main (574ad3fb9) vs feat/nnx-post-train-fixes (574ad3fb9)| Metric | main 574ad3fb9 | feat/nnx-post-train-fixes 574ad3fb9 | Diff (feat/nnx-post-train-fixes − main) |
|---|---|---|---|
| Parameters | 1.104 billion | 1.104 billion | — |
| Final loss | 5.3910 | 5.3910 | 0 |
| TFLOP/s | 91.999 | 92.126 | +0.127 |
| Tok/s | 13867.1 | 13886.3 | +19.181 |
| Avg s/step | 8.613 | 8.605 | -0.008 |
| Memory % | 1.38 | 1.38 | 0 |
| JAX | 0.9.2 | 0.9.2 | — |
Diff = branch value − main value. Green = branch improved. Red = branch regressed.
XPK Start: Sat Apr 18 19:55:54 UTC 2026 PyTorch was not found. Models won't be available and only tokenizers, configuration and file/data utilities can be used. 2026-04-18 19:56:18.585836: 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) I0418 19:56:19.115494 134226523375424 max_utils.py:273] Attempting to initialize the jax distributed system... I0418 19:56:28.157152 134226523375424 distributed.py:149] Starting JAX distributed service on [::]:8482 I0418 19:56:28.159672 134226523375424 distributed.py:172] Connecting to JAX distributed service on mt-16-zarr3-false-rs-jkbfq-slice-job-0-0.mt-16-zarr3-false-rs-jkbfq:8482 I0418 19:56:29.891535 134226523375424 max_utils.py:284] Jax distributed system initialized! I0418 19:56:36.038491 134226523375424 max_utils.py:800] System Information: Jax Version: 0.9.2 I0418 19:56:36.038592 134226523375424 max_utils.py:801] System Information: Jaxlib Version: 0.9.2 I0418 19:56:36.038631 134226523375424 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 I0418 19:56:36.038663 134226523375424 train_utils.py:348] WARNING: Sequence packing is essentially ignored for synthetic data. Please use a real dataset to use sequence packing. I0418 19:56:36.720108 134226523375424 maxtext_utils.py:1551] Num_devices: 32, shape (1, 1, 1, 32, 1, 1, 1, 1, 1, 1, 1, 1, 1) I0418 19:56:36.720384 134226523375424 checkpointing.py:677] Setting up checkpoint logger... I0418 19:56:36.720438 134226523375424 checkpointing.py:233] Creating checkpoint manager with ocdbt=True and zarr3=False I0418 19:56:36.720481 134226523375424 pytree_checkpoint_handler.py:592] save_device_host_concurrent_bytes=None I0418 19:56:36.720825 134226523375424 base_pytree_checkpoint_handler.py:441] Created BasePyTreeCheckpointHandler: use_ocdbt=True, use_zarr3=False, pytree_metadata_options=PyTreeMetadataOptions(support_rich_types=False), array_metadata_store=<orbax.checkpoint._src.metadata.array_metadata_store.Store object at 0x7a136e519070>, enable_pinned_host_transfer=False, save_concurrent_bytes: 96000000000 (89.4 GiB), restore_concurrent_bytes: 96000000000 (89.4 GiB) I0418 19:56:38.301365 134226523375424 checkpointing.py:265] Enabling policy for fixed interval checkpointing. I0418 19:56:38.301596 134226523375424 checkpoint_manager.py:708] [process=0][thread=MainThread] CheckpointManager init: checkpointers=None, item_names=('items',), item_handlers={'items': <orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeCheckpointHandler object at 0x79ff281966f0>}, handler_registry=None I0418 19:56:38.301845 134226523375424 composite_checkpoint_handler.py:237] Deferred registration for item: "items". Adding handler `<orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeCheckpointHandler object at 0x79ff281966f0>` 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`. I0418 19:56:38.301893 134226523375424 composite_checkpoint_handler.py:237] Deferred registration for item: "metrics". Adding handler `<orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonCheckpointHandler object at 0x79ff28198b60>` 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`. I0418 19:56:38.301928 134226523375424 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 0x79ff281966f0>, ('items', <class 'orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeRestoreArgs'>): <orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeCheckpointHandler object at 0x79ff281966f0>, ('metrics', <class 'orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonSaveArgs'>): <orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonCheckpointHandler object at 0x79ff28198b60>, ('metrics', <class 'orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonRestoreArgs'>): <orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonCheckpointHandler object at 0x79ff28198b60>}). I0418 19:56:38.302253 134226523375424 abstract_checkpointer.py:35] orbax-checkpoint version: 0.11.34 I0418 19:56:38.302323 134226523375424 async_checkpointer.py:192] [process=0][thread=MainThread] Using barrier_sync_fn: <function get_barrier_sync_fn.<locals>._fn at 0x79fef84db880> timeout: 1200 secs and primary_host=0 for async checkpoint writes I0418 19:56:41.615752 134103978473216 checkpoint.py:228] Read Metadata={'item_handlers': {'items': 'orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeCheckpointHandler'}, 'metrics': {}, 'performance_metrics': {}, 'init_timestamp_nsecs': 1776542030492889275, 'commit_timestamp_nsecs': 1776542077797214067, 'custom_metadata': {}} from gs://lance-maxtext/linen_ckpt_xpk_main_20260418_180002/linen_xpk_main_20260418_180002_16_zarr3_false/checkpoints/9/_CHECKPOINT_METADATA I0418 19:56:41.864759 134103978473216 checkpoint_manager.py:1871] Missing metrics for step 9 I0418 19:56:42.708612 134226523375424 checkpoint_manager.py:1812] Found 3 checkpoint steps in gs://lance-maxtext/linen_ckpt_xpk_main_20260418_180002/linen_xpk_main_20260418_180002_16_zarr3_false/checkpoints I0418 19:56:43.106492 134226523375424 checkpoint_manager.py:929] [process=0][thread=MainThread] CheckpointManager created, primary_host=0, CheckpointManagerOptions=CheckpointManagerOptions(save_interval_steps=1, max_to_keep=None, keep_time_interval=None, keep_period=None, should_keep_fn=None, best_fn=None, best_mode='max', keep_checkpoints_without_metrics=True, step_prefix=None, step_format_fixed_length=None, step_name_format=None, create=True, cleanup_tmp_directories=False, save_on_steps=frozenset(), single_host_load_and_broadcast=False, todelete_subdir=None, todelete_full_path=None, enable_background_delete=False, read_only=False, enable_async_checkpointing=True, async_options=None, multiprocessing_options=MultiprocessingOptions(primary_host=0, active_processes=None, barrier_sync_key_prefix=None), should_save_fn=None, file_options=FileOptions(path_permission_mode=None), save_root_metadata=True, temporary_path_class=None, save_decision_policy=FixedIntervalPolicy(interval=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_20260418_180002/linen_xpk_main_20260418_180002_16_zarr3_false/checkpoints: <orbax.checkpoint.checkpoint_manager.CheckpointManager object at 0x79ff28375880> I0418 19:56:43.106657 134226523375424 checkpointing.py:301] Checkpoint manager created! I0418 19:56:44.011914 134226523375424 nnx_wrappers.py:437] Unknown Logical: bfloat16[32,2048,2048]...................................... ('activation_batch', 'activation_norm_length', 'activation_embed'). I0418 19:56:44.012024 134226523375424 nnx_wrappers.py:437] Unknown Physical: bfloat16[32,2048,2048]...................................... ('fsdp', None, None). I0418 19:56:44.391348 134226523375424 attentions.py:1088] attentions/inputs_q Logical: bfloat16[32,2048,2048]...................................... ('activation_batch', 'activation_attn_length', 'activation_attn_embed'). I0418 19:56:44.391495 134226523375424 attentions.py:1088] attentions/inputs_q Physical: bfloat16[32,2048,2048]...................................... ('fsdp', None, None). I0418 19:56:44.407781 134226523375424 attentions.py:1089] attentions/inputs_kv Logical: bfloat16[32,2048,2048]...................................... ('activation_batch', 'activation_attn_length', 'activation_attn_embed'). I0418 19:56:44.407863 134226523375424 attentions.py:1089] attentions/inputs_kv Physical: bfloat16[32,2048,2048]...................................... ('fsdp', None, None). I0418 19:56:44.431795 134226523375424 attentions.py:1154] attentions/query Logical: bfloat16[32,2048,16,128].................................... ('activation_kv_batch', 'activation_attn_length', 'activation_kv_heads', 'activation_kv_head_dim'). I0418 19:56:44.431871 134226523375424 attentions.py:1154] attentions/query Physical: bfloat16[32,2048,16,128].................................... ('fsdp', None, None, None). I0418 19:56:44.448015 134226523375424 attentions.py:1155] attentions/key Logical: bfloat16[32,2048,16,128].................................... ('activation_kv_batch', 'activation_attn_length', 'activation_kv_heads', 'activation_kv_head_dim'). I0418 19:56:44.448076 134226523375424 attentions.py:1155] attentions/key Physical: bfloat16[32,2048,16,128].................................... ('fsdp', None, None, None). I0418 19:56:44.464211 134226523375424 attentions.py:1156] attentions/value Logical: bfloat16[32,2048,16,128].................................... ('activation_kv_batch', 'activation_attn_length', 'activation_kv_heads', 'activation_kv_head_dim'). I0418 19:56:44.464271 134226523375424 attentions.py:1156] attentions/value Physical: bfloat16[32,2048,16,128].................................... ('fsdp', None, None, None). I0418 19:56:44.488595 134226523375424 attentions.py:1197] attentions/out Logical: bfloat16[32,2048,16,128].................................... ('activation_batch', 'activation_attn_length', 'activation_heads', 'activation_kv'). I0418 19:56:44.488664 134226523375424 attentions.py:1197] attentions/out Physical: bfloat16[32,2048,16,128].................................... ('fsdp', None, None, None). I0418 19:56:44.509698 134226523375424 linears.py:525] linears/x Logical: bfloat16[32,2048,7168]...................................... ('activation_batch', 'activation_length', 'activation_mlp'). I0418 19:56:44.509761 134226523375424 linears.py:525] linears/x Physical: bfloat16[32,2048,7168]...................................... ('fsdp', None, None). I0418 19:56:44.724377 134226523375424 checkpointing.py:577] checkpoint manager exists so trying to load this run's existing checkpoint I0418 19:56:44.724479 134226523375424 checkpointing.py:581] restoring from this run's directory step 9 I0418 19:56:45.325993 134226523375424 event_tracking.py:70] [process=0] [sync] Started load checkpoint @ gs://lance-maxtext/linen_ckpt_xpk_main_20260418_180002/linen_xpk_main_20260418_180002_16_zarr3_false/checkpoints/9. I0418 19:56:45.572578 134226523375424 checkpointer.py:307] Restoring checkpoint from gs://lance-maxtext/linen_ckpt_xpk_main_20260418_180002/linen_xpk_main_20260418_180002_16_zarr3_false/checkpoints/9. I0418 19:56:45.573018 134226523375424 event_tracking.py:125] [process=0] [sync] Finished blocking load in 0.25 seconds. Continuing load @ gs://lance-maxtext/linen_ckpt_xpk_main_20260418_180002/linen_xpk_main_20260418_180002_16_zarr3_false/checkpoints/9. I0418 19:56:47.000221 1463 google_auth_provider.cc:181] Running on GCE, using service account 562977990677-compute@developer.gserviceaccount.com I0418 19:56:53.658679 134226523375424 jax_array_handlers.py:843] [process=0] /jax/orbax/read/worker/io/requested throughput: 237.127 MiB/s (total gbytes: 1.5 GiB) (time elapsed: 6.6642420291900635 s) (per-host) I0418 19:56:53.663620 134226523375424 base_pytree_checkpoint_handler.py:130] [process=0] /jax/checkpoint/read/gbytes_per_sec: 223.152 MiB/s (total gbytes: 1.5 GiB) (time elapsed: 7.091994047164917 s) (per-host) I0418 19:56:56.141288 134226523375424 event_tracking.py:138] [process=0] [sync] Finished load in 10.82 seconds @ gs://lance-maxtext/linen_ckpt_xpk_main_20260418_180002/linen_xpk_main_20260418_180002_16_zarr3_false/checkpoints/9 I0418 19:56:56.141445 134226523375424 standard_logger.py:34] {'step': 9, 'event_type': 'restore', 'directory': 'gs://lance-maxtext/linen_ckpt_xpk_main_20260418_180002/linen_xpk_main_20260418_180002_16_zarr3_false/checkpoints', 'checkpointer_start_time': 1776542205.3258939, 'checkpointer_duration_secs': 10.815519571304321, 'checkpoint_manager_start_time': 1776542204.7247267, 'checkpoint_manager_duration_secs': 11.416688919067383} fsdp: 32 I0418 19:56:56.551715 134226523375424 maxtext_utils.py:1654] params/params/decoder/decoder_norm/scale Shape: float32[2048] Logical: P('norm',) Physical: (None,) I0418 19:56:56.551851 134226523375424 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, None) I0418 19:56:56.551905 134226523375424 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, None) I0418 19:56:56.551961 134226523375424 maxtext_utils.py:1654] params/params/decoder/layers/mlp/wo/kernel Shape: float32[7168,16,2048] Logical: P('mlp', 'layers', 'embed') Physical: (None, None, 'fsdp') I0418 19:56:56.552013 134226523375424 maxtext_utils.py:1654] params/params/decoder/layers/post_self_attention_layer_norm/scale Shape: float32[2048,16] Logical: P('norm', 'layers') Physical: (None, None) I0418 19:56:56.552050 134226523375424 maxtext_utils.py:1654] params/params/decoder/layers/pre_self_attention_layer_norm/scale Shape: float32[2048,16] Logical: P('norm', 'layers') Physical: (None, None) I0418 19:56:56.552100 134226523375424 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, None, None) I0418 19:56:56.552155 134226523375424 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: (None, None, None, 'fsdp') I0418 19:56:56.552196 134226523375424 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, None, None) I0418 19:56:56.552232 134226523375424 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, None, None) I0418 19:56:56.552277 134226523375424 maxtext_utils.py:1654] params/params/decoder/logits_dense/kernel Shape: float32[2048,32000] Logical: P('embed_vocab', 'vocab') Physical: ('fsdp', None) I0418 19:56:56.552324 134226523375424 maxtext_utils.py:1654] params/params/token_embedder/embedding Shape: float32[32000,2048] Logical: P('vocab', 'embed_vocab') Physical: (None, 'fsdp') I0418 19:56:57.040140 134226523375424 train.py:155] train/xent Logical: float32[32,2048]............................................ ('activation_embed_and_logits_batch', 'activation_length'). I0418 19:56:57.040234 134226523375424 train.py:155] train/xent Physical: float32[32,2048]............................................ ('fsdp', None). I0418 19:56:57.055527 134226523375424 train.py:162] train/z_loss Logical: float32[32,2048]............................................ ('activation_embed_and_logits_batch', 'activation_length'). I0418 19:56:57.055586 134226523375424 train.py:162] train/z_loss Physical: float32[32,2048]............................................ ('fsdp', None). I0418 19:57:08.054440 134226523375424 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. I0418 19:57:08.062511 134226523375424 metric_logger.py:301] number parameters: 1.104 billion I0418 19:57:19.559443 134226523375424 checkpointing.py:772] Waiting for step 10 to finish before checkpoint... I0418 19:57:19.715522 134226523375424 checkpointing.py:776] Waited 0.15605616569519043 seconds for step 10 to finish before starting checkpointing. I0418 19:57:19.718290 134226523375424 checkpoint_manager.py:2009] [process=0][thread=MainThread][wait_until_finished] No Save Finalize thread to wait for. Returning. I0418 19:57:19.720352 134226523375424 checkpoint_manager.py:1512] [process=0] Saving checkpoint at step 10 I0418 19:57:19.722018 134226523375424 event_tracking.py:70] [process=0] [async] Started save checkpoint @ gs://lance-maxtext/linen_ckpt_xpk_main_20260418_180002/linen_xpk_main_20260418_180002_16_zarr3_false/checkpoints/10. I0418 19:57:20.465840 134226523375424 signaling_client.py:364] Using JaxDistributedSignalingClient I0418 19:57:20.478746 134226523375424 jax_array_handlers.py:360] Scheduling D2H of 39 prioritized jax.Array. I0418 19:57:20.478824 134226523375424 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 I0418 19:57:20.738430 134102850189056 atomicity.py:140] Creating tmp directory gs://lance-maxtext/linen_ckpt_xpk_main_20260418_180002/linen_xpk_main_20260418_180002_16_zarr3_false/checkpoints/10 I0418 19:57:20.753794 134226523375424 base_pytree_checkpoint_handler.py:154] [process=0][thread=MainThread] Initiated "orbax.checkpoint._src.serialization.jax_array_handlers.ArrayHandler".serialize. Time taken: 0.275773s I0418 19:57:20.754633 134226523375424 base_pytree_checkpoint_handler.py:130] [process=0] /jax/orbax/write/blocking_gbytes_per_sec: 5.495 GiB/s (total gbytes: 1.5 GiB) (time elapsed: 0.28072667121887207 s) (per-host) I0418 19:57:20.754710 134226523375424 base_pytree_checkpoint_handler.py:768] [process=0][thread=MainThread] Initiated Pytree async_save. Time taken: 0.280828s (batch_requests_ready=0.002150s, total_serialization_initiated=0.277927s, others=0.000752s) I0418 19:57:20.754788 134226523375424 composite_checkpoint_handler.py:715] [process=0][thread=MainThread] Initiated CompositeCheckpointHandler.async_save. Time taken: 0.284862s (all_items=0.000018s, per_item={'items': '0.00001812'}, temp_paths=0.284843) I0418 19:57:20.755528 134226523375424 event_tracking.py:125] [process=0] [async] Finished blocking save in 1.04 seconds. Continuing save @ gs://lance-maxtext/linen_ckpt_xpk_main_20260418_180002/linen_xpk_main_20260418_180002_16_zarr3_false/checkpoints/10. I0418 19:57:20.755888 134103944902400 async_checkpointer.py:76] [process=0][thread=async_save] Background save thread started. Deadline for this save operation is 2026-04-18 20:17:20.755850 I0418 19:57:20.764258 134226523375424 checkpoint_manager.py:1560] [process=0][thread=MainThread][step=10] Starting CheckpointManager Save Finalize thread=save_finalize I0418 19:57:20.764524 134098592978688 async_checkpointer.py:280] [process=0][thread=save_finalize] Waiting for background save thread=async_save. I0418 19:57:20.764672 134226523375424 standard_logger.py:34] {'step': 10, 'event_type': 'save', 'directory': 'gs://lance-maxtext/linen_ckpt_xpk_main_20260418_180002/linen_xpk_main_20260418_180002_16_zarr3_false/checkpoints', 'reached_preemption': False, 'preemption_received_at': None, 'synchronous': False, 'wait_for_prev_start_time': 1776542239.718272, 'wait_for_prev_duration_secs': 6.365776062011719e-05, 'time_between_consecutive_saves_sec': None, 'checkpointer_blocking_start_time': 1776542239.7203898, 'checkpointer_blocking_duration_secs': 1.0356519222259521, 'get_old_steps_start_time': 1776542240.7560658, 'get_old_steps_duration_secs': 2.956390380859375e-05, 'checkpoint_manager_blocking_start_time': 1776542239.716053, 'checkpoint_manager_blocking_duration_secs': 1.048583984375} I0418 19:57:20.764771 134226523375424 checkpointing.py:408] Started an asynchronous checkpoint save for step 10 I0418 19:57:20.764832 134226523375424 max_utils.py:750] Memstats: After params initialized: I0418 19:57:20.764881 134226523375424 max_utils.py:756] Using (GB) 0.43 / 31.25 (1.376000%) on TPU_0(process=0,(0,0,0,0)) I0418 19:57:20.764915 134226523375424 max_utils.py:756] Using (GB) 0.43 / 31.25 (1.376000%) on TPU_1(process=0,(1,0,0,0)) I0418 19:57:20.764942 134226523375424 max_utils.py:756] Using (GB) 0.43 / 31.25 (1.376000%) on TPU_4(process=0,(0,1,0,0)) I0418 19:57:20.764966 134226523375424 max_utils.py:756] Using (GB) 0.43 / 31.25 (1.376000%) on TPU_5(process=0,(1,1,0,0)) I0418 19:57:21.116702 134226523375424 metric_logger.py:196] completed step: 10, seconds: 11.182, TFLOP/s/device: 1.215, Tokens/s/device: 183.153, total_weights: 65536, loss: 6.833, lm_loss: 6.833, perplexity: 927.869 I0418 19:57:21.261960 134226523375424 metric_logger.py:196] completed step: 11, seconds: 1.555, TFLOP/s/device: 8.736, Tokens/s/device: 1316.719, total_weights: 65536, loss: 6.527, lm_loss: 6.527, perplexity: 683.659 I0418 19:57:21.409368 134226523375424 metric_logger.py:196] completed step: 12, seconds: 0.011, TFLOP/s/device: 1260.519, Tokens/s/device: 189999.072, total_weights: 65536, loss: 6.265, lm_loss: 6.265, perplexity: 526.068 I0418 19:57:21.556746 134226523375424 metric_logger.py:196] completed step: 13, seconds: 0.146, TFLOP/s/device: 93.276, Tokens/s/device: 14059.561, total_weights: 65536, loss: 6.045, lm_loss: 6.045, perplexity: 422.001 I0418 19:57:21.565860 134226523375424 checkpointing.py:772] Waiting for step 15 to finish before checkpoint... I0418 19:57:21.850849 134226523375424 checkpointing.py:776] Waited 0.28478002548217773 seconds for step 15 to finish before starting checkpointing. I0418 19:57:21.854637 134226523375424 checkpoint_manager.py:2020] [process=0][thread=MainThread][step=10][wait_until_finished] Waiting for Save Finalize thread (save_finalize) to complete. I0418 19:57:22.957235 134100180653824 atomicity.py:140] Creating tmp directory gs://lance-maxtext/linen_ckpt_xpk_main_20260418_180002/linen_xpk_main_20260418_180002_16_zarr3_false/checkpoints/10/items I0418 19:57:23.074592 134098584585984 checkpoint.py:188] Wrote Metadata={'item_handlers': None, 'metrics': {}, 'performance_metrics': {}, 'init_timestamp_nsecs': 1776542242682333868, 'commit_timestamp_nsecs': None, 'custom_metadata': {}}, json={"item_handlers": null, "metrics": {}, "performance_metrics": {}, "init_timestamp_nsecs": 1776542242682333868, "commit_timestamp_nsecs": null, "custom_metadata": {}} to gs://lance-maxtext/linen_ckpt_xpk_main_20260418_180002/linen_xpk_main_20260418_180002_16_zarr3_false/checkpoints/10/_CHECKPOINT_METADATA I0418 19:57:26.623372 134098609764096 array_metadata_store.py:203] [process=0][thread=array_type_handler] Wrote 39 array_metadata.ArrayMetadata to gs://lance-maxtext/linen_ckpt_xpk_main_20260418_180002/linen_xpk_main_20260418_180002_16_zarr3_false/checkpoints/10/items/array_metadatas/process_0 I0418 19:58:06.058773 134098601371392 base_pytree_checkpoint_handler.py:1282] [process=0][thread=write_metadata_after_commits] Commit + Array metadata written. Time taken: 41.595345s (commit=40.378070s, array_metadata_write=1.217275s) I0418 19:58:06.060342 134103944902400 base_pytree_checkpoint_handler.py:130] [process=0] /jax/orbax/write/gbytes_per_sec: 34.651 MiB/s (total gbytes: 1.5 GiB) (time elapsed: 45.58639669418335 s) (per-host) I0418 19:58:06.060513 134103944902400 async_checkpointer.py:90] [process=0][thread=async_save] 3 Handler Commit operations completed. Time taken: 45.304507s. I0418 19:58:08.377839 134103944902400 array_metadata_store.py:411] [process=0][thread=async_save] Validated ArrayMetadata from all 8 hosts. Time taken: 0.000400s. I0418 19:58:08.529413 134098584585984 checkpoint.py:247] Updated Metadata={'item_handlers': {'items': 'orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeCheckpointHandler'}, 'metrics': {}, 'performance_metrics': {}, 'init_timestamp_nsecs': 1776542242682333868, 'commit_timestamp_nsecs': None, 'custom_metadata': {}} to gs://lance-maxtext/linen_ckpt_xpk_main_20260418_180002/linen_xpk_main_20260418_180002_16_zarr3_false/checkpoints/10/_CHECKPOINT_METADATA I0418 19:58:09.283387 134103944902400 base_pytree_checkpoint_handler.py:1406] [process=0][thread=async_save] Pytree save finalize (merge_ocdbt + ArrayMetadata validation) completed. Time taken: 1.934017s. use_zarr3=False, enable_post_merge_validation=True, directory=gs://lance-maxtext/linen_ckpt_xpk_main_20260418_180002/linen_xpk_main_20260418_180002_16_zarr3_false/checkpoints/10/items I0418 19:58:09.285452 134103944902400 atomicity.py:666] Finalizing gs://lance-maxtext/linen_ckpt_xpk_main_20260418_180002/linen_xpk_main_20260418_180002_16_zarr3_false/checkpoints/10/items I0418 19:58:10.242336 134103944902400 atomicity.py:666] Finalizing gs://lance-maxtext/linen_ckpt_xpk_main_20260418_180002/linen_xpk_main_20260418_180002_16_zarr3_false/checkpoints/10 I0418 19:58:12.061011 134103944902400 atomicity.py:847] [process=0][thread=async_save] Finished saving checkpoint (finalized tmp dir) to `gs://lance-maxtext/linen_ckpt_xpk_main_20260418_180002/linen_xpk_main_20260418_180002_16_zarr3_false/checkpoints/10`. I0418 19:58:12.062125 134103944902400 event_tracking.py:138] [process=0] [async] Finished save (blocking + background) in 52.34 seconds @ gs://lance-maxtext/linen_ckpt_xpk_main_20260418_180002/linen_xpk_main_20260418_180002_16_zarr3_false/checkpoints/10 I0418 19:58:12.063731 134103944902400 async_checkpointer.py:160] [process=0][thread=async_save] Background save thread done. Time taken: 51.307721s. I0418 19:58:12.064057 134098592978688 async_checkpointer.py:288] [process=0][thread=save_finalize] Done with waiting for background save thread=async_save. I0418 19:58:12.064234 134098592978688 async_checkpointer.py:298] [process=0][thread=save_finalize] No errors found in background save thread=async_save. I0418 19:58:12.064345 134098592978688 checkpoint_manager.py:2137] [process=0][thread=save_finalize][step=10] CheckpointManager Save Finalize is syncing with other hosts... I0418 19:58:12.065942 134098592978688 checkpoint_manager.py:2146] [process=0][thread=save_finalize][step=10] CheckpointManager Save Finalize is done on all hosts. I0418 19:58:12.066217 134226523375424 checkpoint_manager.py:2032] [process=0][thread=MainThread][step=10][wait_until_finished] Done waiting for Save Finalize thread (save_finalize) running at step=10. W0418 19:58:12.066344 134226523375424 checkpoint_manager.py:1452] Waiting for previous save to complete took 50.211715 seconds. If this number is high, consider checkpointing less frequently. I0418 19:58:12.068187 134226523375424 checkpoint_manager.py:1512] [process=0] Saving checkpoint at step 15 I0418 19:58:12.070496 134226523375424 event_tracking.py:70] [process=0] [async] Started save checkpoint @ gs://lance-maxtext/linen_ckpt_xpk_main_20260418_180002/linen_xpk_main_20260418_180002_16_zarr3_false/checkpoints/15. I0418 19:58:12.797055 134226523375424 jax_array_handlers.py:360] Scheduling D2H of 39 prioritized jax.Array. I0418 19:58:12.797197 134226523375424 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 I0418 19:58:12.826432 134226523375424 base_pytree_checkpoint_handler.py:154] [process=0][thread=MainThread] Initiated "orbax.checkpoint._src.serialization.jax_array_handlers.ArrayHandler".serialize. Time taken: 0.030430s I0418 19:58:12.827342 134226523375424 base_pytree_checkpoint_handler.py:130] [process=0] /jax/orbax/write/blocking_gbytes_per_sec: 39.641 GiB/s (total gbytes: 1.5 GiB) (time elapsed: 0.03891348838806152 s) (per-host) I0418 19:58:12.827439 134226523375424 base_pytree_checkpoint_handler.py:768] [process=0][thread=MainThread] Initiated Pytree async_save. Time taken: 0.039033s (batch_requests_ready=0.004020s, total_serialization_initiated=0.034179s, others=0.000833s) I0418 19:58:12.827556 134226523375424 composite_checkpoint_handler.py:715] [process=0][thread=MainThread] Initiated CompositeCheckpointHandler.async_save. Time taken: 0.043999s (all_items=0.000023s, per_item={'items': '0.00002265'}, temp_paths=0.043977) I0418 19:58:12.828488 134226523375424 event_tracking.py:125] [process=0] [async] Finished blocking save in 0.76 seconds. Continuing save @ gs://lance-maxtext/linen_ckpt_xpk_main_20260418_180002/linen_xpk_main_20260418_180002_16_zarr3_false/checkpoints/15. I0418 19:58:12.828870 134098592978688 async_checkpointer.py:76] [process=0][thread=async_save] Background save thread started. Deadline for this save operation is 2026-04-18 20:18:12.828833 I0418 19:58:12.831358 134226523375424 checkpoint_manager.py:1560] [process=0][thread=MainThread][step=15] Starting CheckpointManager Save Finalize thread=save_finalize I0418 19:58:12.831667 134098609764096 async_checkpointer.py:280] [process=0][thread=save_finalize] Waiting for background save thread=async_save. I0418 19:58:12.831858 134226523375424 standard_logger.py:34] {'step': 15, 'event_type': 'save', 'directory': 'gs://lance-maxtext/linen_ckpt_xpk_main_20260418_180002/linen_xpk_main_20260418_180002_16_zarr3_false/checkpoints', 'reached_preemption': False, 'preemption_received_at': None, 'synchronous': False, 'wait_for_prev_start_time': 1776542241.8545923, 'wait_for_prev_duration_secs': 50.21171522140503, 'time_between_consecutive_saves_sec': None, 'checkpointer_blocking_start_time': 1776542292.0682523, 'checkpointer_blocking_duration_secs': 0.7607996463775635, 'get_old_steps_start_time': 1776542292.829082, 'get_old_steps_duration_secs': 5.030632019042969e-05, 'checkpoint_manager_blocking_start_time': 1776542241.8515317, 'checkpoint_manager_blocking_duration_secs': 50.980263471603394} I0418 19:58:12.832008 134226523375424 checkpointing.py:408] Started an asynchronous checkpoint save for step 15 I0418 19:58:13.025667 134103944902400 atomicity.py:140] Creating tmp directory gs://lance-maxtext/linen_ckpt_xpk_main_20260418_180002/linen_xpk_main_20260418_180002_16_zarr3_false/checkpoints/15 I0418 19:58:16.069210 134100203591424 atomicity.py:140] Creating tmp directory gs://lance-maxtext/linen_ckpt_xpk_main_20260418_180002/linen_xpk_main_20260418_180002_16_zarr3_false/checkpoints/15/items I0418 19:58:19.752691 134100180653824 array_metadata_store.py:203] [process=0][thread=array_type_handler] Wrote 39 array_metadata.ArrayMetadata to gs://lance-maxtext/linen_ckpt_xpk_main_20260418_180002/linen_xpk_main_20260418_180002_16_zarr3_false/checkpoints/15/items/array_metadatas/process_0 I0418 19:58:34.201574 134226523375424 metric_logger.py:196] completed step: 14, seconds: 0.147, TFLOP/s/device: 92.343, Tokens/s/device: 13918.906, total_weights: 65536, loss: 5.864, lm_loss: 5.864, perplexity: 352.193 I0418 19:58:34.210745 134226523375424 metric_logger.py:196] completed step: 15, seconds: 0.147, TFLOP/s/device: 92.238, Tokens/s/device: 13903.126, total_weights: 65536, loss: 5.719, lm_loss: 5.719, perplexity: 304.661 I0418 19:58:34.356463 134226523375424 metric_logger.py:196] completed step: 16, seconds: 72.643, TFLOP/s/device: 0.187, Tokens/s/device: 28.193, total_weights: 65536, loss: 5.605, lm_loss: 5.605, perplexity: 271.872 I0418 19:58:34.504003 134226523375424 metric_logger.py:196] completed step: 17, seconds: 0.007, TFLOP/s/device: 1898.174, Tokens/s/device: 286113.440, total_weights: 65536, loss: 5.517, lm_loss: 5.517, perplexity: 248.975 I0418 19:58:34.651571 134226523375424 metric_logger.py:196] completed step: 18, seconds: 0.147, TFLOP/s/device: 92.597, Tokens/s/device: 13957.229, total_weights: 65536, loss: 5.448, lm_loss: 5.448, perplexity: 232.333 I0418 19:58:34.798094 134226523375424 checkpointing.py:772] Waiting for step 19 to finish before checkpoint... I0418 19:58:34.798763 134226523375424 checkpointing.py:776] Waited 0.0006899833679199219 seconds for step 19 to finish before starting checkpointing. I0418 19:58:34.800886 134226523375424 checkpoint_manager.py:2020] [process=0][thread=MainThread][step=15][wait_until_finished] Waiting for Save Finalize thread (save_finalize) to complete. I0418 19:58:59.051157 134100172261120 base_pytree_checkpoint_handler.py:1282] [process=0][thread=write_metadata_after_commits] Commit + Array metadata written. Time taken: 41.872876s (commit=40.242225s, array_metadata_write=1.630651s) I0418 19:58:59.052181 134098592978688 base_pytree_checkpoint_handler.py:130] [process=0] /jax/orbax/write/gbytes_per_sec: 34.143 MiB/s (total gbytes: 1.5 GiB) (time elapsed: 46.26373052597046 s) (per-host) I0418 19:58:59.052311 134098592978688 async_checkpointer.py:90] [process=0][thread=async_save] 3 Handler Commit operations completed. Time taken: 46.223294s. I0418 19:59:01.276639 134098592978688 array_metadata_store.py:411] [process=0][thread=async_save] Validated ArrayMetadata from all 8 hosts. Time taken: 0.000226s. I0418 19:59:01.903919 134098592978688 base_pytree_checkpoint_handler.py:1406] [process=0][thread=async_save] Pytree save finalize (merge_ocdbt + ArrayMetadata validation) completed. Time taken: 1.929476s. use_zarr3=False, enable_post_merge_validation=True, directory=gs://lance-maxtext/linen_ckpt_xpk_main_20260418_180002/linen_xpk_main_20260418_180002_16_zarr3_false/checkpoints/15/items I0418 19:59:01.905708 134098592978688 atomicity.py:666] Finalizing gs://lance-maxtext/linen_ckpt_xpk_main_20260418_180002/linen_xpk_main_20260418_180002_16_zarr3_false/checkpoints/15/items I0418 19:59:02.496953 134098592978688 atomicity.py:666] Finalizing gs://lance-maxtext/linen_ckpt_xpk_main_20260418_180002/linen_xpk_main_20260418_180002_16_zarr3_false/checkpoints/15 I0418 19:59:03.896874 134098592978688 atomicity.py:847] [process=0][thread=async_save] Finished saving checkpoint (finalized tmp dir) to `gs://lance-maxtext/linen_ckpt_xpk_main_20260418_180002/linen_xpk_main_20260418_180002_16_zarr3_false/checkpoints/15`. I0418 19:59:03.897704 134098592978688 event_tracking.py:138] [process=0] [async] Finished save (blocking + background) in 51.83 seconds @ gs://lance-maxtext/linen_ckpt_xpk_main_20260418_180002/linen_xpk_main_20260418_180002_16_zarr3_false/checkpoints/15 I0418 19:59:03.899243 134098592978688 async_checkpointer.py:160] [process=0][thread=async_save] Background save thread done. Time taken: 51.070223s. I0418 19:59:03.899484 134098609764096 async_checkpointer.py:288] [process=0][thread=save_finalize] Done with waiting for background save thread=async_save. I0418 19:59:03.899611 134098609764096 async_checkpointer.py:298] [process=0][thread=save_finalize] No errors found in background save thread=async_save. I0418 19:59:03.899678 134098609764096 checkpoint_manager.py:2137] [process=0][thread=save_finalize][step=15] CheckpointManager Save Finalize is syncing with other hosts... I0418 19:59:03.901486 134098609764096 checkpoint_manager.py:2146] [process=0][thread=save_finalize][step=15] CheckpointManager Save Finalize is done on all hosts. I0418 19:59:03.901650 134226523375424 checkpoint_manager.py:2032] [process=0][thread=MainThread][step=15][wait_until_finished] Done waiting for Save Finalize thread (save_finalize) running at step=15. W0418 19:59:03.901776 134226523375424 checkpoint_manager.py:1452] Waiting for previous save to complete took 29.100894 seconds. If this number is high, consider checkpointing less frequently. I0418 19:59:03.903314 134226523375424 checkpoint_manager.py:1512] [process=0] Saving checkpoint at step 19 I0418 19:59:03.905287 134226523375424 event_tracking.py:70] [process=0] [async] Started save checkpoint @ gs://lance-maxtext/linen_ckpt_xpk_main_20260418_180002/linen_xpk_main_20260418_180002_16_zarr3_false/checkpoints/19. I0418 19:59:05.033886 134226523375424 jax_array_handlers.py:360] Scheduling D2H of 39 prioritized jax.Array. I0418 19:59:05.033980 134226523375424 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 I0418 19:59:05.059832 134226523375424 base_pytree_checkpoint_handler.py:154] [process=0][thread=MainThread] Initiated "orbax.checkpoint._src.serialization.jax_array_handlers.ArrayHandler".serialize. Time taken: 0.026871s I0418 19:59:05.060689 134226523375424 base_pytree_checkpoint_handler.py:130] [process=0] /jax/orbax/write/blocking_gbytes_per_sec: 49.244 GiB/s (total gbytes: 1.5 GiB) (time elapsed: 0.03132486343383789 s) (per-host) I0418 19:59:05.060763 134226523375424 base_pytree_checkpoint_handler.py:768] [process=0][thread=MainThread] Initiated Pytree async_save. Time taken: 0.031419s (batch_requests_ready=0.001786s, total_serialization_initiated=0.028862s, others=0.000771s) I0418 19:59:05.060857 134226523375424 composite_checkpoint_handler.py:715] [process=0][thread=MainThread] Initiated CompositeCheckpointHandler.async_save. Time taken: 0.035501s (all_items=0.000016s, per_item={'items': '0.00001621'}, temp_paths=0.035485) I0418 19:59:05.061496 134226523375424 event_tracking.py:125] [process=0] [async] Finished blocking save in 1.16 seconds. Continuing save @ gs://lance-maxtext/linen_ckpt_xpk_main_20260418_180002/linen_xpk_main_20260418_180002_16_zarr3_false/checkpoints/19. I0418 19:59:05.061822 134100203591424 async_checkpointer.py:76] [process=0][thread=async_save] Background save thread started. Deadline for this save operation is 2026-04-18 20:19:05.061765 I0418 19:59:05.063438 134226523375424 checkpoint_manager.py:1560] [process=0][thread=MainThread][step=19] Starting CheckpointManager Save Finalize thread=save_finalize I0418 19:59:05.063709 134098609764096 async_checkpointer.py:280] [process=0][thread=save_finalize] Waiting for background save thread=async_save. I0418 19:59:05.063878 134226523375424 standard_logger.py:34] {'step': 19, 'event_type': 'save', 'directory': 'gs://lance-maxtext/linen_ckpt_xpk_main_20260418_180002/linen_xpk_main_20260418_180002_16_zarr3_false/checkpoints', 'reached_preemption': False, 'preemption_received_at': None, 'synchronous': False, 'wait_for_prev_start_time': 1776542314.8008523, 'wait_for_prev_duration_secs': 29.10089421272278, 'time_between_consecutive_saves_sec': 22.734842777252197, 'checkpointer_blocking_start_time': 1776542343.903355, 'checkpointer_blocking_duration_secs': 1.158611536026001, 'get_old_steps_start_time': 1776542345.0619817, 'get_old_steps_duration_secs': 2.765655517578125e-05, 'checkpoint_manager_blocking_start_time': 1776542314.7990124, 'checkpoint_manager_blocking_duration_secs': 30.2648344039917} I0418 19:59:05.063986 134226523375424 checkpointing.py:408] Started an asynchronous checkpoint save for step 19 I0418 19:59:05.064027 134226523375424 checkpoint_manager.py:2020] [process=0][thread=MainThread][step=19][wait_until_finished] Waiting for Save Finalize thread (save_finalize) to complete. I0418 19:59:05.265986 134098592978688 atomicity.py:140] Creating tmp directory gs://lance-maxtext/linen_ckpt_xpk_main_20260418_180002/linen_xpk_main_20260418_180002_16_zarr3_false/checkpoints/19 I0418 19:59:07.426480 134102850189056 atomicity.py:140] Creating tmp directory gs://lance-maxtext/linen_ckpt_xpk_main_20260418_180002/linen_xpk_main_20260418_180002_16_zarr3_false/checkpoints/19/items I0418 19:59:10.665931 134103936509696 array_metadata_store.py:203] [process=0][thread=array_type_handler] Wrote 39 array_metadata.ArrayMetadata to gs://lance-maxtext/linen_ckpt_xpk_main_20260418_180002/linen_xpk_main_20260418_180002_16_zarr3_false/checkpoints/19/items/array_metadatas/process_0 I0418 19:59:50.521384 134102367860480 base_pytree_checkpoint_handler.py:1282] [process=0][thread=write_metadata_after_commits] Commit + Array metadata written. Time taken: 41.999682s (commit=40.405861s, array_metadata_write=1.593821s) I0418 19:59:50.522416 134100203591424 base_pytree_checkpoint_handler.py:130] [process=0] /jax/orbax/write/gbytes_per_sec: 34.722 MiB/s (total gbytes: 1.5 GiB) (time elapsed: 45.493045806884766 s) (per-host) I0418 19:59:50.522498 134100203591424 async_checkpointer.py:90] [process=0][thread=async_save] 3 Handler Commit operations completed. Time taken: 45.460559s. I0418 19:59:52.790468 134100203591424 array_metadata_store.py:411] [process=0][thread=async_save] Validated ArrayMetadata from all 8 hosts. Time taken: 0.000225s. I0418 19:59:53.955210 134100203591424 base_pytree_checkpoint_handler.py:1406] [process=0][thread=async_save] Pytree save finalize (merge_ocdbt + ArrayMetadata validation) completed. Time taken: 2.151744s. use_zarr3=False, enable_post_merge_validation=True, directory=gs://lance-maxtext/linen_ckpt_xpk_main_20260418_180002/linen_xpk_main_20260418_180002_16_zarr3_false/checkpoints/19/items I0418 19:59:53.956887 134100203591424 atomicity.py:666] Finalizing gs://lance-maxtext/linen_ckpt_xpk_main_20260418_180002/linen_xpk_main_20260418_180002_16_zarr3_false/checkpoints/19/items I0418 19:59:54.473652 134100203591424 atomicity.py:666] Finalizing gs://lance-maxtext/linen_ckpt_xpk_main_20260418_180002/linen_xpk_main_20260418_180002_16_zarr3_false/checkpoints/19 I0418 19:59:55.877516 134100203591424 atomicity.py:847] [process=0][thread=async_save] Finished saving checkpoint (finalized tmp dir) to `gs://lance-maxtext/linen_ckpt_xpk_main_20260418_180002/linen_xpk_main_20260418_180002_16_zarr3_false/checkpoints/19`. I0418 19:59:55.878356 134100203591424 event_tracking.py:138] [process=0] [async] Finished save (blocking + background) in 51.97 seconds @ gs://lance-maxtext/linen_ckpt_xpk_main_20260418_180002/linen_xpk_main_20260418_180002_16_zarr3_false/checkpoints/19 I0418 19:59:55.879791 134100203591424 async_checkpointer.py:160] [process=0][thread=async_save] Background save thread done. Time taken: 50.817832s. I0418 19:59:55.880036 134098609764096 async_checkpointer.py:288] [process=0][thread=save_finalize] Done with waiting for background save thread=async_save. I0418 19:59:55.880116 134098609764096 async_checkpointer.py:298] [process=0][thread=save_finalize] No errors found in background save thread=async_save. I0418 19:59:55.880174 134098609764096 checkpoint_manager.py:2137] [process=0][thread=save_finalize][step=19] CheckpointManager Save Finalize is syncing with other hosts... I0418 19:59:55.882199 134098609764096 checkpoint_manager.py:2146] [process=0][thread=save_finalize][step=19] CheckpointManager Save Finalize is done on all hosts. I0418 19:59:55.882384 134226523375424 checkpoint_manager.py:2032] [process=0][thread=MainThread][step=19][wait_until_finished] Done waiting for Save Finalize thread (save_finalize) running at step=19. I0418 19:59:55.882535 134226523375424 checkpoint_manager.py:2009] [process=0][thread=MainThread][wait_until_finished] No Save Finalize thread to wait for. Returning. I0418 19:59:55.883387 134226523375424 metric_logger.py:196] completed step: 19, seconds: 0.148, TFLOP/s/device: 91.999, Tokens/s/device: 13867.071, total_weights: 65536, loss: 5.391, lm_loss: 5.391, perplexity: 219.349 Per train step: Total TFLOPs: 13.59 split as 93.93% learnable weight flops and 6.07% attention flops XPK End: Sat Apr 18 20:00:09 UTC 2026 EXIT_CODE=0
XPK Start: Sun Apr 19 01:20:06 UTC 2026 PyTorch was not found. Models won't be available and only tokenizers, configuration and file/data utilities can be used. 2026-04-19 01:20:30.542340: 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) I0419 01:20:30.720520 133141823870784 max_utils.py:273] Attempting to initialize the jax distributed system... I0419 01:20:39.761923 133141823870784 distributed.py:149] Starting JAX distributed service on [::]:8482 I0419 01:20:39.764328 133141823870784 distributed.py:172] Connecting to JAX distributed service on mt-16-zarr3-false-rs-pw3bw-slice-job-0-0.mt-16-zarr3-false-rs-pw3bw:8482 I0419 01:20:41.272413 133141823870784 max_utils.py:284] Jax distributed system initialized! I0419 01:20:47.156626 133141823870784 max_utils.py:800] System Information: Jax Version: 0.9.2 I0419 01:20:47.156728 133141823870784 max_utils.py:801] System Information: Jaxlib Version: 0.9.2 I0419 01:20:47.156776 133141823870784 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 I0419 01:20:47.156829 133141823870784 train_utils.py:378] WARNING: Sequence packing is essentially ignored for synthetic data. Please use a real dataset to use sequence packing. I0419 01:20:47.844090 133141823870784 maxtext_utils.py:1718] Num_devices: 32, shape (1, 1, 1, 32, 1, 1, 1, 1, 1, 1, 1, 1, 1) I0419 01:20:47.844672 133141823870784 maxtext_utils.py:1718] Num_devices: 32, shape (1, 1, 1, 32, 1, 1, 1, 1, 1, 1, 1, 1, 1) I0419 01:20:47.844875 133141823870784 checkpointing.py:688] Setting up checkpoint logger... I0419 01:20:47.844930 133141823870784 checkpointing.py:234] Creating checkpoint manager with ocdbt=True and zarr3=False I0419 01:20:47.844973 133141823870784 pytree_checkpoint_handler.py:592] save_device_host_concurrent_bytes=None I0419 01:20:47.845306 133141823870784 base_pytree_checkpoint_handler.py:441] Created BasePyTreeCheckpointHandler: use_ocdbt=True, use_zarr3=False, pytree_metadata_options=PyTreeMetadataOptions(support_rich_types=False), array_metadata_store=<orbax.checkpoint._src.metadata.array_metadata_store.Store object at 0x7916c7750260>, enable_pinned_host_transfer=False, save_concurrent_bytes: 96000000000 (89.4 GiB), restore_concurrent_bytes: 96000000000 (89.4 GiB) I0419 01:20:49.966414 133141823870784 checkpointing.py:266] Enabling policy for fixed interval checkpointing. I0419 01:20:49.966645 133141823870784 checkpoint_manager.py:708] [process=0][thread=MainThread] CheckpointManager init: checkpointers=None, item_names=('items',), item_handlers={'items': <orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeCheckpointHandler object at 0x7903206ce180>}, handler_registry=None I0419 01:20:49.966899 133141823870784 composite_checkpoint_handler.py:237] Deferred registration for item: "items". Adding handler `<orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeCheckpointHandler object at 0x7903206ce180>` 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`. I0419 01:20:49.966950 133141823870784 composite_checkpoint_handler.py:237] Deferred registration for item: "metrics". Adding handler `<orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonCheckpointHandler object at 0x7903206d1be0>` 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`. I0419 01:20:49.966986 133141823870784 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 0x7903206ce180>, ('items', <class 'orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeRestoreArgs'>): <orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeCheckpointHandler object at 0x7903206ce180>, ('metrics', <class 'orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonSaveArgs'>): <orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonCheckpointHandler object at 0x7903206d1be0>, ('metrics', <class 'orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonRestoreArgs'>): <orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonCheckpointHandler object at 0x7903206d1be0>}). I0419 01:20:49.967301 133141823870784 abstract_checkpointer.py:35] orbax-checkpoint version: 0.11.34 I0419 01:20:49.967372 133141823870784 async_checkpointer.py:192] [process=0][thread=MainThread] Using barrier_sync_fn: <function get_barrier_sync_fn.<locals>._fn at 0x79026028f740> timeout: 1200 secs and primary_host=0 for async checkpoint writes I0419 01:20:53.078234 133023827736320 checkpoint.py:228] Read Metadata={'item_handlers': {'items': 'orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeCheckpointHandler'}, 'metrics': {}, 'performance_metrics': {}, 'init_timestamp_nsecs': 1776561427326095839, 'commit_timestamp_nsecs': 1776561476751034969, 'custom_metadata': {}} from gs://lance-maxtext/linen_ckpt_xpk_feat_nnx_post_train_fixes_20260418_235007/linen_xpk_feat_nnx_post_train_fixes_20260418_235007_16_zarr3_false/checkpoints/5/_CHECKPOINT_METADATA I0419 01:20:53.327504 133023827736320 checkpoint_manager.py:1871] Missing metrics for step 5 I0419 01:20:53.767640 133141823870784 checkpoint_manager.py:1812] Found 3 checkpoint steps in gs://lance-maxtext/linen_ckpt_xpk_feat_nnx_post_train_fixes_20260418_235007/linen_xpk_feat_nnx_post_train_fixes_20260418_235007_16_zarr3_false/checkpoints I0419 01:20:55.482841 133141823870784 checkpoint_manager.py:929] [process=0][thread=MainThread] CheckpointManager created, primary_host=0, CheckpointManagerOptions=CheckpointManagerOptions(save_interval_steps=1, max_to_keep=None, keep_time_interval=None, keep_period=None, should_keep_fn=None, best_fn=None, best_mode='max', keep_checkpoints_without_metrics=True, step_prefix=None, step_format_fixed_length=None, step_name_format=None, create=True, cleanup_tmp_directories=False, save_on_steps=frozenset(), single_host_load_and_broadcast=False, todelete_subdir=None, todelete_full_path=None, enable_background_delete=False, read_only=False, enable_async_checkpointing=True, async_options=None, multiprocessing_options=MultiprocessingOptions(primary_host=0, active_processes=None, barrier_sync_key_prefix=None), should_save_fn=None, file_options=FileOptions(path_permission_mode=None), save_root_metadata=True, temporary_path_class=None, save_decision_policy=FixedIntervalPolicy(interval=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_feat_nnx_post_train_fixes_20260418_235007/linen_xpk_feat_nnx_post_train_fixes_20260418_235007_16_zarr3_false/checkpoints: <orbax.checkpoint.checkpoint_manager.CheckpointManager object at 0x7903206cf0e0> I0419 01:20:55.483013 133141823870784 checkpointing.py:302] Checkpoint manager created! I0419 01:20:56.396967 133141823870784 nnx_wrappers.py:437] Unknown Logical: bfloat16[32,2048,2048]...................................... ('activation_batch', 'activation_norm_length', 'activation_embed'). I0419 01:20:56.397072 133141823870784 nnx_wrappers.py:437] Unknown Physical: bfloat16[32,2048,2048]...................................... ('fsdp', None, None). I0419 01:20:56.779620 133141823870784 attentions.py:1088] attentions/inputs_q Logical: bfloat16[32,2048,2048]...................................... ('activation_batch', 'activation_attn_length', 'activation_attn_embed'). I0419 01:20:56.779709 133141823870784 attentions.py:1088] attentions/inputs_q Physical: bfloat16[32,2048,2048]...................................... ('fsdp', None, None). I0419 01:20:56.795898 133141823870784 attentions.py:1089] attentions/inputs_kv Logical: bfloat16[32,2048,2048]...................................... ('activation_batch', 'activation_attn_length', 'activation_attn_embed'). I0419 01:20:56.795957 133141823870784 attentions.py:1089] attentions/inputs_kv Physical: bfloat16[32,2048,2048]...................................... ('fsdp', None, None). I0419 01:20:56.819380 133141823870784 attentions.py:1154] attentions/query Logical: bfloat16[32,2048,16,128].................................... ('activation_kv_batch', 'activation_attn_length', 'activation_kv_heads', 'activation_kv_head_dim'). I0419 01:20:56.819447 133141823870784 attentions.py:1154] attentions/query Physical: bfloat16[32,2048,16,128].................................... ('fsdp', None, None, None). I0419 01:20:56.835681 133141823870784 attentions.py:1155] attentions/key Logical: bfloat16[32,2048,16,128].................................... ('activation_kv_batch', 'activation_attn_length', 'activation_kv_heads', 'activation_kv_head_dim'). I0419 01:20:56.835741 133141823870784 attentions.py:1155] attentions/key Physical: bfloat16[32,2048,16,128].................................... ('fsdp', None, None, None). I0419 01:20:56.851845 133141823870784 attentions.py:1156] attentions/value Logical: bfloat16[32,2048,16,128].................................... ('activation_kv_batch', 'activation_attn_length', 'activation_kv_heads', 'activation_kv_head_dim'). I0419 01:20:56.851904 133141823870784 attentions.py:1156] attentions/value Physical: bfloat16[32,2048,16,128].................................... ('fsdp', None, None, None). I0419 01:20:56.876178 133141823870784 attentions.py:1197] attentions/out Logical: bfloat16[32,2048,16,128].................................... ('activation_batch', 'activation_attn_length', 'activation_heads', 'activation_kv'). I0419 01:20:56.876245 133141823870784 attentions.py:1197] attentions/out Physical: bfloat16[32,2048,16,128].................................... ('fsdp', None, None, None). I0419 01:20:56.897192 133141823870784 linears.py:525] linears/x Logical: bfloat16[32,2048,7168]...................................... ('activation_batch', 'activation_length', 'activation_mlp'). I0419 01:20:56.897257 133141823870784 linears.py:525] linears/x Physical: bfloat16[32,2048,7168]...................................... ('fsdp', None, None). I0419 01:20:57.109116 133141823870784 checkpointing.py:578] checkpoint manager exists so trying to load this run's existing checkpoint I0419 01:20:57.109220 133141823870784 checkpointing.py:582] restoring from this run's directory step 9 I0419 01:20:57.356637 133141823870784 event_tracking.py:70] [process=0] [sync] Started load checkpoint @ gs://lance-maxtext/linen_ckpt_xpk_feat_nnx_post_train_fixes_20260418_235007/linen_xpk_feat_nnx_post_train_fixes_20260418_235007_16_zarr3_false/checkpoints/9. I0419 01:20:57.603670 133141823870784 checkpointer.py:307] Restoring checkpoint from gs://lance-maxtext/linen_ckpt_xpk_feat_nnx_post_train_fixes_20260418_235007/linen_xpk_feat_nnx_post_train_fixes_20260418_235007_16_zarr3_false/checkpoints/9. I0419 01:20:57.604118 133141823870784 event_tracking.py:125] [process=0] [sync] Finished blocking load in 0.25 seconds. Continuing load @ gs://lance-maxtext/linen_ckpt_xpk_feat_nnx_post_train_fixes_20260418_235007/linen_xpk_feat_nnx_post_train_fixes_20260418_235007_16_zarr3_false/checkpoints/9. I0419 01:20:59.045442 1417 google_auth_provider.cc:181] Running on GCE, using service account 562977990677-compute@developer.gserviceaccount.com I0419 01:21:05.657505 133141823870784 jax_array_handlers.py:843] [process=0] /jax/orbax/read/worker/io/requested throughput: 238.787 MiB/s (total gbytes: 1.5 GiB) (time elapsed: 6.617924213409424 s) (per-host) I0419 01:21:05.662395 133141823870784 base_pytree_checkpoint_handler.py:130] [process=0] /jax/checkpoint/read/gbytes_per_sec: 224.091 MiB/s (total gbytes: 1.5 GiB) (time elapsed: 7.0622968673706055 s) (per-host) I0419 01:21:10.107423 133141823870784 event_tracking.py:138] [process=0] [sync] Finished load in 12.75 seconds @ gs://lance-maxtext/linen_ckpt_xpk_feat_nnx_post_train_fixes_20260418_235007/linen_xpk_feat_nnx_post_train_fixes_20260418_235007_16_zarr3_false/checkpoints/9 I0419 01:21:10.107590 133141823870784 standard_logger.py:34] {'step': 9, 'event_type': 'restore', 'directory': 'gs://lance-maxtext/linen_ckpt_xpk_feat_nnx_post_train_fixes_20260418_235007/linen_xpk_feat_nnx_post_train_fixes_20260418_235007_16_zarr3_false/checkpoints', 'checkpointer_start_time': 1776561657.3565428, 'checkpointer_duration_secs': 12.751016616821289, 'checkpoint_manager_start_time': 1776561657.1094666, 'checkpoint_manager_duration_secs': 12.998094081878662} fsdp: 32 I0419 01:21:10.518571 133141823870784 maxtext_utils.py:1827] params/params/decoder/decoder_norm/scale Shape: float32[2048] Logical: P('norm',) Physical: (None,) I0419 01:21:10.518697 133141823870784 maxtext_utils.py:1827] params/params/decoder/layers/mlp/wi_0/kernel Shape: float32[2048,16,7168] Logical: P('embed', 'layers', 'mlp') Physical: ('fsdp', None, None) I0419 01:21:10.518750 133141823870784 maxtext_utils.py:1827] params/params/decoder/layers/mlp/wi_1/kernel Shape: float32[2048,16,7168] Logical: P('embed', 'layers', 'mlp') Physical: ('fsdp', None, None) I0419 01:21:10.518856 133141823870784 maxtext_utils.py:1827] params/params/decoder/layers/mlp/wo/kernel Shape: float32[7168,16,2048] Logical: P('mlp', 'layers', 'embed') Physical: (None, None, 'fsdp') I0419 01:21:10.518918 133141823870784 maxtext_utils.py:1827] params/params/decoder/layers/post_self_attention_layer_norm/scale Shape: float32[2048,16] Logical: P('norm', 'layers') Physical: (None, None) I0419 01:21:10.518960 133141823870784 maxtext_utils.py:1827] params/params/decoder/layers/pre_self_attention_layer_norm/scale Shape: float32[2048,16] Logical: P('norm', 'layers') Physical: (None, None) I0419 01:21:10.519015 133141823870784 maxtext_utils.py:1827] 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) I0419 01:21:10.519068 133141823870784 maxtext_utils.py:1827] 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') I0419 01:21:10.519108 133141823870784 maxtext_utils.py:1827] 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) I0419 01:21:10.519145 133141823870784 maxtext_utils.py:1827] 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) I0419 01:21:10.519192 133141823870784 maxtext_utils.py:1827] params/params/decoder/logits_dense/kernel Shape: float32[2048,32000] Logical: P('embed_vocab', 'vocab') Physical: ('fsdp', None) I0419 01:21:10.519237 133141823870784 maxtext_utils.py:1827] params/params/token_embedder/embedding Shape: float32[32000,2048] Logical: P('vocab', 'embed_vocab') Physical: (None, 'fsdp') I0419 01:21:11.004665 133141823870784 train.py:157] train/xent Logical: float32[32,2048]............................................ ('activation_embed_and_logits_batch', 'activation_length'). I0419 01:21:11.004761 133141823870784 train.py:157] train/xent Physical: float32[32,2048]............................................ ('fsdp', None). I0419 01:21:11.020124 133141823870784 train.py:164] train/z_loss Logical: float32[32,2048]............................................ ('activation_embed_and_logits_batch', 'activation_length'). I0419 01:21:11.020182 133141823870784 train.py:164] train/z_loss Physical: float32[32,2048]............................................ ('fsdp', None). I0419 01:21:21.981070 133141823870784 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. I0419 01:21:21.989516 133141823870784 metric_logger.py:301] number parameters: 1.104 billion I0419 01:21:33.609582 133141823870784 checkpointing.py:794] Waiting for step 10 to finish before checkpoint... I0419 01:21:33.764917 133141823870784 checkpointing.py:798] Waited 0.15531301498413086 seconds for step 10 to finish before starting checkpointing. I0419 01:21:33.767208 133141823870784 checkpoint_manager.py:2009] [process=0][thread=MainThread][wait_until_finished] No Save Finalize thread to wait for. Returning. I0419 01:21:33.768922 133141823870784 checkpoint_manager.py:1512] [process=0] Saving checkpoint at step 10 I0419 01:21:33.772098 133141823870784 event_tracking.py:70] [process=0] [async] Started save checkpoint @ gs://lance-maxtext/linen_ckpt_xpk_feat_nnx_post_train_fixes_20260418_235007/linen_xpk_feat_nnx_post_train_fixes_20260418_235007_16_zarr3_false/checkpoints/10. I0419 01:21:34.913678 133141823870784 signaling_client.py:364] Using JaxDistributedSignalingClient I0419 01:21:34.926431 133141823870784 jax_array_handlers.py:360] Scheduling D2H of 39 prioritized jax.Array. I0419 01:21:34.926500 133141823870784 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 I0419 01:21:35.169896 133022217123584 atomicity.py:140] Creating tmp directory gs://lance-maxtext/linen_ckpt_xpk_feat_nnx_post_train_fixes_20260418_235007/linen_xpk_feat_nnx_post_train_fixes_20260418_235007_16_zarr3_false/checkpoints/10 I0419 01:21:35.199188 133141823870784 base_pytree_checkpoint_handler.py:154] [process=0][thread=MainThread] Initiated "orbax.checkpoint._src.serialization.jax_array_handlers.ArrayHandler".serialize. Time taken: 0.273661s I0419 01:21:35.199938 133141823870784 base_pytree_checkpoint_handler.py:130] [process=0] /jax/orbax/write/blocking_gbytes_per_sec: 5.547 GiB/s (total gbytes: 1.5 GiB) (time elapsed: 0.2781100273132324 s) (per-host) I0419 01:21:35.200010 133141823870784 base_pytree_checkpoint_handler.py:768] [process=0][thread=MainThread] Initiated Pytree async_save. Time taken: 0.278201s (batch_requests_ready=0.001974s, total_serialization_initiated=0.275554s, others=0.000673s) I0419 01:21:35.200088 133141823870784 composite_checkpoint_handler.py:715] [process=0][thread=MainThread] Initiated CompositeCheckpointHandler.async_save. Time taken: 0.282385s (all_items=0.000016s, per_item={'items': '0.00001645'}, temp_paths=0.282369) I0419 01:21:35.200795 133141823870784 event_tracking.py:125] [process=0] [async] Finished blocking save in 1.43 seconds. Continuing save @ gs://lance-maxtext/linen_ckpt_xpk_feat_nnx_post_train_fixes_20260418_235007/linen_xpk_feat_nnx_post_train_fixes_20260418_235007_16_zarr3_false/checkpoints/10. I0419 01:21:35.201141 133023282472704 async_checkpointer.py:76] [process=0][thread=async_save] Background save thread started. Deadline for this save operation is 2026-04-19 01:41:35.201104 I0419 01:21:35.216428 133141823870784 checkpoint_manager.py:1560] [process=0][thread=MainThread][step=10] Starting CheckpointManager Save Finalize thread=save_finalize I0419 01:21:35.216699 133018451330816 async_checkpointer.py:280] [process=0][thread=save_finalize] Waiting for background save thread=async_save. I0419 01:21:35.216863 133141823870784 standard_logger.py:34] {'step': 10, 'event_type': 'save', 'directory': 'gs://lance-maxtext/linen_ckpt_xpk_feat_nnx_post_train_fixes_20260418_235007/linen_xpk_feat_nnx_post_train_fixes_20260418_235007_16_zarr3_false/checkpoints', 'reached_preemption': False, 'preemption_received_at': None, 'synchronous': False, 'wait_for_prev_start_time': 1776561693.767191, 'wait_for_prev_duration_secs': 5.6743621826171875e-05, 'time_between_consecutive_saves_sec': None, 'checkpointer_blocking_start_time': 1776561693.7689621, 'checkpointer_blocking_duration_secs': 1.4323296546936035, 'get_old_steps_start_time': 1776561695.2013154, 'get_old_steps_duration_secs': 2.8133392333984375e-05, 'checkpoint_manager_blocking_start_time': 1776561693.7654397, 'checkpoint_manager_blocking_duration_secs': 1.4513888359069824} I0419 01:21:35.216979 133141823870784 checkpointing.py:409] Started an asynchronous checkpoint save for step 10 I0419 01:21:35.217032 133141823870784 max_utils.py:750] Memstats: After params initialized: I0419 01:21:35.217081 133141823870784 max_utils.py:756] Using (GB) 0.43 / 31.25 (1.376000%) on TPU_0(process=0,(0,0,0,0)) I0419 01:21:35.217115 133141823870784 max_utils.py:756] Using (GB) 0.43 / 31.25 (1.376000%) on TPU_1(process=0,(1,0,0,0)) I0419 01:21:35.217141 133141823870784 max_utils.py:756] Using (GB) 0.43 / 31.25 (1.376000%) on TPU_4(process=0,(0,1,0,0)) I0419 01:21:35.217165 133141823870784 max_utils.py:756] Using (GB) 0.43 / 31.25 (1.376000%) on TPU_5(process=0,(1,1,0,0)) I0419 01:21:35.553492 133141823870784 metric_logger.py:196] completed step: 10, seconds: 11.308, TFLOP/s/device: 1.202, Tokens/s/device: 181.110, total_weights: 65536, loss: 6.833, lm_loss: 6.833, perplexity: 927.869 I0419 01:21:35.698784 133141823870784 metric_logger.py:196] completed step: 11, seconds: 1.942, TFLOP/s/device: 6.997, Tokens/s/device: 1054.595, total_weights: 65536, loss: 6.527, lm_loss: 6.527, perplexity: 683.659 I0419 01:21:35.846518 133141823870784 metric_logger.py:196] completed step: 12, seconds: 0.011, TFLOP/s/device: 1260.752, Tokens/s/device: 190034.332, total_weights: 65536, loss: 6.265, lm_loss: 6.265, perplexity: 526.068 I0419 01:21:35.993941 133141823870784 metric_logger.py:196] completed step: 13, seconds: 0.146, TFLOP/s/device: 93.302, Tokens/s/device: 14063.423, total_weights: 65536, loss: 6.045, lm_loss: 6.045, perplexity: 422.001 I0419 01:21:36.003099 133141823870784 checkpointing.py:794] Waiting for step 15 to finish before checkpoint... I0419 01:21:36.288149 133141823870784 checkpointing.py:798] Waited 0.2848825454711914 seconds for step 15 to finish before starting checkpointing. I0419 01:21:36.291222 133141823870784 checkpoint_manager.py:2020] [process=0][thread=MainThread][step=10][wait_until_finished] Waiting for Save Finalize thread (save_finalize) to complete. I0419 01:21:37.384752 133019029468928 atomicity.py:140] Creating tmp directory gs://lance-maxtext/linen_ckpt_xpk_feat_nnx_post_train_fixes_20260418_235007/linen_xpk_feat_nnx_post_train_fixes_20260418_235007_16_zarr3_false/checkpoints/10/items I0419 01:21:37.543571 133018442938112 checkpoint.py:188] Wrote Metadata={'item_handlers': None, 'metrics': {}, 'performance_metrics': {}, 'init_timestamp_nsecs': 1776561697137920817, 'commit_timestamp_nsecs': None, 'custom_metadata': {}}, json={"item_handlers": null, "metrics": {}, "performance_metrics": {}, "init_timestamp_nsecs": 1776561697137920817, "commit_timestamp_nsecs": null, "custom_metadata": {}} to gs://lance-maxtext/linen_ckpt_xpk_feat_nnx_post_train_fixes_20260418_235007/linen_xpk_feat_nnx_post_train_fixes_20260418_235007_16_zarr3_false/checkpoints/10/_CHECKPOINT_METADATA I0419 01:21:40.713163 133019012683520 array_metadata_store.py:203] [process=0][thread=array_type_handler] Wrote 39 array_metadata.ArrayMetadata to gs://lance-maxtext/linen_ckpt_xpk_feat_nnx_post_train_fixes_20260418_235007/linen_xpk_feat_nnx_post_train_fixes_20260418_235007_16_zarr3_false/checkpoints/10/items/array_metadatas/process_0 I0419 01:22:21.160737 133018974508800 base_pytree_checkpoint_handler.py:1282] [process=0][thread=write_metadata_after_commits] Commit + Array metadata written. Time taken: 42.224693s (commit=40.639484s, array_metadata_write=1.585209s) I0419 01:22:21.162199 133023282472704 base_pytree_checkpoint_handler.py:130] [process=0] /jax/orbax/write/gbytes_per_sec: 34.161 MiB/s (total gbytes: 1.5 GiB) (time elapsed: 46.24032950401306 s) (per-host) I0419 01:22:21.162369 133023282472704 async_checkpointer.py:90] [process=0][thread=async_save] 3 Handler Commit operations completed. Time taken: 45.961110s. I0419 01:22:23.061681 133018442938112 checkpoint.py:247] Updated Metadata={'item_handlers': {'items': 'orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeCheckpointHandler'}, 'metrics': {}, 'performance_metrics': {}, 'init_timestamp_nsecs': 1776561697137920817, 'commit_timestamp_nsecs': None, 'custom_metadata': {}} to gs://lance-maxtext/linen_ckpt_xpk_feat_nnx_post_train_fixes_20260418_235007/linen_xpk_feat_nnx_post_train_fixes_20260418_235007_16_zarr3_false/checkpoints/10/_CHECKPOINT_METADATA I0419 01:22:23.517437 133023282472704 array_metadata_store.py:411] [process=0][thread=async_save] Validated ArrayMetadata from all 8 hosts. Time taken: 0.000408s. I0419 01:22:24.592230 133023282472704 base_pytree_checkpoint_handler.py:1406] [process=0][thread=async_save] Pytree save finalize (merge_ocdbt + ArrayMetadata validation) completed. Time taken: 2.167064s. use_zarr3=False, enable_post_merge_validation=True, directory=gs://lance-maxtext/linen_ckpt_xpk_feat_nnx_post_train_fixes_20260418_235007/linen_xpk_feat_nnx_post_train_fixes_20260418_235007_16_zarr3_false/checkpoints/10/items I0419 01:22:24.594506 133023282472704 atomicity.py:666] Finalizing gs://lance-maxtext/linen_ckpt_xpk_feat_nnx_post_train_fixes_20260418_235007/linen_xpk_feat_nnx_post_train_fixes_20260418_235007_16_zarr3_false/checkpoints/10/items I0419 01:22:25.154975 133023282472704 atomicity.py:666] Finalizing gs://lance-maxtext/linen_ckpt_xpk_feat_nnx_post_train_fixes_20260418_235007/linen_xpk_feat_nnx_post_train_fixes_20260418_235007_16_zarr3_false/checkpoints/10 I0419 01:22:26.500334 133023282472704 atomicity.py:847] [process=0][thread=async_save] Finished saving checkpoint (finalized tmp dir) to `gs://lance-maxtext/linen_ckpt_xpk_feat_nnx_post_train_fixes_20260418_235007/linen_xpk_feat_nnx_post_train_fixes_20260418_235007_16_zarr3_false/checkpoints/10`. I0419 01:22:26.501497 133023282472704 event_tracking.py:138] [process=0] [async] Finished save (blocking + background) in 52.73 seconds @ gs://lance-maxtext/linen_ckpt_xpk_feat_nnx_post_train_fixes_20260418_235007/linen_xpk_feat_nnx_post_train_fixes_20260418_235007_16_zarr3_false/checkpoints/10 I0419 01:22:26.502924 133023282472704 async_checkpointer.py:160] [process=0][thread=async_save] Background save thread done. Time taken: 51.301661s. I0419 01:22:26.503133 133018451330816 async_checkpointer.py:288] [process=0][thread=save_finalize] Done with waiting for background save thread=async_save. I0419 01:22:26.503229 133018451330816 async_checkpointer.py:298] [process=0][thread=save_finalize] No errors found in background save thread=async_save. I0419 01:22:26.503331 133018451330816 checkpoint_manager.py:2137] [process=0][thread=save_finalize][step=10] CheckpointManager Save Finalize is syncing with other hosts... I0419 01:22:26.505014 133018451330816 checkpoint_manager.py:2146] [process=0][thread=save_finalize][step=10] CheckpointManager Save Finalize is done on all hosts. I0419 01:22:26.505280 133141823870784 checkpoint_manager.py:2032] [process=0][thread=MainThread][step=10][wait_until_finished] Done waiting for Save Finalize thread (save_finalize) running at step=10. W0419 01:22:26.505464 133141823870784 checkpoint_manager.py:1452] Waiting for previous save to complete took 50.214244 seconds. If this number is high, consider checkpointing less frequently. I0419 01:22:26.507289 133141823870784 checkpoint_manager.py:1512] [process=0] Saving checkpoint at step 15 I0419 01:22:26.509739 133141823870784 event_tracking.py:70] [process=0] [async] Started save checkpoint @ gs://lance-maxtext/linen_ckpt_xpk_feat_nnx_post_train_fixes_20260418_235007/linen_xpk_feat_nnx_post_train_fixes_20260418_235007_16_zarr3_false/checkpoints/15. I0419 01:22:27.224370 133141823870784 jax_array_handlers.py:360] Scheduling D2H of 39 prioritized jax.Array. I0419 01:22:27.224513 133141823870784 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 I0419 01:22:27.253375 133141823870784 base_pytree_checkpoint_handler.py:154] [process=0][thread=MainThread] Initiated "orbax.checkpoint._src.serialization.jax_array_handlers.ArrayHandler".serialize. Time taken: 0.030029s I0419 01:22:27.254239 133141823870784 base_pytree_checkpoint_handler.py:130] [process=0] /jax/orbax/write/blocking_gbytes_per_sec: 39.684 GiB/s (total gbytes: 1.5 GiB) (time elapsed: 0.03887128829956055 s) (per-host) I0419 01:22:27.254350 133141823870784 base_pytree_checkpoint_handler.py:768] [process=0][thread=MainThread] Initiated Pytree async_save. Time taken: 0.039005s (batch_requests_ready=0.004029s, total_serialization_initiated=0.034174s, others=0.000801s) I0419 01:22:27.254475 133141823870784 composite_checkpoint_handler.py:715] [process=0][thread=MainThread] Initiated CompositeCheckpointHandler.async_save. Time taken: 0.043950s (all_items=0.000024s, per_item={'items': '0.00002360'}, temp_paths=0.043926) I0419 01:22:27.255409 133141823870784 event_tracking.py:125] [process=0] [async] Finished blocking save in 0.75 seconds. Continuing save @ gs://lance-maxtext/linen_ckpt_xpk_feat_nnx_post_train_fixes_20260418_235007/linen_xpk_feat_nnx_post_train_fixes_20260418_235007_16_zarr3_false/checkpoints/15. I0419 01:22:27.255786 133018451330816 async_checkpointer.py:76] [process=0][thread=async_save] Background save thread started. Deadline for this save operation is 2026-04-19 01:42:27.255748 I0419 01:22:27.258215 133141823870784 checkpoint_manager.py:1560] [process=0][thread=MainThread][step=15] Starting CheckpointManager Save Finalize thread=save_finalize I0419 01:22:27.258533 133019021076224 async_checkpointer.py:280] [process=0][thread=save_finalize] Waiting for background save thread=async_save. I0419 01:22:27.258685 133141823870784 standard_logger.py:34] {'step': 15, 'event_type': 'save', 'directory': 'gs://lance-maxtext/linen_ckpt_xpk_feat_nnx_post_train_fixes_20260418_235007/linen_xpk_feat_nnx_post_train_fixes_20260418_235007_16_zarr3_false/checkpoints', 'reached_preemption': False, 'preemption_received_at': None, 'synchronous': False, 'wait_for_prev_start_time': 1776561696.2911768, 'wait_for_prev_duration_secs': 50.21424412727356, 'time_between_consecutive_saves_sec': None, 'checkpointer_blocking_start_time': 1776561746.5073562, 'checkpointer_blocking_duration_secs': 0.7486221790313721, 'get_old_steps_start_time': 1776561747.2560089, 'get_old_steps_duration_secs': 4.410743713378906e-05, 'checkpoint_manager_blocking_start_time': 1776561696.2888322, 'checkpoint_manager_blocking_duration_secs': 50.96980047225952} I0419 01:22:27.258905 133141823870784 checkpointing.py:409] Started an asynchronous checkpoint save for step 15 I0419 01:22:27.457969 133023282472704 atomicity.py:140] Creating tmp directory gs://lance-maxtext/linen_ckpt_xpk_feat_nnx_post_train_fixes_20260418_235007/linen_xpk_feat_nnx_post_train_fixes_20260418_235007_16_zarr3_false/checkpoints/15 I0419 01:22:30.070996 133017947334400 atomicity.py:140] Creating tmp directory gs://lance-maxtext/linen_ckpt_xpk_feat_nnx_post_train_fixes_20260418_235007/linen_xpk_feat_nnx_post_train_fixes_20260418_235007_16_zarr3_false/checkpoints/15/items I0419 01:22:32.917827 133019524376320 array_metadata_store.py:203] [process=0][thread=array_type_handler] Wrote 39 array_metadata.ArrayMetadata to gs://lance-maxtext/linen_ckpt_xpk_feat_nnx_post_train_fixes_20260418_235007/linen_xpk_feat_nnx_post_train_fixes_20260418_235007_16_zarr3_false/checkpoints/15/items/array_metadatas/process_0 I0419 01:22:48.043692 133141823870784 metric_logger.py:196] completed step: 14, seconds: 0.148, TFLOP/s/device: 91.985, Tokens/s/device: 13865.006, total_weights: 65536, loss: 5.864, lm_loss: 5.864, perplexity: 352.193 I0419 01:22:48.052929 133141823870784 metric_logger.py:196] completed step: 15, seconds: 0.147, TFLOP/s/device: 92.245, Tokens/s/device: 13904.164, total_weights: 65536, loss: 5.719, lm_loss: 5.719, perplexity: 304.661 I0419 01:22:48.198247 133141823870784 metric_logger.py:196] completed step: 16, seconds: 72.048, TFLOP/s/device: 0.189, Tokens/s/device: 28.425, total_weights: 65536, loss: 5.605, lm_loss: 5.605, perplexity: 271.872 I0419 01:22:48.345830 133141823870784 metric_logger.py:196] completed step: 17, seconds: 0.007, TFLOP/s/device: 1916.109, Tokens/s/device: 288816.810, total_weights: 65536, loss: 5.517, lm_loss: 5.517, perplexity: 248.975 I0419 01:22:48.493310 133141823870784 metric_logger.py:196] completed step: 18, seconds: 0.146, TFLOP/s/device: 92.837, Tokens/s/device: 13993.372, total_weights: 65536, loss: 5.448, lm_loss: 5.448, perplexity: 232.333 I0419 01:22:48.640094 133141823870784 checkpointing.py:794] Waiting for step 19 to finish before checkpoint... I0419 01:22:48.640867 133141823870784 checkpointing.py:798] Waited 0.00079345703125 seconds for step 19 to finish before starting checkpointing. I0419 01:22:48.643517 133141823870784 checkpoint_manager.py:2020] [process=0][thread=MainThread][step=15][wait_until_finished] Waiting for Save Finalize thread (save_finalize) to complete. I0419 01:23:14.289867 133019029468928 base_pytree_checkpoint_handler.py:1282] [process=0][thread=write_metadata_after_commits] Commit + Array metadata written. Time taken: 43.116885s (commit=41.478056s, array_metadata_write=1.638829s) I0419 01:23:14.290979 133018451330816 base_pytree_checkpoint_handler.py:130] [process=0] /jax/orbax/write/gbytes_per_sec: 33.554 MiB/s (total gbytes: 1.5 GiB) (time elapsed: 47.075592279434204 s) (per-host) I0419 01:23:14.291115 133018451330816 async_checkpointer.py:90] [process=0][thread=async_save] 3 Handler Commit operations completed. Time taken: 47.035169s. I0419 01:23:16.250550 133018451330816 array_metadata_store.py:411] [process=0][thread=async_save] Validated ArrayMetadata from all 8 hosts. Time taken: 0.000219s. I0419 01:23:17.357667 133018451330816 base_pytree_checkpoint_handler.py:1406] [process=0][thread=async_save] Pytree save finalize (merge_ocdbt + ArrayMetadata validation) completed. Time taken: 2.149861s. use_zarr3=False, enable_post_merge_validation=True, directory=gs://lance-maxtext/linen_ckpt_xpk_feat_nnx_post_train_fixes_20260418_235007/linen_xpk_feat_nnx_post_train_fixes_20260418_235007_16_zarr3_false/checkpoints/15/items I0419 01:23:17.359303 133018451330816 atomicity.py:666] Finalizing gs://lance-maxtext/linen_ckpt_xpk_feat_nnx_post_train_fixes_20260418_235007/linen_xpk_feat_nnx_post_train_fixes_20260418_235007_16_zarr3_false/checkpoints/15/items I0419 01:23:17.890680 133018451330816 atomicity.py:666] Finalizing gs://lance-maxtext/linen_ckpt_xpk_feat_nnx_post_train_fixes_20260418_235007/linen_xpk_feat_nnx_post_train_fixes_20260418_235007_16_zarr3_false/checkpoints/15 I0419 01:23:20.098751 133018451330816 atomicity.py:847] [process=0][thread=async_save] Finished saving checkpoint (finalized tmp dir) to `gs://lance-maxtext/linen_ckpt_xpk_feat_nnx_post_train_fixes_20260418_235007/linen_xpk_feat_nnx_post_train_fixes_20260418_235007_16_zarr3_false/checkpoints/15`. I0419 01:23:20.099597 133018451330816 event_tracking.py:138] [process=0] [async] Finished save (blocking + background) in 53.59 seconds @ gs://lance-maxtext/linen_ckpt_xpk_feat_nnx_post_train_fixes_20260418_235007/linen_xpk_feat_nnx_post_train_fixes_20260418_235007_16_zarr3_false/checkpoints/15 I0419 01:23:20.100983 133018451330816 async_checkpointer.py:160] [process=0][thread=async_save] Background save thread done. Time taken: 52.845034s. I0419 01:23:20.101146 133019021076224 async_checkpointer.py:288] [process=0][thread=save_finalize] Done with waiting for background save thread=async_save. I0419 01:23:20.101232 133019021076224 async_checkpointer.py:298] [process=0][thread=save_finalize] No errors found in background save thread=async_save. I0419 01:23:20.101295 133019021076224 checkpoint_manager.py:2137] [process=0][thread=save_finalize][step=15] CheckpointManager Save Finalize is syncing with other hosts... I0419 01:23:20.102977 133019021076224 checkpoint_manager.py:2146] [process=0][thread=save_finalize][step=15] CheckpointManager Save Finalize is done on all hosts. I0419 01:23:20.103136 133141823870784 checkpoint_manager.py:2032] [process=0][thread=MainThread][step=15][wait_until_finished] Done waiting for Save Finalize thread (save_finalize) running at step=15. W0419 01:23:20.103264 133141823870784 checkpoint_manager.py:1452] Waiting for previous save to complete took 31.459751 seconds. If this number is high, consider checkpointing less frequently. I0419 01:23:20.104723 133141823870784 checkpoint_manager.py:1512] [process=0] Saving checkpoint at step 19 I0419 01:23:20.106854 133141823870784 event_tracking.py:70] [process=0] [async] Started save checkpoint @ gs://lance-maxtext/linen_ckpt_xpk_feat_nnx_post_train_fixes_20260418_235007/linen_xpk_feat_nnx_post_train_fixes_20260418_235007_16_zarr3_false/checkpoints/19. I0419 01:23:20.855707 133141823870784 jax_array_handlers.py:360] Scheduling D2H of 39 prioritized jax.Array. I0419 01:23:20.855802 133141823870784 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 I0419 01:23:20.881945 133141823870784 base_pytree_checkpoint_handler.py:154] [process=0][thread=MainThread] Initiated "orbax.checkpoint._src.serialization.jax_array_handlers.ArrayHandler".serialize. Time taken: 0.027133s I0419 01:23:20.882739 133141823870784 base_pytree_checkpoint_handler.py:130] [process=0] /jax/orbax/write/blocking_gbytes_per_sec: 48.752 GiB/s (total gbytes: 1.5 GiB) (time elapsed: 0.03164100646972656 s) (per-host) I0419 01:23:20.882819 133141823870784 base_pytree_checkpoint_handler.py:768] [process=0][thread=MainThread] Initiated Pytree async_save. Time taken: 0.031725s (batch_requests_ready=0.001932s, total_serialization_initiated=0.029082s, others=0.000712s) I0419 01:23:20.882899 133141823870784 composite_checkpoint_handler.py:715] [process=0][thread=MainThread] Initiated CompositeCheckpointHandler.async_save. Time taken: 0.035777s (all_items=0.000015s, per_item={'items': '0.00001502'}, temp_paths=0.035762) I0419 01:23:20.883520 133141823870784 event_tracking.py:125] [process=0] [async] Finished blocking save in 0.78 seconds. Continuing save @ gs://lance-maxtext/linen_ckpt_xpk_feat_nnx_post_train_fixes_20260418_235007/linen_xpk_feat_nnx_post_train_fixes_20260418_235007_16_zarr3_false/checkpoints/19. I0419 01:23:20.883825 133022200338176 async_checkpointer.py:76] [process=0][thread=async_save] Background save thread started. Deadline for this save operation is 2026-04-19 01:43:20.883776 I0419 01:23:20.887283 133141823870784 checkpoint_manager.py:1560] [process=0][thread=MainThread][step=19] Starting CheckpointManager Save Finalize thread=save_finalize I0419 01:23:20.887501 133019021076224 async_checkpointer.py:280] [process=0][thread=save_finalize] Waiting for background save thread=async_save. I0419 01:23:20.887622 133141823870784 standard_logger.py:34] {'step': 19, 'event_type': 'save', 'directory': 'gs://lance-maxtext/linen_ckpt_xpk_feat_nnx_post_train_fixes_20260418_235007/linen_xpk_feat_nnx_post_train_fixes_20260418_235007_16_zarr3_false/checkpoints', 'reached_preemption': False, 'preemption_received_at': None, 'synchronous': False, 'wait_for_prev_start_time': 1776561768.643484, 'wait_for_prev_duration_secs': 31.459750652313232, 'time_between_consecutive_saves_sec': 22.138405561447144, 'checkpointer_blocking_start_time': 1776561800.1047626, 'checkpointer_blocking_duration_secs': 0.77919602394104, 'get_old_steps_start_time': 1776561800.8839753, 'get_old_steps_duration_secs': 2.7179718017578125e-05, 'checkpoint_manager_blocking_start_time': 1776561768.6410995, 'checkpoint_manager_blocking_duration_secs': 32.24649000167847} I0419 01:23:20.887731 133141823870784 checkpointing.py:409] Started an asynchronous checkpoint save for step 19 I0419 01:23:20.887774 133141823870784 checkpoint_manager.py:2020] [process=0][thread=MainThread][step=19][wait_until_finished] Waiting for Save Finalize thread (save_finalize) to complete. I0419 01:23:21.095030 133018451330816 atomicity.py:140] Creating tmp directory gs://lance-maxtext/linen_ckpt_xpk_feat_nnx_post_train_fixes_20260418_235007/linen_xpk_feat_nnx_post_train_fixes_20260418_235007_16_zarr3_false/checkpoints/19 I0419 01:23:23.686602 133017382229760 atomicity.py:140] Creating tmp directory gs://lance-maxtext/linen_ckpt_xpk_feat_nnx_post_train_fixes_20260418_235007/linen_xpk_feat_nnx_post_train_fixes_20260418_235007_16_zarr3_false/checkpoints/19/items I0419 01:23:28.058275 133022217123584 array_metadata_store.py:203] [process=0][thread=array_type_handler] Wrote 39 array_metadata.ArrayMetadata to gs://lance-maxtext/linen_ckpt_xpk_feat_nnx_post_train_fixes_20260418_235007/linen_xpk_feat_nnx_post_train_fixes_20260418_235007_16_zarr3_false/checkpoints/19/items/array_metadatas/process_0 I0419 01:24:08.359788 133022208730880 base_pytree_checkpoint_handler.py:1282] [process=0][thread=write_metadata_after_commits] Commit + Array metadata written. Time taken: 43.135339s (commit=41.622161s, array_metadata_write=1.513179s) I0419 01:24:08.360907 133022200338176 base_pytree_checkpoint_handler.py:130] [process=0] /jax/orbax/write/gbytes_per_sec: 33.248 MiB/s (total gbytes: 1.5 GiB) (time elapsed: 47.509774684906006 s) (per-host) I0419 01:24:08.361022 133022200338176 async_checkpointer.py:90] [process=0][thread=async_save] 3 Handler Commit operations completed. Time taken: 47.477087s. I0419 01:24:10.707729 133022200338176 array_metadata_store.py:411] [process=0][thread=async_save] Validated ArrayMetadata from all 8 hosts. Time taken: 0.000235s. I0419 01:24:11.895335 133022200338176 base_pytree_checkpoint_handler.py:1406] [process=0][thread=async_save] Pytree save finalize (merge_ocdbt + ArrayMetadata validation) completed. Time taken: 2.222064s. use_zarr3=False, enable_post_merge_validation=True, directory=gs://lance-maxtext/linen_ckpt_xpk_feat_nnx_post_train_fixes_20260418_235007/linen_xpk_feat_nnx_post_train_fixes_20260418_235007_16_zarr3_false/checkpoints/19/items I0419 01:24:11.896966 133022200338176 atomicity.py:666] Finalizing gs://lance-maxtext/linen_ckpt_xpk_feat_nnx_post_train_fixes_20260418_235007/linen_xpk_feat_nnx_post_train_fixes_20260418_235007_16_zarr3_false/checkpoints/19/items I0419 01:24:12.406919 133022200338176 atomicity.py:666] Finalizing gs://lance-maxtext/linen_ckpt_xpk_feat_nnx_post_train_fixes_20260418_235007/linen_xpk_feat_nnx_post_train_fixes_20260418_235007_16_zarr3_false/checkpoints/19 I0419 01:24:14.234000 133022200338176 atomicity.py:847] [process=0][thread=async_save] Finished saving checkpoint (finalized tmp dir) to `gs://lance-maxtext/linen_ckpt_xpk_feat_nnx_post_train_fixes_20260418_235007/linen_xpk_feat_nnx_post_train_fixes_20260418_235007_16_zarr3_false/checkpoints/19`. I0419 01:24:14.234887 133022200338176 event_tracking.py:138] [process=0] [async] Finished save (blocking + background) in 54.13 seconds @ gs://lance-maxtext/linen_ckpt_xpk_feat_nnx_post_train_fixes_20260418_235007/linen_xpk_feat_nnx_post_train_fixes_20260418_235007_16_zarr3_false/checkpoints/19 I0419 01:24:14.236235 133022200338176 async_checkpointer.py:160] [process=0][thread=async_save] Background save thread done. Time taken: 53.352296s. I0419 01:24:14.236438 133019021076224 async_checkpointer.py:288] [process=0][thread=save_finalize] Done with waiting for background save thread=async_save. I0419 01:24:14.236548 133019021076224 async_checkpointer.py:298] [process=0][thread=save_finalize] No errors found in background save thread=async_save. I0419 01:24:14.236600 133019021076224 checkpoint_manager.py:2137] [process=0][thread=save_finalize][step=19] CheckpointManager Save Finalize is syncing with other hosts... I0419 01:24:14.238842 133019021076224 checkpoint_manager.py:2146] [process=0][thread=save_finalize][step=19] CheckpointManager Save Finalize is done on all hosts. I0419 01:24:14.238960 133141823870784 checkpoint_manager.py:2032] [process=0][thread=MainThread][step=19][wait_until_finished] Done waiting for Save Finalize thread (save_finalize) running at step=19. I0419 01:24:14.239055 133141823870784 checkpoint_manager.py:2009] [process=0][thread=MainThread][wait_until_finished] No Save Finalize thread to wait for. Returning. I0419 01:24:14.239962 133141823870784 metric_logger.py:196] completed step: 19, seconds: 0.147, TFLOP/s/device: 92.126, Tokens/s/device: 13886.252, total_weights: 65536, loss: 5.391, lm_loss: 5.391, perplexity: 219.349 Per train step: Total TFLOPs: 13.59 split as 93.93% learnable weight flops and 6.07% attention flops XPK End: Sun Apr 19 01:24:27 UTC 2026 EXIT_CODE=0