MaxView

Timing Test Reference Updates
‹ 08_checkpoint_async_true_saveCase: 09_pdb_lt_110_shardy_false ›

Metrics: main (574ad3fb9) vs test/pipeline-scan-nnx (e27fc1e97)

Metricmain  574ad3fb9test/pipeline-scan-nnx  e27fc1e97Diff (test/pipeline-scan-nnx − main)
Parameters1.104 billion1.104 billion—
Final loss5.89205.89200
TFLOP/s58.45158.053-0.398
Tok/s8810.48750.3-60.079
Avg s/step3.4042.819-0.585
Memory %1.381.380
JAX0.9.20.8.1—

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

main  ·  574ad3fb9  ·  main_20260420_060028  ·  full log 
XPK Start: Mon Apr 20 06:32:16 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 06:32:40.300508: 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 06:32:40.825050 135121984083776 max_utils.py:273] Attempting to initialize the jax distributed system...
I0420 06:32:49.866224 135121984083776 distributed.py:149] Starting JAX distributed service on [::]:8482
I0420 06:32:49.868501 135121984083776 distributed.py:172] Connecting to JAX distributed service on mt-09-pdb-lt-1-q8xwz-slice-job-0-0.mt-09-pdb-lt-1-q8xwz:8482
I0420 06:32:51.182766 135121984083776 max_utils.py:284] Jax distributed system initialized!
I0420 06:32:56.504998 135121984083776 max_utils.py:800] System Information: Jax Version: 0.9.2
I0420 06:32:56.505098 135121984083776 max_utils.py:801] System Information: Jaxlib Version: 0.9.2
I0420 06:32:56.505138 135121984083776 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 06:32:56.505172 135121984083776 train_utils.py:348] WARNING: Sequence packing is essentially ignored for synthetic data. Please use a real dataset to use sequence packing.
I0420 06:32:57.197817 135121984083776 maxtext_utils.py:1551] Num_devices: 32, shape (1, 1, 1, 8, 1, 1, 1, 1, 4, 1, 1, 1, 1)
I0420 06:32:57.198091 135121984083776 checkpointing.py:677] Setting up checkpoint logger...
I0420 06:32:57.198144 135121984083776 checkpointing.py:233] Creating checkpoint manager with ocdbt=True and zarr3=True
I0420 06:32:57.198187 135121984083776 pytree_checkpoint_handler.py:592] save_device_host_concurrent_bytes=None
I0420 06:32:57.198549 135121984083776 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 0x7ae3ee51ccb0>, enable_pinned_host_transfer=False, save_concurrent_bytes: 96000000000 (89.4 GiB), restore_concurrent_bytes: 96000000000 (89.4 GiB)
I0420 06:33:00.575733 135121984083776 checkpointing.py:265] Enabling policy for fixed interval checkpointing.
I0420 06:33:00.575959 135121984083776 checkpoint_manager.py:708] [process=4][thread=MainThread] CheckpointManager init: checkpointers=None, item_names=('items',), item_handlers={'items': <orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeCheckpointHandler object at 0x7ad05c7b7860>}, handler_registry=None
I0420 06:33:00.576191 135121984083776 composite_checkpoint_handler.py:237] Deferred registration for item: "items". Adding handler `<orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeCheckpointHandler object at 0x7ad05c7b7860>` 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 06:33:00.576240 135121984083776 composite_checkpoint_handler.py:237] Deferred registration for item: "metrics". Adding handler `<orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonCheckpointHandler object at 0x7ad05c7b9be0>` 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 06:33:00.576275 135121984083776 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 0x7ad05c7b7860>, ('items', <class 'orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeRestoreArgs'>): <orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeCheckpointHandler object at 0x7ad05c7b7860>, ('metrics', <class 'orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonSaveArgs'>): <orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonCheckpointHandler object at 0x7ad05c7b9be0>, ('metrics', <class 'orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonRestoreArgs'>): <orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonCheckpointHandler object at 0x7ad05c7b9be0>}).
I0420 06:33:00.576617 135121984083776 abstract_checkpointer.py:35] orbax-checkpoint version: 0.11.34
I0420 06:33:00.576689 135121984083776 async_checkpointer.py:192] [process=4][thread=MainThread] Using barrier_sync_fn: <function get_barrier_sync_fn.<locals>._fn at 0x7ad05c3eb920> timeout: 1200 secs and primary_host=0 for async checkpoint writes
I0420 06:33:01.685627 135121984083776 checkpoint_manager.py:1812] Found 0 checkpoint steps in gs://lance-maxtext/linen_ckpt_xpk_main_20260420_060028/linen_xpk_main_20260420_060028_09_pdb_lt_1/checkpoints
I0420 06:33:01.687796 135121984083776 checkpoint_manager.py:929] [process=4][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/linen_ckpt_xpk_main_20260420_060028/linen_xpk_main_20260420_060028_09_pdb_lt_1/checkpoints: <orbax.checkpoint.checkpoint_manager.CheckpointManager object at 0x7ad05c7b69c0>
I0420 06:33:01.687907 135121984083776 checkpointing.py:301] Checkpoint manager created!
I0420 06:33:02.605948 135121984083776 nnx_wrappers.py:437] Unknown Logical: bfloat16[8,2048,2048]....................................... ('activation_batch', 'activation_norm_length', 'activation_embed').
I0420 06:33:02.606057 135121984083776 nnx_wrappers.py:437] Unknown Physical: bfloat16[8,2048,2048]....................................... ('fsdp', None, 'tensor').
I0420 06:33:02.985940 135121984083776 attentions.py:1088] attentions/inputs_q Logical: bfloat16[8,2048,2048]....................................... ('activation_batch', 'activation_attn_length', 'activation_attn_embed').
I0420 06:33:02.986028 135121984083776 attentions.py:1088] attentions/inputs_q Physical: bfloat16[8,2048,2048]....................................... ('fsdp', None, 'tensor').
I0420 06:33:03.002871 135121984083776 attentions.py:1089] attentions/inputs_kv Logical: bfloat16[8,2048,2048]....................................... ('activation_batch', 'activation_attn_length', 'activation_attn_embed').
I0420 06:33:03.002928 135121984083776 attentions.py:1089] attentions/inputs_kv Physical: bfloat16[8,2048,2048]....................................... ('fsdp', None, 'tensor').
I0420 06:33:03.026705 135121984083776 attentions.py:1154] attentions/query Logical: bfloat16[8,2048,16,128]..................................... ('activation_kv_batch', 'activation_attn_length', 'activation_kv_heads', 'activation_kv_head_dim').
I0420 06:33:03.026769 135121984083776 attentions.py:1154] attentions/query Physical: bfloat16[8,2048,16,128]..................................... ('fsdp', None, 'tensor', None).
I0420 06:33:03.043116 135121984083776 attentions.py:1155] attentions/key Logical: bfloat16[8,2048,16,128]..................................... ('activation_kv_batch', 'activation_attn_length', 'activation_kv_heads', 'activation_kv_head_dim').
I0420 06:33:03.043175 135121984083776 attentions.py:1155] attentions/key Physical: bfloat16[8,2048,16,128]..................................... ('fsdp', None, 'tensor', None).
I0420 06:33:03.059733 135121984083776 attentions.py:1156] attentions/value Logical: bfloat16[8,2048,16,128]..................................... ('activation_kv_batch', 'activation_attn_length', 'activation_kv_heads', 'activation_kv_head_dim').
I0420 06:33:03.059800 135121984083776 attentions.py:1156] attentions/value Physical: bfloat16[8,2048,16,128]..................................... ('fsdp', None, 'tensor', None).
I0420 06:33:03.085382 135121984083776 attentions.py:1197] attentions/out Logical: bfloat16[8,2048,16,128]..................................... ('activation_batch', 'activation_attn_length', 'activation_heads', 'activation_kv').
I0420 06:33:03.085456 135121984083776 attentions.py:1197] attentions/out Physical: bfloat16[8,2048,16,128]..................................... ('fsdp', None, 'tensor', None).
I0420 06:33:03.106225 135121984083776 linears.py:525] linears/x Logical: bfloat16[8,2048,7168]....................................... ('activation_batch', 'activation_length', 'activation_mlp').
I0420 06:33:03.106286 135121984083776 linears.py:525] linears/x Physical: bfloat16[8,2048,7168]....................................... ('fsdp', None, 'tensor').
I0420 06:33:03.320221 135121984083776 checkpointing.py:577] checkpoint manager exists so trying to load this run's existing checkpoint
I0420 06:33:03.320325 135121984083776 checkpointing.py:665] No existing checkpoints found, not restoring checkpoint.
fsdp: 8
tensor: 4
I0420 06:33:04.752044 135121984083776 maxtext_utils.py:1654]  params/params/decoder/decoder_norm/scale
    Shape:     float32[2048]
    Logical:   P('norm',)
    Physical:  ('tensor',)
I0420 06:33:04.752164 135121984083776 maxtext_utils.py:1654]  params/params/decoder/layers/mlp/wi_0/kernel
    Shape:     float32[2048,16,7168]
    Logical:   P('embed', 'layers', 'mlp')
    Physical:  ('fsdp', None, 'tensor')
I0420 06:33:04.752216 135121984083776 maxtext_utils.py:1654]  params/params/decoder/layers/mlp/wi_1/kernel
    Shape:     float32[2048,16,7168]
    Logical:   P('embed', 'layers', 'mlp')
    Physical:  ('fsdp', None, 'tensor')
I0420 06:33:04.752272 135121984083776 maxtext_utils.py:1654]  params/params/decoder/layers/mlp/wo/kernel
    Shape:     float32[7168,16,2048]
    Logical:   P('mlp', 'layers', 'embed')
    Physical:  ('tensor', None, 'fsdp')
I0420 06:33:04.752323 135121984083776 maxtext_utils.py:1654]  params/params/decoder/layers/post_self_attention_layer_norm/scale
    Shape:     float32[2048,16]
    Logical:   P('norm', 'layers')
    Physical:  ('tensor', None)
I0420 06:33:04.752394 135121984083776 maxtext_utils.py:1654]  params/params/decoder/layers/pre_self_attention_layer_norm/scale
    Shape:     float32[2048,16]
    Logical:   P('norm', 'layers')
    Physical:  ('tensor', None)
I0420 06:33:04.752449 135121984083776 maxtext_utils.py:1654]  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, 'tensor', None)
I0420 06:33:04.752571 135121984083776 maxtext_utils.py:1654]  params/params/decoder/layers/self_attention/out/kernel
    Shape:     float32[16,16,128,2048]
    Logical:   P('heads', 'layers', 'kv', 'embed')
    Physical:  ('tensor', None, None, 'fsdp')
I0420 06:33:04.752646 135121984083776 maxtext_utils.py:1654]  params/params/decoder/layers/self_attention/query/kernel
    Shape:     float32[2048,16,16,128]
    Logical:   P('embed', 'layers', 'q_heads', 'kv')
    Physical:  ('fsdp', None, 'tensor', None)
I0420 06:33:04.752694 135121984083776 maxtext_utils.py:1654]  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, 'tensor', None)
I0420 06:33:04.752751 135121984083776 maxtext_utils.py:1654]  params/params/decoder/logits_dense/kernel
    Shape:     float32[2048,32000]
    Logical:   P('embed_vocab', 'vocab')
    Physical:  ('fsdp', 'tensor')
I0420 06:33:04.752805 135121984083776 maxtext_utils.py:1654]  params/params/token_embedder/embedding
    Shape:     float32[32000,2048]
    Logical:   P('vocab', 'embed_vocab')
    Physical:  ('tensor', 'fsdp')

I0420 06:33:05.244531 135121984083776 train.py:155] train/xent Logical: float32[8,2048]............................................. ('activation_embed_and_logits_batch', 'activation_length').
I0420 06:33:05.244627 135121984083776 train.py:155] train/xent Physical: float32[8,2048]............................................. ('fsdp', None).
I0420 06:33:05.259930 135121984083776 train.py:162] train/z_loss Logical: float32[8,2048]............................................. ('activation_embed_and_logits_batch', 'activation_length').
I0420 06:33:05.259990 135121984083776 train.py:162] train/z_loss Physical: float32[8,2048]............................................. ('fsdp', None).
I0420 06:33:20.348328 135121984083776 max_utils.py:791] Total memory size: 0.9 GB, Output size: 0.4 GB, Temp size: 0.5 GB, Argument size: 0.4 GB, Host temp size: 0.0 GB.
I0420 06:33:20.349161 135121984083776 metric_logger.py:301] number parameters: 1.104 billion
I0420 06:33:21.496765 135121984083776 checkpointing.py:772] Waiting for step 0 to finish before checkpoint...
I0420 06:33:36.020965 135121984083776 checkpointing.py:776] Waited 14.524179697036743 seconds for step 0 to finish before starting checkpointing.
I0420 06:33:36.023397 135121984083776 checkpoint_manager.py:2009] [process=4][thread=MainThread][wait_until_finished] No Save Finalize thread to wait for. Returning.
I0420 06:33:36.025492 135121984083776 checkpoint_manager.py:1512] [process=4] Saving checkpoint at step 0
I0420 06:33:36.026950 135121984083776 event_tracking.py:70] [process=4] [async] Started save checkpoint @ gs://lance-maxtext/linen_ckpt_xpk_main_20260420_060028/linen_xpk_main_20260420_060028_09_pdb_lt_1/checkpoints/0.
I0420 06:33:36.370064 135121984083776 signaling_client.py:364] Using JaxDistributedSignalingClient
I0420 06:33:36.371099 135121984083776 jax_array_handlers.py:360] Scheduling D2H of 39 prioritized jax.Array.
I0420 06:33:36.371155 135121984083776 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 06:33:36.635266 135121984083776 base_pytree_checkpoint_handler.py:154] [process=4][thread=MainThread] Initiated "orbax.checkpoint._src.serialization.jax_array_handlers.ArrayHandler".serialize. Time taken: 0.265274s
I0420 06:33:36.635464 135121984083776 base_pytree_checkpoint_handler.py:130] [process=4] /jax/orbax/write/blocking_gbytes_per_sec: 5.699 GiB/s (total gbytes: 1.5 GiB) (time elapsed: 0.27068042755126953 s) (per-host)
I0420 06:33:36.635517 135121984083776 base_pytree_checkpoint_handler.py:768] [process=4][thread=MainThread] Initiated Pytree async_save. Time taken: 0.270744s (batch_requests_ready=0.002205s, total_serialization_initiated=0.268466s, others=0.000073s)
I0420 06:33:36.635614 135121984083776 composite_checkpoint_handler.py:715] [process=4][thread=MainThread] Initiated CompositeCheckpointHandler.async_save. Time taken: 0.274798s (all_items=0.000018s, per_item={'items': '0.00001812'}, temp_paths=0.274780)
I0420 06:33:36.636422 135121984083776 event_tracking.py:125] [process=4] [async] Finished blocking save in 0.61 seconds. Continuing save @ gs://lance-maxtext/linen_ckpt_xpk_main_20260420_060028/linen_xpk_main_20260420_060028_09_pdb_lt_1/checkpoints/0.
I0420 06:33:36.636763 134997417813760 async_checkpointer.py:76] [process=4][thread=async_save] Background save thread started. Deadline for this save operation is 2026-04-20 06:53:36.636724
I0420 06:33:36.658661 135121984083776 checkpoint_manager.py:1560] [process=4][thread=MainThread][step=0] Starting CheckpointManager Save Finalize thread=save_finalize
I0420 06:33:36.658954 134997289707264 async_checkpointer.py:280] [process=4][thread=save_finalize] Waiting for background save thread=async_save.
I0420 06:33:36.659110 135121984083776 standard_logger.py:34] {'step': 0, 'event_type': 'save', 'directory': 'gs://lance-maxtext/linen_ckpt_xpk_main_20260420_060028/linen_xpk_main_20260420_060028_09_pdb_lt_1/checkpoints', 'reached_preemption': False, 'preemption_received_at': None, 'synchronous': False, 'wait_for_prev_start_time': 1776666816.023378, 'wait_for_prev_duration_secs': 6.246566772460938e-05, 'time_between_consecutive_saves_sec': None, 'checkpointer_blocking_start_time': 1776666816.0255334, 'checkpointer_blocking_duration_secs': 0.6113917827606201, 'get_old_steps_start_time': 1776666816.636952, 'get_old_steps_duration_secs': 2.8848648071289062e-05, 'checkpoint_manager_blocking_start_time': 1776666816.0214949, 'checkpoint_manager_blocking_duration_secs': 0.6375751495361328}
I0420 06:33:36.659213 135121984083776 checkpointing.py:408] Started an asynchronous checkpoint save for step 0
I0420 06:33:36.659261 135121984083776 max_utils.py:750] 
Memstats: After params initialized:
I0420 06:33:36.659312 135121984083776 max_utils.py:756] 	Using (GB) 0.43 / 31.25 (1.376000%) on TPU_16(process=4,(0,4,0,0))
I0420 06:33:36.659358 135121984083776 max_utils.py:756] 	Using (GB) 0.43 / 31.25 (1.376000%) on TPU_17(process=4,(1,4,0,0))
I0420 06:33:36.659386 135121984083776 max_utils.py:756] 	Using (GB) 0.43 / 31.25 (1.376000%) on TPU_20(process=4,(0,5,0,0))
I0420 06:33:36.659411 135121984083776 max_utils.py:756] 	Using (GB) 0.43 / 31.25 (1.376000%) on TPU_21(process=4,(1,5,0,0))
I0420 06:33:36.968011 135121984083776 metric_logger.py:196] completed step: 0, seconds: 1.147, TFLOP/s/device: 2.960, Tokens/s/device: 446.190, total_weights: 16384, loss: 10.862, lm_loss: 10.862, perplexity: 52153.723
I0420 06:33:37.065304 135121984083776 metric_logger.py:196] completed step: 1, seconds: 15.470, TFLOP/s/device: 0.220, Tokens/s/device: 33.097, total_weights: 16384, loss: 10.862, lm_loss: 10.862, perplexity: 52153.723
I0420 06:33:37.473737 135121984083776 metric_logger.py:196] completed step: 2, seconds: 0.039, TFLOP/s/device: 87.571, Tokens/s/device: 13199.618, total_weights: 16384, loss: 9.763, lm_loss: 9.763, perplexity: 17373.129
I0420 06:33:37.532399 135121984083776 metric_logger.py:196] completed step: 3, seconds: 0.409, TFLOP/s/device: 8.312, Tokens/s/device: 1252.930, total_weights: 16384, loss: 8.823, lm_loss: 8.823, perplexity: 6791.817
I0420 06:33:37.649544 135121984083776 metric_logger.py:196] completed step: 4, seconds: 0.064, TFLOP/s/device: 52.932, Tokens/s/device: 7978.558, total_weights: 16384, loss: 7.947, lm_loss: 7.947, perplexity: 2828.417
I0420 06:33:37.655258 135121984083776 metric_logger.py:196] completed step: 5, seconds: 0.059, TFLOP/s/device: 57.596, Tokens/s/device: 8681.498, total_weights: 16384, loss: 7.212, lm_loss: 7.212, perplexity: 1355.339
I0420 06:33:39.522024    2502 google_auth_provider.cc:181] Running on GCE, using service account 562977990677-compute@developer.gserviceaccount.com
I0420 06:33:41.608364 134997309224704 array_metadata_store.py:203] [process=4][thread=array_type_handler] Wrote 39 array_metadata.ArrayMetadata to gs://lance-maxtext/linen_ckpt_xpk_main_20260420_060028/linen_xpk_main_20260420_060028_09_pdb_lt_1/checkpoints/0/items/array_metadatas/process_4
I0420 06:33:52.066782 135121984083776 metric_logger.py:196] completed step: 6, seconds: 0.117, TFLOP/s/device: 29.009, Tokens/s/device: 4372.518, total_weights: 16384, loss: 6.652, lm_loss: 6.652, perplexity: 774.659
I0420 06:33:52.125154 135121984083776 metric_logger.py:196] completed step: 7, seconds: 14.354, TFLOP/s/device: 0.237, Tokens/s/device: 35.670, total_weights: 16384, loss: 6.269, lm_loss: 6.269, perplexity: 527.696
I0420 06:33:52.183559 135121984083776 metric_logger.py:196] completed step: 8, seconds: 0.063, TFLOP/s/device: 53.661, Tokens/s/device: 8088.340, total_weights: 16384, loss: 6.031, lm_loss: 6.031, perplexity: 416.165
I0420 06:33:52.241223 135121984083776 checkpointing.py:772] Waiting for step 9 to finish before checkpoint...
I0420 06:33:52.241857 135121984083776 checkpointing.py:776] Waited 0.0006487369537353516 seconds for step 9 to finish before starting checkpointing.
I0420 06:33:52.243901 135121984083776 checkpoint_manager.py:2020] [process=4][thread=MainThread][step=0][wait_until_finished] Waiting for Save Finalize thread (save_finalize) to complete.
I0420 06:34:11.785105 134997417813760 base_pytree_checkpoint_handler.py:130] [process=4] /jax/orbax/write/gbytes_per_sec: 44.596 MiB/s (total gbytes: 1.5 GiB) (time elapsed: 35.4202835559845 s) (per-host)
I0420 06:34:11.785224 134997417813760 async_checkpointer.py:90] [process=4][thread=async_save] 3 Handler Commit operations completed. Time taken: 35.148345s.
I0420 06:34:20.315673 134997417813760 async_checkpointer.py:160] [process=4][thread=async_save] Background save thread done. Time taken: 43.678779s.
I0420 06:34:20.315975 134997289707264 async_checkpointer.py:288] [process=4][thread=save_finalize] Done with waiting for background save thread=async_save.
I0420 06:34:20.316082 134997289707264 async_checkpointer.py:298] [process=4][thread=save_finalize] No errors found in background save thread=async_save.
I0420 06:34:20.316130 134997289707264 checkpoint_manager.py:2137] [process=4][thread=save_finalize][step=0] CheckpointManager Save Finalize is syncing with other hosts...
I0420 06:34:20.317592 134997289707264 checkpoint_manager.py:2146] [process=4][thread=save_finalize][step=0] CheckpointManager Save Finalize is done on all hosts.
I0420 06:34:20.317759 135121984083776 checkpoint_manager.py:2032] [process=4][thread=MainThread][step=0][wait_until_finished] Done waiting for Save Finalize thread (save_finalize) running at step=0.
W0420 06:34:20.317892 135121984083776 checkpoint_manager.py:1452] Waiting for previous save to complete took 28.073993 seconds. If this number is high, consider checkpointing less frequently.
I0420 06:34:20.319863 135121984083776 checkpoint_manager.py:1512] [process=4] Saving checkpoint at step 9
I0420 06:34:20.322047 135121984083776 event_tracking.py:70] [process=4] [async] Started save checkpoint @ gs://lance-maxtext/linen_ckpt_xpk_main_20260420_060028/linen_xpk_main_20260420_060028_09_pdb_lt_1/checkpoints/9.
I0420 06:34:20.897707 135121984083776 jax_array_handlers.py:360] Scheduling D2H of 39 prioritized jax.Array.
I0420 06:34:20.897817 135121984083776 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 06:34:20.924247 135121984083776 base_pytree_checkpoint_handler.py:154] [process=4][thread=MainThread] Initiated "orbax.checkpoint._src.serialization.jax_array_handlers.ArrayHandler".serialize. Time taken: 0.027678s
I0420 06:34:20.924403 135121984083776 base_pytree_checkpoint_handler.py:130] [process=4] /jax/orbax/write/blocking_gbytes_per_sec: 49.730 GiB/s (total gbytes: 1.5 GiB) (time elapsed: 0.031019210815429688 s) (per-host)
I0420 06:34:20.924455 135121984083776 base_pytree_checkpoint_handler.py:768] [process=4][thread=MainThread] Initiated Pytree async_save. Time taken: 0.031081s (batch_requests_ready=0.001724s, total_serialization_initiated=0.029288s, others=0.000069s)
I0420 06:34:20.924533 135121984083776 composite_checkpoint_handler.py:715] [process=4][thread=MainThread] Initiated CompositeCheckpointHandler.async_save. Time taken: 0.035440s (all_items=0.000015s, per_item={'items': '0.00001478'}, temp_paths=0.035425)
I0420 06:34:20.925176 135121984083776 event_tracking.py:125] [process=4] [async] Finished blocking save in 0.61 seconds. Continuing save @ gs://lance-maxtext/linen_ckpt_xpk_main_20260420_060028/linen_xpk_main_20260420_060028_09_pdb_lt_1/checkpoints/9.
I0420 06:34:20.925469 134997289707264 async_checkpointer.py:76] [process=4][thread=async_save] Background save thread started. Deadline for this save operation is 2026-04-20 06:54:20.925440
I0420 06:34:20.928016 135121984083776 checkpoint_manager.py:1560] [process=4][thread=MainThread][step=9] Starting CheckpointManager Save Finalize thread=save_finalize
I0420 06:34:20.928291 134990461388544 async_checkpointer.py:280] [process=4][thread=save_finalize] Waiting for background save thread=async_save.
I0420 06:34:20.928465 135121984083776 standard_logger.py:34] {'step': 9, 'event_type': 'save', 'directory': 'gs://lance-maxtext/linen_ckpt_xpk_main_20260420_060028/linen_xpk_main_20260420_060028_09_pdb_lt_1/checkpoints', 'reached_preemption': False, 'preemption_received_at': None, 'synchronous': False, 'wait_for_prev_start_time': 1776666832.2438707, 'wait_for_prev_duration_secs': 28.07399320602417, 'time_between_consecutive_saves_sec': None, 'checkpointer_blocking_start_time': 1776666860.3199012, 'checkpointer_blocking_duration_secs': 0.6056861877441406, 'get_old_steps_start_time': 1776666860.9256096, 'get_old_steps_duration_secs': 2.956390380859375e-05, 'checkpoint_manager_blocking_start_time': 1776666832.2420788, 'checkpoint_manager_blocking_duration_secs': 28.686352491378784}
I0420 06:34:20.928581 135121984083776 checkpointing.py:408] Started an asynchronous checkpoint save for step 9
I0420 06:34:20.928627 135121984083776 checkpoint_manager.py:2020] [process=4][thread=MainThread][step=9][wait_until_finished] Waiting for Save Finalize thread (save_finalize) to complete.
I0420 06:34:26.051063 134997417813760 array_metadata_store.py:203] [process=4][thread=array_type_handler] Wrote 39 array_metadata.ArrayMetadata to gs://lance-maxtext/linen_ckpt_xpk_main_20260420_060028/linen_xpk_main_20260420_060028_09_pdb_lt_1/checkpoints/9/items/array_metadatas/process_4
I0420 06:35:01.925260 134997289707264 base_pytree_checkpoint_handler.py:130] [process=4] /jax/orbax/write/gbytes_per_sec: 38.497 MiB/s (total gbytes: 1.5 GiB) (time elapsed: 41.03185725212097 s) (per-host)
I0420 06:35:01.925378 134997289707264 async_checkpointer.py:90] [process=4][thread=async_save] 3 Handler Commit operations completed. Time taken: 40.999815s.
I0420 06:35:11.060609 134997289707264 async_checkpointer.py:160] [process=4][thread=async_save] Background save thread done. Time taken: 50.135030s.
I0420 06:35:11.060909 134990461388544 async_checkpointer.py:288] [process=4][thread=save_finalize] Done with waiting for background save thread=async_save.
I0420 06:35:11.061045 134990461388544 async_checkpointer.py:298] [process=4][thread=save_finalize] No errors found in background save thread=async_save.
I0420 06:35:11.061108 134990461388544 checkpoint_manager.py:2137] [process=4][thread=save_finalize][step=9] CheckpointManager Save Finalize is syncing with other hosts...
I0420 06:35:11.062370 134990461388544 checkpoint_manager.py:2146] [process=4][thread=save_finalize][step=9] CheckpointManager Save Finalize is done on all hosts.
I0420 06:35:11.062543 135121984083776 checkpoint_manager.py:2032] [process=4][thread=MainThread][step=9][wait_until_finished] Done waiting for Save Finalize thread (save_finalize) running at step=9.
I0420 06:35:11.062695 135121984083776 checkpoint_manager.py:2009] [process=4][thread=MainThread][wait_until_finished] No Save Finalize thread to wait for. Returning.
I0420 06:35:11.063640 135121984083776 metric_logger.py:196] completed step: 9, seconds: 0.058, TFLOP/s/device: 58.451, Tokens/s/device: 8810.421, total_weights: 16384, loss: 5.892, lm_loss: 5.892, perplexity: 361.955
Per train step:
 Total TFLOPs: 3.40 
 split as 93.93% learnable weight flops and 6.07% attention flops
XPK End: Mon Apr 20 06:35:22 UTC 2026
EXIT_CODE=0
test/pipeline-scan-nnx  ·  e27fc1e97  ·  test_pipeline_scan_nnx_20260421_005327  ·  full log 
XPK Start: Tue Apr 21 01:24:44 UTC 2026
2026-04-21 01:24:48.605916: E external/local_xla/xla/stream_executor/cuda/cuda_fft.cc:467] Unable to register cuFFT factory: Attempting to register factory for plugin cuFFT when one has already been registered
WARNING: All log messages before absl::InitializeLog() is called are written to STDERR
E0000 00:00:1776734688.618916      10 cuda_dnn.cc:8579] Unable to register cuDNN factory: Attempting to register factory for plugin cuDNN when one has already been registered
E0000 00:00:1776734688.622614      10 cuda_blas.cc:1407] Unable to register cuBLAS factory: Attempting to register factory for plugin cuBLAS when one has already been registered
W0000 00:00:1776734688.633953      10 computation_placer.cc:177] computation placer already registered. Please check linkage and avoid linking the same target more than once.
W0000 00:00:1776734688.633971      10 computation_placer.cc:177] computation placer already registered. Please check linkage and avoid linking the same target more than once.
W0000 00:00:1776734688.633974      10 computation_placer.cc:177] computation placer already registered. Please check linkage and avoid linking the same target more than once.
W0000 00:00:1776734688.633976      10 computation_placer.cc:177] computation placer already registered. Please check linkage and avoid linking the same target more than once.
2026-04-21 01:25:07.803623: E external/local_xla/xla/stream_executor/cuda/cuda_platform.cc:51] failed call to cuInit: INTERNAL: CUDA error: Failed call to cuInit: UNKNOWN ERROR (303)
I0421 01:25:08.315373 138795075704640 max_utils.py:273] Attempting to initialize the jax distributed system...
INFO:2026-04-21 01:25:17,357:jax._src.distributed:140: Starting JAX distributed service on [::]:8482
I0421 01:25:17.357925 138795075704640 distributed.py:140] Starting JAX distributed service on [::]:8482
INFO:2026-04-21 01:25:17,360:jax._src.distributed:157: Connecting to JAX distributed service on mt-09-pdb-lt-1-c1dr4-slice-job-0-0.mt-09-pdb-lt-1-c1dr4:8482
I0421 01:25:17.360253 138795075704640 distributed.py:157] Connecting to JAX distributed service on mt-09-pdb-lt-1-c1dr4-slice-job-0-0.mt-09-pdb-lt-1-c1dr4:8482
I0421 01:25:18.320336 138795075704640 max_utils.py:284] Jax distributed system initialized!
I0421 01:25:25.080946 138795075704640 max_utils.py:800] System Information: Jax Version: 0.8.1
I0421 01:25:25.081052 138795075704640 max_utils.py:801] System Information: Jaxlib Version: 0.8.1
I0421 01:25:25.081093 138795075704640 max_utils.py:802] System Information: Jax Backend: PJRT C API
TFRT TPU v6 lite
Built on Nov 12 2025 14:16:36 (1762985796) cl/831091709
I0421 01:25:25.081130 138795075704640 train_utils.py:347] WARNING: Sequence packing is essentially ignored for synthetic data. Please use a real dataset to use sequence packing.
I0421 01:25:25.697260 138795075704640 maxtext_utils.py:1398] Num_devices: 32, shape (1, 1, 1, 8, 1, 1, 1, 1, 4, 1, 1, 1, 1)
I0421 01:25:25.697557 138795075704640 checkpointing.py:677] Setting up checkpoint logger...
I0421 01:25:25.697614 138795075704640 checkpointing.py:233] Creating checkpoint manager with ocdbt=True and zarr3=True
I0421 01:25:25.697659 138795075704640 pytree_checkpoint_handler.py:589] save_device_host_concurrent_bytes=None
I0421 01:25:25.698004 138795075704640 base_pytree_checkpoint_handler.py:415] 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 0x7e3b4fbb3740>, enable_pinned_host_transfer=False, save_concurrent_bytes: 96000000000 (89.4 GiB), restore_concurrent_bytes: 96000000000 (89.4 GiB)
I0421 01:25:29.016196 138795075704640 checkpointing.py:265] Enabling policy for fixed interval checkpointing.
I0421 01:25:29.016456 138795075704640 checkpoint_manager.py:709] [process=1][thread=MainThread] CheckpointManager init: checkpointers=None, item_names=('items',), item_handlers={'items': <orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeCheckpointHandler object at 0x7e39f9e661b0>}, handler_registry=None
I0421 01:25:29.016865 138795075704640 composite_checkpoint_handler.py:237] Deferred registration for item: "items". Adding handler `<orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeCheckpointHandler object at 0x7e39f9e661b0>` for item "items" and save args `<class 'orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeSaveArgs'>` and restore args `<class 'orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeRestoreArgs'>` to `_handler_registry`.
I0421 01:25:29.017013 138795075704640 composite_checkpoint_handler.py:237] Deferred registration for item: "metrics". Adding handler `<orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonCheckpointHandler object at 0x7e2797bc35c0>` for item "metrics" and save args `<class 'orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonSaveArgs'>` and restore args `<class 'orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonRestoreArgs'>` to `_handler_registry`.
I0421 01:25:29.017055 138795075704640 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 0x7e39f9e661b0>, ('items', <class 'orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeRestoreArgs'>): <orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeCheckpointHandler object at 0x7e39f9e661b0>, ('metrics', <class 'orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonSaveArgs'>): <orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonCheckpointHandler object at 0x7e2797bc35c0>, ('metrics', <class 'orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonRestoreArgs'>): <orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonCheckpointHandler object at 0x7e2797bc35c0>}).
I0421 01:25:29.017587 138795075704640 abstract_checkpointer.py:35] orbax-checkpoint version: 0.11.33
I0421 01:25:29.017682 138795075704640 async_checkpointer.py:177] [process=1][thread=MainThread] Using barrier_sync_fn: <function get_barrier_sync_fn.<locals>._fn at 0x7e26b4040040> timeout: 600 secs and primary_host=0 for async checkpoint writes
I0421 01:25:30.132915 138795075704640 checkpoint_manager.py:1818] Found 0 checkpoint steps in gs://lance-maxtext/linen_ckpt_xpk_test_pipeline_scan_nnx_20260421_005327/linen_xpk_test_pipeline_scan_nnx_20260421_005327_09_pdb_lt_1/checkpoints
I0421 01:25:30.590529 138795075704640 checkpoint_manager.py:929] [process=1][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/linen_ckpt_xpk_test_pipeline_scan_nnx_20260421_005327/linen_xpk_test_pipeline_scan_nnx_20260421_005327_09_pdb_lt_1/checkpoints: <orbax.checkpoint.checkpoint_manager.CheckpointManager object at 0x7e2797bbcda0>
I0421 01:25:30.590702 138795075704640 checkpointing.py:301] Checkpoint manager created!
I0421 01:25:31.489848 138795075704640 nnx_wrappers.py:453] Unknown Logical: bfloat16[8,2048,2048]....................................... ('activation_batch', 'activation_norm_length', 'activation_embed').
I0421 01:25:31.489959 138795075704640 nnx_wrappers.py:453] Unknown Physical: bfloat16[8,2048,2048]....................................... ('fsdp', None, 'tensor').
I0421 01:25:31.862391 138795075704640 attentions.py:1084] attentions/inputs_q Logical: bfloat16[8,2048,2048]....................................... ('activation_batch', 'activation_attn_length', 'activation_attn_embed').
I0421 01:25:31.862494 138795075704640 attentions.py:1084] attentions/inputs_q Physical: bfloat16[8,2048,2048]....................................... ('fsdp', None, 'tensor').
I0421 01:25:31.878864 138795075704640 attentions.py:1085] attentions/inputs_kv Logical: bfloat16[8,2048,2048]....................................... ('activation_batch', 'activation_attn_length', 'activation_attn_embed').
I0421 01:25:31.878931 138795075704640 attentions.py:1085] attentions/inputs_kv Physical: bfloat16[8,2048,2048]....................................... ('fsdp', None, 'tensor').
I0421 01:25:31.907438 138795075704640 attentions.py:1150] attentions/query Logical: bfloat16[8,2048,16,128]..................................... ('activation_kv_batch', 'activation_attn_length', 'activation_kv_heads', 'activation_kv_head_dim').
I0421 01:25:31.907544 138795075704640 attentions.py:1150] attentions/query Physical: bfloat16[8,2048,16,128]..................................... ('fsdp', None, 'tensor', None).
I0421 01:25:31.924492 138795075704640 attentions.py:1151] attentions/key Logical: bfloat16[8,2048,16,128]..................................... ('activation_kv_batch', 'activation_attn_length', 'activation_kv_heads', 'activation_kv_head_dim').
I0421 01:25:31.924555 138795075704640 attentions.py:1151] attentions/key Physical: bfloat16[8,2048,16,128]..................................... ('fsdp', None, 'tensor', None).
I0421 01:25:31.941072 138795075704640 attentions.py:1152] attentions/value Logical: bfloat16[8,2048,16,128]..................................... ('activation_kv_batch', 'activation_attn_length', 'activation_kv_heads', 'activation_kv_head_dim').
I0421 01:25:31.941140 138795075704640 attentions.py:1152] attentions/value Physical: bfloat16[8,2048,16,128]..................................... ('fsdp', None, 'tensor', None).
I0421 01:25:31.966799 138795075704640 attentions.py:1193] attentions/out Logical: bfloat16[8,2048,16,128]..................................... ('activation_batch', 'activation_attn_length', 'activation_heads', 'activation_kv').
I0421 01:25:31.966871 138795075704640 attentions.py:1193] attentions/out Physical: bfloat16[8,2048,16,128]..................................... ('fsdp', None, 'tensor', None).
I0421 01:25:31.986941 138795075704640 linears.py:525] linears/x Logical: bfloat16[8,2048,7168]....................................... ('activation_batch', 'activation_length', 'activation_mlp').
I0421 01:25:31.987006 138795075704640 linears.py:525] linears/x Physical: bfloat16[8,2048,7168]....................................... ('fsdp', None, 'tensor').
I0421 01:25:32.181922 138795075704640 checkpointing.py:577] checkpoint manager exists so trying to load this run's existing checkpoint
I0421 01:25:32.182028 138795075704640 checkpointing.py:665] No existing checkpoints found, not restoring checkpoint.
fsdp: 8
tensor: 4
I0421 01:25:33.611262 138795075704640 maxtext_utils.py:1501]  params/params/decoder/decoder_norm/scale
    Shape:     float32[2048]
    Logical:   PartitionSpec('norm',)
    Physical:  ('tensor',)
I0421 01:25:33.611393 138795075704640 maxtext_utils.py:1501]  params/params/decoder/layers/mlp/wi_0/kernel
    Shape:     float32[2048,16,7168]
    Logical:   PartitionSpec('embed', 'layers', 'mlp')
    Physical:  ('fsdp', None, 'tensor')
I0421 01:25:33.611467 138795075704640 maxtext_utils.py:1501]  params/params/decoder/layers/mlp/wi_1/kernel
    Shape:     float32[2048,16,7168]
    Logical:   PartitionSpec('embed', 'layers', 'mlp')
    Physical:  ('fsdp', None, 'tensor')
I0421 01:25:33.611528 138795075704640 maxtext_utils.py:1501]  params/params/decoder/layers/mlp/wo/kernel
    Shape:     float32[7168,16,2048]
    Logical:   PartitionSpec('mlp', 'layers', 'embed')
    Physical:  ('tensor', None, 'fsdp')
I0421 01:25:33.611582 138795075704640 maxtext_utils.py:1501]  params/params/decoder/layers/post_self_attention_layer_norm/scale
    Shape:     float32[2048,16]
    Logical:   PartitionSpec('norm', 'layers')
    Physical:  ('tensor', None)
I0421 01:25:33.611622 138795075704640 maxtext_utils.py:1501]  params/params/decoder/layers/pre_self_attention_layer_norm/scale
    Shape:     float32[2048,16]
    Logical:   PartitionSpec('norm', 'layers')
    Physical:  ('tensor', None)
I0421 01:25:33.611675 138795075704640 maxtext_utils.py:1501]  params/params/decoder/layers/self_attention/key/kernel
    Shape:     float32[2048,16,16,128]
    Logical:   PartitionSpec('embed', 'layers', 'kv_heads', 'kv_head_dim')
    Physical:  ('fsdp', None, 'tensor', None)
I0421 01:25:33.611726 138795075704640 maxtext_utils.py:1501]  params/params/decoder/layers/self_attention/out/kernel
    Shape:     float32[16,16,128,2048]
    Logical:   PartitionSpec('heads', 'layers', 'kv', 'embed')
    Physical:  ('tensor', None, None, 'fsdp')
I0421 01:25:33.611767 138795075704640 maxtext_utils.py:1501]  params/params/decoder/layers/self_attention/query/kernel
    Shape:     float32[2048,16,16,128]
    Logical:   PartitionSpec('embed', 'layers', 'q_heads', 'kv')
    Physical:  ('fsdp', None, 'tensor', None)
I0421 01:25:33.611804 138795075704640 maxtext_utils.py:1501]  params/params/decoder/layers/self_attention/value/kernel
    Shape:     float32[2048,16,16,128]
    Logical:   PartitionSpec('embed', 'layers', 'kv_heads', 'kv_head_dim')
    Physical:  ('fsdp', None, 'tensor', None)
I0421 01:25:33.611856 138795075704640 maxtext_utils.py:1501]  params/params/decoder/logits_dense/kernel
    Shape:     float32[2048,32000]
    Logical:   PartitionSpec('embed', 'vocab')
    Physical:  ('fsdp', 'tensor')

I0421 01:25:33.611904 138795075704640 maxtext_utils.py:1501]  params/params/token_embedder/embedding
    Shape:     float32[32000,2048]
    Logical:   PartitionSpec('vocab', 'embed')
    Physical:  ('tensor', 'fsdp')
I0421 01:25:34.117737 138795075704640 train.py:156] train/xent Logical: float32[8,2048]............................................. ('activation_embed_and_logits_batch', 'activation_length').
I0421 01:25:34.117835 138795075704640 train.py:156] train/xent Physical: float32[8,2048]............................................. ('fsdp', None).
I0421 01:25:34.133501 138795075704640 train.py:163] train/z_loss Logical: float32[8,2048]............................................. ('activation_embed_and_logits_batch', 'activation_length').
I0421 01:25:34.133561 138795075704640 train.py:163] train/z_loss Physical: float32[8,2048]............................................. ('fsdp', None).
I0421 01:25:43.920309 138795075704640 max_utils.py:791] Total memory size: 0.9 GB, Output size: 0.4 GB, Temp size: 0.5 GB, Argument size: 0.4 GB, Host temp size: 0.0 GB.
I0421 01:25:43.921199 138795075704640 metric_logger.py:289] number parameters: 1.104 billion
I0421 01:25:45.457610 138795075704640 checkpointing.py:772] Waiting for step 0 to finish before checkpoint...
I0421 01:25:54.259028 138795075704640 checkpointing.py:776] Waited 8.801398038864136 seconds for step 0 to finish before starting checkpointing.
I0421 01:25:54.261614 138795075704640 checkpoint_manager.py:2013] [process=1][thread=MainThread][wait_until_finished] No Save Finalize thread to wait for. Returning.
I0421 01:25:54.263329 138795075704640 checkpoint_manager.py:1518] [process=1] Saving checkpoint at step 0
I0421 01:25:54.266079 138795075704640 async_checkpointer.py:452] [process=1] Started async saving checkpoint to gs://lance-maxtext/linen_ckpt_xpk_test_pipeline_scan_nnx_20260421_005327/linen_xpk_test_pipeline_scan_nnx_20260421_005327_09_pdb_lt_1/checkpoints/0.
I0421 01:25:56.073190 138795075704640 signaling_client.py:364] Using JaxDistributedSignalingClient
I0421 01:25:56.074290 138795075704640 jax_array_handlers.py:358] Scheduling D2H of 39 prioritized jax.Array.
I0421 01:25:56.074347 138795075704640 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
I0421 01:25:56.295672 138795075704640 base_pytree_checkpoint_handler.py:153] [process=1][thread=MainThread] Initiated "orbax.checkpoint._src.serialization.jax_array_handlers.ArrayHandler".serialize. Time taken: 0.222560s
I0421 01:25:56.295842 138795075704640 base_pytree_checkpoint_handler.py:129] [process=1] /jax/checkpoint/write/blocking_gbytes_per_sec: 1.710 GiB/s (total gbytes: 1.5 GiB) (time elapsed: 0.9022216796875 s) (per-host)
I0421 01:25:56.295895 138795075704640 base_pytree_checkpoint_handler.py:737] [process=1][thread=MainThread] Initiated Pytree async_save. Time taken: 0.902287s (batch_requests_ready=0.676064s, total_serialization_initiated=0.226150s, others=0.000073s)
I0421 01:25:56.296011 138795075704640 composite_checkpoint_handler.py:715] [process=1][thread=MainThread] Initiated CompositeCheckpointHandler.async_save. Time taken: 0.906289s (all_items=0.000020s, per_item={'items': '0.00001955'}, temp_paths=0.906269)
I0421 01:25:56.296993 138672229816064 async_checkpointer.py:79] [process=1][thread=async_save] Background save thread started.
I0421 01:25:56.297177 138795075704640 async_checkpointer.py:561] Finished blocking save. Time taken: 2.033767s. Continuing background save to gs://lance-maxtext/linen_ckpt_xpk_test_pipeline_scan_nnx_20260421_005327/linen_xpk_test_pipeline_scan_nnx_20260421_005327_09_pdb_lt_1/checkpoints/0.
I0421 01:25:56.328059 138795075704640 checkpoint_manager.py:1566] [process=1][thread=MainThread][step=0] Starting CheckpointManager Save Finalize thread=save_finalize
I0421 01:25:56.328390 138669545461504 async_checkpointer.py:265] [process=1][thread=save_finalize] Waiting for background save thread=async_save.
I0421 01:25:56.328569 138795075704640 standard_logger.py:34] {'step': 0, 'event_type': 'save', 'directory': 'gs://lance-maxtext/linen_ckpt_xpk_test_pipeline_scan_nnx_20260421_005327/linen_xpk_test_pipeline_scan_nnx_20260421_005327_09_pdb_lt_1/checkpoints', 'reached_preemption': False, 'preemption_received_at': None, 'synchronous': False, 'wait_for_prev_start_time': 1776734754.2615962, 'wait_for_prev_duration_secs': 6.175041198730469e-05, 'time_between_consecutive_saves_sec': None, 'checkpointer_blocking_start_time': 1776734754.2633712, 'checkpointer_blocking_duration_secs': 2.0339102745056152, 'get_old_steps_start_time': 1776734756.2973008, 'get_old_steps_duration_secs': 2.5272369384765625e-05, 'checkpoint_manager_blocking_start_time': 1776734754.2595732, 'checkpoint_manager_blocking_duration_secs': 2.068955898284912}
I0421 01:25:56.328674 138795075704640 checkpointing.py:408] Started an asynchronous checkpoint save for step 0
I0421 01:25:56.328726 138795075704640 max_utils.py:750] 
Memstats: After params initialized:
I0421 01:25:56.328782 138795075704640 max_utils.py:756] 	Using (GB) 0.43 / 31.25 (1.376000%) on TPU_2(process=1,(2,0,0,0))
I0421 01:25:56.328813 138795075704640 max_utils.py:756] 	Using (GB) 0.43 / 31.25 (1.376000%) on TPU_3(process=1,(3,0,0,0))
I0421 01:25:56.328841 138795075704640 max_utils.py:756] 	Using (GB) 0.43 / 31.25 (1.376000%) on TPU_6(process=1,(2,1,0,0))
I0421 01:25:56.328866 138795075704640 max_utils.py:756] 	Using (GB) 0.43 / 31.25 (1.376000%) on TPU_7(process=1,(3,1,0,0))
I0421 01:25:56.589340 138795075704640 metric_logger.py:185] completed step: 0, seconds: 1.536, TFLOP/s/device: 2.211, Tokens/s/device: 333.268, total_weights: 16384, loss: 10.862
I0421 01:25:56.674119 138795075704640 metric_logger.py:185] completed step: 1, seconds: 11.131, TFLOP/s/device: 0.305, Tokens/s/device: 46.000, total_weights: 16384, loss: 10.862
I0421 01:25:57.102787 138795075704640 metric_logger.py:185] completed step: 2, seconds: 0.027, TFLOP/s/device: 127.747, Tokens/s/device: 19255.359, total_weights: 16384, loss: 9.763
I0421 01:25:57.161453 138795075704640 metric_logger.py:185] completed step: 3, seconds: 0.429, TFLOP/s/device: 7.920, Tokens/s/device: 1193.815, total_weights: 16384, loss: 8.823
I0421 01:25:57.279020 138795075704640 metric_logger.py:185] completed step: 4, seconds: 0.064, TFLOP/s/device: 53.429, Tokens/s/device: 8053.353, total_weights: 16384, loss: 7.947
I0421 01:25:57.284608 138795075704640 metric_logger.py:185] completed step: 5, seconds: 0.059, TFLOP/s/device: 58.006, Tokens/s/device: 8743.319, total_weights: 16384, loss: 7.212
I0421 01:26:10.765708 138795075704640 metric_logger.py:185] completed step: 6, seconds: 0.118, TFLOP/s/device: 28.759, Tokens/s/device: 4334.942, total_weights: 16384, loss: 6.652
I0421 01:26:10.824429 138795075704640 metric_logger.py:185] completed step: 7, seconds: 13.423, TFLOP/s/device: 0.253, Tokens/s/device: 38.143, total_weights: 16384, loss: 6.269
I0421 01:26:10.882898 138795075704640 metric_logger.py:185] completed step: 8, seconds: 0.063, TFLOP/s/device: 53.799, Tokens/s/device: 8109.221, total_weights: 16384, loss: 6.031
I0421 01:26:10.941192 138795075704640 checkpointing.py:772] Waiting for step 9 to finish before checkpoint...
I0421 01:26:10.941832 138795075704640 checkpointing.py:776] Waited 0.0006511211395263672 seconds for step 9 to finish before starting checkpointing.
I0421 01:26:10.943798 138795075704640 checkpoint_manager.py:2024] [process=1][thread=MainThread][step=0][wait_until_finished] Waiting for Save Finalize thread (save_finalize) to complete.
I0421 01:26:12.388076    2396 google_auth_provider.cc:181] Running on GCE, using service account 562977990677-compute@developer.gserviceaccount.com
I0421 01:26:14.787727 138672238208768 array_metadata_store.py:203] [process=1][thread=array_type_handler] Wrote 39 array_metadata.ArrayMetadata to gs://lance-maxtext/linen_ckpt_xpk_test_pipeline_scan_nnx_20260421_005327/linen_xpk_test_pipeline_scan_nnx_20260421_005327_09_pdb_lt_1/checkpoints/0/items/array_metadatas/process_1
I0421 01:26:44.955770 138672229816064 base_pytree_checkpoint_handler.py:129] [process=1] /jax/checkpoint/write/gbytes_per_sec: 31.871 MiB/s (total gbytes: 1.5 GiB) (time elapsed: 49.56211185455322 s) (per-host)
I0421 01:26:44.955897 138672229816064 async_checkpointer.py:90] [process=1][thread=async_save] 3 Handler Commit operations completed. Time taken: 48.658788s.
I0421 01:26:55.520773 138672229816064 async_checkpointer.py:144] [process=1][thread=async_save] Background save thread done. Time taken: 59.223644s.
I0421 01:26:55.521076 138669545461504 async_checkpointer.py:273] [process=1][thread=save_finalize] Done with waiting for background save thread=async_save.
I0421 01:26:55.521208 138669545461504 async_checkpointer.py:283] [process=1][thread=save_finalize] No errors found in background save thread=async_save.
I0421 01:26:55.521258 138669545461504 checkpoint_manager.py:2133] [process=1][thread=save_finalize][step=0] CheckpointManager Save Finalize is syncing with other hosts...
I0421 01:26:55.523086 138669545461504 checkpoint_manager.py:2142] [process=1][thread=save_finalize][step=0] CheckpointManager Save Finalize is done on all hosts.
I0421 01:26:55.523259 138795075704640 checkpoint_manager.py:2036] [process=1][thread=MainThread][step=0][wait_until_finished] Done waiting for Save Finalize thread (save_finalize) running at step=0.
W0421 01:26:55.523387 138795075704640 checkpoint_manager.py:1458] Waiting for previous save to complete took 44.579591 seconds. If this number is high, consider checkpointing less frequently.
I0421 01:26:55.525360 138795075704640 checkpoint_manager.py:1518] [process=1] Saving checkpoint at step 9
I0421 01:26:55.528682 138795075704640 async_checkpointer.py:452] [process=1] Started async saving checkpoint to gs://lance-maxtext/linen_ckpt_xpk_test_pipeline_scan_nnx_20260421_005327/linen_xpk_test_pipeline_scan_nnx_20260421_005327_09_pdb_lt_1/checkpoints/9.
I0421 01:26:56.503277 138795075704640 jax_array_handlers.py:358] Scheduling D2H of 39 prioritized jax.Array.
I0421 01:26:56.503376 138795075704640 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
I0421 01:26:56.546857 138795075704640 base_pytree_checkpoint_handler.py:153] [process=1][thread=MainThread] Initiated "orbax.checkpoint._src.serialization.jax_array_handlers.ArrayHandler".serialize. Time taken: 0.044706s
I0421 01:26:56.547014 138795075704640 base_pytree_checkpoint_handler.py:129] [process=1] /jax/checkpoint/write/blocking_gbytes_per_sec: 5.061 GiB/s (total gbytes: 1.5 GiB) (time elapsed: 0.30478358268737793 s) (per-host)
I0421 01:26:56.547066 138795075704640 base_pytree_checkpoint_handler.py:737] [process=1][thread=MainThread] Initiated Pytree async_save. Time taken: 0.304843s (batch_requests_ready=0.258305s, total_serialization_initiated=0.046471s, others=0.000066s)
I0421 01:26:56.547158 138795075704640 composite_checkpoint_handler.py:715] [process=1][thread=MainThread] Initiated CompositeCheckpointHandler.async_save. Time taken: 0.308995s (all_items=0.000015s, per_item={'items': '0.00001502'}, temp_paths=0.308980)
I0421 01:26:56.548016 138669058946816 async_checkpointer.py:79] [process=1][thread=async_save] Background save thread started.
I0421 01:26:56.548177 138795075704640 async_checkpointer.py:561] Finished blocking save. Time taken: 1.022733s. Continuing background save to gs://lance-maxtext/linen_ckpt_xpk_test_pipeline_scan_nnx_20260421_005327/linen_xpk_test_pipeline_scan_nnx_20260421_005327_09_pdb_lt_1/checkpoints/9.
I0421 01:26:56.995819 138795075704640 checkpoint_manager.py:1566] [process=1][thread=MainThread][step=9] Starting CheckpointManager Save Finalize thread=save_finalize
I0421 01:26:56.996213 138669545461504 async_checkpointer.py:265] [process=1][thread=save_finalize] Waiting for background save thread=async_save.
I0421 01:26:56.996382 138795075704640 standard_logger.py:34] {'step': 9, 'event_type': 'save', 'directory': 'gs://lance-maxtext/linen_ckpt_xpk_test_pipeline_scan_nnx_20260421_005327/linen_xpk_test_pipeline_scan_nnx_20260421_005327_09_pdb_lt_1/checkpoints', 'reached_preemption': False, 'preemption_received_at': None, 'synchronous': False, 'wait_for_prev_start_time': 1776734770.9437714, 'wait_for_prev_duration_secs': 44.579590797424316, 'time_between_consecutive_saves_sec': None, 'checkpointer_blocking_start_time': 1776734815.5254114, 'checkpointer_blocking_duration_secs': 1.0228681564331055, 'get_old_steps_start_time': 1776734816.548297, 'get_old_steps_duration_secs': 2.5033950805664062e-05, 'checkpoint_manager_blocking_start_time': 1776734770.9420605, 'checkpoint_manager_blocking_duration_secs': 46.054285526275635}
I0421 01:26:56.996504 138795075704640 checkpointing.py:408] Started an asynchronous checkpoint save for step 9
I0421 01:26:56.996550 138795075704640 checkpoint_manager.py:2024] [process=1][thread=MainThread][step=9][wait_until_finished] Waiting for Save Finalize thread (save_finalize) to complete.
I0421 01:27:03.916422 138672238208768 array_metadata_store.py:203] [process=1][thread=array_type_handler] Wrote 39 array_metadata.ArrayMetadata to gs://lance-maxtext/linen_ckpt_xpk_test_pipeline_scan_nnx_20260421_005327/linen_xpk_test_pipeline_scan_nnx_20260421_005327_09_pdb_lt_1/checkpoints/9/items/array_metadatas/process_1
I0421 01:27:40.178784 138669058946816 base_pytree_checkpoint_handler.py:129] [process=1] /jax/checkpoint/write/gbytes_per_sec: 35.952 MiB/s (total gbytes: 1.5 GiB) (time elapsed: 43.936511516571045 s) (per-host)
I0421 01:27:40.178911 138669058946816 async_checkpointer.py:90] [process=1][thread=async_save] 3 Handler Commit operations completed. Time taken: 43.630780s.
I0421 01:27:47.696953 138669058946816 async_checkpointer.py:144] [process=1][thread=async_save] Background save thread done. Time taken: 51.148806s.
I0421 01:27:47.697203 138669545461504 async_checkpointer.py:273] [process=1][thread=save_finalize] Done with waiting for background save thread=async_save.
I0421 01:27:47.697261 138669545461504 async_checkpointer.py:283] [process=1][thread=save_finalize] No errors found in background save thread=async_save.
I0421 01:27:47.697313 138669545461504 checkpoint_manager.py:2133] [process=1][thread=save_finalize][step=9] CheckpointManager Save Finalize is syncing with other hosts...
I0421 01:27:47.698914 138669545461504 checkpoint_manager.py:2142] [process=1][thread=save_finalize][step=9] CheckpointManager Save Finalize is done on all hosts.
I0421 01:27:47.699073 138795075704640 checkpoint_manager.py:2036] [process=1][thread=MainThread][step=9][wait_until_finished] Done waiting for Save Finalize thread (save_finalize) running at step=9.
I0421 01:27:47.699218 138795075704640 checkpoint_manager.py:2013] [process=1][thread=MainThread][wait_until_finished] No Save Finalize thread to wait for. Returning.
I0421 01:27:47.699845 138795075704640 metric_logger.py:185] completed step: 9, seconds: 0.059, TFLOP/s/device: 58.053, Tokens/s/device: 8750.342, total_weights: 16384, loss: 5.892
Per train step:
 Total TFLOPs: 3.40 
 split as 93.93% learnable weight flops and 6.07% attention flops
XPK End: Tue Apr 21 01:28:00 UTC 2026
EXIT_CODE=0