MaxView

‹ 08_checkpoint_async_true_resumeCase: 08_checkpoint_async_true_save09_pdb_lt_1 ›

Metrics: main (586e69205) vs test/pipeline-scan-nnx (5b21b66fc)

Metricmain  586e69205test/pipeline-scan-nnx  5b21b66fcDiff (test/pipeline-scan-nnx − main)
Parameters1.104 billion1.104 billion
Final loss7.18107.18100
TFLOP/s92.21992.150-0.069
Tok/s13900.213889.9-10.276
Avg s/step8.0327.659-0.373
Memory %1.381.380
JAX0.9.20.9.2

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

main  ·  586e69205  ·  main_20260423_071538  ·  full log
XPK Start: Thu Apr 23 07:41:15 UTC 2026
PyTorch was not found. Models won't be available and only tokenizers, configuration and file/data utilities can be used.
`rope_parameters`'s factor field must be a float >= 1, got 40
`rope_parameters`'s beta_fast field must be a float, got 32
`rope_parameters`'s beta_slow field must be a float, got 1
DeepseekV32Config got `key=rope_scaling` in kwargs but hasn't set it as attribute. For RoPE standardization you need to set `self.rope_parameters` in model's config. 
2026-04-23 07:41:39.900351: 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)
I0423 07:41:40.112861 138875909756736 max_utils.py:273] Attempting to initialize the jax distributed system...
I0423 07:41:49.153935 138875909756736 distributed.py:149] Starting JAX distributed service on [::]:8482
I0423 07:41:49.156529 138875909756736 distributed.py:172] Connecting to JAX distributed service on mt-08-checkpoint-async--j2ilr-slice-job-0-0.mt-08-checkpoint-async--j2ilr:8482
I0423 07:41:50.348064 138875909756736 max_utils.py:284] Jax distributed system initialized!
I0423 07:41:56.468802 138875909756736 max_utils.py:800] System Information: Jax Version: 0.9.2
I0423 07:41:56.468911 138875909756736 max_utils.py:801] System Information: Jaxlib Version: 0.9.2
I0423 07:41:56.468952 138875909756736 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
I0423 07:41:56.468989 138875909756736 train_utils.py:361] WARNING: Sequence packing is essentially ignored for synthetic data. Please use a real dataset to use sequence packing.
I0423 07:41:57.166580 138875909756736 maxtext_utils.py:1604] Num_devices: 32, shape (1, 1, 1, 32, 1, 1, 1, 1, 1, 1, 1, 1, 1)
I0423 07:41:57.166879 138875909756736 checkpointing.py:677] Setting up checkpoint logger...
I0423 07:41:57.166936 138875909756736 checkpointing.py:233] Creating checkpoint manager with ocdbt=True and zarr3=True
I0423 07:41:57.166981 138875909756736 pytree_checkpoint_handler.py:592] save_device_host_concurrent_bytes=None
I0423 07:41:57.167326 138875909756736 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 0x7e4dd9b3c5f0>, enable_pinned_host_transfer=False, save_concurrent_bytes: 96000000000 (89.4 GiB), restore_concurrent_bytes: 96000000000 (89.4 GiB)
I0423 07:42:00.360559 138875909756736 checkpointing.py:265] Enabling policy for fixed interval checkpointing.
I0423 07:42:00.360805 138875909756736 checkpoint_manager.py:708] [process=5][thread=MainThread] CheckpointManager init: checkpointers=None, item_names=('items',), item_handlers={'items': <orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeCheckpointHandler object at 0x7e39a4579d90>}, handler_registry=None
I0423 07:42:00.361050 138875909756736 composite_checkpoint_handler.py:237] Deferred registration for item: "items". Adding handler `<orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeCheckpointHandler object at 0x7e39a4579d90>` 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`.
I0423 07:42:00.361101 138875909756736 composite_checkpoint_handler.py:237] Deferred registration for item: "metrics". Adding handler `<orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonCheckpointHandler object at 0x7e39a457b530>` 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`.
I0423 07:42:00.361138 138875909756736 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 0x7e39a4579d90>, ('items', <class 'orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeRestoreArgs'>): <orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeCheckpointHandler object at 0x7e39a4579d90>, ('metrics', <class 'orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonSaveArgs'>): <orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonCheckpointHandler object at 0x7e39a457b530>, ('metrics', <class 'orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonRestoreArgs'>): <orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonCheckpointHandler object at 0x7e39a457b530>}).
I0423 07:42:00.361466 138875909756736 abstract_checkpointer.py:35] orbax-checkpoint version: 0.11.34
I0423 07:42:00.361539 138875909756736 async_checkpointer.py:192] [process=5][thread=MainThread] Using barrier_sync_fn: <function get_barrier_sync_fn.<locals>._fn at 0x7e392418d580> timeout: 1200 secs and primary_host=0 for async checkpoint writes
I0423 07:42:01.912228 138875909756736 checkpoint_manager.py:1812] Found 0 checkpoint steps in gs://lance-maxtext/linen_ckpt_xpk_main_20260423_071538/linen_xpk_main_20260423_071538_08_checkpoint_async_true/checkpoints
I0423 07:42:02.316749 138875909756736 checkpoint_manager.py:929] [process=5][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=5), 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/linen_ckpt_xpk_main_20260423_071538/linen_xpk_main_20260423_071538_08_checkpoint_async_true/checkpoints: <orbax.checkpoint.checkpoint_manager.CheckpointManager object at 0x7e39a457a3c0>
I0423 07:42:02.316922 138875909756736 checkpointing.py:301] Checkpoint manager created!
I0423 07:42:03.248366 138875909756736 nnx_wrappers.py:437] Unknown Logical: bfloat16[32,2048,2048]...................................... ('activation_batch', 'activation_norm_length', 'activation_embed').
I0423 07:42:03.248480 138875909756736 nnx_wrappers.py:437] Unknown Physical: bfloat16[32,2048,2048]...................................... ('fsdp', None, None).
I0423 07:42:03.628087 138875909756736 attentions.py:1088] attentions/inputs_q Logical: bfloat16[32,2048,2048]...................................... ('activation_batch', 'activation_attn_length', 'activation_attn_embed').
I0423 07:42:03.628178 138875909756736 attentions.py:1088] attentions/inputs_q Physical: bfloat16[32,2048,2048]...................................... ('fsdp', None, None).
I0423 07:42:03.644791 138875909756736 attentions.py:1089] attentions/inputs_kv Logical: bfloat16[32,2048,2048]...................................... ('activation_batch', 'activation_attn_length', 'activation_attn_embed').
I0423 07:42:03.644848 138875909756736 attentions.py:1089] attentions/inputs_kv Physical: bfloat16[32,2048,2048]...................................... ('fsdp', None, None).
I0423 07:42:03.668772 138875909756736 attentions.py:1154] attentions/query Logical: bfloat16[32,2048,16,128].................................... ('activation_kv_batch', 'activation_attn_length', 'activation_kv_heads', 'activation_kv_head_dim').
I0423 07:42:03.668836 138875909756736 attentions.py:1154] attentions/query Physical: bfloat16[32,2048,16,128].................................... ('fsdp', None, None, None).
I0423 07:42:03.685406 138875909756736 attentions.py:1155] attentions/key Logical: bfloat16[32,2048,16,128].................................... ('activation_kv_batch', 'activation_attn_length', 'activation_kv_heads', 'activation_kv_head_dim').
I0423 07:42:03.685467 138875909756736 attentions.py:1155] attentions/key Physical: bfloat16[32,2048,16,128].................................... ('fsdp', None, None, None).
I0423 07:42:03.701905 138875909756736 attentions.py:1156] attentions/value Logical: bfloat16[32,2048,16,128].................................... ('activation_kv_batch', 'activation_attn_length', 'activation_kv_heads', 'activation_kv_head_dim').
I0423 07:42:03.701970 138875909756736 attentions.py:1156] attentions/value Physical: bfloat16[32,2048,16,128].................................... ('fsdp', None, None, None).
I0423 07:42:03.726828 138875909756736 attentions.py:1198] attentions/out Logical: bfloat16[32,2048,16,128].................................... ('activation_batch', 'activation_attn_length', 'activation_heads', 'activation_kv').
I0423 07:42:03.726897 138875909756736 attentions.py:1198] attentions/out Physical: bfloat16[32,2048,16,128].................................... ('fsdp', None, None, None).
I0423 07:42:03.748562 138875909756736 linears.py:525] linears/x Logical: bfloat16[32,2048,7168]...................................... ('activation_batch', 'activation_length', 'activation_mlp').
I0423 07:42:03.748630 138875909756736 linears.py:525] linears/x Physical: bfloat16[32,2048,7168]...................................... ('fsdp', None, None).
I0423 07:42:03.960738 138875909756736 checkpointing.py:577] checkpoint manager exists so trying to load this run's existing checkpoint
I0423 07:42:03.960850 138875909756736 checkpointing.py:665] No existing checkpoints found, not restoring checkpoint.
fsdp: 32
I0423 07:42:05.404704 138875909756736 maxtext_utils.py:1707]  params/params/decoder/decoder_norm/scale
    Shape:     float32[2048]
    Logical:   P('norm',)
    Physical:  (None,)
I0423 07:42:05.404838 138875909756736 maxtext_utils.py:1707]  params/params/decoder/layers/mlp/wi_0/kernel
    Shape:     float32[2048,16,7168]
    Logical:   P('embed', 'layers', 'mlp')
    Physical:  ('fsdp', None, None)
I0423 07:42:05.404891 138875909756736 maxtext_utils.py:1707]  params/params/decoder/layers/mlp/wi_1/kernel
    Shape:     float32[2048,16,7168]
    Logical:   P('embed', 'layers', 'mlp')
    Physical:  ('fsdp', None, None)
I0423 07:42:05.404947 138875909756736 maxtext_utils.py:1707]  params/params/decoder/layers/mlp/wo/kernel
    Shape:     float32[7168,16,2048]
    Logical:   P('mlp', 'layers', 'embed')
    Physical:  (None, None, 'fsdp')
I0423 07:42:05.404998 138875909756736 maxtext_utils.py:1707]  params/params/decoder/layers/post_self_attention_layer_norm/scale
    Shape:     float32[2048,16]
    Logical:   P('norm', 'layers')
    Physical:  (None, None)
I0423 07:42:05.405038 138875909756736 maxtext_utils.py:1707]  params/params/decoder/layers/pre_self_attention_layer_norm/scale
    Shape:     float32[2048,16]
    Logical:   P('norm', 'layers')
    Physical:  (None, None)
I0423 07:42:05.405089 138875909756736 maxtext_utils.py:1707]  params/params/decoder/layers/self_attention/key/kernel
    Shape:     float32[2048,16,16,128]
    Logical:   P('embed', 'layers', 'kv_heads', 'kv_head_dim')
    Physical:  ('fsdp', None, None, None)
I0423 07:42:05.405138 138875909756736 maxtext_utils.py:1707]  params/params/decoder/layers/self_attention/out/kernel
    Shape:     float32[16,16,128,2048]
    Logical:   P('heads', 'layers', 'kv', 'embed')
    Physical:  (None, None, None, 'fsdp')
I0423 07:42:05.405177 138875909756736 maxtext_utils.py:1707]  params/params/decoder/layers/self_attention/query/kernel
    Shape:     float32[2048,16,16,128]
    Logical:   P('embed', 'layers', 'q_heads', 'kv')
    Physical:  ('fsdp', None, None, None)
I0423 07:42:05.405218 138875909756736 maxtext_utils.py:1707]  params/params/decoder/layers/self_attention/value/kernel
    Shape:     float32[2048,16,16,128]
    Logical:   P('embed', 'layers', 'kv_heads', 'kv_head_dim')
    Physical:  ('fsdp', None, None, None)
I0423 07:42:05.405266 138875909756736 maxtext_utils.py:1707]  params/params/decoder/logits_dense/kernel
    Shape:     float32[2048,32000]
    Logical:   P('embed_vocab', 'vocab')
    Physical:  ('fsdp', None)
I0423 07:42:05.405313 138875909756736 maxtext_utils.py:1707]  params/params/token_embedder/embedding
    Shape:     float32[32000,2048]
    Logical:   P('vocab', 'embed_vocab')
    Physical:  (None, 'fsdp')

I0423 07:42:05.900256 138875909756736 train.py:155] train/xent Logical: float32[32,2048]............................................ ('activation_embed_and_logits_batch', 'activation_length').
I0423 07:42:05.900353 138875909756736 train.py:155] train/xent Physical: float32[32,2048]............................................ ('fsdp', None).
I0423 07:42:05.916827 138875909756736 train.py:162] train/z_loss Logical: float32[32,2048]............................................ ('activation_embed_and_logits_batch', 'activation_length').
I0423 07:42:05.916896 138875909756736 train.py:162] train/z_loss Physical: float32[32,2048]............................................ ('fsdp', None).
I0423 07:42:16.874771 138875909756736 max_utils.py:791] Total memory size: 1.5 GB, Output size: 0.4 GB, Temp size: 1.1 GB, Argument size: 0.4 GB, Host temp size: 0.0 GB.
I0423 07:42:16.875586 138875909756736 metric_logger.py:301] number parameters: 1.104 billion
I0423 07:42:28.067526 138875909756736 checkpointing.py:772] Waiting for step 0 to finish before checkpoint...
I0423 07:42:28.222130 138875909756736 checkpointing.py:776] Waited 0.15458011627197266 seconds for step 0 to finish before starting checkpointing.
I0423 07:42:28.224519 138875909756736 checkpoint_manager.py:2009] [process=5][thread=MainThread][wait_until_finished] No Save Finalize thread to wait for. Returning.
I0423 07:42:28.226288 138875909756736 checkpoint_manager.py:1512] [process=5] Saving checkpoint at step 0
I0423 07:42:28.227676 138875909756736 event_tracking.py:70] [process=5] [async] Started save checkpoint @ gs://lance-maxtext/linen_ckpt_xpk_main_20260423_071538/linen_xpk_main_20260423_071538_08_checkpoint_async_true/checkpoints/0.
I0423 07:42:28.972394 138875909756736 signaling_client.py:364] Using JaxDistributedSignalingClient
I0423 07:42:28.973414 138875909756736 jax_array_handlers.py:360] Scheduling D2H of 39 prioritized jax.Array.
I0423 07:42:28.973471 138875909756736 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
I0423 07:42:29.246111 138875909756736 base_pytree_checkpoint_handler.py:154] [process=5][thread=MainThread] Initiated "orbax.checkpoint._src.serialization.jax_array_handlers.ArrayHandler".serialize. Time taken: 0.273787s
I0423 07:42:29.246287 138875909756736 base_pytree_checkpoint_handler.py:130] [process=5] /jax/orbax/write/blocking_gbytes_per_sec: 5.518 GiB/s (total gbytes: 1.5 GiB) (time elapsed: 0.27956485748291016 s) (per-host)
I0423 07:42:29.246341 138875909756736 base_pytree_checkpoint_handler.py:768] [process=5][thread=MainThread] Initiated Pytree async_save. Time taken: 0.279630s (batch_requests_ready=0.002615s, total_serialization_initiated=0.276941s, others=0.000074s)
I0423 07:42:29.246432 138875909756736 composite_checkpoint_handler.py:715] [process=5][thread=MainThread] Initiated CompositeCheckpointHandler.async_save. Time taken: 0.283833s (all_items=0.000018s, per_item={'items': '0.00001788'}, temp_paths=0.283815)
I0423 07:42:29.247247 138875909756736 event_tracking.py:125] [process=5] [async] Finished blocking save in 1.02 seconds. Continuing save @ gs://lance-maxtext/linen_ckpt_xpk_main_20260423_071538/linen_xpk_main_20260423_071538_08_checkpoint_async_true/checkpoints/0.
I0423 07:42:29.247573 138747860731648 async_checkpointer.py:76] [process=5][thread=async_save] Background save thread started. Deadline for this save operation is 2026-04-23 08:02:29.247535
I0423 07:42:29.258000 138875909756736 checkpoint_manager.py:1560] [process=5][thread=MainThread][step=0] Starting CheckpointManager Save Finalize thread=save_finalize
I0423 07:42:29.258304 138746808755968 async_checkpointer.py:280] [process=5][thread=save_finalize] Waiting for background save thread=async_save.
I0423 07:42:29.258455 138875909756736 standard_logger.py:34] {'step': 0, 'event_type': 'save', 'directory': 'gs://lance-maxtext/linen_ckpt_xpk_main_20260423_071538/linen_xpk_main_20260423_071538_08_checkpoint_async_true/checkpoints', 'reached_preemption': False, 'preemption_received_at': None, 'synchronous': False, 'wait_for_prev_start_time': 1776930148.2245016, 'wait_for_prev_duration_secs': 5.984306335449219e-05, 'time_between_consecutive_saves_sec': None, 'checkpointer_blocking_start_time': 1776930148.2263272, 'checkpointer_blocking_duration_secs': 1.021355152130127, 'get_old_steps_start_time': 1776930149.2477083, 'get_old_steps_duration_secs': 3.0994415283203125e-05, 'checkpoint_manager_blocking_start_time': 1776930148.2226734, 'checkpoint_manager_blocking_duration_secs': 1.0357434749603271}
I0423 07:42:29.258561 138875909756736 checkpointing.py:408] Started an asynchronous checkpoint save for step 0
I0423 07:42:29.258613 138875909756736 max_utils.py:750] 
Memstats: After params initialized:
I0423 07:42:29.258675 138875909756736 max_utils.py:756] 	Using (GB) 0.43 / 31.25 (1.376000%) on TPU_18(process=5,(2,4,0,0))
I0423 07:42:29.258711 138875909756736 max_utils.py:756] 	Using (GB) 0.43 / 31.25 (1.376000%) on TPU_19(process=5,(3,4,0,0))
I0423 07:42:29.258739 138875909756736 max_utils.py:756] 	Using (GB) 0.43 / 31.25 (1.376000%) on TPU_22(process=5,(2,5,0,0))
I0423 07:42:29.258766 138875909756736 max_utils.py:756] 	Using (GB) 0.43 / 31.25 (1.376000%) on TPU_23(process=5,(3,5,0,0))
I0423 07:42:29.569679 138875909756736 metric_logger.py:196] completed step: 0, seconds: 11.192, TFLOP/s/device: 1.214, Tokens/s/device: 182.991, total_weights: 65536, loss: 10.874, lm_loss: 10.874, perplexity: 52779.875
I0423 07:42:29.747335 138875909756736 metric_logger.py:196] completed step: 1, seconds: 1.501, TFLOP/s/device: 9.054, Tokens/s/device: 1364.749, total_weights: 65536, loss: 10.874, lm_loss: 10.874, perplexity: 52779.875
I0423 07:42:30.137381 138875909756736 metric_logger.py:196] completed step: 2, seconds: 0.030, TFLOP/s/device: 446.695, Tokens/s/device: 67330.769, total_weights: 65536, loss: 10.560, lm_loss: 10.560, perplexity: 38547.719
I0423 07:42:30.284742 138875909756736 metric_logger.py:196] completed step: 3, seconds: 0.390, TFLOP/s/device: 34.841, Tokens/s/device: 5251.646, total_weights: 65536, loss: 9.980, lm_loss: 9.980, perplexity: 21590.506
I0423 07:42:30.289165 138875909756736 checkpointing.py:772] Waiting for step 5 to finish before checkpoint...
I0423 07:42:30.579004 138875909756736 checkpointing.py:776] Waited 0.289792537689209 seconds for step 5 to finish before starting checkpointing.
I0423 07:42:30.581923 138875909756736 checkpoint_manager.py:2020] [process=5][thread=MainThread][step=0][wait_until_finished] Waiting for Save Finalize thread (save_finalize) to complete.
I0423 07:42:32.595059    2814 google_auth_provider.cc:181] Running on GCE, using service account 562977990677-compute@developer.gserviceaccount.com
I0423 07:42:35.552093 138747318155008 array_metadata_store.py:203] [process=5][thread=array_type_handler] Wrote 39 array_metadata.ArrayMetadata to gs://lance-maxtext/linen_ckpt_xpk_main_20260423_071538/linen_xpk_main_20260423_071538_08_checkpoint_async_true/checkpoints/0/items/array_metadatas/process_5
I0423 07:43:05.752702 138747860731648 base_pytree_checkpoint_handler.py:130] [process=5] /jax/orbax/write/gbytes_per_sec: 42.940 MiB/s (total gbytes: 1.5 GiB) (time elapsed: 36.78593039512634 s) (per-host)
I0423 07:43:05.752829 138747860731648 async_checkpointer.py:90] [process=5][thread=async_save] 3 Handler Commit operations completed. Time taken: 36.504611s.
I0423 07:43:16.247272 138747860731648 async_checkpointer.py:160] [process=5][thread=async_save] Background save thread done. Time taken: 46.999028s.
I0423 07:43:16.247534 138746808755968 async_checkpointer.py:288] [process=5][thread=save_finalize] Done with waiting for background save thread=async_save.
I0423 07:43:16.247667 138746808755968 async_checkpointer.py:298] [process=5][thread=save_finalize] No errors found in background save thread=async_save.
I0423 07:43:16.247732 138746808755968 checkpoint_manager.py:2137] [process=5][thread=save_finalize][step=0] CheckpointManager Save Finalize is syncing with other hosts...
I0423 07:43:16.249551 138746808755968 checkpoint_manager.py:2146] [process=5][thread=save_finalize][step=0] CheckpointManager Save Finalize is done on all hosts.
I0423 07:43:16.249762 138875909756736 checkpoint_manager.py:2032] [process=5][thread=MainThread][step=0][wait_until_finished] Done waiting for Save Finalize thread (save_finalize) running at step=0.
W0423 07:43:16.249902 138875909756736 checkpoint_manager.py:1452] Waiting for previous save to complete took 45.667977 seconds. If this number is high, consider checkpointing less frequently.
I0423 07:43:16.251707 138875909756736 checkpoint_manager.py:1512] [process=5] Saving checkpoint at step 5
I0423 07:43:16.254156 138875909756736 event_tracking.py:70] [process=5] [async] Started save checkpoint @ gs://lance-maxtext/linen_ckpt_xpk_main_20260423_071538/linen_xpk_main_20260423_071538_08_checkpoint_async_true/checkpoints/5.
I0423 07:43:16.965642 138875909756736 jax_array_handlers.py:360] Scheduling D2H of 39 prioritized jax.Array.
I0423 07:43:16.965748 138875909756736 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
I0423 07:43:17.001144 138875909756736 base_pytree_checkpoint_handler.py:154] [process=5][thread=MainThread] Initiated "orbax.checkpoint._src.serialization.jax_array_handlers.ArrayHandler".serialize. Time taken: 0.036613s
I0423 07:43:17.001311 138875909756736 base_pytree_checkpoint_handler.py:130] [process=5] /jax/orbax/write/blocking_gbytes_per_sec: 38.432 GiB/s (total gbytes: 1.5 GiB) (time elapsed: 0.040137529373168945 s) (per-host)
I0423 07:43:17.001363 138875909756736 base_pytree_checkpoint_handler.py:768] [process=5][thread=MainThread] Initiated Pytree async_save. Time taken: 0.040198s (batch_requests_ready=0.001883s, total_serialization_initiated=0.038250s, others=0.000066s)
I0423 07:43:17.001456 138875909756736 composite_checkpoint_handler.py:715] [process=5][thread=MainThread] Initiated CompositeCheckpointHandler.async_save. Time taken: 0.045178s (all_items=0.000015s, per_item={'items': '0.00001502'}, temp_paths=0.045163)
I0423 07:43:17.002158 138875909756736 event_tracking.py:125] [process=5] [async] Finished blocking save in 0.75 seconds. Continuing save @ gs://lance-maxtext/linen_ckpt_xpk_main_20260423_071538/linen_xpk_main_20260423_071538_08_checkpoint_async_true/checkpoints/5.
I0423 07:43:17.002448 138746808755968 async_checkpointer.py:76] [process=5][thread=async_save] Background save thread started. Deadline for this save operation is 2026-04-23 08:03:17.002415
I0423 07:43:17.009958 138875909756736 checkpoint_manager.py:1560] [process=5][thread=MainThread][step=5] Starting CheckpointManager Save Finalize thread=save_finalize
I0423 07:43:17.010257 138722133665536 async_checkpointer.py:280] [process=5][thread=save_finalize] Waiting for background save thread=async_save.
I0423 07:43:17.010422 138875909756736 standard_logger.py:34] {'step': 5, 'event_type': 'save', 'directory': 'gs://lance-maxtext/linen_ckpt_xpk_main_20260423_071538/linen_xpk_main_20260423_071538_08_checkpoint_async_true/checkpoints', 'reached_preemption': False, 'preemption_received_at': None, 'synchronous': False, 'wait_for_prev_start_time': 1776930150.5818937, 'wait_for_prev_duration_secs': 45.66797733306885, 'time_between_consecutive_saves_sec': None, 'checkpointer_blocking_start_time': 1776930196.251744, 'checkpointer_blocking_duration_secs': 0.7508463859558105, 'get_old_steps_start_time': 1776930197.002614, 'get_old_steps_duration_secs': 2.8848648071289062e-05, 'checkpoint_manager_blocking_start_time': 1776930150.5796425, 'checkpoint_manager_blocking_duration_secs': 46.43074083328247}
I0423 07:43:17.010538 138875909756736 checkpointing.py:408] Started an asynchronous checkpoint save for step 5
I0423 07:43:17.011364 138875909756736 metric_logger.py:196] completed step: 4, seconds: 0.153, TFLOP/s/device: 88.660, Tokens/s/device: 13363.785, total_weights: 65536, loss: 9.460, lm_loss: 9.460, perplexity: 12840.355
I0423 07:43:17.017503 138875909756736 metric_logger.py:196] completed step: 5, seconds: 0.147, TFLOP/s/device: 92.274, Tokens/s/device: 13908.603, total_weights: 65536, loss: 8.959, lm_loss: 8.959, perplexity: 7776.467
I0423 07:43:22.288951 138747318155008 array_metadata_store.py:203] [process=5][thread=array_type_handler] Wrote 39 array_metadata.ArrayMetadata to gs://lance-maxtext/linen_ckpt_xpk_main_20260423_071538/linen_xpk_main_20260423_071538_08_checkpoint_async_true/checkpoints/5/items/array_metadatas/process_5
I0423 07:43:40.204716 138875909756736 metric_logger.py:196] completed step: 6, seconds: 46.727, TFLOP/s/device: 0.291, Tokens/s/device: 43.829, total_weights: 65536, loss: 8.469, lm_loss: 8.469, perplexity: 4762.740
I0423 07:43:40.352141 138875909756736 metric_logger.py:196] completed step: 7, seconds: 23.041, TFLOP/s/device: 0.590, Tokens/s/device: 88.884, total_weights: 65536, loss: 8.003, lm_loss: 8.003, perplexity: 2989.151
I0423 07:43:40.499492 138875909756736 metric_logger.py:196] completed step: 8, seconds: 0.152, TFLOP/s/device: 89.235, Tokens/s/device: 13450.500, total_weights: 65536, loss: 7.572, lm_loss: 7.572, perplexity: 1943.098
I0423 07:43:40.646207 138875909756736 checkpointing.py:772] Waiting for step 9 to finish before checkpoint...
I0423 07:43:40.646923 138875909756736 checkpointing.py:776] Waited 0.0007386207580566406 seconds for step 9 to finish before starting checkpointing.
I0423 07:43:40.649009 138875909756736 checkpoint_manager.py:2020] [process=5][thread=MainThread][step=5][wait_until_finished] Waiting for Save Finalize thread (save_finalize) to complete.
I0423 07:43:58.684573 138746808755968 base_pytree_checkpoint_handler.py:130] [process=5] /jax/orbax/write/gbytes_per_sec: 37.859 MiB/s (total gbytes: 1.5 GiB) (time elapsed: 41.72335863113403 s) (per-host)
I0423 07:43:58.684711 138746808755968 async_checkpointer.py:90] [process=5][thread=async_save] 3 Handler Commit operations completed. Time taken: 41.682153s.
I0423 07:44:07.629317 138746808755968 async_checkpointer.py:160] [process=5][thread=async_save] Background save thread done. Time taken: 50.626743s.
I0423 07:44:07.629608 138722133665536 async_checkpointer.py:288] [process=5][thread=save_finalize] Done with waiting for background save thread=async_save.
I0423 07:44:07.629752 138722133665536 async_checkpointer.py:298] [process=5][thread=save_finalize] No errors found in background save thread=async_save.
I0423 07:44:07.629804 138722133665536 checkpoint_manager.py:2137] [process=5][thread=save_finalize][step=5] CheckpointManager Save Finalize is syncing with other hosts...
I0423 07:44:07.631323 138722133665536 checkpoint_manager.py:2146] [process=5][thread=save_finalize][step=5] CheckpointManager Save Finalize is done on all hosts.
I0423 07:44:07.631491 138875909756736 checkpoint_manager.py:2032] [process=5][thread=MainThread][step=5][wait_until_finished] Done waiting for Save Finalize thread (save_finalize) running at step=5.
W0423 07:44:07.631616 138875909756736 checkpoint_manager.py:1452] Waiting for previous save to complete took 26.982614 seconds. If this number is high, consider checkpointing less frequently.
I0423 07:44:07.633059 138875909756736 checkpoint_manager.py:1512] [process=5] Saving checkpoint at step 9
I0423 07:44:07.635217 138875909756736 event_tracking.py:70] [process=5] [async] Started save checkpoint @ gs://lance-maxtext/linen_ckpt_xpk_main_20260423_071538/linen_xpk_main_20260423_071538_08_checkpoint_async_true/checkpoints/9.
I0423 07:44:08.343324 138875909756736 jax_array_handlers.py:360] Scheduling D2H of 39 prioritized jax.Array.
I0423 07:44:08.343420 138875909756736 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
I0423 07:44:08.378567 138875909756736 base_pytree_checkpoint_handler.py:154] [process=5][thread=MainThread] Initiated "orbax.checkpoint._src.serialization.jax_array_handlers.ArrayHandler".serialize. Time taken: 0.036123s
I0423 07:44:08.378751 138875909756736 base_pytree_checkpoint_handler.py:130] [process=5] /jax/orbax/write/blocking_gbytes_per_sec: 38.726 GiB/s (total gbytes: 1.5 GiB) (time elapsed: 0.03983354568481445 s) (per-host)
I0423 07:44:08.378806 138875909756736 base_pytree_checkpoint_handler.py:768] [process=5][thread=MainThread] Initiated Pytree async_save. Time taken: 0.039899s (batch_requests_ready=0.001838s, total_serialization_initiated=0.037990s, others=0.000070s)
I0423 07:44:08.378897 138875909756736 composite_checkpoint_handler.py:715] [process=5][thread=MainThread] Initiated CompositeCheckpointHandler.async_save. Time taken: 0.044271s (all_items=0.000017s, per_item={'items': '0.00001669'}, temp_paths=0.044255)
I0423 07:44:08.379626 138875909756736 event_tracking.py:125] [process=5] [async] Finished blocking save in 0.75 seconds. Continuing save @ gs://lance-maxtext/linen_ckpt_xpk_main_20260423_071538/linen_xpk_main_20260423_071538_08_checkpoint_async_true/checkpoints/9.
I0423 07:44:08.379940 138722133665536 async_checkpointer.py:76] [process=5][thread=async_save] Background save thread started. Deadline for this save operation is 2026-04-23 08:04:08.379907
I0423 07:44:08.381875 138875909756736 checkpoint_manager.py:1560] [process=5][thread=MainThread][step=9] Starting CheckpointManager Save Finalize thread=save_finalize
I0423 07:44:08.382177 138747318155008 async_checkpointer.py:280] [process=5][thread=save_finalize] Waiting for background save thread=async_save.
I0423 07:44:08.382336 138875909756736 standard_logger.py:34] {'step': 9, 'event_type': 'save', 'directory': 'gs://lance-maxtext/linen_ckpt_xpk_main_20260423_071538/linen_xpk_main_20260423_071538_08_checkpoint_async_true/checkpoints', 'reached_preemption': False, 'preemption_received_at': None, 'synchronous': False, 'wait_for_prev_start_time': 1776930220.6489778, 'wait_for_prev_duration_secs': 26.982613801956177, 'time_between_consecutive_saves_sec': 24.399385452270508, 'checkpointer_blocking_start_time': 1776930247.6330974, 'checkpointer_blocking_duration_secs': 0.7469871044158936, 'get_old_steps_start_time': 1776930248.3801017, 'get_old_steps_duration_secs': 2.5272369384765625e-05, 'checkpoint_manager_blocking_start_time': 1776930220.6471543, 'checkpoint_manager_blocking_duration_secs': 27.735148429870605}
I0423 07:44:08.382457 138875909756736 checkpointing.py:408] Started an asynchronous checkpoint save for step 9
I0423 07:44:08.382502 138875909756736 checkpoint_manager.py:2020] [process=5][thread=MainThread][step=9][wait_until_finished] Waiting for Save Finalize thread (save_finalize) to complete.
I0423 07:44:14.530970 138743080023808 array_metadata_store.py:203] [process=5][thread=array_type_handler] Wrote 39 array_metadata.ArrayMetadata to gs://lance-maxtext/linen_ckpt_xpk_main_20260423_071538/linen_xpk_main_20260423_071538_08_checkpoint_async_true/checkpoints/9/items/array_metadatas/process_5
I0423 07:44:50.429761 138722133665536 base_pytree_checkpoint_handler.py:130] [process=5] /jax/orbax/write/gbytes_per_sec: 37.528 MiB/s (total gbytes: 1.5 GiB) (time elapsed: 42.09081411361694 s) (per-host)
I0423 07:44:50.429863 138722133665536 async_checkpointer.py:90] [process=5][thread=async_save] 3 Handler Commit operations completed. Time taken: 42.049812s.
I0423 07:44:59.908580 138722133665536 async_checkpointer.py:160] [process=5][thread=async_save] Background save thread done. Time taken: 51.528512s.
I0423 07:44:59.908873 138747318155008 async_checkpointer.py:288] [process=5][thread=save_finalize] Done with waiting for background save thread=async_save.
I0423 07:44:59.909100 138747318155008 async_checkpointer.py:298] [process=5][thread=save_finalize] No errors found in background save thread=async_save.
I0423 07:44:59.909165 138747318155008 checkpoint_manager.py:2137] [process=5][thread=save_finalize][step=9] CheckpointManager Save Finalize is syncing with other hosts...
I0423 07:44:59.910493 138747318155008 checkpoint_manager.py:2146] [process=5][thread=save_finalize][step=9] CheckpointManager Save Finalize is done on all hosts.
I0423 07:44:59.910673 138875909756736 checkpoint_manager.py:2032] [process=5][thread=MainThread][step=9][wait_until_finished] Done waiting for Save Finalize thread (save_finalize) running at step=9.
I0423 07:44:59.910822 138875909756736 checkpoint_manager.py:2009] [process=5][thread=MainThread][wait_until_finished] No Save Finalize thread to wait for. Returning.
I0423 07:44:59.911701 138875909756736 metric_logger.py:196] completed step: 9, seconds: 0.147, TFLOP/s/device: 92.219, Tokens/s/device: 13900.201, total_weights: 65536, loss: 7.181, lm_loss: 7.181, perplexity: 1314.495
Per train step:
 Total TFLOPs: 13.59 
 split as 93.93% learnable weight flops and 6.07% attention flops
XPK End: Thu Apr 23 07:45:10 UTC 2026
EXIT_CODE=0
XPK Start: Thu Apr 23 21:37:48 UTC 2026
PyTorch was not found. Models won't be available and only tokenizers, configuration and file/data utilities can be used.
`rope_parameters`'s factor field must be a float >= 1, got 40
`rope_parameters`'s beta_fast field must be a float, got 32
`rope_parameters`'s beta_slow field must be a float, got 1
DeepseekV32Config got `key=rope_scaling` in kwargs but hasn't set it as attribute. For RoPE standardization you need to set `self.rope_parameters` in model's config. 
2026-04-23 21:38:13.727879: 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)
I0423 21:38:13.941118 136838485276480 max_utils.py:273] Attempting to initialize the jax distributed system...
I0423 21:38:22.980386 136838485276480 distributed.py:149] Starting JAX distributed service on [::]:8482
I0423 21:38:22.982725 136838485276480 distributed.py:172] Connecting to JAX distributed service on mt-08-checkpoint-async--mkctd-slice-job-0-0.mt-08-checkpoint-async--mkctd:8482
I0423 21:38:24.185115 136838485276480 max_utils.py:284] Jax distributed system initialized!
I0423 21:38:30.265588 136838485276480 max_utils.py:800] System Information: Jax Version: 0.9.2
I0423 21:38:30.265704 136838485276480 max_utils.py:801] System Information: Jaxlib Version: 0.9.2
I0423 21:38:30.265745 136838485276480 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
I0423 21:38:30.265782 136838485276480 train_utils.py:361] WARNING: Sequence packing is essentially ignored for synthetic data. Please use a real dataset to use sequence packing.
I0423 21:38:30.956221 136838485276480 maxtext_utils.py:1565] Num_devices: 32, shape (1, 1, 1, 32, 1, 1, 1, 1, 1, 1, 1, 1, 1)
I0423 21:38:30.956498 136838485276480 checkpointing.py:677] Setting up checkpoint logger...
I0423 21:38:30.956552 136838485276480 checkpointing.py:233] Creating checkpoint manager with ocdbt=True and zarr3=True
I0423 21:38:30.956596 136838485276480 pytree_checkpoint_handler.py:592] save_device_host_concurrent_bytes=None
I0423 21:38:30.956964 136838485276480 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 0x7c7379d3cda0>, enable_pinned_host_transfer=False, save_concurrent_bytes: 96000000000 (89.4 GiB), restore_concurrent_bytes: 96000000000 (89.4 GiB)
I0423 21:38:33.745365 136838485276480 checkpointing.py:265] Enabling policy for fixed interval checkpointing.
I0423 21:38:33.745604 136838485276480 checkpoint_manager.py:708] [process=5][thread=MainThread] CheckpointManager init: checkpointers=None, item_names=('items',), item_handlers={'items': <orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeCheckpointHandler object at 0x7c5e8456b680>}, handler_registry=None
I0423 21:38:33.745960 136838485276480 composite_checkpoint_handler.py:237] Deferred registration for item: "items". Adding handler `<orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeCheckpointHandler object at 0x7c5e8456b680>` 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`.
I0423 21:38:33.746112 136838485276480 composite_checkpoint_handler.py:237] Deferred registration for item: "metrics". Adding handler `<orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonCheckpointHandler object at 0x7c5e8456da30>` 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`.
I0423 21:38:33.746160 136838485276480 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 0x7c5e8456b680>, ('items', <class 'orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeRestoreArgs'>): <orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeCheckpointHandler object at 0x7c5e8456b680>, ('metrics', <class 'orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonSaveArgs'>): <orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonCheckpointHandler object at 0x7c5e8456da30>, ('metrics', <class 'orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonRestoreArgs'>): <orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonCheckpointHandler object at 0x7c5e8456da30>}).
I0423 21:38:33.746524 136838485276480 abstract_checkpointer.py:35] orbax-checkpoint version: 0.11.34
I0423 21:38:33.746613 136838485276480 async_checkpointer.py:192] [process=5][thread=MainThread] Using barrier_sync_fn: <function get_barrier_sync_fn.<locals>._fn at 0x7c5e8438f560> timeout: 1200 secs and primary_host=0 for async checkpoint writes
I0423 21:38:35.001644 136838485276480 checkpoint_manager.py:1812] Found 0 checkpoint steps in gs://lance-maxtext/linen_ckpt_xpk_test_pipeline_scan_nnx_20260423_211617/linen_xpk_test_pipeline_scan_nnx_20260423_211617_08_checkpoint_async_true/checkpoints
I0423 21:38:35.419423 136838485276480 checkpoint_manager.py:929] [process=5][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=5), 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/linen_ckpt_xpk_test_pipeline_scan_nnx_20260423_211617/linen_xpk_test_pipeline_scan_nnx_20260423_211617_08_checkpoint_async_true/checkpoints: <orbax.checkpoint.checkpoint_manager.CheckpointManager object at 0x7c5e8456a7e0>
I0423 21:38:35.419602 136838485276480 checkpointing.py:301] Checkpoint manager created!
I0423 21:38:36.353072 136838485276480 nnx_wrappers.py:453] Unknown Logical: bfloat16[32,2048,2048]...................................... ('activation_batch', 'activation_norm_length', 'activation_embed').
I0423 21:38:36.353190 136838485276480 nnx_wrappers.py:453] Unknown Physical: bfloat16[32,2048,2048]...................................... ('fsdp', None, None).
I0423 21:38:36.733722 136838485276480 attentions.py:1088] attentions/inputs_q Logical: bfloat16[32,2048,2048]...................................... ('activation_batch', 'activation_attn_length', 'activation_attn_embed').
I0423 21:38:36.733813 136838485276480 attentions.py:1088] attentions/inputs_q Physical: bfloat16[32,2048,2048]...................................... ('fsdp', None, None).
I0423 21:38:36.750179 136838485276480 attentions.py:1089] attentions/inputs_kv Logical: bfloat16[32,2048,2048]...................................... ('activation_batch', 'activation_attn_length', 'activation_attn_embed').
I0423 21:38:36.750242 136838485276480 attentions.py:1089] attentions/inputs_kv Physical: bfloat16[32,2048,2048]...................................... ('fsdp', None, None).
I0423 21:38:36.773962 136838485276480 attentions.py:1154] attentions/query Logical: bfloat16[32,2048,16,128].................................... ('activation_kv_batch', 'activation_attn_length', 'activation_kv_heads', 'activation_kv_head_dim').
I0423 21:38:36.774035 136838485276480 attentions.py:1154] attentions/query Physical: bfloat16[32,2048,16,128].................................... ('fsdp', None, None, None).
I0423 21:38:36.790695 136838485276480 attentions.py:1155] attentions/key Logical: bfloat16[32,2048,16,128].................................... ('activation_kv_batch', 'activation_attn_length', 'activation_kv_heads', 'activation_kv_head_dim').
I0423 21:38:36.790759 136838485276480 attentions.py:1155] attentions/key Physical: bfloat16[32,2048,16,128].................................... ('fsdp', None, None, None).
I0423 21:38:36.807245 136838485276480 attentions.py:1156] attentions/value Logical: bfloat16[32,2048,16,128].................................... ('activation_kv_batch', 'activation_attn_length', 'activation_kv_heads', 'activation_kv_head_dim').
I0423 21:38:36.807301 136838485276480 attentions.py:1156] attentions/value Physical: bfloat16[32,2048,16,128].................................... ('fsdp', None, None, None).
I0423 21:38:36.832256 136838485276480 attentions.py:1198] attentions/out Logical: bfloat16[32,2048,16,128].................................... ('activation_batch', 'activation_attn_length', 'activation_heads', 'activation_kv').
I0423 21:38:36.832334 136838485276480 attentions.py:1198] attentions/out Physical: bfloat16[32,2048,16,128].................................... ('fsdp', None, None, None).
I0423 21:38:36.853298 136838485276480 linears.py:525] linears/x Logical: bfloat16[32,2048,7168]...................................... ('activation_batch', 'activation_length', 'activation_mlp').
I0423 21:38:36.853371 136838485276480 linears.py:525] linears/x Physical: bfloat16[32,2048,7168]...................................... ('fsdp', None, None).
I0423 21:38:37.063373 136838485276480 checkpointing.py:577] checkpoint manager exists so trying to load this run's existing checkpoint
I0423 21:38:37.063484 136838485276480 checkpointing.py:665] No existing checkpoints found, not restoring checkpoint.
fsdp: 32
I0423 21:38:38.506061 136838485276480 maxtext_utils.py:1668]  params/params/decoder/decoder_norm/scale
    Shape:     float32[2048]
    Logical:   P('norm',)
    Physical:  (None,)
I0423 21:38:38.506190 136838485276480 maxtext_utils.py:1668]  params/params/decoder/layers/mlp/wi_0/kernel
    Shape:     float32[2048,16,7168]
    Logical:   P('embed', 'layers', 'mlp')
    Physical:  ('fsdp', None, None)
I0423 21:38:38.506243 136838485276480 maxtext_utils.py:1668]  params/params/decoder/layers/mlp/wi_1/kernel
    Shape:     float32[2048,16,7168]
    Logical:   P('embed', 'layers', 'mlp')
    Physical:  ('fsdp', None, None)
I0423 21:38:38.506302 136838485276480 maxtext_utils.py:1668]  params/params/decoder/layers/mlp/wo/kernel
    Shape:     float32[7168,16,2048]
    Logical:   P('mlp', 'layers', 'embed')
    Physical:  (None, None, 'fsdp')
I0423 21:38:38.506354 136838485276480 maxtext_utils.py:1668]  params/params/decoder/layers/post_self_attention_layer_norm/scale
    Shape:     float32[2048,16]
    Logical:   P('norm', 'layers')
    Physical:  (None, None)
I0423 21:38:38.506392 136838485276480 maxtext_utils.py:1668]  params/params/decoder/layers/pre_self_attention_layer_norm/scale
    Shape:     float32[2048,16]
    Logical:   P('norm', 'layers')
    Physical:  (None, None)
I0423 21:38:38.506446 136838485276480 maxtext_utils.py:1668]  params/params/decoder/layers/self_attention/key/kernel
    Shape:     float32[2048,16,16,128]
    Logical:   P('embed', 'layers', 'kv_heads', 'kv_head_dim')
    Physical:  ('fsdp', None, None, None)
I0423 21:38:38.506499 136838485276480 maxtext_utils.py:1668]  params/params/decoder/layers/self_attention/out/kernel
    Shape:     float32[16,16,128,2048]
    Logical:   P('heads', 'layers', 'kv', 'embed')
    Physical:  (None, None, None, 'fsdp')
I0423 21:38:38.506540 136838485276480 maxtext_utils.py:1668]  params/params/decoder/layers/self_attention/query/kernel
    Shape:     float32[2048,16,16,128]
    Logical:   P('embed', 'layers', 'q_heads', 'kv')
    Physical:  ('fsdp', None, None, None)
I0423 21:38:38.506577 136838485276480 maxtext_utils.py:1668]  params/params/decoder/layers/self_attention/value/kernel
    Shape:     float32[2048,16,16,128]
    Logical:   P('embed', 'layers', 'kv_heads', 'kv_head_dim')
    Physical:  ('fsdp', None, None, None)
I0423 21:38:38.506626 136838485276480 maxtext_utils.py:1668]  params/params/decoder/logits_dense/kernel
    Shape:     float32[2048,32000]
    Logical:   P('embed_vocab', 'vocab')
    Physical:  ('fsdp', None)
I0423 21:38:38.506688 136838485276480 maxtext_utils.py:1668]  params/params/token_embedder/embedding
    Shape:     float32[32000,2048]

    Logical:   P('vocab', 'embed_vocab')
    Physical:  (None, 'fsdp')
I0423 21:38:39.001312 136838485276480 train.py:155] train/xent Logical: float32[32,2048]............................................ ('activation_embed_and_logits_batch', 'activation_length').
I0423 21:38:39.001407 136838485276480 train.py:155] train/xent Physical: float32[32,2048]............................................ ('fsdp', None).
I0423 21:38:39.016864 136838485276480 train.py:162] train/z_loss Logical: float32[32,2048]............................................ ('activation_embed_and_logits_batch', 'activation_length').
I0423 21:38:39.016926 136838485276480 train.py:162] train/z_loss Physical: float32[32,2048]............................................ ('fsdp', None).
I0423 21:38:49.932245 136838485276480 max_utils.py:791] Total memory size: 1.5 GB, Output size: 0.4 GB, Temp size: 1.1 GB, Argument size: 0.4 GB, Host temp size: 0.0 GB.
I0423 21:38:49.933043 136838485276480 metric_logger.py:301] number parameters: 1.104 billion
I0423 21:39:01.548878 136838485276480 checkpointing.py:772] Waiting for step 0 to finish before checkpoint...
I0423 21:39:01.701643 136838485276480 checkpointing.py:776] Waited 0.15274262428283691 seconds for step 0 to finish before starting checkpointing.
I0423 21:39:01.704280 136838485276480 checkpoint_manager.py:2009] [process=5][thread=MainThread][wait_until_finished] No Save Finalize thread to wait for. Returning.
I0423 21:39:01.706097 136838485276480 checkpoint_manager.py:1512] [process=5] Saving checkpoint at step 0
I0423 21:39:01.707690 136838485276480 event_tracking.py:70] [process=5] [async] Started save checkpoint @ gs://lance-maxtext/linen_ckpt_xpk_test_pipeline_scan_nnx_20260423_211617/linen_xpk_test_pipeline_scan_nnx_20260423_211617_08_checkpoint_async_true/checkpoints/0.
I0423 21:39:01.997588 136838485276480 signaling_client.py:364] Using JaxDistributedSignalingClient
I0423 21:39:01.998638 136838485276480 jax_array_handlers.py:360] Scheduling D2H of 39 prioritized jax.Array.
I0423 21:39:01.998731 136838485276480 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
I0423 21:39:02.276366 136838485276480 base_pytree_checkpoint_handler.py:154] [process=5][thread=MainThread] Initiated "orbax.checkpoint._src.serialization.jax_array_handlers.ArrayHandler".serialize. Time taken: 0.278860s
I0423 21:39:02.276536 136838485276480 base_pytree_checkpoint_handler.py:130] [process=5] /jax/orbax/write/blocking_gbytes_per_sec: 5.417 GiB/s (total gbytes: 1.5 GiB) (time elapsed: 0.2847433090209961 s) (per-host)
I0423 21:39:02.276589 136838485276480 base_pytree_checkpoint_handler.py:768] [process=5][thread=MainThread] Initiated Pytree async_save. Time taken: 0.284805s (batch_requests_ready=0.002453s, total_serialization_initiated=0.282282s, others=0.000070s)
I0423 21:39:02.276698 136838485276480 composite_checkpoint_handler.py:715] [process=5][thread=MainThread] Initiated CompositeCheckpointHandler.async_save. Time taken: 0.289049s (all_items=0.000024s, per_item={'items': '0.00002408'}, temp_paths=0.289025)
I0423 21:39:02.277520 136838485276480 event_tracking.py:125] [process=5] [async] Finished blocking save in 0.57 seconds. Continuing save @ gs://lance-maxtext/linen_ckpt_xpk_test_pipeline_scan_nnx_20260423_211617/linen_xpk_test_pipeline_scan_nnx_20260423_211617_08_checkpoint_async_true/checkpoints/0.
I0423 21:39:02.277811 136708905072384 async_checkpointer.py:76] [process=5][thread=async_save] Background save thread started. Deadline for this save operation is 2026-04-23 21:59:02.277778
I0423 21:39:02.279754 136838485276480 checkpoint_manager.py:1560] [process=5][thread=MainThread][step=0] Starting CheckpointManager Save Finalize thread=save_finalize
I0423 21:39:02.279976 136707861112576 async_checkpointer.py:280] [process=5][thread=save_finalize] Waiting for background save thread=async_save.
I0423 21:39:02.280069 136838485276480 standard_logger.py:34] {'step': 0, 'event_type': 'save', 'directory': 'gs://lance-maxtext/linen_ckpt_xpk_test_pipeline_scan_nnx_20260423_211617/linen_xpk_test_pipeline_scan_nnx_20260423_211617_08_checkpoint_async_true/checkpoints', 'reached_preemption': False, 'preemption_received_at': None, 'synchronous': False, 'wait_for_prev_start_time': 1776980341.7042613, 'wait_for_prev_duration_secs': 6.031990051269531e-05, 'time_between_consecutive_saves_sec': None, 'checkpointer_blocking_start_time': 1776980341.7061353, 'checkpointer_blocking_duration_secs': 0.5718293190002441, 'get_old_steps_start_time': 1776980342.277989, 'get_old_steps_duration_secs': 3.0279159545898438e-05, 'checkpoint_manager_blocking_start_time': 1776980341.7021563, 'checkpoint_manager_blocking_duration_secs': 0.5778775215148926}
I0423 21:39:02.280229 136838485276480 checkpointing.py:408] Started an asynchronous checkpoint save for step 0
I0423 21:39:02.280282 136838485276480 max_utils.py:750] 
Memstats: After params initialized:
I0423 21:39:02.280329 136838485276480 max_utils.py:756] 	Using (GB) 0.43 / 31.25 (1.376000%) on TPU_18(process=5,(2,4,0,0))
I0423 21:39:02.280363 136838485276480 max_utils.py:756] 	Using (GB) 0.43 / 31.25 (1.376000%) on TPU_19(process=5,(3,4,0,0))
I0423 21:39:02.280390 136838485276480 max_utils.py:756] 	Using (GB) 0.43 / 31.25 (1.376000%) on TPU_22(process=5,(2,5,0,0))
I0423 21:39:02.280416 136838485276480 max_utils.py:756] 	Using (GB) 0.43 / 31.25 (1.376000%) on TPU_23(process=5,(3,5,0,0))
I0423 21:39:02.592238 136838485276480 metric_logger.py:196] completed step: 0, seconds: 11.616, TFLOP/s/device: 1.170, Tokens/s/device: 176.313, total_weights: 65536, loss: 10.874, lm_loss: 10.874, perplexity: 52779.875
I0423 21:39:02.763148 136838485276480 metric_logger.py:196] completed step: 1, seconds: 1.042, TFLOP/s/device: 13.041, Tokens/s/device: 1965.730, total_weights: 65536, loss: 10.874, lm_loss: 10.874, perplexity: 52779.875
I0423 21:39:03.174485 136838485276480 metric_logger.py:196] completed step: 2, seconds: 0.024, TFLOP/s/device: 570.217, Tokens/s/device: 85949.303, total_weights: 65536, loss: 10.560, lm_loss: 10.560, perplexity: 38547.719
I0423 21:39:03.321741 136838485276480 metric_logger.py:196] completed step: 3, seconds: 0.412, TFLOP/s/device: 33.010, Tokens/s/device: 4975.656, total_weights: 65536, loss: 9.980, lm_loss: 9.980, perplexity: 21590.506
I0423 21:39:03.326296 136838485276480 checkpointing.py:772] Waiting for step 5 to finish before checkpoint...
I0423 21:39:03.616072 136838485276480 checkpointing.py:776] Waited 0.289747953414917 seconds for step 5 to finish before starting checkpointing.
I0423 21:39:03.619019 136838485276480 checkpoint_manager.py:2020] [process=5][thread=MainThread][step=0][wait_until_finished] Waiting for Save Finalize thread (save_finalize) to complete.
I0423 21:39:05.392065    2834 google_auth_provider.cc:181] Running on GCE, using service account 562977990677-compute@developer.gserviceaccount.com
I0423 21:39:07.717158 136707869505280 array_metadata_store.py:203] [process=5][thread=array_type_handler] Wrote 39 array_metadata.ArrayMetadata to gs://lance-maxtext/linen_ckpt_xpk_test_pipeline_scan_nnx_20260423_211617/linen_xpk_test_pipeline_scan_nnx_20260423_211617_08_checkpoint_async_true/checkpoints/0/items/array_metadatas/process_5
I0423 21:39:38.551301 136708905072384 base_pytree_checkpoint_handler.py:130] [process=5] /jax/orbax/write/gbytes_per_sec: 43.206 MiB/s (total gbytes: 1.5 GiB) (time elapsed: 36.55945110321045 s) (per-host)
I0423 21:39:38.551423 136708905072384 async_checkpointer.py:90] [process=5][thread=async_save] 3 Handler Commit operations completed. Time taken: 36.273500s.
I0423 21:39:47.641228 136708905072384 async_checkpointer.py:160] [process=5][thread=async_save] Background save thread done. Time taken: 45.363284s.
I0423 21:39:47.641524 136707861112576 async_checkpointer.py:288] [process=5][thread=save_finalize] Done with waiting for background save thread=async_save.
I0423 21:39:47.641644 136707861112576 async_checkpointer.py:298] [process=5][thread=save_finalize] No errors found in background save thread=async_save.
I0423 21:39:47.641715 136707861112576 checkpoint_manager.py:2137] [process=5][thread=save_finalize][step=0] CheckpointManager Save Finalize is syncing with other hosts...
I0423 21:39:47.644175 136707861112576 checkpoint_manager.py:2146] [process=5][thread=save_finalize][step=0] CheckpointManager Save Finalize is done on all hosts.
I0423 21:39:47.644340 136838485276480 checkpoint_manager.py:2032] [process=5][thread=MainThread][step=0][wait_until_finished] Done waiting for Save Finalize thread (save_finalize) running at step=0.
W0423 21:39:47.644477 136838485276480 checkpoint_manager.py:1452] Waiting for previous save to complete took 44.025458 seconds. If this number is high, consider checkpointing less frequently.
I0423 21:39:47.646288 136838485276480 checkpoint_manager.py:1512] [process=5] Saving checkpoint at step 5
I0423 21:39:47.648642 136838485276480 event_tracking.py:70] [process=5] [async] Started save checkpoint @ gs://lance-maxtext/linen_ckpt_xpk_test_pipeline_scan_nnx_20260423_211617/linen_xpk_test_pipeline_scan_nnx_20260423_211617_08_checkpoint_async_true/checkpoints/5.
I0423 21:39:47.958004 136838485276480 jax_array_handlers.py:360] Scheduling D2H of 39 prioritized jax.Array.
I0423 21:39:47.958097 136838485276480 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
I0423 21:39:48.006838 136838485276480 base_pytree_checkpoint_handler.py:154] [process=5][thread=MainThread] Initiated "orbax.checkpoint._src.serialization.jax_array_handlers.ArrayHandler".serialize. Time taken: 0.049916s
I0423 21:39:48.007010 136838485276480 base_pytree_checkpoint_handler.py:130] [process=5] /jax/orbax/write/blocking_gbytes_per_sec: 28.835 GiB/s (total gbytes: 1.5 GiB) (time elapsed: 0.0534970760345459 s) (per-host)
I0423 21:39:48.007060 136838485276480 base_pytree_checkpoint_handler.py:768] [process=5][thread=MainThread] Initiated Pytree async_save. Time taken: 0.053557s (batch_requests_ready=0.001904s, total_serialization_initiated=0.051585s, others=0.000068s)
I0423 21:39:48.007152 136838485276480 composite_checkpoint_handler.py:715] [process=5][thread=MainThread] Initiated CompositeCheckpointHandler.async_save. Time taken: 0.058627s (all_items=0.000016s, per_item={'items': '0.00001597'}, temp_paths=0.058611)
I0423 21:39:48.007789 136838485276480 event_tracking.py:125] [process=5] [async] Finished blocking save in 0.36 seconds. Continuing save @ gs://lance-maxtext/linen_ckpt_xpk_test_pipeline_scan_nnx_20260423_211617/linen_xpk_test_pipeline_scan_nnx_20260423_211617_08_checkpoint_async_true/checkpoints/5.
I0423 21:39:48.008148 136707861112576 async_checkpointer.py:76] [process=5][thread=async_save] Background save thread started. Deadline for this save operation is 2026-04-23 21:59:48.008109
I0423 21:39:48.010064 136838485276480 checkpoint_manager.py:1560] [process=5][thread=MainThread][step=5] Starting CheckpointManager Save Finalize thread=save_finalize
I0423 21:39:48.010349 136687644559104 async_checkpointer.py:280] [process=5][thread=save_finalize] Waiting for background save thread=async_save.
I0423 21:39:48.010483 136838485276480 standard_logger.py:34] {'step': 5, 'event_type': 'save', 'directory': 'gs://lance-maxtext/linen_ckpt_xpk_test_pipeline_scan_nnx_20260423_211617/linen_xpk_test_pipeline_scan_nnx_20260423_211617_08_checkpoint_async_true/checkpoints', 'reached_preemption': False, 'preemption_received_at': None, 'synchronous': False, 'wait_for_prev_start_time': 1776980343.618988, 'wait_for_prev_duration_secs': 44.025458097457886, 'time_between_consecutive_saves_sec': None, 'checkpointer_blocking_start_time': 1776980387.6463265, 'checkpointer_blocking_duration_secs': 0.36197900772094727, 'get_old_steps_start_time': 1776980388.008326, 'get_old_steps_duration_secs': 2.86102294921875e-05, 'checkpoint_manager_blocking_start_time': 1776980343.616722, 'checkpoint_manager_blocking_duration_secs': 44.393728256225586}
I0423 21:39:48.010666 136838485276480 checkpointing.py:408] Started an asynchronous checkpoint save for step 5
I0423 21:39:48.011515 136838485276480 metric_logger.py:196] completed step: 4, seconds: 0.153, TFLOP/s/device: 88.879, Tokens/s/device: 13396.829, total_weights: 65536, loss: 9.460, lm_loss: 9.460, perplexity: 12840.355
I0423 21:39:48.017678 136838485276480 metric_logger.py:196] completed step: 5, seconds: 0.147, TFLOP/s/device: 92.339, Tokens/s/device: 13918.338, total_weights: 65536, loss: 8.959, lm_loss: 8.959, perplexity: 7776.467
I0423 21:39:53.465388 136702847928064 array_metadata_store.py:203] [process=5][thread=array_type_handler] Wrote 39 array_metadata.ArrayMetadata to gs://lance-maxtext/linen_ckpt_xpk_test_pipeline_scan_nnx_20260423_211617/linen_xpk_test_pipeline_scan_nnx_20260423_211617_08_checkpoint_async_true/checkpoints/5/items/array_metadatas/process_5
I0423 21:40:10.326790 136838485276480 metric_logger.py:196] completed step: 6, seconds: 44.690, TFLOP/s/device: 0.304, Tokens/s/device: 45.827, total_weights: 65536, loss: 8.469, lm_loss: 8.469, perplexity: 4762.740
I0423 21:40:10.474260 136838485276480 metric_logger.py:196] completed step: 7, seconds: 22.163, TFLOP/s/device: 0.613, Tokens/s/device: 92.406, total_weights: 65536, loss: 8.003, lm_loss: 8.003, perplexity: 2989.151
I0423 21:40:10.621586 136838485276480 metric_logger.py:196] completed step: 8, seconds: 0.152, TFLOP/s/device: 89.331, Tokens/s/device: 13464.914, total_weights: 65536, loss: 7.572, lm_loss: 7.572, perplexity: 1943.098
I0423 21:40:10.768277 136838485276480 checkpointing.py:772] Waiting for step 9 to finish before checkpoint...
I0423 21:40:10.768928 136838485276480 checkpointing.py:776] Waited 0.0006704330444335938 seconds for step 9 to finish before starting checkpointing.
I0423 21:40:10.770977 136838485276480 checkpoint_manager.py:2020] [process=5][thread=MainThread][step=5][wait_until_finished] Waiting for Save Finalize thread (save_finalize) to complete.
I0423 21:40:29.948011 136707861112576 base_pytree_checkpoint_handler.py:130] [process=5] /jax/orbax/write/gbytes_per_sec: 37.614 MiB/s (total gbytes: 1.5 GiB) (time elapsed: 41.99445652961731 s) (per-host)
I0423 21:40:29.948130 136707861112576 async_checkpointer.py:90] [process=5][thread=async_save] 3 Handler Commit operations completed. Time taken: 41.939862s.
I0423 21:40:39.799294 136707861112576 async_checkpointer.py:160] [process=5][thread=async_save] Background save thread done. Time taken: 51.791011s.
I0423 21:40:39.799599 136687644559104 async_checkpointer.py:288] [process=5][thread=save_finalize] Done with waiting for background save thread=async_save.
I0423 21:40:39.799739 136687644559104 async_checkpointer.py:298] [process=5][thread=save_finalize] No errors found in background save thread=async_save.
I0423 21:40:39.799787 136687644559104 checkpoint_manager.py:2137] [process=5][thread=save_finalize][step=5] CheckpointManager Save Finalize is syncing with other hosts...
I0423 21:40:39.801995 136687644559104 checkpoint_manager.py:2146] [process=5][thread=save_finalize][step=5] CheckpointManager Save Finalize is done on all hosts.
I0423 21:40:39.802157 136838485276480 checkpoint_manager.py:2032] [process=5][thread=MainThread][step=5][wait_until_finished] Done waiting for Save Finalize thread (save_finalize) running at step=5.
W0423 21:40:39.802282 136838485276480 checkpoint_manager.py:1452] Waiting for previous save to complete took 29.031311 seconds. If this number is high, consider checkpointing less frequently.
I0423 21:40:39.803746 136838485276480 checkpoint_manager.py:1512] [process=5] Saving checkpoint at step 9
I0423 21:40:39.805721 136838485276480 event_tracking.py:70] [process=5] [async] Started save checkpoint @ gs://lance-maxtext/linen_ckpt_xpk_test_pipeline_scan_nnx_20260423_211617/linen_xpk_test_pipeline_scan_nnx_20260423_211617_08_checkpoint_async_true/checkpoints/9.
I0423 21:40:40.100301 136838485276480 jax_array_handlers.py:360] Scheduling D2H of 39 prioritized jax.Array.
I0423 21:40:40.100389 136838485276480 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
I0423 21:40:40.142964 136838485276480 base_pytree_checkpoint_handler.py:154] [process=5][thread=MainThread] Initiated "orbax.checkpoint._src.serialization.jax_array_handlers.ArrayHandler".serialize. Time taken: 0.043503s
I0423 21:40:40.143115 136838485276480 base_pytree_checkpoint_handler.py:130] [process=5] /jax/orbax/write/blocking_gbytes_per_sec: 32.747 GiB/s (total gbytes: 1.5 GiB) (time elapsed: 0.047106027603149414 s) (per-host)
I0423 21:40:40.143162 136838485276480 base_pytree_checkpoint_handler.py:768] [process=5][thread=MainThread] Initiated Pytree async_save. Time taken: 0.047163s (batch_requests_ready=0.001809s, total_serialization_initiated=0.045293s, others=0.000061s)
I0423 21:40:40.143251 136838485276480 composite_checkpoint_handler.py:715] [process=5][thread=MainThread] Initiated CompositeCheckpointHandler.async_save. Time taken: 0.051466s (all_items=0.000010s, per_item={'items': '0.00001049'}, temp_paths=0.051456)
I0423 21:40:40.143986 136838485276480 event_tracking.py:125] [process=5] [async] Finished blocking save in 0.34 seconds. Continuing save @ gs://lance-maxtext/linen_ckpt_xpk_test_pipeline_scan_nnx_20260423_211617/linen_xpk_test_pipeline_scan_nnx_20260423_211617_08_checkpoint_async_true/checkpoints/9.
I0423 21:40:40.144326 136702847928064 async_checkpointer.py:76] [process=5][thread=async_save] Background save thread started. Deadline for this save operation is 2026-04-23 22:00:40.144288
I0423 21:40:40.146297 136838485276480 checkpoint_manager.py:1560] [process=5][thread=MainThread][step=9] Starting CheckpointManager Save Finalize thread=save_finalize
I0423 21:40:40.146589 136687644559104 async_checkpointer.py:280] [process=5][thread=save_finalize] Waiting for background save thread=async_save.
I0423 21:40:40.146730 136838485276480 standard_logger.py:34] {'step': 9, 'event_type': 'save', 'directory': 'gs://lance-maxtext/linen_ckpt_xpk_test_pipeline_scan_nnx_20260423_211617/linen_xpk_test_pipeline_scan_nnx_20260423_211617_08_checkpoint_async_true/checkpoints', 'reached_preemption': False, 'preemption_received_at': None, 'synchronous': False, 'wait_for_prev_start_time': 1776980410.7709453, 'wait_for_prev_duration_secs': 29.031310558319092, 'time_between_consecutive_saves_sec': 23.126729011535645, 'checkpointer_blocking_start_time': 1776980439.8037841, 'checkpointer_blocking_duration_secs': 0.3406863212585449, 'get_old_steps_start_time': 1776980440.1444888, 'get_old_steps_duration_secs': 2.288818359375e-05, 'checkpoint_manager_blocking_start_time': 1776980410.7691538, 'checkpoint_manager_blocking_duration_secs': 29.377547025680542}
I0423 21:40:40.146900 136838485276480 checkpointing.py:408] Started an asynchronous checkpoint save for step 9
I0423 21:40:40.146944 136838485276480 checkpoint_manager.py:2020] [process=5][thread=MainThread][step=9][wait_until_finished] Waiting for Save Finalize thread (save_finalize) to complete.
I0423 21:40:45.905086 136688491804416 array_metadata_store.py:203] [process=5][thread=array_type_handler] Wrote 39 array_metadata.ArrayMetadata to gs://lance-maxtext/linen_ckpt_xpk_test_pipeline_scan_nnx_20260423_211617/linen_xpk_test_pipeline_scan_nnx_20260423_211617_08_checkpoint_async_true/checkpoints/9/items/array_metadatas/process_5
I0423 21:41:23.000763 136702847928064 base_pytree_checkpoint_handler.py:130] [process=5] /jax/orbax/write/gbytes_per_sec: 36.816 MiB/s (total gbytes: 1.5 GiB) (time elapsed: 42.90471267700195 s) (per-host)
I0423 21:41:23.000885 136702847928064 async_checkpointer.py:90] [process=5][thread=async_save] 3 Handler Commit operations completed. Time taken: 42.856445s.
I0423 21:41:31.336550 136702847928064 async_checkpointer.py:160] [process=5][thread=async_save] Background save thread done. Time taken: 51.192095s.
I0423 21:41:31.336870 136687644559104 async_checkpointer.py:288] [process=5][thread=save_finalize] Done with waiting for background save thread=async_save.
I0423 21:41:31.336990 136687644559104 async_checkpointer.py:298] [process=5][thread=save_finalize] No errors found in background save thread=async_save.
I0423 21:41:31.337031 136687644559104 checkpoint_manager.py:2137] [process=5][thread=save_finalize][step=9] CheckpointManager Save Finalize is syncing with other hosts...
I0423 21:41:31.338688 136687644559104 checkpoint_manager.py:2146] [process=5][thread=save_finalize][step=9] CheckpointManager Save Finalize is done on all hosts.
I0423 21:41:31.338876 136838485276480 checkpoint_manager.py:2032] [process=5][thread=MainThread][step=9][wait_until_finished] Done waiting for Save Finalize thread (save_finalize) running at step=9.
I0423 21:41:31.339025 136838485276480 checkpoint_manager.py:2009] [process=5][thread=MainThread][wait_until_finished] No Save Finalize thread to wait for. Returning.
I0423 21:41:31.340001 136838485276480 metric_logger.py:196] completed step: 9, seconds: 0.147, TFLOP/s/device: 92.150, Tokens/s/device: 13889.925, total_weights: 65536, loss: 7.181, lm_loss: 7.181, perplexity: 1314.495
Per train step:
 Total TFLOPs: 13.59 
 split as 93.93% learnable weight flops and 6.07% attention flops
XPK End: Thu Apr 23 21:41:44 UTC 2026
EXIT_CODE=0