W0618 14:50:43.668000 66664 site-packages/torch/distributed/run.py:862] 
W0618 14:50:43.668000 66664 site-packages/torch/distributed/run.py:862] *****************************************
W0618 14:50:43.668000 66664 site-packages/torch/distributed/run.py:862] Setting OMP_NUM_THREADS environment variable for each process to be 1 in default, to avoid your system being overloaded, please further tune the variable for optimal performance in your application as needed. 
W0618 14:50:43.668000 66664 site-packages/torch/distributed/run.py:862] *****************************************
W0618 14:51:12.952000 66732 site-packages/torch/utils/_pytree.py:630] <enum 'KernelPreference'> is an Enum subclass and is now natively supported by torch.compile as an opaque value type. Calling register_constant() on Enum subclasses is deprecated and will be an error in a future release.
W0618 14:51:12.958000 66738 site-packages/torch/utils/_pytree.py:630] <enum 'KernelPreference'> is an Enum subclass and is now natively supported by torch.compile as an opaque value type. Calling register_constant() on Enum subclasses is deprecated and will be an error in a future release.
W0618 14:51:13.012000 66732 site-packages/torch/utils/_pytree.py:630] <enum 'ScaleCalculationMode'> is an Enum subclass and is now natively supported by torch.compile as an opaque value type. Calling register_constant() on Enum subclasses is deprecated and will be an error in a future release.
W0618 14:51:13.018000 66738 site-packages/torch/utils/_pytree.py:630] <enum 'ScaleCalculationMode'> is an Enum subclass and is now natively supported by torch.compile as an opaque value type. Calling register_constant() on Enum subclasses is deprecated and will be an error in a future release.
W0618 14:51:13.052000 66734 site-packages/torch/utils/_pytree.py:630] <enum 'KernelPreference'> is an Enum subclass and is now natively supported by torch.compile as an opaque value type. Calling register_constant() on Enum subclasses is deprecated and will be an error in a future release.
W0618 14:51:13.077000 66734 site-packages/torch/utils/_pytree.py:630] <enum 'ScaleCalculationMode'> is an Enum subclass and is now natively supported by torch.compile as an opaque value type. Calling register_constant() on Enum subclasses is deprecated and will be an error in a future release.
W0618 14:51:13.111000 66739 site-packages/torch/utils/_pytree.py:630] <enum 'KernelPreference'> is an Enum subclass and is now natively supported by torch.compile as an opaque value type. Calling register_constant() on Enum subclasses is deprecated and will be an error in a future release.
W0618 14:51:13.134000 66739 site-packages/torch/utils/_pytree.py:630] <enum 'ScaleCalculationMode'> is an Enum subclass and is now natively supported by torch.compile as an opaque value type. Calling register_constant() on Enum subclasses is deprecated and will be an error in a future release.
W0618 14:51:13.353000 66737 site-packages/torch/utils/_pytree.py:630] <enum 'KernelPreference'> is an Enum subclass and is now natively supported by torch.compile as an opaque value type. Calling register_constant() on Enum subclasses is deprecated and will be an error in a future release.
W0618 14:51:13.413000 66737 site-packages/torch/utils/_pytree.py:630] <enum 'ScaleCalculationMode'> is an Enum subclass and is now natively supported by torch.compile as an opaque value type. Calling register_constant() on Enum subclasses is deprecated and will be an error in a future release.
W0618 14:51:13.471000 66736 site-packages/torch/utils/_pytree.py:630] <enum 'KernelPreference'> is an Enum subclass and is now natively supported by torch.compile as an opaque value type. Calling register_constant() on Enum subclasses is deprecated and will be an error in a future release.
W0618 14:51:13.501000 66735 site-packages/torch/utils/_pytree.py:630] <enum 'KernelPreference'> is an Enum subclass and is now natively supported by torch.compile as an opaque value type. Calling register_constant() on Enum subclasses is deprecated and will be an error in a future release.
W0618 14:51:13.532000 66736 site-packages/torch/utils/_pytree.py:630] <enum 'ScaleCalculationMode'> is an Enum subclass and is now natively supported by torch.compile as an opaque value type. Calling register_constant() on Enum subclasses is deprecated and will be an error in a future release.
W0618 14:51:13.541000 66733 site-packages/torch/utils/_pytree.py:630] <enum 'KernelPreference'> is an Enum subclass and is now natively supported by torch.compile as an opaque value type. Calling register_constant() on Enum subclasses is deprecated and will be an error in a future release.
W0618 14:51:13.560000 66735 site-packages/torch/utils/_pytree.py:630] <enum 'ScaleCalculationMode'> is an Enum subclass and is now natively supported by torch.compile as an opaque value type. Calling register_constant() on Enum subclasses is deprecated and will be an error in a future release.
W0618 14:51:13.591000 66733 site-packages/torch/utils/_pytree.py:630] <enum 'ScaleCalculationMode'> is an Enum subclass and is now natively supported by torch.compile as an opaque value type. Calling register_constant() on Enum subclasses is deprecated and will be an error in a future release.
cfg-path: examples/speculative/p-eagle/p_eagle_kv_reuse.yaml
cfg-path: examples/speculative/p-eagle/p_eagle_kv_reuse.yaml
cfg-path: examples/speculative/p-eagle/p_eagle_kv_reuse.yamlcfg-path: examples/speculative/p-eagle/p_eagle_kv_reuse.yamlcfg-path: examples/speculative/p-eagle/p_eagle_kv_reuse.yaml


cfg-path: examples/speculative/p-eagle/p_eagle_kv_reuse.yaml
cfg-path: examples/speculative/p-eagle/p_eagle_kv_reuse.yaml
cfg-path: examples/speculative/p-eagle/p_eagle_kv_reuse.yaml
> initializing torch distributed with 32 workers.
h-liuchonghan-translate-a8nodes-main-m-0:66732:66732 [0] NCCL INFO ENV/Plugin: Could not find: libnccl-env.so
h-liuchonghan-translate-a8nodes-main-m-0:66732:66732 [0] NCCL INFO NCCL_SOCKET_IFNAME set by environment to eth0
h-liuchonghan-translate-a8nodes-main-m-0:66732:66732 [0] NCCL INFO Bootstrap: Using eth0:10.178.140.145<0>
h-liuchonghan-translate-a8nodes-main-m-0:66732:66732 [0] NCCL INFO cudaDriverVersion 13000
h-liuchonghan-translate-a8nodes-main-m-0:66735:66735 [3] NCCL INFO ENV/Plugin: Could not find: libnccl-env.so
h-liuchonghan-translate-a8nodes-main-m-0:66735:66735 [3] NCCL INFO cudaDriverVersion 13000
h-liuchonghan-translate-a8nodes-main-m-0:66732:66732 [0] NCCL INFO NCCL version 2.29.7+cuda13.2
h-liuchonghan-translate-a8nodes-main-m-0:66732:66732 [0] NCCL INFO NCCL git version stable b81d6a5a3
h-liuchonghan-translate-a8nodes-main-m-0:66736:66736 [4] NCCL INFO ENV/Plugin: Could not find: libnccl-env.so
h-liuchonghan-translate-a8nodes-main-m-0:66736:66736 [4] NCCL INFO cudaDriverVersion 13000
h-liuchonghan-translate-a8nodes-main-m-0:66735:66735 [3] NCCL INFO NCCL_SOCKET_IFNAME set by environment to eth0
h-liuchonghan-translate-a8nodes-main-m-0:66735:66735 [3] NCCL INFO Bootstrap: Using eth0:10.178.140.145<0>
h-liuchonghan-translate-a8nodes-main-m-0:66735:66735 [3] NCCL INFO NCCL version 2.29.7+cuda13.2
h-liuchonghan-translate-a8nodes-main-m-0:66735:66735 [3] NCCL INFO NCCL git version stable b81d6a5a3
h-liuchonghan-translate-a8nodes-main-m-0:66736:66736 [4] NCCL INFO NCCL_SOCKET_IFNAME set by environment to eth0
h-liuchonghan-translate-a8nodes-main-m-0:66736:66736 [4] NCCL INFO Bootstrap: Using eth0:10.178.140.145<0>
h-liuchonghan-translate-a8nodes-main-m-0:66736:66736 [4] NCCL INFO NCCL version 2.29.7+cuda13.2
h-liuchonghan-translate-a8nodes-main-m-0:66736:66736 [4] NCCL INFO NCCL git version stable b81d6a5a3
h-liuchonghan-translate-a8nodes-main-m-0:66732:66732 [0] NCCL INFO Comm config Blocking set to 1
h-liuchonghan-translate-a8nodes-main-m-0:66737:66737 [5] NCCL INFO ENV/Plugin: Could not find: libnccl-env.so
h-liuchonghan-translate-a8nodes-main-m-0:66737:66737 [5] NCCL INFO cudaDriverVersion 13000
h-liuchonghan-translate-a8nodes-main-m-0:66737:66737 [5] NCCL INFO NCCL_SOCKET_IFNAME set by environment to eth0
h-liuchonghan-translate-a8nodes-main-m-0:66737:66737 [5] NCCL INFO Bootstrap: Using eth0:10.178.140.145<0>
h-liuchonghan-translate-a8nodes-main-m-0:66737:66737 [5] NCCL INFO NCCL version 2.29.7+cuda13.2
h-liuchonghan-translate-a8nodes-main-m-0:66737:66737 [5] NCCL INFO NCCL git version stable b81d6a5a3
h-liuchonghan-translate-a8nodes-main-m-0:66735:66735 [3] NCCL INFO Comm config Blocking set to 1
h-liuchonghan-translate-a8nodes-main-m-0:66736:66736 [4] NCCL INFO Comm config Blocking set to 1
h-liuchonghan-translate-a8nodes-main-m-0:66737:66737 [5] NCCL INFO Comm config Blocking set to 1
h-liuchonghan-translate-a8nodes-main-m-0:66733:66733 [1] NCCL INFO ENV/Plugin: Could not find: libnccl-env.so
h-liuchonghan-translate-a8nodes-main-m-0:66733:66733 [1] NCCL INFO cudaDriverVersion 13000
h-liuchonghan-translate-a8nodes-main-m-0:66733:66733 [1] NCCL INFO NCCL_SOCKET_IFNAME set by environment to eth0
h-liuchonghan-translate-a8nodes-main-m-0:66733:66733 [1] NCCL INFO Bootstrap: Using eth0:10.178.140.145<0>
h-liuchonghan-translate-a8nodes-main-m-0:66733:66733 [1] NCCL INFO NCCL version 2.29.7+cuda13.2
h-liuchonghan-translate-a8nodes-main-m-0:66733:66733 [1] NCCL INFO NCCL git version stable b81d6a5a3
h-liuchonghan-translate-a8nodes-main-m-0:66733:66733 [1] NCCL INFO Comm config Blocking set to 1
h-liuchonghan-translate-a8nodes-main-m-0:66739:66739 [7] NCCL INFO ENV/Plugin: Could not find: libnccl-env.so
h-liuchonghan-translate-a8nodes-main-m-0:66739:66739 [7] NCCL INFO cudaDriverVersion 13000
h-liuchonghan-translate-a8nodes-main-m-0:66739:66739 [7] NCCL INFO NCCL_SOCKET_IFNAME set by environment to eth0
h-liuchonghan-translate-a8nodes-main-m-0:66739:66739 [7] NCCL INFO Bootstrap: Using eth0:10.178.140.145<0>
h-liuchonghan-translate-a8nodes-main-m-0:66739:66739 [7] NCCL INFO NCCL version 2.29.7+cuda13.2
h-liuchonghan-translate-a8nodes-main-m-0:66739:66739 [7] NCCL INFO NCCL git version stable b81d6a5a3
h-liuchonghan-translate-a8nodes-main-m-0:66739:66739 [7] NCCL INFO Comm config Blocking set to 1
h-liuchonghan-translate-a8nodes-main-m-0:66734:66734 [2] NCCL INFO ENV/Plugin: Could not find: libnccl-env.so
h-liuchonghan-translate-a8nodes-main-m-0:66734:66734 [2] NCCL INFO cudaDriverVersion 13000
h-liuchonghan-translate-a8nodes-main-m-0:66738:66738 [6] NCCL INFO ENV/Plugin: Could not find: libnccl-env.so
h-liuchonghan-translate-a8nodes-main-m-0:66738:66738 [6] NCCL INFO cudaDriverVersion 13000
h-liuchonghan-translate-a8nodes-main-m-0:66734:66734 [2] NCCL INFO NCCL_SOCKET_IFNAME set by environment to eth0
h-liuchonghan-translate-a8nodes-main-m-0:66738:66738 [6] NCCL INFO NCCL_SOCKET_IFNAME set by environment to eth0
h-liuchonghan-translate-a8nodes-main-m-0:66734:66734 [2] NCCL INFO Bootstrap: Using eth0:10.178.140.145<0>
h-liuchonghan-translate-a8nodes-main-m-0:66738:66738 [6] NCCL INFO Bootstrap: Using eth0:10.178.140.145<0>
h-liuchonghan-translate-a8nodes-main-m-0:66734:66734 [2] NCCL INFO NCCL version 2.29.7+cuda13.2
h-liuchonghan-translate-a8nodes-main-m-0:66738:66738 [6] NCCL INFO NCCL version 2.29.7+cuda13.2
h-liuchonghan-translate-a8nodes-main-m-0:66734:66734 [2] NCCL INFO NCCL git version stable b81d6a5a3
h-liuchonghan-translate-a8nodes-main-m-0:66738:66738 [6] NCCL INFO NCCL git version stable b81d6a5a3
h-liuchonghan-translate-a8nodes-main-m-0:66734:66734 [2] NCCL INFO Comm config Blocking set to 1
h-liuchonghan-translate-a8nodes-main-m-0:66738:66738 [6] NCCL INFO Comm config Blocking set to 1
h-liuchonghan-translate-a8nodes-main-m-0:66739:66739 [7] NCCL INFO NET/Plugin: Could not find: libnccl-net.so
h-liuchonghan-translate-a8nodes-main-m-0:66739:66739 [7] NCCL INFO NCCL_IB_DISABLE set by environment to 1.
h-liuchonghan-translate-a8nodes-main-m-0:66739:66739 [7] NCCL INFO transport/net_ib/init.cc:396 -> 3
h-liuchonghan-translate-a8nodes-main-m-0:66739:66739 [7] NCCL INFO Failed to initialize NET plugin IB
h-liuchonghan-translate-a8nodes-main-m-0:66739:66739 [7] NCCL INFO NCCL_SOCKET_IFNAME set by environment to eth0
h-liuchonghan-translate-a8nodes-main-m-0:66739:66739 [7] NCCL INFO NET/Socket : Using [0]eth0:10.178.140.145<0>
h-liuchonghan-translate-a8nodes-main-m-0:66739:66739 [7] NCCL INFO Initialized NET plugin Socket
h-liuchonghan-translate-a8nodes-main-m-0:66739:66739 [7] NCCL INFO Assigned NET plugin Socket to comm
h-liuchonghan-translate-a8nodes-main-m-0:66739:66739 [7] NCCL INFO GIN/Plugin: Could not find: libnccl-gin.so
h-liuchonghan-translate-a8nodes-main-m-0:66739:66739 [7] NCCL INFO Failed to initialize any GIN plugin
h-liuchonghan-translate-a8nodes-main-m-0:66739:66739 [7] NCCL INFO Using network Socket
h-liuchonghan-translate-a8nodes-main-m-0:66736:66736 [4] NCCL INFO NET/Plugin: Could not find: libnccl-net.so
h-liuchonghan-translate-a8nodes-main-m-0:66736:66736 [4] NCCL INFO NCCL_IB_DISABLE set by environment to 1.
h-liuchonghan-translate-a8nodes-main-m-0:66736:66736 [4] NCCL INFO transport/net_ib/init.cc:396 -> 3
h-liuchonghan-translate-a8nodes-main-m-0:66736:66736 [4] NCCL INFO Failed to initialize NET plugin IB
h-liuchonghan-translate-a8nodes-main-m-0:66736:66736 [4] NCCL INFO NCCL_SOCKET_IFNAME set by environment to eth0
h-liuchonghan-translate-a8nodes-main-m-0:66736:66736 [4] NCCL INFO NET/Socket : Using [0]eth0:10.178.140.145<0>
h-liuchonghan-translate-a8nodes-main-m-0:66736:66736 [4] NCCL INFO Initialized NET plugin Socket
h-liuchonghan-translate-a8nodes-main-m-0:66736:66736 [4] NCCL INFO Assigned NET plugin Socket to comm
h-liuchonghan-translate-a8nodes-main-m-0:66732:66732 [0] NCCL INFO NET/Plugin: Could not find: libnccl-net.so
h-liuchonghan-translate-a8nodes-main-m-0:66732:66732 [0] NCCL INFO NCCL_IB_DISABLE set by environment to 1.
h-liuchonghan-translate-a8nodes-main-m-0:66732:66732 [0] NCCL INFO transport/net_ib/init.cc:396 -> 3
h-liuchonghan-translate-a8nodes-main-m-0:66732:66732 [0] NCCL INFO Failed to initialize NET plugin IB
h-liuchonghan-translate-a8nodes-main-m-0:66736:66736 [4] NCCL INFO GIN/Plugin: Could not find: libnccl-gin.so
h-liuchonghan-translate-a8nodes-main-m-0:66732:66732 [0] NCCL INFO NCCL_SOCKET_IFNAME set by environment to eth0
h-liuchonghan-translate-a8nodes-main-m-0:66736:66736 [4] NCCL INFO Failed to initialize any GIN plugin
h-liuchonghan-translate-a8nodes-main-m-0:66736:66736 [4] NCCL INFO Using network Socket
h-liuchonghan-translate-a8nodes-main-m-0:66732:66732 [0] NCCL INFO NET/Socket : Using [0]eth0:10.178.140.145<0>
h-liuchonghan-translate-a8nodes-main-m-0:66732:66732 [0] NCCL INFO Initialized NET plugin Socket
h-liuchonghan-translate-a8nodes-main-m-0:66732:66732 [0] NCCL INFO Assigned NET plugin Socket to comm
h-liuchonghan-translate-a8nodes-main-m-0:66732:66732 [0] NCCL INFO GIN/Plugin: Could not find: libnccl-gin.so
h-liuchonghan-translate-a8nodes-main-m-0:66732:66732 [0] NCCL INFO Failed to initialize any GIN plugin
h-liuchonghan-translate-a8nodes-main-m-0:66732:66732 [0] NCCL INFO Using network Socket
h-liuchonghan-translate-a8nodes-main-m-0:66739:66739 [7] NCCL INFO [Rank 7] ncclCommInitRankConfig comm 0x564f0f810dd0 rank 7 nranks 32 cudaDev 7 nvmlDev 7 busId e0000 commId 0x7a425254b479deb7 - Init START
h-liuchonghan-translate-a8nodes-main-m-0:66733:66733 [1] NCCL INFO NET/Plugin: Could not find: libnccl-net.so
h-liuchonghan-translate-a8nodes-main-m-0:66733:66733 [1] NCCL INFO NCCL_IB_DISABLE set by environment to 1.
h-liuchonghan-translate-a8nodes-main-m-0:66733:66733 [1] NCCL INFO transport/net_ib/init.cc:396 -> 3
h-liuchonghan-translate-a8nodes-main-m-0:66733:66733 [1] NCCL INFO Failed to initialize NET plugin IB
h-liuchonghan-translate-a8nodes-main-m-0:66733:66733 [1] NCCL INFO NCCL_SOCKET_IFNAME set by environment to eth0
h-liuchonghan-translate-a8nodes-main-m-0:66733:66733 [1] NCCL INFO NET/Socket : Using [0]eth0:10.178.140.145<0>
h-liuchonghan-translate-a8nodes-main-m-0:66733:66733 [1] NCCL INFO Initialized NET plugin Socket
h-liuchonghan-translate-a8nodes-main-m-0:66733:66733 [1] NCCL INFO Assigned NET plugin Socket to comm
h-liuchonghan-translate-a8nodes-main-m-0:66733:66733 [1] NCCL INFO GIN/Plugin: Could not find: libnccl-gin.so
h-liuchonghan-translate-a8nodes-main-m-0:66733:66733 [1] NCCL INFO Failed to initialize any GIN plugin
h-liuchonghan-translate-a8nodes-main-m-0:66733:66733 [1] NCCL INFO Using network Socket
h-liuchonghan-translate-a8nodes-main-m-0:66736:66736 [4] NCCL INFO [Rank 4] ncclCommInitRankConfig comm 0x562913d20eb0 rank 4 nranks 32 cudaDev 4 nvmlDev 4 busId 9b000 commId 0x7a425254b479deb7 - Init START
h-liuchonghan-translate-a8nodes-main-m-0:66732:66732 [0] NCCL INFO [Rank 0] ncclCommInitRankConfig comm 0x55d1c3b89b70 rank 0 nranks 32 cudaDev 0 nvmlDev 0 busId b000 commId 0x7a425254b479deb7 - Init START
h-liuchonghan-translate-a8nodes-main-m-0:66735:66735 [3] NCCL INFO NET/Plugin: Could not find: libnccl-net.so
h-liuchonghan-translate-a8nodes-main-m-0:66735:66735 [3] NCCL INFO NCCL_IB_DISABLE set by environment to 1.
h-liuchonghan-translate-a8nodes-main-m-0:66735:66735 [3] NCCL INFO transport/net_ib/init.cc:396 -> 3
h-liuchonghan-translate-a8nodes-main-m-0:66735:66735 [3] NCCL INFO Failed to initialize NET plugin IB
h-liuchonghan-translate-a8nodes-main-m-0:66735:66735 [3] NCCL INFO NCCL_SOCKET_IFNAME set by environment to eth0
h-liuchonghan-translate-a8nodes-main-m-0:66735:66735 [3] NCCL INFO NET/Socket : Using [0]eth0:10.178.140.145<0>
h-liuchonghan-translate-a8nodes-main-m-0:66735:66735 [3] NCCL INFO Initialized NET plugin Socket
h-liuchonghan-translate-a8nodes-main-m-0:66735:66735 [3] NCCL INFO Assigned NET plugin Socket to comm
h-liuchonghan-translate-a8nodes-main-m-0:66735:66735 [3] NCCL INFO GIN/Plugin: Could not find: libnccl-gin.so
h-liuchonghan-translate-a8nodes-main-m-0:66735:66735 [3] NCCL INFO Failed to initialize any GIN plugin
h-liuchonghan-translate-a8nodes-main-m-0:66735:66735 [3] NCCL INFO Using network Socket
h-liuchonghan-translate-a8nodes-main-m-0:66733:66733 [1] NCCL INFO [Rank 1] ncclCommInitRankConfig comm 0x55a35ad250b0 rank 1 nranks 32 cudaDev 1 nvmlDev 1 busId 10000 commId 0x7a425254b479deb7 - Init START
h-liuchonghan-translate-a8nodes-main-m-0:66735:66735 [3] NCCL INFO [Rank 3] ncclCommInitRankConfig comm 0x559159d41560 rank 3 nranks 32 cudaDev 3 nvmlDev 3 busId 4e000 commId 0x7a425254b479deb7 - Init START
h-liuchonghan-translate-a8nodes-main-m-0:66737:66737 [5] NCCL INFO NET/Plugin: Could not find: libnccl-net.so
h-liuchonghan-translate-a8nodes-main-m-0:66737:66737 [5] NCCL INFO NCCL_IB_DISABLE set by environment to 1.
h-liuchonghan-translate-a8nodes-main-m-0:66737:66737 [5] NCCL INFO transport/net_ib/init.cc:396 -> 3
h-liuchonghan-translate-a8nodes-main-m-0:66737:66737 [5] NCCL INFO Failed to initialize NET plugin IB
h-liuchonghan-translate-a8nodes-main-m-0:66737:66737 [5] NCCL INFO NCCL_SOCKET_IFNAME set by environment to eth0
h-liuchonghan-translate-a8nodes-main-m-0:66737:66737 [5] NCCL INFO NET/Socket : Using [0]eth0:10.178.140.145<0>
h-liuchonghan-translate-a8nodes-main-m-0:66737:66737 [5] NCCL INFO Initialized NET plugin Socket
h-liuchonghan-translate-a8nodes-main-m-0:66737:66737 [5] NCCL INFO Assigned NET plugin Socket to comm
h-liuchonghan-translate-a8nodes-main-m-0:66737:66737 [5] NCCL INFO GIN/Plugin: Could not find: libnccl-gin.so
h-liuchonghan-translate-a8nodes-main-m-0:66737:66737 [5] NCCL INFO Failed to initialize any GIN plugin
h-liuchonghan-translate-a8nodes-main-m-0:66737:66737 [5] NCCL INFO Using network Socket
h-liuchonghan-translate-a8nodes-main-m-0:66737:66737 [5] NCCL INFO [Rank 5] ncclCommInitRankConfig comm 0x55f6153f0f30 rank 5 nranks 32 cudaDev 5 nvmlDev 5 busId a1000 commId 0x7a425254b479deb7 - Init START
h-liuchonghan-translate-a8nodes-main-m-0:66738:66738 [6] NCCL INFO NET/Plugin: Could not find: libnccl-net.so
h-liuchonghan-translate-a8nodes-main-m-0:66738:66738 [6] NCCL INFO NCCL_IB_DISABLE set by environment to 1.
h-liuchonghan-translate-a8nodes-main-m-0:66738:66738 [6] NCCL INFO transport/net_ib/init.cc:396 -> 3
h-liuchonghan-translate-a8nodes-main-m-0:66738:66738 [6] NCCL INFO Failed to initialize NET plugin IB
h-liuchonghan-translate-a8nodes-main-m-0:66738:66738 [6] NCCL INFO NCCL_SOCKET_IFNAME set by environment to eth0
h-liuchonghan-translate-a8nodes-main-m-0:66738:66738 [6] NCCL INFO NET/Socket : Using [0]eth0:10.178.140.145<0>
h-liuchonghan-translate-a8nodes-main-m-0:66738:66738 [6] NCCL INFO Initialized NET plugin Socket
h-liuchonghan-translate-a8nodes-main-m-0:66736:66736 [4] NCCL INFO RAS client listening socket at 127.0.0.1<28028>
h-liuchonghan-translate-a8nodes-main-m-0:66738:66738 [6] NCCL INFO Assigned NET plugin Socket to comm
h-liuchonghan-translate-a8nodes-main-m-0:66738:66738 [6] NCCL INFO GIN/Plugin: Could not find: libnccl-gin.so
h-liuchonghan-translate-a8nodes-main-m-0:66738:66738 [6] NCCL INFO Failed to initialize any GIN plugin
h-liuchonghan-translate-a8nodes-main-m-0:66738:66738 [6] NCCL INFO Using network Socket
h-liuchonghan-translate-a8nodes-main-m-0:66738:66738 [6] NCCL INFO [Rank 6] ncclCommInitRankConfig comm 0x55904d46bc60 rank 6 nranks 32 cudaDev 6 nvmlDev 6 busId db000 commId 0x7a425254b479deb7 - Init START
h-liuchonghan-translate-a8nodes-main-m-0:66738:66738 [6] NCCL INFO RAS client listening socket at 127.0.0.1<28028>
h-liuchonghan-translate-a8nodes-main-m-0:66737:66737 [5] NCCL INFO RAS client listening socket at 127.0.0.1<28028>
h-liuchonghan-translate-a8nodes-main-m-0:66734:66734 [2] NCCL INFO NET/Plugin: Could not find: libnccl-net.so
h-liuchonghan-translate-a8nodes-main-m-0:66734:66734 [2] NCCL INFO NCCL_IB_DISABLE set by environment to 1.
h-liuchonghan-translate-a8nodes-main-m-0:66734:66734 [2] NCCL INFO transport/net_ib/init.cc:396 -> 3
h-liuchonghan-translate-a8nodes-main-m-0:66734:66734 [2] NCCL INFO Failed to initialize NET plugin IB
h-liuchonghan-translate-a8nodes-main-m-0:66734:66734 [2] NCCL INFO NCCL_SOCKET_IFNAME set by environment to eth0
h-liuchonghan-translate-a8nodes-main-m-0:66734:66734 [2] NCCL INFO NET/Socket : Using [0]eth0:10.178.140.145<0>
h-liuchonghan-translate-a8nodes-main-m-0:66734:66734 [2] NCCL INFO Initialized NET plugin Socket
h-liuchonghan-translate-a8nodes-main-m-0:66734:66734 [2] NCCL INFO Assigned NET plugin Socket to comm
h-liuchonghan-translate-a8nodes-main-m-0:66734:66734 [2] NCCL INFO GIN/Plugin: Could not find: libnccl-gin.so
h-liuchonghan-translate-a8nodes-main-m-0:66734:66734 [2] NCCL INFO Failed to initialize any GIN plugin
h-liuchonghan-translate-a8nodes-main-m-0:66734:66734 [2] NCCL INFO Using network Socket
h-liuchonghan-translate-a8nodes-main-m-0:66734:66734 [2] NCCL INFO [Rank 2] ncclCommInitRankConfig comm 0x55985c7ad020 rank 2 nranks 32 cudaDev 2 nvmlDev 2 busId 48000 commId 0x7a425254b479deb7 - Init START
h-liuchonghan-translate-a8nodes-main-m-0:66734:66734 [2] NCCL INFO RAS client listening socket at 127.0.0.1<28028>
h-liuchonghan-translate-a8nodes-main-m-0:66733:66733 [1] NCCL INFO RAS client listening socket at 127.0.0.1<28028>
h-liuchonghan-translate-a8nodes-main-m-0:66735:66735 [3] NCCL INFO RAS client listening socket at 127.0.0.1<28028>
h-liuchonghan-translate-a8nodes-main-m-0:66732:66732 [0] NCCL INFO RAS client listening socket at 127.0.0.1<28028>
h-liuchonghan-translate-a8nodes-main-m-0:66739:66739 [7] NCCL INFO RAS client listening socket at 127.0.0.1<28028>
h-liuchonghan-translate-a8nodes-main-m-0:66739:66739 [7] NCCL INFO Bootstrap timings total 22.627825 (create 0.000136, send 0.000230, recv 22.294864, ring 0.331058, delay 0.000004)
h-liuchonghan-translate-a8nodes-main-m-0:66738:66738 [6] NCCL INFO Bootstrap timings total 22.597964 (create 0.000075, send 0.000161, recv 0.000231, ring 22.596825, delay 0.000002)
h-liuchonghan-translate-a8nodes-main-m-0:66737:66737 [5] NCCL INFO Bootstrap timings total 22.604521 (create 0.000091, send 0.000171, recv 0.006695, ring 22.596681, delay 0.000002)
h-liuchonghan-translate-a8nodes-main-m-0:66735:66735 [3] NCCL INFO Bootstrap timings total 22.614470 (create 0.000062, send 0.000132, recv 0.000186, ring 22.577084, delay 0.000003)
h-liuchonghan-translate-a8nodes-main-m-0:66736:66736 [4] NCCL INFO Bootstrap timings total 22.624510 (create 0.000108, send 0.000193, recv 0.019933, ring 22.603118, delay 0.000003)
h-liuchonghan-translate-a8nodes-main-m-0:66732:66732 [0] NCCL INFO Bootstrap timings total 22.624430 (create 0.000080, send 0.000134, recv 0.006920, ring 1.625886, delay 0.000001)
h-liuchonghan-translate-a8nodes-main-m-0:66734:66734 [2] NCCL INFO Bootstrap timings total 22.578677 (create 0.000070, send 0.000148, recv 0.000305, ring 22.577406, delay 0.000002)
h-liuchonghan-translate-a8nodes-main-m-0:66733:66733 [1] NCCL INFO Bootstrap timings total 22.617699 (create 0.000083, send 0.000139, recv 0.039232, ring 22.577355, delay 0.000002)
h-liuchonghan-translate-a8nodes-main-m-0:66738:66738 [6] NCCL INFO ncclTopoGetCpuAffinity: Affinity for GPU 6 is 64-127,192-254. (GPU affinity = 64-127,192-254 ; CPU affinity = 0-254).
h-liuchonghan-translate-a8nodes-main-m-0:66738:66738 [6] NCCL INFO NVLS multicast support is not available on dev 6 (NVLS_NCHANNELS 0)
h-liuchonghan-translate-a8nodes-main-m-0:66734:66734 [2] NCCL INFO ncclTopoGetCpuAffinity: Affinity for GPU 2 is 0-63,128-191. (GPU affinity = 0-63,128-191 ; CPU affinity = 0-254).
h-liuchonghan-translate-a8nodes-main-m-0:66734:66734 [2] NCCL INFO NVLS multicast support is not available on dev 2 (NVLS_NCHANNELS 0)
h-liuchonghan-translate-a8nodes-main-m-0:66733:66733 [1] NCCL INFO ncclTopoGetCpuAffinity: Affinity for GPU 1 is 0-63,128-191. (GPU affinity = 0-63,128-191 ; CPU affinity = 0-254).
h-liuchonghan-translate-a8nodes-main-m-0:66733:66733 [1] NCCL INFO NVLS multicast support is not available on dev 1 (NVLS_NCHANNELS 0)
h-liuchonghan-translate-a8nodes-main-m-0:66732:66732 [0] NCCL INFO ncclTopoGetCpuAffinity: Affinity for GPU 0 is 0-63,128-191. (GPU affinity = 0-63,128-191 ; CPU affinity = 0-254).
h-liuchonghan-translate-a8nodes-main-m-0:66739:66739 [7] NCCL INFO ncclTopoGetCpuAffinity: Affinity for GPU 7 is 64-127,192-254. (GPU affinity = 64-127,192-254 ; CPU affinity = 0-254).
h-liuchonghan-translate-a8nodes-main-m-0:66732:66732 [0] NCCL INFO NVLS multicast support is not available on dev 0 (NVLS_NCHANNELS 0)
h-liuchonghan-translate-a8nodes-main-m-0:66739:66739 [7] NCCL INFO NVLS multicast support is not available on dev 7 (NVLS_NCHANNELS 0)
h-liuchonghan-translate-a8nodes-main-m-0:66737:66737 [5] NCCL INFO ncclTopoGetCpuAffinity: Affinity for GPU 5 is 64-127,192-254. (GPU affinity = 64-127,192-254 ; CPU affinity = 0-254).
h-liuchonghan-translate-a8nodes-main-m-0:66737:66737 [5] NCCL INFO NVLS multicast support is not available on dev 5 (NVLS_NCHANNELS 0)
h-liuchonghan-translate-a8nodes-main-m-0:66736:66736 [4] NCCL INFO ncclTopoGetCpuAffinity: Affinity for GPU 4 is 64-127,192-254. (GPU affinity = 64-127,192-254 ; CPU affinity = 0-254).
h-liuchonghan-translate-a8nodes-main-m-0:66736:66736 [4] NCCL INFO NVLS multicast support is not available on dev 4 (NVLS_NCHANNELS 0)
h-liuchonghan-translate-a8nodes-main-m-0:66735:66735 [3] NCCL INFO ncclTopoGetCpuAffinity: Affinity for GPU 3 is 0-63,128-191. (GPU affinity = 0-63,128-191 ; CPU affinity = 0-254).
h-liuchonghan-translate-a8nodes-main-m-0:66735:66735 [3] NCCL INFO NVLS multicast support is not available on dev 3 (NVLS_NCHANNELS 0)
h-liuchonghan-translate-a8nodes-main-m-0:66732:66732 [0] NCCL INFO comm 0x55d1c3b89b70 rank 0 nRanks 32 nNodes 4 localRanks 8 localRank 0 MNNVL 0
h-liuchonghan-translate-a8nodes-main-m-0:66733:66733 [1] NCCL INFO comm 0x55a35ad250b0 rank 1 nRanks 32 nNodes 4 localRanks 8 localRank 1 MNNVL 0
h-liuchonghan-translate-a8nodes-main-m-0:66734:66734 [2] NCCL INFO comm 0x55985c7ad020 rank 2 nRanks 32 nNodes 4 localRanks 8 localRank 2 MNNVL 0
h-liuchonghan-translate-a8nodes-main-m-0:66735:66735 [3] NCCL INFO comm 0x559159d41560 rank 3 nRanks 32 nNodes 4 localRanks 8 localRank 3 MNNVL 0
h-liuchonghan-translate-a8nodes-main-m-0:66732:66732 [0] NCCL INFO Channel 00/02 :  0  7  6  5  4  3  2  1  8 15 14 13 12 11 10  9 16 23 22 21 20 19 18 17 24 31 30 29 28 27 26 25
h-liuchonghan-translate-a8nodes-main-m-0:66733:66733 [1] NCCL INFO Trees [0] 2/-1/-1->1->0 [1] 2/-1/-1->1->0
h-liuchonghan-translate-a8nodes-main-m-0:66732:66732 [0] NCCL INFO Channel 01/02 :  0  7  6  5  4  3  2  1  8 15 14 13 12 11 10  9 16 23 22 21 20 19 18 17 24 31 30 29 28 27 26 25
h-liuchonghan-translate-a8nodes-main-m-0:66733:66733 [1] NCCL INFO P2P Chunksize set to 131072
h-liuchonghan-translate-a8nodes-main-m-0:66732:66732 [0] NCCL INFO Trees [0] 1/16/-1->0->-1 [1] 1/-1/-1->0->8
h-liuchonghan-translate-a8nodes-main-m-0:66732:66732 [0] NCCL INFO P2P Chunksize set to 131072
h-liuchonghan-translate-a8nodes-main-m-0:66734:66734 [2] NCCL INFO Trees [0] 3/-1/-1->2->1 [1] 3/-1/-1->2->1
h-liuchonghan-translate-a8nodes-main-m-0:66734:66734 [2] NCCL INFO P2P Chunksize set to 131072
h-liuchonghan-translate-a8nodes-main-m-0:66735:66735 [3] NCCL INFO Trees [0] 4/-1/-1->3->2 [1] 4/-1/-1->3->2
h-liuchonghan-translate-a8nodes-main-m-0:66735:66735 [3] NCCL INFO P2P Chunksize set to 131072
h-liuchonghan-translate-a8nodes-main-m-0:66736:66736 [4] NCCL INFO comm 0x562913d20eb0 rank 4 nRanks 32 nNodes 4 localRanks 8 localRank 4 MNNVL 0
h-liuchonghan-translate-a8nodes-main-m-0:66737:66737 [5] NCCL INFO comm 0x55f6153f0f30 rank 5 nRanks 32 nNodes 4 localRanks 8 localRank 5 MNNVL 0
h-liuchonghan-translate-a8nodes-main-m-0:66736:66736 [4] NCCL INFO Trees [0] 5/-1/-1->4->3 [1] 5/-1/-1->4->3
h-liuchonghan-translate-a8nodes-main-m-0:66736:66736 [4] NCCL INFO P2P Chunksize set to 131072
h-liuchonghan-translate-a8nodes-main-m-0:66737:66737 [5] NCCL INFO Trees [0] 6/-1/-1->5->4 [1] 6/-1/-1->5->4
h-liuchonghan-translate-a8nodes-main-m-0:66737:66737 [5] NCCL INFO P2P Chunksize set to 131072
h-liuchonghan-translate-a8nodes-main-m-0:66738:66738 [6] NCCL INFO comm 0x55904d46bc60 rank 6 nRanks 32 nNodes 4 localRanks 8 localRank 6 MNNVL 0
h-liuchonghan-translate-a8nodes-main-m-0:66739:66739 [7] NCCL INFO comm 0x564f0f810dd0 rank 7 nRanks 32 nNodes 4 localRanks 8 localRank 7 MNNVL 0
h-liuchonghan-translate-a8nodes-main-m-0:66738:66738 [6] NCCL INFO Trees [0] 7/-1/-1->6->5 [1] 7/-1/-1->6->5
h-liuchonghan-translate-a8nodes-main-m-0:66738:66738 [6] NCCL INFO P2P Chunksize set to 131072
h-liuchonghan-translate-a8nodes-main-m-0:66739:66739 [7] NCCL INFO Trees [0] -1/-1/-1->7->6 [1] -1/-1/-1->7->6
h-liuchonghan-translate-a8nodes-main-m-0:66739:66739 [7] NCCL INFO P2P Chunksize set to 131072
h-liuchonghan-translate-a8nodes-main-m-0:66732:66732 [0] NCCL INFO PROFILER/Plugin: Could not find: libnccl-profiler.so
h-liuchonghan-translate-a8nodes-main-m-0:66732:66732 [0] NCCL INFO Check P2P Type isAllDirectP2p 1 directMode 0 isAllCudaP2p 1
h-liuchonghan-translate-a8nodes-main-m-0:66735:66735 [3] NCCL INFO PROFILER/Plugin: Could not find: libnccl-profiler.so
h-liuchonghan-translate-a8nodes-main-m-0:66735:66735 [3] NCCL INFO Check P2P Type isAllDirectP2p 1 directMode 0 isAllCudaP2p 1
h-liuchonghan-translate-a8nodes-main-m-0:66734:66734 [2] NCCL INFO PROFILER/Plugin: Could not find: libnccl-profiler.so
h-liuchonghan-translate-a8nodes-main-m-0:66734:66734 [2] NCCL INFO Check P2P Type isAllDirectP2p 1 directMode 0 isAllCudaP2p 1
h-liuchonghan-translate-a8nodes-main-m-0:66733:66733 [1] NCCL INFO PROFILER/Plugin: Could not find: libnccl-profiler.so
h-liuchonghan-translate-a8nodes-main-m-0:66733:66733 [1] NCCL INFO Check P2P Type isAllDirectP2p 1 directMode 0 isAllCudaP2p 1
h-liuchonghan-translate-a8nodes-main-m-0:66736:66736 [4] NCCL INFO PROFILER/Plugin: Could not find: libnccl-profiler.so
h-liuchonghan-translate-a8nodes-main-m-0:66732:66824 [0] NCCL INFO [Proxy Service] Device 0 CPU core 155
h-liuchonghan-translate-a8nodes-main-m-0:66736:66736 [4] NCCL INFO Check P2P Type isAllDirectP2p 1 directMode 0 isAllCudaP2p 1
h-liuchonghan-translate-a8nodes-main-m-0:66737:66737 [5] NCCL INFO PROFILER/Plugin: Could not find: libnccl-profiler.so
h-liuchonghan-translate-a8nodes-main-m-0:66737:66737 [5] NCCL INFO Check P2P Type isAllDirectP2p 1 directMode 0 isAllCudaP2p 1
h-liuchonghan-translate-a8nodes-main-m-0:66738:66738 [6] NCCL INFO PROFILER/Plugin: Could not find: libnccl-profiler.so
h-liuchonghan-translate-a8nodes-main-m-0:66738:66738 [6] NCCL INFO Check P2P Type isAllDirectP2p 1 directMode 0 isAllCudaP2p 1
h-liuchonghan-translate-a8nodes-main-m-0:66739:66739 [7] NCCL INFO PROFILER/Plugin: Could not find: libnccl-profiler.so
h-liuchonghan-translate-a8nodes-main-m-0:66739:66739 [7] NCCL INFO Check P2P Type isAllDirectP2p 1 directMode 0 isAllCudaP2p 1
h-liuchonghan-translate-a8nodes-main-m-0:66734:66827 [0] NCCL INFO [Proxy Service] Device 2 CPU core 2
h-liuchonghan-translate-a8nodes-main-m-0:66732:66826 [0] NCCL INFO [Proxy Service UDS] Device 0 CPU core 4
h-liuchonghan-translate-a8nodes-main-m-0:66734:66830 [0] NCCL INFO [Proxy Service UDS] Device 2 CPU core 22
h-liuchonghan-translate-a8nodes-main-m-0:66735:66828 [0] NCCL INFO [Proxy Service UDS] Device 3 CPU core 3
h-liuchonghan-translate-a8nodes-main-m-0:66738:66834 [0] NCCL INFO [Proxy Service] Device 6 CPU core 232
h-liuchonghan-translate-a8nodes-main-m-0:66735:66825 [0] NCCL INFO [Proxy Service] Device 3 CPU core 160
h-liuchonghan-translate-a8nodes-main-m-0:66733:66831 [0] NCCL INFO [Proxy Service UDS] Device 1 CPU core 165
h-liuchonghan-translate-a8nodes-main-m-0:66736:66832 [0] NCCL INFO [Proxy Service] Device 4 CPU core 66
h-liuchonghan-translate-a8nodes-main-m-0:66737:66833 [0] NCCL INFO [Proxy Service] Device 5 CPU core 193
h-liuchonghan-translate-a8nodes-main-m-0:66738:66838 [0] NCCL INFO [Proxy Service UDS] Device 6 CPU core 93
h-liuchonghan-translate-a8nodes-main-m-0:66736:66836 [0] NCCL INFO [Proxy Service UDS] Device 4 CPU core 199
h-liuchonghan-translate-a8nodes-main-m-0:66737:66835 [0] NCCL INFO [Proxy Service UDS] Device 5 CPU core 70
h-liuchonghan-translate-a8nodes-main-m-0:66733:66829 [0] NCCL INFO [Proxy Service] Device 1 CPU core 138
h-liuchonghan-translate-a8nodes-main-m-0:66739:66837 [0] NCCL INFO [Proxy Service] Device 7 CPU core 205
h-liuchonghan-translate-a8nodes-main-m-0:66739:66839 [0] NCCL INFO [Proxy Service UDS] Device 7 CPU core 96
h-liuchonghan-translate-a8nodes-main-m-0:66735:66735 [3] NCCL INFO TUNER/Plugin: Could not find: libnccl-tuner.so
h-liuchonghan-translate-a8nodes-main-m-0:66733:66733 [1] NCCL INFO TUNER/Plugin: Could not find: libnccl-tuner.so
h-liuchonghan-translate-a8nodes-main-m-0:66733:66733 [1] NCCL INFO threadThresholds 8/8/64 | 256/8/64 | 512 | 512
h-liuchonghan-translate-a8nodes-main-m-0:66733:66733 [1] NCCL INFO 2 coll channels, 2 collnet channels, 0 nvls channels, 2 p2p channels, 1 p2p channels per peer
h-liuchonghan-translate-a8nodes-main-m-0:66735:66735 [3] NCCL INFO threadThresholds 8/8/64 | 256/8/64 | 512 | 512
h-liuchonghan-translate-a8nodes-main-m-0:66733:66733 [1] NCCL INFO Symmetric VA size=80GB
h-liuchonghan-translate-a8nodes-main-m-0:66735:66735 [3] NCCL INFO 2 coll channels, 2 collnet channels, 0 nvls channels, 2 p2p channels, 1 p2p channels per peer
h-liuchonghan-translate-a8nodes-main-m-0:66733:66733 [1] NCCL INFO Symmetric memory is not supported. cuMemEnable 1, globalGinSupport 0, globalNicFused 0 cuMemGdrSupport 1
h-liuchonghan-translate-a8nodes-main-m-0:66735:66735 [3] NCCL INFO Symmetric VA size=80GB
h-liuchonghan-translate-a8nodes-main-m-0:66735:66735 [3] NCCL INFO Symmetric memory is not supported. cuMemEnable 1, globalGinSupport 0, globalNicFused 0 cuMemGdrSupport 1
h-liuchonghan-translate-a8nodes-main-m-0:66739:66739 [7] NCCL INFO TUNER/Plugin: Could not find: libnccl-tuner.so
h-liuchonghan-translate-a8nodes-main-m-0:66739:66739 [7] NCCL INFO threadThresholds 8/8/64 | 256/8/64 | 512 | 512
h-liuchonghan-translate-a8nodes-main-m-0:66739:66739 [7] NCCL INFO 2 coll channels, 2 collnet channels, 0 nvls channels, 2 p2p channels, 1 p2p channels per peer
h-liuchonghan-translate-a8nodes-main-m-0:66739:66739 [7] NCCL INFO Symmetric VA size=80GB
h-liuchonghan-translate-a8nodes-main-m-0:66739:66739 [7] NCCL INFO Symmetric memory is not supported. cuMemEnable 1, globalGinSupport 0, globalNicFused 0 cuMemGdrSupport 1
h-liuchonghan-translate-a8nodes-main-m-0:66734:66734 [2] NCCL INFO TUNER/Plugin: Could not find: libnccl-tuner.so
h-liuchonghan-translate-a8nodes-main-m-0:66734:66734 [2] NCCL INFO threadThresholds 8/8/64 | 256/8/64 | 512 | 512
h-liuchonghan-translate-a8nodes-main-m-0:66734:66734 [2] NCCL INFO 2 coll channels, 2 collnet channels, 0 nvls channels, 2 p2p channels, 1 p2p channels per peer
h-liuchonghan-translate-a8nodes-main-m-0:66734:66734 [2] NCCL INFO Symmetric VA size=80GB
h-liuchonghan-translate-a8nodes-main-m-0:66734:66734 [2] NCCL INFO Symmetric memory is not supported. cuMemEnable 1, globalGinSupport 0, globalNicFused 0 cuMemGdrSupport 1
h-liuchonghan-translate-a8nodes-main-m-0:66737:66737 [5] NCCL INFO TUNER/Plugin: Could not find: libnccl-tuner.so
h-liuchonghan-translate-a8nodes-main-m-0:66737:66737 [5] NCCL INFO threadThresholds 8/8/64 | 256/8/64 | 512 | 512
h-liuchonghan-translate-a8nodes-main-m-0:66737:66737 [5] NCCL INFO 2 coll channels, 2 collnet channels, 0 nvls channels, 2 p2p channels, 1 p2p channels per peer
h-liuchonghan-translate-a8nodes-main-m-0:66737:66737 [5] NCCL INFO Symmetric VA size=80GB
h-liuchonghan-translate-a8nodes-main-m-0:66737:66737 [5] NCCL INFO Symmetric memory is not supported. cuMemEnable 1, globalGinSupport 0, globalNicFused 0 cuMemGdrSupport 1
h-liuchonghan-translate-a8nodes-main-m-0:66732:66732 [0] NCCL INFO TUNER/Plugin: Could not find: libnccl-tuner.so
h-liuchonghan-translate-a8nodes-main-m-0:66732:66732 [0] NCCL INFO threadThresholds 8/8/64 | 256/8/64 | 512 | 512
h-liuchonghan-translate-a8nodes-main-m-0:66732:66732 [0] NCCL INFO 2 coll channels, 2 collnet channels, 0 nvls channels, 2 p2p channels, 1 p2p channels per peer
h-liuchonghan-translate-a8nodes-main-m-0:66732:66732 [0] NCCL INFO Symmetric VA size=80GB
h-liuchonghan-translate-a8nodes-main-m-0:66732:66732 [0] NCCL INFO Symmetric memory is not supported. cuMemEnable 1, globalGinSupport 0, globalNicFused 0 cuMemGdrSupport 1
h-liuchonghan-translate-a8nodes-main-m-0:66736:66736 [4] NCCL INFO TUNER/Plugin: Could not find: libnccl-tuner.so
h-liuchonghan-translate-a8nodes-main-m-0:66736:66736 [4] NCCL INFO threadThresholds 8/8/64 | 256/8/64 | 512 | 512
h-liuchonghan-translate-a8nodes-main-m-0:66736:66736 [4] NCCL INFO 2 coll channels, 2 collnet channels, 0 nvls channels, 2 p2p channels, 1 p2p channels per peer
h-liuchonghan-translate-a8nodes-main-m-0:66736:66736 [4] NCCL INFO Symmetric VA size=80GB
h-liuchonghan-translate-a8nodes-main-m-0:66736:66736 [4] NCCL INFO Symmetric memory is not supported. cuMemEnable 1, globalGinSupport 0, globalNicFused 0 cuMemGdrSupport 1
h-liuchonghan-translate-a8nodes-main-m-0:66738:66738 [6] NCCL INFO TUNER/Plugin: Could not find: libnccl-tuner.so
h-liuchonghan-translate-a8nodes-main-m-0:66738:66738 [6] NCCL INFO threadThresholds 8/8/64 | 256/8/64 | 512 | 512
h-liuchonghan-translate-a8nodes-main-m-0:66738:66738 [6] NCCL INFO 2 coll channels, 2 collnet channels, 0 nvls channels, 2 p2p channels, 1 p2p channels per peer
h-liuchonghan-translate-a8nodes-main-m-0:66738:66738 [6] NCCL INFO Symmetric VA size=80GB
h-liuchonghan-translate-a8nodes-main-m-0:66738:66738 [6] NCCL INFO Symmetric memory is not supported. cuMemEnable 1, globalGinSupport 0, globalNicFused 0 cuMemGdrSupport 1
h-liuchonghan-translate-a8nodes-main-m-0:66732:66732 [0] NCCL INFO CC Off, workFifoBytes 1048576
h-liuchonghan-translate-a8nodes-main-m-0:66737:66737 [5] NCCL INFO ncclCommInitRankConfig comm 0x55f6153f0f30 rank 5 nranks 32 cudaDev 5 nvmlDev 5 busId a1000 commId 0x7a425254b479deb7 - Init COMPLETE
h-liuchonghan-translate-a8nodes-main-m-0:66733:66733 [1] NCCL INFO ncclCommInitRankConfig comm 0x55a35ad250b0 rank 1 nranks 32 cudaDev 1 nvmlDev 1 busId 10000 commId 0x7a425254b479deb7 - Init COMPLETE
h-liuchonghan-translate-a8nodes-main-m-0:66737:66737 [5] NCCL INFO Init timings - ncclCommInitRankConfig: rank 5 nranks 32 total 23.13 (kernels 0.20, alloc 0.01, bootstrap 22.60, allgathers 0.08, topo 0.22, graphs 0.00, connections 0.01, rest 0.00)
h-liuchonghan-translate-a8nodes-main-m-0:66733:66733 [1] NCCL INFO Init timings - ncclCommInitRankConfig: rank 1 nranks 32 total 23.12 (kernels 0.17, alloc 0.01, bootstrap 22.62, allgathers 0.08, topo 0.22, graphs 0.00, connections 0.01, rest 0.01)
h-liuchonghan-translate-a8nodes-main-m-0:66739:66739 [7] NCCL INFO ncclCommInitRankConfig comm 0x564f0f810dd0 rank 7 nranks 32 cudaDev 7 nvmlDev 7 busId e0000 commId 0x7a425254b479deb7 - Init COMPLETE
h-liuchonghan-translate-a8nodes-main-m-0:66736:66736 [4] NCCL INFO ncclCommInitRankConfig comm 0x562913d20eb0 rank 4 nranks 32 cudaDev 4 nvmlDev 4 busId 9b000 commId 0x7a425254b479deb7 - Init COMPLETE
h-liuchonghan-translate-a8nodes-main-m-0:66735:66735 [3] NCCL INFO ncclCommInitRankConfig comm 0x559159d41560 rank 3 nranks 32 cudaDev 3 nvmlDev 3 busId 4e000 commId 0x7a425254b479deb7 - Init COMPLETE
h-liuchonghan-translate-a8nodes-main-m-0:66739:66739 [7] NCCL INFO Init timings - ncclCommInitRankConfig: rank 7 nranks 32 total 23.12 (kernels 0.15, alloc 0.02, bootstrap 22.63, allgathers 0.08, topo 0.22, graphs 0.00, connections 0.01, rest 0.01)
h-liuchonghan-translate-a8nodes-main-m-0:66732:66732 [0] NCCL INFO ncclCommInitRankConfig comm 0x55d1c3b89b70 rank 0 nranks 32 cudaDev 0 nvmlDev 0 busId b000 commId 0x7a425254b479deb7 - Init COMPLETE
h-liuchonghan-translate-a8nodes-main-m-0:66736:66736 [4] NCCL INFO Init timings - ncclCommInitRankConfig: rank 4 nranks 32 total 23.13 (kernels 0.17, alloc 0.02, bootstrap 22.62, allgathers 0.08, topo 0.22, graphs 0.00, connections 0.01, rest 0.00)
h-liuchonghan-translate-a8nodes-main-m-0:66732:66732 [0] NCCL INFO Init timings - ncclCommInitRankConfig: rank 0 nranks 32 total 23.14 (kernels 0.17, alloc 0.02, bootstrap 22.62, allgathers 0.08, topo 0.22, graphs 0.00, connections 0.01, rest 0.00)
h-liuchonghan-translate-a8nodes-main-m-0:66735:66735 [3] NCCL INFO Init timings - ncclCommInitRankConfig: rank 3 nranks 32 total 23.13 (kernels 0.19, alloc 0.01, bootstrap 22.61, allgathers 0.08, topo 0.22, graphs 0.00, connections 0.01, rest 0.00)
h-liuchonghan-translate-a8nodes-main-m-0:66734:66734 [2] NCCL INFO ncclCommInitRankConfig comm 0x55985c7ad020 rank 2 nranks 32 cudaDev 2 nvmlDev 2 busId 48000 commId 0x7a425254b479deb7 - Init COMPLETE
h-liuchonghan-translate-a8nodes-main-m-0:66738:66738 [6] NCCL INFO ncclCommInitRankConfig comm 0x55904d46bc60 rank 6 nranks 32 cudaDev 6 nvmlDev 6 busId db000 commId 0x7a425254b479deb7 - Init COMPLETE
h-liuchonghan-translate-a8nodes-main-m-0:66734:66734 [2] NCCL INFO Init timings - ncclCommInitRankConfig: rank 2 nranks 32 total 23.10 (kernels 0.19, alloc 0.01, bootstrap 22.58, allgathers 0.08, topo 0.22, graphs 0.00, connections 0.01, rest 0.01)
h-liuchonghan-translate-a8nodes-main-m-0:66738:66738 [6] NCCL INFO Init timings - ncclCommInitRankConfig: rank 6 nranks 32 total 23.10 (kernels 0.17, alloc 0.01, bootstrap 22.60, allgathers 0.09, topo 0.22, graphs 0.00, connections 0.02, rest 0.00)
h-liuchonghan-translate-a8nodes-main-m-0:66734:66734 [2] NCCL INFO Channel 00/0 : 2[2] -> 3[3] via P2P/CUMEM/read
h-liuchonghan-translate-a8nodes-main-m-0:66737:66737 [5] NCCL INFO Channel 00/0 : 5[5] -> 6[6] via P2P/CUMEM/read
h-liuchonghan-translate-a8nodes-main-m-0:66738:66738 [6] NCCL INFO Channel 00/0 : 6[6] -> 7[7] via P2P/CUMEM/read
h-liuchonghan-translate-a8nodes-main-m-0:66734:66734 [2] NCCL INFO Channel 01/0 : 2[2] -> 3[3] via P2P/CUMEM/read
h-liuchonghan-translate-a8nodes-main-m-0:66733:66733 [1] NCCL INFO Channel 00/0 : 1[1] -> 2[2] via P2P/CUMEM/read
h-liuchonghan-translate-a8nodes-main-m-0:66737:66737 [5] NCCL INFO Channel 01/0 : 5[5] -> 6[6] via P2P/CUMEM/read
h-liuchonghan-translate-a8nodes-main-m-0:66733:66733 [1] NCCL INFO Channel 01/0 : 1[1] -> 2[2] via P2P/CUMEM/read
h-liuchonghan-translate-a8nodes-main-m-0:66738:66738 [6] NCCL INFO Channel 01/0 : 6[6] -> 7[7] via P2P/CUMEM/read
h-liuchonghan-translate-a8nodes-main-m-0:66739:66739 [7] NCCL INFO Channel 00/0 : 7[7] -> 6[6] via P2P/CUMEM/read
h-liuchonghan-translate-a8nodes-main-m-0:66732:66732 [0] NCCL INFO Channel 00/0 : 0[0] -> 1[1] via P2P/CUMEM/read
h-liuchonghan-translate-a8nodes-main-m-0:66739:66739 [7] NCCL INFO Channel 01/0 : 7[7] -> 6[6] via P2P/CUMEM/read
h-liuchonghan-translate-a8nodes-main-m-0:66735:66735 [3] NCCL INFO Channel 00/0 : 3[3] -> 4[4] via P2P/CUMEM/read
h-liuchonghan-translate-a8nodes-main-m-0:66732:66732 [0] NCCL INFO Channel 01/0 : 0[0] -> 1[1] via P2P/CUMEM/read
h-liuchonghan-translate-a8nodes-main-m-0:66738:66738 [6] NCCL INFO Channel 00/0 : 6[6] -> 5[5] via P2P/CUMEM/read
h-liuchonghan-translate-a8nodes-main-m-0:66736:66736 [4] NCCL INFO Channel 00/0 : 4[4] -> 5[5] via P2P/CUMEM/read
h-liuchonghan-translate-a8nodes-main-m-0:66735:66735 [3] NCCL INFO Channel 01/0 : 3[3] -> 4[4] via P2P/CUMEM/read
h-liuchonghan-translate-a8nodes-main-m-0:66738:66738 [6] NCCL INFO Channel 01/0 : 6[6] -> 5[5] via P2P/CUMEM/read
h-liuchonghan-translate-a8nodes-main-m-0:66736:66736 [4] NCCL INFO Channel 01/0 : 4[4] -> 5[5] via P2P/CUMEM/read
h-liuchonghan-translate-a8nodes-main-m-0:66733:66733 [1] NCCL INFO Channel 00/0 : 1[1] -> 0[0] via P2P/CUMEM/read
h-liuchonghan-translate-a8nodes-main-m-0:66733:66733 [1] NCCL INFO Channel 01/0 : 1[1] -> 0[0] via P2P/CUMEM/read
h-liuchonghan-translate-a8nodes-main-m-0:66734:66734 [2] NCCL INFO Channel 00/0 : 2[2] -> 1[1] via P2P/CUMEM/read
h-liuchonghan-translate-a8nodes-main-m-0:66735:66735 [3] NCCL INFO Channel 00/0 : 3[3] -> 2[2] via P2P/CUMEM/read
h-liuchonghan-translate-a8nodes-main-m-0:66737:66737 [5] NCCL INFO Channel 00/0 : 5[5] -> 4[4] via P2P/CUMEM/read
h-liuchonghan-translate-a8nodes-main-m-0:66736:66736 [4] NCCL INFO Channel 00/0 : 4[4] -> 3[3] via P2P/CUMEM/read
h-liuchonghan-translate-a8nodes-main-m-0:66734:66734 [2] NCCL INFO Channel 01/0 : 2[2] -> 1[1] via P2P/CUMEM/read
h-liuchonghan-translate-a8nodes-main-m-0:66737:66737 [5] NCCL INFO Channel 01/0 : 5[5] -> 4[4] via P2P/CUMEM/read
h-liuchonghan-translate-a8nodes-main-m-0:66735:66735 [3] NCCL INFO Channel 01/0 : 3[3] -> 2[2] via P2P/CUMEM/read
h-liuchonghan-translate-a8nodes-main-m-0:66736:66736 [4] NCCL INFO Channel 01/0 : 4[4] -> 3[3] via P2P/CUMEM/read
h-liuchonghan-translate-a8nodes-main-m-0:66732:66840 [0] NCCL INFO [Proxy Progress] Device 0 CPU core 8
h-liuchonghan-translate-a8nodes-main-m-0:66732:66732 [0] NCCL INFO Channel 01/0 : 0[0] -> 8[0] [send] via NET/Socket/0
h-liuchonghan-translate-a8nodes-main-m-0:66732:66732 [0] NCCL INFO Channel 00/0 : 16[0] -> 0[0] [receive] via NET/Socket/0
h-liuchonghan-translate-a8nodes-main-m-0:66732:66732 [0] NCCL INFO Channel 00/0 : 0[0] -> 16[0] [send] via NET/Socket/0
h-liuchonghan-translate-a8nodes-main-m-0:66732:66732 [0] NCCL INFO Channel 01/0 : 8[0] -> 0[0] [receive] via NET/Socket/0
h-liuchonghan-translate-a8nodes-main-m-0:66736:66736 [4] NCCL INFO Connected all trees
h-liuchonghan-translate-a8nodes-main-m-0:66737:66737 [5] NCCL INFO Connected all trees
h-liuchonghan-translate-a8nodes-main-m-0:66738:66738 [6] NCCL INFO Connected all trees
h-liuchonghan-translate-a8nodes-main-m-0:66739:66739 [7] NCCL INFO Connected all trees
h-liuchonghan-translate-a8nodes-main-m-0:66735:66735 [3] NCCL INFO Connected all trees
h-liuchonghan-translate-a8nodes-main-m-0:66734:66734 [2] NCCL INFO Connected all trees
h-liuchonghan-translate-a8nodes-main-m-0:66732:66732 [0] NCCL INFO Connected all trees
h-liuchonghan-translate-a8nodes-main-m-0:66733:66733 [1] NCCL INFO Connected all trees
2026-06-18 14:51:42 | INFO | nemo_automodel.components.loggers.log_utils | Setting logging level to 20
[transformers] `torch_dtype` is deprecated! Use `dtype` instead!

Loading weights:   0%|          | 0/399 [00:00<?, ?it/s]
Loading weights:   0%|          | 0/399 [00:00<?, ?it/s]
Loading weights:   0%|          | 0/399 [00:00<?, ?it/s]
Loading weights:   0%|          | 0/399 [00:00<?, ?it/s]
Loading weights:   0%|          | 0/399 [00:00<?, ?it/s]
Loading weights:   0%|          | 0/399 [00:00<?, ?it/s]
Loading weights:   0%|          | 0/399 [00:00<?, ?it/s]
Loading weights:   0%|          | 0/399 [00:00<?, ?it/s]
Loading weights: 100%|██████████| 399/399 [00:00<00:00, 4048.52it/s]

Loading weights:  94%|█████████▍| 376/399 [00:00<00:00, 3758.61it/s]
Loading weights: 100%|██████████| 399/399 [00:00<00:00, 3949.12it/s]

Loading weights: 100%|██████████| 399/399 [00:00<00:00, 4326.75it/s]

Loading weights:  78%|███████▊  | 313/399 [00:00<00:00, 3129.39it/s]
Loading weights: 100%|██████████| 399/399 [00:00<00:00, 3864.30it/s]

Loading weights: 100%|██████████| 399/399 [00:00<00:00, 4034.67it/s]

Loading weights: 100%|██████████| 399/399 [00:00<00:00, 4258.24it/s]

Loading weights:  69%|██████▊   | 274/399 [00:00<00:00, 2738.11it/s]
Loading weights: 100%|██████████| 399/399 [00:00<00:00, 3797.20it/s]

Loading weights:  71%|███████   | 283/399 [00:00<00:00, 2828.55it/s]
Loading weights: 100%|██████████| 399/399 [00:00<00:00, 3724.31it/s]
2026-06-18 14:51:43 | WARNING | nemo_automodel._transformers.kernel_patches | Asked to use Liger Kernel, but could not import
2026-06-18 14:51:43 | INFO | nemo_automodel._transformers.kernel_patches | Patched model with SDPA method= [<SDPBackend.CUDNN_ATTENTION: 3>, <SDPBackend.FLASH_ATTENTION: 1>, <SDPBackend.EFFICIENT_ATTENTION: 2>, <SDPBackend.MATH: 0>]
2026-06-18 14:52:35 | INFO | root | load_model: 15.26 GB loaded in 51.70s (0.30 GB/s overall | disk read 0.15s, distribute 51.55s)
2026-06-18 14:52:36 | INFO | root | Model summary:
2026-06-18 14:52:36 | INFO | root | --------------------------------
2026-06-18 14:52:36 | INFO | root | Trainable parameters: 8,190,735,360
2026-06-18 14:52:36 | INFO | root | Total parameters: 8,190,735,360
2026-06-18 14:52:36 | INFO | root | Trainable parameters percentage: 100.00%
2026-06-18 14:52:36 | INFO | root | Param L2 norm: 2346.2788
2026-06-18 14:52:36 | INFO | root | --------------------------------
2026-06-18 14:52:36 | INFO | nemo_automodel.components.datasets.llm.chat_dataset | Loading local dataset from /llm-align/liuchonghan/Automodel_lao/cache/dataset/perfectblend-qwen3-8b-regen-messages via load_dataset
h-liuchonghan-translate-a8nodes-main-m-0:66732:66732 [0] NCCL INFO Channel 00/0 : 25[1] -> 0[0] [receive] via NET/Socket/0
h-liuchonghan-translate-a8nodes-main-m-0:66732:66732 [0] NCCL INFO Channel 01/0 : 25[1] -> 0[0] [receive] via NET/Socket/0
h-liuchonghan-translate-a8nodes-main-m-0:66732:66732 [0] NCCL INFO Channel 00/0 : 0[0] -> 7[7] via P2P/CUMEM/read
h-liuchonghan-translate-a8nodes-main-m-0:66732:66732 [0] NCCL INFO Channel 01/0 : 0[0] -> 7[7] via P2P/CUMEM/read
h-liuchonghan-translate-a8nodes-main-m-0:66733:67182 [0] NCCL INFO [Proxy Progress] Device 1 CPU core 186
h-liuchonghan-translate-a8nodes-main-m-0:66733:66733 [1] NCCL INFO Channel 00/0 : 1[1] -> 8[0] [send] via NET/Socket/0
h-liuchonghan-translate-a8nodes-main-m-0:66733:66733 [1] NCCL INFO Channel 01/0 : 1[1] -> 8[0] [send] via NET/Socket/0
h-liuchonghan-translate-a8nodes-main-m-0:66732:66732 [0] NCCL INFO Connected all rings, use ring PXN 0 GDR 0
h-liuchonghan-translate-a8nodes-main-m-0:66733:66733 [1] NCCL INFO Connected all rings, use ring PXN 0 GDR 0
h-liuchonghan-translate-a8nodes-main-m-0:66735:66735 [3] NCCL INFO Connected all rings, use ring PXN 0 GDR 0
h-liuchonghan-translate-a8nodes-main-m-0:66734:66734 [2] NCCL INFO Connected all rings, use ring PXN 0 GDR 0
h-liuchonghan-translate-a8nodes-main-m-0:66736:66736 [4] NCCL INFO Connected all rings, use ring PXN 0 GDR 0
h-liuchonghan-translate-a8nodes-main-m-0:66737:66737 [5] NCCL INFO Connected all rings, use ring PXN 0 GDR 0
h-liuchonghan-translate-a8nodes-main-m-0:66739:66739 [7] NCCL INFO Connected all rings, use ring PXN 0 GDR 0
h-liuchonghan-translate-a8nodes-main-m-0:66738:66738 [6] NCCL INFO Connected all rings, use ring PXN 0 GDR 0
2026-06-18 14:52:55 | INFO | nemo_automodel.components.datasets.llm.eagle3 | Loaded EAGLE-3 draft vocab (32000 ids) from cache /llm-align/liuchonghan/Automodel_lao/outputs/peagle_qwen3_8b_perfectblend/draft_token_map_151936.pt
2026-06-18 14:53:16 | INFO | root | Draft summary:
2026-06-18 14:53:16 | INFO | root | --------------------------------
2026-06-18 14:53:16 | INFO | root | Trainable parameters: 2,003,374,080
2026-06-18 14:53:16 | INFO | root | Total parameters: 2,003,374,080
2026-06-18 14:53:16 | INFO | root | Trainable parameters percentage: 100.00%
2026-06-18 14:53:16 | INFO | root | Param L2 norm: 964.2572
2026-06-18 14:53:16 | INFO | root | --------------------------------
2026-06-18 14:53:26 | WARNING | root | save_consolidated=every but v4_compatible=False; checkpoint assets may be not compatible with transformers v4; [experimental] set --checkpoint.v4_compatible=True to enable
2026-06-18 14:53:26 | WARNING | nemo_automodel.components.checkpoint.checkpointing | checkpoint.save_consolidated=every exports HuggingFace safetensors during every checkpoint save and can leave GPUs idle during consolidation and filesystem writes. Recommended: checkpoint.save_consolidated=final, or checkpoint.save_consolidated=false and run bash <checkpoint>/model/consolidate.sh after training.
2026-06-18 14:53:27 | INFO | __main__ | Training start: start_epoch=0 num_epochs=2 batches_per_epoch=44326 grad_accum=8 log_every=50 total_optim_steps=11082 warmup_steps=166 peak_lr=1.000e-04 min_lr_ratio=0.1

Training:   0%|          | 0/11082 [00:00<?, ?step/s]
Training:   0%|          | 1/11082 [00:37<116:19:53, 37.79s/step]
Training:   0%|          | 2/11082 [00:41<55:25:58, 18.01s/step] 
Training:   0%|          | 3/11082 [00:53<46:46:33, 15.20s/step]
Training:   0%|          | 4/11082 [01:05<42:37:57, 13.85s/step]
Training:   0%|          | 5/11082 [01:16<39:13:17, 12.75s/step]
Training:   0%|          | 6/11082 [01:28<38:14:23, 12.43s/step]
Training:   0%|          | 7/11082 [01:39<37:33:23, 12.21s/step]
Training:   0%|          | 8/11082 [01:51<36:58:28, 12.02s/step]
Training:   0%|          | 9/11082 [02:03<36:30:16, 11.87s/step]
Training:   0%|          | 10/11082 [02:15<37:07:08, 12.07s/step]
Training:   0%|          | 11/11082 [02:27<36:53:34, 12.00s/step]
Training:   0%|          | 12/11082 [02:38<36:21:42, 11.82s/step]
Training:   0%|          | 13/11082 [02:50<35:43:05, 11.62s/step]
Training:   0%|          | 14/11082 [03:01<35:34:31, 11.57s/step]
Training:   0%|          | 15/11082 [03:13<35:34:38, 11.57s/step]
Training:   0%|          | 16/11082 [03:24<35:42:21, 11.62s/step]
Training:   0%|          | 17/11082 [03:36<35:47:12, 11.64s/step]
Training:   0%|          | 18/11082 [03:48<35:53:47, 11.68s/step]
Training:   0%|          | 19/11082 [03:59<35:34:21, 11.58s/step]
Training:   0%|          | 20/11082 [04:11<35:31:23, 11.56s/step]
Training:   0%|          | 21/11082 [04:22<35:30:24, 11.56s/step]
Training:   0%|          | 22/11082 [04:34<35:49:14, 11.66s/step]
Training:   0%|          | 23/11082 [04:45<35:21:56, 11.51s/step]
Training:   0%|          | 24/11082 [04:57<35:34:07, 11.58s/step]
Training:   0%|          | 25/11082 [05:09<36:06:08, 11.75s/step]
Training:   0%|          | 26/11082 [05:20<35:37:30, 11.60s/step]
Training:   0%|          | 27/11082 [05:33<36:24:32, 11.86s/step]
Training:   0%|          | 28/11082 [05:44<35:57:11, 11.71s/step]
Training:   0%|          | 29/11082 [05:56<35:44:05, 11.64s/step]
Training:   0%|          | 30/11082 [06:07<35:48:26, 11.66s/step]
Training:   0%|          | 31/11082 [06:19<35:35:31, 11.59s/step]
Training:   0%|          | 32/11082 [06:31<35:56:04, 11.71s/step]
Training:   0%|          | 33/11082 [06:42<35:43:47, 11.64s/step]
Training:   0%|          | 34/11082 [06:53<35:15:40, 11.49s/step]
Training:   0%|          | 35/11082 [07:06<35:51:59, 11.69s/step]
Training:   0%|          | 36/11082 [07:17<35:40:44, 11.63s/step]
Training:   0%|          | 37/11082 [07:28<35:28:47, 11.56s/step]
Training:   0%|          | 38/11082 [07:40<35:29:21, 11.57s/step]
Training:   0%|          | 39/11082 [07:52<35:24:48, 11.54s/step]
Training:   0%|          | 40/11082 [08:03<35:11:22, 11.47s/step]
Training:   0%|          | 41/11082 [08:14<35:20:59, 11.53s/step]
Training:   0%|          | 42/11082 [08:26<35:01:11, 11.42s/step]
Training:   0%|          | 43/11082 [08:38<35:29:52, 11.58s/step]
Training:   0%|          | 44/11082 [08:49<35:40:08, 11.63s/step]
Training:   0%|          | 45/11082 [09:01<35:19:40, 11.52s/step]
Training:   0%|          | 46/11082 [09:12<35:18:54, 11.52s/step]
Training:   0%|          | 47/11082 [09:24<35:25:02, 11.55s/step]
Training:   0%|          | 48/11082 [09:35<35:32:22, 11.60s/step]
Training:   0%|          | 49/11082 [09:47<35:38:07, 11.63s/step]
Training:   0%|          | 50/11082 [09:58<35:04:48, 11.45s/step]
Training:   0%|          | 50/11082 [10:06<35:04:48, 11.45s/step, acc=0.0346, loss=8.8541, lr=3.07e-05]2026-06-18 15:03:34 | INFO | __main__ | epoch=0 step=50 train_loss=8.854098 train_acc=0.034634 lr=3.072e-05 grad_norm=2.0469 acc_d0=0.0212 acc_d1=0.0374 acc_d2=0.0374 acc_d3=0.0376 acc_d4=0.0376 acc_d5=0.0374 acc_d6=0.0374 acc_d7=0.0371

Training:   0%|          | 51/11082 [10:10<35:31:07, 11.59s/step, acc=0.0346, loss=8.8541, lr=3.07e-05]
Training:   0%|          | 52/11082 [10:22<36:06:23, 11.78s/step, acc=0.0346, loss=8.8541, lr=3.07e-05]
Training:   0%|          | 53/11082 [10:34<36:04:50, 11.78s/step, acc=0.0346, loss=8.8541, lr=3.07e-05]
Training:   0%|          | 54/11082 [10:46<35:45:09, 11.67s/step, acc=0.0346, loss=8.8541, lr=3.07e-05]
Training:   0%|          | 55/11082 [10:57<35:43:08, 11.66s/step, acc=0.0346, loss=8.8541, lr=3.07e-05]
Training:   1%|          | 56/11082 [11:09<35:46:24, 11.68s/step, acc=0.0346, loss=8.8541, lr=3.07e-05]
Training:   1%|          | 57/11082 [11:21<35:41:49, 11.66s/step, acc=0.0346, loss=8.8541, lr=3.07e-05]
Training:   1%|          | 58/11082 [11:32<35:31:27, 11.60s/step, acc=0.0346, loss=8.8541, lr=3.07e-05]
Training:   1%|          | 59/11082 [11:44<35:38:33, 11.64s/step, acc=0.0346, loss=8.8541, lr=3.07e-05]
Training:   1%|          | 60/11082 [11:55<35:20:26, 11.54s/step, acc=0.0346, loss=8.8541, lr=3.07e-05]
Training:   1%|          | 61/11082 [12:07<35:30:48, 11.60s/step, acc=0.0346, loss=8.8541, lr=3.07e-05]
Training:   1%|          | 62/11082 [12:19<35:44:24, 11.68s/step, acc=0.0346, loss=8.8541, lr=3.07e-05]
Training:   1%|          | 63/11082 [12:31<36:22:53, 11.89s/step, acc=0.0346, loss=8.8541, lr=3.07e-05]
Training:   1%|          | 64/11082 [12:42<35:51:46, 11.72s/step, acc=0.0346, loss=8.8541, lr=3.07e-05]
Training:   1%|          | 65/11082 [12:54<35:35:11, 11.63s/step, acc=0.0346, loss=8.8541, lr=3.07e-05]
Training:   1%|          | 66/11082 [13:05<35:14:41, 11.52s/step, acc=0.0346, loss=8.8541, lr=3.07e-05]
Training:   1%|          | 67/11082 [13:17<35:25:35, 11.58s/step, acc=0.0346, loss=8.8541, lr=3.07e-05]
Training:   1%|          | 68/11082 [13:28<35:12:14, 11.51s/step, acc=0.0346, loss=8.8541, lr=3.07e-05]
Training:   1%|          | 69/11082 [13:40<35:22:15, 11.56s/step, acc=0.0346, loss=8.8541, lr=3.07e-05]
Training:   1%|          | 70/11082 [13:51<35:21:33, 11.56s/step, acc=0.0346, loss=8.8541, lr=3.07e-05]
Training:   1%|          | 71/11082 [14:03<35:03:10, 11.46s/step, acc=0.0346, loss=8.8541, lr=3.07e-05]
Training:   1%|          | 72/11082 [14:14<35:25:19, 11.58s/step, acc=0.0346, loss=8.8541, lr=3.07e-05]
Training:   1%|          | 73/11082 [14:26<35:25:28, 11.58s/step, acc=0.0346, loss=8.8541, lr=3.07e-05]
Training:   1%|          | 74/11082 [14:38<35:26:49, 11.59s/step, acc=0.0346, loss=8.8541, lr=3.07e-05]
Training:   1%|          | 75/11082 [14:49<35:31:35, 11.62s/step, acc=0.0346, loss=8.8541, lr=3.07e-05]
Training:   1%|          | 76/11082 [15:00<34:58:21, 11.44s/step, acc=0.0346, loss=8.8541, lr=3.07e-05]
Training:   1%|          | 77/11082 [15:12<35:23:50, 11.58s/step, acc=0.0346, loss=8.8541, lr=3.07e-05]
Training:   1%|          | 78/11082 [15:24<35:09:37, 11.50s/step, acc=0.0346, loss=8.8541, lr=3.07e-05]
Training:   1%|          | 79/11082 [15:35<35:22:01, 11.57s/step, acc=0.0346, loss=8.8541, lr=3.07e-05]
Training:   1%|          | 80/11082 [15:47<35:24:47, 11.59s/step, acc=0.0346, loss=8.8541, lr=3.07e-05]
Training:   1%|          | 81/11082 [15:58<35:15:21, 11.54s/step, acc=0.0346, loss=8.8541, lr=3.07e-05]
Training:   1%|          | 82/11082 [16:10<35:19:52, 11.56s/step, acc=0.0346, loss=8.8541, lr=3.07e-05]
Training:   1%|          | 83/11082 [16:22<35:30:24, 11.62s/step, acc=0.0346, loss=8.8541, lr=3.07e-05]
Training:   1%|          | 84/11082 [16:33<35:00:12, 11.46s/step, acc=0.0346, loss=8.8541, lr=3.07e-05]
Training:   1%|          | 85/11082 [16:44<35:13:56, 11.53s/step, acc=0.0346, loss=8.8541, lr=3.07e-05]
Training:   1%|          | 86/11082 [16:56<35:05:31, 11.49s/step, acc=0.0346, loss=8.8541, lr=3.07e-05]
Training:   1%|          | 87/11082 [17:08<35:25:06, 11.60s/step, acc=0.0346, loss=8.8541, lr=3.07e-05]
Training:   1%|          | 88/11082 [17:19<35:21:09, 11.58s/step, acc=0.0346, loss=8.8541, lr=3.07e-05]
Training:   1%|          | 89/11082 [17:31<35:17:12, 11.56s/step, acc=0.0346, loss=8.8541, lr=3.07e-05]
Training:   1%|          | 90/11082 [17:42<34:58:29, 11.45s/step, acc=0.0346, loss=8.8541, lr=3.07e-05]
Training:   1%|          | 91/11082 [17:54<35:30:29, 11.63s/step, acc=0.0346, loss=8.8541, lr=3.07e-05]
Training:   1%|          | 92/11082 [18:05<34:52:43, 11.43s/step, acc=0.0346, loss=8.8541, lr=3.07e-05]
Training:   1%|          | 93/11082 [18:17<35:06:33, 11.50s/step, acc=0.0346, loss=8.8541, lr=3.07e-05]
Training:   1%|          | 94/11082 [18:28<35:12:16, 11.53s/step, acc=0.0346, loss=8.8541, lr=3.07e-05]
Training:   1%|          | 95/11082 [18:40<35:09:39, 11.52s/step, acc=0.0346, loss=8.8541, lr=3.07e-05]
Training:   1%|          | 96/11082 [18:51<35:16:13, 11.56s/step, acc=0.0346, loss=8.8541, lr=3.07e-05]
Training:   1%|          | 97/11082 [19:03<35:46:39, 11.73s/step, acc=0.0346, loss=8.8541, lr=3.07e-05]
Training:   1%|          | 98/11082 [19:15<35:26:35, 11.62s/step, acc=0.0346, loss=8.8541, lr=3.07e-05]
Training:   1%|          | 99/11082 [19:27<35:41:02, 11.70s/step, acc=0.0346, loss=8.8541, lr=3.07e-05]
Training:   1%|          | 100/11082 [19:38<34:59:52, 11.47s/step, acc=0.0346, loss=8.8541, lr=3.07e-05]
Training:   1%|          | 100/11082 [19:45<34:59:52, 11.47s/step, acc=0.1041, loss=5.9822, lr=6.08e-05]2026-06-18 15:13:13 | INFO | __main__ | epoch=0 step=100 train_loss=5.982246 train_acc=0.104147 lr=6.084e-05 grad_norm=2.5156 acc_d0=0.1952 acc_d1=0.0820 acc_d2=0.0780 acc_d3=0.0737 acc_d4=0.0714 acc_d5=0.0703 acc_d6=0.0684 acc_d7=0.0683

Training:   1%|          | 101/11082 [19:50<35:21:22, 11.59s/step, acc=0.1041, loss=5.9822, lr=6.08e-05]
Training:   1%|          | 102/11082 [20:01<35:26:48, 11.62s/step, acc=0.1041, loss=5.9822, lr=6.08e-05]
Training:   1%|          | 103/11082 [20:13<35:20:27, 11.59s/step, acc=0.1041, loss=5.9822, lr=6.08e-05]
Training:   1%|          | 104/11082 [20:24<35:07:36, 11.52s/step, acc=0.1041, loss=5.9822, lr=6.08e-05]
Training:   1%|          | 105/11082 [20:36<35:21:05, 11.59s/step, acc=0.1041, loss=5.9822, lr=6.08e-05]
Training:   1%|          | 106/11082 [20:47<35:18:02, 11.58s/step, acc=0.1041, loss=5.9822, lr=6.08e-05]
Training:   1%|          | 107/11082 [20:59<35:09:43, 11.53s/step, acc=0.1041, loss=5.9822, lr=6.08e-05]
Training:   1%|          | 108/11082 [21:10<34:44:45, 11.40s/step, acc=0.1041, loss=5.9822, lr=6.08e-05]
Training:   1%|          | 109/11082 [21:22<35:31:22, 11.65s/step, acc=0.1041, loss=5.9822, lr=6.08e-05]
Training:   1%|          | 110/11082 [21:33<35:09:48, 11.54s/step, acc=0.1041, loss=5.9822, lr=6.08e-05]
Training:   1%|          | 111/11082 [21:45<34:59:45, 11.48s/step, acc=0.1041, loss=5.9822, lr=6.08e-05]
Training:   1%|          | 112/11082 [21:56<34:49:59, 11.43s/step, acc=0.1041, loss=5.9822, lr=6.08e-05]
Training:   1%|          | 113/11082 [22:08<35:24:36, 11.62s/step, acc=0.1041, loss=5.9822, lr=6.08e-05]
Training:   1%|          | 114/11082 [22:20<35:16:15, 11.58s/step, acc=0.1041, loss=5.9822, lr=6.08e-05]
Training:   1%|          | 115/11082 [22:31<35:16:24, 11.58s/step, acc=0.1041, loss=5.9822, lr=6.08e-05]
Training:   1%|          | 116/11082 [22:43<35:02:12, 11.50s/step, acc=0.1041, loss=5.9822, lr=6.08e-05]
Training:   1%|          | 117/11082 [22:54<34:56:28, 11.47s/step, acc=0.1041, loss=5.9822, lr=6.08e-05]
Training:   1%|          | 118/11082 [23:06<35:24:32, 11.63s/step, acc=0.1041, loss=5.9822, lr=6.08e-05]
Training:   1%|          | 119/11082 [23:17<35:02:01, 11.50s/step, acc=0.1041, loss=5.9822, lr=6.08e-05]
Training:   1%|          | 120/11082 [23:29<35:02:10, 11.51s/step, acc=0.1041, loss=5.9822, lr=6.08e-05]
Training:   1%|          | 121/11082 [23:41<35:30:40, 11.66s/step, acc=0.1041, loss=5.9822, lr=6.08e-05]
Training:   1%|          | 122/11082 [23:52<35:18:30, 11.60s/step, acc=0.1041, loss=5.9822, lr=6.08e-05]
Training:   1%|          | 123/11082 [24:04<35:05:00, 11.52s/step, acc=0.1041, loss=5.9822, lr=6.08e-05]
Training:   1%|          | 124/11082 [24:15<35:09:18, 11.55s/step, acc=0.1041, loss=5.9822, lr=6.08e-05]
Training:   1%|          | 125/11082 [24:27<35:10:04, 11.55s/step, acc=0.1041, loss=5.9822, lr=6.08e-05]
Training:   1%|          | 126/11082 [24:38<35:11:45, 11.56s/step, acc=0.1041, loss=5.9822, lr=6.08e-05]
Training:   1%|          | 127/11082 [24:49<34:48:14, 11.44s/step, acc=0.1041, loss=5.9822, lr=6.08e-05]
Training:   1%|          | 128/11082 [25:01<34:52:49, 11.46s/step, acc=0.1041, loss=5.9822, lr=6.08e-05]
Training:   1%|          | 129/11082 [25:13<35:18:42, 11.61s/step, acc=0.1041, loss=5.9822, lr=6.08e-05]
Training:   1%|          | 130/11082 [25:24<34:52:22, 11.46s/step, acc=0.1041, loss=5.9822, lr=6.08e-05]
Training:   1%|          | 131/11082 [25:36<35:28:14, 11.66s/step, acc=0.1041, loss=5.9822, lr=6.08e-05]
Training:   1%|          | 132/11082 [25:48<35:42:24, 11.74s/step, acc=0.1041, loss=5.9822, lr=6.08e-05]
Training:   1%|          | 133/11082 [26:00<35:39:58, 11.73s/step, acc=0.1041, loss=5.9822, lr=6.08e-05]
Training:   1%|          | 134/11082 [26:12<36:01:29, 11.85s/step, acc=0.1041, loss=5.9822, lr=6.08e-05]
Training:   1%|          | 135/11082 [26:23<35:19:14, 11.62s/step, acc=0.1041, loss=5.9822, lr=6.08e-05]
Training:   1%|          | 136/11082 [26:35<35:20:06, 11.62s/step, acc=0.1041, loss=5.9822, lr=6.08e-05]
Training:   1%|          | 137/11082 [26:46<35:28:36, 11.67s/step, acc=0.1041, loss=5.9822, lr=6.08e-05]
Training:   1%|          | 138/11082 [26:58<35:22:53, 11.64s/step, acc=0.1041, loss=5.9822, lr=6.08e-05]
Training:   1%|▏         | 139/11082 [27:09<35:18:27, 11.62s/step, acc=0.1041, loss=5.9822, lr=6.08e-05]
Training:   1%|▏         | 140/11082 [27:21<35:08:07, 11.56s/step, acc=0.1041, loss=5.9822, lr=6.08e-05]
Training:   1%|▏         | 141/11082 [27:32<34:47:31, 11.45s/step, acc=0.1041, loss=5.9822, lr=6.08e-05]
Training:   1%|▏         | 142/11082 [27:44<35:11:44, 11.58s/step, acc=0.1041, loss=5.9822, lr=6.08e-05]
Training:   1%|▏         | 143/11082 [27:56<35:15:32, 11.60s/step, acc=0.1041, loss=5.9822, lr=6.08e-05]
Training:   1%|▏         | 144/11082 [28:07<35:10:50, 11.58s/step, acc=0.1041, loss=5.9822, lr=6.08e-05]
Training:   1%|▏         | 145/11082 [28:19<35:12:27, 11.59s/step, acc=0.1041, loss=5.9822, lr=6.08e-05]
Training:   1%|▏         | 146/11082 [28:30<34:53:23, 11.49s/step, acc=0.1041, loss=5.9822, lr=6.08e-05]
Training:   1%|▏         | 147/11082 [28:42<35:13:40, 11.60s/step, acc=0.1041, loss=5.9822, lr=6.08e-05]
Training:   1%|▏         | 148/11082 [28:53<34:55:53, 11.50s/step, acc=0.1041, loss=5.9822, lr=6.08e-05]
Training:   1%|▏         | 149/11082 [29:05<34:59:09, 11.52s/step, acc=0.1041, loss=5.9822, lr=6.08e-05]
Training:   1%|▏         | 150/11082 [29:16<34:58:49, 11.52s/step, acc=0.1041, loss=5.9822, lr=6.08e-05]
Training:   1%|▏         | 150/11082 [29:23<34:58:49, 11.52s/step, acc=0.1514, loss=5.2607, lr=9.10e-05]2026-06-18 15:22:51 | INFO | __main__ | epoch=0 step=150 train_loss=5.260693 train_acc=0.151374 lr=9.096e-05 grad_norm=3.5625 acc_d0=0.3205 acc_d1=0.1381 acc_d2=0.1091 acc_d3=0.0885 acc_d4=0.0789 acc_d5=0.0720 acc_d6=0.0692 acc_d7=0.0670

Training:   1%|▏         | 151/11082 [29:28<34:47:52, 11.46s/step, acc=0.1514, loss=5.2607, lr=9.10e-05]
Training:   1%|▏         | 152/11082 [29:39<34:56:10, 11.51s/step, acc=0.1514, loss=5.2607, lr=9.10e-05]
Training:   1%|▏         | 153/11082 [29:51<35:11:43, 11.59s/step, acc=0.1514, loss=5.2607, lr=9.10e-05]
Training:   1%|▏         | 154/11082 [30:03<35:17:21, 11.63s/step, acc=0.1514, loss=5.2607, lr=9.10e-05]
Training:   1%|▏         | 155/11082 [30:14<35:02:03, 11.54s/step, acc=0.1514, loss=5.2607, lr=9.10e-05]
Training:   1%|▏         | 156/11082 [30:26<34:58:23, 11.52s/step, acc=0.1514, loss=5.2607, lr=9.10e-05]
Training:   1%|▏         | 157/11082 [30:37<35:05:41, 11.56s/step, acc=0.1514, loss=5.2607, lr=9.10e-05]
Training:   1%|▏         | 158/11082 [30:49<35:29:44, 11.70s/step, acc=0.1514, loss=5.2607, lr=9.10e-05]
Training:   1%|▏         | 159/11082 [31:00<34:53:10, 11.50s/step, acc=0.1514, loss=5.2607, lr=9.10e-05]
Training:   1%|▏         | 160/11082 [31:12<35:00:49, 11.54s/step, acc=0.1514, loss=5.2607, lr=9.10e-05]
Training:   1%|▏         | 161/11082 [31:23<35:03:48, 11.56s/step, acc=0.1514, loss=5.2607, lr=9.10e-05]
Training:   1%|▏         | 162/11082 [31:35<34:49:33, 11.48s/step, acc=0.1514, loss=5.2607, lr=9.10e-05]
Training:   1%|▏         | 163/11082 [31:46<35:02:48, 11.55s/step, acc=0.1514, loss=5.2607, lr=9.10e-05]
Training:   1%|▏         | 164/11082 [31:58<34:48:19, 11.48s/step, acc=0.1514, loss=5.2607, lr=9.10e-05]
Training:   1%|▏         | 165/11082 [32:09<34:58:22, 11.53s/step, acc=0.1514, loss=5.2607, lr=9.10e-05]
Training:   1%|▏         | 166/11082 [32:22<35:31:38, 11.72s/step, acc=0.1514, loss=5.2607, lr=9.10e-05]
Training:   2%|▏         | 167/11082 [32:33<35:30:46, 11.71s/step, acc=0.1514, loss=5.2607, lr=9.10e-05]
Training:   2%|▏         | 168/11082 [32:45<35:34:59, 11.74s/step, acc=0.1514, loss=5.2607, lr=9.10e-05]
Training:   2%|▏         | 169/11082 [32:56<35:13:18, 11.62s/step, acc=0.1514, loss=5.2607, lr=9.10e-05]
Training:   2%|▏         | 170/11082 [33:08<35:16:37, 11.64s/step, acc=0.1514, loss=5.2607, lr=9.10e-05]
Training:   2%|▏         | 171/11082 [33:19<34:47:04, 11.48s/step, acc=0.1514, loss=5.2607, lr=9.10e-05]
Training:   2%|▏         | 172/11082 [33:31<34:51:01, 11.50s/step, acc=0.1514, loss=5.2607, lr=9.10e-05]
Training:   2%|▏         | 173/11082 [33:42<34:41:53, 11.45s/step, acc=0.1514, loss=5.2607, lr=9.10e-05]
Training:   2%|▏         | 174/11082 [33:54<34:39:16, 11.44s/step, acc=0.1514, loss=5.2607, lr=9.10e-05]
Training:   2%|▏         | 175/11082 [34:06<35:15:25, 11.64s/step, acc=0.1514, loss=5.2607, lr=9.10e-05]
Training:   2%|▏         | 176/11082 [34:17<34:52:48, 11.51s/step, acc=0.1514, loss=5.2607, lr=9.10e-05]
Training:   2%|▏         | 177/11082 [34:28<35:00:36, 11.56s/step, acc=0.1514, loss=5.2607, lr=9.10e-05]
Training:   2%|▏         | 178/11082 [34:39<34:21:40, 11.34s/step, acc=0.1514, loss=5.2607, lr=9.10e-05]
Training:   2%|▏         | 179/11082 [34:51<35:00:08, 11.56s/step, acc=0.1514, loss=5.2607, lr=9.10e-05]
Training:   2%|▏         | 180/11082 [35:03<35:06:05, 11.59s/step, acc=0.1514, loss=5.2607, lr=9.10e-05]
Training:   2%|▏         | 181/11082 [35:15<35:01:30, 11.57s/step, acc=0.1514, loss=5.2607, lr=9.10e-05]
Training:   2%|▏         | 182/11082 [35:26<35:05:27, 11.59s/step, acc=0.1514, loss=5.2607, lr=9.10e-05]
Training:   2%|▏         | 183/11082 [35:38<35:18:31, 11.66s/step, acc=0.1514, loss=5.2607, lr=9.10e-05]
Training:   2%|▏         | 184/11082 [35:50<35:24:46, 11.70s/step, acc=0.1514, loss=5.2607, lr=9.10e-05]
Training:   2%|▏         | 185/11082 [36:01<35:11:53, 11.63s/step, acc=0.1514, loss=5.2607, lr=9.10e-05]
Training:   2%|▏         | 186/11082 [36:12<34:34:26, 11.42s/step, acc=0.1514, loss=5.2607, lr=9.10e-05]
Training:   2%|▏         | 187/11082 [36:24<35:06:20, 11.60s/step, acc=0.1514, loss=5.2607, lr=9.10e-05]
Training:   2%|▏         | 188/11082 [36:36<34:53:18, 11.53s/step, acc=0.1514, loss=5.2607, lr=9.10e-05]
Training:   2%|▏         | 189/11082 [36:47<34:50:11, 11.51s/step, acc=0.1514, loss=5.2607, lr=9.10e-05]
Training:   2%|▏         | 190/11082 [36:58<34:43:40, 11.48s/step, acc=0.1514, loss=5.2607, lr=9.10e-05]
Training:   2%|▏         | 191/11082 [37:10<35:07:40, 11.61s/step, acc=0.1514, loss=5.2607, lr=9.10e-05]
Training:   2%|▏         | 192/11082 [37:22<34:54:49, 11.54s/step, acc=0.1514, loss=5.2607, lr=9.10e-05]
Training:   2%|▏         | 193/11082 [37:34<35:20:21, 11.68s/step, acc=0.1514, loss=5.2607, lr=9.10e-05]
Training:   2%|▏         | 194/11082 [37:45<34:44:59, 11.49s/step, acc=0.1514, loss=5.2607, lr=9.10e-05]
Training:   2%|▏         | 195/11082 [37:57<35:08:51, 11.62s/step, acc=0.1514, loss=5.2607, lr=9.10e-05]
Training:   2%|▏         | 196/11082 [38:08<35:10:22, 11.63s/step, acc=0.1514, loss=5.2607, lr=9.10e-05]
Training:   2%|▏         | 197/11082 [38:20<34:43:21, 11.48s/step, acc=0.1514, loss=5.2607, lr=9.10e-05]
Training:   2%|▏         | 198/11082 [38:31<34:24:34, 11.38s/step, acc=0.1514, loss=5.2607, lr=9.10e-05]
Training:   2%|▏         | 199/11082 [38:42<34:43:23, 11.49s/step, acc=0.1514, loss=5.2607, lr=9.10e-05]
Training:   2%|▏         | 200/11082 [38:54<34:48:49, 11.52s/step, acc=0.1514, loss=5.2607, lr=9.10e-05]
Training:   2%|▏         | 200/11082 [39:01<34:48:49, 11.52s/step, acc=0.1803, loss=4.8810, lr=1.00e-04]2026-06-18 15:32:29 | INFO | __main__ | epoch=0 step=200 train_loss=4.880984 train_acc=0.180307 lr=1.000e-04 grad_norm=2.2969 acc_d0=0.3850 acc_d1=0.1851 acc_d2=0.1281 acc_d3=0.1006 acc_d4=0.0864 acc_d5=0.0786 acc_d6=0.0727 acc_d7=0.0698

Training:   2%|▏         | 201/11082 [39:05<34:21:05, 11.37s/step, acc=0.1803, loss=4.8810, lr=1.00e-04]
Training:   2%|▏         | 202/11082 [39:17<35:10:22, 11.64s/step, acc=0.1803, loss=4.8810, lr=1.00e-04]
Training:   2%|▏         | 203/11082 [39:29<35:05:17, 11.61s/step, acc=0.1803, loss=4.8810, lr=1.00e-04]
Training:   2%|▏         | 204/11082 [39:40<34:58:30, 11.57s/step, acc=0.1803, loss=4.8810, lr=1.00e-04]
Training:   2%|▏         | 205/11082 [39:52<34:54:42, 11.55s/step, acc=0.1803, loss=4.8810, lr=1.00e-04]
Training:   2%|▏         | 206/11082 [40:03<34:50:11, 11.53s/step, acc=0.1803, loss=4.8810, lr=1.00e-04]
Training:   2%|▏         | 207/11082 [40:15<35:00:23, 11.59s/step, acc=0.1803, loss=4.8810, lr=1.00e-04]
Training:   2%|▏         | 208/11082 [40:27<34:52:29, 11.55s/step, acc=0.1803, loss=4.8810, lr=1.00e-04]
Training:   2%|▏         | 209/11082 [40:38<35:09:34, 11.64s/step, acc=0.1803, loss=4.8810, lr=1.00e-04]
Training:   2%|▏         | 210/11082 [40:50<35:09:02, 11.64s/step, acc=0.1803, loss=4.8810, lr=1.00e-04]
Training:   2%|▏         | 211/11082 [41:01<34:49:16, 11.53s/step, acc=0.1803, loss=4.8810, lr=1.00e-04]
Training:   2%|▏         | 212/11082 [41:13<35:23:54, 11.72s/step, acc=0.1803, loss=4.8810, lr=1.00e-04]
Training:   2%|▏         | 213/11082 [41:25<35:38:03, 11.80s/step, acc=0.1803, loss=4.8810, lr=1.00e-04]
Training:   2%|▏         | 214/11082 [41:37<35:08:04, 11.64s/step, acc=0.1803, loss=4.8810, lr=1.00e-04]
Training:   2%|▏         | 215/11082 [41:48<35:15:44, 11.68s/step, acc=0.1803, loss=4.8810, lr=1.00e-04]
Training:   2%|▏         | 216/11082 [42:00<35:01:55, 11.61s/step, acc=0.1803, loss=4.8810, lr=1.00e-04]
Training:   2%|▏         | 217/11082 [42:11<35:00:16, 11.60s/step, acc=0.1803, loss=4.8810, lr=1.00e-04]
Training:   2%|▏         | 218/11082 [42:23<35:12:27, 11.67s/step, acc=0.1803, loss=4.8810, lr=1.00e-04]
Training:   2%|▏         | 219/11082 [42:35<34:50:50, 11.55s/step, acc=0.1803, loss=4.8810, lr=1.00e-04]
Training:   2%|▏         | 220/11082 [42:46<34:42:48, 11.51s/step, acc=0.1803, loss=4.8810, lr=1.00e-04]
Training:   2%|▏         | 221/11082 [42:58<34:52:41, 11.56s/step, acc=0.1803, loss=4.8810, lr=1.00e-04]
Training:   2%|▏         | 222/11082 [43:09<34:47:44, 11.53s/step, acc=0.1803, loss=4.8810, lr=1.00e-04]
Training:   2%|▏         | 223/11082 [43:21<35:03:27, 11.62s/step, acc=0.1803, loss=4.8810, lr=1.00e-04]
Training:   2%|▏         | 224/11082 [43:33<34:57:29, 11.59s/step, acc=0.1803, loss=4.8810, lr=1.00e-04]
Training:   2%|▏         | 225/11082 [43:44<35:02:13, 11.62s/step, acc=0.1803, loss=4.8810, lr=1.00e-04]
Training:   2%|▏         | 226/11082 [43:56<34:49:36, 11.55s/step, acc=0.1803, loss=4.8810, lr=1.00e-04]
Training:   2%|▏         | 227/11082 [44:07<34:58:33, 11.60s/step, acc=0.1803, loss=4.8810, lr=1.00e-04]
Training:   2%|▏         | 228/11082 [44:19<34:38:02, 11.49s/step, acc=0.1803, loss=4.8810, lr=1.00e-04]
Training:   2%|▏         | 229/11082 [44:30<34:38:09, 11.49s/step, acc=0.1803, loss=4.8810, lr=1.00e-04]
Training:   2%|▏         | 230/11082 [44:42<34:57:40, 11.60s/step, acc=0.1803, loss=4.8810, lr=1.00e-04]
Training:   2%|▏         | 231/11082 [44:53<34:50:43, 11.56s/step, acc=0.1803, loss=4.8810, lr=1.00e-04]
Training:   2%|▏         | 232/11082 [45:05<35:07:54, 11.66s/step, acc=0.1803, loss=4.8810, lr=1.00e-04]
Training:   2%|▏         | 233/11082 [45:16<34:36:07, 11.48s/step, acc=0.1803, loss=4.8810, lr=1.00e-04]
Training:   2%|▏         | 234/11082 [45:28<35:10:39, 11.67s/step, acc=0.1803, loss=4.8810, lr=1.00e-04]
Training:   2%|▏         | 235/11082 [45:40<34:48:37, 11.55s/step, acc=0.1803, loss=4.8810, lr=1.00e-04]
Training:   2%|▏         | 236/11082 [45:51<34:28:58, 11.45s/step, acc=0.1803, loss=4.8810, lr=1.00e-04]
Training:   2%|▏         | 237/11082 [46:02<34:37:20, 11.49s/step, acc=0.1803, loss=4.8810, lr=1.00e-04]
Training:   2%|▏         | 238/11082 [46:14<34:54:38, 11.59s/step, acc=0.1803, loss=4.8810, lr=1.00e-04]
Training:   2%|▏         | 239/11082 [46:26<35:01:15, 11.63s/step, acc=0.1803, loss=4.8810, lr=1.00e-04]
Training:   2%|▏         | 240/11082 [46:38<35:03:51, 11.64s/step, acc=0.1803, loss=4.8810, lr=1.00e-04]
Training:   2%|▏         | 241/11082 [46:49<34:38:52, 11.51s/step, acc=0.1803, loss=4.8810, lr=1.00e-04]
Training:   2%|▏         | 242/11082 [47:01<35:05:37, 11.65s/step, acc=0.1803, loss=4.8810, lr=1.00e-04]
Training:   2%|▏         | 243/11082 [47:13<35:07:38, 11.67s/step, acc=0.1803, loss=4.8810, lr=1.00e-04]
Training:   2%|▏         | 244/11082 [47:24<34:58:23, 11.62s/step, acc=0.1803, loss=4.8810, lr=1.00e-04]
Training:   2%|▏         | 245/11082 [47:35<34:38:41, 11.51s/step, acc=0.1803, loss=4.8810, lr=1.00e-04]
Training:   2%|▏         | 246/11082 [47:47<34:43:44, 11.54s/step, acc=0.1803, loss=4.8810, lr=1.00e-04]
Training:   2%|▏         | 247/11082 [47:59<34:46:43, 11.56s/step, acc=0.1803, loss=4.8810, lr=1.00e-04]
Training:   2%|▏         | 248/11082 [48:10<34:46:06, 11.55s/step, acc=0.1803, loss=4.8810, lr=1.00e-04]
Training:   2%|▏         | 249/11082 [48:21<34:26:58, 11.45s/step, acc=0.1803, loss=4.8810, lr=1.00e-04]
Training:   2%|▏         | 250/11082 [48:33<35:00:31, 11.64s/step, acc=0.1803, loss=4.8810, lr=1.00e-04]
Training:   2%|▏         | 250/11082 [48:40<35:00:31, 11.64s/step, acc=0.2134, loss=4.5025, lr=1.00e-04]2026-06-18 15:42:08 | INFO | __main__ | epoch=0 step=250 train_loss=4.502458 train_acc=0.213420 lr=9.999e-05 grad_norm=1.6797 acc_d0=0.4262 acc_d1=0.2335 acc_d2=0.1549 acc_d3=0.1175 acc_d4=0.0956 acc_d5=0.0854 acc_d6=0.0782 acc_d7=0.0728

Training:   2%|▏         | 251/11082 [48:45<34:42:49, 11.54s/step, acc=0.2134, loss=4.5025, lr=1.00e-04]
Training:   2%|▏         | 252/11082 [48:56<34:14:22, 11.38s/step, acc=0.2134, loss=4.5025, lr=1.00e-04]
Training:   2%|▏         | 253/11082 [49:08<34:51:02, 11.59s/step, acc=0.2134, loss=4.5025, lr=1.00e-04]
Training:   2%|▏         | 254/11082 [49:19<34:31:52, 11.48s/step, acc=0.2134, loss=4.5025, lr=1.00e-04]
Training:   2%|▏         | 255/11082 [49:30<34:32:51, 11.49s/step, acc=0.2134, loss=4.5025, lr=1.00e-04]
Training:   2%|▏         | 256/11082 [49:42<35:00:19, 11.64s/step, acc=0.2134, loss=4.5025, lr=1.00e-04]
Training:   2%|▏         | 257/11082 [49:54<35:07:22, 11.68s/step, acc=0.2134, loss=4.5025, lr=1.00e-04]
Training:   2%|▏         | 258/11082 [50:06<35:12:24, 11.71s/step, acc=0.2134, loss=4.5025, lr=1.00e-04]
Training:   2%|▏         | 259/11082 [50:18<35:07:08, 11.68s/step, acc=0.2134, loss=4.5025, lr=1.00e-04]
Training:   2%|▏         | 260/11082 [50:29<34:52:31, 11.60s/step, acc=0.2134, loss=4.5025, lr=1.00e-04]
Training:   2%|▏         | 261/11082 [50:40<34:35:35, 11.51s/step, acc=0.2134, loss=4.5025, lr=1.00e-04]
Training:   2%|▏         | 262/11082 [50:52<35:07:35, 11.69s/step, acc=0.2134, loss=4.5025, lr=1.00e-04]
Training:   2%|▏         | 263/11082 [51:04<34:48:23, 11.58s/step, acc=0.2134, loss=4.5025, lr=1.00e-04]
Training:   2%|▏         | 264/11082 [51:15<34:23:49, 11.45s/step, acc=0.2134, loss=4.5025, lr=1.00e-04]
Training:   2%|▏         | 265/11082 [51:27<34:48:28, 11.58s/step, acc=0.2134, loss=4.5025, lr=1.00e-04]
Training:   2%|▏         | 266/11082 [51:38<34:45:26, 11.57s/step, acc=0.2134, loss=4.5025, lr=1.00e-04]
Training:   2%|▏         | 267/11082 [51:50<34:24:25, 11.45s/step, acc=0.2134, loss=4.5025, lr=1.00e-04]
Training:   2%|▏         | 268/11082 [52:01<34:49:06, 11.59s/step, acc=0.2134, loss=4.5025, lr=1.00e-04]
Training:   2%|▏         | 269/11082 [52:13<34:24:33, 11.46s/step, acc=0.2134, loss=4.5025, lr=1.00e-04]
Training:   2%|▏         | 270/11082 [52:24<34:47:35, 11.58s/step, acc=0.2134, loss=4.5025, lr=1.00e-04]
Training:   2%|▏         | 271/11082 [52:36<34:31:31, 11.50s/step, acc=0.2134, loss=4.5025, lr=1.00e-04]
Training:   2%|▏         | 272/11082 [52:47<34:39:32, 11.54s/step, acc=0.2134, loss=4.5025, lr=1.00e-04]
Training:   2%|▏         | 273/11082 [52:59<34:48:33, 11.59s/step, acc=0.2134, loss=4.5025, lr=1.00e-04]
Training:   2%|▏         | 274/11082 [53:11<34:51:39, 11.61s/step, acc=0.2134, loss=4.5025, lr=1.00e-04]
Training:   2%|▏         | 275/11082 [53:22<34:32:08, 11.50s/step, acc=0.2134, loss=4.5025, lr=1.00e-04]
Training:   2%|▏         | 276/11082 [53:34<34:33:43, 11.51s/step, acc=0.2134, loss=4.5025, lr=1.00e-04]
Training:   2%|▏         | 277/11082 [53:45<34:43:28, 11.57s/step, acc=0.2134, loss=4.5025, lr=1.00e-04]
Training:   3%|▎         | 278/11082 [53:56<34:18:45, 11.43s/step, acc=0.2134, loss=4.5025, lr=1.00e-04]
Training:   3%|▎         | 279/11082 [54:08<34:51:02, 11.61s/step, acc=0.2134, loss=4.5025, lr=1.00e-04]
Training:   3%|▎         | 280/11082 [54:20<34:35:49, 11.53s/step, acc=0.2134, loss=4.5025, lr=1.00e-04]
Training:   3%|▎         | 281/11082 [54:31<34:22:08, 11.46s/step, acc=0.2134, loss=4.5025, lr=1.00e-04]
Training:   3%|▎         | 282/11082 [54:43<34:41:43, 11.57s/step, acc=0.2134, loss=4.5025, lr=1.00e-04]
Training:   3%|▎         | 283/11082 [54:54<34:37:35, 11.54s/step, acc=0.2134, loss=4.5025, lr=1.00e-04]
Training:   3%|▎         | 284/11082 [55:06<34:30:23, 11.50s/step, acc=0.2134, loss=4.5025, lr=1.00e-04]
Training:   3%|▎         | 285/11082 [55:18<34:43:38, 11.58s/step, acc=0.2134, loss=4.5025, lr=1.00e-04]
Training:   3%|▎         | 286/11082 [55:29<34:21:07, 11.45s/step, acc=0.2134, loss=4.5025, lr=1.00e-04]
Training:   3%|▎         | 287/11082 [55:40<34:27:55, 11.49s/step, acc=0.2134, loss=4.5025, lr=1.00e-04]
Training:   3%|▎         | 288/11082 [55:52<34:43:47, 11.58s/step, acc=0.2134, loss=4.5025, lr=1.00e-04]
Training:   3%|▎         | 289/11082 [56:03<34:15:33, 11.43s/step, acc=0.2134, loss=4.5025, lr=1.00e-04]
Training:   3%|▎         | 290/11082 [56:15<34:35:12, 11.54s/step, acc=0.2134, loss=4.5025, lr=1.00e-04]
Training:   3%|▎         | 291/11082 [56:26<33:59:34, 11.34s/step, acc=0.2134, loss=4.5025, lr=1.00e-04]
Training:   3%|▎         | 292/11082 [56:38<34:39:01, 11.56s/step, acc=0.2134, loss=4.5025, lr=1.00e-04]
Training:   3%|▎         | 293/11082 [56:50<35:15:59, 11.77s/step, acc=0.2134, loss=4.5025, lr=1.00e-04]
Training:   3%|▎         | 294/11082 [57:02<35:02:05, 11.69s/step, acc=0.2134, loss=4.5025, lr=1.00e-04]
Training:   3%|▎         | 295/11082 [57:13<35:04:27, 11.71s/step, acc=0.2134, loss=4.5025, lr=1.00e-04]
Training:   3%|▎         | 296/11082 [57:25<34:47:36, 11.61s/step, acc=0.2134, loss=4.5025, lr=1.00e-04]
Training:   3%|▎         | 297/11082 [57:36<34:25:35, 11.49s/step, acc=0.2134, loss=4.5025, lr=1.00e-04]
Training:   3%|▎         | 298/11082 [57:48<34:46:03, 11.61s/step, acc=0.2134, loss=4.5025, lr=1.00e-04]
Training:   3%|▎         | 299/11082 [58:00<35:03:03, 11.70s/step, acc=0.2134, loss=4.5025, lr=1.00e-04]
Training:   3%|▎         | 300/11082 [58:11<34:22:07, 11.48s/step, acc=0.2134, loss=4.5025, lr=1.00e-04]
Training:   3%|▎         | 300/11082 [58:18<34:22:07, 11.48s/step, acc=0.2423, loss=4.1980, lr=1.00e-04]2026-06-18 15:51:46 | INFO | __main__ | epoch=0 step=300 train_loss=4.197972 train_acc=0.242276 lr=9.997e-05 grad_norm=1.8672 acc_d0=0.4881 acc_d1=0.2889 acc_d2=0.1865 acc_d3=0.1376 acc_d4=0.1090 acc_d5=0.0950 acc_d6=0.0853 acc_d7=0.0789

Training:   3%|▎         | 301/11082 [58:23<34:42:44, 11.59s/step, acc=0.2423, loss=4.1980, lr=1.00e-04]
Training:   3%|▎         | 302/11082 [58:34<34:38:59, 11.57s/step, acc=0.2423, loss=4.1980, lr=1.00e-04]
Training:   3%|▎         | 303/11082 [58:45<34:12:51, 11.43s/step, acc=0.2423, loss=4.1980, lr=1.00e-04]
Training:   3%|▎         | 304/11082 [58:57<34:10:44, 11.42s/step, acc=0.2423, loss=4.1980, lr=1.00e-04]
Training:   3%|▎         | 305/11082 [59:08<34:35:24, 11.55s/step, acc=0.2423, loss=4.1980, lr=1.00e-04]
Training:   3%|▎         | 306/11082 [59:20<34:46:27, 11.62s/step, acc=0.2423, loss=4.1980, lr=1.00e-04]
Training:   3%|▎         | 307/11082 [59:32<34:49:36, 11.64s/step, acc=0.2423, loss=4.1980, lr=1.00e-04]
Training:   3%|▎         | 308/11082 [59:43<34:17:23, 11.46s/step, acc=0.2423, loss=4.1980, lr=1.00e-04]
Training:   3%|▎         | 309/11082 [59:55<34:38:32, 11.58s/step, acc=0.2423, loss=4.1980, lr=1.00e-04]
Training:   3%|▎         | 310/11082 [1:00:06<34:16:01, 11.45s/step, acc=0.2423, loss=4.1980, lr=1.00e-04]
Training:   3%|▎         | 311/11082 [1:00:18<35:01:01, 11.70s/step, acc=0.2423, loss=4.1980, lr=1.00e-04]
Training:   3%|▎         | 312/11082 [1:00:30<35:02:36, 11.71s/step, acc=0.2423, loss=4.1980, lr=1.00e-04]
Training:   3%|▎         | 313/11082 [1:00:41<34:45:02, 11.62s/step, acc=0.2423, loss=4.1980, lr=1.00e-04]
Training:   3%|▎         | 314/11082 [1:00:53<34:41:47, 11.60s/step, acc=0.2423, loss=4.1980, lr=1.00e-04]
Training:   3%|▎         | 315/11082 [1:01:05<34:43:01, 11.61s/step, acc=0.2423, loss=4.1980, lr=1.00e-04]
Training:   3%|▎         | 316/11082 [1:01:16<34:17:46, 11.47s/step, acc=0.2423, loss=4.1980, lr=1.00e-04]
Training:   3%|▎         | 317/11082 [1:01:28<34:51:11, 11.66s/step, acc=0.2423, loss=4.1980, lr=1.00e-04]
Training:   3%|▎         | 318/11082 [1:01:39<34:27:30, 11.52s/step, acc=0.2423, loss=4.1980, lr=1.00e-04]
Training:   3%|▎         | 319/11082 [1:01:51<34:31:48, 11.55s/step, acc=0.2423, loss=4.1980, lr=1.00e-04]
Training:   3%|▎         | 320/11082 [1:02:02<34:16:31, 11.47s/step, acc=0.2423, loss=4.1980, lr=1.00e-04]
Training:   3%|▎         | 321/11082 [1:02:14<34:50:33, 11.66s/step, acc=0.2423, loss=4.1980, lr=1.00e-04]
Training:   3%|▎         | 322/11082 [1:02:25<33:57:36, 11.36s/step, acc=0.2423, loss=4.1980, lr=1.00e-04]
Training:   3%|▎         | 323/11082 [1:02:36<34:01:50, 11.39s/step, acc=0.2423, loss=4.1980, lr=1.00e-04]
Training:   3%|▎         | 324/11082 [1:02:48<34:14:53, 11.46s/step, acc=0.2423, loss=4.1980, lr=1.00e-04]
Training:   3%|▎         | 325/11082 [1:02:59<34:24:46, 11.52s/step, acc=0.2423, loss=4.1980, lr=1.00e-04]
Training:   3%|▎         | 326/11082 [1:03:11<34:05:01, 11.41s/step, acc=0.2423, loss=4.1980, lr=1.00e-04]
Training:   3%|▎         | 327/11082 [1:03:23<34:47:05, 11.64s/step, acc=0.2423, loss=4.1980, lr=1.00e-04]
Training:   3%|▎         | 328/11082 [1:03:34<34:37:33, 11.59s/step, acc=0.2423, loss=4.1980, lr=1.00e-04]
Training:   3%|▎         | 329/11082 [1:03:46<34:23:27, 11.51s/step, acc=0.2423, loss=4.1980, lr=1.00e-04]
Training:   3%|▎         | 330/11082 [1:03:57<34:03:56, 11.41s/step, acc=0.2423, loss=4.1980, lr=1.00e-04]
Training:   3%|▎         | 331/11082 [1:04:09<34:29:36, 11.55s/step, acc=0.2423, loss=4.1980, lr=1.00e-04]
Training:   3%|▎         | 332/11082 [1:04:20<34:03:02, 11.40s/step, acc=0.2423, loss=4.1980, lr=1.00e-04]
Training:   3%|▎         | 333/11082 [1:04:32<34:41:06, 11.62s/step, acc=0.2423, loss=4.1980, lr=1.00e-04]
Training:   3%|▎         | 334/11082 [1:04:43<34:41:25, 11.62s/step, acc=0.2423, loss=4.1980, lr=1.00e-04]
Training:   3%|▎         | 335/11082 [1:04:55<34:19:51, 11.50s/step, acc=0.2423, loss=4.1980, lr=1.00e-04]
Training:   3%|▎         | 336/11082 [1:05:06<34:18:46, 11.50s/step, acc=0.2423, loss=4.1980, lr=1.00e-04]
Training:   3%|▎         | 337/11082 [1:05:18<34:31:08, 11.57s/step, acc=0.2423, loss=4.1980, lr=1.00e-04]
Training:   3%|▎         | 338/11082 [1:05:29<34:25:49, 11.54s/step, acc=0.2423, loss=4.1980, lr=1.00e-04]
Training:   3%|▎         | 339/11082 [1:05:41<34:19:33, 11.50s/step, acc=0.2423, loss=4.1980, lr=1.00e-04]
Training:   3%|▎         | 340/11082 [1:05:52<34:19:25, 11.50s/step, acc=0.2423, loss=4.1980, lr=1.00e-04]
Training:   3%|▎         | 341/11082 [1:06:04<34:17:51, 11.50s/step, acc=0.2423, loss=4.1980, lr=1.00e-04]
Training:   3%|▎         | 342/11082 [1:06:16<34:35:09, 11.59s/step, acc=0.2423, loss=4.1980, lr=1.00e-04]
Training:   3%|▎         | 343/11082 [1:06:27<34:26:07, 11.54s/step, acc=0.2423, loss=4.1980, lr=1.00e-04]
Training:   3%|▎         | 344/11082 [1:06:38<34:12:18, 11.47s/step, acc=0.2423, loss=4.1980, lr=1.00e-04]
Training:   3%|▎         | 345/11082 [1:06:50<34:23:45, 11.53s/step, acc=0.2423, loss=4.1980, lr=1.00e-04]
Training:   3%|▎         | 346/11082 [1:07:01<34:21:45, 11.52s/step, acc=0.2423, loss=4.1980, lr=1.00e-04]
Training:   3%|▎         | 347/11082 [1:07:13<34:39:48, 11.62s/step, acc=0.2423, loss=4.1980, lr=1.00e-04]
Training:   3%|▎         | 348/11082 [1:07:25<34:19:11, 11.51s/step, acc=0.2423, loss=4.1980, lr=1.00e-04]
Training:   3%|▎         | 349/11082 [1:07:36<34:29:20, 11.57s/step, acc=0.2423, loss=4.1980, lr=1.00e-04]
Training:   3%|▎         | 350/11082 [1:07:47<33:59:02, 11.40s/step, acc=0.2423, loss=4.1980, lr=1.00e-04]
Training:   3%|▎         | 350/11082 [1:07:55<33:59:02, 11.40s/step, acc=0.2585, loss=4.0104, lr=9.99e-05]2026-06-18 16:01:23 | INFO | __main__ | epoch=0 step=350 train_loss=4.010356 train_acc=0.258486 lr=9.994e-05 grad_norm=2.3438 acc_d0=0.4916 acc_d1=0.3056 acc_d2=0.1978 acc_d3=0.1429 acc_d4=0.1121 acc_d5=0.0965 acc_d6=0.0873 acc_d7=0.0778

Training:   3%|▎         | 351/11082 [1:07:59<34:34:14, 11.60s/step, acc=0.2585, loss=4.0104, lr=9.99e-05]
Training:   3%|▎         | 352/11082 [1:08:11<34:22:04, 11.53s/step, acc=0.2585, loss=4.0104, lr=9.99e-05]
Training:   3%|▎         | 353/11082 [1:08:22<34:29:46, 11.57s/step, acc=0.2585, loss=4.0104, lr=9.99e-05]
Training:   3%|▎         | 354/11082 [1:08:34<34:36:01, 11.61s/step, acc=0.2585, loss=4.0104, lr=9.99e-05]
Training:   3%|▎         | 355/11082 [1:08:46<34:29:25, 11.58s/step, acc=0.2585, loss=4.0104, lr=9.99e-05]
Training:   3%|▎         | 356/11082 [1:08:57<34:04:39, 11.44s/step, acc=0.2585, loss=4.0104, lr=9.99e-05]
Training:   3%|▎         | 357/11082 [1:09:09<34:41:59, 11.65s/step, acc=0.2585, loss=4.0104, lr=9.99e-05]
Training:   3%|▎         | 358/11082 [1:09:20<34:03:03, 11.43s/step, acc=0.2585, loss=4.0104, lr=9.99e-05]
Training:   3%|▎         | 359/11082 [1:09:31<34:16:18, 11.51s/step, acc=0.2585, loss=4.0104, lr=9.99e-05]
Training:   3%|▎         | 360/11082 [1:09:43<34:17:50, 11.52s/step, acc=0.2585, loss=4.0104, lr=9.99e-05]
Training:   3%|▎         | 361/11082 [1:09:55<34:22:08, 11.54s/step, acc=0.2585, loss=4.0104, lr=9.99e-05]
Training:   3%|▎         | 362/11082 [1:10:06<34:08:14, 11.46s/step, acc=0.2585, loss=4.0104, lr=9.99e-05]
Training:   3%|▎         | 363/11082 [1:10:17<34:14:04, 11.50s/step, acc=0.2585, loss=4.0104, lr=9.99e-05]
Training:   3%|▎         | 364/11082 [1:10:29<34:35:52, 11.62s/step, acc=0.2585, loss=4.0104, lr=9.99e-05]
Training:   3%|▎         | 365/11082 [1:10:41<34:37:59, 11.63s/step, acc=0.2585, loss=4.0104, lr=9.99e-05]
Training:   3%|▎         | 366/11082 [1:10:52<34:11:50, 11.49s/step, acc=0.2585, loss=4.0104, lr=9.99e-05]
Training:   3%|▎         | 367/11082 [1:11:04<34:32:38, 11.61s/step, acc=0.2585, loss=4.0104, lr=9.99e-05]
Training:   3%|▎         | 368/11082 [1:11:16<34:29:12, 11.59s/step, acc=0.2585, loss=4.0104, lr=9.99e-05]
Training:   3%|▎         | 369/11082 [1:11:27<34:13:48, 11.50s/step, acc=0.2585, loss=4.0104, lr=9.99e-05]
Training:   3%|▎         | 370/11082 [1:11:38<34:05:47, 11.46s/step, acc=0.2585, loss=4.0104, lr=9.99e-05]
Training:   3%|▎         | 371/11082 [1:11:50<34:20:12, 11.54s/step, acc=0.2585, loss=4.0104, lr=9.99e-05]
Training:   3%|▎         | 372/11082 [1:12:02<34:37:57, 11.64s/step, acc=0.2585, loss=4.0104, lr=9.99e-05]
Training:   3%|▎         | 373/11082 [1:12:13<34:16:57, 11.52s/step, acc=0.2585, loss=4.0104, lr=9.99e-05]
Training:   3%|▎         | 374/11082 [1:12:25<34:15:00, 11.51s/step, acc=0.2585, loss=4.0104, lr=9.99e-05]
Training:   3%|▎         | 375/11082 [1:12:36<34:34:02, 11.62s/step, acc=0.2585, loss=4.0104, lr=9.99e-05]
Training:   3%|▎         | 376/11082 [1:12:48<34:38:52, 11.65s/step, acc=0.2585, loss=4.0104, lr=9.99e-05]
Training:   3%|▎         | 377/11082 [1:13:00<34:32:51, 11.62s/step, acc=0.2585, loss=4.0104, lr=9.99e-05]
Training:   3%|▎         | 378/11082 [1:13:12<34:49:09, 11.71s/step, acc=0.2585, loss=4.0104, lr=9.99e-05]
Training:   3%|▎         | 379/11082 [1:13:23<34:21:36, 11.56s/step, acc=0.2585, loss=4.0104, lr=9.99e-05]
Training:   3%|▎         | 380/11082 [1:13:34<34:13:51, 11.51s/step, acc=0.2585, loss=4.0104, lr=9.99e-05]
Training:   3%|▎         | 381/11082 [1:13:46<34:29:42, 11.60s/step, acc=0.2585, loss=4.0104, lr=9.99e-05]
Training:   3%|▎         | 382/11082 [1:13:58<34:34:09, 11.63s/step, acc=0.2585, loss=4.0104, lr=9.99e-05]
Training:   3%|▎         | 383/11082 [1:14:10<34:51:46, 11.73s/step, acc=0.2585, loss=4.0104, lr=9.99e-05]
Training:   3%|▎         | 384/11082 [1:14:21<34:20:13, 11.55s/step, acc=0.2585, loss=4.0104, lr=9.99e-05]
Training:   3%|▎         | 385/11082 [1:14:33<34:49:01, 11.72s/step, acc=0.2585, loss=4.0104, lr=9.99e-05]
Training:   3%|▎         | 386/11082 [1:14:44<34:12:44, 11.51s/step, acc=0.2585, loss=4.0104, lr=9.99e-05]
Training:   3%|▎         | 387/11082 [1:14:55<34:06:03, 11.48s/step, acc=0.2585, loss=4.0104, lr=9.99e-05]
Training:   4%|▎         | 388/11082 [1:15:07<34:18:17, 11.55s/step, acc=0.2585, loss=4.0104, lr=9.99e-05]
Training:   4%|▎         | 389/11082 [1:15:18<34:05:38, 11.48s/step, acc=0.2585, loss=4.0104, lr=9.99e-05]
Training:   4%|▎         | 390/11082 [1:15:30<34:21:29, 11.57s/step, acc=0.2585, loss=4.0104, lr=9.99e-05]
Training:   4%|▎         | 391/11082 [1:15:41<34:01:33, 11.46s/step, acc=0.2585, loss=4.0104, lr=9.99e-05]
Training:   4%|▎         | 392/11082 [1:15:53<34:18:19, 11.55s/step, acc=0.2585, loss=4.0104, lr=9.99e-05]
Training:   4%|▎         | 393/11082 [1:16:04<34:01:15, 11.46s/step, acc=0.2585, loss=4.0104, lr=9.99e-05]
Training:   4%|▎         | 394/11082 [1:16:16<34:14:17, 11.53s/step, acc=0.2585, loss=4.0104, lr=9.99e-05]
Training:   4%|▎         | 395/11082 [1:16:28<34:29:40, 11.62s/step, acc=0.2585, loss=4.0104, lr=9.99e-05]
Training:   4%|▎         | 396/11082 [1:16:39<34:19:42, 11.56s/step, acc=0.2585, loss=4.0104, lr=9.99e-05]
Training:   4%|▎         | 397/11082 [1:16:51<34:23:38, 11.59s/step, acc=0.2585, loss=4.0104, lr=9.99e-05]
Training:   4%|▎         | 398/11082 [1:17:03<34:20:25, 11.57s/step, acc=0.2585, loss=4.0104, lr=9.99e-05]
Training:   4%|▎         | 399/11082 [1:17:14<34:16:34, 11.55s/step, acc=0.2585, loss=4.0104, lr=9.99e-05]
Training:   4%|▎         | 400/11082 [1:17:25<34:10:01, 11.51s/step, acc=0.2585, loss=4.0104, lr=9.99e-05]
Training:   4%|▎         | 400/11082 [1:17:33<34:10:01, 11.51s/step, acc=0.2748, loss=3.8521, lr=9.99e-05]2026-06-18 16:11:01 | INFO | __main__ | epoch=0 step=400 train_loss=3.852075 train_acc=0.274752 lr=9.990e-05 grad_norm=1.9922 acc_d0=0.5325 acc_d1=0.3424 acc_d2=0.2246 acc_d3=0.1601 acc_d4=0.1256 acc_d5=0.1052 acc_d6=0.0948 acc_d7=0.0833

Training:   4%|▎         | 401/11082 [1:17:37<34:11:42, 11.53s/step, acc=0.2748, loss=3.8521, lr=9.99e-05]
Training:   4%|▎         | 402/11082 [1:17:49<34:23:49, 11.59s/step, acc=0.2748, loss=3.8521, lr=9.99e-05]
Training:   4%|▎         | 403/11082 [1:18:00<34:27:23, 11.62s/step, acc=0.2748, loss=3.8521, lr=9.99e-05]
Training:   4%|▎         | 404/11082 [1:18:12<34:09:32, 11.52s/step, acc=0.2748, loss=3.8521, lr=9.99e-05]
Training:   4%|▎         | 405/11082 [1:18:23<34:06:37, 11.50s/step, acc=0.2748, loss=3.8521, lr=9.99e-05]
Training:   4%|▎         | 406/11082 [1:18:35<34:13:07, 11.54s/step, acc=0.2748, loss=3.8521, lr=9.99e-05]
Training:   4%|▎         | 407/11082 [1:18:47<34:21:29, 11.59s/step, acc=0.2748, loss=3.8521, lr=9.99e-05]
Training:   4%|▎         | 408/11082 [1:18:58<33:49:22, 11.41s/step, acc=0.2748, loss=3.8521, lr=9.99e-05]
Training:   4%|▎         | 409/11082 [1:19:10<34:30:42, 11.64s/step, acc=0.2748, loss=3.8521, lr=9.99e-05]
Training:   4%|▎         | 410/11082 [1:19:21<34:13:28, 11.54s/step, acc=0.2748, loss=3.8521, lr=9.99e-05]
Training:   4%|▎         | 411/11082 [1:19:33<34:22:36, 11.60s/step, acc=0.2748, loss=3.8521, lr=9.99e-05]
Training:   4%|▎         | 412/11082 [1:19:44<34:29:31, 11.64s/step, acc=0.2748, loss=3.8521, lr=9.99e-05]
Training:   4%|▎         | 413/11082 [1:19:55<33:43:46, 11.38s/step, acc=0.2748, loss=3.8521, lr=9.99e-05]
Training:   4%|▎         | 414/11082 [1:20:07<34:18:29, 11.58s/step, acc=0.2748, loss=3.8521, lr=9.99e-05]
Training:   4%|▎         | 415/11082 [1:20:19<34:14:13, 11.55s/step, acc=0.2748, loss=3.8521, lr=9.99e-05]
Training:   4%|▍         | 416/11082 [1:20:30<34:15:21, 11.56s/step, acc=0.2748, loss=3.8521, lr=9.99e-05]
Training:   4%|▍         | 417/11082 [1:20:42<34:34:57, 11.67s/step, acc=0.2748, loss=3.8521, lr=9.99e-05]
Training:   4%|▍         | 418/11082 [1:20:53<34:00:53, 11.48s/step, acc=0.2748, loss=3.8521, lr=9.99e-05]
Training:   4%|▍         | 419/11082 [1:21:05<34:16:21, 11.57s/step, acc=0.2748, loss=3.8521, lr=9.99e-05]
Training:   4%|▍         | 420/11082 [1:21:17<34:15:22, 11.57s/step, acc=0.2748, loss=3.8521, lr=9.99e-05]
Training:   4%|▍         | 421/11082 [1:21:28<33:59:52, 11.48s/step, acc=0.2748, loss=3.8521, lr=9.99e-05]
Training:   4%|▍         | 422/11082 [1:21:40<34:19:52, 11.59s/step, acc=0.2748, loss=3.8521, lr=9.99e-05]
Training:   4%|▍         | 423/11082 [1:21:51<34:08:12, 11.53s/step, acc=0.2748, loss=3.8521, lr=9.99e-05]
Training:   4%|▍         | 424/11082 [1:22:03<34:13:16, 11.56s/step, acc=0.2748, loss=3.8521, lr=9.99e-05]
Training:   4%|▍         | 425/11082 [1:22:14<34:09:54, 11.54s/step, acc=0.2748, loss=3.8521, lr=9.99e-05]
Training:   4%|▍         | 426/11082 [1:22:26<34:17:15, 11.58s/step, acc=0.2748, loss=3.8521, lr=9.99e-05]
Training:   4%|▍         | 427/11082 [1:22:37<34:09:47, 11.54s/step, acc=0.2748, loss=3.8521, lr=9.99e-05]
Training:   4%|▍         | 428/11082 [1:22:49<33:57:55, 11.48s/step, acc=0.2748, loss=3.8521, lr=9.99e-05]
Training:   4%|▍         | 429/11082 [1:23:01<34:11:56, 11.56s/step, acc=0.2748, loss=3.8521, lr=9.99e-05]
Training:   4%|▍         | 430/11082 [1:23:12<34:14:38, 11.57s/step, acc=0.2748, loss=3.8521, lr=9.99e-05]
Training:   4%|▍         | 431/11082 [1:23:23<33:57:03, 11.48s/step, acc=0.2748, loss=3.8521, lr=9.99e-05]
Training:   4%|▍         | 432/11082 [1:23:35<34:08:45, 11.54s/step, acc=0.2748, loss=3.8521, lr=9.99e-05]
Training:   4%|▍         | 433/11082 [1:23:47<34:08:52, 11.54s/step, acc=0.2748, loss=3.8521, lr=9.99e-05]
Training:   4%|▍         | 434/11082 [1:23:58<34:02:27, 11.51s/step, acc=0.2748, loss=3.8521, lr=9.99e-05]
Training:   4%|▍         | 435/11082 [1:24:10<34:15:27, 11.58s/step, acc=0.2748, loss=3.8521, lr=9.99e-05]
Training:   4%|▍         | 436/11082 [1:24:21<34:04:27, 11.52s/step, acc=0.2748, loss=3.8521, lr=9.99e-05]
Training:   4%|▍         | 437/11082 [1:24:33<34:02:48, 11.51s/step, acc=0.2748, loss=3.8521, lr=9.99e-05]
Training:   4%|▍         | 438/11082 [1:24:44<33:56:05, 11.48s/step, acc=0.2748, loss=3.8521, lr=9.99e-05]
Training:   4%|▍         | 439/11082 [1:24:56<34:03:12, 11.52s/step, acc=0.2748, loss=3.8521, lr=9.99e-05]
Training:   4%|▍         | 440/11082 [1:25:07<33:53:30, 11.46s/step, acc=0.2748, loss=3.8521, lr=9.99e-05]
Training:   4%|▍         | 441/11082 [1:25:19<34:03:35, 11.52s/step, acc=0.2748, loss=3.8521, lr=9.99e-05]
Training:   4%|▍         | 442/11082 [1:25:30<34:09:09, 11.56s/step, acc=0.2748, loss=3.8521, lr=9.99e-05]
Training:   4%|▍         | 443/11082 [1:25:41<33:47:08, 11.43s/step, acc=0.2748, loss=3.8521, lr=9.99e-05]
Training:   4%|▍         | 444/11082 [1:25:53<34:10:24, 11.56s/step, acc=0.2748, loss=3.8521, lr=9.99e-05]
Training:   4%|▍         | 445/11082 [1:26:05<34:07:34, 11.55s/step, acc=0.2748, loss=3.8521, lr=9.99e-05]
Training:   4%|▍         | 446/11082 [1:26:16<34:02:20, 11.52s/step, acc=0.2748, loss=3.8521, lr=9.99e-05]
Training:   4%|▍         | 447/11082 [1:26:28<33:56:33, 11.49s/step, acc=0.2748, loss=3.8521, lr=9.99e-05]
Training:   4%|▍         | 448/11082 [1:26:39<34:06:38, 11.55s/step, acc=0.2748, loss=3.8521, lr=9.99e-05]
Training:   4%|▍         | 449/11082 [1:26:51<34:08:25, 11.56s/step, acc=0.2748, loss=3.8521, lr=9.99e-05]
Training:   4%|▍         | 450/11082 [1:27:03<34:09:09, 11.56s/step, acc=0.2748, loss=3.8521, lr=9.99e-05]
Training:   4%|▍         | 450/11082 [1:27:10<34:09:09, 11.56s/step, acc=0.2910, loss=3.6869, lr=9.98e-05]2026-06-18 16:20:38 | INFO | __main__ | epoch=0 step=450 train_loss=3.686922 train_acc=0.291025 lr=9.985e-05 grad_norm=1.4375 acc_d0=0.5534 acc_d1=0.3726 acc_d2=0.2483 acc_d3=0.1763 acc_d4=0.1365 acc_d5=0.1143 acc_d6=0.1002 acc_d7=0.0908

Training:   4%|▍         | 451/11082 [1:27:14<34:12:49, 11.59s/step, acc=0.2910, loss=3.6869, lr=9.98e-05]
Training:   4%|▍         | 452/11082 [1:27:25<33:53:22, 11.48s/step, acc=0.2910, loss=3.6869, lr=9.98e-05]
Training:   4%|▍         | 453/11082 [1:27:37<34:09:54, 11.57s/step, acc=0.2910, loss=3.6869, lr=9.98e-05]
Training:   4%|▍         | 454/11082 [1:27:49<34:17:39, 11.62s/step, acc=0.2910, loss=3.6869, lr=9.98e-05]
Training:   4%|▍         | 455/11082 [1:28:01<34:31:22, 11.69s/step, acc=0.2910, loss=3.6869, lr=9.98e-05]
Training:   4%|▍         | 456/11082 [1:28:13<34:48:15, 11.79s/step, acc=0.2910, loss=3.6869, lr=9.98e-05]
Training:   4%|▍         | 457/11082 [1:28:24<34:29:48, 11.69s/step, acc=0.2910, loss=3.6869, lr=9.98e-05]
Training:   4%|▍         | 458/11082 [1:28:36<34:48:19, 11.79s/step, acc=0.2910, loss=3.6869, lr=9.98e-05]
Training:   4%|▍         | 459/11082 [1:28:48<34:26:52, 11.67s/step, acc=0.2910, loss=3.6869, lr=9.98e-05]
Training:   4%|▍         | 460/11082 [1:29:00<34:36:04, 11.73s/step, acc=0.2910, loss=3.6869, lr=9.98e-05]
Training:   4%|▍         | 461/11082 [1:29:11<34:21:09, 11.64s/step, acc=0.2910, loss=3.6869, lr=9.98e-05]
Training:   4%|▍         | 462/11082 [1:29:23<34:16:36, 11.62s/step, acc=0.2910, loss=3.6869, lr=9.98e-05]
Training:   4%|▍         | 463/11082 [1:29:34<33:57:33, 11.51s/step, acc=0.2910, loss=3.6869, lr=9.98e-05]
Training:   4%|▍         | 464/11082 [1:29:45<34:03:38, 11.55s/step, acc=0.2910, loss=3.6869, lr=9.98e-05]
Training:   4%|▍         | 465/11082 [1:29:57<34:01:04, 11.53s/step, acc=0.2910, loss=3.6869, lr=9.98e-05]
Training:   4%|▍         | 466/11082 [1:30:09<34:15:46, 11.62s/step, acc=0.2910, loss=3.6869, lr=9.98e-05]
Training:   4%|▍         | 467/11082 [1:30:20<34:00:14, 11.53s/step, acc=0.2910, loss=3.6869, lr=9.98e-05]
Training:   4%|▍         | 468/11082 [1:30:32<34:00:53, 11.54s/step, acc=0.2910, loss=3.6869, lr=9.98e-05]
Training:   4%|▍         | 469/11082 [1:30:43<33:53:10, 11.49s/step, acc=0.2910, loss=3.6869, lr=9.98e-05]
Training:   4%|▍         | 470/11082 [1:30:55<34:11:52, 11.60s/step, acc=0.2910, loss=3.6869, lr=9.98e-05]
Training:   4%|▍         | 471/11082 [1:31:06<34:00:16, 11.54s/step, acc=0.2910, loss=3.6869, lr=9.98e-05]
Training:   4%|▍         | 472/11082 [1:31:17<33:40:41, 11.43s/step, acc=0.2910, loss=3.6869, lr=9.98e-05]
Training:   4%|▍         | 473/11082 [1:31:29<33:56:00, 11.51s/step, acc=0.2910, loss=3.6869, lr=9.98e-05]
Training:   4%|▍         | 474/11082 [1:31:41<34:09:59, 11.59s/step, acc=0.2910, loss=3.6869, lr=9.98e-05]
Training:   4%|▍         | 475/11082 [1:31:52<34:05:11, 11.57s/step, acc=0.2910, loss=3.6869, lr=9.98e-05]
Training:   4%|▍         | 476/11082 [1:32:04<33:46:42, 11.47s/step, acc=0.2910, loss=3.6869, lr=9.98e-05]
Training:   4%|▍         | 477/11082 [1:32:16<34:17:21, 11.64s/step, acc=0.2910, loss=3.6869, lr=9.98e-05]
Training:   4%|▍         | 478/11082 [1:32:27<33:55:37, 11.52s/step, acc=0.2910, loss=3.6869, lr=9.98e-05]
Training:   4%|▍         | 479/11082 [1:32:39<34:20:55, 11.66s/step, acc=0.2910, loss=3.6869, lr=9.98e-05]
Training:   4%|▍         | 480/11082 [1:32:50<34:02:44, 11.56s/step, acc=0.2910, loss=3.6869, lr=9.98e-05]
Training:   4%|▍         | 481/11082 [1:33:03<34:51:11, 11.84s/step, acc=0.2910, loss=3.6869, lr=9.98e-05]
Training:   4%|▍         | 482/11082 [1:33:14<34:13:26, 11.62s/step, acc=0.2910, loss=3.6869, lr=9.98e-05]
Training:   4%|▍         | 483/11082 [1:33:25<34:03:21, 11.57s/step, acc=0.2910, loss=3.6869, lr=9.98e-05]
Training:   4%|▍         | 484/11082 [1:33:37<33:51:32, 11.50s/step, acc=0.2910, loss=3.6869, lr=9.98e-05]
Training:   4%|▍         | 485/11082 [1:33:49<34:15:14, 11.64s/step, acc=0.2910, loss=3.6869, lr=9.98e-05]
Training:   4%|▍         | 486/11082 [1:34:00<34:06:18, 11.59s/step, acc=0.2910, loss=3.6869, lr=9.98e-05]
Training:   4%|▍         | 487/11082 [1:34:11<33:45:48, 11.47s/step, acc=0.2910, loss=3.6869, lr=9.98e-05]
Training:   4%|▍         | 488/11082 [1:34:23<33:56:53, 11.54s/step, acc=0.2910, loss=3.6869, lr=9.98e-05]
Training:   4%|▍         | 489/11082 [1:34:34<33:52:37, 11.51s/step, acc=0.2910, loss=3.6869, lr=9.98e-05]
Training:   4%|▍         | 490/11082 [1:34:46<33:59:40, 11.55s/step, acc=0.2910, loss=3.6869, lr=9.98e-05]
Training:   4%|▍         | 491/11082 [1:34:58<34:06:28, 11.59s/step, acc=0.2910, loss=3.6869, lr=9.98e-05]
Training:   4%|▍         | 492/11082 [1:35:09<33:56:55, 11.54s/step, acc=0.2910, loss=3.6869, lr=9.98e-05]
Training:   4%|▍         | 493/11082 [1:35:21<33:52:58, 11.52s/step, acc=0.2910, loss=3.6869, lr=9.98e-05]
Training:   4%|▍         | 494/11082 [1:35:32<33:44:56, 11.47s/step, acc=0.2910, loss=3.6869, lr=9.98e-05]
Training:   4%|▍         | 495/11082 [1:35:44<34:14:52, 11.65s/step, acc=0.2910, loss=3.6869, lr=9.98e-05]
Training:   4%|▍         | 496/11082 [1:35:55<33:47:06, 11.49s/step, acc=0.2910, loss=3.6869, lr=9.98e-05]
Training:   4%|▍         | 497/11082 [1:36:07<34:07:57, 11.61s/step, acc=0.2910, loss=3.6869, lr=9.98e-05]
Training:   4%|▍         | 498/11082 [1:36:19<34:03:33, 11.58s/step, acc=0.2910, loss=3.6869, lr=9.98e-05]
Training:   5%|▍         | 499/11082 [1:36:30<34:14:20, 11.65s/step, acc=0.2910, loss=3.6869, lr=9.98e-05]
Training:   5%|▍         | 500/11082 [1:36:42<33:43:25, 11.47s/step, acc=0.2910, loss=3.6869, lr=9.98e-05]
Training:   5%|▍         | 500/11082 [1:36:49<33:43:25, 11.47s/step, acc=0.3052, loss=3.5573, lr=9.98e-05]2026-06-18 16:30:17 | INFO | __main__ | epoch=0 step=500 train_loss=3.557290 train_acc=0.305217 lr=9.979e-05 grad_norm=1.9062 acc_d0=0.5767 acc_d1=0.3948 acc_d2=0.2670 acc_d3=0.1902 acc_d4=0.1451 acc_d5=0.1206 acc_d6=0.1061 acc_d7=0.0932

Training:   5%|▍         | 501/11082 [1:36:53<33:37:14, 11.44s/step, acc=0.3052, loss=3.5573, lr=9.98e-05]
Training:   5%|▍         | 502/11082 [1:37:05<34:04:31, 11.59s/step, acc=0.3052, loss=3.5573, lr=9.98e-05]
Training:   5%|▍         | 503/11082 [1:37:16<34:00:55, 11.58s/step, acc=0.3052, loss=3.5573, lr=9.98e-05]
Training:   5%|▍         | 504/11082 [1:37:28<33:40:34, 11.46s/step, acc=0.3052, loss=3.5573, lr=9.98e-05]
Training:   5%|▍         | 505/11082 [1:37:39<34:02:54, 11.59s/step, acc=0.3052, loss=3.5573, lr=9.98e-05]
Training:   5%|▍         | 506/11082 [1:37:51<33:45:15, 11.49s/step, acc=0.3052, loss=3.5573, lr=9.98e-05]
Training:   5%|▍         | 507/11082 [1:38:02<33:46:17, 11.50s/step, acc=0.3052, loss=3.5573, lr=9.98e-05]
Training:   5%|▍         | 508/11082 [1:38:14<34:08:01, 11.62s/step, acc=0.3052, loss=3.5573, lr=9.98e-05]
Training:   5%|▍         | 509/11082 [1:38:25<33:45:55, 11.50s/step, acc=0.3052, loss=3.5573, lr=9.98e-05]
Training:   5%|▍         | 510/11082 [1:38:37<34:06:00, 11.61s/step, acc=0.3052, loss=3.5573, lr=9.98e-05]
Training:   5%|▍         | 511/11082 [1:38:49<33:49:59, 11.52s/step, acc=0.3052, loss=3.5573, lr=9.98e-05]
Training:   5%|▍         | 512/11082 [1:39:00<33:41:23, 11.47s/step, acc=0.3052, loss=3.5573, lr=9.98e-05]
Training:   5%|▍         | 513/11082 [1:39:11<33:30:34, 11.41s/step, acc=0.3052, loss=3.5573, lr=9.98e-05]
Training:   5%|▍         | 514/11082 [1:39:23<33:52:33, 11.54s/step, acc=0.3052, loss=3.5573, lr=9.98e-05]
Training:   5%|▍         | 515/11082 [1:39:35<34:00:31, 11.59s/step, acc=0.3052, loss=3.5573, lr=9.98e-05]
Training:   5%|▍         | 516/11082 [1:39:46<34:06:38, 11.62s/step, acc=0.3052, loss=3.5573, lr=9.98e-05]
Training:   5%|▍         | 517/11082 [1:39:58<33:41:58, 11.48s/step, acc=0.3052, loss=3.5573, lr=9.98e-05]
Training:   5%|▍         | 518/11082 [1:40:09<33:56:39, 11.57s/step, acc=0.3052, loss=3.5573, lr=9.98e-05]
Training:   5%|▍         | 519/11082 [1:40:21<33:58:14, 11.58s/step, acc=0.3052, loss=3.5573, lr=9.98e-05]
Training:   5%|▍         | 520/11082 [1:40:32<33:47:07, 11.52s/step, acc=0.3052, loss=3.5573, lr=9.98e-05]
Training:   5%|▍         | 521/11082 [1:40:44<33:53:42, 11.55s/step, acc=0.3052, loss=3.5573, lr=9.98e-05]
Training:   5%|▍         | 522/11082 [1:40:56<34:02:27, 11.60s/step, acc=0.3052, loss=3.5573, lr=9.98e-05]
Training:   5%|▍         | 523/11082 [1:41:07<33:43:00, 11.50s/step, acc=0.3052, loss=3.5573, lr=9.98e-05]
Training:   5%|▍         | 524/11082 [1:41:19<33:56:03, 11.57s/step, acc=0.3052, loss=3.5573, lr=9.98e-05]
Training:   5%|▍         | 525/11082 [1:41:30<33:39:47, 11.48s/step, acc=0.3052, loss=3.5573, lr=9.98e-05]
Training:   5%|▍         | 526/11082 [1:41:41<33:37:45, 11.47s/step, acc=0.3052, loss=3.5573, lr=9.98e-05]
Training:   5%|▍         | 527/11082 [1:41:53<33:57:20, 11.58s/step, acc=0.3052, loss=3.5573, lr=9.98e-05]
Training:   5%|▍         | 528/11082 [1:42:05<33:48:42, 11.53s/step, acc=0.3052, loss=3.5573, lr=9.98e-05]
Training:   5%|▍         | 529/11082 [1:42:16<34:05:19, 11.63s/step, acc=0.3052, loss=3.5573, lr=9.98e-05]
Training:   5%|▍         | 530/11082 [1:42:28<33:57:25, 11.59s/step, acc=0.3052, loss=3.5573, lr=9.98e-05]
Training:   5%|▍         | 531/11082 [1:42:39<33:52:17, 11.56s/step, acc=0.3052, loss=3.5573, lr=9.98e-05]
Training:   5%|▍         | 532/11082 [1:42:51<33:41:01, 11.49s/step, acc=0.3052, loss=3.5573, lr=9.98e-05]
Training:   5%|▍         | 533/11082 [1:43:03<33:58:10, 11.59s/step, acc=0.3052, loss=3.5573, lr=9.98e-05]
Training:   5%|▍         | 534/11082 [1:43:14<33:58:18, 11.59s/step, acc=0.3052, loss=3.5573, lr=9.98e-05]
Training:   5%|▍         | 535/11082 [1:43:26<34:06:09, 11.64s/step, acc=0.3052, loss=3.5573, lr=9.98e-05]
Training:   5%|▍         | 536/11082 [1:43:38<34:23:23, 11.74s/step, acc=0.3052, loss=3.5573, lr=9.98e-05]
Training:   5%|▍         | 537/11082 [1:43:49<34:00:07, 11.61s/step, acc=0.3052, loss=3.5573, lr=9.98e-05]
Training:   5%|▍         | 538/11082 [1:44:01<33:58:14, 11.60s/step, acc=0.3052, loss=3.5573, lr=9.98e-05]
Training:   5%|▍         | 539/11082 [1:44:13<34:21:45, 11.73s/step, acc=0.3052, loss=3.5573, lr=9.98e-05]
Training:   5%|▍         | 540/11082 [1:44:24<34:03:36, 11.63s/step, acc=0.3052, loss=3.5573, lr=9.98e-05]
Training:   5%|▍         | 541/11082 [1:44:36<34:02:26, 11.63s/step, acc=0.3052, loss=3.5573, lr=9.98e-05]
Training:   5%|▍         | 542/11082 [1:44:48<34:08:26, 11.66s/step, acc=0.3052, loss=3.5573, lr=9.98e-05]
Training:   5%|▍         | 543/11082 [1:44:59<33:53:58, 11.58s/step, acc=0.3052, loss=3.5573, lr=9.98e-05]
Training:   5%|▍         | 544/11082 [1:45:11<33:59:57, 11.61s/step, acc=0.3052, loss=3.5573, lr=9.98e-05]
Training:   5%|▍         | 545/11082 [1:45:22<33:49:25, 11.56s/step, acc=0.3052, loss=3.5573, lr=9.98e-05]
Training:   5%|▍         | 546/11082 [1:45:34<33:54:36, 11.59s/step, acc=0.3052, loss=3.5573, lr=9.98e-05]
Training:   5%|▍         | 547/11082 [1:45:45<33:36:48, 11.49s/step, acc=0.3052, loss=3.5573, lr=9.98e-05]
Training:   5%|▍         | 548/11082 [1:45:56<33:26:22, 11.43s/step, acc=0.3052, loss=3.5573, lr=9.98e-05]
Training:   5%|▍         | 549/11082 [1:46:08<33:53:45, 11.59s/step, acc=0.3052, loss=3.5573, lr=9.98e-05]
Training:   5%|▍         | 550/11082 [1:46:20<33:43:53, 11.53s/step, acc=0.3052, loss=3.5573, lr=9.98e-05]
Training:   5%|▍         | 550/11082 [1:46:27<33:43:53, 11.53s/step, acc=0.3133, loss=3.4794, lr=9.97e-05]2026-06-18 16:39:55 | INFO | __main__ | epoch=0 step=550 train_loss=3.479364 train_acc=0.313255 lr=9.973e-05 grad_norm=1.4453 acc_d0=0.5690 acc_d1=0.3937 acc_d2=0.2669 acc_d3=0.1879 acc_d4=0.1441 acc_d5=0.1186 acc_d6=0.1031 acc_d7=0.0915

Training:   5%|▍         | 551/11082 [1:46:32<34:01:04, 11.63s/step, acc=0.3133, loss=3.4794, lr=9.97e-05]
Training:   5%|▍         | 552/11082 [1:46:43<33:50:08, 11.57s/step, acc=0.3133, loss=3.4794, lr=9.97e-05]
Training:   5%|▍         | 553/11082 [1:46:55<33:55:20, 11.60s/step, acc=0.3133, loss=3.4794, lr=9.97e-05]
Training:   5%|▍         | 554/11082 [1:47:06<33:56:19, 11.61s/step, acc=0.3133, loss=3.4794, lr=9.97e-05]
Training:   5%|▌         | 555/11082 [1:47:17<33:17:29, 11.38s/step, acc=0.3133, loss=3.4794, lr=9.97e-05]
Training:   5%|▌         | 556/11082 [1:47:29<33:45:32, 11.55s/step, acc=0.3133, loss=3.4794, lr=9.97e-05]
Training:   5%|▌         | 557/11082 [1:47:40<33:30:48, 11.46s/step, acc=0.3133, loss=3.4794, lr=9.97e-05]
Training:   5%|▌         | 558/11082 [1:47:52<33:37:25, 11.50s/step, acc=0.3133, loss=3.4794, lr=9.97e-05]
Training:   5%|▌         | 559/11082 [1:48:04<33:54:40, 11.60s/step, acc=0.3133, loss=3.4794, lr=9.97e-05]
Training:   5%|▌         | 560/11082 [1:48:14<33:06:59, 11.33s/step, acc=0.3133, loss=3.4794, lr=9.97e-05]
Training:   5%|▌         | 561/11082 [1:48:26<33:45:00, 11.55s/step, acc=0.3133, loss=3.4794, lr=9.97e-05]
Training:   5%|▌         | 562/11082 [1:48:38<33:54:10, 11.60s/step, acc=0.3133, loss=3.4794, lr=9.97e-05]
Training:   5%|▌         | 563/11082 [1:48:49<33:31:27, 11.47s/step, acc=0.3133, loss=3.4794, lr=9.97e-05]
Training:   5%|▌         | 564/11082 [1:49:01<33:50:11, 11.58s/step, acc=0.3133, loss=3.4794, lr=9.97e-05]
Training:   5%|▌         | 565/11082 [1:49:13<33:41:33, 11.53s/step, acc=0.3133, loss=3.4794, lr=9.97e-05]
Training:   5%|▌         | 566/11082 [1:49:25<33:59:58, 11.64s/step, acc=0.3133, loss=3.4794, lr=9.97e-05]
Training:   5%|▌         | 567/11082 [1:49:36<33:49:00, 11.58s/step, acc=0.3133, loss=3.4794, lr=9.97e-05]
Training:   5%|▌         | 568/11082 [1:49:47<33:39:17, 11.52s/step, acc=0.3133, loss=3.4794, lr=9.97e-05]
Training:   5%|▌         | 569/11082 [1:49:59<33:25:00, 11.44s/step, acc=0.3133, loss=3.4794, lr=9.97e-05]
Training:   5%|▌         | 570/11082 [1:50:10<33:25:36, 11.45s/step, acc=0.3133, loss=3.4794, lr=9.97e-05]
Training:   5%|▌         | 571/11082 [1:50:21<33:17:27, 11.40s/step, acc=0.3133, loss=3.4794, lr=9.97e-05]
Training:   5%|▌         | 572/11082 [1:50:33<33:40:16, 11.53s/step, acc=0.3133, loss=3.4794, lr=9.97e-05]
Training:   5%|▌         | 573/11082 [1:50:45<33:39:19, 11.53s/step, acc=0.3133, loss=3.4794, lr=9.97e-05]
Training:   5%|▌         | 574/11082 [1:50:56<33:31:25, 11.49s/step, acc=0.3133, loss=3.4794, lr=9.97e-05]
Training:   5%|▌         | 575/11082 [1:51:08<33:37:34, 11.52s/step, acc=0.3133, loss=3.4794, lr=9.97e-05]
Training:   5%|▌         | 576/11082 [1:51:20<33:57:37, 11.64s/step, acc=0.3133, loss=3.4794, lr=9.97e-05]
Training:   5%|▌         | 577/11082 [1:51:31<33:36:46, 11.52s/step, acc=0.3133, loss=3.4794, lr=9.97e-05]
Training:   5%|▌         | 578/11082 [1:51:42<33:23:14, 11.44s/step, acc=0.3133, loss=3.4794, lr=9.97e-05]
Training:   5%|▌         | 579/11082 [1:51:54<33:30:00, 11.48s/step, acc=0.3133, loss=3.4794, lr=9.97e-05]
Training:   5%|▌         | 580/11082 [1:52:06<33:55:07, 11.63s/step, acc=0.3133, loss=3.4794, lr=9.97e-05]
Training:   5%|▌         | 581/11082 [1:52:17<33:31:00, 11.49s/step, acc=0.3133, loss=3.4794, lr=9.97e-05]
Training:   5%|▌         | 582/11082 [1:52:28<33:19:49, 11.43s/step, acc=0.3133, loss=3.4794, lr=9.97e-05]
Training:   5%|▌         | 583/11082 [1:52:40<33:37:02, 11.53s/step, acc=0.3133, loss=3.4794, lr=9.97e-05]
Training:   5%|▌         | 584/11082 [1:52:51<33:40:44, 11.55s/step, acc=0.3133, loss=3.4794, lr=9.97e-05]
Training:   5%|▌         | 585/11082 [1:53:03<33:37:14, 11.53s/step, acc=0.3133, loss=3.4794, lr=9.97e-05]
Training:   5%|▌         | 586/11082 [1:53:15<33:44:32, 11.57s/step, acc=0.3133, loss=3.4794, lr=9.97e-05]
Training:   5%|▌         | 587/11082 [1:53:26<33:39:39, 11.55s/step, acc=0.3133, loss=3.4794, lr=9.97e-05]
Training:   5%|▌         | 588/11082 [1:53:38<33:54:20, 11.63s/step, acc=0.3133, loss=3.4794, lr=9.97e-05]
Training:   5%|▌         | 589/11082 [1:53:49<33:33:33, 11.51s/step, acc=0.3133, loss=3.4794, lr=9.97e-05]
Training:   5%|▌         | 590/11082 [1:54:01<33:31:55, 11.51s/step, acc=0.3133, loss=3.4794, lr=9.97e-05]
Training:   5%|▌         | 591/11082 [1:54:12<33:47:00, 11.59s/step, acc=0.3133, loss=3.4794, lr=9.97e-05]
Training:   5%|▌         | 592/11082 [1:54:24<33:29:57, 11.50s/step, acc=0.3133, loss=3.4794, lr=9.97e-05]
Training:   5%|▌         | 593/11082 [1:54:35<33:17:28, 11.43s/step, acc=0.3133, loss=3.4794, lr=9.97e-05]
Training:   5%|▌         | 594/11082 [1:54:47<33:51:00, 11.62s/step, acc=0.3133, loss=3.4794, lr=9.97e-05]
Training:   5%|▌         | 595/11082 [1:54:58<33:37:25, 11.54s/step, acc=0.3133, loss=3.4794, lr=9.97e-05]
Training:   5%|▌         | 596/11082 [1:55:10<33:20:02, 11.44s/step, acc=0.3133, loss=3.4794, lr=9.97e-05]
Training:   5%|▌         | 597/11082 [1:55:22<33:41:19, 11.57s/step, acc=0.3133, loss=3.4794, lr=9.97e-05]
Training:   5%|▌         | 598/11082 [1:55:33<33:38:52, 11.55s/step, acc=0.3133, loss=3.4794, lr=9.97e-05]
Training:   5%|▌         | 599/11082 [1:55:45<33:50:52, 11.62s/step, acc=0.3133, loss=3.4794, lr=9.97e-05]
Training:   5%|▌         | 600/11082 [1:55:57<33:55:01, 11.65s/step, acc=0.3133, loss=3.4794, lr=9.97e-05]
Training:   5%|▌         | 600/11082 [1:56:04<33:55:01, 11.65s/step, acc=0.3216, loss=3.4014, lr=9.96e-05]2026-06-18 16:49:32 | INFO | __main__ | epoch=0 step=600 train_loss=3.401385 train_acc=0.321583 lr=9.965e-05 grad_norm=1.7578 acc_d0=0.5726 acc_d1=0.4011 acc_d2=0.2741 acc_d3=0.1951 acc_d4=0.1499 acc_d5=0.1235 acc_d6=0.1074 acc_d7=0.0948

Training:   5%|▌         | 601/11082 [1:56:08<33:52:41, 11.64s/step, acc=0.3216, loss=3.4014, lr=9.96e-05]
Training:   5%|▌         | 602/11082 [1:56:20<34:06:07, 11.71s/step, acc=0.3216, loss=3.4014, lr=9.96e-05]
Training:   5%|▌         | 603/11082 [1:56:32<34:03:18, 11.70s/step, acc=0.3216, loss=3.4014, lr=9.96e-05]
Training:   5%|▌         | 604/11082 [1:56:43<33:22:14, 11.47s/step, acc=0.3216, loss=3.4014, lr=9.96e-05]
Training:   5%|▌         | 605/11082 [1:56:54<33:35:31, 11.54s/step, acc=0.3216, loss=3.4014, lr=9.96e-05]
Training:   5%|▌         | 606/11082 [1:57:06<33:19:46, 11.45s/step, acc=0.3216, loss=3.4014, lr=9.96e-05]
Training:   5%|▌         | 607/11082 [1:57:18<33:47:32, 11.61s/step, acc=0.3216, loss=3.4014, lr=9.96e-05]
Training:   5%|▌         | 608/11082 [1:57:29<33:31:19, 11.52s/step, acc=0.3216, loss=3.4014, lr=9.96e-05]
Training:   5%|▌         | 609/11082 [1:57:40<33:29:13, 11.51s/step, acc=0.3216, loss=3.4014, lr=9.96e-05]
Training:   6%|▌         | 610/11082 [1:57:52<33:47:56, 11.62s/step, acc=0.3216, loss=3.4014, lr=9.96e-05]
Training:   6%|▌         | 611/11082 [1:58:04<33:48:40, 11.62s/step, acc=0.3216, loss=3.4014, lr=9.96e-05]
Training:   6%|▌         | 612/11082 [1:58:15<33:26:28, 11.50s/step, acc=0.3216, loss=3.4014, lr=9.96e-05]
Training:   6%|▌         | 613/11082 [1:58:27<33:24:36, 11.49s/step, acc=0.3216, loss=3.4014, lr=9.96e-05]
Training:   6%|▌         | 614/11082 [1:58:38<33:34:08, 11.54s/step, acc=0.3216, loss=3.4014, lr=9.96e-05]
Training:   6%|▌         | 615/11082 [1:58:50<33:27:10, 11.51s/step, acc=0.3216, loss=3.4014, lr=9.96e-05]
Training:   6%|▌         | 616/11082 [1:59:01<33:30:59, 11.53s/step, acc=0.3216, loss=3.4014, lr=9.96e-05]
Training:   6%|▌         | 617/11082 [1:59:13<33:37:07, 11.56s/step, acc=0.3216, loss=3.4014, lr=9.96e-05]
Training:   6%|▌         | 618/11082 [1:59:24<33:34:26, 11.55s/step, acc=0.3216, loss=3.4014, lr=9.96e-05]
Training:   6%|▌         | 619/11082 [1:59:36<33:51:39, 11.65s/step, acc=0.3216, loss=3.4014, lr=9.96e-05]
Training:   6%|▌         | 620/11082 [1:59:47<33:28:49, 11.52s/step, acc=0.3216, loss=3.4014, lr=9.96e-05]
Training:   6%|▌         | 621/11082 [2:00:00<34:26:01, 11.85s/step, acc=0.3216, loss=3.4014, lr=9.96e-05]
Training:   6%|▌         | 622/11082 [2:00:11<33:56:01, 11.68s/step, acc=0.3216, loss=3.4014, lr=9.96e-05]
Training:   6%|▌         | 623/11082 [2:00:23<33:26:39, 11.51s/step, acc=0.3216, loss=3.4014, lr=9.96e-05]
Training:   6%|▌         | 624/11082 [2:00:35<33:52:15, 11.66s/step, acc=0.3216, loss=3.4014, lr=9.96e-05]
Training:   6%|▌         | 625/11082 [2:00:46<33:58:33, 11.70s/step, acc=0.3216, loss=3.4014, lr=9.96e-05]
Training:   6%|▌         | 626/11082 [2:00:58<33:48:27, 11.64s/step, acc=0.3216, loss=3.4014, lr=9.96e-05]
Training:   6%|▌         | 627/11082 [2:01:09<33:14:47, 11.45s/step, acc=0.3216, loss=3.4014, lr=9.96e-05]
Training:   6%|▌         | 628/11082 [2:01:20<33:18:38, 11.47s/step, acc=0.3216, loss=3.4014, lr=9.96e-05]
Training:   6%|▌         | 629/11082 [2:01:32<33:35:50, 11.57s/step, acc=0.3216, loss=3.4014, lr=9.96e-05]
Training:   6%|▌         | 630/11082 [2:01:44<33:39:44, 11.59s/step, acc=0.3216, loss=3.4014, lr=9.96e-05]
Training:   6%|▌         | 631/11082 [2:01:56<33:47:18, 11.64s/step, acc=0.3216, loss=3.4014, lr=9.96e-05]
Training:   6%|▌         | 632/11082 [2:02:07<33:17:41, 11.47s/step, acc=0.3216, loss=3.4014, lr=9.96e-05]
Training:   6%|▌         | 633/11082 [2:02:18<33:31:40, 11.55s/step, acc=0.3216, loss=3.4014, lr=9.96e-05]
Training:   6%|▌         | 634/11082 [2:02:30<33:40:35, 11.60s/step, acc=0.3216, loss=3.4014, lr=9.96e-05]
Training:   6%|▌         | 635/11082 [2:02:41<33:17:11, 11.47s/step, acc=0.3216, loss=3.4014, lr=9.96e-05]
Training:   6%|▌         | 636/11082 [2:02:53<33:29:47, 11.54s/step, acc=0.3216, loss=3.4014, lr=9.96e-05]
Training:   6%|▌         | 637/11082 [2:03:05<33:32:30, 11.56s/step, acc=0.3216, loss=3.4014, lr=9.96e-05]
Training:   6%|▌         | 638/11082 [2:03:16<33:13:12, 11.45s/step, acc=0.3216, loss=3.4014, lr=9.96e-05]
Training:   6%|▌         | 639/11082 [2:03:28<33:43:27, 11.63s/step, acc=0.3216, loss=3.4014, lr=9.96e-05]
Training:   6%|▌         | 640/11082 [2:03:39<33:24:31, 11.52s/step, acc=0.3216, loss=3.4014, lr=9.96e-05]
Training:   6%|▌         | 641/11082 [2:03:50<33:19:57, 11.49s/step, acc=0.3216, loss=3.4014, lr=9.96e-05]
Training:   6%|▌         | 642/11082 [2:04:02<33:34:43, 11.58s/step, acc=0.3216, loss=3.4014, lr=9.96e-05]
Training:   6%|▌         | 643/11082 [2:04:14<33:53:31, 11.69s/step, acc=0.3216, loss=3.4014, lr=9.96e-05]
Training:   6%|▌         | 644/11082 [2:04:25<33:18:24, 11.49s/step, acc=0.3216, loss=3.4014, lr=9.96e-05]
Training:   6%|▌         | 645/11082 [2:04:37<33:36:50, 11.59s/step, acc=0.3216, loss=3.4014, lr=9.96e-05]
Training:   6%|▌         | 646/11082 [2:04:48<33:22:43, 11.51s/step, acc=0.3216, loss=3.4014, lr=9.96e-05]
Training:   6%|▌         | 647/11082 [2:05:00<33:11:31, 11.45s/step, acc=0.3216, loss=3.4014, lr=9.96e-05]
Training:   6%|▌         | 648/11082 [2:05:11<33:25:23, 11.53s/step, acc=0.3216, loss=3.4014, lr=9.96e-05]
Training:   6%|▌         | 649/11082 [2:05:23<33:37:32, 11.60s/step, acc=0.3216, loss=3.4014, lr=9.96e-05]
Training:   6%|▌         | 650/11082 [2:05:35<33:32:33, 11.58s/step, acc=0.3216, loss=3.4014, lr=9.96e-05]
Training:   6%|▌         | 650/11082 [2:05:42<33:32:33, 11.58s/step, acc=0.3313, loss=3.3095, lr=9.96e-05]2026-06-18 16:59:10 | INFO | __main__ | epoch=0 step=650 train_loss=3.309494 train_acc=0.331318 lr=9.956e-05 grad_norm=3.9062 acc_d0=0.5912 acc_d1=0.4176 acc_d2=0.2881 acc_d3=0.2027 acc_d4=0.1543 acc_d5=0.1262 acc_d6=0.1079 acc_d7=0.0970

Training:   6%|▌         | 651/11082 [2:05:46<33:14:14, 11.47s/step, acc=0.3313, loss=3.3095, lr=9.96e-05]
Training:   6%|▌         | 652/11082 [2:05:57<33:13:50, 11.47s/step, acc=0.3313, loss=3.3095, lr=9.96e-05]
Training:   6%|▌         | 653/11082 [2:06:09<33:28:03, 11.55s/step, acc=0.3313, loss=3.3095, lr=9.96e-05]
Training:   6%|▌         | 654/11082 [2:06:21<33:29:37, 11.56s/step, acc=0.3313, loss=3.3095, lr=9.96e-05]
Training:   6%|▌         | 655/11082 [2:06:33<33:40:34, 11.63s/step, acc=0.3313, loss=3.3095, lr=9.96e-05]
Training:   6%|▌         | 656/11082 [2:06:44<33:35:14, 11.60s/step, acc=0.3313, loss=3.3095, lr=9.96e-05]
Training:   6%|▌         | 657/11082 [2:06:55<33:05:49, 11.43s/step, acc=0.3313, loss=3.3095, lr=9.96e-05]
Training:   6%|▌         | 658/11082 [2:07:07<33:25:24, 11.54s/step, acc=0.3313, loss=3.3095, lr=9.96e-05]
Training:   6%|▌         | 659/11082 [2:07:18<33:19:20, 11.51s/step, acc=0.3313, loss=3.3095, lr=9.96e-05]
Training:   6%|▌         | 660/11082 [2:07:30<33:35:07, 11.60s/step, acc=0.3313, loss=3.3095, lr=9.96e-05]
Training:   6%|▌         | 661/11082 [2:07:41<33:18:23, 11.51s/step, acc=0.3313, loss=3.3095, lr=9.96e-05]
Training:   6%|▌         | 662/11082 [2:07:53<33:25:02, 11.55s/step, acc=0.3313, loss=3.3095, lr=9.96e-05]
Training:   6%|▌         | 663/11082 [2:08:04<33:16:26, 11.50s/step, acc=0.3313, loss=3.3095, lr=9.96e-05]
Training:   6%|▌         | 664/11082 [2:08:16<33:38:39, 11.63s/step, acc=0.3313, loss=3.3095, lr=9.96e-05]
Training:   6%|▌         | 665/11082 [2:08:27<33:06:02, 11.44s/step, acc=0.3313, loss=3.3095, lr=9.96e-05]
Training:   6%|▌         | 666/11082 [2:08:39<33:06:09, 11.44s/step, acc=0.3313, loss=3.3095, lr=9.96e-05]
Training:   6%|▌         | 667/11082 [2:08:51<33:45:33, 11.67s/step, acc=0.3313, loss=3.3095, lr=9.96e-05]
Training:   6%|▌         | 668/11082 [2:09:02<33:36:11, 11.62s/step, acc=0.3313, loss=3.3095, lr=9.96e-05]
Training:   6%|▌         | 669/11082 [2:09:14<33:20:06, 11.52s/step, acc=0.3313, loss=3.3095, lr=9.96e-05]
Training:   6%|▌         | 670/11082 [2:09:26<33:37:32, 11.63s/step, acc=0.3313, loss=3.3095, lr=9.96e-05]
Training:   6%|▌         | 671/11082 [2:09:37<33:44:52, 11.67s/step, acc=0.3313, loss=3.3095, lr=9.96e-05]
Training:   6%|▌         | 672/11082 [2:09:49<33:45:28, 11.67s/step, acc=0.3313, loss=3.3095, lr=9.96e-05]
Training:   6%|▌         | 673/11082 [2:10:01<33:46:44, 11.68s/step, acc=0.3313, loss=3.3095, lr=9.96e-05]
Training:   6%|▌         | 674/11082 [2:10:12<33:40:54, 11.65s/step, acc=0.3313, loss=3.3095, lr=9.96e-05]
Training:   6%|▌         | 675/11082 [2:10:24<33:36:11, 11.62s/step, acc=0.3313, loss=3.3095, lr=9.96e-05]
Training:   6%|▌         | 676/11082 [2:10:35<33:15:16, 11.50s/step, acc=0.3313, loss=3.3095, lr=9.96e-05]
Training:   6%|▌         | 677/11082 [2:10:47<33:32:03, 11.60s/step, acc=0.3313, loss=3.3095, lr=9.96e-05]
Training:   6%|▌         | 678/11082 [2:10:59<33:35:28, 11.62s/step, acc=0.3313, loss=3.3095, lr=9.96e-05]
Training:   6%|▌         | 679/11082 [2:11:10<33:07:34, 11.46s/step, acc=0.3313, loss=3.3095, lr=9.96e-05]
Training:   6%|▌         | 680/11082 [2:11:21<33:18:57, 11.53s/step, acc=0.3313, loss=3.3095, lr=9.96e-05]
Training:   6%|▌         | 681/11082 [2:11:33<33:25:38, 11.57s/step, acc=0.3313, loss=3.3095, lr=9.96e-05]
Training:   6%|▌         | 682/11082 [2:11:45<33:17:01, 11.52s/step, acc=0.3313, loss=3.3095, lr=9.96e-05]
Training:   6%|▌         | 683/11082 [2:11:56<32:59:52, 11.42s/step, acc=0.3313, loss=3.3095, lr=9.96e-05]
Training:   6%|▌         | 684/11082 [2:12:08<33:17:33, 11.53s/step, acc=0.3313, loss=3.3095, lr=9.96e-05]
Training:   6%|▌         | 685/11082 [2:12:19<33:07:45, 11.47s/step, acc=0.3313, loss=3.3095, lr=9.96e-05]
Training:   6%|▌         | 686/11082 [2:12:31<33:20:03, 11.54s/step, acc=0.3313, loss=3.3095, lr=9.96e-05]
Training:   6%|▌         | 687/11082 [2:12:42<33:22:54, 11.56s/step, acc=0.3313, loss=3.3095, lr=9.96e-05]
Training:   6%|▌         | 688/11082 [2:12:54<33:29:33, 11.60s/step, acc=0.3313, loss=3.3095, lr=9.96e-05]
Training:   6%|▌         | 689/11082 [2:13:05<33:15:44, 11.52s/step, acc=0.3313, loss=3.3095, lr=9.96e-05]
Training:   6%|▌         | 690/11082 [2:13:17<33:17:04, 11.53s/step, acc=0.3313, loss=3.3095, lr=9.96e-05]
Training:   6%|▌         | 691/11082 [2:13:28<33:13:01, 11.51s/step, acc=0.3313, loss=3.3095, lr=9.96e-05]
Training:   6%|▌         | 692/11082 [2:13:40<33:17:25, 11.53s/step, acc=0.3313, loss=3.3095, lr=9.96e-05]
Training:   6%|▋         | 693/11082 [2:13:51<33:13:20, 11.51s/step, acc=0.3313, loss=3.3095, lr=9.96e-05]
Training:   6%|▋         | 694/11082 [2:14:03<33:11:13, 11.50s/step, acc=0.3313, loss=3.3095, lr=9.96e-05]
Training:   6%|▋         | 695/11082 [2:14:14<33:13:24, 11.51s/step, acc=0.3313, loss=3.3095, lr=9.96e-05]
Training:   6%|▋         | 696/11082 [2:14:25<32:56:25, 11.42s/step, acc=0.3313, loss=3.3095, lr=9.96e-05]
Training:   6%|▋         | 697/11082 [2:14:37<33:22:04, 11.57s/step, acc=0.3313, loss=3.3095, lr=9.96e-05]
Training:   6%|▋         | 698/11082 [2:14:49<33:33:21, 11.63s/step, acc=0.3313, loss=3.3095, lr=9.96e-05]
Training:   6%|▋         | 699/11082 [2:15:01<33:33:31, 11.64s/step, acc=0.3313, loss=3.3095, lr=9.96e-05]
Training:   6%|▋         | 700/11082 [2:15:13<33:51:06, 11.74s/step, acc=0.3313, loss=3.3095, lr=9.96e-05]
Training:   6%|▋         | 700/11082 [2:15:20<33:51:06, 11.74s/step, acc=0.3366, loss=3.2720, lr=9.95e-05]2026-06-18 17:08:48 | INFO | __main__ | epoch=0 step=700 train_loss=3.271951 train_acc=0.336584 lr=9.947e-05 grad_norm=3.8125 acc_d0=0.5939 acc_d1=0.4210 acc_d2=0.2906 acc_d3=0.2073 acc_d4=0.1575 acc_d5=0.1294 acc_d6=0.1108 acc_d7=0.0977

Training:   6%|▋         | 701/11082 [2:15:24<33:45:17, 11.71s/step, acc=0.3366, loss=3.2720, lr=9.95e-05]
Training:   6%|▋         | 702/11082 [2:15:36<33:51:01, 11.74s/step, acc=0.3366, loss=3.2720, lr=9.95e-05]
Training:   6%|▋         | 703/11082 [2:15:48<33:38:32, 11.67s/step, acc=0.3366, loss=3.2720, lr=9.95e-05]
Training:   6%|▋         | 704/11082 [2:16:00<33:43:53, 11.70s/step, acc=0.3366, loss=3.2720, lr=9.95e-05]
Training:   6%|▋         | 705/11082 [2:16:11<33:14:21, 11.53s/step, acc=0.3366, loss=3.2720, lr=9.95e-05]
Training:   6%|▋         | 706/11082 [2:16:23<33:38:29, 11.67s/step, acc=0.3366, loss=3.2720, lr=9.95e-05]
Training:   6%|▋         | 707/11082 [2:16:34<33:12:59, 11.53s/step, acc=0.3366, loss=3.2720, lr=9.95e-05]
Training:   6%|▋         | 708/11082 [2:16:45<33:13:06, 11.53s/step, acc=0.3366, loss=3.2720, lr=9.95e-05]
Training:   6%|▋         | 709/11082 [2:16:57<33:02:52, 11.47s/step, acc=0.3366, loss=3.2720, lr=9.95e-05]
Training:   6%|▋         | 710/11082 [2:17:08<32:51:44, 11.41s/step, acc=0.3366, loss=3.2720, lr=9.95e-05]
Training:   6%|▋         | 711/11082 [2:17:20<33:23:13, 11.59s/step, acc=0.3366, loss=3.2720, lr=9.95e-05]
Training:   6%|▋         | 712/11082 [2:17:31<33:17:20, 11.56s/step, acc=0.3366, loss=3.2720, lr=9.95e-05]
Training:   6%|▋         | 713/11082 [2:17:43<33:19:50, 11.57s/step, acc=0.3366, loss=3.2720, lr=9.95e-05]
Training:   6%|▋         | 714/11082 [2:17:54<33:03:15, 11.48s/step, acc=0.3366, loss=3.2720, lr=9.95e-05]
Training:   6%|▋         | 715/11082 [2:18:06<33:02:57, 11.48s/step, acc=0.3366, loss=3.2720, lr=9.95e-05]
Training:   6%|▋         | 716/11082 [2:18:18<33:14:43, 11.55s/step, acc=0.3366, loss=3.2720, lr=9.95e-05]
Training:   6%|▋         | 717/11082 [2:18:29<33:05:29, 11.49s/step, acc=0.3366, loss=3.2720, lr=9.95e-05]
Training:   6%|▋         | 718/11082 [2:18:41<33:20:09, 11.58s/step, acc=0.3366, loss=3.2720, lr=9.95e-05]
Training:   6%|▋         | 719/11082 [2:18:52<33:02:28, 11.48s/step, acc=0.3366, loss=3.2720, lr=9.95e-05]
Training:   6%|▋         | 720/11082 [2:19:04<33:15:42, 11.56s/step, acc=0.3366, loss=3.2720, lr=9.95e-05]
Training:   7%|▋         | 721/11082 [2:19:15<33:25:52, 11.62s/step, acc=0.3366, loss=3.2720, lr=9.95e-05]
Training:   7%|▋         | 722/11082 [2:19:27<33:16:25, 11.56s/step, acc=0.3366, loss=3.2720, lr=9.95e-05]
Training:   7%|▋         | 723/11082 [2:19:38<33:11:24, 11.53s/step, acc=0.3366, loss=3.2720, lr=9.95e-05]
Training:   7%|▋         | 724/11082 [2:19:50<33:11:08, 11.53s/step, acc=0.3366, loss=3.2720, lr=9.95e-05]
Training:   7%|▋         | 725/11082 [2:20:02<33:22:05, 11.60s/step, acc=0.3366, loss=3.2720, lr=9.95e-05]
Training:   7%|▋         | 726/11082 [2:20:13<33:24:54, 11.62s/step, acc=0.3366, loss=3.2720, lr=9.95e-05]
Training:   7%|▋         | 727/11082 [2:20:24<32:49:23, 11.41s/step, acc=0.3366, loss=3.2720, lr=9.95e-05]
Training:   7%|▋         | 728/11082 [2:20:36<33:17:23, 11.57s/step, acc=0.3366, loss=3.2720, lr=9.95e-05]
Training:   7%|▋         | 729/11082 [2:20:48<33:09:58, 11.53s/step, acc=0.3366, loss=3.2720, lr=9.95e-05]
Training:   7%|▋         | 730/11082 [2:20:59<33:07:52, 11.52s/step, acc=0.3366, loss=3.2720, lr=9.95e-05]
Training:   7%|▋         | 731/11082 [2:21:11<33:18:07, 11.58s/step, acc=0.3366, loss=3.2720, lr=9.95e-05]
Training:   7%|▋         | 732/11082 [2:21:22<32:45:08, 11.39s/step, acc=0.3366, loss=3.2720, lr=9.95e-05]
Training:   7%|▋         | 733/11082 [2:21:34<33:12:55, 11.55s/step, acc=0.3366, loss=3.2720, lr=9.95e-05]
Training:   7%|▋         | 734/11082 [2:21:45<33:15:43, 11.57s/step, acc=0.3366, loss=3.2720, lr=9.95e-05]
Training:   7%|▋         | 735/11082 [2:21:57<33:13:18, 11.56s/step, acc=0.3366, loss=3.2720, lr=9.95e-05]
Training:   7%|▋         | 736/11082 [2:22:08<33:07:59, 11.53s/step, acc=0.3366, loss=3.2720, lr=9.95e-05]
Training:   7%|▋         | 737/11082 [2:22:20<33:15:10, 11.57s/step, acc=0.3366, loss=3.2720, lr=9.95e-05]
Training:   7%|▋         | 738/11082 [2:22:31<33:13:31, 11.56s/step, acc=0.3366, loss=3.2720, lr=9.95e-05]
Training:   7%|▋         | 739/11082 [2:22:43<33:07:05, 11.53s/step, acc=0.3366, loss=3.2720, lr=9.95e-05]
Training:   7%|▋         | 740/11082 [2:22:55<33:22:06, 11.62s/step, acc=0.3366, loss=3.2720, lr=9.95e-05]
Training:   7%|▋         | 741/11082 [2:23:07<33:30:18, 11.66s/step, acc=0.3366, loss=3.2720, lr=9.95e-05]
Training:   7%|▋         | 742/11082 [2:23:18<33:29:39, 11.66s/step, acc=0.3366, loss=3.2720, lr=9.95e-05]
Training:   7%|▋         | 743/11082 [2:23:30<33:22:18, 11.62s/step, acc=0.3366, loss=3.2720, lr=9.95e-05]
Training:   7%|▋         | 744/11082 [2:23:42<33:34:57, 11.69s/step, acc=0.3366, loss=3.2720, lr=9.95e-05]
Training:   7%|▋         | 745/11082 [2:23:53<33:16:08, 11.59s/step, acc=0.3366, loss=3.2720, lr=9.95e-05]
Training:   7%|▋         | 746/11082 [2:24:04<33:02:09, 11.51s/step, acc=0.3366, loss=3.2720, lr=9.95e-05]
Training:   7%|▋         | 747/11082 [2:24:16<33:20:47, 11.62s/step, acc=0.3366, loss=3.2720, lr=9.95e-05]
Training:   7%|▋         | 748/11082 [2:24:27<33:05:44, 11.53s/step, acc=0.3366, loss=3.2720, lr=9.95e-05]
Training:   7%|▋         | 749/11082 [2:24:39<32:50:48, 11.44s/step, acc=0.3366, loss=3.2720, lr=9.95e-05]
Training:   7%|▋         | 750/11082 [2:24:50<32:55:51, 11.47s/step, acc=0.3366, loss=3.2720, lr=9.95e-05]
Training:   7%|▋         | 750/11082 [2:24:57<32:55:51, 11.47s/step, acc=0.3416, loss=3.2140, lr=9.94e-05]2026-06-18 17:18:25 | INFO | __main__ | epoch=0 step=750 train_loss=3.213996 train_acc=0.341606 lr=9.937e-05 grad_norm=1.7812 acc_d0=0.6149 acc_d1=0.4434 acc_d2=0.3091 acc_d3=0.2209 acc_d4=0.1680 acc_d5=0.1365 acc_d6=0.1177 acc_d7=0.1011

Training:   7%|▋         | 751/11082 [2:25:02<33:10:23, 11.56s/step, acc=0.3416, loss=3.2140, lr=9.94e-05]
Training:   7%|▋         | 752/11082 [2:25:13<32:59:17, 11.50s/step, acc=0.3416, loss=3.2140, lr=9.94e-05]
Training:   7%|▋         | 753/11082 [2:25:25<32:58:24, 11.49s/step, acc=0.3416, loss=3.2140, lr=9.94e-05]
Training:   7%|▋         | 754/11082 [2:25:37<33:24:49, 11.65s/step, acc=0.3416, loss=3.2140, lr=9.94e-05]
Training:   7%|▋         | 755/11082 [2:25:48<33:02:48, 11.52s/step, acc=0.3416, loss=3.2140, lr=9.94e-05]
Training:   7%|▋         | 756/11082 [2:26:00<33:07:55, 11.55s/step, acc=0.3416, loss=3.2140, lr=9.94e-05]
Training:   7%|▋         | 757/11082 [2:26:11<33:11:24, 11.57s/step, acc=0.3416, loss=3.2140, lr=9.94e-05]
Training:   7%|▋         | 758/11082 [2:26:22<32:44:08, 11.42s/step, acc=0.3416, loss=3.2140, lr=9.94e-05]
Training:   7%|▋         | 759/11082 [2:26:34<33:14:52, 11.59s/step, acc=0.3416, loss=3.2140, lr=9.94e-05]
Training:   7%|▋         | 760/11082 [2:26:46<33:05:10, 11.54s/step, acc=0.3416, loss=3.2140, lr=9.94e-05]
Training:   7%|▋         | 761/11082 [2:26:57<32:53:29, 11.47s/step, acc=0.3416, loss=3.2140, lr=9.94e-05]
Training:   7%|▋         | 762/11082 [2:27:09<33:23:30, 11.65s/step, acc=0.3416, loss=3.2140, lr=9.94e-05]
Training:   7%|▋         | 763/11082 [2:27:21<33:20:20, 11.63s/step, acc=0.3416, loss=3.2140, lr=9.94e-05]
Training:   7%|▋         | 764/11082 [2:27:32<32:58:11, 11.50s/step, acc=0.3416, loss=3.2140, lr=9.94e-05]
Training:   7%|▋         | 765/11082 [2:27:43<32:40:26, 11.40s/step, acc=0.3416, loss=3.2140, lr=9.94e-05]
Training:   7%|▋         | 766/11082 [2:27:55<33:05:03, 11.55s/step, acc=0.3416, loss=3.2140, lr=9.94e-05]
Training:   7%|▋         | 767/11082 [2:28:06<32:41:36, 11.41s/step, acc=0.3416, loss=3.2140, lr=9.94e-05]
Training:   7%|▋         | 768/11082 [2:28:18<32:47:56, 11.45s/step, acc=0.3416, loss=3.2140, lr=9.94e-05]
Training:   7%|▋         | 769/11082 [2:28:29<33:07:41, 11.56s/step, acc=0.3416, loss=3.2140, lr=9.94e-05]
Training:   7%|▋         | 770/11082 [2:28:41<33:09:38, 11.58s/step, acc=0.3416, loss=3.2140, lr=9.94e-05]
Training:   7%|▋         | 771/11082 [2:28:53<33:05:16, 11.55s/step, acc=0.3416, loss=3.2140, lr=9.94e-05]
Training:   7%|▋         | 772/11082 [2:29:04<33:06:59, 11.56s/step, acc=0.3416, loss=3.2140, lr=9.94e-05]
Training:   7%|▋         | 773/11082 [2:29:16<32:57:48, 11.51s/step, acc=0.3416, loss=3.2140, lr=9.94e-05]
Training:   7%|▋         | 774/11082 [2:29:27<33:02:46, 11.54s/step, acc=0.3416, loss=3.2140, lr=9.94e-05]
Training:   7%|▋         | 775/11082 [2:29:38<32:42:02, 11.42s/step, acc=0.3416, loss=3.2140, lr=9.94e-05]
Training:   7%|▋         | 776/11082 [2:29:50<32:50:47, 11.47s/step, acc=0.3416, loss=3.2140, lr=9.94e-05]
Training:   7%|▋         | 777/11082 [2:30:01<32:51:38, 11.48s/step, acc=0.3416, loss=3.2140, lr=9.94e-05]
Training:   7%|▋         | 778/11082 [2:30:13<33:19:57, 11.65s/step, acc=0.3416, loss=3.2140, lr=9.94e-05]
Training:   7%|▋         | 779/11082 [2:30:25<33:05:58, 11.57s/step, acc=0.3416, loss=3.2140, lr=9.94e-05]
Training:   7%|▋         | 780/11082 [2:30:36<33:02:22, 11.55s/step, acc=0.3416, loss=3.2140, lr=9.94e-05]
Training:   7%|▋         | 781/11082 [2:30:48<33:14:40, 11.62s/step, acc=0.3416, loss=3.2140, lr=9.94e-05]
Training:   7%|▋         | 782/11082 [2:31:00<33:20:24, 11.65s/step, acc=0.3416, loss=3.2140, lr=9.94e-05]
Training:   7%|▋         | 783/11082 [2:31:12<33:30:57, 11.72s/step, acc=0.3416, loss=3.2140, lr=9.94e-05]
Training:   7%|▋         | 784/11082 [2:31:23<33:06:29, 11.57s/step, acc=0.3416, loss=3.2140, lr=9.94e-05]
Training:   7%|▋         | 785/11082 [2:31:35<33:15:01, 11.62s/step, acc=0.3416, loss=3.2140, lr=9.94e-05]
Training:   7%|▋         | 786/11082 [2:31:46<32:59:23, 11.53s/step, acc=0.3416, loss=3.2140, lr=9.94e-05]
Training:   7%|▋         | 787/11082 [2:31:58<33:11:24, 11.61s/step, acc=0.3416, loss=3.2140, lr=9.94e-05]
Training:   7%|▋         | 788/11082 [2:32:09<32:59:25, 11.54s/step, acc=0.3416, loss=3.2140, lr=9.94e-05]
Training:   7%|▋         | 789/11082 [2:32:21<33:29:01, 11.71s/step, acc=0.3416, loss=3.2140, lr=9.94e-05]
Training:   7%|▋         | 790/11082 [2:32:32<33:03:05, 11.56s/step, acc=0.3416, loss=3.2140, lr=9.94e-05]
Training:   7%|▋         | 791/11082 [2:32:44<32:46:19, 11.46s/step, acc=0.3416, loss=3.2140, lr=9.94e-05]
Training:   7%|▋         | 792/11082 [2:32:56<33:05:30, 11.58s/step, acc=0.3416, loss=3.2140, lr=9.94e-05]
Training:   7%|▋         | 793/11082 [2:33:07<32:58:58, 11.54s/step, acc=0.3416, loss=3.2140, lr=9.94e-05]
Training:   7%|▋         | 794/11082 [2:33:19<33:07:16, 11.59s/step, acc=0.3416, loss=3.2140, lr=9.94e-05]
Training:   7%|▋         | 795/11082 [2:33:30<32:57:58, 11.54s/step, acc=0.3416, loss=3.2140, lr=9.94e-05]
Training:   7%|▋         | 796/11082 [2:33:42<32:56:45, 11.53s/step, acc=0.3416, loss=3.2140, lr=9.94e-05]
Training:   7%|▋         | 797/11082 [2:33:53<33:10:50, 11.61s/step, acc=0.3416, loss=3.2140, lr=9.94e-05]
Training:   7%|▋         | 798/11082 [2:34:05<32:53:07, 11.51s/step, acc=0.3416, loss=3.2140, lr=9.94e-05]
Training:   7%|▋         | 799/11082 [2:34:17<33:13:28, 11.63s/step, acc=0.3416, loss=3.2140, lr=9.94e-05]
Training:   7%|▋         | 800/11082 [2:34:28<32:59:14, 11.55s/step, acc=0.3416, loss=3.2140, lr=9.94e-05]
Training:   7%|▋         | 800/11082 [2:34:35<32:59:14, 11.55s/step, acc=0.3511, loss=3.1374, lr=9.93e-05]2026-06-18 17:28:03 | INFO | __main__ | epoch=0 step=800 train_loss=3.137382 train_acc=0.351069 lr=9.925e-05 grad_norm=1.9609 acc_d0=0.6017 acc_d1=0.4336 acc_d2=0.3034 acc_d3=0.2162 acc_d4=0.1659 acc_d5=0.1337 acc_d6=0.1143 acc_d7=0.1023

Training:   7%|▋         | 801/11082 [2:34:40<33:04:36, 11.58s/step, acc=0.3511, loss=3.1374, lr=9.93e-05]
Training:   7%|▋         | 802/11082 [2:34:51<32:58:19, 11.55s/step, acc=0.3511, loss=3.1374, lr=9.93e-05]
Training:   7%|▋         | 803/11082 [2:35:02<32:49:49, 11.50s/step, acc=0.3511, loss=3.1374, lr=9.93e-05]
Training:   7%|▋         | 804/11082 [2:35:14<33:08:14, 11.61s/step, acc=0.3511, loss=3.1374, lr=9.93e-05]
Training:   7%|▋         | 805/11082 [2:35:26<32:57:50, 11.55s/step, acc=0.3511, loss=3.1374, lr=9.93e-05]
Training:   7%|▋         | 806/11082 [2:35:37<32:57:29, 11.55s/step, acc=0.3511, loss=3.1374, lr=9.93e-05]
Training:   7%|▋         | 807/11082 [2:35:49<32:53:30, 11.52s/step, acc=0.3511, loss=3.1374, lr=9.93e-05]
Training:   7%|▋         | 808/11082 [2:36:00<33:02:03, 11.58s/step, acc=0.3511, loss=3.1374, lr=9.93e-05]
Training:   7%|▋         | 809/11082 [2:36:12<32:52:11, 11.52s/step, acc=0.3511, loss=3.1374, lr=9.93e-05]
Training:   7%|▋         | 810/11082 [2:36:23<32:36:42, 11.43s/step, acc=0.3511, loss=3.1374, lr=9.93e-05]
Training:   7%|▋         | 811/11082 [2:36:35<33:20:20, 11.69s/step, acc=0.3511, loss=3.1374, lr=9.93e-05]
Training:   7%|▋         | 812/11082 [2:36:46<32:40:45, 11.46s/step, acc=0.3511, loss=3.1374, lr=9.93e-05]
Training:   7%|▋         | 813/11082 [2:36:58<33:14:42, 11.65s/step, acc=0.3511, loss=3.1374, lr=9.93e-05]
Training:   7%|▋         | 814/11082 [2:37:10<33:02:23, 11.58s/step, acc=0.3511, loss=3.1374, lr=9.93e-05]
Training:   7%|▋         | 815/11082 [2:37:21<33:07:45, 11.62s/step, acc=0.3511, loss=3.1374, lr=9.93e-05]
Training:   7%|▋         | 816/11082 [2:37:33<33:23:32, 11.71s/step, acc=0.3511, loss=3.1374, lr=9.93e-05]
Training:   7%|▋         | 817/11082 [2:37:45<33:26:21, 11.73s/step, acc=0.3511, loss=3.1374, lr=9.93e-05]
Training:   7%|▋         | 818/11082 [2:37:57<33:12:19, 11.65s/step, acc=0.3511, loss=3.1374, lr=9.93e-05]
Training:   7%|▋         | 819/11082 [2:38:08<33:09:27, 11.63s/step, acc=0.3511, loss=3.1374, lr=9.93e-05]
Training:   7%|▋         | 820/11082 [2:38:20<33:03:53, 11.60s/step, acc=0.3511, loss=3.1374, lr=9.93e-05]
Training:   7%|▋         | 821/11082 [2:38:32<33:14:21, 11.66s/step, acc=0.3511, loss=3.1374, lr=9.93e-05]
Training:   7%|▋         | 822/11082 [2:38:43<32:55:40, 11.55s/step, acc=0.3511, loss=3.1374, lr=9.93e-05]
Training:   7%|▋         | 823/11082 [2:38:54<32:51:17, 11.53s/step, acc=0.3511, loss=3.1374, lr=9.93e-05]
Training:   7%|▋         | 824/11082 [2:39:06<32:55:14, 11.55s/step, acc=0.3511, loss=3.1374, lr=9.93e-05]
Training:   7%|▋         | 825/11082 [2:39:18<33:01:28, 11.59s/step, acc=0.3511, loss=3.1374, lr=9.93e-05]
Training:   7%|▋         | 826/11082 [2:39:29<32:51:31, 11.53s/step, acc=0.3511, loss=3.1374, lr=9.93e-05]
Training:   7%|▋         | 827/11082 [2:39:41<33:00:29, 11.59s/step, acc=0.3511, loss=3.1374, lr=9.93e-05]
Training:   7%|▋         | 828/11082 [2:39:53<33:13:02, 11.66s/step, acc=0.3511, loss=3.1374, lr=9.93e-05]
Training:   7%|▋         | 829/11082 [2:40:04<33:03:32, 11.61s/step, acc=0.3511, loss=3.1374, lr=9.93e-05]
Training:   7%|▋         | 830/11082 [2:40:16<33:12:41, 11.66s/step, acc=0.3511, loss=3.1374, lr=9.93e-05]
Training:   7%|▋         | 831/11082 [2:40:27<33:03:10, 11.61s/step, acc=0.3511, loss=3.1374, lr=9.93e-05]
Training:   8%|▊         | 832/11082 [2:40:38<32:37:24, 11.46s/step, acc=0.3511, loss=3.1374, lr=9.93e-05]
Training:   8%|▊         | 833/11082 [2:40:50<32:31:07, 11.42s/step, acc=0.3511, loss=3.1374, lr=9.93e-05]
Training:   8%|▊         | 834/11082 [2:41:01<32:45:35, 11.51s/step, acc=0.3511, loss=3.1374, lr=9.93e-05]
Training:   8%|▊         | 835/11082 [2:41:13<33:00:00, 11.59s/step, acc=0.3511, loss=3.1374, lr=9.93e-05]
Training:   8%|▊         | 836/11082 [2:41:25<32:52:17, 11.55s/step, acc=0.3511, loss=3.1374, lr=9.93e-05]
Training:   8%|▊         | 837/11082 [2:41:36<32:50:06, 11.54s/step, acc=0.3511, loss=3.1374, lr=9.93e-05]
Training:   8%|▊         | 838/11082 [2:41:48<33:02:49, 11.61s/step, acc=0.3511, loss=3.1374, lr=9.93e-05]
Training:   8%|▊         | 839/11082 [2:41:59<32:46:05, 11.52s/step, acc=0.3511, loss=3.1374, lr=9.93e-05]
Training:   8%|▊         | 840/11082 [2:42:11<32:36:35, 11.46s/step, acc=0.3511, loss=3.1374, lr=9.93e-05]
Training:   8%|▊         | 841/11082 [2:42:22<32:45:37, 11.52s/step, acc=0.3511, loss=3.1374, lr=9.93e-05]
Training:   8%|▊         | 842/11082 [2:42:33<32:23:20, 11.39s/step, acc=0.3511, loss=3.1374, lr=9.93e-05]
Training:   8%|▊         | 843/11082 [2:42:45<32:48:57, 11.54s/step, acc=0.3511, loss=3.1374, lr=9.93e-05]
Training:   8%|▊         | 844/11082 [2:42:57<32:44:45, 11.51s/step, acc=0.3511, loss=3.1374, lr=9.93e-05]
Training:   8%|▊         | 845/11082 [2:43:08<32:36:58, 11.47s/step, acc=0.3511, loss=3.1374, lr=9.93e-05]
Training:   8%|▊         | 846/11082 [2:43:20<32:34:31, 11.46s/step, acc=0.3511, loss=3.1374, lr=9.93e-05]
Training:   8%|▊         | 847/11082 [2:43:31<32:40:26, 11.49s/step, acc=0.3511, loss=3.1374, lr=9.93e-05]
Training:   8%|▊         | 848/11082 [2:43:43<32:41:13, 11.50s/step, acc=0.3511, loss=3.1374, lr=9.93e-05]
Training:   8%|▊         | 849/11082 [2:43:54<32:20:59, 11.38s/step, acc=0.3511, loss=3.1374, lr=9.93e-05]
Training:   8%|▊         | 850/11082 [2:44:06<33:03:52, 11.63s/step, acc=0.3511, loss=3.1374, lr=9.93e-05]
Training:   8%|▊         | 850/11082 [2:44:13<33:03:52, 11.63s/step, acc=0.3565, loss=3.0899, lr=9.91e-05]2026-06-18 17:37:41 | INFO | __main__ | epoch=0 step=850 train_loss=3.089944 train_acc=0.356536 lr=9.913e-05 grad_norm=1.8984 acc_d0=0.6185 acc_d1=0.4507 acc_d2=0.3204 acc_d3=0.2306 acc_d4=0.1764 acc_d5=0.1413 acc_d6=0.1196 acc_d7=0.1041

Training:   8%|▊         | 851/11082 [2:44:17<32:57:10, 11.60s/step, acc=0.3565, loss=3.0899, lr=9.91e-05]
Training:   8%|▊         | 852/11082 [2:44:29<32:48:11, 11.54s/step, acc=0.3565, loss=3.0899, lr=9.91e-05]
Training:   8%|▊         | 853/11082 [2:44:40<32:39:27, 11.49s/step, acc=0.3565, loss=3.0899, lr=9.91e-05]
Training:   8%|▊         | 854/11082 [2:44:52<32:45:03, 11.53s/step, acc=0.3565, loss=3.0899, lr=9.91e-05]
Training:   8%|▊         | 855/11082 [2:45:03<32:37:39, 11.49s/step, acc=0.3565, loss=3.0899, lr=9.91e-05]
Training:   8%|▊         | 856/11082 [2:45:15<33:04:24, 11.64s/step, acc=0.3565, loss=3.0899, lr=9.91e-05]
Training:   8%|▊         | 857/11082 [2:45:26<32:33:34, 11.46s/step, acc=0.3565, loss=3.0899, lr=9.91e-05]
Training:   8%|▊         | 858/11082 [2:45:38<32:56:19, 11.60s/step, acc=0.3565, loss=3.0899, lr=9.91e-05]
Training:   8%|▊         | 859/11082 [2:45:50<32:50:27, 11.56s/step, acc=0.3565, loss=3.0899, lr=9.91e-05]
Training:   8%|▊         | 860/11082 [2:46:02<33:05:56, 11.66s/step, acc=0.3565, loss=3.0899, lr=9.91e-05]
Training:   8%|▊         | 861/11082 [2:46:13<33:07:34, 11.67s/step, acc=0.3565, loss=3.0899, lr=9.91e-05]
Training:   8%|▊         | 862/11082 [2:46:25<32:48:18, 11.56s/step, acc=0.3565, loss=3.0899, lr=9.91e-05]
Training:   8%|▊         | 863/11082 [2:46:37<33:28:34, 11.79s/step, acc=0.3565, loss=3.0899, lr=9.91e-05]
Training:   8%|▊         | 864/11082 [2:46:48<33:18:23, 11.73s/step, acc=0.3565, loss=3.0899, lr=9.91e-05]
Training:   8%|▊         | 865/11082 [2:47:00<33:28:04, 11.79s/step, acc=0.3565, loss=3.0899, lr=9.91e-05]
Training:   8%|▊         | 866/11082 [2:47:12<33:08:09, 11.68s/step, acc=0.3565, loss=3.0899, lr=9.91e-05]
Training:   8%|▊         | 867/11082 [2:47:24<33:08:22, 11.68s/step, acc=0.3565, loss=3.0899, lr=9.91e-05]
Training:   8%|▊         | 868/11082 [2:47:35<33:12:07, 11.70s/step, acc=0.3565, loss=3.0899, lr=9.91e-05]
Training:   8%|▊         | 869/11082 [2:47:47<32:55:50, 11.61s/step, acc=0.3565, loss=3.0899, lr=9.91e-05]
Training:   8%|▊         | 870/11082 [2:47:58<32:41:37, 11.53s/step, acc=0.3565, loss=3.0899, lr=9.91e-05]
Training:   8%|▊         | 871/11082 [2:48:10<32:50:52, 11.58s/step, acc=0.3565, loss=3.0899, lr=9.91e-05]
Training:   8%|▊         | 872/11082 [2:48:21<32:34:36, 11.49s/step, acc=0.3565, loss=3.0899, lr=9.91e-05]
Training:   8%|▊         | 873/11082 [2:48:33<32:38:56, 11.51s/step, acc=0.3565, loss=3.0899, lr=9.91e-05]
Training:   8%|▊         | 874/11082 [2:48:44<32:36:34, 11.50s/step, acc=0.3565, loss=3.0899, lr=9.91e-05]
Training:   8%|▊         | 875/11082 [2:48:56<32:43:45, 11.54s/step, acc=0.3565, loss=3.0899, lr=9.91e-05]
Training:   8%|▊         | 876/11082 [2:49:07<32:51:14, 11.59s/step, acc=0.3565, loss=3.0899, lr=9.91e-05]
Training:   8%|▊         | 877/11082 [2:49:19<32:45:08, 11.55s/step, acc=0.3565, loss=3.0899, lr=9.91e-05]
Training:   8%|▊         | 878/11082 [2:49:30<32:38:40, 11.52s/step, acc=0.3565, loss=3.0899, lr=9.91e-05]
Training:   8%|▊         | 879/11082 [2:49:42<32:40:41, 11.53s/step, acc=0.3565, loss=3.0899, lr=9.91e-05]
Training:   8%|▊         | 880/11082 [2:49:54<32:52:46, 11.60s/step, acc=0.3565, loss=3.0899, lr=9.91e-05]
Training:   8%|▊         | 881/11082 [2:50:05<32:49:16, 11.58s/step, acc=0.3565, loss=3.0899, lr=9.91e-05]
Training:   8%|▊         | 882/11082 [2:50:16<32:30:21, 11.47s/step, acc=0.3565, loss=3.0899, lr=9.91e-05]
Training:   8%|▊         | 883/11082 [2:50:28<32:27:58, 11.46s/step, acc=0.3565, loss=3.0899, lr=9.91e-05]
Training:   8%|▊         | 884/11082 [2:50:40<33:07:39, 11.69s/step, acc=0.3565, loss=3.0899, lr=9.91e-05]
Training:   8%|▊         | 885/11082 [2:50:51<32:52:41, 11.61s/step, acc=0.3565, loss=3.0899, lr=9.91e-05]
Training:   8%|▊         | 886/11082 [2:51:03<32:59:01, 11.65s/step, acc=0.3565, loss=3.0899, lr=9.91e-05]
Training:   8%|▊         | 887/11082 [2:51:14<32:34:05, 11.50s/step, acc=0.3565, loss=3.0899, lr=9.91e-05]
Training:   8%|▊         | 888/11082 [2:51:26<32:44:00, 11.56s/step, acc=0.3565, loss=3.0899, lr=9.91e-05]
Training:   8%|▊         | 889/11082 [2:51:38<33:13:35, 11.74s/step, acc=0.3565, loss=3.0899, lr=9.91e-05]
Training:   8%|▊         | 890/11082 [2:51:49<32:40:20, 11.54s/step, acc=0.3565, loss=3.0899, lr=9.91e-05]
Training:   8%|▊         | 891/11082 [2:52:01<32:43:30, 11.56s/step, acc=0.3565, loss=3.0899, lr=9.91e-05]
Training:   8%|▊         | 892/11082 [2:52:13<32:56:30, 11.64s/step, acc=0.3565, loss=3.0899, lr=9.91e-05]
Training:   8%|▊         | 893/11082 [2:52:24<32:45:09, 11.57s/step, acc=0.3565, loss=3.0899, lr=9.91e-05]
Training:   8%|▊         | 894/11082 [2:52:36<32:39:52, 11.54s/step, acc=0.3565, loss=3.0899, lr=9.91e-05]
Training:   8%|▊         | 895/11082 [2:52:47<32:50:23, 11.61s/step, acc=0.3565, loss=3.0899, lr=9.91e-05]
Training:   8%|▊         | 896/11082 [2:52:59<33:00:21, 11.67s/step, acc=0.3565, loss=3.0899, lr=9.91e-05]
Training:   8%|▊         | 897/11082 [2:53:10<32:28:13, 11.48s/step, acc=0.3565, loss=3.0899, lr=9.91e-05]
Training:   8%|▊         | 898/11082 [2:53:22<32:41:54, 11.56s/step, acc=0.3565, loss=3.0899, lr=9.91e-05]
Training:   8%|▊         | 899/11082 [2:53:33<32:40:34, 11.55s/step, acc=0.3565, loss=3.0899, lr=9.91e-05]
Training:   8%|▊         | 900/11082 [2:53:45<32:41:30, 11.56s/step, acc=0.3565, loss=3.0899, lr=9.91e-05]
Training:   8%|▊         | 900/11082 [2:53:52<32:41:30, 11.56s/step, acc=0.3602, loss=3.0572, lr=9.90e-05]2026-06-18 17:47:20 | INFO | __main__ | epoch=0 step=900 train_loss=3.057157 train_acc=0.360190 lr=9.900e-05 grad_norm=1.8438 acc_d0=0.6345 acc_d1=0.4650 acc_d2=0.3322 acc_d3=0.2373 acc_d4=0.1812 acc_d5=0.1447 acc_d6=0.1234 acc_d7=0.1054

Training:   8%|▊         | 901/11082 [2:53:57<32:45:03, 11.58s/step, acc=0.3602, loss=3.0572, lr=9.90e-05]
Training:   8%|▊         | 902/11082 [2:54:08<32:38:32, 11.54s/step, acc=0.3602, loss=3.0572, lr=9.90e-05]
Training:   8%|▊         | 903/11082 [2:54:20<32:37:56, 11.54s/step, acc=0.3602, loss=3.0572, lr=9.90e-05]
Training:   8%|▊         | 904/11082 [2:54:31<32:36:07, 11.53s/step, acc=0.3602, loss=3.0572, lr=9.90e-05]
Training:   8%|▊         | 905/11082 [2:54:43<32:27:28, 11.48s/step, acc=0.3602, loss=3.0572, lr=9.90e-05]
Training:   8%|▊         | 906/11082 [2:54:54<32:42:16, 11.57s/step, acc=0.3602, loss=3.0572, lr=9.90e-05]
Training:   8%|▊         | 907/11082 [2:55:06<32:43:39, 11.58s/step, acc=0.3602, loss=3.0572, lr=9.90e-05]
Training:   8%|▊         | 908/11082 [2:55:17<32:43:48, 11.58s/step, acc=0.3602, loss=3.0572, lr=9.90e-05]
Training:   8%|▊         | 909/11082 [2:55:29<32:43:30, 11.58s/step, acc=0.3602, loss=3.0572, lr=9.90e-05]
Training:   8%|▊         | 910/11082 [2:55:40<32:28:52, 11.50s/step, acc=0.3602, loss=3.0572, lr=9.90e-05]
Training:   8%|▊         | 911/11082 [2:55:52<32:45:11, 11.59s/step, acc=0.3602, loss=3.0572, lr=9.90e-05]
Training:   8%|▊         | 912/11082 [2:56:03<32:25:47, 11.48s/step, acc=0.3602, loss=3.0572, lr=9.90e-05]
Training:   8%|▊         | 913/11082 [2:56:15<32:32:09, 11.52s/step, acc=0.3602, loss=3.0572, lr=9.90e-05]
Training:   8%|▊         | 914/11082 [2:56:27<32:55:50, 11.66s/step, acc=0.3602, loss=3.0572, lr=9.90e-05]
Training:   8%|▊         | 915/11082 [2:56:38<32:36:21, 11.55s/step, acc=0.3602, loss=3.0572, lr=9.90e-05]
Training:   8%|▊         | 916/11082 [2:56:50<32:23:17, 11.47s/step, acc=0.3602, loss=3.0572, lr=9.90e-05]
Training:   8%|▊         | 917/11082 [2:57:01<32:36:20, 11.55s/step, acc=0.3602, loss=3.0572, lr=9.90e-05]
Training:   8%|▊         | 918/11082 [2:57:13<32:40:00, 11.57s/step, acc=0.3602, loss=3.0572, lr=9.90e-05]
Training:   8%|▊         | 919/11082 [2:57:25<32:54:20, 11.66s/step, acc=0.3602, loss=3.0572, lr=9.90e-05]
Training:   8%|▊         | 920/11082 [2:57:36<32:43:11, 11.59s/step, acc=0.3602, loss=3.0572, lr=9.90e-05]
Training:   8%|▊         | 921/11082 [2:57:47<32:26:09, 11.49s/step, acc=0.3602, loss=3.0572, lr=9.90e-05]
Training:   8%|▊         | 922/11082 [2:57:59<32:36:52, 11.56s/step, acc=0.3602, loss=3.0572, lr=9.90e-05]
Training:   8%|▊         | 923/11082 [2:58:11<32:33:40, 11.54s/step, acc=0.3602, loss=3.0572, lr=9.90e-05]
Training:   8%|▊         | 924/11082 [2:58:22<32:22:55, 11.48s/step, acc=0.3602, loss=3.0572, lr=9.90e-05]
Training:   8%|▊         | 925/11082 [2:58:33<32:12:57, 11.42s/step, acc=0.3602, loss=3.0572, lr=9.90e-05]
Training:   8%|▊         | 926/11082 [2:58:45<32:33:05, 11.54s/step, acc=0.3602, loss=3.0572, lr=9.90e-05]
Training:   8%|▊         | 927/11082 [2:58:57<32:30:00, 11.52s/step, acc=0.3602, loss=3.0572, lr=9.90e-05]
Training:   8%|▊         | 928/11082 [2:59:08<32:22:10, 11.48s/step, acc=0.3602, loss=3.0572, lr=9.90e-05]
Training:   8%|▊         | 929/11082 [2:59:20<32:26:28, 11.50s/step, acc=0.3602, loss=3.0572, lr=9.90e-05]
Training:   8%|▊         | 930/11082 [2:59:31<32:23:59, 11.49s/step, acc=0.3602, loss=3.0572, lr=9.90e-05]
Training:   8%|▊         | 931/11082 [2:59:42<32:07:57, 11.40s/step, acc=0.3602, loss=3.0572, lr=9.90e-05]
Training:   8%|▊         | 932/11082 [2:59:54<32:40:59, 11.59s/step, acc=0.3602, loss=3.0572, lr=9.90e-05]
Training:   8%|▊         | 933/11082 [3:00:05<31:56:09, 11.33s/step, acc=0.3602, loss=3.0572, lr=9.90e-05]
Training:   8%|▊         | 934/11082 [3:00:17<32:37:37, 11.57s/step, acc=0.3602, loss=3.0572, lr=9.90e-05]
Training:   8%|▊         | 935/11082 [3:00:28<32:21:07, 11.48s/step, acc=0.3602, loss=3.0572, lr=9.90e-05]
Training:   8%|▊         | 936/11082 [3:00:40<32:12:11, 11.43s/step, acc=0.3602, loss=3.0572, lr=9.90e-05]
Training:   8%|▊         | 937/11082 [3:00:52<32:46:35, 11.63s/step, acc=0.3602, loss=3.0572, lr=9.90e-05]
Training:   8%|▊         | 938/11082 [3:01:03<32:16:37, 11.45s/step, acc=0.3602, loss=3.0572, lr=9.90e-05]
Training:   8%|▊         | 939/11082 [3:01:14<32:24:10, 11.50s/step, acc=0.3602, loss=3.0572, lr=9.90e-05]
Training:   8%|▊         | 940/11082 [3:01:26<32:43:40, 11.62s/step, acc=0.3602, loss=3.0572, lr=9.90e-05]
Training:   8%|▊         | 941/11082 [3:01:38<32:44:46, 11.62s/step, acc=0.3602, loss=3.0572, lr=9.90e-05]
Training:   9%|▊         | 942/11082 [3:01:50<32:46:11, 11.63s/step, acc=0.3602, loss=3.0572, lr=9.90e-05]
Training:   9%|▊         | 943/11082 [3:02:01<32:45:53, 11.63s/step, acc=0.3602, loss=3.0572, lr=9.90e-05]
Training:   9%|▊         | 944/11082 [3:02:13<33:11:22, 11.79s/step, acc=0.3602, loss=3.0572, lr=9.90e-05]
Training:   9%|▊         | 945/11082 [3:02:25<33:11:29, 11.79s/step, acc=0.3602, loss=3.0572, lr=9.90e-05]
Training:   9%|▊         | 946/11082 [3:02:36<32:46:10, 11.64s/step, acc=0.3602, loss=3.0572, lr=9.90e-05]
Training:   9%|▊         | 947/11082 [3:02:47<32:16:42, 11.47s/step, acc=0.3602, loss=3.0572, lr=9.90e-05]
Training:   9%|▊         | 948/11082 [3:02:59<32:31:54, 11.56s/step, acc=0.3602, loss=3.0572, lr=9.90e-05]
Training:   9%|▊         | 949/11082 [3:03:10<32:12:45, 11.44s/step, acc=0.3602, loss=3.0572, lr=9.90e-05]
Training:   9%|▊         | 950/11082 [3:03:23<32:56:27, 11.70s/step, acc=0.3602, loss=3.0572, lr=9.90e-05]
Training:   9%|▊         | 950/11082 [3:03:30<32:56:27, 11.70s/step, acc=0.3643, loss=3.0199, lr=9.89e-05]2026-06-18 17:56:57 | INFO | __main__ | epoch=0 step=950 train_loss=3.019870 train_acc=0.364333 lr=9.886e-05 grad_norm=1.6172 acc_d0=0.6367 acc_d1=0.4725 acc_d2=0.3399 acc_d3=0.2463 acc_d4=0.1873 acc_d5=0.1505 acc_d6=0.1271 acc_d7=0.1108

Training:   9%|▊         | 951/11082 [3:03:34<32:47:01, 11.65s/step, acc=0.3643, loss=3.0199, lr=9.89e-05]
Training:   9%|▊         | 952/11082 [3:03:46<32:29:54, 11.55s/step, acc=0.3643, loss=3.0199, lr=9.89e-05]
Training:   9%|▊         | 953/11082 [3:03:57<32:32:47, 11.57s/step, acc=0.3643, loss=3.0199, lr=9.89e-05]
Training:   9%|▊         | 954/11082 [3:04:09<32:34:18, 11.58s/step, acc=0.3643, loss=3.0199, lr=9.89e-05]
Training:   9%|▊         | 955/11082 [3:04:20<32:17:11, 11.48s/step, acc=0.3643, loss=3.0199, lr=9.89e-05]
Training:   9%|▊         | 956/11082 [3:04:32<32:38:22, 11.60s/step, acc=0.3643, loss=3.0199, lr=9.89e-05]
Training:   9%|▊         | 957/11082 [3:04:43<32:06:37, 11.42s/step, acc=0.3643, loss=3.0199, lr=9.89e-05]
Training:   9%|▊         | 958/11082 [3:04:55<32:26:16, 11.53s/step, acc=0.3643, loss=3.0199, lr=9.89e-05]
Training:   9%|▊         | 959/11082 [3:05:06<32:34:15, 11.58s/step, acc=0.3643, loss=3.0199, lr=9.89e-05]
Training:   9%|▊         | 960/11082 [3:05:18<32:27:35, 11.54s/step, acc=0.3643, loss=3.0199, lr=9.89e-05]
Training:   9%|▊         | 961/11082 [3:05:30<32:54:13, 11.70s/step, acc=0.3643, loss=3.0199, lr=9.89e-05]
Training:   9%|▊         | 962/11082 [3:05:41<32:27:10, 11.54s/step, acc=0.3643, loss=3.0199, lr=9.89e-05]
Training:   9%|▊         | 963/11082 [3:05:53<32:45:01, 11.65s/step, acc=0.3643, loss=3.0199, lr=9.89e-05]
Training:   9%|▊         | 964/11082 [3:06:05<32:38:47, 11.62s/step, acc=0.3643, loss=3.0199, lr=9.89e-05]
Training:   9%|▊         | 965/11082 [3:06:16<32:22:45, 11.52s/step, acc=0.3643, loss=3.0199, lr=9.89e-05]
Training:   9%|▊         | 966/11082 [3:06:27<32:02:49, 11.40s/step, acc=0.3643, loss=3.0199, lr=9.89e-05]
Training:   9%|▊         | 967/11082 [3:06:39<32:51:06, 11.69s/step, acc=0.3643, loss=3.0199, lr=9.89e-05]
Training:   9%|▊         | 968/11082 [3:06:51<32:33:50, 11.59s/step, acc=0.3643, loss=3.0199, lr=9.89e-05]
Training:   9%|▊         | 969/11082 [3:07:02<32:14:59, 11.48s/step, acc=0.3643, loss=3.0199, lr=9.89e-05]
Training:   9%|▉         | 970/11082 [3:07:13<32:17:07, 11.49s/step, acc=0.3643, loss=3.0199, lr=9.89e-05]
Training:   9%|▉         | 971/11082 [3:07:25<32:16:16, 11.49s/step, acc=0.3643, loss=3.0199, lr=9.89e-05]
Training:   9%|▉         | 972/11082 [3:07:37<32:28:45, 11.57s/step, acc=0.3643, loss=3.0199, lr=9.89e-05]
Training:   9%|▉         | 973/11082 [3:07:49<32:41:11, 11.64s/step, acc=0.3643, loss=3.0199, lr=9.89e-05]
Training:   9%|▉         | 974/11082 [3:08:00<32:29:51, 11.57s/step, acc=0.3643, loss=3.0199, lr=9.89e-05]
Training:   9%|▉         | 975/11082 [3:08:11<32:27:23, 11.56s/step, acc=0.3643, loss=3.0199, lr=9.89e-05]
Training:   9%|▉         | 976/11082 [3:08:23<32:02:12, 11.41s/step, acc=0.3643, loss=3.0199, lr=9.89e-05]
Training:   9%|▉         | 977/11082 [3:08:34<32:18:59, 11.51s/step, acc=0.3643, loss=3.0199, lr=9.89e-05]
Training:   9%|▉         | 978/11082 [3:08:46<32:33:17, 11.60s/step, acc=0.3643, loss=3.0199, lr=9.89e-05]
Training:   9%|▉         | 979/11082 [3:08:57<32:24:30, 11.55s/step, acc=0.3643, loss=3.0199, lr=9.89e-05]
Training:   9%|▉         | 980/11082 [3:09:09<32:37:46, 11.63s/step, acc=0.3643, loss=3.0199, lr=9.89e-05]
Training:   9%|▉         | 981/11082 [3:09:21<32:32:11, 11.60s/step, acc=0.3643, loss=3.0199, lr=9.89e-05]
Training:   9%|▉         | 982/11082 [3:09:32<32:16:38, 11.50s/step, acc=0.3643, loss=3.0199, lr=9.89e-05]
Training:   9%|▉         | 983/11082 [3:09:44<32:17:15, 11.51s/step, acc=0.3643, loss=3.0199, lr=9.89e-05]
Training:   9%|▉         | 984/11082 [3:09:55<32:30:22, 11.59s/step, acc=0.3643, loss=3.0199, lr=9.89e-05]
Training:   9%|▉         | 985/11082 [3:10:07<32:22:50, 11.55s/step, acc=0.3643, loss=3.0199, lr=9.89e-05]
Training:   9%|▉         | 986/11082 [3:10:18<32:02:56, 11.43s/step, acc=0.3643, loss=3.0199, lr=9.89e-05]
Training:   9%|▉         | 987/11082 [3:10:30<32:25:10, 11.56s/step, acc=0.3643, loss=3.0199, lr=9.89e-05]
Training:   9%|▉         | 988/11082 [3:10:41<32:16:23, 11.51s/step, acc=0.3643, loss=3.0199, lr=9.89e-05]
Training:   9%|▉         | 989/11082 [3:10:53<32:40:20, 11.65s/step, acc=0.3643, loss=3.0199, lr=9.89e-05]
Training:   9%|▉         | 990/11082 [3:11:04<32:10:12, 11.48s/step, acc=0.3643, loss=3.0199, lr=9.89e-05]
Training:   9%|▉         | 991/11082 [3:11:16<32:23:05, 11.55s/step, acc=0.3643, loss=3.0199, lr=9.89e-05]
Training:   9%|▉         | 992/11082 [3:11:28<32:33:16, 11.62s/step, acc=0.3643, loss=3.0199, lr=9.89e-05]
Training:   9%|▉         | 993/11082 [3:11:39<32:01:51, 11.43s/step, acc=0.3643, loss=3.0199, lr=9.89e-05]
Training:   9%|▉         | 994/11082 [3:11:51<32:32:52, 11.62s/step, acc=0.3643, loss=3.0199, lr=9.89e-05]
Training:   9%|▉         | 995/11082 [3:12:02<32:06:33, 11.46s/step, acc=0.3643, loss=3.0199, lr=9.89e-05]
Training:   9%|▉         | 996/11082 [3:12:13<32:03:00, 11.44s/step, acc=0.3643, loss=3.0199, lr=9.89e-05]
Training:   9%|▉         | 997/11082 [3:12:25<32:23:23, 11.56s/step, acc=0.3643, loss=3.0199, lr=9.89e-05]
Training:   9%|▉         | 998/11082 [3:12:37<32:38:40, 11.65s/step, acc=0.3643, loss=3.0199, lr=9.89e-05]
Training:   9%|▉         | 999/11082 [3:12:48<32:21:52, 11.56s/step, acc=0.3643, loss=3.0199, lr=9.89e-05]
Training:   9%|▉         | 1000/11082 [3:13:00<32:12:22, 11.50s/step, acc=0.3643, loss=3.0199, lr=9.89e-05]
Training:   9%|▉         | 1000/11082 [3:13:07<32:12:22, 11.50s/step, acc=0.3721, loss=2.9612, lr=9.87e-05]2026-06-18 18:06:35 | INFO | __main__ | epoch=0 step=1000 train_loss=2.961208 train_acc=0.372122 lr=9.871e-05 grad_norm=2.2500 acc_d0=0.6415 acc_d1=0.4733 acc_d2=0.3395 acc_d3=0.2445 acc_d4=0.1851 acc_d5=0.1474 acc_d6=0.1237 acc_d7=0.1086

Training:   9%|▉         | 1001/11082 [3:13:12<32:27:54, 11.59s/step, acc=0.3721, loss=2.9612, lr=9.87e-05]
Training:   9%|▉         | 1002/11082 [3:13:23<32:20:06, 11.55s/step, acc=0.3721, loss=2.9612, lr=9.87e-05]
Training:   9%|▉         | 1003/11082 [3:13:35<32:23:47, 11.57s/step, acc=0.3721, loss=2.9612, lr=9.87e-05]
Training:   9%|▉         | 1004/11082 [3:13:46<31:59:23, 11.43s/step, acc=0.3721, loss=2.9612, lr=9.87e-05]
Training:   9%|▉         | 1005/11082 [3:13:57<32:07:41, 11.48s/step, acc=0.3721, loss=2.9612, lr=9.87e-05]
Training:   9%|▉         | 1006/11082 [3:14:09<32:34:01, 11.64s/step, acc=0.3721, loss=2.9612, lr=9.87e-05]
Training:   9%|▉         | 1007/11082 [3:14:21<32:11:10, 11.50s/step, acc=0.3721, loss=2.9612, lr=9.87e-05]
Training:   9%|▉         | 1008/11082 [3:14:32<32:08:54, 11.49s/step, acc=0.3721, loss=2.9612, lr=9.87e-05]
Training:   9%|▉         | 1009/11082 [3:14:44<32:27:20, 11.60s/step, acc=0.3721, loss=2.9612, lr=9.87e-05]
Training:   9%|▉         | 1010/11082 [3:14:55<32:16:07, 11.53s/step, acc=0.3721, loss=2.9612, lr=9.87e-05]
Training:   9%|▉         | 1011/11082 [3:15:07<32:33:19, 11.64s/step, acc=0.3721, loss=2.9612, lr=9.87e-05]
Training:   9%|▉         | 1012/11082 [3:15:19<32:22:26, 11.57s/step, acc=0.3721, loss=2.9612, lr=9.87e-05]
Training:   9%|▉         | 1013/11082 [3:15:30<32:40:56, 11.69s/step, acc=0.3721, loss=2.9612, lr=9.87e-05]
Training:   9%|▉         | 1014/11082 [3:15:42<32:08:29, 11.49s/step, acc=0.3721, loss=2.9612, lr=9.87e-05]
Training:   9%|▉         | 1015/11082 [3:15:53<32:30:23, 11.62s/step, acc=0.3721, loss=2.9612, lr=9.87e-05]
Training:   9%|▉         | 1016/11082 [3:16:05<32:16:16, 11.54s/step, acc=0.3721, loss=2.9612, lr=9.87e-05]
Training:   9%|▉         | 1017/11082 [3:16:16<32:04:14, 11.47s/step, acc=0.3721, loss=2.9612, lr=9.87e-05]
Training:   9%|▉         | 1018/11082 [3:16:28<32:13:40, 11.53s/step, acc=0.3721, loss=2.9612, lr=9.87e-05]
Training:   9%|▉         | 1019/11082 [3:16:39<32:08:56, 11.50s/step, acc=0.3721, loss=2.9612, lr=9.87e-05]
Training:   9%|▉         | 1020/11082 [3:16:51<32:29:20, 11.62s/step, acc=0.3721, loss=2.9612, lr=9.87e-05]
Training:   9%|▉         | 1021/11082 [3:17:03<32:21:24, 11.58s/step, acc=0.3721, loss=2.9612, lr=9.87e-05]
Training:   9%|▉         | 1022/11082 [3:17:14<32:12:04, 11.52s/step, acc=0.3721, loss=2.9612, lr=9.87e-05]
Training:   9%|▉         | 1023/11082 [3:17:25<31:55:42, 11.43s/step, acc=0.3721, loss=2.9612, lr=9.87e-05]
Training:   9%|▉         | 1024/11082 [3:17:38<33:06:08, 11.85s/step, acc=0.3721, loss=2.9612, lr=9.87e-05]
Training:   9%|▉         | 1025/11082 [3:17:50<32:57:29, 11.80s/step, acc=0.3721, loss=2.9612, lr=9.87e-05]
Training:   9%|▉         | 1026/11082 [3:18:01<32:33:05, 11.65s/step, acc=0.3721, loss=2.9612, lr=9.87e-05]
Training:   9%|▉         | 1027/11082 [3:18:13<32:26:18, 11.61s/step, acc=0.3721, loss=2.9612, lr=9.87e-05]
Training:   9%|▉         | 1028/11082 [3:18:24<32:14:30, 11.54s/step, acc=0.3721, loss=2.9612, lr=9.87e-05]
Training:   9%|▉         | 1029/11082 [3:18:36<32:32:28, 11.65s/step, acc=0.3721, loss=2.9612, lr=9.87e-05]
Training:   9%|▉         | 1030/11082 [3:18:47<32:19:05, 11.57s/step, acc=0.3721, loss=2.9612, lr=9.87e-05]
Training:   9%|▉         | 1031/11082 [3:18:59<32:20:07, 11.58s/step, acc=0.3721, loss=2.9612, lr=9.87e-05]
Training:   9%|▉         | 1032/11082 [3:19:11<32:28:48, 11.63s/step, acc=0.3721, loss=2.9612, lr=9.87e-05]
Training:   9%|▉         | 1033/11082 [3:19:22<32:22:51, 11.60s/step, acc=0.3721, loss=2.9612, lr=9.87e-05]
Training:   9%|▉         | 1034/11082 [3:19:33<32:07:49, 11.51s/step, acc=0.3721, loss=2.9612, lr=9.87e-05]
Training:   9%|▉         | 1035/11082 [3:19:45<32:05:47, 11.50s/step, acc=0.3721, loss=2.9612, lr=9.87e-05]
Training:   9%|▉         | 1036/11082 [3:19:56<31:49:50, 11.41s/step, acc=0.3721, loss=2.9612, lr=9.87e-05]
Training:   9%|▉         | 1037/11082 [3:20:08<32:06:35, 11.51s/step, acc=0.3721, loss=2.9612, lr=9.87e-05]
Training:   9%|▉         | 1038/11082 [3:20:20<32:20:52, 11.59s/step, acc=0.3721, loss=2.9612, lr=9.87e-05]
Training:   9%|▉         | 1039/11082 [3:20:31<32:07:56, 11.52s/step, acc=0.3721, loss=2.9612, lr=9.87e-05]
Training:   9%|▉         | 1040/11082 [3:20:42<31:57:40, 11.46s/step, acc=0.3721, loss=2.9612, lr=9.87e-05]
Training:   9%|▉         | 1041/11082 [3:20:54<32:14:59, 11.56s/step, acc=0.3721, loss=2.9612, lr=9.87e-05]
Training:   9%|▉         | 1042/11082 [3:21:06<32:24:36, 11.62s/step, acc=0.3721, loss=2.9612, lr=9.87e-05]
Training:   9%|▉         | 1043/11082 [3:21:17<32:10:35, 11.54s/step, acc=0.3721, loss=2.9612, lr=9.87e-05]
Training:   9%|▉         | 1044/11082 [3:21:29<32:19:34, 11.59s/step, acc=0.3721, loss=2.9612, lr=9.87e-05]
Training:   9%|▉         | 1045/11082 [3:21:40<32:12:51, 11.55s/step, acc=0.3721, loss=2.9612, lr=9.87e-05]
Training:   9%|▉         | 1046/11082 [3:21:52<32:05:41, 11.51s/step, acc=0.3721, loss=2.9612, lr=9.87e-05]
Training:   9%|▉         | 1047/11082 [3:22:03<31:57:52, 11.47s/step, acc=0.3721, loss=2.9612, lr=9.87e-05]
Training:   9%|▉         | 1048/11082 [3:22:15<32:01:08, 11.49s/step, acc=0.3721, loss=2.9612, lr=9.87e-05]
Training:   9%|▉         | 1049/11082 [3:22:26<32:17:36, 11.59s/step, acc=0.3721, loss=2.9612, lr=9.87e-05]
Training:   9%|▉         | 1050/11082 [3:22:38<31:53:10, 11.44s/step, acc=0.3721, loss=2.9612, lr=9.87e-05]
Training:   9%|▉         | 1050/11082 [3:22:45<31:53:10, 11.44s/step, acc=0.3713, loss=2.9637, lr=9.86e-05]2026-06-18 18:16:13 | INFO | __main__ | epoch=0 step=1050 train_loss=2.963656 train_acc=0.371281 lr=9.855e-05 grad_norm=1.4531 acc_d0=0.6408 acc_d1=0.4716 acc_d2=0.3392 acc_d3=0.2457 acc_d4=0.1862 acc_d5=0.1496 acc_d6=0.1249 acc_d7=0.1084

Training:   9%|▉         | 1051/11082 [3:22:50<32:23:28, 11.62s/step, acc=0.3713, loss=2.9637, lr=9.86e-05]
Training:   9%|▉         | 1052/11082 [3:23:01<32:02:03, 11.50s/step, acc=0.3713, loss=2.9637, lr=9.86e-05]
Training:  10%|▉         | 1053/11082 [3:23:12<32:02:50, 11.50s/step, acc=0.3713, loss=2.9637, lr=9.86e-05]
Training:  10%|▉         | 1054/11082 [3:23:24<32:03:30, 11.51s/step, acc=0.3713, loss=2.9637, lr=9.86e-05]
Training:  10%|▉         | 1055/11082 [3:23:35<31:55:37, 11.46s/step, acc=0.3713, loss=2.9637, lr=9.86e-05]
Training:  10%|▉         | 1056/11082 [3:23:47<32:10:00, 11.55s/step, acc=0.3713, loss=2.9637, lr=9.86e-05]
Training:  10%|▉         | 1057/11082 [3:23:59<32:20:13, 11.61s/step, acc=0.3713, loss=2.9637, lr=9.86e-05]
Training:  10%|▉         | 1058/11082 [3:24:10<31:53:56, 11.46s/step, acc=0.3713, loss=2.9637, lr=9.86e-05]
Training:  10%|▉         | 1059/11082 [3:24:22<32:09:44, 11.55s/step, acc=0.3713, loss=2.9637, lr=9.86e-05]
Training:  10%|▉         | 1060/11082 [3:24:33<32:06:21, 11.53s/step, acc=0.3713, loss=2.9637, lr=9.86e-05]
Training:  10%|▉         | 1061/11082 [3:24:44<31:48:18, 11.43s/step, acc=0.3713, loss=2.9637, lr=9.86e-05]
Training:  10%|▉         | 1062/11082 [3:24:56<32:18:40, 11.61s/step, acc=0.3713, loss=2.9637, lr=9.86e-05]
Training:  10%|▉         | 1063/11082 [3:25:07<31:53:55, 11.46s/step, acc=0.3713, loss=2.9637, lr=9.86e-05]
Training:  10%|▉         | 1064/11082 [3:25:19<32:01:49, 11.51s/step, acc=0.3713, loss=2.9637, lr=9.86e-05]
Training:  10%|▉         | 1065/11082 [3:25:31<32:07:57, 11.55s/step, acc=0.3713, loss=2.9637, lr=9.86e-05]
Training:  10%|▉         | 1066/11082 [3:25:42<31:50:07, 11.44s/step, acc=0.3713, loss=2.9637, lr=9.86e-05]
Training:  10%|▉         | 1067/11082 [3:25:54<32:17:48, 11.61s/step, acc=0.3713, loss=2.9637, lr=9.86e-05]
Training:  10%|▉         | 1068/11082 [3:26:05<32:06:25, 11.54s/step, acc=0.3713, loss=2.9637, lr=9.86e-05]
Training:  10%|▉         | 1069/11082 [3:26:17<31:59:42, 11.50s/step, acc=0.3713, loss=2.9637, lr=9.86e-05]
Training:  10%|▉         | 1070/11082 [3:26:28<32:03:34, 11.53s/step, acc=0.3713, loss=2.9637, lr=9.86e-05]
Training:  10%|▉         | 1071/11082 [3:26:40<31:58:40, 11.50s/step, acc=0.3713, loss=2.9637, lr=9.86e-05]
Training:  10%|▉         | 1072/11082 [3:26:51<32:12:16, 11.58s/step, acc=0.3713, loss=2.9637, lr=9.86e-05]
Training:  10%|▉         | 1073/11082 [3:27:02<31:43:14, 11.41s/step, acc=0.3713, loss=2.9637, lr=9.86e-05]
Training:  10%|▉         | 1074/11082 [3:27:14<32:06:55, 11.55s/step, acc=0.3713, loss=2.9637, lr=9.86e-05]
Training:  10%|▉         | 1075/11082 [3:27:26<32:17:25, 11.62s/step, acc=0.3713, loss=2.9637, lr=9.86e-05]
Training:  10%|▉         | 1076/11082 [3:27:37<31:33:57, 11.36s/step, acc=0.3713, loss=2.9637, lr=9.86e-05]
Training:  10%|▉         | 1077/11082 [3:27:49<32:03:00, 11.53s/step, acc=0.3713, loss=2.9637, lr=9.86e-05]
Training:  10%|▉         | 1078/11082 [3:28:00<32:02:38, 11.53s/step, acc=0.3713, loss=2.9637, lr=9.86e-05]
Training:  10%|▉         | 1079/11082 [3:28:12<31:52:48, 11.47s/step, acc=0.3713, loss=2.9637, lr=9.86e-05]
Training:  10%|▉         | 1080/11082 [3:28:24<32:11:18, 11.59s/step, acc=0.3713, loss=2.9637, lr=9.86e-05]
Training:  10%|▉         | 1081/11082 [3:28:35<31:53:49, 11.48s/step, acc=0.3713, loss=2.9637, lr=9.86e-05]
Training:  10%|▉         | 1082/11082 [3:28:46<31:59:44, 11.52s/step, acc=0.3713, loss=2.9637, lr=9.86e-05]
Training:  10%|▉         | 1083/11082 [3:28:58<32:07:17, 11.56s/step, acc=0.3713, loss=2.9637, lr=9.86e-05]
Training:  10%|▉         | 1084/11082 [3:29:09<31:43:45, 11.42s/step, acc=0.3713, loss=2.9637, lr=9.86e-05]
Training:  10%|▉         | 1085/11082 [3:29:21<32:01:27, 11.53s/step, acc=0.3713, loss=2.9637, lr=9.86e-05]
Training:  10%|▉         | 1086/11082 [3:29:33<32:09:36, 11.58s/step, acc=0.3713, loss=2.9637, lr=9.86e-05]
Training:  10%|▉         | 1087/11082 [3:29:44<32:09:04, 11.58s/step, acc=0.3713, loss=2.9637, lr=9.86e-05]
Training:  10%|▉         | 1088/11082 [3:29:56<31:56:00, 11.50s/step, acc=0.3713, loss=2.9637, lr=9.86e-05]
Training:  10%|▉         | 1089/11082 [3:30:07<31:58:05, 11.52s/step, acc=0.3713, loss=2.9637, lr=9.86e-05]
Training:  10%|▉         | 1090/11082 [3:30:18<31:49:48, 11.47s/step, acc=0.3713, loss=2.9637, lr=9.86e-05]
Training:  10%|▉         | 1091/11082 [3:30:30<31:42:38, 11.43s/step, acc=0.3713, loss=2.9637, lr=9.86e-05]
Training:  10%|▉         | 1092/11082 [3:30:41<31:43:52, 11.43s/step, acc=0.3713, loss=2.9637, lr=9.86e-05]
Training:  10%|▉         | 1093/11082 [3:30:53<32:00:14, 11.53s/step, acc=0.3713, loss=2.9637, lr=9.86e-05]
Training:  10%|▉         | 1094/11082 [3:31:05<32:08:31, 11.59s/step, acc=0.3713, loss=2.9637, lr=9.86e-05]
Training:  10%|▉         | 1095/11082 [3:31:16<32:05:00, 11.57s/step, acc=0.3713, loss=2.9637, lr=9.86e-05]
Training:  10%|▉         | 1096/11082 [3:31:28<32:14:10, 11.62s/step, acc=0.3713, loss=2.9637, lr=9.86e-05]
Training:  10%|▉         | 1097/11082 [3:31:39<31:50:22, 11.48s/step, acc=0.3713, loss=2.9637, lr=9.86e-05]
Training:  10%|▉         | 1098/11082 [3:31:50<31:41:33, 11.43s/step, acc=0.3713, loss=2.9637, lr=9.86e-05]
Training:  10%|▉         | 1099/11082 [3:32:03<32:19:24, 11.66s/step, acc=0.3713, loss=2.9637, lr=9.86e-05]
Training:  10%|▉         | 1100/11082 [3:32:14<32:01:36, 11.55s/step, acc=0.3713, loss=2.9637, lr=9.86e-05]
Training:  10%|▉         | 1100/11082 [3:32:21<32:01:36, 11.55s/step, acc=0.3764, loss=2.9213, lr=9.84e-05]2026-06-18 18:25:49 | INFO | __main__ | epoch=0 step=1100 train_loss=2.921294 train_acc=0.376433 lr=9.838e-05 grad_norm=1.2422 acc_d0=0.6362 acc_d1=0.4701 acc_d2=0.3393 acc_d3=0.2465 acc_d4=0.1884 acc_d5=0.1508 acc_d6=0.1269 acc_d7=0.1103

Training:  10%|▉         | 1101/11082 [3:32:26<32:07:19, 11.59s/step, acc=0.3764, loss=2.9213, lr=9.84e-05]
Training:  10%|▉         | 1102/11082 [3:32:38<32:32:24, 11.74s/step, acc=0.3764, loss=2.9213, lr=9.84e-05]
Training:  10%|▉         | 1103/11082 [3:32:49<32:01:40, 11.55s/step, acc=0.3764, loss=2.9213, lr=9.84e-05]
Training:  10%|▉         | 1104/11082 [3:33:01<32:32:04, 11.74s/step, acc=0.3764, loss=2.9213, lr=9.84e-05]
Training:  10%|▉         | 1105/11082 [3:33:13<32:33:08, 11.75s/step, acc=0.3764, loss=2.9213, lr=9.84e-05]
Training:  10%|▉         | 1106/11082 [3:33:24<32:07:27, 11.59s/step, acc=0.3764, loss=2.9213, lr=9.84e-05]
Training:  10%|▉         | 1107/11082 [3:33:36<32:52:28, 11.86s/step, acc=0.3764, loss=2.9213, lr=9.84e-05]
Training:  10%|▉         | 1108/11082 [3:33:48<32:32:41, 11.75s/step, acc=0.3764, loss=2.9213, lr=9.84e-05]
Training:  10%|█         | 1109/11082 [3:34:00<32:23:46, 11.69s/step, acc=0.3764, loss=2.9213, lr=9.84e-05]
Training:  10%|█         | 1110/11082 [3:34:11<32:26:27, 11.71s/step, acc=0.3764, loss=2.9213, lr=9.84e-05]
Training:  10%|█         | 1111/11082 [3:34:22<32:01:22, 11.56s/step, acc=0.3764, loss=2.9213, lr=9.84e-05]
Training:  10%|█         | 1112/11082 [3:34:34<32:08:43, 11.61s/step, acc=0.3764, loss=2.9213, lr=9.84e-05]
Training:  10%|█         | 1113/11082 [3:34:46<31:58:50, 11.55s/step, acc=0.3764, loss=2.9213, lr=9.84e-05]
Training:  10%|█         | 1114/11082 [3:34:57<31:44:57, 11.47s/step, acc=0.3764, loss=2.9213, lr=9.84e-05]
Training:  10%|█         | 1115/11082 [3:35:09<32:04:33, 11.59s/step, acc=0.3764, loss=2.9213, lr=9.84e-05]
Training:  10%|█         | 1116/11082 [3:35:20<32:03:18, 11.58s/step, acc=0.3764, loss=2.9213, lr=9.84e-05]
Training:  10%|█         | 1117/11082 [3:35:32<31:45:42, 11.47s/step, acc=0.3764, loss=2.9213, lr=9.84e-05]
Training:  10%|█         | 1118/11082 [3:35:43<31:38:06, 11.43s/step, acc=0.3764, loss=2.9213, lr=9.84e-05]
Training:  10%|█         | 1119/11082 [3:35:55<31:52:57, 11.52s/step, acc=0.3764, loss=2.9213, lr=9.84e-05]
Training:  10%|█         | 1120/11082 [3:36:06<31:44:12, 11.47s/step, acc=0.3764, loss=2.9213, lr=9.84e-05]
Training:  10%|█         | 1121/11082 [3:36:18<32:14:53, 11.65s/step, acc=0.3764, loss=2.9213, lr=9.84e-05]
Training:  10%|█         | 1122/11082 [3:36:29<31:58:30, 11.56s/step, acc=0.3764, loss=2.9213, lr=9.84e-05]
Training:  10%|█         | 1123/11082 [3:36:41<31:44:03, 11.47s/step, acc=0.3764, loss=2.9213, lr=9.84e-05]
Training:  10%|█         | 1124/11082 [3:36:52<31:52:55, 11.53s/step, acc=0.3764, loss=2.9213, lr=9.84e-05]
Training:  10%|█         | 1125/11082 [3:37:04<32:09:29, 11.63s/step, acc=0.3764, loss=2.9213, lr=9.84e-05]
Training:  10%|█         | 1126/11082 [3:37:16<31:57:05, 11.55s/step, acc=0.3764, loss=2.9213, lr=9.84e-05]
Training:  10%|█         | 1127/11082 [3:37:28<32:18:49, 11.69s/step, acc=0.3764, loss=2.9213, lr=9.84e-05]
Training:  10%|█         | 1128/11082 [3:37:38<31:40:44, 11.46s/step, acc=0.3764, loss=2.9213, lr=9.84e-05]
Training:  10%|█         | 1129/11082 [3:37:50<31:58:12, 11.56s/step, acc=0.3764, loss=2.9213, lr=9.84e-05]
Training:  10%|█         | 1130/11082 [3:38:02<31:52:03, 11.53s/step, acc=0.3764, loss=2.9213, lr=9.84e-05]
Training:  10%|█         | 1131/11082 [3:38:13<31:51:08, 11.52s/step, acc=0.3764, loss=2.9213, lr=9.84e-05]
Training:  10%|█         | 1132/11082 [3:38:25<32:05:44, 11.61s/step, acc=0.3764, loss=2.9213, lr=9.84e-05]
Training:  10%|█         | 1133/11082 [3:38:37<32:03:03, 11.60s/step, acc=0.3764, loss=2.9213, lr=9.84e-05]
Training:  10%|█         | 1134/11082 [3:38:48<31:42:46, 11.48s/step, acc=0.3764, loss=2.9213, lr=9.84e-05]
Training:  10%|█         | 1135/11082 [3:38:59<31:49:34, 11.52s/step, acc=0.3764, loss=2.9213, lr=9.84e-05]
Training:  10%|█         | 1136/11082 [3:39:11<31:54:26, 11.55s/step, acc=0.3764, loss=2.9213, lr=9.84e-05]
Training:  10%|█         | 1137/11082 [3:39:23<31:57:08, 11.57s/step, acc=0.3764, loss=2.9213, lr=9.84e-05]
Training:  10%|█         | 1138/11082 [3:39:34<31:36:53, 11.45s/step, acc=0.3764, loss=2.9213, lr=9.84e-05]
Training:  10%|█         | 1139/11082 [3:39:45<31:49:12, 11.52s/step, acc=0.3764, loss=2.9213, lr=9.84e-05]
Training:  10%|█         | 1140/11082 [3:39:57<32:12:31, 11.66s/step, acc=0.3764, loss=2.9213, lr=9.84e-05]
Training:  10%|█         | 1141/11082 [3:40:08<31:24:36, 11.37s/step, acc=0.3764, loss=2.9213, lr=9.84e-05]
Training:  10%|█         | 1142/11082 [3:40:20<31:50:59, 11.54s/step, acc=0.3764, loss=2.9213, lr=9.84e-05]
Training:  10%|█         | 1143/11082 [3:40:32<31:58:10, 11.58s/step, acc=0.3764, loss=2.9213, lr=9.84e-05]
Training:  10%|█         | 1144/11082 [3:40:43<31:56:35, 11.57s/step, acc=0.3764, loss=2.9213, lr=9.84e-05]
Training:  10%|█         | 1145/11082 [3:40:54<31:25:30, 11.38s/step, acc=0.3764, loss=2.9213, lr=9.84e-05]
Training:  10%|█         | 1146/11082 [3:41:06<32:02:07, 11.61s/step, acc=0.3764, loss=2.9213, lr=9.84e-05]
Training:  10%|█         | 1147/11082 [3:41:18<31:56:56, 11.58s/step, acc=0.3764, loss=2.9213, lr=9.84e-05]
Training:  10%|█         | 1148/11082 [3:41:30<32:06:16, 11.63s/step, acc=0.3764, loss=2.9213, lr=9.84e-05]
Training:  10%|█         | 1149/11082 [3:41:41<31:59:48, 11.60s/step, acc=0.3764, loss=2.9213, lr=9.84e-05]
Training:  10%|█         | 1150/11082 [3:41:53<31:45:47, 11.51s/step, acc=0.3764, loss=2.9213, lr=9.84e-05]
Training:  10%|█         | 1150/11082 [3:42:00<31:45:47, 11.51s/step, acc=0.3830, loss=2.8727, lr=9.82e-05]2026-06-18 18:35:27 | INFO | __main__ | epoch=0 step=1150 train_loss=2.872731 train_acc=0.383038 lr=9.821e-05 grad_norm=1.6953 acc_d0=0.6555 acc_d1=0.4897 acc_d2=0.3573 acc_d3=0.2608 acc_d4=0.1990 acc_d5=0.1607 acc_d6=0.1344 acc_d7=0.1155

Training:  10%|█         | 1151/11082 [3:42:04<32:01:55, 11.61s/step, acc=0.3830, loss=2.8727, lr=9.82e-05]
Training:  10%|█         | 1152/11082 [3:42:15<31:34:04, 11.44s/step, acc=0.3830, loss=2.8727, lr=9.82e-05]
Training:  10%|█         | 1153/11082 [3:42:27<31:41:41, 11.49s/step, acc=0.3830, loss=2.8727, lr=9.82e-05]
Training:  10%|█         | 1154/11082 [3:42:39<31:53:07, 11.56s/step, acc=0.3830, loss=2.8727, lr=9.82e-05]
Training:  10%|█         | 1155/11082 [3:42:50<31:30:42, 11.43s/step, acc=0.3830, loss=2.8727, lr=9.82e-05]
Training:  10%|█         | 1156/11082 [3:43:01<31:39:48, 11.48s/step, acc=0.3830, loss=2.8727, lr=9.82e-05]
Training:  10%|█         | 1157/11082 [3:43:13<31:56:03, 11.58s/step, acc=0.3830, loss=2.8727, lr=9.82e-05]
Training:  10%|█         | 1158/11082 [3:43:24<31:26:02, 11.40s/step, acc=0.3830, loss=2.8727, lr=9.82e-05]
Training:  10%|█         | 1159/11082 [3:43:36<32:02:48, 11.63s/step, acc=0.3830, loss=2.8727, lr=9.82e-05]
Training:  10%|█         | 1160/11082 [3:43:48<31:39:15, 11.49s/step, acc=0.3830, loss=2.8727, lr=9.82e-05]
Training:  10%|█         | 1161/11082 [3:43:59<31:44:33, 11.52s/step, acc=0.3830, loss=2.8727, lr=9.82e-05]
Training:  10%|█         | 1162/11082 [3:44:11<31:50:16, 11.55s/step, acc=0.3830, loss=2.8727, lr=9.82e-05]
Training:  10%|█         | 1163/11082 [3:44:22<31:34:34, 11.46s/step, acc=0.3830, loss=2.8727, lr=9.82e-05]
Training:  11%|█         | 1164/11082 [3:44:34<31:35:31, 11.47s/step, acc=0.3830, loss=2.8727, lr=9.82e-05]
Training:  11%|█         | 1165/11082 [3:44:45<31:18:35, 11.37s/step, acc=0.3830, loss=2.8727, lr=9.82e-05]
Training:  11%|█         | 1166/11082 [3:44:57<31:56:25, 11.60s/step, acc=0.3830, loss=2.8727, lr=9.82e-05]
Training:  11%|█         | 1167/11082 [3:45:08<31:39:26, 11.49s/step, acc=0.3830, loss=2.8727, lr=9.82e-05]
Training:  11%|█         | 1168/11082 [3:45:20<31:44:07, 11.52s/step, acc=0.3830, loss=2.8727, lr=9.82e-05]
Training:  11%|█         | 1169/11082 [3:45:31<32:00:56, 11.63s/step, acc=0.3830, loss=2.8727, lr=9.82e-05]
Training:  11%|█         | 1170/11082 [3:45:43<31:55:25, 11.59s/step, acc=0.3830, loss=2.8727, lr=9.82e-05]
Training:  11%|█         | 1171/11082 [3:45:54<31:46:19, 11.54s/step, acc=0.3830, loss=2.8727, lr=9.82e-05]
Training:  11%|█         | 1172/11082 [3:46:06<31:45:15, 11.54s/step, acc=0.3830, loss=2.8727, lr=9.82e-05]
Training:  11%|█         | 1173/11082 [3:46:17<31:27:54, 11.43s/step, acc=0.3830, loss=2.8727, lr=9.82e-05]
Training:  11%|█         | 1174/11082 [3:46:29<31:38:51, 11.50s/step, acc=0.3830, loss=2.8727, lr=9.82e-05]
Training:  11%|█         | 1175/11082 [3:46:40<31:44:04, 11.53s/step, acc=0.3830, loss=2.8727, lr=9.82e-05]
Training:  11%|█         | 1176/11082 [3:46:52<32:01:32, 11.64s/step, acc=0.3830, loss=2.8727, lr=9.82e-05]
Training:  11%|█         | 1177/11082 [3:47:04<32:13:38, 11.71s/step, acc=0.3830, loss=2.8727, lr=9.82e-05]
Training:  11%|█         | 1178/11082 [3:47:15<31:46:41, 11.55s/step, acc=0.3830, loss=2.8727, lr=9.82e-05]
Training:  11%|█         | 1179/11082 [3:47:27<32:02:36, 11.65s/step, acc=0.3830, loss=2.8727, lr=9.82e-05]
Training:  11%|█         | 1180/11082 [3:47:38<31:24:42, 11.42s/step, acc=0.3830, loss=2.8727, lr=9.82e-05]
Training:  11%|█         | 1181/11082 [3:47:50<32:00:27, 11.64s/step, acc=0.3830, loss=2.8727, lr=9.82e-05]
Training:  11%|█         | 1182/11082 [3:48:02<31:44:22, 11.54s/step, acc=0.3830, loss=2.8727, lr=9.82e-05]
Training:  11%|█         | 1183/11082 [3:48:13<31:43:54, 11.54s/step, acc=0.3830, loss=2.8727, lr=9.82e-05]
Training:  11%|█         | 1184/11082 [3:48:24<31:29:07, 11.45s/step, acc=0.3830, loss=2.8727, lr=9.82e-05]
Training:  11%|█         | 1185/11082 [3:48:36<31:43:48, 11.54s/step, acc=0.3830, loss=2.8727, lr=9.82e-05]
Training:  11%|█         | 1186/11082 [3:48:48<31:37:33, 11.50s/step, acc=0.3830, loss=2.8727, lr=9.82e-05]
Training:  11%|█         | 1187/11082 [3:49:00<32:09:06, 11.70s/step, acc=0.3830, loss=2.8727, lr=9.82e-05]
Training:  11%|█         | 1188/11082 [3:49:11<31:54:31, 11.61s/step, acc=0.3830, loss=2.8727, lr=9.82e-05]
Training:  11%|█         | 1189/11082 [3:49:23<31:58:48, 11.64s/step, acc=0.3830, loss=2.8727, lr=9.82e-05]
Training:  11%|█         | 1190/11082 [3:49:34<31:57:53, 11.63s/step, acc=0.3830, loss=2.8727, lr=9.82e-05]
Training:  11%|█         | 1191/11082 [3:49:46<31:52:19, 11.60s/step, acc=0.3830, loss=2.8727, lr=9.82e-05]
Training:  11%|█         | 1192/11082 [3:49:58<32:00:17, 11.65s/step, acc=0.3830, loss=2.8727, lr=9.82e-05]
Training:  11%|█         | 1193/11082 [3:50:09<31:35:31, 11.50s/step, acc=0.3830, loss=2.8727, lr=9.82e-05]
Training:  11%|█         | 1194/11082 [3:50:21<31:50:35, 11.59s/step, acc=0.3830, loss=2.8727, lr=9.82e-05]
Training:  11%|█         | 1195/11082 [3:50:32<31:47:19, 11.57s/step, acc=0.3830, loss=2.8727, lr=9.82e-05]
Training:  11%|█         | 1196/11082 [3:50:43<31:16:29, 11.39s/step, acc=0.3830, loss=2.8727, lr=9.82e-05]
Training:  11%|█         | 1197/11082 [3:50:55<31:45:08, 11.56s/step, acc=0.3830, loss=2.8727, lr=9.82e-05]
Training:  11%|█         | 1198/11082 [3:51:07<31:54:00, 11.62s/step, acc=0.3830, loss=2.8727, lr=9.82e-05]
Training:  11%|█         | 1199/11082 [3:51:18<31:46:07, 11.57s/step, acc=0.3830, loss=2.8727, lr=9.82e-05]
Training:  11%|█         | 1200/11082 [3:51:30<31:53:10, 11.62s/step, acc=0.3830, loss=2.8727, lr=9.82e-05]
Training:  11%|█         | 1200/11082 [3:51:37<31:53:10, 11.62s/step, acc=0.3863, loss=2.8375, lr=9.80e-05]2026-06-18 18:45:05 | INFO | __main__ | epoch=0 step=1200 train_loss=2.837458 train_acc=0.386283 lr=9.802e-05 grad_norm=1.7656 acc_d0=0.6574 acc_d1=0.4940 acc_d2=0.3618 acc_d3=0.2650 acc_d4=0.2020 acc_d5=0.1608 acc_d6=0.1336 acc_d7=0.1156

Training:  11%|█         | 1201/11082 [3:51:41<31:41:04, 11.54s/step, acc=0.3863, loss=2.8375, lr=9.80e-05]
Training:  11%|█         | 1202/11082 [3:51:53<31:23:49, 11.44s/step, acc=0.3863, loss=2.8375, lr=9.80e-05]
Training:  11%|█         | 1203/11082 [3:52:04<31:15:43, 11.39s/step, acc=0.3863, loss=2.8375, lr=9.80e-05]
Training:  11%|█         | 1204/11082 [3:52:16<31:41:55, 11.55s/step, acc=0.3863, loss=2.8375, lr=9.80e-05]
Training:  11%|█         | 1205/11082 [3:52:27<31:37:22, 11.53s/step, acc=0.3863, loss=2.8375, lr=9.80e-05]
Training:  11%|█         | 1206/11082 [3:52:39<31:36:58, 11.52s/step, acc=0.3863, loss=2.8375, lr=9.80e-05]
Training:  11%|█         | 1207/11082 [3:52:51<31:53:10, 11.62s/step, acc=0.3863, loss=2.8375, lr=9.80e-05]
Training:  11%|█         | 1208/11082 [3:53:02<31:30:47, 11.49s/step, acc=0.3863, loss=2.8375, lr=9.80e-05]
Training:  11%|█         | 1209/11082 [3:53:14<31:41:37, 11.56s/step, acc=0.3863, loss=2.8375, lr=9.80e-05]
Training:  11%|█         | 1210/11082 [3:53:25<31:50:34, 11.61s/step, acc=0.3863, loss=2.8375, lr=9.80e-05]
Training:  11%|█         | 1211/11082 [3:53:37<31:31:46, 11.50s/step, acc=0.3863, loss=2.8375, lr=9.80e-05]
Training:  11%|█         | 1212/11082 [3:53:48<31:32:27, 11.50s/step, acc=0.3863, loss=2.8375, lr=9.80e-05]
Training:  11%|█         | 1213/11082 [3:54:00<31:34:36, 11.52s/step, acc=0.3863, loss=2.8375, lr=9.80e-05]
Training:  11%|█         | 1214/11082 [3:54:11<31:35:33, 11.53s/step, acc=0.3863, loss=2.8375, lr=9.80e-05]
Training:  11%|█         | 1215/11082 [3:54:23<31:46:53, 11.60s/step, acc=0.3863, loss=2.8375, lr=9.80e-05]
Training:  11%|█         | 1216/11082 [3:54:34<31:35:07, 11.53s/step, acc=0.3863, loss=2.8375, lr=9.80e-05]
Training:  11%|█         | 1217/11082 [3:54:45<31:11:38, 11.38s/step, acc=0.3863, loss=2.8375, lr=9.80e-05]
Training:  11%|█         | 1218/11082 [3:54:57<31:43:23, 11.58s/step, acc=0.3863, loss=2.8375, lr=9.80e-05]
Training:  11%|█         | 1219/11082 [3:55:09<31:23:11, 11.46s/step, acc=0.3863, loss=2.8375, lr=9.80e-05]
Training:  11%|█         | 1220/11082 [3:55:20<31:34:20, 11.53s/step, acc=0.3863, loss=2.8375, lr=9.80e-05]
Training:  11%|█         | 1221/11082 [3:55:31<31:18:34, 11.43s/step, acc=0.3863, loss=2.8375, lr=9.80e-05]
Training:  11%|█         | 1222/11082 [3:55:43<31:48:15, 11.61s/step, acc=0.3863, loss=2.8375, lr=9.80e-05]
Training:  11%|█         | 1223/11082 [3:55:55<31:42:28, 11.58s/step, acc=0.3863, loss=2.8375, lr=9.80e-05]
Training:  11%|█         | 1224/11082 [3:56:07<31:51:18, 11.63s/step, acc=0.3863, loss=2.8375, lr=9.80e-05]
Training:  11%|█         | 1225/11082 [3:56:18<31:41:03, 11.57s/step, acc=0.3863, loss=2.8375, lr=9.80e-05]
Training:  11%|█         | 1226/11082 [3:56:30<31:44:05, 11.59s/step, acc=0.3863, loss=2.8375, lr=9.80e-05]
Training:  11%|█         | 1227/11082 [3:56:41<31:39:44, 11.57s/step, acc=0.3863, loss=2.8375, lr=9.80e-05]
Training:  11%|█         | 1228/11082 [3:56:53<31:41:26, 11.58s/step, acc=0.3863, loss=2.8375, lr=9.80e-05]
Training:  11%|█         | 1229/11082 [3:57:04<31:28:24, 11.50s/step, acc=0.3863, loss=2.8375, lr=9.80e-05]
Training:  11%|█         | 1230/11082 [3:57:16<31:48:05, 11.62s/step, acc=0.3863, loss=2.8375, lr=9.80e-05]
Training:  11%|█         | 1231/11082 [3:57:27<31:30:30, 11.51s/step, acc=0.3863, loss=2.8375, lr=9.80e-05]
Training:  11%|█         | 1232/11082 [3:57:39<31:25:03, 11.48s/step, acc=0.3863, loss=2.8375, lr=9.80e-05]
Training:  11%|█         | 1233/11082 [3:57:50<31:21:11, 11.46s/step, acc=0.3863, loss=2.8375, lr=9.80e-05]
Training:  11%|█         | 1234/11082 [3:58:02<31:33:14, 11.53s/step, acc=0.3863, loss=2.8375, lr=9.80e-05]
Training:  11%|█         | 1235/11082 [3:58:13<31:34:35, 11.54s/step, acc=0.3863, loss=2.8375, lr=9.80e-05]
Training:  11%|█         | 1236/11082 [3:58:25<31:30:47, 11.52s/step, acc=0.3863, loss=2.8375, lr=9.80e-05]
Training:  11%|█         | 1237/11082 [3:58:37<31:47:05, 11.62s/step, acc=0.3863, loss=2.8375, lr=9.80e-05]
Training:  11%|█         | 1238/11082 [3:58:48<31:39:51, 11.58s/step, acc=0.3863, loss=2.8375, lr=9.80e-05]
Training:  11%|█         | 1239/11082 [3:59:00<31:26:06, 11.50s/step, acc=0.3863, loss=2.8375, lr=9.80e-05]
Training:  11%|█         | 1240/11082 [3:59:11<31:37:42, 11.57s/step, acc=0.3863, loss=2.8375, lr=9.80e-05]
Training:  11%|█         | 1241/11082 [3:59:23<31:29:31, 11.52s/step, acc=0.3863, loss=2.8375, lr=9.80e-05]
Training:  11%|█         | 1242/11082 [3:59:34<31:29:39, 11.52s/step, acc=0.3863, loss=2.8375, lr=9.80e-05]
Training:  11%|█         | 1243/11082 [3:59:46<31:47:48, 11.63s/step, acc=0.3863, loss=2.8375, lr=9.80e-05]
Training:  11%|█         | 1244/11082 [3:59:57<31:23:00, 11.48s/step, acc=0.3863, loss=2.8375, lr=9.80e-05]
Training:  11%|█         | 1245/11082 [4:00:09<31:32:48, 11.55s/step, acc=0.3863, loss=2.8375, lr=9.80e-05]
Training:  11%|█         | 1246/11082 [4:00:21<31:58:58, 11.71s/step, acc=0.3863, loss=2.8375, lr=9.80e-05]
Training:  11%|█▏        | 1247/11082 [4:00:32<31:38:20, 11.58s/step, acc=0.3863, loss=2.8375, lr=9.80e-05]
Training:  11%|█▏        | 1248/11082 [4:00:44<31:47:56, 11.64s/step, acc=0.3863, loss=2.8375, lr=9.80e-05]
Training:  11%|█▏        | 1249/11082 [4:00:56<32:04:47, 11.74s/step, acc=0.3863, loss=2.8375, lr=9.80e-05]
Training:  11%|█▏        | 1250/11082 [4:01:08<32:01:02, 11.72s/step, acc=0.3863, loss=2.8375, lr=9.80e-05]
Training:  11%|█▏        | 1250/11082 [4:01:15<32:01:02, 11.72s/step, acc=0.3888, loss=2.8203, lr=9.78e-05]2026-06-18 18:54:43 | INFO | __main__ | epoch=0 step=1250 train_loss=2.820264 train_acc=0.388828 lr=9.783e-05 grad_norm=1.9219 acc_d0=0.6618 acc_d1=0.4934 acc_d2=0.3611 acc_d3=0.2647 acc_d4=0.2015 acc_d5=0.1611 acc_d6=0.1338 acc_d7=0.1143

Training:  11%|█▏        | 1251/11082 [4:01:19<31:48:20, 11.65s/step, acc=0.3888, loss=2.8203, lr=9.78e-05]
Training:  11%|█▏        | 1252/11082 [4:01:30<31:01:14, 11.36s/step, acc=0.3888, loss=2.8203, lr=9.78e-05]
Training:  11%|█▏        | 1253/11082 [4:01:42<31:43:48, 11.62s/step, acc=0.3888, loss=2.8203, lr=9.78e-05]
Training:  11%|█▏        | 1254/11082 [4:01:54<31:52:21, 11.67s/step, acc=0.3888, loss=2.8203, lr=9.78e-05]
Training:  11%|█▏        | 1255/11082 [4:02:05<31:41:31, 11.61s/step, acc=0.3888, loss=2.8203, lr=9.78e-05]
Training:  11%|█▏        | 1256/11082 [4:02:17<31:23:42, 11.50s/step, acc=0.3888, loss=2.8203, lr=9.78e-05]
Training:  11%|█▏        | 1257/11082 [4:02:29<31:41:29, 11.61s/step, acc=0.3888, loss=2.8203, lr=9.78e-05]
Training:  11%|█▏        | 1258/11082 [4:02:40<31:36:29, 11.58s/step, acc=0.3888, loss=2.8203, lr=9.78e-05]
Training:  11%|█▏        | 1259/11082 [4:02:51<31:25:33, 11.52s/step, acc=0.3888, loss=2.8203, lr=9.78e-05]
Training:  11%|█▏        | 1260/11082 [4:03:03<31:22:19, 11.50s/step, acc=0.3888, loss=2.8203, lr=9.78e-05]
Training:  11%|█▏        | 1261/11082 [4:03:15<31:32:23, 11.56s/step, acc=0.3888, loss=2.8203, lr=9.78e-05]
Training:  11%|█▏        | 1262/11082 [4:03:26<31:22:36, 11.50s/step, acc=0.3888, loss=2.8203, lr=9.78e-05]
Training:  11%|█▏        | 1263/11082 [4:03:37<31:19:18, 11.48s/step, acc=0.3888, loss=2.8203, lr=9.78e-05]
Training:  11%|█▏        | 1264/11082 [4:03:49<31:26:40, 11.53s/step, acc=0.3888, loss=2.8203, lr=9.78e-05]
Training:  11%|█▏        | 1265/11082 [4:04:01<31:56:46, 11.71s/step, acc=0.3888, loss=2.8203, lr=9.78e-05]
Training:  11%|█▏        | 1266/11082 [4:04:13<31:53:29, 11.70s/step, acc=0.3888, loss=2.8203, lr=9.78e-05]
Training:  11%|█▏        | 1267/11082 [4:04:25<32:06:23, 11.78s/step, acc=0.3888, loss=2.8203, lr=9.78e-05]
Training:  11%|█▏        | 1268/11082 [4:04:36<31:55:52, 11.71s/step, acc=0.3888, loss=2.8203, lr=9.78e-05]
Training:  11%|█▏        | 1269/11082 [4:04:48<31:53:35, 11.70s/step, acc=0.3888, loss=2.8203, lr=9.78e-05]
Training:  11%|█▏        | 1270/11082 [4:04:59<31:41:06, 11.63s/step, acc=0.3888, loss=2.8203, lr=9.78e-05]
Training:  11%|█▏        | 1271/11082 [4:05:11<31:46:18, 11.66s/step, acc=0.3888, loss=2.8203, lr=9.78e-05]
Training:  11%|█▏        | 1272/11082 [4:05:23<31:32:36, 11.58s/step, acc=0.3888, loss=2.8203, lr=9.78e-05]
Training:  11%|█▏        | 1273/11082 [4:05:34<31:39:41, 11.62s/step, acc=0.3888, loss=2.8203, lr=9.78e-05]
Training:  11%|█▏        | 1274/11082 [4:05:46<31:40:48, 11.63s/step, acc=0.3888, loss=2.8203, lr=9.78e-05]
Training:  12%|█▏        | 1275/11082 [4:05:57<31:31:15, 11.57s/step, acc=0.3888, loss=2.8203, lr=9.78e-05]
Training:  12%|█▏        | 1276/11082 [4:06:09<31:19:29, 11.50s/step, acc=0.3888, loss=2.8203, lr=9.78e-05]
Training:  12%|█▏        | 1277/11082 [4:06:21<31:46:03, 11.66s/step, acc=0.3888, loss=2.8203, lr=9.78e-05]
Training:  12%|█▏        | 1278/11082 [4:06:32<31:12:29, 11.46s/step, acc=0.3888, loss=2.8203, lr=9.78e-05]
Training:  12%|█▏        | 1279/11082 [4:06:43<31:06:19, 11.42s/step, acc=0.3888, loss=2.8203, lr=9.78e-05]
Training:  12%|█▏        | 1280/11082 [4:06:55<31:38:37, 11.62s/step, acc=0.3888, loss=2.8203, lr=9.78e-05]
Training:  12%|█▏        | 1281/11082 [4:07:07<31:40:08, 11.63s/step, acc=0.3888, loss=2.8203, lr=9.78e-05]
Training:  12%|█▏        | 1282/11082 [4:07:18<31:15:24, 11.48s/step, acc=0.3888, loss=2.8203, lr=9.78e-05]
Training:  12%|█▏        | 1283/11082 [4:07:30<31:38:33, 11.62s/step, acc=0.3888, loss=2.8203, lr=9.78e-05]
Training:  12%|█▏        | 1284/11082 [4:07:41<31:34:14, 11.60s/step, acc=0.3888, loss=2.8203, lr=9.78e-05]
Training:  12%|█▏        | 1285/11082 [4:07:53<31:26:11, 11.55s/step, acc=0.3888, loss=2.8203, lr=9.78e-05]
Training:  12%|█▏        | 1286/11082 [4:08:04<31:07:03, 11.44s/step, acc=0.3888, loss=2.8203, lr=9.78e-05]
Training:  12%|█▏        | 1287/11082 [4:08:16<31:08:53, 11.45s/step, acc=0.3888, loss=2.8203, lr=9.78e-05]
Training:  12%|█▏        | 1288/11082 [4:08:27<31:27:58, 11.57s/step, acc=0.3888, loss=2.8203, lr=9.78e-05]
Training:  12%|█▏        | 1289/11082 [4:08:39<31:34:44, 11.61s/step, acc=0.3888, loss=2.8203, lr=9.78e-05]
Training:  12%|█▏        | 1290/11082 [4:08:50<31:17:29, 11.50s/step, acc=0.3888, loss=2.8203, lr=9.78e-05]
Training:  12%|█▏        | 1291/11082 [4:09:02<31:31:48, 11.59s/step, acc=0.3888, loss=2.8203, lr=9.78e-05]
Training:  12%|█▏        | 1292/11082 [4:09:14<31:21:02, 11.53s/step, acc=0.3888, loss=2.8203, lr=9.78e-05]
Training:  12%|█▏        | 1293/11082 [4:09:25<31:36:30, 11.62s/step, acc=0.3888, loss=2.8203, lr=9.78e-05]
Training:  12%|█▏        | 1294/11082 [4:09:37<31:23:09, 11.54s/step, acc=0.3888, loss=2.8203, lr=9.78e-05]
Training:  12%|█▏        | 1295/11082 [4:09:48<31:23:48, 11.55s/step, acc=0.3888, loss=2.8203, lr=9.78e-05]
Training:  12%|█▏        | 1296/11082 [4:09:59<31:04:09, 11.43s/step, acc=0.3888, loss=2.8203, lr=9.78e-05]
Training:  12%|█▏        | 1297/11082 [4:10:11<31:22:31, 11.54s/step, acc=0.3888, loss=2.8203, lr=9.78e-05]
Training:  12%|█▏        | 1298/11082 [4:10:23<31:30:10, 11.59s/step, acc=0.3888, loss=2.8203, lr=9.78e-05]
Training:  12%|█▏        | 1299/11082 [4:10:34<31:22:09, 11.54s/step, acc=0.3888, loss=2.8203, lr=9.78e-05]
Training:  12%|█▏        | 1300/11082 [4:10:46<31:23:24, 11.55s/step, acc=0.3888, loss=2.8203, lr=9.78e-05]
Training:  12%|█▏        | 1300/11082 [4:10:53<31:23:24, 11.55s/step, acc=0.3920, loss=2.7949, lr=9.76e-05]2026-06-18 19:04:21 | INFO | __main__ | epoch=0 step=1300 train_loss=2.794868 train_acc=0.392013 lr=9.762e-05 grad_norm=1.3984 acc_d0=0.6625 acc_d1=0.4959 acc_d2=0.3636 acc_d3=0.2639 acc_d4=0.2016 acc_d5=0.1596 acc_d6=0.1311 acc_d7=0.1137

Training:  12%|█▏        | 1301/11082 [4:10:58<31:33:17, 11.61s/step, acc=0.3920, loss=2.7949, lr=9.76e-05]
Training:  12%|█▏        | 1302/11082 [4:11:09<31:29:24, 11.59s/step, acc=0.3920, loss=2.7949, lr=9.76e-05]
Training:  12%|█▏        | 1303/11082 [4:11:20<31:05:48, 11.45s/step, acc=0.3920, loss=2.7949, lr=9.76e-05]
Training:  12%|█▏        | 1304/11082 [4:11:32<31:00:30, 11.42s/step, acc=0.3920, loss=2.7949, lr=9.76e-05]
Training:  12%|█▏        | 1305/11082 [4:11:43<30:57:32, 11.40s/step, acc=0.3920, loss=2.7949, lr=9.76e-05]
Training:  12%|█▏        | 1306/11082 [4:11:55<31:39:33, 11.66s/step, acc=0.3920, loss=2.7949, lr=9.76e-05]
Training:  12%|█▏        | 1307/11082 [4:12:06<31:06:09, 11.45s/step, acc=0.3920, loss=2.7949, lr=9.76e-05]
Training:  12%|█▏        | 1308/11082 [4:12:18<31:24:59, 11.57s/step, acc=0.3920, loss=2.7949, lr=9.76e-05]
Training:  12%|█▏        | 1309/11082 [4:12:30<31:33:45, 11.63s/step, acc=0.3920, loss=2.7949, lr=9.76e-05]
Training:  12%|█▏        | 1310/11082 [4:12:41<31:26:40, 11.58s/step, acc=0.3920, loss=2.7949, lr=9.76e-05]
Training:  12%|█▏        | 1311/11082 [4:12:53<31:42:55, 11.69s/step, acc=0.3920, loss=2.7949, lr=9.76e-05]
Training:  12%|█▏        | 1312/11082 [4:13:04<31:05:47, 11.46s/step, acc=0.3920, loss=2.7949, lr=9.76e-05]
Training:  12%|█▏        | 1313/11082 [4:13:16<31:13:23, 11.51s/step, acc=0.3920, loss=2.7949, lr=9.76e-05]
Training:  12%|█▏        | 1314/11082 [4:13:28<31:25:24, 11.58s/step, acc=0.3920, loss=2.7949, lr=9.76e-05]
Training:  12%|█▏        | 1315/11082 [4:13:39<31:20:40, 11.55s/step, acc=0.3920, loss=2.7949, lr=9.76e-05]
Training:  12%|█▏        | 1316/11082 [4:13:51<31:19:41, 11.55s/step, acc=0.3920, loss=2.7949, lr=9.76e-05]
Training:  12%|█▏        | 1317/11082 [4:14:02<31:14:00, 11.51s/step, acc=0.3920, loss=2.7949, lr=9.76e-05]
Training:  12%|█▏        | 1318/11082 [4:14:14<31:32:27, 11.63s/step, acc=0.3920, loss=2.7949, lr=9.76e-05]
Training:  12%|█▏        | 1319/11082 [4:14:26<31:32:05, 11.63s/step, acc=0.3920, loss=2.7949, lr=9.76e-05]
Training:  12%|█▏        | 1320/11082 [4:14:37<31:14:04, 11.52s/step, acc=0.3920, loss=2.7949, lr=9.76e-05]
Training:  12%|█▏        | 1321/11082 [4:14:48<31:10:23, 11.50s/step, acc=0.3920, loss=2.7949, lr=9.76e-05]
Training:  12%|█▏        | 1322/11082 [4:15:00<31:33:57, 11.64s/step, acc=0.3920, loss=2.7949, lr=9.76e-05]
Training:  12%|█▏        | 1323/11082 [4:15:12<31:23:44, 11.58s/step, acc=0.3920, loss=2.7949, lr=9.76e-05]
Training:  12%|█▏        | 1324/11082 [4:15:23<31:02:36, 11.45s/step, acc=0.3920, loss=2.7949, lr=9.76e-05]
Training:  12%|█▏        | 1325/11082 [4:15:35<31:09:32, 11.50s/step, acc=0.3920, loss=2.7949, lr=9.76e-05]
Training:  12%|█▏        | 1326/11082 [4:15:46<31:25:05, 11.59s/step, acc=0.3920, loss=2.7949, lr=9.76e-05]
Training:  12%|█▏        | 1327/11082 [4:15:58<31:17:22, 11.55s/step, acc=0.3920, loss=2.7949, lr=9.76e-05]
Training:  12%|█▏        | 1328/11082 [4:16:10<31:29:17, 11.62s/step, acc=0.3920, loss=2.7949, lr=9.76e-05]
Training:  12%|█▏        | 1329/11082 [4:16:21<31:06:51, 11.48s/step, acc=0.3920, loss=2.7949, lr=9.76e-05]
Training:  12%|█▏        | 1330/11082 [4:16:32<30:58:02, 11.43s/step, acc=0.3920, loss=2.7949, lr=9.76e-05]
Training:  12%|█▏        | 1331/11082 [4:16:44<31:25:45, 11.60s/step, acc=0.3920, loss=2.7949, lr=9.76e-05]
Training:  12%|█▏        | 1332/11082 [4:16:55<31:16:53, 11.55s/step, acc=0.3920, loss=2.7949, lr=9.76e-05]
Training:  12%|█▏        | 1333/11082 [4:17:07<31:19:28, 11.57s/step, acc=0.3920, loss=2.7949, lr=9.76e-05]
Training:  12%|█▏        | 1334/11082 [4:17:19<31:29:40, 11.63s/step, acc=0.3920, loss=2.7949, lr=9.76e-05]
Training:  12%|█▏        | 1335/11082 [4:17:30<31:28:45, 11.63s/step, acc=0.3920, loss=2.7949, lr=9.76e-05]
Training:  12%|█▏        | 1336/11082 [4:17:42<31:31:45, 11.65s/step, acc=0.3920, loss=2.7949, lr=9.76e-05]
Training:  12%|█▏        | 1337/11082 [4:17:53<30:53:09, 11.41s/step, acc=0.3920, loss=2.7949, lr=9.76e-05]
Training:  12%|█▏        | 1338/11082 [4:18:05<31:00:29, 11.46s/step, acc=0.3920, loss=2.7949, lr=9.76e-05]
Training:  12%|█▏        | 1339/11082 [4:18:16<31:20:22, 11.58s/step, acc=0.3920, loss=2.7949, lr=9.76e-05]
Training:  12%|█▏        | 1340/11082 [4:18:28<31:13:01, 11.54s/step, acc=0.3920, loss=2.7949, lr=9.76e-05]
Training:  12%|█▏        | 1341/11082 [4:18:40<31:16:54, 11.56s/step, acc=0.3920, loss=2.7949, lr=9.76e-05]
Training:  12%|█▏        | 1342/11082 [4:18:51<31:24:18, 11.61s/step, acc=0.3920, loss=2.7949, lr=9.76e-05]
Training:  12%|█▏        | 1343/11082 [4:19:02<31:03:08, 11.48s/step, acc=0.3920, loss=2.7949, lr=9.76e-05]
Training:  12%|█▏        | 1344/11082 [4:19:14<31:28:19, 11.63s/step, acc=0.3920, loss=2.7949, lr=9.76e-05]
Training:  12%|█▏        | 1345/11082 [4:19:26<31:25:09, 11.62s/step, acc=0.3920, loss=2.7949, lr=9.76e-05]
Training:  12%|█▏        | 1346/11082 [4:19:38<31:46:20, 11.75s/step, acc=0.3920, loss=2.7949, lr=9.76e-05]
Training:  12%|█▏        | 1347/11082 [4:19:49<31:22:20, 11.60s/step, acc=0.3920, loss=2.7949, lr=9.76e-05]
Training:  12%|█▏        | 1348/11082 [4:20:01<31:36:04, 11.69s/step, acc=0.3920, loss=2.7949, lr=9.76e-05]
Training:  12%|█▏        | 1349/11082 [4:20:13<31:30:54, 11.66s/step, acc=0.3920, loss=2.7949, lr=9.76e-05]
Training:  12%|█▏        | 1350/11082 [4:20:24<31:30:58, 11.66s/step, acc=0.3920, loss=2.7949, lr=9.76e-05]
Training:  12%|█▏        | 1350/11082 [4:20:32<31:30:58, 11.66s/step, acc=0.3971, loss=2.7609, lr=9.74e-05]2026-06-18 19:14:00 | INFO | __main__ | epoch=0 step=1350 train_loss=2.760950 train_acc=0.397068 lr=9.741e-05 grad_norm=1.2969 acc_d0=0.6730 acc_d1=0.5111 acc_d2=0.3783 acc_d3=0.2807 acc_d4=0.2138 acc_d5=0.1707 acc_d6=0.1420 acc_d7=0.1192

Training:  12%|█▏        | 1351/11082 [4:20:36<31:38:05, 11.70s/step, acc=0.3971, loss=2.7609, lr=9.74e-05]
Training:  12%|█▏        | 1352/11082 [4:20:48<31:20:38, 11.60s/step, acc=0.3971, loss=2.7609, lr=9.74e-05]
Training:  12%|█▏        | 1353/11082 [4:20:59<31:34:08, 11.68s/step, acc=0.3971, loss=2.7609, lr=9.74e-05]
Training:  12%|█▏        | 1354/11082 [4:21:11<31:36:42, 11.70s/step, acc=0.3971, loss=2.7609, lr=9.74e-05]
Training:  12%|█▏        | 1355/11082 [4:21:23<31:32:46, 11.68s/step, acc=0.3971, loss=2.7609, lr=9.74e-05]
Training:  12%|█▏        | 1356/11082 [4:21:34<31:09:28, 11.53s/step, acc=0.3971, loss=2.7609, lr=9.74e-05]
Training:  12%|█▏        | 1357/11082 [4:21:46<31:07:09, 11.52s/step, acc=0.3971, loss=2.7609, lr=9.74e-05]
Training:  12%|█▏        | 1358/11082 [4:21:58<31:32:52, 11.68s/step, acc=0.3971, loss=2.7609, lr=9.74e-05]
Training:  12%|█▏        | 1359/11082 [4:22:09<31:11:04, 11.55s/step, acc=0.3971, loss=2.7609, lr=9.74e-05]
Training:  12%|█▏        | 1360/11082 [4:22:20<31:05:51, 11.52s/step, acc=0.3971, loss=2.7609, lr=9.74e-05]
Training:  12%|█▏        | 1361/11082 [4:22:32<31:04:03, 11.51s/step, acc=0.3971, loss=2.7609, lr=9.74e-05]
Training:  12%|█▏        | 1362/11082 [4:22:44<31:22:07, 11.62s/step, acc=0.3971, loss=2.7609, lr=9.74e-05]
Training:  12%|█▏        | 1363/11082 [4:22:55<30:57:45, 11.47s/step, acc=0.3971, loss=2.7609, lr=9.74e-05]
Training:  12%|█▏        | 1364/11082 [4:23:06<30:45:59, 11.40s/step, acc=0.3971, loss=2.7609, lr=9.74e-05]
Training:  12%|█▏        | 1365/11082 [4:23:18<31:17:08, 11.59s/step, acc=0.3971, loss=2.7609, lr=9.74e-05]
Training:  12%|█▏        | 1366/11082 [4:23:29<31:07:14, 11.53s/step, acc=0.3971, loss=2.7609, lr=9.74e-05]
Training:  12%|█▏        | 1367/11082 [4:23:41<31:22:32, 11.63s/step, acc=0.3971, loss=2.7609, lr=9.74e-05]
Training:  12%|█▏        | 1368/11082 [4:23:53<31:15:18, 11.58s/step, acc=0.3971, loss=2.7609, lr=9.74e-05]
Training:  12%|█▏        | 1369/11082 [4:24:04<31:02:58, 11.51s/step, acc=0.3971, loss=2.7609, lr=9.74e-05]
Training:  12%|█▏        | 1370/11082 [4:24:16<31:14:58, 11.58s/step, acc=0.3971, loss=2.7609, lr=9.74e-05]
Training:  12%|█▏        | 1371/11082 [4:24:27<31:08:59, 11.55s/step, acc=0.3971, loss=2.7609, lr=9.74e-05]
Training:  12%|█▏        | 1372/11082 [4:24:39<31:05:47, 11.53s/step, acc=0.3971, loss=2.7609, lr=9.74e-05]
Training:  12%|█▏        | 1373/11082 [4:24:50<31:13:16, 11.58s/step, acc=0.3971, loss=2.7609, lr=9.74e-05]
Training:  12%|█▏        | 1374/11082 [4:25:02<30:59:46, 11.49s/step, acc=0.3971, loss=2.7609, lr=9.74e-05]
Training:  12%|█▏        | 1375/11082 [4:25:13<31:10:29, 11.56s/step, acc=0.3971, loss=2.7609, lr=9.74e-05]
Training:  12%|█▏        | 1376/11082 [4:25:25<31:08:46, 11.55s/step, acc=0.3971, loss=2.7609, lr=9.74e-05]
Training:  12%|█▏        | 1377/11082 [4:25:37<31:06:35, 11.54s/step, acc=0.3971, loss=2.7609, lr=9.74e-05]
Training:  12%|█▏        | 1378/11082 [4:25:48<31:09:48, 11.56s/step, acc=0.3971, loss=2.7609, lr=9.74e-05]
Training:  12%|█▏        | 1379/11082 [4:26:00<31:10:15, 11.57s/step, acc=0.3971, loss=2.7609, lr=9.74e-05]
Training:  12%|█▏        | 1380/11082 [4:26:11<30:49:13, 11.44s/step, acc=0.3971, loss=2.7609, lr=9.74e-05]
Training:  12%|█▏        | 1381/11082 [4:26:23<31:04:01, 11.53s/step, acc=0.3971, loss=2.7609, lr=9.74e-05]
Training:  12%|█▏        | 1382/11082 [4:26:34<31:08:02, 11.55s/step, acc=0.3971, loss=2.7609, lr=9.74e-05]
Training:  12%|█▏        | 1383/11082 [4:26:46<31:09:31, 11.57s/step, acc=0.3971, loss=2.7609, lr=9.74e-05]
Training:  12%|█▏        | 1384/11082 [4:26:58<31:18:43, 11.62s/step, acc=0.3971, loss=2.7609, lr=9.74e-05]
Training:  12%|█▏        | 1385/11082 [4:27:09<31:07:06, 11.55s/step, acc=0.3971, loss=2.7609, lr=9.74e-05]
Training:  13%|█▎        | 1386/11082 [4:27:20<30:58:13, 11.50s/step, acc=0.3971, loss=2.7609, lr=9.74e-05]
Training:  13%|█▎        | 1387/11082 [4:27:32<31:15:26, 11.61s/step, acc=0.3971, loss=2.7609, lr=9.74e-05]
Training:  13%|█▎        | 1388/11082 [4:27:44<31:08:54, 11.57s/step, acc=0.3971, loss=2.7609, lr=9.74e-05]
Training:  13%|█▎        | 1389/11082 [4:27:55<31:21:35, 11.65s/step, acc=0.3971, loss=2.7609, lr=9.74e-05]
Training:  13%|█▎        | 1390/11082 [4:28:07<31:10:45, 11.58s/step, acc=0.3971, loss=2.7609, lr=9.74e-05]
Training:  13%|█▎        | 1391/11082 [4:28:19<31:14:02, 11.60s/step, acc=0.3971, loss=2.7609, lr=9.74e-05]
Training:  13%|█▎        | 1392/11082 [4:28:30<31:15:29, 11.61s/step, acc=0.3971, loss=2.7609, lr=9.74e-05]
Training:  13%|█▎        | 1393/11082 [4:28:42<31:31:03, 11.71s/step, acc=0.3971, loss=2.7609, lr=9.74e-05]
Training:  13%|█▎        | 1394/11082 [4:28:54<31:32:03, 11.72s/step, acc=0.3971, loss=2.7609, lr=9.74e-05]
Training:  13%|█▎        | 1395/11082 [4:29:05<31:17:41, 11.63s/step, acc=0.3971, loss=2.7609, lr=9.74e-05]
Training:  13%|█▎        | 1396/11082 [4:29:17<30:57:27, 11.51s/step, acc=0.3971, loss=2.7609, lr=9.74e-05]
Training:  13%|█▎        | 1397/11082 [4:29:28<30:55:28, 11.49s/step, acc=0.3971, loss=2.7609, lr=9.74e-05]
Training:  13%|█▎        | 1398/11082 [4:29:40<30:58:20, 11.51s/step, acc=0.3971, loss=2.7609, lr=9.74e-05]
Training:  13%|█▎        | 1399/11082 [4:29:52<31:28:28, 11.70s/step, acc=0.3971, loss=2.7609, lr=9.74e-05]
Training:  13%|█▎        | 1400/11082 [4:30:03<31:15:18, 11.62s/step, acc=0.3971, loss=2.7609, lr=9.74e-05]
Training:  13%|█▎        | 1400/11082 [4:30:10<31:15:18, 11.62s/step, acc=0.3983, loss=2.7508, lr=9.72e-05]2026-06-18 19:23:38 | INFO | __main__ | epoch=0 step=1400 train_loss=2.750768 train_acc=0.398296 lr=9.719e-05 grad_norm=1.4844 acc_d0=0.6639 acc_d1=0.4985 acc_d2=0.3645 acc_d3=0.2698 acc_d4=0.2044 acc_d5=0.1640 acc_d6=0.1362 acc_d7=0.1161

Training:  13%|█▎        | 1401/11082 [4:30:15<31:07:46, 11.58s/step, acc=0.3983, loss=2.7508, lr=9.72e-05]
Training:  13%|█▎        | 1402/11082 [4:30:26<31:09:27, 11.59s/step, acc=0.3983, loss=2.7508, lr=9.72e-05]
Training:  13%|█▎        | 1403/11082 [4:30:38<31:07:48, 11.58s/step, acc=0.3983, loss=2.7508, lr=9.72e-05]
Training:  13%|█▎        | 1404/11082 [4:30:49<30:52:45, 11.49s/step, acc=0.3983, loss=2.7508, lr=9.72e-05]
Training:  13%|█▎        | 1405/11082 [4:31:01<31:16:53, 11.64s/step, acc=0.3983, loss=2.7508, lr=9.72e-05]
Training:  13%|█▎        | 1406/11082 [4:31:12<30:39:10, 11.40s/step, acc=0.3983, loss=2.7508, lr=9.72e-05]
Training:  13%|█▎        | 1407/11082 [4:31:24<31:06:08, 11.57s/step, acc=0.3983, loss=2.7508, lr=9.72e-05]
Training:  13%|█▎        | 1408/11082 [4:31:35<31:07:49, 11.58s/step, acc=0.3983, loss=2.7508, lr=9.72e-05]
Training:  13%|█▎        | 1409/11082 [4:31:47<30:55:00, 11.51s/step, acc=0.3983, loss=2.7508, lr=9.72e-05]
Training:  13%|█▎        | 1410/11082 [4:31:58<31:01:04, 11.55s/step, acc=0.3983, loss=2.7508, lr=9.72e-05]
Training:  13%|█▎        | 1411/11082 [4:32:10<31:00:09, 11.54s/step, acc=0.3983, loss=2.7508, lr=9.72e-05]
Training:  13%|█▎        | 1412/11082 [4:32:21<30:55:49, 11.51s/step, acc=0.3983, loss=2.7508, lr=9.72e-05]
Training:  13%|█▎        | 1413/11082 [4:32:33<30:42:07, 11.43s/step, acc=0.3983, loss=2.7508, lr=9.72e-05]
Training:  13%|█▎        | 1414/11082 [4:32:44<30:44:45, 11.45s/step, acc=0.3983, loss=2.7508, lr=9.72e-05]
Training:  13%|█▎        | 1415/11082 [4:32:56<30:59:57, 11.54s/step, acc=0.3983, loss=2.7508, lr=9.72e-05]
Training:  13%|█▎        | 1416/11082 [4:33:07<30:49:51, 11.48s/step, acc=0.3983, loss=2.7508, lr=9.72e-05]
Training:  13%|█▎        | 1417/11082 [4:33:19<30:52:06, 11.50s/step, acc=0.3983, loss=2.7508, lr=9.72e-05]
Training:  13%|█▎        | 1418/11082 [4:33:30<30:42:12, 11.44s/step, acc=0.3983, loss=2.7508, lr=9.72e-05]
Training:  13%|█▎        | 1419/11082 [4:33:42<31:15:09, 11.64s/step, acc=0.3983, loss=2.7508, lr=9.72e-05]
Training:  13%|█▎        | 1420/11082 [4:33:54<31:03:53, 11.57s/step, acc=0.3983, loss=2.7508, lr=9.72e-05]
Training:  13%|█▎        | 1421/11082 [4:34:05<31:12:15, 11.63s/step, acc=0.3983, loss=2.7508, lr=9.72e-05]
Training:  13%|█▎        | 1422/11082 [4:34:17<30:52:04, 11.50s/step, acc=0.3983, loss=2.7508, lr=9.72e-05]
Training:  13%|█▎        | 1423/11082 [4:34:28<31:06:18, 11.59s/step, acc=0.3983, loss=2.7508, lr=9.72e-05]
Training:  13%|█▎        | 1424/11082 [4:34:39<30:37:59, 11.42s/step, acc=0.3983, loss=2.7508, lr=9.72e-05]
Training:  13%|█▎        | 1425/11082 [4:34:51<31:03:22, 11.58s/step, acc=0.3983, loss=2.7508, lr=9.72e-05]
Training:  13%|█▎        | 1426/11082 [4:35:02<30:43:18, 11.45s/step, acc=0.3983, loss=2.7508, lr=9.72e-05]
Training:  13%|█▎        | 1427/11082 [4:35:14<31:06:46, 11.60s/step, acc=0.3983, loss=2.7508, lr=9.72e-05]
Training:  13%|█▎        | 1428/11082 [4:35:25<30:39:44, 11.43s/step, acc=0.3983, loss=2.7508, lr=9.72e-05]
Training:  13%|█▎        | 1429/11082 [4:35:37<30:55:59, 11.54s/step, acc=0.3983, loss=2.7508, lr=9.72e-05]
Training:  13%|█▎        | 1430/11082 [4:35:50<31:34:05, 11.77s/step, acc=0.3983, loss=2.7508, lr=9.72e-05]
Training:  13%|█▎        | 1431/11082 [4:36:01<31:13:56, 11.65s/step, acc=0.3983, loss=2.7508, lr=9.72e-05]
Training:  13%|█▎        | 1432/11082 [4:36:13<31:17:42, 11.67s/step, acc=0.3983, loss=2.7508, lr=9.72e-05]
Training:  13%|█▎        | 1433/11082 [4:36:24<30:57:42, 11.55s/step, acc=0.3983, loss=2.7508, lr=9.72e-05]
Training:  13%|█▎        | 1434/11082 [4:36:36<31:05:36, 11.60s/step, acc=0.3983, loss=2.7508, lr=9.72e-05]
Training:  13%|█▎        | 1435/11082 [4:36:48<31:28:52, 11.75s/step, acc=0.3983, loss=2.7508, lr=9.72e-05]
Training:  13%|█▎        | 1436/11082 [4:36:59<31:00:18, 11.57s/step, acc=0.3983, loss=2.7508, lr=9.72e-05]
Training:  13%|█▎        | 1437/11082 [4:37:11<31:33:29, 11.78s/step, acc=0.3983, loss=2.7508, lr=9.72e-05]
Training:  13%|█▎        | 1438/11082 [4:37:22<31:07:13, 11.62s/step, acc=0.3983, loss=2.7508, lr=9.72e-05]
Training:  13%|█▎        | 1439/11082 [4:37:34<30:58:36, 11.56s/step, acc=0.3983, loss=2.7508, lr=9.72e-05]
Training:  13%|█▎        | 1440/11082 [4:37:45<30:49:29, 11.51s/step, acc=0.3983, loss=2.7508, lr=9.72e-05]
Training:  13%|█▎        | 1441/11082 [4:37:57<31:09:07, 11.63s/step, acc=0.3983, loss=2.7508, lr=9.72e-05]
Training:  13%|█▎        | 1442/11082 [4:38:08<30:44:32, 11.48s/step, acc=0.3983, loss=2.7508, lr=9.72e-05]
Training:  13%|█▎        | 1443/11082 [4:38:20<30:37:55, 11.44s/step, acc=0.3983, loss=2.7508, lr=9.72e-05]
Training:  13%|█▎        | 1444/11082 [4:38:31<30:46:41, 11.50s/step, acc=0.3983, loss=2.7508, lr=9.72e-05]
Training:  13%|█▎        | 1445/11082 [4:38:43<30:46:48, 11.50s/step, acc=0.3983, loss=2.7508, lr=9.72e-05]
Training:  13%|█▎        | 1446/11082 [4:38:54<30:27:53, 11.38s/step, acc=0.3983, loss=2.7508, lr=9.72e-05]
Training:  13%|█▎        | 1447/11082 [4:39:06<31:14:13, 11.67s/step, acc=0.3983, loss=2.7508, lr=9.72e-05]
Training:  13%|█▎        | 1448/11082 [4:39:18<31:10:15, 11.65s/step, acc=0.3983, loss=2.7508, lr=9.72e-05]
Training:  13%|█▎        | 1449/11082 [4:39:29<31:06:51, 11.63s/step, acc=0.3983, loss=2.7508, lr=9.72e-05]
Training:  13%|█▎        | 1450/11082 [4:39:41<30:42:03, 11.47s/step, acc=0.3983, loss=2.7508, lr=9.72e-05]
Training:  13%|█▎        | 1450/11082 [4:39:48<30:42:03, 11.47s/step, acc=0.4004, loss=2.7245, lr=9.70e-05]2026-06-18 19:33:16 | INFO | __main__ | epoch=0 step=1450 train_loss=2.724548 train_acc=0.400395 lr=9.696e-05 grad_norm=1.3672 acc_d0=0.6581 acc_d1=0.4924 acc_d2=0.3626 acc_d3=0.2677 acc_d4=0.2041 acc_d5=0.1636 acc_d6=0.1367 acc_d7=0.1167

Training:  13%|█▎        | 1451/11082 [4:39:52<31:02:03, 11.60s/step, acc=0.4004, loss=2.7245, lr=9.70e-05]
Training:  13%|█▎        | 1452/11082 [4:40:03<30:33:08, 11.42s/step, acc=0.4004, loss=2.7245, lr=9.70e-05]
Training:  13%|█▎        | 1453/11082 [4:40:15<31:01:55, 11.60s/step, acc=0.4004, loss=2.7245, lr=9.70e-05]
Training:  13%|█▎        | 1454/11082 [4:40:27<30:58:08, 11.58s/step, acc=0.4004, loss=2.7245, lr=9.70e-05]
Training:  13%|█▎        | 1455/11082 [4:40:38<30:35:04, 11.44s/step, acc=0.4004, loss=2.7245, lr=9.70e-05]
Training:  13%|█▎        | 1456/11082 [4:40:50<31:04:45, 11.62s/step, acc=0.4004, loss=2.7245, lr=9.70e-05]
Training:  13%|█▎        | 1457/11082 [4:41:01<30:46:28, 11.51s/step, acc=0.4004, loss=2.7245, lr=9.70e-05]
Training:  13%|█▎        | 1458/11082 [4:41:13<31:03:08, 11.62s/step, acc=0.4004, loss=2.7245, lr=9.70e-05]
Training:  13%|█▎        | 1459/11082 [4:41:25<30:53:58, 11.56s/step, acc=0.4004, loss=2.7245, lr=9.70e-05]
Training:  13%|█▎        | 1460/11082 [4:41:36<30:55:05, 11.57s/step, acc=0.4004, loss=2.7245, lr=9.70e-05]
Training:  13%|█▎        | 1461/11082 [4:41:48<30:51:59, 11.55s/step, acc=0.4004, loss=2.7245, lr=9.70e-05]
Training:  13%|█▎        | 1462/11082 [4:42:00<31:09:24, 11.66s/step, acc=0.4004, loss=2.7245, lr=9.70e-05]
Training:  13%|█▎        | 1463/11082 [4:42:11<30:38:24, 11.47s/step, acc=0.4004, loss=2.7245, lr=9.70e-05]
Training:  13%|█▎        | 1464/11082 [4:42:23<30:56:00, 11.58s/step, acc=0.4004, loss=2.7245, lr=9.70e-05]
Training:  13%|█▎        | 1465/11082 [4:42:34<30:57:35, 11.59s/step, acc=0.4004, loss=2.7245, lr=9.70e-05]
Training:  13%|█▎        | 1466/11082 [4:42:46<30:58:28, 11.60s/step, acc=0.4004, loss=2.7245, lr=9.70e-05]
Training:  13%|█▎        | 1467/11082 [4:42:57<30:48:40, 11.54s/step, acc=0.4004, loss=2.7245, lr=9.70e-05]
Training:  13%|█▎        | 1468/11082 [4:43:08<30:35:54, 11.46s/step, acc=0.4004, loss=2.7245, lr=9.70e-05]
Training:  13%|█▎        | 1469/11082 [4:43:20<30:43:08, 11.50s/step, acc=0.4004, loss=2.7245, lr=9.70e-05]
Training:  13%|█▎        | 1470/11082 [4:43:32<30:58:54, 11.60s/step, acc=0.4004, loss=2.7245, lr=9.70e-05]
Training:  13%|█▎        | 1471/11082 [4:43:43<30:57:51, 11.60s/step, acc=0.4004, loss=2.7245, lr=9.70e-05]
Training:  13%|█▎        | 1472/11082 [4:43:55<30:46:19, 11.53s/step, acc=0.4004, loss=2.7245, lr=9.70e-05]
Training:  13%|█▎        | 1473/11082 [4:44:07<30:58:26, 11.60s/step, acc=0.4004, loss=2.7245, lr=9.70e-05]
Training:  13%|█▎        | 1474/11082 [4:44:18<30:42:39, 11.51s/step, acc=0.4004, loss=2.7245, lr=9.70e-05]
Training:  13%|█▎        | 1475/11082 [4:44:29<30:25:34, 11.40s/step, acc=0.4004, loss=2.7245, lr=9.70e-05]
Training:  13%|█▎        | 1476/11082 [4:44:41<30:42:51, 11.51s/step, acc=0.4004, loss=2.7245, lr=9.70e-05]
Training:  13%|█▎        | 1477/11082 [4:44:52<30:50:47, 11.56s/step, acc=0.4004, loss=2.7245, lr=9.70e-05]
Training:  13%|█▎        | 1478/11082 [4:45:04<30:54:47, 11.59s/step, acc=0.4004, loss=2.7245, lr=9.70e-05]
Training:  13%|█▎        | 1479/11082 [4:45:15<30:32:50, 11.45s/step, acc=0.4004, loss=2.7245, lr=9.70e-05]
Training:  13%|█▎        | 1480/11082 [4:45:27<30:51:29, 11.57s/step, acc=0.4004, loss=2.7245, lr=9.70e-05]
Training:  13%|█▎        | 1481/11082 [4:45:38<30:33:45, 11.46s/step, acc=0.4004, loss=2.7245, lr=9.70e-05]
Training:  13%|█▎        | 1482/11082 [4:45:50<30:39:43, 11.50s/step, acc=0.4004, loss=2.7245, lr=9.70e-05]
Training:  13%|█▎        | 1483/11082 [4:46:02<30:51:41, 11.57s/step, acc=0.4004, loss=2.7245, lr=9.70e-05]
Training:  13%|█▎        | 1484/11082 [4:46:13<30:50:01, 11.57s/step, acc=0.4004, loss=2.7245, lr=9.70e-05]
Training:  13%|█▎        | 1485/11082 [4:46:25<30:39:27, 11.50s/step, acc=0.4004, loss=2.7245, lr=9.70e-05]
Training:  13%|█▎        | 1486/11082 [4:46:36<30:50:59, 11.57s/step, acc=0.4004, loss=2.7245, lr=9.70e-05]
Training:  13%|█▎        | 1487/11082 [4:46:48<30:58:50, 11.62s/step, acc=0.4004, loss=2.7245, lr=9.70e-05]
Training:  13%|█▎        | 1488/11082 [4:46:59<30:25:23, 11.42s/step, acc=0.4004, loss=2.7245, lr=9.70e-05]
Training:  13%|█▎        | 1489/11082 [4:47:11<30:56:29, 11.61s/step, acc=0.4004, loss=2.7245, lr=9.70e-05]
Training:  13%|█▎        | 1490/11082 [4:47:23<31:12:27, 11.71s/step, acc=0.4004, loss=2.7245, lr=9.70e-05]
Training:  13%|█▎        | 1491/11082 [4:47:34<31:02:36, 11.65s/step, acc=0.4004, loss=2.7245, lr=9.70e-05]
Training:  13%|█▎        | 1492/11082 [4:47:46<30:57:00, 11.62s/step, acc=0.4004, loss=2.7245, lr=9.70e-05]
Training:  13%|█▎        | 1493/11082 [4:47:58<30:58:07, 11.63s/step, acc=0.4004, loss=2.7245, lr=9.70e-05]
Training:  13%|█▎        | 1494/11082 [4:48:09<30:37:01, 11.50s/step, acc=0.4004, loss=2.7245, lr=9.70e-05]
Training:  13%|█▎        | 1495/11082 [4:48:21<30:45:30, 11.55s/step, acc=0.4004, loss=2.7245, lr=9.70e-05]
Training:  13%|█▎        | 1496/11082 [4:48:32<30:21:44, 11.40s/step, acc=0.4004, loss=2.7245, lr=9.70e-05]
Training:  14%|█▎        | 1497/11082 [4:48:44<30:56:31, 11.62s/step, acc=0.4004, loss=2.7245, lr=9.70e-05]
Training:  14%|█▎        | 1498/11082 [4:48:55<30:45:41, 11.55s/step, acc=0.4004, loss=2.7245, lr=9.70e-05]
Training:  14%|█▎        | 1499/11082 [4:49:07<30:52:34, 11.60s/step, acc=0.4004, loss=2.7245, lr=9.70e-05]
Training:  14%|█▎        | 1500/11082 [4:49:18<30:49:15, 11.58s/step, acc=0.4004, loss=2.7245, lr=9.70e-05]
Training:  14%|█▎        | 1500/11082 [4:49:25<30:49:15, 11.58s/step, acc=0.4051, loss=2.6986, lr=9.67e-05]2026-06-18 19:42:53 | INFO | __main__ | epoch=0 step=1500 train_loss=2.698628 train_acc=0.405066 lr=9.672e-05 grad_norm=1.4141 acc_d0=0.6759 acc_d1=0.5101 acc_d2=0.3770 acc_d3=0.2789 acc_d4=0.2119 acc_d5=0.1705 acc_d6=0.1414 acc_d7=0.1202

Training:  14%|█▎        | 1501/11082 [4:49:29<30:27:15, 11.44s/step, acc=0.4051, loss=2.6986, lr=9.67e-05]
Training:  14%|█▎        | 1502/11082 [4:49:41<30:39:39, 11.52s/step, acc=0.4051, loss=2.6986, lr=9.67e-05]
Training:  14%|█▎        | 1503/11082 [4:49:52<30:22:16, 11.41s/step, acc=0.4051, loss=2.6986, lr=9.67e-05]
Training:  14%|█▎        | 1504/11082 [4:50:05<31:10:36, 11.72s/step, acc=0.4051, loss=2.6986, lr=9.67e-05]
Training:  14%|█▎        | 1505/11082 [4:50:16<30:44:39, 11.56s/step, acc=0.4051, loss=2.6986, lr=9.67e-05]
Training:  14%|█▎        | 1506/11082 [4:50:28<31:01:48, 11.67s/step, acc=0.4051, loss=2.6986, lr=9.67e-05]
Training:  14%|█▎        | 1507/11082 [4:50:39<30:34:26, 11.50s/step, acc=0.4051, loss=2.6986, lr=9.67e-05]
Training:  14%|█▎        | 1508/11082 [4:50:52<31:24:58, 11.81s/step, acc=0.4051, loss=2.6986, lr=9.67e-05]
Training:  14%|█▎        | 1509/11082 [4:51:03<31:18:00, 11.77s/step, acc=0.4051, loss=2.6986, lr=9.67e-05]
Training:  14%|█▎        | 1510/11082 [4:51:15<31:07:54, 11.71s/step, acc=0.4051, loss=2.6986, lr=9.67e-05]
Training:  14%|█▎        | 1511/11082 [4:51:27<31:23:58, 11.81s/step, acc=0.4051, loss=2.6986, lr=9.67e-05]
Training:  14%|█▎        | 1512/11082 [4:51:38<31:01:15, 11.67s/step, acc=0.4051, loss=2.6986, lr=9.67e-05]
Training:  14%|█▎        | 1513/11082 [4:51:50<30:51:57, 11.61s/step, acc=0.4051, loss=2.6986, lr=9.67e-05]
Training:  14%|█▎        | 1514/11082 [4:52:02<31:17:32, 11.77s/step, acc=0.4051, loss=2.6986, lr=9.67e-05]
Training:  14%|█▎        | 1515/11082 [4:52:13<31:02:15, 11.68s/step, acc=0.4051, loss=2.6986, lr=9.67e-05]
Training:  14%|█▎        | 1516/11082 [4:52:25<31:09:48, 11.73s/step, acc=0.4051, loss=2.6986, lr=9.67e-05]
Training:  14%|█▎        | 1517/11082 [4:52:37<31:07:19, 11.71s/step, acc=0.4051, loss=2.6986, lr=9.67e-05]
Training:  14%|█▎        | 1518/11082 [4:52:48<30:35:44, 11.52s/step, acc=0.4051, loss=2.6986, lr=9.67e-05]
Training:  14%|█▎        | 1519/11082 [4:52:59<30:36:31, 11.52s/step, acc=0.4051, loss=2.6986, lr=9.67e-05]
Training:  14%|█▎        | 1520/11082 [4:53:11<30:29:55, 11.48s/step, acc=0.4051, loss=2.6986, lr=9.67e-05]
Training:  14%|█▎        | 1521/11082 [4:53:22<30:41:18, 11.56s/step, acc=0.4051, loss=2.6986, lr=9.67e-05]
Training:  14%|█▎        | 1522/11082 [4:53:34<30:52:14, 11.62s/step, acc=0.4051, loss=2.6986, lr=9.67e-05]
Training:  14%|█▎        | 1523/11082 [4:53:45<30:26:31, 11.46s/step, acc=0.4051, loss=2.6986, lr=9.67e-05]
Training:  14%|█▍        | 1524/11082 [4:53:57<30:24:13, 11.45s/step, acc=0.4051, loss=2.6986, lr=9.67e-05]
Training:  14%|█▍        | 1525/11082 [4:54:08<30:23:28, 11.45s/step, acc=0.4051, loss=2.6986, lr=9.67e-05]
Training:  14%|█▍        | 1526/11082 [4:54:20<30:33:31, 11.51s/step, acc=0.4051, loss=2.6986, lr=9.67e-05]
Training:  14%|█▍        | 1527/11082 [4:54:32<30:57:29, 11.66s/step, acc=0.4051, loss=2.6986, lr=9.67e-05]
Training:  14%|█▍        | 1528/11082 [4:54:43<30:40:38, 11.56s/step, acc=0.4051, loss=2.6986, lr=9.67e-05]
Training:  14%|█▍        | 1529/11082 [4:54:55<30:45:42, 11.59s/step, acc=0.4051, loss=2.6986, lr=9.67e-05]
Training:  14%|█▍        | 1530/11082 [4:55:06<30:33:50, 11.52s/step, acc=0.4051, loss=2.6986, lr=9.67e-05]
Training:  14%|█▍        | 1531/11082 [4:55:18<30:35:09, 11.53s/step, acc=0.4051, loss=2.6986, lr=9.67e-05]
Training:  14%|█▍        | 1532/11082 [4:55:30<30:48:41, 11.61s/step, acc=0.4051, loss=2.6986, lr=9.67e-05]
Training:  14%|█▍        | 1533/11082 [4:55:41<30:35:48, 11.54s/step, acc=0.4051, loss=2.6986, lr=9.67e-05]
Training:  14%|█▍        | 1534/11082 [4:55:52<30:36:27, 11.54s/step, acc=0.4051, loss=2.6986, lr=9.67e-05]
Training:  14%|█▍        | 1535/11082 [4:56:04<30:25:52, 11.48s/step, acc=0.4051, loss=2.6986, lr=9.67e-05]
Training:  14%|█▍        | 1536/11082 [4:56:16<30:58:07, 11.68s/step, acc=0.4051, loss=2.6986, lr=9.67e-05]
Training:  14%|█▍        | 1537/11082 [4:56:27<30:30:58, 11.51s/step, acc=0.4051, loss=2.6986, lr=9.67e-05]
Training:  14%|█▍        | 1538/11082 [4:56:39<30:51:49, 11.64s/step, acc=0.4051, loss=2.6986, lr=9.67e-05]
Training:  14%|█▍        | 1539/11082 [4:56:50<30:39:58, 11.57s/step, acc=0.4051, loss=2.6986, lr=9.67e-05]
Training:  14%|█▍        | 1540/11082 [4:57:02<30:55:44, 11.67s/step, acc=0.4051, loss=2.6986, lr=9.67e-05]
Training:  14%|█▍        | 1541/11082 [4:57:13<30:28:59, 11.50s/step, acc=0.4051, loss=2.6986, lr=9.67e-05]
Training:  14%|█▍        | 1542/11082 [4:57:25<30:38:39, 11.56s/step, acc=0.4051, loss=2.6986, lr=9.67e-05]
Training:  14%|█▍        | 1543/11082 [4:57:37<30:49:53, 11.64s/step, acc=0.4051, loss=2.6986, lr=9.67e-05]
Training:  14%|█▍        | 1544/11082 [4:57:48<30:18:57, 11.44s/step, acc=0.4051, loss=2.6986, lr=9.67e-05]
Training:  14%|█▍        | 1545/11082 [4:57:59<30:23:39, 11.47s/step, acc=0.4051, loss=2.6986, lr=9.67e-05]
Training:  14%|█▍        | 1546/11082 [4:58:11<30:40:03, 11.58s/step, acc=0.4051, loss=2.6986, lr=9.67e-05]
Training:  14%|█▍        | 1547/11082 [4:58:23<30:40:05, 11.58s/step, acc=0.4051, loss=2.6986, lr=9.67e-05]
Training:  14%|█▍        | 1548/11082 [4:58:34<30:36:12, 11.56s/step, acc=0.4051, loss=2.6986, lr=9.67e-05]
Training:  14%|█▍        | 1549/11082 [4:58:46<30:40:45, 11.59s/step, acc=0.4051, loss=2.6986, lr=9.67e-05]
Training:  14%|█▍        | 1550/11082 [4:58:57<30:26:12, 11.50s/step, acc=0.4051, loss=2.6986, lr=9.67e-05]
Training:  14%|█▍        | 1550/11082 [4:59:05<30:26:12, 11.50s/step, acc=0.4093, loss=2.6629, lr=9.65e-05]2026-06-18 19:52:32 | INFO | __main__ | epoch=0 step=1550 train_loss=2.662875 train_acc=0.409260 lr=9.648e-05 grad_norm=1.7812 acc_d0=0.6897 acc_d1=0.5280 acc_d2=0.3961 acc_d3=0.2943 acc_d4=0.2250 acc_d5=0.1776 acc_d6=0.1461 acc_d7=0.1232

Training:  14%|█▍        | 1551/11082 [4:59:09<30:27:06, 11.50s/step, acc=0.4093, loss=2.6629, lr=9.65e-05]
Training:  14%|█▍        | 1552/11082 [4:59:20<30:26:23, 11.50s/step, acc=0.4093, loss=2.6629, lr=9.65e-05]
Training:  14%|█▍        | 1553/11082 [4:59:32<30:35:51, 11.56s/step, acc=0.4093, loss=2.6629, lr=9.65e-05]
Training:  14%|█▍        | 1554/11082 [4:59:43<30:13:06, 11.42s/step, acc=0.4093, loss=2.6629, lr=9.65e-05]
Training:  14%|█▍        | 1555/11082 [4:59:55<30:25:14, 11.50s/step, acc=0.4093, loss=2.6629, lr=9.65e-05]
Training:  14%|█▍        | 1556/11082 [5:00:07<30:52:19, 11.67s/step, acc=0.4093, loss=2.6629, lr=9.65e-05]
Training:  14%|█▍        | 1557/11082 [5:00:18<30:41:57, 11.60s/step, acc=0.4093, loss=2.6629, lr=9.65e-05]
Training:  14%|█▍        | 1558/11082 [5:00:29<30:15:24, 11.44s/step, acc=0.4093, loss=2.6629, lr=9.65e-05]
Training:  14%|█▍        | 1559/11082 [5:00:41<30:32:44, 11.55s/step, acc=0.4093, loss=2.6629, lr=9.65e-05]
Training:  14%|█▍        | 1560/11082 [5:00:53<30:49:27, 11.65s/step, acc=0.4093, loss=2.6629, lr=9.65e-05]
Training:  14%|█▍        | 1561/11082 [5:01:05<30:48:26, 11.65s/step, acc=0.4093, loss=2.6629, lr=9.65e-05]
Training:  14%|█▍        | 1562/11082 [5:01:16<30:26:48, 11.51s/step, acc=0.4093, loss=2.6629, lr=9.65e-05]
Training:  14%|█▍        | 1563/11082 [5:01:27<30:25:47, 11.51s/step, acc=0.4093, loss=2.6629, lr=9.65e-05]
Training:  14%|█▍        | 1564/11082 [5:01:39<30:23:17, 11.49s/step, acc=0.4093, loss=2.6629, lr=9.65e-05]
Training:  14%|█▍        | 1565/11082 [5:01:51<30:36:01, 11.58s/step, acc=0.4093, loss=2.6629, lr=9.65e-05]
Training:  14%|█▍        | 1566/11082 [5:02:02<30:20:19, 11.48s/step, acc=0.4093, loss=2.6629, lr=9.65e-05]
Training:  14%|█▍        | 1567/11082 [5:02:13<30:06:12, 11.39s/step, acc=0.4093, loss=2.6629, lr=9.65e-05]
Training:  14%|█▍        | 1568/11082 [5:02:25<30:24:43, 11.51s/step, acc=0.4093, loss=2.6629, lr=9.65e-05]
Training:  14%|█▍        | 1569/11082 [5:02:36<30:24:23, 11.51s/step, acc=0.4093, loss=2.6629, lr=9.65e-05]
Training:  14%|█▍        | 1570/11082 [5:02:48<30:47:24, 11.65s/step, acc=0.4093, loss=2.6629, lr=9.65e-05]
Training:  14%|█▍        | 1571/11082 [5:02:59<30:21:04, 11.49s/step, acc=0.4093, loss=2.6629, lr=9.65e-05]
Training:  14%|█▍        | 1572/11082 [5:03:11<30:16:39, 11.46s/step, acc=0.4093, loss=2.6629, lr=9.65e-05]
Training:  14%|█▍        | 1573/11082 [5:03:23<30:30:43, 11.55s/step, acc=0.4093, loss=2.6629, lr=9.65e-05]
Training:  14%|█▍        | 1574/11082 [5:03:34<30:34:45, 11.58s/step, acc=0.4093, loss=2.6629, lr=9.65e-05]
Training:  14%|█▍        | 1575/11082 [5:03:46<30:29:06, 11.54s/step, acc=0.4093, loss=2.6629, lr=9.65e-05]
Training:  14%|█▍        | 1576/11082 [5:03:57<30:25:41, 11.52s/step, acc=0.4093, loss=2.6629, lr=9.65e-05]
Training:  14%|█▍        | 1577/11082 [5:04:09<30:26:04, 11.53s/step, acc=0.4093, loss=2.6629, lr=9.65e-05]
Training:  14%|█▍        | 1578/11082 [5:04:20<30:36:00, 11.59s/step, acc=0.4093, loss=2.6629, lr=9.65e-05]
Training:  14%|█▍        | 1579/11082 [5:04:32<30:27:54, 11.54s/step, acc=0.4093, loss=2.6629, lr=9.65e-05]
Training:  14%|█▍        | 1580/11082 [5:04:44<30:35:44, 11.59s/step, acc=0.4093, loss=2.6629, lr=9.65e-05]
Training:  14%|█▍        | 1581/11082 [5:04:55<30:28:08, 11.54s/step, acc=0.4093, loss=2.6629, lr=9.65e-05]
Training:  14%|█▍        | 1582/11082 [5:05:07<30:30:53, 11.56s/step, acc=0.4093, loss=2.6629, lr=9.65e-05]
Training:  14%|█▍        | 1583/11082 [5:05:18<30:31:02, 11.57s/step, acc=0.4093, loss=2.6629, lr=9.65e-05]
Training:  14%|█▍        | 1584/11082 [5:05:30<30:21:11, 11.50s/step, acc=0.4093, loss=2.6629, lr=9.65e-05]
Training:  14%|█▍        | 1585/11082 [5:05:41<30:30:35, 11.57s/step, acc=0.4093, loss=2.6629, lr=9.65e-05]
Training:  14%|█▍        | 1586/11082 [5:05:53<30:24:08, 11.53s/step, acc=0.4093, loss=2.6629, lr=9.65e-05]
Training:  14%|█▍        | 1587/11082 [5:06:04<30:31:14, 11.57s/step, acc=0.4093, loss=2.6629, lr=9.65e-05]
Training:  14%|█▍        | 1588/11082 [5:06:16<30:31:58, 11.58s/step, acc=0.4093, loss=2.6629, lr=9.65e-05]
Training:  14%|█▍        | 1589/11082 [5:06:28<30:49:27, 11.69s/step, acc=0.4093, loss=2.6629, lr=9.65e-05]
Training:  14%|█▍        | 1590/11082 [5:06:39<30:31:04, 11.57s/step, acc=0.4093, loss=2.6629, lr=9.65e-05]
Training:  14%|█▍        | 1591/11082 [5:06:51<30:52:36, 11.71s/step, acc=0.4093, loss=2.6629, lr=9.65e-05]
Training:  14%|█▍        | 1592/11082 [5:07:03<31:13:24, 11.84s/step, acc=0.4093, loss=2.6629, lr=9.65e-05]
Training:  14%|█▍        | 1593/11082 [5:07:15<30:42:15, 11.65s/step, acc=0.4093, loss=2.6629, lr=9.65e-05]
Training:  14%|█▍        | 1594/11082 [5:07:26<30:45:14, 11.67s/step, acc=0.4093, loss=2.6629, lr=9.65e-05]
Training:  14%|█▍        | 1595/11082 [5:07:38<30:23:57, 11.54s/step, acc=0.4093, loss=2.6629, lr=9.65e-05]
Training:  14%|█▍        | 1596/11082 [5:07:49<30:40:17, 11.64s/step, acc=0.4093, loss=2.6629, lr=9.65e-05]
Training:  14%|█▍        | 1597/11082 [5:08:01<30:45:46, 11.68s/step, acc=0.4093, loss=2.6629, lr=9.65e-05]
Training:  14%|█▍        | 1598/11082 [5:08:13<30:52:26, 11.72s/step, acc=0.4093, loss=2.6629, lr=9.65e-05]
Training:  14%|█▍        | 1599/11082 [5:08:25<30:50:18, 11.71s/step, acc=0.4093, loss=2.6629, lr=9.65e-05]
Training:  14%|█▍        | 1600/11082 [5:08:36<30:18:29, 11.51s/step, acc=0.4093, loss=2.6629, lr=9.65e-05]
Training:  14%|█▍        | 1600/11082 [5:08:43<30:18:29, 11.51s/step, acc=0.4127, loss=2.6304, lr=9.62e-05]2026-06-18 20:02:11 | INFO | __main__ | epoch=0 step=1600 train_loss=2.630362 train_acc=0.412733 lr=9.622e-05 grad_norm=1.6250 acc_d0=0.6917 acc_d1=0.5286 acc_d2=0.3945 acc_d3=0.2930 acc_d4=0.2235 acc_d5=0.1770 acc_d6=0.1466 acc_d7=0.1236

Training:  14%|█▍        | 1601/11082 [5:08:48<30:42:03, 11.66s/step, acc=0.4127, loss=2.6304, lr=9.62e-05]
Training:  14%|█▍        | 1602/11082 [5:08:59<30:00:10, 11.39s/step, acc=0.4127, loss=2.6304, lr=9.62e-05]
Training:  14%|█▍        | 1603/11082 [5:09:11<30:39:18, 11.64s/step, acc=0.4127, loss=2.6304, lr=9.62e-05]
Training:  14%|█▍        | 1604/11082 [5:09:23<30:52:43, 11.73s/step, acc=0.4127, loss=2.6304, lr=9.62e-05]
Training:  14%|█▍        | 1605/11082 [5:09:34<30:53:50, 11.74s/step, acc=0.4127, loss=2.6304, lr=9.62e-05]
Training:  14%|█▍        | 1606/11082 [5:09:46<30:34:12, 11.61s/step, acc=0.4127, loss=2.6304, lr=9.62e-05]
Training:  15%|█▍        | 1607/11082 [5:09:57<30:14:15, 11.49s/step, acc=0.4127, loss=2.6304, lr=9.62e-05]
Training:  15%|█▍        | 1608/11082 [5:10:09<30:20:46, 11.53s/step, acc=0.4127, loss=2.6304, lr=9.62e-05]
Training:  15%|█▍        | 1609/11082 [5:10:21<30:38:04, 11.64s/step, acc=0.4127, loss=2.6304, lr=9.62e-05]
Training:  15%|█▍        | 1610/11082 [5:10:32<30:43:30, 11.68s/step, acc=0.4127, loss=2.6304, lr=9.62e-05]
Training:  15%|█▍        | 1611/11082 [5:10:44<30:29:22, 11.59s/step, acc=0.4127, loss=2.6304, lr=9.62e-05]
Training:  15%|█▍        | 1612/11082 [5:10:55<30:36:57, 11.64s/step, acc=0.4127, loss=2.6304, lr=9.62e-05]
Training:  15%|█▍        | 1613/11082 [5:11:07<30:26:07, 11.57s/step, acc=0.4127, loss=2.6304, lr=9.62e-05]
Training:  15%|█▍        | 1614/11082 [5:11:19<30:35:19, 11.63s/step, acc=0.4127, loss=2.6304, lr=9.62e-05]
Training:  15%|█▍        | 1615/11082 [5:11:30<30:08:01, 11.46s/step, acc=0.4127, loss=2.6304, lr=9.62e-05]
Training:  15%|█▍        | 1616/11082 [5:11:42<30:36:04, 11.64s/step, acc=0.4127, loss=2.6304, lr=9.62e-05]
Training:  15%|█▍        | 1617/11082 [5:11:53<30:25:53, 11.57s/step, acc=0.4127, loss=2.6304, lr=9.62e-05]
Training:  15%|█▍        | 1618/11082 [5:12:04<30:01:10, 11.42s/step, acc=0.4127, loss=2.6304, lr=9.62e-05]
Training:  15%|█▍        | 1619/11082 [5:12:16<30:26:25, 11.58s/step, acc=0.4127, loss=2.6304, lr=9.62e-05]
Training:  15%|█▍        | 1620/11082 [5:12:27<30:11:26, 11.49s/step, acc=0.4127, loss=2.6304, lr=9.62e-05]
Training:  15%|█▍        | 1621/11082 [5:12:39<30:18:10, 11.53s/step, acc=0.4127, loss=2.6304, lr=9.62e-05]
Training:  15%|█▍        | 1622/11082 [5:12:51<30:14:36, 11.51s/step, acc=0.4127, loss=2.6304, lr=9.62e-05]
Training:  15%|█▍        | 1623/11082 [5:13:02<30:10:09, 11.48s/step, acc=0.4127, loss=2.6304, lr=9.62e-05]
Training:  15%|█▍        | 1624/11082 [5:13:14<30:21:10, 11.55s/step, acc=0.4127, loss=2.6304, lr=9.62e-05]
Training:  15%|█▍        | 1625/11082 [5:13:25<30:25:03, 11.58s/step, acc=0.4127, loss=2.6304, lr=9.62e-05]
Training:  15%|█▍        | 1626/11082 [5:13:37<30:09:52, 11.48s/step, acc=0.4127, loss=2.6304, lr=9.62e-05]
Training:  15%|█▍        | 1627/11082 [5:13:48<30:26:11, 11.59s/step, acc=0.4127, loss=2.6304, lr=9.62e-05]
Training:  15%|█▍        | 1628/11082 [5:14:00<30:08:57, 11.48s/step, acc=0.4127, loss=2.6304, lr=9.62e-05]
Training:  15%|█▍        | 1629/11082 [5:14:11<30:25:11, 11.58s/step, acc=0.4127, loss=2.6304, lr=9.62e-05]
Training:  15%|█▍        | 1630/11082 [5:14:23<30:17:07, 11.53s/step, acc=0.4127, loss=2.6304, lr=9.62e-05]
Training:  15%|█▍        | 1631/11082 [5:14:34<30:10:42, 11.50s/step, acc=0.4127, loss=2.6304, lr=9.62e-05]
Training:  15%|█▍        | 1632/11082 [5:14:46<30:21:04, 11.56s/step, acc=0.4127, loss=2.6304, lr=9.62e-05]
Training:  15%|█▍        | 1633/11082 [5:14:57<30:00:46, 11.43s/step, acc=0.4127, loss=2.6304, lr=9.62e-05]
Training:  15%|█▍        | 1634/11082 [5:15:09<30:29:57, 11.62s/step, acc=0.4127, loss=2.6304, lr=9.62e-05]
Training:  15%|█▍        | 1635/11082 [5:15:21<30:18:04, 11.55s/step, acc=0.4127, loss=2.6304, lr=9.62e-05]
Training:  15%|█▍        | 1636/11082 [5:15:32<30:15:04, 11.53s/step, acc=0.4127, loss=2.6304, lr=9.62e-05]
Training:  15%|█▍        | 1637/11082 [5:15:44<30:28:16, 11.61s/step, acc=0.4127, loss=2.6304, lr=9.62e-05]
Training:  15%|█▍        | 1638/11082 [5:15:55<30:29:40, 11.62s/step, acc=0.4127, loss=2.6304, lr=9.62e-05]
Training:  15%|█▍        | 1639/11082 [5:16:07<30:17:58, 11.55s/step, acc=0.4127, loss=2.6304, lr=9.62e-05]
Training:  15%|█▍        | 1640/11082 [5:16:19<30:28:43, 11.62s/step, acc=0.4127, loss=2.6304, lr=9.62e-05]
Training:  15%|█▍        | 1641/11082 [5:16:30<30:18:12, 11.56s/step, acc=0.4127, loss=2.6304, lr=9.62e-05]
Training:  15%|█▍        | 1642/11082 [5:16:41<30:09:13, 11.50s/step, acc=0.4127, loss=2.6304, lr=9.62e-05]
Training:  15%|█▍        | 1643/11082 [5:16:53<30:17:54, 11.56s/step, acc=0.4127, loss=2.6304, lr=9.62e-05]
Training:  15%|█▍        | 1644/11082 [5:17:05<30:21:34, 11.58s/step, acc=0.4127, loss=2.6304, lr=9.62e-05]
Training:  15%|█▍        | 1645/11082 [5:17:16<29:58:36, 11.44s/step, acc=0.4127, loss=2.6304, lr=9.62e-05]
Training:  15%|█▍        | 1646/11082 [5:17:28<30:23:57, 11.60s/step, acc=0.4127, loss=2.6304, lr=9.62e-05]
Training:  15%|█▍        | 1647/11082 [5:17:39<30:19:41, 11.57s/step, acc=0.4127, loss=2.6304, lr=9.62e-05]
Training:  15%|█▍        | 1648/11082 [5:17:51<30:01:52, 11.46s/step, acc=0.4127, loss=2.6304, lr=9.62e-05]
Training:  15%|█▍        | 1649/11082 [5:18:03<30:26:23, 11.62s/step, acc=0.4127, loss=2.6304, lr=9.62e-05]
Training:  15%|█▍        | 1650/11082 [5:18:14<30:04:02, 11.48s/step, acc=0.4127, loss=2.6304, lr=9.62e-05]
Training:  15%|█▍        | 1650/11082 [5:18:21<30:04:02, 11.48s/step, acc=0.4141, loss=2.6270, lr=9.60e-05]2026-06-18 20:11:49 | INFO | __main__ | epoch=0 step=1650 train_loss=2.626969 train_acc=0.414058 lr=9.596e-05 grad_norm=1.4688 acc_d0=0.6879 acc_d1=0.5267 acc_d2=0.3954 acc_d3=0.2960 acc_d4=0.2260 acc_d5=0.1800 acc_d6=0.1487 acc_d7=0.1253

Training:  15%|█▍        | 1651/11082 [5:18:26<30:23:29, 11.60s/step, acc=0.4141, loss=2.6270, lr=9.60e-05]
Training:  15%|█▍        | 1652/11082 [5:18:37<30:15:06, 11.55s/step, acc=0.4141, loss=2.6270, lr=9.60e-05]
Training:  15%|█▍        | 1653/11082 [5:18:48<30:06:58, 11.50s/step, acc=0.4141, loss=2.6270, lr=9.60e-05]
Training:  15%|█▍        | 1654/11082 [5:19:00<30:15:44, 11.56s/step, acc=0.4141, loss=2.6270, lr=9.60e-05]
Training:  15%|█▍        | 1655/11082 [5:19:11<30:00:47, 11.46s/step, acc=0.4141, loss=2.6270, lr=9.60e-05]
Training:  15%|█▍        | 1656/11082 [5:19:23<29:59:51, 11.46s/step, acc=0.4141, loss=2.6270, lr=9.60e-05]
Training:  15%|█▍        | 1657/11082 [5:19:35<30:25:57, 11.62s/step, acc=0.4141, loss=2.6270, lr=9.60e-05]
Training:  15%|█▍        | 1658/11082 [5:19:46<30:12:59, 11.54s/step, acc=0.4141, loss=2.6270, lr=9.60e-05]
Training:  15%|█▍        | 1659/11082 [5:19:58<30:11:34, 11.53s/step, acc=0.4141, loss=2.6270, lr=9.60e-05]
Training:  15%|█▍        | 1660/11082 [5:20:09<30:13:03, 11.55s/step, acc=0.4141, loss=2.6270, lr=9.60e-05]
Training:  15%|█▍        | 1661/11082 [5:20:20<29:53:28, 11.42s/step, acc=0.4141, loss=2.6270, lr=9.60e-05]
Training:  15%|█▍        | 1662/11082 [5:20:32<30:08:17, 11.52s/step, acc=0.4141, loss=2.6270, lr=9.60e-05]
Training:  15%|█▌        | 1663/11082 [5:20:43<29:56:27, 11.44s/step, acc=0.4141, loss=2.6270, lr=9.60e-05]
Training:  15%|█▌        | 1664/11082 [5:20:55<30:06:14, 11.51s/step, acc=0.4141, loss=2.6270, lr=9.60e-05]
Training:  15%|█▌        | 1665/11082 [5:21:07<30:07:00, 11.51s/step, acc=0.4141, loss=2.6270, lr=9.60e-05]
Training:  15%|█▌        | 1666/11082 [5:21:18<29:56:08, 11.45s/step, acc=0.4141, loss=2.6270, lr=9.60e-05]
Training:  15%|█▌        | 1667/11082 [5:21:30<30:34:35, 11.69s/step, acc=0.4141, loss=2.6270, lr=9.60e-05]
Training:  15%|█▌        | 1668/11082 [5:21:42<30:36:20, 11.70s/step, acc=0.4141, loss=2.6270, lr=9.60e-05]
Training:  15%|█▌        | 1669/11082 [5:21:54<30:36:16, 11.70s/step, acc=0.4141, loss=2.6270, lr=9.60e-05]
Training:  15%|█▌        | 1670/11082 [5:22:05<30:17:46, 11.59s/step, acc=0.4141, loss=2.6270, lr=9.60e-05]
Training:  15%|█▌        | 1671/11082 [5:22:17<30:31:43, 11.68s/step, acc=0.4141, loss=2.6270, lr=9.60e-05]
Training:  15%|█▌        | 1672/11082 [5:22:29<30:45:09, 11.77s/step, acc=0.4141, loss=2.6270, lr=9.60e-05]
Training:  15%|█▌        | 1673/11082 [5:22:40<30:42:25, 11.75s/step, acc=0.4141, loss=2.6270, lr=9.60e-05]
Training:  15%|█▌        | 1674/11082 [5:22:52<30:19:43, 11.61s/step, acc=0.4141, loss=2.6270, lr=9.60e-05]
Training:  15%|█▌        | 1675/11082 [5:23:03<30:12:27, 11.56s/step, acc=0.4141, loss=2.6270, lr=9.60e-05]
Training:  15%|█▌        | 1676/11082 [5:23:15<30:11:17, 11.55s/step, acc=0.4141, loss=2.6270, lr=9.60e-05]
Training:  15%|█▌        | 1677/11082 [5:23:27<30:33:46, 11.70s/step, acc=0.4141, loss=2.6270, lr=9.60e-05]
Training:  15%|█▌        | 1678/11082 [5:23:38<30:25:18, 11.65s/step, acc=0.4141, loss=2.6270, lr=9.60e-05]
Training:  15%|█▌        | 1679/11082 [5:23:50<30:22:40, 11.63s/step, acc=0.4141, loss=2.6270, lr=9.60e-05]
Training:  15%|█▌        | 1680/11082 [5:24:01<30:09:39, 11.55s/step, acc=0.4141, loss=2.6270, lr=9.60e-05]
Training:  15%|█▌        | 1681/11082 [5:24:13<30:20:35, 11.62s/step, acc=0.4141, loss=2.6270, lr=9.60e-05]
Training:  15%|█▌        | 1682/11082 [5:24:25<30:19:06, 11.61s/step, acc=0.4141, loss=2.6270, lr=9.60e-05]
Training:  15%|█▌        | 1683/11082 [5:24:37<30:36:53, 11.73s/step, acc=0.4141, loss=2.6270, lr=9.60e-05]
Training:  15%|█▌        | 1684/11082 [5:24:48<30:22:59, 11.64s/step, acc=0.4141, loss=2.6270, lr=9.60e-05]
Training:  15%|█▌        | 1685/11082 [5:24:59<30:13:45, 11.58s/step, acc=0.4141, loss=2.6270, lr=9.60e-05]
Training:  15%|█▌        | 1686/11082 [5:25:11<29:59:22, 11.49s/step, acc=0.4141, loss=2.6270, lr=9.60e-05]
Training:  15%|█▌        | 1687/11082 [5:25:22<29:46:00, 11.41s/step, acc=0.4141, loss=2.6270, lr=9.60e-05]
Training:  15%|█▌        | 1688/11082 [5:25:34<30:07:33, 11.54s/step, acc=0.4141, loss=2.6270, lr=9.60e-05]
Training:  15%|█▌        | 1689/11082 [5:25:45<29:46:04, 11.41s/step, acc=0.4141, loss=2.6270, lr=9.60e-05]
Training:  15%|█▌        | 1690/11082 [5:25:57<30:07:54, 11.55s/step, acc=0.4141, loss=2.6270, lr=9.60e-05]
Training:  15%|█▌        | 1691/11082 [5:26:08<30:04:45, 11.53s/step, acc=0.4141, loss=2.6270, lr=9.60e-05]
Training:  15%|█▌        | 1692/11082 [5:26:20<30:19:48, 11.63s/step, acc=0.4141, loss=2.6270, lr=9.60e-05]
Training:  15%|█▌        | 1693/11082 [5:26:31<30:04:03, 11.53s/step, acc=0.4141, loss=2.6270, lr=9.60e-05]
Training:  15%|█▌        | 1694/11082 [5:26:43<30:06:37, 11.55s/step, acc=0.4141, loss=2.6270, lr=9.60e-05]
Training:  15%|█▌        | 1695/11082 [5:26:54<29:58:33, 11.50s/step, acc=0.4141, loss=2.6270, lr=9.60e-05]
Training:  15%|█▌        | 1696/11082 [5:27:06<30:09:12, 11.57s/step, acc=0.4141, loss=2.6270, lr=9.60e-05]
Training:  15%|█▌        | 1697/11082 [5:27:17<29:58:59, 11.50s/step, acc=0.4141, loss=2.6270, lr=9.60e-05]
Training:  15%|█▌        | 1698/11082 [5:27:29<29:56:21, 11.49s/step, acc=0.4141, loss=2.6270, lr=9.60e-05]
Training:  15%|█▌        | 1699/11082 [5:27:40<29:57:42, 11.50s/step, acc=0.4141, loss=2.6270, lr=9.60e-05]
Training:  15%|█▌        | 1700/11082 [5:27:52<29:59:51, 11.51s/step, acc=0.4141, loss=2.6270, lr=9.60e-05]
Training:  15%|█▌        | 1700/11082 [5:27:59<29:59:51, 11.51s/step, acc=0.4159, loss=2.6109, lr=9.57e-05]2026-06-18 20:21:27 | INFO | __main__ | epoch=0 step=1700 train_loss=2.610929 train_acc=0.415893 lr=9.569e-05 grad_norm=1.3281 acc_d0=0.6850 acc_d1=0.5225 acc_d2=0.3910 acc_d3=0.2926 acc_d4=0.2239 acc_d5=0.1783 acc_d6=0.1468 acc_d7=0.1240

Training:  15%|█▌        | 1701/11082 [5:28:04<30:06:42, 11.56s/step, acc=0.4159, loss=2.6109, lr=9.57e-05]
Training:  15%|█▌        | 1702/11082 [5:28:15<29:36:36, 11.36s/step, acc=0.4159, loss=2.6109, lr=9.57e-05]
Training:  15%|█▌        | 1703/11082 [5:28:26<29:53:27, 11.47s/step, acc=0.4159, loss=2.6109, lr=9.57e-05]
Training:  15%|█▌        | 1704/11082 [5:28:38<30:16:58, 11.62s/step, acc=0.4159, loss=2.6109, lr=9.57e-05]
Training:  15%|█▌        | 1705/11082 [5:28:50<30:14:08, 11.61s/step, acc=0.4159, loss=2.6109, lr=9.57e-05]
Training:  15%|█▌        | 1706/11082 [5:29:01<30:13:46, 11.61s/step, acc=0.4159, loss=2.6109, lr=9.57e-05]
Training:  15%|█▌        | 1707/11082 [5:29:13<29:57:15, 11.50s/step, acc=0.4159, loss=2.6109, lr=9.57e-05]
Training:  15%|█▌        | 1708/11082 [5:29:24<29:57:08, 11.50s/step, acc=0.4159, loss=2.6109, lr=9.57e-05]
Training:  15%|█▌        | 1709/11082 [5:29:36<30:11:24, 11.60s/step, acc=0.4159, loss=2.6109, lr=9.57e-05]
Training:  15%|█▌        | 1710/11082 [5:29:47<30:03:20, 11.55s/step, acc=0.4159, loss=2.6109, lr=9.57e-05]
Training:  15%|█▌        | 1711/11082 [5:29:59<30:14:25, 11.62s/step, acc=0.4159, loss=2.6109, lr=9.57e-05]
Training:  15%|█▌        | 1712/11082 [5:30:10<29:56:47, 11.51s/step, acc=0.4159, loss=2.6109, lr=9.57e-05]
Training:  15%|█▌        | 1713/11082 [5:30:22<30:00:54, 11.53s/step, acc=0.4159, loss=2.6109, lr=9.57e-05]
Training:  15%|█▌        | 1714/11082 [5:30:34<30:05:33, 11.56s/step, acc=0.4159, loss=2.6109, lr=9.57e-05]
Training:  15%|█▌        | 1715/11082 [5:30:45<30:06:51, 11.57s/step, acc=0.4159, loss=2.6109, lr=9.57e-05]
Training:  15%|█▌        | 1716/11082 [5:30:57<30:08:50, 11.59s/step, acc=0.4159, loss=2.6109, lr=9.57e-05]
Training:  15%|█▌        | 1717/11082 [5:31:08<29:34:28, 11.37s/step, acc=0.4159, loss=2.6109, lr=9.57e-05]
Training:  16%|█▌        | 1718/11082 [5:31:20<29:57:59, 11.52s/step, acc=0.4159, loss=2.6109, lr=9.57e-05]
Training:  16%|█▌        | 1719/11082 [5:31:31<30:04:57, 11.57s/step, acc=0.4159, loss=2.6109, lr=9.57e-05]
Training:  16%|█▌        | 1720/11082 [5:31:43<30:07:01, 11.58s/step, acc=0.4159, loss=2.6109, lr=9.57e-05]
Training:  16%|█▌        | 1721/11082 [5:31:54<30:06:31, 11.58s/step, acc=0.4159, loss=2.6109, lr=9.57e-05]
Training:  16%|█▌        | 1722/11082 [5:32:06<29:57:29, 11.52s/step, acc=0.4159, loss=2.6109, lr=9.57e-05]
Training:  16%|█▌        | 1723/11082 [5:32:17<29:27:47, 11.33s/step, acc=0.4159, loss=2.6109, lr=9.57e-05]
Training:  16%|█▌        | 1724/11082 [5:32:29<30:00:50, 11.55s/step, acc=0.4159, loss=2.6109, lr=9.57e-05]
Training:  16%|█▌        | 1725/11082 [5:32:41<30:19:00, 11.66s/step, acc=0.4159, loss=2.6109, lr=9.57e-05]
Training:  16%|█▌        | 1726/11082 [5:32:52<30:06:20, 11.58s/step, acc=0.4159, loss=2.6109, lr=9.57e-05]
Training:  16%|█▌        | 1727/11082 [5:33:04<30:11:40, 11.62s/step, acc=0.4159, loss=2.6109, lr=9.57e-05]
Training:  16%|█▌        | 1728/11082 [5:33:15<29:46:25, 11.46s/step, acc=0.4159, loss=2.6109, lr=9.57e-05]
Training:  16%|█▌        | 1729/11082 [5:33:27<30:04:30, 11.58s/step, acc=0.4159, loss=2.6109, lr=9.57e-05]
Training:  16%|█▌        | 1730/11082 [5:33:38<29:45:45, 11.46s/step, acc=0.4159, loss=2.6109, lr=9.57e-05]
Training:  16%|█▌        | 1731/11082 [5:33:50<30:05:34, 11.59s/step, acc=0.4159, loss=2.6109, lr=9.57e-05]
Training:  16%|█▌        | 1732/11082 [5:34:01<30:06:52, 11.59s/step, acc=0.4159, loss=2.6109, lr=9.57e-05]
Training:  16%|█▌        | 1733/11082 [5:34:13<29:43:49, 11.45s/step, acc=0.4159, loss=2.6109, lr=9.57e-05]
Training:  16%|█▌        | 1734/11082 [5:34:24<29:53:41, 11.51s/step, acc=0.4159, loss=2.6109, lr=9.57e-05]
Training:  16%|█▌        | 1735/11082 [5:34:36<30:09:00, 11.61s/step, acc=0.4159, loss=2.6109, lr=9.57e-05]
Training:  16%|█▌        | 1736/11082 [5:34:48<30:07:05, 11.60s/step, acc=0.4159, loss=2.6109, lr=9.57e-05]
Training:  16%|█▌        | 1737/11082 [5:34:59<29:59:59, 11.56s/step, acc=0.4159, loss=2.6109, lr=9.57e-05]
Training:  16%|█▌        | 1738/11082 [5:35:10<29:42:48, 11.45s/step, acc=0.4159, loss=2.6109, lr=9.57e-05]
Training:  16%|█▌        | 1739/11082 [5:35:22<30:07:19, 11.61s/step, acc=0.4159, loss=2.6109, lr=9.57e-05]
Training:  16%|█▌        | 1740/11082 [5:35:33<29:44:26, 11.46s/step, acc=0.4159, loss=2.6109, lr=9.57e-05]
Training:  16%|█▌        | 1741/11082 [5:35:45<30:00:57, 11.57s/step, acc=0.4159, loss=2.6109, lr=9.57e-05]
Training:  16%|█▌        | 1742/11082 [5:35:57<29:57:13, 11.55s/step, acc=0.4159, loss=2.6109, lr=9.57e-05]
Training:  16%|█▌        | 1743/11082 [5:36:08<29:40:58, 11.44s/step, acc=0.4159, loss=2.6109, lr=9.57e-05]
Training:  16%|█▌        | 1744/11082 [5:36:20<30:19:22, 11.69s/step, acc=0.4159, loss=2.6109, lr=9.57e-05]
Training:  16%|█▌        | 1745/11082 [5:36:31<29:59:59, 11.57s/step, acc=0.4159, loss=2.6109, lr=9.57e-05]
Training:  16%|█▌        | 1746/11082 [5:36:43<30:05:37, 11.60s/step, acc=0.4159, loss=2.6109, lr=9.57e-05]
Training:  16%|█▌        | 1747/11082 [5:36:54<29:50:46, 11.51s/step, acc=0.4159, loss=2.6109, lr=9.57e-05]
Training:  16%|█▌        | 1748/11082 [5:37:06<30:01:44, 11.58s/step, acc=0.4159, loss=2.6109, lr=9.57e-05]
Training:  16%|█▌        | 1749/11082 [5:37:18<29:56:20, 11.55s/step, acc=0.4159, loss=2.6109, lr=9.57e-05]
Training:  16%|█▌        | 1750/11082 [5:37:29<30:04:31, 11.60s/step, acc=0.4159, loss=2.6109, lr=9.57e-05]
Training:  16%|█▌        | 1750/11082 [5:37:37<30:04:31, 11.60s/step, acc=0.4170, loss=2.6029, lr=9.54e-05]2026-06-18 20:31:04 | INFO | __main__ | epoch=0 step=1750 train_loss=2.602861 train_acc=0.416978 lr=9.540e-05 grad_norm=1.2422 acc_d0=0.6775 acc_d1=0.5161 acc_d2=0.3873 acc_d3=0.2888 acc_d4=0.2216 acc_d5=0.1761 acc_d6=0.1457 acc_d7=0.1253

Training:  16%|█▌        | 1751/11082 [5:37:40<29:41:00, 11.45s/step, acc=0.4170, loss=2.6029, lr=9.54e-05]
Training:  16%|█▌        | 1752/11082 [5:37:53<30:18:42, 11.70s/step, acc=0.4170, loss=2.6029, lr=9.54e-05]
Training:  16%|█▌        | 1753/11082 [5:38:04<30:13:29, 11.66s/step, acc=0.4170, loss=2.6029, lr=9.54e-05]
Training:  16%|█▌        | 1754/11082 [5:38:16<30:22:35, 11.72s/step, acc=0.4170, loss=2.6029, lr=9.54e-05]
Training:  16%|█▌        | 1755/11082 [5:38:27<29:48:25, 11.50s/step, acc=0.4170, loss=2.6029, lr=9.54e-05]
Training:  16%|█▌        | 1756/11082 [5:38:39<30:09:35, 11.64s/step, acc=0.4170, loss=2.6029, lr=9.54e-05]
Training:  16%|█▌        | 1757/11082 [5:38:51<30:25:52, 11.75s/step, acc=0.4170, loss=2.6029, lr=9.54e-05]
Training:  16%|█▌        | 1758/11082 [5:39:03<30:08:37, 11.64s/step, acc=0.4170, loss=2.6029, lr=9.54e-05]
Training:  16%|█▌        | 1759/11082 [5:39:15<30:31:29, 11.79s/step, acc=0.4170, loss=2.6029, lr=9.54e-05]
Training:  16%|█▌        | 1760/11082 [5:39:26<30:30:30, 11.78s/step, acc=0.4170, loss=2.6029, lr=9.54e-05]
Training:  16%|█▌        | 1761/11082 [5:39:38<30:35:20, 11.81s/step, acc=0.4170, loss=2.6029, lr=9.54e-05]
Training:  16%|█▌        | 1762/11082 [5:39:50<30:09:49, 11.65s/step, acc=0.4170, loss=2.6029, lr=9.54e-05]
Training:  16%|█▌        | 1763/11082 [5:40:01<30:12:47, 11.67s/step, acc=0.4170, loss=2.6029, lr=9.54e-05]
Training:  16%|█▌        | 1764/11082 [5:40:13<30:16:22, 11.70s/step, acc=0.4170, loss=2.6029, lr=9.54e-05]
Training:  16%|█▌        | 1765/11082 [5:40:24<29:45:08, 11.50s/step, acc=0.4170, loss=2.6029, lr=9.54e-05]
Training:  16%|█▌        | 1766/11082 [5:40:36<29:58:21, 11.58s/step, acc=0.4170, loss=2.6029, lr=9.54e-05]
Training:  16%|█▌        | 1767/11082 [5:40:47<29:50:37, 11.53s/step, acc=0.4170, loss=2.6029, lr=9.54e-05]
Training:  16%|█▌        | 1768/11082 [5:40:59<29:53:07, 11.55s/step, acc=0.4170, loss=2.6029, lr=9.54e-05]
Training:  16%|█▌        | 1769/11082 [5:41:11<30:01:36, 11.61s/step, acc=0.4170, loss=2.6029, lr=9.54e-05]
Training:  16%|█▌        | 1770/11082 [5:41:22<29:59:15, 11.59s/step, acc=0.4170, loss=2.6029, lr=9.54e-05]
Training:  16%|█▌        | 1771/11082 [5:41:34<29:56:02, 11.57s/step, acc=0.4170, loss=2.6029, lr=9.54e-05]
Training:  16%|█▌        | 1772/11082 [5:41:45<30:01:49, 11.61s/step, acc=0.4170, loss=2.6029, lr=9.54e-05]
Training:  16%|█▌        | 1773/11082 [5:41:57<29:57:02, 11.58s/step, acc=0.4170, loss=2.6029, lr=9.54e-05]
Training:  16%|█▌        | 1774/11082 [5:42:09<30:04:11, 11.63s/step, acc=0.4170, loss=2.6029, lr=9.54e-05]
Training:  16%|█▌        | 1775/11082 [5:42:20<29:48:18, 11.53s/step, acc=0.4170, loss=2.6029, lr=9.54e-05]
Training:  16%|█▌        | 1776/11082 [5:42:32<30:12:00, 11.68s/step, acc=0.4170, loss=2.6029, lr=9.54e-05]
Training:  16%|█▌        | 1777/11082 [5:42:44<30:12:37, 11.69s/step, acc=0.4170, loss=2.6029, lr=9.54e-05]
Training:  16%|█▌        | 1778/11082 [5:42:55<29:59:45, 11.61s/step, acc=0.4170, loss=2.6029, lr=9.54e-05]
Training:  16%|█▌        | 1779/11082 [5:43:07<29:51:48, 11.56s/step, acc=0.4170, loss=2.6029, lr=9.54e-05]
Training:  16%|█▌        | 1780/11082 [5:43:18<29:50:14, 11.55s/step, acc=0.4170, loss=2.6029, lr=9.54e-05]
Training:  16%|█▌        | 1781/11082 [5:43:30<29:51:48, 11.56s/step, acc=0.4170, loss=2.6029, lr=9.54e-05]
Training:  16%|█▌        | 1782/11082 [5:43:41<29:32:31, 11.44s/step, acc=0.4170, loss=2.6029, lr=9.54e-05]
Training:  16%|█▌        | 1783/11082 [5:43:52<29:24:29, 11.39s/step, acc=0.4170, loss=2.6029, lr=9.54e-05]
Training:  16%|█▌        | 1784/11082 [5:44:04<30:02:21, 11.63s/step, acc=0.4170, loss=2.6029, lr=9.54e-05]
Training:  16%|█▌        | 1785/11082 [5:44:16<29:42:27, 11.50s/step, acc=0.4170, loss=2.6029, lr=9.54e-05]
Training:  16%|█▌        | 1786/11082 [5:44:28<30:08:45, 11.67s/step, acc=0.4170, loss=2.6029, lr=9.54e-05]
Training:  16%|█▌        | 1787/11082 [5:44:39<29:54:52, 11.59s/step, acc=0.4170, loss=2.6029, lr=9.54e-05]
Training:  16%|█▌        | 1788/11082 [5:44:50<29:40:29, 11.49s/step, acc=0.4170, loss=2.6029, lr=9.54e-05]
Training:  16%|█▌        | 1789/11082 [5:45:02<29:31:13, 11.44s/step, acc=0.4170, loss=2.6029, lr=9.54e-05]
Training:  16%|█▌        | 1790/11082 [5:45:13<29:54:00, 11.58s/step, acc=0.4170, loss=2.6029, lr=9.54e-05]
Training:  16%|█▌        | 1791/11082 [5:45:25<29:55:40, 11.60s/step, acc=0.4170, loss=2.6029, lr=9.54e-05]
Training:  16%|█▌        | 1792/11082 [5:45:37<29:53:06, 11.58s/step, acc=0.4170, loss=2.6029, lr=9.54e-05]
Training:  16%|█▌        | 1793/11082 [5:45:48<29:37:46, 11.48s/step, acc=0.4170, loss=2.6029, lr=9.54e-05]
Training:  16%|█▌        | 1794/11082 [5:45:59<29:37:53, 11.49s/step, acc=0.4170, loss=2.6029, lr=9.54e-05]
Training:  16%|█▌        | 1795/11082 [5:46:11<30:00:37, 11.63s/step, acc=0.4170, loss=2.6029, lr=9.54e-05]
Training:  16%|█▌        | 1796/11082 [5:46:23<29:40:37, 11.51s/step, acc=0.4170, loss=2.6029, lr=9.54e-05]
Training:  16%|█▌        | 1797/11082 [5:46:34<29:44:59, 11.53s/step, acc=0.4170, loss=2.6029, lr=9.54e-05]
Training:  16%|█▌        | 1798/11082 [5:46:46<29:45:48, 11.54s/step, acc=0.4170, loss=2.6029, lr=9.54e-05]
Training:  16%|█▌        | 1799/11082 [5:46:57<29:38:05, 11.49s/step, acc=0.4170, loss=2.6029, lr=9.54e-05]
Training:  16%|█▌        | 1800/11082 [5:47:09<29:57:42, 11.62s/step, acc=0.4170, loss=2.6029, lr=9.54e-05]
Training:  16%|█▌        | 1800/11082 [5:47:16<29:57:42, 11.62s/step, acc=0.4195, loss=2.5785, lr=9.51e-05]2026-06-18 20:40:44 | INFO | __main__ | epoch=0 step=1800 train_loss=2.578512 train_acc=0.419535 lr=9.512e-05 grad_norm=1.5234 acc_d0=0.6815 acc_d1=0.5172 acc_d2=0.3851 acc_d3=0.2861 acc_d4=0.2199 acc_d5=0.1736 acc_d6=0.1428 acc_d7=0.1213

Training:  16%|█▋        | 1801/11082 [5:47:21<29:51:58, 11.58s/step, acc=0.4195, loss=2.5785, lr=9.51e-05]
Training:  16%|█▋        | 1802/11082 [5:47:32<29:41:23, 11.52s/step, acc=0.4195, loss=2.5785, lr=9.51e-05]
Training:  16%|█▋        | 1803/11082 [5:47:43<29:07:17, 11.30s/step, acc=0.4195, loss=2.5785, lr=9.51e-05]
Training:  16%|█▋        | 1804/11082 [5:47:55<29:35:17, 11.48s/step, acc=0.4195, loss=2.5785, lr=9.51e-05]
Training:  16%|█▋        | 1805/11082 [5:48:06<29:28:39, 11.44s/step, acc=0.4195, loss=2.5785, lr=9.51e-05]
Training:  16%|█▋        | 1806/11082 [5:48:18<29:46:46, 11.56s/step, acc=0.4195, loss=2.5785, lr=9.51e-05]
Training:  16%|█▋        | 1807/11082 [5:48:29<29:47:49, 11.57s/step, acc=0.4195, loss=2.5785, lr=9.51e-05]
Training:  16%|█▋        | 1808/11082 [5:48:40<29:25:48, 11.42s/step, acc=0.4195, loss=2.5785, lr=9.51e-05]
Training:  16%|█▋        | 1809/11082 [5:48:52<29:43:09, 11.54s/step, acc=0.4195, loss=2.5785, lr=9.51e-05]
Training:  16%|█▋        | 1810/11082 [5:49:04<29:44:46, 11.55s/step, acc=0.4195, loss=2.5785, lr=9.51e-05]
Training:  16%|█▋        | 1811/11082 [5:49:15<29:40:15, 11.52s/step, acc=0.4195, loss=2.5785, lr=9.51e-05]
Training:  16%|█▋        | 1812/11082 [5:49:27<29:37:27, 11.50s/step, acc=0.4195, loss=2.5785, lr=9.51e-05]
Training:  16%|█▋        | 1813/11082 [5:49:38<29:29:22, 11.45s/step, acc=0.4195, loss=2.5785, lr=9.51e-05]
Training:  16%|█▋        | 1814/11082 [5:49:50<29:46:19, 11.56s/step, acc=0.4195, loss=2.5785, lr=9.51e-05]
Training:  16%|█▋        | 1815/11082 [5:50:01<29:28:30, 11.45s/step, acc=0.4195, loss=2.5785, lr=9.51e-05]
Training:  16%|█▋        | 1816/11082 [5:50:13<29:40:25, 11.53s/step, acc=0.4195, loss=2.5785, lr=9.51e-05]
Training:  16%|█▋        | 1817/11082 [5:50:24<29:46:23, 11.57s/step, acc=0.4195, loss=2.5785, lr=9.51e-05]
Training:  16%|█▋        | 1818/11082 [5:50:36<29:33:44, 11.49s/step, acc=0.4195, loss=2.5785, lr=9.51e-05]
Training:  16%|█▋        | 1819/11082 [5:50:48<29:47:09, 11.58s/step, acc=0.4195, loss=2.5785, lr=9.51e-05]
Training:  16%|█▋        | 1820/11082 [5:50:59<29:43:59, 11.56s/step, acc=0.4195, loss=2.5785, lr=9.51e-05]
Training:  16%|█▋        | 1821/11082 [5:51:11<29:49:57, 11.60s/step, acc=0.4195, loss=2.5785, lr=9.51e-05]
Training:  16%|█▋        | 1822/11082 [5:51:22<29:40:44, 11.54s/step, acc=0.4195, loss=2.5785, lr=9.51e-05]
Training:  16%|█▋        | 1823/11082 [5:51:34<29:40:22, 11.54s/step, acc=0.4195, loss=2.5785, lr=9.51e-05]
Training:  16%|█▋        | 1824/11082 [5:51:45<29:27:17, 11.45s/step, acc=0.4195, loss=2.5785, lr=9.51e-05]
Training:  16%|█▋        | 1825/11082 [5:51:57<29:46:17, 11.58s/step, acc=0.4195, loss=2.5785, lr=9.51e-05]
Training:  16%|█▋        | 1826/11082 [5:52:09<30:00:32, 11.67s/step, acc=0.4195, loss=2.5785, lr=9.51e-05]
Training:  16%|█▋        | 1827/11082 [5:52:20<29:47:18, 11.59s/step, acc=0.4195, loss=2.5785, lr=9.51e-05]
Training:  16%|█▋        | 1828/11082 [5:52:32<29:56:18, 11.65s/step, acc=0.4195, loss=2.5785, lr=9.51e-05]
Training:  17%|█▋        | 1829/11082 [5:52:43<29:30:36, 11.48s/step, acc=0.4195, loss=2.5785, lr=9.51e-05]
Training:  17%|█▋        | 1830/11082 [5:52:55<29:52:30, 11.62s/step, acc=0.4195, loss=2.5785, lr=9.51e-05]
Training:  17%|█▋        | 1831/11082 [5:53:06<29:43:10, 11.57s/step, acc=0.4195, loss=2.5785, lr=9.51e-05]
Training:  17%|█▋        | 1832/11082 [5:53:18<29:37:46, 11.53s/step, acc=0.4195, loss=2.5785, lr=9.51e-05]
Training:  17%|█▋        | 1833/11082 [5:53:30<29:59:36, 11.67s/step, acc=0.4195, loss=2.5785, lr=9.51e-05]
Training:  17%|█▋        | 1834/11082 [5:53:41<29:48:58, 11.61s/step, acc=0.4195, loss=2.5785, lr=9.51e-05]
Training:  17%|█▋        | 1835/11082 [5:53:53<29:59:03, 11.67s/step, acc=0.4195, loss=2.5785, lr=9.51e-05]
Training:  17%|█▋        | 1836/11082 [5:54:05<29:50:48, 11.62s/step, acc=0.4195, loss=2.5785, lr=9.51e-05]
Training:  17%|█▋        | 1837/11082 [5:54:16<30:02:57, 11.70s/step, acc=0.4195, loss=2.5785, lr=9.51e-05]
Training:  17%|█▋        | 1838/11082 [5:54:28<29:41:48, 11.57s/step, acc=0.4195, loss=2.5785, lr=9.51e-05]
Training:  17%|█▋        | 1839/11082 [5:54:40<29:50:57, 11.63s/step, acc=0.4195, loss=2.5785, lr=9.51e-05]
Training:  17%|█▋        | 1840/11082 [5:54:51<29:45:43, 11.59s/step, acc=0.4195, loss=2.5785, lr=9.51e-05]
Training:  17%|█▋        | 1841/11082 [5:55:02<29:29:14, 11.49s/step, acc=0.4195, loss=2.5785, lr=9.51e-05]
Training:  17%|█▋        | 1842/11082 [5:55:14<30:02:06, 11.70s/step, acc=0.4195, loss=2.5785, lr=9.51e-05]
Training:  17%|█▋        | 1843/11082 [5:55:26<29:42:51, 11.58s/step, acc=0.4195, loss=2.5785, lr=9.51e-05]
Training:  17%|█▋        | 1844/11082 [5:55:37<29:33:17, 11.52s/step, acc=0.4195, loss=2.5785, lr=9.51e-05]
Training:  17%|█▋        | 1845/11082 [5:55:49<29:42:07, 11.58s/step, acc=0.4195, loss=2.5785, lr=9.51e-05]
Training:  17%|█▋        | 1846/11082 [5:56:00<29:29:42, 11.50s/step, acc=0.4195, loss=2.5785, lr=9.51e-05]
Training:  17%|█▋        | 1847/11082 [5:56:12<29:46:13, 11.61s/step, acc=0.4195, loss=2.5785, lr=9.51e-05]
Training:  17%|█▋        | 1848/11082 [5:56:23<29:27:11, 11.48s/step, acc=0.4195, loss=2.5785, lr=9.51e-05]
Training:  17%|█▋        | 1849/11082 [5:56:35<29:45:26, 11.60s/step, acc=0.4195, loss=2.5785, lr=9.51e-05]
Training:  17%|█▋        | 1850/11082 [5:56:46<29:16:33, 11.42s/step, acc=0.4195, loss=2.5785, lr=9.51e-05]
Training:  17%|█▋        | 1850/11082 [5:56:54<29:16:33, 11.42s/step, acc=0.4210, loss=2.5713, lr=9.48e-05]2026-06-18 20:50:21 | INFO | __main__ | epoch=0 step=1850 train_loss=2.571262 train_acc=0.421048 lr=9.482e-05 grad_norm=1.5703 acc_d0=0.7046 acc_d1=0.5375 acc_d2=0.4036 acc_d3=0.3017 acc_d4=0.2297 acc_d5=0.1838 acc_d6=0.1500 acc_d7=0.1271

Training:  17%|█▋        | 1851/11082 [5:56:58<29:19:27, 11.44s/step, acc=0.4210, loss=2.5713, lr=9.48e-05]
Training:  17%|█▋        | 1852/11082 [5:57:09<29:38:38, 11.56s/step, acc=0.4210, loss=2.5713, lr=9.48e-05]
Training:  17%|█▋        | 1853/11082 [5:57:21<29:52:41, 11.65s/step, acc=0.4210, loss=2.5713, lr=9.48e-05]
Training:  17%|█▋        | 1854/11082 [5:57:33<29:56:03, 11.68s/step, acc=0.4210, loss=2.5713, lr=9.48e-05]
Training:  17%|█▋        | 1855/11082 [5:57:44<29:34:20, 11.54s/step, acc=0.4210, loss=2.5713, lr=9.48e-05]
Training:  17%|█▋        | 1856/11082 [5:57:55<29:10:19, 11.38s/step, acc=0.4210, loss=2.5713, lr=9.48e-05]
Training:  17%|█▋        | 1857/11082 [5:58:08<29:50:34, 11.65s/step, acc=0.4210, loss=2.5713, lr=9.48e-05]
Training:  17%|█▋        | 1858/11082 [5:58:19<29:29:22, 11.51s/step, acc=0.4210, loss=2.5713, lr=9.48e-05]
Training:  17%|█▋        | 1859/11082 [5:58:31<29:44:01, 11.61s/step, acc=0.4210, loss=2.5713, lr=9.48e-05]
Training:  17%|█▋        | 1860/11082 [5:58:42<29:44:21, 11.61s/step, acc=0.4210, loss=2.5713, lr=9.48e-05]
Training:  17%|█▋        | 1861/11082 [5:58:53<29:21:37, 11.46s/step, acc=0.4210, loss=2.5713, lr=9.48e-05]
Training:  17%|█▋        | 1862/11082 [5:59:05<29:19:58, 11.45s/step, acc=0.4210, loss=2.5713, lr=9.48e-05]
Training:  17%|█▋        | 1863/11082 [5:59:16<29:20:02, 11.45s/step, acc=0.4210, loss=2.5713, lr=9.48e-05]
Training:  17%|█▋        | 1864/11082 [5:59:28<29:40:05, 11.59s/step, acc=0.4210, loss=2.5713, lr=9.48e-05]
Training:  17%|█▋        | 1865/11082 [5:59:39<29:29:11, 11.52s/step, acc=0.4210, loss=2.5713, lr=9.48e-05]
Training:  17%|█▋        | 1866/11082 [5:59:51<29:22:57, 11.48s/step, acc=0.4210, loss=2.5713, lr=9.48e-05]
Training:  17%|█▋        | 1867/11082 [6:00:03<29:42:56, 11.61s/step, acc=0.4210, loss=2.5713, lr=9.48e-05]
Training:  17%|█▋        | 1868/11082 [6:00:14<29:24:42, 11.49s/step, acc=0.4210, loss=2.5713, lr=9.48e-05]
Training:  17%|█▋        | 1869/11082 [6:00:26<29:28:40, 11.52s/step, acc=0.4210, loss=2.5713, lr=9.48e-05]
Training:  17%|█▋        | 1870/11082 [6:00:37<29:36:15, 11.57s/step, acc=0.4210, loss=2.5713, lr=9.48e-05]
Training:  17%|█▋        | 1871/11082 [6:00:48<29:11:07, 11.41s/step, acc=0.4210, loss=2.5713, lr=9.48e-05]
Training:  17%|█▋        | 1872/11082 [6:01:00<29:36:52, 11.58s/step, acc=0.4210, loss=2.5713, lr=9.48e-05]
Training:  17%|█▋        | 1873/11082 [6:01:11<29:02:50, 11.36s/step, acc=0.4210, loss=2.5713, lr=9.48e-05]
Training:  17%|█▋        | 1874/11082 [6:01:23<29:34:16, 11.56s/step, acc=0.4210, loss=2.5713, lr=9.48e-05]
Training:  17%|█▋        | 1875/11082 [6:01:34<29:10:58, 11.41s/step, acc=0.4210, loss=2.5713, lr=9.48e-05]
Training:  17%|█▋        | 1876/11082 [6:01:46<29:28:19, 11.53s/step, acc=0.4210, loss=2.5713, lr=9.48e-05]
Training:  17%|█▋        | 1877/11082 [6:01:58<29:39:18, 11.60s/step, acc=0.4210, loss=2.5713, lr=9.48e-05]
Training:  17%|█▋        | 1878/11082 [6:02:09<29:19:03, 11.47s/step, acc=0.4210, loss=2.5713, lr=9.48e-05]
Training:  17%|█▋        | 1879/11082 [6:02:20<29:26:59, 11.52s/step, acc=0.4210, loss=2.5713, lr=9.48e-05]
Training:  17%|█▋        | 1880/11082 [6:02:32<29:29:13, 11.54s/step, acc=0.4210, loss=2.5713, lr=9.48e-05]
Training:  17%|█▋        | 1881/11082 [6:02:43<29:21:08, 11.48s/step, acc=0.4210, loss=2.5713, lr=9.48e-05]
Training:  17%|█▋        | 1882/11082 [6:02:55<29:23:54, 11.50s/step, acc=0.4210, loss=2.5713, lr=9.48e-05]
Training:  17%|█▋        | 1883/11082 [6:03:07<29:31:20, 11.55s/step, acc=0.4210, loss=2.5713, lr=9.48e-05]
Training:  17%|█▋        | 1884/11082 [6:03:18<29:35:04, 11.58s/step, acc=0.4210, loss=2.5713, lr=9.48e-05]
Training:  17%|█▋        | 1885/11082 [6:03:30<29:22:55, 11.50s/step, acc=0.4210, loss=2.5713, lr=9.48e-05]
Training:  17%|█▋        | 1886/11082 [6:03:42<29:40:49, 11.62s/step, acc=0.4210, loss=2.5713, lr=9.48e-05]
Training:  17%|█▋        | 1887/11082 [6:03:53<29:26:06, 11.52s/step, acc=0.4210, loss=2.5713, lr=9.48e-05]
Training:  17%|█▋        | 1888/11082 [6:04:04<29:29:50, 11.55s/step, acc=0.4210, loss=2.5713, lr=9.48e-05]
Training:  17%|█▋        | 1889/11082 [6:04:16<29:19:24, 11.48s/step, acc=0.4210, loss=2.5713, lr=9.48e-05]
Training:  17%|█▋        | 1890/11082 [6:04:28<29:56:45, 11.73s/step, acc=0.4210, loss=2.5713, lr=9.48e-05]
Training:  17%|█▋        | 1891/11082 [6:04:39<29:25:18, 11.52s/step, acc=0.4210, loss=2.5713, lr=9.48e-05]
Training:  17%|█▋        | 1892/11082 [6:04:51<29:42:24, 11.64s/step, acc=0.4210, loss=2.5713, lr=9.48e-05]
Training:  17%|█▋        | 1893/11082 [6:05:03<30:01:48, 11.76s/step, acc=0.4210, loss=2.5713, lr=9.48e-05]
Training:  17%|█▋        | 1894/11082 [6:05:15<29:56:04, 11.73s/step, acc=0.4210, loss=2.5713, lr=9.48e-05]
Training:  17%|█▋        | 1895/11082 [6:05:26<29:48:23, 11.68s/step, acc=0.4210, loss=2.5713, lr=9.48e-05]
Training:  17%|█▋        | 1896/11082 [6:05:37<29:23:35, 11.52s/step, acc=0.4210, loss=2.5713, lr=9.48e-05]
Training:  17%|█▋        | 1897/11082 [6:05:49<29:27:41, 11.55s/step, acc=0.4210, loss=2.5713, lr=9.48e-05]
Training:  17%|█▋        | 1898/11082 [6:06:01<29:33:14, 11.58s/step, acc=0.4210, loss=2.5713, lr=9.48e-05]
Training:  17%|█▋        | 1899/11082 [6:06:13<29:53:03, 11.72s/step, acc=0.4210, loss=2.5713, lr=9.48e-05]
Training:  17%|█▋        | 1900/11082 [6:06:24<29:45:11, 11.67s/step, acc=0.4210, loss=2.5713, lr=9.48e-05]
Training:  17%|█▋        | 1900/11082 [6:06:31<29:45:11, 11.67s/step, acc=0.4244, loss=2.5462, lr=9.45e-05]2026-06-18 20:59:59 | INFO | __main__ | epoch=0 step=1900 train_loss=2.546180 train_acc=0.424386 lr=9.451e-05 grad_norm=1.1328 acc_d0=0.6852 acc_d1=0.5212 acc_d2=0.3871 acc_d3=0.2894 acc_d4=0.2216 acc_d5=0.1755 acc_d6=0.1469 acc_d7=0.1224

Training:  17%|█▋        | 1901/11082 [6:06:36<29:31:15, 11.58s/step, acc=0.4244, loss=2.5462, lr=9.45e-05]
Training:  17%|█▋        | 1902/11082 [6:06:47<29:00:16, 11.37s/step, acc=0.4244, loss=2.5462, lr=9.45e-05]
Training:  17%|█▋        | 1903/11082 [6:06:59<29:35:25, 11.61s/step, acc=0.4244, loss=2.5462, lr=9.45e-05]
Training:  17%|█▋        | 1904/11082 [6:07:10<29:27:20, 11.55s/step, acc=0.4244, loss=2.5462, lr=9.45e-05]
Training:  17%|█▋        | 1905/11082 [6:07:22<29:29:43, 11.57s/step, acc=0.4244, loss=2.5462, lr=9.45e-05]
Training:  17%|█▋        | 1906/11082 [6:07:33<29:14:59, 11.48s/step, acc=0.4244, loss=2.5462, lr=9.45e-05]
Training:  17%|█▋        | 1907/11082 [6:07:44<29:14:46, 11.48s/step, acc=0.4244, loss=2.5462, lr=9.45e-05]
Training:  17%|█▋        | 1908/11082 [6:07:56<29:24:28, 11.54s/step, acc=0.4244, loss=2.5462, lr=9.45e-05]
Training:  17%|█▋        | 1909/11082 [6:08:08<29:25:37, 11.55s/step, acc=0.4244, loss=2.5462, lr=9.45e-05]
Training:  17%|█▋        | 1910/11082 [6:08:19<29:27:27, 11.56s/step, acc=0.4244, loss=2.5462, lr=9.45e-05]
Training:  17%|█▋        | 1911/11082 [6:08:30<29:07:18, 11.43s/step, acc=0.4244, loss=2.5462, lr=9.45e-05]
Training:  17%|█▋        | 1912/11082 [6:08:42<29:31:10, 11.59s/step, acc=0.4244, loss=2.5462, lr=9.45e-05]
Training:  17%|█▋        | 1913/11082 [6:08:53<29:07:22, 11.43s/step, acc=0.4244, loss=2.5462, lr=9.45e-05]
Training:  17%|█▋        | 1914/11082 [6:09:05<29:33:34, 11.61s/step, acc=0.4244, loss=2.5462, lr=9.45e-05]
Training:  17%|█▋        | 1915/11082 [6:09:17<29:20:09, 11.52s/step, acc=0.4244, loss=2.5462, lr=9.45e-05]
Training:  17%|█▋        | 1916/11082 [6:09:29<29:51:13, 11.73s/step, acc=0.4244, loss=2.5462, lr=9.45e-05]
Training:  17%|█▋        | 1917/11082 [6:09:40<29:35:44, 11.63s/step, acc=0.4244, loss=2.5462, lr=9.45e-05]
Training:  17%|█▋        | 1918/11082 [6:09:52<29:30:10, 11.59s/step, acc=0.4244, loss=2.5462, lr=9.45e-05]
Training:  17%|█▋        | 1919/11082 [6:10:03<29:16:14, 11.50s/step, acc=0.4244, loss=2.5462, lr=9.45e-05]
Training:  17%|█▋        | 1920/11082 [6:10:15<29:30:23, 11.59s/step, acc=0.4244, loss=2.5462, lr=9.45e-05]
Training:  17%|█▋        | 1921/11082 [6:10:27<29:33:11, 11.61s/step, acc=0.4244, loss=2.5462, lr=9.45e-05]
Training:  17%|█▋        | 1922/11082 [6:10:38<29:42:44, 11.68s/step, acc=0.4244, loss=2.5462, lr=9.45e-05]
Training:  17%|█▋        | 1923/11082 [6:10:50<29:24:35, 11.56s/step, acc=0.4244, loss=2.5462, lr=9.45e-05]
Training:  17%|█▋        | 1924/11082 [6:11:02<30:07:40, 11.84s/step, acc=0.4244, loss=2.5462, lr=9.45e-05]
Training:  17%|█▋        | 1925/11082 [6:11:14<29:51:32, 11.74s/step, acc=0.4244, loss=2.5462, lr=9.45e-05]
Training:  17%|█▋        | 1926/11082 [6:11:25<29:46:37, 11.71s/step, acc=0.4244, loss=2.5462, lr=9.45e-05]
Training:  17%|█▋        | 1927/11082 [6:11:37<29:20:39, 11.54s/step, acc=0.4244, loss=2.5462, lr=9.45e-05]
Training:  17%|█▋        | 1928/11082 [6:11:48<29:28:07, 11.59s/step, acc=0.4244, loss=2.5462, lr=9.45e-05]
Training:  17%|█▋        | 1929/11082 [6:11:59<29:00:32, 11.41s/step, acc=0.4244, loss=2.5462, lr=9.45e-05]
Training:  17%|█▋        | 1930/11082 [6:12:11<29:28:54, 11.60s/step, acc=0.4244, loss=2.5462, lr=9.45e-05]
Training:  17%|█▋        | 1931/11082 [6:12:23<29:27:04, 11.59s/step, acc=0.4244, loss=2.5462, lr=9.45e-05]
Training:  17%|█▋        | 1932/11082 [6:12:34<29:18:43, 11.53s/step, acc=0.4244, loss=2.5462, lr=9.45e-05]
Training:  17%|█▋        | 1933/11082 [6:12:46<29:08:59, 11.47s/step, acc=0.4244, loss=2.5462, lr=9.45e-05]
Training:  17%|█▋        | 1934/11082 [6:12:57<29:27:52, 11.60s/step, acc=0.4244, loss=2.5462, lr=9.45e-05]
Training:  17%|█▋        | 1935/11082 [6:13:09<29:20:08, 11.55s/step, acc=0.4244, loss=2.5462, lr=9.45e-05]
Training:  17%|█▋        | 1936/11082 [6:13:21<29:30:13, 11.61s/step, acc=0.4244, loss=2.5462, lr=9.45e-05]
Training:  17%|█▋        | 1937/11082 [6:13:31<28:53:42, 11.37s/step, acc=0.4244, loss=2.5462, lr=9.45e-05]
Training:  17%|█▋        | 1938/11082 [6:13:43<29:11:53, 11.50s/step, acc=0.4244, loss=2.5462, lr=9.45e-05]
Training:  17%|█▋        | 1939/11082 [6:13:55<29:30:51, 11.62s/step, acc=0.4244, loss=2.5462, lr=9.45e-05]
Training:  18%|█▊        | 1940/11082 [6:14:07<29:19:50, 11.55s/step, acc=0.4244, loss=2.5462, lr=9.45e-05]
Training:  18%|█▊        | 1941/11082 [6:14:18<29:27:44, 11.60s/step, acc=0.4244, loss=2.5462, lr=9.45e-05]
Training:  18%|█▊        | 1942/11082 [6:14:30<29:10:28, 11.49s/step, acc=0.4244, loss=2.5462, lr=9.45e-05]
Training:  18%|█▊        | 1943/11082 [6:14:41<29:09:48, 11.49s/step, acc=0.4244, loss=2.5462, lr=9.45e-05]
Training:  18%|█▊        | 1944/11082 [6:14:52<28:55:33, 11.40s/step, acc=0.4244, loss=2.5462, lr=9.45e-05]
Training:  18%|█▊        | 1945/11082 [6:15:04<29:03:19, 11.45s/step, acc=0.4244, loss=2.5462, lr=9.45e-05]
Training:  18%|█▊        | 1946/11082 [6:15:15<29:16:39, 11.54s/step, acc=0.4244, loss=2.5462, lr=9.45e-05]
Training:  18%|█▊        | 1947/11082 [6:15:27<29:38:10, 11.68s/step, acc=0.4244, loss=2.5462, lr=9.45e-05]
Training:  18%|█▊        | 1948/11082 [6:15:39<29:31:05, 11.63s/step, acc=0.4244, loss=2.5462, lr=9.45e-05]
Training:  18%|█▊        | 1949/11082 [6:15:50<29:03:08, 11.45s/step, acc=0.4244, loss=2.5462, lr=9.45e-05]
Training:  18%|█▊        | 1950/11082 [6:16:02<29:11:41, 11.51s/step, acc=0.4244, loss=2.5462, lr=9.45e-05]
Training:  18%|█▊        | 1950/11082 [6:16:09<29:11:41, 11.51s/step, acc=0.4261, loss=2.5343, lr=9.42e-05]2026-06-18 21:09:37 | INFO | __main__ | epoch=0 step=1950 train_loss=2.534290 train_acc=0.426136 lr=9.420e-05 grad_norm=1.3047 acc_d0=0.7027 acc_d1=0.5399 acc_d2=0.4057 acc_d3=0.3037 acc_d4=0.2306 acc_d5=0.1831 acc_d6=0.1497 acc_d7=0.1280

Training:  18%|█▊        | 1951/11082 [6:16:13<29:12:05, 11.51s/step, acc=0.4261, loss=2.5343, lr=9.42e-05]
Training:  18%|█▊        | 1952/11082 [6:16:25<29:15:22, 11.54s/step, acc=0.4261, loss=2.5343, lr=9.42e-05]
Training:  18%|█▊        | 1953/11082 [6:16:37<29:26:17, 11.61s/step, acc=0.4261, loss=2.5343, lr=9.42e-05]
Training:  18%|█▊        | 1954/11082 [6:16:48<29:10:06, 11.50s/step, acc=0.4261, loss=2.5343, lr=9.42e-05]
Training:  18%|█▊        | 1955/11082 [6:17:00<29:19:54, 11.57s/step, acc=0.4261, loss=2.5343, lr=9.42e-05]
Training:  18%|█▊        | 1956/11082 [6:17:11<29:35:59, 11.68s/step, acc=0.4261, loss=2.5343, lr=9.42e-05]
Training:  18%|█▊        | 1957/11082 [6:17:23<29:19:22, 11.57s/step, acc=0.4261, loss=2.5343, lr=9.42e-05]
Training:  18%|█▊        | 1958/11082 [6:17:34<29:16:20, 11.55s/step, acc=0.4261, loss=2.5343, lr=9.42e-05]
Training:  18%|█▊        | 1959/11082 [6:17:46<29:08:09, 11.50s/step, acc=0.4261, loss=2.5343, lr=9.42e-05]
Training:  18%|█▊        | 1960/11082 [6:17:57<29:07:12, 11.49s/step, acc=0.4261, loss=2.5343, lr=9.42e-05]
Training:  18%|█▊        | 1961/11082 [6:18:09<29:21:29, 11.59s/step, acc=0.4261, loss=2.5343, lr=9.42e-05]
Training:  18%|█▊        | 1962/11082 [6:18:20<28:59:10, 11.44s/step, acc=0.4261, loss=2.5343, lr=9.42e-05]
Training:  18%|█▊        | 1963/11082 [6:18:32<29:34:06, 11.67s/step, acc=0.4261, loss=2.5343, lr=9.42e-05]
Training:  18%|█▊        | 1964/11082 [6:18:44<29:36:39, 11.69s/step, acc=0.4261, loss=2.5343, lr=9.42e-05]
Training:  18%|█▊        | 1965/11082 [6:18:56<29:34:53, 11.68s/step, acc=0.4261, loss=2.5343, lr=9.42e-05]
Training:  18%|█▊        | 1966/11082 [6:19:07<29:15:19, 11.55s/step, acc=0.4261, loss=2.5343, lr=9.42e-05]
Training:  18%|█▊        | 1967/11082 [6:19:19<29:19:16, 11.58s/step, acc=0.4261, loss=2.5343, lr=9.42e-05]
Training:  18%|█▊        | 1968/11082 [6:19:30<29:26:12, 11.63s/step, acc=0.4261, loss=2.5343, lr=9.42e-05]
Training:  18%|█▊        | 1969/11082 [6:19:42<29:32:30, 11.67s/step, acc=0.4261, loss=2.5343, lr=9.42e-05]
Training:  18%|█▊        | 1970/11082 [6:19:54<29:50:02, 11.79s/step, acc=0.4261, loss=2.5343, lr=9.42e-05]
Training:  18%|█▊        | 1971/11082 [6:20:05<29:28:14, 11.64s/step, acc=0.4261, loss=2.5343, lr=9.42e-05]
Training:  18%|█▊        | 1972/11082 [6:20:17<29:24:46, 11.62s/step, acc=0.4261, loss=2.5343, lr=9.42e-05]
Training:  18%|█▊        | 1973/11082 [6:20:29<29:46:01, 11.76s/step, acc=0.4261, loss=2.5343, lr=9.42e-05]
Training:  18%|█▊        | 1974/11082 [6:20:40<29:14:10, 11.56s/step, acc=0.4261, loss=2.5343, lr=9.42e-05]
Training:  18%|█▊        | 1975/11082 [6:20:52<29:13:04, 11.55s/step, acc=0.4261, loss=2.5343, lr=9.42e-05]
Training:  18%|█▊        | 1976/11082 [6:21:03<28:58:38, 11.46s/step, acc=0.4261, loss=2.5343, lr=9.42e-05]
Training:  18%|█▊        | 1977/11082 [6:21:15<29:34:38, 11.69s/step, acc=0.4261, loss=2.5343, lr=9.42e-05]
Training:  18%|█▊        | 1978/11082 [6:21:26<29:06:51, 11.51s/step, acc=0.4261, loss=2.5343, lr=9.42e-05]
Training:  18%|█▊        | 1979/11082 [6:21:38<29:06:39, 11.51s/step, acc=0.4261, loss=2.5343, lr=9.42e-05]
Training:  18%|█▊        | 1980/11082 [6:21:49<29:09:06, 11.53s/step, acc=0.4261, loss=2.5343, lr=9.42e-05]
Training:  18%|█▊        | 1981/11082 [6:22:01<29:18:15, 11.59s/step, acc=0.4261, loss=2.5343, lr=9.42e-05]
Training:  18%|█▊        | 1982/11082 [6:22:13<29:22:49, 11.62s/step, acc=0.4261, loss=2.5343, lr=9.42e-05]
Training:  18%|█▊        | 1983/11082 [6:22:24<29:11:05, 11.55s/step, acc=0.4261, loss=2.5343, lr=9.42e-05]
Training:  18%|█▊        | 1984/11082 [6:22:36<29:08:35, 11.53s/step, acc=0.4261, loss=2.5343, lr=9.42e-05]
Training:  18%|█▊        | 1985/11082 [6:22:47<29:03:32, 11.50s/step, acc=0.4261, loss=2.5343, lr=9.42e-05]
Training:  18%|█▊        | 1986/11082 [6:22:59<29:01:39, 11.49s/step, acc=0.4261, loss=2.5343, lr=9.42e-05]
Training:  18%|█▊        | 1987/11082 [6:23:10<29:19:43, 11.61s/step, acc=0.4261, loss=2.5343, lr=9.42e-05]
Training:  18%|█▊        | 1988/11082 [6:23:22<29:12:15, 11.56s/step, acc=0.4261, loss=2.5343, lr=9.42e-05]
Training:  18%|█▊        | 1989/11082 [6:23:34<29:19:32, 11.61s/step, acc=0.4261, loss=2.5343, lr=9.42e-05]
Training:  18%|█▊        | 1990/11082 [6:23:45<29:04:19, 11.51s/step, acc=0.4261, loss=2.5343, lr=9.42e-05]
Training:  18%|█▊        | 1991/11082 [6:23:57<29:22:12, 11.63s/step, acc=0.4261, loss=2.5343, lr=9.42e-05]
Training:  18%|█▊        | 1992/11082 [6:24:08<29:12:32, 11.57s/step, acc=0.4261, loss=2.5343, lr=9.42e-05]
Training:  18%|█▊        | 1993/11082 [6:24:20<29:21:47, 11.63s/step, acc=0.4261, loss=2.5343, lr=9.42e-05]
Training:  18%|█▊        | 1994/11082 [6:24:32<29:30:48, 11.69s/step, acc=0.4261, loss=2.5343, lr=9.42e-05]
Training:  18%|█▊        | 1995/11082 [6:24:43<29:11:07, 11.56s/step, acc=0.4261, loss=2.5343, lr=9.42e-05]
Training:  18%|█▊        | 1996/11082 [6:24:55<29:17:09, 11.60s/step, acc=0.4261, loss=2.5343, lr=9.42e-05]
Training:  18%|█▊        | 1997/11082 [6:25:06<29:17:47, 11.61s/step, acc=0.4261, loss=2.5343, lr=9.42e-05]
Training:  18%|█▊        | 1998/11082 [6:25:18<29:25:55, 11.66s/step, acc=0.4261, loss=2.5343, lr=9.42e-05]
Training:  18%|█▊        | 1999/11082 [6:25:30<29:12:12, 11.57s/step, acc=0.4261, loss=2.5343, lr=9.42e-05]
Training:  18%|█▊        | 2000/11082 [6:25:41<29:20:34, 11.63s/step, acc=0.4261, loss=2.5343, lr=9.42e-05]
Training:  18%|█▊        | 2000/11082 [6:25:48<29:20:34, 11.63s/step, acc=0.4274, loss=2.5268, lr=9.39e-05]2026-06-18 21:19:16 | INFO | __main__ | epoch=0 step=2000 train_loss=2.526793 train_acc=0.427409 lr=9.388e-05 grad_norm=1.3906 acc_d0=0.7112 acc_d1=0.5496 acc_d2=0.4164 acc_d3=0.3162 acc_d4=0.2419 acc_d5=0.1917 acc_d6=0.1564 acc_d7=0.1318

Training:  18%|█▊        | 2001/11082 [6:25:53<29:10:57, 11.57s/step, acc=0.4274, loss=2.5268, lr=9.39e-05]
Training:  18%|█▊        | 2002/11082 [6:26:04<29:01:55, 11.51s/step, acc=0.4274, loss=2.5268, lr=9.39e-05]
Training:  18%|█▊        | 2003/11082 [6:26:15<28:50:31, 11.44s/step, acc=0.4274, loss=2.5268, lr=9.39e-05]
Training:  18%|█▊        | 2004/11082 [6:26:27<29:04:53, 11.53s/step, acc=0.4274, loss=2.5268, lr=9.39e-05]
Training:  18%|█▊        | 2005/11082 [6:26:39<29:06:17, 11.54s/step, acc=0.4274, loss=2.5268, lr=9.39e-05]
Training:  18%|█▊        | 2006/11082 [6:26:50<29:14:49, 11.60s/step, acc=0.4274, loss=2.5268, lr=9.39e-05]
Training:  18%|█▊        | 2007/11082 [6:27:02<29:32:03, 11.72s/step, acc=0.4274, loss=2.5268, lr=9.39e-05]
Training:  18%|█▊        | 2008/11082 [6:27:13<28:58:29, 11.50s/step, acc=0.4274, loss=2.5268, lr=9.39e-05]
Training:  18%|█▊        | 2009/11082 [6:27:25<29:19:29, 11.64s/step, acc=0.4274, loss=2.5268, lr=9.39e-05]
Training:  18%|█▊        | 2010/11082 [6:27:37<29:16:07, 11.61s/step, acc=0.4274, loss=2.5268, lr=9.39e-05]
Training:  18%|█▊        | 2011/11082 [6:27:48<29:07:00, 11.56s/step, acc=0.4274, loss=2.5268, lr=9.39e-05]
Training:  18%|█▊        | 2012/11082 [6:28:00<28:50:04, 11.44s/step, acc=0.4274, loss=2.5268, lr=9.39e-05]
Training:  18%|█▊        | 2013/11082 [6:28:12<29:22:07, 11.66s/step, acc=0.4274, loss=2.5268, lr=9.39e-05]
Training:  18%|█▊        | 2014/11082 [6:28:23<28:51:49, 11.46s/step, acc=0.4274, loss=2.5268, lr=9.39e-05]
Training:  18%|█▊        | 2015/11082 [6:28:34<29:01:36, 11.52s/step, acc=0.4274, loss=2.5268, lr=9.39e-05]
Training:  18%|█▊        | 2016/11082 [6:28:46<29:26:38, 11.69s/step, acc=0.4274, loss=2.5268, lr=9.39e-05]
Training:  18%|█▊        | 2017/11082 [6:28:58<29:18:18, 11.64s/step, acc=0.4274, loss=2.5268, lr=9.39e-05]
Training:  18%|█▊        | 2018/11082 [6:29:09<29:08:01, 11.57s/step, acc=0.4274, loss=2.5268, lr=9.39e-05]
Training:  18%|█▊        | 2019/11082 [6:29:21<28:54:53, 11.49s/step, acc=0.4274, loss=2.5268, lr=9.39e-05]
Training:  18%|█▊        | 2020/11082 [6:29:32<28:58:18, 11.51s/step, acc=0.4274, loss=2.5268, lr=9.39e-05]
Training:  18%|█▊        | 2021/11082 [6:29:44<28:50:21, 11.46s/step, acc=0.4274, loss=2.5268, lr=9.39e-05]
Training:  18%|█▊        | 2022/11082 [6:29:55<29:04:04, 11.55s/step, acc=0.4274, loss=2.5268, lr=9.39e-05]
Training:  18%|█▊        | 2023/11082 [6:30:07<29:04:14, 11.55s/step, acc=0.4274, loss=2.5268, lr=9.39e-05]
Training:  18%|█▊        | 2024/11082 [6:30:18<28:54:26, 11.49s/step, acc=0.4274, loss=2.5268, lr=9.39e-05]
Training:  18%|█▊        | 2025/11082 [6:30:30<28:58:56, 11.52s/step, acc=0.4274, loss=2.5268, lr=9.39e-05]
Training:  18%|█▊        | 2026/11082 [6:30:42<29:04:46, 11.56s/step, acc=0.4274, loss=2.5268, lr=9.39e-05]
Training:  18%|█▊        | 2027/11082 [6:30:53<28:54:26, 11.49s/step, acc=0.4274, loss=2.5268, lr=9.39e-05]
Training:  18%|█▊        | 2028/11082 [6:31:04<28:59:30, 11.53s/step, acc=0.4274, loss=2.5268, lr=9.39e-05]
Training:  18%|█▊        | 2029/11082 [6:31:16<28:45:59, 11.44s/step, acc=0.4274, loss=2.5268, lr=9.39e-05]
Training:  18%|█▊        | 2030/11082 [6:31:28<29:06:46, 11.58s/step, acc=0.4274, loss=2.5268, lr=9.39e-05]
Training:  18%|█▊        | 2031/11082 [6:31:39<28:51:53, 11.48s/step, acc=0.4274, loss=2.5268, lr=9.39e-05]
Training:  18%|█▊        | 2032/11082 [6:31:51<29:10:48, 11.61s/step, acc=0.4274, loss=2.5268, lr=9.39e-05]
Training:  18%|█▊        | 2033/11082 [6:32:02<29:06:38, 11.58s/step, acc=0.4274, loss=2.5268, lr=9.39e-05]
Training:  18%|█▊        | 2034/11082 [6:32:14<29:10:47, 11.61s/step, acc=0.4274, loss=2.5268, lr=9.39e-05]
Training:  18%|█▊        | 2035/11082 [6:32:25<28:52:23, 11.49s/step, acc=0.4274, loss=2.5268, lr=9.39e-05]
Training:  18%|█▊        | 2036/11082 [6:32:37<28:53:36, 11.50s/step, acc=0.4274, loss=2.5268, lr=9.39e-05]
Training:  18%|█▊        | 2037/11082 [6:32:48<28:55:37, 11.51s/step, acc=0.4274, loss=2.5268, lr=9.39e-05]
Training:  18%|█▊        | 2038/11082 [6:33:00<29:16:57, 11.66s/step, acc=0.4274, loss=2.5268, lr=9.39e-05]
Training:  18%|█▊        | 2039/11082 [6:33:12<29:12:46, 11.63s/step, acc=0.4274, loss=2.5268, lr=9.39e-05]
Training:  18%|█▊        | 2040/11082 [6:33:24<29:21:34, 11.69s/step, acc=0.4274, loss=2.5268, lr=9.39e-05]
Training:  18%|█▊        | 2041/11082 [6:33:34<28:42:20, 11.43s/step, acc=0.4274, loss=2.5268, lr=9.39e-05]
Training:  18%|█▊        | 2042/11082 [6:33:47<29:31:23, 11.76s/step, acc=0.4274, loss=2.5268, lr=9.39e-05]
Training:  18%|█▊        | 2043/11082 [6:33:58<29:12:02, 11.63s/step, acc=0.4274, loss=2.5268, lr=9.39e-05]
Training:  18%|█▊        | 2044/11082 [6:34:10<29:12:29, 11.63s/step, acc=0.4274, loss=2.5268, lr=9.39e-05]
Training:  18%|█▊        | 2045/11082 [6:34:22<29:17:13, 11.67s/step, acc=0.4274, loss=2.5268, lr=9.39e-05]
Training:  18%|█▊        | 2046/11082 [6:34:32<28:37:57, 11.41s/step, acc=0.4274, loss=2.5268, lr=9.39e-05]
Training:  18%|█▊        | 2047/11082 [6:34:44<28:49:56, 11.49s/step, acc=0.4274, loss=2.5268, lr=9.39e-05]
Training:  18%|█▊        | 2048/11082 [6:34:56<29:15:42, 11.66s/step, acc=0.4274, loss=2.5268, lr=9.39e-05]
Training:  18%|█▊        | 2049/11082 [6:35:08<28:59:46, 11.56s/step, acc=0.4274, loss=2.5268, lr=9.39e-05]
Training:  18%|█▊        | 2050/11082 [6:35:19<28:48:47, 11.48s/step, acc=0.4274, loss=2.5268, lr=9.39e-05]
Training:  18%|█▊        | 2050/11082 [6:35:26<28:48:47, 11.48s/step, acc=0.4313, loss=2.4960, lr=9.35e-05]2026-06-18 21:28:54 | INFO | __main__ | epoch=0 step=2050 train_loss=2.495962 train_acc=0.431310 lr=9.355e-05 grad_norm=1.2109 acc_d0=0.6992 acc_d1=0.5404 acc_d2=0.4096 acc_d3=0.3099 acc_d4=0.2373 acc_d5=0.1888 acc_d6=0.1551 acc_d7=0.1298

Training:  19%|█▊        | 2051/11082 [6:35:30<28:52:01, 11.51s/step, acc=0.4313, loss=2.4960, lr=9.35e-05]
Training:  19%|█▊        | 2052/11082 [6:35:42<28:59:10, 11.56s/step, acc=0.4313, loss=2.4960, lr=9.35e-05]
Training:  19%|█▊        | 2053/11082 [6:35:53<28:48:51, 11.49s/step, acc=0.4313, loss=2.4960, lr=9.35e-05]
Training:  19%|█▊        | 2054/11082 [6:36:05<29:05:39, 11.60s/step, acc=0.4313, loss=2.4960, lr=9.35e-05]
Training:  19%|█▊        | 2055/11082 [6:36:16<28:41:54, 11.45s/step, acc=0.4313, loss=2.4960, lr=9.35e-05]
Training:  19%|█▊        | 2056/11082 [6:36:28<28:57:23, 11.55s/step, acc=0.4313, loss=2.4960, lr=9.35e-05]
Training:  19%|█▊        | 2057/11082 [6:36:40<28:58:38, 11.56s/step, acc=0.4313, loss=2.4960, lr=9.35e-05]
Training:  19%|█▊        | 2058/11082 [6:36:51<29:04:44, 11.60s/step, acc=0.4313, loss=2.4960, lr=9.35e-05]
Training:  19%|█▊        | 2059/11082 [6:37:03<29:13:29, 11.66s/step, acc=0.4313, loss=2.4960, lr=9.35e-05]
Training:  19%|█▊        | 2060/11082 [6:37:15<28:58:03, 11.56s/step, acc=0.4313, loss=2.4960, lr=9.35e-05]
Training:  19%|█▊        | 2061/11082 [6:37:26<28:55:12, 11.54s/step, acc=0.4313, loss=2.4960, lr=9.35e-05]
Training:  19%|█▊        | 2062/11082 [6:37:38<28:59:59, 11.57s/step, acc=0.4313, loss=2.4960, lr=9.35e-05]
Training:  19%|█▊        | 2063/11082 [6:37:50<29:10:36, 11.65s/step, acc=0.4313, loss=2.4960, lr=9.35e-05]
Training:  19%|█▊        | 2064/11082 [6:38:01<28:59:08, 11.57s/step, acc=0.4313, loss=2.4960, lr=9.35e-05]
Training:  19%|█▊        | 2065/11082 [6:38:12<28:41:08, 11.45s/step, acc=0.4313, loss=2.4960, lr=9.35e-05]
Training:  19%|█▊        | 2066/11082 [6:38:24<28:56:03, 11.55s/step, acc=0.4313, loss=2.4960, lr=9.35e-05]
Training:  19%|█▊        | 2067/11082 [6:38:36<29:13:16, 11.67s/step, acc=0.4313, loss=2.4960, lr=9.35e-05]
Training:  19%|█▊        | 2068/11082 [6:38:47<28:34:14, 11.41s/step, acc=0.4313, loss=2.4960, lr=9.35e-05]
Training:  19%|█▊        | 2069/11082 [6:38:58<28:49:16, 11.51s/step, acc=0.4313, loss=2.4960, lr=9.35e-05]
Training:  19%|█▊        | 2070/11082 [6:39:10<29:01:11, 11.59s/step, acc=0.4313, loss=2.4960, lr=9.35e-05]
Training:  19%|█▊        | 2071/11082 [6:39:21<28:49:40, 11.52s/step, acc=0.4313, loss=2.4960, lr=9.35e-05]
Training:  19%|█▊        | 2072/11082 [6:39:33<28:37:47, 11.44s/step, acc=0.4313, loss=2.4960, lr=9.35e-05]
Training:  19%|█▊        | 2073/11082 [6:39:45<29:05:21, 11.62s/step, acc=0.4313, loss=2.4960, lr=9.35e-05]
Training:  19%|█▊        | 2074/11082 [6:39:57<29:13:26, 11.68s/step, acc=0.4313, loss=2.4960, lr=9.35e-05]
Training:  19%|█▊        | 2075/11082 [6:40:08<29:01:20, 11.60s/step, acc=0.4313, loss=2.4960, lr=9.35e-05]
Training:  19%|█▊        | 2076/11082 [6:40:20<29:30:36, 11.80s/step, acc=0.4313, loss=2.4960, lr=9.35e-05]
Training:  19%|█▊        | 2077/11082 [6:40:32<29:18:45, 11.72s/step, acc=0.4313, loss=2.4960, lr=9.35e-05]
Training:  19%|█▉        | 2078/11082 [6:40:44<29:27:34, 11.78s/step, acc=0.4313, loss=2.4960, lr=9.35e-05]
Training:  19%|█▉        | 2079/11082 [6:40:55<29:02:09, 11.61s/step, acc=0.4313, loss=2.4960, lr=9.35e-05]
Training:  19%|█▉        | 2080/11082 [6:41:07<29:12:49, 11.68s/step, acc=0.4313, loss=2.4960, lr=9.35e-05]
Training:  19%|█▉        | 2081/11082 [6:41:18<28:52:05, 11.55s/step, acc=0.4313, loss=2.4960, lr=9.35e-05]
Training:  19%|█▉        | 2082/11082 [6:41:30<29:04:51, 11.63s/step, acc=0.4313, loss=2.4960, lr=9.35e-05]
Training:  19%|█▉        | 2083/11082 [6:41:41<29:01:11, 11.61s/step, acc=0.4313, loss=2.4960, lr=9.35e-05]
Training:  19%|█▉        | 2084/11082 [6:41:53<29:02:26, 11.62s/step, acc=0.4313, loss=2.4960, lr=9.35e-05]
Training:  19%|█▉        | 2085/11082 [6:42:05<28:59:34, 11.60s/step, acc=0.4313, loss=2.4960, lr=9.35e-05]
Training:  19%|█▉        | 2086/11082 [6:42:16<28:43:57, 11.50s/step, acc=0.4313, loss=2.4960, lr=9.35e-05]
Training:  19%|█▉        | 2087/11082 [6:42:28<29:25:17, 11.78s/step, acc=0.4313, loss=2.4960, lr=9.35e-05]
Training:  19%|█▉        | 2088/11082 [6:42:39<28:47:00, 11.52s/step, acc=0.4313, loss=2.4960, lr=9.35e-05]
Training:  19%|█▉        | 2089/11082 [6:42:51<28:54:48, 11.57s/step, acc=0.4313, loss=2.4960, lr=9.35e-05]
Training:  19%|█▉        | 2090/11082 [6:43:03<29:04:36, 11.64s/step, acc=0.4313, loss=2.4960, lr=9.35e-05]
Training:  19%|█▉        | 2091/11082 [6:43:14<28:38:52, 11.47s/step, acc=0.4313, loss=2.4960, lr=9.35e-05]
Training:  19%|█▉        | 2092/11082 [6:43:26<28:55:55, 11.59s/step, acc=0.4313, loss=2.4960, lr=9.35e-05]
Training:  19%|█▉        | 2093/11082 [6:43:37<28:51:50, 11.56s/step, acc=0.4313, loss=2.4960, lr=9.35e-05]
Training:  19%|█▉        | 2094/11082 [6:43:49<28:55:49, 11.59s/step, acc=0.4313, loss=2.4960, lr=9.35e-05]
Training:  19%|█▉        | 2095/11082 [6:44:00<28:45:39, 11.52s/step, acc=0.4313, loss=2.4960, lr=9.35e-05]
Training:  19%|█▉        | 2096/11082 [6:44:11<28:34:42, 11.45s/step, acc=0.4313, loss=2.4960, lr=9.35e-05]
Training:  19%|█▉        | 2097/11082 [6:44:23<28:48:32, 11.54s/step, acc=0.4313, loss=2.4960, lr=9.35e-05]
Training:  19%|█▉        | 2098/11082 [6:44:35<28:47:11, 11.54s/step, acc=0.4313, loss=2.4960, lr=9.35e-05]
Training:  19%|█▉        | 2099/11082 [6:44:46<28:23:48, 11.38s/step, acc=0.4313, loss=2.4960, lr=9.35e-05]
Training:  19%|█▉        | 2100/11082 [6:44:58<28:56:03, 11.60s/step, acc=0.4313, loss=2.4960, lr=9.35e-05]
Training:  19%|█▉        | 2100/11082 [6:45:05<28:56:03, 11.60s/step, acc=0.4329, loss=2.4848, lr=9.32e-05]2026-06-18 21:38:33 | INFO | __main__ | epoch=0 step=2100 train_loss=2.484812 train_acc=0.432911 lr=9.321e-05 grad_norm=1.1719 acc_d0=0.6967 acc_d1=0.5340 acc_d2=0.4036 acc_d3=0.3061 acc_d4=0.2374 acc_d5=0.1893 acc_d6=0.1552 acc_d7=0.1305

Training:  19%|█▉        | 2101/11082 [6:45:09<28:54:38, 11.59s/step, acc=0.4329, loss=2.4848, lr=9.32e-05]
Training:  19%|█▉        | 2102/11082 [6:45:20<28:23:10, 11.38s/step, acc=0.4329, loss=2.4848, lr=9.32e-05]
Training:  19%|█▉        | 2103/11082 [6:45:32<28:45:26, 11.53s/step, acc=0.4329, loss=2.4848, lr=9.32e-05]
Training:  19%|█▉        | 2104/11082 [6:45:44<28:39:47, 11.49s/step, acc=0.4329, loss=2.4848, lr=9.32e-05]
Training:  19%|█▉        | 2105/11082 [6:45:55<28:48:09, 11.55s/step, acc=0.4329, loss=2.4848, lr=9.32e-05]
Training:  19%|█▉        | 2106/11082 [6:46:07<28:35:16, 11.47s/step, acc=0.4329, loss=2.4848, lr=9.32e-05]
Training:  19%|█▉        | 2107/11082 [6:46:19<29:07:59, 11.69s/step, acc=0.4329, loss=2.4848, lr=9.32e-05]
Training:  19%|█▉        | 2108/11082 [6:46:30<29:06:26, 11.68s/step, acc=0.4329, loss=2.4848, lr=9.32e-05]
Training:  19%|█▉        | 2109/11082 [6:46:42<28:52:29, 11.58s/step, acc=0.4329, loss=2.4848, lr=9.32e-05]
Training:  19%|█▉        | 2110/11082 [6:46:53<28:54:23, 11.60s/step, acc=0.4329, loss=2.4848, lr=9.32e-05]
Training:  19%|█▉        | 2111/11082 [6:47:05<28:44:45, 11.54s/step, acc=0.4329, loss=2.4848, lr=9.32e-05]
Training:  19%|█▉        | 2112/11082 [6:47:17<28:52:54, 11.59s/step, acc=0.4329, loss=2.4848, lr=9.32e-05]
Training:  19%|█▉        | 2113/11082 [6:47:28<28:45:05, 11.54s/step, acc=0.4329, loss=2.4848, lr=9.32e-05]
Training:  19%|█▉        | 2114/11082 [6:47:40<28:48:16, 11.56s/step, acc=0.4329, loss=2.4848, lr=9.32e-05]
Training:  19%|█▉        | 2115/11082 [6:47:51<28:36:09, 11.48s/step, acc=0.4329, loss=2.4848, lr=9.32e-05]
Training:  19%|█▉        | 2116/11082 [6:48:03<28:55:01, 11.61s/step, acc=0.4329, loss=2.4848, lr=9.32e-05]
Training:  19%|█▉        | 2117/11082 [6:48:14<28:52:14, 11.59s/step, acc=0.4329, loss=2.4848, lr=9.32e-05]
Training:  19%|█▉        | 2118/11082 [6:48:26<28:41:46, 11.52s/step, acc=0.4329, loss=2.4848, lr=9.32e-05]
Training:  19%|█▉        | 2119/11082 [6:48:37<28:54:07, 11.61s/step, acc=0.4329, loss=2.4848, lr=9.32e-05]
Training:  19%|█▉        | 2120/11082 [6:48:49<28:45:58, 11.56s/step, acc=0.4329, loss=2.4848, lr=9.32e-05]
Training:  19%|█▉        | 2121/11082 [6:49:00<28:42:34, 11.53s/step, acc=0.4329, loss=2.4848, lr=9.32e-05]
Training:  19%|█▉        | 2122/11082 [6:49:12<28:22:50, 11.40s/step, acc=0.4329, loss=2.4848, lr=9.32e-05]
Training:  19%|█▉        | 2123/11082 [6:49:24<28:50:48, 11.59s/step, acc=0.4329, loss=2.4848, lr=9.32e-05]
Training:  19%|█▉        | 2124/11082 [6:49:35<28:37:16, 11.50s/step, acc=0.4329, loss=2.4848, lr=9.32e-05]
Training:  19%|█▉        | 2125/11082 [6:49:46<28:36:36, 11.50s/step, acc=0.4329, loss=2.4848, lr=9.32e-05]
Training:  19%|█▉        | 2126/11082 [6:49:58<29:00:43, 11.66s/step, acc=0.4329, loss=2.4848, lr=9.32e-05]
Training:  19%|█▉        | 2127/11082 [6:50:10<28:39:12, 11.52s/step, acc=0.4329, loss=2.4848, lr=9.32e-05]
Training:  19%|█▉        | 2128/11082 [6:50:21<28:46:53, 11.57s/step, acc=0.4329, loss=2.4848, lr=9.32e-05]
Training:  19%|█▉        | 2129/11082 [6:50:33<28:46:05, 11.57s/step, acc=0.4329, loss=2.4848, lr=9.32e-05]
Training:  19%|█▉        | 2130/11082 [6:50:44<28:36:16, 11.50s/step, acc=0.4329, loss=2.4848, lr=9.32e-05]
Training:  19%|█▉        | 2131/11082 [6:50:55<28:11:58, 11.34s/step, acc=0.4329, loss=2.4848, lr=9.32e-05]
Training:  19%|█▉        | 2132/11082 [6:51:07<28:42:02, 11.54s/step, acc=0.4329, loss=2.4848, lr=9.32e-05]
Training:  19%|█▉        | 2133/11082 [6:51:19<28:39:47, 11.53s/step, acc=0.4329, loss=2.4848, lr=9.32e-05]
Training:  19%|█▉        | 2134/11082 [6:51:31<28:59:17, 11.66s/step, acc=0.4329, loss=2.4848, lr=9.32e-05]
Training:  19%|█▉        | 2135/11082 [6:51:42<28:41:35, 11.55s/step, acc=0.4329, loss=2.4848, lr=9.32e-05]
Training:  19%|█▉        | 2136/11082 [6:51:54<28:56:59, 11.65s/step, acc=0.4329, loss=2.4848, lr=9.32e-05]
Training:  19%|█▉        | 2137/11082 [6:52:05<28:36:29, 11.51s/step, acc=0.4329, loss=2.4848, lr=9.32e-05]
Training:  19%|█▉        | 2138/11082 [6:52:16<28:26:53, 11.45s/step, acc=0.4329, loss=2.4848, lr=9.32e-05]
Training:  19%|█▉        | 2139/11082 [6:52:28<28:30:45, 11.48s/step, acc=0.4329, loss=2.4848, lr=9.32e-05]
Training:  19%|█▉        | 2140/11082 [6:52:40<28:58:46, 11.67s/step, acc=0.4329, loss=2.4848, lr=9.32e-05]
Training:  19%|█▉        | 2141/11082 [6:52:51<28:35:54, 11.51s/step, acc=0.4329, loss=2.4848, lr=9.32e-05]
Training:  19%|█▉        | 2142/11082 [6:53:03<28:40:52, 11.55s/step, acc=0.4329, loss=2.4848, lr=9.32e-05]
Training:  19%|█▉        | 2143/11082 [6:53:14<28:49:22, 11.61s/step, acc=0.4329, loss=2.4848, lr=9.32e-05]
Training:  19%|█▉        | 2144/11082 [6:53:26<28:37:41, 11.53s/step, acc=0.4329, loss=2.4848, lr=9.32e-05]
Training:  19%|█▉        | 2145/11082 [6:53:37<28:28:58, 11.47s/step, acc=0.4329, loss=2.4848, lr=9.32e-05]
Training:  19%|█▉        | 2146/11082 [6:53:49<28:49:35, 11.61s/step, acc=0.4329, loss=2.4848, lr=9.32e-05]
Training:  19%|█▉        | 2147/11082 [6:54:00<28:31:51, 11.50s/step, acc=0.4329, loss=2.4848, lr=9.32e-05]
Training:  19%|█▉        | 2148/11082 [6:54:12<28:45:49, 11.59s/step, acc=0.4329, loss=2.4848, lr=9.32e-05]
Training:  19%|█▉        | 2149/11082 [6:54:23<28:34:44, 11.52s/step, acc=0.4329, loss=2.4848, lr=9.32e-05]
Training:  19%|█▉        | 2150/11082 [6:54:35<28:32:56, 11.51s/step, acc=0.4329, loss=2.4848, lr=9.32e-05]
Training:  19%|█▉        | 2150/11082 [6:54:42<28:32:56, 11.51s/step, acc=0.4360, loss=2.4695, lr=9.29e-05]2026-06-18 21:48:10 | INFO | __main__ | epoch=0 step=2150 train_loss=2.469546 train_acc=0.435956 lr=9.286e-05 grad_norm=1.8125 acc_d0=0.6978 acc_d1=0.5389 acc_d2=0.4082 acc_d3=0.3098 acc_d4=0.2384 acc_d5=0.1910 acc_d6=0.1562 acc_d7=0.1306

Training:  19%|█▉        | 2151/11082 [6:54:47<28:36:54, 11.53s/step, acc=0.4360, loss=2.4695, lr=9.29e-05]
Training:  19%|█▉        | 2152/11082 [6:54:58<28:48:51, 11.62s/step, acc=0.4360, loss=2.4695, lr=9.29e-05]
Training:  19%|█▉        | 2153/11082 [6:55:10<28:45:16, 11.59s/step, acc=0.4360, loss=2.4695, lr=9.29e-05]
Training:  19%|█▉        | 2154/11082 [6:55:21<28:37:42, 11.54s/step, acc=0.4360, loss=2.4695, lr=9.29e-05]
Training:  19%|█▉        | 2155/11082 [6:55:33<28:45:25, 11.60s/step, acc=0.4360, loss=2.4695, lr=9.29e-05]
Training:  19%|█▉        | 2156/11082 [6:55:44<28:35:32, 11.53s/step, acc=0.4360, loss=2.4695, lr=9.29e-05]
Training:  19%|█▉        | 2157/11082 [6:55:56<28:54:56, 11.66s/step, acc=0.4360, loss=2.4695, lr=9.29e-05]
Training:  19%|█▉        | 2158/11082 [6:56:08<28:34:12, 11.53s/step, acc=0.4360, loss=2.4695, lr=9.29e-05]
Training:  19%|█▉        | 2159/11082 [6:56:19<28:40:20, 11.57s/step, acc=0.4360, loss=2.4695, lr=9.29e-05]
Training:  19%|█▉        | 2160/11082 [6:56:31<28:29:46, 11.50s/step, acc=0.4360, loss=2.4695, lr=9.29e-05]
Training:  20%|█▉        | 2161/11082 [6:56:42<28:36:12, 11.54s/step, acc=0.4360, loss=2.4695, lr=9.29e-05]
Training:  20%|█▉        | 2162/11082 [6:56:54<28:30:12, 11.50s/step, acc=0.4360, loss=2.4695, lr=9.29e-05]
Training:  20%|█▉        | 2163/11082 [6:57:06<28:52:22, 11.65s/step, acc=0.4360, loss=2.4695, lr=9.29e-05]
Training:  20%|█▉        | 2164/11082 [6:57:17<28:35:04, 11.54s/step, acc=0.4360, loss=2.4695, lr=9.29e-05]
Training:  20%|█▉        | 2165/11082 [6:57:29<28:49:31, 11.64s/step, acc=0.4360, loss=2.4695, lr=9.29e-05]
Training:  20%|█▉        | 2166/11082 [6:57:40<28:48:13, 11.63s/step, acc=0.4360, loss=2.4695, lr=9.29e-05]
Training:  20%|█▉        | 2167/11082 [6:57:52<28:43:00, 11.60s/step, acc=0.4360, loss=2.4695, lr=9.29e-05]
Training:  20%|█▉        | 2168/11082 [6:58:04<29:02:49, 11.73s/step, acc=0.4360, loss=2.4695, lr=9.29e-05]
Training:  20%|█▉        | 2169/11082 [6:58:16<29:00:49, 11.72s/step, acc=0.4360, loss=2.4695, lr=9.29e-05]
Training:  20%|█▉        | 2170/11082 [6:58:27<28:48:13, 11.64s/step, acc=0.4360, loss=2.4695, lr=9.29e-05]
Training:  20%|█▉        | 2171/11082 [6:58:38<28:23:36, 11.47s/step, acc=0.4360, loss=2.4695, lr=9.29e-05]
Training:  20%|█▉        | 2172/11082 [6:58:50<28:37:43, 11.57s/step, acc=0.4360, loss=2.4695, lr=9.29e-05]
Training:  20%|█▉        | 2173/11082 [6:59:02<28:44:22, 11.61s/step, acc=0.4360, loss=2.4695, lr=9.29e-05]
Training:  20%|█▉        | 2174/11082 [6:59:13<28:25:30, 11.49s/step, acc=0.4360, loss=2.4695, lr=9.29e-05]
Training:  20%|█▉        | 2175/11082 [6:59:25<28:41:06, 11.59s/step, acc=0.4360, loss=2.4695, lr=9.29e-05]
Training:  20%|█▉        | 2176/11082 [6:59:36<28:25:51, 11.49s/step, acc=0.4360, loss=2.4695, lr=9.29e-05]
Training:  20%|█▉        | 2177/11082 [6:59:47<28:17:49, 11.44s/step, acc=0.4360, loss=2.4695, lr=9.29e-05]
Training:  20%|█▉        | 2178/11082 [6:59:59<28:48:11, 11.65s/step, acc=0.4360, loss=2.4695, lr=9.29e-05]
Training:  20%|█▉        | 2179/11082 [7:00:11<28:42:41, 11.61s/step, acc=0.4360, loss=2.4695, lr=9.29e-05]
Training:  20%|█▉        | 2180/11082 [7:00:21<27:52:09, 11.27s/step, acc=0.4360, loss=2.4695, lr=9.29e-05]
Training:  20%|█▉        | 2181/11082 [7:00:34<28:29:09, 11.52s/step, acc=0.4360, loss=2.4695, lr=9.29e-05]
Training:  20%|█▉        | 2182/11082 [7:00:46<28:51:10, 11.67s/step, acc=0.4360, loss=2.4695, lr=9.29e-05]
Training:  20%|█▉        | 2183/11082 [7:00:57<28:47:40, 11.65s/step, acc=0.4360, loss=2.4695, lr=9.29e-05]
Training:  20%|█▉        | 2184/11082 [7:01:08<28:32:09, 11.55s/step, acc=0.4360, loss=2.4695, lr=9.29e-05]
Training:  20%|█▉        | 2185/11082 [7:01:20<28:32:59, 11.55s/step, acc=0.4360, loss=2.4695, lr=9.29e-05]
Training:  20%|█▉        | 2186/11082 [7:01:32<28:43:32, 11.62s/step, acc=0.4360, loss=2.4695, lr=9.29e-05]
Training:  20%|█▉        | 2187/11082 [7:01:43<28:31:38, 11.55s/step, acc=0.4360, loss=2.4695, lr=9.29e-05]
Training:  20%|█▉        | 2188/11082 [7:01:55<28:34:10, 11.56s/step, acc=0.4360, loss=2.4695, lr=9.29e-05]
Training:  20%|█▉        | 2189/11082 [7:02:06<28:27:07, 11.52s/step, acc=0.4360, loss=2.4695, lr=9.29e-05]
Training:  20%|█▉        | 2190/11082 [7:02:18<28:40:30, 11.61s/step, acc=0.4360, loss=2.4695, lr=9.29e-05]
Training:  20%|█▉        | 2191/11082 [7:02:29<28:33:22, 11.56s/step, acc=0.4360, loss=2.4695, lr=9.29e-05]
Training:  20%|█▉        | 2192/11082 [7:02:41<28:42:25, 11.62s/step, acc=0.4360, loss=2.4695, lr=9.29e-05]
Training:  20%|█▉        | 2193/11082 [7:02:52<28:20:49, 11.48s/step, acc=0.4360, loss=2.4695, lr=9.29e-05]
Training:  20%|█▉        | 2194/11082 [7:03:04<28:38:11, 11.60s/step, acc=0.4360, loss=2.4695, lr=9.29e-05]
Training:  20%|█▉        | 2195/11082 [7:03:15<28:20:08, 11.48s/step, acc=0.4360, loss=2.4695, lr=9.29e-05]
Training:  20%|█▉        | 2196/11082 [7:03:27<28:27:21, 11.53s/step, acc=0.4360, loss=2.4695, lr=9.29e-05]
Training:  20%|█▉        | 2197/11082 [7:03:39<28:52:29, 11.70s/step, acc=0.4360, loss=2.4695, lr=9.29e-05]
Training:  20%|█▉        | 2198/11082 [7:03:50<28:29:03, 11.54s/step, acc=0.4360, loss=2.4695, lr=9.29e-05]
Training:  20%|█▉        | 2199/11082 [7:04:02<28:22:09, 11.50s/step, acc=0.4360, loss=2.4695, lr=9.29e-05]
Training:  20%|█▉        | 2200/11082 [7:04:14<28:35:03, 11.59s/step, acc=0.4360, loss=2.4695, lr=9.29e-05]
Training:  20%|█▉        | 2200/11082 [7:04:21<28:35:03, 11.59s/step, acc=0.4339, loss=2.4769, lr=9.25e-05]2026-06-18 21:57:48 | INFO | __main__ | epoch=0 step=2200 train_loss=2.476931 train_acc=0.433862 lr=9.251e-05 grad_norm=1.5469 acc_d0=0.7092 acc_d1=0.5427 acc_d2=0.4087 acc_d3=0.3060 acc_d4=0.2357 acc_d5=0.1843 acc_d6=0.1506 acc_d7=0.1283

Training:  20%|█▉        | 2201/11082 [7:04:25<28:10:42, 11.42s/step, acc=0.4339, loss=2.4769, lr=9.25e-05]
Training:  20%|█▉        | 2202/11082 [7:04:36<28:25:31, 11.52s/step, acc=0.4339, loss=2.4769, lr=9.25e-05]
Training:  20%|█▉        | 2203/11082 [7:04:48<28:13:16, 11.44s/step, acc=0.4339, loss=2.4769, lr=9.25e-05]
Training:  20%|█▉        | 2204/11082 [7:04:59<28:28:10, 11.54s/step, acc=0.4339, loss=2.4769, lr=9.25e-05]
Training:  20%|█▉        | 2205/11082 [7:05:11<28:34:22, 11.59s/step, acc=0.4339, loss=2.4769, lr=9.25e-05]
Training:  20%|█▉        | 2206/11082 [7:05:23<28:31:16, 11.57s/step, acc=0.4339, loss=2.4769, lr=9.25e-05]
Training:  20%|█▉        | 2207/11082 [7:05:34<28:43:14, 11.65s/step, acc=0.4339, loss=2.4769, lr=9.25e-05]
Training:  20%|█▉        | 2208/11082 [7:05:46<28:25:14, 11.53s/step, acc=0.4339, loss=2.4769, lr=9.25e-05]
Training:  20%|█▉        | 2209/11082 [7:05:58<28:38:20, 11.62s/step, acc=0.4339, loss=2.4769, lr=9.25e-05]
Training:  20%|█▉        | 2210/11082 [7:06:08<28:04:16, 11.39s/step, acc=0.4339, loss=2.4769, lr=9.25e-05]
Training:  20%|█▉        | 2211/11082 [7:06:21<28:43:15, 11.66s/step, acc=0.4339, loss=2.4769, lr=9.25e-05]
Training:  20%|█▉        | 2212/11082 [7:06:32<28:34:27, 11.60s/step, acc=0.4339, loss=2.4769, lr=9.25e-05]
Training:  20%|█▉        | 2213/11082 [7:06:44<28:26:23, 11.54s/step, acc=0.4339, loss=2.4769, lr=9.25e-05]
Training:  20%|█▉        | 2214/11082 [7:06:55<28:34:40, 11.60s/step, acc=0.4339, loss=2.4769, lr=9.25e-05]
Training:  20%|█▉        | 2215/11082 [7:07:07<28:26:16, 11.55s/step, acc=0.4339, loss=2.4769, lr=9.25e-05]
Training:  20%|█▉        | 2216/11082 [7:07:18<28:15:45, 11.48s/step, acc=0.4339, loss=2.4769, lr=9.25e-05]
Training:  20%|██        | 2217/11082 [7:07:30<28:34:03, 11.60s/step, acc=0.4339, loss=2.4769, lr=9.25e-05]
Training:  20%|██        | 2218/11082 [7:07:41<28:27:01, 11.55s/step, acc=0.4339, loss=2.4769, lr=9.25e-05]
Training:  20%|██        | 2219/11082 [7:07:53<28:27:51, 11.56s/step, acc=0.4339, loss=2.4769, lr=9.25e-05]
Training:  20%|██        | 2220/11082 [7:08:05<28:30:46, 11.58s/step, acc=0.4339, loss=2.4769, lr=9.25e-05]
Training:  20%|██        | 2221/11082 [7:08:16<28:20:19, 11.51s/step, acc=0.4339, loss=2.4769, lr=9.25e-05]
Training:  20%|██        | 2222/11082 [7:08:28<28:34:43, 11.61s/step, acc=0.4339, loss=2.4769, lr=9.25e-05]
Training:  20%|██        | 2223/11082 [7:08:39<28:11:58, 11.46s/step, acc=0.4339, loss=2.4769, lr=9.25e-05]
Training:  20%|██        | 2224/11082 [7:08:51<28:21:48, 11.53s/step, acc=0.4339, loss=2.4769, lr=9.25e-05]
Training:  20%|██        | 2225/11082 [7:09:02<28:28:21, 11.57s/step, acc=0.4339, loss=2.4769, lr=9.25e-05]
Training:  20%|██        | 2226/11082 [7:09:14<28:39:18, 11.65s/step, acc=0.4339, loss=2.4769, lr=9.25e-05]
Training:  20%|██        | 2227/11082 [7:09:25<28:29:07, 11.58s/step, acc=0.4339, loss=2.4769, lr=9.25e-05]
Training:  20%|██        | 2228/11082 [7:09:37<28:29:16, 11.58s/step, acc=0.4339, loss=2.4769, lr=9.25e-05]
Training:  20%|██        | 2229/11082 [7:09:49<28:24:30, 11.55s/step, acc=0.4339, loss=2.4769, lr=9.25e-05]
Training:  20%|██        | 2230/11082 [7:10:00<28:12:06, 11.47s/step, acc=0.4339, loss=2.4769, lr=9.25e-05]
Training:  20%|██        | 2231/11082 [7:10:11<28:08:24, 11.45s/step, acc=0.4339, loss=2.4769, lr=9.25e-05]
Training:  20%|██        | 2232/11082 [7:10:23<28:26:14, 11.57s/step, acc=0.4339, loss=2.4769, lr=9.25e-05]
Training:  20%|██        | 2233/11082 [7:10:34<28:18:42, 11.52s/step, acc=0.4339, loss=2.4769, lr=9.25e-05]
Training:  20%|██        | 2234/11082 [7:10:45<27:49:34, 11.32s/step, acc=0.4339, loss=2.4769, lr=9.25e-05]
Training:  20%|██        | 2235/11082 [7:10:58<28:44:18, 11.69s/step, acc=0.4339, loss=2.4769, lr=9.25e-05]
Training:  20%|██        | 2236/11082 [7:11:09<28:25:44, 11.57s/step, acc=0.4339, loss=2.4769, lr=9.25e-05]
Training:  20%|██        | 2237/11082 [7:11:21<28:47:14, 11.72s/step, acc=0.4339, loss=2.4769, lr=9.25e-05]
Training:  20%|██        | 2238/11082 [7:11:32<28:21:20, 11.54s/step, acc=0.4339, loss=2.4769, lr=9.25e-05]
Training:  20%|██        | 2239/11082 [7:11:44<28:22:52, 11.55s/step, acc=0.4339, loss=2.4769, lr=9.25e-05]
Training:  20%|██        | 2240/11082 [7:11:56<28:49:30, 11.74s/step, acc=0.4339, loss=2.4769, lr=9.25e-05]
Training:  20%|██        | 2241/11082 [7:12:07<28:15:38, 11.51s/step, acc=0.4339, loss=2.4769, lr=9.25e-05]
Training:  20%|██        | 2242/11082 [7:12:19<28:35:11, 11.64s/step, acc=0.4339, loss=2.4769, lr=9.25e-05]
Training:  20%|██        | 2243/11082 [7:12:31<28:40:39, 11.68s/step, acc=0.4339, loss=2.4769, lr=9.25e-05]
Training:  20%|██        | 2244/11082 [7:12:43<28:52:41, 11.76s/step, acc=0.4339, loss=2.4769, lr=9.25e-05]
Training:  20%|██        | 2245/11082 [7:12:54<28:47:56, 11.73s/step, acc=0.4339, loss=2.4769, lr=9.25e-05]
Training:  20%|██        | 2246/11082 [7:13:06<28:39:37, 11.68s/step, acc=0.4339, loss=2.4769, lr=9.25e-05]
Training:  20%|██        | 2247/11082 [7:13:18<28:32:59, 11.63s/step, acc=0.4339, loss=2.4769, lr=9.25e-05]
Training:  20%|██        | 2248/11082 [7:13:30<29:05:43, 11.86s/step, acc=0.4339, loss=2.4769, lr=9.25e-05]
Training:  20%|██        | 2249/11082 [7:13:42<28:55:42, 11.79s/step, acc=0.4339, loss=2.4769, lr=9.25e-05]
Training:  20%|██        | 2250/11082 [7:13:53<28:41:37, 11.70s/step, acc=0.4339, loss=2.4769, lr=9.25e-05]
Training:  20%|██        | 2250/11082 [7:14:00<28:41:37, 11.70s/step, acc=0.4367, loss=2.4622, lr=9.21e-05]2026-06-18 22:07:28 | INFO | __main__ | epoch=0 step=2250 train_loss=2.462176 train_acc=0.436739 lr=9.215e-05 grad_norm=1.0859 acc_d0=0.7091 acc_d1=0.5487 acc_d2=0.4194 acc_d3=0.3188 acc_d4=0.2460 acc_d5=0.1958 acc_d6=0.1609 acc_d7=0.1350

Training:  20%|██        | 2251/11082 [7:14:04<28:20:13, 11.55s/step, acc=0.4367, loss=2.4622, lr=9.21e-05]
Training:  20%|██        | 2252/11082 [7:14:16<28:42:39, 11.71s/step, acc=0.4367, loss=2.4622, lr=9.21e-05]
Training:  20%|██        | 2253/11082 [7:14:28<28:21:39, 11.56s/step, acc=0.4367, loss=2.4622, lr=9.21e-05]
Training:  20%|██        | 2254/11082 [7:14:39<28:32:48, 11.64s/step, acc=0.4367, loss=2.4622, lr=9.21e-05]
Training:  20%|██        | 2255/11082 [7:14:50<28:11:42, 11.50s/step, acc=0.4367, loss=2.4622, lr=9.21e-05]
Training:  20%|██        | 2256/11082 [7:15:02<28:11:01, 11.50s/step, acc=0.4367, loss=2.4622, lr=9.21e-05]
Training:  20%|██        | 2257/11082 [7:15:14<28:21:46, 11.57s/step, acc=0.4367, loss=2.4622, lr=9.21e-05]
Training:  20%|██        | 2258/11082 [7:15:25<28:23:04, 11.58s/step, acc=0.4367, loss=2.4622, lr=9.21e-05]
Training:  20%|██        | 2259/11082 [7:15:37<28:23:10, 11.58s/step, acc=0.4367, loss=2.4622, lr=9.21e-05]
Training:  20%|██        | 2260/11082 [7:15:49<28:25:51, 11.60s/step, acc=0.4367, loss=2.4622, lr=9.21e-05]
Training:  20%|██        | 2261/11082 [7:16:00<28:23:26, 11.59s/step, acc=0.4367, loss=2.4622, lr=9.21e-05]
Training:  20%|██        | 2262/11082 [7:16:12<28:31:56, 11.65s/step, acc=0.4367, loss=2.4622, lr=9.21e-05]
Training:  20%|██        | 2263/11082 [7:16:24<28:30:52, 11.64s/step, acc=0.4367, loss=2.4622, lr=9.21e-05]
Training:  20%|██        | 2264/11082 [7:16:35<28:34:14, 11.66s/step, acc=0.4367, loss=2.4622, lr=9.21e-05]
Training:  20%|██        | 2265/11082 [7:16:47<28:27:32, 11.62s/step, acc=0.4367, loss=2.4622, lr=9.21e-05]
Training:  20%|██        | 2266/11082 [7:16:58<28:32:09, 11.65s/step, acc=0.4367, loss=2.4622, lr=9.21e-05]
Training:  20%|██        | 2267/11082 [7:17:10<28:09:49, 11.50s/step, acc=0.4367, loss=2.4622, lr=9.21e-05]
Training:  20%|██        | 2268/11082 [7:17:21<28:05:54, 11.48s/step, acc=0.4367, loss=2.4622, lr=9.21e-05]
Training:  20%|██        | 2269/11082 [7:17:33<28:06:11, 11.48s/step, acc=0.4367, loss=2.4622, lr=9.21e-05]
Training:  20%|██        | 2270/11082 [7:17:44<28:00:50, 11.44s/step, acc=0.4367, loss=2.4622, lr=9.21e-05]
Training:  20%|██        | 2271/11082 [7:17:56<28:14:43, 11.54s/step, acc=0.4367, loss=2.4622, lr=9.21e-05]
Training:  21%|██        | 2272/11082 [7:18:07<28:13:20, 11.53s/step, acc=0.4367, loss=2.4622, lr=9.21e-05]
Training:  21%|██        | 2273/11082 [7:18:19<28:31:44, 11.66s/step, acc=0.4367, loss=2.4622, lr=9.21e-05]
Training:  21%|██        | 2274/11082 [7:18:30<27:59:29, 11.44s/step, acc=0.4367, loss=2.4622, lr=9.21e-05]
Training:  21%|██        | 2275/11082 [7:18:42<28:12:06, 11.53s/step, acc=0.4367, loss=2.4622, lr=9.21e-05]
Training:  21%|██        | 2276/11082 [7:18:54<28:26:52, 11.63s/step, acc=0.4367, loss=2.4622, lr=9.21e-05]
Training:  21%|██        | 2277/11082 [7:19:05<28:20:29, 11.59s/step, acc=0.4367, loss=2.4622, lr=9.21e-05]
Training:  21%|██        | 2278/11082 [7:19:17<28:12:31, 11.53s/step, acc=0.4367, loss=2.4622, lr=9.21e-05]
Training:  21%|██        | 2279/11082 [7:19:28<28:05:25, 11.49s/step, acc=0.4367, loss=2.4622, lr=9.21e-05]
Training:  21%|██        | 2280/11082 [7:19:40<28:11:17, 11.53s/step, acc=0.4367, loss=2.4622, lr=9.21e-05]
Training:  21%|██        | 2281/11082 [7:19:51<28:20:10, 11.59s/step, acc=0.4367, loss=2.4622, lr=9.21e-05]
Training:  21%|██        | 2282/11082 [7:20:03<28:02:18, 11.47s/step, acc=0.4367, loss=2.4622, lr=9.21e-05]
Training:  21%|██        | 2283/11082 [7:20:14<28:08:59, 11.52s/step, acc=0.4367, loss=2.4622, lr=9.21e-05]
Training:  21%|██        | 2284/11082 [7:20:26<28:07:22, 11.51s/step, acc=0.4367, loss=2.4622, lr=9.21e-05]
Training:  21%|██        | 2285/11082 [7:20:38<28:26:47, 11.64s/step, acc=0.4367, loss=2.4622, lr=9.21e-05]
Training:  21%|██        | 2286/11082 [7:20:49<28:18:24, 11.59s/step, acc=0.4367, loss=2.4622, lr=9.21e-05]
Training:  21%|██        | 2287/11082 [7:21:01<28:14:49, 11.56s/step, acc=0.4367, loss=2.4622, lr=9.21e-05]
Training:  21%|██        | 2288/11082 [7:21:12<28:19:30, 11.60s/step, acc=0.4367, loss=2.4622, lr=9.21e-05]
Training:  21%|██        | 2289/11082 [7:21:23<28:02:35, 11.48s/step, acc=0.4367, loss=2.4622, lr=9.21e-05]
Training:  21%|██        | 2290/11082 [7:21:35<27:56:30, 11.44s/step, acc=0.4367, loss=2.4622, lr=9.21e-05]
Training:  21%|██        | 2291/11082 [7:21:47<28:11:20, 11.54s/step, acc=0.4367, loss=2.4622, lr=9.21e-05]
Training:  21%|██        | 2292/11082 [7:21:58<28:18:17, 11.59s/step, acc=0.4367, loss=2.4622, lr=9.21e-05]
Training:  21%|██        | 2293/11082 [7:22:10<28:07:38, 11.52s/step, acc=0.4367, loss=2.4622, lr=9.21e-05]
Training:  21%|██        | 2294/11082 [7:22:21<28:13:01, 11.56s/step, acc=0.4367, loss=2.4622, lr=9.21e-05]
Training:  21%|██        | 2295/11082 [7:22:32<27:46:13, 11.38s/step, acc=0.4367, loss=2.4622, lr=9.21e-05]
Training:  21%|██        | 2296/11082 [7:22:44<28:22:47, 11.63s/step, acc=0.4367, loss=2.4622, lr=9.21e-05]
Training:  21%|██        | 2297/11082 [7:22:56<28:20:23, 11.61s/step, acc=0.4367, loss=2.4622, lr=9.21e-05]
Training:  21%|██        | 2298/11082 [7:23:07<28:03:36, 11.50s/step, acc=0.4367, loss=2.4622, lr=9.21e-05]
Training:  21%|██        | 2299/11082 [7:23:19<27:54:32, 11.44s/step, acc=0.4367, loss=2.4622, lr=9.21e-05]
Training:  21%|██        | 2300/11082 [7:23:31<28:17:54, 11.60s/step, acc=0.4367, loss=2.4622, lr=9.21e-05]
Training:  21%|██        | 2300/11082 [7:23:37<28:17:54, 11.60s/step, acc=0.4401, loss=2.4356, lr=9.18e-05]2026-06-18 22:17:05 | INFO | __main__ | epoch=0 step=2300 train_loss=2.435604 train_acc=0.440059 lr=9.178e-05 grad_norm=1.3984 acc_d0=0.7166 acc_d1=0.5558 acc_d2=0.4243 acc_d3=0.3222 acc_d4=0.2500 acc_d5=0.1980 acc_d6=0.1622 acc_d7=0.1367

Training:  21%|██        | 2301/11082 [7:23:42<27:52:35, 11.43s/step, acc=0.4401, loss=2.4356, lr=9.18e-05]
Training:  21%|██        | 2302/11082 [7:23:53<27:57:37, 11.46s/step, acc=0.4401, loss=2.4356, lr=9.18e-05]
Training:  21%|██        | 2303/11082 [7:24:05<28:13:03, 11.57s/step, acc=0.4401, loss=2.4356, lr=9.18e-05]
Training:  21%|██        | 2304/11082 [7:24:16<27:44:26, 11.38s/step, acc=0.4401, loss=2.4356, lr=9.18e-05]
Training:  21%|██        | 2305/11082 [7:24:28<28:03:32, 11.51s/step, acc=0.4401, loss=2.4356, lr=9.18e-05]
Training:  21%|██        | 2306/11082 [7:24:39<28:04:02, 11.51s/step, acc=0.4401, loss=2.4356, lr=9.18e-05]
Training:  21%|██        | 2307/11082 [7:24:51<28:20:14, 11.63s/step, acc=0.4401, loss=2.4356, lr=9.18e-05]
Training:  21%|██        | 2308/11082 [7:25:02<28:02:09, 11.50s/step, acc=0.4401, loss=2.4356, lr=9.18e-05]
Training:  21%|██        | 2309/11082 [7:25:14<28:14:38, 11.59s/step, acc=0.4401, loss=2.4356, lr=9.18e-05]
Training:  21%|██        | 2310/11082 [7:25:26<28:22:54, 11.65s/step, acc=0.4401, loss=2.4356, lr=9.18e-05]
Training:  21%|██        | 2311/11082 [7:25:37<27:44:30, 11.39s/step, acc=0.4401, loss=2.4356, lr=9.18e-05]
Training:  21%|██        | 2312/11082 [7:25:49<28:09:03, 11.56s/step, acc=0.4401, loss=2.4356, lr=9.18e-05]
Training:  21%|██        | 2313/11082 [7:26:00<28:03:39, 11.52s/step, acc=0.4401, loss=2.4356, lr=9.18e-05]
Training:  21%|██        | 2314/11082 [7:26:12<28:02:04, 11.51s/step, acc=0.4401, loss=2.4356, lr=9.18e-05]
Training:  21%|██        | 2315/11082 [7:26:23<28:18:35, 11.62s/step, acc=0.4401, loss=2.4356, lr=9.18e-05]
Training:  21%|██        | 2316/11082 [7:26:35<28:07:03, 11.55s/step, acc=0.4401, loss=2.4356, lr=9.18e-05]
Training:  21%|██        | 2317/11082 [7:26:46<28:00:19, 11.50s/step, acc=0.4401, loss=2.4356, lr=9.18e-05]
Training:  21%|██        | 2318/11082 [7:26:58<28:27:05, 11.69s/step, acc=0.4401, loss=2.4356, lr=9.18e-05]
Training:  21%|██        | 2319/11082 [7:27:10<28:39:04, 11.77s/step, acc=0.4401, loss=2.4356, lr=9.18e-05]
Training:  21%|██        | 2320/11082 [7:27:22<28:35:16, 11.75s/step, acc=0.4401, loss=2.4356, lr=9.18e-05]
Training:  21%|██        | 2321/11082 [7:27:33<28:18:47, 11.63s/step, acc=0.4401, loss=2.4356, lr=9.18e-05]
Training:  21%|██        | 2322/11082 [7:27:45<28:27:12, 11.69s/step, acc=0.4401, loss=2.4356, lr=9.18e-05]
Training:  21%|██        | 2323/11082 [7:27:56<28:00:43, 11.51s/step, acc=0.4401, loss=2.4356, lr=9.18e-05]
Training:  21%|██        | 2324/11082 [7:28:08<28:23:34, 11.67s/step, acc=0.4401, loss=2.4356, lr=9.18e-05]
Training:  21%|██        | 2325/11082 [7:28:20<28:18:09, 11.64s/step, acc=0.4401, loss=2.4356, lr=9.18e-05]
Training:  21%|██        | 2326/11082 [7:28:31<27:56:36, 11.49s/step, acc=0.4401, loss=2.4356, lr=9.18e-05]
Training:  21%|██        | 2327/11082 [7:28:43<28:12:17, 11.60s/step, acc=0.4401, loss=2.4356, lr=9.18e-05]
Training:  21%|██        | 2328/11082 [7:28:55<28:24:47, 11.68s/step, acc=0.4401, loss=2.4356, lr=9.18e-05]
Training:  21%|██        | 2329/11082 [7:29:06<28:22:53, 11.67s/step, acc=0.4401, loss=2.4356, lr=9.18e-05]
Training:  21%|██        | 2330/11082 [7:29:18<28:30:54, 11.73s/step, acc=0.4401, loss=2.4356, lr=9.18e-05]
Training:  21%|██        | 2331/11082 [7:29:30<28:32:56, 11.74s/step, acc=0.4401, loss=2.4356, lr=9.18e-05]
Training:  21%|██        | 2332/11082 [7:29:41<28:19:08, 11.65s/step, acc=0.4401, loss=2.4356, lr=9.18e-05]
Training:  21%|██        | 2333/11082 [7:29:53<28:20:29, 11.66s/step, acc=0.4401, loss=2.4356, lr=9.18e-05]
Training:  21%|██        | 2334/11082 [7:30:05<28:11:37, 11.60s/step, acc=0.4401, loss=2.4356, lr=9.18e-05]
Training:  21%|██        | 2335/11082 [7:30:16<28:09:16, 11.59s/step, acc=0.4401, loss=2.4356, lr=9.18e-05]
Training:  21%|██        | 2336/11082 [7:30:28<28:09:47, 11.59s/step, acc=0.4401, loss=2.4356, lr=9.18e-05]
Training:  21%|██        | 2337/11082 [7:30:39<28:03:39, 11.55s/step, acc=0.4401, loss=2.4356, lr=9.18e-05]
Training:  21%|██        | 2338/11082 [7:30:51<28:06:56, 11.58s/step, acc=0.4401, loss=2.4356, lr=9.18e-05]
Training:  21%|██        | 2339/11082 [7:31:02<28:05:02, 11.56s/step, acc=0.4401, loss=2.4356, lr=9.18e-05]
Training:  21%|██        | 2340/11082 [7:31:14<28:11:12, 11.61s/step, acc=0.4401, loss=2.4356, lr=9.18e-05]
Training:  21%|██        | 2341/11082 [7:31:25<27:54:50, 11.50s/step, acc=0.4401, loss=2.4356, lr=9.18e-05]
Training:  21%|██        | 2342/11082 [7:31:37<28:05:55, 11.57s/step, acc=0.4401, loss=2.4356, lr=9.18e-05]
Training:  21%|██        | 2343/11082 [7:31:48<27:58:11, 11.52s/step, acc=0.4401, loss=2.4356, lr=9.18e-05]
Training:  21%|██        | 2344/11082 [7:32:00<28:04:48, 11.57s/step, acc=0.4401, loss=2.4356, lr=9.18e-05]
Training:  21%|██        | 2345/11082 [7:32:12<28:02:59, 11.56s/step, acc=0.4401, loss=2.4356, lr=9.18e-05]
Training:  21%|██        | 2346/11082 [7:32:23<28:02:20, 11.55s/step, acc=0.4401, loss=2.4356, lr=9.18e-05]
Training:  21%|██        | 2347/11082 [7:32:35<27:53:55, 11.50s/step, acc=0.4401, loss=2.4356, lr=9.18e-05]
Training:  21%|██        | 2348/11082 [7:32:46<28:03:00, 11.56s/step, acc=0.4401, loss=2.4356, lr=9.18e-05]
Training:  21%|██        | 2349/11082 [7:32:58<28:17:56, 11.67s/step, acc=0.4401, loss=2.4356, lr=9.18e-05]
Training:  21%|██        | 2350/11082 [7:33:09<27:35:55, 11.38s/step, acc=0.4401, loss=2.4356, lr=9.18e-05]
Training:  21%|██        | 2350/11082 [7:33:17<27:35:55, 11.38s/step, acc=0.4407, loss=2.4311, lr=9.14e-05]2026-06-18 22:26:44 | INFO | __main__ | epoch=0 step=2350 train_loss=2.431056 train_acc=0.440740 lr=9.140e-05 grad_norm=1.4219 acc_d0=0.7211 acc_d1=0.5641 acc_d2=0.4338 acc_d3=0.3331 acc_d4=0.2583 acc_d5=0.2046 acc_d6=0.1687 acc_d7=0.1399

Training:  21%|██        | 2351/11082 [7:33:21<28:04:37, 11.58s/step, acc=0.4407, loss=2.4311, lr=9.14e-05]
Training:  21%|██        | 2352/11082 [7:33:33<28:12:31, 11.63s/step, acc=0.4407, loss=2.4311, lr=9.14e-05]
Training:  21%|██        | 2353/11082 [7:33:44<27:58:11, 11.54s/step, acc=0.4407, loss=2.4311, lr=9.14e-05]
Training:  21%|██        | 2354/11082 [7:33:55<27:52:17, 11.50s/step, acc=0.4407, loss=2.4311, lr=9.14e-05]
Training:  21%|██▏       | 2355/11082 [7:34:07<27:58:34, 11.54s/step, acc=0.4407, loss=2.4311, lr=9.14e-05]
Training:  21%|██▏       | 2356/11082 [7:34:18<27:40:01, 11.41s/step, acc=0.4407, loss=2.4311, lr=9.14e-05]
Training:  21%|██▏       | 2357/11082 [7:34:30<27:56:34, 11.53s/step, acc=0.4407, loss=2.4311, lr=9.14e-05]
Training:  21%|██▏       | 2358/11082 [7:34:41<27:44:15, 11.45s/step, acc=0.4407, loss=2.4311, lr=9.14e-05]
Training:  21%|██▏       | 2359/11082 [7:34:53<28:09:48, 11.62s/step, acc=0.4407, loss=2.4311, lr=9.14e-05]
Training:  21%|██▏       | 2360/11082 [7:35:05<28:07:40, 11.61s/step, acc=0.4407, loss=2.4311, lr=9.14e-05]
Training:  21%|██▏       | 2361/11082 [7:35:16<27:43:15, 11.44s/step, acc=0.4407, loss=2.4311, lr=9.14e-05]
Training:  21%|██▏       | 2362/11082 [7:35:28<28:02:26, 11.58s/step, acc=0.4407, loss=2.4311, lr=9.14e-05]
Training:  21%|██▏       | 2363/11082 [7:35:39<27:50:34, 11.50s/step, acc=0.4407, loss=2.4311, lr=9.14e-05]
Training:  21%|██▏       | 2364/11082 [7:35:51<27:52:48, 11.51s/step, acc=0.4407, loss=2.4311, lr=9.14e-05]
Training:  21%|██▏       | 2365/11082 [7:36:02<27:42:21, 11.44s/step, acc=0.4407, loss=2.4311, lr=9.14e-05]
Training:  21%|██▏       | 2366/11082 [7:36:14<27:55:41, 11.54s/step, acc=0.4407, loss=2.4311, lr=9.14e-05]
Training:  21%|██▏       | 2367/11082 [7:36:24<27:12:22, 11.24s/step, acc=0.4407, loss=2.4311, lr=9.14e-05]
Training:  21%|██▏       | 2368/11082 [7:36:37<28:09:27, 11.63s/step, acc=0.4407, loss=2.4311, lr=9.14e-05]
Training:  21%|██▏       | 2369/11082 [7:36:48<28:07:47, 11.62s/step, acc=0.4407, loss=2.4311, lr=9.14e-05]
Training:  21%|██▏       | 2370/11082 [7:37:00<28:02:59, 11.59s/step, acc=0.4407, loss=2.4311, lr=9.14e-05]
Training:  21%|██▏       | 2371/11082 [7:37:11<27:56:22, 11.55s/step, acc=0.4407, loss=2.4311, lr=9.14e-05]
Training:  21%|██▏       | 2372/11082 [7:37:23<27:52:02, 11.52s/step, acc=0.4407, loss=2.4311, lr=9.14e-05]
Training:  21%|██▏       | 2373/11082 [7:37:35<28:03:46, 11.60s/step, acc=0.4407, loss=2.4311, lr=9.14e-05]
Training:  21%|██▏       | 2374/11082 [7:37:46<27:47:31, 11.49s/step, acc=0.4407, loss=2.4311, lr=9.14e-05]
Training:  21%|██▏       | 2375/11082 [7:37:57<27:41:27, 11.45s/step, acc=0.4407, loss=2.4311, lr=9.14e-05]
Training:  21%|██▏       | 2376/11082 [7:38:09<27:36:10, 11.41s/step, acc=0.4407, loss=2.4311, lr=9.14e-05]
Training:  21%|██▏       | 2377/11082 [7:38:20<27:52:22, 11.53s/step, acc=0.4407, loss=2.4311, lr=9.14e-05]
Training:  21%|██▏       | 2378/11082 [7:38:32<27:58:13, 11.57s/step, acc=0.4407, loss=2.4311, lr=9.14e-05]
Training:  21%|██▏       | 2379/11082 [7:38:44<28:13:15, 11.67s/step, acc=0.4407, loss=2.4311, lr=9.14e-05]
Training:  21%|██▏       | 2380/11082 [7:38:56<28:13:12, 11.67s/step, acc=0.4407, loss=2.4311, lr=9.14e-05]
Training:  21%|██▏       | 2381/11082 [7:39:07<27:47:11, 11.50s/step, acc=0.4407, loss=2.4311, lr=9.14e-05]
Training:  21%|██▏       | 2382/11082 [7:39:18<27:50:58, 11.52s/step, acc=0.4407, loss=2.4311, lr=9.14e-05]
Training:  22%|██▏       | 2383/11082 [7:39:30<27:58:32, 11.58s/step, acc=0.4407, loss=2.4311, lr=9.14e-05]
Training:  22%|██▏       | 2384/11082 [7:39:41<27:49:04, 11.51s/step, acc=0.4407, loss=2.4311, lr=9.14e-05]
Training:  22%|██▏       | 2385/11082 [7:39:53<27:50:37, 11.53s/step, acc=0.4407, loss=2.4311, lr=9.14e-05]
Training:  22%|██▏       | 2386/11082 [7:40:05<28:01:36, 11.60s/step, acc=0.4407, loss=2.4311, lr=9.14e-05]
Training:  22%|██▏       | 2387/11082 [7:40:16<27:48:52, 11.52s/step, acc=0.4407, loss=2.4311, lr=9.14e-05]
Training:  22%|██▏       | 2388/11082 [7:40:28<28:07:18, 11.64s/step, acc=0.4407, loss=2.4311, lr=9.14e-05]
Training:  22%|██▏       | 2389/11082 [7:40:39<27:32:46, 11.41s/step, acc=0.4407, loss=2.4311, lr=9.14e-05]
Training:  22%|██▏       | 2390/11082 [7:40:51<27:55:34, 11.57s/step, acc=0.4407, loss=2.4311, lr=9.14e-05]
Training:  22%|██▏       | 2391/11082 [7:41:02<27:45:34, 11.50s/step, acc=0.4407, loss=2.4311, lr=9.14e-05]
Training:  22%|██▏       | 2392/11082 [7:41:14<28:16:52, 11.72s/step, acc=0.4407, loss=2.4311, lr=9.14e-05]
Training:  22%|██▏       | 2393/11082 [7:41:26<28:00:55, 11.61s/step, acc=0.4407, loss=2.4311, lr=9.14e-05]
Training:  22%|██▏       | 2394/11082 [7:41:37<27:59:40, 11.60s/step, acc=0.4407, loss=2.4311, lr=9.14e-05]
Training:  22%|██▏       | 2395/11082 [7:41:49<27:56:55, 11.58s/step, acc=0.4407, loss=2.4311, lr=9.14e-05]
Training:  22%|██▏       | 2396/11082 [7:42:01<28:29:10, 11.81s/step, acc=0.4407, loss=2.4311, lr=9.14e-05]
Training:  22%|██▏       | 2397/11082 [7:42:13<28:28:37, 11.80s/step, acc=0.4407, loss=2.4311, lr=9.14e-05]
Training:  22%|██▏       | 2398/11082 [7:42:24<28:09:10, 11.67s/step, acc=0.4407, loss=2.4311, lr=9.14e-05]
Training:  22%|██▏       | 2399/11082 [7:42:36<28:13:23, 11.70s/step, acc=0.4407, loss=2.4311, lr=9.14e-05]
Training:  22%|██▏       | 2400/11082 [7:42:48<28:17:19, 11.73s/step, acc=0.4407, loss=2.4311, lr=9.14e-05]
Training:  22%|██▏       | 2400/11082 [7:42:55<28:17:19, 11.73s/step, acc=0.4395, loss=2.4414, lr=9.10e-05]2026-06-18 22:36:23 | INFO | __main__ | epoch=0 step=2400 train_loss=2.441408 train_acc=0.439482 lr=9.102e-05 grad_norm=1.2578 acc_d0=0.7167 acc_d1=0.5569 acc_d2=0.4273 acc_d3=0.3235 acc_d4=0.2505 acc_d5=0.1988 acc_d6=0.1619 acc_d7=0.1362

Training:  22%|██▏       | 2401/11082 [7:43:00<28:31:57, 11.83s/step, acc=0.4395, loss=2.4414, lr=9.10e-05]
Training:  22%|██▏       | 2402/11082 [7:43:11<28:12:26, 11.70s/step, acc=0.4395, loss=2.4414, lr=9.10e-05]
Training:  22%|██▏       | 2403/11082 [7:43:23<28:14:11, 11.71s/step, acc=0.4395, loss=2.4414, lr=9.10e-05]
Training:  22%|██▏       | 2404/11082 [7:43:35<28:09:51, 11.68s/step, acc=0.4395, loss=2.4414, lr=9.10e-05]
Training:  22%|██▏       | 2405/11082 [7:43:46<27:47:52, 11.53s/step, acc=0.4395, loss=2.4414, lr=9.10e-05]
Training:  22%|██▏       | 2406/11082 [7:43:58<28:15:56, 11.73s/step, acc=0.4395, loss=2.4414, lr=9.10e-05]
Training:  22%|██▏       | 2407/11082 [7:44:10<28:18:04, 11.74s/step, acc=0.4395, loss=2.4414, lr=9.10e-05]
Training:  22%|██▏       | 2408/11082 [7:44:21<28:00:10, 11.62s/step, acc=0.4395, loss=2.4414, lr=9.10e-05]
Training:  22%|██▏       | 2409/11082 [7:44:33<28:04:28, 11.65s/step, acc=0.4395, loss=2.4414, lr=9.10e-05]
Training:  22%|██▏       | 2410/11082 [7:44:45<28:08:57, 11.69s/step, acc=0.4395, loss=2.4414, lr=9.10e-05]
Training:  22%|██▏       | 2411/11082 [7:44:56<28:09:31, 11.69s/step, acc=0.4395, loss=2.4414, lr=9.10e-05]
Training:  22%|██▏       | 2412/11082 [7:45:08<27:56:13, 11.60s/step, acc=0.4395, loss=2.4414, lr=9.10e-05]
Training:  22%|██▏       | 2413/11082 [7:45:19<27:51:11, 11.57s/step, acc=0.4395, loss=2.4414, lr=9.10e-05]
Training:  22%|██▏       | 2414/11082 [7:45:31<27:47:55, 11.55s/step, acc=0.4395, loss=2.4414, lr=9.10e-05]
Training:  22%|██▏       | 2415/11082 [7:45:43<28:09:25, 11.70s/step, acc=0.4395, loss=2.4414, lr=9.10e-05]
Training:  22%|██▏       | 2416/11082 [7:45:54<27:58:26, 11.62s/step, acc=0.4395, loss=2.4414, lr=9.10e-05]
Training:  22%|██▏       | 2417/11082 [7:46:06<27:51:03, 11.57s/step, acc=0.4395, loss=2.4414, lr=9.10e-05]
Training:  22%|██▏       | 2418/11082 [7:46:17<27:36:14, 11.47s/step, acc=0.4395, loss=2.4414, lr=9.10e-05]
Training:  22%|██▏       | 2419/11082 [7:46:28<27:38:15, 11.49s/step, acc=0.4395, loss=2.4414, lr=9.10e-05]
Training:  22%|██▏       | 2420/11082 [7:46:40<27:55:57, 11.61s/step, acc=0.4395, loss=2.4414, lr=9.10e-05]
Training:  22%|██▏       | 2421/11082 [7:46:51<27:31:35, 11.44s/step, acc=0.4395, loss=2.4414, lr=9.10e-05]
Training:  22%|██▏       | 2422/11082 [7:47:03<27:51:27, 11.58s/step, acc=0.4395, loss=2.4414, lr=9.10e-05]
Training:  22%|██▏       | 2423/11082 [7:47:15<27:42:57, 11.52s/step, acc=0.4395, loss=2.4414, lr=9.10e-05]
Training:  22%|██▏       | 2424/11082 [7:47:27<28:07:41, 11.70s/step, acc=0.4395, loss=2.4414, lr=9.10e-05]
Training:  22%|██▏       | 2425/11082 [7:47:38<27:40:07, 11.51s/step, acc=0.4395, loss=2.4414, lr=9.10e-05]
Training:  22%|██▏       | 2426/11082 [7:47:49<27:42:05, 11.52s/step, acc=0.4395, loss=2.4414, lr=9.10e-05]
Training:  22%|██▏       | 2427/11082 [7:48:01<27:33:26, 11.46s/step, acc=0.4395, loss=2.4414, lr=9.10e-05]
Training:  22%|██▏       | 2428/11082 [7:48:13<27:53:58, 11.61s/step, acc=0.4395, loss=2.4414, lr=9.10e-05]
Training:  22%|██▏       | 2429/11082 [7:48:24<27:41:19, 11.52s/step, acc=0.4395, loss=2.4414, lr=9.10e-05]
Training:  22%|██▏       | 2430/11082 [7:48:35<27:31:45, 11.45s/step, acc=0.4395, loss=2.4414, lr=9.10e-05]
Training:  22%|██▏       | 2431/11082 [7:48:46<27:23:29, 11.40s/step, acc=0.4395, loss=2.4414, lr=9.10e-05]
Training:  22%|██▏       | 2432/11082 [7:48:59<28:01:05, 11.66s/step, acc=0.4395, loss=2.4414, lr=9.10e-05]
Training:  22%|██▏       | 2433/11082 [7:49:10<27:46:35, 11.56s/step, acc=0.4395, loss=2.4414, lr=9.10e-05]
Training:  22%|██▏       | 2434/11082 [7:49:21<27:40:44, 11.52s/step, acc=0.4395, loss=2.4414, lr=9.10e-05]
Training:  22%|██▏       | 2435/11082 [7:49:33<27:37:26, 11.50s/step, acc=0.4395, loss=2.4414, lr=9.10e-05]
Training:  22%|██▏       | 2436/11082 [7:49:45<27:48:38, 11.58s/step, acc=0.4395, loss=2.4414, lr=9.10e-05]
Training:  22%|██▏       | 2437/11082 [7:49:56<27:39:55, 11.52s/step, acc=0.4395, loss=2.4414, lr=9.10e-05]
Training:  22%|██▏       | 2438/11082 [7:50:08<27:42:59, 11.54s/step, acc=0.4395, loss=2.4414, lr=9.10e-05]
Training:  22%|██▏       | 2439/11082 [7:50:19<27:51:12, 11.60s/step, acc=0.4395, loss=2.4414, lr=9.10e-05]
Training:  22%|██▏       | 2440/11082 [7:50:31<27:40:16, 11.53s/step, acc=0.4395, loss=2.4414, lr=9.10e-05]
Training:  22%|██▏       | 2441/11082 [7:50:42<27:41:54, 11.54s/step, acc=0.4395, loss=2.4414, lr=9.10e-05]
Training:  22%|██▏       | 2442/11082 [7:50:54<27:32:19, 11.47s/step, acc=0.4395, loss=2.4414, lr=9.10e-05]
Training:  22%|██▏       | 2443/11082 [7:51:05<27:37:26, 11.51s/step, acc=0.4395, loss=2.4414, lr=9.10e-05]
Training:  22%|██▏       | 2444/11082 [7:51:17<27:40:01, 11.53s/step, acc=0.4395, loss=2.4414, lr=9.10e-05]
Training:  22%|██▏       | 2445/11082 [7:51:28<27:39:09, 11.53s/step, acc=0.4395, loss=2.4414, lr=9.10e-05]
Training:  22%|██▏       | 2446/11082 [7:51:40<27:56:43, 11.65s/step, acc=0.4395, loss=2.4414, lr=9.10e-05]
Training:  22%|██▏       | 2447/11082 [7:51:51<27:29:00, 11.46s/step, acc=0.4395, loss=2.4414, lr=9.10e-05]
Training:  22%|██▏       | 2448/11082 [7:52:03<27:50:29, 11.61s/step, acc=0.4395, loss=2.4414, lr=9.10e-05]
Training:  22%|██▏       | 2449/11082 [7:52:15<27:52:49, 11.63s/step, acc=0.4395, loss=2.4414, lr=9.10e-05]
Training:  22%|██▏       | 2450/11082 [7:52:26<27:35:04, 11.50s/step, acc=0.4395, loss=2.4414, lr=9.10e-05]
Training:  22%|██▏       | 2450/11082 [7:52:34<27:35:04, 11.50s/step, acc=0.4440, loss=2.4107, lr=9.06e-05]2026-06-18 22:46:01 | INFO | __main__ | epoch=0 step=2450 train_loss=2.410707 train_acc=0.443960 lr=9.062e-05 grad_norm=0.9492 acc_d0=0.7129 acc_d1=0.5523 acc_d2=0.4211 acc_d3=0.3193 acc_d4=0.2468 acc_d5=0.1967 acc_d6=0.1608 acc_d7=0.1333

Training:  22%|██▏       | 2451/11082 [7:52:38<27:47:17, 11.59s/step, acc=0.4440, loss=2.4107, lr=9.06e-05]
Training:  22%|██▏       | 2452/11082 [7:52:49<27:31:53, 11.48s/step, acc=0.4440, loss=2.4107, lr=9.06e-05]
Training:  22%|██▏       | 2453/11082 [7:53:01<28:04:13, 11.71s/step, acc=0.4440, loss=2.4107, lr=9.06e-05]
Training:  22%|██▏       | 2454/11082 [7:53:13<27:36:42, 11.52s/step, acc=0.4440, loss=2.4107, lr=9.06e-05]
Training:  22%|██▏       | 2455/11082 [7:53:24<27:44:51, 11.58s/step, acc=0.4440, loss=2.4107, lr=9.06e-05]
Training:  22%|██▏       | 2456/11082 [7:53:36<27:41:10, 11.55s/step, acc=0.4440, loss=2.4107, lr=9.06e-05]
Training:  22%|██▏       | 2457/11082 [7:53:47<27:40:32, 11.55s/step, acc=0.4440, loss=2.4107, lr=9.06e-05]
Training:  22%|██▏       | 2458/11082 [7:53:59<27:47:15, 11.60s/step, acc=0.4440, loss=2.4107, lr=9.06e-05]
Training:  22%|██▏       | 2459/11082 [7:54:10<27:33:25, 11.50s/step, acc=0.4440, loss=2.4107, lr=9.06e-05]
Training:  22%|██▏       | 2460/11082 [7:54:22<27:35:35, 11.52s/step, acc=0.4440, loss=2.4107, lr=9.06e-05]
Training:  22%|██▏       | 2461/11082 [7:54:34<27:45:01, 11.59s/step, acc=0.4440, loss=2.4107, lr=9.06e-05]
Training:  22%|██▏       | 2462/11082 [7:54:45<27:48:02, 11.61s/step, acc=0.4440, loss=2.4107, lr=9.06e-05]
Training:  22%|██▏       | 2463/11082 [7:54:57<28:02:56, 11.72s/step, acc=0.4440, loss=2.4107, lr=9.06e-05]
Training:  22%|██▏       | 2464/11082 [7:55:08<27:43:27, 11.58s/step, acc=0.4440, loss=2.4107, lr=9.06e-05]
Training:  22%|██▏       | 2465/11082 [7:55:20<27:40:02, 11.56s/step, acc=0.4440, loss=2.4107, lr=9.06e-05]
Training:  22%|██▏       | 2466/11082 [7:55:32<27:59:11, 11.69s/step, acc=0.4440, loss=2.4107, lr=9.06e-05]
Training:  22%|██▏       | 2467/11082 [7:55:44<27:59:35, 11.70s/step, acc=0.4440, loss=2.4107, lr=9.06e-05]
Training:  22%|██▏       | 2468/11082 [7:55:55<27:45:37, 11.60s/step, acc=0.4440, loss=2.4107, lr=9.06e-05]
Training:  22%|██▏       | 2469/11082 [7:56:06<27:18:04, 11.41s/step, acc=0.4440, loss=2.4107, lr=9.06e-05]
Training:  22%|██▏       | 2470/11082 [7:56:18<27:50:53, 11.64s/step, acc=0.4440, loss=2.4107, lr=9.06e-05]
Training:  22%|██▏       | 2471/11082 [7:56:30<27:37:21, 11.55s/step, acc=0.4440, loss=2.4107, lr=9.06e-05]
Training:  22%|██▏       | 2472/11082 [7:56:41<27:39:14, 11.56s/step, acc=0.4440, loss=2.4107, lr=9.06e-05]
Training:  22%|██▏       | 2473/11082 [7:56:53<27:33:36, 11.52s/step, acc=0.4440, loss=2.4107, lr=9.06e-05]
Training:  22%|██▏       | 2474/11082 [7:57:04<27:37:59, 11.56s/step, acc=0.4440, loss=2.4107, lr=9.06e-05]
Training:  22%|██▏       | 2475/11082 [7:57:16<27:30:18, 11.50s/step, acc=0.4440, loss=2.4107, lr=9.06e-05]
Training:  22%|██▏       | 2476/11082 [7:57:27<27:39:09, 11.57s/step, acc=0.4440, loss=2.4107, lr=9.06e-05]
Training:  22%|██▏       | 2477/11082 [7:57:39<27:40:09, 11.58s/step, acc=0.4440, loss=2.4107, lr=9.06e-05]
Training:  22%|██▏       | 2478/11082 [7:57:50<27:30:10, 11.51s/step, acc=0.4440, loss=2.4107, lr=9.06e-05]
Training:  22%|██▏       | 2479/11082 [7:58:03<28:05:07, 11.75s/step, acc=0.4440, loss=2.4107, lr=9.06e-05]
Training:  22%|██▏       | 2480/11082 [7:58:14<27:59:41, 11.72s/step, acc=0.4440, loss=2.4107, lr=9.06e-05]
Training:  22%|██▏       | 2481/11082 [7:58:25<27:40:02, 11.58s/step, acc=0.4440, loss=2.4107, lr=9.06e-05]
Training:  22%|██▏       | 2482/11082 [7:58:38<28:02:03, 11.74s/step, acc=0.4440, loss=2.4107, lr=9.06e-05]
Training:  22%|██▏       | 2483/11082 [7:58:49<27:55:07, 11.69s/step, acc=0.4440, loss=2.4107, lr=9.06e-05]
Training:  22%|██▏       | 2484/11082 [7:59:01<27:44:39, 11.62s/step, acc=0.4440, loss=2.4107, lr=9.06e-05]
Training:  22%|██▏       | 2485/11082 [7:59:12<27:38:59, 11.58s/step, acc=0.4440, loss=2.4107, lr=9.06e-05]
Training:  22%|██▏       | 2486/11082 [7:59:24<27:40:13, 11.59s/step, acc=0.4440, loss=2.4107, lr=9.06e-05]
Training:  22%|██▏       | 2487/11082 [7:59:35<27:45:51, 11.63s/step, acc=0.4440, loss=2.4107, lr=9.06e-05]
Training:  22%|██▏       | 2488/11082 [7:59:47<27:46:34, 11.64s/step, acc=0.4440, loss=2.4107, lr=9.06e-05]
Training:  22%|██▏       | 2489/11082 [7:59:58<27:26:19, 11.50s/step, acc=0.4440, loss=2.4107, lr=9.06e-05]
Training:  22%|██▏       | 2490/11082 [8:00:10<27:39:04, 11.59s/step, acc=0.4440, loss=2.4107, lr=9.06e-05]
Training:  22%|██▏       | 2491/11082 [8:00:22<27:43:49, 11.62s/step, acc=0.4440, loss=2.4107, lr=9.06e-05]
Training:  22%|██▏       | 2492/11082 [8:00:33<27:40:16, 11.60s/step, acc=0.4440, loss=2.4107, lr=9.06e-05]
Training:  22%|██▏       | 2493/11082 [8:00:45<27:39:51, 11.60s/step, acc=0.4440, loss=2.4107, lr=9.06e-05]
Training:  23%|██▎       | 2494/11082 [8:00:56<27:37:01, 11.58s/step, acc=0.4440, loss=2.4107, lr=9.06e-05]
Training:  23%|██▎       | 2495/11082 [8:01:08<27:25:47, 11.50s/step, acc=0.4440, loss=2.4107, lr=9.06e-05]
Training:  23%|██▎       | 2496/11082 [8:01:20<27:48:08, 11.66s/step, acc=0.4440, loss=2.4107, lr=9.06e-05]
Training:  23%|██▎       | 2497/11082 [8:01:31<27:40:23, 11.60s/step, acc=0.4440, loss=2.4107, lr=9.06e-05]
Training:  23%|██▎       | 2498/11082 [8:01:43<27:32:04, 11.55s/step, acc=0.4440, loss=2.4107, lr=9.06e-05]
Training:  23%|██▎       | 2499/11082 [8:01:54<27:40:17, 11.61s/step, acc=0.4440, loss=2.4107, lr=9.06e-05]
Training:  23%|██▎       | 2500/11082 [8:02:06<27:23:36, 11.49s/step, acc=0.4440, loss=2.4107, lr=9.06e-05]
Training:  23%|██▎       | 2500/11082 [8:02:13<27:23:36, 11.49s/step, acc=0.4473, loss=2.3853, lr=9.02e-05]2026-06-18 22:55:40 | INFO | __main__ | epoch=0 step=2500 train_loss=2.385312 train_acc=0.447315 lr=9.022e-05 grad_norm=1.4766 acc_d0=0.7159 acc_d1=0.5595 acc_d2=0.4324 acc_d3=0.3323 acc_d4=0.2584 acc_d5=0.2064 acc_d6=0.1679 acc_d7=0.1408

Training:  23%|██▎       | 2501/11082 [8:02:17<27:20:45, 11.47s/step, acc=0.4473, loss=2.3853, lr=9.02e-05]
Training:  23%|██▎       | 2502/11082 [8:02:28<27:17:57, 11.45s/step, acc=0.4473, loss=2.3853, lr=9.02e-05]
Training:  23%|██▎       | 2503/11082 [8:02:40<27:32:57, 11.56s/step, acc=0.4473, loss=2.3853, lr=9.02e-05]
Training:  23%|██▎       | 2504/11082 [8:02:51<27:08:02, 11.39s/step, acc=0.4473, loss=2.3853, lr=9.02e-05]
Training:  23%|██▎       | 2505/11082 [8:03:03<27:30:46, 11.55s/step, acc=0.4473, loss=2.3853, lr=9.02e-05]
Training:  23%|██▎       | 2506/11082 [8:03:14<27:20:35, 11.48s/step, acc=0.4473, loss=2.3853, lr=9.02e-05]
Training:  23%|██▎       | 2507/11082 [8:03:26<27:35:54, 11.59s/step, acc=0.4473, loss=2.3853, lr=9.02e-05]
Training:  23%|██▎       | 2508/11082 [8:03:38<27:33:00, 11.57s/step, acc=0.4473, loss=2.3853, lr=9.02e-05]
Training:  23%|██▎       | 2509/11082 [8:03:49<27:35:23, 11.59s/step, acc=0.4473, loss=2.3853, lr=9.02e-05]
Training:  23%|██▎       | 2510/11082 [8:04:01<27:30:13, 11.55s/step, acc=0.4473, loss=2.3853, lr=9.02e-05]
Training:  23%|██▎       | 2511/11082 [8:04:13<27:37:58, 11.61s/step, acc=0.4473, loss=2.3853, lr=9.02e-05]
Training:  23%|██▎       | 2512/11082 [8:04:24<27:39:01, 11.62s/step, acc=0.4473, loss=2.3853, lr=9.02e-05]
Training:  23%|██▎       | 2513/11082 [8:04:36<27:25:53, 11.52s/step, acc=0.4473, loss=2.3853, lr=9.02e-05]
Training:  23%|██▎       | 2514/11082 [8:04:47<27:10:08, 11.42s/step, acc=0.4473, loss=2.3853, lr=9.02e-05]
Training:  23%|██▎       | 2515/11082 [8:04:58<27:19:58, 11.49s/step, acc=0.4473, loss=2.3853, lr=9.02e-05]
Training:  23%|██▎       | 2516/11082 [8:05:10<27:25:01, 11.52s/step, acc=0.4473, loss=2.3853, lr=9.02e-05]
Training:  23%|██▎       | 2517/11082 [8:05:22<27:36:20, 11.60s/step, acc=0.4473, loss=2.3853, lr=9.02e-05]
Training:  23%|██▎       | 2518/11082 [8:05:33<27:26:45, 11.54s/step, acc=0.4473, loss=2.3853, lr=9.02e-05]
Training:  23%|██▎       | 2519/11082 [8:05:45<27:20:15, 11.49s/step, acc=0.4473, loss=2.3853, lr=9.02e-05]
Training:  23%|██▎       | 2520/11082 [8:05:56<27:32:46, 11.58s/step, acc=0.4473, loss=2.3853, lr=9.02e-05]
Training:  23%|██▎       | 2521/11082 [8:06:08<27:13:50, 11.45s/step, acc=0.4473, loss=2.3853, lr=9.02e-05]
Training:  23%|██▎       | 2522/11082 [8:06:19<27:15:27, 11.46s/step, acc=0.4473, loss=2.3853, lr=9.02e-05]
Training:  23%|██▎       | 2523/11082 [8:06:31<27:44:43, 11.67s/step, acc=0.4473, loss=2.3853, lr=9.02e-05]
Training:  23%|██▎       | 2524/11082 [8:06:43<27:30:57, 11.57s/step, acc=0.4473, loss=2.3853, lr=9.02e-05]
Training:  23%|██▎       | 2525/11082 [8:06:54<27:13:29, 11.45s/step, acc=0.4473, loss=2.3853, lr=9.02e-05]
Training:  23%|██▎       | 2526/11082 [8:07:05<27:26:11, 11.54s/step, acc=0.4473, loss=2.3853, lr=9.02e-05]
Training:  23%|██▎       | 2527/11082 [8:07:17<27:44:23, 11.67s/step, acc=0.4473, loss=2.3853, lr=9.02e-05]
Training:  23%|██▎       | 2528/11082 [8:07:29<27:42:50, 11.66s/step, acc=0.4473, loss=2.3853, lr=9.02e-05]
Training:  23%|██▎       | 2529/11082 [8:07:40<26:59:47, 11.36s/step, acc=0.4473, loss=2.3853, lr=9.02e-05]
Training:  23%|██▎       | 2530/11082 [8:07:52<27:42:13, 11.66s/step, acc=0.4473, loss=2.3853, lr=9.02e-05]
Training:  23%|██▎       | 2531/11082 [8:08:03<27:27:11, 11.56s/step, acc=0.4473, loss=2.3853, lr=9.02e-05]
Training:  23%|██▎       | 2532/11082 [8:08:15<27:09:42, 11.44s/step, acc=0.4473, loss=2.3853, lr=9.02e-05]
Training:  23%|██▎       | 2533/11082 [8:08:27<27:41:58, 11.66s/step, acc=0.4473, loss=2.3853, lr=9.02e-05]
Training:  23%|██▎       | 2534/11082 [8:08:38<27:32:29, 11.60s/step, acc=0.4473, loss=2.3853, lr=9.02e-05]
Training:  23%|██▎       | 2535/11082 [8:08:50<27:29:30, 11.58s/step, acc=0.4473, loss=2.3853, lr=9.02e-05]
Training:  23%|██▎       | 2536/11082 [8:09:01<27:34:18, 11.61s/step, acc=0.4473, loss=2.3853, lr=9.02e-05]
Training:  23%|██▎       | 2537/11082 [8:09:13<27:37:18, 11.64s/step, acc=0.4473, loss=2.3853, lr=9.02e-05]
Training:  23%|██▎       | 2538/11082 [8:09:24<27:23:54, 11.54s/step, acc=0.4473, loss=2.3853, lr=9.02e-05]
Training:  23%|██▎       | 2539/11082 [8:09:36<27:29:54, 11.59s/step, acc=0.4473, loss=2.3853, lr=9.02e-05]
Training:  23%|██▎       | 2540/11082 [8:09:48<27:44:38, 11.69s/step, acc=0.4473, loss=2.3853, lr=9.02e-05]
Training:  23%|██▎       | 2541/11082 [8:10:00<27:39:11, 11.66s/step, acc=0.4473, loss=2.3853, lr=9.02e-05]
Training:  23%|██▎       | 2542/11082 [8:10:11<27:36:06, 11.64s/step, acc=0.4473, loss=2.3853, lr=9.02e-05]
Training:  23%|██▎       | 2543/11082 [8:10:22<27:14:42, 11.49s/step, acc=0.4473, loss=2.3853, lr=9.02e-05]
Training:  23%|██▎       | 2544/11082 [8:10:34<27:07:27, 11.44s/step, acc=0.4473, loss=2.3853, lr=9.02e-05]
Training:  23%|██▎       | 2545/11082 [8:10:46<27:28:33, 11.59s/step, acc=0.4473, loss=2.3853, lr=9.02e-05]
Training:  23%|██▎       | 2546/11082 [8:10:58<27:45:51, 11.71s/step, acc=0.4473, loss=2.3853, lr=9.02e-05]
Training:  23%|██▎       | 2547/11082 [8:11:09<27:32:40, 11.62s/step, acc=0.4473, loss=2.3853, lr=9.02e-05]
Training:  23%|██▎       | 2548/11082 [8:11:21<27:34:54, 11.64s/step, acc=0.4473, loss=2.3853, lr=9.02e-05]
Training:  23%|██▎       | 2549/11082 [8:11:32<27:31:00, 11.61s/step, acc=0.4473, loss=2.3853, lr=9.02e-05]
Training:  23%|██▎       | 2550/11082 [8:11:43<27:08:24, 11.45s/step, acc=0.4473, loss=2.3853, lr=9.02e-05]
Training:  23%|██▎       | 2550/11082 [8:11:51<27:08:24, 11.45s/step, acc=0.4469, loss=2.3884, lr=8.98e-05]2026-06-18 23:05:19 | INFO | __main__ | epoch=0 step=2550 train_loss=2.388392 train_acc=0.446906 lr=8.982e-05 grad_norm=1.1016 acc_d0=0.7117 acc_d1=0.5508 acc_d2=0.4209 acc_d3=0.3213 acc_d4=0.2485 acc_d5=0.1975 acc_d6=0.1613 acc_d7=0.1329

Training:  23%|██▎       | 2551/11082 [8:11:56<27:50:53, 11.75s/step, acc=0.4469, loss=2.3884, lr=8.98e-05]
Training:  23%|██▎       | 2552/11082 [8:12:07<27:33:51, 11.63s/step, acc=0.4469, loss=2.3884, lr=8.98e-05]
Training:  23%|██▎       | 2553/11082 [8:12:19<27:23:58, 11.57s/step, acc=0.4469, loss=2.3884, lr=8.98e-05]
Training:  23%|██▎       | 2554/11082 [8:12:30<27:22:16, 11.55s/step, acc=0.4469, loss=2.3884, lr=8.98e-05]
Training:  23%|██▎       | 2555/11082 [8:12:42<27:20:14, 11.54s/step, acc=0.4469, loss=2.3884, lr=8.98e-05]
Training:  23%|██▎       | 2556/11082 [8:12:53<27:17:25, 11.52s/step, acc=0.4469, loss=2.3884, lr=8.98e-05]
Training:  23%|██▎       | 2557/11082 [8:13:05<27:40:21, 11.69s/step, acc=0.4469, loss=2.3884, lr=8.98e-05]
Training:  23%|██▎       | 2558/11082 [8:13:16<27:23:17, 11.57s/step, acc=0.4469, loss=2.3884, lr=8.98e-05]
Training:  23%|██▎       | 2559/11082 [8:13:28<27:26:16, 11.59s/step, acc=0.4469, loss=2.3884, lr=8.98e-05]
Training:  23%|██▎       | 2560/11082 [8:13:39<27:16:07, 11.52s/step, acc=0.4469, loss=2.3884, lr=8.98e-05]
Training:  23%|██▎       | 2561/11082 [8:13:51<27:35:16, 11.66s/step, acc=0.4469, loss=2.3884, lr=8.98e-05]
Training:  23%|██▎       | 2562/11082 [8:14:03<27:27:43, 11.60s/step, acc=0.4469, loss=2.3884, lr=8.98e-05]
Training:  23%|██▎       | 2563/11082 [8:14:15<27:45:24, 11.73s/step, acc=0.4469, loss=2.3884, lr=8.98e-05]
Training:  23%|██▎       | 2564/11082 [8:14:26<27:12:17, 11.50s/step, acc=0.4469, loss=2.3884, lr=8.98e-05]
Training:  23%|██▎       | 2565/11082 [8:14:38<27:35:44, 11.66s/step, acc=0.4469, loss=2.3884, lr=8.98e-05]
Training:  23%|██▎       | 2566/11082 [8:14:50<27:32:36, 11.64s/step, acc=0.4469, loss=2.3884, lr=8.98e-05]
Training:  23%|██▎       | 2567/11082 [8:15:01<27:28:29, 11.62s/step, acc=0.4469, loss=2.3884, lr=8.98e-05]
Training:  23%|██▎       | 2568/11082 [8:15:12<27:18:28, 11.55s/step, acc=0.4469, loss=2.3884, lr=8.98e-05]
Training:  23%|██▎       | 2569/11082 [8:15:24<27:27:39, 11.61s/step, acc=0.4469, loss=2.3884, lr=8.98e-05]
Training:  23%|██▎       | 2570/11082 [8:15:36<27:36:45, 11.68s/step, acc=0.4469, loss=2.3884, lr=8.98e-05]
Training:  23%|██▎       | 2571/11082 [8:15:47<27:25:23, 11.60s/step, acc=0.4469, loss=2.3884, lr=8.98e-05]
Training:  23%|██▎       | 2572/11082 [8:15:59<27:12:35, 11.51s/step, acc=0.4469, loss=2.3884, lr=8.98e-05]
Training:  23%|██▎       | 2573/11082 [8:16:11<27:35:51, 11.68s/step, acc=0.4469, loss=2.3884, lr=8.98e-05]
Training:  23%|██▎       | 2574/11082 [8:16:23<27:36:17, 11.68s/step, acc=0.4469, loss=2.3884, lr=8.98e-05]
Training:  23%|██▎       | 2575/11082 [8:16:34<27:44:56, 11.74s/step, acc=0.4469, loss=2.3884, lr=8.98e-05]
Training:  23%|██▎       | 2576/11082 [8:16:45<27:12:41, 11.52s/step, acc=0.4469, loss=2.3884, lr=8.98e-05]
Training:  23%|██▎       | 2577/11082 [8:16:57<27:18:27, 11.56s/step, acc=0.4469, loss=2.3884, lr=8.98e-05]
Training:  23%|██▎       | 2578/11082 [8:17:09<27:30:50, 11.65s/step, acc=0.4469, loss=2.3884, lr=8.98e-05]
Training:  23%|██▎       | 2579/11082 [8:17:20<27:07:08, 11.48s/step, acc=0.4469, loss=2.3884, lr=8.98e-05]
Training:  23%|██▎       | 2580/11082 [8:17:31<27:04:18, 11.46s/step, acc=0.4469, loss=2.3884, lr=8.98e-05]
Training:  23%|██▎       | 2581/11082 [8:17:43<27:16:28, 11.55s/step, acc=0.4469, loss=2.3884, lr=8.98e-05]
Training:  23%|██▎       | 2582/11082 [8:17:55<27:11:16, 11.51s/step, acc=0.4469, loss=2.3884, lr=8.98e-05]
Training:  23%|██▎       | 2583/11082 [8:18:06<27:18:02, 11.56s/step, acc=0.4469, loss=2.3884, lr=8.98e-05]
Training:  23%|██▎       | 2584/11082 [8:18:18<27:15:39, 11.55s/step, acc=0.4469, loss=2.3884, lr=8.98e-05]
Training:  23%|██▎       | 2585/11082 [8:18:29<27:20:14, 11.58s/step, acc=0.4469, loss=2.3884, lr=8.98e-05]
Training:  23%|██▎       | 2586/11082 [8:18:41<27:05:16, 11.48s/step, acc=0.4469, loss=2.3884, lr=8.98e-05]
Training:  23%|██▎       | 2587/11082 [8:18:53<27:19:55, 11.58s/step, acc=0.4469, loss=2.3884, lr=8.98e-05]
Training:  23%|██▎       | 2588/11082 [8:19:04<27:14:32, 11.55s/step, acc=0.4469, loss=2.3884, lr=8.98e-05]
Training:  23%|██▎       | 2589/11082 [8:19:15<27:06:49, 11.49s/step, acc=0.4469, loss=2.3884, lr=8.98e-05]
Training:  23%|██▎       | 2590/11082 [8:19:27<27:03:28, 11.47s/step, acc=0.4469, loss=2.3884, lr=8.98e-05]
Training:  23%|██▎       | 2591/11082 [8:19:38<26:58:40, 11.44s/step, acc=0.4469, loss=2.3884, lr=8.98e-05]
Training:  23%|██▎       | 2592/11082 [8:19:50<27:21:12, 11.60s/step, acc=0.4469, loss=2.3884, lr=8.98e-05]
Training:  23%|██▎       | 2593/11082 [8:20:02<27:15:38, 11.56s/step, acc=0.4469, loss=2.3884, lr=8.98e-05]
Training:  23%|██▎       | 2594/11082 [8:20:13<27:02:05, 11.47s/step, acc=0.4469, loss=2.3884, lr=8.98e-05]
Training:  23%|██▎       | 2595/11082 [8:20:25<27:25:13, 11.63s/step, acc=0.4469, loss=2.3884, lr=8.98e-05]
Training:  23%|██▎       | 2596/11082 [8:20:36<27:24:15, 11.63s/step, acc=0.4469, loss=2.3884, lr=8.98e-05]
Training:  23%|██▎       | 2597/11082 [8:20:48<27:14:02, 11.55s/step, acc=0.4469, loss=2.3884, lr=8.98e-05]
Training:  23%|██▎       | 2598/11082 [8:20:59<27:15:11, 11.56s/step, acc=0.4469, loss=2.3884, lr=8.98e-05]
Training:  23%|██▎       | 2599/11082 [8:21:11<27:11:04, 11.54s/step, acc=0.4469, loss=2.3884, lr=8.98e-05]
Training:  23%|██▎       | 2600/11082 [8:21:23<27:22:12, 11.62s/step, acc=0.4469, loss=2.3884, lr=8.98e-05]
Training:  23%|██▎       | 2600/11082 [8:21:30<27:22:12, 11.62s/step, acc=0.4474, loss=2.3846, lr=8.94e-05]2026-06-18 23:14:58 | INFO | __main__ | epoch=0 step=2600 train_loss=2.384589 train_acc=0.447379 lr=8.940e-05 grad_norm=1.3203 acc_d0=0.7188 acc_d1=0.5581 acc_d2=0.4265 acc_d3=0.3232 acc_d4=0.2499 acc_d5=0.1982 acc_d6=0.1634 acc_d7=0.1351

Training:  23%|██▎       | 2601/11082 [8:21:34<27:07:25, 11.51s/step, acc=0.4474, loss=2.3846, lr=8.94e-05]
Training:  23%|██▎       | 2602/11082 [8:21:45<26:58:59, 11.46s/step, acc=0.4474, loss=2.3846, lr=8.94e-05]
Training:  23%|██▎       | 2603/11082 [8:21:57<27:19:37, 11.60s/step, acc=0.4474, loss=2.3846, lr=8.94e-05]
Training:  23%|██▎       | 2604/11082 [8:22:09<27:11:35, 11.55s/step, acc=0.4474, loss=2.3846, lr=8.94e-05]
Training:  24%|██▎       | 2605/11082 [8:22:20<26:57:59, 11.45s/step, acc=0.4474, loss=2.3846, lr=8.94e-05]
Training:  24%|██▎       | 2606/11082 [8:22:32<27:11:55, 11.55s/step, acc=0.4474, loss=2.3846, lr=8.94e-05]
Training:  24%|██▎       | 2607/11082 [8:22:43<27:13:48, 11.57s/step, acc=0.4474, loss=2.3846, lr=8.94e-05]
Training:  24%|██▎       | 2608/11082 [8:22:55<27:28:47, 11.67s/step, acc=0.4474, loss=2.3846, lr=8.94e-05]
Training:  24%|██▎       | 2609/11082 [8:23:06<27:09:06, 11.54s/step, acc=0.4474, loss=2.3846, lr=8.94e-05]
Training:  24%|██▎       | 2610/11082 [8:23:18<27:03:57, 11.50s/step, acc=0.4474, loss=2.3846, lr=8.94e-05]
Training:  24%|██▎       | 2611/11082 [8:23:29<27:01:38, 11.49s/step, acc=0.4474, loss=2.3846, lr=8.94e-05]
Training:  24%|██▎       | 2612/11082 [8:23:41<27:22:15, 11.63s/step, acc=0.4474, loss=2.3846, lr=8.94e-05]
Training:  24%|██▎       | 2613/11082 [8:23:53<27:16:35, 11.59s/step, acc=0.4474, loss=2.3846, lr=8.94e-05]
Training:  24%|██▎       | 2614/11082 [8:24:04<27:17:48, 11.60s/step, acc=0.4474, loss=2.3846, lr=8.94e-05]
Training:  24%|██▎       | 2615/11082 [8:24:16<27:05:57, 11.52s/step, acc=0.4474, loss=2.3846, lr=8.94e-05]
Training:  24%|██▎       | 2616/11082 [8:24:28<27:19:08, 11.62s/step, acc=0.4474, loss=2.3846, lr=8.94e-05]
Training:  24%|██▎       | 2617/11082 [8:24:39<27:17:50, 11.61s/step, acc=0.4474, loss=2.3846, lr=8.94e-05]
Training:  24%|██▎       | 2618/11082 [8:24:51<27:21:45, 11.64s/step, acc=0.4474, loss=2.3846, lr=8.94e-05]
Training:  24%|██▎       | 2619/11082 [8:25:03<27:22:51, 11.65s/step, acc=0.4474, loss=2.3846, lr=8.94e-05]
Training:  24%|██▎       | 2620/11082 [8:25:14<27:11:56, 11.57s/step, acc=0.4474, loss=2.3846, lr=8.94e-05]
Training:  24%|██▎       | 2621/11082 [8:25:25<27:06:56, 11.54s/step, acc=0.4474, loss=2.3846, lr=8.94e-05]
Training:  24%|██▎       | 2622/11082 [8:25:37<26:58:12, 11.48s/step, acc=0.4474, loss=2.3846, lr=8.94e-05]
Training:  24%|██▎       | 2623/11082 [8:25:49<27:13:20, 11.59s/step, acc=0.4474, loss=2.3846, lr=8.94e-05]
Training:  24%|██▎       | 2624/11082 [8:26:00<27:06:55, 11.54s/step, acc=0.4474, loss=2.3846, lr=8.94e-05]
Training:  24%|██▎       | 2625/11082 [8:26:12<27:07:36, 11.55s/step, acc=0.4474, loss=2.3846, lr=8.94e-05]
Training:  24%|██▎       | 2626/11082 [8:26:23<27:17:56, 11.62s/step, acc=0.4474, loss=2.3846, lr=8.94e-05]
Training:  24%|██▎       | 2627/11082 [8:26:35<27:15:07, 11.60s/step, acc=0.4474, loss=2.3846, lr=8.94e-05]
Training:  24%|██▎       | 2628/11082 [8:26:46<27:12:54, 11.59s/step, acc=0.4474, loss=2.3846, lr=8.94e-05]
Training:  24%|██▎       | 2629/11082 [8:26:58<27:13:46, 11.60s/step, acc=0.4474, loss=2.3846, lr=8.94e-05]
Training:  24%|██▎       | 2630/11082 [8:27:09<27:02:01, 11.51s/step, acc=0.4474, loss=2.3846, lr=8.94e-05]
Training:  24%|██▎       | 2631/11082 [8:27:21<27:11:22, 11.58s/step, acc=0.4474, loss=2.3846, lr=8.94e-05]
Training:  24%|██▍       | 2632/11082 [8:27:32<26:49:21, 11.43s/step, acc=0.4474, loss=2.3846, lr=8.94e-05]
Training:  24%|██▍       | 2633/11082 [8:27:44<27:07:57, 11.56s/step, acc=0.4474, loss=2.3846, lr=8.94e-05]
Training:  24%|██▍       | 2634/11082 [8:27:55<26:53:11, 11.46s/step, acc=0.4474, loss=2.3846, lr=8.94e-05]
Training:  24%|██▍       | 2635/11082 [8:28:07<27:15:49, 11.62s/step, acc=0.4474, loss=2.3846, lr=8.94e-05]
Training:  24%|██▍       | 2636/11082 [8:28:19<27:15:57, 11.62s/step, acc=0.4474, loss=2.3846, lr=8.94e-05]
Training:  24%|██▍       | 2637/11082 [8:28:30<27:10:28, 11.58s/step, acc=0.4474, loss=2.3846, lr=8.94e-05]
Training:  24%|██▍       | 2638/11082 [8:28:42<27:01:51, 11.52s/step, acc=0.4474, loss=2.3846, lr=8.94e-05]
Training:  24%|██▍       | 2639/11082 [8:28:53<27:08:02, 11.57s/step, acc=0.4474, loss=2.3846, lr=8.94e-05]
Training:  24%|██▍       | 2640/11082 [8:29:05<27:08:19, 11.57s/step, acc=0.4474, loss=2.3846, lr=8.94e-05]
Training:  24%|██▍       | 2641/11082 [8:29:17<27:17:44, 11.64s/step, acc=0.4474, loss=2.3846, lr=8.94e-05]
Training:  24%|██▍       | 2642/11082 [8:29:29<27:22:10, 11.67s/step, acc=0.4474, loss=2.3846, lr=8.94e-05]
Training:  24%|██▍       | 2643/11082 [8:29:41<27:41:03, 11.81s/step, acc=0.4474, loss=2.3846, lr=8.94e-05]
Training:  24%|██▍       | 2644/11082 [8:29:52<27:32:01, 11.75s/step, acc=0.4474, loss=2.3846, lr=8.94e-05]
Training:  24%|██▍       | 2645/11082 [8:30:03<27:02:18, 11.54s/step, acc=0.4474, loss=2.3846, lr=8.94e-05]
Training:  24%|██▍       | 2646/11082 [8:30:15<27:25:03, 11.70s/step, acc=0.4474, loss=2.3846, lr=8.94e-05]
Training:  24%|██▍       | 2647/11082 [8:30:27<27:12:23, 11.61s/step, acc=0.4474, loss=2.3846, lr=8.94e-05]
Training:  24%|██▍       | 2648/11082 [8:30:38<27:08:14, 11.58s/step, acc=0.4474, loss=2.3846, lr=8.94e-05]
Training:  24%|██▍       | 2649/11082 [8:30:50<27:11:14, 11.61s/step, acc=0.4474, loss=2.3846, lr=8.94e-05]
Training:  24%|██▍       | 2650/11082 [8:31:02<27:15:54, 11.64s/step, acc=0.4474, loss=2.3846, lr=8.94e-05]
Training:  24%|██▍       | 2650/11082 [8:31:09<27:15:54, 11.64s/step, acc=0.4505, loss=2.3655, lr=8.90e-05]2026-06-18 23:24:36 | INFO | __main__ | epoch=0 step=2650 train_loss=2.365492 train_acc=0.450475 lr=8.898e-05 grad_norm=1.3594 acc_d0=0.7157 acc_d1=0.5561 acc_d2=0.4257 acc_d3=0.3263 acc_d4=0.2534 acc_d5=0.2007 acc_d6=0.1658 acc_d7=0.1380

Training:  24%|██▍       | 2651/11082 [8:31:13<26:53:53, 11.49s/step, acc=0.4505, loss=2.3655, lr=8.90e-05]
Training:  24%|██▍       | 2652/11082 [8:31:25<27:13:51, 11.63s/step, acc=0.4505, loss=2.3655, lr=8.90e-05]
Training:  24%|██▍       | 2653/11082 [8:31:36<26:55:45, 11.50s/step, acc=0.4505, loss=2.3655, lr=8.90e-05]
Training:  24%|██▍       | 2654/11082 [8:31:48<27:03:09, 11.56s/step, acc=0.4505, loss=2.3655, lr=8.90e-05]
Training:  24%|██▍       | 2655/11082 [8:31:59<26:43:05, 11.41s/step, acc=0.4505, loss=2.3655, lr=8.90e-05]
Training:  24%|██▍       | 2656/11082 [8:32:11<27:19:52, 11.68s/step, acc=0.4505, loss=2.3655, lr=8.90e-05]
Training:  24%|██▍       | 2657/11082 [8:32:22<27:02:19, 11.55s/step, acc=0.4505, loss=2.3655, lr=8.90e-05]
Training:  24%|██▍       | 2658/11082 [8:32:34<27:19:18, 11.68s/step, acc=0.4505, loss=2.3655, lr=8.90e-05]
Training:  24%|██▍       | 2659/11082 [8:32:46<27:02:40, 11.56s/step, acc=0.4505, loss=2.3655, lr=8.90e-05]
Training:  24%|██▍       | 2660/11082 [8:32:57<27:09:21, 11.61s/step, acc=0.4505, loss=2.3655, lr=8.90e-05]
Training:  24%|██▍       | 2661/11082 [8:33:09<26:57:11, 11.52s/step, acc=0.4505, loss=2.3655, lr=8.90e-05]
Training:  24%|██▍       | 2662/11082 [8:33:20<27:05:23, 11.58s/step, acc=0.4505, loss=2.3655, lr=8.90e-05]
Training:  24%|██▍       | 2663/11082 [8:33:32<27:10:13, 11.62s/step, acc=0.4505, loss=2.3655, lr=8.90e-05]
Training:  24%|██▍       | 2664/11082 [8:33:43<26:57:42, 11.53s/step, acc=0.4505, loss=2.3655, lr=8.90e-05]
Training:  24%|██▍       | 2665/11082 [8:33:55<26:46:10, 11.45s/step, acc=0.4505, loss=2.3655, lr=8.90e-05]
Training:  24%|██▍       | 2666/11082 [8:34:07<27:09:44, 11.62s/step, acc=0.4505, loss=2.3655, lr=8.90e-05]
Training:  24%|██▍       | 2667/11082 [8:34:18<27:02:50, 11.57s/step, acc=0.4505, loss=2.3655, lr=8.90e-05]
Training:  24%|██▍       | 2668/11082 [8:34:29<26:52:15, 11.50s/step, acc=0.4505, loss=2.3655, lr=8.90e-05]
Training:  24%|██▍       | 2669/11082 [8:34:41<27:10:06, 11.63s/step, acc=0.4505, loss=2.3655, lr=8.90e-05]
Training:  24%|██▍       | 2670/11082 [8:34:53<26:55:47, 11.52s/step, acc=0.4505, loss=2.3655, lr=8.90e-05]
Training:  24%|██▍       | 2671/11082 [8:35:04<26:43:04, 11.44s/step, acc=0.4505, loss=2.3655, lr=8.90e-05]
Training:  24%|██▍       | 2672/11082 [8:35:16<27:06:19, 11.60s/step, acc=0.4505, loss=2.3655, lr=8.90e-05]
Training:  24%|██▍       | 2673/11082 [8:35:27<27:00:39, 11.56s/step, acc=0.4505, loss=2.3655, lr=8.90e-05]
Training:  24%|██▍       | 2674/11082 [8:35:39<26:57:24, 11.54s/step, acc=0.4505, loss=2.3655, lr=8.90e-05]
Training:  24%|██▍       | 2675/11082 [8:35:50<26:48:26, 11.48s/step, acc=0.4505, loss=2.3655, lr=8.90e-05]
Training:  24%|██▍       | 2676/11082 [8:36:02<26:59:19, 11.56s/step, acc=0.4505, loss=2.3655, lr=8.90e-05]
Training:  24%|██▍       | 2677/11082 [8:36:13<26:53:30, 11.52s/step, acc=0.4505, loss=2.3655, lr=8.90e-05]
Training:  24%|██▍       | 2678/11082 [8:36:25<27:17:17, 11.69s/step, acc=0.4505, loss=2.3655, lr=8.90e-05]
Training:  24%|██▍       | 2679/11082 [8:36:37<26:59:19, 11.56s/step, acc=0.4505, loss=2.3655, lr=8.90e-05]
Training:  24%|██▍       | 2680/11082 [8:36:48<27:05:08, 11.61s/step, acc=0.4505, loss=2.3655, lr=8.90e-05]
Training:  24%|██▍       | 2681/11082 [8:37:00<27:04:35, 11.60s/step, acc=0.4505, loss=2.3655, lr=8.90e-05]
Training:  24%|██▍       | 2682/11082 [8:37:11<26:58:28, 11.56s/step, acc=0.4505, loss=2.3655, lr=8.90e-05]
Training:  24%|██▍       | 2683/11082 [8:37:23<27:11:12, 11.65s/step, acc=0.4505, loss=2.3655, lr=8.90e-05]
Training:  24%|██▍       | 2684/11082 [8:37:35<27:09:44, 11.64s/step, acc=0.4505, loss=2.3655, lr=8.90e-05]
Training:  24%|██▍       | 2685/11082 [8:37:47<27:09:37, 11.64s/step, acc=0.4505, loss=2.3655, lr=8.90e-05]
Training:  24%|██▍       | 2686/11082 [8:37:58<27:00:21, 11.58s/step, acc=0.4505, loss=2.3655, lr=8.90e-05]
Training:  24%|██▍       | 2687/11082 [8:38:09<26:45:49, 11.48s/step, acc=0.4505, loss=2.3655, lr=8.90e-05]
Training:  24%|██▍       | 2688/11082 [8:38:21<27:03:48, 11.61s/step, acc=0.4505, loss=2.3655, lr=8.90e-05]
Training:  24%|██▍       | 2689/11082 [8:38:32<26:34:27, 11.40s/step, acc=0.4505, loss=2.3655, lr=8.90e-05]
Training:  24%|██▍       | 2690/11082 [8:38:44<26:50:19, 11.51s/step, acc=0.4505, loss=2.3655, lr=8.90e-05]
Training:  24%|██▍       | 2691/11082 [8:38:55<26:51:55, 11.53s/step, acc=0.4505, loss=2.3655, lr=8.90e-05]
Training:  24%|██▍       | 2692/11082 [8:39:07<26:57:57, 11.57s/step, acc=0.4505, loss=2.3655, lr=8.90e-05]
Training:  24%|██▍       | 2693/11082 [8:39:19<27:00:13, 11.59s/step, acc=0.4505, loss=2.3655, lr=8.90e-05]
Training:  24%|██▍       | 2694/11082 [8:39:30<26:47:20, 11.50s/step, acc=0.4505, loss=2.3655, lr=8.90e-05]
Training:  24%|██▍       | 2695/11082 [8:39:42<26:48:51, 11.51s/step, acc=0.4505, loss=2.3655, lr=8.90e-05]
Training:  24%|██▍       | 2696/11082 [8:39:53<27:04:05, 11.62s/step, acc=0.4505, loss=2.3655, lr=8.90e-05]
Training:  24%|██▍       | 2697/11082 [8:40:05<26:55:25, 11.56s/step, acc=0.4505, loss=2.3655, lr=8.90e-05]
Training:  24%|██▍       | 2698/11082 [8:40:16<26:52:04, 11.54s/step, acc=0.4505, loss=2.3655, lr=8.90e-05]
Training:  24%|██▍       | 2699/11082 [8:40:28<26:52:03, 11.54s/step, acc=0.4505, loss=2.3655, lr=8.90e-05]
Training:  24%|██▍       | 2700/11082 [8:40:39<26:50:49, 11.53s/step, acc=0.4505, loss=2.3655, lr=8.90e-05]
Training:  24%|██▍       | 2700/11082 [8:40:47<26:50:49, 11.53s/step, acc=0.4524, loss=2.3451, lr=8.86e-05]2026-06-18 23:34:15 | INFO | __main__ | epoch=0 step=2700 train_loss=2.345138 train_acc=0.452351 lr=8.855e-05 grad_norm=1.1484 acc_d0=0.7329 acc_d1=0.5750 acc_d2=0.4443 acc_d3=0.3403 acc_d4=0.2635 acc_d5=0.2087 acc_d6=0.1709 acc_d7=0.1436

Training:  24%|██▍       | 2701/11082 [8:40:51<27:08:59, 11.66s/step, acc=0.4524, loss=2.3451, lr=8.86e-05]
Training:  24%|██▍       | 2702/11082 [8:41:03<26:56:23, 11.57s/step, acc=0.4524, loss=2.3451, lr=8.86e-05]
Training:  24%|██▍       | 2703/11082 [8:41:14<26:59:24, 11.60s/step, acc=0.4524, loss=2.3451, lr=8.86e-05]
Training:  24%|██▍       | 2704/11082 [8:41:26<26:53:11, 11.55s/step, acc=0.4524, loss=2.3451, lr=8.86e-05]
Training:  24%|██▍       | 2705/11082 [8:41:37<26:43:34, 11.49s/step, acc=0.4524, loss=2.3451, lr=8.86e-05]
Training:  24%|██▍       | 2706/11082 [8:41:49<26:43:50, 11.49s/step, acc=0.4524, loss=2.3451, lr=8.86e-05]
Training:  24%|██▍       | 2707/11082 [8:42:00<26:42:31, 11.48s/step, acc=0.4524, loss=2.3451, lr=8.86e-05]
Training:  24%|██▍       | 2708/11082 [8:42:12<26:53:26, 11.56s/step, acc=0.4524, loss=2.3451, lr=8.86e-05]
Training:  24%|██▍       | 2709/11082 [8:42:23<26:48:01, 11.52s/step, acc=0.4524, loss=2.3451, lr=8.86e-05]
Training:  24%|██▍       | 2710/11082 [8:42:35<26:58:31, 11.60s/step, acc=0.4524, loss=2.3451, lr=8.86e-05]
Training:  24%|██▍       | 2711/11082 [8:42:47<26:54:54, 11.58s/step, acc=0.4524, loss=2.3451, lr=8.86e-05]
Training:  24%|██▍       | 2712/11082 [8:42:58<26:55:02, 11.58s/step, acc=0.4524, loss=2.3451, lr=8.86e-05]
Training:  24%|██▍       | 2713/11082 [8:43:10<26:53:47, 11.57s/step, acc=0.4524, loss=2.3451, lr=8.86e-05]
Training:  24%|██▍       | 2714/11082 [8:43:20<26:19:05, 11.32s/step, acc=0.4524, loss=2.3451, lr=8.86e-05]
Training:  24%|██▍       | 2715/11082 [8:43:33<26:54:59, 11.58s/step, acc=0.4524, loss=2.3451, lr=8.86e-05]
Training:  25%|██▍       | 2716/11082 [8:43:44<26:40:07, 11.48s/step, acc=0.4524, loss=2.3451, lr=8.86e-05]
Training:  25%|██▍       | 2717/11082 [8:43:55<26:39:49, 11.48s/step, acc=0.4524, loss=2.3451, lr=8.86e-05]
Training:  25%|██▍       | 2718/11082 [8:44:07<26:34:49, 11.44s/step, acc=0.4524, loss=2.3451, lr=8.86e-05]
Training:  25%|██▍       | 2719/11082 [8:44:19<26:58:52, 11.61s/step, acc=0.4524, loss=2.3451, lr=8.86e-05]
Training:  25%|██▍       | 2720/11082 [8:44:30<26:47:37, 11.54s/step, acc=0.4524, loss=2.3451, lr=8.86e-05]
Training:  25%|██▍       | 2721/11082 [8:44:41<26:29:28, 11.41s/step, acc=0.4524, loss=2.3451, lr=8.86e-05]
Training:  25%|██▍       | 2722/11082 [8:44:53<26:40:03, 11.48s/step, acc=0.4524, loss=2.3451, lr=8.86e-05]
Training:  25%|██▍       | 2723/11082 [8:45:06<27:44:08, 11.94s/step, acc=0.4524, loss=2.3451, lr=8.86e-05]
Training:  25%|██▍       | 2724/11082 [8:45:17<27:08:38, 11.69s/step, acc=0.4524, loss=2.3451, lr=8.86e-05]
Training:  25%|██▍       | 2725/11082 [8:45:29<27:31:04, 11.85s/step, acc=0.4524, loss=2.3451, lr=8.86e-05]
Training:  25%|██▍       | 2726/11082 [8:45:41<27:11:42, 11.72s/step, acc=0.4524, loss=2.3451, lr=8.86e-05]
Training:  25%|██▍       | 2727/11082 [8:45:53<27:18:03, 11.76s/step, acc=0.4524, loss=2.3451, lr=8.86e-05]
Training:  25%|██▍       | 2728/11082 [8:46:04<27:08:40, 11.70s/step, acc=0.4524, loss=2.3451, lr=8.86e-05]
Training:  25%|██▍       | 2729/11082 [8:46:15<26:57:44, 11.62s/step, acc=0.4524, loss=2.3451, lr=8.86e-05]
Training:  25%|██▍       | 2730/11082 [8:46:27<27:08:18, 11.70s/step, acc=0.4524, loss=2.3451, lr=8.86e-05]
Training:  25%|██▍       | 2731/11082 [8:46:39<26:49:26, 11.56s/step, acc=0.4524, loss=2.3451, lr=8.86e-05]
Training:  25%|██▍       | 2732/11082 [8:46:51<27:08:33, 11.70s/step, acc=0.4524, loss=2.3451, lr=8.86e-05]
Training:  25%|██▍       | 2733/11082 [8:47:02<27:04:38, 11.68s/step, acc=0.4524, loss=2.3451, lr=8.86e-05]
Training:  25%|██▍       | 2734/11082 [8:47:13<26:31:34, 11.44s/step, acc=0.4524, loss=2.3451, lr=8.86e-05]
Training:  25%|██▍       | 2735/11082 [8:47:25<26:46:57, 11.55s/step, acc=0.4524, loss=2.3451, lr=8.86e-05]
Training:  25%|██▍       | 2736/11082 [8:47:37<26:52:15, 11.59s/step, acc=0.4524, loss=2.3451, lr=8.86e-05]
Training:  25%|██▍       | 2737/11082 [8:47:48<26:50:27, 11.58s/step, acc=0.4524, loss=2.3451, lr=8.86e-05]
Training:  25%|██▍       | 2738/11082 [8:48:00<26:55:09, 11.61s/step, acc=0.4524, loss=2.3451, lr=8.86e-05]
Training:  25%|██▍       | 2739/11082 [8:48:11<26:47:17, 11.56s/step, acc=0.4524, loss=2.3451, lr=8.86e-05]
Training:  25%|██▍       | 2740/11082 [8:48:23<26:46:22, 11.55s/step, acc=0.4524, loss=2.3451, lr=8.86e-05]
Training:  25%|██▍       | 2741/11082 [8:48:35<26:55:54, 11.62s/step, acc=0.4524, loss=2.3451, lr=8.86e-05]
Training:  25%|██▍       | 2742/11082 [8:48:46<26:44:03, 11.54s/step, acc=0.4524, loss=2.3451, lr=8.86e-05]
Training:  25%|██▍       | 2743/11082 [8:48:58<26:47:26, 11.57s/step, acc=0.4524, loss=2.3451, lr=8.86e-05]
Training:  25%|██▍       | 2744/11082 [8:49:09<26:40:35, 11.52s/step, acc=0.4524, loss=2.3451, lr=8.86e-05]
Training:  25%|██▍       | 2745/11082 [8:49:21<26:48:20, 11.57s/step, acc=0.4524, loss=2.3451, lr=8.86e-05]
Training:  25%|██▍       | 2746/11082 [8:49:32<26:33:23, 11.47s/step, acc=0.4524, loss=2.3451, lr=8.86e-05]
Training:  25%|██▍       | 2747/11082 [8:49:44<26:53:19, 11.61s/step, acc=0.4524, loss=2.3451, lr=8.86e-05]
Training:  25%|██▍       | 2748/11082 [8:49:55<26:49:18, 11.59s/step, acc=0.4524, loss=2.3451, lr=8.86e-05]
Training:  25%|██▍       | 2749/11082 [8:50:07<26:46:28, 11.57s/step, acc=0.4524, loss=2.3451, lr=8.86e-05]
Training:  25%|██▍       | 2750/11082 [8:50:18<26:25:03, 11.41s/step, acc=0.4524, loss=2.3451, lr=8.86e-05]
Training:  25%|██▍       | 2750/11082 [8:50:26<26:25:03, 11.41s/step, acc=0.4513, loss=2.3562, lr=8.81e-05]2026-06-18 23:43:54 | INFO | __main__ | epoch=0 step=2750 train_loss=2.356186 train_acc=0.451332 lr=8.812e-05 grad_norm=1.0547 acc_d0=0.7280 acc_d1=0.5713 acc_d2=0.4408 acc_d3=0.3384 acc_d4=0.2638 acc_d5=0.2083 acc_d6=0.1695 acc_d7=0.1408

Training:  25%|██▍       | 2751/11082 [8:50:30<27:09:13, 11.73s/step, acc=0.4513, loss=2.3562, lr=8.81e-05]
Training:  25%|██▍       | 2752/11082 [8:50:42<26:40:29, 11.53s/step, acc=0.4513, loss=2.3562, lr=8.81e-05]
Training:  25%|██▍       | 2753/11082 [8:50:53<26:51:58, 11.61s/step, acc=0.4513, loss=2.3562, lr=8.81e-05]
Training:  25%|██▍       | 2754/11082 [8:51:05<26:57:03, 11.65s/step, acc=0.4513, loss=2.3562, lr=8.81e-05]
Training:  25%|██▍       | 2755/11082 [8:51:17<27:10:13, 11.75s/step, acc=0.4513, loss=2.3562, lr=8.81e-05]
Training:  25%|██▍       | 2756/11082 [8:51:28<26:44:34, 11.56s/step, acc=0.4513, loss=2.3562, lr=8.81e-05]
Training:  25%|██▍       | 2757/11082 [8:51:40<26:45:11, 11.57s/step, acc=0.4513, loss=2.3562, lr=8.81e-05]
Training:  25%|██▍       | 2758/11082 [8:51:51<26:47:51, 11.59s/step, acc=0.4513, loss=2.3562, lr=8.81e-05]
Training:  25%|██▍       | 2759/11082 [8:52:03<26:52:57, 11.63s/step, acc=0.4513, loss=2.3562, lr=8.81e-05]
Training:  25%|██▍       | 2760/11082 [8:52:15<26:44:48, 11.57s/step, acc=0.4513, loss=2.3562, lr=8.81e-05]
Training:  25%|██▍       | 2761/11082 [8:52:27<27:00:21, 11.68s/step, acc=0.4513, loss=2.3562, lr=8.81e-05]
Training:  25%|██▍       | 2762/11082 [8:52:37<26:30:50, 11.47s/step, acc=0.4513, loss=2.3562, lr=8.81e-05]
Training:  25%|██▍       | 2763/11082 [8:52:50<26:58:55, 11.68s/step, acc=0.4513, loss=2.3562, lr=8.81e-05]
Training:  25%|██▍       | 2764/11082 [8:53:01<26:40:21, 11.54s/step, acc=0.4513, loss=2.3562, lr=8.81e-05]
Training:  25%|██▍       | 2765/11082 [8:53:12<26:30:53, 11.48s/step, acc=0.4513, loss=2.3562, lr=8.81e-05]
Training:  25%|██▍       | 2766/11082 [8:53:24<26:32:36, 11.49s/step, acc=0.4513, loss=2.3562, lr=8.81e-05]
Training:  25%|██▍       | 2767/11082 [8:53:35<26:39:55, 11.54s/step, acc=0.4513, loss=2.3562, lr=8.81e-05]
Training:  25%|██▍       | 2768/11082 [8:53:47<27:02:07, 11.71s/step, acc=0.4513, loss=2.3562, lr=8.81e-05]
Training:  25%|██▍       | 2769/11082 [8:53:59<26:43:19, 11.57s/step, acc=0.4513, loss=2.3562, lr=8.81e-05]
Training:  25%|██▍       | 2770/11082 [8:54:10<26:39:04, 11.54s/step, acc=0.4513, loss=2.3562, lr=8.81e-05]
Training:  25%|██▌       | 2771/11082 [8:54:22<26:39:28, 11.55s/step, acc=0.4513, loss=2.3562, lr=8.81e-05]
Training:  25%|██▌       | 2772/11082 [8:54:33<26:29:21, 11.48s/step, acc=0.4513, loss=2.3562, lr=8.81e-05]
Training:  25%|██▌       | 2773/11082 [8:54:45<26:37:18, 11.53s/step, acc=0.4513, loss=2.3562, lr=8.81e-05]
Training:  25%|██▌       | 2774/11082 [8:54:56<26:41:23, 11.57s/step, acc=0.4513, loss=2.3562, lr=8.81e-05]
Training:  25%|██▌       | 2775/11082 [8:55:08<26:33:18, 11.51s/step, acc=0.4513, loss=2.3562, lr=8.81e-05]
Training:  25%|██▌       | 2776/11082 [8:55:19<26:18:11, 11.40s/step, acc=0.4513, loss=2.3562, lr=8.81e-05]
Training:  25%|██▌       | 2777/11082 [8:55:31<26:38:02, 11.55s/step, acc=0.4513, loss=2.3562, lr=8.81e-05]
Training:  25%|██▌       | 2778/11082 [8:55:42<26:38:18, 11.55s/step, acc=0.4513, loss=2.3562, lr=8.81e-05]
Training:  25%|██▌       | 2779/11082 [8:55:54<26:30:59, 11.50s/step, acc=0.4513, loss=2.3562, lr=8.81e-05]
Training:  25%|██▌       | 2780/11082 [8:56:05<26:35:43, 11.53s/step, acc=0.4513, loss=2.3562, lr=8.81e-05]
Training:  25%|██▌       | 2781/11082 [8:56:17<26:40:45, 11.57s/step, acc=0.4513, loss=2.3562, lr=8.81e-05]
Training:  25%|██▌       | 2782/11082 [8:56:28<26:36:36, 11.54s/step, acc=0.4513, loss=2.3562, lr=8.81e-05]
Training:  25%|██▌       | 2783/11082 [8:56:40<26:23:50, 11.45s/step, acc=0.4513, loss=2.3562, lr=8.81e-05]
Training:  25%|██▌       | 2784/11082 [8:56:51<26:36:52, 11.55s/step, acc=0.4513, loss=2.3562, lr=8.81e-05]
Training:  25%|██▌       | 2785/11082 [8:57:03<26:42:55, 11.59s/step, acc=0.4513, loss=2.3562, lr=8.81e-05]
Training:  25%|██▌       | 2786/11082 [8:57:14<26:31:07, 11.51s/step, acc=0.4513, loss=2.3562, lr=8.81e-05]
Training:  25%|██▌       | 2787/11082 [8:57:26<26:24:44, 11.46s/step, acc=0.4513, loss=2.3562, lr=8.81e-05]
Training:  25%|██▌       | 2788/11082 [8:57:37<26:22:58, 11.45s/step, acc=0.4513, loss=2.3562, lr=8.81e-05]
Training:  25%|██▌       | 2789/11082 [8:57:49<26:42:24, 11.59s/step, acc=0.4513, loss=2.3562, lr=8.81e-05]
Training:  25%|██▌       | 2790/11082 [8:58:01<26:48:02, 11.64s/step, acc=0.4513, loss=2.3562, lr=8.81e-05]
Training:  25%|██▌       | 2791/11082 [8:58:12<26:26:03, 11.48s/step, acc=0.4513, loss=2.3562, lr=8.81e-05]
Training:  25%|██▌       | 2792/11082 [8:58:24<26:46:35, 11.63s/step, acc=0.4513, loss=2.3562, lr=8.81e-05]
Training:  25%|██▌       | 2793/11082 [8:58:35<26:22:56, 11.46s/step, acc=0.4513, loss=2.3562, lr=8.81e-05]
Training:  25%|██▌       | 2794/11082 [8:58:47<26:35:40, 11.55s/step, acc=0.4513, loss=2.3562, lr=8.81e-05]
Training:  25%|██▌       | 2795/11082 [8:58:58<26:13:11, 11.39s/step, acc=0.4513, loss=2.3562, lr=8.81e-05]
Training:  25%|██▌       | 2796/11082 [8:59:10<26:31:08, 11.52s/step, acc=0.4513, loss=2.3562, lr=8.81e-05]
Training:  25%|██▌       | 2797/11082 [8:59:22<26:50:19, 11.66s/step, acc=0.4513, loss=2.3562, lr=8.81e-05]
Training:  25%|██▌       | 2798/11082 [8:59:33<26:34:13, 11.55s/step, acc=0.4513, loss=2.3562, lr=8.81e-05]
Training:  25%|██▌       | 2799/11082 [8:59:45<26:49:35, 11.66s/step, acc=0.4513, loss=2.3562, lr=8.81e-05]
Training:  25%|██▌       | 2800/11082 [8:59:56<26:37:19, 11.57s/step, acc=0.4513, loss=2.3562, lr=8.81e-05]
Training:  25%|██▌       | 2800/11082 [9:00:04<26:37:19, 11.57s/step, acc=0.4548, loss=2.3329, lr=8.77e-05]2026-06-18 23:53:31 | INFO | __main__ | epoch=0 step=2800 train_loss=2.332910 train_acc=0.454795 lr=8.768e-05 grad_norm=1.3672 acc_d0=0.7272 acc_d1=0.5723 acc_d2=0.4442 acc_d3=0.3431 acc_d4=0.2682 acc_d5=0.2141 acc_d6=0.1743 acc_d7=0.1455

Training:  25%|██▌       | 2801/11082 [9:00:08<26:55:24, 11.70s/step, acc=0.4548, loss=2.3329, lr=8.77e-05]
Training:  25%|██▌       | 2802/11082 [9:00:20<26:39:08, 11.59s/step, acc=0.4548, loss=2.3329, lr=8.77e-05]
Training:  25%|██▌       | 2803/11082 [9:00:32<26:53:54, 11.70s/step, acc=0.4548, loss=2.3329, lr=8.77e-05]
Training:  25%|██▌       | 2804/11082 [9:00:43<26:55:53, 11.71s/step, acc=0.4548, loss=2.3329, lr=8.77e-05]
Training:  25%|██▌       | 2805/11082 [9:00:55<26:41:31, 11.61s/step, acc=0.4548, loss=2.3329, lr=8.77e-05]
Training:  25%|██▌       | 2806/11082 [9:01:07<27:01:29, 11.76s/step, acc=0.4548, loss=2.3329, lr=8.77e-05]
Training:  25%|██▌       | 2807/11082 [9:01:18<26:41:25, 11.61s/step, acc=0.4548, loss=2.3329, lr=8.77e-05]
Training:  25%|██▌       | 2808/11082 [9:01:29<26:33:44, 11.56s/step, acc=0.4548, loss=2.3329, lr=8.77e-05]
Training:  25%|██▌       | 2809/11082 [9:01:41<26:20:24, 11.46s/step, acc=0.4548, loss=2.3329, lr=8.77e-05]
Training:  25%|██▌       | 2810/11082 [9:01:52<26:33:09, 11.56s/step, acc=0.4548, loss=2.3329, lr=8.77e-05]
Training:  25%|██▌       | 2811/11082 [9:02:04<26:35:56, 11.58s/step, acc=0.4548, loss=2.3329, lr=8.77e-05]
Training:  25%|██▌       | 2812/11082 [9:02:16<26:38:01, 11.59s/step, acc=0.4548, loss=2.3329, lr=8.77e-05]
Training:  25%|██▌       | 2813/11082 [9:02:27<26:43:45, 11.64s/step, acc=0.4548, loss=2.3329, lr=8.77e-05]
Training:  25%|██▌       | 2814/11082 [9:02:39<26:44:30, 11.64s/step, acc=0.4548, loss=2.3329, lr=8.77e-05]
Training:  25%|██▌       | 2815/11082 [9:02:51<26:38:00, 11.60s/step, acc=0.4548, loss=2.3329, lr=8.77e-05]
Training:  25%|██▌       | 2816/11082 [9:03:02<26:38:13, 11.60s/step, acc=0.4548, loss=2.3329, lr=8.77e-05]
Training:  25%|██▌       | 2817/11082 [9:03:13<26:22:36, 11.49s/step, acc=0.4548, loss=2.3329, lr=8.77e-05]
Training:  25%|██▌       | 2818/11082 [9:03:25<26:35:56, 11.59s/step, acc=0.4548, loss=2.3329, lr=8.77e-05]
Training:  25%|██▌       | 2819/11082 [9:03:37<26:38:16, 11.61s/step, acc=0.4548, loss=2.3329, lr=8.77e-05]
Training:  25%|██▌       | 2820/11082 [9:03:49<26:51:36, 11.70s/step, acc=0.4548, loss=2.3329, lr=8.77e-05]
Training:  25%|██▌       | 2821/11082 [9:04:00<26:30:15, 11.55s/step, acc=0.4548, loss=2.3329, lr=8.77e-05]
Training:  25%|██▌       | 2822/11082 [9:04:12<26:50:44, 11.70s/step, acc=0.4548, loss=2.3329, lr=8.77e-05]
Training:  25%|██▌       | 2823/11082 [9:04:23<26:37:29, 11.61s/step, acc=0.4548, loss=2.3329, lr=8.77e-05]
Training:  25%|██▌       | 2824/11082 [9:04:35<26:29:23, 11.55s/step, acc=0.4548, loss=2.3329, lr=8.77e-05]
Training:  25%|██▌       | 2825/11082 [9:04:47<26:33:16, 11.58s/step, acc=0.4548, loss=2.3329, lr=8.77e-05]
Training:  26%|██▌       | 2826/11082 [9:04:58<26:23:21, 11.51s/step, acc=0.4548, loss=2.3329, lr=8.77e-05]
Training:  26%|██▌       | 2827/11082 [9:05:09<26:23:32, 11.51s/step, acc=0.4548, loss=2.3329, lr=8.77e-05]
Training:  26%|██▌       | 2828/11082 [9:05:21<26:18:53, 11.48s/step, acc=0.4548, loss=2.3329, lr=8.77e-05]
Training:  26%|██▌       | 2829/11082 [9:05:33<26:28:39, 11.55s/step, acc=0.4548, loss=2.3329, lr=8.77e-05]
Training:  26%|██▌       | 2830/11082 [9:05:44<26:12:43, 11.44s/step, acc=0.4548, loss=2.3329, lr=8.77e-05]
Training:  26%|██▌       | 2831/11082 [9:05:55<26:24:51, 11.52s/step, acc=0.4548, loss=2.3329, lr=8.77e-05]
Training:  26%|██▌       | 2832/11082 [9:06:07<26:20:07, 11.49s/step, acc=0.4548, loss=2.3329, lr=8.77e-05]
Training:  26%|██▌       | 2833/11082 [9:06:19<26:27:53, 11.55s/step, acc=0.4548, loss=2.3329, lr=8.77e-05]
Training:  26%|██▌       | 2834/11082 [9:06:30<26:30:46, 11.57s/step, acc=0.4548, loss=2.3329, lr=8.77e-05]
Training:  26%|██▌       | 2835/11082 [9:06:42<26:35:34, 11.61s/step, acc=0.4548, loss=2.3329, lr=8.77e-05]
Training:  26%|██▌       | 2836/11082 [9:06:53<26:35:25, 11.61s/step, acc=0.4548, loss=2.3329, lr=8.77e-05]
Training:  26%|██▌       | 2837/11082 [9:07:05<26:18:28, 11.49s/step, acc=0.4548, loss=2.3329, lr=8.77e-05]
Training:  26%|██▌       | 2838/11082 [9:07:16<26:33:14, 11.60s/step, acc=0.4548, loss=2.3329, lr=8.77e-05]
Training:  26%|██▌       | 2839/11082 [9:07:28<26:19:37, 11.50s/step, acc=0.4548, loss=2.3329, lr=8.77e-05]
Training:  26%|██▌       | 2840/11082 [9:07:40<26:43:49, 11.68s/step, acc=0.4548, loss=2.3329, lr=8.77e-05]
Training:  26%|██▌       | 2841/11082 [9:07:51<26:35:07, 11.61s/step, acc=0.4548, loss=2.3329, lr=8.77e-05]
Training:  26%|██▌       | 2842/11082 [9:08:03<26:37:40, 11.63s/step, acc=0.4548, loss=2.3329, lr=8.77e-05]
Training:  26%|██▌       | 2843/11082 [9:08:15<26:36:21, 11.63s/step, acc=0.4548, loss=2.3329, lr=8.77e-05]
Training:  26%|██▌       | 2844/11082 [9:08:26<26:22:15, 11.52s/step, acc=0.4548, loss=2.3329, lr=8.77e-05]
Training:  26%|██▌       | 2845/11082 [9:08:38<26:30:19, 11.58s/step, acc=0.4548, loss=2.3329, lr=8.77e-05]
Training:  26%|██▌       | 2846/11082 [9:08:49<26:02:57, 11.39s/step, acc=0.4548, loss=2.3329, lr=8.77e-05]
Training:  26%|██▌       | 2847/11082 [9:09:00<26:24:25, 11.54s/step, acc=0.4548, loss=2.3329, lr=8.77e-05]
Training:  26%|██▌       | 2848/11082 [9:09:12<26:24:15, 11.54s/step, acc=0.4548, loss=2.3329, lr=8.77e-05]
Training:  26%|██▌       | 2849/11082 [9:09:24<26:26:07, 11.56s/step, acc=0.4548, loss=2.3329, lr=8.77e-05]
Training:  26%|██▌       | 2850/11082 [9:09:35<26:08:27, 11.43s/step, acc=0.4548, loss=2.3329, lr=8.77e-05]
Training:  26%|██▌       | 2850/11082 [9:09:42<26:08:27, 11.43s/step, acc=0.4547, loss=2.3331, lr=8.72e-05]2026-06-19 00:03:10 | INFO | __main__ | epoch=0 step=2850 train_loss=2.333117 train_acc=0.454694 lr=8.723e-05 grad_norm=1.0078 acc_d0=0.7308 acc_d1=0.5728 acc_d2=0.4425 acc_d3=0.3407 acc_d4=0.2650 acc_d5=0.2105 acc_d6=0.1711 acc_d7=0.1414

Training:  26%|██▌       | 2851/11082 [9:09:47<26:24:53, 11.55s/step, acc=0.4547, loss=2.3331, lr=8.72e-05]
Training:  26%|██▌       | 2852/11082 [9:09:58<26:05:08, 11.41s/step, acc=0.4547, loss=2.3331, lr=8.72e-05]
Training:  26%|██▌       | 2853/11082 [9:10:10<26:31:21, 11.60s/step, acc=0.4547, loss=2.3331, lr=8.72e-05]
Training:  26%|██▌       | 2854/11082 [9:10:21<26:36:57, 11.65s/step, acc=0.4547, loss=2.3331, lr=8.72e-05]
Training:  26%|██▌       | 2855/11082 [9:10:33<26:19:40, 11.52s/step, acc=0.4547, loss=2.3331, lr=8.72e-05]
Training:  26%|██▌       | 2856/11082 [9:10:44<26:22:17, 11.54s/step, acc=0.4547, loss=2.3331, lr=8.72e-05]
Training:  26%|██▌       | 2857/11082 [9:10:56<26:12:10, 11.47s/step, acc=0.4547, loss=2.3331, lr=8.72e-05]
Training:  26%|██▌       | 2858/11082 [9:11:07<26:10:43, 11.46s/step, acc=0.4547, loss=2.3331, lr=8.72e-05]
Training:  26%|██▌       | 2859/11082 [9:11:19<26:31:24, 11.61s/step, acc=0.4547, loss=2.3331, lr=8.72e-05]
Training:  26%|██▌       | 2860/11082 [9:11:31<26:36:06, 11.65s/step, acc=0.4547, loss=2.3331, lr=8.72e-05]
Training:  26%|██▌       | 2861/11082 [9:11:42<26:22:53, 11.55s/step, acc=0.4547, loss=2.3331, lr=8.72e-05]
Training:  26%|██▌       | 2862/11082 [9:11:53<26:09:28, 11.46s/step, acc=0.4547, loss=2.3331, lr=8.72e-05]
Training:  26%|██▌       | 2863/11082 [9:12:05<26:07:33, 11.44s/step, acc=0.4547, loss=2.3331, lr=8.72e-05]
Training:  26%|██▌       | 2864/11082 [9:12:16<26:19:56, 11.54s/step, acc=0.4547, loss=2.3331, lr=8.72e-05]
Training:  26%|██▌       | 2865/11082 [9:12:28<26:29:35, 11.61s/step, acc=0.4547, loss=2.3331, lr=8.72e-05]
Training:  26%|██▌       | 2866/11082 [9:12:40<26:19:20, 11.53s/step, acc=0.4547, loss=2.3331, lr=8.72e-05]
Training:  26%|██▌       | 2867/11082 [9:12:51<26:15:50, 11.51s/step, acc=0.4547, loss=2.3331, lr=8.72e-05]
Training:  26%|██▌       | 2868/11082 [9:13:02<26:11:28, 11.48s/step, acc=0.4547, loss=2.3331, lr=8.72e-05]
Training:  26%|██▌       | 2869/11082 [9:13:14<26:27:20, 11.60s/step, acc=0.4547, loss=2.3331, lr=8.72e-05]
Training:  26%|██▌       | 2870/11082 [9:13:25<26:10:41, 11.48s/step, acc=0.4547, loss=2.3331, lr=8.72e-05]
Training:  26%|██▌       | 2871/11082 [9:13:38<26:34:24, 11.65s/step, acc=0.4547, loss=2.3331, lr=8.72e-05]
Training:  26%|██▌       | 2872/11082 [9:13:49<26:21:12, 11.56s/step, acc=0.4547, loss=2.3331, lr=8.72e-05]
Training:  26%|██▌       | 2873/11082 [9:14:01<26:23:37, 11.57s/step, acc=0.4547, loss=2.3331, lr=8.72e-05]
Training:  26%|██▌       | 2874/11082 [9:14:12<26:11:19, 11.49s/step, acc=0.4547, loss=2.3331, lr=8.72e-05]
Training:  26%|██▌       | 2875/11082 [9:14:23<25:58:59, 11.40s/step, acc=0.4547, loss=2.3331, lr=8.72e-05]
Training:  26%|██▌       | 2876/11082 [9:14:35<26:37:17, 11.68s/step, acc=0.4547, loss=2.3331, lr=8.72e-05]
Training:  26%|██▌       | 2877/11082 [9:14:47<26:23:55, 11.58s/step, acc=0.4547, loss=2.3331, lr=8.72e-05]
Training:  26%|██▌       | 2878/11082 [9:14:58<26:19:24, 11.55s/step, acc=0.4547, loss=2.3331, lr=8.72e-05]
Training:  26%|██▌       | 2879/11082 [9:15:10<26:16:37, 11.53s/step, acc=0.4547, loss=2.3331, lr=8.72e-05]
Training:  26%|██▌       | 2880/11082 [9:15:22<26:37:57, 11.69s/step, acc=0.4547, loss=2.3331, lr=8.72e-05]
Training:  26%|██▌       | 2881/11082 [9:15:33<26:18:57, 11.55s/step, acc=0.4547, loss=2.3331, lr=8.72e-05]
Training:  26%|██▌       | 2882/11082 [9:15:44<25:59:54, 11.41s/step, acc=0.4547, loss=2.3331, lr=8.72e-05]
Training:  26%|██▌       | 2883/11082 [9:15:56<26:24:42, 11.60s/step, acc=0.4547, loss=2.3331, lr=8.72e-05]
Training:  26%|██▌       | 2884/11082 [9:16:07<26:15:08, 11.53s/step, acc=0.4547, loss=2.3331, lr=8.72e-05]
Training:  26%|██▌       | 2885/11082 [9:16:19<26:35:26, 11.68s/step, acc=0.4547, loss=2.3331, lr=8.72e-05]
Training:  26%|██▌       | 2886/11082 [9:16:31<26:40:53, 11.72s/step, acc=0.4547, loss=2.3331, lr=8.72e-05]
Training:  26%|██▌       | 2887/11082 [9:16:43<26:38:23, 11.70s/step, acc=0.4547, loss=2.3331, lr=8.72e-05]
Training:  26%|██▌       | 2888/11082 [9:16:54<26:19:23, 11.57s/step, acc=0.4547, loss=2.3331, lr=8.72e-05]
Training:  26%|██▌       | 2889/11082 [9:17:06<26:23:29, 11.60s/step, acc=0.4547, loss=2.3331, lr=8.72e-05]
Training:  26%|██▌       | 2890/11082 [9:17:18<26:27:38, 11.63s/step, acc=0.4547, loss=2.3331, lr=8.72e-05]
Training:  26%|██▌       | 2891/11082 [9:17:29<26:09:57, 11.50s/step, acc=0.4547, loss=2.3331, lr=8.72e-05]
Training:  26%|██▌       | 2892/11082 [9:17:41<26:26:38, 11.62s/step, acc=0.4547, loss=2.3331, lr=8.72e-05]
Training:  26%|██▌       | 2893/11082 [9:17:52<26:23:08, 11.60s/step, acc=0.4547, loss=2.3331, lr=8.72e-05]
Training:  26%|██▌       | 2894/11082 [9:18:04<26:22:30, 11.60s/step, acc=0.4547, loss=2.3331, lr=8.72e-05]
Training:  26%|██▌       | 2895/11082 [9:18:15<26:25:34, 11.62s/step, acc=0.4547, loss=2.3331, lr=8.72e-05]
Training:  26%|██▌       | 2896/11082 [9:18:27<26:30:03, 11.65s/step, acc=0.4547, loss=2.3331, lr=8.72e-05]
Training:  26%|██▌       | 2897/11082 [9:18:38<26:10:07, 11.51s/step, acc=0.4547, loss=2.3331, lr=8.72e-05]
Training:  26%|██▌       | 2898/11082 [9:18:51<26:37:54, 11.71s/step, acc=0.4547, loss=2.3331, lr=8.72e-05]
Training:  26%|██▌       | 2899/11082 [9:19:02<26:34:56, 11.69s/step, acc=0.4547, loss=2.3331, lr=8.72e-05]
Training:  26%|██▌       | 2900/11082 [9:19:14<26:30:04, 11.66s/step, acc=0.4547, loss=2.3331, lr=8.72e-05]
Training:  26%|██▌       | 2900/11082 [9:19:21<26:30:04, 11.66s/step, acc=0.4540, loss=2.3381, lr=8.68e-05]2026-06-19 00:12:49 | INFO | __main__ | epoch=0 step=2900 train_loss=2.338086 train_acc=0.453978 lr=8.677e-05 grad_norm=1.3672 acc_d0=0.7379 acc_d1=0.5839 acc_d2=0.4548 acc_d3=0.3530 acc_d4=0.2763 acc_d5=0.2189 acc_d6=0.1791 acc_d7=0.1492

Training:  26%|██▌       | 2901/11082 [9:19:26<26:39:14, 11.73s/step, acc=0.4540, loss=2.3381, lr=8.68e-05]
Training:  26%|██▌       | 2902/11082 [9:19:37<26:25:14, 11.63s/step, acc=0.4540, loss=2.3381, lr=8.68e-05]
Training:  26%|██▌       | 2903/11082 [9:19:49<26:25:50, 11.63s/step, acc=0.4540, loss=2.3381, lr=8.68e-05]
Training:  26%|██▌       | 2904/11082 [9:20:01<26:36:16, 11.71s/step, acc=0.4540, loss=2.3381, lr=8.68e-05]
Training:  26%|██▌       | 2905/11082 [9:20:12<26:22:10, 11.61s/step, acc=0.4540, loss=2.3381, lr=8.68e-05]
Training:  26%|██▌       | 2906/11082 [9:20:24<26:21:06, 11.60s/step, acc=0.4540, loss=2.3381, lr=8.68e-05]
Training:  26%|██▌       | 2907/11082 [9:20:34<25:45:26, 11.34s/step, acc=0.4540, loss=2.3381, lr=8.68e-05]
Training:  26%|██▌       | 2908/11082 [9:20:47<26:32:03, 11.69s/step, acc=0.4540, loss=2.3381, lr=8.68e-05]
Training:  26%|██▌       | 2909/11082 [9:20:59<26:40:07, 11.75s/step, acc=0.4540, loss=2.3381, lr=8.68e-05]
Training:  26%|██▋       | 2910/11082 [9:21:10<26:24:33, 11.63s/step, acc=0.4540, loss=2.3381, lr=8.68e-05]
Training:  26%|██▋       | 2911/11082 [9:21:22<26:23:51, 11.63s/step, acc=0.4540, loss=2.3381, lr=8.68e-05]
Training:  26%|██▋       | 2912/11082 [9:21:33<26:04:13, 11.49s/step, acc=0.4540, loss=2.3381, lr=8.68e-05]
Training:  26%|██▋       | 2913/11082 [9:21:45<26:16:00, 11.58s/step, acc=0.4540, loss=2.3381, lr=8.68e-05]
Training:  26%|██▋       | 2914/11082 [9:21:56<26:14:59, 11.57s/step, acc=0.4540, loss=2.3381, lr=8.68e-05]
Training:  26%|██▋       | 2915/11082 [9:22:08<26:06:07, 11.51s/step, acc=0.4540, loss=2.3381, lr=8.68e-05]
Training:  26%|██▋       | 2916/11082 [9:22:19<26:11:01, 11.54s/step, acc=0.4540, loss=2.3381, lr=8.68e-05]
Training:  26%|██▋       | 2917/11082 [9:22:30<26:00:30, 11.47s/step, acc=0.4540, loss=2.3381, lr=8.68e-05]
Training:  26%|██▋       | 2918/11082 [9:22:42<26:04:27, 11.50s/step, acc=0.4540, loss=2.3381, lr=8.68e-05]
Training:  26%|██▋       | 2919/11082 [9:22:54<26:12:47, 11.56s/step, acc=0.4540, loss=2.3381, lr=8.68e-05]
Training:  26%|██▋       | 2920/11082 [9:23:05<26:06:01, 11.51s/step, acc=0.4540, loss=2.3381, lr=8.68e-05]
Training:  26%|██▋       | 2921/11082 [9:23:17<26:05:37, 11.51s/step, acc=0.4540, loss=2.3381, lr=8.68e-05]
Training:  26%|██▋       | 2922/11082 [9:23:29<26:23:02, 11.64s/step, acc=0.4540, loss=2.3381, lr=8.68e-05]
Training:  26%|██▋       | 2923/11082 [9:23:40<26:27:34, 11.67s/step, acc=0.4540, loss=2.3381, lr=8.68e-05]
Training:  26%|██▋       | 2924/11082 [9:23:52<26:14:10, 11.58s/step, acc=0.4540, loss=2.3381, lr=8.68e-05]
Training:  26%|██▋       | 2925/11082 [9:24:03<26:15:09, 11.59s/step, acc=0.4540, loss=2.3381, lr=8.68e-05]
Training:  26%|██▋       | 2926/11082 [9:24:15<26:09:41, 11.55s/step, acc=0.4540, loss=2.3381, lr=8.68e-05]
Training:  26%|██▋       | 2927/11082 [9:24:26<26:10:33, 11.56s/step, acc=0.4540, loss=2.3381, lr=8.68e-05]
Training:  26%|██▋       | 2928/11082 [9:24:38<26:11:27, 11.56s/step, acc=0.4540, loss=2.3381, lr=8.68e-05]
Training:  26%|██▋       | 2929/11082 [9:24:50<26:16:54, 11.60s/step, acc=0.4540, loss=2.3381, lr=8.68e-05]
Training:  26%|██▋       | 2930/11082 [9:25:00<25:48:09, 11.39s/step, acc=0.4540, loss=2.3381, lr=8.68e-05]
Training:  26%|██▋       | 2931/11082 [9:25:12<25:50:47, 11.42s/step, acc=0.4540, loss=2.3381, lr=8.68e-05]
Training:  26%|██▋       | 2932/11082 [9:25:24<26:21:49, 11.65s/step, acc=0.4540, loss=2.3381, lr=8.68e-05]
Training:  26%|██▋       | 2933/11082 [9:25:35<26:01:10, 11.49s/step, acc=0.4540, loss=2.3381, lr=8.68e-05]
Training:  26%|██▋       | 2934/11082 [9:25:47<26:02:06, 11.50s/step, acc=0.4540, loss=2.3381, lr=8.68e-05]
Training:  26%|██▋       | 2935/11082 [9:25:58<26:03:23, 11.51s/step, acc=0.4540, loss=2.3381, lr=8.68e-05]
Training:  26%|██▋       | 2936/11082 [9:26:10<26:18:14, 11.62s/step, acc=0.4540, loss=2.3381, lr=8.68e-05]
Training:  27%|██▋       | 2937/11082 [9:26:22<26:06:18, 11.54s/step, acc=0.4540, loss=2.3381, lr=8.68e-05]
Training:  27%|██▋       | 2938/11082 [9:26:33<26:06:12, 11.54s/step, acc=0.4540, loss=2.3381, lr=8.68e-05]
Training:  27%|██▋       | 2939/11082 [9:26:45<26:05:17, 11.53s/step, acc=0.4540, loss=2.3381, lr=8.68e-05]
Training:  27%|██▋       | 2940/11082 [9:26:57<26:22:39, 11.66s/step, acc=0.4540, loss=2.3381, lr=8.68e-05]
Training:  27%|██▋       | 2941/11082 [9:27:08<26:10:33, 11.58s/step, acc=0.4540, loss=2.3381, lr=8.68e-05]
Training:  27%|██▋       | 2942/11082 [9:27:19<26:02:05, 11.51s/step, acc=0.4540, loss=2.3381, lr=8.68e-05]
Training:  27%|██▋       | 2943/11082 [9:27:31<25:55:48, 11.47s/step, acc=0.4540, loss=2.3381, lr=8.68e-05]
Training:  27%|██▋       | 2944/11082 [9:27:42<26:00:17, 11.50s/step, acc=0.4540, loss=2.3381, lr=8.68e-05]
Training:  27%|██▋       | 2945/11082 [9:27:54<26:17:41, 11.63s/step, acc=0.4540, loss=2.3381, lr=8.68e-05]
Training:  27%|██▋       | 2946/11082 [9:28:05<26:02:47, 11.52s/step, acc=0.4540, loss=2.3381, lr=8.68e-05]
Training:  27%|██▋       | 2947/11082 [9:28:17<26:13:50, 11.61s/step, acc=0.4540, loss=2.3381, lr=8.68e-05]
Training:  27%|██▋       | 2948/11082 [9:28:29<26:07:30, 11.56s/step, acc=0.4540, loss=2.3381, lr=8.68e-05]
Training:  27%|██▋       | 2949/11082 [9:28:40<26:08:01, 11.57s/step, acc=0.4540, loss=2.3381, lr=8.68e-05]
Training:  27%|██▋       | 2950/11082 [9:28:52<25:57:28, 11.49s/step, acc=0.4540, loss=2.3381, lr=8.68e-05]
Training:  27%|██▋       | 2950/11082 [9:28:59<25:57:28, 11.49s/step, acc=0.4564, loss=2.3279, lr=8.63e-05]2026-06-19 00:22:27 | INFO | __main__ | epoch=0 step=2950 train_loss=2.327921 train_acc=0.456362 lr=8.631e-05 grad_norm=0.9727 acc_d0=0.7319 acc_d1=0.5774 acc_d2=0.4493 acc_d3=0.3464 acc_d4=0.2705 acc_d5=0.2165 acc_d6=0.1742 acc_d7=0.1449

Training:  27%|██▋       | 2951/11082 [9:29:03<25:59:03, 11.50s/step, acc=0.4564, loss=2.3279, lr=8.63e-05]
Training:  27%|██▋       | 2952/11082 [9:29:15<26:08:30, 11.58s/step, acc=0.4564, loss=2.3279, lr=8.63e-05]
Training:  27%|██▋       | 2953/11082 [9:29:26<25:55:38, 11.48s/step, acc=0.4564, loss=2.3279, lr=8.63e-05]
Training:  27%|██▋       | 2954/11082 [9:29:38<26:04:29, 11.55s/step, acc=0.4564, loss=2.3279, lr=8.63e-05]
Training:  27%|██▋       | 2955/11082 [9:29:50<26:12:12, 11.61s/step, acc=0.4564, loss=2.3279, lr=8.63e-05]
Training:  27%|██▋       | 2956/11082 [9:30:01<25:56:16, 11.49s/step, acc=0.4564, loss=2.3279, lr=8.63e-05]
Training:  27%|██▋       | 2957/11082 [9:30:13<26:11:50, 11.61s/step, acc=0.4564, loss=2.3279, lr=8.63e-05]
Training:  27%|██▋       | 2958/11082 [9:30:24<25:56:56, 11.50s/step, acc=0.4564, loss=2.3279, lr=8.63e-05]
Training:  27%|██▋       | 2959/11082 [9:30:35<25:48:37, 11.44s/step, acc=0.4564, loss=2.3279, lr=8.63e-05]
Training:  27%|██▋       | 2960/11082 [9:30:47<26:08:31, 11.59s/step, acc=0.4564, loss=2.3279, lr=8.63e-05]
Training:  27%|██▋       | 2961/11082 [9:30:59<26:13:06, 11.62s/step, acc=0.4564, loss=2.3279, lr=8.63e-05]
Training:  27%|██▋       | 2962/11082 [9:31:10<25:58:49, 11.52s/step, acc=0.4564, loss=2.3279, lr=8.63e-05]
Training:  27%|██▋       | 2963/11082 [9:31:22<26:01:24, 11.54s/step, acc=0.4564, loss=2.3279, lr=8.63e-05]
Training:  27%|██▋       | 2964/11082 [9:31:33<25:59:05, 11.52s/step, acc=0.4564, loss=2.3279, lr=8.63e-05]
Training:  27%|██▋       | 2965/11082 [9:31:45<26:01:04, 11.54s/step, acc=0.4564, loss=2.3279, lr=8.63e-05]
Training:  27%|██▋       | 2966/11082 [9:31:57<26:14:47, 11.64s/step, acc=0.4564, loss=2.3279, lr=8.63e-05]
Training:  27%|██▋       | 2967/11082 [9:32:09<26:28:16, 11.74s/step, acc=0.4564, loss=2.3279, lr=8.63e-05]
Training:  27%|██▋       | 2968/11082 [9:32:20<26:04:55, 11.57s/step, acc=0.4564, loss=2.3279, lr=8.63e-05]
Training:  27%|██▋       | 2969/11082 [9:32:31<26:06:43, 11.59s/step, acc=0.4564, loss=2.3279, lr=8.63e-05]
Training:  27%|██▋       | 2970/11082 [9:32:43<26:02:10, 11.55s/step, acc=0.4564, loss=2.3279, lr=8.63e-05]
Training:  27%|██▋       | 2971/11082 [9:32:55<26:15:56, 11.66s/step, acc=0.4564, loss=2.3279, lr=8.63e-05]
Training:  27%|██▋       | 2972/11082 [9:33:06<25:55:45, 11.51s/step, acc=0.4564, loss=2.3279, lr=8.63e-05]
Training:  27%|██▋       | 2973/11082 [9:33:18<26:11:59, 11.63s/step, acc=0.4564, loss=2.3279, lr=8.63e-05]
Training:  27%|██▋       | 2974/11082 [9:33:29<26:03:38, 11.57s/step, acc=0.4564, loss=2.3279, lr=8.63e-05]
Training:  27%|██▋       | 2975/11082 [9:33:41<25:54:08, 11.50s/step, acc=0.4564, loss=2.3279, lr=8.63e-05]
Training:  27%|██▋       | 2976/11082 [9:33:53<26:29:08, 11.76s/step, acc=0.4564, loss=2.3279, lr=8.63e-05]
Training:  27%|██▋       | 2977/11082 [9:34:05<26:15:38, 11.66s/step, acc=0.4564, loss=2.3279, lr=8.63e-05]
Training:  27%|██▋       | 2978/11082 [9:34:16<26:04:03, 11.58s/step, acc=0.4564, loss=2.3279, lr=8.63e-05]
Training:  27%|██▋       | 2979/11082 [9:34:27<25:53:38, 11.50s/step, acc=0.4564, loss=2.3279, lr=8.63e-05]
Training:  27%|██▋       | 2980/11082 [9:34:39<26:04:28, 11.59s/step, acc=0.4564, loss=2.3279, lr=8.63e-05]
Training:  27%|██▋       | 2981/11082 [9:34:51<26:01:23, 11.56s/step, acc=0.4564, loss=2.3279, lr=8.63e-05]
Training:  27%|██▋       | 2982/11082 [9:35:03<26:19:22, 11.70s/step, acc=0.4564, loss=2.3279, lr=8.63e-05]
Training:  27%|██▋       | 2983/11082 [9:35:14<26:18:31, 11.69s/step, acc=0.4564, loss=2.3279, lr=8.63e-05]
Training:  27%|██▋       | 2984/11082 [9:35:25<25:51:39, 11.50s/step, acc=0.4564, loss=2.3279, lr=8.63e-05]
Training:  27%|██▋       | 2985/11082 [9:35:38<26:25:45, 11.75s/step, acc=0.4564, loss=2.3279, lr=8.63e-05]
Training:  27%|██▋       | 2986/11082 [9:35:49<25:57:27, 11.54s/step, acc=0.4564, loss=2.3279, lr=8.63e-05]
Training:  27%|██▋       | 2987/11082 [9:36:00<26:08:12, 11.62s/step, acc=0.4564, loss=2.3279, lr=8.63e-05]
Training:  27%|██▋       | 2988/11082 [9:36:12<25:58:04, 11.55s/step, acc=0.4564, loss=2.3279, lr=8.63e-05]
Training:  27%|██▋       | 2989/11082 [9:36:24<26:09:52, 11.64s/step, acc=0.4564, loss=2.3279, lr=8.63e-05]
Training:  27%|██▋       | 2990/11082 [9:36:35<26:16:36, 11.69s/step, acc=0.4564, loss=2.3279, lr=8.63e-05]
Training:  27%|██▋       | 2991/11082 [9:36:47<25:59:10, 11.56s/step, acc=0.4564, loss=2.3279, lr=8.63e-05]
Training:  27%|██▋       | 2992/11082 [9:36:58<26:04:25, 11.60s/step, acc=0.4564, loss=2.3279, lr=8.63e-05]
Training:  27%|██▋       | 2993/11082 [9:37:10<25:59:53, 11.57s/step, acc=0.4564, loss=2.3279, lr=8.63e-05]
Training:  27%|██▋       | 2994/11082 [9:37:21<25:53:54, 11.53s/step, acc=0.4564, loss=2.3279, lr=8.63e-05]
Training:  27%|██▋       | 2995/11082 [9:37:32<25:27:40, 11.33s/step, acc=0.4564, loss=2.3279, lr=8.63e-05]
Training:  27%|██▋       | 2996/11082 [9:37:44<26:00:59, 11.58s/step, acc=0.4564, loss=2.3279, lr=8.63e-05]
Training:  27%|██▋       | 2997/11082 [9:37:56<25:44:08, 11.46s/step, acc=0.4564, loss=2.3279, lr=8.63e-05]
Training:  27%|██▋       | 2998/11082 [9:38:07<25:39:41, 11.43s/step, acc=0.4564, loss=2.3279, lr=8.63e-05]
Training:  27%|██▋       | 2999/11082 [9:38:19<26:02:17, 11.60s/step, acc=0.4564, loss=2.3279, lr=8.63e-05]
Training:  27%|██▋       | 3000/11082 [9:38:30<25:57:37, 11.56s/step, acc=0.4564, loss=2.3279, lr=8.63e-05]
Training:  27%|██▋       | 3000/11082 [9:38:37<25:57:37, 11.56s/step, acc=0.4575, loss=2.3178, lr=8.58e-05]2026-06-19 00:32:05 | INFO | __main__ | epoch=0 step=3000 train_loss=2.317765 train_acc=0.457519 lr=8.584e-05 grad_norm=1.1406 acc_d0=0.7409 acc_d1=0.5847 acc_d2=0.4539 acc_d3=0.3498 acc_d4=0.2724 acc_d5=0.2151 acc_d6=0.1747 acc_d7=0.1448

Training:  27%|██▋       | 3001/11082 [9:38:42<25:47:34, 11.49s/step, acc=0.4575, loss=2.3178, lr=8.58e-05]
Training:  27%|██▋       | 3002/11082 [9:38:53<25:53:58, 11.54s/step, acc=0.4575, loss=2.3178, lr=8.58e-05]
Training:  27%|██▋       | 3003/11082 [9:39:04<25:34:09, 11.39s/step, acc=0.4575, loss=2.3178, lr=8.58e-05]
Training:  27%|██▋       | 3004/11082 [9:39:16<25:47:05, 11.49s/step, acc=0.4575, loss=2.3178, lr=8.58e-05]
Training:  27%|██▋       | 3005/11082 [9:39:28<25:48:17, 11.50s/step, acc=0.4575, loss=2.3178, lr=8.58e-05]
Training:  27%|██▋       | 3006/11082 [9:39:39<25:50:54, 11.52s/step, acc=0.4575, loss=2.3178, lr=8.58e-05]
Training:  27%|██▋       | 3007/11082 [9:39:51<25:51:34, 11.53s/step, acc=0.4575, loss=2.3178, lr=8.58e-05]
Training:  27%|██▋       | 3008/11082 [9:40:02<25:51:29, 11.53s/step, acc=0.4575, loss=2.3178, lr=8.58e-05]
Training:  27%|██▋       | 3009/11082 [9:40:13<25:30:53, 11.38s/step, acc=0.4575, loss=2.3178, lr=8.58e-05]
Training:  27%|██▋       | 3010/11082 [9:40:25<25:37:45, 11.43s/step, acc=0.4575, loss=2.3178, lr=8.58e-05]
Training:  27%|██▋       | 3011/11082 [9:40:37<26:03:33, 11.62s/step, acc=0.4575, loss=2.3178, lr=8.58e-05]
Training:  27%|██▋       | 3012/11082 [9:40:48<25:57:51, 11.58s/step, acc=0.4575, loss=2.3178, lr=8.58e-05]
Training:  27%|██▋       | 3013/11082 [9:41:00<26:02:39, 11.62s/step, acc=0.4575, loss=2.3178, lr=8.58e-05]
Training:  27%|██▋       | 3014/11082 [9:41:11<25:49:50, 11.53s/step, acc=0.4575, loss=2.3178, lr=8.58e-05]
Training:  27%|██▋       | 3015/11082 [9:41:23<26:00:06, 11.60s/step, acc=0.4575, loss=2.3178, lr=8.58e-05]
Training:  27%|██▋       | 3016/11082 [9:41:35<25:55:31, 11.57s/step, acc=0.4575, loss=2.3178, lr=8.58e-05]
Training:  27%|██▋       | 3017/11082 [9:41:46<25:53:15, 11.56s/step, acc=0.4575, loss=2.3178, lr=8.58e-05]
Training:  27%|██▋       | 3018/11082 [9:41:57<25:31:23, 11.39s/step, acc=0.4575, loss=2.3178, lr=8.58e-05]
Training:  27%|██▋       | 3019/11082 [9:42:09<25:43:08, 11.48s/step, acc=0.4575, loss=2.3178, lr=8.58e-05]
Training:  27%|██▋       | 3020/11082 [9:42:21<25:46:12, 11.51s/step, acc=0.4575, loss=2.3178, lr=8.58e-05]
Training:  27%|██▋       | 3021/11082 [9:42:32<25:38:56, 11.45s/step, acc=0.4575, loss=2.3178, lr=8.58e-05]
Training:  27%|██▋       | 3022/11082 [9:42:44<25:56:13, 11.58s/step, acc=0.4575, loss=2.3178, lr=8.58e-05]
Training:  27%|██▋       | 3023/11082 [9:42:55<25:49:56, 11.54s/step, acc=0.4575, loss=2.3178, lr=8.58e-05]
Training:  27%|██▋       | 3024/11082 [9:43:07<25:50:27, 11.54s/step, acc=0.4575, loss=2.3178, lr=8.58e-05]
Training:  27%|██▋       | 3025/11082 [9:43:18<25:43:23, 11.49s/step, acc=0.4575, loss=2.3178, lr=8.58e-05]
Training:  27%|██▋       | 3026/11082 [9:43:30<25:44:34, 11.50s/step, acc=0.4575, loss=2.3178, lr=8.58e-05]
Training:  27%|██▋       | 3027/11082 [9:43:42<26:07:08, 11.67s/step, acc=0.4575, loss=2.3178, lr=8.58e-05]
Training:  27%|██▋       | 3028/11082 [9:43:53<25:37:45, 11.46s/step, acc=0.4575, loss=2.3178, lr=8.58e-05]
Training:  27%|██▋       | 3029/11082 [9:44:05<25:58:13, 11.61s/step, acc=0.4575, loss=2.3178, lr=8.58e-05]
Training:  27%|██▋       | 3030/11082 [9:44:16<25:33:16, 11.43s/step, acc=0.4575, loss=2.3178, lr=8.58e-05]
Training:  27%|██▋       | 3031/11082 [9:44:27<25:44:38, 11.51s/step, acc=0.4575, loss=2.3178, lr=8.58e-05]
Training:  27%|██▋       | 3032/11082 [9:44:39<25:55:58, 11.60s/step, acc=0.4575, loss=2.3178, lr=8.58e-05]
Training:  27%|██▋       | 3033/11082 [9:44:50<25:42:34, 11.50s/step, acc=0.4575, loss=2.3178, lr=8.58e-05]
Training:  27%|██▋       | 3034/11082 [9:45:02<25:58:05, 11.62s/step, acc=0.4575, loss=2.3178, lr=8.58e-05]
Training:  27%|██▋       | 3035/11082 [9:45:13<25:33:51, 11.44s/step, acc=0.4575, loss=2.3178, lr=8.58e-05]
Training:  27%|██▋       | 3036/11082 [9:45:25<25:51:09, 11.57s/step, acc=0.4575, loss=2.3178, lr=8.58e-05]
Training:  27%|██▋       | 3037/11082 [9:45:37<25:40:20, 11.49s/step, acc=0.4575, loss=2.3178, lr=8.58e-05]
Training:  27%|██▋       | 3038/11082 [9:45:48<25:45:05, 11.52s/step, acc=0.4575, loss=2.3178, lr=8.58e-05]
Training:  27%|██▋       | 3039/11082 [9:46:00<25:39:22, 11.48s/step, acc=0.4575, loss=2.3178, lr=8.58e-05]
Training:  27%|██▋       | 3040/11082 [9:46:11<25:56:53, 11.62s/step, acc=0.4575, loss=2.3178, lr=8.58e-05]
Training:  27%|██▋       | 3041/11082 [9:46:23<25:50:12, 11.57s/step, acc=0.4575, loss=2.3178, lr=8.58e-05]
Training:  27%|██▋       | 3042/11082 [9:46:35<26:06:54, 11.69s/step, acc=0.4575, loss=2.3178, lr=8.58e-05]
Training:  27%|██▋       | 3043/11082 [9:46:46<25:53:30, 11.59s/step, acc=0.4575, loss=2.3178, lr=8.58e-05]
Training:  27%|██▋       | 3044/11082 [9:46:58<25:43:11, 11.52s/step, acc=0.4575, loss=2.3178, lr=8.58e-05]
Training:  27%|██▋       | 3045/11082 [9:47:10<26:03:03, 11.67s/step, acc=0.4575, loss=2.3178, lr=8.58e-05]
Training:  27%|██▋       | 3046/11082 [9:47:21<26:02:52, 11.67s/step, acc=0.4575, loss=2.3178, lr=8.58e-05]
Training:  27%|██▋       | 3047/11082 [9:47:33<26:01:10, 11.66s/step, acc=0.4575, loss=2.3178, lr=8.58e-05]
Training:  28%|██▊       | 3048/11082 [9:47:44<25:54:34, 11.61s/step, acc=0.4575, loss=2.3178, lr=8.58e-05]
Training:  28%|██▊       | 3049/11082 [9:47:56<25:46:58, 11.55s/step, acc=0.4575, loss=2.3178, lr=8.58e-05]
Training:  28%|██▊       | 3050/11082 [9:48:07<25:42:52, 11.53s/step, acc=0.4575, loss=2.3178, lr=8.58e-05]
Training:  28%|██▊       | 3050/11082 [9:48:15<25:42:52, 11.53s/step, acc=0.4611, loss=2.2981, lr=8.54e-05]2026-06-19 00:41:43 | INFO | __main__ | epoch=0 step=3050 train_loss=2.298071 train_acc=0.461107 lr=8.537e-05 grad_norm=2.0625 acc_d0=0.7295 acc_d1=0.5742 acc_d2=0.4474 acc_d3=0.3469 acc_d4=0.2712 acc_d5=0.2165 acc_d6=0.1770 acc_d7=0.1484

Training:  28%|██▊       | 3051/11082 [9:48:19<26:06:28, 11.70s/step, acc=0.4611, loss=2.2981, lr=8.54e-05]
Training:  28%|██▊       | 3052/11082 [9:48:31<25:47:06, 11.56s/step, acc=0.4611, loss=2.2981, lr=8.54e-05]
Training:  28%|██▊       | 3053/11082 [9:48:43<26:00:58, 11.67s/step, acc=0.4611, loss=2.2981, lr=8.54e-05]
Training:  28%|██▊       | 3054/11082 [9:48:54<25:47:35, 11.57s/step, acc=0.4611, loss=2.2981, lr=8.54e-05]
Training:  28%|██▊       | 3055/11082 [9:49:06<25:50:01, 11.59s/step, acc=0.4611, loss=2.2981, lr=8.54e-05]
Training:  28%|██▊       | 3056/11082 [9:49:17<25:48:44, 11.58s/step, acc=0.4611, loss=2.2981, lr=8.54e-05]
Training:  28%|██▊       | 3057/11082 [9:49:29<25:50:33, 11.59s/step, acc=0.4611, loss=2.2981, lr=8.54e-05]
Training:  28%|██▊       | 3058/11082 [9:49:40<25:54:29, 11.62s/step, acc=0.4611, loss=2.2981, lr=8.54e-05]
Training:  28%|██▊       | 3059/11082 [9:49:52<25:40:08, 11.52s/step, acc=0.4611, loss=2.2981, lr=8.54e-05]
Training:  28%|██▊       | 3060/11082 [9:50:03<25:50:12, 11.59s/step, acc=0.4611, loss=2.2981, lr=8.54e-05]
Training:  28%|██▊       | 3061/11082 [9:50:15<25:49:43, 11.59s/step, acc=0.4611, loss=2.2981, lr=8.54e-05]
Training:  28%|██▊       | 3062/11082 [9:50:27<25:48:30, 11.58s/step, acc=0.4611, loss=2.2981, lr=8.54e-05]
Training:  28%|██▊       | 3063/11082 [9:50:38<25:54:14, 11.63s/step, acc=0.4611, loss=2.2981, lr=8.54e-05]
Training:  28%|██▊       | 3064/11082 [9:50:50<25:52:21, 11.62s/step, acc=0.4611, loss=2.2981, lr=8.54e-05]
Training:  28%|██▊       | 3065/11082 [9:51:01<25:35:50, 11.49s/step, acc=0.4611, loss=2.2981, lr=8.54e-05]
Training:  28%|██▊       | 3066/11082 [9:51:13<25:46:07, 11.57s/step, acc=0.4611, loss=2.2981, lr=8.54e-05]
Training:  28%|██▊       | 3067/11082 [9:51:24<25:40:29, 11.53s/step, acc=0.4611, loss=2.2981, lr=8.54e-05]
Training:  28%|██▊       | 3068/11082 [9:51:36<25:38:12, 11.52s/step, acc=0.4611, loss=2.2981, lr=8.54e-05]
Training:  28%|██▊       | 3069/11082 [9:51:48<25:47:18, 11.59s/step, acc=0.4611, loss=2.2981, lr=8.54e-05]
Training:  28%|██▊       | 3070/11082 [9:51:59<25:42:11, 11.55s/step, acc=0.4611, loss=2.2981, lr=8.54e-05]
Training:  28%|██▊       | 3071/11082 [9:52:11<25:43:09, 11.56s/step, acc=0.4611, loss=2.2981, lr=8.54e-05]
Training:  28%|██▊       | 3072/11082 [9:52:22<25:31:44, 11.47s/step, acc=0.4611, loss=2.2981, lr=8.54e-05]
Training:  28%|██▊       | 3073/11082 [9:52:34<25:46:22, 11.58s/step, acc=0.4611, loss=2.2981, lr=8.54e-05]
Training:  28%|██▊       | 3074/11082 [9:52:45<25:39:36, 11.54s/step, acc=0.4611, loss=2.2981, lr=8.54e-05]
Training:  28%|██▊       | 3075/11082 [9:52:57<25:48:32, 11.60s/step, acc=0.4611, loss=2.2981, lr=8.54e-05]
Training:  28%|██▊       | 3076/11082 [9:53:08<25:42:07, 11.56s/step, acc=0.4611, loss=2.2981, lr=8.54e-05]
Training:  28%|██▊       | 3077/11082 [9:53:20<25:49:10, 11.61s/step, acc=0.4611, loss=2.2981, lr=8.54e-05]
Training:  28%|██▊       | 3078/11082 [9:53:32<25:44:23, 11.58s/step, acc=0.4611, loss=2.2981, lr=8.54e-05]
Training:  28%|██▊       | 3079/11082 [9:53:43<25:25:24, 11.44s/step, acc=0.4611, loss=2.2981, lr=8.54e-05]
Training:  28%|██▊       | 3080/11082 [9:53:55<25:57:28, 11.68s/step, acc=0.4611, loss=2.2981, lr=8.54e-05]
Training:  28%|██▊       | 3081/11082 [9:54:06<25:50:16, 11.63s/step, acc=0.4611, loss=2.2981, lr=8.54e-05]
Training:  28%|██▊       | 3082/11082 [9:54:17<25:20:23, 11.40s/step, acc=0.4611, loss=2.2981, lr=8.54e-05]
Training:  28%|██▊       | 3083/11082 [9:54:29<25:50:17, 11.63s/step, acc=0.4611, loss=2.2981, lr=8.54e-05]
Training:  28%|██▊       | 3084/11082 [9:54:41<25:28:56, 11.47s/step, acc=0.4611, loss=2.2981, lr=8.54e-05]
Training:  28%|██▊       | 3085/11082 [9:54:52<25:36:20, 11.53s/step, acc=0.4611, loss=2.2981, lr=8.54e-05]
Training:  28%|██▊       | 3086/11082 [9:55:04<25:33:29, 11.51s/step, acc=0.4611, loss=2.2981, lr=8.54e-05]
Training:  28%|██▊       | 3087/11082 [9:55:15<25:41:00, 11.56s/step, acc=0.4611, loss=2.2981, lr=8.54e-05]
Training:  28%|██▊       | 3088/11082 [9:55:27<25:31:43, 11.50s/step, acc=0.4611, loss=2.2981, lr=8.54e-05]
Training:  28%|██▊       | 3089/11082 [9:55:38<25:36:13, 11.53s/step, acc=0.4611, loss=2.2981, lr=8.54e-05]
Training:  28%|██▊       | 3090/11082 [9:55:50<25:43:50, 11.59s/step, acc=0.4611, loss=2.2981, lr=8.54e-05]
Training:  28%|██▊       | 3091/11082 [9:56:01<25:35:11, 11.53s/step, acc=0.4611, loss=2.2981, lr=8.54e-05]
Training:  28%|██▊       | 3092/11082 [9:56:13<25:17:27, 11.40s/step, acc=0.4611, loss=2.2981, lr=8.54e-05]
Training:  28%|██▊       | 3093/11082 [9:56:24<25:39:48, 11.56s/step, acc=0.4611, loss=2.2981, lr=8.54e-05]
Training:  28%|██▊       | 3094/11082 [9:56:36<25:25:45, 11.46s/step, acc=0.4611, loss=2.2981, lr=8.54e-05]
Training:  28%|██▊       | 3095/11082 [9:56:47<25:27:20, 11.47s/step, acc=0.4611, loss=2.2981, lr=8.54e-05]
Training:  28%|██▊       | 3096/11082 [9:56:59<25:38:51, 11.56s/step, acc=0.4611, loss=2.2981, lr=8.54e-05]
Training:  28%|██▊       | 3097/11082 [9:57:11<25:45:21, 11.61s/step, acc=0.4611, loss=2.2981, lr=8.54e-05]
Training:  28%|██▊       | 3098/11082 [9:57:22<25:41:18, 11.58s/step, acc=0.4611, loss=2.2981, lr=8.54e-05]
Training:  28%|██▊       | 3099/11082 [9:57:34<25:33:26, 11.53s/step, acc=0.4611, loss=2.2981, lr=8.54e-05]
Training:  28%|██▊       | 3100/11082 [9:57:45<25:30:29, 11.50s/step, acc=0.4611, loss=2.2981, lr=8.54e-05]
Training:  28%|██▊       | 3100/11082 [9:57:52<25:30:29, 11.50s/step, acc=0.4623, loss=2.2850, lr=8.49e-05]2026-06-19 00:51:20 | INFO | __main__ | epoch=0 step=3100 train_loss=2.285034 train_acc=0.462346 lr=8.489e-05 grad_norm=1.6172 acc_d0=0.7399 acc_d1=0.5858 acc_d2=0.4563 acc_d3=0.3526 acc_d4=0.2747 acc_d5=0.2174 acc_d6=0.1754 acc_d7=0.1473

Training:  28%|██▊       | 3101/11082 [9:57:57<25:31:46, 11.52s/step, acc=0.4623, loss=2.2850, lr=8.49e-05]
Training:  28%|██▊       | 3102/11082 [9:58:08<25:23:14, 11.45s/step, acc=0.4623, loss=2.2850, lr=8.49e-05]
Training:  28%|██▊       | 3103/11082 [9:58:20<25:58:11, 11.72s/step, acc=0.4623, loss=2.2850, lr=8.49e-05]
Training:  28%|██▊       | 3104/11082 [9:58:32<25:39:19, 11.58s/step, acc=0.4623, loss=2.2850, lr=8.49e-05]
Training:  28%|██▊       | 3105/11082 [9:58:43<25:37:15, 11.56s/step, acc=0.4623, loss=2.2850, lr=8.49e-05]
Training:  28%|██▊       | 3106/11082 [9:58:55<25:43:01, 11.61s/step, acc=0.4623, loss=2.2850, lr=8.49e-05]
Training:  28%|██▊       | 3107/11082 [9:59:06<25:26:30, 11.48s/step, acc=0.4623, loss=2.2850, lr=8.49e-05]
Training:  28%|██▊       | 3108/11082 [9:59:17<25:20:51, 11.44s/step, acc=0.4623, loss=2.2850, lr=8.49e-05]
Training:  28%|██▊       | 3109/11082 [9:59:29<25:47:26, 11.65s/step, acc=0.4623, loss=2.2850, lr=8.49e-05]
Training:  28%|██▊       | 3110/11082 [9:59:41<25:30:35, 11.52s/step, acc=0.4623, loss=2.2850, lr=8.49e-05]
Training:  28%|██▊       | 3111/11082 [9:59:53<25:46:47, 11.64s/step, acc=0.4623, loss=2.2850, lr=8.49e-05]
Training:  28%|██▊       | 3112/11082 [10:00:04<25:29:33, 11.51s/step, acc=0.4623, loss=2.2850, lr=8.49e-05]
Training:  28%|██▊       | 3113/11082 [10:00:15<25:30:53, 11.53s/step, acc=0.4623, loss=2.2850, lr=8.49e-05]
Training:  28%|██▊       | 3114/11082 [10:00:27<25:26:38, 11.50s/step, acc=0.4623, loss=2.2850, lr=8.49e-05]
Training:  28%|██▊       | 3115/11082 [10:00:39<25:42:07, 11.61s/step, acc=0.4623, loss=2.2850, lr=8.49e-05]
Training:  28%|██▊       | 3116/11082 [10:00:50<25:46:57, 11.65s/step, acc=0.4623, loss=2.2850, lr=8.49e-05]
Training:  28%|██▊       | 3117/11082 [10:01:01<25:22:13, 11.47s/step, acc=0.4623, loss=2.2850, lr=8.49e-05]
Training:  28%|██▊       | 3118/11082 [10:01:13<25:29:08, 11.52s/step, acc=0.4623, loss=2.2850, lr=8.49e-05]
Training:  28%|██▊       | 3119/11082 [10:01:25<25:36:09, 11.57s/step, acc=0.4623, loss=2.2850, lr=8.49e-05]
Training:  28%|██▊       | 3120/11082 [10:01:36<25:35:15, 11.57s/step, acc=0.4623, loss=2.2850, lr=8.49e-05]
Training:  28%|██▊       | 3121/11082 [10:01:48<25:37:21, 11.59s/step, acc=0.4623, loss=2.2850, lr=8.49e-05]
Training:  28%|██▊       | 3122/11082 [10:01:59<25:34:37, 11.57s/step, acc=0.4623, loss=2.2850, lr=8.49e-05]
Training:  28%|██▊       | 3123/11082 [10:02:11<25:32:43, 11.55s/step, acc=0.4623, loss=2.2850, lr=8.49e-05]
Training:  28%|██▊       | 3124/11082 [10:02:22<25:14:07, 11.42s/step, acc=0.4623, loss=2.2850, lr=8.49e-05]
Training:  28%|██▊       | 3125/11082 [10:02:34<25:48:08, 11.67s/step, acc=0.4623, loss=2.2850, lr=8.49e-05]
Training:  28%|██▊       | 3126/11082 [10:02:46<25:32:29, 11.56s/step, acc=0.4623, loss=2.2850, lr=8.49e-05]
Training:  28%|██▊       | 3127/11082 [10:02:57<25:32:02, 11.56s/step, acc=0.4623, loss=2.2850, lr=8.49e-05]
Training:  28%|██▊       | 3128/11082 [10:03:09<25:49:41, 11.69s/step, acc=0.4623, loss=2.2850, lr=8.49e-05]
Training:  28%|██▊       | 3129/11082 [10:03:21<25:47:57, 11.68s/step, acc=0.4623, loss=2.2850, lr=8.49e-05]
Training:  28%|██▊       | 3130/11082 [10:03:33<25:50:17, 11.70s/step, acc=0.4623, loss=2.2850, lr=8.49e-05]
Training:  28%|██▊       | 3131/11082 [10:03:44<25:29:29, 11.54s/step, acc=0.4623, loss=2.2850, lr=8.49e-05]
Training:  28%|██▊       | 3132/11082 [10:03:56<25:54:17, 11.73s/step, acc=0.4623, loss=2.2850, lr=8.49e-05]
Training:  28%|██▊       | 3133/11082 [10:04:07<25:41:56, 11.64s/step, acc=0.4623, loss=2.2850, lr=8.49e-05]
Training:  28%|██▊       | 3134/11082 [10:04:19<25:43:53, 11.65s/step, acc=0.4623, loss=2.2850, lr=8.49e-05]
Training:  28%|██▊       | 3135/11082 [10:04:30<25:30:03, 11.55s/step, acc=0.4623, loss=2.2850, lr=8.49e-05]
Training:  28%|██▊       | 3136/11082 [10:04:42<25:30:55, 11.56s/step, acc=0.4623, loss=2.2850, lr=8.49e-05]
Training:  28%|██▊       | 3137/11082 [10:04:54<25:37:42, 11.61s/step, acc=0.4623, loss=2.2850, lr=8.49e-05]
Training:  28%|██▊       | 3138/11082 [10:05:05<25:33:22, 11.58s/step, acc=0.4623, loss=2.2850, lr=8.49e-05]
Training:  28%|██▊       | 3139/11082 [10:05:17<25:49:00, 11.70s/step, acc=0.4623, loss=2.2850, lr=8.49e-05]
Training:  28%|██▊       | 3140/11082 [10:05:29<25:50:12, 11.71s/step, acc=0.4623, loss=2.2850, lr=8.49e-05]
Training:  28%|██▊       | 3141/11082 [10:05:40<25:08:06, 11.39s/step, acc=0.4623, loss=2.2850, lr=8.49e-05]
Training:  28%|██▊       | 3142/11082 [10:05:52<25:46:04, 11.68s/step, acc=0.4623, loss=2.2850, lr=8.49e-05]
Training:  28%|██▊       | 3143/11082 [10:06:03<25:14:22, 11.45s/step, acc=0.4623, loss=2.2850, lr=8.49e-05]
Training:  28%|██▊       | 3144/11082 [10:06:16<26:02:56, 11.81s/step, acc=0.4623, loss=2.2850, lr=8.49e-05]
Training:  28%|██▊       | 3145/11082 [10:06:27<25:54:38, 11.75s/step, acc=0.4623, loss=2.2850, lr=8.49e-05]
Training:  28%|██▊       | 3146/11082 [10:06:38<25:39:57, 11.64s/step, acc=0.4623, loss=2.2850, lr=8.49e-05]
Training:  28%|██▊       | 3147/11082 [10:06:50<25:23:58, 11.52s/step, acc=0.4623, loss=2.2850, lr=8.49e-05]
Training:  28%|██▊       | 3148/11082 [10:07:02<25:33:27, 11.60s/step, acc=0.4623, loss=2.2850, lr=8.49e-05]
Training:  28%|██▊       | 3149/11082 [10:07:13<25:35:14, 11.61s/step, acc=0.4623, loss=2.2850, lr=8.49e-05]
Training:  28%|██▊       | 3150/11082 [10:07:25<25:32:27, 11.59s/step, acc=0.4623, loss=2.2850, lr=8.49e-05]
Training:  28%|██▊       | 3150/11082 [10:07:32<25:32:27, 11.59s/step, acc=0.4635, loss=2.2827, lr=8.44e-05]2026-06-19 01:01:00 | INFO | __main__ | epoch=0 step=3150 train_loss=2.282695 train_acc=0.463455 lr=8.440e-05 grad_norm=1.1172 acc_d0=0.7284 acc_d1=0.5691 acc_d2=0.4398 acc_d3=0.3379 acc_d4=0.2638 acc_d5=0.2105 acc_d6=0.1721 acc_d7=0.1443

Training:  28%|██▊       | 3151/11082 [10:07:37<25:42:35, 11.67s/step, acc=0.4635, loss=2.2827, lr=8.44e-05]
Training:  28%|██▊       | 3152/11082 [10:07:48<25:14:30, 11.46s/step, acc=0.4635, loss=2.2827, lr=8.44e-05]
Training:  28%|██▊       | 3153/11082 [10:07:59<25:27:53, 11.56s/step, acc=0.4635, loss=2.2827, lr=8.44e-05]
Training:  28%|██▊       | 3154/11082 [10:08:11<25:29:58, 11.58s/step, acc=0.4635, loss=2.2827, lr=8.44e-05]
Training:  28%|██▊       | 3155/11082 [10:08:23<25:31:31, 11.59s/step, acc=0.4635, loss=2.2827, lr=8.44e-05]
Training:  28%|██▊       | 3156/11082 [10:08:34<25:17:34, 11.49s/step, acc=0.4635, loss=2.2827, lr=8.44e-05]
Training:  28%|██▊       | 3157/11082 [10:08:46<25:37:56, 11.64s/step, acc=0.4635, loss=2.2827, lr=8.44e-05]
Training:  28%|██▊       | 3158/11082 [10:08:57<25:23:19, 11.53s/step, acc=0.4635, loss=2.2827, lr=8.44e-05]
Training:  29%|██▊       | 3159/11082 [10:09:09<25:26:44, 11.56s/step, acc=0.4635, loss=2.2827, lr=8.44e-05]
Training:  29%|██▊       | 3160/11082 [10:09:20<25:20:58, 11.52s/step, acc=0.4635, loss=2.2827, lr=8.44e-05]
Training:  29%|██▊       | 3161/11082 [10:09:32<25:15:03, 11.48s/step, acc=0.4635, loss=2.2827, lr=8.44e-05]
Training:  29%|██▊       | 3162/11082 [10:09:43<25:11:50, 11.45s/step, acc=0.4635, loss=2.2827, lr=8.44e-05]
Training:  29%|██▊       | 3163/11082 [10:09:55<25:20:52, 11.52s/step, acc=0.4635, loss=2.2827, lr=8.44e-05]
Training:  29%|██▊       | 3164/11082 [10:10:06<25:33:14, 11.62s/step, acc=0.4635, loss=2.2827, lr=8.44e-05]
Training:  29%|██▊       | 3165/11082 [10:10:18<25:28:52, 11.59s/step, acc=0.4635, loss=2.2827, lr=8.44e-05]
Training:  29%|██▊       | 3166/11082 [10:10:30<25:27:42, 11.58s/step, acc=0.4635, loss=2.2827, lr=8.44e-05]
Training:  29%|██▊       | 3167/11082 [10:10:41<25:15:27, 11.49s/step, acc=0.4635, loss=2.2827, lr=8.44e-05]
Training:  29%|██▊       | 3168/11082 [10:10:52<25:11:58, 11.46s/step, acc=0.4635, loss=2.2827, lr=8.44e-05]
Training:  29%|██▊       | 3169/11082 [10:11:04<25:20:08, 11.53s/step, acc=0.4635, loss=2.2827, lr=8.44e-05]
Training:  29%|██▊       | 3170/11082 [10:11:15<25:10:07, 11.45s/step, acc=0.4635, loss=2.2827, lr=8.44e-05]
Training:  29%|██▊       | 3171/11082 [10:11:27<25:26:37, 11.58s/step, acc=0.4635, loss=2.2827, lr=8.44e-05]
Training:  29%|██▊       | 3172/11082 [10:11:39<25:35:17, 11.65s/step, acc=0.4635, loss=2.2827, lr=8.44e-05]
Training:  29%|██▊       | 3173/11082 [10:11:50<25:26:51, 11.58s/step, acc=0.4635, loss=2.2827, lr=8.44e-05]
Training:  29%|██▊       | 3174/11082 [10:12:01<24:58:31, 11.37s/step, acc=0.4635, loss=2.2827, lr=8.44e-05]
Training:  29%|██▊       | 3175/11082 [10:12:13<25:14:14, 11.49s/step, acc=0.4635, loss=2.2827, lr=8.44e-05]
Training:  29%|██▊       | 3176/11082 [10:12:25<25:33:18, 11.64s/step, acc=0.4635, loss=2.2827, lr=8.44e-05]
Training:  29%|██▊       | 3177/11082 [10:12:36<25:23:56, 11.57s/step, acc=0.4635, loss=2.2827, lr=8.44e-05]
Training:  29%|██▊       | 3178/11082 [10:12:48<25:18:45, 11.53s/step, acc=0.4635, loss=2.2827, lr=8.44e-05]
Training:  29%|██▊       | 3179/11082 [10:13:00<25:28:33, 11.60s/step, acc=0.4635, loss=2.2827, lr=8.44e-05]
Training:  29%|██▊       | 3180/11082 [10:13:11<25:14:04, 11.50s/step, acc=0.4635, loss=2.2827, lr=8.44e-05]
Training:  29%|██▊       | 3181/11082 [10:13:23<25:33:04, 11.64s/step, acc=0.4635, loss=2.2827, lr=8.44e-05]
Training:  29%|██▊       | 3182/11082 [10:13:34<25:20:03, 11.54s/step, acc=0.4635, loss=2.2827, lr=8.44e-05]
Training:  29%|██▊       | 3183/11082 [10:13:46<25:17:32, 11.53s/step, acc=0.4635, loss=2.2827, lr=8.44e-05]
Training:  29%|██▊       | 3184/11082 [10:13:57<25:08:15, 11.46s/step, acc=0.4635, loss=2.2827, lr=8.44e-05]
Training:  29%|██▊       | 3185/11082 [10:14:08<25:12:52, 11.49s/step, acc=0.4635, loss=2.2827, lr=8.44e-05]
Training:  29%|██▊       | 3186/11082 [10:14:20<25:08:27, 11.46s/step, acc=0.4635, loss=2.2827, lr=8.44e-05]
Training:  29%|██▉       | 3187/11082 [10:14:32<25:33:58, 11.66s/step, acc=0.4635, loss=2.2827, lr=8.44e-05]
Training:  29%|██▉       | 3188/11082 [10:14:43<25:25:28, 11.59s/step, acc=0.4635, loss=2.2827, lr=8.44e-05]
Training:  29%|██▉       | 3189/11082 [10:14:55<25:15:36, 11.52s/step, acc=0.4635, loss=2.2827, lr=8.44e-05]
Training:  29%|██▉       | 3190/11082 [10:15:07<25:33:00, 11.65s/step, acc=0.4635, loss=2.2827, lr=8.44e-05]
Training:  29%|██▉       | 3191/11082 [10:15:18<25:23:01, 11.58s/step, acc=0.4635, loss=2.2827, lr=8.44e-05]
Training:  29%|██▉       | 3192/11082 [10:15:29<25:06:57, 11.46s/step, acc=0.4635, loss=2.2827, lr=8.44e-05]
Training:  29%|██▉       | 3193/11082 [10:15:41<25:22:12, 11.58s/step, acc=0.4635, loss=2.2827, lr=8.44e-05]
Training:  29%|██▉       | 3194/11082 [10:15:53<25:15:43, 11.53s/step, acc=0.4635, loss=2.2827, lr=8.44e-05]
Training:  29%|██▉       | 3195/11082 [10:16:04<25:06:36, 11.46s/step, acc=0.4635, loss=2.2827, lr=8.44e-05]
Training:  29%|██▉       | 3196/11082 [10:16:16<25:18:57, 11.56s/step, acc=0.4635, loss=2.2827, lr=8.44e-05]
Training:  29%|██▉       | 3197/11082 [10:16:27<25:19:59, 11.57s/step, acc=0.4635, loss=2.2827, lr=8.44e-05]
Training:  29%|██▉       | 3198/11082 [10:16:39<25:16:38, 11.54s/step, acc=0.4635, loss=2.2827, lr=8.44e-05]
Training:  29%|██▉       | 3199/11082 [10:16:50<25:15:24, 11.53s/step, acc=0.4635, loss=2.2827, lr=8.44e-05]
Training:  29%|██▉       | 3200/11082 [10:17:02<25:06:35, 11.47s/step, acc=0.4635, loss=2.2827, lr=8.44e-05]
Training:  29%|██▉       | 3200/11082 [10:17:09<25:06:35, 11.47s/step, acc=0.4647, loss=2.2729, lr=8.39e-05]2026-06-19 01:10:37 | INFO | __main__ | epoch=0 step=3200 train_loss=2.272906 train_acc=0.464715 lr=8.391e-05 grad_norm=0.8516 acc_d0=0.7451 acc_d1=0.5931 acc_d2=0.4660 acc_d3=0.3617 acc_d4=0.2842 acc_d5=0.2248 acc_d6=0.1842 acc_d7=0.1538

Training:  29%|██▉       | 3201/11082 [10:17:14<25:29:05, 11.64s/step, acc=0.4647, loss=2.2729, lr=8.39e-05]
Training:  29%|██▉       | 3202/11082 [10:17:25<25:23:54, 11.60s/step, acc=0.4647, loss=2.2729, lr=8.39e-05]
Training:  29%|██▉       | 3203/11082 [10:17:36<25:13:48, 11.53s/step, acc=0.4647, loss=2.2729, lr=8.39e-05]
Training:  29%|██▉       | 3204/11082 [10:17:48<25:30:29, 11.66s/step, acc=0.4647, loss=2.2729, lr=8.39e-05]
Training:  29%|██▉       | 3205/11082 [10:18:00<25:27:14, 11.63s/step, acc=0.4647, loss=2.2729, lr=8.39e-05]
Training:  29%|██▉       | 3206/11082 [10:18:12<25:23:06, 11.60s/step, acc=0.4647, loss=2.2729, lr=8.39e-05]
Training:  29%|██▉       | 3207/11082 [10:18:23<25:24:38, 11.62s/step, acc=0.4647, loss=2.2729, lr=8.39e-05]
Training:  29%|██▉       | 3208/11082 [10:18:34<25:07:35, 11.49s/step, acc=0.4647, loss=2.2729, lr=8.39e-05]
Training:  29%|██▉       | 3209/11082 [10:18:46<25:26:15, 11.63s/step, acc=0.4647, loss=2.2729, lr=8.39e-05]
Training:  29%|██▉       | 3210/11082 [10:18:58<25:22:05, 11.60s/step, acc=0.4647, loss=2.2729, lr=8.39e-05]
Training:  29%|██▉       | 3211/11082 [10:19:09<25:17:23, 11.57s/step, acc=0.4647, loss=2.2729, lr=8.39e-05]
Training:  29%|██▉       | 3212/11082 [10:19:20<24:50:39, 11.36s/step, acc=0.4647, loss=2.2729, lr=8.39e-05]
Training:  29%|██▉       | 3213/11082 [10:19:32<25:15:38, 11.56s/step, acc=0.4647, loss=2.2729, lr=8.39e-05]
Training:  29%|██▉       | 3214/11082 [10:19:44<25:27:14, 11.65s/step, acc=0.4647, loss=2.2729, lr=8.39e-05]
Training:  29%|██▉       | 3215/11082 [10:19:55<25:16:03, 11.56s/step, acc=0.4647, loss=2.2729, lr=8.39e-05]
Training:  29%|██▉       | 3216/11082 [10:20:07<25:22:32, 11.61s/step, acc=0.4647, loss=2.2729, lr=8.39e-05]
Training:  29%|██▉       | 3217/11082 [10:20:18<25:10:14, 11.52s/step, acc=0.4647, loss=2.2729, lr=8.39e-05]
Training:  29%|██▉       | 3218/11082 [10:20:30<25:11:41, 11.53s/step, acc=0.4647, loss=2.2729, lr=8.39e-05]
Training:  29%|██▉       | 3219/11082 [10:20:42<25:19:57, 11.60s/step, acc=0.4647, loss=2.2729, lr=8.39e-05]
Training:  29%|██▉       | 3220/11082 [10:20:53<25:05:06, 11.49s/step, acc=0.4647, loss=2.2729, lr=8.39e-05]
Training:  29%|██▉       | 3221/11082 [10:21:05<25:13:51, 11.55s/step, acc=0.4647, loss=2.2729, lr=8.39e-05]
Training:  29%|██▉       | 3222/11082 [10:21:16<25:13:18, 11.55s/step, acc=0.4647, loss=2.2729, lr=8.39e-05]
Training:  29%|██▉       | 3223/11082 [10:21:28<25:02:12, 11.47s/step, acc=0.4647, loss=2.2729, lr=8.39e-05]
Training:  29%|██▉       | 3224/11082 [10:21:39<25:09:08, 11.52s/step, acc=0.4647, loss=2.2729, lr=8.39e-05]
Training:  29%|██▉       | 3225/11082 [10:21:51<25:12:48, 11.55s/step, acc=0.4647, loss=2.2729, lr=8.39e-05]
Training:  29%|██▉       | 3226/11082 [10:22:03<25:20:17, 11.61s/step, acc=0.4647, loss=2.2729, lr=8.39e-05]
Training:  29%|██▉       | 3227/11082 [10:22:14<25:16:59, 11.59s/step, acc=0.4647, loss=2.2729, lr=8.39e-05]
Training:  29%|██▉       | 3228/11082 [10:22:25<24:57:34, 11.44s/step, acc=0.4647, loss=2.2729, lr=8.39e-05]
Training:  29%|██▉       | 3229/11082 [10:22:37<25:11:39, 11.55s/step, acc=0.4647, loss=2.2729, lr=8.39e-05]
Training:  29%|██▉       | 3230/11082 [10:22:48<25:04:12, 11.49s/step, acc=0.4647, loss=2.2729, lr=8.39e-05]
Training:  29%|██▉       | 3231/11082 [10:23:00<25:05:57, 11.51s/step, acc=0.4647, loss=2.2729, lr=8.39e-05]
Training:  29%|██▉       | 3232/11082 [10:23:11<24:54:12, 11.42s/step, acc=0.4647, loss=2.2729, lr=8.39e-05]
Training:  29%|██▉       | 3233/11082 [10:23:23<25:03:55, 11.50s/step, acc=0.4647, loss=2.2729, lr=8.39e-05]
Training:  29%|██▉       | 3234/11082 [10:23:35<25:17:31, 11.60s/step, acc=0.4647, loss=2.2729, lr=8.39e-05]
Training:  29%|██▉       | 3235/11082 [10:23:46<25:03:00, 11.49s/step, acc=0.4647, loss=2.2729, lr=8.39e-05]
Training:  29%|██▉       | 3236/11082 [10:23:58<25:17:23, 11.60s/step, acc=0.4647, loss=2.2729, lr=8.39e-05]
Training:  29%|██▉       | 3237/11082 [10:24:09<25:08:01, 11.53s/step, acc=0.4647, loss=2.2729, lr=8.39e-05]
Training:  29%|██▉       | 3238/11082 [10:24:20<24:59:41, 11.47s/step, acc=0.4647, loss=2.2729, lr=8.39e-05]
Training:  29%|██▉       | 3239/11082 [10:24:32<25:04:28, 11.51s/step, acc=0.4647, loss=2.2729, lr=8.39e-05]
Training:  29%|██▉       | 3240/11082 [10:24:44<25:12:46, 11.57s/step, acc=0.4647, loss=2.2729, lr=8.39e-05]
Training:  29%|██▉       | 3241/11082 [10:24:55<25:01:10, 11.49s/step, acc=0.4647, loss=2.2729, lr=8.39e-05]
Training:  29%|██▉       | 3242/11082 [10:25:06<24:48:42, 11.39s/step, acc=0.4647, loss=2.2729, lr=8.39e-05]
Training:  29%|██▉       | 3243/11082 [10:25:18<25:11:57, 11.57s/step, acc=0.4647, loss=2.2729, lr=8.39e-05]
Training:  29%|██▉       | 3244/11082 [10:25:30<25:09:55, 11.56s/step, acc=0.4647, loss=2.2729, lr=8.39e-05]
Training:  29%|██▉       | 3245/11082 [10:25:41<25:08:57, 11.55s/step, acc=0.4647, loss=2.2729, lr=8.39e-05]
Training:  29%|██▉       | 3246/11082 [10:25:53<25:12:28, 11.58s/step, acc=0.4647, loss=2.2729, lr=8.39e-05]
Training:  29%|██▉       | 3247/11082 [10:26:05<25:12:30, 11.58s/step, acc=0.4647, loss=2.2729, lr=8.39e-05]
Training:  29%|██▉       | 3248/11082 [10:26:16<24:57:59, 11.47s/step, acc=0.4647, loss=2.2729, lr=8.39e-05]
Training:  29%|██▉       | 3249/11082 [10:26:27<24:57:42, 11.47s/step, acc=0.4647, loss=2.2729, lr=8.39e-05]
Training:  29%|██▉       | 3250/11082 [10:26:39<25:05:01, 11.53s/step, acc=0.4647, loss=2.2729, lr=8.39e-05]
Training:  29%|██▉       | 3250/11082 [10:26:46<25:05:01, 11.53s/step, acc=0.4632, loss=2.2816, lr=8.34e-05]2026-06-19 01:20:14 | INFO | __main__ | epoch=0 step=3250 train_loss=2.281553 train_acc=0.463170 lr=8.341e-05 grad_norm=1.8047 acc_d0=0.7305 acc_d1=0.5744 acc_d2=0.4467 acc_d3=0.3469 acc_d4=0.2706 acc_d5=0.2135 acc_d6=0.1729 acc_d7=0.1420

Training:  29%|██▉       | 3251/11082 [10:26:50<24:58:07, 11.48s/step, acc=0.4632, loss=2.2816, lr=8.34e-05]
Training:  29%|██▉       | 3252/11082 [10:27:02<24:59:01, 11.49s/step, acc=0.4632, loss=2.2816, lr=8.34e-05]
Training:  29%|██▉       | 3253/11082 [10:27:13<25:05:52, 11.54s/step, acc=0.4632, loss=2.2816, lr=8.34e-05]
Training:  29%|██▉       | 3254/11082 [10:27:26<25:28:15, 11.71s/step, acc=0.4632, loss=2.2816, lr=8.34e-05]
Training:  29%|██▉       | 3255/11082 [10:27:37<25:16:22, 11.62s/step, acc=0.4632, loss=2.2816, lr=8.34e-05]
Training:  29%|██▉       | 3256/11082 [10:27:49<25:14:27, 11.61s/step, acc=0.4632, loss=2.2816, lr=8.34e-05]
Training:  29%|██▉       | 3257/11082 [10:28:00<25:11:22, 11.59s/step, acc=0.4632, loss=2.2816, lr=8.34e-05]
Training:  29%|██▉       | 3258/11082 [10:28:11<24:51:51, 11.44s/step, acc=0.4632, loss=2.2816, lr=8.34e-05]
Training:  29%|██▉       | 3259/11082 [10:28:24<25:30:47, 11.74s/step, acc=0.4632, loss=2.2816, lr=8.34e-05]
Training:  29%|██▉       | 3260/11082 [10:28:35<25:22:27, 11.68s/step, acc=0.4632, loss=2.2816, lr=8.34e-05]
Training:  29%|██▉       | 3261/11082 [10:28:46<25:07:12, 11.56s/step, acc=0.4632, loss=2.2816, lr=8.34e-05]
Training:  29%|██▉       | 3262/11082 [10:28:58<24:57:18, 11.49s/step, acc=0.4632, loss=2.2816, lr=8.34e-05]
Training:  29%|██▉       | 3263/11082 [10:29:10<25:12:47, 11.61s/step, acc=0.4632, loss=2.2816, lr=8.34e-05]
Training:  29%|██▉       | 3264/11082 [10:29:21<25:12:39, 11.61s/step, acc=0.4632, loss=2.2816, lr=8.34e-05]
Training:  29%|██▉       | 3265/11082 [10:29:32<24:55:46, 11.48s/step, acc=0.4632, loss=2.2816, lr=8.34e-05]
Training:  29%|██▉       | 3266/11082 [10:29:44<24:57:56, 11.50s/step, acc=0.4632, loss=2.2816, lr=8.34e-05]
Training:  29%|██▉       | 3267/11082 [10:29:56<25:04:44, 11.55s/step, acc=0.4632, loss=2.2816, lr=8.34e-05]
Training:  29%|██▉       | 3268/11082 [10:30:07<25:16:26, 11.64s/step, acc=0.4632, loss=2.2816, lr=8.34e-05]
Training:  29%|██▉       | 3269/11082 [10:30:19<25:20:24, 11.68s/step, acc=0.4632, loss=2.2816, lr=8.34e-05]
Training:  30%|██▉       | 3270/11082 [10:30:30<24:58:03, 11.51s/step, acc=0.4632, loss=2.2816, lr=8.34e-05]
Training:  30%|██▉       | 3271/11082 [10:30:42<24:56:46, 11.50s/step, acc=0.4632, loss=2.2816, lr=8.34e-05]
Training:  30%|██▉       | 3272/11082 [10:30:54<25:10:46, 11.61s/step, acc=0.4632, loss=2.2816, lr=8.34e-05]
Training:  30%|██▉       | 3273/11082 [10:31:06<25:22:02, 11.69s/step, acc=0.4632, loss=2.2816, lr=8.34e-05]
Training:  30%|██▉       | 3274/11082 [10:31:17<25:05:04, 11.57s/step, acc=0.4632, loss=2.2816, lr=8.34e-05]
Training:  30%|██▉       | 3275/11082 [10:31:29<25:09:33, 11.60s/step, acc=0.4632, loss=2.2816, lr=8.34e-05]
Training:  30%|██▉       | 3276/11082 [10:31:40<25:11:39, 11.62s/step, acc=0.4632, loss=2.2816, lr=8.34e-05]
Training:  30%|██▉       | 3277/11082 [10:31:52<25:10:00, 11.61s/step, acc=0.4632, loss=2.2816, lr=8.34e-05]
Training:  30%|██▉       | 3278/11082 [10:32:03<25:13:25, 11.64s/step, acc=0.4632, loss=2.2816, lr=8.34e-05]
Training:  30%|██▉       | 3279/11082 [10:32:15<24:59:41, 11.53s/step, acc=0.4632, loss=2.2816, lr=8.34e-05]
Training:  30%|██▉       | 3280/11082 [10:32:27<25:08:48, 11.60s/step, acc=0.4632, loss=2.2816, lr=8.34e-05]
Training:  30%|██▉       | 3281/11082 [10:32:38<25:10:19, 11.62s/step, acc=0.4632, loss=2.2816, lr=8.34e-05]
Training:  30%|██▉       | 3282/11082 [10:32:49<24:52:28, 11.48s/step, acc=0.4632, loss=2.2816, lr=8.34e-05]
Training:  30%|██▉       | 3283/11082 [10:33:01<24:59:13, 11.53s/step, acc=0.4632, loss=2.2816, lr=8.34e-05]
Training:  30%|██▉       | 3284/11082 [10:33:12<24:52:45, 11.49s/step, acc=0.4632, loss=2.2816, lr=8.34e-05]
Training:  30%|██▉       | 3285/11082 [10:33:24<24:47:51, 11.45s/step, acc=0.4632, loss=2.2816, lr=8.34e-05]
Training:  30%|██▉       | 3286/11082 [10:33:35<24:54:50, 11.50s/step, acc=0.4632, loss=2.2816, lr=8.34e-05]
Training:  30%|██▉       | 3287/11082 [10:33:47<25:07:46, 11.61s/step, acc=0.4632, loss=2.2816, lr=8.34e-05]
Training:  30%|██▉       | 3288/11082 [10:33:59<25:01:44, 11.56s/step, acc=0.4632, loss=2.2816, lr=8.34e-05]
Training:  30%|██▉       | 3289/11082 [10:34:10<25:01:01, 11.56s/step, acc=0.4632, loss=2.2816, lr=8.34e-05]
Training:  30%|██▉       | 3290/11082 [10:34:22<24:54:06, 11.50s/step, acc=0.4632, loss=2.2816, lr=8.34e-05]
Training:  30%|██▉       | 3291/11082 [10:34:34<25:20:46, 11.71s/step, acc=0.4632, loss=2.2816, lr=8.34e-05]
Training:  30%|██▉       | 3292/11082 [10:34:46<25:31:29, 11.80s/step, acc=0.4632, loss=2.2816, lr=8.34e-05]
Training:  30%|██▉       | 3293/11082 [10:34:57<24:58:07, 11.54s/step, acc=0.4632, loss=2.2816, lr=8.34e-05]
Training:  30%|██▉       | 3294/11082 [10:35:09<25:25:19, 11.75s/step, acc=0.4632, loss=2.2816, lr=8.34e-05]
Training:  30%|██▉       | 3295/11082 [10:35:21<25:23:15, 11.74s/step, acc=0.4632, loss=2.2816, lr=8.34e-05]
Training:  30%|██▉       | 3296/11082 [10:35:32<25:11:43, 11.65s/step, acc=0.4632, loss=2.2816, lr=8.34e-05]
Training:  30%|██▉       | 3297/11082 [10:35:43<25:00:27, 11.56s/step, acc=0.4632, loss=2.2816, lr=8.34e-05]
Training:  30%|██▉       | 3298/11082 [10:35:55<25:12:14, 11.66s/step, acc=0.4632, loss=2.2816, lr=8.34e-05]
Training:  30%|██▉       | 3299/11082 [10:36:07<25:08:24, 11.63s/step, acc=0.4632, loss=2.2816, lr=8.34e-05]
Training:  30%|██▉       | 3300/11082 [10:36:19<25:19:19, 11.71s/step, acc=0.4632, loss=2.2816, lr=8.34e-05]
Training:  30%|██▉       | 3300/11082 [10:36:26<25:19:19, 11.71s/step, acc=0.4639, loss=2.2738, lr=8.29e-05]2026-06-19 01:29:53 | INFO | __main__ | epoch=0 step=3300 train_loss=2.273759 train_acc=0.463861 lr=8.290e-05 grad_norm=1.2812 acc_d0=0.7233 acc_d1=0.5665 acc_d2=0.4371 acc_d3=0.3376 acc_d4=0.2646 acc_d5=0.2107 acc_d6=0.1704 acc_d7=0.1417

Training:  30%|██▉       | 3301/11082 [10:36:30<24:59:08, 11.56s/step, acc=0.4639, loss=2.2738, lr=8.29e-05]
Training:  30%|██▉       | 3302/11082 [10:36:41<24:50:52, 11.50s/step, acc=0.4639, loss=2.2738, lr=8.29e-05]
Training:  30%|██▉       | 3303/11082 [10:36:53<25:05:13, 11.61s/step, acc=0.4639, loss=2.2738, lr=8.29e-05]
Training:  30%|██▉       | 3304/11082 [10:37:05<24:52:32, 11.51s/step, acc=0.4639, loss=2.2738, lr=8.29e-05]
Training:  30%|██▉       | 3305/11082 [10:37:16<24:51:14, 11.51s/step, acc=0.4639, loss=2.2738, lr=8.29e-05]
Training:  30%|██▉       | 3306/11082 [10:37:28<25:08:04, 11.64s/step, acc=0.4639, loss=2.2738, lr=8.29e-05]
Training:  30%|██▉       | 3307/11082 [10:37:40<25:17:37, 11.71s/step, acc=0.4639, loss=2.2738, lr=8.29e-05]
Training:  30%|██▉       | 3308/11082 [10:37:51<25:07:16, 11.63s/step, acc=0.4639, loss=2.2738, lr=8.29e-05]
Training:  30%|██▉       | 3309/11082 [10:38:03<25:06:00, 11.62s/step, acc=0.4639, loss=2.2738, lr=8.29e-05]
Training:  30%|██▉       | 3310/11082 [10:38:14<24:55:05, 11.54s/step, acc=0.4639, loss=2.2738, lr=8.29e-05]
Training:  30%|██▉       | 3311/11082 [10:38:26<24:52:01, 11.52s/step, acc=0.4639, loss=2.2738, lr=8.29e-05]
Training:  30%|██▉       | 3312/11082 [10:38:38<25:06:20, 11.63s/step, acc=0.4639, loss=2.2738, lr=8.29e-05]
Training:  30%|██▉       | 3313/11082 [10:38:49<24:56:39, 11.56s/step, acc=0.4639, loss=2.2738, lr=8.29e-05]
Training:  30%|██▉       | 3314/11082 [10:39:01<24:58:33, 11.57s/step, acc=0.4639, loss=2.2738, lr=8.29e-05]
Training:  30%|██▉       | 3315/11082 [10:39:12<24:56:49, 11.56s/step, acc=0.4639, loss=2.2738, lr=8.29e-05]
Training:  30%|██▉       | 3316/11082 [10:39:23<24:46:01, 11.48s/step, acc=0.4639, loss=2.2738, lr=8.29e-05]
Training:  30%|██▉       | 3317/11082 [10:39:35<24:50:58, 11.52s/step, acc=0.4639, loss=2.2738, lr=8.29e-05]
Training:  30%|██▉       | 3318/11082 [10:39:47<24:57:32, 11.57s/step, acc=0.4639, loss=2.2738, lr=8.29e-05]
Training:  30%|██▉       | 3319/11082 [10:39:58<24:42:52, 11.46s/step, acc=0.4639, loss=2.2738, lr=8.29e-05]
Training:  30%|██▉       | 3320/11082 [10:40:10<25:01:59, 11.61s/step, acc=0.4639, loss=2.2738, lr=8.29e-05]
Training:  30%|██▉       | 3321/11082 [10:40:21<24:53:17, 11.54s/step, acc=0.4639, loss=2.2738, lr=8.29e-05]
Training:  30%|██▉       | 3322/11082 [10:40:33<25:04:33, 11.63s/step, acc=0.4639, loss=2.2738, lr=8.29e-05]
Training:  30%|██▉       | 3323/11082 [10:40:44<24:50:45, 11.53s/step, acc=0.4639, loss=2.2738, lr=8.29e-05]
Training:  30%|██▉       | 3324/11082 [10:40:56<24:56:26, 11.57s/step, acc=0.4639, loss=2.2738, lr=8.29e-05]
Training:  30%|███       | 3325/11082 [10:41:08<25:00:24, 11.61s/step, acc=0.4639, loss=2.2738, lr=8.29e-05]
Training:  30%|███       | 3326/11082 [10:41:19<24:45:08, 11.49s/step, acc=0.4639, loss=2.2738, lr=8.29e-05]
Training:  30%|███       | 3327/11082 [10:41:31<24:53:41, 11.56s/step, acc=0.4639, loss=2.2738, lr=8.29e-05]
Training:  30%|███       | 3328/11082 [10:41:42<24:59:02, 11.60s/step, acc=0.4639, loss=2.2738, lr=8.29e-05]
Training:  30%|███       | 3329/11082 [10:41:54<24:58:56, 11.60s/step, acc=0.4639, loss=2.2738, lr=8.29e-05]
Training:  30%|███       | 3330/11082 [10:42:06<25:02:48, 11.63s/step, acc=0.4639, loss=2.2738, lr=8.29e-05]
Training:  30%|███       | 3331/11082 [10:42:17<24:47:01, 11.51s/step, acc=0.4639, loss=2.2738, lr=8.29e-05]
Training:  30%|███       | 3332/11082 [10:42:28<24:42:21, 11.48s/step, acc=0.4639, loss=2.2738, lr=8.29e-05]
Training:  30%|███       | 3333/11082 [10:42:40<24:58:19, 11.60s/step, acc=0.4639, loss=2.2738, lr=8.29e-05]
Training:  30%|███       | 3334/11082 [10:42:52<24:52:02, 11.55s/step, acc=0.4639, loss=2.2738, lr=8.29e-05]
Training:  30%|███       | 3335/11082 [10:43:03<24:28:00, 11.37s/step, acc=0.4639, loss=2.2738, lr=8.29e-05]
Training:  30%|███       | 3336/11082 [10:43:15<24:57:23, 11.60s/step, acc=0.4639, loss=2.2738, lr=8.29e-05]
Training:  30%|███       | 3337/11082 [10:43:26<24:45:48, 11.51s/step, acc=0.4639, loss=2.2738, lr=8.29e-05]
Training:  30%|███       | 3338/11082 [10:43:37<24:41:03, 11.48s/step, acc=0.4639, loss=2.2738, lr=8.29e-05]
Training:  30%|███       | 3339/11082 [10:43:49<24:49:21, 11.54s/step, acc=0.4639, loss=2.2738, lr=8.29e-05]
Training:  30%|███       | 3340/11082 [10:44:01<24:54:17, 11.58s/step, acc=0.4639, loss=2.2738, lr=8.29e-05]
Training:  30%|███       | 3341/11082 [10:44:12<24:51:21, 11.56s/step, acc=0.4639, loss=2.2738, lr=8.29e-05]
Training:  30%|███       | 3342/11082 [10:44:24<24:47:18, 11.53s/step, acc=0.4639, loss=2.2738, lr=8.29e-05]
Training:  30%|███       | 3343/11082 [10:44:36<24:54:56, 11.59s/step, acc=0.4639, loss=2.2738, lr=8.29e-05]
Training:  30%|███       | 3344/11082 [10:44:47<24:58:18, 11.62s/step, acc=0.4639, loss=2.2738, lr=8.29e-05]
Training:  30%|███       | 3345/11082 [10:44:59<24:56:44, 11.61s/step, acc=0.4639, loss=2.2738, lr=8.29e-05]
Training:  30%|███       | 3346/11082 [10:45:11<25:07:23, 11.69s/step, acc=0.4639, loss=2.2738, lr=8.29e-05]
Training:  30%|███       | 3347/11082 [10:45:22<24:55:38, 11.60s/step, acc=0.4639, loss=2.2738, lr=8.29e-05]
Training:  30%|███       | 3348/11082 [10:45:34<24:50:17, 11.56s/step, acc=0.4639, loss=2.2738, lr=8.29e-05]
Training:  30%|███       | 3349/11082 [10:45:45<24:43:03, 11.51s/step, acc=0.4639, loss=2.2738, lr=8.29e-05]
Training:  30%|███       | 3350/11082 [10:45:57<24:48:21, 11.55s/step, acc=0.4639, loss=2.2738, lr=8.29e-05]
Training:  30%|███       | 3350/11082 [10:46:04<24:48:21, 11.55s/step, acc=0.4649, loss=2.2706, lr=8.24e-05]2026-06-19 01:39:32 | INFO | __main__ | epoch=0 step=3350 train_loss=2.270606 train_acc=0.464902 lr=8.239e-05 grad_norm=1.2031 acc_d0=0.7446 acc_d1=0.5935 acc_d2=0.4654 acc_d3=0.3644 acc_d4=0.2850 acc_d5=0.2279 acc_d6=0.1829 acc_d7=0.1522

Training:  30%|███       | 3351/11082 [10:46:08<24:47:19, 11.54s/step, acc=0.4649, loss=2.2706, lr=8.24e-05]
Training:  30%|███       | 3352/11082 [10:46:20<24:50:37, 11.57s/step, acc=0.4649, loss=2.2706, lr=8.24e-05]
Training:  30%|███       | 3353/11082 [10:46:31<24:32:28, 11.43s/step, acc=0.4649, loss=2.2706, lr=8.24e-05]
Training:  30%|███       | 3354/11082 [10:46:43<24:52:32, 11.59s/step, acc=0.4649, loss=2.2706, lr=8.24e-05]
Training:  30%|███       | 3355/11082 [10:46:54<24:49:11, 11.56s/step, acc=0.4649, loss=2.2706, lr=8.24e-05]
Training:  30%|███       | 3356/11082 [10:47:06<24:34:30, 11.45s/step, acc=0.4649, loss=2.2706, lr=8.24e-05]
Training:  30%|███       | 3357/11082 [10:47:17<24:52:03, 11.59s/step, acc=0.4649, loss=2.2706, lr=8.24e-05]
Training:  30%|███       | 3358/11082 [10:47:29<24:55:51, 11.62s/step, acc=0.4649, loss=2.2706, lr=8.24e-05]
Training:  30%|███       | 3359/11082 [10:47:41<24:54:19, 11.61s/step, acc=0.4649, loss=2.2706, lr=8.24e-05]
Training:  30%|███       | 3360/11082 [10:47:52<24:45:45, 11.54s/step, acc=0.4649, loss=2.2706, lr=8.24e-05]
Training:  30%|███       | 3361/11082 [10:48:03<24:31:36, 11.44s/step, acc=0.4649, loss=2.2706, lr=8.24e-05]
Training:  30%|███       | 3362/11082 [10:48:15<24:43:38, 11.53s/step, acc=0.4649, loss=2.2706, lr=8.24e-05]
Training:  30%|███       | 3363/11082 [10:48:27<24:45:30, 11.55s/step, acc=0.4649, loss=2.2706, lr=8.24e-05]
Training:  30%|███       | 3364/11082 [10:48:38<24:52:26, 11.60s/step, acc=0.4649, loss=2.2706, lr=8.24e-05]
Training:  30%|███       | 3365/11082 [10:48:50<24:55:31, 11.63s/step, acc=0.4649, loss=2.2706, lr=8.24e-05]
Training:  30%|███       | 3366/11082 [10:49:01<24:34:20, 11.46s/step, acc=0.4649, loss=2.2706, lr=8.24e-05]
Training:  30%|███       | 3367/11082 [10:49:13<24:39:13, 11.50s/step, acc=0.4649, loss=2.2706, lr=8.24e-05]
Training:  30%|███       | 3368/11082 [10:49:24<24:33:10, 11.46s/step, acc=0.4649, loss=2.2706, lr=8.24e-05]
Training:  30%|███       | 3369/11082 [10:49:36<24:50:25, 11.59s/step, acc=0.4649, loss=2.2706, lr=8.24e-05]
Training:  30%|███       | 3370/11082 [10:49:48<24:54:56, 11.63s/step, acc=0.4649, loss=2.2706, lr=8.24e-05]
Training:  30%|███       | 3371/11082 [10:49:59<24:57:13, 11.65s/step, acc=0.4649, loss=2.2706, lr=8.24e-05]
Training:  30%|███       | 3372/11082 [10:50:11<24:59:02, 11.67s/step, acc=0.4649, loss=2.2706, lr=8.24e-05]
Training:  30%|███       | 3373/11082 [10:50:23<24:53:37, 11.63s/step, acc=0.4649, loss=2.2706, lr=8.24e-05]
Training:  30%|███       | 3374/11082 [10:50:34<24:45:19, 11.56s/step, acc=0.4649, loss=2.2706, lr=8.24e-05]
Training:  30%|███       | 3375/11082 [10:50:45<24:41:22, 11.53s/step, acc=0.4649, loss=2.2706, lr=8.24e-05]
Training:  30%|███       | 3376/11082 [10:50:57<24:47:15, 11.58s/step, acc=0.4649, loss=2.2706, lr=8.24e-05]
Training:  30%|███       | 3377/11082 [10:51:09<24:54:10, 11.64s/step, acc=0.4649, loss=2.2706, lr=8.24e-05]
Training:  30%|███       | 3378/11082 [10:51:20<24:32:31, 11.47s/step, acc=0.4649, loss=2.2706, lr=8.24e-05]
Training:  30%|███       | 3379/11082 [10:51:32<24:48:07, 11.59s/step, acc=0.4649, loss=2.2706, lr=8.24e-05]
Training:  30%|███       | 3380/11082 [10:51:44<24:49:19, 11.60s/step, acc=0.4649, loss=2.2706, lr=8.24e-05]
Training:  31%|███       | 3381/11082 [10:51:55<24:54:24, 11.64s/step, acc=0.4649, loss=2.2706, lr=8.24e-05]
Training:  31%|███       | 3382/11082 [10:52:07<24:46:56, 11.59s/step, acc=0.4649, loss=2.2706, lr=8.24e-05]
Training:  31%|███       | 3383/11082 [10:52:18<24:43:57, 11.56s/step, acc=0.4649, loss=2.2706, lr=8.24e-05]
Training:  31%|███       | 3384/11082 [10:52:30<24:33:31, 11.49s/step, acc=0.4649, loss=2.2706, lr=8.24e-05]
Training:  31%|███       | 3385/11082 [10:52:41<24:38:15, 11.52s/step, acc=0.4649, loss=2.2706, lr=8.24e-05]
Training:  31%|███       | 3386/11082 [10:52:53<24:32:02, 11.48s/step, acc=0.4649, loss=2.2706, lr=8.24e-05]
Training:  31%|███       | 3387/11082 [10:53:05<25:03:16, 11.72s/step, acc=0.4649, loss=2.2706, lr=8.24e-05]
Training:  31%|███       | 3388/11082 [10:53:16<24:54:53, 11.66s/step, acc=0.4649, loss=2.2706, lr=8.24e-05]
Training:  31%|███       | 3389/11082 [10:53:28<25:00:01, 11.70s/step, acc=0.4649, loss=2.2706, lr=8.24e-05]
Training:  31%|███       | 3390/11082 [10:53:40<24:49:20, 11.62s/step, acc=0.4649, loss=2.2706, lr=8.24e-05]
Training:  31%|███       | 3391/11082 [10:53:51<24:50:16, 11.63s/step, acc=0.4649, loss=2.2706, lr=8.24e-05]
Training:  31%|███       | 3392/11082 [10:54:02<24:31:13, 11.48s/step, acc=0.4649, loss=2.2706, lr=8.24e-05]
Training:  31%|███       | 3393/11082 [10:54:14<24:46:59, 11.60s/step, acc=0.4649, loss=2.2706, lr=8.24e-05]
Training:  31%|███       | 3394/11082 [10:54:26<24:43:08, 11.58s/step, acc=0.4649, loss=2.2706, lr=8.24e-05]
Training:  31%|███       | 3395/11082 [10:54:37<24:42:20, 11.57s/step, acc=0.4649, loss=2.2706, lr=8.24e-05]
Training:  31%|███       | 3396/11082 [10:54:49<24:33:10, 11.50s/step, acc=0.4649, loss=2.2706, lr=8.24e-05]
Training:  31%|███       | 3397/11082 [10:55:00<24:34:47, 11.51s/step, acc=0.4649, loss=2.2706, lr=8.24e-05]
Training:  31%|███       | 3398/11082 [10:55:11<24:23:06, 11.42s/step, acc=0.4649, loss=2.2706, lr=8.24e-05]
Training:  31%|███       | 3399/11082 [10:55:23<24:36:09, 11.53s/step, acc=0.4649, loss=2.2706, lr=8.24e-05]
Training:  31%|███       | 3400/11082 [10:55:34<24:28:18, 11.47s/step, acc=0.4649, loss=2.2706, lr=8.24e-05]
Training:  31%|███       | 3400/11082 [10:55:42<24:28:18, 11.47s/step, acc=0.4652, loss=2.2664, lr=8.19e-05]2026-06-19 01:49:10 | INFO | __main__ | epoch=0 step=3400 train_loss=2.266440 train_acc=0.465201 lr=8.188e-05 grad_norm=1.0312 acc_d0=0.7226 acc_d1=0.5646 acc_d2=0.4365 acc_d3=0.3352 acc_d4=0.2595 acc_d5=0.2059 acc_d6=0.1687 acc_d7=0.1382

Training:  31%|███       | 3401/11082 [10:55:47<24:52:46, 11.66s/step, acc=0.4652, loss=2.2664, lr=8.19e-05]
Training:  31%|███       | 3402/11082 [10:55:58<24:38:37, 11.55s/step, acc=0.4652, loss=2.2664, lr=8.19e-05]
Training:  31%|███       | 3403/11082 [10:56:10<24:49:46, 11.64s/step, acc=0.4652, loss=2.2664, lr=8.19e-05]
Training:  31%|███       | 3404/11082 [10:56:21<24:19:24, 11.40s/step, acc=0.4652, loss=2.2664, lr=8.19e-05]
Training:  31%|███       | 3405/11082 [10:56:33<24:44:38, 11.60s/step, acc=0.4652, loss=2.2664, lr=8.19e-05]
Training:  31%|███       | 3406/11082 [10:56:45<24:57:18, 11.70s/step, acc=0.4652, loss=2.2664, lr=8.19e-05]
Training:  31%|███       | 3407/11082 [10:56:56<24:36:56, 11.55s/step, acc=0.4652, loss=2.2664, lr=8.19e-05]
Training:  31%|███       | 3408/11082 [10:57:07<24:40:01, 11.57s/step, acc=0.4652, loss=2.2664, lr=8.19e-05]
Training:  31%|███       | 3409/11082 [10:57:19<24:33:43, 11.52s/step, acc=0.4652, loss=2.2664, lr=8.19e-05]
Training:  31%|███       | 3410/11082 [10:57:30<24:28:07, 11.48s/step, acc=0.4652, loss=2.2664, lr=8.19e-05]
Training:  31%|███       | 3411/11082 [10:57:42<24:37:13, 11.55s/step, acc=0.4652, loss=2.2664, lr=8.19e-05]
Training:  31%|███       | 3412/11082 [10:57:53<24:30:47, 11.51s/step, acc=0.4652, loss=2.2664, lr=8.19e-05]
Training:  31%|███       | 3413/11082 [10:58:05<24:29:43, 11.50s/step, acc=0.4652, loss=2.2664, lr=8.19e-05]
Training:  31%|███       | 3414/11082 [10:58:17<24:39:12, 11.57s/step, acc=0.4652, loss=2.2664, lr=8.19e-05]
Training:  31%|███       | 3415/11082 [10:58:28<24:18:52, 11.42s/step, acc=0.4652, loss=2.2664, lr=8.19e-05]
Training:  31%|███       | 3416/11082 [10:58:39<24:29:08, 11.50s/step, acc=0.4652, loss=2.2664, lr=8.19e-05]
Training:  31%|███       | 3417/11082 [10:58:51<24:34:38, 11.54s/step, acc=0.4652, loss=2.2664, lr=8.19e-05]
Training:  31%|███       | 3418/11082 [10:59:03<24:41:14, 11.60s/step, acc=0.4652, loss=2.2664, lr=8.19e-05]
Training:  31%|███       | 3419/11082 [10:59:14<24:42:08, 11.60s/step, acc=0.4652, loss=2.2664, lr=8.19e-05]
Training:  31%|███       | 3420/11082 [10:59:26<24:32:32, 11.53s/step, acc=0.4652, loss=2.2664, lr=8.19e-05]
Training:  31%|███       | 3421/11082 [10:59:37<24:27:09, 11.49s/step, acc=0.4652, loss=2.2664, lr=8.19e-05]
Training:  31%|███       | 3422/11082 [10:59:49<24:48:17, 11.66s/step, acc=0.4652, loss=2.2664, lr=8.19e-05]
Training:  31%|███       | 3423/11082 [11:00:00<24:36:37, 11.57s/step, acc=0.4652, loss=2.2664, lr=8.19e-05]
Training:  31%|███       | 3424/11082 [11:00:12<24:34:35, 11.55s/step, acc=0.4652, loss=2.2664, lr=8.19e-05]
Training:  31%|███       | 3425/11082 [11:00:23<24:32:18, 11.54s/step, acc=0.4652, loss=2.2664, lr=8.19e-05]
Training:  31%|███       | 3426/11082 [11:00:35<24:38:17, 11.59s/step, acc=0.4652, loss=2.2664, lr=8.19e-05]
Training:  31%|███       | 3427/11082 [11:00:46<24:26:10, 11.49s/step, acc=0.4652, loss=2.2664, lr=8.19e-05]
Training:  31%|███       | 3428/11082 [11:00:58<24:31:46, 11.54s/step, acc=0.4652, loss=2.2664, lr=8.19e-05]
Training:  31%|███       | 3429/11082 [11:01:10<24:42:34, 11.62s/step, acc=0.4652, loss=2.2664, lr=8.19e-05]
Training:  31%|███       | 3430/11082 [11:01:21<24:36:26, 11.58s/step, acc=0.4652, loss=2.2664, lr=8.19e-05]
Training:  31%|███       | 3431/11082 [11:01:33<24:27:56, 11.51s/step, acc=0.4652, loss=2.2664, lr=8.19e-05]
Training:  31%|███       | 3432/11082 [11:01:44<24:20:25, 11.45s/step, acc=0.4652, loss=2.2664, lr=8.19e-05]
Training:  31%|███       | 3433/11082 [11:01:56<24:40:48, 11.62s/step, acc=0.4652, loss=2.2664, lr=8.19e-05]
Training:  31%|███       | 3434/11082 [11:02:07<24:32:32, 11.55s/step, acc=0.4652, loss=2.2664, lr=8.19e-05]
Training:  31%|███       | 3435/11082 [11:02:19<24:20:57, 11.46s/step, acc=0.4652, loss=2.2664, lr=8.19e-05]
Training:  31%|███       | 3436/11082 [11:02:31<24:39:14, 11.61s/step, acc=0.4652, loss=2.2664, lr=8.19e-05]
Training:  31%|███       | 3437/11082 [11:02:42<24:16:50, 11.43s/step, acc=0.4652, loss=2.2664, lr=8.19e-05]
Training:  31%|███       | 3438/11082 [11:02:53<24:24:03, 11.49s/step, acc=0.4652, loss=2.2664, lr=8.19e-05]
Training:  31%|███       | 3439/11082 [11:03:05<24:31:05, 11.55s/step, acc=0.4652, loss=2.2664, lr=8.19e-05]
Training:  31%|███       | 3440/11082 [11:03:17<24:46:45, 11.67s/step, acc=0.4652, loss=2.2664, lr=8.19e-05]
Training:  31%|███       | 3441/11082 [11:03:28<24:33:02, 11.57s/step, acc=0.4652, loss=2.2664, lr=8.19e-05]
Training:  31%|███       | 3442/11082 [11:03:40<24:28:45, 11.53s/step, acc=0.4652, loss=2.2664, lr=8.19e-05]
Training:  31%|███       | 3443/11082 [11:03:51<24:26:09, 11.52s/step, acc=0.4652, loss=2.2664, lr=8.19e-05]
Training:  31%|███       | 3444/11082 [11:04:03<24:38:51, 11.62s/step, acc=0.4652, loss=2.2664, lr=8.19e-05]
Training:  31%|███       | 3445/11082 [11:04:14<24:26:24, 11.52s/step, acc=0.4652, loss=2.2664, lr=8.19e-05]
Training:  31%|███       | 3446/11082 [11:04:26<24:24:30, 11.51s/step, acc=0.4652, loss=2.2664, lr=8.19e-05]
Training:  31%|███       | 3447/11082 [11:04:38<24:38:20, 11.62s/step, acc=0.4652, loss=2.2664, lr=8.19e-05]
Training:  31%|███       | 3448/11082 [11:04:48<24:06:56, 11.37s/step, acc=0.4652, loss=2.2664, lr=8.19e-05]
Training:  31%|███       | 3449/11082 [11:05:01<24:40:50, 11.64s/step, acc=0.4652, loss=2.2664, lr=8.19e-05]
Training:  31%|███       | 3450/11082 [11:05:12<24:37:34, 11.62s/step, acc=0.4652, loss=2.2664, lr=8.19e-05]
Training:  31%|███       | 3450/11082 [11:05:20<24:37:34, 11.62s/step, acc=0.4694, loss=2.2409, lr=8.14e-05]2026-06-19 01:58:47 | INFO | __main__ | epoch=0 step=3450 train_loss=2.240905 train_acc=0.469359 lr=8.135e-05 grad_norm=1.2109 acc_d0=0.7367 acc_d1=0.5838 acc_d2=0.4592 acc_d3=0.3592 acc_d4=0.2829 acc_d5=0.2249 acc_d6=0.1842 acc_d7=0.1530

Training:  31%|███       | 3451/11082 [11:05:24<24:54:32, 11.75s/step, acc=0.4694, loss=2.2409, lr=8.14e-05]
Training:  31%|███       | 3452/11082 [11:05:35<24:30:03, 11.56s/step, acc=0.4694, loss=2.2409, lr=8.14e-05]
Training:  31%|███       | 3453/11082 [11:05:47<24:34:27, 11.60s/step, acc=0.4694, loss=2.2409, lr=8.14e-05]
Training:  31%|███       | 3454/11082 [11:05:59<24:31:08, 11.57s/step, acc=0.4694, loss=2.2409, lr=8.14e-05]
Training:  31%|███       | 3455/11082 [11:06:10<24:23:01, 11.51s/step, acc=0.4694, loss=2.2409, lr=8.14e-05]
Training:  31%|███       | 3456/11082 [11:06:22<24:33:14, 11.59s/step, acc=0.4694, loss=2.2409, lr=8.14e-05]
Training:  31%|███       | 3457/11082 [11:06:34<24:43:45, 11.68s/step, acc=0.4694, loss=2.2409, lr=8.14e-05]
Training:  31%|███       | 3458/11082 [11:06:46<24:49:34, 11.72s/step, acc=0.4694, loss=2.2409, lr=8.14e-05]
Training:  31%|███       | 3459/11082 [11:06:57<24:21:37, 11.50s/step, acc=0.4694, loss=2.2409, lr=8.14e-05]
Training:  31%|███       | 3460/11082 [11:07:08<24:32:54, 11.59s/step, acc=0.4694, loss=2.2409, lr=8.14e-05]
Training:  31%|███       | 3461/11082 [11:07:20<24:40:51, 11.66s/step, acc=0.4694, loss=2.2409, lr=8.14e-05]
Training:  31%|███       | 3462/11082 [11:07:32<24:28:48, 11.57s/step, acc=0.4694, loss=2.2409, lr=8.14e-05]
Training:  31%|███       | 3463/11082 [11:07:43<24:37:53, 11.64s/step, acc=0.4694, loss=2.2409, lr=8.14e-05]
Training:  31%|███▏      | 3464/11082 [11:07:55<24:25:13, 11.54s/step, acc=0.4694, loss=2.2409, lr=8.14e-05]
Training:  31%|███▏      | 3465/11082 [11:08:06<24:25:51, 11.55s/step, acc=0.4694, loss=2.2409, lr=8.14e-05]
Training:  31%|███▏      | 3466/11082 [11:08:18<24:24:52, 11.54s/step, acc=0.4694, loss=2.2409, lr=8.14e-05]
Training:  31%|███▏      | 3467/11082 [11:08:29<24:10:03, 11.43s/step, acc=0.4694, loss=2.2409, lr=8.14e-05]
Training:  31%|███▏      | 3468/11082 [11:08:41<24:49:59, 11.74s/step, acc=0.4694, loss=2.2409, lr=8.14e-05]
Training:  31%|███▏      | 3469/11082 [11:08:53<24:56:31, 11.79s/step, acc=0.4694, loss=2.2409, lr=8.14e-05]
Training:  31%|███▏      | 3470/11082 [11:09:04<24:34:39, 11.62s/step, acc=0.4694, loss=2.2409, lr=8.14e-05]
Training:  31%|███▏      | 3471/11082 [11:09:16<24:46:05, 11.72s/step, acc=0.4694, loss=2.2409, lr=8.14e-05]
Training:  31%|███▏      | 3472/11082 [11:09:28<24:30:39, 11.60s/step, acc=0.4694, loss=2.2409, lr=8.14e-05]
Training:  31%|███▏      | 3473/11082 [11:09:40<24:45:29, 11.71s/step, acc=0.4694, loss=2.2409, lr=8.14e-05]
Training:  31%|███▏      | 3474/11082 [11:09:51<24:34:21, 11.63s/step, acc=0.4694, loss=2.2409, lr=8.14e-05]
Training:  31%|███▏      | 3475/11082 [11:10:03<24:42:10, 11.69s/step, acc=0.4694, loss=2.2409, lr=8.14e-05]
Training:  31%|███▏      | 3476/11082 [11:10:14<24:32:44, 11.62s/step, acc=0.4694, loss=2.2409, lr=8.14e-05]
Training:  31%|███▏      | 3477/11082 [11:10:26<24:28:58, 11.59s/step, acc=0.4694, loss=2.2409, lr=8.14e-05]
Training:  31%|███▏      | 3478/11082 [11:10:37<24:19:33, 11.52s/step, acc=0.4694, loss=2.2409, lr=8.14e-05]
Training:  31%|███▏      | 3479/11082 [11:10:49<24:32:18, 11.62s/step, acc=0.4694, loss=2.2409, lr=8.14e-05]
Training:  31%|███▏      | 3480/11082 [11:11:00<24:17:22, 11.50s/step, acc=0.4694, loss=2.2409, lr=8.14e-05]
Training:  31%|███▏      | 3481/11082 [11:11:12<24:29:57, 11.60s/step, acc=0.4694, loss=2.2409, lr=8.14e-05]
Training:  31%|███▏      | 3482/11082 [11:11:23<24:10:57, 11.45s/step, acc=0.4694, loss=2.2409, lr=8.14e-05]
Training:  31%|███▏      | 3483/11082 [11:11:35<24:16:17, 11.50s/step, acc=0.4694, loss=2.2409, lr=8.14e-05]
Training:  31%|███▏      | 3484/11082 [11:11:47<24:22:07, 11.55s/step, acc=0.4694, loss=2.2409, lr=8.14e-05]
Training:  31%|███▏      | 3485/11082 [11:11:58<24:27:44, 11.59s/step, acc=0.4694, loss=2.2409, lr=8.14e-05]
Training:  31%|███▏      | 3486/11082 [11:12:10<24:36:01, 11.66s/step, acc=0.4694, loss=2.2409, lr=8.14e-05]
Training:  31%|███▏      | 3487/11082 [11:12:22<24:43:13, 11.72s/step, acc=0.4694, loss=2.2409, lr=8.14e-05]
Training:  31%|███▏      | 3488/11082 [11:12:33<24:31:33, 11.63s/step, acc=0.4694, loss=2.2409, lr=8.14e-05]
Training:  31%|███▏      | 3489/11082 [11:12:45<24:20:10, 11.54s/step, acc=0.4694, loss=2.2409, lr=8.14e-05]
Training:  31%|███▏      | 3490/11082 [11:12:56<24:20:59, 11.55s/step, acc=0.4694, loss=2.2409, lr=8.14e-05]
Training:  32%|███▏      | 3491/11082 [11:13:08<24:31:26, 11.63s/step, acc=0.4694, loss=2.2409, lr=8.14e-05]
Training:  32%|███▏      | 3492/11082 [11:13:19<24:22:04, 11.56s/step, acc=0.4694, loss=2.2409, lr=8.14e-05]
Training:  32%|███▏      | 3493/11082 [11:13:31<24:27:09, 11.60s/step, acc=0.4694, loss=2.2409, lr=8.14e-05]
Training:  32%|███▏      | 3494/11082 [11:13:42<24:11:50, 11.48s/step, acc=0.4694, loss=2.2409, lr=8.14e-05]
Training:  32%|███▏      | 3495/11082 [11:13:54<24:14:49, 11.51s/step, acc=0.4694, loss=2.2409, lr=8.14e-05]
Training:  32%|███▏      | 3496/11082 [11:14:06<24:22:53, 11.57s/step, acc=0.4694, loss=2.2409, lr=8.14e-05]
Training:  32%|███▏      | 3497/11082 [11:14:17<24:11:32, 11.48s/step, acc=0.4694, loss=2.2409, lr=8.14e-05]
Training:  32%|███▏      | 3498/11082 [11:14:29<24:18:01, 11.53s/step, acc=0.4694, loss=2.2409, lr=8.14e-05]
Training:  32%|███▏      | 3499/11082 [11:14:40<24:18:58, 11.54s/step, acc=0.4694, loss=2.2409, lr=8.14e-05]
Training:  32%|███▏      | 3500/11082 [11:14:52<24:17:58, 11.54s/step, acc=0.4694, loss=2.2409, lr=8.14e-05]
Training:  32%|███▏      | 3500/11082 [11:14:59<24:17:58, 11.54s/step, acc=0.4691, loss=2.2440, lr=8.08e-05]2026-06-19 02:08:27 | INFO | __main__ | epoch=0 step=3500 train_loss=2.244041 train_acc=0.469136 lr=8.083e-05 grad_norm=1.2891 acc_d0=0.7309 acc_d1=0.5775 acc_d2=0.4544 acc_d3=0.3547 acc_d4=0.2780 acc_d5=0.2227 acc_d6=0.1806 acc_d7=0.1499

Training:  32%|███▏      | 3501/11082 [11:15:04<24:28:32, 11.62s/step, acc=0.4691, loss=2.2440, lr=8.08e-05]
Training:  32%|███▏      | 3502/11082 [11:15:15<24:08:52, 11.47s/step, acc=0.4691, loss=2.2440, lr=8.08e-05]
Training:  32%|███▏      | 3503/11082 [11:15:26<24:22:13, 11.58s/step, acc=0.4691, loss=2.2440, lr=8.08e-05]
Training:  32%|███▏      | 3504/11082 [11:15:37<23:58:01, 11.39s/step, acc=0.4691, loss=2.2440, lr=8.08e-05]
Training:  32%|███▏      | 3505/11082 [11:15:50<24:26:19, 11.61s/step, acc=0.4691, loss=2.2440, lr=8.08e-05]
Training:  32%|███▏      | 3506/11082 [11:16:01<24:17:16, 11.54s/step, acc=0.4691, loss=2.2440, lr=8.08e-05]
Training:  32%|███▏      | 3507/11082 [11:16:13<24:21:10, 11.57s/step, acc=0.4691, loss=2.2440, lr=8.08e-05]
Training:  32%|███▏      | 3508/11082 [11:16:24<24:03:25, 11.43s/step, acc=0.4691, loss=2.2440, lr=8.08e-05]
Training:  32%|███▏      | 3509/11082 [11:16:35<24:08:42, 11.48s/step, acc=0.4691, loss=2.2440, lr=8.08e-05]
Training:  32%|███▏      | 3510/11082 [11:16:47<24:28:33, 11.64s/step, acc=0.4691, loss=2.2440, lr=8.08e-05]
Training:  32%|███▏      | 3511/11082 [11:16:59<24:28:55, 11.64s/step, acc=0.4691, loss=2.2440, lr=8.08e-05]
Training:  32%|███▏      | 3512/11082 [11:17:10<24:20:03, 11.57s/step, acc=0.4691, loss=2.2440, lr=8.08e-05]
Training:  32%|███▏      | 3513/11082 [11:17:22<24:13:22, 11.52s/step, acc=0.4691, loss=2.2440, lr=8.08e-05]
Training:  32%|███▏      | 3514/11082 [11:17:33<24:16:14, 11.55s/step, acc=0.4691, loss=2.2440, lr=8.08e-05]
Training:  32%|███▏      | 3515/11082 [11:17:45<24:19:57, 11.58s/step, acc=0.4691, loss=2.2440, lr=8.08e-05]
Training:  32%|███▏      | 3516/11082 [11:17:56<24:14:40, 11.54s/step, acc=0.4691, loss=2.2440, lr=8.08e-05]
Training:  32%|███▏      | 3517/11082 [11:18:08<24:23:07, 11.60s/step, acc=0.4691, loss=2.2440, lr=8.08e-05]
Training:  32%|███▏      | 3518/11082 [11:18:19<24:11:59, 11.52s/step, acc=0.4691, loss=2.2440, lr=8.08e-05]
Training:  32%|███▏      | 3519/11082 [11:18:31<24:21:58, 11.60s/step, acc=0.4691, loss=2.2440, lr=8.08e-05]
Training:  32%|███▏      | 3520/11082 [11:18:42<24:05:50, 11.47s/step, acc=0.4691, loss=2.2440, lr=8.08e-05]
Training:  32%|███▏      | 3521/11082 [11:18:54<23:59:08, 11.42s/step, acc=0.4691, loss=2.2440, lr=8.08e-05]
Training:  32%|███▏      | 3522/11082 [11:19:06<24:21:25, 11.60s/step, acc=0.4691, loss=2.2440, lr=8.08e-05]
Training:  32%|███▏      | 3523/11082 [11:19:17<24:14:55, 11.55s/step, acc=0.4691, loss=2.2440, lr=8.08e-05]
Training:  32%|███▏      | 3524/11082 [11:19:29<24:18:10, 11.58s/step, acc=0.4691, loss=2.2440, lr=8.08e-05]
Training:  32%|███▏      | 3525/11082 [11:19:40<23:48:34, 11.34s/step, acc=0.4691, loss=2.2440, lr=8.08e-05]
Training:  32%|███▏      | 3526/11082 [11:19:52<24:15:52, 11.56s/step, acc=0.4691, loss=2.2440, lr=8.08e-05]
Training:  32%|███▏      | 3527/11082 [11:20:03<24:23:22, 11.62s/step, acc=0.4691, loss=2.2440, lr=8.08e-05]
Training:  32%|███▏      | 3528/11082 [11:20:15<24:16:19, 11.57s/step, acc=0.4691, loss=2.2440, lr=8.08e-05]
Training:  32%|███▏      | 3529/11082 [11:20:27<24:20:34, 11.60s/step, acc=0.4691, loss=2.2440, lr=8.08e-05]
Training:  32%|███▏      | 3530/11082 [11:20:38<24:14:42, 11.56s/step, acc=0.4691, loss=2.2440, lr=8.08e-05]
Training:  32%|███▏      | 3531/11082 [11:20:49<24:05:42, 11.49s/step, acc=0.4691, loss=2.2440, lr=8.08e-05]
Training:  32%|███▏      | 3532/11082 [11:21:01<23:53:20, 11.39s/step, acc=0.4691, loss=2.2440, lr=8.08e-05]
Training:  32%|███▏      | 3533/11082 [11:21:12<24:10:01, 11.52s/step, acc=0.4691, loss=2.2440, lr=8.08e-05]
Training:  32%|███▏      | 3534/11082 [11:21:24<24:20:09, 11.61s/step, acc=0.4691, loss=2.2440, lr=8.08e-05]
Training:  32%|███▏      | 3535/11082 [11:21:35<24:01:11, 11.46s/step, acc=0.4691, loss=2.2440, lr=8.08e-05]
Training:  32%|███▏      | 3536/11082 [11:21:47<24:25:05, 11.65s/step, acc=0.4691, loss=2.2440, lr=8.08e-05]
Training:  32%|███▏      | 3537/11082 [11:21:59<24:12:50, 11.55s/step, acc=0.4691, loss=2.2440, lr=8.08e-05]
Training:  32%|███▏      | 3538/11082 [11:22:10<24:16:04, 11.58s/step, acc=0.4691, loss=2.2440, lr=8.08e-05]
Training:  32%|███▏      | 3539/11082 [11:22:22<24:16:45, 11.59s/step, acc=0.4691, loss=2.2440, lr=8.08e-05]
Training:  32%|███▏      | 3540/11082 [11:22:34<24:14:37, 11.57s/step, acc=0.4691, loss=2.2440, lr=8.08e-05]
Training:  32%|███▏      | 3541/11082 [11:22:45<23:56:00, 11.43s/step, acc=0.4691, loss=2.2440, lr=8.08e-05]
Training:  32%|███▏      | 3542/11082 [11:22:56<23:56:32, 11.43s/step, acc=0.4691, loss=2.2440, lr=8.08e-05]
Training:  32%|███▏      | 3543/11082 [11:23:08<24:08:33, 11.53s/step, acc=0.4691, loss=2.2440, lr=8.08e-05]
Training:  32%|███▏      | 3544/11082 [11:23:20<24:16:30, 11.59s/step, acc=0.4691, loss=2.2440, lr=8.08e-05]
Training:  32%|███▏      | 3545/11082 [11:23:31<24:14:44, 11.58s/step, acc=0.4691, loss=2.2440, lr=8.08e-05]
Training:  32%|███▏      | 3546/11082 [11:23:43<24:25:48, 11.67s/step, acc=0.4691, loss=2.2440, lr=8.08e-05]
Training:  32%|███▏      | 3547/11082 [11:23:55<24:21:45, 11.64s/step, acc=0.4691, loss=2.2440, lr=8.08e-05]
Training:  32%|███▏      | 3548/11082 [11:24:06<24:11:42, 11.56s/step, acc=0.4691, loss=2.2440, lr=8.08e-05]
Training:  32%|███▏      | 3549/11082 [11:24:18<24:18:10, 11.61s/step, acc=0.4691, loss=2.2440, lr=8.08e-05]
Training:  32%|███▏      | 3550/11082 [11:24:29<24:17:02, 11.61s/step, acc=0.4691, loss=2.2440, lr=8.08e-05]
Training:  32%|███▏      | 3550/11082 [11:24:37<24:17:02, 11.61s/step, acc=0.4708, loss=2.2304, lr=8.03e-05]2026-06-19 02:18:05 | INFO | __main__ | epoch=0 step=3550 train_loss=2.230385 train_acc=0.470831 lr=8.029e-05 grad_norm=1.1172 acc_d0=0.7458 acc_d1=0.5906 acc_d2=0.4624 acc_d3=0.3598 acc_d4=0.2840 acc_d5=0.2261 acc_d6=0.1834 acc_d7=0.1511

Training:  32%|███▏      | 3551/11082 [11:24:42<24:42:32, 11.81s/step, acc=0.4708, loss=2.2304, lr=8.03e-05]
Training:  32%|███▏      | 3552/11082 [11:24:53<24:35:56, 11.76s/step, acc=0.4708, loss=2.2304, lr=8.03e-05]
Training:  32%|███▏      | 3553/11082 [11:25:05<24:32:16, 11.73s/step, acc=0.4708, loss=2.2304, lr=8.03e-05]
Training:  32%|███▏      | 3554/11082 [11:25:16<24:24:04, 11.67s/step, acc=0.4708, loss=2.2304, lr=8.03e-05]
Training:  32%|███▏      | 3555/11082 [11:25:28<24:20:52, 11.65s/step, acc=0.4708, loss=2.2304, lr=8.03e-05]
Training:  32%|███▏      | 3556/11082 [11:25:39<24:09:15, 11.55s/step, acc=0.4708, loss=2.2304, lr=8.03e-05]
Training:  32%|███▏      | 3557/11082 [11:25:52<24:34:23, 11.76s/step, acc=0.4708, loss=2.2304, lr=8.03e-05]
Training:  32%|███▏      | 3558/11082 [11:26:03<24:10:42, 11.57s/step, acc=0.4708, loss=2.2304, lr=8.03e-05]
Training:  32%|███▏      | 3559/11082 [11:26:14<24:05:54, 11.53s/step, acc=0.4708, loss=2.2304, lr=8.03e-05]
Training:  32%|███▏      | 3560/11082 [11:26:25<23:59:31, 11.48s/step, acc=0.4708, loss=2.2304, lr=8.03e-05]
Training:  32%|███▏      | 3561/11082 [11:26:37<24:08:38, 11.56s/step, acc=0.4708, loss=2.2304, lr=8.03e-05]
Training:  32%|███▏      | 3562/11082 [11:26:49<24:06:43, 11.54s/step, acc=0.4708, loss=2.2304, lr=8.03e-05]
Training:  32%|███▏      | 3563/11082 [11:27:00<23:59:57, 11.49s/step, acc=0.4708, loss=2.2304, lr=8.03e-05]
Training:  32%|███▏      | 3564/11082 [11:27:12<24:01:02, 11.50s/step, acc=0.4708, loss=2.2304, lr=8.03e-05]
Training:  32%|███▏      | 3565/11082 [11:27:23<24:04:17, 11.53s/step, acc=0.4708, loss=2.2304, lr=8.03e-05]
Training:  32%|███▏      | 3566/11082 [11:27:35<24:08:36, 11.56s/step, acc=0.4708, loss=2.2304, lr=8.03e-05]
Training:  32%|███▏      | 3567/11082 [11:27:47<24:12:52, 11.60s/step, acc=0.4708, loss=2.2304, lr=8.03e-05]
Training:  32%|███▏      | 3568/11082 [11:27:58<24:14:15, 11.61s/step, acc=0.4708, loss=2.2304, lr=8.03e-05]
Training:  32%|███▏      | 3569/11082 [11:28:10<24:06:55, 11.56s/step, acc=0.4708, loss=2.2304, lr=8.03e-05]
Training:  32%|███▏      | 3570/11082 [11:28:21<24:10:29, 11.59s/step, acc=0.4708, loss=2.2304, lr=8.03e-05]
Training:  32%|███▏      | 3571/11082 [11:28:33<23:59:00, 11.50s/step, acc=0.4708, loss=2.2304, lr=8.03e-05]
Training:  32%|███▏      | 3572/11082 [11:28:44<24:03:39, 11.53s/step, acc=0.4708, loss=2.2304, lr=8.03e-05]
Training:  32%|███▏      | 3573/11082 [11:28:56<24:02:50, 11.53s/step, acc=0.4708, loss=2.2304, lr=8.03e-05]
Training:  32%|███▏      | 3574/11082 [11:29:07<23:46:12, 11.40s/step, acc=0.4708, loss=2.2304, lr=8.03e-05]
Training:  32%|███▏      | 3575/11082 [11:29:18<23:57:56, 11.49s/step, acc=0.4708, loss=2.2304, lr=8.03e-05]
Training:  32%|███▏      | 3576/11082 [11:29:31<24:20:21, 11.67s/step, acc=0.4708, loss=2.2304, lr=8.03e-05]
Training:  32%|███▏      | 3577/11082 [11:29:42<24:08:32, 11.58s/step, acc=0.4708, loss=2.2304, lr=8.03e-05]
Training:  32%|███▏      | 3578/11082 [11:29:53<23:58:46, 11.50s/step, acc=0.4708, loss=2.2304, lr=8.03e-05]
Training:  32%|███▏      | 3579/11082 [11:30:05<24:04:32, 11.55s/step, acc=0.4708, loss=2.2304, lr=8.03e-05]
Training:  32%|███▏      | 3580/11082 [11:30:17<24:07:40, 11.58s/step, acc=0.4708, loss=2.2304, lr=8.03e-05]
Training:  32%|███▏      | 3581/11082 [11:30:28<24:02:43, 11.54s/step, acc=0.4708, loss=2.2304, lr=8.03e-05]
Training:  32%|███▏      | 3582/11082 [11:30:39<24:00:18, 11.52s/step, acc=0.4708, loss=2.2304, lr=8.03e-05]
Training:  32%|███▏      | 3583/11082 [11:30:51<24:10:24, 11.60s/step, acc=0.4708, loss=2.2304, lr=8.03e-05]
Training:  32%|███▏      | 3584/11082 [11:31:03<23:58:27, 11.51s/step, acc=0.4708, loss=2.2304, lr=8.03e-05]
Training:  32%|███▏      | 3585/11082 [11:31:14<24:04:54, 11.56s/step, acc=0.4708, loss=2.2304, lr=8.03e-05]
Training:  32%|███▏      | 3586/11082 [11:31:26<24:11:16, 11.62s/step, acc=0.4708, loss=2.2304, lr=8.03e-05]
Training:  32%|███▏      | 3587/11082 [11:31:37<23:39:09, 11.36s/step, acc=0.4708, loss=2.2304, lr=8.03e-05]
Training:  32%|███▏      | 3588/11082 [11:31:49<24:08:46, 11.60s/step, acc=0.4708, loss=2.2304, lr=8.03e-05]
Training:  32%|███▏      | 3589/11082 [11:32:00<24:00:32, 11.54s/step, acc=0.4708, loss=2.2304, lr=8.03e-05]
Training:  32%|███▏      | 3590/11082 [11:32:12<24:02:06, 11.55s/step, acc=0.4708, loss=2.2304, lr=8.03e-05]
Training:  32%|███▏      | 3591/11082 [11:32:24<24:12:32, 11.63s/step, acc=0.4708, loss=2.2304, lr=8.03e-05]
Training:  32%|███▏      | 3592/11082 [11:32:35<23:57:34, 11.52s/step, acc=0.4708, loss=2.2304, lr=8.03e-05]
Training:  32%|███▏      | 3593/11082 [11:32:47<24:02:55, 11.56s/step, acc=0.4708, loss=2.2304, lr=8.03e-05]
Training:  32%|███▏      | 3594/11082 [11:32:58<23:51:39, 11.47s/step, acc=0.4708, loss=2.2304, lr=8.03e-05]
Training:  32%|███▏      | 3595/11082 [11:33:10<24:03:25, 11.57s/step, acc=0.4708, loss=2.2304, lr=8.03e-05]
Training:  32%|███▏      | 3596/11082 [11:33:21<23:45:07, 11.42s/step, acc=0.4708, loss=2.2304, lr=8.03e-05]
Training:  32%|███▏      | 3597/11082 [11:33:32<23:55:04, 11.50s/step, acc=0.4708, loss=2.2304, lr=8.03e-05]
Training:  32%|███▏      | 3598/11082 [11:33:45<24:16:46, 11.68s/step, acc=0.4708, loss=2.2304, lr=8.03e-05]
Training:  32%|███▏      | 3599/11082 [11:33:56<24:12:51, 11.65s/step, acc=0.4708, loss=2.2304, lr=8.03e-05]
Training:  32%|███▏      | 3600/11082 [11:34:07<23:52:19, 11.49s/step, acc=0.4708, loss=2.2304, lr=8.03e-05]
Training:  32%|███▏      | 3600/11082 [11:34:15<23:52:19, 11.49s/step, acc=0.4724, loss=2.2233, lr=7.98e-05]2026-06-19 02:27:43 | INFO | __main__ | epoch=0 step=3600 train_loss=2.223295 train_acc=0.472355 lr=7.976e-05 grad_norm=1.2109 acc_d0=0.7430 acc_d1=0.5882 acc_d2=0.4605 acc_d3=0.3593 acc_d4=0.2819 acc_d5=0.2240 acc_d6=0.1810 acc_d7=0.1517

Training:  32%|███▏      | 3601/11082 [11:34:19<24:12:36, 11.65s/step, acc=0.4724, loss=2.2233, lr=7.98e-05]
Training:  33%|███▎      | 3602/11082 [11:34:31<24:12:03, 11.65s/step, acc=0.4724, loss=2.2233, lr=7.98e-05]
Training:  33%|███▎      | 3603/11082 [11:34:42<23:53:54, 11.50s/step, acc=0.4724, loss=2.2233, lr=7.98e-05]
Training:  33%|███▎      | 3604/11082 [11:34:53<23:50:18, 11.48s/step, acc=0.4724, loss=2.2233, lr=7.98e-05]
Training:  33%|███▎      | 3605/11082 [11:35:06<24:20:27, 11.72s/step, acc=0.4724, loss=2.2233, lr=7.98e-05]
Training:  33%|███▎      | 3606/11082 [11:35:17<24:15:19, 11.68s/step, acc=0.4724, loss=2.2233, lr=7.98e-05]
Training:  33%|███▎      | 3607/11082 [11:35:29<23:58:44, 11.55s/step, acc=0.4724, loss=2.2233, lr=7.98e-05]
Training:  33%|███▎      | 3608/11082 [11:35:40<24:03:35, 11.59s/step, acc=0.4724, loss=2.2233, lr=7.98e-05]
Training:  33%|███▎      | 3609/11082 [11:35:52<24:05:43, 11.61s/step, acc=0.4724, loss=2.2233, lr=7.98e-05]
Training:  33%|███▎      | 3610/11082 [11:36:04<24:14:20, 11.68s/step, acc=0.4724, loss=2.2233, lr=7.98e-05]
Training:  33%|███▎      | 3611/11082 [11:36:15<24:00:08, 11.57s/step, acc=0.4724, loss=2.2233, lr=7.98e-05]
Training:  33%|███▎      | 3612/11082 [11:36:27<24:18:55, 11.72s/step, acc=0.4724, loss=2.2233, lr=7.98e-05]
Training:  33%|███▎      | 3613/11082 [11:36:39<24:12:30, 11.67s/step, acc=0.4724, loss=2.2233, lr=7.98e-05]
Training:  33%|███▎      | 3614/11082 [11:36:50<24:05:28, 11.61s/step, acc=0.4724, loss=2.2233, lr=7.98e-05]
Training:  33%|███▎      | 3615/11082 [11:37:02<24:15:03, 11.69s/step, acc=0.4724, loss=2.2233, lr=7.98e-05]
Training:  33%|███▎      | 3616/11082 [11:37:14<24:13:10, 11.68s/step, acc=0.4724, loss=2.2233, lr=7.98e-05]
Training:  33%|███▎      | 3617/11082 [11:37:26<24:17:06, 11.71s/step, acc=0.4724, loss=2.2233, lr=7.98e-05]
Training:  33%|███▎      | 3618/11082 [11:37:37<24:13:54, 11.69s/step, acc=0.4724, loss=2.2233, lr=7.98e-05]
Training:  33%|███▎      | 3619/11082 [11:37:48<23:50:20, 11.50s/step, acc=0.4724, loss=2.2233, lr=7.98e-05]
Training:  33%|███▎      | 3620/11082 [11:38:00<23:57:18, 11.56s/step, acc=0.4724, loss=2.2233, lr=7.98e-05]
Training:  33%|███▎      | 3621/11082 [11:38:12<24:15:43, 11.71s/step, acc=0.4724, loss=2.2233, lr=7.98e-05]
Training:  33%|███▎      | 3622/11082 [11:38:24<24:17:29, 11.72s/step, acc=0.4724, loss=2.2233, lr=7.98e-05]
Training:  33%|███▎      | 3623/11082 [11:38:35<23:58:36, 11.57s/step, acc=0.4724, loss=2.2233, lr=7.98e-05]
Training:  33%|███▎      | 3624/11082 [11:38:46<23:52:04, 11.52s/step, acc=0.4724, loss=2.2233, lr=7.98e-05]
Training:  33%|███▎      | 3625/11082 [11:38:58<23:53:21, 11.53s/step, acc=0.4724, loss=2.2233, lr=7.98e-05]
Training:  33%|███▎      | 3626/11082 [11:39:10<23:57:37, 11.57s/step, acc=0.4724, loss=2.2233, lr=7.98e-05]
Training:  33%|███▎      | 3627/11082 [11:39:22<24:19:27, 11.75s/step, acc=0.4724, loss=2.2233, lr=7.98e-05]
Training:  33%|███▎      | 3628/11082 [11:39:33<23:53:43, 11.54s/step, acc=0.4724, loss=2.2233, lr=7.98e-05]
Training:  33%|███▎      | 3629/11082 [11:39:44<23:46:01, 11.48s/step, acc=0.4724, loss=2.2233, lr=7.98e-05]
Training:  33%|███▎      | 3630/11082 [11:39:56<23:53:20, 11.54s/step, acc=0.4724, loss=2.2233, lr=7.98e-05]
Training:  33%|███▎      | 3631/11082 [11:40:08<24:16:47, 11.73s/step, acc=0.4724, loss=2.2233, lr=7.98e-05]
Training:  33%|███▎      | 3632/11082 [11:40:20<24:09:36, 11.67s/step, acc=0.4724, loss=2.2233, lr=7.98e-05]
Training:  33%|███▎      | 3633/11082 [11:40:31<24:05:14, 11.64s/step, acc=0.4724, loss=2.2233, lr=7.98e-05]
Training:  33%|███▎      | 3634/11082 [11:40:43<24:11:45, 11.70s/step, acc=0.4724, loss=2.2233, lr=7.98e-05]
Training:  33%|███▎      | 3635/11082 [11:40:54<24:00:34, 11.61s/step, acc=0.4724, loss=2.2233, lr=7.98e-05]
Training:  33%|███▎      | 3636/11082 [11:41:06<24:03:34, 11.63s/step, acc=0.4724, loss=2.2233, lr=7.98e-05]
Training:  33%|███▎      | 3637/11082 [11:41:17<23:47:06, 11.50s/step, acc=0.4724, loss=2.2233, lr=7.98e-05]
Training:  33%|███▎      | 3638/11082 [11:41:29<23:58:29, 11.59s/step, acc=0.4724, loss=2.2233, lr=7.98e-05]
Training:  33%|███▎      | 3639/11082 [11:41:40<23:41:39, 11.46s/step, acc=0.4724, loss=2.2233, lr=7.98e-05]
Training:  33%|███▎      | 3640/11082 [11:41:52<23:54:33, 11.57s/step, acc=0.4724, loss=2.2233, lr=7.98e-05]
Training:  33%|███▎      | 3641/11082 [11:42:03<23:52:01, 11.55s/step, acc=0.4724, loss=2.2233, lr=7.98e-05]
Training:  33%|███▎      | 3642/11082 [11:42:15<23:44:39, 11.49s/step, acc=0.4724, loss=2.2233, lr=7.98e-05]
Training:  33%|███▎      | 3643/11082 [11:42:27<23:59:58, 11.61s/step, acc=0.4724, loss=2.2233, lr=7.98e-05]
Training:  33%|███▎      | 3644/11082 [11:42:39<24:11:11, 11.71s/step, acc=0.4724, loss=2.2233, lr=7.98e-05]
Training:  33%|███▎      | 3645/11082 [11:42:50<23:50:26, 11.54s/step, acc=0.4724, loss=2.2233, lr=7.98e-05]
Training:  33%|███▎      | 3646/11082 [11:43:01<23:51:53, 11.55s/step, acc=0.4724, loss=2.2233, lr=7.98e-05]
Training:  33%|███▎      | 3647/11082 [11:43:13<23:48:54, 11.53s/step, acc=0.4724, loss=2.2233, lr=7.98e-05]
Training:  33%|███▎      | 3648/11082 [11:43:24<23:43:43, 11.49s/step, acc=0.4724, loss=2.2233, lr=7.98e-05]
Training:  33%|███▎      | 3649/11082 [11:43:36<23:57:04, 11.60s/step, acc=0.4724, loss=2.2233, lr=7.98e-05]
Training:  33%|███▎      | 3650/11082 [11:43:48<23:59:26, 11.62s/step, acc=0.4724, loss=2.2233, lr=7.98e-05]
Training:  33%|███▎      | 3650/11082 [11:43:55<23:59:26, 11.62s/step, acc=0.4724, loss=2.2204, lr=7.92e-05]2026-06-19 02:37:22 | INFO | __main__ | epoch=0 step=3650 train_loss=2.220425 train_acc=0.472450 lr=7.921e-05 grad_norm=0.8906 acc_d0=0.7444 acc_d1=0.5915 acc_d2=0.4643 acc_d3=0.3623 acc_d4=0.2850 acc_d5=0.2265 acc_d6=0.1838 acc_d7=0.1534

Training:  33%|███▎      | 3651/11082 [11:43:59<23:51:59, 11.56s/step, acc=0.4724, loss=2.2204, lr=7.92e-05]
Training:  33%|███▎      | 3652/11082 [11:44:11<23:44:56, 11.51s/step, acc=0.4724, loss=2.2204, lr=7.92e-05]
Training:  33%|███▎      | 3653/11082 [11:44:22<23:46:46, 11.52s/step, acc=0.4724, loss=2.2204, lr=7.92e-05]
Training:  33%|███▎      | 3654/11082 [11:44:34<23:46:29, 11.52s/step, acc=0.4724, loss=2.2204, lr=7.92e-05]
Training:  33%|███▎      | 3655/11082 [11:44:45<23:43:19, 11.50s/step, acc=0.4724, loss=2.2204, lr=7.92e-05]
Training:  33%|███▎      | 3656/11082 [11:44:57<23:56:35, 11.61s/step, acc=0.4724, loss=2.2204, lr=7.92e-05]
Training:  33%|███▎      | 3657/11082 [11:45:08<23:46:56, 11.53s/step, acc=0.4724, loss=2.2204, lr=7.92e-05]
Training:  33%|███▎      | 3658/11082 [11:45:20<23:47:42, 11.54s/step, acc=0.4724, loss=2.2204, lr=7.92e-05]
Training:  33%|███▎      | 3659/11082 [11:45:31<23:35:06, 11.44s/step, acc=0.4724, loss=2.2204, lr=7.92e-05]
Training:  33%|███▎      | 3660/11082 [11:45:43<23:37:06, 11.46s/step, acc=0.4724, loss=2.2204, lr=7.92e-05]
Training:  33%|███▎      | 3661/11082 [11:45:54<23:53:47, 11.59s/step, acc=0.4724, loss=2.2204, lr=7.92e-05]
Training:  33%|███▎      | 3662/11082 [11:46:06<23:40:36, 11.49s/step, acc=0.4724, loss=2.2204, lr=7.92e-05]
Training:  33%|███▎      | 3663/11082 [11:46:17<23:29:37, 11.40s/step, acc=0.4724, loss=2.2204, lr=7.92e-05]
Training:  33%|███▎      | 3664/11082 [11:46:29<23:48:27, 11.55s/step, acc=0.4724, loss=2.2204, lr=7.92e-05]
Training:  33%|███▎      | 3665/11082 [11:46:40<23:44:36, 11.52s/step, acc=0.4724, loss=2.2204, lr=7.92e-05]
Training:  33%|███▎      | 3666/11082 [11:46:51<23:32:28, 11.43s/step, acc=0.4724, loss=2.2204, lr=7.92e-05]
Training:  33%|███▎      | 3667/11082 [11:47:03<23:39:03, 11.48s/step, acc=0.4724, loss=2.2204, lr=7.92e-05]
Training:  33%|███▎      | 3668/11082 [11:47:15<23:59:40, 11.65s/step, acc=0.4724, loss=2.2204, lr=7.92e-05]
Training:  33%|███▎      | 3669/11082 [11:47:26<23:40:18, 11.50s/step, acc=0.4724, loss=2.2204, lr=7.92e-05]
Training:  33%|███▎      | 3670/11082 [11:47:38<23:52:32, 11.60s/step, acc=0.4724, loss=2.2204, lr=7.92e-05]
Training:  33%|███▎      | 3671/11082 [11:47:49<23:31:51, 11.43s/step, acc=0.4724, loss=2.2204, lr=7.92e-05]
Training:  33%|███▎      | 3672/11082 [11:48:01<23:44:33, 11.53s/step, acc=0.4724, loss=2.2204, lr=7.92e-05]
Training:  33%|███▎      | 3673/11082 [11:48:13<23:54:32, 11.62s/step, acc=0.4724, loss=2.2204, lr=7.92e-05]
Training:  33%|███▎      | 3674/11082 [11:48:24<23:42:56, 11.52s/step, acc=0.4724, loss=2.2204, lr=7.92e-05]
Training:  33%|███▎      | 3675/11082 [11:48:36<23:56:33, 11.64s/step, acc=0.4724, loss=2.2204, lr=7.92e-05]
Training:  33%|███▎      | 3676/11082 [11:48:47<23:40:28, 11.51s/step, acc=0.4724, loss=2.2204, lr=7.92e-05]
Training:  33%|███▎      | 3677/11082 [11:48:58<23:34:06, 11.46s/step, acc=0.4724, loss=2.2204, lr=7.92e-05]
Training:  33%|███▎      | 3678/11082 [11:49:11<23:54:30, 11.62s/step, acc=0.4724, loss=2.2204, lr=7.92e-05]
Training:  33%|███▎      | 3679/11082 [11:49:21<23:28:00, 11.41s/step, acc=0.4724, loss=2.2204, lr=7.92e-05]
Training:  33%|███▎      | 3680/11082 [11:49:33<23:43:52, 11.54s/step, acc=0.4724, loss=2.2204, lr=7.92e-05]
Training:  33%|███▎      | 3681/11082 [11:49:45<23:40:31, 11.52s/step, acc=0.4724, loss=2.2204, lr=7.92e-05]
Training:  33%|███▎      | 3682/11082 [11:49:56<23:27:31, 11.41s/step, acc=0.4724, loss=2.2204, lr=7.92e-05]
Training:  33%|███▎      | 3683/11082 [11:50:08<23:53:09, 11.62s/step, acc=0.4724, loss=2.2204, lr=7.92e-05]
Training:  33%|███▎      | 3684/11082 [11:50:20<23:54:41, 11.64s/step, acc=0.4724, loss=2.2204, lr=7.92e-05]
Training:  33%|███▎      | 3685/11082 [11:50:31<23:54:48, 11.64s/step, acc=0.4724, loss=2.2204, lr=7.92e-05]
Training:  33%|███▎      | 3686/11082 [11:50:43<23:48:19, 11.59s/step, acc=0.4724, loss=2.2204, lr=7.92e-05]
Training:  33%|███▎      | 3687/11082 [11:50:54<23:45:52, 11.57s/step, acc=0.4724, loss=2.2204, lr=7.92e-05]
Training:  33%|███▎      | 3688/11082 [11:51:06<23:54:19, 11.64s/step, acc=0.4724, loss=2.2204, lr=7.92e-05]
Training:  33%|███▎      | 3689/11082 [11:51:18<23:45:11, 11.57s/step, acc=0.4724, loss=2.2204, lr=7.92e-05]
Training:  33%|███▎      | 3690/11082 [11:51:29<23:58:16, 11.67s/step, acc=0.4724, loss=2.2204, lr=7.92e-05]
Training:  33%|███▎      | 3691/11082 [11:51:41<23:55:20, 11.65s/step, acc=0.4724, loss=2.2204, lr=7.92e-05]
Training:  33%|███▎      | 3692/11082 [11:51:53<23:55:37, 11.66s/step, acc=0.4724, loss=2.2204, lr=7.92e-05]
Training:  33%|███▎      | 3693/11082 [11:52:05<24:02:03, 11.71s/step, acc=0.4724, loss=2.2204, lr=7.92e-05]
Training:  33%|███▎      | 3694/11082 [11:52:16<23:57:40, 11.68s/step, acc=0.4724, loss=2.2204, lr=7.92e-05]
Training:  33%|███▎      | 3695/11082 [11:52:28<23:50:35, 11.62s/step, acc=0.4724, loss=2.2204, lr=7.92e-05]
Training:  33%|███▎      | 3696/11082 [11:52:39<23:56:40, 11.67s/step, acc=0.4724, loss=2.2204, lr=7.92e-05]
Training:  33%|███▎      | 3697/11082 [11:52:51<23:40:51, 11.54s/step, acc=0.4724, loss=2.2204, lr=7.92e-05]
Training:  33%|███▎      | 3698/11082 [11:53:02<23:37:41, 11.52s/step, acc=0.4724, loss=2.2204, lr=7.92e-05]
Training:  33%|███▎      | 3699/11082 [11:53:14<23:45:24, 11.58s/step, acc=0.4724, loss=2.2204, lr=7.92e-05]
Training:  33%|███▎      | 3700/11082 [11:53:26<23:51:48, 11.64s/step, acc=0.4724, loss=2.2204, lr=7.92e-05]
Training:  33%|███▎      | 3700/11082 [11:53:33<23:51:48, 11.64s/step, acc=0.4738, loss=2.2114, lr=7.87e-05]2026-06-19 02:47:01 | INFO | __main__ | epoch=0 step=3700 train_loss=2.211444 train_acc=0.473756 lr=7.866e-05 grad_norm=1.1797 acc_d0=0.7492 acc_d1=0.6005 acc_d2=0.4756 acc_d3=0.3737 acc_d4=0.2972 acc_d5=0.2372 acc_d6=0.1919 acc_d7=0.1567

Training:  33%|███▎      | 3701/11082 [11:53:38<24:07:40, 11.77s/step, acc=0.4738, loss=2.2114, lr=7.87e-05]
Training:  33%|███▎      | 3702/11082 [11:53:49<23:47:56, 11.61s/step, acc=0.4738, loss=2.2114, lr=7.87e-05]
Training:  33%|███▎      | 3703/11082 [11:54:00<23:38:41, 11.54s/step, acc=0.4738, loss=2.2114, lr=7.87e-05]
Training:  33%|███▎      | 3704/11082 [11:54:12<23:47:50, 11.61s/step, acc=0.4738, loss=2.2114, lr=7.87e-05]
Training:  33%|███▎      | 3705/11082 [11:54:24<23:43:07, 11.57s/step, acc=0.4738, loss=2.2114, lr=7.87e-05]
Training:  33%|███▎      | 3706/11082 [11:54:35<23:26:26, 11.44s/step, acc=0.4738, loss=2.2114, lr=7.87e-05]
Training:  33%|███▎      | 3707/11082 [11:54:46<23:19:41, 11.39s/step, acc=0.4738, loss=2.2114, lr=7.87e-05]
Training:  33%|███▎      | 3708/11082 [11:54:59<24:17:31, 11.86s/step, acc=0.4738, loss=2.2114, lr=7.87e-05]
Training:  33%|███▎      | 3709/11082 [11:55:10<23:51:19, 11.65s/step, acc=0.4738, loss=2.2114, lr=7.87e-05]
Training:  33%|███▎      | 3710/11082 [11:55:22<23:50:35, 11.64s/step, acc=0.4738, loss=2.2114, lr=7.87e-05]
Training:  33%|███▎      | 3711/11082 [11:55:33<23:35:34, 11.52s/step, acc=0.4738, loss=2.2114, lr=7.87e-05]
Training:  33%|███▎      | 3712/11082 [11:55:44<23:34:22, 11.51s/step, acc=0.4738, loss=2.2114, lr=7.87e-05]
Training:  34%|███▎      | 3713/11082 [11:55:57<23:55:32, 11.69s/step, acc=0.4738, loss=2.2114, lr=7.87e-05]
Training:  34%|███▎      | 3714/11082 [11:56:08<23:50:34, 11.65s/step, acc=0.4738, loss=2.2114, lr=7.87e-05]
Training:  34%|███▎      | 3715/11082 [11:56:20<24:12:58, 11.83s/step, acc=0.4738, loss=2.2114, lr=7.87e-05]
Training:  34%|███▎      | 3716/11082 [11:56:32<24:01:55, 11.75s/step, acc=0.4738, loss=2.2114, lr=7.87e-05]
Training:  34%|███▎      | 3717/11082 [11:56:44<23:56:44, 11.70s/step, acc=0.4738, loss=2.2114, lr=7.87e-05]
Training:  34%|███▎      | 3718/11082 [11:56:55<23:41:51, 11.58s/step, acc=0.4738, loss=2.2114, lr=7.87e-05]
Training:  34%|███▎      | 3719/11082 [11:57:06<23:35:04, 11.53s/step, acc=0.4738, loss=2.2114, lr=7.87e-05]
Training:  34%|███▎      | 3720/11082 [11:57:18<23:43:57, 11.61s/step, acc=0.4738, loss=2.2114, lr=7.87e-05]
Training:  34%|███▎      | 3721/11082 [11:57:30<23:42:12, 11.59s/step, acc=0.4738, loss=2.2114, lr=7.87e-05]
Training:  34%|███▎      | 3722/11082 [11:57:41<23:30:30, 11.50s/step, acc=0.4738, loss=2.2114, lr=7.87e-05]
Training:  34%|███▎      | 3723/11082 [11:57:52<23:32:57, 11.52s/step, acc=0.4738, loss=2.2114, lr=7.87e-05]
Training:  34%|███▎      | 3724/11082 [11:58:04<23:42:52, 11.60s/step, acc=0.4738, loss=2.2114, lr=7.87e-05]
Training:  34%|███▎      | 3725/11082 [11:58:16<23:38:42, 11.57s/step, acc=0.4738, loss=2.2114, lr=7.87e-05]
Training:  34%|███▎      | 3726/11082 [11:58:27<23:29:49, 11.50s/step, acc=0.4738, loss=2.2114, lr=7.87e-05]
Training:  34%|███▎      | 3727/11082 [11:58:38<23:13:53, 11.37s/step, acc=0.4738, loss=2.2114, lr=7.87e-05]
Training:  34%|███▎      | 3728/11082 [11:58:50<23:15:04, 11.38s/step, acc=0.4738, loss=2.2114, lr=7.87e-05]
Training:  34%|███▎      | 3729/11082 [11:59:02<23:42:39, 11.61s/step, acc=0.4738, loss=2.2114, lr=7.87e-05]
Training:  34%|███▎      | 3730/11082 [11:59:13<23:26:41, 11.48s/step, acc=0.4738, loss=2.2114, lr=7.87e-05]
Training:  34%|███▎      | 3731/11082 [11:59:24<23:14:46, 11.38s/step, acc=0.4738, loss=2.2114, lr=7.87e-05]
Training:  34%|███▎      | 3732/11082 [11:59:36<23:21:53, 11.44s/step, acc=0.4738, loss=2.2114, lr=7.87e-05]
Training:  34%|███▎      | 3733/11082 [11:59:47<23:32:30, 11.53s/step, acc=0.4738, loss=2.2114, lr=7.87e-05]
Training:  34%|███▎      | 3734/11082 [11:59:59<23:38:57, 11.59s/step, acc=0.4738, loss=2.2114, lr=7.87e-05]
Training:  34%|███▎      | 3735/11082 [12:00:10<23:22:27, 11.45s/step, acc=0.4738, loss=2.2114, lr=7.87e-05]
Training:  34%|███▎      | 3736/11082 [12:00:22<23:38:50, 11.59s/step, acc=0.4738, loss=2.2114, lr=7.87e-05]
Training:  34%|███▎      | 3737/11082 [12:00:33<23:22:12, 11.45s/step, acc=0.4738, loss=2.2114, lr=7.87e-05]
Training:  34%|███▎      | 3738/11082 [12:00:45<23:41:15, 11.61s/step, acc=0.4738, loss=2.2114, lr=7.87e-05]
Training:  34%|███▎      | 3739/11082 [12:00:57<23:30:16, 11.52s/step, acc=0.4738, loss=2.2114, lr=7.87e-05]
Training:  34%|███▎      | 3740/11082 [12:01:08<23:19:49, 11.44s/step, acc=0.4738, loss=2.2114, lr=7.87e-05]
Training:  34%|███▍      | 3741/11082 [12:01:19<23:29:12, 11.52s/step, acc=0.4738, loss=2.2114, lr=7.87e-05]
Training:  34%|███▍      | 3742/11082 [12:01:31<23:25:42, 11.49s/step, acc=0.4738, loss=2.2114, lr=7.87e-05]
Training:  34%|███▍      | 3743/11082 [12:01:43<23:36:00, 11.58s/step, acc=0.4738, loss=2.2114, lr=7.87e-05]
Training:  34%|███▍      | 3744/11082 [12:01:54<23:37:58, 11.59s/step, acc=0.4738, loss=2.2114, lr=7.87e-05]
Training:  34%|███▍      | 3745/11082 [12:02:06<23:24:35, 11.49s/step, acc=0.4738, loss=2.2114, lr=7.87e-05]
Training:  34%|███▍      | 3746/11082 [12:02:17<23:27:25, 11.51s/step, acc=0.4738, loss=2.2114, lr=7.87e-05]
Training:  34%|███▍      | 3747/11082 [12:02:29<23:23:43, 11.48s/step, acc=0.4738, loss=2.2114, lr=7.87e-05]
Training:  34%|███▍      | 3748/11082 [12:02:40<23:36:34, 11.59s/step, acc=0.4738, loss=2.2114, lr=7.87e-05]
Training:  34%|███▍      | 3749/11082 [12:02:52<23:19:53, 11.45s/step, acc=0.4738, loss=2.2114, lr=7.87e-05]
Training:  34%|███▍      | 3750/11082 [12:03:03<23:22:30, 11.48s/step, acc=0.4738, loss=2.2114, lr=7.87e-05]
Training:  34%|███▍      | 3750/11082 [12:03:10<23:22:30, 11.48s/step, acc=0.4728, loss=2.2237, lr=7.81e-05]2026-06-19 02:56:38 | INFO | __main__ | epoch=0 step=3750 train_loss=2.223712 train_acc=0.472823 lr=7.811e-05 grad_norm=1.2422 acc_d0=0.7446 acc_d1=0.5931 acc_d2=0.4670 acc_d3=0.3662 acc_d4=0.2880 acc_d5=0.2288 acc_d6=0.1860 acc_d7=0.1544

Training:  34%|███▍      | 3751/11082 [12:03:15<23:25:23, 11.50s/step, acc=0.4728, loss=2.2237, lr=7.81e-05]
Training:  34%|███▍      | 3752/11082 [12:03:26<23:30:26, 11.55s/step, acc=0.4728, loss=2.2237, lr=7.81e-05]
Training:  34%|███▍      | 3753/11082 [12:03:38<23:21:28, 11.47s/step, acc=0.4728, loss=2.2237, lr=7.81e-05]
Training:  34%|███▍      | 3754/11082 [12:03:50<23:42:34, 11.65s/step, acc=0.4728, loss=2.2237, lr=7.81e-05]
Training:  34%|███▍      | 3755/11082 [12:04:01<23:23:29, 11.49s/step, acc=0.4728, loss=2.2237, lr=7.81e-05]
Training:  34%|███▍      | 3756/11082 [12:04:13<23:34:29, 11.58s/step, acc=0.4728, loss=2.2237, lr=7.81e-05]
Training:  34%|███▍      | 3757/11082 [12:04:24<23:36:29, 11.60s/step, acc=0.4728, loss=2.2237, lr=7.81e-05]
Training:  34%|███▍      | 3758/11082 [12:04:35<23:24:05, 11.50s/step, acc=0.4728, loss=2.2237, lr=7.81e-05]
Training:  34%|███▍      | 3759/11082 [12:04:47<23:19:49, 11.47s/step, acc=0.4728, loss=2.2237, lr=7.81e-05]
Training:  34%|███▍      | 3760/11082 [12:04:59<23:30:43, 11.56s/step, acc=0.4728, loss=2.2237, lr=7.81e-05]
Training:  34%|███▍      | 3761/11082 [12:05:10<23:38:10, 11.62s/step, acc=0.4728, loss=2.2237, lr=7.81e-05]
Training:  34%|███▍      | 3762/11082 [12:05:22<23:35:37, 11.60s/step, acc=0.4728, loss=2.2237, lr=7.81e-05]
Training:  34%|███▍      | 3763/11082 [12:05:33<23:32:39, 11.58s/step, acc=0.4728, loss=2.2237, lr=7.81e-05]
Training:  34%|███▍      | 3764/11082 [12:05:45<23:18:31, 11.47s/step, acc=0.4728, loss=2.2237, lr=7.81e-05]
Training:  34%|███▍      | 3765/11082 [12:05:56<23:24:27, 11.52s/step, acc=0.4728, loss=2.2237, lr=7.81e-05]
Training:  34%|███▍      | 3766/11082 [12:06:08<23:30:02, 11.56s/step, acc=0.4728, loss=2.2237, lr=7.81e-05]
Training:  34%|███▍      | 3767/11082 [12:06:20<23:32:03, 11.58s/step, acc=0.4728, loss=2.2237, lr=7.81e-05]
Training:  34%|███▍      | 3768/11082 [12:06:31<23:16:24, 11.46s/step, acc=0.4728, loss=2.2237, lr=7.81e-05]
Training:  34%|███▍      | 3769/11082 [12:06:43<23:35:53, 11.62s/step, acc=0.4728, loss=2.2237, lr=7.81e-05]
Training:  34%|███▍      | 3770/11082 [12:06:55<23:43:08, 11.68s/step, acc=0.4728, loss=2.2237, lr=7.81e-05]
Training:  34%|███▍      | 3771/11082 [12:07:06<23:46:52, 11.71s/step, acc=0.4728, loss=2.2237, lr=7.81e-05]
Training:  34%|███▍      | 3772/11082 [12:07:18<23:26:45, 11.55s/step, acc=0.4728, loss=2.2237, lr=7.81e-05]
Training:  34%|███▍      | 3773/11082 [12:07:29<23:34:48, 11.61s/step, acc=0.4728, loss=2.2237, lr=7.81e-05]
Training:  34%|███▍      | 3774/11082 [12:07:41<23:35:16, 11.62s/step, acc=0.4728, loss=2.2237, lr=7.81e-05]
Training:  34%|███▍      | 3775/11082 [12:07:52<23:26:12, 11.55s/step, acc=0.4728, loss=2.2237, lr=7.81e-05]
Training:  34%|███▍      | 3776/11082 [12:08:05<23:55:57, 11.79s/step, acc=0.4728, loss=2.2237, lr=7.81e-05]
Training:  34%|███▍      | 3777/11082 [12:08:16<23:31:57, 11.60s/step, acc=0.4728, loss=2.2237, lr=7.81e-05]
Training:  34%|███▍      | 3778/11082 [12:08:28<23:41:52, 11.68s/step, acc=0.4728, loss=2.2237, lr=7.81e-05]
Training:  34%|███▍      | 3779/11082 [12:08:39<23:43:46, 11.70s/step, acc=0.4728, loss=2.2237, lr=7.81e-05]
Training:  34%|███▍      | 3780/11082 [12:08:50<23:18:25, 11.49s/step, acc=0.4728, loss=2.2237, lr=7.81e-05]
Training:  34%|███▍      | 3781/11082 [12:09:03<23:43:47, 11.70s/step, acc=0.4728, loss=2.2237, lr=7.81e-05]
Training:  34%|███▍      | 3782/11082 [12:09:14<23:26:58, 11.56s/step, acc=0.4728, loss=2.2237, lr=7.81e-05]
Training:  34%|███▍      | 3783/11082 [12:09:26<23:39:11, 11.67s/step, acc=0.4728, loss=2.2237, lr=7.81e-05]
Training:  34%|███▍      | 3784/11082 [12:09:37<23:19:25, 11.51s/step, acc=0.4728, loss=2.2237, lr=7.81e-05]
Training:  34%|███▍      | 3785/11082 [12:09:48<23:15:49, 11.48s/step, acc=0.4728, loss=2.2237, lr=7.81e-05]
Training:  34%|███▍      | 3786/11082 [12:10:00<23:10:57, 11.44s/step, acc=0.4728, loss=2.2237, lr=7.81e-05]
Training:  34%|███▍      | 3787/11082 [12:10:12<23:44:04, 11.71s/step, acc=0.4728, loss=2.2237, lr=7.81e-05]
Training:  34%|███▍      | 3788/11082 [12:10:23<23:35:23, 11.64s/step, acc=0.4728, loss=2.2237, lr=7.81e-05]
Training:  34%|███▍      | 3789/11082 [12:10:35<23:19:29, 11.51s/step, acc=0.4728, loss=2.2237, lr=7.81e-05]
Training:  34%|███▍      | 3790/11082 [12:10:47<23:33:04, 11.63s/step, acc=0.4728, loss=2.2237, lr=7.81e-05]
Training:  34%|███▍      | 3791/11082 [12:10:58<23:36:53, 11.66s/step, acc=0.4728, loss=2.2237, lr=7.81e-05]
Training:  34%|███▍      | 3792/11082 [12:11:09<23:02:58, 11.38s/step, acc=0.4728, loss=2.2237, lr=7.81e-05]
Training:  34%|███▍      | 3793/11082 [12:11:21<23:35:01, 11.65s/step, acc=0.4728, loss=2.2237, lr=7.81e-05]
Training:  34%|███▍      | 3794/11082 [12:11:33<23:26:48, 11.58s/step, acc=0.4728, loss=2.2237, lr=7.81e-05]
Training:  34%|███▍      | 3795/11082 [12:11:45<23:32:25, 11.63s/step, acc=0.4728, loss=2.2237, lr=7.81e-05]
Training:  34%|███▍      | 3796/11082 [12:11:56<23:33:18, 11.64s/step, acc=0.4728, loss=2.2237, lr=7.81e-05]
Training:  34%|███▍      | 3797/11082 [12:12:08<23:27:00, 11.59s/step, acc=0.4728, loss=2.2237, lr=7.81e-05]
Training:  34%|███▍      | 3798/11082 [12:12:19<23:21:38, 11.55s/step, acc=0.4728, loss=2.2237, lr=7.81e-05]
Training:  34%|███▍      | 3799/11082 [12:12:30<23:12:07, 11.47s/step, acc=0.4728, loss=2.2237, lr=7.81e-05]
Training:  34%|███▍      | 3800/11082 [12:12:42<23:28:19, 11.60s/step, acc=0.4728, loss=2.2237, lr=7.81e-05]
Training:  34%|███▍      | 3800/11082 [12:12:49<23:28:19, 11.60s/step, acc=0.4729, loss=2.2192, lr=7.76e-05]2026-06-19 03:06:17 | INFO | __main__ | epoch=0 step=3800 train_loss=2.219207 train_acc=0.472884 lr=7.755e-05 grad_norm=1.3047 acc_d0=0.7325 acc_d1=0.5783 acc_d2=0.4523 acc_d3=0.3527 acc_d4=0.2765 acc_d5=0.2207 acc_d6=0.1792 acc_d7=0.1485

Training:  34%|███▍      | 3801/11082 [12:12:54<23:21:33, 11.55s/step, acc=0.4729, loss=2.2192, lr=7.76e-05]
Training:  34%|███▍      | 3802/11082 [12:13:05<23:25:00, 11.58s/step, acc=0.4729, loss=2.2192, lr=7.76e-05]
Training:  34%|███▍      | 3803/11082 [12:13:17<23:22:06, 11.56s/step, acc=0.4729, loss=2.2192, lr=7.76e-05]
Training:  34%|███▍      | 3804/11082 [12:13:28<22:58:37, 11.37s/step, acc=0.4729, loss=2.2192, lr=7.76e-05]
Training:  34%|███▍      | 3805/11082 [12:13:39<23:02:38, 11.40s/step, acc=0.4729, loss=2.2192, lr=7.76e-05]
Training:  34%|███▍      | 3806/11082 [12:13:51<23:24:24, 11.58s/step, acc=0.4729, loss=2.2192, lr=7.76e-05]
Training:  34%|███▍      | 3807/11082 [12:14:03<23:24:18, 11.58s/step, acc=0.4729, loss=2.2192, lr=7.76e-05]
Training:  34%|███▍      | 3808/11082 [12:14:14<23:22:10, 11.57s/step, acc=0.4729, loss=2.2192, lr=7.76e-05]
Training:  34%|███▍      | 3809/11082 [12:14:26<23:20:51, 11.56s/step, acc=0.4729, loss=2.2192, lr=7.76e-05]
Training:  34%|███▍      | 3810/11082 [12:14:37<23:20:58, 11.56s/step, acc=0.4729, loss=2.2192, lr=7.76e-05]
Training:  34%|███▍      | 3811/11082 [12:14:49<23:28:01, 11.62s/step, acc=0.4729, loss=2.2192, lr=7.76e-05]
Training:  34%|███▍      | 3812/11082 [12:15:01<23:16:29, 11.53s/step, acc=0.4729, loss=2.2192, lr=7.76e-05]
Training:  34%|███▍      | 3813/11082 [12:15:12<23:09:06, 11.47s/step, acc=0.4729, loss=2.2192, lr=7.76e-05]
Training:  34%|███▍      | 3814/11082 [12:15:24<23:29:18, 11.63s/step, acc=0.4729, loss=2.2192, lr=7.76e-05]
Training:  34%|███▍      | 3815/11082 [12:15:35<23:13:36, 11.51s/step, acc=0.4729, loss=2.2192, lr=7.76e-05]
Training:  34%|███▍      | 3816/11082 [12:15:47<23:16:22, 11.53s/step, acc=0.4729, loss=2.2192, lr=7.76e-05]
Training:  34%|███▍      | 3817/11082 [12:15:58<23:18:05, 11.55s/step, acc=0.4729, loss=2.2192, lr=7.76e-05]
Training:  34%|███▍      | 3818/11082 [12:16:10<23:11:19, 11.49s/step, acc=0.4729, loss=2.2192, lr=7.76e-05]
Training:  34%|███▍      | 3819/11082 [12:16:22<23:31:16, 11.66s/step, acc=0.4729, loss=2.2192, lr=7.76e-05]
Training:  34%|███▍      | 3820/11082 [12:16:33<23:07:45, 11.47s/step, acc=0.4729, loss=2.2192, lr=7.76e-05]
Training:  34%|███▍      | 3821/11082 [12:16:44<23:15:06, 11.53s/step, acc=0.4729, loss=2.2192, lr=7.76e-05]
Training:  34%|███▍      | 3822/11082 [12:16:56<23:10:50, 11.49s/step, acc=0.4729, loss=2.2192, lr=7.76e-05]
Training:  34%|███▍      | 3823/11082 [12:17:08<23:23:08, 11.60s/step, acc=0.4729, loss=2.2192, lr=7.76e-05]
Training:  35%|███▍      | 3824/11082 [12:17:19<23:23:39, 11.60s/step, acc=0.4729, loss=2.2192, lr=7.76e-05]
Training:  35%|███▍      | 3825/11082 [12:17:31<23:24:54, 11.62s/step, acc=0.4729, loss=2.2192, lr=7.76e-05]
Training:  35%|███▍      | 3826/11082 [12:17:42<23:03:44, 11.44s/step, acc=0.4729, loss=2.2192, lr=7.76e-05]
Training:  35%|███▍      | 3827/11082 [12:17:54<23:25:40, 11.63s/step, acc=0.4729, loss=2.2192, lr=7.76e-05]
Training:  35%|███▍      | 3828/11082 [12:18:06<23:23:33, 11.61s/step, acc=0.4729, loss=2.2192, lr=7.76e-05]
Training:  35%|███▍      | 3829/11082 [12:18:17<23:19:54, 11.58s/step, acc=0.4729, loss=2.2192, lr=7.76e-05]
Training:  35%|███▍      | 3830/11082 [12:18:29<23:17:59, 11.57s/step, acc=0.4729, loss=2.2192, lr=7.76e-05]
Training:  35%|███▍      | 3831/11082 [12:18:40<23:16:02, 11.55s/step, acc=0.4729, loss=2.2192, lr=7.76e-05]
Training:  35%|███▍      | 3832/11082 [12:18:52<23:17:21, 11.56s/step, acc=0.4729, loss=2.2192, lr=7.76e-05]
Training:  35%|███▍      | 3833/11082 [12:19:03<23:21:32, 11.60s/step, acc=0.4729, loss=2.2192, lr=7.76e-05]
Training:  35%|███▍      | 3834/11082 [12:19:15<23:20:59, 11.60s/step, acc=0.4729, loss=2.2192, lr=7.76e-05]
Training:  35%|███▍      | 3835/11082 [12:19:27<23:17:34, 11.57s/step, acc=0.4729, loss=2.2192, lr=7.76e-05]
Training:  35%|███▍      | 3836/11082 [12:19:38<23:01:08, 11.44s/step, acc=0.4729, loss=2.2192, lr=7.76e-05]
Training:  35%|███▍      | 3837/11082 [12:19:50<23:25:32, 11.64s/step, acc=0.4729, loss=2.2192, lr=7.76e-05]
Training:  35%|███▍      | 3838/11082 [12:20:01<23:17:32, 11.58s/step, acc=0.4729, loss=2.2192, lr=7.76e-05]
Training:  35%|███▍      | 3839/11082 [12:20:12<23:07:21, 11.49s/step, acc=0.4729, loss=2.2192, lr=7.76e-05]
Training:  35%|███▍      | 3840/11082 [12:20:24<23:25:36, 11.65s/step, acc=0.4729, loss=2.2192, lr=7.76e-05]
Training:  35%|███▍      | 3841/11082 [12:20:36<23:15:19, 11.56s/step, acc=0.4729, loss=2.2192, lr=7.76e-05]
Training:  35%|███▍      | 3842/11082 [12:20:48<23:24:00, 11.64s/step, acc=0.4729, loss=2.2192, lr=7.76e-05]
Training:  35%|███▍      | 3843/11082 [12:20:59<23:18:16, 11.59s/step, acc=0.4729, loss=2.2192, lr=7.76e-05]
Training:  35%|███▍      | 3844/11082 [12:21:11<23:10:36, 11.53s/step, acc=0.4729, loss=2.2192, lr=7.76e-05]
Training:  35%|███▍      | 3845/11082 [12:21:22<23:09:06, 11.52s/step, acc=0.4729, loss=2.2192, lr=7.76e-05]
Training:  35%|███▍      | 3846/11082 [12:21:33<22:42:57, 11.30s/step, acc=0.4729, loss=2.2192, lr=7.76e-05]
Training:  35%|███▍      | 3847/11082 [12:21:45<23:25:19, 11.65s/step, acc=0.4729, loss=2.2192, lr=7.76e-05]
Training:  35%|███▍      | 3848/11082 [12:21:56<23:09:05, 11.52s/step, acc=0.4729, loss=2.2192, lr=7.76e-05]
Training:  35%|███▍      | 3849/11082 [12:22:08<23:07:13, 11.51s/step, acc=0.4729, loss=2.2192, lr=7.76e-05]
Training:  35%|███▍      | 3850/11082 [12:22:20<23:27:07, 11.67s/step, acc=0.4729, loss=2.2192, lr=7.76e-05]
Training:  35%|███▍      | 3850/11082 [12:22:27<23:27:07, 11.67s/step, acc=0.4758, loss=2.2017, lr=7.70e-05]2026-06-19 03:15:55 | INFO | __main__ | epoch=0 step=3850 train_loss=2.201670 train_acc=0.475805 lr=7.699e-05 grad_norm=0.9375 acc_d0=0.7406 acc_d1=0.5879 acc_d2=0.4624 acc_d3=0.3629 acc_d4=0.2854 acc_d5=0.2295 acc_d6=0.1869 acc_d7=0.1549

Training:  35%|███▍      | 3851/11082 [12:22:31<23:12:47, 11.56s/step, acc=0.4758, loss=2.2017, lr=7.70e-05]
Training:  35%|███▍      | 3852/11082 [12:22:43<23:26:49, 11.67s/step, acc=0.4758, loss=2.2017, lr=7.70e-05]
Training:  35%|███▍      | 3853/11082 [12:22:55<23:21:22, 11.63s/step, acc=0.4758, loss=2.2017, lr=7.70e-05]
Training:  35%|███▍      | 3854/11082 [12:23:06<23:17:11, 11.60s/step, acc=0.4758, loss=2.2017, lr=7.70e-05]
Training:  35%|███▍      | 3855/11082 [12:23:17<22:55:40, 11.42s/step, acc=0.4758, loss=2.2017, lr=7.70e-05]
Training:  35%|███▍      | 3856/11082 [12:23:30<23:25:03, 11.67s/step, acc=0.4758, loss=2.2017, lr=7.70e-05]
Training:  35%|███▍      | 3857/11082 [12:23:41<23:21:22, 11.64s/step, acc=0.4758, loss=2.2017, lr=7.70e-05]
Training:  35%|███▍      | 3858/11082 [12:23:53<23:41:29, 11.81s/step, acc=0.4758, loss=2.2017, lr=7.70e-05]
Training:  35%|███▍      | 3859/11082 [12:24:04<23:13:33, 11.58s/step, acc=0.4758, loss=2.2017, lr=7.70e-05]
Training:  35%|███▍      | 3860/11082 [12:24:16<23:15:50, 11.60s/step, acc=0.4758, loss=2.2017, lr=7.70e-05]
Training:  35%|███▍      | 3861/11082 [12:24:28<23:22:06, 11.65s/step, acc=0.4758, loss=2.2017, lr=7.70e-05]
Training:  35%|███▍      | 3862/11082 [12:24:39<23:00:38, 11.47s/step, acc=0.4758, loss=2.2017, lr=7.70e-05]
Training:  35%|███▍      | 3863/11082 [12:24:50<23:00:24, 11.47s/step, acc=0.4758, loss=2.2017, lr=7.70e-05]
Training:  35%|███▍      | 3864/11082 [12:25:02<23:06:09, 11.52s/step, acc=0.4758, loss=2.2017, lr=7.70e-05]
Training:  35%|███▍      | 3865/11082 [12:25:13<22:59:09, 11.47s/step, acc=0.4758, loss=2.2017, lr=7.70e-05]
Training:  35%|███▍      | 3866/11082 [12:25:25<23:22:22, 11.66s/step, acc=0.4758, loss=2.2017, lr=7.70e-05]
Training:  35%|███▍      | 3867/11082 [12:25:37<23:07:47, 11.54s/step, acc=0.4758, loss=2.2017, lr=7.70e-05]
Training:  35%|███▍      | 3868/11082 [12:25:48<23:12:36, 11.58s/step, acc=0.4758, loss=2.2017, lr=7.70e-05]
Training:  35%|███▍      | 3869/11082 [12:25:59<22:54:36, 11.43s/step, acc=0.4758, loss=2.2017, lr=7.70e-05]
Training:  35%|███▍      | 3870/11082 [12:26:11<23:01:37, 11.49s/step, acc=0.4758, loss=2.2017, lr=7.70e-05]
Training:  35%|███▍      | 3871/11082 [12:26:23<22:59:36, 11.48s/step, acc=0.4758, loss=2.2017, lr=7.70e-05]
Training:  35%|███▍      | 3872/11082 [12:26:35<23:30:12, 11.74s/step, acc=0.4758, loss=2.2017, lr=7.70e-05]
Training:  35%|███▍      | 3873/11082 [12:26:46<23:13:19, 11.60s/step, acc=0.4758, loss=2.2017, lr=7.70e-05]
Training:  35%|███▍      | 3874/11082 [12:26:57<23:00:02, 11.49s/step, acc=0.4758, loss=2.2017, lr=7.70e-05]
Training:  35%|███▍      | 3875/11082 [12:27:09<23:03:09, 11.52s/step, acc=0.4758, loss=2.2017, lr=7.70e-05]
Training:  35%|███▍      | 3876/11082 [12:27:21<23:16:18, 11.63s/step, acc=0.4758, loss=2.2017, lr=7.70e-05]
Training:  35%|███▍      | 3877/11082 [12:27:32<23:13:33, 11.60s/step, acc=0.4758, loss=2.2017, lr=7.70e-05]
Training:  35%|███▍      | 3878/11082 [12:27:44<23:28:24, 11.73s/step, acc=0.4758, loss=2.2017, lr=7.70e-05]
Training:  35%|███▌      | 3879/11082 [12:27:55<22:57:44, 11.48s/step, acc=0.4758, loss=2.2017, lr=7.70e-05]
Training:  35%|███▌      | 3880/11082 [12:28:07<23:18:13, 11.65s/step, acc=0.4758, loss=2.2017, lr=7.70e-05]
Training:  35%|███▌      | 3881/11082 [12:28:18<22:49:51, 11.41s/step, acc=0.4758, loss=2.2017, lr=7.70e-05]
Training:  35%|███▌      | 3882/11082 [12:28:30<23:12:52, 11.61s/step, acc=0.4758, loss=2.2017, lr=7.70e-05]
Training:  35%|███▌      | 3883/11082 [12:28:42<23:01:12, 11.51s/step, acc=0.4758, loss=2.2017, lr=7.70e-05]
Training:  35%|███▌      | 3884/11082 [12:28:53<23:09:21, 11.58s/step, acc=0.4758, loss=2.2017, lr=7.70e-05]
Training:  35%|███▌      | 3885/11082 [12:29:05<22:58:30, 11.49s/step, acc=0.4758, loss=2.2017, lr=7.70e-05]
Training:  35%|███▌      | 3886/11082 [12:29:16<23:13:14, 11.62s/step, acc=0.4758, loss=2.2017, lr=7.70e-05]
Training:  35%|███▌      | 3887/11082 [12:29:28<23:08:43, 11.58s/step, acc=0.4758, loss=2.2017, lr=7.70e-05]
Training:  35%|███▌      | 3888/11082 [12:29:39<23:04:22, 11.55s/step, acc=0.4758, loss=2.2017, lr=7.70e-05]
Training:  35%|███▌      | 3889/11082 [12:29:51<23:00:49, 11.52s/step, acc=0.4758, loss=2.2017, lr=7.70e-05]
Training:  35%|███▌      | 3890/11082 [12:30:02<22:57:58, 11.50s/step, acc=0.4758, loss=2.2017, lr=7.70e-05]
Training:  35%|███▌      | 3891/11082 [12:30:14<22:59:25, 11.51s/step, acc=0.4758, loss=2.2017, lr=7.70e-05]
Training:  35%|███▌      | 3892/11082 [12:30:26<23:05:04, 11.56s/step, acc=0.4758, loss=2.2017, lr=7.70e-05]
Training:  35%|███▌      | 3893/11082 [12:30:36<22:42:12, 11.37s/step, acc=0.4758, loss=2.2017, lr=7.70e-05]
Training:  35%|███▌      | 3894/11082 [12:30:49<23:10:31, 11.61s/step, acc=0.4758, loss=2.2017, lr=7.70e-05]
Training:  35%|███▌      | 3895/11082 [12:31:00<22:46:25, 11.41s/step, acc=0.4758, loss=2.2017, lr=7.70e-05]
Training:  35%|███▌      | 3896/11082 [12:31:12<23:15:08, 11.65s/step, acc=0.4758, loss=2.2017, lr=7.70e-05]
Training:  35%|███▌      | 3897/11082 [12:31:23<22:50:52, 11.45s/step, acc=0.4758, loss=2.2017, lr=7.70e-05]
Training:  35%|███▌      | 3898/11082 [12:31:35<23:02:19, 11.55s/step, acc=0.4758, loss=2.2017, lr=7.70e-05]
Training:  35%|███▌      | 3899/11082 [12:31:46<22:59:28, 11.52s/step, acc=0.4758, loss=2.2017, lr=7.70e-05]
Training:  35%|███▌      | 3900/11082 [12:31:58<23:00:48, 11.54s/step, acc=0.4758, loss=2.2017, lr=7.70e-05]
Training:  35%|███▌      | 3900/11082 [12:32:05<23:00:48, 11.54s/step, acc=0.4766, loss=2.1961, lr=7.64e-05]2026-06-19 03:25:33 | INFO | __main__ | epoch=0 step=3900 train_loss=2.196139 train_acc=0.476646 lr=7.642e-05 grad_norm=0.9219 acc_d0=0.7430 acc_d1=0.5931 acc_d2=0.4686 acc_d3=0.3680 acc_d4=0.2928 acc_d5=0.2334 acc_d6=0.1890 acc_d7=0.1577

Training:  35%|███▌      | 3901/11082 [12:32:10<23:16:56, 11.67s/step, acc=0.4766, loss=2.1961, lr=7.64e-05]
Training:  35%|███▌      | 3902/11082 [12:32:21<22:59:23, 11.53s/step, acc=0.4766, loss=2.1961, lr=7.64e-05]
Training:  35%|███▌      | 3903/11082 [12:32:32<22:50:36, 11.46s/step, acc=0.4766, loss=2.1961, lr=7.64e-05]
Training:  35%|███▌      | 3904/11082 [12:32:44<22:58:39, 11.52s/step, acc=0.4766, loss=2.1961, lr=7.64e-05]
Training:  35%|███▌      | 3905/11082 [12:32:56<23:26:37, 11.76s/step, acc=0.4766, loss=2.1961, lr=7.64e-05]
Training:  35%|███▌      | 3906/11082 [12:33:07<23:10:41, 11.63s/step, acc=0.4766, loss=2.1961, lr=7.64e-05]
Training:  35%|███▌      | 3907/11082 [12:33:19<23:06:25, 11.59s/step, acc=0.4766, loss=2.1961, lr=7.64e-05]
Training:  35%|███▌      | 3908/11082 [12:33:30<23:06:04, 11.59s/step, acc=0.4766, loss=2.1961, lr=7.64e-05]
Training:  35%|███▌      | 3909/11082 [12:33:42<22:59:44, 11.54s/step, acc=0.4766, loss=2.1961, lr=7.64e-05]
Training:  35%|███▌      | 3910/11082 [12:33:54<23:02:52, 11.57s/step, acc=0.4766, loss=2.1961, lr=7.64e-05]
Training:  35%|███▌      | 3911/11082 [12:34:05<22:59:56, 11.55s/step, acc=0.4766, loss=2.1961, lr=7.64e-05]
Training:  35%|███▌      | 3912/11082 [12:34:17<22:58:58, 11.54s/step, acc=0.4766, loss=2.1961, lr=7.64e-05]
Training:  35%|███▌      | 3913/11082 [12:34:28<23:02:58, 11.57s/step, acc=0.4766, loss=2.1961, lr=7.64e-05]
Training:  35%|███▌      | 3914/11082 [12:34:39<22:52:18, 11.49s/step, acc=0.4766, loss=2.1961, lr=7.64e-05]
Training:  35%|███▌      | 3915/11082 [12:34:51<22:56:36, 11.52s/step, acc=0.4766, loss=2.1961, lr=7.64e-05]
Training:  35%|███▌      | 3916/11082 [12:35:03<23:14:37, 11.68s/step, acc=0.4766, loss=2.1961, lr=7.64e-05]
Training:  35%|███▌      | 3917/11082 [12:35:15<23:09:55, 11.64s/step, acc=0.4766, loss=2.1961, lr=7.64e-05]
Training:  35%|███▌      | 3918/11082 [12:35:26<22:59:43, 11.56s/step, acc=0.4766, loss=2.1961, lr=7.64e-05]
Training:  35%|███▌      | 3919/11082 [12:35:38<22:59:12, 11.55s/step, acc=0.4766, loss=2.1961, lr=7.64e-05]
Training:  35%|███▌      | 3920/11082 [12:35:49<23:08:50, 11.64s/step, acc=0.4766, loss=2.1961, lr=7.64e-05]
Training:  35%|███▌      | 3921/11082 [12:36:01<22:51:34, 11.49s/step, acc=0.4766, loss=2.1961, lr=7.64e-05]
Training:  35%|███▌      | 3922/11082 [12:36:13<23:16:07, 11.70s/step, acc=0.4766, loss=2.1961, lr=7.64e-05]
Training:  35%|███▌      | 3923/11082 [12:36:24<22:53:00, 11.51s/step, acc=0.4766, loss=2.1961, lr=7.64e-05]
Training:  35%|███▌      | 3924/11082 [12:36:36<23:03:36, 11.60s/step, acc=0.4766, loss=2.1961, lr=7.64e-05]
Training:  35%|███▌      | 3925/11082 [12:36:47<22:57:34, 11.55s/step, acc=0.4766, loss=2.1961, lr=7.64e-05]
Training:  35%|███▌      | 3926/11082 [12:36:59<23:03:17, 11.60s/step, acc=0.4766, loss=2.1961, lr=7.64e-05]
Training:  35%|███▌      | 3927/11082 [12:37:10<22:56:03, 11.54s/step, acc=0.4766, loss=2.1961, lr=7.64e-05]
Training:  35%|███▌      | 3928/11082 [12:37:22<23:00:46, 11.58s/step, acc=0.4766, loss=2.1961, lr=7.64e-05]
Training:  35%|███▌      | 3929/11082 [12:37:33<22:55:14, 11.54s/step, acc=0.4766, loss=2.1961, lr=7.64e-05]
Training:  35%|███▌      | 3930/11082 [12:37:45<22:53:53, 11.53s/step, acc=0.4766, loss=2.1961, lr=7.64e-05]
Training:  35%|███▌      | 3931/11082 [12:37:56<22:56:47, 11.55s/step, acc=0.4766, loss=2.1961, lr=7.64e-05]
Training:  35%|███▌      | 3932/11082 [12:38:08<22:50:07, 11.50s/step, acc=0.4766, loss=2.1961, lr=7.64e-05]
Training:  35%|███▌      | 3933/11082 [12:38:20<23:00:08, 11.58s/step, acc=0.4766, loss=2.1961, lr=7.64e-05]
Training:  35%|███▌      | 3934/11082 [12:38:32<23:30:32, 11.84s/step, acc=0.4766, loss=2.1961, lr=7.64e-05]
Training:  36%|███▌      | 3935/11082 [12:38:43<23:04:00, 11.62s/step, acc=0.4766, loss=2.1961, lr=7.64e-05]
Training:  36%|███▌      | 3936/11082 [12:38:55<23:04:43, 11.63s/step, acc=0.4766, loss=2.1961, lr=7.64e-05]
Training:  36%|███▌      | 3937/11082 [12:39:07<23:12:30, 11.69s/step, acc=0.4766, loss=2.1961, lr=7.64e-05]
Training:  36%|███▌      | 3938/11082 [12:39:18<22:58:30, 11.58s/step, acc=0.4766, loss=2.1961, lr=7.64e-05]
Training:  36%|███▌      | 3939/11082 [12:39:30<23:17:02, 11.73s/step, acc=0.4766, loss=2.1961, lr=7.64e-05]
Training:  36%|███▌      | 3940/11082 [12:39:41<23:07:42, 11.66s/step, acc=0.4766, loss=2.1961, lr=7.64e-05]
Training:  36%|███▌      | 3941/11082 [12:39:53<23:04:57, 11.64s/step, acc=0.4766, loss=2.1961, lr=7.64e-05]
Training:  36%|███▌      | 3942/11082 [12:40:05<23:05:39, 11.64s/step, acc=0.4766, loss=2.1961, lr=7.64e-05]
Training:  36%|███▌      | 3943/11082 [12:40:16<22:51:48, 11.53s/step, acc=0.4766, loss=2.1961, lr=7.64e-05]
Training:  36%|███▌      | 3944/11082 [12:40:28<23:03:50, 11.63s/step, acc=0.4766, loss=2.1961, lr=7.64e-05]
Training:  36%|███▌      | 3945/11082 [12:40:39<22:36:18, 11.40s/step, acc=0.4766, loss=2.1961, lr=7.64e-05]
Training:  36%|███▌      | 3946/11082 [12:40:51<23:09:37, 11.68s/step, acc=0.4766, loss=2.1961, lr=7.64e-05]
Training:  36%|███▌      | 3947/11082 [12:41:03<23:02:54, 11.63s/step, acc=0.4766, loss=2.1961, lr=7.64e-05]
Training:  36%|███▌      | 3948/11082 [12:41:14<23:01:49, 11.62s/step, acc=0.4766, loss=2.1961, lr=7.64e-05]
Training:  36%|███▌      | 3949/11082 [12:41:25<22:39:58, 11.44s/step, acc=0.4766, loss=2.1961, lr=7.64e-05]
Training:  36%|███▌      | 3950/11082 [12:41:37<22:53:27, 11.55s/step, acc=0.4766, loss=2.1961, lr=7.64e-05]
Training:  36%|███▌      | 3950/11082 [12:41:44<22:53:27, 11.55s/step, acc=0.4771, loss=2.1988, lr=7.59e-05]2026-06-19 03:35:12 | INFO | __main__ | epoch=0 step=3950 train_loss=2.198832 train_acc=0.477070 lr=7.585e-05 grad_norm=0.9805 acc_d0=0.7613 acc_d1=0.6131 acc_d2=0.4866 acc_d3=0.3863 acc_d4=0.3064 acc_d5=0.2466 acc_d6=0.2009 acc_d7=0.1636

Training:  36%|███▌      | 3951/11082 [12:41:49<22:56:16, 11.58s/step, acc=0.4771, loss=2.1988, lr=7.59e-05]
Training:  36%|███▌      | 3952/11082 [12:42:00<22:53:34, 11.56s/step, acc=0.4771, loss=2.1988, lr=7.59e-05]
Training:  36%|███▌      | 3953/11082 [12:42:12<22:56:59, 11.59s/step, acc=0.4771, loss=2.1988, lr=7.59e-05]
Training:  36%|███▌      | 3954/11082 [12:42:23<22:55:09, 11.58s/step, acc=0.4771, loss=2.1988, lr=7.59e-05]
Training:  36%|███▌      | 3955/11082 [12:42:35<22:43:05, 11.48s/step, acc=0.4771, loss=2.1988, lr=7.59e-05]
Training:  36%|███▌      | 3956/11082 [12:42:46<22:51:58, 11.55s/step, acc=0.4771, loss=2.1988, lr=7.59e-05]
Training:  36%|███▌      | 3957/11082 [12:42:58<23:00:19, 11.62s/step, acc=0.4771, loss=2.1988, lr=7.59e-05]
Training:  36%|███▌      | 3958/11082 [12:43:09<22:41:39, 11.47s/step, acc=0.4771, loss=2.1988, lr=7.59e-05]
Training:  36%|███▌      | 3959/11082 [12:43:21<23:08:15, 11.69s/step, acc=0.4771, loss=2.1988, lr=7.59e-05]
Training:  36%|███▌      | 3960/11082 [12:43:33<23:19:41, 11.79s/step, acc=0.4771, loss=2.1988, lr=7.59e-05]
Training:  36%|███▌      | 3961/11082 [12:43:45<23:11:28, 11.72s/step, acc=0.4771, loss=2.1988, lr=7.59e-05]
Training:  36%|███▌      | 3962/11082 [12:43:56<22:50:16, 11.55s/step, acc=0.4771, loss=2.1988, lr=7.59e-05]
Training:  36%|███▌      | 3963/11082 [12:44:08<23:04:11, 11.67s/step, acc=0.4771, loss=2.1988, lr=7.59e-05]
Training:  36%|███▌      | 3964/11082 [12:44:20<23:13:15, 11.74s/step, acc=0.4771, loss=2.1988, lr=7.59e-05]
Training:  36%|███▌      | 3965/11082 [12:44:31<22:51:09, 11.56s/step, acc=0.4771, loss=2.1988, lr=7.59e-05]
Training:  36%|███▌      | 3966/11082 [12:44:43<22:52:17, 11.57s/step, acc=0.4771, loss=2.1988, lr=7.59e-05]
Training:  36%|███▌      | 3967/11082 [12:44:55<23:04:05, 11.67s/step, acc=0.4771, loss=2.1988, lr=7.59e-05]
Training:  36%|███▌      | 3968/11082 [12:45:06<22:54:23, 11.59s/step, acc=0.4771, loss=2.1988, lr=7.59e-05]
Training:  36%|███▌      | 3969/11082 [12:45:17<22:42:28, 11.49s/step, acc=0.4771, loss=2.1988, lr=7.59e-05]
Training:  36%|███▌      | 3970/11082 [12:45:29<22:51:10, 11.57s/step, acc=0.4771, loss=2.1988, lr=7.59e-05]
Training:  36%|███▌      | 3971/11082 [12:45:41<22:56:02, 11.61s/step, acc=0.4771, loss=2.1988, lr=7.59e-05]
Training:  36%|███▌      | 3972/11082 [12:45:53<23:03:16, 11.67s/step, acc=0.4771, loss=2.1988, lr=7.59e-05]
Training:  36%|███▌      | 3973/11082 [12:46:04<22:39:57, 11.48s/step, acc=0.4771, loss=2.1988, lr=7.59e-05]
Training:  36%|███▌      | 3974/11082 [12:46:16<22:57:46, 11.63s/step, acc=0.4771, loss=2.1988, lr=7.59e-05]
Training:  36%|███▌      | 3975/11082 [12:46:27<22:50:47, 11.57s/step, acc=0.4771, loss=2.1988, lr=7.59e-05]
Training:  36%|███▌      | 3976/11082 [12:46:39<22:51:06, 11.58s/step, acc=0.4771, loss=2.1988, lr=7.59e-05]
Training:  36%|███▌      | 3977/11082 [12:46:50<22:41:55, 11.50s/step, acc=0.4771, loss=2.1988, lr=7.59e-05]
Training:  36%|███▌      | 3978/11082 [12:47:02<22:44:56, 11.53s/step, acc=0.4771, loss=2.1988, lr=7.59e-05]
Training:  36%|███▌      | 3979/11082 [12:47:13<22:48:47, 11.56s/step, acc=0.4771, loss=2.1988, lr=7.59e-05]
Training:  36%|███▌      | 3980/11082 [12:47:25<22:47:04, 11.55s/step, acc=0.4771, loss=2.1988, lr=7.59e-05]
Training:  36%|███▌      | 3981/11082 [12:47:36<22:39:44, 11.49s/step, acc=0.4771, loss=2.1988, lr=7.59e-05]
Training:  36%|███▌      | 3982/11082 [12:47:48<22:41:07, 11.50s/step, acc=0.4771, loss=2.1988, lr=7.59e-05]
Training:  36%|███▌      | 3983/11082 [12:47:59<22:44:46, 11.54s/step, acc=0.4771, loss=2.1988, lr=7.59e-05]
Training:  36%|███▌      | 3984/11082 [12:48:10<22:35:32, 11.46s/step, acc=0.4771, loss=2.1988, lr=7.59e-05]
Training:  36%|███▌      | 3985/11082 [12:48:22<22:48:44, 11.57s/step, acc=0.4771, loss=2.1988, lr=7.59e-05]
Training:  36%|███▌      | 3986/11082 [12:48:34<22:43:06, 11.53s/step, acc=0.4771, loss=2.1988, lr=7.59e-05]
Training:  36%|███▌      | 3987/11082 [12:48:45<22:44:13, 11.54s/step, acc=0.4771, loss=2.1988, lr=7.59e-05]
Training:  36%|███▌      | 3988/11082 [12:48:57<22:36:10, 11.47s/step, acc=0.4771, loss=2.1988, lr=7.59e-05]
Training:  36%|███▌      | 3989/11082 [12:49:08<22:33:01, 11.45s/step, acc=0.4771, loss=2.1988, lr=7.59e-05]
Training:  36%|███▌      | 3990/11082 [12:49:20<22:39:39, 11.50s/step, acc=0.4771, loss=2.1988, lr=7.59e-05]
Training:  36%|███▌      | 3991/11082 [12:49:31<22:51:00, 11.60s/step, acc=0.4771, loss=2.1988, lr=7.59e-05]
Training:  36%|███▌      | 3992/11082 [12:49:43<22:40:51, 11.52s/step, acc=0.4771, loss=2.1988, lr=7.59e-05]
Training:  36%|███▌      | 3993/11082 [12:49:54<22:36:24, 11.48s/step, acc=0.4771, loss=2.1988, lr=7.59e-05]
Training:  36%|███▌      | 3994/11082 [12:50:05<22:29:44, 11.43s/step, acc=0.4771, loss=2.1988, lr=7.59e-05]
Training:  36%|███▌      | 3995/11082 [12:50:18<23:00:36, 11.69s/step, acc=0.4771, loss=2.1988, lr=7.59e-05]
Training:  36%|███▌      | 3996/11082 [12:50:29<22:43:16, 11.54s/step, acc=0.4771, loss=2.1988, lr=7.59e-05]
Training:  36%|███▌      | 3997/11082 [12:50:41<22:56:41, 11.66s/step, acc=0.4771, loss=2.1988, lr=7.59e-05]
Training:  36%|███▌      | 3998/11082 [12:50:52<22:38:35, 11.51s/step, acc=0.4771, loss=2.1988, lr=7.59e-05]
Training:  36%|███▌      | 3999/11082 [12:51:04<22:53:03, 11.63s/step, acc=0.4771, loss=2.1988, lr=7.59e-05]
Training:  36%|███▌      | 4000/11082 [12:51:15<22:36:56, 11.50s/step, acc=0.4771, loss=2.1988, lr=7.59e-05]
Training:  36%|███▌      | 4000/11082 [12:51:22<22:36:56, 11.50s/step, acc=0.4806, loss=2.1721, lr=7.53e-05]2026-06-19 03:44:50 | INFO | __main__ | epoch=0 step=4000 train_loss=2.172070 train_acc=0.480562 lr=7.527e-05 grad_norm=2.0469 acc_d0=0.7499 acc_d1=0.6011 acc_d2=0.4767 acc_d3=0.3774 acc_d4=0.3010 acc_d5=0.2409 acc_d6=0.1965 acc_d7=0.1607

Training:  36%|███▌      | 4001/11082 [12:51:27<22:59:30, 11.69s/step, acc=0.4806, loss=2.1721, lr=7.53e-05]
Training:  36%|███▌      | 4002/11082 [12:51:38<22:28:48, 11.43s/step, acc=0.4806, loss=2.1721, lr=7.53e-05]
Training:  36%|███▌      | 4003/11082 [12:51:50<22:43:41, 11.56s/step, acc=0.4806, loss=2.1721, lr=7.53e-05]
Training:  36%|███▌      | 4004/11082 [12:52:02<22:53:08, 11.64s/step, acc=0.4806, loss=2.1721, lr=7.53e-05]
Training:  36%|███▌      | 4005/11082 [12:52:13<22:44:46, 11.57s/step, acc=0.4806, loss=2.1721, lr=7.53e-05]
Training:  36%|███▌      | 4006/11082 [12:52:25<22:39:43, 11.53s/step, acc=0.4806, loss=2.1721, lr=7.53e-05]
Training:  36%|███▌      | 4007/11082 [12:52:36<22:47:37, 11.60s/step, acc=0.4806, loss=2.1721, lr=7.53e-05]
Training:  36%|███▌      | 4008/11082 [12:52:48<22:36:01, 11.50s/step, acc=0.4806, loss=2.1721, lr=7.53e-05]
Training:  36%|███▌      | 4009/11082 [12:53:00<22:47:21, 11.60s/step, acc=0.4806, loss=2.1721, lr=7.53e-05]
Training:  36%|███▌      | 4010/11082 [12:53:11<22:39:48, 11.54s/step, acc=0.4806, loss=2.1721, lr=7.53e-05]
Training:  36%|███▌      | 4011/11082 [12:53:22<22:34:52, 11.50s/step, acc=0.4806, loss=2.1721, lr=7.53e-05]
Training:  36%|███▌      | 4012/11082 [12:53:34<22:36:45, 11.51s/step, acc=0.4806, loss=2.1721, lr=7.53e-05]
Training:  36%|███▌      | 4013/11082 [12:53:46<22:41:30, 11.56s/step, acc=0.4806, loss=2.1721, lr=7.53e-05]
Training:  36%|███▌      | 4014/11082 [12:53:57<22:44:48, 11.59s/step, acc=0.4806, loss=2.1721, lr=7.53e-05]
Training:  36%|███▌      | 4015/11082 [12:54:09<22:54:14, 11.67s/step, acc=0.4806, loss=2.1721, lr=7.53e-05]
Training:  36%|███▌      | 4016/11082 [12:54:20<22:38:47, 11.54s/step, acc=0.4806, loss=2.1721, lr=7.53e-05]
Training:  36%|███▌      | 4017/11082 [12:54:33<23:06:36, 11.78s/step, acc=0.4806, loss=2.1721, lr=7.53e-05]
Training:  36%|███▋      | 4018/11082 [12:54:44<23:01:01, 11.73s/step, acc=0.4806, loss=2.1721, lr=7.53e-05]
Training:  36%|███▋      | 4019/11082 [12:54:56<22:44:42, 11.59s/step, acc=0.4806, loss=2.1721, lr=7.53e-05]
Training:  36%|███▋      | 4020/11082 [12:55:07<22:50:38, 11.65s/step, acc=0.4806, loss=2.1721, lr=7.53e-05]
Training:  36%|███▋      | 4021/11082 [12:55:19<22:56:56, 11.70s/step, acc=0.4806, loss=2.1721, lr=7.53e-05]
Training:  36%|███▋      | 4022/11082 [12:55:31<22:48:28, 11.63s/step, acc=0.4806, loss=2.1721, lr=7.53e-05]
Training:  36%|███▋      | 4023/11082 [12:55:42<22:47:40, 11.62s/step, acc=0.4806, loss=2.1721, lr=7.53e-05]
Training:  36%|███▋      | 4024/11082 [12:55:54<22:49:50, 11.65s/step, acc=0.4806, loss=2.1721, lr=7.53e-05]
Training:  36%|███▋      | 4025/11082 [12:56:06<22:51:12, 11.66s/step, acc=0.4806, loss=2.1721, lr=7.53e-05]
Training:  36%|███▋      | 4026/11082 [12:56:17<22:34:46, 11.52s/step, acc=0.4806, loss=2.1721, lr=7.53e-05]
Training:  36%|███▋      | 4027/11082 [12:56:28<22:29:01, 11.47s/step, acc=0.4806, loss=2.1721, lr=7.53e-05]
Training:  36%|███▋      | 4028/11082 [12:56:40<22:42:08, 11.59s/step, acc=0.4806, loss=2.1721, lr=7.53e-05]
Training:  36%|███▋      | 4029/11082 [12:56:51<22:30:48, 11.49s/step, acc=0.4806, loss=2.1721, lr=7.53e-05]
Training:  36%|███▋      | 4030/11082 [12:57:02<22:22:05, 11.42s/step, acc=0.4806, loss=2.1721, lr=7.53e-05]
Training:  36%|███▋      | 4031/11082 [12:57:15<22:53:08, 11.68s/step, acc=0.4806, loss=2.1721, lr=7.53e-05]
Training:  36%|███▋      | 4032/11082 [12:57:26<22:41:15, 11.59s/step, acc=0.4806, loss=2.1721, lr=7.53e-05]
Training:  36%|███▋      | 4033/11082 [12:57:37<22:31:59, 11.51s/step, acc=0.4806, loss=2.1721, lr=7.53e-05]
Training:  36%|███▋      | 4034/11082 [12:57:49<22:35:55, 11.54s/step, acc=0.4806, loss=2.1721, lr=7.53e-05]
Training:  36%|███▋      | 4035/11082 [12:58:01<22:37:05, 11.55s/step, acc=0.4806, loss=2.1721, lr=7.53e-05]
Training:  36%|███▋      | 4036/11082 [12:58:12<22:39:51, 11.58s/step, acc=0.4806, loss=2.1721, lr=7.53e-05]
Training:  36%|███▋      | 4037/11082 [12:58:24<22:26:41, 11.47s/step, acc=0.4806, loss=2.1721, lr=7.53e-05]
Training:  36%|███▋      | 4038/11082 [12:58:36<22:48:19, 11.66s/step, acc=0.4806, loss=2.1721, lr=7.53e-05]
Training:  36%|███▋      | 4039/11082 [12:58:47<22:42:00, 11.60s/step, acc=0.4806, loss=2.1721, lr=7.53e-05]
Training:  36%|███▋      | 4040/11082 [12:58:59<22:42:33, 11.61s/step, acc=0.4806, loss=2.1721, lr=7.53e-05]
Training:  36%|███▋      | 4041/11082 [12:59:11<22:49:37, 11.67s/step, acc=0.4806, loss=2.1721, lr=7.53e-05]
Training:  36%|███▋      | 4042/11082 [12:59:22<22:40:39, 11.60s/step, acc=0.4806, loss=2.1721, lr=7.53e-05]
Training:  36%|███▋      | 4043/11082 [12:59:33<22:29:55, 11.51s/step, acc=0.4806, loss=2.1721, lr=7.53e-05]
Training:  36%|███▋      | 4044/11082 [12:59:45<22:49:29, 11.68s/step, acc=0.4806, loss=2.1721, lr=7.53e-05]
Training:  37%|███▋      | 4045/11082 [12:59:57<22:32:33, 11.53s/step, acc=0.4806, loss=2.1721, lr=7.53e-05]
Training:  37%|███▋      | 4046/11082 [13:00:08<22:22:51, 11.45s/step, acc=0.4806, loss=2.1721, lr=7.53e-05]
Training:  37%|███▋      | 4047/11082 [13:00:20<22:37:14, 11.58s/step, acc=0.4806, loss=2.1721, lr=7.53e-05]
Training:  37%|███▋      | 4048/11082 [13:00:31<22:33:36, 11.55s/step, acc=0.4806, loss=2.1721, lr=7.53e-05]
Training:  37%|███▋      | 4049/11082 [13:00:43<22:51:38, 11.70s/step, acc=0.4806, loss=2.1721, lr=7.53e-05]
Training:  37%|███▋      | 4050/11082 [13:00:55<22:48:25, 11.68s/step, acc=0.4806, loss=2.1721, lr=7.53e-05]
Training:  37%|███▋      | 4050/11082 [13:01:02<22:48:25, 11.68s/step, acc=0.4794, loss=2.1762, lr=7.47e-05]2026-06-19 03:54:30 | INFO | __main__ | epoch=0 step=4050 train_loss=2.176193 train_acc=0.479404 lr=7.469e-05 grad_norm=0.7852 acc_d0=0.7487 acc_d1=0.5973 acc_d2=0.4724 acc_d3=0.3725 acc_d4=0.2936 acc_d5=0.2345 acc_d6=0.1911 acc_d7=0.1593

Training:  37%|███▋      | 4051/11082 [13:01:06<22:20:02, 11.44s/step, acc=0.4794, loss=2.1762, lr=7.47e-05]
Training:  37%|███▋      | 4052/11082 [13:01:17<22:15:58, 11.40s/step, acc=0.4794, loss=2.1762, lr=7.47e-05]
Training:  37%|███▋      | 4053/11082 [13:01:29<22:47:55, 11.68s/step, acc=0.4794, loss=2.1762, lr=7.47e-05]
Training:  37%|███▋      | 4054/11082 [13:01:40<22:27:52, 11.51s/step, acc=0.4794, loss=2.1762, lr=7.47e-05]
Training:  37%|███▋      | 4055/11082 [13:01:52<22:34:02, 11.56s/step, acc=0.4794, loss=2.1762, lr=7.47e-05]
Training:  37%|███▋      | 4056/11082 [13:02:04<22:46:49, 11.67s/step, acc=0.4794, loss=2.1762, lr=7.47e-05]
Training:  37%|███▋      | 4057/11082 [13:02:15<22:36:37, 11.59s/step, acc=0.4794, loss=2.1762, lr=7.47e-05]
Training:  37%|███▋      | 4058/11082 [13:02:27<22:35:51, 11.58s/step, acc=0.4794, loss=2.1762, lr=7.47e-05]
Training:  37%|███▋      | 4059/11082 [13:02:38<22:18:46, 11.44s/step, acc=0.4794, loss=2.1762, lr=7.47e-05]
Training:  37%|███▋      | 4060/11082 [13:02:50<22:34:48, 11.58s/step, acc=0.4794, loss=2.1762, lr=7.47e-05]
Training:  37%|███▋      | 4061/11082 [13:03:02<22:35:04, 11.58s/step, acc=0.4794, loss=2.1762, lr=7.47e-05]
Training:  37%|███▋      | 4062/11082 [13:03:13<22:30:50, 11.55s/step, acc=0.4794, loss=2.1762, lr=7.47e-05]
Training:  37%|███▋      | 4063/11082 [13:03:25<22:33:23, 11.57s/step, acc=0.4794, loss=2.1762, lr=7.47e-05]
Training:  37%|███▋      | 4064/11082 [13:03:36<22:26:56, 11.52s/step, acc=0.4794, loss=2.1762, lr=7.47e-05]
Training:  37%|███▋      | 4065/11082 [13:03:48<22:30:01, 11.54s/step, acc=0.4794, loss=2.1762, lr=7.47e-05]
Training:  37%|███▋      | 4066/11082 [13:03:59<22:14:40, 11.41s/step, acc=0.4794, loss=2.1762, lr=7.47e-05]
Training:  37%|███▋      | 4067/11082 [13:04:11<22:24:26, 11.50s/step, acc=0.4794, loss=2.1762, lr=7.47e-05]
Training:  37%|███▋      | 4068/11082 [13:04:22<22:32:01, 11.57s/step, acc=0.4794, loss=2.1762, lr=7.47e-05]
Training:  37%|███▋      | 4069/11082 [13:04:34<22:29:52, 11.55s/step, acc=0.4794, loss=2.1762, lr=7.47e-05]
Training:  37%|███▋      | 4070/11082 [13:04:45<22:33:06, 11.58s/step, acc=0.4794, loss=2.1762, lr=7.47e-05]
Training:  37%|███▋      | 4071/11082 [13:04:57<22:34:34, 11.59s/step, acc=0.4794, loss=2.1762, lr=7.47e-05]
Training:  37%|███▋      | 4072/11082 [13:05:08<22:19:24, 11.46s/step, acc=0.4794, loss=2.1762, lr=7.47e-05]
Training:  37%|███▋      | 4073/11082 [13:05:20<22:36:38, 11.61s/step, acc=0.4794, loss=2.1762, lr=7.47e-05]
Training:  37%|███▋      | 4074/11082 [13:05:32<22:34:07, 11.59s/step, acc=0.4794, loss=2.1762, lr=7.47e-05]
Training:  37%|███▋      | 4075/11082 [13:05:43<22:21:24, 11.49s/step, acc=0.4794, loss=2.1762, lr=7.47e-05]
Training:  37%|███▋      | 4076/11082 [13:05:54<22:08:02, 11.37s/step, acc=0.4794, loss=2.1762, lr=7.47e-05]
Training:  37%|███▋      | 4077/11082 [13:06:06<22:35:30, 11.61s/step, acc=0.4794, loss=2.1762, lr=7.47e-05]
Training:  37%|███▋      | 4078/11082 [13:06:18<22:31:29, 11.58s/step, acc=0.4794, loss=2.1762, lr=7.47e-05]
Training:  37%|███▋      | 4079/11082 [13:06:29<22:20:16, 11.48s/step, acc=0.4794, loss=2.1762, lr=7.47e-05]
Training:  37%|███▋      | 4080/11082 [13:06:41<22:32:24, 11.59s/step, acc=0.4794, loss=2.1762, lr=7.47e-05]
Training:  37%|███▋      | 4081/11082 [13:06:53<22:38:32, 11.64s/step, acc=0.4794, loss=2.1762, lr=7.47e-05]
Training:  37%|███▋      | 4082/11082 [13:07:04<22:41:15, 11.67s/step, acc=0.4794, loss=2.1762, lr=7.47e-05]
Training:  37%|███▋      | 4083/11082 [13:07:16<22:25:21, 11.53s/step, acc=0.4794, loss=2.1762, lr=7.47e-05]
Training:  37%|███▋      | 4084/11082 [13:07:27<22:28:42, 11.56s/step, acc=0.4794, loss=2.1762, lr=7.47e-05]
Training:  37%|███▋      | 4085/11082 [13:07:39<22:35:50, 11.63s/step, acc=0.4794, loss=2.1762, lr=7.47e-05]
Training:  37%|███▋      | 4086/11082 [13:07:50<22:19:25, 11.49s/step, acc=0.4794, loss=2.1762, lr=7.47e-05]
Training:  37%|███▋      | 4087/11082 [13:08:01<22:15:16, 11.45s/step, acc=0.4794, loss=2.1762, lr=7.47e-05]
Training:  37%|███▋      | 4088/11082 [13:08:13<22:34:28, 11.62s/step, acc=0.4794, loss=2.1762, lr=7.47e-05]
Training:  37%|███▋      | 4089/11082 [13:08:25<22:15:17, 11.46s/step, acc=0.4794, loss=2.1762, lr=7.47e-05]
Training:  37%|███▋      | 4090/11082 [13:08:37<22:38:14, 11.66s/step, acc=0.4794, loss=2.1762, lr=7.47e-05]
Training:  37%|███▋      | 4091/11082 [13:08:48<22:17:23, 11.48s/step, acc=0.4794, loss=2.1762, lr=7.47e-05]
Training:  37%|███▋      | 4092/11082 [13:09:00<22:47:38, 11.74s/step, acc=0.4794, loss=2.1762, lr=7.47e-05]
Training:  37%|███▋      | 4093/11082 [13:09:11<22:28:40, 11.58s/step, acc=0.4794, loss=2.1762, lr=7.47e-05]
Training:  37%|███▋      | 4094/11082 [13:09:23<22:34:59, 11.63s/step, acc=0.4794, loss=2.1762, lr=7.47e-05]
Training:  37%|███▋      | 4095/11082 [13:09:35<22:45:32, 11.73s/step, acc=0.4794, loss=2.1762, lr=7.47e-05]
Training:  37%|███▋      | 4096/11082 [13:09:46<22:25:33, 11.56s/step, acc=0.4794, loss=2.1762, lr=7.47e-05]
Training:  37%|███▋      | 4097/11082 [13:09:58<22:22:45, 11.53s/step, acc=0.4794, loss=2.1762, lr=7.47e-05]
Training:  37%|███▋      | 4098/11082 [13:10:09<22:22:54, 11.54s/step, acc=0.4794, loss=2.1762, lr=7.47e-05]
Training:  37%|███▋      | 4099/11082 [13:10:21<22:29:46, 11.60s/step, acc=0.4794, loss=2.1762, lr=7.47e-05]
Training:  37%|███▋      | 4100/11082 [13:10:33<22:29:43, 11.60s/step, acc=0.4794, loss=2.1762, lr=7.47e-05]
Training:  37%|███▋      | 4100/11082 [13:10:40<22:29:43, 11.60s/step, acc=0.4806, loss=2.1731, lr=7.41e-05]2026-06-19 04:04:07 | INFO | __main__ | epoch=0 step=4100 train_loss=2.173053 train_acc=0.480595 lr=7.411e-05 grad_norm=1.4141 acc_d0=0.7482 acc_d1=0.5987 acc_d2=0.4731 acc_d3=0.3712 acc_d4=0.2938 acc_d5=0.2341 acc_d6=0.1882 acc_d7=0.1567

Training:  37%|███▋      | 4101/11082 [13:10:44<22:31:14, 11.61s/step, acc=0.4806, loss=2.1731, lr=7.41e-05]
Training:  37%|███▋      | 4102/11082 [13:10:56<22:33:48, 11.64s/step, acc=0.4806, loss=2.1731, lr=7.41e-05]
Training:  37%|███▋      | 4103/11082 [13:11:07<22:15:56, 11.49s/step, acc=0.4806, loss=2.1731, lr=7.41e-05]
Training:  37%|███▋      | 4104/11082 [13:11:19<22:26:17, 11.58s/step, acc=0.4806, loss=2.1731, lr=7.41e-05]
Training:  37%|███▋      | 4105/11082 [13:11:31<22:45:09, 11.74s/step, acc=0.4806, loss=2.1731, lr=7.41e-05]
Training:  37%|███▋      | 4106/11082 [13:11:43<22:41:47, 11.71s/step, acc=0.4806, loss=2.1731, lr=7.41e-05]
Training:  37%|███▋      | 4107/11082 [13:11:54<22:19:51, 11.53s/step, acc=0.4806, loss=2.1731, lr=7.41e-05]
Training:  37%|███▋      | 4108/11082 [13:12:05<22:21:37, 11.54s/step, acc=0.4806, loss=2.1731, lr=7.41e-05]
Training:  37%|███▋      | 4109/11082 [13:12:17<22:31:13, 11.63s/step, acc=0.4806, loss=2.1731, lr=7.41e-05]
Training:  37%|███▋      | 4110/11082 [13:12:28<22:19:37, 11.53s/step, acc=0.4806, loss=2.1731, lr=7.41e-05]
Training:  37%|███▋      | 4111/11082 [13:12:40<22:22:26, 11.55s/step, acc=0.4806, loss=2.1731, lr=7.41e-05]
Training:  37%|███▋      | 4112/11082 [13:12:52<22:31:59, 11.64s/step, acc=0.4806, loss=2.1731, lr=7.41e-05]
Training:  37%|███▋      | 4113/11082 [13:13:03<22:20:50, 11.54s/step, acc=0.4806, loss=2.1731, lr=7.41e-05]
Training:  37%|███▋      | 4114/11082 [13:13:15<22:23:44, 11.57s/step, acc=0.4806, loss=2.1731, lr=7.41e-05]
Training:  37%|███▋      | 4115/11082 [13:13:27<22:34:32, 11.67s/step, acc=0.4806, loss=2.1731, lr=7.41e-05]
Training:  37%|███▋      | 4116/11082 [13:13:38<22:37:46, 11.69s/step, acc=0.4806, loss=2.1731, lr=7.41e-05]
Training:  37%|███▋      | 4117/11082 [13:13:50<22:23:43, 11.58s/step, acc=0.4806, loss=2.1731, lr=7.41e-05]
Training:  37%|███▋      | 4118/11082 [13:14:01<22:20:35, 11.55s/step, acc=0.4806, loss=2.1731, lr=7.41e-05]
Training:  37%|███▋      | 4119/11082 [13:14:13<22:25:00, 11.59s/step, acc=0.4806, loss=2.1731, lr=7.41e-05]
Training:  37%|███▋      | 4120/11082 [13:14:24<22:24:54, 11.59s/step, acc=0.4806, loss=2.1731, lr=7.41e-05]
Training:  37%|███▋      | 4121/11082 [13:14:36<22:09:48, 11.46s/step, acc=0.4806, loss=2.1731, lr=7.41e-05]
Training:  37%|███▋      | 4122/11082 [13:14:48<22:25:17, 11.60s/step, acc=0.4806, loss=2.1731, lr=7.41e-05]
Training:  37%|███▋      | 4123/11082 [13:15:00<22:43:14, 11.75s/step, acc=0.4806, loss=2.1731, lr=7.41e-05]
Training:  37%|███▋      | 4124/11082 [13:15:11<22:43:27, 11.76s/step, acc=0.4806, loss=2.1731, lr=7.41e-05]
Training:  37%|███▋      | 4125/11082 [13:15:22<22:19:16, 11.55s/step, acc=0.4806, loss=2.1731, lr=7.41e-05]
Training:  37%|███▋      | 4126/11082 [13:15:34<22:14:13, 11.51s/step, acc=0.4806, loss=2.1731, lr=7.41e-05]
Training:  37%|███▋      | 4127/11082 [13:15:45<22:07:58, 11.46s/step, acc=0.4806, loss=2.1731, lr=7.41e-05]
Training:  37%|███▋      | 4128/11082 [13:15:58<22:41:00, 11.74s/step, acc=0.4806, loss=2.1731, lr=7.41e-05]
Training:  37%|███▋      | 4129/11082 [13:16:08<22:09:59, 11.48s/step, acc=0.4806, loss=2.1731, lr=7.41e-05]
Training:  37%|███▋      | 4130/11082 [13:16:20<22:23:20, 11.59s/step, acc=0.4806, loss=2.1731, lr=7.41e-05]
Training:  37%|███▋      | 4131/11082 [13:16:32<22:15:13, 11.53s/step, acc=0.4806, loss=2.1731, lr=7.41e-05]
Training:  37%|███▋      | 4132/11082 [13:16:44<22:25:14, 11.61s/step, acc=0.4806, loss=2.1731, lr=7.41e-05]
Training:  37%|███▋      | 4133/11082 [13:16:55<22:12:58, 11.51s/step, acc=0.4806, loss=2.1731, lr=7.41e-05]
Training:  37%|███▋      | 4134/11082 [13:17:06<22:18:11, 11.56s/step, acc=0.4806, loss=2.1731, lr=7.41e-05]
Training:  37%|███▋      | 4135/11082 [13:17:19<22:39:56, 11.75s/step, acc=0.4806, loss=2.1731, lr=7.41e-05]
Training:  37%|███▋      | 4136/11082 [13:17:30<22:12:29, 11.51s/step, acc=0.4806, loss=2.1731, lr=7.41e-05]
Training:  37%|███▋      | 4137/11082 [13:17:42<22:32:14, 11.68s/step, acc=0.4806, loss=2.1731, lr=7.41e-05]
Training:  37%|███▋      | 4138/11082 [13:17:53<22:13:22, 11.52s/step, acc=0.4806, loss=2.1731, lr=7.41e-05]
Training:  37%|███▋      | 4139/11082 [13:18:05<22:23:22, 11.61s/step, acc=0.4806, loss=2.1731, lr=7.41e-05]
Training:  37%|███▋      | 4140/11082 [13:18:16<22:00:02, 11.41s/step, acc=0.4806, loss=2.1731, lr=7.41e-05]
Training:  37%|███▋      | 4141/11082 [13:18:28<22:27:42, 11.65s/step, acc=0.4806, loss=2.1731, lr=7.41e-05]
Training:  37%|███▋      | 4142/11082 [13:18:39<22:14:07, 11.53s/step, acc=0.4806, loss=2.1731, lr=7.41e-05]
Training:  37%|███▋      | 4143/11082 [13:18:50<22:04:53, 11.46s/step, acc=0.4806, loss=2.1731, lr=7.41e-05]
Training:  37%|███▋      | 4144/11082 [13:19:02<22:08:07, 11.49s/step, acc=0.4806, loss=2.1731, lr=7.41e-05]
Training:  37%|███▋      | 4145/11082 [13:19:13<22:07:45, 11.48s/step, acc=0.4806, loss=2.1731, lr=7.41e-05]
Training:  37%|███▋      | 4146/11082 [13:19:25<22:23:52, 11.63s/step, acc=0.4806, loss=2.1731, lr=7.41e-05]
Training:  37%|███▋      | 4147/11082 [13:19:36<22:03:01, 11.45s/step, acc=0.4806, loss=2.1731, lr=7.41e-05]
Training:  37%|███▋      | 4148/11082 [13:19:48<22:13:49, 11.54s/step, acc=0.4806, loss=2.1731, lr=7.41e-05]
Training:  37%|███▋      | 4149/11082 [13:20:00<22:31:46, 11.70s/step, acc=0.4806, loss=2.1731, lr=7.41e-05]
Training:  37%|███▋      | 4150/11082 [13:20:11<22:14:56, 11.55s/step, acc=0.4806, loss=2.1731, lr=7.41e-05]
Training:  37%|███▋      | 4150/11082 [13:20:19<22:14:56, 11.55s/step, acc=0.4812, loss=2.1655, lr=7.35e-05]2026-06-19 04:13:47 | INFO | __main__ | epoch=0 step=4150 train_loss=2.165518 train_acc=0.481155 lr=7.352e-05 grad_norm=1.0938 acc_d0=0.7491 acc_d1=0.5967 acc_d2=0.4713 acc_d3=0.3702 acc_d4=0.2933 acc_d5=0.2332 acc_d6=0.1892 acc_d7=0.1568

Training:  37%|███▋      | 4151/11082 [13:20:23<22:23:52, 11.63s/step, acc=0.4812, loss=2.1655, lr=7.35e-05]
Training:  37%|███▋      | 4152/11082 [13:20:34<22:09:35, 11.51s/step, acc=0.4812, loss=2.1655, lr=7.35e-05]
Training:  37%|███▋      | 4153/11082 [13:20:46<22:01:17, 11.44s/step, acc=0.4812, loss=2.1655, lr=7.35e-05]
Training:  37%|███▋      | 4154/11082 [13:20:57<22:09:26, 11.51s/step, acc=0.4812, loss=2.1655, lr=7.35e-05]
Training:  37%|███▋      | 4155/11082 [13:21:09<22:26:32, 11.66s/step, acc=0.4812, loss=2.1655, lr=7.35e-05]
Training:  38%|███▊      | 4156/11082 [13:21:21<22:17:49, 11.59s/step, acc=0.4812, loss=2.1655, lr=7.35e-05]
Training:  38%|███▊      | 4157/11082 [13:21:32<22:10:06, 11.52s/step, acc=0.4812, loss=2.1655, lr=7.35e-05]
Training:  38%|███▊      | 4158/11082 [13:21:44<22:16:21, 11.58s/step, acc=0.4812, loss=2.1655, lr=7.35e-05]
Training:  38%|███▊      | 4159/11082 [13:21:55<22:14:39, 11.57s/step, acc=0.4812, loss=2.1655, lr=7.35e-05]
Training:  38%|███▊      | 4160/11082 [13:22:07<22:18:51, 11.61s/step, acc=0.4812, loss=2.1655, lr=7.35e-05]
Training:  38%|███▊      | 4161/11082 [13:22:19<22:27:15, 11.68s/step, acc=0.4812, loss=2.1655, lr=7.35e-05]
Training:  38%|███▊      | 4162/11082 [13:22:30<21:55:21, 11.40s/step, acc=0.4812, loss=2.1655, lr=7.35e-05]
Training:  38%|███▊      | 4163/11082 [13:22:42<22:19:21, 11.61s/step, acc=0.4812, loss=2.1655, lr=7.35e-05]
Training:  38%|███▊      | 4164/11082 [13:22:53<22:07:19, 11.51s/step, acc=0.4812, loss=2.1655, lr=7.35e-05]
Training:  38%|███▊      | 4165/11082 [13:23:05<22:11:13, 11.55s/step, acc=0.4812, loss=2.1655, lr=7.35e-05]
Training:  38%|███▊      | 4166/11082 [13:23:16<21:59:16, 11.45s/step, acc=0.4812, loss=2.1655, lr=7.35e-05]
Training:  38%|███▊      | 4167/11082 [13:23:28<22:15:47, 11.59s/step, acc=0.4812, loss=2.1655, lr=7.35e-05]
Training:  38%|███▊      | 4168/11082 [13:23:39<22:04:49, 11.50s/step, acc=0.4812, loss=2.1655, lr=7.35e-05]
Training:  38%|███▊      | 4169/11082 [13:23:51<22:05:13, 11.50s/step, acc=0.4812, loss=2.1655, lr=7.35e-05]
Training:  38%|███▊      | 4170/11082 [13:24:02<22:09:47, 11.54s/step, acc=0.4812, loss=2.1655, lr=7.35e-05]
Training:  38%|███▊      | 4171/11082 [13:24:14<22:13:30, 11.58s/step, acc=0.4812, loss=2.1655, lr=7.35e-05]
Training:  38%|███▊      | 4172/11082 [13:24:26<22:15:46, 11.60s/step, acc=0.4812, loss=2.1655, lr=7.35e-05]
Training:  38%|███▊      | 4173/11082 [13:24:37<22:10:06, 11.55s/step, acc=0.4812, loss=2.1655, lr=7.35e-05]
Training:  38%|███▊      | 4174/11082 [13:24:49<22:04:15, 11.50s/step, acc=0.4812, loss=2.1655, lr=7.35e-05]
Training:  38%|███▊      | 4175/11082 [13:25:00<22:07:52, 11.54s/step, acc=0.4812, loss=2.1655, lr=7.35e-05]
Training:  38%|███▊      | 4176/11082 [13:25:11<22:02:09, 11.49s/step, acc=0.4812, loss=2.1655, lr=7.35e-05]
Training:  38%|███▊      | 4177/11082 [13:25:24<22:20:55, 11.65s/step, acc=0.4812, loss=2.1655, lr=7.35e-05]
Training:  38%|███▊      | 4178/11082 [13:25:35<22:23:25, 11.68s/step, acc=0.4812, loss=2.1655, lr=7.35e-05]
Training:  38%|███▊      | 4179/11082 [13:25:47<22:19:36, 11.64s/step, acc=0.4812, loss=2.1655, lr=7.35e-05]
Training:  38%|███▊      | 4180/11082 [13:25:58<22:13:17, 11.59s/step, acc=0.4812, loss=2.1655, lr=7.35e-05]
Training:  38%|███▊      | 4181/11082 [13:26:10<22:28:44, 11.73s/step, acc=0.4812, loss=2.1655, lr=7.35e-05]
Training:  38%|███▊      | 4182/11082 [13:26:22<22:36:00, 11.79s/step, acc=0.4812, loss=2.1655, lr=7.35e-05]
Training:  38%|███▊      | 4183/11082 [13:26:33<22:11:18, 11.58s/step, acc=0.4812, loss=2.1655, lr=7.35e-05]
Training:  38%|███▊      | 4184/11082 [13:26:45<22:20:50, 11.66s/step, acc=0.4812, loss=2.1655, lr=7.35e-05]
Training:  38%|███▊      | 4185/11082 [13:26:57<22:21:08, 11.67s/step, acc=0.4812, loss=2.1655, lr=7.35e-05]
Training:  38%|███▊      | 4186/11082 [13:27:09<22:29:00, 11.74s/step, acc=0.4812, loss=2.1655, lr=7.35e-05]
Training:  38%|███▊      | 4187/11082 [13:27:20<22:14:32, 11.61s/step, acc=0.4812, loss=2.1655, lr=7.35e-05]
Training:  38%|███▊      | 4188/11082 [13:27:31<22:03:28, 11.52s/step, acc=0.4812, loss=2.1655, lr=7.35e-05]
Training:  38%|███▊      | 4189/11082 [13:27:43<21:59:35, 11.49s/step, acc=0.4812, loss=2.1655, lr=7.35e-05]
Training:  38%|███▊      | 4190/11082 [13:27:55<22:16:03, 11.63s/step, acc=0.4812, loss=2.1655, lr=7.35e-05]
Training:  38%|███▊      | 4191/11082 [13:28:06<22:11:45, 11.60s/step, acc=0.4812, loss=2.1655, lr=7.35e-05]
Training:  38%|███▊      | 4192/11082 [13:28:18<21:57:49, 11.48s/step, acc=0.4812, loss=2.1655, lr=7.35e-05]
Training:  38%|███▊      | 4193/11082 [13:28:29<22:13:36, 11.62s/step, acc=0.4812, loss=2.1655, lr=7.35e-05]
Training:  38%|███▊      | 4194/11082 [13:28:41<21:57:47, 11.48s/step, acc=0.4812, loss=2.1655, lr=7.35e-05]
Training:  38%|███▊      | 4195/11082 [13:28:52<22:07:53, 11.57s/step, acc=0.4812, loss=2.1655, lr=7.35e-05]
Training:  38%|███▊      | 4196/11082 [13:29:04<22:10:22, 11.59s/step, acc=0.4812, loss=2.1655, lr=7.35e-05]
Training:  38%|███▊      | 4197/11082 [13:29:15<22:04:26, 11.54s/step, acc=0.4812, loss=2.1655, lr=7.35e-05]
Training:  38%|███▊      | 4198/11082 [13:29:27<21:55:47, 11.47s/step, acc=0.4812, loss=2.1655, lr=7.35e-05]
Training:  38%|███▊      | 4199/11082 [13:29:38<22:01:11, 11.52s/step, acc=0.4812, loss=2.1655, lr=7.35e-05]
Training:  38%|███▊      | 4200/11082 [13:29:50<22:09:38, 11.59s/step, acc=0.4812, loss=2.1655, lr=7.35e-05]
Training:  38%|███▊      | 4200/11082 [13:29:57<22:09:38, 11.59s/step, acc=0.4801, loss=2.1780, lr=7.29e-05]2026-06-19 04:23:25 | INFO | __main__ | epoch=0 step=4200 train_loss=2.178048 train_acc=0.480134 lr=7.293e-05 grad_norm=1.3047 acc_d0=0.7519 acc_d1=0.6017 acc_d2=0.4772 acc_d3=0.3769 acc_d4=0.2989 acc_d5=0.2398 acc_d6=0.1952 acc_d7=0.1618

Training:  38%|███▊      | 4201/11082 [13:30:02<22:07:33, 11.58s/step, acc=0.4801, loss=2.1780, lr=7.29e-05]
Training:  38%|███▊      | 4202/11082 [13:30:13<22:08:06, 11.58s/step, acc=0.4801, loss=2.1780, lr=7.29e-05]
Training:  38%|███▊      | 4203/11082 [13:30:25<22:04:07, 11.55s/step, acc=0.4801, loss=2.1780, lr=7.29e-05]
Training:  38%|███▊      | 4204/11082 [13:30:36<22:02:53, 11.54s/step, acc=0.4801, loss=2.1780, lr=7.29e-05]
Training:  38%|███▊      | 4205/11082 [13:30:48<22:10:27, 11.61s/step, acc=0.4801, loss=2.1780, lr=7.29e-05]
Training:  38%|███▊      | 4206/11082 [13:31:00<22:16:47, 11.66s/step, acc=0.4801, loss=2.1780, lr=7.29e-05]
Training:  38%|███▊      | 4207/11082 [13:31:11<22:04:04, 11.56s/step, acc=0.4801, loss=2.1780, lr=7.29e-05]
Training:  38%|███▊      | 4208/11082 [13:31:23<22:12:31, 11.63s/step, acc=0.4801, loss=2.1780, lr=7.29e-05]
Training:  38%|███▊      | 4209/11082 [13:31:34<21:54:29, 11.48s/step, acc=0.4801, loss=2.1780, lr=7.29e-05]
Training:  38%|███▊      | 4210/11082 [13:31:46<21:58:41, 11.51s/step, acc=0.4801, loss=2.1780, lr=7.29e-05]
Training:  38%|███▊      | 4211/11082 [13:31:57<21:55:39, 11.49s/step, acc=0.4801, loss=2.1780, lr=7.29e-05]
Training:  38%|███▊      | 4212/11082 [13:32:09<22:04:50, 11.57s/step, acc=0.4801, loss=2.1780, lr=7.29e-05]
Training:  38%|███▊      | 4213/11082 [13:32:20<21:50:46, 11.45s/step, acc=0.4801, loss=2.1780, lr=7.29e-05]
Training:  38%|███▊      | 4214/11082 [13:32:32<21:58:08, 11.52s/step, acc=0.4801, loss=2.1780, lr=7.29e-05]
Training:  38%|███▊      | 4215/11082 [13:32:43<22:03:20, 11.56s/step, acc=0.4801, loss=2.1780, lr=7.29e-05]
Training:  38%|███▊      | 4216/11082 [13:32:55<21:59:23, 11.53s/step, acc=0.4801, loss=2.1780, lr=7.29e-05]
Training:  38%|███▊      | 4217/11082 [13:33:06<22:01:02, 11.55s/step, acc=0.4801, loss=2.1780, lr=7.29e-05]
Training:  38%|███▊      | 4218/11082 [13:33:18<22:08:03, 11.61s/step, acc=0.4801, loss=2.1780, lr=7.29e-05]
Training:  38%|███▊      | 4219/11082 [13:33:29<21:55:57, 11.50s/step, acc=0.4801, loss=2.1780, lr=7.29e-05]
Training:  38%|███▊      | 4220/11082 [13:33:41<22:06:36, 11.60s/step, acc=0.4801, loss=2.1780, lr=7.29e-05]
Training:  38%|███▊      | 4221/11082 [13:33:53<21:59:36, 11.54s/step, acc=0.4801, loss=2.1780, lr=7.29e-05]
Training:  38%|███▊      | 4222/11082 [13:34:04<22:01:54, 11.56s/step, acc=0.4801, loss=2.1780, lr=7.29e-05]
Training:  38%|███▊      | 4223/11082 [13:34:16<22:00:07, 11.55s/step, acc=0.4801, loss=2.1780, lr=7.29e-05]
Training:  38%|███▊      | 4224/11082 [13:34:27<21:57:05, 11.52s/step, acc=0.4801, loss=2.1780, lr=7.29e-05]
Training:  38%|███▊      | 4225/11082 [13:34:38<21:47:06, 11.44s/step, acc=0.4801, loss=2.1780, lr=7.29e-05]
Training:  38%|███▊      | 4226/11082 [13:34:51<22:12:19, 11.66s/step, acc=0.4801, loss=2.1780, lr=7.29e-05]
Training:  38%|███▊      | 4227/11082 [13:35:02<22:04:30, 11.59s/step, acc=0.4801, loss=2.1780, lr=7.29e-05]
Training:  38%|███▊      | 4228/11082 [13:35:13<21:53:23, 11.50s/step, acc=0.4801, loss=2.1780, lr=7.29e-05]
Training:  38%|███▊      | 4229/11082 [13:35:25<22:06:03, 11.61s/step, acc=0.4801, loss=2.1780, lr=7.29e-05]
Training:  38%|███▊      | 4230/11082 [13:35:37<22:08:18, 11.63s/step, acc=0.4801, loss=2.1780, lr=7.29e-05]
Training:  38%|███▊      | 4231/11082 [13:35:48<21:44:21, 11.42s/step, acc=0.4801, loss=2.1780, lr=7.29e-05]
Training:  38%|███▊      | 4232/11082 [13:36:00<21:58:39, 11.55s/step, acc=0.4801, loss=2.1780, lr=7.29e-05]
Training:  38%|███▊      | 4233/11082 [13:36:11<21:42:55, 11.41s/step, acc=0.4801, loss=2.1780, lr=7.29e-05]
Training:  38%|███▊      | 4234/11082 [13:36:23<21:52:58, 11.50s/step, acc=0.4801, loss=2.1780, lr=7.29e-05]
Training:  38%|███▊      | 4235/11082 [13:36:34<22:04:33, 11.61s/step, acc=0.4801, loss=2.1780, lr=7.29e-05]
Training:  38%|███▊      | 4236/11082 [13:36:46<22:16:03, 11.71s/step, acc=0.4801, loss=2.1780, lr=7.29e-05]
Training:  38%|███▊      | 4237/11082 [13:36:57<21:44:36, 11.44s/step, acc=0.4801, loss=2.1780, lr=7.29e-05]
Training:  38%|███▊      | 4238/11082 [13:37:09<21:59:24, 11.57s/step, acc=0.4801, loss=2.1780, lr=7.29e-05]
Training:  38%|███▊      | 4239/11082 [13:37:20<21:51:07, 11.50s/step, acc=0.4801, loss=2.1780, lr=7.29e-05]
Training:  38%|███▊      | 4240/11082 [13:37:32<21:57:26, 11.55s/step, acc=0.4801, loss=2.1780, lr=7.29e-05]
Training:  38%|███▊      | 4241/11082 [13:37:44<21:56:02, 11.54s/step, acc=0.4801, loss=2.1780, lr=7.29e-05]
Training:  38%|███▊      | 4242/11082 [13:37:55<22:04:22, 11.62s/step, acc=0.4801, loss=2.1780, lr=7.29e-05]
Training:  38%|███▊      | 4243/11082 [13:38:07<21:57:40, 11.56s/step, acc=0.4801, loss=2.1780, lr=7.29e-05]
Training:  38%|███▊      | 4244/11082 [13:38:18<21:58:13, 11.57s/step, acc=0.4801, loss=2.1780, lr=7.29e-05]
Training:  38%|███▊      | 4245/11082 [13:38:30<21:51:52, 11.51s/step, acc=0.4801, loss=2.1780, lr=7.29e-05]
Training:  38%|███▊      | 4246/11082 [13:38:41<21:30:16, 11.32s/step, acc=0.4801, loss=2.1780, lr=7.29e-05]
Training:  38%|███▊      | 4247/11082 [13:38:53<22:12:17, 11.70s/step, acc=0.4801, loss=2.1780, lr=7.29e-05]
Training:  38%|███▊      | 4248/11082 [13:39:04<21:47:55, 11.48s/step, acc=0.4801, loss=2.1780, lr=7.29e-05]
Training:  38%|███▊      | 4249/11082 [13:39:16<21:51:45, 11.52s/step, acc=0.4801, loss=2.1780, lr=7.29e-05]
Training:  38%|███▊      | 4250/11082 [13:39:28<22:00:36, 11.60s/step, acc=0.4801, loss=2.1780, lr=7.29e-05]
Training:  38%|███▊      | 4250/11082 [13:39:34<22:00:36, 11.60s/step, acc=0.4820, loss=2.1608, lr=7.23e-05]2026-06-19 04:33:02 | INFO | __main__ | epoch=0 step=4250 train_loss=2.160774 train_acc=0.482018 lr=7.233e-05 grad_norm=0.8516 acc_d0=0.7503 acc_d1=0.6007 acc_d2=0.4744 acc_d3=0.3733 acc_d4=0.2969 acc_d5=0.2366 acc_d6=0.1916 acc_d7=0.1590

Training:  38%|███▊      | 4251/11082 [13:39:39<21:47:12, 11.48s/step, acc=0.4820, loss=2.1608, lr=7.23e-05]
Training:  38%|███▊      | 4252/11082 [13:39:50<21:52:25, 11.53s/step, acc=0.4820, loss=2.1608, lr=7.23e-05]
Training:  38%|███▊      | 4253/11082 [13:40:02<21:56:53, 11.57s/step, acc=0.4820, loss=2.1608, lr=7.23e-05]
Training:  38%|███▊      | 4254/11082 [13:40:14<21:57:45, 11.58s/step, acc=0.4820, loss=2.1608, lr=7.23e-05]
Training:  38%|███▊      | 4255/11082 [13:40:25<21:54:56, 11.56s/step, acc=0.4820, loss=2.1608, lr=7.23e-05]
Training:  38%|███▊      | 4256/11082 [13:40:37<22:04:12, 11.64s/step, acc=0.4820, loss=2.1608, lr=7.23e-05]
Training:  38%|███▊      | 4257/11082 [13:40:49<22:04:17, 11.64s/step, acc=0.4820, loss=2.1608, lr=7.23e-05]
Training:  38%|███▊      | 4258/11082 [13:41:00<22:01:14, 11.62s/step, acc=0.4820, loss=2.1608, lr=7.23e-05]
Training:  38%|███▊      | 4259/11082 [13:41:12<22:08:42, 11.68s/step, acc=0.4820, loss=2.1608, lr=7.23e-05]
Training:  38%|███▊      | 4260/11082 [13:41:24<22:18:45, 11.77s/step, acc=0.4820, loss=2.1608, lr=7.23e-05]
Training:  38%|███▊      | 4261/11082 [13:41:35<21:59:01, 11.60s/step, acc=0.4820, loss=2.1608, lr=7.23e-05]
Training:  38%|███▊      | 4262/11082 [13:41:47<22:00:12, 11.61s/step, acc=0.4820, loss=2.1608, lr=7.23e-05]
Training:  38%|███▊      | 4263/11082 [13:41:59<22:01:11, 11.63s/step, acc=0.4820, loss=2.1608, lr=7.23e-05]
Training:  38%|███▊      | 4264/11082 [13:42:10<22:04:54, 11.66s/step, acc=0.4820, loss=2.1608, lr=7.23e-05]
Training:  38%|███▊      | 4265/11082 [13:42:22<22:04:01, 11.65s/step, acc=0.4820, loss=2.1608, lr=7.23e-05]
Training:  38%|███▊      | 4266/11082 [13:42:33<21:44:26, 11.48s/step, acc=0.4820, loss=2.1608, lr=7.23e-05]
Training:  39%|███▊      | 4267/11082 [13:42:45<21:55:37, 11.58s/step, acc=0.4820, loss=2.1608, lr=7.23e-05]
Training:  39%|███▊      | 4268/11082 [13:42:56<21:49:13, 11.53s/step, acc=0.4820, loss=2.1608, lr=7.23e-05]
Training:  39%|███▊      | 4269/11082 [13:43:08<22:05:27, 11.67s/step, acc=0.4820, loss=2.1608, lr=7.23e-05]
Training:  39%|███▊      | 4270/11082 [13:43:20<21:52:46, 11.56s/step, acc=0.4820, loss=2.1608, lr=7.23e-05]
Training:  39%|███▊      | 4271/11082 [13:43:31<21:55:12, 11.59s/step, acc=0.4820, loss=2.1608, lr=7.23e-05]
Training:  39%|███▊      | 4272/11082 [13:43:43<21:54:22, 11.58s/step, acc=0.4820, loss=2.1608, lr=7.23e-05]
Training:  39%|███▊      | 4273/11082 [13:43:54<21:44:17, 11.49s/step, acc=0.4820, loss=2.1608, lr=7.23e-05]
Training:  39%|███▊      | 4274/11082 [13:44:06<21:48:32, 11.53s/step, acc=0.4820, loss=2.1608, lr=7.23e-05]
Training:  39%|███▊      | 4275/11082 [13:44:17<21:44:41, 11.50s/step, acc=0.4820, loss=2.1608, lr=7.23e-05]
Training:  39%|███▊      | 4276/11082 [13:44:29<21:51:23, 11.56s/step, acc=0.4820, loss=2.1608, lr=7.23e-05]
Training:  39%|███▊      | 4277/11082 [13:44:40<21:53:56, 11.59s/step, acc=0.4820, loss=2.1608, lr=7.23e-05]
Training:  39%|███▊      | 4278/11082 [13:44:52<21:46:53, 11.52s/step, acc=0.4820, loss=2.1608, lr=7.23e-05]
Training:  39%|███▊      | 4279/11082 [13:45:03<21:43:53, 11.50s/step, acc=0.4820, loss=2.1608, lr=7.23e-05]
Training:  39%|███▊      | 4280/11082 [13:45:15<21:56:38, 11.61s/step, acc=0.4820, loss=2.1608, lr=7.23e-05]
Training:  39%|███▊      | 4281/11082 [13:45:27<21:52:47, 11.58s/step, acc=0.4820, loss=2.1608, lr=7.23e-05]
Training:  39%|███▊      | 4282/11082 [13:45:38<21:48:50, 11.55s/step, acc=0.4820, loss=2.1608, lr=7.23e-05]
Training:  39%|███▊      | 4283/11082 [13:45:50<21:46:50, 11.53s/step, acc=0.4820, loss=2.1608, lr=7.23e-05]
Training:  39%|███▊      | 4284/11082 [13:46:01<21:50:46, 11.57s/step, acc=0.4820, loss=2.1608, lr=7.23e-05]
Training:  39%|███▊      | 4285/11082 [13:46:13<22:00:02, 11.65s/step, acc=0.4820, loss=2.1608, lr=7.23e-05]
Training:  39%|███▊      | 4286/11082 [13:46:25<22:06:12, 11.71s/step, acc=0.4820, loss=2.1608, lr=7.23e-05]
Training:  39%|███▊      | 4287/11082 [13:46:36<21:57:56, 11.64s/step, acc=0.4820, loss=2.1608, lr=7.23e-05]
Training:  39%|███▊      | 4288/11082 [13:46:48<21:46:23, 11.54s/step, acc=0.4820, loss=2.1608, lr=7.23e-05]
Training:  39%|███▊      | 4289/11082 [13:46:59<21:43:04, 11.51s/step, acc=0.4820, loss=2.1608, lr=7.23e-05]
Training:  39%|███▊      | 4290/11082 [13:47:11<21:42:59, 11.51s/step, acc=0.4820, loss=2.1608, lr=7.23e-05]
Training:  39%|███▊      | 4291/11082 [13:47:22<21:49:26, 11.57s/step, acc=0.4820, loss=2.1608, lr=7.23e-05]
Training:  39%|███▊      | 4292/11082 [13:47:34<21:50:19, 11.58s/step, acc=0.4820, loss=2.1608, lr=7.23e-05]
Training:  39%|███▊      | 4293/11082 [13:47:45<21:32:38, 11.42s/step, acc=0.4820, loss=2.1608, lr=7.23e-05]
Training:  39%|███▊      | 4294/11082 [13:47:57<21:45:16, 11.54s/step, acc=0.4820, loss=2.1608, lr=7.23e-05]
Training:  39%|███▉      | 4295/11082 [13:48:09<21:51:54, 11.60s/step, acc=0.4820, loss=2.1608, lr=7.23e-05]
Training:  39%|███▉      | 4296/11082 [13:48:20<21:45:05, 11.54s/step, acc=0.4820, loss=2.1608, lr=7.23e-05]
Training:  39%|███▉      | 4297/11082 [13:48:32<21:49:37, 11.58s/step, acc=0.4820, loss=2.1608, lr=7.23e-05]
Training:  39%|███▉      | 4298/11082 [13:48:43<21:42:25, 11.52s/step, acc=0.4820, loss=2.1608, lr=7.23e-05]
Training:  39%|███▉      | 4299/11082 [13:48:55<21:56:08, 11.64s/step, acc=0.4820, loss=2.1608, lr=7.23e-05]
Training:  39%|███▉      | 4300/11082 [13:49:06<21:39:10, 11.49s/step, acc=0.4820, loss=2.1608, lr=7.23e-05]
Training:  39%|███▉      | 4300/11082 [13:49:13<21:39:10, 11.49s/step, acc=0.4832, loss=2.1570, lr=7.17e-05]2026-06-19 04:42:41 | INFO | __main__ | epoch=0 step=4300 train_loss=2.157027 train_acc=0.483171 lr=7.174e-05 grad_norm=1.4062 acc_d0=0.7523 acc_d1=0.6026 acc_d2=0.4784 acc_d3=0.3772 acc_d4=0.2994 acc_d5=0.2392 acc_d6=0.1945 acc_d7=0.1610

Training:  39%|███▉      | 4301/11082 [13:49:18<21:40:42, 11.51s/step, acc=0.4832, loss=2.1570, lr=7.17e-05]
Training:  39%|███▉      | 4302/11082 [13:49:29<21:39:40, 11.50s/step, acc=0.4832, loss=2.1570, lr=7.17e-05]
Training:  39%|███▉      | 4303/11082 [13:49:41<21:40:47, 11.51s/step, acc=0.4832, loss=2.1570, lr=7.17e-05]
Training:  39%|███▉      | 4304/11082 [13:49:53<21:57:16, 11.66s/step, acc=0.4832, loss=2.1570, lr=7.17e-05]
Training:  39%|███▉      | 4305/11082 [13:50:03<21:28:09, 11.40s/step, acc=0.4832, loss=2.1570, lr=7.17e-05]
Training:  39%|███▉      | 4306/11082 [13:50:15<21:37:55, 11.49s/step, acc=0.4832, loss=2.1570, lr=7.17e-05]
Training:  39%|███▉      | 4307/11082 [13:50:27<21:35:24, 11.47s/step, acc=0.4832, loss=2.1570, lr=7.17e-05]
Training:  39%|███▉      | 4308/11082 [13:50:38<21:38:34, 11.50s/step, acc=0.4832, loss=2.1570, lr=7.17e-05]
Training:  39%|███▉      | 4309/11082 [13:50:50<21:37:40, 11.50s/step, acc=0.4832, loss=2.1570, lr=7.17e-05]
Training:  39%|███▉      | 4310/11082 [13:51:01<21:29:44, 11.43s/step, acc=0.4832, loss=2.1570, lr=7.17e-05]
Training:  39%|███▉      | 4311/11082 [13:51:13<21:44:39, 11.56s/step, acc=0.4832, loss=2.1570, lr=7.17e-05]
Training:  39%|███▉      | 4312/11082 [13:51:25<22:03:20, 11.73s/step, acc=0.4832, loss=2.1570, lr=7.17e-05]
Training:  39%|███▉      | 4313/11082 [13:51:36<21:50:45, 11.62s/step, acc=0.4832, loss=2.1570, lr=7.17e-05]
Training:  39%|███▉      | 4314/11082 [13:51:47<21:36:26, 11.49s/step, acc=0.4832, loss=2.1570, lr=7.17e-05]
Training:  39%|███▉      | 4315/11082 [13:51:59<21:48:05, 11.60s/step, acc=0.4832, loss=2.1570, lr=7.17e-05]
Training:  39%|███▉      | 4316/11082 [13:52:11<21:47:53, 11.60s/step, acc=0.4832, loss=2.1570, lr=7.17e-05]
Training:  39%|███▉      | 4317/11082 [13:52:22<21:25:28, 11.40s/step, acc=0.4832, loss=2.1570, lr=7.17e-05]
Training:  39%|███▉      | 4318/11082 [13:52:34<21:46:18, 11.59s/step, acc=0.4832, loss=2.1570, lr=7.17e-05]
Training:  39%|███▉      | 4319/11082 [13:52:45<21:42:03, 11.55s/step, acc=0.4832, loss=2.1570, lr=7.17e-05]
Training:  39%|███▉      | 4320/11082 [13:52:57<21:43:26, 11.57s/step, acc=0.4832, loss=2.1570, lr=7.17e-05]
Training:  39%|███▉      | 4321/11082 [13:53:08<21:34:47, 11.49s/step, acc=0.4832, loss=2.1570, lr=7.17e-05]
Training:  39%|███▉      | 4322/11082 [13:53:20<21:38:51, 11.53s/step, acc=0.4832, loss=2.1570, lr=7.17e-05]
Training:  39%|███▉      | 4323/11082 [13:53:31<21:38:01, 11.52s/step, acc=0.4832, loss=2.1570, lr=7.17e-05]
Training:  39%|███▉      | 4324/11082 [13:53:43<21:26:24, 11.42s/step, acc=0.4832, loss=2.1570, lr=7.17e-05]
Training:  39%|███▉      | 4325/11082 [13:53:54<21:33:34, 11.49s/step, acc=0.4832, loss=2.1570, lr=7.17e-05]
Training:  39%|███▉      | 4326/11082 [13:54:06<21:27:57, 11.44s/step, acc=0.4832, loss=2.1570, lr=7.17e-05]
Training:  39%|███▉      | 4327/11082 [13:54:17<21:31:25, 11.47s/step, acc=0.4832, loss=2.1570, lr=7.17e-05]
Training:  39%|███▉      | 4328/11082 [13:54:28<21:23:12, 11.40s/step, acc=0.4832, loss=2.1570, lr=7.17e-05]
Training:  39%|███▉      | 4329/11082 [13:54:40<21:44:04, 11.59s/step, acc=0.4832, loss=2.1570, lr=7.17e-05]
Training:  39%|███▉      | 4330/11082 [13:54:52<21:31:29, 11.48s/step, acc=0.4832, loss=2.1570, lr=7.17e-05]
Training:  39%|███▉      | 4331/11082 [13:55:03<21:28:51, 11.45s/step, acc=0.4832, loss=2.1570, lr=7.17e-05]
Training:  39%|███▉      | 4332/11082 [13:55:15<21:46:33, 11.61s/step, acc=0.4832, loss=2.1570, lr=7.17e-05]
Training:  39%|███▉      | 4333/11082 [13:55:26<21:38:03, 11.54s/step, acc=0.4832, loss=2.1570, lr=7.17e-05]
Training:  39%|███▉      | 4334/11082 [13:55:38<21:38:03, 11.54s/step, acc=0.4832, loss=2.1570, lr=7.17e-05]
Training:  39%|███▉      | 4335/11082 [13:55:49<21:32:04, 11.49s/step, acc=0.4832, loss=2.1570, lr=7.17e-05]
Training:  39%|███▉      | 4336/11082 [13:56:01<21:33:39, 11.51s/step, acc=0.4832, loss=2.1570, lr=7.17e-05]
Training:  39%|███▉      | 4337/11082 [13:56:12<21:30:06, 11.48s/step, acc=0.4832, loss=2.1570, lr=7.17e-05]
Training:  39%|███▉      | 4338/11082 [13:56:24<21:44:58, 11.61s/step, acc=0.4832, loss=2.1570, lr=7.17e-05]
Training:  39%|███▉      | 4339/11082 [13:56:36<21:40:08, 11.57s/step, acc=0.4832, loss=2.1570, lr=7.17e-05]
Training:  39%|███▉      | 4340/11082 [13:56:47<21:31:23, 11.49s/step, acc=0.4832, loss=2.1570, lr=7.17e-05]
Training:  39%|███▉      | 4341/11082 [13:56:58<21:33:21, 11.51s/step, acc=0.4832, loss=2.1570, lr=7.17e-05]
Training:  39%|███▉      | 4342/11082 [13:57:10<21:40:17, 11.58s/step, acc=0.4832, loss=2.1570, lr=7.17e-05]
Training:  39%|███▉      | 4343/11082 [13:57:22<21:43:47, 11.61s/step, acc=0.4832, loss=2.1570, lr=7.17e-05]
Training:  39%|███▉      | 4344/11082 [13:57:33<21:39:56, 11.58s/step, acc=0.4832, loss=2.1570, lr=7.17e-05]
Training:  39%|███▉      | 4345/11082 [13:57:45<21:55:37, 11.72s/step, acc=0.4832, loss=2.1570, lr=7.17e-05]
Training:  39%|███▉      | 4346/11082 [13:57:57<21:48:38, 11.66s/step, acc=0.4832, loss=2.1570, lr=7.17e-05]
Training:  39%|███▉      | 4347/11082 [13:58:08<21:36:13, 11.55s/step, acc=0.4832, loss=2.1570, lr=7.17e-05]
Training:  39%|███▉      | 4348/11082 [13:58:20<21:38:35, 11.57s/step, acc=0.4832, loss=2.1570, lr=7.17e-05]
Training:  39%|███▉      | 4349/11082 [13:58:31<21:38:45, 11.57s/step, acc=0.4832, loss=2.1570, lr=7.17e-05]
Training:  39%|███▉      | 4350/11082 [13:58:43<21:48:28, 11.66s/step, acc=0.4832, loss=2.1570, lr=7.17e-05]
Training:  39%|███▉      | 4350/11082 [13:58:51<21:48:28, 11.66s/step, acc=0.4837, loss=2.1535, lr=7.11e-05]2026-06-19 04:52:19 | INFO | __main__ | epoch=0 step=4350 train_loss=2.153488 train_acc=0.483705 lr=7.113e-05 grad_norm=0.8398 acc_d0=0.7552 acc_d1=0.6044 acc_d2=0.4783 acc_d3=0.3757 acc_d4=0.2982 acc_d5=0.2360 acc_d6=0.1915 acc_d7=0.1575

Training:  39%|███▉      | 4351/11082 [13:58:55<21:59:52, 11.77s/step, acc=0.4837, loss=2.1535, lr=7.11e-05]
Training:  39%|███▉      | 4352/11082 [13:59:07<21:48:02, 11.66s/step, acc=0.4837, loss=2.1535, lr=7.11e-05]
Training:  39%|███▉      | 4353/11082 [13:59:18<21:45:06, 11.64s/step, acc=0.4837, loss=2.1535, lr=7.11e-05]
Training:  39%|███▉      | 4354/11082 [13:59:30<21:49:50, 11.68s/step, acc=0.4837, loss=2.1535, lr=7.11e-05]
Training:  39%|███▉      | 4355/11082 [13:59:41<21:27:20, 11.48s/step, acc=0.4837, loss=2.1535, lr=7.11e-05]
Training:  39%|███▉      | 4356/11082 [13:59:53<21:35:17, 11.55s/step, acc=0.4837, loss=2.1535, lr=7.11e-05]
Training:  39%|███▉      | 4357/11082 [14:00:05<21:42:35, 11.62s/step, acc=0.4837, loss=2.1535, lr=7.11e-05]
Training:  39%|███▉      | 4358/11082 [14:00:16<21:29:38, 11.51s/step, acc=0.4837, loss=2.1535, lr=7.11e-05]
Training:  39%|███▉      | 4359/11082 [14:00:27<21:20:22, 11.43s/step, acc=0.4837, loss=2.1535, lr=7.11e-05]
Training:  39%|███▉      | 4360/11082 [14:00:39<21:34:46, 11.56s/step, acc=0.4837, loss=2.1535, lr=7.11e-05]
Training:  39%|███▉      | 4361/11082 [14:00:51<21:49:05, 11.69s/step, acc=0.4837, loss=2.1535, lr=7.11e-05]
Training:  39%|███▉      | 4362/11082 [14:01:02<21:32:43, 11.54s/step, acc=0.4837, loss=2.1535, lr=7.11e-05]
Training:  39%|███▉      | 4363/11082 [14:01:14<21:46:46, 11.67s/step, acc=0.4837, loss=2.1535, lr=7.11e-05]
Training:  39%|███▉      | 4364/11082 [14:01:25<21:31:15, 11.53s/step, acc=0.4837, loss=2.1535, lr=7.11e-05]
Training:  39%|███▉      | 4365/11082 [14:01:36<21:13:20, 11.37s/step, acc=0.4837, loss=2.1535, lr=7.11e-05]
Training:  39%|███▉      | 4366/11082 [14:01:48<21:40:16, 11.62s/step, acc=0.4837, loss=2.1535, lr=7.11e-05]
Training:  39%|███▉      | 4367/11082 [14:02:00<21:38:42, 11.60s/step, acc=0.4837, loss=2.1535, lr=7.11e-05]
Training:  39%|███▉      | 4368/11082 [14:02:12<21:46:31, 11.68s/step, acc=0.4837, loss=2.1535, lr=7.11e-05]
Training:  39%|███▉      | 4369/11082 [14:02:24<21:46:53, 11.68s/step, acc=0.4837, loss=2.1535, lr=7.11e-05]
Training:  39%|███▉      | 4370/11082 [14:02:35<21:31:06, 11.54s/step, acc=0.4837, loss=2.1535, lr=7.11e-05]
Training:  39%|███▉      | 4371/11082 [14:02:46<21:35:29, 11.58s/step, acc=0.4837, loss=2.1535, lr=7.11e-05]
Training:  39%|███▉      | 4372/11082 [14:02:58<21:36:17, 11.59s/step, acc=0.4837, loss=2.1535, lr=7.11e-05]
Training:  39%|███▉      | 4373/11082 [14:03:10<21:29:38, 11.53s/step, acc=0.4837, loss=2.1535, lr=7.11e-05]
Training:  39%|███▉      | 4374/11082 [14:03:21<21:20:48, 11.46s/step, acc=0.4837, loss=2.1535, lr=7.11e-05]
Training:  39%|███▉      | 4375/11082 [14:03:32<21:27:33, 11.52s/step, acc=0.4837, loss=2.1535, lr=7.11e-05]
Training:  39%|███▉      | 4376/11082 [14:03:44<21:26:36, 11.51s/step, acc=0.4837, loss=2.1535, lr=7.11e-05]
Training:  39%|███▉      | 4377/11082 [14:03:55<21:19:03, 11.45s/step, acc=0.4837, loss=2.1535, lr=7.11e-05]
Training:  40%|███▉      | 4378/11082 [14:04:07<21:29:00, 11.54s/step, acc=0.4837, loss=2.1535, lr=7.11e-05]
Training:  40%|███▉      | 4379/11082 [14:04:19<21:41:20, 11.65s/step, acc=0.4837, loss=2.1535, lr=7.11e-05]
Training:  40%|███▉      | 4380/11082 [14:04:30<21:31:08, 11.56s/step, acc=0.4837, loss=2.1535, lr=7.11e-05]
Training:  40%|███▉      | 4381/11082 [14:04:42<21:41:01, 11.65s/step, acc=0.4837, loss=2.1535, lr=7.11e-05]
Training:  40%|███▉      | 4382/11082 [14:04:53<21:30:09, 11.55s/step, acc=0.4837, loss=2.1535, lr=7.11e-05]
Training:  40%|███▉      | 4383/11082 [14:05:05<21:26:16, 11.52s/step, acc=0.4837, loss=2.1535, lr=7.11e-05]
Training:  40%|███▉      | 4384/11082 [14:05:17<21:29:48, 11.55s/step, acc=0.4837, loss=2.1535, lr=7.11e-05]
Training:  40%|███▉      | 4385/11082 [14:05:28<21:21:28, 11.48s/step, acc=0.4837, loss=2.1535, lr=7.11e-05]
Training:  40%|███▉      | 4386/11082 [14:05:39<21:27:18, 11.54s/step, acc=0.4837, loss=2.1535, lr=7.11e-05]
Training:  40%|███▉      | 4387/11082 [14:05:51<21:15:16, 11.43s/step, acc=0.4837, loss=2.1535, lr=7.11e-05]
Training:  40%|███▉      | 4388/11082 [14:06:03<21:39:50, 11.65s/step, acc=0.4837, loss=2.1535, lr=7.11e-05]
Training:  40%|███▉      | 4389/11082 [14:06:14<21:33:56, 11.60s/step, acc=0.4837, loss=2.1535, lr=7.11e-05]
Training:  40%|███▉      | 4390/11082 [14:06:25<21:18:54, 11.47s/step, acc=0.4837, loss=2.1535, lr=7.11e-05]
Training:  40%|███▉      | 4391/11082 [14:06:37<21:07:13, 11.36s/step, acc=0.4837, loss=2.1535, lr=7.11e-05]
Training:  40%|███▉      | 4392/11082 [14:06:49<21:30:30, 11.57s/step, acc=0.4837, loss=2.1535, lr=7.11e-05]
Training:  40%|███▉      | 4393/11082 [14:07:00<21:32:39, 11.60s/step, acc=0.4837, loss=2.1535, lr=7.11e-05]
Training:  40%|███▉      | 4394/11082 [14:07:12<21:24:30, 11.52s/step, acc=0.4837, loss=2.1535, lr=7.11e-05]
Training:  40%|███▉      | 4395/11082 [14:07:23<21:31:47, 11.59s/step, acc=0.4837, loss=2.1535, lr=7.11e-05]
Training:  40%|███▉      | 4396/11082 [14:07:35<21:40:03, 11.67s/step, acc=0.4837, loss=2.1535, lr=7.11e-05]
Training:  40%|███▉      | 4397/11082 [14:07:46<21:25:35, 11.54s/step, acc=0.4837, loss=2.1535, lr=7.11e-05]
Training:  40%|███▉      | 4398/11082 [14:07:58<21:26:32, 11.55s/step, acc=0.4837, loss=2.1535, lr=7.11e-05]
Training:  40%|███▉      | 4399/11082 [14:08:10<21:28:34, 11.57s/step, acc=0.4837, loss=2.1535, lr=7.11e-05]
Training:  40%|███▉      | 4400/11082 [14:08:21<21:27:57, 11.57s/step, acc=0.4837, loss=2.1535, lr=7.11e-05]
Training:  40%|███▉      | 4400/11082 [14:08:28<21:27:57, 11.57s/step, acc=0.4837, loss=2.1526, lr=7.05e-05]2026-06-19 05:01:56 | INFO | __main__ | epoch=0 step=4400 train_loss=2.152599 train_acc=0.483703 lr=7.053e-05 grad_norm=0.8867 acc_d0=0.7480 acc_d1=0.5974 acc_d2=0.4728 acc_d3=0.3715 acc_d4=0.2936 acc_d5=0.2346 acc_d6=0.1913 acc_d7=0.1586

Training:  40%|███▉      | 4401/11082 [14:08:33<21:19:56, 11.49s/step, acc=0.4837, loss=2.1526, lr=7.05e-05]
Training:  40%|███▉      | 4402/11082 [14:08:44<21:23:57, 11.53s/step, acc=0.4837, loss=2.1526, lr=7.05e-05]
Training:  40%|███▉      | 4403/11082 [14:08:55<21:14:29, 11.45s/step, acc=0.4837, loss=2.1526, lr=7.05e-05]
Training:  40%|███▉      | 4404/11082 [14:09:07<21:11:20, 11.42s/step, acc=0.4837, loss=2.1526, lr=7.05e-05]
Training:  40%|███▉      | 4405/11082 [14:09:18<21:20:24, 11.51s/step, acc=0.4837, loss=2.1526, lr=7.05e-05]
Training:  40%|███▉      | 4406/11082 [14:09:31<21:43:24, 11.71s/step, acc=0.4837, loss=2.1526, lr=7.05e-05]
Training:  40%|███▉      | 4407/11082 [14:09:42<21:37:55, 11.67s/step, acc=0.4837, loss=2.1526, lr=7.05e-05]
Training:  40%|███▉      | 4408/11082 [14:09:54<21:28:51, 11.59s/step, acc=0.4837, loss=2.1526, lr=7.05e-05]
Training:  40%|███▉      | 4409/11082 [14:10:05<21:15:42, 11.47s/step, acc=0.4837, loss=2.1526, lr=7.05e-05]
Training:  40%|███▉      | 4410/11082 [14:10:17<21:36:05, 11.66s/step, acc=0.4837, loss=2.1526, lr=7.05e-05]
Training:  40%|███▉      | 4411/11082 [14:10:28<21:26:50, 11.57s/step, acc=0.4837, loss=2.1526, lr=7.05e-05]
Training:  40%|███▉      | 4412/11082 [14:10:40<21:23:29, 11.55s/step, acc=0.4837, loss=2.1526, lr=7.05e-05]
Training:  40%|███▉      | 4413/11082 [14:10:51<21:18:25, 11.50s/step, acc=0.4837, loss=2.1526, lr=7.05e-05]
Training:  40%|███▉      | 4414/11082 [14:11:03<21:38:38, 11.69s/step, acc=0.4837, loss=2.1526, lr=7.05e-05]
Training:  40%|███▉      | 4415/11082 [14:11:15<21:33:11, 11.64s/step, acc=0.4837, loss=2.1526, lr=7.05e-05]
Training:  40%|███▉      | 4416/11082 [14:11:26<21:28:10, 11.59s/step, acc=0.4837, loss=2.1526, lr=7.05e-05]
Training:  40%|███▉      | 4417/11082 [14:11:38<21:25:19, 11.57s/step, acc=0.4837, loss=2.1526, lr=7.05e-05]
Training:  40%|███▉      | 4418/11082 [14:11:49<21:26:17, 11.58s/step, acc=0.4837, loss=2.1526, lr=7.05e-05]
Training:  40%|███▉      | 4419/11082 [14:12:01<21:21:30, 11.54s/step, acc=0.4837, loss=2.1526, lr=7.05e-05]
Training:  40%|███▉      | 4420/11082 [14:12:13<21:42:22, 11.73s/step, acc=0.4837, loss=2.1526, lr=7.05e-05]
Training:  40%|███▉      | 4421/11082 [14:12:24<21:28:50, 11.61s/step, acc=0.4837, loss=2.1526, lr=7.05e-05]
Training:  40%|███▉      | 4422/11082 [14:12:35<21:10:37, 11.45s/step, acc=0.4837, loss=2.1526, lr=7.05e-05]
Training:  40%|███▉      | 4423/11082 [14:12:48<21:39:55, 11.71s/step, acc=0.4837, loss=2.1526, lr=7.05e-05]
Training:  40%|███▉      | 4424/11082 [14:12:59<21:31:31, 11.64s/step, acc=0.4837, loss=2.1526, lr=7.05e-05]
Training:  40%|███▉      | 4425/11082 [14:13:11<21:44:08, 11.75s/step, acc=0.4837, loss=2.1526, lr=7.05e-05]
Training:  40%|███▉      | 4426/11082 [14:13:23<21:44:22, 11.76s/step, acc=0.4837, loss=2.1526, lr=7.05e-05]
Training:  40%|███▉      | 4427/11082 [14:13:34<21:30:24, 11.63s/step, acc=0.4837, loss=2.1526, lr=7.05e-05]
Training:  40%|███▉      | 4428/11082 [14:13:46<21:38:35, 11.71s/step, acc=0.4837, loss=2.1526, lr=7.05e-05]
Training:  40%|███▉      | 4429/11082 [14:13:58<21:25:23, 11.59s/step, acc=0.4837, loss=2.1526, lr=7.05e-05]
Training:  40%|███▉      | 4430/11082 [14:14:09<21:23:29, 11.58s/step, acc=0.4837, loss=2.1526, lr=7.05e-05]
Training:  40%|███▉      | 4431/11082 [14:14:21<21:29:29, 11.63s/step, acc=0.4837, loss=2.1526, lr=7.05e-05]
Training:  40%|███▉      | 4432/11082 [14:14:32<21:21:23, 11.56s/step, acc=0.4837, loss=2.1526, lr=7.05e-05]
Training:  40%|████      | 4433/11082 [14:14:44<21:21:23, 11.56s/step, acc=0.4837, loss=2.1526, lr=7.05e-05]
Training:  40%|████      | 4434/11082 [14:14:56<21:24:44, 11.60s/step, acc=0.4837, loss=2.1526, lr=7.05e-05]
Training:  40%|████      | 4435/11082 [14:15:07<21:29:54, 11.64s/step, acc=0.4837, loss=2.1526, lr=7.05e-05]
Training:  40%|████      | 4436/11082 [14:15:19<21:24:39, 11.60s/step, acc=0.4837, loss=2.1526, lr=7.05e-05]
Training:  40%|████      | 4437/11082 [14:15:30<21:27:45, 11.63s/step, acc=0.4837, loss=2.1526, lr=7.05e-05]
Training:  40%|████      | 4438/11082 [14:15:41<20:56:11, 11.34s/step, acc=0.4837, loss=2.1526, lr=7.05e-05]
Training:  40%|████      | 4439/11082 [14:15:53<21:20:50, 11.57s/step, acc=0.4837, loss=2.1526, lr=7.05e-05]
Training:  40%|████      | 4440/11082 [14:16:05<21:14:46, 11.52s/step, acc=0.4837, loss=2.1526, lr=7.05e-05]
Training:  40%|████      | 4441/11082 [14:16:16<21:10:15, 11.48s/step, acc=0.4837, loss=2.1526, lr=7.05e-05]
Training:  40%|████      | 4442/11082 [14:16:27<20:55:43, 11.35s/step, acc=0.4837, loss=2.1526, lr=7.05e-05]
Training:  40%|████      | 4443/11082 [14:16:40<21:32:34, 11.68s/step, acc=0.4837, loss=2.1526, lr=7.05e-05]
Training:  40%|████      | 4444/11082 [14:16:51<21:21:27, 11.58s/step, acc=0.4837, loss=2.1526, lr=7.05e-05]
Training:  40%|████      | 4445/11082 [14:17:02<21:13:04, 11.51s/step, acc=0.4837, loss=2.1526, lr=7.05e-05]
Training:  40%|████      | 4446/11082 [14:17:13<21:02:13, 11.41s/step, acc=0.4837, loss=2.1526, lr=7.05e-05]
Training:  40%|████      | 4447/11082 [14:17:26<21:33:43, 11.70s/step, acc=0.4837, loss=2.1526, lr=7.05e-05]
Training:  40%|████      | 4448/11082 [14:17:37<21:24:03, 11.61s/step, acc=0.4837, loss=2.1526, lr=7.05e-05]
Training:  40%|████      | 4449/11082 [14:17:49<21:29:57, 11.67s/step, acc=0.4837, loss=2.1526, lr=7.05e-05]
Training:  40%|████      | 4450/11082 [14:18:00<20:59:07, 11.39s/step, acc=0.4837, loss=2.1526, lr=7.05e-05]
Training:  40%|████      | 4450/11082 [14:18:07<20:59:07, 11.39s/step, acc=0.4851, loss=2.1447, lr=6.99e-05]2026-06-19 05:11:35 | INFO | __main__ | epoch=0 step=4450 train_loss=2.144738 train_acc=0.485125 lr=6.992e-05 grad_norm=1.1094 acc_d0=0.7495 acc_d1=0.5981 acc_d2=0.4736 acc_d3=0.3718 acc_d4=0.2944 acc_d5=0.2351 acc_d6=0.1893 acc_d7=0.1585

Training:  40%|████      | 4451/11082 [14:18:12<21:23:54, 11.62s/step, acc=0.4851, loss=2.1447, lr=6.99e-05]
Training:  40%|████      | 4452/11082 [14:18:23<21:21:34, 11.60s/step, acc=0.4851, loss=2.1447, lr=6.99e-05]
Training:  40%|████      | 4453/11082 [14:18:35<21:11:33, 11.51s/step, acc=0.4851, loss=2.1447, lr=6.99e-05]
Training:  40%|████      | 4454/11082 [14:18:47<21:21:42, 11.60s/step, acc=0.4851, loss=2.1447, lr=6.99e-05]
Training:  40%|████      | 4455/11082 [14:18:58<21:21:30, 11.60s/step, acc=0.4851, loss=2.1447, lr=6.99e-05]
Training:  40%|████      | 4456/11082 [14:19:09<21:12:02, 11.52s/step, acc=0.4851, loss=2.1447, lr=6.99e-05]
Training:  40%|████      | 4457/11082 [14:19:21<21:25:19, 11.64s/step, acc=0.4851, loss=2.1447, lr=6.99e-05]
Training:  40%|████      | 4458/11082 [14:19:33<21:09:26, 11.50s/step, acc=0.4851, loss=2.1447, lr=6.99e-05]
Training:  40%|████      | 4459/11082 [14:19:44<21:16:55, 11.57s/step, acc=0.4851, loss=2.1447, lr=6.99e-05]
Training:  40%|████      | 4460/11082 [14:19:56<21:13:03, 11.53s/step, acc=0.4851, loss=2.1447, lr=6.99e-05]
Training:  40%|████      | 4461/11082 [14:20:07<21:09:41, 11.51s/step, acc=0.4851, loss=2.1447, lr=6.99e-05]
Training:  40%|████      | 4462/11082 [14:20:19<21:18:28, 11.59s/step, acc=0.4851, loss=2.1447, lr=6.99e-05]
Training:  40%|████      | 4463/11082 [14:20:31<21:25:05, 11.65s/step, acc=0.4851, loss=2.1447, lr=6.99e-05]
Training:  40%|████      | 4464/11082 [14:20:42<21:11:40, 11.53s/step, acc=0.4851, loss=2.1447, lr=6.99e-05]
Training:  40%|████      | 4465/11082 [14:20:53<21:01:01, 11.43s/step, acc=0.4851, loss=2.1447, lr=6.99e-05]
Training:  40%|████      | 4466/11082 [14:21:05<21:19:05, 11.60s/step, acc=0.4851, loss=2.1447, lr=6.99e-05]
Training:  40%|████      | 4467/11082 [14:21:16<21:06:30, 11.49s/step, acc=0.4851, loss=2.1447, lr=6.99e-05]
Training:  40%|████      | 4468/11082 [14:21:28<21:20:06, 11.61s/step, acc=0.4851, loss=2.1447, lr=6.99e-05]
Training:  40%|████      | 4469/11082 [14:21:40<21:08:05, 11.51s/step, acc=0.4851, loss=2.1447, lr=6.99e-05]
Training:  40%|████      | 4470/11082 [14:21:51<21:17:46, 11.59s/step, acc=0.4851, loss=2.1447, lr=6.99e-05]
Training:  40%|████      | 4471/11082 [14:22:03<21:31:23, 11.72s/step, acc=0.4851, loss=2.1447, lr=6.99e-05]
Training:  40%|████      | 4472/11082 [14:22:15<21:18:10, 11.60s/step, acc=0.4851, loss=2.1447, lr=6.99e-05]
Training:  40%|████      | 4473/11082 [14:22:26<21:15:37, 11.58s/step, acc=0.4851, loss=2.1447, lr=6.99e-05]
Training:  40%|████      | 4474/11082 [14:22:38<21:15:46, 11.58s/step, acc=0.4851, loss=2.1447, lr=6.99e-05]
Training:  40%|████      | 4475/11082 [14:22:49<21:05:05, 11.49s/step, acc=0.4851, loss=2.1447, lr=6.99e-05]
Training:  40%|████      | 4476/11082 [14:23:01<21:08:23, 11.52s/step, acc=0.4851, loss=2.1447, lr=6.99e-05]
Training:  40%|████      | 4477/11082 [14:23:12<21:10:48, 11.54s/step, acc=0.4851, loss=2.1447, lr=6.99e-05]
Training:  40%|████      | 4478/11082 [14:23:24<21:12:14, 11.56s/step, acc=0.4851, loss=2.1447, lr=6.99e-05]
Training:  40%|████      | 4479/11082 [14:23:35<21:07:57, 11.52s/step, acc=0.4851, loss=2.1447, lr=6.99e-05]
Training:  40%|████      | 4480/11082 [14:23:47<21:13:22, 11.57s/step, acc=0.4851, loss=2.1447, lr=6.99e-05]
Training:  40%|████      | 4481/11082 [14:23:59<21:17:49, 11.61s/step, acc=0.4851, loss=2.1447, lr=6.99e-05]
Training:  40%|████      | 4482/11082 [14:24:10<21:01:09, 11.47s/step, acc=0.4851, loss=2.1447, lr=6.99e-05]
Training:  40%|████      | 4483/11082 [14:24:22<21:05:48, 11.51s/step, acc=0.4851, loss=2.1447, lr=6.99e-05]
Training:  40%|████      | 4484/11082 [14:24:33<21:16:02, 11.60s/step, acc=0.4851, loss=2.1447, lr=6.99e-05]
Training:  40%|████      | 4485/11082 [14:24:45<21:07:32, 11.53s/step, acc=0.4851, loss=2.1447, lr=6.99e-05]
Training:  40%|████      | 4486/11082 [14:24:56<21:10:34, 11.56s/step, acc=0.4851, loss=2.1447, lr=6.99e-05]
Training:  40%|████      | 4487/11082 [14:25:08<21:12:00, 11.57s/step, acc=0.4851, loss=2.1447, lr=6.99e-05]
Training:  40%|████      | 4488/11082 [14:25:19<21:07:06, 11.53s/step, acc=0.4851, loss=2.1447, lr=6.99e-05]
Training:  41%|████      | 4489/11082 [14:25:31<21:14:39, 11.60s/step, acc=0.4851, loss=2.1447, lr=6.99e-05]
Training:  41%|████      | 4490/11082 [14:25:43<21:14:20, 11.60s/step, acc=0.4851, loss=2.1447, lr=6.99e-05]
Training:  41%|████      | 4491/11082 [14:25:54<20:51:00, 11.39s/step, acc=0.4851, loss=2.1447, lr=6.99e-05]
Training:  41%|████      | 4492/11082 [14:26:06<21:09:29, 11.56s/step, acc=0.4851, loss=2.1447, lr=6.99e-05]
Training:  41%|████      | 4493/11082 [14:26:17<21:06:39, 11.53s/step, acc=0.4851, loss=2.1447, lr=6.99e-05]
Training:  41%|████      | 4494/11082 [14:26:29<21:09:51, 11.57s/step, acc=0.4851, loss=2.1447, lr=6.99e-05]
Training:  41%|████      | 4495/11082 [14:26:40<21:10:02, 11.57s/step, acc=0.4851, loss=2.1447, lr=6.99e-05]
Training:  41%|████      | 4496/11082 [14:26:52<21:04:34, 11.52s/step, acc=0.4851, loss=2.1447, lr=6.99e-05]
Training:  41%|████      | 4497/11082 [14:27:03<20:56:15, 11.45s/step, acc=0.4851, loss=2.1447, lr=6.99e-05]
Training:  41%|████      | 4498/11082 [14:27:15<21:06:45, 11.54s/step, acc=0.4851, loss=2.1447, lr=6.99e-05]
Training:  41%|████      | 4499/11082 [14:27:27<21:17:55, 11.65s/step, acc=0.4851, loss=2.1447, lr=6.99e-05]
Training:  41%|████      | 4500/11082 [14:27:38<21:22:36, 11.69s/step, acc=0.4851, loss=2.1447, lr=6.99e-05]
Training:  41%|████      | 4500/11082 [14:27:45<21:22:36, 11.69s/step, acc=0.4863, loss=2.1402, lr=6.93e-05]2026-06-19 05:21:13 | INFO | __main__ | epoch=0 step=4500 train_loss=2.140226 train_acc=0.486318 lr=6.930e-05 grad_norm=1.0312 acc_d0=0.7601 acc_d1=0.6151 acc_d2=0.4913 acc_d3=0.3905 acc_d4=0.3112 acc_d5=0.2494 acc_d6=0.2031 acc_d7=0.1673

Training:  41%|████      | 4501/11082 [14:27:50<21:07:13, 11.55s/step, acc=0.4863, loss=2.1402, lr=6.93e-05]
Training:  41%|████      | 4502/11082 [14:28:01<20:58:46, 11.48s/step, acc=0.4863, loss=2.1402, lr=6.93e-05]
Training:  41%|████      | 4503/11082 [14:28:13<21:19:24, 11.67s/step, acc=0.4863, loss=2.1402, lr=6.93e-05]
Training:  41%|████      | 4504/11082 [14:28:24<21:09:56, 11.58s/step, acc=0.4863, loss=2.1402, lr=6.93e-05]
Training:  41%|████      | 4505/11082 [14:28:36<21:22:53, 11.70s/step, acc=0.4863, loss=2.1402, lr=6.93e-05]
Training:  41%|████      | 4506/11082 [14:28:48<21:18:18, 11.66s/step, acc=0.4863, loss=2.1402, lr=6.93e-05]
Training:  41%|████      | 4507/11082 [14:28:59<21:07:50, 11.57s/step, acc=0.4863, loss=2.1402, lr=6.93e-05]
Training:  41%|████      | 4508/11082 [14:29:11<21:15:04, 11.64s/step, acc=0.4863, loss=2.1402, lr=6.93e-05]
Training:  41%|████      | 4509/11082 [14:29:22<21:00:39, 11.51s/step, acc=0.4863, loss=2.1402, lr=6.93e-05]
Training:  41%|████      | 4510/11082 [14:29:34<21:16:39, 11.66s/step, acc=0.4863, loss=2.1402, lr=6.93e-05]
Training:  41%|████      | 4511/11082 [14:29:46<21:26:50, 11.75s/step, acc=0.4863, loss=2.1402, lr=6.93e-05]
Training:  41%|████      | 4512/11082 [14:29:58<21:10:44, 11.60s/step, acc=0.4863, loss=2.1402, lr=6.93e-05]
Training:  41%|████      | 4513/11082 [14:30:09<21:11:21, 11.61s/step, acc=0.4863, loss=2.1402, lr=6.93e-05]
Training:  41%|████      | 4514/11082 [14:30:21<21:12:09, 11.62s/step, acc=0.4863, loss=2.1402, lr=6.93e-05]
Training:  41%|████      | 4515/11082 [14:30:32<21:05:56, 11.57s/step, acc=0.4863, loss=2.1402, lr=6.93e-05]
Training:  41%|████      | 4516/11082 [14:30:44<21:13:39, 11.64s/step, acc=0.4863, loss=2.1402, lr=6.93e-05]
Training:  41%|████      | 4517/11082 [14:30:56<21:15:24, 11.66s/step, acc=0.4863, loss=2.1402, lr=6.93e-05]
Training:  41%|████      | 4518/11082 [14:31:07<21:05:09, 11.56s/step, acc=0.4863, loss=2.1402, lr=6.93e-05]
Training:  41%|████      | 4519/11082 [14:31:19<21:05:18, 11.57s/step, acc=0.4863, loss=2.1402, lr=6.93e-05]
Training:  41%|████      | 4520/11082 [14:31:30<21:03:30, 11.55s/step, acc=0.4863, loss=2.1402, lr=6.93e-05]
Training:  41%|████      | 4521/11082 [14:31:42<20:57:35, 11.50s/step, acc=0.4863, loss=2.1402, lr=6.93e-05]
Training:  41%|████      | 4522/11082 [14:31:53<21:02:05, 11.54s/step, acc=0.4863, loss=2.1402, lr=6.93e-05]
Training:  41%|████      | 4523/11082 [14:32:05<20:53:41, 11.47s/step, acc=0.4863, loss=2.1402, lr=6.93e-05]
Training:  41%|████      | 4524/11082 [14:32:16<20:50:20, 11.44s/step, acc=0.4863, loss=2.1402, lr=6.93e-05]
Training:  41%|████      | 4525/11082 [14:32:27<20:52:14, 11.46s/step, acc=0.4863, loss=2.1402, lr=6.93e-05]
Training:  41%|████      | 4526/11082 [14:32:40<21:15:47, 11.68s/step, acc=0.4863, loss=2.1402, lr=6.93e-05]
Training:  41%|████      | 4527/11082 [14:32:51<21:07:13, 11.60s/step, acc=0.4863, loss=2.1402, lr=6.93e-05]
Training:  41%|████      | 4528/11082 [14:33:02<20:58:57, 11.53s/step, acc=0.4863, loss=2.1402, lr=6.93e-05]
Training:  41%|████      | 4529/11082 [14:33:14<20:55:38, 11.50s/step, acc=0.4863, loss=2.1402, lr=6.93e-05]
Training:  41%|████      | 4530/11082 [14:33:26<21:02:41, 11.56s/step, acc=0.4863, loss=2.1402, lr=6.93e-05]
Training:  41%|████      | 4531/11082 [14:33:37<21:05:36, 11.59s/step, acc=0.4863, loss=2.1402, lr=6.93e-05]
Training:  41%|████      | 4532/11082 [14:33:49<21:02:26, 11.56s/step, acc=0.4863, loss=2.1402, lr=6.93e-05]
Training:  41%|████      | 4533/11082 [14:34:00<20:45:52, 11.41s/step, acc=0.4863, loss=2.1402, lr=6.93e-05]
Training:  41%|████      | 4534/11082 [14:34:12<21:12:15, 11.66s/step, acc=0.4863, loss=2.1402, lr=6.93e-05]
Training:  41%|████      | 4535/11082 [14:34:24<21:11:58, 11.66s/step, acc=0.4863, loss=2.1402, lr=6.93e-05]
Training:  41%|████      | 4536/11082 [14:34:35<20:48:23, 11.44s/step, acc=0.4863, loss=2.1402, lr=6.93e-05]
Training:  41%|████      | 4537/11082 [14:34:47<21:04:55, 11.60s/step, acc=0.4863, loss=2.1402, lr=6.93e-05]
Training:  41%|████      | 4538/11082 [14:34:58<21:01:26, 11.57s/step, acc=0.4863, loss=2.1402, lr=6.93e-05]
Training:  41%|████      | 4539/11082 [14:35:10<21:07:34, 11.62s/step, acc=0.4863, loss=2.1402, lr=6.93e-05]
Training:  41%|████      | 4540/11082 [14:35:21<20:57:45, 11.54s/step, acc=0.4863, loss=2.1402, lr=6.93e-05]
Training:  41%|████      | 4541/11082 [14:35:33<21:01:25, 11.57s/step, acc=0.4863, loss=2.1402, lr=6.93e-05]
Training:  41%|████      | 4542/11082 [14:35:43<20:32:01, 11.30s/step, acc=0.4863, loss=2.1402, lr=6.93e-05]
Training:  41%|████      | 4543/11082 [14:35:56<21:10:19, 11.66s/step, acc=0.4863, loss=2.1402, lr=6.93e-05]
Training:  41%|████      | 4544/11082 [14:36:08<21:08:11, 11.64s/step, acc=0.4863, loss=2.1402, lr=6.93e-05]
Training:  41%|████      | 4545/11082 [14:36:19<20:52:10, 11.49s/step, acc=0.4863, loss=2.1402, lr=6.93e-05]
Training:  41%|████      | 4546/11082 [14:36:30<20:53:26, 11.51s/step, acc=0.4863, loss=2.1402, lr=6.93e-05]
Training:  41%|████      | 4547/11082 [14:36:42<20:54:24, 11.52s/step, acc=0.4863, loss=2.1402, lr=6.93e-05]
Training:  41%|████      | 4548/11082 [14:36:53<20:59:16, 11.56s/step, acc=0.4863, loss=2.1402, lr=6.93e-05]
Training:  41%|████      | 4549/11082 [14:37:05<21:00:24, 11.58s/step, acc=0.4863, loss=2.1402, lr=6.93e-05]
Training:  41%|████      | 4550/11082 [14:37:17<21:11:28, 11.68s/step, acc=0.4863, loss=2.1402, lr=6.93e-05]
Training:  41%|████      | 4550/11082 [14:37:24<21:11:28, 11.68s/step, acc=0.4864, loss=2.1372, lr=6.87e-05]2026-06-19 05:30:52 | INFO | __main__ | epoch=0 step=4550 train_loss=2.137201 train_acc=0.486437 lr=6.869e-05 grad_norm=1.2109 acc_d0=0.7455 acc_d1=0.5920 acc_d2=0.4660 acc_d3=0.3655 acc_d4=0.2871 acc_d5=0.2296 acc_d6=0.1856 acc_d7=0.1522

Training:  41%|████      | 4551/11082 [14:37:29<21:08:41, 11.66s/step, acc=0.4864, loss=2.1372, lr=6.87e-05]
Training:  41%|████      | 4552/11082 [14:37:40<21:05:56, 11.63s/step, acc=0.4864, loss=2.1372, lr=6.87e-05]
Training:  41%|████      | 4553/11082 [14:37:52<21:00:05, 11.58s/step, acc=0.4864, loss=2.1372, lr=6.87e-05]
Training:  41%|████      | 4554/11082 [14:38:03<21:06:12, 11.64s/step, acc=0.4864, loss=2.1372, lr=6.87e-05]
Training:  41%|████      | 4555/11082 [14:38:15<21:04:34, 11.62s/step, acc=0.4864, loss=2.1372, lr=6.87e-05]
Training:  41%|████      | 4556/11082 [14:38:27<21:05:26, 11.63s/step, acc=0.4864, loss=2.1372, lr=6.87e-05]
Training:  41%|████      | 4557/11082 [14:38:38<20:42:25, 11.42s/step, acc=0.4864, loss=2.1372, lr=6.87e-05]
Training:  41%|████      | 4558/11082 [14:38:49<20:50:18, 11.50s/step, acc=0.4864, loss=2.1372, lr=6.87e-05]
Training:  41%|████      | 4559/11082 [14:39:01<21:03:35, 11.62s/step, acc=0.4864, loss=2.1372, lr=6.87e-05]
Training:  41%|████      | 4560/11082 [14:39:13<20:57:09, 11.57s/step, acc=0.4864, loss=2.1372, lr=6.87e-05]
Training:  41%|████      | 4561/11082 [14:39:24<20:49:03, 11.49s/step, acc=0.4864, loss=2.1372, lr=6.87e-05]
Training:  41%|████      | 4562/11082 [14:39:35<20:49:42, 11.50s/step, acc=0.4864, loss=2.1372, lr=6.87e-05]
Training:  41%|████      | 4563/11082 [14:39:47<20:51:56, 11.52s/step, acc=0.4864, loss=2.1372, lr=6.87e-05]
Training:  41%|████      | 4564/11082 [14:39:58<20:50:01, 11.51s/step, acc=0.4864, loss=2.1372, lr=6.87e-05]
Training:  41%|████      | 4565/11082 [14:40:10<20:46:22, 11.47s/step, acc=0.4864, loss=2.1372, lr=6.87e-05]
Training:  41%|████      | 4566/11082 [14:40:22<21:07:47, 11.67s/step, acc=0.4864, loss=2.1372, lr=6.87e-05]
Training:  41%|████      | 4567/11082 [14:40:33<20:58:47, 11.59s/step, acc=0.4864, loss=2.1372, lr=6.87e-05]
Training:  41%|████      | 4568/11082 [14:40:45<21:00:02, 11.61s/step, acc=0.4864, loss=2.1372, lr=6.87e-05]
Training:  41%|████      | 4569/11082 [14:40:56<20:52:32, 11.54s/step, acc=0.4864, loss=2.1372, lr=6.87e-05]
Training:  41%|████      | 4570/11082 [14:41:08<20:44:43, 11.47s/step, acc=0.4864, loss=2.1372, lr=6.87e-05]
Training:  41%|████      | 4571/11082 [14:41:20<21:04:03, 11.65s/step, acc=0.4864, loss=2.1372, lr=6.87e-05]
Training:  41%|████▏     | 4572/11082 [14:41:31<20:44:50, 11.47s/step, acc=0.4864, loss=2.1372, lr=6.87e-05]
Training:  41%|████▏     | 4573/11082 [14:41:43<20:53:59, 11.56s/step, acc=0.4864, loss=2.1372, lr=6.87e-05]
Training:  41%|████▏     | 4574/11082 [14:41:54<21:04:00, 11.65s/step, acc=0.4864, loss=2.1372, lr=6.87e-05]
Training:  41%|████▏     | 4575/11082 [14:42:06<21:02:19, 11.64s/step, acc=0.4864, loss=2.1372, lr=6.87e-05]
Training:  41%|████▏     | 4576/11082 [14:42:18<20:55:16, 11.58s/step, acc=0.4864, loss=2.1372, lr=6.87e-05]
Training:  41%|████▏     | 4577/11082 [14:42:29<20:41:21, 11.45s/step, acc=0.4864, loss=2.1372, lr=6.87e-05]
Training:  41%|████▏     | 4578/11082 [14:42:40<20:43:16, 11.47s/step, acc=0.4864, loss=2.1372, lr=6.87e-05]
Training:  41%|████▏     | 4579/11082 [14:42:52<20:56:52, 11.60s/step, acc=0.4864, loss=2.1372, lr=6.87e-05]
Training:  41%|████▏     | 4580/11082 [14:43:04<20:53:53, 11.57s/step, acc=0.4864, loss=2.1372, lr=6.87e-05]
Training:  41%|████▏     | 4581/11082 [14:43:16<21:07:35, 11.70s/step, acc=0.4864, loss=2.1372, lr=6.87e-05]
Training:  41%|████▏     | 4582/11082 [14:43:27<20:59:40, 11.63s/step, acc=0.4864, loss=2.1372, lr=6.87e-05]
Training:  41%|████▏     | 4583/11082 [14:43:39<21:00:41, 11.64s/step, acc=0.4864, loss=2.1372, lr=6.87e-05]
Training:  41%|████▏     | 4584/11082 [14:43:50<21:03:38, 11.67s/step, acc=0.4864, loss=2.1372, lr=6.87e-05]
Training:  41%|████▏     | 4585/11082 [14:44:02<20:57:19, 11.61s/step, acc=0.4864, loss=2.1372, lr=6.87e-05]
Training:  41%|████▏     | 4586/11082 [14:44:14<21:12:01, 11.75s/step, acc=0.4864, loss=2.1372, lr=6.87e-05]
Training:  41%|████▏     | 4587/11082 [14:44:25<20:57:44, 11.62s/step, acc=0.4864, loss=2.1372, lr=6.87e-05]
Training:  41%|████▏     | 4588/11082 [14:44:37<20:59:27, 11.64s/step, acc=0.4864, loss=2.1372, lr=6.87e-05]
Training:  41%|████▏     | 4589/11082 [14:44:49<21:06:19, 11.70s/step, acc=0.4864, loss=2.1372, lr=6.87e-05]
Training:  41%|████▏     | 4590/11082 [14:45:01<21:10:36, 11.74s/step, acc=0.4864, loss=2.1372, lr=6.87e-05]
Training:  41%|████▏     | 4591/11082 [14:45:12<20:48:40, 11.54s/step, acc=0.4864, loss=2.1372, lr=6.87e-05]
Training:  41%|████▏     | 4592/11082 [14:45:24<21:01:58, 11.67s/step, acc=0.4864, loss=2.1372, lr=6.87e-05]
Training:  41%|████▏     | 4593/11082 [14:45:35<20:59:14, 11.64s/step, acc=0.4864, loss=2.1372, lr=6.87e-05]
Training:  41%|████▏     | 4594/11082 [14:45:47<20:47:07, 11.53s/step, acc=0.4864, loss=2.1372, lr=6.87e-05]
Training:  41%|████▏     | 4595/11082 [14:45:58<20:45:22, 11.52s/step, acc=0.4864, loss=2.1372, lr=6.87e-05]
Training:  41%|████▏     | 4596/11082 [14:46:10<20:45:20, 11.52s/step, acc=0.4864, loss=2.1372, lr=6.87e-05]
Training:  41%|████▏     | 4597/11082 [14:46:21<20:53:29, 11.60s/step, acc=0.4864, loss=2.1372, lr=6.87e-05]
Training:  41%|████▏     | 4598/11082 [14:46:33<20:51:37, 11.58s/step, acc=0.4864, loss=2.1372, lr=6.87e-05]
Training:  41%|████▏     | 4599/11082 [14:46:44<20:33:27, 11.42s/step, acc=0.4864, loss=2.1372, lr=6.87e-05]
Training:  42%|████▏     | 4600/11082 [14:46:56<21:02:51, 11.69s/step, acc=0.4864, loss=2.1372, lr=6.87e-05]
Training:  42%|████▏     | 4600/11082 [14:47:03<21:02:51, 11.69s/step, acc=0.4876, loss=2.1273, lr=6.81e-05]2026-06-19 05:40:31 | INFO | __main__ | epoch=0 step=4600 train_loss=2.127311 train_acc=0.487632 lr=6.807e-05 grad_norm=1.3047 acc_d0=0.7479 acc_d1=0.5967 acc_d2=0.4729 acc_d3=0.3721 acc_d4=0.2944 acc_d5=0.2319 acc_d6=0.1889 acc_d7=0.1552

Training:  42%|████▏     | 4601/11082 [14:47:07<20:43:37, 11.51s/step, acc=0.4876, loss=2.1273, lr=6.81e-05]
Training:  42%|████▏     | 4602/11082 [14:47:19<20:43:15, 11.51s/step, acc=0.4876, loss=2.1273, lr=6.81e-05]
Training:  42%|████▏     | 4603/11082 [14:47:31<20:53:49, 11.61s/step, acc=0.4876, loss=2.1273, lr=6.81e-05]
Training:  42%|████▏     | 4604/11082 [14:47:42<20:58:50, 11.66s/step, acc=0.4876, loss=2.1273, lr=6.81e-05]
Training:  42%|████▏     | 4605/11082 [14:47:53<20:32:07, 11.41s/step, acc=0.4876, loss=2.1273, lr=6.81e-05]
Training:  42%|████▏     | 4606/11082 [14:48:05<20:53:05, 11.61s/step, acc=0.4876, loss=2.1273, lr=6.81e-05]
Training:  42%|████▏     | 4607/11082 [14:48:17<20:43:10, 11.52s/step, acc=0.4876, loss=2.1273, lr=6.81e-05]
Training:  42%|████▏     | 4608/11082 [14:48:29<20:59:42, 11.67s/step, acc=0.4876, loss=2.1273, lr=6.81e-05]
Training:  42%|████▏     | 4609/11082 [14:48:40<20:50:46, 11.59s/step, acc=0.4876, loss=2.1273, lr=6.81e-05]
Training:  42%|████▏     | 4610/11082 [14:48:52<20:58:00, 11.66s/step, acc=0.4876, loss=2.1273, lr=6.81e-05]
Training:  42%|████▏     | 4611/11082 [14:49:04<21:03:16, 11.71s/step, acc=0.4876, loss=2.1273, lr=6.81e-05]
Training:  42%|████▏     | 4612/11082 [14:49:16<21:06:14, 11.74s/step, acc=0.4876, loss=2.1273, lr=6.81e-05]
Training:  42%|████▏     | 4613/11082 [14:49:27<20:56:01, 11.65s/step, acc=0.4876, loss=2.1273, lr=6.81e-05]
Training:  42%|████▏     | 4614/11082 [14:49:39<21:02:43, 11.71s/step, acc=0.4876, loss=2.1273, lr=6.81e-05]
Training:  42%|████▏     | 4615/11082 [14:49:50<20:40:32, 11.51s/step, acc=0.4876, loss=2.1273, lr=6.81e-05]
Training:  42%|████▏     | 4616/11082 [14:50:02<20:46:31, 11.57s/step, acc=0.4876, loss=2.1273, lr=6.81e-05]
Training:  42%|████▏     | 4617/11082 [14:50:13<20:46:14, 11.57s/step, acc=0.4876, loss=2.1273, lr=6.81e-05]
Training:  42%|████▏     | 4618/11082 [14:50:25<20:49:35, 11.60s/step, acc=0.4876, loss=2.1273, lr=6.81e-05]
Training:  42%|████▏     | 4619/11082 [14:50:36<20:31:57, 11.44s/step, acc=0.4876, loss=2.1273, lr=6.81e-05]
Training:  42%|████▏     | 4620/11082 [14:50:48<20:47:05, 11.58s/step, acc=0.4876, loss=2.1273, lr=6.81e-05]
Training:  42%|████▏     | 4621/11082 [14:50:59<20:32:43, 11.45s/step, acc=0.4876, loss=2.1273, lr=6.81e-05]
Training:  42%|████▏     | 4622/11082 [14:51:11<20:49:13, 11.60s/step, acc=0.4876, loss=2.1273, lr=6.81e-05]
Training:  42%|████▏     | 4623/11082 [14:51:22<20:46:07, 11.58s/step, acc=0.4876, loss=2.1273, lr=6.81e-05]
Training:  42%|████▏     | 4624/11082 [14:51:34<20:39:50, 11.52s/step, acc=0.4876, loss=2.1273, lr=6.81e-05]
Training:  42%|████▏     | 4625/11082 [14:51:46<20:55:49, 11.67s/step, acc=0.4876, loss=2.1273, lr=6.81e-05]
Training:  42%|████▏     | 4626/11082 [14:51:57<20:40:26, 11.53s/step, acc=0.4876, loss=2.1273, lr=6.81e-05]
Training:  42%|████▏     | 4627/11082 [14:52:09<20:49:15, 11.61s/step, acc=0.4876, loss=2.1273, lr=6.81e-05]
Training:  42%|████▏     | 4628/11082 [14:52:20<20:33:46, 11.47s/step, acc=0.4876, loss=2.1273, lr=6.81e-05]
Training:  42%|████▏     | 4629/11082 [14:52:32<20:38:41, 11.52s/step, acc=0.4876, loss=2.1273, lr=6.81e-05]
Training:  42%|████▏     | 4630/11082 [14:52:43<20:37:24, 11.51s/step, acc=0.4876, loss=2.1273, lr=6.81e-05]
Training:  42%|████▏     | 4631/11082 [14:52:55<20:37:28, 11.51s/step, acc=0.4876, loss=2.1273, lr=6.81e-05]
Training:  42%|████▏     | 4632/11082 [14:53:06<20:36:00, 11.50s/step, acc=0.4876, loss=2.1273, lr=6.81e-05]
Training:  42%|████▏     | 4633/11082 [14:53:18<20:40:44, 11.54s/step, acc=0.4876, loss=2.1273, lr=6.81e-05]
Training:  42%|████▏     | 4634/11082 [14:53:29<20:30:17, 11.45s/step, acc=0.4876, loss=2.1273, lr=6.81e-05]
Training:  42%|████▏     | 4635/11082 [14:53:41<20:55:54, 11.69s/step, acc=0.4876, loss=2.1273, lr=6.81e-05]
Training:  42%|████▏     | 4636/11082 [14:53:53<20:49:32, 11.63s/step, acc=0.4876, loss=2.1273, lr=6.81e-05]
Training:  42%|████▏     | 4637/11082 [14:54:04<20:37:01, 11.52s/step, acc=0.4876, loss=2.1273, lr=6.81e-05]
Training:  42%|████▏     | 4638/11082 [14:54:16<20:38:38, 11.53s/step, acc=0.4876, loss=2.1273, lr=6.81e-05]
Training:  42%|████▏     | 4639/11082 [14:54:28<20:53:34, 11.67s/step, acc=0.4876, loss=2.1273, lr=6.81e-05]
Training:  42%|████▏     | 4640/11082 [14:54:39<20:39:30, 11.54s/step, acc=0.4876, loss=2.1273, lr=6.81e-05]
Training:  42%|████▏     | 4641/11082 [14:54:51<20:52:10, 11.66s/step, acc=0.4876, loss=2.1273, lr=6.81e-05]
Training:  42%|████▏     | 4642/11082 [14:55:02<20:38:15, 11.54s/step, acc=0.4876, loss=2.1273, lr=6.81e-05]
Training:  42%|████▏     | 4643/11082 [14:55:14<20:41:10, 11.57s/step, acc=0.4876, loss=2.1273, lr=6.81e-05]
Training:  42%|████▏     | 4644/11082 [14:55:25<20:46:59, 11.62s/step, acc=0.4876, loss=2.1273, lr=6.81e-05]
Training:  42%|████▏     | 4645/11082 [14:55:37<20:40:16, 11.56s/step, acc=0.4876, loss=2.1273, lr=6.81e-05]
Training:  42%|████▏     | 4646/11082 [14:55:49<20:50:58, 11.66s/step, acc=0.4876, loss=2.1273, lr=6.81e-05]
Training:  42%|████▏     | 4647/11082 [14:56:00<20:32:43, 11.49s/step, acc=0.4876, loss=2.1273, lr=6.81e-05]
Training:  42%|████▏     | 4648/11082 [14:56:11<20:34:55, 11.52s/step, acc=0.4876, loss=2.1273, lr=6.81e-05]
Training:  42%|████▏     | 4649/11082 [14:56:23<20:46:05, 11.62s/step, acc=0.4876, loss=2.1273, lr=6.81e-05]
Training:  42%|████▏     | 4650/11082 [14:56:35<20:43:49, 11.60s/step, acc=0.4876, loss=2.1273, lr=6.81e-05]
Training:  42%|████▏     | 4650/11082 [14:56:42<20:43:49, 11.60s/step, acc=0.4894, loss=2.1213, lr=6.74e-05]2026-06-19 05:50:09 | INFO | __main__ | epoch=0 step=4650 train_loss=2.121306 train_acc=0.489373 lr=6.745e-05 grad_norm=0.9180 acc_d0=0.7609 acc_d1=0.6148 acc_d2=0.4916 acc_d3=0.3913 acc_d4=0.3117 acc_d5=0.2495 acc_d6=0.2008 acc_d7=0.1664

Training:  42%|████▏     | 4651/11082 [14:56:46<20:20:43, 11.39s/step, acc=0.4894, loss=2.1213, lr=6.74e-05]
Training:  42%|████▏     | 4652/11082 [14:56:57<20:30:46, 11.48s/step, acc=0.4894, loss=2.1213, lr=6.74e-05]
Training:  42%|████▏     | 4653/11082 [14:57:09<20:27:08, 11.45s/step, acc=0.4894, loss=2.1213, lr=6.74e-05]
Training:  42%|████▏     | 4654/11082 [14:57:21<20:44:34, 11.62s/step, acc=0.4894, loss=2.1213, lr=6.74e-05]
Training:  42%|████▏     | 4655/11082 [14:57:32<20:42:38, 11.60s/step, acc=0.4894, loss=2.1213, lr=6.74e-05]
Training:  42%|████▏     | 4656/11082 [14:57:44<20:28:50, 11.47s/step, acc=0.4894, loss=2.1213, lr=6.74e-05]
Training:  42%|████▏     | 4657/11082 [14:57:55<20:36:47, 11.55s/step, acc=0.4894, loss=2.1213, lr=6.74e-05]
Training:  42%|████▏     | 4658/11082 [14:58:07<20:36:40, 11.55s/step, acc=0.4894, loss=2.1213, lr=6.74e-05]
Training:  42%|████▏     | 4659/11082 [14:58:18<20:39:04, 11.57s/step, acc=0.4894, loss=2.1213, lr=6.74e-05]
Training:  42%|████▏     | 4660/11082 [14:58:30<20:38:10, 11.57s/step, acc=0.4894, loss=2.1213, lr=6.74e-05]
Training:  42%|████▏     | 4661/11082 [14:58:41<20:34:31, 11.54s/step, acc=0.4894, loss=2.1213, lr=6.74e-05]
Training:  42%|████▏     | 4662/11082 [14:58:54<20:53:34, 11.72s/step, acc=0.4894, loss=2.1213, lr=6.74e-05]
Training:  42%|████▏     | 4663/11082 [14:59:05<20:29:40, 11.49s/step, acc=0.4894, loss=2.1213, lr=6.74e-05]
Training:  42%|████▏     | 4664/11082 [14:59:16<20:41:48, 11.61s/step, acc=0.4894, loss=2.1213, lr=6.74e-05]
Training:  42%|████▏     | 4665/11082 [14:59:28<20:54:44, 11.73s/step, acc=0.4894, loss=2.1213, lr=6.74e-05]
Training:  42%|████▏     | 4666/11082 [14:59:40<20:45:50, 11.65s/step, acc=0.4894, loss=2.1213, lr=6.74e-05]
Training:  42%|████▏     | 4667/11082 [14:59:51<20:36:43, 11.57s/step, acc=0.4894, loss=2.1213, lr=6.74e-05]
Training:  42%|████▏     | 4668/11082 [15:00:03<20:44:17, 11.64s/step, acc=0.4894, loss=2.1213, lr=6.74e-05]
Training:  42%|████▏     | 4669/11082 [15:00:15<20:49:56, 11.69s/step, acc=0.4894, loss=2.1213, lr=6.74e-05]
Training:  42%|████▏     | 4670/11082 [15:00:26<20:33:05, 11.54s/step, acc=0.4894, loss=2.1213, lr=6.74e-05]
Training:  42%|████▏     | 4671/11082 [15:00:38<20:34:19, 11.55s/step, acc=0.4894, loss=2.1213, lr=6.74e-05]
Training:  42%|████▏     | 4672/11082 [15:00:49<20:34:46, 11.56s/step, acc=0.4894, loss=2.1213, lr=6.74e-05]
Training:  42%|████▏     | 4673/11082 [15:01:01<20:36:09, 11.57s/step, acc=0.4894, loss=2.1213, lr=6.74e-05]
Training:  42%|████▏     | 4674/11082 [15:01:12<20:35:21, 11.57s/step, acc=0.4894, loss=2.1213, lr=6.74e-05]
Training:  42%|████▏     | 4675/11082 [15:01:24<20:32:44, 11.54s/step, acc=0.4894, loss=2.1213, lr=6.74e-05]
Training:  42%|████▏     | 4676/11082 [15:01:36<20:45:19, 11.66s/step, acc=0.4894, loss=2.1213, lr=6.74e-05]
Training:  42%|████▏     | 4677/11082 [15:01:47<20:25:19, 11.48s/step, acc=0.4894, loss=2.1213, lr=6.74e-05]
Training:  42%|████▏     | 4678/11082 [15:01:59<20:48:20, 11.70s/step, acc=0.4894, loss=2.1213, lr=6.74e-05]
Training:  42%|████▏     | 4679/11082 [15:02:11<20:43:48, 11.66s/step, acc=0.4894, loss=2.1213, lr=6.74e-05]
Training:  42%|████▏     | 4680/11082 [15:02:22<20:21:07, 11.44s/step, acc=0.4894, loss=2.1213, lr=6.74e-05]
Training:  42%|████▏     | 4681/11082 [15:02:33<20:34:28, 11.57s/step, acc=0.4894, loss=2.1213, lr=6.74e-05]
Training:  42%|████▏     | 4682/11082 [15:02:45<20:42:13, 11.65s/step, acc=0.4894, loss=2.1213, lr=6.74e-05]
Training:  42%|████▏     | 4683/11082 [15:02:57<20:38:55, 11.62s/step, acc=0.4894, loss=2.1213, lr=6.74e-05]
Training:  42%|████▏     | 4684/11082 [15:03:08<20:32:44, 11.56s/step, acc=0.4894, loss=2.1213, lr=6.74e-05]
Training:  42%|████▏     | 4685/11082 [15:03:19<20:17:23, 11.42s/step, acc=0.4894, loss=2.1213, lr=6.74e-05]
Training:  42%|████▏     | 4686/11082 [15:03:31<20:29:33, 11.53s/step, acc=0.4894, loss=2.1213, lr=6.74e-05]
Training:  42%|████▏     | 4687/11082 [15:03:43<20:47:47, 11.71s/step, acc=0.4894, loss=2.1213, lr=6.74e-05]
Training:  42%|████▏     | 4688/11082 [15:03:55<20:41:58, 11.65s/step, acc=0.4894, loss=2.1213, lr=6.74e-05]
Training:  42%|████▏     | 4689/11082 [15:04:06<20:29:32, 11.54s/step, acc=0.4894, loss=2.1213, lr=6.74e-05]
Training:  42%|████▏     | 4690/11082 [15:04:18<20:32:29, 11.57s/step, acc=0.4894, loss=2.1213, lr=6.74e-05]
Training:  42%|████▏     | 4691/11082 [15:04:29<20:35:17, 11.60s/step, acc=0.4894, loss=2.1213, lr=6.74e-05]
Training:  42%|████▏     | 4692/11082 [15:04:41<20:33:07, 11.58s/step, acc=0.4894, loss=2.1213, lr=6.74e-05]
Training:  42%|████▏     | 4693/11082 [15:04:53<20:36:35, 11.61s/step, acc=0.4894, loss=2.1213, lr=6.74e-05]
Training:  42%|████▏     | 4694/11082 [15:05:04<20:39:09, 11.64s/step, acc=0.4894, loss=2.1213, lr=6.74e-05]
Training:  42%|████▏     | 4695/11082 [15:05:16<20:36:13, 11.61s/step, acc=0.4894, loss=2.1213, lr=6.74e-05]
Training:  42%|████▏     | 4696/11082 [15:05:27<20:32:52, 11.58s/step, acc=0.4894, loss=2.1213, lr=6.74e-05]
Training:  42%|████▏     | 4697/11082 [15:05:39<20:41:09, 11.66s/step, acc=0.4894, loss=2.1213, lr=6.74e-05]
Training:  42%|████▏     | 4698/11082 [15:05:51<20:35:02, 11.61s/step, acc=0.4894, loss=2.1213, lr=6.74e-05]
Training:  42%|████▏     | 4699/11082 [15:06:02<20:17:06, 11.44s/step, acc=0.4894, loss=2.1213, lr=6.74e-05]
Training:  42%|████▏     | 4700/11082 [15:06:14<20:27:45, 11.54s/step, acc=0.4894, loss=2.1213, lr=6.74e-05]
Training:  42%|████▏     | 4700/11082 [15:06:21<20:27:45, 11.54s/step, acc=0.4890, loss=2.1217, lr=6.68e-05]2026-06-19 05:59:49 | INFO | __main__ | epoch=0 step=4700 train_loss=2.121674 train_acc=0.489028 lr=6.683e-05 grad_norm=1.2344 acc_d0=0.7548 acc_d1=0.6062 acc_d2=0.4813 acc_d3=0.3812 acc_d4=0.3017 acc_d5=0.2404 acc_d6=0.1950 acc_d7=0.1629

Training:  42%|████▏     | 4701/11082 [15:06:25<20:33:01, 11.59s/step, acc=0.4890, loss=2.1217, lr=6.68e-05]
Training:  42%|████▏     | 4702/11082 [15:06:37<20:26:03, 11.53s/step, acc=0.4890, loss=2.1217, lr=6.68e-05]
Training:  42%|████▏     | 4703/11082 [15:06:48<20:28:33, 11.56s/step, acc=0.4890, loss=2.1217, lr=6.68e-05]
Training:  42%|████▏     | 4704/11082 [15:07:00<20:30:31, 11.58s/step, acc=0.4890, loss=2.1217, lr=6.68e-05]
Training:  42%|████▏     | 4705/11082 [15:07:12<20:36:10, 11.63s/step, acc=0.4890, loss=2.1217, lr=6.68e-05]
Training:  42%|████▏     | 4706/11082 [15:07:23<20:33:37, 11.61s/step, acc=0.4890, loss=2.1217, lr=6.68e-05]
Training:  42%|████▏     | 4707/11082 [15:07:35<20:27:37, 11.55s/step, acc=0.4890, loss=2.1217, lr=6.68e-05]
Training:  42%|████▏     | 4708/11082 [15:07:46<20:11:58, 11.41s/step, acc=0.4890, loss=2.1217, lr=6.68e-05]
Training:  42%|████▏     | 4709/11082 [15:07:58<20:36:37, 11.64s/step, acc=0.4890, loss=2.1217, lr=6.68e-05]
Training:  43%|████▎     | 4710/11082 [15:08:09<20:25:15, 11.54s/step, acc=0.4890, loss=2.1217, lr=6.68e-05]
Training:  43%|████▎     | 4711/11082 [15:08:21<20:23:21, 11.52s/step, acc=0.4890, loss=2.1217, lr=6.68e-05]
Training:  43%|████▎     | 4712/11082 [15:08:32<20:24:47, 11.54s/step, acc=0.4890, loss=2.1217, lr=6.68e-05]
Training:  43%|████▎     | 4713/11082 [15:08:44<20:18:34, 11.48s/step, acc=0.4890, loss=2.1217, lr=6.68e-05]
Training:  43%|████▎     | 4714/11082 [15:08:56<20:37:32, 11.66s/step, acc=0.4890, loss=2.1217, lr=6.68e-05]
Training:  43%|████▎     | 4715/11082 [15:09:07<20:24:49, 11.54s/step, acc=0.4890, loss=2.1217, lr=6.68e-05]
Training:  43%|████▎     | 4716/11082 [15:09:19<20:32:58, 11.62s/step, acc=0.4890, loss=2.1217, lr=6.68e-05]
Training:  43%|████▎     | 4717/11082 [15:09:30<20:20:09, 11.50s/step, acc=0.4890, loss=2.1217, lr=6.68e-05]
Training:  43%|████▎     | 4718/11082 [15:09:41<20:16:09, 11.47s/step, acc=0.4890, loss=2.1217, lr=6.68e-05]
Training:  43%|████▎     | 4719/11082 [15:09:53<20:14:11, 11.45s/step, acc=0.4890, loss=2.1217, lr=6.68e-05]
Training:  43%|████▎     | 4720/11082 [15:10:05<20:31:23, 11.61s/step, acc=0.4890, loss=2.1217, lr=6.68e-05]
Training:  43%|████▎     | 4721/11082 [15:10:16<20:21:57, 11.53s/step, acc=0.4890, loss=2.1217, lr=6.68e-05]
Training:  43%|████▎     | 4722/11082 [15:10:28<20:34:54, 11.65s/step, acc=0.4890, loss=2.1217, lr=6.68e-05]
Training:  43%|████▎     | 4723/11082 [15:10:40<20:32:54, 11.63s/step, acc=0.4890, loss=2.1217, lr=6.68e-05]
Training:  43%|████▎     | 4724/11082 [15:10:51<20:17:59, 11.49s/step, acc=0.4890, loss=2.1217, lr=6.68e-05]
Training:  43%|████▎     | 4725/11082 [15:11:03<20:35:03, 11.66s/step, acc=0.4890, loss=2.1217, lr=6.68e-05]
Training:  43%|████▎     | 4726/11082 [15:11:14<20:22:09, 11.54s/step, acc=0.4890, loss=2.1217, lr=6.68e-05]
Training:  43%|████▎     | 4727/11082 [15:11:25<20:10:22, 11.43s/step, acc=0.4890, loss=2.1217, lr=6.68e-05]
Training:  43%|████▎     | 4728/11082 [15:11:37<20:21:55, 11.54s/step, acc=0.4890, loss=2.1217, lr=6.68e-05]
Training:  43%|████▎     | 4729/11082 [15:11:48<20:11:48, 11.44s/step, acc=0.4890, loss=2.1217, lr=6.68e-05]
Training:  43%|████▎     | 4730/11082 [15:12:00<20:13:18, 11.46s/step, acc=0.4890, loss=2.1217, lr=6.68e-05]
Training:  43%|████▎     | 4731/11082 [15:12:11<20:21:45, 11.54s/step, acc=0.4890, loss=2.1217, lr=6.68e-05]
Training:  43%|████▎     | 4732/11082 [15:12:23<20:21:15, 11.54s/step, acc=0.4890, loss=2.1217, lr=6.68e-05]
Training:  43%|████▎     | 4733/11082 [15:12:35<20:21:10, 11.54s/step, acc=0.4890, loss=2.1217, lr=6.68e-05]
Training:  43%|████▎     | 4734/11082 [15:12:46<20:28:10, 11.61s/step, acc=0.4890, loss=2.1217, lr=6.68e-05]
Training:  43%|████▎     | 4735/11082 [15:12:58<20:32:40, 11.65s/step, acc=0.4890, loss=2.1217, lr=6.68e-05]
Training:  43%|████▎     | 4736/11082 [15:13:09<20:22:47, 11.56s/step, acc=0.4890, loss=2.1217, lr=6.68e-05]
Training:  43%|████▎     | 4737/11082 [15:13:21<20:16:42, 11.51s/step, acc=0.4890, loss=2.1217, lr=6.68e-05]
Training:  43%|████▎     | 4738/11082 [15:13:33<20:23:52, 11.58s/step, acc=0.4890, loss=2.1217, lr=6.68e-05]
Training:  43%|████▎     | 4739/11082 [15:13:44<20:13:22, 11.48s/step, acc=0.4890, loss=2.1217, lr=6.68e-05]
Training:  43%|████▎     | 4740/11082 [15:13:56<20:31:24, 11.65s/step, acc=0.4890, loss=2.1217, lr=6.68e-05]
Training:  43%|████▎     | 4741/11082 [15:14:07<20:21:25, 11.56s/step, acc=0.4890, loss=2.1217, lr=6.68e-05]
Training:  43%|████▎     | 4742/11082 [15:14:19<20:30:15, 11.64s/step, acc=0.4890, loss=2.1217, lr=6.68e-05]
Training:  43%|████▎     | 4743/11082 [15:14:30<20:25:15, 11.60s/step, acc=0.4890, loss=2.1217, lr=6.68e-05]
Training:  43%|████▎     | 4744/11082 [15:14:42<20:22:27, 11.57s/step, acc=0.4890, loss=2.1217, lr=6.68e-05]
Training:  43%|████▎     | 4745/11082 [15:14:54<20:26:21, 11.61s/step, acc=0.4890, loss=2.1217, lr=6.68e-05]
Training:  43%|████▎     | 4746/11082 [15:15:05<20:28:45, 11.64s/step, acc=0.4890, loss=2.1217, lr=6.68e-05]
Training:  43%|████▎     | 4747/11082 [15:15:17<20:30:39, 11.66s/step, acc=0.4890, loss=2.1217, lr=6.68e-05]
Training:  43%|████▎     | 4748/11082 [15:15:29<20:25:55, 11.61s/step, acc=0.4890, loss=2.1217, lr=6.68e-05]
Training:  43%|████▎     | 4749/11082 [15:15:40<20:33:38, 11.69s/step, acc=0.4890, loss=2.1217, lr=6.68e-05]
Training:  43%|████▎     | 4750/11082 [15:15:52<20:36:26, 11.72s/step, acc=0.4890, loss=2.1217, lr=6.68e-05]
Training:  43%|████▎     | 4750/11082 [15:15:59<20:36:26, 11.72s/step, acc=0.4861, loss=2.1411, lr=6.62e-05]2026-06-19 06:09:27 | INFO | __main__ | epoch=0 step=4750 train_loss=2.141063 train_acc=0.486072 lr=6.620e-05 grad_norm=1.4297 acc_d0=0.7574 acc_d1=0.6095 acc_d2=0.4868 acc_d3=0.3864 acc_d4=0.3086 acc_d5=0.2481 acc_d6=0.2020 acc_d7=0.1661

Training:  43%|████▎     | 4751/11082 [15:16:03<20:17:33, 11.54s/step, acc=0.4861, loss=2.1411, lr=6.62e-05]
Training:  43%|████▎     | 4752/11082 [15:16:15<20:09:10, 11.46s/step, acc=0.4861, loss=2.1411, lr=6.62e-05]
Training:  43%|████▎     | 4753/11082 [15:16:27<20:33:12, 11.69s/step, acc=0.4861, loss=2.1411, lr=6.62e-05]
Training:  43%|████▎     | 4754/11082 [15:16:38<20:16:52, 11.54s/step, acc=0.4861, loss=2.1411, lr=6.62e-05]
Training:  43%|████▎     | 4755/11082 [15:16:50<20:18:25, 11.55s/step, acc=0.4861, loss=2.1411, lr=6.62e-05]
Training:  43%|████▎     | 4756/11082 [15:17:01<20:24:17, 11.61s/step, acc=0.4861, loss=2.1411, lr=6.62e-05]
Training:  43%|████▎     | 4757/11082 [15:17:14<20:40:59, 11.77s/step, acc=0.4861, loss=2.1411, lr=6.62e-05]
Training:  43%|████▎     | 4758/11082 [15:17:25<20:29:18, 11.66s/step, acc=0.4861, loss=2.1411, lr=6.62e-05]
Training:  43%|████▎     | 4759/11082 [15:17:37<20:28:06, 11.65s/step, acc=0.4861, loss=2.1411, lr=6.62e-05]
Training:  43%|████▎     | 4760/11082 [15:17:48<20:19:41, 11.58s/step, acc=0.4861, loss=2.1411, lr=6.62e-05]
Training:  43%|████▎     | 4761/11082 [15:18:00<20:29:05, 11.67s/step, acc=0.4861, loss=2.1411, lr=6.62e-05]
Training:  43%|████▎     | 4762/11082 [15:18:11<20:23:38, 11.62s/step, acc=0.4861, loss=2.1411, lr=6.62e-05]
Training:  43%|████▎     | 4763/11082 [15:18:23<20:19:12, 11.58s/step, acc=0.4861, loss=2.1411, lr=6.62e-05]
Training:  43%|████▎     | 4764/11082 [15:18:34<20:07:36, 11.47s/step, acc=0.4861, loss=2.1411, lr=6.62e-05]
Training:  43%|████▎     | 4765/11082 [15:18:46<20:17:38, 11.57s/step, acc=0.4861, loss=2.1411, lr=6.62e-05]
Training:  43%|████▎     | 4766/11082 [15:18:57<20:06:55, 11.47s/step, acc=0.4861, loss=2.1411, lr=6.62e-05]
Training:  43%|████▎     | 4767/11082 [15:19:09<20:12:41, 11.52s/step, acc=0.4861, loss=2.1411, lr=6.62e-05]
Training:  43%|████▎     | 4768/11082 [15:19:20<20:16:54, 11.56s/step, acc=0.4861, loss=2.1411, lr=6.62e-05]
Training:  43%|████▎     | 4769/11082 [15:19:32<20:02:29, 11.43s/step, acc=0.4861, loss=2.1411, lr=6.62e-05]
Training:  43%|████▎     | 4770/11082 [15:19:43<20:13:29, 11.54s/step, acc=0.4861, loss=2.1411, lr=6.62e-05]
Training:  43%|████▎     | 4771/11082 [15:19:55<20:20:16, 11.60s/step, acc=0.4861, loss=2.1411, lr=6.62e-05]
Training:  43%|████▎     | 4772/11082 [15:20:06<20:13:47, 11.54s/step, acc=0.4861, loss=2.1411, lr=6.62e-05]
Training:  43%|████▎     | 4773/11082 [15:20:18<20:05:18, 11.46s/step, acc=0.4861, loss=2.1411, lr=6.62e-05]
Training:  43%|████▎     | 4774/11082 [15:20:30<20:34:17, 11.74s/step, acc=0.4861, loss=2.1411, lr=6.62e-05]
Training:  43%|████▎     | 4775/11082 [15:20:42<20:31:24, 11.71s/step, acc=0.4861, loss=2.1411, lr=6.62e-05]
Training:  43%|████▎     | 4776/11082 [15:20:53<20:26:24, 11.67s/step, acc=0.4861, loss=2.1411, lr=6.62e-05]
Training:  43%|████▎     | 4777/11082 [15:21:05<20:30:02, 11.71s/step, acc=0.4861, loss=2.1411, lr=6.62e-05]
Training:  43%|████▎     | 4778/11082 [15:21:17<20:23:24, 11.64s/step, acc=0.4861, loss=2.1411, lr=6.62e-05]
Training:  43%|████▎     | 4779/11082 [15:21:28<20:18:31, 11.60s/step, acc=0.4861, loss=2.1411, lr=6.62e-05]
Training:  43%|████▎     | 4780/11082 [15:21:39<20:06:51, 11.49s/step, acc=0.4861, loss=2.1411, lr=6.62e-05]
Training:  43%|████▎     | 4781/11082 [15:21:51<20:17:27, 11.59s/step, acc=0.4861, loss=2.1411, lr=6.62e-05]
Training:  43%|████▎     | 4782/11082 [15:22:03<20:22:07, 11.64s/step, acc=0.4861, loss=2.1411, lr=6.62e-05]
Training:  43%|████▎     | 4783/11082 [15:22:14<20:18:14, 11.60s/step, acc=0.4861, loss=2.1411, lr=6.62e-05]
Training:  43%|████▎     | 4784/11082 [15:22:26<20:18:55, 11.61s/step, acc=0.4861, loss=2.1411, lr=6.62e-05]
Training:  43%|████▎     | 4785/11082 [15:22:37<20:09:11, 11.52s/step, acc=0.4861, loss=2.1411, lr=6.62e-05]
Training:  43%|████▎     | 4786/11082 [15:22:49<20:13:19, 11.56s/step, acc=0.4861, loss=2.1411, lr=6.62e-05]
Training:  43%|████▎     | 4787/11082 [15:23:01<20:21:48, 11.65s/step, acc=0.4861, loss=2.1411, lr=6.62e-05]
Training:  43%|████▎     | 4788/11082 [15:23:12<20:00:44, 11.45s/step, acc=0.4861, loss=2.1411, lr=6.62e-05]
Training:  43%|████▎     | 4789/11082 [15:23:23<20:02:44, 11.47s/step, acc=0.4861, loss=2.1411, lr=6.62e-05]
Training:  43%|████▎     | 4790/11082 [15:23:35<20:08:04, 11.52s/step, acc=0.4861, loss=2.1411, lr=6.62e-05]
Training:  43%|████▎     | 4791/11082 [15:23:46<20:03:23, 11.48s/step, acc=0.4861, loss=2.1411, lr=6.62e-05]
Training:  43%|████▎     | 4792/11082 [15:23:58<20:21:25, 11.65s/step, acc=0.4861, loss=2.1411, lr=6.62e-05]
Training:  43%|████▎     | 4793/11082 [15:24:10<20:14:56, 11.59s/step, acc=0.4861, loss=2.1411, lr=6.62e-05]
Training:  43%|████▎     | 4794/11082 [15:24:21<20:04:59, 11.50s/step, acc=0.4861, loss=2.1411, lr=6.62e-05]
Training:  43%|████▎     | 4795/11082 [15:24:32<19:56:07, 11.42s/step, acc=0.4861, loss=2.1411, lr=6.62e-05]
Training:  43%|████▎     | 4796/11082 [15:24:44<20:14:20, 11.59s/step, acc=0.4861, loss=2.1411, lr=6.62e-05]
Training:  43%|████▎     | 4797/11082 [15:24:56<20:14:56, 11.60s/step, acc=0.4861, loss=2.1411, lr=6.62e-05]
Training:  43%|████▎     | 4798/11082 [15:25:07<20:00:36, 11.46s/step, acc=0.4861, loss=2.1411, lr=6.62e-05]
Training:  43%|████▎     | 4799/11082 [15:25:19<20:01:54, 11.48s/step, acc=0.4861, loss=2.1411, lr=6.62e-05]
Training:  43%|████▎     | 4800/11082 [15:25:31<20:13:27, 11.59s/step, acc=0.4861, loss=2.1411, lr=6.62e-05]
Training:  43%|████▎     | 4800/11082 [15:25:38<20:13:27, 11.59s/step, acc=0.4871, loss=2.1334, lr=6.56e-05]2026-06-19 06:19:05 | INFO | __main__ | epoch=0 step=4800 train_loss=2.133448 train_acc=0.487138 lr=6.557e-05 grad_norm=0.8125 acc_d0=0.7541 acc_d1=0.6036 acc_d2=0.4793 acc_d3=0.3785 acc_d4=0.2984 acc_d5=0.2396 acc_d6=0.1931 acc_d7=0.1610

Training:  43%|████▎     | 4801/11082 [15:25:42<20:05:20, 11.51s/step, acc=0.4871, loss=2.1334, lr=6.56e-05]
Training:  43%|████▎     | 4802/11082 [15:25:54<20:08:09, 11.54s/step, acc=0.4871, loss=2.1334, lr=6.56e-05]
Training:  43%|████▎     | 4803/11082 [15:26:05<19:57:30, 11.44s/step, acc=0.4871, loss=2.1334, lr=6.56e-05]
Training:  43%|████▎     | 4804/11082 [15:26:17<20:11:26, 11.58s/step, acc=0.4871, loss=2.1334, lr=6.56e-05]
Training:  43%|████▎     | 4805/11082 [15:26:28<20:05:55, 11.53s/step, acc=0.4871, loss=2.1334, lr=6.56e-05]
Training:  43%|████▎     | 4806/11082 [15:26:39<20:00:34, 11.48s/step, acc=0.4871, loss=2.1334, lr=6.56e-05]
Training:  43%|████▎     | 4807/11082 [15:26:51<20:01:28, 11.49s/step, acc=0.4871, loss=2.1334, lr=6.56e-05]
Training:  43%|████▎     | 4808/11082 [15:27:03<20:23:28, 11.70s/step, acc=0.4871, loss=2.1334, lr=6.56e-05]
Training:  43%|████▎     | 4809/11082 [15:27:15<20:16:53, 11.64s/step, acc=0.4871, loss=2.1334, lr=6.56e-05]
Training:  43%|████▎     | 4810/11082 [15:27:25<19:48:49, 11.37s/step, acc=0.4871, loss=2.1334, lr=6.56e-05]
Training:  43%|████▎     | 4811/11082 [15:27:37<20:06:38, 11.55s/step, acc=0.4871, loss=2.1334, lr=6.56e-05]
Training:  43%|████▎     | 4812/11082 [15:27:49<20:05:16, 11.53s/step, acc=0.4871, loss=2.1334, lr=6.56e-05]
Training:  43%|████▎     | 4813/11082 [15:28:00<20:07:41, 11.56s/step, acc=0.4871, loss=2.1334, lr=6.56e-05]
Training:  43%|████▎     | 4814/11082 [15:28:12<20:10:18, 11.59s/step, acc=0.4871, loss=2.1334, lr=6.56e-05]
Training:  43%|████▎     | 4815/11082 [15:28:24<20:09:51, 11.58s/step, acc=0.4871, loss=2.1334, lr=6.56e-05]
Training:  43%|████▎     | 4816/11082 [15:28:35<20:01:00, 11.50s/step, acc=0.4871, loss=2.1334, lr=6.56e-05]
Training:  43%|████▎     | 4817/11082 [15:28:46<19:56:12, 11.46s/step, acc=0.4871, loss=2.1334, lr=6.56e-05]
Training:  43%|████▎     | 4818/11082 [15:28:58<20:00:04, 11.49s/step, acc=0.4871, loss=2.1334, lr=6.56e-05]
Training:  43%|████▎     | 4819/11082 [15:29:10<20:12:23, 11.61s/step, acc=0.4871, loss=2.1334, lr=6.56e-05]
Training:  43%|████▎     | 4820/11082 [15:29:21<20:05:50, 11.55s/step, acc=0.4871, loss=2.1334, lr=6.56e-05]
Training:  44%|████▎     | 4821/11082 [15:29:33<20:00:38, 11.51s/step, acc=0.4871, loss=2.1334, lr=6.56e-05]
Training:  44%|████▎     | 4822/11082 [15:29:44<20:06:05, 11.56s/step, acc=0.4871, loss=2.1334, lr=6.56e-05]
Training:  44%|████▎     | 4823/11082 [15:29:56<20:11:32, 11.61s/step, acc=0.4871, loss=2.1334, lr=6.56e-05]
Training:  44%|████▎     | 4824/11082 [15:30:08<20:15:23, 11.65s/step, acc=0.4871, loss=2.1334, lr=6.56e-05]
Training:  44%|████▎     | 4825/11082 [15:30:19<20:14:36, 11.65s/step, acc=0.4871, loss=2.1334, lr=6.56e-05]
Training:  44%|████▎     | 4826/11082 [15:30:31<20:14:18, 11.65s/step, acc=0.4871, loss=2.1334, lr=6.56e-05]
Training:  44%|████▎     | 4827/11082 [15:30:43<20:13:33, 11.64s/step, acc=0.4871, loss=2.1334, lr=6.56e-05]
Training:  44%|████▎     | 4828/11082 [15:30:54<19:51:02, 11.43s/step, acc=0.4871, loss=2.1334, lr=6.56e-05]
Training:  44%|████▎     | 4829/11082 [15:31:06<20:11:18, 11.62s/step, acc=0.4871, loss=2.1334, lr=6.56e-05]
Training:  44%|████▎     | 4830/11082 [15:31:17<20:03:51, 11.55s/step, acc=0.4871, loss=2.1334, lr=6.56e-05]
Training:  44%|████▎     | 4831/11082 [15:31:29<20:16:10, 11.67s/step, acc=0.4871, loss=2.1334, lr=6.56e-05]
Training:  44%|████▎     | 4832/11082 [15:31:40<20:07:38, 11.59s/step, acc=0.4871, loss=2.1334, lr=6.56e-05]
Training:  44%|████▎     | 4833/11082 [15:31:52<20:14:35, 11.66s/step, acc=0.4871, loss=2.1334, lr=6.56e-05]
Training:  44%|████▎     | 4834/11082 [15:32:03<19:52:28, 11.45s/step, acc=0.4871, loss=2.1334, lr=6.56e-05]
Training:  44%|████▎     | 4835/11082 [15:32:16<20:19:20, 11.71s/step, acc=0.4871, loss=2.1334, lr=6.56e-05]
Training:  44%|████▎     | 4836/11082 [15:32:27<20:06:00, 11.59s/step, acc=0.4871, loss=2.1334, lr=6.56e-05]
Training:  44%|████▎     | 4837/11082 [15:32:39<20:11:53, 11.64s/step, acc=0.4871, loss=2.1334, lr=6.56e-05]
Training:  44%|████▎     | 4838/11082 [15:32:50<20:05:59, 11.59s/step, acc=0.4871, loss=2.1334, lr=6.56e-05]
Training:  44%|████▎     | 4839/11082 [15:33:02<20:02:06, 11.55s/step, acc=0.4871, loss=2.1334, lr=6.56e-05]
Training:  44%|████▎     | 4840/11082 [15:33:13<20:12:14, 11.65s/step, acc=0.4871, loss=2.1334, lr=6.56e-05]
Training:  44%|████▎     | 4841/11082 [15:33:25<20:21:09, 11.74s/step, acc=0.4871, loss=2.1334, lr=6.56e-05]
Training:  44%|████▎     | 4842/11082 [15:33:37<20:04:07, 11.58s/step, acc=0.4871, loss=2.1334, lr=6.56e-05]
Training:  44%|████▎     | 4843/11082 [15:33:48<20:05:10, 11.59s/step, acc=0.4871, loss=2.1334, lr=6.56e-05]
Training:  44%|████▎     | 4844/11082 [15:34:00<20:01:37, 11.56s/step, acc=0.4871, loss=2.1334, lr=6.56e-05]
Training:  44%|████▎     | 4845/11082 [15:34:11<19:49:33, 11.44s/step, acc=0.4871, loss=2.1334, lr=6.56e-05]
Training:  44%|████▎     | 4846/11082 [15:34:23<20:02:51, 11.57s/step, acc=0.4871, loss=2.1334, lr=6.56e-05]
Training:  44%|████▎     | 4847/11082 [15:34:34<19:59:06, 11.54s/step, acc=0.4871, loss=2.1334, lr=6.56e-05]
Training:  44%|████▎     | 4848/11082 [15:34:46<19:57:01, 11.52s/step, acc=0.4871, loss=2.1334, lr=6.56e-05]
Training:  44%|████▍     | 4849/11082 [15:34:57<19:49:43, 11.45s/step, acc=0.4871, loss=2.1334, lr=6.56e-05]
Training:  44%|████▍     | 4850/11082 [15:35:09<19:59:59, 11.55s/step, acc=0.4871, loss=2.1334, lr=6.56e-05]
Training:  44%|████▍     | 4850/11082 [15:35:16<19:59:59, 11.55s/step, acc=0.4894, loss=2.1230, lr=6.49e-05]2026-06-19 06:28:44 | INFO | __main__ | epoch=0 step=4850 train_loss=2.123006 train_acc=0.489383 lr=6.494e-05 grad_norm=0.8828 acc_d0=0.7489 acc_d1=0.6023 acc_d2=0.4809 acc_d3=0.3820 acc_d4=0.3053 acc_d5=0.2464 acc_d6=0.2005 acc_d7=0.1660

Training:  44%|████▍     | 4851/11082 [15:35:21<20:12:09, 11.67s/step, acc=0.4894, loss=2.1230, lr=6.49e-05]
Training:  44%|████▍     | 4852/11082 [15:35:32<19:54:40, 11.51s/step, acc=0.4894, loss=2.1230, lr=6.49e-05]
Training:  44%|████▍     | 4853/11082 [15:35:44<20:02:11, 11.58s/step, acc=0.4894, loss=2.1230, lr=6.49e-05]
Training:  44%|████▍     | 4854/11082 [15:35:55<20:02:06, 11.58s/step, acc=0.4894, loss=2.1230, lr=6.49e-05]
Training:  44%|████▍     | 4855/11082 [15:36:06<19:52:21, 11.49s/step, acc=0.4894, loss=2.1230, lr=6.49e-05]
Training:  44%|████▍     | 4856/11082 [15:36:18<19:54:00, 11.51s/step, acc=0.4894, loss=2.1230, lr=6.49e-05]
Training:  44%|████▍     | 4857/11082 [15:36:30<20:17:56, 11.74s/step, acc=0.4894, loss=2.1230, lr=6.49e-05]
Training:  44%|████▍     | 4858/11082 [15:36:41<20:00:56, 11.58s/step, acc=0.4894, loss=2.1230, lr=6.49e-05]
Training:  44%|████▍     | 4859/11082 [15:36:52<19:43:46, 11.41s/step, acc=0.4894, loss=2.1230, lr=6.49e-05]
Training:  44%|████▍     | 4860/11082 [15:37:04<19:51:50, 11.49s/step, acc=0.4894, loss=2.1230, lr=6.49e-05]
Training:  44%|████▍     | 4861/11082 [15:37:16<20:04:06, 11.61s/step, acc=0.4894, loss=2.1230, lr=6.49e-05]
Training:  44%|████▍     | 4862/11082 [15:37:27<19:55:23, 11.53s/step, acc=0.4894, loss=2.1230, lr=6.49e-05]
Training:  44%|████▍     | 4863/11082 [15:37:39<19:48:17, 11.46s/step, acc=0.4894, loss=2.1230, lr=6.49e-05]
Training:  44%|████▍     | 4864/11082 [15:37:51<20:03:54, 11.62s/step, acc=0.4894, loss=2.1230, lr=6.49e-05]
Training:  44%|████▍     | 4865/11082 [15:38:02<20:09:31, 11.67s/step, acc=0.4894, loss=2.1230, lr=6.49e-05]
Training:  44%|████▍     | 4866/11082 [15:38:14<19:56:06, 11.55s/step, acc=0.4894, loss=2.1230, lr=6.49e-05]
Training:  44%|████▍     | 4867/11082 [15:38:25<20:00:01, 11.59s/step, acc=0.4894, loss=2.1230, lr=6.49e-05]
Training:  44%|████▍     | 4868/11082 [15:38:37<19:53:04, 11.52s/step, acc=0.4894, loss=2.1230, lr=6.49e-05]
Training:  44%|████▍     | 4869/11082 [15:38:48<19:48:33, 11.48s/step, acc=0.4894, loss=2.1230, lr=6.49e-05]
Training:  44%|████▍     | 4870/11082 [15:39:00<20:04:34, 11.63s/step, acc=0.4894, loss=2.1230, lr=6.49e-05]
Training:  44%|████▍     | 4871/11082 [15:39:11<19:53:06, 11.53s/step, acc=0.4894, loss=2.1230, lr=6.49e-05]
Training:  44%|████▍     | 4872/11082 [15:39:23<20:00:12, 11.60s/step, acc=0.4894, loss=2.1230, lr=6.49e-05]
Training:  44%|████▍     | 4873/11082 [15:39:35<19:52:03, 11.52s/step, acc=0.4894, loss=2.1230, lr=6.49e-05]
Training:  44%|████▍     | 4874/11082 [15:39:46<19:59:21, 11.59s/step, acc=0.4894, loss=2.1230, lr=6.49e-05]
Training:  44%|████▍     | 4875/11082 [15:39:58<19:59:16, 11.59s/step, acc=0.4894, loss=2.1230, lr=6.49e-05]
Training:  44%|████▍     | 4876/11082 [15:40:10<20:01:08, 11.61s/step, acc=0.4894, loss=2.1230, lr=6.49e-05]
Training:  44%|████▍     | 4877/11082 [15:40:21<19:49:38, 11.50s/step, acc=0.4894, loss=2.1230, lr=6.49e-05]
Training:  44%|████▍     | 4878/11082 [15:40:32<19:51:20, 11.52s/step, acc=0.4894, loss=2.1230, lr=6.49e-05]
Training:  44%|████▍     | 4879/11082 [15:40:44<19:41:36, 11.43s/step, acc=0.4894, loss=2.1230, lr=6.49e-05]
Training:  44%|████▍     | 4880/11082 [15:40:55<19:49:19, 11.51s/step, acc=0.4894, loss=2.1230, lr=6.49e-05]
Training:  44%|████▍     | 4881/11082 [15:41:07<19:44:57, 11.47s/step, acc=0.4894, loss=2.1230, lr=6.49e-05]
Training:  44%|████▍     | 4882/11082 [15:41:18<19:42:35, 11.44s/step, acc=0.4894, loss=2.1230, lr=6.49e-05]
Training:  44%|████▍     | 4883/11082 [15:41:30<19:47:42, 11.50s/step, acc=0.4894, loss=2.1230, lr=6.49e-05]
Training:  44%|████▍     | 4884/11082 [15:41:41<19:43:31, 11.46s/step, acc=0.4894, loss=2.1230, lr=6.49e-05]
Training:  44%|████▍     | 4885/11082 [15:41:53<19:45:32, 11.48s/step, acc=0.4894, loss=2.1230, lr=6.49e-05]
Training:  44%|████▍     | 4886/11082 [15:42:04<19:53:25, 11.56s/step, acc=0.4894, loss=2.1230, lr=6.49e-05]
Training:  44%|████▍     | 4887/11082 [15:42:16<19:50:31, 11.53s/step, acc=0.4894, loss=2.1230, lr=6.49e-05]
Training:  44%|████▍     | 4888/11082 [15:42:27<19:50:38, 11.53s/step, acc=0.4894, loss=2.1230, lr=6.49e-05]
Training:  44%|████▍     | 4889/11082 [15:42:39<19:48:18, 11.51s/step, acc=0.4894, loss=2.1230, lr=6.49e-05]
Training:  44%|████▍     | 4890/11082 [15:42:50<19:49:34, 11.53s/step, acc=0.4894, loss=2.1230, lr=6.49e-05]
Training:  44%|████▍     | 4891/11082 [15:43:02<19:43:55, 11.47s/step, acc=0.4894, loss=2.1230, lr=6.49e-05]
Training:  44%|████▍     | 4892/11082 [15:43:14<20:01:09, 11.64s/step, acc=0.4894, loss=2.1230, lr=6.49e-05]
Training:  44%|████▍     | 4893/11082 [15:43:25<19:54:21, 11.58s/step, acc=0.4894, loss=2.1230, lr=6.49e-05]
Training:  44%|████▍     | 4894/11082 [15:43:37<19:51:55, 11.56s/step, acc=0.4894, loss=2.1230, lr=6.49e-05]
Training:  44%|████▍     | 4895/11082 [15:43:48<19:32:32, 11.37s/step, acc=0.4894, loss=2.1230, lr=6.49e-05]
Training:  44%|████▍     | 4896/11082 [15:44:00<19:58:19, 11.62s/step, acc=0.4894, loss=2.1230, lr=6.49e-05]
Training:  44%|████▍     | 4897/11082 [15:44:11<19:37:51, 11.43s/step, acc=0.4894, loss=2.1230, lr=6.49e-05]
Training:  44%|████▍     | 4898/11082 [15:44:22<19:47:24, 11.52s/step, acc=0.4894, loss=2.1230, lr=6.49e-05]
Training:  44%|████▍     | 4899/11082 [15:44:34<19:47:16, 11.52s/step, acc=0.4894, loss=2.1230, lr=6.49e-05]
Training:  44%|████▍     | 4900/11082 [15:44:46<19:51:46, 11.57s/step, acc=0.4894, loss=2.1230, lr=6.49e-05]
Training:  44%|████▍     | 4900/11082 [15:44:53<19:51:46, 11.57s/step, acc=0.4901, loss=2.1163, lr=6.43e-05]2026-06-19 06:38:21 | INFO | __main__ | epoch=0 step=4900 train_loss=2.116301 train_acc=0.490143 lr=6.431e-05 grad_norm=0.9258 acc_d0=0.7597 acc_d1=0.6149 acc_d2=0.4927 acc_d3=0.3924 acc_d4=0.3139 acc_d5=0.2520 acc_d6=0.2035 acc_d7=0.1676

Training:  44%|████▍     | 4901/11082 [15:44:57<19:57:46, 11.63s/step, acc=0.4901, loss=2.1163, lr=6.43e-05]
Training:  44%|████▍     | 4902/11082 [15:45:09<19:48:21, 11.54s/step, acc=0.4901, loss=2.1163, lr=6.43e-05]
Training:  44%|████▍     | 4903/11082 [15:45:20<19:48:20, 11.54s/step, acc=0.4901, loss=2.1163, lr=6.43e-05]
Training:  44%|████▍     | 4904/11082 [15:45:32<20:05:09, 11.70s/step, acc=0.4901, loss=2.1163, lr=6.43e-05]
Training:  44%|████▍     | 4905/11082 [15:45:44<20:00:20, 11.66s/step, acc=0.4901, loss=2.1163, lr=6.43e-05]
Training:  44%|████▍     | 4906/11082 [15:45:56<20:05:22, 11.71s/step, acc=0.4901, loss=2.1163, lr=6.43e-05]
Training:  44%|████▍     | 4907/11082 [15:46:07<19:53:06, 11.59s/step, acc=0.4901, loss=2.1163, lr=6.43e-05]
Training:  44%|████▍     | 4908/11082 [15:46:19<20:05:11, 11.71s/step, acc=0.4901, loss=2.1163, lr=6.43e-05]
Training:  44%|████▍     | 4909/11082 [15:46:30<19:40:21, 11.47s/step, acc=0.4901, loss=2.1163, lr=6.43e-05]
Training:  44%|████▍     | 4910/11082 [15:46:42<20:02:22, 11.69s/step, acc=0.4901, loss=2.1163, lr=6.43e-05]
Training:  44%|████▍     | 4911/11082 [15:46:53<19:41:41, 11.49s/step, acc=0.4901, loss=2.1163, lr=6.43e-05]
Training:  44%|████▍     | 4912/11082 [15:47:05<19:50:43, 11.58s/step, acc=0.4901, loss=2.1163, lr=6.43e-05]
Training:  44%|████▍     | 4913/11082 [15:47:17<20:01:00, 11.68s/step, acc=0.4901, loss=2.1163, lr=6.43e-05]
Training:  44%|████▍     | 4914/11082 [15:47:28<19:40:47, 11.49s/step, acc=0.4901, loss=2.1163, lr=6.43e-05]
Training:  44%|████▍     | 4915/11082 [15:47:40<19:45:20, 11.53s/step, acc=0.4901, loss=2.1163, lr=6.43e-05]
Training:  44%|████▍     | 4916/11082 [15:47:52<20:09:36, 11.77s/step, acc=0.4901, loss=2.1163, lr=6.43e-05]
Training:  44%|████▍     | 4917/11082 [15:48:03<19:59:20, 11.67s/step, acc=0.4901, loss=2.1163, lr=6.43e-05]
Training:  44%|████▍     | 4918/11082 [15:48:15<19:52:27, 11.61s/step, acc=0.4901, loss=2.1163, lr=6.43e-05]
Training:  44%|████▍     | 4919/11082 [15:48:26<19:43:59, 11.53s/step, acc=0.4901, loss=2.1163, lr=6.43e-05]
Training:  44%|████▍     | 4920/11082 [15:48:38<19:53:23, 11.62s/step, acc=0.4901, loss=2.1163, lr=6.43e-05]
Training:  44%|████▍     | 4921/11082 [15:48:50<19:52:28, 11.61s/step, acc=0.4901, loss=2.1163, lr=6.43e-05]
Training:  44%|████▍     | 4922/11082 [15:49:01<19:49:15, 11.58s/step, acc=0.4901, loss=2.1163, lr=6.43e-05]
Training:  44%|████▍     | 4923/11082 [15:49:13<19:48:16, 11.58s/step, acc=0.4901, loss=2.1163, lr=6.43e-05]
Training:  44%|████▍     | 4924/11082 [15:49:24<19:49:00, 11.58s/step, acc=0.4901, loss=2.1163, lr=6.43e-05]
Training:  44%|████▍     | 4925/11082 [15:49:36<19:54:30, 11.64s/step, acc=0.4901, loss=2.1163, lr=6.43e-05]
Training:  44%|████▍     | 4926/11082 [15:49:47<19:40:59, 11.51s/step, acc=0.4901, loss=2.1163, lr=6.43e-05]
Training:  44%|████▍     | 4927/11082 [15:49:59<19:34:21, 11.45s/step, acc=0.4901, loss=2.1163, lr=6.43e-05]
Training:  44%|████▍     | 4928/11082 [15:50:11<19:50:56, 11.61s/step, acc=0.4901, loss=2.1163, lr=6.43e-05]
Training:  44%|████▍     | 4929/11082 [15:50:22<19:46:39, 11.57s/step, acc=0.4901, loss=2.1163, lr=6.43e-05]
Training:  44%|████▍     | 4930/11082 [15:50:34<19:52:25, 11.63s/step, acc=0.4901, loss=2.1163, lr=6.43e-05]
Training:  44%|████▍     | 4931/11082 [15:50:46<19:58:29, 11.69s/step, acc=0.4901, loss=2.1163, lr=6.43e-05]
Training:  45%|████▍     | 4932/11082 [15:50:57<19:34:18, 11.46s/step, acc=0.4901, loss=2.1163, lr=6.43e-05]
Training:  45%|████▍     | 4933/11082 [15:51:09<19:51:03, 11.62s/step, acc=0.4901, loss=2.1163, lr=6.43e-05]
Training:  45%|████▍     | 4934/11082 [15:51:20<19:44:57, 11.56s/step, acc=0.4901, loss=2.1163, lr=6.43e-05]
Training:  45%|████▍     | 4935/11082 [15:51:31<19:36:27, 11.48s/step, acc=0.4901, loss=2.1163, lr=6.43e-05]
Training:  45%|████▍     | 4936/11082 [15:51:43<19:43:03, 11.55s/step, acc=0.4901, loss=2.1163, lr=6.43e-05]
Training:  45%|████▍     | 4937/11082 [15:51:55<19:46:15, 11.58s/step, acc=0.4901, loss=2.1163, lr=6.43e-05]
Training:  45%|████▍     | 4938/11082 [15:52:06<19:41:09, 11.53s/step, acc=0.4901, loss=2.1163, lr=6.43e-05]
Training:  45%|████▍     | 4939/11082 [15:52:17<19:37:55, 11.51s/step, acc=0.4901, loss=2.1163, lr=6.43e-05]
Training:  45%|████▍     | 4940/11082 [15:52:29<19:45:11, 11.58s/step, acc=0.4901, loss=2.1163, lr=6.43e-05]
Training:  45%|████▍     | 4941/11082 [15:52:40<19:35:16, 11.48s/step, acc=0.4901, loss=2.1163, lr=6.43e-05]
Training:  45%|████▍     | 4942/11082 [15:52:52<19:41:56, 11.55s/step, acc=0.4901, loss=2.1163, lr=6.43e-05]
Training:  45%|████▍     | 4943/11082 [15:53:04<19:41:30, 11.55s/step, acc=0.4901, loss=2.1163, lr=6.43e-05]
Training:  45%|████▍     | 4944/11082 [15:53:15<19:33:57, 11.48s/step, acc=0.4901, loss=2.1163, lr=6.43e-05]
Training:  45%|████▍     | 4945/11082 [15:53:27<19:47:25, 11.61s/step, acc=0.4901, loss=2.1163, lr=6.43e-05]
Training:  45%|████▍     | 4946/11082 [15:53:38<19:35:55, 11.50s/step, acc=0.4901, loss=2.1163, lr=6.43e-05]
Training:  45%|████▍     | 4947/11082 [15:53:50<19:43:25, 11.57s/step, acc=0.4901, loss=2.1163, lr=6.43e-05]
Training:  45%|████▍     | 4948/11082 [15:54:02<19:43:05, 11.57s/step, acc=0.4901, loss=2.1163, lr=6.43e-05]
Training:  45%|████▍     | 4949/11082 [15:54:13<19:46:43, 11.61s/step, acc=0.4901, loss=2.1163, lr=6.43e-05]
Training:  45%|████▍     | 4950/11082 [15:54:25<19:38:41, 11.53s/step, acc=0.4901, loss=2.1163, lr=6.43e-05]
Training:  45%|████▍     | 4950/11082 [15:54:32<19:38:41, 11.53s/step, acc=0.4904, loss=2.1186, lr=6.37e-05]2026-06-19 06:48:00 | INFO | __main__ | epoch=0 step=4950 train_loss=2.118603 train_acc=0.490436 lr=6.367e-05 grad_norm=0.8594 acc_d0=0.7513 acc_d1=0.6012 acc_d2=0.4775 acc_d3=0.3781 acc_d4=0.3003 acc_d5=0.2412 acc_d6=0.1960 acc_d7=0.1605

Training:  45%|████▍     | 4951/11082 [15:54:36<19:45:43, 11.60s/step, acc=0.4904, loss=2.1186, lr=6.37e-05]
Training:  45%|████▍     | 4952/11082 [15:54:48<19:39:03, 11.54s/step, acc=0.4904, loss=2.1186, lr=6.37e-05]
Training:  45%|████▍     | 4953/11082 [15:54:59<19:41:58, 11.57s/step, acc=0.4904, loss=2.1186, lr=6.37e-05]
Training:  45%|████▍     | 4954/11082 [15:55:11<19:43:30, 11.59s/step, acc=0.4904, loss=2.1186, lr=6.37e-05]
Training:  45%|████▍     | 4955/11082 [15:55:22<19:23:40, 11.40s/step, acc=0.4904, loss=2.1186, lr=6.37e-05]
Training:  45%|████▍     | 4956/11082 [15:55:33<19:21:24, 11.38s/step, acc=0.4904, loss=2.1186, lr=6.37e-05]
Training:  45%|████▍     | 4957/11082 [15:55:45<19:41:09, 11.57s/step, acc=0.4904, loss=2.1186, lr=6.37e-05]
Training:  45%|████▍     | 4958/11082 [15:55:57<19:37:52, 11.54s/step, acc=0.4904, loss=2.1186, lr=6.37e-05]
Training:  45%|████▍     | 4959/11082 [15:56:08<19:42:53, 11.59s/step, acc=0.4904, loss=2.1186, lr=6.37e-05]
Training:  45%|████▍     | 4960/11082 [15:56:20<19:32:35, 11.49s/step, acc=0.4904, loss=2.1186, lr=6.37e-05]
Training:  45%|████▍     | 4961/11082 [15:56:32<19:44:19, 11.61s/step, acc=0.4904, loss=2.1186, lr=6.37e-05]
Training:  45%|████▍     | 4962/11082 [15:56:43<19:37:47, 11.55s/step, acc=0.4904, loss=2.1186, lr=6.37e-05]
Training:  45%|████▍     | 4963/11082 [15:56:55<19:35:20, 11.52s/step, acc=0.4904, loss=2.1186, lr=6.37e-05]
Training:  45%|████▍     | 4964/11082 [15:57:06<19:29:34, 11.47s/step, acc=0.4904, loss=2.1186, lr=6.37e-05]
Training:  45%|████▍     | 4965/11082 [15:57:17<19:30:44, 11.48s/step, acc=0.4904, loss=2.1186, lr=6.37e-05]
Training:  45%|████▍     | 4966/11082 [15:57:29<19:37:54, 11.56s/step, acc=0.4904, loss=2.1186, lr=6.37e-05]
Training:  45%|████▍     | 4967/11082 [15:57:41<19:38:15, 11.56s/step, acc=0.4904, loss=2.1186, lr=6.37e-05]
Training:  45%|████▍     | 4968/11082 [15:57:53<19:48:23, 11.66s/step, acc=0.4904, loss=2.1186, lr=6.37e-05]
Training:  45%|████▍     | 4969/11082 [15:58:04<19:25:55, 11.44s/step, acc=0.4904, loss=2.1186, lr=6.37e-05]
Training:  45%|████▍     | 4970/11082 [15:58:15<19:38:30, 11.57s/step, acc=0.4904, loss=2.1186, lr=6.37e-05]
Training:  45%|████▍     | 4971/11082 [15:58:27<19:37:09, 11.56s/step, acc=0.4904, loss=2.1186, lr=6.37e-05]
Training:  45%|████▍     | 4972/11082 [15:58:38<19:30:44, 11.50s/step, acc=0.4904, loss=2.1186, lr=6.37e-05]
Training:  45%|████▍     | 4973/11082 [15:58:50<19:27:48, 11.47s/step, acc=0.4904, loss=2.1186, lr=6.37e-05]
Training:  45%|████▍     | 4974/11082 [15:59:02<19:43:11, 11.62s/step, acc=0.4904, loss=2.1186, lr=6.37e-05]
Training:  45%|████▍     | 4975/11082 [15:59:12<19:19:16, 11.39s/step, acc=0.4904, loss=2.1186, lr=6.37e-05]
Training:  45%|████▍     | 4976/11082 [15:59:24<19:31:10, 11.51s/step, acc=0.4904, loss=2.1186, lr=6.37e-05]
Training:  45%|████▍     | 4977/11082 [15:59:36<19:43:50, 11.63s/step, acc=0.4904, loss=2.1186, lr=6.37e-05]
Training:  45%|████▍     | 4978/11082 [15:59:48<19:44:23, 11.64s/step, acc=0.4904, loss=2.1186, lr=6.37e-05]
Training:  45%|████▍     | 4979/11082 [15:59:59<19:29:27, 11.50s/step, acc=0.4904, loss=2.1186, lr=6.37e-05]
Training:  45%|████▍     | 4980/11082 [16:00:11<19:30:46, 11.51s/step, acc=0.4904, loss=2.1186, lr=6.37e-05]
Training:  45%|████▍     | 4981/11082 [16:00:22<19:23:20, 11.44s/step, acc=0.4904, loss=2.1186, lr=6.37e-05]
Training:  45%|████▍     | 4982/11082 [16:00:33<19:21:07, 11.42s/step, acc=0.4904, loss=2.1186, lr=6.37e-05]
Training:  45%|████▍     | 4983/11082 [16:00:45<19:36:54, 11.58s/step, acc=0.4904, loss=2.1186, lr=6.37e-05]
Training:  45%|████▍     | 4984/11082 [16:00:57<19:35:33, 11.57s/step, acc=0.4904, loss=2.1186, lr=6.37e-05]
Training:  45%|████▍     | 4985/11082 [16:01:08<19:31:45, 11.53s/step, acc=0.4904, loss=2.1186, lr=6.37e-05]
Training:  45%|████▍     | 4986/11082 [16:01:20<19:38:22, 11.60s/step, acc=0.4904, loss=2.1186, lr=6.37e-05]
Training:  45%|████▌     | 4987/11082 [16:01:32<19:40:23, 11.62s/step, acc=0.4904, loss=2.1186, lr=6.37e-05]
Training:  45%|████▌     | 4988/11082 [16:01:43<19:43:11, 11.65s/step, acc=0.4904, loss=2.1186, lr=6.37e-05]
Training:  45%|████▌     | 4989/11082 [16:01:54<19:24:53, 11.47s/step, acc=0.4904, loss=2.1186, lr=6.37e-05]
Training:  45%|████▌     | 4990/11082 [16:02:06<19:42:10, 11.64s/step, acc=0.4904, loss=2.1186, lr=6.37e-05]
Training:  45%|████▌     | 4991/11082 [16:02:18<19:40:10, 11.63s/step, acc=0.4904, loss=2.1186, lr=6.37e-05]
Training:  45%|████▌     | 4992/11082 [16:02:30<19:45:21, 11.68s/step, acc=0.4904, loss=2.1186, lr=6.37e-05]
Training:  45%|████▌     | 4993/11082 [16:02:41<19:38:09, 11.61s/step, acc=0.4904, loss=2.1186, lr=6.37e-05]
Training:  45%|████▌     | 4994/11082 [16:02:53<19:37:49, 11.61s/step, acc=0.4904, loss=2.1186, lr=6.37e-05]
Training:  45%|████▌     | 4995/11082 [16:03:04<19:30:11, 11.53s/step, acc=0.4904, loss=2.1186, lr=6.37e-05]
Training:  45%|████▌     | 4996/11082 [16:03:16<19:41:08, 11.64s/step, acc=0.4904, loss=2.1186, lr=6.37e-05]
Training:  45%|████▌     | 4997/11082 [16:03:28<19:35:18, 11.59s/step, acc=0.4904, loss=2.1186, lr=6.37e-05]
Training:  45%|████▌     | 4998/11082 [16:03:39<19:43:20, 11.67s/step, acc=0.4904, loss=2.1186, lr=6.37e-05]
Training:  45%|████▌     | 4999/11082 [16:03:51<19:45:19, 11.69s/step, acc=0.4904, loss=2.1186, lr=6.37e-05]
Training:  45%|████▌     | 5000/11082 [16:04:03<19:41:12, 11.65s/step, acc=0.4904, loss=2.1186, lr=6.37e-05]
Training:  45%|████▌     | 5000/11082 [16:04:10<19:41:12, 11.65s/step, acc=0.4913, loss=2.1111, lr=6.30e-05]2026-06-19 06:57:38 | INFO | __main__ | epoch=0 step=5000 train_loss=2.111106 train_acc=0.491343 lr=6.304e-05 grad_norm=0.8359 acc_d0=0.7576 acc_d1=0.6080 acc_d2=0.4842 acc_d3=0.3834 acc_d4=0.3058 acc_d5=0.2435 acc_d6=0.1963 acc_d7=0.1624

Training:  45%|████▌     | 5001/11082 [16:04:14<19:37:54, 11.62s/step, acc=0.4913, loss=2.1111, lr=6.30e-05]
Training:  45%|████▌     | 5002/11082 [16:04:26<19:37:30, 11.62s/step, acc=0.4913, loss=2.1111, lr=6.30e-05]
Training:  45%|████▌     | 5003/11082 [16:04:37<19:29:07, 11.54s/step, acc=0.4913, loss=2.1111, lr=6.30e-05]
Training:  45%|████▌     | 5004/11082 [16:04:48<19:14:34, 11.40s/step, acc=0.4913, loss=2.1111, lr=6.30e-05]
Training:  45%|████▌     | 5005/11082 [16:05:00<19:25:51, 11.51s/step, acc=0.4913, loss=2.1111, lr=6.30e-05]
Training:  45%|████▌     | 5006/11082 [16:05:11<19:20:22, 11.46s/step, acc=0.4913, loss=2.1111, lr=6.30e-05]
Training:  45%|████▌     | 5007/11082 [16:05:23<19:24:43, 11.50s/step, acc=0.4913, loss=2.1111, lr=6.30e-05]
Training:  45%|████▌     | 5008/11082 [16:05:35<19:36:20, 11.62s/step, acc=0.4913, loss=2.1111, lr=6.30e-05]
Training:  45%|████▌     | 5009/11082 [16:05:46<19:34:30, 11.60s/step, acc=0.4913, loss=2.1111, lr=6.30e-05]
Training:  45%|████▌     | 5010/11082 [16:05:58<19:32:09, 11.58s/step, acc=0.4913, loss=2.1111, lr=6.30e-05]
Training:  45%|████▌     | 5011/11082 [16:06:10<19:32:32, 11.59s/step, acc=0.4913, loss=2.1111, lr=6.30e-05]
Training:  45%|████▌     | 5012/11082 [16:06:21<19:23:39, 11.50s/step, acc=0.4913, loss=2.1111, lr=6.30e-05]
Training:  45%|████▌     | 5013/11082 [16:06:33<19:38:14, 11.65s/step, acc=0.4913, loss=2.1111, lr=6.30e-05]
Training:  45%|████▌     | 5014/11082 [16:06:44<19:19:53, 11.47s/step, acc=0.4913, loss=2.1111, lr=6.30e-05]
Training:  45%|████▌     | 5015/11082 [16:06:55<19:15:37, 11.43s/step, acc=0.4913, loss=2.1111, lr=6.30e-05]
Training:  45%|████▌     | 5016/11082 [16:07:08<19:43:45, 11.71s/step, acc=0.4913, loss=2.1111, lr=6.30e-05]
Training:  45%|████▌     | 5017/11082 [16:07:19<19:40:42, 11.68s/step, acc=0.4913, loss=2.1111, lr=6.30e-05]
Training:  45%|████▌     | 5018/11082 [16:07:31<19:37:29, 11.65s/step, acc=0.4913, loss=2.1111, lr=6.30e-05]
Training:  45%|████▌     | 5019/11082 [16:07:42<19:32:30, 11.60s/step, acc=0.4913, loss=2.1111, lr=6.30e-05]
Training:  45%|████▌     | 5020/11082 [16:07:54<19:42:08, 11.70s/step, acc=0.4913, loss=2.1111, lr=6.30e-05]
Training:  45%|████▌     | 5021/11082 [16:08:06<19:33:00, 11.61s/step, acc=0.4913, loss=2.1111, lr=6.30e-05]
Training:  45%|████▌     | 5022/11082 [16:08:17<19:23:16, 11.52s/step, acc=0.4913, loss=2.1111, lr=6.30e-05]
Training:  45%|████▌     | 5023/11082 [16:08:29<19:30:31, 11.59s/step, acc=0.4913, loss=2.1111, lr=6.30e-05]
Training:  45%|████▌     | 5024/11082 [16:08:40<19:32:04, 11.61s/step, acc=0.4913, loss=2.1111, lr=6.30e-05]
Training:  45%|████▌     | 5025/11082 [16:08:52<19:30:01, 11.59s/step, acc=0.4913, loss=2.1111, lr=6.30e-05]
Training:  45%|████▌     | 5026/11082 [16:09:03<19:19:15, 11.49s/step, acc=0.4913, loss=2.1111, lr=6.30e-05]
Training:  45%|████▌     | 5027/11082 [16:09:15<19:22:48, 11.52s/step, acc=0.4913, loss=2.1111, lr=6.30e-05]
Training:  45%|████▌     | 5028/11082 [16:09:26<19:15:50, 11.46s/step, acc=0.4913, loss=2.1111, lr=6.30e-05]
Training:  45%|████▌     | 5029/11082 [16:09:38<19:25:09, 11.55s/step, acc=0.4913, loss=2.1111, lr=6.30e-05]
Training:  45%|████▌     | 5030/11082 [16:09:49<19:14:23, 11.44s/step, acc=0.4913, loss=2.1111, lr=6.30e-05]
Training:  45%|████▌     | 5031/11082 [16:10:01<19:27:57, 11.58s/step, acc=0.4913, loss=2.1111, lr=6.30e-05]
Training:  45%|████▌     | 5032/11082 [16:10:13<19:29:38, 11.60s/step, acc=0.4913, loss=2.1111, lr=6.30e-05]
Training:  45%|████▌     | 5033/11082 [16:10:24<19:22:06, 11.53s/step, acc=0.4913, loss=2.1111, lr=6.30e-05]
Training:  45%|████▌     | 5034/11082 [16:10:35<19:20:36, 11.51s/step, acc=0.4913, loss=2.1111, lr=6.30e-05]
Training:  45%|████▌     | 5035/11082 [16:10:46<19:06:22, 11.37s/step, acc=0.4913, loss=2.1111, lr=6.30e-05]
Training:  45%|████▌     | 5036/11082 [16:10:58<19:14:30, 11.46s/step, acc=0.4913, loss=2.1111, lr=6.30e-05]
Training:  45%|████▌     | 5037/11082 [16:11:10<19:25:17, 11.57s/step, acc=0.4913, loss=2.1111, lr=6.30e-05]
Training:  45%|████▌     | 5038/11082 [16:11:21<19:20:04, 11.52s/step, acc=0.4913, loss=2.1111, lr=6.30e-05]
Training:  45%|████▌     | 5039/11082 [16:11:33<19:30:17, 11.62s/step, acc=0.4913, loss=2.1111, lr=6.30e-05]
Training:  45%|████▌     | 5040/11082 [16:11:45<19:30:44, 11.63s/step, acc=0.4913, loss=2.1111, lr=6.30e-05]
Training:  45%|████▌     | 5041/11082 [16:11:56<19:19:08, 11.51s/step, acc=0.4913, loss=2.1111, lr=6.30e-05]
Training:  45%|████▌     | 5042/11082 [16:12:08<19:30:19, 11.63s/step, acc=0.4913, loss=2.1111, lr=6.30e-05]
Training:  46%|████▌     | 5043/11082 [16:12:20<19:29:06, 11.62s/step, acc=0.4913, loss=2.1111, lr=6.30e-05]
Training:  46%|████▌     | 5044/11082 [16:12:31<19:35:17, 11.68s/step, acc=0.4913, loss=2.1111, lr=6.30e-05]
Training:  46%|████▌     | 5045/11082 [16:12:43<19:27:58, 11.61s/step, acc=0.4913, loss=2.1111, lr=6.30e-05]
Training:  46%|████▌     | 5046/11082 [16:12:55<19:30:33, 11.64s/step, acc=0.4913, loss=2.1111, lr=6.30e-05]
Training:  46%|████▌     | 5047/11082 [16:13:05<19:03:33, 11.37s/step, acc=0.4913, loss=2.1111, lr=6.30e-05]
Training:  46%|████▌     | 5048/11082 [16:13:17<19:15:09, 11.49s/step, acc=0.4913, loss=2.1111, lr=6.30e-05]
Training:  46%|████▌     | 5049/11082 [16:13:29<19:29:54, 11.64s/step, acc=0.4913, loss=2.1111, lr=6.30e-05]
Training:  46%|████▌     | 5050/11082 [16:13:41<19:26:55, 11.61s/step, acc=0.4913, loss=2.1111, lr=6.30e-05]
Training:  46%|████▌     | 5050/11082 [16:13:47<19:26:55, 11.61s/step, acc=0.4918, loss=2.1059, lr=6.24e-05]2026-06-19 07:07:15 | INFO | __main__ | epoch=0 step=5050 train_loss=2.105939 train_acc=0.491831 lr=6.240e-05 grad_norm=0.8711 acc_d0=0.7532 acc_d1=0.6067 acc_d2=0.4843 acc_d3=0.3854 acc_d4=0.3080 acc_d5=0.2469 acc_d6=0.2000 acc_d7=0.1653

Training:  46%|████▌     | 5051/11082 [16:13:52<19:19:12, 11.53s/step, acc=0.4918, loss=2.1059, lr=6.24e-05]
Training:  46%|████▌     | 5052/11082 [16:14:04<19:20:00, 11.54s/step, acc=0.4918, loss=2.1059, lr=6.24e-05]
Training:  46%|████▌     | 5053/11082 [16:14:15<19:27:43, 11.62s/step, acc=0.4918, loss=2.1059, lr=6.24e-05]
Training:  46%|████▌     | 5054/11082 [16:14:27<19:24:01, 11.59s/step, acc=0.4918, loss=2.1059, lr=6.24e-05]
Training:  46%|████▌     | 5055/11082 [16:14:39<19:31:04, 11.66s/step, acc=0.4918, loss=2.1059, lr=6.24e-05]
Training:  46%|████▌     | 5056/11082 [16:14:50<19:31:47, 11.67s/step, acc=0.4918, loss=2.1059, lr=6.24e-05]
Training:  46%|████▌     | 5057/11082 [16:15:02<19:18:02, 11.53s/step, acc=0.4918, loss=2.1059, lr=6.24e-05]
Training:  46%|████▌     | 5058/11082 [16:15:13<19:21:38, 11.57s/step, acc=0.4918, loss=2.1059, lr=6.24e-05]
Training:  46%|████▌     | 5059/11082 [16:15:25<19:20:02, 11.56s/step, acc=0.4918, loss=2.1059, lr=6.24e-05]
Training:  46%|████▌     | 5060/11082 [16:15:36<19:09:04, 11.45s/step, acc=0.4918, loss=2.1059, lr=6.24e-05]
Training:  46%|████▌     | 5061/11082 [16:15:48<19:13:12, 11.49s/step, acc=0.4918, loss=2.1059, lr=6.24e-05]
Training:  46%|████▌     | 5062/11082 [16:15:59<19:15:21, 11.52s/step, acc=0.4918, loss=2.1059, lr=6.24e-05]
Training:  46%|████▌     | 5063/11082 [16:16:11<19:17:29, 11.54s/step, acc=0.4918, loss=2.1059, lr=6.24e-05]
Training:  46%|████▌     | 5064/11082 [16:16:22<19:21:45, 11.58s/step, acc=0.4918, loss=2.1059, lr=6.24e-05]
Training:  46%|████▌     | 5065/11082 [16:16:34<19:29:15, 11.66s/step, acc=0.4918, loss=2.1059, lr=6.24e-05]
Training:  46%|████▌     | 5066/11082 [16:16:45<19:15:27, 11.52s/step, acc=0.4918, loss=2.1059, lr=6.24e-05]
Training:  46%|████▌     | 5067/11082 [16:16:57<19:25:11, 11.62s/step, acc=0.4918, loss=2.1059, lr=6.24e-05]
Training:  46%|████▌     | 5068/11082 [16:17:09<19:14:53, 11.52s/step, acc=0.4918, loss=2.1059, lr=6.24e-05]
Training:  46%|████▌     | 5069/11082 [16:17:20<19:17:42, 11.55s/step, acc=0.4918, loss=2.1059, lr=6.24e-05]
Training:  46%|████▌     | 5070/11082 [16:17:32<19:22:47, 11.60s/step, acc=0.4918, loss=2.1059, lr=6.24e-05]
Training:  46%|████▌     | 5071/11082 [16:17:44<19:26:35, 11.64s/step, acc=0.4918, loss=2.1059, lr=6.24e-05]
Training:  46%|████▌     | 5072/11082 [16:17:54<19:00:46, 11.39s/step, acc=0.4918, loss=2.1059, lr=6.24e-05]
Training:  46%|████▌     | 5073/11082 [16:18:06<19:19:08, 11.57s/step, acc=0.4918, loss=2.1059, lr=6.24e-05]
Training:  46%|████▌     | 5074/11082 [16:18:19<19:40:08, 11.79s/step, acc=0.4918, loss=2.1059, lr=6.24e-05]
Training:  46%|████▌     | 5075/11082 [16:18:30<19:35:41, 11.74s/step, acc=0.4918, loss=2.1059, lr=6.24e-05]
Training:  46%|████▌     | 5076/11082 [16:18:42<19:21:03, 11.60s/step, acc=0.4918, loss=2.1059, lr=6.24e-05]
Training:  46%|████▌     | 5077/11082 [16:18:53<19:26:23, 11.65s/step, acc=0.4918, loss=2.1059, lr=6.24e-05]
Training:  46%|████▌     | 5078/11082 [16:19:05<19:22:31, 11.62s/step, acc=0.4918, loss=2.1059, lr=6.24e-05]
Training:  46%|████▌     | 5079/11082 [16:19:16<19:17:08, 11.57s/step, acc=0.4918, loss=2.1059, lr=6.24e-05]
Training:  46%|████▌     | 5080/11082 [16:19:28<19:08:25, 11.48s/step, acc=0.4918, loss=2.1059, lr=6.24e-05]
Training:  46%|████▌     | 5081/11082 [16:19:40<19:22:45, 11.63s/step, acc=0.4918, loss=2.1059, lr=6.24e-05]
Training:  46%|████▌     | 5082/11082 [16:19:51<19:22:21, 11.62s/step, acc=0.4918, loss=2.1059, lr=6.24e-05]
Training:  46%|████▌     | 5083/11082 [16:20:03<19:25:49, 11.66s/step, acc=0.4918, loss=2.1059, lr=6.24e-05]
Training:  46%|████▌     | 5084/11082 [16:20:15<19:30:54, 11.71s/step, acc=0.4918, loss=2.1059, lr=6.24e-05]
Training:  46%|████▌     | 5085/11082 [16:20:26<19:15:29, 11.56s/step, acc=0.4918, loss=2.1059, lr=6.24e-05]
Training:  46%|████▌     | 5086/11082 [16:20:38<19:19:40, 11.60s/step, acc=0.4918, loss=2.1059, lr=6.24e-05]
Training:  46%|████▌     | 5087/11082 [16:20:49<19:21:24, 11.62s/step, acc=0.4918, loss=2.1059, lr=6.24e-05]
Training:  46%|████▌     | 5088/11082 [16:21:00<19:03:21, 11.45s/step, acc=0.4918, loss=2.1059, lr=6.24e-05]
Training:  46%|████▌     | 5089/11082 [16:21:12<19:08:30, 11.50s/step, acc=0.4918, loss=2.1059, lr=6.24e-05]
Training:  46%|████▌     | 5090/11082 [16:21:24<19:18:37, 11.60s/step, acc=0.4918, loss=2.1059, lr=6.24e-05]
Training:  46%|████▌     | 5091/11082 [16:21:35<19:11:22, 11.53s/step, acc=0.4918, loss=2.1059, lr=6.24e-05]
Training:  46%|████▌     | 5092/11082 [16:21:47<19:12:56, 11.55s/step, acc=0.4918, loss=2.1059, lr=6.24e-05]
Training:  46%|████▌     | 5093/11082 [16:21:58<19:09:51, 11.52s/step, acc=0.4918, loss=2.1059, lr=6.24e-05]
Training:  46%|████▌     | 5094/11082 [16:22:10<19:18:30, 11.61s/step, acc=0.4918, loss=2.1059, lr=6.24e-05]
Training:  46%|████▌     | 5095/11082 [16:22:22<19:18:19, 11.61s/step, acc=0.4918, loss=2.1059, lr=6.24e-05]
Training:  46%|████▌     | 5096/11082 [16:22:33<19:07:19, 11.50s/step, acc=0.4918, loss=2.1059, lr=6.24e-05]
Training:  46%|████▌     | 5097/11082 [16:22:44<18:57:18, 11.40s/step, acc=0.4918, loss=2.1059, lr=6.24e-05]
Training:  46%|████▌     | 5098/11082 [16:22:56<19:03:12, 11.46s/step, acc=0.4918, loss=2.1059, lr=6.24e-05]
Training:  46%|████▌     | 5099/11082 [16:23:08<19:21:42, 11.65s/step, acc=0.4918, loss=2.1059, lr=6.24e-05]
Training:  46%|████▌     | 5100/11082 [16:23:19<19:20:56, 11.64s/step, acc=0.4918, loss=2.1059, lr=6.24e-05]
Training:  46%|████▌     | 5100/11082 [16:23:26<19:20:56, 11.64s/step, acc=0.4913, loss=2.1086, lr=6.18e-05]2026-06-19 07:16:54 | INFO | __main__ | epoch=0 step=5100 train_loss=2.108571 train_acc=0.491276 lr=6.176e-05 grad_norm=1.1562 acc_d0=0.7633 acc_d1=0.6193 acc_d2=0.4986 acc_d3=0.4006 acc_d4=0.3211 acc_d5=0.2576 acc_d6=0.2086 acc_d7=0.1726

Training:  46%|████▌     | 5101/11082 [16:23:31<19:24:33, 11.68s/step, acc=0.4913, loss=2.1086, lr=6.18e-05]
Training:  46%|████▌     | 5102/11082 [16:23:43<19:14:05, 11.58s/step, acc=0.4913, loss=2.1086, lr=6.18e-05]
Training:  46%|████▌     | 5103/11082 [16:23:54<19:09:51, 11.54s/step, acc=0.4913, loss=2.1086, lr=6.18e-05]
Training:  46%|████▌     | 5104/11082 [16:24:05<19:00:16, 11.44s/step, acc=0.4913, loss=2.1086, lr=6.18e-05]
Training:  46%|████▌     | 5105/11082 [16:24:17<19:10:06, 11.55s/step, acc=0.4913, loss=2.1086, lr=6.18e-05]
Training:  46%|████▌     | 5106/11082 [16:24:28<18:56:50, 11.41s/step, acc=0.4913, loss=2.1086, lr=6.18e-05]
Training:  46%|████▌     | 5107/11082 [16:24:40<19:05:01, 11.50s/step, acc=0.4913, loss=2.1086, lr=6.18e-05]
Training:  46%|████▌     | 5108/11082 [16:24:52<19:18:42, 11.64s/step, acc=0.4913, loss=2.1086, lr=6.18e-05]
Training:  46%|████▌     | 5109/11082 [16:25:03<19:19:35, 11.65s/step, acc=0.4913, loss=2.1086, lr=6.18e-05]
Training:  46%|████▌     | 5110/11082 [16:25:15<19:05:56, 11.51s/step, acc=0.4913, loss=2.1086, lr=6.18e-05]
Training:  46%|████▌     | 5111/11082 [16:25:27<19:15:44, 11.61s/step, acc=0.4913, loss=2.1086, lr=6.18e-05]
Training:  46%|████▌     | 5112/11082 [16:25:38<19:15:25, 11.61s/step, acc=0.4913, loss=2.1086, lr=6.18e-05]
Training:  46%|████▌     | 5113/11082 [16:25:49<18:59:58, 11.46s/step, acc=0.4913, loss=2.1086, lr=6.18e-05]
Training:  46%|████▌     | 5114/11082 [16:26:01<19:11:04, 11.57s/step, acc=0.4913, loss=2.1086, lr=6.18e-05]
Training:  46%|████▌     | 5115/11082 [16:26:13<19:12:21, 11.59s/step, acc=0.4913, loss=2.1086, lr=6.18e-05]
Training:  46%|████▌     | 5116/11082 [16:26:24<19:13:33, 11.60s/step, acc=0.4913, loss=2.1086, lr=6.18e-05]
Training:  46%|████▌     | 5117/11082 [16:26:36<19:04:01, 11.51s/step, acc=0.4913, loss=2.1086, lr=6.18e-05]
Training:  46%|████▌     | 5118/11082 [16:26:47<18:59:49, 11.47s/step, acc=0.4913, loss=2.1086, lr=6.18e-05]
Training:  46%|████▌     | 5119/11082 [16:26:59<19:17:35, 11.65s/step, acc=0.4913, loss=2.1086, lr=6.18e-05]
Training:  46%|████▌     | 5120/11082 [16:27:10<19:01:38, 11.49s/step, acc=0.4913, loss=2.1086, lr=6.18e-05]
Training:  46%|████▌     | 5121/11082 [16:27:22<19:13:26, 11.61s/step, acc=0.4913, loss=2.1086, lr=6.18e-05]
Training:  46%|████▌     | 5122/11082 [16:27:33<19:01:42, 11.49s/step, acc=0.4913, loss=2.1086, lr=6.18e-05]
Training:  46%|████▌     | 5123/11082 [16:27:45<18:57:18, 11.45s/step, acc=0.4913, loss=2.1086, lr=6.18e-05]
Training:  46%|████▌     | 5124/11082 [16:27:56<19:00:59, 11.49s/step, acc=0.4913, loss=2.1086, lr=6.18e-05]
Training:  46%|████▌     | 5125/11082 [16:28:08<19:06:47, 11.55s/step, acc=0.4913, loss=2.1086, lr=6.18e-05]
Training:  46%|████▋     | 5126/11082 [16:28:19<19:02:28, 11.51s/step, acc=0.4913, loss=2.1086, lr=6.18e-05]
Training:  46%|████▋     | 5127/11082 [16:28:31<19:11:39, 11.60s/step, acc=0.4913, loss=2.1086, lr=6.18e-05]
Training:  46%|████▋     | 5128/11082 [16:28:43<19:10:58, 11.60s/step, acc=0.4913, loss=2.1086, lr=6.18e-05]
Training:  46%|████▋     | 5129/11082 [16:28:54<19:05:34, 11.55s/step, acc=0.4913, loss=2.1086, lr=6.18e-05]
Training:  46%|████▋     | 5130/11082 [16:29:06<19:11:15, 11.61s/step, acc=0.4913, loss=2.1086, lr=6.18e-05]
Training:  46%|████▋     | 5131/11082 [16:29:18<19:15:12, 11.65s/step, acc=0.4913, loss=2.1086, lr=6.18e-05]
Training:  46%|████▋     | 5132/11082 [16:29:29<19:08:52, 11.59s/step, acc=0.4913, loss=2.1086, lr=6.18e-05]
Training:  46%|████▋     | 5133/11082 [16:29:41<19:14:28, 11.64s/step, acc=0.4913, loss=2.1086, lr=6.18e-05]
Training:  46%|████▋     | 5134/11082 [16:29:52<19:08:38, 11.59s/step, acc=0.4913, loss=2.1086, lr=6.18e-05]
Training:  46%|████▋     | 5135/11082 [16:30:04<19:16:57, 11.67s/step, acc=0.4913, loss=2.1086, lr=6.18e-05]
Training:  46%|████▋     | 5136/11082 [16:30:16<19:06:53, 11.57s/step, acc=0.4913, loss=2.1086, lr=6.18e-05]
Training:  46%|████▋     | 5137/11082 [16:30:27<19:07:22, 11.58s/step, acc=0.4913, loss=2.1086, lr=6.18e-05]
Training:  46%|████▋     | 5138/11082 [16:30:38<19:00:31, 11.51s/step, acc=0.4913, loss=2.1086, lr=6.18e-05]
Training:  46%|████▋     | 5139/11082 [16:30:51<19:16:45, 11.68s/step, acc=0.4913, loss=2.1086, lr=6.18e-05]
Training:  46%|████▋     | 5140/11082 [16:31:02<19:09:26, 11.61s/step, acc=0.4913, loss=2.1086, lr=6.18e-05]
Training:  46%|████▋     | 5141/11082 [16:31:13<19:05:43, 11.57s/step, acc=0.4913, loss=2.1086, lr=6.18e-05]
Training:  46%|████▋     | 5142/11082 [16:31:25<19:05:17, 11.57s/step, acc=0.4913, loss=2.1086, lr=6.18e-05]
Training:  46%|████▋     | 5143/11082 [16:31:36<18:55:22, 11.47s/step, acc=0.4913, loss=2.1086, lr=6.18e-05]
Training:  46%|████▋     | 5144/11082 [16:31:47<18:40:44, 11.32s/step, acc=0.4913, loss=2.1086, lr=6.18e-05]
Training:  46%|████▋     | 5145/11082 [16:32:00<19:11:10, 11.63s/step, acc=0.4913, loss=2.1086, lr=6.18e-05]
Training:  46%|████▋     | 5146/11082 [16:32:11<19:12:23, 11.65s/step, acc=0.4913, loss=2.1086, lr=6.18e-05]
Training:  46%|████▋     | 5147/11082 [16:32:23<19:12:20, 11.65s/step, acc=0.4913, loss=2.1086, lr=6.18e-05]
Training:  46%|████▋     | 5148/11082 [16:32:34<19:05:08, 11.58s/step, acc=0.4913, loss=2.1086, lr=6.18e-05]
Training:  46%|████▋     | 5149/11082 [16:32:46<19:00:11, 11.53s/step, acc=0.4913, loss=2.1086, lr=6.18e-05]
Training:  46%|████▋     | 5150/11082 [16:32:58<19:07:28, 11.61s/step, acc=0.4913, loss=2.1086, lr=6.18e-05]
Training:  46%|████▋     | 5150/11082 [16:33:05<19:07:28, 11.61s/step, acc=0.4939, loss=2.0921, lr=6.11e-05]2026-06-19 07:26:32 | INFO | __main__ | epoch=0 step=5150 train_loss=2.092119 train_acc=0.493880 lr=6.112e-05 grad_norm=0.8047 acc_d0=0.7616 acc_d1=0.6168 acc_d2=0.4936 acc_d3=0.3936 acc_d4=0.3120 acc_d5=0.2511 acc_d6=0.2025 acc_d7=0.1656

Training:  46%|████▋     | 5151/11082 [16:33:09<19:06:42, 11.60s/step, acc=0.4939, loss=2.0921, lr=6.11e-05]
Training:  46%|████▋     | 5152/11082 [16:33:21<19:09:31, 11.63s/step, acc=0.4939, loss=2.0921, lr=6.11e-05]
Training:  46%|████▋     | 5153/11082 [16:33:32<19:05:03, 11.59s/step, acc=0.4939, loss=2.0921, lr=6.11e-05]
Training:  47%|████▋     | 5154/11082 [16:33:44<19:13:21, 11.67s/step, acc=0.4939, loss=2.0921, lr=6.11e-05]
Training:  47%|████▋     | 5155/11082 [16:33:56<19:02:22, 11.56s/step, acc=0.4939, loss=2.0921, lr=6.11e-05]
Training:  47%|████▋     | 5156/11082 [16:34:07<19:11:09, 11.66s/step, acc=0.4939, loss=2.0921, lr=6.11e-05]
Training:  47%|████▋     | 5157/11082 [16:34:19<19:07:53, 11.62s/step, acc=0.4939, loss=2.0921, lr=6.11e-05]
Training:  47%|████▋     | 5158/11082 [16:34:31<19:06:26, 11.61s/step, acc=0.4939, loss=2.0921, lr=6.11e-05]
Training:  47%|████▋     | 5159/11082 [16:34:42<19:00:59, 11.56s/step, acc=0.4939, loss=2.0921, lr=6.11e-05]
Training:  47%|████▋     | 5160/11082 [16:34:54<19:02:23, 11.57s/step, acc=0.4939, loss=2.0921, lr=6.11e-05]
Training:  47%|████▋     | 5161/11082 [16:35:05<18:58:07, 11.53s/step, acc=0.4939, loss=2.0921, lr=6.11e-05]
Training:  47%|████▋     | 5162/11082 [16:35:16<18:55:04, 11.50s/step, acc=0.4939, loss=2.0921, lr=6.11e-05]
Training:  47%|████▋     | 5163/11082 [16:35:28<19:00:25, 11.56s/step, acc=0.4939, loss=2.0921, lr=6.11e-05]
Training:  47%|████▋     | 5164/11082 [16:35:40<19:10:43, 11.67s/step, acc=0.4939, loss=2.0921, lr=6.11e-05]
Training:  47%|████▋     | 5165/11082 [16:35:51<19:03:37, 11.60s/step, acc=0.4939, loss=2.0921, lr=6.11e-05]
Training:  47%|████▋     | 5166/11082 [16:36:03<18:51:10, 11.47s/step, acc=0.4939, loss=2.0921, lr=6.11e-05]
Training:  47%|████▋     | 5167/11082 [16:36:14<18:59:12, 11.56s/step, acc=0.4939, loss=2.0921, lr=6.11e-05]
Training:  47%|████▋     | 5168/11082 [16:36:25<18:34:00, 11.30s/step, acc=0.4939, loss=2.0921, lr=6.11e-05]
Training:  47%|████▋     | 5169/11082 [16:36:38<19:05:55, 11.63s/step, acc=0.4939, loss=2.0921, lr=6.11e-05]
Training:  47%|████▋     | 5170/11082 [16:36:49<18:59:42, 11.57s/step, acc=0.4939, loss=2.0921, lr=6.11e-05]
Training:  47%|████▋     | 5171/11082 [16:37:00<18:57:47, 11.55s/step, acc=0.4939, loss=2.0921, lr=6.11e-05]
Training:  47%|████▋     | 5172/11082 [16:37:12<18:49:47, 11.47s/step, acc=0.4939, loss=2.0921, lr=6.11e-05]
Training:  47%|████▋     | 5173/11082 [16:37:24<19:00:49, 11.58s/step, acc=0.4939, loss=2.0921, lr=6.11e-05]
Training:  47%|████▋     | 5174/11082 [16:37:35<18:52:14, 11.50s/step, acc=0.4939, loss=2.0921, lr=6.11e-05]
Training:  47%|████▋     | 5175/11082 [16:37:47<18:59:36, 11.58s/step, acc=0.4939, loss=2.0921, lr=6.11e-05]
Training:  47%|████▋     | 5176/11082 [16:37:58<18:57:49, 11.56s/step, acc=0.4939, loss=2.0921, lr=6.11e-05]
Training:  47%|████▋     | 5177/11082 [16:38:10<19:01:16, 11.60s/step, acc=0.4939, loss=2.0921, lr=6.11e-05]
Training:  47%|████▋     | 5178/11082 [16:38:21<18:48:27, 11.47s/step, acc=0.4939, loss=2.0921, lr=6.11e-05]
Training:  47%|████▋     | 5179/11082 [16:38:33<19:04:23, 11.63s/step, acc=0.4939, loss=2.0921, lr=6.11e-05]
Training:  47%|████▋     | 5180/11082 [16:38:44<18:57:46, 11.57s/step, acc=0.4939, loss=2.0921, lr=6.11e-05]
Training:  47%|████▋     | 5181/11082 [16:38:56<18:57:08, 11.56s/step, acc=0.4939, loss=2.0921, lr=6.11e-05]
Training:  47%|████▋     | 5182/11082 [16:39:08<18:55:57, 11.55s/step, acc=0.4939, loss=2.0921, lr=6.11e-05]
Training:  47%|████▋     | 5183/11082 [16:39:19<18:54:22, 11.54s/step, acc=0.4939, loss=2.0921, lr=6.11e-05]
Training:  47%|████▋     | 5184/11082 [16:39:30<18:50:53, 11.50s/step, acc=0.4939, loss=2.0921, lr=6.11e-05]
Training:  47%|████▋     | 5185/11082 [16:39:42<18:55:54, 11.56s/step, acc=0.4939, loss=2.0921, lr=6.11e-05]
Training:  47%|████▋     | 5186/11082 [16:39:54<18:51:03, 11.51s/step, acc=0.4939, loss=2.0921, lr=6.11e-05]
Training:  47%|████▋     | 5187/11082 [16:40:05<19:00:08, 11.60s/step, acc=0.4939, loss=2.0921, lr=6.11e-05]
Training:  47%|████▋     | 5188/11082 [16:40:17<19:05:14, 11.66s/step, acc=0.4939, loss=2.0921, lr=6.11e-05]
Training:  47%|████▋     | 5189/11082 [16:40:28<18:51:36, 11.52s/step, acc=0.4939, loss=2.0921, lr=6.11e-05]
Training:  47%|████▋     | 5190/11082 [16:40:40<18:58:16, 11.59s/step, acc=0.4939, loss=2.0921, lr=6.11e-05]
Training:  47%|████▋     | 5191/11082 [16:40:52<19:02:58, 11.64s/step, acc=0.4939, loss=2.0921, lr=6.11e-05]
Training:  47%|████▋     | 5192/11082 [16:41:03<18:48:33, 11.50s/step, acc=0.4939, loss=2.0921, lr=6.11e-05]
Training:  47%|████▋     | 5193/11082 [16:41:15<18:57:08, 11.59s/step, acc=0.4939, loss=2.0921, lr=6.11e-05]
Training:  47%|████▋     | 5194/11082 [16:41:27<19:04:11, 11.66s/step, acc=0.4939, loss=2.0921, lr=6.11e-05]
Training:  47%|████▋     | 5195/11082 [16:41:38<19:00:10, 11.62s/step, acc=0.4939, loss=2.0921, lr=6.11e-05]
Training:  47%|████▋     | 5196/11082 [16:41:49<18:50:15, 11.52s/step, acc=0.4939, loss=2.0921, lr=6.11e-05]
Training:  47%|████▋     | 5197/11082 [16:42:01<18:41:48, 11.44s/step, acc=0.4939, loss=2.0921, lr=6.11e-05]
Training:  47%|████▋     | 5198/11082 [16:42:13<19:02:23, 11.65s/step, acc=0.4939, loss=2.0921, lr=6.11e-05]
Training:  47%|████▋     | 5199/11082 [16:42:24<18:51:04, 11.54s/step, acc=0.4939, loss=2.0921, lr=6.11e-05]
Training:  47%|████▋     | 5200/11082 [16:42:36<18:57:57, 11.61s/step, acc=0.4939, loss=2.0921, lr=6.11e-05]
Training:  47%|████▋     | 5200/11082 [16:42:43<18:57:57, 11.61s/step, acc=0.4903, loss=2.1165, lr=6.05e-05]2026-06-19 07:36:11 | INFO | __main__ | epoch=0 step=5200 train_loss=2.116479 train_acc=0.490265 lr=6.048e-05 grad_norm=1.2500 acc_d0=0.7384 acc_d1=0.5898 acc_d2=0.4695 acc_d3=0.3726 acc_d4=0.2965 acc_d5=0.2402 acc_d6=0.1948 acc_d7=0.1585

Training:  47%|████▋     | 5201/11082 [16:42:47<18:49:25, 11.52s/step, acc=0.4903, loss=2.1165, lr=6.05e-05]
Training:  47%|████▋     | 5202/11082 [16:42:59<18:56:27, 11.60s/step, acc=0.4903, loss=2.1165, lr=6.05e-05]
Training:  47%|████▋     | 5203/11082 [16:43:10<18:47:00, 11.50s/step, acc=0.4903, loss=2.1165, lr=6.05e-05]
Training:  47%|████▋     | 5204/11082 [16:43:22<18:42:51, 11.46s/step, acc=0.4903, loss=2.1165, lr=6.05e-05]
Training:  47%|████▋     | 5205/11082 [16:43:34<18:54:59, 11.59s/step, acc=0.4903, loss=2.1165, lr=6.05e-05]
Training:  47%|████▋     | 5206/11082 [16:43:45<18:48:20, 11.52s/step, acc=0.4903, loss=2.1165, lr=6.05e-05]
Training:  47%|████▋     | 5207/11082 [16:43:57<18:53:15, 11.57s/step, acc=0.4903, loss=2.1165, lr=6.05e-05]
Training:  47%|████▋     | 5208/11082 [16:44:08<18:42:58, 11.47s/step, acc=0.4903, loss=2.1165, lr=6.05e-05]
Training:  47%|████▋     | 5209/11082 [16:44:20<18:54:38, 11.59s/step, acc=0.4903, loss=2.1165, lr=6.05e-05]
Training:  47%|████▋     | 5210/11082 [16:44:31<18:49:47, 11.54s/step, acc=0.4903, loss=2.1165, lr=6.05e-05]
Training:  47%|████▋     | 5211/11082 [16:44:43<18:55:53, 11.61s/step, acc=0.4903, loss=2.1165, lr=6.05e-05]
Training:  47%|████▋     | 5212/11082 [16:44:54<18:54:16, 11.59s/step, acc=0.4903, loss=2.1165, lr=6.05e-05]
Training:  47%|████▋     | 5213/11082 [16:45:06<18:48:44, 11.54s/step, acc=0.4903, loss=2.1165, lr=6.05e-05]
Training:  47%|████▋     | 5214/11082 [16:45:18<18:54:27, 11.60s/step, acc=0.4903, loss=2.1165, lr=6.05e-05]
Training:  47%|████▋     | 5215/11082 [16:45:29<18:50:55, 11.57s/step, acc=0.4903, loss=2.1165, lr=6.05e-05]
Training:  47%|████▋     | 5216/11082 [16:45:41<18:47:08, 11.53s/step, acc=0.4903, loss=2.1165, lr=6.05e-05]
Training:  47%|████▋     | 5217/11082 [16:45:53<19:06:03, 11.72s/step, acc=0.4903, loss=2.1165, lr=6.05e-05]
Training:  47%|████▋     | 5218/11082 [16:46:04<18:52:58, 11.59s/step, acc=0.4903, loss=2.1165, lr=6.05e-05]
Training:  47%|████▋     | 5219/11082 [16:46:16<18:56:23, 11.63s/step, acc=0.4903, loss=2.1165, lr=6.05e-05]
Training:  47%|████▋     | 5220/11082 [16:46:27<18:55:47, 11.63s/step, acc=0.4903, loss=2.1165, lr=6.05e-05]
Training:  47%|████▋     | 5221/11082 [16:46:38<18:36:26, 11.43s/step, acc=0.4903, loss=2.1165, lr=6.05e-05]
Training:  47%|████▋     | 5222/11082 [16:46:50<18:56:21, 11.64s/step, acc=0.4903, loss=2.1165, lr=6.05e-05]
Training:  47%|████▋     | 5223/11082 [16:47:01<18:28:20, 11.35s/step, acc=0.4903, loss=2.1165, lr=6.05e-05]
Training:  47%|████▋     | 5224/11082 [16:47:13<18:52:41, 11.60s/step, acc=0.4903, loss=2.1165, lr=6.05e-05]
Training:  47%|████▋     | 5225/11082 [16:47:25<18:48:09, 11.56s/step, acc=0.4903, loss=2.1165, lr=6.05e-05]
Training:  47%|████▋     | 5226/11082 [16:47:36<18:29:58, 11.37s/step, acc=0.4903, loss=2.1165, lr=6.05e-05]
Training:  47%|████▋     | 5227/11082 [16:47:48<18:57:54, 11.66s/step, acc=0.4903, loss=2.1165, lr=6.05e-05]
Training:  47%|████▋     | 5228/11082 [16:48:00<18:54:13, 11.63s/step, acc=0.4903, loss=2.1165, lr=6.05e-05]
Training:  47%|████▋     | 5229/11082 [16:48:11<18:58:19, 11.67s/step, acc=0.4903, loss=2.1165, lr=6.05e-05]
Training:  47%|████▋     | 5230/11082 [16:48:23<18:57:48, 11.67s/step, acc=0.4903, loss=2.1165, lr=6.05e-05]
Training:  47%|████▋     | 5231/11082 [16:48:34<18:45:51, 11.55s/step, acc=0.4903, loss=2.1165, lr=6.05e-05]
Training:  47%|████▋     | 5232/11082 [16:48:46<18:37:59, 11.47s/step, acc=0.4903, loss=2.1165, lr=6.05e-05]
Training:  47%|████▋     | 5233/11082 [16:48:58<18:55:17, 11.65s/step, acc=0.4903, loss=2.1165, lr=6.05e-05]
Training:  47%|████▋     | 5234/11082 [16:49:10<19:04:21, 11.74s/step, acc=0.4903, loss=2.1165, lr=6.05e-05]
Training:  47%|████▋     | 5235/11082 [16:49:21<18:56:47, 11.67s/step, acc=0.4903, loss=2.1165, lr=6.05e-05]
Training:  47%|████▋     | 5236/11082 [16:49:33<18:53:12, 11.63s/step, acc=0.4903, loss=2.1165, lr=6.05e-05]
Training:  47%|████▋     | 5237/11082 [16:49:44<18:54:03, 11.64s/step, acc=0.4903, loss=2.1165, lr=6.05e-05]
Training:  47%|████▋     | 5238/11082 [16:49:56<18:54:57, 11.65s/step, acc=0.4903, loss=2.1165, lr=6.05e-05]
Training:  47%|████▋     | 5239/11082 [16:50:07<18:36:05, 11.46s/step, acc=0.4903, loss=2.1165, lr=6.05e-05]
Training:  47%|████▋     | 5240/11082 [16:50:19<18:45:02, 11.55s/step, acc=0.4903, loss=2.1165, lr=6.05e-05]
Training:  47%|████▋     | 5241/11082 [16:50:30<18:43:21, 11.54s/step, acc=0.4903, loss=2.1165, lr=6.05e-05]
Training:  47%|████▋     | 5242/11082 [16:50:42<18:54:50, 11.66s/step, acc=0.4903, loss=2.1165, lr=6.05e-05]
Training:  47%|████▋     | 5243/11082 [16:50:54<18:45:03, 11.56s/step, acc=0.4903, loss=2.1165, lr=6.05e-05]
Training:  47%|████▋     | 5244/11082 [16:51:05<18:44:08, 11.55s/step, acc=0.4903, loss=2.1165, lr=6.05e-05]
Training:  47%|████▋     | 5245/11082 [16:51:17<18:43:18, 11.55s/step, acc=0.4903, loss=2.1165, lr=6.05e-05]
Training:  47%|████▋     | 5246/11082 [16:51:28<18:40:37, 11.52s/step, acc=0.4903, loss=2.1165, lr=6.05e-05]
Training:  47%|████▋     | 5247/11082 [16:51:39<18:30:45, 11.42s/step, acc=0.4903, loss=2.1165, lr=6.05e-05]
Training:  47%|████▋     | 5248/11082 [16:51:51<18:33:55, 11.46s/step, acc=0.4903, loss=2.1165, lr=6.05e-05]
Training:  47%|████▋     | 5249/11082 [16:52:03<18:44:45, 11.57s/step, acc=0.4903, loss=2.1165, lr=6.05e-05]
Training:  47%|████▋     | 5250/11082 [16:52:14<18:39:08, 11.51s/step, acc=0.4903, loss=2.1165, lr=6.05e-05]
Training:  47%|████▋     | 5250/11082 [16:52:21<18:39:08, 11.51s/step, acc=0.4923, loss=2.1067, lr=5.98e-05]2026-06-19 07:45:49 | INFO | __main__ | epoch=0 step=5250 train_loss=2.106651 train_acc=0.492286 lr=5.983e-05 grad_norm=0.7461 acc_d0=0.7592 acc_d1=0.6128 acc_d2=0.4913 acc_d3=0.3916 acc_d4=0.3130 acc_d5=0.2515 acc_d6=0.2038 acc_d7=0.1690

Training:  47%|████▋     | 5251/11082 [16:52:25<18:29:02, 11.41s/step, acc=0.4923, loss=2.1067, lr=5.98e-05]
Training:  47%|████▋     | 5252/11082 [16:52:37<18:44:48, 11.58s/step, acc=0.4923, loss=2.1067, lr=5.98e-05]
Training:  47%|████▋     | 5253/11082 [16:52:48<18:38:37, 11.51s/step, acc=0.4923, loss=2.1067, lr=5.98e-05]
Training:  47%|████▋     | 5254/11082 [16:53:00<18:38:35, 11.52s/step, acc=0.4923, loss=2.1067, lr=5.98e-05]
Training:  47%|████▋     | 5255/11082 [16:53:12<18:58:12, 11.72s/step, acc=0.4923, loss=2.1067, lr=5.98e-05]
Training:  47%|████▋     | 5256/11082 [16:53:24<18:46:32, 11.60s/step, acc=0.4923, loss=2.1067, lr=5.98e-05]
Training:  47%|████▋     | 5257/11082 [16:53:35<18:40:42, 11.54s/step, acc=0.4923, loss=2.1067, lr=5.98e-05]
Training:  47%|████▋     | 5258/11082 [16:53:46<18:29:34, 11.43s/step, acc=0.4923, loss=2.1067, lr=5.98e-05]
Training:  47%|████▋     | 5259/11082 [16:53:58<18:32:03, 11.46s/step, acc=0.4923, loss=2.1067, lr=5.98e-05]
Training:  47%|████▋     | 5260/11082 [16:54:10<18:54:14, 11.69s/step, acc=0.4923, loss=2.1067, lr=5.98e-05]
Training:  47%|████▋     | 5261/11082 [16:54:21<18:47:01, 11.62s/step, acc=0.4923, loss=2.1067, lr=5.98e-05]
Training:  47%|████▋     | 5262/11082 [16:54:33<18:50:29, 11.65s/step, acc=0.4923, loss=2.1067, lr=5.98e-05]
Training:  47%|████▋     | 5263/11082 [16:54:45<18:45:45, 11.61s/step, acc=0.4923, loss=2.1067, lr=5.98e-05]
Training:  48%|████▊     | 5264/11082 [16:54:56<18:38:22, 11.53s/step, acc=0.4923, loss=2.1067, lr=5.98e-05]
Training:  48%|████▊     | 5265/11082 [16:55:07<18:38:36, 11.54s/step, acc=0.4923, loss=2.1067, lr=5.98e-05]
Training:  48%|████▊     | 5266/11082 [16:55:19<18:34:35, 11.50s/step, acc=0.4923, loss=2.1067, lr=5.98e-05]
Training:  48%|████▊     | 5267/11082 [16:55:31<18:50:18, 11.66s/step, acc=0.4923, loss=2.1067, lr=5.98e-05]
Training:  48%|████▊     | 5268/11082 [16:55:42<18:41:55, 11.58s/step, acc=0.4923, loss=2.1067, lr=5.98e-05]
Training:  48%|████▊     | 5269/11082 [16:55:54<18:36:46, 11.53s/step, acc=0.4923, loss=2.1067, lr=5.98e-05]
Training:  48%|████▊     | 5270/11082 [16:56:06<18:45:54, 11.62s/step, acc=0.4923, loss=2.1067, lr=5.98e-05]
Training:  48%|████▊     | 5271/11082 [16:56:17<18:37:51, 11.54s/step, acc=0.4923, loss=2.1067, lr=5.98e-05]
Training:  48%|████▊     | 5272/11082 [16:56:29<18:45:50, 11.63s/step, acc=0.4923, loss=2.1067, lr=5.98e-05]
Training:  48%|████▊     | 5273/11082 [16:56:40<18:43:48, 11.61s/step, acc=0.4923, loss=2.1067, lr=5.98e-05]
Training:  48%|████▊     | 5274/11082 [16:56:52<18:34:23, 11.51s/step, acc=0.4923, loss=2.1067, lr=5.98e-05]
Training:  48%|████▊     | 5275/11082 [16:57:03<18:35:27, 11.53s/step, acc=0.4923, loss=2.1067, lr=5.98e-05]
Training:  48%|████▊     | 5276/11082 [16:57:15<18:44:08, 11.62s/step, acc=0.4923, loss=2.1067, lr=5.98e-05]
Training:  48%|████▊     | 5277/11082 [16:57:26<18:37:14, 11.55s/step, acc=0.4923, loss=2.1067, lr=5.98e-05]
Training:  48%|████▊     | 5278/11082 [16:57:38<18:35:10, 11.53s/step, acc=0.4923, loss=2.1067, lr=5.98e-05]
Training:  48%|████▊     | 5279/11082 [16:57:49<18:34:30, 11.52s/step, acc=0.4923, loss=2.1067, lr=5.98e-05]
Training:  48%|████▊     | 5280/11082 [16:58:01<18:30:15, 11.48s/step, acc=0.4923, loss=2.1067, lr=5.98e-05]
Training:  48%|████▊     | 5281/11082 [16:58:13<18:39:02, 11.57s/step, acc=0.4923, loss=2.1067, lr=5.98e-05]
Training:  48%|████▊     | 5282/11082 [16:58:24<18:33:33, 11.52s/step, acc=0.4923, loss=2.1067, lr=5.98e-05]
Training:  48%|████▊     | 5283/11082 [16:58:35<18:35:26, 11.54s/step, acc=0.4923, loss=2.1067, lr=5.98e-05]
Training:  48%|████▊     | 5284/11082 [16:58:47<18:36:10, 11.55s/step, acc=0.4923, loss=2.1067, lr=5.98e-05]
Training:  48%|████▊     | 5285/11082 [16:58:58<18:29:03, 11.48s/step, acc=0.4923, loss=2.1067, lr=5.98e-05]
Training:  48%|████▊     | 5286/11082 [16:59:10<18:24:49, 11.44s/step, acc=0.4923, loss=2.1067, lr=5.98e-05]
Training:  48%|████▊     | 5287/11082 [16:59:22<18:40:12, 11.60s/step, acc=0.4923, loss=2.1067, lr=5.98e-05]
Training:  48%|████▊     | 5288/11082 [16:59:33<18:42:29, 11.62s/step, acc=0.4923, loss=2.1067, lr=5.98e-05]
Training:  48%|████▊     | 5289/11082 [16:59:45<18:42:30, 11.63s/step, acc=0.4923, loss=2.1067, lr=5.98e-05]
Training:  48%|████▊     | 5290/11082 [16:59:57<18:49:10, 11.70s/step, acc=0.4923, loss=2.1067, lr=5.98e-05]
Training:  48%|████▊     | 5291/11082 [17:00:08<18:42:06, 11.63s/step, acc=0.4923, loss=2.1067, lr=5.98e-05]
Training:  48%|████▊     | 5292/11082 [17:00:20<18:31:33, 11.52s/step, acc=0.4923, loss=2.1067, lr=5.98e-05]
Training:  48%|████▊     | 5293/11082 [17:00:32<18:49:36, 11.71s/step, acc=0.4923, loss=2.1067, lr=5.98e-05]
Training:  48%|████▊     | 5294/11082 [17:00:43<18:32:30, 11.53s/step, acc=0.4923, loss=2.1067, lr=5.98e-05]
Training:  48%|████▊     | 5295/11082 [17:00:54<18:29:56, 11.51s/step, acc=0.4923, loss=2.1067, lr=5.98e-05]
Training:  48%|████▊     | 5296/11082 [17:01:06<18:37:45, 11.59s/step, acc=0.4923, loss=2.1067, lr=5.98e-05]
Training:  48%|████▊     | 5297/11082 [17:01:18<18:43:54, 11.66s/step, acc=0.4923, loss=2.1067, lr=5.98e-05]
Training:  48%|████▊     | 5298/11082 [17:01:29<18:32:19, 11.54s/step, acc=0.4923, loss=2.1067, lr=5.98e-05]
Training:  48%|████▊     | 5299/11082 [17:01:41<18:26:48, 11.48s/step, acc=0.4923, loss=2.1067, lr=5.98e-05]
Training:  48%|████▊     | 5300/11082 [17:01:52<18:31:02, 11.53s/step, acc=0.4923, loss=2.1067, lr=5.98e-05]
Training:  48%|████▊     | 5300/11082 [17:01:59<18:31:02, 11.53s/step, acc=0.4942, loss=2.0905, lr=5.92e-05]2026-06-19 07:55:27 | INFO | __main__ | epoch=0 step=5300 train_loss=2.090510 train_acc=0.494223 lr=5.919e-05 grad_norm=0.8594 acc_d0=0.7480 acc_d1=0.6023 acc_d2=0.4806 acc_d3=0.3820 acc_d4=0.3071 acc_d5=0.2457 acc_d6=0.1999 acc_d7=0.1665

Training:  48%|████▊     | 5301/11082 [17:02:04<18:27:21, 11.49s/step, acc=0.4942, loss=2.0905, lr=5.92e-05]
Training:  48%|████▊     | 5302/11082 [17:02:15<18:30:11, 11.52s/step, acc=0.4942, loss=2.0905, lr=5.92e-05]
Training:  48%|████▊     | 5303/11082 [17:02:27<18:35:36, 11.58s/step, acc=0.4942, loss=2.0905, lr=5.92e-05]
Training:  48%|████▊     | 5304/11082 [17:02:38<18:28:46, 11.51s/step, acc=0.4942, loss=2.0905, lr=5.92e-05]
Training:  48%|████▊     | 5305/11082 [17:02:50<18:27:32, 11.50s/step, acc=0.4942, loss=2.0905, lr=5.92e-05]
Training:  48%|████▊     | 5306/11082 [17:03:01<18:17:19, 11.40s/step, acc=0.4942, loss=2.0905, lr=5.92e-05]
Training:  48%|████▊     | 5307/11082 [17:03:13<18:37:18, 11.61s/step, acc=0.4942, loss=2.0905, lr=5.92e-05]
Training:  48%|████▊     | 5308/11082 [17:03:25<18:35:36, 11.59s/step, acc=0.4942, loss=2.0905, lr=5.92e-05]
Training:  48%|████▊     | 5309/11082 [17:03:36<18:39:18, 11.63s/step, acc=0.4942, loss=2.0905, lr=5.92e-05]
Training:  48%|████▊     | 5310/11082 [17:03:48<18:28:42, 11.52s/step, acc=0.4942, loss=2.0905, lr=5.92e-05]
Training:  48%|████▊     | 5311/11082 [17:04:00<18:42:27, 11.67s/step, acc=0.4942, loss=2.0905, lr=5.92e-05]
Training:  48%|████▊     | 5312/11082 [17:04:11<18:34:26, 11.59s/step, acc=0.4942, loss=2.0905, lr=5.92e-05]
Training:  48%|████▊     | 5313/11082 [17:04:23<18:34:22, 11.59s/step, acc=0.4942, loss=2.0905, lr=5.92e-05]
Training:  48%|████▊     | 5314/11082 [17:04:34<18:35:34, 11.60s/step, acc=0.4942, loss=2.0905, lr=5.92e-05]
Training:  48%|████▊     | 5315/11082 [17:04:46<18:48:08, 11.74s/step, acc=0.4942, loss=2.0905, lr=5.92e-05]
Training:  48%|████▊     | 5316/11082 [17:04:58<18:37:56, 11.63s/step, acc=0.4942, loss=2.0905, lr=5.92e-05]
Training:  48%|████▊     | 5317/11082 [17:05:09<18:31:38, 11.57s/step, acc=0.4942, loss=2.0905, lr=5.92e-05]
Training:  48%|████▊     | 5318/11082 [17:05:20<18:24:43, 11.50s/step, acc=0.4942, loss=2.0905, lr=5.92e-05]
Training:  48%|████▊     | 5319/11082 [17:05:32<18:29:13, 11.55s/step, acc=0.4942, loss=2.0905, lr=5.92e-05]
Training:  48%|████▊     | 5320/11082 [17:05:43<18:26:28, 11.52s/step, acc=0.4942, loss=2.0905, lr=5.92e-05]
Training:  48%|████▊     | 5321/11082 [17:05:55<18:22:43, 11.48s/step, acc=0.4942, loss=2.0905, lr=5.92e-05]
Training:  48%|████▊     | 5322/11082 [17:06:07<18:32:55, 11.59s/step, acc=0.4942, loss=2.0905, lr=5.92e-05]
Training:  48%|████▊     | 5323/11082 [17:06:18<18:25:34, 11.52s/step, acc=0.4942, loss=2.0905, lr=5.92e-05]
Training:  48%|████▊     | 5324/11082 [17:06:30<18:37:42, 11.65s/step, acc=0.4942, loss=2.0905, lr=5.92e-05]
Training:  48%|████▊     | 5325/11082 [17:06:41<18:18:59, 11.45s/step, acc=0.4942, loss=2.0905, lr=5.92e-05]
Training:  48%|████▊     | 5326/11082 [17:06:53<18:28:39, 11.56s/step, acc=0.4942, loss=2.0905, lr=5.92e-05]
Training:  48%|████▊     | 5327/11082 [17:07:05<18:42:59, 11.71s/step, acc=0.4942, loss=2.0905, lr=5.92e-05]
Training:  48%|████▊     | 5328/11082 [17:07:16<18:34:43, 11.62s/step, acc=0.4942, loss=2.0905, lr=5.92e-05]
Training:  48%|████▊     | 5329/11082 [17:07:28<18:26:24, 11.54s/step, acc=0.4942, loss=2.0905, lr=5.92e-05]
Training:  48%|████▊     | 5330/11082 [17:07:39<18:25:51, 11.54s/step, acc=0.4942, loss=2.0905, lr=5.92e-05]
Training:  48%|████▊     | 5331/11082 [17:07:51<18:27:08, 11.55s/step, acc=0.4942, loss=2.0905, lr=5.92e-05]
Training:  48%|████▊     | 5332/11082 [17:08:03<18:33:44, 11.62s/step, acc=0.4942, loss=2.0905, lr=5.92e-05]
Training:  48%|████▊     | 5333/11082 [17:08:14<18:25:38, 11.54s/step, acc=0.4942, loss=2.0905, lr=5.92e-05]
Training:  48%|████▊     | 5334/11082 [17:08:26<18:34:10, 11.63s/step, acc=0.4942, loss=2.0905, lr=5.92e-05]
Training:  48%|████▊     | 5335/11082 [17:08:37<18:30:20, 11.59s/step, acc=0.4942, loss=2.0905, lr=5.92e-05]
Training:  48%|████▊     | 5336/11082 [17:08:49<18:24:23, 11.53s/step, acc=0.4942, loss=2.0905, lr=5.92e-05]
Training:  48%|████▊     | 5337/11082 [17:09:00<18:29:51, 11.59s/step, acc=0.4942, loss=2.0905, lr=5.92e-05]
Training:  48%|████▊     | 5338/11082 [17:09:12<18:23:28, 11.53s/step, acc=0.4942, loss=2.0905, lr=5.92e-05]
Training:  48%|████▊     | 5339/11082 [17:09:23<18:17:20, 11.46s/step, acc=0.4942, loss=2.0905, lr=5.92e-05]
Training:  48%|████▊     | 5340/11082 [17:09:35<18:25:47, 11.55s/step, acc=0.4942, loss=2.0905, lr=5.92e-05]
Training:  48%|████▊     | 5341/11082 [17:09:46<18:26:44, 11.57s/step, acc=0.4942, loss=2.0905, lr=5.92e-05]
Training:  48%|████▊     | 5342/11082 [17:09:58<18:32:39, 11.63s/step, acc=0.4942, loss=2.0905, lr=5.92e-05]
Training:  48%|████▊     | 5343/11082 [17:10:10<18:24:48, 11.55s/step, acc=0.4942, loss=2.0905, lr=5.92e-05]
Training:  48%|████▊     | 5344/11082 [17:10:21<18:30:59, 11.62s/step, acc=0.4942, loss=2.0905, lr=5.92e-05]
Training:  48%|████▊     | 5345/11082 [17:10:33<18:36:29, 11.68s/step, acc=0.4942, loss=2.0905, lr=5.92e-05]
Training:  48%|████▊     | 5346/11082 [17:10:45<18:29:18, 11.60s/step, acc=0.4942, loss=2.0905, lr=5.92e-05]
Training:  48%|████▊     | 5347/11082 [17:10:56<18:21:44, 11.53s/step, acc=0.4942, loss=2.0905, lr=5.92e-05]
Training:  48%|████▊     | 5348/11082 [17:11:08<18:26:55, 11.58s/step, acc=0.4942, loss=2.0905, lr=5.92e-05]
Training:  48%|████▊     | 5349/11082 [17:11:19<18:18:23, 11.50s/step, acc=0.4942, loss=2.0905, lr=5.92e-05]
Training:  48%|████▊     | 5350/11082 [17:11:30<18:13:26, 11.45s/step, acc=0.4942, loss=2.0905, lr=5.92e-05]
Training:  48%|████▊     | 5350/11082 [17:11:38<18:13:26, 11.45s/step, acc=0.4959, loss=2.0837, lr=5.85e-05]2026-06-19 08:05:05 | INFO | __main__ | epoch=0 step=5350 train_loss=2.083749 train_acc=0.495886 lr=5.854e-05 grad_norm=1.0156 acc_d0=0.7682 acc_d1=0.6250 acc_d2=0.5019 acc_d3=0.4032 acc_d4=0.3219 acc_d5=0.2583 acc_d6=0.2076 acc_d7=0.1713

Training:  48%|████▊     | 5351/11082 [17:11:42<18:30:10, 11.62s/step, acc=0.4959, loss=2.0837, lr=5.85e-05]
Training:  48%|████▊     | 5352/11082 [17:11:54<18:23:29, 11.55s/step, acc=0.4959, loss=2.0837, lr=5.85e-05]
Training:  48%|████▊     | 5353/11082 [17:12:05<18:27:51, 11.60s/step, acc=0.4959, loss=2.0837, lr=5.85e-05]
Training:  48%|████▊     | 5354/11082 [17:12:17<18:29:16, 11.62s/step, acc=0.4959, loss=2.0837, lr=5.85e-05]
Training:  48%|████▊     | 5355/11082 [17:12:28<18:14:27, 11.47s/step, acc=0.4959, loss=2.0837, lr=5.85e-05]
Training:  48%|████▊     | 5356/11082 [17:12:40<18:18:45, 11.51s/step, acc=0.4959, loss=2.0837, lr=5.85e-05]
Training:  48%|████▊     | 5357/11082 [17:12:51<18:13:58, 11.47s/step, acc=0.4959, loss=2.0837, lr=5.85e-05]
Training:  48%|████▊     | 5358/11082 [17:13:03<18:25:34, 11.59s/step, acc=0.4959, loss=2.0837, lr=5.85e-05]
Training:  48%|████▊     | 5359/11082 [17:13:15<18:27:11, 11.61s/step, acc=0.4959, loss=2.0837, lr=5.85e-05]
Training:  48%|████▊     | 5360/11082 [17:13:26<18:29:15, 11.63s/step, acc=0.4959, loss=2.0837, lr=5.85e-05]
Training:  48%|████▊     | 5361/11082 [17:13:38<18:27:06, 11.61s/step, acc=0.4959, loss=2.0837, lr=5.85e-05]
Training:  48%|████▊     | 5362/11082 [17:13:49<18:22:13, 11.56s/step, acc=0.4959, loss=2.0837, lr=5.85e-05]
Training:  48%|████▊     | 5363/11082 [17:14:02<18:37:55, 11.73s/step, acc=0.4959, loss=2.0837, lr=5.85e-05]
Training:  48%|████▊     | 5364/11082 [17:14:13<18:18:58, 11.53s/step, acc=0.4959, loss=2.0837, lr=5.85e-05]
Training:  48%|████▊     | 5365/11082 [17:14:24<18:26:59, 11.62s/step, acc=0.4959, loss=2.0837, lr=5.85e-05]
Training:  48%|████▊     | 5366/11082 [17:14:36<18:21:35, 11.56s/step, acc=0.4959, loss=2.0837, lr=5.85e-05]
Training:  48%|████▊     | 5367/11082 [17:14:47<18:13:04, 11.48s/step, acc=0.4959, loss=2.0837, lr=5.85e-05]
Training:  48%|████▊     | 5368/11082 [17:14:59<18:28:34, 11.64s/step, acc=0.4959, loss=2.0837, lr=5.85e-05]
Training:  48%|████▊     | 5369/11082 [17:15:10<18:09:13, 11.44s/step, acc=0.4959, loss=2.0837, lr=5.85e-05]
Training:  48%|████▊     | 5370/11082 [17:15:22<18:32:41, 11.69s/step, acc=0.4959, loss=2.0837, lr=5.85e-05]
Training:  48%|████▊     | 5371/11082 [17:15:34<18:29:42, 11.66s/step, acc=0.4959, loss=2.0837, lr=5.85e-05]
Training:  48%|████▊     | 5372/11082 [17:15:45<18:13:11, 11.49s/step, acc=0.4959, loss=2.0837, lr=5.85e-05]
Training:  48%|████▊     | 5373/11082 [17:15:57<18:18:13, 11.54s/step, acc=0.4959, loss=2.0837, lr=5.85e-05]
Training:  48%|████▊     | 5374/11082 [17:16:08<18:09:17, 11.45s/step, acc=0.4959, loss=2.0837, lr=5.85e-05]
Training:  49%|████▊     | 5375/11082 [17:16:19<18:11:24, 11.47s/step, acc=0.4959, loss=2.0837, lr=5.85e-05]
Training:  49%|████▊     | 5376/11082 [17:16:31<18:09:56, 11.46s/step, acc=0.4959, loss=2.0837, lr=5.85e-05]
Training:  49%|████▊     | 5377/11082 [17:16:43<18:23:25, 11.60s/step, acc=0.4959, loss=2.0837, lr=5.85e-05]
Training:  49%|████▊     | 5378/11082 [17:16:54<18:19:16, 11.56s/step, acc=0.4959, loss=2.0837, lr=5.85e-05]
Training:  49%|████▊     | 5379/11082 [17:17:06<18:11:52, 11.49s/step, acc=0.4959, loss=2.0837, lr=5.85e-05]
Training:  49%|████▊     | 5380/11082 [17:17:17<18:15:51, 11.53s/step, acc=0.4959, loss=2.0837, lr=5.85e-05]
Training:  49%|████▊     | 5381/11082 [17:17:29<18:16:21, 11.54s/step, acc=0.4959, loss=2.0837, lr=5.85e-05]
Training:  49%|████▊     | 5382/11082 [17:17:40<18:17:41, 11.55s/step, acc=0.4959, loss=2.0837, lr=5.85e-05]
Training:  49%|████▊     | 5383/11082 [17:17:52<18:12:32, 11.50s/step, acc=0.4959, loss=2.0837, lr=5.85e-05]
Training:  49%|████▊     | 5384/11082 [17:18:04<18:19:09, 11.57s/step, acc=0.4959, loss=2.0837, lr=5.85e-05]
Training:  49%|████▊     | 5385/11082 [17:18:15<18:14:20, 11.53s/step, acc=0.4959, loss=2.0837, lr=5.85e-05]
Training:  49%|████▊     | 5386/11082 [17:18:26<18:10:50, 11.49s/step, acc=0.4959, loss=2.0837, lr=5.85e-05]
Training:  49%|████▊     | 5387/11082 [17:18:39<18:29:43, 11.69s/step, acc=0.4959, loss=2.0837, lr=5.85e-05]
Training:  49%|████▊     | 5388/11082 [17:18:50<18:19:14, 11.58s/step, acc=0.4959, loss=2.0837, lr=5.85e-05]
Training:  49%|████▊     | 5389/11082 [17:19:01<18:20:57, 11.60s/step, acc=0.4959, loss=2.0837, lr=5.85e-05]
Training:  49%|████▊     | 5390/11082 [17:19:12<18:02:11, 11.41s/step, acc=0.4959, loss=2.0837, lr=5.85e-05]
Training:  49%|████▊     | 5391/11082 [17:19:24<18:15:31, 11.55s/step, acc=0.4959, loss=2.0837, lr=5.85e-05]
Training:  49%|████▊     | 5392/11082 [17:19:36<18:16:33, 11.56s/step, acc=0.4959, loss=2.0837, lr=5.85e-05]
Training:  49%|████▊     | 5393/11082 [17:19:48<18:19:49, 11.60s/step, acc=0.4959, loss=2.0837, lr=5.85e-05]
Training:  49%|████▊     | 5394/11082 [17:19:59<18:15:09, 11.55s/step, acc=0.4959, loss=2.0837, lr=5.85e-05]
Training:  49%|████▊     | 5395/11082 [17:20:11<18:17:35, 11.58s/step, acc=0.4959, loss=2.0837, lr=5.85e-05]
Training:  49%|████▊     | 5396/11082 [17:20:22<18:04:50, 11.45s/step, acc=0.4959, loss=2.0837, lr=5.85e-05]
Training:  49%|████▊     | 5397/11082 [17:20:34<18:23:22, 11.65s/step, acc=0.4959, loss=2.0837, lr=5.85e-05]
Training:  49%|████▊     | 5398/11082 [17:20:45<18:15:47, 11.57s/step, acc=0.4959, loss=2.0837, lr=5.85e-05]
Training:  49%|████▊     | 5399/11082 [17:20:56<18:00:36, 11.41s/step, acc=0.4959, loss=2.0837, lr=5.85e-05]
Training:  49%|████▊     | 5400/11082 [17:21:08<18:06:09, 11.47s/step, acc=0.4959, loss=2.0837, lr=5.85e-05]
Training:  49%|████▊     | 5400/11082 [17:21:15<18:06:09, 11.47s/step, acc=0.4952, loss=2.0891, lr=5.79e-05]2026-06-19 08:14:43 | INFO | __main__ | epoch=0 step=5400 train_loss=2.089141 train_acc=0.495241 lr=5.790e-05 grad_norm=0.9609 acc_d0=0.7593 acc_d1=0.6124 acc_d2=0.4898 acc_d3=0.3910 acc_d4=0.3125 acc_d5=0.2515 acc_d6=0.2058 acc_d7=0.1698

Training:  49%|████▊     | 5401/11082 [17:21:20<18:13:43, 11.55s/step, acc=0.4952, loss=2.0891, lr=5.79e-05]
Training:  49%|████▊     | 5402/11082 [17:21:31<18:15:17, 11.57s/step, acc=0.4952, loss=2.0891, lr=5.79e-05]
Training:  49%|████▉     | 5403/11082 [17:21:43<18:30:08, 11.73s/step, acc=0.4952, loss=2.0891, lr=5.79e-05]
Training:  49%|████▉     | 5404/11082 [17:21:55<18:27:33, 11.70s/step, acc=0.4952, loss=2.0891, lr=5.79e-05]
Training:  49%|████▉     | 5405/11082 [17:22:06<18:13:29, 11.56s/step, acc=0.4952, loss=2.0891, lr=5.79e-05]
Training:  49%|████▉     | 5406/11082 [17:22:18<18:24:04, 11.67s/step, acc=0.4952, loss=2.0891, lr=5.79e-05]
Training:  49%|████▉     | 5407/11082 [17:22:30<18:18:06, 11.61s/step, acc=0.4952, loss=2.0891, lr=5.79e-05]
Training:  49%|████▉     | 5408/11082 [17:22:41<18:14:58, 11.58s/step, acc=0.4952, loss=2.0891, lr=5.79e-05]
Training:  49%|████▉     | 5409/11082 [17:22:52<18:04:44, 11.47s/step, acc=0.4952, loss=2.0891, lr=5.79e-05]
Training:  49%|████▉     | 5410/11082 [17:23:04<18:05:01, 11.48s/step, acc=0.4952, loss=2.0891, lr=5.79e-05]
Training:  49%|████▉     | 5411/11082 [17:23:15<18:07:07, 11.50s/step, acc=0.4952, loss=2.0891, lr=5.79e-05]
Training:  49%|████▉     | 5412/11082 [17:23:27<18:12:14, 11.56s/step, acc=0.4952, loss=2.0891, lr=5.79e-05]
Training:  49%|████▉     | 5413/11082 [17:23:39<18:19:10, 11.63s/step, acc=0.4952, loss=2.0891, lr=5.79e-05]
Training:  49%|████▉     | 5414/11082 [17:23:50<18:03:16, 11.47s/step, acc=0.4952, loss=2.0891, lr=5.79e-05]
Training:  49%|████▉     | 5415/11082 [17:24:02<18:16:02, 11.60s/step, acc=0.4952, loss=2.0891, lr=5.79e-05]
Training:  49%|████▉     | 5416/11082 [17:24:13<18:12:38, 11.57s/step, acc=0.4952, loss=2.0891, lr=5.79e-05]
Training:  49%|████▉     | 5417/11082 [17:24:25<18:06:38, 11.51s/step, acc=0.4952, loss=2.0891, lr=5.79e-05]
Training:  49%|████▉     | 5418/11082 [17:24:37<18:11:06, 11.56s/step, acc=0.4952, loss=2.0891, lr=5.79e-05]
Training:  49%|████▉     | 5419/11082 [17:24:48<18:05:12, 11.50s/step, acc=0.4952, loss=2.0891, lr=5.79e-05]
Training:  49%|████▉     | 5420/11082 [17:25:00<18:10:23, 11.55s/step, acc=0.4952, loss=2.0891, lr=5.79e-05]
Training:  49%|████▉     | 5421/11082 [17:25:11<17:57:18, 11.42s/step, acc=0.4952, loss=2.0891, lr=5.79e-05]
Training:  49%|████▉     | 5422/11082 [17:25:23<18:17:20, 11.63s/step, acc=0.4952, loss=2.0891, lr=5.79e-05]
Training:  49%|████▉     | 5423/11082 [17:25:34<18:09:19, 11.55s/step, acc=0.4952, loss=2.0891, lr=5.79e-05]
Training:  49%|████▉     | 5424/11082 [17:25:46<18:10:11, 11.56s/step, acc=0.4952, loss=2.0891, lr=5.79e-05]
Training:  49%|████▉     | 5425/11082 [17:25:57<18:07:02, 11.53s/step, acc=0.4952, loss=2.0891, lr=5.79e-05]
Training:  49%|████▉     | 5426/11082 [17:26:09<18:17:26, 11.64s/step, acc=0.4952, loss=2.0891, lr=5.79e-05]
Training:  49%|████▉     | 5427/11082 [17:26:20<18:08:50, 11.55s/step, acc=0.4952, loss=2.0891, lr=5.79e-05]
Training:  49%|████▉     | 5428/11082 [17:26:32<18:20:48, 11.68s/step, acc=0.4952, loss=2.0891, lr=5.79e-05]
Training:  49%|████▉     | 5429/11082 [17:26:44<18:11:06, 11.58s/step, acc=0.4952, loss=2.0891, lr=5.79e-05]
Training:  49%|████▉     | 5430/11082 [17:26:55<18:00:04, 11.47s/step, acc=0.4952, loss=2.0891, lr=5.79e-05]
Training:  49%|████▉     | 5431/11082 [17:27:07<18:02:25, 11.49s/step, acc=0.4952, loss=2.0891, lr=5.79e-05]
Training:  49%|████▉     | 5432/11082 [17:27:18<18:09:39, 11.57s/step, acc=0.4952, loss=2.0891, lr=5.79e-05]
Training:  49%|████▉     | 5433/11082 [17:27:30<18:03:00, 11.50s/step, acc=0.4952, loss=2.0891, lr=5.79e-05]
Training:  49%|████▉     | 5434/11082 [17:27:42<18:19:05, 11.68s/step, acc=0.4952, loss=2.0891, lr=5.79e-05]
Training:  49%|████▉     | 5435/11082 [17:27:54<18:27:39, 11.77s/step, acc=0.4952, loss=2.0891, lr=5.79e-05]
Training:  49%|████▉     | 5436/11082 [17:28:05<18:23:06, 11.72s/step, acc=0.4952, loss=2.0891, lr=5.79e-05]
Training:  49%|████▉     | 5437/11082 [17:28:17<18:12:46, 11.61s/step, acc=0.4952, loss=2.0891, lr=5.79e-05]
Training:  49%|████▉     | 5438/11082 [17:28:28<18:09:31, 11.58s/step, acc=0.4952, loss=2.0891, lr=5.79e-05]
Training:  49%|████▉     | 5439/11082 [17:28:40<18:08:57, 11.58s/step, acc=0.4952, loss=2.0891, lr=5.79e-05]
Training:  49%|████▉     | 5440/11082 [17:28:51<17:55:26, 11.44s/step, acc=0.4952, loss=2.0891, lr=5.79e-05]
Training:  49%|████▉     | 5441/11082 [17:29:03<18:04:19, 11.53s/step, acc=0.4952, loss=2.0891, lr=5.79e-05]
Training:  49%|████▉     | 5442/11082 [17:29:14<18:04:25, 11.54s/step, acc=0.4952, loss=2.0891, lr=5.79e-05]
Training:  49%|████▉     | 5443/11082 [17:29:25<17:59:00, 11.48s/step, acc=0.4952, loss=2.0891, lr=5.79e-05]
Training:  49%|████▉     | 5444/11082 [17:29:37<18:07:30, 11.57s/step, acc=0.4952, loss=2.0891, lr=5.79e-05]
Training:  49%|████▉     | 5445/11082 [17:29:49<17:58:11, 11.48s/step, acc=0.4952, loss=2.0891, lr=5.79e-05]
Training:  49%|████▉     | 5446/11082 [17:30:00<18:08:17, 11.59s/step, acc=0.4952, loss=2.0891, lr=5.79e-05]
Training:  49%|████▉     | 5447/11082 [17:30:12<18:08:43, 11.59s/step, acc=0.4952, loss=2.0891, lr=5.79e-05]
Training:  49%|████▉     | 5448/11082 [17:30:23<17:51:51, 11.41s/step, acc=0.4952, loss=2.0891, lr=5.79e-05]
Training:  49%|████▉     | 5449/11082 [17:30:35<18:09:14, 11.60s/step, acc=0.4952, loss=2.0891, lr=5.79e-05]
Training:  49%|████▉     | 5450/11082 [17:30:46<17:56:28, 11.47s/step, acc=0.4952, loss=2.0891, lr=5.79e-05]
Training:  49%|████▉     | 5450/11082 [17:30:54<17:56:28, 11.47s/step, acc=0.4970, loss=2.0762, lr=5.73e-05]2026-06-19 08:24:21 | INFO | __main__ | epoch=0 step=5450 train_loss=2.076246 train_acc=0.496974 lr=5.725e-05 grad_norm=1.1875 acc_d0=0.7781 acc_d1=0.6332 acc_d2=0.5097 acc_d3=0.4083 acc_d4=0.3259 acc_d5=0.2606 acc_d6=0.2097 acc_d7=0.1716

Training:  49%|████▉     | 5451/11082 [17:30:58<18:08:19, 11.60s/step, acc=0.4970, loss=2.0762, lr=5.73e-05]
Training:  49%|████▉     | 5452/11082 [17:31:10<18:04:07, 11.55s/step, acc=0.4970, loss=2.0762, lr=5.73e-05]
Training:  49%|████▉     | 5453/11082 [17:31:22<18:17:18, 11.70s/step, acc=0.4970, loss=2.0762, lr=5.73e-05]
Training:  49%|████▉     | 5454/11082 [17:31:33<18:07:21, 11.59s/step, acc=0.4970, loss=2.0762, lr=5.73e-05]
Training:  49%|████▉     | 5455/11082 [17:31:44<17:58:45, 11.50s/step, acc=0.4970, loss=2.0762, lr=5.73e-05]
Training:  49%|████▉     | 5456/11082 [17:31:56<17:57:39, 11.49s/step, acc=0.4970, loss=2.0762, lr=5.73e-05]
Training:  49%|████▉     | 5457/11082 [17:32:08<18:09:29, 11.62s/step, acc=0.4970, loss=2.0762, lr=5.73e-05]
Training:  49%|████▉     | 5458/11082 [17:32:19<18:08:22, 11.61s/step, acc=0.4970, loss=2.0762, lr=5.73e-05]
Training:  49%|████▉     | 5459/11082 [17:32:31<18:04:44, 11.57s/step, acc=0.4970, loss=2.0762, lr=5.73e-05]
Training:  49%|████▉     | 5460/11082 [17:32:42<17:48:12, 11.40s/step, acc=0.4970, loss=2.0762, lr=5.73e-05]
Training:  49%|████▉     | 5461/11082 [17:32:53<17:54:04, 11.47s/step, acc=0.4970, loss=2.0762, lr=5.73e-05]
Training:  49%|████▉     | 5462/11082 [17:33:05<17:49:11, 11.41s/step, acc=0.4970, loss=2.0762, lr=5.73e-05]
Training:  49%|████▉     | 5463/11082 [17:33:16<18:03:23, 11.57s/step, acc=0.4970, loss=2.0762, lr=5.73e-05]
Training:  49%|████▉     | 5464/11082 [17:33:28<18:00:41, 11.54s/step, acc=0.4970, loss=2.0762, lr=5.73e-05]
Training:  49%|████▉     | 5465/11082 [17:33:40<18:05:02, 11.59s/step, acc=0.4970, loss=2.0762, lr=5.73e-05]
Training:  49%|████▉     | 5466/11082 [17:33:51<18:01:57, 11.56s/step, acc=0.4970, loss=2.0762, lr=5.73e-05]
Training:  49%|████▉     | 5467/11082 [17:34:03<18:02:49, 11.57s/step, acc=0.4970, loss=2.0762, lr=5.73e-05]
Training:  49%|████▉     | 5468/11082 [17:34:14<18:02:08, 11.57s/step, acc=0.4970, loss=2.0762, lr=5.73e-05]
Training:  49%|████▉     | 5469/11082 [17:34:26<18:02:45, 11.57s/step, acc=0.4970, loss=2.0762, lr=5.73e-05]
Training:  49%|████▉     | 5470/11082 [17:34:37<17:53:28, 11.48s/step, acc=0.4970, loss=2.0762, lr=5.73e-05]
Training:  49%|████▉     | 5471/11082 [17:34:49<18:12:39, 11.68s/step, acc=0.4970, loss=2.0762, lr=5.73e-05]
Training:  49%|████▉     | 5472/11082 [17:35:00<17:56:49, 11.52s/step, acc=0.4970, loss=2.0762, lr=5.73e-05]
Training:  49%|████▉     | 5473/11082 [17:35:12<18:02:55, 11.58s/step, acc=0.4970, loss=2.0762, lr=5.73e-05]
Training:  49%|████▉     | 5474/11082 [17:35:24<18:11:22, 11.68s/step, acc=0.4970, loss=2.0762, lr=5.73e-05]
Training:  49%|████▉     | 5475/11082 [17:35:36<18:12:16, 11.69s/step, acc=0.4970, loss=2.0762, lr=5.73e-05]
Training:  49%|████▉     | 5476/11082 [17:35:47<18:00:19, 11.56s/step, acc=0.4970, loss=2.0762, lr=5.73e-05]
Training:  49%|████▉     | 5477/11082 [17:35:59<18:11:50, 11.69s/step, acc=0.4970, loss=2.0762, lr=5.73e-05]
Training:  49%|████▉     | 5478/11082 [17:36:10<17:57:45, 11.54s/step, acc=0.4970, loss=2.0762, lr=5.73e-05]
Training:  49%|████▉     | 5479/11082 [17:36:23<18:19:39, 11.78s/step, acc=0.4970, loss=2.0762, lr=5.73e-05]
Training:  49%|████▉     | 5480/11082 [17:36:34<18:10:43, 11.68s/step, acc=0.4970, loss=2.0762, lr=5.73e-05]
Training:  49%|████▉     | 5481/11082 [17:36:45<17:58:25, 11.55s/step, acc=0.4970, loss=2.0762, lr=5.73e-05]
Training:  49%|████▉     | 5482/11082 [17:36:57<18:01:21, 11.59s/step, acc=0.4970, loss=2.0762, lr=5.73e-05]
Training:  49%|████▉     | 5483/11082 [17:37:08<17:59:42, 11.57s/step, acc=0.4970, loss=2.0762, lr=5.73e-05]
Training:  49%|████▉     | 5484/11082 [17:37:20<18:08:47, 11.67s/step, acc=0.4970, loss=2.0762, lr=5.73e-05]
Training:  49%|████▉     | 5485/11082 [17:37:32<18:03:19, 11.61s/step, acc=0.4970, loss=2.0762, lr=5.73e-05]
Training:  50%|████▉     | 5486/11082 [17:37:43<17:58:56, 11.57s/step, acc=0.4970, loss=2.0762, lr=5.73e-05]
Training:  50%|████▉     | 5487/11082 [17:37:55<18:09:50, 11.69s/step, acc=0.4970, loss=2.0762, lr=5.73e-05]
Training:  50%|████▉     | 5488/11082 [17:38:07<18:07:12, 11.66s/step, acc=0.4970, loss=2.0762, lr=5.73e-05]
Training:  50%|████▉     | 5489/11082 [17:38:19<18:05:29, 11.64s/step, acc=0.4970, loss=2.0762, lr=5.73e-05]
Training:  50%|████▉     | 5490/11082 [17:38:30<18:08:19, 11.68s/step, acc=0.4970, loss=2.0762, lr=5.73e-05]
Training:  50%|████▉     | 5491/11082 [17:38:42<17:56:49, 11.56s/step, acc=0.4970, loss=2.0762, lr=5.73e-05]
Training:  50%|████▉     | 5492/11082 [17:38:53<17:57:21, 11.56s/step, acc=0.4970, loss=2.0762, lr=5.73e-05]
Training:  50%|████▉     | 5493/11082 [17:39:05<17:53:35, 11.53s/step, acc=0.4970, loss=2.0762, lr=5.73e-05]
Training:  50%|████▉     | 5494/11082 [17:39:16<17:57:06, 11.57s/step, acc=0.4970, loss=2.0762, lr=5.73e-05]
Training:  50%|████▉     | 5495/11082 [17:39:28<17:53:52, 11.53s/step, acc=0.4970, loss=2.0762, lr=5.73e-05]
Training:  50%|████▉     | 5496/11082 [17:39:39<17:59:01, 11.59s/step, acc=0.4970, loss=2.0762, lr=5.73e-05]
Training:  50%|████▉     | 5497/11082 [17:39:51<17:51:32, 11.51s/step, acc=0.4970, loss=2.0762, lr=5.73e-05]
Training:  50%|████▉     | 5498/11082 [17:40:03<18:06:43, 11.68s/step, acc=0.4970, loss=2.0762, lr=5.73e-05]
Training:  50%|████▉     | 5499/11082 [17:40:14<17:50:20, 11.50s/step, acc=0.4970, loss=2.0762, lr=5.73e-05]
Training:  50%|████▉     | 5500/11082 [17:40:25<17:50:26, 11.51s/step, acc=0.4970, loss=2.0762, lr=5.73e-05]
Training:  50%|████▉     | 5500/11082 [17:40:33<17:50:26, 11.51s/step, acc=0.4977, loss=2.0688, lr=5.66e-05]2026-06-19 08:34:01 | INFO | __main__ | epoch=0 step=5500 train_loss=2.068821 train_acc=0.497699 lr=5.661e-05 grad_norm=0.8398 acc_d0=0.7719 acc_d1=0.6261 acc_d2=0.5033 acc_d3=0.4027 acc_d4=0.3211 acc_d5=0.2576 acc_d6=0.2086 acc_d7=0.1703

Training:  50%|████▉     | 5501/11082 [17:40:37<17:52:20, 11.53s/step, acc=0.4977, loss=2.0688, lr=5.66e-05]
Training:  50%|████▉     | 5502/11082 [17:40:49<17:54:44, 11.56s/step, acc=0.4977, loss=2.0688, lr=5.66e-05]
Training:  50%|████▉     | 5503/11082 [17:41:00<18:02:43, 11.64s/step, acc=0.4977, loss=2.0688, lr=5.66e-05]
Training:  50%|████▉     | 5504/11082 [17:41:12<18:08:35, 11.71s/step, acc=0.4977, loss=2.0688, lr=5.66e-05]
Training:  50%|████▉     | 5505/11082 [17:41:24<18:09:51, 11.73s/step, acc=0.4977, loss=2.0688, lr=5.66e-05]
Training:  50%|████▉     | 5506/11082 [17:41:36<18:03:52, 11.66s/step, acc=0.4977, loss=2.0688, lr=5.66e-05]
Training:  50%|████▉     | 5507/11082 [17:41:47<17:58:11, 11.60s/step, acc=0.4977, loss=2.0688, lr=5.66e-05]
Training:  50%|████▉     | 5508/11082 [17:41:59<18:00:54, 11.64s/step, acc=0.4977, loss=2.0688, lr=5.66e-05]
Training:  50%|████▉     | 5509/11082 [17:42:10<17:49:01, 11.51s/step, acc=0.4977, loss=2.0688, lr=5.66e-05]
Training:  50%|████▉     | 5510/11082 [17:42:22<18:01:57, 11.65s/step, acc=0.4977, loss=2.0688, lr=5.66e-05]
Training:  50%|████▉     | 5511/11082 [17:42:34<18:04:25, 11.68s/step, acc=0.4977, loss=2.0688, lr=5.66e-05]
Training:  50%|████▉     | 5512/11082 [17:42:45<17:59:17, 11.63s/step, acc=0.4977, loss=2.0688, lr=5.66e-05]
Training:  50%|████▉     | 5513/11082 [17:42:56<17:50:00, 11.53s/step, acc=0.4977, loss=2.0688, lr=5.66e-05]
Training:  50%|████▉     | 5514/11082 [17:43:08<17:54:23, 11.58s/step, acc=0.4977, loss=2.0688, lr=5.66e-05]
Training:  50%|████▉     | 5515/11082 [17:43:19<17:46:15, 11.49s/step, acc=0.4977, loss=2.0688, lr=5.66e-05]
Training:  50%|████▉     | 5516/11082 [17:43:31<17:50:18, 11.54s/step, acc=0.4977, loss=2.0688, lr=5.66e-05]
Training:  50%|████▉     | 5517/11082 [17:43:42<17:39:14, 11.42s/step, acc=0.4977, loss=2.0688, lr=5.66e-05]
Training:  50%|████▉     | 5518/11082 [17:43:54<17:43:38, 11.47s/step, acc=0.4977, loss=2.0688, lr=5.66e-05]
Training:  50%|████▉     | 5519/11082 [17:44:06<17:57:15, 11.62s/step, acc=0.4977, loss=2.0688, lr=5.66e-05]
Training:  50%|████▉     | 5520/11082 [17:44:17<17:50:04, 11.54s/step, acc=0.4977, loss=2.0688, lr=5.66e-05]
Training:  50%|████▉     | 5521/11082 [17:44:29<17:49:08, 11.54s/step, acc=0.4977, loss=2.0688, lr=5.66e-05]
Training:  50%|████▉     | 5522/11082 [17:44:41<17:58:46, 11.64s/step, acc=0.4977, loss=2.0688, lr=5.66e-05]
Training:  50%|████▉     | 5523/11082 [17:44:52<17:48:40, 11.53s/step, acc=0.4977, loss=2.0688, lr=5.66e-05]
Training:  50%|████▉     | 5524/11082 [17:45:03<17:41:23, 11.46s/step, acc=0.4977, loss=2.0688, lr=5.66e-05]
Training:  50%|████▉     | 5525/11082 [17:45:15<17:40:07, 11.45s/step, acc=0.4977, loss=2.0688, lr=5.66e-05]
Training:  50%|████▉     | 5526/11082 [17:45:27<17:57:34, 11.64s/step, acc=0.4977, loss=2.0688, lr=5.66e-05]
Training:  50%|████▉     | 5527/11082 [17:45:38<17:48:27, 11.54s/step, acc=0.4977, loss=2.0688, lr=5.66e-05]
Training:  50%|████▉     | 5528/11082 [17:45:49<17:42:48, 11.48s/step, acc=0.4977, loss=2.0688, lr=5.66e-05]
Training:  50%|████▉     | 5529/11082 [17:46:01<17:58:27, 11.65s/step, acc=0.4977, loss=2.0688, lr=5.66e-05]
Training:  50%|████▉     | 5530/11082 [17:46:13<17:44:15, 11.50s/step, acc=0.4977, loss=2.0688, lr=5.66e-05]
Training:  50%|████▉     | 5531/11082 [17:46:24<17:53:46, 11.61s/step, acc=0.4977, loss=2.0688, lr=5.66e-05]
Training:  50%|████▉     | 5532/11082 [17:46:36<17:56:03, 11.63s/step, acc=0.4977, loss=2.0688, lr=5.66e-05]
Training:  50%|████▉     | 5533/11082 [17:46:47<17:46:30, 11.53s/step, acc=0.4977, loss=2.0688, lr=5.66e-05]
Training:  50%|████▉     | 5534/11082 [17:46:59<17:56:30, 11.64s/step, acc=0.4977, loss=2.0688, lr=5.66e-05]
Training:  50%|████▉     | 5535/11082 [17:47:11<17:50:49, 11.58s/step, acc=0.4977, loss=2.0688, lr=5.66e-05]
Training:  50%|████▉     | 5536/11082 [17:47:22<17:49:22, 11.57s/step, acc=0.4977, loss=2.0688, lr=5.66e-05]
Training:  50%|████▉     | 5537/11082 [17:47:34<17:52:19, 11.60s/step, acc=0.4977, loss=2.0688, lr=5.66e-05]
Training:  50%|████▉     | 5538/11082 [17:47:45<17:44:14, 11.52s/step, acc=0.4977, loss=2.0688, lr=5.66e-05]
Training:  50%|████▉     | 5539/11082 [17:47:56<17:36:40, 11.44s/step, acc=0.4977, loss=2.0688, lr=5.66e-05]
Training:  50%|████▉     | 5540/11082 [17:48:08<17:39:57, 11.48s/step, acc=0.4977, loss=2.0688, lr=5.66e-05]
Training:  50%|█████     | 5541/11082 [17:48:19<17:16:44, 11.23s/step, acc=0.4977, loss=2.0688, lr=5.66e-05]2026-06-19 08:41:54 | INFO | __main__ | epoch=0 step=5541 train_loss=2.076469 train_acc=0.496263 lr=5.607e-05 grad_norm=0.8242 acc_d0=0.7575 acc_d1=0.6124 acc_d2=0.4908 acc_d3=0.3926 acc_d4=0.3138 acc_d5=0.2522 acc_d6=0.2059 acc_d7=0.1692 (trailing flush)
2026-06-19 08:41:54 | INFO | __main__ | Epoch 0 done: total_batches_seen=44326 global_step=5541
/llm-align/liuchonghan/env/envs/automodel/lib/python3.12/site-packages/torch/distributed/c10d_logger.py:83: UserWarning: barrier(): using the device under current context. You can specify `device_id` in `init_process_group` to mute this warning.
  return func(*args, **kwargs)
h-liuchonghan-translate-a8nodes-main-m-0:66737:66737 [5] NCCL INFO Channel 00/1 : 5[5] -> 0[0] via P2P/CUMEM/read
h-liuchonghan-translate-a8nodes-main-m-0:66736:66736 [4] NCCL INFO Channel 00/1 : 4[4] -> 0[0] via P2P/CUMEM/read
h-liuchonghan-translate-a8nodes-main-m-0:66738:66738 [6] NCCL INFO Channel 00/1 : 6[6] -> 0[0] via P2P/CUMEM/read
h-liuchonghan-translate-a8nodes-main-m-0:66735:66735 [3] NCCL INFO Channel 00/1 : 3[3] -> 0[0] via P2P/CUMEM/read
h-liuchonghan-translate-a8nodes-main-m-0:66739:66739 [7] NCCL INFO Channel 00/1 : 7[7] -> 0[0] via P2P/CUMEM/read
h-liuchonghan-translate-a8nodes-main-m-0:66734:66734 [2] NCCL INFO Channel 00/1 : 2[2] -> 0[0] via P2P/CUMEM/read
h-liuchonghan-translate-a8nodes-main-m-0:66733:66733 [1] NCCL INFO Channel 00/1 : 1[1] -> 0[0] via P2P/CUMEM/read
h-liuchonghan-translate-a8nodes-main-m-0:66732:66732 [0] NCCL INFO Channel 01/1 : 31[7] -> 0[0] [receive] via NET/Socket/0/Shared
h-liuchonghan-translate-a8nodes-main-m-0:66732:66732 [0] NCCL INFO Channel 01/1 : 30[6] -> 0[0] [receive] via NET/Socket/0/Shared
h-liuchonghan-translate-a8nodes-main-m-0:66732:66732 [0] NCCL INFO Channel 01/1 : 29[5] -> 0[0] [receive] via NET/Socket/0/Shared
h-liuchonghan-translate-a8nodes-main-m-0:66732:66732 [0] NCCL INFO Channel 01/1 : 28[4] -> 0[0] [receive] via NET/Socket/0/Shared
h-liuchonghan-translate-a8nodes-main-m-0:66732:66732 [0] NCCL INFO Channel 01/1 : 27[3] -> 0[0] [receive] via NET/Socket/0/Shared
h-liuchonghan-translate-a8nodes-main-m-0:66732:66732 [0] NCCL INFO Channel 01/1 : 26[2] -> 0[0] [receive] via NET/Socket/0/Shared
h-liuchonghan-translate-a8nodes-main-m-0:66732:66732 [0] NCCL INFO Channel 01/1 : 25[1] -> 0[0] [receive] via NET/Socket/0/Shared
h-liuchonghan-translate-a8nodes-main-m-0:66732:66732 [0] NCCL INFO Channel 01/1 : 24[0] -> 0[0] [receive] via NET/Socket/0/Shared
h-liuchonghan-translate-a8nodes-main-m-0:66732:66732 [0] NCCL INFO Channel 01/1 : 23[7] -> 0[0] [receive] via NET/Socket/0/Shared
h-liuchonghan-translate-a8nodes-main-m-0:66732:66732 [0] NCCL INFO Channel 01/1 : 22[6] -> 0[0] [receive] via NET/Socket/0/Shared
h-liuchonghan-translate-a8nodes-main-m-0:66732:66732 [0] NCCL INFO Channel 01/1 : 21[5] -> 0[0] [receive] via NET/Socket/0/Shared
h-liuchonghan-translate-a8nodes-main-m-0:66732:66732 [0] NCCL INFO Channel 01/1 : 20[4] -> 0[0] [receive] via NET/Socket/0/Shared
h-liuchonghan-translate-a8nodes-main-m-0:66732:66732 [0] NCCL INFO Channel 01/1 : 19[3] -> 0[0] [receive] via NET/Socket/0/Shared
h-liuchonghan-translate-a8nodes-main-m-0:66732:66732 [0] NCCL INFO Channel 01/1 : 18[2] -> 0[0] [receive] via NET/Socket/0/Shared
h-liuchonghan-translate-a8nodes-main-m-0:66732:66732 [0] NCCL INFO Channel 01/1 : 17[1] -> 0[0] [receive] via NET/Socket/0/Shared
h-liuchonghan-translate-a8nodes-main-m-0:66732:66732 [0] NCCL INFO Channel 01/1 : 16[0] -> 0[0] [receive] via NET/Socket/0/Shared
h-liuchonghan-translate-a8nodes-main-m-0:66732:66732 [0] NCCL INFO Channel 00/1 : 15[7] -> 0[0] [receive] via NET/Socket/0/Shared
h-liuchonghan-translate-a8nodes-main-m-0:66732:66732 [0] NCCL INFO Channel 00/1 : 14[6] -> 0[0] [receive] via NET/Socket/0/Shared
h-liuchonghan-translate-a8nodes-main-m-0:66732:66732 [0] NCCL INFO Channel 00/1 : 13[5] -> 0[0] [receive] via NET/Socket/0/Shared
h-liuchonghan-translate-a8nodes-main-m-0:66732:66732 [0] NCCL INFO Channel 00/1 : 12[4] -> 0[0] [receive] via NET/Socket/0/Shared
h-liuchonghan-translate-a8nodes-main-m-0:66732:66732 [0] NCCL INFO Channel 00/1 : 11[3] -> 0[0] [receive] via NET/Socket/0/Shared
h-liuchonghan-translate-a8nodes-main-m-0:66732:66732 [0] NCCL INFO Channel 00/1 : 10[2] -> 0[0] [receive] via NET/Socket/0/Shared
h-liuchonghan-translate-a8nodes-main-m-0:66732:66732 [0] NCCL INFO Channel 00/1 : 9[1] -> 0[0] [receive] via NET/Socket/0/Shared
h-liuchonghan-translate-a8nodes-main-m-0:66732:66732 [0] NCCL INFO Channel 00/1 : 8[0] -> 0[0] [receive] via NET/Socket/0/Shared
h-liuchonghan-translate-a8nodes-main-m-0:66732:66732 [0] NCCL INFO Channel 00/1 : 0[0] -> 1[1] via P2P/CUMEM/read
h-liuchonghan-translate-a8nodes-main-m-0:66732:66732 [0] NCCL INFO Channel 00/1 : 0[0] -> 2[2] via P2P/CUMEM/read
h-liuchonghan-translate-a8nodes-main-m-0:66732:66732 [0] NCCL INFO Channel 00/1 : 0[0] -> 3[3] via P2P/CUMEM/read
h-liuchonghan-translate-a8nodes-main-m-0:66732:66732 [0] NCCL INFO Channel 00/1 : 0[0] -> 4[4] via P2P/CUMEM/read
h-liuchonghan-translate-a8nodes-main-m-0:66732:66732 [0] NCCL INFO Channel 00/1 : 0[0] -> 5[5] via P2P/CUMEM/read
h-liuchonghan-translate-a8nodes-main-m-0:66732:66732 [0] NCCL INFO Channel 00/1 : 0[0] -> 6[6] via P2P/CUMEM/read
h-liuchonghan-translate-a8nodes-main-m-0:66732:66732 [0] NCCL INFO Channel 00/1 : 0[0] -> 7[7] via P2P/CUMEM/read
h-liuchonghan-translate-a8nodes-main-m-0:66732:66732 [0] NCCL INFO Channel 01/1 : 0[0] -> 8[0] [send] via NET/Socket/0/Shared
h-liuchonghan-translate-a8nodes-main-m-0:66732:66732 [0] NCCL INFO Channel 01/1 : 0[0] -> 9[1] [send] via NET/Socket/0/Shared
h-liuchonghan-translate-a8nodes-main-m-0:66732:66732 [0] NCCL INFO Channel 01/1 : 0[0] -> 10[2] [send] via NET/Socket/0/Shared
h-liuchonghan-translate-a8nodes-main-m-0:66732:66732 [0] NCCL INFO Channel 01/1 : 0[0] -> 11[3] [send] via NET/Socket/0/Shared
h-liuchonghan-translate-a8nodes-main-m-0:66732:66732 [0] NCCL INFO Channel 01/1 : 0[0] -> 12[4] [send] via NET/Socket/0/Shared
h-liuchonghan-translate-a8nodes-main-m-0:66732:66732 [0] NCCL INFO Channel 01/1 : 0[0] -> 13[5] [send] via NET/Socket/0/Shared
h-liuchonghan-translate-a8nodes-main-m-0:66732:66732 [0] NCCL INFO Channel 01/1 : 0[0] -> 14[6] [send] via NET/Socket/0/Shared
h-liuchonghan-translate-a8nodes-main-m-0:66732:66732 [0] NCCL INFO Channel 01/1 : 0[0] -> 15[7] [send] via NET/Socket/0/Shared
h-liuchonghan-translate-a8nodes-main-m-0:66732:66732 [0] NCCL INFO Channel 01/1 : 0[0] -> 16[0] [send] via NET/Socket/0/Shared
h-liuchonghan-translate-a8nodes-main-m-0:66732:66732 [0] NCCL INFO Channel 01/1 : 0[0] -> 17[1] [send] via NET/Socket/0/Shared
h-liuchonghan-translate-a8nodes-main-m-0:66732:66732 [0] NCCL INFO Channel 01/1 : 0[0] -> 18[2] [send] via NET/Socket/0/Shared
h-liuchonghan-translate-a8nodes-main-m-0:66732:66732 [0] NCCL INFO Channel 01/1 : 0[0] -> 19[3] [send] via NET/Socket/0/Shared
h-liuchonghan-translate-a8nodes-main-m-0:66732:66732 [0] NCCL INFO Channel 01/1 : 0[0] -> 20[4] [send] via NET/Socket/0/Shared
h-liuchonghan-translate-a8nodes-main-m-0:66732:66732 [0] NCCL INFO Channel 01/1 : 0[0] -> 21[5] [send] via NET/Socket/0/Shared
h-liuchonghan-translate-a8nodes-main-m-0:66732:66732 [0] NCCL INFO Channel 01/1 : 0[0] -> 22[6] [send] via NET/Socket/0/Shared
h-liuchonghan-translate-a8nodes-main-m-0:66732:66732 [0] NCCL INFO Channel 01/1 : 0[0] -> 23[7] [send] via NET/Socket/0/Shared
h-liuchonghan-translate-a8nodes-main-m-0:66732:66732 [0] NCCL INFO Channel 00/1 : 0[0] -> 24[0] [send] via NET/Socket/0/Shared
h-liuchonghan-translate-a8nodes-main-m-0:66732:66732 [0] NCCL INFO Channel 00/1 : 0[0] -> 25[1] [send] via NET/Socket/0/Shared
h-liuchonghan-translate-a8nodes-main-m-0:66732:66732 [0] NCCL INFO Channel 00/1 : 0[0] -> 26[2] [send] via NET/Socket/0/Shared
h-liuchonghan-translate-a8nodes-main-m-0:66732:66732 [0] NCCL INFO Channel 00/1 : 0[0] -> 27[3] [send] via NET/Socket/0/Shared
h-liuchonghan-translate-a8nodes-main-m-0:66732:66732 [0] NCCL INFO Channel 00/1 : 0[0] -> 28[4] [send] via NET/Socket/0/Shared
h-liuchonghan-translate-a8nodes-main-m-0:66732:66732 [0] NCCL INFO Channel 00/1 : 0[0] -> 29[5] [send] via NET/Socket/0/Shared
h-liuchonghan-translate-a8nodes-main-m-0:66732:66732 [0] NCCL INFO Channel 00/1 : 0[0] -> 30[6] [send] via NET/Socket/0/Shared
h-liuchonghan-translate-a8nodes-main-m-0:66732:66732 [0] NCCL INFO Channel 00/1 : 0[0] -> 31[7] [send] via NET/Socket/0/Shared
2026-06-19 08:41:56 | INFO | nemo_automodel.components.checkpoint._backports.consolidate_hf_safetensors | Consolidating safetensors files from /llm-align/liuchonghan/runs/p_eagle_kv_reuse/checkpoints/epoch_1_step_5541/model to /llm-align/liuchonghan/runs/p_eagle_kv_reuse/checkpoints/epoch_1_step_5541/model/consolidated.
2026-06-19 08:42:02 | INFO | nemo_automodel.components.checkpoint.checkpointing | Successfully exported consolidated HF safetensors to /llm-align/liuchonghan/runs/p_eagle_kv_reuse/checkpoints/epoch_1_step_5541/model/consolidated.
2026-06-19 08:42:05 | INFO | __main__ | Saved epoch checkpoint to /llm-align/liuchonghan/runs/p_eagle_kv_reuse/checkpoints/epoch_1_step_5541

Training:  50%|█████     | 5542/11082 [17:48:42<22:48:22, 14.82s/step, acc=0.4977, loss=2.0688, lr=5.66e-05]
Training:  50%|█████     | 5543/11082 [17:48:54<21:36:46, 14.05s/step, acc=0.4977, loss=2.0688, lr=5.66e-05]
Training:  50%|█████     | 5544/11082 [17:49:07<20:55:17, 13.60s/step, acc=0.4977, loss=2.0688, lr=5.66e-05]
Training:  50%|█████     | 5545/11082 [17:49:18<19:54:02, 12.94s/step, acc=0.4977, loss=2.0688, lr=5.66e-05]
Training:  50%|█████     | 5546/11082 [17:49:30<19:15:23, 12.52s/step, acc=0.4977, loss=2.0688, lr=5.66e-05]
Training:  50%|█████     | 5547/11082 [17:49:41<18:55:45, 12.31s/step, acc=0.4977, loss=2.0688, lr=5.66e-05]
Training:  50%|█████     | 5548/11082 [17:49:53<18:25:33, 11.99s/step, acc=0.4977, loss=2.0688, lr=5.66e-05]
Training:  50%|█████     | 5549/11082 [17:50:04<18:17:02, 11.90s/step, acc=0.4977, loss=2.0688, lr=5.66e-05]
Training:  50%|█████     | 5550/11082 [17:50:16<18:07:16, 11.79s/step, acc=0.4977, loss=2.0688, lr=5.66e-05]
Training:  50%|█████     | 5550/11082 [17:50:23<18:07:16, 11.79s/step, acc=0.4957, loss=2.0893, lr=5.60e-05]2026-06-19 08:43:51 | INFO | __main__ | epoch=1 step=5550 train_loss=2.089281 train_acc=0.495748 lr=5.596e-05 grad_norm=0.7695 acc_d0=0.7677 acc_d1=0.6199 acc_d2=0.4974 acc_d3=0.3932 acc_d4=0.3149 acc_d5=0.2543 acc_d6=0.2052 acc_d7=0.1678

Training:  50%|█████     | 5551/11082 [17:50:27<17:59:14, 11.71s/step, acc=0.4957, loss=2.0893, lr=5.60e-05]
Training:  50%|█████     | 5552/11082 [17:50:39<18:07:39, 11.80s/step, acc=0.4957, loss=2.0893, lr=5.60e-05]
Training:  50%|█████     | 5553/11082 [17:50:51<17:47:34, 11.59s/step, acc=0.4957, loss=2.0893, lr=5.60e-05]
Training:  50%|█████     | 5554/11082 [17:51:02<17:55:52, 11.68s/step, acc=0.4957, loss=2.0893, lr=5.60e-05]
Training:  50%|█████     | 5555/11082 [17:51:14<17:53:49, 11.66s/step, acc=0.4957, loss=2.0893, lr=5.60e-05]
Training:  50%|█████     | 5556/11082 [17:51:26<18:14:38, 11.89s/step, acc=0.4957, loss=2.0893, lr=5.60e-05]
Training:  50%|█████     | 5557/11082 [17:51:38<17:58:44, 11.71s/step, acc=0.4957, loss=2.0893, lr=5.60e-05]
Training:  50%|█████     | 5558/11082 [17:51:49<17:56:26, 11.69s/step, acc=0.4957, loss=2.0893, lr=5.60e-05]
Training:  50%|█████     | 5559/11082 [17:52:01<17:54:41, 11.68s/step, acc=0.4957, loss=2.0893, lr=5.60e-05]
Training:  50%|█████     | 5560/11082 [17:52:13<17:49:00, 11.62s/step, acc=0.4957, loss=2.0893, lr=5.60e-05]
Training:  50%|█████     | 5561/11082 [17:52:24<17:45:00, 11.57s/step, acc=0.4957, loss=2.0893, lr=5.60e-05]
Training:  50%|█████     | 5562/11082 [17:52:35<17:41:01, 11.53s/step, acc=0.4957, loss=2.0893, lr=5.60e-05]
Training:  50%|█████     | 5563/11082 [17:52:47<17:43:41, 11.56s/step, acc=0.4957, loss=2.0893, lr=5.60e-05]
Training:  50%|█████     | 5564/11082 [17:52:58<17:34:00, 11.46s/step, acc=0.4957, loss=2.0893, lr=5.60e-05]
Training:  50%|█████     | 5565/11082 [17:53:10<17:49:08, 11.63s/step, acc=0.4957, loss=2.0893, lr=5.60e-05]
Training:  50%|█████     | 5566/11082 [17:53:22<17:45:37, 11.59s/step, acc=0.4957, loss=2.0893, lr=5.60e-05]
Training:  50%|█████     | 5567/11082 [17:53:33<17:39:19, 11.52s/step, acc=0.4957, loss=2.0893, lr=5.60e-05]
Training:  50%|█████     | 5568/11082 [17:53:44<17:31:51, 11.45s/step, acc=0.4957, loss=2.0893, lr=5.60e-05]
Training:  50%|█████     | 5569/11082 [17:53:56<17:47:04, 11.61s/step, acc=0.4957, loss=2.0893, lr=5.60e-05]
Training:  50%|█████     | 5570/11082 [17:54:08<17:42:04, 11.56s/step, acc=0.4957, loss=2.0893, lr=5.60e-05]
Training:  50%|█████     | 5571/11082 [17:54:19<17:32:08, 11.46s/step, acc=0.4957, loss=2.0893, lr=5.60e-05]
Training:  50%|█████     | 5572/11082 [17:54:31<17:38:51, 11.53s/step, acc=0.4957, loss=2.0893, lr=5.60e-05]
Training:  50%|█████     | 5573/11082 [17:54:43<17:47:39, 11.63s/step, acc=0.4957, loss=2.0893, lr=5.60e-05]
Training:  50%|█████     | 5574/11082 [17:54:54<17:38:55, 11.54s/step, acc=0.4957, loss=2.0893, lr=5.60e-05]
Training:  50%|█████     | 5575/11082 [17:55:06<17:41:35, 11.57s/step, acc=0.4957, loss=2.0893, lr=5.60e-05]
Training:  50%|█████     | 5576/11082 [17:55:17<17:40:38, 11.56s/step, acc=0.4957, loss=2.0893, lr=5.60e-05]
Training:  50%|█████     | 5577/11082 [17:55:29<17:42:01, 11.58s/step, acc=0.4957, loss=2.0893, lr=5.60e-05]
Training:  50%|█████     | 5578/11082 [17:55:41<17:45:40, 11.62s/step, acc=0.4957, loss=2.0893, lr=5.60e-05]
Training:  50%|█████     | 5579/11082 [17:55:52<17:36:21, 11.52s/step, acc=0.4957, loss=2.0893, lr=5.60e-05]
Training:  50%|█████     | 5580/11082 [17:56:03<17:38:31, 11.54s/step, acc=0.4957, loss=2.0893, lr=5.60e-05]
Training:  50%|█████     | 5581/11082 [17:56:15<17:41:24, 11.58s/step, acc=0.4957, loss=2.0893, lr=5.60e-05]
Training:  50%|█████     | 5582/11082 [17:56:27<17:39:20, 11.56s/step, acc=0.4957, loss=2.0893, lr=5.60e-05]
Training:  50%|█████     | 5583/11082 [17:56:39<17:51:50, 11.69s/step, acc=0.4957, loss=2.0893, lr=5.60e-05]
Training:  50%|█████     | 5584/11082 [17:56:50<17:50:24, 11.68s/step, acc=0.4957, loss=2.0893, lr=5.60e-05]
Training:  50%|█████     | 5585/11082 [17:57:02<17:47:04, 11.65s/step, acc=0.4957, loss=2.0893, lr=5.60e-05]
Training:  50%|█████     | 5586/11082 [17:57:13<17:37:39, 11.55s/step, acc=0.4957, loss=2.0893, lr=5.60e-05]
Training:  50%|█████     | 5587/11082 [17:57:25<17:45:27, 11.63s/step, acc=0.4957, loss=2.0893, lr=5.60e-05]
Training:  50%|█████     | 5588/11082 [17:57:36<17:41:26, 11.59s/step, acc=0.4957, loss=2.0893, lr=5.60e-05]
Training:  50%|█████     | 5589/11082 [17:57:48<17:34:59, 11.52s/step, acc=0.4957, loss=2.0893, lr=5.60e-05]
Training:  50%|█████     | 5590/11082 [17:58:00<17:40:58, 11.59s/step, acc=0.4957, loss=2.0893, lr=5.60e-05]
Training:  50%|█████     | 5591/11082 [17:58:11<17:32:57, 11.51s/step, acc=0.4957, loss=2.0893, lr=5.60e-05]
Training:  50%|█████     | 5592/11082 [17:58:22<17:36:01, 11.54s/step, acc=0.4957, loss=2.0893, lr=5.60e-05]
Training:  50%|█████     | 5593/11082 [17:58:34<17:38:33, 11.57s/step, acc=0.4957, loss=2.0893, lr=5.60e-05]
Training:  50%|█████     | 5594/11082 [17:58:45<17:31:24, 11.50s/step, acc=0.4957, loss=2.0893, lr=5.60e-05]
Training:  50%|█████     | 5595/11082 [17:58:57<17:40:52, 11.60s/step, acc=0.4957, loss=2.0893, lr=5.60e-05]
Training:  50%|█████     | 5596/11082 [17:59:09<17:31:08, 11.50s/step, acc=0.4957, loss=2.0893, lr=5.60e-05]
Training:  51%|█████     | 5597/11082 [17:59:20<17:33:45, 11.53s/step, acc=0.4957, loss=2.0893, lr=5.60e-05]
Training:  51%|█████     | 5598/11082 [17:59:31<17:22:05, 11.40s/step, acc=0.4957, loss=2.0893, lr=5.60e-05]
Training:  51%|█████     | 5599/11082 [17:59:43<17:37:22, 11.57s/step, acc=0.4957, loss=2.0893, lr=5.60e-05]
Training:  51%|█████     | 5600/11082 [17:59:55<17:31:08, 11.50s/step, acc=0.4957, loss=2.0893, lr=5.60e-05]
Training:  51%|█████     | 5600/11082 [18:00:02<17:31:08, 11.50s/step, acc=0.4972, loss=2.0715, lr=5.53e-05]2026-06-19 08:53:30 | INFO | __main__ | epoch=1 step=5600 train_loss=2.071546 train_acc=0.497198 lr=5.531e-05 grad_norm=0.8281 acc_d0=0.7579 acc_d1=0.6113 acc_d2=0.4900 acc_d3=0.3903 acc_d4=0.3122 acc_d5=0.2514 acc_d6=0.2041 acc_d7=0.1676

Training:  51%|█████     | 5601/11082 [18:00:06<17:35:45, 11.56s/step, acc=0.4972, loss=2.0715, lr=5.53e-05]
Training:  51%|█████     | 5602/11082 [18:00:18<17:32:40, 11.53s/step, acc=0.4972, loss=2.0715, lr=5.53e-05]
Training:  51%|█████     | 5603/11082 [18:00:29<17:32:17, 11.52s/step, acc=0.4972, loss=2.0715, lr=5.53e-05]
Training:  51%|█████     | 5604/11082 [18:00:40<17:25:14, 11.45s/step, acc=0.4972, loss=2.0715, lr=5.53e-05]
Training:  51%|█████     | 5605/11082 [18:00:52<17:38:03, 11.59s/step, acc=0.4972, loss=2.0715, lr=5.53e-05]
Training:  51%|█████     | 5606/11082 [18:01:04<17:29:14, 11.50s/step, acc=0.4972, loss=2.0715, lr=5.53e-05]
Training:  51%|█████     | 5607/11082 [18:01:15<17:24:48, 11.45s/step, acc=0.4972, loss=2.0715, lr=5.53e-05]
Training:  51%|█████     | 5608/11082 [18:01:27<17:44:02, 11.66s/step, acc=0.4972, loss=2.0715, lr=5.53e-05]
Training:  51%|█████     | 5609/11082 [18:01:38<17:29:52, 11.51s/step, acc=0.4972, loss=2.0715, lr=5.53e-05]
Training:  51%|█████     | 5610/11082 [18:01:50<17:24:19, 11.45s/step, acc=0.4972, loss=2.0715, lr=5.53e-05]
Training:  51%|█████     | 5611/11082 [18:02:02<17:39:38, 11.62s/step, acc=0.4972, loss=2.0715, lr=5.53e-05]
Training:  51%|█████     | 5612/11082 [18:02:13<17:26:35, 11.48s/step, acc=0.4972, loss=2.0715, lr=5.53e-05]
Training:  51%|█████     | 5613/11082 [18:02:25<17:33:11, 11.55s/step, acc=0.4972, loss=2.0715, lr=5.53e-05]
Training:  51%|█████     | 5614/11082 [18:02:36<17:32:07, 11.54s/step, acc=0.4972, loss=2.0715, lr=5.53e-05]
Training:  51%|█████     | 5615/11082 [18:02:48<17:36:25, 11.59s/step, acc=0.4972, loss=2.0715, lr=5.53e-05]
Training:  51%|█████     | 5616/11082 [18:02:59<17:32:39, 11.56s/step, acc=0.4972, loss=2.0715, lr=5.53e-05]
Training:  51%|█████     | 5617/11082 [18:03:11<17:35:43, 11.59s/step, acc=0.4972, loss=2.0715, lr=5.53e-05]
Training:  51%|█████     | 5618/11082 [18:03:22<17:28:20, 11.51s/step, acc=0.4972, loss=2.0715, lr=5.53e-05]
Training:  51%|█████     | 5619/11082 [18:03:34<17:38:09, 11.62s/step, acc=0.4972, loss=2.0715, lr=5.53e-05]
Training:  51%|█████     | 5620/11082 [18:03:46<17:37:25, 11.62s/step, acc=0.4972, loss=2.0715, lr=5.53e-05]
Training:  51%|█████     | 5621/11082 [18:03:57<17:18:18, 11.41s/step, acc=0.4972, loss=2.0715, lr=5.53e-05]
Training:  51%|█████     | 5622/11082 [18:04:08<17:25:53, 11.49s/step, acc=0.4972, loss=2.0715, lr=5.53e-05]
Training:  51%|█████     | 5623/11082 [18:04:20<17:31:36, 11.56s/step, acc=0.4972, loss=2.0715, lr=5.53e-05]
Training:  51%|█████     | 5624/11082 [18:04:31<17:17:01, 11.40s/step, acc=0.4972, loss=2.0715, lr=5.53e-05]
Training:  51%|█████     | 5625/11082 [18:04:43<17:40:29, 11.66s/step, acc=0.4972, loss=2.0715, lr=5.53e-05]
Training:  51%|█████     | 5626/11082 [18:04:55<17:29:01, 11.54s/step, acc=0.4972, loss=2.0715, lr=5.53e-05]
Training:  51%|█████     | 5627/11082 [18:05:06<17:35:44, 11.61s/step, acc=0.4972, loss=2.0715, lr=5.53e-05]
Training:  51%|█████     | 5628/11082 [18:05:18<17:39:12, 11.65s/step, acc=0.4972, loss=2.0715, lr=5.53e-05]
Training:  51%|█████     | 5629/11082 [18:05:30<17:33:51, 11.60s/step, acc=0.4972, loss=2.0715, lr=5.53e-05]
Training:  51%|█████     | 5630/11082 [18:05:42<17:43:25, 11.70s/step, acc=0.4972, loss=2.0715, lr=5.53e-05]
Training:  51%|█████     | 5631/11082 [18:05:54<17:50:36, 11.78s/step, acc=0.4972, loss=2.0715, lr=5.53e-05]
Training:  51%|█████     | 5632/11082 [18:06:05<17:35:18, 11.62s/step, acc=0.4972, loss=2.0715, lr=5.53e-05]
Training:  51%|█████     | 5633/11082 [18:06:17<17:43:23, 11.71s/step, acc=0.4972, loss=2.0715, lr=5.53e-05]
Training:  51%|█████     | 5634/11082 [18:06:28<17:36:55, 11.64s/step, acc=0.4972, loss=2.0715, lr=5.53e-05]
Training:  51%|█████     | 5635/11082 [18:06:40<17:36:40, 11.64s/step, acc=0.4972, loss=2.0715, lr=5.53e-05]
Training:  51%|█████     | 5636/11082 [18:06:52<17:44:41, 11.73s/step, acc=0.4972, loss=2.0715, lr=5.53e-05]
Training:  51%|█████     | 5637/11082 [18:07:03<17:36:07, 11.64s/step, acc=0.4972, loss=2.0715, lr=5.53e-05]
Training:  51%|█████     | 5638/11082 [18:07:15<17:38:25, 11.67s/step, acc=0.4972, loss=2.0715, lr=5.53e-05]
Training:  51%|█████     | 5639/11082 [18:07:27<17:38:43, 11.67s/step, acc=0.4972, loss=2.0715, lr=5.53e-05]
Training:  51%|█████     | 5640/11082 [18:07:38<17:40:52, 11.70s/step, acc=0.4972, loss=2.0715, lr=5.53e-05]
Training:  51%|█████     | 5641/11082 [18:07:50<17:28:49, 11.57s/step, acc=0.4972, loss=2.0715, lr=5.53e-05]
Training:  51%|█████     | 5642/11082 [18:08:01<17:35:20, 11.64s/step, acc=0.4972, loss=2.0715, lr=5.53e-05]
Training:  51%|█████     | 5643/11082 [18:08:13<17:27:49, 11.56s/step, acc=0.4972, loss=2.0715, lr=5.53e-05]
Training:  51%|█████     | 5644/11082 [18:08:24<17:16:16, 11.43s/step, acc=0.4972, loss=2.0715, lr=5.53e-05]
Training:  51%|█████     | 5645/11082 [18:08:36<17:33:03, 11.62s/step, acc=0.4972, loss=2.0715, lr=5.53e-05]
Training:  51%|█████     | 5646/11082 [18:08:48<17:32:19, 11.62s/step, acc=0.4972, loss=2.0715, lr=5.53e-05]
Training:  51%|█████     | 5647/11082 [18:08:59<17:36:11, 11.66s/step, acc=0.4972, loss=2.0715, lr=5.53e-05]
Training:  51%|█████     | 5648/11082 [18:09:10<17:21:41, 11.50s/step, acc=0.4972, loss=2.0715, lr=5.53e-05]
Training:  51%|█████     | 5649/11082 [18:09:22<17:24:25, 11.53s/step, acc=0.4972, loss=2.0715, lr=5.53e-05]
Training:  51%|█████     | 5650/11082 [18:09:33<17:18:06, 11.47s/step, acc=0.4972, loss=2.0715, lr=5.53e-05]
Training:  51%|█████     | 5650/11082 [18:09:41<17:18:06, 11.47s/step, acc=0.4958, loss=2.0810, lr=5.47e-05]2026-06-19 09:03:09 | INFO | __main__ | epoch=1 step=5650 train_loss=2.080997 train_acc=0.495814 lr=5.466e-05 grad_norm=0.8086 acc_d0=0.7579 acc_d1=0.6132 acc_d2=0.4918 acc_d3=0.3939 acc_d4=0.3137 acc_d5=0.2530 acc_d6=0.2054 acc_d7=0.1668

Training:  51%|█████     | 5651/11082 [18:09:45<17:29:38, 11.60s/step, acc=0.4958, loss=2.0810, lr=5.47e-05]
Training:  51%|█████     | 5652/11082 [18:09:56<17:14:52, 11.44s/step, acc=0.4958, loss=2.0810, lr=5.47e-05]
Training:  51%|█████     | 5653/11082 [18:10:08<17:07:56, 11.36s/step, acc=0.4958, loss=2.0810, lr=5.47e-05]
Training:  51%|█████     | 5654/11082 [18:10:19<17:24:05, 11.54s/step, acc=0.4958, loss=2.0810, lr=5.47e-05]
Training:  51%|█████     | 5655/11082 [18:10:31<17:28:53, 11.60s/step, acc=0.4958, loss=2.0810, lr=5.47e-05]
Training:  51%|█████     | 5656/11082 [18:10:43<17:29:59, 11.61s/step, acc=0.4958, loss=2.0810, lr=5.47e-05]
Training:  51%|█████     | 5657/11082 [18:10:55<17:30:32, 11.62s/step, acc=0.4958, loss=2.0810, lr=5.47e-05]
Training:  51%|█████     | 5658/11082 [18:11:06<17:24:08, 11.55s/step, acc=0.4958, loss=2.0810, lr=5.47e-05]
Training:  51%|█████     | 5659/11082 [18:11:18<17:32:12, 11.64s/step, acc=0.4958, loss=2.0810, lr=5.47e-05]
Training:  51%|█████     | 5660/11082 [18:11:29<17:10:13, 11.40s/step, acc=0.4958, loss=2.0810, lr=5.47e-05]
Training:  51%|█████     | 5661/11082 [18:11:40<17:23:34, 11.55s/step, acc=0.4958, loss=2.0810, lr=5.47e-05]
Training:  51%|█████     | 5662/11082 [18:11:52<17:34:36, 11.67s/step, acc=0.4958, loss=2.0810, lr=5.47e-05]
Training:  51%|█████     | 5663/11082 [18:12:04<17:37:12, 11.71s/step, acc=0.4958, loss=2.0810, lr=5.47e-05]
Training:  51%|█████     | 5664/11082 [18:12:16<17:33:38, 11.67s/step, acc=0.4958, loss=2.0810, lr=5.47e-05]
Training:  51%|█████     | 5665/11082 [18:12:27<17:30:31, 11.64s/step, acc=0.4958, loss=2.0810, lr=5.47e-05]
Training:  51%|█████     | 5666/11082 [18:12:39<17:21:44, 11.54s/step, acc=0.4958, loss=2.0810, lr=5.47e-05]
Training:  51%|█████     | 5667/11082 [18:12:51<17:31:53, 11.66s/step, acc=0.4958, loss=2.0810, lr=5.47e-05]
Training:  51%|█████     | 5668/11082 [18:13:02<17:32:18, 11.66s/step, acc=0.4958, loss=2.0810, lr=5.47e-05]
Training:  51%|█████     | 5669/11082 [18:13:14<17:32:47, 11.67s/step, acc=0.4958, loss=2.0810, lr=5.47e-05]
Training:  51%|█████     | 5670/11082 [18:13:25<17:19:41, 11.53s/step, acc=0.4958, loss=2.0810, lr=5.47e-05]
Training:  51%|█████     | 5671/11082 [18:13:37<17:26:06, 11.60s/step, acc=0.4958, loss=2.0810, lr=5.47e-05]
Training:  51%|█████     | 5672/11082 [18:13:48<17:17:29, 11.51s/step, acc=0.4958, loss=2.0810, lr=5.47e-05]
Training:  51%|█████     | 5673/11082 [18:14:00<17:18:31, 11.52s/step, acc=0.4958, loss=2.0810, lr=5.47e-05]
Training:  51%|█████     | 5674/11082 [18:14:11<17:14:36, 11.48s/step, acc=0.4958, loss=2.0810, lr=5.47e-05]
Training:  51%|█████     | 5675/11082 [18:14:23<17:23:21, 11.58s/step, acc=0.4958, loss=2.0810, lr=5.47e-05]
Training:  51%|█████     | 5676/11082 [18:14:35<17:27:22, 11.62s/step, acc=0.4958, loss=2.0810, lr=5.47e-05]
Training:  51%|█████     | 5677/11082 [18:14:46<17:21:04, 11.56s/step, acc=0.4958, loss=2.0810, lr=5.47e-05]
Training:  51%|█████     | 5678/11082 [18:14:58<17:20:47, 11.56s/step, acc=0.4958, loss=2.0810, lr=5.47e-05]
Training:  51%|█████     | 5679/11082 [18:15:09<17:25:29, 11.61s/step, acc=0.4958, loss=2.0810, lr=5.47e-05]
Training:  51%|█████▏    | 5680/11082 [18:15:21<17:17:13, 11.52s/step, acc=0.4958, loss=2.0810, lr=5.47e-05]
Training:  51%|█████▏    | 5681/11082 [18:15:32<17:05:45, 11.40s/step, acc=0.4958, loss=2.0810, lr=5.47e-05]
Training:  51%|█████▏    | 5682/11082 [18:15:44<17:17:25, 11.53s/step, acc=0.4958, loss=2.0810, lr=5.47e-05]
Training:  51%|█████▏    | 5683/11082 [18:15:56<17:26:16, 11.63s/step, acc=0.4958, loss=2.0810, lr=5.47e-05]
Training:  51%|█████▏    | 5684/11082 [18:16:07<17:20:29, 11.57s/step, acc=0.4958, loss=2.0810, lr=5.47e-05]
Training:  51%|█████▏    | 5685/11082 [18:16:18<17:14:25, 11.50s/step, acc=0.4958, loss=2.0810, lr=5.47e-05]
Training:  51%|█████▏    | 5686/11082 [18:16:30<17:20:03, 11.56s/step, acc=0.4958, loss=2.0810, lr=5.47e-05]
Training:  51%|█████▏    | 5687/11082 [18:16:42<17:27:01, 11.64s/step, acc=0.4958, loss=2.0810, lr=5.47e-05]
Training:  51%|█████▏    | 5688/11082 [18:16:53<17:08:17, 11.44s/step, acc=0.4958, loss=2.0810, lr=5.47e-05]
Training:  51%|█████▏    | 5689/11082 [18:17:05<17:18:01, 11.55s/step, acc=0.4958, loss=2.0810, lr=5.47e-05]
Training:  51%|█████▏    | 5690/11082 [18:17:16<17:22:35, 11.60s/step, acc=0.4958, loss=2.0810, lr=5.47e-05]
Training:  51%|█████▏    | 5691/11082 [18:17:27<17:09:06, 11.45s/step, acc=0.4958, loss=2.0810, lr=5.47e-05]
Training:  51%|█████▏    | 5692/11082 [18:17:39<17:00:08, 11.36s/step, acc=0.4958, loss=2.0810, lr=5.47e-05]
Training:  51%|█████▏    | 5693/11082 [18:17:51<17:16:30, 11.54s/step, acc=0.4958, loss=2.0810, lr=5.47e-05]
Training:  51%|█████▏    | 5694/11082 [18:18:02<17:16:46, 11.55s/step, acc=0.4958, loss=2.0810, lr=5.47e-05]
Training:  51%|█████▏    | 5695/11082 [18:18:14<17:29:33, 11.69s/step, acc=0.4958, loss=2.0810, lr=5.47e-05]
Training:  51%|█████▏    | 5696/11082 [18:18:25<17:19:46, 11.58s/step, acc=0.4958, loss=2.0810, lr=5.47e-05]
Training:  51%|█████▏    | 5697/11082 [18:18:37<17:20:02, 11.59s/step, acc=0.4958, loss=2.0810, lr=5.47e-05]
Training:  51%|█████▏    | 5698/11082 [18:18:49<17:21:47, 11.61s/step, acc=0.4958, loss=2.0810, lr=5.47e-05]
Training:  51%|█████▏    | 5699/11082 [18:19:00<17:18:43, 11.58s/step, acc=0.4958, loss=2.0810, lr=5.47e-05]
Training:  51%|█████▏    | 5700/11082 [18:19:12<17:13:18, 11.52s/step, acc=0.4958, loss=2.0810, lr=5.47e-05]
Training:  51%|█████▏    | 5700/11082 [18:19:19<17:13:18, 11.52s/step, acc=0.4990, loss=2.0626, lr=5.40e-05]2026-06-19 09:12:47 | INFO | __main__ | epoch=1 step=5700 train_loss=2.062589 train_acc=0.499043 lr=5.402e-05 grad_norm=0.9531 acc_d0=0.7683 acc_d1=0.6265 acc_d2=0.5070 acc_d3=0.4098 acc_d4=0.3313 acc_d5=0.2668 acc_d6=0.2151 acc_d7=0.1791

Training:  51%|█████▏    | 5701/11082 [18:19:23<17:11:19, 11.50s/step, acc=0.4990, loss=2.0626, lr=5.40e-05]
Training:  51%|█████▏    | 5702/11082 [18:19:35<17:15:23, 11.55s/step, acc=0.4990, loss=2.0626, lr=5.40e-05]
Training:  51%|█████▏    | 5703/11082 [18:19:46<17:19:10, 11.59s/step, acc=0.4990, loss=2.0626, lr=5.40e-05]
Training:  51%|█████▏    | 5704/11082 [18:19:58<17:20:46, 11.61s/step, acc=0.4990, loss=2.0626, lr=5.40e-05]
Training:  51%|█████▏    | 5705/11082 [18:20:09<17:13:05, 11.53s/step, acc=0.4990, loss=2.0626, lr=5.40e-05]
Training:  51%|█████▏    | 5706/11082 [18:20:20<17:00:21, 11.39s/step, acc=0.4990, loss=2.0626, lr=5.40e-05]
Training:  51%|█████▏    | 5707/11082 [18:20:32<17:05:51, 11.45s/step, acc=0.4990, loss=2.0626, lr=5.40e-05]
Training:  52%|█████▏    | 5708/11082 [18:20:44<17:11:05, 11.51s/step, acc=0.4990, loss=2.0626, lr=5.40e-05]
Training:  52%|█████▏    | 5709/11082 [18:20:55<17:11:56, 11.52s/step, acc=0.4990, loss=2.0626, lr=5.40e-05]
Training:  52%|█████▏    | 5710/11082 [18:21:07<17:26:09, 11.68s/step, acc=0.4990, loss=2.0626, lr=5.40e-05]
Training:  52%|█████▏    | 5711/11082 [18:21:19<17:24:15, 11.67s/step, acc=0.4990, loss=2.0626, lr=5.40e-05]
Training:  52%|█████▏    | 5712/11082 [18:21:31<17:29:05, 11.72s/step, acc=0.4990, loss=2.0626, lr=5.40e-05]
Training:  52%|█████▏    | 5713/11082 [18:21:42<17:28:54, 11.72s/step, acc=0.4990, loss=2.0626, lr=5.40e-05]
Training:  52%|█████▏    | 5714/11082 [18:21:54<17:15:46, 11.58s/step, acc=0.4990, loss=2.0626, lr=5.40e-05]
Training:  52%|█████▏    | 5715/11082 [18:22:05<17:17:15, 11.60s/step, acc=0.4990, loss=2.0626, lr=5.40e-05]
Training:  52%|█████▏    | 5716/11082 [18:22:17<17:22:24, 11.66s/step, acc=0.4990, loss=2.0626, lr=5.40e-05]
Training:  52%|█████▏    | 5717/11082 [18:22:29<17:15:33, 11.58s/step, acc=0.4990, loss=2.0626, lr=5.40e-05]
Training:  52%|█████▏    | 5718/11082 [18:22:41<17:29:26, 11.74s/step, acc=0.4990, loss=2.0626, lr=5.40e-05]
Training:  52%|█████▏    | 5719/11082 [18:22:52<17:13:39, 11.56s/step, acc=0.4990, loss=2.0626, lr=5.40e-05]
Training:  52%|█████▏    | 5720/11082 [18:23:03<17:05:38, 11.48s/step, acc=0.4990, loss=2.0626, lr=5.40e-05]
Training:  52%|█████▏    | 5721/11082 [18:23:15<17:13:57, 11.57s/step, acc=0.4990, loss=2.0626, lr=5.40e-05]
Training:  52%|█████▏    | 5722/11082 [18:23:27<17:15:25, 11.59s/step, acc=0.4990, loss=2.0626, lr=5.40e-05]
Training:  52%|█████▏    | 5723/11082 [18:23:38<17:09:07, 11.52s/step, acc=0.4990, loss=2.0626, lr=5.40e-05]
Training:  52%|█████▏    | 5724/11082 [18:23:50<17:17:16, 11.62s/step, acc=0.4990, loss=2.0626, lr=5.40e-05]
Training:  52%|█████▏    | 5725/11082 [18:24:01<17:13:51, 11.58s/step, acc=0.4990, loss=2.0626, lr=5.40e-05]
Training:  52%|█████▏    | 5726/11082 [18:24:13<17:09:14, 11.53s/step, acc=0.4990, loss=2.0626, lr=5.40e-05]
Training:  52%|█████▏    | 5727/11082 [18:24:24<16:51:43, 11.34s/step, acc=0.4990, loss=2.0626, lr=5.40e-05]
Training:  52%|█████▏    | 5728/11082 [18:24:36<17:12:58, 11.58s/step, acc=0.4990, loss=2.0626, lr=5.40e-05]
Training:  52%|█████▏    | 5729/11082 [18:24:47<17:16:49, 11.62s/step, acc=0.4990, loss=2.0626, lr=5.40e-05]
Training:  52%|█████▏    | 5730/11082 [18:24:58<17:00:38, 11.44s/step, acc=0.4990, loss=2.0626, lr=5.40e-05]
Training:  52%|█████▏    | 5731/11082 [18:25:10<17:07:35, 11.52s/step, acc=0.4990, loss=2.0626, lr=5.40e-05]
Training:  52%|█████▏    | 5732/11082 [18:25:22<17:15:44, 11.62s/step, acc=0.4990, loss=2.0626, lr=5.40e-05]
Training:  52%|█████▏    | 5733/11082 [18:25:34<17:18:58, 11.65s/step, acc=0.4990, loss=2.0626, lr=5.40e-05]
Training:  52%|█████▏    | 5734/11082 [18:25:45<17:01:58, 11.47s/step, acc=0.4990, loss=2.0626, lr=5.40e-05]
Training:  52%|█████▏    | 5735/11082 [18:25:56<17:09:15, 11.55s/step, acc=0.4990, loss=2.0626, lr=5.40e-05]
Training:  52%|█████▏    | 5736/11082 [18:26:08<17:14:11, 11.61s/step, acc=0.4990, loss=2.0626, lr=5.40e-05]
Training:  52%|█████▏    | 5737/11082 [18:26:19<17:02:25, 11.48s/step, acc=0.4990, loss=2.0626, lr=5.40e-05]
Training:  52%|█████▏    | 5738/11082 [18:26:31<17:17:34, 11.65s/step, acc=0.4990, loss=2.0626, lr=5.40e-05]
Training:  52%|█████▏    | 5739/11082 [18:26:43<17:06:47, 11.53s/step, acc=0.4990, loss=2.0626, lr=5.40e-05]
Training:  52%|█████▏    | 5740/11082 [18:26:54<17:10:46, 11.58s/step, acc=0.4990, loss=2.0626, lr=5.40e-05]
Training:  52%|█████▏    | 5741/11082 [18:27:06<17:15:07, 11.63s/step, acc=0.4990, loss=2.0626, lr=5.40e-05]
Training:  52%|█████▏    | 5742/11082 [18:27:18<17:15:07, 11.63s/step, acc=0.4990, loss=2.0626, lr=5.40e-05]
Training:  52%|█████▏    | 5743/11082 [18:27:29<17:01:34, 11.48s/step, acc=0.4990, loss=2.0626, lr=5.40e-05]
Training:  52%|█████▏    | 5744/11082 [18:27:41<17:14:12, 11.62s/step, acc=0.4990, loss=2.0626, lr=5.40e-05]
Training:  52%|█████▏    | 5745/11082 [18:27:52<17:10:02, 11.58s/step, acc=0.4990, loss=2.0626, lr=5.40e-05]
Training:  52%|█████▏    | 5746/11082 [18:28:04<17:07:37, 11.56s/step, acc=0.4990, loss=2.0626, lr=5.40e-05]
Training:  52%|█████▏    | 5747/11082 [18:28:16<17:11:50, 11.60s/step, acc=0.4990, loss=2.0626, lr=5.40e-05]
Training:  52%|█████▏    | 5748/11082 [18:28:27<17:19:47, 11.70s/step, acc=0.4990, loss=2.0626, lr=5.40e-05]
Training:  52%|█████▏    | 5749/11082 [18:28:39<17:09:18, 11.58s/step, acc=0.4990, loss=2.0626, lr=5.40e-05]
Training:  52%|█████▏    | 5750/11082 [18:28:50<17:05:26, 11.54s/step, acc=0.4990, loss=2.0626, lr=5.40e-05]
Training:  52%|█████▏    | 5750/11082 [18:28:57<17:05:26, 11.54s/step, acc=0.4984, loss=2.0638, lr=5.34e-05]2026-06-19 09:22:25 | INFO | __main__ | epoch=1 step=5750 train_loss=2.063845 train_acc=0.498370 lr=5.337e-05 grad_norm=0.7617 acc_d0=0.7633 acc_d1=0.6185 acc_d2=0.4966 acc_d3=0.3954 acc_d4=0.3160 acc_d5=0.2547 acc_d6=0.2047 acc_d7=0.1689

Training:  52%|█████▏    | 5751/11082 [18:29:01<16:54:07, 11.41s/step, acc=0.4984, loss=2.0638, lr=5.34e-05]
Training:  52%|█████▏    | 5752/11082 [18:29:13<17:09:38, 11.59s/step, acc=0.4984, loss=2.0638, lr=5.34e-05]
Training:  52%|█████▏    | 5753/11082 [18:29:25<17:10:10, 11.60s/step, acc=0.4984, loss=2.0638, lr=5.34e-05]
Training:  52%|█████▏    | 5754/11082 [18:29:37<17:10:36, 11.61s/step, acc=0.4984, loss=2.0638, lr=5.34e-05]
Training:  52%|█████▏    | 5755/11082 [18:29:48<17:01:51, 11.51s/step, acc=0.4984, loss=2.0638, lr=5.34e-05]
Training:  52%|█████▏    | 5756/11082 [18:29:59<17:03:04, 11.53s/step, acc=0.4984, loss=2.0638, lr=5.34e-05]
Training:  52%|█████▏    | 5757/11082 [18:30:11<17:01:14, 11.51s/step, acc=0.4984, loss=2.0638, lr=5.34e-05]
Training:  52%|█████▏    | 5758/11082 [18:30:22<16:59:17, 11.49s/step, acc=0.4984, loss=2.0638, lr=5.34e-05]
Training:  52%|█████▏    | 5759/11082 [18:30:34<17:04:58, 11.55s/step, acc=0.4984, loss=2.0638, lr=5.34e-05]
Training:  52%|█████▏    | 5760/11082 [18:30:46<17:07:29, 11.58s/step, acc=0.4984, loss=2.0638, lr=5.34e-05]
Training:  52%|█████▏    | 5761/11082 [18:30:57<17:00:37, 11.51s/step, acc=0.4984, loss=2.0638, lr=5.34e-05]
Training:  52%|█████▏    | 5762/11082 [18:31:09<17:04:51, 11.56s/step, acc=0.4984, loss=2.0638, lr=5.34e-05]
Training:  52%|█████▏    | 5763/11082 [18:31:20<16:59:20, 11.50s/step, acc=0.4984, loss=2.0638, lr=5.34e-05]
Training:  52%|█████▏    | 5764/11082 [18:31:32<17:07:00, 11.59s/step, acc=0.4984, loss=2.0638, lr=5.34e-05]
Training:  52%|█████▏    | 5765/11082 [18:31:43<17:03:06, 11.55s/step, acc=0.4984, loss=2.0638, lr=5.34e-05]
Training:  52%|█████▏    | 5766/11082 [18:31:55<17:08:29, 11.61s/step, acc=0.4984, loss=2.0638, lr=5.34e-05]
Training:  52%|█████▏    | 5767/11082 [18:32:06<17:01:55, 11.54s/step, acc=0.4984, loss=2.0638, lr=5.34e-05]
Training:  52%|█████▏    | 5768/11082 [18:32:18<16:59:07, 11.51s/step, acc=0.4984, loss=2.0638, lr=5.34e-05]
Training:  52%|█████▏    | 5769/11082 [18:32:29<16:59:41, 11.52s/step, acc=0.4984, loss=2.0638, lr=5.34e-05]
Training:  52%|█████▏    | 5770/11082 [18:32:41<16:59:16, 11.51s/step, acc=0.4984, loss=2.0638, lr=5.34e-05]
Training:  52%|█████▏    | 5771/11082 [18:32:53<17:07:13, 11.60s/step, acc=0.4984, loss=2.0638, lr=5.34e-05]
Training:  52%|█████▏    | 5772/11082 [18:33:05<17:12:38, 11.67s/step, acc=0.4984, loss=2.0638, lr=5.34e-05]
Training:  52%|█████▏    | 5773/11082 [18:33:16<17:08:39, 11.63s/step, acc=0.4984, loss=2.0638, lr=5.34e-05]
Training:  52%|█████▏    | 5774/11082 [18:33:28<17:04:28, 11.58s/step, acc=0.4984, loss=2.0638, lr=5.34e-05]
Training:  52%|█████▏    | 5775/11082 [18:33:39<17:04:30, 11.58s/step, acc=0.4984, loss=2.0638, lr=5.34e-05]
Training:  52%|█████▏    | 5776/11082 [18:33:50<16:50:32, 11.43s/step, acc=0.4984, loss=2.0638, lr=5.34e-05]
Training:  52%|█████▏    | 5777/11082 [18:34:02<16:48:40, 11.41s/step, acc=0.4984, loss=2.0638, lr=5.34e-05]
Training:  52%|█████▏    | 5778/11082 [18:34:13<17:00:49, 11.55s/step, acc=0.4984, loss=2.0638, lr=5.34e-05]
Training:  52%|█████▏    | 5779/11082 [18:34:25<17:05:29, 11.60s/step, acc=0.4984, loss=2.0638, lr=5.34e-05]
Training:  52%|█████▏    | 5780/11082 [18:34:36<16:51:25, 11.45s/step, acc=0.4984, loss=2.0638, lr=5.34e-05]
Training:  52%|█████▏    | 5781/11082 [18:34:48<17:05:35, 11.61s/step, acc=0.4984, loss=2.0638, lr=5.34e-05]
Training:  52%|█████▏    | 5782/11082 [18:35:00<17:00:43, 11.56s/step, acc=0.4984, loss=2.0638, lr=5.34e-05]
Training:  52%|█████▏    | 5783/11082 [18:35:11<17:02:06, 11.57s/step, acc=0.4984, loss=2.0638, lr=5.34e-05]
Training:  52%|█████▏    | 5784/11082 [18:35:23<16:59:56, 11.55s/step, acc=0.4984, loss=2.0638, lr=5.34e-05]
Training:  52%|█████▏    | 5785/11082 [18:35:34<16:52:51, 11.47s/step, acc=0.4984, loss=2.0638, lr=5.34e-05]
Training:  52%|█████▏    | 5786/11082 [18:35:46<17:01:23, 11.57s/step, acc=0.4984, loss=2.0638, lr=5.34e-05]
Training:  52%|█████▏    | 5787/11082 [18:35:58<17:03:06, 11.59s/step, acc=0.4984, loss=2.0638, lr=5.34e-05]
Training:  52%|█████▏    | 5788/11082 [18:36:09<17:02:40, 11.59s/step, acc=0.4984, loss=2.0638, lr=5.34e-05]
Training:  52%|█████▏    | 5789/11082 [18:36:20<16:47:10, 11.42s/step, acc=0.4984, loss=2.0638, lr=5.34e-05]
Training:  52%|█████▏    | 5790/11082 [18:36:32<16:54:56, 11.51s/step, acc=0.4984, loss=2.0638, lr=5.34e-05]
Training:  52%|█████▏    | 5791/11082 [18:36:44<17:13:27, 11.72s/step, acc=0.4984, loss=2.0638, lr=5.34e-05]
Training:  52%|█████▏    | 5792/11082 [18:36:55<17:01:38, 11.59s/step, acc=0.4984, loss=2.0638, lr=5.34e-05]
Training:  52%|█████▏    | 5793/11082 [18:37:07<16:54:59, 11.51s/step, acc=0.4984, loss=2.0638, lr=5.34e-05]
Training:  52%|█████▏    | 5794/11082 [18:37:19<17:09:46, 11.68s/step, acc=0.4984, loss=2.0638, lr=5.34e-05]
Training:  52%|█████▏    | 5795/11082 [18:37:30<17:00:28, 11.58s/step, acc=0.4984, loss=2.0638, lr=5.34e-05]
Training:  52%|█████▏    | 5796/11082 [18:37:42<17:04:31, 11.63s/step, acc=0.4984, loss=2.0638, lr=5.34e-05]
Training:  52%|█████▏    | 5797/11082 [18:37:53<16:58:01, 11.56s/step, acc=0.4984, loss=2.0638, lr=5.34e-05]
Training:  52%|█████▏    | 5798/11082 [18:38:05<16:52:39, 11.50s/step, acc=0.4984, loss=2.0638, lr=5.34e-05]
Training:  52%|█████▏    | 5799/11082 [18:38:16<16:56:02, 11.54s/step, acc=0.4984, loss=2.0638, lr=5.34e-05]
Training:  52%|█████▏    | 5800/11082 [18:38:28<17:02:21, 11.61s/step, acc=0.4984, loss=2.0638, lr=5.34e-05]
Training:  52%|█████▏    | 5800/11082 [18:38:35<17:02:21, 11.61s/step, acc=0.4988, loss=2.0597, lr=5.27e-05]2026-06-19 09:32:03 | INFO | __main__ | epoch=1 step=5800 train_loss=2.059695 train_acc=0.498773 lr=5.272e-05 grad_norm=0.8477 acc_d0=0.7637 acc_d1=0.6210 acc_d2=0.5034 acc_d3=0.4049 acc_d4=0.3284 acc_d5=0.2659 acc_d6=0.2169 acc_d7=0.1797

Training:  52%|█████▏    | 5801/11082 [18:38:40<17:05:58, 11.66s/step, acc=0.4988, loss=2.0597, lr=5.27e-05]
Training:  52%|█████▏    | 5802/11082 [18:38:51<17:03:41, 11.63s/step, acc=0.4988, loss=2.0597, lr=5.27e-05]
Training:  52%|█████▏    | 5803/11082 [18:39:03<17:01:06, 11.61s/step, acc=0.4988, loss=2.0597, lr=5.27e-05]
Training:  52%|█████▏    | 5804/11082 [18:39:14<16:52:08, 11.51s/step, acc=0.4988, loss=2.0597, lr=5.27e-05]
Training:  52%|█████▏    | 5805/11082 [18:39:26<16:54:20, 11.53s/step, acc=0.4988, loss=2.0597, lr=5.27e-05]
Training:  52%|█████▏    | 5806/11082 [18:39:37<16:54:13, 11.53s/step, acc=0.4988, loss=2.0597, lr=5.27e-05]
Training:  52%|█████▏    | 5807/11082 [18:39:49<17:03:49, 11.65s/step, acc=0.4988, loss=2.0597, lr=5.27e-05]
Training:  52%|█████▏    | 5808/11082 [18:40:01<17:06:50, 11.68s/step, acc=0.4988, loss=2.0597, lr=5.27e-05]
Training:  52%|█████▏    | 5809/11082 [18:40:12<16:50:15, 11.50s/step, acc=0.4988, loss=2.0597, lr=5.27e-05]
Training:  52%|█████▏    | 5810/11082 [18:40:24<16:58:43, 11.59s/step, acc=0.4988, loss=2.0597, lr=5.27e-05]
Training:  52%|█████▏    | 5811/11082 [18:40:36<17:22:00, 11.86s/step, acc=0.4988, loss=2.0597, lr=5.27e-05]
Training:  52%|█████▏    | 5812/11082 [18:40:48<17:10:19, 11.73s/step, acc=0.4988, loss=2.0597, lr=5.27e-05]
Training:  52%|█████▏    | 5813/11082 [18:40:59<16:58:40, 11.60s/step, acc=0.4988, loss=2.0597, lr=5.27e-05]
Training:  52%|█████▏    | 5814/11082 [18:41:11<16:56:41, 11.58s/step, acc=0.4988, loss=2.0597, lr=5.27e-05]
Training:  52%|█████▏    | 5815/11082 [18:41:22<17:04:30, 11.67s/step, acc=0.4988, loss=2.0597, lr=5.27e-05]
Training:  52%|█████▏    | 5816/11082 [18:41:34<16:56:02, 11.58s/step, acc=0.4988, loss=2.0597, lr=5.27e-05]
Training:  52%|█████▏    | 5817/11082 [18:41:45<16:55:33, 11.57s/step, acc=0.4988, loss=2.0597, lr=5.27e-05]
Training:  52%|█████▏    | 5818/11082 [18:41:57<16:54:26, 11.56s/step, acc=0.4988, loss=2.0597, lr=5.27e-05]
Training:  53%|█████▎    | 5819/11082 [18:42:09<16:58:48, 11.61s/step, acc=0.4988, loss=2.0597, lr=5.27e-05]
Training:  53%|█████▎    | 5820/11082 [18:42:20<16:54:49, 11.57s/step, acc=0.4988, loss=2.0597, lr=5.27e-05]
Training:  53%|█████▎    | 5821/11082 [18:42:31<16:43:26, 11.44s/step, acc=0.4988, loss=2.0597, lr=5.27e-05]
Training:  53%|█████▎    | 5822/11082 [18:42:43<16:49:21, 11.51s/step, acc=0.4988, loss=2.0597, lr=5.27e-05]
Training:  53%|█████▎    | 5823/11082 [18:42:55<16:59:04, 11.63s/step, acc=0.4988, loss=2.0597, lr=5.27e-05]
Training:  53%|█████▎    | 5824/11082 [18:43:06<16:49:03, 11.51s/step, acc=0.4988, loss=2.0597, lr=5.27e-05]
Training:  53%|█████▎    | 5825/11082 [18:43:18<16:54:38, 11.58s/step, acc=0.4988, loss=2.0597, lr=5.27e-05]
Training:  53%|█████▎    | 5826/11082 [18:43:30<17:02:09, 11.67s/step, acc=0.4988, loss=2.0597, lr=5.27e-05]
Training:  53%|█████▎    | 5827/11082 [18:43:41<16:48:47, 11.52s/step, acc=0.4988, loss=2.0597, lr=5.27e-05]
Training:  53%|█████▎    | 5828/11082 [18:43:53<16:55:56, 11.60s/step, acc=0.4988, loss=2.0597, lr=5.27e-05]
Training:  53%|█████▎    | 5829/11082 [18:44:04<16:46:21, 11.49s/step, acc=0.4988, loss=2.0597, lr=5.27e-05]
Training:  53%|█████▎    | 5830/11082 [18:44:16<17:03:45, 11.70s/step, acc=0.4988, loss=2.0597, lr=5.27e-05]
Training:  53%|█████▎    | 5831/11082 [18:44:27<16:52:21, 11.57s/step, acc=0.4988, loss=2.0597, lr=5.27e-05]
Training:  53%|█████▎    | 5832/11082 [18:44:40<17:07:58, 11.75s/step, acc=0.4988, loss=2.0597, lr=5.27e-05]
Training:  53%|█████▎    | 5833/11082 [18:44:51<16:53:28, 11.58s/step, acc=0.4988, loss=2.0597, lr=5.27e-05]
Training:  53%|█████▎    | 5834/11082 [18:45:02<16:48:40, 11.53s/step, acc=0.4988, loss=2.0597, lr=5.27e-05]
Training:  53%|█████▎    | 5835/11082 [18:45:14<16:57:22, 11.63s/step, acc=0.4988, loss=2.0597, lr=5.27e-05]
Training:  53%|█████▎    | 5836/11082 [18:45:25<16:49:34, 11.55s/step, acc=0.4988, loss=2.0597, lr=5.27e-05]
Training:  53%|█████▎    | 5837/11082 [18:45:37<16:46:39, 11.52s/step, acc=0.4988, loss=2.0597, lr=5.27e-05]
Training:  53%|█████▎    | 5838/11082 [18:45:48<16:49:50, 11.55s/step, acc=0.4988, loss=2.0597, lr=5.27e-05]
Training:  53%|█████▎    | 5839/11082 [18:46:00<16:41:13, 11.46s/step, acc=0.4988, loss=2.0597, lr=5.27e-05]
Training:  53%|█████▎    | 5840/11082 [18:46:12<16:53:49, 11.60s/step, acc=0.4988, loss=2.0597, lr=5.27e-05]
Training:  53%|█████▎    | 5841/11082 [18:46:23<16:47:28, 11.53s/step, acc=0.4988, loss=2.0597, lr=5.27e-05]
Training:  53%|█████▎    | 5842/11082 [18:46:35<16:49:05, 11.55s/step, acc=0.4988, loss=2.0597, lr=5.27e-05]
Training:  53%|█████▎    | 5843/11082 [18:46:46<16:52:58, 11.60s/step, acc=0.4988, loss=2.0597, lr=5.27e-05]
Training:  53%|█████▎    | 5844/11082 [18:46:58<16:49:22, 11.56s/step, acc=0.4988, loss=2.0597, lr=5.27e-05]
Training:  53%|█████▎    | 5845/11082 [18:47:10<16:55:33, 11.64s/step, acc=0.4988, loss=2.0597, lr=5.27e-05]
Training:  53%|█████▎    | 5846/11082 [18:47:21<16:39:42, 11.46s/step, acc=0.4988, loss=2.0597, lr=5.27e-05]
Training:  53%|█████▎    | 5847/11082 [18:47:32<16:42:04, 11.49s/step, acc=0.4988, loss=2.0597, lr=5.27e-05]
Training:  53%|█████▎    | 5848/11082 [18:47:44<16:49:37, 11.57s/step, acc=0.4988, loss=2.0597, lr=5.27e-05]
Training:  53%|█████▎    | 5849/11082 [18:47:55<16:49:04, 11.57s/step, acc=0.4988, loss=2.0597, lr=5.27e-05]
Training:  53%|█████▎    | 5850/11082 [18:48:07<16:48:28, 11.57s/step, acc=0.4988, loss=2.0597, lr=5.27e-05]
Training:  53%|█████▎    | 5850/11082 [18:48:14<16:48:28, 11.57s/step, acc=0.4995, loss=2.0579, lr=5.21e-05]2026-06-19 09:41:42 | INFO | __main__ | epoch=1 step=5850 train_loss=2.057929 train_acc=0.499481 lr=5.208e-05 grad_norm=0.8047 acc_d0=0.7592 acc_d1=0.6139 acc_d2=0.4920 acc_d3=0.3946 acc_d4=0.3147 acc_d5=0.2523 acc_d6=0.2050 acc_d7=0.1688

Training:  53%|█████▎    | 5851/11082 [18:48:18<16:43:51, 11.51s/step, acc=0.4995, loss=2.0579, lr=5.21e-05]
Training:  53%|█████▎    | 5852/11082 [18:48:30<16:42:59, 11.51s/step, acc=0.4995, loss=2.0579, lr=5.21e-05]
Training:  53%|█████▎    | 5853/11082 [18:48:41<16:44:11, 11.52s/step, acc=0.4995, loss=2.0579, lr=5.21e-05]
Training:  53%|█████▎    | 5854/11082 [18:48:53<16:49:01, 11.58s/step, acc=0.4995, loss=2.0579, lr=5.21e-05]
Training:  53%|█████▎    | 5855/11082 [18:49:04<16:40:07, 11.48s/step, acc=0.4995, loss=2.0579, lr=5.21e-05]
Training:  53%|█████▎    | 5856/11082 [18:49:16<16:49:00, 11.58s/step, acc=0.4995, loss=2.0579, lr=5.21e-05]
Training:  53%|█████▎    | 5857/11082 [18:49:28<16:48:23, 11.58s/step, acc=0.4995, loss=2.0579, lr=5.21e-05]
Training:  53%|█████▎    | 5858/11082 [18:49:39<16:45:51, 11.55s/step, acc=0.4995, loss=2.0579, lr=5.21e-05]
Training:  53%|█████▎    | 5859/11082 [18:49:51<16:52:55, 11.64s/step, acc=0.4995, loss=2.0579, lr=5.21e-05]
Training:  53%|█████▎    | 5860/11082 [18:50:03<16:53:27, 11.64s/step, acc=0.4995, loss=2.0579, lr=5.21e-05]
Training:  53%|█████▎    | 5861/11082 [18:50:14<16:37:10, 11.46s/step, acc=0.4995, loss=2.0579, lr=5.21e-05]
Training:  53%|█████▎    | 5862/11082 [18:50:26<16:50:31, 11.62s/step, acc=0.4995, loss=2.0579, lr=5.21e-05]
Training:  53%|█████▎    | 5863/11082 [18:50:38<16:51:19, 11.63s/step, acc=0.4995, loss=2.0579, lr=5.21e-05]
Training:  53%|█████▎    | 5864/11082 [18:50:49<16:50:30, 11.62s/step, acc=0.4995, loss=2.0579, lr=5.21e-05]
Training:  53%|█████▎    | 5865/11082 [18:51:01<16:47:08, 11.58s/step, acc=0.4995, loss=2.0579, lr=5.21e-05]
Training:  53%|█████▎    | 5866/11082 [18:51:12<16:47:14, 11.59s/step, acc=0.4995, loss=2.0579, lr=5.21e-05]
Training:  53%|█████▎    | 5867/11082 [18:51:24<16:43:27, 11.55s/step, acc=0.4995, loss=2.0579, lr=5.21e-05]
Training:  53%|█████▎    | 5868/11082 [18:51:35<16:42:33, 11.54s/step, acc=0.4995, loss=2.0579, lr=5.21e-05]
Training:  53%|█████▎    | 5869/11082 [18:51:46<16:37:00, 11.48s/step, acc=0.4995, loss=2.0579, lr=5.21e-05]
Training:  53%|█████▎    | 5870/11082 [18:51:58<16:42:43, 11.54s/step, acc=0.4995, loss=2.0579, lr=5.21e-05]
Training:  53%|█████▎    | 5871/11082 [18:52:10<16:37:52, 11.49s/step, acc=0.4995, loss=2.0579, lr=5.21e-05]
Training:  53%|█████▎    | 5872/11082 [18:52:21<16:35:59, 11.47s/step, acc=0.4995, loss=2.0579, lr=5.21e-05]
Training:  53%|█████▎    | 5873/11082 [18:52:33<16:47:54, 11.61s/step, acc=0.4995, loss=2.0579, lr=5.21e-05]
Training:  53%|█████▎    | 5874/11082 [18:52:45<17:00:13, 11.75s/step, acc=0.4995, loss=2.0579, lr=5.21e-05]
Training:  53%|█████▎    | 5875/11082 [18:52:57<16:54:12, 11.69s/step, acc=0.4995, loss=2.0579, lr=5.21e-05]
Training:  53%|█████▎    | 5876/11082 [18:53:08<16:46:21, 11.60s/step, acc=0.4995, loss=2.0579, lr=5.21e-05]
Training:  53%|█████▎    | 5877/11082 [18:53:20<16:57:34, 11.73s/step, acc=0.4995, loss=2.0579, lr=5.21e-05]
Training:  53%|█████▎    | 5878/11082 [18:53:31<16:49:46, 11.64s/step, acc=0.4995, loss=2.0579, lr=5.21e-05]
Training:  53%|█████▎    | 5879/11082 [18:53:43<16:55:51, 11.71s/step, acc=0.4995, loss=2.0579, lr=5.21e-05]
Training:  53%|█████▎    | 5880/11082 [18:53:55<16:46:14, 11.61s/step, acc=0.4995, loss=2.0579, lr=5.21e-05]
Training:  53%|█████▎    | 5881/11082 [18:54:07<16:56:40, 11.73s/step, acc=0.4995, loss=2.0579, lr=5.21e-05]
Training:  53%|█████▎    | 5882/11082 [18:54:18<16:46:58, 11.62s/step, acc=0.4995, loss=2.0579, lr=5.21e-05]
Training:  53%|█████▎    | 5883/11082 [18:54:29<16:42:32, 11.57s/step, acc=0.4995, loss=2.0579, lr=5.21e-05]
Training:  53%|█████▎    | 5884/11082 [18:54:42<17:00:13, 11.78s/step, acc=0.4995, loss=2.0579, lr=5.21e-05]
Training:  53%|█████▎    | 5885/11082 [18:54:53<16:43:10, 11.58s/step, acc=0.4995, loss=2.0579, lr=5.21e-05]
Training:  53%|█████▎    | 5886/11082 [18:55:04<16:41:01, 11.56s/step, acc=0.4995, loss=2.0579, lr=5.21e-05]
Training:  53%|█████▎    | 5887/11082 [18:55:16<16:39:14, 11.54s/step, acc=0.4995, loss=2.0579, lr=5.21e-05]
Training:  53%|█████▎    | 5888/11082 [18:55:28<16:44:05, 11.60s/step, acc=0.4995, loss=2.0579, lr=5.21e-05]
Training:  53%|█████▎    | 5889/11082 [18:55:39<16:41:54, 11.58s/step, acc=0.4995, loss=2.0579, lr=5.21e-05]
Training:  53%|█████▎    | 5890/11082 [18:55:51<16:46:28, 11.63s/step, acc=0.4995, loss=2.0579, lr=5.21e-05]
Training:  53%|█████▎    | 5891/11082 [18:56:03<16:46:52, 11.64s/step, acc=0.4995, loss=2.0579, lr=5.21e-05]
Training:  53%|█████▎    | 5892/11082 [18:56:14<16:42:25, 11.59s/step, acc=0.4995, loss=2.0579, lr=5.21e-05]
Training:  53%|█████▎    | 5893/11082 [18:56:25<16:36:45, 11.53s/step, acc=0.4995, loss=2.0579, lr=5.21e-05]
Training:  53%|█████▎    | 5894/11082 [18:56:38<16:52:56, 11.71s/step, acc=0.4995, loss=2.0579, lr=5.21e-05]
Training:  53%|█████▎    | 5895/11082 [18:56:49<16:48:38, 11.67s/step, acc=0.4995, loss=2.0579, lr=5.21e-05]
Training:  53%|█████▎    | 5896/11082 [18:57:01<16:50:16, 11.69s/step, acc=0.4995, loss=2.0579, lr=5.21e-05]
Training:  53%|█████▎    | 5897/11082 [18:57:12<16:46:23, 11.65s/step, acc=0.4995, loss=2.0579, lr=5.21e-05]
Training:  53%|█████▎    | 5898/11082 [18:57:24<16:32:30, 11.49s/step, acc=0.4995, loss=2.0579, lr=5.21e-05]
Training:  53%|█████▎    | 5899/11082 [18:57:35<16:38:07, 11.55s/step, acc=0.4995, loss=2.0579, lr=5.21e-05]
Training:  53%|█████▎    | 5900/11082 [18:57:46<16:27:44, 11.44s/step, acc=0.4995, loss=2.0579, lr=5.21e-05]
Training:  53%|█████▎    | 5900/11082 [18:57:54<16:27:44, 11.44s/step, acc=0.5001, loss=2.0562, lr=5.14e-05]2026-06-19 09:51:22 | INFO | __main__ | epoch=1 step=5900 train_loss=2.056248 train_acc=0.500143 lr=5.143e-05 grad_norm=0.8789 acc_d0=0.7722 acc_d1=0.6304 acc_d2=0.5101 acc_d3=0.4119 acc_d4=0.3318 acc_d5=0.2679 acc_d6=0.2169 acc_d7=0.1788

Training:  53%|█████▎    | 5901/11082 [18:57:58<16:25:17, 11.41s/step, acc=0.5001, loss=2.0562, lr=5.14e-05]
Training:  53%|█████▎    | 5902/11082 [18:58:10<16:38:40, 11.57s/step, acc=0.5001, loss=2.0562, lr=5.14e-05]
Training:  53%|█████▎    | 5903/11082 [18:58:21<16:34:17, 11.52s/step, acc=0.5001, loss=2.0562, lr=5.14e-05]
Training:  53%|█████▎    | 5904/11082 [18:58:33<16:37:01, 11.55s/step, acc=0.5001, loss=2.0562, lr=5.14e-05]
Training:  53%|█████▎    | 5905/11082 [18:58:45<16:44:21, 11.64s/step, acc=0.5001, loss=2.0562, lr=5.14e-05]
Training:  53%|█████▎    | 5906/11082 [18:58:56<16:44:49, 11.65s/step, acc=0.5001, loss=2.0562, lr=5.14e-05]
Training:  53%|█████▎    | 5907/11082 [18:59:07<16:28:43, 11.46s/step, acc=0.5001, loss=2.0562, lr=5.14e-05]
Training:  53%|█████▎    | 5908/11082 [18:59:19<16:40:17, 11.60s/step, acc=0.5001, loss=2.0562, lr=5.14e-05]
Training:  53%|█████▎    | 5909/11082 [18:59:31<16:34:05, 11.53s/step, acc=0.5001, loss=2.0562, lr=5.14e-05]
Training:  53%|█████▎    | 5910/11082 [18:59:42<16:33:38, 11.53s/step, acc=0.5001, loss=2.0562, lr=5.14e-05]
Training:  53%|█████▎    | 5911/11082 [18:59:54<16:44:18, 11.65s/step, acc=0.5001, loss=2.0562, lr=5.14e-05]
Training:  53%|█████▎    | 5912/11082 [19:00:06<16:46:53, 11.69s/step, acc=0.5001, loss=2.0562, lr=5.14e-05]
Training:  53%|█████▎    | 5913/11082 [19:00:17<16:27:43, 11.47s/step, acc=0.5001, loss=2.0562, lr=5.14e-05]
Training:  53%|█████▎    | 5914/11082 [19:00:29<16:40:39, 11.62s/step, acc=0.5001, loss=2.0562, lr=5.14e-05]
Training:  53%|█████▎    | 5915/11082 [19:00:40<16:28:54, 11.48s/step, acc=0.5001, loss=2.0562, lr=5.14e-05]
Training:  53%|█████▎    | 5916/11082 [19:00:52<16:42:52, 11.65s/step, acc=0.5001, loss=2.0562, lr=5.14e-05]
Training:  53%|█████▎    | 5917/11082 [19:01:03<16:37:44, 11.59s/step, acc=0.5001, loss=2.0562, lr=5.14e-05]
Training:  53%|█████▎    | 5918/11082 [19:01:15<16:29:52, 11.50s/step, acc=0.5001, loss=2.0562, lr=5.14e-05]
Training:  53%|█████▎    | 5919/11082 [19:01:26<16:35:16, 11.57s/step, acc=0.5001, loss=2.0562, lr=5.14e-05]
Training:  53%|█████▎    | 5920/11082 [19:01:38<16:29:43, 11.50s/step, acc=0.5001, loss=2.0562, lr=5.14e-05]
Training:  53%|█████▎    | 5921/11082 [19:01:49<16:32:48, 11.54s/step, acc=0.5001, loss=2.0562, lr=5.14e-05]
Training:  53%|█████▎    | 5922/11082 [19:02:00<16:19:03, 11.38s/step, acc=0.5001, loss=2.0562, lr=5.14e-05]
Training:  53%|█████▎    | 5923/11082 [19:02:12<16:38:21, 11.61s/step, acc=0.5001, loss=2.0562, lr=5.14e-05]
Training:  53%|█████▎    | 5924/11082 [19:02:24<16:36:36, 11.59s/step, acc=0.5001, loss=2.0562, lr=5.14e-05]
Training:  53%|█████▎    | 5925/11082 [19:02:36<16:36:13, 11.59s/step, acc=0.5001, loss=2.0562, lr=5.14e-05]
Training:  53%|█████▎    | 5926/11082 [19:02:47<16:30:08, 11.52s/step, acc=0.5001, loss=2.0562, lr=5.14e-05]
Training:  53%|█████▎    | 5927/11082 [19:02:58<16:27:32, 11.49s/step, acc=0.5001, loss=2.0562, lr=5.14e-05]
Training:  53%|█████▎    | 5928/11082 [19:03:10<16:33:33, 11.57s/step, acc=0.5001, loss=2.0562, lr=5.14e-05]
Training:  54%|█████▎    | 5929/11082 [19:03:22<16:31:20, 11.54s/step, acc=0.5001, loss=2.0562, lr=5.14e-05]
Training:  54%|█████▎    | 5930/11082 [19:03:33<16:30:30, 11.54s/step, acc=0.5001, loss=2.0562, lr=5.14e-05]
Training:  54%|█████▎    | 5931/11082 [19:03:45<16:27:06, 11.50s/step, acc=0.5001, loss=2.0562, lr=5.14e-05]
Training:  54%|█████▎    | 5932/11082 [19:03:56<16:35:11, 11.59s/step, acc=0.5001, loss=2.0562, lr=5.14e-05]
Training:  54%|█████▎    | 5933/11082 [19:04:07<16:21:32, 11.44s/step, acc=0.5001, loss=2.0562, lr=5.14e-05]
Training:  54%|█████▎    | 5934/11082 [19:04:19<16:33:27, 11.58s/step, acc=0.5001, loss=2.0562, lr=5.14e-05]
Training:  54%|█████▎    | 5935/11082 [19:04:31<16:33:25, 11.58s/step, acc=0.5001, loss=2.0562, lr=5.14e-05]
Training:  54%|█████▎    | 5936/11082 [19:04:42<16:30:31, 11.55s/step, acc=0.5001, loss=2.0562, lr=5.14e-05]
Training:  54%|█████▎    | 5937/11082 [19:04:54<16:30:19, 11.55s/step, acc=0.5001, loss=2.0562, lr=5.14e-05]
Training:  54%|█████▎    | 5938/11082 [19:05:06<16:41:43, 11.68s/step, acc=0.5001, loss=2.0562, lr=5.14e-05]
Training:  54%|█████▎    | 5939/11082 [19:05:17<16:30:39, 11.56s/step, acc=0.5001, loss=2.0562, lr=5.14e-05]
Training:  54%|█████▎    | 5940/11082 [19:05:29<16:35:17, 11.61s/step, acc=0.5001, loss=2.0562, lr=5.14e-05]
Training:  54%|█████▎    | 5941/11082 [19:05:40<16:20:52, 11.45s/step, acc=0.5001, loss=2.0562, lr=5.14e-05]
Training:  54%|█████▎    | 5942/11082 [19:05:52<16:37:15, 11.64s/step, acc=0.5001, loss=2.0562, lr=5.14e-05]
Training:  54%|█████▎    | 5943/11082 [19:06:03<16:24:30, 11.49s/step, acc=0.5001, loss=2.0562, lr=5.14e-05]
Training:  54%|█████▎    | 5944/11082 [19:06:15<16:25:52, 11.51s/step, acc=0.5001, loss=2.0562, lr=5.14e-05]
Training:  54%|█████▎    | 5945/11082 [19:06:26<16:18:53, 11.43s/step, acc=0.5001, loss=2.0562, lr=5.14e-05]
Training:  54%|█████▎    | 5946/11082 [19:06:38<16:24:35, 11.50s/step, acc=0.5001, loss=2.0562, lr=5.14e-05]
Training:  54%|█████▎    | 5947/11082 [19:06:49<16:28:23, 11.55s/step, acc=0.5001, loss=2.0562, lr=5.14e-05]
Training:  54%|█████▎    | 5948/11082 [19:07:01<16:23:07, 11.49s/step, acc=0.5001, loss=2.0562, lr=5.14e-05]
Training:  54%|█████▎    | 5949/11082 [19:07:12<16:20:46, 11.46s/step, acc=0.5001, loss=2.0562, lr=5.14e-05]
Training:  54%|█████▎    | 5950/11082 [19:07:24<16:36:06, 11.65s/step, acc=0.5001, loss=2.0562, lr=5.14e-05]
Training:  54%|█████▎    | 5950/11082 [19:07:31<16:36:06, 11.65s/step, acc=0.4992, loss=2.0599, lr=5.08e-05]2026-06-19 10:00:59 | INFO | __main__ | epoch=1 step=5950 train_loss=2.059850 train_acc=0.499160 lr=5.078e-05 grad_norm=0.6758 acc_d0=0.7534 acc_d1=0.6049 acc_d2=0.4844 acc_d3=0.3875 acc_d4=0.3088 acc_d5=0.2463 acc_d6=0.1993 acc_d7=0.1632

Training:  54%|█████▎    | 5951/11082 [19:07:36<16:26:19, 11.53s/step, acc=0.4992, loss=2.0599, lr=5.08e-05]
Training:  54%|█████▎    | 5952/11082 [19:07:47<16:18:01, 11.44s/step, acc=0.4992, loss=2.0599, lr=5.08e-05]
Training:  54%|█████▎    | 5953/11082 [19:07:59<16:39:15, 11.69s/step, acc=0.4992, loss=2.0599, lr=5.08e-05]
Training:  54%|█████▎    | 5954/11082 [19:08:11<16:42:38, 11.73s/step, acc=0.4992, loss=2.0599, lr=5.08e-05]
Training:  54%|█████▎    | 5955/11082 [19:08:22<16:31:12, 11.60s/step, acc=0.4992, loss=2.0599, lr=5.08e-05]
Training:  54%|█████▎    | 5956/11082 [19:08:34<16:29:22, 11.58s/step, acc=0.4992, loss=2.0599, lr=5.08e-05]
Training:  54%|█████▍    | 5957/11082 [19:08:45<16:19:11, 11.46s/step, acc=0.4992, loss=2.0599, lr=5.08e-05]
Training:  54%|█████▍    | 5958/11082 [19:08:57<16:34:09, 11.64s/step, acc=0.4992, loss=2.0599, lr=5.08e-05]
Training:  54%|█████▍    | 5959/11082 [19:09:08<16:31:13, 11.61s/step, acc=0.4992, loss=2.0599, lr=5.08e-05]
Training:  54%|█████▍    | 5960/11082 [19:09:20<16:34:14, 11.65s/step, acc=0.4992, loss=2.0599, lr=5.08e-05]
Training:  54%|█████▍    | 5961/11082 [19:09:31<16:24:26, 11.53s/step, acc=0.4992, loss=2.0599, lr=5.08e-05]
Training:  54%|█████▍    | 5962/11082 [19:09:44<16:38:28, 11.70s/step, acc=0.4992, loss=2.0599, lr=5.08e-05]
Training:  54%|█████▍    | 5963/11082 [19:09:55<16:32:12, 11.63s/step, acc=0.4992, loss=2.0599, lr=5.08e-05]
Training:  54%|█████▍    | 5964/11082 [19:10:07<16:38:00, 11.70s/step, acc=0.4992, loss=2.0599, lr=5.08e-05]
Training:  54%|█████▍    | 5965/11082 [19:10:18<16:35:55, 11.68s/step, acc=0.4992, loss=2.0599, lr=5.08e-05]
Training:  54%|█████▍    | 5966/11082 [19:10:30<16:32:01, 11.63s/step, acc=0.4992, loss=2.0599, lr=5.08e-05]
Training:  54%|█████▍    | 5967/11082 [19:10:42<16:32:34, 11.64s/step, acc=0.4992, loss=2.0599, lr=5.08e-05]
Training:  54%|█████▍    | 5968/11082 [19:10:53<16:24:18, 11.55s/step, acc=0.4992, loss=2.0599, lr=5.08e-05]
Training:  54%|█████▍    | 5969/11082 [19:11:04<16:21:16, 11.52s/step, acc=0.4992, loss=2.0599, lr=5.08e-05]
Training:  54%|█████▍    | 5970/11082 [19:11:16<16:22:39, 11.53s/step, acc=0.4992, loss=2.0599, lr=5.08e-05]
Training:  54%|█████▍    | 5971/11082 [19:11:28<16:21:28, 11.52s/step, acc=0.4992, loss=2.0599, lr=5.08e-05]
Training:  54%|█████▍    | 5972/11082 [19:11:39<16:25:22, 11.57s/step, acc=0.4992, loss=2.0599, lr=5.08e-05]
Training:  54%|█████▍    | 5973/11082 [19:11:51<16:33:06, 11.66s/step, acc=0.4992, loss=2.0599, lr=5.08e-05]
Training:  54%|█████▍    | 5974/11082 [19:12:03<16:41:58, 11.77s/step, acc=0.4992, loss=2.0599, lr=5.08e-05]
Training:  54%|█████▍    | 5975/11082 [19:12:14<16:30:05, 11.63s/step, acc=0.4992, loss=2.0599, lr=5.08e-05]
Training:  54%|█████▍    | 5976/11082 [19:12:26<16:22:53, 11.55s/step, acc=0.4992, loss=2.0599, lr=5.08e-05]
Training:  54%|█████▍    | 5977/11082 [19:12:37<16:25:15, 11.58s/step, acc=0.4992, loss=2.0599, lr=5.08e-05]
Training:  54%|█████▍    | 5978/11082 [19:12:49<16:35:37, 11.70s/step, acc=0.4992, loss=2.0599, lr=5.08e-05]
Training:  54%|█████▍    | 5979/11082 [19:13:01<16:27:20, 11.61s/step, acc=0.4992, loss=2.0599, lr=5.08e-05]
Training:  54%|█████▍    | 5980/11082 [19:13:12<16:26:12, 11.60s/step, acc=0.4992, loss=2.0599, lr=5.08e-05]
Training:  54%|█████▍    | 5981/11082 [19:13:24<16:21:04, 11.54s/step, acc=0.4992, loss=2.0599, lr=5.08e-05]
Training:  54%|█████▍    | 5982/11082 [19:13:35<16:23:06, 11.57s/step, acc=0.4992, loss=2.0599, lr=5.08e-05]
Training:  54%|█████▍    | 5983/11082 [19:13:47<16:23:15, 11.57s/step, acc=0.4992, loss=2.0599, lr=5.08e-05]
Training:  54%|█████▍    | 5984/11082 [19:13:59<16:23:27, 11.57s/step, acc=0.4992, loss=2.0599, lr=5.08e-05]
Training:  54%|█████▍    | 5985/11082 [19:14:10<16:23:33, 11.58s/step, acc=0.4992, loss=2.0599, lr=5.08e-05]
Training:  54%|█████▍    | 5986/11082 [19:14:22<16:20:00, 11.54s/step, acc=0.4992, loss=2.0599, lr=5.08e-05]
Training:  54%|█████▍    | 5987/11082 [19:14:33<16:23:27, 11.58s/step, acc=0.4992, loss=2.0599, lr=5.08e-05]
Training:  54%|█████▍    | 5988/11082 [19:14:45<16:21:10, 11.56s/step, acc=0.4992, loss=2.0599, lr=5.08e-05]
Training:  54%|█████▍    | 5989/11082 [19:14:56<16:22:49, 11.58s/step, acc=0.4992, loss=2.0599, lr=5.08e-05]
Training:  54%|█████▍    | 5990/11082 [19:15:08<16:34:47, 11.72s/step, acc=0.4992, loss=2.0599, lr=5.08e-05]
Training:  54%|█████▍    | 5991/11082 [19:15:20<16:27:41, 11.64s/step, acc=0.4992, loss=2.0599, lr=5.08e-05]
Training:  54%|█████▍    | 5992/11082 [19:15:32<16:29:25, 11.66s/step, acc=0.4992, loss=2.0599, lr=5.08e-05]
Training:  54%|█████▍    | 5993/11082 [19:15:43<16:20:18, 11.56s/step, acc=0.4992, loss=2.0599, lr=5.08e-05]
Training:  54%|█████▍    | 5994/11082 [19:15:55<16:44:15, 11.84s/step, acc=0.4992, loss=2.0599, lr=5.08e-05]
Training:  54%|█████▍    | 5995/11082 [19:16:07<16:33:18, 11.72s/step, acc=0.4992, loss=2.0599, lr=5.08e-05]
Training:  54%|█████▍    | 5996/11082 [19:16:18<16:23:14, 11.60s/step, acc=0.4992, loss=2.0599, lr=5.08e-05]
Training:  54%|█████▍    | 5997/11082 [19:16:30<16:24:47, 11.62s/step, acc=0.4992, loss=2.0599, lr=5.08e-05]
Training:  54%|█████▍    | 5998/11082 [19:16:42<16:25:03, 11.63s/step, acc=0.4992, loss=2.0599, lr=5.08e-05]
Training:  54%|█████▍    | 5999/11082 [19:16:53<16:20:15, 11.57s/step, acc=0.4992, loss=2.0599, lr=5.08e-05]
Training:  54%|█████▍    | 6000/11082 [19:17:05<16:21:22, 11.59s/step, acc=0.4992, loss=2.0599, lr=5.08e-05]
Training:  54%|█████▍    | 6000/11082 [19:17:12<16:21:22, 11.59s/step, acc=0.5004, loss=2.0581, lr=5.01e-05]2026-06-19 10:10:40 | INFO | __main__ | epoch=1 step=6000 train_loss=2.058050 train_acc=0.500357 lr=5.014e-05 grad_norm=0.6562 acc_d0=0.7689 acc_d1=0.6246 acc_d2=0.5027 acc_d3=0.4044 acc_d4=0.3259 acc_d5=0.2647 acc_d6=0.2135 acc_d7=0.1728

Training:  54%|█████▍    | 6001/11082 [19:17:16<16:22:17, 11.60s/step, acc=0.5004, loss=2.0581, lr=5.01e-05]
Training:  54%|█████▍    | 6002/11082 [19:17:28<16:21:23, 11.59s/step, acc=0.5004, loss=2.0581, lr=5.01e-05]
Training:  54%|█████▍    | 6003/11082 [19:17:40<16:28:39, 11.68s/step, acc=0.5004, loss=2.0581, lr=5.01e-05]
Training:  54%|█████▍    | 6004/11082 [19:17:51<16:20:40, 11.59s/step, acc=0.5004, loss=2.0581, lr=5.01e-05]
Training:  54%|█████▍    | 6005/11082 [19:18:02<16:13:15, 11.50s/step, acc=0.5004, loss=2.0581, lr=5.01e-05]
Training:  54%|█████▍    | 6006/11082 [19:18:14<16:17:23, 11.55s/step, acc=0.5004, loss=2.0581, lr=5.01e-05]
Training:  54%|█████▍    | 6007/11082 [19:18:26<16:21:49, 11.61s/step, acc=0.5004, loss=2.0581, lr=5.01e-05]
Training:  54%|█████▍    | 6008/11082 [19:18:37<16:14:01, 11.52s/step, acc=0.5004, loss=2.0581, lr=5.01e-05]
Training:  54%|█████▍    | 6009/11082 [19:18:49<16:25:42, 11.66s/step, acc=0.5004, loss=2.0581, lr=5.01e-05]
Training:  54%|█████▍    | 6010/11082 [19:19:01<16:25:27, 11.66s/step, acc=0.5004, loss=2.0581, lr=5.01e-05]
Training:  54%|█████▍    | 6011/11082 [19:19:12<16:17:59, 11.57s/step, acc=0.5004, loss=2.0581, lr=5.01e-05]
Training:  54%|█████▍    | 6012/11082 [19:19:24<16:17:28, 11.57s/step, acc=0.5004, loss=2.0581, lr=5.01e-05]
Training:  54%|█████▍    | 6013/11082 [19:19:35<16:15:28, 11.55s/step, acc=0.5004, loss=2.0581, lr=5.01e-05]
Training:  54%|█████▍    | 6014/11082 [19:19:47<16:22:53, 11.64s/step, acc=0.5004, loss=2.0581, lr=5.01e-05]
Training:  54%|█████▍    | 6015/11082 [19:19:58<16:00:09, 11.37s/step, acc=0.5004, loss=2.0581, lr=5.01e-05]
Training:  54%|█████▍    | 6016/11082 [19:20:10<16:23:33, 11.65s/step, acc=0.5004, loss=2.0581, lr=5.01e-05]
Training:  54%|█████▍    | 6017/11082 [19:20:21<16:09:24, 11.48s/step, acc=0.5004, loss=2.0581, lr=5.01e-05]
Training:  54%|█████▍    | 6018/11082 [19:20:33<16:08:33, 11.48s/step, acc=0.5004, loss=2.0581, lr=5.01e-05]
Training:  54%|█████▍    | 6019/11082 [19:20:44<16:06:50, 11.46s/step, acc=0.5004, loss=2.0581, lr=5.01e-05]
Training:  54%|█████▍    | 6020/11082 [19:20:56<16:11:20, 11.51s/step, acc=0.5004, loss=2.0581, lr=5.01e-05]
Training:  54%|█████▍    | 6021/11082 [19:21:08<16:20:19, 11.62s/step, acc=0.5004, loss=2.0581, lr=5.01e-05]
Training:  54%|█████▍    | 6022/11082 [19:21:19<16:15:31, 11.57s/step, acc=0.5004, loss=2.0581, lr=5.01e-05]
Training:  54%|█████▍    | 6023/11082 [19:21:31<16:20:23, 11.63s/step, acc=0.5004, loss=2.0581, lr=5.01e-05]
Training:  54%|█████▍    | 6024/11082 [19:21:42<16:15:30, 11.57s/step, acc=0.5004, loss=2.0581, lr=5.01e-05]
Training:  54%|█████▍    | 6025/11082 [19:21:54<16:22:29, 11.66s/step, acc=0.5004, loss=2.0581, lr=5.01e-05]
Training:  54%|█████▍    | 6026/11082 [19:22:05<16:14:58, 11.57s/step, acc=0.5004, loss=2.0581, lr=5.01e-05]
Training:  54%|█████▍    | 6027/11082 [19:22:17<16:12:42, 11.55s/step, acc=0.5004, loss=2.0581, lr=5.01e-05]
Training:  54%|█████▍    | 6028/11082 [19:22:29<16:22:21, 11.66s/step, acc=0.5004, loss=2.0581, lr=5.01e-05]
Training:  54%|█████▍    | 6029/11082 [19:22:40<16:03:18, 11.44s/step, acc=0.5004, loss=2.0581, lr=5.01e-05]
Training:  54%|█████▍    | 6030/11082 [19:22:52<16:13:53, 11.57s/step, acc=0.5004, loss=2.0581, lr=5.01e-05]
Training:  54%|█████▍    | 6031/11082 [19:23:03<16:18:46, 11.63s/step, acc=0.5004, loss=2.0581, lr=5.01e-05]
Training:  54%|█████▍    | 6032/11082 [19:23:15<16:12:10, 11.55s/step, acc=0.5004, loss=2.0581, lr=5.01e-05]
Training:  54%|█████▍    | 6033/11082 [19:23:26<16:12:48, 11.56s/step, acc=0.5004, loss=2.0581, lr=5.01e-05]
Training:  54%|█████▍    | 6034/11082 [19:23:38<16:13:43, 11.57s/step, acc=0.5004, loss=2.0581, lr=5.01e-05]
Training:  54%|█████▍    | 6035/11082 [19:23:49<16:11:15, 11.55s/step, acc=0.5004, loss=2.0581, lr=5.01e-05]
Training:  54%|█████▍    | 6036/11082 [19:24:01<16:14:17, 11.58s/step, acc=0.5004, loss=2.0581, lr=5.01e-05]
Training:  54%|█████▍    | 6037/11082 [19:24:12<16:06:28, 11.49s/step, acc=0.5004, loss=2.0581, lr=5.01e-05]
Training:  54%|█████▍    | 6038/11082 [19:24:24<16:09:34, 11.53s/step, acc=0.5004, loss=2.0581, lr=5.01e-05]
Training:  54%|█████▍    | 6039/11082 [19:24:36<16:21:35, 11.68s/step, acc=0.5004, loss=2.0581, lr=5.01e-05]
Training:  55%|█████▍    | 6040/11082 [19:24:48<16:23:13, 11.70s/step, acc=0.5004, loss=2.0581, lr=5.01e-05]
Training:  55%|█████▍    | 6041/11082 [19:24:59<16:19:02, 11.65s/step, acc=0.5004, loss=2.0581, lr=5.01e-05]
Training:  55%|█████▍    | 6042/11082 [19:25:11<16:12:53, 11.58s/step, acc=0.5004, loss=2.0581, lr=5.01e-05]
Training:  55%|█████▍    | 6043/11082 [19:25:22<16:06:11, 11.50s/step, acc=0.5004, loss=2.0581, lr=5.01e-05]
Training:  55%|█████▍    | 6044/11082 [19:25:34<16:08:59, 11.54s/step, acc=0.5004, loss=2.0581, lr=5.01e-05]
Training:  55%|█████▍    | 6045/11082 [19:25:46<16:18:06, 11.65s/step, acc=0.5004, loss=2.0581, lr=5.01e-05]
Training:  55%|█████▍    | 6046/11082 [19:25:57<16:12:10, 11.58s/step, acc=0.5004, loss=2.0581, lr=5.01e-05]
Training:  55%|█████▍    | 6047/11082 [19:26:09<16:11:05, 11.57s/step, acc=0.5004, loss=2.0581, lr=5.01e-05]
Training:  55%|█████▍    | 6048/11082 [19:26:20<16:15:29, 11.63s/step, acc=0.5004, loss=2.0581, lr=5.01e-05]
Training:  55%|█████▍    | 6049/11082 [19:26:32<16:09:00, 11.55s/step, acc=0.5004, loss=2.0581, lr=5.01e-05]
Training:  55%|█████▍    | 6050/11082 [19:26:43<16:03:41, 11.49s/step, acc=0.5004, loss=2.0581, lr=5.01e-05]
Training:  55%|█████▍    | 6050/11082 [19:26:50<16:03:41, 11.49s/step, acc=0.5002, loss=2.0547, lr=4.95e-05]2026-06-19 10:20:18 | INFO | __main__ | epoch=1 step=6050 train_loss=2.054736 train_acc=0.500198 lr=4.950e-05 grad_norm=0.7891 acc_d0=0.7621 acc_d1=0.6166 acc_d2=0.4961 acc_d3=0.3970 acc_d4=0.3189 acc_d5=0.2551 acc_d6=0.2057 acc_d7=0.1702

Training:  55%|█████▍    | 6051/11082 [19:26:55<16:11:17, 11.58s/step, acc=0.5002, loss=2.0547, lr=4.95e-05]
Training:  55%|█████▍    | 6052/11082 [19:27:06<16:10:25, 11.58s/step, acc=0.5002, loss=2.0547, lr=4.95e-05]
Training:  55%|█████▍    | 6053/11082 [19:27:18<16:04:05, 11.50s/step, acc=0.5002, loss=2.0547, lr=4.95e-05]
Training:  55%|█████▍    | 6054/11082 [19:27:30<16:21:45, 11.72s/step, acc=0.5002, loss=2.0547, lr=4.95e-05]
Training:  55%|█████▍    | 6055/11082 [19:27:41<16:10:51, 11.59s/step, acc=0.5002, loss=2.0547, lr=4.95e-05]
Training:  55%|█████▍    | 6056/11082 [19:27:53<16:09:53, 11.58s/step, acc=0.5002, loss=2.0547, lr=4.95e-05]
Training:  55%|█████▍    | 6057/11082 [19:28:04<16:07:01, 11.55s/step, acc=0.5002, loss=2.0547, lr=4.95e-05]
Training:  55%|█████▍    | 6058/11082 [19:28:16<16:04:12, 11.52s/step, acc=0.5002, loss=2.0547, lr=4.95e-05]
Training:  55%|█████▍    | 6059/11082 [19:28:27<16:03:40, 11.51s/step, acc=0.5002, loss=2.0547, lr=4.95e-05]
Training:  55%|█████▍    | 6060/11082 [19:28:39<16:07:41, 11.56s/step, acc=0.5002, loss=2.0547, lr=4.95e-05]
Training:  55%|█████▍    | 6061/11082 [19:28:51<16:15:39, 11.66s/step, acc=0.5002, loss=2.0547, lr=4.95e-05]
Training:  55%|█████▍    | 6062/11082 [19:29:02<16:15:10, 11.66s/step, acc=0.5002, loss=2.0547, lr=4.95e-05]
Training:  55%|█████▍    | 6063/11082 [19:29:14<16:02:24, 11.51s/step, acc=0.5002, loss=2.0547, lr=4.95e-05]
Training:  55%|█████▍    | 6064/11082 [19:29:25<16:07:21, 11.57s/step, acc=0.5002, loss=2.0547, lr=4.95e-05]
Training:  55%|█████▍    | 6065/11082 [19:29:37<16:22:38, 11.75s/step, acc=0.5002, loss=2.0547, lr=4.95e-05]
Training:  55%|█████▍    | 6066/11082 [19:29:48<15:56:31, 11.44s/step, acc=0.5002, loss=2.0547, lr=4.95e-05]
Training:  55%|█████▍    | 6067/11082 [19:30:00<16:06:23, 11.56s/step, acc=0.5002, loss=2.0547, lr=4.95e-05]
Training:  55%|█████▍    | 6068/11082 [19:30:11<15:55:01, 11.43s/step, acc=0.5002, loss=2.0547, lr=4.95e-05]
Training:  55%|█████▍    | 6069/11082 [19:30:23<16:08:15, 11.59s/step, acc=0.5002, loss=2.0547, lr=4.95e-05]
Training:  55%|█████▍    | 6070/11082 [19:30:34<16:01:28, 11.51s/step, acc=0.5002, loss=2.0547, lr=4.95e-05]
Training:  55%|█████▍    | 6071/11082 [19:30:46<16:07:43, 11.59s/step, acc=0.5002, loss=2.0547, lr=4.95e-05]
Training:  55%|█████▍    | 6072/11082 [19:30:58<16:02:11, 11.52s/step, acc=0.5002, loss=2.0547, lr=4.95e-05]
Training:  55%|█████▍    | 6073/11082 [19:31:09<16:11:05, 11.63s/step, acc=0.5002, loss=2.0547, lr=4.95e-05]
Training:  55%|█████▍    | 6074/11082 [19:31:21<16:04:01, 11.55s/step, acc=0.5002, loss=2.0547, lr=4.95e-05]
Training:  55%|█████▍    | 6075/11082 [19:31:33<16:09:51, 11.62s/step, acc=0.5002, loss=2.0547, lr=4.95e-05]
Training:  55%|█████▍    | 6076/11082 [19:31:44<16:08:41, 11.61s/step, acc=0.5002, loss=2.0547, lr=4.95e-05]
Training:  55%|█████▍    | 6077/11082 [19:31:56<16:04:43, 11.57s/step, acc=0.5002, loss=2.0547, lr=4.95e-05]
Training:  55%|█████▍    | 6078/11082 [19:32:07<16:07:44, 11.60s/step, acc=0.5002, loss=2.0547, lr=4.95e-05]
Training:  55%|█████▍    | 6079/11082 [19:32:19<16:08:38, 11.62s/step, acc=0.5002, loss=2.0547, lr=4.95e-05]
Training:  55%|█████▍    | 6080/11082 [19:32:30<15:56:12, 11.47s/step, acc=0.5002, loss=2.0547, lr=4.95e-05]
Training:  55%|█████▍    | 6081/11082 [19:32:42<16:13:34, 11.68s/step, acc=0.5002, loss=2.0547, lr=4.95e-05]
Training:  55%|█████▍    | 6082/11082 [19:32:54<16:08:38, 11.62s/step, acc=0.5002, loss=2.0547, lr=4.95e-05]
Training:  55%|█████▍    | 6083/11082 [19:33:05<16:04:41, 11.58s/step, acc=0.5002, loss=2.0547, lr=4.95e-05]
Training:  55%|█████▍    | 6084/11082 [19:33:17<16:09:37, 11.64s/step, acc=0.5002, loss=2.0547, lr=4.95e-05]
Training:  55%|█████▍    | 6085/11082 [19:33:29<16:06:55, 11.61s/step, acc=0.5002, loss=2.0547, lr=4.95e-05]
Training:  55%|█████▍    | 6086/11082 [19:33:40<16:05:31, 11.60s/step, acc=0.5002, loss=2.0547, lr=4.95e-05]
Training:  55%|█████▍    | 6087/11082 [19:33:52<16:15:40, 11.72s/step, acc=0.5002, loss=2.0547, lr=4.95e-05]
Training:  55%|█████▍    | 6088/11082 [19:34:03<16:05:55, 11.61s/step, acc=0.5002, loss=2.0547, lr=4.95e-05]
Training:  55%|█████▍    | 6089/11082 [19:34:15<16:04:49, 11.59s/step, acc=0.5002, loss=2.0547, lr=4.95e-05]
Training:  55%|█████▍    | 6090/11082 [19:34:27<16:03:21, 11.58s/step, acc=0.5002, loss=2.0547, lr=4.95e-05]
Training:  55%|█████▍    | 6091/11082 [19:34:38<16:06:35, 11.62s/step, acc=0.5002, loss=2.0547, lr=4.95e-05]
Training:  55%|█████▍    | 6092/11082 [19:34:50<15:57:31, 11.51s/step, acc=0.5002, loss=2.0547, lr=4.95e-05]
Training:  55%|█████▍    | 6093/11082 [19:35:01<15:50:55, 11.44s/step, acc=0.5002, loss=2.0547, lr=4.95e-05]
Training:  55%|█████▍    | 6094/11082 [19:35:13<16:02:29, 11.58s/step, acc=0.5002, loss=2.0547, lr=4.95e-05]
Training:  55%|█████▍    | 6095/11082 [19:35:24<16:06:04, 11.62s/step, acc=0.5002, loss=2.0547, lr=4.95e-05]
Training:  55%|█████▌    | 6096/11082 [19:35:36<16:04:11, 11.60s/step, acc=0.5002, loss=2.0547, lr=4.95e-05]
Training:  55%|█████▌    | 6097/11082 [19:35:47<15:56:00, 11.51s/step, acc=0.5002, loss=2.0547, lr=4.95e-05]
Training:  55%|█████▌    | 6098/11082 [19:35:59<16:03:08, 11.59s/step, acc=0.5002, loss=2.0547, lr=4.95e-05]
Training:  55%|█████▌    | 6099/11082 [19:36:10<15:54:57, 11.50s/step, acc=0.5002, loss=2.0547, lr=4.95e-05]
Training:  55%|█████▌    | 6100/11082 [19:36:22<15:50:36, 11.45s/step, acc=0.5002, loss=2.0547, lr=4.95e-05]
Training:  55%|█████▌    | 6100/11082 [19:36:29<15:50:36, 11.45s/step, acc=0.4992, loss=2.0603, lr=4.89e-05]2026-06-19 10:29:57 | INFO | __main__ | epoch=1 step=6100 train_loss=2.060317 train_acc=0.499166 lr=4.885e-05 grad_norm=0.9531 acc_d0=0.7647 acc_d1=0.6229 acc_d2=0.5039 acc_d3=0.4059 acc_d4=0.3269 acc_d5=0.2632 acc_d6=0.2147 acc_d7=0.1749

Training:  55%|█████▌    | 6101/11082 [19:36:33<15:55:55, 11.51s/step, acc=0.4992, loss=2.0603, lr=4.89e-05]
Training:  55%|█████▌    | 6102/11082 [19:36:45<15:58:59, 11.55s/step, acc=0.4992, loss=2.0603, lr=4.89e-05]
Training:  55%|█████▌    | 6103/11082 [19:36:56<15:52:00, 11.47s/step, acc=0.4992, loss=2.0603, lr=4.89e-05]
Training:  55%|█████▌    | 6104/11082 [19:37:07<15:45:11, 11.39s/step, acc=0.4992, loss=2.0603, lr=4.89e-05]
Training:  55%|█████▌    | 6105/11082 [19:37:20<16:11:14, 11.71s/step, acc=0.4992, loss=2.0603, lr=4.89e-05]
Training:  55%|█████▌    | 6106/11082 [19:37:31<15:58:24, 11.56s/step, acc=0.4992, loss=2.0603, lr=4.89e-05]
Training:  55%|█████▌    | 6107/11082 [19:37:43<15:55:40, 11.53s/step, acc=0.4992, loss=2.0603, lr=4.89e-05]
Training:  55%|█████▌    | 6108/11082 [19:37:54<15:53:05, 11.50s/step, acc=0.4992, loss=2.0603, lr=4.89e-05]
Training:  55%|█████▌    | 6109/11082 [19:38:06<16:01:10, 11.60s/step, acc=0.4992, loss=2.0603, lr=4.89e-05]
Training:  55%|█████▌    | 6110/11082 [19:38:18<16:03:43, 11.63s/step, acc=0.4992, loss=2.0603, lr=4.89e-05]
Training:  55%|█████▌    | 6111/11082 [19:38:29<15:58:10, 11.57s/step, acc=0.4992, loss=2.0603, lr=4.89e-05]
Training:  55%|█████▌    | 6112/11082 [19:38:40<15:52:37, 11.50s/step, acc=0.4992, loss=2.0603, lr=4.89e-05]
Training:  55%|█████▌    | 6113/11082 [19:38:52<15:51:08, 11.49s/step, acc=0.4992, loss=2.0603, lr=4.89e-05]
Training:  55%|█████▌    | 6114/11082 [19:39:04<16:04:20, 11.65s/step, acc=0.4992, loss=2.0603, lr=4.89e-05]
Training:  55%|█████▌    | 6115/11082 [19:39:15<15:54:46, 11.53s/step, acc=0.4992, loss=2.0603, lr=4.89e-05]
Training:  55%|█████▌    | 6116/11082 [19:39:27<16:03:36, 11.64s/step, acc=0.4992, loss=2.0603, lr=4.89e-05]
Training:  55%|█████▌    | 6117/11082 [19:39:38<15:56:15, 11.56s/step, acc=0.4992, loss=2.0603, lr=4.89e-05]
Training:  55%|█████▌    | 6118/11082 [19:39:50<15:49:44, 11.48s/step, acc=0.4992, loss=2.0603, lr=4.89e-05]
Training:  55%|█████▌    | 6119/11082 [19:40:02<16:10:53, 11.74s/step, acc=0.4992, loss=2.0603, lr=4.89e-05]
Training:  55%|█████▌    | 6120/11082 [19:40:13<15:56:39, 11.57s/step, acc=0.4992, loss=2.0603, lr=4.89e-05]
Training:  55%|█████▌    | 6121/11082 [19:40:25<16:14:55, 11.79s/step, acc=0.4992, loss=2.0603, lr=4.89e-05]
Training:  55%|█████▌    | 6122/11082 [19:40:37<16:08:19, 11.71s/step, acc=0.4992, loss=2.0603, lr=4.89e-05]
Training:  55%|█████▌    | 6123/11082 [19:40:48<15:49:06, 11.48s/step, acc=0.4992, loss=2.0603, lr=4.89e-05]
Training:  55%|█████▌    | 6124/11082 [19:41:00<15:53:19, 11.54s/step, acc=0.4992, loss=2.0603, lr=4.89e-05]
Training:  55%|█████▌    | 6125/11082 [19:41:11<15:56:03, 11.57s/step, acc=0.4992, loss=2.0603, lr=4.89e-05]
Training:  55%|█████▌    | 6126/11082 [19:41:23<15:58:55, 11.61s/step, acc=0.4992, loss=2.0603, lr=4.89e-05]
Training:  55%|█████▌    | 6127/11082 [19:41:35<16:07:12, 11.71s/step, acc=0.4992, loss=2.0603, lr=4.89e-05]
Training:  55%|█████▌    | 6128/11082 [19:41:46<16:01:32, 11.65s/step, acc=0.4992, loss=2.0603, lr=4.89e-05]
Training:  55%|█████▌    | 6129/11082 [19:41:58<15:57:45, 11.60s/step, acc=0.4992, loss=2.0603, lr=4.89e-05]
Training:  55%|█████▌    | 6130/11082 [19:42:10<15:58:29, 11.61s/step, acc=0.4992, loss=2.0603, lr=4.89e-05]
Training:  55%|█████▌    | 6131/11082 [19:42:21<16:02:11, 11.66s/step, acc=0.4992, loss=2.0603, lr=4.89e-05]
Training:  55%|█████▌    | 6132/11082 [19:42:33<15:55:33, 11.58s/step, acc=0.4992, loss=2.0603, lr=4.89e-05]
Training:  55%|█████▌    | 6133/11082 [19:42:44<15:47:58, 11.49s/step, acc=0.4992, loss=2.0603, lr=4.89e-05]
Training:  55%|█████▌    | 6134/11082 [19:42:55<15:48:06, 11.50s/step, acc=0.4992, loss=2.0603, lr=4.89e-05]
Training:  55%|█████▌    | 6135/11082 [19:43:08<16:00:56, 11.65s/step, acc=0.4992, loss=2.0603, lr=4.89e-05]
Training:  55%|█████▌    | 6136/11082 [19:43:19<15:57:24, 11.61s/step, acc=0.4992, loss=2.0603, lr=4.89e-05]
Training:  55%|█████▌    | 6137/11082 [19:43:31<15:54:10, 11.58s/step, acc=0.4992, loss=2.0603, lr=4.89e-05]
Training:  55%|█████▌    | 6138/11082 [19:43:42<15:54:24, 11.58s/step, acc=0.4992, loss=2.0603, lr=4.89e-05]
Training:  55%|█████▌    | 6139/11082 [19:43:53<15:40:51, 11.42s/step, acc=0.4992, loss=2.0603, lr=4.89e-05]
Training:  55%|█████▌    | 6140/11082 [19:44:05<15:57:15, 11.62s/step, acc=0.4992, loss=2.0603, lr=4.89e-05]
Training:  55%|█████▌    | 6141/11082 [19:44:16<15:43:11, 11.45s/step, acc=0.4992, loss=2.0603, lr=4.89e-05]
Training:  55%|█████▌    | 6142/11082 [19:44:28<15:51:06, 11.55s/step, acc=0.4992, loss=2.0603, lr=4.89e-05]
Training:  55%|█████▌    | 6143/11082 [19:44:39<15:46:24, 11.50s/step, acc=0.4992, loss=2.0603, lr=4.89e-05]
Training:  55%|█████▌    | 6144/11082 [19:44:51<15:53:05, 11.58s/step, acc=0.4992, loss=2.0603, lr=4.89e-05]
Training:  55%|█████▌    | 6145/11082 [19:45:03<15:48:11, 11.52s/step, acc=0.4992, loss=2.0603, lr=4.89e-05]
Training:  55%|█████▌    | 6146/11082 [19:45:14<15:52:26, 11.58s/step, acc=0.4992, loss=2.0603, lr=4.89e-05]
Training:  55%|█████▌    | 6147/11082 [19:45:26<15:47:12, 11.52s/step, acc=0.4992, loss=2.0603, lr=4.89e-05]
Training:  55%|█████▌    | 6148/11082 [19:45:37<15:46:59, 11.52s/step, acc=0.4992, loss=2.0603, lr=4.89e-05]
Training:  55%|█████▌    | 6149/11082 [19:45:49<15:45:19, 11.50s/step, acc=0.4992, loss=2.0603, lr=4.89e-05]
Training:  55%|█████▌    | 6150/11082 [19:46:00<15:51:48, 11.58s/step, acc=0.4992, loss=2.0603, lr=4.89e-05]
Training:  55%|█████▌    | 6150/11082 [19:46:07<15:51:48, 11.58s/step, acc=0.5004, loss=2.0526, lr=4.82e-05]2026-06-19 10:39:35 | INFO | __main__ | epoch=1 step=6150 train_loss=2.052562 train_acc=0.500385 lr=4.821e-05 grad_norm=0.8906 acc_d0=0.7642 acc_d1=0.6197 acc_d2=0.4986 acc_d3=0.4004 acc_d4=0.3214 acc_d5=0.2587 acc_d6=0.2101 acc_d7=0.1720

Training:  56%|█████▌    | 6151/11082 [19:46:11<15:38:16, 11.42s/step, acc=0.5004, loss=2.0526, lr=4.82e-05]
Training:  56%|█████▌    | 6152/11082 [19:46:23<15:48:38, 11.55s/step, acc=0.5004, loss=2.0526, lr=4.82e-05]
Training:  56%|█████▌    | 6153/11082 [19:46:35<15:54:13, 11.62s/step, acc=0.5004, loss=2.0526, lr=4.82e-05]
Training:  56%|█████▌    | 6154/11082 [19:46:47<15:50:47, 11.58s/step, acc=0.5004, loss=2.0526, lr=4.82e-05]
Training:  56%|█████▌    | 6155/11082 [19:46:58<15:49:45, 11.57s/step, acc=0.5004, loss=2.0526, lr=4.82e-05]
Training:  56%|█████▌    | 6156/11082 [19:47:10<15:59:48, 11.69s/step, acc=0.5004, loss=2.0526, lr=4.82e-05]
Training:  56%|█████▌    | 6157/11082 [19:47:21<15:50:39, 11.58s/step, acc=0.5004, loss=2.0526, lr=4.82e-05]
Training:  56%|█████▌    | 6158/11082 [19:47:33<15:41:16, 11.47s/step, acc=0.5004, loss=2.0526, lr=4.82e-05]
Training:  56%|█████▌    | 6159/11082 [19:47:44<15:48:46, 11.56s/step, acc=0.5004, loss=2.0526, lr=4.82e-05]
Training:  56%|█████▌    | 6160/11082 [19:47:56<15:42:21, 11.49s/step, acc=0.5004, loss=2.0526, lr=4.82e-05]
Training:  56%|█████▌    | 6161/11082 [19:48:08<15:49:51, 11.58s/step, acc=0.5004, loss=2.0526, lr=4.82e-05]
Training:  56%|█████▌    | 6162/11082 [19:48:19<15:50:06, 11.59s/step, acc=0.5004, loss=2.0526, lr=4.82e-05]
Training:  56%|█████▌    | 6163/11082 [19:48:31<15:45:49, 11.54s/step, acc=0.5004, loss=2.0526, lr=4.82e-05]
Training:  56%|█████▌    | 6164/11082 [19:48:43<15:57:54, 11.69s/step, acc=0.5004, loss=2.0526, lr=4.82e-05]
Training:  56%|█████▌    | 6165/11082 [19:48:54<15:51:22, 11.61s/step, acc=0.5004, loss=2.0526, lr=4.82e-05]
Training:  56%|█████▌    | 6166/11082 [19:49:05<15:45:14, 11.54s/step, acc=0.5004, loss=2.0526, lr=4.82e-05]
Training:  56%|█████▌    | 6167/11082 [19:49:17<15:50:09, 11.60s/step, acc=0.5004, loss=2.0526, lr=4.82e-05]
Training:  56%|█████▌    | 6168/11082 [19:49:29<15:54:08, 11.65s/step, acc=0.5004, loss=2.0526, lr=4.82e-05]
Training:  56%|█████▌    | 6169/11082 [19:49:40<15:47:23, 11.57s/step, acc=0.5004, loss=2.0526, lr=4.82e-05]
Training:  56%|█████▌    | 6170/11082 [19:49:52<15:45:41, 11.55s/step, acc=0.5004, loss=2.0526, lr=4.82e-05]
Training:  56%|█████▌    | 6171/11082 [19:50:04<15:50:05, 11.61s/step, acc=0.5004, loss=2.0526, lr=4.82e-05]
Training:  56%|█████▌    | 6172/11082 [19:50:15<15:38:58, 11.47s/step, acc=0.5004, loss=2.0526, lr=4.82e-05]
Training:  56%|█████▌    | 6173/11082 [19:50:26<15:36:44, 11.45s/step, acc=0.5004, loss=2.0526, lr=4.82e-05]
Training:  56%|█████▌    | 6174/11082 [19:50:38<15:52:41, 11.65s/step, acc=0.5004, loss=2.0526, lr=4.82e-05]
Training:  56%|█████▌    | 6175/11082 [19:50:50<15:47:45, 11.59s/step, acc=0.5004, loss=2.0526, lr=4.82e-05]
Training:  56%|█████▌    | 6176/11082 [19:51:01<15:42:30, 11.53s/step, acc=0.5004, loss=2.0526, lr=4.82e-05]
Training:  56%|█████▌    | 6177/11082 [19:51:13<15:40:54, 11.51s/step, acc=0.5004, loss=2.0526, lr=4.82e-05]
Training:  56%|█████▌    | 6178/11082 [19:51:24<15:46:34, 11.58s/step, acc=0.5004, loss=2.0526, lr=4.82e-05]
Training:  56%|█████▌    | 6179/11082 [19:51:36<15:41:16, 11.52s/step, acc=0.5004, loss=2.0526, lr=4.82e-05]
Training:  56%|█████▌    | 6180/11082 [19:51:47<15:30:42, 11.39s/step, acc=0.5004, loss=2.0526, lr=4.82e-05]
Training:  56%|█████▌    | 6181/11082 [19:51:58<15:30:51, 11.40s/step, acc=0.5004, loss=2.0526, lr=4.82e-05]
Training:  56%|█████▌    | 6182/11082 [19:52:10<15:44:15, 11.56s/step, acc=0.5004, loss=2.0526, lr=4.82e-05]
Training:  56%|█████▌    | 6183/11082 [19:52:21<15:37:58, 11.49s/step, acc=0.5004, loss=2.0526, lr=4.82e-05]
Training:  56%|█████▌    | 6184/11082 [19:52:33<15:37:25, 11.48s/step, acc=0.5004, loss=2.0526, lr=4.82e-05]
Training:  56%|█████▌    | 6185/11082 [19:52:44<15:39:33, 11.51s/step, acc=0.5004, loss=2.0526, lr=4.82e-05]
Training:  56%|█████▌    | 6186/11082 [19:52:56<15:45:55, 11.59s/step, acc=0.5004, loss=2.0526, lr=4.82e-05]
Training:  56%|█████▌    | 6187/11082 [19:53:08<15:43:54, 11.57s/step, acc=0.5004, loss=2.0526, lr=4.82e-05]
Training:  56%|█████▌    | 6188/11082 [19:53:19<15:40:57, 11.54s/step, acc=0.5004, loss=2.0526, lr=4.82e-05]
Training:  56%|█████▌    | 6189/11082 [19:53:30<15:33:08, 11.44s/step, acc=0.5004, loss=2.0526, lr=4.82e-05]
Training:  56%|█████▌    | 6190/11082 [19:53:43<15:51:13, 11.67s/step, acc=0.5004, loss=2.0526, lr=4.82e-05]
Training:  56%|█████▌    | 6191/11082 [19:53:54<15:45:58, 11.60s/step, acc=0.5004, loss=2.0526, lr=4.82e-05]
Training:  56%|█████▌    | 6192/11082 [19:54:05<15:30:57, 11.42s/step, acc=0.5004, loss=2.0526, lr=4.82e-05]
Training:  56%|█████▌    | 6193/11082 [19:54:17<15:51:59, 11.68s/step, acc=0.5004, loss=2.0526, lr=4.82e-05]
Training:  56%|█████▌    | 6194/11082 [19:54:29<15:46:25, 11.62s/step, acc=0.5004, loss=2.0526, lr=4.82e-05]
Training:  56%|█████▌    | 6195/11082 [19:54:40<15:46:39, 11.62s/step, acc=0.5004, loss=2.0526, lr=4.82e-05]
Training:  56%|█████▌    | 6196/11082 [19:54:52<15:49:29, 11.66s/step, acc=0.5004, loss=2.0526, lr=4.82e-05]
Training:  56%|█████▌    | 6197/11082 [19:55:04<15:52:55, 11.70s/step, acc=0.5004, loss=2.0526, lr=4.82e-05]
Training:  56%|█████▌    | 6198/11082 [19:55:16<15:48:41, 11.65s/step, acc=0.5004, loss=2.0526, lr=4.82e-05]
Training:  56%|█████▌    | 6199/11082 [19:55:27<15:40:24, 11.56s/step, acc=0.5004, loss=2.0526, lr=4.82e-05]
Training:  56%|█████▌    | 6200/11082 [19:55:38<15:41:47, 11.57s/step, acc=0.5004, loss=2.0526, lr=4.82e-05]
Training:  56%|█████▌    | 6200/11082 [19:55:46<15:41:47, 11.57s/step, acc=0.5004, loss=2.0542, lr=4.76e-05]2026-06-19 10:49:14 | INFO | __main__ | epoch=1 step=6200 train_loss=2.054209 train_acc=0.500401 lr=4.757e-05 grad_norm=0.9336 acc_d0=0.7588 acc_d1=0.6153 acc_d2=0.4935 acc_d3=0.3967 acc_d4=0.3190 acc_d5=0.2567 acc_d6=0.2107 acc_d7=0.1696

Training:  56%|█████▌    | 6201/11082 [19:55:50<15:45:51, 11.63s/step, acc=0.5004, loss=2.0542, lr=4.76e-05]
Training:  56%|█████▌    | 6202/11082 [19:56:02<15:40:15, 11.56s/step, acc=0.5004, loss=2.0542, lr=4.76e-05]
Training:  56%|█████▌    | 6203/11082 [19:56:14<15:55:08, 11.75s/step, acc=0.5004, loss=2.0542, lr=4.76e-05]
Training:  56%|█████▌    | 6204/11082 [19:56:25<15:46:51, 11.65s/step, acc=0.5004, loss=2.0542, lr=4.76e-05]
Training:  56%|█████▌    | 6205/11082 [19:56:37<15:50:26, 11.69s/step, acc=0.5004, loss=2.0542, lr=4.76e-05]
Training:  56%|█████▌    | 6206/11082 [19:56:49<15:51:49, 11.71s/step, acc=0.5004, loss=2.0542, lr=4.76e-05]
Training:  56%|█████▌    | 6207/11082 [19:57:00<15:44:32, 11.63s/step, acc=0.5004, loss=2.0542, lr=4.76e-05]
Training:  56%|█████▌    | 6208/11082 [19:57:12<15:51:10, 11.71s/step, acc=0.5004, loss=2.0542, lr=4.76e-05]
Training:  56%|█████▌    | 6209/11082 [19:57:24<15:47:08, 11.66s/step, acc=0.5004, loss=2.0542, lr=4.76e-05]
Training:  56%|█████▌    | 6210/11082 [19:57:35<15:49:16, 11.69s/step, acc=0.5004, loss=2.0542, lr=4.76e-05]
Training:  56%|█████▌    | 6211/11082 [19:57:47<15:41:02, 11.59s/step, acc=0.5004, loss=2.0542, lr=4.76e-05]
Training:  56%|█████▌    | 6212/11082 [19:57:58<15:36:07, 11.53s/step, acc=0.5004, loss=2.0542, lr=4.76e-05]
Training:  56%|█████▌    | 6213/11082 [19:58:10<15:32:11, 11.49s/step, acc=0.5004, loss=2.0542, lr=4.76e-05]
Training:  56%|█████▌    | 6214/11082 [19:58:21<15:40:23, 11.59s/step, acc=0.5004, loss=2.0542, lr=4.76e-05]
Training:  56%|█████▌    | 6215/11082 [19:58:33<15:32:26, 11.50s/step, acc=0.5004, loss=2.0542, lr=4.76e-05]
Training:  56%|█████▌    | 6216/11082 [19:58:45<15:40:34, 11.60s/step, acc=0.5004, loss=2.0542, lr=4.76e-05]
Training:  56%|█████▌    | 6217/11082 [19:58:56<15:39:27, 11.59s/step, acc=0.5004, loss=2.0542, lr=4.76e-05]
Training:  56%|█████▌    | 6218/11082 [19:59:07<15:26:14, 11.43s/step, acc=0.5004, loss=2.0542, lr=4.76e-05]
Training:  56%|█████▌    | 6219/11082 [19:59:19<15:40:03, 11.60s/step, acc=0.5004, loss=2.0542, lr=4.76e-05]
Training:  56%|█████▌    | 6220/11082 [19:59:30<15:29:21, 11.47s/step, acc=0.5004, loss=2.0542, lr=4.76e-05]
Training:  56%|█████▌    | 6221/11082 [19:59:42<15:44:37, 11.66s/step, acc=0.5004, loss=2.0542, lr=4.76e-05]
Training:  56%|█████▌    | 6222/11082 [19:59:54<15:44:32, 11.66s/step, acc=0.5004, loss=2.0542, lr=4.76e-05]
Training:  56%|█████▌    | 6223/11082 [20:00:05<15:30:31, 11.49s/step, acc=0.5004, loss=2.0542, lr=4.76e-05]
Training:  56%|█████▌    | 6224/11082 [20:00:17<15:38:48, 11.59s/step, acc=0.5004, loss=2.0542, lr=4.76e-05]
Training:  56%|█████▌    | 6225/11082 [20:00:28<15:33:15, 11.53s/step, acc=0.5004, loss=2.0542, lr=4.76e-05]
Training:  56%|█████▌    | 6226/11082 [20:00:40<15:38:16, 11.59s/step, acc=0.5004, loss=2.0542, lr=4.76e-05]
Training:  56%|█████▌    | 6227/11082 [20:00:52<15:37:43, 11.59s/step, acc=0.5004, loss=2.0542, lr=4.76e-05]
Training:  56%|█████▌    | 6228/11082 [20:01:03<15:26:52, 11.46s/step, acc=0.5004, loss=2.0542, lr=4.76e-05]
Training:  56%|█████▌    | 6229/11082 [20:01:15<15:37:43, 11.59s/step, acc=0.5004, loss=2.0542, lr=4.76e-05]
Training:  56%|█████▌    | 6230/11082 [20:01:26<15:39:32, 11.62s/step, acc=0.5004, loss=2.0542, lr=4.76e-05]
Training:  56%|█████▌    | 6231/11082 [20:01:38<15:42:33, 11.66s/step, acc=0.5004, loss=2.0542, lr=4.76e-05]
Training:  56%|█████▌    | 6232/11082 [20:01:50<15:38:14, 11.61s/step, acc=0.5004, loss=2.0542, lr=4.76e-05]
Training:  56%|█████▌    | 6233/11082 [20:02:01<15:35:02, 11.57s/step, acc=0.5004, loss=2.0542, lr=4.76e-05]
Training:  56%|█████▋    | 6234/11082 [20:02:13<15:32:55, 11.55s/step, acc=0.5004, loss=2.0542, lr=4.76e-05]
Training:  56%|█████▋    | 6235/11082 [20:02:24<15:33:32, 11.56s/step, acc=0.5004, loss=2.0542, lr=4.76e-05]
Training:  56%|█████▋    | 6236/11082 [20:02:35<15:26:36, 11.47s/step, acc=0.5004, loss=2.0542, lr=4.76e-05]
Training:  56%|█████▋    | 6237/11082 [20:02:47<15:26:07, 11.47s/step, acc=0.5004, loss=2.0542, lr=4.76e-05]
Training:  56%|█████▋    | 6238/11082 [20:02:59<15:37:14, 11.61s/step, acc=0.5004, loss=2.0542, lr=4.76e-05]
Training:  56%|█████▋    | 6239/11082 [20:03:10<15:35:41, 11.59s/step, acc=0.5004, loss=2.0542, lr=4.76e-05]
Training:  56%|█████▋    | 6240/11082 [20:03:22<15:37:59, 11.62s/step, acc=0.5004, loss=2.0542, lr=4.76e-05]
Training:  56%|█████▋    | 6241/11082 [20:03:34<15:33:55, 11.58s/step, acc=0.5004, loss=2.0542, lr=4.76e-05]
Training:  56%|█████▋    | 6242/11082 [20:03:45<15:35:03, 11.59s/step, acc=0.5004, loss=2.0542, lr=4.76e-05]
Training:  56%|█████▋    | 6243/11082 [20:03:57<15:31:42, 11.55s/step, acc=0.5004, loss=2.0542, lr=4.76e-05]
Training:  56%|█████▋    | 6244/11082 [20:04:09<15:41:58, 11.68s/step, acc=0.5004, loss=2.0542, lr=4.76e-05]
Training:  56%|█████▋    | 6245/11082 [20:04:20<15:33:46, 11.58s/step, acc=0.5004, loss=2.0542, lr=4.76e-05]
Training:  56%|█████▋    | 6246/11082 [20:04:31<15:30:21, 11.54s/step, acc=0.5004, loss=2.0542, lr=4.76e-05]
Training:  56%|█████▋    | 6247/11082 [20:04:43<15:29:07, 11.53s/step, acc=0.5004, loss=2.0542, lr=4.76e-05]
Training:  56%|█████▋    | 6248/11082 [20:04:54<15:25:05, 11.48s/step, acc=0.5004, loss=2.0542, lr=4.76e-05]
Training:  56%|█████▋    | 6249/11082 [20:05:06<15:27:07, 11.51s/step, acc=0.5004, loss=2.0542, lr=4.76e-05]
Training:  56%|█████▋    | 6250/11082 [20:05:18<15:33:14, 11.59s/step, acc=0.5004, loss=2.0542, lr=4.76e-05]
Training:  56%|█████▋    | 6250/11082 [20:05:25<15:33:14, 11.59s/step, acc=0.5011, loss=2.0486, lr=4.69e-05]2026-06-19 10:58:52 | INFO | __main__ | epoch=1 step=6250 train_loss=2.048631 train_acc=0.501143 lr=4.694e-05 grad_norm=0.7266 acc_d0=0.7535 acc_d1=0.6054 acc_d2=0.4841 acc_d3=0.3850 acc_d4=0.3071 acc_d5=0.2482 acc_d6=0.2005 acc_d7=0.1669

Training:  56%|█████▋    | 6251/11082 [20:05:29<15:32:02, 11.58s/step, acc=0.5011, loss=2.0486, lr=4.69e-05]
Training:  56%|█████▋    | 6252/11082 [20:05:41<15:33:05, 11.59s/step, acc=0.5011, loss=2.0486, lr=4.69e-05]
Training:  56%|█████▋    | 6253/11082 [20:05:52<15:27:49, 11.53s/step, acc=0.5011, loss=2.0486, lr=4.69e-05]
Training:  56%|█████▋    | 6254/11082 [20:06:04<15:24:12, 11.49s/step, acc=0.5011, loss=2.0486, lr=4.69e-05]
Training:  56%|█████▋    | 6255/11082 [20:06:15<15:24:09, 11.49s/step, acc=0.5011, loss=2.0486, lr=4.69e-05]
Training:  56%|█████▋    | 6256/11082 [20:06:27<15:29:14, 11.55s/step, acc=0.5011, loss=2.0486, lr=4.69e-05]
Training:  56%|█████▋    | 6257/11082 [20:06:38<15:28:40, 11.55s/step, acc=0.5011, loss=2.0486, lr=4.69e-05]
Training:  56%|█████▋    | 6258/11082 [20:06:50<15:22:43, 11.48s/step, acc=0.5011, loss=2.0486, lr=4.69e-05]
Training:  56%|█████▋    | 6259/11082 [20:07:01<15:26:18, 11.52s/step, acc=0.5011, loss=2.0486, lr=4.69e-05]
Training:  56%|█████▋    | 6260/11082 [20:07:13<15:23:53, 11.50s/step, acc=0.5011, loss=2.0486, lr=4.69e-05]
Training:  56%|█████▋    | 6261/11082 [20:07:25<15:33:16, 11.62s/step, acc=0.5011, loss=2.0486, lr=4.69e-05]
Training:  57%|█████▋    | 6262/11082 [20:07:36<15:29:42, 11.57s/step, acc=0.5011, loss=2.0486, lr=4.69e-05]
Training:  57%|█████▋    | 6263/11082 [20:07:48<15:29:09, 11.57s/step, acc=0.5011, loss=2.0486, lr=4.69e-05]
Training:  57%|█████▋    | 6264/11082 [20:07:59<15:24:51, 11.52s/step, acc=0.5011, loss=2.0486, lr=4.69e-05]
Training:  57%|█████▋    | 6265/11082 [20:08:11<15:24:49, 11.52s/step, acc=0.5011, loss=2.0486, lr=4.69e-05]
Training:  57%|█████▋    | 6266/11082 [20:08:22<15:21:39, 11.48s/step, acc=0.5011, loss=2.0486, lr=4.69e-05]
Training:  57%|█████▋    | 6267/11082 [20:08:34<15:36:37, 11.67s/step, acc=0.5011, loss=2.0486, lr=4.69e-05]
Training:  57%|█████▋    | 6268/11082 [20:08:46<15:32:33, 11.62s/step, acc=0.5011, loss=2.0486, lr=4.69e-05]
Training:  57%|█████▋    | 6269/11082 [20:08:57<15:27:15, 11.56s/step, acc=0.5011, loss=2.0486, lr=4.69e-05]
Training:  57%|█████▋    | 6270/11082 [20:09:08<15:15:58, 11.42s/step, acc=0.5011, loss=2.0486, lr=4.69e-05]
Training:  57%|█████▋    | 6271/11082 [20:09:20<15:33:05, 11.64s/step, acc=0.5011, loss=2.0486, lr=4.69e-05]
Training:  57%|█████▋    | 6272/11082 [20:09:32<15:23:42, 11.52s/step, acc=0.5011, loss=2.0486, lr=4.69e-05]
Training:  57%|█████▋    | 6273/11082 [20:09:43<15:25:17, 11.54s/step, acc=0.5011, loss=2.0486, lr=4.69e-05]
Training:  57%|█████▋    | 6274/11082 [20:09:55<15:23:17, 11.52s/step, acc=0.5011, loss=2.0486, lr=4.69e-05]
Training:  57%|█████▋    | 6275/11082 [20:10:06<15:24:09, 11.54s/step, acc=0.5011, loss=2.0486, lr=4.69e-05]
Training:  57%|█████▋    | 6276/11082 [20:10:18<15:19:47, 11.48s/step, acc=0.5011, loss=2.0486, lr=4.69e-05]
Training:  57%|█████▋    | 6277/11082 [20:10:29<15:31:28, 11.63s/step, acc=0.5011, loss=2.0486, lr=4.69e-05]
Training:  57%|█████▋    | 6278/11082 [20:10:41<15:32:21, 11.64s/step, acc=0.5011, loss=2.0486, lr=4.69e-05]
Training:  57%|█████▋    | 6279/11082 [20:10:52<15:24:34, 11.55s/step, acc=0.5011, loss=2.0486, lr=4.69e-05]
Training:  57%|█████▋    | 6280/11082 [20:11:04<15:34:43, 11.68s/step, acc=0.5011, loss=2.0486, lr=4.69e-05]
Training:  57%|█████▋    | 6281/11082 [20:11:16<15:22:23, 11.53s/step, acc=0.5011, loss=2.0486, lr=4.69e-05]
Training:  57%|█████▋    | 6282/11082 [20:11:28<15:32:49, 11.66s/step, acc=0.5011, loss=2.0486, lr=4.69e-05]
Training:  57%|█████▋    | 6283/11082 [20:11:39<15:32:03, 11.65s/step, acc=0.5011, loss=2.0486, lr=4.69e-05]
Training:  57%|█████▋    | 6284/11082 [20:11:51<15:34:14, 11.68s/step, acc=0.5011, loss=2.0486, lr=4.69e-05]
Training:  57%|█████▋    | 6285/11082 [20:12:02<15:26:46, 11.59s/step, acc=0.5011, loss=2.0486, lr=4.69e-05]
Training:  57%|█████▋    | 6286/11082 [20:12:14<15:22:51, 11.55s/step, acc=0.5011, loss=2.0486, lr=4.69e-05]
Training:  57%|█████▋    | 6287/11082 [20:12:25<15:20:31, 11.52s/step, acc=0.5011, loss=2.0486, lr=4.69e-05]
Training:  57%|█████▋    | 6288/11082 [20:12:37<15:17:51, 11.49s/step, acc=0.5011, loss=2.0486, lr=4.69e-05]
Training:  57%|█████▋    | 6289/11082 [20:12:48<15:20:59, 11.53s/step, acc=0.5011, loss=2.0486, lr=4.69e-05]
Training:  57%|█████▋    | 6290/11082 [20:13:00<15:31:01, 11.66s/step, acc=0.5011, loss=2.0486, lr=4.69e-05]
Training:  57%|█████▋    | 6291/11082 [20:13:12<15:24:39, 11.58s/step, acc=0.5011, loss=2.0486, lr=4.69e-05]
Training:  57%|█████▋    | 6292/11082 [20:13:24<15:30:13, 11.65s/step, acc=0.5011, loss=2.0486, lr=4.69e-05]
Training:  57%|█████▋    | 6293/11082 [20:13:35<15:20:36, 11.53s/step, acc=0.5011, loss=2.0486, lr=4.69e-05]
Training:  57%|█████▋    | 6294/11082 [20:13:46<15:18:21, 11.51s/step, acc=0.5011, loss=2.0486, lr=4.69e-05]
Training:  57%|█████▋    | 6295/11082 [20:13:58<15:18:52, 11.52s/step, acc=0.5011, loss=2.0486, lr=4.69e-05]
Training:  57%|█████▋    | 6296/11082 [20:14:10<15:26:09, 11.61s/step, acc=0.5011, loss=2.0486, lr=4.69e-05]
Training:  57%|█████▋    | 6297/11082 [20:14:21<15:25:51, 11.61s/step, acc=0.5011, loss=2.0486, lr=4.69e-05]
Training:  57%|█████▋    | 6298/11082 [20:14:33<15:20:01, 11.54s/step, acc=0.5011, loss=2.0486, lr=4.69e-05]
Training:  57%|█████▋    | 6299/11082 [20:14:44<15:21:44, 11.56s/step, acc=0.5011, loss=2.0486, lr=4.69e-05]
Training:  57%|█████▋    | 6300/11082 [20:14:56<15:29:16, 11.66s/step, acc=0.5011, loss=2.0486, lr=4.69e-05]
Training:  57%|█████▋    | 6300/11082 [20:15:03<15:29:16, 11.66s/step, acc=0.4998, loss=2.0567, lr=4.63e-05]2026-06-19 11:08:31 | INFO | __main__ | epoch=1 step=6300 train_loss=2.056704 train_acc=0.499778 lr=4.630e-05 grad_norm=0.7305 acc_d0=0.7684 acc_d1=0.6233 acc_d2=0.5027 acc_d3=0.4047 acc_d4=0.3239 acc_d5=0.2611 acc_d6=0.2107 acc_d7=0.1721

Training:  57%|█████▋    | 6301/11082 [20:15:08<15:25:01, 11.61s/step, acc=0.4998, loss=2.0567, lr=4.63e-05]
Training:  57%|█████▋    | 6302/11082 [20:15:19<15:17:35, 11.52s/step, acc=0.4998, loss=2.0567, lr=4.63e-05]
Training:  57%|█████▋    | 6303/11082 [20:15:30<15:20:19, 11.55s/step, acc=0.4998, loss=2.0567, lr=4.63e-05]
Training:  57%|█████▋    | 6304/11082 [20:15:42<15:20:25, 11.56s/step, acc=0.4998, loss=2.0567, lr=4.63e-05]
Training:  57%|█████▋    | 6305/11082 [20:15:54<15:21:22, 11.57s/step, acc=0.4998, loss=2.0567, lr=4.63e-05]
Training:  57%|█████▋    | 6306/11082 [20:16:05<15:23:38, 11.60s/step, acc=0.4998, loss=2.0567, lr=4.63e-05]
Training:  57%|█████▋    | 6307/11082 [20:16:17<15:21:19, 11.58s/step, acc=0.4998, loss=2.0567, lr=4.63e-05]
Training:  57%|█████▋    | 6308/11082 [20:16:28<15:10:02, 11.44s/step, acc=0.4998, loss=2.0567, lr=4.63e-05]
Training:  57%|█████▋    | 6309/11082 [20:16:40<15:15:08, 11.50s/step, acc=0.4998, loss=2.0567, lr=4.63e-05]
Training:  57%|█████▋    | 6310/11082 [20:16:51<15:20:02, 11.57s/step, acc=0.4998, loss=2.0567, lr=4.63e-05]
Training:  57%|█████▋    | 6311/11082 [20:17:03<15:25:46, 11.64s/step, acc=0.4998, loss=2.0567, lr=4.63e-05]
Training:  57%|█████▋    | 6312/11082 [20:17:15<15:20:03, 11.57s/step, acc=0.4998, loss=2.0567, lr=4.63e-05]
Training:  57%|█████▋    | 6313/11082 [20:17:26<15:23:18, 11.62s/step, acc=0.4998, loss=2.0567, lr=4.63e-05]
Training:  57%|█████▋    | 6314/11082 [20:17:38<15:22:30, 11.61s/step, acc=0.4998, loss=2.0567, lr=4.63e-05]
Training:  57%|█████▋    | 6315/11082 [20:17:50<15:27:29, 11.67s/step, acc=0.4998, loss=2.0567, lr=4.63e-05]
Training:  57%|█████▋    | 6316/11082 [20:18:01<15:21:01, 11.59s/step, acc=0.4998, loss=2.0567, lr=4.63e-05]
Training:  57%|█████▋    | 6317/11082 [20:18:13<15:26:07, 11.66s/step, acc=0.4998, loss=2.0567, lr=4.63e-05]
Training:  57%|█████▋    | 6318/11082 [20:18:24<15:19:47, 11.58s/step, acc=0.4998, loss=2.0567, lr=4.63e-05]
Training:  57%|█████▋    | 6319/11082 [20:18:36<15:13:32, 11.51s/step, acc=0.4998, loss=2.0567, lr=4.63e-05]
Training:  57%|█████▋    | 6320/11082 [20:18:48<15:22:40, 11.63s/step, acc=0.4998, loss=2.0567, lr=4.63e-05]
Training:  57%|█████▋    | 6321/11082 [20:18:59<15:24:43, 11.65s/step, acc=0.4998, loss=2.0567, lr=4.63e-05]
Training:  57%|█████▋    | 6322/11082 [20:19:11<15:22:36, 11.63s/step, acc=0.4998, loss=2.0567, lr=4.63e-05]
Training:  57%|█████▋    | 6323/11082 [20:19:22<15:07:29, 11.44s/step, acc=0.4998, loss=2.0567, lr=4.63e-05]
Training:  57%|█████▋    | 6324/11082 [20:19:33<15:11:01, 11.49s/step, acc=0.4998, loss=2.0567, lr=4.63e-05]
Training:  57%|█████▋    | 6325/11082 [20:19:45<15:21:33, 11.62s/step, acc=0.4998, loss=2.0567, lr=4.63e-05]
Training:  57%|█████▋    | 6326/11082 [20:19:57<15:15:06, 11.54s/step, acc=0.4998, loss=2.0567, lr=4.63e-05]
Training:  57%|█████▋    | 6327/11082 [20:20:08<15:15:54, 11.56s/step, acc=0.4998, loss=2.0567, lr=4.63e-05]
Training:  57%|█████▋    | 6328/11082 [20:20:20<15:16:16, 11.56s/step, acc=0.4998, loss=2.0567, lr=4.63e-05]
Training:  57%|█████▋    | 6329/11082 [20:20:31<15:13:35, 11.53s/step, acc=0.4998, loss=2.0567, lr=4.63e-05]
Training:  57%|█████▋    | 6330/11082 [20:20:43<15:08:43, 11.47s/step, acc=0.4998, loss=2.0567, lr=4.63e-05]
Training:  57%|█████▋    | 6331/11082 [20:20:54<15:08:12, 11.47s/step, acc=0.4998, loss=2.0567, lr=4.63e-05]
Training:  57%|█████▋    | 6332/11082 [20:21:06<15:15:04, 11.56s/step, acc=0.4998, loss=2.0567, lr=4.63e-05]
Training:  57%|█████▋    | 6333/11082 [20:21:17<15:14:13, 11.55s/step, acc=0.4998, loss=2.0567, lr=4.63e-05]
Training:  57%|█████▋    | 6334/11082 [20:21:29<15:12:24, 11.53s/step, acc=0.4998, loss=2.0567, lr=4.63e-05]
Training:  57%|█████▋    | 6335/11082 [20:21:40<15:09:55, 11.50s/step, acc=0.4998, loss=2.0567, lr=4.63e-05]
Training:  57%|█████▋    | 6336/11082 [20:21:52<15:14:28, 11.56s/step, acc=0.4998, loss=2.0567, lr=4.63e-05]
Training:  57%|█████▋    | 6337/11082 [20:22:04<15:11:53, 11.53s/step, acc=0.4998, loss=2.0567, lr=4.63e-05]
Training:  57%|█████▋    | 6338/11082 [20:22:15<15:17:28, 11.60s/step, acc=0.4998, loss=2.0567, lr=4.63e-05]
Training:  57%|█████▋    | 6339/11082 [20:22:27<15:24:11, 11.69s/step, acc=0.4998, loss=2.0567, lr=4.63e-05]
Training:  57%|█████▋    | 6340/11082 [20:22:38<15:06:06, 11.46s/step, acc=0.4998, loss=2.0567, lr=4.63e-05]
Training:  57%|█████▋    | 6341/11082 [20:22:50<15:11:54, 11.54s/step, acc=0.4998, loss=2.0567, lr=4.63e-05]
Training:  57%|█████▋    | 6342/11082 [20:23:01<15:13:11, 11.56s/step, acc=0.4998, loss=2.0567, lr=4.63e-05]
Training:  57%|█████▋    | 6343/11082 [20:23:13<15:05:08, 11.46s/step, acc=0.4998, loss=2.0567, lr=4.63e-05]
Training:  57%|█████▋    | 6344/11082 [20:23:24<15:06:48, 11.48s/step, acc=0.4998, loss=2.0567, lr=4.63e-05]
Training:  57%|█████▋    | 6345/11082 [20:23:36<15:11:47, 11.55s/step, acc=0.4998, loss=2.0567, lr=4.63e-05]
Training:  57%|█████▋    | 6346/11082 [20:23:48<15:14:57, 11.59s/step, acc=0.4998, loss=2.0567, lr=4.63e-05]
Training:  57%|█████▋    | 6347/11082 [20:23:59<15:08:08, 11.51s/step, acc=0.4998, loss=2.0567, lr=4.63e-05]
Training:  57%|█████▋    | 6348/11082 [20:24:11<15:15:41, 11.61s/step, acc=0.4998, loss=2.0567, lr=4.63e-05]
Training:  57%|█████▋    | 6349/11082 [20:24:22<15:10:42, 11.54s/step, acc=0.4998, loss=2.0567, lr=4.63e-05]
Training:  57%|█████▋    | 6350/11082 [20:24:34<15:09:22, 11.53s/step, acc=0.4998, loss=2.0567, lr=4.63e-05]
Training:  57%|█████▋    | 6350/11082 [20:24:41<15:09:22, 11.53s/step, acc=0.5009, loss=2.0507, lr=4.57e-05]2026-06-19 11:18:09 | INFO | __main__ | epoch=1 step=6350 train_loss=2.050748 train_acc=0.500941 lr=4.567e-05 grad_norm=0.7266 acc_d0=0.7556 acc_d1=0.6100 acc_d2=0.4873 acc_d3=0.3917 acc_d4=0.3155 acc_d5=0.2529 acc_d6=0.2048 acc_d7=0.1676

Training:  57%|█████▋    | 6351/11082 [20:24:46<15:18:14, 11.65s/step, acc=0.5009, loss=2.0507, lr=4.57e-05]
Training:  57%|█████▋    | 6352/11082 [20:24:57<15:14:12, 11.60s/step, acc=0.5009, loss=2.0507, lr=4.57e-05]
Training:  57%|█████▋    | 6353/11082 [20:25:09<15:09:42, 11.54s/step, acc=0.5009, loss=2.0507, lr=4.57e-05]
Training:  57%|█████▋    | 6354/11082 [20:25:20<15:16:50, 11.64s/step, acc=0.5009, loss=2.0507, lr=4.57e-05]
Training:  57%|█████▋    | 6355/11082 [20:25:32<15:14:19, 11.61s/step, acc=0.5009, loss=2.0507, lr=4.57e-05]
Training:  57%|█████▋    | 6356/11082 [20:25:43<15:12:53, 11.59s/step, acc=0.5009, loss=2.0507, lr=4.57e-05]
Training:  57%|█████▋    | 6357/11082 [20:25:55<15:13:57, 11.61s/step, acc=0.5009, loss=2.0507, lr=4.57e-05]
Training:  57%|█████▋    | 6358/11082 [20:26:06<15:07:58, 11.53s/step, acc=0.5009, loss=2.0507, lr=4.57e-05]
Training:  57%|█████▋    | 6359/11082 [20:26:18<15:04:28, 11.49s/step, acc=0.5009, loss=2.0507, lr=4.57e-05]
Training:  57%|█████▋    | 6360/11082 [20:26:30<15:09:48, 11.56s/step, acc=0.5009, loss=2.0507, lr=4.57e-05]
Training:  57%|█████▋    | 6361/11082 [20:26:41<15:11:33, 11.59s/step, acc=0.5009, loss=2.0507, lr=4.57e-05]
Training:  57%|█████▋    | 6362/11082 [20:26:53<15:14:07, 11.62s/step, acc=0.5009, loss=2.0507, lr=4.57e-05]
Training:  57%|█████▋    | 6363/11082 [20:27:05<15:16:17, 11.65s/step, acc=0.5009, loss=2.0507, lr=4.57e-05]
Training:  57%|█████▋    | 6364/11082 [20:27:17<15:22:38, 11.73s/step, acc=0.5009, loss=2.0507, lr=4.57e-05]
Training:  57%|█████▋    | 6365/11082 [20:27:28<15:15:54, 11.65s/step, acc=0.5009, loss=2.0507, lr=4.57e-05]
Training:  57%|█████▋    | 6366/11082 [20:27:40<15:11:50, 11.60s/step, acc=0.5009, loss=2.0507, lr=4.57e-05]
Training:  57%|█████▋    | 6367/11082 [20:27:51<15:12:31, 11.61s/step, acc=0.5009, loss=2.0507, lr=4.57e-05]
Training:  57%|█████▋    | 6368/11082 [20:28:02<15:04:20, 11.51s/step, acc=0.5009, loss=2.0507, lr=4.57e-05]
Training:  57%|█████▋    | 6369/11082 [20:28:14<15:05:21, 11.53s/step, acc=0.5009, loss=2.0507, lr=4.57e-05]
Training:  57%|█████▋    | 6370/11082 [20:28:26<15:09:07, 11.58s/step, acc=0.5009, loss=2.0507, lr=4.57e-05]
Training:  57%|█████▋    | 6371/11082 [20:28:37<15:07:45, 11.56s/step, acc=0.5009, loss=2.0507, lr=4.57e-05]
Training:  57%|█████▋    | 6372/11082 [20:28:49<15:09:11, 11.58s/step, acc=0.5009, loss=2.0507, lr=4.57e-05]
Training:  58%|█████▊    | 6373/11082 [20:29:01<15:13:49, 11.64s/step, acc=0.5009, loss=2.0507, lr=4.57e-05]
Training:  58%|█████▊    | 6374/11082 [20:29:12<15:15:10, 11.66s/step, acc=0.5009, loss=2.0507, lr=4.57e-05]
Training:  58%|█████▊    | 6375/11082 [20:29:24<15:07:34, 11.57s/step, acc=0.5009, loss=2.0507, lr=4.57e-05]
Training:  58%|█████▊    | 6376/11082 [20:29:35<15:04:14, 11.53s/step, acc=0.5009, loss=2.0507, lr=4.57e-05]
Training:  58%|█████▊    | 6377/11082 [20:29:47<15:08:38, 11.59s/step, acc=0.5009, loss=2.0507, lr=4.57e-05]
Training:  58%|█████▊    | 6378/11082 [20:29:59<15:10:36, 11.61s/step, acc=0.5009, loss=2.0507, lr=4.57e-05]
Training:  58%|█████▊    | 6379/11082 [20:30:10<15:05:40, 11.55s/step, acc=0.5009, loss=2.0507, lr=4.57e-05]
Training:  58%|█████▊    | 6380/11082 [20:30:22<15:06:28, 11.57s/step, acc=0.5009, loss=2.0507, lr=4.57e-05]
Training:  58%|█████▊    | 6381/11082 [20:30:33<15:03:54, 11.54s/step, acc=0.5009, loss=2.0507, lr=4.57e-05]
Training:  58%|█████▊    | 6382/11082 [20:30:45<15:08:04, 11.59s/step, acc=0.5009, loss=2.0507, lr=4.57e-05]
Training:  58%|█████▊    | 6383/11082 [20:30:56<15:04:23, 11.55s/step, acc=0.5009, loss=2.0507, lr=4.57e-05]
Training:  58%|█████▊    | 6384/11082 [20:31:08<15:06:35, 11.58s/step, acc=0.5009, loss=2.0507, lr=4.57e-05]
Training:  58%|█████▊    | 6385/11082 [20:31:19<15:08:39, 11.61s/step, acc=0.5009, loss=2.0507, lr=4.57e-05]
Training:  58%|█████▊    | 6386/11082 [20:31:31<15:05:44, 11.57s/step, acc=0.5009, loss=2.0507, lr=4.57e-05]
Training:  58%|█████▊    | 6387/11082 [20:31:42<15:00:40, 11.51s/step, acc=0.5009, loss=2.0507, lr=4.57e-05]
Training:  58%|█████▊    | 6388/11082 [20:31:54<15:15:08, 11.70s/step, acc=0.5009, loss=2.0507, lr=4.57e-05]
Training:  58%|█████▊    | 6389/11082 [20:32:06<15:12:02, 11.66s/step, acc=0.5009, loss=2.0507, lr=4.57e-05]
Training:  58%|█████▊    | 6390/11082 [20:32:18<15:07:58, 11.61s/step, acc=0.5009, loss=2.0507, lr=4.57e-05]
Training:  58%|█████▊    | 6391/11082 [20:32:29<15:05:14, 11.58s/step, acc=0.5009, loss=2.0507, lr=4.57e-05]
Training:  58%|█████▊    | 6392/11082 [20:32:40<15:00:51, 11.52s/step, acc=0.5009, loss=2.0507, lr=4.57e-05]
Training:  58%|█████▊    | 6393/11082 [20:32:52<14:52:28, 11.42s/step, acc=0.5009, loss=2.0507, lr=4.57e-05]
Training:  58%|█████▊    | 6394/11082 [20:33:04<15:07:10, 11.61s/step, acc=0.5009, loss=2.0507, lr=4.57e-05]
Training:  58%|█████▊    | 6395/11082 [20:33:15<14:57:47, 11.49s/step, acc=0.5009, loss=2.0507, lr=4.57e-05]
Training:  58%|█████▊    | 6396/11082 [20:33:27<15:05:59, 11.60s/step, acc=0.5009, loss=2.0507, lr=4.57e-05]
Training:  58%|█████▊    | 6397/11082 [20:33:38<15:06:01, 11.60s/step, acc=0.5009, loss=2.0507, lr=4.57e-05]
Training:  58%|█████▊    | 6398/11082 [20:33:50<15:07:27, 11.62s/step, acc=0.5009, loss=2.0507, lr=4.57e-05]
Training:  58%|█████▊    | 6399/11082 [20:34:02<15:09:01, 11.65s/step, acc=0.5009, loss=2.0507, lr=4.57e-05]
Training:  58%|█████▊    | 6400/11082 [20:34:13<15:05:20, 11.60s/step, acc=0.5009, loss=2.0507, lr=4.57e-05]
Training:  58%|█████▊    | 6400/11082 [20:34:20<15:05:20, 11.60s/step, acc=0.5009, loss=2.0551, lr=4.50e-05]2026-06-19 11:27:48 | INFO | __main__ | epoch=1 step=6400 train_loss=2.055142 train_acc=0.500930 lr=4.503e-05 grad_norm=0.8633 acc_d0=0.7575 acc_d1=0.6104 acc_d2=0.4880 acc_d3=0.3913 acc_d4=0.3131 acc_d5=0.2526 acc_d6=0.2048 acc_d7=0.1691

Training:  58%|█████▊    | 6401/11082 [20:34:25<15:05:36, 11.61s/step, acc=0.5009, loss=2.0551, lr=4.50e-05]
Training:  58%|█████▊    | 6402/11082 [20:34:36<14:55:32, 11.48s/step, acc=0.5009, loss=2.0551, lr=4.50e-05]
Training:  58%|█████▊    | 6403/11082 [20:34:48<14:56:15, 11.49s/step, acc=0.5009, loss=2.0551, lr=4.50e-05]
Training:  58%|█████▊    | 6404/11082 [20:34:59<14:58:45, 11.53s/step, acc=0.5009, loss=2.0551, lr=4.50e-05]
Training:  58%|█████▊    | 6405/11082 [20:35:11<15:01:01, 11.56s/step, acc=0.5009, loss=2.0551, lr=4.50e-05]
Training:  58%|█████▊    | 6406/11082 [20:35:22<14:55:15, 11.49s/step, acc=0.5009, loss=2.0551, lr=4.50e-05]
Training:  58%|█████▊    | 6407/11082 [20:35:33<14:50:28, 11.43s/step, acc=0.5009, loss=2.0551, lr=4.50e-05]
Training:  58%|█████▊    | 6408/11082 [20:35:45<14:50:52, 11.44s/step, acc=0.5009, loss=2.0551, lr=4.50e-05]
Training:  58%|█████▊    | 6409/11082 [20:35:57<14:59:06, 11.54s/step, acc=0.5009, loss=2.0551, lr=4.50e-05]
Training:  58%|█████▊    | 6410/11082 [20:36:08<14:59:32, 11.55s/step, acc=0.5009, loss=2.0551, lr=4.50e-05]
Training:  58%|█████▊    | 6411/11082 [20:36:20<15:13:49, 11.74s/step, acc=0.5009, loss=2.0551, lr=4.50e-05]
Training:  58%|█████▊    | 6412/11082 [20:36:32<15:12:41, 11.73s/step, acc=0.5009, loss=2.0551, lr=4.50e-05]
Training:  58%|█████▊    | 6413/11082 [20:36:43<15:04:48, 11.63s/step, acc=0.5009, loss=2.0551, lr=4.50e-05]
Training:  58%|█████▊    | 6414/11082 [20:36:55<14:58:39, 11.55s/step, acc=0.5009, loss=2.0551, lr=4.50e-05]
Training:  58%|█████▊    | 6415/11082 [20:37:07<15:02:29, 11.60s/step, acc=0.5009, loss=2.0551, lr=4.50e-05]
Training:  58%|█████▊    | 6416/11082 [20:37:18<14:55:36, 11.52s/step, acc=0.5009, loss=2.0551, lr=4.50e-05]
Training:  58%|█████▊    | 6417/11082 [20:37:30<14:59:35, 11.57s/step, acc=0.5009, loss=2.0551, lr=4.50e-05]
Training:  58%|█████▊    | 6418/11082 [20:37:41<14:56:23, 11.53s/step, acc=0.5009, loss=2.0551, lr=4.50e-05]
Training:  58%|█████▊    | 6419/11082 [20:37:52<14:51:33, 11.47s/step, acc=0.5009, loss=2.0551, lr=4.50e-05]
Training:  58%|█████▊    | 6420/11082 [20:38:04<14:55:44, 11.53s/step, acc=0.5009, loss=2.0551, lr=4.50e-05]
Training:  58%|█████▊    | 6421/11082 [20:38:16<14:55:13, 11.52s/step, acc=0.5009, loss=2.0551, lr=4.50e-05]
Training:  58%|█████▊    | 6422/11082 [20:38:27<14:59:50, 11.59s/step, acc=0.5009, loss=2.0551, lr=4.50e-05]
Training:  58%|█████▊    | 6423/11082 [20:38:39<14:54:51, 11.52s/step, acc=0.5009, loss=2.0551, lr=4.50e-05]
Training:  58%|█████▊    | 6424/11082 [20:38:50<14:57:08, 11.56s/step, acc=0.5009, loss=2.0551, lr=4.50e-05]
Training:  58%|█████▊    | 6425/11082 [20:39:02<14:51:59, 11.49s/step, acc=0.5009, loss=2.0551, lr=4.50e-05]
Training:  58%|█████▊    | 6426/11082 [20:39:13<14:53:38, 11.52s/step, acc=0.5009, loss=2.0551, lr=4.50e-05]
Training:  58%|█████▊    | 6427/11082 [20:39:25<14:54:22, 11.53s/step, acc=0.5009, loss=2.0551, lr=4.50e-05]
Training:  58%|█████▊    | 6428/11082 [20:39:36<14:47:22, 11.44s/step, acc=0.5009, loss=2.0551, lr=4.50e-05]
Training:  58%|█████▊    | 6429/11082 [20:39:48<14:51:56, 11.50s/step, acc=0.5009, loss=2.0551, lr=4.50e-05]
Training:  58%|█████▊    | 6430/11082 [20:39:59<14:55:53, 11.55s/step, acc=0.5009, loss=2.0551, lr=4.50e-05]
Training:  58%|█████▊    | 6431/11082 [20:40:11<14:50:15, 11.48s/step, acc=0.5009, loss=2.0551, lr=4.50e-05]
Training:  58%|█████▊    | 6432/11082 [20:40:22<14:41:13, 11.37s/step, acc=0.5009, loss=2.0551, lr=4.50e-05]
Training:  58%|█████▊    | 6433/11082 [20:40:34<14:55:11, 11.55s/step, acc=0.5009, loss=2.0551, lr=4.50e-05]
Training:  58%|█████▊    | 6434/11082 [20:40:45<14:56:00, 11.57s/step, acc=0.5009, loss=2.0551, lr=4.50e-05]
Training:  58%|█████▊    | 6435/11082 [20:40:57<15:00:50, 11.63s/step, acc=0.5009, loss=2.0551, lr=4.50e-05]
Training:  58%|█████▊    | 6436/11082 [20:41:09<15:00:38, 11.63s/step, acc=0.5009, loss=2.0551, lr=4.50e-05]
Training:  58%|█████▊    | 6437/11082 [20:41:20<14:50:39, 11.50s/step, acc=0.5009, loss=2.0551, lr=4.50e-05]
Training:  58%|█████▊    | 6438/11082 [20:41:32<15:01:12, 11.64s/step, acc=0.5009, loss=2.0551, lr=4.50e-05]
Training:  58%|█████▊    | 6439/11082 [20:41:43<14:50:01, 11.50s/step, acc=0.5009, loss=2.0551, lr=4.50e-05]
Training:  58%|█████▊    | 6440/11082 [20:41:55<14:57:11, 11.60s/step, acc=0.5009, loss=2.0551, lr=4.50e-05]
Training:  58%|█████▊    | 6441/11082 [20:42:07<15:03:24, 11.68s/step, acc=0.5009, loss=2.0551, lr=4.50e-05]
Training:  58%|█████▊    | 6442/11082 [20:42:18<15:00:03, 11.64s/step, acc=0.5009, loss=2.0551, lr=4.50e-05]
Training:  58%|█████▊    | 6443/11082 [20:42:30<15:03:38, 11.69s/step, acc=0.5009, loss=2.0551, lr=4.50e-05]
Training:  58%|█████▊    | 6444/11082 [20:42:41<14:56:19, 11.60s/step, acc=0.5009, loss=2.0551, lr=4.50e-05]
Training:  58%|█████▊    | 6445/11082 [20:42:53<14:57:52, 11.62s/step, acc=0.5009, loss=2.0551, lr=4.50e-05]
Training:  58%|█████▊    | 6446/11082 [20:43:05<14:58:10, 11.62s/step, acc=0.5009, loss=2.0551, lr=4.50e-05]
Training:  58%|█████▊    | 6447/11082 [20:43:17<15:09:19, 11.77s/step, acc=0.5009, loss=2.0551, lr=4.50e-05]
Training:  58%|█████▊    | 6448/11082 [20:43:28<14:56:30, 11.61s/step, acc=0.5009, loss=2.0551, lr=4.50e-05]
Training:  58%|█████▊    | 6449/11082 [20:43:40<14:56:57, 11.62s/step, acc=0.5009, loss=2.0551, lr=4.50e-05]
Training:  58%|█████▊    | 6450/11082 [20:43:52<14:59:39, 11.65s/step, acc=0.5009, loss=2.0551, lr=4.50e-05]
Training:  58%|█████▊    | 6450/11082 [20:43:58<14:59:39, 11.65s/step, acc=0.5021, loss=2.0433, lr=4.44e-05]2026-06-19 11:37:26 | INFO | __main__ | epoch=1 step=6450 train_loss=2.043274 train_acc=0.502113 lr=4.440e-05 grad_norm=0.7422 acc_d0=0.7619 acc_d1=0.6163 acc_d2=0.4963 acc_d3=0.3993 acc_d4=0.3216 acc_d5=0.2591 acc_d6=0.2117 acc_d7=0.1724

Training:  58%|█████▊    | 6451/11082 [20:44:03<14:49:58, 11.53s/step, acc=0.5021, loss=2.0433, lr=4.44e-05]
Training:  58%|█████▊    | 6452/11082 [20:44:14<14:54:21, 11.59s/step, acc=0.5021, loss=2.0433, lr=4.44e-05]
Training:  58%|█████▊    | 6453/11082 [20:44:26<14:55:33, 11.61s/step, acc=0.5021, loss=2.0433, lr=4.44e-05]
Training:  58%|█████▊    | 6454/11082 [20:44:38<14:56:18, 11.62s/step, acc=0.5021, loss=2.0433, lr=4.44e-05]
Training:  58%|█████▊    | 6455/11082 [20:44:49<14:49:55, 11.54s/step, acc=0.5021, loss=2.0433, lr=4.44e-05]
Training:  58%|█████▊    | 6456/11082 [20:45:01<14:52:27, 11.58s/step, acc=0.5021, loss=2.0433, lr=4.44e-05]
Training:  58%|█████▊    | 6457/11082 [20:45:13<14:57:49, 11.65s/step, acc=0.5021, loss=2.0433, lr=4.44e-05]
Training:  58%|█████▊    | 6458/11082 [20:45:24<14:54:31, 11.61s/step, acc=0.5021, loss=2.0433, lr=4.44e-05]
Training:  58%|█████▊    | 6459/11082 [20:45:35<14:41:09, 11.44s/step, acc=0.5021, loss=2.0433, lr=4.44e-05]
Training:  58%|█████▊    | 6460/11082 [20:45:47<14:51:28, 11.57s/step, acc=0.5021, loss=2.0433, lr=4.44e-05]
Training:  58%|█████▊    | 6461/11082 [20:45:59<14:56:06, 11.64s/step, acc=0.5021, loss=2.0433, lr=4.44e-05]
Training:  58%|█████▊    | 6462/11082 [20:46:10<14:52:42, 11.59s/step, acc=0.5021, loss=2.0433, lr=4.44e-05]
Training:  58%|█████▊    | 6463/11082 [20:46:22<14:53:09, 11.60s/step, acc=0.5021, loss=2.0433, lr=4.44e-05]
Training:  58%|█████▊    | 6464/11082 [20:46:34<14:52:51, 11.60s/step, acc=0.5021, loss=2.0433, lr=4.44e-05]
Training:  58%|█████▊    | 6465/11082 [20:46:45<14:52:28, 11.60s/step, acc=0.5021, loss=2.0433, lr=4.44e-05]
Training:  58%|█████▊    | 6466/11082 [20:46:56<14:41:34, 11.46s/step, acc=0.5021, loss=2.0433, lr=4.44e-05]
Training:  58%|█████▊    | 6467/11082 [20:47:08<14:47:22, 11.54s/step, acc=0.5021, loss=2.0433, lr=4.44e-05]
Training:  58%|█████▊    | 6468/11082 [20:47:19<14:43:31, 11.49s/step, acc=0.5021, loss=2.0433, lr=4.44e-05]
Training:  58%|█████▊    | 6469/11082 [20:47:31<14:51:08, 11.59s/step, acc=0.5021, loss=2.0433, lr=4.44e-05]
Training:  58%|█████▊    | 6470/11082 [20:47:43<14:46:49, 11.54s/step, acc=0.5021, loss=2.0433, lr=4.44e-05]
Training:  58%|█████▊    | 6471/11082 [20:47:54<14:43:09, 11.49s/step, acc=0.5021, loss=2.0433, lr=4.44e-05]
Training:  58%|█████▊    | 6472/11082 [20:48:06<14:46:46, 11.54s/step, acc=0.5021, loss=2.0433, lr=4.44e-05]
Training:  58%|█████▊    | 6473/11082 [20:48:17<14:45:18, 11.52s/step, acc=0.5021, loss=2.0433, lr=4.44e-05]
Training:  58%|█████▊    | 6474/11082 [20:48:28<14:33:25, 11.37s/step, acc=0.5021, loss=2.0433, lr=4.44e-05]
Training:  58%|█████▊    | 6475/11082 [20:48:40<14:47:00, 11.55s/step, acc=0.5021, loss=2.0433, lr=4.44e-05]
Training:  58%|█████▊    | 6476/11082 [20:48:52<14:50:42, 11.60s/step, acc=0.5021, loss=2.0433, lr=4.44e-05]
Training:  58%|█████▊    | 6477/11082 [20:49:04<14:52:03, 11.62s/step, acc=0.5021, loss=2.0433, lr=4.44e-05]
Training:  58%|█████▊    | 6478/11082 [20:49:14<14:36:45, 11.43s/step, acc=0.5021, loss=2.0433, lr=4.44e-05]
Training:  58%|█████▊    | 6479/11082 [20:49:27<14:56:00, 11.68s/step, acc=0.5021, loss=2.0433, lr=4.44e-05]
Training:  58%|█████▊    | 6480/11082 [20:49:38<14:50:42, 11.61s/step, acc=0.5021, loss=2.0433, lr=4.44e-05]
Training:  58%|█████▊    | 6481/11082 [20:49:50<14:51:05, 11.62s/step, acc=0.5021, loss=2.0433, lr=4.44e-05]
Training:  58%|█████▊    | 6482/11082 [20:50:01<14:48:52, 11.59s/step, acc=0.5021, loss=2.0433, lr=4.44e-05]
Training:  59%|█████▊    | 6483/11082 [20:50:13<14:48:46, 11.60s/step, acc=0.5021, loss=2.0433, lr=4.44e-05]
Training:  59%|█████▊    | 6484/11082 [20:50:24<14:39:10, 11.47s/step, acc=0.5021, loss=2.0433, lr=4.44e-05]
Training:  59%|█████▊    | 6485/11082 [20:50:36<14:57:58, 11.72s/step, acc=0.5021, loss=2.0433, lr=4.44e-05]
Training:  59%|█████▊    | 6486/11082 [20:50:48<14:47:21, 11.58s/step, acc=0.5021, loss=2.0433, lr=4.44e-05]
Training:  59%|█████▊    | 6487/11082 [20:51:00<14:53:58, 11.67s/step, acc=0.5021, loss=2.0433, lr=4.44e-05]
Training:  59%|█████▊    | 6488/11082 [20:51:11<14:48:01, 11.60s/step, acc=0.5021, loss=2.0433, lr=4.44e-05]
Training:  59%|█████▊    | 6489/11082 [20:51:23<14:53:30, 11.67s/step, acc=0.5021, loss=2.0433, lr=4.44e-05]
Training:  59%|█████▊    | 6490/11082 [20:51:34<14:41:04, 11.51s/step, acc=0.5021, loss=2.0433, lr=4.44e-05]
Training:  59%|█████▊    | 6491/11082 [20:51:46<14:47:25, 11.60s/step, acc=0.5021, loss=2.0433, lr=4.44e-05]
Training:  59%|█████▊    | 6492/11082 [20:51:57<14:35:58, 11.45s/step, acc=0.5021, loss=2.0433, lr=4.44e-05]
Training:  59%|█████▊    | 6493/11082 [20:52:09<14:46:15, 11.59s/step, acc=0.5021, loss=2.0433, lr=4.44e-05]
Training:  59%|█████▊    | 6494/11082 [20:52:20<14:34:08, 11.43s/step, acc=0.5021, loss=2.0433, lr=4.44e-05]
Training:  59%|█████▊    | 6495/11082 [20:52:32<14:43:32, 11.56s/step, acc=0.5021, loss=2.0433, lr=4.44e-05]
Training:  59%|█████▊    | 6496/11082 [20:52:43<14:46:46, 11.60s/step, acc=0.5021, loss=2.0433, lr=4.44e-05]
Training:  59%|█████▊    | 6497/11082 [20:52:55<14:44:58, 11.58s/step, acc=0.5021, loss=2.0433, lr=4.44e-05]
Training:  59%|█████▊    | 6498/11082 [20:53:07<14:55:20, 11.72s/step, acc=0.5021, loss=2.0433, lr=4.44e-05]
Training:  59%|█████▊    | 6499/11082 [20:53:18<14:47:56, 11.62s/step, acc=0.5021, loss=2.0433, lr=4.44e-05]
Training:  59%|█████▊    | 6500/11082 [20:53:29<14:33:46, 11.44s/step, acc=0.5021, loss=2.0433, lr=4.44e-05]
Training:  59%|█████▊    | 6500/11082 [20:53:37<14:33:46, 11.44s/step, acc=0.5002, loss=2.0540, lr=4.38e-05]2026-06-19 11:47:05 | INFO | __main__ | epoch=1 step=6500 train_loss=2.054017 train_acc=0.500217 lr=4.377e-05 grad_norm=0.6758 acc_d0=0.7579 acc_d1=0.6129 acc_d2=0.4912 acc_d3=0.3921 acc_d4=0.3146 acc_d5=0.2536 acc_d6=0.2065 acc_d7=0.1708

Training:  59%|█████▊    | 6501/11082 [20:53:41<14:43:08, 11.57s/step, acc=0.5002, loss=2.0540, lr=4.38e-05]
Training:  59%|█████▊    | 6502/11082 [20:53:53<14:47:49, 11.63s/step, acc=0.5002, loss=2.0540, lr=4.38e-05]
Training:  59%|█████▊    | 6503/11082 [20:54:04<14:31:23, 11.42s/step, acc=0.5002, loss=2.0540, lr=4.38e-05]
Training:  59%|█████▊    | 6504/11082 [20:54:16<14:33:22, 11.45s/step, acc=0.5002, loss=2.0540, lr=4.38e-05]
Training:  59%|█████▊    | 6505/11082 [20:54:27<14:43:39, 11.58s/step, acc=0.5002, loss=2.0540, lr=4.38e-05]
Training:  59%|█████▊    | 6506/11082 [20:54:39<14:37:54, 11.51s/step, acc=0.5002, loss=2.0540, lr=4.38e-05]
Training:  59%|█████▊    | 6507/11082 [20:54:50<14:41:42, 11.56s/step, acc=0.5002, loss=2.0540, lr=4.38e-05]
Training:  59%|█████▊    | 6508/11082 [20:55:02<14:34:00, 11.47s/step, acc=0.5002, loss=2.0540, lr=4.38e-05]
Training:  59%|█████▊    | 6509/11082 [20:55:13<14:28:46, 11.40s/step, acc=0.5002, loss=2.0540, lr=4.38e-05]
Training:  59%|█████▊    | 6510/11082 [20:55:25<14:51:44, 11.70s/step, acc=0.5002, loss=2.0540, lr=4.38e-05]
Training:  59%|█████▉    | 6511/11082 [20:55:37<14:47:32, 11.65s/step, acc=0.5002, loss=2.0540, lr=4.38e-05]
Training:  59%|█████▉    | 6512/11082 [20:55:48<14:44:09, 11.61s/step, acc=0.5002, loss=2.0540, lr=4.38e-05]
Training:  59%|█████▉    | 6513/11082 [20:55:59<14:32:12, 11.45s/step, acc=0.5002, loss=2.0540, lr=4.38e-05]
Training:  59%|█████▉    | 6514/11082 [20:56:11<14:32:32, 11.46s/step, acc=0.5002, loss=2.0540, lr=4.38e-05]
Training:  59%|█████▉    | 6515/11082 [20:56:23<14:42:47, 11.60s/step, acc=0.5002, loss=2.0540, lr=4.38e-05]
Training:  59%|█████▉    | 6516/11082 [20:56:34<14:37:32, 11.53s/step, acc=0.5002, loss=2.0540, lr=4.38e-05]
Training:  59%|█████▉    | 6517/11082 [20:56:46<14:42:36, 11.60s/step, acc=0.5002, loss=2.0540, lr=4.38e-05]
Training:  59%|█████▉    | 6518/11082 [20:56:58<14:51:51, 11.72s/step, acc=0.5002, loss=2.0540, lr=4.38e-05]
Training:  59%|█████▉    | 6519/11082 [20:57:09<14:40:39, 11.58s/step, acc=0.5002, loss=2.0540, lr=4.38e-05]
Training:  59%|█████▉    | 6520/11082 [20:57:21<14:41:53, 11.60s/step, acc=0.5002, loss=2.0540, lr=4.38e-05]
Training:  59%|█████▉    | 6521/11082 [20:57:33<14:43:17, 11.62s/step, acc=0.5002, loss=2.0540, lr=4.38e-05]
Training:  59%|█████▉    | 6522/11082 [20:57:44<14:39:39, 11.57s/step, acc=0.5002, loss=2.0540, lr=4.38e-05]
Training:  59%|█████▉    | 6523/11082 [20:57:56<14:36:43, 11.54s/step, acc=0.5002, loss=2.0540, lr=4.38e-05]
Training:  59%|█████▉    | 6524/11082 [20:58:07<14:39:39, 11.58s/step, acc=0.5002, loss=2.0540, lr=4.38e-05]
Training:  59%|█████▉    | 6525/11082 [20:58:18<14:31:20, 11.47s/step, acc=0.5002, loss=2.0540, lr=4.38e-05]
Training:  59%|█████▉    | 6526/11082 [20:58:30<14:30:13, 11.46s/step, acc=0.5002, loss=2.0540, lr=4.38e-05]
Training:  59%|█████▉    | 6527/11082 [20:58:41<14:34:11, 11.52s/step, acc=0.5002, loss=2.0540, lr=4.38e-05]
Training:  59%|█████▉    | 6528/11082 [20:58:53<14:43:18, 11.64s/step, acc=0.5002, loss=2.0540, lr=4.38e-05]
Training:  59%|█████▉    | 6529/11082 [20:59:05<14:48:48, 11.71s/step, acc=0.5002, loss=2.0540, lr=4.38e-05]
Training:  59%|█████▉    | 6530/11082 [20:59:17<14:42:22, 11.63s/step, acc=0.5002, loss=2.0540, lr=4.38e-05]
Training:  59%|█████▉    | 6531/11082 [20:59:28<14:44:01, 11.65s/step, acc=0.5002, loss=2.0540, lr=4.38e-05]
Training:  59%|█████▉    | 6532/11082 [20:59:40<14:41:01, 11.62s/step, acc=0.5002, loss=2.0540, lr=4.38e-05]
Training:  59%|█████▉    | 6533/11082 [20:59:52<14:46:33, 11.69s/step, acc=0.5002, loss=2.0540, lr=4.38e-05]
Training:  59%|█████▉    | 6534/11082 [21:00:03<14:34:16, 11.53s/step, acc=0.5002, loss=2.0540, lr=4.38e-05]
Training:  59%|█████▉    | 6535/11082 [21:00:15<14:49:12, 11.73s/step, acc=0.5002, loss=2.0540, lr=4.38e-05]
Training:  59%|█████▉    | 6536/11082 [21:00:26<14:36:13, 11.56s/step, acc=0.5002, loss=2.0540, lr=4.38e-05]
Training:  59%|█████▉    | 6537/11082 [21:00:38<14:43:10, 11.66s/step, acc=0.5002, loss=2.0540, lr=4.38e-05]
Training:  59%|█████▉    | 6538/11082 [21:00:50<14:44:43, 11.68s/step, acc=0.5002, loss=2.0540, lr=4.38e-05]
Training:  59%|█████▉    | 6539/11082 [21:01:01<14:38:35, 11.60s/step, acc=0.5002, loss=2.0540, lr=4.38e-05]
Training:  59%|█████▉    | 6540/11082 [21:01:13<14:32:52, 11.53s/step, acc=0.5002, loss=2.0540, lr=4.38e-05]
Training:  59%|█████▉    | 6541/11082 [21:01:24<14:35:13, 11.56s/step, acc=0.5002, loss=2.0540, lr=4.38e-05]
Training:  59%|█████▉    | 6542/11082 [21:01:36<14:44:02, 11.68s/step, acc=0.5002, loss=2.0540, lr=4.38e-05]
Training:  59%|█████▉    | 6543/11082 [21:01:48<14:34:07, 11.55s/step, acc=0.5002, loss=2.0540, lr=4.38e-05]
Training:  59%|█████▉    | 6544/11082 [21:01:59<14:39:38, 11.63s/step, acc=0.5002, loss=2.0540, lr=4.38e-05]
Training:  59%|█████▉    | 6545/11082 [21:02:11<14:39:19, 11.63s/step, acc=0.5002, loss=2.0540, lr=4.38e-05]
Training:  59%|█████▉    | 6546/11082 [21:02:23<14:41:41, 11.66s/step, acc=0.5002, loss=2.0540, lr=4.38e-05]
Training:  59%|█████▉    | 6547/11082 [21:02:35<14:45:56, 11.72s/step, acc=0.5002, loss=2.0540, lr=4.38e-05]
Training:  59%|█████▉    | 6548/11082 [21:02:46<14:36:05, 11.59s/step, acc=0.5002, loss=2.0540, lr=4.38e-05]
Training:  59%|█████▉    | 6549/11082 [21:02:58<14:37:25, 11.61s/step, acc=0.5002, loss=2.0540, lr=4.38e-05]
Training:  59%|█████▉    | 6550/11082 [21:03:09<14:31:56, 11.54s/step, acc=0.5002, loss=2.0540, lr=4.38e-05]
Training:  59%|█████▉    | 6550/11082 [21:03:16<14:31:56, 11.54s/step, acc=0.5010, loss=2.0489, lr=4.31e-05]2026-06-19 11:56:44 | INFO | __main__ | epoch=1 step=6550 train_loss=2.048941 train_acc=0.500957 lr=4.315e-05 grad_norm=0.7617 acc_d0=0.7621 acc_d1=0.6182 acc_d2=0.4971 acc_d3=0.3991 acc_d4=0.3200 acc_d5=0.2582 acc_d6=0.2100 acc_d7=0.1716

Training:  59%|█████▉    | 6551/11082 [21:03:21<14:33:28, 11.57s/step, acc=0.5010, loss=2.0489, lr=4.31e-05]
Training:  59%|█████▉    | 6552/11082 [21:03:32<14:37:34, 11.62s/step, acc=0.5010, loss=2.0489, lr=4.31e-05]
Training:  59%|█████▉    | 6553/11082 [21:03:44<14:32:42, 11.56s/step, acc=0.5010, loss=2.0489, lr=4.31e-05]
Training:  59%|█████▉    | 6554/11082 [21:03:55<14:31:56, 11.55s/step, acc=0.5010, loss=2.0489, lr=4.31e-05]
Training:  59%|█████▉    | 6555/11082 [21:04:06<14:20:36, 11.41s/step, acc=0.5010, loss=2.0489, lr=4.31e-05]
Training:  59%|█████▉    | 6556/11082 [21:04:18<14:28:46, 11.52s/step, acc=0.5010, loss=2.0489, lr=4.31e-05]
Training:  59%|█████▉    | 6557/11082 [21:04:30<14:34:34, 11.60s/step, acc=0.5010, loss=2.0489, lr=4.31e-05]
Training:  59%|█████▉    | 6558/11082 [21:04:42<14:38:58, 11.66s/step, acc=0.5010, loss=2.0489, lr=4.31e-05]
Training:  59%|█████▉    | 6559/11082 [21:04:53<14:40:09, 11.68s/step, acc=0.5010, loss=2.0489, lr=4.31e-05]
Training:  59%|█████▉    | 6560/11082 [21:05:05<14:36:48, 11.63s/step, acc=0.5010, loss=2.0489, lr=4.31e-05]
Training:  59%|█████▉    | 6561/11082 [21:05:16<14:32:46, 11.58s/step, acc=0.5010, loss=2.0489, lr=4.31e-05]
Training:  59%|█████▉    | 6562/11082 [21:05:28<14:27:53, 11.52s/step, acc=0.5010, loss=2.0489, lr=4.31e-05]
Training:  59%|█████▉    | 6563/11082 [21:05:39<14:28:53, 11.54s/step, acc=0.5010, loss=2.0489, lr=4.31e-05]
Training:  59%|█████▉    | 6564/11082 [21:05:51<14:29:55, 11.55s/step, acc=0.5010, loss=2.0489, lr=4.31e-05]
Training:  59%|█████▉    | 6565/11082 [21:06:02<14:24:50, 11.49s/step, acc=0.5010, loss=2.0489, lr=4.31e-05]
Training:  59%|█████▉    | 6566/11082 [21:06:14<14:37:18, 11.66s/step, acc=0.5010, loss=2.0489, lr=4.31e-05]
Training:  59%|█████▉    | 6567/11082 [21:06:26<14:35:29, 11.63s/step, acc=0.5010, loss=2.0489, lr=4.31e-05]
Training:  59%|█████▉    | 6568/11082 [21:06:37<14:27:38, 11.53s/step, acc=0.5010, loss=2.0489, lr=4.31e-05]
Training:  59%|█████▉    | 6569/11082 [21:06:49<14:32:18, 11.60s/step, acc=0.5010, loss=2.0489, lr=4.31e-05]
Training:  59%|█████▉    | 6570/11082 [21:07:00<14:29:17, 11.56s/step, acc=0.5010, loss=2.0489, lr=4.31e-05]
Training:  59%|█████▉    | 6571/11082 [21:07:12<14:27:35, 11.54s/step, acc=0.5010, loss=2.0489, lr=4.31e-05]
Training:  59%|█████▉    | 6572/11082 [21:07:24<14:40:55, 11.72s/step, acc=0.5010, loss=2.0489, lr=4.31e-05]
Training:  59%|█████▉    | 6573/11082 [21:07:35<14:24:35, 11.50s/step, acc=0.5010, loss=2.0489, lr=4.31e-05]
Training:  59%|█████▉    | 6574/11082 [21:07:47<14:31:27, 11.60s/step, acc=0.5010, loss=2.0489, lr=4.31e-05]
Training:  59%|█████▉    | 6575/11082 [21:07:58<14:24:08, 11.50s/step, acc=0.5010, loss=2.0489, lr=4.31e-05]
Training:  59%|█████▉    | 6576/11082 [21:08:10<14:32:37, 11.62s/step, acc=0.5010, loss=2.0489, lr=4.31e-05]
Training:  59%|█████▉    | 6577/11082 [21:08:22<14:28:35, 11.57s/step, acc=0.5010, loss=2.0489, lr=4.31e-05]
Training:  59%|█████▉    | 6578/11082 [21:08:33<14:18:11, 11.43s/step, acc=0.5010, loss=2.0489, lr=4.31e-05]
Training:  59%|█████▉    | 6579/11082 [21:08:44<14:19:38, 11.45s/step, acc=0.5010, loss=2.0489, lr=4.31e-05]
Training:  59%|█████▉    | 6580/11082 [21:08:56<14:32:19, 11.63s/step, acc=0.5010, loss=2.0489, lr=4.31e-05]
Training:  59%|█████▉    | 6581/11082 [21:09:07<14:22:45, 11.50s/step, acc=0.5010, loss=2.0489, lr=4.31e-05]
Training:  59%|█████▉    | 6582/11082 [21:09:19<14:28:13, 11.58s/step, acc=0.5010, loss=2.0489, lr=4.31e-05]
Training:  59%|█████▉    | 6583/11082 [21:09:31<14:29:49, 11.60s/step, acc=0.5010, loss=2.0489, lr=4.31e-05]
Training:  59%|█████▉    | 6584/11082 [21:09:42<14:20:06, 11.47s/step, acc=0.5010, loss=2.0489, lr=4.31e-05]
Training:  59%|█████▉    | 6585/11082 [21:09:54<14:21:49, 11.50s/step, acc=0.5010, loss=2.0489, lr=4.31e-05]
Training:  59%|█████▉    | 6586/11082 [21:10:05<14:21:46, 11.50s/step, acc=0.5010, loss=2.0489, lr=4.31e-05]
Training:  59%|█████▉    | 6587/11082 [21:10:16<14:19:22, 11.47s/step, acc=0.5010, loss=2.0489, lr=4.31e-05]
Training:  59%|█████▉    | 6588/11082 [21:10:28<14:22:51, 11.52s/step, acc=0.5010, loss=2.0489, lr=4.31e-05]
Training:  59%|█████▉    | 6589/11082 [21:10:40<14:27:24, 11.58s/step, acc=0.5010, loss=2.0489, lr=4.31e-05]
Training:  59%|█████▉    | 6590/11082 [21:10:51<14:23:44, 11.54s/step, acc=0.5010, loss=2.0489, lr=4.31e-05]
Training:  59%|█████▉    | 6591/11082 [21:11:03<14:33:32, 11.67s/step, acc=0.5010, loss=2.0489, lr=4.31e-05]
Training:  59%|█████▉    | 6592/11082 [21:11:14<14:23:33, 11.54s/step, acc=0.5010, loss=2.0489, lr=4.31e-05]
Training:  59%|█████▉    | 6593/11082 [21:11:26<14:27:28, 11.59s/step, acc=0.5010, loss=2.0489, lr=4.31e-05]
Training:  60%|█████▉    | 6594/11082 [21:11:38<14:23:33, 11.54s/step, acc=0.5010, loss=2.0489, lr=4.31e-05]
Training:  60%|█████▉    | 6595/11082 [21:11:49<14:21:19, 11.52s/step, acc=0.5010, loss=2.0489, lr=4.31e-05]
Training:  60%|█████▉    | 6596/11082 [21:12:01<14:24:18, 11.56s/step, acc=0.5010, loss=2.0489, lr=4.31e-05]
Training:  60%|█████▉    | 6597/11082 [21:12:12<14:21:52, 11.53s/step, acc=0.5010, loss=2.0489, lr=4.31e-05]
Training:  60%|█████▉    | 6598/11082 [21:12:24<14:24:38, 11.57s/step, acc=0.5010, loss=2.0489, lr=4.31e-05]
Training:  60%|█████▉    | 6599/11082 [21:12:35<14:25:03, 11.58s/step, acc=0.5010, loss=2.0489, lr=4.31e-05]
Training:  60%|█████▉    | 6600/11082 [21:12:47<14:21:52, 11.54s/step, acc=0.5010, loss=2.0489, lr=4.31e-05]
Training:  60%|█████▉    | 6600/11082 [21:12:54<14:21:52, 11.54s/step, acc=0.5038, loss=2.0349, lr=4.25e-05]2026-06-19 12:06:22 | INFO | __main__ | epoch=1 step=6600 train_loss=2.034856 train_acc=0.503822 lr=4.253e-05 grad_norm=0.8008 acc_d0=0.7659 acc_d1=0.6227 acc_d2=0.5016 acc_d3=0.4029 acc_d4=0.3248 acc_d5=0.2610 acc_d6=0.2125 acc_d7=0.1729

Training:  60%|█████▉    | 6601/11082 [21:12:58<14:18:45, 11.50s/step, acc=0.5038, loss=2.0349, lr=4.25e-05]
Training:  60%|█████▉    | 6602/11082 [21:13:10<14:14:29, 11.44s/step, acc=0.5038, loss=2.0349, lr=4.25e-05]
Training:  60%|█████▉    | 6603/11082 [21:13:21<14:18:03, 11.49s/step, acc=0.5038, loss=2.0349, lr=4.25e-05]
Training:  60%|█████▉    | 6604/11082 [21:13:33<14:24:04, 11.58s/step, acc=0.5038, loss=2.0349, lr=4.25e-05]
Training:  60%|█████▉    | 6605/11082 [21:13:45<14:27:48, 11.63s/step, acc=0.5038, loss=2.0349, lr=4.25e-05]
Training:  60%|█████▉    | 6606/11082 [21:13:56<14:28:24, 11.64s/step, acc=0.5038, loss=2.0349, lr=4.25e-05]
Training:  60%|█████▉    | 6607/11082 [21:14:08<14:36:33, 11.75s/step, acc=0.5038, loss=2.0349, lr=4.25e-05]
Training:  60%|█████▉    | 6608/11082 [21:14:20<14:35:44, 11.74s/step, acc=0.5038, loss=2.0349, lr=4.25e-05]
Training:  60%|█████▉    | 6609/11082 [21:14:31<14:22:30, 11.57s/step, acc=0.5038, loss=2.0349, lr=4.25e-05]
Training:  60%|█████▉    | 6610/11082 [21:14:43<14:27:23, 11.64s/step, acc=0.5038, loss=2.0349, lr=4.25e-05]
Training:  60%|█████▉    | 6611/11082 [21:14:55<14:25:25, 11.61s/step, acc=0.5038, loss=2.0349, lr=4.25e-05]
Training:  60%|█████▉    | 6612/11082 [21:15:06<14:26:56, 11.64s/step, acc=0.5038, loss=2.0349, lr=4.25e-05]
Training:  60%|█████▉    | 6613/11082 [21:15:18<14:18:44, 11.53s/step, acc=0.5038, loss=2.0349, lr=4.25e-05]
Training:  60%|█████▉    | 6614/11082 [21:15:29<14:22:02, 11.58s/step, acc=0.5038, loss=2.0349, lr=4.25e-05]
Training:  60%|█████▉    | 6615/11082 [21:15:41<14:13:45, 11.47s/step, acc=0.5038, loss=2.0349, lr=4.25e-05]
Training:  60%|█████▉    | 6616/11082 [21:15:52<14:19:47, 11.55s/step, acc=0.5038, loss=2.0349, lr=4.25e-05]
Training:  60%|█████▉    | 6617/11082 [21:16:04<14:27:03, 11.65s/step, acc=0.5038, loss=2.0349, lr=4.25e-05]
Training:  60%|█████▉    | 6618/11082 [21:16:15<14:17:59, 11.53s/step, acc=0.5038, loss=2.0349, lr=4.25e-05]
Training:  60%|█████▉    | 6619/11082 [21:16:27<14:15:44, 11.50s/step, acc=0.5038, loss=2.0349, lr=4.25e-05]
Training:  60%|█████▉    | 6620/11082 [21:16:39<14:18:39, 11.55s/step, acc=0.5038, loss=2.0349, lr=4.25e-05]
Training:  60%|█████▉    | 6621/11082 [21:16:50<14:08:59, 11.42s/step, acc=0.5038, loss=2.0349, lr=4.25e-05]
Training:  60%|█████▉    | 6622/11082 [21:17:02<14:18:48, 11.55s/step, acc=0.5038, loss=2.0349, lr=4.25e-05]
Training:  60%|█████▉    | 6623/11082 [21:17:13<14:19:10, 11.56s/step, acc=0.5038, loss=2.0349, lr=4.25e-05]
Training:  60%|█████▉    | 6624/11082 [21:17:24<14:09:38, 11.44s/step, acc=0.5038, loss=2.0349, lr=4.25e-05]
Training:  60%|█████▉    | 6625/11082 [21:17:36<14:14:28, 11.50s/step, acc=0.5038, loss=2.0349, lr=4.25e-05]
Training:  60%|█████▉    | 6626/11082 [21:17:48<14:19:36, 11.57s/step, acc=0.5038, loss=2.0349, lr=4.25e-05]
Training:  60%|█████▉    | 6627/11082 [21:17:59<14:19:51, 11.58s/step, acc=0.5038, loss=2.0349, lr=4.25e-05]
Training:  60%|█████▉    | 6628/11082 [21:18:11<14:27:27, 11.69s/step, acc=0.5038, loss=2.0349, lr=4.25e-05]
Training:  60%|█████▉    | 6629/11082 [21:18:22<14:18:03, 11.56s/step, acc=0.5038, loss=2.0349, lr=4.25e-05]
Training:  60%|█████▉    | 6630/11082 [21:18:34<14:21:04, 11.60s/step, acc=0.5038, loss=2.0349, lr=4.25e-05]
Training:  60%|█████▉    | 6631/11082 [21:18:46<14:21:04, 11.61s/step, acc=0.5038, loss=2.0349, lr=4.25e-05]
Training:  60%|█████▉    | 6632/11082 [21:18:57<14:22:18, 11.63s/step, acc=0.5038, loss=2.0349, lr=4.25e-05]
Training:  60%|█████▉    | 6633/11082 [21:19:09<14:19:13, 11.59s/step, acc=0.5038, loss=2.0349, lr=4.25e-05]
Training:  60%|█████▉    | 6634/11082 [21:19:20<14:12:10, 11.50s/step, acc=0.5038, loss=2.0349, lr=4.25e-05]
Training:  60%|█████▉    | 6635/11082 [21:19:32<14:08:18, 11.45s/step, acc=0.5038, loss=2.0349, lr=4.25e-05]
Training:  60%|█████▉    | 6636/11082 [21:19:43<14:16:02, 11.55s/step, acc=0.5038, loss=2.0349, lr=4.25e-05]
Training:  60%|█████▉    | 6637/11082 [21:19:55<14:19:58, 11.61s/step, acc=0.5038, loss=2.0349, lr=4.25e-05]
Training:  60%|█████▉    | 6638/11082 [21:20:07<14:22:56, 11.65s/step, acc=0.5038, loss=2.0349, lr=4.25e-05]
Training:  60%|█████▉    | 6639/11082 [21:20:19<14:23:29, 11.66s/step, acc=0.5038, loss=2.0349, lr=4.25e-05]
Training:  60%|█████▉    | 6640/11082 [21:20:30<14:12:56, 11.52s/step, acc=0.5038, loss=2.0349, lr=4.25e-05]
Training:  60%|█████▉    | 6641/11082 [21:20:42<14:21:43, 11.64s/step, acc=0.5038, loss=2.0349, lr=4.25e-05]
Training:  60%|█████▉    | 6642/11082 [21:20:53<14:14:27, 11.55s/step, acc=0.5038, loss=2.0349, lr=4.25e-05]
Training:  60%|█████▉    | 6643/11082 [21:21:04<14:12:13, 11.52s/step, acc=0.5038, loss=2.0349, lr=4.25e-05]
Training:  60%|█████▉    | 6644/11082 [21:21:16<14:18:50, 11.61s/step, acc=0.5038, loss=2.0349, lr=4.25e-05]
Training:  60%|█████▉    | 6645/11082 [21:21:28<14:12:06, 11.52s/step, acc=0.5038, loss=2.0349, lr=4.25e-05]
Training:  60%|█████▉    | 6646/11082 [21:21:39<14:09:17, 11.49s/step, acc=0.5038, loss=2.0349, lr=4.25e-05]
Training:  60%|█████▉    | 6647/11082 [21:21:51<14:13:01, 11.54s/step, acc=0.5038, loss=2.0349, lr=4.25e-05]
Training:  60%|█████▉    | 6648/11082 [21:22:02<14:14:03, 11.56s/step, acc=0.5038, loss=2.0349, lr=4.25e-05]
Training:  60%|█████▉    | 6649/11082 [21:22:14<14:18:00, 11.61s/step, acc=0.5038, loss=2.0349, lr=4.25e-05]
Training:  60%|██████    | 6650/11082 [21:22:25<14:14:43, 11.57s/step, acc=0.5038, loss=2.0349, lr=4.25e-05]
Training:  60%|██████    | 6650/11082 [21:22:33<14:14:43, 11.57s/step, acc=0.5014, loss=2.0514, lr=4.19e-05]2026-06-19 12:16:01 | INFO | __main__ | epoch=1 step=6650 train_loss=2.051406 train_acc=0.501389 lr=4.190e-05 grad_norm=0.7656 acc_d0=0.7597 acc_d1=0.6155 acc_d2=0.4939 acc_d3=0.3968 acc_d4=0.3180 acc_d5=0.2555 acc_d6=0.2085 acc_d7=0.1709

Training:  60%|██████    | 6651/11082 [21:22:37<14:16:36, 11.60s/step, acc=0.5014, loss=2.0514, lr=4.19e-05]
Training:  60%|██████    | 6652/11082 [21:22:49<14:20:54, 11.66s/step, acc=0.5014, loss=2.0514, lr=4.19e-05]
Training:  60%|██████    | 6653/11082 [21:23:00<14:18:03, 11.62s/step, acc=0.5014, loss=2.0514, lr=4.19e-05]
Training:  60%|██████    | 6654/11082 [21:23:12<14:13:17, 11.56s/step, acc=0.5014, loss=2.0514, lr=4.19e-05]
Training:  60%|██████    | 6655/11082 [21:23:24<14:14:50, 11.59s/step, acc=0.5014, loss=2.0514, lr=4.19e-05]
Training:  60%|██████    | 6656/11082 [21:23:35<14:17:43, 11.63s/step, acc=0.5014, loss=2.0514, lr=4.19e-05]
Training:  60%|██████    | 6657/11082 [21:23:46<14:03:50, 11.44s/step, acc=0.5014, loss=2.0514, lr=4.19e-05]
Training:  60%|██████    | 6658/11082 [21:23:58<14:11:15, 11.55s/step, acc=0.5014, loss=2.0514, lr=4.19e-05]
Training:  60%|██████    | 6659/11082 [21:24:10<14:10:44, 11.54s/step, acc=0.5014, loss=2.0514, lr=4.19e-05]
Training:  60%|██████    | 6660/11082 [21:24:21<14:09:19, 11.52s/step, acc=0.5014, loss=2.0514, lr=4.19e-05]
Training:  60%|██████    | 6661/11082 [21:24:33<14:08:06, 11.51s/step, acc=0.5014, loss=2.0514, lr=4.19e-05]
Training:  60%|██████    | 6662/11082 [21:24:44<14:09:25, 11.53s/step, acc=0.5014, loss=2.0514, lr=4.19e-05]
Training:  60%|██████    | 6663/11082 [21:24:56<14:12:43, 11.58s/step, acc=0.5014, loss=2.0514, lr=4.19e-05]
Training:  60%|██████    | 6664/11082 [21:25:07<14:09:02, 11.53s/step, acc=0.5014, loss=2.0514, lr=4.19e-05]
Training:  60%|██████    | 6665/11082 [21:25:19<14:15:46, 11.62s/step, acc=0.5014, loss=2.0514, lr=4.19e-05]
Training:  60%|██████    | 6666/11082 [21:25:31<14:12:48, 11.59s/step, acc=0.5014, loss=2.0514, lr=4.19e-05]
Training:  60%|██████    | 6667/11082 [21:25:42<14:07:46, 11.52s/step, acc=0.5014, loss=2.0514, lr=4.19e-05]
Training:  60%|██████    | 6668/11082 [21:25:54<14:11:45, 11.58s/step, acc=0.5014, loss=2.0514, lr=4.19e-05]
Training:  60%|██████    | 6669/11082 [21:26:05<14:12:53, 11.60s/step, acc=0.5014, loss=2.0514, lr=4.19e-05]
Training:  60%|██████    | 6670/11082 [21:26:17<14:10:04, 11.56s/step, acc=0.5014, loss=2.0514, lr=4.19e-05]
Training:  60%|██████    | 6671/11082 [21:26:28<14:04:00, 11.48s/step, acc=0.5014, loss=2.0514, lr=4.19e-05]
Training:  60%|██████    | 6672/11082 [21:26:40<14:09:30, 11.56s/step, acc=0.5014, loss=2.0514, lr=4.19e-05]
Training:  60%|██████    | 6673/11082 [21:26:51<14:11:50, 11.59s/step, acc=0.5014, loss=2.0514, lr=4.19e-05]
Training:  60%|██████    | 6674/11082 [21:27:03<14:16:57, 11.66s/step, acc=0.5014, loss=2.0514, lr=4.19e-05]
Training:  60%|██████    | 6675/11082 [21:27:14<14:04:56, 11.50s/step, acc=0.5014, loss=2.0514, lr=4.19e-05]
Training:  60%|██████    | 6676/11082 [21:27:26<14:02:12, 11.47s/step, acc=0.5014, loss=2.0514, lr=4.19e-05]
Training:  60%|██████    | 6677/11082 [21:27:37<14:07:00, 11.54s/step, acc=0.5014, loss=2.0514, lr=4.19e-05]
Training:  60%|██████    | 6678/11082 [21:27:49<14:07:29, 11.55s/step, acc=0.5014, loss=2.0514, lr=4.19e-05]
Training:  60%|██████    | 6679/11082 [21:28:01<14:11:14, 11.60s/step, acc=0.5014, loss=2.0514, lr=4.19e-05]
Training:  60%|██████    | 6680/11082 [21:28:12<14:12:20, 11.62s/step, acc=0.5014, loss=2.0514, lr=4.19e-05]
Training:  60%|██████    | 6681/11082 [21:28:24<14:05:53, 11.53s/step, acc=0.5014, loss=2.0514, lr=4.19e-05]
Training:  60%|██████    | 6682/11082 [21:28:35<14:08:46, 11.57s/step, acc=0.5014, loss=2.0514, lr=4.19e-05]
Training:  60%|██████    | 6683/11082 [21:28:47<14:06:04, 11.54s/step, acc=0.5014, loss=2.0514, lr=4.19e-05]
Training:  60%|██████    | 6684/11082 [21:28:58<13:59:39, 11.46s/step, acc=0.5014, loss=2.0514, lr=4.19e-05]
Training:  60%|██████    | 6685/11082 [21:29:10<14:04:59, 11.53s/step, acc=0.5014, loss=2.0514, lr=4.19e-05]
Training:  60%|██████    | 6686/11082 [21:29:21<14:06:25, 11.55s/step, acc=0.5014, loss=2.0514, lr=4.19e-05]
Training:  60%|██████    | 6687/11082 [21:29:33<14:11:50, 11.63s/step, acc=0.5014, loss=2.0514, lr=4.19e-05]
Training:  60%|██████    | 6688/11082 [21:29:45<14:08:58, 11.59s/step, acc=0.5014, loss=2.0514, lr=4.19e-05]
Training:  60%|██████    | 6689/11082 [21:29:56<14:04:46, 11.54s/step, acc=0.5014, loss=2.0514, lr=4.19e-05]
Training:  60%|██████    | 6690/11082 [21:30:08<14:08:25, 11.59s/step, acc=0.5014, loss=2.0514, lr=4.19e-05]
Training:  60%|██████    | 6691/11082 [21:30:20<14:11:58, 11.64s/step, acc=0.5014, loss=2.0514, lr=4.19e-05]
Training:  60%|██████    | 6692/11082 [21:30:32<14:19:40, 11.75s/step, acc=0.5014, loss=2.0514, lr=4.19e-05]
Training:  60%|██████    | 6693/11082 [21:30:43<14:08:06, 11.59s/step, acc=0.5014, loss=2.0514, lr=4.19e-05]
Training:  60%|██████    | 6694/11082 [21:30:55<14:13:10, 11.67s/step, acc=0.5014, loss=2.0514, lr=4.19e-05]
Training:  60%|██████    | 6695/11082 [21:31:06<14:05:39, 11.57s/step, acc=0.5014, loss=2.0514, lr=4.19e-05]
Training:  60%|██████    | 6696/11082 [21:31:18<14:12:27, 11.66s/step, acc=0.5014, loss=2.0514, lr=4.19e-05]
Training:  60%|██████    | 6697/11082 [21:31:29<14:00:54, 11.51s/step, acc=0.5014, loss=2.0514, lr=4.19e-05]
Training:  60%|██████    | 6698/11082 [21:31:41<14:16:40, 11.72s/step, acc=0.5014, loss=2.0514, lr=4.19e-05]
Training:  60%|██████    | 6699/11082 [21:31:53<14:15:14, 11.71s/step, acc=0.5014, loss=2.0514, lr=4.19e-05]
Training:  60%|██████    | 6700/11082 [21:32:05<14:14:06, 11.69s/step, acc=0.5014, loss=2.0514, lr=4.19e-05]
Training:  60%|██████    | 6700/11082 [21:32:12<14:14:06, 11.69s/step, acc=0.5038, loss=2.0338, lr=4.13e-05]2026-06-19 12:25:40 | INFO | __main__ | epoch=1 step=6700 train_loss=2.033751 train_acc=0.503754 lr=4.129e-05 grad_norm=0.7461 acc_d0=0.7656 acc_d1=0.6225 acc_d2=0.5037 acc_d3=0.4052 acc_d4=0.3265 acc_d5=0.2620 acc_d6=0.2128 acc_d7=0.1755

Training:  60%|██████    | 6701/11082 [21:32:16<14:07:07, 11.60s/step, acc=0.5038, loss=2.0338, lr=4.13e-05]
Training:  60%|██████    | 6702/11082 [21:32:28<14:03:34, 11.56s/step, acc=0.5038, loss=2.0338, lr=4.13e-05]
Training:  60%|██████    | 6703/11082 [21:32:39<14:06:05, 11.59s/step, acc=0.5038, loss=2.0338, lr=4.13e-05]
Training:  60%|██████    | 6704/11082 [21:32:51<14:17:47, 11.76s/step, acc=0.5038, loss=2.0338, lr=4.13e-05]
Training:  61%|██████    | 6705/11082 [21:33:02<13:53:05, 11.42s/step, acc=0.5038, loss=2.0338, lr=4.13e-05]
Training:  61%|██████    | 6706/11082 [21:33:14<14:06:44, 11.61s/step, acc=0.5038, loss=2.0338, lr=4.13e-05]
Training:  61%|██████    | 6707/11082 [21:33:26<14:04:44, 11.58s/step, acc=0.5038, loss=2.0338, lr=4.13e-05]
Training:  61%|██████    | 6708/11082 [21:33:38<14:13:09, 11.70s/step, acc=0.5038, loss=2.0338, lr=4.13e-05]
Training:  61%|██████    | 6709/11082 [21:33:49<14:07:25, 11.63s/step, acc=0.5038, loss=2.0338, lr=4.13e-05]
Training:  61%|██████    | 6710/11082 [21:34:00<14:05:11, 11.60s/step, acc=0.5038, loss=2.0338, lr=4.13e-05]
Training:  61%|██████    | 6711/11082 [21:34:12<14:01:24, 11.55s/step, acc=0.5038, loss=2.0338, lr=4.13e-05]
Training:  61%|██████    | 6712/11082 [21:34:24<14:11:13, 11.69s/step, acc=0.5038, loss=2.0338, lr=4.13e-05]
Training:  61%|██████    | 6713/11082 [21:34:36<14:08:38, 11.65s/step, acc=0.5038, loss=2.0338, lr=4.13e-05]
Training:  61%|██████    | 6714/11082 [21:34:47<14:12:20, 11.71s/step, acc=0.5038, loss=2.0338, lr=4.13e-05]
Training:  61%|██████    | 6715/11082 [21:34:58<13:50:39, 11.41s/step, acc=0.5038, loss=2.0338, lr=4.13e-05]
Training:  61%|██████    | 6716/11082 [21:35:10<14:05:39, 11.62s/step, acc=0.5038, loss=2.0338, lr=4.13e-05]
Training:  61%|██████    | 6717/11082 [21:35:22<14:06:43, 11.64s/step, acc=0.5038, loss=2.0338, lr=4.13e-05]
Training:  61%|██████    | 6718/11082 [21:35:33<14:00:43, 11.56s/step, acc=0.5038, loss=2.0338, lr=4.13e-05]
Training:  61%|██████    | 6719/11082 [21:35:45<13:55:45, 11.49s/step, acc=0.5038, loss=2.0338, lr=4.13e-05]
Training:  61%|██████    | 6720/11082 [21:35:56<14:00:14, 11.56s/step, acc=0.5038, loss=2.0338, lr=4.13e-05]
Training:  61%|██████    | 6721/11082 [21:36:08<14:01:43, 11.58s/step, acc=0.5038, loss=2.0338, lr=4.13e-05]
Training:  61%|██████    | 6722/11082 [21:36:19<13:58:17, 11.54s/step, acc=0.5038, loss=2.0338, lr=4.13e-05]
Training:  61%|██████    | 6723/11082 [21:36:31<13:56:49, 11.52s/step, acc=0.5038, loss=2.0338, lr=4.13e-05]
Training:  61%|██████    | 6724/11082 [21:36:42<13:54:31, 11.49s/step, acc=0.5038, loss=2.0338, lr=4.13e-05]
Training:  61%|██████    | 6725/11082 [21:36:54<13:51:54, 11.46s/step, acc=0.5038, loss=2.0338, lr=4.13e-05]
Training:  61%|██████    | 6726/11082 [21:37:05<13:55:55, 11.51s/step, acc=0.5038, loss=2.0338, lr=4.13e-05]
Training:  61%|██████    | 6727/11082 [21:37:17<14:02:49, 11.61s/step, acc=0.5038, loss=2.0338, lr=4.13e-05]
Training:  61%|██████    | 6728/11082 [21:37:28<13:48:45, 11.42s/step, acc=0.5038, loss=2.0338, lr=4.13e-05]
Training:  61%|██████    | 6729/11082 [21:37:40<13:55:10, 11.51s/step, acc=0.5038, loss=2.0338, lr=4.13e-05]
Training:  61%|██████    | 6730/11082 [21:37:52<14:02:35, 11.62s/step, acc=0.5038, loss=2.0338, lr=4.13e-05]
Training:  61%|██████    | 6731/11082 [21:38:03<14:02:42, 11.62s/step, acc=0.5038, loss=2.0338, lr=4.13e-05]
Training:  61%|██████    | 6732/11082 [21:38:15<14:02:07, 11.62s/step, acc=0.5038, loss=2.0338, lr=4.13e-05]
Training:  61%|██████    | 6733/11082 [21:38:26<13:52:37, 11.49s/step, acc=0.5038, loss=2.0338, lr=4.13e-05]
Training:  61%|██████    | 6734/11082 [21:38:38<14:03:54, 11.65s/step, acc=0.5038, loss=2.0338, lr=4.13e-05]
Training:  61%|██████    | 6735/11082 [21:38:50<14:03:53, 11.65s/step, acc=0.5038, loss=2.0338, lr=4.13e-05]
Training:  61%|██████    | 6736/11082 [21:39:01<13:57:22, 11.56s/step, acc=0.5038, loss=2.0338, lr=4.13e-05]
Training:  61%|██████    | 6737/11082 [21:39:13<14:04:52, 11.67s/step, acc=0.5038, loss=2.0338, lr=4.13e-05]
Training:  61%|██████    | 6738/11082 [21:39:25<14:02:27, 11.64s/step, acc=0.5038, loss=2.0338, lr=4.13e-05]
Training:  61%|██████    | 6739/11082 [21:39:36<13:55:29, 11.54s/step, acc=0.5038, loss=2.0338, lr=4.13e-05]
Training:  61%|██████    | 6740/11082 [21:39:47<13:55:52, 11.55s/step, acc=0.5038, loss=2.0338, lr=4.13e-05]
Training:  61%|██████    | 6741/11082 [21:39:59<13:57:41, 11.58s/step, acc=0.5038, loss=2.0338, lr=4.13e-05]
Training:  61%|██████    | 6742/11082 [21:40:10<13:46:54, 11.43s/step, acc=0.5038, loss=2.0338, lr=4.13e-05]
Training:  61%|██████    | 6743/11082 [21:40:22<14:01:00, 11.63s/step, acc=0.5038, loss=2.0338, lr=4.13e-05]
Training:  61%|██████    | 6744/11082 [21:40:34<14:00:39, 11.63s/step, acc=0.5038, loss=2.0338, lr=4.13e-05]
Training:  61%|██████    | 6745/11082 [21:40:45<13:54:16, 11.54s/step, acc=0.5038, loss=2.0338, lr=4.13e-05]
Training:  61%|██████    | 6746/11082 [21:40:57<13:53:58, 11.54s/step, acc=0.5038, loss=2.0338, lr=4.13e-05]
Training:  61%|██████    | 6747/11082 [21:41:08<13:56:25, 11.58s/step, acc=0.5038, loss=2.0338, lr=4.13e-05]
Training:  61%|██████    | 6748/11082 [21:41:20<13:51:57, 11.52s/step, acc=0.5038, loss=2.0338, lr=4.13e-05]
Training:  61%|██████    | 6749/11082 [21:41:31<13:47:12, 11.45s/step, acc=0.5038, loss=2.0338, lr=4.13e-05]
Training:  61%|██████    | 6750/11082 [21:41:43<13:55:39, 11.57s/step, acc=0.5038, loss=2.0338, lr=4.13e-05]
Training:  61%|██████    | 6750/11082 [21:41:50<13:55:39, 11.57s/step, acc=0.5024, loss=2.0409, lr=4.07e-05]2026-06-19 12:35:18 | INFO | __main__ | epoch=1 step=6750 train_loss=2.040946 train_acc=0.502436 lr=4.067e-05 grad_norm=0.6914 acc_d0=0.7632 acc_d1=0.6191 acc_d2=0.4992 acc_d3=0.4010 acc_d4=0.3247 acc_d5=0.2600 acc_d6=0.2117 acc_d7=0.1713

Training:  61%|██████    | 6751/11082 [21:41:55<13:55:23, 11.57s/step, acc=0.5024, loss=2.0409, lr=4.07e-05]
Training:  61%|██████    | 6752/11082 [21:42:06<13:52:30, 11.54s/step, acc=0.5024, loss=2.0409, lr=4.07e-05]
Training:  61%|██████    | 6753/11082 [21:42:18<13:54:02, 11.56s/step, acc=0.5024, loss=2.0409, lr=4.07e-05]
Training:  61%|██████    | 6754/11082 [21:42:29<13:55:59, 11.59s/step, acc=0.5024, loss=2.0409, lr=4.07e-05]
Training:  61%|██████    | 6755/11082 [21:42:41<13:47:58, 11.48s/step, acc=0.5024, loss=2.0409, lr=4.07e-05]
Training:  61%|██████    | 6756/11082 [21:42:52<13:46:32, 11.46s/step, acc=0.5024, loss=2.0409, lr=4.07e-05]
Training:  61%|██████    | 6757/11082 [21:43:04<13:51:23, 11.53s/step, acc=0.5024, loss=2.0409, lr=4.07e-05]
Training:  61%|██████    | 6758/11082 [21:43:16<14:02:07, 11.69s/step, acc=0.5024, loss=2.0409, lr=4.07e-05]
Training:  61%|██████    | 6759/11082 [21:43:27<13:51:45, 11.54s/step, acc=0.5024, loss=2.0409, lr=4.07e-05]
Training:  61%|██████    | 6760/11082 [21:43:38<13:42:56, 11.42s/step, acc=0.5024, loss=2.0409, lr=4.07e-05]
Training:  61%|██████    | 6761/11082 [21:43:50<13:57:05, 11.62s/step, acc=0.5024, loss=2.0409, lr=4.07e-05]
Training:  61%|██████    | 6762/11082 [21:44:02<13:51:26, 11.55s/step, acc=0.5024, loss=2.0409, lr=4.07e-05]
Training:  61%|██████    | 6763/11082 [21:44:13<13:57:12, 11.63s/step, acc=0.5024, loss=2.0409, lr=4.07e-05]
Training:  61%|██████    | 6764/11082 [21:44:25<14:01:12, 11.69s/step, acc=0.5024, loss=2.0409, lr=4.07e-05]
Training:  61%|██████    | 6765/11082 [21:44:37<13:57:05, 11.63s/step, acc=0.5024, loss=2.0409, lr=4.07e-05]
Training:  61%|██████    | 6766/11082 [21:44:48<13:51:45, 11.56s/step, acc=0.5024, loss=2.0409, lr=4.07e-05]
Training:  61%|██████    | 6767/11082 [21:44:59<13:47:18, 11.50s/step, acc=0.5024, loss=2.0409, lr=4.07e-05]
Training:  61%|██████    | 6768/11082 [21:45:11<13:46:56, 11.50s/step, acc=0.5024, loss=2.0409, lr=4.07e-05]
Training:  61%|██████    | 6769/11082 [21:45:22<13:45:41, 11.49s/step, acc=0.5024, loss=2.0409, lr=4.07e-05]
Training:  61%|██████    | 6770/11082 [21:45:34<13:55:04, 11.62s/step, acc=0.5024, loss=2.0409, lr=4.07e-05]
Training:  61%|██████    | 6771/11082 [21:45:46<13:47:50, 11.52s/step, acc=0.5024, loss=2.0409, lr=4.07e-05]
Training:  61%|██████    | 6772/11082 [21:45:58<13:59:50, 11.69s/step, acc=0.5024, loss=2.0409, lr=4.07e-05]
Training:  61%|██████    | 6773/11082 [21:46:09<13:56:13, 11.64s/step, acc=0.5024, loss=2.0409, lr=4.07e-05]
Training:  61%|██████    | 6774/11082 [21:46:20<13:42:56, 11.46s/step, acc=0.5024, loss=2.0409, lr=4.07e-05]
Training:  61%|██████    | 6775/11082 [21:46:32<13:50:36, 11.57s/step, acc=0.5024, loss=2.0409, lr=4.07e-05]
Training:  61%|██████    | 6776/11082 [21:46:44<13:54:15, 11.62s/step, acc=0.5024, loss=2.0409, lr=4.07e-05]
Training:  61%|██████    | 6777/11082 [21:46:55<13:54:20, 11.63s/step, acc=0.5024, loss=2.0409, lr=4.07e-05]
Training:  61%|██████    | 6778/11082 [21:47:07<13:55:53, 11.65s/step, acc=0.5024, loss=2.0409, lr=4.07e-05]
Training:  61%|██████    | 6779/11082 [21:47:19<13:58:49, 11.70s/step, acc=0.5024, loss=2.0409, lr=4.07e-05]
Training:  61%|██████    | 6780/11082 [21:47:31<13:57:16, 11.68s/step, acc=0.5024, loss=2.0409, lr=4.07e-05]
Training:  61%|██████    | 6781/11082 [21:47:42<13:53:12, 11.62s/step, acc=0.5024, loss=2.0409, lr=4.07e-05]
Training:  61%|██████    | 6782/11082 [21:47:54<13:49:28, 11.57s/step, acc=0.5024, loss=2.0409, lr=4.07e-05]
Training:  61%|██████    | 6783/11082 [21:48:05<13:53:37, 11.63s/step, acc=0.5024, loss=2.0409, lr=4.07e-05]
Training:  61%|██████    | 6784/11082 [21:48:17<13:47:15, 11.55s/step, acc=0.5024, loss=2.0409, lr=4.07e-05]
Training:  61%|██████    | 6785/11082 [21:48:28<13:46:54, 11.55s/step, acc=0.5024, loss=2.0409, lr=4.07e-05]
Training:  61%|██████    | 6786/11082 [21:48:39<13:38:26, 11.43s/step, acc=0.5024, loss=2.0409, lr=4.07e-05]
Training:  61%|██████    | 6787/11082 [21:48:51<13:41:26, 11.48s/step, acc=0.5024, loss=2.0409, lr=4.07e-05]
Training:  61%|██████▏   | 6788/11082 [21:49:03<13:52:34, 11.63s/step, acc=0.5024, loss=2.0409, lr=4.07e-05]
Training:  61%|██████▏   | 6789/11082 [21:49:15<13:53:21, 11.65s/step, acc=0.5024, loss=2.0409, lr=4.07e-05]
Training:  61%|██████▏   | 6790/11082 [21:49:26<13:47:16, 11.56s/step, acc=0.5024, loss=2.0409, lr=4.07e-05]
Training:  61%|██████▏   | 6791/11082 [21:49:38<13:45:34, 11.54s/step, acc=0.5024, loss=2.0409, lr=4.07e-05]
Training:  61%|██████▏   | 6792/11082 [21:49:49<13:40:38, 11.48s/step, acc=0.5024, loss=2.0409, lr=4.07e-05]
Training:  61%|██████▏   | 6793/11082 [21:50:01<13:55:05, 11.68s/step, acc=0.5024, loss=2.0409, lr=4.07e-05]
Training:  61%|██████▏   | 6794/11082 [21:50:12<13:40:05, 11.48s/step, acc=0.5024, loss=2.0409, lr=4.07e-05]
Training:  61%|██████▏   | 6795/11082 [21:50:24<13:50:39, 11.63s/step, acc=0.5024, loss=2.0409, lr=4.07e-05]
Training:  61%|██████▏   | 6796/11082 [21:50:35<13:43:59, 11.54s/step, acc=0.5024, loss=2.0409, lr=4.07e-05]
Training:  61%|██████▏   | 6797/11082 [21:50:47<13:40:35, 11.49s/step, acc=0.5024, loss=2.0409, lr=4.07e-05]
Training:  61%|██████▏   | 6798/11082 [21:50:58<13:43:17, 11.53s/step, acc=0.5024, loss=2.0409, lr=4.07e-05]
Training:  61%|██████▏   | 6799/11082 [21:51:10<13:50:19, 11.63s/step, acc=0.5024, loss=2.0409, lr=4.07e-05]
Training:  61%|██████▏   | 6800/11082 [21:51:22<13:43:48, 11.54s/step, acc=0.5024, loss=2.0409, lr=4.07e-05]
Training:  61%|██████▏   | 6800/11082 [21:51:29<13:43:48, 11.54s/step, acc=0.5041, loss=2.0312, lr=4.01e-05]2026-06-19 12:44:57 | INFO | __main__ | epoch=1 step=6800 train_loss=2.031151 train_acc=0.504145 lr=4.006e-05 grad_norm=0.6484 acc_d0=0.7657 acc_d1=0.6211 acc_d2=0.5007 acc_d3=0.4031 acc_d4=0.3243 acc_d5=0.2602 acc_d6=0.2106 acc_d7=0.1722

Training:  61%|██████▏   | 6801/11082 [21:51:33<13:52:10, 11.66s/step, acc=0.5041, loss=2.0312, lr=4.01e-05]
Training:  61%|██████▏   | 6802/11082 [21:51:45<13:41:35, 11.52s/step, acc=0.5041, loss=2.0312, lr=4.01e-05]
Training:  61%|██████▏   | 6803/11082 [21:51:57<13:53:30, 11.69s/step, acc=0.5041, loss=2.0312, lr=4.01e-05]
Training:  61%|██████▏   | 6804/11082 [21:52:08<13:39:55, 11.50s/step, acc=0.5041, loss=2.0312, lr=4.01e-05]
Training:  61%|██████▏   | 6805/11082 [21:52:20<13:47:28, 11.61s/step, acc=0.5041, loss=2.0312, lr=4.01e-05]
Training:  61%|██████▏   | 6806/11082 [21:52:31<13:48:26, 11.62s/step, acc=0.5041, loss=2.0312, lr=4.01e-05]
Training:  61%|██████▏   | 6807/11082 [21:52:43<13:48:45, 11.63s/step, acc=0.5041, loss=2.0312, lr=4.01e-05]
Training:  61%|██████▏   | 6808/11082 [21:52:54<13:42:05, 11.54s/step, acc=0.5041, loss=2.0312, lr=4.01e-05]
Training:  61%|██████▏   | 6809/11082 [21:53:06<13:48:51, 11.64s/step, acc=0.5041, loss=2.0312, lr=4.01e-05]
Training:  61%|██████▏   | 6810/11082 [21:53:18<13:48:21, 11.63s/step, acc=0.5041, loss=2.0312, lr=4.01e-05]
Training:  61%|██████▏   | 6811/11082 [21:53:30<13:50:11, 11.66s/step, acc=0.5041, loss=2.0312, lr=4.01e-05]
Training:  61%|██████▏   | 6812/11082 [21:53:41<13:48:55, 11.65s/step, acc=0.5041, loss=2.0312, lr=4.01e-05]
Training:  61%|██████▏   | 6813/11082 [21:53:53<13:59:42, 11.80s/step, acc=0.5041, loss=2.0312, lr=4.01e-05]
Training:  61%|██████▏   | 6814/11082 [21:54:05<13:57:02, 11.77s/step, acc=0.5041, loss=2.0312, lr=4.01e-05]
Training:  61%|██████▏   | 6815/11082 [21:54:16<13:46:02, 11.62s/step, acc=0.5041, loss=2.0312, lr=4.01e-05]
Training:  62%|██████▏   | 6816/11082 [21:54:27<13:36:10, 11.48s/step, acc=0.5041, loss=2.0312, lr=4.01e-05]
Training:  62%|██████▏   | 6817/11082 [21:54:40<13:55:00, 11.75s/step, acc=0.5041, loss=2.0312, lr=4.01e-05]
Training:  62%|██████▏   | 6818/11082 [21:54:51<13:34:35, 11.46s/step, acc=0.5041, loss=2.0312, lr=4.01e-05]
Training:  62%|██████▏   | 6819/11082 [21:55:02<13:43:08, 11.59s/step, acc=0.5041, loss=2.0312, lr=4.01e-05]
Training:  62%|██████▏   | 6820/11082 [21:55:14<13:39:03, 11.53s/step, acc=0.5041, loss=2.0312, lr=4.01e-05]
Training:  62%|██████▏   | 6821/11082 [21:55:26<13:43:38, 11.60s/step, acc=0.5041, loss=2.0312, lr=4.01e-05]
Training:  62%|██████▏   | 6822/11082 [21:55:37<13:48:50, 11.67s/step, acc=0.5041, loss=2.0312, lr=4.01e-05]
Training:  62%|██████▏   | 6823/11082 [21:55:48<13:34:14, 11.47s/step, acc=0.5041, loss=2.0312, lr=4.01e-05]
Training:  62%|██████▏   | 6824/11082 [21:56:00<13:34:40, 11.48s/step, acc=0.5041, loss=2.0312, lr=4.01e-05]
Training:  62%|██████▏   | 6825/11082 [21:56:12<13:39:15, 11.55s/step, acc=0.5041, loss=2.0312, lr=4.01e-05]
Training:  62%|██████▏   | 6826/11082 [21:56:23<13:37:56, 11.53s/step, acc=0.5041, loss=2.0312, lr=4.01e-05]
Training:  62%|██████▏   | 6827/11082 [21:56:35<13:43:09, 11.61s/step, acc=0.5041, loss=2.0312, lr=4.01e-05]
Training:  62%|██████▏   | 6828/11082 [21:56:47<13:45:12, 11.64s/step, acc=0.5041, loss=2.0312, lr=4.01e-05]
Training:  62%|██████▏   | 6829/11082 [21:56:58<13:43:05, 11.61s/step, acc=0.5041, loss=2.0312, lr=4.01e-05]
Training:  62%|██████▏   | 6830/11082 [21:57:09<13:35:17, 11.50s/step, acc=0.5041, loss=2.0312, lr=4.01e-05]
Training:  62%|██████▏   | 6831/11082 [21:57:21<13:39:01, 11.56s/step, acc=0.5041, loss=2.0312, lr=4.01e-05]
Training:  62%|██████▏   | 6832/11082 [21:57:32<13:29:28, 11.43s/step, acc=0.5041, loss=2.0312, lr=4.01e-05]
Training:  62%|██████▏   | 6833/11082 [21:57:44<13:40:03, 11.58s/step, acc=0.5041, loss=2.0312, lr=4.01e-05]
Training:  62%|██████▏   | 6834/11082 [21:57:56<13:38:25, 11.56s/step, acc=0.5041, loss=2.0312, lr=4.01e-05]
Training:  62%|██████▏   | 6835/11082 [21:58:07<13:34:06, 11.50s/step, acc=0.5041, loss=2.0312, lr=4.01e-05]
Training:  62%|██████▏   | 6836/11082 [21:58:19<13:35:22, 11.52s/step, acc=0.5041, loss=2.0312, lr=4.01e-05]
Training:  62%|██████▏   | 6837/11082 [21:58:30<13:34:05, 11.51s/step, acc=0.5041, loss=2.0312, lr=4.01e-05]
Training:  62%|██████▏   | 6838/11082 [21:58:42<13:37:24, 11.56s/step, acc=0.5041, loss=2.0312, lr=4.01e-05]
Training:  62%|██████▏   | 6839/11082 [21:58:54<13:43:48, 11.65s/step, acc=0.5041, loss=2.0312, lr=4.01e-05]
Training:  62%|██████▏   | 6840/11082 [21:59:05<13:35:04, 11.53s/step, acc=0.5041, loss=2.0312, lr=4.01e-05]
Training:  62%|██████▏   | 6841/11082 [21:59:17<13:36:49, 11.56s/step, acc=0.5041, loss=2.0312, lr=4.01e-05]
Training:  62%|██████▏   | 6842/11082 [21:59:28<13:39:17, 11.59s/step, acc=0.5041, loss=2.0312, lr=4.01e-05]
Training:  62%|██████▏   | 6843/11082 [21:59:39<13:29:02, 11.45s/step, acc=0.5041, loss=2.0312, lr=4.01e-05]
Training:  62%|██████▏   | 6844/11082 [21:59:51<13:41:18, 11.63s/step, acc=0.5041, loss=2.0312, lr=4.01e-05]
Training:  62%|██████▏   | 6845/11082 [22:00:03<13:47:45, 11.72s/step, acc=0.5041, loss=2.0312, lr=4.01e-05]
Training:  62%|██████▏   | 6846/11082 [22:00:15<13:38:56, 11.60s/step, acc=0.5041, loss=2.0312, lr=4.01e-05]
Training:  62%|██████▏   | 6847/11082 [22:00:26<13:42:52, 11.66s/step, acc=0.5041, loss=2.0312, lr=4.01e-05]
Training:  62%|██████▏   | 6848/11082 [22:00:38<13:46:25, 11.71s/step, acc=0.5041, loss=2.0312, lr=4.01e-05]
Training:  62%|██████▏   | 6849/11082 [22:00:49<13:31:25, 11.50s/step, acc=0.5041, loss=2.0312, lr=4.01e-05]
Training:  62%|██████▏   | 6850/11082 [22:01:01<13:42:11, 11.66s/step, acc=0.5041, loss=2.0312, lr=4.01e-05]
Training:  62%|██████▏   | 6850/11082 [22:01:08<13:42:11, 11.66s/step, acc=0.5042, loss=2.0357, lr=3.94e-05]2026-06-19 12:54:36 | INFO | __main__ | epoch=1 step=6850 train_loss=2.035658 train_acc=0.504204 lr=3.945e-05 grad_norm=0.7734 acc_d0=0.7656 acc_d1=0.6213 acc_d2=0.5031 acc_d3=0.4032 acc_d4=0.3255 acc_d5=0.2600 acc_d6=0.2125 acc_d7=0.1742

Training:  62%|██████▏   | 6851/11082 [22:01:13<13:33:46, 11.54s/step, acc=0.5042, loss=2.0357, lr=3.94e-05]
Training:  62%|██████▏   | 6852/11082 [22:01:24<13:37:35, 11.60s/step, acc=0.5042, loss=2.0357, lr=3.94e-05]
Training:  62%|██████▏   | 6853/11082 [22:01:36<13:45:47, 11.72s/step, acc=0.5042, loss=2.0357, lr=3.94e-05]
Training:  62%|██████▏   | 6854/11082 [22:01:48<13:37:12, 11.60s/step, acc=0.5042, loss=2.0357, lr=3.94e-05]
Training:  62%|██████▏   | 6855/11082 [22:01:59<13:41:11, 11.66s/step, acc=0.5042, loss=2.0357, lr=3.94e-05]
Training:  62%|██████▏   | 6856/11082 [22:02:11<13:36:56, 11.60s/step, acc=0.5042, loss=2.0357, lr=3.94e-05]
Training:  62%|██████▏   | 6857/11082 [22:02:22<13:35:56, 11.59s/step, acc=0.5042, loss=2.0357, lr=3.94e-05]
Training:  62%|██████▏   | 6858/11082 [22:02:34<13:32:02, 11.53s/step, acc=0.5042, loss=2.0357, lr=3.94e-05]
Training:  62%|██████▏   | 6859/11082 [22:02:45<13:31:38, 11.53s/step, acc=0.5042, loss=2.0357, lr=3.94e-05]
Training:  62%|██████▏   | 6860/11082 [22:02:57<13:30:09, 11.51s/step, acc=0.5042, loss=2.0357, lr=3.94e-05]
Training:  62%|██████▏   | 6861/11082 [22:03:09<13:37:00, 11.61s/step, acc=0.5042, loss=2.0357, lr=3.94e-05]
Training:  62%|██████▏   | 6862/11082 [22:03:21<13:43:48, 11.71s/step, acc=0.5042, loss=2.0357, lr=3.94e-05]
Training:  62%|██████▏   | 6863/11082 [22:03:32<13:32:34, 11.56s/step, acc=0.5042, loss=2.0357, lr=3.94e-05]
Training:  62%|██████▏   | 6864/11082 [22:03:44<13:37:35, 11.63s/step, acc=0.5042, loss=2.0357, lr=3.94e-05]
Training:  62%|██████▏   | 6865/11082 [22:03:55<13:31:39, 11.55s/step, acc=0.5042, loss=2.0357, lr=3.94e-05]
Training:  62%|██████▏   | 6866/11082 [22:04:06<13:26:46, 11.48s/step, acc=0.5042, loss=2.0357, lr=3.94e-05]
Training:  62%|██████▏   | 6867/11082 [22:04:18<13:32:08, 11.56s/step, acc=0.5042, loss=2.0357, lr=3.94e-05]
Training:  62%|██████▏   | 6868/11082 [22:04:30<13:36:42, 11.63s/step, acc=0.5042, loss=2.0357, lr=3.94e-05]
Training:  62%|██████▏   | 6869/11082 [22:04:41<13:27:36, 11.50s/step, acc=0.5042, loss=2.0357, lr=3.94e-05]
Training:  62%|██████▏   | 6870/11082 [22:04:53<13:30:22, 11.54s/step, acc=0.5042, loss=2.0357, lr=3.94e-05]
Training:  62%|██████▏   | 6871/11082 [22:05:04<13:26:07, 11.49s/step, acc=0.5042, loss=2.0357, lr=3.94e-05]
Training:  62%|██████▏   | 6872/11082 [22:05:16<13:33:10, 11.59s/step, acc=0.5042, loss=2.0357, lr=3.94e-05]
Training:  62%|██████▏   | 6873/11082 [22:05:27<13:30:17, 11.55s/step, acc=0.5042, loss=2.0357, lr=3.94e-05]
Training:  62%|██████▏   | 6874/11082 [22:05:39<13:37:14, 11.65s/step, acc=0.5042, loss=2.0357, lr=3.94e-05]
Training:  62%|██████▏   | 6875/11082 [22:05:50<13:25:44, 11.49s/step, acc=0.5042, loss=2.0357, lr=3.94e-05]
Training:  62%|██████▏   | 6876/11082 [22:06:02<13:30:42, 11.57s/step, acc=0.5042, loss=2.0357, lr=3.94e-05]
Training:  62%|██████▏   | 6877/11082 [22:06:13<13:20:24, 11.42s/step, acc=0.5042, loss=2.0357, lr=3.94e-05]
Training:  62%|██████▏   | 6878/11082 [22:06:25<13:35:34, 11.64s/step, acc=0.5042, loss=2.0357, lr=3.94e-05]
Training:  62%|██████▏   | 6879/11082 [22:06:37<13:36:18, 11.65s/step, acc=0.5042, loss=2.0357, lr=3.94e-05]
Training:  62%|██████▏   | 6880/11082 [22:06:48<13:30:42, 11.58s/step, acc=0.5042, loss=2.0357, lr=3.94e-05]
Training:  62%|██████▏   | 6881/11082 [22:07:00<13:25:06, 11.50s/step, acc=0.5042, loss=2.0357, lr=3.94e-05]
Training:  62%|██████▏   | 6882/11082 [22:07:12<13:39:04, 11.70s/step, acc=0.5042, loss=2.0357, lr=3.94e-05]
Training:  62%|██████▏   | 6883/11082 [22:07:23<13:29:26, 11.57s/step, acc=0.5042, loss=2.0357, lr=3.94e-05]
Training:  62%|██████▏   | 6884/11082 [22:07:35<13:30:54, 11.59s/step, acc=0.5042, loss=2.0357, lr=3.94e-05]
Training:  62%|██████▏   | 6885/11082 [22:07:46<13:27:34, 11.55s/step, acc=0.5042, loss=2.0357, lr=3.94e-05]
Training:  62%|██████▏   | 6886/11082 [22:07:58<13:37:46, 11.69s/step, acc=0.5042, loss=2.0357, lr=3.94e-05]
Training:  62%|██████▏   | 6887/11082 [22:08:09<13:24:56, 11.51s/step, acc=0.5042, loss=2.0357, lr=3.94e-05]
Training:  62%|██████▏   | 6888/11082 [22:08:21<13:27:45, 11.56s/step, acc=0.5042, loss=2.0357, lr=3.94e-05]
Training:  62%|██████▏   | 6889/11082 [22:08:33<13:30:50, 11.60s/step, acc=0.5042, loss=2.0357, lr=3.94e-05]
Training:  62%|██████▏   | 6890/11082 [22:08:44<13:24:02, 11.51s/step, acc=0.5042, loss=2.0357, lr=3.94e-05]
Training:  62%|██████▏   | 6891/11082 [22:08:56<13:25:43, 11.54s/step, acc=0.5042, loss=2.0357, lr=3.94e-05]
Training:  62%|██████▏   | 6892/11082 [22:09:07<13:23:15, 11.50s/step, acc=0.5042, loss=2.0357, lr=3.94e-05]
Training:  62%|██████▏   | 6893/11082 [22:09:19<13:30:01, 11.60s/step, acc=0.5042, loss=2.0357, lr=3.94e-05]
Training:  62%|██████▏   | 6894/11082 [22:09:30<13:21:35, 11.48s/step, acc=0.5042, loss=2.0357, lr=3.94e-05]
Training:  62%|██████▏   | 6895/11082 [22:09:42<13:30:29, 11.61s/step, acc=0.5042, loss=2.0357, lr=3.94e-05]
Training:  62%|██████▏   | 6896/11082 [22:09:53<13:28:12, 11.58s/step, acc=0.5042, loss=2.0357, lr=3.94e-05]
Training:  62%|██████▏   | 6897/11082 [22:10:05<13:31:04, 11.63s/step, acc=0.5042, loss=2.0357, lr=3.94e-05]
Training:  62%|██████▏   | 6898/11082 [22:10:17<13:31:43, 11.64s/step, acc=0.5042, loss=2.0357, lr=3.94e-05]
Training:  62%|██████▏   | 6899/11082 [22:10:28<13:30:27, 11.63s/step, acc=0.5042, loss=2.0357, lr=3.94e-05]
Training:  62%|██████▏   | 6900/11082 [22:10:40<13:23:55, 11.53s/step, acc=0.5042, loss=2.0357, lr=3.94e-05]
Training:  62%|██████▏   | 6900/11082 [22:10:47<13:23:55, 11.53s/step, acc=0.5027, loss=2.0420, lr=3.88e-05]2026-06-19 13:04:15 | INFO | __main__ | epoch=1 step=6900 train_loss=2.042012 train_acc=0.502683 lr=3.884e-05 grad_norm=0.7930 acc_d0=0.7651 acc_d1=0.6197 acc_d2=0.4983 acc_d3=0.4003 acc_d4=0.3207 acc_d5=0.2590 acc_d6=0.2089 acc_d7=0.1721

Training:  62%|██████▏   | 6901/11082 [22:10:52<13:28:17, 11.60s/step, acc=0.5027, loss=2.0420, lr=3.88e-05]
Training:  62%|██████▏   | 6902/11082 [22:11:03<13:26:56, 11.58s/step, acc=0.5027, loss=2.0420, lr=3.88e-05]
Training:  62%|██████▏   | 6903/11082 [22:11:14<13:15:49, 11.43s/step, acc=0.5027, loss=2.0420, lr=3.88e-05]
Training:  62%|██████▏   | 6904/11082 [22:11:26<13:22:06, 11.52s/step, acc=0.5027, loss=2.0420, lr=3.88e-05]
Training:  62%|██████▏   | 6905/11082 [22:11:38<13:24:04, 11.55s/step, acc=0.5027, loss=2.0420, lr=3.88e-05]
Training:  62%|██████▏   | 6906/11082 [22:11:49<13:27:03, 11.60s/step, acc=0.5027, loss=2.0420, lr=3.88e-05]
Training:  62%|██████▏   | 6907/11082 [22:12:00<13:18:50, 11.48s/step, acc=0.5027, loss=2.0420, lr=3.88e-05]
Training:  62%|██████▏   | 6908/11082 [22:12:12<13:23:23, 11.55s/step, acc=0.5027, loss=2.0420, lr=3.88e-05]
Training:  62%|██████▏   | 6909/11082 [22:12:24<13:21:05, 11.52s/step, acc=0.5027, loss=2.0420, lr=3.88e-05]
Training:  62%|██████▏   | 6910/11082 [22:12:35<13:22:12, 11.54s/step, acc=0.5027, loss=2.0420, lr=3.88e-05]
Training:  62%|██████▏   | 6911/11082 [22:12:47<13:25:33, 11.59s/step, acc=0.5027, loss=2.0420, lr=3.88e-05]
Training:  62%|██████▏   | 6912/11082 [22:12:58<13:19:22, 11.50s/step, acc=0.5027, loss=2.0420, lr=3.88e-05]
Training:  62%|██████▏   | 6913/11082 [22:13:10<13:25:05, 11.59s/step, acc=0.5027, loss=2.0420, lr=3.88e-05]
Training:  62%|██████▏   | 6914/11082 [22:13:21<13:21:14, 11.53s/step, acc=0.5027, loss=2.0420, lr=3.88e-05]
Training:  62%|██████▏   | 6915/11082 [22:13:33<13:22:16, 11.55s/step, acc=0.5027, loss=2.0420, lr=3.88e-05]
Training:  62%|██████▏   | 6916/11082 [22:13:45<13:27:07, 11.62s/step, acc=0.5027, loss=2.0420, lr=3.88e-05]
Training:  62%|██████▏   | 6917/11082 [22:13:56<13:18:10, 11.50s/step, acc=0.5027, loss=2.0420, lr=3.88e-05]
Training:  62%|██████▏   | 6918/11082 [22:14:08<13:24:32, 11.59s/step, acc=0.5027, loss=2.0420, lr=3.88e-05]
Training:  62%|██████▏   | 6919/11082 [22:14:19<13:21:22, 11.55s/step, acc=0.5027, loss=2.0420, lr=3.88e-05]
Training:  62%|██████▏   | 6920/11082 [22:14:31<13:31:57, 11.71s/step, acc=0.5027, loss=2.0420, lr=3.88e-05]
Training:  62%|██████▏   | 6921/11082 [22:14:43<13:23:13, 11.58s/step, acc=0.5027, loss=2.0420, lr=3.88e-05]
Training:  62%|██████▏   | 6922/11082 [22:14:54<13:26:14, 11.63s/step, acc=0.5027, loss=2.0420, lr=3.88e-05]
Training:  62%|██████▏   | 6923/11082 [22:15:05<13:12:46, 11.44s/step, acc=0.5027, loss=2.0420, lr=3.88e-05]
Training:  62%|██████▏   | 6924/11082 [22:15:17<13:23:47, 11.60s/step, acc=0.5027, loss=2.0420, lr=3.88e-05]
Training:  62%|██████▏   | 6925/11082 [22:15:29<13:19:35, 11.54s/step, acc=0.5027, loss=2.0420, lr=3.88e-05]
Training:  62%|██████▏   | 6926/11082 [22:15:40<13:19:20, 11.54s/step, acc=0.5027, loss=2.0420, lr=3.88e-05]
Training:  63%|██████▎   | 6927/11082 [22:15:51<13:12:20, 11.44s/step, acc=0.5027, loss=2.0420, lr=3.88e-05]
Training:  63%|██████▎   | 6928/11082 [22:16:03<13:21:39, 11.58s/step, acc=0.5027, loss=2.0420, lr=3.88e-05]
Training:  63%|██████▎   | 6929/11082 [22:16:15<13:15:51, 11.50s/step, acc=0.5027, loss=2.0420, lr=3.88e-05]
Training:  63%|██████▎   | 6930/11082 [22:16:26<13:17:31, 11.52s/step, acc=0.5027, loss=2.0420, lr=3.88e-05]
Training:  63%|██████▎   | 6931/11082 [22:16:38<13:27:06, 11.67s/step, acc=0.5027, loss=2.0420, lr=3.88e-05]
Training:  63%|██████▎   | 6932/11082 [22:16:50<13:22:38, 11.60s/step, acc=0.5027, loss=2.0420, lr=3.88e-05]
Training:  63%|██████▎   | 6933/11082 [22:17:01<13:26:05, 11.66s/step, acc=0.5027, loss=2.0420, lr=3.88e-05]
Training:  63%|██████▎   | 6934/11082 [22:17:13<13:15:40, 11.51s/step, acc=0.5027, loss=2.0420, lr=3.88e-05]
Training:  63%|██████▎   | 6935/11082 [22:17:24<13:18:44, 11.56s/step, acc=0.5027, loss=2.0420, lr=3.88e-05]
Training:  63%|██████▎   | 6936/11082 [22:17:36<13:18:48, 11.56s/step, acc=0.5027, loss=2.0420, lr=3.88e-05]
Training:  63%|██████▎   | 6937/11082 [22:17:47<13:04:17, 11.35s/step, acc=0.5027, loss=2.0420, lr=3.88e-05]
Training:  63%|██████▎   | 6938/11082 [22:17:59<13:23:28, 11.63s/step, acc=0.5027, loss=2.0420, lr=3.88e-05]
Training:  63%|██████▎   | 6939/11082 [22:18:11<13:21:39, 11.61s/step, acc=0.5027, loss=2.0420, lr=3.88e-05]
Training:  63%|██████▎   | 6940/11082 [22:18:22<13:14:34, 11.51s/step, acc=0.5027, loss=2.0420, lr=3.88e-05]
Training:  63%|██████▎   | 6941/11082 [22:18:34<13:26:41, 11.69s/step, acc=0.5027, loss=2.0420, lr=3.88e-05]
Training:  63%|██████▎   | 6942/11082 [22:18:45<13:13:52, 11.51s/step, acc=0.5027, loss=2.0420, lr=3.88e-05]
Training:  63%|██████▎   | 6943/11082 [22:18:57<13:24:04, 11.66s/step, acc=0.5027, loss=2.0420, lr=3.88e-05]
Training:  63%|██████▎   | 6944/11082 [22:19:08<13:14:32, 11.52s/step, acc=0.5027, loss=2.0420, lr=3.88e-05]
Training:  63%|██████▎   | 6945/11082 [22:19:20<13:23:20, 11.65s/step, acc=0.5027, loss=2.0420, lr=3.88e-05]
Training:  63%|██████▎   | 6946/11082 [22:19:32<13:20:46, 11.62s/step, acc=0.5027, loss=2.0420, lr=3.88e-05]
Training:  63%|██████▎   | 6947/11082 [22:19:43<13:23:07, 11.65s/step, acc=0.5027, loss=2.0420, lr=3.88e-05]
Training:  63%|██████▎   | 6948/11082 [22:19:55<13:23:53, 11.67s/step, acc=0.5027, loss=2.0420, lr=3.88e-05]
Training:  63%|██████▎   | 6949/11082 [22:20:07<13:20:34, 11.62s/step, acc=0.5027, loss=2.0420, lr=3.88e-05]
Training:  63%|██████▎   | 6950/11082 [22:20:18<13:15:40, 11.55s/step, acc=0.5027, loss=2.0420, lr=3.88e-05]
Training:  63%|██████▎   | 6950/11082 [22:20:25<13:15:40, 11.55s/step, acc=0.5057, loss=2.0198, lr=3.82e-05]2026-06-19 13:13:53 | INFO | __main__ | epoch=1 step=6950 train_loss=2.019793 train_acc=0.505734 lr=3.824e-05 grad_norm=0.7812 acc_d0=0.7682 acc_d1=0.6234 acc_d2=0.5021 acc_d3=0.4038 acc_d4=0.3246 acc_d5=0.2611 acc_d6=0.2112 acc_d7=0.1717

Training:  63%|██████▎   | 6951/11082 [22:20:29<13:10:16, 11.48s/step, acc=0.5057, loss=2.0198, lr=3.82e-05]
Training:  63%|██████▎   | 6952/11082 [22:20:41<13:20:56, 11.64s/step, acc=0.5057, loss=2.0198, lr=3.82e-05]
Training:  63%|██████▎   | 6953/11082 [22:20:53<13:23:15, 11.67s/step, acc=0.5057, loss=2.0198, lr=3.82e-05]
Training:  63%|██████▎   | 6954/11082 [22:21:05<13:21:16, 11.65s/step, acc=0.5057, loss=2.0198, lr=3.82e-05]
Training:  63%|██████▎   | 6955/11082 [22:21:16<13:11:03, 11.50s/step, acc=0.5057, loss=2.0198, lr=3.82e-05]
Training:  63%|██████▎   | 6956/11082 [22:21:28<13:13:05, 11.53s/step, acc=0.5057, loss=2.0198, lr=3.82e-05]
Training:  63%|██████▎   | 6957/11082 [22:21:39<13:13:36, 11.54s/step, acc=0.5057, loss=2.0198, lr=3.82e-05]
Training:  63%|██████▎   | 6958/11082 [22:21:51<13:20:57, 11.65s/step, acc=0.5057, loss=2.0198, lr=3.82e-05]
Training:  63%|██████▎   | 6959/11082 [22:22:03<13:21:16, 11.66s/step, acc=0.5057, loss=2.0198, lr=3.82e-05]
Training:  63%|██████▎   | 6960/11082 [22:22:14<13:11:25, 11.52s/step, acc=0.5057, loss=2.0198, lr=3.82e-05]
Training:  63%|██████▎   | 6961/11082 [22:22:26<13:16:24, 11.60s/step, acc=0.5057, loss=2.0198, lr=3.82e-05]
Training:  63%|██████▎   | 6962/11082 [22:22:37<13:14:26, 11.57s/step, acc=0.5057, loss=2.0198, lr=3.82e-05]
Training:  63%|██████▎   | 6963/11082 [22:22:49<13:20:51, 11.67s/step, acc=0.5057, loss=2.0198, lr=3.82e-05]
Training:  63%|██████▎   | 6964/11082 [22:23:00<13:05:00, 11.44s/step, acc=0.5057, loss=2.0198, lr=3.82e-05]
Training:  63%|██████▎   | 6965/11082 [22:23:12<13:12:39, 11.55s/step, acc=0.5057, loss=2.0198, lr=3.82e-05]
Training:  63%|██████▎   | 6966/11082 [22:23:24<13:19:16, 11.65s/step, acc=0.5057, loss=2.0198, lr=3.82e-05]
Training:  63%|██████▎   | 6967/11082 [22:23:35<13:13:34, 11.57s/step, acc=0.5057, loss=2.0198, lr=3.82e-05]
Training:  63%|██████▎   | 6968/11082 [22:23:46<13:09:37, 11.52s/step, acc=0.5057, loss=2.0198, lr=3.82e-05]
Training:  63%|██████▎   | 6969/11082 [22:23:58<13:15:54, 11.61s/step, acc=0.5057, loss=2.0198, lr=3.82e-05]
Training:  63%|██████▎   | 6970/11082 [22:24:10<13:14:25, 11.59s/step, acc=0.5057, loss=2.0198, lr=3.82e-05]
Training:  63%|██████▎   | 6971/11082 [22:24:21<13:14:47, 11.60s/step, acc=0.5057, loss=2.0198, lr=3.82e-05]
Training:  63%|██████▎   | 6972/11082 [22:24:33<13:10:34, 11.54s/step, acc=0.5057, loss=2.0198, lr=3.82e-05]
Training:  63%|██████▎   | 6973/11082 [22:24:45<13:23:31, 11.73s/step, acc=0.5057, loss=2.0198, lr=3.82e-05]
Training:  63%|██████▎   | 6974/11082 [22:24:56<13:16:01, 11.63s/step, acc=0.5057, loss=2.0198, lr=3.82e-05]
Training:  63%|██████▎   | 6975/11082 [22:25:08<13:09:19, 11.53s/step, acc=0.5057, loss=2.0198, lr=3.82e-05]
Training:  63%|██████▎   | 6976/11082 [22:25:20<13:24:02, 11.75s/step, acc=0.5057, loss=2.0198, lr=3.82e-05]
Training:  63%|██████▎   | 6977/11082 [22:25:31<13:15:25, 11.63s/step, acc=0.5057, loss=2.0198, lr=3.82e-05]
Training:  63%|██████▎   | 6978/11082 [22:25:44<13:30:35, 11.85s/step, acc=0.5057, loss=2.0198, lr=3.82e-05]
Training:  63%|██████▎   | 6979/11082 [22:25:55<13:11:51, 11.58s/step, acc=0.5057, loss=2.0198, lr=3.82e-05]
Training:  63%|██████▎   | 6980/11082 [22:26:07<13:20:16, 11.71s/step, acc=0.5057, loss=2.0198, lr=3.82e-05]
Training:  63%|██████▎   | 6981/11082 [22:26:18<13:07:43, 11.52s/step, acc=0.5057, loss=2.0198, lr=3.82e-05]
Training:  63%|██████▎   | 6982/11082 [22:26:29<13:06:34, 11.51s/step, acc=0.5057, loss=2.0198, lr=3.82e-05]
Training:  63%|██████▎   | 6983/11082 [22:26:41<13:11:55, 11.59s/step, acc=0.5057, loss=2.0198, lr=3.82e-05]
Training:  63%|██████▎   | 6984/11082 [22:26:53<13:14:25, 11.63s/step, acc=0.5057, loss=2.0198, lr=3.82e-05]
Training:  63%|██████▎   | 6985/11082 [22:27:04<13:11:59, 11.60s/step, acc=0.5057, loss=2.0198, lr=3.82e-05]
Training:  63%|██████▎   | 6986/11082 [22:27:16<13:09:51, 11.57s/step, acc=0.5057, loss=2.0198, lr=3.82e-05]
Training:  63%|██████▎   | 6987/11082 [22:27:27<13:08:31, 11.55s/step, acc=0.5057, loss=2.0198, lr=3.82e-05]
Training:  63%|██████▎   | 6988/11082 [22:27:39<13:11:28, 11.60s/step, acc=0.5057, loss=2.0198, lr=3.82e-05]
Training:  63%|██████▎   | 6989/11082 [22:27:51<13:15:30, 11.66s/step, acc=0.5057, loss=2.0198, lr=3.82e-05]
Training:  63%|██████▎   | 6990/11082 [22:28:02<13:11:44, 11.61s/step, acc=0.5057, loss=2.0198, lr=3.82e-05]
Training:  63%|██████▎   | 6991/11082 [22:28:14<13:08:27, 11.56s/step, acc=0.5057, loss=2.0198, lr=3.82e-05]
Training:  63%|██████▎   | 6992/11082 [22:28:25<13:09:17, 11.58s/step, acc=0.5057, loss=2.0198, lr=3.82e-05]
Training:  63%|██████▎   | 6993/11082 [22:28:37<13:13:35, 11.64s/step, acc=0.5057, loss=2.0198, lr=3.82e-05]
Training:  63%|██████▎   | 6994/11082 [22:28:49<13:12:48, 11.64s/step, acc=0.5057, loss=2.0198, lr=3.82e-05]
Training:  63%|██████▎   | 6995/11082 [22:29:00<13:02:50, 11.49s/step, acc=0.5057, loss=2.0198, lr=3.82e-05]
Training:  63%|██████▎   | 6996/11082 [22:29:11<13:03:19, 11.50s/step, acc=0.5057, loss=2.0198, lr=3.82e-05]
Training:  63%|██████▎   | 6997/11082 [22:29:23<13:06:36, 11.55s/step, acc=0.5057, loss=2.0198, lr=3.82e-05]
Training:  63%|██████▎   | 6998/11082 [22:29:35<13:08:49, 11.59s/step, acc=0.5057, loss=2.0198, lr=3.82e-05]
Training:  63%|██████▎   | 6999/11082 [22:29:46<13:05:31, 11.54s/step, acc=0.5057, loss=2.0198, lr=3.82e-05]
Training:  63%|██████▎   | 7000/11082 [22:29:58<13:06:24, 11.56s/step, acc=0.5057, loss=2.0198, lr=3.82e-05]
Training:  63%|██████▎   | 7000/11082 [22:30:05<13:06:24, 11.56s/step, acc=0.5012, loss=2.0533, lr=3.76e-05]2026-06-19 13:23:33 | INFO | __main__ | epoch=1 step=7000 train_loss=2.053304 train_acc=0.501168 lr=3.764e-05 grad_norm=0.6875 acc_d0=0.7563 acc_d1=0.6119 acc_d2=0.4900 acc_d3=0.3924 acc_d4=0.3166 acc_d5=0.2554 acc_d6=0.2069 acc_d7=0.1693

Training:  63%|██████▎   | 7001/11082 [22:30:09<13:05:43, 11.55s/step, acc=0.5012, loss=2.0533, lr=3.76e-05]
Training:  63%|██████▎   | 7002/11082 [22:30:21<13:06:45, 11.57s/step, acc=0.5012, loss=2.0533, lr=3.76e-05]
Training:  63%|██████▎   | 7003/11082 [22:30:32<13:03:15, 11.52s/step, acc=0.5012, loss=2.0533, lr=3.76e-05]
Training:  63%|██████▎   | 7004/11082 [22:30:44<13:14:29, 11.69s/step, acc=0.5012, loss=2.0533, lr=3.76e-05]
Training:  63%|██████▎   | 7005/11082 [22:30:56<13:15:10, 11.70s/step, acc=0.5012, loss=2.0533, lr=3.76e-05]
Training:  63%|██████▎   | 7006/11082 [22:31:08<13:09:46, 11.63s/step, acc=0.5012, loss=2.0533, lr=3.76e-05]
Training:  63%|██████▎   | 7007/11082 [22:31:19<13:06:22, 11.58s/step, acc=0.5012, loss=2.0533, lr=3.76e-05]
Training:  63%|██████▎   | 7008/11082 [22:31:30<12:57:28, 11.45s/step, acc=0.5012, loss=2.0533, lr=3.76e-05]
Training:  63%|██████▎   | 7009/11082 [22:31:42<13:01:53, 11.52s/step, acc=0.5012, loss=2.0533, lr=3.76e-05]
Training:  63%|██████▎   | 7010/11082 [22:31:53<13:01:11, 11.51s/step, acc=0.5012, loss=2.0533, lr=3.76e-05]
Training:  63%|██████▎   | 7011/11082 [22:32:05<13:08:00, 11.61s/step, acc=0.5012, loss=2.0533, lr=3.76e-05]
Training:  63%|██████▎   | 7012/11082 [22:32:17<13:05:34, 11.58s/step, acc=0.5012, loss=2.0533, lr=3.76e-05]
Training:  63%|██████▎   | 7013/11082 [22:32:29<13:11:12, 11.67s/step, acc=0.5012, loss=2.0533, lr=3.76e-05]
Training:  63%|██████▎   | 7014/11082 [22:32:41<13:18:37, 11.78s/step, acc=0.5012, loss=2.0533, lr=3.76e-05]
Training:  63%|██████▎   | 7015/11082 [22:32:52<13:15:44, 11.74s/step, acc=0.5012, loss=2.0533, lr=3.76e-05]
Training:  63%|██████▎   | 7016/11082 [22:33:04<13:09:12, 11.65s/step, acc=0.5012, loss=2.0533, lr=3.76e-05]
Training:  63%|██████▎   | 7017/11082 [22:33:15<13:07:02, 11.62s/step, acc=0.5012, loss=2.0533, lr=3.76e-05]
Training:  63%|██████▎   | 7018/11082 [22:33:27<13:06:34, 11.61s/step, acc=0.5012, loss=2.0533, lr=3.76e-05]
Training:  63%|██████▎   | 7019/11082 [22:33:39<13:11:29, 11.69s/step, acc=0.5012, loss=2.0533, lr=3.76e-05]
Training:  63%|██████▎   | 7020/11082 [22:33:50<13:06:35, 11.62s/step, acc=0.5012, loss=2.0533, lr=3.76e-05]
Training:  63%|██████▎   | 7021/11082 [22:34:02<13:05:11, 11.60s/step, acc=0.5012, loss=2.0533, lr=3.76e-05]
Training:  63%|██████▎   | 7022/11082 [22:34:13<13:06:31, 11.62s/step, acc=0.5012, loss=2.0533, lr=3.76e-05]
Training:  63%|██████▎   | 7023/11082 [22:34:25<13:01:21, 11.55s/step, acc=0.5012, loss=2.0533, lr=3.76e-05]
Training:  63%|██████▎   | 7024/11082 [22:34:36<13:01:37, 11.56s/step, acc=0.5012, loss=2.0533, lr=3.76e-05]
Training:  63%|██████▎   | 7025/11082 [22:34:48<13:08:01, 11.65s/step, acc=0.5012, loss=2.0533, lr=3.76e-05]
Training:  63%|██████▎   | 7026/11082 [22:35:00<13:05:54, 11.63s/step, acc=0.5012, loss=2.0533, lr=3.76e-05]
Training:  63%|██████▎   | 7027/11082 [22:35:11<12:57:52, 11.51s/step, acc=0.5012, loss=2.0533, lr=3.76e-05]
Training:  63%|██████▎   | 7028/11082 [22:35:23<13:03:49, 11.60s/step, acc=0.5012, loss=2.0533, lr=3.76e-05]
Training:  63%|██████▎   | 7029/11082 [22:35:35<13:13:58, 11.75s/step, acc=0.5012, loss=2.0533, lr=3.76e-05]
Training:  63%|██████▎   | 7030/11082 [22:35:46<13:07:44, 11.66s/step, acc=0.5012, loss=2.0533, lr=3.76e-05]
Training:  63%|██████▎   | 7031/11082 [22:35:58<13:08:27, 11.68s/step, acc=0.5012, loss=2.0533, lr=3.76e-05]
Training:  63%|██████▎   | 7032/11082 [22:36:09<12:59:44, 11.55s/step, acc=0.5012, loss=2.0533, lr=3.76e-05]
Training:  63%|██████▎   | 7033/11082 [22:36:21<13:08:21, 11.68s/step, acc=0.5012, loss=2.0533, lr=3.76e-05]
Training:  63%|██████▎   | 7034/11082 [22:36:33<13:02:28, 11.60s/step, acc=0.5012, loss=2.0533, lr=3.76e-05]
Training:  63%|██████▎   | 7035/11082 [22:36:45<13:16:27, 11.81s/step, acc=0.5012, loss=2.0533, lr=3.76e-05]
Training:  63%|██████▎   | 7036/11082 [22:36:56<12:59:39, 11.56s/step, acc=0.5012, loss=2.0533, lr=3.76e-05]
Training:  63%|██████▎   | 7037/11082 [22:37:08<13:02:15, 11.60s/step, acc=0.5012, loss=2.0533, lr=3.76e-05]
Training:  64%|██████▎   | 7038/11082 [22:37:19<13:01:18, 11.59s/step, acc=0.5012, loss=2.0533, lr=3.76e-05]
Training:  64%|██████▎   | 7039/11082 [22:37:31<12:58:04, 11.55s/step, acc=0.5012, loss=2.0533, lr=3.76e-05]
Training:  64%|██████▎   | 7040/11082 [22:37:43<13:06:42, 11.68s/step, acc=0.5012, loss=2.0533, lr=3.76e-05]
Training:  64%|██████▎   | 7041/11082 [22:37:54<13:01:03, 11.60s/step, acc=0.5012, loss=2.0533, lr=3.76e-05]
Training:  64%|██████▎   | 7042/11082 [22:38:06<12:59:04, 11.57s/step, acc=0.5012, loss=2.0533, lr=3.76e-05]
Training:  64%|██████▎   | 7043/11082 [22:38:17<12:56:43, 11.54s/step, acc=0.5012, loss=2.0533, lr=3.76e-05]
Training:  64%|██████▎   | 7044/11082 [22:38:29<12:56:37, 11.54s/step, acc=0.5012, loss=2.0533, lr=3.76e-05]
Training:  64%|██████▎   | 7045/11082 [22:38:40<12:57:58, 11.56s/step, acc=0.5012, loss=2.0533, lr=3.76e-05]
Training:  64%|██████▎   | 7046/11082 [22:38:52<12:53:48, 11.50s/step, acc=0.5012, loss=2.0533, lr=3.76e-05]
Training:  64%|██████▎   | 7047/11082 [22:39:03<12:58:16, 11.57s/step, acc=0.5012, loss=2.0533, lr=3.76e-05]
Training:  64%|██████▎   | 7048/11082 [22:39:15<12:50:53, 11.47s/step, acc=0.5012, loss=2.0533, lr=3.76e-05]
Training:  64%|██████▎   | 7049/11082 [22:39:27<13:06:27, 11.70s/step, acc=0.5012, loss=2.0533, lr=3.76e-05]
Training:  64%|██████▎   | 7050/11082 [22:39:38<12:57:24, 11.57s/step, acc=0.5012, loss=2.0533, lr=3.76e-05]
Training:  64%|██████▎   | 7050/11082 [22:39:45<12:57:24, 11.57s/step, acc=0.5057, loss=2.0222, lr=3.70e-05]2026-06-19 13:33:13 | INFO | __main__ | epoch=1 step=7050 train_loss=2.022230 train_acc=0.505714 lr=3.705e-05 grad_norm=0.6367 acc_d0=0.7652 acc_d1=0.6230 acc_d2=0.5032 acc_d3=0.4046 acc_d4=0.3273 acc_d5=0.2656 acc_d6=0.2164 acc_d7=0.1782

Training:  64%|██████▎   | 7051/11082 [22:39:49<12:50:12, 11.46s/step, acc=0.5057, loss=2.0222, lr=3.70e-05]
Training:  64%|██████▎   | 7052/11082 [22:40:01<12:53:57, 11.52s/step, acc=0.5057, loss=2.0222, lr=3.70e-05]
Training:  64%|██████▎   | 7053/11082 [22:40:13<13:03:17, 11.66s/step, acc=0.5057, loss=2.0222, lr=3.70e-05]
Training:  64%|██████▎   | 7054/11082 [22:40:24<12:53:52, 11.53s/step, acc=0.5057, loss=2.0222, lr=3.70e-05]
Training:  64%|██████▎   | 7055/11082 [22:40:35<12:46:08, 11.42s/step, acc=0.5057, loss=2.0222, lr=3.70e-05]
Training:  64%|██████▎   | 7056/11082 [22:40:47<13:00:12, 11.63s/step, acc=0.5057, loss=2.0222, lr=3.70e-05]
Training:  64%|██████▎   | 7057/11082 [22:40:59<12:52:21, 11.51s/step, acc=0.5057, loss=2.0222, lr=3.70e-05]
Training:  64%|██████▎   | 7058/11082 [22:41:10<12:53:45, 11.54s/step, acc=0.5057, loss=2.0222, lr=3.70e-05]
Training:  64%|██████▎   | 7059/11082 [22:41:22<12:55:59, 11.57s/step, acc=0.5057, loss=2.0222, lr=3.70e-05]
Training:  64%|██████▎   | 7060/11082 [22:41:34<13:02:01, 11.67s/step, acc=0.5057, loss=2.0222, lr=3.70e-05]
Training:  64%|██████▎   | 7061/11082 [22:41:45<13:00:56, 11.65s/step, acc=0.5057, loss=2.0222, lr=3.70e-05]
Training:  64%|██████▎   | 7062/11082 [22:41:57<12:57:54, 11.61s/step, acc=0.5057, loss=2.0222, lr=3.70e-05]
Training:  64%|██████▎   | 7063/11082 [22:42:08<12:55:21, 11.58s/step, acc=0.5057, loss=2.0222, lr=3.70e-05]
Training:  64%|██████▎   | 7064/11082 [22:42:20<12:52:32, 11.54s/step, acc=0.5057, loss=2.0222, lr=3.70e-05]
Training:  64%|██████▍   | 7065/11082 [22:42:32<13:00:18, 11.66s/step, acc=0.5057, loss=2.0222, lr=3.70e-05]
Training:  64%|██████▍   | 7066/11082 [22:42:43<12:55:33, 11.59s/step, acc=0.5057, loss=2.0222, lr=3.70e-05]
Training:  64%|██████▍   | 7067/11082 [22:42:55<12:58:10, 11.63s/step, acc=0.5057, loss=2.0222, lr=3.70e-05]
Training:  64%|██████▍   | 7068/11082 [22:43:07<12:55:18, 11.59s/step, acc=0.5057, loss=2.0222, lr=3.70e-05]
Training:  64%|██████▍   | 7069/11082 [22:43:18<12:58:49, 11.64s/step, acc=0.5057, loss=2.0222, lr=3.70e-05]
Training:  64%|██████▍   | 7070/11082 [22:43:29<12:48:20, 11.49s/step, acc=0.5057, loss=2.0222, lr=3.70e-05]
Training:  64%|██████▍   | 7071/11082 [22:43:41<12:54:26, 11.58s/step, acc=0.5057, loss=2.0222, lr=3.70e-05]
Training:  64%|██████▍   | 7072/11082 [22:43:52<12:45:04, 11.45s/step, acc=0.5057, loss=2.0222, lr=3.70e-05]
Training:  64%|██████▍   | 7073/11082 [22:44:04<12:53:19, 11.57s/step, acc=0.5057, loss=2.0222, lr=3.70e-05]
Training:  64%|██████▍   | 7074/11082 [22:44:16<12:50:58, 11.54s/step, acc=0.5057, loss=2.0222, lr=3.70e-05]
Training:  64%|██████▍   | 7075/11082 [22:44:27<12:50:58, 11.54s/step, acc=0.5057, loss=2.0222, lr=3.70e-05]
Training:  64%|██████▍   | 7076/11082 [22:44:39<12:50:52, 11.55s/step, acc=0.5057, loss=2.0222, lr=3.70e-05]
Training:  64%|██████▍   | 7077/11082 [22:44:50<12:50:44, 11.55s/step, acc=0.5057, loss=2.0222, lr=3.70e-05]
Training:  64%|██████▍   | 7078/11082 [22:45:02<12:44:00, 11.45s/step, acc=0.5057, loss=2.0222, lr=3.70e-05]
Training:  64%|██████▍   | 7079/11082 [22:45:13<12:50:17, 11.55s/step, acc=0.5057, loss=2.0222, lr=3.70e-05]
Training:  64%|██████▍   | 7080/11082 [22:45:25<12:53:30, 11.60s/step, acc=0.5057, loss=2.0222, lr=3.70e-05]
Training:  64%|██████▍   | 7081/11082 [22:45:36<12:47:23, 11.51s/step, acc=0.5057, loss=2.0222, lr=3.70e-05]
Training:  64%|██████▍   | 7082/11082 [22:45:48<12:46:46, 11.50s/step, acc=0.5057, loss=2.0222, lr=3.70e-05]
Training:  64%|██████▍   | 7083/11082 [22:45:59<12:47:21, 11.51s/step, acc=0.5057, loss=2.0222, lr=3.70e-05]
Training:  64%|██████▍   | 7084/11082 [22:46:11<12:45:44, 11.49s/step, acc=0.5057, loss=2.0222, lr=3.70e-05]
Training:  64%|██████▍   | 7085/11082 [22:46:22<12:41:50, 11.44s/step, acc=0.5057, loss=2.0222, lr=3.70e-05]
Training:  64%|██████▍   | 7086/11082 [22:46:34<12:45:20, 11.49s/step, acc=0.5057, loss=2.0222, lr=3.70e-05]
Training:  64%|██████▍   | 7087/11082 [22:46:45<12:44:08, 11.48s/step, acc=0.5057, loss=2.0222, lr=3.70e-05]
Training:  64%|██████▍   | 7088/11082 [22:46:57<12:44:27, 11.48s/step, acc=0.5057, loss=2.0222, lr=3.70e-05]
Training:  64%|██████▍   | 7089/11082 [22:47:08<12:49:05, 11.56s/step, acc=0.5057, loss=2.0222, lr=3.70e-05]
Training:  64%|██████▍   | 7090/11082 [22:47:19<12:37:52, 11.39s/step, acc=0.5057, loss=2.0222, lr=3.70e-05]
Training:  64%|██████▍   | 7091/11082 [22:47:31<12:48:40, 11.56s/step, acc=0.5057, loss=2.0222, lr=3.70e-05]
Training:  64%|██████▍   | 7092/11082 [22:47:43<12:44:53, 11.50s/step, acc=0.5057, loss=2.0222, lr=3.70e-05]
Training:  64%|██████▍   | 7093/11082 [22:47:55<12:51:58, 11.61s/step, acc=0.5057, loss=2.0222, lr=3.70e-05]
Training:  64%|██████▍   | 7094/11082 [22:48:06<12:51:28, 11.61s/step, acc=0.5057, loss=2.0222, lr=3.70e-05]
Training:  64%|██████▍   | 7095/11082 [22:48:18<13:02:38, 11.78s/step, acc=0.5057, loss=2.0222, lr=3.70e-05]
Training:  64%|██████▍   | 7096/11082 [22:48:30<12:53:50, 11.65s/step, acc=0.5057, loss=2.0222, lr=3.70e-05]
Training:  64%|██████▍   | 7097/11082 [22:48:41<12:51:21, 11.61s/step, acc=0.5057, loss=2.0222, lr=3.70e-05]
Training:  64%|██████▍   | 7098/11082 [22:48:52<12:43:15, 11.49s/step, acc=0.5057, loss=2.0222, lr=3.70e-05]
Training:  64%|██████▍   | 7099/11082 [22:49:04<12:47:29, 11.56s/step, acc=0.5057, loss=2.0222, lr=3.70e-05]
Training:  64%|██████▍   | 7100/11082 [22:49:16<12:54:28, 11.67s/step, acc=0.5057, loss=2.0222, lr=3.70e-05]
Training:  64%|██████▍   | 7100/11082 [22:49:23<12:54:28, 11.67s/step, acc=0.5044, loss=2.0324, lr=3.65e-05]2026-06-19 13:42:51 | INFO | __main__ | epoch=1 step=7100 train_loss=2.032363 train_acc=0.504446 lr=3.645e-05 grad_norm=0.6953 acc_d0=0.7492 acc_d1=0.6050 acc_d2=0.4847 acc_d3=0.3879 acc_d4=0.3099 acc_d5=0.2478 acc_d6=0.2005 acc_d7=0.1632

Training:  64%|██████▍   | 7101/11082 [22:49:28<12:50:25, 11.61s/step, acc=0.5044, loss=2.0324, lr=3.65e-05]
Training:  64%|██████▍   | 7102/11082 [22:49:40<12:56:53, 11.71s/step, acc=0.5044, loss=2.0324, lr=3.65e-05]
Training:  64%|██████▍   | 7103/11082 [22:49:51<12:54:45, 11.68s/step, acc=0.5044, loss=2.0324, lr=3.65e-05]
Training:  64%|██████▍   | 7104/11082 [22:50:03<12:53:46, 11.67s/step, acc=0.5044, loss=2.0324, lr=3.65e-05]
Training:  64%|██████▍   | 7105/11082 [22:50:14<12:48:01, 11.59s/step, acc=0.5044, loss=2.0324, lr=3.65e-05]
Training:  64%|██████▍   | 7106/11082 [22:50:26<12:56:06, 11.71s/step, acc=0.5044, loss=2.0324, lr=3.65e-05]
Training:  64%|██████▍   | 7107/11082 [22:50:38<12:55:18, 11.70s/step, acc=0.5044, loss=2.0324, lr=3.65e-05]
Training:  64%|██████▍   | 7108/11082 [22:50:49<12:47:19, 11.59s/step, acc=0.5044, loss=2.0324, lr=3.65e-05]
Training:  64%|██████▍   | 7109/11082 [22:51:00<12:40:27, 11.48s/step, acc=0.5044, loss=2.0324, lr=3.65e-05]
Training:  64%|██████▍   | 7110/11082 [22:51:12<12:43:01, 11.53s/step, acc=0.5044, loss=2.0324, lr=3.65e-05]
Training:  64%|██████▍   | 7111/11082 [22:51:24<12:46:54, 11.59s/step, acc=0.5044, loss=2.0324, lr=3.65e-05]
Training:  64%|██████▍   | 7112/11082 [22:51:35<12:46:07, 11.58s/step, acc=0.5044, loss=2.0324, lr=3.65e-05]
Training:  64%|██████▍   | 7113/11082 [22:51:46<12:32:56, 11.38s/step, acc=0.5044, loss=2.0324, lr=3.65e-05]
Training:  64%|██████▍   | 7114/11082 [22:51:58<12:39:07, 11.48s/step, acc=0.5044, loss=2.0324, lr=3.65e-05]
Training:  64%|██████▍   | 7115/11082 [22:52:10<12:42:50, 11.54s/step, acc=0.5044, loss=2.0324, lr=3.65e-05]
Training:  64%|██████▍   | 7116/11082 [22:52:21<12:48:14, 11.62s/step, acc=0.5044, loss=2.0324, lr=3.65e-05]
Training:  64%|██████▍   | 7117/11082 [22:52:33<12:49:31, 11.64s/step, acc=0.5044, loss=2.0324, lr=3.65e-05]
Training:  64%|██████▍   | 7118/11082 [22:52:44<12:40:54, 11.52s/step, acc=0.5044, loss=2.0324, lr=3.65e-05]
Training:  64%|██████▍   | 7119/11082 [22:52:56<12:40:18, 11.51s/step, acc=0.5044, loss=2.0324, lr=3.65e-05]
Training:  64%|██████▍   | 7120/11082 [22:53:08<12:47:53, 11.63s/step, acc=0.5044, loss=2.0324, lr=3.65e-05]
Training:  64%|██████▍   | 7121/11082 [22:53:19<12:47:20, 11.62s/step, acc=0.5044, loss=2.0324, lr=3.65e-05]
Training:  64%|██████▍   | 7122/11082 [22:53:31<12:45:15, 11.59s/step, acc=0.5044, loss=2.0324, lr=3.65e-05]
Training:  64%|██████▍   | 7123/11082 [22:53:42<12:44:25, 11.59s/step, acc=0.5044, loss=2.0324, lr=3.65e-05]
Training:  64%|██████▍   | 7124/11082 [22:53:54<12:46:56, 11.63s/step, acc=0.5044, loss=2.0324, lr=3.65e-05]
Training:  64%|██████▍   | 7125/11082 [22:54:06<12:40:33, 11.53s/step, acc=0.5044, loss=2.0324, lr=3.65e-05]
Training:  64%|██████▍   | 7126/11082 [22:54:17<12:38:56, 11.51s/step, acc=0.5044, loss=2.0324, lr=3.65e-05]
Training:  64%|██████▍   | 7127/11082 [22:54:29<12:41:49, 11.56s/step, acc=0.5044, loss=2.0324, lr=3.65e-05]
Training:  64%|██████▍   | 7128/11082 [22:54:40<12:35:04, 11.46s/step, acc=0.5044, loss=2.0324, lr=3.65e-05]
Training:  64%|██████▍   | 7129/11082 [22:54:52<12:44:55, 11.61s/step, acc=0.5044, loss=2.0324, lr=3.65e-05]
Training:  64%|██████▍   | 7130/11082 [22:55:03<12:43:04, 11.59s/step, acc=0.5044, loss=2.0324, lr=3.65e-05]
Training:  64%|██████▍   | 7131/11082 [22:55:15<12:45:12, 11.62s/step, acc=0.5044, loss=2.0324, lr=3.65e-05]
Training:  64%|██████▍   | 7132/11082 [22:55:26<12:41:07, 11.56s/step, acc=0.5044, loss=2.0324, lr=3.65e-05]
Training:  64%|██████▍   | 7133/11082 [22:55:38<12:44:21, 11.61s/step, acc=0.5044, loss=2.0324, lr=3.65e-05]
Training:  64%|██████▍   | 7134/11082 [22:55:50<12:45:36, 11.64s/step, acc=0.5044, loss=2.0324, lr=3.65e-05]
Training:  64%|██████▍   | 7135/11082 [22:56:01<12:36:09, 11.49s/step, acc=0.5044, loss=2.0324, lr=3.65e-05]
Training:  64%|██████▍   | 7136/11082 [22:56:13<12:45:11, 11.63s/step, acc=0.5044, loss=2.0324, lr=3.65e-05]
Training:  64%|██████▍   | 7137/11082 [22:56:24<12:37:52, 11.53s/step, acc=0.5044, loss=2.0324, lr=3.65e-05]
Training:  64%|██████▍   | 7138/11082 [22:56:36<12:40:48, 11.57s/step, acc=0.5044, loss=2.0324, lr=3.65e-05]
Training:  64%|██████▍   | 7139/11082 [22:56:47<12:38:49, 11.55s/step, acc=0.5044, loss=2.0324, lr=3.65e-05]
Training:  64%|██████▍   | 7140/11082 [22:56:59<12:37:12, 11.53s/step, acc=0.5044, loss=2.0324, lr=3.65e-05]
Training:  64%|██████▍   | 7141/11082 [22:57:11<12:43:23, 11.62s/step, acc=0.5044, loss=2.0324, lr=3.65e-05]
Training:  64%|██████▍   | 7142/11082 [22:57:22<12:36:07, 11.51s/step, acc=0.5044, loss=2.0324, lr=3.65e-05]
Training:  64%|██████▍   | 7143/11082 [22:57:34<12:40:46, 11.59s/step, acc=0.5044, loss=2.0324, lr=3.65e-05]
Training:  64%|██████▍   | 7144/11082 [22:57:45<12:41:11, 11.60s/step, acc=0.5044, loss=2.0324, lr=3.65e-05]
Training:  64%|██████▍   | 7145/11082 [22:57:57<12:40:50, 11.60s/step, acc=0.5044, loss=2.0324, lr=3.65e-05]
Training:  64%|██████▍   | 7146/11082 [22:58:08<12:33:05, 11.48s/step, acc=0.5044, loss=2.0324, lr=3.65e-05]
Training:  64%|██████▍   | 7147/11082 [22:58:20<12:46:15, 11.68s/step, acc=0.5044, loss=2.0324, lr=3.65e-05]
Training:  65%|██████▍   | 7148/11082 [22:58:32<12:41:11, 11.61s/step, acc=0.5044, loss=2.0324, lr=3.65e-05]
Training:  65%|██████▍   | 7149/11082 [22:58:43<12:37:01, 11.55s/step, acc=0.5044, loss=2.0324, lr=3.65e-05]
Training:  65%|██████▍   | 7150/11082 [22:58:55<12:40:56, 11.61s/step, acc=0.5044, loss=2.0324, lr=3.65e-05]
Training:  65%|██████▍   | 7150/11082 [22:59:02<12:40:56, 11.61s/step, acc=0.5044, loss=2.0282, lr=3.59e-05]2026-06-19 13:52:30 | INFO | __main__ | epoch=1 step=7150 train_loss=2.028162 train_acc=0.504445 lr=3.587e-05 grad_norm=0.6836 acc_d0=0.7548 acc_d1=0.6080 acc_d2=0.4868 acc_d3=0.3886 acc_d4=0.3111 acc_d5=0.2506 acc_d6=0.2031 acc_d7=0.1658

Training:  65%|██████▍   | 7151/11082 [22:59:06<12:33:20, 11.50s/step, acc=0.5044, loss=2.0282, lr=3.59e-05]
Training:  65%|██████▍   | 7152/11082 [22:59:18<12:29:25, 11.44s/step, acc=0.5044, loss=2.0282, lr=3.59e-05]
Training:  65%|██████▍   | 7153/11082 [22:59:29<12:35:47, 11.54s/step, acc=0.5044, loss=2.0282, lr=3.59e-05]
Training:  65%|██████▍   | 7154/11082 [22:59:41<12:35:05, 11.53s/step, acc=0.5044, loss=2.0282, lr=3.59e-05]
Training:  65%|██████▍   | 7155/11082 [22:59:53<12:41:07, 11.63s/step, acc=0.5044, loss=2.0282, lr=3.59e-05]
Training:  65%|██████▍   | 7156/11082 [23:00:04<12:34:09, 11.53s/step, acc=0.5044, loss=2.0282, lr=3.59e-05]
Training:  65%|██████▍   | 7157/11082 [23:00:15<12:33:15, 11.51s/step, acc=0.5044, loss=2.0282, lr=3.59e-05]
Training:  65%|██████▍   | 7158/11082 [23:00:27<12:26:49, 11.42s/step, acc=0.5044, loss=2.0282, lr=3.59e-05]
Training:  65%|██████▍   | 7159/11082 [23:00:38<12:22:46, 11.36s/step, acc=0.5044, loss=2.0282, lr=3.59e-05]
Training:  65%|██████▍   | 7160/11082 [23:00:50<12:43:38, 11.68s/step, acc=0.5044, loss=2.0282, lr=3.59e-05]
Training:  65%|██████▍   | 7161/11082 [23:01:02<12:33:58, 11.54s/step, acc=0.5044, loss=2.0282, lr=3.59e-05]
Training:  65%|██████▍   | 7162/11082 [23:01:13<12:23:00, 11.37s/step, acc=0.5044, loss=2.0282, lr=3.59e-05]
Training:  65%|██████▍   | 7163/11082 [23:01:25<12:43:13, 11.69s/step, acc=0.5044, loss=2.0282, lr=3.59e-05]
Training:  65%|██████▍   | 7164/11082 [23:01:36<12:39:07, 11.63s/step, acc=0.5044, loss=2.0282, lr=3.59e-05]
Training:  65%|██████▍   | 7165/11082 [23:01:48<12:33:35, 11.54s/step, acc=0.5044, loss=2.0282, lr=3.59e-05]
Training:  65%|██████▍   | 7166/11082 [23:01:59<12:33:17, 11.54s/step, acc=0.5044, loss=2.0282, lr=3.59e-05]
Training:  65%|██████▍   | 7167/11082 [23:02:11<12:32:17, 11.53s/step, acc=0.5044, loss=2.0282, lr=3.59e-05]
Training:  65%|██████▍   | 7168/11082 [23:02:22<12:32:52, 11.54s/step, acc=0.5044, loss=2.0282, lr=3.59e-05]
Training:  65%|██████▍   | 7169/11082 [23:02:34<12:41:14, 11.67s/step, acc=0.5044, loss=2.0282, lr=3.59e-05]
Training:  65%|██████▍   | 7170/11082 [23:02:46<12:31:39, 11.53s/step, acc=0.5044, loss=2.0282, lr=3.59e-05]
Training:  65%|██████▍   | 7171/11082 [23:02:57<12:32:42, 11.55s/step, acc=0.5044, loss=2.0282, lr=3.59e-05]
Training:  65%|██████▍   | 7172/11082 [23:03:09<12:35:43, 11.60s/step, acc=0.5044, loss=2.0282, lr=3.59e-05]
Training:  65%|██████▍   | 7173/11082 [23:03:20<12:29:31, 11.50s/step, acc=0.5044, loss=2.0282, lr=3.59e-05]
Training:  65%|██████▍   | 7174/11082 [23:03:32<12:28:45, 11.50s/step, acc=0.5044, loss=2.0282, lr=3.59e-05]
Training:  65%|██████▍   | 7175/11082 [23:03:43<12:26:57, 11.47s/step, acc=0.5044, loss=2.0282, lr=3.59e-05]
Training:  65%|██████▍   | 7176/11082 [23:03:55<12:34:23, 11.59s/step, acc=0.5044, loss=2.0282, lr=3.59e-05]
Training:  65%|██████▍   | 7177/11082 [23:04:06<12:32:44, 11.57s/step, acc=0.5044, loss=2.0282, lr=3.59e-05]
Training:  65%|██████▍   | 7178/11082 [23:04:18<12:34:34, 11.60s/step, acc=0.5044, loss=2.0282, lr=3.59e-05]
Training:  65%|██████▍   | 7179/11082 [23:04:30<12:36:58, 11.64s/step, acc=0.5044, loss=2.0282, lr=3.59e-05]
Training:  65%|██████▍   | 7180/11082 [23:04:41<12:37:44, 11.65s/step, acc=0.5044, loss=2.0282, lr=3.59e-05]
Training:  65%|██████▍   | 7181/11082 [23:04:53<12:33:04, 11.58s/step, acc=0.5044, loss=2.0282, lr=3.59e-05]
Training:  65%|██████▍   | 7182/11082 [23:05:05<12:38:53, 11.68s/step, acc=0.5044, loss=2.0282, lr=3.59e-05]
Training:  65%|██████▍   | 7183/11082 [23:05:16<12:36:47, 11.65s/step, acc=0.5044, loss=2.0282, lr=3.59e-05]
Training:  65%|██████▍   | 7184/11082 [23:05:28<12:36:51, 11.65s/step, acc=0.5044, loss=2.0282, lr=3.59e-05]
Training:  65%|██████▍   | 7185/11082 [23:05:40<12:34:55, 11.62s/step, acc=0.5044, loss=2.0282, lr=3.59e-05]
Training:  65%|██████▍   | 7186/11082 [23:05:51<12:36:06, 11.64s/step, acc=0.5044, loss=2.0282, lr=3.59e-05]
Training:  65%|██████▍   | 7187/11082 [23:06:03<12:43:11, 11.76s/step, acc=0.5044, loss=2.0282, lr=3.59e-05]
Training:  65%|██████▍   | 7188/11082 [23:06:14<12:28:10, 11.53s/step, acc=0.5044, loss=2.0282, lr=3.59e-05]
Training:  65%|██████▍   | 7189/11082 [23:06:26<12:27:07, 11.51s/step, acc=0.5044, loss=2.0282, lr=3.59e-05]
Training:  65%|██████▍   | 7190/11082 [23:06:38<12:33:41, 11.62s/step, acc=0.5044, loss=2.0282, lr=3.59e-05]
Training:  65%|██████▍   | 7191/11082 [23:06:50<12:39:33, 11.71s/step, acc=0.5044, loss=2.0282, lr=3.59e-05]
Training:  65%|██████▍   | 7192/11082 [23:07:01<12:31:55, 11.60s/step, acc=0.5044, loss=2.0282, lr=3.59e-05]
Training:  65%|██████▍   | 7193/11082 [23:07:12<12:30:09, 11.57s/step, acc=0.5044, loss=2.0282, lr=3.59e-05]
Training:  65%|██████▍   | 7194/11082 [23:07:24<12:27:57, 11.54s/step, acc=0.5044, loss=2.0282, lr=3.59e-05]
Training:  65%|██████▍   | 7195/11082 [23:07:36<12:37:25, 11.69s/step, acc=0.5044, loss=2.0282, lr=3.59e-05]
Training:  65%|██████▍   | 7196/11082 [23:07:47<12:20:08, 11.43s/step, acc=0.5044, loss=2.0282, lr=3.59e-05]
Training:  65%|██████▍   | 7197/11082 [23:07:59<12:33:35, 11.64s/step, acc=0.5044, loss=2.0282, lr=3.59e-05]
Training:  65%|██████▍   | 7198/11082 [23:08:10<12:32:26, 11.62s/step, acc=0.5044, loss=2.0282, lr=3.59e-05]
Training:  65%|██████▍   | 7199/11082 [23:08:22<12:29:04, 11.57s/step, acc=0.5044, loss=2.0282, lr=3.59e-05]
Training:  65%|██████▍   | 7200/11082 [23:08:33<12:27:07, 11.55s/step, acc=0.5044, loss=2.0282, lr=3.59e-05]
Training:  65%|██████▍   | 7200/11082 [23:08:41<12:27:07, 11.55s/step, acc=0.5060, loss=2.0202, lr=3.53e-05]2026-06-19 14:02:08 | INFO | __main__ | epoch=1 step=7200 train_loss=2.020195 train_acc=0.506037 lr=3.528e-05 grad_norm=0.7383 acc_d0=0.7702 acc_d1=0.6267 acc_d2=0.5034 acc_d3=0.4032 acc_d4=0.3259 acc_d5=0.2615 acc_d6=0.2123 acc_d7=0.1732

Training:  65%|██████▍   | 7201/11082 [23:08:45<12:27:37, 11.56s/step, acc=0.5060, loss=2.0202, lr=3.53e-05]
Training:  65%|██████▍   | 7202/11082 [23:08:57<12:29:19, 11.59s/step, acc=0.5060, loss=2.0202, lr=3.53e-05]
Training:  65%|██████▍   | 7203/11082 [23:09:08<12:25:21, 11.53s/step, acc=0.5060, loss=2.0202, lr=3.53e-05]
Training:  65%|██████▌   | 7204/11082 [23:09:20<12:30:43, 11.62s/step, acc=0.5060, loss=2.0202, lr=3.53e-05]
Training:  65%|██████▌   | 7205/11082 [23:09:31<12:20:46, 11.46s/step, acc=0.5060, loss=2.0202, lr=3.53e-05]
Training:  65%|██████▌   | 7206/11082 [23:09:43<12:23:03, 11.50s/step, acc=0.5060, loss=2.0202, lr=3.53e-05]
Training:  65%|██████▌   | 7207/11082 [23:09:54<12:17:01, 11.41s/step, acc=0.5060, loss=2.0202, lr=3.53e-05]
Training:  65%|██████▌   | 7208/11082 [23:10:06<12:29:00, 11.60s/step, acc=0.5060, loss=2.0202, lr=3.53e-05]
Training:  65%|██████▌   | 7209/11082 [23:10:17<12:25:44, 11.55s/step, acc=0.5060, loss=2.0202, lr=3.53e-05]
Training:  65%|██████▌   | 7210/11082 [23:10:29<12:29:08, 11.61s/step, acc=0.5060, loss=2.0202, lr=3.53e-05]
Training:  65%|██████▌   | 7211/11082 [23:10:41<12:28:31, 11.60s/step, acc=0.5060, loss=2.0202, lr=3.53e-05]
Training:  65%|██████▌   | 7212/11082 [23:10:52<12:31:02, 11.64s/step, acc=0.5060, loss=2.0202, lr=3.53e-05]
Training:  65%|██████▌   | 7213/11082 [23:11:04<12:23:35, 11.53s/step, acc=0.5060, loss=2.0202, lr=3.53e-05]
Training:  65%|██████▌   | 7214/11082 [23:11:15<12:30:32, 11.64s/step, acc=0.5060, loss=2.0202, lr=3.53e-05]
Training:  65%|██████▌   | 7215/11082 [23:11:27<12:25:08, 11.56s/step, acc=0.5060, loss=2.0202, lr=3.53e-05]
Training:  65%|██████▌   | 7216/11082 [23:11:38<12:22:52, 11.53s/step, acc=0.5060, loss=2.0202, lr=3.53e-05]
Training:  65%|██████▌   | 7217/11082 [23:11:50<12:20:23, 11.49s/step, acc=0.5060, loss=2.0202, lr=3.53e-05]
Training:  65%|██████▌   | 7218/11082 [23:12:01<12:13:22, 11.39s/step, acc=0.5060, loss=2.0202, lr=3.53e-05]
Training:  65%|██████▌   | 7219/11082 [23:12:13<12:23:23, 11.55s/step, acc=0.5060, loss=2.0202, lr=3.53e-05]
Training:  65%|██████▌   | 7220/11082 [23:12:24<12:23:41, 11.55s/step, acc=0.5060, loss=2.0202, lr=3.53e-05]
Training:  65%|██████▌   | 7221/11082 [23:12:36<12:27:28, 11.62s/step, acc=0.5060, loss=2.0202, lr=3.53e-05]
Training:  65%|██████▌   | 7222/11082 [23:12:48<12:26:48, 11.61s/step, acc=0.5060, loss=2.0202, lr=3.53e-05]
Training:  65%|██████▌   | 7223/11082 [23:12:59<12:21:45, 11.53s/step, acc=0.5060, loss=2.0202, lr=3.53e-05]
Training:  65%|██████▌   | 7224/11082 [23:13:11<12:21:36, 11.53s/step, acc=0.5060, loss=2.0202, lr=3.53e-05]
Training:  65%|██████▌   | 7225/11082 [23:13:23<12:29:07, 11.65s/step, acc=0.5060, loss=2.0202, lr=3.53e-05]
Training:  65%|██████▌   | 7226/11082 [23:13:34<12:25:48, 11.60s/step, acc=0.5060, loss=2.0202, lr=3.53e-05]
Training:  65%|██████▌   | 7227/11082 [23:13:46<12:28:36, 11.65s/step, acc=0.5060, loss=2.0202, lr=3.53e-05]
Training:  65%|██████▌   | 7228/11082 [23:13:57<12:26:04, 11.62s/step, acc=0.5060, loss=2.0202, lr=3.53e-05]
Training:  65%|██████▌   | 7229/11082 [23:14:09<12:26:56, 11.63s/step, acc=0.5060, loss=2.0202, lr=3.53e-05]
Training:  65%|██████▌   | 7230/11082 [23:14:20<12:20:26, 11.53s/step, acc=0.5060, loss=2.0202, lr=3.53e-05]
Training:  65%|██████▌   | 7231/11082 [23:14:32<12:16:55, 11.48s/step, acc=0.5060, loss=2.0202, lr=3.53e-05]
Training:  65%|██████▌   | 7232/11082 [23:14:43<12:15:06, 11.46s/step, acc=0.5060, loss=2.0202, lr=3.53e-05]
Training:  65%|██████▌   | 7233/11082 [23:14:55<12:28:14, 11.66s/step, acc=0.5060, loss=2.0202, lr=3.53e-05]
Training:  65%|██████▌   | 7234/11082 [23:15:06<12:19:42, 11.53s/step, acc=0.5060, loss=2.0202, lr=3.53e-05]
Training:  65%|██████▌   | 7235/11082 [23:15:18<12:22:56, 11.59s/step, acc=0.5060, loss=2.0202, lr=3.53e-05]
Training:  65%|██████▌   | 7236/11082 [23:15:29<12:13:29, 11.44s/step, acc=0.5060, loss=2.0202, lr=3.53e-05]
Training:  65%|██████▌   | 7237/11082 [23:15:41<12:25:09, 11.63s/step, acc=0.5060, loss=2.0202, lr=3.53e-05]
Training:  65%|██████▌   | 7238/11082 [23:15:53<12:23:28, 11.60s/step, acc=0.5060, loss=2.0202, lr=3.53e-05]
Training:  65%|██████▌   | 7239/11082 [23:16:04<12:18:55, 11.54s/step, acc=0.5060, loss=2.0202, lr=3.53e-05]
Training:  65%|██████▌   | 7240/11082 [23:16:16<12:21:09, 11.57s/step, acc=0.5060, loss=2.0202, lr=3.53e-05]
Training:  65%|██████▌   | 7241/11082 [23:16:27<12:18:33, 11.54s/step, acc=0.5060, loss=2.0202, lr=3.53e-05]
Training:  65%|██████▌   | 7242/11082 [23:16:39<12:13:47, 11.47s/step, acc=0.5060, loss=2.0202, lr=3.53e-05]
Training:  65%|██████▌   | 7243/11082 [23:16:51<12:23:25, 11.62s/step, acc=0.5060, loss=2.0202, lr=3.53e-05]
Training:  65%|██████▌   | 7244/11082 [23:17:02<12:26:34, 11.67s/step, acc=0.5060, loss=2.0202, lr=3.53e-05]
Training:  65%|██████▌   | 7245/11082 [23:17:14<12:18:55, 11.55s/step, acc=0.5060, loss=2.0202, lr=3.53e-05]
Training:  65%|██████▌   | 7246/11082 [23:17:26<12:26:51, 11.68s/step, acc=0.5060, loss=2.0202, lr=3.53e-05]
Training:  65%|██████▌   | 7247/11082 [23:17:37<12:14:29, 11.49s/step, acc=0.5060, loss=2.0202, lr=3.53e-05]
Training:  65%|██████▌   | 7248/11082 [23:17:49<12:22:14, 11.62s/step, acc=0.5060, loss=2.0202, lr=3.53e-05]
Training:  65%|██████▌   | 7249/11082 [23:18:00<12:15:37, 11.52s/step, acc=0.5060, loss=2.0202, lr=3.53e-05]
Training:  65%|██████▌   | 7250/11082 [23:18:11<12:14:24, 11.50s/step, acc=0.5060, loss=2.0202, lr=3.53e-05]
Training:  65%|██████▌   | 7250/11082 [23:18:19<12:14:24, 11.50s/step, acc=0.5074, loss=2.0149, lr=3.47e-05]2026-06-19 14:11:47 | INFO | __main__ | epoch=1 step=7250 train_loss=2.014915 train_acc=0.507423 lr=3.470e-05 grad_norm=0.6445 acc_d0=0.7638 acc_d1=0.6218 acc_d2=0.5024 acc_d3=0.4044 acc_d4=0.3267 acc_d5=0.2626 acc_d6=0.2126 acc_d7=0.1752

Training:  65%|██████▌   | 7251/11082 [23:18:23<12:23:19, 11.64s/step, acc=0.5074, loss=2.0149, lr=3.47e-05]
Training:  65%|██████▌   | 7252/11082 [23:18:35<12:14:25, 11.51s/step, acc=0.5074, loss=2.0149, lr=3.47e-05]
Training:  65%|██████▌   | 7253/11082 [23:18:46<12:20:01, 11.60s/step, acc=0.5074, loss=2.0149, lr=3.47e-05]
Training:  65%|██████▌   | 7254/11082 [23:18:58<12:14:42, 11.52s/step, acc=0.5074, loss=2.0149, lr=3.47e-05]
Training:  65%|██████▌   | 7255/11082 [23:19:09<12:18:38, 11.58s/step, acc=0.5074, loss=2.0149, lr=3.47e-05]
Training:  65%|██████▌   | 7256/11082 [23:19:21<12:23:50, 11.66s/step, acc=0.5074, loss=2.0149, lr=3.47e-05]
Training:  65%|██████▌   | 7257/11082 [23:19:33<12:23:02, 11.66s/step, acc=0.5074, loss=2.0149, lr=3.47e-05]
Training:  65%|██████▌   | 7258/11082 [23:19:44<12:16:00, 11.55s/step, acc=0.5074, loss=2.0149, lr=3.47e-05]
Training:  66%|██████▌   | 7259/11082 [23:19:56<12:27:19, 11.73s/step, acc=0.5074, loss=2.0149, lr=3.47e-05]
Training:  66%|██████▌   | 7260/11082 [23:20:08<12:27:52, 11.74s/step, acc=0.5074, loss=2.0149, lr=3.47e-05]
Training:  66%|██████▌   | 7261/11082 [23:20:20<12:23:25, 11.67s/step, acc=0.5074, loss=2.0149, lr=3.47e-05]
Training:  66%|██████▌   | 7262/11082 [23:20:31<12:24:36, 11.70s/step, acc=0.5074, loss=2.0149, lr=3.47e-05]
Training:  66%|██████▌   | 7263/11082 [23:20:43<12:19:07, 11.61s/step, acc=0.5074, loss=2.0149, lr=3.47e-05]
Training:  66%|██████▌   | 7264/11082 [23:20:54<12:15:57, 11.57s/step, acc=0.5074, loss=2.0149, lr=3.47e-05]
Training:  66%|██████▌   | 7265/11082 [23:21:06<12:13:11, 11.53s/step, acc=0.5074, loss=2.0149, lr=3.47e-05]
Training:  66%|██████▌   | 7266/11082 [23:21:18<12:21:53, 11.66s/step, acc=0.5074, loss=2.0149, lr=3.47e-05]
Training:  66%|██████▌   | 7267/11082 [23:21:29<12:20:30, 11.65s/step, acc=0.5074, loss=2.0149, lr=3.47e-05]
Training:  66%|██████▌   | 7268/11082 [23:21:41<12:15:29, 11.57s/step, acc=0.5074, loss=2.0149, lr=3.47e-05]
Training:  66%|██████▌   | 7269/11082 [23:21:52<12:15:09, 11.57s/step, acc=0.5074, loss=2.0149, lr=3.47e-05]
Training:  66%|██████▌   | 7270/11082 [23:22:04<12:16:14, 11.59s/step, acc=0.5074, loss=2.0149, lr=3.47e-05]
Training:  66%|██████▌   | 7271/11082 [23:22:15<12:12:49, 11.54s/step, acc=0.5074, loss=2.0149, lr=3.47e-05]
Training:  66%|██████▌   | 7272/11082 [23:22:26<12:06:26, 11.44s/step, acc=0.5074, loss=2.0149, lr=3.47e-05]
Training:  66%|██████▌   | 7273/11082 [23:22:38<12:12:43, 11.54s/step, acc=0.5074, loss=2.0149, lr=3.47e-05]
Training:  66%|██████▌   | 7274/11082 [23:22:50<12:13:54, 11.56s/step, acc=0.5074, loss=2.0149, lr=3.47e-05]
Training:  66%|██████▌   | 7275/11082 [23:23:02<12:20:48, 11.68s/step, acc=0.5074, loss=2.0149, lr=3.47e-05]
Training:  66%|██████▌   | 7276/11082 [23:23:13<12:11:54, 11.54s/step, acc=0.5074, loss=2.0149, lr=3.47e-05]
Training:  66%|██████▌   | 7277/11082 [23:23:25<12:16:00, 11.61s/step, acc=0.5074, loss=2.0149, lr=3.47e-05]
Training:  66%|██████▌   | 7278/11082 [23:23:37<12:20:00, 11.67s/step, acc=0.5074, loss=2.0149, lr=3.47e-05]
Training:  66%|██████▌   | 7279/11082 [23:23:48<12:16:09, 11.61s/step, acc=0.5074, loss=2.0149, lr=3.47e-05]
Training:  66%|██████▌   | 7280/11082 [23:24:00<12:18:52, 11.66s/step, acc=0.5074, loss=2.0149, lr=3.47e-05]
Training:  66%|██████▌   | 7281/11082 [23:24:12<12:18:28, 11.66s/step, acc=0.5074, loss=2.0149, lr=3.47e-05]
Training:  66%|██████▌   | 7282/11082 [23:24:23<12:18:14, 11.66s/step, acc=0.5074, loss=2.0149, lr=3.47e-05]
Training:  66%|██████▌   | 7283/11082 [23:24:34<12:07:01, 11.48s/step, acc=0.5074, loss=2.0149, lr=3.47e-05]
Training:  66%|██████▌   | 7284/11082 [23:24:46<12:17:58, 11.66s/step, acc=0.5074, loss=2.0149, lr=3.47e-05]
Training:  66%|██████▌   | 7285/11082 [23:24:57<12:07:24, 11.49s/step, acc=0.5074, loss=2.0149, lr=3.47e-05]
Training:  66%|██████▌   | 7286/11082 [23:25:09<12:09:45, 11.53s/step, acc=0.5074, loss=2.0149, lr=3.47e-05]
Training:  66%|██████▌   | 7287/11082 [23:25:21<12:09:34, 11.53s/step, acc=0.5074, loss=2.0149, lr=3.47e-05]
Training:  66%|██████▌   | 7288/11082 [23:25:32<12:11:05, 11.56s/step, acc=0.5074, loss=2.0149, lr=3.47e-05]
Training:  66%|██████▌   | 7289/11082 [23:25:43<12:03:38, 11.45s/step, acc=0.5074, loss=2.0149, lr=3.47e-05]
Training:  66%|██████▌   | 7290/11082 [23:25:55<12:05:22, 11.48s/step, acc=0.5074, loss=2.0149, lr=3.47e-05]
Training:  66%|██████▌   | 7291/11082 [23:26:07<12:16:15, 11.65s/step, acc=0.5074, loss=2.0149, lr=3.47e-05]
Training:  66%|██████▌   | 7292/11082 [23:26:19<12:17:37, 11.68s/step, acc=0.5074, loss=2.0149, lr=3.47e-05]
Training:  66%|██████▌   | 7293/11082 [23:26:30<12:12:42, 11.60s/step, acc=0.5074, loss=2.0149, lr=3.47e-05]
Training:  66%|██████▌   | 7294/11082 [23:26:42<12:16:43, 11.67s/step, acc=0.5074, loss=2.0149, lr=3.47e-05]
Training:  66%|██████▌   | 7295/11082 [23:26:54<12:17:04, 11.68s/step, acc=0.5074, loss=2.0149, lr=3.47e-05]
Training:  66%|██████▌   | 7296/11082 [23:27:05<12:10:48, 11.58s/step, acc=0.5074, loss=2.0149, lr=3.47e-05]
Training:  66%|██████▌   | 7297/11082 [23:27:17<12:12:13, 11.61s/step, acc=0.5074, loss=2.0149, lr=3.47e-05]
Training:  66%|██████▌   | 7298/11082 [23:27:28<12:04:50, 11.49s/step, acc=0.5074, loss=2.0149, lr=3.47e-05]
Training:  66%|██████▌   | 7299/11082 [23:27:39<12:05:17, 11.50s/step, acc=0.5074, loss=2.0149, lr=3.47e-05]
Training:  66%|██████▌   | 7300/11082 [23:27:51<12:08:54, 11.56s/step, acc=0.5074, loss=2.0149, lr=3.47e-05]
Training:  66%|██████▌   | 7300/11082 [23:27:58<12:08:54, 11.56s/step, acc=0.5070, loss=2.0145, lr=3.41e-05]2026-06-19 14:21:26 | INFO | __main__ | epoch=1 step=7300 train_loss=2.014516 train_acc=0.507030 lr=3.413e-05 grad_norm=0.7539 acc_d0=0.7717 acc_d1=0.6288 acc_d2=0.5087 acc_d3=0.4088 acc_d4=0.3284 acc_d5=0.2624 acc_d6=0.2111 acc_d7=0.1740

Training:  66%|██████▌   | 7301/11082 [23:28:03<12:08:47, 11.57s/step, acc=0.5070, loss=2.0145, lr=3.41e-05]
Training:  66%|██████▌   | 7302/11082 [23:28:14<12:02:23, 11.47s/step, acc=0.5070, loss=2.0145, lr=3.41e-05]
Training:  66%|██████▌   | 7303/11082 [23:28:26<12:09:57, 11.59s/step, acc=0.5070, loss=2.0145, lr=3.41e-05]
Training:  66%|██████▌   | 7304/11082 [23:28:37<12:00:16, 11.44s/step, acc=0.5070, loss=2.0145, lr=3.41e-05]
Training:  66%|██████▌   | 7305/11082 [23:28:49<12:07:41, 11.56s/step, acc=0.5070, loss=2.0145, lr=3.41e-05]
Training:  66%|██████▌   | 7306/11082 [23:29:01<12:11:27, 11.62s/step, acc=0.5070, loss=2.0145, lr=3.41e-05]
Training:  66%|██████▌   | 7307/11082 [23:29:12<12:12:48, 11.65s/step, acc=0.5070, loss=2.0145, lr=3.41e-05]
Training:  66%|██████▌   | 7308/11082 [23:29:24<12:10:45, 11.62s/step, acc=0.5070, loss=2.0145, lr=3.41e-05]
Training:  66%|██████▌   | 7309/11082 [23:29:35<12:03:24, 11.50s/step, acc=0.5070, loss=2.0145, lr=3.41e-05]
Training:  66%|██████▌   | 7310/11082 [23:29:47<12:02:25, 11.49s/step, acc=0.5070, loss=2.0145, lr=3.41e-05]
Training:  66%|██████▌   | 7311/11082 [23:29:58<12:10:10, 11.62s/step, acc=0.5070, loss=2.0145, lr=3.41e-05]
Training:  66%|██████▌   | 7312/11082 [23:30:10<12:14:25, 11.69s/step, acc=0.5070, loss=2.0145, lr=3.41e-05]
Training:  66%|██████▌   | 7313/11082 [23:30:21<12:00:54, 11.48s/step, acc=0.5070, loss=2.0145, lr=3.41e-05]
Training:  66%|██████▌   | 7314/11082 [23:30:33<11:56:44, 11.41s/step, acc=0.5070, loss=2.0145, lr=3.41e-05]
Training:  66%|██████▌   | 7315/11082 [23:30:45<12:07:29, 11.59s/step, acc=0.5070, loss=2.0145, lr=3.41e-05]
Training:  66%|██████▌   | 7316/11082 [23:30:56<12:09:48, 11.63s/step, acc=0.5070, loss=2.0145, lr=3.41e-05]
Training:  66%|██████▌   | 7317/11082 [23:31:08<12:05:02, 11.55s/step, acc=0.5070, loss=2.0145, lr=3.41e-05]
Training:  66%|██████▌   | 7318/11082 [23:31:19<12:06:15, 11.58s/step, acc=0.5070, loss=2.0145, lr=3.41e-05]
Training:  66%|██████▌   | 7319/11082 [23:31:31<12:07:09, 11.59s/step, acc=0.5070, loss=2.0145, lr=3.41e-05]
Training:  66%|██████▌   | 7320/11082 [23:31:42<12:06:59, 11.59s/step, acc=0.5070, loss=2.0145, lr=3.41e-05]
Training:  66%|██████▌   | 7321/11082 [23:31:54<12:12:20, 11.68s/step, acc=0.5070, loss=2.0145, lr=3.41e-05]
Training:  66%|██████▌   | 7322/11082 [23:32:06<12:06:21, 11.59s/step, acc=0.5070, loss=2.0145, lr=3.41e-05]
Training:  66%|██████▌   | 7323/11082 [23:32:17<12:04:43, 11.57s/step, acc=0.5070, loss=2.0145, lr=3.41e-05]
Training:  66%|██████▌   | 7324/11082 [23:32:29<12:03:04, 11.54s/step, acc=0.5070, loss=2.0145, lr=3.41e-05]
Training:  66%|██████▌   | 7325/11082 [23:32:40<12:03:19, 11.55s/step, acc=0.5070, loss=2.0145, lr=3.41e-05]
Training:  66%|██████▌   | 7326/11082 [23:32:51<11:54:37, 11.42s/step, acc=0.5070, loss=2.0145, lr=3.41e-05]
Training:  66%|██████▌   | 7327/11082 [23:33:03<12:03:01, 11.55s/step, acc=0.5070, loss=2.0145, lr=3.41e-05]
Training:  66%|██████▌   | 7328/11082 [23:33:15<12:05:52, 11.60s/step, acc=0.5070, loss=2.0145, lr=3.41e-05]
Training:  66%|██████▌   | 7329/11082 [23:33:26<12:01:13, 11.53s/step, acc=0.5070, loss=2.0145, lr=3.41e-05]
Training:  66%|██████▌   | 7330/11082 [23:33:38<11:56:49, 11.46s/step, acc=0.5070, loss=2.0145, lr=3.41e-05]
Training:  66%|██████▌   | 7331/11082 [23:33:49<12:00:49, 11.53s/step, acc=0.5070, loss=2.0145, lr=3.41e-05]
Training:  66%|██████▌   | 7332/11082 [23:34:01<11:55:08, 11.44s/step, acc=0.5070, loss=2.0145, lr=3.41e-05]
Training:  66%|██████▌   | 7333/11082 [23:34:12<11:56:02, 11.46s/step, acc=0.5070, loss=2.0145, lr=3.41e-05]
Training:  66%|██████▌   | 7334/11082 [23:34:24<12:04:01, 11.59s/step, acc=0.5070, loss=2.0145, lr=3.41e-05]
Training:  66%|██████▌   | 7335/11082 [23:34:35<11:59:59, 11.53s/step, acc=0.5070, loss=2.0145, lr=3.41e-05]
Training:  66%|██████▌   | 7336/11082 [23:34:47<12:04:12, 11.60s/step, acc=0.5070, loss=2.0145, lr=3.41e-05]
Training:  66%|██████▌   | 7337/11082 [23:34:58<11:56:14, 11.48s/step, acc=0.5070, loss=2.0145, lr=3.41e-05]
Training:  66%|██████▌   | 7338/11082 [23:35:11<12:10:43, 11.71s/step, acc=0.5070, loss=2.0145, lr=3.41e-05]
Training:  66%|██████▌   | 7339/11082 [23:35:22<11:58:30, 11.52s/step, acc=0.5070, loss=2.0145, lr=3.41e-05]
Training:  66%|██████▌   | 7340/11082 [23:35:34<12:06:48, 11.65s/step, acc=0.5070, loss=2.0145, lr=3.41e-05]
Training:  66%|██████▌   | 7341/11082 [23:35:45<12:00:21, 11.55s/step, acc=0.5070, loss=2.0145, lr=3.41e-05]
Training:  66%|██████▋   | 7342/11082 [23:35:56<11:58:28, 11.53s/step, acc=0.5070, loss=2.0145, lr=3.41e-05]
Training:  66%|██████▋   | 7343/11082 [23:36:08<12:08:07, 11.68s/step, acc=0.5070, loss=2.0145, lr=3.41e-05]
Training:  66%|██████▋   | 7344/11082 [23:36:20<11:57:02, 11.51s/step, acc=0.5070, loss=2.0145, lr=3.41e-05]
Training:  66%|██████▋   | 7345/11082 [23:36:31<12:00:03, 11.56s/step, acc=0.5070, loss=2.0145, lr=3.41e-05]
Training:  66%|██████▋   | 7346/11082 [23:36:43<12:00:43, 11.57s/step, acc=0.5070, loss=2.0145, lr=3.41e-05]
Training:  66%|██████▋   | 7347/11082 [23:36:54<11:57:12, 11.52s/step, acc=0.5070, loss=2.0145, lr=3.41e-05]
Training:  66%|██████▋   | 7348/11082 [23:37:06<11:55:43, 11.50s/step, acc=0.5070, loss=2.0145, lr=3.41e-05]
Training:  66%|██████▋   | 7349/11082 [23:37:18<12:06:00, 11.67s/step, acc=0.5070, loss=2.0145, lr=3.41e-05]
Training:  66%|██████▋   | 7350/11082 [23:37:29<12:02:15, 11.61s/step, acc=0.5070, loss=2.0145, lr=3.41e-05]
Training:  66%|██████▋   | 7350/11082 [23:37:36<12:02:15, 11.61s/step, acc=0.5069, loss=2.0156, lr=3.36e-05]2026-06-19 14:31:04 | INFO | __main__ | epoch=1 step=7350 train_loss=2.015641 train_acc=0.506945 lr=3.355e-05 grad_norm=0.6484 acc_d0=0.7821 acc_d1=0.6410 acc_d2=0.5233 acc_d3=0.4245 acc_d4=0.3431 acc_d5=0.2778 acc_d6=0.2233 acc_d7=0.1827

Training:  66%|██████▋   | 7351/11082 [23:37:41<12:02:34, 11.62s/step, acc=0.5069, loss=2.0156, lr=3.36e-05]
Training:  66%|██████▋   | 7352/11082 [23:37:53<12:05:37, 11.67s/step, acc=0.5069, loss=2.0156, lr=3.36e-05]
Training:  66%|██████▋   | 7353/11082 [23:38:04<12:04:41, 11.66s/step, acc=0.5069, loss=2.0156, lr=3.36e-05]
Training:  66%|██████▋   | 7354/11082 [23:38:15<11:53:36, 11.49s/step, acc=0.5069, loss=2.0156, lr=3.36e-05]
Training:  66%|██████▋   | 7355/11082 [23:38:28<12:05:48, 11.68s/step, acc=0.5069, loss=2.0156, lr=3.36e-05]
Training:  66%|██████▋   | 7356/11082 [23:38:39<11:58:32, 11.57s/step, acc=0.5069, loss=2.0156, lr=3.36e-05]
Training:  66%|██████▋   | 7357/11082 [23:38:51<12:00:04, 11.60s/step, acc=0.5069, loss=2.0156, lr=3.36e-05]
Training:  66%|██████▋   | 7358/11082 [23:39:02<11:55:50, 11.53s/step, acc=0.5069, loss=2.0156, lr=3.36e-05]
Training:  66%|██████▋   | 7359/11082 [23:39:14<11:59:11, 11.59s/step, acc=0.5069, loss=2.0156, lr=3.36e-05]
Training:  66%|██████▋   | 7360/11082 [23:39:25<11:59:26, 11.60s/step, acc=0.5069, loss=2.0156, lr=3.36e-05]
Training:  66%|██████▋   | 7361/11082 [23:39:37<11:54:04, 11.51s/step, acc=0.5069, loss=2.0156, lr=3.36e-05]
Training:  66%|██████▋   | 7362/11082 [23:39:48<11:55:20, 11.54s/step, acc=0.5069, loss=2.0156, lr=3.36e-05]
Training:  66%|██████▋   | 7363/11082 [23:39:59<11:48:02, 11.42s/step, acc=0.5069, loss=2.0156, lr=3.36e-05]
Training:  66%|██████▋   | 7364/11082 [23:40:11<11:59:57, 11.62s/step, acc=0.5069, loss=2.0156, lr=3.36e-05]
Training:  66%|██████▋   | 7365/11082 [23:40:23<11:53:55, 11.52s/step, acc=0.5069, loss=2.0156, lr=3.36e-05]
Training:  66%|██████▋   | 7366/11082 [23:40:34<11:57:48, 11.59s/step, acc=0.5069, loss=2.0156, lr=3.36e-05]
Training:  66%|██████▋   | 7367/11082 [23:40:46<11:58:15, 11.60s/step, acc=0.5069, loss=2.0156, lr=3.36e-05]
Training:  66%|██████▋   | 7368/11082 [23:40:58<12:00:25, 11.64s/step, acc=0.5069, loss=2.0156, lr=3.36e-05]
Training:  66%|██████▋   | 7369/11082 [23:41:09<11:57:00, 11.59s/step, acc=0.5069, loss=2.0156, lr=3.36e-05]
Training:  67%|██████▋   | 7370/11082 [23:41:21<11:59:34, 11.63s/step, acc=0.5069, loss=2.0156, lr=3.36e-05]
Training:  67%|██████▋   | 7371/11082 [23:41:32<11:54:19, 11.55s/step, acc=0.5069, loss=2.0156, lr=3.36e-05]
Training:  67%|██████▋   | 7372/11082 [23:41:44<11:57:19, 11.60s/step, acc=0.5069, loss=2.0156, lr=3.36e-05]
Training:  67%|██████▋   | 7373/11082 [23:41:55<11:50:33, 11.49s/step, acc=0.5069, loss=2.0156, lr=3.36e-05]
Training:  67%|██████▋   | 7374/11082 [23:42:07<11:53:31, 11.55s/step, acc=0.5069, loss=2.0156, lr=3.36e-05]
Training:  67%|██████▋   | 7375/11082 [23:42:19<11:57:43, 11.62s/step, acc=0.5069, loss=2.0156, lr=3.36e-05]
Training:  67%|██████▋   | 7376/11082 [23:42:30<11:49:14, 11.48s/step, acc=0.5069, loss=2.0156, lr=3.36e-05]
Training:  67%|██████▋   | 7377/11082 [23:42:41<11:49:47, 11.49s/step, acc=0.5069, loss=2.0156, lr=3.36e-05]
Training:  67%|██████▋   | 7378/11082 [23:42:53<11:55:03, 11.58s/step, acc=0.5069, loss=2.0156, lr=3.36e-05]
Training:  67%|██████▋   | 7379/11082 [23:43:05<11:52:11, 11.54s/step, acc=0.5069, loss=2.0156, lr=3.36e-05]
Training:  67%|██████▋   | 7380/11082 [23:43:16<11:54:06, 11.57s/step, acc=0.5069, loss=2.0156, lr=3.36e-05]
Training:  67%|██████▋   | 7381/11082 [23:43:28<11:49:02, 11.49s/step, acc=0.5069, loss=2.0156, lr=3.36e-05]
Training:  67%|██████▋   | 7382/11082 [23:43:39<11:41:05, 11.37s/step, acc=0.5069, loss=2.0156, lr=3.36e-05]
Training:  67%|██████▋   | 7383/11082 [23:43:51<11:52:20, 11.55s/step, acc=0.5069, loss=2.0156, lr=3.36e-05]
Training:  67%|██████▋   | 7384/11082 [23:44:02<11:49:47, 11.52s/step, acc=0.5069, loss=2.0156, lr=3.36e-05]
Training:  67%|██████▋   | 7385/11082 [23:44:14<11:52:39, 11.57s/step, acc=0.5069, loss=2.0156, lr=3.36e-05]
Training:  67%|██████▋   | 7386/11082 [23:44:25<11:52:59, 11.57s/step, acc=0.5069, loss=2.0156, lr=3.36e-05]
Training:  67%|██████▋   | 7387/11082 [23:44:37<11:51:00, 11.55s/step, acc=0.5069, loss=2.0156, lr=3.36e-05]
Training:  67%|██████▋   | 7388/11082 [23:44:49<11:56:44, 11.64s/step, acc=0.5069, loss=2.0156, lr=3.36e-05]
Training:  67%|██████▋   | 7389/11082 [23:45:01<12:00:27, 11.71s/step, acc=0.5069, loss=2.0156, lr=3.36e-05]
Training:  67%|██████▋   | 7390/11082 [23:45:12<11:52:32, 11.58s/step, acc=0.5069, loss=2.0156, lr=3.36e-05]
Training:  67%|██████▋   | 7391/11082 [23:45:24<11:57:41, 11.67s/step, acc=0.5069, loss=2.0156, lr=3.36e-05]
Training:  67%|██████▋   | 7392/11082 [23:45:35<11:54:40, 11.62s/step, acc=0.5069, loss=2.0156, lr=3.36e-05]
Training:  67%|██████▋   | 7393/11082 [23:45:47<11:58:12, 11.68s/step, acc=0.5069, loss=2.0156, lr=3.36e-05]
Training:  67%|██████▋   | 7394/11082 [23:45:59<11:54:27, 11.62s/step, acc=0.5069, loss=2.0156, lr=3.36e-05]
Training:  67%|██████▋   | 7395/11082 [23:46:10<11:51:53, 11.58s/step, acc=0.5069, loss=2.0156, lr=3.36e-05]
Training:  67%|██████▋   | 7396/11082 [23:46:22<11:54:19, 11.63s/step, acc=0.5069, loss=2.0156, lr=3.36e-05]
Training:  67%|██████▋   | 7397/11082 [23:46:33<11:48:17, 11.53s/step, acc=0.5069, loss=2.0156, lr=3.36e-05]
Training:  67%|██████▋   | 7398/11082 [23:46:44<11:45:06, 11.48s/step, acc=0.5069, loss=2.0156, lr=3.36e-05]
Training:  67%|██████▋   | 7399/11082 [23:46:57<11:54:42, 11.64s/step, acc=0.5069, loss=2.0156, lr=3.36e-05]
Training:  67%|██████▋   | 7400/11082 [23:47:07<11:41:17, 11.43s/step, acc=0.5069, loss=2.0156, lr=3.36e-05]
Training:  67%|██████▋   | 7400/11082 [23:47:15<11:41:17, 11.43s/step, acc=0.5044, loss=2.0316, lr=3.30e-05]2026-06-19 14:40:43 | INFO | __main__ | epoch=1 step=7400 train_loss=2.031569 train_acc=0.504448 lr=3.299e-05 grad_norm=0.7305 acc_d0=0.7676 acc_d1=0.6259 acc_d2=0.5064 acc_d3=0.4087 acc_d4=0.3312 acc_d5=0.2678 acc_d6=0.2186 acc_d7=0.1803

Training:  67%|██████▋   | 7401/11082 [23:47:19<11:48:02, 11.54s/step, acc=0.5044, loss=2.0316, lr=3.30e-05]
Training:  67%|██████▋   | 7402/11082 [23:47:31<11:49:36, 11.57s/step, acc=0.5044, loss=2.0316, lr=3.30e-05]
Training:  67%|██████▋   | 7403/11082 [23:47:43<11:51:20, 11.60s/step, acc=0.5044, loss=2.0316, lr=3.30e-05]
Training:  67%|██████▋   | 7404/11082 [23:47:54<11:46:28, 11.52s/step, acc=0.5044, loss=2.0316, lr=3.30e-05]
Training:  67%|██████▋   | 7405/11082 [23:48:06<11:48:59, 11.57s/step, acc=0.5044, loss=2.0316, lr=3.30e-05]
Training:  67%|██████▋   | 7406/11082 [23:48:17<11:44:57, 11.51s/step, acc=0.5044, loss=2.0316, lr=3.30e-05]
Training:  67%|██████▋   | 7407/11082 [23:48:29<11:52:02, 11.63s/step, acc=0.5044, loss=2.0316, lr=3.30e-05]
Training:  67%|██████▋   | 7408/11082 [23:48:40<11:49:43, 11.59s/step, acc=0.5044, loss=2.0316, lr=3.30e-05]
Training:  67%|██████▋   | 7409/11082 [23:48:52<11:49:48, 11.59s/step, acc=0.5044, loss=2.0316, lr=3.30e-05]
Training:  67%|██████▋   | 7410/11082 [23:49:04<11:51:16, 11.62s/step, acc=0.5044, loss=2.0316, lr=3.30e-05]
Training:  67%|██████▋   | 7411/11082 [23:49:15<11:42:59, 11.49s/step, acc=0.5044, loss=2.0316, lr=3.30e-05]
Training:  67%|██████▋   | 7412/11082 [23:49:27<11:49:57, 11.61s/step, acc=0.5044, loss=2.0316, lr=3.30e-05]
Training:  67%|██████▋   | 7413/11082 [23:49:38<11:39:00, 11.43s/step, acc=0.5044, loss=2.0316, lr=3.30e-05]
Training:  67%|██████▋   | 7414/11082 [23:49:50<11:47:59, 11.58s/step, acc=0.5044, loss=2.0316, lr=3.30e-05]
Training:  67%|██████▋   | 7415/11082 [23:50:01<11:37:44, 11.42s/step, acc=0.5044, loss=2.0316, lr=3.30e-05]
Training:  67%|██████▋   | 7416/11082 [23:50:12<11:39:40, 11.45s/step, acc=0.5044, loss=2.0316, lr=3.30e-05]
Training:  67%|██████▋   | 7417/11082 [23:50:24<11:48:49, 11.60s/step, acc=0.5044, loss=2.0316, lr=3.30e-05]
Training:  67%|██████▋   | 7418/11082 [23:50:36<11:44:41, 11.54s/step, acc=0.5044, loss=2.0316, lr=3.30e-05]
Training:  67%|██████▋   | 7419/11082 [23:50:47<11:47:52, 11.60s/step, acc=0.5044, loss=2.0316, lr=3.30e-05]
Training:  67%|██████▋   | 7420/11082 [23:50:59<11:42:16, 11.51s/step, acc=0.5044, loss=2.0316, lr=3.30e-05]
Training:  67%|██████▋   | 7421/11082 [23:51:10<11:43:04, 11.52s/step, acc=0.5044, loss=2.0316, lr=3.30e-05]
Training:  67%|██████▋   | 7422/11082 [23:51:22<11:51:55, 11.67s/step, acc=0.5044, loss=2.0316, lr=3.30e-05]
Training:  67%|██████▋   | 7423/11082 [23:51:34<11:47:19, 11.60s/step, acc=0.5044, loss=2.0316, lr=3.30e-05]
Training:  67%|██████▋   | 7424/11082 [23:51:45<11:42:30, 11.52s/step, acc=0.5044, loss=2.0316, lr=3.30e-05]
Training:  67%|██████▋   | 7425/11082 [23:51:57<11:50:55, 11.66s/step, acc=0.5044, loss=2.0316, lr=3.30e-05]
Training:  67%|██████▋   | 7426/11082 [23:52:08<11:45:28, 11.58s/step, acc=0.5044, loss=2.0316, lr=3.30e-05]
Training:  67%|██████▋   | 7427/11082 [23:52:20<11:43:41, 11.55s/step, acc=0.5044, loss=2.0316, lr=3.30e-05]
Training:  67%|██████▋   | 7428/11082 [23:52:31<11:44:26, 11.57s/step, acc=0.5044, loss=2.0316, lr=3.30e-05]
Training:  67%|██████▋   | 7429/11082 [23:52:43<11:45:11, 11.58s/step, acc=0.5044, loss=2.0316, lr=3.30e-05]
Training:  67%|██████▋   | 7430/11082 [23:52:54<11:39:54, 11.50s/step, acc=0.5044, loss=2.0316, lr=3.30e-05]
Training:  67%|██████▋   | 7431/11082 [23:53:06<11:44:38, 11.58s/step, acc=0.5044, loss=2.0316, lr=3.30e-05]
Training:  67%|██████▋   | 7432/11082 [23:53:18<11:45:27, 11.60s/step, acc=0.5044, loss=2.0316, lr=3.30e-05]
Training:  67%|██████▋   | 7433/11082 [23:53:29<11:41:37, 11.54s/step, acc=0.5044, loss=2.0316, lr=3.30e-05]
Training:  67%|██████▋   | 7434/11082 [23:53:41<11:49:33, 11.67s/step, acc=0.5044, loss=2.0316, lr=3.30e-05]
Training:  67%|██████▋   | 7435/11082 [23:53:52<11:38:23, 11.49s/step, acc=0.5044, loss=2.0316, lr=3.30e-05]
Training:  67%|██████▋   | 7436/11082 [23:54:04<11:49:16, 11.67s/step, acc=0.5044, loss=2.0316, lr=3.30e-05]
Training:  67%|██████▋   | 7437/11082 [23:54:16<11:48:04, 11.66s/step, acc=0.5044, loss=2.0316, lr=3.30e-05]
Training:  67%|██████▋   | 7438/11082 [23:54:28<11:47:21, 11.65s/step, acc=0.5044, loss=2.0316, lr=3.30e-05]
Training:  67%|██████▋   | 7439/11082 [23:54:39<11:36:12, 11.47s/step, acc=0.5044, loss=2.0316, lr=3.30e-05]
Training:  67%|██████▋   | 7440/11082 [23:54:50<11:43:55, 11.60s/step, acc=0.5044, loss=2.0316, lr=3.30e-05]
Training:  67%|██████▋   | 7441/11082 [23:55:02<11:39:42, 11.53s/step, acc=0.5044, loss=2.0316, lr=3.30e-05]
Training:  67%|██████▋   | 7442/11082 [23:55:13<11:40:26, 11.55s/step, acc=0.5044, loss=2.0316, lr=3.30e-05]
Training:  67%|██████▋   | 7443/11082 [23:55:25<11:47:58, 11.67s/step, acc=0.5044, loss=2.0316, lr=3.30e-05]
Training:  67%|██████▋   | 7444/11082 [23:55:37<11:41:33, 11.57s/step, acc=0.5044, loss=2.0316, lr=3.30e-05]
Training:  67%|██████▋   | 7445/11082 [23:55:48<11:38:36, 11.52s/step, acc=0.5044, loss=2.0316, lr=3.30e-05]
Training:  67%|██████▋   | 7446/11082 [23:56:00<11:41:14, 11.57s/step, acc=0.5044, loss=2.0316, lr=3.30e-05]
Training:  67%|██████▋   | 7447/11082 [23:56:12<11:46:18, 11.66s/step, acc=0.5044, loss=2.0316, lr=3.30e-05]
Training:  67%|██████▋   | 7448/11082 [23:56:23<11:36:49, 11.50s/step, acc=0.5044, loss=2.0316, lr=3.30e-05]
Training:  67%|██████▋   | 7449/11082 [23:56:35<11:45:21, 11.65s/step, acc=0.5044, loss=2.0316, lr=3.30e-05]
Training:  67%|██████▋   | 7450/11082 [23:56:46<11:41:42, 11.59s/step, acc=0.5044, loss=2.0316, lr=3.30e-05]
Training:  67%|██████▋   | 7450/11082 [23:56:53<11:41:42, 11.59s/step, acc=0.5032, loss=2.0357, lr=3.24e-05]2026-06-19 14:50:21 | INFO | __main__ | epoch=1 step=7450 train_loss=2.035734 train_acc=0.503235 lr=3.243e-05 grad_norm=0.6484 acc_d0=0.7692 acc_d1=0.6279 acc_d2=0.5093 acc_d3=0.4102 acc_d4=0.3305 acc_d5=0.2680 acc_d6=0.2172 acc_d7=0.1782

Training:  67%|██████▋   | 7451/11082 [23:56:57<11:32:21, 11.44s/step, acc=0.5032, loss=2.0357, lr=3.24e-05]
Training:  67%|██████▋   | 7452/11082 [23:57:09<11:35:29, 11.50s/step, acc=0.5032, loss=2.0357, lr=3.24e-05]
Training:  67%|██████▋   | 7453/11082 [23:57:20<11:33:07, 11.46s/step, acc=0.5032, loss=2.0357, lr=3.24e-05]
Training:  67%|██████▋   | 7454/11082 [23:57:32<11:37:49, 11.54s/step, acc=0.5032, loss=2.0357, lr=3.24e-05]
Training:  67%|██████▋   | 7455/11082 [23:57:44<11:42:40, 11.62s/step, acc=0.5032, loss=2.0357, lr=3.24e-05]
Training:  67%|██████▋   | 7456/11082 [23:57:55<11:34:34, 11.49s/step, acc=0.5032, loss=2.0357, lr=3.24e-05]
Training:  67%|██████▋   | 7457/11082 [23:58:07<11:35:47, 11.52s/step, acc=0.5032, loss=2.0357, lr=3.24e-05]
Training:  67%|██████▋   | 7458/11082 [23:58:18<11:37:21, 11.55s/step, acc=0.5032, loss=2.0357, lr=3.24e-05]
Training:  67%|██████▋   | 7459/11082 [23:58:30<11:37:16, 11.55s/step, acc=0.5032, loss=2.0357, lr=3.24e-05]
Training:  67%|██████▋   | 7460/11082 [23:58:42<11:41:29, 11.62s/step, acc=0.5032, loss=2.0357, lr=3.24e-05]
Training:  67%|██████▋   | 7461/11082 [23:58:52<11:27:44, 11.40s/step, acc=0.5032, loss=2.0357, lr=3.24e-05]
Training:  67%|██████▋   | 7462/11082 [23:59:04<11:37:49, 11.57s/step, acc=0.5032, loss=2.0357, lr=3.24e-05]
Training:  67%|██████▋   | 7463/11082 [23:59:16<11:36:53, 11.55s/step, acc=0.5032, loss=2.0357, lr=3.24e-05]
Training:  67%|██████▋   | 7464/11082 [23:59:28<11:36:47, 11.56s/step, acc=0.5032, loss=2.0357, lr=3.24e-05]
Training:  67%|██████▋   | 7465/11082 [23:59:39<11:38:47, 11.59s/step, acc=0.5032, loss=2.0357, lr=3.24e-05]
Training:  67%|██████▋   | 7466/11082 [23:59:51<11:36:46, 11.56s/step, acc=0.5032, loss=2.0357, lr=3.24e-05]
Training:  67%|██████▋   | 7467/11082 [24:00:02<11:34:37, 11.53s/step, acc=0.5032, loss=2.0357, lr=3.24e-05]
Training:  67%|██████▋   | 7468/11082 [24:00:14<11:33:52, 11.52s/step, acc=0.5032, loss=2.0357, lr=3.24e-05]
Training:  67%|██████▋   | 7469/11082 [24:00:25<11:37:29, 11.58s/step, acc=0.5032, loss=2.0357, lr=3.24e-05]
Training:  67%|██████▋   | 7470/11082 [24:00:37<11:44:37, 11.70s/step, acc=0.5032, loss=2.0357, lr=3.24e-05]
Training:  67%|██████▋   | 7471/11082 [24:00:49<11:42:01, 11.66s/step, acc=0.5032, loss=2.0357, lr=3.24e-05]
Training:  67%|██████▋   | 7472/11082 [24:01:00<11:31:45, 11.50s/step, acc=0.5032, loss=2.0357, lr=3.24e-05]
Training:  67%|██████▋   | 7473/11082 [24:01:12<11:30:40, 11.48s/step, acc=0.5032, loss=2.0357, lr=3.24e-05]
Training:  67%|██████▋   | 7474/11082 [24:01:23<11:33:58, 11.54s/step, acc=0.5032, loss=2.0357, lr=3.24e-05]
Training:  67%|██████▋   | 7475/11082 [24:01:35<11:39:51, 11.64s/step, acc=0.5032, loss=2.0357, lr=3.24e-05]
Training:  67%|██████▋   | 7476/11082 [24:01:46<11:30:24, 11.49s/step, acc=0.5032, loss=2.0357, lr=3.24e-05]
Training:  67%|██████▋   | 7477/11082 [24:01:58<11:38:11, 11.62s/step, acc=0.5032, loss=2.0357, lr=3.24e-05]
Training:  67%|██████▋   | 7478/11082 [24:02:10<11:37:43, 11.62s/step, acc=0.5032, loss=2.0357, lr=3.24e-05]
Training:  67%|██████▋   | 7479/11082 [24:02:21<11:29:56, 11.49s/step, acc=0.5032, loss=2.0357, lr=3.24e-05]
Training:  67%|██████▋   | 7480/11082 [24:02:33<11:39:04, 11.64s/step, acc=0.5032, loss=2.0357, lr=3.24e-05]
Training:  68%|██████▊   | 7481/11082 [24:02:44<11:34:22, 11.57s/step, acc=0.5032, loss=2.0357, lr=3.24e-05]
Training:  68%|██████▊   | 7482/11082 [24:02:56<11:38:11, 11.64s/step, acc=0.5032, loss=2.0357, lr=3.24e-05]
Training:  68%|██████▊   | 7483/11082 [24:03:07<11:30:58, 11.52s/step, acc=0.5032, loss=2.0357, lr=3.24e-05]
Training:  68%|██████▊   | 7484/11082 [24:03:19<11:40:23, 11.68s/step, acc=0.5032, loss=2.0357, lr=3.24e-05]
Training:  68%|██████▊   | 7485/11082 [24:03:31<11:40:57, 11.69s/step, acc=0.5032, loss=2.0357, lr=3.24e-05]
Training:  68%|██████▊   | 7486/11082 [24:03:43<11:35:39, 11.61s/step, acc=0.5032, loss=2.0357, lr=3.24e-05]
Training:  68%|██████▊   | 7487/11082 [24:03:54<11:32:24, 11.56s/step, acc=0.5032, loss=2.0357, lr=3.24e-05]
Training:  68%|██████▊   | 7488/11082 [24:04:06<11:35:36, 11.61s/step, acc=0.5032, loss=2.0357, lr=3.24e-05]
Training:  68%|██████▊   | 7489/11082 [24:04:17<11:34:03, 11.59s/step, acc=0.5032, loss=2.0357, lr=3.24e-05]
Training:  68%|██████▊   | 7490/11082 [24:04:29<11:36:41, 11.64s/step, acc=0.5032, loss=2.0357, lr=3.24e-05]
Training:  68%|██████▊   | 7491/11082 [24:04:40<11:25:51, 11.46s/step, acc=0.5032, loss=2.0357, lr=3.24e-05]
Training:  68%|██████▊   | 7492/11082 [24:04:52<11:32:44, 11.58s/step, acc=0.5032, loss=2.0357, lr=3.24e-05]
Training:  68%|██████▊   | 7493/11082 [24:05:04<11:36:21, 11.64s/step, acc=0.5032, loss=2.0357, lr=3.24e-05]
Training:  68%|██████▊   | 7494/11082 [24:05:15<11:28:41, 11.52s/step, acc=0.5032, loss=2.0357, lr=3.24e-05]
Training:  68%|██████▊   | 7495/11082 [24:05:26<11:26:45, 11.49s/step, acc=0.5032, loss=2.0357, lr=3.24e-05]
Training:  68%|██████▊   | 7496/11082 [24:05:38<11:29:27, 11.54s/step, acc=0.5032, loss=2.0357, lr=3.24e-05]
Training:  68%|██████▊   | 7497/11082 [24:05:50<11:31:07, 11.57s/step, acc=0.5032, loss=2.0357, lr=3.24e-05]
Training:  68%|██████▊   | 7498/11082 [24:06:01<11:28:33, 11.53s/step, acc=0.5032, loss=2.0357, lr=3.24e-05]
Training:  68%|██████▊   | 7499/11082 [24:06:13<11:28:01, 11.52s/step, acc=0.5032, loss=2.0357, lr=3.24e-05]
Training:  68%|██████▊   | 7500/11082 [24:06:25<11:36:47, 11.67s/step, acc=0.5032, loss=2.0357, lr=3.24e-05]
Training:  68%|██████▊   | 7500/11082 [24:06:32<11:36:47, 11.67s/step, acc=0.5049, loss=2.0304, lr=3.19e-05]2026-06-19 15:00:00 | INFO | __main__ | epoch=1 step=7500 train_loss=2.030421 train_acc=0.504871 lr=3.187e-05 grad_norm=0.6602 acc_d0=0.7672 acc_d1=0.6237 acc_d2=0.5014 acc_d3=0.4033 acc_d4=0.3238 acc_d5=0.2608 acc_d6=0.2108 acc_d7=0.1711

Training:  68%|██████▊   | 7501/11082 [24:06:37<11:42:18, 11.77s/step, acc=0.5049, loss=2.0304, lr=3.19e-05]
Training:  68%|██████▊   | 7502/11082 [24:06:48<11:36:13, 11.67s/step, acc=0.5049, loss=2.0304, lr=3.19e-05]
Training:  68%|██████▊   | 7503/11082 [24:07:00<11:37:16, 11.69s/step, acc=0.5049, loss=2.0304, lr=3.19e-05]
Training:  68%|██████▊   | 7504/11082 [24:07:11<11:25:27, 11.49s/step, acc=0.5049, loss=2.0304, lr=3.19e-05]
Training:  68%|██████▊   | 7505/11082 [24:07:23<11:30:33, 11.58s/step, acc=0.5049, loss=2.0304, lr=3.19e-05]
Training:  68%|██████▊   | 7506/11082 [24:07:34<11:32:38, 11.62s/step, acc=0.5049, loss=2.0304, lr=3.19e-05]
Training:  68%|██████▊   | 7507/11082 [24:07:46<11:32:03, 11.61s/step, acc=0.5049, loss=2.0304, lr=3.19e-05]
Training:  68%|██████▊   | 7508/11082 [24:07:58<11:34:07, 11.65s/step, acc=0.5049, loss=2.0304, lr=3.19e-05]
Training:  68%|██████▊   | 7509/11082 [24:08:09<11:32:06, 11.62s/step, acc=0.5049, loss=2.0304, lr=3.19e-05]
Training:  68%|██████▊   | 7510/11082 [24:08:21<11:26:57, 11.54s/step, acc=0.5049, loss=2.0304, lr=3.19e-05]
Training:  68%|██████▊   | 7511/11082 [24:08:33<11:34:49, 11.67s/step, acc=0.5049, loss=2.0304, lr=3.19e-05]
Training:  68%|██████▊   | 7512/11082 [24:08:44<11:30:28, 11.60s/step, acc=0.5049, loss=2.0304, lr=3.19e-05]
Training:  68%|██████▊   | 7513/11082 [24:08:55<11:24:15, 11.50s/step, acc=0.5049, loss=2.0304, lr=3.19e-05]
Training:  68%|██████▊   | 7514/11082 [24:09:08<11:38:34, 11.75s/step, acc=0.5049, loss=2.0304, lr=3.19e-05]
Training:  68%|██████▊   | 7515/11082 [24:09:19<11:34:55, 11.69s/step, acc=0.5049, loss=2.0304, lr=3.19e-05]
Training:  68%|██████▊   | 7516/11082 [24:09:31<11:32:09, 11.65s/step, acc=0.5049, loss=2.0304, lr=3.19e-05]
Training:  68%|██████▊   | 7517/11082 [24:09:42<11:30:09, 11.62s/step, acc=0.5049, loss=2.0304, lr=3.19e-05]
Training:  68%|██████▊   | 7518/11082 [24:09:54<11:27:51, 11.58s/step, acc=0.5049, loss=2.0304, lr=3.19e-05]
Training:  68%|██████▊   | 7519/11082 [24:10:05<11:29:25, 11.61s/step, acc=0.5049, loss=2.0304, lr=3.19e-05]
Training:  68%|██████▊   | 7520/11082 [24:10:17<11:27:04, 11.57s/step, acc=0.5049, loss=2.0304, lr=3.19e-05]
Training:  68%|██████▊   | 7521/11082 [24:10:28<11:24:24, 11.53s/step, acc=0.5049, loss=2.0304, lr=3.19e-05]
Training:  68%|██████▊   | 7522/11082 [24:10:40<11:21:54, 11.49s/step, acc=0.5049, loss=2.0304, lr=3.19e-05]
Training:  68%|██████▊   | 7523/11082 [24:10:51<11:20:49, 11.48s/step, acc=0.5049, loss=2.0304, lr=3.19e-05]
Training:  68%|██████▊   | 7524/11082 [24:11:03<11:27:25, 11.59s/step, acc=0.5049, loss=2.0304, lr=3.19e-05]
Training:  68%|██████▊   | 7525/11082 [24:11:14<11:24:51, 11.55s/step, acc=0.5049, loss=2.0304, lr=3.19e-05]
Training:  68%|██████▊   | 7526/11082 [24:11:26<11:29:00, 11.63s/step, acc=0.5049, loss=2.0304, lr=3.19e-05]
Training:  68%|██████▊   | 7527/11082 [24:11:38<11:24:06, 11.55s/step, acc=0.5049, loss=2.0304, lr=3.19e-05]
Training:  68%|██████▊   | 7528/11082 [24:11:49<11:25:42, 11.58s/step, acc=0.5049, loss=2.0304, lr=3.19e-05]
Training:  68%|██████▊   | 7529/11082 [24:12:01<11:30:30, 11.66s/step, acc=0.5049, loss=2.0304, lr=3.19e-05]
Training:  68%|██████▊   | 7530/11082 [24:12:12<11:23:39, 11.55s/step, acc=0.5049, loss=2.0304, lr=3.19e-05]
Training:  68%|██████▊   | 7531/11082 [24:12:24<11:20:44, 11.50s/step, acc=0.5049, loss=2.0304, lr=3.19e-05]
Training:  68%|██████▊   | 7532/11082 [24:12:35<11:23:22, 11.55s/step, acc=0.5049, loss=2.0304, lr=3.19e-05]
Training:  68%|██████▊   | 7533/11082 [24:12:47<11:27:55, 11.63s/step, acc=0.5049, loss=2.0304, lr=3.19e-05]
Training:  68%|██████▊   | 7534/11082 [24:12:59<11:25:16, 11.59s/step, acc=0.5049, loss=2.0304, lr=3.19e-05]
Training:  68%|██████▊   | 7535/11082 [24:13:11<11:31:57, 11.71s/step, acc=0.5049, loss=2.0304, lr=3.19e-05]
Training:  68%|██████▊   | 7536/11082 [24:13:22<11:22:35, 11.55s/step, acc=0.5049, loss=2.0304, lr=3.19e-05]
Training:  68%|██████▊   | 7537/11082 [24:13:33<11:22:28, 11.55s/step, acc=0.5049, loss=2.0304, lr=3.19e-05]
Training:  68%|██████▊   | 7538/11082 [24:13:45<11:24:31, 11.59s/step, acc=0.5049, loss=2.0304, lr=3.19e-05]
Training:  68%|██████▊   | 7539/11082 [24:13:57<11:26:06, 11.62s/step, acc=0.5049, loss=2.0304, lr=3.19e-05]
Training:  68%|██████▊   | 7540/11082 [24:14:08<11:25:13, 11.61s/step, acc=0.5049, loss=2.0304, lr=3.19e-05]
Training:  68%|██████▊   | 7541/11082 [24:14:20<11:19:46, 11.52s/step, acc=0.5049, loss=2.0304, lr=3.19e-05]
Training:  68%|██████▊   | 7542/11082 [24:14:31<11:15:28, 11.45s/step, acc=0.5049, loss=2.0304, lr=3.19e-05]
Training:  68%|██████▊   | 7543/11082 [24:14:43<11:25:43, 11.63s/step, acc=0.5049, loss=2.0304, lr=3.19e-05]
Training:  68%|██████▊   | 7544/11082 [24:14:54<11:15:37, 11.46s/step, acc=0.5049, loss=2.0304, lr=3.19e-05]
Training:  68%|██████▊   | 7545/11082 [24:15:06<11:17:43, 11.50s/step, acc=0.5049, loss=2.0304, lr=3.19e-05]
Training:  68%|██████▊   | 7546/11082 [24:15:17<11:20:48, 11.55s/step, acc=0.5049, loss=2.0304, lr=3.19e-05]
Training:  68%|██████▊   | 7547/11082 [24:15:29<11:22:29, 11.58s/step, acc=0.5049, loss=2.0304, lr=3.19e-05]
Training:  68%|██████▊   | 7548/11082 [24:15:40<11:17:07, 11.50s/step, acc=0.5049, loss=2.0304, lr=3.19e-05]
Training:  68%|██████▊   | 7549/11082 [24:15:52<11:22:55, 11.60s/step, acc=0.5049, loss=2.0304, lr=3.19e-05]
Training:  68%|██████▊   | 7550/11082 [24:16:03<11:17:31, 11.51s/step, acc=0.5049, loss=2.0304, lr=3.19e-05]
Training:  68%|██████▊   | 7550/11082 [24:16:11<11:17:31, 11.51s/step, acc=0.5061, loss=2.0211, lr=3.13e-05]2026-06-19 15:09:38 | INFO | __main__ | epoch=1 step=7550 train_loss=2.021138 train_acc=0.506080 lr=3.131e-05 grad_norm=0.6406 acc_d0=0.7610 acc_d1=0.6190 acc_d2=0.4992 acc_d3=0.4008 acc_d4=0.3228 acc_d5=0.2610 acc_d6=0.2102 acc_d7=0.1734

Training:  68%|██████▊   | 7551/11082 [24:16:15<11:20:51, 11.57s/step, acc=0.5061, loss=2.0211, lr=3.13e-05]
Training:  68%|██████▊   | 7552/11082 [24:16:27<11:18:56, 11.54s/step, acc=0.5061, loss=2.0211, lr=3.13e-05]
Training:  68%|██████▊   | 7553/11082 [24:16:39<11:27:07, 11.68s/step, acc=0.5061, loss=2.0211, lr=3.13e-05]
Training:  68%|██████▊   | 7554/11082 [24:16:50<11:19:23, 11.55s/step, acc=0.5061, loss=2.0211, lr=3.13e-05]
Training:  68%|██████▊   | 7555/11082 [24:17:02<11:23:54, 11.63s/step, acc=0.5061, loss=2.0211, lr=3.13e-05]
Training:  68%|██████▊   | 7556/11082 [24:17:13<11:17:59, 11.54s/step, acc=0.5061, loss=2.0211, lr=3.13e-05]
Training:  68%|██████▊   | 7557/11082 [24:17:25<11:18:22, 11.55s/step, acc=0.5061, loss=2.0211, lr=3.13e-05]
Training:  68%|██████▊   | 7558/11082 [24:17:37<11:24:41, 11.66s/step, acc=0.5061, loss=2.0211, lr=3.13e-05]
Training:  68%|██████▊   | 7559/11082 [24:17:48<11:15:36, 11.51s/step, acc=0.5061, loss=2.0211, lr=3.13e-05]
Training:  68%|██████▊   | 7560/11082 [24:18:00<11:24:35, 11.66s/step, acc=0.5061, loss=2.0211, lr=3.13e-05]
Training:  68%|██████▊   | 7561/11082 [24:18:11<11:17:47, 11.55s/step, acc=0.5061, loss=2.0211, lr=3.13e-05]
Training:  68%|██████▊   | 7562/11082 [24:18:22<11:15:43, 11.52s/step, acc=0.5061, loss=2.0211, lr=3.13e-05]
Training:  68%|██████▊   | 7563/11082 [24:18:34<11:18:27, 11.57s/step, acc=0.5061, loss=2.0211, lr=3.13e-05]
Training:  68%|██████▊   | 7564/11082 [24:18:46<11:15:25, 11.52s/step, acc=0.5061, loss=2.0211, lr=3.13e-05]
Training:  68%|██████▊   | 7565/11082 [24:18:57<11:15:18, 11.52s/step, acc=0.5061, loss=2.0211, lr=3.13e-05]
Training:  68%|██████▊   | 7566/11082 [24:19:09<11:13:34, 11.49s/step, acc=0.5061, loss=2.0211, lr=3.13e-05]
Training:  68%|██████▊   | 7567/11082 [24:19:21<11:23:20, 11.66s/step, acc=0.5061, loss=2.0211, lr=3.13e-05]
Training:  68%|██████▊   | 7568/11082 [24:19:32<11:11:57, 11.47s/step, acc=0.5061, loss=2.0211, lr=3.13e-05]
Training:  68%|██████▊   | 7569/11082 [24:19:43<11:16:00, 11.55s/step, acc=0.5061, loss=2.0211, lr=3.13e-05]
Training:  68%|██████▊   | 7570/11082 [24:19:55<11:17:37, 11.58s/step, acc=0.5061, loss=2.0211, lr=3.13e-05]
Training:  68%|██████▊   | 7571/11082 [24:20:06<11:14:09, 11.52s/step, acc=0.5061, loss=2.0211, lr=3.13e-05]
Training:  68%|██████▊   | 7572/11082 [24:20:18<11:12:21, 11.49s/step, acc=0.5061, loss=2.0211, lr=3.13e-05]
Training:  68%|██████▊   | 7573/11082 [24:20:30<11:17:06, 11.58s/step, acc=0.5061, loss=2.0211, lr=3.13e-05]
Training:  68%|██████▊   | 7574/11082 [24:20:41<11:13:03, 11.51s/step, acc=0.5061, loss=2.0211, lr=3.13e-05]
Training:  68%|██████▊   | 7575/11082 [24:20:52<11:10:41, 11.47s/step, acc=0.5061, loss=2.0211, lr=3.13e-05]
Training:  68%|██████▊   | 7576/11082 [24:21:04<11:14:45, 11.55s/step, acc=0.5061, loss=2.0211, lr=3.13e-05]
Training:  68%|██████▊   | 7577/11082 [24:21:16<11:22:21, 11.68s/step, acc=0.5061, loss=2.0211, lr=3.13e-05]
Training:  68%|██████▊   | 7578/11082 [24:21:27<11:14:54, 11.56s/step, acc=0.5061, loss=2.0211, lr=3.13e-05]
Training:  68%|██████▊   | 7579/11082 [24:21:38<11:06:23, 11.41s/step, acc=0.5061, loss=2.0211, lr=3.13e-05]
Training:  68%|██████▊   | 7580/11082 [24:21:50<11:10:58, 11.50s/step, acc=0.5061, loss=2.0211, lr=3.13e-05]
Training:  68%|██████▊   | 7581/11082 [24:22:02<11:15:20, 11.57s/step, acc=0.5061, loss=2.0211, lr=3.13e-05]
Training:  68%|██████▊   | 7582/11082 [24:22:14<11:19:33, 11.65s/step, acc=0.5061, loss=2.0211, lr=3.13e-05]
Training:  68%|██████▊   | 7583/11082 [24:22:25<11:11:55, 11.52s/step, acc=0.5061, loss=2.0211, lr=3.13e-05]
Training:  68%|██████▊   | 7584/11082 [24:22:36<11:10:39, 11.50s/step, acc=0.5061, loss=2.0211, lr=3.13e-05]
Training:  68%|██████▊   | 7585/11082 [24:22:48<11:15:57, 11.60s/step, acc=0.5061, loss=2.0211, lr=3.13e-05]
Training:  68%|██████▊   | 7586/11082 [24:22:59<11:10:59, 11.52s/step, acc=0.5061, loss=2.0211, lr=3.13e-05]
Training:  68%|██████▊   | 7587/11082 [24:23:11<11:13:14, 11.56s/step, acc=0.5061, loss=2.0211, lr=3.13e-05]
Training:  68%|██████▊   | 7588/11082 [24:23:22<11:08:47, 11.48s/step, acc=0.5061, loss=2.0211, lr=3.13e-05]
Training:  68%|██████▊   | 7589/11082 [24:23:34<11:16:51, 11.63s/step, acc=0.5061, loss=2.0211, lr=3.13e-05]
Training:  68%|██████▊   | 7590/11082 [24:23:46<11:10:53, 11.53s/step, acc=0.5061, loss=2.0211, lr=3.13e-05]
Training:  68%|██████▊   | 7591/11082 [24:23:57<11:10:34, 11.53s/step, acc=0.5061, loss=2.0211, lr=3.13e-05]
Training:  69%|██████▊   | 7592/11082 [24:24:09<11:13:13, 11.57s/step, acc=0.5061, loss=2.0211, lr=3.13e-05]
Training:  69%|██████▊   | 7593/11082 [24:24:21<11:17:39, 11.65s/step, acc=0.5061, loss=2.0211, lr=3.13e-05]
Training:  69%|██████▊   | 7594/11082 [24:24:32<11:11:21, 11.55s/step, acc=0.5061, loss=2.0211, lr=3.13e-05]
Training:  69%|██████▊   | 7595/11082 [24:24:44<11:11:50, 11.56s/step, acc=0.5061, loss=2.0211, lr=3.13e-05]
Training:  69%|██████▊   | 7596/11082 [24:24:55<11:12:58, 11.58s/step, acc=0.5061, loss=2.0211, lr=3.13e-05]
Training:  69%|██████▊   | 7597/11082 [24:25:07<11:13:52, 11.60s/step, acc=0.5061, loss=2.0211, lr=3.13e-05]
Training:  69%|██████▊   | 7598/11082 [24:25:18<11:11:57, 11.57s/step, acc=0.5061, loss=2.0211, lr=3.13e-05]
Training:  69%|██████▊   | 7599/11082 [24:25:30<11:13:34, 11.60s/step, acc=0.5061, loss=2.0211, lr=3.13e-05]
Training:  69%|██████▊   | 7600/11082 [24:25:41<11:06:48, 11.49s/step, acc=0.5061, loss=2.0211, lr=3.13e-05]
Training:  69%|██████▊   | 7600/11082 [24:25:48<11:06:48, 11.49s/step, acc=0.5078, loss=2.0130, lr=3.08e-05]2026-06-19 15:19:16 | INFO | __main__ | epoch=1 step=7600 train_loss=2.013047 train_acc=0.507833 lr=3.077e-05 grad_norm=0.6523 acc_d0=0.7651 acc_d1=0.6232 acc_d2=0.5036 acc_d3=0.4063 acc_d4=0.3283 acc_d5=0.2664 acc_d6=0.2170 acc_d7=0.1771

Training:  69%|██████▊   | 7601/11082 [24:25:53<11:07:37, 11.51s/step, acc=0.5078, loss=2.0130, lr=3.08e-05]
Training:  69%|██████▊   | 7602/11082 [24:26:04<11:09:04, 11.54s/step, acc=0.5078, loss=2.0130, lr=3.08e-05]
Training:  69%|██████▊   | 7603/11082 [24:26:16<11:06:52, 11.50s/step, acc=0.5078, loss=2.0130, lr=3.08e-05]
Training:  69%|██████▊   | 7604/11082 [24:26:28<11:12:07, 11.60s/step, acc=0.5078, loss=2.0130, lr=3.08e-05]
Training:  69%|██████▊   | 7605/11082 [24:26:40<11:32:32, 11.95s/step, acc=0.5078, loss=2.0130, lr=3.08e-05]
Training:  69%|██████▊   | 7606/11082 [24:26:52<11:17:57, 11.70s/step, acc=0.5078, loss=2.0130, lr=3.08e-05]
Training:  69%|██████▊   | 7607/11082 [24:27:04<11:23:08, 11.80s/step, acc=0.5078, loss=2.0130, lr=3.08e-05]
Training:  69%|██████▊   | 7608/11082 [24:27:15<11:19:06, 11.73s/step, acc=0.5078, loss=2.0130, lr=3.08e-05]
Training:  69%|██████▊   | 7609/11082 [24:27:27<11:23:26, 11.81s/step, acc=0.5078, loss=2.0130, lr=3.08e-05]
Training:  69%|██████▊   | 7610/11082 [24:27:39<11:20:11, 11.75s/step, acc=0.5078, loss=2.0130, lr=3.08e-05]
Training:  69%|██████▊   | 7611/11082 [24:27:51<11:22:47, 11.80s/step, acc=0.5078, loss=2.0130, lr=3.08e-05]
Training:  69%|██████▊   | 7612/11082 [24:28:02<11:12:13, 11.62s/step, acc=0.5078, loss=2.0130, lr=3.08e-05]
Training:  69%|██████▊   | 7613/11082 [24:28:14<11:15:20, 11.68s/step, acc=0.5078, loss=2.0130, lr=3.08e-05]
Training:  69%|██████▊   | 7614/11082 [24:28:26<11:17:07, 11.71s/step, acc=0.5078, loss=2.0130, lr=3.08e-05]
Training:  69%|██████▊   | 7615/11082 [24:28:37<11:10:59, 11.61s/step, acc=0.5078, loss=2.0130, lr=3.08e-05]
Training:  69%|██████▊   | 7616/11082 [24:28:48<11:08:38, 11.57s/step, acc=0.5078, loss=2.0130, lr=3.08e-05]
Training:  69%|██████▊   | 7617/11082 [24:29:00<11:10:28, 11.61s/step, acc=0.5078, loss=2.0130, lr=3.08e-05]
Training:  69%|██████▊   | 7618/11082 [24:29:12<11:13:12, 11.66s/step, acc=0.5078, loss=2.0130, lr=3.08e-05]
Training:  69%|██████▉   | 7619/11082 [24:29:23<11:10:09, 11.61s/step, acc=0.5078, loss=2.0130, lr=3.08e-05]
Training:  69%|██████▉   | 7620/11082 [24:29:35<11:04:41, 11.52s/step, acc=0.5078, loss=2.0130, lr=3.08e-05]
Training:  69%|██████▉   | 7621/11082 [24:29:46<11:08:06, 11.58s/step, acc=0.5078, loss=2.0130, lr=3.08e-05]
Training:  69%|██████▉   | 7622/11082 [24:29:58<11:06:32, 11.56s/step, acc=0.5078, loss=2.0130, lr=3.08e-05]
Training:  69%|██████▉   | 7623/11082 [24:30:10<11:07:36, 11.58s/step, acc=0.5078, loss=2.0130, lr=3.08e-05]
Training:  69%|██████▉   | 7624/11082 [24:30:22<11:16:41, 11.74s/step, acc=0.5078, loss=2.0130, lr=3.08e-05]
Training:  69%|██████▉   | 7625/11082 [24:30:33<11:07:34, 11.59s/step, acc=0.5078, loss=2.0130, lr=3.08e-05]
Training:  69%|██████▉   | 7626/11082 [24:30:45<11:13:04, 11.69s/step, acc=0.5078, loss=2.0130, lr=3.08e-05]
Training:  69%|██████▉   | 7627/11082 [24:30:56<11:03:04, 11.51s/step, acc=0.5078, loss=2.0130, lr=3.08e-05]
Training:  69%|██████▉   | 7628/11082 [24:31:08<11:12:57, 11.69s/step, acc=0.5078, loss=2.0130, lr=3.08e-05]
Training:  69%|██████▉   | 7629/11082 [24:31:20<11:17:16, 11.77s/step, acc=0.5078, loss=2.0130, lr=3.08e-05]
Training:  69%|██████▉   | 7630/11082 [24:31:31<10:59:40, 11.47s/step, acc=0.5078, loss=2.0130, lr=3.08e-05]
Training:  69%|██████▉   | 7631/11082 [24:31:43<11:13:45, 11.71s/step, acc=0.5078, loss=2.0130, lr=3.08e-05]
Training:  69%|██████▉   | 7632/11082 [24:31:54<11:06:32, 11.59s/step, acc=0.5078, loss=2.0130, lr=3.08e-05]
Training:  69%|██████▉   | 7633/11082 [24:32:06<11:07:58, 11.62s/step, acc=0.5078, loss=2.0130, lr=3.08e-05]
Training:  69%|██████▉   | 7634/11082 [24:32:18<11:08:43, 11.64s/step, acc=0.5078, loss=2.0130, lr=3.08e-05]
Training:  69%|██████▉   | 7635/11082 [24:32:29<11:07:34, 11.62s/step, acc=0.5078, loss=2.0130, lr=3.08e-05]
Training:  69%|██████▉   | 7636/11082 [24:32:41<11:04:39, 11.57s/step, acc=0.5078, loss=2.0130, lr=3.08e-05]
Training:  69%|██████▉   | 7637/11082 [24:32:52<11:07:07, 11.62s/step, acc=0.5078, loss=2.0130, lr=3.08e-05]
Training:  69%|██████▉   | 7638/11082 [24:33:04<11:08:08, 11.64s/step, acc=0.5078, loss=2.0130, lr=3.08e-05]
Training:  69%|██████▉   | 7639/11082 [24:33:16<11:08:52, 11.66s/step, acc=0.5078, loss=2.0130, lr=3.08e-05]
Training:  69%|██████▉   | 7640/11082 [24:33:27<11:07:52, 11.64s/step, acc=0.5078, loss=2.0130, lr=3.08e-05]
Training:  69%|██████▉   | 7641/11082 [24:33:39<11:05:34, 11.61s/step, acc=0.5078, loss=2.0130, lr=3.08e-05]
Training:  69%|██████▉   | 7642/11082 [24:33:51<11:04:50, 11.60s/step, acc=0.5078, loss=2.0130, lr=3.08e-05]
Training:  69%|██████▉   | 7643/11082 [24:34:02<11:02:52, 11.57s/step, acc=0.5078, loss=2.0130, lr=3.08e-05]
Training:  69%|██████▉   | 7644/11082 [24:34:14<11:03:28, 11.58s/step, acc=0.5078, loss=2.0130, lr=3.08e-05]
Training:  69%|██████▉   | 7645/11082 [24:34:25<10:58:45, 11.50s/step, acc=0.5078, loss=2.0130, lr=3.08e-05]
Training:  69%|██████▉   | 7646/11082 [24:34:36<10:55:13, 11.44s/step, acc=0.5078, loss=2.0130, lr=3.08e-05]
Training:  69%|██████▉   | 7647/11082 [24:34:49<11:09:13, 11.69s/step, acc=0.5078, loss=2.0130, lr=3.08e-05]
Training:  69%|██████▉   | 7648/11082 [24:35:00<11:00:33, 11.54s/step, acc=0.5078, loss=2.0130, lr=3.08e-05]
Training:  69%|██████▉   | 7649/11082 [24:35:12<11:07:38, 11.67s/step, acc=0.5078, loss=2.0130, lr=3.08e-05]
Training:  69%|██████▉   | 7650/11082 [24:35:23<11:01:25, 11.56s/step, acc=0.5078, loss=2.0130, lr=3.08e-05]
Training:  69%|██████▉   | 7650/11082 [24:35:30<11:01:25, 11.56s/step, acc=0.5044, loss=2.0288, lr=3.02e-05]2026-06-19 15:28:58 | INFO | __main__ | epoch=1 step=7650 train_loss=2.028823 train_acc=0.504431 lr=3.022e-05 grad_norm=0.6211 acc_d0=0.7689 acc_d1=0.6265 acc_d2=0.5054 acc_d3=0.4084 acc_d4=0.3292 acc_d5=0.2653 acc_d6=0.2142 acc_d7=0.1753

Training:  69%|██████▉   | 7651/11082 [24:35:35<11:07:22, 11.67s/step, acc=0.5044, loss=2.0288, lr=3.02e-05]
Training:  69%|██████▉   | 7652/11082 [24:35:47<11:10:08, 11.72s/step, acc=0.5044, loss=2.0288, lr=3.02e-05]
Training:  69%|██████▉   | 7653/11082 [24:35:58<10:59:40, 11.54s/step, acc=0.5044, loss=2.0288, lr=3.02e-05]
Training:  69%|██████▉   | 7654/11082 [24:36:10<11:07:56, 11.69s/step, acc=0.5044, loss=2.0288, lr=3.02e-05]
Training:  69%|██████▉   | 7655/11082 [24:36:22<11:06:11, 11.66s/step, acc=0.5044, loss=2.0288, lr=3.02e-05]
Training:  69%|██████▉   | 7656/11082 [24:36:33<10:57:47, 11.52s/step, acc=0.5044, loss=2.0288, lr=3.02e-05]
Training:  69%|██████▉   | 7657/11082 [24:36:45<11:03:57, 11.63s/step, acc=0.5044, loss=2.0288, lr=3.02e-05]
Training:  69%|██████▉   | 7658/11082 [24:36:56<11:01:01, 11.58s/step, acc=0.5044, loss=2.0288, lr=3.02e-05]
Training:  69%|██████▉   | 7659/11082 [24:37:08<11:02:09, 11.61s/step, acc=0.5044, loss=2.0288, lr=3.02e-05]
Training:  69%|██████▉   | 7660/11082 [24:37:19<11:01:47, 11.60s/step, acc=0.5044, loss=2.0288, lr=3.02e-05]
Training:  69%|██████▉   | 7661/11082 [24:37:31<11:02:45, 11.62s/step, acc=0.5044, loss=2.0288, lr=3.02e-05]
Training:  69%|██████▉   | 7662/11082 [24:37:42<10:54:25, 11.48s/step, acc=0.5044, loss=2.0288, lr=3.02e-05]
Training:  69%|██████▉   | 7663/11082 [24:37:54<11:05:29, 11.68s/step, acc=0.5044, loss=2.0288, lr=3.02e-05]
Training:  69%|██████▉   | 7664/11082 [24:38:06<11:04:08, 11.66s/step, acc=0.5044, loss=2.0288, lr=3.02e-05]
Training:  69%|██████▉   | 7665/11082 [24:38:17<10:58:52, 11.57s/step, acc=0.5044, loss=2.0288, lr=3.02e-05]
Training:  69%|██████▉   | 7666/11082 [24:38:29<11:01:02, 11.61s/step, acc=0.5044, loss=2.0288, lr=3.02e-05]
Training:  69%|██████▉   | 7667/11082 [24:38:41<11:01:15, 11.62s/step, acc=0.5044, loss=2.0288, lr=3.02e-05]
Training:  69%|██████▉   | 7668/11082 [24:38:52<10:58:54, 11.58s/step, acc=0.5044, loss=2.0288, lr=3.02e-05]
Training:  69%|██████▉   | 7669/11082 [24:39:04<10:58:31, 11.58s/step, acc=0.5044, loss=2.0288, lr=3.02e-05]
Training:  69%|██████▉   | 7670/11082 [24:39:15<10:54:54, 11.52s/step, acc=0.5044, loss=2.0288, lr=3.02e-05]
Training:  69%|██████▉   | 7671/11082 [24:39:26<10:51:29, 11.46s/step, acc=0.5044, loss=2.0288, lr=3.02e-05]
Training:  69%|██████▉   | 7672/11082 [24:39:39<11:05:18, 11.71s/step, acc=0.5044, loss=2.0288, lr=3.02e-05]
Training:  69%|██████▉   | 7673/11082 [24:39:50<10:53:11, 11.50s/step, acc=0.5044, loss=2.0288, lr=3.02e-05]
Training:  69%|██████▉   | 7674/11082 [24:40:02<11:00:23, 11.63s/step, acc=0.5044, loss=2.0288, lr=3.02e-05]
Training:  69%|██████▉   | 7675/11082 [24:40:13<10:58:22, 11.59s/step, acc=0.5044, loss=2.0288, lr=3.02e-05]
Training:  69%|██████▉   | 7676/11082 [24:40:25<11:02:26, 11.67s/step, acc=0.5044, loss=2.0288, lr=3.02e-05]
Training:  69%|██████▉   | 7677/11082 [24:40:37<11:02:39, 11.68s/step, acc=0.5044, loss=2.0288, lr=3.02e-05]
Training:  69%|██████▉   | 7678/11082 [24:40:48<11:02:06, 11.67s/step, acc=0.5044, loss=2.0288, lr=3.02e-05]
Training:  69%|██████▉   | 7679/11082 [24:41:00<10:58:15, 11.61s/step, acc=0.5044, loss=2.0288, lr=3.02e-05]
Training:  69%|██████▉   | 7680/11082 [24:41:11<10:57:27, 11.60s/step, acc=0.5044, loss=2.0288, lr=3.02e-05]
Training:  69%|██████▉   | 7681/11082 [24:41:23<10:56:05, 11.57s/step, acc=0.5044, loss=2.0288, lr=3.02e-05]
Training:  69%|██████▉   | 7682/11082 [24:41:34<10:55:36, 11.57s/step, acc=0.5044, loss=2.0288, lr=3.02e-05]
Training:  69%|██████▉   | 7683/11082 [24:41:46<10:55:18, 11.57s/step, acc=0.5044, loss=2.0288, lr=3.02e-05]
Training:  69%|██████▉   | 7684/11082 [24:41:58<11:03:11, 11.71s/step, acc=0.5044, loss=2.0288, lr=3.02e-05]
Training:  69%|██████▉   | 7685/11082 [24:42:09<10:52:32, 11.53s/step, acc=0.5044, loss=2.0288, lr=3.02e-05]
Training:  69%|██████▉   | 7686/11082 [24:42:20<10:46:38, 11.42s/step, acc=0.5044, loss=2.0288, lr=3.02e-05]
Training:  69%|██████▉   | 7687/11082 [24:42:32<10:59:14, 11.65s/step, acc=0.5044, loss=2.0288, lr=3.02e-05]
Training:  69%|██████▉   | 7688/11082 [24:42:44<10:59:31, 11.66s/step, acc=0.5044, loss=2.0288, lr=3.02e-05]
Training:  69%|██████▉   | 7689/11082 [24:42:56<10:56:06, 11.60s/step, acc=0.5044, loss=2.0288, lr=3.02e-05]
Training:  69%|██████▉   | 7690/11082 [24:43:08<11:01:59, 11.71s/step, acc=0.5044, loss=2.0288, lr=3.02e-05]
Training:  69%|██████▉   | 7691/11082 [24:43:19<10:58:18, 11.65s/step, acc=0.5044, loss=2.0288, lr=3.02e-05]
Training:  69%|██████▉   | 7692/11082 [24:43:30<10:53:28, 11.57s/step, acc=0.5044, loss=2.0288, lr=3.02e-05]
Training:  69%|██████▉   | 7693/11082 [24:43:42<10:57:53, 11.65s/step, acc=0.5044, loss=2.0288, lr=3.02e-05]
Training:  69%|██████▉   | 7694/11082 [24:43:54<10:59:01, 11.67s/step, acc=0.5044, loss=2.0288, lr=3.02e-05]
Training:  69%|██████▉   | 7695/11082 [24:44:05<10:49:43, 11.51s/step, acc=0.5044, loss=2.0288, lr=3.02e-05]
Training:  69%|██████▉   | 7696/11082 [24:44:17<10:56:01, 11.62s/step, acc=0.5044, loss=2.0288, lr=3.02e-05]
Training:  69%|██████▉   | 7697/11082 [24:44:29<10:57:11, 11.65s/step, acc=0.5044, loss=2.0288, lr=3.02e-05]
Training:  69%|██████▉   | 7698/11082 [24:44:40<10:51:16, 11.55s/step, acc=0.5044, loss=2.0288, lr=3.02e-05]
Training:  69%|██████▉   | 7699/11082 [24:44:52<10:48:53, 11.51s/step, acc=0.5044, loss=2.0288, lr=3.02e-05]
Training:  69%|██████▉   | 7700/11082 [24:45:03<10:54:03, 11.60s/step, acc=0.5044, loss=2.0288, lr=3.02e-05]
Training:  69%|██████▉   | 7700/11082 [24:45:10<10:54:03, 11.60s/step, acc=0.5049, loss=2.0299, lr=2.97e-05]2026-06-19 15:38:38 | INFO | __main__ | epoch=1 step=7700 train_loss=2.029898 train_acc=0.504949 lr=2.969e-05 grad_norm=0.5938 acc_d0=0.7690 acc_d1=0.6294 acc_d2=0.5100 acc_d3=0.4122 acc_d4=0.3331 acc_d5=0.2672 acc_d6=0.2173 acc_d7=0.1776

Training:  69%|██████▉   | 7701/11082 [24:45:15<10:53:03, 11.59s/step, acc=0.5049, loss=2.0299, lr=2.97e-05]
Training:  70%|██████▉   | 7702/11082 [24:45:27<10:55:57, 11.64s/step, acc=0.5049, loss=2.0299, lr=2.97e-05]
Training:  70%|██████▉   | 7703/11082 [24:45:38<10:53:22, 11.60s/step, acc=0.5049, loss=2.0299, lr=2.97e-05]
Training:  70%|██████▉   | 7704/11082 [24:45:50<10:52:41, 11.59s/step, acc=0.5049, loss=2.0299, lr=2.97e-05]
Training:  70%|██████▉   | 7705/11082 [24:46:01<10:54:13, 11.62s/step, acc=0.5049, loss=2.0299, lr=2.97e-05]
Training:  70%|██████▉   | 7706/11082 [24:46:13<10:45:32, 11.47s/step, acc=0.5049, loss=2.0299, lr=2.97e-05]
Training:  70%|██████▉   | 7707/11082 [24:46:24<10:53:18, 11.61s/step, acc=0.5049, loss=2.0299, lr=2.97e-05]
Training:  70%|██████▉   | 7708/11082 [24:46:36<10:49:33, 11.55s/step, acc=0.5049, loss=2.0299, lr=2.97e-05]
Training:  70%|██████▉   | 7709/11082 [24:46:48<10:58:01, 11.71s/step, acc=0.5049, loss=2.0299, lr=2.97e-05]
Training:  70%|██████▉   | 7710/11082 [24:46:59<10:54:38, 11.65s/step, acc=0.5049, loss=2.0299, lr=2.97e-05]
Training:  70%|██████▉   | 7711/11082 [24:47:11<10:51:15, 11.59s/step, acc=0.5049, loss=2.0299, lr=2.97e-05]
Training:  70%|██████▉   | 7712/11082 [24:47:22<10:46:13, 11.51s/step, acc=0.5049, loss=2.0299, lr=2.97e-05]
Training:  70%|██████▉   | 7713/11082 [24:47:34<10:52:49, 11.63s/step, acc=0.5049, loss=2.0299, lr=2.97e-05]
Training:  70%|██████▉   | 7714/11082 [24:47:45<10:45:10, 11.49s/step, acc=0.5049, loss=2.0299, lr=2.97e-05]
Training:  70%|██████▉   | 7715/11082 [24:47:58<10:56:20, 11.70s/step, acc=0.5049, loss=2.0299, lr=2.97e-05]
Training:  70%|██████▉   | 7716/11082 [24:48:09<10:54:12, 11.66s/step, acc=0.5049, loss=2.0299, lr=2.97e-05]
Training:  70%|██████▉   | 7717/11082 [24:48:21<10:54:58, 11.68s/step, acc=0.5049, loss=2.0299, lr=2.97e-05]
Training:  70%|██████▉   | 7718/11082 [24:48:32<10:51:02, 11.61s/step, acc=0.5049, loss=2.0299, lr=2.97e-05]
Training:  70%|██████▉   | 7719/11082 [24:48:44<10:49:58, 11.60s/step, acc=0.5049, loss=2.0299, lr=2.97e-05]
Training:  70%|██████▉   | 7720/11082 [24:48:55<10:49:42, 11.60s/step, acc=0.5049, loss=2.0299, lr=2.97e-05]
Training:  70%|██████▉   | 7721/11082 [24:49:07<10:51:21, 11.63s/step, acc=0.5049, loss=2.0299, lr=2.97e-05]
Training:  70%|██████▉   | 7722/11082 [24:49:19<10:51:50, 11.64s/step, acc=0.5049, loss=2.0299, lr=2.97e-05]
Training:  70%|██████▉   | 7723/11082 [24:49:31<10:53:56, 11.68s/step, acc=0.5049, loss=2.0299, lr=2.97e-05]
Training:  70%|██████▉   | 7724/11082 [24:49:42<10:44:53, 11.52s/step, acc=0.5049, loss=2.0299, lr=2.97e-05]
Training:  70%|██████▉   | 7725/11082 [24:49:53<10:47:01, 11.56s/step, acc=0.5049, loss=2.0299, lr=2.97e-05]
Training:  70%|██████▉   | 7726/11082 [24:50:05<10:49:58, 11.62s/step, acc=0.5049, loss=2.0299, lr=2.97e-05]
Training:  70%|██████▉   | 7727/11082 [24:50:17<10:50:45, 11.64s/step, acc=0.5049, loss=2.0299, lr=2.97e-05]
Training:  70%|██████▉   | 7728/11082 [24:50:29<10:54:10, 11.70s/step, acc=0.5049, loss=2.0299, lr=2.97e-05]
Training:  70%|██████▉   | 7729/11082 [24:50:40<10:51:09, 11.65s/step, acc=0.5049, loss=2.0299, lr=2.97e-05]
Training:  70%|██████▉   | 7730/11082 [24:50:52<10:48:24, 11.61s/step, acc=0.5049, loss=2.0299, lr=2.97e-05]
Training:  70%|██████▉   | 7731/11082 [24:51:04<10:53:46, 11.71s/step, acc=0.5049, loss=2.0299, lr=2.97e-05]
Training:  70%|██████▉   | 7732/11082 [24:51:15<10:48:23, 11.61s/step, acc=0.5049, loss=2.0299, lr=2.97e-05]
Training:  70%|██████▉   | 7733/11082 [24:51:27<10:50:52, 11.66s/step, acc=0.5049, loss=2.0299, lr=2.97e-05]
Training:  70%|██████▉   | 7734/11082 [24:51:38<10:39:32, 11.46s/step, acc=0.5049, loss=2.0299, lr=2.97e-05]
Training:  70%|██████▉   | 7735/11082 [24:51:50<10:47:02, 11.60s/step, acc=0.5049, loss=2.0299, lr=2.97e-05]
Training:  70%|██████▉   | 7736/11082 [24:52:02<10:55:14, 11.75s/step, acc=0.5049, loss=2.0299, lr=2.97e-05]
Training:  70%|██████▉   | 7737/11082 [24:52:13<10:50:02, 11.66s/step, acc=0.5049, loss=2.0299, lr=2.97e-05]
Training:  70%|██████▉   | 7738/11082 [24:52:25<10:44:20, 11.56s/step, acc=0.5049, loss=2.0299, lr=2.97e-05]
Training:  70%|██████▉   | 7739/11082 [24:52:37<10:51:17, 11.69s/step, acc=0.5049, loss=2.0299, lr=2.97e-05]
Training:  70%|██████▉   | 7740/11082 [24:52:48<10:43:42, 11.56s/step, acc=0.5049, loss=2.0299, lr=2.97e-05]
Training:  70%|██████▉   | 7741/11082 [24:52:59<10:41:35, 11.52s/step, acc=0.5049, loss=2.0299, lr=2.97e-05]
Training:  70%|██████▉   | 7742/11082 [24:53:11<10:38:20, 11.47s/step, acc=0.5049, loss=2.0299, lr=2.97e-05]
Training:  70%|██████▉   | 7743/11082 [24:53:23<10:45:58, 11.61s/step, acc=0.5049, loss=2.0299, lr=2.97e-05]
Training:  70%|██████▉   | 7744/11082 [24:53:35<10:51:29, 11.71s/step, acc=0.5049, loss=2.0299, lr=2.97e-05]
Training:  70%|██████▉   | 7745/11082 [24:53:46<10:52:44, 11.74s/step, acc=0.5049, loss=2.0299, lr=2.97e-05]
Training:  70%|██████▉   | 7746/11082 [24:53:58<10:45:57, 11.62s/step, acc=0.5049, loss=2.0299, lr=2.97e-05]
Training:  70%|██████▉   | 7747/11082 [24:54:09<10:43:44, 11.58s/step, acc=0.5049, loss=2.0299, lr=2.97e-05]
Training:  70%|██████▉   | 7748/11082 [24:54:21<10:41:19, 11.54s/step, acc=0.5049, loss=2.0299, lr=2.97e-05]
Training:  70%|██████▉   | 7749/11082 [24:54:32<10:43:46, 11.59s/step, acc=0.5049, loss=2.0299, lr=2.97e-05]
Training:  70%|██████▉   | 7750/11082 [24:54:44<10:44:09, 11.60s/step, acc=0.5049, loss=2.0299, lr=2.97e-05]
Training:  70%|██████▉   | 7750/11082 [24:54:51<10:44:09, 11.60s/step, acc=0.5052, loss=2.0298, lr=2.92e-05]2026-06-19 15:48:19 | INFO | __main__ | epoch=1 step=7750 train_loss=2.029767 train_acc=0.505206 lr=2.915e-05 grad_norm=0.5859 acc_d0=0.7698 acc_d1=0.6301 acc_d2=0.5111 acc_d3=0.4134 acc_d4=0.3351 acc_d5=0.2704 acc_d6=0.2194 acc_d7=0.1793

Training:  70%|██████▉   | 7751/11082 [24:54:55<10:33:13, 11.41s/step, acc=0.5052, loss=2.0298, lr=2.92e-05]
Training:  70%|██████▉   | 7752/11082 [24:55:07<10:39:02, 11.51s/step, acc=0.5052, loss=2.0298, lr=2.92e-05]
Training:  70%|██████▉   | 7753/11082 [24:55:19<10:50:04, 11.72s/step, acc=0.5052, loss=2.0298, lr=2.92e-05]
Training:  70%|██████▉   | 7754/11082 [24:55:30<10:38:22, 11.51s/step, acc=0.5052, loss=2.0298, lr=2.92e-05]
Training:  70%|██████▉   | 7755/11082 [24:55:42<10:48:05, 11.69s/step, acc=0.5052, loss=2.0298, lr=2.92e-05]
Training:  70%|██████▉   | 7756/11082 [24:55:53<10:43:58, 11.62s/step, acc=0.5052, loss=2.0298, lr=2.92e-05]
Training:  70%|██████▉   | 7757/11082 [24:56:05<10:42:35, 11.60s/step, acc=0.5052, loss=2.0298, lr=2.92e-05]
Training:  70%|███████   | 7758/11082 [24:56:17<10:46:09, 11.66s/step, acc=0.5052, loss=2.0298, lr=2.92e-05]
Training:  70%|███████   | 7759/11082 [24:56:28<10:44:31, 11.64s/step, acc=0.5052, loss=2.0298, lr=2.92e-05]
Training:  70%|███████   | 7760/11082 [24:56:40<10:44:23, 11.64s/step, acc=0.5052, loss=2.0298, lr=2.92e-05]
Training:  70%|███████   | 7761/11082 [24:56:52<10:45:42, 11.67s/step, acc=0.5052, loss=2.0298, lr=2.92e-05]
Training:  70%|███████   | 7762/11082 [24:57:04<10:49:31, 11.74s/step, acc=0.5052, loss=2.0298, lr=2.92e-05]
Training:  70%|███████   | 7763/11082 [24:57:16<10:51:53, 11.78s/step, acc=0.5052, loss=2.0298, lr=2.92e-05]
Training:  70%|███████   | 7764/11082 [24:57:27<10:50:38, 11.77s/step, acc=0.5052, loss=2.0298, lr=2.92e-05]
Training:  70%|███████   | 7765/11082 [24:57:39<10:46:26, 11.69s/step, acc=0.5052, loss=2.0298, lr=2.92e-05]
Training:  70%|███████   | 7766/11082 [24:57:50<10:38:44, 11.56s/step, acc=0.5052, loss=2.0298, lr=2.92e-05]
Training:  70%|███████   | 7767/11082 [24:58:01<10:32:18, 11.44s/step, acc=0.5052, loss=2.0298, lr=2.92e-05]
Training:  70%|███████   | 7768/11082 [24:58:13<10:36:40, 11.53s/step, acc=0.5052, loss=2.0298, lr=2.92e-05]
Training:  70%|███████   | 7769/11082 [24:58:25<10:40:45, 11.60s/step, acc=0.5052, loss=2.0298, lr=2.92e-05]
Training:  70%|███████   | 7770/11082 [24:58:36<10:41:05, 11.61s/step, acc=0.5052, loss=2.0298, lr=2.92e-05]
Training:  70%|███████   | 7771/11082 [24:58:48<10:46:17, 11.71s/step, acc=0.5052, loss=2.0298, lr=2.92e-05]
Training:  70%|███████   | 7772/11082 [24:58:59<10:37:40, 11.56s/step, acc=0.5052, loss=2.0298, lr=2.92e-05]
Training:  70%|███████   | 7773/11082 [24:59:11<10:34:55, 11.51s/step, acc=0.5052, loss=2.0298, lr=2.92e-05]
Training:  70%|███████   | 7774/11082 [24:59:23<10:37:15, 11.56s/step, acc=0.5052, loss=2.0298, lr=2.92e-05]
Training:  70%|███████   | 7775/11082 [24:59:34<10:42:23, 11.66s/step, acc=0.5052, loss=2.0298, lr=2.92e-05]
Training:  70%|███████   | 7776/11082 [24:59:46<10:44:15, 11.69s/step, acc=0.5052, loss=2.0298, lr=2.92e-05]
Training:  70%|███████   | 7777/11082 [24:59:57<10:36:18, 11.55s/step, acc=0.5052, loss=2.0298, lr=2.92e-05]
Training:  70%|███████   | 7778/11082 [25:00:09<10:33:04, 11.50s/step, acc=0.5052, loss=2.0298, lr=2.92e-05]
Training:  70%|███████   | 7779/11082 [25:00:21<10:39:49, 11.62s/step, acc=0.5052, loss=2.0298, lr=2.92e-05]
Training:  70%|███████   | 7780/11082 [25:00:32<10:39:29, 11.62s/step, acc=0.5052, loss=2.0298, lr=2.92e-05]
Training:  70%|███████   | 7781/11082 [25:00:44<10:37:55, 11.60s/step, acc=0.5052, loss=2.0298, lr=2.92e-05]
Training:  70%|███████   | 7782/11082 [25:00:56<10:38:46, 11.61s/step, acc=0.5052, loss=2.0298, lr=2.92e-05]
Training:  70%|███████   | 7783/11082 [25:01:07<10:28:29, 11.43s/step, acc=0.5052, loss=2.0298, lr=2.92e-05]
Training:  70%|███████   | 7784/11082 [25:01:19<10:43:04, 11.70s/step, acc=0.5052, loss=2.0298, lr=2.92e-05]
Training:  70%|███████   | 7785/11082 [25:01:30<10:41:16, 11.67s/step, acc=0.5052, loss=2.0298, lr=2.92e-05]
Training:  70%|███████   | 7786/11082 [25:01:42<10:35:36, 11.57s/step, acc=0.5052, loss=2.0298, lr=2.92e-05]
Training:  70%|███████   | 7787/11082 [25:01:53<10:36:29, 11.59s/step, acc=0.5052, loss=2.0298, lr=2.92e-05]
Training:  70%|███████   | 7788/11082 [25:02:05<10:36:33, 11.60s/step, acc=0.5052, loss=2.0298, lr=2.92e-05]
Training:  70%|███████   | 7789/11082 [25:02:17<10:35:58, 11.59s/step, acc=0.5052, loss=2.0298, lr=2.92e-05]
Training:  70%|███████   | 7790/11082 [25:02:28<10:40:23, 11.67s/step, acc=0.5052, loss=2.0298, lr=2.92e-05]
Training:  70%|███████   | 7791/11082 [25:02:40<10:33:50, 11.56s/step, acc=0.5052, loss=2.0298, lr=2.92e-05]
Training:  70%|███████   | 7792/11082 [25:02:52<10:37:57, 11.63s/step, acc=0.5052, loss=2.0298, lr=2.92e-05]
Training:  70%|███████   | 7793/11082 [25:03:03<10:37:31, 11.63s/step, acc=0.5052, loss=2.0298, lr=2.92e-05]
Training:  70%|███████   | 7794/11082 [25:03:15<10:40:35, 11.69s/step, acc=0.5052, loss=2.0298, lr=2.92e-05]
Training:  70%|███████   | 7795/11082 [25:03:27<10:37:36, 11.64s/step, acc=0.5052, loss=2.0298, lr=2.92e-05]
Training:  70%|███████   | 7796/11082 [25:03:38<10:35:05, 11.60s/step, acc=0.5052, loss=2.0298, lr=2.92e-05]
Training:  70%|███████   | 7797/11082 [25:03:50<10:36:48, 11.63s/step, acc=0.5052, loss=2.0298, lr=2.92e-05]
Training:  70%|███████   | 7798/11082 [25:04:01<10:32:34, 11.56s/step, acc=0.5052, loss=2.0298, lr=2.92e-05]
Training:  70%|███████   | 7799/11082 [25:04:13<10:42:10, 11.74s/step, acc=0.5052, loss=2.0298, lr=2.92e-05]
Training:  70%|███████   | 7800/11082 [25:04:25<10:44:19, 11.78s/step, acc=0.5052, loss=2.0298, lr=2.92e-05]
Training:  70%|███████   | 7800/11082 [25:04:32<10:44:19, 11.78s/step, acc=0.5055, loss=2.0248, lr=2.86e-05]2026-06-19 15:58:00 | INFO | __main__ | epoch=1 step=7800 train_loss=2.024818 train_acc=0.505481 lr=2.863e-05 grad_norm=0.6406 acc_d0=0.7638 acc_d1=0.6202 acc_d2=0.5014 acc_d3=0.4043 acc_d4=0.3268 acc_d5=0.2645 acc_d6=0.2154 acc_d7=0.1779

Training:  70%|███████   | 7801/11082 [25:04:37<10:48:54, 11.87s/step, acc=0.5055, loss=2.0248, lr=2.86e-05]
Training:  70%|███████   | 7802/11082 [25:04:48<10:37:20, 11.66s/step, acc=0.5055, loss=2.0248, lr=2.86e-05]
Training:  70%|███████   | 7803/11082 [25:05:00<10:39:19, 11.70s/step, acc=0.5055, loss=2.0248, lr=2.86e-05]
Training:  70%|███████   | 7804/11082 [25:05:11<10:29:36, 11.52s/step, acc=0.5055, loss=2.0248, lr=2.86e-05]
Training:  70%|███████   | 7805/11082 [25:05:23<10:34:01, 11.61s/step, acc=0.5055, loss=2.0248, lr=2.86e-05]
Training:  70%|███████   | 7806/11082 [25:05:35<10:36:08, 11.65s/step, acc=0.5055, loss=2.0248, lr=2.86e-05]
Training:  70%|███████   | 7807/11082 [25:05:46<10:28:51, 11.52s/step, acc=0.5055, loss=2.0248, lr=2.86e-05]
Training:  70%|███████   | 7808/11082 [25:05:58<10:32:44, 11.60s/step, acc=0.5055, loss=2.0248, lr=2.86e-05]
Training:  70%|███████   | 7809/11082 [25:06:10<10:34:14, 11.63s/step, acc=0.5055, loss=2.0248, lr=2.86e-05]
Training:  70%|███████   | 7810/11082 [25:06:21<10:23:05, 11.43s/step, acc=0.5055, loss=2.0248, lr=2.86e-05]
Training:  70%|███████   | 7811/11082 [25:06:32<10:30:27, 11.56s/step, acc=0.5055, loss=2.0248, lr=2.86e-05]
Training:  70%|███████   | 7812/11082 [25:06:44<10:32:10, 11.60s/step, acc=0.5055, loss=2.0248, lr=2.86e-05]
Training:  71%|███████   | 7813/11082 [25:06:55<10:25:26, 11.48s/step, acc=0.5055, loss=2.0248, lr=2.86e-05]
Training:  71%|███████   | 7814/11082 [25:07:08<10:37:19, 11.70s/step, acc=0.5055, loss=2.0248, lr=2.86e-05]
Training:  71%|███████   | 7815/11082 [25:07:18<10:25:29, 11.49s/step, acc=0.5055, loss=2.0248, lr=2.86e-05]
Training:  71%|███████   | 7816/11082 [25:07:31<10:38:13, 11.72s/step, acc=0.5055, loss=2.0248, lr=2.86e-05]
Training:  71%|███████   | 7817/11082 [25:07:42<10:35:27, 11.68s/step, acc=0.5055, loss=2.0248, lr=2.86e-05]
Training:  71%|███████   | 7818/11082 [25:07:54<10:28:36, 11.56s/step, acc=0.5055, loss=2.0248, lr=2.86e-05]
Training:  71%|███████   | 7819/11082 [25:08:06<10:37:10, 11.72s/step, acc=0.5055, loss=2.0248, lr=2.86e-05]
Training:  71%|███████   | 7820/11082 [25:08:18<10:38:44, 11.75s/step, acc=0.5055, loss=2.0248, lr=2.86e-05]
Training:  71%|███████   | 7821/11082 [25:08:29<10:28:27, 11.56s/step, acc=0.5055, loss=2.0248, lr=2.86e-05]
Training:  71%|███████   | 7822/11082 [25:08:40<10:23:46, 11.48s/step, acc=0.5055, loss=2.0248, lr=2.86e-05]
Training:  71%|███████   | 7823/11082 [25:08:52<10:30:16, 11.60s/step, acc=0.5055, loss=2.0248, lr=2.86e-05]
Training:  71%|███████   | 7824/11082 [25:09:04<10:31:47, 11.64s/step, acc=0.5055, loss=2.0248, lr=2.86e-05]
Training:  71%|███████   | 7825/11082 [25:09:15<10:27:00, 11.55s/step, acc=0.5055, loss=2.0248, lr=2.86e-05]
Training:  71%|███████   | 7826/11082 [25:09:26<10:27:11, 11.56s/step, acc=0.5055, loss=2.0248, lr=2.86e-05]
Training:  71%|███████   | 7827/11082 [25:09:38<10:33:43, 11.68s/step, acc=0.5055, loss=2.0248, lr=2.86e-05]
Training:  71%|███████   | 7828/11082 [25:09:51<10:41:11, 11.82s/step, acc=0.5055, loss=2.0248, lr=2.86e-05]
Training:  71%|███████   | 7829/11082 [25:10:02<10:33:38, 11.69s/step, acc=0.5055, loss=2.0248, lr=2.86e-05]
Training:  71%|███████   | 7830/11082 [25:10:14<10:38:54, 11.79s/step, acc=0.5055, loss=2.0248, lr=2.86e-05]
Training:  71%|███████   | 7831/11082 [25:10:25<10:33:40, 11.70s/step, acc=0.5055, loss=2.0248, lr=2.86e-05]
Training:  71%|███████   | 7832/11082 [25:10:37<10:38:00, 11.78s/step, acc=0.5055, loss=2.0248, lr=2.86e-05]
Training:  71%|███████   | 7833/11082 [25:10:49<10:32:34, 11.68s/step, acc=0.5055, loss=2.0248, lr=2.86e-05]
Training:  71%|███████   | 7834/11082 [25:11:00<10:25:09, 11.55s/step, acc=0.5055, loss=2.0248, lr=2.86e-05]
Training:  71%|███████   | 7835/11082 [25:11:12<10:31:38, 11.67s/step, acc=0.5055, loss=2.0248, lr=2.86e-05]
Training:  71%|███████   | 7836/11082 [25:11:24<10:32:47, 11.70s/step, acc=0.5055, loss=2.0248, lr=2.86e-05]
Training:  71%|███████   | 7837/11082 [25:11:36<10:34:53, 11.74s/step, acc=0.5055, loss=2.0248, lr=2.86e-05]
Training:  71%|███████   | 7838/11082 [25:11:47<10:30:32, 11.66s/step, acc=0.5055, loss=2.0248, lr=2.86e-05]
Training:  71%|███████   | 7839/11082 [25:11:59<10:27:11, 11.60s/step, acc=0.5055, loss=2.0248, lr=2.86e-05]
Training:  71%|███████   | 7840/11082 [25:12:11<10:33:58, 11.73s/step, acc=0.5055, loss=2.0248, lr=2.86e-05]
Training:  71%|███████   | 7841/11082 [25:12:22<10:31:49, 11.70s/step, acc=0.5055, loss=2.0248, lr=2.86e-05]
Training:  71%|███████   | 7842/11082 [25:12:34<10:26:40, 11.60s/step, acc=0.5055, loss=2.0248, lr=2.86e-05]
Training:  71%|███████   | 7843/11082 [25:12:46<10:33:27, 11.73s/step, acc=0.5055, loss=2.0248, lr=2.86e-05]
Training:  71%|███████   | 7844/11082 [25:12:58<10:36:31, 11.79s/step, acc=0.5055, loss=2.0248, lr=2.86e-05]
Training:  71%|███████   | 7845/11082 [25:13:09<10:22:14, 11.53s/step, acc=0.5055, loss=2.0248, lr=2.86e-05]
Training:  71%|███████   | 7846/11082 [25:13:20<10:26:11, 11.61s/step, acc=0.5055, loss=2.0248, lr=2.86e-05]
Training:  71%|███████   | 7847/11082 [25:13:32<10:29:45, 11.68s/step, acc=0.5055, loss=2.0248, lr=2.86e-05]
Training:  71%|███████   | 7848/11082 [25:13:44<10:28:57, 11.67s/step, acc=0.5055, loss=2.0248, lr=2.86e-05]
Training:  71%|███████   | 7849/11082 [25:13:55<10:24:10, 11.58s/step, acc=0.5055, loss=2.0248, lr=2.86e-05]
Training:  71%|███████   | 7850/11082 [25:14:07<10:26:35, 11.63s/step, acc=0.5055, loss=2.0248, lr=2.86e-05]
Training:  71%|███████   | 7850/11082 [25:14:14<10:26:35, 11.63s/step, acc=0.5052, loss=2.0329, lr=2.81e-05]2026-06-19 16:07:42 | INFO | __main__ | epoch=1 step=7850 train_loss=2.032901 train_acc=0.505239 lr=2.810e-05 grad_norm=0.6758 acc_d0=0.7654 acc_d1=0.6228 acc_d2=0.5045 acc_d3=0.4075 acc_d4=0.3289 acc_d5=0.2676 acc_d6=0.2189 acc_d7=0.1795

Training:  71%|███████   | 7851/11082 [25:14:18<10:15:09, 11.42s/step, acc=0.5052, loss=2.0329, lr=2.81e-05]
Training:  71%|███████   | 7852/11082 [25:14:31<10:34:28, 11.79s/step, acc=0.5052, loss=2.0329, lr=2.81e-05]
Training:  71%|███████   | 7853/11082 [25:14:42<10:27:42, 11.66s/step, acc=0.5052, loss=2.0329, lr=2.81e-05]
Training:  71%|███████   | 7854/11082 [25:14:53<10:22:34, 11.57s/step, acc=0.5052, loss=2.0329, lr=2.81e-05]
Training:  71%|███████   | 7855/11082 [25:15:05<10:19:22, 11.52s/step, acc=0.5052, loss=2.0329, lr=2.81e-05]
Training:  71%|███████   | 7856/11082 [25:15:16<10:22:27, 11.58s/step, acc=0.5052, loss=2.0329, lr=2.81e-05]
Training:  71%|███████   | 7857/11082 [25:15:28<10:19:52, 11.53s/step, acc=0.5052, loss=2.0329, lr=2.81e-05]
Training:  71%|███████   | 7858/11082 [25:15:39<10:20:22, 11.55s/step, acc=0.5052, loss=2.0329, lr=2.81e-05]
Training:  71%|███████   | 7859/11082 [25:15:51<10:20:13, 11.55s/step, acc=0.5052, loss=2.0329, lr=2.81e-05]
Training:  71%|███████   | 7860/11082 [25:16:03<10:31:00, 11.75s/step, acc=0.5052, loss=2.0329, lr=2.81e-05]
Training:  71%|███████   | 7861/11082 [25:16:15<10:24:04, 11.63s/step, acc=0.5052, loss=2.0329, lr=2.81e-05]
Training:  71%|███████   | 7862/11082 [25:16:26<10:29:44, 11.73s/step, acc=0.5052, loss=2.0329, lr=2.81e-05]
Training:  71%|███████   | 7863/11082 [25:16:38<10:26:23, 11.68s/step, acc=0.5052, loss=2.0329, lr=2.81e-05]
Training:  71%|███████   | 7864/11082 [25:16:49<10:19:06, 11.54s/step, acc=0.5052, loss=2.0329, lr=2.81e-05]
Training:  71%|███████   | 7865/11082 [25:17:01<10:26:33, 11.69s/step, acc=0.5052, loss=2.0329, lr=2.81e-05]
Training:  71%|███████   | 7866/11082 [25:17:13<10:20:32, 11.58s/step, acc=0.5052, loss=2.0329, lr=2.81e-05]
Training:  71%|███████   | 7867/11082 [25:17:24<10:18:36, 11.54s/step, acc=0.5052, loss=2.0329, lr=2.81e-05]
Training:  71%|███████   | 7868/11082 [25:17:36<10:17:23, 11.53s/step, acc=0.5052, loss=2.0329, lr=2.81e-05]
Training:  71%|███████   | 7869/11082 [25:17:48<10:25:15, 11.68s/step, acc=0.5052, loss=2.0329, lr=2.81e-05]
Training:  71%|███████   | 7870/11082 [25:17:59<10:14:42, 11.48s/step, acc=0.5052, loss=2.0329, lr=2.81e-05]
Training:  71%|███████   | 7871/11082 [25:18:10<10:19:30, 11.58s/step, acc=0.5052, loss=2.0329, lr=2.81e-05]
Training:  71%|███████   | 7872/11082 [25:18:22<10:27:09, 11.72s/step, acc=0.5052, loss=2.0329, lr=2.81e-05]
Training:  71%|███████   | 7873/11082 [25:18:34<10:24:14, 11.67s/step, acc=0.5052, loss=2.0329, lr=2.81e-05]
Training:  71%|███████   | 7874/11082 [25:18:46<10:25:11, 11.69s/step, acc=0.5052, loss=2.0329, lr=2.81e-05]
Training:  71%|███████   | 7875/11082 [25:18:57<10:24:17, 11.68s/step, acc=0.5052, loss=2.0329, lr=2.81e-05]
Training:  71%|███████   | 7876/11082 [25:19:09<10:17:24, 11.55s/step, acc=0.5052, loss=2.0329, lr=2.81e-05]
Training:  71%|███████   | 7877/11082 [25:19:20<10:18:56, 11.59s/step, acc=0.5052, loss=2.0329, lr=2.81e-05]
Training:  71%|███████   | 7878/11082 [25:19:32<10:19:30, 11.60s/step, acc=0.5052, loss=2.0329, lr=2.81e-05]
Training:  71%|███████   | 7879/11082 [25:19:44<10:25:01, 11.71s/step, acc=0.5052, loss=2.0329, lr=2.81e-05]
Training:  71%|███████   | 7880/11082 [25:19:56<10:23:19, 11.68s/step, acc=0.5052, loss=2.0329, lr=2.81e-05]
Training:  71%|███████   | 7881/11082 [25:20:07<10:22:59, 11.68s/step, acc=0.5052, loss=2.0329, lr=2.81e-05]
Training:  71%|███████   | 7882/11082 [25:20:18<10:15:41, 11.54s/step, acc=0.5052, loss=2.0329, lr=2.81e-05]
Training:  71%|███████   | 7883/11082 [25:20:30<10:22:56, 11.68s/step, acc=0.5052, loss=2.0329, lr=2.81e-05]
Training:  71%|███████   | 7884/11082 [25:20:42<10:23:49, 11.70s/step, acc=0.5052, loss=2.0329, lr=2.81e-05]
Training:  71%|███████   | 7885/11082 [25:20:54<10:17:19, 11.59s/step, acc=0.5052, loss=2.0329, lr=2.81e-05]
Training:  71%|███████   | 7886/11082 [25:21:05<10:13:11, 11.51s/step, acc=0.5052, loss=2.0329, lr=2.81e-05]
Training:  71%|███████   | 7887/11082 [25:21:17<10:22:54, 11.70s/step, acc=0.5052, loss=2.0329, lr=2.81e-05]
Training:  71%|███████   | 7888/11082 [25:21:28<10:14:34, 11.54s/step, acc=0.5052, loss=2.0329, lr=2.81e-05]
Training:  71%|███████   | 7889/11082 [25:21:40<10:14:45, 11.55s/step, acc=0.5052, loss=2.0329, lr=2.81e-05]
Training:  71%|███████   | 7890/11082 [25:21:51<10:17:10, 11.60s/step, acc=0.5052, loss=2.0329, lr=2.81e-05]
Training:  71%|███████   | 7891/11082 [25:22:03<10:20:04, 11.66s/step, acc=0.5052, loss=2.0329, lr=2.81e-05]
Training:  71%|███████   | 7892/11082 [25:22:15<10:19:46, 11.66s/step, acc=0.5052, loss=2.0329, lr=2.81e-05]
Training:  71%|███████   | 7893/11082 [25:22:27<10:20:04, 11.67s/step, acc=0.5052, loss=2.0329, lr=2.81e-05]
Training:  71%|███████   | 7894/11082 [25:22:38<10:08:21, 11.45s/step, acc=0.5052, loss=2.0329, lr=2.81e-05]
Training:  71%|███████   | 7895/11082 [25:22:49<10:05:48, 11.41s/step, acc=0.5052, loss=2.0329, lr=2.81e-05]
Training:  71%|███████▏  | 7896/11082 [25:23:01<10:20:27, 11.68s/step, acc=0.5052, loss=2.0329, lr=2.81e-05]
Training:  71%|███████▏  | 7897/11082 [25:23:13<10:16:21, 11.61s/step, acc=0.5052, loss=2.0329, lr=2.81e-05]
Training:  71%|███████▏  | 7898/11082 [25:23:25<10:20:37, 11.70s/step, acc=0.5052, loss=2.0329, lr=2.81e-05]
Training:  71%|███████▏  | 7899/11082 [25:23:36<10:19:57, 11.69s/step, acc=0.5052, loss=2.0329, lr=2.81e-05]
Training:  71%|███████▏  | 7900/11082 [25:23:48<10:14:57, 11.60s/step, acc=0.5052, loss=2.0329, lr=2.81e-05]
Training:  71%|███████▏  | 7900/11082 [25:23:55<10:14:57, 11.60s/step, acc=0.5060, loss=2.0216, lr=2.76e-05]2026-06-19 16:17:23 | INFO | __main__ | epoch=1 step=7900 train_loss=2.021644 train_acc=0.505991 lr=2.759e-05 grad_norm=0.6367 acc_d0=0.7706 acc_d1=0.6293 acc_d2=0.5101 acc_d3=0.4119 acc_d4=0.3314 acc_d5=0.2678 acc_d6=0.2172 acc_d7=0.1795

Training:  71%|███████▏  | 7901/11082 [25:23:59<10:18:35, 11.67s/step, acc=0.5060, loss=2.0216, lr=2.76e-05]
Training:  71%|███████▏  | 7902/11082 [25:24:11<10:17:41, 11.65s/step, acc=0.5060, loss=2.0216, lr=2.76e-05]
Training:  71%|███████▏  | 7903/11082 [25:24:23<10:19:26, 11.69s/step, acc=0.5060, loss=2.0216, lr=2.76e-05]
Training:  71%|███████▏  | 7904/11082 [25:24:34<10:16:56, 11.65s/step, acc=0.5060, loss=2.0216, lr=2.76e-05]
Training:  71%|███████▏  | 7905/11082 [25:24:46<10:16:28, 11.64s/step, acc=0.5060, loss=2.0216, lr=2.76e-05]
Training:  71%|███████▏  | 7906/11082 [25:24:58<10:20:29, 11.72s/step, acc=0.5060, loss=2.0216, lr=2.76e-05]
Training:  71%|███████▏  | 7907/11082 [25:25:10<10:19:10, 11.70s/step, acc=0.5060, loss=2.0216, lr=2.76e-05]
Training:  71%|███████▏  | 7908/11082 [25:25:21<10:07:54, 11.49s/step, acc=0.5060, loss=2.0216, lr=2.76e-05]
Training:  71%|███████▏  | 7909/11082 [25:25:33<10:15:29, 11.64s/step, acc=0.5060, loss=2.0216, lr=2.76e-05]
Training:  71%|███████▏  | 7910/11082 [25:25:44<10:13:22, 11.60s/step, acc=0.5060, loss=2.0216, lr=2.76e-05]
Training:  71%|███████▏  | 7911/11082 [25:25:56<10:16:27, 11.66s/step, acc=0.5060, loss=2.0216, lr=2.76e-05]
Training:  71%|███████▏  | 7912/11082 [25:26:08<10:17:49, 11.69s/step, acc=0.5060, loss=2.0216, lr=2.76e-05]
Training:  71%|███████▏  | 7913/11082 [25:26:19<10:06:31, 11.48s/step, acc=0.5060, loss=2.0216, lr=2.76e-05]
Training:  71%|███████▏  | 7914/11082 [25:26:31<10:15:32, 11.66s/step, acc=0.5060, loss=2.0216, lr=2.76e-05]
Training:  71%|███████▏  | 7915/11082 [25:26:42<10:09:33, 11.55s/step, acc=0.5060, loss=2.0216, lr=2.76e-05]
Training:  71%|███████▏  | 7916/11082 [25:26:54<10:14:18, 11.64s/step, acc=0.5060, loss=2.0216, lr=2.76e-05]
Training:  71%|███████▏  | 7917/11082 [25:27:05<10:09:23, 11.55s/step, acc=0.5060, loss=2.0216, lr=2.76e-05]
Training:  71%|███████▏  | 7918/11082 [25:27:17<10:13:07, 11.63s/step, acc=0.5060, loss=2.0216, lr=2.76e-05]
Training:  71%|███████▏  | 7919/11082 [25:27:29<10:14:14, 11.65s/step, acc=0.5060, loss=2.0216, lr=2.76e-05]
Training:  71%|███████▏  | 7920/11082 [25:27:40<10:08:54, 11.55s/step, acc=0.5060, loss=2.0216, lr=2.76e-05]
Training:  71%|███████▏  | 7921/11082 [25:27:52<10:18:56, 11.75s/step, acc=0.5060, loss=2.0216, lr=2.76e-05]
Training:  71%|███████▏  | 7922/11082 [25:28:04<10:12:27, 11.63s/step, acc=0.5060, loss=2.0216, lr=2.76e-05]
Training:  71%|███████▏  | 7923/11082 [25:28:15<10:12:39, 11.64s/step, acc=0.5060, loss=2.0216, lr=2.76e-05]
Training:  72%|███████▏  | 7924/11082 [25:28:27<10:15:12, 11.69s/step, acc=0.5060, loss=2.0216, lr=2.76e-05]
Training:  72%|███████▏  | 7925/11082 [25:28:38<10:08:23, 11.56s/step, acc=0.5060, loss=2.0216, lr=2.76e-05]
Training:  72%|███████▏  | 7926/11082 [25:28:50<10:10:43, 11.61s/step, acc=0.5060, loss=2.0216, lr=2.76e-05]
Training:  72%|███████▏  | 7927/11082 [25:29:02<10:09:12, 11.59s/step, acc=0.5060, loss=2.0216, lr=2.76e-05]
Training:  72%|███████▏  | 7928/11082 [25:29:13<10:08:05, 11.57s/step, acc=0.5060, loss=2.0216, lr=2.76e-05]
Training:  72%|███████▏  | 7929/11082 [25:29:25<10:16:40, 11.73s/step, acc=0.5060, loss=2.0216, lr=2.76e-05]
Training:  72%|███████▏  | 7930/11082 [25:29:37<10:15:39, 11.72s/step, acc=0.5060, loss=2.0216, lr=2.76e-05]
Training:  72%|███████▏  | 7931/11082 [25:29:48<10:08:09, 11.58s/step, acc=0.5060, loss=2.0216, lr=2.76e-05]
Training:  72%|███████▏  | 7932/11082 [25:30:00<10:10:10, 11.62s/step, acc=0.5060, loss=2.0216, lr=2.76e-05]
Training:  72%|███████▏  | 7933/11082 [25:30:11<10:08:26, 11.59s/step, acc=0.5060, loss=2.0216, lr=2.76e-05]
Training:  72%|███████▏  | 7934/11082 [25:30:23<10:14:34, 11.71s/step, acc=0.5060, loss=2.0216, lr=2.76e-05]
Training:  72%|███████▏  | 7935/11082 [25:30:35<10:06:07, 11.56s/step, acc=0.5060, loss=2.0216, lr=2.76e-05]
Training:  72%|███████▏  | 7936/11082 [25:30:46<10:10:31, 11.64s/step, acc=0.5060, loss=2.0216, lr=2.76e-05]
Training:  72%|███████▏  | 7937/11082 [25:30:58<10:09:07, 11.62s/step, acc=0.5060, loss=2.0216, lr=2.76e-05]
Training:  72%|███████▏  | 7938/11082 [25:31:10<10:10:52, 11.66s/step, acc=0.5060, loss=2.0216, lr=2.76e-05]
Training:  72%|███████▏  | 7939/11082 [25:31:21<10:00:06, 11.46s/step, acc=0.5060, loss=2.0216, lr=2.76e-05]
Training:  72%|███████▏  | 7940/11082 [25:31:33<10:06:14, 11.58s/step, acc=0.5060, loss=2.0216, lr=2.76e-05]
Training:  72%|███████▏  | 7941/11082 [25:31:44<10:09:36, 11.64s/step, acc=0.5060, loss=2.0216, lr=2.76e-05]
Training:  72%|███████▏  | 7942/11082 [25:31:56<10:05:23, 11.57s/step, acc=0.5060, loss=2.0216, lr=2.76e-05]
Training:  72%|███████▏  | 7943/11082 [25:32:08<10:10:49, 11.68s/step, acc=0.5060, loss=2.0216, lr=2.76e-05]
Training:  72%|███████▏  | 7944/11082 [25:32:19<10:04:44, 11.56s/step, acc=0.5060, loss=2.0216, lr=2.76e-05]
Training:  72%|███████▏  | 7945/11082 [25:32:31<10:06:53, 11.61s/step, acc=0.5060, loss=2.0216, lr=2.76e-05]
Training:  72%|███████▏  | 7946/11082 [25:32:42<10:05:26, 11.58s/step, acc=0.5060, loss=2.0216, lr=2.76e-05]
Training:  72%|███████▏  | 7947/11082 [25:32:54<10:10:34, 11.69s/step, acc=0.5060, loss=2.0216, lr=2.76e-05]
Training:  72%|███████▏  | 7948/11082 [25:33:06<10:05:45, 11.60s/step, acc=0.5060, loss=2.0216, lr=2.76e-05]
Training:  72%|███████▏  | 7949/11082 [25:33:17<10:10:15, 11.69s/step, acc=0.5060, loss=2.0216, lr=2.76e-05]
Training:  72%|███████▏  | 7950/11082 [25:33:29<10:01:18, 11.52s/step, acc=0.5060, loss=2.0216, lr=2.76e-05]
Training:  72%|███████▏  | 7950/11082 [25:33:36<10:01:18, 11.52s/step, acc=0.5067, loss=2.0152, lr=2.71e-05]2026-06-19 16:27:04 | INFO | __main__ | epoch=1 step=7950 train_loss=2.015203 train_acc=0.506704 lr=2.708e-05 grad_norm=0.5742 acc_d0=0.7669 acc_d1=0.6244 acc_d2=0.5047 acc_d3=0.4083 acc_d4=0.3288 acc_d5=0.2644 acc_d6=0.2172 acc_d7=0.1768

Training:  72%|███████▏  | 7951/11082 [25:33:41<10:10:30, 11.70s/step, acc=0.5067, loss=2.0152, lr=2.71e-05]
Training:  72%|███████▏  | 7952/11082 [25:33:52<10:05:15, 11.60s/step, acc=0.5067, loss=2.0152, lr=2.71e-05]
Training:  72%|███████▏  | 7953/11082 [25:34:04<10:07:55, 11.66s/step, acc=0.5067, loss=2.0152, lr=2.71e-05]
Training:  72%|███████▏  | 7954/11082 [25:34:15<10:01:50, 11.54s/step, acc=0.5067, loss=2.0152, lr=2.71e-05]
Training:  72%|███████▏  | 7955/11082 [25:34:27<10:02:48, 11.57s/step, acc=0.5067, loss=2.0152, lr=2.71e-05]
Training:  72%|███████▏  | 7956/11082 [25:34:38<10:02:44, 11.57s/step, acc=0.5067, loss=2.0152, lr=2.71e-05]
Training:  72%|███████▏  | 7957/11082 [25:34:50<10:06:00, 11.64s/step, acc=0.5067, loss=2.0152, lr=2.71e-05]
Training:  72%|███████▏  | 7958/11082 [25:35:02<10:05:08, 11.62s/step, acc=0.5067, loss=2.0152, lr=2.71e-05]
Training:  72%|███████▏  | 7959/11082 [25:35:13<10:01:49, 11.56s/step, acc=0.5067, loss=2.0152, lr=2.71e-05]
Training:  72%|███████▏  | 7960/11082 [25:35:25<10:00:19, 11.54s/step, acc=0.5067, loss=2.0152, lr=2.71e-05]
Training:  72%|███████▏  | 7961/11082 [25:35:37<10:07:29, 11.68s/step, acc=0.5067, loss=2.0152, lr=2.71e-05]
Training:  72%|███████▏  | 7962/11082 [25:35:49<10:12:17, 11.77s/step, acc=0.5067, loss=2.0152, lr=2.71e-05]
Training:  72%|███████▏  | 7963/11082 [25:36:00<10:10:12, 11.74s/step, acc=0.5067, loss=2.0152, lr=2.71e-05]
Training:  72%|███████▏  | 7964/11082 [25:36:11<10:02:14, 11.59s/step, acc=0.5067, loss=2.0152, lr=2.71e-05]
Training:  72%|███████▏  | 7965/11082 [25:36:23<10:01:48, 11.58s/step, acc=0.5067, loss=2.0152, lr=2.71e-05]
Training:  72%|███████▏  | 7966/11082 [25:36:35<10:08:41, 11.72s/step, acc=0.5067, loss=2.0152, lr=2.71e-05]
Training:  72%|███████▏  | 7967/11082 [25:36:46<10:02:46, 11.61s/step, acc=0.5067, loss=2.0152, lr=2.71e-05]
Training:  72%|███████▏  | 7968/11082 [25:36:58<10:01:23, 11.59s/step, acc=0.5067, loss=2.0152, lr=2.71e-05]
Training:  72%|███████▏  | 7969/11082 [25:37:09<9:55:43, 11.48s/step, acc=0.5067, loss=2.0152, lr=2.71e-05] 
Training:  72%|███████▏  | 7970/11082 [25:37:21<9:57:21, 11.52s/step, acc=0.5067, loss=2.0152, lr=2.71e-05]
Training:  72%|███████▏  | 7971/11082 [25:37:32<9:55:23, 11.48s/step, acc=0.5067, loss=2.0152, lr=2.71e-05]
Training:  72%|███████▏  | 7972/11082 [25:37:44<10:02:19, 11.62s/step, acc=0.5067, loss=2.0152, lr=2.71e-05]
Training:  72%|███████▏  | 7973/11082 [25:37:56<10:00:55, 11.60s/step, acc=0.5067, loss=2.0152, lr=2.71e-05]
Training:  72%|███████▏  | 7974/11082 [25:38:07<9:57:02, 11.53s/step, acc=0.5067, loss=2.0152, lr=2.71e-05] 
Training:  72%|███████▏  | 7975/11082 [25:38:19<10:03:16, 11.65s/step, acc=0.5067, loss=2.0152, lr=2.71e-05]
Training:  72%|███████▏  | 7976/11082 [25:38:31<10:00:44, 11.60s/step, acc=0.5067, loss=2.0152, lr=2.71e-05]
Training:  72%|███████▏  | 7977/11082 [25:38:42<10:00:00, 11.59s/step, acc=0.5067, loss=2.0152, lr=2.71e-05]
Training:  72%|███████▏  | 7978/11082 [25:38:54<10:00:50, 11.61s/step, acc=0.5067, loss=2.0152, lr=2.71e-05]
Training:  72%|███████▏  | 7979/11082 [25:39:05<10:00:39, 11.61s/step, acc=0.5067, loss=2.0152, lr=2.71e-05]
Training:  72%|███████▏  | 7980/11082 [25:39:17<9:59:38, 11.60s/step, acc=0.5067, loss=2.0152, lr=2.71e-05] 
Training:  72%|███████▏  | 7981/11082 [25:39:28<9:58:57, 11.59s/step, acc=0.5067, loss=2.0152, lr=2.71e-05]
Training:  72%|███████▏  | 7982/11082 [25:39:40<9:54:42, 11.51s/step, acc=0.5067, loss=2.0152, lr=2.71e-05]
Training:  72%|███████▏  | 7983/11082 [25:39:51<9:52:30, 11.47s/step, acc=0.5067, loss=2.0152, lr=2.71e-05]
Training:  72%|███████▏  | 7984/11082 [25:40:03<10:01:03, 11.64s/step, acc=0.5067, loss=2.0152, lr=2.71e-05]
Training:  72%|███████▏  | 7985/11082 [25:40:15<9:59:21, 11.61s/step, acc=0.5067, loss=2.0152, lr=2.71e-05] 
Training:  72%|███████▏  | 7986/11082 [25:40:26<9:57:15, 11.57s/step, acc=0.5067, loss=2.0152, lr=2.71e-05]
Training:  72%|███████▏  | 7987/11082 [25:40:38<9:55:39, 11.55s/step, acc=0.5067, loss=2.0152, lr=2.71e-05]
Training:  72%|███████▏  | 7988/11082 [25:40:50<10:06:29, 11.76s/step, acc=0.5067, loss=2.0152, lr=2.71e-05]
Training:  72%|███████▏  | 7989/11082 [25:41:02<10:03:09, 11.70s/step, acc=0.5067, loss=2.0152, lr=2.71e-05]
Training:  72%|███████▏  | 7990/11082 [25:41:13<10:01:55, 11.68s/step, acc=0.5067, loss=2.0152, lr=2.71e-05]
Training:  72%|███████▏  | 7991/11082 [25:41:25<9:59:25, 11.64s/step, acc=0.5067, loss=2.0152, lr=2.71e-05] 
Training:  72%|███████▏  | 7992/11082 [25:41:37<10:03:17, 11.71s/step, acc=0.5067, loss=2.0152, lr=2.71e-05]
Training:  72%|███████▏  | 7993/11082 [25:41:48<9:58:42, 11.63s/step, acc=0.5067, loss=2.0152, lr=2.71e-05] 
Training:  72%|███████▏  | 7994/11082 [25:41:59<9:52:56, 11.52s/step, acc=0.5067, loss=2.0152, lr=2.71e-05]
Training:  72%|███████▏  | 7995/11082 [25:42:11<9:56:13, 11.59s/step, acc=0.5067, loss=2.0152, lr=2.71e-05]
Training:  72%|███████▏  | 7996/11082 [25:42:23<9:58:34, 11.64s/step, acc=0.5067, loss=2.0152, lr=2.71e-05]
Training:  72%|███████▏  | 7997/11082 [25:42:34<9:54:42, 11.57s/step, acc=0.5067, loss=2.0152, lr=2.71e-05]
Training:  72%|███████▏  | 7998/11082 [25:42:46<9:56:53, 11.61s/step, acc=0.5067, loss=2.0152, lr=2.71e-05]
Training:  72%|███████▏  | 7999/11082 [25:42:58<9:57:41, 11.63s/step, acc=0.5067, loss=2.0152, lr=2.71e-05]
Training:  72%|███████▏  | 8000/11082 [25:43:09<9:56:55, 11.62s/step, acc=0.5067, loss=2.0152, lr=2.71e-05]
Training:  72%|███████▏  | 8000/11082 [25:43:16<9:56:55, 11.62s/step, acc=0.5063, loss=2.0202, lr=2.66e-05]2026-06-19 16:36:44 | INFO | __main__ | epoch=1 step=8000 train_loss=2.020244 train_acc=0.506318 lr=2.657e-05 grad_norm=0.5781 acc_d0=0.7697 acc_d1=0.6278 acc_d2=0.5070 acc_d3=0.4095 acc_d4=0.3291 acc_d5=0.2661 acc_d6=0.2167 acc_d7=0.1780

Training:  72%|███████▏  | 8001/11082 [25:43:21<9:54:50, 11.58s/step, acc=0.5063, loss=2.0202, lr=2.66e-05]
Training:  72%|███████▏  | 8002/11082 [25:43:32<9:56:07, 11.61s/step, acc=0.5063, loss=2.0202, lr=2.66e-05]
Training:  72%|███████▏  | 8003/11082 [25:43:44<10:01:52, 11.73s/step, acc=0.5063, loss=2.0202, lr=2.66e-05]
Training:  72%|███████▏  | 8004/11082 [25:43:56<10:05:49, 11.81s/step, acc=0.5063, loss=2.0202, lr=2.66e-05]
Training:  72%|███████▏  | 8005/11082 [25:44:08<9:59:45, 11.69s/step, acc=0.5063, loss=2.0202, lr=2.66e-05] 
Training:  72%|███████▏  | 8006/11082 [25:44:19<9:57:14, 11.65s/step, acc=0.5063, loss=2.0202, lr=2.66e-05]
Training:  72%|███████▏  | 8007/11082 [25:44:31<9:59:46, 11.70s/step, acc=0.5063, loss=2.0202, lr=2.66e-05]
Training:  72%|███████▏  | 8008/11082 [25:44:43<9:54:36, 11.61s/step, acc=0.5063, loss=2.0202, lr=2.66e-05]
Training:  72%|███████▏  | 8009/11082 [25:44:54<9:56:21, 11.64s/step, acc=0.5063, loss=2.0202, lr=2.66e-05]
Training:  72%|███████▏  | 8010/11082 [25:45:06<9:58:01, 11.68s/step, acc=0.5063, loss=2.0202, lr=2.66e-05]
Training:  72%|███████▏  | 8011/11082 [25:45:18<9:55:47, 11.64s/step, acc=0.5063, loss=2.0202, lr=2.66e-05]
Training:  72%|███████▏  | 8012/11082 [25:45:29<9:53:20, 11.60s/step, acc=0.5063, loss=2.0202, lr=2.66e-05]
Training:  72%|███████▏  | 8013/11082 [25:45:40<9:45:24, 11.44s/step, acc=0.5063, loss=2.0202, lr=2.66e-05]
Training:  72%|███████▏  | 8014/11082 [25:45:52<9:47:53, 11.50s/step, acc=0.5063, loss=2.0202, lr=2.66e-05]
Training:  72%|███████▏  | 8015/11082 [25:46:04<9:52:34, 11.59s/step, acc=0.5063, loss=2.0202, lr=2.66e-05]
Training:  72%|███████▏  | 8016/11082 [25:46:15<9:52:22, 11.59s/step, acc=0.5063, loss=2.0202, lr=2.66e-05]
Training:  72%|███████▏  | 8017/11082 [25:46:27<9:56:21, 11.67s/step, acc=0.5063, loss=2.0202, lr=2.66e-05]
Training:  72%|███████▏  | 8018/11082 [25:46:39<9:52:37, 11.60s/step, acc=0.5063, loss=2.0202, lr=2.66e-05]
Training:  72%|███████▏  | 8019/11082 [25:46:51<9:59:13, 11.74s/step, acc=0.5063, loss=2.0202, lr=2.66e-05]
Training:  72%|███████▏  | 8020/11082 [25:47:02<9:54:04, 11.64s/step, acc=0.5063, loss=2.0202, lr=2.66e-05]
Training:  72%|███████▏  | 8021/11082 [25:47:13<9:50:24, 11.57s/step, acc=0.5063, loss=2.0202, lr=2.66e-05]
Training:  72%|███████▏  | 8022/11082 [25:47:25<9:53:28, 11.64s/step, acc=0.5063, loss=2.0202, lr=2.66e-05]
Training:  72%|███████▏  | 8023/11082 [25:47:37<9:48:52, 11.55s/step, acc=0.5063, loss=2.0202, lr=2.66e-05]
Training:  72%|███████▏  | 8024/11082 [25:47:48<9:43:50, 11.46s/step, acc=0.5063, loss=2.0202, lr=2.66e-05]
Training:  72%|███████▏  | 8025/11082 [25:47:59<9:45:59, 11.50s/step, acc=0.5063, loss=2.0202, lr=2.66e-05]
Training:  72%|███████▏  | 8026/11082 [25:48:11<9:52:20, 11.63s/step, acc=0.5063, loss=2.0202, lr=2.66e-05]
Training:  72%|███████▏  | 8027/11082 [25:48:22<9:44:44, 11.48s/step, acc=0.5063, loss=2.0202, lr=2.66e-05]
Training:  72%|███████▏  | 8028/11082 [25:48:35<9:56:32, 11.72s/step, acc=0.5063, loss=2.0202, lr=2.66e-05]
Training:  72%|███████▏  | 8029/11082 [25:48:46<9:44:51, 11.49s/step, acc=0.5063, loss=2.0202, lr=2.66e-05]
Training:  72%|███████▏  | 8030/11082 [25:48:58<9:53:50, 11.67s/step, acc=0.5063, loss=2.0202, lr=2.66e-05]
Training:  72%|███████▏  | 8031/11082 [25:49:09<9:49:57, 11.60s/step, acc=0.5063, loss=2.0202, lr=2.66e-05]
Training:  72%|███████▏  | 8032/11082 [25:49:20<9:42:03, 11.45s/step, acc=0.5063, loss=2.0202, lr=2.66e-05]
Training:  72%|███████▏  | 8033/11082 [25:49:32<9:52:39, 11.66s/step, acc=0.5063, loss=2.0202, lr=2.66e-05]
Training:  72%|███████▏  | 8034/11082 [25:49:44<9:53:07, 11.68s/step, acc=0.5063, loss=2.0202, lr=2.66e-05]
Training:  73%|███████▎  | 8035/11082 [25:49:55<9:43:37, 11.49s/step, acc=0.5063, loss=2.0202, lr=2.66e-05]
Training:  73%|███████▎  | 8036/11082 [25:50:07<9:49:56, 11.62s/step, acc=0.5063, loss=2.0202, lr=2.66e-05]
Training:  73%|███████▎  | 8037/11082 [25:50:18<9:44:36, 11.52s/step, acc=0.5063, loss=2.0202, lr=2.66e-05]
Training:  73%|███████▎  | 8038/11082 [25:50:30<9:49:16, 11.62s/step, acc=0.5063, loss=2.0202, lr=2.66e-05]
Training:  73%|███████▎  | 8039/11082 [25:50:42<9:48:37, 11.61s/step, acc=0.5063, loss=2.0202, lr=2.66e-05]
Training:  73%|███████▎  | 8040/11082 [25:50:54<9:49:48, 11.63s/step, acc=0.5063, loss=2.0202, lr=2.66e-05]
Training:  73%|███████▎  | 8041/11082 [25:51:05<9:43:58, 11.52s/step, acc=0.5063, loss=2.0202, lr=2.66e-05]
Training:  73%|███████▎  | 8042/11082 [25:51:17<9:46:05, 11.57s/step, acc=0.5063, loss=2.0202, lr=2.66e-05]
Training:  73%|███████▎  | 8043/11082 [25:51:28<9:46:50, 11.59s/step, acc=0.5063, loss=2.0202, lr=2.66e-05]
Training:  73%|███████▎  | 8044/11082 [25:51:40<9:47:33, 11.60s/step, acc=0.5063, loss=2.0202, lr=2.66e-05]
Training:  73%|███████▎  | 8045/11082 [25:51:52<9:57:57, 11.81s/step, acc=0.5063, loss=2.0202, lr=2.66e-05]
Training:  73%|███████▎  | 8046/11082 [25:52:03<9:50:34, 11.67s/step, acc=0.5063, loss=2.0202, lr=2.66e-05]
Training:  73%|███████▎  | 8047/11082 [25:52:15<9:54:05, 11.74s/step, acc=0.5063, loss=2.0202, lr=2.66e-05]
Training:  73%|███████▎  | 8048/11082 [25:52:27<9:48:05, 11.63s/step, acc=0.5063, loss=2.0202, lr=2.66e-05]
Training:  73%|███████▎  | 8049/11082 [25:52:38<9:48:41, 11.65s/step, acc=0.5063, loss=2.0202, lr=2.66e-05]
Training:  73%|███████▎  | 8050/11082 [25:52:49<9:39:16, 11.46s/step, acc=0.5063, loss=2.0202, lr=2.66e-05]
Training:  73%|███████▎  | 8050/11082 [25:52:57<9:39:16, 11.46s/step, acc=0.5063, loss=2.0185, lr=2.61e-05]2026-06-19 16:46:25 | INFO | __main__ | epoch=1 step=8050 train_loss=2.018520 train_acc=0.506270 lr=2.607e-05 grad_norm=0.7031 acc_d0=0.7696 acc_d1=0.6291 acc_d2=0.5111 acc_d3=0.4143 acc_d4=0.3360 acc_d5=0.2718 acc_d6=0.2192 acc_d7=0.1804

Training:  73%|███████▎  | 8051/11082 [25:53:02<9:49:50, 11.68s/step, acc=0.5063, loss=2.0185, lr=2.61e-05]
Training:  73%|███████▎  | 8052/11082 [25:53:13<9:43:51, 11.56s/step, acc=0.5063, loss=2.0185, lr=2.61e-05]
Training:  73%|███████▎  | 8053/11082 [25:53:25<9:45:46, 11.60s/step, acc=0.5063, loss=2.0185, lr=2.61e-05]
Training:  73%|███████▎  | 8054/11082 [25:53:36<9:42:28, 11.54s/step, acc=0.5063, loss=2.0185, lr=2.61e-05]
Training:  73%|███████▎  | 8055/11082 [25:53:47<9:41:07, 11.52s/step, acc=0.5063, loss=2.0185, lr=2.61e-05]
Training:  73%|███████▎  | 8056/11082 [25:53:59<9:48:18, 11.67s/step, acc=0.5063, loss=2.0185, lr=2.61e-05]
Training:  73%|███████▎  | 8057/11082 [25:54:11<9:44:06, 11.59s/step, acc=0.5063, loss=2.0185, lr=2.61e-05]
Training:  73%|███████▎  | 8058/11082 [25:54:22<9:40:46, 11.52s/step, acc=0.5063, loss=2.0185, lr=2.61e-05]
Training:  73%|███████▎  | 8059/11082 [25:54:34<9:46:47, 11.65s/step, acc=0.5063, loss=2.0185, lr=2.61e-05]
Training:  73%|███████▎  | 8060/11082 [25:54:46<9:47:54, 11.67s/step, acc=0.5063, loss=2.0185, lr=2.61e-05]
Training:  73%|███████▎  | 8061/11082 [25:54:57<9:43:30, 11.59s/step, acc=0.5063, loss=2.0185, lr=2.61e-05]
Training:  73%|███████▎  | 8062/11082 [25:55:09<9:40:26, 11.53s/step, acc=0.5063, loss=2.0185, lr=2.61e-05]
Training:  73%|███████▎  | 8063/11082 [25:55:20<9:43:48, 11.60s/step, acc=0.5063, loss=2.0185, lr=2.61e-05]
Training:  73%|███████▎  | 8064/11082 [25:55:32<9:37:10, 11.47s/step, acc=0.5063, loss=2.0185, lr=2.61e-05]
Training:  73%|███████▎  | 8065/11082 [25:55:44<9:46:38, 11.67s/step, acc=0.5063, loss=2.0185, lr=2.61e-05]
Training:  73%|███████▎  | 8066/11082 [25:55:55<9:42:44, 11.59s/step, acc=0.5063, loss=2.0185, lr=2.61e-05]
Training:  73%|███████▎  | 8067/11082 [25:56:07<9:38:26, 11.51s/step, acc=0.5063, loss=2.0185, lr=2.61e-05]
Training:  73%|███████▎  | 8068/11082 [25:56:18<9:35:47, 11.46s/step, acc=0.5063, loss=2.0185, lr=2.61e-05]
Training:  73%|███████▎  | 8069/11082 [25:56:30<9:44:58, 11.65s/step, acc=0.5063, loss=2.0185, lr=2.61e-05]
Training:  73%|███████▎  | 8070/11082 [25:56:41<9:39:51, 11.55s/step, acc=0.5063, loss=2.0185, lr=2.61e-05]
Training:  73%|███████▎  | 8071/11082 [25:56:53<9:47:32, 11.71s/step, acc=0.5063, loss=2.0185, lr=2.61e-05]
Training:  73%|███████▎  | 8072/11082 [25:57:05<9:41:26, 11.59s/step, acc=0.5063, loss=2.0185, lr=2.61e-05]
Training:  73%|███████▎  | 8073/11082 [25:57:17<9:51:39, 11.80s/step, acc=0.5063, loss=2.0185, lr=2.61e-05]
Training:  73%|███████▎  | 8074/11082 [25:57:28<9:44:33, 11.66s/step, acc=0.5063, loss=2.0185, lr=2.61e-05]
Training:  73%|███████▎  | 8075/11082 [25:57:40<9:42:17, 11.62s/step, acc=0.5063, loss=2.0185, lr=2.61e-05]
Training:  73%|███████▎  | 8076/11082 [25:57:51<9:35:21, 11.48s/step, acc=0.5063, loss=2.0185, lr=2.61e-05]
Training:  73%|███████▎  | 8077/11082 [25:58:03<9:43:13, 11.65s/step, acc=0.5063, loss=2.0185, lr=2.61e-05]
Training:  73%|███████▎  | 8078/11082 [25:58:15<9:46:56, 11.72s/step, acc=0.5063, loss=2.0185, lr=2.61e-05]
Training:  73%|███████▎  | 8079/11082 [25:58:27<9:45:58, 11.71s/step, acc=0.5063, loss=2.0185, lr=2.61e-05]
Training:  73%|███████▎  | 8080/11082 [25:58:38<9:37:45, 11.55s/step, acc=0.5063, loss=2.0185, lr=2.61e-05]
Training:  73%|███████▎  | 8081/11082 [25:58:50<9:44:31, 11.69s/step, acc=0.5063, loss=2.0185, lr=2.61e-05]
Training:  73%|███████▎  | 8082/11082 [25:59:01<9:41:08, 11.62s/step, acc=0.5063, loss=2.0185, lr=2.61e-05]
Training:  73%|███████▎  | 8083/11082 [25:59:13<9:43:25, 11.67s/step, acc=0.5063, loss=2.0185, lr=2.61e-05]
Training:  73%|███████▎  | 8084/11082 [25:59:25<9:43:43, 11.68s/step, acc=0.5063, loss=2.0185, lr=2.61e-05]
Training:  73%|███████▎  | 8085/11082 [25:59:36<9:44:14, 11.70s/step, acc=0.5063, loss=2.0185, lr=2.61e-05]
Training:  73%|███████▎  | 8086/11082 [25:59:48<9:40:44, 11.63s/step, acc=0.5063, loss=2.0185, lr=2.61e-05]
Training:  73%|███████▎  | 8087/11082 [25:59:59<9:37:51, 11.58s/step, acc=0.5063, loss=2.0185, lr=2.61e-05]
Training:  73%|███████▎  | 8088/11082 [26:00:11<9:39:41, 11.62s/step, acc=0.5063, loss=2.0185, lr=2.61e-05]
Training:  73%|███████▎  | 8089/11082 [26:00:23<9:39:15, 11.61s/step, acc=0.5063, loss=2.0185, lr=2.61e-05]
Training:  73%|███████▎  | 8090/11082 [26:00:34<9:34:11, 11.51s/step, acc=0.5063, loss=2.0185, lr=2.61e-05]
Training:  73%|███████▎  | 8091/11082 [26:00:45<9:33:57, 11.51s/step, acc=0.5063, loss=2.0185, lr=2.61e-05]
Training:  73%|███████▎  | 8092/11082 [26:00:57<9:38:16, 11.60s/step, acc=0.5063, loss=2.0185, lr=2.61e-05]
Training:  73%|███████▎  | 8093/11082 [26:01:09<9:42:00, 11.68s/step, acc=0.5063, loss=2.0185, lr=2.61e-05]
Training:  73%|███████▎  | 8094/11082 [26:01:21<9:37:01, 11.59s/step, acc=0.5063, loss=2.0185, lr=2.61e-05]
Training:  73%|███████▎  | 8095/11082 [26:01:32<9:33:29, 11.52s/step, acc=0.5063, loss=2.0185, lr=2.61e-05]
Training:  73%|███████▎  | 8096/11082 [26:01:44<9:36:25, 11.58s/step, acc=0.5063, loss=2.0185, lr=2.61e-05]
Training:  73%|███████▎  | 8097/11082 [26:01:55<9:29:40, 11.45s/step, acc=0.5063, loss=2.0185, lr=2.61e-05]
Training:  73%|███████▎  | 8098/11082 [26:02:07<9:43:48, 11.74s/step, acc=0.5063, loss=2.0185, lr=2.61e-05]
Training:  73%|███████▎  | 8099/11082 [26:02:19<9:38:52, 11.64s/step, acc=0.5063, loss=2.0185, lr=2.61e-05]
Training:  73%|███████▎  | 8100/11082 [26:02:30<9:37:23, 11.62s/step, acc=0.5063, loss=2.0185, lr=2.61e-05]
Training:  73%|███████▎  | 8100/11082 [26:02:37<9:37:23, 11.62s/step, acc=0.5064, loss=2.0222, lr=2.56e-05]2026-06-19 16:56:05 | INFO | __main__ | epoch=1 step=8100 train_loss=2.022205 train_acc=0.506443 lr=2.558e-05 grad_norm=0.6445 acc_d0=0.7672 acc_d1=0.6271 acc_d2=0.5088 acc_d3=0.4113 acc_d4=0.3331 acc_d5=0.2700 acc_d6=0.2182 acc_d7=0.1794

Training:  73%|███████▎  | 8101/11082 [26:02:42<9:39:09, 11.66s/step, acc=0.5064, loss=2.0222, lr=2.56e-05]
Training:  73%|███████▎  | 8102/11082 [26:02:53<9:33:30, 11.55s/step, acc=0.5064, loss=2.0222, lr=2.56e-05]
Training:  73%|███████▎  | 8103/11082 [26:03:05<9:36:45, 11.62s/step, acc=0.5064, loss=2.0222, lr=2.56e-05]
Training:  73%|███████▎  | 8104/11082 [26:03:16<9:30:54, 11.50s/step, acc=0.5064, loss=2.0222, lr=2.56e-05]
Training:  73%|███████▎  | 8105/11082 [26:03:28<9:36:14, 11.61s/step, acc=0.5064, loss=2.0222, lr=2.56e-05]
Training:  73%|███████▎  | 8106/11082 [26:03:39<9:31:27, 11.52s/step, acc=0.5064, loss=2.0222, lr=2.56e-05]
Training:  73%|███████▎  | 8107/11082 [26:03:51<9:29:31, 11.49s/step, acc=0.5064, loss=2.0222, lr=2.56e-05]
Training:  73%|███████▎  | 8108/11082 [26:04:03<9:41:46, 11.74s/step, acc=0.5064, loss=2.0222, lr=2.56e-05]
Training:  73%|███████▎  | 8109/11082 [26:04:14<9:31:52, 11.54s/step, acc=0.5064, loss=2.0222, lr=2.56e-05]
Training:  73%|███████▎  | 8110/11082 [26:04:26<9:38:04, 11.67s/step, acc=0.5064, loss=2.0222, lr=2.56e-05]
Training:  73%|███████▎  | 8111/11082 [26:04:38<9:35:25, 11.62s/step, acc=0.5064, loss=2.0222, lr=2.56e-05]
Training:  73%|███████▎  | 8112/11082 [26:04:49<9:37:37, 11.67s/step, acc=0.5064, loss=2.0222, lr=2.56e-05]
Training:  73%|███████▎  | 8113/11082 [26:05:01<9:31:01, 11.54s/step, acc=0.5064, loss=2.0222, lr=2.56e-05]
Training:  73%|███████▎  | 8114/11082 [26:05:13<9:36:49, 11.66s/step, acc=0.5064, loss=2.0222, lr=2.56e-05]
Training:  73%|███████▎  | 8115/11082 [26:05:24<9:30:07, 11.53s/step, acc=0.5064, loss=2.0222, lr=2.56e-05]
Training:  73%|███████▎  | 8116/11082 [26:05:36<9:34:03, 11.61s/step, acc=0.5064, loss=2.0222, lr=2.56e-05]
Training:  73%|███████▎  | 8117/11082 [26:05:48<9:37:26, 11.69s/step, acc=0.5064, loss=2.0222, lr=2.56e-05]
Training:  73%|███████▎  | 8118/11082 [26:05:59<9:29:30, 11.53s/step, acc=0.5064, loss=2.0222, lr=2.56e-05]
Training:  73%|███████▎  | 8119/11082 [26:06:10<9:31:26, 11.57s/step, acc=0.5064, loss=2.0222, lr=2.56e-05]
Training:  73%|███████▎  | 8120/11082 [26:06:22<9:31:52, 11.58s/step, acc=0.5064, loss=2.0222, lr=2.56e-05]
Training:  73%|███████▎  | 8121/11082 [26:06:34<9:33:51, 11.63s/step, acc=0.5064, loss=2.0222, lr=2.56e-05]
Training:  73%|███████▎  | 8122/11082 [26:06:45<9:35:18, 11.66s/step, acc=0.5064, loss=2.0222, lr=2.56e-05]
Training:  73%|███████▎  | 8123/11082 [26:06:57<9:32:52, 11.62s/step, acc=0.5064, loss=2.0222, lr=2.56e-05]
Training:  73%|███████▎  | 8124/11082 [26:07:08<9:31:16, 11.59s/step, acc=0.5064, loss=2.0222, lr=2.56e-05]
Training:  73%|███████▎  | 8125/11082 [26:07:20<9:27:43, 11.52s/step, acc=0.5064, loss=2.0222, lr=2.56e-05]
Training:  73%|███████▎  | 8126/11082 [26:07:32<9:30:10, 11.57s/step, acc=0.5064, loss=2.0222, lr=2.56e-05]
Training:  73%|███████▎  | 8127/11082 [26:07:43<9:32:55, 11.63s/step, acc=0.5064, loss=2.0222, lr=2.56e-05]
Training:  73%|███████▎  | 8128/11082 [26:07:54<9:25:58, 11.50s/step, acc=0.5064, loss=2.0222, lr=2.56e-05]
Training:  73%|███████▎  | 8129/11082 [26:08:06<9:29:01, 11.56s/step, acc=0.5064, loss=2.0222, lr=2.56e-05]
Training:  73%|███████▎  | 8130/11082 [26:08:18<9:27:05, 11.53s/step, acc=0.5064, loss=2.0222, lr=2.56e-05]
Training:  73%|███████▎  | 8131/11082 [26:08:29<9:25:19, 11.49s/step, acc=0.5064, loss=2.0222, lr=2.56e-05]
Training:  73%|███████▎  | 8132/11082 [26:08:41<9:29:54, 11.59s/step, acc=0.5064, loss=2.0222, lr=2.56e-05]
Training:  73%|███████▎  | 8133/11082 [26:08:52<9:26:48, 11.53s/step, acc=0.5064, loss=2.0222, lr=2.56e-05]
Training:  73%|███████▎  | 8134/11082 [26:09:04<9:30:27, 11.61s/step, acc=0.5064, loss=2.0222, lr=2.56e-05]
Training:  73%|███████▎  | 8135/11082 [26:09:16<9:34:59, 11.71s/step, acc=0.5064, loss=2.0222, lr=2.56e-05]
Training:  73%|███████▎  | 8136/11082 [26:09:27<9:26:52, 11.55s/step, acc=0.5064, loss=2.0222, lr=2.56e-05]
Training:  73%|███████▎  | 8137/11082 [26:09:39<9:34:22, 11.70s/step, acc=0.5064, loss=2.0222, lr=2.56e-05]
Training:  73%|███████▎  | 8138/11082 [26:09:51<9:28:34, 11.59s/step, acc=0.5064, loss=2.0222, lr=2.56e-05]
Training:  73%|███████▎  | 8139/11082 [26:10:02<9:30:35, 11.63s/step, acc=0.5064, loss=2.0222, lr=2.56e-05]
Training:  73%|███████▎  | 8140/11082 [26:10:14<9:25:52, 11.54s/step, acc=0.5064, loss=2.0222, lr=2.56e-05]
Training:  73%|███████▎  | 8141/11082 [26:10:26<9:36:35, 11.76s/step, acc=0.5064, loss=2.0222, lr=2.56e-05]
Training:  73%|███████▎  | 8142/11082 [26:10:37<9:32:45, 11.69s/step, acc=0.5064, loss=2.0222, lr=2.56e-05]
Training:  73%|███████▎  | 8143/11082 [26:10:49<9:26:16, 11.56s/step, acc=0.5064, loss=2.0222, lr=2.56e-05]
Training:  73%|███████▎  | 8144/11082 [26:11:00<9:27:50, 11.60s/step, acc=0.5064, loss=2.0222, lr=2.56e-05]
Training:  73%|███████▎  | 8145/11082 [26:11:12<9:30:40, 11.66s/step, acc=0.5064, loss=2.0222, lr=2.56e-05]
Training:  74%|███████▎  | 8146/11082 [26:11:24<9:27:58, 11.61s/step, acc=0.5064, loss=2.0222, lr=2.56e-05]
Training:  74%|███████▎  | 8147/11082 [26:11:35<9:27:58, 11.61s/step, acc=0.5064, loss=2.0222, lr=2.56e-05]
Training:  74%|███████▎  | 8148/11082 [26:11:46<9:21:14, 11.48s/step, acc=0.5064, loss=2.0222, lr=2.56e-05]
Training:  74%|███████▎  | 8149/11082 [26:11:58<9:27:50, 11.62s/step, acc=0.5064, loss=2.0222, lr=2.56e-05]
Training:  74%|███████▎  | 8150/11082 [26:12:10<9:26:23, 11.59s/step, acc=0.5064, loss=2.0222, lr=2.56e-05]
Training:  74%|███████▎  | 8150/11082 [26:12:17<9:26:23, 11.59s/step, acc=0.5065, loss=2.0192, lr=2.51e-05]2026-06-19 17:05:45 | INFO | __main__ | epoch=1 step=8150 train_loss=2.019228 train_acc=0.506520 lr=2.509e-05 grad_norm=0.6406 acc_d0=0.7622 acc_d1=0.6170 acc_d2=0.4952 acc_d3=0.3964 acc_d4=0.3174 acc_d5=0.2557 acc_d6=0.2055 acc_d7=0.1692

Training:  74%|███████▎  | 8151/11082 [26:12:22<9:28:05, 11.63s/step, acc=0.5065, loss=2.0192, lr=2.51e-05]
Training:  74%|███████▎  | 8152/11082 [26:12:33<9:22:00, 11.51s/step, acc=0.5065, loss=2.0192, lr=2.51e-05]
Training:  74%|███████▎  | 8153/11082 [26:12:45<9:26:58, 11.61s/step, acc=0.5065, loss=2.0192, lr=2.51e-05]
Training:  74%|███████▎  | 8154/11082 [26:12:57<9:29:24, 11.67s/step, acc=0.5065, loss=2.0192, lr=2.51e-05]
Training:  74%|███████▎  | 8155/11082 [26:13:08<9:24:39, 11.57s/step, acc=0.5065, loss=2.0192, lr=2.51e-05]
Training:  74%|███████▎  | 8156/11082 [26:13:20<9:26:09, 11.61s/step, acc=0.5065, loss=2.0192, lr=2.51e-05]
Training:  74%|███████▎  | 8157/11082 [26:13:31<9:28:04, 11.65s/step, acc=0.5065, loss=2.0192, lr=2.51e-05]
Training:  74%|███████▎  | 8158/11082 [26:13:42<9:19:48, 11.49s/step, acc=0.5065, loss=2.0192, lr=2.51e-05]
Training:  74%|███████▎  | 8159/11082 [26:13:54<9:25:48, 11.61s/step, acc=0.5065, loss=2.0192, lr=2.51e-05]
Training:  74%|███████▎  | 8160/11082 [26:14:06<9:25:52, 11.62s/step, acc=0.5065, loss=2.0192, lr=2.51e-05]
Training:  74%|███████▎  | 8161/11082 [26:14:18<9:31:59, 11.75s/step, acc=0.5065, loss=2.0192, lr=2.51e-05]
Training:  74%|███████▎  | 8162/11082 [26:14:29<9:27:09, 11.65s/step, acc=0.5065, loss=2.0192, lr=2.51e-05]
Training:  74%|███████▎  | 8163/11082 [26:14:41<9:24:15, 11.60s/step, acc=0.5065, loss=2.0192, lr=2.51e-05]
Training:  74%|███████▎  | 8164/11082 [26:14:53<9:31:29, 11.75s/step, acc=0.5065, loss=2.0192, lr=2.51e-05]
Training:  74%|███████▎  | 8165/11082 [26:15:04<9:26:07, 11.64s/step, acc=0.5065, loss=2.0192, lr=2.51e-05]
Training:  74%|███████▎  | 8166/11082 [26:15:17<9:42:06, 11.98s/step, acc=0.5065, loss=2.0192, lr=2.51e-05]
Training:  74%|███████▎  | 8167/11082 [26:15:29<9:35:06, 11.84s/step, acc=0.5065, loss=2.0192, lr=2.51e-05]
Training:  74%|███████▎  | 8168/11082 [26:15:40<9:25:46, 11.65s/step, acc=0.5065, loss=2.0192, lr=2.51e-05]
Training:  74%|███████▎  | 8169/11082 [26:15:51<9:24:27, 11.63s/step, acc=0.5065, loss=2.0192, lr=2.51e-05]
Training:  74%|███████▎  | 8170/11082 [26:16:03<9:23:15, 11.61s/step, acc=0.5065, loss=2.0192, lr=2.51e-05]
Training:  74%|███████▎  | 8171/11082 [26:16:15<9:27:16, 11.69s/step, acc=0.5065, loss=2.0192, lr=2.51e-05]
Training:  74%|███████▎  | 8172/11082 [26:16:26<9:21:30, 11.58s/step, acc=0.5065, loss=2.0192, lr=2.51e-05]
Training:  74%|███████▍  | 8173/11082 [26:16:38<9:23:15, 11.62s/step, acc=0.5065, loss=2.0192, lr=2.51e-05]
Training:  74%|███████▍  | 8174/11082 [26:16:49<9:21:52, 11.59s/step, acc=0.5065, loss=2.0192, lr=2.51e-05]
Training:  74%|███████▍  | 8175/11082 [26:17:01<9:21:03, 11.58s/step, acc=0.5065, loss=2.0192, lr=2.51e-05]
Training:  74%|███████▍  | 8176/11082 [26:17:12<9:19:08, 11.54s/step, acc=0.5065, loss=2.0192, lr=2.51e-05]
Training:  74%|███████▍  | 8177/11082 [26:17:24<9:11:33, 11.39s/step, acc=0.5065, loss=2.0192, lr=2.51e-05]
Training:  74%|███████▍  | 8178/11082 [26:17:35<9:19:55, 11.57s/step, acc=0.5065, loss=2.0192, lr=2.51e-05]
Training:  74%|███████▍  | 8179/11082 [26:17:47<9:22:38, 11.63s/step, acc=0.5065, loss=2.0192, lr=2.51e-05]
Training:  74%|███████▍  | 8180/11082 [26:17:59<9:23:29, 11.65s/step, acc=0.5065, loss=2.0192, lr=2.51e-05]
Training:  74%|███████▍  | 8181/11082 [26:18:11<9:28:11, 11.75s/step, acc=0.5065, loss=2.0192, lr=2.51e-05]
Training:  74%|███████▍  | 8182/11082 [26:18:22<9:17:51, 11.54s/step, acc=0.5065, loss=2.0192, lr=2.51e-05]
Training:  74%|███████▍  | 8183/11082 [26:18:34<9:22:45, 11.65s/step, acc=0.5065, loss=2.0192, lr=2.51e-05]
Training:  74%|███████▍  | 8184/11082 [26:18:45<9:20:37, 11.61s/step, acc=0.5065, loss=2.0192, lr=2.51e-05]
Training:  74%|███████▍  | 8185/11082 [26:18:57<9:20:25, 11.61s/step, acc=0.5065, loss=2.0192, lr=2.51e-05]
Training:  74%|███████▍  | 8186/11082 [26:19:09<9:23:23, 11.67s/step, acc=0.5065, loss=2.0192, lr=2.51e-05]
Training:  74%|███████▍  | 8187/11082 [26:19:20<9:18:14, 11.57s/step, acc=0.5065, loss=2.0192, lr=2.51e-05]
Training:  74%|███████▍  | 8188/11082 [26:19:32<9:19:33, 11.60s/step, acc=0.5065, loss=2.0192, lr=2.51e-05]
Training:  74%|███████▍  | 8189/11082 [26:19:44<9:26:10, 11.74s/step, acc=0.5065, loss=2.0192, lr=2.51e-05]
Training:  74%|███████▍  | 8190/11082 [26:19:55<9:19:01, 11.60s/step, acc=0.5065, loss=2.0192, lr=2.51e-05]
Training:  74%|███████▍  | 8191/11082 [26:20:07<9:22:59, 11.68s/step, acc=0.5065, loss=2.0192, lr=2.51e-05]
Training:  74%|███████▍  | 8192/11082 [26:20:18<9:18:27, 11.59s/step, acc=0.5065, loss=2.0192, lr=2.51e-05]
Training:  74%|███████▍  | 8193/11082 [26:20:30<9:17:34, 11.58s/step, acc=0.5065, loss=2.0192, lr=2.51e-05]
Training:  74%|███████▍  | 8194/11082 [26:20:42<9:22:42, 11.69s/step, acc=0.5065, loss=2.0192, lr=2.51e-05]
Training:  74%|███████▍  | 8195/11082 [26:20:54<9:22:18, 11.69s/step, acc=0.5065, loss=2.0192, lr=2.51e-05]
Training:  74%|███████▍  | 8196/11082 [26:21:05<9:19:01, 11.62s/step, acc=0.5065, loss=2.0192, lr=2.51e-05]
Training:  74%|███████▍  | 8197/11082 [26:21:16<9:15:33, 11.55s/step, acc=0.5065, loss=2.0192, lr=2.51e-05]
Training:  74%|███████▍  | 8198/11082 [26:21:28<9:17:58, 11.61s/step, acc=0.5065, loss=2.0192, lr=2.51e-05]
Training:  74%|███████▍  | 8199/11082 [26:21:40<9:17:50, 11.61s/step, acc=0.5065, loss=2.0192, lr=2.51e-05]
Training:  74%|███████▍  | 8200/11082 [26:21:51<9:10:38, 11.46s/step, acc=0.5065, loss=2.0192, lr=2.51e-05]
Training:  74%|███████▍  | 8200/11082 [26:21:59<9:10:38, 11.46s/step, acc=0.5070, loss=2.0169, lr=2.46e-05]2026-06-19 17:15:27 | INFO | __main__ | epoch=1 step=8200 train_loss=2.016894 train_acc=0.507035 lr=2.461e-05 grad_norm=0.6016 acc_d0=0.7635 acc_d1=0.6201 acc_d2=0.5005 acc_d3=0.4040 acc_d4=0.3272 acc_d5=0.2642 acc_d6=0.2150 acc_d7=0.1752

Training:  74%|███████▍  | 8201/11082 [26:22:03<9:17:47, 11.62s/step, acc=0.5070, loss=2.0169, lr=2.46e-05]
Training:  74%|███████▍  | 8202/11082 [26:22:15<9:22:53, 11.73s/step, acc=0.5070, loss=2.0169, lr=2.46e-05]
Training:  74%|███████▍  | 8203/11082 [26:22:26<9:18:38, 11.64s/step, acc=0.5070, loss=2.0169, lr=2.46e-05]
Training:  74%|███████▍  | 8204/11082 [26:22:38<9:18:35, 11.65s/step, acc=0.5070, loss=2.0169, lr=2.46e-05]
Training:  74%|███████▍  | 8205/11082 [26:22:50<9:17:42, 11.63s/step, acc=0.5070, loss=2.0169, lr=2.46e-05]
Training:  74%|███████▍  | 8206/11082 [26:23:01<9:18:37, 11.65s/step, acc=0.5070, loss=2.0169, lr=2.46e-05]
Training:  74%|███████▍  | 8207/11082 [26:23:13<9:16:37, 11.62s/step, acc=0.5070, loss=2.0169, lr=2.46e-05]
Training:  74%|███████▍  | 8208/11082 [26:23:24<9:14:10, 11.57s/step, acc=0.5070, loss=2.0169, lr=2.46e-05]
Training:  74%|███████▍  | 8209/11082 [26:23:36<9:15:28, 11.60s/step, acc=0.5070, loss=2.0169, lr=2.46e-05]
Training:  74%|███████▍  | 8210/11082 [26:23:47<9:06:09, 11.41s/step, acc=0.5070, loss=2.0169, lr=2.46e-05]
Training:  74%|███████▍  | 8211/11082 [26:23:59<9:12:44, 11.55s/step, acc=0.5070, loss=2.0169, lr=2.46e-05]
Training:  74%|███████▍  | 8212/11082 [26:24:10<9:13:28, 11.57s/step, acc=0.5070, loss=2.0169, lr=2.46e-05]
Training:  74%|███████▍  | 8213/11082 [26:24:22<9:16:36, 11.64s/step, acc=0.5070, loss=2.0169, lr=2.46e-05]
Training:  74%|███████▍  | 8214/11082 [26:24:34<9:18:17, 11.68s/step, acc=0.5070, loss=2.0169, lr=2.46e-05]
Training:  74%|███████▍  | 8215/11082 [26:24:45<9:07:14, 11.45s/step, acc=0.5070, loss=2.0169, lr=2.46e-05]
Training:  74%|███████▍  | 8216/11082 [26:24:57<9:15:13, 11.62s/step, acc=0.5070, loss=2.0169, lr=2.46e-05]
Training:  74%|███████▍  | 8217/11082 [26:25:09<9:16:55, 11.66s/step, acc=0.5070, loss=2.0169, lr=2.46e-05]
Training:  74%|███████▍  | 8218/11082 [26:25:21<9:18:55, 11.71s/step, acc=0.5070, loss=2.0169, lr=2.46e-05]
Training:  74%|███████▍  | 8219/11082 [26:25:32<9:20:41, 11.75s/step, acc=0.5070, loss=2.0169, lr=2.46e-05]
Training:  74%|███████▍  | 8220/11082 [26:25:44<9:17:06, 11.68s/step, acc=0.5070, loss=2.0169, lr=2.46e-05]
Training:  74%|███████▍  | 8221/11082 [26:25:55<9:11:06, 11.56s/step, acc=0.5070, loss=2.0169, lr=2.46e-05]
Training:  74%|███████▍  | 8222/11082 [26:26:07<9:12:41, 11.60s/step, acc=0.5070, loss=2.0169, lr=2.46e-05]
Training:  74%|███████▍  | 8223/11082 [26:26:19<9:16:57, 11.69s/step, acc=0.5070, loss=2.0169, lr=2.46e-05]
Training:  74%|███████▍  | 8224/11082 [26:26:31<9:20:11, 11.76s/step, acc=0.5070, loss=2.0169, lr=2.46e-05]
Training:  74%|███████▍  | 8225/11082 [26:26:41<9:05:39, 11.46s/step, acc=0.5070, loss=2.0169, lr=2.46e-05]
Training:  74%|███████▍  | 8226/11082 [26:26:53<9:11:48, 11.59s/step, acc=0.5070, loss=2.0169, lr=2.46e-05]
Training:  74%|███████▍  | 8227/11082 [26:27:05<9:09:57, 11.56s/step, acc=0.5070, loss=2.0169, lr=2.46e-05]
Training:  74%|███████▍  | 8228/11082 [26:27:17<9:12:16, 11.61s/step, acc=0.5070, loss=2.0169, lr=2.46e-05]
Training:  74%|███████▍  | 8229/11082 [26:27:28<9:13:34, 11.64s/step, acc=0.5070, loss=2.0169, lr=2.46e-05]
Training:  74%|███████▍  | 8230/11082 [26:27:39<9:06:42, 11.50s/step, acc=0.5070, loss=2.0169, lr=2.46e-05]
Training:  74%|███████▍  | 8231/11082 [26:27:51<9:13:06, 11.64s/step, acc=0.5070, loss=2.0169, lr=2.46e-05]
Training:  74%|███████▍  | 8232/11082 [26:28:03<9:11:57, 11.62s/step, acc=0.5070, loss=2.0169, lr=2.46e-05]
Training:  74%|███████▍  | 8233/11082 [26:28:15<9:13:24, 11.65s/step, acc=0.5070, loss=2.0169, lr=2.46e-05]
Training:  74%|███████▍  | 8234/11082 [26:28:27<9:18:13, 11.76s/step, acc=0.5070, loss=2.0169, lr=2.46e-05]
Training:  74%|███████▍  | 8235/11082 [26:28:38<9:11:23, 11.62s/step, acc=0.5070, loss=2.0169, lr=2.46e-05]
Training:  74%|███████▍  | 8236/11082 [26:28:50<9:12:04, 11.64s/step, acc=0.5070, loss=2.0169, lr=2.46e-05]
Training:  74%|███████▍  | 8237/11082 [26:29:01<9:12:35, 11.65s/step, acc=0.5070, loss=2.0169, lr=2.46e-05]
Training:  74%|███████▍  | 8238/11082 [26:29:13<9:07:56, 11.56s/step, acc=0.5070, loss=2.0169, lr=2.46e-05]
Training:  74%|███████▍  | 8239/11082 [26:29:25<9:10:40, 11.62s/step, acc=0.5070, loss=2.0169, lr=2.46e-05]
Training:  74%|███████▍  | 8240/11082 [26:29:36<9:15:47, 11.73s/step, acc=0.5070, loss=2.0169, lr=2.46e-05]
Training:  74%|███████▍  | 8241/11082 [26:29:48<9:10:02, 11.62s/step, acc=0.5070, loss=2.0169, lr=2.46e-05]
Training:  74%|███████▍  | 8242/11082 [26:30:00<9:10:30, 11.63s/step, acc=0.5070, loss=2.0169, lr=2.46e-05]
Training:  74%|███████▍  | 8243/11082 [26:30:11<9:10:55, 11.64s/step, acc=0.5070, loss=2.0169, lr=2.46e-05]
Training:  74%|███████▍  | 8244/11082 [26:30:23<9:11:52, 11.67s/step, acc=0.5070, loss=2.0169, lr=2.46e-05]
Training:  74%|███████▍  | 8245/11082 [26:30:34<9:06:12, 11.55s/step, acc=0.5070, loss=2.0169, lr=2.46e-05]
Training:  74%|███████▍  | 8246/11082 [26:30:46<9:06:24, 11.56s/step, acc=0.5070, loss=2.0169, lr=2.46e-05]
Training:  74%|███████▍  | 8247/11082 [26:30:58<9:16:06, 11.77s/step, acc=0.5070, loss=2.0169, lr=2.46e-05]
Training:  74%|███████▍  | 8248/11082 [26:31:09<9:09:01, 11.62s/step, acc=0.5070, loss=2.0169, lr=2.46e-05]
Training:  74%|███████▍  | 8249/11082 [26:31:21<9:05:12, 11.55s/step, acc=0.5070, loss=2.0169, lr=2.46e-05]
Training:  74%|███████▍  | 8250/11082 [26:31:33<9:09:11, 11.64s/step, acc=0.5070, loss=2.0169, lr=2.46e-05]
Training:  74%|███████▍  | 8250/11082 [26:31:40<9:09:11, 11.64s/step, acc=0.5066, loss=2.0206, lr=2.41e-05]2026-06-19 17:25:07 | INFO | __main__ | epoch=1 step=8250 train_loss=2.020558 train_acc=0.506604 lr=2.414e-05 grad_norm=0.6680 acc_d0=0.7724 acc_d1=0.6341 acc_d2=0.5157 acc_d3=0.4185 acc_d4=0.3393 acc_d5=0.2731 acc_d6=0.2235 acc_d7=0.1828

Training:  74%|███████▍  | 8251/11082 [26:31:44<9:11:11, 11.68s/step, acc=0.5066, loss=2.0206, lr=2.41e-05]
Training:  74%|███████▍  | 8252/11082 [26:31:55<9:03:10, 11.52s/step, acc=0.5066, loss=2.0206, lr=2.41e-05]
Training:  74%|███████▍  | 8253/11082 [26:32:07<9:02:33, 11.51s/step, acc=0.5066, loss=2.0206, lr=2.41e-05]
Training:  74%|███████▍  | 8254/11082 [26:32:19<9:05:49, 11.58s/step, acc=0.5066, loss=2.0206, lr=2.41e-05]
Training:  74%|███████▍  | 8255/11082 [26:32:30<9:07:01, 11.61s/step, acc=0.5066, loss=2.0206, lr=2.41e-05]
Training:  74%|███████▍  | 8256/11082 [26:32:42<9:05:53, 11.59s/step, acc=0.5066, loss=2.0206, lr=2.41e-05]
Training:  75%|███████▍  | 8257/11082 [26:32:54<9:12:57, 11.74s/step, acc=0.5066, loss=2.0206, lr=2.41e-05]
Training:  75%|███████▍  | 8258/11082 [26:33:05<9:05:04, 11.58s/step, acc=0.5066, loss=2.0206, lr=2.41e-05]
Training:  75%|███████▍  | 8259/11082 [26:33:16<8:59:17, 11.46s/step, acc=0.5066, loss=2.0206, lr=2.41e-05]
Training:  75%|███████▍  | 8260/11082 [26:33:28<9:05:49, 11.61s/step, acc=0.5066, loss=2.0206, lr=2.41e-05]
Training:  75%|███████▍  | 8261/11082 [26:33:40<9:06:58, 11.63s/step, acc=0.5066, loss=2.0206, lr=2.41e-05]
Training:  75%|███████▍  | 8262/11082 [26:33:52<9:12:10, 11.75s/step, acc=0.5066, loss=2.0206, lr=2.41e-05]
Training:  75%|███████▍  | 8263/11082 [26:34:03<9:05:55, 11.62s/step, acc=0.5066, loss=2.0206, lr=2.41e-05]
Training:  75%|███████▍  | 8264/11082 [26:34:15<9:08:46, 11.68s/step, acc=0.5066, loss=2.0206, lr=2.41e-05]
Training:  75%|███████▍  | 8265/11082 [26:34:27<9:06:16, 11.64s/step, acc=0.5066, loss=2.0206, lr=2.41e-05]
Training:  75%|███████▍  | 8266/11082 [26:34:38<8:59:24, 11.49s/step, acc=0.5066, loss=2.0206, lr=2.41e-05]
Training:  75%|███████▍  | 8267/11082 [26:34:50<9:06:11, 11.64s/step, acc=0.5066, loss=2.0206, lr=2.41e-05]
Training:  75%|███████▍  | 8268/11082 [26:35:01<9:03:26, 11.59s/step, acc=0.5066, loss=2.0206, lr=2.41e-05]
Training:  75%|███████▍  | 8269/11082 [26:35:13<9:07:24, 11.68s/step, acc=0.5066, loss=2.0206, lr=2.41e-05]
Training:  75%|███████▍  | 8270/11082 [26:35:24<8:58:46, 11.50s/step, acc=0.5066, loss=2.0206, lr=2.41e-05]
Training:  75%|███████▍  | 8271/11082 [26:35:36<8:59:42, 11.52s/step, acc=0.5066, loss=2.0206, lr=2.41e-05]
Training:  75%|███████▍  | 8272/11082 [26:35:48<9:04:08, 11.62s/step, acc=0.5066, loss=2.0206, lr=2.41e-05]
Training:  75%|███████▍  | 8273/11082 [26:35:59<9:01:33, 11.57s/step, acc=0.5066, loss=2.0206, lr=2.41e-05]
Training:  75%|███████▍  | 8274/11082 [26:36:11<9:06:57, 11.69s/step, acc=0.5066, loss=2.0206, lr=2.41e-05]
Training:  75%|███████▍  | 8275/11082 [26:36:23<9:04:01, 11.63s/step, acc=0.5066, loss=2.0206, lr=2.41e-05]
Training:  75%|███████▍  | 8276/11082 [26:36:34<8:56:22, 11.47s/step, acc=0.5066, loss=2.0206, lr=2.41e-05]
Training:  75%|███████▍  | 8277/11082 [26:36:45<8:59:08, 11.53s/step, acc=0.5066, loss=2.0206, lr=2.41e-05]
Training:  75%|███████▍  | 8278/11082 [26:36:57<9:01:24, 11.58s/step, acc=0.5066, loss=2.0206, lr=2.41e-05]
Training:  75%|███████▍  | 8279/11082 [26:37:08<8:57:33, 11.51s/step, acc=0.5066, loss=2.0206, lr=2.41e-05]
Training:  75%|███████▍  | 8280/11082 [26:37:20<9:04:35, 11.66s/step, acc=0.5066, loss=2.0206, lr=2.41e-05]
Training:  75%|███████▍  | 8281/11082 [26:37:32<8:56:29, 11.49s/step, acc=0.5066, loss=2.0206, lr=2.41e-05]
Training:  75%|███████▍  | 8282/11082 [26:37:44<9:07:15, 11.73s/step, acc=0.5066, loss=2.0206, lr=2.41e-05]
Training:  75%|███████▍  | 8283/11082 [26:37:55<8:59:44, 11.57s/step, acc=0.5066, loss=2.0206, lr=2.41e-05]
Training:  75%|███████▍  | 8284/11082 [26:38:07<9:07:54, 11.75s/step, acc=0.5066, loss=2.0206, lr=2.41e-05]
Training:  75%|███████▍  | 8285/11082 [26:38:18<8:56:13, 11.50s/step, acc=0.5066, loss=2.0206, lr=2.41e-05]
Training:  75%|███████▍  | 8286/11082 [26:38:30<9:04:07, 11.68s/step, acc=0.5066, loss=2.0206, lr=2.41e-05]
Training:  75%|███████▍  | 8287/11082 [26:38:42<8:59:43, 11.59s/step, acc=0.5066, loss=2.0206, lr=2.41e-05]
Training:  75%|███████▍  | 8288/11082 [26:38:53<8:56:26, 11.52s/step, acc=0.5066, loss=2.0206, lr=2.41e-05]
Training:  75%|███████▍  | 8289/11082 [26:39:05<9:02:48, 11.66s/step, acc=0.5066, loss=2.0206, lr=2.41e-05]
Training:  75%|███████▍  | 8290/11082 [26:39:17<9:08:28, 11.79s/step, acc=0.5066, loss=2.0206, lr=2.41e-05]
Training:  75%|███████▍  | 8291/11082 [26:39:28<9:03:51, 11.69s/step, acc=0.5066, loss=2.0206, lr=2.41e-05]
Training:  75%|███████▍  | 8292/11082 [26:39:40<9:05:41, 11.74s/step, acc=0.5066, loss=2.0206, lr=2.41e-05]
Training:  75%|███████▍  | 8293/11082 [26:39:52<9:04:14, 11.71s/step, acc=0.5066, loss=2.0206, lr=2.41e-05]
Training:  75%|███████▍  | 8294/11082 [26:40:03<8:52:17, 11.46s/step, acc=0.5066, loss=2.0206, lr=2.41e-05]
Training:  75%|███████▍  | 8295/11082 [26:40:15<8:57:30, 11.57s/step, acc=0.5066, loss=2.0206, lr=2.41e-05]
Training:  75%|███████▍  | 8296/11082 [26:40:27<9:03:29, 11.70s/step, acc=0.5066, loss=2.0206, lr=2.41e-05]
Training:  75%|███████▍  | 8297/11082 [26:40:38<9:03:02, 11.70s/step, acc=0.5066, loss=2.0206, lr=2.41e-05]
Training:  75%|███████▍  | 8298/11082 [26:40:50<8:59:30, 11.63s/step, acc=0.5066, loss=2.0206, lr=2.41e-05]
Training:  75%|███████▍  | 8299/11082 [26:41:01<8:58:30, 11.61s/step, acc=0.5066, loss=2.0206, lr=2.41e-05]
Training:  75%|███████▍  | 8300/11082 [26:41:13<8:54:49, 11.53s/step, acc=0.5066, loss=2.0206, lr=2.41e-05]
Training:  75%|███████▍  | 8300/11082 [26:41:20<8:54:49, 11.53s/step, acc=0.5074, loss=2.0141, lr=2.37e-05]2026-06-19 17:34:48 | INFO | __main__ | epoch=1 step=8300 train_loss=2.014094 train_acc=0.507442 lr=2.367e-05 grad_norm=0.6055 acc_d0=0.7654 acc_d1=0.6255 acc_d2=0.5071 acc_d3=0.4093 acc_d4=0.3318 acc_d5=0.2679 acc_d6=0.2187 acc_d7=0.1793

Training:  75%|███████▍  | 8301/11082 [26:41:25<8:58:51, 11.63s/step, acc=0.5074, loss=2.0141, lr=2.37e-05]
Training:  75%|███████▍  | 8302/11082 [26:41:36<8:49:35, 11.43s/step, acc=0.5074, loss=2.0141, lr=2.37e-05]
Training:  75%|███████▍  | 8303/11082 [26:41:48<8:59:20, 11.64s/step, acc=0.5074, loss=2.0141, lr=2.37e-05]
Training:  75%|███████▍  | 8304/11082 [26:41:59<8:57:29, 11.61s/step, acc=0.5074, loss=2.0141, lr=2.37e-05]
Training:  75%|███████▍  | 8305/11082 [26:42:10<8:50:02, 11.45s/step, acc=0.5074, loss=2.0141, lr=2.37e-05]
Training:  75%|███████▍  | 8306/11082 [26:42:22<8:55:17, 11.57s/step, acc=0.5074, loss=2.0141, lr=2.37e-05]
Training:  75%|███████▍  | 8307/11082 [26:42:34<8:54:24, 11.55s/step, acc=0.5074, loss=2.0141, lr=2.37e-05]
Training:  75%|███████▍  | 8308/11082 [26:42:45<8:54:41, 11.56s/step, acc=0.5074, loss=2.0141, lr=2.37e-05]
Training:  75%|███████▍  | 8309/11082 [26:42:57<8:56:54, 11.62s/step, acc=0.5074, loss=2.0141, lr=2.37e-05]
Training:  75%|███████▍  | 8310/11082 [26:43:08<8:47:35, 11.42s/step, acc=0.5074, loss=2.0141, lr=2.37e-05]
Training:  75%|███████▍  | 8311/11082 [26:43:20<8:54:40, 11.58s/step, acc=0.5074, loss=2.0141, lr=2.37e-05]
Training:  75%|███████▌  | 8312/11082 [26:43:31<8:52:32, 11.54s/step, acc=0.5074, loss=2.0141, lr=2.37e-05]
Training:  75%|███████▌  | 8313/11082 [26:43:43<8:53:34, 11.56s/step, acc=0.5074, loss=2.0141, lr=2.37e-05]
Training:  75%|███████▌  | 8314/11082 [26:43:55<9:04:54, 11.81s/step, acc=0.5074, loss=2.0141, lr=2.37e-05]
Training:  75%|███████▌  | 8315/11082 [26:44:07<8:59:58, 11.71s/step, acc=0.5074, loss=2.0141, lr=2.37e-05]
Training:  75%|███████▌  | 8316/11082 [26:44:19<8:59:32, 11.70s/step, acc=0.5074, loss=2.0141, lr=2.37e-05]
Training:  75%|███████▌  | 8317/11082 [26:44:30<8:55:00, 11.61s/step, acc=0.5074, loss=2.0141, lr=2.37e-05]
Training:  75%|███████▌  | 8318/11082 [26:44:41<8:54:00, 11.59s/step, acc=0.5074, loss=2.0141, lr=2.37e-05]
Training:  75%|███████▌  | 8319/11082 [26:44:53<8:57:52, 11.68s/step, acc=0.5074, loss=2.0141, lr=2.37e-05]
Training:  75%|███████▌  | 8320/11082 [26:45:05<8:57:44, 11.68s/step, acc=0.5074, loss=2.0141, lr=2.37e-05]
Training:  75%|███████▌  | 8321/11082 [26:45:16<8:51:47, 11.56s/step, acc=0.5074, loss=2.0141, lr=2.37e-05]
Training:  75%|███████▌  | 8322/11082 [26:45:28<8:56:36, 11.67s/step, acc=0.5074, loss=2.0141, lr=2.37e-05]
Training:  75%|███████▌  | 8323/11082 [26:45:40<8:53:26, 11.60s/step, acc=0.5074, loss=2.0141, lr=2.37e-05]
Training:  75%|███████▌  | 8324/11082 [26:45:52<9:00:48, 11.77s/step, acc=0.5074, loss=2.0141, lr=2.37e-05]
Training:  75%|███████▌  | 8325/11082 [26:46:03<8:56:40, 11.68s/step, acc=0.5074, loss=2.0141, lr=2.37e-05]
Training:  75%|███████▌  | 8326/11082 [26:46:15<8:58:28, 11.72s/step, acc=0.5074, loss=2.0141, lr=2.37e-05]
Training:  75%|███████▌  | 8327/11082 [26:46:27<8:55:25, 11.66s/step, acc=0.5074, loss=2.0141, lr=2.37e-05]
Training:  75%|███████▌  | 8328/11082 [26:46:38<8:57:04, 11.70s/step, acc=0.5074, loss=2.0141, lr=2.37e-05]
Training:  75%|███████▌  | 8329/11082 [26:46:50<8:53:08, 11.62s/step, acc=0.5074, loss=2.0141, lr=2.37e-05]
Training:  75%|███████▌  | 8330/11082 [26:47:01<8:51:43, 11.59s/step, acc=0.5074, loss=2.0141, lr=2.37e-05]
Training:  75%|███████▌  | 8331/11082 [26:47:13<8:52:19, 11.61s/step, acc=0.5074, loss=2.0141, lr=2.37e-05]
Training:  75%|███████▌  | 8332/11082 [26:47:25<8:51:38, 11.60s/step, acc=0.5074, loss=2.0141, lr=2.37e-05]
Training:  75%|███████▌  | 8333/11082 [26:47:36<8:52:03, 11.61s/step, acc=0.5074, loss=2.0141, lr=2.37e-05]
Training:  75%|███████▌  | 8334/11082 [26:47:48<8:51:21, 11.60s/step, acc=0.5074, loss=2.0141, lr=2.37e-05]
Training:  75%|███████▌  | 8335/11082 [26:47:59<8:50:48, 11.59s/step, acc=0.5074, loss=2.0141, lr=2.37e-05]
Training:  75%|███████▌  | 8336/11082 [26:48:11<8:52:44, 11.64s/step, acc=0.5074, loss=2.0141, lr=2.37e-05]
Training:  75%|███████▌  | 8337/11082 [26:48:22<8:46:32, 11.51s/step, acc=0.5074, loss=2.0141, lr=2.37e-05]
Training:  75%|███████▌  | 8338/11082 [26:48:34<8:51:11, 11.62s/step, acc=0.5074, loss=2.0141, lr=2.37e-05]
Training:  75%|███████▌  | 8339/11082 [26:48:46<8:48:12, 11.55s/step, acc=0.5074, loss=2.0141, lr=2.37e-05]
Training:  75%|███████▌  | 8340/11082 [26:48:57<8:49:14, 11.58s/step, acc=0.5074, loss=2.0141, lr=2.37e-05]
Training:  75%|███████▌  | 8341/11082 [26:49:09<8:47:52, 11.55s/step, acc=0.5074, loss=2.0141, lr=2.37e-05]
Training:  75%|███████▌  | 8342/11082 [26:49:21<8:52:17, 11.66s/step, acc=0.5074, loss=2.0141, lr=2.37e-05]
Training:  75%|███████▌  | 8343/11082 [26:49:32<8:52:01, 11.65s/step, acc=0.5074, loss=2.0141, lr=2.37e-05]
Training:  75%|███████▌  | 8344/11082 [26:49:44<8:50:51, 11.63s/step, acc=0.5074, loss=2.0141, lr=2.37e-05]
Training:  75%|███████▌  | 8345/11082 [26:49:55<8:45:31, 11.52s/step, acc=0.5074, loss=2.0141, lr=2.37e-05]
Training:  75%|███████▌  | 8346/11082 [26:50:07<8:51:02, 11.65s/step, acc=0.5074, loss=2.0141, lr=2.37e-05]
Training:  75%|███████▌  | 8347/11082 [26:50:18<8:47:13, 11.57s/step, acc=0.5074, loss=2.0141, lr=2.37e-05]
Training:  75%|███████▌  | 8348/11082 [26:50:30<8:50:21, 11.64s/step, acc=0.5074, loss=2.0141, lr=2.37e-05]
Training:  75%|███████▌  | 8349/11082 [26:50:42<8:56:31, 11.78s/step, acc=0.5074, loss=2.0141, lr=2.37e-05]
Training:  75%|███████▌  | 8350/11082 [26:50:53<8:44:24, 11.52s/step, acc=0.5074, loss=2.0141, lr=2.37e-05]
Training:  75%|███████▌  | 8350/11082 [26:51:01<8:44:24, 11.52s/step, acc=0.5060, loss=2.0182, lr=2.32e-05]2026-06-19 17:44:28 | INFO | __main__ | epoch=1 step=8350 train_loss=2.018230 train_acc=0.505975 lr=2.321e-05 grad_norm=0.5625 acc_d0=0.7617 acc_d1=0.6208 acc_d2=0.5022 acc_d3=0.4070 acc_d4=0.3291 acc_d5=0.2677 acc_d6=0.2174 acc_d7=0.1784

Training:  75%|███████▌  | 8351/11082 [26:51:05<8:44:50, 11.53s/step, acc=0.5060, loss=2.0182, lr=2.32e-05]
Training:  75%|███████▌  | 8352/11082 [26:51:17<8:47:39, 11.60s/step, acc=0.5060, loss=2.0182, lr=2.32e-05]
Training:  75%|███████▌  | 8353/11082 [26:51:29<8:51:55, 11.70s/step, acc=0.5060, loss=2.0182, lr=2.32e-05]
Training:  75%|███████▌  | 8354/11082 [26:51:40<8:48:07, 11.62s/step, acc=0.5060, loss=2.0182, lr=2.32e-05]
Training:  75%|███████▌  | 8355/11082 [26:51:51<8:42:59, 11.51s/step, acc=0.5060, loss=2.0182, lr=2.32e-05]
Training:  75%|███████▌  | 8356/11082 [26:52:03<8:44:37, 11.55s/step, acc=0.5060, loss=2.0182, lr=2.32e-05]
Training:  75%|███████▌  | 8357/11082 [26:52:15<8:47:55, 11.62s/step, acc=0.5060, loss=2.0182, lr=2.32e-05]
Training:  75%|███████▌  | 8358/11082 [26:52:26<8:47:26, 11.62s/step, acc=0.5060, loss=2.0182, lr=2.32e-05]
Training:  75%|███████▌  | 8359/11082 [26:52:38<8:48:37, 11.65s/step, acc=0.5060, loss=2.0182, lr=2.32e-05]
Training:  75%|███████▌  | 8360/11082 [26:52:49<8:41:11, 11.49s/step, acc=0.5060, loss=2.0182, lr=2.32e-05]
Training:  75%|███████▌  | 8361/11082 [26:53:01<8:46:42, 11.61s/step, acc=0.5060, loss=2.0182, lr=2.32e-05]
Training:  75%|███████▌  | 8362/11082 [26:53:13<8:48:05, 11.65s/step, acc=0.5060, loss=2.0182, lr=2.32e-05]
Training:  75%|███████▌  | 8363/11082 [26:53:24<8:44:24, 11.57s/step, acc=0.5060, loss=2.0182, lr=2.32e-05]
Training:  75%|███████▌  | 8364/11082 [26:53:36<8:47:18, 11.64s/step, acc=0.5060, loss=2.0182, lr=2.32e-05]
Training:  75%|███████▌  | 8365/11082 [26:53:48<8:47:12, 11.64s/step, acc=0.5060, loss=2.0182, lr=2.32e-05]
Training:  75%|███████▌  | 8366/11082 [26:53:59<8:45:55, 11.62s/step, acc=0.5060, loss=2.0182, lr=2.32e-05]
Training:  76%|███████▌  | 8367/11082 [26:54:11<8:42:38, 11.55s/step, acc=0.5060, loss=2.0182, lr=2.32e-05]
Training:  76%|███████▌  | 8368/11082 [26:54:22<8:34:58, 11.38s/step, acc=0.5060, loss=2.0182, lr=2.32e-05]
Training:  76%|███████▌  | 8369/11082 [26:54:33<8:42:08, 11.55s/step, acc=0.5060, loss=2.0182, lr=2.32e-05]
Training:  76%|███████▌  | 8370/11082 [26:54:45<8:39:09, 11.49s/step, acc=0.5060, loss=2.0182, lr=2.32e-05]
Training:  76%|███████▌  | 8371/11082 [26:54:57<8:45:08, 11.62s/step, acc=0.5060, loss=2.0182, lr=2.32e-05]
Training:  76%|███████▌  | 8372/11082 [26:55:09<8:47:07, 11.67s/step, acc=0.5060, loss=2.0182, lr=2.32e-05]
Training:  76%|███████▌  | 8373/11082 [26:55:20<8:43:24, 11.59s/step, acc=0.5060, loss=2.0182, lr=2.32e-05]
Training:  76%|███████▌  | 8374/11082 [26:55:31<8:42:23, 11.57s/step, acc=0.5060, loss=2.0182, lr=2.32e-05]
Training:  76%|███████▌  | 8375/11082 [26:55:43<8:46:45, 11.68s/step, acc=0.5060, loss=2.0182, lr=2.32e-05]
Training:  76%|███████▌  | 8376/11082 [26:55:55<8:45:08, 11.64s/step, acc=0.5060, loss=2.0182, lr=2.32e-05]
Training:  76%|███████▌  | 8377/11082 [26:56:06<8:36:26, 11.46s/step, acc=0.5060, loss=2.0182, lr=2.32e-05]
Training:  76%|███████▌  | 8378/11082 [26:56:18<8:47:21, 11.70s/step, acc=0.5060, loss=2.0182, lr=2.32e-05]
Training:  76%|███████▌  | 8379/11082 [26:56:30<8:43:59, 11.63s/step, acc=0.5060, loss=2.0182, lr=2.32e-05]
Training:  76%|███████▌  | 8380/11082 [26:56:42<8:48:51, 11.74s/step, acc=0.5060, loss=2.0182, lr=2.32e-05]
Training:  76%|███████▌  | 8381/11082 [26:56:53<8:44:47, 11.66s/step, acc=0.5060, loss=2.0182, lr=2.32e-05]
Training:  76%|███████▌  | 8382/11082 [26:57:05<8:45:45, 11.68s/step, acc=0.5060, loss=2.0182, lr=2.32e-05]
Training:  76%|███████▌  | 8383/11082 [26:57:17<8:44:46, 11.67s/step, acc=0.5060, loss=2.0182, lr=2.32e-05]
Training:  76%|███████▌  | 8384/11082 [26:57:28<8:37:10, 11.50s/step, acc=0.5060, loss=2.0182, lr=2.32e-05]
Training:  76%|███████▌  | 8385/11082 [26:57:40<8:41:45, 11.61s/step, acc=0.5060, loss=2.0182, lr=2.32e-05]
Training:  76%|███████▌  | 8386/11082 [26:57:51<8:41:40, 11.61s/step, acc=0.5060, loss=2.0182, lr=2.32e-05]
Training:  76%|███████▌  | 8387/11082 [26:58:03<8:40:17, 11.58s/step, acc=0.5060, loss=2.0182, lr=2.32e-05]
Training:  76%|███████▌  | 8388/11082 [26:58:14<8:38:41, 11.55s/step, acc=0.5060, loss=2.0182, lr=2.32e-05]
Training:  76%|███████▌  | 8389/11082 [26:58:26<8:39:15, 11.57s/step, acc=0.5060, loss=2.0182, lr=2.32e-05]
Training:  76%|███████▌  | 8390/11082 [26:58:37<8:37:09, 11.53s/step, acc=0.5060, loss=2.0182, lr=2.32e-05]
Training:  76%|███████▌  | 8391/11082 [26:58:49<8:37:04, 11.53s/step, acc=0.5060, loss=2.0182, lr=2.32e-05]
Training:  76%|███████▌  | 8392/11082 [26:59:01<8:41:03, 11.62s/step, acc=0.5060, loss=2.0182, lr=2.32e-05]
Training:  76%|███████▌  | 8393/11082 [26:59:12<8:37:49, 11.55s/step, acc=0.5060, loss=2.0182, lr=2.32e-05]
Training:  76%|███████▌  | 8394/11082 [26:59:23<8:33:17, 11.46s/step, acc=0.5060, loss=2.0182, lr=2.32e-05]
Training:  76%|███████▌  | 8395/11082 [26:59:35<8:41:17, 11.64s/step, acc=0.5060, loss=2.0182, lr=2.32e-05]
Training:  76%|███████▌  | 8396/11082 [26:59:47<8:41:13, 11.64s/step, acc=0.5060, loss=2.0182, lr=2.32e-05]
Training:  76%|███████▌  | 8397/11082 [26:59:59<8:44:53, 11.73s/step, acc=0.5060, loss=2.0182, lr=2.32e-05]
Training:  76%|███████▌  | 8398/11082 [27:00:10<8:38:25, 11.59s/step, acc=0.5060, loss=2.0182, lr=2.32e-05]
Training:  76%|███████▌  | 8399/11082 [27:00:22<8:45:49, 11.76s/step, acc=0.5060, loss=2.0182, lr=2.32e-05]
Training:  76%|███████▌  | 8400/11082 [27:00:34<8:43:50, 11.72s/step, acc=0.5060, loss=2.0182, lr=2.32e-05]
Training:  76%|███████▌  | 8400/11082 [27:00:41<8:43:50, 11.72s/step, acc=0.5082, loss=2.0082, lr=2.28e-05]2026-06-19 17:54:08 | INFO | __main__ | epoch=1 step=8400 train_loss=2.008153 train_acc=0.508189 lr=2.275e-05 grad_norm=0.5859 acc_d0=0.7588 acc_d1=0.6153 acc_d2=0.4954 acc_d3=0.3983 acc_d4=0.3215 acc_d5=0.2558 acc_d6=0.2089 acc_d7=0.1693

Training:  76%|███████▌  | 8401/11082 [27:00:45<8:38:03, 11.59s/step, acc=0.5082, loss=2.0082, lr=2.28e-05]
Training:  76%|███████▌  | 8402/11082 [27:00:56<8:32:09, 11.47s/step, acc=0.5082, loss=2.0082, lr=2.28e-05]
Training:  76%|███████▌  | 8403/11082 [27:01:08<8:39:30, 11.64s/step, acc=0.5082, loss=2.0082, lr=2.28e-05]
Training:  76%|███████▌  | 8404/11082 [27:01:20<8:37:07, 11.59s/step, acc=0.5082, loss=2.0082, lr=2.28e-05]
Training:  76%|███████▌  | 8405/11082 [27:01:31<8:36:18, 11.57s/step, acc=0.5082, loss=2.0082, lr=2.28e-05]
Training:  76%|███████▌  | 8406/11082 [27:01:43<8:33:39, 11.52s/step, acc=0.5082, loss=2.0082, lr=2.28e-05]
Training:  76%|███████▌  | 8407/11082 [27:01:54<8:35:42, 11.57s/step, acc=0.5082, loss=2.0082, lr=2.28e-05]
Training:  76%|███████▌  | 8408/11082 [27:02:06<8:37:43, 11.62s/step, acc=0.5082, loss=2.0082, lr=2.28e-05]
Training:  76%|███████▌  | 8409/11082 [27:02:18<8:38:19, 11.63s/step, acc=0.5082, loss=2.0082, lr=2.28e-05]
Training:  76%|███████▌  | 8410/11082 [27:02:30<8:40:16, 11.68s/step, acc=0.5082, loss=2.0082, lr=2.28e-05]
Training:  76%|███████▌  | 8411/11082 [27:02:41<8:41:25, 11.71s/step, acc=0.5082, loss=2.0082, lr=2.28e-05]
Training:  76%|███████▌  | 8412/11082 [27:02:53<8:35:47, 11.59s/step, acc=0.5082, loss=2.0082, lr=2.28e-05]
Training:  76%|███████▌  | 8413/11082 [27:03:04<8:34:25, 11.56s/step, acc=0.5082, loss=2.0082, lr=2.28e-05]
Training:  76%|███████▌  | 8414/11082 [27:03:17<8:43:29, 11.77s/step, acc=0.5082, loss=2.0082, lr=2.28e-05]
Training:  76%|███████▌  | 8415/11082 [27:03:28<8:38:07, 11.66s/step, acc=0.5082, loss=2.0082, lr=2.28e-05]
Training:  76%|███████▌  | 8416/11082 [27:03:39<8:36:05, 11.61s/step, acc=0.5082, loss=2.0082, lr=2.28e-05]
Training:  76%|███████▌  | 8417/11082 [27:03:51<8:37:13, 11.64s/step, acc=0.5082, loss=2.0082, lr=2.28e-05]
Training:  76%|███████▌  | 8418/11082 [27:04:02<8:32:41, 11.55s/step, acc=0.5082, loss=2.0082, lr=2.28e-05]
Training:  76%|███████▌  | 8419/11082 [27:04:14<8:37:05, 11.65s/step, acc=0.5082, loss=2.0082, lr=2.28e-05]
Training:  76%|███████▌  | 8420/11082 [27:04:26<8:33:48, 11.58s/step, acc=0.5082, loss=2.0082, lr=2.28e-05]
Training:  76%|███████▌  | 8421/11082 [27:04:38<8:36:23, 11.64s/step, acc=0.5082, loss=2.0082, lr=2.28e-05]
Training:  76%|███████▌  | 8422/11082 [27:04:49<8:32:27, 11.56s/step, acc=0.5082, loss=2.0082, lr=2.28e-05]
Training:  76%|███████▌  | 8423/11082 [27:05:00<8:31:40, 11.55s/step, acc=0.5082, loss=2.0082, lr=2.28e-05]
Training:  76%|███████▌  | 8424/11082 [27:05:12<8:32:47, 11.58s/step, acc=0.5082, loss=2.0082, lr=2.28e-05]
Training:  76%|███████▌  | 8425/11082 [27:05:24<8:36:25, 11.66s/step, acc=0.5082, loss=2.0082, lr=2.28e-05]
Training:  76%|███████▌  | 8426/11082 [27:05:36<8:35:19, 11.64s/step, acc=0.5082, loss=2.0082, lr=2.28e-05]
Training:  76%|███████▌  | 8427/11082 [27:05:47<8:33:24, 11.60s/step, acc=0.5082, loss=2.0082, lr=2.28e-05]
Training:  76%|███████▌  | 8428/11082 [27:05:58<8:29:20, 11.51s/step, acc=0.5082, loss=2.0082, lr=2.28e-05]
Training:  76%|███████▌  | 8429/11082 [27:06:10<8:32:37, 11.59s/step, acc=0.5082, loss=2.0082, lr=2.28e-05]
Training:  76%|███████▌  | 8430/11082 [27:06:22<8:35:04, 11.65s/step, acc=0.5082, loss=2.0082, lr=2.28e-05]
Training:  76%|███████▌  | 8431/11082 [27:06:33<8:28:02, 11.50s/step, acc=0.5082, loss=2.0082, lr=2.28e-05]
Training:  76%|███████▌  | 8432/11082 [27:06:45<8:31:48, 11.59s/step, acc=0.5082, loss=2.0082, lr=2.28e-05]
Training:  76%|███████▌  | 8433/11082 [27:06:56<8:31:57, 11.60s/step, acc=0.5082, loss=2.0082, lr=2.28e-05]
Training:  76%|███████▌  | 8434/11082 [27:07:08<8:33:30, 11.64s/step, acc=0.5082, loss=2.0082, lr=2.28e-05]
Training:  76%|███████▌  | 8435/11082 [27:07:20<8:33:09, 11.63s/step, acc=0.5082, loss=2.0082, lr=2.28e-05]
Training:  76%|███████▌  | 8436/11082 [27:07:31<8:31:41, 11.60s/step, acc=0.5082, loss=2.0082, lr=2.28e-05]
Training:  76%|███████▌  | 8437/11082 [27:07:43<8:35:44, 11.70s/step, acc=0.5082, loss=2.0082, lr=2.28e-05]
Training:  76%|███████▌  | 8438/11082 [27:07:55<8:36:13, 11.71s/step, acc=0.5082, loss=2.0082, lr=2.28e-05]
Training:  76%|███████▌  | 8439/11082 [27:08:07<8:33:05, 11.65s/step, acc=0.5082, loss=2.0082, lr=2.28e-05]
Training:  76%|███████▌  | 8440/11082 [27:08:18<8:34:31, 11.68s/step, acc=0.5082, loss=2.0082, lr=2.28e-05]
Training:  76%|███████▌  | 8441/11082 [27:08:30<8:30:02, 11.59s/step, acc=0.5082, loss=2.0082, lr=2.28e-05]
Training:  76%|███████▌  | 8442/11082 [27:08:41<8:31:28, 11.62s/step, acc=0.5082, loss=2.0082, lr=2.28e-05]
Training:  76%|███████▌  | 8443/11082 [27:08:53<8:33:32, 11.68s/step, acc=0.5082, loss=2.0082, lr=2.28e-05]
Training:  76%|███████▌  | 8444/11082 [27:09:04<8:28:12, 11.56s/step, acc=0.5082, loss=2.0082, lr=2.28e-05]
Training:  76%|███████▌  | 8445/11082 [27:09:16<8:27:47, 11.55s/step, acc=0.5082, loss=2.0082, lr=2.28e-05]
Training:  76%|███████▌  | 8446/11082 [27:09:28<8:30:40, 11.62s/step, acc=0.5082, loss=2.0082, lr=2.28e-05]
Training:  76%|███████▌  | 8447/11082 [27:09:39<8:19:19, 11.37s/step, acc=0.5082, loss=2.0082, lr=2.28e-05]
Training:  76%|███████▌  | 8448/11082 [27:09:51<8:30:55, 11.64s/step, acc=0.5082, loss=2.0082, lr=2.28e-05]
Training:  76%|███████▌  | 8449/11082 [27:10:02<8:23:17, 11.47s/step, acc=0.5082, loss=2.0082, lr=2.28e-05]
Training:  76%|███████▌  | 8450/11082 [27:10:14<8:31:12, 11.65s/step, acc=0.5082, loss=2.0082, lr=2.28e-05]
Training:  76%|███████▌  | 8450/11082 [27:10:21<8:31:12, 11.65s/step, acc=0.5045, loss=2.0311, lr=2.23e-05]2026-06-19 18:03:49 | INFO | __main__ | epoch=1 step=8450 train_loss=2.031122 train_acc=0.504452 lr=2.230e-05 grad_norm=0.6758 acc_d0=0.7532 acc_d1=0.6057 acc_d2=0.4852 acc_d3=0.3878 acc_d4=0.3115 acc_d5=0.2488 acc_d6=0.2012 acc_d7=0.1659

Training:  76%|███████▋  | 8451/11082 [27:10:25<8:26:02, 11.54s/step, acc=0.5045, loss=2.0311, lr=2.23e-05]
Training:  76%|███████▋  | 8452/11082 [27:10:37<8:25:29, 11.53s/step, acc=0.5045, loss=2.0311, lr=2.23e-05]
Training:  76%|███████▋  | 8453/11082 [27:10:48<8:27:56, 11.59s/step, acc=0.5045, loss=2.0311, lr=2.23e-05]
Training:  76%|███████▋  | 8454/11082 [27:11:01<8:35:34, 11.77s/step, acc=0.5045, loss=2.0311, lr=2.23e-05]
Training:  76%|███████▋  | 8455/11082 [27:11:13<8:41:10, 11.90s/step, acc=0.5045, loss=2.0311, lr=2.23e-05]
Training:  76%|███████▋  | 8456/11082 [27:11:24<8:32:00, 11.70s/step, acc=0.5045, loss=2.0311, lr=2.23e-05]
Training:  76%|███████▋  | 8457/11082 [27:11:36<8:32:29, 11.71s/step, acc=0.5045, loss=2.0311, lr=2.23e-05]
Training:  76%|███████▋  | 8458/11082 [27:11:47<8:24:28, 11.54s/step, acc=0.5045, loss=2.0311, lr=2.23e-05]
Training:  76%|███████▋  | 8459/11082 [27:11:58<8:23:58, 11.53s/step, acc=0.5045, loss=2.0311, lr=2.23e-05]
Training:  76%|███████▋  | 8460/11082 [27:12:10<8:24:20, 11.54s/step, acc=0.5045, loss=2.0311, lr=2.23e-05]
Training:  76%|███████▋  | 8461/11082 [27:12:22<8:28:08, 11.63s/step, acc=0.5045, loss=2.0311, lr=2.23e-05]
Training:  76%|███████▋  | 8462/11082 [27:12:33<8:24:32, 11.55s/step, acc=0.5045, loss=2.0311, lr=2.23e-05]
Training:  76%|███████▋  | 8463/11082 [27:12:45<8:25:02, 11.57s/step, acc=0.5045, loss=2.0311, lr=2.23e-05]
Training:  76%|███████▋  | 8464/11082 [27:12:56<8:24:10, 11.55s/step, acc=0.5045, loss=2.0311, lr=2.23e-05]
Training:  76%|███████▋  | 8465/11082 [27:13:08<8:22:04, 11.51s/step, acc=0.5045, loss=2.0311, lr=2.23e-05]
Training:  76%|███████▋  | 8466/11082 [27:13:20<8:30:26, 11.71s/step, acc=0.5045, loss=2.0311, lr=2.23e-05]
Training:  76%|███████▋  | 8467/11082 [27:13:32<8:27:52, 11.65s/step, acc=0.5045, loss=2.0311, lr=2.23e-05]
Training:  76%|███████▋  | 8468/11082 [27:13:43<8:22:39, 11.54s/step, acc=0.5045, loss=2.0311, lr=2.23e-05]
Training:  76%|███████▋  | 8469/11082 [27:13:54<8:22:50, 11.55s/step, acc=0.5045, loss=2.0311, lr=2.23e-05]
Training:  76%|███████▋  | 8470/11082 [27:14:06<8:24:46, 11.59s/step, acc=0.5045, loss=2.0311, lr=2.23e-05]
Training:  76%|███████▋  | 8471/11082 [27:14:18<8:25:23, 11.61s/step, acc=0.5045, loss=2.0311, lr=2.23e-05]
Training:  76%|███████▋  | 8472/11082 [27:14:29<8:22:23, 11.55s/step, acc=0.5045, loss=2.0311, lr=2.23e-05]
Training:  76%|███████▋  | 8473/11082 [27:14:41<8:24:04, 11.59s/step, acc=0.5045, loss=2.0311, lr=2.23e-05]
Training:  76%|███████▋  | 8474/11082 [27:14:52<8:17:04, 11.44s/step, acc=0.5045, loss=2.0311, lr=2.23e-05]
Training:  76%|███████▋  | 8475/11082 [27:15:04<8:25:16, 11.63s/step, acc=0.5045, loss=2.0311, lr=2.23e-05]
Training:  76%|███████▋  | 8476/11082 [27:15:16<8:26:49, 11.67s/step, acc=0.5045, loss=2.0311, lr=2.23e-05]
Training:  76%|███████▋  | 8477/11082 [27:15:27<8:26:58, 11.68s/step, acc=0.5045, loss=2.0311, lr=2.23e-05]
Training:  77%|███████▋  | 8478/11082 [27:15:39<8:20:54, 11.54s/step, acc=0.5045, loss=2.0311, lr=2.23e-05]
Training:  77%|███████▋  | 8479/11082 [27:15:50<8:23:35, 11.61s/step, acc=0.5045, loss=2.0311, lr=2.23e-05]
Training:  77%|███████▋  | 8480/11082 [27:16:02<8:23:47, 11.62s/step, acc=0.5045, loss=2.0311, lr=2.23e-05]
Training:  77%|███████▋  | 8481/11082 [27:16:14<8:23:48, 11.62s/step, acc=0.5045, loss=2.0311, lr=2.23e-05]
Training:  77%|███████▋  | 8482/11082 [27:16:26<8:27:24, 11.71s/step, acc=0.5045, loss=2.0311, lr=2.23e-05]
Training:  77%|███████▋  | 8483/11082 [27:16:37<8:21:09, 11.57s/step, acc=0.5045, loss=2.0311, lr=2.23e-05]
Training:  77%|███████▋  | 8484/11082 [27:16:49<8:26:27, 11.70s/step, acc=0.5045, loss=2.0311, lr=2.23e-05]
Training:  77%|███████▋  | 8485/11082 [27:17:00<8:25:30, 11.68s/step, acc=0.5045, loss=2.0311, lr=2.23e-05]
Training:  77%|███████▋  | 8486/11082 [27:17:12<8:24:16, 11.65s/step, acc=0.5045, loss=2.0311, lr=2.23e-05]
Training:  77%|███████▋  | 8487/11082 [27:17:23<8:20:07, 11.56s/step, acc=0.5045, loss=2.0311, lr=2.23e-05]
Training:  77%|███████▋  | 8488/11082 [27:17:35<8:20:39, 11.58s/step, acc=0.5045, loss=2.0311, lr=2.23e-05]
Training:  77%|███████▋  | 8489/11082 [27:17:47<8:21:55, 11.61s/step, acc=0.5045, loss=2.0311, lr=2.23e-05]
Training:  77%|███████▋  | 8490/11082 [27:17:58<8:20:26, 11.58s/step, acc=0.5045, loss=2.0311, lr=2.23e-05]
Training:  77%|███████▋  | 8491/11082 [27:18:10<8:24:16, 11.68s/step, acc=0.5045, loss=2.0311, lr=2.23e-05]
Training:  77%|███████▋  | 8492/11082 [27:18:22<8:27:23, 11.75s/step, acc=0.5045, loss=2.0311, lr=2.23e-05]
Training:  77%|███████▋  | 8493/11082 [27:18:33<8:20:31, 11.60s/step, acc=0.5045, loss=2.0311, lr=2.23e-05]
Training:  77%|███████▋  | 8494/11082 [27:18:45<8:23:14, 11.67s/step, acc=0.5045, loss=2.0311, lr=2.23e-05]
Training:  77%|███████▋  | 8495/11082 [27:18:57<8:20:29, 11.61s/step, acc=0.5045, loss=2.0311, lr=2.23e-05]
Training:  77%|███████▋  | 8496/11082 [27:19:09<8:25:20, 11.72s/step, acc=0.5045, loss=2.0311, lr=2.23e-05]
Training:  77%|███████▋  | 8497/11082 [27:19:20<8:20:08, 11.61s/step, acc=0.5045, loss=2.0311, lr=2.23e-05]
Training:  77%|███████▋  | 8498/11082 [27:19:32<8:21:17, 11.64s/step, acc=0.5045, loss=2.0311, lr=2.23e-05]
Training:  77%|███████▋  | 8499/11082 [27:19:43<8:19:01, 11.59s/step, acc=0.5045, loss=2.0311, lr=2.23e-05]
Training:  77%|███████▋  | 8500/11082 [27:19:55<8:19:53, 11.62s/step, acc=0.5045, loss=2.0311, lr=2.23e-05]
Training:  77%|███████▋  | 8500/11082 [27:20:02<8:19:53, 11.62s/step, acc=0.5064, loss=2.0182, lr=2.19e-05]2026-06-19 18:13:30 | INFO | __main__ | epoch=1 step=8500 train_loss=2.018247 train_acc=0.506389 lr=2.186e-05 grad_norm=0.5664 acc_d0=0.7625 acc_d1=0.6194 acc_d2=0.5004 acc_d3=0.4050 acc_d4=0.3257 acc_d5=0.2654 acc_d6=0.2164 acc_d7=0.1790

Training:  77%|███████▋  | 8501/11082 [27:20:07<8:21:01, 11.65s/step, acc=0.5064, loss=2.0182, lr=2.19e-05]
Training:  77%|███████▋  | 8502/11082 [27:20:18<8:17:34, 11.57s/step, acc=0.5064, loss=2.0182, lr=2.19e-05]
Training:  77%|███████▋  | 8503/11082 [27:20:29<8:15:21, 11.52s/step, acc=0.5064, loss=2.0182, lr=2.19e-05]
Training:  77%|███████▋  | 8504/11082 [27:20:41<8:20:05, 11.64s/step, acc=0.5064, loss=2.0182, lr=2.19e-05]
Training:  77%|███████▋  | 8505/11082 [27:20:53<8:15:11, 11.53s/step, acc=0.5064, loss=2.0182, lr=2.19e-05]
Training:  77%|███████▋  | 8506/11082 [27:21:04<8:19:34, 11.64s/step, acc=0.5064, loss=2.0182, lr=2.19e-05]
Training:  77%|███████▋  | 8507/11082 [27:21:16<8:20:42, 11.67s/step, acc=0.5064, loss=2.0182, lr=2.19e-05]
Training:  77%|███████▋  | 8508/11082 [27:21:28<8:18:58, 11.63s/step, acc=0.5064, loss=2.0182, lr=2.19e-05]
Training:  77%|███████▋  | 8509/11082 [27:21:39<8:18:13, 11.62s/step, acc=0.5064, loss=2.0182, lr=2.19e-05]
Training:  77%|███████▋  | 8510/11082 [27:21:51<8:17:48, 11.61s/step, acc=0.5064, loss=2.0182, lr=2.19e-05]
Training:  77%|███████▋  | 8511/11082 [27:22:03<8:18:53, 11.64s/step, acc=0.5064, loss=2.0182, lr=2.19e-05]
Training:  77%|███████▋  | 8512/11082 [27:22:14<8:17:08, 11.61s/step, acc=0.5064, loss=2.0182, lr=2.19e-05]
Training:  77%|███████▋  | 8513/11082 [27:22:26<8:17:52, 11.63s/step, acc=0.5064, loss=2.0182, lr=2.19e-05]
Training:  77%|███████▋  | 8514/11082 [27:22:37<8:14:42, 11.56s/step, acc=0.5064, loss=2.0182, lr=2.19e-05]
Training:  77%|███████▋  | 8515/11082 [27:22:49<8:14:17, 11.55s/step, acc=0.5064, loss=2.0182, lr=2.19e-05]
Training:  77%|███████▋  | 8516/11082 [27:23:00<8:12:26, 11.51s/step, acc=0.5064, loss=2.0182, lr=2.19e-05]
Training:  77%|███████▋  | 8517/11082 [27:23:12<8:21:25, 11.73s/step, acc=0.5064, loss=2.0182, lr=2.19e-05]
Training:  77%|███████▋  | 8518/11082 [27:23:24<8:18:46, 11.67s/step, acc=0.5064, loss=2.0182, lr=2.19e-05]
Training:  77%|███████▋  | 8519/11082 [27:23:35<8:14:02, 11.57s/step, acc=0.5064, loss=2.0182, lr=2.19e-05]
Training:  77%|███████▋  | 8520/11082 [27:23:47<8:16:39, 11.63s/step, acc=0.5064, loss=2.0182, lr=2.19e-05]
Training:  77%|███████▋  | 8521/11082 [27:23:58<8:13:57, 11.57s/step, acc=0.5064, loss=2.0182, lr=2.19e-05]
Training:  77%|███████▋  | 8522/11082 [27:24:10<8:12:28, 11.54s/step, acc=0.5064, loss=2.0182, lr=2.19e-05]
Training:  77%|███████▋  | 8523/11082 [27:24:22<8:13:29, 11.57s/step, acc=0.5064, loss=2.0182, lr=2.19e-05]
Training:  77%|███████▋  | 8524/11082 [27:24:33<8:16:35, 11.65s/step, acc=0.5064, loss=2.0182, lr=2.19e-05]
Training:  77%|███████▋  | 8525/11082 [27:24:45<8:16:22, 11.65s/step, acc=0.5064, loss=2.0182, lr=2.19e-05]
Training:  77%|███████▋  | 8526/11082 [27:24:56<8:08:02, 11.46s/step, acc=0.5064, loss=2.0182, lr=2.19e-05]
Training:  77%|███████▋  | 8527/11082 [27:25:08<8:15:12, 11.63s/step, acc=0.5064, loss=2.0182, lr=2.19e-05]
Training:  77%|███████▋  | 8528/11082 [27:25:19<8:02:22, 11.33s/step, acc=0.5064, loss=2.0182, lr=2.19e-05]
Training:  77%|███████▋  | 8529/11082 [27:25:31<8:16:24, 11.67s/step, acc=0.5064, loss=2.0182, lr=2.19e-05]
Training:  77%|███████▋  | 8530/11082 [27:25:42<8:07:12, 11.45s/step, acc=0.5064, loss=2.0182, lr=2.19e-05]
Training:  77%|███████▋  | 8531/11082 [27:25:54<8:13:52, 11.62s/step, acc=0.5064, loss=2.0182, lr=2.19e-05]
Training:  77%|███████▋  | 8532/11082 [27:26:06<8:21:34, 11.80s/step, acc=0.5064, loss=2.0182, lr=2.19e-05]
Training:  77%|███████▋  | 8533/11082 [27:26:18<8:13:18, 11.61s/step, acc=0.5064, loss=2.0182, lr=2.19e-05]
Training:  77%|███████▋  | 8534/11082 [27:26:29<8:09:04, 11.52s/step, acc=0.5064, loss=2.0182, lr=2.19e-05]
Training:  77%|███████▋  | 8535/11082 [27:26:41<8:12:43, 11.61s/step, acc=0.5064, loss=2.0182, lr=2.19e-05]
Training:  77%|███████▋  | 8536/11082 [27:26:52<8:14:51, 11.66s/step, acc=0.5064, loss=2.0182, lr=2.19e-05]
Training:  77%|███████▋  | 8537/11082 [27:27:04<8:13:35, 11.64s/step, acc=0.5064, loss=2.0182, lr=2.19e-05]
Training:  77%|███████▋  | 8538/11082 [27:27:16<8:12:00, 11.60s/step, acc=0.5064, loss=2.0182, lr=2.19e-05]
Training:  77%|███████▋  | 8539/11082 [27:27:27<8:11:31, 11.60s/step, acc=0.5064, loss=2.0182, lr=2.19e-05]
Training:  77%|███████▋  | 8540/11082 [27:27:39<8:11:57, 11.61s/step, acc=0.5064, loss=2.0182, lr=2.19e-05]
Training:  77%|███████▋  | 8541/11082 [27:27:50<8:13:08, 11.64s/step, acc=0.5064, loss=2.0182, lr=2.19e-05]
Training:  77%|███████▋  | 8542/11082 [27:28:02<8:14:06, 11.67s/step, acc=0.5064, loss=2.0182, lr=2.19e-05]
Training:  77%|███████▋  | 8543/11082 [27:28:13<8:08:43, 11.55s/step, acc=0.5064, loss=2.0182, lr=2.19e-05]
Training:  77%|███████▋  | 8544/11082 [27:28:25<8:12:45, 11.65s/step, acc=0.5064, loss=2.0182, lr=2.19e-05]
Training:  77%|███████▋  | 8545/11082 [27:28:37<8:14:25, 11.69s/step, acc=0.5064, loss=2.0182, lr=2.19e-05]
Training:  77%|███████▋  | 8546/11082 [27:28:49<8:11:38, 11.63s/step, acc=0.5064, loss=2.0182, lr=2.19e-05]
Training:  77%|███████▋  | 8547/11082 [27:29:01<8:17:03, 11.76s/step, acc=0.5064, loss=2.0182, lr=2.19e-05]
Training:  77%|███████▋  | 8548/11082 [27:29:12<8:05:34, 11.50s/step, acc=0.5064, loss=2.0182, lr=2.19e-05]
Training:  77%|███████▋  | 8549/11082 [27:29:24<8:15:54, 11.75s/step, acc=0.5064, loss=2.0182, lr=2.19e-05]
Training:  77%|███████▋  | 8550/11082 [27:29:35<8:09:07, 11.59s/step, acc=0.5064, loss=2.0182, lr=2.19e-05]
Training:  77%|███████▋  | 8550/11082 [27:29:42<8:09:07, 11.59s/step, acc=0.5064, loss=2.0238, lr=2.14e-05]2026-06-19 18:23:10 | INFO | __main__ | epoch=1 step=8550 train_loss=2.023831 train_acc=0.506431 lr=2.143e-05 grad_norm=0.5742 acc_d0=0.7600 acc_d1=0.6149 acc_d2=0.4932 acc_d3=0.3963 acc_d4=0.3175 acc_d5=0.2554 acc_d6=0.2071 acc_d7=0.1708

Training:  77%|███████▋  | 8551/11082 [27:29:47<8:10:14, 11.62s/step, acc=0.5064, loss=2.0238, lr=2.14e-05]
Training:  77%|███████▋  | 8552/11082 [27:29:58<8:09:54, 11.62s/step, acc=0.5064, loss=2.0238, lr=2.14e-05]
Training:  77%|███████▋  | 8553/11082 [27:30:10<8:08:19, 11.59s/step, acc=0.5064, loss=2.0238, lr=2.14e-05]
Training:  77%|███████▋  | 8554/11082 [27:30:21<8:04:28, 11.50s/step, acc=0.5064, loss=2.0238, lr=2.14e-05]
Training:  77%|███████▋  | 8555/11082 [27:30:34<8:13:48, 11.72s/step, acc=0.5064, loss=2.0238, lr=2.14e-05]
Training:  77%|███████▋  | 8556/11082 [27:30:45<8:07:23, 11.58s/step, acc=0.5064, loss=2.0238, lr=2.14e-05]
Training:  77%|███████▋  | 8557/11082 [27:30:57<8:13:50, 11.74s/step, acc=0.5064, loss=2.0238, lr=2.14e-05]
Training:  77%|███████▋  | 8558/11082 [27:31:08<8:07:30, 11.59s/step, acc=0.5064, loss=2.0238, lr=2.14e-05]
Training:  77%|███████▋  | 8559/11082 [27:31:20<8:14:11, 11.75s/step, acc=0.5064, loss=2.0238, lr=2.14e-05]
Training:  77%|███████▋  | 8560/11082 [27:31:31<8:06:43, 11.58s/step, acc=0.5064, loss=2.0238, lr=2.14e-05]
Training:  77%|███████▋  | 8561/11082 [27:31:43<8:05:54, 11.56s/step, acc=0.5064, loss=2.0238, lr=2.14e-05]
Training:  77%|███████▋  | 8562/11082 [27:31:55<8:08:23, 11.63s/step, acc=0.5064, loss=2.0238, lr=2.14e-05]
Training:  77%|███████▋  | 8563/11082 [27:32:06<8:08:57, 11.65s/step, acc=0.5064, loss=2.0238, lr=2.14e-05]
Training:  77%|███████▋  | 8564/11082 [27:32:18<8:08:12, 11.63s/step, acc=0.5064, loss=2.0238, lr=2.14e-05]
Training:  77%|███████▋  | 8565/11082 [27:32:29<8:05:06, 11.56s/step, acc=0.5064, loss=2.0238, lr=2.14e-05]
Training:  77%|███████▋  | 8566/11082 [27:32:41<8:02:17, 11.50s/step, acc=0.5064, loss=2.0238, lr=2.14e-05]
Training:  77%|███████▋  | 8567/11082 [27:32:53<8:08:03, 11.64s/step, acc=0.5064, loss=2.0238, lr=2.14e-05]
Training:  77%|███████▋  | 8568/11082 [27:33:04<8:08:58, 11.67s/step, acc=0.5064, loss=2.0238, lr=2.14e-05]
Training:  77%|███████▋  | 8569/11082 [27:33:16<8:06:45, 11.62s/step, acc=0.5064, loss=2.0238, lr=2.14e-05]
Training:  77%|███████▋  | 8570/11082 [27:33:27<8:05:15, 11.59s/step, acc=0.5064, loss=2.0238, lr=2.14e-05]
Training:  77%|███████▋  | 8571/11082 [27:33:39<8:06:34, 11.63s/step, acc=0.5064, loss=2.0238, lr=2.14e-05]
Training:  77%|███████▋  | 8572/11082 [27:33:51<8:11:14, 11.74s/step, acc=0.5064, loss=2.0238, lr=2.14e-05]
Training:  77%|███████▋  | 8573/11082 [27:34:03<8:05:37, 11.61s/step, acc=0.5064, loss=2.0238, lr=2.14e-05]
Training:  77%|███████▋  | 8574/11082 [27:34:14<8:06:20, 11.63s/step, acc=0.5064, loss=2.0238, lr=2.14e-05]
Training:  77%|███████▋  | 8575/11082 [27:34:26<8:02:28, 11.55s/step, acc=0.5064, loss=2.0238, lr=2.14e-05]
Training:  77%|███████▋  | 8576/11082 [27:34:37<8:03:01, 11.57s/step, acc=0.5064, loss=2.0238, lr=2.14e-05]
Training:  77%|███████▋  | 8577/11082 [27:34:49<8:06:10, 11.64s/step, acc=0.5064, loss=2.0238, lr=2.14e-05]
Training:  77%|███████▋  | 8578/11082 [27:35:01<8:06:11, 11.65s/step, acc=0.5064, loss=2.0238, lr=2.14e-05]
Training:  77%|███████▋  | 8579/11082 [27:35:12<8:06:00, 11.65s/step, acc=0.5064, loss=2.0238, lr=2.14e-05]
Training:  77%|███████▋  | 8580/11082 [27:35:23<8:00:01, 11.51s/step, acc=0.5064, loss=2.0238, lr=2.14e-05]
Training:  77%|███████▋  | 8581/11082 [27:35:36<8:07:51, 11.70s/step, acc=0.5064, loss=2.0238, lr=2.14e-05]
Training:  77%|███████▋  | 8582/11082 [27:35:47<8:08:31, 11.72s/step, acc=0.5064, loss=2.0238, lr=2.14e-05]
Training:  77%|███████▋  | 8583/11082 [27:35:59<8:01:20, 11.56s/step, acc=0.5064, loss=2.0238, lr=2.14e-05]
Training:  77%|███████▋  | 8584/11082 [27:36:10<7:59:31, 11.52s/step, acc=0.5064, loss=2.0238, lr=2.14e-05]
Training:  77%|███████▋  | 8585/11082 [27:36:21<7:55:45, 11.43s/step, acc=0.5064, loss=2.0238, lr=2.14e-05]
Training:  77%|███████▋  | 8586/11082 [27:36:34<8:06:23, 11.69s/step, acc=0.5064, loss=2.0238, lr=2.14e-05]
Training:  77%|███████▋  | 8587/11082 [27:36:45<8:02:03, 11.59s/step, acc=0.5064, loss=2.0238, lr=2.14e-05]
Training:  77%|███████▋  | 8588/11082 [27:36:56<8:01:10, 11.58s/step, acc=0.5064, loss=2.0238, lr=2.14e-05]
Training:  78%|███████▊  | 8589/11082 [27:37:08<7:56:50, 11.48s/step, acc=0.5064, loss=2.0238, lr=2.14e-05]
Training:  78%|███████▊  | 8590/11082 [27:37:19<7:55:53, 11.46s/step, acc=0.5064, loss=2.0238, lr=2.14e-05]
Training:  78%|███████▊  | 8591/11082 [27:37:31<8:01:18, 11.59s/step, acc=0.5064, loss=2.0238, lr=2.14e-05]
Training:  78%|███████▊  | 8592/11082 [27:37:43<8:03:32, 11.65s/step, acc=0.5064, loss=2.0238, lr=2.14e-05]
Training:  78%|███████▊  | 8593/11082 [27:37:54<7:59:05, 11.55s/step, acc=0.5064, loss=2.0238, lr=2.14e-05]
Training:  78%|███████▊  | 8594/11082 [27:38:05<7:55:55, 11.48s/step, acc=0.5064, loss=2.0238, lr=2.14e-05]
Training:  78%|███████▊  | 8595/11082 [27:38:17<8:01:08, 11.61s/step, acc=0.5064, loss=2.0238, lr=2.14e-05]
Training:  78%|███████▊  | 8596/11082 [27:38:29<7:57:55, 11.53s/step, acc=0.5064, loss=2.0238, lr=2.14e-05]
Training:  78%|███████▊  | 8597/11082 [27:38:40<7:59:27, 11.58s/step, acc=0.5064, loss=2.0238, lr=2.14e-05]
Training:  78%|███████▊  | 8598/11082 [27:38:52<8:00:41, 11.61s/step, acc=0.5064, loss=2.0238, lr=2.14e-05]
Training:  78%|███████▊  | 8599/11082 [27:39:04<7:59:24, 11.58s/step, acc=0.5064, loss=2.0238, lr=2.14e-05]
Training:  78%|███████▊  | 8600/11082 [27:39:15<8:00:14, 11.61s/step, acc=0.5064, loss=2.0238, lr=2.14e-05]
Training:  78%|███████▊  | 8600/11082 [27:39:22<8:00:14, 11.61s/step, acc=0.5070, loss=2.0175, lr=2.10e-05]2026-06-19 18:32:50 | INFO | __main__ | epoch=1 step=8600 train_loss=2.017467 train_acc=0.506979 lr=2.100e-05 grad_norm=0.6016 acc_d0=0.7595 acc_d1=0.6163 acc_d2=0.4957 acc_d3=0.3993 acc_d4=0.3220 acc_d5=0.2593 acc_d6=0.2117 acc_d7=0.1739

Training:  78%|███████▊  | 8601/11082 [27:39:27<8:02:59, 11.68s/step, acc=0.5070, loss=2.0175, lr=2.10e-05]
Training:  78%|███████▊  | 8602/11082 [27:39:38<7:55:42, 11.51s/step, acc=0.5070, loss=2.0175, lr=2.10e-05]
Training:  78%|███████▊  | 8603/11082 [27:39:50<7:55:37, 11.51s/step, acc=0.5070, loss=2.0175, lr=2.10e-05]
Training:  78%|███████▊  | 8604/11082 [27:40:01<7:53:24, 11.46s/step, acc=0.5070, loss=2.0175, lr=2.10e-05]
Training:  78%|███████▊  | 8605/11082 [27:40:13<8:04:33, 11.74s/step, acc=0.5070, loss=2.0175, lr=2.10e-05]
Training:  78%|███████▊  | 8606/11082 [27:40:25<7:56:09, 11.54s/step, acc=0.5070, loss=2.0175, lr=2.10e-05]
Training:  78%|███████▊  | 8607/11082 [27:40:36<7:59:02, 11.61s/step, acc=0.5070, loss=2.0175, lr=2.10e-05]
Training:  78%|███████▊  | 8608/11082 [27:40:47<7:52:31, 11.46s/step, acc=0.5070, loss=2.0175, lr=2.10e-05]
Training:  78%|███████▊  | 8609/11082 [27:40:59<7:55:27, 11.54s/step, acc=0.5070, loss=2.0175, lr=2.10e-05]
Training:  78%|███████▊  | 8610/11082 [27:41:11<7:53:51, 11.50s/step, acc=0.5070, loss=2.0175, lr=2.10e-05]
Training:  78%|███████▊  | 8611/11082 [27:41:22<7:57:15, 11.59s/step, acc=0.5070, loss=2.0175, lr=2.10e-05]
Training:  78%|███████▊  | 8612/11082 [27:41:34<7:57:51, 11.61s/step, acc=0.5070, loss=2.0175, lr=2.10e-05]
Training:  78%|███████▊  | 8613/11082 [27:41:45<7:53:19, 11.50s/step, acc=0.5070, loss=2.0175, lr=2.10e-05]
Training:  78%|███████▊  | 8614/11082 [27:41:57<7:54:57, 11.55s/step, acc=0.5070, loss=2.0175, lr=2.10e-05]
Training:  78%|███████▊  | 8615/11082 [27:42:09<7:58:27, 11.64s/step, acc=0.5070, loss=2.0175, lr=2.10e-05]
Training:  78%|███████▊  | 8616/11082 [27:42:20<7:53:42, 11.53s/step, acc=0.5070, loss=2.0175, lr=2.10e-05]
Training:  78%|███████▊  | 8617/11082 [27:42:32<7:57:41, 11.63s/step, acc=0.5070, loss=2.0175, lr=2.10e-05]
Training:  78%|███████▊  | 8618/11082 [27:42:44<7:57:33, 11.63s/step, acc=0.5070, loss=2.0175, lr=2.10e-05]
Training:  78%|███████▊  | 8619/11082 [27:42:55<7:53:04, 11.52s/step, acc=0.5070, loss=2.0175, lr=2.10e-05]
Training:  78%|███████▊  | 8620/11082 [27:43:07<7:57:44, 11.64s/step, acc=0.5070, loss=2.0175, lr=2.10e-05]
Training:  78%|███████▊  | 8621/11082 [27:43:18<7:58:36, 11.67s/step, acc=0.5070, loss=2.0175, lr=2.10e-05]
Training:  78%|███████▊  | 8622/11082 [27:43:29<7:48:19, 11.42s/step, acc=0.5070, loss=2.0175, lr=2.10e-05]
Training:  78%|███████▊  | 8623/11082 [27:43:41<7:53:19, 11.55s/step, acc=0.5070, loss=2.0175, lr=2.10e-05]
Training:  78%|███████▊  | 8624/11082 [27:43:54<8:03:39, 11.81s/step, acc=0.5070, loss=2.0175, lr=2.10e-05]
Training:  78%|███████▊  | 8625/11082 [27:44:05<7:56:27, 11.63s/step, acc=0.5070, loss=2.0175, lr=2.10e-05]
Training:  78%|███████▊  | 8626/11082 [27:44:16<7:55:29, 11.62s/step, acc=0.5070, loss=2.0175, lr=2.10e-05]
Training:  78%|███████▊  | 8627/11082 [27:44:28<7:54:32, 11.60s/step, acc=0.5070, loss=2.0175, lr=2.10e-05]
Training:  78%|███████▊  | 8628/11082 [27:44:39<7:54:21, 11.60s/step, acc=0.5070, loss=2.0175, lr=2.10e-05]
Training:  78%|███████▊  | 8629/11082 [27:44:51<7:55:47, 11.64s/step, acc=0.5070, loss=2.0175, lr=2.10e-05]
Training:  78%|███████▊  | 8630/11082 [27:45:03<7:56:53, 11.67s/step, acc=0.5070, loss=2.0175, lr=2.10e-05]
Training:  78%|███████▊  | 8631/11082 [27:45:14<7:54:41, 11.62s/step, acc=0.5070, loss=2.0175, lr=2.10e-05]
Training:  78%|███████▊  | 8632/11082 [27:45:26<7:53:37, 11.60s/step, acc=0.5070, loss=2.0175, lr=2.10e-05]
Training:  78%|███████▊  | 8633/11082 [27:45:37<7:49:38, 11.51s/step, acc=0.5070, loss=2.0175, lr=2.10e-05]
Training:  78%|███████▊  | 8634/11082 [27:45:49<7:49:22, 11.50s/step, acc=0.5070, loss=2.0175, lr=2.10e-05]
Training:  78%|███████▊  | 8635/11082 [27:46:00<7:50:56, 11.55s/step, acc=0.5070, loss=2.0175, lr=2.10e-05]
Training:  78%|███████▊  | 8636/11082 [27:46:12<7:52:53, 11.60s/step, acc=0.5070, loss=2.0175, lr=2.10e-05]
Training:  78%|███████▊  | 8637/11082 [27:46:24<7:55:04, 11.66s/step, acc=0.5070, loss=2.0175, lr=2.10e-05]
Training:  78%|███████▊  | 8638/11082 [27:46:35<7:52:06, 11.59s/step, acc=0.5070, loss=2.0175, lr=2.10e-05]
Training:  78%|███████▊  | 8639/11082 [27:46:47<7:56:45, 11.71s/step, acc=0.5070, loss=2.0175, lr=2.10e-05]
Training:  78%|███████▊  | 8640/11082 [27:46:59<7:51:37, 11.59s/step, acc=0.5070, loss=2.0175, lr=2.10e-05]
Training:  78%|███████▊  | 8641/11082 [27:47:11<7:55:28, 11.69s/step, acc=0.5070, loss=2.0175, lr=2.10e-05]
Training:  78%|███████▊  | 8642/11082 [27:47:22<7:48:35, 11.52s/step, acc=0.5070, loss=2.0175, lr=2.10e-05]
Training:  78%|███████▊  | 8643/11082 [27:47:34<7:56:50, 11.73s/step, acc=0.5070, loss=2.0175, lr=2.10e-05]
Training:  78%|███████▊  | 8644/11082 [27:47:46<7:57:35, 11.75s/step, acc=0.5070, loss=2.0175, lr=2.10e-05]
Training:  78%|███████▊  | 8645/11082 [27:47:57<7:50:10, 11.58s/step, acc=0.5070, loss=2.0175, lr=2.10e-05]
Training:  78%|███████▊  | 8646/11082 [27:48:09<7:55:52, 11.72s/step, acc=0.5070, loss=2.0175, lr=2.10e-05]
Training:  78%|███████▊  | 8647/11082 [27:48:20<7:48:56, 11.56s/step, acc=0.5070, loss=2.0175, lr=2.10e-05]
Training:  78%|███████▊  | 8648/11082 [27:48:32<7:49:47, 11.58s/step, acc=0.5070, loss=2.0175, lr=2.10e-05]
Training:  78%|███████▊  | 8649/11082 [27:48:44<7:51:01, 11.62s/step, acc=0.5070, loss=2.0175, lr=2.10e-05]
Training:  78%|███████▊  | 8650/11082 [27:48:54<7:42:12, 11.40s/step, acc=0.5070, loss=2.0175, lr=2.10e-05]
Training:  78%|███████▊  | 8650/11082 [27:49:02<7:42:12, 11.40s/step, acc=0.5057, loss=2.0250, lr=2.06e-05]2026-06-19 18:42:30 | INFO | __main__ | epoch=1 step=8650 train_loss=2.024967 train_acc=0.505708 lr=2.058e-05 grad_norm=0.6094 acc_d0=0.7629 acc_d1=0.6192 acc_d2=0.5009 acc_d3=0.4037 acc_d4=0.3254 acc_d5=0.2646 acc_d6=0.2156 acc_d7=0.1770

Training:  78%|███████▊  | 8651/11082 [27:49:07<7:53:20, 11.68s/step, acc=0.5057, loss=2.0250, lr=2.06e-05]
Training:  78%|███████▊  | 8652/11082 [27:49:18<7:48:17, 11.56s/step, acc=0.5057, loss=2.0250, lr=2.06e-05]
Training:  78%|███████▊  | 8653/11082 [27:49:30<7:54:13, 11.71s/step, acc=0.5057, loss=2.0250, lr=2.06e-05]
Training:  78%|███████▊  | 8654/11082 [27:49:42<7:52:41, 11.68s/step, acc=0.5057, loss=2.0250, lr=2.06e-05]
Training:  78%|███████▊  | 8655/11082 [27:49:54<7:56:27, 11.78s/step, acc=0.5057, loss=2.0250, lr=2.06e-05]
Training:  78%|███████▊  | 8656/11082 [27:50:05<7:52:05, 11.68s/step, acc=0.5057, loss=2.0250, lr=2.06e-05]
Training:  78%|███████▊  | 8657/11082 [27:50:16<7:46:04, 11.53s/step, acc=0.5057, loss=2.0250, lr=2.06e-05]
Training:  78%|███████▊  | 8658/11082 [27:50:28<7:51:23, 11.67s/step, acc=0.5057, loss=2.0250, lr=2.06e-05]
Training:  78%|███████▊  | 8659/11082 [27:50:40<7:55:25, 11.77s/step, acc=0.5057, loss=2.0250, lr=2.06e-05]
Training:  78%|███████▊  | 8660/11082 [27:50:51<7:45:41, 11.54s/step, acc=0.5057, loss=2.0250, lr=2.06e-05]
Training:  78%|███████▊  | 8661/11082 [27:51:03<7:50:50, 11.67s/step, acc=0.5057, loss=2.0250, lr=2.06e-05]
Training:  78%|███████▊  | 8662/11082 [27:51:15<7:50:03, 11.65s/step, acc=0.5057, loss=2.0250, lr=2.06e-05]
Training:  78%|███████▊  | 8663/11082 [27:51:26<7:47:12, 11.59s/step, acc=0.5057, loss=2.0250, lr=2.06e-05]
Training:  78%|███████▊  | 8664/11082 [27:51:38<7:43:09, 11.49s/step, acc=0.5057, loss=2.0250, lr=2.06e-05]
Training:  78%|███████▊  | 8665/11082 [27:51:49<7:46:38, 11.58s/step, acc=0.5057, loss=2.0250, lr=2.06e-05]
Training:  78%|███████▊  | 8666/11082 [27:52:01<7:47:15, 11.60s/step, acc=0.5057, loss=2.0250, lr=2.06e-05]
Training:  78%|███████▊  | 8667/11082 [27:52:13<7:46:58, 11.60s/step, acc=0.5057, loss=2.0250, lr=2.06e-05]
Training:  78%|███████▊  | 8668/11082 [27:52:24<7:44:51, 11.55s/step, acc=0.5057, loss=2.0250, lr=2.06e-05]
Training:  78%|███████▊  | 8669/11082 [27:52:36<7:45:22, 11.57s/step, acc=0.5057, loss=2.0250, lr=2.06e-05]
Training:  78%|███████▊  | 8670/11082 [27:52:48<7:55:29, 11.83s/step, acc=0.5057, loss=2.0250, lr=2.06e-05]
Training:  78%|███████▊  | 8671/11082 [27:52:59<7:41:41, 11.49s/step, acc=0.5057, loss=2.0250, lr=2.06e-05]
Training:  78%|███████▊  | 8672/11082 [27:53:11<7:49:56, 11.70s/step, acc=0.5057, loss=2.0250, lr=2.06e-05]
Training:  78%|███████▊  | 8673/11082 [27:53:22<7:41:07, 11.48s/step, acc=0.5057, loss=2.0250, lr=2.06e-05]
Training:  78%|███████▊  | 8674/11082 [27:53:34<7:45:36, 11.60s/step, acc=0.5057, loss=2.0250, lr=2.06e-05]
Training:  78%|███████▊  | 8675/11082 [27:53:46<7:46:01, 11.62s/step, acc=0.5057, loss=2.0250, lr=2.06e-05]
Training:  78%|███████▊  | 8676/11082 [27:53:57<7:44:53, 11.59s/step, acc=0.5057, loss=2.0250, lr=2.06e-05]
Training:  78%|███████▊  | 8677/11082 [27:54:09<7:45:20, 11.61s/step, acc=0.5057, loss=2.0250, lr=2.06e-05]
Training:  78%|███████▊  | 8678/11082 [27:54:21<7:47:45, 11.67s/step, acc=0.5057, loss=2.0250, lr=2.06e-05]
Training:  78%|███████▊  | 8679/11082 [27:54:32<7:43:19, 11.57s/step, acc=0.5057, loss=2.0250, lr=2.06e-05]
Training:  78%|███████▊  | 8680/11082 [27:54:44<7:45:13, 11.62s/step, acc=0.5057, loss=2.0250, lr=2.06e-05]
Training:  78%|███████▊  | 8681/11082 [27:54:55<7:44:47, 11.61s/step, acc=0.5057, loss=2.0250, lr=2.06e-05]
Training:  78%|███████▊  | 8682/11082 [27:55:07<7:41:03, 11.53s/step, acc=0.5057, loss=2.0250, lr=2.06e-05]
Training:  78%|███████▊  | 8683/11082 [27:55:18<7:45:41, 11.65s/step, acc=0.5057, loss=2.0250, lr=2.06e-05]
Training:  78%|███████▊  | 8684/11082 [27:55:30<7:43:22, 11.59s/step, acc=0.5057, loss=2.0250, lr=2.06e-05]
Training:  78%|███████▊  | 8685/11082 [27:55:41<7:41:04, 11.54s/step, acc=0.5057, loss=2.0250, lr=2.06e-05]
Training:  78%|███████▊  | 8686/11082 [27:55:53<7:40:45, 11.54s/step, acc=0.5057, loss=2.0250, lr=2.06e-05]
Training:  78%|███████▊  | 8687/11082 [27:56:04<7:40:21, 11.53s/step, acc=0.5057, loss=2.0250, lr=2.06e-05]
Training:  78%|███████▊  | 8688/11082 [27:56:16<7:40:00, 11.53s/step, acc=0.5057, loss=2.0250, lr=2.06e-05]
Training:  78%|███████▊  | 8689/11082 [27:56:28<7:45:06, 11.66s/step, acc=0.5057, loss=2.0250, lr=2.06e-05]
Training:  78%|███████▊  | 8690/11082 [27:56:39<7:42:50, 11.61s/step, acc=0.5057, loss=2.0250, lr=2.06e-05]
Training:  78%|███████▊  | 8691/11082 [27:56:51<7:42:29, 11.61s/step, acc=0.5057, loss=2.0250, lr=2.06e-05]
Training:  78%|███████▊  | 8692/11082 [27:57:03<7:45:39, 11.69s/step, acc=0.5057, loss=2.0250, lr=2.06e-05]
Training:  78%|███████▊  | 8693/11082 [27:57:14<7:40:52, 11.58s/step, acc=0.5057, loss=2.0250, lr=2.06e-05]
Training:  78%|███████▊  | 8694/11082 [27:57:26<7:42:17, 11.62s/step, acc=0.5057, loss=2.0250, lr=2.06e-05]
Training:  78%|███████▊  | 8695/11082 [27:57:38<7:44:11, 11.67s/step, acc=0.5057, loss=2.0250, lr=2.06e-05]
Training:  78%|███████▊  | 8696/11082 [27:57:49<7:43:27, 11.65s/step, acc=0.5057, loss=2.0250, lr=2.06e-05]
Training:  78%|███████▊  | 8697/11082 [27:58:01<7:42:38, 11.64s/step, acc=0.5057, loss=2.0250, lr=2.06e-05]
Training:  78%|███████▊  | 8698/11082 [27:58:13<7:41:50, 11.62s/step, acc=0.5057, loss=2.0250, lr=2.06e-05]
Training:  78%|███████▊  | 8699/11082 [27:58:24<7:38:49, 11.55s/step, acc=0.5057, loss=2.0250, lr=2.06e-05]
Training:  79%|███████▊  | 8700/11082 [27:58:36<7:43:03, 11.66s/step, acc=0.5057, loss=2.0250, lr=2.06e-05]
Training:  79%|███████▊  | 8700/11082 [27:58:43<7:43:03, 11.66s/step, acc=0.5085, loss=2.0099, lr=2.02e-05]2026-06-19 18:52:11 | INFO | __main__ | epoch=1 step=8700 train_loss=2.009910 train_acc=0.508516 lr=2.017e-05 grad_norm=0.5547 acc_d0=0.7675 acc_d1=0.6255 acc_d2=0.5078 acc_d3=0.4092 acc_d4=0.3319 acc_d5=0.2680 acc_d6=0.2176 acc_d7=0.1791

Training:  79%|███████▊  | 8701/11082 [27:58:47<7:40:40, 11.61s/step, acc=0.5085, loss=2.0099, lr=2.02e-05]
Training:  79%|███████▊  | 8702/11082 [27:58:59<7:36:37, 11.51s/step, acc=0.5085, loss=2.0099, lr=2.02e-05]
Training:  79%|███████▊  | 8703/11082 [27:59:10<7:40:38, 11.62s/step, acc=0.5085, loss=2.0099, lr=2.02e-05]
Training:  79%|███████▊  | 8704/11082 [27:59:22<7:35:53, 11.50s/step, acc=0.5085, loss=2.0099, lr=2.02e-05]
Training:  79%|███████▊  | 8705/11082 [27:59:33<7:36:57, 11.53s/step, acc=0.5085, loss=2.0099, lr=2.02e-05]
Training:  79%|███████▊  | 8706/11082 [27:59:45<7:41:49, 11.66s/step, acc=0.5085, loss=2.0099, lr=2.02e-05]
Training:  79%|███████▊  | 8707/11082 [27:59:56<7:35:54, 11.52s/step, acc=0.5085, loss=2.0099, lr=2.02e-05]
Training:  79%|███████▊  | 8708/11082 [28:00:08<7:37:19, 11.56s/step, acc=0.5085, loss=2.0099, lr=2.02e-05]
Training:  79%|███████▊  | 8709/11082 [28:00:20<7:41:29, 11.67s/step, acc=0.5085, loss=2.0099, lr=2.02e-05]
Training:  79%|███████▊  | 8710/11082 [28:00:31<7:34:21, 11.49s/step, acc=0.5085, loss=2.0099, lr=2.02e-05]
Training:  79%|███████▊  | 8711/11082 [28:00:43<7:39:36, 11.63s/step, acc=0.5085, loss=2.0099, lr=2.02e-05]
Training:  79%|███████▊  | 8712/11082 [28:00:55<7:41:34, 11.69s/step, acc=0.5085, loss=2.0099, lr=2.02e-05]
Training:  79%|███████▊  | 8713/11082 [28:01:06<7:33:34, 11.49s/step, acc=0.5085, loss=2.0099, lr=2.02e-05]
Training:  79%|███████▊  | 8714/11082 [28:01:17<7:30:32, 11.42s/step, acc=0.5085, loss=2.0099, lr=2.02e-05]
Training:  79%|███████▊  | 8715/11082 [28:01:30<7:44:18, 11.77s/step, acc=0.5085, loss=2.0099, lr=2.02e-05]
Training:  79%|███████▊  | 8716/11082 [28:01:41<7:39:02, 11.64s/step, acc=0.5085, loss=2.0099, lr=2.02e-05]
Training:  79%|███████▊  | 8717/11082 [28:01:53<7:38:47, 11.64s/step, acc=0.5085, loss=2.0099, lr=2.02e-05]
Training:  79%|███████▊  | 8718/11082 [28:02:04<7:36:30, 11.59s/step, acc=0.5085, loss=2.0099, lr=2.02e-05]
Training:  79%|███████▊  | 8719/11082 [28:02:15<7:33:14, 11.51s/step, acc=0.5085, loss=2.0099, lr=2.02e-05]
Training:  79%|███████▊  | 8720/11082 [28:02:27<7:33:43, 11.53s/step, acc=0.5085, loss=2.0099, lr=2.02e-05]
Training:  79%|███████▊  | 8721/11082 [28:02:39<7:38:58, 11.66s/step, acc=0.5085, loss=2.0099, lr=2.02e-05]
Training:  79%|███████▊  | 8722/11082 [28:02:51<7:39:40, 11.69s/step, acc=0.5085, loss=2.0099, lr=2.02e-05]
Training:  79%|███████▊  | 8723/11082 [28:03:02<7:37:21, 11.63s/step, acc=0.5085, loss=2.0099, lr=2.02e-05]
Training:  79%|███████▊  | 8724/11082 [28:03:14<7:37:08, 11.63s/step, acc=0.5085, loss=2.0099, lr=2.02e-05]
Training:  79%|███████▊  | 8725/11082 [28:03:26<7:38:31, 11.67s/step, acc=0.5085, loss=2.0099, lr=2.02e-05]
Training:  79%|███████▊  | 8726/11082 [28:03:38<7:42:53, 11.79s/step, acc=0.5085, loss=2.0099, lr=2.02e-05]
Training:  79%|███████▊  | 8727/11082 [28:03:49<7:40:59, 11.75s/step, acc=0.5085, loss=2.0099, lr=2.02e-05]
Training:  79%|███████▉  | 8728/11082 [28:04:01<7:37:04, 11.65s/step, acc=0.5085, loss=2.0099, lr=2.02e-05]
Training:  79%|███████▉  | 8729/11082 [28:04:12<7:34:38, 11.59s/step, acc=0.5085, loss=2.0099, lr=2.02e-05]
Training:  79%|███████▉  | 8730/11082 [28:04:24<7:38:43, 11.70s/step, acc=0.5085, loss=2.0099, lr=2.02e-05]
Training:  79%|███████▉  | 8731/11082 [28:04:35<7:32:26, 11.55s/step, acc=0.5085, loss=2.0099, lr=2.02e-05]
Training:  79%|███████▉  | 8732/11082 [28:04:47<7:33:05, 11.57s/step, acc=0.5085, loss=2.0099, lr=2.02e-05]
Training:  79%|███████▉  | 8733/11082 [28:04:59<7:34:46, 11.62s/step, acc=0.5085, loss=2.0099, lr=2.02e-05]
Training:  79%|███████▉  | 8734/11082 [28:05:10<7:27:21, 11.43s/step, acc=0.5085, loss=2.0099, lr=2.02e-05]
Training:  79%|███████▉  | 8735/11082 [28:05:22<7:37:16, 11.69s/step, acc=0.5085, loss=2.0099, lr=2.02e-05]
Training:  79%|███████▉  | 8736/11082 [28:05:33<7:33:07, 11.59s/step, acc=0.5085, loss=2.0099, lr=2.02e-05]
Training:  79%|███████▉  | 8737/11082 [28:05:45<7:35:26, 11.65s/step, acc=0.5085, loss=2.0099, lr=2.02e-05]
Training:  79%|███████▉  | 8738/11082 [28:05:57<7:34:19, 11.63s/step, acc=0.5085, loss=2.0099, lr=2.02e-05]
Training:  79%|███████▉  | 8739/11082 [28:06:08<7:34:31, 11.64s/step, acc=0.5085, loss=2.0099, lr=2.02e-05]
Training:  79%|███████▉  | 8740/11082 [28:06:20<7:30:16, 11.54s/step, acc=0.5085, loss=2.0099, lr=2.02e-05]
Training:  79%|███████▉  | 8741/11082 [28:06:32<7:33:16, 11.62s/step, acc=0.5085, loss=2.0099, lr=2.02e-05]
Training:  79%|███████▉  | 8742/11082 [28:06:43<7:34:34, 11.66s/step, acc=0.5085, loss=2.0099, lr=2.02e-05]
Training:  79%|███████▉  | 8743/11082 [28:06:55<7:34:30, 11.66s/step, acc=0.5085, loss=2.0099, lr=2.02e-05]
Training:  79%|███████▉  | 8744/11082 [28:07:06<7:29:46, 11.54s/step, acc=0.5085, loss=2.0099, lr=2.02e-05]
Training:  79%|███████▉  | 8745/11082 [28:07:18<7:28:58, 11.53s/step, acc=0.5085, loss=2.0099, lr=2.02e-05]
Training:  79%|███████▉  | 8746/11082 [28:07:30<7:35:50, 11.71s/step, acc=0.5085, loss=2.0099, lr=2.02e-05]
Training:  79%|███████▉  | 8747/11082 [28:07:41<7:29:59, 11.56s/step, acc=0.5085, loss=2.0099, lr=2.02e-05]
Training:  79%|███████▉  | 8748/11082 [28:07:53<7:30:48, 11.59s/step, acc=0.5085, loss=2.0099, lr=2.02e-05]
Training:  79%|███████▉  | 8749/11082 [28:08:04<7:32:07, 11.63s/step, acc=0.5085, loss=2.0099, lr=2.02e-05]
Training:  79%|███████▉  | 8750/11082 [28:08:16<7:33:18, 11.66s/step, acc=0.5085, loss=2.0099, lr=2.02e-05]
Training:  79%|███████▉  | 8750/11082 [28:08:23<7:33:18, 11.66s/step, acc=0.5057, loss=2.0228, lr=1.98e-05]2026-06-19 19:01:51 | INFO | __main__ | epoch=1 step=8750 train_loss=2.022839 train_acc=0.505727 lr=1.976e-05 grad_norm=0.5547 acc_d0=0.7612 acc_d1=0.6180 acc_d2=0.4977 acc_d3=0.3998 acc_d4=0.3214 acc_d5=0.2584 acc_d6=0.2107 acc_d7=0.1740

Training:  79%|███████▉  | 8751/11082 [28:08:28<7:32:07, 11.64s/step, acc=0.5057, loss=2.0228, lr=1.98e-05]
Training:  79%|███████▉  | 8752/11082 [28:08:39<7:31:50, 11.64s/step, acc=0.5057, loss=2.0228, lr=1.98e-05]
Training:  79%|███████▉  | 8753/11082 [28:08:51<7:33:02, 11.67s/step, acc=0.5057, loss=2.0228, lr=1.98e-05]
Training:  79%|███████▉  | 8754/11082 [28:09:02<7:26:40, 11.51s/step, acc=0.5057, loss=2.0228, lr=1.98e-05]
Training:  79%|███████▉  | 8755/11082 [28:09:14<7:33:26, 11.69s/step, acc=0.5057, loss=2.0228, lr=1.98e-05]
Training:  79%|███████▉  | 8756/11082 [28:09:26<7:27:31, 11.54s/step, acc=0.5057, loss=2.0228, lr=1.98e-05]
Training:  79%|███████▉  | 8757/11082 [28:09:37<7:28:05, 11.56s/step, acc=0.5057, loss=2.0228, lr=1.98e-05]
Training:  79%|███████▉  | 8758/11082 [28:09:49<7:33:44, 11.71s/step, acc=0.5057, loss=2.0228, lr=1.98e-05]
Training:  79%|███████▉  | 8759/11082 [28:10:01<7:29:24, 11.61s/step, acc=0.5057, loss=2.0228, lr=1.98e-05]
Training:  79%|███████▉  | 8760/11082 [28:10:12<7:31:29, 11.67s/step, acc=0.5057, loss=2.0228, lr=1.98e-05]
Training:  79%|███████▉  | 8761/11082 [28:10:23<7:20:19, 11.38s/step, acc=0.5057, loss=2.0228, lr=1.98e-05]
Training:  79%|███████▉  | 8762/11082 [28:10:35<7:30:38, 11.65s/step, acc=0.5057, loss=2.0228, lr=1.98e-05]
Training:  79%|███████▉  | 8763/11082 [28:10:47<7:27:34, 11.58s/step, acc=0.5057, loss=2.0228, lr=1.98e-05]
Training:  79%|███████▉  | 8764/11082 [28:10:59<7:28:51, 11.62s/step, acc=0.5057, loss=2.0228, lr=1.98e-05]
Training:  79%|███████▉  | 8765/11082 [28:11:10<7:24:14, 11.50s/step, acc=0.5057, loss=2.0228, lr=1.98e-05]
Training:  79%|███████▉  | 8766/11082 [28:11:22<7:26:32, 11.57s/step, acc=0.5057, loss=2.0228, lr=1.98e-05]
Training:  79%|███████▉  | 8767/11082 [28:11:33<7:28:29, 11.62s/step, acc=0.5057, loss=2.0228, lr=1.98e-05]
Training:  79%|███████▉  | 8768/11082 [28:11:45<7:29:32, 11.66s/step, acc=0.5057, loss=2.0228, lr=1.98e-05]
Training:  79%|███████▉  | 8769/11082 [28:11:56<7:26:38, 11.59s/step, acc=0.5057, loss=2.0228, lr=1.98e-05]
Training:  79%|███████▉  | 8770/11082 [28:12:08<7:29:01, 11.65s/step, acc=0.5057, loss=2.0228, lr=1.98e-05]
Training:  79%|███████▉  | 8771/11082 [28:12:20<7:28:49, 11.65s/step, acc=0.5057, loss=2.0228, lr=1.98e-05]
Training:  79%|███████▉  | 8772/11082 [28:12:31<7:21:51, 11.48s/step, acc=0.5057, loss=2.0228, lr=1.98e-05]
Training:  79%|███████▉  | 8773/11082 [28:12:43<7:26:19, 11.60s/step, acc=0.5057, loss=2.0228, lr=1.98e-05]
Training:  79%|███████▉  | 8774/11082 [28:12:55<7:28:02, 11.65s/step, acc=0.5057, loss=2.0228, lr=1.98e-05]
Training:  79%|███████▉  | 8775/11082 [28:13:06<7:29:48, 11.70s/step, acc=0.5057, loss=2.0228, lr=1.98e-05]
Training:  79%|███████▉  | 8776/11082 [28:13:18<7:27:13, 11.64s/step, acc=0.5057, loss=2.0228, lr=1.98e-05]
Training:  79%|███████▉  | 8777/11082 [28:13:30<7:26:56, 11.63s/step, acc=0.5057, loss=2.0228, lr=1.98e-05]
Training:  79%|███████▉  | 8778/11082 [28:13:41<7:25:45, 11.61s/step, acc=0.5057, loss=2.0228, lr=1.98e-05]
Training:  79%|███████▉  | 8779/11082 [28:13:53<7:23:20, 11.55s/step, acc=0.5057, loss=2.0228, lr=1.98e-05]
Training:  79%|███████▉  | 8780/11082 [28:14:05<7:28:33, 11.69s/step, acc=0.5057, loss=2.0228, lr=1.98e-05]
Training:  79%|███████▉  | 8781/11082 [28:14:16<7:25:00, 11.60s/step, acc=0.5057, loss=2.0228, lr=1.98e-05]
Training:  79%|███████▉  | 8782/11082 [28:14:28<7:24:34, 11.60s/step, acc=0.5057, loss=2.0228, lr=1.98e-05]
Training:  79%|███████▉  | 8783/11082 [28:14:39<7:24:30, 11.60s/step, acc=0.5057, loss=2.0228, lr=1.98e-05]
Training:  79%|███████▉  | 8784/11082 [28:14:51<7:29:05, 11.73s/step, acc=0.5057, loss=2.0228, lr=1.98e-05]
Training:  79%|███████▉  | 8785/11082 [28:15:02<7:24:16, 11.61s/step, acc=0.5057, loss=2.0228, lr=1.98e-05]
Training:  79%|███████▉  | 8786/11082 [28:15:14<7:26:17, 11.66s/step, acc=0.5057, loss=2.0228, lr=1.98e-05]
Training:  79%|███████▉  | 8787/11082 [28:15:26<7:26:26, 11.67s/step, acc=0.5057, loss=2.0228, lr=1.98e-05]
Training:  79%|███████▉  | 8788/11082 [28:15:37<7:21:49, 11.56s/step, acc=0.5057, loss=2.0228, lr=1.98e-05]
Training:  79%|███████▉  | 8789/11082 [28:15:49<7:27:54, 11.72s/step, acc=0.5057, loss=2.0228, lr=1.98e-05]
Training:  79%|███████▉  | 8790/11082 [28:16:01<7:22:56, 11.60s/step, acc=0.5057, loss=2.0228, lr=1.98e-05]
Training:  79%|███████▉  | 8791/11082 [28:16:13<7:26:23, 11.69s/step, acc=0.5057, loss=2.0228, lr=1.98e-05]
Training:  79%|███████▉  | 8792/11082 [28:16:24<7:23:59, 11.63s/step, acc=0.5057, loss=2.0228, lr=1.98e-05]
Training:  79%|███████▉  | 8793/11082 [28:16:36<7:27:42, 11.74s/step, acc=0.5057, loss=2.0228, lr=1.98e-05]
Training:  79%|███████▉  | 8794/11082 [28:16:47<7:22:10, 11.60s/step, acc=0.5057, loss=2.0228, lr=1.98e-05]
Training:  79%|███████▉  | 8795/11082 [28:16:59<7:21:09, 11.57s/step, acc=0.5057, loss=2.0228, lr=1.98e-05]
Training:  79%|███████▉  | 8796/11082 [28:17:11<7:22:53, 11.62s/step, acc=0.5057, loss=2.0228, lr=1.98e-05]
Training:  79%|███████▉  | 8797/11082 [28:17:22<7:24:07, 11.66s/step, acc=0.5057, loss=2.0228, lr=1.98e-05]
Training:  79%|███████▉  | 8798/11082 [28:17:34<7:24:02, 11.66s/step, acc=0.5057, loss=2.0228, lr=1.98e-05]
Training:  79%|███████▉  | 8799/11082 [28:17:46<7:22:18, 11.62s/step, acc=0.5057, loss=2.0228, lr=1.98e-05]
Training:  79%|███████▉  | 8800/11082 [28:17:57<7:22:07, 11.62s/step, acc=0.5057, loss=2.0228, lr=1.98e-05]
Training:  79%|███████▉  | 8800/11082 [28:18:04<7:22:07, 11.62s/step, acc=0.5070, loss=2.0153, lr=1.94e-05]2026-06-19 19:11:32 | INFO | __main__ | epoch=1 step=8800 train_loss=2.015319 train_acc=0.507031 lr=1.936e-05 grad_norm=0.5117 acc_d0=0.7682 acc_d1=0.6270 acc_d2=0.5067 acc_d3=0.4089 acc_d4=0.3292 acc_d5=0.2658 acc_d6=0.2167 acc_d7=0.1780

Training:  79%|███████▉  | 8801/11082 [28:18:09<7:22:38, 11.64s/step, acc=0.5070, loss=2.0153, lr=1.94e-05]
Training:  79%|███████▉  | 8802/11082 [28:18:20<7:18:04, 11.53s/step, acc=0.5070, loss=2.0153, lr=1.94e-05]
Training:  79%|███████▉  | 8803/11082 [28:18:32<7:24:53, 11.71s/step, acc=0.5070, loss=2.0153, lr=1.94e-05]
Training:  79%|███████▉  | 8804/11082 [28:18:44<7:22:02, 11.64s/step, acc=0.5070, loss=2.0153, lr=1.94e-05]
Training:  79%|███████▉  | 8805/11082 [28:18:55<7:20:51, 11.62s/step, acc=0.5070, loss=2.0153, lr=1.94e-05]
Training:  79%|███████▉  | 8806/11082 [28:19:07<7:18:36, 11.56s/step, acc=0.5070, loss=2.0153, lr=1.94e-05]
Training:  79%|███████▉  | 8807/11082 [28:19:19<7:28:05, 11.82s/step, acc=0.5070, loss=2.0153, lr=1.94e-05]
Training:  79%|███████▉  | 8808/11082 [28:19:30<7:21:12, 11.64s/step, acc=0.5070, loss=2.0153, lr=1.94e-05]
Training:  79%|███████▉  | 8809/11082 [28:19:42<7:20:30, 11.63s/step, acc=0.5070, loss=2.0153, lr=1.94e-05]
Training:  79%|███████▉  | 8810/11082 [28:19:54<7:21:31, 11.66s/step, acc=0.5070, loss=2.0153, lr=1.94e-05]
Training:  80%|███████▉  | 8811/11082 [28:20:05<7:20:39, 11.64s/step, acc=0.5070, loss=2.0153, lr=1.94e-05]
Training:  80%|███████▉  | 8812/11082 [28:20:16<7:14:35, 11.49s/step, acc=0.5070, loss=2.0153, lr=1.94e-05]
Training:  80%|███████▉  | 8813/11082 [28:20:29<7:26:13, 11.80s/step, acc=0.5070, loss=2.0153, lr=1.94e-05]
Training:  80%|███████▉  | 8814/11082 [28:20:41<7:23:25, 11.73s/step, acc=0.5070, loss=2.0153, lr=1.94e-05]
Training:  80%|███████▉  | 8815/11082 [28:20:52<7:20:00, 11.65s/step, acc=0.5070, loss=2.0153, lr=1.94e-05]
Training:  80%|███████▉  | 8816/11082 [28:21:03<7:12:50, 11.46s/step, acc=0.5070, loss=2.0153, lr=1.94e-05]
Training:  80%|███████▉  | 8817/11082 [28:21:15<7:24:16, 11.77s/step, acc=0.5070, loss=2.0153, lr=1.94e-05]
Training:  80%|███████▉  | 8818/11082 [28:21:27<7:20:09, 11.66s/step, acc=0.5070, loss=2.0153, lr=1.94e-05]
Training:  80%|███████▉  | 8819/11082 [28:21:39<7:19:59, 11.67s/step, acc=0.5070, loss=2.0153, lr=1.94e-05]
Training:  80%|███████▉  | 8820/11082 [28:21:50<7:19:23, 11.65s/step, acc=0.5070, loss=2.0153, lr=1.94e-05]
Training:  80%|███████▉  | 8821/11082 [28:22:02<7:19:57, 11.67s/step, acc=0.5070, loss=2.0153, lr=1.94e-05]
Training:  80%|███████▉  | 8822/11082 [28:22:13<7:16:16, 11.58s/step, acc=0.5070, loss=2.0153, lr=1.94e-05]
Training:  80%|███████▉  | 8823/11082 [28:22:25<7:18:45, 11.65s/step, acc=0.5070, loss=2.0153, lr=1.94e-05]
Training:  80%|███████▉  | 8824/11082 [28:22:36<7:14:27, 11.54s/step, acc=0.5070, loss=2.0153, lr=1.94e-05]
Training:  80%|███████▉  | 8825/11082 [28:22:48<7:18:27, 11.66s/step, acc=0.5070, loss=2.0153, lr=1.94e-05]
Training:  80%|███████▉  | 8826/11082 [28:23:00<7:16:32, 11.61s/step, acc=0.5070, loss=2.0153, lr=1.94e-05]
Training:  80%|███████▉  | 8827/11082 [28:23:11<7:15:25, 11.59s/step, acc=0.5070, loss=2.0153, lr=1.94e-05]
Training:  80%|███████▉  | 8828/11082 [28:23:23<7:17:08, 11.64s/step, acc=0.5070, loss=2.0153, lr=1.94e-05]
Training:  80%|███████▉  | 8829/11082 [28:23:35<7:17:21, 11.65s/step, acc=0.5070, loss=2.0153, lr=1.94e-05]
Training:  80%|███████▉  | 8830/11082 [28:23:46<7:13:25, 11.55s/step, acc=0.5070, loss=2.0153, lr=1.94e-05]
Training:  80%|███████▉  | 8831/11082 [28:23:58<7:14:35, 11.58s/step, acc=0.5070, loss=2.0153, lr=1.94e-05]
Training:  80%|███████▉  | 8832/11082 [28:24:10<7:17:23, 11.66s/step, acc=0.5070, loss=2.0153, lr=1.94e-05]
Training:  80%|███████▉  | 8833/11082 [28:24:21<7:14:12, 11.58s/step, acc=0.5070, loss=2.0153, lr=1.94e-05]
Training:  80%|███████▉  | 8834/11082 [28:24:32<7:10:49, 11.50s/step, acc=0.5070, loss=2.0153, lr=1.94e-05]
Training:  80%|███████▉  | 8835/11082 [28:24:44<7:15:37, 11.63s/step, acc=0.5070, loss=2.0153, lr=1.94e-05]
Training:  80%|███████▉  | 8836/11082 [28:24:56<7:14:23, 11.60s/step, acc=0.5070, loss=2.0153, lr=1.94e-05]
Training:  80%|███████▉  | 8837/11082 [28:25:07<7:12:54, 11.57s/step, acc=0.5070, loss=2.0153, lr=1.94e-05]
Training:  80%|███████▉  | 8838/11082 [28:25:19<7:09:39, 11.49s/step, acc=0.5070, loss=2.0153, lr=1.94e-05]
Training:  80%|███████▉  | 8839/11082 [28:25:30<7:13:19, 11.59s/step, acc=0.5070, loss=2.0153, lr=1.94e-05]
Training:  80%|███████▉  | 8840/11082 [28:25:42<7:13:54, 11.61s/step, acc=0.5070, loss=2.0153, lr=1.94e-05]
Training:  80%|███████▉  | 8841/11082 [28:25:54<7:12:44, 11.59s/step, acc=0.5070, loss=2.0153, lr=1.94e-05]
Training:  80%|███████▉  | 8842/11082 [28:26:05<7:13:47, 11.62s/step, acc=0.5070, loss=2.0153, lr=1.94e-05]
Training:  80%|███████▉  | 8843/11082 [28:26:17<7:18:23, 11.75s/step, acc=0.5070, loss=2.0153, lr=1.94e-05]
Training:  80%|███████▉  | 8844/11082 [28:26:28<7:11:19, 11.56s/step, acc=0.5070, loss=2.0153, lr=1.94e-05]
Training:  80%|███████▉  | 8845/11082 [28:26:40<7:07:19, 11.46s/step, acc=0.5070, loss=2.0153, lr=1.94e-05]
Training:  80%|███████▉  | 8846/11082 [28:26:51<7:09:00, 11.51s/step, acc=0.5070, loss=2.0153, lr=1.94e-05]
Training:  80%|███████▉  | 8847/11082 [28:27:03<7:09:22, 11.53s/step, acc=0.5070, loss=2.0153, lr=1.94e-05]
Training:  80%|███████▉  | 8848/11082 [28:27:15<7:16:41, 11.73s/step, acc=0.5070, loss=2.0153, lr=1.94e-05]
Training:  80%|███████▉  | 8849/11082 [28:27:27<7:13:41, 11.65s/step, acc=0.5070, loss=2.0153, lr=1.94e-05]
Training:  80%|███████▉  | 8850/11082 [28:27:38<7:13:51, 11.66s/step, acc=0.5070, loss=2.0153, lr=1.94e-05]
Training:  80%|███████▉  | 8850/11082 [28:27:45<7:13:51, 11.66s/step, acc=0.5087, loss=2.0020, lr=1.90e-05]2026-06-19 19:21:13 | INFO | __main__ | epoch=1 step=8850 train_loss=2.001999 train_acc=0.508671 lr=1.897e-05 grad_norm=0.5312 acc_d0=0.7657 acc_d1=0.6222 acc_d2=0.5013 acc_d3=0.4056 acc_d4=0.3271 acc_d5=0.2648 acc_d6=0.2151 acc_d7=0.1769

Training:  80%|███████▉  | 8851/11082 [28:27:49<7:05:43, 11.45s/step, acc=0.5087, loss=2.0020, lr=1.90e-05]
Training:  80%|███████▉  | 8852/11082 [28:28:02<7:22:31, 11.91s/step, acc=0.5087, loss=2.0020, lr=1.90e-05]
Training:  80%|███████▉  | 8853/11082 [28:28:13<7:12:48, 11.65s/step, acc=0.5087, loss=2.0020, lr=1.90e-05]
Training:  80%|███████▉  | 8854/11082 [28:28:25<7:10:27, 11.59s/step, acc=0.5087, loss=2.0020, lr=1.90e-05]
Training:  80%|███████▉  | 8855/11082 [28:28:36<7:11:33, 11.63s/step, acc=0.5087, loss=2.0020, lr=1.90e-05]
Training:  80%|███████▉  | 8856/11082 [28:28:48<7:12:01, 11.64s/step, acc=0.5087, loss=2.0020, lr=1.90e-05]
Training:  80%|███████▉  | 8857/11082 [28:29:00<7:11:18, 11.63s/step, acc=0.5087, loss=2.0020, lr=1.90e-05]
Training:  80%|███████▉  | 8858/11082 [28:29:11<7:08:24, 11.56s/step, acc=0.5087, loss=2.0020, lr=1.90e-05]
Training:  80%|███████▉  | 8859/11082 [28:29:23<7:11:34, 11.65s/step, acc=0.5087, loss=2.0020, lr=1.90e-05]
Training:  80%|███████▉  | 8860/11082 [28:29:34<7:08:58, 11.58s/step, acc=0.5087, loss=2.0020, lr=1.90e-05]
Training:  80%|███████▉  | 8861/11082 [28:29:46<7:12:25, 11.68s/step, acc=0.5087, loss=2.0020, lr=1.90e-05]
Training:  80%|███████▉  | 8862/11082 [28:29:57<7:06:25, 11.53s/step, acc=0.5087, loss=2.0020, lr=1.90e-05]
Training:  80%|███████▉  | 8863/11082 [28:30:09<7:08:26, 11.58s/step, acc=0.5087, loss=2.0020, lr=1.90e-05]
Training:  80%|███████▉  | 8864/11082 [28:30:21<7:10:56, 11.66s/step, acc=0.5087, loss=2.0020, lr=1.90e-05]
Training:  80%|███████▉  | 8865/11082 [28:30:32<7:08:29, 11.60s/step, acc=0.5087, loss=2.0020, lr=1.90e-05]
Training:  80%|████████  | 8866/11082 [28:30:44<7:11:24, 11.68s/step, acc=0.5087, loss=2.0020, lr=1.90e-05]
Training:  80%|████████  | 8867/11082 [28:30:56<7:07:51, 11.59s/step, acc=0.5087, loss=2.0020, lr=1.90e-05]
Training:  80%|████████  | 8868/11082 [28:31:07<7:06:47, 11.57s/step, acc=0.5087, loss=2.0020, lr=1.90e-05]
Training:  80%|████████  | 8869/11082 [28:31:19<7:05:28, 11.54s/step, acc=0.5087, loss=2.0020, lr=1.90e-05]
Training:  80%|████████  | 8870/11082 [28:31:31<7:09:05, 11.64s/step, acc=0.5087, loss=2.0020, lr=1.90e-05]
Training:  80%|████████  | 8871/11082 [28:31:42<7:06:17, 11.57s/step, acc=0.5087, loss=2.0020, lr=1.90e-05]
Training:  80%|████████  | 8872/11082 [28:31:54<7:08:15, 11.63s/step, acc=0.5087, loss=2.0020, lr=1.90e-05]
Training:  80%|████████  | 8873/11082 [28:32:05<7:08:10, 11.63s/step, acc=0.5087, loss=2.0020, lr=1.90e-05]
Training:  80%|████████  | 8874/11082 [28:32:16<6:57:45, 11.35s/step, acc=0.5087, loss=2.0020, lr=1.90e-05]
Training:  80%|████████  | 8875/11082 [28:32:28<7:07:51, 11.63s/step, acc=0.5087, loss=2.0020, lr=1.90e-05]
Training:  80%|████████  | 8876/11082 [28:32:40<7:04:06, 11.54s/step, acc=0.5087, loss=2.0020, lr=1.90e-05]
Training:  80%|████████  | 8877/11082 [28:32:52<7:09:17, 11.68s/step, acc=0.5087, loss=2.0020, lr=1.90e-05]
Training:  80%|████████  | 8878/11082 [28:33:03<7:04:58, 11.57s/step, acc=0.5087, loss=2.0020, lr=1.90e-05]
Training:  80%|████████  | 8879/11082 [28:33:15<7:11:13, 11.74s/step, acc=0.5087, loss=2.0020, lr=1.90e-05]
Training:  80%|████████  | 8880/11082 [28:33:27<7:07:02, 11.64s/step, acc=0.5087, loss=2.0020, lr=1.90e-05]
Training:  80%|████████  | 8881/11082 [28:33:38<7:00:59, 11.48s/step, acc=0.5087, loss=2.0020, lr=1.90e-05]
Training:  80%|████████  | 8882/11082 [28:33:49<7:00:10, 11.46s/step, acc=0.5087, loss=2.0020, lr=1.90e-05]
Training:  80%|████████  | 8883/11082 [28:34:01<7:03:30, 11.56s/step, acc=0.5087, loss=2.0020, lr=1.90e-05]
Training:  80%|████████  | 8884/11082 [28:34:13<7:06:21, 11.64s/step, acc=0.5087, loss=2.0020, lr=1.90e-05]
Training:  80%|████████  | 8885/11082 [28:34:24<7:05:22, 11.62s/step, acc=0.5087, loss=2.0020, lr=1.90e-05]
Training:  80%|████████  | 8886/11082 [28:34:36<7:02:57, 11.56s/step, acc=0.5087, loss=2.0020, lr=1.90e-05]
Training:  80%|████████  | 8887/11082 [28:34:47<7:03:25, 11.57s/step, acc=0.5087, loss=2.0020, lr=1.90e-05]
Training:  80%|████████  | 8888/11082 [28:34:59<7:06:49, 11.67s/step, acc=0.5087, loss=2.0020, lr=1.90e-05]
Training:  80%|████████  | 8889/11082 [28:35:10<7:01:19, 11.53s/step, acc=0.5087, loss=2.0020, lr=1.90e-05]
Training:  80%|████████  | 8890/11082 [28:35:22<7:00:12, 11.50s/step, acc=0.5087, loss=2.0020, lr=1.90e-05]
Training:  80%|████████  | 8891/11082 [28:35:33<7:00:10, 11.51s/step, acc=0.5087, loss=2.0020, lr=1.90e-05]
Training:  80%|████████  | 8892/11082 [28:35:45<6:58:51, 11.48s/step, acc=0.5087, loss=2.0020, lr=1.90e-05]
Training:  80%|████████  | 8893/11082 [28:35:56<6:58:29, 11.47s/step, acc=0.5087, loss=2.0020, lr=1.90e-05]
Training:  80%|████████  | 8894/11082 [28:36:08<6:58:40, 11.48s/step, acc=0.5087, loss=2.0020, lr=1.90e-05]
Training:  80%|████████  | 8895/11082 [28:36:19<7:02:13, 11.58s/step, acc=0.5087, loss=2.0020, lr=1.90e-05]
Training:  80%|████████  | 8896/11082 [28:36:31<7:01:09, 11.56s/step, acc=0.5087, loss=2.0020, lr=1.90e-05]
Training:  80%|████████  | 8897/11082 [28:36:43<7:01:42, 11.58s/step, acc=0.5087, loss=2.0020, lr=1.90e-05]
Training:  80%|████████  | 8898/11082 [28:36:54<7:03:44, 11.64s/step, acc=0.5087, loss=2.0020, lr=1.90e-05]
Training:  80%|████████  | 8899/11082 [28:37:06<7:01:24, 11.58s/step, acc=0.5087, loss=2.0020, lr=1.90e-05]
Training:  80%|████████  | 8900/11082 [28:37:17<7:00:21, 11.56s/step, acc=0.5087, loss=2.0020, lr=1.90e-05]
Training:  80%|████████  | 8900/11082 [28:37:25<7:00:21, 11.56s/step, acc=0.5058, loss=2.0266, lr=1.86e-05]2026-06-19 19:30:52 | INFO | __main__ | epoch=1 step=8900 train_loss=2.026572 train_acc=0.505783 lr=1.859e-05 grad_norm=0.5820 acc_d0=0.7659 acc_d1=0.6253 acc_d2=0.5073 acc_d3=0.4122 acc_d4=0.3342 acc_d5=0.2697 acc_d6=0.2228 acc_d7=0.1817

Training:  80%|████████  | 8901/11082 [28:37:29<7:03:20, 11.65s/step, acc=0.5058, loss=2.0266, lr=1.86e-05]
Training:  80%|████████  | 8902/11082 [28:37:40<6:57:15, 11.48s/step, acc=0.5058, loss=2.0266, lr=1.86e-05]
Training:  80%|████████  | 8903/11082 [28:37:52<6:56:18, 11.46s/step, acc=0.5058, loss=2.0266, lr=1.86e-05]
Training:  80%|████████  | 8904/11082 [28:38:03<6:57:58, 11.51s/step, acc=0.5058, loss=2.0266, lr=1.86e-05]
Training:  80%|████████  | 8905/11082 [28:38:15<6:58:56, 11.55s/step, acc=0.5058, loss=2.0266, lr=1.86e-05]
Training:  80%|████████  | 8906/11082 [28:38:27<7:01:09, 11.61s/step, acc=0.5058, loss=2.0266, lr=1.86e-05]
Training:  80%|████████  | 8907/11082 [28:38:39<7:03:13, 11.68s/step, acc=0.5058, loss=2.0266, lr=1.86e-05]
Training:  80%|████████  | 8908/11082 [28:38:50<6:57:41, 11.53s/step, acc=0.5058, loss=2.0266, lr=1.86e-05]
Training:  80%|████████  | 8909/11082 [28:39:01<6:54:59, 11.46s/step, acc=0.5058, loss=2.0266, lr=1.86e-05]
Training:  80%|████████  | 8910/11082 [28:39:13<6:57:57, 11.55s/step, acc=0.5058, loss=2.0266, lr=1.86e-05]
Training:  80%|████████  | 8911/11082 [28:39:25<7:00:18, 11.62s/step, acc=0.5058, loss=2.0266, lr=1.86e-05]
Training:  80%|████████  | 8912/11082 [28:39:36<6:59:55, 11.61s/step, acc=0.5058, loss=2.0266, lr=1.86e-05]
Training:  80%|████████  | 8913/11082 [28:39:47<6:52:47, 11.42s/step, acc=0.5058, loss=2.0266, lr=1.86e-05]
Training:  80%|████████  | 8914/11082 [28:39:59<6:58:32, 11.58s/step, acc=0.5058, loss=2.0266, lr=1.86e-05]
Training:  80%|████████  | 8915/11082 [28:40:11<7:02:23, 11.70s/step, acc=0.5058, loss=2.0266, lr=1.86e-05]
Training:  80%|████████  | 8916/11082 [28:40:22<6:56:24, 11.53s/step, acc=0.5058, loss=2.0266, lr=1.86e-05]
Training:  80%|████████  | 8917/11082 [28:40:34<6:59:31, 11.63s/step, acc=0.5058, loss=2.0266, lr=1.86e-05]
Training:  80%|████████  | 8918/11082 [28:40:46<6:59:15, 11.62s/step, acc=0.5058, loss=2.0266, lr=1.86e-05]
Training:  80%|████████  | 8919/11082 [28:40:57<7:00:12, 11.66s/step, acc=0.5058, loss=2.0266, lr=1.86e-05]
Training:  80%|████████  | 8920/11082 [28:41:09<6:56:06, 11.55s/step, acc=0.5058, loss=2.0266, lr=1.86e-05]
Training:  80%|████████  | 8921/11082 [28:41:21<6:59:33, 11.65s/step, acc=0.5058, loss=2.0266, lr=1.86e-05]
Training:  81%|████████  | 8922/11082 [28:41:32<6:59:50, 11.66s/step, acc=0.5058, loss=2.0266, lr=1.86e-05]
Training:  81%|████████  | 8923/11082 [28:41:44<7:00:32, 11.69s/step, acc=0.5058, loss=2.0266, lr=1.86e-05]
Training:  81%|████████  | 8924/11082 [28:41:55<6:52:27, 11.47s/step, acc=0.5058, loss=2.0266, lr=1.86e-05]
Training:  81%|████████  | 8925/11082 [28:42:06<6:52:45, 11.48s/step, acc=0.5058, loss=2.0266, lr=1.86e-05]
Training:  81%|████████  | 8926/11082 [28:42:18<6:53:26, 11.51s/step, acc=0.5058, loss=2.0266, lr=1.86e-05]
Training:  81%|████████  | 8927/11082 [28:42:30<6:57:27, 11.62s/step, acc=0.5058, loss=2.0266, lr=1.86e-05]
Training:  81%|████████  | 8928/11082 [28:42:41<6:51:42, 11.47s/step, acc=0.5058, loss=2.0266, lr=1.86e-05]
Training:  81%|████████  | 8929/11082 [28:42:53<6:53:19, 11.52s/step, acc=0.5058, loss=2.0266, lr=1.86e-05]
Training:  81%|████████  | 8930/11082 [28:43:04<6:55:40, 11.59s/step, acc=0.5058, loss=2.0266, lr=1.86e-05]
Training:  81%|████████  | 8931/11082 [28:43:16<6:54:56, 11.57s/step, acc=0.5058, loss=2.0266, lr=1.86e-05]
Training:  81%|████████  | 8932/11082 [28:43:28<6:54:02, 11.55s/step, acc=0.5058, loss=2.0266, lr=1.86e-05]
Training:  81%|████████  | 8933/11082 [28:43:39<6:51:02, 11.48s/step, acc=0.5058, loss=2.0266, lr=1.86e-05]
Training:  81%|████████  | 8934/11082 [28:43:51<6:55:10, 11.60s/step, acc=0.5058, loss=2.0266, lr=1.86e-05]
Training:  81%|████████  | 8935/11082 [28:44:02<6:56:16, 11.63s/step, acc=0.5058, loss=2.0266, lr=1.86e-05]
Training:  81%|████████  | 8936/11082 [28:44:14<6:56:00, 11.63s/step, acc=0.5058, loss=2.0266, lr=1.86e-05]
Training:  81%|████████  | 8937/11082 [28:44:26<6:55:42, 11.63s/step, acc=0.5058, loss=2.0266, lr=1.86e-05]
Training:  81%|████████  | 8938/11082 [28:44:37<6:51:43, 11.52s/step, acc=0.5058, loss=2.0266, lr=1.86e-05]
Training:  81%|████████  | 8939/11082 [28:44:48<6:50:51, 11.50s/step, acc=0.5058, loss=2.0266, lr=1.86e-05]
Training:  81%|████████  | 8940/11082 [28:45:00<6:54:19, 11.61s/step, acc=0.5058, loss=2.0266, lr=1.86e-05]
Training:  81%|████████  | 8941/11082 [28:45:12<6:53:49, 11.60s/step, acc=0.5058, loss=2.0266, lr=1.86e-05]
Training:  81%|████████  | 8942/11082 [28:45:23<6:51:32, 11.54s/step, acc=0.5058, loss=2.0266, lr=1.86e-05]
Training:  81%|████████  | 8943/11082 [28:45:35<6:55:15, 11.65s/step, acc=0.5058, loss=2.0266, lr=1.86e-05]
Training:  81%|████████  | 8944/11082 [28:45:46<6:52:07, 11.57s/step, acc=0.5058, loss=2.0266, lr=1.86e-05]
Training:  81%|████████  | 8945/11082 [28:45:58<6:51:28, 11.55s/step, acc=0.5058, loss=2.0266, lr=1.86e-05]
Training:  81%|████████  | 8946/11082 [28:46:10<6:50:45, 11.54s/step, acc=0.5058, loss=2.0266, lr=1.86e-05]
Training:  81%|████████  | 8947/11082 [28:46:21<6:52:09, 11.58s/step, acc=0.5058, loss=2.0266, lr=1.86e-05]
Training:  81%|████████  | 8948/11082 [28:46:33<6:50:51, 11.55s/step, acc=0.5058, loss=2.0266, lr=1.86e-05]
Training:  81%|████████  | 8949/11082 [28:46:44<6:50:42, 11.55s/step, acc=0.5058, loss=2.0266, lr=1.86e-05]
Training:  81%|████████  | 8950/11082 [28:46:56<6:49:38, 11.53s/step, acc=0.5058, loss=2.0266, lr=1.86e-05]
Training:  81%|████████  | 8950/11082 [28:47:03<6:49:38, 11.53s/step, acc=0.5074, loss=2.0166, lr=1.82e-05]2026-06-19 19:40:31 | INFO | __main__ | epoch=1 step=8950 train_loss=2.016556 train_acc=0.507421 lr=1.821e-05 grad_norm=0.5898 acc_d0=0.7642 acc_d1=0.6229 acc_d2=0.5047 acc_d3=0.4086 acc_d4=0.3325 acc_d5=0.2695 acc_d6=0.2192 acc_d7=0.1794

Training:  81%|████████  | 8951/11082 [28:47:07<6:52:01, 11.60s/step, acc=0.5074, loss=2.0166, lr=1.82e-05]
Training:  81%|████████  | 8952/11082 [28:47:19<6:51:32, 11.59s/step, acc=0.5074, loss=2.0166, lr=1.82e-05]
Training:  81%|████████  | 8953/11082 [28:47:30<6:47:38, 11.49s/step, acc=0.5074, loss=2.0166, lr=1.82e-05]
Training:  81%|████████  | 8954/11082 [28:47:42<6:48:43, 11.52s/step, acc=0.5074, loss=2.0166, lr=1.82e-05]
Training:  81%|████████  | 8955/11082 [28:47:53<6:45:31, 11.44s/step, acc=0.5074, loss=2.0166, lr=1.82e-05]
Training:  81%|████████  | 8956/11082 [28:48:05<6:49:55, 11.57s/step, acc=0.5074, loss=2.0166, lr=1.82e-05]
Training:  81%|████████  | 8957/11082 [28:48:17<6:51:01, 11.61s/step, acc=0.5074, loss=2.0166, lr=1.82e-05]
Training:  81%|████████  | 8958/11082 [28:48:28<6:45:48, 11.46s/step, acc=0.5074, loss=2.0166, lr=1.82e-05]
Training:  81%|████████  | 8959/11082 [28:48:39<6:46:01, 11.48s/step, acc=0.5074, loss=2.0166, lr=1.82e-05]
Training:  81%|████████  | 8960/11082 [28:48:51<6:48:33, 11.55s/step, acc=0.5074, loss=2.0166, lr=1.82e-05]
Training:  81%|████████  | 8961/11082 [28:49:03<6:47:17, 11.52s/step, acc=0.5074, loss=2.0166, lr=1.82e-05]
Training:  81%|████████  | 8962/11082 [28:49:14<6:45:20, 11.47s/step, acc=0.5074, loss=2.0166, lr=1.82e-05]
Training:  81%|████████  | 8963/11082 [28:49:26<6:47:26, 11.54s/step, acc=0.5074, loss=2.0166, lr=1.82e-05]
Training:  81%|████████  | 8964/11082 [28:49:37<6:48:56, 11.58s/step, acc=0.5074, loss=2.0166, lr=1.82e-05]
Training:  81%|████████  | 8965/11082 [28:49:49<6:47:10, 11.54s/step, acc=0.5074, loss=2.0166, lr=1.82e-05]
Training:  81%|████████  | 8966/11082 [28:50:00<6:48:48, 11.59s/step, acc=0.5074, loss=2.0166, lr=1.82e-05]
Training:  81%|████████  | 8967/11082 [28:50:12<6:47:36, 11.56s/step, acc=0.5074, loss=2.0166, lr=1.82e-05]
Training:  81%|████████  | 8968/11082 [28:50:24<6:48:24, 11.59s/step, acc=0.5074, loss=2.0166, lr=1.82e-05]
Training:  81%|████████  | 8969/11082 [28:50:35<6:46:18, 11.54s/step, acc=0.5074, loss=2.0166, lr=1.82e-05]
Training:  81%|████████  | 8970/11082 [28:50:47<6:49:43, 11.64s/step, acc=0.5074, loss=2.0166, lr=1.82e-05]
Training:  81%|████████  | 8971/11082 [28:50:59<6:50:23, 11.66s/step, acc=0.5074, loss=2.0166, lr=1.82e-05]
Training:  81%|████████  | 8972/11082 [28:51:10<6:49:35, 11.65s/step, acc=0.5074, loss=2.0166, lr=1.82e-05]
Training:  81%|████████  | 8973/11082 [28:51:22<6:46:09, 11.55s/step, acc=0.5074, loss=2.0166, lr=1.82e-05]
Training:  81%|████████  | 8974/11082 [28:51:33<6:45:41, 11.55s/step, acc=0.5074, loss=2.0166, lr=1.82e-05]
Training:  81%|████████  | 8975/11082 [28:51:45<6:45:15, 11.54s/step, acc=0.5074, loss=2.0166, lr=1.82e-05]
Training:  81%|████████  | 8976/11082 [28:51:56<6:45:15, 11.55s/step, acc=0.5074, loss=2.0166, lr=1.82e-05]
Training:  81%|████████  | 8977/11082 [28:52:08<6:48:17, 11.64s/step, acc=0.5074, loss=2.0166, lr=1.82e-05]
Training:  81%|████████  | 8978/11082 [28:52:19<6:42:07, 11.47s/step, acc=0.5074, loss=2.0166, lr=1.82e-05]
Training:  81%|████████  | 8979/11082 [28:52:31<6:50:13, 11.70s/step, acc=0.5074, loss=2.0166, lr=1.82e-05]
Training:  81%|████████  | 8980/11082 [28:52:43<6:44:58, 11.56s/step, acc=0.5074, loss=2.0166, lr=1.82e-05]
Training:  81%|████████  | 8981/11082 [28:52:54<6:44:38, 11.56s/step, acc=0.5074, loss=2.0166, lr=1.82e-05]
Training:  81%|████████  | 8982/11082 [28:53:06<6:44:38, 11.56s/step, acc=0.5074, loss=2.0166, lr=1.82e-05]
Training:  81%|████████  | 8983/11082 [28:53:17<6:42:27, 11.50s/step, acc=0.5074, loss=2.0166, lr=1.82e-05]
Training:  81%|████████  | 8984/11082 [28:53:29<6:43:08, 11.53s/step, acc=0.5074, loss=2.0166, lr=1.82e-05]
Training:  81%|████████  | 8985/11082 [28:53:41<6:48:37, 11.69s/step, acc=0.5074, loss=2.0166, lr=1.82e-05]
Training:  81%|████████  | 8986/11082 [28:53:52<6:47:37, 11.67s/step, acc=0.5074, loss=2.0166, lr=1.82e-05]
Training:  81%|████████  | 8987/11082 [28:54:04<6:46:04, 11.63s/step, acc=0.5074, loss=2.0166, lr=1.82e-05]
Training:  81%|████████  | 8988/11082 [28:54:16<6:46:53, 11.66s/step, acc=0.5074, loss=2.0166, lr=1.82e-05]
Training:  81%|████████  | 8989/11082 [28:54:27<6:46:23, 11.65s/step, acc=0.5074, loss=2.0166, lr=1.82e-05]
Training:  81%|████████  | 8990/11082 [28:54:38<6:41:37, 11.52s/step, acc=0.5074, loss=2.0166, lr=1.82e-05]
Training:  81%|████████  | 8991/11082 [28:54:50<6:43:10, 11.57s/step, acc=0.5074, loss=2.0166, lr=1.82e-05]
Training:  81%|████████  | 8992/11082 [28:55:02<6:43:06, 11.57s/step, acc=0.5074, loss=2.0166, lr=1.82e-05]
Training:  81%|████████  | 8993/11082 [28:55:13<6:41:36, 11.53s/step, acc=0.5074, loss=2.0166, lr=1.82e-05]
Training:  81%|████████  | 8994/11082 [28:55:25<6:45:56, 11.67s/step, acc=0.5074, loss=2.0166, lr=1.82e-05]
Training:  81%|████████  | 8995/11082 [28:55:37<6:43:19, 11.60s/step, acc=0.5074, loss=2.0166, lr=1.82e-05]
Training:  81%|████████  | 8996/11082 [28:55:48<6:39:58, 11.50s/step, acc=0.5074, loss=2.0166, lr=1.82e-05]
Training:  81%|████████  | 8997/11082 [28:56:00<6:44:38, 11.64s/step, acc=0.5074, loss=2.0166, lr=1.82e-05]
Training:  81%|████████  | 8998/11082 [28:56:11<6:43:29, 11.62s/step, acc=0.5074, loss=2.0166, lr=1.82e-05]
Training:  81%|████████  | 8999/11082 [28:56:23<6:41:13, 11.56s/step, acc=0.5074, loss=2.0166, lr=1.82e-05]
Training:  81%|████████  | 9000/11082 [28:56:34<6:38:29, 11.48s/step, acc=0.5074, loss=2.0166, lr=1.82e-05]
Training:  81%|████████  | 9000/11082 [28:56:41<6:38:29, 11.48s/step, acc=0.5074, loss=2.0164, lr=1.78e-05]2026-06-19 19:50:09 | INFO | __main__ | epoch=1 step=9000 train_loss=2.016360 train_acc=0.507362 lr=1.784e-05 grad_norm=0.4980 acc_d0=0.7609 acc_d1=0.6175 acc_d2=0.4995 acc_d3=0.4015 acc_d4=0.3249 acc_d5=0.2619 acc_d6=0.2109 acc_d7=0.1727

Training:  81%|████████  | 9001/11082 [28:56:46<6:38:58, 11.50s/step, acc=0.5074, loss=2.0164, lr=1.78e-05]
Training:  81%|████████  | 9002/11082 [28:56:57<6:36:43, 11.44s/step, acc=0.5074, loss=2.0164, lr=1.78e-05]
Training:  81%|████████  | 9003/11082 [28:57:09<6:39:19, 11.52s/step, acc=0.5074, loss=2.0164, lr=1.78e-05]
Training:  81%|████████  | 9004/11082 [28:57:20<6:39:38, 11.54s/step, acc=0.5074, loss=2.0164, lr=1.78e-05]
Training:  81%|████████▏ | 9005/11082 [28:57:32<6:41:18, 11.59s/step, acc=0.5074, loss=2.0164, lr=1.78e-05]
Training:  81%|████████▏ | 9006/11082 [28:57:44<6:41:19, 11.60s/step, acc=0.5074, loss=2.0164, lr=1.78e-05]
Training:  81%|████████▏ | 9007/11082 [28:57:55<6:43:18, 11.66s/step, acc=0.5074, loss=2.0164, lr=1.78e-05]
Training:  81%|████████▏ | 9008/11082 [28:58:06<6:36:15, 11.46s/step, acc=0.5074, loss=2.0164, lr=1.78e-05]
Training:  81%|████████▏ | 9009/11082 [28:58:18<6:42:43, 11.66s/step, acc=0.5074, loss=2.0164, lr=1.78e-05]
Training:  81%|████████▏ | 9010/11082 [28:58:30<6:39:57, 11.58s/step, acc=0.5074, loss=2.0164, lr=1.78e-05]
Training:  81%|████████▏ | 9011/11082 [28:58:42<6:42:04, 11.65s/step, acc=0.5074, loss=2.0164, lr=1.78e-05]
Training:  81%|████████▏ | 9012/11082 [28:58:53<6:36:14, 11.49s/step, acc=0.5074, loss=2.0164, lr=1.78e-05]
Training:  81%|████████▏ | 9013/11082 [28:59:05<6:39:00, 11.57s/step, acc=0.5074, loss=2.0164, lr=1.78e-05]
Training:  81%|████████▏ | 9014/11082 [28:59:16<6:40:41, 11.63s/step, acc=0.5074, loss=2.0164, lr=1.78e-05]
Training:  81%|████████▏ | 9015/11082 [28:59:28<6:39:01, 11.58s/step, acc=0.5074, loss=2.0164, lr=1.78e-05]
Training:  81%|████████▏ | 9016/11082 [28:59:40<6:40:42, 11.64s/step, acc=0.5074, loss=2.0164, lr=1.78e-05]
Training:  81%|████████▏ | 9017/11082 [28:59:51<6:39:55, 11.62s/step, acc=0.5074, loss=2.0164, lr=1.78e-05]
Training:  81%|████████▏ | 9018/11082 [29:00:03<6:41:31, 11.67s/step, acc=0.5074, loss=2.0164, lr=1.78e-05]
Training:  81%|████████▏ | 9019/11082 [29:00:14<6:37:18, 11.56s/step, acc=0.5074, loss=2.0164, lr=1.78e-05]
Training:  81%|████████▏ | 9020/11082 [29:00:26<6:36:59, 11.55s/step, acc=0.5074, loss=2.0164, lr=1.78e-05]
Training:  81%|████████▏ | 9021/11082 [29:00:38<6:39:21, 11.63s/step, acc=0.5074, loss=2.0164, lr=1.78e-05]
Training:  81%|████████▏ | 9022/11082 [29:00:48<6:31:38, 11.41s/step, acc=0.5074, loss=2.0164, lr=1.78e-05]
Training:  81%|████████▏ | 9023/11082 [29:01:00<6:37:48, 11.59s/step, acc=0.5074, loss=2.0164, lr=1.78e-05]
Training:  81%|████████▏ | 9024/11082 [29:01:12<6:37:50, 11.60s/step, acc=0.5074, loss=2.0164, lr=1.78e-05]
Training:  81%|████████▏ | 9025/11082 [29:01:24<6:36:06, 11.55s/step, acc=0.5074, loss=2.0164, lr=1.78e-05]
Training:  81%|████████▏ | 9026/11082 [29:01:35<6:37:27, 11.60s/step, acc=0.5074, loss=2.0164, lr=1.78e-05]
Training:  81%|████████▏ | 9027/11082 [29:01:47<6:40:03, 11.68s/step, acc=0.5074, loss=2.0164, lr=1.78e-05]
Training:  81%|████████▏ | 9028/11082 [29:01:59<6:39:24, 11.67s/step, acc=0.5074, loss=2.0164, lr=1.78e-05]
Training:  81%|████████▏ | 9029/11082 [29:02:10<6:37:58, 11.63s/step, acc=0.5074, loss=2.0164, lr=1.78e-05]
Training:  81%|████████▏ | 9030/11082 [29:02:22<6:34:20, 11.53s/step, acc=0.5074, loss=2.0164, lr=1.78e-05]
Training:  81%|████████▏ | 9031/11082 [29:02:33<6:37:02, 11.62s/step, acc=0.5074, loss=2.0164, lr=1.78e-05]
Training:  82%|████████▏ | 9032/11082 [29:02:45<6:36:29, 11.60s/step, acc=0.5074, loss=2.0164, lr=1.78e-05]
Training:  82%|████████▏ | 9033/11082 [29:02:57<6:35:56, 11.59s/step, acc=0.5074, loss=2.0164, lr=1.78e-05]
Training:  82%|████████▏ | 9034/11082 [29:03:08<6:32:55, 11.51s/step, acc=0.5074, loss=2.0164, lr=1.78e-05]
Training:  82%|████████▏ | 9035/11082 [29:03:19<6:32:55, 11.52s/step, acc=0.5074, loss=2.0164, lr=1.78e-05]
Training:  82%|████████▏ | 9036/11082 [29:03:31<6:33:47, 11.55s/step, acc=0.5074, loss=2.0164, lr=1.78e-05]
Training:  82%|████████▏ | 9037/11082 [29:03:42<6:30:49, 11.47s/step, acc=0.5074, loss=2.0164, lr=1.78e-05]
Training:  82%|████████▏ | 9038/11082 [29:03:54<6:34:49, 11.59s/step, acc=0.5074, loss=2.0164, lr=1.78e-05]
Training:  82%|████████▏ | 9039/11082 [29:04:06<6:34:12, 11.58s/step, acc=0.5074, loss=2.0164, lr=1.78e-05]
Training:  82%|████████▏ | 9040/11082 [29:04:17<6:32:52, 11.54s/step, acc=0.5074, loss=2.0164, lr=1.78e-05]
Training:  82%|████████▏ | 9041/11082 [29:04:29<6:35:22, 11.62s/step, acc=0.5074, loss=2.0164, lr=1.78e-05]
Training:  82%|████████▏ | 9042/11082 [29:04:40<6:30:59, 11.50s/step, acc=0.5074, loss=2.0164, lr=1.78e-05]
Training:  82%|████████▏ | 9043/11082 [29:04:52<6:32:33, 11.55s/step, acc=0.5074, loss=2.0164, lr=1.78e-05]
Training:  82%|████████▏ | 9044/11082 [29:05:03<6:31:11, 11.52s/step, acc=0.5074, loss=2.0164, lr=1.78e-05]
Training:  82%|████████▏ | 9045/11082 [29:05:15<6:35:42, 11.66s/step, acc=0.5074, loss=2.0164, lr=1.78e-05]
Training:  82%|████████▏ | 9046/11082 [29:05:26<6:29:15, 11.47s/step, acc=0.5074, loss=2.0164, lr=1.78e-05]
Training:  82%|████████▏ | 9047/11082 [29:05:38<6:33:34, 11.60s/step, acc=0.5074, loss=2.0164, lr=1.78e-05]
Training:  82%|████████▏ | 9048/11082 [29:05:49<6:24:45, 11.35s/step, acc=0.5074, loss=2.0164, lr=1.78e-05]
Training:  82%|████████▏ | 9049/11082 [29:06:01<6:30:57, 11.54s/step, acc=0.5074, loss=2.0164, lr=1.78e-05]
Training:  82%|████████▏ | 9050/11082 [29:06:13<6:30:51, 11.54s/step, acc=0.5074, loss=2.0164, lr=1.78e-05]
Training:  82%|████████▏ | 9050/11082 [29:06:20<6:30:51, 11.54s/step, acc=0.5065, loss=2.0221, lr=1.75e-05]2026-06-19 19:59:48 | INFO | __main__ | epoch=1 step=9050 train_loss=2.022101 train_acc=0.506529 lr=1.748e-05 grad_norm=0.5000 acc_d0=0.7618 acc_d1=0.6190 acc_d2=0.5000 acc_d3=0.4023 acc_d4=0.3245 acc_d5=0.2616 acc_d6=0.2143 acc_d7=0.1758

Training:  82%|████████▏ | 9051/11082 [29:06:24<6:30:57, 11.55s/step, acc=0.5065, loss=2.0221, lr=1.75e-05]
Training:  82%|████████▏ | 9052/11082 [29:06:36<6:31:50, 11.58s/step, acc=0.5065, loss=2.0221, lr=1.75e-05]
Training:  82%|████████▏ | 9053/11082 [29:06:47<6:27:13, 11.45s/step, acc=0.5065, loss=2.0221, lr=1.75e-05]
Training:  82%|████████▏ | 9054/11082 [29:06:58<6:25:00, 11.39s/step, acc=0.5065, loss=2.0221, lr=1.75e-05]
Training:  82%|████████▏ | 9055/11082 [29:07:10<6:29:53, 11.54s/step, acc=0.5065, loss=2.0221, lr=1.75e-05]
Training:  82%|████████▏ | 9056/11082 [29:07:22<6:31:52, 11.61s/step, acc=0.5065, loss=2.0221, lr=1.75e-05]
Training:  82%|████████▏ | 9057/11082 [29:07:33<6:27:14, 11.47s/step, acc=0.5065, loss=2.0221, lr=1.75e-05]
Training:  82%|████████▏ | 9058/11082 [29:07:45<6:28:39, 11.52s/step, acc=0.5065, loss=2.0221, lr=1.75e-05]
Training:  82%|████████▏ | 9059/11082 [29:07:56<6:26:57, 11.48s/step, acc=0.5065, loss=2.0221, lr=1.75e-05]
Training:  82%|████████▏ | 9060/11082 [29:08:08<6:30:55, 11.60s/step, acc=0.5065, loss=2.0221, lr=1.75e-05]
Training:  82%|████████▏ | 9061/11082 [29:08:20<6:31:56, 11.64s/step, acc=0.5065, loss=2.0221, lr=1.75e-05]
Training:  82%|████████▏ | 9062/11082 [29:08:31<6:31:24, 11.63s/step, acc=0.5065, loss=2.0221, lr=1.75e-05]
Training:  82%|████████▏ | 9063/11082 [29:08:43<6:30:20, 11.60s/step, acc=0.5065, loss=2.0221, lr=1.75e-05]
Training:  82%|████████▏ | 9064/11082 [29:08:54<6:31:09, 11.63s/step, acc=0.5065, loss=2.0221, lr=1.75e-05]
Training:  82%|████████▏ | 9065/11082 [29:09:06<6:33:15, 11.70s/step, acc=0.5065, loss=2.0221, lr=1.75e-05]
Training:  82%|████████▏ | 9066/11082 [29:09:17<6:27:31, 11.53s/step, acc=0.5065, loss=2.0221, lr=1.75e-05]
Training:  82%|████████▏ | 9067/11082 [29:09:29<6:28:32, 11.57s/step, acc=0.5065, loss=2.0221, lr=1.75e-05]
Training:  82%|████████▏ | 9068/11082 [29:09:40<6:26:33, 11.52s/step, acc=0.5065, loss=2.0221, lr=1.75e-05]
Training:  82%|████████▏ | 9069/11082 [29:09:52<6:21:41, 11.38s/step, acc=0.5065, loss=2.0221, lr=1.75e-05]
Training:  82%|████████▏ | 9070/11082 [29:10:04<6:28:53, 11.60s/step, acc=0.5065, loss=2.0221, lr=1.75e-05]
Training:  82%|████████▏ | 9071/11082 [29:10:15<6:25:59, 11.52s/step, acc=0.5065, loss=2.0221, lr=1.75e-05]
Training:  82%|████████▏ | 9072/11082 [29:10:27<6:31:13, 11.68s/step, acc=0.5065, loss=2.0221, lr=1.75e-05]
Training:  82%|████████▏ | 9073/11082 [29:10:38<6:23:55, 11.47s/step, acc=0.5065, loss=2.0221, lr=1.75e-05]
Training:  82%|████████▏ | 9074/11082 [29:10:49<6:23:12, 11.45s/step, acc=0.5065, loss=2.0221, lr=1.75e-05]
Training:  82%|████████▏ | 9075/11082 [29:11:01<6:26:11, 11.55s/step, acc=0.5065, loss=2.0221, lr=1.75e-05]
Training:  82%|████████▏ | 9076/11082 [29:11:13<6:26:34, 11.56s/step, acc=0.5065, loss=2.0221, lr=1.75e-05]
Training:  82%|████████▏ | 9077/11082 [29:11:24<6:25:16, 11.53s/step, acc=0.5065, loss=2.0221, lr=1.75e-05]
Training:  82%|████████▏ | 9078/11082 [29:11:36<6:23:43, 11.49s/step, acc=0.5065, loss=2.0221, lr=1.75e-05]
Training:  82%|████████▏ | 9079/11082 [29:11:48<6:30:10, 11.69s/step, acc=0.5065, loss=2.0221, lr=1.75e-05]
Training:  82%|████████▏ | 9080/11082 [29:11:59<6:29:31, 11.67s/step, acc=0.5065, loss=2.0221, lr=1.75e-05]
Training:  82%|████████▏ | 9081/11082 [29:12:11<6:30:52, 11.72s/step, acc=0.5065, loss=2.0221, lr=1.75e-05]
Training:  82%|████████▏ | 9082/11082 [29:12:23<6:28:28, 11.65s/step, acc=0.5065, loss=2.0221, lr=1.75e-05]
Training:  82%|████████▏ | 9083/11082 [29:12:34<6:19:39, 11.40s/step, acc=0.5065, loss=2.0221, lr=1.75e-05]
Training:  82%|████████▏ | 9084/11082 [29:12:45<6:19:02, 11.38s/step, acc=0.5065, loss=2.0221, lr=1.75e-05]
Training:  82%|████████▏ | 9085/11082 [29:12:57<6:21:24, 11.46s/step, acc=0.5065, loss=2.0221, lr=1.75e-05]
Training:  82%|████████▏ | 9086/11082 [29:13:09<6:26:33, 11.62s/step, acc=0.5065, loss=2.0221, lr=1.75e-05]
Training:  82%|████████▏ | 9087/11082 [29:13:20<6:28:38, 11.69s/step, acc=0.5065, loss=2.0221, lr=1.75e-05]
Training:  82%|████████▏ | 9088/11082 [29:13:32<6:25:41, 11.61s/step, acc=0.5065, loss=2.0221, lr=1.75e-05]
Training:  82%|████████▏ | 9089/11082 [29:13:43<6:23:07, 11.53s/step, acc=0.5065, loss=2.0221, lr=1.75e-05]
Training:  82%|████████▏ | 9090/11082 [29:13:55<6:22:27, 11.52s/step, acc=0.5065, loss=2.0221, lr=1.75e-05]
Training:  82%|████████▏ | 9091/11082 [29:14:06<6:24:45, 11.59s/step, acc=0.5065, loss=2.0221, lr=1.75e-05]
Training:  82%|████████▏ | 9092/11082 [29:14:18<6:21:11, 11.49s/step, acc=0.5065, loss=2.0221, lr=1.75e-05]
Training:  82%|████████▏ | 9093/11082 [29:14:29<6:18:12, 11.41s/step, acc=0.5065, loss=2.0221, lr=1.75e-05]
Training:  82%|████████▏ | 9094/11082 [29:14:41<6:21:30, 11.51s/step, acc=0.5065, loss=2.0221, lr=1.75e-05]
Training:  82%|████████▏ | 9095/11082 [29:14:52<6:21:18, 11.51s/step, acc=0.5065, loss=2.0221, lr=1.75e-05]
Training:  82%|████████▏ | 9096/11082 [29:15:04<6:24:48, 11.63s/step, acc=0.5065, loss=2.0221, lr=1.75e-05]
Training:  82%|████████▏ | 9097/11082 [29:15:16<6:27:01, 11.70s/step, acc=0.5065, loss=2.0221, lr=1.75e-05]
Training:  82%|████████▏ | 9098/11082 [29:15:27<6:21:51, 11.55s/step, acc=0.5065, loss=2.0221, lr=1.75e-05]
Training:  82%|████████▏ | 9099/11082 [29:15:38<6:20:11, 11.50s/step, acc=0.5065, loss=2.0221, lr=1.75e-05]
Training:  82%|████████▏ | 9100/11082 [29:15:50<6:22:51, 11.59s/step, acc=0.5065, loss=2.0221, lr=1.75e-05]
Training:  82%|████████▏ | 9100/11082 [29:15:57<6:22:51, 11.59s/step, acc=0.5060, loss=2.0237, lr=1.71e-05]2026-06-19 20:09:25 | INFO | __main__ | epoch=1 step=9100 train_loss=2.023732 train_acc=0.505993 lr=1.712e-05 grad_norm=0.4883 acc_d0=0.7674 acc_d1=0.6250 acc_d2=0.5049 acc_d3=0.4081 acc_d4=0.3292 acc_d5=0.2671 acc_d6=0.2154 acc_d7=0.1767

Training:  82%|████████▏ | 9101/11082 [29:16:02<6:23:04, 11.60s/step, acc=0.5060, loss=2.0237, lr=1.71e-05]
Training:  82%|████████▏ | 9102/11082 [29:16:13<6:22:05, 11.58s/step, acc=0.5060, loss=2.0237, lr=1.71e-05]
Training:  82%|████████▏ | 9103/11082 [29:16:25<6:19:44, 11.51s/step, acc=0.5060, loss=2.0237, lr=1.71e-05]
Training:  82%|████████▏ | 9104/11082 [29:16:36<6:20:03, 11.53s/step, acc=0.5060, loss=2.0237, lr=1.71e-05]
Training:  82%|████████▏ | 9105/11082 [29:16:48<6:21:44, 11.59s/step, acc=0.5060, loss=2.0237, lr=1.71e-05]
Training:  82%|████████▏ | 9106/11082 [29:17:00<6:20:31, 11.55s/step, acc=0.5060, loss=2.0237, lr=1.71e-05]
Training:  82%|████████▏ | 9107/11082 [29:17:11<6:19:50, 11.54s/step, acc=0.5060, loss=2.0237, lr=1.71e-05]
Training:  82%|████████▏ | 9108/11082 [29:17:23<6:21:16, 11.59s/step, acc=0.5060, loss=2.0237, lr=1.71e-05]
Training:  82%|████████▏ | 9109/11082 [29:17:35<6:27:40, 11.79s/step, acc=0.5060, loss=2.0237, lr=1.71e-05]
Training:  82%|████████▏ | 9110/11082 [29:17:47<6:26:51, 11.77s/step, acc=0.5060, loss=2.0237, lr=1.71e-05]
Training:  82%|████████▏ | 9111/11082 [29:17:58<6:21:28, 11.61s/step, acc=0.5060, loss=2.0237, lr=1.71e-05]
Training:  82%|████████▏ | 9112/11082 [29:18:10<6:24:29, 11.71s/step, acc=0.5060, loss=2.0237, lr=1.71e-05]
Training:  82%|████████▏ | 9113/11082 [29:18:21<6:22:39, 11.66s/step, acc=0.5060, loss=2.0237, lr=1.71e-05]
Training:  82%|████████▏ | 9114/11082 [29:18:33<6:17:01, 11.49s/step, acc=0.5060, loss=2.0237, lr=1.71e-05]
Training:  82%|████████▏ | 9115/11082 [29:18:45<6:22:14, 11.66s/step, acc=0.5060, loss=2.0237, lr=1.71e-05]
Training:  82%|████████▏ | 9116/11082 [29:18:56<6:15:46, 11.47s/step, acc=0.5060, loss=2.0237, lr=1.71e-05]
Training:  82%|████████▏ | 9117/11082 [29:19:08<6:19:53, 11.60s/step, acc=0.5060, loss=2.0237, lr=1.71e-05]
Training:  82%|████████▏ | 9118/11082 [29:19:19<6:19:47, 11.60s/step, acc=0.5060, loss=2.0237, lr=1.71e-05]
Training:  82%|████████▏ | 9119/11082 [29:19:31<6:18:26, 11.57s/step, acc=0.5060, loss=2.0237, lr=1.71e-05]
Training:  82%|████████▏ | 9120/11082 [29:19:42<6:18:45, 11.58s/step, acc=0.5060, loss=2.0237, lr=1.71e-05]
Training:  82%|████████▏ | 9121/11082 [29:19:53<6:13:02, 11.41s/step, acc=0.5060, loss=2.0237, lr=1.71e-05]
Training:  82%|████████▏ | 9122/11082 [29:20:05<6:18:21, 11.58s/step, acc=0.5060, loss=2.0237, lr=1.71e-05]
Training:  82%|████████▏ | 9123/11082 [29:20:17<6:19:06, 11.61s/step, acc=0.5060, loss=2.0237, lr=1.71e-05]
Training:  82%|████████▏ | 9124/11082 [29:20:28<6:14:57, 11.49s/step, acc=0.5060, loss=2.0237, lr=1.71e-05]
Training:  82%|████████▏ | 9125/11082 [29:20:40<6:13:40, 11.46s/step, acc=0.5060, loss=2.0237, lr=1.71e-05]
Training:  82%|████████▏ | 9126/11082 [29:20:51<6:17:02, 11.57s/step, acc=0.5060, loss=2.0237, lr=1.71e-05]
Training:  82%|████████▏ | 9127/11082 [29:21:03<6:15:01, 11.51s/step, acc=0.5060, loss=2.0237, lr=1.71e-05]
Training:  82%|████████▏ | 9128/11082 [29:21:14<6:13:26, 11.47s/step, acc=0.5060, loss=2.0237, lr=1.71e-05]
Training:  82%|████████▏ | 9129/11082 [29:21:26<6:19:23, 11.66s/step, acc=0.5060, loss=2.0237, lr=1.71e-05]
Training:  82%|████████▏ | 9130/11082 [29:21:38<6:16:02, 11.56s/step, acc=0.5060, loss=2.0237, lr=1.71e-05]
Training:  82%|████████▏ | 9131/11082 [29:21:49<6:16:15, 11.57s/step, acc=0.5060, loss=2.0237, lr=1.71e-05]
Training:  82%|████████▏ | 9132/11082 [29:22:01<6:21:21, 11.73s/step, acc=0.5060, loss=2.0237, lr=1.71e-05]
Training:  82%|████████▏ | 9133/11082 [29:22:13<6:17:40, 11.63s/step, acc=0.5060, loss=2.0237, lr=1.71e-05]
Training:  82%|████████▏ | 9134/11082 [29:22:24<6:16:09, 11.59s/step, acc=0.5060, loss=2.0237, lr=1.71e-05]
Training:  82%|████████▏ | 9135/11082 [29:22:36<6:20:46, 11.73s/step, acc=0.5060, loss=2.0237, lr=1.71e-05]
Training:  82%|████████▏ | 9136/11082 [29:22:48<6:19:03, 11.69s/step, acc=0.5060, loss=2.0237, lr=1.71e-05]
Training:  82%|████████▏ | 9137/11082 [29:22:59<6:16:17, 11.61s/step, acc=0.5060, loss=2.0237, lr=1.71e-05]
Training:  82%|████████▏ | 9138/11082 [29:23:11<6:14:38, 11.56s/step, acc=0.5060, loss=2.0237, lr=1.71e-05]
Training:  82%|████████▏ | 9139/11082 [29:23:23<6:17:50, 11.67s/step, acc=0.5060, loss=2.0237, lr=1.71e-05]
Training:  82%|████████▏ | 9140/11082 [29:23:34<6:14:50, 11.58s/step, acc=0.5060, loss=2.0237, lr=1.71e-05]
Training:  82%|████████▏ | 9141/11082 [29:23:46<6:17:08, 11.66s/step, acc=0.5060, loss=2.0237, lr=1.71e-05]
Training:  82%|████████▏ | 9142/11082 [29:23:58<6:20:41, 11.77s/step, acc=0.5060, loss=2.0237, lr=1.71e-05]
Training:  83%|████████▎ | 9143/11082 [29:24:09<6:17:47, 11.69s/step, acc=0.5060, loss=2.0237, lr=1.71e-05]
Training:  83%|████████▎ | 9144/11082 [29:24:21<6:18:00, 11.70s/step, acc=0.5060, loss=2.0237, lr=1.71e-05]
Training:  83%|████████▎ | 9145/11082 [29:24:33<6:16:20, 11.66s/step, acc=0.5060, loss=2.0237, lr=1.71e-05]
Training:  83%|████████▎ | 9146/11082 [29:24:44<6:12:40, 11.55s/step, acc=0.5060, loss=2.0237, lr=1.71e-05]
Training:  83%|████████▎ | 9147/11082 [29:24:56<6:13:33, 11.58s/step, acc=0.5060, loss=2.0237, lr=1.71e-05]
Training:  83%|████████▎ | 9148/11082 [29:25:07<6:10:36, 11.50s/step, acc=0.5060, loss=2.0237, lr=1.71e-05]
Training:  83%|████████▎ | 9149/11082 [29:25:19<6:13:25, 11.59s/step, acc=0.5060, loss=2.0237, lr=1.71e-05]
Training:  83%|████████▎ | 9150/11082 [29:25:30<6:13:21, 11.59s/step, acc=0.5060, loss=2.0237, lr=1.71e-05]
Training:  83%|████████▎ | 9150/11082 [29:25:37<6:13:21, 11.59s/step, acc=0.5070, loss=2.0119, lr=1.68e-05]2026-06-19 20:19:05 | INFO | __main__ | epoch=1 step=9150 train_loss=2.011914 train_acc=0.506980 lr=1.678e-05 grad_norm=0.5156 acc_d0=0.7705 acc_d1=0.6288 acc_d2=0.5100 acc_d3=0.4127 acc_d4=0.3346 acc_d5=0.2706 acc_d6=0.2205 acc_d7=0.1803

Training:  83%|████████▎ | 9151/11082 [29:25:42<6:11:38, 11.55s/step, acc=0.5070, loss=2.0119, lr=1.68e-05]
Training:  83%|████████▎ | 9152/11082 [29:25:53<6:10:49, 11.53s/step, acc=0.5070, loss=2.0119, lr=1.68e-05]
Training:  83%|████████▎ | 9153/11082 [29:26:05<6:14:14, 11.64s/step, acc=0.5070, loss=2.0119, lr=1.68e-05]
Training:  83%|████████▎ | 9154/11082 [29:26:17<6:13:22, 11.62s/step, acc=0.5070, loss=2.0119, lr=1.68e-05]
Training:  83%|████████▎ | 9155/11082 [29:26:28<6:13:28, 11.63s/step, acc=0.5070, loss=2.0119, lr=1.68e-05]
Training:  83%|████████▎ | 9156/11082 [29:26:40<6:10:48, 11.55s/step, acc=0.5070, loss=2.0119, lr=1.68e-05]
Training:  83%|████████▎ | 9157/11082 [29:26:51<6:08:37, 11.49s/step, acc=0.5070, loss=2.0119, lr=1.68e-05]
Training:  83%|████████▎ | 9158/11082 [29:27:03<6:10:50, 11.56s/step, acc=0.5070, loss=2.0119, lr=1.68e-05]
Training:  83%|████████▎ | 9159/11082 [29:27:14<6:09:15, 11.52s/step, acc=0.5070, loss=2.0119, lr=1.68e-05]
Training:  83%|████████▎ | 9160/11082 [29:27:26<6:11:44, 11.61s/step, acc=0.5070, loss=2.0119, lr=1.68e-05]
Training:  83%|████████▎ | 9161/11082 [29:27:38<6:11:02, 11.59s/step, acc=0.5070, loss=2.0119, lr=1.68e-05]
Training:  83%|████████▎ | 9162/11082 [29:27:50<6:14:56, 11.72s/step, acc=0.5070, loss=2.0119, lr=1.68e-05]
Training:  83%|████████▎ | 9163/11082 [29:28:01<6:10:56, 11.60s/step, acc=0.5070, loss=2.0119, lr=1.68e-05]
Training:  83%|████████▎ | 9164/11082 [29:28:13<6:12:20, 11.65s/step, acc=0.5070, loss=2.0119, lr=1.68e-05]
Training:  83%|████████▎ | 9165/11082 [29:28:24<6:11:45, 11.64s/step, acc=0.5070, loss=2.0119, lr=1.68e-05]
Training:  83%|████████▎ | 9166/11082 [29:28:36<6:08:23, 11.54s/step, acc=0.5070, loss=2.0119, lr=1.68e-05]
Training:  83%|████████▎ | 9167/11082 [29:28:47<6:10:07, 11.60s/step, acc=0.5070, loss=2.0119, lr=1.68e-05]
Training:  83%|████████▎ | 9168/11082 [29:28:59<6:12:28, 11.68s/step, acc=0.5070, loss=2.0119, lr=1.68e-05]
Training:  83%|████████▎ | 9169/11082 [29:29:10<6:07:27, 11.53s/step, acc=0.5070, loss=2.0119, lr=1.68e-05]
Training:  83%|████████▎ | 9170/11082 [29:29:22<6:06:10, 11.49s/step, acc=0.5070, loss=2.0119, lr=1.68e-05]
Training:  83%|████████▎ | 9171/11082 [29:29:34<6:08:33, 11.57s/step, acc=0.5070, loss=2.0119, lr=1.68e-05]
Training:  83%|████████▎ | 9172/11082 [29:29:45<6:08:49, 11.59s/step, acc=0.5070, loss=2.0119, lr=1.68e-05]
Training:  83%|████████▎ | 9173/11082 [29:29:57<6:10:08, 11.63s/step, acc=0.5070, loss=2.0119, lr=1.68e-05]
Training:  83%|████████▎ | 9174/11082 [29:30:08<6:06:51, 11.54s/step, acc=0.5070, loss=2.0119, lr=1.68e-05]
Training:  83%|████████▎ | 9175/11082 [29:30:20<6:06:34, 11.53s/step, acc=0.5070, loss=2.0119, lr=1.68e-05]
Training:  83%|████████▎ | 9176/11082 [29:30:31<6:08:33, 11.60s/step, acc=0.5070, loss=2.0119, lr=1.68e-05]
Training:  83%|████████▎ | 9177/11082 [29:30:43<6:05:21, 11.51s/step, acc=0.5070, loss=2.0119, lr=1.68e-05]
Training:  83%|████████▎ | 9178/11082 [29:30:54<6:04:01, 11.47s/step, acc=0.5070, loss=2.0119, lr=1.68e-05]
Training:  83%|████████▎ | 9179/11082 [29:31:06<6:05:39, 11.53s/step, acc=0.5070, loss=2.0119, lr=1.68e-05]
Training:  83%|████████▎ | 9180/11082 [29:31:18<6:11:49, 11.73s/step, acc=0.5070, loss=2.0119, lr=1.68e-05]
Training:  83%|████████▎ | 9181/11082 [29:31:29<6:04:59, 11.52s/step, acc=0.5070, loss=2.0119, lr=1.68e-05]
Training:  83%|████████▎ | 9182/11082 [29:31:40<6:01:13, 11.41s/step, acc=0.5070, loss=2.0119, lr=1.68e-05]
Training:  83%|████████▎ | 9183/11082 [29:31:52<6:07:53, 11.62s/step, acc=0.5070, loss=2.0119, lr=1.68e-05]
Training:  83%|████████▎ | 9184/11082 [29:32:04<6:03:54, 11.50s/step, acc=0.5070, loss=2.0119, lr=1.68e-05]
Training:  83%|████████▎ | 9185/11082 [29:32:15<6:04:25, 11.53s/step, acc=0.5070, loss=2.0119, lr=1.68e-05]
Training:  83%|████████▎ | 9186/11082 [29:32:27<6:04:31, 11.54s/step, acc=0.5070, loss=2.0119, lr=1.68e-05]
Training:  83%|████████▎ | 9187/11082 [29:32:38<6:01:54, 11.46s/step, acc=0.5070, loss=2.0119, lr=1.68e-05]
Training:  83%|████████▎ | 9188/11082 [29:32:49<5:58:20, 11.35s/step, acc=0.5070, loss=2.0119, lr=1.68e-05]
Training:  83%|████████▎ | 9189/11082 [29:33:01<6:04:50, 11.56s/step, acc=0.5070, loss=2.0119, lr=1.68e-05]
Training:  83%|████████▎ | 9190/11082 [29:33:12<6:02:08, 11.48s/step, acc=0.5070, loss=2.0119, lr=1.68e-05]
Training:  83%|████████▎ | 9191/11082 [29:33:24<6:07:05, 11.65s/step, acc=0.5070, loss=2.0119, lr=1.68e-05]
Training:  83%|████████▎ | 9192/11082 [29:33:36<6:03:07, 11.53s/step, acc=0.5070, loss=2.0119, lr=1.68e-05]
Training:  83%|████████▎ | 9193/11082 [29:33:47<6:02:29, 11.51s/step, acc=0.5070, loss=2.0119, lr=1.68e-05]
Training:  83%|████████▎ | 9194/11082 [29:33:59<6:03:12, 11.54s/step, acc=0.5070, loss=2.0119, lr=1.68e-05]
Training:  83%|████████▎ | 9195/11082 [29:34:10<6:04:40, 11.60s/step, acc=0.5070, loss=2.0119, lr=1.68e-05]
Training:  83%|████████▎ | 9196/11082 [29:34:22<6:03:17, 11.56s/step, acc=0.5070, loss=2.0119, lr=1.68e-05]
Training:  83%|████████▎ | 9197/11082 [29:34:33<6:00:00, 11.46s/step, acc=0.5070, loss=2.0119, lr=1.68e-05]
Training:  83%|████████▎ | 9198/11082 [29:34:45<6:00:59, 11.50s/step, acc=0.5070, loss=2.0119, lr=1.68e-05]
Training:  83%|████████▎ | 9199/11082 [29:34:56<6:01:59, 11.53s/step, acc=0.5070, loss=2.0119, lr=1.68e-05]
Training:  83%|████████▎ | 9200/11082 [29:35:08<5:59:49, 11.47s/step, acc=0.5070, loss=2.0119, lr=1.68e-05]
Training:  83%|████████▎ | 9200/11082 [29:35:15<5:59:49, 11.47s/step, acc=0.5076, loss=2.0137, lr=1.64e-05]2026-06-19 20:28:43 | INFO | __main__ | epoch=1 step=9200 train_loss=2.013710 train_acc=0.507556 lr=1.644e-05 grad_norm=0.5195 acc_d0=0.7649 acc_d1=0.6248 acc_d2=0.5063 acc_d3=0.4107 acc_d4=0.3319 acc_d5=0.2699 acc_d6=0.2200 acc_d7=0.1800

Training:  83%|████████▎ | 9201/11082 [29:35:19<5:57:37, 11.41s/step, acc=0.5076, loss=2.0137, lr=1.64e-05]
Training:  83%|████████▎ | 9202/11082 [29:35:31<6:00:16, 11.50s/step, acc=0.5076, loss=2.0137, lr=1.64e-05]
Training:  83%|████████▎ | 9203/11082 [29:35:42<5:59:41, 11.49s/step, acc=0.5076, loss=2.0137, lr=1.64e-05]
Training:  83%|████████▎ | 9204/11082 [29:35:54<6:00:26, 11.52s/step, acc=0.5076, loss=2.0137, lr=1.64e-05]
Training:  83%|████████▎ | 9205/11082 [29:36:05<6:01:55, 11.57s/step, acc=0.5076, loss=2.0137, lr=1.64e-05]
Training:  83%|████████▎ | 9206/11082 [29:36:17<6:02:54, 11.61s/step, acc=0.5076, loss=2.0137, lr=1.64e-05]
Training:  83%|████████▎ | 9207/11082 [29:36:28<6:00:15, 11.53s/step, acc=0.5076, loss=2.0137, lr=1.64e-05]
Training:  83%|████████▎ | 9208/11082 [29:36:40<6:02:11, 11.60s/step, acc=0.5076, loss=2.0137, lr=1.64e-05]
Training:  83%|████████▎ | 9209/11082 [29:36:52<6:04:23, 11.67s/step, acc=0.5076, loss=2.0137, lr=1.64e-05]
Training:  83%|████████▎ | 9210/11082 [29:37:04<6:04:26, 11.68s/step, acc=0.5076, loss=2.0137, lr=1.64e-05]
Training:  83%|████████▎ | 9211/11082 [29:37:15<6:03:59, 11.67s/step, acc=0.5076, loss=2.0137, lr=1.64e-05]
Training:  83%|████████▎ | 9212/11082 [29:37:27<6:02:41, 11.64s/step, acc=0.5076, loss=2.0137, lr=1.64e-05]
Training:  83%|████████▎ | 9213/11082 [29:37:39<6:05:38, 11.74s/step, acc=0.5076, loss=2.0137, lr=1.64e-05]
Training:  83%|████████▎ | 9214/11082 [29:37:50<6:02:03, 11.63s/step, acc=0.5076, loss=2.0137, lr=1.64e-05]
Training:  83%|████████▎ | 9215/11082 [29:38:02<6:01:56, 11.63s/step, acc=0.5076, loss=2.0137, lr=1.64e-05]
Training:  83%|████████▎ | 9216/11082 [29:38:13<5:57:37, 11.50s/step, acc=0.5076, loss=2.0137, lr=1.64e-05]
Training:  83%|████████▎ | 9217/11082 [29:38:25<6:02:38, 11.67s/step, acc=0.5076, loss=2.0137, lr=1.64e-05]
Training:  83%|████████▎ | 9218/11082 [29:38:37<6:01:10, 11.63s/step, acc=0.5076, loss=2.0137, lr=1.64e-05]
Training:  83%|████████▎ | 9219/11082 [29:38:48<5:59:31, 11.58s/step, acc=0.5076, loss=2.0137, lr=1.64e-05]
Training:  83%|████████▎ | 9220/11082 [29:39:00<5:58:21, 11.55s/step, acc=0.5076, loss=2.0137, lr=1.64e-05]
Training:  83%|████████▎ | 9221/11082 [29:39:12<6:00:39, 11.63s/step, acc=0.5076, loss=2.0137, lr=1.64e-05]
Training:  83%|████████▎ | 9222/11082 [29:39:23<6:00:52, 11.64s/step, acc=0.5076, loss=2.0137, lr=1.64e-05]
Training:  83%|████████▎ | 9223/11082 [29:39:35<6:01:13, 11.66s/step, acc=0.5076, loss=2.0137, lr=1.64e-05]
Training:  83%|████████▎ | 9224/11082 [29:39:46<5:59:32, 11.61s/step, acc=0.5076, loss=2.0137, lr=1.64e-05]
Training:  83%|████████▎ | 9225/11082 [29:39:58<6:00:09, 11.64s/step, acc=0.5076, loss=2.0137, lr=1.64e-05]
Training:  83%|████████▎ | 9226/11082 [29:40:09<5:55:40, 11.50s/step, acc=0.5076, loss=2.0137, lr=1.64e-05]
Training:  83%|████████▎ | 9227/11082 [29:40:21<5:56:29, 11.53s/step, acc=0.5076, loss=2.0137, lr=1.64e-05]
Training:  83%|████████▎ | 9228/11082 [29:40:32<5:56:29, 11.54s/step, acc=0.5076, loss=2.0137, lr=1.64e-05]
Training:  83%|████████▎ | 9229/11082 [29:40:44<5:56:28, 11.54s/step, acc=0.5076, loss=2.0137, lr=1.64e-05]
Training:  83%|████████▎ | 9230/11082 [29:40:55<5:55:09, 11.51s/step, acc=0.5076, loss=2.0137, lr=1.64e-05]
Training:  83%|████████▎ | 9231/11082 [29:41:07<5:59:56, 11.67s/step, acc=0.5076, loss=2.0137, lr=1.64e-05]
Training:  83%|████████▎ | 9232/11082 [29:41:19<5:57:11, 11.58s/step, acc=0.5076, loss=2.0137, lr=1.64e-05]
Training:  83%|████████▎ | 9233/11082 [29:41:30<5:55:44, 11.54s/step, acc=0.5076, loss=2.0137, lr=1.64e-05]
Training:  83%|████████▎ | 9234/11082 [29:41:42<5:54:18, 11.50s/step, acc=0.5076, loss=2.0137, lr=1.64e-05]
Training:  83%|████████▎ | 9235/11082 [29:41:54<5:58:02, 11.63s/step, acc=0.5076, loss=2.0137, lr=1.64e-05]
Training:  83%|████████▎ | 9236/11082 [29:42:05<5:58:51, 11.66s/step, acc=0.5076, loss=2.0137, lr=1.64e-05]
Training:  83%|████████▎ | 9237/11082 [29:42:16<5:51:16, 11.42s/step, acc=0.5076, loss=2.0137, lr=1.64e-05]
Training:  83%|████████▎ | 9238/11082 [29:42:28<5:55:20, 11.56s/step, acc=0.5076, loss=2.0137, lr=1.64e-05]
Training:  83%|████████▎ | 9239/11082 [29:42:40<5:56:21, 11.60s/step, acc=0.5076, loss=2.0137, lr=1.64e-05]
Training:  83%|████████▎ | 9240/11082 [29:42:51<5:55:18, 11.57s/step, acc=0.5076, loss=2.0137, lr=1.64e-05]
Training:  83%|████████▎ | 9241/11082 [29:43:03<5:53:36, 11.52s/step, acc=0.5076, loss=2.0137, lr=1.64e-05]
Training:  83%|████████▎ | 9242/11082 [29:43:14<5:53:06, 11.51s/step, acc=0.5076, loss=2.0137, lr=1.64e-05]
Training:  83%|████████▎ | 9243/11082 [29:43:26<5:53:57, 11.55s/step, acc=0.5076, loss=2.0137, lr=1.64e-05]
Training:  83%|████████▎ | 9244/11082 [29:43:37<5:54:18, 11.57s/step, acc=0.5076, loss=2.0137, lr=1.64e-05]
Training:  83%|████████▎ | 9245/11082 [29:43:49<5:52:48, 11.52s/step, acc=0.5076, loss=2.0137, lr=1.64e-05]
Training:  83%|████████▎ | 9246/11082 [29:44:00<5:51:01, 11.47s/step, acc=0.5076, loss=2.0137, lr=1.64e-05]
Training:  83%|████████▎ | 9247/11082 [29:44:12<5:53:24, 11.56s/step, acc=0.5076, loss=2.0137, lr=1.64e-05]
Training:  83%|████████▎ | 9248/11082 [29:44:23<5:52:36, 11.54s/step, acc=0.5076, loss=2.0137, lr=1.64e-05]
Training:  83%|████████▎ | 9249/11082 [29:44:35<5:52:41, 11.54s/step, acc=0.5076, loss=2.0137, lr=1.64e-05]
Training:  83%|████████▎ | 9250/11082 [29:44:47<5:52:18, 11.54s/step, acc=0.5076, loss=2.0137, lr=1.64e-05]
Training:  83%|████████▎ | 9250/11082 [29:44:54<5:52:18, 11.54s/step, acc=0.5084, loss=2.0108, lr=1.61e-05]2026-06-19 20:38:22 | INFO | __main__ | epoch=1 step=9250 train_loss=2.010809 train_acc=0.508374 lr=1.611e-05 grad_norm=0.5352 acc_d0=0.7738 acc_d1=0.6355 acc_d2=0.5181 acc_d3=0.4208 acc_d4=0.3415 acc_d5=0.2784 acc_d6=0.2267 acc_d7=0.1851

Training:  83%|████████▎ | 9251/11082 [29:44:58<5:53:56, 11.60s/step, acc=0.5084, loss=2.0108, lr=1.61e-05]
Training:  83%|████████▎ | 9252/11082 [29:45:10<5:54:46, 11.63s/step, acc=0.5084, loss=2.0108, lr=1.61e-05]
Training:  83%|████████▎ | 9253/11082 [29:45:21<5:51:22, 11.53s/step, acc=0.5084, loss=2.0108, lr=1.61e-05]
Training:  84%|████████▎ | 9254/11082 [29:45:33<5:56:48, 11.71s/step, acc=0.5084, loss=2.0108, lr=1.61e-05]
Training:  84%|████████▎ | 9255/11082 [29:45:45<5:53:11, 11.60s/step, acc=0.5084, loss=2.0108, lr=1.61e-05]
Training:  84%|████████▎ | 9256/11082 [29:45:56<5:52:10, 11.57s/step, acc=0.5084, loss=2.0108, lr=1.61e-05]
Training:  84%|████████▎ | 9257/11082 [29:46:08<5:52:40, 11.59s/step, acc=0.5084, loss=2.0108, lr=1.61e-05]
Training:  84%|████████▎ | 9258/11082 [29:46:19<5:49:01, 11.48s/step, acc=0.5084, loss=2.0108, lr=1.61e-05]
Training:  84%|████████▎ | 9259/11082 [29:46:31<5:49:39, 11.51s/step, acc=0.5084, loss=2.0108, lr=1.61e-05]
Training:  84%|████████▎ | 9260/11082 [29:46:42<5:52:09, 11.60s/step, acc=0.5084, loss=2.0108, lr=1.61e-05]
Training:  84%|████████▎ | 9261/11082 [29:46:54<5:50:56, 11.56s/step, acc=0.5084, loss=2.0108, lr=1.61e-05]
Training:  84%|████████▎ | 9262/11082 [29:47:06<5:52:47, 11.63s/step, acc=0.5084, loss=2.0108, lr=1.61e-05]
Training:  84%|████████▎ | 9263/11082 [29:47:17<5:47:33, 11.46s/step, acc=0.5084, loss=2.0108, lr=1.61e-05]
Training:  84%|████████▎ | 9264/11082 [29:47:28<5:46:22, 11.43s/step, acc=0.5084, loss=2.0108, lr=1.61e-05]
Training:  84%|████████▎ | 9265/11082 [29:47:40<5:48:11, 11.50s/step, acc=0.5084, loss=2.0108, lr=1.61e-05]
Training:  84%|████████▎ | 9266/11082 [29:47:51<5:46:38, 11.45s/step, acc=0.5084, loss=2.0108, lr=1.61e-05]
Training:  84%|████████▎ | 9267/11082 [29:48:03<5:51:35, 11.62s/step, acc=0.5084, loss=2.0108, lr=1.61e-05]
Training:  84%|████████▎ | 9268/11082 [29:48:15<5:49:29, 11.56s/step, acc=0.5084, loss=2.0108, lr=1.61e-05]
Training:  84%|████████▎ | 9269/11082 [29:48:26<5:48:38, 11.54s/step, acc=0.5084, loss=2.0108, lr=1.61e-05]
Training:  84%|████████▎ | 9270/11082 [29:48:37<5:44:55, 11.42s/step, acc=0.5084, loss=2.0108, lr=1.61e-05]
Training:  84%|████████▎ | 9271/11082 [29:48:49<5:49:08, 11.57s/step, acc=0.5084, loss=2.0108, lr=1.61e-05]
Training:  84%|████████▎ | 9272/11082 [29:49:01<5:49:30, 11.59s/step, acc=0.5084, loss=2.0108, lr=1.61e-05]
Training:  84%|████████▎ | 9273/11082 [29:49:12<5:45:31, 11.46s/step, acc=0.5084, loss=2.0108, lr=1.61e-05]
Training:  84%|████████▎ | 9274/11082 [29:49:24<5:50:40, 11.64s/step, acc=0.5084, loss=2.0108, lr=1.61e-05]
Training:  84%|████████▎ | 9275/11082 [29:49:36<5:53:50, 11.75s/step, acc=0.5084, loss=2.0108, lr=1.61e-05]
Training:  84%|████████▎ | 9276/11082 [29:49:47<5:49:15, 11.60s/step, acc=0.5084, loss=2.0108, lr=1.61e-05]
Training:  84%|████████▎ | 9277/11082 [29:49:59<5:45:44, 11.49s/step, acc=0.5084, loss=2.0108, lr=1.61e-05]
Training:  84%|████████▎ | 9278/11082 [29:50:10<5:48:51, 11.60s/step, acc=0.5084, loss=2.0108, lr=1.61e-05]
Training:  84%|████████▎ | 9279/11082 [29:50:22<5:48:29, 11.60s/step, acc=0.5084, loss=2.0108, lr=1.61e-05]
Training:  84%|████████▎ | 9280/11082 [29:50:34<5:47:56, 11.59s/step, acc=0.5084, loss=2.0108, lr=1.61e-05]
Training:  84%|████████▎ | 9281/11082 [29:50:45<5:45:24, 11.51s/step, acc=0.5084, loss=2.0108, lr=1.61e-05]
Training:  84%|████████▍ | 9282/11082 [29:50:57<5:48:10, 11.61s/step, acc=0.5084, loss=2.0108, lr=1.61e-05]
Training:  84%|████████▍ | 9283/11082 [29:51:08<5:47:54, 11.60s/step, acc=0.5084, loss=2.0108, lr=1.61e-05]
Training:  84%|████████▍ | 9284/11082 [29:51:20<5:47:35, 11.60s/step, acc=0.5084, loss=2.0108, lr=1.61e-05]
Training:  84%|████████▍ | 9285/11082 [29:51:32<5:48:17, 11.63s/step, acc=0.5084, loss=2.0108, lr=1.61e-05]
Training:  84%|████████▍ | 9286/11082 [29:51:43<5:45:42, 11.55s/step, acc=0.5084, loss=2.0108, lr=1.61e-05]
Training:  84%|████████▍ | 9287/11082 [29:51:55<5:47:14, 11.61s/step, acc=0.5084, loss=2.0108, lr=1.61e-05]
Training:  84%|████████▍ | 9288/11082 [29:52:06<5:43:27, 11.49s/step, acc=0.5084, loss=2.0108, lr=1.61e-05]
Training:  84%|████████▍ | 9289/11082 [29:52:17<5:43:40, 11.50s/step, acc=0.5084, loss=2.0108, lr=1.61e-05]
Training:  84%|████████▍ | 9290/11082 [29:52:29<5:42:20, 11.46s/step, acc=0.5084, loss=2.0108, lr=1.61e-05]
Training:  84%|████████▍ | 9291/11082 [29:52:41<5:48:16, 11.67s/step, acc=0.5084, loss=2.0108, lr=1.61e-05]
Training:  84%|████████▍ | 9292/11082 [29:52:52<5:43:11, 11.50s/step, acc=0.5084, loss=2.0108, lr=1.61e-05]
Training:  84%|████████▍ | 9293/11082 [29:53:04<5:44:29, 11.55s/step, acc=0.5084, loss=2.0108, lr=1.61e-05]
Training:  84%|████████▍ | 9294/11082 [29:53:15<5:44:10, 11.55s/step, acc=0.5084, loss=2.0108, lr=1.61e-05]
Training:  84%|████████▍ | 9295/11082 [29:53:27<5:46:44, 11.64s/step, acc=0.5084, loss=2.0108, lr=1.61e-05]
Training:  84%|████████▍ | 9296/11082 [29:53:38<5:42:38, 11.51s/step, acc=0.5084, loss=2.0108, lr=1.61e-05]
Training:  84%|████████▍ | 9297/11082 [29:53:50<5:42:23, 11.51s/step, acc=0.5084, loss=2.0108, lr=1.61e-05]
Training:  84%|████████▍ | 9298/11082 [29:54:02<5:43:58, 11.57s/step, acc=0.5084, loss=2.0108, lr=1.61e-05]
Training:  84%|████████▍ | 9299/11082 [29:54:13<5:43:17, 11.55s/step, acc=0.5084, loss=2.0108, lr=1.61e-05]
Training:  84%|████████▍ | 9300/11082 [29:54:24<5:40:56, 11.48s/step, acc=0.5084, loss=2.0108, lr=1.61e-05]
Training:  84%|████████▍ | 9300/11082 [29:54:32<5:40:56, 11.48s/step, acc=0.5056, loss=2.0277, lr=1.58e-05]2026-06-19 20:47:59 | INFO | __main__ | epoch=1 step=9300 train_loss=2.027659 train_acc=0.505634 lr=1.579e-05 grad_norm=0.5156 acc_d0=0.7627 acc_d1=0.6204 acc_d2=0.4997 acc_d3=0.4028 acc_d4=0.3263 acc_d5=0.2630 acc_d6=0.2129 acc_d7=0.1740

Training:  84%|████████▍ | 9301/11082 [29:54:36<5:42:23, 11.53s/step, acc=0.5056, loss=2.0277, lr=1.58e-05]
Training:  84%|████████▍ | 9302/11082 [29:54:47<5:39:52, 11.46s/step, acc=0.5056, loss=2.0277, lr=1.58e-05]
Training:  84%|████████▍ | 9303/11082 [29:54:59<5:44:43, 11.63s/step, acc=0.5056, loss=2.0277, lr=1.58e-05]
Training:  84%|████████▍ | 9304/11082 [29:55:11<5:42:08, 11.55s/step, acc=0.5056, loss=2.0277, lr=1.58e-05]
Training:  84%|████████▍ | 9305/11082 [29:55:23<5:44:47, 11.64s/step, acc=0.5056, loss=2.0277, lr=1.58e-05]
Training:  84%|████████▍ | 9306/11082 [29:55:34<5:38:42, 11.44s/step, acc=0.5056, loss=2.0277, lr=1.58e-05]
Training:  84%|████████▍ | 9307/11082 [29:55:46<5:43:40, 11.62s/step, acc=0.5056, loss=2.0277, lr=1.58e-05]
Training:  84%|████████▍ | 9308/11082 [29:55:57<5:42:12, 11.57s/step, acc=0.5056, loss=2.0277, lr=1.58e-05]
Training:  84%|████████▍ | 9309/11082 [29:56:09<5:41:35, 11.56s/step, acc=0.5056, loss=2.0277, lr=1.58e-05]
Training:  84%|████████▍ | 9310/11082 [29:56:20<5:39:28, 11.49s/step, acc=0.5056, loss=2.0277, lr=1.58e-05]
Training:  84%|████████▍ | 9311/11082 [29:56:31<5:39:14, 11.49s/step, acc=0.5056, loss=2.0277, lr=1.58e-05]
Training:  84%|████████▍ | 9312/11082 [29:56:43<5:43:41, 11.65s/step, acc=0.5056, loss=2.0277, lr=1.58e-05]
Training:  84%|████████▍ | 9313/11082 [29:56:55<5:38:59, 11.50s/step, acc=0.5056, loss=2.0277, lr=1.58e-05]
Training:  84%|████████▍ | 9314/11082 [29:57:06<5:37:22, 11.45s/step, acc=0.5056, loss=2.0277, lr=1.58e-05]
Training:  84%|████████▍ | 9315/11082 [29:57:18<5:41:05, 11.58s/step, acc=0.5056, loss=2.0277, lr=1.58e-05]
Training:  84%|████████▍ | 9316/11082 [29:57:29<5:40:12, 11.56s/step, acc=0.5056, loss=2.0277, lr=1.58e-05]
Training:  84%|████████▍ | 9317/11082 [29:57:41<5:40:42, 11.58s/step, acc=0.5056, loss=2.0277, lr=1.58e-05]
Training:  84%|████████▍ | 9318/11082 [29:57:52<5:39:53, 11.56s/step, acc=0.5056, loss=2.0277, lr=1.58e-05]
Training:  84%|████████▍ | 9319/11082 [29:58:04<5:40:23, 11.58s/step, acc=0.5056, loss=2.0277, lr=1.58e-05]
Training:  84%|████████▍ | 9320/11082 [29:58:16<5:42:41, 11.67s/step, acc=0.5056, loss=2.0277, lr=1.58e-05]
Training:  84%|████████▍ | 9321/11082 [29:58:27<5:38:49, 11.54s/step, acc=0.5056, loss=2.0277, lr=1.58e-05]
Training:  84%|████████▍ | 9322/11082 [29:58:39<5:43:00, 11.69s/step, acc=0.5056, loss=2.0277, lr=1.58e-05]
Training:  84%|████████▍ | 9323/11082 [29:58:51<5:40:53, 11.63s/step, acc=0.5056, loss=2.0277, lr=1.58e-05]
Training:  84%|████████▍ | 9324/11082 [29:59:02<5:39:35, 11.59s/step, acc=0.5056, loss=2.0277, lr=1.58e-05]
Training:  84%|████████▍ | 9325/11082 [29:59:14<5:39:29, 11.59s/step, acc=0.5056, loss=2.0277, lr=1.58e-05]
Training:  84%|████████▍ | 9326/11082 [29:59:25<5:39:29, 11.60s/step, acc=0.5056, loss=2.0277, lr=1.58e-05]
Training:  84%|████████▍ | 9327/11082 [29:59:37<5:38:12, 11.56s/step, acc=0.5056, loss=2.0277, lr=1.58e-05]
Training:  84%|████████▍ | 9328/11082 [29:59:49<5:38:56, 11.59s/step, acc=0.5056, loss=2.0277, lr=1.58e-05]
Training:  84%|████████▍ | 9329/11082 [30:00:00<5:37:21, 11.55s/step, acc=0.5056, loss=2.0277, lr=1.58e-05]
Training:  84%|████████▍ | 9330/11082 [30:00:11<5:35:50, 11.50s/step, acc=0.5056, loss=2.0277, lr=1.58e-05]
Training:  84%|████████▍ | 9331/11082 [30:00:23<5:35:34, 11.50s/step, acc=0.5056, loss=2.0277, lr=1.58e-05]
Training:  84%|████████▍ | 9332/11082 [30:00:34<5:33:49, 11.45s/step, acc=0.5056, loss=2.0277, lr=1.58e-05]
Training:  84%|████████▍ | 9333/11082 [30:00:46<5:37:50, 11.59s/step, acc=0.5056, loss=2.0277, lr=1.58e-05]
Training:  84%|████████▍ | 9334/11082 [30:00:57<5:33:40, 11.45s/step, acc=0.5056, loss=2.0277, lr=1.58e-05]
Training:  84%|████████▍ | 9335/11082 [30:01:09<5:36:52, 11.57s/step, acc=0.5056, loss=2.0277, lr=1.58e-05]
Training:  84%|████████▍ | 9336/11082 [30:01:21<5:35:07, 11.52s/step, acc=0.5056, loss=2.0277, lr=1.58e-05]
Training:  84%|████████▍ | 9337/11082 [30:01:32<5:36:35, 11.57s/step, acc=0.5056, loss=2.0277, lr=1.58e-05]
Training:  84%|████████▍ | 9338/11082 [30:01:43<5:32:42, 11.45s/step, acc=0.5056, loss=2.0277, lr=1.58e-05]
Training:  84%|████████▍ | 9339/11082 [30:01:55<5:37:16, 11.61s/step, acc=0.5056, loss=2.0277, lr=1.58e-05]
Training:  84%|████████▍ | 9340/11082 [30:02:07<5:34:38, 11.53s/step, acc=0.5056, loss=2.0277, lr=1.58e-05]
Training:  84%|████████▍ | 9341/11082 [30:02:18<5:32:56, 11.47s/step, acc=0.5056, loss=2.0277, lr=1.58e-05]
Training:  84%|████████▍ | 9342/11082 [30:02:29<5:32:20, 11.46s/step, acc=0.5056, loss=2.0277, lr=1.58e-05]
Training:  84%|████████▍ | 9343/11082 [30:02:41<5:36:42, 11.62s/step, acc=0.5056, loss=2.0277, lr=1.58e-05]
Training:  84%|████████▍ | 9344/11082 [30:02:53<5:34:17, 11.54s/step, acc=0.5056, loss=2.0277, lr=1.58e-05]
Training:  84%|████████▍ | 9345/11082 [30:03:05<5:38:19, 11.69s/step, acc=0.5056, loss=2.0277, lr=1.58e-05]
Training:  84%|████████▍ | 9346/11082 [30:03:16<5:34:48, 11.57s/step, acc=0.5056, loss=2.0277, lr=1.58e-05]
Training:  84%|████████▍ | 9347/11082 [30:03:27<5:30:02, 11.41s/step, acc=0.5056, loss=2.0277, lr=1.58e-05]
Training:  84%|████████▍ | 9348/11082 [30:03:39<5:32:55, 11.52s/step, acc=0.5056, loss=2.0277, lr=1.58e-05]
Training:  84%|████████▍ | 9349/11082 [30:03:51<5:37:50, 11.70s/step, acc=0.5056, loss=2.0277, lr=1.58e-05]
Training:  84%|████████▍ | 9350/11082 [30:04:02<5:33:30, 11.55s/step, acc=0.5056, loss=2.0277, lr=1.58e-05]
Training:  84%|████████▍ | 9350/11082 [30:04:09<5:33:30, 11.55s/step, acc=0.5083, loss=2.0107, lr=1.55e-05]2026-06-19 20:57:37 | INFO | __main__ | epoch=1 step=9350 train_loss=2.010670 train_acc=0.508290 lr=1.548e-05 grad_norm=0.5430 acc_d0=0.7575 acc_d1=0.6147 acc_d2=0.4945 acc_d3=0.3989 acc_d4=0.3217 acc_d5=0.2595 acc_d6=0.2093 acc_d7=0.1740

Training:  84%|████████▍ | 9351/11082 [30:04:14<5:33:48, 11.57s/step, acc=0.5083, loss=2.0107, lr=1.55e-05]
Training:  84%|████████▍ | 9352/11082 [30:04:25<5:31:56, 11.51s/step, acc=0.5083, loss=2.0107, lr=1.55e-05]
Training:  84%|████████▍ | 9353/11082 [30:04:37<5:32:45, 11.55s/step, acc=0.5083, loss=2.0107, lr=1.55e-05]
Training:  84%|████████▍ | 9354/11082 [30:04:49<5:34:26, 11.61s/step, acc=0.5083, loss=2.0107, lr=1.55e-05]
Training:  84%|████████▍ | 9355/11082 [30:05:00<5:35:17, 11.65s/step, acc=0.5083, loss=2.0107, lr=1.55e-05]
Training:  84%|████████▍ | 9356/11082 [30:05:13<5:40:53, 11.85s/step, acc=0.5083, loss=2.0107, lr=1.55e-05]
Training:  84%|████████▍ | 9357/11082 [30:05:24<5:31:57, 11.55s/step, acc=0.5083, loss=2.0107, lr=1.55e-05]
Training:  84%|████████▍ | 9358/11082 [30:05:35<5:33:52, 11.62s/step, acc=0.5083, loss=2.0107, lr=1.55e-05]
Training:  84%|████████▍ | 9359/11082 [30:05:47<5:34:59, 11.67s/step, acc=0.5083, loss=2.0107, lr=1.55e-05]
Training:  84%|████████▍ | 9360/11082 [30:05:59<5:32:55, 11.60s/step, acc=0.5083, loss=2.0107, lr=1.55e-05]
Training:  84%|████████▍ | 9361/11082 [30:06:10<5:35:12, 11.69s/step, acc=0.5083, loss=2.0107, lr=1.55e-05]
Training:  84%|████████▍ | 9362/11082 [30:06:22<5:30:21, 11.52s/step, acc=0.5083, loss=2.0107, lr=1.55e-05]
Training:  84%|████████▍ | 9363/11082 [30:06:33<5:30:05, 11.52s/step, acc=0.5083, loss=2.0107, lr=1.55e-05]
Training:  84%|████████▍ | 9364/11082 [30:06:45<5:29:51, 11.52s/step, acc=0.5083, loss=2.0107, lr=1.55e-05]
Training:  85%|████████▍ | 9365/11082 [30:06:56<5:32:22, 11.61s/step, acc=0.5083, loss=2.0107, lr=1.55e-05]
Training:  85%|████████▍ | 9366/11082 [30:07:08<5:29:47, 11.53s/step, acc=0.5083, loss=2.0107, lr=1.55e-05]
Training:  85%|████████▍ | 9367/11082 [30:07:19<5:29:46, 11.54s/step, acc=0.5083, loss=2.0107, lr=1.55e-05]
Training:  85%|████████▍ | 9368/11082 [30:07:31<5:26:33, 11.43s/step, acc=0.5083, loss=2.0107, lr=1.55e-05]
Training:  85%|████████▍ | 9369/11082 [30:07:43<5:31:42, 11.62s/step, acc=0.5083, loss=2.0107, lr=1.55e-05]
Training:  85%|████████▍ | 9370/11082 [30:07:54<5:30:13, 11.57s/step, acc=0.5083, loss=2.0107, lr=1.55e-05]
Training:  85%|████████▍ | 9371/11082 [30:08:06<5:32:04, 11.65s/step, acc=0.5083, loss=2.0107, lr=1.55e-05]
Training:  85%|████████▍ | 9372/11082 [30:08:17<5:29:55, 11.58s/step, acc=0.5083, loss=2.0107, lr=1.55e-05]
Training:  85%|████████▍ | 9373/11082 [30:08:29<5:32:17, 11.67s/step, acc=0.5083, loss=2.0107, lr=1.55e-05]
Training:  85%|████████▍ | 9374/11082 [30:08:41<5:31:43, 11.65s/step, acc=0.5083, loss=2.0107, lr=1.55e-05]
Training:  85%|████████▍ | 9375/11082 [30:08:52<5:26:34, 11.48s/step, acc=0.5083, loss=2.0107, lr=1.55e-05]
Training:  85%|████████▍ | 9376/11082 [30:09:04<5:30:19, 11.62s/step, acc=0.5083, loss=2.0107, lr=1.55e-05]
Training:  85%|████████▍ | 9377/11082 [30:09:15<5:30:17, 11.62s/step, acc=0.5083, loss=2.0107, lr=1.55e-05]
Training:  85%|████████▍ | 9378/11082 [30:09:27<5:29:52, 11.62s/step, acc=0.5083, loss=2.0107, lr=1.55e-05]
Training:  85%|████████▍ | 9379/11082 [30:09:39<5:29:36, 11.61s/step, acc=0.5083, loss=2.0107, lr=1.55e-05]
Training:  85%|████████▍ | 9380/11082 [30:09:50<5:30:19, 11.64s/step, acc=0.5083, loss=2.0107, lr=1.55e-05]
Training:  85%|████████▍ | 9381/11082 [30:10:02<5:27:46, 11.56s/step, acc=0.5083, loss=2.0107, lr=1.55e-05]
Training:  85%|████████▍ | 9382/11082 [30:10:14<5:29:24, 11.63s/step, acc=0.5083, loss=2.0107, lr=1.55e-05]
Training:  85%|████████▍ | 9383/11082 [30:10:25<5:24:38, 11.46s/step, acc=0.5083, loss=2.0107, lr=1.55e-05]
Training:  85%|████████▍ | 9384/11082 [30:10:36<5:26:43, 11.55s/step, acc=0.5083, loss=2.0107, lr=1.55e-05]
Training:  85%|████████▍ | 9385/11082 [30:10:48<5:27:58, 11.60s/step, acc=0.5083, loss=2.0107, lr=1.55e-05]
Training:  85%|████████▍ | 9386/11082 [30:10:59<5:23:31, 11.45s/step, acc=0.5083, loss=2.0107, lr=1.55e-05]
Training:  85%|████████▍ | 9387/11082 [30:11:11<5:28:05, 11.61s/step, acc=0.5083, loss=2.0107, lr=1.55e-05]
Training:  85%|████████▍ | 9388/11082 [30:11:23<5:27:30, 11.60s/step, acc=0.5083, loss=2.0107, lr=1.55e-05]
Training:  85%|████████▍ | 9389/11082 [30:11:34<5:24:28, 11.50s/step, acc=0.5083, loss=2.0107, lr=1.55e-05]
Training:  85%|████████▍ | 9390/11082 [30:11:45<5:20:18, 11.36s/step, acc=0.5083, loss=2.0107, lr=1.55e-05]
Training:  85%|████████▍ | 9391/11082 [30:11:57<5:24:54, 11.53s/step, acc=0.5083, loss=2.0107, lr=1.55e-05]
Training:  85%|████████▍ | 9392/11082 [30:12:09<5:28:27, 11.66s/step, acc=0.5083, loss=2.0107, lr=1.55e-05]
Training:  85%|████████▍ | 9393/11082 [30:12:21<5:28:31, 11.67s/step, acc=0.5083, loss=2.0107, lr=1.55e-05]
Training:  85%|████████▍ | 9394/11082 [30:12:32<5:22:24, 11.46s/step, acc=0.5083, loss=2.0107, lr=1.55e-05]
Training:  85%|████████▍ | 9395/11082 [30:12:44<5:26:54, 11.63s/step, acc=0.5083, loss=2.0107, lr=1.55e-05]
Training:  85%|████████▍ | 9396/11082 [30:12:55<5:23:17, 11.50s/step, acc=0.5083, loss=2.0107, lr=1.55e-05]
Training:  85%|████████▍ | 9397/11082 [30:13:07<5:25:03, 11.58s/step, acc=0.5083, loss=2.0107, lr=1.55e-05]
Training:  85%|████████▍ | 9398/11082 [30:13:18<5:24:13, 11.55s/step, acc=0.5083, loss=2.0107, lr=1.55e-05]
Training:  85%|████████▍ | 9399/11082 [30:13:29<5:20:46, 11.44s/step, acc=0.5083, loss=2.0107, lr=1.55e-05]
Training:  85%|████████▍ | 9400/11082 [30:13:41<5:22:57, 11.52s/step, acc=0.5083, loss=2.0107, lr=1.55e-05]
Training:  85%|████████▍ | 9400/11082 [30:13:48<5:22:57, 11.52s/step, acc=0.5061, loss=2.0205, lr=1.52e-05]2026-06-19 21:07:16 | INFO | __main__ | epoch=1 step=9400 train_loss=2.020518 train_acc=0.506095 lr=1.517e-05 grad_norm=0.5391 acc_d0=0.7747 acc_d1=0.6347 acc_d2=0.5149 acc_d3=0.4178 acc_d4=0.3390 acc_d5=0.2750 acc_d6=0.2227 acc_d7=0.1841

Training:  85%|████████▍ | 9401/11082 [30:13:53<5:23:26, 11.54s/step, acc=0.5061, loss=2.0205, lr=1.52e-05]
Training:  85%|████████▍ | 9402/11082 [30:14:04<5:19:16, 11.40s/step, acc=0.5061, loss=2.0205, lr=1.52e-05]
Training:  85%|████████▍ | 9403/11082 [30:14:15<5:21:53, 11.50s/step, acc=0.5061, loss=2.0205, lr=1.52e-05]
Training:  85%|████████▍ | 9404/11082 [30:14:27<5:26:26, 11.67s/step, acc=0.5061, loss=2.0205, lr=1.52e-05]
Training:  85%|████████▍ | 9405/11082 [30:14:39<5:22:53, 11.55s/step, acc=0.5061, loss=2.0205, lr=1.52e-05]
Training:  85%|████████▍ | 9406/11082 [30:14:50<5:24:00, 11.60s/step, acc=0.5061, loss=2.0205, lr=1.52e-05]
Training:  85%|████████▍ | 9407/11082 [30:15:02<5:22:40, 11.56s/step, acc=0.5061, loss=2.0205, lr=1.52e-05]
Training:  85%|████████▍ | 9408/11082 [30:15:13<5:18:26, 11.41s/step, acc=0.5061, loss=2.0205, lr=1.52e-05]
Training:  85%|████████▍ | 9409/11082 [30:15:25<5:23:14, 11.59s/step, acc=0.5061, loss=2.0205, lr=1.52e-05]
Training:  85%|████████▍ | 9410/11082 [30:15:36<5:21:13, 11.53s/step, acc=0.5061, loss=2.0205, lr=1.52e-05]
Training:  85%|████████▍ | 9411/11082 [30:15:48<5:22:38, 11.59s/step, acc=0.5061, loss=2.0205, lr=1.52e-05]
Training:  85%|████████▍ | 9412/11082 [30:15:59<5:21:00, 11.53s/step, acc=0.5061, loss=2.0205, lr=1.52e-05]
Training:  85%|████████▍ | 9413/11082 [30:16:11<5:21:09, 11.55s/step, acc=0.5061, loss=2.0205, lr=1.52e-05]
Training:  85%|████████▍ | 9414/11082 [30:16:22<5:20:07, 11.52s/step, acc=0.5061, loss=2.0205, lr=1.52e-05]
Training:  85%|████████▍ | 9415/11082 [30:16:33<5:15:23, 11.35s/step, acc=0.5061, loss=2.0205, lr=1.52e-05]
Training:  85%|████████▍ | 9416/11082 [30:16:45<5:17:57, 11.45s/step, acc=0.5061, loss=2.0205, lr=1.52e-05]
Training:  85%|████████▍ | 9417/11082 [30:16:57<5:20:37, 11.55s/step, acc=0.5061, loss=2.0205, lr=1.52e-05]
Training:  85%|████████▍ | 9418/11082 [30:17:08<5:19:59, 11.54s/step, acc=0.5061, loss=2.0205, lr=1.52e-05]
Training:  85%|████████▍ | 9419/11082 [30:17:20<5:19:42, 11.53s/step, acc=0.5061, loss=2.0205, lr=1.52e-05]
Training:  85%|████████▌ | 9420/11082 [30:17:32<5:20:23, 11.57s/step, acc=0.5061, loss=2.0205, lr=1.52e-05]
Training:  85%|████████▌ | 9421/11082 [30:17:43<5:19:38, 11.55s/step, acc=0.5061, loss=2.0205, lr=1.52e-05]
Training:  85%|████████▌ | 9422/11082 [30:17:55<5:18:48, 11.52s/step, acc=0.5061, loss=2.0205, lr=1.52e-05]
Training:  85%|████████▌ | 9423/11082 [30:18:06<5:14:58, 11.39s/step, acc=0.5061, loss=2.0205, lr=1.52e-05]
Training:  85%|████████▌ | 9424/11082 [30:18:18<5:20:59, 11.62s/step, acc=0.5061, loss=2.0205, lr=1.52e-05]
Training:  85%|████████▌ | 9425/11082 [30:18:29<5:19:22, 11.56s/step, acc=0.5061, loss=2.0205, lr=1.52e-05]
Training:  85%|████████▌ | 9426/11082 [30:18:40<5:16:59, 11.49s/step, acc=0.5061, loss=2.0205, lr=1.52e-05]
Training:  85%|████████▌ | 9427/11082 [30:18:52<5:14:40, 11.41s/step, acc=0.5061, loss=2.0205, lr=1.52e-05]
Training:  85%|████████▌ | 9428/11082 [30:19:04<5:20:02, 11.61s/step, acc=0.5061, loss=2.0205, lr=1.52e-05]
Training:  85%|████████▌ | 9429/11082 [30:19:15<5:17:56, 11.54s/step, acc=0.5061, loss=2.0205, lr=1.52e-05]
Training:  85%|████████▌ | 9430/11082 [30:19:27<5:17:52, 11.55s/step, acc=0.5061, loss=2.0205, lr=1.52e-05]
Training:  85%|████████▌ | 9431/11082 [30:19:39<5:20:18, 11.64s/step, acc=0.5061, loss=2.0205, lr=1.52e-05]
Training:  85%|████████▌ | 9432/11082 [30:19:50<5:17:08, 11.53s/step, acc=0.5061, loss=2.0205, lr=1.52e-05]
Training:  85%|████████▌ | 9433/11082 [30:20:02<5:19:29, 11.63s/step, acc=0.5061, loss=2.0205, lr=1.52e-05]
Training:  85%|████████▌ | 9434/11082 [30:20:13<5:15:55, 11.50s/step, acc=0.5061, loss=2.0205, lr=1.52e-05]
Training:  85%|████████▌ | 9435/11082 [30:20:25<5:16:53, 11.54s/step, acc=0.5061, loss=2.0205, lr=1.52e-05]
Training:  85%|████████▌ | 9436/11082 [30:20:36<5:15:49, 11.51s/step, acc=0.5061, loss=2.0205, lr=1.52e-05]
Training:  85%|████████▌ | 9437/11082 [30:20:48<5:18:33, 11.62s/step, acc=0.5061, loss=2.0205, lr=1.52e-05]
Training:  85%|████████▌ | 9438/11082 [30:21:00<5:20:30, 11.70s/step, acc=0.5061, loss=2.0205, lr=1.52e-05]
Training:  85%|████████▌ | 9439/11082 [30:21:12<5:20:57, 11.72s/step, acc=0.5061, loss=2.0205, lr=1.52e-05]
Training:  85%|████████▌ | 9440/11082 [30:21:22<5:14:02, 11.48s/step, acc=0.5061, loss=2.0205, lr=1.52e-05]
Training:  85%|████████▌ | 9441/11082 [30:21:34<5:15:52, 11.55s/step, acc=0.5061, loss=2.0205, lr=1.52e-05]
Training:  85%|████████▌ | 9442/11082 [30:21:46<5:16:01, 11.56s/step, acc=0.5061, loss=2.0205, lr=1.52e-05]
Training:  85%|████████▌ | 9443/11082 [30:21:58<5:17:56, 11.64s/step, acc=0.5061, loss=2.0205, lr=1.52e-05]
Training:  85%|████████▌ | 9444/11082 [30:22:09<5:15:51, 11.57s/step, acc=0.5061, loss=2.0205, lr=1.52e-05]
Training:  85%|████████▌ | 9445/11082 [30:22:21<5:16:28, 11.60s/step, acc=0.5061, loss=2.0205, lr=1.52e-05]
Training:  85%|████████▌ | 9446/11082 [30:22:32<5:16:42, 11.62s/step, acc=0.5061, loss=2.0205, lr=1.52e-05]
Training:  85%|████████▌ | 9447/11082 [30:22:44<5:16:16, 11.61s/step, acc=0.5061, loss=2.0205, lr=1.52e-05]
Training:  85%|████████▌ | 9448/11082 [30:22:55<5:14:41, 11.56s/step, acc=0.5061, loss=2.0205, lr=1.52e-05]
Training:  85%|████████▌ | 9449/11082 [30:23:07<5:12:57, 11.50s/step, acc=0.5061, loss=2.0205, lr=1.52e-05]
Training:  85%|████████▌ | 9450/11082 [30:23:19<5:16:16, 11.63s/step, acc=0.5061, loss=2.0205, lr=1.52e-05]
Training:  85%|████████▌ | 9450/11082 [30:23:26<5:16:16, 11.63s/step, acc=0.5052, loss=2.0302, lr=1.49e-05]2026-06-19 21:16:53 | INFO | __main__ | epoch=1 step=9450 train_loss=2.030225 train_acc=0.505241 lr=1.487e-05 grad_norm=0.5469 acc_d0=0.7619 acc_d1=0.6215 acc_d2=0.5038 acc_d3=0.4067 acc_d4=0.3289 acc_d5=0.2656 acc_d6=0.2157 acc_d7=0.1787

Training:  85%|████████▌ | 9451/11082 [30:23:30<5:14:28, 11.57s/step, acc=0.5052, loss=2.0302, lr=1.49e-05]
Training:  85%|████████▌ | 9452/11082 [30:23:42<5:13:37, 11.54s/step, acc=0.5052, loss=2.0302, lr=1.49e-05]
Training:  85%|████████▌ | 9453/11082 [30:23:53<5:13:53, 11.56s/step, acc=0.5052, loss=2.0302, lr=1.49e-05]
Training:  85%|████████▌ | 9454/11082 [30:24:05<5:13:59, 11.57s/step, acc=0.5052, loss=2.0302, lr=1.49e-05]
Training:  85%|████████▌ | 9455/11082 [30:24:16<5:12:37, 11.53s/step, acc=0.5052, loss=2.0302, lr=1.49e-05]
Training:  85%|████████▌ | 9456/11082 [30:24:28<5:13:32, 11.57s/step, acc=0.5052, loss=2.0302, lr=1.49e-05]
Training:  85%|████████▌ | 9457/11082 [30:24:39<5:11:58, 11.52s/step, acc=0.5052, loss=2.0302, lr=1.49e-05]
Training:  85%|████████▌ | 9458/11082 [30:24:51<5:12:27, 11.54s/step, acc=0.5052, loss=2.0302, lr=1.49e-05]
Training:  85%|████████▌ | 9459/11082 [30:25:03<5:13:51, 11.60s/step, acc=0.5052, loss=2.0302, lr=1.49e-05]
Training:  85%|████████▌ | 9460/11082 [30:25:14<5:10:16, 11.48s/step, acc=0.5052, loss=2.0302, lr=1.49e-05]
Training:  85%|████████▌ | 9461/11082 [30:25:26<5:16:30, 11.72s/step, acc=0.5052, loss=2.0302, lr=1.49e-05]
Training:  85%|████████▌ | 9462/11082 [30:25:38<5:15:52, 11.70s/step, acc=0.5052, loss=2.0302, lr=1.49e-05]
Training:  85%|████████▌ | 9463/11082 [30:25:49<5:13:32, 11.62s/step, acc=0.5052, loss=2.0302, lr=1.49e-05]
Training:  85%|████████▌ | 9464/11082 [30:26:01<5:13:39, 11.63s/step, acc=0.5052, loss=2.0302, lr=1.49e-05]
Training:  85%|████████▌ | 9465/11082 [30:26:12<5:09:56, 11.50s/step, acc=0.5052, loss=2.0302, lr=1.49e-05]
Training:  85%|████████▌ | 9466/11082 [30:26:24<5:16:04, 11.74s/step, acc=0.5052, loss=2.0302, lr=1.49e-05]
Training:  85%|████████▌ | 9467/11082 [30:26:36<5:13:14, 11.64s/step, acc=0.5052, loss=2.0302, lr=1.49e-05]
Training:  85%|████████▌ | 9468/11082 [30:26:48<5:15:43, 11.74s/step, acc=0.5052, loss=2.0302, lr=1.49e-05]
Training:  85%|████████▌ | 9469/11082 [30:26:59<5:12:19, 11.62s/step, acc=0.5052, loss=2.0302, lr=1.49e-05]
Training:  85%|████████▌ | 9470/11082 [30:27:11<5:12:35, 11.63s/step, acc=0.5052, loss=2.0302, lr=1.49e-05]
Training:  85%|████████▌ | 9471/11082 [30:27:23<5:14:40, 11.72s/step, acc=0.5052, loss=2.0302, lr=1.49e-05]
Training:  85%|████████▌ | 9472/11082 [30:27:34<5:15:01, 11.74s/step, acc=0.5052, loss=2.0302, lr=1.49e-05]
Training:  85%|████████▌ | 9473/11082 [30:27:45<5:09:54, 11.56s/step, acc=0.5052, loss=2.0302, lr=1.49e-05]
Training:  85%|████████▌ | 9474/11082 [30:27:57<5:10:13, 11.58s/step, acc=0.5052, loss=2.0302, lr=1.49e-05]
Training:  85%|████████▌ | 9475/11082 [30:28:09<5:11:15, 11.62s/step, acc=0.5052, loss=2.0302, lr=1.49e-05]
Training:  86%|████████▌ | 9476/11082 [30:28:21<5:11:23, 11.63s/step, acc=0.5052, loss=2.0302, lr=1.49e-05]
Training:  86%|████████▌ | 9477/11082 [30:28:32<5:06:45, 11.47s/step, acc=0.5052, loss=2.0302, lr=1.49e-05]
Training:  86%|████████▌ | 9478/11082 [30:28:44<5:11:39, 11.66s/step, acc=0.5052, loss=2.0302, lr=1.49e-05]
Training:  86%|████████▌ | 9479/11082 [30:28:55<5:08:04, 11.53s/step, acc=0.5052, loss=2.0302, lr=1.49e-05]
Training:  86%|████████▌ | 9480/11082 [30:29:07<5:10:06, 11.61s/step, acc=0.5052, loss=2.0302, lr=1.49e-05]
Training:  86%|████████▌ | 9481/11082 [30:29:18<5:08:11, 11.55s/step, acc=0.5052, loss=2.0302, lr=1.49e-05]
Training:  86%|████████▌ | 9482/11082 [30:29:30<5:07:25, 11.53s/step, acc=0.5052, loss=2.0302, lr=1.49e-05]
Training:  86%|████████▌ | 9483/11082 [30:29:41<5:08:13, 11.57s/step, acc=0.5052, loss=2.0302, lr=1.49e-05]
Training:  86%|████████▌ | 9484/11082 [30:29:52<5:04:23, 11.43s/step, acc=0.5052, loss=2.0302, lr=1.49e-05]
Training:  86%|████████▌ | 9485/11082 [30:30:04<5:05:12, 11.47s/step, acc=0.5052, loss=2.0302, lr=1.49e-05]
Training:  86%|████████▌ | 9486/11082 [30:30:16<5:08:29, 11.60s/step, acc=0.5052, loss=2.0302, lr=1.49e-05]
Training:  86%|████████▌ | 9487/11082 [30:30:27<5:06:32, 11.53s/step, acc=0.5052, loss=2.0302, lr=1.49e-05]
Training:  86%|████████▌ | 9488/11082 [30:30:39<5:07:23, 11.57s/step, acc=0.5052, loss=2.0302, lr=1.49e-05]
Training:  86%|████████▌ | 9489/11082 [30:30:50<5:05:58, 11.52s/step, acc=0.5052, loss=2.0302, lr=1.49e-05]
Training:  86%|████████▌ | 9490/11082 [30:31:02<5:06:25, 11.55s/step, acc=0.5052, loss=2.0302, lr=1.49e-05]
Training:  86%|████████▌ | 9491/11082 [30:31:13<5:03:12, 11.43s/step, acc=0.5052, loss=2.0302, lr=1.49e-05]
Training:  86%|████████▌ | 9492/11082 [30:31:25<5:03:48, 11.46s/step, acc=0.5052, loss=2.0302, lr=1.49e-05]
Training:  86%|████████▌ | 9493/11082 [30:31:37<5:09:34, 11.69s/step, acc=0.5052, loss=2.0302, lr=1.49e-05]
Training:  86%|████████▌ | 9494/11082 [30:31:48<5:06:25, 11.58s/step, acc=0.5052, loss=2.0302, lr=1.49e-05]
Training:  86%|████████▌ | 9495/11082 [30:31:59<5:03:51, 11.49s/step, acc=0.5052, loss=2.0302, lr=1.49e-05]
Training:  86%|████████▌ | 9496/11082 [30:32:11<5:03:08, 11.47s/step, acc=0.5052, loss=2.0302, lr=1.49e-05]
Training:  86%|████████▌ | 9497/11082 [30:32:22<5:04:25, 11.52s/step, acc=0.5052, loss=2.0302, lr=1.49e-05]
Training:  86%|████████▌ | 9498/11082 [30:32:34<5:04:43, 11.54s/step, acc=0.5052, loss=2.0302, lr=1.49e-05]
Training:  86%|████████▌ | 9499/11082 [30:32:46<5:04:20, 11.54s/step, acc=0.5052, loss=2.0302, lr=1.49e-05]
Training:  86%|████████▌ | 9500/11082 [30:32:57<5:04:13, 11.54s/step, acc=0.5052, loss=2.0302, lr=1.49e-05]
Training:  86%|████████▌ | 9500/11082 [30:33:05<5:04:13, 11.54s/step, acc=0.5050, loss=2.0292, lr=1.46e-05]2026-06-19 21:26:32 | INFO | __main__ | epoch=1 step=9500 train_loss=2.029235 train_acc=0.505015 lr=1.458e-05 grad_norm=0.5391 acc_d0=0.7607 acc_d1=0.6173 acc_d2=0.4972 acc_d3=0.4007 acc_d4=0.3233 acc_d5=0.2602 acc_d6=0.2131 acc_d7=0.1731

Training:  86%|████████▌ | 9501/11082 [30:33:09<5:04:06, 11.54s/step, acc=0.5050, loss=2.0292, lr=1.46e-05]
Training:  86%|████████▌ | 9502/11082 [30:33:21<5:08:21, 11.71s/step, acc=0.5050, loss=2.0292, lr=1.46e-05]
Training:  86%|████████▌ | 9503/11082 [30:33:32<5:05:49, 11.62s/step, acc=0.5050, loss=2.0292, lr=1.46e-05]
Training:  86%|████████▌ | 9504/11082 [30:33:44<5:03:49, 11.55s/step, acc=0.5050, loss=2.0292, lr=1.46e-05]
Training:  86%|████████▌ | 9505/11082 [30:33:55<4:59:06, 11.38s/step, acc=0.5050, loss=2.0292, lr=1.46e-05]
Training:  86%|████████▌ | 9506/11082 [30:34:07<5:05:04, 11.61s/step, acc=0.5050, loss=2.0292, lr=1.46e-05]
Training:  86%|████████▌ | 9507/11082 [30:34:18<5:02:51, 11.54s/step, acc=0.5050, loss=2.0292, lr=1.46e-05]
Training:  86%|████████▌ | 9508/11082 [30:34:30<5:04:37, 11.61s/step, acc=0.5050, loss=2.0292, lr=1.46e-05]
Training:  86%|████████▌ | 9509/11082 [30:34:41<5:03:51, 11.59s/step, acc=0.5050, loss=2.0292, lr=1.46e-05]
Training:  86%|████████▌ | 9510/11082 [30:34:53<5:04:33, 11.62s/step, acc=0.5050, loss=2.0292, lr=1.46e-05]
Training:  86%|████████▌ | 9511/11082 [30:35:04<4:59:28, 11.44s/step, acc=0.5050, loss=2.0292, lr=1.46e-05]
Training:  86%|████████▌ | 9512/11082 [30:35:16<5:01:20, 11.52s/step, acc=0.5050, loss=2.0292, lr=1.46e-05]
Training:  86%|████████▌ | 9513/11082 [30:35:28<5:03:05, 11.59s/step, acc=0.5050, loss=2.0292, lr=1.46e-05]
Training:  86%|████████▌ | 9514/11082 [30:35:39<5:01:09, 11.52s/step, acc=0.5050, loss=2.0292, lr=1.46e-05]
Training:  86%|████████▌ | 9515/11082 [30:35:50<4:59:43, 11.48s/step, acc=0.5050, loss=2.0292, lr=1.46e-05]
Training:  86%|████████▌ | 9516/11082 [30:36:02<5:02:15, 11.58s/step, acc=0.5050, loss=2.0292, lr=1.46e-05]
Training:  86%|████████▌ | 9517/11082 [30:36:13<4:59:01, 11.46s/step, acc=0.5050, loss=2.0292, lr=1.46e-05]
Training:  86%|████████▌ | 9518/11082 [30:36:25<5:01:11, 11.55s/step, acc=0.5050, loss=2.0292, lr=1.46e-05]
Training:  86%|████████▌ | 9519/11082 [30:36:37<5:01:19, 11.57s/step, acc=0.5050, loss=2.0292, lr=1.46e-05]
Training:  86%|████████▌ | 9520/11082 [30:36:48<5:02:48, 11.63s/step, acc=0.5050, loss=2.0292, lr=1.46e-05]
Training:  86%|████████▌ | 9521/11082 [30:37:00<5:01:23, 11.58s/step, acc=0.5050, loss=2.0292, lr=1.46e-05]
Training:  86%|████████▌ | 9522/11082 [30:37:11<4:59:50, 11.53s/step, acc=0.5050, loss=2.0292, lr=1.46e-05]
Training:  86%|████████▌ | 9523/11082 [30:37:23<5:02:00, 11.62s/step, acc=0.5050, loss=2.0292, lr=1.46e-05]
Training:  86%|████████▌ | 9524/11082 [30:37:35<5:02:48, 11.66s/step, acc=0.5050, loss=2.0292, lr=1.46e-05]
Training:  86%|████████▌ | 9525/11082 [30:37:46<5:00:04, 11.56s/step, acc=0.5050, loss=2.0292, lr=1.46e-05]
Training:  86%|████████▌ | 9526/11082 [30:37:58<5:00:12, 11.58s/step, acc=0.5050, loss=2.0292, lr=1.46e-05]
Training:  86%|████████▌ | 9527/11082 [30:38:09<4:55:04, 11.39s/step, acc=0.5050, loss=2.0292, lr=1.46e-05]
Training:  86%|████████▌ | 9528/11082 [30:38:21<4:57:29, 11.49s/step, acc=0.5050, loss=2.0292, lr=1.46e-05]
Training:  86%|████████▌ | 9529/11082 [30:38:32<4:58:49, 11.55s/step, acc=0.5050, loss=2.0292, lr=1.46e-05]
Training:  86%|████████▌ | 9530/11082 [30:38:43<4:55:39, 11.43s/step, acc=0.5050, loss=2.0292, lr=1.46e-05]
Training:  86%|████████▌ | 9531/11082 [30:38:55<4:58:41, 11.55s/step, acc=0.5050, loss=2.0292, lr=1.46e-05]
Training:  86%|████████▌ | 9532/11082 [30:39:07<4:57:58, 11.53s/step, acc=0.5050, loss=2.0292, lr=1.46e-05]
Training:  86%|████████▌ | 9533/11082 [30:39:18<4:59:17, 11.59s/step, acc=0.5050, loss=2.0292, lr=1.46e-05]
Training:  86%|████████▌ | 9534/11082 [30:39:30<4:58:30, 11.57s/step, acc=0.5050, loss=2.0292, lr=1.46e-05]
Training:  86%|████████▌ | 9535/11082 [30:39:42<5:00:54, 11.67s/step, acc=0.5050, loss=2.0292, lr=1.46e-05]
Training:  86%|████████▌ | 9536/11082 [30:39:54<5:01:15, 11.69s/step, acc=0.5050, loss=2.0292, lr=1.46e-05]
Training:  86%|████████▌ | 9537/11082 [30:40:05<4:57:17, 11.55s/step, acc=0.5050, loss=2.0292, lr=1.46e-05]
Training:  86%|████████▌ | 9538/11082 [30:40:16<4:58:05, 11.58s/step, acc=0.5050, loss=2.0292, lr=1.46e-05]
Training:  86%|████████▌ | 9539/11082 [30:40:29<5:01:45, 11.73s/step, acc=0.5050, loss=2.0292, lr=1.46e-05]
Training:  86%|████████▌ | 9540/11082 [30:40:40<4:57:21, 11.57s/step, acc=0.5050, loss=2.0292, lr=1.46e-05]
Training:  86%|████████▌ | 9541/11082 [30:40:51<4:57:48, 11.60s/step, acc=0.5050, loss=2.0292, lr=1.46e-05]
Training:  86%|████████▌ | 9542/11082 [30:41:03<4:58:32, 11.63s/step, acc=0.5050, loss=2.0292, lr=1.46e-05]
Training:  86%|████████▌ | 9543/11082 [30:41:14<4:53:13, 11.43s/step, acc=0.5050, loss=2.0292, lr=1.46e-05]
Training:  86%|████████▌ | 9544/11082 [30:41:26<4:58:31, 11.65s/step, acc=0.5050, loss=2.0292, lr=1.46e-05]
Training:  86%|████████▌ | 9545/11082 [30:41:37<4:52:38, 11.42s/step, acc=0.5050, loss=2.0292, lr=1.46e-05]
Training:  86%|████████▌ | 9546/11082 [30:41:49<4:56:55, 11.60s/step, acc=0.5050, loss=2.0292, lr=1.46e-05]
Training:  86%|████████▌ | 9547/11082 [30:42:01<4:56:58, 11.61s/step, acc=0.5050, loss=2.0292, lr=1.46e-05]
Training:  86%|████████▌ | 9548/11082 [30:42:13<4:57:38, 11.64s/step, acc=0.5050, loss=2.0292, lr=1.46e-05]
Training:  86%|████████▌ | 9549/11082 [30:42:23<4:51:21, 11.40s/step, acc=0.5050, loss=2.0292, lr=1.46e-05]
Training:  86%|████████▌ | 9550/11082 [30:42:35<4:54:47, 11.55s/step, acc=0.5050, loss=2.0292, lr=1.46e-05]
Training:  86%|████████▌ | 9550/11082 [30:42:43<4:54:47, 11.55s/step, acc=0.5076, loss=2.0148, lr=1.43e-05]2026-06-19 21:36:11 | INFO | __main__ | epoch=1 step=9550 train_loss=2.014821 train_acc=0.507585 lr=1.430e-05 grad_norm=0.5312 acc_d0=0.7772 acc_d1=0.6369 acc_d2=0.5195 acc_d3=0.4201 acc_d4=0.3403 acc_d5=0.2763 acc_d6=0.2229 acc_d7=0.1840

Training:  86%|████████▌ | 9551/11082 [30:42:47<4:55:33, 11.58s/step, acc=0.5076, loss=2.0148, lr=1.43e-05]
Training:  86%|████████▌ | 9552/11082 [30:42:58<4:53:46, 11.52s/step, acc=0.5076, loss=2.0148, lr=1.43e-05]
Training:  86%|████████▌ | 9553/11082 [30:43:10<4:56:20, 11.63s/step, acc=0.5076, loss=2.0148, lr=1.43e-05]
Training:  86%|████████▌ | 9554/11082 [30:43:22<4:58:18, 11.71s/step, acc=0.5076, loss=2.0148, lr=1.43e-05]
Training:  86%|████████▌ | 9555/11082 [30:43:33<4:52:40, 11.50s/step, acc=0.5076, loss=2.0148, lr=1.43e-05]
Training:  86%|████████▌ | 9556/11082 [30:43:45<4:57:09, 11.68s/step, acc=0.5076, loss=2.0148, lr=1.43e-05]
Training:  86%|████████▌ | 9557/11082 [30:43:57<4:54:21, 11.58s/step, acc=0.5076, loss=2.0148, lr=1.43e-05]
Training:  86%|████████▌ | 9558/11082 [30:44:08<4:52:02, 11.50s/step, acc=0.5076, loss=2.0148, lr=1.43e-05]
Training:  86%|████████▋ | 9559/11082 [30:44:20<4:53:42, 11.57s/step, acc=0.5076, loss=2.0148, lr=1.43e-05]
Training:  86%|████████▋ | 9560/11082 [30:44:31<4:52:59, 11.55s/step, acc=0.5076, loss=2.0148, lr=1.43e-05]
Training:  86%|████████▋ | 9561/11082 [30:44:43<4:51:54, 11.51s/step, acc=0.5076, loss=2.0148, lr=1.43e-05]
Training:  86%|████████▋ | 9562/11082 [30:44:54<4:52:25, 11.54s/step, acc=0.5076, loss=2.0148, lr=1.43e-05]
Training:  86%|████████▋ | 9563/11082 [30:45:06<4:54:52, 11.65s/step, acc=0.5076, loss=2.0148, lr=1.43e-05]
Training:  86%|████████▋ | 9564/11082 [30:45:18<4:53:47, 11.61s/step, acc=0.5076, loss=2.0148, lr=1.43e-05]
Training:  86%|████████▋ | 9565/11082 [30:45:28<4:46:43, 11.34s/step, acc=0.5076, loss=2.0148, lr=1.43e-05]
Training:  86%|████████▋ | 9566/11082 [30:45:40<4:50:27, 11.50s/step, acc=0.5076, loss=2.0148, lr=1.43e-05]
Training:  86%|████████▋ | 9567/11082 [30:45:52<4:54:17, 11.66s/step, acc=0.5076, loss=2.0148, lr=1.43e-05]
Training:  86%|████████▋ | 9568/11082 [30:46:04<4:52:30, 11.59s/step, acc=0.5076, loss=2.0148, lr=1.43e-05]
Training:  86%|████████▋ | 9569/11082 [30:46:15<4:52:29, 11.60s/step, acc=0.5076, loss=2.0148, lr=1.43e-05]
Training:  86%|████████▋ | 9570/11082 [30:46:27<4:50:34, 11.53s/step, acc=0.5076, loss=2.0148, lr=1.43e-05]
Training:  86%|████████▋ | 9571/11082 [30:46:38<4:53:12, 11.64s/step, acc=0.5076, loss=2.0148, lr=1.43e-05]
Training:  86%|████████▋ | 9572/11082 [30:46:50<4:52:17, 11.61s/step, acc=0.5076, loss=2.0148, lr=1.43e-05]
Training:  86%|████████▋ | 9573/11082 [30:47:01<4:50:57, 11.57s/step, acc=0.5076, loss=2.0148, lr=1.43e-05]
Training:  86%|████████▋ | 9574/11082 [30:47:14<4:54:20, 11.71s/step, acc=0.5076, loss=2.0148, lr=1.43e-05]
Training:  86%|████████▋ | 9575/11082 [30:47:24<4:46:08, 11.39s/step, acc=0.5076, loss=2.0148, lr=1.43e-05]
Training:  86%|████████▋ | 9576/11082 [30:47:37<4:54:15, 11.72s/step, acc=0.5076, loss=2.0148, lr=1.43e-05]
Training:  86%|████████▋ | 9577/11082 [30:47:48<4:48:51, 11.52s/step, acc=0.5076, loss=2.0148, lr=1.43e-05]
Training:  86%|████████▋ | 9578/11082 [30:48:00<4:51:48, 11.64s/step, acc=0.5076, loss=2.0148, lr=1.43e-05]
Training:  86%|████████▋ | 9579/11082 [30:48:11<4:50:53, 11.61s/step, acc=0.5076, loss=2.0148, lr=1.43e-05]
Training:  86%|████████▋ | 9580/11082 [30:48:22<4:45:20, 11.40s/step, acc=0.5076, loss=2.0148, lr=1.43e-05]
Training:  86%|████████▋ | 9581/11082 [30:48:34<4:50:57, 11.63s/step, acc=0.5076, loss=2.0148, lr=1.43e-05]
Training:  86%|████████▋ | 9582/11082 [30:48:45<4:47:30, 11.50s/step, acc=0.5076, loss=2.0148, lr=1.43e-05]
Training:  86%|████████▋ | 9583/11082 [30:48:57<4:50:14, 11.62s/step, acc=0.5076, loss=2.0148, lr=1.43e-05]
Training:  86%|████████▋ | 9584/11082 [30:49:09<4:46:48, 11.49s/step, acc=0.5076, loss=2.0148, lr=1.43e-05]
Training:  86%|████████▋ | 9585/11082 [30:49:20<4:48:02, 11.54s/step, acc=0.5076, loss=2.0148, lr=1.43e-05]
Training:  87%|████████▋ | 9586/11082 [30:49:32<4:48:53, 11.59s/step, acc=0.5076, loss=2.0148, lr=1.43e-05]
Training:  87%|████████▋ | 9587/11082 [30:49:44<4:49:33, 11.62s/step, acc=0.5076, loss=2.0148, lr=1.43e-05]
Training:  87%|████████▋ | 9588/11082 [30:49:55<4:46:21, 11.50s/step, acc=0.5076, loss=2.0148, lr=1.43e-05]
Training:  87%|████████▋ | 9589/11082 [30:50:07<4:47:48, 11.57s/step, acc=0.5076, loss=2.0148, lr=1.43e-05]
Training:  87%|████████▋ | 9590/11082 [30:50:18<4:48:53, 11.62s/step, acc=0.5076, loss=2.0148, lr=1.43e-05]
Training:  87%|████████▋ | 9591/11082 [30:50:30<4:47:15, 11.56s/step, acc=0.5076, loss=2.0148, lr=1.43e-05]
Training:  87%|████████▋ | 9592/11082 [30:50:42<4:49:13, 11.65s/step, acc=0.5076, loss=2.0148, lr=1.43e-05]
Training:  87%|████████▋ | 9593/11082 [30:50:53<4:46:57, 11.56s/step, acc=0.5076, loss=2.0148, lr=1.43e-05]
Training:  87%|████████▋ | 9594/11082 [30:51:04<4:46:17, 11.54s/step, acc=0.5076, loss=2.0148, lr=1.43e-05]
Training:  87%|████████▋ | 9595/11082 [30:51:16<4:47:05, 11.58s/step, acc=0.5076, loss=2.0148, lr=1.43e-05]
Training:  87%|████████▋ | 9596/11082 [30:51:27<4:45:33, 11.53s/step, acc=0.5076, loss=2.0148, lr=1.43e-05]
Training:  87%|████████▋ | 9597/11082 [30:51:39<4:42:24, 11.41s/step, acc=0.5076, loss=2.0148, lr=1.43e-05]
Training:  87%|████████▋ | 9598/11082 [30:51:50<4:43:07, 11.45s/step, acc=0.5076, loss=2.0148, lr=1.43e-05]
Training:  87%|████████▋ | 9599/11082 [30:52:02<4:45:04, 11.53s/step, acc=0.5076, loss=2.0148, lr=1.43e-05]
Training:  87%|████████▋ | 9600/11082 [30:52:13<4:43:29, 11.48s/step, acc=0.5076, loss=2.0148, lr=1.43e-05]
Training:  87%|████████▋ | 9600/11082 [30:52:21<4:43:29, 11.48s/step, acc=0.5069, loss=2.0179, lr=1.40e-05]2026-06-19 21:45:49 | INFO | __main__ | epoch=1 step=9600 train_loss=2.017928 train_acc=0.506913 lr=1.403e-05 grad_norm=0.5117 acc_d0=0.7574 acc_d1=0.6126 acc_d2=0.4917 acc_d3=0.3950 acc_d4=0.3170 acc_d5=0.2575 acc_d6=0.2082 acc_d7=0.1696

Training:  87%|████████▋ | 9601/11082 [30:52:25<4:48:11, 11.68s/step, acc=0.5069, loss=2.0179, lr=1.40e-05]
Training:  87%|████████▋ | 9602/11082 [30:52:37<4:47:19, 11.65s/step, acc=0.5069, loss=2.0179, lr=1.40e-05]
Training:  87%|████████▋ | 9603/11082 [30:52:49<4:46:28, 11.62s/step, acc=0.5069, loss=2.0179, lr=1.40e-05]
Training:  87%|████████▋ | 9604/11082 [30:53:00<4:44:32, 11.55s/step, acc=0.5069, loss=2.0179, lr=1.40e-05]
Training:  87%|████████▋ | 9605/11082 [30:53:12<4:47:07, 11.66s/step, acc=0.5069, loss=2.0179, lr=1.40e-05]
Training:  87%|████████▋ | 9606/11082 [30:53:23<4:45:45, 11.62s/step, acc=0.5069, loss=2.0179, lr=1.40e-05]
Training:  87%|████████▋ | 9607/11082 [30:53:35<4:45:25, 11.61s/step, acc=0.5069, loss=2.0179, lr=1.40e-05]
Training:  87%|████████▋ | 9608/11082 [30:53:47<4:46:32, 11.66s/step, acc=0.5069, loss=2.0179, lr=1.40e-05]
Training:  87%|████████▋ | 9609/11082 [30:53:58<4:46:06, 11.65s/step, acc=0.5069, loss=2.0179, lr=1.40e-05]
Training:  87%|████████▋ | 9610/11082 [30:54:10<4:44:07, 11.58s/step, acc=0.5069, loss=2.0179, lr=1.40e-05]
Training:  87%|████████▋ | 9611/11082 [30:54:21<4:44:49, 11.62s/step, acc=0.5069, loss=2.0179, lr=1.40e-05]
Training:  87%|████████▋ | 9612/11082 [30:54:33<4:41:52, 11.50s/step, acc=0.5069, loss=2.0179, lr=1.40e-05]
Training:  87%|████████▋ | 9613/11082 [30:54:45<4:44:53, 11.64s/step, acc=0.5069, loss=2.0179, lr=1.40e-05]
Training:  87%|████████▋ | 9614/11082 [30:54:56<4:43:35, 11.59s/step, acc=0.5069, loss=2.0179, lr=1.40e-05]
Training:  87%|████████▋ | 9615/11082 [30:55:07<4:40:06, 11.46s/step, acc=0.5069, loss=2.0179, lr=1.40e-05]
Training:  87%|████████▋ | 9616/11082 [30:55:19<4:44:43, 11.65s/step, acc=0.5069, loss=2.0179, lr=1.40e-05]
Training:  87%|████████▋ | 9617/11082 [30:55:31<4:43:06, 11.59s/step, acc=0.5069, loss=2.0179, lr=1.40e-05]
Training:  87%|████████▋ | 9618/11082 [30:55:43<4:44:56, 11.68s/step, acc=0.5069, loss=2.0179, lr=1.40e-05]
Training:  87%|████████▋ | 9619/11082 [30:55:54<4:40:42, 11.51s/step, acc=0.5069, loss=2.0179, lr=1.40e-05]
Training:  87%|████████▋ | 9620/11082 [30:56:06<4:44:14, 11.67s/step, acc=0.5069, loss=2.0179, lr=1.40e-05]
Training:  87%|████████▋ | 9621/11082 [30:56:17<4:39:47, 11.49s/step, acc=0.5069, loss=2.0179, lr=1.40e-05]
Training:  87%|████████▋ | 9622/11082 [30:56:29<4:43:05, 11.63s/step, acc=0.5069, loss=2.0179, lr=1.40e-05]
Training:  87%|████████▋ | 9623/11082 [30:56:40<4:38:07, 11.44s/step, acc=0.5069, loss=2.0179, lr=1.40e-05]
Training:  87%|████████▋ | 9624/11082 [30:56:52<4:40:11, 11.53s/step, acc=0.5069, loss=2.0179, lr=1.40e-05]
Training:  87%|████████▋ | 9625/11082 [30:57:03<4:41:51, 11.61s/step, acc=0.5069, loss=2.0179, lr=1.40e-05]
Training:  87%|████████▋ | 9626/11082 [30:57:15<4:38:45, 11.49s/step, acc=0.5069, loss=2.0179, lr=1.40e-05]
Training:  87%|████████▋ | 9627/11082 [30:57:26<4:37:10, 11.43s/step, acc=0.5069, loss=2.0179, lr=1.40e-05]
Training:  87%|████████▋ | 9628/11082 [30:57:37<4:36:34, 11.41s/step, acc=0.5069, loss=2.0179, lr=1.40e-05]
Training:  87%|████████▋ | 9629/11082 [30:57:50<4:42:44, 11.68s/step, acc=0.5069, loss=2.0179, lr=1.40e-05]
Training:  87%|████████▋ | 9630/11082 [30:58:01<4:43:29, 11.71s/step, acc=0.5069, loss=2.0179, lr=1.40e-05]
Training:  87%|████████▋ | 9631/11082 [30:58:13<4:42:42, 11.69s/step, acc=0.5069, loss=2.0179, lr=1.40e-05]
Training:  87%|████████▋ | 9632/11082 [30:58:24<4:40:44, 11.62s/step, acc=0.5069, loss=2.0179, lr=1.40e-05]
Training:  87%|████████▋ | 9633/11082 [30:58:36<4:41:42, 11.67s/step, acc=0.5069, loss=2.0179, lr=1.40e-05]
Training:  87%|████████▋ | 9634/11082 [30:58:48<4:42:20, 11.70s/step, acc=0.5069, loss=2.0179, lr=1.40e-05]
Training:  87%|████████▋ | 9635/11082 [30:58:59<4:38:05, 11.53s/step, acc=0.5069, loss=2.0179, lr=1.40e-05]
Training:  87%|████████▋ | 9636/11082 [30:59:11<4:39:05, 11.58s/step, acc=0.5069, loss=2.0179, lr=1.40e-05]
Training:  87%|████████▋ | 9637/11082 [30:59:22<4:39:00, 11.59s/step, acc=0.5069, loss=2.0179, lr=1.40e-05]
Training:  87%|████████▋ | 9638/11082 [30:59:34<4:36:21, 11.48s/step, acc=0.5069, loss=2.0179, lr=1.40e-05]
Training:  87%|████████▋ | 9639/11082 [30:59:45<4:35:07, 11.44s/step, acc=0.5069, loss=2.0179, lr=1.40e-05]
Training:  87%|████████▋ | 9640/11082 [30:59:57<4:39:00, 11.61s/step, acc=0.5069, loss=2.0179, lr=1.40e-05]
Training:  87%|████████▋ | 9641/11082 [31:00:08<4:37:33, 11.56s/step, acc=0.5069, loss=2.0179, lr=1.40e-05]
Training:  87%|████████▋ | 9642/11082 [31:00:20<4:38:39, 11.61s/step, acc=0.5069, loss=2.0179, lr=1.40e-05]
Training:  87%|████████▋ | 9643/11082 [31:00:32<4:38:01, 11.59s/step, acc=0.5069, loss=2.0179, lr=1.40e-05]
Training:  87%|████████▋ | 9644/11082 [31:00:44<4:39:11, 11.65s/step, acc=0.5069, loss=2.0179, lr=1.40e-05]
Training:  87%|████████▋ | 9645/11082 [31:00:55<4:37:40, 11.59s/step, acc=0.5069, loss=2.0179, lr=1.40e-05]
Training:  87%|████████▋ | 9646/11082 [31:01:06<4:33:23, 11.42s/step, acc=0.5069, loss=2.0179, lr=1.40e-05]
Training:  87%|████████▋ | 9647/11082 [31:01:18<4:35:46, 11.53s/step, acc=0.5069, loss=2.0179, lr=1.40e-05]
Training:  87%|████████▋ | 9648/11082 [31:01:30<4:36:55, 11.59s/step, acc=0.5069, loss=2.0179, lr=1.40e-05]
Training:  87%|████████▋ | 9649/11082 [31:01:41<4:34:43, 11.50s/step, acc=0.5069, loss=2.0179, lr=1.40e-05]
Training:  87%|████████▋ | 9650/11082 [31:01:53<4:36:19, 11.58s/step, acc=0.5069, loss=2.0179, lr=1.40e-05]
Training:  87%|████████▋ | 9650/11082 [31:02:00<4:36:19, 11.58s/step, acc=0.5082, loss=2.0104, lr=1.38e-05]2026-06-19 21:55:28 | INFO | __main__ | epoch=1 step=9650 train_loss=2.010383 train_acc=0.508206 lr=1.377e-05 grad_norm=0.5156 acc_d0=0.7608 acc_d1=0.6163 acc_d2=0.4972 acc_d3=0.3998 acc_d4=0.3223 acc_d5=0.2599 acc_d6=0.2108 acc_d7=0.1694

Training:  87%|████████▋ | 9651/11082 [31:02:04<4:33:27, 11.47s/step, acc=0.5082, loss=2.0104, lr=1.38e-05]
Training:  87%|████████▋ | 9652/11082 [31:02:16<4:39:11, 11.71s/step, acc=0.5082, loss=2.0104, lr=1.38e-05]
Training:  87%|████████▋ | 9653/11082 [31:02:28<4:37:12, 11.64s/step, acc=0.5082, loss=2.0104, lr=1.38e-05]
Training:  87%|████████▋ | 9654/11082 [31:02:38<4:31:54, 11.42s/step, acc=0.5082, loss=2.0104, lr=1.38e-05]
Training:  87%|████████▋ | 9655/11082 [31:02:50<4:35:04, 11.57s/step, acc=0.5082, loss=2.0104, lr=1.38e-05]
Training:  87%|████████▋ | 9656/11082 [31:03:02<4:34:28, 11.55s/step, acc=0.5082, loss=2.0104, lr=1.38e-05]
Training:  87%|████████▋ | 9657/11082 [31:03:13<4:33:44, 11.53s/step, acc=0.5082, loss=2.0104, lr=1.38e-05]
Training:  87%|████████▋ | 9658/11082 [31:03:25<4:33:31, 11.53s/step, acc=0.5082, loss=2.0104, lr=1.38e-05]
Training:  87%|████████▋ | 9659/11082 [31:03:36<4:33:38, 11.54s/step, acc=0.5082, loss=2.0104, lr=1.38e-05]
Training:  87%|████████▋ | 9660/11082 [31:03:48<4:32:57, 11.52s/step, acc=0.5082, loss=2.0104, lr=1.38e-05]
Training:  87%|████████▋ | 9661/11082 [31:03:59<4:32:16, 11.50s/step, acc=0.5082, loss=2.0104, lr=1.38e-05]
Training:  87%|████████▋ | 9662/11082 [31:04:11<4:30:45, 11.44s/step, acc=0.5082, loss=2.0104, lr=1.38e-05]
Training:  87%|████████▋ | 9663/11082 [31:04:22<4:32:50, 11.54s/step, acc=0.5082, loss=2.0104, lr=1.38e-05]
Training:  87%|████████▋ | 9664/11082 [31:04:34<4:30:48, 11.46s/step, acc=0.5082, loss=2.0104, lr=1.38e-05]
Training:  87%|████████▋ | 9665/11082 [31:04:46<4:36:12, 11.70s/step, acc=0.5082, loss=2.0104, lr=1.38e-05]
Training:  87%|████████▋ | 9666/11082 [31:04:57<4:34:15, 11.62s/step, acc=0.5082, loss=2.0104, lr=1.38e-05]
Training:  87%|████████▋ | 9667/11082 [31:05:09<4:34:27, 11.64s/step, acc=0.5082, loss=2.0104, lr=1.38e-05]
Training:  87%|████████▋ | 9668/11082 [31:05:20<4:29:29, 11.44s/step, acc=0.5082, loss=2.0104, lr=1.38e-05]
Training:  87%|████████▋ | 9669/11082 [31:05:32<4:34:37, 11.66s/step, acc=0.5082, loss=2.0104, lr=1.38e-05]
Training:  87%|████████▋ | 9670/11082 [31:05:44<4:31:56, 11.56s/step, acc=0.5082, loss=2.0104, lr=1.38e-05]
Training:  87%|████████▋ | 9671/11082 [31:05:55<4:29:13, 11.45s/step, acc=0.5082, loss=2.0104, lr=1.38e-05]
Training:  87%|████████▋ | 9672/11082 [31:06:07<4:35:07, 11.71s/step, acc=0.5082, loss=2.0104, lr=1.38e-05]
Training:  87%|████████▋ | 9673/11082 [31:06:19<4:33:11, 11.63s/step, acc=0.5082, loss=2.0104, lr=1.38e-05]
Training:  87%|████████▋ | 9674/11082 [31:06:30<4:30:39, 11.53s/step, acc=0.5082, loss=2.0104, lr=1.38e-05]
Training:  87%|████████▋ | 9675/11082 [31:06:41<4:30:29, 11.53s/step, acc=0.5082, loss=2.0104, lr=1.38e-05]
Training:  87%|████████▋ | 9676/11082 [31:06:53<4:32:27, 11.63s/step, acc=0.5082, loss=2.0104, lr=1.38e-05]
Training:  87%|████████▋ | 9677/11082 [31:07:04<4:27:59, 11.44s/step, acc=0.5082, loss=2.0104, lr=1.38e-05]
Training:  87%|████████▋ | 9678/11082 [31:07:16<4:29:28, 11.52s/step, acc=0.5082, loss=2.0104, lr=1.38e-05]
Training:  87%|████████▋ | 9679/11082 [31:07:28<4:30:11, 11.56s/step, acc=0.5082, loss=2.0104, lr=1.38e-05]
Training:  87%|████████▋ | 9680/11082 [31:07:39<4:30:00, 11.56s/step, acc=0.5082, loss=2.0104, lr=1.38e-05]
Training:  87%|████████▋ | 9681/11082 [31:07:51<4:29:57, 11.56s/step, acc=0.5082, loss=2.0104, lr=1.38e-05]
Training:  87%|████████▋ | 9682/11082 [31:08:02<4:29:08, 11.53s/step, acc=0.5082, loss=2.0104, lr=1.38e-05]
Training:  87%|████████▋ | 9683/11082 [31:08:14<4:28:45, 11.53s/step, acc=0.5082, loss=2.0104, lr=1.38e-05]
Training:  87%|████████▋ | 9684/11082 [31:08:25<4:27:34, 11.48s/step, acc=0.5082, loss=2.0104, lr=1.38e-05]
Training:  87%|████████▋ | 9685/11082 [31:08:36<4:26:52, 11.46s/step, acc=0.5082, loss=2.0104, lr=1.38e-05]
Training:  87%|████████▋ | 9686/11082 [31:08:48<4:25:51, 11.43s/step, acc=0.5082, loss=2.0104, lr=1.38e-05]
Training:  87%|████████▋ | 9687/11082 [31:09:00<4:32:29, 11.72s/step, acc=0.5082, loss=2.0104, lr=1.38e-05]
Training:  87%|████████▋ | 9688/11082 [31:09:12<4:32:07, 11.71s/step, acc=0.5082, loss=2.0104, lr=1.38e-05]
Training:  87%|████████▋ | 9689/11082 [31:09:23<4:27:21, 11.52s/step, acc=0.5082, loss=2.0104, lr=1.38e-05]
Training:  87%|████████▋ | 9690/11082 [31:09:35<4:28:05, 11.56s/step, acc=0.5082, loss=2.0104, lr=1.38e-05]
Training:  87%|████████▋ | 9691/11082 [31:09:46<4:29:27, 11.62s/step, acc=0.5082, loss=2.0104, lr=1.38e-05]
Training:  87%|████████▋ | 9692/11082 [31:09:58<4:29:49, 11.65s/step, acc=0.5082, loss=2.0104, lr=1.38e-05]
Training:  87%|████████▋ | 9693/11082 [31:10:09<4:25:08, 11.45s/step, acc=0.5082, loss=2.0104, lr=1.38e-05]
Training:  87%|████████▋ | 9694/11082 [31:10:21<4:28:06, 11.59s/step, acc=0.5082, loss=2.0104, lr=1.38e-05]
Training:  87%|████████▋ | 9695/11082 [31:10:33<4:28:39, 11.62s/step, acc=0.5082, loss=2.0104, lr=1.38e-05]
Training:  87%|████████▋ | 9696/11082 [31:10:44<4:24:16, 11.44s/step, acc=0.5082, loss=2.0104, lr=1.38e-05]
Training:  88%|████████▊ | 9697/11082 [31:10:56<4:26:59, 11.57s/step, acc=0.5082, loss=2.0104, lr=1.38e-05]
Training:  88%|████████▊ | 9698/11082 [31:11:07<4:27:54, 11.61s/step, acc=0.5082, loss=2.0104, lr=1.38e-05]
Training:  88%|████████▊ | 9699/11082 [31:11:19<4:27:18, 11.60s/step, acc=0.5082, loss=2.0104, lr=1.38e-05]
Training:  88%|████████▊ | 9700/11082 [31:11:30<4:26:40, 11.58s/step, acc=0.5082, loss=2.0104, lr=1.38e-05]
Training:  88%|████████▊ | 9700/11082 [31:11:38<4:26:40, 11.58s/step, acc=0.5069, loss=2.0202, lr=1.35e-05]2026-06-19 22:05:06 | INFO | __main__ | epoch=1 step=9700 train_loss=2.020244 train_acc=0.506878 lr=1.351e-05 grad_norm=0.4961 acc_d0=0.7557 acc_d1=0.6120 acc_d2=0.4912 acc_d3=0.3943 acc_d4=0.3170 acc_d5=0.2579 acc_d6=0.2071 acc_d7=0.1689

Training:  88%|████████▊ | 9701/11082 [31:11:42<4:28:28, 11.66s/step, acc=0.5069, loss=2.0202, lr=1.35e-05]
Training:  88%|████████▊ | 9702/11082 [31:11:53<4:23:24, 11.45s/step, acc=0.5069, loss=2.0202, lr=1.35e-05]
Training:  88%|████████▊ | 9703/11082 [31:12:05<4:26:21, 11.59s/step, acc=0.5069, loss=2.0202, lr=1.35e-05]
Training:  88%|████████▊ | 9704/11082 [31:12:17<4:27:49, 11.66s/step, acc=0.5069, loss=2.0202, lr=1.35e-05]
Training:  88%|████████▊ | 9705/11082 [31:12:29<4:28:21, 11.69s/step, acc=0.5069, loss=2.0202, lr=1.35e-05]
Training:  88%|████████▊ | 9706/11082 [31:12:40<4:27:13, 11.65s/step, acc=0.5069, loss=2.0202, lr=1.35e-05]
Training:  88%|████████▊ | 9707/11082 [31:12:52<4:25:30, 11.59s/step, acc=0.5069, loss=2.0202, lr=1.35e-05]
Training:  88%|████████▊ | 9708/11082 [31:13:03<4:24:51, 11.57s/step, acc=0.5069, loss=2.0202, lr=1.35e-05]
Training:  88%|████████▊ | 9709/11082 [31:13:14<4:21:47, 11.44s/step, acc=0.5069, loss=2.0202, lr=1.35e-05]
Training:  88%|████████▊ | 9710/11082 [31:13:26<4:23:49, 11.54s/step, acc=0.5069, loss=2.0202, lr=1.35e-05]
Training:  88%|████████▊ | 9711/11082 [31:13:38<4:26:43, 11.67s/step, acc=0.5069, loss=2.0202, lr=1.35e-05]
Training:  88%|████████▊ | 9712/11082 [31:13:50<4:25:25, 11.62s/step, acc=0.5069, loss=2.0202, lr=1.35e-05]
Training:  88%|████████▊ | 9713/11082 [31:14:01<4:23:44, 11.56s/step, acc=0.5069, loss=2.0202, lr=1.35e-05]
Training:  88%|████████▊ | 9714/11082 [31:14:13<4:23:13, 11.55s/step, acc=0.5069, loss=2.0202, lr=1.35e-05]
Training:  88%|████████▊ | 9715/11082 [31:14:25<4:26:55, 11.72s/step, acc=0.5069, loss=2.0202, lr=1.35e-05]
Training:  88%|████████▊ | 9716/11082 [31:14:36<4:24:04, 11.60s/step, acc=0.5069, loss=2.0202, lr=1.35e-05]
Training:  88%|████████▊ | 9717/11082 [31:14:48<4:25:23, 11.67s/step, acc=0.5069, loss=2.0202, lr=1.35e-05]
Training:  88%|████████▊ | 9718/11082 [31:14:59<4:22:11, 11.53s/step, acc=0.5069, loss=2.0202, lr=1.35e-05]
Training:  88%|████████▊ | 9719/11082 [31:15:11<4:25:17, 11.68s/step, acc=0.5069, loss=2.0202, lr=1.35e-05]
Training:  88%|████████▊ | 9720/11082 [31:15:23<4:25:31, 11.70s/step, acc=0.5069, loss=2.0202, lr=1.35e-05]
Training:  88%|████████▊ | 9721/11082 [31:15:35<4:25:21, 11.70s/step, acc=0.5069, loss=2.0202, lr=1.35e-05]
Training:  88%|████████▊ | 9722/11082 [31:15:46<4:23:02, 11.60s/step, acc=0.5069, loss=2.0202, lr=1.35e-05]
Training:  88%|████████▊ | 9723/11082 [31:15:57<4:21:34, 11.55s/step, acc=0.5069, loss=2.0202, lr=1.35e-05]
Training:  88%|████████▊ | 9724/11082 [31:16:09<4:22:24, 11.59s/step, acc=0.5069, loss=2.0202, lr=1.35e-05]
Training:  88%|████████▊ | 9725/11082 [31:16:20<4:20:45, 11.53s/step, acc=0.5069, loss=2.0202, lr=1.35e-05]
Training:  88%|████████▊ | 9726/11082 [31:16:31<4:16:38, 11.36s/step, acc=0.5069, loss=2.0202, lr=1.35e-05]
Training:  88%|████████▊ | 9727/11082 [31:16:43<4:20:38, 11.54s/step, acc=0.5069, loss=2.0202, lr=1.35e-05]
Training:  88%|████████▊ | 9728/11082 [31:16:55<4:20:10, 11.53s/step, acc=0.5069, loss=2.0202, lr=1.35e-05]
Training:  88%|████████▊ | 9729/11082 [31:17:06<4:19:55, 11.53s/step, acc=0.5069, loss=2.0202, lr=1.35e-05]
Training:  88%|████████▊ | 9730/11082 [31:17:18<4:21:36, 11.61s/step, acc=0.5069, loss=2.0202, lr=1.35e-05]
Training:  88%|████████▊ | 9731/11082 [31:17:29<4:17:38, 11.44s/step, acc=0.5069, loss=2.0202, lr=1.35e-05]
Training:  88%|████████▊ | 9732/11082 [31:17:41<4:20:30, 11.58s/step, acc=0.5069, loss=2.0202, lr=1.35e-05]
Training:  88%|████████▊ | 9733/11082 [31:17:52<4:18:12, 11.48s/step, acc=0.5069, loss=2.0202, lr=1.35e-05]
Training:  88%|████████▊ | 9734/11082 [31:18:04<4:20:02, 11.57s/step, acc=0.5069, loss=2.0202, lr=1.35e-05]
Training:  88%|████████▊ | 9735/11082 [31:18:16<4:23:13, 11.72s/step, acc=0.5069, loss=2.0202, lr=1.35e-05]
Training:  88%|████████▊ | 9736/11082 [31:18:28<4:22:01, 11.68s/step, acc=0.5069, loss=2.0202, lr=1.35e-05]
Training:  88%|████████▊ | 9737/11082 [31:18:39<4:18:20, 11.52s/step, acc=0.5069, loss=2.0202, lr=1.35e-05]
Training:  88%|████████▊ | 9738/11082 [31:18:51<4:22:12, 11.71s/step, acc=0.5069, loss=2.0202, lr=1.35e-05]
Training:  88%|████████▊ | 9739/11082 [31:19:03<4:21:21, 11.68s/step, acc=0.5069, loss=2.0202, lr=1.35e-05]
Training:  88%|████████▊ | 9740/11082 [31:19:14<4:18:16, 11.55s/step, acc=0.5069, loss=2.0202, lr=1.35e-05]
Training:  88%|████████▊ | 9741/11082 [31:19:26<4:19:28, 11.61s/step, acc=0.5069, loss=2.0202, lr=1.35e-05]
Training:  88%|████████▊ | 9742/11082 [31:19:37<4:16:31, 11.49s/step, acc=0.5069, loss=2.0202, lr=1.35e-05]
Training:  88%|████████▊ | 9743/11082 [31:19:49<4:18:06, 11.57s/step, acc=0.5069, loss=2.0202, lr=1.35e-05]
Training:  88%|████████▊ | 9744/11082 [31:20:00<4:19:53, 11.65s/step, acc=0.5069, loss=2.0202, lr=1.35e-05]
Training:  88%|████████▊ | 9745/11082 [31:20:12<4:16:28, 11.51s/step, acc=0.5069, loss=2.0202, lr=1.35e-05]
Training:  88%|████████▊ | 9746/11082 [31:20:23<4:15:12, 11.46s/step, acc=0.5069, loss=2.0202, lr=1.35e-05]
Training:  88%|████████▊ | 9747/11082 [31:20:35<4:19:04, 11.64s/step, acc=0.5069, loss=2.0202, lr=1.35e-05]
Training:  88%|████████▊ | 9748/11082 [31:20:46<4:15:32, 11.49s/step, acc=0.5069, loss=2.0202, lr=1.35e-05]
Training:  88%|████████▊ | 9749/11082 [31:20:58<4:17:45, 11.60s/step, acc=0.5069, loss=2.0202, lr=1.35e-05]
Training:  88%|████████▊ | 9750/11082 [31:21:10<4:17:09, 11.58s/step, acc=0.5069, loss=2.0202, lr=1.35e-05]
Training:  88%|████████▊ | 9750/11082 [31:21:17<4:17:09, 11.58s/step, acc=0.5077, loss=2.0124, lr=1.33e-05]2026-06-19 22:14:45 | INFO | __main__ | epoch=1 step=9750 train_loss=2.012401 train_acc=0.507706 lr=1.327e-05 grad_norm=0.8008 acc_d0=0.7697 acc_d1=0.6286 acc_d2=0.5088 acc_d3=0.4123 acc_d4=0.3343 acc_d5=0.2694 acc_d6=0.2183 acc_d7=0.1770

Training:  88%|████████▊ | 9751/11082 [31:21:21<4:13:42, 11.44s/step, acc=0.5077, loss=2.0124, lr=1.33e-05]
Training:  88%|████████▊ | 9752/11082 [31:21:32<4:15:44, 11.54s/step, acc=0.5077, loss=2.0124, lr=1.33e-05]
Training:  88%|████████▊ | 9753/11082 [31:21:44<4:18:11, 11.66s/step, acc=0.5077, loss=2.0124, lr=1.33e-05]
Training:  88%|████████▊ | 9754/11082 [31:21:56<4:15:18, 11.54s/step, acc=0.5077, loss=2.0124, lr=1.33e-05]
Training:  88%|████████▊ | 9755/11082 [31:22:07<4:14:25, 11.50s/step, acc=0.5077, loss=2.0124, lr=1.33e-05]
Training:  88%|████████▊ | 9756/11082 [31:22:19<4:15:30, 11.56s/step, acc=0.5077, loss=2.0124, lr=1.33e-05]
Training:  88%|████████▊ | 9757/11082 [31:22:31<4:16:27, 11.61s/step, acc=0.5077, loss=2.0124, lr=1.33e-05]
Training:  88%|████████▊ | 9758/11082 [31:22:42<4:16:42, 11.63s/step, acc=0.5077, loss=2.0124, lr=1.33e-05]
Training:  88%|████████▊ | 9759/11082 [31:22:54<4:15:34, 11.59s/step, acc=0.5077, loss=2.0124, lr=1.33e-05]
Training:  88%|████████▊ | 9760/11082 [31:23:05<4:10:37, 11.37s/step, acc=0.5077, loss=2.0124, lr=1.33e-05]
Training:  88%|████████▊ | 9761/11082 [31:23:16<4:13:00, 11.49s/step, acc=0.5077, loss=2.0124, lr=1.33e-05]
Training:  88%|████████▊ | 9762/11082 [31:23:28<4:14:33, 11.57s/step, acc=0.5077, loss=2.0124, lr=1.33e-05]
Training:  88%|████████▊ | 9763/11082 [31:23:40<4:14:38, 11.58s/step, acc=0.5077, loss=2.0124, lr=1.33e-05]
Training:  88%|████████▊ | 9764/11082 [31:23:51<4:13:41, 11.55s/step, acc=0.5077, loss=2.0124, lr=1.33e-05]
Training:  88%|████████▊ | 9765/11082 [31:24:03<4:17:53, 11.75s/step, acc=0.5077, loss=2.0124, lr=1.33e-05]
Training:  88%|████████▊ | 9766/11082 [31:24:15<4:17:13, 11.73s/step, acc=0.5077, loss=2.0124, lr=1.33e-05]
Training:  88%|████████▊ | 9767/11082 [31:24:27<4:17:05, 11.73s/step, acc=0.5077, loss=2.0124, lr=1.33e-05]
Training:  88%|████████▊ | 9768/11082 [31:24:38<4:15:46, 11.68s/step, acc=0.5077, loss=2.0124, lr=1.33e-05]
Training:  88%|████████▊ | 9769/11082 [31:24:50<4:15:29, 11.68s/step, acc=0.5077, loss=2.0124, lr=1.33e-05]
Training:  88%|████████▊ | 9770/11082 [31:25:01<4:13:43, 11.60s/step, acc=0.5077, loss=2.0124, lr=1.33e-05]
Training:  88%|████████▊ | 9771/11082 [31:25:13<4:13:39, 11.61s/step, acc=0.5077, loss=2.0124, lr=1.33e-05]
Training:  88%|████████▊ | 9772/11082 [31:25:25<4:13:57, 11.63s/step, acc=0.5077, loss=2.0124, lr=1.33e-05]
Training:  88%|████████▊ | 9773/11082 [31:25:37<4:15:25, 11.71s/step, acc=0.5077, loss=2.0124, lr=1.33e-05]
Training:  88%|████████▊ | 9774/11082 [31:25:48<4:12:29, 11.58s/step, acc=0.5077, loss=2.0124, lr=1.33e-05]
Training:  88%|████████▊ | 9775/11082 [31:26:00<4:12:09, 11.58s/step, acc=0.5077, loss=2.0124, lr=1.33e-05]
Training:  88%|████████▊ | 9776/11082 [31:26:11<4:08:42, 11.43s/step, acc=0.5077, loss=2.0124, lr=1.33e-05]
Training:  88%|████████▊ | 9777/11082 [31:26:22<4:11:00, 11.54s/step, acc=0.5077, loss=2.0124, lr=1.33e-05]
Training:  88%|████████▊ | 9778/11082 [31:26:34<4:09:56, 11.50s/step, acc=0.5077, loss=2.0124, lr=1.33e-05]
Training:  88%|████████▊ | 9779/11082 [31:26:45<4:10:25, 11.53s/step, acc=0.5077, loss=2.0124, lr=1.33e-05]
Training:  88%|████████▊ | 9780/11082 [31:26:57<4:10:39, 11.55s/step, acc=0.5077, loss=2.0124, lr=1.33e-05]
Training:  88%|████████▊ | 9781/11082 [31:27:09<4:10:43, 11.56s/step, acc=0.5077, loss=2.0124, lr=1.33e-05]
Training:  88%|████████▊ | 9782/11082 [31:27:20<4:11:28, 11.61s/step, acc=0.5077, loss=2.0124, lr=1.33e-05]
Training:  88%|████████▊ | 9783/11082 [31:27:32<4:09:24, 11.52s/step, acc=0.5077, loss=2.0124, lr=1.33e-05]
Training:  88%|████████▊ | 9784/11082 [31:27:43<4:07:51, 11.46s/step, acc=0.5077, loss=2.0124, lr=1.33e-05]
Training:  88%|████████▊ | 9785/11082 [31:27:54<4:07:38, 11.46s/step, acc=0.5077, loss=2.0124, lr=1.33e-05]
Training:  88%|████████▊ | 9786/11082 [31:28:06<4:09:57, 11.57s/step, acc=0.5077, loss=2.0124, lr=1.33e-05]
Training:  88%|████████▊ | 9787/11082 [31:28:18<4:09:02, 11.54s/step, acc=0.5077, loss=2.0124, lr=1.33e-05]
Training:  88%|████████▊ | 9788/11082 [31:28:30<4:10:41, 11.62s/step, acc=0.5077, loss=2.0124, lr=1.33e-05]
Training:  88%|████████▊ | 9789/11082 [31:28:41<4:11:21, 11.66s/step, acc=0.5077, loss=2.0124, lr=1.33e-05]
Training:  88%|████████▊ | 9790/11082 [31:28:53<4:08:33, 11.54s/step, acc=0.5077, loss=2.0124, lr=1.33e-05]
Training:  88%|████████▊ | 9791/11082 [31:29:04<4:07:45, 11.52s/step, acc=0.5077, loss=2.0124, lr=1.33e-05]
Training:  88%|████████▊ | 9792/11082 [31:29:15<4:06:35, 11.47s/step, acc=0.5077, loss=2.0124, lr=1.33e-05]
Training:  88%|████████▊ | 9793/11082 [31:29:27<4:06:57, 11.50s/step, acc=0.5077, loss=2.0124, lr=1.33e-05]
Training:  88%|████████▊ | 9794/11082 [31:29:39<4:09:26, 11.62s/step, acc=0.5077, loss=2.0124, lr=1.33e-05]
Training:  88%|████████▊ | 9795/11082 [31:29:50<4:05:41, 11.45s/step, acc=0.5077, loss=2.0124, lr=1.33e-05]
Training:  88%|████████▊ | 9796/11082 [31:30:02<4:06:40, 11.51s/step, acc=0.5077, loss=2.0124, lr=1.33e-05]
Training:  88%|████████▊ | 9797/11082 [31:30:13<4:08:30, 11.60s/step, acc=0.5077, loss=2.0124, lr=1.33e-05]
Training:  88%|████████▊ | 9798/11082 [31:30:25<4:07:39, 11.57s/step, acc=0.5077, loss=2.0124, lr=1.33e-05]
Training:  88%|████████▊ | 9799/11082 [31:30:36<4:05:01, 11.46s/step, acc=0.5077, loss=2.0124, lr=1.33e-05]
Training:  88%|████████▊ | 9800/11082 [31:30:48<4:08:39, 11.64s/step, acc=0.5077, loss=2.0124, lr=1.33e-05]
Training:  88%|████████▊ | 9800/11082 [31:30:55<4:08:39, 11.64s/step, acc=0.5053, loss=2.0280, lr=1.30e-05]2026-06-19 22:24:23 | INFO | __main__ | epoch=1 step=9800 train_loss=2.028037 train_acc=0.505299 lr=1.303e-05 grad_norm=0.5039 acc_d0=0.7731 acc_d1=0.6340 acc_d2=0.5172 acc_d3=0.4201 acc_d4=0.3413 acc_d5=0.2751 acc_d6=0.2245 acc_d7=0.1840

Training:  88%|████████▊ | 9801/11082 [31:30:59<4:06:35, 11.55s/step, acc=0.5053, loss=2.0280, lr=1.30e-05]
Training:  88%|████████▊ | 9802/11082 [31:31:11<4:05:35, 11.51s/step, acc=0.5053, loss=2.0280, lr=1.30e-05]
Training:  88%|████████▊ | 9803/11082 [31:31:23<4:06:30, 11.56s/step, acc=0.5053, loss=2.0280, lr=1.30e-05]
Training:  88%|████████▊ | 9804/11082 [31:31:34<4:05:48, 11.54s/step, acc=0.5053, loss=2.0280, lr=1.30e-05]
Training:  88%|████████▊ | 9805/11082 [31:31:46<4:08:26, 11.67s/step, acc=0.5053, loss=2.0280, lr=1.30e-05]
Training:  88%|████████▊ | 9806/11082 [31:31:57<4:04:59, 11.52s/step, acc=0.5053, loss=2.0280, lr=1.30e-05]
Training:  88%|████████▊ | 9807/11082 [31:32:09<4:08:13, 11.68s/step, acc=0.5053, loss=2.0280, lr=1.30e-05]
Training:  89%|████████▊ | 9808/11082 [31:32:20<4:03:28, 11.47s/step, acc=0.5053, loss=2.0280, lr=1.30e-05]
Training:  89%|████████▊ | 9809/11082 [31:32:32<4:06:34, 11.62s/step, acc=0.5053, loss=2.0280, lr=1.30e-05]
Training:  89%|████████▊ | 9810/11082 [31:32:44<4:04:43, 11.54s/step, acc=0.5053, loss=2.0280, lr=1.30e-05]
Training:  89%|████████▊ | 9811/11082 [31:32:55<4:05:20, 11.58s/step, acc=0.5053, loss=2.0280, lr=1.30e-05]
Training:  89%|████████▊ | 9812/11082 [31:33:07<4:08:08, 11.72s/step, acc=0.5053, loss=2.0280, lr=1.30e-05]
Training:  89%|████████▊ | 9813/11082 [31:33:19<4:10:52, 11.86s/step, acc=0.5053, loss=2.0280, lr=1.30e-05]
Training:  89%|████████▊ | 9814/11082 [31:33:31<4:05:42, 11.63s/step, acc=0.5053, loss=2.0280, lr=1.30e-05]
Training:  89%|████████▊ | 9815/11082 [31:33:42<4:05:57, 11.65s/step, acc=0.5053, loss=2.0280, lr=1.30e-05]
Training:  89%|████████▊ | 9816/11082 [31:33:54<4:04:48, 11.60s/step, acc=0.5053, loss=2.0280, lr=1.30e-05]
Training:  89%|████████▊ | 9817/11082 [31:34:05<4:03:50, 11.57s/step, acc=0.5053, loss=2.0280, lr=1.30e-05]
Training:  89%|████████▊ | 9818/11082 [31:34:17<4:05:03, 11.63s/step, acc=0.5053, loss=2.0280, lr=1.30e-05]
Training:  89%|████████▊ | 9819/11082 [31:34:29<4:04:37, 11.62s/step, acc=0.5053, loss=2.0280, lr=1.30e-05]
Training:  89%|████████▊ | 9820/11082 [31:34:40<4:03:32, 11.58s/step, acc=0.5053, loss=2.0280, lr=1.30e-05]
Training:  89%|████████▊ | 9821/11082 [31:34:52<4:02:43, 11.55s/step, acc=0.5053, loss=2.0280, lr=1.30e-05]
Training:  89%|████████▊ | 9822/11082 [31:35:03<4:02:32, 11.55s/step, acc=0.5053, loss=2.0280, lr=1.30e-05]
Training:  89%|████████▊ | 9823/11082 [31:35:15<4:04:03, 11.63s/step, acc=0.5053, loss=2.0280, lr=1.30e-05]
Training:  89%|████████▊ | 9824/11082 [31:35:26<4:02:33, 11.57s/step, acc=0.5053, loss=2.0280, lr=1.30e-05]
Training:  89%|████████▊ | 9825/11082 [31:35:38<4:00:32, 11.48s/step, acc=0.5053, loss=2.0280, lr=1.30e-05]
Training:  89%|████████▊ | 9826/11082 [31:35:49<4:01:15, 11.53s/step, acc=0.5053, loss=2.0280, lr=1.30e-05]
Training:  89%|████████▊ | 9827/11082 [31:36:01<4:02:18, 11.58s/step, acc=0.5053, loss=2.0280, lr=1.30e-05]
Training:  89%|████████▊ | 9828/11082 [31:36:12<4:01:11, 11.54s/step, acc=0.5053, loss=2.0280, lr=1.30e-05]
Training:  89%|████████▊ | 9829/11082 [31:36:24<4:03:11, 11.65s/step, acc=0.5053, loss=2.0280, lr=1.30e-05]
Training:  89%|████████▊ | 9830/11082 [31:36:36<4:01:25, 11.57s/step, acc=0.5053, loss=2.0280, lr=1.30e-05]
Training:  89%|████████▊ | 9831/11082 [31:36:47<4:02:23, 11.63s/step, acc=0.5053, loss=2.0280, lr=1.30e-05]
Training:  89%|████████▊ | 9832/11082 [31:36:59<4:02:49, 11.66s/step, acc=0.5053, loss=2.0280, lr=1.30e-05]
Training:  89%|████████▊ | 9833/11082 [31:37:10<3:59:44, 11.52s/step, acc=0.5053, loss=2.0280, lr=1.30e-05]
Training:  89%|████████▊ | 9834/11082 [31:37:22<3:57:44, 11.43s/step, acc=0.5053, loss=2.0280, lr=1.30e-05]
Training:  89%|████████▊ | 9835/11082 [31:37:33<3:59:36, 11.53s/step, acc=0.5053, loss=2.0280, lr=1.30e-05]
Training:  89%|████████▉ | 9836/11082 [31:37:45<3:58:21, 11.48s/step, acc=0.5053, loss=2.0280, lr=1.30e-05]
Training:  89%|████████▉ | 9837/11082 [31:37:57<4:00:33, 11.59s/step, acc=0.5053, loss=2.0280, lr=1.30e-05]
Training:  89%|████████▉ | 9838/11082 [31:38:08<4:00:57, 11.62s/step, acc=0.5053, loss=2.0280, lr=1.30e-05]
Training:  89%|████████▉ | 9839/11082 [31:38:20<4:02:42, 11.72s/step, acc=0.5053, loss=2.0280, lr=1.30e-05]
Training:  89%|████████▉ | 9840/11082 [31:38:31<3:58:49, 11.54s/step, acc=0.5053, loss=2.0280, lr=1.30e-05]
Training:  89%|████████▉ | 9841/11082 [31:38:43<4:00:33, 11.63s/step, acc=0.5053, loss=2.0280, lr=1.30e-05]
Training:  89%|████████▉ | 9842/11082 [31:38:54<3:57:38, 11.50s/step, acc=0.5053, loss=2.0280, lr=1.30e-05]
Training:  89%|████████▉ | 9843/11082 [31:39:06<3:58:52, 11.57s/step, acc=0.5053, loss=2.0280, lr=1.30e-05]
Training:  89%|████████▉ | 9844/11082 [31:39:17<3:56:47, 11.48s/step, acc=0.5053, loss=2.0280, lr=1.30e-05]
Training:  89%|████████▉ | 9845/11082 [31:39:29<3:59:09, 11.60s/step, acc=0.5053, loss=2.0280, lr=1.30e-05]
Training:  89%|████████▉ | 9846/11082 [31:39:41<3:57:33, 11.53s/step, acc=0.5053, loss=2.0280, lr=1.30e-05]
Training:  89%|████████▉ | 9847/11082 [31:39:52<3:58:39, 11.60s/step, acc=0.5053, loss=2.0280, lr=1.30e-05]
Training:  89%|████████▉ | 9848/11082 [31:40:04<3:58:49, 11.61s/step, acc=0.5053, loss=2.0280, lr=1.30e-05]
Training:  89%|████████▉ | 9849/11082 [31:40:16<4:00:56, 11.72s/step, acc=0.5053, loss=2.0280, lr=1.30e-05]
Training:  89%|████████▉ | 9850/11082 [31:40:28<3:59:43, 11.68s/step, acc=0.5053, loss=2.0280, lr=1.30e-05]
Training:  89%|████████▉ | 9850/11082 [31:40:34<3:59:43, 11.68s/step, acc=0.5088, loss=2.0117, lr=1.28e-05]2026-06-19 22:34:02 | INFO | __main__ | epoch=1 step=9850 train_loss=2.011693 train_acc=0.508790 lr=1.280e-05 grad_norm=0.4824 acc_d0=0.7646 acc_d1=0.6241 acc_d2=0.5073 acc_d3=0.4111 acc_d4=0.3331 acc_d5=0.2698 acc_d6=0.2185 acc_d7=0.1801

Training:  89%|████████▉ | 9851/11082 [31:40:39<3:59:28, 11.67s/step, acc=0.5088, loss=2.0117, lr=1.28e-05]
Training:  89%|████████▉ | 9852/11082 [31:40:51<3:58:22, 11.63s/step, acc=0.5088, loss=2.0117, lr=1.28e-05]
Training:  89%|████████▉ | 9853/11082 [31:41:02<3:56:09, 11.53s/step, acc=0.5088, loss=2.0117, lr=1.28e-05]
Training:  89%|████████▉ | 9854/11082 [31:41:13<3:54:11, 11.44s/step, acc=0.5088, loss=2.0117, lr=1.28e-05]
Training:  89%|████████▉ | 9855/11082 [31:41:25<3:56:33, 11.57s/step, acc=0.5088, loss=2.0117, lr=1.28e-05]
Training:  89%|████████▉ | 9856/11082 [31:41:36<3:52:48, 11.39s/step, acc=0.5088, loss=2.0117, lr=1.28e-05]
Training:  89%|████████▉ | 9857/11082 [31:41:48<3:55:50, 11.55s/step, acc=0.5088, loss=2.0117, lr=1.28e-05]
Training:  89%|████████▉ | 9858/11082 [31:41:59<3:54:43, 11.51s/step, acc=0.5088, loss=2.0117, lr=1.28e-05]
Training:  89%|████████▉ | 9859/11082 [31:42:11<3:55:38, 11.56s/step, acc=0.5088, loss=2.0117, lr=1.28e-05]
Training:  89%|████████▉ | 9860/11082 [31:42:23<3:56:04, 11.59s/step, acc=0.5088, loss=2.0117, lr=1.28e-05]
Training:  89%|████████▉ | 9861/11082 [31:42:34<3:55:06, 11.55s/step, acc=0.5088, loss=2.0117, lr=1.28e-05]
Training:  89%|████████▉ | 9862/11082 [31:42:46<3:55:16, 11.57s/step, acc=0.5088, loss=2.0117, lr=1.28e-05]
Training:  89%|████████▉ | 9863/11082 [31:42:58<3:56:26, 11.64s/step, acc=0.5088, loss=2.0117, lr=1.28e-05]
Training:  89%|████████▉ | 9864/11082 [31:43:09<3:54:20, 11.54s/step, acc=0.5088, loss=2.0117, lr=1.28e-05]
Training:  89%|████████▉ | 9865/11082 [31:43:21<3:56:55, 11.68s/step, acc=0.5088, loss=2.0117, lr=1.28e-05]
Training:  89%|████████▉ | 9866/11082 [31:43:32<3:53:20, 11.51s/step, acc=0.5088, loss=2.0117, lr=1.28e-05]
Training:  89%|████████▉ | 9867/11082 [31:43:44<3:57:09, 11.71s/step, acc=0.5088, loss=2.0117, lr=1.28e-05]
Training:  89%|████████▉ | 9868/11082 [31:43:56<3:56:46, 11.70s/step, acc=0.5088, loss=2.0117, lr=1.28e-05]
Training:  89%|████████▉ | 9869/11082 [31:44:07<3:53:51, 11.57s/step, acc=0.5088, loss=2.0117, lr=1.28e-05]
Training:  89%|████████▉ | 9870/11082 [31:44:19<3:54:49, 11.62s/step, acc=0.5088, loss=2.0117, lr=1.28e-05]
Training:  89%|████████▉ | 9871/11082 [31:44:31<3:54:06, 11.60s/step, acc=0.5088, loss=2.0117, lr=1.28e-05]
Training:  89%|████████▉ | 9872/11082 [31:44:42<3:54:34, 11.63s/step, acc=0.5088, loss=2.0117, lr=1.28e-05]
Training:  89%|████████▉ | 9873/11082 [31:44:54<3:52:23, 11.53s/step, acc=0.5088, loss=2.0117, lr=1.28e-05]
Training:  89%|████████▉ | 9874/11082 [31:45:05<3:54:40, 11.66s/step, acc=0.5088, loss=2.0117, lr=1.28e-05]
Training:  89%|████████▉ | 9875/11082 [31:45:17<3:54:56, 11.68s/step, acc=0.5088, loss=2.0117, lr=1.28e-05]
Training:  89%|████████▉ | 9876/11082 [31:45:28<3:51:58, 11.54s/step, acc=0.5088, loss=2.0117, lr=1.28e-05]
Training:  89%|████████▉ | 9877/11082 [31:45:40<3:52:50, 11.59s/step, acc=0.5088, loss=2.0117, lr=1.28e-05]
Training:  89%|████████▉ | 9878/11082 [31:45:52<3:53:56, 11.66s/step, acc=0.5088, loss=2.0117, lr=1.28e-05]
Training:  89%|████████▉ | 9879/11082 [31:46:03<3:53:01, 11.62s/step, acc=0.5088, loss=2.0117, lr=1.28e-05]
Training:  89%|████████▉ | 9880/11082 [31:46:15<3:50:56, 11.53s/step, acc=0.5088, loss=2.0117, lr=1.28e-05]
Training:  89%|████████▉ | 9881/11082 [31:46:26<3:51:21, 11.56s/step, acc=0.5088, loss=2.0117, lr=1.28e-05]
Training:  89%|████████▉ | 9882/11082 [31:46:38<3:53:21, 11.67s/step, acc=0.5088, loss=2.0117, lr=1.28e-05]
Training:  89%|████████▉ | 9883/11082 [31:46:50<3:52:54, 11.66s/step, acc=0.5088, loss=2.0117, lr=1.28e-05]
Training:  89%|████████▉ | 9884/11082 [31:47:01<3:50:32, 11.55s/step, acc=0.5088, loss=2.0117, lr=1.28e-05]
Training:  89%|████████▉ | 9885/11082 [31:47:14<3:55:05, 11.78s/step, acc=0.5088, loss=2.0117, lr=1.28e-05]
Training:  89%|████████▉ | 9886/11082 [31:47:25<3:51:14, 11.60s/step, acc=0.5088, loss=2.0117, lr=1.28e-05]
Training:  89%|████████▉ | 9887/11082 [31:47:36<3:47:48, 11.44s/step, acc=0.5088, loss=2.0117, lr=1.28e-05]
Training:  89%|████████▉ | 9888/11082 [31:47:48<3:49:21, 11.53s/step, acc=0.5088, loss=2.0117, lr=1.28e-05]
Training:  89%|████████▉ | 9889/11082 [31:48:00<3:52:45, 11.71s/step, acc=0.5088, loss=2.0117, lr=1.28e-05]
Training:  89%|████████▉ | 9890/11082 [31:48:11<3:47:22, 11.44s/step, acc=0.5088, loss=2.0117, lr=1.28e-05]
Training:  89%|████████▉ | 9891/11082 [31:48:22<3:49:22, 11.56s/step, acc=0.5088, loss=2.0117, lr=1.28e-05]
Training:  89%|████████▉ | 9892/11082 [31:48:34<3:50:25, 11.62s/step, acc=0.5088, loss=2.0117, lr=1.28e-05]
Training:  89%|████████▉ | 9893/11082 [31:48:46<3:49:49, 11.60s/step, acc=0.5088, loss=2.0117, lr=1.28e-05]
Training:  89%|████████▉ | 9894/11082 [31:48:57<3:49:31, 11.59s/step, acc=0.5088, loss=2.0117, lr=1.28e-05]
Training:  89%|████████▉ | 9895/11082 [31:49:09<3:49:15, 11.59s/step, acc=0.5088, loss=2.0117, lr=1.28e-05]
Training:  89%|████████▉ | 9896/11082 [31:49:21<3:49:51, 11.63s/step, acc=0.5088, loss=2.0117, lr=1.28e-05]
Training:  89%|████████▉ | 9897/11082 [31:49:32<3:48:21, 11.56s/step, acc=0.5088, loss=2.0117, lr=1.28e-05]
Training:  89%|████████▉ | 9898/11082 [31:49:43<3:47:15, 11.52s/step, acc=0.5088, loss=2.0117, lr=1.28e-05]
Training:  89%|████████▉ | 9899/11082 [31:49:55<3:49:35, 11.64s/step, acc=0.5088, loss=2.0117, lr=1.28e-05]
Training:  89%|████████▉ | 9900/11082 [31:50:07<3:50:03, 11.68s/step, acc=0.5088, loss=2.0117, lr=1.28e-05]
Training:  89%|████████▉ | 9900/11082 [31:50:14<3:50:03, 11.68s/step, acc=0.5065, loss=2.0205, lr=1.26e-05]2026-06-19 22:43:42 | INFO | __main__ | epoch=1 step=9900 train_loss=2.020453 train_acc=0.506489 lr=1.258e-05 grad_norm=0.4902 acc_d0=0.7642 acc_d1=0.6200 acc_d2=0.4981 acc_d3=0.4008 acc_d4=0.3223 acc_d5=0.2589 acc_d6=0.2095 acc_d7=0.1718

Training:  89%|████████▉ | 9901/11082 [31:50:18<3:47:52, 11.58s/step, acc=0.5065, loss=2.0205, lr=1.26e-05]
Training:  89%|████████▉ | 9902/11082 [31:50:30<3:47:56, 11.59s/step, acc=0.5065, loss=2.0205, lr=1.26e-05]
Training:  89%|████████▉ | 9903/11082 [31:50:42<3:48:52, 11.65s/step, acc=0.5065, loss=2.0205, lr=1.26e-05]
Training:  89%|████████▉ | 9904/11082 [31:50:53<3:46:41, 11.55s/step, acc=0.5065, loss=2.0205, lr=1.26e-05]
Training:  89%|████████▉ | 9905/11082 [31:51:05<3:48:09, 11.63s/step, acc=0.5065, loss=2.0205, lr=1.26e-05]
Training:  89%|████████▉ | 9906/11082 [31:51:16<3:41:45, 11.31s/step, acc=0.5065, loss=2.0205, lr=1.26e-05]
Training:  89%|████████▉ | 9907/11082 [31:51:28<3:47:08, 11.60s/step, acc=0.5065, loss=2.0205, lr=1.26e-05]
Training:  89%|████████▉ | 9908/11082 [31:51:39<3:46:28, 11.57s/step, acc=0.5065, loss=2.0205, lr=1.26e-05]
Training:  89%|████████▉ | 9909/11082 [31:51:51<3:47:16, 11.63s/step, acc=0.5065, loss=2.0205, lr=1.26e-05]
Training:  89%|████████▉ | 9910/11082 [31:52:03<3:46:09, 11.58s/step, acc=0.5065, loss=2.0205, lr=1.26e-05]
Training:  89%|████████▉ | 9911/11082 [31:52:14<3:44:03, 11.48s/step, acc=0.5065, loss=2.0205, lr=1.26e-05]
Training:  89%|████████▉ | 9912/11082 [31:52:25<3:43:27, 11.46s/step, acc=0.5065, loss=2.0205, lr=1.26e-05]
Training:  89%|████████▉ | 9913/11082 [31:52:37<3:43:41, 11.48s/step, acc=0.5065, loss=2.0205, lr=1.26e-05]
Training:  89%|████████▉ | 9914/11082 [31:52:49<3:46:58, 11.66s/step, acc=0.5065, loss=2.0205, lr=1.26e-05]
Training:  89%|████████▉ | 9915/11082 [31:53:00<3:43:41, 11.50s/step, acc=0.5065, loss=2.0205, lr=1.26e-05]
Training:  89%|████████▉ | 9916/11082 [31:53:11<3:43:08, 11.48s/step, acc=0.5065, loss=2.0205, lr=1.26e-05]
Training:  89%|████████▉ | 9917/11082 [31:53:23<3:42:49, 11.48s/step, acc=0.5065, loss=2.0205, lr=1.26e-05]
Training:  89%|████████▉ | 9918/11082 [31:53:35<3:44:10, 11.56s/step, acc=0.5065, loss=2.0205, lr=1.26e-05]
Training:  90%|████████▉ | 9919/11082 [31:53:46<3:44:05, 11.56s/step, acc=0.5065, loss=2.0205, lr=1.26e-05]
Training:  90%|████████▉ | 9920/11082 [31:53:58<3:45:22, 11.64s/step, acc=0.5065, loss=2.0205, lr=1.26e-05]
Training:  90%|████████▉ | 9921/11082 [31:54:09<3:44:19, 11.59s/step, acc=0.5065, loss=2.0205, lr=1.26e-05]
Training:  90%|████████▉ | 9922/11082 [31:54:21<3:42:50, 11.53s/step, acc=0.5065, loss=2.0205, lr=1.26e-05]
Training:  90%|████████▉ | 9923/11082 [31:54:32<3:42:44, 11.53s/step, acc=0.5065, loss=2.0205, lr=1.26e-05]
Training:  90%|████████▉ | 9924/11082 [31:54:44<3:40:45, 11.44s/step, acc=0.5065, loss=2.0205, lr=1.26e-05]
Training:  90%|████████▉ | 9925/11082 [31:54:55<3:40:26, 11.43s/step, acc=0.5065, loss=2.0205, lr=1.26e-05]
Training:  90%|████████▉ | 9926/11082 [31:55:07<3:41:08, 11.48s/step, acc=0.5065, loss=2.0205, lr=1.26e-05]
Training:  90%|████████▉ | 9927/11082 [31:55:18<3:42:09, 11.54s/step, acc=0.5065, loss=2.0205, lr=1.26e-05]
Training:  90%|████████▉ | 9928/11082 [31:55:30<3:44:09, 11.65s/step, acc=0.5065, loss=2.0205, lr=1.26e-05]
Training:  90%|████████▉ | 9929/11082 [31:55:41<3:40:01, 11.45s/step, acc=0.5065, loss=2.0205, lr=1.26e-05]
Training:  90%|████████▉ | 9930/11082 [31:55:53<3:43:13, 11.63s/step, acc=0.5065, loss=2.0205, lr=1.26e-05]
Training:  90%|████████▉ | 9931/11082 [31:56:05<3:43:22, 11.64s/step, acc=0.5065, loss=2.0205, lr=1.26e-05]
Training:  90%|████████▉ | 9932/11082 [31:56:16<3:41:27, 11.55s/step, acc=0.5065, loss=2.0205, lr=1.26e-05]
Training:  90%|████████▉ | 9933/11082 [31:56:27<3:39:39, 11.47s/step, acc=0.5065, loss=2.0205, lr=1.26e-05]
Training:  90%|████████▉ | 9934/11082 [31:56:39<3:41:25, 11.57s/step, acc=0.5065, loss=2.0205, lr=1.26e-05]
Training:  90%|████████▉ | 9935/11082 [31:56:51<3:41:29, 11.59s/step, acc=0.5065, loss=2.0205, lr=1.26e-05]
Training:  90%|████████▉ | 9936/11082 [31:57:02<3:35:34, 11.29s/step, acc=0.5065, loss=2.0205, lr=1.26e-05]
Training:  90%|████████▉ | 9937/11082 [31:57:14<3:42:03, 11.64s/step, acc=0.5065, loss=2.0205, lr=1.26e-05]
Training:  90%|████████▉ | 9938/11082 [31:57:25<3:40:18, 11.55s/step, acc=0.5065, loss=2.0205, lr=1.26e-05]
Training:  90%|████████▉ | 9939/11082 [31:57:37<3:41:10, 11.61s/step, acc=0.5065, loss=2.0205, lr=1.26e-05]
Training:  90%|████████▉ | 9940/11082 [31:57:48<3:38:26, 11.48s/step, acc=0.5065, loss=2.0205, lr=1.26e-05]
Training:  90%|████████▉ | 9941/11082 [31:58:00<3:37:30, 11.44s/step, acc=0.5065, loss=2.0205, lr=1.26e-05]
Training:  90%|████████▉ | 9942/11082 [31:58:12<3:41:13, 11.64s/step, acc=0.5065, loss=2.0205, lr=1.26e-05]
Training:  90%|████████▉ | 9943/11082 [31:58:23<3:38:34, 11.51s/step, acc=0.5065, loss=2.0205, lr=1.26e-05]
Training:  90%|████████▉ | 9944/11082 [31:58:35<3:39:13, 11.56s/step, acc=0.5065, loss=2.0205, lr=1.26e-05]
Training:  90%|████████▉ | 9945/11082 [31:58:46<3:37:15, 11.46s/step, acc=0.5065, loss=2.0205, lr=1.26e-05]
Training:  90%|████████▉ | 9946/11082 [31:58:58<3:39:11, 11.58s/step, acc=0.5065, loss=2.0205, lr=1.26e-05]
Training:  90%|████████▉ | 9947/11082 [31:59:09<3:38:07, 11.53s/step, acc=0.5065, loss=2.0205, lr=1.26e-05]
Training:  90%|████████▉ | 9948/11082 [31:59:21<3:39:19, 11.60s/step, acc=0.5065, loss=2.0205, lr=1.26e-05]
Training:  90%|████████▉ | 9949/11082 [31:59:32<3:38:25, 11.57s/step, acc=0.5065, loss=2.0205, lr=1.26e-05]
Training:  90%|████████▉ | 9950/11082 [31:59:44<3:37:28, 11.53s/step, acc=0.5065, loss=2.0205, lr=1.26e-05]
Training:  90%|████████▉ | 9950/11082 [31:59:51<3:37:28, 11.53s/step, acc=0.5084, loss=2.0109, lr=1.24e-05]2026-06-19 22:53:19 | INFO | __main__ | epoch=1 step=9950 train_loss=2.010915 train_acc=0.508406 lr=1.237e-05 grad_norm=0.4980 acc_d0=0.7681 acc_d1=0.6269 acc_d2=0.5080 acc_d3=0.4112 acc_d4=0.3312 acc_d5=0.2684 acc_d6=0.2170 acc_d7=0.1783

Training:  90%|████████▉ | 9951/11082 [31:59:55<3:37:59, 11.56s/step, acc=0.5084, loss=2.0109, lr=1.24e-05]
Training:  90%|████████▉ | 9952/11082 [32:00:07<3:37:34, 11.55s/step, acc=0.5084, loss=2.0109, lr=1.24e-05]
Training:  90%|████████▉ | 9953/11082 [32:00:19<3:37:50, 11.58s/step, acc=0.5084, loss=2.0109, lr=1.24e-05]
Training:  90%|████████▉ | 9954/11082 [32:00:30<3:38:32, 11.62s/step, acc=0.5084, loss=2.0109, lr=1.24e-05]
Training:  90%|████████▉ | 9955/11082 [32:00:42<3:39:05, 11.66s/step, acc=0.5084, loss=2.0109, lr=1.24e-05]
Training:  90%|████████▉ | 9956/11082 [32:00:54<3:40:07, 11.73s/step, acc=0.5084, loss=2.0109, lr=1.24e-05]
Training:  90%|████████▉ | 9957/11082 [32:01:05<3:34:14, 11.43s/step, acc=0.5084, loss=2.0109, lr=1.24e-05]
Training:  90%|████████▉ | 9958/11082 [32:01:17<3:37:59, 11.64s/step, acc=0.5084, loss=2.0109, lr=1.24e-05]
Training:  90%|████████▉ | 9959/11082 [32:01:28<3:37:25, 11.62s/step, acc=0.5084, loss=2.0109, lr=1.24e-05]
Training:  90%|████████▉ | 9960/11082 [32:01:40<3:36:55, 11.60s/step, acc=0.5084, loss=2.0109, lr=1.24e-05]
Training:  90%|████████▉ | 9961/11082 [32:01:52<3:37:03, 11.62s/step, acc=0.5084, loss=2.0109, lr=1.24e-05]
Training:  90%|████████▉ | 9962/11082 [32:02:03<3:36:25, 11.59s/step, acc=0.5084, loss=2.0109, lr=1.24e-05]
Training:  90%|████████▉ | 9963/11082 [32:02:15<3:36:48, 11.63s/step, acc=0.5084, loss=2.0109, lr=1.24e-05]
Training:  90%|████████▉ | 9964/11082 [32:02:26<3:35:45, 11.58s/step, acc=0.5084, loss=2.0109, lr=1.24e-05]
Training:  90%|████████▉ | 9965/11082 [32:02:37<3:33:08, 11.45s/step, acc=0.5084, loss=2.0109, lr=1.24e-05]
Training:  90%|████████▉ | 9966/11082 [32:02:49<3:35:23, 11.58s/step, acc=0.5084, loss=2.0109, lr=1.24e-05]
Training:  90%|████████▉ | 9967/11082 [32:03:01<3:35:06, 11.58s/step, acc=0.5084, loss=2.0109, lr=1.24e-05]
Training:  90%|████████▉ | 9968/11082 [32:03:12<3:33:27, 11.50s/step, acc=0.5084, loss=2.0109, lr=1.24e-05]
Training:  90%|████████▉ | 9969/11082 [32:03:24<3:33:08, 11.49s/step, acc=0.5084, loss=2.0109, lr=1.24e-05]
Training:  90%|████████▉ | 9970/11082 [32:03:35<3:32:00, 11.44s/step, acc=0.5084, loss=2.0109, lr=1.24e-05]
Training:  90%|████████▉ | 9971/11082 [32:03:47<3:34:19, 11.57s/step, acc=0.5084, loss=2.0109, lr=1.24e-05]
Training:  90%|████████▉ | 9972/11082 [32:03:58<3:34:00, 11.57s/step, acc=0.5084, loss=2.0109, lr=1.24e-05]
Training:  90%|████████▉ | 9973/11082 [32:04:10<3:34:10, 11.59s/step, acc=0.5084, loss=2.0109, lr=1.24e-05]
Training:  90%|█████████ | 9974/11082 [32:04:21<3:32:36, 11.51s/step, acc=0.5084, loss=2.0109, lr=1.24e-05]
Training:  90%|█████████ | 9975/11082 [32:04:33<3:33:49, 11.59s/step, acc=0.5084, loss=2.0109, lr=1.24e-05]
Training:  90%|█████████ | 9976/11082 [32:04:45<3:32:10, 11.51s/step, acc=0.5084, loss=2.0109, lr=1.24e-05]
Training:  90%|█████████ | 9977/11082 [32:04:56<3:33:59, 11.62s/step, acc=0.5084, loss=2.0109, lr=1.24e-05]
Training:  90%|█████████ | 9978/11082 [32:05:08<3:31:00, 11.47s/step, acc=0.5084, loss=2.0109, lr=1.24e-05]
Training:  90%|█████████ | 9979/11082 [32:05:20<3:35:42, 11.73s/step, acc=0.5084, loss=2.0109, lr=1.24e-05]
Training:  90%|█████████ | 9980/11082 [32:05:31<3:32:45, 11.58s/step, acc=0.5084, loss=2.0109, lr=1.24e-05]
Training:  90%|█████████ | 9981/11082 [32:05:43<3:31:59, 11.55s/step, acc=0.5084, loss=2.0109, lr=1.24e-05]
Training:  90%|█████████ | 9982/11082 [32:05:54<3:33:30, 11.65s/step, acc=0.5084, loss=2.0109, lr=1.24e-05]
Training:  90%|█████████ | 9983/11082 [32:06:06<3:32:30, 11.60s/step, acc=0.5084, loss=2.0109, lr=1.24e-05]
Training:  90%|█████████ | 9984/11082 [32:06:17<3:31:14, 11.54s/step, acc=0.5084, loss=2.0109, lr=1.24e-05]
Training:  90%|█████████ | 9985/11082 [32:06:29<3:31:00, 11.54s/step, acc=0.5084, loss=2.0109, lr=1.24e-05]
Training:  90%|█████████ | 9986/11082 [32:06:40<3:30:33, 11.53s/step, acc=0.5084, loss=2.0109, lr=1.24e-05]
Training:  90%|█████████ | 9987/11082 [32:06:52<3:28:59, 11.45s/step, acc=0.5084, loss=2.0109, lr=1.24e-05]
Training:  90%|█████████ | 9988/11082 [32:07:04<3:31:15, 11.59s/step, acc=0.5084, loss=2.0109, lr=1.24e-05]
Training:  90%|█████████ | 9989/11082 [32:07:15<3:29:18, 11.49s/step, acc=0.5084, loss=2.0109, lr=1.24e-05]
Training:  90%|█████████ | 9990/11082 [32:07:27<3:31:11, 11.60s/step, acc=0.5084, loss=2.0109, lr=1.24e-05]
Training:  90%|█████████ | 9991/11082 [32:07:38<3:31:35, 11.64s/step, acc=0.5084, loss=2.0109, lr=1.24e-05]
Training:  90%|█████████ | 9992/11082 [32:07:50<3:30:48, 11.60s/step, acc=0.5084, loss=2.0109, lr=1.24e-05]
Training:  90%|█████████ | 9993/11082 [32:08:01<3:28:44, 11.50s/step, acc=0.5084, loss=2.0109, lr=1.24e-05]
Training:  90%|█████████ | 9994/11082 [32:08:13<3:29:16, 11.54s/step, acc=0.5084, loss=2.0109, lr=1.24e-05]
Training:  90%|█████████ | 9995/11082 [32:08:24<3:26:09, 11.38s/step, acc=0.5084, loss=2.0109, lr=1.24e-05]
Training:  90%|█████████ | 9996/11082 [32:08:36<3:29:03, 11.55s/step, acc=0.5084, loss=2.0109, lr=1.24e-05]
Training:  90%|█████████ | 9997/11082 [32:08:47<3:28:52, 11.55s/step, acc=0.5084, loss=2.0109, lr=1.24e-05]
Training:  90%|█████████ | 9998/11082 [32:08:59<3:27:29, 11.48s/step, acc=0.5084, loss=2.0109, lr=1.24e-05]
Training:  90%|█████████ | 9999/11082 [32:09:11<3:29:33, 11.61s/step, acc=0.5084, loss=2.0109, lr=1.24e-05]
Training:  90%|█████████ | 10000/11082 [32:09:22<3:27:39, 11.52s/step, acc=0.5084, loss=2.0109, lr=1.24e-05]2026-06-19 23:02:59 | INFO | nemo_automodel.components.checkpoint._backports.consolidate_hf_safetensors | Consolidating safetensors files from /llm-align/liuchonghan/runs/p_eagle_kv_reuse/checkpoints/epoch_1_step_10000/model to /llm-align/liuchonghan/runs/p_eagle_kv_reuse/checkpoints/epoch_1_step_10000/model/consolidated.
2026-06-19 23:03:05 | INFO | nemo_automodel.components.checkpoint.checkpointing | Successfully exported consolidated HF safetensors to /llm-align/liuchonghan/runs/p_eagle_kv_reuse/checkpoints/epoch_1_step_10000/model/consolidated.
2026-06-19 23:03:08 | INFO | __main__ | Saved step checkpoint to /llm-align/liuchonghan/runs/p_eagle_kv_reuse/checkpoints/epoch_1_step_10000

Training:  90%|█████████ | 10000/11082 [32:09:40<3:27:39, 11.52s/step, acc=0.5093, loss=2.0048, lr=1.22e-05]2026-06-19 23:03:08 | INFO | __main__ | epoch=1 step=10000 train_loss=2.004774 train_acc=0.509313 lr=1.216e-05 grad_norm=0.4824 acc_d0=0.7738 acc_d1=0.6332 acc_d2=0.5146 acc_d3=0.4173 acc_d4=0.3381 acc_d5=0.2746 acc_d6=0.2230 acc_d7=0.1814

Training:  90%|█████████ | 10001/11082 [32:09:44<4:27:31, 14.85s/step, acc=0.5093, loss=2.0048, lr=1.22e-05]
Training:  90%|█████████ | 10002/11082 [32:09:55<4:05:43, 13.65s/step, acc=0.5093, loss=2.0048, lr=1.22e-05]
Training:  90%|█████████ | 10003/11082 [32:10:08<3:58:27, 13.26s/step, acc=0.5093, loss=2.0048, lr=1.22e-05]
Training:  90%|█████████ | 10004/11082 [32:10:19<3:49:11, 12.76s/step, acc=0.5093, loss=2.0048, lr=1.22e-05]
Training:  90%|█████████ | 10005/11082 [32:10:31<3:41:09, 12.32s/step, acc=0.5093, loss=2.0048, lr=1.22e-05]
Training:  90%|█████████ | 10006/11082 [32:10:42<3:37:31, 12.13s/step, acc=0.5093, loss=2.0048, lr=1.22e-05]
Training:  90%|█████████ | 10007/11082 [32:10:54<3:33:26, 11.91s/step, acc=0.5093, loss=2.0048, lr=1.22e-05]
Training:  90%|█████████ | 10008/11082 [32:11:06<3:32:59, 11.90s/step, acc=0.5093, loss=2.0048, lr=1.22e-05]
Training:  90%|█████████ | 10009/11082 [32:11:17<3:32:41, 11.89s/step, acc=0.5093, loss=2.0048, lr=1.22e-05]
Training:  90%|█████████ | 10010/11082 [32:11:29<3:32:16, 11.88s/step, acc=0.5093, loss=2.0048, lr=1.22e-05]
Training:  90%|█████████ | 10011/11082 [32:11:40<3:27:32, 11.63s/step, acc=0.5093, loss=2.0048, lr=1.22e-05]
Training:  90%|█████████ | 10012/11082 [32:11:52<3:26:29, 11.58s/step, acc=0.5093, loss=2.0048, lr=1.22e-05]
Training:  90%|█████████ | 10013/11082 [32:12:04<3:27:31, 11.65s/step, acc=0.5093, loss=2.0048, lr=1.22e-05]
Training:  90%|█████████ | 10014/11082 [32:12:15<3:26:40, 11.61s/step, acc=0.5093, loss=2.0048, lr=1.22e-05]
Training:  90%|█████████ | 10015/11082 [32:12:27<3:27:18, 11.66s/step, acc=0.5093, loss=2.0048, lr=1.22e-05]
Training:  90%|█████████ | 10016/11082 [32:12:38<3:23:55, 11.48s/step, acc=0.5093, loss=2.0048, lr=1.22e-05]
Training:  90%|█████████ | 10017/11082 [32:12:50<3:26:57, 11.66s/step, acc=0.5093, loss=2.0048, lr=1.22e-05]
Training:  90%|█████████ | 10018/11082 [32:13:01<3:23:18, 11.46s/step, acc=0.5093, loss=2.0048, lr=1.22e-05]
Training:  90%|█████████ | 10019/11082 [32:13:13<3:26:11, 11.64s/step, acc=0.5093, loss=2.0048, lr=1.22e-05]
Training:  90%|█████████ | 10020/11082 [32:13:24<3:21:57, 11.41s/step, acc=0.5093, loss=2.0048, lr=1.22e-05]
Training:  90%|█████████ | 10021/11082 [32:13:36<3:23:07, 11.49s/step, acc=0.5093, loss=2.0048, lr=1.22e-05]
Training:  90%|█████████ | 10022/11082 [32:13:47<3:24:25, 11.57s/step, acc=0.5093, loss=2.0048, lr=1.22e-05]
Training:  90%|█████████ | 10023/11082 [32:13:59<3:25:40, 11.65s/step, acc=0.5093, loss=2.0048, lr=1.22e-05]
Training:  90%|█████████ | 10024/11082 [32:14:11<3:24:17, 11.59s/step, acc=0.5093, loss=2.0048, lr=1.22e-05]
Training:  90%|█████████ | 10025/11082 [32:14:23<3:26:49, 11.74s/step, acc=0.5093, loss=2.0048, lr=1.22e-05]
Training:  90%|█████████ | 10026/11082 [32:14:34<3:23:32, 11.56s/step, acc=0.5093, loss=2.0048, lr=1.22e-05]
Training:  90%|█████████ | 10027/11082 [32:14:46<3:23:39, 11.58s/step, acc=0.5093, loss=2.0048, lr=1.22e-05]
Training:  90%|█████████ | 10028/11082 [32:14:57<3:23:14, 11.57s/step, acc=0.5093, loss=2.0048, lr=1.22e-05]
Training:  90%|█████████ | 10029/11082 [32:15:07<3:17:00, 11.23s/step, acc=0.5093, loss=2.0048, lr=1.22e-05]
Training:  91%|█████████ | 10030/11082 [32:15:20<3:22:14, 11.53s/step, acc=0.5093, loss=2.0048, lr=1.22e-05]
Training:  91%|█████████ | 10031/11082 [32:15:31<3:21:41, 11.51s/step, acc=0.5093, loss=2.0048, lr=1.22e-05]
Training:  91%|█████████ | 10032/11082 [32:15:43<3:24:16, 11.67s/step, acc=0.5093, loss=2.0048, lr=1.22e-05]
Training:  91%|█████████ | 10033/11082 [32:15:55<3:24:18, 11.69s/step, acc=0.5093, loss=2.0048, lr=1.22e-05]
Training:  91%|█████████ | 10034/11082 [32:16:07<3:23:20, 11.64s/step, acc=0.5093, loss=2.0048, lr=1.22e-05]
Training:  91%|█████████ | 10035/11082 [32:16:18<3:20:31, 11.49s/step, acc=0.5093, loss=2.0048, lr=1.22e-05]
Training:  91%|█████████ | 10036/11082 [32:16:29<3:21:34, 11.56s/step, acc=0.5093, loss=2.0048, lr=1.22e-05]
Training:  91%|█████████ | 10037/11082 [32:16:41<3:24:03, 11.72s/step, acc=0.5093, loss=2.0048, lr=1.22e-05]
Training:  91%|█████████ | 10038/11082 [32:16:53<3:22:10, 11.62s/step, acc=0.5093, loss=2.0048, lr=1.22e-05]
Training:  91%|█████████ | 10039/11082 [32:17:04<3:22:07, 11.63s/step, acc=0.5093, loss=2.0048, lr=1.22e-05]
Training:  91%|█████████ | 10040/11082 [32:17:16<3:21:06, 11.58s/step, acc=0.5093, loss=2.0048, lr=1.22e-05]
Training:  91%|█████████ | 10041/11082 [32:17:27<3:20:10, 11.54s/step, acc=0.5093, loss=2.0048, lr=1.22e-05]
Training:  91%|█████████ | 10042/11082 [32:17:39<3:20:33, 11.57s/step, acc=0.5093, loss=2.0048, lr=1.22e-05]
Training:  91%|█████████ | 10043/11082 [32:17:51<3:20:25, 11.57s/step, acc=0.5093, loss=2.0048, lr=1.22e-05]
Training:  91%|█████████ | 10044/11082 [32:18:02<3:19:30, 11.53s/step, acc=0.5093, loss=2.0048, lr=1.22e-05]
Training:  91%|█████████ | 10045/11082 [32:18:13<3:18:36, 11.49s/step, acc=0.5093, loss=2.0048, lr=1.22e-05]
Training:  91%|█████████ | 10046/11082 [32:18:25<3:20:45, 11.63s/step, acc=0.5093, loss=2.0048, lr=1.22e-05]
Training:  91%|█████████ | 10047/11082 [32:18:37<3:19:15, 11.55s/step, acc=0.5093, loss=2.0048, lr=1.22e-05]
Training:  91%|█████████ | 10048/11082 [32:18:48<3:19:03, 11.55s/step, acc=0.5093, loss=2.0048, lr=1.22e-05]
Training:  91%|█████████ | 10049/11082 [32:19:00<3:19:03, 11.56s/step, acc=0.5093, loss=2.0048, lr=1.22e-05]
Training:  91%|█████████ | 10050/11082 [32:19:12<3:19:15, 11.58s/step, acc=0.5093, loss=2.0048, lr=1.22e-05]
Training:  91%|█████████ | 10050/11082 [32:19:18<3:19:15, 11.58s/step, acc=0.5072, loss=2.0200, lr=1.20e-05]2026-06-19 23:12:46 | INFO | __main__ | epoch=1 step=10050 train_loss=2.020016 train_acc=0.507243 lr=1.197e-05 grad_norm=0.4473 acc_d0=0.7630 acc_d1=0.6197 acc_d2=0.5006 acc_d3=0.4043 acc_d4=0.3283 acc_d5=0.2639 acc_d6=0.2138 acc_d7=0.1766

Training:  91%|█████████ | 10051/11082 [32:19:23<3:18:30, 11.55s/step, acc=0.5072, loss=2.0200, lr=1.20e-05]
Training:  91%|█████████ | 10052/11082 [32:19:34<3:16:34, 11.45s/step, acc=0.5072, loss=2.0200, lr=1.20e-05]
Training:  91%|█████████ | 10053/11082 [32:19:46<3:18:25, 11.57s/step, acc=0.5072, loss=2.0200, lr=1.20e-05]
Training:  91%|█████████ | 10054/11082 [32:19:58<3:17:35, 11.53s/step, acc=0.5072, loss=2.0200, lr=1.20e-05]
Training:  91%|█████████ | 10055/11082 [32:20:09<3:15:47, 11.44s/step, acc=0.5072, loss=2.0200, lr=1.20e-05]
Training:  91%|█████████ | 10056/11082 [32:20:20<3:16:38, 11.50s/step, acc=0.5072, loss=2.0200, lr=1.20e-05]
Training:  91%|█████████ | 10057/11082 [32:20:32<3:15:18, 11.43s/step, acc=0.5072, loss=2.0200, lr=1.20e-05]
Training:  91%|█████████ | 10058/11082 [32:20:44<3:17:43, 11.59s/step, acc=0.5072, loss=2.0200, lr=1.20e-05]
Training:  91%|█████████ | 10059/11082 [32:20:56<3:19:10, 11.68s/step, acc=0.5072, loss=2.0200, lr=1.20e-05]
Training:  91%|█████████ | 10060/11082 [32:21:07<3:18:13, 11.64s/step, acc=0.5072, loss=2.0200, lr=1.20e-05]
Training:  91%|█████████ | 10061/11082 [32:21:18<3:15:55, 11.51s/step, acc=0.5072, loss=2.0200, lr=1.20e-05]
Training:  91%|█████████ | 10062/11082 [32:21:30<3:15:18, 11.49s/step, acc=0.5072, loss=2.0200, lr=1.20e-05]
Training:  91%|█████████ | 10063/11082 [32:21:41<3:15:03, 11.49s/step, acc=0.5072, loss=2.0200, lr=1.20e-05]
Training:  91%|█████████ | 10064/11082 [32:21:53<3:15:40, 11.53s/step, acc=0.5072, loss=2.0200, lr=1.20e-05]
Training:  91%|█████████ | 10065/11082 [32:22:04<3:14:45, 11.49s/step, acc=0.5072, loss=2.0200, lr=1.20e-05]
Training:  91%|█████████ | 10066/11082 [32:22:16<3:14:27, 11.48s/step, acc=0.5072, loss=2.0200, lr=1.20e-05]
Training:  91%|█████████ | 10067/11082 [32:22:27<3:14:53, 11.52s/step, acc=0.5072, loss=2.0200, lr=1.20e-05]
Training:  91%|█████████ | 10068/11082 [32:22:39<3:14:39, 11.52s/step, acc=0.5072, loss=2.0200, lr=1.20e-05]
Training:  91%|█████████ | 10069/11082 [32:22:51<3:15:31, 11.58s/step, acc=0.5072, loss=2.0200, lr=1.20e-05]
Training:  91%|█████████ | 10070/11082 [32:23:02<3:13:05, 11.45s/step, acc=0.5072, loss=2.0200, lr=1.20e-05]
Training:  91%|█████████ | 10071/11082 [32:23:14<3:16:31, 11.66s/step, acc=0.5072, loss=2.0200, lr=1.20e-05]
Training:  91%|█████████ | 10072/11082 [32:23:25<3:14:55, 11.58s/step, acc=0.5072, loss=2.0200, lr=1.20e-05]
Training:  91%|█████████ | 10073/11082 [32:23:37<3:14:46, 11.58s/step, acc=0.5072, loss=2.0200, lr=1.20e-05]
Training:  91%|█████████ | 10074/11082 [32:23:48<3:11:52, 11.42s/step, acc=0.5072, loss=2.0200, lr=1.20e-05]
Training:  91%|█████████ | 10075/11082 [32:24:00<3:14:56, 11.62s/step, acc=0.5072, loss=2.0200, lr=1.20e-05]
Training:  91%|█████████ | 10076/11082 [32:24:11<3:13:48, 11.56s/step, acc=0.5072, loss=2.0200, lr=1.20e-05]
Training:  91%|█████████ | 10077/11082 [32:24:23<3:15:52, 11.69s/step, acc=0.5072, loss=2.0200, lr=1.20e-05]
Training:  91%|█████████ | 10078/11082 [32:24:34<3:12:08, 11.48s/step, acc=0.5072, loss=2.0200, lr=1.20e-05]
Training:  91%|█████████ | 10079/11082 [32:24:47<3:15:21, 11.69s/step, acc=0.5072, loss=2.0200, lr=1.20e-05]
Training:  91%|█████████ | 10080/11082 [32:24:58<3:14:24, 11.64s/step, acc=0.5072, loss=2.0200, lr=1.20e-05]
Training:  91%|█████████ | 10081/11082 [32:25:10<3:13:56, 11.63s/step, acc=0.5072, loss=2.0200, lr=1.20e-05]
Training:  91%|█████████ | 10082/11082 [32:25:21<3:14:47, 11.69s/step, acc=0.5072, loss=2.0200, lr=1.20e-05]
Training:  91%|█████████ | 10083/11082 [32:25:33<3:11:45, 11.52s/step, acc=0.5072, loss=2.0200, lr=1.20e-05]
Training:  91%|█████████ | 10084/11082 [32:25:44<3:11:14, 11.50s/step, acc=0.5072, loss=2.0200, lr=1.20e-05]
Training:  91%|█████████ | 10085/11082 [32:25:55<3:10:24, 11.46s/step, acc=0.5072, loss=2.0200, lr=1.20e-05]
Training:  91%|█████████ | 10086/11082 [32:26:07<3:11:10, 11.52s/step, acc=0.5072, loss=2.0200, lr=1.20e-05]
Training:  91%|█████████ | 10087/11082 [32:26:19<3:11:54, 11.57s/step, acc=0.5072, loss=2.0200, lr=1.20e-05]
Training:  91%|█████████ | 10088/11082 [32:26:30<3:11:17, 11.55s/step, acc=0.5072, loss=2.0200, lr=1.20e-05]
Training:  91%|█████████ | 10089/11082 [32:26:42<3:10:40, 11.52s/step, acc=0.5072, loss=2.0200, lr=1.20e-05]
Training:  91%|█████████ | 10090/11082 [32:26:53<3:10:44, 11.54s/step, acc=0.5072, loss=2.0200, lr=1.20e-05]
Training:  91%|█████████ | 10091/11082 [32:27:05<3:10:05, 11.51s/step, acc=0.5072, loss=2.0200, lr=1.20e-05]
Training:  91%|█████████ | 10092/11082 [32:27:16<3:10:23, 11.54s/step, acc=0.5072, loss=2.0200, lr=1.20e-05]
Training:  91%|█████████ | 10093/11082 [32:27:28<3:12:42, 11.69s/step, acc=0.5072, loss=2.0200, lr=1.20e-05]
Training:  91%|█████████ | 10094/11082 [32:27:40<3:10:30, 11.57s/step, acc=0.5072, loss=2.0200, lr=1.20e-05]
Training:  91%|█████████ | 10095/11082 [32:27:51<3:09:28, 11.52s/step, acc=0.5072, loss=2.0200, lr=1.20e-05]
Training:  91%|█████████ | 10096/11082 [32:28:03<3:11:06, 11.63s/step, acc=0.5072, loss=2.0200, lr=1.20e-05]
Training:  91%|█████████ | 10097/11082 [32:28:14<3:09:19, 11.53s/step, acc=0.5072, loss=2.0200, lr=1.20e-05]
Training:  91%|█████████ | 10098/11082 [32:28:26<3:10:08, 11.59s/step, acc=0.5072, loss=2.0200, lr=1.20e-05]
Training:  91%|█████████ | 10099/11082 [32:28:37<3:07:57, 11.47s/step, acc=0.5072, loss=2.0200, lr=1.20e-05]
Training:  91%|█████████ | 10100/11082 [32:28:49<3:11:34, 11.70s/step, acc=0.5072, loss=2.0200, lr=1.20e-05]
Training:  91%|█████████ | 10100/11082 [32:28:57<3:11:34, 11.70s/step, acc=0.5063, loss=2.0235, lr=1.18e-05]2026-06-19 23:22:24 | INFO | __main__ | epoch=1 step=10100 train_loss=2.023454 train_acc=0.506348 lr=1.179e-05 grad_norm=0.4922 acc_d0=0.7638 acc_d1=0.6191 acc_d2=0.4984 acc_d3=0.4011 acc_d4=0.3255 acc_d5=0.2628 acc_d6=0.2129 acc_d7=0.1751

Training:  91%|█████████ | 10101/11082 [32:29:01<3:12:30, 11.77s/step, acc=0.5063, loss=2.0235, lr=1.18e-05]
Training:  91%|█████████ | 10102/11082 [32:29:13<3:09:09, 11.58s/step, acc=0.5063, loss=2.0235, lr=1.18e-05]
Training:  91%|█████████ | 10103/11082 [32:29:24<3:07:01, 11.46s/step, acc=0.5063, loss=2.0235, lr=1.18e-05]
Training:  91%|█████████ | 10104/11082 [32:29:36<3:12:37, 11.82s/step, acc=0.5063, loss=2.0235, lr=1.18e-05]
Training:  91%|█████████ | 10105/11082 [32:29:48<3:10:45, 11.72s/step, acc=0.5063, loss=2.0235, lr=1.18e-05]
Training:  91%|█████████ | 10106/11082 [32:30:00<3:10:49, 11.73s/step, acc=0.5063, loss=2.0235, lr=1.18e-05]
Training:  91%|█████████ | 10107/11082 [32:30:11<3:08:13, 11.58s/step, acc=0.5063, loss=2.0235, lr=1.18e-05]
Training:  91%|█████████ | 10108/11082 [32:30:23<3:09:39, 11.68s/step, acc=0.5063, loss=2.0235, lr=1.18e-05]
Training:  91%|█████████ | 10109/11082 [32:30:35<3:10:28, 11.75s/step, acc=0.5063, loss=2.0235, lr=1.18e-05]
Training:  91%|█████████ | 10110/11082 [32:30:46<3:06:42, 11.53s/step, acc=0.5063, loss=2.0235, lr=1.18e-05]
Training:  91%|█████████ | 10111/11082 [32:30:58<3:08:41, 11.66s/step, acc=0.5063, loss=2.0235, lr=1.18e-05]
Training:  91%|█████████ | 10112/11082 [32:31:09<3:06:17, 11.52s/step, acc=0.5063, loss=2.0235, lr=1.18e-05]
Training:  91%|█████████▏| 10113/11082 [32:31:20<3:05:56, 11.51s/step, acc=0.5063, loss=2.0235, lr=1.18e-05]
Training:  91%|█████████▏| 10114/11082 [32:31:32<3:06:59, 11.59s/step, acc=0.5063, loss=2.0235, lr=1.18e-05]
Training:  91%|█████████▏| 10115/11082 [32:31:44<3:06:47, 11.59s/step, acc=0.5063, loss=2.0235, lr=1.18e-05]
Training:  91%|█████████▏| 10116/11082 [32:31:55<3:06:07, 11.56s/step, acc=0.5063, loss=2.0235, lr=1.18e-05]
Training:  91%|█████████▏| 10117/11082 [32:32:07<3:08:00, 11.69s/step, acc=0.5063, loss=2.0235, lr=1.18e-05]
Training:  91%|█████████▏| 10118/11082 [32:32:19<3:07:21, 11.66s/step, acc=0.5063, loss=2.0235, lr=1.18e-05]
Training:  91%|█████████▏| 10119/11082 [32:32:30<3:06:22, 11.61s/step, acc=0.5063, loss=2.0235, lr=1.18e-05]
Training:  91%|█████████▏| 10120/11082 [32:32:42<3:08:29, 11.76s/step, acc=0.5063, loss=2.0235, lr=1.18e-05]
Training:  91%|█████████▏| 10121/11082 [32:32:54<3:06:30, 11.64s/step, acc=0.5063, loss=2.0235, lr=1.18e-05]
Training:  91%|█████████▏| 10122/11082 [32:33:06<3:07:08, 11.70s/step, acc=0.5063, loss=2.0235, lr=1.18e-05]
Training:  91%|█████████▏| 10123/11082 [32:33:17<3:04:47, 11.56s/step, acc=0.5063, loss=2.0235, lr=1.18e-05]
Training:  91%|█████████▏| 10124/11082 [32:33:29<3:06:14, 11.66s/step, acc=0.5063, loss=2.0235, lr=1.18e-05]
Training:  91%|█████████▏| 10125/11082 [32:33:40<3:05:01, 11.60s/step, acc=0.5063, loss=2.0235, lr=1.18e-05]
Training:  91%|█████████▏| 10126/11082 [32:33:52<3:06:08, 11.68s/step, acc=0.5063, loss=2.0235, lr=1.18e-05]
Training:  91%|█████████▏| 10127/11082 [32:34:03<3:04:52, 11.62s/step, acc=0.5063, loss=2.0235, lr=1.18e-05]
Training:  91%|█████████▏| 10128/11082 [32:34:15<3:03:29, 11.54s/step, acc=0.5063, loss=2.0235, lr=1.18e-05]
Training:  91%|█████████▏| 10129/11082 [32:34:26<3:03:36, 11.56s/step, acc=0.5063, loss=2.0235, lr=1.18e-05]
Training:  91%|█████████▏| 10130/11082 [32:34:38<3:03:15, 11.55s/step, acc=0.5063, loss=2.0235, lr=1.18e-05]
Training:  91%|█████████▏| 10131/11082 [32:34:50<3:03:10, 11.56s/step, acc=0.5063, loss=2.0235, lr=1.18e-05]
Training:  91%|█████████▏| 10132/11082 [32:35:01<3:03:44, 11.60s/step, acc=0.5063, loss=2.0235, lr=1.18e-05]
Training:  91%|█████████▏| 10133/11082 [32:35:13<3:03:37, 11.61s/step, acc=0.5063, loss=2.0235, lr=1.18e-05]
Training:  91%|█████████▏| 10134/11082 [32:35:25<3:04:01, 11.65s/step, acc=0.5063, loss=2.0235, lr=1.18e-05]
Training:  91%|█████████▏| 10135/11082 [32:35:36<3:03:37, 11.63s/step, acc=0.5063, loss=2.0235, lr=1.18e-05]
Training:  91%|█████████▏| 10136/11082 [32:35:48<3:01:48, 11.53s/step, acc=0.5063, loss=2.0235, lr=1.18e-05]
Training:  91%|█████████▏| 10137/11082 [32:35:59<3:03:07, 11.63s/step, acc=0.5063, loss=2.0235, lr=1.18e-05]
Training:  91%|█████████▏| 10138/11082 [32:36:11<3:02:37, 11.61s/step, acc=0.5063, loss=2.0235, lr=1.18e-05]
Training:  91%|█████████▏| 10139/11082 [32:36:22<3:01:28, 11.55s/step, acc=0.5063, loss=2.0235, lr=1.18e-05]
Training:  91%|█████████▏| 10140/11082 [32:36:34<3:01:16, 11.55s/step, acc=0.5063, loss=2.0235, lr=1.18e-05]
Training:  92%|█████████▏| 10141/11082 [32:36:45<3:01:10, 11.55s/step, acc=0.5063, loss=2.0235, lr=1.18e-05]
Training:  92%|█████████▏| 10142/11082 [32:36:57<2:59:51, 11.48s/step, acc=0.5063, loss=2.0235, lr=1.18e-05]
Training:  92%|█████████▏| 10143/11082 [32:37:09<3:02:24, 11.66s/step, acc=0.5063, loss=2.0235, lr=1.18e-05]
Training:  92%|█████████▏| 10144/11082 [32:37:20<3:01:47, 11.63s/step, acc=0.5063, loss=2.0235, lr=1.18e-05]
Training:  92%|█████████▏| 10145/11082 [32:37:32<3:01:09, 11.60s/step, acc=0.5063, loss=2.0235, lr=1.18e-05]
Training:  92%|█████████▏| 10146/11082 [32:37:43<3:00:28, 11.57s/step, acc=0.5063, loss=2.0235, lr=1.18e-05]
Training:  92%|█████████▏| 10147/11082 [32:37:55<3:00:11, 11.56s/step, acc=0.5063, loss=2.0235, lr=1.18e-05]
Training:  92%|█████████▏| 10148/11082 [32:38:07<2:59:57, 11.56s/step, acc=0.5063, loss=2.0235, lr=1.18e-05]
Training:  92%|█████████▏| 10149/11082 [32:38:18<3:01:15, 11.66s/step, acc=0.5063, loss=2.0235, lr=1.18e-05]
Training:  92%|█████████▏| 10150/11082 [32:38:30<2:59:06, 11.53s/step, acc=0.5063, loss=2.0235, lr=1.18e-05]
Training:  92%|█████████▏| 10150/11082 [32:38:37<2:59:06, 11.53s/step, acc=0.5071, loss=2.0174, lr=1.16e-05]2026-06-19 23:32:05 | INFO | __main__ | epoch=1 step=10150 train_loss=2.017352 train_acc=0.507142 lr=1.161e-05 grad_norm=0.4902 acc_d0=0.7728 acc_d1=0.6322 acc_d2=0.5112 acc_d3=0.4132 acc_d4=0.3329 acc_d5=0.2691 acc_d6=0.2184 acc_d7=0.1791

Training:  92%|█████████▏| 10151/11082 [32:38:41<2:59:55, 11.60s/step, acc=0.5071, loss=2.0174, lr=1.16e-05]
Training:  92%|█████████▏| 10152/11082 [32:38:53<2:58:34, 11.52s/step, acc=0.5071, loss=2.0174, lr=1.16e-05]
Training:  92%|█████████▏| 10153/11082 [32:39:04<2:58:15, 11.51s/step, acc=0.5071, loss=2.0174, lr=1.16e-05]
Training:  92%|█████████▏| 10154/11082 [32:39:16<2:59:25, 11.60s/step, acc=0.5071, loss=2.0174, lr=1.16e-05]
Training:  92%|█████████▏| 10155/11082 [32:39:28<2:59:27, 11.62s/step, acc=0.5071, loss=2.0174, lr=1.16e-05]
Training:  92%|█████████▏| 10156/11082 [32:39:39<2:56:55, 11.46s/step, acc=0.5071, loss=2.0174, lr=1.16e-05]
Training:  92%|█████████▏| 10157/11082 [32:39:50<2:56:54, 11.47s/step, acc=0.5071, loss=2.0174, lr=1.16e-05]
Training:  92%|█████████▏| 10158/11082 [32:40:02<2:58:44, 11.61s/step, acc=0.5071, loss=2.0174, lr=1.16e-05]
Training:  92%|█████████▏| 10159/11082 [32:40:14<2:59:45, 11.69s/step, acc=0.5071, loss=2.0174, lr=1.16e-05]
Training:  92%|█████████▏| 10160/11082 [32:40:26<2:58:48, 11.64s/step, acc=0.5071, loss=2.0174, lr=1.16e-05]
Training:  92%|█████████▏| 10161/11082 [32:40:37<2:58:09, 11.61s/step, acc=0.5071, loss=2.0174, lr=1.16e-05]
Training:  92%|█████████▏| 10162/11082 [32:40:48<2:56:11, 11.49s/step, acc=0.5071, loss=2.0174, lr=1.16e-05]
Training:  92%|█████████▏| 10163/11082 [32:41:00<2:56:25, 11.52s/step, acc=0.5071, loss=2.0174, lr=1.16e-05]
Training:  92%|█████████▏| 10164/11082 [32:41:12<2:58:31, 11.67s/step, acc=0.5071, loss=2.0174, lr=1.16e-05]
Training:  92%|█████████▏| 10165/11082 [32:41:24<2:59:22, 11.74s/step, acc=0.5071, loss=2.0174, lr=1.16e-05]
Training:  92%|█████████▏| 10166/11082 [32:41:35<2:57:06, 11.60s/step, acc=0.5071, loss=2.0174, lr=1.16e-05]
Training:  92%|█████████▏| 10167/11082 [32:41:47<2:56:27, 11.57s/step, acc=0.5071, loss=2.0174, lr=1.16e-05]
Training:  92%|█████████▏| 10168/11082 [32:41:58<2:56:22, 11.58s/step, acc=0.5071, loss=2.0174, lr=1.16e-05]
Training:  92%|█████████▏| 10169/11082 [32:42:10<2:55:48, 11.55s/step, acc=0.5071, loss=2.0174, lr=1.16e-05]
Training:  92%|█████████▏| 10170/11082 [32:42:21<2:54:41, 11.49s/step, acc=0.5071, loss=2.0174, lr=1.16e-05]
Training:  92%|█████████▏| 10171/11082 [32:42:33<2:57:24, 11.68s/step, acc=0.5071, loss=2.0174, lr=1.16e-05]
Training:  92%|█████████▏| 10172/11082 [32:42:45<2:56:36, 11.64s/step, acc=0.5071, loss=2.0174, lr=1.16e-05]
Training:  92%|█████████▏| 10173/11082 [32:42:56<2:53:41, 11.47s/step, acc=0.5071, loss=2.0174, lr=1.16e-05]
Training:  92%|█████████▏| 10174/11082 [32:43:08<2:55:56, 11.63s/step, acc=0.5071, loss=2.0174, lr=1.16e-05]
Training:  92%|█████████▏| 10175/11082 [32:43:19<2:55:48, 11.63s/step, acc=0.5071, loss=2.0174, lr=1.16e-05]
Training:  92%|█████████▏| 10176/11082 [32:43:31<2:55:23, 11.62s/step, acc=0.5071, loss=2.0174, lr=1.16e-05]
Training:  92%|█████████▏| 10177/11082 [32:43:43<2:55:36, 11.64s/step, acc=0.5071, loss=2.0174, lr=1.16e-05]
Training:  92%|█████████▏| 10178/11082 [32:43:54<2:54:04, 11.55s/step, acc=0.5071, loss=2.0174, lr=1.16e-05]
Training:  92%|█████████▏| 10179/11082 [32:44:05<2:52:53, 11.49s/step, acc=0.5071, loss=2.0174, lr=1.16e-05]
Training:  92%|█████████▏| 10180/11082 [32:44:17<2:52:45, 11.49s/step, acc=0.5071, loss=2.0174, lr=1.16e-05]
Training:  92%|█████████▏| 10181/11082 [32:44:29<2:53:29, 11.55s/step, acc=0.5071, loss=2.0174, lr=1.16e-05]
Training:  92%|█████████▏| 10182/11082 [32:44:41<2:55:01, 11.67s/step, acc=0.5071, loss=2.0174, lr=1.16e-05]
Training:  92%|█████████▏| 10183/11082 [32:44:52<2:56:01, 11.75s/step, acc=0.5071, loss=2.0174, lr=1.16e-05]
Training:  92%|█████████▏| 10184/11082 [32:45:04<2:54:16, 11.64s/step, acc=0.5071, loss=2.0174, lr=1.16e-05]
Training:  92%|█████████▏| 10185/11082 [32:45:16<2:54:14, 11.65s/step, acc=0.5071, loss=2.0174, lr=1.16e-05]
Training:  92%|█████████▏| 10186/11082 [32:45:27<2:54:58, 11.72s/step, acc=0.5071, loss=2.0174, lr=1.16e-05]
Training:  92%|█████████▏| 10187/11082 [32:45:39<2:54:22, 11.69s/step, acc=0.5071, loss=2.0174, lr=1.16e-05]
Training:  92%|█████████▏| 10188/11082 [32:45:51<2:53:24, 11.64s/step, acc=0.5071, loss=2.0174, lr=1.16e-05]
Training:  92%|█████████▏| 10189/11082 [32:46:02<2:53:09, 11.63s/step, acc=0.5071, loss=2.0174, lr=1.16e-05]
Training:  92%|█████████▏| 10190/11082 [32:46:14<2:53:43, 11.69s/step, acc=0.5071, loss=2.0174, lr=1.16e-05]
Training:  92%|█████████▏| 10191/11082 [32:46:26<2:53:32, 11.69s/step, acc=0.5071, loss=2.0174, lr=1.16e-05]
Training:  92%|█████████▏| 10192/11082 [32:46:37<2:51:14, 11.54s/step, acc=0.5071, loss=2.0174, lr=1.16e-05]
Training:  92%|█████████▏| 10193/11082 [32:46:49<2:53:07, 11.68s/step, acc=0.5071, loss=2.0174, lr=1.16e-05]
Training:  92%|█████████▏| 10194/11082 [32:47:00<2:51:01, 11.56s/step, acc=0.5071, loss=2.0174, lr=1.16e-05]
Training:  92%|█████████▏| 10195/11082 [32:47:12<2:51:09, 11.58s/step, acc=0.5071, loss=2.0174, lr=1.16e-05]
Training:  92%|█████████▏| 10196/11082 [32:47:24<2:52:02, 11.65s/step, acc=0.5071, loss=2.0174, lr=1.16e-05]
Training:  92%|█████████▏| 10197/11082 [32:47:35<2:49:40, 11.50s/step, acc=0.5071, loss=2.0174, lr=1.16e-05]
Training:  92%|█████████▏| 10198/11082 [32:47:46<2:50:00, 11.54s/step, acc=0.5071, loss=2.0174, lr=1.16e-05]
Training:  92%|█████████▏| 10199/11082 [32:47:58<2:48:34, 11.45s/step, acc=0.5071, loss=2.0174, lr=1.16e-05]
Training:  92%|█████████▏| 10200/11082 [32:48:09<2:49:38, 11.54s/step, acc=0.5071, loss=2.0174, lr=1.16e-05]
Training:  92%|█████████▏| 10200/11082 [32:48:16<2:49:38, 11.54s/step, acc=0.5086, loss=2.0081, lr=1.14e-05]2026-06-19 23:41:44 | INFO | __main__ | epoch=1 step=10200 train_loss=2.008116 train_acc=0.508618 lr=1.144e-05 grad_norm=0.4766 acc_d0=0.7629 acc_d1=0.6189 acc_d2=0.4984 acc_d3=0.4019 acc_d4=0.3266 acc_d5=0.2635 acc_d6=0.2131 acc_d7=0.1756

Training:  92%|█████████▏| 10201/11082 [32:48:21<2:49:53, 11.57s/step, acc=0.5086, loss=2.0081, lr=1.14e-05]
Training:  92%|█████████▏| 10202/11082 [32:48:32<2:47:48, 11.44s/step, acc=0.5086, loss=2.0081, lr=1.14e-05]
Training:  92%|█████████▏| 10203/11082 [32:48:44<2:49:37, 11.58s/step, acc=0.5086, loss=2.0081, lr=1.14e-05]
Training:  92%|█████████▏| 10204/11082 [32:48:56<2:49:49, 11.61s/step, acc=0.5086, loss=2.0081, lr=1.14e-05]
Training:  92%|█████████▏| 10205/11082 [32:49:07<2:48:46, 11.55s/step, acc=0.5086, loss=2.0081, lr=1.14e-05]
Training:  92%|█████████▏| 10206/11082 [32:49:19<2:47:48, 11.49s/step, acc=0.5086, loss=2.0081, lr=1.14e-05]
Training:  92%|█████████▏| 10207/11082 [32:49:30<2:47:58, 11.52s/step, acc=0.5086, loss=2.0081, lr=1.14e-05]
Training:  92%|█████████▏| 10208/11082 [32:49:42<2:48:22, 11.56s/step, acc=0.5086, loss=2.0081, lr=1.14e-05]
Training:  92%|█████████▏| 10209/11082 [32:49:53<2:48:51, 11.61s/step, acc=0.5086, loss=2.0081, lr=1.14e-05]
Training:  92%|█████████▏| 10210/11082 [32:50:05<2:48:57, 11.63s/step, acc=0.5086, loss=2.0081, lr=1.14e-05]
Training:  92%|█████████▏| 10211/11082 [32:50:17<2:50:33, 11.75s/step, acc=0.5086, loss=2.0081, lr=1.14e-05]
Training:  92%|█████████▏| 10212/11082 [32:50:28<2:48:20, 11.61s/step, acc=0.5086, loss=2.0081, lr=1.14e-05]
Training:  92%|█████████▏| 10213/11082 [32:50:40<2:49:33, 11.71s/step, acc=0.5086, loss=2.0081, lr=1.14e-05]
Training:  92%|█████████▏| 10214/11082 [32:50:52<2:48:12, 11.63s/step, acc=0.5086, loss=2.0081, lr=1.14e-05]
Training:  92%|█████████▏| 10215/11082 [32:51:03<2:47:11, 11.57s/step, acc=0.5086, loss=2.0081, lr=1.14e-05]
Training:  92%|█████████▏| 10216/11082 [32:51:15<2:46:26, 11.53s/step, acc=0.5086, loss=2.0081, lr=1.14e-05]
Training:  92%|█████████▏| 10217/11082 [32:51:26<2:46:10, 11.53s/step, acc=0.5086, loss=2.0081, lr=1.14e-05]
Training:  92%|█████████▏| 10218/11082 [32:51:38<2:45:18, 11.48s/step, acc=0.5086, loss=2.0081, lr=1.14e-05]
Training:  92%|█████████▏| 10219/11082 [32:51:50<2:48:45, 11.73s/step, acc=0.5086, loss=2.0081, lr=1.14e-05]
Training:  92%|█████████▏| 10220/11082 [32:52:01<2:46:17, 11.57s/step, acc=0.5086, loss=2.0081, lr=1.14e-05]
Training:  92%|█████████▏| 10221/11082 [32:52:13<2:45:58, 11.57s/step, acc=0.5086, loss=2.0081, lr=1.14e-05]
Training:  92%|█████████▏| 10222/11082 [32:52:24<2:45:22, 11.54s/step, acc=0.5086, loss=2.0081, lr=1.14e-05]
Training:  92%|█████████▏| 10223/11082 [32:52:36<2:46:19, 11.62s/step, acc=0.5086, loss=2.0081, lr=1.14e-05]
Training:  92%|█████████▏| 10224/11082 [32:52:47<2:45:24, 11.57s/step, acc=0.5086, loss=2.0081, lr=1.14e-05]
Training:  92%|█████████▏| 10225/11082 [32:52:59<2:44:19, 11.50s/step, acc=0.5086, loss=2.0081, lr=1.14e-05]
Training:  92%|█████████▏| 10226/11082 [32:53:11<2:45:11, 11.58s/step, acc=0.5086, loss=2.0081, lr=1.14e-05]
Training:  92%|█████████▏| 10227/11082 [32:53:22<2:45:36, 11.62s/step, acc=0.5086, loss=2.0081, lr=1.14e-05]
Training:  92%|█████████▏| 10228/11082 [32:53:34<2:44:37, 11.57s/step, acc=0.5086, loss=2.0081, lr=1.14e-05]
Training:  92%|█████████▏| 10229/11082 [32:53:45<2:45:30, 11.64s/step, acc=0.5086, loss=2.0081, lr=1.14e-05]
Training:  92%|█████████▏| 10230/11082 [32:53:57<2:44:44, 11.60s/step, acc=0.5086, loss=2.0081, lr=1.14e-05]
Training:  92%|█████████▏| 10231/11082 [32:54:08<2:41:51, 11.41s/step, acc=0.5086, loss=2.0081, lr=1.14e-05]
Training:  92%|█████████▏| 10232/11082 [32:54:20<2:44:31, 11.61s/step, acc=0.5086, loss=2.0081, lr=1.14e-05]
Training:  92%|█████████▏| 10233/11082 [32:54:31<2:41:46, 11.43s/step, acc=0.5086, loss=2.0081, lr=1.14e-05]
Training:  92%|█████████▏| 10234/11082 [32:54:43<2:42:54, 11.53s/step, acc=0.5086, loss=2.0081, lr=1.14e-05]
Training:  92%|█████████▏| 10235/11082 [32:54:54<2:42:17, 11.50s/step, acc=0.5086, loss=2.0081, lr=1.14e-05]
Training:  92%|█████████▏| 10236/11082 [32:55:06<2:42:55, 11.56s/step, acc=0.5086, loss=2.0081, lr=1.14e-05]
Training:  92%|█████████▏| 10237/11082 [32:55:17<2:42:46, 11.56s/step, acc=0.5086, loss=2.0081, lr=1.14e-05]
Training:  92%|█████████▏| 10238/11082 [32:55:29<2:42:36, 11.56s/step, acc=0.5086, loss=2.0081, lr=1.14e-05]
Training:  92%|█████████▏| 10239/11082 [32:55:40<2:40:59, 11.46s/step, acc=0.5086, loss=2.0081, lr=1.14e-05]
Training:  92%|█████████▏| 10240/11082 [32:55:52<2:43:33, 11.66s/step, acc=0.5086, loss=2.0081, lr=1.14e-05]
Training:  92%|█████████▏| 10241/11082 [32:56:04<2:43:21, 11.65s/step, acc=0.5086, loss=2.0081, lr=1.14e-05]
Training:  92%|█████████▏| 10242/11082 [32:56:15<2:41:18, 11.52s/step, acc=0.5086, loss=2.0081, lr=1.14e-05]
Training:  92%|█████████▏| 10243/11082 [32:56:27<2:43:14, 11.67s/step, acc=0.5086, loss=2.0081, lr=1.14e-05]
Training:  92%|█████████▏| 10244/11082 [32:56:39<2:41:30, 11.56s/step, acc=0.5086, loss=2.0081, lr=1.14e-05]
Training:  92%|█████████▏| 10245/11082 [32:56:50<2:40:15, 11.49s/step, acc=0.5086, loss=2.0081, lr=1.14e-05]
Training:  92%|█████████▏| 10246/11082 [32:57:02<2:41:50, 11.62s/step, acc=0.5086, loss=2.0081, lr=1.14e-05]
Training:  92%|█████████▏| 10247/11082 [32:57:13<2:39:40, 11.47s/step, acc=0.5086, loss=2.0081, lr=1.14e-05]
Training:  92%|█████████▏| 10248/11082 [32:57:24<2:38:43, 11.42s/step, acc=0.5086, loss=2.0081, lr=1.14e-05]
Training:  92%|█████████▏| 10249/11082 [32:57:36<2:39:59, 11.52s/step, acc=0.5086, loss=2.0081, lr=1.14e-05]
Training:  92%|█████████▏| 10250/11082 [32:57:47<2:38:57, 11.46s/step, acc=0.5086, loss=2.0081, lr=1.14e-05]
Training:  92%|█████████▏| 10250/11082 [32:57:54<2:38:57, 11.46s/step, acc=0.5089, loss=2.0108, lr=1.13e-05]2026-06-19 23:51:22 | INFO | __main__ | epoch=1 step=10250 train_loss=2.010793 train_acc=0.508888 lr=1.128e-05 grad_norm=0.5312 acc_d0=0.7600 acc_d1=0.6188 acc_d2=0.5005 acc_d3=0.4036 acc_d4=0.3261 acc_d5=0.2635 acc_d6=0.2140 acc_d7=0.1753

Training:  93%|█████████▎| 10251/11082 [32:57:59<2:39:56, 11.55s/step, acc=0.5089, loss=2.0108, lr=1.13e-05]
Training:  93%|█████████▎| 10252/11082 [32:58:11<2:39:36, 11.54s/step, acc=0.5089, loss=2.0108, lr=1.13e-05]
Training:  93%|█████████▎| 10253/11082 [32:58:22<2:40:12, 11.60s/step, acc=0.5089, loss=2.0108, lr=1.13e-05]
Training:  93%|█████████▎| 10254/11082 [32:58:33<2:37:30, 11.41s/step, acc=0.5089, loss=2.0108, lr=1.13e-05]
Training:  93%|█████████▎| 10255/11082 [32:58:45<2:38:16, 11.48s/step, acc=0.5089, loss=2.0108, lr=1.13e-05]
Training:  93%|█████████▎| 10256/11082 [32:58:57<2:39:14, 11.57s/step, acc=0.5089, loss=2.0108, lr=1.13e-05]
Training:  93%|█████████▎| 10257/11082 [32:59:08<2:39:35, 11.61s/step, acc=0.5089, loss=2.0108, lr=1.13e-05]
Training:  93%|█████████▎| 10258/11082 [32:59:20<2:40:27, 11.68s/step, acc=0.5089, loss=2.0108, lr=1.13e-05]
Training:  93%|█████████▎| 10259/11082 [32:59:32<2:39:08, 11.60s/step, acc=0.5089, loss=2.0108, lr=1.13e-05]
Training:  93%|█████████▎| 10260/11082 [32:59:43<2:37:43, 11.51s/step, acc=0.5089, loss=2.0108, lr=1.13e-05]
Training:  93%|█████████▎| 10261/11082 [32:59:55<2:38:22, 11.57s/step, acc=0.5089, loss=2.0108, lr=1.13e-05]
Training:  93%|█████████▎| 10262/11082 [33:00:06<2:38:08, 11.57s/step, acc=0.5089, loss=2.0108, lr=1.13e-05]
Training:  93%|█████████▎| 10263/11082 [33:00:17<2:35:50, 11.42s/step, acc=0.5089, loss=2.0108, lr=1.13e-05]
Training:  93%|█████████▎| 10264/11082 [33:00:29<2:35:17, 11.39s/step, acc=0.5089, loss=2.0108, lr=1.13e-05]
Training:  93%|█████████▎| 10265/11082 [33:00:40<2:36:18, 11.48s/step, acc=0.5089, loss=2.0108, lr=1.13e-05]
Training:  93%|█████████▎| 10266/11082 [33:00:52<2:36:33, 11.51s/step, acc=0.5089, loss=2.0108, lr=1.13e-05]
Training:  93%|█████████▎| 10267/11082 [33:01:03<2:36:27, 11.52s/step, acc=0.5089, loss=2.0108, lr=1.13e-05]
Training:  93%|█████████▎| 10268/11082 [33:01:15<2:37:39, 11.62s/step, acc=0.5089, loss=2.0108, lr=1.13e-05]
Training:  93%|█████████▎| 10269/11082 [33:01:26<2:34:36, 11.41s/step, acc=0.5089, loss=2.0108, lr=1.13e-05]
Training:  93%|█████████▎| 10270/11082 [33:01:38<2:35:40, 11.50s/step, acc=0.5089, loss=2.0108, lr=1.13e-05]
Training:  93%|█████████▎| 10271/11082 [33:01:49<2:34:50, 11.46s/step, acc=0.5089, loss=2.0108, lr=1.13e-05]
Training:  93%|█████████▎| 10272/11082 [33:02:01<2:35:25, 11.51s/step, acc=0.5089, loss=2.0108, lr=1.13e-05]
Training:  93%|█████████▎| 10273/11082 [33:02:13<2:37:32, 11.68s/step, acc=0.5089, loss=2.0108, lr=1.13e-05]
Training:  93%|█████████▎| 10274/11082 [33:02:25<2:37:18, 11.68s/step, acc=0.5089, loss=2.0108, lr=1.13e-05]
Training:  93%|█████████▎| 10275/11082 [33:02:36<2:35:17, 11.55s/step, acc=0.5089, loss=2.0108, lr=1.13e-05]
Training:  93%|█████████▎| 10276/11082 [33:02:47<2:34:39, 11.51s/step, acc=0.5089, loss=2.0108, lr=1.13e-05]
Training:  93%|█████████▎| 10277/11082 [33:02:59<2:36:24, 11.66s/step, acc=0.5089, loss=2.0108, lr=1.13e-05]
Training:  93%|█████████▎| 10278/11082 [33:03:10<2:33:59, 11.49s/step, acc=0.5089, loss=2.0108, lr=1.13e-05]
Training:  93%|█████████▎| 10279/11082 [33:03:22<2:34:46, 11.57s/step, acc=0.5089, loss=2.0108, lr=1.13e-05]
Training:  93%|█████████▎| 10280/11082 [33:03:34<2:35:22, 11.62s/step, acc=0.5089, loss=2.0108, lr=1.13e-05]
Training:  93%|█████████▎| 10281/11082 [33:03:45<2:34:09, 11.55s/step, acc=0.5089, loss=2.0108, lr=1.13e-05]
Training:  93%|█████████▎| 10282/11082 [33:03:58<2:36:28, 11.74s/step, acc=0.5089, loss=2.0108, lr=1.13e-05]
Training:  93%|█████████▎| 10283/11082 [33:04:09<2:34:44, 11.62s/step, acc=0.5089, loss=2.0108, lr=1.13e-05]
Training:  93%|█████████▎| 10284/11082 [33:04:21<2:38:06, 11.89s/step, acc=0.5089, loss=2.0108, lr=1.13e-05]
Training:  93%|█████████▎| 10285/11082 [33:04:33<2:35:36, 11.72s/step, acc=0.5089, loss=2.0108, lr=1.13e-05]
Training:  93%|█████████▎| 10286/11082 [33:04:44<2:34:18, 11.63s/step, acc=0.5089, loss=2.0108, lr=1.13e-05]
Training:  93%|█████████▎| 10287/11082 [33:04:55<2:32:08, 11.48s/step, acc=0.5089, loss=2.0108, lr=1.13e-05]
Training:  93%|█████████▎| 10288/11082 [33:05:07<2:34:01, 11.64s/step, acc=0.5089, loss=2.0108, lr=1.13e-05]
Training:  93%|█████████▎| 10289/11082 [33:05:19<2:32:33, 11.54s/step, acc=0.5089, loss=2.0108, lr=1.13e-05]
Training:  93%|█████████▎| 10290/11082 [33:05:30<2:33:30, 11.63s/step, acc=0.5089, loss=2.0108, lr=1.13e-05]
Training:  93%|█████████▎| 10291/11082 [33:05:42<2:31:31, 11.49s/step, acc=0.5089, loss=2.0108, lr=1.13e-05]
Training:  93%|█████████▎| 10292/11082 [33:05:53<2:31:17, 11.49s/step, acc=0.5089, loss=2.0108, lr=1.13e-05]
Training:  93%|█████████▎| 10293/11082 [33:06:05<2:31:37, 11.53s/step, acc=0.5089, loss=2.0108, lr=1.13e-05]
Training:  93%|█████████▎| 10294/11082 [33:06:16<2:31:01, 11.50s/step, acc=0.5089, loss=2.0108, lr=1.13e-05]
Training:  93%|█████████▎| 10295/11082 [33:06:28<2:31:36, 11.56s/step, acc=0.5089, loss=2.0108, lr=1.13e-05]
Training:  93%|█████████▎| 10296/11082 [33:06:40<2:32:13, 11.62s/step, acc=0.5089, loss=2.0108, lr=1.13e-05]
Training:  93%|█████████▎| 10297/11082 [33:06:51<2:32:34, 11.66s/step, acc=0.5089, loss=2.0108, lr=1.13e-05]
Training:  93%|█████████▎| 10298/11082 [33:07:03<2:32:12, 11.65s/step, acc=0.5089, loss=2.0108, lr=1.13e-05]
Training:  93%|█████████▎| 10299/11082 [33:07:15<2:31:54, 11.64s/step, acc=0.5089, loss=2.0108, lr=1.13e-05]
Training:  93%|█████████▎| 10300/11082 [33:07:26<2:30:39, 11.56s/step, acc=0.5089, loss=2.0108, lr=1.13e-05]
Training:  93%|█████████▎| 10300/11082 [33:07:33<2:30:39, 11.56s/step, acc=0.5065, loss=2.0205, lr=1.11e-05]2026-06-20 00:01:01 | INFO | __main__ | epoch=1 step=10300 train_loss=2.020533 train_acc=0.506468 lr=1.113e-05 grad_norm=0.5234 acc_d0=0.7647 acc_d1=0.6219 acc_d2=0.5024 acc_d3=0.4042 acc_d4=0.3265 acc_d5=0.2644 acc_d6=0.2143 acc_d7=0.1767

Training:  93%|█████████▎| 10301/11082 [33:07:37<2:29:16, 11.47s/step, acc=0.5065, loss=2.0205, lr=1.11e-05]
Training:  93%|█████████▎| 10302/11082 [33:07:49<2:30:48, 11.60s/step, acc=0.5065, loss=2.0205, lr=1.11e-05]
Training:  93%|█████████▎| 10303/11082 [33:08:01<2:31:19, 11.66s/step, acc=0.5065, loss=2.0205, lr=1.11e-05]
Training:  93%|█████████▎| 10304/11082 [33:08:12<2:30:12, 11.58s/step, acc=0.5065, loss=2.0205, lr=1.11e-05]
Training:  93%|█████████▎| 10305/11082 [33:08:24<2:30:32, 11.63s/step, acc=0.5065, loss=2.0205, lr=1.11e-05]
Training:  93%|█████████▎| 10306/11082 [33:08:36<2:30:07, 11.61s/step, acc=0.5065, loss=2.0205, lr=1.11e-05]
Training:  93%|█████████▎| 10307/11082 [33:08:47<2:29:56, 11.61s/step, acc=0.5065, loss=2.0205, lr=1.11e-05]
Training:  93%|█████████▎| 10308/11082 [33:08:58<2:28:09, 11.49s/step, acc=0.5065, loss=2.0205, lr=1.11e-05]
Training:  93%|█████████▎| 10309/11082 [33:09:11<2:30:43, 11.70s/step, acc=0.5065, loss=2.0205, lr=1.11e-05]
Training:  93%|█████████▎| 10310/11082 [33:09:22<2:29:49, 11.64s/step, acc=0.5065, loss=2.0205, lr=1.11e-05]
Training:  93%|█████████▎| 10311/11082 [33:09:34<2:29:51, 11.66s/step, acc=0.5065, loss=2.0205, lr=1.11e-05]
Training:  93%|█████████▎| 10312/11082 [33:09:45<2:27:29, 11.49s/step, acc=0.5065, loss=2.0205, lr=1.11e-05]
Training:  93%|█████████▎| 10313/11082 [33:09:57<2:28:46, 11.61s/step, acc=0.5065, loss=2.0205, lr=1.11e-05]
Training:  93%|█████████▎| 10314/11082 [33:10:09<2:29:21, 11.67s/step, acc=0.5065, loss=2.0205, lr=1.11e-05]
Training:  93%|█████████▎| 10315/11082 [33:10:20<2:26:31, 11.46s/step, acc=0.5065, loss=2.0205, lr=1.11e-05]
Training:  93%|█████████▎| 10316/11082 [33:10:31<2:27:39, 11.57s/step, acc=0.5065, loss=2.0205, lr=1.11e-05]
Training:  93%|█████████▎| 10317/11082 [33:10:43<2:28:25, 11.64s/step, acc=0.5065, loss=2.0205, lr=1.11e-05]
Training:  93%|█████████▎| 10318/11082 [33:10:54<2:26:41, 11.52s/step, acc=0.5065, loss=2.0205, lr=1.11e-05]
Training:  93%|█████████▎| 10319/11082 [33:11:06<2:26:52, 11.55s/step, acc=0.5065, loss=2.0205, lr=1.11e-05]
Training:  93%|█████████▎| 10320/11082 [33:11:17<2:25:46, 11.48s/step, acc=0.5065, loss=2.0205, lr=1.11e-05]
Training:  93%|█████████▎| 10321/11082 [33:11:30<2:28:05, 11.68s/step, acc=0.5065, loss=2.0205, lr=1.11e-05]
Training:  93%|█████████▎| 10322/11082 [33:11:41<2:26:05, 11.53s/step, acc=0.5065, loss=2.0205, lr=1.11e-05]
Training:  93%|█████████▎| 10323/11082 [33:11:52<2:26:41, 11.60s/step, acc=0.5065, loss=2.0205, lr=1.11e-05]
Training:  93%|█████████▎| 10324/11082 [33:12:04<2:26:22, 11.59s/step, acc=0.5065, loss=2.0205, lr=1.11e-05]
Training:  93%|█████████▎| 10325/11082 [33:12:16<2:26:38, 11.62s/step, acc=0.5065, loss=2.0205, lr=1.11e-05]
Training:  93%|█████████▎| 10326/11082 [33:12:27<2:26:10, 11.60s/step, acc=0.5065, loss=2.0205, lr=1.11e-05]
Training:  93%|█████████▎| 10327/11082 [33:12:39<2:26:15, 11.62s/step, acc=0.5065, loss=2.0205, lr=1.11e-05]
Training:  93%|█████████▎| 10328/11082 [33:12:51<2:26:15, 11.64s/step, acc=0.5065, loss=2.0205, lr=1.11e-05]
Training:  93%|█████████▎| 10329/11082 [33:13:02<2:23:42, 11.45s/step, acc=0.5065, loss=2.0205, lr=1.11e-05]
Training:  93%|█████████▎| 10330/11082 [33:13:14<2:25:08, 11.58s/step, acc=0.5065, loss=2.0205, lr=1.11e-05]
Training:  93%|█████████▎| 10331/11082 [33:13:25<2:25:14, 11.60s/step, acc=0.5065, loss=2.0205, lr=1.11e-05]
Training:  93%|█████████▎| 10332/11082 [33:13:36<2:23:15, 11.46s/step, acc=0.5065, loss=2.0205, lr=1.11e-05]
Training:  93%|█████████▎| 10333/11082 [33:13:48<2:25:21, 11.64s/step, acc=0.5065, loss=2.0205, lr=1.11e-05]
Training:  93%|█████████▎| 10334/11082 [33:14:00<2:24:29, 11.59s/step, acc=0.5065, loss=2.0205, lr=1.11e-05]
Training:  93%|█████████▎| 10335/11082 [33:14:11<2:22:38, 11.46s/step, acc=0.5065, loss=2.0205, lr=1.11e-05]
Training:  93%|█████████▎| 10336/11082 [33:14:23<2:24:49, 11.65s/step, acc=0.5065, loss=2.0205, lr=1.11e-05]
Training:  93%|█████████▎| 10337/11082 [33:14:35<2:26:00, 11.76s/step, acc=0.5065, loss=2.0205, lr=1.11e-05]
Training:  93%|█████████▎| 10338/11082 [33:14:47<2:25:22, 11.72s/step, acc=0.5065, loss=2.0205, lr=1.11e-05]
Training:  93%|█████████▎| 10339/11082 [33:14:58<2:23:56, 11.62s/step, acc=0.5065, loss=2.0205, lr=1.11e-05]
Training:  93%|█████████▎| 10340/11082 [33:15:10<2:24:05, 11.65s/step, acc=0.5065, loss=2.0205, lr=1.11e-05]
Training:  93%|█████████▎| 10341/11082 [33:15:21<2:23:44, 11.64s/step, acc=0.5065, loss=2.0205, lr=1.11e-05]
Training:  93%|█████████▎| 10342/11082 [33:15:33<2:23:33, 11.64s/step, acc=0.5065, loss=2.0205, lr=1.11e-05]
Training:  93%|█████████▎| 10343/11082 [33:15:45<2:22:40, 11.58s/step, acc=0.5065, loss=2.0205, lr=1.11e-05]
Training:  93%|█████████▎| 10344/11082 [33:15:56<2:21:51, 11.53s/step, acc=0.5065, loss=2.0205, lr=1.11e-05]
Training:  93%|█████████▎| 10345/11082 [33:16:07<2:20:50, 11.47s/step, acc=0.5065, loss=2.0205, lr=1.11e-05]
Training:  93%|█████████▎| 10346/11082 [33:16:19<2:21:17, 11.52s/step, acc=0.5065, loss=2.0205, lr=1.11e-05]
Training:  93%|█████████▎| 10347/11082 [33:16:30<2:21:07, 11.52s/step, acc=0.5065, loss=2.0205, lr=1.11e-05]
Training:  93%|█████████▎| 10348/11082 [33:16:42<2:22:23, 11.64s/step, acc=0.5065, loss=2.0205, lr=1.11e-05]
Training:  93%|█████████▎| 10349/11082 [33:16:53<2:19:53, 11.45s/step, acc=0.5065, loss=2.0205, lr=1.11e-05]
Training:  93%|█████████▎| 10350/11082 [33:17:05<2:20:04, 11.48s/step, acc=0.5065, loss=2.0205, lr=1.11e-05]
Training:  93%|█████████▎| 10350/11082 [33:17:12<2:20:04, 11.48s/step, acc=0.5076, loss=2.0151, lr=1.10e-05]2026-06-20 00:10:40 | INFO | __main__ | epoch=1 step=10350 train_loss=2.015111 train_acc=0.507570 lr=1.099e-05 grad_norm=0.4805 acc_d0=0.7589 acc_d1=0.6149 acc_d2=0.4947 acc_d3=0.3968 acc_d4=0.3182 acc_d5=0.2561 acc_d6=0.2079 acc_d7=0.1709

Training:  93%|█████████▎| 10351/11082 [33:17:17<2:20:49, 11.56s/step, acc=0.5076, loss=2.0151, lr=1.10e-05]
Training:  93%|█████████▎| 10352/11082 [33:17:28<2:20:04, 11.51s/step, acc=0.5076, loss=2.0151, lr=1.10e-05]
Training:  93%|█████████▎| 10353/11082 [33:17:40<2:20:42, 11.58s/step, acc=0.5076, loss=2.0151, lr=1.10e-05]
Training:  93%|█████████▎| 10354/11082 [33:17:52<2:21:23, 11.65s/step, acc=0.5076, loss=2.0151, lr=1.10e-05]
Training:  93%|█████████▎| 10355/11082 [33:18:03<2:19:12, 11.49s/step, acc=0.5076, loss=2.0151, lr=1.10e-05]
Training:  93%|█████████▎| 10356/11082 [33:18:15<2:20:00, 11.57s/step, acc=0.5076, loss=2.0151, lr=1.10e-05]
Training:  93%|█████████▎| 10357/11082 [33:18:26<2:19:46, 11.57s/step, acc=0.5076, loss=2.0151, lr=1.10e-05]
Training:  93%|█████████▎| 10358/11082 [33:18:38<2:19:47, 11.59s/step, acc=0.5076, loss=2.0151, lr=1.10e-05]
Training:  93%|█████████▎| 10359/11082 [33:18:49<2:18:14, 11.47s/step, acc=0.5076, loss=2.0151, lr=1.10e-05]
Training:  93%|█████████▎| 10360/11082 [33:19:01<2:19:27, 11.59s/step, acc=0.5076, loss=2.0151, lr=1.10e-05]
Training:  93%|█████████▎| 10361/11082 [33:19:12<2:17:46, 11.47s/step, acc=0.5076, loss=2.0151, lr=1.10e-05]
Training:  94%|█████████▎| 10362/11082 [33:19:24<2:19:14, 11.60s/step, acc=0.5076, loss=2.0151, lr=1.10e-05]
Training:  94%|█████████▎| 10363/11082 [33:19:36<2:19:08, 11.61s/step, acc=0.5076, loss=2.0151, lr=1.10e-05]
Training:  94%|█████████▎| 10364/11082 [33:19:47<2:18:13, 11.55s/step, acc=0.5076, loss=2.0151, lr=1.10e-05]
Training:  94%|█████████▎| 10365/11082 [33:19:59<2:18:19, 11.57s/step, acc=0.5076, loss=2.0151, lr=1.10e-05]
Training:  94%|█████████▎| 10366/11082 [33:20:10<2:18:52, 11.64s/step, acc=0.5076, loss=2.0151, lr=1.10e-05]
Training:  94%|█████████▎| 10367/11082 [33:20:22<2:18:38, 11.63s/step, acc=0.5076, loss=2.0151, lr=1.10e-05]
Training:  94%|█████████▎| 10368/11082 [33:20:33<2:17:40, 11.57s/step, acc=0.5076, loss=2.0151, lr=1.10e-05]
Training:  94%|█████████▎| 10369/11082 [33:20:45<2:16:31, 11.49s/step, acc=0.5076, loss=2.0151, lr=1.10e-05]
Training:  94%|█████████▎| 10370/11082 [33:20:56<2:16:28, 11.50s/step, acc=0.5076, loss=2.0151, lr=1.10e-05]
Training:  94%|█████████▎| 10371/11082 [33:21:08<2:18:24, 11.68s/step, acc=0.5076, loss=2.0151, lr=1.10e-05]
Training:  94%|█████████▎| 10372/11082 [33:21:19<2:16:09, 11.51s/step, acc=0.5076, loss=2.0151, lr=1.10e-05]
Training:  94%|█████████▎| 10373/11082 [33:21:31<2:16:21, 11.54s/step, acc=0.5076, loss=2.0151, lr=1.10e-05]
Training:  94%|█████████▎| 10374/11082 [33:21:43<2:16:10, 11.54s/step, acc=0.5076, loss=2.0151, lr=1.10e-05]
Training:  94%|█████████▎| 10375/11082 [33:21:54<2:16:33, 11.59s/step, acc=0.5076, loss=2.0151, lr=1.10e-05]
Training:  94%|█████████▎| 10376/11082 [33:22:06<2:16:57, 11.64s/step, acc=0.5076, loss=2.0151, lr=1.10e-05]
Training:  94%|█████████▎| 10377/11082 [33:22:17<2:15:28, 11.53s/step, acc=0.5076, loss=2.0151, lr=1.10e-05]
Training:  94%|█████████▎| 10378/11082 [33:22:29<2:15:25, 11.54s/step, acc=0.5076, loss=2.0151, lr=1.10e-05]
Training:  94%|█████████▎| 10379/11082 [33:22:41<2:15:58, 11.61s/step, acc=0.5076, loss=2.0151, lr=1.10e-05]
Training:  94%|█████████▎| 10380/11082 [33:22:52<2:14:32, 11.50s/step, acc=0.5076, loss=2.0151, lr=1.10e-05]
Training:  94%|█████████▎| 10381/11082 [33:23:03<2:14:20, 11.50s/step, acc=0.5076, loss=2.0151, lr=1.10e-05]
Training:  94%|█████████▎| 10382/11082 [33:23:15<2:16:00, 11.66s/step, acc=0.5076, loss=2.0151, lr=1.10e-05]
Training:  94%|█████████▎| 10383/11082 [33:23:27<2:14:19, 11.53s/step, acc=0.5076, loss=2.0151, lr=1.10e-05]
Training:  94%|█████████▎| 10384/11082 [33:23:38<2:13:45, 11.50s/step, acc=0.5076, loss=2.0151, lr=1.10e-05]
Training:  94%|█████████▎| 10385/11082 [33:23:51<2:17:17, 11.82s/step, acc=0.5076, loss=2.0151, lr=1.10e-05]
Training:  94%|█████████▎| 10386/11082 [33:24:02<2:15:14, 11.66s/step, acc=0.5076, loss=2.0151, lr=1.10e-05]
Training:  94%|█████████▎| 10387/11082 [33:24:13<2:14:39, 11.63s/step, acc=0.5076, loss=2.0151, lr=1.10e-05]
Training:  94%|█████████▎| 10388/11082 [33:24:25<2:15:10, 11.69s/step, acc=0.5076, loss=2.0151, lr=1.10e-05]
Training:  94%|█████████▎| 10389/11082 [33:24:37<2:14:49, 11.67s/step, acc=0.5076, loss=2.0151, lr=1.10e-05]
Training:  94%|█████████▍| 10390/11082 [33:24:48<2:14:03, 11.62s/step, acc=0.5076, loss=2.0151, lr=1.10e-05]
Training:  94%|█████████▍| 10391/11082 [33:25:00<2:13:55, 11.63s/step, acc=0.5076, loss=2.0151, lr=1.10e-05]
Training:  94%|█████████▍| 10392/11082 [33:25:11<2:12:27, 11.52s/step, acc=0.5076, loss=2.0151, lr=1.10e-05]
Training:  94%|█████████▍| 10393/11082 [33:25:23<2:12:24, 11.53s/step, acc=0.5076, loss=2.0151, lr=1.10e-05]
Training:  94%|█████████▍| 10394/11082 [33:25:34<2:10:34, 11.39s/step, acc=0.5076, loss=2.0151, lr=1.10e-05]
Training:  94%|█████████▍| 10395/11082 [33:25:46<2:11:43, 11.50s/step, acc=0.5076, loss=2.0151, lr=1.10e-05]
Training:  94%|█████████▍| 10396/11082 [33:25:57<2:10:17, 11.40s/step, acc=0.5076, loss=2.0151, lr=1.10e-05]
Training:  94%|█████████▍| 10397/11082 [33:26:08<2:10:41, 11.45s/step, acc=0.5076, loss=2.0151, lr=1.10e-05]
Training:  94%|█████████▍| 10398/11082 [33:26:20<2:12:16, 11.60s/step, acc=0.5076, loss=2.0151, lr=1.10e-05]
Training:  94%|█████████▍| 10399/11082 [33:26:32<2:10:38, 11.48s/step, acc=0.5076, loss=2.0151, lr=1.10e-05]
Training:  94%|█████████▍| 10400/11082 [33:26:43<2:11:31, 11.57s/step, acc=0.5076, loss=2.0151, lr=1.10e-05]
Training:  94%|█████████▍| 10400/11082 [33:26:50<2:11:31, 11.57s/step, acc=0.5088, loss=2.0072, lr=1.09e-05]2026-06-20 00:20:18 | INFO | __main__ | epoch=1 step=10400 train_loss=2.007155 train_acc=0.508832 lr=1.086e-05 grad_norm=0.4961 acc_d0=0.7665 acc_d1=0.6225 acc_d2=0.5048 acc_d3=0.4072 acc_d4=0.3286 acc_d5=0.2642 acc_d6=0.2144 acc_d7=0.1767

Training:  94%|█████████▍| 10401/11082 [33:26:55<2:10:59, 11.54s/step, acc=0.5088, loss=2.0072, lr=1.09e-05]
Training:  94%|█████████▍| 10402/11082 [33:27:06<2:10:14, 11.49s/step, acc=0.5088, loss=2.0072, lr=1.09e-05]
Training:  94%|█████████▍| 10403/11082 [33:27:18<2:11:52, 11.65s/step, acc=0.5088, loss=2.0072, lr=1.09e-05]
Training:  94%|█████████▍| 10404/11082 [33:27:30<2:10:45, 11.57s/step, acc=0.5088, loss=2.0072, lr=1.09e-05]
Training:  94%|█████████▍| 10405/11082 [33:27:41<2:09:33, 11.48s/step, acc=0.5088, loss=2.0072, lr=1.09e-05]
Training:  94%|█████████▍| 10406/11082 [33:27:53<2:10:27, 11.58s/step, acc=0.5088, loss=2.0072, lr=1.09e-05]
Training:  94%|█████████▍| 10407/11082 [33:28:04<2:09:19, 11.50s/step, acc=0.5088, loss=2.0072, lr=1.09e-05]
Training:  94%|█████████▍| 10408/11082 [33:28:15<2:08:37, 11.45s/step, acc=0.5088, loss=2.0072, lr=1.09e-05]
Training:  94%|█████████▍| 10409/11082 [33:28:27<2:09:39, 11.56s/step, acc=0.5088, loss=2.0072, lr=1.09e-05]
Training:  94%|█████████▍| 10410/11082 [33:28:39<2:09:16, 11.54s/step, acc=0.5088, loss=2.0072, lr=1.09e-05]
Training:  94%|█████████▍| 10411/11082 [33:28:50<2:09:48, 11.61s/step, acc=0.5088, loss=2.0072, lr=1.09e-05]
Training:  94%|█████████▍| 10412/11082 [33:29:02<2:09:18, 11.58s/step, acc=0.5088, loss=2.0072, lr=1.09e-05]
Training:  94%|█████████▍| 10413/11082 [33:29:13<2:08:55, 11.56s/step, acc=0.5088, loss=2.0072, lr=1.09e-05]
Training:  94%|█████████▍| 10414/11082 [33:29:25<2:08:41, 11.56s/step, acc=0.5088, loss=2.0072, lr=1.09e-05]
Training:  94%|█████████▍| 10415/11082 [33:29:36<2:07:49, 11.50s/step, acc=0.5088, loss=2.0072, lr=1.09e-05]
Training:  94%|█████████▍| 10416/11082 [33:29:48<2:06:37, 11.41s/step, acc=0.5088, loss=2.0072, lr=1.09e-05]
Training:  94%|█████████▍| 10417/11082 [33:29:59<2:07:38, 11.52s/step, acc=0.5088, loss=2.0072, lr=1.09e-05]
Training:  94%|█████████▍| 10418/11082 [33:30:11<2:08:27, 11.61s/step, acc=0.5088, loss=2.0072, lr=1.09e-05]
Training:  94%|█████████▍| 10419/11082 [33:30:23<2:09:05, 11.68s/step, acc=0.5088, loss=2.0072, lr=1.09e-05]
Training:  94%|█████████▍| 10420/11082 [33:30:35<2:09:10, 11.71s/step, acc=0.5088, loss=2.0072, lr=1.09e-05]
Training:  94%|█████████▍| 10421/11082 [33:30:46<2:07:50, 11.61s/step, acc=0.5088, loss=2.0072, lr=1.09e-05]
Training:  94%|█████████▍| 10422/11082 [33:30:58<2:08:06, 11.65s/step, acc=0.5088, loss=2.0072, lr=1.09e-05]
Training:  94%|█████████▍| 10423/11082 [33:31:10<2:07:57, 11.65s/step, acc=0.5088, loss=2.0072, lr=1.09e-05]
Training:  94%|█████████▍| 10424/11082 [33:31:21<2:08:00, 11.67s/step, acc=0.5088, loss=2.0072, lr=1.09e-05]
Training:  94%|█████████▍| 10425/11082 [33:31:33<2:07:56, 11.68s/step, acc=0.5088, loss=2.0072, lr=1.09e-05]
Training:  94%|█████████▍| 10426/11082 [33:31:44<2:06:51, 11.60s/step, acc=0.5088, loss=2.0072, lr=1.09e-05]
Training:  94%|█████████▍| 10427/11082 [33:31:56<2:06:38, 11.60s/step, acc=0.5088, loss=2.0072, lr=1.09e-05]
Training:  94%|█████████▍| 10428/11082 [33:32:08<2:07:03, 11.66s/step, acc=0.5088, loss=2.0072, lr=1.09e-05]
Training:  94%|█████████▍| 10429/11082 [33:32:19<2:05:11, 11.50s/step, acc=0.5088, loss=2.0072, lr=1.09e-05]
Training:  94%|█████████▍| 10430/11082 [33:32:31<2:06:14, 11.62s/step, acc=0.5088, loss=2.0072, lr=1.09e-05]
Training:  94%|█████████▍| 10431/11082 [33:32:42<2:05:14, 11.54s/step, acc=0.5088, loss=2.0072, lr=1.09e-05]
Training:  94%|█████████▍| 10432/11082 [33:32:54<2:05:09, 11.55s/step, acc=0.5088, loss=2.0072, lr=1.09e-05]
Training:  94%|█████████▍| 10433/11082 [33:33:05<2:03:32, 11.42s/step, acc=0.5088, loss=2.0072, lr=1.09e-05]
Training:  94%|█████████▍| 10434/11082 [33:33:17<2:05:07, 11.59s/step, acc=0.5088, loss=2.0072, lr=1.09e-05]
Training:  94%|█████████▍| 10435/11082 [33:33:29<2:05:34, 11.65s/step, acc=0.5088, loss=2.0072, lr=1.09e-05]
Training:  94%|█████████▍| 10436/11082 [33:33:40<2:04:39, 11.58s/step, acc=0.5088, loss=2.0072, lr=1.09e-05]
Training:  94%|█████████▍| 10437/11082 [33:33:51<2:03:28, 11.49s/step, acc=0.5088, loss=2.0072, lr=1.09e-05]
Training:  94%|█████████▍| 10438/11082 [33:34:03<2:02:26, 11.41s/step, acc=0.5088, loss=2.0072, lr=1.09e-05]
Training:  94%|█████████▍| 10439/11082 [33:34:15<2:04:36, 11.63s/step, acc=0.5088, loss=2.0072, lr=1.09e-05]
Training:  94%|█████████▍| 10440/11082 [33:34:26<2:03:27, 11.54s/step, acc=0.5088, loss=2.0072, lr=1.09e-05]
Training:  94%|█████████▍| 10441/11082 [33:34:38<2:03:51, 11.59s/step, acc=0.5088, loss=2.0072, lr=1.09e-05]
Training:  94%|█████████▍| 10442/11082 [33:34:49<2:03:35, 11.59s/step, acc=0.5088, loss=2.0072, lr=1.09e-05]
Training:  94%|█████████▍| 10443/11082 [33:35:01<2:03:22, 11.59s/step, acc=0.5088, loss=2.0072, lr=1.09e-05]
Training:  94%|█████████▍| 10444/11082 [33:35:13<2:03:45, 11.64s/step, acc=0.5088, loss=2.0072, lr=1.09e-05]
Training:  94%|█████████▍| 10445/11082 [33:35:24<2:03:08, 11.60s/step, acc=0.5088, loss=2.0072, lr=1.09e-05]
Training:  94%|█████████▍| 10446/11082 [33:35:36<2:02:34, 11.56s/step, acc=0.5088, loss=2.0072, lr=1.09e-05]
Training:  94%|█████████▍| 10447/11082 [33:35:48<2:03:27, 11.67s/step, acc=0.5088, loss=2.0072, lr=1.09e-05]
Training:  94%|█████████▍| 10448/11082 [33:35:59<2:02:08, 11.56s/step, acc=0.5088, loss=2.0072, lr=1.09e-05]
Training:  94%|█████████▍| 10449/11082 [33:36:10<2:02:08, 11.58s/step, acc=0.5088, loss=2.0072, lr=1.09e-05]
Training:  94%|█████████▍| 10450/11082 [33:36:22<2:02:37, 11.64s/step, acc=0.5088, loss=2.0072, lr=1.09e-05]
Training:  94%|█████████▍| 10450/11082 [33:36:29<2:02:37, 11.64s/step, acc=0.5072, loss=2.0134, lr=1.07e-05]2026-06-20 00:29:57 | INFO | __main__ | epoch=1 step=10450 train_loss=2.013388 train_acc=0.507183 lr=1.074e-05 grad_norm=0.4961 acc_d0=0.7633 acc_d1=0.6216 acc_d2=0.5030 acc_d3=0.4061 acc_d4=0.3280 acc_d5=0.2648 acc_d6=0.2156 acc_d7=0.1773

Training:  94%|█████████▍| 10451/11082 [33:36:34<2:02:21, 11.63s/step, acc=0.5072, loss=2.0134, lr=1.07e-05]
Training:  94%|█████████▍| 10452/11082 [33:36:45<2:01:38, 11.59s/step, acc=0.5072, loss=2.0134, lr=1.07e-05]
Training:  94%|█████████▍| 10453/11082 [33:36:57<2:00:30, 11.49s/step, acc=0.5072, loss=2.0134, lr=1.07e-05]
Training:  94%|█████████▍| 10454/11082 [33:37:08<2:00:16, 11.49s/step, acc=0.5072, loss=2.0134, lr=1.07e-05]
Training:  94%|█████████▍| 10455/11082 [33:37:20<2:01:15, 11.60s/step, acc=0.5072, loss=2.0134, lr=1.07e-05]
Training:  94%|█████████▍| 10456/11082 [33:37:31<2:00:02, 11.51s/step, acc=0.5072, loss=2.0134, lr=1.07e-05]
Training:  94%|█████████▍| 10457/11082 [33:37:43<1:59:26, 11.47s/step, acc=0.5072, loss=2.0134, lr=1.07e-05]
Training:  94%|█████████▍| 10458/11082 [33:37:55<2:00:46, 11.61s/step, acc=0.5072, loss=2.0134, lr=1.07e-05]
Training:  94%|█████████▍| 10459/11082 [33:38:06<1:59:38, 11.52s/step, acc=0.5072, loss=2.0134, lr=1.07e-05]
Training:  94%|█████████▍| 10460/11082 [33:38:18<2:00:02, 11.58s/step, acc=0.5072, loss=2.0134, lr=1.07e-05]
Training:  94%|█████████▍| 10461/11082 [33:38:29<1:58:05, 11.41s/step, acc=0.5072, loss=2.0134, lr=1.07e-05]
Training:  94%|█████████▍| 10462/11082 [33:38:40<1:58:59, 11.52s/step, acc=0.5072, loss=2.0134, lr=1.07e-05]
Training:  94%|█████████▍| 10463/11082 [33:38:52<1:59:23, 11.57s/step, acc=0.5072, loss=2.0134, lr=1.07e-05]
Training:  94%|█████████▍| 10464/11082 [33:39:04<1:58:42, 11.53s/step, acc=0.5072, loss=2.0134, lr=1.07e-05]
Training:  94%|█████████▍| 10465/11082 [33:39:15<1:58:47, 11.55s/step, acc=0.5072, loss=2.0134, lr=1.07e-05]
Training:  94%|█████████▍| 10466/11082 [33:39:27<1:58:41, 11.56s/step, acc=0.5072, loss=2.0134, lr=1.07e-05]
Training:  94%|█████████▍| 10467/11082 [33:39:39<1:59:24, 11.65s/step, acc=0.5072, loss=2.0134, lr=1.07e-05]
Training:  94%|█████████▍| 10468/11082 [33:39:50<1:58:25, 11.57s/step, acc=0.5072, loss=2.0134, lr=1.07e-05]
Training:  94%|█████████▍| 10469/11082 [33:40:01<1:56:55, 11.44s/step, acc=0.5072, loss=2.0134, lr=1.07e-05]
Training:  94%|█████████▍| 10470/11082 [33:40:13<1:57:26, 11.51s/step, acc=0.5072, loss=2.0134, lr=1.07e-05]
Training:  94%|█████████▍| 10471/11082 [33:40:25<1:57:50, 11.57s/step, acc=0.5072, loss=2.0134, lr=1.07e-05]
Training:  94%|█████████▍| 10472/11082 [33:40:36<1:58:16, 11.63s/step, acc=0.5072, loss=2.0134, lr=1.07e-05]
Training:  95%|█████████▍| 10473/11082 [33:40:48<1:57:19, 11.56s/step, acc=0.5072, loss=2.0134, lr=1.07e-05]
Training:  95%|█████████▍| 10474/11082 [33:40:59<1:56:36, 11.51s/step, acc=0.5072, loss=2.0134, lr=1.07e-05]
Training:  95%|█████████▍| 10475/11082 [33:41:11<1:57:09, 11.58s/step, acc=0.5072, loss=2.0134, lr=1.07e-05]
Training:  95%|█████████▍| 10476/11082 [33:41:22<1:55:41, 11.45s/step, acc=0.5072, loss=2.0134, lr=1.07e-05]
Training:  95%|█████████▍| 10477/11082 [33:41:33<1:55:04, 11.41s/step, acc=0.5072, loss=2.0134, lr=1.07e-05]
Training:  95%|█████████▍| 10478/11082 [33:41:45<1:56:08, 11.54s/step, acc=0.5072, loss=2.0134, lr=1.07e-05]
Training:  95%|█████████▍| 10479/11082 [33:41:57<1:56:46, 11.62s/step, acc=0.5072, loss=2.0134, lr=1.07e-05]
Training:  95%|█████████▍| 10480/11082 [33:42:08<1:55:55, 11.55s/step, acc=0.5072, loss=2.0134, lr=1.07e-05]
Training:  95%|█████████▍| 10481/11082 [33:42:20<1:56:17, 11.61s/step, acc=0.5072, loss=2.0134, lr=1.07e-05]
Training:  95%|█████████▍| 10482/11082 [33:42:32<1:56:21, 11.64s/step, acc=0.5072, loss=2.0134, lr=1.07e-05]
Training:  95%|█████████▍| 10483/11082 [33:42:42<1:53:27, 11.36s/step, acc=0.5072, loss=2.0134, lr=1.07e-05]
Training:  95%|█████████▍| 10484/11082 [33:42:55<1:56:23, 11.68s/step, acc=0.5072, loss=2.0134, lr=1.07e-05]
Training:  95%|█████████▍| 10485/11082 [33:43:06<1:54:43, 11.53s/step, acc=0.5072, loss=2.0134, lr=1.07e-05]
Training:  95%|█████████▍| 10486/11082 [33:43:18<1:54:45, 11.55s/step, acc=0.5072, loss=2.0134, lr=1.07e-05]
Training:  95%|█████████▍| 10487/11082 [33:43:29<1:54:11, 11.52s/step, acc=0.5072, loss=2.0134, lr=1.07e-05]
Training:  95%|█████████▍| 10488/11082 [33:43:41<1:54:03, 11.52s/step, acc=0.5072, loss=2.0134, lr=1.07e-05]
Training:  95%|█████████▍| 10489/11082 [33:43:52<1:52:16, 11.36s/step, acc=0.5072, loss=2.0134, lr=1.07e-05]
Training:  95%|█████████▍| 10490/11082 [33:44:04<1:54:10, 11.57s/step, acc=0.5072, loss=2.0134, lr=1.07e-05]
Training:  95%|█████████▍| 10491/11082 [33:44:15<1:54:04, 11.58s/step, acc=0.5072, loss=2.0134, lr=1.07e-05]
Training:  95%|█████████▍| 10492/11082 [33:44:27<1:53:42, 11.56s/step, acc=0.5072, loss=2.0134, lr=1.07e-05]
Training:  95%|█████████▍| 10493/11082 [33:44:38<1:51:56, 11.40s/step, acc=0.5072, loss=2.0134, lr=1.07e-05]
Training:  95%|█████████▍| 10494/11082 [33:44:50<1:53:26, 11.58s/step, acc=0.5072, loss=2.0134, lr=1.07e-05]
Training:  95%|█████████▍| 10495/11082 [33:45:01<1:51:17, 11.38s/step, acc=0.5072, loss=2.0134, lr=1.07e-05]
Training:  95%|█████████▍| 10496/11082 [33:45:13<1:52:46, 11.55s/step, acc=0.5072, loss=2.0134, lr=1.07e-05]
Training:  95%|█████████▍| 10497/11082 [33:45:24<1:51:51, 11.47s/step, acc=0.5072, loss=2.0134, lr=1.07e-05]
Training:  95%|█████████▍| 10498/11082 [33:45:36<1:52:18, 11.54s/step, acc=0.5072, loss=2.0134, lr=1.07e-05]
Training:  95%|█████████▍| 10499/11082 [33:45:47<1:52:24, 11.57s/step, acc=0.5072, loss=2.0134, lr=1.07e-05]
Training:  95%|█████████▍| 10500/11082 [33:45:59<1:52:23, 11.59s/step, acc=0.5072, loss=2.0134, lr=1.07e-05]
Training:  95%|█████████▍| 10500/11082 [33:46:06<1:52:23, 11.59s/step, acc=0.5084, loss=2.0088, lr=1.06e-05]2026-06-20 00:39:34 | INFO | __main__ | epoch=1 step=10500 train_loss=2.008789 train_acc=0.508448 lr=1.063e-05 grad_norm=0.5039 acc_d0=0.7642 acc_d1=0.6221 acc_d2=0.5020 acc_d3=0.4056 acc_d4=0.3278 acc_d5=0.2647 acc_d6=0.2162 acc_d7=0.1763

Training:  95%|█████████▍| 10501/11082 [33:46:11<1:52:45, 11.64s/step, acc=0.5084, loss=2.0088, lr=1.06e-05]
Training:  95%|█████████▍| 10502/11082 [33:46:22<1:50:21, 11.42s/step, acc=0.5084, loss=2.0088, lr=1.06e-05]
Training:  95%|█████████▍| 10503/11082 [33:46:34<1:53:08, 11.72s/step, acc=0.5084, loss=2.0088, lr=1.06e-05]
Training:  95%|█████████▍| 10504/11082 [33:46:45<1:52:05, 11.64s/step, acc=0.5084, loss=2.0088, lr=1.06e-05]
Training:  95%|█████████▍| 10505/11082 [33:46:57<1:51:39, 11.61s/step, acc=0.5084, loss=2.0088, lr=1.06e-05]
Training:  95%|█████████▍| 10506/11082 [33:47:09<1:51:45, 11.64s/step, acc=0.5084, loss=2.0088, lr=1.06e-05]
Training:  95%|█████████▍| 10507/11082 [33:47:20<1:51:11, 11.60s/step, acc=0.5084, loss=2.0088, lr=1.06e-05]
Training:  95%|█████████▍| 10508/11082 [33:47:32<1:50:48, 11.58s/step, acc=0.5084, loss=2.0088, lr=1.06e-05]
Training:  95%|█████████▍| 10509/11082 [33:47:43<1:48:48, 11.39s/step, acc=0.5084, loss=2.0088, lr=1.06e-05]
Training:  95%|█████████▍| 10510/11082 [33:47:55<1:50:51, 11.63s/step, acc=0.5084, loss=2.0088, lr=1.06e-05]
Training:  95%|█████████▍| 10511/11082 [33:48:06<1:50:02, 11.56s/step, acc=0.5084, loss=2.0088, lr=1.06e-05]
Training:  95%|█████████▍| 10512/11082 [33:48:18<1:49:03, 11.48s/step, acc=0.5084, loss=2.0088, lr=1.06e-05]
Training:  95%|█████████▍| 10513/11082 [33:48:29<1:48:33, 11.45s/step, acc=0.5084, loss=2.0088, lr=1.06e-05]
Training:  95%|█████████▍| 10514/11082 [33:48:41<1:50:12, 11.64s/step, acc=0.5084, loss=2.0088, lr=1.06e-05]
Training:  95%|█████████▍| 10515/11082 [33:48:52<1:49:05, 11.54s/step, acc=0.5084, loss=2.0088, lr=1.06e-05]
Training:  95%|█████████▍| 10516/11082 [33:49:05<1:50:26, 11.71s/step, acc=0.5084, loss=2.0088, lr=1.06e-05]
Training:  95%|█████████▍| 10517/11082 [33:49:16<1:48:35, 11.53s/step, acc=0.5084, loss=2.0088, lr=1.06e-05]
Training:  95%|█████████▍| 10518/11082 [33:49:27<1:48:49, 11.58s/step, acc=0.5084, loss=2.0088, lr=1.06e-05]
Training:  95%|█████████▍| 10519/11082 [33:49:39<1:48:46, 11.59s/step, acc=0.5084, loss=2.0088, lr=1.06e-05]
Training:  95%|█████████▍| 10520/11082 [33:49:51<1:49:13, 11.66s/step, acc=0.5084, loss=2.0088, lr=1.06e-05]
Training:  95%|█████████▍| 10521/11082 [33:50:02<1:47:28, 11.49s/step, acc=0.5084, loss=2.0088, lr=1.06e-05]
Training:  95%|█████████▍| 10522/11082 [33:50:13<1:47:14, 11.49s/step, acc=0.5084, loss=2.0088, lr=1.06e-05]
Training:  95%|█████████▍| 10523/11082 [33:50:25<1:48:00, 11.59s/step, acc=0.5084, loss=2.0088, lr=1.06e-05]
Training:  95%|█████████▍| 10524/11082 [33:50:37<1:48:32, 11.67s/step, acc=0.5084, loss=2.0088, lr=1.06e-05]
Training:  95%|█████████▍| 10525/11082 [33:50:49<1:48:16, 11.66s/step, acc=0.5084, loss=2.0088, lr=1.06e-05]
Training:  95%|█████████▍| 10526/11082 [33:51:00<1:46:50, 11.53s/step, acc=0.5084, loss=2.0088, lr=1.06e-05]
Training:  95%|█████████▍| 10527/11082 [33:51:12<1:47:07, 11.58s/step, acc=0.5084, loss=2.0088, lr=1.06e-05]
Training:  95%|█████████▌| 10528/11082 [33:51:23<1:47:34, 11.65s/step, acc=0.5084, loss=2.0088, lr=1.06e-05]
Training:  95%|█████████▌| 10529/11082 [33:51:35<1:46:47, 11.59s/step, acc=0.5084, loss=2.0088, lr=1.06e-05]
Training:  95%|█████████▌| 10530/11082 [33:51:47<1:46:49, 11.61s/step, acc=0.5084, loss=2.0088, lr=1.06e-05]
Training:  95%|█████████▌| 10531/11082 [33:51:58<1:45:41, 11.51s/step, acc=0.5084, loss=2.0088, lr=1.06e-05]
Training:  95%|█████████▌| 10532/11082 [33:52:10<1:46:38, 11.63s/step, acc=0.5084, loss=2.0088, lr=1.06e-05]
Training:  95%|█████████▌| 10533/11082 [33:52:21<1:45:48, 11.56s/step, acc=0.5084, loss=2.0088, lr=1.06e-05]
Training:  95%|█████████▌| 10534/11082 [33:52:32<1:43:23, 11.32s/step, acc=0.5084, loss=2.0088, lr=1.06e-05]
Training:  95%|█████████▌| 10535/11082 [33:52:44<1:46:00, 11.63s/step, acc=0.5084, loss=2.0088, lr=1.06e-05]
Training:  95%|█████████▌| 10536/11082 [33:52:56<1:46:20, 11.69s/step, acc=0.5084, loss=2.0088, lr=1.06e-05]
Training:  95%|█████████▌| 10537/11082 [33:53:08<1:46:13, 11.69s/step, acc=0.5084, loss=2.0088, lr=1.06e-05]
Training:  95%|█████████▌| 10538/11082 [33:53:19<1:44:25, 11.52s/step, acc=0.5084, loss=2.0088, lr=1.06e-05]
Training:  95%|█████████▌| 10539/11082 [33:53:31<1:45:20, 11.64s/step, acc=0.5084, loss=2.0088, lr=1.06e-05]
Training:  95%|█████████▌| 10540/11082 [33:53:42<1:44:33, 11.58s/step, acc=0.5084, loss=2.0088, lr=1.06e-05]
Training:  95%|█████████▌| 10541/11082 [33:53:54<1:43:57, 11.53s/step, acc=0.5084, loss=2.0088, lr=1.06e-05]
Training:  95%|█████████▌| 10542/11082 [33:54:05<1:43:54, 11.54s/step, acc=0.5084, loss=2.0088, lr=1.06e-05]
Training:  95%|█████████▌| 10543/11082 [33:54:16<1:42:30, 11.41s/step, acc=0.5084, loss=2.0088, lr=1.06e-05]
Training:  95%|█████████▌| 10544/11082 [33:54:28<1:43:18, 11.52s/step, acc=0.5084, loss=2.0088, lr=1.06e-05]
Training:  95%|█████████▌| 10545/11082 [33:54:39<1:41:54, 11.39s/step, acc=0.5084, loss=2.0088, lr=1.06e-05]
Training:  95%|█████████▌| 10546/11082 [33:54:51<1:43:36, 11.60s/step, acc=0.5084, loss=2.0088, lr=1.06e-05]
Training:  95%|█████████▌| 10547/11082 [33:55:03<1:42:57, 11.55s/step, acc=0.5084, loss=2.0088, lr=1.06e-05]
Training:  95%|█████████▌| 10548/11082 [33:55:14<1:42:57, 11.57s/step, acc=0.5084, loss=2.0088, lr=1.06e-05]
Training:  95%|█████████▌| 10549/11082 [33:55:26<1:42:45, 11.57s/step, acc=0.5084, loss=2.0088, lr=1.06e-05]
Training:  95%|█████████▌| 10550/11082 [33:55:37<1:41:49, 11.48s/step, acc=0.5084, loss=2.0088, lr=1.06e-05]
Training:  95%|█████████▌| 10550/11082 [33:55:44<1:41:49, 11.48s/step, acc=0.5070, loss=2.0221, lr=1.05e-05]2026-06-20 00:49:12 | INFO | __main__ | epoch=1 step=10550 train_loss=2.022118 train_acc=0.506961 lr=1.053e-05 grad_norm=0.4980 acc_d0=0.7677 acc_d1=0.6257 acc_d2=0.5072 acc_d3=0.4110 acc_d4=0.3321 acc_d5=0.2693 acc_d6=0.2183 acc_d7=0.1763

Training:  95%|█████████▌| 10551/11082 [33:55:49<1:41:28, 11.47s/step, acc=0.5070, loss=2.0221, lr=1.05e-05]
Training:  95%|█████████▌| 10552/11082 [33:56:00<1:41:24, 11.48s/step, acc=0.5070, loss=2.0221, lr=1.05e-05]
Training:  95%|█████████▌| 10553/11082 [33:56:12<1:42:26, 11.62s/step, acc=0.5070, loss=2.0221, lr=1.05e-05]
Training:  95%|█████████▌| 10554/11082 [33:56:23<1:41:25, 11.52s/step, acc=0.5070, loss=2.0221, lr=1.05e-05]
Training:  95%|█████████▌| 10555/11082 [33:56:35<1:41:21, 11.54s/step, acc=0.5070, loss=2.0221, lr=1.05e-05]
Training:  95%|█████████▌| 10556/11082 [33:56:47<1:41:21, 11.56s/step, acc=0.5070, loss=2.0221, lr=1.05e-05]
Training:  95%|█████████▌| 10557/11082 [33:56:58<1:41:25, 11.59s/step, acc=0.5070, loss=2.0221, lr=1.05e-05]
Training:  95%|█████████▌| 10558/11082 [33:57:10<1:42:16, 11.71s/step, acc=0.5070, loss=2.0221, lr=1.05e-05]
Training:  95%|█████████▌| 10559/11082 [33:57:21<1:41:00, 11.59s/step, acc=0.5070, loss=2.0221, lr=1.05e-05]
Training:  95%|█████████▌| 10560/11082 [33:57:33<1:40:46, 11.58s/step, acc=0.5070, loss=2.0221, lr=1.05e-05]
Training:  95%|█████████▌| 10561/11082 [33:57:44<1:39:05, 11.41s/step, acc=0.5070, loss=2.0221, lr=1.05e-05]
Training:  95%|█████████▌| 10562/11082 [33:57:56<1:40:16, 11.57s/step, acc=0.5070, loss=2.0221, lr=1.05e-05]
Training:  95%|█████████▌| 10563/11082 [33:58:07<1:38:41, 11.41s/step, acc=0.5070, loss=2.0221, lr=1.05e-05]
Training:  95%|█████████▌| 10564/11082 [33:58:19<1:40:09, 11.60s/step, acc=0.5070, loss=2.0221, lr=1.05e-05]
Training:  95%|█████████▌| 10565/11082 [33:58:30<1:39:17, 11.52s/step, acc=0.5070, loss=2.0221, lr=1.05e-05]
Training:  95%|█████████▌| 10566/11082 [33:58:42<1:39:06, 11.52s/step, acc=0.5070, loss=2.0221, lr=1.05e-05]
Training:  95%|█████████▌| 10567/11082 [33:58:54<1:39:15, 11.56s/step, acc=0.5070, loss=2.0221, lr=1.05e-05]
Training:  95%|█████████▌| 10568/11082 [33:59:05<1:38:37, 11.51s/step, acc=0.5070, loss=2.0221, lr=1.05e-05]
Training:  95%|█████████▌| 10569/11082 [33:59:16<1:37:32, 11.41s/step, acc=0.5070, loss=2.0221, lr=1.05e-05]
Training:  95%|█████████▌| 10570/11082 [33:59:28<1:39:06, 11.61s/step, acc=0.5070, loss=2.0221, lr=1.05e-05]
Training:  95%|█████████▌| 10571/11082 [33:59:40<1:40:03, 11.75s/step, acc=0.5070, loss=2.0221, lr=1.05e-05]
Training:  95%|█████████▌| 10572/11082 [33:59:52<1:39:02, 11.65s/step, acc=0.5070, loss=2.0221, lr=1.05e-05]
Training:  95%|█████████▌| 10573/11082 [34:00:03<1:38:10, 11.57s/step, acc=0.5070, loss=2.0221, lr=1.05e-05]
Training:  95%|█████████▌| 10574/11082 [34:00:15<1:37:49, 11.55s/step, acc=0.5070, loss=2.0221, lr=1.05e-05]
Training:  95%|█████████▌| 10575/11082 [34:00:26<1:38:07, 11.61s/step, acc=0.5070, loss=2.0221, lr=1.05e-05]
Training:  95%|█████████▌| 10576/11082 [34:00:38<1:37:29, 11.56s/step, acc=0.5070, loss=2.0221, lr=1.05e-05]
Training:  95%|█████████▌| 10577/11082 [34:00:49<1:37:13, 11.55s/step, acc=0.5070, loss=2.0221, lr=1.05e-05]
Training:  95%|█████████▌| 10578/11082 [34:01:01<1:36:46, 11.52s/step, acc=0.5070, loss=2.0221, lr=1.05e-05]
Training:  95%|█████████▌| 10579/11082 [34:01:12<1:35:54, 11.44s/step, acc=0.5070, loss=2.0221, lr=1.05e-05]
Training:  95%|█████████▌| 10580/11082 [34:01:24<1:35:58, 11.47s/step, acc=0.5070, loss=2.0221, lr=1.05e-05]
Training:  95%|█████████▌| 10581/11082 [34:01:36<1:36:54, 11.61s/step, acc=0.5070, loss=2.0221, lr=1.05e-05]
Training:  95%|█████████▌| 10582/11082 [34:01:47<1:36:50, 11.62s/step, acc=0.5070, loss=2.0221, lr=1.05e-05]
Training:  95%|█████████▌| 10583/11082 [34:01:59<1:37:04, 11.67s/step, acc=0.5070, loss=2.0221, lr=1.05e-05]
Training:  96%|█████████▌| 10584/11082 [34:02:11<1:36:47, 11.66s/step, acc=0.5070, loss=2.0221, lr=1.05e-05]
Training:  96%|█████████▌| 10585/11082 [34:02:23<1:37:29, 11.77s/step, acc=0.5070, loss=2.0221, lr=1.05e-05]
Training:  96%|█████████▌| 10586/11082 [34:02:34<1:36:39, 11.69s/step, acc=0.5070, loss=2.0221, lr=1.05e-05]
Training:  96%|█████████▌| 10587/11082 [34:02:46<1:35:45, 11.61s/step, acc=0.5070, loss=2.0221, lr=1.05e-05]
Training:  96%|█████████▌| 10588/11082 [34:02:57<1:35:53, 11.65s/step, acc=0.5070, loss=2.0221, lr=1.05e-05]
Training:  96%|█████████▌| 10589/11082 [34:03:09<1:35:42, 11.65s/step, acc=0.5070, loss=2.0221, lr=1.05e-05]
Training:  96%|█████████▌| 10590/11082 [34:03:20<1:35:03, 11.59s/step, acc=0.5070, loss=2.0221, lr=1.05e-05]
Training:  96%|█████████▌| 10591/11082 [34:03:32<1:35:18, 11.65s/step, acc=0.5070, loss=2.0221, lr=1.05e-05]
Training:  96%|█████████▌| 10592/11082 [34:03:43<1:33:28, 11.45s/step, acc=0.5070, loss=2.0221, lr=1.05e-05]
Training:  96%|█████████▌| 10593/11082 [34:03:55<1:33:40, 11.49s/step, acc=0.5070, loss=2.0221, lr=1.05e-05]
Training:  96%|█████████▌| 10594/11082 [34:04:07<1:34:59, 11.68s/step, acc=0.5070, loss=2.0221, lr=1.05e-05]
Training:  96%|█████████▌| 10595/11082 [34:04:18<1:34:05, 11.59s/step, acc=0.5070, loss=2.0221, lr=1.05e-05]
Training:  96%|█████████▌| 10596/11082 [34:04:30<1:33:51, 11.59s/step, acc=0.5070, loss=2.0221, lr=1.05e-05]
Training:  96%|█████████▌| 10597/11082 [34:04:41<1:32:55, 11.50s/step, acc=0.5070, loss=2.0221, lr=1.05e-05]
Training:  96%|█████████▌| 10598/11082 [34:04:53<1:33:07, 11.54s/step, acc=0.5070, loss=2.0221, lr=1.05e-05]
Training:  96%|█████████▌| 10599/11082 [34:05:04<1:33:13, 11.58s/step, acc=0.5070, loss=2.0221, lr=1.05e-05]
Training:  96%|█████████▌| 10600/11082 [34:05:16<1:32:09, 11.47s/step, acc=0.5070, loss=2.0221, lr=1.05e-05]
Training:  96%|█████████▌| 10600/11082 [34:05:23<1:32:09, 11.47s/step, acc=0.5077, loss=2.0107, lr=1.04e-05]2026-06-20 00:58:51 | INFO | __main__ | epoch=1 step=10600 train_loss=2.010651 train_acc=0.507714 lr=1.043e-05 grad_norm=0.4863 acc_d0=0.7657 acc_d1=0.6229 acc_d2=0.5042 acc_d3=0.4063 acc_d4=0.3294 acc_d5=0.2655 acc_d6=0.2154 acc_d7=0.1773

Training:  96%|█████████▌| 10601/11082 [34:05:27<1:32:30, 11.54s/step, acc=0.5077, loss=2.0107, lr=1.04e-05]
Training:  96%|█████████▌| 10602/11082 [34:05:39<1:33:01, 11.63s/step, acc=0.5077, loss=2.0107, lr=1.04e-05]
Training:  96%|█████████▌| 10603/11082 [34:05:50<1:31:47, 11.50s/step, acc=0.5077, loss=2.0107, lr=1.04e-05]
Training:  96%|█████████▌| 10604/11082 [34:06:03<1:33:16, 11.71s/step, acc=0.5077, loss=2.0107, lr=1.04e-05]
Training:  96%|█████████▌| 10605/11082 [34:06:14<1:32:32, 11.64s/step, acc=0.5077, loss=2.0107, lr=1.04e-05]
Training:  96%|█████████▌| 10606/11082 [34:06:26<1:32:25, 11.65s/step, acc=0.5077, loss=2.0107, lr=1.04e-05]
Training:  96%|█████████▌| 10607/11082 [34:06:37<1:31:19, 11.54s/step, acc=0.5077, loss=2.0107, lr=1.04e-05]
Training:  96%|█████████▌| 10608/11082 [34:06:49<1:31:42, 11.61s/step, acc=0.5077, loss=2.0107, lr=1.04e-05]
Training:  96%|█████████▌| 10609/11082 [34:07:01<1:32:15, 11.70s/step, acc=0.5077, loss=2.0107, lr=1.04e-05]
Training:  96%|█████████▌| 10610/11082 [34:07:13<1:32:17, 11.73s/step, acc=0.5077, loss=2.0107, lr=1.04e-05]
Training:  96%|█████████▌| 10611/11082 [34:07:24<1:30:51, 11.58s/step, acc=0.5077, loss=2.0107, lr=1.04e-05]
Training:  96%|█████████▌| 10612/11082 [34:07:35<1:30:36, 11.57s/step, acc=0.5077, loss=2.0107, lr=1.04e-05]
Training:  96%|█████████▌| 10613/11082 [34:07:47<1:31:00, 11.64s/step, acc=0.5077, loss=2.0107, lr=1.04e-05]
Training:  96%|█████████▌| 10614/11082 [34:07:59<1:30:46, 11.64s/step, acc=0.5077, loss=2.0107, lr=1.04e-05]
Training:  96%|█████████▌| 10615/11082 [34:08:10<1:30:55, 11.68s/step, acc=0.5077, loss=2.0107, lr=1.04e-05]
Training:  96%|█████████▌| 10616/11082 [34:08:22<1:30:10, 11.61s/step, acc=0.5077, loss=2.0107, lr=1.04e-05]
Training:  96%|█████████▌| 10617/11082 [34:08:33<1:29:25, 11.54s/step, acc=0.5077, loss=2.0107, lr=1.04e-05]
Training:  96%|█████████▌| 10618/11082 [34:08:45<1:29:46, 11.61s/step, acc=0.5077, loss=2.0107, lr=1.04e-05]
Training:  96%|█████████▌| 10619/11082 [34:08:56<1:28:24, 11.46s/step, acc=0.5077, loss=2.0107, lr=1.04e-05]
Training:  96%|█████████▌| 10620/11082 [34:09:08<1:29:37, 11.64s/step, acc=0.5077, loss=2.0107, lr=1.04e-05]
Training:  96%|█████████▌| 10621/11082 [34:09:20<1:29:18, 11.62s/step, acc=0.5077, loss=2.0107, lr=1.04e-05]
Training:  96%|█████████▌| 10622/11082 [34:09:32<1:29:17, 11.65s/step, acc=0.5077, loss=2.0107, lr=1.04e-05]
Training:  96%|█████████▌| 10623/11082 [34:09:43<1:29:06, 11.65s/step, acc=0.5077, loss=2.0107, lr=1.04e-05]
Training:  96%|█████████▌| 10624/11082 [34:09:55<1:28:24, 11.58s/step, acc=0.5077, loss=2.0107, lr=1.04e-05]
Training:  96%|█████████▌| 10625/11082 [34:10:07<1:28:56, 11.68s/step, acc=0.5077, loss=2.0107, lr=1.04e-05]
Training:  96%|█████████▌| 10626/11082 [34:10:18<1:27:40, 11.54s/step, acc=0.5077, loss=2.0107, lr=1.04e-05]
Training:  96%|█████████▌| 10627/11082 [34:10:29<1:27:54, 11.59s/step, acc=0.5077, loss=2.0107, lr=1.04e-05]
Training:  96%|█████████▌| 10628/11082 [34:10:41<1:27:08, 11.52s/step, acc=0.5077, loss=2.0107, lr=1.04e-05]
Training:  96%|█████████▌| 10629/11082 [34:10:53<1:27:40, 11.61s/step, acc=0.5077, loss=2.0107, lr=1.04e-05]
Training:  96%|█████████▌| 10630/11082 [34:11:04<1:27:40, 11.64s/step, acc=0.5077, loss=2.0107, lr=1.04e-05]
Training:  96%|█████████▌| 10631/11082 [34:11:16<1:26:32, 11.51s/step, acc=0.5077, loss=2.0107, lr=1.04e-05]
Training:  96%|█████████▌| 10632/11082 [34:11:28<1:27:40, 11.69s/step, acc=0.5077, loss=2.0107, lr=1.04e-05]
Training:  96%|█████████▌| 10633/11082 [34:11:38<1:25:27, 11.42s/step, acc=0.5077, loss=2.0107, lr=1.04e-05]
Training:  96%|█████████▌| 10634/11082 [34:11:50<1:25:46, 11.49s/step, acc=0.5077, loss=2.0107, lr=1.04e-05]
Training:  96%|█████████▌| 10635/11082 [34:12:02<1:27:14, 11.71s/step, acc=0.5077, loss=2.0107, lr=1.04e-05]
Training:  96%|█████████▌| 10636/11082 [34:12:13<1:25:51, 11.55s/step, acc=0.5077, loss=2.0107, lr=1.04e-05]
Training:  96%|█████████▌| 10637/11082 [34:12:25<1:26:24, 11.65s/step, acc=0.5077, loss=2.0107, lr=1.04e-05]
Training:  96%|█████████▌| 10638/11082 [34:12:37<1:26:06, 11.64s/step, acc=0.5077, loss=2.0107, lr=1.04e-05]
Training:  96%|█████████▌| 10639/11082 [34:12:48<1:24:47, 11.48s/step, acc=0.5077, loss=2.0107, lr=1.04e-05]
Training:  96%|█████████▌| 10640/11082 [34:13:00<1:25:55, 11.66s/step, acc=0.5077, loss=2.0107, lr=1.04e-05]
Training:  96%|█████████▌| 10641/11082 [34:13:11<1:24:10, 11.45s/step, acc=0.5077, loss=2.0107, lr=1.04e-05]
Training:  96%|█████████▌| 10642/11082 [34:13:22<1:23:31, 11.39s/step, acc=0.5077, loss=2.0107, lr=1.04e-05]
Training:  96%|█████████▌| 10643/11082 [34:13:34<1:24:53, 11.60s/step, acc=0.5077, loss=2.0107, lr=1.04e-05]
Training:  96%|█████████▌| 10644/11082 [34:13:46<1:23:38, 11.46s/step, acc=0.5077, loss=2.0107, lr=1.04e-05]
Training:  96%|█████████▌| 10645/11082 [34:13:57<1:23:20, 11.44s/step, acc=0.5077, loss=2.0107, lr=1.04e-05]
Training:  96%|█████████▌| 10646/11082 [34:14:09<1:23:55, 11.55s/step, acc=0.5077, loss=2.0107, lr=1.04e-05]
Training:  96%|█████████▌| 10647/11082 [34:14:21<1:24:18, 11.63s/step, acc=0.5077, loss=2.0107, lr=1.04e-05]
Training:  96%|█████████▌| 10648/11082 [34:14:32<1:23:09, 11.50s/step, acc=0.5077, loss=2.0107, lr=1.04e-05]
Training:  96%|█████████▌| 10649/11082 [34:14:43<1:22:37, 11.45s/step, acc=0.5077, loss=2.0107, lr=1.04e-05]
Training:  96%|█████████▌| 10650/11082 [34:14:56<1:24:55, 11.79s/step, acc=0.5077, loss=2.0107, lr=1.04e-05]
Training:  96%|█████████▌| 10650/11082 [34:15:02<1:24:55, 11.79s/step, acc=0.5080, loss=2.0118, lr=1.03e-05]2026-06-20 01:08:30 | INFO | __main__ | epoch=1 step=10650 train_loss=2.011760 train_acc=0.507993 lr=1.035e-05 grad_norm=0.5664 acc_d0=0.7640 acc_d1=0.6206 acc_d2=0.5016 acc_d3=0.4049 acc_d4=0.3270 acc_d5=0.2636 acc_d6=0.2161 acc_d7=0.1762

Training:  96%|█████████▌| 10651/11082 [34:15:06<1:22:26, 11.48s/step, acc=0.5080, loss=2.0118, lr=1.03e-05]
Training:  96%|█████████▌| 10652/11082 [34:15:19<1:24:08, 11.74s/step, acc=0.5080, loss=2.0118, lr=1.03e-05]
Training:  96%|█████████▌| 10653/11082 [34:15:30<1:23:27, 11.67s/step, acc=0.5080, loss=2.0118, lr=1.03e-05]
Training:  96%|█████████▌| 10654/11082 [34:15:42<1:22:57, 11.63s/step, acc=0.5080, loss=2.0118, lr=1.03e-05]
Training:  96%|█████████▌| 10655/11082 [34:15:53<1:21:46, 11.49s/step, acc=0.5080, loss=2.0118, lr=1.03e-05]
Training:  96%|█████████▌| 10656/11082 [34:16:05<1:22:10, 11.57s/step, acc=0.5080, loss=2.0118, lr=1.03e-05]
Training:  96%|█████████▌| 10657/11082 [34:16:16<1:21:00, 11.44s/step, acc=0.5080, loss=2.0118, lr=1.03e-05]
Training:  96%|█████████▌| 10658/11082 [34:16:28<1:21:19, 11.51s/step, acc=0.5080, loss=2.0118, lr=1.03e-05]
Training:  96%|█████████▌| 10659/11082 [34:16:39<1:20:42, 11.45s/step, acc=0.5080, loss=2.0118, lr=1.03e-05]
Training:  96%|█████████▌| 10660/11082 [34:16:50<1:20:43, 11.48s/step, acc=0.5080, loss=2.0118, lr=1.03e-05]
Training:  96%|█████████▌| 10661/11082 [34:17:02<1:20:51, 11.52s/step, acc=0.5080, loss=2.0118, lr=1.03e-05]
Training:  96%|█████████▌| 10662/11082 [34:17:14<1:22:16, 11.75s/step, acc=0.5080, loss=2.0118, lr=1.03e-05]
Training:  96%|█████████▌| 10663/11082 [34:17:26<1:21:40, 11.70s/step, acc=0.5080, loss=2.0118, lr=1.03e-05]
Training:  96%|█████████▌| 10664/11082 [34:17:37<1:20:57, 11.62s/step, acc=0.5080, loss=2.0118, lr=1.03e-05]
Training:  96%|█████████▌| 10665/11082 [34:17:49<1:20:50, 11.63s/step, acc=0.5080, loss=2.0118, lr=1.03e-05]
Training:  96%|█████████▌| 10666/11082 [34:18:01<1:20:30, 11.61s/step, acc=0.5080, loss=2.0118, lr=1.03e-05]
Training:  96%|█████████▋| 10667/11082 [34:18:12<1:20:29, 11.64s/step, acc=0.5080, loss=2.0118, lr=1.03e-05]
Training:  96%|█████████▋| 10668/11082 [34:18:24<1:20:11, 11.62s/step, acc=0.5080, loss=2.0118, lr=1.03e-05]
Training:  96%|█████████▋| 10669/11082 [34:18:36<1:20:36, 11.71s/step, acc=0.5080, loss=2.0118, lr=1.03e-05]
Training:  96%|█████████▋| 10670/11082 [34:18:47<1:19:11, 11.53s/step, acc=0.5080, loss=2.0118, lr=1.03e-05]
Training:  96%|█████████▋| 10671/11082 [34:18:59<1:19:12, 11.56s/step, acc=0.5080, loss=2.0118, lr=1.03e-05]
Training:  96%|█████████▋| 10672/11082 [34:19:10<1:19:24, 11.62s/step, acc=0.5080, loss=2.0118, lr=1.03e-05]
Training:  96%|█████████▋| 10673/11082 [34:19:22<1:19:07, 11.61s/step, acc=0.5080, loss=2.0118, lr=1.03e-05]
Training:  96%|█████████▋| 10674/11082 [34:19:34<1:19:47, 11.73s/step, acc=0.5080, loss=2.0118, lr=1.03e-05]
Training:  96%|█████████▋| 10675/11082 [34:19:45<1:18:57, 11.64s/step, acc=0.5080, loss=2.0118, lr=1.03e-05]
Training:  96%|█████████▋| 10676/11082 [34:19:57<1:18:42, 11.63s/step, acc=0.5080, loss=2.0118, lr=1.03e-05]
Training:  96%|█████████▋| 10677/11082 [34:20:08<1:18:13, 11.59s/step, acc=0.5080, loss=2.0118, lr=1.03e-05]
Training:  96%|█████████▋| 10678/11082 [34:20:20<1:17:20, 11.49s/step, acc=0.5080, loss=2.0118, lr=1.03e-05]
Training:  96%|█████████▋| 10679/11082 [34:20:31<1:17:19, 11.51s/step, acc=0.5080, loss=2.0118, lr=1.03e-05]
Training:  96%|█████████▋| 10680/11082 [34:20:43<1:17:54, 11.63s/step, acc=0.5080, loss=2.0118, lr=1.03e-05]
Training:  96%|█████████▋| 10681/11082 [34:20:55<1:17:58, 11.67s/step, acc=0.5080, loss=2.0118, lr=1.03e-05]
Training:  96%|█████████▋| 10682/11082 [34:21:06<1:17:12, 11.58s/step, acc=0.5080, loss=2.0118, lr=1.03e-05]
Training:  96%|█████████▋| 10683/11082 [34:21:18<1:16:55, 11.57s/step, acc=0.5080, loss=2.0118, lr=1.03e-05]
Training:  96%|█████████▋| 10684/11082 [34:21:29<1:16:06, 11.47s/step, acc=0.5080, loss=2.0118, lr=1.03e-05]
Training:  96%|█████████▋| 10685/11082 [34:21:41<1:16:48, 11.61s/step, acc=0.5080, loss=2.0118, lr=1.03e-05]
Training:  96%|█████████▋| 10686/11082 [34:21:52<1:16:03, 11.52s/step, acc=0.5080, loss=2.0118, lr=1.03e-05]
Training:  96%|█████████▋| 10687/11082 [34:22:04<1:16:45, 11.66s/step, acc=0.5080, loss=2.0118, lr=1.03e-05]
Training:  96%|█████████▋| 10688/11082 [34:22:16<1:16:27, 11.64s/step, acc=0.5080, loss=2.0118, lr=1.03e-05]
Training:  96%|█████████▋| 10689/11082 [34:22:27<1:15:24, 11.51s/step, acc=0.5080, loss=2.0118, lr=1.03e-05]
Training:  96%|█████████▋| 10690/11082 [34:22:39<1:15:26, 11.55s/step, acc=0.5080, loss=2.0118, lr=1.03e-05]
Training:  96%|█████████▋| 10691/11082 [34:22:50<1:15:19, 11.56s/step, acc=0.5080, loss=2.0118, lr=1.03e-05]
Training:  96%|█████████▋| 10692/11082 [34:23:02<1:15:50, 11.67s/step, acc=0.5080, loss=2.0118, lr=1.03e-05]
Training:  96%|█████████▋| 10693/11082 [34:23:14<1:15:12, 11.60s/step, acc=0.5080, loss=2.0118, lr=1.03e-05]
Training:  96%|█████████▋| 10694/11082 [34:23:25<1:15:18, 11.64s/step, acc=0.5080, loss=2.0118, lr=1.03e-05]
Training:  97%|█████████▋| 10695/11082 [34:23:37<1:15:03, 11.64s/step, acc=0.5080, loss=2.0118, lr=1.03e-05]
Training:  97%|█████████▋| 10696/11082 [34:23:49<1:14:33, 11.59s/step, acc=0.5080, loss=2.0118, lr=1.03e-05]
Training:  97%|█████████▋| 10697/11082 [34:24:00<1:13:18, 11.42s/step, acc=0.5080, loss=2.0118, lr=1.03e-05]
Training:  97%|█████████▋| 10698/11082 [34:24:12<1:14:02, 11.57s/step, acc=0.5080, loss=2.0118, lr=1.03e-05]
Training:  97%|█████████▋| 10699/11082 [34:24:23<1:13:33, 11.52s/step, acc=0.5080, loss=2.0118, lr=1.03e-05]
Training:  97%|█████████▋| 10700/11082 [34:24:35<1:14:06, 11.64s/step, acc=0.5080, loss=2.0118, lr=1.03e-05]
Training:  97%|█████████▋| 10700/11082 [34:24:42<1:14:06, 11.64s/step, acc=0.5072, loss=2.0156, lr=1.03e-05]2026-06-20 01:18:10 | INFO | __main__ | epoch=1 step=10700 train_loss=2.015563 train_acc=0.507153 lr=1.027e-05 grad_norm=0.4629 acc_d0=0.7688 acc_d1=0.6246 acc_d2=0.5034 acc_d3=0.4047 acc_d4=0.3259 acc_d5=0.2625 acc_d6=0.2129 acc_d7=0.1743

Training:  97%|█████████▋| 10701/11082 [34:24:46<1:13:11, 11.53s/step, acc=0.5072, loss=2.0156, lr=1.03e-05]
Training:  97%|█████████▋| 10702/11082 [34:24:57<1:12:29, 11.45s/step, acc=0.5072, loss=2.0156, lr=1.03e-05]
Training:  97%|█████████▋| 10703/11082 [34:25:09<1:13:20, 11.61s/step, acc=0.5072, loss=2.0156, lr=1.03e-05]
Training:  97%|█████████▋| 10704/11082 [34:25:21<1:13:33, 11.67s/step, acc=0.5072, loss=2.0156, lr=1.03e-05]
Training:  97%|█████████▋| 10705/11082 [34:25:32<1:11:56, 11.45s/step, acc=0.5072, loss=2.0156, lr=1.03e-05]
Training:  97%|█████████▋| 10706/11082 [34:25:44<1:12:43, 11.61s/step, acc=0.5072, loss=2.0156, lr=1.03e-05]
Training:  97%|█████████▋| 10707/11082 [34:25:55<1:11:57, 11.51s/step, acc=0.5072, loss=2.0156, lr=1.03e-05]
Training:  97%|█████████▋| 10708/11082 [34:26:07<1:12:16, 11.60s/step, acc=0.5072, loss=2.0156, lr=1.03e-05]
Training:  97%|█████████▋| 10709/11082 [34:26:18<1:11:22, 11.48s/step, acc=0.5072, loss=2.0156, lr=1.03e-05]
Training:  97%|█████████▋| 10710/11082 [34:26:30<1:12:04, 11.62s/step, acc=0.5072, loss=2.0156, lr=1.03e-05]
Training:  97%|█████████▋| 10711/11082 [34:26:41<1:10:50, 11.46s/step, acc=0.5072, loss=2.0156, lr=1.03e-05]
Training:  97%|█████████▋| 10712/11082 [34:26:54<1:11:51, 11.65s/step, acc=0.5072, loss=2.0156, lr=1.03e-05]
Training:  97%|█████████▋| 10713/11082 [34:27:05<1:11:40, 11.65s/step, acc=0.5072, loss=2.0156, lr=1.03e-05]
Training:  97%|█████████▋| 10714/11082 [34:27:17<1:11:27, 11.65s/step, acc=0.5072, loss=2.0156, lr=1.03e-05]
Training:  97%|█████████▋| 10715/11082 [34:27:28<1:10:09, 11.47s/step, acc=0.5072, loss=2.0156, lr=1.03e-05]
Training:  97%|█████████▋| 10716/11082 [34:27:40<1:10:22, 11.54s/step, acc=0.5072, loss=2.0156, lr=1.03e-05]
Training:  97%|█████████▋| 10717/11082 [34:27:51<1:09:56, 11.50s/step, acc=0.5072, loss=2.0156, lr=1.03e-05]
Training:  97%|█████████▋| 10718/11082 [34:28:02<1:09:25, 11.44s/step, acc=0.5072, loss=2.0156, lr=1.03e-05]
Training:  97%|█████████▋| 10719/11082 [34:28:14<1:10:35, 11.67s/step, acc=0.5072, loss=2.0156, lr=1.03e-05]
Training:  97%|█████████▋| 10720/11082 [34:28:26<1:09:36, 11.54s/step, acc=0.5072, loss=2.0156, lr=1.03e-05]
Training:  97%|█████████▋| 10721/11082 [34:28:37<1:09:00, 11.47s/step, acc=0.5072, loss=2.0156, lr=1.03e-05]
Training:  97%|█████████▋| 10722/11082 [34:28:48<1:08:38, 11.44s/step, acc=0.5072, loss=2.0156, lr=1.03e-05]
Training:  97%|█████████▋| 10723/11082 [34:29:00<1:09:00, 11.53s/step, acc=0.5072, loss=2.0156, lr=1.03e-05]
Training:  97%|█████████▋| 10724/11082 [34:29:11<1:08:26, 11.47s/step, acc=0.5072, loss=2.0156, lr=1.03e-05]
Training:  97%|█████████▋| 10725/11082 [34:29:23<1:08:39, 11.54s/step, acc=0.5072, loss=2.0156, lr=1.03e-05]
Training:  97%|█████████▋| 10726/11082 [34:29:35<1:08:15, 11.50s/step, acc=0.5072, loss=2.0156, lr=1.03e-05]
Training:  97%|█████████▋| 10727/11082 [34:29:46<1:08:27, 11.57s/step, acc=0.5072, loss=2.0156, lr=1.03e-05]
Training:  97%|█████████▋| 10728/11082 [34:29:58<1:08:05, 11.54s/step, acc=0.5072, loss=2.0156, lr=1.03e-05]
Training:  97%|█████████▋| 10729/11082 [34:30:09<1:08:06, 11.58s/step, acc=0.5072, loss=2.0156, lr=1.03e-05]
Training:  97%|█████████▋| 10730/11082 [34:30:21<1:07:11, 11.45s/step, acc=0.5072, loss=2.0156, lr=1.03e-05]
Training:  97%|█████████▋| 10731/11082 [34:30:32<1:07:13, 11.49s/step, acc=0.5072, loss=2.0156, lr=1.03e-05]
Training:  97%|█████████▋| 10732/11082 [34:30:44<1:06:48, 11.45s/step, acc=0.5072, loss=2.0156, lr=1.03e-05]
Training:  97%|█████████▋| 10733/11082 [34:30:56<1:07:34, 11.62s/step, acc=0.5072, loss=2.0156, lr=1.03e-05]
Training:  97%|█████████▋| 10734/11082 [34:31:07<1:07:26, 11.63s/step, acc=0.5072, loss=2.0156, lr=1.03e-05]
Training:  97%|█████████▋| 10735/11082 [34:31:19<1:06:43, 11.54s/step, acc=0.5072, loss=2.0156, lr=1.03e-05]
Training:  97%|█████████▋| 10736/11082 [34:31:30<1:06:46, 11.58s/step, acc=0.5072, loss=2.0156, lr=1.03e-05]
Training:  97%|█████████▋| 10737/11082 [34:31:42<1:06:16, 11.53s/step, acc=0.5072, loss=2.0156, lr=1.03e-05]
Training:  97%|█████████▋| 10738/11082 [34:31:53<1:06:22, 11.58s/step, acc=0.5072, loss=2.0156, lr=1.03e-05]
Training:  97%|█████████▋| 10739/11082 [34:32:05<1:05:55, 11.53s/step, acc=0.5072, loss=2.0156, lr=1.03e-05]
Training:  97%|█████████▋| 10740/11082 [34:32:16<1:05:19, 11.46s/step, acc=0.5072, loss=2.0156, lr=1.03e-05]
Training:  97%|█████████▋| 10741/11082 [34:32:28<1:06:26, 11.69s/step, acc=0.5072, loss=2.0156, lr=1.03e-05]
Training:  97%|█████████▋| 10742/11082 [34:32:39<1:05:26, 11.55s/step, acc=0.5072, loss=2.0156, lr=1.03e-05]
Training:  97%|█████████▋| 10743/11082 [34:32:51<1:04:44, 11.46s/step, acc=0.5072, loss=2.0156, lr=1.03e-05]
Training:  97%|█████████▋| 10744/11082 [34:33:02<1:04:33, 11.46s/step, acc=0.5072, loss=2.0156, lr=1.03e-05]
Training:  97%|█████████▋| 10745/11082 [34:33:14<1:04:27, 11.48s/step, acc=0.5072, loss=2.0156, lr=1.03e-05]
Training:  97%|█████████▋| 10746/11082 [34:33:26<1:04:56, 11.60s/step, acc=0.5072, loss=2.0156, lr=1.03e-05]
Training:  97%|█████████▋| 10747/11082 [34:33:37<1:04:04, 11.48s/step, acc=0.5072, loss=2.0156, lr=1.03e-05]
Training:  97%|█████████▋| 10748/11082 [34:33:49<1:05:09, 11.71s/step, acc=0.5072, loss=2.0156, lr=1.03e-05]
Training:  97%|█████████▋| 10749/11082 [34:34:00<1:03:46, 11.49s/step, acc=0.5072, loss=2.0156, lr=1.03e-05]
Training:  97%|█████████▋| 10750/11082 [34:34:12<1:04:41, 11.69s/step, acc=0.5072, loss=2.0156, lr=1.03e-05]
Training:  97%|█████████▋| 10750/11082 [34:34:19<1:04:41, 11.69s/step, acc=0.5096, loss=1.9997, lr=1.02e-05]2026-06-20 01:27:47 | INFO | __main__ | epoch=1 step=10750 train_loss=1.999733 train_acc=0.509584 lr=1.021e-05 grad_norm=0.4941 acc_d0=0.7530 acc_d1=0.6075 acc_d2=0.4885 acc_d3=0.3922 acc_d4=0.3185 acc_d5=0.2584 acc_d6=0.2090 acc_d7=0.1742

Training:  97%|█████████▋| 10751/11082 [34:34:24<1:04:18, 11.66s/step, acc=0.5096, loss=1.9997, lr=1.02e-05]
Training:  97%|█████████▋| 10752/11082 [34:34:35<1:03:57, 11.63s/step, acc=0.5096, loss=1.9997, lr=1.02e-05]
Training:  97%|█████████▋| 10753/11082 [34:34:47<1:03:35, 11.60s/step, acc=0.5096, loss=1.9997, lr=1.02e-05]
Training:  97%|█████████▋| 10754/11082 [34:34:58<1:03:09, 11.55s/step, acc=0.5096, loss=1.9997, lr=1.02e-05]
Training:  97%|█████████▋| 10755/11082 [34:35:10<1:03:03, 11.57s/step, acc=0.5096, loss=1.9997, lr=1.02e-05]
Training:  97%|█████████▋| 10756/11082 [34:35:22<1:03:52, 11.76s/step, acc=0.5096, loss=1.9997, lr=1.02e-05]
Training:  97%|█████████▋| 10757/11082 [34:35:34<1:03:29, 11.72s/step, acc=0.5096, loss=1.9997, lr=1.02e-05]
Training:  97%|█████████▋| 10758/11082 [34:35:45<1:02:58, 11.66s/step, acc=0.5096, loss=1.9997, lr=1.02e-05]
Training:  97%|█████████▋| 10759/11082 [34:35:57<1:02:49, 11.67s/step, acc=0.5096, loss=1.9997, lr=1.02e-05]
Training:  97%|█████████▋| 10760/11082 [34:36:09<1:02:37, 11.67s/step, acc=0.5096, loss=1.9997, lr=1.02e-05]
Training:  97%|█████████▋| 10761/11082 [34:36:20<1:02:24, 11.67s/step, acc=0.5096, loss=1.9997, lr=1.02e-05]
Training:  97%|█████████▋| 10762/11082 [34:36:31<1:01:29, 11.53s/step, acc=0.5096, loss=1.9997, lr=1.02e-05]
Training:  97%|█████████▋| 10763/11082 [34:36:43<1:01:19, 11.53s/step, acc=0.5096, loss=1.9997, lr=1.02e-05]
Training:  97%|█████████▋| 10764/11082 [34:36:55<1:01:11, 11.55s/step, acc=0.5096, loss=1.9997, lr=1.02e-05]
Training:  97%|█████████▋| 10765/11082 [34:37:06<1:00:58, 11.54s/step, acc=0.5096, loss=1.9997, lr=1.02e-05]
Training:  97%|█████████▋| 10766/11082 [34:37:17<1:00:25, 11.47s/step, acc=0.5096, loss=1.9997, lr=1.02e-05]
Training:  97%|█████████▋| 10767/11082 [34:37:29<1:00:38, 11.55s/step, acc=0.5096, loss=1.9997, lr=1.02e-05]
Training:  97%|█████████▋| 10768/11082 [34:37:41<1:00:16, 11.52s/step, acc=0.5096, loss=1.9997, lr=1.02e-05]
Training:  97%|█████████▋| 10769/11082 [34:37:52<1:00:13, 11.54s/step, acc=0.5096, loss=1.9997, lr=1.02e-05]
Training:  97%|█████████▋| 10770/11082 [34:38:04<1:00:29, 11.63s/step, acc=0.5096, loss=1.9997, lr=1.02e-05]
Training:  97%|█████████▋| 10771/11082 [34:38:15<59:53, 11.56s/step, acc=0.5096, loss=1.9997, lr=1.02e-05]  
Training:  97%|█████████▋| 10772/11082 [34:38:27<59:48, 11.58s/step, acc=0.5096, loss=1.9997, lr=1.02e-05]
Training:  97%|█████████▋| 10773/11082 [34:38:38<59:20, 11.52s/step, acc=0.5096, loss=1.9997, lr=1.02e-05]
Training:  97%|█████████▋| 10774/11082 [34:38:50<59:39, 11.62s/step, acc=0.5096, loss=1.9997, lr=1.02e-05]
Training:  97%|█████████▋| 10775/11082 [34:39:02<59:17, 11.59s/step, acc=0.5096, loss=1.9997, lr=1.02e-05]
Training:  97%|█████████▋| 10776/11082 [34:39:13<58:19, 11.44s/step, acc=0.5096, loss=1.9997, lr=1.02e-05]
Training:  97%|█████████▋| 10777/11082 [34:39:24<58:10, 11.44s/step, acc=0.5096, loss=1.9997, lr=1.02e-05]
Training:  97%|█████████▋| 10778/11082 [34:39:36<58:47, 11.60s/step, acc=0.5096, loss=1.9997, lr=1.02e-05]
Training:  97%|█████████▋| 10779/11082 [34:39:47<57:42, 11.43s/step, acc=0.5096, loss=1.9997, lr=1.02e-05]
Training:  97%|█████████▋| 10780/11082 [34:40:00<58:43, 11.67s/step, acc=0.5096, loss=1.9997, lr=1.02e-05]
Training:  97%|█████████▋| 10781/11082 [34:40:11<57:51, 11.53s/step, acc=0.5096, loss=1.9997, lr=1.02e-05]
Training:  97%|█████████▋| 10782/11082 [34:40:23<58:14, 11.65s/step, acc=0.5096, loss=1.9997, lr=1.02e-05]
Training:  97%|█████████▋| 10783/11082 [34:40:34<57:29, 11.54s/step, acc=0.5096, loss=1.9997, lr=1.02e-05]
Training:  97%|█████████▋| 10784/11082 [34:40:45<56:45, 11.43s/step, acc=0.5096, loss=1.9997, lr=1.02e-05]
Training:  97%|█████████▋| 10785/11082 [34:40:57<57:41, 11.66s/step, acc=0.5096, loss=1.9997, lr=1.02e-05]
Training:  97%|█████████▋| 10786/11082 [34:41:09<56:53, 11.53s/step, acc=0.5096, loss=1.9997, lr=1.02e-05]
Training:  97%|█████████▋| 10787/11082 [34:41:20<57:05, 11.61s/step, acc=0.5096, loss=1.9997, lr=1.02e-05]
Training:  97%|█████████▋| 10788/11082 [34:41:32<56:28, 11.53s/step, acc=0.5096, loss=1.9997, lr=1.02e-05]
Training:  97%|█████████▋| 10789/11082 [34:41:43<55:49, 11.43s/step, acc=0.5096, loss=1.9997, lr=1.02e-05]
Training:  97%|█████████▋| 10790/11082 [34:41:55<56:26, 11.60s/step, acc=0.5096, loss=1.9997, lr=1.02e-05]
Training:  97%|█████████▋| 10791/11082 [34:42:06<56:05, 11.57s/step, acc=0.5096, loss=1.9997, lr=1.02e-05]
Training:  97%|█████████▋| 10792/11082 [34:42:18<55:28, 11.48s/step, acc=0.5096, loss=1.9997, lr=1.02e-05]
Training:  97%|█████████▋| 10793/11082 [34:42:29<55:33, 11.54s/step, acc=0.5096, loss=1.9997, lr=1.02e-05]
Training:  97%|█████████▋| 10794/11082 [34:42:41<55:22, 11.54s/step, acc=0.5096, loss=1.9997, lr=1.02e-05]
Training:  97%|█████████▋| 10795/11082 [34:42:53<55:26, 11.59s/step, acc=0.5096, loss=1.9997, lr=1.02e-05]
Training:  97%|█████████▋| 10796/11082 [34:43:04<55:16, 11.60s/step, acc=0.5096, loss=1.9997, lr=1.02e-05]
Training:  97%|█████████▋| 10797/11082 [34:43:16<54:49, 11.54s/step, acc=0.5096, loss=1.9997, lr=1.02e-05]
Training:  97%|█████████▋| 10798/11082 [34:43:27<55:05, 11.64s/step, acc=0.5096, loss=1.9997, lr=1.02e-05]
Training:  97%|█████████▋| 10799/11082 [34:43:39<54:24, 11.53s/step, acc=0.5096, loss=1.9997, lr=1.02e-05]
Training:  97%|█████████▋| 10800/11082 [34:43:50<53:35, 11.40s/step, acc=0.5096, loss=1.9997, lr=1.02e-05]
Training:  97%|█████████▋| 10800/11082 [34:43:57<53:35, 11.40s/step, acc=0.5059, loss=2.0261, lr=1.01e-05]2026-06-20 01:37:25 | INFO | __main__ | epoch=1 step=10800 train_loss=2.026077 train_acc=0.505931 lr=1.015e-05 grad_norm=0.4863 acc_d0=0.7669 acc_d1=0.6235 acc_d2=0.5052 acc_d3=0.4096 acc_d4=0.3313 acc_d5=0.2686 acc_d6=0.2188 acc_d7=0.1778

Training:  97%|█████████▋| 10801/11082 [34:44:02<54:23, 11.61s/step, acc=0.5059, loss=2.0261, lr=1.01e-05]
Training:  97%|█████████▋| 10802/11082 [34:44:13<53:51, 11.54s/step, acc=0.5059, loss=2.0261, lr=1.01e-05]
Training:  97%|█████████▋| 10803/11082 [34:44:25<53:35, 11.53s/step, acc=0.5059, loss=2.0261, lr=1.01e-05]
Training:  97%|█████████▋| 10804/11082 [34:44:37<54:01, 11.66s/step, acc=0.5059, loss=2.0261, lr=1.01e-05]
Training:  98%|█████████▊| 10805/11082 [34:44:48<53:07, 11.51s/step, acc=0.5059, loss=2.0261, lr=1.01e-05]
Training:  98%|█████████▊| 10806/11082 [34:45:00<53:00, 11.52s/step, acc=0.5059, loss=2.0261, lr=1.01e-05]
Training:  98%|█████████▊| 10807/11082 [34:45:11<53:03, 11.58s/step, acc=0.5059, loss=2.0261, lr=1.01e-05]
Training:  98%|█████████▊| 10808/11082 [34:45:22<52:17, 11.45s/step, acc=0.5059, loss=2.0261, lr=1.01e-05]
Training:  98%|█████████▊| 10809/11082 [34:45:34<52:55, 11.63s/step, acc=0.5059, loss=2.0261, lr=1.01e-05]
Training:  98%|█████████▊| 10810/11082 [34:45:46<51:59, 11.47s/step, acc=0.5059, loss=2.0261, lr=1.01e-05]
Training:  98%|█████████▊| 10811/11082 [34:45:57<52:05, 11.53s/step, acc=0.5059, loss=2.0261, lr=1.01e-05]
Training:  98%|█████████▊| 10812/11082 [34:46:09<52:36, 11.69s/step, acc=0.5059, loss=2.0261, lr=1.01e-05]
Training:  98%|█████████▊| 10813/11082 [34:46:20<51:14, 11.43s/step, acc=0.5059, loss=2.0261, lr=1.01e-05]
Training:  98%|█████████▊| 10814/11082 [34:46:32<51:17, 11.48s/step, acc=0.5059, loss=2.0261, lr=1.01e-05]
Training:  98%|█████████▊| 10815/11082 [34:46:43<51:14, 11.51s/step, acc=0.5059, loss=2.0261, lr=1.01e-05]
Training:  98%|█████████▊| 10816/11082 [34:46:55<51:14, 11.56s/step, acc=0.5059, loss=2.0261, lr=1.01e-05]
Training:  98%|█████████▊| 10817/11082 [34:47:06<50:38, 11.47s/step, acc=0.5059, loss=2.0261, lr=1.01e-05]
Training:  98%|█████████▊| 10818/11082 [34:47:18<50:31, 11.48s/step, acc=0.5059, loss=2.0261, lr=1.01e-05]
Training:  98%|█████████▊| 10819/11082 [34:47:30<51:10, 11.68s/step, acc=0.5059, loss=2.0261, lr=1.01e-05]
Training:  98%|█████████▊| 10820/11082 [34:47:41<50:21, 11.53s/step, acc=0.5059, loss=2.0261, lr=1.01e-05]
Training:  98%|█████████▊| 10821/11082 [34:47:53<50:38, 11.64s/step, acc=0.5059, loss=2.0261, lr=1.01e-05]
Training:  98%|█████████▊| 10822/11082 [34:48:05<50:22, 11.62s/step, acc=0.5059, loss=2.0261, lr=1.01e-05]
Training:  98%|█████████▊| 10823/11082 [34:48:16<50:11, 11.63s/step, acc=0.5059, loss=2.0261, lr=1.01e-05]
Training:  98%|█████████▊| 10824/11082 [34:48:27<49:14, 11.45s/step, acc=0.5059, loss=2.0261, lr=1.01e-05]
Training:  98%|█████████▊| 10825/11082 [34:48:39<50:07, 11.70s/step, acc=0.5059, loss=2.0261, lr=1.01e-05]
Training:  98%|█████████▊| 10826/11082 [34:48:51<49:21, 11.57s/step, acc=0.5059, loss=2.0261, lr=1.01e-05]
Training:  98%|█████████▊| 10827/11082 [34:49:02<49:19, 11.61s/step, acc=0.5059, loss=2.0261, lr=1.01e-05]
Training:  98%|█████████▊| 10828/11082 [34:49:15<49:46, 11.76s/step, acc=0.5059, loss=2.0261, lr=1.01e-05]
Training:  98%|█████████▊| 10829/11082 [34:49:26<49:38, 11.77s/step, acc=0.5059, loss=2.0261, lr=1.01e-05]
Training:  98%|█████████▊| 10830/11082 [34:49:38<49:20, 11.75s/step, acc=0.5059, loss=2.0261, lr=1.01e-05]
Training:  98%|█████████▊| 10831/11082 [34:49:50<49:01, 11.72s/step, acc=0.5059, loss=2.0261, lr=1.01e-05]
Training:  98%|█████████▊| 10832/11082 [34:50:02<49:18, 11.83s/step, acc=0.5059, loss=2.0261, lr=1.01e-05]
Training:  98%|█████████▊| 10833/11082 [34:50:13<48:42, 11.74s/step, acc=0.5059, loss=2.0261, lr=1.01e-05]
Training:  98%|█████████▊| 10834/11082 [34:50:24<47:50, 11.57s/step, acc=0.5059, loss=2.0261, lr=1.01e-05]
Training:  98%|█████████▊| 10835/11082 [34:50:36<48:00, 11.66s/step, acc=0.5059, loss=2.0261, lr=1.01e-05]
Training:  98%|█████████▊| 10836/11082 [34:50:48<47:33, 11.60s/step, acc=0.5059, loss=2.0261, lr=1.01e-05]
Training:  98%|█████████▊| 10837/11082 [34:50:59<47:21, 11.60s/step, acc=0.5059, loss=2.0261, lr=1.01e-05]
Training:  98%|█████████▊| 10838/11082 [34:51:11<47:00, 11.56s/step, acc=0.5059, loss=2.0261, lr=1.01e-05]
Training:  98%|█████████▊| 10839/11082 [34:51:23<46:57, 11.59s/step, acc=0.5059, loss=2.0261, lr=1.01e-05]
Training:  98%|█████████▊| 10840/11082 [34:51:34<46:18, 11.48s/step, acc=0.5059, loss=2.0261, lr=1.01e-05]
Training:  98%|█████████▊| 10841/11082 [34:51:45<46:21, 11.54s/step, acc=0.5059, loss=2.0261, lr=1.01e-05]
Training:  98%|█████████▊| 10842/11082 [34:51:57<46:17, 11.57s/step, acc=0.5059, loss=2.0261, lr=1.01e-05]
Training:  98%|█████████▊| 10843/11082 [34:52:08<45:46, 11.49s/step, acc=0.5059, loss=2.0261, lr=1.01e-05]
Training:  98%|█████████▊| 10844/11082 [34:52:20<45:54, 11.57s/step, acc=0.5059, loss=2.0261, lr=1.01e-05]
Training:  98%|█████████▊| 10845/11082 [34:52:32<45:48, 11.60s/step, acc=0.5059, loss=2.0261, lr=1.01e-05]
Training:  98%|█████████▊| 10846/11082 [34:52:43<45:21, 11.53s/step, acc=0.5059, loss=2.0261, lr=1.01e-05]
Training:  98%|█████████▊| 10847/11082 [34:52:55<45:22, 11.59s/step, acc=0.5059, loss=2.0261, lr=1.01e-05]
Training:  98%|█████████▊| 10848/11082 [34:53:06<45:07, 11.57s/step, acc=0.5059, loss=2.0261, lr=1.01e-05]
Training:  98%|█████████▊| 10849/11082 [34:53:18<44:49, 11.54s/step, acc=0.5059, loss=2.0261, lr=1.01e-05]
Training:  98%|█████████▊| 10850/11082 [34:53:29<44:36, 11.53s/step, acc=0.5059, loss=2.0261, lr=1.01e-05]
Training:  98%|█████████▊| 10850/11082 [34:53:37<44:36, 11.53s/step, acc=0.5091, loss=2.0071, lr=1.01e-05]2026-06-20 01:47:05 | INFO | __main__ | epoch=1 step=10850 train_loss=2.007061 train_acc=0.509057 lr=1.010e-05 grad_norm=0.4785 acc_d0=0.7585 acc_d1=0.6188 acc_d2=0.5011 acc_d3=0.4072 acc_d4=0.3301 acc_d5=0.2697 acc_d6=0.2201 acc_d7=0.1795

Training:  98%|█████████▊| 10851/11082 [34:53:41<44:52, 11.66s/step, acc=0.5091, loss=2.0071, lr=1.01e-05]
Training:  98%|█████████▊| 10852/11082 [34:53:53<44:25, 11.59s/step, acc=0.5091, loss=2.0071, lr=1.01e-05]
Training:  98%|█████████▊| 10853/11082 [34:54:04<43:57, 11.52s/step, acc=0.5091, loss=2.0071, lr=1.01e-05]
Training:  98%|█████████▊| 10854/11082 [34:54:16<44:03, 11.60s/step, acc=0.5091, loss=2.0071, lr=1.01e-05]
Training:  98%|█████████▊| 10855/11082 [34:54:28<44:16, 11.70s/step, acc=0.5091, loss=2.0071, lr=1.01e-05]
Training:  98%|█████████▊| 10856/11082 [34:54:39<43:26, 11.53s/step, acc=0.5091, loss=2.0071, lr=1.01e-05]
Training:  98%|█████████▊| 10857/11082 [34:54:51<43:20, 11.56s/step, acc=0.5091, loss=2.0071, lr=1.01e-05]
Training:  98%|█████████▊| 10858/11082 [34:55:02<43:07, 11.55s/step, acc=0.5091, loss=2.0071, lr=1.01e-05]
Training:  98%|█████████▊| 10859/11082 [34:55:14<43:28, 11.70s/step, acc=0.5091, loss=2.0071, lr=1.01e-05]
Training:  98%|█████████▊| 10860/11082 [34:55:26<43:10, 11.67s/step, acc=0.5091, loss=2.0071, lr=1.01e-05]
Training:  98%|█████████▊| 10861/11082 [34:55:37<42:34, 11.56s/step, acc=0.5091, loss=2.0071, lr=1.01e-05]
Training:  98%|█████████▊| 10862/11082 [34:55:48<41:57, 11.45s/step, acc=0.5091, loss=2.0071, lr=1.01e-05]
Training:  98%|█████████▊| 10863/11082 [34:56:00<42:08, 11.54s/step, acc=0.5091, loss=2.0071, lr=1.01e-05]
Training:  98%|█████████▊| 10864/11082 [34:56:11<41:42, 11.48s/step, acc=0.5091, loss=2.0071, lr=1.01e-05]
Training:  98%|█████████▊| 10865/11082 [34:56:23<41:41, 11.53s/step, acc=0.5091, loss=2.0071, lr=1.01e-05]
Training:  98%|█████████▊| 10866/11082 [34:56:35<41:41, 11.58s/step, acc=0.5091, loss=2.0071, lr=1.01e-05]
Training:  98%|█████████▊| 10867/11082 [34:56:46<41:27, 11.57s/step, acc=0.5091, loss=2.0071, lr=1.01e-05]
Training:  98%|█████████▊| 10868/11082 [34:56:58<41:13, 11.56s/step, acc=0.5091, loss=2.0071, lr=1.01e-05]
Training:  98%|█████████▊| 10869/11082 [34:57:09<40:57, 11.54s/step, acc=0.5091, loss=2.0071, lr=1.01e-05]
Training:  98%|█████████▊| 10870/11082 [34:57:21<40:55, 11.58s/step, acc=0.5091, loss=2.0071, lr=1.01e-05]
Training:  98%|█████████▊| 10871/11082 [34:57:33<40:51, 11.62s/step, acc=0.5091, loss=2.0071, lr=1.01e-05]
Training:  98%|█████████▊| 10872/11082 [34:57:44<39:51, 11.39s/step, acc=0.5091, loss=2.0071, lr=1.01e-05]
Training:  98%|█████████▊| 10873/11082 [34:57:56<40:23, 11.60s/step, acc=0.5091, loss=2.0071, lr=1.01e-05]
Training:  98%|█████████▊| 10874/11082 [34:58:07<39:38, 11.44s/step, acc=0.5091, loss=2.0071, lr=1.01e-05]
Training:  98%|█████████▊| 10875/11082 [34:58:18<39:49, 11.54s/step, acc=0.5091, loss=2.0071, lr=1.01e-05]
Training:  98%|█████████▊| 10876/11082 [34:58:30<39:38, 11.54s/step, acc=0.5091, loss=2.0071, lr=1.01e-05]
Training:  98%|█████████▊| 10877/11082 [34:58:41<39:15, 11.49s/step, acc=0.5091, loss=2.0071, lr=1.01e-05]
Training:  98%|█████████▊| 10878/11082 [34:58:52<38:33, 11.34s/step, acc=0.5091, loss=2.0071, lr=1.01e-05]
Training:  98%|█████████▊| 10879/11082 [34:59:05<39:11, 11.58s/step, acc=0.5091, loss=2.0071, lr=1.01e-05]
Training:  98%|█████████▊| 10880/11082 [34:59:16<39:09, 11.63s/step, acc=0.5091, loss=2.0071, lr=1.01e-05]
Training:  98%|█████████▊| 10881/11082 [34:59:28<38:49, 11.59s/step, acc=0.5091, loss=2.0071, lr=1.01e-05]
Training:  98%|█████████▊| 10882/11082 [34:59:39<38:10, 11.45s/step, acc=0.5091, loss=2.0071, lr=1.01e-05]
Training:  98%|█████████▊| 10883/11082 [34:59:51<38:26, 11.59s/step, acc=0.5091, loss=2.0071, lr=1.01e-05]
Training:  98%|█████████▊| 10884/11082 [35:00:02<38:08, 11.56s/step, acc=0.5091, loss=2.0071, lr=1.01e-05]
Training:  98%|█████████▊| 10885/11082 [35:00:14<37:45, 11.50s/step, acc=0.5091, loss=2.0071, lr=1.01e-05]
Training:  98%|█████████▊| 10886/11082 [35:00:25<37:21, 11.44s/step, acc=0.5091, loss=2.0071, lr=1.01e-05]
Training:  98%|█████████▊| 10887/11082 [35:00:37<37:44, 11.61s/step, acc=0.5091, loss=2.0071, lr=1.01e-05]
Training:  98%|█████████▊| 10888/11082 [35:00:48<36:37, 11.33s/step, acc=0.5091, loss=2.0071, lr=1.01e-05]
Training:  98%|█████████▊| 10889/11082 [35:01:00<37:31, 11.67s/step, acc=0.5091, loss=2.0071, lr=1.01e-05]
Training:  98%|█████████▊| 10890/11082 [35:01:11<36:34, 11.43s/step, acc=0.5091, loss=2.0071, lr=1.01e-05]
Training:  98%|█████████▊| 10891/11082 [35:01:23<37:21, 11.73s/step, acc=0.5091, loss=2.0071, lr=1.01e-05]
Training:  98%|█████████▊| 10892/11082 [35:01:34<36:23, 11.49s/step, acc=0.5091, loss=2.0071, lr=1.01e-05]
Training:  98%|█████████▊| 10893/11082 [35:01:46<36:32, 11.60s/step, acc=0.5091, loss=2.0071, lr=1.01e-05]
Training:  98%|█████████▊| 10894/11082 [35:01:58<36:31, 11.65s/step, acc=0.5091, loss=2.0071, lr=1.01e-05]
Training:  98%|█████████▊| 10895/11082 [35:02:09<35:55, 11.53s/step, acc=0.5091, loss=2.0071, lr=1.01e-05]
Training:  98%|█████████▊| 10896/11082 [35:02:21<36:01, 11.62s/step, acc=0.5091, loss=2.0071, lr=1.01e-05]
Training:  98%|█████████▊| 10897/11082 [35:02:33<35:57, 11.66s/step, acc=0.5091, loss=2.0071, lr=1.01e-05]
Training:  98%|█████████▊| 10898/11082 [35:02:44<35:22, 11.54s/step, acc=0.5091, loss=2.0071, lr=1.01e-05]
Training:  98%|█████████▊| 10899/11082 [35:02:56<35:23, 11.61s/step, acc=0.5091, loss=2.0071, lr=1.01e-05]
Training:  98%|█████████▊| 10900/11082 [35:03:07<34:59, 11.54s/step, acc=0.5091, loss=2.0071, lr=1.01e-05]
Training:  98%|█████████▊| 10900/11082 [35:03:15<34:59, 11.54s/step, acc=0.5076, loss=2.0119, lr=1.01e-05]2026-06-20 01:56:43 | INFO | __main__ | epoch=1 step=10900 train_loss=2.011875 train_acc=0.507624 lr=1.006e-05 grad_norm=0.5195 acc_d0=0.7762 acc_d1=0.6360 acc_d2=0.5167 acc_d3=0.4211 acc_d4=0.3410 acc_d5=0.2777 acc_d6=0.2256 acc_d7=0.1839

Training:  98%|█████████▊| 10901/11082 [35:03:19<34:52, 11.56s/step, acc=0.5076, loss=2.0119, lr=1.01e-05]
Training:  98%|█████████▊| 10902/11082 [35:03:31<34:55, 11.64s/step, acc=0.5076, loss=2.0119, lr=1.01e-05]
Training:  98%|█████████▊| 10903/11082 [35:03:42<34:40, 11.62s/step, acc=0.5076, loss=2.0119, lr=1.01e-05]
Training:  98%|█████████▊| 10904/11082 [35:03:54<34:20, 11.58s/step, acc=0.5076, loss=2.0119, lr=1.01e-05]
Training:  98%|█████████▊| 10905/11082 [35:04:05<34:01, 11.53s/step, acc=0.5076, loss=2.0119, lr=1.01e-05]
Training:  98%|█████████▊| 10906/11082 [35:04:17<34:05, 11.62s/step, acc=0.5076, loss=2.0119, lr=1.01e-05]
Training:  98%|█████████▊| 10907/11082 [35:04:29<34:00, 11.66s/step, acc=0.5076, loss=2.0119, lr=1.01e-05]
Training:  98%|█████████▊| 10908/11082 [35:04:40<33:38, 11.60s/step, acc=0.5076, loss=2.0119, lr=1.01e-05]
Training:  98%|█████████▊| 10909/11082 [35:04:51<32:41, 11.34s/step, acc=0.5076, loss=2.0119, lr=1.01e-05]
Training:  98%|█████████▊| 10910/11082 [35:05:03<32:53, 11.47s/step, acc=0.5076, loss=2.0119, lr=1.01e-05]
Training:  98%|█████████▊| 10911/11082 [35:05:15<33:07, 11.62s/step, acc=0.5076, loss=2.0119, lr=1.01e-05]
Training:  98%|█████████▊| 10912/11082 [35:05:26<32:47, 11.57s/step, acc=0.5076, loss=2.0119, lr=1.01e-05]
Training:  98%|█████████▊| 10913/11082 [35:05:38<33:11, 11.79s/step, acc=0.5076, loss=2.0119, lr=1.01e-05]
Training:  98%|█████████▊| 10914/11082 [35:05:50<32:31, 11.62s/step, acc=0.5076, loss=2.0119, lr=1.01e-05]
Training:  98%|█████████▊| 10915/11082 [35:06:01<32:19, 11.61s/step, acc=0.5076, loss=2.0119, lr=1.01e-05]
Training:  99%|█████████▊| 10916/11082 [35:06:13<32:08, 11.62s/step, acc=0.5076, loss=2.0119, lr=1.01e-05]
Training:  99%|█████████▊| 10917/11082 [35:06:25<32:06, 11.67s/step, acc=0.5076, loss=2.0119, lr=1.01e-05]
Training:  99%|█████████▊| 10918/11082 [35:06:36<31:34, 11.55s/step, acc=0.5076, loss=2.0119, lr=1.01e-05]
Training:  99%|█████████▊| 10919/11082 [35:06:47<31:06, 11.45s/step, acc=0.5076, loss=2.0119, lr=1.01e-05]
Training:  99%|█████████▊| 10920/11082 [35:06:59<31:13, 11.57s/step, acc=0.5076, loss=2.0119, lr=1.01e-05]
Training:  99%|█████████▊| 10921/11082 [35:07:11<31:09, 11.61s/step, acc=0.5076, loss=2.0119, lr=1.01e-05]
Training:  99%|█████████▊| 10922/11082 [35:07:22<30:55, 11.60s/step, acc=0.5076, loss=2.0119, lr=1.01e-05]
Training:  99%|█████████▊| 10923/11082 [35:07:34<30:42, 11.59s/step, acc=0.5076, loss=2.0119, lr=1.01e-05]
Training:  99%|█████████▊| 10924/11082 [35:07:46<30:37, 11.63s/step, acc=0.5076, loss=2.0119, lr=1.01e-05]
Training:  99%|█████████▊| 10925/11082 [35:07:57<30:11, 11.54s/step, acc=0.5076, loss=2.0119, lr=1.01e-05]
Training:  99%|█████████▊| 10926/11082 [35:08:08<29:57, 11.52s/step, acc=0.5076, loss=2.0119, lr=1.01e-05]
Training:  99%|█████████▊| 10927/11082 [35:08:20<29:46, 11.53s/step, acc=0.5076, loss=2.0119, lr=1.01e-05]
Training:  99%|█████████▊| 10928/11082 [35:08:32<29:44, 11.59s/step, acc=0.5076, loss=2.0119, lr=1.01e-05]
Training:  99%|█████████▊| 10929/11082 [35:08:43<29:27, 11.55s/step, acc=0.5076, loss=2.0119, lr=1.01e-05]
Training:  99%|█████████▊| 10930/11082 [35:08:55<29:19, 11.58s/step, acc=0.5076, loss=2.0119, lr=1.01e-05]
Training:  99%|█████████▊| 10931/11082 [35:09:06<28:59, 11.52s/step, acc=0.5076, loss=2.0119, lr=1.01e-05]
Training:  99%|█████████▊| 10932/11082 [35:09:18<28:46, 11.51s/step, acc=0.5076, loss=2.0119, lr=1.01e-05]
Training:  99%|█████████▊| 10933/11082 [35:09:29<28:40, 11.55s/step, acc=0.5076, loss=2.0119, lr=1.01e-05]
Training:  99%|█████████▊| 10934/11082 [35:09:41<28:24, 11.52s/step, acc=0.5076, loss=2.0119, lr=1.01e-05]
Training:  99%|█████████▊| 10935/11082 [35:09:52<28:11, 11.51s/step, acc=0.5076, loss=2.0119, lr=1.01e-05]
Training:  99%|█████████▊| 10936/11082 [35:10:04<28:19, 11.64s/step, acc=0.5076, loss=2.0119, lr=1.01e-05]
Training:  99%|█████████▊| 10937/11082 [35:10:16<28:07, 11.64s/step, acc=0.5076, loss=2.0119, lr=1.01e-05]
Training:  99%|█████████▊| 10938/11082 [35:10:27<27:43, 11.55s/step, acc=0.5076, loss=2.0119, lr=1.01e-05]
Training:  99%|█████████▊| 10939/11082 [35:10:38<27:17, 11.45s/step, acc=0.5076, loss=2.0119, lr=1.01e-05]
Training:  99%|█████████▊| 10940/11082 [35:10:50<27:27, 11.60s/step, acc=0.5076, loss=2.0119, lr=1.01e-05]
Training:  99%|█████████▊| 10941/11082 [35:11:01<26:55, 11.46s/step, acc=0.5076, loss=2.0119, lr=1.01e-05]
Training:  99%|█████████▊| 10942/11082 [35:11:13<26:52, 11.52s/step, acc=0.5076, loss=2.0119, lr=1.01e-05]
Training:  99%|█████████▊| 10943/11082 [35:11:25<26:56, 11.63s/step, acc=0.5076, loss=2.0119, lr=1.01e-05]
Training:  99%|█████████▉| 10944/11082 [35:11:36<26:41, 11.60s/step, acc=0.5076, loss=2.0119, lr=1.01e-05]
Training:  99%|█████████▉| 10945/11082 [35:11:48<26:37, 11.66s/step, acc=0.5076, loss=2.0119, lr=1.01e-05]
Training:  99%|█████████▉| 10946/11082 [35:12:00<26:10, 11.55s/step, acc=0.5076, loss=2.0119, lr=1.01e-05]
Training:  99%|█████████▉| 10947/11082 [35:12:11<25:52, 11.50s/step, acc=0.5076, loss=2.0119, lr=1.01e-05]
Training:  99%|█████████▉| 10948/11082 [35:12:23<26:02, 11.66s/step, acc=0.5076, loss=2.0119, lr=1.01e-05]
Training:  99%|█████████▉| 10949/11082 [35:12:34<25:35, 11.55s/step, acc=0.5076, loss=2.0119, lr=1.01e-05]
Training:  99%|█████████▉| 10950/11082 [35:12:46<25:37, 11.65s/step, acc=0.5076, loss=2.0119, lr=1.01e-05]
Training:  99%|█████████▉| 10950/11082 [35:12:53<25:37, 11.65s/step, acc=0.5087, loss=2.0076, lr=1.00e-05]2026-06-20 02:06:21 | INFO | __main__ | epoch=1 step=10950 train_loss=2.007593 train_acc=0.508671 lr=1.003e-05 grad_norm=0.4902 acc_d0=0.7587 acc_d1=0.6143 acc_d2=0.4962 acc_d3=0.3993 acc_d4=0.3204 acc_d5=0.2582 acc_d6=0.2101 acc_d7=0.1737

Training:  99%|█████████▉| 10951/11082 [35:12:58<25:14, 11.56s/step, acc=0.5087, loss=2.0076, lr=1.00e-05]
Training:  99%|█████████▉| 10952/11082 [35:13:09<25:16, 11.67s/step, acc=0.5087, loss=2.0076, lr=1.00e-05]
Training:  99%|█████████▉| 10953/11082 [35:13:21<25:02, 11.65s/step, acc=0.5087, loss=2.0076, lr=1.00e-05]
Training:  99%|█████████▉| 10954/11082 [35:13:32<24:32, 11.50s/step, acc=0.5087, loss=2.0076, lr=1.00e-05]
Training:  99%|█████████▉| 10955/11082 [35:13:44<24:47, 11.71s/step, acc=0.5087, loss=2.0076, lr=1.00e-05]
Training:  99%|█████████▉| 10956/11082 [35:13:56<24:31, 11.68s/step, acc=0.5087, loss=2.0076, lr=1.00e-05]
Training:  99%|█████████▉| 10957/11082 [35:14:07<24:08, 11.59s/step, acc=0.5087, loss=2.0076, lr=1.00e-05]
Training:  99%|█████████▉| 10958/11082 [35:14:19<23:46, 11.50s/step, acc=0.5087, loss=2.0076, lr=1.00e-05]
Training:  99%|█████████▉| 10959/11082 [35:14:30<23:35, 11.51s/step, acc=0.5087, loss=2.0076, lr=1.00e-05]
Training:  99%|█████████▉| 10960/11082 [35:14:42<23:38, 11.63s/step, acc=0.5087, loss=2.0076, lr=1.00e-05]
Training:  99%|█████████▉| 10961/11082 [35:14:53<23:15, 11.53s/step, acc=0.5087, loss=2.0076, lr=1.00e-05]
Training:  99%|█████████▉| 10962/11082 [35:15:05<23:10, 11.59s/step, acc=0.5087, loss=2.0076, lr=1.00e-05]
Training:  99%|█████████▉| 10963/11082 [35:15:17<22:59, 11.59s/step, acc=0.5087, loss=2.0076, lr=1.00e-05]
Training:  99%|█████████▉| 10964/11082 [35:15:28<22:40, 11.53s/step, acc=0.5087, loss=2.0076, lr=1.00e-05]
Training:  99%|█████████▉| 10965/11082 [35:15:40<22:36, 11.59s/step, acc=0.5087, loss=2.0076, lr=1.00e-05]
Training:  99%|█████████▉| 10966/11082 [35:15:51<22:23, 11.58s/step, acc=0.5087, loss=2.0076, lr=1.00e-05]
Training:  99%|█████████▉| 10967/11082 [35:16:03<22:18, 11.64s/step, acc=0.5087, loss=2.0076, lr=1.00e-05]
Training:  99%|█████████▉| 10968/11082 [35:16:15<22:01, 11.59s/step, acc=0.5087, loss=2.0076, lr=1.00e-05]
Training:  99%|█████████▉| 10969/11082 [35:16:26<21:54, 11.64s/step, acc=0.5087, loss=2.0076, lr=1.00e-05]
Training:  99%|█████████▉| 10970/11082 [35:16:38<21:40, 11.61s/step, acc=0.5087, loss=2.0076, lr=1.00e-05]
Training:  99%|█████████▉| 10971/11082 [35:16:49<21:14, 11.48s/step, acc=0.5087, loss=2.0076, lr=1.00e-05]
Training:  99%|█████████▉| 10972/11082 [35:17:01<21:13, 11.58s/step, acc=0.5087, loss=2.0076, lr=1.00e-05]
Training:  99%|█████████▉| 10973/11082 [35:17:12<20:49, 11.46s/step, acc=0.5087, loss=2.0076, lr=1.00e-05]
Training:  99%|█████████▉| 10974/11082 [35:17:24<20:59, 11.66s/step, acc=0.5087, loss=2.0076, lr=1.00e-05]
Training:  99%|█████████▉| 10975/11082 [35:17:35<20:32, 11.52s/step, acc=0.5087, loss=2.0076, lr=1.00e-05]
Training:  99%|█████████▉| 10976/11082 [35:17:47<20:23, 11.54s/step, acc=0.5087, loss=2.0076, lr=1.00e-05]
Training:  99%|█████████▉| 10977/11082 [35:17:59<20:19, 11.61s/step, acc=0.5087, loss=2.0076, lr=1.00e-05]
Training:  99%|█████████▉| 10978/11082 [35:18:10<20:06, 11.60s/step, acc=0.5087, loss=2.0076, lr=1.00e-05]
Training:  99%|█████████▉| 10979/11082 [35:18:22<19:50, 11.56s/step, acc=0.5087, loss=2.0076, lr=1.00e-05]
Training:  99%|█████████▉| 10980/11082 [35:18:34<19:43, 11.61s/step, acc=0.5087, loss=2.0076, lr=1.00e-05]
Training:  99%|█████████▉| 10981/11082 [35:18:45<19:26, 11.55s/step, acc=0.5087, loss=2.0076, lr=1.00e-05]
Training:  99%|█████████▉| 10982/11082 [35:18:56<19:10, 11.51s/step, acc=0.5087, loss=2.0076, lr=1.00e-05]
Training:  99%|█████████▉| 10983/11082 [35:19:08<19:01, 11.53s/step, acc=0.5087, loss=2.0076, lr=1.00e-05]
Training:  99%|█████████▉| 10984/11082 [35:19:20<18:52, 11.56s/step, acc=0.5087, loss=2.0076, lr=1.00e-05]
Training:  99%|█████████▉| 10985/11082 [35:19:31<18:35, 11.51s/step, acc=0.5087, loss=2.0076, lr=1.00e-05]
Training:  99%|█████████▉| 10986/11082 [35:19:42<18:16, 11.42s/step, acc=0.5087, loss=2.0076, lr=1.00e-05]
Training:  99%|█████████▉| 10987/11082 [35:19:54<18:09, 11.47s/step, acc=0.5087, loss=2.0076, lr=1.00e-05]
Training:  99%|█████████▉| 10988/11082 [35:20:05<18:03, 11.52s/step, acc=0.5087, loss=2.0076, lr=1.00e-05]
Training:  99%|█████████▉| 10989/11082 [35:20:17<17:54, 11.56s/step, acc=0.5087, loss=2.0076, lr=1.00e-05]
Training:  99%|█████████▉| 10990/11082 [35:20:29<17:39, 11.52s/step, acc=0.5087, loss=2.0076, lr=1.00e-05]
Training:  99%|█████████▉| 10991/11082 [35:20:40<17:31, 11.55s/step, acc=0.5087, loss=2.0076, lr=1.00e-05]
Training:  99%|█████████▉| 10992/11082 [35:20:52<17:25, 11.62s/step, acc=0.5087, loss=2.0076, lr=1.00e-05]
Training:  99%|█████████▉| 10993/11082 [35:21:03<17:11, 11.59s/step, acc=0.5087, loss=2.0076, lr=1.00e-05]
Training:  99%|█████████▉| 10994/11082 [35:21:15<16:56, 11.55s/step, acc=0.5087, loss=2.0076, lr=1.00e-05]
Training:  99%|█████████▉| 10995/11082 [35:21:27<16:51, 11.63s/step, acc=0.5087, loss=2.0076, lr=1.00e-05]
Training:  99%|█████████▉| 10996/11082 [35:21:39<16:47, 11.71s/step, acc=0.5087, loss=2.0076, lr=1.00e-05]
Training:  99%|█████████▉| 10997/11082 [35:21:50<16:28, 11.63s/step, acc=0.5087, loss=2.0076, lr=1.00e-05]
Training:  99%|█████████▉| 10998/11082 [35:22:01<16:04, 11.48s/step, acc=0.5087, loss=2.0076, lr=1.00e-05]
Training:  99%|█████████▉| 10999/11082 [35:22:13<16:00, 11.57s/step, acc=0.5087, loss=2.0076, lr=1.00e-05]
Training:  99%|█████████▉| 11000/11082 [35:22:24<15:46, 11.54s/step, acc=0.5087, loss=2.0076, lr=1.00e-05]
Training:  99%|█████████▉| 11000/11082 [35:22:31<15:46, 11.54s/step, acc=0.5073, loss=2.0182, lr=1.00e-05]2026-06-20 02:15:59 | INFO | __main__ | epoch=1 step=11000 train_loss=2.018178 train_acc=0.507328 lr=1.001e-05 grad_norm=0.5039 acc_d0=0.7673 acc_d1=0.6261 acc_d2=0.5079 acc_d3=0.4113 acc_d4=0.3346 acc_d5=0.2710 acc_d6=0.2208 acc_d7=0.1813

Training:  99%|█████████▉| 11001/11082 [35:22:36<15:38, 11.59s/step, acc=0.5073, loss=2.0182, lr=1.00e-05]
Training:  99%|█████████▉| 11002/11082 [35:22:48<15:22, 11.53s/step, acc=0.5073, loss=2.0182, lr=1.00e-05]
Training:  99%|█████████▉| 11003/11082 [35:22:59<15:05, 11.46s/step, acc=0.5073, loss=2.0182, lr=1.00e-05]
Training:  99%|█████████▉| 11004/11082 [35:23:11<14:59, 11.53s/step, acc=0.5073, loss=2.0182, lr=1.00e-05]
Training:  99%|█████████▉| 11005/11082 [35:23:22<14:45, 11.51s/step, acc=0.5073, loss=2.0182, lr=1.00e-05]
Training:  99%|█████████▉| 11006/11082 [35:23:34<14:39, 11.58s/step, acc=0.5073, loss=2.0182, lr=1.00e-05]
Training:  99%|█████████▉| 11007/11082 [35:23:45<14:24, 11.53s/step, acc=0.5073, loss=2.0182, lr=1.00e-05]
Training:  99%|█████████▉| 11008/11082 [35:23:57<14:18, 11.61s/step, acc=0.5073, loss=2.0182, lr=1.00e-05]
Training:  99%|█████████▉| 11009/11082 [35:24:09<14:08, 11.63s/step, acc=0.5073, loss=2.0182, lr=1.00e-05]
Training:  99%|█████████▉| 11010/11082 [35:24:20<13:55, 11.60s/step, acc=0.5073, loss=2.0182, lr=1.00e-05]
Training:  99%|█████████▉| 11011/11082 [35:24:31<13:38, 11.53s/step, acc=0.5073, loss=2.0182, lr=1.00e-05]
Training:  99%|█████████▉| 11012/11082 [35:24:43<13:29, 11.56s/step, acc=0.5073, loss=2.0182, lr=1.00e-05]
Training:  99%|█████████▉| 11013/11082 [35:24:55<13:17, 11.56s/step, acc=0.5073, loss=2.0182, lr=1.00e-05]
Training:  99%|█████████▉| 11014/11082 [35:25:07<13:12, 11.66s/step, acc=0.5073, loss=2.0182, lr=1.00e-05]
Training:  99%|█████████▉| 11015/11082 [35:25:18<12:58, 11.62s/step, acc=0.5073, loss=2.0182, lr=1.00e-05]
Training:  99%|█████████▉| 11016/11082 [35:25:30<12:59, 11.81s/step, acc=0.5073, loss=2.0182, lr=1.00e-05]
Training:  99%|█████████▉| 11017/11082 [35:25:42<12:36, 11.64s/step, acc=0.5073, loss=2.0182, lr=1.00e-05]
Training:  99%|█████████▉| 11018/11082 [35:25:53<12:28, 11.70s/step, acc=0.5073, loss=2.0182, lr=1.00e-05]
Training:  99%|█████████▉| 11019/11082 [35:26:05<12:06, 11.54s/step, acc=0.5073, loss=2.0182, lr=1.00e-05]
Training:  99%|█████████▉| 11020/11082 [35:26:16<11:46, 11.39s/step, acc=0.5073, loss=2.0182, lr=1.00e-05]
Training:  99%|█████████▉| 11021/11082 [35:26:28<11:50, 11.64s/step, acc=0.5073, loss=2.0182, lr=1.00e-05]
Training:  99%|█████████▉| 11022/11082 [35:26:39<11:29, 11.49s/step, acc=0.5073, loss=2.0182, lr=1.00e-05]
Training:  99%|█████████▉| 11023/11082 [35:26:51<11:29, 11.69s/step, acc=0.5073, loss=2.0182, lr=1.00e-05]
Training:  99%|█████████▉| 11024/11082 [35:27:03<11:16, 11.66s/step, acc=0.5073, loss=2.0182, lr=1.00e-05]
Training:  99%|█████████▉| 11025/11082 [35:27:14<11:01, 11.60s/step, acc=0.5073, loss=2.0182, lr=1.00e-05]
Training:  99%|█████████▉| 11026/11082 [35:27:26<10:53, 11.66s/step, acc=0.5073, loss=2.0182, lr=1.00e-05]
Training: 100%|█████████▉| 11027/11082 [35:27:37<10:38, 11.61s/step, acc=0.5073, loss=2.0182, lr=1.00e-05]
Training: 100%|█████████▉| 11028/11082 [35:27:49<10:20, 11.48s/step, acc=0.5073, loss=2.0182, lr=1.00e-05]
Training: 100%|█████████▉| 11029/11082 [35:28:01<10:14, 11.59s/step, acc=0.5073, loss=2.0182, lr=1.00e-05]
Training: 100%|█████████▉| 11030/11082 [35:28:12<10:03, 11.61s/step, acc=0.5073, loss=2.0182, lr=1.00e-05]
Training: 100%|█████████▉| 11031/11082 [35:28:24<09:49, 11.56s/step, acc=0.5073, loss=2.0182, lr=1.00e-05]
Training: 100%|█████████▉| 11032/11082 [35:28:35<09:37, 11.54s/step, acc=0.5073, loss=2.0182, lr=1.00e-05]
Training: 100%|█████████▉| 11033/11082 [35:28:47<09:26, 11.57s/step, acc=0.5073, loss=2.0182, lr=1.00e-05]
Training: 100%|█████████▉| 11034/11082 [35:28:58<09:13, 11.54s/step, acc=0.5073, loss=2.0182, lr=1.00e-05]
Training: 100%|█████████▉| 11035/11082 [35:29:10<09:02, 11.54s/step, acc=0.5073, loss=2.0182, lr=1.00e-05]
Training: 100%|█████████▉| 11036/11082 [35:29:21<08:47, 11.46s/step, acc=0.5073, loss=2.0182, lr=1.00e-05]
Training: 100%|█████████▉| 11037/11082 [35:29:33<08:37, 11.49s/step, acc=0.5073, loss=2.0182, lr=1.00e-05]
Training: 100%|█████████▉| 11038/11082 [35:29:45<08:33, 11.67s/step, acc=0.5073, loss=2.0182, lr=1.00e-05]
Training: 100%|█████████▉| 11039/11082 [35:29:56<08:19, 11.61s/step, acc=0.5073, loss=2.0182, lr=1.00e-05]
Training: 100%|█████████▉| 11040/11082 [35:30:08<08:08, 11.64s/step, acc=0.5073, loss=2.0182, lr=1.00e-05]
Training: 100%|█████████▉| 11041/11082 [35:30:19<07:53, 11.54s/step, acc=0.5073, loss=2.0182, lr=1.00e-05]
Training: 100%|█████████▉| 11042/11082 [35:30:31<07:45, 11.64s/step, acc=0.5073, loss=2.0182, lr=1.00e-05]
Training: 100%|█████████▉| 11043/11082 [35:30:42<07:29, 11.52s/step, acc=0.5073, loss=2.0182, lr=1.00e-05]
Training: 100%|█████████▉| 11044/11082 [35:30:54<07:19, 11.55s/step, acc=0.5073, loss=2.0182, lr=1.00e-05]
Training: 100%|█████████▉| 11045/11082 [35:31:05<07:07, 11.56s/step, acc=0.5073, loss=2.0182, lr=1.00e-05]
Training: 100%|█████████▉| 11046/11082 [35:31:17<07:00, 11.68s/step, acc=0.5073, loss=2.0182, lr=1.00e-05]
Training: 100%|█████████▉| 11047/11082 [35:31:28<06:40, 11.45s/step, acc=0.5073, loss=2.0182, lr=1.00e-05]
Training: 100%|█████████▉| 11048/11082 [35:31:40<06:33, 11.58s/step, acc=0.5073, loss=2.0182, lr=1.00e-05]
Training: 100%|█████████▉| 11049/11082 [35:31:52<06:22, 11.59s/step, acc=0.5073, loss=2.0182, lr=1.00e-05]
Training: 100%|█████████▉| 11050/11082 [35:32:03<06:08, 11.52s/step, acc=0.5073, loss=2.0182, lr=1.00e-05]
Training: 100%|█████████▉| 11050/11082 [35:32:10<06:08, 11.52s/step, acc=0.5055, loss=2.0278, lr=1.00e-05]2026-06-20 02:25:38 | INFO | __main__ | epoch=1 step=11050 train_loss=2.027792 train_acc=0.505501 lr=1.000e-05 grad_norm=0.5039 acc_d0=0.7641 acc_d1=0.6244 acc_d2=0.5072 acc_d3=0.4098 acc_d4=0.3318 acc_d5=0.2679 acc_d6=0.2148 acc_d7=0.1759

Training: 100%|█████████▉| 11051/11082 [35:32:15<05:57, 11.53s/step, acc=0.5055, loss=2.0278, lr=1.00e-05]
Training: 100%|█████████▉| 11052/11082 [35:32:26<05:45, 11.51s/step, acc=0.5055, loss=2.0278, lr=1.00e-05]
Training: 100%|█████████▉| 11053/11082 [35:32:38<05:33, 11.52s/step, acc=0.5055, loss=2.0278, lr=1.00e-05]
Training: 100%|█████████▉| 11054/11082 [35:32:49<05:22, 11.51s/step, acc=0.5055, loss=2.0278, lr=1.00e-05]
Training: 100%|█████████▉| 11055/11082 [35:33:01<05:10, 11.51s/step, acc=0.5055, loss=2.0278, lr=1.00e-05]
Training: 100%|█████████▉| 11056/11082 [35:33:12<04:58, 11.47s/step, acc=0.5055, loss=2.0278, lr=1.00e-05]
Training: 100%|█████████▉| 11057/11082 [35:33:24<04:50, 11.61s/step, acc=0.5055, loss=2.0278, lr=1.00e-05]
Training: 100%|█████████▉| 11058/11082 [35:33:35<04:37, 11.55s/step, acc=0.5055, loss=2.0278, lr=1.00e-05]
Training: 100%|█████████▉| 11059/11082 [35:33:47<04:28, 11.65s/step, acc=0.5055, loss=2.0278, lr=1.00e-05]
Training: 100%|█████████▉| 11060/11082 [35:33:59<04:13, 11.51s/step, acc=0.5055, loss=2.0278, lr=1.00e-05]
Training: 100%|█████████▉| 11061/11082 [35:34:10<04:02, 11.54s/step, acc=0.5055, loss=2.0278, lr=1.00e-05]
Training: 100%|█████████▉| 11062/11082 [35:34:22<03:50, 11.53s/step, acc=0.5055, loss=2.0278, lr=1.00e-05]
Training: 100%|█████████▉| 11063/11082 [35:34:33<03:40, 11.61s/step, acc=0.5055, loss=2.0278, lr=1.00e-05]
Training: 100%|█████████▉| 11064/11082 [35:34:45<03:28, 11.61s/step, acc=0.5055, loss=2.0278, lr=1.00e-05]
Training: 100%|█████████▉| 11065/11082 [35:34:57<03:17, 11.59s/step, acc=0.5055, loss=2.0278, lr=1.00e-05]
Training: 100%|█████████▉| 11066/11082 [35:35:08<03:03, 11.48s/step, acc=0.5055, loss=2.0278, lr=1.00e-05]
Training: 100%|█████████▉| 11067/11082 [35:35:19<02:52, 11.47s/step, acc=0.5055, loss=2.0278, lr=1.00e-05]
Training: 100%|█████████▉| 11068/11082 [35:35:31<02:40, 11.47s/step, acc=0.5055, loss=2.0278, lr=1.00e-05]
Training: 100%|█████████▉| 11069/11082 [35:35:42<02:30, 11.55s/step, acc=0.5055, loss=2.0278, lr=1.00e-05]
Training: 100%|█████████▉| 11070/11082 [35:35:54<02:19, 11.59s/step, acc=0.5055, loss=2.0278, lr=1.00e-05]
Training: 100%|█████████▉| 11071/11082 [35:36:05<02:06, 11.50s/step, acc=0.5055, loss=2.0278, lr=1.00e-05]
Training: 100%|█████████▉| 11072/11082 [35:36:17<01:55, 11.57s/step, acc=0.5055, loss=2.0278, lr=1.00e-05]
Training: 100%|█████████▉| 11073/11082 [35:36:29<01:43, 11.56s/step, acc=0.5055, loss=2.0278, lr=1.00e-05]
Training: 100%|█████████▉| 11074/11082 [35:36:40<01:31, 11.47s/step, acc=0.5055, loss=2.0278, lr=1.00e-05]
Training: 100%|█████████▉| 11075/11082 [35:36:52<01:21, 11.57s/step, acc=0.5055, loss=2.0278, lr=1.00e-05]
Training: 100%|█████████▉| 11076/11082 [35:37:04<01:10, 11.73s/step, acc=0.5055, loss=2.0278, lr=1.00e-05]
Training: 100%|█████████▉| 11077/11082 [35:37:15<00:58, 11.67s/step, acc=0.5055, loss=2.0278, lr=1.00e-05]
Training: 100%|█████████▉| 11078/11082 [35:37:27<00:46, 11.64s/step, acc=0.5055, loss=2.0278, lr=1.00e-05]
Training: 100%|█████████▉| 11079/11082 [35:37:39<00:35, 11.67s/step, acc=0.5055, loss=2.0278, lr=1.00e-05]
Training: 100%|█████████▉| 11080/11082 [35:37:50<00:23, 11.67s/step, acc=0.5055, loss=2.0278, lr=1.00e-05]
Training: 100%|█████████▉| 11081/11082 [35:38:02<00:11, 11.61s/step, acc=0.5055, loss=2.0278, lr=1.00e-05]
Training: 100%|██████████| 11082/11082 [35:38:12<00:00, 11.18s/step, acc=0.5055, loss=2.0278, lr=1.00e-05]2026-06-20 02:31:47 | INFO | __main__ | epoch=1 step=11082 train_loss=1.996471 train_acc=0.510854 lr=1.000e-05 grad_norm=0.5508 acc_d0=0.7599 acc_d1=0.6167 acc_d2=0.4966 acc_d3=0.4019 acc_d4=0.3247 acc_d5=0.2638 acc_d6=0.2141 acc_d7=0.1788 (trailing flush)
2026-06-20 02:31:47 | INFO | __main__ | Epoch 1 done: total_batches_seen=44326 global_step=11082
2026-06-20 02:31:49 | INFO | nemo_automodel.components.checkpoint._backports.consolidate_hf_safetensors | Consolidating safetensors files from /llm-align/liuchonghan/runs/p_eagle_kv_reuse/checkpoints/epoch_2_step_11082/model to /llm-align/liuchonghan/runs/p_eagle_kv_reuse/checkpoints/epoch_2_step_11082/model/consolidated.
2026-06-20 02:31:56 | INFO | nemo_automodel.components.checkpoint.checkpointing | Successfully exported consolidated HF safetensors to /llm-align/liuchonghan/runs/p_eagle_kv_reuse/checkpoints/epoch_2_step_11082/model/consolidated.
2026-06-20 02:31:58 | INFO | __main__ | Saved epoch checkpoint to /llm-align/liuchonghan/runs/p_eagle_kv_reuse/checkpoints/epoch_2_step_11082
2026-06-20 02:31:58 | INFO | __main__ | Training complete: global_step=11082

Training: 100%|██████████| 11082/11082 [35:38:30<00:00, 11.58s/step, acc=0.5055, loss=2.0278, lr=1.00e-05]
h-liuchonghan-translate-a8nodes-main-m-0:66739:66739 [7] NCCL INFO comm 0x564f0f810dd0 rank 7 nranks 32 cudaDev 7 busId e0000 - Destroy COMPLETE
h-liuchonghan-translate-a8nodes-main-m-0:66737:66737 [5] NCCL INFO comm 0x55f6153f0f30 rank 5 nranks 32 cudaDev 5 busId a1000 - Destroy COMPLETE
h-liuchonghan-translate-a8nodes-main-m-0:66736:66736 [4] NCCL INFO comm 0x562913d20eb0 rank 4 nranks 32 cudaDev 4 busId 9b000 - Destroy COMPLETE
h-liuchonghan-translate-a8nodes-main-m-0:66732:66732 [0] NCCL INFO comm 0x55d1c3b89b70 rank 0 nranks 32 cudaDev 0 busId b000 - Destroy COMPLETE
h-liuchonghan-translate-a8nodes-main-m-0:66738:66738 [6] NCCL INFO comm 0x55904d46bc60 rank 6 nranks 32 cudaDev 6 busId db000 - Destroy COMPLETE
h-liuchonghan-translate-a8nodes-main-m-0:66735:66735 [3] NCCL INFO comm 0x559159d41560 rank 3 nranks 32 cudaDev 3 busId 4e000 - Destroy COMPLETE
h-liuchonghan-translate-a8nodes-main-m-0:66734:66734 [2] NCCL INFO comm 0x55985c7ad020 rank 2 nranks 32 cudaDev 2 busId 48000 - Destroy COMPLETE
h-liuchonghan-translate-a8nodes-main-m-0:66733:66733 [1] NCCL INFO comm 0x55a35ad250b0 rank 1 nranks 32 cudaDev 1 busId 10000 - Destroy COMPLETE
h-liuchonghan-translate-a8nodes-main-m-0:66739:66739 [7] NCCL INFO ENV/Plugin: Closing env plugin ncclEnvDefault
h-liuchonghan-translate-a8nodes-main-m-0:66737:66737 [5] NCCL INFO ENV/Plugin: Closing env plugin ncclEnvDefault
h-liuchonghan-translate-a8nodes-main-m-0:66736:66736 [4] NCCL INFO ENV/Plugin: Closing env plugin ncclEnvDefault
h-liuchonghan-translate-a8nodes-main-m-0:66735:66735 [3] NCCL INFO ENV/Plugin: Closing env plugin ncclEnvDefault
h-liuchonghan-translate-a8nodes-main-m-0:66738:66738 [6] NCCL INFO ENV/Plugin: Closing env plugin ncclEnvDefault
h-liuchonghan-translate-a8nodes-main-m-0:66734:66734 [2] NCCL INFO ENV/Plugin: Closing env plugin ncclEnvDefault
h-liuchonghan-translate-a8nodes-main-m-0:66733:66733 [1] NCCL INFO ENV/Plugin: Closing env plugin ncclEnvDefault
h-liuchonghan-translate-a8nodes-main-m-0:66732:66732 [0] NCCL INFO ENV/Plugin: Closing env plugin ncclEnvDefault