socket.cc:828 -> 3
+t-20260513223132-g9wrc-worker-0:10261:10752 [6] NCCL INFO misc/socket.cc:64 -> 3
+t-20260513223132-g9wrc-worker-0:10261:10752 [6] NCCL INFO misc/socket.cc:80 -> 3
+t-20260513223132-g9wrc-worker-0:10261:10752 [6] NCCL INFO misc/socket.cc:828 -> 3
+t-20260513223132-g9wrc-worker-0:10261:10752 [6] NCCL INFO misc/socket.cc:64 -> 3
+t-20260513223132-g9wrc-worker-0:10261:10752 [6] NCCL INFO misc/socket.cc:80 -> 3
+t-20260513223132-g9wrc-worker-0:10261:10752 [6] NCCL INFO misc/socket.cc:828 -> 3
+t-20260513223132-g9wrc-worker-0:10261:10752 [6] NCCL INFO misc/socket.cc:64 -> 3
+t-20260513223132-g9wrc-worker-0:10261:10752 [6] NCCL INFO misc/socket.cc:80 -> 3
+t-20260513223132-g9wrc-worker-0:10261:10752 [6] NCCL INFO misc/socket.cc:828 -> 3
+t-20260513223132-g9wrc-worker-0:10261:10411 [6] NCCL INFO misc/socket.cc:880 -> 3
+t-20260513223132-g9wrc-worker-0:10257:10754 [2] NCCL INFO misc/socket.cc:64 -> 3
+t-20260513223132-g9wrc-worker-0:10257:10754 [2] NCCL INFO misc/socket.cc:80 -> 3
+t-20260513223132-g9wrc-worker-0:10257:10754 [2] NCCL INFO misc/socket.cc:828 -> 3
+t-20260513223132-g9wrc-worker-0:10257:10754 [2] NCCL INFO misc/socket.cc:64 -> 3
+t-20260513223132-g9wrc-worker-0:10257:10754 [2] NCCL INFO misc/socket.cc:80 -> 3
+t-20260513223132-g9wrc-worker-0:10257:10754 [2] NCCL INFO misc/socket.cc:828 -> 3
+t-20260513223132-g9wrc-worker-0:10257:10754 [2] NCCL INFO misc/socket.cc:64 -> 3
+t-20260513223132-g9wrc-worker-0:10257:10754 [2] NCCL INFO misc/socket.cc:80 -> 3
+t-20260513223132-g9wrc-worker-0:10257:10754 [2] NCCL INFO misc/socket.cc:828 -> 3
+t-20260513223132-g9wrc-worker-0:10256:10409 [1] NCCL INFO misc/socket.cc:880 -> 3
+t-20260513223132-g9wrc-worker-0:10257:10754 [2] NCCL INFO misc/socket.cc:64 -> 3
+t-20260513223132-g9wrc-worker-0:10257:10754 [2] NCCL INFO misc/socket.cc:80 -> 3
+t-20260513223132-g9wrc-worker-0:10257:10754 [2] NCCL INFO misc/socket.cc:828 -> 3
+t-20260513223132-g9wrc-worker-0:10257:10407 [2] NCCL INFO misc/socket.cc:880 -> 3
+t-20260513223132-g9wrc-worker-0:10260:10756 [5] NCCL INFO misc/socket.cc:64 -> 3
+t-20260513223132-g9wrc-worker-0:10260:10756 [5] NCCL INFO misc/socket.cc:80 -> 3
+t-20260513223132-g9wrc-worker-0:10260:10756 [5] NCCL INFO misc/socket.cc:828 -> 3
+t-20260513223132-g9wrc-worker-0:10260:10756 [5] NCCL INFO misc/socket.cc:64 -> 3
+t-20260513223132-g9wrc-worker-0:10260:10756 [5] NCCL INFO misc/socket.cc:80 -> 3
+t-20260513223132-g9wrc-worker-0:10260:10756 [5] NCCL INFO misc/socket.cc:828 -> 3
+t-20260513223132-g9wrc-worker-0:10260:10756 [5] NCCL INFO misc/socket.cc:64 -> 3
+t-20260513223132-g9wrc-worker-0:10260:10756 [5] NCCL INFO misc/socket.cc:80 -> 3
+t-20260513223132-g9wrc-worker-0:10260:10756 [5] NCCL INFO misc/socket.cc:828 -> 3
+t-20260513223132-g9wrc-worker-0:10260:10756 [5] NCCL INFO misc/socket.cc:64 -> 3
+t-20260513223132-g9wrc-worker-0:10260:10756 [5] NCCL INFO misc/socket.cc:80 -> 3
+t-20260513223132-g9wrc-worker-0:10260:10756 [5] NCCL INFO misc/socket.cc:828 -> 3
+t-20260513223132-g9wrc-worker-0:10260:10414 [5] NCCL INFO misc/socket.cc:880 -> 3
+t-20260513223132-g9wrc-worker-0:10261:10411 [6] NCCL INFO misc/socket.cc:880 -> 3
+t-20260513223132-g9wrc-worker-0:10258:10758 [3] NCCL INFO misc/socket.cc:64 -> 3
+t-20260513223132-g9wrc-worker-0:10258:10758 [3] NCCL INFO misc/socket.cc:80 -> 3
+t-20260513223132-g9wrc-worker-0:10258:10758 [3] NCCL INFO misc/socket.cc:828 -> 3
+t-20260513223132-g9wrc-worker-0:10258:10758 [3] NCCL INFO misc/socket.cc:64 -> 3
+t-20260513223132-g9wrc-worker-0:10258:10758 [3] NCCL INFO misc/socket.cc:80 -> 3
+t-20260513223132-g9wrc-worker-0:10258:10758 [3] NCCL INFO misc/socket.cc:828 -> 3
+t-20260513223132-g9wrc-worker-0:10258:10758 [3] NCCL INFO misc/socket.cc:64 -> 3
+t-20260513223132-g9wrc-worker-0:10257:10407 [2] NCCL INFO misc/socket.cc:880 -> 3
+t-20260513223132-g9wrc-worker-0:10258:10758 [3] NCCL INFO misc/socket.cc:80 -> 3
+t-20260513223132-g9wrc-worker-0:10258:10758 [3] NCCL INFO misc/socket.cc:828 -> 3
+t-20260513223132-g9wrc-worker-0:10258:10758 [3] NCCL INFO misc/socket.cc:64 -> 3
+t-20260513223132-g9wrc-worker-0:10258:10758 [3] NCCL INFO misc/socket.cc:80 -> 3
+t-20260513223132-g9wrc-worker-0:10258:10758 [3] NCCL INFO misc/socket.cc:828 -> 3
+t-20260513223132-g9wrc-worker-0:10258:10418 [3] NCCL INFO misc/socket.cc:880 -> 3
+t-20260513223132-g9wrc-worker-0:10262:10760 [7] NCCL INFO misc/socket.cc:64 -> 3
+t-20260513223132-g9wrc-worker-0:10262:10760 [7] NCCL INFO misc/socket.cc:80 -> 3
+t-20260513223132-g9wrc-worker-0:10262:10760 [7] NCCL INFO misc/socket.cc:828 -> 3
+t-20260513223132-g9wrc-worker-0:10262:10760 [7] NCCL INFO misc/socket.cc:64 -> 3
+t-20260513223132-g9wrc-worker-0:10262:10760 [7] NCCL INFO misc/socket.cc:80 -> 3
+t-20260513223132-g9wrc-worker-0:10262:10760 [7] NCCL INFO misc/socket.cc:828 -> 3
+t-20260513223132-g9wrc-worker-0:10262:10760 [7] NCCL INFO misc/socket.cc:64 -> 3
+t-20260513223132-g9wrc-worker-0:10262:10760 [7] NCCL INFO misc/socket.cc:80 -> 3
+t-20260513223132-g9wrc-worker-0:10262:10760 [7] NCCL INFO misc/socket.cc:828 -> 3
+t-20260513223132-g9wrc-worker-0:10262:10413 [7] NCCL INFO misc/socket.cc:880 -> 3
+t-20260513223132-g9wrc-worker-0:10259:10762 [4] NCCL INFO misc/socket.cc:64 -> 3
+t-20260513223132-g9wrc-worker-0:10259:10762 [4] NCCL INFO misc/socket.cc:80 -> 3
+t-20260513223132-g9wrc-worker-0:10259:10762 [4] NCCL INFO misc/socket.cc:828 -> 3
+t-20260513223132-g9wrc-worker-0:10259:10762 [4] NCCL INFO misc/socket.cc:64 -> 3
+t-20260513223132-g9wrc-worker-0:10259:10762 [4] NCCL INFO misc/socket.cc:80 -> 3
+t-20260513223132-g9wrc-worker-0:10259:10762 [4] NCCL INFO misc/socket.cc:828 -> 3
+t-20260513223132-g9wrc-worker-0:10258:10418 [3] NCCL INFO misc/socket.cc:880 -> 3
+t-20260513223132-g9wrc-worker-0:10259:10762 [4] NCCL INFO misc/socket.cc:64 -> 3
+t-20260513223132-g9wrc-worker-0:10259:10762 [4] NCCL INFO misc/socket.cc:80 -> 3
+t-20260513223132-g9wrc-worker-0:10259:10762 [4] NCCL INFO misc/socket.cc:828 -> 3
+t-20260513223132-g9wrc-worker-0:10259:10762 [4] NCCL INFO misc/socket.cc:64 -> 3
+t-20260513223132-g9wrc-worker-0:10259:10762 [4] NCCL INFO misc/socket.cc:80 -> 3
+t-20260513223132-g9wrc-worker-0:10259:10762 [4] NCCL INFO misc/socket.cc:828 -> 3
+t-20260513223132-g9wrc-worker-0:10259:10415 [4] NCCL INFO misc/socket.cc:880 -> 3
+t-20260513223132-g9wrc-worker-0:10260:10414 [5] NCCL INFO misc/socket.cc:880 -> 3
+[rank0]:[W513 14:32:46.158776236 ProcessGroupNCCL.cpp:1487] Warning: WARNING: destroy_process_group() was not called before program exit, which can leak resources. For more info, please see https://pytorch.org/docs/stable/distributed.html#shutdown (function operator())
+t-20260513223132-g9wrc-worker-0:10255:10764 [0] NCCL INFO misc/socket.cc:64 -> 3
+t-20260513223132-g9wrc-worker-0:10255:10764 [0] NCCL INFO misc/socket.cc:80 -> 3
+t-20260513223132-g9wrc-worker-0:10255:10764 [0] NCCL INFO misc/socket.cc:828 -> 3
+t-20260513223132-g9wrc-worker-0:10255:10421 [0] NCCL INFO misc/socket.cc:880 -> 3
+t-20260513223132-g9wrc-worker-0:10255:10764 [0] NCCL INFO misc/socket.cc:64 -> 3
+t-20260513223132-g9wrc-worker-0:10255:10764 [0] NCCL INFO misc/socket.cc:80 -> 3
+t-20260513223132-g9wrc-worker-0:10255:10764 [0] NCCL INFO misc/socket.cc:828 -> 3
+t-20260513223132-g9wrc-worker-0:10255:10764 [0] NCCL INFO misc/socket.cc:64 -> 3
+t-20260513223132-g9wrc-worker-0:10255:10764 [0] NCCL INFO misc/socket.cc:80 -> 3
+t-20260513223132-g9wrc-worker-0:10255:10764 [0] NCCL INFO misc/socket.cc:828 -> 3
+t-20260513223132-g9wrc-worker-0:10262:10413 [7] NCCL INFO misc/socket.cc:880 -> 3
+t-20260513223132-g9wrc-worker-0:10257:10754 [2] NCCL INFO comm 0xb0afb80 rank 2 nranks 8 cudaDev 2 busId 69020 - Abort COMPLETE
+t-20260513223132-g9wrc-worker-0:10256:10750 [1] NCCL INFO comm 0x9e75e40 rank 1 nranks 8 cudaDev 1 busId 67020 - Abort COMPLETE
+t-20260513223132-g9wrc-worker-0:10260:10756 [5] NCCL INFO comm 0xa89c920 rank 5 nranks 8 cudaDev 5 busId 71020 - Abort COMPLETE
+t-20260513223132-g9wrc-worker-0:10258:10758 [3] NCCL INFO comm 0xb5ef120 rank 3 nranks 8 cudaDev 3 busId 6b020 - Abort COMPLETE
+t-20260513223132-g9wrc-worker-0:10261:10752 [6] NCCL INFO comm 0xb044c70 rank 6 nranks 8 cudaDev 6 busId 73020 - Abort COMPLETE
+t-20260513223132-g9wrc-worker-0:10262:10760 [7] NCCL INFO comm 0xab21280 rank 7 nranks 8 cudaDev 7 busId 75020 - Abort COMPLETE
+t-20260513223132-g9wrc-worker-0:10259:10762 [4] NCCL INFO comm 0xb1d1a20 rank 4 nranks 8 cudaDev 4 busId 6f020 - Abort COMPLETE
+t-20260513223132-g9wrc-worker-0:10255:10764 [0] NCCL INFO comm 0xb5c2750 rank 0 nranks 8 cudaDev 0 busId 65040 - Abort COMPLETE
+W0513 14:32:47.004000 10251 torch/distributed/elastic/multiprocessing/api.py:898] Sending process 10255 closing signal SIGTERM
+W0513 14:32:47.005000 10251 torch/distributed/elastic/multiprocessing/api.py:898] Sending process 10258 closing signal SIGTERM
+W0513 14:32:47.005000 10251 torch/distributed/elastic/multiprocessing/api.py:898] Sending process 10259 closing signal SIGTERM
+W0513 14:32:47.006000 10251 torch/distributed/elastic/multiprocessing/api.py:898] Sending process 10260 closing signal SIGTERM
+W0513 14:32:47.006000 10251 torch/distributed/elastic/multiprocessing/api.py:898] Sending process 10261 closing signal SIGTERM
+W0513 14:32:47.007000 10251 torch/distributed/elastic/multiprocessing/api.py:898] Sending process 10262 closing signal SIGTERM
+E0513 14:32:47.522000 10251 torch/distributed/elastic/multiprocessing/api.py:870] failed (exitcode: 1) local_rank: 1 (pid: 10256) of binary: /usr/bin/python
+Traceback (most recent call last):
+  File "<frozen runpy>", line 198, in _run_module_as_main
+  File "<frozen runpy>", line 88, in _run_code
+  File "/usr/local/lib/python3.12/dist-packages/torch/distributed/run.py", line 922, in <module>
+    main()
+  File "/usr/local/lib/python3.12/dist-packages/torch/distributed/elastic/multiprocessing/errors/__init__.py", line 355, in wrapper
+    return f(*args, **kwargs)
+           ^^^^^^^^^^^^^^^^^^
+  File "/usr/local/lib/python3.12/dist-packages/torch/distributed/run.py", line 918, in main
+    run(args)
+  File "/usr/local/lib/python3.12/dist-packages/torch/distributed/run.py", line 909, in run
+    elastic_launch(
+  File "/usr/local/lib/python3.12/dist-packages/torch/distributed/launcher/api.py", line 139, in __call__
+    return launch_agent(self._config, self._entrypoint, list(args))
+           ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
+  File "/usr/local/lib/python3.12/dist-packages/torch/distributed/launcher/api.py", line 270, in launch_agent
+    raise ChildFailedError(
+torch.distributed.elastic.multiprocessing.errors.ChildFailedError: 
+============================================================
+train.py FAILED
+------------------------------------------------------------
+Failures:
+[1]:
+  time      : 2026-05-13_14:32:47
+  host      : t-20260513223132-g9wrc-worker-0.t-20260513223132-g9wrc-worker.mlplatform-customtask.svc.cluster.local
+  rank      : 2 (local_rank: 2)
+  exitcode  : 1 (pid: 10257)
+  error_file: <N/A>
+  traceback : To enable traceback see: https://pytorch.org/docs/stable/elastic/errors.html
+------------------------------------------------------------
+Root Cause (first observed failure):
+[0]:
+  time      : 2026-05-13_14:32:47
+  host      : t-20260513223132-g9wrc-worker-0.t-20260513223132-g9wrc-worker.mlplatform-customtask.svc.cluster.local
+  rank      : 1 (local_rank: 1)
+  exitcode  : 1 (pid: 10256)
+  error_file: <N/A>
+  traceback : To enable traceback see: https://pytorch.org/docs/stable/elastic/errors.html
+============================================================

LTA_openwebtext_dualt/logs/owt_bert_absrope_time4_C1_to_1024_mask1_sameT_dualline_watch/infer_lta_owt_bert_absrope_time4_dirichlet_len1024_C1_to_1024_mask1_sameT_gbs512_b32_8gpu_1m_save10k_20260526_step_0040000.log

@@ -0,0 +1,398 @@
+[watch-gumbel] 2026-05-26_08:08:35 infer runs/lta_owt_bert_absrope_time4_dirichlet_len1024_C1_to_1024_mask1_sameT_gbs512_b32_8gpu_1m_save10k_20260526/step_0040000.pt -> docs/lta_samples/metrics_20260526/owt_bert_absrope_time4_C1_to_1024_mask1_sameT_dualline_dirres_c1_1024_n128/lta_owt_bert_absrope_time4_dirichlet_len1024_C1_to_1024_mask1_sameT_gbs512_b32_8gpu_1m_save10k_20260526/step_0040000
+[watch-gumbel] 2026-05-26_08:08:37 infer runs/lta_owt_bert_absrope_time4_dirichlet_len1024_C1_to_1024_mask1_sameT_gbs512_b32_8gpu_1m_save10k_20260526/step_0040000.pt -> docs/lta_samples/metrics_20260526/owt_bert_absrope_time4_C1_to_1024_mask1_sameT_dualline_dirres_c1_1024_n128/lta_owt_bert_absrope_time4_dirichlet_len1024_C1_to_1024_mask1_sameT_gbs512_b32_8gpu_1m_save10k_20260526/step_0040000
+[decode] max_len=1024 generated=1/128
+[decode] max_len=1024 generated=1/128
+[decode] max_len=1024 generated=2/128
+[decode] max_len=1024 generated=2/128
+[decode] max_len=1024 generated=3/128
+[decode] max_len=1024 generated=3/128
+[decode] max_len=1024 generated=4/128
+[decode] max_len=1024 generated=4/128
+[decode] max_len=1024 generated=5/128
+[decode] max_len=1024 generated=5/128
+[decode] max_len=1024 generated=6/128
+[decode] max_len=1024 generated=6/128
+[decode] max_len=1024 generated=7/128
+[decode] max_len=1024 generated=7/128
+[decode] max_len=1024 generated=8/128
+[decode] max_len=1024 generated=8/128
+[decode] max_len=1024 generated=9/128
+[decode] max_len=1024 generated=9/128
+[decode] max_len=1024 generated=10/128
+[decode] max_len=1024 generated=10/128
+[decode] max_len=1024 generated=11/128
+[decode] max_len=1024 generated=11/128
+[decode] max_len=1024 generated=12/128
+[decode] max_len=1024 generated=12/128
+[decode] max_len=1024 generated=13/128
+[decode] max_len=1024 generated=13/128
+[decode] max_len=1024 generated=14/128
+[decode] max_len=1024 generated=14/128
+[decode] max_len=1024 generated=15/128
+[decode] max_len=1024 generated=15/128
+[decode] max_len=1024 generated=16/128
+[decode] max_len=1024 generated=16/128
+[decode] max_len=1024 generated=17/128
+[decode] max_len=1024 generated=17/128
+[decode] max_len=1024 generated=18/128
+[decode] max_len=1024 generated=18/128
+[decode] max_len=1024 generated=19/128
+[decode] max_len=1024 generated=19/128
+[decode] max_len=1024 generated=20/128
+[decode] max_len=1024 generated=20/128
+[decode] max_len=1024 generated=21/128
+[decode] max_len=1024 generated=21/128
+[decode] max_len=1024 generated=22/128
+[decode] max_len=1024 generated=22/128
+[decode] max_len=1024 generated=23/128
+[decode] max_len=1024 generated=23/128
+[decode] max_len=1024 generated=24/128
+[decode] max_len=1024 generated=24/128
+[decode] max_len=1024 generated=25/128
+[decode] max_len=1024 generated=25/128
+[decode] max_len=1024 generated=26/128
+[decode] max_len=1024 generated=26/128
+[decode] max_len=1024 generated=27/128
+[decode] max_len=1024 generated=27/128
+[decode] max_len=1024 generated=28/128
+[decode] max_len=1024 generated=28/128
+[decode] max_len=1024 generated=29/128
+[decode] max_len=1024 generated=29/128
+[decode] max_len=1024 generated=30/128
+[decode] max_len=1024 generated=30/128
+[decode] max_len=1024 generated=31/128
+[decode] max_len=1024 generated=31/128
+[decode] max_len=1024 generated=32/128
+[decode] max_len=1024 generated=32/128
+[decode] max_len=1024 generated=33/128
+[decode] max_len=1024 generated=33/128
+[decode] max_len=1024 generated=34/128
+[decode] max_len=1024 generated=34/128
+[decode] max_len=1024 generated=35/128
+[decode] max_len=1024 generated=35/128
+[decode] max_len=1024 generated=36/128
+[decode] max_len=1024 generated=36/128
+[decode] max_len=1024 generated=37/128
+[decode] max_len=1024 generated=37/128
+[decode] max_len=1024 generated=38/128
+[decode] max_len=1024 generated=38/128
+[decode] max_len=1024 generated=39/128
+[decode] max_len=1024 generated=39/128
+[decode] max_len=1024 generated=40/128
+[decode] max_len=1024 generated=40/128
+[decode] max_len=1024 generated=41/128
+[decode] max_len=1024 generated=41/128
+[decode] max_len=1024 generated=42/128
+[decode] max_len=1024 generated=42/128
+[decode] max_len=1024 generated=43/128
+[decode] max_len=1024 generated=43/128
+[decode] max_len=1024 generated=44/128
+[decode] max_len=1024 generated=44/128
+[decode] max_len=1024 generated=45/128
+[decode] max_len=1024 generated=45/128
+[decode] max_len=1024 generated=46/128
+[decode] max_len=1024 generated=46/128
+[decode] max_len=1024 generated=47/128
+[decode] max_len=1024 generated=47/128
+[decode] max_len=1024 generated=48/128
+[decode] max_len=1024 generated=48/128
+[decode] max_len=1024 generated=49/128
+[decode] max_len=1024 generated=49/128
+[decode] max_len=1024 generated=50/128
+[decode] max_len=1024 generated=50/128
+[decode] max_len=1024 generated=51/128
+[decode] max_len=1024 generated=51/128
+[decode] max_len=1024 generated=52/128
+[decode] max_len=1024 generated=52/128
+[decode] max_len=1024 generated=53/128
+[decode] max_len=1024 generated=53/128
+[decode] max_len=1024 generated=54/128
+[decode] max_len=1024 generated=54/128
+[decode] max_len=1024 generated=55/128
+[decode] max_len=1024 generated=55/128
+[decode] max_len=1024 generated=56/128
+[decode] max_len=1024 generated=56/128
+[decode] max_len=1024 generated=57/128
+[decode] max_len=1024 generated=57/128
+[decode] max_len=1024 generated=58/128
+[decode] max_len=1024 generated=58/128
+[decode] max_len=1024 generated=59/128
+[decode] max_len=1024 generated=59/128
+[decode] max_len=1024 generated=60/128
+[decode] max_len=1024 generated=60/128
+[decode] max_len=1024 generated=61/128
+[decode] max_len=1024 generated=61/128
+[decode] max_len=1024 generated=62/128
+[decode] max_len=1024 generated=62/128
+[decode] max_len=1024 generated=63/128
+[decode] max_len=1024 generated=63/128
+[decode] max_len=1024 generated=64/128
+[decode] max_len=1024 generated=64/128
+[decode] max_len=1024 generated=65/128
+[decode] max_len=1024 generated=65/128
+[decode] max_len=1024 generated=66/128
+[decode] max_len=1024 generated=66/128
+[decode] max_len=1024 generated=67/128
+[decode] max_len=1024 generated=67/128
+[decode] max_len=1024 generated=68/128
+[decode] max_len=1024 generated=68/128
+[decode] max_len=1024 generated=69/128
+[decode] max_len=1024 generated=69/128
+[decode] max_len=1024 generated=70/128
+[decode] max_len=1024 generated=70/128
+[decode] max_len=1024 generated=71/128
+[decode] max_len=1024 generated=71/128
+[decode] max_len=1024 generated=72/128
+[decode] max_len=1024 generated=72/128
+[decode] max_len=1024 generated=73/128
+[decode] max_len=1024 generated=73/128
+[decode] max_len=1024 generated=74/128
+[decode] max_len=1024 generated=74/128
+[decode] max_len=1024 generated=75/128
+[decode] max_len=1024 generated=75/128
+[decode] max_len=1024 generated=76/128
+[decode] max_len=1024 generated=76/128
+[decode] max_len=1024 generated=77/128
+[decode] max_len=1024 generated=77/128
+[decode] max_len=1024 generated=78/128
+[decode] max_len=1024 generated=78/128
+[decode] max_len=1024 generated=79/128
+[decode] max_len=1024 generated=79/128
+[decode] max_len=1024 generated=80/128
+[decode] max_len=1024 generated=80/128
+[decode] max_len=1024 generated=81/128
+[decode] max_len=1024 generated=81/128
+[decode] max_len=1024 generated=82/128
+[decode] max_len=1024 generated=82/128
+[decode] max_len=1024 generated=83/128
+[decode] max_len=1024 generated=83/128
+[decode] max_len=1024 generated=84/128
+[decode] max_len=1024 generated=84/128
+[decode] max_len=1024 generated=85/128
+[decode] max_len=1024 generated=85/128
+[decode] max_len=1024 generated=86/128
+[decode] max_len=1024 generated=86/128
+[decode] max_len=1024 generated=87/128
+[decode] max_len=1024 generated=87/128
+[decode] max_len=1024 generated=88/128
+[decode] max_len=1024 generated=88/128
+[decode] max_len=1024 generated=89/128
+[decode] max_len=1024 generated=89/128
+[decode] max_len=1024 generated=90/128
+[decode] max_len=1024 generated=90/128
+[decode] max_len=1024 generated=91/128
+[decode] max_len=1024 generated=91/128
+[decode] max_len=1024 generated=92/128
+[decode] max_len=1024 generated=92/128
+[decode] max_len=1024 generated=93/128
+[decode] max_len=1024 generated=93/128
+[decode] max_len=1024 generated=94/128
+[decode] max_len=1024 generated=94/128
+[decode] max_len=1024 generated=95/128
+[decode] max_len=1024 generated=95/128
+[decode] max_len=1024 generated=96/128
+[decode] max_len=1024 generated=96/128
+[decode] max_len=1024 generated=97/128
+[decode] max_len=1024 generated=97/128
+[decode] max_len=1024 generated=98/128
+[decode] max_len=1024 generated=98/128
+[decode] max_len=1024 generated=99/128
+[decode] max_len=1024 generated=99/128
+[decode] max_len=1024 generated=100/128
+[decode] max_len=1024 generated=100/128
+[decode] max_len=1024 generated=101/128
+[decode] max_len=1024 generated=101/128
+[decode] max_len=1024 generated=102/128
+[decode] max_len=1024 generated=103/128
+[decode] max_len=1024 generated=102/128
+[decode] max_len=1024 generated=104/128
+[decode] max_len=1024 generated=103/128
+[decode] max_len=1024 generated=105/128
+[decode] max_len=1024 generated=104/128
+[decode] max_len=1024 generated=106/128
+[decode] max_len=1024 generated=105/128
+[decode] max_len=1024 generated=107/128
+[decode] max_len=1024 generated=106/128
+[decode] max_len=1024 generated=108/128
+[decode] max_len=1024 generated=107/128
+[decode] max_len=1024 generated=109/128
+[decode] max_len=1024 generated=108/128
+[decode] max_len=1024 generated=110/128
+[decode] max_len=1024 generated=109/128
+[decode] max_len=1024 generated=111/128
+[decode] max_len=1024 generated=110/128
+[decode] max_len=1024 generated=112/128
+[decode] max_len=1024 generated=111/128
+[decode] max_len=1024 generated=113/128
+[decode] max_len=1024 generated=112/128
+[decode] max_len=1024 generated=114/128
+[decode] max_len=1024 generated=113/128
+[decode] max_len=1024 generated=115/128
+[decode] max_len=1024 generated=114/128
+[decode] max_len=1024 generated=116/128
+[decode] max_len=1024 generated=115/128
+[decode] max_len=1024 generated=117/128
+[decode] max_len=1024 generated=116/128
+[decode] max_len=1024 generated=118/128
+[decode] max_len=1024 generated=117/128
+[decode] max_len=1024 generated=119/128
+[decode] max_len=1024 generated=118/128
+[decode] max_len=1024 generated=120/128
+[decode] max_len=1024 generated=119/128
+[decode] max_len=1024 generated=121/128
+[decode] max_len=1024 generated=120/128
+[decode] max_len=1024 generated=122/128
+[decode] max_len=1024 generated=121/128
+[decode] max_len=1024 generated=123/128
+[decode] max_len=1024 generated=122/128
+[decode] max_len=1024 generated=124/128
+[decode] max_len=1024 generated=123/128
+[decode] max_len=1024 generated=125/128
+[decode] max_len=1024 generated=124/128
+[decode] max_len=1024 generated=126/128
+[decode] max_len=1024 generated=125/128
+[decode] max_len=1024 generated=127/128
+[decode] max_len=1024 generated=126/128
+[decode] max_len=1024 generated=128/128
+[decode] max_len=1024 generated=127/128
+[decode] max_len=1024 generated=128/128
+[
+  {
+    "checkpoint": "runs/lta_owt_bert_absrope_time4_dirichlet_len1024_C1_to_1024_mask1_sameT_gbs512_b32_8gpu_1m_save10k_20260526/step_0040000.pt",
+    "ckpt_step": 40000,
+    "max_len": 1024,
+    "decode_rule": "dual_line_resample",
+    "support_power": 1.0,
+    "semantic_power": 1.5,
+    "steps": 128,
+    "c_min": 1.0,
+    "c_max": 1024.0,
+    "anchor_mode": "state",
+    "model_t_mode": "flow",
+    "time_schedule": "uniform",
+    "time_logit_mean": -1.5,
+    "time_logit_std": 0.8,
+    "time_power": 2.0,
+    "input_noise_scale": 0.0,
+    "input_noise_until": 1.0,
+    "input_noise_dirichlet_concentration": 1.0,
+    "endpoint_softening": "none",
+    "endpoint_soft_power": 2.0,
+    "endpoint_soft_min_conf": 0.0,
+    "endpoint_soft_max_conf": 1.0,
+    "soft_target_decode_mode": "off",
+    "soft_target_power": 1.0,
+    "soft_target_min_conf": 0.0,
+    "soft_target_max_conf": 1.0,
+    "soft_target_debias_start": 0.7,
+    "final_from": "blend",
+    "final_decode": "argmax",
+    "final_sample_temp": 1.0,
+    "final_top_k": 0,
+    "final_top_p": 1.0,
+    "commit_mode": "off",
+    "commit_conf_threshold": 0.0,
+    "commit_margin_threshold": 0.0,
+    "commit_start": 0.0,
+    "commit_min_ratio": 0.0,
+    "commit_max_ratio": 1.0,
+    "commit_power": 2.0,
+    "commit_freq_max_frac": 0.08,
+    "early_temp": 2.8,
+    "late_temp": 1.45,
+    "temp_end": 0.55,
+    "temp_power": 1.5,
+    "pos_extend": "repeat",
+    "fixed_first_token_id": null,
+    "fixed_first_token_text": "",
+    "fixed_first_initial_argmax": false,
+    "use_ema": false,
+    "n_samples": 128,
+    "sample_entropy": 0.5497349080951995,
+    "unique_tokens": 467,
+    "token_count": 131072,
+    "distinct_1": 0.00356292724609375,
+    "distinct_2": 0.018824841153470186,
+    "top_token_mass": 0.7095108032226562,
+    "texts_preview": [
+      "???????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????? —????????????? —??????????????????????? —???????????????????????????????????????????????????????????? —?? —??????? —??????????????? —??????? —? —???? — —?????????????????????????????????????? — — — — — —? —? — — —?????? —? —?? —? — —?????? —? —????? —????????????? —????????? —??????????????????????????????????????? —??????????????????????????????????????? —????????????? —? —?????? —???????? —??????????????? —?????????????????????????????????????????",
+      "[CLS]. korika, kojima kojima, kojima. korika. korika, kojima koi. korika, kojima kojima. korika. korika, - kojima kojima, kodaleni. kojima. koma kontinen, - - kojima korika, - kodaleni. kodaleni. kodaleni. kojima. korika kojima - korika. ko kojima, kona korika. kodalena. kodaleni kojima - kojima. kodaleni kojima. kodaleni. kodalena, - koanum. koriki, ko - kodaleni. ko kodaleni. ko kodaleni. kodalena. kodalena. kodalena. kojima, ko - kojima. kodalena, kojima - kojima, kojima. ko kodaleni. ko - koma. kodaleni. kojima. kodaleni. ko kodaleni. kodalena. ko kodaleni. kodalena. kodalenna. kodalena. kojima - kojima. kojima, kojima - ko - kojima. kojima -, -, - - - - - - -, - - - - - -,, - - - - - - -, - - - - -,, - - - -,, - ko - - -, - ko, - - - -, - - ko -, - - - -,,, ko ko - -, -,, ko -, - - - -, ko, - - - - - - - -, - ko - - -, ko - -, -, ko - - - - -,,, - - - - - - - - - - - - - - - - - - - - - - - -,, - - -, ko,,,,, ko - - - -,, - koa,, - - - - -,,, ko - - -, -,, koa - -, - - - -,,, - ko - - - - - -, - ko - - - - -, - ko - -,,, - - - - - - - - - - -,,, - - -,, ko -,,, ko - - -, -, ko,,,, ko -,,, - - - ko -,,,, ko, ko - - -, - - ko - - - - - -,, ko -, - -, -, - - - - - -, -, -, ko - - -, -, ko, ko - -, - - ko -, ko, - - -, -,,,, ko - - - - -, - - - -,, koma -ma, -, - - - ko -,,, ko -, -, ko,, ko - - -, ko,, ko - - -, ko, - - -, -, ko,, - ko - -,, -, ko - - -,,,, ko ko - - - - - -,, -, - ko - -,, ko ko ko -, - - -, -, ko -.ka, ko, ko - chei,.,.,, :, koka - -,, - - - ko, -, - - koma, - ko - - ko - -, ko, koku kokuma -,,, ko, koku - ko -, ko, - -, ko,, -,ma - - -,, - ko -, ko - ko ko - -,,ka, ko, - - ko - ko, -, - -, -, - - - - - - -, ko ko, koa, - -, -, ko, ko,, - - - - ko -,,, ko, ko ko,. -, ko, ko, -, -, -, ko, - - -,, ko.,. -, -, - - - - - - - ko, - - - -,., ko - koka,. - ko ko ko -,ma -... - ko - - - -, -,,, - - - ko -, -, ko ko - -ma, -, ko -, -, ko, - - - - ko ko - - - -,, - - - -, ko - - - - -,, - - ko ko -, -, ko - - - - - -, - - - - ko,, ko ko - koa, -, - - -,,, ko, ko - - - - ko -,. -, -,, -, ko, ko - - ko, - [SEP]",
+      "??????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????? —??????? — —?????????????????????????????????????????????????????????????????????????????????????????????????????? —??? —??????????????????? —???? —????????????????????????? —?????????????????????????????????????????????????????????????????? — —? — — —? —????????????????????? —? —? —??? —?? —???? —??????? —?? —???? — —?? —??? — — —????????????? —? —?? — —? —??????? —??????? — —?????? —???? —???? —????? — —??? — — — — —? —? — — — —??? —? —? —??????? — —?????? — —?????? —? —???? — — — — —? — — — — — — — —??? —?? — —? — — — — — — — — —? —?? —? —?? —?????????????? — —??? —??? —??????????????????????? —??????? —????????? —????????????? —?????????????? —????????? —?? —??? —??????? —????? —? — —??? —?? —??????? — —?????? —?? — —??? —?????????????????????????????????",
+      "?????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????? —?????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????"
+    ],
+    "gen_ppl": 4.906654522033406,
+    "gen_nll": 1.590592349500496,
+    "gen_tokens": 46838
+  }
+]
+[watch-gumbel] 2026-05-26_08:15:10 done step_0040000
+[
+  {
+    "checkpoint": "runs/lta_owt_bert_absrope_time4_dirichlet_len1024_C1_to_1024_mask1_sameT_gbs512_b32_8gpu_1m_save10k_20260526/step_0040000.pt",
+    "ckpt_step": 40000,
+    "max_len": 1024,
+    "decode_rule": "dual_line_resample",
+    "support_power": 1.0,
+    "semantic_power": 1.5,
+    "steps": 128,
+    "c_min": 1.0,
+    "c_max": 1024.0,
+    "anchor_mode": "state",
+    "model_t_mode": "flow",
+    "time_schedule": "uniform",
+    "time_logit_mean": -1.5,
+    "time_logit_std": 0.8,
+    "time_power": 2.0,
+    "input_noise_scale": 0.0,
+    "input_noise_until": 1.0,
+    "input_noise_dirichlet_concentration": 1.0,
+    "endpoint_softening": "none",
+    "endpoint_soft_power": 2.0,
+    "endpoint_soft_min_conf": 0.0,
+    "endpoint_soft_max_conf": 1.0,
+    "soft_target_decode_mode": "off",
+    "soft_target_power": 1.0,
+    "soft_target_min_conf": 0.0,
+    "soft_target_max_conf": 1.0,
+    "soft_target_debias_start": 0.7,
+    "final_from": "blend",
+    "final_decode": "argmax",
+    "final_sample_temp": 1.0,
+    "final_top_k": 0,
+    "final_top_p": 1.0,
+    "commit_mode": "off",
+    "commit_conf_threshold": 0.0,
+    "commit_margin_threshold": 0.0,
+    "commit_start": 0.0,
+    "commit_min_ratio": 0.0,
+    "commit_max_ratio": 1.0,
+    "commit_power": 2.0,
+    "commit_freq_max_frac": 0.08,
+    "early_temp": 2.8,
+    "late_temp": 1.45,
+    "temp_end": 0.55,
+    "temp_power": 1.5,
+    "pos_extend": "repeat",
+    "fixed_first_token_id": null,
+    "fixed_first_token_text": "",
+    "fixed_first_initial_argmax": false,
+    "use_ema": false,
+    "n_samples": 128,
+    "sample_entropy": 0.5497349080951995,
+    "unique_tokens": 467,
+    "token_count": 131072,
+    "distinct_1": 0.00356292724609375,
+    "distinct_2": 0.018824841153470186,
+    "top_token_mass": 0.7095108032226562,
+    "texts_preview": [
+      "???????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????? —????????????? —??????????????????????? —???????????????????????????????????????????????????????????? —?? —??????? —??????????????? —??????? —? —???? — —?????????????????????????????????????? — — — — — —? —? — — —?????? —? —?? —? — —?????? —? —????? —????????????? —????????? —??????????????????????????????????????? —??????????????????????????????????????? —????????????? —? —?????? —???????? —??????????????? —?????????????????????????????????????????",
+      "[CLS]. korika, kojima kojima, kojima. korika. korika, kojima koi. korika, kojima kojima. korika. korika, - kojima kojima, kodaleni. kojima. koma kontinen, - - kojima korika, - kodaleni. kodaleni. kodaleni. kojima. korika kojima - korika. ko kojima, kona korika. kodalena. kodaleni kojima - kojima. kodaleni kojima. kodaleni. kodalena, - koanum. koriki, ko - kodaleni. ko kodaleni. ko kodaleni. kodalena. kodalena. kodalena. kojima, ko - kojima. kodalena, kojima - kojima, kojima. ko kodaleni. ko - koma. kodaleni. kojima. kodaleni. ko kodaleni. kodalena. ko kodaleni. kodalena. kodalenna. kodalena. kojima - kojima. kojima, kojima - ko - kojima. kojima -, -, - - - - - - -, - - - - - -,, - - - - - - -, - - - - -,, - - - -,, - ko - - -, - ko, - - - -, - - ko -, - - - -,,, ko ko - -, -,, ko -, - - - -, ko, - - - - - - - -, - ko - - -, ko - -, -, ko - - - - -,,, - - - - - - - - - - - - - - - - - - - - - - - -,, - - -, ko,,,,, ko - - - -,, - koa,, - - - - -,,, ko - - -, -,, koa - -, - - - -,,, - ko - - - - - -, - ko - - - - -, - ko - -,,, - - - - - - - - - - -,,, - - -,, ko -,,, ko - - -, -, ko,,,, ko -,,, - - - ko -,,,, ko, ko - - -, - - ko - - - - - -,, ko -, - -, -, - - - - - -, -, -, ko - - -, -, ko, ko - -, - - ko -, ko, - - -, -,,,, ko - - - - -, - - - -,, koma -ma, -, - - - ko -,,, ko -, -, ko,, ko - - -, ko,, ko - - -, ko, - - -, -, ko,, - ko - -,, -, ko - - -,,,, ko ko - - - - - -,, -, - ko - -,, ko ko ko -, - - -, -, ko -.ka, ko, ko - chei,.,.,, :, koka - -,, - - - ko, -, - - koma, - ko - - ko - -, ko, koku kokuma -,,, ko, koku - ko -, ko, - -, ko,, -,ma - - -,, - ko -, ko - ko ko - -,,ka, ko, - - ko - ko, -, - -, -, - - - - - - -, ko ko, koa, - -, -, ko, ko,, - - - - ko -,,, ko, ko ko,. -, ko, ko, -, -, -, ko, - - -,, ko.,. -, -, - - - - - - - ko, - - - -,., ko - koka,. - ko ko ko -,ma -... - ko - - - -, -,,, - - - ko -, -, ko ko - -ma, -, ko -, -, ko, - - - - ko ko - - - -,, - - - -, ko - - - - -,, - - ko ko -, -, ko - - - - - -, - - - - ko,, ko ko - koa, -, - - -,,, ko, ko - - - - ko -,. -, -,, -, ko, ko - - ko, - [SEP]",
+      "??????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????? —??????? — —?????????????????????????????????????????????????????????????????????????????????????????????????????? —??? —??????????????????? —???? —????????????????????????? —?????????????????????????????????????????????????????????????????? — —? — — —? —????????????????????? —? —? —??? —?? —???? —??????? —?? —???? — —?? —??? — — —????????????? —? —?? — —? —??????? —??????? — —?????? —???? —???? —????? — —??? — — — — —? —? — — — —??? —? —? —??????? — —?????? — —?????? —? —???? — — — — —? — — — — — — — —??? —?? — —? — — — — — — — — —? —?? —? —?? —?????????????? — —??? —??? —??????????????????????? —??????? —????????? —????????????? —?????????????? —????????? —?? —??? —??????? —????? —? — —??? —?? —??????? — —?????? —?? — —??? —?????????????????????????????????",
+      "?????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????? —?????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????"
+    ],
+    "gen_ppl": 4.906654522033406,
+    "gen_nll": 1.590592349500496,
+    "gen_tokens": 46838
+  }
+]
+[watch-gumbel] 2026-05-26_08:15:15 done step_0040000

LTA_openwebtext_dualt/logs/owt_bert_absrope_time4_C1_to_1024_mask1_sameT_dualline_watch/infer_lta_owt_bert_absrope_time4_dirichlet_len1024_C1_to_1024_mask1_sameT_gbs512_b32_8gpu_1m_save10k_20260526_step_0050000.log

@@ -0,0 +1,398 @@
+[watch-gumbel] 2026-05-26_09:07:11 infer runs/lta_owt_bert_absrope_time4_dirichlet_len1024_C1_to_1024_mask1_sameT_gbs512_b32_8gpu_1m_save10k_20260526/step_0050000.pt -> docs/lta_samples/metrics_20260526/owt_bert_absrope_time4_C1_to_1024_mask1_sameT_dualline_dirres_c1_1024_n128/lta_owt_bert_absrope_time4_dirichlet_len1024_C1_to_1024_mask1_sameT_gbs512_b32_8gpu_1m_save10k_20260526/step_0050000
+[watch-gumbel] 2026-05-26_09:07:17 infer runs/lta_owt_bert_absrope_time4_dirichlet_len1024_C1_to_1024_mask1_sameT_gbs512_b32_8gpu_1m_save10k_20260526/step_0050000.pt -> docs/lta_samples/metrics_20260526/owt_bert_absrope_time4_C1_to_1024_mask1_sameT_dualline_dirres_c1_1024_n128/lta_owt_bert_absrope_time4_dirichlet_len1024_C1_to_1024_mask1_sameT_gbs512_b32_8gpu_1m_save10k_20260526/step_0050000
+[decode] max_len=1024 generated=1/128
+[decode] max_len=1024 generated=2/128
+[decode] max_len=1024 generated=3/128
+[decode] max_len=1024 generated=1/128
+[decode] max_len=1024 generated=4/128
+[decode] max_len=1024 generated=2/128
+[decode] max_len=1024 generated=5/128
+[decode] max_len=1024 generated=3/128
+[decode] max_len=1024 generated=6/128
+[decode] max_len=1024 generated=4/128
+[decode] max_len=1024 generated=7/128
+[decode] max_len=1024 generated=5/128
+[decode] max_len=1024 generated=8/128
+[decode] max_len=1024 generated=6/128
+[decode] max_len=1024 generated=9/128
+[decode] max_len=1024 generated=7/128
+[decode] max_len=1024 generated=10/128
+[decode] max_len=1024 generated=8/128
+[decode] max_len=1024 generated=11/128
+[decode] max_len=1024 generated=9/128
+[decode] max_len=1024 generated=12/128
+[decode] max_len=1024 generated=10/128
+[decode] max_len=1024 generated=13/128
+[decode] max_len=1024 generated=11/128
+[decode] max_len=1024 generated=14/128
+[decode] max_len=1024 generated=12/128
+[decode] max_len=1024 generated=15/128
+[decode] max_len=1024 generated=13/128
+[decode] max_len=1024 generated=16/128
+[decode] max_len=1024 generated=14/128
+[decode] max_len=1024 generated=17/128
+[decode] max_len=1024 generated=15/128
+[decode] max_len=1024 generated=18/128
+[decode] max_len=1024 generated=16/128
+[decode] max_len=1024 generated=19/128
+[decode] max_len=1024 generated=17/128
+[decode] max_len=1024 generated=20/128
+[decode] max_len=1024 generated=18/128
+[decode] max_len=1024 generated=21/128
+[decode] max_len=1024 generated=19/128
+[decode] max_len=1024 generated=22/128
+[decode] max_len=1024 generated=20/128
+[decode] max_len=1024 generated=23/128
+[decode] max_len=1024 generated=21/128
+[decode] max_len=1024 generated=24/128
+[decode] max_len=1024 generated=22/128
+[decode] max_len=1024 generated=25/128
+[decode] max_len=1024 generated=23/128
+[decode] max_len=1024 generated=26/128
+[decode] max_len=1024 generated=24/128
+[decode] max_len=1024 generated=27/128
+[decode] max_len=1024 generated=25/128
+[decode] max_len=1024 generated=28/128
+[decode] max_len=1024 generated=26/128
+[decode] max_len=1024 generated=29/128
+[decode] max_len=1024 generated=27/128
+[decode] max_len=1024 generated=30/128
+[decode] max_len=1024 generated=28/128
+[decode] max_len=1024 generated=31/128
+[decode] max_len=1024 generated=29/128
+[decode] max_len=1024 generated=32/128
+[decode] max_len=1024 generated=30/128
+[decode] max_len=1024 generated=33/128
+[decode] max_len=1024 generated=31/128
+[decode] max_len=1024 generated=34/128
+[decode] max_len=1024 generated=32/128
+[decode] max_len=1024 generated=35/128
+[decode] max_len=1024 generated=33/128
+[decode] max_len=1024 generated=36/128
+[decode] max_len=1024 generated=34/128
+[decode] max_len=1024 generated=37/128
+[decode] max_len=1024 generated=35/128
+[decode] max_len=1024 generated=38/128
+[decode] max_len=1024 generated=36/128
+[decode] max_len=1024 generated=39/128
+[decode] max_len=1024 generated=37/128
+[decode] max_len=1024 generated=40/128
+[decode] max_len=1024 generated=38/128
+[decode] max_len=1024 generated=41/128
+[decode] max_len=1024 generated=39/128
+[decode] max_len=1024 generated=42/128
+[decode] max_len=1024 generated=40/128
+[decode] max_len=1024 generated=43/128
+[decode] max_len=1024 generated=41/128
+[decode] max_len=1024 generated=44/128
+[decode] max_len=1024 generated=42/128
+[decode] max_len=1024 generated=45/128
+[decode] max_len=1024 generated=43/128
+[decode] max_len=1024 generated=46/128
+[decode] max_len=1024 generated=44/128
+[decode] max_len=1024 generated=47/128
+[decode] max_len=1024 generated=45/128
+[decode] max_len=1024 generated=48/128
+[decode] max_len=1024 generated=46/128
+[decode] max_len=1024 generated=49/128
+[decode] max_len=1024 generated=47/128
+[decode] max_len=1024 generated=50/128
+[decode] max_len=1024 generated=48/128
+[decode] max_len=1024 generated=51/128
+[decode] max_len=1024 generated=49/128
+[decode] max_len=1024 generated=52/128
+[decode] max_len=1024 generated=50/128
+[decode] max_len=1024 generated=53/128
+[decode] max_len=1024 generated=51/128
+[decode] max_len=1024 generated=54/128
+[decode] max_len=1024 generated=52/128
+[decode] max_len=1024 generated=55/128
+[decode] max_len=1024 generated=53/128
+[decode] max_len=1024 generated=56/128
+[decode] max_len=1024 generated=54/128
+[decode] max_len=1024 generated=57/128
+[decode] max_len=1024 generated=55/128
+[decode] max_len=1024 generated=58/128
+[decode] max_len=1024 generated=56/128
+[decode] max_len=1024 generated=59/128
+[decode] max_len=1024 generated=57/128
+[decode] max_len=1024 generated=60/128
+[decode] max_len=1024 generated=58/128
+[decode] max_len=1024 generated=61/128
+[decode] max_len=1024 generated=59/128
+[decode] max_len=1024 generated=62/128
+[decode] max_len=1024 generated=60/128
+[decode] max_len=1024 generated=63/128
+[decode] max_len=1024 generated=61/128
+[decode] max_len=1024 generated=64/128
+[decode] max_len=1024 generated=62/128
+[decode] max_len=1024 generated=65/128
+[decode] max_len=1024 generated=63/128
+[decode] max_len=1024 generated=66/128
+[decode] max_len=1024 generated=64/128
+[decode] max_len=1024 generated=67/128
+[decode] max_len=1024 generated=65/128
+[decode] max_len=1024 generated=68/128
+[decode] max_len=1024 generated=66/128
+[decode] max_len=1024 generated=69/128
+[decode] max_len=1024 generated=67/128
+[decode] max_len=1024 generated=70/128
+[decode] max_len=1024 generated=68/128
+[decode] max_len=1024 generated=71/128
+[decode] max_len=1024 generated=69/128
+[decode] max_len=1024 generated=72/128
+[decode] max_len=1024 generated=70/128
+[decode] max_len=1024 generated=73/128
+[decode] max_len=1024 generated=71/128
+[decode] max_len=1024 generated=74/128
+[decode] max_len=1024 generated=72/128
+[decode] max_len=1024 generated=75/128
+[decode] max_len=1024 generated=73/128
+[decode] max_len=1024 generated=76/128
+[decode] max_len=1024 generated=74/128
+[decode] max_len=1024 generated=77/128
+[decode] max_len=1024 generated=75/128
+[decode] max_len=1024 generated=78/128
+[decode] max_len=1024 generated=76/128
+[decode] max_len=1024 generated=79/128
+[decode] max_len=1024 generated=77/128
+[decode] max_len=1024 generated=80/128
+[decode] max_len=1024 generated=78/128
+[decode] max_len=1024 generated=81/128
+[decode] max_len=1024 generated=79/128
+[decode] max_len=1024 generated=82/128
+[decode] max_len=1024 generated=80/128
+[decode] max_len=1024 generated=83/128
+[decode] max_len=1024 generated=81/128
+[decode] max_len=1024 generated=84/128
+[decode] max_len=1024 generated=82/128
+[decode] max_len=1024 generated=85/128
+[decode] max_len=1024 generated=83/128
+[decode] max_len=1024 generated=86/128
+[decode] max_len=1024 generated=84/128
+[decode] max_len=1024 generated=87/128
+[decode] max_len=1024 generated=85/128
+[decode] max_len=1024 generated=88/128
+[decode] max_len=1024 generated=86/128
+[decode] max_len=1024 generated=89/128
+[decode] max_len=1024 generated=87/128
+[decode] max_len=1024 generated=90/128
+[decode] max_len=1024 generated=88/128
+[decode] max_len=1024 generated=91/128
+[decode] max_len=1024 generated=89/128
+[decode] max_len=1024 generated=92/128
+[decode] max_len=1024 generated=90/128
+[decode] max_len=1024 generated=93/128
+[decode] max_len=1024 generated=91/128
+[decode] max_len=1024 generated=94/128
+[decode] max_len=1024 generated=92/128
+[decode] max_len=1024 generated=95/128
+[decode] max_len=1024 generated=93/128
+[decode] max_len=1024 generated=96/128
+[decode] max_len=1024 generated=94/128
+[decode] max_len=1024 generated=97/128
+[decode] max_len=1024 generated=95/128
+[decode] max_len=1024 generated=98/128
+[decode] max_len=1024 generated=96/128
+[decode] max_len=1024 generated=99/128
+[decode] max_len=1024 generated=97/128
+[decode] max_len=1024 generated=100/128
+[decode] max_len=1024 generated=98/128
+[decode] max_len=1024 generated=101/128
+[decode] max_len=1024 generated=99/128
+[decode] max_len=1024 generated=102/128
+[decode] max_len=1024 generated=100/128
+[decode] max_len=1024 generated=103/128
+[decode] max_len=1024 generated=101/128
+[decode] max_len=1024 generated=104/128
+[decode] max_len=1024 generated=102/128
+[decode] max_len=1024 generated=105/128
+[decode] max_len=1024 generated=103/128
+[decode] max_len=1024 generated=106/128
+[decode] max_len=1024 generated=104/128
+[decode] max_len=1024 generated=107/128
+[decode] max_len=1024 generated=105/128
+[decode] max_len=1024 generated=108/128
+[decode] max_len=1024 generated=106/128
+[decode] max_len=1024 generated=109/128
+[decode] max_len=1024 generated=107/128
+[decode] max_len=1024 generated=110/128
+[decode] max_len=1024 generated=108/128
+[decode] max_len=1024 generated=111/128
+[decode] max_len=1024 generated=109/128
+[decode] max_len=1024 generated=112/128
+[decode] max_len=1024 generated=110/128
+[decode] max_len=1024 generated=113/128
+[decode] max_len=1024 generated=111/128
+[decode] max_len=1024 generated=114/128
+[decode] max_len=1024 generated=112/128
+[decode] max_len=1024 generated=115/128
+[decode] max_len=1024 generated=113/128
+[decode] max_len=1024 generated=116/128
+[decode] max_len=1024 generated=114/128
+[decode] max_len=1024 generated=117/128
+[decode] max_len=1024 generated=115/128
+[decode] max_len=1024 generated=118/128
+[decode] max_len=1024 generated=116/128
+[decode] max_len=1024 generated=119/128
+[decode] max_len=1024 generated=117/128
+[decode] max_len=1024 generated=120/128
+[decode] max_len=1024 generated=118/128
+[decode] max_len=1024 generated=121/128
+[decode] max_len=1024 generated=119/128
+[decode] max_len=1024 generated=122/128
+[decode] max_len=1024 generated=120/128
+[decode] max_len=1024 generated=123/128
+[decode] max_len=1024 generated=121/128
+[decode] max_len=1024 generated=124/128
+[decode] max_len=1024 generated=122/128
+[decode] max_len=1024 generated=125/128
+[decode] max_len=1024 generated=123/128
+[decode] max_len=1024 generated=126/128
+[decode] max_len=1024 generated=124/128
+[decode] max_len=1024 generated=127/128
+[decode] max_len=1024 generated=125/128
+[decode] max_len=1024 generated=128/128
+[decode] max_len=1024 generated=126/128
+[decode] max_len=1024 generated=127/128
+[decode] max_len=1024 generated=128/128
+[
+  {
+    "checkpoint": "runs/lta_owt_bert_absrope_time4_dirichlet_len1024_C1_to_1024_mask1_sameT_gbs512_b32_8gpu_1m_save10k_20260526/step_0050000.pt",
+    "ckpt_step": 50000,
+    "max_len": 1024,
+    "decode_rule": "dual_line_resample",
+    "support_power": 1.0,
+    "semantic_power": 1.5,
+    "steps": 128,
+    "c_min": 1.0,
+    "c_max": 1024.0,
+    "anchor_mode": "state",
+    "model_t_mode": "flow",
+    "time_schedule": "uniform",
+    "time_logit_mean": -1.5,
+    "time_logit_std": 0.8,
+    "time_power": 2.0,
+    "input_noise_scale": 0.0,
+    "input_noise_until": 1.0,
+    "input_noise_dirichlet_concentration": 1.0,
+    "endpoint_softening": "none",
+    "endpoint_soft_power": 2.0,
+    "endpoint_soft_min_conf": 0.0,
+    "endpoint_soft_max_conf": 1.0,
+    "soft_target_decode_mode": "off",
+    "soft_target_power": 1.0,
+    "soft_target_min_conf": 0.0,
+    "soft_target_max_conf": 1.0,
+    "soft_target_debias_start": 0.7,
+    "final_from": "blend",
+    "final_decode": "argmax",
+    "final_sample_temp": 1.0,
+    "final_top_k": 0,
+    "final_top_p": 1.0,
+    "commit_mode": "off",
+    "commit_conf_threshold": 0.0,
+    "commit_margin_threshold": 0.0,
+    "commit_start": 0.0,
+    "commit_min_ratio": 0.0,
+    "commit_max_ratio": 1.0,
+    "commit_power": 2.0,
+    "commit_freq_max_frac": 0.08,
+    "early_temp": 2.8,
+    "late_temp": 1.45,
+    "temp_end": 0.55,
+    "temp_power": 1.5,
+    "pos_extend": "repeat",
+    "fixed_first_token_id": null,
+    "fixed_first_token_text": "",
+    "fixed_first_initial_argmax": false,
+    "use_ema": false,
+    "n_samples": 128,
+    "sample_entropy": 0.5168811757259412,
+    "unique_tokens": 377,
+    "token_count": 131072,
+    "distinct_1": 0.00287628173828125,
+    "distinct_2": 0.017618218475073315,
+    "top_token_mass": 0.7140960693359375,
+    "texts_preview": [
+      "i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i",
+      "[CLS],,,,,,,,,,,,,,,,,,,,,,,,, oxidation oxidation oxidation oxidation oxidation oxidation oxidation oxidation oxidation oxidation oxidation oxidation oxidation oxidation oxidation oxidation oxidation oxidation oxidation oxidation oxidation oxidation oxidation oxidation oxidation oxidation oxidation oxidation,,, oxidation oxidation oxidation,,√やややややややややややや oxidation,やややややややややややややややややややややややややややややχややややややややややや,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,, [SEP],,,,,,,,,,,,,,,,,,,,,,,, [SEP]",
+      "i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i",
+      "i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i"
+    ],
+    "gen_ppl": 1.914934104351525,
+    "gen_nll": 0.6496832117802624,
+    "gen_tokens": 120908
+  }
+]
+[watch-gumbel] 2026-05-26_09:13:45 done step_0050000
+[
+  {
+    "checkpoint": "runs/lta_owt_bert_absrope_time4_dirichlet_len1024_C1_to_1024_mask1_sameT_gbs512_b32_8gpu_1m_save10k_20260526/step_0050000.pt",
+    "ckpt_step": 50000,
+    "max_len": 1024,
+    "decode_rule": "dual_line_resample",
+    "support_power": 1.0,
+    "semantic_power": 1.5,
+    "steps": 128,
+    "c_min": 1.0,
+    "c_max": 1024.0,
+    "anchor_mode": "state",
+    "model_t_mode": "flow",
+    "time_schedule": "uniform",
+    "time_logit_mean": -1.5,
+    "time_logit_std": 0.8,
+    "time_power": 2.0,
+    "input_noise_scale": 0.0,
+    "input_noise_until": 1.0,
+    "input_noise_dirichlet_concentration": 1.0,
+    "endpoint_softening": "none",
+    "endpoint_soft_power": 2.0,
+    "endpoint_soft_min_conf": 0.0,
+    "endpoint_soft_max_conf": 1.0,
+    "soft_target_decode_mode": "off",
+    "soft_target_power": 1.0,
+    "soft_target_min_conf": 0.0,
+    "soft_target_max_conf": 1.0,
+    "soft_target_debias_start": 0.7,
+    "final_from": "blend",
+    "final_decode": "argmax",
+    "final_sample_temp": 1.0,
+    "final_top_k": 0,
+    "final_top_p": 1.0,
+    "commit_mode": "off",
+    "commit_conf_threshold": 0.0,
+    "commit_margin_threshold": 0.0,
+    "commit_start": 0.0,
+    "commit_min_ratio": 0.0,
+    "commit_max_ratio": 1.0,
+    "commit_power": 2.0,
+    "commit_freq_max_frac": 0.08,
+    "early_temp": 2.8,
+    "late_temp": 1.45,
+    "temp_end": 0.55,
+    "temp_power": 1.5,
+    "pos_extend": "repeat",
+    "fixed_first_token_id": null,
+    "fixed_first_token_text": "",
+    "fixed_first_initial_argmax": false,
+    "use_ema": false,
+    "n_samples": 128,
+    "sample_entropy": 0.5168811757259412,
+    "unique_tokens": 377,
+    "token_count": 131072,
+    "distinct_1": 0.00287628173828125,
+    "distinct_2": 0.017618218475073315,
+    "top_token_mass": 0.7140960693359375,
+    "texts_preview": [
+      "i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i",
+      "[CLS],,,,,,,,,,,,,,,,,,,,,,,,, oxidation oxidation oxidation oxidation oxidation oxidation oxidation oxidation oxidation oxidation oxidation oxidation oxidation oxidation oxidation oxidation oxidation oxidation oxidation oxidation oxidation oxidation oxidation oxidation oxidation oxidation oxidation oxidation,,, oxidation oxidation oxidation,,√やややややややややややや oxidation,やややややややややややややややややややややややややややややχややややややややややや,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,, [SEP],,,,,,,,,,,,,,,,,,,,,,,, [SEP]",
+      "i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i",
+      "i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i"
+    ],
+    "gen_ppl": 1.914934104351525,
+    "gen_nll": 0.6496832117802624,
+    "gen_tokens": 120908
+  }
+]
+[watch-gumbel] 2026-05-26_09:13:52 done step_0050000

LTA_openwebtext_dualt/logs/owt_bert_absrope_time4_C1_to_1024_mask1_sameT_dualline_watch/infer_lta_owt_bert_absrope_time4_dirichlet_len1024_C1_to_1024_mask1_sameT_gbs512_b32_8gpu_1m_save10k_20260526_step_0130000.log

@@ -0,0 +1,398 @@
+[watch-gumbel] 2026-05-26_16:53:04 infer runs/lta_owt_bert_absrope_time4_dirichlet_len1024_C1_to_1024_mask1_sameT_gbs512_b32_8gpu_1m_save10k_20260526/step_0130000.pt -> docs/lta_samples/metrics_20260526/owt_bert_absrope_time4_C1_to_1024_mask1_sameT_dualline_dirres_c1_1024_n128/lta_owt_bert_absrope_time4_dirichlet_len1024_C1_to_1024_mask1_sameT_gbs512_b32_8gpu_1m_save10k_20260526/step_0130000
+[watch-gumbel] 2026-05-26_16:53:08 infer runs/lta_owt_bert_absrope_time4_dirichlet_len1024_C1_to_1024_mask1_sameT_gbs512_b32_8gpu_1m_save10k_20260526/step_0130000.pt -> docs/lta_samples/metrics_20260526/owt_bert_absrope_time4_C1_to_1024_mask1_sameT_dualline_dirres_c1_1024_n128/lta_owt_bert_absrope_time4_dirichlet_len1024_C1_to_1024_mask1_sameT_gbs512_b32_8gpu_1m_save10k_20260526/step_0130000
+[decode] max_len=1024 generated=1/128
+[decode] max_len=1024 generated=2/128
+[decode] max_len=1024 generated=1/128
+[decode] max_len=1024 generated=3/128
+[decode] max_len=1024 generated=2/128
+[decode] max_len=1024 generated=4/128
+[decode] max_len=1024 generated=3/128
+[decode] max_len=1024 generated=5/128
+[decode] max_len=1024 generated=4/128
+[decode] max_len=1024 generated=6/128
+[decode] max_len=1024 generated=5/128
+[decode] max_len=1024 generated=7/128
+[decode] max_len=1024 generated=6/128
+[decode] max_len=1024 generated=8/128
+[decode] max_len=1024 generated=7/128
+[decode] max_len=1024 generated=9/128
+[decode] max_len=1024 generated=8/128
+[decode] max_len=1024 generated=10/128
+[decode] max_len=1024 generated=9/128
+[decode] max_len=1024 generated=11/128
+[decode] max_len=1024 generated=10/128
+[decode] max_len=1024 generated=12/128
+[decode] max_len=1024 generated=11/128
+[decode] max_len=1024 generated=13/128
+[decode] max_len=1024 generated=12/128
+[decode] max_len=1024 generated=14/128
+[decode] max_len=1024 generated=13/128
+[decode] max_len=1024 generated=15/128
+[decode] max_len=1024 generated=14/128
+[decode] max_len=1024 generated=16/128
+[decode] max_len=1024 generated=15/128
+[decode] max_len=1024 generated=17/128
+[decode] max_len=1024 generated=16/128
+[decode] max_len=1024 generated=18/128
+[decode] max_len=1024 generated=17/128
+[decode] max_len=1024 generated=19/128
+[decode] max_len=1024 generated=18/128
+[decode] max_len=1024 generated=20/128
+[decode] max_len=1024 generated=19/128
+[decode] max_len=1024 generated=21/128
+[decode] max_len=1024 generated=20/128
+[decode] max_len=1024 generated=22/128
+[decode] max_len=1024 generated=21/128
+[decode] max_len=1024 generated=23/128
+[decode] max_len=1024 generated=22/128
+[decode] max_len=1024 generated=24/128
+[decode] max_len=1024 generated=23/128
+[decode] max_len=1024 generated=25/128
+[decode] max_len=1024 generated=24/128
+[decode] max_len=1024 generated=26/128
+[decode] max_len=1024 generated=25/128
+[decode] max_len=1024 generated=27/128
+[decode] max_len=1024 generated=26/128
+[decode] max_len=1024 generated=28/128
+[decode] max_len=1024 generated=27/128
+[decode] max_len=1024 generated=29/128
+[decode] max_len=1024 generated=28/128
+[decode] max_len=1024 generated=30/128
+[decode] max_len=1024 generated=29/128
+[decode] max_len=1024 generated=31/128
+[decode] max_len=1024 generated=30/128
+[decode] max_len=1024 generated=32/128
+[decode] max_len=1024 generated=31/128
+[decode] max_len=1024 generated=33/128
+[decode] max_len=1024 generated=32/128
+[decode] max_len=1024 generated=34/128
+[decode] max_len=1024 generated=33/128
+[decode] max_len=1024 generated=35/128
+[decode] max_len=1024 generated=34/128
+[decode] max_len=1024 generated=36/128
+[decode] max_len=1024 generated=35/128
+[decode] max_len=1024 generated=37/128
+[decode] max_len=1024 generated=36/128
+[decode] max_len=1024 generated=38/128
+[decode] max_len=1024 generated=37/128
+[decode] max_len=1024 generated=39/128
+[decode] max_len=1024 generated=38/128
+[decode] max_len=1024 generated=40/128
+[decode] max_len=1024 generated=39/128
+[decode] max_len=1024 generated=41/128
+[decode] max_len=1024 generated=40/128
+[decode] max_len=1024 generated=42/128
+[decode] max_len=1024 generated=41/128
+[decode] max_len=1024 generated=43/128
+[decode] max_len=1024 generated=42/128
+[decode] max_len=1024 generated=44/128
+[decode] max_len=1024 generated=43/128
+[decode] max_len=1024 generated=45/128
+[decode] max_len=1024 generated=44/128
+[decode] max_len=1024 generated=46/128
+[decode] max_len=1024 generated=45/128
+[decode] max_len=1024 generated=47/128
+[decode] max_len=1024 generated=46/128
+[decode] max_len=1024 generated=48/128
+[decode] max_len=1024 generated=47/128
+[decode] max_len=1024 generated=49/128
+[decode] max_len=1024 generated=48/128
+[decode] max_len=1024 generated=50/128
+[decode] max_len=1024 generated=49/128
+[decode] max_len=1024 generated=51/128
+[decode] max_len=1024 generated=50/128
+[decode] max_len=1024 generated=52/128
+[decode] max_len=1024 generated=51/128
+[decode] max_len=1024 generated=53/128
+[decode] max_len=1024 generated=52/128
+[decode] max_len=1024 generated=54/128
+[decode] max_len=1024 generated=53/128
+[decode] max_len=1024 generated=55/128
+[decode] max_len=1024 generated=54/128
+[decode] max_len=1024 generated=56/128
+[decode] max_len=1024 generated=55/128
+[decode] max_len=1024 generated=57/128
+[decode] max_len=1024 generated=56/128
+[decode] max_len=1024 generated=58/128
+[decode] max_len=1024 generated=57/128
+[decode] max_len=1024 generated=59/128
+[decode] max_len=1024 generated=58/128
+[decode] max_len=1024 generated=60/128
+[decode] max_len=1024 generated=59/128
+[decode] max_len=1024 generated=61/128
+[decode] max_len=1024 generated=60/128
+[decode] max_len=1024 generated=61/128
+[decode] max_len=1024 generated=62/128
+[decode] max_len=1024 generated=62/128
+[decode] max_len=1024 generated=63/128
+[decode] max_len=1024 generated=63/128
+[decode] max_len=1024 generated=64/128
+[decode] max_len=1024 generated=64/128
+[decode] max_len=1024 generated=65/128
+[decode] max_len=1024 generated=65/128
+[decode] max_len=1024 generated=66/128
+[decode] max_len=1024 generated=66/128
+[decode] max_len=1024 generated=67/128
+[decode] max_len=1024 generated=67/128
+[decode] max_len=1024 generated=68/128
+[decode] max_len=1024 generated=68/128
+[decode] max_len=1024 generated=69/128
+[decode] max_len=1024 generated=69/128
+[decode] max_len=1024 generated=70/128
+[decode] max_len=1024 generated=70/128
+[decode] max_len=1024 generated=71/128
+[decode] max_len=1024 generated=71/128
+[decode] max_len=1024 generated=72/128
+[decode] max_len=1024 generated=72/128
+[decode] max_len=1024 generated=73/128
+[decode] max_len=1024 generated=73/128
+[decode] max_len=1024 generated=74/128
+[decode] max_len=1024 generated=74/128
+[decode] max_len=1024 generated=75/128
+[decode] max_len=1024 generated=75/128
+[decode] max_len=1024 generated=76/128
+[decode] max_len=1024 generated=76/128
+[decode] max_len=1024 generated=77/128
+[decode] max_len=1024 generated=77/128
+[decode] max_len=1024 generated=78/128
+[decode] max_len=1024 generated=78/128
+[decode] max_len=1024 generated=79/128
+[decode] max_len=1024 generated=79/128
+[decode] max_len=1024 generated=80/128
+[decode] max_len=1024 generated=80/128
+[decode] max_len=1024 generated=81/128
+[decode] max_len=1024 generated=81/128
+[decode] max_len=1024 generated=82/128
+[decode] max_len=1024 generated=82/128
+[decode] max_len=1024 generated=83/128
+[decode] max_len=1024 generated=83/128
+[decode] max_len=1024 generated=84/128
+[decode] max_len=1024 generated=84/128
+[decode] max_len=1024 generated=85/128
+[decode] max_len=1024 generated=85/128
+[decode] max_len=1024 generated=86/128
+[decode] max_len=1024 generated=86/128
+[decode] max_len=1024 generated=87/128
+[decode] max_len=1024 generated=87/128
+[decode] max_len=1024 generated=88/128
+[decode] max_len=1024 generated=88/128
+[decode] max_len=1024 generated=89/128
+[decode] max_len=1024 generated=89/128
+[decode] max_len=1024 generated=90/128
+[decode] max_len=1024 generated=90/128
+[decode] max_len=1024 generated=91/128
+[decode] max_len=1024 generated=91/128
+[decode] max_len=1024 generated=92/128
+[decode] max_len=1024 generated=92/128
+[decode] max_len=1024 generated=93/128
+[decode] max_len=1024 generated=93/128
+[decode] max_len=1024 generated=94/128
+[decode] max_len=1024 generated=94/128
+[decode] max_len=1024 generated=95/128
+[decode] max_len=1024 generated=95/128
+[decode] max_len=1024 generated=96/128
+[decode] max_len=1024 generated=96/128
+[decode] max_len=1024 generated=97/128
+[decode] max_len=1024 generated=97/128
+[decode] max_len=1024 generated=98/128
+[decode] max_len=1024 generated=98/128
+[decode] max_len=1024 generated=99/128
+[decode] max_len=1024 generated=99/128
+[decode] max_len=1024 generated=100/128
+[decode] max_len=1024 generated=100/128
+[decode] max_len=1024 generated=101/128
+[decode] max_len=1024 generated=101/128
+[decode] max_len=1024 generated=102/128
+[decode] max_len=1024 generated=102/128
+[decode] max_len=1024 generated=103/128
+[decode] max_len=1024 generated=103/128
+[decode] max_len=1024 generated=104/128
+[decode] max_len=1024 generated=104/128
+[decode] max_len=1024 generated=105/128
+[decode] max_len=1024 generated=105/128
+[decode] max_len=1024 generated=106/128
+[decode] max_len=1024 generated=106/128
+[decode] max_len=1024 generated=107/128
+[decode] max_len=1024 generated=107/128
+[decode] max_len=1024 generated=108/128
+[decode] max_len=1024 generated=108/128
+[decode] max_len=1024 generated=109/128
+[decode] max_len=1024 generated=109/128
+[decode] max_len=1024 generated=110/128
+[decode] max_len=1024 generated=110/128
+[decode] max_len=1024 generated=111/128
+[decode] max_len=1024 generated=111/128
+[decode] max_len=1024 generated=112/128
+[decode] max_len=1024 generated=112/128
+[decode] max_len=1024 generated=113/128
+[decode] max_len=1024 generated=113/128
+[decode] max_len=1024 generated=114/128
+[decode] max_len=1024 generated=114/128
+[decode] max_len=1024 generated=115/128
+[decode] max_len=1024 generated=115/128
+[decode] max_len=1024 generated=116/128
+[decode] max_len=1024 generated=116/128
+[decode] max_len=1024 generated=117/128
+[decode] max_len=1024 generated=117/128
+[decode] max_len=1024 generated=118/128
+[decode] max_len=1024 generated=118/128
+[decode] max_len=1024 generated=119/128
+[decode] max_len=1024 generated=119/128
+[decode] max_len=1024 generated=120/128
+[decode] max_len=1024 generated=120/128
+[decode] max_len=1024 generated=121/128
+[decode] max_len=1024 generated=121/128
+[decode] max_len=1024 generated=122/128
+[decode] max_len=1024 generated=122/128
+[decode] max_len=1024 generated=123/128
+[decode] max_len=1024 generated=123/128
+[decode] max_len=1024 generated=124/128
+[decode] max_len=1024 generated=124/128
+[decode] max_len=1024 generated=125/128
+[decode] max_len=1024 generated=125/128
+[decode] max_len=1024 generated=126/128
+[decode] max_len=1024 generated=126/128
+[decode] max_len=1024 generated=127/128
+[decode] max_len=1024 generated=127/128
+[decode] max_len=1024 generated=128/128
+[decode] max_len=1024 generated=128/128
+[
+  {
+    "checkpoint": "runs/lta_owt_bert_absrope_time4_dirichlet_len1024_C1_to_1024_mask1_sameT_gbs512_b32_8gpu_1m_save10k_20260526/step_0130000.pt",
+    "ckpt_step": 130000,
+    "max_len": 1024,
+    "decode_rule": "dual_line_resample",
+    "support_power": 1.0,
+    "semantic_power": 1.5,
+    "steps": 128,
+    "c_min": 1.0,
+    "c_max": 1024.0,
+    "anchor_mode": "state",
+    "model_t_mode": "flow",
+    "time_schedule": "uniform",
+    "time_logit_mean": -1.5,
+    "time_logit_std": 0.8,
+    "time_power": 2.0,
+    "input_noise_scale": 0.0,
+    "input_noise_until": 1.0,
+    "input_noise_dirichlet_concentration": 1.0,
+    "endpoint_softening": "none",
+    "endpoint_soft_power": 2.0,
+    "endpoint_soft_min_conf": 0.0,
+    "endpoint_soft_max_conf": 1.0,
+    "soft_target_decode_mode": "off",
+    "soft_target_power": 1.0,
+    "soft_target_min_conf": 0.0,
+    "soft_target_max_conf": 1.0,
+    "soft_target_debias_start": 0.7,
+    "final_from": "blend",
+    "final_decode": "argmax",
+    "final_sample_temp": 1.0,
+    "final_top_k": 0,
+    "final_top_p": 1.0,
+    "commit_mode": "off",
+    "commit_conf_threshold": 0.0,
+    "commit_margin_threshold": 0.0,
+    "commit_start": 0.0,
+    "commit_min_ratio": 0.0,
+    "commit_max_ratio": 1.0,
+    "commit_power": 2.0,
+    "commit_freq_max_frac": 0.08,
+    "early_temp": 2.8,
+    "late_temp": 1.45,
+    "temp_end": 0.55,
+    "temp_power": 1.5,
+    "pos_extend": "repeat",
+    "fixed_first_token_id": null,
+    "fixed_first_token_text": "",
+    "fixed_first_initial_argmax": false,
+    "use_ema": false,
+    "n_samples": 128,
+    "sample_entropy": 0.4655874822927602,
+    "unique_tokens": 353,
+    "token_count": 131072,
+    "distinct_1": 0.00269317626953125,
+    "distinct_2": 0.013876160801564027,
+    "top_token_mass": 0.3938446044921875,
+    "texts_preview": [
+      "[CLS] / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / stations / stations stations stations stations stations stations stations stations stations stations stations stations stations stations stations stations stations stations stations stations stations stations stations stations stations stations stations stations stations stations stations stations stations m m m m m m m m m m m m m m m m m m m m m m m m m m m m m m m m m m m m m m m m m m m m m m m m m m m m m m m m m com com com com com com com com com com com com com com com com com com com com com com me m m m m m m m m m m m m m m m m m m m m m m m m m m m m m m m m m m m m m m m m m m m m m m m m m m m m m m m m m m m m m m m m / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / stations stations stations stations stations stations stations stations stations stations stations stations stations stations m m m m m m m m m m m m m m m m m m m m m m m m m m m m m m m m m m m m m m m [SEP]",
+      "[SEP] 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 [SEP]",
+      "i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i",
+      "[CLS] 1, 1, 1 / 1 1 / 1, 1 / 1 /, 1, 1 / 1, 1 1, 1 / 1, / 1, 1, / / / 1, 1 / 1, / 1, 1, 1, 1, 1, 1, 1, 1 / 1 1, 1 / 1, 1 / 1 /, / 1, 1 / 1, 1 / 1, 1 / / 1, 1 / 1 / 1, 1, 1, 1 / 1, 1 1 / 1, 1 / 1 / 1, 1, 1 / / 1, 1 / / /, 1 / 1, 1 /, 1 / 1 / 1, / 1 / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / munoz munoz munoz munoz munoz munoz munoz munoz munoz munoz munoz munoz munoz munoz munoz munoz munoz munoz munoz munoz munoz munoz munoz munoz munoz munoz munoz munoz munoz munoz munoz munoz munoz munoz / / munoz / munoz / / munoz munoz / / / munoz munoz munoz / munoz / munoz munoz / munoz munoz munoz munoz munoz munoz munoz munoz munoz munoz / munoz munoz / munoz munoz munoz munoz munoz / munoz / / / munoz munoz munoz munoz munoz munoz munoz munoz munoz munoz munoz / munoz munoz / munoz munoz munoz / / munoz / munoz munoz munoz / / munoz munoz / munoz / / munoz / munoz / munoz munoz munoz munoz munoz / munoz munoz / munoz munoz / munoz munoz munoz munoz / munoz munoz munoz munoz munoz munoz munoz munoz munoz munoz munoz munoz / / munoz munoz munoz munoz / munoz munoz munoz munoz munoz munoz munoz / munoz munoz munoz munoz munoz / / munoz munoz munoz / munoz munoz / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / [SEP]"
+    ],
+    "gen_ppl": 1.3137663186017279,
+    "gen_nll": 0.27289806452234733,
+    "gen_tokens": 129120
+  }
+]
+[
+  {
+    "checkpoint": "runs/lta_owt_bert_absrope_time4_dirichlet_len1024_C1_to_1024_mask1_sameT_gbs512_b32_8gpu_1m_save10k_20260526/step_0130000.pt",
+    "ckpt_step": 130000,
+    "max_len": 1024,
+    "decode_rule": "dual_line_resample",
+    "support_power": 1.0,
+    "semantic_power": 1.5,
+    "steps": 128,
+    "c_min": 1.0,
+    "c_max": 1024.0,
+    "anchor_mode": "state",
+    "model_t_mode": "flow",
+    "time_schedule": "uniform",
+    "time_logit_mean": -1.5,
+    "time_logit_std": 0.8,
+    "time_power": 2.0,
+    "input_noise_scale": 0.0,
+    "input_noise_until": 1.0,
+    "input_noise_dirichlet_concentration": 1.0,
+    "endpoint_softening": "none",
+    "endpoint_soft_power": 2.0,
+    "endpoint_soft_min_conf": 0.0,
+    "endpoint_soft_max_conf": 1.0,
+    "soft_target_decode_mode": "off",
+    "soft_target_power": 1.0,
+    "soft_target_min_conf": 0.0,
+    "soft_target_max_conf": 1.0,
+    "soft_target_debias_start": 0.7,
+    "final_from": "blend",
+    "final_decode": "argmax",
+    "final_sample_temp": 1.0,
+    "final_top_k": 0,
+    "final_top_p": 1.0,
+    "commit_mode": "off",
+    "commit_conf_threshold": 0.0,
+    "commit_margin_threshold": 0.0,
+    "commit_start": 0.0,
+    "commit_min_ratio": 0.0,
+    "commit_max_ratio": 1.0,
+    "commit_power": 2.0,
+    "commit_freq_max_frac": 0.08,
+    "early_temp": 2.8,
+    "late_temp": 1.45,
+    "temp_end": 0.55,
+    "temp_power": 1.5,
+    "pos_extend": "repeat",
+    "fixed_first_token_id": null,
+    "fixed_first_token_text": "",
+    "fixed_first_initial_argmax": false,
+    "use_ema": false,
+    "n_samples": 128,
+    "sample_entropy": 0.4655874822927602,
+    "unique_tokens": 353,
+    "token_count": 131072,
+    "distinct_1": 0.00269317626953125,
+    "distinct_2": 0.013876160801564027,
+    "top_token_mass": 0.3938446044921875,
+    "texts_preview": [
+      "[CLS] / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / stations / stations stations stations stations stations stations stations stations stations stations stations stations stations stations stations stations stations stations stations stations stations stations stations stations stations stations stations stations stations stations stations stations stations m m m m m m m m m m m m m m m m m m m m m m m m m m m m m m m m m m m m m m m m m m m m m m m m m m m m m m m m m com com com com com com com com com com com com com com com com com com com com com com me m m m m m m m m m m m m m m m m m m m m m m m m m m m m m m m m m m m m m m m m m m m m m m m m m m m m m m m m m m m m m m m m / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / stations stations stations stations stations stations stations stations stations stations stations stations stations stations m m m m m m m m m m m m m m m m m m m m m m m m m m m m m m m m m m m m m m m [SEP]",
+      "[SEP] 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 [SEP]",
+      "i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i",
+      "[CLS] 1, 1, 1 / 1 1 / 1, 1 / 1 /, 1, 1 / 1, 1 1, 1 / 1, / 1, 1, / / / 1, 1 / 1, / 1, 1, 1, 1, 1, 1, 1, 1 / 1 1, 1 / 1, 1 / 1 /, / 1, 1 / 1, 1 / 1, 1 / / 1, 1 / 1 / 1, 1, 1, 1 / 1, 1 1 / 1, 1 / 1 / 1, 1, 1 / / 1, 1 / / /, 1 / 1, 1 /, 1 / 1 / 1, / 1 / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / munoz munoz munoz munoz munoz munoz munoz munoz munoz munoz munoz munoz munoz munoz munoz munoz munoz munoz munoz munoz munoz munoz munoz munoz munoz munoz munoz munoz munoz munoz munoz munoz munoz munoz / / munoz / munoz / / munoz munoz / / / munoz munoz munoz / munoz / munoz munoz / munoz munoz munoz munoz munoz munoz munoz munoz munoz munoz / munoz munoz / munoz munoz munoz munoz munoz / munoz / / / munoz munoz munoz munoz munoz munoz munoz munoz munoz munoz munoz / munoz munoz / munoz munoz munoz / / munoz / munoz munoz munoz / / munoz munoz / munoz / / munoz / munoz / munoz munoz munoz munoz munoz / munoz munoz / munoz munoz / munoz munoz munoz munoz / munoz munoz munoz munoz munoz munoz munoz munoz munoz munoz munoz munoz / / munoz munoz munoz munoz / munoz munoz munoz munoz munoz munoz munoz / munoz munoz munoz munoz munoz / / munoz munoz munoz / munoz munoz / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / [SEP]"
+    ],
+    "gen_ppl": 1.3137663186017279,
+    "gen_nll": 0.27289806452234733,
+    "gen_tokens": 129120
+  }
+]
+[watch-gumbel] 2026-05-26_16:59:43 done step_0130000
+[watch-gumbel] 2026-05-26_16:59:44 done step_0130000

LTA_openwebtext_dualt/logs/smoke_lta_openwebtext_dirichlet_dualt_tsched_len1024_1gpu.log

@@ -0,0 +1,31 @@
+{
+  "device": "cuda:0",
+  "rank": 0,
+  "world_size": 1,
+  "samples": "wrapped_streaming",
+  "vocab_size": 50257,
+  "save_dir": "runs/smoke_lta_openwebtext_dirichlet_dualt_tsched_len1024_1gpu",
+  "batch_size": 1,
+  "grad_accum": 1,
+  "effective_batch_size": 1,
+  "global_batch_size": 1,
+  "lr_schedule": "constant_warmup",
+  "warmup_steps": 3,
+  "model_type": "ddit",
+  "dual_t": true,
+  "corrupt_t_mode": "independent",
+  "corrupt_min_t": 0.0,
+  "corrupt_max_t": 1.0,
+  "torch_compile": false,
+  "compile_mode": "max-autotune",
+  "state_format": "prob",
+  "target_loss": "soft_ce",
+  "meanflow_weight": 0.0,
+  "bridge_noise_init": "logistic_normal",
+  "noise_sigma": -1.0,
+  "wrap": true,
+  "num_workers": 0
+}
+step=1 micro_steps=1 elapsed=1.4s lr=2.000000e-04 loss_all=10.8249 acc_all=0.0000 loss_corrupt=10.8249 acc_corrupt=0.0000 corrupt_frac=0.2363 loss=10.8249 loss_recon=10.8249 loss_meanflow=0.0000 mean_model_t=0.0669 mean_corrupt_t=0.6396 wrong_frac=0.4050 init_acc_corrupt=0.5950 init_gold_top10=0.5950 init_gold_top100=0.5950
+step=2 micro_steps=2 elapsed=0.1s lr=3.000000e-04 loss_all=10.8041 acc_all=0.3027 loss_corrupt=10.8101 acc_corrupt=0.2199 corrupt_frac=0.3330 loss=10.8101 loss_recon=10.8101 loss_meanflow=0.0000 mean_model_t=0.7491 mean_corrupt_t=0.7061 wrong_frac=0.3050 init_acc_corrupt=0.6950 init_gold_top10=0.6950 init_gold_top100=0.6950
+step=3 micro_steps=3 elapsed=0.0s lr=3.000000e-04 loss_all=10.7836 acc_all=0.2432 loss_corrupt=10.7924 acc_corrupt=0.1938 corrupt_frac=0.8516 loss=10.7924 loss_recon=10.7924 loss_meanflow=0.0000 mean_model_t=0.2118 mean_corrupt_t=0.3812 wrong_frac=0.6250 init_acc_corrupt=0.3635 init_gold_top10=0.3750 init_gold_top100=0.3750

LTA_openwebtext_dualt/logs/train8ctx8_allcorrupt/driver.log

@@ -0,0 +1,670 @@
+[allcorrupt] start train8_n8_allcorrupt_hard_ce_20260517_train8ctx8_allcorrupt Sun May 17 00:22:10 UTC 2026
+[launch] gpt2 cached OWT soft-endpoint m/n pilot
+[launch] run_name=train8_n8_allcorrupt_hard_ce_20260517_train8ctx8_allcorrupt
+[launch] save_dir=runs/train8_n8_allcorrupt_hard_ce_20260517_train8ctx8_allcorrupt
+[launch] n=8 m=0 clean_state_mode=onehot
+[launch] mask_mixture lowk=0 all=1
+[launch] target_loss=hard_ce conf=0.0->1.0 power=1.0
+[launch] cache=/e2e-data/evad-tech-vla/wanghan58/data/small_benchmarks/langflow_2604_11748/openwebtext_lta_cached_chunks/gpt2_len8_train8_overfit
+NCCL version 2.25.1+cuda12.8
+{
+  "device": "cuda:0",
+  "rank": 0,
+  "world_size": 4,
+  "samples": "owt_cached_chunks:8",
+  "vocab_size": 50257,
+  "tokenizer_vocab_size": 50257,
+  "save_dir": "runs/train8_n8_allcorrupt_hard_ce_20260517_train8ctx8_allcorrupt",
+  "batch_size": 1,
+  "grad_accum": 1,
+  "effective_batch_size": 4,
+  "global_batch_size": 4,
+  "lr_schedule": "constant_warmup",
+  "optimizer": "muon",
+  "epochs": 0.0,
+  "steps_per_epoch": 2,
+  "total_steps": 500,
+  "warmup_steps": 10,
+  "warmup_epochs": -1.0,
+  "min_lr": 0.0,
+  "weight_decay": 0.1,
+  "output_weight_decay": -1.0,
+  "adamw_param_groups": "nanogpt",
+  "adam_beta1": 0.9,
+  "adam_beta2": 0.95,
+  "adam_eps": 1e-08,
+  "muon_impl": "legacy",
+  "muon_momentum": 0.95,
+  "muon_ns_steps": 5,
+  "muon_update_scale": 1.0,
+  "muon_nesterov": false,
+  "muon_width_scale": false,
+  "muon_grouping": "legacy_dim_ge_2",
+  "muon_param_count": 169453056,
+  "muon_adam_param_count": 122368,
+  "muon_param_names": [
+    "vocab_embed.embedding",
+    "sigma_map.net.0.weight",
+    "sigma_map.net.2.weight",
+    "blocks.0.attn_qkv.weight",
+    "blocks.0.attn_out.weight",
+    "blocks.0.mlp.0.weight",
+    "blocks.0.mlp.2.weight",
+    "blocks.0.adaLN_modulation.weight",
+    "blocks.1.attn_qkv.weight",
+    "blocks.1.attn_out.weight",
+    "blocks.1.mlp.0.weight",
+    "blocks.1.mlp.2.weight",
+    "blocks.1.adaLN_modulation.weight",
+    "blocks.2.attn_qkv.weight",
+    "blocks.2.attn_out.weight",
+    "blocks.2.mlp.0.weight",
+    "blocks.2.mlp.2.weight",
+    "blocks.2.adaLN_modulation.weight",
+    "blocks.3.attn_qkv.weight",
+    "blocks.3.attn_out.weight",
+    "blocks.3.mlp.0.weight",
+    "blocks.3.mlp.2.weight",
+    "blocks.3.adaLN_modulation.weight",
+    "blocks.4.attn_qkv.weight",
+    "blocks.4.attn_out.weight",
+    "blocks.4.mlp.0.weight",
+    "blocks.4.mlp.2.weight",
+    "blocks.4.adaLN_modulation.weight",
+    "blocks.5.attn_qkv.weight",
+    "blocks.5.attn_out.weight",
+    "blocks.5.mlp.0.weight",
+    "blocks.5.mlp.2.weight",
+    "blocks.5.adaLN_modulation.weight",
+    "blocks.6.attn_qkv.weight",
+    "blocks.6.attn_out.weight",
+    "blocks.6.mlp.0.weight",
+    "blocks.6.mlp.2.weight",
+    "blocks.6.adaLN_modulation.weight",
+    "blocks.7.attn_qkv.weight",
+    "blocks.7.attn_out.weight",
+    "blocks.7.mlp.0.weight",
+    "blocks.7.mlp.2.weight",
+    "blocks.7.adaLN_modulation.weight",
+    "blocks.8.attn_qkv.weight",
+    "blocks.8.attn_out.weight",
+    "blocks.8.mlp.0.weight",
+    "blocks.8.mlp.2.weight",
+    "blocks.8.adaLN_modulation.weight",
+    "blocks.9.attn_qkv.weight",
+    "blocks.9.attn_out.weight",
+    "blocks.9.mlp.0.weight",
+    "blocks.9.mlp.2.weight",
+    "blocks.9.adaLN_modulation.weight",
+    "blocks.10.attn_qkv.weight",
+    "blocks.10.attn_out.weight",
+    "blocks.10.mlp.0.weight",
+    "blocks.10.mlp.2.weight",
+    "blocks.10.adaLN_modulation.weight",
+    "blocks.11.attn_qkv.weight",
+    "blocks.11.attn_out.weight",
+    "blocks.11.mlp.0.weight",
+    "blocks.11.mlp.2.weight",
+    "blocks.11.adaLN_modulation.weight",
+    "output_layer.linear.weight",
+    "output_layer.adaLN_modulation.weight"
+  ],
+  "muon_adam_param_names": [
+    "sigma_map.net.0.bias",
+    "sigma_map.net.2.bias",
+    "blocks.0.norm1.weight",
+    "blocks.0.norm2.weight",
+    "blocks.0.mlp.0.bias",
+    "blocks.0.mlp.2.bias",
+    "blocks.0.adaLN_modulation.bias",
+    "blocks.1.norm1.weight",
+    "blocks.1.norm2.weight",
+    "blocks.1.mlp.0.bias",
+    "blocks.1.mlp.2.bias",
+    "blocks.1.adaLN_modulation.bias",
+    "blocks.2.norm1.weight",
+    "blocks.2.norm2.weight",
+    "blocks.2.mlp.0.bias",
+    "blocks.2.mlp.2.bias",
+    "blocks.2.adaLN_modulation.bias",
+    "blocks.3.norm1.weight",
+    "blocks.3.norm2.weight",
+    "blocks.3.mlp.0.bias",
+    "blocks.3.mlp.2.bias",
+    "blocks.3.adaLN_modulation.bias",
+    "blocks.4.norm1.weight",
+    "blocks.4.norm2.weight",
+    "blocks.4.mlp.0.bias",
+    "blocks.4.mlp.2.bias",
+    "blocks.4.adaLN_modulation.bias",
+    "blocks.5.norm1.weight",
+    "blocks.5.norm2.weight",
+    "blocks.5.mlp.0.bias",
+    "blocks.5.mlp.2.bias",
+    "blocks.5.adaLN_modulation.bias",
+    "blocks.6.norm1.weight",
+    "blocks.6.norm2.weight",
+    "blocks.6.mlp.0.bias",
+    "blocks.6.mlp.2.bias",
+    "blocks.6.adaLN_modulation.bias",
+    "blocks.7.norm1.weight",
+    "blocks.7.norm2.weight",
+    "blocks.7.mlp.0.bias",
+    "blocks.7.mlp.2.bias",
+    "blocks.7.adaLN_modulation.bias",
+    "blocks.8.norm1.weight",
+    "blocks.8.norm2.weight",
+    "blocks.8.mlp.0.bias",
+    "blocks.8.mlp.2.bias",
+    "blocks.8.adaLN_modulation.bias",
+    "blocks.9.norm1.weight",
+    "blocks.9.norm2.weight",
+    "blocks.9.mlp.0.bias",
+    "blocks.9.mlp.2.bias",
+    "blocks.9.adaLN_modulation.bias",
+    "blocks.10.norm1.weight",
+    "blocks.10.norm2.weight",
+    "blocks.10.mlp.0.bias",
+    "blocks.10.mlp.2.bias",
+    "blocks.10.adaLN_modulation.bias",
+    "blocks.11.norm1.weight",
+    "blocks.11.norm2.weight",
+    "blocks.11.mlp.0.bias",
+    "blocks.11.mlp.2.bias",
+    "blocks.11.adaLN_modulation.bias",
+    "output_layer.norm_final.weight",
+    "output_layer.adaLN_modulation.bias"
+  ],
+  "muon_effective_nesterov": false,
+  "muon_effective_width_scale": false,
+  "muon_effective_weight_decay": 0.1,
+  "muon_adam_fallback_nesterov": false,
+  "muon_adam_fallback_weight_decay": 0.1,
+  "ema_decay": 0.9999,
+  "ema_start_step": 0,
+  "model_type": "ddit",
+  "elf_num_time_tokens": 4,
+  "elf_num_model_mode_tokens": 0,
+  "qk_norm": true,
+  "output_bias": false,
+  "output_init_std": -1.0,
+  "norm_type": "rmsnorm",
+  "target_loss": "hard_ce",
+  "linear_soft_target_power": 1.0,
+  "linear_soft_target_min_conf": 0.0,
+  "linear_soft_target_max_conf": 1.0,
+  "t_sampling_mode": "logit_normal",
+  "t_sampling_power": 1.0,
+  "t_sampling_eps": 0.0001,
+  "t_sampling_logit_mean": -1.5,
+  "t_sampling_logit_std": 0.8,
+  "dual_t": true,
+  "corrupt_t_mode": "same",
+  "corrupt_min_t": 0.0,
+  "corrupt_max_t": 1.0,
+  "prefix_block_prob": 0.0,
+  "prefix_block_len": 128,
+  "mask_ratio_floor_schedule": "none",
+  "dirichlet_endpoint_mode": "categorical_dual_t",
+  "dirichlet_semantic_t_mode": "same",
+  "dirichlet_semantic_t_value": 0.0,
+  "dirichlet_semantic_t_curve": "linear",
+  "dirichlet_semantic_t_power": 1.0,
+  "endpoint_sequence_random_prob_alpha": 0.0,
+  "categorical_wrong_from_full_vocab": true,
+  "categorical_wrong_from_batch_valid_tokens": false,
+  "categorical_wrong_basin_token_ids": "",
+  "categorical_wrong_basin_prob": 0.0,
+  "categorical_wrong_unigram_prob": 0.0,
+  "categorical_wrong_uniform_prob": 0.0,
+  "categorical_wrong_corpus_unigram_path": "",
+  "categorical_wrong_corpus_unigram_alpha": 1.0,
+  "categorical_wrong_basin_shared_prob": 0.0,
+  "categorical_wrong_unigram_shared_prob": 0.0,
+  "mask_mixture_original_prob": 0.0,
+  "mask_mixture_lowk_prob": 0.0,
+  "mask_mixture_lowcorrupt_prob": 0.0,
+  "mask_mixture_block_prob": 0.0,
+  "mask_mixture_all_prob": 1.0,
+  "mask_mixture_lowk_clean_tokens": "0",
+  "mask_mixture_lowcorrupt_tokens": "1,2,4,8,16,32,64",
+  "mask_mixture_block_tokens": "64,128",
+  "simplex_bridge_sampler": "dirichlet",
+  "logistic_normal_sigma_min": 0.18,
+  "logistic_normal_sigma_max": 2.2,
+  "logistic_normal_tau_min": 0.65,
+  "logistic_normal_tau_max": 1.15,
+  "torch_compile": false,
+  "compile_mode": "max-autotune",
+  "state_format": "prob",
+  "meanflow_weight": 0.0,
+  "rollout_train_prob": 0.0,
+  "rollout_train_steps": 1,
+  "rollout_train_infer_steps": 64,
+  "rollout_train_temp": 1.45,
+  "rollout_train_max_gamma": 1.0,
+  "rollout_train_corrupt_only": true,
+  "rollout_train_samplewise": false,
+  "rollout_train_compute_always": false,
+  "bridge_noise_init": "logistic_normal",
+  "noise_sigma": -1.0,
+  "allow_tf32": true,
+  "activation_checkpointing": false,
+  "activation_checkpoint_interval": 1,
+  "activation_checkpoint_scope": "block",
+  "ddp_static_graph": false,
+  "ddp_gradient_as_bucket_view": true,
+  "blocking_data_transfer": false,
+  "dataloader_prefetch_factor": 4,
+  "full_train_stats": false,
+  "tokenized_hf": false,
+  "tokenized_pad_token": "pad",
+  "elf_conditional_hf": false,
+  "record_pad_truncate": false,
+  "record_add_eos": false,
+  "record_add_special_tokens": false,
+  "record_pad_token": "pad",
+  "record_shuffle_buffer": 10000,
+  "wrap": true,
+  "wrap_mode": "stream",
+  "wrap_record_buffer_size": 200,
+  "owt_cached_chunks": true,
+  "owt_chunk_cache_dir": "/e2e-data/evad-tech-vla/wanghan58/data/small_benchmarks/langflow_2604_11748/openwebtext_lta_cached_chunks/gpt2_len8_train8_overfit",
+  "owt_chunk_cache_rebuild": false,
+  "owt_chunk_cache_write_batch": 4096,
+  "owt_exact_repeat_per_chunk": 0,
+  "online_chunk_shuffle": false,
+  "online_chunk_shuffle_buffer": 10000,
+  "openwebtext_split": "train_minus_100k",
+  "detokenizer": "auto",
+  "resolved_detokenizer": null,
+  "num_workers": 0,
+  "latest_every": 10,
+  "resume_path": ""
+}
+step=10 epoch=5/250 epoch_step=2/2 micro_steps=10 elapsed=1.8s lr=2.000000e-03 loss=10.8055 loss_recon=10.8055 loss_meanflow=0.0000 mean_model_t=0.2827 mean_corrupt_t=0.2827 mean_loss_t_weight=1.0000 linear_soft_target_mean_conf=0.0000 prior_center_loss_beta=0.0000 rollout_train_applied=0.0000 grad_enabled_before_rollout=1.0000 grad_enabled_after_rollout=1.0000 logits_requires_grad=1.0000 raw_loss_requires_grad=1.0000 acc_all=0.1625 corrupt_frac=1.0000 acc_corrupt=0.1625 loss_corrupt=10.8055 wrong_frac=0.7500 init_acc_corrupt=0.1625 acc_corrupt_t_0p0_0p2=0.0938 corrupt_frac_t_0p0_0p2=1.0000 out_w_norm=0.0155 out_g_norm=4.9894 acc_corrupt_t_0p6_0p8=0.0000 corrupt_frac_t_0p6_0p8=1.0000 acc_corrupt_t_0p2_0p4=0.1875 corrupt_frac_t_0p2_0p4=1.0000 acc_corrupt_t_0p4_0p6=0.5000 corrupt_frac_t_0p4_0p6=1.0000 loss_all=10.7734 init_gold_top10=0.3750 init_gold_top100=0.3750
+step=20 epoch=10/250 epoch_step=2/2 micro_steps=20 elapsed=5.5s lr=2.000000e-03 loss=10.7273 loss_recon=10.7273 loss_meanflow=0.0000 mean_model_t=0.2379 mean_corrupt_t=0.2379 mean_loss_t_weight=1.0000 linear_soft_target_mean_conf=0.0000 prior_center_loss_beta=0.0000 rollout_train_applied=0.0000 grad_enabled_before_rollout=1.0000 grad_enabled_after_rollout=1.0000 logits_requires_grad=1.0000 raw_loss_requires_grad=1.0000 acc_all=0.1625 corrupt_frac=1.0000 acc_corrupt=0.1625 loss_corrupt=10.7273 wrong_frac=0.7250 init_acc_corrupt=0.1625 acc_corrupt_t_0p0_0p2=0.1250 corrupt_frac_t_0p0_0p2=1.0000 out_w_norm=0.0995 out_g_norm=6.0688 acc_corrupt_t_0p2_0p4=0.1875 corrupt_frac_t_0p2_0p4=1.0000 acc_corrupt_t_0p6_0p8=0.3750 corrupt_frac_t_0p6_0p8=1.0000 acc_corrupt_t_0p4_0p6=0.1250 corrupt_frac_t_0p4_0p6=1.0000 loss_all=10.7188 init_gold_top10=0.1250 init_gold_top100=0.1250
+step=30 epoch=15/250 epoch_step=2/2 micro_steps=30 elapsed=5.2s lr=2.000000e-03 loss=10.6273 loss_recon=10.6273 loss_meanflow=0.0000 mean_model_t=0.2564 mean_corrupt_t=0.2564 mean_loss_t_weight=1.0000 linear_soft_target_mean_conf=0.0000 prior_center_loss_beta=0.0000 rollout_train_applied=0.0000 grad_enabled_before_rollout=1.0000 grad_enabled_after_rollout=1.0000 logits_requires_grad=1.0000 raw_loss_requires_grad=1.0000 acc_all=0.1500 corrupt_frac=1.0000 acc_corrupt=0.1500 loss_corrupt=10.6273 wrong_frac=0.7750 init_acc_corrupt=0.1750 acc_corrupt_t_0p2_0p4=0.1500 corrupt_frac_t_0p2_0p4=1.0000 out_w_norm=0.2176 out_g_norm=6.9701 acc_corrupt_t_0p4_0p6=0.1250 corrupt_frac_t_0p4_0p6=1.0000 acc_corrupt_t_0p0_0p2=0.1667 corrupt_frac_t_0p0_0p2=1.0000 loss_all=10.6328 init_gold_top10=0.0000 init_gold_top100=0.1250
+step=40 epoch=20/250 epoch_step=2/2 micro_steps=40 elapsed=4.5s lr=2.000000e-03 loss=10.5273 loss_recon=10.5273 loss_meanflow=0.0000 mean_model_t=0.1906 mean_corrupt_t=0.1906 mean_loss_t_weight=1.0000 linear_soft_target_mean_conf=0.0000 prior_center_loss_beta=0.0000 rollout_train_applied=0.0000 grad_enabled_before_rollout=1.0000 grad_enabled_after_rollout=1.0000 logits_requires_grad=1.0000 raw_loss_requires_grad=1.0000 acc_all=0.1375 corrupt_frac=1.0000 acc_corrupt=0.1375 loss_corrupt=10.5273 wrong_frac=0.8625 init_acc_corrupt=0.0875 acc_corrupt_t_0p2_0p4=0.1250 corrupt_frac_t_0p2_0p4=1.0000 out_w_norm=0.3264 out_g_norm=7.2241 acc_corrupt_t_0p4_0p6=0.1250 corrupt_frac_t_0p4_0p6=1.0000 acc_corrupt_t_0p0_0p2=0.1458 corrupt_frac_t_0p0_0p2=1.0000 loss_all=10.4922 init_gold_top10=0.0000 init_gold_top100=0.1250
+step=50 epoch=25/250 epoch_step=2/2 micro_steps=50 elapsed=5.5s lr=2.000000e-03 loss=10.4398 loss_recon=10.4398 loss_meanflow=0.0000 mean_model_t=0.1646 mean_corrupt_t=0.1646 mean_loss_t_weight=1.0000 linear_soft_target_mean_conf=0.0000 prior_center_loss_beta=0.0000 rollout_train_applied=0.0000 grad_enabled_before_rollout=1.0000 grad_enabled_after_rollout=1.0000 logits_requires_grad=1.0000 raw_loss_requires_grad=1.0000 acc_all=0.1375 corrupt_frac=1.0000 acc_corrupt=0.1375 loss_corrupt=10.4398 wrong_frac=0.8875 init_acc_corrupt=0.0250 acc_corrupt_t_0p0_0p2=0.1458 corrupt_frac_t_0p0_0p2=1.0000 out_w_norm=0.4272 out_g_norm=7.5896 acc_corrupt_t_0p2_0p4=0.1250 corrupt_frac_t_0p2_0p4=1.0000 loss_all=10.5156 init_gold_top10=0.0000 init_gold_top100=0.2500
+step=60 epoch=30/250 epoch_step=2/2 micro_steps=60 elapsed=4.4s lr=2.000000e-03 loss=10.2430 loss_recon=10.2430 loss_meanflow=0.0000 mean_model_t=0.2362 mean_corrupt_t=0.2362 mean_loss_t_weight=1.0000 linear_soft_target_mean_conf=0.0000 prior_center_loss_beta=0.0000 rollout_train_applied=0.0000 grad_enabled_before_rollout=1.0000 grad_enabled_after_rollout=1.0000 logits_requires_grad=1.0000 raw_loss_requires_grad=1.0000 acc_all=0.1500 corrupt_frac=1.0000 acc_corrupt=0.1500 loss_corrupt=10.2430 wrong_frac=0.7250 init_acc_corrupt=0.0875 acc_corrupt_t_0p0_0p2=0.1250 corrupt_frac_t_0p0_0p2=1.0000 out_w_norm=0.5347 out_g_norm=7.9975 acc_corrupt_t_0p4_0p6=0.1250 corrupt_frac_t_0p4_0p6=1.0000 acc_corrupt_t_0p2_0p4=0.1750 corrupt_frac_t_0p2_0p4=1.0000 loss_all=10.0469 init_gold_top10=0.1250 init_gold_top100=0.2500
+step=70 epoch=35/250 epoch_step=2/2 micro_steps=70 elapsed=5.2s lr=2.000000e-03 loss=10.1438 loss_recon=10.1438 loss_meanflow=0.0000 mean_model_t=0.1956 mean_corrupt_t=0.1956 mean_loss_t_weight=1.0000 linear_soft_target_mean_conf=0.0000 prior_center_loss_beta=0.0000 rollout_train_applied=0.0000 grad_enabled_before_rollout=1.0000 grad_enabled_after_rollout=1.0000 logits_requires_grad=1.0000 raw_loss_requires_grad=1.0000 acc_all=0.1500 corrupt_frac=1.0000 acc_corrupt=0.1500 loss_corrupt=10.1438 wrong_frac=0.8375 init_acc_corrupt=0.0375 acc_corrupt_t_0p2_0p4=0.1667 corrupt_frac_t_0p2_0p4=1.0000 out_w_norm=0.6329 out_g_norm=8.0782 acc_corrupt_t_0p0_0p2=0.1250 corrupt_frac_t_0p0_0p2=1.0000 loss_all=10.3750 init_gold_top10=0.1250 init_gold_top100=0.1250
+step=80 epoch=40/250 epoch_step=2/2 micro_steps=80 elapsed=4.6s lr=2.000000e-03 loss=9.7523 loss_recon=9.7523 loss_meanflow=0.0000 mean_model_t=0.2455 mean_corrupt_t=0.2455 mean_loss_t_weight=1.0000 linear_soft_target_mean_conf=0.0000 prior_center_loss_beta=0.0000 rollout_train_applied=0.0000 grad_enabled_before_rollout=1.0000 grad_enabled_after_rollout=1.0000 logits_requires_grad=1.0000 raw_loss_requires_grad=1.0000 acc_all=0.2750 corrupt_frac=1.0000 acc_corrupt=0.2750 loss_corrupt=9.7523 wrong_frac=0.7250 init_acc_corrupt=0.2000 acc_corrupt_t_0p4_0p6=0.5417 corrupt_frac_t_0p4_0p6=1.0000 out_w_norm=0.7131 out_g_norm=8.2600 acc_corrupt_t_0p0_0p2=0.1250 corrupt_frac_t_0p0_0p2=1.0000 acc_corrupt_t_0p2_0p4=0.2083 corrupt_frac_t_0p2_0p4=1.0000 loss_all=9.1094 init_gold_top10=0.5000 init_gold_top100=0.5000
+step=90 epoch=45/250 epoch_step=2/2 micro_steps=90 elapsed=4.6s lr=2.000000e-03 loss=9.8027 loss_recon=9.8027 loss_meanflow=0.0000 mean_model_t=0.2205 mean_corrupt_t=0.2205 mean_loss_t_weight=1.0000 linear_soft_target_mean_conf=0.0000 prior_center_loss_beta=0.0000 rollout_train_applied=0.0000 grad_enabled_before_rollout=1.0000 grad_enabled_after_rollout=1.0000 logits_requires_grad=1.0000 raw_loss_requires_grad=1.0000 acc_all=0.1500 corrupt_frac=1.0000 acc_corrupt=0.1500 loss_corrupt=9.8027 wrong_frac=0.7750 init_acc_corrupt=0.1375 acc_corrupt_t_0p0_0p2=0.1250 corrupt_frac_t_0p0_0p2=1.0000 out_w_norm=0.7917 out_g_norm=8.6682 acc_corrupt_t_0p2_0p4=0.1667 corrupt_frac_t_0p2_0p4=1.0000 acc_corrupt_t_0p4_0p6=0.1875 corrupt_frac_t_0p4_0p6=1.0000 loss_all=9.4961 init_gold_top10=0.3750 init_gold_top100=0.3750
+step=100 epoch=50/250 epoch_step=2/2 micro_steps=100 elapsed=5.5s lr=2.000000e-03 loss=9.5336 loss_recon=9.5336 loss_meanflow=0.0000 mean_model_t=0.1800 mean_corrupt_t=0.1800 mean_loss_t_weight=1.0000 linear_soft_target_mean_conf=0.0000 prior_center_loss_beta=0.0000 rollout_train_applied=0.0000 grad_enabled_before_rollout=1.0000 grad_enabled_after_rollout=1.0000 logits_requires_grad=1.0000 raw_loss_requires_grad=1.0000 acc_all=0.1750 corrupt_frac=1.0000 acc_corrupt=0.1750 loss_corrupt=9.5336 wrong_frac=0.8375 init_acc_corrupt=0.0875 acc_corrupt_t_0p0_0p2=0.1458 corrupt_frac_t_0p0_0p2=1.0000 out_w_norm=0.8749 out_g_norm=8.9338 acc_corrupt_t_0p2_0p4=0.2188 corrupt_frac_t_0p2_0p4=1.0000 loss_all=10.0547 init_gold_top10=0.1250 init_gold_top100=0.2500
+step=110 epoch=55/250 epoch_step=2/2 micro_steps=110 elapsed=4.6s lr=2.000000e-03 loss=9.2227 loss_recon=9.2227 loss_meanflow=0.0000 mean_model_t=0.2276 mean_corrupt_t=0.2276 mean_loss_t_weight=1.0000 linear_soft_target_mean_conf=0.0000 prior_center_loss_beta=0.0000 rollout_train_applied=0.0000 grad_enabled_before_rollout=1.0000 grad_enabled_after_rollout=1.0000 logits_requires_grad=1.0000 raw_loss_requires_grad=1.0000 acc_all=0.2000 corrupt_frac=1.0000 acc_corrupt=0.2000 loss_corrupt=9.2227 wrong_frac=0.8375 init_acc_corrupt=0.1375 acc_corrupt_t_0p0_0p2=0.1667 corrupt_frac_t_0p0_0p2=1.0000 out_w_norm=0.9618 out_g_norm=9.2988 acc_corrupt_t_0p4_0p6=0.2500 corrupt_frac_t_0p4_0p6=1.0000 acc_corrupt_t_0p2_0p4=0.2500 corrupt_frac_t_0p2_0p4=1.0000 loss_all=9.2617 init_gold_top10=0.0000 init_gold_top100=0.1250
+step=120 epoch=60/250 epoch_step=2/2 micro_steps=120 elapsed=4.2s lr=2.000000e-03 loss=8.8680 loss_recon=8.8680 loss_meanflow=0.0000 mean_model_t=0.2553 mean_corrupt_t=0.2553 mean_loss_t_weight=1.0000 linear_soft_target_mean_conf=0.0000 prior_center_loss_beta=0.0000 rollout_train_applied=0.0000 grad_enabled_before_rollout=1.0000 grad_enabled_after_rollout=1.0000 logits_requires_grad=1.0000 raw_loss_requires_grad=1.0000 acc_all=0.2750 corrupt_frac=1.0000 acc_corrupt=0.2750 loss_corrupt=8.8680 wrong_frac=0.6750 init_acc_corrupt=0.2250 acc_corrupt_t_0p2_0p4=0.3214 corrupt_frac_t_0p2_0p4=1.0000 out_w_norm=1.0520 out_g_norm=9.2110 acc_corrupt_t_0p0_0p2=0.1667 corrupt_frac_t_0p0_0p2=1.0000 loss_all=9.4043 init_gold_top10=0.5000 init_gold_top100=0.5000
+step=130 epoch=65/250 epoch_step=2/2 micro_steps=130 elapsed=5.2s lr=2.000000e-03 loss=8.7281 loss_recon=8.7281 loss_meanflow=0.0000 mean_model_t=0.2286 mean_corrupt_t=0.2286 mean_loss_t_weight=1.0000 linear_soft_target_mean_conf=0.0000 prior_center_loss_beta=0.0000 rollout_train_applied=0.0000 grad_enabled_before_rollout=1.0000 grad_enabled_after_rollout=1.0000 logits_requires_grad=1.0000 raw_loss_requires_grad=1.0000 acc_all=0.2375 corrupt_frac=1.0000 acc_corrupt=0.2375 loss_corrupt=8.7281 wrong_frac=0.7875 init_acc_corrupt=0.1375 acc_corrupt_t_0p0_0p2=0.1667 corrupt_frac_t_0p0_0p2=1.0000 out_w_norm=1.1458 out_g_norm=9.8395 acc_corrupt_t_0p2_0p4=0.2679 corrupt_frac_t_0p2_0p4=1.0000 loss_all=9.1562 init_gold_top10=0.1250 init_gold_top100=0.1250
+step=140 epoch=70/250 epoch_step=2/2 micro_steps=140 elapsed=4.6s lr=2.000000e-03 loss=8.8338 loss_recon=8.8338 loss_meanflow=0.0000 mean_model_t=0.1682 mean_corrupt_t=0.1682 mean_loss_t_weight=1.0000 linear_soft_target_mean_conf=0.0000 prior_center_loss_beta=0.0000 rollout_train_applied=0.0000 grad_enabled_before_rollout=1.0000 grad_enabled_after_rollout=1.0000 logits_requires_grad=1.0000 raw_loss_requires_grad=1.0000 acc_all=0.1625 corrupt_frac=1.0000 acc_corrupt=0.1625 loss_corrupt=8.8338 wrong_frac=0.8375 init_acc_corrupt=0.0500 acc_corrupt_t_0p0_0p2=0.1250 corrupt_frac_t_0p0_0p2=1.0000 out_w_norm=1.2408 out_g_norm=9.9622 acc_corrupt_t_0p2_0p4=0.2500 corrupt_frac_t_0p2_0p4=1.0000 loss_all=9.0508 init_gold_top10=0.1250 init_gold_top100=0.1250
+step=150 epoch=75/250 epoch_step=2/2 micro_steps=150 elapsed=4.2s lr=2.000000e-03 loss=8.2799 loss_recon=8.2799 loss_meanflow=0.0000 mean_model_t=0.2326 mean_corrupt_t=0.2326 mean_loss_t_weight=1.0000 linear_soft_target_mean_conf=0.0000 prior_center_loss_beta=0.0000 rollout_train_applied=0.0000 grad_enabled_before_rollout=1.0000 grad_enabled_after_rollout=1.0000 logits_requires_grad=1.0000 raw_loss_requires_grad=1.0000 acc_all=0.2500 corrupt_frac=1.0000 acc_corrupt=0.2500 loss_corrupt=8.2799 wrong_frac=0.7625 init_acc_corrupt=0.1375 acc_corrupt_t_0p0_0p2=0.1750 corrupt_frac_t_0p0_0p2=1.0000 out_w_norm=1.3332 out_g_norm=10.2737 acc_corrupt_t_0p2_0p4=0.2812 corrupt_frac_t_0p2_0p4=1.0000 acc_corrupt_t_0p4_0p6=0.5000 corrupt_frac_t_0p4_0p6=1.0000 loss_all=8.1875 init_gold_top10=0.2500 init_gold_top100=0.2500
+step=160 epoch=80/250 epoch_step=2/2 micro_steps=160 elapsed=4.7s lr=2.000000e-03 loss=7.4644 loss_recon=7.4644 loss_meanflow=0.0000 mean_model_t=0.1882 mean_corrupt_t=0.1882 mean_loss_t_weight=1.0000 linear_soft_target_mean_conf=0.0000 prior_center_loss_beta=0.0000 rollout_train_applied=0.0000 grad_enabled_before_rollout=1.0000 grad_enabled_after_rollout=1.0000 logits_requires_grad=1.0000 raw_loss_requires_grad=1.0000 acc_all=0.3125 corrupt_frac=1.0000 acc_corrupt=0.3125 loss_corrupt=7.4644 wrong_frac=0.7500 init_acc_corrupt=0.1750 acc_corrupt_t_0p2_0p4=0.4750 corrupt_frac_t_0p2_0p4=1.0000 out_w_norm=1.4282 out_g_norm=10.1290 acc_corrupt_t_0p0_0p2=0.1500 corrupt_frac_t_0p0_0p2=1.0000 loss_all=8.6543 init_gold_top10=0.0000 init_gold_top100=0.0000
+step=170 epoch=85/250 epoch_step=2/2 micro_steps=170 elapsed=4.5s lr=2.000000e-03 loss=7.5713 loss_recon=7.5713 loss_meanflow=0.0000 mean_model_t=0.2017 mean_corrupt_t=0.2017 mean_loss_t_weight=1.0000 linear_soft_target_mean_conf=0.0000 prior_center_loss_beta=0.0000 rollout_train_applied=0.0000 grad_enabled_before_rollout=1.0000 grad_enabled_after_rollout=1.0000 logits_requires_grad=1.0000 raw_loss_requires_grad=1.0000 acc_all=0.2500 corrupt_frac=1.0000 acc_corrupt=0.2500 loss_corrupt=7.5713 wrong_frac=0.7250 init_acc_corrupt=0.1000 acc_corrupt_t_0p2_0p4=0.4000 corrupt_frac_t_0p2_0p4=1.0000 out_w_norm=1.5251 out_g_norm=10.4161 acc_corrupt_t_0p0_0p2=0.1000 corrupt_frac_t_0p0_0p2=1.0000 loss_all=5.2729 init_gold_top10=0.5000 init_gold_top100=0.6250
+step=180 epoch=90/250 epoch_step=2/2 micro_steps=180 elapsed=4.2s lr=2.000000e-03 loss=7.6099 loss_recon=7.6099 loss_meanflow=0.0000 mean_model_t=0.2081 mean_corrupt_t=0.2081 mean_loss_t_weight=1.0000 linear_soft_target_mean_conf=0.0000 prior_center_loss_beta=0.0000 rollout_train_applied=0.0000 grad_enabled_before_rollout=1.0000 grad_enabled_after_rollout=1.0000 logits_requires_grad=1.0000 raw_loss_requires_grad=1.0000 acc_all=0.2250 corrupt_frac=1.0000 acc_corrupt=0.2250 loss_corrupt=7.6099 wrong_frac=0.7875 init_acc_corrupt=0.1250 acc_corrupt_t_0p0_0p2=0.1458 corrupt_frac_t_0p0_0p2=1.0000 out_w_norm=1.6215 out_g_norm=9.9961 acc_corrupt_t_0p2_0p4=0.3438 corrupt_frac_t_0p2_0p4=1.0000 loss_all=7.9766 init_gold_top10=0.3750 init_gold_top100=0.5000
+step=190 epoch=95/250 epoch_step=2/2 micro_steps=190 elapsed=4.6s lr=2.000000e-03 loss=7.4357 loss_recon=7.4357 loss_meanflow=0.0000 mean_model_t=0.1942 mean_corrupt_t=0.1942 mean_loss_t_weight=1.0000 linear_soft_target_mean_conf=0.0000 prior_center_loss_beta=0.0000 rollout_train_applied=0.0000 grad_enabled_before_rollout=1.0000 grad_enabled_after_rollout=1.0000 logits_requires_grad=1.0000 raw_loss_requires_grad=1.0000 acc_all=0.2250 corrupt_frac=1.0000 acc_corrupt=0.2250 loss_corrupt=7.4357 wrong_frac=0.8625 init_acc_corrupt=0.0625 acc_corrupt_t_0p0_0p2=0.1667 corrupt_frac_t_0p0_0p2=1.0000 out_w_norm=1.7179 out_g_norm=10.3147 acc_corrupt_t_0p2_0p4=0.3125 corrupt_frac_t_0p2_0p4=1.0000 loss_all=8.7559 init_gold_top10=0.1250 init_gold_top100=0.1250
+step=200 epoch=100/250 epoch_step=2/2 micro_steps=200 elapsed=4.6s lr=2.000000e-03 loss=7.6464 loss_recon=7.6464 loss_meanflow=0.0000 mean_model_t=0.1970 mean_corrupt_t=0.1970 mean_loss_t_weight=1.0000 linear_soft_target_mean_conf=0.0000 prior_center_loss_beta=0.0000 rollout_train_applied=0.0000 grad_enabled_before_rollout=1.0000 grad_enabled_after_rollout=1.0000 logits_requires_grad=1.0000 raw_loss_requires_grad=1.0000 acc_all=0.1625 corrupt_frac=1.0000 acc_corrupt=0.1625 loss_corrupt=7.6464 wrong_frac=0.8250 init_acc_corrupt=0.0625 acc_corrupt_t_0p0_0p2=0.1071 corrupt_frac_t_0p0_0p2=1.0000 out_w_norm=1.8126 out_g_norm=10.9069 acc_corrupt_t_0p2_0p4=0.2917 corrupt_frac_t_0p2_0p4=1.0000 loss_all=7.8730 init_gold_top10=0.0000 init_gold_top100=0.3750
+step=210 epoch=105/250 epoch_step=2/2 micro_steps=210 elapsed=4.2s lr=2.000000e-03 loss=6.5814 loss_recon=6.5814 loss_meanflow=0.0000 mean_model_t=0.1880 mean_corrupt_t=0.1880 mean_loss_t_weight=1.0000 linear_soft_target_mean_conf=0.0000 prior_center_loss_beta=0.0000 rollout_train_applied=0.0000 grad_enabled_before_rollout=1.0000 grad_enabled_after_rollout=1.0000 logits_requires_grad=1.0000 raw_loss_requires_grad=1.0000 acc_all=0.2375 corrupt_frac=1.0000 acc_corrupt=0.2375 loss_corrupt=6.5814 wrong_frac=0.8500 init_acc_corrupt=0.0875 acc_corrupt_t_0p2_0p4=0.2812 corrupt_frac_t_0p2_0p4=1.0000 out_w_norm=1.9072 out_g_norm=9.6478 acc_corrupt_t_0p0_0p2=0.2083 corrupt_frac_t_0p0_0p2=1.0000 loss_all=4.4038 init_gold_top10=0.2500 init_gold_top100=0.3750
+step=220 epoch=110/250 epoch_step=2/2 micro_steps=220 elapsed=4.6s lr=2.000000e-03 loss=7.1286 loss_recon=7.1286 loss_meanflow=0.0000 mean_model_t=0.1718 mean_corrupt_t=0.1718 mean_loss_t_weight=1.0000 linear_soft_target_mean_conf=0.0000 prior_center_loss_beta=0.0000 rollout_train_applied=0.0000 grad_enabled_before_rollout=1.0000 grad_enabled_after_rollout=1.0000 logits_requires_grad=1.0000 raw_loss_requires_grad=1.0000 acc_all=0.1625 corrupt_frac=1.0000 acc_corrupt=0.1625 loss_corrupt=7.1286 wrong_frac=0.8500 init_acc_corrupt=0.0375 acc_corrupt_t_0p0_0p2=0.1667 corrupt_frac_t_0p0_0p2=1.0000 out_w_norm=1.9991 out_g_norm=10.3675 acc_corrupt_t_0p2_0p4=0.1562 corrupt_frac_t_0p2_0p4=1.0000 loss_all=7.7090 init_gold_top10=0.1250 init_gold_top100=0.1250
+step=230 epoch=115/250 epoch_step=2/2 micro_steps=230 elapsed=4.6s lr=2.000000e-03 loss=4.9539 loss_recon=4.9539 loss_meanflow=0.0000 mean_model_t=0.3049 mean_corrupt_t=0.3049 mean_loss_t_weight=1.0000 linear_soft_target_mean_conf=0.0000 prior_center_loss_beta=0.0000 rollout_train_applied=0.0000 grad_enabled_before_rollout=1.0000 grad_enabled_after_rollout=1.0000 logits_requires_grad=1.0000 raw_loss_requires_grad=1.0000 acc_all=0.4250 corrupt_frac=1.0000 acc_corrupt=0.4250 loss_corrupt=4.9539 wrong_frac=0.6500 init_acc_corrupt=0.2875 acc_corrupt_t_0p4_0p6=0.7917 corrupt_frac_t_0p4_0p6=1.0000 out_w_norm=2.0831 out_g_norm=9.7181 acc_corrupt_t_0p2_0p4=0.4167 corrupt_frac_t_0p2_0p4=1.0000 acc_corrupt_t_0p0_0p2=0.1562 corrupt_frac_t_0p0_0p2=1.0000 loss_all=7.3464 init_gold_top10=0.2500 init_gold_top100=0.2500
+step=240 epoch=120/250 epoch_step=2/2 micro_steps=240 elapsed=4.1s lr=2.000000e-03 loss=5.3579 loss_recon=5.3579 loss_meanflow=0.0000 mean_model_t=0.2693 mean_corrupt_t=0.2693 mean_loss_t_weight=1.0000 linear_soft_target_mean_conf=0.0000 prior_center_loss_beta=0.0000 rollout_train_applied=0.0000 grad_enabled_before_rollout=1.0000 grad_enabled_after_rollout=1.0000 logits_requires_grad=1.0000 raw_loss_requires_grad=1.0000 acc_all=0.3500 corrupt_frac=1.0000 acc_corrupt=0.3500 loss_corrupt=5.3579 wrong_frac=0.7125 init_acc_corrupt=0.2000 acc_corrupt_t_0p0_0p2=0.1562 corrupt_frac_t_0p0_0p2=1.0000 out_w_norm=2.1655 out_g_norm=10.1470 acc_corrupt_t_0p4_0p6=0.6250 corrupt_frac_t_0p4_0p6=1.0000 acc_corrupt_t_0p2_0p4=0.4062 corrupt_frac_t_0p2_0p4=1.0000 loss_all=4.7973 init_gold_top10=0.2500 init_gold_top100=0.2500
+step=250 epoch=125/250 epoch_step=2/2 micro_steps=250 elapsed=4.6s lr=2.000000e-03 loss=5.4397 loss_recon=5.4397 loss_meanflow=0.0000 mean_model_t=0.1964 mean_corrupt_t=0.1964 mean_loss_t_weight=1.0000 linear_soft_target_mean_conf=0.0000 prior_center_loss_beta=0.0000 rollout_train_applied=0.0000 grad_enabled_before_rollout=1.0000 grad_enabled_after_rollout=1.0000 logits_requires_grad=1.0000 raw_loss_requires_grad=1.0000 acc_all=0.3000 corrupt_frac=1.0000 acc_corrupt=0.3000 loss_corrupt=5.4397 wrong_frac=0.7500 init_acc_corrupt=0.1250 acc_corrupt_t_0p0_0p2=0.1875 corrupt_frac_t_0p0_0p2=1.0000 out_w_norm=2.2462 out_g_norm=10.4779 acc_corrupt_t_0p6_0p8=0.8750 corrupt_frac_t_0p6_0p8=1.0000 acc_corrupt_t_0p2_0p4=0.6250 corrupt_frac_t_0p2_0p4=1.0000 loss_all=2.5704 init_gold_top10=0.5000 init_gold_top100=0.5000
+step=260 epoch=130/250 epoch_step=2/2 micro_steps=260 elapsed=4.5s lr=2.000000e-03 loss=6.4952 loss_recon=6.4952 loss_meanflow=0.0000 mean_model_t=0.1517 mean_corrupt_t=0.1517 mean_loss_t_weight=1.0000 linear_soft_target_mean_conf=0.0000 prior_center_loss_beta=0.0000 rollout_train_applied=0.0000 grad_enabled_before_rollout=1.0000 grad_enabled_after_rollout=1.0000 logits_requires_grad=1.0000 raw_loss_requires_grad=1.0000 acc_all=0.1375 corrupt_frac=1.0000 acc_corrupt=0.1375 loss_corrupt=6.4952 wrong_frac=0.8750 init_acc_corrupt=0.0250 acc_corrupt_t_0p0_0p2=0.1250 corrupt_frac_t_0p0_0p2=1.0000 out_w_norm=2.3208 out_g_norm=10.8209 acc_corrupt_t_0p2_0p4=0.1667 corrupt_frac_t_0p2_0p4=1.0000 loss_all=4.7949 init_gold_top10=0.1250 init_gold_top100=0.5000
+step=270 epoch=135/250 epoch_step=2/2 micro_steps=270 elapsed=4.2s lr=2.000000e-03 loss=5.5522 loss_recon=5.5522 loss_meanflow=0.0000 mean_model_t=0.1781 mean_corrupt_t=0.1781 mean_loss_t_weight=1.0000 linear_soft_target_mean_conf=0.0000 prior_center_loss_beta=0.0000 rollout_train_applied=0.0000 grad_enabled_before_rollout=1.0000 grad_enabled_after_rollout=1.0000 logits_requires_grad=1.0000 raw_loss_requires_grad=1.0000 acc_all=0.2625 corrupt_frac=1.0000 acc_corrupt=0.2625 loss_corrupt=5.5522 wrong_frac=0.7750 init_acc_corrupt=0.0875 acc_corrupt_t_0p0_0p2=0.1964 corrupt_frac_t_0p0_0p2=1.0000 out_w_norm=2.3848 out_g_norm=10.4100 acc_corrupt_t_0p2_0p4=0.4167 corrupt_frac_t_0p2_0p4=1.0000 loss_all=4.8256 init_gold_top10=0.3750 init_gold_top100=0.3750
+step=280 epoch=140/250 epoch_step=2/2 micro_steps=280 elapsed=5.0s lr=2.000000e-03 loss=5.8259 loss_recon=5.8259 loss_meanflow=0.0000 mean_model_t=0.2199 mean_corrupt_t=0.2199 mean_loss_t_weight=1.0000 linear_soft_target_mean_conf=0.0000 prior_center_loss_beta=0.0000 rollout_train_applied=0.0000 grad_enabled_before_rollout=1.0000 grad_enabled_after_rollout=1.0000 logits_requires_grad=1.0000 raw_loss_requires_grad=1.0000 acc_all=0.1625 corrupt_frac=1.0000 acc_corrupt=0.1625 loss_corrupt=5.8259 wrong_frac=0.8000 init_acc_corrupt=0.0625 acc_corrupt_t_0p4_0p6=0.1250 corrupt_frac_t_0p4_0p6=1.0000 out_w_norm=2.4410 out_g_norm=10.8257 acc_corrupt_t_0p0_0p2=0.1250 corrupt_frac_t_0p0_0p2=1.0000 acc_corrupt_t_0p2_0p4=0.2188 corrupt_frac_t_0p2_0p4=1.0000 loss_all=5.9141 init_gold_top10=0.1250 init_gold_top100=0.2500
+step=290 epoch=145/250 epoch_step=2/2 micro_steps=290 elapsed=4.5s lr=2.000000e-03 loss=4.2819 loss_recon=4.2819 loss_meanflow=0.0000 mean_model_t=0.2100 mean_corrupt_t=0.2100 mean_loss_t_weight=1.0000 linear_soft_target_mean_conf=0.0000 prior_center_loss_beta=0.0000 rollout_train_applied=0.0000 grad_enabled_before_rollout=1.0000 grad_enabled_after_rollout=1.0000 logits_requires_grad=1.0000 raw_loss_requires_grad=1.0000 acc_all=0.3250 corrupt_frac=1.0000 acc_corrupt=0.3250 loss_corrupt=4.2819 wrong_frac=0.7625 init_acc_corrupt=0.1625 acc_corrupt_t_0p2_0p4=0.4500 corrupt_frac_t_0p2_0p4=1.0000 out_w_norm=2.4965 out_g_norm=10.5429 acc_corrupt_t_0p0_0p2=0.2000 corrupt_frac_t_0p0_0p2=1.0000 loss_all=4.9616 init_gold_top10=0.2500 init_gold_top100=0.2500
+step=300 epoch=150/250 epoch_step=2/2 micro_steps=300 elapsed=4.2s lr=2.000000e-03 loss=4.6966 loss_recon=4.6966 loss_meanflow=0.0000 mean_model_t=0.2037 mean_corrupt_t=0.2037 mean_loss_t_weight=1.0000 linear_soft_target_mean_conf=0.0000 prior_center_loss_beta=0.0000 rollout_train_applied=0.0000 grad_enabled_before_rollout=1.0000 grad_enabled_after_rollout=1.0000 logits_requires_grad=1.0000 raw_loss_requires_grad=1.0000 acc_all=0.3000 corrupt_frac=1.0000 acc_corrupt=0.3000 loss_corrupt=4.6966 wrong_frac=0.7250 init_acc_corrupt=0.0875 acc_corrupt_t_0p0_0p2=0.2500 corrupt_frac_t_0p0_0p2=1.0000 out_w_norm=2.5467 out_g_norm=10.7409 acc_corrupt_t_0p2_0p4=0.3500 corrupt_frac_t_0p2_0p4=1.0000 loss_all=5.9668 init_gold_top10=0.1250 init_gold_top100=0.5000
+step=310 epoch=155/250 epoch_step=2/2 micro_steps=310 elapsed=5.0s lr=2.000000e-03 loss=4.8928 loss_recon=4.8928 loss_meanflow=0.0000 mean_model_t=0.1660 mean_corrupt_t=0.1660 mean_loss_t_weight=1.0000 linear_soft_target_mean_conf=0.0000 prior_center_loss_beta=0.0000 rollout_train_applied=0.0000 grad_enabled_before_rollout=1.0000 grad_enabled_after_rollout=1.0000 logits_requires_grad=1.0000 raw_loss_requires_grad=1.0000 acc_all=0.2750 corrupt_frac=1.0000 acc_corrupt=0.2750 loss_corrupt=4.8928 wrong_frac=0.8500 init_acc_corrupt=0.0875 acc_corrupt_t_0p0_0p2=0.2321 corrupt_frac_t_0p0_0p2=1.0000 out_w_norm=2.5929 out_g_norm=10.7503 acc_corrupt_t_0p4_0p6=0.8750 corrupt_frac_t_0p4_0p6=1.0000 acc_corrupt_t_0p2_0p4=0.1250 corrupt_frac_t_0p2_0p4=1.0000 loss_all=6.3398 init_gold_top10=0.0000 init_gold_top100=0.0000
+step=320 epoch=160/250 epoch_step=2/2 micro_steps=320 elapsed=4.5s lr=2.000000e-03 loss=4.6516 loss_recon=4.6516 loss_meanflow=0.0000 mean_model_t=0.2203 mean_corrupt_t=0.2203 mean_loss_t_weight=1.0000 linear_soft_target_mean_conf=0.0000 prior_center_loss_beta=0.0000 rollout_train_applied=0.0000 grad_enabled_before_rollout=1.0000 grad_enabled_after_rollout=1.0000 logits_requires_grad=1.0000 raw_loss_requires_grad=1.0000 acc_all=0.2125 corrupt_frac=1.0000 acc_corrupt=0.2125 loss_corrupt=4.6516 wrong_frac=0.8250 init_acc_corrupt=0.1125 acc_corrupt_t_0p0_0p2=0.1458 corrupt_frac_t_0p0_0p2=1.0000 out_w_norm=2.6345 out_g_norm=10.9623 acc_corrupt_t_0p4_0p6=0.2500 corrupt_frac_t_0p4_0p6=1.0000 acc_corrupt_t_0p2_0p4=0.3333 corrupt_frac_t_0p2_0p4=1.0000 loss_all=4.9951 init_gold_top10=0.0000 init_gold_top100=0.3750
+step=330 epoch=165/250 epoch_step=2/2 micro_steps=330 elapsed=4.2s lr=2.000000e-03 loss=3.6249 loss_recon=3.6249 loss_meanflow=0.0000 mean_model_t=0.2293 mean_corrupt_t=0.2293 mean_loss_t_weight=1.0000 linear_soft_target_mean_conf=0.0000 prior_center_loss_beta=0.0000 rollout_train_applied=0.0000 grad_enabled_before_rollout=1.0000 grad_enabled_after_rollout=1.0000 logits_requires_grad=1.0000 raw_loss_requires_grad=1.0000 acc_all=0.3625 corrupt_frac=1.0000 acc_corrupt=0.3625 loss_corrupt=3.6249 wrong_frac=0.6750 init_acc_corrupt=0.2125 acc_corrupt_t_0p0_0p2=0.1562 corrupt_frac_t_0p0_0p2=1.0000 out_w_norm=2.6737 out_g_norm=11.0498 acc_corrupt_t_0p2_0p4=0.5000 corrupt_frac_t_0p2_0p4=1.0000 loss_all=1.3865 init_gold_top10=0.6250 init_gold_top100=0.6250
+step=340 epoch=170/250 epoch_step=2/2 micro_steps=340 elapsed=5.1s lr=2.000000e-03 loss=3.6906 loss_recon=3.6906 loss_meanflow=0.0000 mean_model_t=0.2682 mean_corrupt_t=0.2682 mean_loss_t_weight=1.0000 linear_soft_target_mean_conf=0.0000 prior_center_loss_beta=0.0000 rollout_train_applied=0.0000 grad_enabled_before_rollout=1.0000 grad_enabled_after_rollout=1.0000 logits_requires_grad=1.0000 raw_loss_requires_grad=1.0000 acc_all=0.4000 corrupt_frac=1.0000 acc_corrupt=0.4000 loss_corrupt=3.6906 wrong_frac=0.6875 init_acc_corrupt=0.2375 acc_corrupt_t_0p0_0p2=0.1250 corrupt_frac_t_0p0_0p2=1.0000 out_w_norm=2.7084 out_g_norm=10.1418 acc_corrupt_t_0p2_0p4=0.5417 corrupt_frac_t_0p2_0p4=1.0000 acc_corrupt_t_0p4_0p6=0.7500 corrupt_frac_t_0p4_0p6=1.0000 acc_corrupt_t_0p6_0p8=1.0000 corrupt_frac_t_0p6_0p8=1.0000 loss_all=5.2021 init_gold_top10=0.1250 init_gold_top100=0.1250
+step=350 epoch=175/250 epoch_step=2/2 micro_steps=350 elapsed=4.5s lr=2.000000e-03 loss=4.6338 loss_recon=4.6338 loss_meanflow=0.0000 mean_model_t=0.1929 mean_corrupt_t=0.1929 mean_loss_t_weight=1.0000 linear_soft_target_mean_conf=0.0000 prior_center_loss_beta=0.0000 rollout_train_applied=0.0000 grad_enabled_before_rollout=1.0000 grad_enabled_after_rollout=1.0000 logits_requires_grad=1.0000 raw_loss_requires_grad=1.0000 acc_all=0.2000 corrupt_frac=1.0000 acc_corrupt=0.2000 loss_corrupt=4.6338 wrong_frac=0.8375 init_acc_corrupt=0.0750 acc_corrupt_t_0p2_0p4=0.3250 corrupt_frac_t_0p2_0p4=1.0000 out_w_norm=2.7401 out_g_norm=11.0149 acc_corrupt_t_0p0_0p2=0.0750 corrupt_frac_t_0p0_0p2=1.0000 loss_all=3.8555 init_gold_top10=0.1250 init_gold_top100=0.2500
+step=360 epoch=180/250 epoch_step=2/2 micro_steps=360 elapsed=4.2s lr=2.000000e-03 loss=4.2104 loss_recon=4.2104 loss_meanflow=0.0000 mean_model_t=0.1923 mean_corrupt_t=0.1923 mean_loss_t_weight=1.0000 linear_soft_target_mean_conf=0.0000 prior_center_loss_beta=0.0000 rollout_train_applied=0.0000 grad_enabled_before_rollout=1.0000 grad_enabled_after_rollout=1.0000 logits_requires_grad=1.0000 raw_loss_requires_grad=1.0000 acc_all=0.2000 corrupt_frac=1.0000 acc_corrupt=0.2000 loss_corrupt=4.2104 wrong_frac=0.7875 init_acc_corrupt=0.1000 acc_corrupt_t_0p0_0p2=0.1250 corrupt_frac_t_0p0_0p2=1.0000 out_w_norm=2.7663 out_g_norm=11.0627 acc_corrupt_t_0p4_0p6=0.5000 corrupt_frac_t_0p4_0p6=1.0000 loss_all=1.9618 init_gold_top10=0.2500 init_gold_top100=0.2500
+step=370 epoch=185/250 epoch_step=2/2 micro_steps=370 elapsed=5.0s lr=2.000000e-03 loss=3.6980 loss_recon=3.6980 loss_meanflow=0.0000 mean_model_t=0.1757 mean_corrupt_t=0.1757 mean_loss_t_weight=1.0000 linear_soft_target_mean_conf=0.0000 prior_center_loss_beta=0.0000 rollout_train_applied=0.0000 grad_enabled_before_rollout=1.0000 grad_enabled_after_rollout=1.0000 logits_requires_grad=1.0000 raw_loss_requires_grad=1.0000 acc_all=0.3125 corrupt_frac=1.0000 acc_corrupt=0.3125 loss_corrupt=3.6980 wrong_frac=0.8250 init_acc_corrupt=0.0750 acc_corrupt_t_0p2_0p4=0.5625 corrupt_frac_t_0p2_0p4=1.0000 out_w_norm=2.7910 out_g_norm=10.5404 acc_corrupt_t_0p0_0p2=0.2500 corrupt_frac_t_0p0_0p2=1.0000 loss_all=3.2881 init_gold_top10=0.2500 init_gold_top100=0.5000
+step=380 epoch=190/250 epoch_step=2/2 micro_steps=380 elapsed=4.5s lr=2.000000e-03 loss=3.2751 loss_recon=3.2751 loss_meanflow=0.0000 mean_model_t=0.1794 mean_corrupt_t=0.1794 mean_loss_t_weight=1.0000 linear_soft_target_mean_conf=0.0000 prior_center_loss_beta=0.0000 rollout_train_applied=0.0000 grad_enabled_before_rollout=1.0000 grad_enabled_after_rollout=1.0000 logits_requires_grad=1.0000 raw_loss_requires_grad=1.0000 acc_all=0.2625 corrupt_frac=1.0000 acc_corrupt=0.2625 loss_corrupt=3.2751 wrong_frac=0.8375 init_acc_corrupt=0.1000 acc_corrupt_t_0p0_0p2=0.1607 corrupt_frac_t_0p0_0p2=1.0000 out_w_norm=2.8162 out_g_norm=11.1712 acc_corrupt_t_0p4_0p6=0.5000 corrupt_frac_t_0p4_0p6=1.0000 acc_corrupt_t_0p2_0p4=0.5000 corrupt_frac_t_0p2_0p4=1.0000 loss_all=2.6837 init_gold_top10=0.2500 init_gold_top100=0.2500
+step=390 epoch=195/250 epoch_step=2/2 micro_steps=390 elapsed=4.2s lr=2.000000e-03 loss=4.0428 loss_recon=4.0428 loss_meanflow=0.0000 mean_model_t=0.1873 mean_corrupt_t=0.1873 mean_loss_t_weight=1.0000 linear_soft_target_mean_conf=0.0000 prior_center_loss_beta=0.0000 rollout_train_applied=0.0000 grad_enabled_before_rollout=1.0000 grad_enabled_after_rollout=1.0000 logits_requires_grad=1.0000 raw_loss_requires_grad=1.0000 acc_all=0.2250 corrupt_frac=1.0000 acc_corrupt=0.2250 loss_corrupt=4.0428 wrong_frac=0.8125 init_acc_corrupt=0.0750 acc_corrupt_t_0p2_0p4=0.2812 corrupt_frac_t_0p2_0p4=1.0000 out_w_norm=2.8404 out_g_norm=12.1307 acc_corrupt_t_0p0_0p2=0.1875 corrupt_frac_t_0p0_0p2=1.0000 loss_all=5.0107 init_gold_top10=0.1250 init_gold_top100=0.5000
+step=400 epoch=200/250 epoch_step=2/2 micro_steps=400 elapsed=5.0s lr=2.000000e-03 loss=3.0034 loss_recon=3.0034 loss_meanflow=0.0000 mean_model_t=0.3288 mean_corrupt_t=0.3288 mean_loss_t_weight=1.0000 linear_soft_target_mean_conf=0.0000 prior_center_loss_beta=0.0000 rollout_train_applied=0.0000 grad_enabled_before_rollout=1.0000 grad_enabled_after_rollout=1.0000 logits_requires_grad=1.0000 raw_loss_requires_grad=1.0000 acc_all=0.3625 corrupt_frac=1.0000 acc_corrupt=0.3625 loss_corrupt=3.0034 wrong_frac=0.7250 init_acc_corrupt=0.2250 acc_corrupt_t_0p0_0p2=0.1250 corrupt_frac_t_0p0_0p2=1.0000 out_w_norm=2.8626 out_g_norm=10.6248 acc_corrupt_t_0p2_0p4=0.2188 corrupt_frac_t_0p2_0p4=1.0000 acc_corrupt_t_0p6_0p8=0.8750 corrupt_frac_t_0p6_0p8=1.0000 acc_corrupt_t_0p4_0p6=0.7500 corrupt_frac_t_0p4_0p6=1.0000 loss_all=0.2258 init_gold_top10=0.7500 init_gold_top100=0.7500
+step=410 epoch=205/250 epoch_step=2/2 micro_steps=410 elapsed=4.6s lr=2.000000e-03 loss=2.5523 loss_recon=2.5523 loss_meanflow=0.0000 mean_model_t=0.2749 mean_corrupt_t=0.2749 mean_loss_t_weight=1.0000 linear_soft_target_mean_conf=0.0000 prior_center_loss_beta=0.0000 rollout_train_applied=0.0000 grad_enabled_before_rollout=1.0000 grad_enabled_after_rollout=1.0000 logits_requires_grad=1.0000 raw_loss_requires_grad=1.0000 acc_all=0.4500 corrupt_frac=1.0000 acc_corrupt=0.4500 loss_corrupt=2.5523 wrong_frac=0.6750 init_acc_corrupt=0.2375 acc_corrupt_t_0p0_0p2=0.1250 corrupt_frac_t_0p0_0p2=1.0000 out_w_norm=2.8829 out_g_norm=11.8955 acc_corrupt_t_0p2_0p4=0.5750 corrupt_frac_t_0p2_0p4=1.0000 acc_corrupt_t_0p4_0p6=0.6250 corrupt_frac_t_0p4_0p6=1.0000 loss_all=1.8015 init_gold_top10=0.3750 init_gold_top100=0.5000
+step=420 epoch=210/250 epoch_step=2/2 micro_steps=420 elapsed=4.1s lr=2.000000e-03 loss=3.2439 loss_recon=3.2439 loss_meanflow=0.0000 mean_model_t=0.2123 mean_corrupt_t=0.2123 mean_loss_t_weight=1.0000 linear_soft_target_mean_conf=0.0000 prior_center_loss_beta=0.0000 rollout_train_applied=0.0000 grad_enabled_before_rollout=1.0000 grad_enabled_after_rollout=1.0000 logits_requires_grad=1.0000 raw_loss_requires_grad=1.0000 acc_all=0.2750 corrupt_frac=1.0000 acc_corrupt=0.2750 loss_corrupt=3.2439 wrong_frac=0.7750 init_acc_corrupt=0.1000 acc_corrupt_t_0p2_0p4=0.3333 corrupt_frac_t_0p2_0p4=1.0000 out_w_norm=2.8965 out_g_norm=11.8560 acc_corrupt_t_0p0_0p2=0.1875 corrupt_frac_t_0p0_0p2=1.0000 loss_all=4.2246 init_gold_top10=0.1250 init_gold_top100=0.3750
+step=430 epoch=215/250 epoch_step=2/2 micro_steps=430 elapsed=5.0s lr=2.000000e-03 loss=3.1938 loss_recon=3.1938 loss_meanflow=0.0000 mean_model_t=0.1772 mean_corrupt_t=0.1772 mean_loss_t_weight=1.0000 linear_soft_target_mean_conf=0.0000 prior_center_loss_beta=0.0000 rollout_train_applied=0.0000 grad_enabled_before_rollout=1.0000 grad_enabled_after_rollout=1.0000 logits_requires_grad=1.0000 raw_loss_requires_grad=1.0000 acc_all=0.2625 corrupt_frac=1.0000 acc_corrupt=0.2625 loss_corrupt=3.1938 wrong_frac=0.8250 init_acc_corrupt=0.0875 acc_corrupt_t_0p0_0p2=0.2000 corrupt_frac_t_0p0_0p2=1.0000 out_w_norm=2.9060 out_g_norm=11.6931 acc_corrupt_t_0p2_0p4=0.3250 corrupt_frac_t_0p2_0p4=1.0000 loss_all=3.3418 init_gold_top10=0.1250 init_gold_top100=0.3750
+step=440 epoch=220/250 epoch_step=2/2 micro_steps=440 elapsed=4.6s lr=2.000000e-03 loss=2.7462 loss_recon=2.7462 loss_meanflow=0.0000 mean_model_t=0.2333 mean_corrupt_t=0.2333 mean_loss_t_weight=1.0000 linear_soft_target_mean_conf=0.0000 prior_center_loss_beta=0.0000 rollout_train_applied=0.0000 grad_enabled_before_rollout=1.0000 grad_enabled_after_rollout=1.0000 logits_requires_grad=1.0000 raw_loss_requires_grad=1.0000 acc_all=0.3625 corrupt_frac=1.0000 acc_corrupt=0.3625 loss_corrupt=2.7462 wrong_frac=0.6875 init_acc_corrupt=0.1875 acc_corrupt_t_0p0_0p2=0.1250 corrupt_frac_t_0p0_0p2=1.0000 out_w_norm=2.9144 out_g_norm=11.5046 acc_corrupt_t_0p2_0p4=0.5208 corrupt_frac_t_0p2_0p4=1.0000 loss_all=3.6408 init_gold_top10=0.3750 init_gold_top100=0.6250
+step=450 epoch=225/250 epoch_step=2/2 micro_steps=450 elapsed=4.2s lr=2.000000e-03 loss=3.8439 loss_recon=3.8439 loss_meanflow=0.0000 mean_model_t=0.1578 mean_corrupt_t=0.1578 mean_loss_t_weight=1.0000 linear_soft_target_mean_conf=0.0000 prior_center_loss_beta=0.0000 rollout_train_applied=0.0000 grad_enabled_before_rollout=1.0000 grad_enabled_after_rollout=1.0000 logits_requires_grad=1.0000 raw_loss_requires_grad=1.0000 acc_all=0.1625 corrupt_frac=1.0000 acc_corrupt=0.1625 loss_corrupt=3.8439 wrong_frac=0.8375 init_acc_corrupt=0.0500 acc_corrupt_t_0p0_0p2=0.1250 corrupt_frac_t_0p0_0p2=1.0000 out_w_norm=2.9216 out_g_norm=10.7069 acc_corrupt_t_0p2_0p4=0.2500 corrupt_frac_t_0p2_0p4=1.0000 loss_all=4.7910 init_gold_top10=0.0000 init_gold_top100=0.0000
+step=460 epoch=230/250 epoch_step=2/2 micro_steps=460 elapsed=5.0s lr=2.000000e-03 loss=3.0667 loss_recon=3.0667 loss_meanflow=0.0000 mean_model_t=0.1884 mean_corrupt_t=0.1884 mean_loss_t_weight=1.0000 linear_soft_target_mean_conf=0.0000 prior_center_loss_beta=0.0000 rollout_train_applied=0.0000 grad_enabled_before_rollout=1.0000 grad_enabled_after_rollout=1.0000 logits_requires_grad=1.0000 raw_loss_requires_grad=1.0000 acc_all=0.3000 corrupt_frac=1.0000 acc_corrupt=0.3000 loss_corrupt=3.0667 wrong_frac=0.8250 init_acc_corrupt=0.1000 acc_corrupt_t_0p0_0p2=0.1429 corrupt_frac_t_0p0_0p2=1.0000 out_w_norm=2.9302 out_g_norm=11.1555 acc_corrupt_t_0p4_0p6=0.7500 corrupt_frac_t_0p4_0p6=1.0000 acc_corrupt_t_0p2_0p4=0.5000 corrupt_frac_t_0p2_0p4=1.0000 loss_all=4.3457 init_gold_top10=0.0000 init_gold_top100=0.1250
+step=470 epoch=235/250 epoch_step=2/2 micro_steps=470 elapsed=4.5s lr=2.000000e-03 loss=3.0813 loss_recon=3.0813 loss_meanflow=0.0000 mean_model_t=0.1817 mean_corrupt_t=0.1817 mean_loss_t_weight=1.0000 linear_soft_target_mean_conf=0.0000 prior_center_loss_beta=0.0000 rollout_train_applied=0.0000 grad_enabled_before_rollout=1.0000 grad_enabled_after_rollout=1.0000 logits_requires_grad=1.0000 raw_loss_requires_grad=1.0000 acc_all=0.3000 corrupt_frac=1.0000 acc_corrupt=0.3000 loss_corrupt=3.0813 wrong_frac=0.8625 init_acc_corrupt=0.0750 acc_corrupt_t_0p0_0p2=0.1964 corrupt_frac_t_0p0_0p2=1.0000 out_w_norm=2.9401 out_g_norm=13.3409 acc_corrupt_t_0p2_0p4=0.5000 corrupt_frac_t_0p2_0p4=1.0000 acc_corrupt_t_0p4_0p6=0.6250 corrupt_frac_t_0p4_0p6=1.0000 loss_all=4.3584 init_gold_top10=0.0000 init_gold_top100=0.0000
+step=480 epoch=240/250 epoch_step=2/2 micro_steps=480 elapsed=4.2s lr=2.000000e-03 loss=2.3354 loss_recon=2.3354 loss_meanflow=0.0000 mean_model_t=0.2314 mean_corrupt_t=0.2314 mean_loss_t_weight=1.0000 linear_soft_target_mean_conf=0.0000 prior_center_loss_beta=0.0000 rollout_train_applied=0.0000 grad_enabled_before_rollout=1.0000 grad_enabled_after_rollout=1.0000 logits_requires_grad=1.0000 raw_loss_requires_grad=1.0000 acc_all=0.4375 corrupt_frac=1.0000 acc_corrupt=0.4375 loss_corrupt=2.3354 wrong_frac=0.6750 init_acc_corrupt=0.1625 acc_corrupt_t_0p0_0p2=0.2250 corrupt_frac_t_0p0_0p2=1.0000 out_w_norm=2.9456 out_g_norm=12.1579 acc_corrupt_t_0p2_0p4=0.6250 corrupt_frac_t_0p2_0p4=1.0000 acc_corrupt_t_0p4_0p6=0.7500 corrupt_frac_t_0p4_0p6=1.0000 loss_all=1.3855 init_gold_top10=0.3750 init_gold_top100=0.5000
+step=490 epoch=245/250 epoch_step=2/2 micro_steps=490 elapsed=5.1s lr=2.000000e-03 loss=2.5818 loss_recon=2.5818 loss_meanflow=0.0000 mean_model_t=0.2122 mean_corrupt_t=0.2122 mean_loss_t_weight=1.0000 linear_soft_target_mean_conf=0.0000 prior_center_loss_beta=0.0000 rollout_train_applied=0.0000 grad_enabled_before_rollout=1.0000 grad_enabled_after_rollout=1.0000 logits_requires_grad=1.0000 raw_loss_requires_grad=1.0000 acc_all=0.3875 corrupt_frac=1.0000 acc_corrupt=0.3875 loss_corrupt=2.5818 wrong_frac=0.7250 init_acc_corrupt=0.1375 acc_corrupt_t_0p0_0p2=0.3125 corrupt_frac_t_0p0_0p2=1.0000 out_w_norm=2.9460 out_g_norm=11.7760 acc_corrupt_t_0p2_0p4=0.3500 corrupt_frac_t_0p2_0p4=1.0000 acc_corrupt_t_0p4_0p6=0.8750 corrupt_frac_t_0p4_0p6=1.0000 loss_all=2.2825 init_gold_top10=0.2500 init_gold_top100=0.2500
+step=500 epoch=250/250 epoch_step=2/2 micro_steps=500 elapsed=4.5s lr=2.000000e-03 loss=3.2203 loss_recon=3.2203 loss_meanflow=0.0000 mean_model_t=0.1335 mean_corrupt_t=0.1335 mean_loss_t_weight=1.0000 linear_soft_target_mean_conf=0.0000 prior_center_loss_beta=0.0000 rollout_train_applied=0.0000 grad_enabled_before_rollout=1.0000 grad_enabled_after_rollout=1.0000 logits_requires_grad=1.0000 raw_loss_requires_grad=1.0000 acc_all=0.2125 corrupt_frac=1.0000 acc_corrupt=0.2125 loss_corrupt=3.2203 wrong_frac=0.8875 init_acc_corrupt=0.0500 acc_corrupt_t_0p2_0p4=0.2500 corrupt_frac_t_0p2_0p4=1.0000 out_w_norm=2.9466 out_g_norm=11.3624 acc_corrupt_t_0p0_0p2=0.2031 corrupt_frac_t_0p0_0p2=1.0000 loss_all=2.3523 init_gold_top10=0.2500 init_gold_top100=0.2500
+[allcorrupt] done train8_n8_allcorrupt_hard_ce_20260517_train8ctx8_allcorrupt Sun May 17 00:26:13 UTC 2026
+[allcorrupt] start train8_n8_allcorrupt_linear_soft_kl_20260517_train8ctx8_allcorrupt Sun May 17 00:26:13 UTC 2026
+[launch] gpt2 cached OWT soft-endpoint m/n pilot
+[launch] run_name=train8_n8_allcorrupt_linear_soft_kl_20260517_train8ctx8_allcorrupt
+[launch] save_dir=runs/train8_n8_allcorrupt_linear_soft_kl_20260517_train8ctx8_allcorrupt
+[launch] n=8 m=0 clean_state_mode=onehot
+[launch] mask_mixture lowk=0 all=1
+[launch] target_loss=linear_soft_kl conf=0.0->1.0 power=1.0
+[launch] cache=/e2e-data/evad-tech-vla/wanghan58/data/small_benchmarks/langflow_2604_11748/openwebtext_lta_cached_chunks/gpt2_len8_train8_overfit
+NCCL version 2.25.1+cuda12.8
+{
+  "device": "cuda:0",
+  "rank": 0,
+  "world_size": 4,
+  "samples": "owt_cached_chunks:8",
+  "vocab_size": 50257,
+  "tokenizer_vocab_size": 50257,
+  "save_dir": "runs/train8_n8_allcorrupt_linear_soft_kl_20260517_train8ctx8_allcorrupt",
+  "batch_size": 1,
+  "grad_accum": 1,
+  "effective_batch_size": 4,
+  "global_batch_size": 4,
+  "lr_schedule": "constant_warmup",
+  "optimizer": "muon",
+  "epochs": 0.0,
+  "steps_per_epoch": 2,
+  "total_steps": 500,
+  "warmup_steps": 10,
+  "warmup_epochs": -1.0,
+  "min_lr": 0.0,
+  "weight_decay": 0.1,
+  "output_weight_decay": -1.0,
+  "adamw_param_groups": "nanogpt",
+  "adam_beta1": 0.9,
+  "adam_beta2": 0.95,
+  "adam_eps": 1e-08,
+  "muon_impl": "legacy",
+  "muon_momentum": 0.95,
+  "muon_ns_steps": 5,
+  "muon_update_scale": 1.0,
+  "muon_nesterov": false,
+  "muon_width_scale": false,
+  "muon_grouping": "legacy_dim_ge_2",
+  "muon_param_count": 169453056,
+  "muon_adam_param_count": 122368,
+  "muon_param_names": [
+    "vocab_embed.embedding",
+    "sigma_map.net.0.weight",
+    "sigma_map.net.2.weight",
+    "blocks.0.attn_qkv.weight",
+    "blocks.0.attn_out.weight",
+    "blocks.0.mlp.0.weight",
+    "blocks.0.mlp.2.weight",
+    "blocks.0.adaLN_modulation.weight",
+    "blocks.1.attn_qkv.weight",
+    "blocks.1.attn_out.weight",
+    "blocks.1.mlp.0.weight",
+    "blocks.1.mlp.2.weight",
+    "blocks.1.adaLN_modulation.weight",
+    "blocks.2.attn_qkv.weight",
+    "blocks.2.attn_out.weight",
+    "blocks.2.mlp.0.weight",
+    "blocks.2.mlp.2.weight",
+    "blocks.2.adaLN_modulation.weight",
+    "blocks.3.attn_qkv.weight",
+    "blocks.3.attn_out.weight",
+    "blocks.3.mlp.0.weight",
+    "blocks.3.mlp.2.weight",
+    "blocks.3.adaLN_modulation.weight",
+    "blocks.4.attn_qkv.weight",
+    "blocks.4.attn_out.weight",
+    "blocks.4.mlp.0.weight",
+    "blocks.4.mlp.2.weight",
+    "blocks.4.adaLN_modulation.weight",
+    "blocks.5.attn_qkv.weight",
+    "blocks.5.attn_out.weight",
+    "blocks.5.mlp.0.weight",
+    "blocks.5.mlp.2.weight",
+    "blocks.5.adaLN_modulation.weight",
+    "blocks.6.attn_qkv.weight",
+    "blocks.6.attn_out.weight",
+    "blocks.6.mlp.0.weight",
+    "blocks.6.mlp.2.weight",
+    "blocks.6.adaLN_modulation.weight",
+    "blocks.7.attn_qkv.weight",
+    "blocks.7.attn_out.weight",
+    "blocks.7.mlp.0.weight",
+    "blocks.7.mlp.2.weight",
+    "blocks.7.adaLN_modulation.weight",
+    "blocks.8.attn_qkv.weight",
+    "blocks.8.attn_out.weight",
+    "blocks.8.mlp.0.weight",
+    "blocks.8.mlp.2.weight",
+    "blocks.8.adaLN_modulation.weight",
+    "blocks.9.attn_qkv.weight",
+    "blocks.9.attn_out.weight",
+    "blocks.9.mlp.0.weight",
+    "blocks.9.mlp.2.weight",
+    "blocks.9.adaLN_modulation.weight",
+    "blocks.10.attn_qkv.weight",
+    "blocks.10.attn_out.weight",
+    "blocks.10.mlp.0.weight",
+    "blocks.10.mlp.2.weight",
+    "blocks.10.adaLN_modulation.weight",
+    "blocks.11.attn_qkv.weight",
+    "blocks.11.attn_out.weight",
+    "blocks.11.mlp.0.weight",
+    "blocks.11.mlp.2.weight",
+    "blocks.11.adaLN_modulation.weight",
+    "output_layer.linear.weight",
+    "output_layer.adaLN_modulation.weight"
+  ],
+  "muon_adam_param_names": [
+    "sigma_map.net.0.bias",
+    "sigma_map.net.2.bias",
+    "blocks.0.norm1.weight",
+    "blocks.0.norm2.weight",
+    "blocks.0.mlp.0.bias",
+    "blocks.0.mlp.2.bias",
+    "blocks.0.adaLN_modulation.bias",
+    "blocks.1.norm1.weight",
+    "blocks.1.norm2.weight",
+    "blocks.1.mlp.0.bias",
+    "blocks.1.mlp.2.bias",
+    "blocks.1.adaLN_modulation.bias",
+    "blocks.2.norm1.weight",
+    "blocks.2.norm2.weight",
+    "blocks.2.mlp.0.bias",
+    "blocks.2.mlp.2.bias",
+    "blocks.2.adaLN_modulation.bias",
+    "blocks.3.norm1.weight",
+    "blocks.3.norm2.weight",
+    "blocks.3.mlp.0.bias",
+    "blocks.3.mlp.2.bias",
+    "blocks.3.adaLN_modulation.bias",
+    "blocks.4.norm1.weight",
+    "blocks.4.norm2.weight",
+    "blocks.4.mlp.0.bias",
+    "blocks.4.mlp.2.bias",
+    "blocks.4.adaLN_modulation.bias",
+    "blocks.5.norm1.weight",
+    "blocks.5.norm2.weight",
+    "blocks.5.mlp.0.bias",
+    "blocks.5.mlp.2.bias",
+    "blocks.5.adaLN_modulation.bias",
+    "blocks.6.norm1.weight",
+    "blocks.6.norm2.weight",
+    "blocks.6.mlp.0.bias",
+    "blocks.6.mlp.2.bias",
+    "blocks.6.adaLN_modulation.bias",
+    "blocks.7.norm1.weight",
+    "blocks.7.norm2.weight",
+    "blocks.7.mlp.0.bias",
+    "blocks.7.mlp.2.bias",
+    "blocks.7.adaLN_modulation.bias",
+    "blocks.8.norm1.weight",
+    "blocks.8.norm2.weight",
+    "blocks.8.mlp.0.bias",
+    "blocks.8.mlp.2.bias",
+    "blocks.8.adaLN_modulation.bias",
+    "blocks.9.norm1.weight",
+    "blocks.9.norm2.weight",
+    "blocks.9.mlp.0.bias",
+    "blocks.9.mlp.2.bias",
+    "blocks.9.adaLN_modulation.bias",
+    "blocks.10.norm1.weight",
+    "blocks.10.norm2.weight",
+    "blocks.10.mlp.0.bias",
+    "blocks.10.mlp.2.bias",
+    "blocks.10.adaLN_modulation.bias",
+    "blocks.11.norm1.weight",
+    "blocks.11.norm2.weight",
+    "blocks.11.mlp.0.bias",
+    "blocks.11.mlp.2.bias",
+    "blocks.11.adaLN_modulation.bias",
+    "output_layer.norm_final.weight",
+    "output_layer.adaLN_modulation.bias"
+  ],
+  "muon_effective_nesterov": false,
+  "muon_effective_width_scale": false,
+  "muon_effective_weight_decay": 0.1,
+  "muon_adam_fallback_nesterov": false,
+  "muon_adam_fallback_weight_decay": 0.1,
+  "ema_decay": 0.9999,
+  "ema_start_step": 0,
+  "model_type": "ddit",
+  "elf_num_time_tokens": 4,
+  "elf_num_model_mode_tokens": 0,
+  "qk_norm": true,
+  "output_bias": false,
+  "output_init_std": -1.0,
+  "norm_type": "rmsnorm",
+  "target_loss": "linear_soft_kl",
+  "linear_soft_target_power": 1.0,
+  "linear_soft_target_min_conf": 0.0,
+  "linear_soft_target_max_conf": 1.0,
+  "t_sampling_mode": "logit_normal",
+  "t_sampling_power": 1.0,
+  "t_sampling_eps": 0.0001,
+  "t_sampling_logit_mean": -1.5,
+  "t_sampling_logit_std": 0.8,
+  "dual_t": true,
+  "corrupt_t_mode": "same",
+  "corrupt_min_t": 0.0,
+  "corrupt_max_t": 1.0,
+  "prefix_block_prob": 0.0,
+  "prefix_block_len": 128,
+  "mask_ratio_floor_schedule": "none",
+  "dirichlet_endpoint_mode": "categorical_dual_t",
+  "dirichlet_semantic_t_mode": "same",
+  "dirichlet_semantic_t_value": 0.0,
+  "dirichlet_semantic_t_curve": "linear",
+  "dirichlet_semantic_t_power": 1.0,
+  "endpoint_sequence_random_prob_alpha": 0.0,
+  "categorical_wrong_from_full_vocab": true,
+  "categorical_wrong_from_batch_valid_tokens": false,
+  "categorical_wrong_basin_token_ids": "",
+  "categorical_wrong_basin_prob": 0.0,
+  "categorical_wrong_unigram_prob": 0.0,
+  "categorical_wrong_uniform_prob": 0.0,
+  "categorical_wrong_corpus_unigram_path": "",
+  "categorical_wrong_corpus_unigram_alpha": 1.0,
+  "categorical_wrong_basin_shared_prob": 0.0,
+  "categorical_wrong_unigram_shared_prob": 0.0,
+  "mask_mixture_original_prob": 0.0,
+  "mask_mixture_lowk_prob": 0.0,
+  "mask_mixture_lowcorrupt_prob": 0.0,
+  "mask_mixture_block_prob": 0.0,
+  "mask_mixture_all_prob": 1.0,
+  "mask_mixture_lowk_clean_tokens": "0",
+  "mask_mixture_lowcorrupt_tokens": "1,2,4,8,16,32,64",
+  "mask_mixture_block_tokens": "64,128",
+  "simplex_bridge_sampler": "dirichlet",
+  "logistic_normal_sigma_min": 0.18,
+  "logistic_normal_sigma_max": 2.2,
+  "logistic_normal_tau_min": 0.65,
+  "logistic_normal_tau_max": 1.15,
+  "torch_compile": false,
+  "compile_mode": "max-autotune",
+  "state_format": "prob",
+  "meanflow_weight": 0.0,
+  "rollout_train_prob": 0.0,
+  "rollout_train_steps": 1,
+  "rollout_train_infer_steps": 64,
+  "rollout_train_temp": 1.45,
+  "rollout_train_max_gamma": 1.0,
+  "rollout_train_corrupt_only": true,
+  "rollout_train_samplewise": false,
+  "rollout_train_compute_always": false,
+  "bridge_noise_init": "logistic_normal",
+  "noise_sigma": -1.0,
+  "allow_tf32": true,
+  "activation_checkpointing": false,
+  "activation_checkpoint_interval": 1,
+  "activation_checkpoint_scope": "block",
+  "ddp_static_graph": false,
+  "ddp_gradient_as_bucket_view": true,
+  "blocking_data_transfer": false,
+  "dataloader_prefetch_factor": 4,
+  "full_train_stats": false,
+  "tokenized_hf": false,
+  "tokenized_pad_token": "pad",
+  "elf_conditional_hf": false,
+  "record_pad_truncate": false,
+  "record_add_eos": false,
+  "record_add_special_tokens": false,
+  "record_pad_token": "pad",
+  "record_shuffle_buffer": 10000,
+  "wrap": true,
+  "wrap_mode": "stream",
+  "wrap_record_buffer_size": 200,
+  "owt_cached_chunks": true,
+  "owt_chunk_cache_dir": "/e2e-data/evad-tech-vla/wanghan58/data/small_benchmarks/langflow_2604_11748/openwebtext_lta_cached_chunks/gpt2_len8_train8_overfit",
+  "owt_chunk_cache_rebuild": false,
+  "owt_chunk_cache_write_batch": 4096,
+  "owt_exact_repeat_per_chunk": 0,
+  "online_chunk_shuffle": false,
+  "online_chunk_shuffle_buffer": 10000,
+  "openwebtext_split": "train_minus_100k",
+  "detokenizer": "auto",
+  "resolved_detokenizer": null,
+  "num_workers": 0,
+  "latest_every": 10,
+  "resume_path": ""
+}
+step=10 epoch=5/250 epoch_step=2/2 micro_steps=10 elapsed=1.9s lr=2.000000e-03 loss=2.5338 loss_recon=2.5338 loss_meanflow=0.0000 mean_model_t=0.2827 mean_corrupt_t=0.2827 mean_loss_t_weight=1.0000 linear_soft_target_mean_conf=0.2827 prior_center_loss_beta=0.0000 rollout_train_applied=0.0000 grad_enabled_before_rollout=1.0000 grad_enabled_after_rollout=1.0000 logits_requires_grad=1.0000 raw_loss_requires_grad=1.0000 acc_all=0.1750 corrupt_frac=1.0000 acc_corrupt=0.1750 loss_corrupt=3.1158 wrong_frac=0.7500 init_acc_corrupt=0.1625 acc_corrupt_t_0p0_0p2=0.0938 corrupt_frac_t_0p0_0p2=1.0000 out_w_norm=0.0172 out_g_norm=1.3628 acc_corrupt_t_0p6_0p8=0.1250 corrupt_frac_t_0p6_0p8=1.0000 acc_corrupt_t_0p2_0p4=0.1875 corrupt_frac_t_0p2_0p4=1.0000 acc_corrupt_t_0p4_0p6=0.5000 corrupt_frac_t_0p4_0p6=1.0000 loss_all=10.7812 init_gold_top10=0.3750 init_gold_top100=0.3750
+step=20 epoch=10/250 epoch_step=2/2 micro_steps=20 elapsed=7.0s lr=2.000000e-03 loss=2.0841 loss_recon=2.0841 loss_meanflow=0.0000 mean_model_t=0.2379 mean_corrupt_t=0.2379 mean_loss_t_weight=1.0000 linear_soft_target_mean_conf=0.2379 prior_center_loss_beta=0.0000 rollout_train_applied=0.0000 grad_enabled_before_rollout=1.0000 grad_enabled_after_rollout=1.0000 logits_requires_grad=1.0000 raw_loss_requires_grad=1.0000 acc_all=0.1750 corrupt_frac=1.0000 acc_corrupt=0.1750 loss_corrupt=3.0768 wrong_frac=0.7250 init_acc_corrupt=0.1625 acc_corrupt_t_0p0_0p2=0.1250 corrupt_frac_t_0p0_0p2=1.0000 out_w_norm=0.1058 out_g_norm=1.5830 acc_corrupt_t_0p2_0p4=0.1875 corrupt_frac_t_0p2_0p4=1.0000 acc_corrupt_t_0p6_0p8=0.5000 corrupt_frac_t_0p6_0p8=1.0000 acc_corrupt_t_0p4_0p6=0.1250 corrupt_frac_t_0p4_0p6=1.0000 loss_all=10.7266 init_gold_top10=0.1250 init_gold_top100=0.1250
+step=30 epoch=15/250 epoch_step=2/2 micro_steps=30 elapsed=6.7s lr=2.000000e-03 loss=2.1873 loss_recon=2.1873 loss_meanflow=0.0000 mean_model_t=0.2564 mean_corrupt_t=0.2564 mean_loss_t_weight=1.0000 linear_soft_target_mean_conf=0.2564 prior_center_loss_beta=0.0000 rollout_train_applied=0.0000 grad_enabled_before_rollout=1.0000 grad_enabled_after_rollout=1.0000 logits_requires_grad=1.0000 raw_loss_requires_grad=1.0000 acc_all=0.1625 corrupt_frac=1.0000 acc_corrupt=0.1625 loss_corrupt=3.0765 wrong_frac=0.7750 init_acc_corrupt=0.1750 acc_corrupt_t_0p2_0p4=0.1750 corrupt_frac_t_0p2_0p4=1.0000 out_w_norm=0.2196 out_g_norm=1.5462 acc_corrupt_t_0p4_0p6=0.1250 corrupt_frac_t_0p4_0p6=1.0000 acc_corrupt_t_0p0_0p2=0.1667 corrupt_frac_t_0p0_0p2=1.0000 loss_all=10.6484 init_gold_top10=0.0000 init_gold_top100=0.1250
+step=40 epoch=20/250 epoch_step=2/2 micro_steps=40 elapsed=6.1s lr=2.000000e-03 loss=1.5748 loss_recon=1.5748 loss_meanflow=0.0000 mean_model_t=0.1906 mean_corrupt_t=0.1906 mean_loss_t_weight=1.0000 linear_soft_target_mean_conf=0.1906 prior_center_loss_beta=0.0000 rollout_train_applied=0.0000 grad_enabled_before_rollout=1.0000 grad_enabled_after_rollout=1.0000 logits_requires_grad=1.0000 raw_loss_requires_grad=1.0000 acc_all=0.1500 corrupt_frac=1.0000 acc_corrupt=0.1500 loss_corrupt=2.5082 wrong_frac=0.8625 init_acc_corrupt=0.0875 acc_corrupt_t_0p2_0p4=0.1250 corrupt_frac_t_0p2_0p4=1.0000 out_w_norm=0.3274 out_g_norm=1.4375 acc_corrupt_t_0p4_0p6=0.2500 corrupt_frac_t_0p4_0p6=1.0000 acc_corrupt_t_0p0_0p2=0.1458 corrupt_frac_t_0p0_0p2=1.0000 loss_all=10.5859 init_gold_top10=0.0000 init_gold_top100=0.1250
+step=50 epoch=25/250 epoch_step=2/2 micro_steps=50 elapsed=5.6s lr=2.000000e-03 loss=1.2940 loss_recon=1.2940 loss_meanflow=0.0000 mean_model_t=0.1646 mean_corrupt_t=0.1646 mean_loss_t_weight=1.0000 linear_soft_target_mean_conf=0.1646 prior_center_loss_beta=0.0000 rollout_train_applied=0.0000 grad_enabled_before_rollout=1.0000 grad_enabled_after_rollout=1.0000 logits_requires_grad=1.0000 raw_loss_requires_grad=1.0000 acc_all=0.1625 corrupt_frac=1.0000 acc_corrupt=0.1625 loss_corrupt=2.2712 wrong_frac=0.8875 init_acc_corrupt=0.0250 acc_corrupt_t_0p0_0p2=0.1458 corrupt_frac_t_0p0_0p2=1.0000 out_w_norm=0.4224 out_g_norm=1.5049 acc_corrupt_t_0p2_0p4=0.1875 corrupt_frac_t_0p2_0p4=1.0000 loss_all=10.6875 init_gold_top10=0.0000 init_gold_top100=0.2500
+step=60 epoch=30/250 epoch_step=2/2 micro_steps=60 elapsed=4.7s lr=2.000000e-03 loss=1.8988 loss_recon=1.8988 loss_meanflow=0.0000 mean_model_t=0.2362 mean_corrupt_t=0.2362 mean_loss_t_weight=1.0000 linear_soft_target_mean_conf=0.2362 prior_center_loss_beta=0.0000 rollout_train_applied=0.0000 grad_enabled_before_rollout=1.0000 grad_enabled_after_rollout=1.0000 logits_requires_grad=1.0000 raw_loss_requires_grad=1.0000 acc_all=0.1625 corrupt_frac=1.0000 acc_corrupt=0.1625 loss_corrupt=2.7098 wrong_frac=0.7250 init_acc_corrupt=0.0875 acc_corrupt_t_0p0_0p2=0.1250 corrupt_frac_t_0p0_0p2=1.0000 out_w_norm=0.5076 out_g_norm=1.8652 acc_corrupt_t_0p4_0p6=0.2500 corrupt_frac_t_0p4_0p6=1.0000 acc_corrupt_t_0p2_0p4=0.1750 corrupt_frac_t_0p2_0p4=1.0000 loss_all=10.1016 init_gold_top10=0.1250 init_gold_top100=0.2500
+step=70 epoch=35/250 epoch_step=2/2 micro_steps=70 elapsed=5.2s lr=2.000000e-03 loss=1.4906 loss_recon=1.4906 loss_meanflow=0.0000 mean_model_t=0.1956 mean_corrupt_t=0.1956 mean_loss_t_weight=1.0000 linear_soft_target_mean_conf=0.1956 prior_center_loss_beta=0.0000 rollout_train_applied=0.0000 grad_enabled_before_rollout=1.0000 grad_enabled_after_rollout=1.0000 logits_requires_grad=1.0000 raw_loss_requires_grad=1.0000 acc_all=0.1500 corrupt_frac=1.0000 acc_corrupt=0.1500 loss_corrupt=2.3653 wrong_frac=0.8375 init_acc_corrupt=0.0375 acc_corrupt_t_0p2_0p4=0.1667 corrupt_frac_t_0p2_0p4=1.0000 out_w_norm=0.5888 out_g_norm=1.8518 acc_corrupt_t_0p0_0p2=0.1250 corrupt_frac_t_0p0_0p2=1.0000 loss_all=10.3281 init_gold_top10=0.1250 init_gold_top100=0.1250
+step=80 epoch=40/250 epoch_step=2/2 micro_steps=80 elapsed=4.5s lr=2.000000e-03 loss=1.8378 loss_recon=1.8378 loss_meanflow=0.0000 mean_model_t=0.2455 mean_corrupt_t=0.2455 mean_loss_t_weight=1.0000 linear_soft_target_mean_conf=0.2455 prior_center_loss_beta=0.0000 rollout_train_applied=0.0000 grad_enabled_before_rollout=1.0000 grad_enabled_after_rollout=1.0000 logits_requires_grad=1.0000 raw_loss_requires_grad=1.0000 acc_all=0.2875 corrupt_frac=1.0000 acc_corrupt=0.2875 loss_corrupt=2.5135 wrong_frac=0.7250 init_acc_corrupt=0.2000 acc_corrupt_t_0p4_0p6=0.5417 corrupt_frac_t_0p4_0p6=1.0000 out_w_norm=0.6751 out_g_norm=1.7835 acc_corrupt_t_0p0_0p2=0.1250 corrupt_frac_t_0p0_0p2=1.0000 acc_corrupt_t_0p2_0p4=0.2500 corrupt_frac_t_0p2_0p4=1.0000 loss_all=9.0312 init_gold_top10=0.5000 init_gold_top100=0.5000
+step=90 epoch=45/250 epoch_step=2/2 micro_steps=90 elapsed=4.1s lr=2.000000e-03 loss=1.6532 loss_recon=1.6532 loss_meanflow=0.0000 mean_model_t=0.2205 mean_corrupt_t=0.2205 mean_loss_t_weight=1.0000 linear_soft_target_mean_conf=0.2205 prior_center_loss_beta=0.0000 rollout_train_applied=0.0000 grad_enabled_before_rollout=1.0000 grad_enabled_after_rollout=1.0000 logits_requires_grad=1.0000 raw_loss_requires_grad=1.0000 acc_all=0.1500 corrupt_frac=1.0000 acc_corrupt=0.1500 loss_corrupt=2.2576 wrong_frac=0.7750 init_acc_corrupt=0.1375 acc_corrupt_t_0p0_0p2=0.1250 corrupt_frac_t_0p0_0p2=1.0000 out_w_norm=0.7642 out_g_norm=2.1013 acc_corrupt_t_0p2_0p4=0.1667 corrupt_frac_t_0p2_0p4=1.0000 acc_corrupt_t_0p4_0p6=0.1875 corrupt_frac_t_0p4_0p6=1.0000 loss_all=9.3867 init_gold_top10=0.3750 init_gold_top100=0.3750
+step=100 epoch=50/250 epoch_step=2/2 micro_steps=100 elapsed=5.3s lr=2.000000e-03 loss=1.2407 loss_recon=1.2407 loss_meanflow=0.0000 mean_model_t=0.1800 mean_corrupt_t=0.1800 mean_loss_t_weight=1.0000 linear_soft_target_mean_conf=0.1800 prior_center_loss_beta=0.0000 rollout_train_applied=0.0000 grad_enabled_before_rollout=1.0000 grad_enabled_after_rollout=1.0000 logits_requires_grad=1.0000 raw_loss_requires_grad=1.0000 acc_all=0.2125 corrupt_frac=1.0000 acc_corrupt=0.2125 loss_corrupt=2.1568 wrong_frac=0.8375 init_acc_corrupt=0.0875 acc_corrupt_t_0p0_0p2=0.1458 corrupt_frac_t_0p0_0p2=1.0000 out_w_norm=0.8532 out_g_norm=1.9523 acc_corrupt_t_0p2_0p4=0.3125 corrupt_frac_t_0p2_0p4=1.0000 loss_all=10.0547 init_gold_top10=0.1250 init_gold_top100=0.2500
+step=110 epoch=55/250 epoch_step=2/2 micro_steps=110 elapsed=4.4s lr=2.000000e-03 loss=1.5690 loss_recon=1.5690 loss_meanflow=0.0000 mean_model_t=0.2276 mean_corrupt_t=0.2276 mean_loss_t_weight=1.0000 linear_soft_target_mean_conf=0.2276 prior_center_loss_beta=0.0000 rollout_train_applied=0.0000 grad_enabled_before_rollout=1.0000 grad_enabled_after_rollout=1.0000 logits_requires_grad=1.0000 raw_loss_requires_grad=1.0000 acc_all=0.2375 corrupt_frac=1.0000 acc_corrupt=0.2375 loss_corrupt=2.4548 wrong_frac=0.8375 init_acc_corrupt=0.1375 acc_corrupt_t_0p0_0p2=0.1458 corrupt_frac_t_0p0_0p2=1.0000 out_w_norm=0.9430 out_g_norm=2.0350 acc_corrupt_t_0p4_0p6=0.3750 corrupt_frac_t_0p4_0p6=1.0000 acc_corrupt_t_0p2_0p4=0.3750 corrupt_frac_t_0p2_0p4=1.0000 loss_all=9.4180 init_gold_top10=0.0000 init_gold_top100=0.1250
+step=120 epoch=60/250 epoch_step=2/2 micro_steps=120 elapsed=4.1s lr=2.000000e-03 loss=1.6581 loss_recon=1.6581 loss_meanflow=0.0000 mean_model_t=0.2553 mean_corrupt_t=0.2553 mean_loss_t_weight=1.0000 linear_soft_target_mean_conf=0.2553 prior_center_loss_beta=0.0000 rollout_train_applied=0.0000 grad_enabled_before_rollout=1.0000 grad_enabled_after_rollout=1.0000 logits_requires_grad=1.0000 raw_loss_requires_grad=1.0000 acc_all=0.3250 corrupt_frac=1.0000 acc_corrupt=0.3250 loss_corrupt=2.4954 wrong_frac=0.6750 init_acc_corrupt=0.2250 acc_corrupt_t_0p2_0p4=0.3929 corrupt_frac_t_0p2_0p4=1.0000 out_w_norm=1.0350 out_g_norm=2.3059 acc_corrupt_t_0p0_0p2=0.1667 corrupt_frac_t_0p0_0p2=1.0000 loss_all=9.1484 init_gold_top10=0.5000 init_gold_top100=0.5000
+step=130 epoch=65/250 epoch_step=2/2 micro_steps=130 elapsed=4.6s lr=2.000000e-03 loss=1.4130 loss_recon=1.4130 loss_meanflow=0.0000 mean_model_t=0.2286 mean_corrupt_t=0.2286 mean_loss_t_weight=1.0000 linear_soft_target_mean_conf=0.2286 prior_center_loss_beta=0.0000 rollout_train_applied=0.0000 grad_enabled_before_rollout=1.0000 grad_enabled_after_rollout=1.0000 logits_requires_grad=1.0000 raw_loss_requires_grad=1.0000 acc_all=0.2375 corrupt_frac=1.0000 acc_corrupt=0.2375 loss_corrupt=2.1573 wrong_frac=0.7875 init_acc_corrupt=0.1375 acc_corrupt_t_0p0_0p2=0.1667 corrupt_frac_t_0p0_0p2=1.0000 out_w_norm=1.1300 out_g_norm=2.4490 acc_corrupt_t_0p2_0p4=0.2679 corrupt_frac_t_0p2_0p4=1.0000 loss_all=8.8047 init_gold_top10=0.1250 init_gold_top100=0.1250
+step=140 epoch=70/250 epoch_step=2/2 micro_steps=140 elapsed=4.4s lr=2.000000e-03 loss=1.0509 loss_recon=1.0509 loss_meanflow=0.0000 mean_model_t=0.1682 mean_corrupt_t=0.1682 mean_loss_t_weight=1.0000 linear_soft_target_mean_conf=0.1682 prior_center_loss_beta=0.0000 rollout_train_applied=0.0000 grad_enabled_before_rollout=1.0000 grad_enabled_after_rollout=1.0000 logits_requires_grad=1.0000 raw_loss_requires_grad=1.0000 acc_all=0.1875 corrupt_frac=1.0000 acc_corrupt=0.1875 loss_corrupt=1.9215 wrong_frac=0.8375 init_acc_corrupt=0.0500 acc_corrupt_t_0p0_0p2=0.1250 corrupt_frac_t_0p0_0p2=1.0000 out_w_norm=1.2307 out_g_norm=2.0366 acc_corrupt_t_0p2_0p4=0.3333 corrupt_frac_t_0p2_0p4=1.0000 loss_all=9.1719 init_gold_top10=0.1250 init_gold_top100=0.1250
+step=150 epoch=75/250 epoch_step=2/2 micro_steps=150 elapsed=4.1s lr=2.000000e-03 loss=1.3402 loss_recon=1.3402 loss_meanflow=0.0000 mean_model_t=0.2326 mean_corrupt_t=0.2326 mean_loss_t_weight=1.0000 linear_soft_target_mean_conf=0.2326 prior_center_loss_beta=0.0000 rollout_train_applied=0.0000 grad_enabled_before_rollout=1.0000 grad_enabled_after_rollout=1.0000 logits_requires_grad=1.0000 raw_loss_requires_grad=1.0000 acc_all=0.2625 corrupt_frac=1.0000 acc_corrupt=0.2625 loss_corrupt=1.9238 wrong_frac=0.7625 init_acc_corrupt=0.1375 acc_corrupt_t_0p0_0p2=0.1750 corrupt_frac_t_0p0_0p2=1.0000 out_w_norm=1.3304 out_g_norm=1.9295 acc_corrupt_t_0p2_0p4=0.3125 corrupt_frac_t_0p2_0p4=1.0000 acc_corrupt_t_0p4_0p6=0.5000 corrupt_frac_t_0p4_0p6=1.0000 loss_all=8.0742 init_gold_top10=0.2500 init_gold_top100=0.2500
+step=160 epoch=80/250 epoch_step=2/2 micro_steps=160 elapsed=4.5s lr=2.000000e-03 loss=0.8337 loss_recon=0.8337 loss_meanflow=0.0000 mean_model_t=0.1882 mean_corrupt_t=0.1882 mean_loss_t_weight=1.0000 linear_soft_target_mean_conf=0.1882 prior_center_loss_beta=0.0000 rollout_train_applied=0.0000 grad_enabled_before_rollout=1.0000 grad_enabled_after_rollout=1.0000 logits_requires_grad=1.0000 raw_loss_requires_grad=1.0000 acc_all=0.3375 corrupt_frac=1.0000 acc_corrupt=0.3375 loss_corrupt=1.6679 wrong_frac=0.7500 init_acc_corrupt=0.1750 acc_corrupt_t_0p2_0p4=0.5000 corrupt_frac_t_0p2_0p4=1.0000 out_w_norm=1.4319 out_g_norm=1.8421 acc_corrupt_t_0p0_0p2=0.1750 corrupt_frac_t_0p0_0p2=1.0000 loss_all=8.7930 init_gold_top10=0.0000 init_gold_top100=0.0000
+step=170 epoch=85/250 epoch_step=2/2 micro_steps=170 elapsed=5.3s lr=2.000000e-03 loss=0.9756 loss_recon=0.9756 loss_meanflow=0.0000 mean_model_t=0.2017 mean_corrupt_t=0.2017 mean_loss_t_weight=1.0000 linear_soft_target_mean_conf=0.2017 prior_center_loss_beta=0.0000 rollout_train_applied=0.0000 grad_enabled_before_rollout=1.0000 grad_enabled_after_rollout=1.0000 logits_requires_grad=1.0000 raw_loss_requires_grad=1.0000 acc_all=0.2250 corrupt_frac=1.0000 acc_corrupt=0.2250 loss_corrupt=1.3851 wrong_frac=0.7250 init_acc_corrupt=0.1000 acc_corrupt_t_0p2_0p4=0.3750 corrupt_frac_t_0p2_0p4=1.0000 out_w_norm=1.5338 out_g_norm=1.6552 acc_corrupt_t_0p0_0p2=0.0750 corrupt_frac_t_0p0_0p2=1.0000 loss_all=4.8604 init_gold_top10=0.5000 init_gold_top100=0.6250
+step=180 epoch=90/250 epoch_step=2/2 micro_steps=180 elapsed=4.1s lr=2.000000e-03 loss=1.0499 loss_recon=1.0499 loss_meanflow=0.0000 mean_model_t=0.2081 mean_corrupt_t=0.2081 mean_loss_t_weight=1.0000 linear_soft_target_mean_conf=0.2081 prior_center_loss_beta=0.0000 rollout_train_applied=0.0000 grad_enabled_before_rollout=1.0000 grad_enabled_after_rollout=1.0000 logits_requires_grad=1.0000 raw_loss_requires_grad=1.0000 acc_all=0.2375 corrupt_frac=1.0000 acc_corrupt=0.2375 loss_corrupt=1.7474 wrong_frac=0.7875 init_acc_corrupt=0.1250 acc_corrupt_t_0p0_0p2=0.1667 corrupt_frac_t_0p0_0p2=1.0000 out_w_norm=1.6309 out_g_norm=1.7936 acc_corrupt_t_0p2_0p4=0.3438 corrupt_frac_t_0p2_0p4=1.0000 loss_all=8.0332 init_gold_top10=0.3750 init_gold_top100=0.5000
+step=190 epoch=95/250 epoch_step=2/2 micro_steps=190 elapsed=4.5s lr=2.000000e-03 loss=0.9665 loss_recon=0.9665 loss_meanflow=0.0000 mean_model_t=0.1942 mean_corrupt_t=0.1942 mean_loss_t_weight=1.0000 linear_soft_target_mean_conf=0.1942 prior_center_loss_beta=0.0000 rollout_train_applied=0.0000 grad_enabled_before_rollout=1.0000 grad_enabled_after_rollout=1.0000 logits_requires_grad=1.0000 raw_loss_requires_grad=1.0000 acc_all=0.2250 corrupt_frac=1.0000 acc_corrupt=0.2250 loss_corrupt=1.7928 wrong_frac=0.8625 init_acc_corrupt=0.0625 acc_corrupt_t_0p0_0p2=0.1458 corrupt_frac_t_0p0_0p2=1.0000 out_w_norm=1.7201 out_g_norm=2.0560 acc_corrupt_t_0p2_0p4=0.3438 corrupt_frac_t_0p2_0p4=1.0000 loss_all=8.8516 init_gold_top10=0.1250 init_gold_top100=0.1250
+step=200 epoch=100/250 epoch_step=2/2 micro_steps=200 elapsed=4.4s lr=2.000000e-03 loss=1.0310 loss_recon=1.0310 loss_meanflow=0.0000 mean_model_t=0.1970 mean_corrupt_t=0.1970 mean_loss_t_weight=1.0000 linear_soft_target_mean_conf=0.1970 prior_center_loss_beta=0.0000 rollout_train_applied=0.0000 grad_enabled_before_rollout=1.0000 grad_enabled_after_rollout=1.0000 logits_requires_grad=1.0000 raw_loss_requires_grad=1.0000 acc_all=0.1625 corrupt_frac=1.0000 acc_corrupt=0.1625 loss_corrupt=1.8039 wrong_frac=0.8250 init_acc_corrupt=0.0625 acc_corrupt_t_0p0_0p2=0.1250 corrupt_frac_t_0p0_0p2=1.0000 out_w_norm=1.8020 out_g_norm=2.0184 acc_corrupt_t_0p2_0p4=0.2500 corrupt_frac_t_0p2_0p4=1.0000 loss_all=7.9961 init_gold_top10=0.0000 init_gold_top100=0.3750
+step=210 epoch=105/250 epoch_step=2/2 micro_steps=210 elapsed=4.1s lr=2.000000e-03 loss=0.8071 loss_recon=0.8071 loss_meanflow=0.0000 mean_model_t=0.1880 mean_corrupt_t=0.1880 mean_loss_t_weight=1.0000 linear_soft_target_mean_conf=0.1880 prior_center_loss_beta=0.0000 rollout_train_applied=0.0000 grad_enabled_before_rollout=1.0000 grad_enabled_after_rollout=1.0000 logits_requires_grad=1.0000 raw_loss_requires_grad=1.0000 acc_all=0.2500 corrupt_frac=1.0000 acc_corrupt=0.2500 loss_corrupt=1.2439 wrong_frac=0.8500 init_acc_corrupt=0.0875 acc_corrupt_t_0p2_0p4=0.3125 corrupt_frac_t_0p2_0p4=1.0000 out_w_norm=1.8702 out_g_norm=2.0898 acc_corrupt_t_0p0_0p2=0.2083 corrupt_frac_t_0p0_0p2=1.0000 loss_all=4.9170 init_gold_top10=0.2500 init_gold_top100=0.3750
+step=220 epoch=110/250 epoch_step=2/2 micro_steps=220 elapsed=4.5s lr=2.000000e-03 loss=0.8528 loss_recon=0.8528 loss_meanflow=0.0000 mean_model_t=0.1718 mean_corrupt_t=0.1718 mean_loss_t_weight=1.0000 linear_soft_target_mean_conf=0.1718 prior_center_loss_beta=0.0000 rollout_train_applied=0.0000 grad_enabled_before_rollout=1.0000 grad_enabled_after_rollout=1.0000 logits_requires_grad=1.0000 raw_loss_requires_grad=1.0000 acc_all=0.2000 corrupt_frac=1.0000 acc_corrupt=0.2000 loss_corrupt=1.6071 wrong_frac=0.8500 init_acc_corrupt=0.0375 acc_corrupt_t_0p0_0p2=0.2083 corrupt_frac_t_0p0_0p2=1.0000 out_w_norm=1.9181 out_g_norm=1.7059 acc_corrupt_t_0p2_0p4=0.1875 corrupt_frac_t_0p2_0p4=1.0000 loss_all=8.7812 init_gold_top10=0.1250 init_gold_top100=0.1250
+step=230 epoch=115/250 epoch_step=2/2 micro_steps=230 elapsed=4.4s lr=2.000000e-03 loss=0.8167 loss_recon=0.8167 loss_meanflow=0.0000 mean_model_t=0.3049 mean_corrupt_t=0.3049 mean_loss_t_weight=1.0000 linear_soft_target_mean_conf=0.3049 prior_center_loss_beta=0.0000 rollout_train_applied=0.0000 grad_enabled_before_rollout=1.0000 grad_enabled_after_rollout=1.0000 logits_requires_grad=1.0000 raw_loss_requires_grad=1.0000 acc_all=0.4375 corrupt_frac=1.0000 acc_corrupt=0.4375 loss_corrupt=1.4772 wrong_frac=0.6500 init_acc_corrupt=0.2875 acc_corrupt_t_0p4_0p6=0.7917 corrupt_frac_t_0p4_0p6=1.0000 out_w_norm=1.9604 out_g_norm=2.2768 acc_corrupt_t_0p2_0p4=0.4583 corrupt_frac_t_0p2_0p4=1.0000 acc_corrupt_t_0p0_0p2=0.1562 corrupt_frac_t_0p0_0p2=1.0000 loss_all=7.4824 init_gold_top10=0.2500 init_gold_top100=0.2500
+step=240 epoch=120/250 epoch_step=2/2 micro_steps=240 elapsed=4.1s lr=2.000000e-03 loss=0.8203 loss_recon=0.8203 loss_meanflow=0.0000 mean_model_t=0.2693 mean_corrupt_t=0.2693 mean_loss_t_weight=1.0000 linear_soft_target_mean_conf=0.2693 prior_center_loss_beta=0.0000 rollout_train_applied=0.0000 grad_enabled_before_rollout=1.0000 grad_enabled_after_rollout=1.0000 logits_requires_grad=1.0000 raw_loss_requires_grad=1.0000 acc_all=0.3375 corrupt_frac=1.0000 acc_corrupt=0.3375 loss_corrupt=1.1419 wrong_frac=0.7125 init_acc_corrupt=0.2000 acc_corrupt_t_0p0_0p2=0.1875 corrupt_frac_t_0p0_0p2=1.0000 out_w_norm=2.0042 out_g_norm=2.2456 acc_corrupt_t_0p4_0p6=0.6250 corrupt_frac_t_0p4_0p6=1.0000 acc_corrupt_t_0p2_0p4=0.3438 corrupt_frac_t_0p2_0p4=1.0000 loss_all=4.4014 init_gold_top10=0.2500 init_gold_top100=0.2500
+step=250 epoch=125/250 epoch_step=2/2 micro_steps=250 elapsed=4.5s lr=2.000000e-03 loss=0.5640 loss_recon=0.5640 loss_meanflow=0.0000 mean_model_t=0.1964 mean_corrupt_t=0.1964 mean_loss_t_weight=1.0000 linear_soft_target_mean_conf=0.1964 prior_center_loss_beta=0.0000 rollout_train_applied=0.0000 grad_enabled_before_rollout=1.0000 grad_enabled_after_rollout=1.0000 logits_requires_grad=1.0000 raw_loss_requires_grad=1.0000 acc_all=0.2875 corrupt_frac=1.0000 acc_corrupt=0.2875 loss_corrupt=0.8368 wrong_frac=0.7500 init_acc_corrupt=0.1250 acc_corrupt_t_0p0_0p2=0.1719 corrupt_frac_t_0p0_0p2=1.0000 out_w_norm=2.0422 out_g_norm=1.8608 acc_corrupt_t_0p6_0p8=0.8750 corrupt_frac_t_0p6_0p8=1.0000 acc_corrupt_t_0p2_0p4=0.6250 corrupt_frac_t_0p2_0p4=1.0000 loss_all=3.4316 init_gold_top10=0.5000 init_gold_top100=0.5000
+step=260 epoch=130/250 epoch_step=2/2 micro_steps=260 elapsed=4.4s lr=2.000000e-03 loss=0.6435 loss_recon=0.6435 loss_meanflow=0.0000 mean_model_t=0.1517 mean_corrupt_t=0.1517 mean_loss_t_weight=1.0000 linear_soft_target_mean_conf=0.1517 prior_center_loss_beta=0.0000 rollout_train_applied=0.0000 grad_enabled_before_rollout=1.0000 grad_enabled_after_rollout=1.0000 logits_requires_grad=1.0000 raw_loss_requires_grad=1.0000 acc_all=0.1500 corrupt_frac=1.0000 acc_corrupt=0.1500 loss_corrupt=1.1853 wrong_frac=0.8750 init_acc_corrupt=0.0250 acc_corrupt_t_0p0_0p2=0.1429 corrupt_frac_t_0p0_0p2=1.0000 out_w_norm=2.0589 out_g_norm=1.9505 acc_corrupt_t_0p2_0p4=0.1667 corrupt_frac_t_0p2_0p4=1.0000 loss_all=5.7266 init_gold_top10=0.1250 init_gold_top100=0.5000
+step=270 epoch=135/250 epoch_step=2/2 micro_steps=270 elapsed=4.0s lr=2.000000e-03 loss=0.6368 loss_recon=0.6368 loss_meanflow=0.0000 mean_model_t=0.1781 mean_corrupt_t=0.1781 mean_loss_t_weight=1.0000 linear_soft_target_mean_conf=0.1781 prior_center_loss_beta=0.0000 rollout_train_applied=0.0000 grad_enabled_before_rollout=1.0000 grad_enabled_after_rollout=1.0000 logits_requires_grad=1.0000 raw_loss_requires_grad=1.0000 acc_all=0.2500 corrupt_frac=1.0000 acc_corrupt=0.2500 loss_corrupt=1.0227 wrong_frac=0.7750 init_acc_corrupt=0.0875 acc_corrupt_t_0p0_0p2=0.1786 corrupt_frac_t_0p0_0p2=1.0000 out_w_norm=2.0719 out_g_norm=2.0863 acc_corrupt_t_0p2_0p4=0.4167 corrupt_frac_t_0p2_0p4=1.0000 loss_all=4.4688 init_gold_top10=0.3750 init_gold_top100=0.3750
+step=280 epoch=140/250 epoch_step=2/2 micro_steps=280 elapsed=4.5s lr=2.000000e-03 loss=0.8962 loss_recon=0.8962 loss_meanflow=0.0000 mean_model_t=0.2199 mean_corrupt_t=0.2199 mean_loss_t_weight=1.0000 linear_soft_target_mean_conf=0.2199 prior_center_loss_beta=0.0000 rollout_train_applied=0.0000 grad_enabled_before_rollout=1.0000 grad_enabled_after_rollout=1.0000 logits_requires_grad=1.0000 raw_loss_requires_grad=1.0000 acc_all=0.1750 corrupt_frac=1.0000 acc_corrupt=0.1750 loss_corrupt=1.4513 wrong_frac=0.8000 init_acc_corrupt=0.0625 acc_corrupt_t_0p4_0p6=0.0000 corrupt_frac_t_0p4_0p6=1.0000 out_w_norm=2.0855 out_g_norm=2.2083 acc_corrupt_t_0p0_0p2=0.1000 corrupt_frac_t_0p0_0p2=1.0000 acc_corrupt_t_0p2_0p4=0.3125 corrupt_frac_t_0p2_0p4=1.0000 loss_all=6.1973 init_gold_top10=0.1250 init_gold_top100=0.2500
+step=290 epoch=145/250 epoch_step=2/2 micro_steps=290 elapsed=4.4s lr=2.000000e-03 loss=0.6349 loss_recon=0.6349 loss_meanflow=0.0000 mean_model_t=0.2100 mean_corrupt_t=0.2100 mean_loss_t_weight=1.0000 linear_soft_target_mean_conf=0.2100 prior_center_loss_beta=0.0000 rollout_train_applied=0.0000 grad_enabled_before_rollout=1.0000 grad_enabled_after_rollout=1.0000 logits_requires_grad=1.0000 raw_loss_requires_grad=1.0000 acc_all=0.3000 corrupt_frac=1.0000 acc_corrupt=0.3000 loss_corrupt=1.0752 wrong_frac=0.7625 init_acc_corrupt=0.1625 acc_corrupt_t_0p2_0p4=0.3750 corrupt_frac_t_0p2_0p4=1.0000 out_w_norm=2.1006 out_g_norm=2.0542 acc_corrupt_t_0p0_0p2=0.2250 corrupt_frac_t_0p0_0p2=1.0000 loss_all=5.5759 init_gold_top10=0.2500 init_gold_top100=0.2500
+step=300 epoch=150/250 epoch_step=2/2 micro_steps=300 elapsed=4.1s lr=2.000000e-03 loss=0.6497 loss_recon=0.6497 loss_meanflow=0.0000 mean_model_t=0.2037 mean_corrupt_t=0.2037 mean_loss_t_weight=1.0000 linear_soft_target_mean_conf=0.2037 prior_center_loss_beta=0.0000 rollout_train_applied=0.0000 grad_enabled_before_rollout=1.0000 grad_enabled_after_rollout=1.0000 logits_requires_grad=1.0000 raw_loss_requires_grad=1.0000 acc_all=0.2625 corrupt_frac=1.0000 acc_corrupt=0.2625 loss_corrupt=1.3369 wrong_frac=0.7250 init_acc_corrupt=0.0875 acc_corrupt_t_0p0_0p2=0.1750 corrupt_frac_t_0p0_0p2=1.0000 out_w_norm=2.1204 out_g_norm=2.0059 acc_corrupt_t_0p2_0p4=0.3500 corrupt_frac_t_0p2_0p4=1.0000 loss_all=7.2656 init_gold_top10=0.1250 init_gold_top100=0.5000
+step=310 epoch=155/250 epoch_step=2/2 micro_steps=310 elapsed=4.5s lr=2.000000e-03 loss=0.5915 loss_recon=0.5915 loss_meanflow=0.0000 mean_model_t=0.1660 mean_corrupt_t=0.1660 mean_loss_t_weight=1.0000 linear_soft_target_mean_conf=0.1660 prior_center_loss_beta=0.0000 rollout_train_applied=0.0000 grad_enabled_before_rollout=1.0000 grad_enabled_after_rollout=1.0000 logits_requires_grad=1.0000 raw_loss_requires_grad=1.0000 acc_all=0.2125 corrupt_frac=1.0000 acc_corrupt=0.2125 loss_corrupt=1.2310 wrong_frac=0.8500 init_acc_corrupt=0.0875 acc_corrupt_t_0p0_0p2=0.1607 corrupt_frac_t_0p0_0p2=1.0000 out_w_norm=2.1404 out_g_norm=2.0428 acc_corrupt_t_0p4_0p6=0.8750 corrupt_frac_t_0p4_0p6=1.0000 acc_corrupt_t_0p2_0p4=0.0625 corrupt_frac_t_0p2_0p4=1.0000 loss_all=8.0625 init_gold_top10=0.0000 init_gold_top100=0.0000
+step=320 epoch=160/250 epoch_step=2/2 micro_steps=320 elapsed=4.4s lr=2.000000e-03 loss=0.7207 loss_recon=0.7207 loss_meanflow=0.0000 mean_model_t=0.2203 mean_corrupt_t=0.2203 mean_loss_t_weight=1.0000 linear_soft_target_mean_conf=0.2203 prior_center_loss_beta=0.0000 rollout_train_applied=0.0000 grad_enabled_before_rollout=1.0000 grad_enabled_after_rollout=1.0000 logits_requires_grad=1.0000 raw_loss_requires_grad=1.0000 acc_all=0.2250 corrupt_frac=1.0000 acc_corrupt=0.2250 loss_corrupt=1.3024 wrong_frac=0.8250 init_acc_corrupt=0.1125 acc_corrupt_t_0p0_0p2=0.1458 corrupt_frac_t_0p0_0p2=1.0000 out_w_norm=2.1608 out_g_norm=2.2383 acc_corrupt_t_0p4_0p6=0.2500 corrupt_frac_t_0p4_0p6=1.0000 acc_corrupt_t_0p2_0p4=0.3750 corrupt_frac_t_0p2_0p4=1.0000 loss_all=6.6562 init_gold_top10=0.0000 init_gold_top100=0.3750
+step=330 epoch=165/250 epoch_step=2/2 micro_steps=330 elapsed=4.0s lr=2.000000e-03 loss=0.6010 loss_recon=0.6010 loss_meanflow=0.0000 mean_model_t=0.2293 mean_corrupt_t=0.2293 mean_loss_t_weight=1.0000 linear_soft_target_mean_conf=0.2293 prior_center_loss_beta=0.0000 rollout_train_applied=0.0000 grad_enabled_before_rollout=1.0000 grad_enabled_after_rollout=1.0000 logits_requires_grad=1.0000 raw_loss_requires_grad=1.0000 acc_all=0.3000 corrupt_frac=1.0000 acc_corrupt=0.3000 loss_corrupt=0.7656 wrong_frac=0.6750 init_acc_corrupt=0.2125 acc_corrupt_t_0p0_0p2=0.0938 corrupt_frac_t_0p0_0p2=1.0000 out_w_norm=2.1767 out_g_norm=2.3064 acc_corrupt_t_0p2_0p4=0.4375 corrupt_frac_t_0p2_0p4=1.0000 loss_all=2.0283 init_gold_top10=0.6250 init_gold_top100=0.6250
+step=340 epoch=170/250 epoch_step=2/2 micro_steps=340 elapsed=4.4s lr=2.000000e-03 loss=0.5942 loss_recon=0.5942 loss_meanflow=0.0000 mean_model_t=0.2682 mean_corrupt_t=0.2682 mean_loss_t_weight=1.0000 linear_soft_target_mean_conf=0.2682 prior_center_loss_beta=0.0000 rollout_train_applied=0.0000 grad_enabled_before_rollout=1.0000 grad_enabled_after_rollout=1.0000 logits_requires_grad=1.0000 raw_loss_requires_grad=1.0000 acc_all=0.3375 corrupt_frac=1.0000 acc_corrupt=0.3375 loss_corrupt=1.0234 wrong_frac=0.6875 init_acc_corrupt=0.2375 acc_corrupt_t_0p0_0p2=0.1250 corrupt_frac_t_0p0_0p2=1.0000 out_w_norm=2.1904 out_g_norm=2.4344 acc_corrupt_t_0p2_0p4=0.4167 corrupt_frac_t_0p2_0p4=1.0000 acc_corrupt_t_0p4_0p6=0.5000 corrupt_frac_t_0p4_0p6=1.0000 acc_corrupt_t_0p6_0p8=1.0000 corrupt_frac_t_0p6_0p8=1.0000 loss_all=4.6797 init_gold_top10=0.1250 init_gold_top100=0.1250
+step=350 epoch=175/250 epoch_step=2/2 micro_steps=350 elapsed=4.4s lr=2.000000e-03 loss=0.6995 loss_recon=0.6995 loss_meanflow=0.0000 mean_model_t=0.1929 mean_corrupt_t=0.1929 mean_loss_t_weight=1.0000 linear_soft_target_mean_conf=0.1929 prior_center_loss_beta=0.0000 rollout_train_applied=0.0000 grad_enabled_before_rollout=1.0000 grad_enabled_after_rollout=1.0000 logits_requires_grad=1.0000 raw_loss_requires_grad=1.0000 acc_all=0.1750 corrupt_frac=1.0000 acc_corrupt=0.1750 loss_corrupt=1.1853 wrong_frac=0.8375 init_acc_corrupt=0.0750 acc_corrupt_t_0p2_0p4=0.2500 corrupt_frac_t_0p2_0p4=1.0000 out_w_norm=2.1949 out_g_norm=2.2119 acc_corrupt_t_0p0_0p2=0.1000 corrupt_frac_t_0p0_0p2=1.0000 loss_all=5.4766 init_gold_top10=0.1250 init_gold_top100=0.2500
+step=360 epoch=180/250 epoch_step=2/2 micro_steps=360 elapsed=4.1s lr=2.000000e-03 loss=0.6252 loss_recon=0.6252 loss_meanflow=0.0000 mean_model_t=0.1923 mean_corrupt_t=0.1923 mean_loss_t_weight=1.0000 linear_soft_target_mean_conf=0.1923 prior_center_loss_beta=0.0000 rollout_train_applied=0.0000 grad_enabled_before_rollout=1.0000 grad_enabled_after_rollout=1.0000 logits_requires_grad=1.0000 raw_loss_requires_grad=1.0000 acc_all=0.2000 corrupt_frac=1.0000 acc_corrupt=0.2000 loss_corrupt=0.9015 wrong_frac=0.7875 init_acc_corrupt=0.1000 acc_corrupt_t_0p0_0p2=0.1250 corrupt_frac_t_0p0_0p2=1.0000 out_w_norm=2.1944 out_g_norm=1.9623 acc_corrupt_t_0p4_0p6=0.5000 corrupt_frac_t_0p4_0p6=1.0000 loss_all=4.1462 init_gold_top10=0.2500 init_gold_top100=0.2500
+step=370 epoch=185/250 epoch_step=2/2 micro_steps=370 elapsed=4.5s lr=2.000000e-03 loss=0.4709 loss_recon=0.4709 loss_meanflow=0.0000 mean_model_t=0.1757 mean_corrupt_t=0.1757 mean_loss_t_weight=1.0000 linear_soft_target_mean_conf=0.1757 prior_center_loss_beta=0.0000 rollout_train_applied=0.0000 grad_enabled_before_rollout=1.0000 grad_enabled_after_rollout=1.0000 logits_requires_grad=1.0000 raw_loss_requires_grad=1.0000 acc_all=0.2750 corrupt_frac=1.0000 acc_corrupt=0.2750 loss_corrupt=0.8083 wrong_frac=0.8250 init_acc_corrupt=0.0750 acc_corrupt_t_0p2_0p4=0.3750 corrupt_frac_t_0p2_0p4=1.0000 out_w_norm=2.1991 out_g_norm=2.0457 acc_corrupt_t_0p0_0p2=0.2500 corrupt_frac_t_0p0_0p2=1.0000 loss_all=3.7109 init_gold_top10=0.2500 init_gold_top100=0.5000
+step=380 epoch=190/250 epoch_step=2/2 micro_steps=380 elapsed=4.4s lr=2.000000e-03 loss=0.4684 loss_recon=0.4684 loss_meanflow=0.0000 mean_model_t=0.1794 mean_corrupt_t=0.1794 mean_loss_t_weight=1.0000 linear_soft_target_mean_conf=0.1794 prior_center_loss_beta=0.0000 rollout_train_applied=0.0000 grad_enabled_before_rollout=1.0000 grad_enabled_after_rollout=1.0000 logits_requires_grad=1.0000 raw_loss_requires_grad=1.0000 acc_all=0.2375 corrupt_frac=1.0000 acc_corrupt=0.2375 loss_corrupt=0.7979 wrong_frac=0.8375 init_acc_corrupt=0.1000 acc_corrupt_t_0p0_0p2=0.1250 corrupt_frac_t_0p0_0p2=1.0000 out_w_norm=2.2059 out_g_norm=2.0713 acc_corrupt_t_0p4_0p6=0.5000 corrupt_frac_t_0p4_0p6=1.0000 acc_corrupt_t_0p2_0p4=0.5000 corrupt_frac_t_0p2_0p4=1.0000 loss_all=4.0498 init_gold_top10=0.2500 init_gold_top100=0.2500
+step=390 epoch=195/250 epoch_step=2/2 micro_steps=390 elapsed=4.1s lr=2.000000e-03 loss=0.6511 loss_recon=0.6511 loss_meanflow=0.0000 mean_model_t=0.1873 mean_corrupt_t=0.1873 mean_loss_t_weight=1.0000 linear_soft_target_mean_conf=0.1873 prior_center_loss_beta=0.0000 rollout_train_applied=0.0000 grad_enabled_before_rollout=1.0000 grad_enabled_after_rollout=1.0000 logits_requires_grad=1.0000 raw_loss_requires_grad=1.0000 acc_all=0.2500 corrupt_frac=1.0000 acc_corrupt=0.2500 loss_corrupt=1.1307 wrong_frac=0.8125 init_acc_corrupt=0.0750 acc_corrupt_t_0p2_0p4=0.3125 corrupt_frac_t_0p2_0p4=1.0000 out_w_norm=2.2128 out_g_norm=2.4076 acc_corrupt_t_0p0_0p2=0.2083 corrupt_frac_t_0p0_0p2=1.0000 loss_all=5.4785 init_gold_top10=0.1250 init_gold_top100=0.5000
+step=400 epoch=200/250 epoch_step=2/2 micro_steps=400 elapsed=4.5s lr=2.000000e-03 loss=0.6558 loss_recon=0.6558 loss_meanflow=0.0000 mean_model_t=0.3288 mean_corrupt_t=0.3288 mean_loss_t_weight=1.0000 linear_soft_target_mean_conf=0.3288 prior_center_loss_beta=0.0000 rollout_train_applied=0.0000 grad_enabled_before_rollout=1.0000 grad_enabled_after_rollout=1.0000 logits_requires_grad=1.0000 raw_loss_requires_grad=1.0000 acc_all=0.3125 corrupt_frac=1.0000 acc_corrupt=0.3125 loss_corrupt=0.7121 wrong_frac=0.7250 init_acc_corrupt=0.2250 acc_corrupt_t_0p0_0p2=0.0833 corrupt_frac_t_0p0_0p2=1.0000 out_w_norm=2.2082 out_g_norm=2.2056 acc_corrupt_t_0p2_0p4=0.2188 corrupt_frac_t_0p2_0p4=1.0000 acc_corrupt_t_0p6_0p8=0.6250 corrupt_frac_t_0p6_0p8=1.0000 acc_corrupt_t_0p4_0p6=0.6875 corrupt_frac_t_0p4_0p6=1.0000 loss_all=1.0971 init_gold_top10=0.7500 init_gold_top100=0.7500
+step=410 epoch=205/250 epoch_step=2/2 micro_steps=410 elapsed=5.9s lr=2.000000e-03 loss=0.6098 loss_recon=0.6098 loss_meanflow=0.0000 mean_model_t=0.2749 mean_corrupt_t=0.2749 mean_loss_t_weight=1.0000 linear_soft_target_mean_conf=0.2749 prior_center_loss_beta=0.0000 rollout_train_applied=0.0000 grad_enabled_before_rollout=1.0000 grad_enabled_after_rollout=1.0000 logits_requires_grad=1.0000 raw_loss_requires_grad=1.0000 acc_all=0.4000 corrupt_frac=1.0000 acc_corrupt=0.4000 loss_corrupt=0.8961 wrong_frac=0.6750 init_acc_corrupt=0.2375 acc_corrupt_t_0p0_0p2=0.1667 corrupt_frac_t_0p0_0p2=1.0000 out_w_norm=2.2129 out_g_norm=2.1538 acc_corrupt_t_0p2_0p4=0.4500 corrupt_frac_t_0p2_0p4=1.0000 acc_corrupt_t_0p4_0p6=0.6250 corrupt_frac_t_0p4_0p6=1.0000 loss_all=3.2842 init_gold_top10=0.3750 init_gold_top100=0.5000
+step=420 epoch=210/250 epoch_step=2/2 micro_steps=420 elapsed=5.5s lr=2.000000e-03 loss=0.6001 loss_recon=0.6001 loss_meanflow=0.0000 mean_model_t=0.2123 mean_corrupt_t=0.2123 mean_loss_t_weight=1.0000 linear_soft_target_mean_conf=0.2123 prior_center_loss_beta=0.0000 rollout_train_applied=0.0000 grad_enabled_before_rollout=1.0000 grad_enabled_after_rollout=1.0000 logits_requires_grad=1.0000 raw_loss_requires_grad=1.0000 acc_all=0.2625 corrupt_frac=1.0000 acc_corrupt=0.2625 loss_corrupt=1.0973 wrong_frac=0.7750 init_acc_corrupt=0.1000 acc_corrupt_t_0p2_0p4=0.3125 corrupt_frac_t_0p2_0p4=1.0000 out_w_norm=2.2078 out_g_norm=2.0003 acc_corrupt_t_0p0_0p2=0.1875 corrupt_frac_t_0p0_0p2=1.0000 loss_all=5.4766 init_gold_top10=0.1250 init_gold_top100=0.3750
+step=430 epoch=215/250 epoch_step=2/2 micro_steps=430 elapsed=6.0s lr=2.000000e-03 loss=0.5659 loss_recon=0.5659 loss_meanflow=0.0000 mean_model_t=0.1772 mean_corrupt_t=0.1772 mean_loss_t_weight=1.0000 linear_soft_target_mean_conf=0.1772 prior_center_loss_beta=0.0000 rollout_train_applied=0.0000 grad_enabled_before_rollout=1.0000 grad_enabled_after_rollout=1.0000 logits_requires_grad=1.0000 raw_loss_requires_grad=1.0000 acc_all=0.2750 corrupt_frac=1.0000 acc_corrupt=0.2750 loss_corrupt=0.9336 wrong_frac=0.8250 init_acc_corrupt=0.0875 acc_corrupt_t_0p0_0p2=0.2250 corrupt_frac_t_0p0_0p2=1.0000 out_w_norm=2.1960 out_g_norm=1.8503 acc_corrupt_t_0p2_0p4=0.3250 corrupt_frac_t_0p2_0p4=1.0000 loss_all=4.2871 init_gold_top10=0.1250 init_gold_top100=0.3750
+step=440 epoch=220/250 epoch_step=2/2 micro_steps=440 elapsed=5.4s lr=2.000000e-03 loss=0.5390 loss_recon=0.5390 loss_meanflow=0.0000 mean_model_t=0.2333 mean_corrupt_t=0.2333 mean_loss_t_weight=1.0000 linear_soft_target_mean_conf=0.2333 prior_center_loss_beta=0.0000 rollout_train_applied=0.0000 grad_enabled_before_rollout=1.0000 grad_enabled_after_rollout=1.0000 logits_requires_grad=1.0000 raw_loss_requires_grad=1.0000 acc_all=0.3625 corrupt_frac=1.0000 acc_corrupt=0.3625 loss_corrupt=0.9075 wrong_frac=0.6875 init_acc_corrupt=0.1875 acc_corrupt_t_0p0_0p2=0.1562 corrupt_frac_t_0p0_0p2=1.0000 out_w_norm=2.1950 out_g_norm=1.9730 acc_corrupt_t_0p2_0p4=0.5000 corrupt_frac_t_0p2_0p4=1.0000 loss_all=4.2441 init_gold_top10=0.3750 init_gold_top100=0.6250
+step=450 epoch=225/250 epoch_step=2/2 micro_steps=450 elapsed=4.3s lr=2.000000e-03 loss=0.5868 loss_recon=0.5868 loss_meanflow=0.0000 mean_model_t=0.1578 mean_corrupt_t=0.1578 mean_loss_t_weight=1.0000 linear_soft_target_mean_conf=0.1578 prior_center_loss_beta=0.0000 rollout_train_applied=0.0000 grad_enabled_before_rollout=1.0000 grad_enabled_after_rollout=1.0000 logits_requires_grad=1.0000 raw_loss_requires_grad=1.0000 acc_all=0.1500 corrupt_frac=1.0000 acc_corrupt=0.1500 loss_corrupt=1.2287 wrong_frac=0.8375 init_acc_corrupt=0.0500 acc_corrupt_t_0p0_0p2=0.1250 corrupt_frac_t_0p0_0p2=1.0000 out_w_norm=2.1938 out_g_norm=2.1070 acc_corrupt_t_0p2_0p4=0.2083 corrupt_frac_t_0p2_0p4=1.0000 loss_all=6.6562 init_gold_top10=0.0000 init_gold_top100=0.0000
+step=460 epoch=230/250 epoch_step=2/2 micro_steps=460 elapsed=5.0s lr=2.000000e-03 loss=0.4614 loss_recon=0.4614 loss_meanflow=0.0000 mean_model_t=0.1884 mean_corrupt_t=0.1884 mean_loss_t_weight=1.0000 linear_soft_target_mean_conf=0.1884 prior_center_loss_beta=0.0000 rollout_train_applied=0.0000 grad_enabled_before_rollout=1.0000 grad_enabled_after_rollout=1.0000 logits_requires_grad=1.0000 raw_loss_requires_grad=1.0000 acc_all=0.1875 corrupt_frac=1.0000 acc_corrupt=0.1875 loss_corrupt=0.9782 wrong_frac=0.8250 init_acc_corrupt=0.1000 acc_corrupt_t_0p0_0p2=0.0179 corrupt_frac_t_0p0_0p2=1.0000 out_w_norm=2.1963 out_g_norm=2.1456 acc_corrupt_t_0p4_0p6=0.5625 corrupt_frac_t_0p4_0p6=1.0000 acc_corrupt_t_0p2_0p4=0.6250 corrupt_frac_t_0p2_0p4=1.0000 loss_all=5.5859 init_gold_top10=0.0000 init_gold_top100=0.1250
+step=470 epoch=235/250 epoch_step=2/2 micro_steps=470 elapsed=5.8s lr=2.000000e-03 loss=0.5090 loss_recon=0.5090 loss_meanflow=0.0000 mean_model_t=0.1817 mean_corrupt_t=0.1817 mean_loss_t_weight=1.0000 linear_soft_target_mean_conf=0.1817 prior_center_loss_beta=0.0000 rollout_train_applied=0.0000 grad_enabled_before_rollout=1.0000 grad_enabled_after_rollout=1.0000 logits_requires_grad=1.0000 raw_loss_requires_grad=1.0000 acc_all=0.2375 corrupt_frac=1.0000 acc_corrupt=0.2375 loss_corrupt=1.1135 wrong_frac=0.8625 init_acc_corrupt=0.0750 acc_corrupt_t_0p0_0p2=0.1786 corrupt_frac_t_0p0_0p2=1.0000 out_w_norm=2.2051 out_g_norm=2.3825 acc_corrupt_t_0p2_0p4=0.3750 corrupt_frac_t_0p2_0p4=1.0000 acc_corrupt_t_0p4_0p6=0.3750 corrupt_frac_t_0p4_0p6=1.0000 loss_all=6.2793 init_gold_top10=0.0000 init_gold_top100=0.0000
+step=480 epoch=240/250 epoch_step=2/2 micro_steps=480 elapsed=5.6s lr=2.000000e-03 loss=0.5577 loss_recon=0.5577 loss_meanflow=0.0000 mean_model_t=0.2314 mean_corrupt_t=0.2314 mean_loss_t_weight=1.0000 linear_soft_target_mean_conf=0.2314 prior_center_loss_beta=0.0000 rollout_train_applied=0.0000 grad_enabled_before_rollout=1.0000 grad_enabled_after_rollout=1.0000 logits_requires_grad=1.0000 raw_loss_requires_grad=1.0000 acc_all=0.3125 corrupt_frac=1.0000 acc_corrupt=0.3125 loss_corrupt=0.8093 wrong_frac=0.6750 init_acc_corrupt=0.1625 acc_corrupt_t_0p0_0p2=0.1000 corrupt_frac_t_0p0_0p2=1.0000 out_w_norm=2.2094 out_g_norm=2.1309 acc_corrupt_t_0p2_0p4=0.4688 corrupt_frac_t_0p2_0p4=1.0000 acc_corrupt_t_0p4_0p6=0.7500 corrupt_frac_t_0p4_0p6=1.0000 loss_all=2.8984 init_gold_top10=0.3750 init_gold_top100=0.5000
+step=490 epoch=245/250 epoch_step=2/2 micro_steps=490 elapsed=6.4s lr=2.000000e-03 loss=0.5394 loss_recon=0.5394 loss_meanflow=0.0000 mean_model_t=0.2122 mean_corrupt_t=0.2122 mean_loss_t_weight=1.0000 linear_soft_target_mean_conf=0.2122 prior_center_loss_beta=0.0000 rollout_train_applied=0.0000 grad_enabled_before_rollout=1.0000 grad_enabled_after_rollout=1.0000 logits_requires_grad=1.0000 raw_loss_requires_grad=1.0000 acc_all=0.2750 corrupt_frac=1.0000 acc_corrupt=0.2750 loss_corrupt=0.8618 wrong_frac=0.7250 init_acc_corrupt=0.1375 acc_corrupt_t_0p0_0p2=0.1875 corrupt_frac_t_0p0_0p2=1.0000 out_w_norm=2.2091 out_g_norm=2.0059 acc_corrupt_t_0p2_0p4=0.2250 corrupt_frac_t_0p2_0p4=1.0000 acc_corrupt_t_0p4_0p6=0.8750 corrupt_frac_t_0p4_0p6=1.0000 loss_all=3.6338 init_gold_top10=0.2500 init_gold_top100=0.2500
+step=500 epoch=250/250 epoch_step=2/2 micro_steps=500 elapsed=4.4s lr=2.000000e-03 loss=0.4201 loss_recon=0.4201 loss_meanflow=0.0000 mean_model_t=0.1335 mean_corrupt_t=0.1335 mean_loss_t_weight=1.0000 linear_soft_target_mean_conf=0.1335 prior_center_loss_beta=0.0000 rollout_train_applied=0.0000 grad_enabled_before_rollout=1.0000 grad_enabled_after_rollout=1.0000 logits_requires_grad=1.0000 raw_loss_requires_grad=1.0000 acc_all=0.1875 corrupt_frac=1.0000 acc_corrupt=0.1875 loss_corrupt=0.7494 wrong_frac=0.8875 init_acc_corrupt=0.0500 acc_corrupt_t_0p2_0p4=0.1875 corrupt_frac_t_0p2_0p4=1.0000 out_w_norm=2.2072 out_g_norm=1.6554 acc_corrupt_t_0p0_0p2=0.1875 corrupt_frac_t_0p0_0p2=1.0000 loss_all=3.9180 init_gold_top10=0.2500 init_gold_top100=0.2500
+[allcorrupt] done train8_n8_allcorrupt_linear_soft_kl_20260517_train8ctx8_allcorrupt Sun May 17 00:30:22 UTC 2026

LTA_openwebtext_dualt/logs/watch_lta_owt_classic_fullvocab_bert_c1024_len1024_lr2e4_gbs2048_2node8gpu_1m_save10k_t-20260522043432-f7vrv_latest1k_gpu3_b4.nohup.log

@@ -0,0 +1,291 @@
+[watch-owt-len1024-lr2e4] run_glob=runs/lta_owt_classic_fullvocab_bert_c1024_len1024_lr2e4_gbs2048_2node8gpu_1m_save10k_*
+[watch-owt-len1024-lr2e4] explicit_run_dir=runs/lta_owt_classic_fullvocab_bert_c1024_len1024_lr2e4_gbs2048_2node8gpu_1m_save10k_t-20260522043432-f7vrv
+[watch-owt-len1024-lr2e4] out_root=docs/lta_samples/metrics_20260521/owt_classic_fullvocab_len1024_lr2e4_gbs2048_latest_every1k_normal_steps_state_t1p45_c1024_n1024
+[watch-owt-len1024-lr2e4] decode=normal_steps_sweep steps=128 cmax=1024 temp=1.45 final_from=state n=1024 max_len=1024
+[watch-owt-len1024-lr2e4] source=latest.pt snapshot_each=1000 decode_batch=4 score_batch=4
+[watch-owt-len1024-lr2e4] 2026-05-21_22:00:41 snapshot latest step_0004000 -> runs/lta_owt_classic_fullvocab_bert_c1024_len1024_lr2e4_gbs2048_2node8gpu_1m_save10k_t-20260522043432-f7vrv/latest_snapshots_1k/step_0004000.pt
+[watch-owt-len1024-lr2e4] 2026-05-21_22:00:44 infer runs/lta_owt_classic_fullvocab_bert_c1024_len1024_lr2e4_gbs2048_2node8gpu_1m_save10k_t-20260522043432-f7vrv/latest_snapshots_1k/step_0004000.pt -> docs/lta_samples/metrics_20260521/owt_classic_fullvocab_len1024_lr2e4_gbs2048_latest_every1k_normal_steps_state_t1p45_c1024_n1024/lta_owt_classic_fullvocab_bert_c1024_len1024_lr2e4_gbs2048_2node8gpu_1m_save10k_t-20260522043432-f7vrv/step_0004000
+[ckpt] runs/lta_owt_classic_fullvocab_bert_c1024_len1024_lr2e4_gbs2048_2node8gpu_1m_save10k_t-20260522043432-f7vrv/latest_snapshots_1k/step_0004000.pt step=4000
+[decode] steps128_c1024_t1p45 generated 4/1024
+[decode] steps128_c1024_t1p45 generated 8/1024
+[decode] steps128_c1024_t1p45 generated 12/1024
+[decode] steps128_c1024_t1p45 generated 16/1024
+[decode] steps128_c1024_t1p45 generated 20/1024
+[decode] steps128_c1024_t1p45 generated 24/1024
+[decode] steps128_c1024_t1p45 generated 28/1024
+[decode] steps128_c1024_t1p45 generated 32/1024
+[decode] steps128_c1024_t1p45 generated 36/1024
+[decode] steps128_c1024_t1p45 generated 40/1024
+[decode] steps128_c1024_t1p45 generated 44/1024
+[decode] steps128_c1024_t1p45 generated 48/1024
+[decode] steps128_c1024_t1p45 generated 52/1024
+[decode] steps128_c1024_t1p45 generated 56/1024
+[decode] steps128_c1024_t1p45 generated 60/1024
+[decode] steps128_c1024_t1p45 generated 64/1024
+[decode] steps128_c1024_t1p45 generated 68/1024
+[decode] steps128_c1024_t1p45 generated 72/1024
+[decode] steps128_c1024_t1p45 generated 76/1024
+[decode] steps128_c1024_t1p45 generated 80/1024
+[decode] steps128_c1024_t1p45 generated 84/1024
+[decode] steps128_c1024_t1p45 generated 88/1024
+[decode] steps128_c1024_t1p45 generated 92/1024
+[decode] steps128_c1024_t1p45 generated 96/1024
+[decode] steps128_c1024_t1p45 generated 100/1024
+[decode] steps128_c1024_t1p45 generated 104/1024
+[decode] steps128_c1024_t1p45 generated 108/1024
+[decode] steps128_c1024_t1p45 generated 112/1024
+[decode] steps128_c1024_t1p45 generated 116/1024
+[decode] steps128_c1024_t1p45 generated 120/1024
+[decode] steps128_c1024_t1p45 generated 124/1024
+[decode] steps128_c1024_t1p45 generated 128/1024
+[decode] steps128_c1024_t1p45 generated 132/1024
+[decode] steps128_c1024_t1p45 generated 136/1024
+[decode] steps128_c1024_t1p45 generated 140/1024
+[decode] steps128_c1024_t1p45 generated 144/1024
+[decode] steps128_c1024_t1p45 generated 148/1024
+[decode] steps128_c1024_t1p45 generated 152/1024
+[decode] steps128_c1024_t1p45 generated 156/1024
+[decode] steps128_c1024_t1p45 generated 160/1024
+[decode] steps128_c1024_t1p45 generated 164/1024
+[decode] steps128_c1024_t1p45 generated 168/1024
+[decode] steps128_c1024_t1p45 generated 172/1024
+[decode] steps128_c1024_t1p45 generated 176/1024
+[decode] steps128_c1024_t1p45 generated 180/1024
+[decode] steps128_c1024_t1p45 generated 184/1024
+[decode] steps128_c1024_t1p45 generated 188/1024
+[decode] steps128_c1024_t1p45 generated 192/1024
+[decode] steps128_c1024_t1p45 generated 196/1024
+[decode] steps128_c1024_t1p45 generated 200/1024
+[decode] steps128_c1024_t1p45 generated 204/1024
+[decode] steps128_c1024_t1p45 generated 208/1024
+[decode] steps128_c1024_t1p45 generated 212/1024
+[decode] steps128_c1024_t1p45 generated 216/1024
+[decode] steps128_c1024_t1p45 generated 220/1024
+[decode] steps128_c1024_t1p45 generated 224/1024
+[decode] steps128_c1024_t1p45 generated 228/1024
+[decode] steps128_c1024_t1p45 generated 232/1024
+[decode] steps128_c1024_t1p45 generated 236/1024
+[decode] steps128_c1024_t1p45 generated 240/1024
+[decode] steps128_c1024_t1p45 generated 244/1024
+[decode] steps128_c1024_t1p45 generated 248/1024
+[decode] steps128_c1024_t1p45 generated 252/1024
+[decode] steps128_c1024_t1p45 generated 256/1024
+[decode] steps128_c1024_t1p45 generated 260/1024
+[decode] steps128_c1024_t1p45 generated 264/1024
+[decode] steps128_c1024_t1p45 generated 268/1024
+[decode] steps128_c1024_t1p45 generated 272/1024
+[decode] steps128_c1024_t1p45 generated 276/1024
+[decode] steps128_c1024_t1p45 generated 280/1024
+[decode] steps128_c1024_t1p45 generated 284/1024
+[decode] steps128_c1024_t1p45 generated 288/1024
+[decode] steps128_c1024_t1p45 generated 292/1024
+[decode] steps128_c1024_t1p45 generated 296/1024
+[decode] steps128_c1024_t1p45 generated 300/1024
+[decode] steps128_c1024_t1p45 generated 304/1024
+[decode] steps128_c1024_t1p45 generated 308/1024
+[decode] steps128_c1024_t1p45 generated 312/1024
+[decode] steps128_c1024_t1p45 generated 316/1024
+[decode] steps128_c1024_t1p45 generated 320/1024
+[decode] steps128_c1024_t1p45 generated 324/1024
+[decode] steps128_c1024_t1p45 generated 328/1024
+[decode] steps128_c1024_t1p45 generated 332/1024
+[decode] steps128_c1024_t1p45 generated 336/1024
+[decode] steps128_c1024_t1p45 generated 340/1024
+[decode] steps128_c1024_t1p45 generated 344/1024
+[decode] steps128_c1024_t1p45 generated 348/1024
+[decode] steps128_c1024_t1p45 generated 352/1024
+[decode] steps128_c1024_t1p45 generated 356/1024
+[decode] steps128_c1024_t1p45 generated 360/1024
+[decode] steps128_c1024_t1p45 generated 364/1024
+[decode] steps128_c1024_t1p45 generated 368/1024
+[decode] steps128_c1024_t1p45 generated 372/1024
+[decode] steps128_c1024_t1p45 generated 376/1024
+[decode] steps128_c1024_t1p45 generated 380/1024
+[decode] steps128_c1024_t1p45 generated 384/1024
+[decode] steps128_c1024_t1p45 generated 388/1024
+[decode] steps128_c1024_t1p45 generated 392/1024
+[decode] steps128_c1024_t1p45 generated 396/1024
+[decode] steps128_c1024_t1p45 generated 400/1024
+[decode] steps128_c1024_t1p45 generated 404/1024
+[decode] steps128_c1024_t1p45 generated 408/1024
+[decode] steps128_c1024_t1p45 generated 412/1024
+[decode] steps128_c1024_t1p45 generated 416/1024
+[decode] steps128_c1024_t1p45 generated 420/1024
+[decode] steps128_c1024_t1p45 generated 424/1024
+[decode] steps128_c1024_t1p45 generated 428/1024
+[decode] steps128_c1024_t1p45 generated 432/1024
+[decode] steps128_c1024_t1p45 generated 436/1024
+[decode] steps128_c1024_t1p45 generated 440/1024
+[decode] steps128_c1024_t1p45 generated 444/1024
+[decode] steps128_c1024_t1p45 generated 448/1024
+[decode] steps128_c1024_t1p45 generated 452/1024
+[decode] steps128_c1024_t1p45 generated 456/1024
+[decode] steps128_c1024_t1p45 generated 460/1024
+[decode] steps128_c1024_t1p45 generated 464/1024
+[decode] steps128_c1024_t1p45 generated 468/1024
+[decode] steps128_c1024_t1p45 generated 472/1024
+[decode] steps128_c1024_t1p45 generated 476/1024
+[decode] steps128_c1024_t1p45 generated 480/1024
+[decode] steps128_c1024_t1p45 generated 484/1024
+[decode] steps128_c1024_t1p45 generated 488/1024
+[decode] steps128_c1024_t1p45 generated 492/1024
+[decode] steps128_c1024_t1p45 generated 496/1024
+[decode] steps128_c1024_t1p45 generated 500/1024
+[decode] steps128_c1024_t1p45 generated 504/1024
+[decode] steps128_c1024_t1p45 generated 508/1024
+[decode] steps128_c1024_t1p45 generated 512/1024
+[decode] steps128_c1024_t1p45 generated 516/1024
+[decode] steps128_c1024_t1p45 generated 520/1024
+[decode] steps128_c1024_t1p45 generated 524/1024
+[decode] steps128_c1024_t1p45 generated 528/1024
+[decode] steps128_c1024_t1p45 generated 532/1024
+[decode] steps128_c1024_t1p45 generated 536/1024
+[decode] steps128_c1024_t1p45 generated 540/1024
+[decode] steps128_c1024_t1p45 generated 544/1024
+[decode] steps128_c1024_t1p45 generated 548/1024
+[decode] steps128_c1024_t1p45 generated 552/1024
+[decode] steps128_c1024_t1p45 generated 556/1024
+[decode] steps128_c1024_t1p45 generated 560/1024
+[decode] steps128_c1024_t1p45 generated 564/1024
+[decode] steps128_c1024_t1p45 generated 568/1024
+[decode] steps128_c1024_t1p45 generated 572/1024
+[decode] steps128_c1024_t1p45 generated 576/1024
+[decode] steps128_c1024_t1p45 generated 580/1024
+[decode] steps128_c1024_t1p45 generated 584/1024
+[decode] steps128_c1024_t1p45 generated 588/1024
+[decode] steps128_c1024_t1p45 generated 592/1024
+[decode] steps128_c1024_t1p45 generated 596/1024
+[decode] steps128_c1024_t1p45 generated 600/1024
+[decode] steps128_c1024_t1p45 generated 604/1024
+[decode] steps128_c1024_t1p45 generated 608/1024
+[decode] steps128_c1024_t1p45 generated 612/1024
+[decode] steps128_c1024_t1p45 generated 616/1024
+[decode] steps128_c1024_t1p45 generated 620/1024
+[decode] steps128_c1024_t1p45 generated 624/1024
+[decode] steps128_c1024_t1p45 generated 628/1024
+[decode] steps128_c1024_t1p45 generated 632/1024
+[decode] steps128_c1024_t1p45 generated 636/1024
+[decode] steps128_c1024_t1p45 generated 640/1024
+[decode] steps128_c1024_t1p45 generated 644/1024
+[decode] steps128_c1024_t1p45 generated 648/1024
+[decode] steps128_c1024_t1p45 generated 652/1024
+[decode] steps128_c1024_t1p45 generated 656/1024
+[decode] steps128_c1024_t1p45 generated 660/1024
+[decode] steps128_c1024_t1p45 generated 664/1024
+[decode] steps128_c1024_t1p45 generated 668/1024
+[decode] steps128_c1024_t1p45 generated 672/1024
+[decode] steps128_c1024_t1p45 generated 676/1024
+[decode] steps128_c1024_t1p45 generated 680/1024
+[decode] steps128_c1024_t1p45 generated 684/1024
+[decode] steps128_c1024_t1p45 generated 688/1024
+[decode] steps128_c1024_t1p45 generated 692/1024
+[decode] steps128_c1024_t1p45 generated 696/1024
+[decode] steps128_c1024_t1p45 generated 700/1024
+[decode] steps128_c1024_t1p45 generated 704/1024
+[decode] steps128_c1024_t1p45 generated 708/1024
+[decode] steps128_c1024_t1p45 generated 712/1024
+[decode] steps128_c1024_t1p45 generated 716/1024
+[decode] steps128_c1024_t1p45 generated 720/1024
+[decode] steps128_c1024_t1p45 generated 724/1024
+[decode] steps128_c1024_t1p45 generated 728/1024
+[decode] steps128_c1024_t1p45 generated 732/1024
+[decode] steps128_c1024_t1p45 generated 736/1024
+[decode] steps128_c1024_t1p45 generated 740/1024
+[decode] steps128_c1024_t1p45 generated 744/1024
+[decode] steps128_c1024_t1p45 generated 748/1024
+[decode] steps128_c1024_t1p45 generated 752/1024
+[decode] steps128_c1024_t1p45 generated 756/1024
+[decode] steps128_c1024_t1p45 generated 760/1024
+[decode] steps128_c1024_t1p45 generated 764/1024
+[decode] steps128_c1024_t1p45 generated 768/1024
+[decode] steps128_c1024_t1p45 generated 772/1024
+[decode] steps128_c1024_t1p45 generated 776/1024
+[decode] steps128_c1024_t1p45 generated 780/1024
+[decode] steps128_c1024_t1p45 generated 784/1024
+[decode] steps128_c1024_t1p45 generated 788/1024
+[decode] steps128_c1024_t1p45 generated 792/1024
+[decode] steps128_c1024_t1p45 generated 796/1024
+[decode] steps128_c1024_t1p45 generated 800/1024
+[decode] steps128_c1024_t1p45 generated 804/1024
+[decode] steps128_c1024_t1p45 generated 808/1024
+[decode] steps128_c1024_t1p45 generated 812/1024
+[decode] steps128_c1024_t1p45 generated 816/1024
+[decode] steps128_c1024_t1p45 generated 820/1024
+[decode] steps128_c1024_t1p45 generated 824/1024
+[decode] steps128_c1024_t1p45 generated 828/1024
+[decode] steps128_c1024_t1p45 generated 832/1024
+[decode] steps128_c1024_t1p45 generated 836/1024
+[decode] steps128_c1024_t1p45 generated 840/1024
+[decode] steps128_c1024_t1p45 generated 844/1024
+[decode] steps128_c1024_t1p45 generated 848/1024
+[decode] steps128_c1024_t1p45 generated 852/1024
+[decode] steps128_c1024_t1p45 generated 856/1024
+[decode] steps128_c1024_t1p45 generated 860/1024
+[decode] steps128_c1024_t1p45 generated 864/1024
+[decode] steps128_c1024_t1p45 generated 868/1024
+[decode] steps128_c1024_t1p45 generated 872/1024
+[decode] steps128_c1024_t1p45 generated 876/1024
+[decode] steps128_c1024_t1p45 generated 880/1024
+[decode] steps128_c1024_t1p45 generated 884/1024
+[decode] steps128_c1024_t1p45 generated 888/1024
+[decode] steps128_c1024_t1p45 generated 892/1024
+[decode] steps128_c1024_t1p45 generated 896/1024
+[decode] steps128_c1024_t1p45 generated 900/1024
+[decode] steps128_c1024_t1p45 generated 904/1024
+[decode] steps128_c1024_t1p45 generated 908/1024
+[decode] steps128_c1024_t1p45 generated 912/1024
+[decode] steps128_c1024_t1p45 generated 916/1024
+[decode] steps128_c1024_t1p45 generated 920/1024
+[decode] steps128_c1024_t1p45 generated 924/1024
+[decode] steps128_c1024_t1p45 generated 928/1024
+[decode] steps128_c1024_t1p45 generated 932/1024
+[decode] steps128_c1024_t1p45 generated 936/1024
+[decode] steps128_c1024_t1p45 generated 940/1024
+[decode] steps128_c1024_t1p45 generated 944/1024
+[decode] steps128_c1024_t1p45 generated 948/1024
+[decode] steps128_c1024_t1p45 generated 952/1024
+[decode] steps128_c1024_t1p45 generated 956/1024
+[decode] steps128_c1024_t1p45 generated 960/1024
+[decode] steps128_c1024_t1p45 generated 964/1024
+[decode] steps128_c1024_t1p45 generated 968/1024
+[decode] steps128_c1024_t1p45 generated 972/1024
+[decode] steps128_c1024_t1p45 generated 976/1024
+[decode] steps128_c1024_t1p45 generated 980/1024
+[decode] steps128_c1024_t1p45 generated 984/1024
+[decode] steps128_c1024_t1p45 generated 988/1024
+[decode] steps128_c1024_t1p45 generated 992/1024
+[decode] steps128_c1024_t1p45 generated 996/1024
+[decode] steps128_c1024_t1p45 generated 1000/1024
+[decode] steps128_c1024_t1p45 generated 1004/1024
+[decode] steps128_c1024_t1p45 generated 1008/1024
+[decode] steps128_c1024_t1p45 generated 1012/1024
+[decode] steps128_c1024_t1p45 generated 1016/1024
+[decode] steps128_c1024_t1p45 generated 1020/1024
+[decode] steps128_c1024_t1p45 generated 1024/1024
+[summary] {"name": "steps128_c1024_t1p45", "step": 4000, "decode_steps": 128, "concentration_max": 1024.0, "raw_genppl": 1.0442802434637282, "stripped_genppl": 1.0442802434637282, "sample_entropy": 0.0, "distinct_1": 1.33514404296875e-05, "distinct_2": 1.3364491691104594e-05, "top_token_mass": 0.6484375, "raw_kept": 1024, "stripped_kept": 1024}
+[watch-owt-len1024-lr2e4] 2026-05-21_22:46:31 done step_0004000
+[watch-owt-len1024-lr2e4] 2026-05-21_22:47:54 snapshot latest step_0006000 -> runs/lta_owt_classic_fullvocab_bert_c1024_len1024_lr2e4_gbs2048_2node8gpu_1m_save10k_t-20260522043432-f7vrv/latest_snapshots_1k/step_0006000.pt
+[watch-owt-len1024-lr2e4] 2026-05-21_22:47:57 infer runs/lta_owt_classic_fullvocab_bert_c1024_len1024_lr2e4_gbs2048_2node8gpu_1m_save10k_t-20260522043432-f7vrv/latest_snapshots_1k/step_0006000.pt -> docs/lta_samples/metrics_20260521/owt_classic_fullvocab_len1024_lr2e4_gbs2048_latest_every1k_normal_steps_state_t1p45_c1024_n1024/lta_owt_classic_fullvocab_bert_c1024_len1024_lr2e4_gbs2048_2node8gpu_1m_save10k_t-20260522043432-f7vrv/step_0006000
+[ckpt] runs/lta_owt_classic_fullvocab_bert_c1024_len1024_lr2e4_gbs2048_2node8gpu_1m_save10k_t-20260522043432-f7vrv/latest_snapshots_1k/step_0006000.pt step=6000
+[decode] steps128_c1024_t1p45 generated 4/1024
+[decode] steps128_c1024_t1p45 generated 8/1024
+[decode] steps128_c1024_t1p45 generated 12/1024
+[decode] steps128_c1024_t1p45 generated 16/1024
+[decode] steps128_c1024_t1p45 generated 20/1024
+[decode] steps128_c1024_t1p45 generated 24/1024
+[decode] steps128_c1024_t1p45 generated 28/1024
+[decode] steps128_c1024_t1p45 generated 32/1024
+[decode] steps128_c1024_t1p45 generated 36/1024
+[decode] steps128_c1024_t1p45 generated 40/1024
+[decode] steps128_c1024_t1p45 generated 44/1024
+[decode] steps128_c1024_t1p45 generated 48/1024
+[decode] steps128_c1024_t1p45 generated 52/1024
+[decode] steps128_c1024_t1p45 generated 56/1024
+[decode] steps128_c1024_t1p45 generated 60/1024
+[decode] steps128_c1024_t1p45 generated 64/1024
+[decode] steps128_c1024_t1p45 generated 68/1024
+[decode] steps128_c1024_t1p45 generated 72/1024
+[decode] steps128_c1024_t1p45 generated 76/1024
+[decode] steps128_c1024_t1p45 generated 80/1024
+[decode] steps128_c1024_t1p45 generated 84/1024
+[decode] steps128_c1024_t1p45 generated 88/1024

LTA_openwebtext_dualt/mini_owt_logdirichlet/.venv_qwen35_uv/lib/python3.12/site-packages/numpy/matrixlib/__init__.pyi

@@ -0,0 +1,15 @@
+from numpy._pytesttester import PytestTester
+
+from numpy import (
+    matrix as matrix,
+)
+
+from numpy.matrixlib.defmatrix import (
+    bmat as bmat,
+    mat as mat,
+    asmatrix as asmatrix,
+)
+
+__all__: list[str]
+__path__: list[str]
+test: PytestTester

LTA_openwebtext_dualt/mini_owt_logdirichlet/.venv_qwen35_uv/lib/python3.12/site-packages/numpy/matrixlib/defmatrix.py

@@ -0,0 +1,1114 @@
+__all__ = ['matrix', 'bmat', 'mat', 'asmatrix']
+
+import sys
+import warnings
+import ast
+
+from .._utils import set_module
+import numpy.core.numeric as N
+from numpy.core.numeric import concatenate, isscalar
+# While not in __all__, matrix_power used to be defined here, so we import
+# it for backward compatibility.
+from numpy.linalg import matrix_power
+
+
+def _convert_from_string(data):
+    for char in '[]':
+        data = data.replace(char, '')
+
+    rows = data.split(';')
+    newdata = []
+    count = 0
+    for row in rows:
+        trow = row.split(',')
+        newrow = []
+        for col in trow:
+            temp = col.split()
+            newrow.extend(map(ast.literal_eval, temp))
+        if count == 0:
+            Ncols = len(newrow)
+        elif len(newrow) != Ncols:
+            raise ValueError("Rows not the same size.")
+        count += 1
+        newdata.append(newrow)
+    return newdata
+
+
+@set_module('numpy')
+def asmatrix(data, dtype=None):
+    """
+    Interpret the input as a matrix.
+
+    Unlike `matrix`, `asmatrix` does not make a copy if the input is already
+    a matrix or an ndarray.  Equivalent to ``matrix(data, copy=False)``.
+
+    Parameters
+    ----------
+    data : array_like
+        Input data.
+    dtype : data-type
+       Data-type of the output matrix.
+
+    Returns
+    -------
+    mat : matrix
+        `data` interpreted as a matrix.
+
+    Examples
+    --------
+    >>> x = np.array([[1, 2], [3, 4]])
+
+    >>> m = np.asmatrix(x)
+
+    >>> x[0,0] = 5
+
+    >>> m
+    matrix([[5, 2],
+            [3, 4]])
+
+    """
+    return matrix(data, dtype=dtype, copy=False)
+
+
+@set_module('numpy')
+class matrix(N.ndarray):
+    """
+    matrix(data, dtype=None, copy=True)
+
+    .. note:: It is no longer recommended to use this class, even for linear
+              algebra. Instead use regular arrays. The class may be removed
+              in the future.
+
+    Returns a matrix from an array-like object, or from a string of data.
+    A matrix is a specialized 2-D array that retains its 2-D nature
+    through operations.  It has certain special operators, such as ``*``
+    (matrix multiplication) and ``**`` (matrix power).
+
+    Parameters
+    ----------
+    data : array_like or string
+       If `data` is a string, it is interpreted as a matrix with commas
+       or spaces separating columns, and semicolons separating rows.
+    dtype : data-type
+       Data-type of the output matrix.
+    copy : bool
+       If `data` is already an `ndarray`, then this flag determines
+       whether the data is copied (the default), or whether a view is
+       constructed.
+
+    See Also
+    --------
+    array
+
+    Examples
+    --------
+    >>> a = np.matrix('1 2; 3 4')
+    >>> a
+    matrix([[1, 2],
+            [3, 4]])
+
+    >>> np.matrix([[1, 2], [3, 4]])
+    matrix([[1, 2],
+            [3, 4]])
+
+    """
+    __array_priority__ = 10.0
+    def __new__(subtype, data, dtype=None, copy=True):
+        warnings.warn('the matrix subclass is not the recommended way to '
+                      'represent matrices or deal with linear algebra (see '
+                      'https://docs.scipy.org/doc/numpy/user/'
+                      'numpy-for-matlab-users.html). '
+                      'Please adjust your code to use regular ndarray.',
+                      PendingDeprecationWarning, stacklevel=2)
+        if isinstance(data, matrix):
+            dtype2 = data.dtype
+            if (dtype is None):
+                dtype = dtype2
+            if (dtype2 == dtype) and (not copy):
+                return data
+            return data.astype(dtype)
+
+        if isinstance(data, N.ndarray):
+            if dtype is None:
+                intype = data.dtype
+            else:
+                intype = N.dtype(dtype)
+            new = data.view(subtype)
+            if intype != data.dtype:
+                return new.astype(intype)
+            if copy: return new.copy()
+            else: return new
+
+        if isinstance(data, str):
+            data = _convert_from_string(data)
+
+        # now convert data to an array
+        arr = N.array(data, dtype=dtype, copy=copy)
+        ndim = arr.ndim
+        shape = arr.shape
+        if (ndim > 2):
+            raise ValueError("matrix must be 2-dimensional")
+        elif ndim == 0:
+            shape = (1, 1)
+        elif ndim == 1:
+            shape = (1, shape[0])
+
+        order = 'C'
+        if (ndim == 2) and arr.flags.fortran:
+            order = 'F'
+
+        if not (order or arr.flags.contiguous):
+            arr = arr.copy()
+
+        ret = N.ndarray.__new__(subtype, shape, arr.dtype,
+                                buffer=arr,
+                                order=order)
+        return ret
+
+    def __array_finalize__(self, obj):
+        self._getitem = False
+        if (isinstance(obj, matrix) and obj._getitem): return
+        ndim = self.ndim
+        if (ndim == 2):
+            return
+        if (ndim > 2):
+            newshape = tuple([x for x in self.shape if x > 1])
+            ndim = len(newshape)
+            if ndim == 2:
+                self.shape = newshape
+                return
+            elif (ndim > 2):
+                raise ValueError("shape too large to be a matrix.")
+        else:
+            newshape = self.shape
+        if ndim == 0:
+            self.shape = (1, 1)
+        elif ndim == 1:
+            self.shape = (1, newshape[0])
+        return
+
+    def __getitem__(self, index):
+        self._getitem = True
+
+        try:
+            out = N.ndarray.__getitem__(self, index)
+        finally:
+            self._getitem = False
+
+        if not isinstance(out, N.ndarray):
+            return out
+
+        if out.ndim == 0:
+            return out[()]
+        if out.ndim == 1:
+            sh = out.shape[0]
+            # Determine when we should have a column array
+            try:
+                n = len(index)
+            except Exception:
+                n = 0
+            if n > 1 and isscalar(index[1]):
+                out.shape = (sh, 1)
+            else:
+                out.shape = (1, sh)
+        return out
+
+    def __mul__(self, other):
+        if isinstance(other, (N.ndarray, list, tuple)) :
+            # This promotes 1-D vectors to row vectors
+            return N.dot(self, asmatrix(other))
+        if isscalar(other) or not hasattr(other, '__rmul__') :
+            return N.dot(self, other)
+        return NotImplemented
+
+    def __rmul__(self, other):
+        return N.dot(other, self)
+
+    def __imul__(self, other):
+        self[:] = self * other
+        return self
+
+    def __pow__(self, other):
+        return matrix_power(self, other)
+
+    def __ipow__(self, other):
+        self[:] = self ** other
+        return self
+
+    def __rpow__(self, other):
+        return NotImplemented
+
+    def _align(self, axis):
+        """A convenience function for operations that need to preserve axis
+        orientation.
+        """
+        if axis is None:
+            return self[0, 0]
+        elif axis==0:
+            return self
+        elif axis==1:
+            return self.transpose()
+        else:
+            raise ValueError("unsupported axis")
+
+    def _collapse(self, axis):
+        """A convenience function for operations that want to collapse
+        to a scalar like _align, but are using keepdims=True
+        """
+        if axis is None:
+            return self[0, 0]
+        else:
+            return self
+
+    # Necessary because base-class tolist expects dimension
+    #  reduction by x[0]
+    def tolist(self):
+        """
+        Return the matrix as a (possibly nested) list.
+
+        See `ndarray.tolist` for full documentation.
+
+        See Also
+        --------
+        ndarray.tolist
+
+        Examples
+        --------
+        >>> x = np.matrix(np.arange(12).reshape((3,4))); x
+        matrix([[ 0,  1,  2,  3],
+                [ 4,  5,  6,  7],
+                [ 8,  9, 10, 11]])
+        >>> x.tolist()
+        [[0, 1, 2, 3], [4, 5, 6, 7], [8, 9, 10, 11]]
+
+        """
+        return self.__array__().tolist()
+
+    # To preserve orientation of result...
+    def sum(self, axis=None, dtype=None, out=None):
+        """
+        Returns the sum of the matrix elements, along the given axis.
+
+        Refer to `numpy.sum` for full documentation.
+
+        See Also
+        --------
+        numpy.sum
+
+        Notes
+        -----
+        This is the same as `ndarray.sum`, except that where an `ndarray` would
+        be returned, a `matrix` object is returned instead.
+
+        Examples
+        --------
+        >>> x = np.matrix([[1, 2], [4, 3]])
+        >>> x.sum()
+        10
+        >>> x.sum(axis=1)
+        matrix([[3],
+                [7]])
+        >>> x.sum(axis=1, dtype='float')
+        matrix([[3.],
+                [7.]])
+        >>> out = np.zeros((2, 1), dtype='float')
+        >>> x.sum(axis=1, dtype='float', out=np.asmatrix(out))
+        matrix([[3.],
+                [7.]])
+
+        """
+        return N.ndarray.sum(self, axis, dtype, out, keepdims=True)._collapse(axis)
+
+
+    # To update docstring from array to matrix...
+    def squeeze(self, axis=None):
+        """
+        Return a possibly reshaped matrix.
+
+        Refer to `numpy.squeeze` for more documentation.
+
+        Parameters
+        ----------
+        axis : None or int or tuple of ints, optional
+            Selects a subset of the axes of length one in the shape.
+            If an axis is selected with shape entry greater than one,
+            an error is raised.
+
+        Returns
+        -------
+        squeezed : matrix
+            The matrix, but as a (1, N) matrix if it had shape (N, 1).
+
+        See Also
+        --------
+        numpy.squeeze : related function
+
+        Notes
+        -----
+        If `m` has a single column then that column is returned
+        as the single row of a matrix.  Otherwise `m` is returned.
+        The returned matrix is always either `m` itself or a view into `m`.
+        Supplying an axis keyword argument will not affect the returned matrix
+        but it may cause an error to be raised.
+
+        Examples
+        --------
+        >>> c = np.matrix([[1], [2]])
+        >>> c
+        matrix([[1],
+                [2]])
+        >>> c.squeeze()
+        matrix([[1, 2]])
+        >>> r = c.T
+        >>> r
+        matrix([[1, 2]])
+        >>> r.squeeze()
+        matrix([[1, 2]])
+        >>> m = np.matrix([[1, 2], [3, 4]])
+        >>> m.squeeze()
+        matrix([[1, 2],
+                [3, 4]])
+
+        """
+        return N.ndarray.squeeze(self, axis=axis)
+
+
+    # To update docstring from array to matrix...
+    def flatten(self, order='C'):
+        """
+        Return a flattened copy of the matrix.
+
+        All `N` elements of the matrix are placed into a single row.
+
+        Parameters
+        ----------
+        order : {'C', 'F', 'A', 'K'}, optional
+            'C' means to flatten in row-major (C-style) order. 'F' means to
+            flatten in column-major (Fortran-style) order. 'A' means to
+            flatten in column-major order if `m` is Fortran *contiguous* in
+            memory, row-major order otherwise. 'K' means to flatten `m` in
+            the order the elements occur in memory. The default is 'C'.
+
+        Returns
+        -------
+        y : matrix
+            A copy of the matrix, flattened to a `(1, N)` matrix where `N`
+            is the number of elements in the original matrix.
+
+        See Also
+        --------
+        ravel : Return a flattened array.
+        flat : A 1-D flat iterator over the matrix.
+
+        Examples
+        --------
+        >>> m = np.matrix([[1,2], [3,4]])
+        >>> m.flatten()
+        matrix([[1, 2, 3, 4]])
+        >>> m.flatten('F')
+        matrix([[1, 3, 2, 4]])
+
+        """
+        return N.ndarray.flatten(self, order=order)
+
+    def mean(self, axis=None, dtype=None, out=None):
+        """
+        Returns the average of the matrix elements along the given axis.
+
+        Refer to `numpy.mean` for full documentation.
+
+        See Also
+        --------
+        numpy.mean
+
+        Notes
+        -----
+        Same as `ndarray.mean` except that, where that returns an `ndarray`,
+        this returns a `matrix` object.
+
+        Examples
+        --------
+        >>> x = np.matrix(np.arange(12).reshape((3, 4)))
+        >>> x
+        matrix([[ 0,  1,  2,  3],
+                [ 4,  5,  6,  7],
+                [ 8,  9, 10, 11]])
+        >>> x.mean()
+        5.5
+        >>> x.mean(0)
+        matrix([[4., 5., 6., 7.]])
+        >>> x.mean(1)
+        matrix([[ 1.5],
+                [ 5.5],
+                [ 9.5]])
+
+        """
+        return N.ndarray.mean(self, axis, dtype, out, keepdims=True)._collapse(axis)
+
+    def std(self, axis=None, dtype=None, out=None, ddof=0):
+        """
+        Return the standard deviation of the array elements along the given axis.
+
+        Refer to `numpy.std` for full documentation.
+
+        See Also
+        --------
+        numpy.std
+
+        Notes
+        -----
+        This is the same as `ndarray.std`, except that where an `ndarray` would
+        be returned, a `matrix` object is returned instead.
+
+        Examples
+        --------
+        >>> x = np.matrix(np.arange(12).reshape((3, 4)))
+        >>> x
+        matrix([[ 0,  1,  2,  3],
+                [ 4,  5,  6,  7],
+                [ 8,  9, 10, 11]])
+        >>> x.std()
+        3.4520525295346629 # may vary
+        >>> x.std(0)
+        matrix([[ 3.26598632,  3.26598632,  3.26598632,  3.26598632]]) # may vary
+        >>> x.std(1)
+        matrix([[ 1.11803399],
+                [ 1.11803399],
+                [ 1.11803399]])
+
+        """
+        return N.ndarray.std(self, axis, dtype, out, ddof, keepdims=True)._collapse(axis)
+
+    def var(self, axis=None, dtype=None, out=None, ddof=0):
+        """
+        Returns the variance of the matrix elements, along the given axis.
+
+        Refer to `numpy.var` for full documentation.
+
+        See Also
+        --------
+        numpy.var
+
+        Notes
+        -----
+        This is the same as `ndarray.var`, except that where an `ndarray` would
+        be returned, a `matrix` object is returned instead.
+
+        Examples
+        --------
+        >>> x = np.matrix(np.arange(12).reshape((3, 4)))
+        >>> x
+        matrix([[ 0,  1,  2,  3],
+                [ 4,  5,  6,  7],
+                [ 8,  9, 10, 11]])
+        >>> x.var()
+        11.916666666666666
+        >>> x.var(0)
+        matrix([[ 10.66666667,  10.66666667,  10.66666667,  10.66666667]]) # may vary
+        >>> x.var(1)
+        matrix([[1.25],
+                [1.25],
+                [1.25]])
+
+        """
+        return N.ndarray.var(self, axis, dtype, out, ddof, keepdims=True)._collapse(axis)
+
+    def prod(self, axis=None, dtype=None, out=None):
+        """
+        Return the product of the array elements over the given axis.
+
+        Refer to `prod` for full documentation.
+
+        See Also
+        --------
+        prod, ndarray.prod
+
+        Notes
+        -----
+        Same as `ndarray.prod`, except, where that returns an `ndarray`, this
+        returns a `matrix` object instead.
+
+        Examples
+        --------
+        >>> x = np.matrix(np.arange(12).reshape((3,4))); x
+        matrix([[ 0,  1,  2,  3],
+                [ 4,  5,  6,  7],
+                [ 8,  9, 10, 11]])
+        >>> x.prod()
+        0
+        >>> x.prod(0)
+        matrix([[  0,  45, 120, 231]])
+        >>> x.prod(1)
+        matrix([[   0],
+                [ 840],
+                [7920]])
+
+        """
+        return N.ndarray.prod(self, axis, dtype, out, keepdims=True)._collapse(axis)
+
+    def any(self, axis=None, out=None):
+        """
+        Test whether any array element along a given axis evaluates to True.
+
+        Refer to `numpy.any` for full documentation.
+
+        Parameters
+        ----------
+        axis : int, optional
+            Axis along which logical OR is performed
+        out : ndarray, optional
+            Output to existing array instead of creating new one, must have
+            same shape as expected output
+
+        Returns
+        -------
+            any : bool, ndarray
+                Returns a single bool if `axis` is ``None``; otherwise,
+                returns `ndarray`
+
+        """
+        return N.ndarray.any(self, axis, out, keepdims=True)._collapse(axis)
+
+    def all(self, axis=None, out=None):
+        """
+        Test whether all matrix elements along a given axis evaluate to True.
+
+        Parameters
+        ----------
+        See `numpy.all` for complete descriptions
+
+        See Also
+        --------
+        numpy.all
+
+        Notes
+        -----
+        This is the same as `ndarray.all`, but it returns a `matrix` object.
+
+        Examples
+        --------
+        >>> x = np.matrix(np.arange(12).reshape((3,4))); x
+        matrix([[ 0,  1,  2,  3],
+                [ 4,  5,  6,  7],
+                [ 8,  9, 10, 11]])
+        >>> y = x[0]; y
+        matrix([[0, 1, 2, 3]])
+        >>> (x == y)
+        matrix([[ True,  True,  True,  True],
+                [False, False, False, False],
+                [False, False, False, False]])
+        >>> (x == y).all()
+        False
+        >>> (x == y).all(0)
+        matrix([[False, False, False, False]])
+        >>> (x == y).all(1)
+        matrix([[ True],
+                [False],
+                [False]])
+
+        """
+        return N.ndarray.all(self, axis, out, keepdims=True)._collapse(axis)
+
+    def max(self, axis=None, out=None):
+        """
+        Return the maximum value along an axis.
+
+        Parameters
+        ----------
+        See `amax` for complete descriptions
+
+        See Also
+        --------
+        amax, ndarray.max
+
+        Notes
+        -----
+        This is the same as `ndarray.max`, but returns a `matrix` object
+        where `ndarray.max` would return an ndarray.
+
+        Examples
+        --------
+        >>> x = np.matrix(np.arange(12).reshape((3,4))); x
+        matrix([[ 0,  1,  2,  3],
+                [ 4,  5,  6,  7],
+                [ 8,  9, 10, 11]])
+        >>> x.max()
+        11
+        >>> x.max(0)
+        matrix([[ 8,  9, 10, 11]])
+        >>> x.max(1)
+        matrix([[ 3],
+                [ 7],
+                [11]])
+
+        """
+        return N.ndarray.max(self, axis, out, keepdims=True)._collapse(axis)
+
+    def argmax(self, axis=None, out=None):
+        """
+        Indexes of the maximum values along an axis.
+
+        Return the indexes of the first occurrences of the maximum values
+        along the specified axis.  If axis is None, the index is for the
+        flattened matrix.
+
+        Parameters
+        ----------
+        See `numpy.argmax` for complete descriptions
+
+        See Also
+        --------
+        numpy.argmax
+
+        Notes
+        -----
+        This is the same as `ndarray.argmax`, but returns a `matrix` object
+        where `ndarray.argmax` would return an `ndarray`.
+
+        Examples
+        --------
+        >>> x = np.matrix(np.arange(12).reshape((3,4))); x
+        matrix([[ 0,  1,  2,  3],
+                [ 4,  5,  6,  7],
+                [ 8,  9, 10, 11]])
+        >>> x.argmax()
+        11
+        >>> x.argmax(0)
+        matrix([[2, 2, 2, 2]])
+        >>> x.argmax(1)
+        matrix([[3],
+                [3],
+                [3]])
+
+        """
+        return N.ndarray.argmax(self, axis, out)._align(axis)
+
+    def min(self, axis=None, out=None):
+        """
+        Return the minimum value along an axis.
+
+        Parameters
+        ----------
+        See `amin` for complete descriptions.
+
+        See Also
+        --------
+        amin, ndarray.min
+
+        Notes
+        -----
+        This is the same as `ndarray.min`, but returns a `matrix` object
+        where `ndarray.min` would return an ndarray.
+
+        Examples
+        --------
+        >>> x = -np.matrix(np.arange(12).reshape((3,4))); x
+        matrix([[  0,  -1,  -2,  -3],
+                [ -4,  -5,  -6,  -7],
+                [ -8,  -9, -10, -11]])
+        >>> x.min()
+        -11
+        >>> x.min(0)
+        matrix([[ -8,  -9, -10, -11]])
+        >>> x.min(1)
+        matrix([[ -3],
+                [ -7],
+                [-11]])
+
+        """
+        return N.ndarray.min(self, axis, out, keepdims=True)._collapse(axis)
+
+    def argmin(self, axis=None, out=None):
+        """
+        Indexes of the minimum values along an axis.
+
+        Return the indexes of the first occurrences of the minimum values
+        along the specified axis.  If axis is None, the index is for the
+        flattened matrix.
+
+        Parameters
+        ----------
+        See `numpy.argmin` for complete descriptions.
+
+        See Also
+        --------
+        numpy.argmin
+
+        Notes
+        -----
+        This is the same as `ndarray.argmin`, but returns a `matrix` object
+        where `ndarray.argmin` would return an `ndarray`.
+
+        Examples
+        --------
+        >>> x = -np.matrix(np.arange(12).reshape((3,4))); x
+        matrix([[  0,  -1,  -2,  -3],
+                [ -4,  -5,  -6,  -7],
+                [ -8,  -9, -10, -11]])
+        >>> x.argmin()
+        11
+        >>> x.argmin(0)
+        matrix([[2, 2, 2, 2]])
+        >>> x.argmin(1)
+        matrix([[3],
+                [3],
+                [3]])
+
+        """
+        return N.ndarray.argmin(self, axis, out)._align(axis)
+
+    def ptp(self, axis=None, out=None):
+        """
+        Peak-to-peak (maximum - minimum) value along the given axis.
+
+        Refer to `numpy.ptp` for full documentation.
+
+        See Also
+        --------
+        numpy.ptp
+
+        Notes
+        -----
+        Same as `ndarray.ptp`, except, where that would return an `ndarray` object,
+        this returns a `matrix` object.
+
+        Examples
+        --------
+        >>> x = np.matrix(np.arange(12).reshape((3,4))); x
+        matrix([[ 0,  1,  2,  3],
+                [ 4,  5,  6,  7],
+                [ 8,  9, 10, 11]])
+        >>> x.ptp()
+        11
+        >>> x.ptp(0)
+        matrix([[8, 8, 8, 8]])
+        >>> x.ptp(1)
+        matrix([[3],
+                [3],
+                [3]])
+
+        """
+        return N.ndarray.ptp(self, axis, out)._align(axis)
+
+    @property
+    def I(self):
+        """
+        Returns the (multiplicative) inverse of invertible `self`.
+
+        Parameters
+        ----------
+        None
+
+        Returns
+        -------
+        ret : matrix object
+            If `self` is non-singular, `ret` is such that ``ret * self`` ==
+            ``self * ret`` == ``np.matrix(np.eye(self[0,:].size))`` all return
+            ``True``.
+
+        Raises
+        ------
+        numpy.linalg.LinAlgError: Singular matrix
+            If `self` is singular.
+
+        See Also
+        --------
+        linalg.inv
+
+        Examples
+        --------
+        >>> m = np.matrix('[1, 2; 3, 4]'); m
+        matrix([[1, 2],
+                [3, 4]])
+        >>> m.getI()
+        matrix([[-2. ,  1. ],
+                [ 1.5, -0.5]])
+        >>> m.getI() * m
+        matrix([[ 1.,  0.], # may vary
+                [ 0.,  1.]])
+
+        """
+        M, N = self.shape
+        if M == N:
+            from numpy.linalg import inv as func
+        else:
+            from numpy.linalg import pinv as func
+        return asmatrix(func(self))
+
+    @property
+    def A(self):
+        """
+        Return `self` as an `ndarray` object.
+
+        Equivalent to ``np.asarray(self)``.
+
+        Parameters
+        ----------
+        None
+
+        Returns
+        -------
+        ret : ndarray
+            `self` as an `ndarray`
+
+        Examples
+        --------
+        >>> x = np.matrix(np.arange(12).reshape((3,4))); x
+        matrix([[ 0,  1,  2,  3],
+                [ 4,  5,  6,  7],
+                [ 8,  9, 10, 11]])
+        >>> x.getA()
+        array([[ 0,  1,  2,  3],
+               [ 4,  5,  6,  7],
+               [ 8,  9, 10, 11]])
+
+        """
+        return self.__array__()
+
+    @property
+    def A1(self):
+        """
+        Return `self` as a flattened `ndarray`.
+
+        Equivalent to ``np.asarray(x).ravel()``
+
+        Parameters
+        ----------
+        None
+
+        Returns
+        -------
+        ret : ndarray
+            `self`, 1-D, as an `ndarray`
+
+        Examples
+        --------
+        >>> x = np.matrix(np.arange(12).reshape((3,4))); x
+        matrix([[ 0,  1,  2,  3],
+                [ 4,  5,  6,  7],
+                [ 8,  9, 10, 11]])
+        >>> x.getA1()
+        array([ 0,  1,  2, ...,  9, 10, 11])
+
+
+        """
+        return self.__array__().ravel()
+
+
+    def ravel(self, order='C'):
+        """
+        Return a flattened matrix.
+
+        Refer to `numpy.ravel` for more documentation.
+
+        Parameters
+        ----------
+        order : {'C', 'F', 'A', 'K'}, optional
+            The elements of `m` are read using this index order. 'C' means to
+            index the elements in C-like order, with the last axis index
+            changing fastest, back to the first axis index changing slowest.
+            'F' means to index the elements in Fortran-like index order, with
+            the first index changing fastest, and the last index changing
+            slowest. Note that the 'C' and 'F' options take no account of the
+            memory layout of the underlying array, and only refer to the order
+            of axis indexing.  'A' means to read the elements in Fortran-like
+            index order if `m` is Fortran *contiguous* in memory, C-like order
+            otherwise.  'K' means to read the elements in the order they occur
+            in memory, except for reversing the data when strides are negative.
+            By default, 'C' index order is used.
+
+        Returns
+        -------
+        ret : matrix
+            Return the matrix flattened to shape `(1, N)` where `N`
+            is the number of elements in the original matrix.
+            A copy is made only if necessary.
+
+        See Also
+        --------
+        matrix.flatten : returns a similar output matrix but always a copy
+        matrix.flat : a flat iterator on the array.
+        numpy.ravel : related function which returns an ndarray
+
+        """
+        return N.ndarray.ravel(self, order=order)
+
+    @property
+    def T(self):
+        """
+        Returns the transpose of the matrix.
+
+        Does *not* conjugate!  For the complex conjugate transpose, use ``.H``.
+
+        Parameters
+        ----------
+        None
+
+        Returns
+        -------
+        ret : matrix object
+            The (non-conjugated) transpose of the matrix.
+
+        See Also
+        --------
+        transpose, getH
+
+        Examples
+        --------
+        >>> m = np.matrix('[1, 2; 3, 4]')
+        >>> m
+        matrix([[1, 2],
+                [3, 4]])
+        >>> m.getT()
+        matrix([[1, 3],
+                [2, 4]])
+
+        """
+        return self.transpose()
+
+    @property
+    def H(self):
+        """
+        Returns the (complex) conjugate transpose of `self`.
+
+        Equivalent to ``np.transpose(self)`` if `self` is real-valued.
+
+        Parameters
+        ----------
+        None
+
+        Returns
+        -------
+        ret : matrix object
+            complex conjugate transpose of `self`
+
+        Examples
+        --------
+        >>> x = np.matrix(np.arange(12).reshape((3,4)))
+        >>> z = x - 1j*x; z
+        matrix([[  0. +0.j,   1. -1.j,   2. -2.j,   3. -3.j],
+                [  4. -4.j,   5. -5.j,   6. -6.j,   7. -7.j],
+                [  8. -8.j,   9. -9.j,  10.-10.j,  11.-11.j]])
+        >>> z.getH()
+        matrix([[ 0. -0.j,  4. +4.j,  8. +8.j],
+                [ 1. +1.j,  5. +5.j,  9. +9.j],
+                [ 2. +2.j,  6. +6.j, 10.+10.j],
+                [ 3. +3.j,  7. +7.j, 11.+11.j]])
+
+        """
+        if issubclass(self.dtype.type, N.complexfloating):
+            return self.transpose().conjugate()
+        else:
+            return self.transpose()
+
+    # kept for compatibility
+    getT = T.fget
+    getA = A.fget
+    getA1 = A1.fget
+    getH = H.fget
+    getI = I.fget
+
+def _from_string(str, gdict, ldict):
+    rows = str.split(';')
+    rowtup = []
+    for row in rows:
+        trow = row.split(',')
+        newrow = []
+        for x in trow:
+            newrow.extend(x.split())
+        trow = newrow
+        coltup = []
+        for col in trow:
+            col = col.strip()
+            try:
+                thismat = ldict[col]
+            except KeyError:
+                try:
+                    thismat = gdict[col]
+                except KeyError as e:
+                    raise NameError(f"name {col!r} is not defined") from None
+
+            coltup.append(thismat)
+        rowtup.append(concatenate(coltup, axis=-1))
+    return concatenate(rowtup, axis=0)
+
+
+@set_module('numpy')
+def bmat(obj, ldict=None, gdict=None):
+    """
+    Build a matrix object from a string, nested sequence, or array.
+
+    Parameters
+    ----------
+    obj : str or array_like
+        Input data. If a string, variables in the current scope may be
+        referenced by name.
+    ldict : dict, optional
+        A dictionary that replaces local operands in current frame.
+        Ignored if `obj` is not a string or `gdict` is None.
+    gdict : dict, optional
+        A dictionary that replaces global operands in current frame.
+        Ignored if `obj` is not a string.
+
+    Returns
+    -------
+    out : matrix
+        Returns a matrix object, which is a specialized 2-D array.
+
+    See Also
+    --------
+    block :
+        A generalization of this function for N-d arrays, that returns normal
+        ndarrays.
+
+    Examples
+    --------
+    >>> A = np.mat('1 1; 1 1')
+    >>> B = np.mat('2 2; 2 2')
+    >>> C = np.mat('3 4; 5 6')
+    >>> D = np.mat('7 8; 9 0')
+
+    All the following expressions construct the same block matrix:
+
+    >>> np.bmat([[A, B], [C, D]])
+    matrix([[1, 1, 2, 2],
+            [1, 1, 2, 2],
+            [3, 4, 7, 8],
+            [5, 6, 9, 0]])
+    >>> np.bmat(np.r_[np.c_[A, B], np.c_[C, D]])
+    matrix([[1, 1, 2, 2],
+            [1, 1, 2, 2],
+            [3, 4, 7, 8],
+            [5, 6, 9, 0]])
+    >>> np.bmat('A,B; C,D')
+    matrix([[1, 1, 2, 2],
+            [1, 1, 2, 2],
+            [3, 4, 7, 8],
+            [5, 6, 9, 0]])
+
+    """
+    if isinstance(obj, str):
+        if gdict is None:
+            # get previous frame
+            frame = sys._getframe().f_back
+            glob_dict = frame.f_globals
+            loc_dict = frame.f_locals
+        else:
+            glob_dict = gdict
+            loc_dict = ldict
+
+        return matrix(_from_string(obj, glob_dict, loc_dict))
+
+    if isinstance(obj, (tuple, list)):
+        # [[A,B],[C,D]]
+        arr_rows = []
+        for row in obj:
+            if isinstance(row, N.ndarray):  # not 2-d
+                return matrix(concatenate(obj, axis=-1))
+            else:
+                arr_rows.append(concatenate(row, axis=-1))
+        return matrix(concatenate(arr_rows, axis=0))
+    if isinstance(obj, N.ndarray):
+        return matrix(obj)
+
+mat = asmatrix

LTA_openwebtext_dualt/mini_owt_logdirichlet/.venv_qwen35_uv/lib/python3.12/site-packages/numpy/matrixlib/defmatrix.pyi

@@ -0,0 +1,16 @@
+from collections.abc import Sequence, Mapping
+from typing import Any
+from numpy import matrix as matrix
+from numpy._typing import ArrayLike, DTypeLike, NDArray
+
+__all__: list[str]
+
+def bmat(
+    obj: str | Sequence[ArrayLike] | NDArray[Any],
+    ldict: None | Mapping[str, Any] = ...,
+    gdict: None | Mapping[str, Any] = ...,
+) -> matrix[Any, Any]: ...
+
+def asmatrix(data: ArrayLike, dtype: DTypeLike = ...) -> matrix[Any, Any]: ...
+
+mat = asmatrix

LTA_openwebtext_dualt/mini_owt_logdirichlet/.venv_qwen35_uv/lib/python3.12/site-packages/numpy/matrixlib/setup.py

@@ -0,0 +1,12 @@
+#!/usr/bin/env python3
+def configuration(parent_package='', top_path=None):
+    from numpy.distutils.misc_util import Configuration
+    config = Configuration('matrixlib', parent_package, top_path)
+    config.add_subpackage('tests')
+    config.add_data_files('*.pyi')
+    return config
+
+if __name__ == "__main__":
+    from numpy.distutils.core import setup
+    config = configuration(top_path='').todict()
+    setup(**config)

LTA_openwebtext_dualt/mini_owt_logdirichlet/.venv_qwen35_uv/lib/python3.12/site-packages/transformers/models/roberta/__init__.py

@@ -0,0 +1,28 @@
+# Copyright 2024 The HuggingFace Team. All rights reserved.
+#
+# Licensed under the Apache License, Version 2.0 (the "License");
+# you may not use this file except in compliance with the License.
+# You may obtain a copy of the License at
+#
+#     http://www.apache.org/licenses/LICENSE-2.0
+#
+# Unless required by applicable law or agreed to in writing, software
+# distributed under the License is distributed on an "AS IS" BASIS,
+# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+# See the License for the specific language governing permissions and
+# limitations under the License.
+from typing import TYPE_CHECKING
+
+from ...utils import _LazyModule
+from ...utils.import_utils import define_import_structure
+
+
+if TYPE_CHECKING:
+    from .configuration_roberta import *
+    from .modeling_roberta import *
+    from .tokenization_roberta import *
+else:
+    import sys
+
+    _file = globals()["__file__"]
+    sys.modules[__name__] = _LazyModule(__name__, _file, define_import_structure(_file), module_spec=__spec__)

LTA_openwebtext_dualt/mini_owt_logdirichlet/.venv_qwen35_uv/lib/python3.12/site-packages/transformers/models/roberta/tokenization_roberta_old.py

@@ -0,0 +1,262 @@
+# Copyright 2018 The Open AI Team Authors and The HuggingFace Inc. team.
+#
+# Licensed under the Apache License, Version 2.0 (the "License");
+# you may not use this file except in compliance with the License.
+# You may obtain a copy of the License at
+#
+#     http://www.apache.org/licenses/LICENSE-2.0
+#
+# Unless required by applicable law or agreed to in writing, software
+# distributed under the License is distributed on an "AS IS" BASIS,
+# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+# See the License for the specific language governing permissions and
+# limitations under the License.
+"""Fast Tokenization classes for RoBERTa."""
+
+import json
+
+from tokenizers import processors
+
+from ...tokenization_utils_base import AddedToken, BatchEncoding
+from ...tokenization_utils_tokenizers import PreTrainedTokenizerFast
+from ...utils import logging
+from .tokenization_roberta import RobertaTokenizer
+
+
+logger = logging.get_logger(__name__)
+
+VOCAB_FILES_NAMES = {"vocab_file": "vocab.json", "merges_file": "merges.txt", "tokenizer_file": "tokenizer.json"}
+
+
+class RobertaTokenizerFast(PreTrainedTokenizerFast):
+    """
+    Construct a "fast" RoBERTa tokenizer (backed by HuggingFace's *tokenizers* library), derived from the GPT-2
+    tokenizer, using byte-level Byte-Pair-Encoding.
+
+    This tokenizer has been trained to treat spaces like parts of the tokens (a bit like sentencepiece) so a word will
+    be encoded differently whether it is at the beginning of the sentence (without space) or not:
+
+    ```python
+    >>> from transformers import RobertaTokenizerFast
+
+    >>> tokenizer = RobertaTokenizerFast.from_pretrained("FacebookAI/roberta-base")
+    >>> tokenizer("Hello world")["input_ids"]
+    [0, 31414, 232, 2]
+
+    >>> tokenizer(" Hello world")["input_ids"]
+    [0, 20920, 232, 2]
+    ```
+
+    You can get around that behavior by passing `add_prefix_space=True` when instantiating this tokenizer or when you
+    call it on some text, but since the model was not pretrained this way, it might yield a decrease in performance.
+
+    <Tip>
+
+    When used with `is_split_into_words=True`, this tokenizer needs to be instantiated with `add_prefix_space=True`.
+
+    </Tip>
+
+    This tokenizer inherits from [`PreTrainedTokenizerFast`] which contains most of the main methods. Users should
+    refer to this superclass for more information regarding those methods.
+
+    Args:
+        vocab_file (`str`):
+            Path to the vocabulary file.
+        merges_file (`str`):
+            Path to the merges file.
+        errors (`str`, *optional*, defaults to `"replace"`):
+            Paradigm to follow when decoding bytes to UTF-8. See
+            [bytes.decode](https://docs.python.org/3/library/stdtypes.html#bytes.decode) for more information.
+        bos_token (`str`, *optional*, defaults to `"<s>"`):
+            The beginning of sequence token that was used during pretraining. Can be used a sequence classifier token.
+
+            <Tip>
+
+            When building a sequence using special tokens, this is not the token that is used for the beginning of
+            sequence. The token used is the `cls_token`.
+
+            </Tip>
+
+        eos_token (`str`, *optional*, defaults to `"</s>"`):
+            The end of sequence token.
+
+            <Tip>
+
+            When building a sequence using special tokens, this is not the token that is used for the end of sequence.
+            The token used is the `sep_token`.
+
+            </Tip>
+
+        sep_token (`str`, *optional*, defaults to `"</s>"`):
+            The separator token, which is used when building a sequence from multiple sequences, e.g. two sequences for
+            sequence classification or for a text and a question for question answering. It is also used as the last
+            token of a sequence built with special tokens.
+        cls_token (`str`, *optional*, defaults to `"<s>"`):
+            The classifier token which is used when doing sequence classification (classification of the whole sequence
+            instead of per-token classification). It is the first token of the sequence when built with special tokens.
+        unk_token (`str`, *optional*, defaults to `"<unk>"`):
+            The unknown token. A token that is not in the vocabulary cannot be converted to an ID and is set to be this
+            token instead.
+        pad_token (`str`, *optional*, defaults to `"<pad>"`):
+            The token used for padding, for example when batching sequences of different lengths.
+        mask_token (`str`, *optional*, defaults to `"<mask>"`):
+            The token used for masking values. This is the token used when training this model with masked language
+            modeling. This is the token which the model will try to predict.
+        add_prefix_space (`bool`, *optional*, defaults to `False`):
+            Whether or not to add an initial space to the input. This allows to treat the leading word just as any
+            other word. (RoBERTa tokenizer detect beginning of words by the preceding space).
+        trim_offsets (`bool`, *optional*, defaults to `True`):
+            Whether the post processing step should trim offsets to avoid including whitespaces.
+    """
+
+    vocab_files_names = VOCAB_FILES_NAMES
+    model_input_names = ["input_ids", "attention_mask"]
+    slow_tokenizer_class = RobertaTokenizer
+
+    def __init__(
+        self,
+        vocab_file=None,
+        merges_file=None,
+        tokenizer_file=None,
+        errors="replace",
+        bos_token="<s>",
+        eos_token="</s>",
+        sep_token="</s>",
+        cls_token="<s>",
+        unk_token="<unk>",
+        pad_token="<pad>",
+        mask_token="<mask>",
+        add_prefix_space=False,
+        trim_offsets=True,
+        **kwargs,
+    ):
+        mask_token = (
+            AddedToken(mask_token, lstrip=True, rstrip=False, normalized=False)
+            if isinstance(mask_token, str)
+            else mask_token
+        )
+        super().__init__(
+            vocab_file,
+            merges_file,
+            tokenizer_file=tokenizer_file,
+            errors=errors,
+            bos_token=bos_token,
+            eos_token=eos_token,
+            sep_token=sep_token,
+            cls_token=cls_token,
+            unk_token=unk_token,
+            pad_token=pad_token,
+            mask_token=mask_token,
+            add_prefix_space=add_prefix_space,
+            trim_offsets=trim_offsets,
+            **kwargs,
+        )
+
+        tokenizer_component = "post_processor"
+        tokenizer_component_instance = getattr(self.backend_tokenizer, tokenizer_component, None)
+        if tokenizer_component_instance:
+            state = json.loads(tokenizer_component_instance.__getstate__())
+
+            # The lists 'sep' and 'cls' must be cased in tuples for the object `post_processor_class`
+            if "sep" in state:
+                state["sep"] = tuple(state["sep"])
+            if "cls" in state:
+                state["cls"] = tuple(state["cls"])
+
+            changes_to_apply = False
+
+            if state.get("add_prefix_space", add_prefix_space) != add_prefix_space:
+                state["add_prefix_space"] = add_prefix_space
+                changes_to_apply = True
+
+            if state.get("trim_offsets", trim_offsets) != trim_offsets:
+                state["trim_offsets"] = trim_offsets
+                changes_to_apply = True
+
+            if changes_to_apply:
+                component_class = getattr(processors, state.pop("type"))
+                new_value = component_class(**state)
+                setattr(self.backend_tokenizer, tokenizer_component, new_value)
+
+    @property
+    def mask_token(self) -> str:
+        """
+        `str`: Mask token, to use when training a model with masked-language modeling. Log an error if used while not
+        having been set.
+
+        Roberta tokenizer has a special mask token to be usable in the fill-mask pipeline. The mask token will greedily
+        comprise the space before the *<mask>*.
+        """
+        if self._mask_token is None:
+            if self.verbose:
+                logger.error("Using mask_token, but it is not set yet.")
+            return None
+        return str(self._mask_token)
+
+    @mask_token.setter
+    def mask_token(self, value):
+        """
+        Overriding the default behavior of the mask token to have it eat the space before it.
+
+        This is needed to preserve backward compatibility with all the previously used models based on Roberta.
+        """
+        # Mask token behave like a normal word, i.e. include the space before it
+        # So we set lstrip to True
+        value = AddedToken(value, lstrip=True, rstrip=False) if isinstance(value, str) else value
+        self._mask_token = value
+
+    def _batch_encode_plus(self, *args, **kwargs) -> BatchEncoding:
+        is_split_into_words = kwargs.get("is_split_into_words", False)
+        assert self.add_prefix_space or not is_split_into_words, (
+            f"You need to instantiate {self.__class__.__name__} with add_prefix_space=True "
+            "to use it with pretokenized inputs."
+        )
+
+        return super()._batch_encode_plus(*args, **kwargs)
+
+    def _encode_plus(self, *args, **kwargs) -> BatchEncoding:
+        is_split_into_words = kwargs.get("is_split_into_words", False)
+
+        assert self.add_prefix_space or not is_split_into_words, (
+            f"You need to instantiate {self.__class__.__name__} with add_prefix_space=True "
+            "to use it with pretokenized inputs."
+        )
+
+        return super()._encode_plus(*args, **kwargs)
+
+    def save_vocabulary(self, save_directory: str, filename_prefix: str | None = None) -> tuple[str]:
+        files = self._tokenizer.model.save(save_directory, name=filename_prefix)
+        return tuple(files)
+
+    def build_inputs_with_special_tokens(self, token_ids_0, token_ids_1=None):
+        output = [self.bos_token_id] + token_ids_0 + [self.eos_token_id]
+        if token_ids_1 is None:
+            return output
+
+        return output + [self.eos_token_id] + token_ids_1 + [self.eos_token_id]
+
+    def create_token_type_ids_from_sequences(
+        self, token_ids_0: list[int], token_ids_1: list[int] | None = None
+    ) -> list[int]:
+        """
+        Create a mask from the two sequences passed to be used in a sequence-pair classification task. RoBERTa does not
+        make use of token type ids, therefore a list of zeros is returned.
+
+        Args:
+            token_ids_0 (`list[int]`):
+                List of IDs.
+            token_ids_1 (`list[int]`, *optional*):
+                Optional second list of IDs for sequence pairs.
+
+        Returns:
+            `list[int]`: List of zeros.
+        """
+        sep = [self.sep_token_id]
+        cls = [self.cls_token_id]
+
+        if token_ids_1 is None:
+            return len(cls + token_ids_0 + sep) * [0]
+        return len(cls + token_ids_0 + sep + sep + token_ids_1 + sep) * [0]
+
+
+__all__ = ["RobertaTokenizerFast"]

LTA_openwebtext_dualt/mini_owt_logdirichlet/runs/owt_t5_elftokenized_full_len1024_C1_to_1024_pow1_d768_l12_h12_gbs512_2x8gpu_50ep_lr1e3_ema0p9999_elfopt_not5_bottleneck128_unfixed_norm_stateprobadd_selfcond_ce_fast_20260609_155046/step_043000.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:ac7bff656f9250a0740c708b5e587c9ca008d69c35d5c768b7a406fee08070a9
+size 927700322

LTA_openwebtext_dualt/mini_owt_logdirichlet/runs/owt_t5_elftokenized_full_len1024_C1_to_1024_pow1_d768_l12_h12_gbs512_2x8gpu_50ep_lr1e3_ema0p9999_elfopt_not5_bottleneck128_unfixed_norm_stateprobadd_selfcond_ce_fast_20260609_155046/step_221000.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:96aa79ce03c9d20b3f6f06a5825a6283effdfbbab93664c7101eb8e29986a01e
+size 927700322

LTA_openwebtext_dualt/mini_owt_logdirichlet/runs/owt_t5_elftokenized_full_len1024_C1_to_1024_pow1_d768_l12_h12_gbs512_2x8gpu_50ep_lr3e3_ema0p9999_elfopt_not5_bottleneck64_unfixed_norm_stateprobadd_selfcond_ce_fast_20260612_062542/step_070000.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:ec675624e4500efc98e08e7111d8f96f0b8dec5db3e1f3d8fc1e3b432d4b1215
+size 910478690

LTA_openwebtext_dualt/mini_owt_logdirichlet/runs/owt_t5_elftokenized_full_len1024_C1_to_1024_pow1_d768_l12_h12_gbs512_2x8gpu_50ep_lr3e3_ema0p9999_elfopt_not5_bottleneck64_unfixed_norm_stateprobadd_selfcond_ce_fast_20260612_062542/step_132000.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:846e08456e9700be976316279b12a2fac1215380bc87e6f46b9a6bc54faea76a
+size 910478690

LTA_openwebtext_dualt/mini_owt_logdirichlet/runs/owt_t5_elftokenized_full_len1024_C1_to_1024_pow1_d768_l12_h12_gbs512_2x8gpu_50ep_lr3e3_ema0p9999_elfopt_not5_bottleneck64_unfixed_norm_stateprobadd_selfcond_ce_fast_20260612_062542/step_246000.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:6cf0aec18f63bd1501c079326e9a2f38361d3e33dbbabab03f291d62c8be0257
+size 910478690

LTA_openwebtext_dualt/mini_owt_logdirichlet/runs/owt_t5_elftokenized_full_len1024_C1_to_1024_pow1_d768_l12_h12_gbs512_2x8gpu_50ep_lr3e3_ema0p9999_elfopt_not5_bottleneck64_unfixed_norm_stateprobadd_selfcond_ce_fast_20260612_062542/step_471000.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:7456cd811fac0f03aed9aabb04e9c502c449c9c6a1fa3224cd5bbfd516024255
+size 910478690