MaxView

XPK Start: Mon Apr 20 18:10:23 UTC 2026
PyTorch was not found. Models won't be available and only tokenizers, configuration and file/data utilities can be used.
2026-04-20 18:10:47.839944: 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)
I0420 18:10:48.016764 133136767887168 max_utils.py:273] Attempting to initialize the jax distributed system...
I0420 18:10:57.058392 133136767887168 distributed.py:149] Starting JAX distributed service on [::]:8482
I0420 18:10:57.060806 133136767887168 distributed.py:172] Connecting to JAX distributed service on mt-13-scan-layers-false-9cz8d-slice-job-0-0.mt-13-scan-layers-false-9cz8d:8482
I0420 18:10:58.270118 133136767887168 max_utils.py:284] Jax distributed system initialized!
I0420 18:11:04.683855 133136767887168 max_utils.py:800] System Information: Jax Version: 0.9.2
I0420 18:11:04.683957 133136767887168 max_utils.py:801] System Information: Jaxlib Version: 0.9.2
I0420 18:11:04.683996 133136767887168 max_utils.py:802] System Information: Jax Backend: PJRT C API
TFRT TPU v6 lite
Built on Mar 4 2026 11:32:08 (1772652728) cl/878335365
I0420 18:11:04.684030 133136767887168 train_utils.py:378] WARNING: Sequence packing is essentially ignored for synthetic data. Please use a real dataset to use sequence packing.
I0420 18:11:05.738292 133136767887168 maxtext_utils.py:1718] Num_devices: 32, shape (1, 1, 1, 32, 1, 1, 1, 1, 1, 1, 1, 1, 1)
I0420 18:11:06.583022 133136767887168 checkpointing.py:688] Setting up checkpoint logger...
I0420 18:11:06.583152 133136767887168 checkpointing.py:234] Creating checkpoint manager with ocdbt=True and zarr3=True
I0420 18:11:06.583197 133136767887168 pytree_checkpoint_handler.py:592] save_device_host_concurrent_bytes=None
I0420 18:11:06.583427 133136767887168 base_pytree_checkpoint_handler.py:441] 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 0x7915bafea6c0>, enable_pinned_host_transfer=False, save_concurrent_bytes: 96000000000 (89.4 GiB), restore_concurrent_bytes: 96000000000 (89.4 GiB)
I0420 18:11:09.239200 133136767887168 checkpointing.py:266] Enabling policy for fixed interval checkpointing.
I0420 18:11:09.239435 133136767887168 checkpoint_manager.py:708] [process=0][thread=MainThread] CheckpointManager init: checkpointers=None, item_names=('items',), item_handlers={'items': <orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeCheckpointHandler object at 0x79148e5c1f70>}, handler_registry=None
I0420 18:11:09.239687 133136767887168 composite_checkpoint_handler.py:237] Deferred registration for item: "items". Adding handler `<orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeCheckpointHandler object at 0x79148e5c1f70>` 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`.
I0420 18:11:09.239734 133136767887168 composite_checkpoint_handler.py:237] Deferred registration for item: "metrics". Adding handler `<orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonCheckpointHandler object at 0x7901907972f0>` 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`.
I0420 18:11:09.239769 133136767887168 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 0x79148e5c1f70>, ('items', <class 'orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeRestoreArgs'>): <orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeCheckpointHandler object at 0x79148e5c1f70>, ('metrics', <class 'orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonSaveArgs'>): <orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonCheckpointHandler object at 0x7901907972f0>, ('metrics', <class 'orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonRestoreArgs'>): <orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonCheckpointHandler object at 0x7901907972f0>}).
I0420 18:11:09.240097 133136767887168 abstract_checkpointer.py:35] orbax-checkpoint version: 0.11.34
I0420 18:11:09.240173 133136767887168 async_checkpointer.py:192] [process=0][thread=MainThread] Using barrier_sync_fn: <function get_barrier_sync_fn.<locals>._fn at 0x7901905d4d60> timeout: 1200 secs and primary_host=0 for async checkpoint writes
I0420 18:11:10.737766 133136767887168 checkpoint_manager.py:1812] Found 0 checkpoint steps in gs://lance-maxtext/nnx_ckpt_xpk_feat_nnx_post_train_fixes_20260420_153543/nnx_xpk_feat_nnx_post_train_fixes_20260420_153543_13_scan_layers_false/checkpoints
I0420 18:11:11.214000 133136767887168 checkpoint_manager.py:929] [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_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, lightweight_initialize=False), root_directory=gs://lance-maxtext/nnx_ckpt_xpk_feat_nnx_post_train_fixes_20260420_153543/nnx_xpk_feat_nnx_post_train_fixes_20260420_153543_13_scan_layers_false/checkpoints: <orbax.checkpoint.checkpoint_manager.CheckpointManager object at 0x7901b0493d10>
I0420 18:11:11.214206 133136767887168 checkpointing.py:302] Checkpoint manager created!
I0420 18:11:13.324270 133136767887168 checkpointing.py:578] checkpoint manager exists so trying to load this run's existing checkpoint
I0420 18:11:13.324398 133136767887168 checkpointing.py:676] No existing checkpoints found, not restoring checkpoint.
fsdp: 32
I0420 18:11:21.221192 133136767887168 maxtext_utils.py:1836]  decoder/decoder_norm/scale/value
    Shape:     float32[2048]
    Physical:  (None,)
I0420 18:11:21.221310 133136767887168 maxtext_utils.py:1836]  decoder/layers/0/mlp/wi_0/kernel/value
    Shape:     float32[2048,7168]
    Physical:  ('fsdp', None)
I0420 18:11:21.221361 133136767887168 maxtext_utils.py:1836]  decoder/layers/0/mlp/wi_1/kernel/value
    Shape:     float32[2048,7168]
    Physical:  ('fsdp', None)
I0420 18:11:21.221402 133136767887168 maxtext_utils.py:1836]  decoder/layers/0/mlp/wo/kernel/value
    Shape:     float32[7168,2048]
    Physical:  (None, 'fsdp')
I0420 18:11:21.221439 133136767887168 maxtext_utils.py:1836]  decoder/layers/0/post_self_attention_layer_norm/scale/value
    Shape:     float32[2048]
    Physical:  (None,)
I0420 18:11:21.221473 133136767887168 maxtext_utils.py:1836]  decoder/layers/0/pre_self_attention_layer_norm/scale/value
    Shape:     float32[2048]
    Physical:  (None,)
I0420 18:11:21.221508 133136767887168 maxtext_utils.py:1836]  decoder/layers/0/self_attention/key/kernel/value
    Shape:     float32[2048,16,128]
    Physical:  ('fsdp', None, None)
I0420 18:11:21.221542 133136767887168 maxtext_utils.py:1836]  decoder/layers/0/self_attention/out/kernel/value
    Shape:     float32[16,128,2048]
    Physical:  (None, None, 'fsdp')
I0420 18:11:21.221575 133136767887168 maxtext_utils.py:1836]  decoder/layers/0/self_attention/query/kernel/value
    Shape:     float32[2048,16,128]
    Physical:  ('fsdp', None, None)
I0420 18:11:21.221606 133136767887168 maxtext_utils.py:1836]  decoder/layers/0/self_attention/value/kernel/value
    Shape:     float32[2048,16,128]
    Physical:  ('fsdp', None, None)
I0420 18:11:21.221643 133136767887168 maxtext_utils.py:1836]  decoder/layers/1/mlp/wi_0/kernel/value
    Shape:     float32[2048,7168]
    Physical:  ('fsdp', None)
I0420 18:11:21.221675 133136767887168 maxtext_utils.py:1836]  decoder/layers/1/mlp/wi_1/kernel/value
    Shape:     float32[2048,7168]
    Physical:  ('fsdp', None)
I0420 18:11:21.221705 133136767887168 maxtext_utils.py:1836]  decoder/layers/1/mlp/wo/kernel/value
    Shape:     float32[7168,2048]
    Physical:  (None, 'fsdp')
I0420 18:11:21.221734 133136767887168 maxtext_utils.py:1836]  decoder/layers/1/post_self_attention_layer_norm/scale/value
    Shape:     float32[2048]
    Physical:  (None,)
I0420 18:11:21.221764 133136767887168 maxtext_utils.py:1836]  decoder/layers/1/pre_self_attention_layer_norm/scale/value
    Shape:     float32[2048]
    Physical:  (None,)
I0420 18:11:21.221794 133136767887168 maxtext_utils.py:1836]  decoder/layers/1/self_attention/key/kernel/value
    Shape:     float32[2048,16,128]
    Physical:  ('fsdp', None, None)
I0420 18:11:21.221824 133136767887168 maxtext_utils.py:1836]  decoder/layers/1/self_attention/out/kernel/value
    Shape:     float32[16,128,2048]
    Physical:  (None, None, 'fsdp')
I0420 18:11:21.221853 133136767887168 maxtext_utils.py:1836]  decoder/layers/1/self_attention/query/kernel/value
    Shape:     float32[2048,16,128]
    Physical:  ('fsdp', None, None)
I0420 18:11:21.221884 133136767887168 maxtext_utils.py:1836]  decoder/layers/1/self_attention/value/kernel/value
    Shape:     float32[2048,16,128]
    Physical:  ('fsdp', None, None)
I0420 18:11:21.221914 133136767887168 maxtext_utils.py:1836]  decoder/layers/2/mlp/wi_0/kernel/value
    Shape:     float32[2048,7168]
    Physical:  ('fsdp', None)
I0420 18:11:21.221943 133136767887168 maxtext_utils.py:1836]  decoder/layers/2/mlp/wi_1/kernel/value
    Shape:     float32[2048,7168]
    Physical:  ('fsdp', None)
I0420 18:11:21.221973 133136767887168 maxtext_utils.py:1836]  decoder/layers/2/mlp/wo/kernel/value
    Shape:     float32[7168,2048]
    Physical:  (None, 'fsdp')
I0420 18:11:21.222001 133136767887168 maxtext_utils.py:1836]  decoder/layers/2/post_self_attention_layer_norm/scale/value
    Shape:     float32[2048]
    Physical:  (None,)
I0420 18:11:21.222028 133136767887168 maxtext_utils.py:1836]  decoder/layers/2/pre_self_attention_layer_norm/scale/value
    Shape:     float32[2048]
    Physical:  (None,)
I0420 18:11:21.222057 133136767887168 maxtext_utils.py:1836]  decoder/layers/2/self_attention/key/kernel/value
    Shape:     float32[2048,16,128]
    Physical:  ('fsdp', None, None)
I0420 18:11:21.222098 133136767887168 maxtext_utils.py:1836]  decoder/layers/2/self_attention/out/kernel/value
    Shape:     float32[16,128,2048]
    Physical:  (None, None, 'fsdp')
I0420 18:11:21.222127 133136767887168 maxtext_utils.py:1836]  decoder/layers/2/self_attention/query/kernel/value
    Shape:     float32[2048,16,128]
    Physical:  ('fsdp', None, None)
I0420 18:11:21.222158 133136767887168 maxtext_utils.py:1836]  decoder/layers/2/self_attention/value/kernel/value
    Shape:     float32[2048,16,128]
    Physical:  ('fsdp', None, None)
I0420 18:11:21.222187 133136767887168 maxtext_utils.py:1836]  decoder/layers/3/mlp/wi_0/kernel/value
    Shape:     float32[2048,7168]
    Physical:  ('fsdp', None)
I0420 18:11:21.222216 133136767887168 maxtext_utils.py:1836]  decoder/layers/3/mlp/wi_1/kernel/value
    Shape:     float32[2048,7168]
    Physical:  ('fsdp', None)
I0420 18:11:21.222245 133136767887168 maxtext_utils.py:1836]  decoder/layers/3/mlp/wo/kernel/value
    Shape:     float32[7168,2048]
    Physical:  (None, 'fsdp')
I0420 18:11:21.222273 133136767887168 maxtext_utils.py:1836]  decoder/layers/3/post_self_attention_layer_norm/scale/value
    Shape:     float32[2048]
    Physical:  (None,)
I0420 18:11:21.222301 133136767887168 maxtext_utils.py:1836]  decoder/layers/3/pre_self_attention_layer_norm/scale/value
    Shape:     float32[2048]
    Physical:  (None,)
I0420 18:11:21.222330 133136767887168 maxtext_utils.py:1836]  decoder/layers/3/self_attention/key/kernel/value
    Shape:     float32[2048,16,128]
    Physical:  ('fsdp', None, None)
I0420 18:11:21.222359 133136767887168 maxtext_utils.py:1836]  decoder/layers/3/self_attention/out/kernel/value
    Shape:     float32[16,128,2048]
    Physical:  (None, None, 'fsdp')
I0420 18:11:21.222388 133136767887168 maxtext_utils.py:1836]  decoder/layers/3/self_attention/query/kernel/value
    Shape:     float32[2048,16,128]
    Physical:  ('fsdp', None, None)
I0420 18:11:21.222416 133136767887168 maxtext_utils.py:1836]  decoder/layers/3/self_attention/value/kernel/value
    Shape:     float32[2048,16,128]
    Physical:  ('fsdp', None, None)
I0420 18:11:21.222444 133136767887168 maxtext_utils.py:1836]  decoder/layers/4/mlp/wi_0/kernel/value
    Shape:     float32[2048,7168]
    Physical:  ('fsdp', None)
I0420 18:11:21.222472 133136767887168 maxtext_utils.py:1836]  decoder/layers/4/mlp/wi_1/kernel/value
    Shape:     float32[2048,7168]
    Physical:  ('fsdp', None)
I0420 18:11:21.222500 133136767887168 maxtext_utils.py:1836]  decoder/layers/4/mlp/wo/kernel/value
    Shape:     float32[7168,2048]
    Physical:  (None, 'fsdp')
I0420 18:11:21.222525 133136767887168 maxtext_utils.py:1836]  decoder/layers/4/post_self_attention_layer_norm/scale/value
    Shape:     float32[2048]
    Physical:  (None,)
I0420 18:11:21.222552 133136767887168 maxtext_utils.py:1836]  decoder/layers/4/pre_self_attention_layer_norm/scale/value
    Shape:     float32[2048]
    Physical:  (None,)
I0420 18:11:21.222580 133136767887168 maxtext_utils.py:1836]  decoder/layers/4/self_attention/key/kernel/value
    Shape:     float32[2048,16,128]
    Physical:  ('fsdp', None, None)
I0420 18:11:21.222608 133136767887168 maxtext_utils.py:1836]  decoder/layers/4/self_attention/out/kernel/value
    Shape:     float32[16,128,2048]
    Physical:  (None, None, 'fsdp')
I0420 18:11:21.222648 133136767887168 maxtext_utils.py:1836]  decoder/layers/4/self_attention/query/kernel/value
    Shape:     float32[2048,16,128]
    Physical:  ('fsdp', None, None)
I0420 18:11:21.222677 133136767887168 maxtext_utils.py:1836]  decoder/layers/4/self_attention/value/kernel/value
    Shape:     float32[2048,16,128]
    Physical:  ('fsdp', None, None)
I0420 18:11:21.222705 133136767887168 maxtext_utils.py:1836]  decoder/layers/5/mlp/wi_0/kernel/value
    Shape:     float32[2048,7168]
    Physical:  ('fsdp', None)
I0420 18:11:21.222732 133136767887168 maxtext_utils.py:1836]  decoder/layers/5/mlp/wi_1/kernel/value
    Shape:     float32[2048,7168]
    Physical:  ('fsdp', None)
I0420 18:11:21.222760 133136767887168 maxtext_utils.py:1836]  decoder/layers/5/mlp/wo/kernel/value
    Shape:     float32[7168,2048]
    Physical:  (None, 'fsdp')
I0420 18:11:21.222787 133136767887168 maxtext_utils.py:1836]  decoder/layers/5/post_self_attention_layer_norm/scale/value
    Shape:     float32[2048]
    Physical:  (None,)
I0420 18:11:21.222814 133136767887168 maxtext_utils.py:1836]  decoder/layers/5/pre_self_attention_layer_norm/scale/value
    Shape:     float32[2048]
    Physical:  (None,)
I0420 18:11:21.222843 133136767887168 maxtext_utils.py:1836]  decoder/layers/5/self_attention/key/kernel/value
    Shape:     float32[2048,16,128]
    Physical:  ('fsdp', None, None)
I0420 18:11:21.222871 133136767887168 maxtext_utils.py:1836]  decoder/layers/5/self_attention/out/kernel/value
    Shape:     float32[16,128,2048]
    Physical:  (None, None, 'fsdp')
I0420 18:11:21.222899 133136767887168 maxtext_utils.py:1836]  decoder/layers/5/self_attention/query/kernel/value
    Shape:     float32[2048,16,128]
    Physical:  ('fsdp', None, None)
I0420 18:11:21.222927 133136767887168 maxtext_utils.py:1836]  decoder/layers/5/self_attention/value/kernel/value
    Shape:     float32[2048,16,128]
    Physical:  ('fsdp', None, None)
I0420 18:11:21.222954 133136767887168 maxtext_utils.py:1836]  decoder/layers/6/mlp/wi_0/kernel/value
    Shape:     float32[2048,7168]
    Physical:  ('fsdp', None)
I0420 18:11:21.222982 133136767887168 maxtext_utils.py:1836]  decoder/layers/6/mlp/wi_1/kernel/value
    Shape:     float32[2048,7168]
    Physical:  ('fsdp', None)
I0420 18:11:21.223010 133136767887168 maxtext_utils.py:1836]  decoder/layers/6/mlp/wo/kernel/value
    Shape:     float32[7168,2048]
    Physical:  (None, 'fsdp')
I0420 18:11:21.223037 133136767887168 maxtext_utils.py:1836]  decoder/layers/6/post_self_attention_layer_norm/scale/value
    Shape:     float32[2048]
    Physical:  (None,)
I0420 18:11:21.223063 133136767887168 maxtext_utils.py:1836]  decoder/layers/6/pre_self_attention_layer_norm/scale/value
    Shape:     float32[2048]
    Physical:  (None,)
I0420 18:11:21.223123 133136767887168 maxtext_utils.py:1836]  decoder/layers/6/self_attention/key/kernel/value
    Shape:     float32[2048,16,128]
    Physical:  ('fsdp', None, None)
I0420 18:11:21.223157 133136767887168 maxtext_utils.py:1836]  decoder/layers/6/self_attention/out/kernel/value
    Shape:     float32[16,128,2048]
    Physical:  (None, None, 'fsdp')
I0420 18:11:21.223187 133136767887168 maxtext_utils.py:1836]  decoder/layers/6/self_attention/query/kernel/value
    Shape:     float32[2048,16,128]
    Physical:  ('fsdp', None, None)
I0420 18:11:21.223216 133136767887168 maxtext_utils.py:1836]  decoder/layers/6/self_attention/value/kernel/value
    Shape:     float32[2048,16,128]
    Physical:  ('fsdp', None, None)
I0420 18:11:21.223244 133136767887168 maxtext_utils.py:1836]  decoder/layers/7/mlp/wi_0/kernel/value
    Shape:     float32[2048,7168]
    Physical:  ('fsdp', None)
I0420 18:11:21.223272 133136767887168 maxtext_utils.py:1836]  decoder/layers/7/mlp/wi_1/kernel/value
    Shape:     float32[2048,7168]
    Physical:  ('fsdp', None)
I0420 18:11:21.223299 133136767887168 maxtext_utils.py:1836]  decoder/layers/7/mlp/wo/kernel/value
    Shape:     float32[7168,2048]
    Physical:  (None, 'fsdp')
I0420 18:11:21.223326 133136767887168 maxtext_utils.py:1836]  decoder/layers/7/post_self_attention_layer_norm/scale/value
    Shape:     float32[2048]
    Physical:  (None,)
I0420 18:11:21.223353 133136767887168 maxtext_utils.py:1836]  decoder/layers/7/pre_self_attention_layer_norm/scale/value
    Shape:     float32[2048]
    Physical:  (None,)
I0420 18:11:21.223381 133136767887168 maxtext_utils.py:1836]  decoder/layers/7/self_attention/key/kernel/value
    Shape:     float32[2048,16,128]
    Physical:  ('fsdp', None, None)
I0420 18:11:21.223409 133136767887168 maxtext_utils.py:1836]  decoder/layers/7/self_attention/out/kernel/value
    Shape:     float32[16,128,2048]
    Physical:  (None, None, 'fsdp')
I0420 18:11:21.223436 133136767887168 maxtext_utils.py:1836]  decoder/layers/7/self_attention/query/kernel/value
    Shape:     float32[2048,16,128]
    Physical:  ('fsdp', None, None)
I0420 18:11:21.223464 133136767887168 maxtext_utils.py:1836]  decoder/layers/7/self_attention/value/kernel/value
    Shape:     float32[2048,16,128]
    Physical:  ('fsdp', None, None)
I0420 18:11:21.223491 133136767887168 maxtext_utils.py:1836]  decoder/layers/8/mlp/wi_0/kernel/value
    Shape:     float32[2048,7168]
    Physical:  ('fsdp', None)
I0420 18:11:21.223519 133136767887168 maxtext_utils.py:1836]  decoder/layers/8/mlp/wi_1/kernel/value
    Shape:     float32[2048,7168]
    Physical:  ('fsdp', None)
I0420 18:11:21.223546 133136767887168 maxtext_utils.py:1836]  decoder/layers/8/mlp/wo/kernel/value
    Shape:     float32[7168,2048]
    Physical:  (None, 'fsdp')
I0420 18:11:21.223572 133136767887168 maxtext_utils.py:1836]  decoder/layers/8/post_self_attention_layer_norm/scale/value
    Shape:     float32[2048]
    Physical:  (None,)
I0420 18:11:21.223599 133136767887168 maxtext_utils.py:1836]  decoder/layers/8/pre_self_attention_layer_norm/scale/value
    Shape:     float32[2048]
    Physical:  (None,)
I0420 18:11:21.223631 133136767887168 maxtext_utils.py:1836]  decoder/layers/8/self_attention/key/kernel/value
    Shape:     float32[2048,16,128]
    Physical:  ('fsdp', None, None)
I0420 18:11:21.223660 133136767887168 maxtext_utils.py:1836]  decoder/layers/8/self_attention/out/kernel/value
    Shape:     float32[16,128,2048]
    Physical:  (None, None, 'fsdp')
I0420 18:11:21.223688 133136767887168 maxtext_utils.py:1836]  decoder/layers/8/self_attention/query/kernel/value
    Shape:     float32[2048,16,128]
    Physical:  ('fsdp', None, None)
I0420 18:11:21.223716 133136767887168 maxtext_utils.py:1836]  decoder/layers/8/self_attention/value/kernel/value
    Shape:     float32[2048,16,128]
    Physical:  ('fsdp', None, None)
I0420 18:11:21.223744 133136767887168 maxtext_utils.py:1836]  decoder/layers/9/mlp/wi_0/kernel/value
    Shape:     float32[2048,7168]
    Physical:  ('fsdp', None)
I0420 18:11:21.223772 133136767887168 maxtext_utils.py:1836]  decoder/layers/9/mlp/wi_1/kernel/value
    Shape:     float32[2048,7168]
    Physical:  ('fsdp', None)
I0420 18:11:21.223800 133136767887168 maxtext_utils.py:1836]  decoder/layers/9/mlp/wo/kernel/value
    Shape:     float32[7168,2048]
    Physical:  (None, 'fsdp')
I0420 18:11:21.223825 133136767887168 maxtext_utils.py:1836]  decoder/layers/9/post_self_attention_layer_norm/scale/value
    Shape:     float32[2048]
    Physical:  (None,)
I0420 18:11:21.223852 133136767887168 maxtext_utils.py:1836]  decoder/layers/9/pre_self_attention_layer_norm/scale/value
    Shape:     float32[2048]
    Physical:  (None,)
I0420 18:11:21.223880 133136767887168 maxtext_utils.py:1836]  decoder/layers/9/self_attention/key/kernel/value
    Shape:     float32[2048,16,128]
    Physical:  ('fsdp', None, None)
I0420 18:11:21.223908 133136767887168 maxtext_utils.py:1836]  decoder/layers/9/self_attention/out/kernel/value
    Shape:     float32[16,128,2048]
    Physical:  (None, None, 'fsdp')
I0420 18:11:21.223936 133136767887168 maxtext_utils.py:1836]  decoder/layers/9/self_attention/query/kernel/value
    Shape:     float32[2048,16,128]
    Physical:  ('fsdp', None, None)
I0420 18:11:21.223964 133136767887168 maxtext_utils.py:1836]  decoder/layers/9/self_attention/value/kernel/value
    Shape:     float32[2048,16,128]
    Physical:  ('fsdp', None, None)
I0420 18:11:21.223992 133136767887168 maxtext_utils.py:1836]  decoder/layers/10/mlp/wi_0/kernel/value
    Shape:     float32[2048,7168]
    Physical:  ('fsdp', None)
I0420 18:11:21.224019 133136767887168 maxtext_utils.py:1836]  decoder/layers/10/mlp/wi_1/kernel/value
    Shape:     float32[2048,7168]
    Physical:  ('fsdp', None)
I0420 18:11:21.224046 133136767887168 maxtext_utils.py:1836]  decoder/layers/10/mlp/wo/kernel/value
    Shape:     float32[7168,2048]
    Physical:  (None, 'fsdp')
I0420 18:11:21.224072 133136767887168 maxtext_utils.py:1836]  decoder/layers/10/post_self_attention_layer_norm/scale/value
    Shape:     float32[2048]
    Physical:  (None,)
I0420 18:11:21.224113 133136767887168 maxtext_utils.py:1836]  decoder/layers/10/pre_self_attention_layer_norm/scale/value
    Shape:     float32[2048]
    Physical:  (None,)
I0420 18:11:21.224143 133136767887168 maxtext_utils.py:1836]  decoder/layers/10/self_attention/key/kernel/value
    Shape:     float32[2048,16,128]
    Physical:  ('fsdp', None, None)
I0420 18:11:21.224173 133136767887168 maxtext_utils.py:1836]  decoder/layers/10/self_attention/out/kernel/value
    Shape:     float32[16,128,2048]
    Physical:  (None, None, 'fsdp')
I0420 18:11:21.224201 133136767887168 maxtext_utils.py:1836]  decoder/layers/10/self_attention/query/kernel/value
    Shape:     float32[2048,16,128]
    Physical:  ('fsdp', None, None)
I0420 18:11:21.224229 133136767887168 maxtext_utils.py:1836]  decoder/layers/10/self_attention/value/kernel/value
    Shape:     float32[2048,16,128]
    Physical:  ('fsdp', None, None)
I0420 18:11:21.224256 133136767887168 maxtext_utils.py:1836]  decoder/layers/11/mlp/wi_0/kernel/value
    Shape:     float32[2048,7168]
    Physical:  ('fsdp', None)
I0420 18:11:21.224284 133136767887168 maxtext_utils.py:1836]  decoder/layers/11/mlp/wi_1/kernel/value
    Shape:     float32[2048,7168]
    Physical:  ('fsdp', None)
I0420 18:11:21.224311 133136767887168 maxtext_utils.py:1836]  decoder/layers/11/mlp/wo/kernel/value
    Shape:     float32[7168,2048]
    Physical:  (None, 'fsdp')
I0420 18:11:21.224337 133136767887168 maxtext_utils.py:1836]  decoder/layers/11/post_self_attention_layer_norm/scale/value
    Shape:     float32[2048]
    Physical:  (None,)
I0420 18:11:21.224363 133136767887168 maxtext_utils.py:1836]  decoder/layers/11/pre_self_attention_layer_norm/scale/value
    Shape:     float32[2048]
    Physical:  (None,)
I0420 18:11:21.224391 133136767887168 maxtext_utils.py:1836]  decoder/layers/11/self_attention/key/kernel/value
    Shape:     float32[2048,16,128]
    Physical:  ('fsdp', None, None)
I0420 18:11:21.224419 133136767887168 maxtext_utils.py:1836]  decoder/layers/11/self_attention/out/kernel/value
    Shape:     float32[16,128,2048]
    Physical:  (None, None, 'fsdp')
I0420 18:11:21.224446 133136767887168 maxtext_utils.py:1836]  decoder/layers/11/self_attention/query/kernel/value
    Shape:     float32[2048,16,128]
    Physical:  ('fsdp', None, None)
I0420 18:11:21.224473 133136767887168 maxtext_utils.py:1836]  decoder/layers/11/self_attention/value/kernel/value
    Shape:     float32[2048,16,128]
    Physical:  ('fsdp', None, None)
I0420 18:11:21.224501 133136767887168 maxtext_utils.py:1836]  decoder/layers/12/mlp/wi_0/kernel/value
    Shape:     float32[2048,7168]
    Physical:  ('fsdp', None)
I0420 18:11:21.224528 133136767887168 maxtext_utils.py:1836]  decoder/layers/12/mlp/wi_1/kernel/value
    Shape:     float32[2048,7168]
    Physical:  ('fsdp', None)
I0420 18:11:21.224555 133136767887168 maxtext_utils.py:1836]  decoder/layers/12/mlp/wo/kernel/value
    Shape:     float32[7168,2048]
    Physical:  (None, 'fsdp')
I0420 18:11:21.224580 133136767887168 maxtext_utils.py:1836]  decoder/layers/12/post_self_attention_layer_norm/scale/value
    Shape:     float32[2048]
    Physical:  (None,)
I0420 18:11:21.224606 133136767887168 maxtext_utils.py:1836]  decoder/layers/12/pre_self_attention_layer_norm/scale/value
    Shape:     float32[2048]
    Physical:  (None,)
I0420 18:11:21.224637 133136767887168 maxtext_utils.py:1836]  decoder/layers/12/self_attention/key/kernel/value
    Shape:     float32[2048,16,128]
    Physical:  ('fsdp', None, None)
I0420 18:11:21.224665 133136767887168 maxtext_utils.py:1836]  decoder/layers/12/self_attention/out/kernel/value
    Shape:     float32[16,128,2048]
    Physical:  (None, None, 'fsdp')
I0420 18:11:21.224693 133136767887168 maxtext_utils.py:1836]  decoder/layers/12/self_attention/query/kernel/value
    Shape:     float32[2048,16,128]
    Physical:  ('fsdp', None, None)
I0420 18:11:21.224720 133136767887168 maxtext_utils.py:1836]  decoder/layers/12/self_attention/value/kernel/value
    Shape:     float32[2048,16,128]
    Physical:  ('fsdp', None, None)
I0420 18:11:21.224747 133136767887168 maxtext_utils.py:1836]  decoder/layers/13/mlp/wi_0/kernel/value
    Shape:     float32[2048,7168]
    Physical:  ('fsdp', None)
I0420 18:11:21.224774 133136767887168 maxtext_utils.py:1836]  decoder/layers/13/mlp/wi_1/kernel/value
    Shape:     float32[2048,7168]
    Physical:  ('fsdp', None)
I0420 18:11:21.224802 133136767887168 maxtext_utils.py:1836]  decoder/layers/13/mlp/wo/kernel/value
    Shape:     float32[7168,2048]
    Physical:  (None, 'fsdp')
I0420 18:11:21.224828 133136767887168 maxtext_utils.py:1836]  decoder/layers/13/post_self_attention_layer_norm/scale/value
    Shape:     float32[2048]
    Physical:  (None,)
I0420 18:11:21.224853 133136767887168 maxtext_utils.py:1836]  decoder/layers/13/pre_self_attention_layer_norm/scale/value
    Shape:     float32[2048]
    Physical:  (None,)
I0420 18:11:21.224881 133136767887168 maxtext_utils.py:1836]  decoder/layers/13/self_attention/key/kernel/value
    Shape:     float32[2048,16,128]
    Physical:  ('fsdp', None, None)
I0420 18:11:21.224909 133136767887168 maxtext_utils.py:1836]  decoder/layers/13/self_attention/out/kernel/value
    Shape:     float32[16,128,2048]
    Physical:  (None, None, 'fsdp')
I0420 18:11:21.224936 133136767887168 maxtext_utils.py:1836]  decoder/layers/13/self_attention/query/kernel/value
    Shape:     float32[2048,16,128]
    Physical:  ('fsdp', None, None)
I0420 18:11:21.224963 133136767887168 maxtext_utils.py:1836]  decoder/layers/13/self_attention/value/kernel/value
    Shape:     float32[2048,16,128]
    Physical:  ('fsdp', None, None)
I0420 18:11:21.224991 133136767887168 maxtext_utils.py:1836]  decoder/layers/14/mlp/wi_0/kernel/value
    Shape:     float32[2048,7168]
    Physical:  ('fsdp', None)
I0420 18:11:21.225018 133136767887168 maxtext_utils.py:1836]  decoder/layers/14/mlp/wi_1/kernel/value
    Shape:     float32[2048,7168]
    Physical:  ('fsdp', None)
I0420 18:11:21.225046 133136767887168 maxtext_utils.py:1836]  decoder/layers/14/mlp/wo/kernel/value
    Shape:     float32[7168,2048]
    Physical:  (None, 'fsdp')
I0420 18:11:21.225074 133136767887168 maxtext_utils.py:1836]  decoder/layers/14/post_self_attention_layer_norm/scale/value
    Shape:     float32[2048]
    Physical:  (None,)
I0420 18:11:21.225113 133136767887168 maxtext_utils.py:1836]  decoder/layers/14/pre_self_attention_layer_norm/scale/value
    Shape:     float32[2048]
    Physical:  (None,)
I0420 18:11:21.225144 133136767887168 maxtext_utils.py:1836]  decoder/layers/14/self_attention/key/kernel/value
    Shape:     float32[2048,16,128]
    Physical:  ('fsdp', None, None)
I0420 18:11:21.225172 133136767887168 maxtext_utils.py:1836]  decoder/layers/14/self_attention/out/kernel/value
    Shape:     float32[16,128,2048]
    Physical:  (None, None, 'fsdp')
I0420 18:11:21.225201 133136767887168 maxtext_utils.py:1836]  decoder/layers/14/self_attention/query/kernel/value
    Shape:     float32[2048,16,128]
    Physical:  ('fsdp', None, None)
I0420 18:11:21.225229 133136767887168 maxtext_utils.py:1836]  decoder/layers/14/self_attention/value/kernel/value
    Shape:     float32[2048,16,128]
    Physical:  ('fsdp', None, None)
I0420 18:11:21.225256 133136767887168 maxtext_utils.py:1836]  decoder/layers/15/mlp/wi_0/kernel/value
    Shape:     float32[2048,7168]
    Physical:  ('fsdp', None)
I0420 18:11:21.225284 133136767887168 maxtext_utils.py:1836]  decoder/layers/15/mlp/wi_1/kernel/value
    Shape:     float32[2048,7168]
    Physical:  ('fsdp', None)
I0420 18:11:21.225312 133136767887168 maxtext_utils.py:1836]  decoder/layers/15/mlp/wo/kernel/value
    Shape:     float32[7168,2048]
    Physical:  (None, 'fsdp')
I0420 18:11:21.225338 133136767887168 maxtext_utils.py:1836]  decoder/layers/15/post_self_attention_layer_norm/scale/value
    Shape:     float32[2048]
    Physical:  (None,)
I0420 18:11:21.225365 133136767887168 maxtext_utils.py:1836]  decoder/layers/15/pre_self_attention_layer_norm/scale/value
    Shape:     float32[2048]
    Physical:  (None,)
I0420 18:11:21.225393 133136767887168 maxtext_utils.py:1836]  decoder/layers/15/self_attention/key/kernel/value
    Shape:     float32[2048,16,128]
    Physical:  ('fsdp', None, None)
I0420 18:11:21.225420 133136767887168 maxtext_utils.py:1836]  decoder/layers/15/self_attention/out/kernel/value
    Shape:     float32[16,128,2048]
    Physical:  (None, None, 'fsdp')
I0420 18:11:21.225449 133136767887168 maxtext_utils.py:1836]  decoder/layers/15/self_attention/query/kernel/value
    Shape:     float32[2048,16,128]
    Physical:  ('fsdp', None, None)
I0420 18:11:21.225477 133136767887168 maxtext_utils.py:1836]  decoder/layers/15/self_attention/value/kernel/value
    Shape:     float32[2048,16,128]
    Physical:  ('fsdp', None, None)
I0420 18:11:21.225505 133136767887168 maxtext_utils.py:1836]  decoder/logits_dense/kernel/value
    Shape:     float32[2048,32000]
    Physical:  ('fsdp', None)
I0420 18:11:21.225533 133136767887168 maxtext_utils.py:1836]  token_embedder/embedding/value
    Shape:     float32[32000,2048]
    Physical:  (None, 'fsdp')

I0420 18:11:21.720626 133136767887168 nnx_decoders.py:1023] nnx_decoders/y_in Logical: bfloat16[32,2048,2048]...................................... ('activation_batch', 'activation_norm_length', 'activation_embed').
I0420 18:11:21.720725 133136767887168 nnx_decoders.py:1023] nnx_decoders/y_in Physical: bfloat16[32,2048,2048]...................................... ('fsdp', None, None).
I0420 18:11:21.725327 133136767887168 nnx_decoders.py:1023] Unknown Logical: bfloat16[32,2048,2048]...................................... ('activation_batch', 'activation_norm_length', 'activation_embed').
I0420 18:11:21.725387 133136767887168 nnx_decoders.py:1023] Unknown Physical: bfloat16[32,2048,2048]...................................... ('fsdp', None, None).
I0420 18:11:21.741285 133136767887168 attentions.py:1088] attentions/inputs_q Logical: bfloat16[32,2048,2048]...................................... ('activation_batch', 'activation_attn_length', 'activation_attn_embed').
I0420 18:11:21.741342 133136767887168 attentions.py:1088] attentions/inputs_q Physical: bfloat16[32,2048,2048]...................................... ('fsdp', None, None).
I0420 18:11:21.756678 133136767887168 attentions.py:1089] attentions/inputs_kv Logical: bfloat16[32,2048,2048]...................................... ('activation_batch', 'activation_attn_length', 'activation_attn_embed').
I0420 18:11:21.756736 133136767887168 attentions.py:1089] attentions/inputs_kv Physical: bfloat16[32,2048,2048]...................................... ('fsdp', None, None).
I0420 18:11:21.779033 133136767887168 attentions.py:1154] attentions/query Logical: bfloat16[32,2048,16,128].................................... ('activation_kv_batch', 'activation_attn_length', 'activation_kv_heads', 'activation_kv_head_dim').
I0420 18:11:21.779115 133136767887168 attentions.py:1154] attentions/query Physical: bfloat16[32,2048,16,128].................................... ('fsdp', None, None, None).
I0420 18:11:21.794472 133136767887168 attentions.py:1155] attentions/key Logical: bfloat16[32,2048,16,128].................................... ('activation_kv_batch', 'activation_attn_length', 'activation_kv_heads', 'activation_kv_head_dim').
I0420 18:11:21.794531 133136767887168 attentions.py:1155] attentions/key Physical: bfloat16[32,2048,16,128].................................... ('fsdp', None, None, None).
I0420 18:11:21.809878 133136767887168 attentions.py:1156] attentions/value Logical: bfloat16[32,2048,16,128].................................... ('activation_kv_batch', 'activation_attn_length', 'activation_kv_heads', 'activation_kv_head_dim').
I0420 18:11:21.809935 133136767887168 attentions.py:1156] attentions/value Physical: bfloat16[32,2048,16,128].................................... ('fsdp', None, None, None).
I0420 18:11:21.835927 133136767887168 attentions.py:1197] attentions/out Logical: bfloat16[32,2048,16,128].................................... ('activation_batch', 'activation_attn_length', 'activation_heads', 'activation_kv').
I0420 18:11:21.835995 133136767887168 attentions.py:1197] attentions/out Physical: bfloat16[32,2048,16,128].................................... ('fsdp', None, None, None).
I0420 18:11:21.856068 133136767887168 linears.py:525] linears/x Logical: bfloat16[32,2048,7168]...................................... ('activation_batch', 'activation_length', 'activation_mlp').
I0420 18:11:21.856144 133136767887168 linears.py:525] linears/x Physical: bfloat16[32,2048,7168]...................................... ('fsdp', None, None).
I0420 18:12:22.229855 133136767887168 max_utils.py:791] Total memory size: 1.8 GB, Output size: 0.4 GB, Temp size: 1.4 GB, Argument size: 0.4 GB, Host temp size: 0.0 GB.
I0420 18:12:22.246684 133136767887168 metric_logger.py:301] number parameters: 1.104 billion
I0420 18:13:24.415490 133136767887168 checkpointing.py:794] Waiting for step 0 to finish before checkpoint...
I0420 18:13:24.633675 133136767887168 checkpointing.py:798] Waited 0.21845650672912598 seconds for step 0 to finish before starting checkpointing.
I0420 18:13:24.636942 133136767887168 checkpoint_manager.py:2009] [process=0][thread=MainThread][wait_until_finished] No Save Finalize thread to wait for. Returning.
I0420 18:13:24.638789 133136767887168 checkpoint_manager.py:1512] [process=0] Saving checkpoint at step 0
I0420 18:13:24.640316 133136767887168 event_tracking.py:70] [process=0] [async] Started save checkpoint @ gs://lance-maxtext/nnx_ckpt_xpk_feat_nnx_post_train_fixes_20260420_153543/nnx_xpk_feat_nnx_post_train_fixes_20260420_153543_13_scan_layers_false/checkpoints/0.
I0420 18:13:25.821234 133136767887168 signaling_client.py:364] Using JaxDistributedSignalingClient
I0420 18:13:25.901491 133136767887168 jax_array_handlers.py:360] Scheduling D2H of 648 prioritized jax.Array.
I0420 18:13:25.901983 133136767887168 replica_slices.py:424] 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
I0420 18:13:26.111292 133010022377216 atomicity.py:140] Creating tmp directory gs://lance-maxtext/nnx_ckpt_xpk_feat_nnx_post_train_fixes_20260420_153543/nnx_xpk_feat_nnx_post_train_fixes_20260420_153543_13_scan_layers_false/checkpoints/0
I0420 18:13:26.225178 133136767887168 base_pytree_checkpoint_handler.py:154] [process=0][thread=MainThread] Initiated "orbax.checkpoint._src.serialization.jax_array_handlers.ArrayHandler".serialize. Time taken: 0.329310s
I0420 18:13:26.226016 133136767887168 base_pytree_checkpoint_handler.py:130] [process=0] /jax/orbax/write/blocking_gbytes_per_sec: 3.893 GiB/s (total gbytes: 1.5 GiB) (time elapsed: 0.39629316329956055 s) (per-host)
I0420 18:13:26.226097 133136767887168 base_pytree_checkpoint_handler.py:768] [process=0][thread=MainThread] Initiated Pytree async_save. Time taken: 0.396376s (batch_requests_ready=0.026103s, total_serialization_initiated=0.369517s, others=0.000756s)
I0420 18:13:26.226260 133136767887168 composite_checkpoint_handler.py:715] [process=0][thread=MainThread] Initiated CompositeCheckpointHandler.async_save. Time taken: 0.400527s (all_items=0.000017s, per_item={'items': '0.00001669'}, temp_paths=0.400511)
I0420 18:13:26.227131 133136767887168 event_tracking.py:125] [process=0] [async] Finished blocking save in 1.59 seconds. Continuing save @ gs://lance-maxtext/nnx_ckpt_xpk_feat_nnx_post_train_fixes_20260420_153543/nnx_xpk_feat_nnx_post_train_fixes_20260420_153543_13_scan_layers_false/checkpoints/0.
I0420 18:13:26.227482 133010030769920 async_checkpointer.py:76] [process=0][thread=async_save] Background save thread started. Deadline for this save operation is 2026-04-20 18:33:26.227445
I0420 18:13:26.242489 133136767887168 checkpoint_manager.py:1560] [process=0][thread=MainThread][step=0] Starting CheckpointManager Save Finalize thread=save_finalize
I0420 18:13:26.242767 133009980413696 async_checkpointer.py:280] [process=0][thread=save_finalize] Waiting for background save thread=async_save.
I0420 18:13:26.242924 133136767887168 standard_logger.py:34] {'step': 0, 'event_type': 'save', 'directory': 'gs://lance-maxtext/nnx_ckpt_xpk_feat_nnx_post_train_fixes_20260420_153543/nnx_xpk_feat_nnx_post_train_fixes_20260420_153543_13_scan_layers_false/checkpoints', 'reached_preemption': False, 'preemption_received_at': None, 'synchronous': False, 'wait_for_prev_start_time': 1776708804.6369238, 'wait_for_prev_duration_secs': 6.151199340820312e-05, 'time_between_consecutive_saves_sec': None, 'checkpointer_blocking_start_time': 1776708804.6388276, 'checkpointer_blocking_duration_secs': 1.588804006576538, 'get_old_steps_start_time': 1776708806.2276556, 'get_old_steps_duration_secs': 2.5987625122070312e-05, 'checkpoint_manager_blocking_start_time': 1776708804.6351438, 'checkpoint_manager_blocking_duration_secs': 1.6077420711517334}
I0420 18:13:26.243172 133136767887168 checkpointing.py:409] Started an asynchronous checkpoint save for step 0
I0420 18:13:26.243373 133136767887168 max_utils.py:750] 
Memstats: After params initialized:
I0420 18:13:26.243425 133136767887168 max_utils.py:756] 	Using (GB) 0.82 / 31.25 (2.624000%) on TPU_0(process=0,(0,0,0,0))
I0420 18:13:26.243459 133136767887168 max_utils.py:756] 	Using (GB) 0.82 / 31.25 (2.624000%) on TPU_1(process=0,(1,0,0,0))
I0420 18:13:26.243486 133136767887168 max_utils.py:756] 	Using (GB) 0.82 / 31.25 (2.624000%) on TPU_4(process=0,(0,1,0,0))
I0420 18:13:26.243521 133136767887168 max_utils.py:756] 	Using (GB) 0.82 / 31.25 (2.624000%) on TPU_5(process=0,(1,1,0,0))
I0420 18:13:26.656055 133136767887168 metric_logger.py:196] completed step: 0, seconds: 61.615, TFLOP/s/device: 0.221, Tokens/s/device: 33.239, total_weights: 65536, loss: 10.874, lm_loss: 10.874, perplexity: 52794.566
I0420 18:13:26.659171 133136767887168 metric_logger.py:281] To see full metrics 'tensorboard --logdir=gs://lance-maxtext/nnx_ckpt_xpk_feat_nnx_post_train_fixes_20260420_153543/nnx_xpk_feat_nnx_post_train_fixes_20260420_153543_13_scan_layers_false/tensorboard/'
I0420 18:13:27.179124 133136767887168 metric_logger.py:196] completed step: 1, seconds: 2.217, TFLOP/s/device: 6.128, Tokens/s/device: 923.673, total_weights: 65536, loss: 10.874, lm_loss: 10.874, perplexity: 52794.566
I0420 18:13:27.276608 133136767887168 metric_logger.py:196] completed step: 2, seconds: 0.531, TFLOP/s/device: 25.601, Tokens/s/device: 3858.901, total_weights: 65536, loss: 10.265, lm_loss: 10.265, perplexity: 28706.043
I0420 18:13:27.412240 133136767887168 metric_logger.py:196] completed step: 3, seconds: 0.040, TFLOP/s/device: 337.300, Tokens/s/device: 50841.567, total_weights: 65536, loss: 9.736, lm_loss: 9.736, perplexity: 16919.420
I0420 18:13:28.205409 133010005591808 atomicity.py:140] Creating tmp directory gs://lance-maxtext/nnx_ckpt_xpk_feat_nnx_post_train_fixes_20260420_153543/nnx_xpk_feat_nnx_post_train_fixes_20260420_153543_13_scan_layers_false/checkpoints/0/items
I0420 18:13:28.338520 133009500010240 checkpoint.py:188] Wrote Metadata={'item_handlers': None, 'metrics': {}, 'performance_metrics': {}, 'init_timestamp_nsecs': 1776708807922864779, 'commit_timestamp_nsecs': None, 'custom_metadata': {}}, json={"item_handlers": null, "metrics": {}, "performance_metrics": {}, "init_timestamp_nsecs": 1776708807922864779, "commit_timestamp_nsecs": null, "custom_metadata": {}} to gs://lance-maxtext/nnx_ckpt_xpk_feat_nnx_post_train_fixes_20260420_153543/nnx_xpk_feat_nnx_post_train_fixes_20260420_153543_13_scan_layers_false/checkpoints/0/_CHECKPOINT_METADATA
I0420 18:13:29.854105    2828 google_auth_provider.cc:181] Running on GCE, using service account 562977990677-compute@developer.gserviceaccount.com
I0420 18:13:32.333551 133012184364800 array_metadata_store.py:203] [process=0][thread=array_type_handler] Wrote 648 array_metadata.ArrayMetadata to gs://lance-maxtext/nnx_ckpt_xpk_feat_nnx_post_train_fixes_20260420_153543/nnx_xpk_feat_nnx_post_train_fixes_20260420_153543_13_scan_layers_false/checkpoints/0/items/array_metadatas/process_0
I0420 18:13:50.555874 133136767887168 metric_logger.py:196] completed step: 4, seconds: 0.099, TFLOP/s/device: 137.738, Tokens/s/device: 20761.316, total_weights: 65536, loss: 9.279, lm_loss: 9.279, perplexity: 10712.232
I0420 18:13:50.581939 133136767887168 metric_logger.py:196] completed step: 5, seconds: 0.139, TFLOP/s/device: 97.555, Tokens/s/device: 14704.510, total_weights: 65536, loss: 8.894, lm_loss: 8.894, perplexity: 7286.416
I0420 18:13:50.701299 133136767887168 metric_logger.py:196] completed step: 6, seconds: 23.135, TFLOP/s/device: 0.587, Tokens/s/device: 88.523, total_weights: 65536, loss: 8.595, lm_loss: 8.595, perplexity: 5404.646
I0420 18:13:50.838994 133136767887168 metric_logger.py:196] completed step: 7, seconds: 0.022, TFLOP/s/device: 612.612, Tokens/s/device: 92339.601, total_weights: 65536, loss: 8.387, lm_loss: 8.387, perplexity: 4388.929
I0420 18:13:50.974493 133136767887168 metric_logger.py:196] completed step: 8, seconds: 0.122, TFLOP/s/device: 110.948, Tokens/s/device: 16723.282, total_weights: 65536, loss: 8.258, lm_loss: 8.258, perplexity: 3856.623
I0420 18:13:51.116929 133136767887168 checkpointing.py:794] Waiting for step 9 to finish before checkpoint...
I0420 18:13:51.121474 133136767887168 checkpointing.py:798] Waited 0.004555463790893555 seconds for step 9 to finish before starting checkpointing.
I0420 18:13:51.124368 133136767887168 checkpoint_manager.py:2020] [process=0][thread=MainThread][step=0][wait_until_finished] Waiting for Save Finalize thread (save_finalize) to complete.
I0420 18:14:05.433530 133009988806400 base_pytree_checkpoint_handler.py:1282] [process=0][thread=write_metadata_after_commits] Commit + Array metadata written. Time taken: 35.623430s (commit=33.827135s, array_metadata_write=1.796295s)
I0420 18:14:05.435451 133010030769920 base_pytree_checkpoint_handler.py:130] [process=0] /jax/orbax/write/gbytes_per_sec: 39.883 MiB/s (total gbytes: 1.5 GiB) (time elapsed: 39.60569095611572 s) (per-host)
I0420 18:14:05.435609 133010030769920 async_checkpointer.py:90] [process=0][thread=async_save] 3 Handler Commit operations completed. Time taken: 39.208005s.
I0420 18:14:06.123091 133010030769920 checkpoint.py:228] Read Metadata={'item_handlers': None, 'metrics': {}, 'performance_metrics': {}, 'init_timestamp_nsecs': 1776708807922864779, 'commit_timestamp_nsecs': None, 'custom_metadata': {}} from gs://lance-maxtext/nnx_ckpt_xpk_feat_nnx_post_train_fixes_20260420_153543/nnx_xpk_feat_nnx_post_train_fixes_20260420_153543_13_scan_layers_false/checkpoints/0/_CHECKPOINT_METADATA
I0420 18:14:06.882497 133009500010240 checkpoint.py:247] Updated Metadata={'item_handlers': {'items': 'orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeCheckpointHandler'}, 'metrics': {}, 'performance_metrics': {}, 'init_timestamp_nsecs': 1776708807922864779, 'commit_timestamp_nsecs': None, 'custom_metadata': {}} to gs://lance-maxtext/nnx_ckpt_xpk_feat_nnx_post_train_fixes_20260420_153543/nnx_xpk_feat_nnx_post_train_fixes_20260420_153543_13_scan_layers_false/checkpoints/0/_CHECKPOINT_METADATA
I0420 18:14:07.426698 133010030769920 ocdbt_utils.py:49] Param validation support for Zarr3 will be added later (b/362328389).
I0420 18:14:07.499860 133010030769920 array_metadata_store.py:411] [process=0][thread=async_save] Validated ArrayMetadata from all 8 hosts. Time taken: 0.003433s.
I0420 18:14:08.907687 133010030769920 base_pytree_checkpoint_handler.py:1406] [process=0][thread=async_save] Pytree save finalize (merge_ocdbt + ArrayMetadata validation) completed. Time taken: 2.658303s. use_zarr3=True, enable_post_merge_validation=True, directory=gs://lance-maxtext/nnx_ckpt_xpk_feat_nnx_post_train_fixes_20260420_153543/nnx_xpk_feat_nnx_post_train_fixes_20260420_153543_13_scan_layers_false/checkpoints/0/items
I0420 18:14:08.909422 133010030769920 atomicity.py:666] Finalizing gs://lance-maxtext/nnx_ckpt_xpk_feat_nnx_post_train_fixes_20260420_153543/nnx_xpk_feat_nnx_post_train_fixes_20260420_153543_13_scan_layers_false/checkpoints/0/items
I0420 18:14:09.447751 133010030769920 atomicity.py:666] Finalizing gs://lance-maxtext/nnx_ckpt_xpk_feat_nnx_post_train_fixes_20260420_153543/nnx_xpk_feat_nnx_post_train_fixes_20260420_153543_13_scan_layers_false/checkpoints/0
I0420 18:14:10.904956 133010030769920 atomicity.py:847] [process=0][thread=async_save] Finished saving checkpoint (finalized tmp dir) to `gs://lance-maxtext/nnx_ckpt_xpk_feat_nnx_post_train_fixes_20260420_153543/nnx_xpk_feat_nnx_post_train_fixes_20260420_153543_13_scan_layers_false/checkpoints/0`.
I0420 18:14:10.905666 133010030769920 event_tracking.py:138] [process=0] [async] Finished save (blocking + background) in 46.27 seconds @ gs://lance-maxtext/nnx_ckpt_xpk_feat_nnx_post_train_fixes_20260420_153543/nnx_xpk_feat_nnx_post_train_fixes_20260420_153543_13_scan_layers_false/checkpoints/0
I0420 18:14:10.907032 133010030769920 async_checkpointer.py:160] [process=0][thread=async_save] Background save thread done. Time taken: 44.679432s.
I0420 18:14:10.907228 133009980413696 async_checkpointer.py:288] [process=0][thread=save_finalize] Done with waiting for background save thread=async_save.
I0420 18:14:10.907352 133009980413696 async_checkpointer.py:298] [process=0][thread=save_finalize] No errors found in background save thread=async_save.
I0420 18:14:10.907421 133009980413696 checkpoint_manager.py:2137] [process=0][thread=save_finalize][step=0] CheckpointManager Save Finalize is syncing with other hosts...
I0420 18:14:10.910396 133009980413696 checkpoint_manager.py:2146] [process=0][thread=save_finalize][step=0] CheckpointManager Save Finalize is done on all hosts.
I0420 18:14:10.910578 133136767887168 checkpoint_manager.py:2032] [process=0][thread=MainThread][step=0][wait_until_finished] Done waiting for Save Finalize thread (save_finalize) running at step=0.
W0420 18:14:10.910721 133136767887168 checkpoint_manager.py:1452] Waiting for previous save to complete took 19.786350 seconds. If this number is high, consider checkpointing less frequently.
I0420 18:14:10.912995 133136767887168 checkpoint_manager.py:1512] [process=0] Saving checkpoint at step 9
I0420 18:14:10.914988 133136767887168 event_tracking.py:70] [process=0] [async] Started save checkpoint @ gs://lance-maxtext/nnx_ckpt_xpk_feat_nnx_post_train_fixes_20260420_153543/nnx_xpk_feat_nnx_post_train_fixes_20260420_153543_13_scan_layers_false/checkpoints/9.
I0420 18:14:12.130505 133136767887168 jax_array_handlers.py:360] Scheduling D2H of 648 prioritized jax.Array.
I0420 18:14:12.130653 133136767887168 replica_slices.py:424] 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
I0420 18:14:12.262310 133136767887168 base_pytree_checkpoint_handler.py:154] [process=0][thread=MainThread] Initiated "orbax.checkpoint._src.serialization.jax_array_handlers.ArrayHandler".serialize. Time taken: 0.137521s
I0420 18:14:12.263179 133136767887168 base_pytree_checkpoint_handler.py:130] [process=0] /jax/orbax/write/blocking_gbytes_per_sec: 7.600 GiB/s (total gbytes: 1.5 GiB) (time elapsed: 0.20296955108642578 s) (per-host)
I0420 18:14:12.263240 133136767887168 base_pytree_checkpoint_handler.py:768] [process=0][thread=MainThread] Initiated Pytree async_save. Time taken: 0.203045s (batch_requests_ready=0.025525s, total_serialization_initiated=0.176738s, others=0.000782s)
I0420 18:14:12.263391 133136767887168 composite_checkpoint_handler.py:715] [process=0][thread=MainThread] Initiated CompositeCheckpointHandler.async_save. Time taken: 0.207147s (all_items=0.000016s, per_item={'items': '0.00001645'}, temp_paths=0.207130)
I0420 18:14:12.264157 133136767887168 event_tracking.py:125] [process=0] [async] Finished blocking save in 1.35 seconds. Continuing save @ gs://lance-maxtext/nnx_ckpt_xpk_feat_nnx_post_train_fixes_20260420_153543/nnx_xpk_feat_nnx_post_train_fixes_20260420_153543_13_scan_layers_false/checkpoints/9.
I0420 18:14:12.264528 133009980413696 async_checkpointer.py:76] [process=0][thread=async_save] Background save thread started. Deadline for this save operation is 2026-04-20 18:34:12.264492
I0420 18:14:12.312648 133010030769920 atomicity.py:140] Creating tmp directory gs://lance-maxtext/nnx_ckpt_xpk_feat_nnx_post_train_fixes_20260420_153543/nnx_xpk_feat_nnx_post_train_fixes_20260420_153543_13_scan_layers_false/checkpoints/9
I0420 18:14:12.325991 133136767887168 checkpoint_manager.py:1560] [process=0][thread=MainThread][step=9] Starting CheckpointManager Save Finalize thread=save_finalize
I0420 18:14:12.326350 133010517903104 async_checkpointer.py:280] [process=0][thread=save_finalize] Waiting for background save thread=async_save.
I0420 18:14:12.326521 133136767887168 standard_logger.py:34] {'step': 9, 'event_type': 'save', 'directory': 'gs://lance-maxtext/nnx_ckpt_xpk_feat_nnx_post_train_fixes_20260420_153543/nnx_xpk_feat_nnx_post_train_fixes_20260420_153543_13_scan_layers_false/checkpoints', 'reached_preemption': False, 'preemption_received_at': None, 'synchronous': False, 'wait_for_prev_start_time': 1776708831.1243367, 'wait_for_prev_duration_secs': 19.786349534988403, 'time_between_consecutive_saves_sec': None, 'checkpointer_blocking_start_time': 1776708850.913035, 'checkpointer_blocking_duration_secs': 1.35164213180542, 'get_old_steps_start_time': 1776708852.2647028, 'get_old_steps_duration_secs': 3.0279159545898438e-05, 'checkpoint_manager_blocking_start_time': 1776708831.1224759, 'checkpoint_manager_blocking_duration_secs': 21.204007863998413}
I0420 18:14:12.326713 133136767887168 checkpointing.py:409] Started an asynchronous checkpoint save for step 9
I0420 18:14:12.326762 133136767887168 checkpoint_manager.py:2020] [process=0][thread=MainThread][step=9][wait_until_finished] Waiting for Save Finalize thread (save_finalize) to complete.
I0420 18:14:14.429208 133009474832128 atomicity.py:140] Creating tmp directory gs://lance-maxtext/nnx_ckpt_xpk_feat_nnx_post_train_fixes_20260420_153543/nnx_xpk_feat_nnx_post_train_fixes_20260420_153543_13_scan_layers_false/checkpoints/9/items
I0420 18:14:18.632772 133010560870144 array_metadata_store.py:203] [process=0][thread=array_type_handler] Wrote 648 array_metadata.ArrayMetadata to gs://lance-maxtext/nnx_ckpt_xpk_feat_nnx_post_train_fixes_20260420_153543/nnx_xpk_feat_nnx_post_train_fixes_20260420_153543_13_scan_layers_false/checkpoints/9/items/array_metadatas/process_0
I0420 18:14:58.966774 133010530125568 base_pytree_checkpoint_handler.py:1282] [process=0][thread=write_metadata_after_commits] Commit + Array metadata written. Time taken: 43.369861s (commit=41.513943s, array_metadata_write=1.855918s)
I0420 18:14:58.968794 133009980413696 base_pytree_checkpoint_handler.py:130] [process=0] /jax/orbax/write/gbytes_per_sec: 33.674 MiB/s (total gbytes: 1.5 GiB) (time elapsed: 46.90855121612549 s) (per-host)
I0420 18:14:58.968917 133009980413696 async_checkpointer.py:90] [process=0][thread=async_save] 3 Handler Commit operations completed. Time taken: 46.704270s.
I0420 18:15:00.653718 133009980413696 array_metadata_store.py:411] [process=0][thread=async_save] Validated ArrayMetadata from all 8 hosts. Time taken: 0.002506s.
I0420 18:15:00.841979 133009980413696 ocdbt_utils.py:49] Param validation support for Zarr3 will be added later (b/362328389).
I0420 18:15:02.359116 133009980413696 base_pytree_checkpoint_handler.py:1406] [process=0][thread=async_save] Pytree save finalize (merge_ocdbt + ArrayMetadata validation) completed. Time taken: 2.560735s. use_zarr3=True, enable_post_merge_validation=True, directory=gs://lance-maxtext/nnx_ckpt_xpk_feat_nnx_post_train_fixes_20260420_153543/nnx_xpk_feat_nnx_post_train_fixes_20260420_153543_13_scan_layers_false/checkpoints/9/items
I0420 18:15:02.360700 133009980413696 atomicity.py:666] Finalizing gs://lance-maxtext/nnx_ckpt_xpk_feat_nnx_post_train_fixes_20260420_153543/nnx_xpk_feat_nnx_post_train_fixes_20260420_153543_13_scan_layers_false/checkpoints/9/items
I0420 18:15:02.908524 133009980413696 atomicity.py:666] Finalizing gs://lance-maxtext/nnx_ckpt_xpk_feat_nnx_post_train_fixes_20260420_153543/nnx_xpk_feat_nnx_post_train_fixes_20260420_153543_13_scan_layers_false/checkpoints/9
I0420 18:15:04.337706 133009980413696 atomicity.py:847] [process=0][thread=async_save] Finished saving checkpoint (finalized tmp dir) to `gs://lance-maxtext/nnx_ckpt_xpk_feat_nnx_post_train_fixes_20260420_153543/nnx_xpk_feat_nnx_post_train_fixes_20260420_153543_13_scan_layers_false/checkpoints/9`.
I0420 18:15:04.338420 133009980413696 event_tracking.py:138] [process=0] [async] Finished save (blocking + background) in 53.43 seconds @ gs://lance-maxtext/nnx_ckpt_xpk_feat_nnx_post_train_fixes_20260420_153543/nnx_xpk_feat_nnx_post_train_fixes_20260420_153543_13_scan_layers_false/checkpoints/9
I0420 18:15:04.339814 133009980413696 async_checkpointer.py:160] [process=0][thread=async_save] Background save thread done. Time taken: 52.075166s.
I0420 18:15:04.340008 133010517903104 async_checkpointer.py:288] [process=0][thread=save_finalize] Done with waiting for background save thread=async_save.
I0420 18:15:04.340133 133010517903104 async_checkpointer.py:298] [process=0][thread=save_finalize] No errors found in background save thread=async_save.
I0420 18:15:04.340198 133010517903104 checkpoint_manager.py:2137] [process=0][thread=save_finalize][step=9] CheckpointManager Save Finalize is syncing with other hosts...
I0420 18:15:04.342000 133010517903104 checkpoint_manager.py:2146] [process=0][thread=save_finalize][step=9] CheckpointManager Save Finalize is done on all hosts.
I0420 18:15:04.342184 133136767887168 checkpoint_manager.py:2032] [process=0][thread=MainThread][step=9][wait_until_finished] Done waiting for Save Finalize thread (save_finalize) running at step=9.
I0420 18:15:04.342454 133136767887168 checkpoint_manager.py:2009] [process=0][thread=MainThread][wait_until_finished] No Save Finalize thread to wait for. Returning.
I0420 18:15:04.343413 133136767887168 metric_logger.py:196] completed step: 9, seconds: 0.137, TFLOP/s/device: 99.037, Tokens/s/device: 14927.984, total_weights: 65536, loss: 8.182, lm_loss: 8.182, perplexity: 3574.857
Per train step:
 Total TFLOPs: 13.59 
 split as 93.93% learnable weight flops and 6.07% attention flops
XPK End: Mon Apr 20 18:15:14 UTC 2026
EXIT_CODE=0