main (574ad3fb9) vs feat/nnx-set-defaults-true (574ad3fb9)| Metric | main 574ad3fb9 | feat/nnx-set-defaults-true 574ad3fb9 | Diff (feat/nnx-set-defaults-true − main) |
|---|---|---|---|
| Parameters | 1.104 billion | 1.104 billion | — |
| Final loss | 7.1810 | 7.1810 | 0 |
| TFLOP/s | 92.178 | 91.889 | -0.289 |
| Tok/s | 13894.1 | 13850.6 | -43.506 |
| Avg s/step | 8.384 | 7.532 | -0.852 |
| Memory % | 1.38 | 1.38 | 0 |
| JAX | 0.9.2 | 0.8.1 | — |
Diff = branch value − main value. Green = branch improved. Red = branch regressed.
XPK Start: Sat Apr 18 19:40:31 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 19:40:54.819016: 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 19:40:55.358118 133840646793024 max_utils.py:273] Attempting to initialize the jax distributed system... I0418 19:41:04.398563 133840646793024 distributed.py:149] Starting JAX distributed service on [::]:8482 I0418 19:41:04.401195 133840646793024 distributed.py:172] Connecting to JAX distributed service on mt-16-zarr3-true-sv-k8ltr-slice-job-0-0.mt-16-zarr3-true-sv-k8ltr:8482 I0418 19:41:05.909374 133840646793024 max_utils.py:284] Jax distributed system initialized! I0418 19:41:12.337016 133840646793024 max_utils.py:800] System Information: Jax Version: 0.9.2 I0418 19:41:12.337119 133840646793024 max_utils.py:801] System Information: Jaxlib Version: 0.9.2 I0418 19:41:12.337160 133840646793024 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 19:41:12.337193 133840646793024 train_utils.py:348] WARNING: Sequence packing is essentially ignored for synthetic data. Please use a real dataset to use sequence packing. I0418 19:41:13.030496 133840646793024 maxtext_utils.py:1551] Num_devices: 32, shape (1, 1, 1, 32, 1, 1, 1, 1, 1, 1, 1, 1, 1) I0418 19:41:13.030770 133840646793024 checkpointing.py:677] Setting up checkpoint logger... I0418 19:41:13.030822 133840646793024 checkpointing.py:233] Creating checkpoint manager with ocdbt=True and zarr3=True I0418 19:41:13.030867 133840646793024 pytree_checkpoint_handler.py:592] save_device_host_concurrent_bytes=None I0418 19:41:13.031199 133840646793024 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 0x79b993b140e0>, enable_pinned_host_transfer=False, save_concurrent_bytes: 96000000000 (89.4 GiB), restore_concurrent_bytes: 96000000000 (89.4 GiB) I0418 19:41:16.286580 133840646793024 checkpointing.py:265] Enabling policy for fixed interval checkpointing. I0418 19:41:16.286815 133840646793024 checkpoint_manager.py:708] [process=5][thread=MainThread] CheckpointManager init: checkpointers=None, item_names=('items',), item_handlers={'items': <orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeCheckpointHandler object at 0x79a52069bd10>}, handler_registry=None I0418 19:41:16.287057 133840646793024 composite_checkpoint_handler.py:237] Deferred registration for item: "items". Adding handler `<orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeCheckpointHandler object at 0x79a52069bd10>` 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 19:41:16.287107 133840646793024 composite_checkpoint_handler.py:237] Deferred registration for item: "metrics". Adding handler `<orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonCheckpointHandler object at 0x79a52069d7f0>` 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 19:41:16.287144 133840646793024 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 0x79a52069bd10>, ('items', <class 'orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeRestoreArgs'>): <orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeCheckpointHandler object at 0x79a52069bd10>, ('metrics', <class 'orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonSaveArgs'>): <orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonCheckpointHandler object at 0x79a52069d7f0>, ('metrics', <class 'orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonRestoreArgs'>): <orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonCheckpointHandler object at 0x79a52069d7f0>}). I0418 19:41:16.287483 133840646793024 abstract_checkpointer.py:35] orbax-checkpoint version: 0.11.34 I0418 19:41:16.287556 133840646793024 async_checkpointer.py:192] [process=5][thread=MainThread] Using barrier_sync_fn: <function get_barrier_sync_fn.<locals>._fn at 0x79a5202d3880> timeout: 1200 secs and primary_host=0 for async checkpoint writes I0418 19:41:17.830642 133840646793024 checkpoint_manager.py:1812] Found 0 checkpoint steps in gs://lance-maxtext/linen_ckpt_xpk_main_20260418_180002/linen_xpk_main_20260418_180002_16_zarr3_true/checkpoints I0418 19:41:17.877970 133840646793024 checkpoint_manager.py:929] [process=5][thread=MainThread] CheckpointManager created, primary_host=0, CheckpointManagerOptions=CheckpointManagerOptions(save_interval_steps=1, max_to_keep=None, keep_time_interval=None, keep_period=None, should_keep_fn=None, best_fn=None, best_mode='max', keep_checkpoints_without_metrics=True, step_prefix=None, step_format_fixed_length=None, step_name_format=None, create=True, cleanup_tmp_directories=False, save_on_steps=frozenset(), single_host_load_and_broadcast=False, todelete_subdir=None, todelete_full_path=None, enable_background_delete=False, read_only=False, enable_async_checkpointing=True, async_options=None, multiprocessing_options=MultiprocessingOptions(primary_host=0, active_processes=None, barrier_sync_key_prefix=None), should_save_fn=None, file_options=FileOptions(path_permission_mode=None), save_root_metadata=True, temporary_path_class=None, save_decision_policy=FixedIntervalPolicy(interval=5), preservation_policy=LatestN(n=None), prevent_write_metrics=False, enable_should_save_is_saving_in_progress_check=True, enable_per_process_directory_creation=False, lightweight_initialize=False), root_directory=gs://lance-maxtext/linen_ckpt_xpk_main_20260418_180002/linen_xpk_main_20260418_180002_16_zarr3_true/checkpoints: <orbax.checkpoint.checkpoint_manager.CheckpointManager object at 0x79a52069ae70> I0418 19:41:17.878124 133840646793024 checkpointing.py:301] Checkpoint manager created! I0418 19:41:18.798441 133840646793024 nnx_wrappers.py:437] Unknown Logical: bfloat16[32,2048,2048]...................................... ('activation_batch', 'activation_norm_length', 'activation_embed'). I0418 19:41:18.798566 133840646793024 nnx_wrappers.py:437] Unknown Physical: bfloat16[32,2048,2048]...................................... ('fsdp', None, None). I0418 19:41:19.174302 133840646793024 attentions.py:1088] attentions/inputs_q Logical: bfloat16[32,2048,2048]...................................... ('activation_batch', 'activation_attn_length', 'activation_attn_embed'). I0418 19:41:19.174396 133840646793024 attentions.py:1088] attentions/inputs_q Physical: bfloat16[32,2048,2048]...................................... ('fsdp', None, None). I0418 19:41:19.190680 133840646793024 attentions.py:1089] attentions/inputs_kv Logical: bfloat16[32,2048,2048]...................................... ('activation_batch', 'activation_attn_length', 'activation_attn_embed'). I0418 19:41:19.190748 133840646793024 attentions.py:1089] attentions/inputs_kv Physical: bfloat16[32,2048,2048]...................................... ('fsdp', None, None). I0418 19:41:19.215103 133840646793024 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 19:41:19.215173 133840646793024 attentions.py:1154] attentions/query Physical: bfloat16[32,2048,16,128].................................... ('fsdp', None, None, None). I0418 19:41:19.231564 133840646793024 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 19:41:19.231621 133840646793024 attentions.py:1155] attentions/key Physical: bfloat16[32,2048,16,128].................................... ('fsdp', None, None, None). I0418 19:41:19.247776 133840646793024 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 19:41:19.247834 133840646793024 attentions.py:1156] attentions/value Physical: bfloat16[32,2048,16,128].................................... ('fsdp', None, None, None). I0418 19:41:19.272228 133840646793024 attentions.py:1197] attentions/out Logical: bfloat16[32,2048,16,128].................................... ('activation_batch', 'activation_attn_length', 'activation_heads', 'activation_kv'). I0418 19:41:19.272297 133840646793024 attentions.py:1197] attentions/out Physical: bfloat16[32,2048,16,128].................................... ('fsdp', None, None, None). I0418 19:41:19.293247 133840646793024 linears.py:525] linears/x Logical: bfloat16[32,2048,7168]...................................... ('activation_batch', 'activation_length', 'activation_mlp'). I0418 19:41:19.293314 133840646793024 linears.py:525] linears/x Physical: bfloat16[32,2048,7168]...................................... ('fsdp', None, None). I0418 19:41:19.507429 133840646793024 checkpointing.py:577] checkpoint manager exists so trying to load this run's existing checkpoint I0418 19:41:19.507579 133840646793024 checkpointing.py:665] No existing checkpoints found, not restoring checkpoint. fsdp: 32 I0418 19:41:20.926434 133840646793024 maxtext_utils.py:1654] params/params/decoder/decoder_norm/scale Shape: float32[2048] Logical: P('norm',) Physical: (None,) I0418 19:41:20.926605 133840646793024 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 19:41:20.926656 133840646793024 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 19:41:20.926715 133840646793024 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 19:41:20.926769 133840646793024 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 19:41:20.926807 133840646793024 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 19:41:20.926859 133840646793024 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 19:41:20.926912 133840646793024 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 19:41:20.926949 133840646793024 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 19:41:20.926991 133840646793024 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 19:41:20.927035 133840646793024 maxtext_utils.py:1654] params/params/decoder/logits_dense/kernel Shape: float32[2048,32000] Logical: P('embed_vocab', 'vocab') Physical: ('fsdp', None) I0418 19:41:20.927083 133840646793024 maxtext_utils.py:1654] params/params/token_embedder/embedding Shape: float32[32000,2048] Logical: P('vocab', 'embed_vocab') Physical: (None, 'fsdp') I0418 19:41:21.414688 133840646793024 train.py:155] train/xent Logical: float32[32,2048]............................................ ('activation_embed_and_logits_batch', 'activation_length'). I0418 19:41:21.414781 133840646793024 train.py:155] train/xent Physical: float32[32,2048]............................................ ('fsdp', None). I0418 19:41:21.429993 133840646793024 train.py:162] train/z_loss Logical: float32[32,2048]............................................ ('activation_embed_and_logits_batch', 'activation_length'). I0418 19:41:21.430052 133840646793024 train.py:162] train/z_loss Physical: float32[32,2048]............................................ ('fsdp', None). I0418 19:41:32.417014 133840646793024 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 19:41:32.417875 133840646793024 metric_logger.py:301] number parameters: 1.104 billion I0418 19:41:43.974220 133840646793024 checkpointing.py:772] Waiting for step 0 to finish before checkpoint... I0418 19:41:44.127358 133840646793024 checkpointing.py:776] Waited 0.15311574935913086 seconds for step 0 to finish before starting checkpointing. I0418 19:41:44.130074 133840646793024 checkpoint_manager.py:2009] [process=5][thread=MainThread][wait_until_finished] No Save Finalize thread to wait for. Returning. I0418 19:41:44.131706 133840646793024 checkpoint_manager.py:1512] [process=5] Saving checkpoint at step 0 I0418 19:41:44.133169 133840646793024 event_tracking.py:70] [process=5] [async] Started save checkpoint @ gs://lance-maxtext/linen_ckpt_xpk_main_20260418_180002/linen_xpk_main_20260418_180002_16_zarr3_true/checkpoints/0. I0418 19:41:44.893867 133840646793024 signaling_client.py:364] Using JaxDistributedSignalingClient I0418 19:41:44.894841 133840646793024 jax_array_handlers.py:360] Scheduling D2H of 39 prioritized jax.Array. I0418 19:41:44.894897 133840646793024 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 19:41:45.166440 133840646793024 base_pytree_checkpoint_handler.py:154] [process=5][thread=MainThread] Initiated "orbax.checkpoint._src.serialization.jax_array_handlers.ArrayHandler".serialize. Time taken: 0.272650s I0418 19:41:45.166614 133840646793024 base_pytree_checkpoint_handler.py:130] [process=5] /jax/orbax/write/blocking_gbytes_per_sec: 5.545 GiB/s (total gbytes: 1.5 GiB) (time elapsed: 0.2781999111175537 s) (per-host) I0418 19:41:45.166666 133840646793024 base_pytree_checkpoint_handler.py:768] [process=5][thread=MainThread] Initiated Pytree async_save. Time taken: 0.278263s (batch_requests_ready=0.002319s, total_serialization_initiated=0.275873s, others=0.000071s) I0418 19:41:45.166760 133840646793024 composite_checkpoint_handler.py:715] [process=5][thread=MainThread] Initiated CompositeCheckpointHandler.async_save. Time taken: 0.282609s (all_items=0.000017s, per_item={'items': '0.00001717'}, temp_paths=0.282592) I0418 19:41:45.167543 133840646793024 event_tracking.py:125] [process=5] [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_true/checkpoints/0. I0418 19:41:45.167850 133711766771456 async_checkpointer.py:76] [process=5][thread=async_save] Background save thread started. Deadline for this save operation is 2026-04-18 20:01:45.167815 I0418 19:41:45.178604 133840646793024 checkpoint_manager.py:1560] [process=5][thread=MainThread][step=0] Starting CheckpointManager Save Finalize thread=save_finalize I0418 19:41:45.178866 133710134810368 async_checkpointer.py:280] [process=5][thread=save_finalize] Waiting for background save thread=async_save. I0418 19:41:45.179003 133840646793024 standard_logger.py:34] {'step': 0, 'event_type': 'save', 'directory': 'gs://lance-maxtext/linen_ckpt_xpk_main_20260418_180002/linen_xpk_main_20260418_180002_16_zarr3_true/checkpoints', 'reached_preemption': False, 'preemption_received_at': None, 'synchronous': False, 'wait_for_prev_start_time': 1776541304.1300542, 'wait_for_prev_duration_secs': 6.914138793945312e-05, 'time_between_consecutive_saves_sec': None, 'checkpointer_blocking_start_time': 1776541304.1317475, 'checkpointer_blocking_duration_secs': 1.0362529754638672, 'get_old_steps_start_time': 1776541305.1680255, 'get_old_steps_duration_secs': 3.170967102050781e-05, 'checkpoint_manager_blocking_start_time': 1776541304.1278975, 'checkpoint_manager_blocking_duration_secs': 1.0510647296905518} I0418 19:41:45.179117 133840646793024 checkpointing.py:408] Started an asynchronous checkpoint save for step 0 I0418 19:41:45.179167 133840646793024 max_utils.py:750] Memstats: After params initialized: I0418 19:41:45.179217 133840646793024 max_utils.py:756] Using (GB) 0.43 / 31.25 (1.376000%) on TPU_18(process=5,(2,4,0,0)) I0418 19:41:45.179250 133840646793024 max_utils.py:756] Using (GB) 0.43 / 31.25 (1.376000%) on TPU_19(process=5,(3,4,0,0)) I0418 19:41:45.179276 133840646793024 max_utils.py:756] Using (GB) 0.43 / 31.25 (1.376000%) on TPU_22(process=5,(2,5,0,0)) I0418 19:41:45.179308 133840646793024 max_utils.py:756] Using (GB) 0.43 / 31.25 (1.376000%) on TPU_23(process=5,(3,5,0,0)) I0418 19:41:45.494714 133840646793024 metric_logger.py:196] completed step: 0, seconds: 11.556, TFLOP/s/device: 1.176, Tokens/s/device: 177.221, total_weights: 65536, loss: 10.874, lm_loss: 10.874, perplexity: 52779.875 I0418 19:41:45.664395 133840646793024 metric_logger.py:196] completed step: 1, seconds: 1.519, TFLOP/s/device: 8.945, Tokens/s/device: 1348.256, total_weights: 65536, loss: 10.874, lm_loss: 10.874, perplexity: 52779.875 I0418 19:41:46.047515 133840646793024 metric_logger.py:196] completed step: 2, seconds: 0.022, TFLOP/s/device: 607.572, Tokens/s/device: 91579.842, total_weights: 65536, loss: 10.560, lm_loss: 10.560, perplexity: 38547.719 I0418 19:41:46.194944 133840646793024 metric_logger.py:196] completed step: 3, seconds: 0.383, TFLOP/s/device: 35.467, Tokens/s/device: 5346.002, total_weights: 65536, loss: 9.980, lm_loss: 9.980, perplexity: 21590.506 I0418 19:41:46.199642 133840646793024 checkpointing.py:772] Waiting for step 5 to finish before checkpoint... I0418 19:41:46.488964 133840646793024 checkpointing.py:776] Waited 0.2892897129058838 seconds for step 5 to finish before starting checkpointing. I0418 19:41:46.491967 133840646793024 checkpoint_manager.py:2020] [process=5][thread=MainThread][step=0][wait_until_finished] Waiting for Save Finalize thread (save_finalize) to complete. I0418 19:41:53.559629 2838 google_auth_provider.cc:181] Running on GCE, using service account 562977990677-compute@developer.gserviceaccount.com I0418 19:41:55.615010 133710143203072 array_metadata_store.py:203] [process=5][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_true/checkpoints/0/items/array_metadatas/process_5 I0418 19:42:27.619488 133711766771456 base_pytree_checkpoint_handler.py:130] [process=5] /jax/orbax/write/gbytes_per_sec: 36.966 MiB/s (total gbytes: 1.5 GiB) (time elapsed: 42.731032609939575 s) (per-host) I0418 19:42:27.619601 133711766771456 async_checkpointer.py:90] [process=5][thread=async_save] 3 Handler Commit operations completed. Time taken: 42.451644s. I0418 19:42:36.183617 133711766771456 async_checkpointer.py:160] [process=5][thread=async_save] Background save thread done. Time taken: 51.015644s. I0418 19:42:36.183895 133710134810368 async_checkpointer.py:288] [process=5][thread=save_finalize] Done with waiting for background save thread=async_save. I0418 19:42:36.184018 133710134810368 async_checkpointer.py:298] [process=5][thread=save_finalize] No errors found in background save thread=async_save. I0418 19:42:36.184069 133710134810368 checkpoint_manager.py:2137] [process=5][thread=save_finalize][step=0] CheckpointManager Save Finalize is syncing with other hosts... I0418 19:42:36.187174 133710134810368 checkpoint_manager.py:2146] [process=5][thread=save_finalize][step=0] CheckpointManager Save Finalize is done on all hosts. I0418 19:42:36.187361 133840646793024 checkpoint_manager.py:2032] [process=5][thread=MainThread][step=0][wait_until_finished] Done waiting for Save Finalize thread (save_finalize) running at step=0. W0418 19:42:36.187518 133840646793024 checkpoint_manager.py:1452] Waiting for previous save to complete took 49.695550 seconds. If this number is high, consider checkpointing less frequently. I0418 19:42:36.189364 133840646793024 checkpoint_manager.py:1512] [process=5] Saving checkpoint at step 5 I0418 19:42:36.191747 133840646793024 event_tracking.py:70] [process=5] [async] Started save checkpoint @ gs://lance-maxtext/linen_ckpt_xpk_main_20260418_180002/linen_xpk_main_20260418_180002_16_zarr3_true/checkpoints/5. I0418 19:42:36.921232 133840646793024 jax_array_handlers.py:360] Scheduling D2H of 39 prioritized jax.Array. I0418 19:42:36.921328 133840646793024 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 19:42:36.954641 133840646793024 base_pytree_checkpoint_handler.py:154] [process=5][thread=MainThread] Initiated "orbax.checkpoint._src.serialization.jax_array_handlers.ArrayHandler".serialize. Time taken: 0.034478s I0418 19:42:36.954806 133840646793024 base_pytree_checkpoint_handler.py:130] [process=5] /jax/orbax/write/blocking_gbytes_per_sec: 40.614 GiB/s (total gbytes: 1.5 GiB) (time elapsed: 0.03798174858093262 s) (per-host) I0418 19:42:36.954857 133840646793024 base_pytree_checkpoint_handler.py:768] [process=5][thread=MainThread] Initiated Pytree async_save. Time taken: 0.038041s (batch_requests_ready=0.001878s, total_serialization_initiated=0.036098s, others=0.000066s) I0418 19:42:36.954945 133840646793024 composite_checkpoint_handler.py:715] [process=5][thread=MainThread] Initiated CompositeCheckpointHandler.async_save. Time taken: 0.043204s (all_items=0.000015s, per_item={'items': '0.00001550'}, temp_paths=0.043189) I0418 19:42:36.955520 133840646793024 event_tracking.py:125] [process=5] [async] Finished blocking save in 0.77 seconds. Continuing save @ gs://lance-maxtext/linen_ckpt_xpk_main_20260418_180002/linen_xpk_main_20260418_180002_16_zarr3_true/checkpoints/5. I0418 19:42:36.955807 133689463772928 async_checkpointer.py:76] [process=5][thread=async_save] Background save thread started. Deadline for this save operation is 2026-04-18 20:02:36.955770 I0418 19:42:36.964551 133840646793024 checkpoint_manager.py:1560] [process=5][thread=MainThread][step=5] Starting CheckpointManager Save Finalize thread=save_finalize I0418 19:42:36.964816 133710134810368 async_checkpointer.py:280] [process=5][thread=save_finalize] Waiting for background save thread=async_save. I0418 19:42:36.964977 133840646793024 standard_logger.py:34] {'step': 5, 'event_type': 'save', 'directory': 'gs://lance-maxtext/linen_ckpt_xpk_main_20260418_180002/linen_xpk_main_20260418_180002_16_zarr3_true/checkpoints', 'reached_preemption': False, 'preemption_received_at': None, 'synchronous': False, 'wait_for_prev_start_time': 1776541306.4919386, 'wait_for_prev_duration_secs': 49.695550203323364, 'time_between_consecutive_saves_sec': None, 'checkpointer_blocking_start_time': 1776541356.1894033, 'checkpointer_blocking_duration_secs': 0.766512393951416, 'get_old_steps_start_time': 1776541356.9559362, 'get_old_steps_duration_secs': 2.5510787963867188e-05, 'checkpoint_manager_blocking_start_time': 1776541306.4896312, 'checkpoint_manager_blocking_duration_secs': 50.475311040878296} I0418 19:42:36.965082 133840646793024 checkpointing.py:408] Started an asynchronous checkpoint save for step 5 I0418 19:42:36.965886 133840646793024 metric_logger.py:196] completed step: 4, seconds: 0.154, TFLOP/s/device: 88.488, Tokens/s/device: 13337.849, total_weights: 65536, loss: 9.460, lm_loss: 9.460, perplexity: 12840.355 I0418 19:42:36.971978 133840646793024 metric_logger.py:196] completed step: 5, seconds: 0.147, TFLOP/s/device: 92.227, Tokens/s/device: 13901.427, total_weights: 65536, loss: 8.959, lm_loss: 8.959, perplexity: 7776.467 I0418 19:42:43.338792 133710143203072 array_metadata_store.py:203] [process=5][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_true/checkpoints/5/items/array_metadatas/process_5 I0418 19:42:59.281384 133840646793024 metric_logger.py:196] completed step: 6, seconds: 50.771, TFLOP/s/device: 0.268, Tokens/s/device: 40.338, total_weights: 65536, loss: 8.469, lm_loss: 8.469, perplexity: 4762.740 I0418 19:42:59.428873 133840646793024 metric_logger.py:196] completed step: 7, seconds: 22.163, TFLOP/s/device: 0.613, Tokens/s/device: 92.405, total_weights: 65536, loss: 8.003, lm_loss: 8.003, perplexity: 2989.151 I0418 19:42:59.576227 133840646793024 metric_logger.py:196] completed step: 8, seconds: 0.152, TFLOP/s/device: 89.190, Tokens/s/device: 13443.613, total_weights: 65536, loss: 7.572, lm_loss: 7.572, perplexity: 1943.098 I0418 19:42:59.723091 133840646793024 checkpointing.py:772] Waiting for step 9 to finish before checkpoint... I0418 19:42:59.723869 133840646793024 checkpointing.py:776] Waited 0.0007994174957275391 seconds for step 9 to finish before starting checkpointing. I0418 19:42:59.725897 133840646793024 checkpoint_manager.py:2020] [process=5][thread=MainThread][step=5][wait_until_finished] Waiting for Save Finalize thread (save_finalize) to complete. I0418 19:43:19.857069 133689463772928 base_pytree_checkpoint_handler.py:130] [process=5] /jax/orbax/write/gbytes_per_sec: 36.786 MiB/s (total gbytes: 1.5 GiB) (time elapsed: 42.9402060508728 s) (per-host) I0418 19:43:19.857176 133689463772928 async_checkpointer.py:90] [process=5][thread=async_save] 3 Handler Commit operations completed. Time taken: 42.900815s. I0418 19:43:28.432404 133689463772928 async_checkpointer.py:160] [process=5][thread=async_save] Background save thread done. Time taken: 51.476027s. I0418 19:43:28.432683 133710134810368 async_checkpointer.py:288] [process=5][thread=save_finalize] Done with waiting for background save thread=async_save. I0418 19:43:28.432803 133710134810368 async_checkpointer.py:298] [process=5][thread=save_finalize] No errors found in background save thread=async_save. I0418 19:43:28.432851 133710134810368 checkpoint_manager.py:2137] [process=5][thread=save_finalize][step=5] CheckpointManager Save Finalize is syncing with other hosts... I0418 19:43:28.434412 133710134810368 checkpoint_manager.py:2146] [process=5][thread=save_finalize][step=5] CheckpointManager Save Finalize is done on all hosts. I0418 19:43:28.434615 133840646793024 checkpoint_manager.py:2032] [process=5][thread=MainThread][step=5][wait_until_finished] Done waiting for Save Finalize thread (save_finalize) running at step=5. W0418 19:43:28.434745 133840646793024 checkpoint_manager.py:1452] Waiting for previous save to complete took 28.708856 seconds. If this number is high, consider checkpointing less frequently. I0418 19:43:28.436439 133840646793024 checkpoint_manager.py:1512] [process=5] Saving checkpoint at step 9 I0418 19:43:28.438435 133840646793024 event_tracking.py:70] [process=5] [async] Started save checkpoint @ gs://lance-maxtext/linen_ckpt_xpk_main_20260418_180002/linen_xpk_main_20260418_180002_16_zarr3_true/checkpoints/9. I0418 19:43:29.139147 133840646793024 jax_array_handlers.py:360] Scheduling D2H of 39 prioritized jax.Array. I0418 19:43:29.139243 133840646793024 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 19:43:29.171083 133840646793024 base_pytree_checkpoint_handler.py:154] [process=5][thread=MainThread] Initiated "orbax.checkpoint._src.serialization.jax_array_handlers.ArrayHandler".serialize. Time taken: 0.032752s I0418 19:43:29.171253 133840646793024 base_pytree_checkpoint_handler.py:130] [process=5] /jax/orbax/write/blocking_gbytes_per_sec: 42.501 GiB/s (total gbytes: 1.5 GiB) (time elapsed: 0.03629469871520996 s) (per-host) I0418 19:43:29.171303 133840646793024 base_pytree_checkpoint_handler.py:768] [process=5][thread=MainThread] Initiated Pytree async_save. Time taken: 0.036355s (batch_requests_ready=0.001771s, total_serialization_initiated=0.034519s, others=0.000065s) I0418 19:43:29.171393 133840646793024 composite_checkpoint_handler.py:715] [process=5][thread=MainThread] Initiated CompositeCheckpointHandler.async_save. Time taken: 0.040448s (all_items=0.000010s, per_item={'items': '0.00001025'}, temp_paths=0.040437) I0418 19:43:29.172065 133840646793024 event_tracking.py:125] [process=5] [async] Finished blocking save in 0.74 seconds. Continuing save @ gs://lance-maxtext/linen_ckpt_xpk_main_20260418_180002/linen_xpk_main_20260418_180002_16_zarr3_true/checkpoints/9. I0418 19:43:29.172330 133710134810368 async_checkpointer.py:76] [process=5][thread=async_save] Background save thread started. Deadline for this save operation is 2026-04-18 20:03:29.172305 I0418 19:43:29.176246 133840646793024 checkpoint_manager.py:1560] [process=5][thread=MainThread][step=9] Starting CheckpointManager Save Finalize thread=save_finalize I0418 19:43:29.176495 133710720530176 async_checkpointer.py:280] [process=5][thread=save_finalize] Waiting for background save thread=async_save. I0418 19:43:29.176640 133840646793024 standard_logger.py:34] {'step': 9, 'event_type': 'save', 'directory': 'gs://lance-maxtext/linen_ckpt_xpk_main_20260418_180002/linen_xpk_main_20260418_180002_16_zarr3_true/checkpoints', 'reached_preemption': False, 'preemption_received_at': None, 'synchronous': False, 'wait_for_prev_start_time': 1776541379.7258663, 'wait_for_prev_duration_secs': 28.708855628967285, 'time_between_consecutive_saves_sec': 23.53865146636963, 'checkpointer_blocking_start_time': 1776541408.4364924, 'checkpointer_blocking_duration_secs': 0.7359409332275391, 'get_old_steps_start_time': 1776541409.1724508, 'get_old_steps_duration_secs': 3.5762786865234375e-05, 'checkpoint_manager_blocking_start_time': 1776541379.724125, 'checkpoint_manager_blocking_duration_secs': 29.452481269836426} I0418 19:43:29.176743 133840646793024 checkpointing.py:408] Started an asynchronous checkpoint save for step 9 I0418 19:43:29.176785 133840646793024 checkpoint_manager.py:2020] [process=5][thread=MainThread][step=9][wait_until_finished] Waiting for Save Finalize thread (save_finalize) to complete. I0418 19:43:34.164968 133710143203072 array_metadata_store.py:203] [process=5][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_true/checkpoints/9/items/array_metadatas/process_5 I0418 19:44:11.084803 133710134810368 base_pytree_checkpoint_handler.py:130] [process=5] /jax/orbax/write/gbytes_per_sec: 37.654 MiB/s (total gbytes: 1.5 GiB) (time elapsed: 41.94980573654175 s) (per-host) I0418 19:44:11.084921 133710134810368 async_checkpointer.py:90] [process=5][thread=async_save] 3 Handler Commit operations completed. Time taken: 41.912513s. I0418 19:44:19.574669 133710134810368 async_checkpointer.py:160] [process=5][thread=async_save] Background save thread done. Time taken: 50.402244s. I0418 19:44:19.574945 133710720530176 async_checkpointer.py:288] [process=5][thread=save_finalize] Done with waiting for background save thread=async_save. I0418 19:44:19.575069 133710720530176 async_checkpointer.py:298] [process=5][thread=save_finalize] No errors found in background save thread=async_save. I0418 19:44:19.575111 133710720530176 checkpoint_manager.py:2137] [process=5][thread=save_finalize][step=9] CheckpointManager Save Finalize is syncing with other hosts... I0418 19:44:19.576806 133710720530176 checkpoint_manager.py:2146] [process=5][thread=save_finalize][step=9] CheckpointManager Save Finalize is done on all hosts. I0418 19:44:19.576997 133840646793024 checkpoint_manager.py:2032] [process=5][thread=MainThread][step=9][wait_until_finished] Done waiting for Save Finalize thread (save_finalize) running at step=9. I0418 19:44:19.577149 133840646793024 checkpoint_manager.py:2009] [process=5][thread=MainThread][wait_until_finished] No Save Finalize thread to wait for. Returning. I0418 19:44:19.578088 133840646793024 metric_logger.py:196] completed step: 9, seconds: 0.147, TFLOP/s/device: 92.178, Tokens/s/device: 13894.071, total_weights: 65536, loss: 7.181, lm_loss: 7.181, perplexity: 1314.495 Per train step: Total TFLOPs: 13.59 split as 93.93% learnable weight flops and 6.07% attention flops XPK End: Sat Apr 18 19:44:31 UTC 2026 EXIT_CODE=0
XPK Start: Sun Apr 19 04:36:57 UTC 2026 2026-04-19 04:37:01.594419: 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 00:00:1776573421.607228 10 cuda_dnn.cc:8579] Unable to register cuDNN factory: Attempting to register factory for plugin cuDNN when one has already been registered E0000 00:00:1776573421.610942 10 cuda_blas.cc:1407] Unable to register cuBLAS factory: Attempting to register factory for plugin cuBLAS when one has already been registered W0000 00:00:1776573421.622173 10 computation_placer.cc:177] computation placer already registered. Please check linkage and avoid linking the same target more than once. W0000 00:00:1776573421.622193 10 computation_placer.cc:177] computation placer already registered. Please check linkage and avoid linking the same target more than once. W0000 00:00:1776573421.622195 10 computation_placer.cc:177] computation placer already registered. Please check linkage and avoid linking the same target more than once. W0000 00:00:1776573421.622197 10 computation_placer.cc:177] computation placer already registered. Please check linkage and avoid linking the same target more than once. 2026-04-19 04:37:21.419966: 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 04:37:21.947899 133259682899776 max_utils.py:273] Attempting to initialize the jax distributed system... INFO:2026-04-19 04:37:30,989:jax._src.distributed:140: Starting JAX distributed service on [::]:8482 I0419 04:37:30.989592 133259682899776 distributed.py:140] Starting JAX distributed service on [::]:8482 INFO:2026-04-19 04:37:30,991:jax._src.distributed:157: Connecting to JAX distributed service on mt-16-zarr3-true-sv-ryke5-slice-job-0-0.mt-16-zarr3-true-sv-ryke5:8482 I0419 04:37:30.991935 133259682899776 distributed.py:157] Connecting to JAX distributed service on mt-16-zarr3-true-sv-ryke5-slice-job-0-0.mt-16-zarr3-true-sv-ryke5:8482 I0419 04:37:32.026210 133259682899776 max_utils.py:284] Jax distributed system initialized! I0419 04:37:38.887553 133259682899776 max_utils.py:800] System Information: Jax Version: 0.8.1 I0419 04:37:38.887658 133259682899776 max_utils.py:801] System Information: Jaxlib Version: 0.8.1 I0419 04:37:38.887698 133259682899776 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 I0419 04:37:38.887732 133259682899776 train_utils.py:377] WARNING: Sequence packing is essentially ignored for synthetic data. Please use a real dataset to use sequence packing. I0419 04:37:39.508362 133259682899776 maxtext_utils.py:1631] Num_devices: 32, shape (1, 1, 1, 32, 1, 1, 1, 1, 1, 1, 1, 1, 1) I0419 04:37:39.508957 133259682899776 maxtext_utils.py:1631] Num_devices: 32, shape (1, 1, 1, 32, 1, 1, 1, 1, 1, 1, 1, 1, 1) I0419 04:37:39.509140 133259682899776 checkpointing.py:688] Setting up checkpoint logger... I0419 04:37:39.509190 133259682899776 checkpointing.py:234] Creating checkpoint manager with ocdbt=True and zarr3=True I0419 04:37:39.509248 133259682899776 pytree_checkpoint_handler.py:589] save_device_host_concurrent_bytes=None I0419 04:37:39.509594 133259682899776 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 0x7932673712b0>, enable_pinned_host_transfer=False, save_concurrent_bytes: 96000000000 (89.4 GiB), restore_concurrent_bytes: 96000000000 (89.4 GiB) I0419 04:37:42.393896 133259682899776 checkpointing.py:266] Enabling policy for fixed interval checkpointing. I0419 04:37:42.394257 133259682899776 checkpoint_manager.py:709] [process=6][thread=MainThread] CheckpointManager init: checkpointers=None, item_names=('items',), item_handlers={'items': <orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeCheckpointHandler object at 0x791f59359a00>}, handler_registry=None I0419 04:37:42.394494 133259682899776 composite_checkpoint_handler.py:237] Deferred registration for item: "items". Adding handler `<orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeCheckpointHandler object at 0x791f59359a00>` 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 04:37:42.394550 133259682899776 composite_checkpoint_handler.py:237] Deferred registration for item: "metrics". Adding handler `<orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonCheckpointHandler object at 0x791f5935e840>` 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 04:37:42.394587 133259682899776 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 0x791f59359a00>, ('items', <class 'orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeRestoreArgs'>): <orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeCheckpointHandler object at 0x791f59359a00>, ('metrics', <class 'orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonSaveArgs'>): <orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonCheckpointHandler object at 0x791f5935e840>, ('metrics', <class 'orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonRestoreArgs'>): <orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonCheckpointHandler object at 0x791f5935e840>}). I0419 04:37:42.395085 133259682899776 abstract_checkpointer.py:35] orbax-checkpoint version: 0.11.33 I0419 04:37:42.395164 133259682899776 async_checkpointer.py:177] [process=6][thread=MainThread] Using barrier_sync_fn: <function get_barrier_sync_fn.<locals>._fn at 0x791de4484180> timeout: 600 secs and primary_host=0 for async checkpoint writes I0419 04:37:43.961054 133259682899776 checkpoint_manager.py:1818] Found 0 checkpoint steps in gs://lance-maxtext/linen_ckpt_xpk_feat_nnx_set_defaults_true_20260419_025828/linen_xpk_feat_nnx_set_defaults_true_20260419_025828_16_zarr3_true/checkpoints I0419 04:37:44.375409 133259682899776 checkpoint_manager.py:929] [process=6][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_set_defaults_true_20260419_025828/linen_xpk_feat_nnx_set_defaults_true_20260419_025828_16_zarr3_true/checkpoints: <orbax.checkpoint.checkpoint_manager.CheckpointManager object at 0x791f5935ac60> I0419 04:37:44.375592 133259682899776 checkpointing.py:302] Checkpoint manager created! I0419 04:37:45.265837 133259682899776 nnx_wrappers.py:455] Unknown Logical: bfloat16[32,2048,2048]...................................... ('activation_batch', 'activation_norm_length', 'activation_embed'). I0419 04:37:45.265944 133259682899776 nnx_wrappers.py:455] Unknown Physical: bfloat16[32,2048,2048]...................................... ('fsdp', None, None). I0419 04:37:45.638115 133259682899776 attentions.py:1084] attentions/inputs_q Logical: bfloat16[32,2048,2048]...................................... ('activation_batch', 'activation_attn_length', 'activation_attn_embed'). I0419 04:37:45.638209 133259682899776 attentions.py:1084] attentions/inputs_q Physical: bfloat16[32,2048,2048]...................................... ('fsdp', None, None). I0419 04:37:45.654143 133259682899776 attentions.py:1085] attentions/inputs_kv Logical: bfloat16[32,2048,2048]...................................... ('activation_batch', 'activation_attn_length', 'activation_attn_embed'). I0419 04:37:45.654206 133259682899776 attentions.py:1085] attentions/inputs_kv Physical: bfloat16[32,2048,2048]...................................... ('fsdp', None, None). I0419 04:37:45.676919 133259682899776 attentions.py:1150] attentions/query Logical: bfloat16[32,2048,16,128].................................... ('activation_kv_batch', 'activation_attn_length', 'activation_kv_heads', 'activation_kv_head_dim'). I0419 04:37:45.676990 133259682899776 attentions.py:1150] attentions/query Physical: bfloat16[32,2048,16,128].................................... ('fsdp', None, None, None). I0419 04:37:45.692962 133259682899776 attentions.py:1151] attentions/key Logical: bfloat16[32,2048,16,128].................................... ('activation_kv_batch', 'activation_attn_length', 'activation_kv_heads', 'activation_kv_head_dim'). I0419 04:37:45.693026 133259682899776 attentions.py:1151] attentions/key Physical: bfloat16[32,2048,16,128].................................... ('fsdp', None, None, None). I0419 04:37:45.708788 133259682899776 attentions.py:1152] attentions/value Logical: bfloat16[32,2048,16,128].................................... ('activation_kv_batch', 'activation_attn_length', 'activation_kv_heads', 'activation_kv_head_dim'). I0419 04:37:45.708851 133259682899776 attentions.py:1152] attentions/value Physical: bfloat16[32,2048,16,128].................................... ('fsdp', None, None, None). I0419 04:37:45.733794 133259682899776 attentions.py:1193] attentions/out Logical: bfloat16[32,2048,16,128].................................... ('activation_batch', 'activation_attn_length', 'activation_heads', 'activation_kv'). I0419 04:37:45.733866 133259682899776 attentions.py:1193] attentions/out Physical: bfloat16[32,2048,16,128].................................... ('fsdp', None, None, None). I0419 04:37:45.753325 133259682899776 linears.py:541] linears/x Logical: bfloat16[32,2048,7168]...................................... ('activation_batch', 'activation_length', 'activation_mlp'). I0419 04:37:45.753391 133259682899776 linears.py:541] linears/x Physical: bfloat16[32,2048,7168]...................................... ('fsdp', None, None). I0419 04:37:45.942214 133259682899776 checkpointing.py:578] checkpoint manager exists so trying to load this run's existing checkpoint I0419 04:37:45.942340 133259682899776 checkpointing.py:676] No existing checkpoints found, not restoring checkpoint. fsdp: 32 I0419 04:37:47.371296 133259682899776 maxtext_utils.py:1740] params/params/decoder/decoder_norm/scale Shape: float32[2048] Logical: PartitionSpec('norm',) Physical: (None,) I0419 04:37:47.371438 133259682899776 maxtext_utils.py:1740] params/params/decoder/layers/mlp/wi_0/kernel Shape: float32[2048,16,7168] Logical: PartitionSpec('embed', 'layers', 'mlp') Physical: ('fsdp', None, None) I0419 04:37:47.371497 133259682899776 maxtext_utils.py:1740] params/params/decoder/layers/mlp/wi_1/kernel Shape: float32[2048,16,7168] Logical: PartitionSpec('embed', 'layers', 'mlp') Physical: ('fsdp', None, None) I0419 04:37:47.371559 133259682899776 maxtext_utils.py:1740] params/params/decoder/layers/mlp/wo/kernel Shape: float32[7168,16,2048] Logical: PartitionSpec('mlp', 'layers', 'embed') Physical: (None, None, 'fsdp') I0419 04:37:47.371618 133259682899776 maxtext_utils.py:1740] params/params/decoder/layers/post_self_attention_layer_norm/scale Shape: float32[2048,16] Logical: PartitionSpec('norm', 'layers') Physical: (None, None) I0419 04:37:47.371658 133259682899776 maxtext_utils.py:1740] params/params/decoder/layers/pre_self_attention_layer_norm/scale Shape: float32[2048,16] Logical: PartitionSpec('norm', 'layers') Physical: (None, None) I0419 04:37:47.371712 133259682899776 maxtext_utils.py:1740] 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, None, None) I0419 04:37:47.371767 133259682899776 maxtext_utils.py:1740] params/params/decoder/layers/self_attention/out/kernel Shape: float32[16,16,128,2048] Logical: PartitionSpec('heads', 'layers', 'kv', 'embed') Physical: (None, None, None, 'fsdp') I0419 04:37:47.371809 133259682899776 maxtext_utils.py:1740] params/params/decoder/layers/self_attention/query/kernel Shape: float32[2048,16,16,128] Logical: PartitionSpec('embed', 'layers', 'q_heads', 'kv') Physical: ('fsdp', None, None, None) I0419 04:37:47.371848 133259682899776 maxtext_utils.py:1740] 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, None, None) I0419 04:37:47.371898 133259682899776 maxtext_utils.py:1740] params/params/decoder/logits_dense/kernel Shape: float32[2048,32000] Logical: PartitionSpec('embed', 'vocab') Physical: ('fsdp', None) I0419 04:37:47.371947 133259682899776 maxtext_utils.py:1740] params/params/token_embedder/embedding Shape: float32[32000,2048] Logical: PartitionSpec('vocab', 'embed') Physical: (None, 'fsdp') I0419 04:37:47.876453 133259682899776 train.py:158] train/xent Logical: float32[32,2048]............................................ ('activation_embed_and_logits_batch', 'activation_length'). I0419 04:37:47.876554 133259682899776 train.py:158] train/xent Physical: float32[32,2048]............................................ ('fsdp', None). I0419 04:37:47.891412 133259682899776 train.py:165] train/z_loss Logical: float32[32,2048]............................................ ('activation_embed_and_logits_batch', 'activation_length'). I0419 04:37:47.891474 133259682899776 train.py:165] train/z_loss Physical: float32[32,2048]............................................ ('fsdp', None). I0419 04:37:58.636294 133259682899776 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 04:37:58.637175 133259682899776 metric_logger.py:289] number parameters: 1.104 billion I0419 04:38:09.788065 133259682899776 checkpointing.py:794] Waiting for step 0 to finish before checkpoint... I0419 04:38:10.147322 133259682899776 checkpointing.py:798] Waited 0.3592374324798584 seconds for step 0 to finish before starting checkpointing. I0419 04:38:10.149815 133259682899776 checkpoint_manager.py:2013] [process=6][thread=MainThread][wait_until_finished] No Save Finalize thread to wait for. Returning. I0419 04:38:10.151706 133259682899776 checkpoint_manager.py:1518] [process=6] Saving checkpoint at step 0 I0419 04:38:10.154516 133259682899776 async_checkpointer.py:452] [process=6] Started async saving checkpoint to gs://lance-maxtext/linen_ckpt_xpk_feat_nnx_set_defaults_true_20260419_025828/linen_xpk_feat_nnx_set_defaults_true_20260419_025828_16_zarr3_true/checkpoints/0. I0419 04:38:11.181179 133259682899776 signaling_client.py:364] Using JaxDistributedSignalingClient I0419 04:38:11.182183 133259682899776 jax_array_handlers.py:358] Scheduling D2H of 39 prioritized jax.Array. I0419 04:38:11.182256 133259682899776 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 04:38:11.426133 133259682899776 base_pytree_checkpoint_handler.py:153] [process=6][thread=MainThread] Initiated "orbax.checkpoint._src.serialization.jax_array_handlers.ArrayHandler".serialize. Time taken: 0.245034s I0419 04:38:11.426311 133259682899776 base_pytree_checkpoint_handler.py:129] [process=6] /jax/checkpoint/write/blocking_gbytes_per_sec: 2.950 GiB/s (total gbytes: 1.5 GiB) (time elapsed: 0.5229711532592773 s) (per-host) I0419 04:38:11.426381 133259682899776 base_pytree_checkpoint_handler.py:737] [process=6][thread=MainThread] Initiated Pytree async_save. Time taken: 0.523045s (batch_requests_ready=0.274234s, total_serialization_initiated=0.248730s, others=0.000082s) I0419 04:38:11.426490 133259682899776 composite_checkpoint_handler.py:715] [process=6][thread=MainThread] Initiated CompositeCheckpointHandler.async_save. Time taken: 0.527297s (all_items=0.000020s, per_item={'items': '0.00001979'}, temp_paths=0.527277) I0419 04:38:11.427501 133136369301248 async_checkpointer.py:79] [process=6][thread=async_save] Background save thread started. I0419 04:38:11.427677 133259682899776 async_checkpointer.py:561] Finished blocking save. Time taken: 1.275899s. Continuing background save to gs://lance-maxtext/linen_ckpt_xpk_feat_nnx_set_defaults_true_20260419_025828/linen_xpk_feat_nnx_set_defaults_true_20260419_025828_16_zarr3_true/checkpoints/0. I0419 04:38:11.883494 133259682899776 checkpoint_manager.py:1566] [process=6][thread=MainThread][step=0] Starting CheckpointManager Save Finalize thread=save_finalize I0419 04:38:11.883894 133131545855744 async_checkpointer.py:265] [process=6][thread=save_finalize] Waiting for background save thread=async_save. I0419 04:38:11.884072 133259682899776 standard_logger.py:34] {'step': 0, 'event_type': 'save', 'directory': 'gs://lance-maxtext/linen_ckpt_xpk_feat_nnx_set_defaults_true_20260419_025828/linen_xpk_feat_nnx_set_defaults_true_20260419_025828_16_zarr3_true/checkpoints', 'reached_preemption': False, 'preemption_received_at': None, 'synchronous': False, 'wait_for_prev_start_time': 1776573490.1497955, 'wait_for_prev_duration_secs': 6.651878356933594e-05, 'time_between_consecutive_saves_sec': None, 'checkpointer_blocking_start_time': 1776573490.151745, 'checkpointer_blocking_duration_secs': 1.2760400772094727, 'get_old_steps_start_time': 1776573491.4278076, 'get_old_steps_duration_secs': 2.574920654296875e-05, 'checkpoint_manager_blocking_start_time': 1776573490.1479526, 'checkpoint_manager_blocking_duration_secs': 1.7360773086547852} I0419 04:38:11.884179 133259682899776 checkpointing.py:409] Started an asynchronous checkpoint save for step 0 I0419 04:38:11.884232 133259682899776 max_utils.py:750] Memstats: After params initialized: I0419 04:38:11.884304 133259682899776 max_utils.py:756] Using (GB) 0.43 / 31.25 (1.376000%) on TPU_24(process=6,(0,6,0,0)) I0419 04:38:11.884340 133259682899776 max_utils.py:756] Using (GB) 0.43 / 31.25 (1.376000%) on TPU_25(process=6,(1,6,0,0)) I0419 04:38:11.884368 133259682899776 max_utils.py:756] Using (GB) 0.43 / 31.25 (1.376000%) on TPU_28(process=6,(0,7,0,0)) I0419 04:38:11.884393 133259682899776 max_utils.py:756] Using (GB) 0.43 / 31.25 (1.376000%) on TPU_29(process=6,(1,7,0,0)) I0419 04:38:12.142526 133259682899776 metric_logger.py:185] completed step: 0, seconds: 11.151, TFLOP/s/device: 1.218, Tokens/s/device: 183.664, total_weights: 65536, loss: 10.874 I0419 04:38:12.331426 133259682899776 metric_logger.py:185] completed step: 1, seconds: 2.353, TFLOP/s/device: 5.774, Tokens/s/device: 870.318, total_weights: 65536, loss: 10.874 I0419 04:38:12.723222 133259682899776 metric_logger.py:185] completed step: 2, seconds: 0.042, TFLOP/s/device: 325.136, Tokens/s/device: 49008.112, total_weights: 65536, loss: 10.560 I0419 04:38:12.870841 133259682899776 metric_logger.py:185] completed step: 3, seconds: 0.392, TFLOP/s/device: 34.674, Tokens/s/device: 5226.450, total_weights: 65536, loss: 9.980 I0419 04:38:12.875540 133259682899776 checkpointing.py:794] Waiting for step 5 to finish before checkpoint... I0419 04:38:13.165518 133259682899776 checkpointing.py:798] Waited 0.2899470329284668 seconds for step 5 to finish before starting checkpointing. I0419 04:38:13.168378 133259682899776 checkpoint_manager.py:2024] [process=6][thread=MainThread][step=0][wait_until_finished] Waiting for Save Finalize thread (save_finalize) to complete. I0419 04:38:14.338096 2648 google_auth_provider.cc:181] Running on GCE, using service account 562977990677-compute@developer.gserviceaccount.com I0419 04:38:17.084656 133136377693952 array_metadata_store.py:203] [process=6][thread=array_type_handler] Wrote 39 array_metadata.ArrayMetadata to gs://lance-maxtext/linen_ckpt_xpk_feat_nnx_set_defaults_true_20260419_025828/linen_xpk_feat_nnx_set_defaults_true_20260419_025828_16_zarr3_true/checkpoints/0/items/array_metadatas/process_6 I0419 04:38:48.785707 133136369301248 base_pytree_checkpoint_handler.py:129] [process=6] /jax/checkpoint/write/gbytes_per_sec: 41.697 MiB/s (total gbytes: 1.5 GiB) (time elapsed: 37.88232445716858 s) (per-host) I0419 04:38:48.785828 133136369301248 async_checkpointer.py:90] [process=6][thread=async_save] 3 Handler Commit operations completed. Time taken: 37.358212s. I0419 04:38:56.478066 133136369301248 async_checkpointer.py:144] [process=6][thread=async_save] Background save thread done. Time taken: 45.050433s. I0419 04:38:56.478383 133131545855744 async_checkpointer.py:273] [process=6][thread=save_finalize] Done with waiting for background save thread=async_save. I0419 04:38:56.478506 133131545855744 async_checkpointer.py:283] [process=6][thread=save_finalize] No errors found in background save thread=async_save. I0419 04:38:56.478555 133131545855744 checkpoint_manager.py:2133] [process=6][thread=save_finalize][step=0] CheckpointManager Save Finalize is syncing with other hosts... I0419 04:38:56.480298 133131545855744 checkpoint_manager.py:2142] [process=6][thread=save_finalize][step=0] CheckpointManager Save Finalize is done on all hosts. I0419 04:38:56.480475 133259682899776 checkpoint_manager.py:2036] [process=6][thread=MainThread][step=0][wait_until_finished] Done waiting for Save Finalize thread (save_finalize) running at step=0. W0419 04:38:56.480612 133259682899776 checkpoint_manager.py:1458] Waiting for previous save to complete took 43.312238 seconds. If this number is high, consider checkpointing less frequently. I0419 04:38:56.482690 133259682899776 checkpoint_manager.py:1518] [process=6] Saving checkpoint at step 5 I0419 04:38:56.486824 133259682899776 async_checkpointer.py:452] [process=6] Started async saving checkpoint to gs://lance-maxtext/linen_ckpt_xpk_feat_nnx_set_defaults_true_20260419_025828/linen_xpk_feat_nnx_set_defaults_true_20260419_025828_16_zarr3_true/checkpoints/5. I0419 04:38:57.866074 133259682899776 jax_array_handlers.py:358] Scheduling D2H of 39 prioritized jax.Array. I0419 04:38:57.866174 133259682899776 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 04:38:57.909632 133259682899776 base_pytree_checkpoint_handler.py:153] [process=6][thread=MainThread] Initiated "orbax.checkpoint._src.serialization.jax_array_handlers.ArrayHandler".serialize. Time taken: 0.044931s I0419 04:38:57.909798 133259682899776 base_pytree_checkpoint_handler.py:129] [process=6] /jax/checkpoint/write/blocking_gbytes_per_sec: 5.204 GiB/s (total gbytes: 1.5 GiB) (time elapsed: 0.2964344024658203 s) (per-host) I0419 04:38:57.909852 133259682899776 base_pytree_checkpoint_handler.py:737] [process=6][thread=MainThread] Initiated Pytree async_save. Time taken: 0.296496s (batch_requests_ready=0.249659s, total_serialization_initiated=0.046769s, others=0.000069s) I0419 04:38:57.909943 133259682899776 composite_checkpoint_handler.py:715] [process=6][thread=MainThread] Initiated CompositeCheckpointHandler.async_save. Time taken: 0.301402s (all_items=0.000015s, per_item={'items': '0.00001526'}, temp_paths=0.301387) I0419 04:38:57.910812 133128349808384 async_checkpointer.py:79] [process=6][thread=async_save] Background save thread started. I0419 04:38:57.910972 133259682899776 async_checkpointer.py:561] Finished blocking save. Time taken: 1.428215s. Continuing background save to gs://lance-maxtext/linen_ckpt_xpk_feat_nnx_set_defaults_true_20260419_025828/linen_xpk_feat_nnx_set_defaults_true_20260419_025828_16_zarr3_true/checkpoints/5. I0419 04:38:58.350492 133259682899776 checkpoint_manager.py:1566] [process=6][thread=MainThread][step=5] Starting CheckpointManager Save Finalize thread=save_finalize I0419 04:38:58.350847 133131545855744 async_checkpointer.py:265] [process=6][thread=save_finalize] Waiting for background save thread=async_save. I0419 04:38:58.351016 133259682899776 standard_logger.py:34] {'step': 5, 'event_type': 'save', 'directory': 'gs://lance-maxtext/linen_ckpt_xpk_feat_nnx_set_defaults_true_20260419_025828/linen_xpk_feat_nnx_set_defaults_true_20260419_025828_16_zarr3_true/checkpoints', 'reached_preemption': False, 'preemption_received_at': None, 'synchronous': False, 'wait_for_prev_start_time': 1776573493.1683476, 'wait_for_prev_duration_secs': 43.31223821640015, 'time_between_consecutive_saves_sec': None, 'checkpointer_blocking_start_time': 1776573536.4827285, 'checkpointer_blocking_duration_secs': 1.4283509254455566, 'get_old_steps_start_time': 1776573537.9110992, 'get_old_steps_duration_secs': 2.4080276489257812e-05, 'checkpoint_manager_blocking_start_time': 1776573493.1661892, 'checkpoint_manager_blocking_duration_secs': 45.18478989601135} I0419 04:38:58.351139 133259682899776 checkpointing.py:409] Started an asynchronous checkpoint save for step 5 I0419 04:38:58.351744 133259682899776 metric_logger.py:185] completed step: 4, seconds: 0.153, TFLOP/s/device: 88.674, Tokens/s/device: 13365.878, total_weights: 65536, loss: 9.460 I0419 04:38:58.357678 133259682899776 metric_logger.py:185] completed step: 5, seconds: 0.147, TFLOP/s/device: 92.216, Tokens/s/device: 13899.824, total_weights: 65536, loss: 8.959 I0419 04:39:17.421304 133259682899776 metric_logger.py:185] completed step: 6, seconds: 45.481, TFLOP/s/device: 0.299, Tokens/s/device: 45.030, total_weights: 65536, loss: 8.469 I0419 04:39:17.568938 133259682899776 metric_logger.py:185] completed step: 7, seconds: 18.917, TFLOP/s/device: 0.718, Tokens/s/device: 108.260, total_weights: 65536, loss: 8.003 I0419 04:39:17.716461 133259682899776 metric_logger.py:185] completed step: 8, seconds: 0.152, TFLOP/s/device: 89.327, Tokens/s/device: 13464.383, total_weights: 65536, loss: 7.572 I0419 04:39:17.863472 133259682899776 checkpointing.py:794] Waiting for step 9 to finish before checkpoint... I0419 04:39:17.864125 133259682899776 checkpointing.py:798] Waited 0.0006732940673828125 seconds for step 9 to finish before starting checkpointing. I0419 04:39:17.866085 133259682899776 checkpoint_manager.py:2024] [process=6][thread=MainThread][step=5][wait_until_finished] Waiting for Save Finalize thread (save_finalize) to complete. I0419 04:39:22.199845 133136377693952 array_metadata_store.py:203] [process=6][thread=array_type_handler] Wrote 39 array_metadata.ArrayMetadata to gs://lance-maxtext/linen_ckpt_xpk_feat_nnx_set_defaults_true_20260419_025828/linen_xpk_feat_nnx_set_defaults_true_20260419_025828_16_zarr3_true/checkpoints/5/items/array_metadatas/process_6 I0419 04:39:58.384996 133128349808384 base_pytree_checkpoint_handler.py:129] [process=6] /jax/checkpoint/write/gbytes_per_sec: 25.992 MiB/s (total gbytes: 1.5 GiB) (time elapsed: 60.77159833908081 s) (per-host) I0419 04:39:58.385107 133128349808384 async_checkpointer.py:90] [process=6][thread=async_save] 3 Handler Commit operations completed. Time taken: 60.474180s. I0419 04:40:06.469492 133128349808384 async_checkpointer.py:144] [process=6][thread=async_save] Background save thread done. Time taken: 68.558551s. I0419 04:40:06.469800 133131545855744 async_checkpointer.py:273] [process=6][thread=save_finalize] Done with waiting for background save thread=async_save. I0419 04:40:06.469932 133131545855744 async_checkpointer.py:283] [process=6][thread=save_finalize] No errors found in background save thread=async_save. I0419 04:40:06.469987 133131545855744 checkpoint_manager.py:2133] [process=6][thread=save_finalize][step=5] CheckpointManager Save Finalize is syncing with other hosts... I0419 04:40:06.472488 133131545855744 checkpoint_manager.py:2142] [process=6][thread=save_finalize][step=5] CheckpointManager Save Finalize is done on all hosts. I0419 04:40:06.472659 133259682899776 checkpoint_manager.py:2036] [process=6][thread=MainThread][step=5][wait_until_finished] Done waiting for Save Finalize thread (save_finalize) running at step=5. W0419 04:40:06.472776 133259682899776 checkpoint_manager.py:1458] Waiting for previous save to complete took 48.606704 seconds. If this number is high, consider checkpointing less frequently. I0419 04:40:06.474292 133259682899776 checkpoint_manager.py:1518] [process=6] Saving checkpoint at step 9 I0419 04:40:06.477788 133259682899776 async_checkpointer.py:452] [process=6] Started async saving checkpoint to gs://lance-maxtext/linen_ckpt_xpk_feat_nnx_set_defaults_true_20260419_025828/linen_xpk_feat_nnx_set_defaults_true_20260419_025828_16_zarr3_true/checkpoints/9. I0419 04:40:07.851402 133259682899776 jax_array_handlers.py:358] Scheduling D2H of 39 prioritized jax.Array. I0419 04:40:07.851501 133259682899776 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 04:40:07.894980 133259682899776 base_pytree_checkpoint_handler.py:153] [process=6][thread=MainThread] Initiated "orbax.checkpoint._src.serialization.jax_array_handlers.ArrayHandler".serialize. Time taken: 0.044747s I0419 04:40:07.895147 133259682899776 base_pytree_checkpoint_handler.py:129] [process=6] /jax/checkpoint/write/blocking_gbytes_per_sec: 2.132 GiB/s (total gbytes: 1.5 GiB) (time elapsed: 0.7235231399536133 s) (per-host) I0419 04:40:07.895196 133259682899776 base_pytree_checkpoint_handler.py:737] [process=6][thread=MainThread] Initiated Pytree async_save. Time taken: 0.723581s (batch_requests_ready=0.676707s, total_serialization_initiated=0.046811s, others=0.000064s) I0419 04:40:07.895305 133259682899776 composite_checkpoint_handler.py:715] [process=6][thread=MainThread] Initiated CompositeCheckpointHandler.async_save. Time taken: 0.727749s (all_items=0.000012s, per_item={'items': '0.00001168'}, temp_paths=0.727738) I0419 04:40:07.896169 133136369301248 async_checkpointer.py:79] [process=6][thread=async_save] Background save thread started. I0419 04:40:07.896308 133259682899776 async_checkpointer.py:561] Finished blocking save. Time taken: 1.421947s. Continuing background save to gs://lance-maxtext/linen_ckpt_xpk_feat_nnx_set_defaults_true_20260419_025828/linen_xpk_feat_nnx_set_defaults_true_20260419_025828_16_zarr3_true/checkpoints/9. I0419 04:40:07.898300 133259682899776 checkpoint_manager.py:1566] [process=6][thread=MainThread][step=9] Starting CheckpointManager Save Finalize thread=save_finalize I0419 04:40:07.898492 133131545855744 async_checkpointer.py:265] [process=6][thread=save_finalize] Waiting for background save thread=async_save. I0419 04:40:07.898603 133259682899776 standard_logger.py:34] {'step': 9, 'event_type': 'save', 'directory': 'gs://lance-maxtext/linen_ckpt_xpk_feat_nnx_set_defaults_true_20260419_025828/linen_xpk_feat_nnx_set_defaults_true_20260419_025828_16_zarr3_true/checkpoints', 'reached_preemption': False, 'preemption_received_at': None, 'synchronous': False, 'wait_for_prev_start_time': 1776573557.866055, 'wait_for_prev_duration_secs': 48.60670351982117, 'time_between_consecutive_saves_sec': 21.385717153549194, 'checkpointer_blocking_start_time': 1776573606.47433, 'checkpointer_blocking_duration_secs': 1.4221453666687012, 'get_old_steps_start_time': 1776573607.896488, 'get_old_steps_duration_secs': 1.9550323486328125e-05, 'checkpoint_manager_blocking_start_time': 1776573557.8643851, 'checkpoint_manager_blocking_duration_secs': 50.03418850898743} I0419 04:40:07.898701 133259682899776 checkpointing.py:409] Started an asynchronous checkpoint save for step 9 I0419 04:40:07.898740 133259682899776 checkpoint_manager.py:2024] [process=6][thread=MainThread][step=9][wait_until_finished] Waiting for Save Finalize thread (save_finalize) to complete. I0419 04:40:13.859797 133136377693952 array_metadata_store.py:203] [process=6][thread=array_type_handler] Wrote 39 array_metadata.ArrayMetadata to gs://lance-maxtext/linen_ckpt_xpk_feat_nnx_set_defaults_true_20260419_025828/linen_xpk_feat_nnx_set_defaults_true_20260419_025828_16_zarr3_true/checkpoints/9/items/array_metadatas/process_6 I0419 04:40:50.749446 133136369301248 base_pytree_checkpoint_handler.py:129] [process=6] /jax/checkpoint/write/gbytes_per_sec: 36.248 MiB/s (total gbytes: 1.5 GiB) (time elapsed: 43.57778882980347 s) (per-host) I0419 04:40:50.749558 133136369301248 async_checkpointer.py:90] [process=6][thread=async_save] 3 Handler Commit operations completed. Time taken: 42.853175s. I0419 04:40:57.336490 133136369301248 async_checkpointer.py:144] [process=6][thread=async_save] Background save thread done. Time taken: 49.440091s. I0419 04:40:57.336718 133131545855744 async_checkpointer.py:273] [process=6][thread=save_finalize] Done with waiting for background save thread=async_save. I0419 04:40:57.336776 133131545855744 async_checkpointer.py:283] [process=6][thread=save_finalize] No errors found in background save thread=async_save. I0419 04:40:57.336825 133131545855744 checkpoint_manager.py:2133] [process=6][thread=save_finalize][step=9] CheckpointManager Save Finalize is syncing with other hosts... I0419 04:40:57.338714 133131545855744 checkpoint_manager.py:2142] [process=6][thread=save_finalize][step=9] CheckpointManager Save Finalize is done on all hosts. I0419 04:40:57.338879 133259682899776 checkpoint_manager.py:2036] [process=6][thread=MainThread][step=9][wait_until_finished] Done waiting for Save Finalize thread (save_finalize) running at step=9. I0419 04:40:57.339026 133259682899776 checkpoint_manager.py:2013] [process=6][thread=MainThread][wait_until_finished] No Save Finalize thread to wait for. Returning. I0419 04:40:57.339744 133259682899776 metric_logger.py:185] completed step: 9, seconds: 0.148, TFLOP/s/device: 91.889, Tokens/s/device: 13850.565, total_weights: 65536, loss: 7.181 Per train step: Total TFLOPs: 13.59 split as 93.93% learnable weight flops and 6.07% attention flops XPK End: Sun Apr 19 04:41:08 UTC 2026 EXIT_CODE=0