Compare: 13_scan_layers_false

‹ 11_optimizer_offload_trueCase: 13_scan_layers_false13_scan_layers_true ›

Metrics: `main (b341212ce)` vs `fix/gate-fp8-nan-sanitization (cd5a4f898)`

Metric	main `b341212ce`	fix/gate-fp8-nan-sanitization `cd5a4f898`	Diff (fix/gate-fp8-nan-sanitization − main)
Parameters	1.104 billion	1.104 billion	—
Final loss	5.9300	5.9300	0
TFLOP/s	76.530	77.581	+1.051
Tok/s	11535.4	11693.9	+158.539
Avg s/step	2.377	2.699	+0.322
Memory %	6.85	6.85	0
JAX	0.8.1	0.8.1	—

Diff = branch value − main value. Green = branch improved. Red = branch regressed.

Field	main	fix/gate-fp8-nan-sanitization
JAX	0.8.1	0.8.1
Jaxlib	0.8.1	0.8.1
Hardware	PJRT C API / TFRT TPU v6 lite	PJRT C API / TFRT TPU v6 lite
Devices	8	8

Memory Breakdown

Field	main	fix/gate-fp8-nan-sanitization
Total	3.3 GB	3.3 GB
Output	1.5 GB	1.5 GB
Temp	1.8 GB	1.8 GB
Argument	1.5 GB	1.5 GB

main

Device	Used (GB)	Total (GB)	%
TPU_0(process=0,(0,0,0,0))	2.14	31.25	6.8%
TPU_1(process=0,(1,0,0,0))	2.14	31.25	6.8%
TPU_2(process=0,(0,1,0,0))	2.14	31.25	6.8%
TPU_3(process=0,(1,1,0,0))	2.14	31.25	6.8%
TPU_4(process=0,(0,2,0,0))	2.14	31.25	6.8%
TPU_5(process=0,(1,2,0,0))	2.14	31.25	6.8%
TPU_6(process=0,(0,3,0,0))	2.14	31.25	6.8%
TPU_7(process=0,(1,3,0,0))	2.14	31.25	6.8%

fix/gate-fp8-nan-sanitization

Device	Used (GB)	Total (GB)	%
TPU_0(process=0,(0,0,0,0))	2.14	31.25	6.8%
TPU_1(process=0,(1,0,0,0))	2.14	31.25	6.8%
TPU_2(process=0,(0,1,0,0))	2.14	31.25	6.8%
TPU_3(process=0,(1,1,0,0))	2.14	31.25	6.8%
TPU_4(process=0,(0,2,0,0))	2.14	31.25	6.8%
TPU_5(process=0,(1,2,0,0))	2.14	31.25	6.8%
TPU_6(process=0,(0,3,0,0))	2.14	31.25	6.8%
TPU_7(process=0,(1,3,0,0))	2.14	31.25	6.8%

main TFLOP/s

fix/gate-fp8-nan-sanitization TFLOP/s

Step	main Loss	fix/gate-fp8-nan-sanitization Loss	main TFLOP/s	fix/gate-fp8-nan-sanitization TFLOP/s	TFLOP/s Δ
0	10.887	10.887	0.216	0.205	-0.0110
1	10.887	10.887	8.837	6.584	-2.2530
2	9.787	9.787	564.953	172.325	-392.6280
3	8.853	8.853	92.169	372.679	+280.5100
4	7.982	7.982	203.446	121.667	-81.7790
5	7.248	7.248	123.101	123.116	+0.0150
6	6.690	6.690	0.719	0.630	-0.0890
7	6.307	6.307	71.326	335.982	+264.6560
8	6.070	6.070	59.792	135.033	+75.2410
9	5.930	5.930	76.530	77.581	+1.0510

main · b341212ce · main_20260421_144657 · full log

2026-04-21 15:29:26.119196: E external/local_xla/xla/stream_executor/cuda/cuda_fft.cc:467] Unable to register cuFFT factory: Attempting to register factory for plugin cuFFT when one has already been registered
WARNING: All log messages before absl::InitializeLog() is called are written to STDERR
E0000 00:00:1776785366.129875 1521226 cuda_dnn.cc:8579] Unable to register cuDNN factory: Attempting to register factory for plugin cuDNN when one has already been registered
E0000 00:00:1776785366.133178 1521226 cuda_blas.cc:1407] Unable to register cuBLAS factory: Attempting to register factory for plugin cuBLAS when one has already been registered
W0000 00:00:1776785366.144022 1521226 computation_placer.cc:177] computation placer already registered. Please check linkage and avoid linking the same target more than once.
W0000 00:00:1776785366.144034 1521226 computation_placer.cc:177] computation placer already registered. Please check linkage and avoid linking the same target more than once.
W0000 00:00:1776785366.144036 1521226 computation_placer.cc:177] computation placer already registered. Please check linkage and avoid linking the same target more than once.
W0000 00:00:1776785366.144037 1521226 computation_placer.cc:177] computation placer already registered. Please check linkage and avoid linking the same target more than once.
2026-04-21 15:29:29.637347: E external/local_xla/xla/stream_executor/cuda/cuda_platform.cc:51] failed call to cuInit: INTERNAL: CUDA error: Failed call to cuInit: UNKNOWN ERROR (303)
I0421 15:29:29.753963 131040467287168 max_utils.py:238] Skipping jax distributed system due to skip_jax_distributed_system=True flag.
I0421 15:30:02.224670 131040467287168 max_utils.py:800] System Information: Jax Version: 0.8.1
I0421 15:30:02.224947 131040467287168 max_utils.py:801] System Information: Jaxlib Version: 0.8.1
I0421 15:30:02.224997 131040467287168 max_utils.py:802] System Information: Jax Backend: PJRT C API
TFRT TPU v6 lite
Built on Nov 12 2025 14:16:36 (1762985796) cl/831091709
I0421 15:30:02.225027 131040467287168 train_utils.py:335] WARNING: 'dataset_path' might be pointing your local file system
I0421 15:30:02.225073 131040467287168 train_utils.py:348] WARNING: Sequence packing is essentially ignored for synthetic data. Please use a real dataset to use sequence packing.
I0421 15:30:02.225235 131040467287168 train.py:703] [DECOUPLED NO-OP] skipping cloud diagnostics wrapper.
I0421 15:30:02.581421 131040467287168 maxtext_utils.py:1551] Num_devices: 8, shape (1, 1, 1, 8, 1, 1, 1, 1, 1, 1, 1, 1, 1)
I0421 15:30:02.581799 131040467287168 checkpointing.py:677] Setting up checkpoint logger...
I0421 15:30:02.581848 131040467287168 checkpointing.py:233] Creating checkpoint manager with ocdbt=True and zarr3=True
I0421 15:30:02.581933 131040467287168 pytree_checkpoint_handler.py:577] save_device_host_concurrent_bytes=None
I0421 15:30:02.582747 131040467287168 base_pytree_checkpoint_handler.py:411] Created BasePyTreeCheckpointHandler: use_ocdbt=True, use_zarr3=True, pytree_metadata_options=PyTreeMetadataOptions(support_rich_types=False), array_metadata_store=<orbax.checkpoint._src.metadata.array_metadata_store.Store object at 0x772e09d261b0>, enable_pinned_host_transfer=False, save_concurrent_bytes: 96000000000 (89.4 GiB), restore_concurrent_bytes: 96000000000 (89.4 GiB)
I0421 15:30:05.003249 131040467287168 checkpointing.py:265] Enabling policy for fixed interval checkpointing.
I0421 15:30:05.003520 131040467287168 checkpoint_manager.py:702] [process=0][thread=MainThread] CheckpointManager init: checkpointers=None, item_names=('items',), item_handlers={'items': <orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeCheckpointHandler object at 0x772c47372780>}, handler_registry=None
I0421 15:30:05.003899 131040467287168 composite_checkpoint_handler.py:237] Deferred registration for item: "items". Adding handler `<orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeCheckpointHandler object at 0x772c47372780>` for item "items" and save args `<class 'orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeSaveArgs'>` and restore args `<class 'orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeRestoreArgs'>` to `_handler_registry`.
I0421 15:30:05.003941 131040467287168 composite_checkpoint_handler.py:237] Deferred registration for item: "metrics". Adding handler `<orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonCheckpointHandler object at 0x772c47373050>` for item "metrics" and save args `<class 'orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonSaveArgs'>` and restore args `<class 'orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonRestoreArgs'>` to `_handler_registry`.
I0421 15:30:05.003982 131040467287168 composite_checkpoint_handler.py:505] Initialized registry DefaultCheckpointHandlerRegistry({('items', <class 'orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeSaveArgs'>): <orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeCheckpointHandler object at 0x772c47372780>, ('items', <class 'orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeRestoreArgs'>): <orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeCheckpointHandler object at 0x772c47372780>, ('metrics', <class 'orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonSaveArgs'>): <orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonCheckpointHandler object at 0x772c47373050>, ('metrics', <class 'orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonRestoreArgs'>): <orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonCheckpointHandler object at 0x772c47373050>}).
I0421 15:30:05.004303 131040467287168 abstract_checkpointer.py:35] orbax-checkpoint version: 0.11.28
I0421 15:30:05.004366 131040467287168 async_checkpointer.py:177] [process=0][thread=MainThread] Using barrier_sync_fn: <function get_barrier_sync_fn.<locals>.<lambda> at 0x772c473951c0> timeout: 600 secs and primary_host=0 for async checkpoint writes
I0421 15:30:05.136100 131040467287168 checkpoint_manager.py:1788] Found 0 checkpoint steps in gs://wanglance-maxtext/linen_ckpt_main_20260421_144657/linen_main_20260421_144657_13_scan_layers_false/checkpoints
I0421 15:30:05.136374 131040467287168 checkpoint_manager.py:921] [process=0][thread=MainThread] CheckpointManager created,  primary_host=0, CheckpointManagerOptions=CheckpointManagerOptions(save_interval_steps=1, max_to_keep=None, keep_time_interval=None, keep_period=None, should_keep_fn=None, best_fn=None, best_mode='max', keep_checkpoints_without_metrics=True, step_prefix=None, step_format_fixed_length=None, step_name_format=None, create=True, cleanup_tmp_directories=False, save_on_steps=frozenset(), single_host_load_and_broadcast=False, todelete_subdir=None, todelete_full_path=None, enable_hns=False, enable_background_delete=False, read_only=False, enable_async_checkpointing=True, async_options=None, multiprocessing_options=MultiprocessingOptions(primary_host=0, active_processes=None, barrier_sync_key_prefix=None), should_save_fn=None, file_options=FileOptions(path_permission_mode=None), save_root_metadata=True, temporary_path_class=None, save_decision_policy=FixedIntervalPolicy(interval=10), preservation_policy=LatestN(n=None), prevent_write_metrics=False, enable_should_save_is_saving_in_progress_check=True, enable_per_process_directory_creation=False), root_directory=gs://wanglance-maxtext/linen_ckpt_main_20260421_144657/linen_main_20260421_144657_13_scan_layers_false/checkpoints: <orbax.checkpoint.checkpoint_manager.CheckpointManager object at 0x772c9bd33f20>
I0421 15:30:05.136468 131040467287168 checkpointing.py:301] Checkpoint manager created!
I0421 15:30:05.808710 131040467287168 nnx_wrappers.py:437] Unknown Logical: bfloat16[8,2048,2048]....................................... ('activation_batch', 'activation_norm_length', 'activation_embed').
I0421 15:30:05.808852 131040467287168 nnx_wrappers.py:437] Unknown Physical: bfloat16[8,2048,2048]....................................... ('fsdp', None, None).
I0421 15:30:06.458740 131040467287168 attentions.py:1088] attentions/inputs_q Logical: bfloat16[8,2048,2048]....................................... ('activation_batch', 'activation_attn_length', 'activation_attn_embed').
I0421 15:30:06.458848 131040467287168 attentions.py:1088] attentions/inputs_q Physical: bfloat16[8,2048,2048]....................................... ('fsdp', None, None).
I0421 15:30:06.471009 131040467287168 attentions.py:1089] attentions/inputs_kv Logical: bfloat16[8,2048,2048]....................................... ('activation_batch', 'activation_attn_length', 'activation_attn_embed').
I0421 15:30:06.471110 131040467287168 attentions.py:1089] attentions/inputs_kv Physical: bfloat16[8,2048,2048]....................................... ('fsdp', None, None).
I0421 15:30:06.493095 131040467287168 attentions.py:1154] attentions/query Logical: bfloat16[8,2048,16,128]..................................... ('activation_kv_batch', 'activation_attn_length', 'activation_kv_heads', 'activation_kv_head_dim').
I0421 15:30:06.493178 131040467287168 attentions.py:1154] attentions/query Physical: bfloat16[8,2048,16,128]..................................... ('fsdp', None, None, None).
I0421 15:30:06.505471 131040467287168 attentions.py:1155] attentions/key Logical: bfloat16[8,2048,16,128]..................................... ('activation_kv_batch', 'activation_attn_length', 'activation_kv_heads', 'activation_kv_head_dim').
I0421 15:30:06.505546 131040467287168 attentions.py:1155] attentions/key Physical: bfloat16[8,2048,16,128]..................................... ('fsdp', None, None, None).
I0421 15:30:06.517673 131040467287168 attentions.py:1156] attentions/value Logical: bfloat16[8,2048,16,128]..................................... ('activation_kv_batch', 'activation_attn_length', 'activation_kv_heads', 'activation_kv_head_dim').
I0421 15:30:06.517752 131040467287168 attentions.py:1156] attentions/value Physical: bfloat16[8,2048,16,128]..................................... ('fsdp', None, None, None).
I0421 15:30:06.538085 131040467287168 attentions.py:1197] attentions/out Logical: bfloat16[8,2048,16,128]..................................... ('activation_batch', 'activation_attn_length', 'activation_heads', 'activation_kv').
I0421 15:30:06.538166 131040467287168 attentions.py:1197] attentions/out Physical: bfloat16[8,2048,16,128]..................................... ('fsdp', None, None, None).
I0421 15:30:06.555620 131040467287168 linears.py:525] linears/x Logical: bfloat16[8,2048,7168]....................................... ('activation_batch', 'activation_length', 'activation_mlp').
I0421 15:30:06.555700 131040467287168 linears.py:525] linears/x Physical: bfloat16[8,2048,7168]....................................... ('fsdp', None, None).
I0421 15:30:09.020594 131040467287168 checkpointing.py:577] checkpoint manager exists so trying to load this run's existing checkpoint
I0421 15:30:09.020712 131040467287168 checkpointing.py:665] No existing checkpoints found, not restoring checkpoint.
[DECOUPLED NO-OP] gcs_storage: using stubs.
[DECOUPLED NO-OP] mldiagnostics: using stub.
[DECOUPLED NO-OP] mldiagnostics: using stub.
[DECOUPLED NO-OP] mldiagnostics: using stub.
[DECOUPLED NO-OP] workload_monitor: using stub.
[DECOUPLED NO-OP] vertex_tensorboard: using stub.
fsdp: 8
I0421 15:30:18.057115 131040467287168 maxtext_utils.py:1654]  params/params/decoder/decoder_norm/scale
    Shape:     float32[2048]
    Logical:   PartitionSpec('norm',)
    Physical:  (None,)
I0421 15:30:18.057321 131040467287168 maxtext_utils.py:1654]  params/params/decoder/layers_0/mlp/wi_0/kernel
    Shape:     float32[2048,7168]
    Logical:   PartitionSpec('embed', 'mlp')
    Physical:  ('fsdp', None)
I0421 15:30:18.057364 131040467287168 maxtext_utils.py:1654]  params/params/decoder/layers_0/mlp/wi_1/kernel
    Shape:     float32[2048,7168]
    Logical:   PartitionSpec('embed', 'mlp')
    Physical:  ('fsdp', None)
I0421 15:30:18.057407 131040467287168 maxtext_utils.py:1654]  params/params/decoder/layers_0/mlp/wo/kernel
    Shape:     float32[7168,2048]
    Logical:   PartitionSpec('mlp', 'embed')
    Physical:  (None, 'fsdp')
I0421 15:30:18.057437 131040467287168 maxtext_utils.py:1654]  params/params/decoder/layers_0/post_self_attention_layer_norm/scale
    Shape:     float32[2048]
    Logical:   PartitionSpec('norm',)
    Physical:  (None,)
I0421 15:30:18.057464 131040467287168 maxtext_utils.py:1654]  params/params/decoder/layers_0/pre_self_attention_layer_norm/scale
    Shape:     float32[2048]
    Logical:   PartitionSpec('norm',)
    Physical:  (None,)
I0421 15:30:18.057503 131040467287168 maxtext_utils.py:1654]  params/params/decoder/layers_0/self_attention/key/kernel
    Shape:     float32[2048,16,128]
    Logical:   PartitionSpec('embed', 'kv_heads', 'kv_head_dim')
    Physical:  ('fsdp', None, None)
I0421 15:30:18.057542 131040467287168 maxtext_utils.py:1654]  params/params/decoder/layers_0/self_attention/out/kernel
    Shape:     float32[16,128,2048]
    Logical:   PartitionSpec('heads', 'kv', 'embed')
    Physical:  (None, None, 'fsdp')
I0421 15:30:18.057569 131040467287168 maxtext_utils.py:1654]  params/params/decoder/layers_0/self_attention/query/kernel
    Shape:     float32[2048,16,128]
    Logical:   PartitionSpec('embed', 'q_heads', 'kv')
    Physical:  ('fsdp', None, None)
I0421 15:30:18.057610 131040467287168 maxtext_utils.py:1654]  params/params/decoder/layers_0/self_attention/value/kernel
    Shape:     float32[2048,16,128]
    Logical:   PartitionSpec('embed', 'kv_heads', 'kv_head_dim')
    Physical:  ('fsdp', None, None)
I0421 15:30:18.057636 131040467287168 maxtext_utils.py:1654]  params/params/decoder/layers_1/mlp/wi_0/kernel
    Shape:     float32[2048,7168]
    Logical:   PartitionSpec('embed', 'mlp')
    Physical:  ('fsdp', None)
I0421 15:30:18.057660 131040467287168 maxtext_utils.py:1654]  params/params/decoder/layers_1/mlp/wi_1/kernel
    Shape:     float32[2048,7168]
    Logical:   PartitionSpec('embed', 'mlp')
    Physical:  ('fsdp', None)
I0421 15:30:18.057682 131040467287168 maxtext_utils.py:1654]  params/params/decoder/layers_1/mlp/wo/kernel
    Shape:     float32[7168,2048]
    Logical:   PartitionSpec('mlp', 'embed')
    Physical:  (None, 'fsdp')
I0421 15:30:18.057705 131040467287168 maxtext_utils.py:1654]  params/params/decoder/layers_1/post_self_attention_layer_norm/scale
    Shape:     float32[2048]
    Logical:   PartitionSpec('norm',)
    Physical:  (None,)
I0421 15:30:18.057727 131040467287168 maxtext_utils.py:1654]  params/params/decoder/layers_1/pre_self_attention_layer_norm/scale
    Shape:     float32[2048]
    Logical:   PartitionSpec('norm',)
    Physical:  (None,)
I0421 15:30:18.057751 131040467287168 maxtext_utils.py:1654]  params/params/decoder/layers_1/self_attention/key/kernel
    Shape:     float32[2048,16,128]
    Logical:   PartitionSpec('embed', 'kv_heads', 'kv_head_dim')
    Physical:  ('fsdp', None, None)
I0421 15:30:18.057775 131040467287168 maxtext_utils.py:1654]  params/params/decoder/layers_1/self_attention/out/kernel
    Shape:     float32[16,128,2048]
    Logical:   PartitionSpec('heads', 'kv', 'embed')
    Physical:  (None, None, 'fsdp')
I0421 15:30:18.057799 131040467287168 maxtext_utils.py:1654]  params/params/decoder/layers_1/self_attention/query/kernel
    Shape:     float32[2048,16,128]
    Logical:   PartitionSpec('embed', 'q_heads', 'kv')
    Physical:  ('fsdp', None, None)
I0421 15:30:18.057824 131040467287168 maxtext_utils.py:1654]  params/params/decoder/layers_1/self_attention/value/kernel
    Shape:     float32[2048,16,128]
    Logical:   PartitionSpec('embed', 'kv_heads', 'kv_head_dim')
    Physical:  ('fsdp', None, None)
I0421 15:30:18.057847 131040467287168 maxtext_utils.py:1654]  params/params/decoder/layers_10/mlp/wi_0/kernel
    Shape:     float32[2048,7168]
    Logical:   PartitionSpec('embed', 'mlp')
    Physical:  ('fsdp', None)
I0421 15:30:18.057869 131040467287168 maxtext_utils.py:1654]  params/params/decoder/layers_10/mlp/wi_1/kernel
    Shape:     float32[2048,7168]
    Logical:   PartitionSpec('embed', 'mlp')
    Physical:  ('fsdp', None)
I0421 15:30:18.057891 131040467287168 maxtext_utils.py:1654]  params/params/decoder/layers_10/mlp/wo/kernel
    Shape:     float32[7168,2048]
    Logical:   PartitionSpec('mlp', 'embed')
    Physical:  (None, 'fsdp')
I0421 15:30:18.057912 131040467287168 maxtext_utils.py:1654]  params/params/decoder/layers_10/post_self_attention_layer_norm/scale
    Shape:     float32[2048]
    Logical:   PartitionSpec('norm',)
    Physical:  (None,)
I0421 15:30:18.057933 131040467287168 maxtext_utils.py:1654]  params/params/decoder/layers_10/pre_self_attention_layer_norm/scale
    Shape:     float32[2048]
    Logical:   PartitionSpec('norm',)
    Physical:  (None,)
I0421 15:30:18.057957 131040467287168 maxtext_utils.py:1654]  params/params/decoder/layers_10/self_attention/key/kernel
    Shape:     float32[2048,16,128]
    Logical:   PartitionSpec('embed', 'kv_heads', 'kv_head_dim')
    Physical:  ('fsdp', None, None)
I0421 15:30:18.057988 131040467287168 maxtext_utils.py:1654]  params/params/decoder/layers_10/self_attention/out/kernel
    Shape:     float32[16,128,2048]
    Logical:   PartitionSpec('heads', 'kv', 'embed')
    Physical:  (None, None, 'fsdp')
I0421 15:30:18.058012 131040467287168 maxtext_utils.py:1654]  params/params/decoder/layers_10/self_attention/query/kernel
    Shape:     float32[2048,16,128]
    Logical:   PartitionSpec('embed', 'q_heads', 'kv')
    Physical:  ('fsdp', None, None)
I0421 15:30:18.058033 131040467287168 maxtext_utils.py:1654]  params/params/decoder/layers_10/self_attention/value/kernel
    Shape:     float32[2048,16,128]
    Logical:   PartitionSpec('embed', 'kv_heads', 'kv_head_dim')
    Physical:  ('fsdp', None, None)
I0421 15:30:18.058071 131040467287168 maxtext_utils.py:1654]  params/params/decoder/layers_11/mlp/wi_0/kernel
    Shape:     float32[2048,7168]
    Logical:   PartitionSpec('embed', 'mlp')
    Physical:  ('fsdp', None)
I0421 15:30:18.058095 131040467287168 maxtext_utils.py:1654]  params/params/decoder/layers_11/mlp/wi_1/kernel
    Shape:     float32[2048,7168]
    Logical:   PartitionSpec('embed', 'mlp')
    Physical:  ('fsdp', None)
I0421 15:30:18.058117 131040467287168 maxtext_utils.py:1654]  params/params/decoder/layers_11/mlp/wo/kernel
    Shape:     float32[7168,2048]
    Logical:   PartitionSpec('mlp', 'embed')
    Physical:  (None, 'fsdp')
I0421 15:30:18.058138 131040467287168 maxtext_utils.py:1654]  params/params/decoder/layers_11/post_self_attention_layer_norm/scale
    Shape:     float32[2048]
    Logical:   PartitionSpec('norm',)
    Physical:  (None,)
I0421 15:30:18.058159 131040467287168 maxtext_utils.py:1654]  params/params/decoder/layers_11/pre_self_attention_layer_norm/scale
    Shape:     float32[2048]
    Logical:   PartitionSpec('norm',)
    Physical:  (None,)
I0421 15:30:18.058181 131040467287168 maxtext_utils.py:1654]  params/params/decoder/layers_11/self_attention/key/kernel
    Shape:     float32[2048,16,128]
    Logical:   PartitionSpec('embed', 'kv_heads', 'kv_head_dim')
    Physical:  ('fsdp', None, None)
I0421 15:30:18.058204 131040467287168 maxtext_utils.py:1654]  params/params/decoder/layers_11/self_attention/out/kernel
    Shape:     float32[16,128,2048]
    Logical:   PartitionSpec('heads', 'kv', 'embed')
    Physical:  (None, None, 'fsdp')
I0421 15:30:18.058226 131040467287168 maxtext_utils.py:1654]  params/params/decoder/layers_11/self_attention/query/kernel
    Shape:     float32[2048,16,128]
    Logical:   PartitionSpec('embed', 'q_heads', 'kv')
    Physical:  ('fsdp', None, None)
I0421 15:30:18.058248 131040467287168 maxtext_utils.py:1654]  params/params/decoder/layers_11/self_attention/value/kernel
    Shape:     float32[2048,16,128]
    Logical:   PartitionSpec('embed', 'kv_heads', 'kv_head_dim')
    Physical:  ('fsdp', None, None)
I0421 15:30:18.058271 131040467287168 maxtext_utils.py:1654]  params/params/decoder/layers_12/mlp/wi_0/kernel
    Shape:     float32[2048,7168]
    Logical:   PartitionSpec('embed', 'mlp')
    Physical:  ('fsdp', None)
I0421 15:30:18.058293 131040467287168 maxtext_utils.py:1654]  params/params/decoder/layers_12/mlp/wi_1/kernel
    Shape:     float32[2048,7168]
    Logical:   PartitionSpec('embed', 'mlp')
    Physical:  ('fsdp', None)
I0421 15:30:18.058316 131040467287168 maxtext_utils.py:1654]  params/params/decoder/layers_12/mlp/wo/kernel
    Shape:     float32[7168,2048]
    Logical:   PartitionSpec('mlp', 'embed')
    Physical:  (None, 'fsdp')
I0421 15:30:18.058336 131040467287168 maxtext_utils.py:1654]  params/params/decoder/layers_12/post_self_attention_layer_norm/scale
    Shape:     float32[2048]
    Logical:   PartitionSpec('norm',)
    Physical:  (None,)
I0421 15:30:18.058357 131040467287168 maxtext_utils.py:1654]  params/params/decoder/layers_12/pre_self_attention_layer_norm/scale
    Shape:     float32[2048]
    Logical:   PartitionSpec('norm',)
    Physical:  (None,)
I0421 15:30:18.058379 131040467287168 maxtext_utils.py:1654]  params/params/decoder/layers_12/self_attention/key/kernel
    Shape:     float32[2048,16,128]
    Logical:   PartitionSpec('embed', 'kv_heads', 'kv_head_dim')
    Physical:  ('fsdp', None, None)
I0421 15:30:18.058401 131040467287168 maxtext_utils.py:1654]  params/params/decoder/layers_12/self_attention/out/kernel
    Shape:     float32[16,128,2048]
    Logical:   PartitionSpec('heads', 'kv', 'embed')
    Physical:  (None, None, 'fsdp')
I0421 15:30:18.058423 131040467287168 maxtext_utils.py:1654]  params/params/decoder/layers_12/self_attention/query/kernel
    Shape:     float32[2048,16,128]
    Logical:   PartitionSpec('embed', 'q_heads', 'kv')
    Physical:  ('fsdp', None, None)
I0421 15:30:18.058446 131040467287168 maxtext_utils.py:1654]  params/params/decoder/layers_12/self_attention/value/kernel
    Shape:     float32[2048,16,128]
    Logical:   PartitionSpec('embed', 'kv_heads', 'kv_head_dim')
    Physical:  ('fsdp', None, None)
I0421 15:30:18.058467 131040467287168 maxtext_utils.py:1654]  params/params/decoder/layers_13/mlp/wi_0/kernel
    Shape:     float32[2048,7168]
    Logical:   PartitionSpec('embed', 'mlp')
    Physical:  ('fsdp', None)
I0421 15:30:18.058489 131040467287168 maxtext_utils.py:1654]  params/params/decoder/layers_13/mlp/wi_1/kernel
    Shape:     float32[2048,7168]
    Logical:   PartitionSpec('embed', 'mlp')
    Physical:  ('fsdp', None)
I0421 15:30:18.058511 131040467287168 maxtext_utils.py:1654]  params/params/decoder/layers_13/mlp/wo/kernel
    Shape:     float32[7168,2048]
    Logical:   PartitionSpec('mlp', 'embed')
    Physical:  (None, 'fsdp')
I0421 15:30:18.058532 131040467287168 maxtext_utils.py:1654]  params/params/decoder/layers_13/post_self_attention_layer_norm/scale
    Shape:     float32[2048]
    Logical:   PartitionSpec('norm',)
    Physical:  (None,)
I0421 15:30:18.058552 131040467287168 maxtext_utils.py:1654]  params/params/decoder/layers_13/pre_self_attention_layer_norm/scale
    Shape:     float32[2048]
    Logical:   PartitionSpec('norm',)
    Physical:  (None,)
I0421 15:30:18.058573 131040467287168 maxtext_utils.py:1654]  params/params/decoder/layers_13/self_attention/key/kernel
    Shape:     float32[2048,16,128]
    Logical:   PartitionSpec('embed', 'kv_heads', 'kv_head_dim')
    Physical:  ('fsdp', None, None)
I0421 15:30:18.058596 131040467287168 maxtext_utils.py:1654]  params/params/decoder/layers_13/self_attention/out/kernel
    Shape:     float32[16,128,2048]
    Logical:   PartitionSpec('heads', 'kv', 'embed')
    Physical:  (None, None, 'fsdp')
I0421 15:30:18.058618 131040467287168 maxtext_utils.py:1654]  params/params/decoder/layers_13/self_attention/query/kernel
    Shape:     float32[2048,16,128]
    Logical:   PartitionSpec('embed', 'q_heads', 'kv')
    Physical:  ('fsdp', None, None)
I0421 15:30:18.058640 131040467287168 maxtext_utils.py:1654]  params/params/decoder/layers_13/self_attention/value/kernel
    Shape:     float32[2048,16,128]
    Logical:   PartitionSpec('embed', 'kv_heads', 'kv_head_dim')
    Physical:  ('fsdp', None, None)
I0421 15:30:18.058663 131040467287168 maxtext_utils.py:1654]  params/params/decoder/layers_14/mlp/wi_0/kernel
    Shape:     float32[2048,7168]
    Logical:   PartitionSpec('embed', 'mlp')
    Physical:  ('fsdp', None)
I0421 15:30:18.058684 131040467287168 maxtext_utils.py:1654]  params/params/decoder/layers_14/mlp/wi_1/kernel
    Shape:     float32[2048,7168]
    Logical:   PartitionSpec('embed', 'mlp')
    Physical:  ('fsdp', None)
I0421 15:30:18.058706 131040467287168 maxtext_utils.py:1654]  params/params/decoder/layers_14/mlp/wo/kernel
    Shape:     float32[7168,2048]
    Logical:   PartitionSpec('mlp', 'embed')
    Physical:  (None, 'fsdp')
I0421 15:30:18.058727 131040467287168 maxtext_utils.py:1654]  params/params/decoder/layers_14/post_self_attention_layer_norm/scale
    Shape:     float32[2048]
    Logical:   PartitionSpec('norm',)
    Physical:  (None,)
I0421 15:30:18.058746 131040467287168 maxtext_utils.py:1654]  params/params/decoder/layers_14/pre_self_attention_layer_norm/scale
    Shape:     float32[2048]
    Logical:   PartitionSpec('norm',)
    Physical:  (None,)
I0421 15:30:18.058768 131040467287168 maxtext_utils.py:1654]  params/params/decoder/layers_14/self_attention/key/kernel
    Shape:     float32[2048,16,128]
    Logical:   PartitionSpec('embed', 'kv_heads', 'kv_head_dim')
    Physical:  ('fsdp', None, None)
I0421 15:30:18.058790 131040467287168 maxtext_utils.py:1654]  params/params/decoder/layers_14/self_attention/out/kernel
    Shape:     float32[16,128,2048]
    Logical:   PartitionSpec('heads', 'kv', 'embed')
    Physical:  (None, None, 'fsdp')
I0421 15:30:18.058811 131040467287168 maxtext_utils.py:1654]  params/params/decoder/layers_14/self_attention/query/kernel
    Shape:     float32[2048,16,128]
    Logical:   PartitionSpec('embed', 'q_heads', 'kv')
    Physical:  ('fsdp', None, None)
I0421 15:30:18.058833 131040467287168 maxtext_utils.py:1654]  params/params/decoder/layers_14/self_attention/value/kernel
    Shape:     float32[2048,16,128]
    Logical:   PartitionSpec('embed', 'kv_heads', 'kv_head_dim')
    Physical:  ('fsdp', None, None)
I0421 15:30:18.058855 131040467287168 maxtext_utils.py:1654]  params/params/decoder/layers_15/mlp/wi_0/kernel
    Shape:     float32[2048,7168]
    Logical:   PartitionSpec('embed', 'mlp')
    Physical:  ('fsdp', None)
I0421 15:30:18.058877 131040467287168 maxtext_utils.py:1654]  params/params/decoder/layers_15/mlp/wi_1/kernel
    Shape:     float32[2048,7168]
    Logical:   PartitionSpec('embed', 'mlp')
    Physical:  ('fsdp', None)
I0421 15:30:18.058898 131040467287168 maxtext_utils.py:1654]  params/params/decoder/layers_15/mlp/wo/kernel
    Shape:     float32[7168,2048]
    Logical:   PartitionSpec('mlp', 'embed')
    Physical:  (None, 'fsdp')
I0421 15:30:18.058919 131040467287168 maxtext_utils.py:1654]  params/params/decoder/layers_15/post_self_attention_layer_norm/scale
    Shape:     float32[2048]
    Logical:   PartitionSpec('norm',)
    Physical:  (None,)
I0421 15:30:18.058939 131040467287168 maxtext_utils.py:1654]  params/params/decoder/layers_15/pre_self_attention_layer_norm/scale
    Shape:     float32[2048]
    Logical:   PartitionSpec('norm',)
    Physical:  (None,)
I0421 15:30:18.058961 131040467287168 maxtext_utils.py:1654]  params/params/decoder/layers_15/self_attention/key/kernel
    Shape:     float32[2048,16,128]
    Logical:   PartitionSpec('embed', 'kv_heads', 'kv_head_dim')
    Physical:  ('fsdp', None, None)
I0421 15:30:18.058987 131040467287168 maxtext_utils.py:1654]  params/params/decoder/layers_15/self_attention/out/kernel
    Shape:     float32[16,128,2048]
    Logical:   PartitionSpec('heads', 'kv', 'embed')
    Physical:  (None, None, 'fsdp')
I0421 15:30:18.059009 131040467287168 maxtext_utils.py:1654]  params/params/decoder/layers_15/self_attention/query/kernel
    Shape:     float32[2048,16,128]
    Logical:   PartitionSpec('embed', 'q_heads', 'kv')
    Physical:  ('fsdp', None, None)
I0421 15:30:18.059031 131040467287168 maxtext_utils.py:1654]  params/params/decoder/layers_15/self_attention/value/kernel
    Shape:     float32[2048,16,128]
    Logical:   PartitionSpec('embed', 'kv_heads', 'kv_head_dim')
    Physical:  ('fsdp', None, None)
I0421 15:30:18.059063 131040467287168 maxtext_utils.py:1654]  params/params/decoder/layers_2/mlp/wi_0/kernel
    Shape:     float32[2048,7168]
    Logical:   PartitionSpec('embed', 'mlp')
    Physical:  ('fsdp', None)
I0421 15:30:18.059086 131040467287168 maxtext_utils.py:1654]  params/params/decoder/layers_2/mlp/wi_1/kernel
    Shape:     float32[2048,7168]
    Logical:   PartitionSpec('embed', 'mlp')
    Physical:  ('fsdp', None)
I0421 15:30:18.059107 131040467287168 maxtext_utils.py:1654]  params/params/decoder/layers_2/mlp/wo/kernel
    Shape:     float32[7168,2048]
    Logical:   PartitionSpec('mlp', 'embed')
    Physical:  (None, 'fsdp')
I0421 15:30:18.059128 131040467287168 maxtext_utils.py:1654]  params/params/decoder/layers_2/post_self_attention_layer_norm/scale
    Shape:     float32[2048]
    Logical:   PartitionSpec('norm',)
    Physical:  (None,)
I0421 15:30:18.059149 131040467287168 maxtext_utils.py:1654]  params/params/decoder/layers_2/pre_self_attention_layer_norm/scale
    Shape:     float32[2048]
    Logical:   PartitionSpec('norm',)
    Physical:  (None,)
I0421 15:30:18.059171 131040467287168 maxtext_utils.py:1654]  params/params/decoder/layers_2/self_attention/key/kernel
    Shape:     float32[2048,16,128]
    Logical:   PartitionSpec('embed', 'kv_heads', 'kv_head_dim')
    Physical:  ('fsdp', None, None)
I0421 15:30:18.059193 131040467287168 maxtext_utils.py:1654]  params/params/decoder/layers_2/self_attention/out/kernel
    Shape:     float32[16,128,2048]
    Logical:   PartitionSpec('heads', 'kv', 'embed')
    Physical:  (None, None, 'fsdp')
I0421 15:30:18.059215 131040467287168 maxtext_utils.py:1654]  params/params/decoder/layers_2/self_attention/query/kernel
    Shape:     float32[2048,16,128]
    Logical:   PartitionSpec('embed', 'q_heads', 'kv')
    Physical:  ('fsdp', None, None)
I0421 15:30:18.059237 131040467287168 maxtext_utils.py:1654]  params/params/decoder/layers_2/self_attention/value/kernel
    Shape:     float32[2048,16,128]
    Logical:   PartitionSpec('embed', 'kv_heads', 'kv_head_dim')
    Physical:  ('fsdp', None, None)
I0421 15:30:18.059259 131040467287168 maxtext_utils.py:1654]  params/params/decoder/layers_3/mlp/wi_0/kernel
    Shape:     float32[2048,7168]
    Logical:   PartitionSpec('embed', 'mlp')
    Physical:  ('fsdp', None)
I0421 15:30:18.059280 131040467287168 maxtext_utils.py:1654]  params/params/decoder/layers_3/mlp/wi_1/kernel
    Shape:     float32[2048,7168]
    Logical:   PartitionSpec('embed', 'mlp')
    Physical:  ('fsdp', None)
I0421 15:30:18.059301 131040467287168 maxtext_utils.py:1654]  params/params/decoder/layers_3/mlp/wo/kernel
    Shape:     float32[7168,2048]
    Logical:   PartitionSpec('mlp', 'embed')
    Physical:  (None, 'fsdp')
I0421 15:30:18.059321 131040467287168 maxtext_utils.py:1654]  params/params/decoder/layers_3/post_self_attention_layer_norm/scale
    Shape:     float32[2048]
    Logical:   PartitionSpec('norm',)
    Physical:  (None,)
I0421 15:30:18.059344 131040467287168 maxtext_utils.py:1654]  params/params/decoder/layers_3/pre_self_attention_layer_norm/scale
    Shape:     float32[2048]
    Logical:   PartitionSpec('norm',)
    Physical:  (None,)
I0421 15:30:18.059366 131040467287168 maxtext_utils.py:1654]  params/params/decoder/layers_3/self_attention/key/kernel
    Shape:     float32[2048,16,128]
    Logical:   PartitionSpec('embed', 'kv_heads', 'kv_head_dim')
    Physical:  ('fsdp', None, None)
I0421 15:30:18.059388 131040467287168 maxtext_utils.py:1654]  params/params/decoder/layers_3/self_attention/out/kernel
    Shape:     float32[16,128,2048]
    Logical:   PartitionSpec('heads', 'kv', 'embed')
    Physical:  (None, None, 'fsdp')
I0421 15:30:18.059410 131040467287168 maxtext_utils.py:1654]  params/params/decoder/layers_3/self_attention/query/kernel
    Shape:     float32[2048,16,128]
    Logical:   PartitionSpec('embed', 'q_heads', 'kv')
    Physical:  ('fsdp', None, None)
I0421 15:30:18.059432 131040467287168 maxtext_utils.py:1654]  params/params/decoder/layers_3/self_attention/value/kernel
    Shape:     float32[2048,16,128]
    Logical:   PartitionSpec('embed', 'kv_heads', 'kv_head_dim')
    Physical:  ('fsdp', None, None)
I0421 15:30:18.059453 131040467287168 maxtext_utils.py:1654]  params/params/decoder/layers_4/mlp/wi_0/kernel
    Shape:     float32[2048,7168]
    Logical:   PartitionSpec('embed', 'mlp')
    Physical:  ('fsdp', None)
I0421 15:30:18.059474 131040467287168 maxtext_utils.py:1654]  params/params/decoder/layers_4/mlp/wi_1/kernel
    Shape:     float32[2048,7168]
    Logical:   PartitionSpec('embed', 'mlp')
    Physical:  ('fsdp', None)
I0421 15:30:18.059495 131040467287168 maxtext_utils.py:1654]  params/params/decoder/layers_4/mlp/wo/kernel
    Shape:     float32[7168,2048]
    Logical:   PartitionSpec('mlp', 'embed')
    Physical:  (None, 'fsdp')
I0421 15:30:18.059515 131040467287168 maxtext_utils.py:1654]  params/params/decoder/layers_4/post_self_attention_layer_norm/scale
    Shape:     float32[2048]
    Logical:   PartitionSpec('norm',)
    Physical:  (None,)
I0421 15:30:18.059535 131040467287168 maxtext_utils.py:1654]  params/params/decoder/layers_4/pre_self_attention_layer_norm/scale
    Shape:     float32[2048]
    Logical:   PartitionSpec('norm',)
    Physical:  (None,)
I0421 15:30:18.059556 131040467287168 maxtext_utils.py:1654]  params/params/decoder/layers_4/self_attention/key/kernel
    Shape:     float32[2048,16,128]
    Logical:   PartitionSpec('embed', 'kv_heads', 'kv_head_dim')
    Physical:  ('fsdp', None, None)
I0421 15:30:18.059578 131040467287168 maxtext_utils.py:1654]  params/params/decoder/layers_4/self_attention/out/kernel
    Shape:     float32[16,128,2048]
    Logical:   PartitionSpec('heads', 'kv', 'embed')
    Physical:  (None, None, 'fsdp')
I0421 15:30:18.059599 131040467287168 maxtext_utils.py:1654]  params/params/decoder/layers_4/self_attention/query/kernel
    Shape:     float32[2048,16,128]
    Logical:   PartitionSpec('embed', 'q_heads', 'kv')
    Physical:  ('fsdp', None, None)
I0421 15:30:18.059620 131040467287168 maxtext_utils.py:1654]  params/params/decoder/layers_4/self_attention/value/kernel
    Shape:     float32[2048,16,128]
    Logical:   PartitionSpec('embed', 'kv_heads', 'kv_head_dim')
    Physical:  ('fsdp', None, None)
I0421 15:30:18.059642 131040467287168 maxtext_utils.py:1654]  params/params/decoder/layers_5/mlp/wi_0/kernel
    Shape:     float32[2048,7168]
    Logical:   PartitionSpec('embed', 'mlp')
    Physical:  ('fsdp', None)
I0421 15:30:18.059664 131040467287168 maxtext_utils.py:1654]  params/params/decoder/layers_5/mlp/wi_1/kernel
    Shape:     float32[2048,7168]
    Logical:   PartitionSpec('embed', 'mlp')
    Physical:  ('fsdp', None)
I0421 15:30:18.059685 131040467287168 maxtext_utils.py:1654]  params/params/decoder/layers_5/mlp/wo/kernel
    Shape:     float32[7168,2048]
    Logical:   PartitionSpec('mlp', 'embed')
    Physical:  (None, 'fsdp')
I0421 15:30:18.059705 131040467287168 maxtext_utils.py:1654]  params/params/decoder/layers_5/post_self_attention_layer_norm/scale
    Shape:     float32[2048]
    Logical:   PartitionSpec('norm',)
    Physical:  (None,)
I0421 15:30:18.059725 131040467287168 maxtext_utils.py:1654]  params/params/decoder/layers_5/pre_self_attention_layer_norm/scale
    Shape:     float32[2048]
    Logical:   PartitionSpec('norm',)
    Physical:  (None,)
I0421 15:30:18.059747 131040467287168 maxtext_utils.py:1654]  params/params/decoder/layers_5/self_attention/key/kernel
    Shape:     float32[2048,16,128]
    Logical:   PartitionSpec('embed', 'kv_heads', 'kv_head_dim')
    Physical:  ('fsdp', None, None)
I0421 15:30:18.059768 131040467287168 maxtext_utils.py:1654]  params/params/decoder/layers_5/self_attention/out/kernel
    Shape:     float32[16,128,2048]
    Logical:   PartitionSpec('heads', 'kv', 'embed')
    Physical:  (None, None, 'fsdp')
I0421 15:30:18.059791 131040467287168 maxtext_utils.py:1654]  params/params/decoder/layers_5/self_attention/query/kernel
    Shape:     float32[2048,16,128]
    Logical:   PartitionSpec('embed', 'q_heads', 'kv')
    Physical:  ('fsdp', None, None)
I0421 15:30:18.059812 131040467287168 maxtext_utils.py:1654]  params/params/decoder/layers_5/self_attention/value/kernel
    Shape:     float32[2048,16,128]
    Logical:   PartitionSpec('embed', 'kv_heads', 'kv_head_dim')
    Physical:  ('fsdp', None, None)
I0421 15:30:18.059834 131040467287168 maxtext_utils.py:1654]  params/params/decoder/layers_6/mlp/wi_0/kernel
    Shape:     float32[2048,7168]
    Logical:   PartitionSpec('embed', 'mlp')
    Physical:  ('fsdp', None)
I0421 15:30:18.059855 131040467287168 maxtext_utils.py:1654]  params/params/decoder/layers_6/mlp/wi_1/kernel
    Shape:     float32[2048,7168]
    Logical:   PartitionSpec('embed', 'mlp')
    Physical:  ('fsdp', None)
I0421 15:30:18.059876 131040467287168 maxtext_utils.py:1654]  params/params/decoder/layers_6/mlp/wo/kernel
    Shape:     float32[7168,2048]
    Logical:   PartitionSpec('mlp', 'embed')
    Physical:  (None, 'fsdp')
I0421 15:30:18.059896 131040467287168 maxtext_utils.py:1654]  params/params/decoder/layers_6/post_self_attention_layer_norm/scale
    Shape:     float32[2048]
    Logical:   PartitionSpec('norm',)
    Physical:  (None,)
I0421 15:30:18.059916 131040467287168 maxtext_utils.py:1654]  params/params/decoder/layers_6/pre_self_attention_layer_norm/scale
    Shape:     float32[2048]
    Logical:   PartitionSpec('norm',)
    Physical:  (None,)
I0421 15:30:18.059938 131040467287168 maxtext_utils.py:1654]  params/params/decoder/layers_6/self_attention/key/kernel
    Shape:     float32[2048,16,128]
    Logical:   PartitionSpec('embed', 'kv_heads', 'kv_head_dim')
    Physical:  ('fsdp', None, None)
I0421 15:30:18.059959 131040467287168 maxtext_utils.py:1654]  params/params/decoder/layers_6/self_attention/out/kernel
    Shape:     float32[16,128,2048]
    Logical:   PartitionSpec('heads', 'kv', 'embed')
    Physical:  (None, None, 'fsdp')
I0421 15:30:18.059984 131040467287168 maxtext_utils.py:1654]  params/params/decoder/layers_6/self_attention/query/kernel
    Shape:     float32[2048,16,128]
    Logical:   PartitionSpec('embed', 'q_heads', 'kv')
    Physical:  ('fsdp', None, None)
I0421 15:30:18.060005 131040467287168 maxtext_utils.py:1654]  params/params/decoder/layers_6/self_attention/value/kernel
    Shape:     float32[2048,16,128]
    Logical:   PartitionSpec('embed', 'kv_heads', 'kv_head_dim')
    Physical:  ('fsdp', None, None)
I0421 15:30:18.060027 131040467287168 maxtext_utils.py:1654]  params/params/decoder/layers_7/mlp/wi_0/kernel
    Shape:     float32[2048,7168]
    Logical:   PartitionSpec('embed', 'mlp')
    Physical:  ('fsdp', None)
I0421 15:30:18.060058 131040467287168 maxtext_utils.py:1654]  params/params/decoder/layers_7/mlp/wi_1/kernel
    Shape:     float32[2048,7168]
    Logical:   PartitionSpec('embed', 'mlp')
    Physical:  ('fsdp', None)
I0421 15:30:18.060082 131040467287168 maxtext_utils.py:1654]  params/params/decoder/layers_7/mlp/wo/kernel
    Shape:     float32[7168,2048]
    Logical:   PartitionSpec('mlp', 'embed')
    Physical:  (None, 'fsdp')
I0421 15:30:18.060103 131040467287168 maxtext_utils.py:1654]  params/params/decoder/layers_7/post_self_attention_layer_norm/scale
    Shape:     float32[2048]
    Logical:   PartitionSpec('norm',)
    Physical:  (None,)
I0421 15:30:18.060123 131040467287168 maxtext_utils.py:1654]  params/params/decoder/layers_7/pre_self_attention_layer_norm/scale
    Shape:     float32[2048]
    Logical:   PartitionSpec('norm',)
    Physical:  (None,)
I0421 15:30:18.060145 131040467287168 maxtext_utils.py:1654]  params/params/decoder/layers_7/self_attention/key/kernel
    Shape:     float32[2048,16,128]
    Logical:   PartitionSpec('embed', 'kv_heads', 'kv_head_dim')
    Physical:  ('fsdp', None, None)
I0421 15:30:18.060168 131040467287168 maxtext_utils.py:1654]  params/params/decoder/layers_7/self_attention/out/kernel
    Shape:     float32[16,128,2048]
    Logical:   PartitionSpec('heads', 'kv', 'embed')
    Physical:  (None, None, 'fsdp')
I0421 15:30:18.060191 131040467287168 maxtext_utils.py:1654]  params/params/decoder/layers_7/self_attention/query/kernel
    Shape:     float32[2048,16,128]
    Logical:   PartitionSpec('embed', 'q_heads', 'kv')
    Physical:  ('fsdp', None, None)
I0421 15:30:18.060214 131040467287168 maxtext_utils.py:1654]  params/params/decoder/layers_7/self_attention/value/kernel
    Shape:     float32[2048,16,128]
    Logical:   PartitionSpec('embed', 'kv_heads', 'kv_head_dim')
    Physical:  ('fsdp', None, None)
I0421 15:30:18.060237 131040467287168 maxtext_utils.py:1654]  params/params/decoder/layers_8/mlp/wi_0/kernel
    Shape:     float32[2048,7168]
    Logical:   PartitionSpec('embed', 'mlp')
    Physical:  ('fsdp', None)
I0421 15:30:18.060257 131040467287168 maxtext_utils.py:1654]  params/params/decoder/layers_8/mlp/wi_1/kernel
    Shape:     float32[2048,7168]
    Logical:   PartitionSpec('embed', 'mlp')
    Physical:  ('fsdp', None)
I0421 15:30:18.060278 131040467287168 maxtext_utils.py:1654]  params/params/decoder/layers_8/mlp/wo/kernel
    Shape:     float32[7168,2048]
    Logical:   PartitionSpec('mlp', 'embed')
    Physical:  (None, 'fsdp')
I0421 15:30:18.060298 131040467287168 maxtext_utils.py:1654]  params/params/decoder/layers_8/post_self_attention_layer_norm/scale
    Shape:     float32[2048]
    Logical:   PartitionSpec('norm',)
    Physical:  (None,)
I0421 15:30:18.060319 131040467287168 maxtext_utils.py:1654]  params/params/decoder/layers_8/pre_self_attention_layer_norm/scale
    Shape:     float32[2048]
    Logical:   PartitionSpec('norm',)
    Physical:  (None,)
I0421 15:30:18.060340 131040467287168 maxtext_utils.py:1654]  params/params/decoder/layers_8/self_attention/key/kernel
    Shape:     float32[2048,16,128]
    Logical:   PartitionSpec('embed', 'kv_heads', 'kv_head_dim')
    Physical:  ('fsdp', None, None)
I0421 15:30:18.060362 131040467287168 maxtext_utils.py:1654]  params/params/decoder/layers_8/self_attention/out/kernel
    Shape:     float32[16,128,2048]
    Logical:   PartitionSpec('heads', 'kv', 'embed')
    Physical:  (None, None, 'fsdp')
I0421 15:30:18.060383 131040467287168 maxtext_utils.py:1654]  params/params/decoder/layers_8/self_attention/query/kernel
    Shape:     float32[2048,16,128]
    Logical:   PartitionSpec('embed', 'q_heads', 'kv')
    Physical:  ('fsdp', None, None)
I0421 15:30:18.060405 131040467287168 maxtext_utils.py:1654]  params/params/decoder/layers_8/self_attention/value/kernel
    Shape:     float32[2048,16,128]
    Logical:   PartitionSpec('embed', 'kv_heads', 'kv_head_dim')
    Physical:  ('fsdp', None, None)
I0421 15:30:18.060425 131040467287168 maxtext_utils.py:1654]  params/params/decoder/layers_9/mlp/wi_0/kernel
    Shape:     float32[2048,7168]
    Logical:   PartitionSpec('embed', 'mlp')
    Physical:  ('fsdp', None)
I0421 15:30:18.060446 131040467287168 maxtext_utils.py:1654]  params/params/decoder/layers_9/mlp/wi_1/kernel
    Shape:     float32[2048,7168]
    Logical:   PartitionSpec('embed', 'mlp')
    Physical:  ('fsdp', None)
I0421 15:30:18.060468 131040467287168 maxtext_utils.py:1654]  params/params/decoder/layers_9/mlp/wo/kernel
    Shape:     float32[7168,2048]
    Logical:   PartitionSpec('mlp', 'embed')
    Physical:  (None, 'fsdp')
I0421 15:30:18.060488 131040467287168 maxtext_utils.py:1654]  params/params/decoder/layers_9/post_self_attention_layer_norm/scale
    Shape:     float32[2048]
    Logical:   PartitionSpec('norm',)
    Physical:  (None,)
I0421 15:30:18.060508 131040467287168 maxtext_utils.py:1654]  params/params/decoder/layers_9/pre_self_attention_layer_norm/scale
    Shape:     float32[2048]
    Logical:   PartitionSpec('norm',)
    Physical:  (None,)
I0421 15:30:18.060528 131040467287168 maxtext_utils.py:1654]  params/params/decoder/layers_9/self_attention/key/kernel
    Shape:     float32[2048,16,128]
    Logical:   PartitionSpec('embed', 'kv_heads', 'kv_head_dim')
    Physical:  ('fsdp', None, None)
I0421 15:30:18.060550 131040467287168 maxtext_utils.py:1654]  params/params/decoder/layers_9/self_attention/out/kernel
    Shape:     float32[16,128,2048]
    Logical:   PartitionSpec('heads', 'kv', 'embed')
    Physical:  (None, None, 'fsdp')
I0421 15:30:18.060571 131040467287168 maxtext_utils.py:1654]  params/params/decoder/layers_9/self_attention/query/kernel
    Shape:     float32[2048,16,128]
    Logical:   PartitionSpec('embed', 'q_heads', 'kv')
    Physical:  ('fsdp', None, None)
I0421 15:30:18.060592 131040467287168 maxtext_utils.py:1654]  params/params/decoder/layers_9/self_attention/value/kernel
    Shape:     float32[2048,16,128]
    Logical:   PartitionSpec('embed', 'kv_heads', 'kv_head_dim')
    Physical:  ('fsdp', None, None)
I0421 15:30:18.060631 131040467287168 maxtext_utils.py:1654]  params/params/decoder/logits_dense/kernel
    Shape:     float32[2048,32000]
    Logical:   PartitionSpec('embed_vocab', 'vocab')
    Physical:  ('fsdp', None)
I0421 15:30:18.060665 131040467287168 maxtext_utils.py:1654]  params/params/token_embedder/embedding
    Shape:     float32[32000,2048]
    Logical:   PartitionSpec('vocab', 'embed_vocab')
    Physical:  (None, 'fsdp')

I0421 15:30:21.071177 131040467287168 train.py:155] train/xent Logical: float32[8,2048]............................................. ('activation_embed_and_logits_batch', 'activation_length').
I0421 15:30:21.071344 131040467287168 train.py:155] train/xent Physical: float32[8,2048]............................................. ('fsdp', None).
I0421 15:30:21.082847 131040467287168 train.py:162] train/z_loss Logical: float32[8,2048]............................................. ('activation_embed_and_logits_batch', 'activation_length').
I0421 15:30:21.082896 131040467287168 train.py:162] train/z_loss Physical: float32[8,2048]............................................. ('fsdp', None).
I0421 15:31:19.989683 131040467287168 max_utils.py:791] Total memory size: 3.3 GB, Output size: 1.5 GB, Temp size: 1.8 GB, Argument size: 1.5 GB, Host temp size: 0.0 GB.
I0421 15:31:19.990934 131040467287168 max_utils.py:194] tensorboardX not available; using no-op SummaryWriter.
I0421 15:31:19.991732 131040467287168 metric_logger.py:301] number parameters: 1.104 billion
I0421 15:32:22.840858 131040467287168 checkpointing.py:772] Waiting for step 0 to finish before checkpoint...
I0421 15:32:22.942571 131040467287168 checkpointing.py:776] Waited 0.1017451286315918 seconds for step 0 to finish before starting checkpointing.
I0421 15:32:22.943907 131040467287168 checkpoint_manager.py:1983] [process=0][thread=MainThread][wait_until_finished] No Save Finalize thread to wait for. Returning.
I0421 15:32:22.944127 131040467287168 checkpoint_manager.py:1501] [process=0] Saving checkpoint at step 0
I0421 15:32:22.944737 131040467287168 async_checkpointer.py:452] [process=0] Started async saving checkpoint to gs://wanglance-maxtext/linen_ckpt_main_20260421_144657/linen_main_20260421_144657_13_scan_layers_false/checkpoints/0.
I0421 15:32:23.031416 131040467287168 signaling_client.py:373] Using ThreadSafeKeyValueSignalingClient
I0421 15:32:23.129127 130955568219712 atomicity.py:137] Creating tmp directory gs://wanglance-maxtext/linen_ckpt_main_20260421_144657/linen_main_20260421_144657_13_scan_layers_false/checkpoints/0
I0421 15:32:23.173930 131040467287168 jax_array_handlers.py:347] Scheduling D2H of 444 prioritized jax.Array.
I0421 15:32:23.174114 131040467287168 replica_slices.py:410] Transferring arrays to host memory with options: use_replica_parallel=True, min_slice_bytes_for_replica_parallel=None, max_replicas_for_replica_parallel=None, enable_pinned_host_transfer=False
I0421 15:32:23.856161 131040467287168 base_pytree_checkpoint_handler.py:153] [process=0][thread=MainThread] Initiated "orbax.checkpoint._src.serialization.jax_array_handlers.ArrayHandler".serialize. Time taken: 0.683044s
I0421 15:32:23.856651 131040467287168 base_pytree_checkpoint_handler.py:128] [process=0] /jax/checkpoint/write/blocking_gbytes_per_sec: 14.976 GiB/s (total gbytes: 12.3 GiB) (time elapsed: 824 milliseconds) (per-host)
I0421 15:32:23.856704 131040467287168 base_pytree_checkpoint_handler.py:732] [process=0][thread=MainThread] Initiated Pytree async_save. Time taken: 0.824124s (batch_requests_ready=0.110438s, total_serialization_initiated=0.713307s, others=0.000380s)
I0421 15:32:23.856808 131040467287168 composite_checkpoint_handler.py:715] [process=0][thread=MainThread] Initiated CompositeCheckpointHandler.async_save. Time taken: 0.824836s (all_items=0.000035s, per_item={'items': '0.00003481'}, temp_paths=0.824801)
I0421 15:32:23.857592 130955505305152 async_checkpointer.py:79] [process=0][thread=async_save] Background save thread started.
I0421 15:32:23.857699 131040467287168 async_checkpointer.py:561] Finished blocking save. Time taken: 0.913526s. Continuing background save to gs://wanglance-maxtext/linen_ckpt_main_20260421_144657/linen_main_20260421_144657_13_scan_layers_false/checkpoints/0.
I0421 15:32:23.857884 131040467287168 checkpoint_manager.py:1549] [process=0][thread=MainThread][step=0] Starting CheckpointManager Save Finalize thread=save_finalize
I0421 15:32:23.858100 131040467287168 standard_logger.py:34] {'step': 0, 'event_type': 'save', 'directory': 'gs://wanglance-maxtext/linen_ckpt_main_20260421_144657/linen_main_20260421_144657_13_scan_layers_false/checkpoints', 'reached_preemption': False, 'preemption_received_at': None, 'synchronous': False, 'wait_for_prev_start_time': 1776785542.9438913, 'wait_for_prev_duration_secs': 4.863739013671875e-05, 'time_between_consecutive_saves_sec': None, 'checkpointer_blocking_start_time': 1776785542.9441519, 'checkpointer_blocking_duration_secs': 0.9136562347412109, 'get_old_steps_start_time': 1776785543.8578227, 'get_old_steps_duration_secs': 2.7179718017578125e-05, 'checkpoint_manager_blocking_start_time': 1776785542.9437745, 'checkpoint_manager_blocking_duration_secs': 0.9143021106719971}
I0421 15:32:23.858246 131040467287168 checkpointing.py:408] Started an asynchronous checkpoint save for step 0
I0421 15:32:23.858317 131040467287168 max_utils.py:750] 
Memstats: After params initialized:
I0421 15:32:23.858646 131040467287168 max_utils.py:756] 	Using (GB) 2.14 / 31.25 (6.848000%) on TPU_0(process=0,(0,0,0,0))
I0421 15:32:23.858675 131040467287168 max_utils.py:756] 	Using (GB) 2.14 / 31.25 (6.848000%) on TPU_1(process=0,(1,0,0,0))
I0421 15:32:23.858694 131040467287168 max_utils.py:756] 	Using (GB) 2.14 / 31.25 (6.848000%) on TPU_2(process=0,(0,1,0,0))
I0421 15:32:23.858711 131040467287168 max_utils.py:756] 	Using (GB) 2.14 / 31.25 (6.848000%) on TPU_3(process=0,(1,1,0,0))
I0421 15:32:23.858729 131040467287168 max_utils.py:756] 	Using (GB) 2.14 / 31.25 (6.848000%) on TPU_4(process=0,(0,2,0,0))
I0421 15:32:23.858752 131040467287168 max_utils.py:756] 	Using (GB) 2.14 / 31.25 (6.848000%) on TPU_5(process=0,(1,2,0,0))
I0421 15:32:23.858768 131040467287168 max_utils.py:756] 	Using (GB) 2.14 / 31.25 (6.848000%) on TPU_6(process=0,(0,3,0,0))
I0421 15:32:23.858783 131040467287168 max_utils.py:756] 	Using (GB) 2.14 / 31.25 (6.848000%) on TPU_7(process=0,(1,3,0,0))
I0421 15:32:23.865351 130955557733952 atomicity.py:137] Creating tmp directory gs://wanglance-maxtext/linen_ckpt_main_20260421_144657/linen_main_20260421_144657_13_scan_layers_false/checkpoints/0/items
I0421 15:32:24.012002 130955526276672 checkpoint.py:188] Wrote Metadata={'item_handlers': None, 'metrics': {}, 'performance_metrics': {}, 'init_timestamp_nsecs': 1776785543763540277, 'commit_timestamp_nsecs': None, 'custom_metadata': {}}, json={"item_handlers": null, "metrics": {}, "performance_metrics": {}, "init_timestamp_nsecs": 1776785543763540277, "commit_timestamp_nsecs": null, "custom_metadata": {}} to gs://wanglance-maxtext/linen_ckpt_main_20260421_144657/linen_main_20260421_144657_13_scan_layers_false/checkpoints/0/_CHECKPOINT_METADATA
I0421 15:32:24.012273 130955494819392 async_checkpointer.py:265] [process=0][thread=save_finalize] Waiting for background save thread=async_save.
I0421 15:32:24.394543 131040467287168 metric_logger.py:196] completed step: 0, seconds: 62.848, TFLOP/s/device: 0.216, Tokens/s/device: 32.586, total_weights: 16384, loss: 10.887, lm_loss: 10.887, perplexity: 53501.250
I0421 15:32:24.395517 131040467287168 metric_logger.py:281] To see full metrics 'tensorboard --logdir=gs://wanglance-maxtext/linen_ckpt_main_20260421_144657/linen_main_20260421_144657_13_scan_layers_false/tensorboard/'
I0421 15:32:24.508015 1524337 google_auth_provider.cc:149] Using credentials at ~/.config/gcloud/application_default_credentials.json
I0421 15:32:24.508091 1524337 google_auth_provider.cc:156] Using OAuth2 AuthProvider
I0421 15:32:24.525654 131040467287168 metric_logger.py:196] completed step: 1, seconds: 1.538, TFLOP/s/device: 8.837, Tokens/s/device: 1331.939, total_weights: 16384, loss: 10.887, lm_loss: 10.887, perplexity: 53501.250
I0421 15:32:24.593971 131040467287168 metric_logger.py:196] completed step: 2, seconds: 0.024, TFLOP/s/device: 564.953, Tokens/s/device: 85155.925, total_weights: 16384, loss: 9.787, lm_loss: 9.787, perplexity: 17806.746
I0421 15:32:24.704234 131040467287168 metric_logger.py:196] completed step: 3, seconds: 0.147, TFLOP/s/device: 92.169, Tokens/s/device: 13892.658, total_weights: 16384, loss: 8.853, lm_loss: 8.853, perplexity: 6997.778
I0421 15:32:43.379150 131040467287168 metric_logger.py:196] completed step: 4, seconds: 0.067, TFLOP/s/device: 203.446, Tokens/s/device: 30665.569, total_weights: 16384, loss: 7.982, lm_loss: 7.982, perplexity: 2927.066
I0421 15:32:43.754586 131040467287168 metric_logger.py:196] completed step: 5, seconds: 0.110, TFLOP/s/device: 123.101, Tokens/s/device: 18555.094, total_weights: 16384, loss: 7.248, lm_loss: 7.248, perplexity: 1405.074
I0421 15:32:43.952023 131040467287168 metric_logger.py:196] completed step: 6, seconds: 18.909, TFLOP/s/device: 0.719, Tokens/s/device: 108.309, total_weights: 16384, loss: 6.690, lm_loss: 6.690, perplexity: 804.397
I0421 15:32:43.960903 130955568219712 array_metadata_store.py:203] [process=0][thread=array_type_handler] Wrote 444 array_metadata.ArrayMetadata to gs://wanglance-maxtext/linen_ckpt_main_20260421_144657/linen_main_20260421_144657_13_scan_layers_false/checkpoints/0/items/array_metadatas/process_0
I0421 15:32:44.143380 131040467287168 metric_logger.py:196] completed step: 7, seconds: 0.190, TFLOP/s/device: 71.326, Tokens/s/device: 10750.995, total_weights: 16384, loss: 6.307, lm_loss: 6.307, perplexity: 548.399
I0421 15:32:44.362349 131040467287168 metric_logger.py:196] completed step: 8, seconds: 0.227, TFLOP/s/device: 59.792, Tokens/s/device: 9012.498, total_weights: 16384, loss: 6.070, lm_loss: 6.070, perplexity: 432.506
I0421 15:32:44.369690 131040467287168 checkpointing.py:772] Waiting for step 9 to finish before checkpoint...
I0421 15:32:44.390755 131040467287168 checkpointing.py:776] Waited 0.021050691604614258 seconds for step 9 to finish before starting checkpointing.
I0421 15:32:44.392928 131040467287168 checkpoint_manager.py:1994] [process=0][thread=MainThread][step=0][wait_until_finished] Waiting for Save Finalize thread (save_finalize) to complete.
I0421 15:33:21.583862 130955515790912 base_pytree_checkpoint_handler.py:1217] [process=0][thread=write_metadata_after_commits] Commit + Array metadata written. Time taken: 57.106115s (commit=56.610461s, array_metadata_write=0.495653s)
I0421 15:33:21.585648 130955505305152 base_pytree_checkpoint_handler.py:128] [process=0] /jax/checkpoint/write/gbytes_per_sec: 215.818 MiB/s (total gbytes: 12.3 GiB) (time elapsed: 58 seconds) (per-host)
I0421 15:33:21.585719 130955505305152 async_checkpointer.py:90] [process=0][thread=async_save] 3 Handler Commit operations completed. Time taken: 57.727975s.
I0421 15:33:21.832583 130955505305152 checkpoint.py:228] Read Metadata={'item_handlers': None, 'metrics': {}, 'performance_metrics': {}, 'init_timestamp_nsecs': 1776785543763540277, 'commit_timestamp_nsecs': None, 'custom_metadata': {}} from gs://wanglance-maxtext/linen_ckpt_main_20260421_144657/linen_main_20260421_144657_13_scan_layers_false/checkpoints/0/_CHECKPOINT_METADATA
I0421 15:33:22.017666 130955505305152 array_metadata_store.py:367] [process=0][thread=async_save] Skipped cross-host ArrayMetadata validation because only one process is found: process_index=0.
I0421 15:33:22.224458 130955526276672 checkpoint.py:247] Updated Metadata={'item_handlers': {'items': 'orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeCheckpointHandler'}, 'metrics': {}, 'performance_metrics': {}, 'init_timestamp_nsecs': 1776785543763540277, 'commit_timestamp_nsecs': None, 'custom_metadata': {}} to gs://wanglance-maxtext/linen_ckpt_main_20260421_144657/linen_main_20260421_144657_13_scan_layers_false/checkpoints/0/_CHECKPOINT_METADATA
I0421 15:33:22.485517 130955505305152 ocdbt_utils.py:56] Param validation support for Zarr3 will be added later (b/362328389).
I0421 15:33:22.486534 130955505305152 base_pytree_checkpoint_handler.py:1342] [process=0][thread=async_save] Pytree save finalize (merge_ocdbt + ArrayMetadata validation) completed. Time taken: 0.607257s. use_zarr3=True, enable_post_merge_validation=True, directory=gs://wanglance-maxtext/linen_ckpt_main_20260421_144657/linen_main_20260421_144657_13_scan_layers_false/checkpoints/0/items
I0421 15:33:22.487278 130955505305152 atomicity.py:608] Finalizing gs://wanglance-maxtext/linen_ckpt_main_20260421_144657/linen_main_20260421_144657_13_scan_layers_false/checkpoints/0/items
I0421 15:33:22.728272 130955505305152 atomicity.py:608] Finalizing gs://wanglance-maxtext/linen_ckpt_main_20260421_144657/linen_main_20260421_144657_13_scan_layers_false/checkpoints/0
I0421 15:33:23.456820 130955505305152 atomicity.py:794] [process=0][thread=async_save] Finished saving checkpoint (finalized tmp dir) to `gs://wanglance-maxtext/linen_ckpt_main_20260421_144657/linen_main_20260421_144657_13_scan_layers_false/checkpoints/0`.
I0421 15:33:23.457597 130955505305152 async_checkpointer.py:420] Finished async_save (blocking + background). Time taken: 60.513418s. directory=gs://wanglance-maxtext/linen_ckpt_main_20260421_144657/linen_main_20260421_144657_13_scan_layers_false/checkpoints/0
I0421 15:33:23.457668 130955505305152 async_checkpointer.py:144] [process=0][thread=async_save] Background save thread done. Time taken: 59.599926s.
I0421 15:33:23.457888 130955494819392 async_checkpointer.py:273] [process=0][thread=save_finalize] Done with waiting for background save thread=async_save.
I0421 15:33:23.458012 130955494819392 async_checkpointer.py:283] [process=0][thread=save_finalize] No errors found in background save thread=async_save.
I0421 15:33:23.458091 130955494819392 checkpoint_manager.py:2103] [process=0][thread=save_finalize][step=0] CheckpointManager Save Finalize is syncing with other hosts...
I0421 15:33:23.458136 130955494819392 checkpoint_manager.py:2112] [process=0][thread=save_finalize][step=0] CheckpointManager Save Finalize is done on all hosts.
I0421 15:33:23.458310 131040467287168 checkpoint_manager.py:2006] [process=0][thread=MainThread][step=0][wait_until_finished] Done waiting for Save Finalize thread (save_finalize) running at step=0.
W0421 15:33:23.458448 131040467287168 checkpoint_manager.py:1441] Waiting for previous save to complete took 39.065523 seconds. If this number is high, consider checkpointing less frequently.
I0421 15:33:23.459615 131040467287168 checkpoint_manager.py:1501] [process=0] Saving checkpoint at step 9
I0421 15:33:23.460008 131040467287168 async_checkpointer.py:452] [process=0] Started async saving checkpoint to gs://wanglance-maxtext/linen_ckpt_main_20260421_144657/linen_main_20260421_144657_13_scan_layers_false/checkpoints/9.
I0421 15:33:23.641389 130955494819392 atomicity.py:137] Creating tmp directory gs://wanglance-maxtext/linen_ckpt_main_20260421_144657/linen_main_20260421_144657_13_scan_layers_false/checkpoints/9
I0421 15:33:23.683608 131040467287168 jax_array_handlers.py:347] Scheduling D2H of 444 prioritized jax.Array.
I0421 15:33:23.683803 131040467287168 replica_slices.py:410] Transferring arrays to host memory with options: use_replica_parallel=True, min_slice_bytes_for_replica_parallel=None, max_replicas_for_replica_parallel=None, enable_pinned_host_transfer=False
I0421 15:33:23.963382 131040467287168 base_pytree_checkpoint_handler.py:153] [process=0][thread=MainThread] Initiated "orbax.checkpoint._src.serialization.jax_array_handlers.ArrayHandler".serialize. Time taken: 0.280624s
I0421 15:33:23.963814 131040467287168 base_pytree_checkpoint_handler.py:128] [process=0] /jax/checkpoint/write/blocking_gbytes_per_sec: 29.778 GiB/s (total gbytes: 12.3 GiB) (time elapsed: 414 milliseconds) (per-host)
I0421 15:33:23.963890 131040467287168 base_pytree_checkpoint_handler.py:732] [process=0][thread=MainThread] Initiated Pytree async_save. Time taken: 0.414546s (batch_requests_ready=0.099661s, total_serialization_initiated=0.314577s, others=0.000308s)
I0421 15:33:23.964028 131040467287168 composite_checkpoint_handler.py:715] [process=0][thread=MainThread] Initiated CompositeCheckpointHandler.async_save. Time taken: 0.415608s (all_items=0.000026s, per_item={'items': '0.00002599'}, temp_paths=0.415582)
I0421 15:33:23.964939 130955796809280 async_checkpointer.py:79] [process=0][thread=async_save] Background save thread started.
I0421 15:33:23.965125 131040467287168 async_checkpointer.py:561] Finished blocking save. Time taken: 0.505451s. Continuing background save to gs://wanglance-maxtext/linen_ckpt_main_20260421_144657/linen_main_20260421_144657_13_scan_layers_false/checkpoints/9.
I0421 15:33:23.965299 131040467287168 checkpoint_manager.py:1549] [process=0][thread=MainThread][step=9] Starting CheckpointManager Save Finalize thread=save_finalize
I0421 15:33:23.965631 130955786323520 async_checkpointer.py:265] [process=0][thread=save_finalize] Waiting for background save thread=async_save.
I0421 15:33:23.965797 131040467287168 standard_logger.py:34] {'step': 9, 'event_type': 'save', 'directory': 'gs://wanglance-maxtext/linen_ckpt_main_20260421_144657/linen_main_20260421_144657_13_scan_layers_false/checkpoints', 'reached_preemption': False, 'preemption_received_at': None, 'synchronous': False, 'wait_for_prev_start_time': 1776785564.3929007, 'wait_for_prev_duration_secs': 39.06552290916443, 'time_between_consecutive_saves_sec': None, 'checkpointer_blocking_start_time': 1776785603.4596446, 'checkpointer_blocking_duration_secs': 0.5055577754974365, 'get_old_steps_start_time': 1776785603.9652228, 'get_old_steps_duration_secs': 2.956390380859375e-05, 'checkpoint_manager_blocking_start_time': 1776785564.3928576, 'checkpoint_manager_blocking_duration_secs': 39.57289958000183}
I0421 15:33:23.966061 131040467287168 checkpointing.py:408] Started an asynchronous checkpoint save for step 9
I0421 15:33:23.966105 131040467287168 checkpoint_manager.py:1994] [process=0][thread=MainThread][step=9][wait_until_finished] Waiting for Save Finalize thread (save_finalize) to complete.
I0421 15:33:24.391885 130955396253248 atomicity.py:137] Creating tmp directory gs://wanglance-maxtext/linen_ckpt_main_20260421_144657/linen_main_20260421_144657_13_scan_layers_false/checkpoints/9/items
I0421 15:33:25.737705 130955817780800 array_metadata_store.py:203] [process=0][thread=array_type_handler] Wrote 444 array_metadata.ArrayMetadata to gs://wanglance-maxtext/linen_ckpt_main_20260421_144657/linen_main_20260421_144657_13_scan_layers_false/checkpoints/9/items/array_metadatas/process_0
I0421 15:33:52.689770 130955807295040 base_pytree_checkpoint_handler.py:1217] [process=0][thread=write_metadata_after_commits] Commit + Array metadata written. Time taken: 27.787194s (commit=27.299196s, array_metadata_write=0.487997s)
I0421 15:33:52.692386 130955796809280 base_pytree_checkpoint_handler.py:128] [process=0] /jax/checkpoint/write/gbytes_per_sec: 433.613 MiB/s (total gbytes: 12.3 GiB) (time elapsed: 29 seconds) (per-host)
I0421 15:33:52.692499 130955796809280 async_checkpointer.py:90] [process=0][thread=async_save] 3 Handler Commit operations completed. Time taken: 28.727401s.
I0421 15:33:53.117730 130955796809280 array_metadata_store.py:367] [process=0][thread=async_save] Skipped cross-host ArrayMetadata validation because only one process is found: process_index=0.
I0421 15:33:53.585815 130955796809280 ocdbt_utils.py:56] Param validation support for Zarr3 will be added later (b/362328389).
I0421 15:33:53.587880 130955796809280 base_pytree_checkpoint_handler.py:1342] [process=0][thread=async_save] Pytree save finalize (merge_ocdbt + ArrayMetadata validation) completed. Time taken: 0.619610s. use_zarr3=True, enable_post_merge_validation=True, directory=gs://wanglance-maxtext/linen_ckpt_main_20260421_144657/linen_main_20260421_144657_13_scan_layers_false/checkpoints/9/items
I0421 15:33:53.588741 130955796809280 atomicity.py:608] Finalizing gs://wanglance-maxtext/linen_ckpt_main_20260421_144657/linen_main_20260421_144657_13_scan_layers_false/checkpoints/9/items
I0421 15:33:53.842126 130955796809280 atomicity.py:608] Finalizing gs://wanglance-maxtext/linen_ckpt_main_20260421_144657/linen_main_20260421_144657_13_scan_layers_false/checkpoints/9
I0421 15:33:54.524935 130955796809280 atomicity.py:794] [process=0][thread=async_save] Finished saving checkpoint (finalized tmp dir) to `gs://wanglance-maxtext/linen_ckpt_main_20260421_144657/linen_main_20260421_144657_13_scan_layers_false/checkpoints/9`.
I0421 15:33:54.525774 130955796809280 async_checkpointer.py:420] Finished async_save (blocking + background). Time taken: 31.066111s. directory=gs://wanglance-maxtext/linen_ckpt_main_20260421_144657/linen_main_20260421_144657_13_scan_layers_false/checkpoints/9
I0421 15:33:54.525874 130955796809280 async_checkpointer.py:144] [process=0][thread=async_save] Background save thread done. Time taken: 30.560786s.
I0421 15:33:54.526145 130955786323520 async_checkpointer.py:273] [process=0][thread=save_finalize] Done with waiting for background save thread=async_save.
I0421 15:33:54.526289 130955786323520 async_checkpointer.py:283] [process=0][thread=save_finalize] No errors found in background save thread=async_save.
I0421 15:33:54.526414 130955786323520 checkpoint_manager.py:2103] [process=0][thread=save_finalize][step=9] CheckpointManager Save Finalize is syncing with other hosts...
I0421 15:33:54.526477 130955786323520 checkpoint_manager.py:2112] [process=0][thread=save_finalize][step=9] CheckpointManager Save Finalize is done on all hosts.
I0421 15:33:54.526784 131040467287168 checkpoint_manager.py:2006] [process=0][thread=MainThread][step=9][wait_until_finished] Done waiting for Save Finalize thread (save_finalize) running at step=9.
I0421 15:33:54.526967 131040467287168 checkpoint_manager.py:1983] [process=0][thread=MainThread][wait_until_finished] No Save Finalize thread to wait for. Returning.
I0421 15:33:54.528041 131040467287168 metric_logger.py:196] completed step: 9, seconds: 0.178, TFLOP/s/device: 76.530, Tokens/s/device: 11535.364, total_weights: 16384, loss: 5.930, lm_loss: 5.930, perplexity: 376.179
Per train step:
 Total TFLOPs: 13.59 
 split as 93.93% learnable weight flops and 6.07% attention flops

fix/gate-fp8-nan-sanitization · cd5a4f898 · fix_gate_fp8_nan_sanitization_20260422_174509 · full log

2026-04-22 18:29:31.903592: E external/local_xla/xla/stream_executor/cuda/cuda_fft.cc:467] Unable to register cuFFT factory: Attempting to register factory for plugin cuFFT when one has already been registered
WARNING: All log messages before absl::InitializeLog() is called are written to STDERR
E0000 00:00:1776882571.916902 2065265 cuda_dnn.cc:8579] Unable to register cuDNN factory: Attempting to register factory for plugin cuDNN when one has already been registered
E0000 00:00:1776882571.920733 2065265 cuda_blas.cc:1407] Unable to register cuBLAS factory: Attempting to register factory for plugin cuBLAS when one has already been registered
W0000 00:00:1776882571.932268 2065265 computation_placer.cc:177] computation placer already registered. Please check linkage and avoid linking the same target more than once.
W0000 00:00:1776882571.932285 2065265 computation_placer.cc:177] computation placer already registered. Please check linkage and avoid linking the same target more than once.
W0000 00:00:1776882571.932286 2065265 computation_placer.cc:177] computation placer already registered. Please check linkage and avoid linking the same target more than once.
W0000 00:00:1776882571.932288 2065265 computation_placer.cc:177] computation placer already registered. Please check linkage and avoid linking the same target more than once.
Unrecognized keys in `rope_scaling` for 'rope_type'='yarn': {'rope_theta'}
`rope_scaling`'s factor field must be a float >= 1, got 40
`rope_scaling`'s beta_fast field must be a float, got 32
`rope_scaling`'s beta_slow field must be a float, got 1
Unrecognized keys in `rope_scaling` for 'rope_type'='yarn': {'rope_theta'}
Unrecognized keys in `rope_scaling` for 'rope_type'='yarn': {'rope_theta'}
Unrecognized keys in `rope_scaling` for 'rope_type'='yarn': {'rope_theta'}
2026-04-22 18:29:35.946387: E external/local_xla/xla/stream_executor/cuda/cuda_platform.cc:51] failed call to cuInit: INTERNAL: CUDA error: Failed call to cuInit: UNKNOWN ERROR (303)
I0422 18:29:36.082752 132115681023808 max_utils.py:238] Skipping jax distributed system due to skip_jax_distributed_system=True flag.
I0422 18:30:34.156598 132115681023808 max_utils.py:800] System Information: Jax Version: 0.8.1
I0422 18:30:34.156712 132115681023808 max_utils.py:801] System Information: Jaxlib Version: 0.8.1
I0422 18:30:34.156750 132115681023808 max_utils.py:802] System Information: Jax Backend: PJRT C API
TFRT TPU v6 lite
Built on Nov 12 2025 14:16:36 (1762985796) cl/831091709
I0422 18:30:34.156776 132115681023808 train_utils.py:348] WARNING: 'dataset_path' might be pointing your local file system
I0422 18:30:34.156798 132115681023808 train_utils.py:361] WARNING: Sequence packing is essentially ignored for synthetic data. Please use a real dataset to use sequence packing.
I0422 18:30:34.156901 132115681023808 train.py:705] [DECOUPLED NO-OP] skipping cloud diagnostics wrapper.
I0422 18:30:34.868929 132115681023808 maxtext_utils.py:1604] Num_devices: 8, shape (1, 1, 1, 8, 1, 1, 1, 1, 1, 1, 1, 1, 1)
I0422 18:30:34.869224 132115681023808 checkpointing.py:677] Setting up checkpoint logger...
I0422 18:30:34.869272 132115681023808 checkpointing.py:233] Creating checkpoint manager with ocdbt=True and zarr3=True
I0422 18:30:34.869313 132115681023808 pytree_checkpoint_handler.py:577] save_device_host_concurrent_bytes=None
I0422 18:30:34.869760 132115681023808 base_pytree_checkpoint_handler.py:411] Created BasePyTreeCheckpointHandler: use_ocdbt=True, use_zarr3=True, pytree_metadata_options=PyTreeMetadataOptions(support_rich_types=False), array_metadata_store=<orbax.checkpoint._src.metadata.array_metadata_store.Store object at 0x78284d3b2fc0>, enable_pinned_host_transfer=False, save_concurrent_bytes: 96000000000 (89.4 GiB), restore_concurrent_bytes: 96000000000 (89.4 GiB)
I0422 18:30:37.229361 132115681023808 checkpointing.py:265] Enabling policy for fixed interval checkpointing.
I0422 18:30:37.229742 132115681023808 checkpoint_manager.py:702] [process=0][thread=MainThread] CheckpointManager init: checkpointers=None, item_names=('items',), item_handlers={'items': <orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeCheckpointHandler object at 0x7826ee9af0e0>}, handler_registry=None
I0422 18:30:37.229968 132115681023808 composite_checkpoint_handler.py:237] Deferred registration for item: "items". Adding handler `<orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeCheckpointHandler object at 0x7826ee9af0e0>` for item "items" and save args `<class 'orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeSaveArgs'>` and restore args `<class 'orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeRestoreArgs'>` to `_handler_registry`.
I0422 18:30:37.230008 132115681023808 composite_checkpoint_handler.py:237] Deferred registration for item: "metrics". Adding handler `<orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonCheckpointHandler object at 0x782672f34bf0>` for item "metrics" and save args `<class 'orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonSaveArgs'>` and restore args `<class 'orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonRestoreArgs'>` to `_handler_registry`.
I0422 18:30:37.230039 132115681023808 composite_checkpoint_handler.py:505] Initialized registry DefaultCheckpointHandlerRegistry({('items', <class 'orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeSaveArgs'>): <orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeCheckpointHandler object at 0x7826ee9af0e0>, ('items', <class 'orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeRestoreArgs'>): <orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeCheckpointHandler object at 0x7826ee9af0e0>, ('metrics', <class 'orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonSaveArgs'>): <orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonCheckpointHandler object at 0x782672f34bf0>, ('metrics', <class 'orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonRestoreArgs'>): <orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonCheckpointHandler object at 0x782672f34bf0>}).
I0422 18:30:37.230353 132115681023808 abstract_checkpointer.py:35] orbax-checkpoint version: 0.11.28
I0422 18:30:37.230408 132115681023808 async_checkpointer.py:177] [process=0][thread=MainThread] Using barrier_sync_fn: <function get_barrier_sync_fn.<locals>.<lambda> at 0x782672f44860> timeout: 600 secs and primary_host=0 for async checkpoint writes
I0422 18:30:37.352531 132115681023808 checkpoint_manager.py:1788] Found 0 checkpoint steps in gs://wanglance-maxtext/linen_ckpt_main_20260422_174509/linen_main_20260422_174509_13_scan_layers_false/checkpoints
I0422 18:30:37.352766 132115681023808 checkpoint_manager.py:921] [process=0][thread=MainThread] CheckpointManager created,  primary_host=0, CheckpointManagerOptions=CheckpointManagerOptions(save_interval_steps=1, max_to_keep=None, keep_time_interval=None, keep_period=None, should_keep_fn=None, best_fn=None, best_mode='max', keep_checkpoints_without_metrics=True, step_prefix=None, step_format_fixed_length=None, step_name_format=None, create=True, cleanup_tmp_directories=False, save_on_steps=frozenset(), single_host_load_and_broadcast=False, todelete_subdir=None, todelete_full_path=None, enable_hns=False, enable_background_delete=False, read_only=False, enable_async_checkpointing=True, async_options=None, multiprocessing_options=MultiprocessingOptions(primary_host=0, active_processes=None, barrier_sync_key_prefix=None), should_save_fn=None, file_options=FileOptions(path_permission_mode=None), save_root_metadata=True, temporary_path_class=None, save_decision_policy=FixedIntervalPolicy(interval=10), preservation_policy=LatestN(n=None), prevent_write_metrics=False, enable_should_save_is_saving_in_progress_check=True, enable_per_process_directory_creation=False), root_directory=gs://wanglance-maxtext/linen_ckpt_main_20260422_174509/linen_main_20260422_174509_13_scan_layers_false/checkpoints: <orbax.checkpoint.checkpoint_manager.CheckpointManager object at 0x7826ee9f6570>
I0422 18:30:37.353057 132115681023808 checkpointing.py:301] Checkpoint manager created!
I0422 18:30:37.773524 132115681023808 nnx_wrappers.py:437] Unknown Logical: bfloat16[8,2048,2048]....................................... ('activation_batch', 'activation_norm_length', 'activation_embed').
I0422 18:30:37.773747 132115681023808 nnx_wrappers.py:437] Unknown Physical: bfloat16[8,2048,2048]....................................... ('fsdp', None, None).
I0422 18:30:38.442646 132115681023808 attentions.py:1088] attentions/inputs_q Logical: bfloat16[8,2048,2048]....................................... ('activation_batch', 'activation_attn_length', 'activation_attn_embed').
I0422 18:30:38.442785 132115681023808 attentions.py:1088] attentions/inputs_q Physical: bfloat16[8,2048,2048]....................................... ('fsdp', None, None).
I0422 18:30:38.456306 132115681023808 attentions.py:1089] attentions/inputs_kv Logical: bfloat16[8,2048,2048]....................................... ('activation_batch', 'activation_attn_length', 'activation_attn_embed').
I0422 18:30:38.456381 132115681023808 attentions.py:1089] attentions/inputs_kv Physical: bfloat16[8,2048,2048]....................................... ('fsdp', None, None).
I0422 18:30:38.479240 132115681023808 attentions.py:1154] attentions/query Logical: bfloat16[8,2048,16,128]..................................... ('activation_kv_batch', 'activation_attn_length', 'activation_kv_heads', 'activation_kv_head_dim').
I0422 18:30:38.479323 132115681023808 attentions.py:1154] attentions/query Physical: bfloat16[8,2048,16,128]..................................... ('fsdp', None, None, None).
I0422 18:30:38.492077 132115681023808 attentions.py:1155] attentions/key Logical: bfloat16[8,2048,16,128]..................................... ('activation_kv_batch', 'activation_attn_length', 'activation_kv_heads', 'activation_kv_head_dim').
I0422 18:30:38.492154 132115681023808 attentions.py:1155] attentions/key Physical: bfloat16[8,2048,16,128]..................................... ('fsdp', None, None, None).
I0422 18:30:38.504689 132115681023808 attentions.py:1156] attentions/value Logical: bfloat16[8,2048,16,128]..................................... ('activation_kv_batch', 'activation_attn_length', 'activation_kv_heads', 'activation_kv_head_dim').
I0422 18:30:38.504765 132115681023808 attentions.py:1156] attentions/value Physical: bfloat16[8,2048,16,128]..................................... ('fsdp', None, None, None).
I0422 18:30:38.525119 132115681023808 attentions.py:1198] attentions/out Logical: bfloat16[8,2048,16,128]..................................... ('activation_batch', 'activation_attn_length', 'activation_heads', 'activation_kv').
I0422 18:30:38.525203 132115681023808 attentions.py:1198] attentions/out Physical: bfloat16[8,2048,16,128]..................................... ('fsdp', None, None, None).
I0422 18:30:38.543127 132115681023808 linears.py:525] linears/x Logical: bfloat16[8,2048,7168]....................................... ('activation_batch', 'activation_length', 'activation_mlp').
I0422 18:30:38.543206 132115681023808 linears.py:525] linears/x Physical: bfloat16[8,2048,7168]....................................... ('fsdp', None, None).
I0422 18:30:40.977761 132115681023808 checkpointing.py:577] checkpoint manager exists so trying to load this run's existing checkpoint
I0422 18:30:40.977903 132115681023808 checkpointing.py:665] No existing checkpoints found, not restoring checkpoint.
[DECOUPLED NO-OP] gcs_storage: using stubs.
[DECOUPLED NO-OP] mldiagnostics: using stub.
[DECOUPLED NO-OP] mldiagnostics: using stub.
[DECOUPLED NO-OP] mldiagnostics: using stub.
[DECOUPLED NO-OP] workload_monitor: using stub.
[DECOUPLED NO-OP] vertex_tensorboard: using stub.
fsdp: 8
I0422 18:30:46.531327 132115681023808 maxtext_utils.py:1707]  params/params/decoder/decoder_norm/scale
    Shape:     float32[2048]
    Logical:   PartitionSpec('norm',)
    Physical:  (None,)
I0422 18:30:46.531452 132115681023808 maxtext_utils.py:1707]  params/params/decoder/layers_0/mlp/wi_0/kernel
    Shape:     float32[2048,7168]
    Logical:   PartitionSpec('embed', 'mlp')
    Physical:  ('fsdp', None)
I0422 18:30:46.531496 132115681023808 maxtext_utils.py:1707]  params/params/decoder/layers_0/mlp/wi_1/kernel
    Shape:     float32[2048,7168]
    Logical:   PartitionSpec('embed', 'mlp')
    Physical:  ('fsdp', None)
I0422 18:30:46.531540 132115681023808 maxtext_utils.py:1707]  params/params/decoder/layers_0/mlp/wo/kernel
    Shape:     float32[7168,2048]
    Logical:   PartitionSpec('mlp', 'embed')
    Physical:  (None, 'fsdp')
I0422 18:30:46.531571 132115681023808 maxtext_utils.py:1707]  params/params/decoder/layers_0/post_self_attention_layer_norm/scale
    Shape:     float32[2048]
    Logical:   PartitionSpec('norm',)
    Physical:  (None,)
I0422 18:30:46.531603 132115681023808 maxtext_utils.py:1707]  params/params/decoder/layers_0/pre_self_attention_layer_norm/scale
    Shape:     float32[2048]
    Logical:   PartitionSpec('norm',)
    Physical:  (None,)
I0422 18:30:46.531647 132115681023808 maxtext_utils.py:1707]  params/params/decoder/layers_0/self_attention/key/kernel
    Shape:     float32[2048,16,128]
    Logical:   PartitionSpec('embed', 'kv_heads', 'kv_head_dim')
    Physical:  ('fsdp', None, None)
I0422 18:30:46.531691 132115681023808 maxtext_utils.py:1707]  params/params/decoder/layers_0/self_attention/out/kernel
    Shape:     float32[16,128,2048]
    Logical:   PartitionSpec('heads', 'kv', 'embed')
    Physical:  (None, None, 'fsdp')
I0422 18:30:46.531721 132115681023808 maxtext_utils.py:1707]  params/params/decoder/layers_0/self_attention/query/kernel
    Shape:     float32[2048,16,128]
    Logical:   PartitionSpec('embed', 'q_heads', 'kv')
    Physical:  ('fsdp', None, None)
I0422 18:30:46.531747 132115681023808 maxtext_utils.py:1707]  params/params/decoder/layers_0/self_attention/value/kernel
    Shape:     float32[2048,16,128]
    Logical:   PartitionSpec('embed', 'kv_heads', 'kv_head_dim')
    Physical:  ('fsdp', None, None)
I0422 18:30:46.531772 132115681023808 maxtext_utils.py:1707]  params/params/decoder/layers_1/mlp/wi_0/kernel
    Shape:     float32[2048,7168]
    Logical:   PartitionSpec('embed', 'mlp')
    Physical:  ('fsdp', None)
I0422 18:30:46.531799 132115681023808 maxtext_utils.py:1707]  params/params/decoder/layers_1/mlp/wi_1/kernel
    Shape:     float32[2048,7168]
    Logical:   PartitionSpec('embed', 'mlp')
    Physical:  ('fsdp', None)
I0422 18:30:46.531823 132115681023808 maxtext_utils.py:1707]  params/params/decoder/layers_1/mlp/wo/kernel
    Shape:     float32[7168,2048]
    Logical:   PartitionSpec('mlp', 'embed')
    Physical:  (None, 'fsdp')
I0422 18:30:46.531847 132115681023808 maxtext_utils.py:1707]  params/params/decoder/layers_1/post_self_attention_layer_norm/scale
    Shape:     float32[2048]
    Logical:   PartitionSpec('norm',)
    Physical:  (None,)
I0422 18:30:46.531871 132115681023808 maxtext_utils.py:1707]  params/params/decoder/layers_1/pre_self_attention_layer_norm/scale
    Shape:     float32[2048]
    Logical:   PartitionSpec('norm',)
    Physical:  (None,)
I0422 18:30:46.531895 132115681023808 maxtext_utils.py:1707]  params/params/decoder/layers_1/self_attention/key/kernel
    Shape:     float32[2048,16,128]
    Logical:   PartitionSpec('embed', 'kv_heads', 'kv_head_dim')
    Physical:  ('fsdp', None, None)
I0422 18:30:46.531919 132115681023808 maxtext_utils.py:1707]  params/params/decoder/layers_1/self_attention/out/kernel
    Shape:     float32[16,128,2048]
    Logical:   PartitionSpec('heads', 'kv', 'embed')
    Physical:  (None, None, 'fsdp')
I0422 18:30:46.531944 132115681023808 maxtext_utils.py:1707]  params/params/decoder/layers_1/self_attention/query/kernel
    Shape:     float32[2048,16,128]
    Logical:   PartitionSpec('embed', 'q_heads', 'kv')
    Physical:  ('fsdp', None, None)
I0422 18:30:46.531967 132115681023808 maxtext_utils.py:1707]  params/params/decoder/layers_1/self_attention/value/kernel
    Shape:     float32[2048,16,128]
    Logical:   PartitionSpec('embed', 'kv_heads', 'kv_head_dim')
    Physical:  ('fsdp', None, None)
I0422 18:30:46.531990 132115681023808 maxtext_utils.py:1707]  params/params/decoder/layers_10/mlp/wi_0/kernel
    Shape:     float32[2048,7168]
    Logical:   PartitionSpec('embed', 'mlp')
    Physical:  ('fsdp', None)
I0422 18:30:46.532012 132115681023808 maxtext_utils.py:1707]  params/params/decoder/layers_10/mlp/wi_1/kernel
    Shape:     float32[2048,7168]
    Logical:   PartitionSpec('embed', 'mlp')
    Physical:  ('fsdp', None)
I0422 18:30:46.532034 132115681023808 maxtext_utils.py:1707]  params/params/decoder/layers_10/mlp/wo/kernel
    Shape:     float32[7168,2048]
    Logical:   PartitionSpec('mlp', 'embed')
    Physical:  (None, 'fsdp')
I0422 18:30:46.532077 132115681023808 maxtext_utils.py:1707]  params/params/decoder/layers_10/post_self_attention_layer_norm/scale
    Shape:     float32[2048]
    Logical:   PartitionSpec('norm',)
    Physical:  (None,)
I0422 18:30:46.532099 132115681023808 maxtext_utils.py:1707]  params/params/decoder/layers_10/pre_self_attention_layer_norm/scale
    Shape:     float32[2048]
    Logical:   PartitionSpec('norm',)
    Physical:  (None,)
I0422 18:30:46.532122 132115681023808 maxtext_utils.py:1707]  params/params/decoder/layers_10/self_attention/key/kernel
    Shape:     float32[2048,16,128]
    Logical:   PartitionSpec('embed', 'kv_heads', 'kv_head_dim')
    Physical:  ('fsdp', None, None)
I0422 18:30:46.532145 132115681023808 maxtext_utils.py:1707]  params/params/decoder/layers_10/self_attention/out/kernel
    Shape:     float32[16,128,2048]
    Logical:   PartitionSpec('heads', 'kv', 'embed')
    Physical:  (None, None, 'fsdp')
I0422 18:30:46.532167 132115681023808 maxtext_utils.py:1707]  params/params/decoder/layers_10/self_attention/query/kernel
    Shape:     float32[2048,16,128]
    Logical:   PartitionSpec('embed', 'q_heads', 'kv')
    Physical:  ('fsdp', None, None)
I0422 18:30:46.532189 132115681023808 maxtext_utils.py:1707]  params/params/decoder/layers_10/self_attention/value/kernel
    Shape:     float32[2048,16,128]
    Logical:   PartitionSpec('embed', 'kv_heads', 'kv_head_dim')
    Physical:  ('fsdp', None, None)
I0422 18:30:46.532210 132115681023808 maxtext_utils.py:1707]  params/params/decoder/layers_11/mlp/wi_0/kernel
    Shape:     float32[2048,7168]
    Logical:   PartitionSpec('embed', 'mlp')
    Physical:  ('fsdp', None)
I0422 18:30:46.532232 132115681023808 maxtext_utils.py:1707]  params/params/decoder/layers_11/mlp/wi_1/kernel
    Shape:     float32[2048,7168]
    Logical:   PartitionSpec('embed', 'mlp')
    Physical:  ('fsdp', None)
I0422 18:30:46.532253 132115681023808 maxtext_utils.py:1707]  params/params/decoder/layers_11/mlp/wo/kernel
    Shape:     float32[7168,2048]
    Logical:   PartitionSpec('mlp', 'embed')
    Physical:  (None, 'fsdp')
I0422 18:30:46.532274 132115681023808 maxtext_utils.py:1707]  params/params/decoder/layers_11/post_self_attention_layer_norm/scale
    Shape:     float32[2048]
    Logical:   PartitionSpec('norm',)
    Physical:  (None,)
I0422 18:30:46.532295 132115681023808 maxtext_utils.py:1707]  params/params/decoder/layers_11/pre_self_attention_layer_norm/scale
    Shape:     float32[2048]
    Logical:   PartitionSpec('norm',)
    Physical:  (None,)
I0422 18:30:46.532317 132115681023808 maxtext_utils.py:1707]  params/params/decoder/layers_11/self_attention/key/kernel
    Shape:     float32[2048,16,128]
    Logical:   PartitionSpec('embed', 'kv_heads', 'kv_head_dim')
    Physical:  ('fsdp', None, None)
I0422 18:30:46.532340 132115681023808 maxtext_utils.py:1707]  params/params/decoder/layers_11/self_attention/out/kernel
    Shape:     float32[16,128,2048]
    Logical:   PartitionSpec('heads', 'kv', 'embed')
    Physical:  (None, None, 'fsdp')
I0422 18:30:46.532361 132115681023808 maxtext_utils.py:1707]  params/params/decoder/layers_11/self_attention/query/kernel
    Shape:     float32[2048,16,128]
    Logical:   PartitionSpec('embed', 'q_heads', 'kv')
    Physical:  ('fsdp', None, None)
I0422 18:30:46.532383 132115681023808 maxtext_utils.py:1707]  params/params/decoder/layers_11/self_attention/value/kernel
    Shape:     float32[2048,16,128]
    Logical:   PartitionSpec('embed', 'kv_heads', 'kv_head_dim')
    Physical:  ('fsdp', None, None)
I0422 18:30:46.532405 132115681023808 maxtext_utils.py:1707]  params/params/decoder/layers_12/mlp/wi_0/kernel
    Shape:     float32[2048,7168]
    Logical:   PartitionSpec('embed', 'mlp')
    Physical:  ('fsdp', None)
I0422 18:30:46.532427 132115681023808 maxtext_utils.py:1707]  params/params/decoder/layers_12/mlp/wi_1/kernel
    Shape:     float32[2048,7168]
    Logical:   PartitionSpec('embed', 'mlp')
    Physical:  ('fsdp', None)
I0422 18:30:46.532450 132115681023808 maxtext_utils.py:1707]  params/params/decoder/layers_12/mlp/wo/kernel
    Shape:     float32[7168,2048]
    Logical:   PartitionSpec('mlp', 'embed')
    Physical:  (None, 'fsdp')
I0422 18:30:46.532472 132115681023808 maxtext_utils.py:1707]  params/params/decoder/layers_12/post_self_attention_layer_norm/scale
    Shape:     float32[2048]
    Logical:   PartitionSpec('norm',)
    Physical:  (None,)
I0422 18:30:46.532492 132115681023808 maxtext_utils.py:1707]  params/params/decoder/layers_12/pre_self_attention_layer_norm/scale
    Shape:     float32[2048]
    Logical:   PartitionSpec('norm',)
    Physical:  (None,)
I0422 18:30:46.532514 132115681023808 maxtext_utils.py:1707]  params/params/decoder/layers_12/self_attention/key/kernel
    Shape:     float32[2048,16,128]
    Logical:   PartitionSpec('embed', 'kv_heads', 'kv_head_dim')
    Physical:  ('fsdp', None, None)
I0422 18:30:46.532536 132115681023808 maxtext_utils.py:1707]  params/params/decoder/layers_12/self_attention/out/kernel
    Shape:     float32[16,128,2048]
    Logical:   PartitionSpec('heads', 'kv', 'embed')
    Physical:  (None, None, 'fsdp')
I0422 18:30:46.532557 132115681023808 maxtext_utils.py:1707]  params/params/decoder/layers_12/self_attention/query/kernel
    Shape:     float32[2048,16,128]
    Logical:   PartitionSpec('embed', 'q_heads', 'kv')
    Physical:  ('fsdp', None, None)
I0422 18:30:46.532579 132115681023808 maxtext_utils.py:1707]  params/params/decoder/layers_12/self_attention/value/kernel
    Shape:     float32[2048,16,128]
    Logical:   PartitionSpec('embed', 'kv_heads', 'kv_head_dim')
    Physical:  ('fsdp', None, None)
I0422 18:30:46.532604 132115681023808 maxtext_utils.py:1707]  params/params/decoder/layers_13/mlp/wi_0/kernel
    Shape:     float32[2048,7168]
    Logical:   PartitionSpec('embed', 'mlp')
    Physical:  ('fsdp', None)
I0422 18:30:46.532626 132115681023808 maxtext_utils.py:1707]  params/params/decoder/layers_13/mlp/wi_1/kernel
    Shape:     float32[2048,7168]
    Logical:   PartitionSpec('embed', 'mlp')
    Physical:  ('fsdp', None)
I0422 18:30:46.532649 132115681023808 maxtext_utils.py:1707]  params/params/decoder/layers_13/mlp/wo/kernel
    Shape:     float32[7168,2048]
    Logical:   PartitionSpec('mlp', 'embed')
    Physical:  (None, 'fsdp')
I0422 18:30:46.532669 132115681023808 maxtext_utils.py:1707]  params/params/decoder/layers_13/post_self_attention_layer_norm/scale
    Shape:     float32[2048]
    Logical:   PartitionSpec('norm',)
    Physical:  (None,)
I0422 18:30:46.532690 132115681023808 maxtext_utils.py:1707]  params/params/decoder/layers_13/pre_self_attention_layer_norm/scale
    Shape:     float32[2048]
    Logical:   PartitionSpec('norm',)
    Physical:  (None,)
I0422 18:30:46.532712 132115681023808 maxtext_utils.py:1707]  params/params/decoder/layers_13/self_attention/key/kernel
    Shape:     float32[2048,16,128]
    Logical:   PartitionSpec('embed', 'kv_heads', 'kv_head_dim')
    Physical:  ('fsdp', None, None)
I0422 18:30:46.532734 132115681023808 maxtext_utils.py:1707]  params/params/decoder/layers_13/self_attention/out/kernel
    Shape:     float32[16,128,2048]
    Logical:   PartitionSpec('heads', 'kv', 'embed')
    Physical:  (None, None, 'fsdp')
I0422 18:30:46.532757 132115681023808 maxtext_utils.py:1707]  params/params/decoder/layers_13/self_attention/query/kernel
    Shape:     float32[2048,16,128]
    Logical:   PartitionSpec('embed', 'q_heads', 'kv')
    Physical:  ('fsdp', None, None)
I0422 18:30:46.532779 132115681023808 maxtext_utils.py:1707]  params/params/decoder/layers_13/self_attention/value/kernel
    Shape:     float32[2048,16,128]
    Logical:   PartitionSpec('embed', 'kv_heads', 'kv_head_dim')
    Physical:  ('fsdp', None, None)
I0422 18:30:46.532801 132115681023808 maxtext_utils.py:1707]  params/params/decoder/layers_14/mlp/wi_0/kernel
    Shape:     float32[2048,7168]
    Logical:   PartitionSpec('embed', 'mlp')
    Physical:  ('fsdp', None)
I0422 18:30:46.532823 132115681023808 maxtext_utils.py:1707]  params/params/decoder/layers_14/mlp/wi_1/kernel
    Shape:     float32[2048,7168]
    Logical:   PartitionSpec('embed', 'mlp')
    Physical:  ('fsdp', None)
I0422 18:30:46.532844 132115681023808 maxtext_utils.py:1707]  params/params/decoder/layers_14/mlp/wo/kernel
    Shape:     float32[7168,2048]
    Logical:   PartitionSpec('mlp', 'embed')
    Physical:  (None, 'fsdp')
I0422 18:30:46.532864 132115681023808 maxtext_utils.py:1707]  params/params/decoder/layers_14/post_self_attention_layer_norm/scale
    Shape:     float32[2048]
    Logical:   PartitionSpec('norm',)
    Physical:  (None,)
I0422 18:30:46.532884 132115681023808 maxtext_utils.py:1707]  params/params/decoder/layers_14/pre_self_attention_layer_norm/scale
    Shape:     float32[2048]
    Logical:   PartitionSpec('norm',)
    Physical:  (None,)
I0422 18:30:46.532906 132115681023808 maxtext_utils.py:1707]  params/params/decoder/layers_14/self_attention/key/kernel
    Shape:     float32[2048,16,128]
    Logical:   PartitionSpec('embed', 'kv_heads', 'kv_head_dim')
    Physical:  ('fsdp', None, None)
I0422 18:30:46.532927 132115681023808 maxtext_utils.py:1707]  params/params/decoder/layers_14/self_attention/out/kernel
    Shape:     float32[16,128,2048]
    Logical:   PartitionSpec('heads', 'kv', 'embed')
    Physical:  (None, None, 'fsdp')
I0422 18:30:46.532948 132115681023808 maxtext_utils.py:1707]  params/params/decoder/layers_14/self_attention/query/kernel
    Shape:     float32[2048,16,128]
    Logical:   PartitionSpec('embed', 'q_heads', 'kv')
    Physical:  ('fsdp', None, None)
I0422 18:30:46.532969 132115681023808 maxtext_utils.py:1707]  params/params/decoder/layers_14/self_attention/value/kernel
    Shape:     float32[2048,16,128]
    Logical:   PartitionSpec('embed', 'kv_heads', 'kv_head_dim')
    Physical:  ('fsdp', None, None)
I0422 18:30:46.532990 132115681023808 maxtext_utils.py:1707]  params/params/decoder/layers_15/mlp/wi_0/kernel
    Shape:     float32[2048,7168]
    Logical:   PartitionSpec('embed', 'mlp')
    Physical:  ('fsdp', None)
I0422 18:30:46.533011 132115681023808 maxtext_utils.py:1707]  params/params/decoder/layers_15/mlp/wi_1/kernel
    Shape:     float32[2048,7168]
    Logical:   PartitionSpec('embed', 'mlp')
    Physical:  ('fsdp', None)
I0422 18:30:46.533032 132115681023808 maxtext_utils.py:1707]  params/params/decoder/layers_15/mlp/wo/kernel
    Shape:     float32[7168,2048]
    Logical:   PartitionSpec('mlp', 'embed')
    Physical:  (None, 'fsdp')
I0422 18:30:46.533064 132115681023808 maxtext_utils.py:1707]  params/params/decoder/layers_15/post_self_attention_layer_norm/scale
    Shape:     float32[2048]
    Logical:   PartitionSpec('norm',)
    Physical:  (None,)
I0422 18:30:46.533085 132115681023808 maxtext_utils.py:1707]  params/params/decoder/layers_15/pre_self_attention_layer_norm/scale
    Shape:     float32[2048]
    Logical:   PartitionSpec('norm',)
    Physical:  (None,)
I0422 18:30:46.533106 132115681023808 maxtext_utils.py:1707]  params/params/decoder/layers_15/self_attention/key/kernel
    Shape:     float32[2048,16,128]
    Logical:   PartitionSpec('embed', 'kv_heads', 'kv_head_dim')
    Physical:  ('fsdp', None, None)
I0422 18:30:46.533127 132115681023808 maxtext_utils.py:1707]  params/params/decoder/layers_15/self_attention/out/kernel
    Shape:     float32[16,128,2048]
    Logical:   PartitionSpec('heads', 'kv', 'embed')
    Physical:  (None, None, 'fsdp')
I0422 18:30:46.533149 132115681023808 maxtext_utils.py:1707]  params/params/decoder/layers_15/self_attention/query/kernel
    Shape:     float32[2048,16,128]
    Logical:   PartitionSpec('embed', 'q_heads', 'kv')
    Physical:  ('fsdp', None, None)
I0422 18:30:46.533171 132115681023808 maxtext_utils.py:1707]  params/params/decoder/layers_15/self_attention/value/kernel
    Shape:     float32[2048,16,128]
    Logical:   PartitionSpec('embed', 'kv_heads', 'kv_head_dim')
    Physical:  ('fsdp', None, None)
I0422 18:30:46.533192 132115681023808 maxtext_utils.py:1707]  params/params/decoder/layers_2/mlp/wi_0/kernel
    Shape:     float32[2048,7168]
    Logical:   PartitionSpec('embed', 'mlp')
    Physical:  ('fsdp', None)
I0422 18:30:46.533215 132115681023808 maxtext_utils.py:1707]  params/params/decoder/layers_2/mlp/wi_1/kernel
    Shape:     float32[2048,7168]
    Logical:   PartitionSpec('embed', 'mlp')
    Physical:  ('fsdp', None)
I0422 18:30:46.533236 132115681023808 maxtext_utils.py:1707]  params/params/decoder/layers_2/mlp/wo/kernel
    Shape:     float32[7168,2048]
    Logical:   PartitionSpec('mlp', 'embed')
    Physical:  (None, 'fsdp')
I0422 18:30:46.533257 132115681023808 maxtext_utils.py:1707]  params/params/decoder/layers_2/post_self_attention_layer_norm/scale
    Shape:     float32[2048]
    Logical:   PartitionSpec('norm',)
    Physical:  (None,)
I0422 18:30:46.533278 132115681023808 maxtext_utils.py:1707]  params/params/decoder/layers_2/pre_self_attention_layer_norm/scale
    Shape:     float32[2048]
    Logical:   PartitionSpec('norm',)
    Physical:  (None,)
I0422 18:30:46.533301 132115681023808 maxtext_utils.py:1707]  params/params/decoder/layers_2/self_attention/key/kernel
    Shape:     float32[2048,16,128]
    Logical:   PartitionSpec('embed', 'kv_heads', 'kv_head_dim')
    Physical:  ('fsdp', None, None)
I0422 18:30:46.533323 132115681023808 maxtext_utils.py:1707]  params/params/decoder/layers_2/self_attention/out/kernel
    Shape:     float32[16,128,2048]
    Logical:   PartitionSpec('heads', 'kv', 'embed')
    Physical:  (None, None, 'fsdp')
I0422 18:30:46.533345 132115681023808 maxtext_utils.py:1707]  params/params/decoder/layers_2/self_attention/query/kernel
    Shape:     float32[2048,16,128]
    Logical:   PartitionSpec('embed', 'q_heads', 'kv')
    Physical:  ('fsdp', None, None)
I0422 18:30:46.533368 132115681023808 maxtext_utils.py:1707]  params/params/decoder/layers_2/self_attention/value/kernel
    Shape:     float32[2048,16,128]
    Logical:   PartitionSpec('embed', 'kv_heads', 'kv_head_dim')
    Physical:  ('fsdp', None, None)
I0422 18:30:46.533390 132115681023808 maxtext_utils.py:1707]  params/params/decoder/layers_3/mlp/wi_0/kernel
    Shape:     float32[2048,7168]
    Logical:   PartitionSpec('embed', 'mlp')
    Physical:  ('fsdp', None)
I0422 18:30:46.533411 132115681023808 maxtext_utils.py:1707]  params/params/decoder/layers_3/mlp/wi_1/kernel
    Shape:     float32[2048,7168]
    Logical:   PartitionSpec('embed', 'mlp')
    Physical:  ('fsdp', None)
I0422 18:30:46.533433 132115681023808 maxtext_utils.py:1707]  params/params/decoder/layers_3/mlp/wo/kernel
    Shape:     float32[7168,2048]
    Logical:   PartitionSpec('mlp', 'embed')
    Physical:  (None, 'fsdp')
I0422 18:30:46.533453 132115681023808 maxtext_utils.py:1707]  params/params/decoder/layers_3/post_self_attention_layer_norm/scale
    Shape:     float32[2048]
    Logical:   PartitionSpec('norm',)
    Physical:  (None,)
I0422 18:30:46.533473 132115681023808 maxtext_utils.py:1707]  params/params/decoder/layers_3/pre_self_attention_layer_norm/scale
    Shape:     float32[2048]
    Logical:   PartitionSpec('norm',)
    Physical:  (None,)
I0422 18:30:46.533496 132115681023808 maxtext_utils.py:1707]  params/params/decoder/layers_3/self_attention/key/kernel
    Shape:     float32[2048,16,128]
    Logical:   PartitionSpec('embed', 'kv_heads', 'kv_head_dim')
    Physical:  ('fsdp', None, None)
I0422 18:30:46.533518 132115681023808 maxtext_utils.py:1707]  params/params/decoder/layers_3/self_attention/out/kernel
    Shape:     float32[16,128,2048]
    Logical:   PartitionSpec('heads', 'kv', 'embed')
    Physical:  (None, None, 'fsdp')
I0422 18:30:46.533539 132115681023808 maxtext_utils.py:1707]  params/params/decoder/layers_3/self_attention/query/kernel
    Shape:     float32[2048,16,128]
    Logical:   PartitionSpec('embed', 'q_heads', 'kv')
    Physical:  ('fsdp', None, None)
I0422 18:30:46.533561 132115681023808 maxtext_utils.py:1707]  params/params/decoder/layers_3/self_attention/value/kernel
    Shape:     float32[2048,16,128]
    Logical:   PartitionSpec('embed', 'kv_heads', 'kv_head_dim')
    Physical:  ('fsdp', None, None)
I0422 18:30:46.533582 132115681023808 maxtext_utils.py:1707]  params/params/decoder/layers_4/mlp/wi_0/kernel
    Shape:     float32[2048,7168]
    Logical:   PartitionSpec('embed', 'mlp')
    Physical:  ('fsdp', None)
I0422 18:30:46.533608 132115681023808 maxtext_utils.py:1707]  params/params/decoder/layers_4/mlp/wi_1/kernel
    Shape:     float32[2048,7168]
    Logical:   PartitionSpec('embed', 'mlp')
    Physical:  ('fsdp', None)
I0422 18:30:46.533631 132115681023808 maxtext_utils.py:1707]  params/params/decoder/layers_4/mlp/wo/kernel
    Shape:     float32[7168,2048]
    Logical:   PartitionSpec('mlp', 'embed')
    Physical:  (None, 'fsdp')
I0422 18:30:46.533652 132115681023808 maxtext_utils.py:1707]  params/params/decoder/layers_4/post_self_attention_layer_norm/scale
    Shape:     float32[2048]
    Logical:   PartitionSpec('norm',)
    Physical:  (None,)
I0422 18:30:46.533672 132115681023808 maxtext_utils.py:1707]  params/params/decoder/layers_4/pre_self_attention_layer_norm/scale
    Shape:     float32[2048]
    Logical:   PartitionSpec('norm',)
    Physical:  (None,)
I0422 18:30:46.533693 132115681023808 maxtext_utils.py:1707]  params/params/decoder/layers_4/self_attention/key/kernel
    Shape:     float32[2048,16,128]
    Logical:   PartitionSpec('embed', 'kv_heads', 'kv_head_dim')
    Physical:  ('fsdp', None, None)
I0422 18:30:46.533715 132115681023808 maxtext_utils.py:1707]  params/params/decoder/layers_4/self_attention/out/kernel
    Shape:     float32[16,128,2048]
    Logical:   PartitionSpec('heads', 'kv', 'embed')
    Physical:  (None, None, 'fsdp')
I0422 18:30:46.533736 132115681023808 maxtext_utils.py:1707]  params/params/decoder/layers_4/self_attention/query/kernel
    Shape:     float32[2048,16,128]
    Logical:   PartitionSpec('embed', 'q_heads', 'kv')
    Physical:  ('fsdp', None, None)
I0422 18:30:46.533758 132115681023808 maxtext_utils.py:1707]  params/params/decoder/layers_4/self_attention/value/kernel
    Shape:     float32[2048,16,128]
    Logical:   PartitionSpec('embed', 'kv_heads', 'kv_head_dim')
    Physical:  ('fsdp', None, None)
I0422 18:30:46.533780 132115681023808 maxtext_utils.py:1707]  params/params/decoder/layers_5/mlp/wi_0/kernel
    Shape:     float32[2048,7168]
    Logical:   PartitionSpec('embed', 'mlp')
    Physical:  ('fsdp', None)
I0422 18:30:46.533802 132115681023808 maxtext_utils.py:1707]  params/params/decoder/layers_5/mlp/wi_1/kernel
    Shape:     float32[2048,7168]
    Logical:   PartitionSpec('embed', 'mlp')
    Physical:  ('fsdp', None)
I0422 18:30:46.533823 132115681023808 maxtext_utils.py:1707]  params/params/decoder/layers_5/mlp/wo/kernel
    Shape:     float32[7168,2048]
    Logical:   PartitionSpec('mlp', 'embed')
    Physical:  (None, 'fsdp')
I0422 18:30:46.533843 132115681023808 maxtext_utils.py:1707]  params/params/decoder/layers_5/post_self_attention_layer_norm/scale
    Shape:     float32[2048]
    Logical:   PartitionSpec('norm',)
    Physical:  (None,)
I0422 18:30:46.533863 132115681023808 maxtext_utils.py:1707]  params/params/decoder/layers_5/pre_self_attention_layer_norm/scale
    Shape:     float32[2048]
    Logical:   PartitionSpec('norm',)
    Physical:  (None,)
I0422 18:30:46.533884 132115681023808 maxtext_utils.py:1707]  params/params/decoder/layers_5/self_attention/key/kernel
    Shape:     float32[2048,16,128]
    Logical:   PartitionSpec('embed', 'kv_heads', 'kv_head_dim')
    Physical:  ('fsdp', None, None)
I0422 18:30:46.533905 132115681023808 maxtext_utils.py:1707]  params/params/decoder/layers_5/self_attention/out/kernel
    Shape:     float32[16,128,2048]
    Logical:   PartitionSpec('heads', 'kv', 'embed')
    Physical:  (None, None, 'fsdp')
I0422 18:30:46.533926 132115681023808 maxtext_utils.py:1707]  params/params/decoder/layers_5/self_attention/query/kernel
    Shape:     float32[2048,16,128]
    Logical:   PartitionSpec('embed', 'q_heads', 'kv')
    Physical:  ('fsdp', None, None)
I0422 18:30:46.533948 132115681023808 maxtext_utils.py:1707]  params/params/decoder/layers_5/self_attention/value/kernel
    Shape:     float32[2048,16,128]
    Logical:   PartitionSpec('embed', 'kv_heads', 'kv_head_dim')
    Physical:  ('fsdp', None, None)
I0422 18:30:46.533969 132115681023808 maxtext_utils.py:1707]  params/params/decoder/layers_6/mlp/wi_0/kernel
    Shape:     float32[2048,7168]
    Logical:   PartitionSpec('embed', 'mlp')
    Physical:  ('fsdp', None)
I0422 18:30:46.533989 132115681023808 maxtext_utils.py:1707]  params/params/decoder/layers_6/mlp/wi_1/kernel
    Shape:     float32[2048,7168]
    Logical:   PartitionSpec('embed', 'mlp')
    Physical:  ('fsdp', None)
I0422 18:30:46.534010 132115681023808 maxtext_utils.py:1707]  params/params/decoder/layers_6/mlp/wo/kernel
    Shape:     float32[7168,2048]
    Logical:   PartitionSpec('mlp', 'embed')
    Physical:  (None, 'fsdp')
I0422 18:30:46.534030 132115681023808 maxtext_utils.py:1707]  params/params/decoder/layers_6/post_self_attention_layer_norm/scale
    Shape:     float32[2048]
    Logical:   PartitionSpec('norm',)
    Physical:  (None,)
I0422 18:30:46.534062 132115681023808 maxtext_utils.py:1707]  params/params/decoder/layers_6/pre_self_attention_layer_norm/scale
    Shape:     float32[2048]
    Logical:   PartitionSpec('norm',)
    Physical:  (None,)
I0422 18:30:46.534085 132115681023808 maxtext_utils.py:1707]  params/params/decoder/layers_6/self_attention/key/kernel
    Shape:     float32[2048,16,128]
    Logical:   PartitionSpec('embed', 'kv_heads', 'kv_head_dim')
    Physical:  ('fsdp', None, None)
I0422 18:30:46.534107 132115681023808 maxtext_utils.py:1707]  params/params/decoder/layers_6/self_attention/out/kernel
    Shape:     float32[16,128,2048]
    Logical:   PartitionSpec('heads', 'kv', 'embed')
    Physical:  (None, None, 'fsdp')
I0422 18:30:46.534128 132115681023808 maxtext_utils.py:1707]  params/params/decoder/layers_6/self_attention/query/kernel
    Shape:     float32[2048,16,128]
    Logical:   PartitionSpec('embed', 'q_heads', 'kv')
    Physical:  ('fsdp', None, None)
I0422 18:30:46.534149 132115681023808 maxtext_utils.py:1707]  params/params/decoder/layers_6/self_attention/value/kernel
    Shape:     float32[2048,16,128]
    Logical:   PartitionSpec('embed', 'kv_heads', 'kv_head_dim')
    Physical:  ('fsdp', None, None)
I0422 18:30:46.534170 132115681023808 maxtext_utils.py:1707]  params/params/decoder/layers_7/mlp/wi_0/kernel
    Shape:     float32[2048,7168]
    Logical:   PartitionSpec('embed', 'mlp')
    Physical:  ('fsdp', None)
I0422 18:30:46.534192 132115681023808 maxtext_utils.py:1707]  params/params/decoder/layers_7/mlp/wi_1/kernel
    Shape:     float32[2048,7168]
    Logical:   PartitionSpec('embed', 'mlp')
    Physical:  ('fsdp', None)
I0422 18:30:46.534213 132115681023808 maxtext_utils.py:1707]  params/params/decoder/layers_7/mlp/wo/kernel
    Shape:     float32[7168,2048]
    Logical:   PartitionSpec('mlp', 'embed')
    Physical:  (None, 'fsdp')
I0422 18:30:46.534233 132115681023808 maxtext_utils.py:1707]  params/params/decoder/layers_7/post_self_attention_layer_norm/scale
    Shape:     float32[2048]
    Logical:   PartitionSpec('norm',)
    Physical:  (None,)
I0422 18:30:46.534254 132115681023808 maxtext_utils.py:1707]  params/params/decoder/layers_7/pre_self_attention_layer_norm/scale
    Shape:     float32[2048]
    Logical:   PartitionSpec('norm',)
    Physical:  (None,)
I0422 18:30:46.534275 132115681023808 maxtext_utils.py:1707]  params/params/decoder/layers_7/self_attention/key/kernel
    Shape:     float32[2048,16,128]
    Logical:   PartitionSpec('embed', 'kv_heads', 'kv_head_dim')
    Physical:  ('fsdp', None, None)
I0422 18:30:46.534297 132115681023808 maxtext_utils.py:1707]  params/params/decoder/layers_7/self_attention/out/kernel
    Shape:     float32[16,128,2048]
    Logical:   PartitionSpec('heads', 'kv', 'embed')
    Physical:  (None, None, 'fsdp')
I0422 18:30:46.534318 132115681023808 maxtext_utils.py:1707]  params/params/decoder/layers_7/self_attention/query/kernel
    Shape:     float32[2048,16,128]
    Logical:   PartitionSpec('embed', 'q_heads', 'kv')
    Physical:  ('fsdp', None, None)
I0422 18:30:46.534339 132115681023808 maxtext_utils.py:1707]  params/params/decoder/layers_7/self_attention/value/kernel
    Shape:     float32[2048,16,128]
    Logical:   PartitionSpec('embed', 'kv_heads', 'kv_head_dim')
    Physical:  ('fsdp', None, None)
I0422 18:30:46.534360 132115681023808 maxtext_utils.py:1707]  params/params/decoder/layers_8/mlp/wi_0/kernel
    Shape:     float32[2048,7168]
    Logical:   PartitionSpec('embed', 'mlp')
    Physical:  ('fsdp', None)
I0422 18:30:46.534382 132115681023808 maxtext_utils.py:1707]  params/params/decoder/layers_8/mlp/wi_1/kernel
    Shape:     float32[2048,7168]
    Logical:   PartitionSpec('embed', 'mlp')
    Physical:  ('fsdp', None)
I0422 18:30:46.534402 132115681023808 maxtext_utils.py:1707]  params/params/decoder/layers_8/mlp/wo/kernel
    Shape:     float32[7168,2048]
    Logical:   PartitionSpec('mlp', 'embed')
    Physical:  (None, 'fsdp')
I0422 18:30:46.534422 132115681023808 maxtext_utils.py:1707]  params/params/decoder/layers_8/post_self_attention_layer_norm/scale
    Shape:     float32[2048]
    Logical:   PartitionSpec('norm',)
    Physical:  (None,)
I0422 18:30:46.534442 132115681023808 maxtext_utils.py:1707]  params/params/decoder/layers_8/pre_self_attention_layer_norm/scale
    Shape:     float32[2048]
    Logical:   PartitionSpec('norm',)
    Physical:  (None,)
I0422 18:30:46.534463 132115681023808 maxtext_utils.py:1707]  params/params/decoder/layers_8/self_attention/key/kernel
    Shape:     float32[2048,16,128]
    Logical:   PartitionSpec('embed', 'kv_heads', 'kv_head_dim')
    Physical:  ('fsdp', None, None)
I0422 18:30:46.534484 132115681023808 maxtext_utils.py:1707]  params/params/decoder/layers_8/self_attention/out/kernel
    Shape:     float32[16,128,2048]
    Logical:   PartitionSpec('heads', 'kv', 'embed')
    Physical:  (None, None, 'fsdp')
I0422 18:30:46.534506 132115681023808 maxtext_utils.py:1707]  params/params/decoder/layers_8/self_attention/query/kernel
    Shape:     float32[2048,16,128]
    Logical:   PartitionSpec('embed', 'q_heads', 'kv')
    Physical:  ('fsdp', None, None)
I0422 18:30:46.534528 132115681023808 maxtext_utils.py:1707]  params/params/decoder/layers_8/self_attention/value/kernel
    Shape:     float32[2048,16,128]
    Logical:   PartitionSpec('embed', 'kv_heads', 'kv_head_dim')
    Physical:  ('fsdp', None, None)
I0422 18:30:46.534550 132115681023808 maxtext_utils.py:1707]  params/params/decoder/layers_9/mlp/wi_0/kernel
    Shape:     float32[2048,7168]
    Logical:   PartitionSpec('embed', 'mlp')
    Physical:  ('fsdp', None)
I0422 18:30:46.534572 132115681023808 maxtext_utils.py:1707]  params/params/decoder/layers_9/mlp/wi_1/kernel
    Shape:     float32[2048,7168]
    Logical:   PartitionSpec('embed', 'mlp')
    Physical:  ('fsdp', None)
I0422 18:30:46.534593 132115681023808 maxtext_utils.py:1707]  params/params/decoder/layers_9/mlp/wo/kernel
    Shape:     float32[7168,2048]
    Logical:   PartitionSpec('mlp', 'embed')
    Physical:  (None, 'fsdp')
I0422 18:30:46.534617 132115681023808 maxtext_utils.py:1707]  params/params/decoder/layers_9/post_self_attention_layer_norm/scale
    Shape:     float32[2048]
    Logical:   PartitionSpec('norm',)
    Physical:  (None,)
I0422 18:30:46.534639 132115681023808 maxtext_utils.py:1707]  params/params/decoder/layers_9/pre_self_attention_layer_norm/scale
    Shape:     float32[2048]
    Logical:   PartitionSpec('norm',)
    Physical:  (None,)
I0422 18:30:46.534660 132115681023808 maxtext_utils.py:1707]  params/params/decoder/layers_9/self_attention/key/kernel
    Shape:     float32[2048,16,128]
    Logical:   PartitionSpec('embed', 'kv_heads', 'kv_head_dim')
    Physical:  ('fsdp', None, None)
I0422 18:30:46.534682 132115681023808 maxtext_utils.py:1707]  params/params/decoder/layers_9/self_attention/out/kernel
    Shape:     float32[16,128,2048]
    Logical:   PartitionSpec('heads', 'kv', 'embed')
    Physical:  (None, None, 'fsdp')
I0422 18:30:46.534704 132115681023808 maxtext_utils.py:1707]  params/params/decoder/layers_9/self_attention/query/kernel
    Shape:     float32[2048,16,128]
    Logical:   PartitionSpec('embed', 'q_heads', 'kv')
    Physical:  ('fsdp', None, None)
I0422 18:30:46.534726 132115681023808 maxtext_utils.py:1707]  params/params/decoder/layers_9/self_attention/value/kernel
    Shape:     float32[2048,16,128]
    Logical:   PartitionSpec('embed', 'kv_heads', 'kv_head_dim')
    Physical:  ('fsdp', None, None)
I0422 18:30:46.534764 132115681023808 maxtext_utils.py:1707]  params/params/decoder/logits_dense/kernel
    Shape:     float32[2048,32000]
    Logical:   PartitionSpec('embed_vocab', 'vocab')
    Physical:  ('fsdp', None)
I0422 18:30:46.534797 132115681023808 maxtext_utils.py:1707]  params/params/token_embedder/embedding
    Shape:     float32[32000,2048]
    Logical:   PartitionSpec('vocab', 'embed_vocab')
    Physical:  (None, 'fsdp')

I0422 18:30:49.244097 132115681023808 train.py:155] train/xent Logical: float32[8,2048]............................................. ('activation_embed_and_logits_batch', 'activation_length').
I0422 18:30:49.244179 132115681023808 train.py:155] train/xent Physical: float32[8,2048]............................................. ('fsdp', None).
I0422 18:30:49.255911 132115681023808 train.py:162] train/z_loss Logical: float32[8,2048]............................................. ('activation_embed_and_logits_batch', 'activation_length').
I0422 18:30:49.255961 132115681023808 train.py:162] train/z_loss Physical: float32[8,2048]............................................. ('fsdp', None).
I0422 18:31:49.012380 132115681023808 max_utils.py:791] Total memory size: 3.3 GB, Output size: 1.5 GB, Temp size: 1.8 GB, Argument size: 1.5 GB, Host temp size: 0.0 GB.
I0422 18:31:49.017141 132115681023808 max_utils.py:194] tensorboardX not available; using no-op SummaryWriter.
I0422 18:31:49.018659 132115681023808 metric_logger.py:301] number parameters: 1.104 billion
I0422 18:32:55.269011 132115681023808 checkpointing.py:772] Waiting for step 0 to finish before checkpoint...
I0422 18:32:55.689436 132115681023808 checkpointing.py:776] Waited 0.4204249382019043 seconds for step 0 to finish before starting checkpointing.
I0422 18:32:55.695034 132115681023808 checkpoint_manager.py:1983] [process=0][thread=MainThread][wait_until_finished] No Save Finalize thread to wait for. Returning.
I0422 18:32:55.695245 132115681023808 checkpoint_manager.py:1501] [process=0] Saving checkpoint at step 0
I0422 18:32:55.695745 132115681023808 async_checkpointer.py:452] [process=0] Started async saving checkpoint to gs://wanglance-maxtext/linen_ckpt_main_20260422_174509/linen_main_20260422_174509_13_scan_layers_false/checkpoints/0.
I0422 18:32:55.781653 132115681023808 signaling_client.py:373] Using ThreadSafeKeyValueSignalingClient
I0422 18:32:55.872464 132084737771072 atomicity.py:137] Creating tmp directory gs://wanglance-maxtext/linen_ckpt_main_20260422_174509/linen_main_20260422_174509_13_scan_layers_false/checkpoints/0
I0422 18:32:55.917265 132115681023808 jax_array_handlers.py:347] Scheduling D2H of 444 prioritized jax.Array.
I0422 18:32:55.917417 132115681023808 replica_slices.py:410] Transferring arrays to host memory with options: use_replica_parallel=True, min_slice_bytes_for_replica_parallel=None, max_replicas_for_replica_parallel=None, enable_pinned_host_transfer=False
I0422 18:32:56.599892 132084727285312 atomicity.py:137] Creating tmp directory gs://wanglance-maxtext/linen_ckpt_main_20260422_174509/linen_main_20260422_174509_13_scan_layers_false/checkpoints/0/items
I0422 18:32:56.769068 132083695486528 checkpoint.py:188] Wrote Metadata={'item_handlers': None, 'metrics': {}, 'performance_metrics': {}, 'init_timestamp_nsecs': 1776882776508756564, 'commit_timestamp_nsecs': None, 'custom_metadata': {}}, json={"item_handlers": null, "metrics": {}, "performance_metrics": {}, "init_timestamp_nsecs": 1776882776508756564, "commit_timestamp_nsecs": null, "custom_metadata": {}} to gs://wanglance-maxtext/linen_ckpt_main_20260422_174509/linen_main_20260422_174509_13_scan_layers_false/checkpoints/0/_CHECKPOINT_METADATA
I0422 18:32:56.884667 132115681023808 base_pytree_checkpoint_handler.py:153] [process=0][thread=MainThread] Initiated "orbax.checkpoint._src.serialization.jax_array_handlers.ArrayHandler".serialize. Time taken: 0.968176s
I0422 18:32:56.885130 132115681023808 base_pytree_checkpoint_handler.py:128] [process=0] /jax/checkpoint/write/blocking_gbytes_per_sec: 11.199 GiB/s (total gbytes: 12.3 GiB) (time elapsed: a second) (per-host)
I0422 18:32:56.885185 132115681023808 base_pytree_checkpoint_handler.py:732] [process=0][thread=MainThread] Initiated Pytree async_save. Time taken: 1.102042s (batch_requests_ready=0.102373s, total_serialization_initiated=0.999297s, others=0.000372s)
I0422 18:32:56.885287 132115681023808 composite_checkpoint_handler.py:715] [process=0][thread=MainThread] Initiated CompositeCheckpointHandler.async_save. Time taken: 1.103134s (all_items=0.000039s, per_item={'items': '0.00003886'}, temp_paths=1.103095)
I0422 18:32:56.886169 132083496257088 async_checkpointer.py:79] [process=0][thread=async_save] Background save thread started.
I0422 18:32:56.886265 132115681023808 async_checkpointer.py:561] Finished blocking save. Time taken: 1.190977s. Continuing background save to gs://wanglance-maxtext/linen_ckpt_main_20260422_174509/linen_main_20260422_174509_13_scan_layers_false/checkpoints/0.
I0422 18:32:56.886464 132115681023808 checkpoint_manager.py:1549] [process=0][thread=MainThread][step=0] Starting CheckpointManager Save Finalize thread=save_finalize
I0422 18:32:56.886612 132084091848256 async_checkpointer.py:265] [process=0][thread=save_finalize] Waiting for background save thread=async_save.
I0422 18:32:56.886682 132115681023808 standard_logger.py:34] {'step': 0, 'event_type': 'save', 'directory': 'gs://wanglance-maxtext/linen_ckpt_main_20260422_174509/linen_main_20260422_174509_13_scan_layers_false/checkpoints', 'reached_preemption': False, 'preemption_received_at': None, 'synchronous': False, 'wait_for_prev_start_time': 1776882775.6950188, 'wait_for_prev_duration_secs': 6.0558319091796875e-05, 'time_between_consecutive_saves_sec': None, 'checkpointer_blocking_start_time': 1776882775.695268, 'checkpointer_blocking_duration_secs': 1.1911036968231201, 'get_old_steps_start_time': 1776882776.886387, 'get_old_steps_duration_secs': 4.076957702636719e-05, 'checkpoint_manager_blocking_start_time': 1776882775.6948998, 'checkpoint_manager_blocking_duration_secs': 1.191755771636963}
I0422 18:32:56.886845 132115681023808 checkpointing.py:408] Started an asynchronous checkpoint save for step 0
I0422 18:32:56.886898 132115681023808 max_utils.py:750] 
Memstats: After params initialized:
I0422 18:32:56.887161 132115681023808 max_utils.py:756] 	Using (GB) 2.14 / 31.25 (6.848000%) on TPU_0(process=0,(0,0,0,0))
I0422 18:32:56.887190 132115681023808 max_utils.py:756] 	Using (GB) 2.14 / 31.25 (6.848000%) on TPU_1(process=0,(1,0,0,0))
I0422 18:32:56.887211 132115681023808 max_utils.py:756] 	Using (GB) 2.14 / 31.25 (6.848000%) on TPU_2(process=0,(0,1,0,0))
I0422 18:32:56.887230 132115681023808 max_utils.py:756] 	Using (GB) 2.14 / 31.25 (6.848000%) on TPU_3(process=0,(1,1,0,0))
I0422 18:32:56.887248 132115681023808 max_utils.py:756] 	Using (GB) 2.14 / 31.25 (6.848000%) on TPU_4(process=0,(0,2,0,0))
I0422 18:32:56.887266 132115681023808 max_utils.py:756] 	Using (GB) 2.14 / 31.25 (6.848000%) on TPU_5(process=0,(1,2,0,0))
I0422 18:32:56.887282 132115681023808 max_utils.py:756] 	Using (GB) 2.14 / 31.25 (6.848000%) on TPU_6(process=0,(0,3,0,0))
I0422 18:32:56.887299 132115681023808 max_utils.py:756] 	Using (GB) 2.14 / 31.25 (6.848000%) on TPU_7(process=0,(1,3,0,0))
I0422 18:32:57.096687 2068762 google_auth_provider.cc:149] Using credentials at ~/.config/gcloud/application_default_credentials.json
I0422 18:32:57.096744 2068762 google_auth_provider.cc:156] Using OAuth2 AuthProvider
I0422 18:32:57.401734 132115681023808 metric_logger.py:196] completed step: 0, seconds: 66.249, TFLOP/s/device: 0.205, Tokens/s/device: 30.913, total_weights: 16384, loss: 10.887, lm_loss: 10.887, perplexity: 53501.250
I0422 18:32:57.402841 132115681023808 metric_logger.py:281] To see full metrics 'tensorboard --logdir=gs://wanglance-maxtext/linen_ckpt_main_20260422_174509/linen_main_20260422_174509_13_scan_layers_false/tensorboard/'
I0422 18:32:57.438205 132115681023808 metric_logger.py:196] completed step: 1, seconds: 2.064, TFLOP/s/device: 6.584, Tokens/s/device: 992.386, total_weights: 16384, loss: 10.887, lm_loss: 10.887, perplexity: 53501.250
I0422 18:32:57.549071 132115681023808 metric_logger.py:196] completed step: 2, seconds: 0.079, TFLOP/s/device: 172.325, Tokens/s/device: 25974.685, total_weights: 16384, loss: 9.787, lm_loss: 9.787, perplexity: 17806.746
I0422 18:32:57.659434 132115681023808 metric_logger.py:196] completed step: 3, seconds: 0.036, TFLOP/s/device: 372.679, Tokens/s/device: 56174.228, total_weights: 16384, loss: 8.853, lm_loss: 8.853, perplexity: 6997.778
I0422 18:33:19.160020 132083685000768 array_metadata_store.py:203] [process=0][thread=array_type_handler] Wrote 444 array_metadata.ArrayMetadata to gs://wanglance-maxtext/linen_ckpt_main_20260422_174509/linen_main_20260422_174509_13_scan_layers_false/checkpoints/0/items/array_metadatas/process_0
I0422 18:33:19.203075 132115681023808 metric_logger.py:196] completed step: 4, seconds: 0.112, TFLOP/s/device: 121.667, Tokens/s/device: 18338.930, total_weights: 16384, loss: 7.982, lm_loss: 7.982, perplexity: 2927.066
I0422 18:33:19.271415 132115681023808 metric_logger.py:196] completed step: 5, seconds: 0.110, TFLOP/s/device: 123.116, Tokens/s/device: 18557.448, total_weights: 16384, loss: 7.248, lm_loss: 7.248, perplexity: 1405.074
I0422 18:33:19.355189 132115681023808 metric_logger.py:196] completed step: 6, seconds: 21.576, TFLOP/s/device: 0.630, Tokens/s/device: 94.921, total_weights: 16384, loss: 6.690, lm_loss: 6.690, perplexity: 804.397
I0422 18:33:19.487012 132115681023808 metric_logger.py:196] completed step: 7, seconds: 0.040, TFLOP/s/device: 335.982, Tokens/s/device: 50642.928, total_weights: 16384, loss: 6.307, lm_loss: 6.307, perplexity: 548.399
I0422 18:33:19.588508 132115681023808 metric_logger.py:196] completed step: 8, seconds: 0.101, TFLOP/s/device: 135.033, Tokens/s/device: 20353.604, total_weights: 16384, loss: 6.070, lm_loss: 6.070, perplexity: 432.506
I0422 18:33:19.719120 132115681023808 checkpointing.py:772] Waiting for step 9 to finish before checkpoint...
I0422 18:33:19.731281 132115681023808 checkpointing.py:776] Waited 0.012179136276245117 seconds for step 9 to finish before starting checkpointing.
I0422 18:33:19.732778 132115681023808 checkpoint_manager.py:1994] [process=0][thread=MainThread][step=0][wait_until_finished] Waiting for Save Finalize thread (save_finalize) to complete.
I0422 18:34:05.981730 132084716799552 base_pytree_checkpoint_handler.py:1217] [process=0][thread=write_metadata_after_commits] Commit + Array metadata written. Time taken: 68.889852s (commit=68.377011s, array_metadata_write=0.512841s)
I0422 18:34:05.983567 132083496257088 base_pytree_checkpoint_handler.py:128] [process=0] /jax/checkpoint/write/gbytes_per_sec: 180.010 MiB/s (total gbytes: 12.3 GiB) (time elapsed: a minute) (per-host)
I0422 18:34:05.983625 132083496257088 async_checkpointer.py:90] [process=0][thread=async_save] 3 Handler Commit operations completed. Time taken: 69.097315s.
I0422 18:34:06.281087 132083496257088 checkpoint.py:228] Read Metadata={'item_handlers': None, 'metrics': {}, 'performance_metrics': {}, 'init_timestamp_nsecs': 1776882776508756564, 'commit_timestamp_nsecs': None, 'custom_metadata': {}} from gs://wanglance-maxtext/linen_ckpt_main_20260422_174509/linen_main_20260422_174509_13_scan_layers_false/checkpoints/0/_CHECKPOINT_METADATA
I0422 18:34:06.508404 132083496257088 array_metadata_store.py:367] [process=0][thread=async_save] Skipped cross-host ArrayMetadata validation because only one process is found: process_index=0.
I0422 18:34:06.694689 132083695486528 checkpoint.py:247] Updated Metadata={'item_handlers': {'items': 'orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeCheckpointHandler'}, 'metrics': {}, 'performance_metrics': {}, 'init_timestamp_nsecs': 1776882776508756564, 'commit_timestamp_nsecs': None, 'custom_metadata': {}} to gs://wanglance-maxtext/linen_ckpt_main_20260422_174509/linen_main_20260422_174509_13_scan_layers_false/checkpoints/0/_CHECKPOINT_METADATA
I0422 18:34:06.971626 132083496257088 ocdbt_utils.py:56] Param validation support for Zarr3 will be added later (b/362328389).
I0422 18:34:06.974884 132083496257088 base_pytree_checkpoint_handler.py:1342] [process=0][thread=async_save] Pytree save finalize (merge_ocdbt + ArrayMetadata validation) completed. Time taken: 0.646239s. use_zarr3=True, enable_post_merge_validation=True, directory=gs://wanglance-maxtext/linen_ckpt_main_20260422_174509/linen_main_20260422_174509_13_scan_layers_false/checkpoints/0/items
I0422 18:34:06.975589 132083496257088 atomicity.py:608] Finalizing gs://wanglance-maxtext/linen_ckpt_main_20260422_174509/linen_main_20260422_174509_13_scan_layers_false/checkpoints/0/items
I0422 18:34:07.221095 132083496257088 atomicity.py:608] Finalizing gs://wanglance-maxtext/linen_ckpt_main_20260422_174509/linen_main_20260422_174509_13_scan_layers_false/checkpoints/0
I0422 18:34:07.962502 132083496257088 atomicity.py:794] [process=0][thread=async_save] Finished saving checkpoint (finalized tmp dir) to `gs://wanglance-maxtext/linen_ckpt_main_20260422_174509/linen_main_20260422_174509_13_scan_layers_false/checkpoints/0`.
I0422 18:34:07.963181 132083496257088 async_checkpointer.py:420] Finished async_save (blocking + background). Time taken: 72.267898s. directory=gs://wanglance-maxtext/linen_ckpt_main_20260422_174509/linen_main_20260422_174509_13_scan_layers_false/checkpoints/0
I0422 18:34:07.963256 132083496257088 async_checkpointer.py:144] [process=0][thread=async_save] Background save thread done. Time taken: 71.076949s.
I0422 18:34:07.963405 132084091848256 async_checkpointer.py:273] [process=0][thread=save_finalize] Done with waiting for background save thread=async_save.
I0422 18:34:07.963461 132084091848256 async_checkpointer.py:283] [process=0][thread=save_finalize] No errors found in background save thread=async_save.
I0422 18:34:07.963522 132084091848256 checkpoint_manager.py:2103] [process=0][thread=save_finalize][step=0] CheckpointManager Save Finalize is syncing with other hosts...
I0422 18:34:07.963566 132084091848256 checkpoint_manager.py:2112] [process=0][thread=save_finalize][step=0] CheckpointManager Save Finalize is done on all hosts.
I0422 18:34:07.963776 132115681023808 checkpoint_manager.py:2006] [process=0][thread=MainThread][step=0][wait_until_finished] Done waiting for Save Finalize thread (save_finalize) running at step=0.
W0422 18:34:07.963939 132115681023808 checkpoint_manager.py:1441] Waiting for previous save to complete took 48.231184 seconds. If this number is high, consider checkpointing less frequently.
I0422 18:34:07.965526 132115681023808 checkpoint_manager.py:1501] [process=0] Saving checkpoint at step 9
I0422 18:34:07.965840 132115681023808 async_checkpointer.py:452] [process=0] Started async saving checkpoint to gs://wanglance-maxtext/linen_ckpt_main_20260422_174509/linen_main_20260422_174509_13_scan_layers_false/checkpoints/9.
I0422 18:34:08.182111 132083496257088 atomicity.py:137] Creating tmp directory gs://wanglance-maxtext/linen_ckpt_main_20260422_174509/linen_main_20260422_174509_13_scan_layers_false/checkpoints/9
I0422 18:34:08.184726 132115681023808 jax_array_handlers.py:347] Scheduling D2H of 444 prioritized jax.Array.
I0422 18:34:08.184900 132115681023808 replica_slices.py:410] Transferring arrays to host memory with options: use_replica_parallel=True, min_slice_bytes_for_replica_parallel=None, max_replicas_for_replica_parallel=None, enable_pinned_host_transfer=False
I0422 18:34:08.936688 132115681023808 base_pytree_checkpoint_handler.py:153] [process=0][thread=MainThread] Initiated "orbax.checkpoint._src.serialization.jax_array_handlers.ArrayHandler".serialize. Time taken: 0.752820s
I0422 18:34:08.937140 132115681023808 base_pytree_checkpoint_handler.py:128] [process=0] /jax/checkpoint/write/blocking_gbytes_per_sec: 13.951 GiB/s (total gbytes: 12.3 GiB) (time elapsed: 884 milliseconds) (per-host)
I0422 18:34:08.937205 132115681023808 base_pytree_checkpoint_handler.py:732] [process=0][thread=MainThread] Initiated Pytree async_save. Time taken: 0.884698s (batch_requests_ready=0.102228s, total_serialization_initiated=0.782174s, others=0.000296s)
I0422 18:34:08.937341 132115681023808 composite_checkpoint_handler.py:715] [process=0][thread=MainThread] Initiated CompositeCheckpointHandler.async_save. Time taken: 0.885297s (all_items=0.000016s, per_item={'items': '0.00001597'}, temp_paths=0.885281)
I0422 18:34:08.938953 132084091848256 async_checkpointer.py:79] [process=0][thread=async_save] Background save thread started.
I0422 18:34:08.939111 132115681023808 async_checkpointer.py:561] Finished blocking save. Time taken: 0.973536s. Continuing background save to gs://wanglance-maxtext/linen_ckpt_main_20260422_174509/linen_main_20260422_174509_13_scan_layers_false/checkpoints/9.
I0422 18:34:08.939380 132115681023808 checkpoint_manager.py:1549] [process=0][thread=MainThread][step=9] Starting CheckpointManager Save Finalize thread=save_finalize
I0422 18:34:08.939604 132115681023808 standard_logger.py:34] {'step': 9, 'event_type': 'save', 'directory': 'gs://wanglance-maxtext/linen_ckpt_main_20260422_174509/linen_main_20260422_174509_13_scan_layers_false/checkpoints', 'reached_preemption': False, 'preemption_received_at': None, 'synchronous': False, 'wait_for_prev_start_time': 1776882799.732718, 'wait_for_prev_duration_secs': 48.231183767318726, 'time_between_consecutive_saves_sec': None, 'checkpointer_blocking_start_time': 1776882847.9655497, 'checkpointer_blocking_duration_secs': 0.9737017154693604, 'get_old_steps_start_time': 1776882848.9392755, 'get_old_steps_duration_secs': 5.4836273193359375e-05, 'checkpoint_manager_blocking_start_time': 1776882799.7326658, 'checkpoint_manager_blocking_duration_secs': 49.20690393447876}
I0422 18:34:08.939811 132115681023808 checkpointing.py:408] Started an asynchronous checkpoint save for step 9
I0422 18:34:08.939854 132115681023808 checkpoint_manager.py:1994] [process=0][thread=MainThread][step=9][wait_until_finished] Waiting for Save Finalize thread (save_finalize) to complete.
I0422 18:34:09.008420 132084716799552 atomicity.py:137] Creating tmp directory gs://wanglance-maxtext/linen_ckpt_main_20260422_174509/linen_main_20260422_174509_13_scan_layers_false/checkpoints/9/items
I0422 18:34:09.169915 132084687439424 async_checkpointer.py:265] [process=0][thread=save_finalize] Waiting for background save thread=async_save.
I0422 18:34:11.141338 132084706313792 array_metadata_store.py:203] [process=0][thread=array_type_handler] Wrote 444 array_metadata.ArrayMetadata to gs://wanglance-maxtext/linen_ckpt_main_20260422_174509/linen_main_20260422_174509_13_scan_layers_false/checkpoints/9/items/array_metadatas/process_0
I0422 18:34:58.676969 132083685000768 base_pytree_checkpoint_handler.py:1217] [process=0][thread=write_metadata_after_commits] Commit + Array metadata written. Time taken: 49.170133s (commit=48.640693s, array_metadata_write=0.529440s)
I0422 18:34:58.679527 132084091848256 base_pytree_checkpoint_handler.py:128] [process=0] /jax/checkpoint/write/gbytes_per_sec: 249.606 MiB/s (total gbytes: 12.3 GiB) (time elapsed: 50 seconds) (per-host)
I0422 18:34:58.679601 132084091848256 async_checkpointer.py:90] [process=0][thread=async_save] 3 Handler Commit operations completed. Time taken: 49.740432s.
I0422 18:34:59.250947 132084091848256 array_metadata_store.py:367] [process=0][thread=async_save] Skipped cross-host ArrayMetadata validation because only one process is found: process_index=0.
I0422 18:34:59.720249 132084091848256 ocdbt_utils.py:56] Param validation support for Zarr3 will be added later (b/362328389).
I0422 18:34:59.736734 132084091848256 base_pytree_checkpoint_handler.py:1342] [process=0][thread=async_save] Pytree save finalize (merge_ocdbt + ArrayMetadata validation) completed. Time taken: 0.690778s. use_zarr3=True, enable_post_merge_validation=True, directory=gs://wanglance-maxtext/linen_ckpt_main_20260422_174509/linen_main_20260422_174509_13_scan_layers_false/checkpoints/9/items
I0422 18:34:59.737704 132084091848256 atomicity.py:608] Finalizing gs://wanglance-maxtext/linen_ckpt_main_20260422_174509/linen_main_20260422_174509_13_scan_layers_false/checkpoints/9/items
I0422 18:35:00.004837 132084091848256 atomicity.py:608] Finalizing gs://wanglance-maxtext/linen_ckpt_main_20260422_174509/linen_main_20260422_174509_13_scan_layers_false/checkpoints/9
I0422 18:35:00.746577 132084091848256 atomicity.py:794] [process=0][thread=async_save] Finished saving checkpoint (finalized tmp dir) to `gs://wanglance-maxtext/linen_ckpt_main_20260422_174509/linen_main_20260422_174509_13_scan_layers_false/checkpoints/9`.
I0422 18:35:00.747480 132084091848256 async_checkpointer.py:420] Finished async_save (blocking + background). Time taken: 52.781916s. directory=gs://wanglance-maxtext/linen_ckpt_main_20260422_174509/linen_main_20260422_174509_13_scan_layers_false/checkpoints/9
I0422 18:35:00.747579 132084091848256 async_checkpointer.py:144] [process=0][thread=async_save] Background save thread done. Time taken: 51.808412s.
I0422 18:35:00.747745 132084687439424 async_checkpointer.py:273] [process=0][thread=save_finalize] Done with waiting for background save thread=async_save.
I0422 18:35:00.747807 132084687439424 async_checkpointer.py:283] [process=0][thread=save_finalize] No errors found in background save thread=async_save.
I0422 18:35:00.747894 132084687439424 checkpoint_manager.py:2103] [process=0][thread=save_finalize][step=9] CheckpointManager Save Finalize is syncing with other hosts...
I0422 18:35:00.747945 132084687439424 checkpoint_manager.py:2112] [process=0][thread=save_finalize][step=9] CheckpointManager Save Finalize is done on all hosts.
I0422 18:35:00.755571 132115681023808 checkpoint_manager.py:2006] [process=0][thread=MainThread][step=9][wait_until_finished] Done waiting for Save Finalize thread (save_finalize) running at step=9.
I0422 18:35:00.755751 132115681023808 checkpoint_manager.py:1983] [process=0][thread=MainThread][wait_until_finished] No Save Finalize thread to wait for. Returning.
I0422 18:35:00.756852 132115681023808 metric_logger.py:196] completed step: 9, seconds: 0.175, TFLOP/s/device: 77.581, Tokens/s/device: 11693.903, total_weights: 16384, loss: 5.930, lm_loss: 5.930, perplexity: 376.179
Per train step:
 Total TFLOPs: 13.59 
 split as 93.93% learnable weight flops and 6.07% attention flops

Metrics: main (b341212ce) vs fix/gate-fp8-nan-sanitization (cd5a4f898)

Metrics: `main (b341212ce)` vs `fix/gate-fp8-nan-sanitization (cd5a4f898)`