XPK Start: Fri Apr 17 06:00:30 UTC 2026 PyTorch was not found. Models won't be available and only tokenizers, configuration and file/data utilities can be used. 2026-04-17 06:00:53.896823: 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) I0417 06:00:54.076474 139090603628352 max_utils.py:273] Attempting to initialize the jax distributed system... I0417 06:01:03.117647 139090603628352 distributed.py:149] Starting JAX distributed service on [::]:8482 I0417 06:01:03.119999 139090603628352 distributed.py:172] Connecting to JAX distributed service on mt-15-ocdbt-true-sv-voeok-slice-job-0-0.mt-15-ocdbt-true-sv-voeok:8482 I0417 06:01:04.400866 139090603628352 max_utils.py:284] Jax distributed system initialized! I0417 06:01:10.508671 139090603628352 max_utils.py:800] System Information: Jax Version: 0.9.2 I0417 06:01:10.508775 139090603628352 max_utils.py:801] System Information: Jaxlib Version: 0.9.2 I0417 06:01:10.508816 139090603628352 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 I0417 06:01:10.508869 139090603628352 train_utils.py:377] WARNING: Sequence packing is essentially ignored for synthetic data. Please use a real dataset to use sequence packing. I0417 06:01:11.177375 139090603628352 maxtext_utils.py:1687] Num_devices: 32, shape (1, 1, 1, 32, 1, 1, 1, 1, 1, 1, 1, 1, 1) I0417 06:01:11.290368 139090603628352 checkpointing.py:688] Setting up checkpoint logger... I0417 06:01:11.290481 139090603628352 checkpointing.py:234] Creating checkpoint manager with ocdbt=True and zarr3=True I0417 06:01:11.290526 139090603628352 pytree_checkpoint_handler.py:592] save_device_host_concurrent_bytes=None I0417 06:01:11.290728 139090603628352 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 0x7e7feb118110>, enable_pinned_host_transfer=False, save_concurrent_bytes: 96000000000 (89.4 GiB), restore_concurrent_bytes: 96000000000 (89.4 GiB) I0417 06:01:14.592310 139090603628352 checkpointing.py:266] Enabling policy for fixed interval checkpointing. I0417 06:01:14.592540 139090603628352 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 0x7e6b5c490230>}, handler_registry=None I0417 06:01:14.592782 139090603628352 composite_checkpoint_handler.py:237] Deferred registration for item: "items". Adding handler `<orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeCheckpointHandler object at 0x7e6b5c490230>` 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`. I0417 06:01:14.592830 139090603628352 composite_checkpoint_handler.py:237] Deferred registration for item: "metrics". Adding handler `<orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonCheckpointHandler object at 0x7e6a806266c0>` 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`. I0417 06:01:14.592878 139090603628352 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 0x7e6b5c490230>, ('items', <class 'orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeRestoreArgs'>): <orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeCheckpointHandler object at 0x7e6b5c490230>, ('metrics', <class 'orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonSaveArgs'>): <orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonCheckpointHandler object at 0x7e6a806266c0>, ('metrics', <class 'orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonRestoreArgs'>): <orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonCheckpointHandler object at 0x7e6a806266c0>}). I0417 06:01:14.593197 139090603628352 abstract_checkpointer.py:35] orbax-checkpoint version: 0.11.34 I0417 06:01:14.593269 139090603628352 async_checkpointer.py:192] [process=0][thread=MainThread] Using barrier_sync_fn: <function get_barrier_sync_fn.<locals>._fn at 0x7e6b5c1ca020> timeout: 1200 secs and primary_host=0 for async checkpoint writes I0417 06:01:16.123827 139090603628352 checkpoint_manager.py:1812] Found 0 checkpoint steps in gs://lance-maxtext/nnx_ckpt_xpk_feat_nnx_post_train_fixes_20260417_023710/nnx_xpk_feat_nnx_post_train_fixes_20260417_023710_15_ocdbt_true/checkpoints I0417 06:01:16.609449 139090603628352 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/nnx_ckpt_xpk_feat_nnx_post_train_fixes_20260417_023710/nnx_xpk_feat_nnx_post_train_fixes_20260417_023710_15_ocdbt_true/checkpoints: <orbax.checkpoint.checkpoint_manager.CheckpointManager object at 0x7e6a80677530> I0417 06:01:16.609621 139090603628352 checkpointing.py:302] Checkpoint manager created! I0417 06:01:17.575878 139090603628352 checkpointing.py:578] checkpoint manager exists so trying to load this run's existing checkpoint I0417 06:01:17.575996 139090603628352 checkpointing.py:676] No existing checkpoints found, not restoring checkpoint. fsdp: 32 I0417 06:01:18.986624 139090603628352 maxtext_utils.py:1805] decoder/decoder_norm/scale/value Shape: float32[2048] Physical: (None,) I0417 06:01:18.986737 139090603628352 maxtext_utils.py:1805] decoder/layers/mlp/wi_0/kernel/value Shape: float32[2048,16,7168] Physical: ('fsdp', None, None) I0417 06:01:18.986787 139090603628352 maxtext_utils.py:1805] decoder/layers/mlp/wi_1/kernel/value Shape: float32[2048,16,7168] Physical: ('fsdp', None, None) I0417 06:01:18.986829 139090603628352 maxtext_utils.py:1805] decoder/layers/mlp/wo/kernel/value Shape: float32[7168,16,2048] Physical: (None, None, 'fsdp') I0417 06:01:18.986882 139090603628352 maxtext_utils.py:1805] decoder/layers/post_self_attention_layer_norm/scale/value Shape: float32[2048,16] Physical: (None, None) I0417 06:01:18.986919 139090603628352 maxtext_utils.py:1805] decoder/layers/pre_self_attention_layer_norm/scale/value Shape: float32[2048,16] Physical: (None, None) I0417 06:01:18.986955 139090603628352 maxtext_utils.py:1805] decoder/layers/self_attention/key/kernel/value Shape: float32[2048,16,16,128] Physical: ('fsdp', None, None, None) I0417 06:01:18.986989 139090603628352 maxtext_utils.py:1805] decoder/layers/self_attention/out/kernel/value Shape: float32[16,16,128,2048] Physical: (None, None, None, 'fsdp') I0417 06:01:18.987024 139090603628352 maxtext_utils.py:1805] decoder/layers/self_attention/query/kernel/value Shape: float32[2048,16,16,128] Physical: ('fsdp', None, None, None) I0417 06:01:18.987056 139090603628352 maxtext_utils.py:1805] decoder/layers/self_attention/value/kernel/value Shape: float32[2048,16,16,128] Physical: ('fsdp', None, None, None) I0417 06:01:18.987088 139090603628352 maxtext_utils.py:1805] decoder/logits_dense/kernel/value Shape: float32[2048,32000] Physical: ('fsdp', None) I0417 06:01:18.987119 139090603628352 maxtext_utils.py:1805] token_embedder/embedding/value Shape: float32[32000,2048] Physical: (None, 'fsdp') I0417 06:01:19.389545 139090603628352 nnx_decoders.py:465] nnx_decoders/carry Logical: bfloat16[32,2048,2048]...................................... ('activation_batch', 'activation_norm_length', 'activation_embed'). I0417 06:01:19.389634 139090603628352 nnx_decoders.py:465] nnx_decoders/carry Physical: bfloat16[32,2048,2048]...................................... ('fsdp', None, None). I0417 06:01:19.394885 139090603628352 nnx_decoders.py:465] Unknown Logical: bfloat16[32,2048,2048]...................................... ('activation_batch', 'activation_norm_length', 'activation_embed'). I0417 06:01:19.394940 139090603628352 nnx_decoders.py:465] Unknown Physical: bfloat16[32,2048,2048]...................................... ('fsdp', None, None). I0417 06:01:19.411318 139090603628352 attentions.py:1088] attentions/inputs_q Logical: bfloat16[32,2048,2048]...................................... ('activation_batch', 'activation_attn_length', 'activation_attn_embed'). I0417 06:01:19.411379 139090603628352 attentions.py:1088] attentions/inputs_q Physical: bfloat16[32,2048,2048]...................................... ('fsdp', None, None). I0417 06:01:19.427092 139090603628352 attentions.py:1089] attentions/inputs_kv Logical: bfloat16[32,2048,2048]...................................... ('activation_batch', 'activation_attn_length', 'activation_attn_embed'). I0417 06:01:19.427147 139090603628352 attentions.py:1089] attentions/inputs_kv Physical: bfloat16[32,2048,2048]...................................... ('fsdp', None, None). I0417 06:01:19.451021 139090603628352 attentions.py:1154] attentions/query Logical: bfloat16[32,2048,16,128].................................... ('activation_kv_batch', 'activation_attn_length', 'activation_kv_heads', 'activation_kv_head_dim'). I0417 06:01:19.451083 139090603628352 attentions.py:1154] attentions/query Physical: bfloat16[32,2048,16,128].................................... ('fsdp', None, None, None). I0417 06:01:19.466917 139090603628352 attentions.py:1155] attentions/key Logical: bfloat16[32,2048,16,128].................................... ('activation_kv_batch', 'activation_attn_length', 'activation_kv_heads', 'activation_kv_head_dim'). I0417 06:01:19.466977 139090603628352 attentions.py:1155] attentions/key Physical: bfloat16[32,2048,16,128].................................... ('fsdp', None, None, None). I0417 06:01:19.482809 139090603628352 attentions.py:1156] attentions/value Logical: bfloat16[32,2048,16,128].................................... ('activation_kv_batch', 'activation_attn_length', 'activation_kv_heads', 'activation_kv_head_dim'). I0417 06:01:19.482881 139090603628352 attentions.py:1156] attentions/value Physical: bfloat16[32,2048,16,128].................................... ('fsdp', None, None, None). I0417 06:01:19.512288 139090603628352 attentions.py:1197] attentions/out Logical: bfloat16[32,2048,16,128].................................... ('activation_batch', 'activation_attn_length', 'activation_heads', 'activation_kv'). I0417 06:01:19.512356 139090603628352 attentions.py:1197] attentions/out Physical: bfloat16[32,2048,16,128].................................... ('fsdp', None, None, None). I0417 06:01:19.532159 139090603628352 linears.py:525] linears/x Logical: bfloat16[32,2048,7168]...................................... ('activation_batch', 'activation_length', 'activation_mlp'). I0417 06:01:19.532222 139090603628352 linears.py:525] linears/x Physical: bfloat16[32,2048,7168]...................................... ('fsdp', None, None). I0417 06:01:30.293134 139090603628352 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. I0417 06:01:30.301975 139090603628352 metric_logger.py:289] number parameters: 1.104 billion I0417 06:01:41.518281 139090603628352 checkpointing.py:794] Waiting for step 0 to finish before checkpoint... I0417 06:01:41.670287 139090603628352 checkpointing.py:798] Waited 0.15199017524719238 seconds for step 0 to finish before starting checkpointing. I0417 06:01:41.672733 139090603628352 checkpoint_manager.py:2009] [process=0][thread=MainThread][wait_until_finished] No Save Finalize thread to wait for. Returning. I0417 06:01:41.674272 139090603628352 checkpoint_manager.py:1512] [process=0] Saving checkpoint at step 0 I0417 06:01:41.675750 139090603628352 event_tracking.py:70] [process=0] [async] Started save checkpoint @ gs://lance-maxtext/nnx_ckpt_xpk_feat_nnx_post_train_fixes_20260417_023710/nnx_xpk_feat_nnx_post_train_fixes_20260417_023710_15_ocdbt_true/checkpoints/0. I0417 06:01:42.403687 139090603628352 signaling_client.py:364] Using JaxDistributedSignalingClient I0417 06:01:42.423173 139090603628352 jax_array_handlers.py:360] Scheduling D2H of 69 prioritized jax.Array. I0417 06:01:42.423243 139090603628352 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 I0417 06:01:42.663558 138961242547968 atomicity.py:140] Creating tmp directory gs://lance-maxtext/nnx_ckpt_xpk_feat_nnx_post_train_fixes_20260417_023710/nnx_xpk_feat_nnx_post_train_fixes_20260417_023710_15_ocdbt_true/checkpoints/0 I0417 06:01:42.697564 139090603628352 base_pytree_checkpoint_handler.py:154] [process=0][thread=MainThread] Initiated "orbax.checkpoint._src.serialization.jax_array_handlers.ArrayHandler".serialize. Time taken: 0.275817s I0417 06:01:42.698374 139090603628352 base_pytree_checkpoint_handler.py:130] [process=0] /jax/orbax/write/blocking_gbytes_per_sec: 5.378 GiB/s (total gbytes: 1.5 GiB) (time elapsed: 0.28681206703186035 s) (per-host) I0417 06:01:42.698454 139090603628352 base_pytree_checkpoint_handler.py:768] [process=0][thread=MainThread] Initiated Pytree async_save. Time taken: 0.286912s (batch_requests_ready=0.003126s, total_serialization_initiated=0.283045s, others=0.000741s) I0417 06:01:42.698559 139090603628352 composite_checkpoint_handler.py:715] [process=0][thread=MainThread] Initiated CompositeCheckpointHandler.async_save. Time taken: 0.290941s (all_items=0.000018s, per_item={'items': '0.00001812'}, temp_paths=0.290922) I0417 06:01:42.699409 139090603628352 event_tracking.py:125] [process=0] [async] Finished blocking save in 1.03 seconds. Continuing save @ gs://lance-maxtext/nnx_ckpt_xpk_feat_nnx_post_train_fixes_20260417_023710/nnx_xpk_feat_nnx_post_train_fixes_20260417_023710_15_ocdbt_true/checkpoints/0. I0417 06:01:42.699739 138961250940672 async_checkpointer.py:76] [process=0][thread=async_save] Background save thread started. Deadline for this save operation is 2026-04-17 06:21:42.699702 I0417 06:01:42.706996 139090603628352 checkpoint_manager.py:1560] [process=0][thread=MainThread][step=0] Starting CheckpointManager Save Finalize thread=save_finalize I0417 06:01:42.707271 138959631935232 async_checkpointer.py:280] [process=0][thread=save_finalize] Waiting for background save thread=async_save. I0417 06:01:42.707426 139090603628352 standard_logger.py:34] {'step': 0, 'event_type': 'save', 'directory': 'gs://lance-maxtext/nnx_ckpt_xpk_feat_nnx_post_train_fixes_20260417_023710/nnx_xpk_feat_nnx_post_train_fixes_20260417_023710_15_ocdbt_true/checkpoints', 'reached_preemption': False, 'preemption_received_at': None, 'synchronous': False, 'wait_for_prev_start_time': 1776405701.672715, 'wait_for_prev_duration_secs': 5.841255187988281e-05, 'time_between_consecutive_saves_sec': None, 'checkpointer_blocking_start_time': 1776405701.6743097, 'checkpointer_blocking_duration_secs': 1.025599718093872, 'get_old_steps_start_time': 1776405702.6999407, 'get_old_steps_duration_secs': 2.9325485229492188e-05, 'checkpoint_manager_blocking_start_time': 1776405701.670666, 'checkpoint_manager_blocking_duration_secs': 1.0367207527160645} I0417 06:01:42.707532 139090603628352 checkpointing.py:409] Started an asynchronous checkpoint save for step 0 I0417 06:01:42.707602 139090603628352 max_utils.py:750] Memstats: After params initialized: I0417 06:01:42.707653 139090603628352 max_utils.py:756] Using (GB) 0.43 / 31.25 (1.376000%) on TPU_0(process=0,(0,0,0,0)) I0417 06:01:42.707685 139090603628352 max_utils.py:756] Using (GB) 0.43 / 31.25 (1.376000%) on TPU_1(process=0,(1,0,0,0)) I0417 06:01:42.707714 139090603628352 max_utils.py:756] Using (GB) 0.43 / 31.25 (1.376000%) on TPU_4(process=0,(0,1,0,0)) I0417 06:01:42.707736 139090603628352 max_utils.py:756] Using (GB) 0.43 / 31.25 (1.376000%) on TPU_5(process=0,(1,1,0,0)) I0417 06:01:43.053121 139090603628352 metric_logger.py:185] completed step: 0, seconds: 10.913, TFLOP/s/device: 1.245, Tokens/s/device: 187.666, total_weights: 65536, loss: 10.874 I0417 06:01:43.056455 139090603628352 metric_logger.py:269] To see full metrics 'tensorboard --logdir=gs://lance-maxtext/nnx_ckpt_xpk_feat_nnx_post_train_fixes_20260417_023710/nnx_xpk_feat_nnx_post_train_fixes_20260417_023710_15_ocdbt_true/tensorboard/' I0417 06:01:43.478778 139090603628352 metric_logger.py:185] completed step: 1, seconds: 1.531, TFLOP/s/device: 8.872, Tokens/s/device: 1337.324, total_weights: 65536, loss: 10.874 I0417 06:01:43.620440 139090603628352 metric_logger.py:185] completed step: 2, seconds: 0.427, TFLOP/s/device: 31.846, Tokens/s/device: 4800.165, total_weights: 65536, loss: 10.560 I0417 06:01:43.767907 139090603628352 metric_logger.py:185] completed step: 3, seconds: 0.012, TFLOP/s/device: 1131.695, Tokens/s/device: 170581.376, total_weights: 65536, loss: 9.979 I0417 06:01:43.779918 139090603628352 checkpointing.py:794] Waiting for step 5 to finish before checkpoint... I0417 06:01:44.063272 139090603628352 checkpointing.py:798] Waited 0.2833254337310791 seconds for step 5 to finish before starting checkpointing. I0417 06:01:44.066097 139090603628352 checkpoint_manager.py:2020] [process=0][thread=MainThread][step=0][wait_until_finished] Waiting for Save Finalize thread (save_finalize) to complete. I0417 06:01:44.546665 138961225762560 atomicity.py:140] Creating tmp directory gs://lance-maxtext/nnx_ckpt_xpk_feat_nnx_post_train_fixes_20260417_023710/nnx_xpk_feat_nnx_post_train_fixes_20260417_023710_15_ocdbt_true/checkpoints/0/items I0417 06:01:44.697221 138956362237696 checkpoint.py:188] Wrote Metadata={'item_handlers': None, 'metrics': {}, 'performance_metrics': {}, 'init_timestamp_nsecs': 1776405704290629295, 'commit_timestamp_nsecs': None, 'custom_metadata': {}}, json={"item_handlers": null, "metrics": {}, "performance_metrics": {}, "init_timestamp_nsecs": 1776405704290629295, "commit_timestamp_nsecs": null, "custom_metadata": {}} to gs://lance-maxtext/nnx_ckpt_xpk_feat_nnx_post_train_fixes_20260417_023710/nnx_xpk_feat_nnx_post_train_fixes_20260417_023710_15_ocdbt_true/checkpoints/0/_CHECKPOINT_METADATA I0417 06:01:45.686610 2806 google_auth_provider.cc:181] Running on GCE, using service account 562977990677-compute@developer.gserviceaccount.com I0417 06:01:47.483072 138960655574784 array_metadata_store.py:203] [process=0][thread=array_type_handler] Wrote 69 array_metadata.ArrayMetadata to gs://lance-maxtext/nnx_ckpt_xpk_feat_nnx_post_train_fixes_20260417_023710/nnx_xpk_feat_nnx_post_train_fixes_20260417_023710_15_ocdbt_true/checkpoints/0/items/array_metadatas/process_0 I0417 06:02:22.487445 138959640327936 base_pytree_checkpoint_handler.py:1282] [process=0][thread=write_metadata_after_commits] Commit + Array metadata written. Time taken: 36.809051s (commit=35.338663s, array_metadata_write=1.470388s) I0417 06:02:22.488952 138961250940672 base_pytree_checkpoint_handler.py:130] [process=0] /jax/orbax/write/gbytes_per_sec: 39.414 MiB/s (total gbytes: 1.5 GiB) (time elapsed: 40.07737731933594 s) (per-host) I0417 06:02:22.489056 138961250940672 async_checkpointer.py:90] [process=0][thread=async_save] 3 Handler Commit operations completed. Time taken: 39.789185s. I0417 06:02:23.635725 138961250940672 checkpoint.py:228] Read Metadata={'item_handlers': None, 'metrics': {}, 'performance_metrics': {}, 'init_timestamp_nsecs': 1776405704290629295, 'commit_timestamp_nsecs': None, 'custom_metadata': {}} from gs://lance-maxtext/nnx_ckpt_xpk_feat_nnx_post_train_fixes_20260417_023710/nnx_xpk_feat_nnx_post_train_fixes_20260417_023710_15_ocdbt_true/checkpoints/0/_CHECKPOINT_METADATA I0417 06:02:24.747141 138961250940672 array_metadata_store.py:411] [process=0][thread=async_save] Validated ArrayMetadata from all 8 hosts. Time taken: 0.000755s. I0417 06:02:24.873389 138956362237696 checkpoint.py:247] Updated Metadata={'item_handlers': {'items': 'orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeCheckpointHandler'}, 'metrics': {}, 'performance_metrics': {}, 'init_timestamp_nsecs': 1776405704290629295, 'commit_timestamp_nsecs': None, 'custom_metadata': {}} to gs://lance-maxtext/nnx_ckpt_xpk_feat_nnx_post_train_fixes_20260417_023710/nnx_xpk_feat_nnx_post_train_fixes_20260417_023710_15_ocdbt_true/checkpoints/0/_CHECKPOINT_METADATA I0417 06:02:25.011289 138961250940672 ocdbt_utils.py:49] Param validation support for Zarr3 will be added later (b/362328389). I0417 06:02:25.948072 138961250940672 base_pytree_checkpoint_handler.py:1406] [process=0][thread=async_save] Pytree save finalize (merge_ocdbt + ArrayMetadata validation) completed. Time taken: 2.182252s. use_zarr3=True, enable_post_merge_validation=True, directory=gs://lance-maxtext/nnx_ckpt_xpk_feat_nnx_post_train_fixes_20260417_023710/nnx_xpk_feat_nnx_post_train_fixes_20260417_023710_15_ocdbt_true/checkpoints/0/items I0417 06:02:25.950385 138961250940672 atomicity.py:666] Finalizing gs://lance-maxtext/nnx_ckpt_xpk_feat_nnx_post_train_fixes_20260417_023710/nnx_xpk_feat_nnx_post_train_fixes_20260417_023710_15_ocdbt_true/checkpoints/0/items I0417 06:02:26.574309 138961250940672 atomicity.py:666] Finalizing gs://lance-maxtext/nnx_ckpt_xpk_feat_nnx_post_train_fixes_20260417_023710/nnx_xpk_feat_nnx_post_train_fixes_20260417_023710_15_ocdbt_true/checkpoints/0 I0417 06:02:28.395894 138961250940672 atomicity.py:847] [process=0][thread=async_save] Finished saving checkpoint (finalized tmp dir) to `gs://lance-maxtext/nnx_ckpt_xpk_feat_nnx_post_train_fixes_20260417_023710/nnx_xpk_feat_nnx_post_train_fixes_20260417_023710_15_ocdbt_true/checkpoints/0`. I0417 06:02:28.397065 138961250940672 event_tracking.py:138] [process=0] [async] Finished save (blocking + background) in 46.72 seconds @ gs://lance-maxtext/nnx_ckpt_xpk_feat_nnx_post_train_fixes_20260417_023710/nnx_xpk_feat_nnx_post_train_fixes_20260417_023710_15_ocdbt_true/checkpoints/0 I0417 06:02:28.398610 138961250940672 async_checkpointer.py:160] [process=0][thread=async_save] Background save thread done. Time taken: 45.698735s. I0417 06:02:28.398892 138959631935232 async_checkpointer.py:288] [process=0][thread=save_finalize] Done with waiting for background save thread=async_save. I0417 06:02:28.399008 138959631935232 async_checkpointer.py:298] [process=0][thread=save_finalize] No errors found in background save thread=async_save. I0417 06:02:28.399136 138959631935232 checkpoint_manager.py:2137] [process=0][thread=save_finalize][step=0] CheckpointManager Save Finalize is syncing with other hosts... I0417 06:02:28.400770 138959631935232 checkpoint_manager.py:2146] [process=0][thread=save_finalize][step=0] CheckpointManager Save Finalize is done on all hosts. I0417 06:02:28.401099 139090603628352 checkpoint_manager.py:2032] [process=0][thread=MainThread][step=0][wait_until_finished] Done waiting for Save Finalize thread (save_finalize) running at step=0. W0417 06:02:28.401286 139090603628352 checkpoint_manager.py:1452] Waiting for previous save to complete took 44.335192 seconds. If this number is high, consider checkpointing less frequently. I0417 06:02:28.403006 139090603628352 checkpoint_manager.py:1512] [process=0] Saving checkpoint at step 5 I0417 06:02:28.405452 139090603628352 event_tracking.py:70] [process=0] [async] Started save checkpoint @ gs://lance-maxtext/nnx_ckpt_xpk_feat_nnx_post_train_fixes_20260417_023710/nnx_xpk_feat_nnx_post_train_fixes_20260417_023710_15_ocdbt_true/checkpoints/5. I0417 06:02:29.135264 139090603628352 jax_array_handlers.py:360] Scheduling D2H of 69 prioritized jax.Array. I0417 06:02:29.135397 139090603628352 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 I0417 06:02:29.174408 139090603628352 base_pytree_checkpoint_handler.py:154] [process=0][thread=MainThread] Initiated "orbax.checkpoint._src.serialization.jax_array_handlers.ArrayHandler".serialize. Time taken: 0.042147s I0417 06:02:29.175463 139090603628352 base_pytree_checkpoint_handler.py:130] [process=0] /jax/orbax/write/blocking_gbytes_per_sec: 25.074 GiB/s (total gbytes: 1.5 GiB) (time elapsed: 0.06152153015136719 s) (per-host) I0417 06:02:29.175571 139090603628352 base_pytree_checkpoint_handler.py:768] [process=0][thread=MainThread] Initiated Pytree async_save. Time taken: 0.061653s (batch_requests_ready=0.006238s, total_serialization_initiated=0.054463s, others=0.000952s) I0417 06:02:29.175709 139090603628352 composite_checkpoint_handler.py:715] [process=0][thread=MainThread] Initiated CompositeCheckpointHandler.async_save. Time taken: 0.066627s (all_items=0.000025s, per_item={'items': '0.00002503'}, temp_paths=0.066602) I0417 06:02:29.176651 139090603628352 event_tracking.py:125] [process=0] [async] Finished blocking save in 0.77 seconds. Continuing save @ gs://lance-maxtext/nnx_ckpt_xpk_feat_nnx_post_train_fixes_20260417_023710/nnx_xpk_feat_nnx_post_train_fixes_20260417_023710_15_ocdbt_true/checkpoints/5. I0417 06:02:29.177181 138959631935232 async_checkpointer.py:76] [process=0][thread=async_save] Background save thread started. Deadline for this save operation is 2026-04-17 06:22:29.177141 I0417 06:02:29.179630 139090603628352 checkpoint_manager.py:1560] [process=0][thread=MainThread][step=5] Starting CheckpointManager Save Finalize thread=save_finalize I0417 06:02:29.180048 138959480932096 async_checkpointer.py:280] [process=0][thread=save_finalize] Waiting for background save thread=async_save. I0417 06:02:29.180286 139090603628352 standard_logger.py:34] {'step': 5, 'event_type': 'save', 'directory': 'gs://lance-maxtext/nnx_ckpt_xpk_feat_nnx_post_train_fixes_20260417_023710/nnx_xpk_feat_nnx_post_train_fixes_20260417_023710_15_ocdbt_true/checkpoints', 'reached_preemption': False, 'preemption_received_at': None, 'synchronous': False, 'wait_for_prev_start_time': 1776405704.0660515, 'wait_for_prev_duration_secs': 44.33519172668457, 'time_between_consecutive_saves_sec': None, 'checkpointer_blocking_start_time': 1776405748.403073, 'checkpointer_blocking_duration_secs': 0.7742972373962402, 'get_old_steps_start_time': 1776405749.1773996, 'get_old_steps_duration_secs': 4.172325134277344e-05, 'checkpoint_manager_blocking_start_time': 1776405704.063754, 'checkpoint_manager_blocking_duration_secs': 45.11647915840149} I0417 06:02:29.180431 139090603628352 checkpointing.py:409] Started an asynchronous checkpoint save for step 5 I0417 06:02:29.349340 138961250940672 atomicity.py:140] Creating tmp directory gs://lance-maxtext/nnx_ckpt_xpk_feat_nnx_post_train_fixes_20260417_023710/nnx_xpk_feat_nnx_post_train_fixes_20260417_023710_15_ocdbt_true/checkpoints/5 I0417 06:02:31.246361 138961225762560 atomicity.py:140] Creating tmp directory gs://lance-maxtext/nnx_ckpt_xpk_feat_nnx_post_train_fixes_20260417_023710/nnx_xpk_feat_nnx_post_train_fixes_20260417_023710_15_ocdbt_true/checkpoints/5/items I0417 06:02:34.582157 138959506110208 array_metadata_store.py:203] [process=0][thread=array_type_handler] Wrote 69 array_metadata.ArrayMetadata to gs://lance-maxtext/nnx_ckpt_xpk_feat_nnx_post_train_fixes_20260417_023710/nnx_xpk_feat_nnx_post_train_fixes_20260417_023710_15_ocdbt_true/checkpoints/5/items/array_metadatas/process_0 I0417 06:02:52.378972 139090603628352 metric_logger.py:185] completed step: 4, seconds: 0.143, TFLOP/s/device: 94.998, Tokens/s/device: 14319.075, total_weights: 65536, loss: 9.457 I0417 06:02:52.389579 139090603628352 metric_logger.py:185] completed step: 5, seconds: 0.147, TFLOP/s/device: 92.492, Tokens/s/device: 13941.457, total_weights: 65536, loss: 8.955 I0417 06:02:52.533455 139090603628352 metric_logger.py:185] completed step: 6, seconds: 68.609, TFLOP/s/device: 0.198, Tokens/s/device: 29.850, total_weights: 65536, loss: 8.465 I0417 06:02:52.680743 139090603628352 metric_logger.py:185] completed step: 7, seconds: 0.008, TFLOP/s/device: 1615.208, Tokens/s/device: 243461.721, total_weights: 65536, loss: 7.998 I0417 06:02:52.828637 139090603628352 metric_logger.py:185] completed step: 8, seconds: 0.145, TFLOP/s/device: 93.531, Tokens/s/device: 14097.984, total_weights: 65536, loss: 7.567 I0417 06:02:52.977178 139090603628352 checkpointing.py:794] Waiting for step 9 to finish before checkpoint... I0417 06:02:52.977955 139090603628352 checkpointing.py:798] Waited 0.0007891654968261719 seconds for step 9 to finish before starting checkpointing. I0417 06:02:52.979969 139090603628352 checkpoint_manager.py:2020] [process=0][thread=MainThread][step=5][wait_until_finished] Waiting for Save Finalize thread (save_finalize) to complete. I0417 06:03:14.335078 138959489324800 base_pytree_checkpoint_handler.py:1282] [process=0][thread=write_metadata_after_commits] Commit + Array metadata written. Time taken: 41.542740s (commit=40.315405s, array_metadata_write=1.227335s) I0417 06:03:14.336265 138959631935232 base_pytree_checkpoint_handler.py:130] [process=0] /jax/orbax/write/gbytes_per_sec: 34.930 MiB/s (total gbytes: 1.5 GiB) (time elapsed: 45.22230577468872 s) (per-host) I0417 06:03:14.336377 138959631935232 async_checkpointer.py:90] [process=0][thread=async_save] 3 Handler Commit operations completed. Time taken: 45.159048s. I0417 06:03:16.031352 138959631935232 array_metadata_store.py:411] [process=0][thread=async_save] Validated ArrayMetadata from all 8 hosts. Time taken: 0.000413s. I0417 06:03:16.247659 138959631935232 ocdbt_utils.py:49] Param validation support for Zarr3 will be added later (b/362328389). I0417 06:03:17.191490 138959631935232 base_pytree_checkpoint_handler.py:1406] [process=0][thread=async_save] Pytree save finalize (merge_ocdbt + ArrayMetadata validation) completed. Time taken: 2.048374s. use_zarr3=True, enable_post_merge_validation=True, directory=gs://lance-maxtext/nnx_ckpt_xpk_feat_nnx_post_train_fixes_20260417_023710/nnx_xpk_feat_nnx_post_train_fixes_20260417_023710_15_ocdbt_true/checkpoints/5/items I0417 06:03:17.193149 138959631935232 atomicity.py:666] Finalizing gs://lance-maxtext/nnx_ckpt_xpk_feat_nnx_post_train_fixes_20260417_023710/nnx_xpk_feat_nnx_post_train_fixes_20260417_023710_15_ocdbt_true/checkpoints/5/items I0417 06:03:17.840817 138959631935232 atomicity.py:666] Finalizing gs://lance-maxtext/nnx_ckpt_xpk_feat_nnx_post_train_fixes_20260417_023710/nnx_xpk_feat_nnx_post_train_fixes_20260417_023710_15_ocdbt_true/checkpoints/5 I0417 06:03:20.126954 138959631935232 atomicity.py:847] [process=0][thread=async_save] Finished saving checkpoint (finalized tmp dir) to `gs://lance-maxtext/nnx_ckpt_xpk_feat_nnx_post_train_fixes_20260417_023710/nnx_xpk_feat_nnx_post_train_fixes_20260417_023710_15_ocdbt_true/checkpoints/5`. I0417 06:03:20.127805 138959631935232 event_tracking.py:138] [process=0] [async] Finished save (blocking + background) in 51.72 seconds @ gs://lance-maxtext/nnx_ckpt_xpk_feat_nnx_post_train_fixes_20260417_023710/nnx_xpk_feat_nnx_post_train_fixes_20260417_023710_15_ocdbt_true/checkpoints/5 I0417 06:03:20.129306 138959631935232 async_checkpointer.py:160] [process=0][thread=async_save] Background save thread done. Time taken: 50.951976s. I0417 06:03:20.129512 138959480932096 async_checkpointer.py:288] [process=0][thread=save_finalize] Done with waiting for background save thread=async_save. I0417 06:03:20.129623 138959480932096 async_checkpointer.py:298] [process=0][thread=save_finalize] No errors found in background save thread=async_save. I0417 06:03:20.129687 138959480932096 checkpoint_manager.py:2137] [process=0][thread=save_finalize][step=5] CheckpointManager Save Finalize is syncing with other hosts... I0417 06:03:20.131677 138959480932096 checkpoint_manager.py:2146] [process=0][thread=save_finalize][step=5] CheckpointManager Save Finalize is done on all hosts. I0417 06:03:20.131916 139090603628352 checkpoint_manager.py:2032] [process=0][thread=MainThread][step=5][wait_until_finished] Done waiting for Save Finalize thread (save_finalize) running at step=5. W0417 06:03:20.132045 139090603628352 checkpoint_manager.py:1452] Waiting for previous save to complete took 27.152080 seconds. If this number is high, consider checkpointing less frequently. I0417 06:03:20.133457 139090603628352 checkpoint_manager.py:1512] [process=0] Saving checkpoint at step 9 I0417 06:03:20.135604 139090603628352 event_tracking.py:70] [process=0] [async] Started save checkpoint @ gs://lance-maxtext/nnx_ckpt_xpk_feat_nnx_post_train_fixes_20260417_023710/nnx_xpk_feat_nnx_post_train_fixes_20260417_023710_15_ocdbt_true/checkpoints/9. I0417 06:03:20.861583 139090603628352 jax_array_handlers.py:360] Scheduling D2H of 69 prioritized jax.Array. I0417 06:03:20.861680 139090603628352 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 I0417 06:03:20.889492 139090603628352 base_pytree_checkpoint_handler.py:154] [process=0][thread=MainThread] Initiated "orbax.checkpoint._src.serialization.jax_array_handlers.ArrayHandler".serialize. Time taken: 0.029406s I0417 06:03:20.890320 139090603628352 base_pytree_checkpoint_handler.py:130] [process=0] /jax/orbax/write/blocking_gbytes_per_sec: 40.363 GiB/s (total gbytes: 1.5 GiB) (time elapsed: 0.03821730613708496 s) (per-host) I0417 06:03:20.890393 139090603628352 base_pytree_checkpoint_handler.py:768] [process=0][thread=MainThread] Initiated Pytree async_save. Time taken: 0.038310s (batch_requests_ready=0.002752s, total_serialization_initiated=0.034816s, others=0.000742s) I0417 06:03:20.890480 139090603628352 composite_checkpoint_handler.py:715] [process=0][thread=MainThread] Initiated CompositeCheckpointHandler.async_save. Time taken: 0.042462s (all_items=0.000016s, per_item={'items': '0.00001621'}, temp_paths=0.042446) I0417 06:03:20.891212 139090603628352 event_tracking.py:125] [process=0] [async] Finished blocking save in 0.76 seconds. Continuing save @ gs://lance-maxtext/nnx_ckpt_xpk_feat_nnx_post_train_fixes_20260417_023710/nnx_xpk_feat_nnx_post_train_fixes_20260417_023710_15_ocdbt_true/checkpoints/9. I0417 06:03:20.891546 138959480932096 async_checkpointer.py:76] [process=0][thread=async_save] Background save thread started. Deadline for this save operation is 2026-04-17 06:23:20.891510 I0417 06:03:20.898565 139090603628352 checkpoint_manager.py:1560] [process=0][thread=MainThread][step=9] Starting CheckpointManager Save Finalize thread=save_finalize I0417 06:03:20.898810 138960697284352 async_checkpointer.py:280] [process=0][thread=save_finalize] Waiting for background save thread=async_save. I0417 06:03:20.898968 139090603628352 standard_logger.py:34] {'step': 9, 'event_type': 'save', 'directory': 'gs://lance-maxtext/nnx_ckpt_xpk_feat_nnx_post_train_fixes_20260417_023710/nnx_xpk_feat_nnx_post_train_fixes_20260417_023710_15_ocdbt_true/checkpoints', 'reached_preemption': False, 'preemption_received_at': None, 'synchronous': False, 'wait_for_prev_start_time': 1776405772.9799354, 'wait_for_prev_duration_secs': 27.152080059051514, 'time_between_consecutive_saves_sec': 24.57908272743225, 'checkpointer_blocking_start_time': 1776405800.1334946, 'checkpointer_blocking_duration_secs': 0.7581865787506104, 'get_old_steps_start_time': 1776405800.8917003, 'get_old_steps_duration_secs': 2.7418136596679688e-05, 'checkpoint_manager_blocking_start_time': 1776405772.978187, 'checkpoint_manager_blocking_duration_secs': 27.920748710632324} I0417 06:03:20.899091 139090603628352 checkpointing.py:409] Started an asynchronous checkpoint save for step 9 I0417 06:03:20.899136 139090603628352 checkpoint_manager.py:2020] [process=0][thread=MainThread][step=9][wait_until_finished] Waiting for Save Finalize thread (save_finalize) to complete. I0417 06:03:21.099151 138959631935232 atomicity.py:140] Creating tmp directory gs://lance-maxtext/nnx_ckpt_xpk_feat_nnx_post_train_fixes_20260417_023710/nnx_xpk_feat_nnx_post_train_fixes_20260417_023710_15_ocdbt_true/checkpoints/9 I0417 06:03:22.830550 138959464146688 atomicity.py:140] Creating tmp directory gs://lance-maxtext/nnx_ckpt_xpk_feat_nnx_post_train_fixes_20260417_023710/nnx_xpk_feat_nnx_post_train_fixes_20260417_023710_15_ocdbt_true/checkpoints/9/items I0417 06:03:25.732564 138961250940672 array_metadata_store.py:203] [process=0][thread=array_type_handler] Wrote 69 array_metadata.ArrayMetadata to gs://lance-maxtext/nnx_ckpt_xpk_feat_nnx_post_train_fixes_20260417_023710/nnx_xpk_feat_nnx_post_train_fixes_20260417_023710_15_ocdbt_true/checkpoints/9/items/array_metadatas/process_0 I0417 06:04:05.750830 138961225762560 base_pytree_checkpoint_handler.py:1282] [process=0][thread=write_metadata_after_commits] Commit + Array metadata written. Time taken: 41.790592s (commit=40.575946s, array_metadata_write=1.214646s) I0417 06:04:05.751884 138959480932096 base_pytree_checkpoint_handler.py:130] [process=0] /jax/orbax/write/gbytes_per_sec: 35.181 MiB/s (total gbytes: 1.5 GiB) (time elapsed: 44.89972376823425 s) (per-host) I0417 06:04:05.752016 138959480932096 async_checkpointer.py:90] [process=0][thread=async_save] 3 Handler Commit operations completed. Time taken: 44.860360s. I0417 06:04:07.567037 138959480932096 array_metadata_store.py:411] [process=0][thread=async_save] Validated ArrayMetadata from all 8 hosts. Time taken: 0.000441s. I0417 06:04:07.794740 138959480932096 ocdbt_utils.py:49] Param validation support for Zarr3 will be added later (b/362328389). I0417 06:04:08.762270 138959480932096 base_pytree_checkpoint_handler.py:1406] [process=0][thread=async_save] Pytree save finalize (merge_ocdbt + ArrayMetadata validation) completed. Time taken: 2.196487s. use_zarr3=True, enable_post_merge_validation=True, directory=gs://lance-maxtext/nnx_ckpt_xpk_feat_nnx_post_train_fixes_20260417_023710/nnx_xpk_feat_nnx_post_train_fixes_20260417_023710_15_ocdbt_true/checkpoints/9/items I0417 06:04:08.764006 138959480932096 atomicity.py:666] Finalizing gs://lance-maxtext/nnx_ckpt_xpk_feat_nnx_post_train_fixes_20260417_023710/nnx_xpk_feat_nnx_post_train_fixes_20260417_023710_15_ocdbt_true/checkpoints/9/items I0417 06:04:09.309815 138959480932096 atomicity.py:666] Finalizing gs://lance-maxtext/nnx_ckpt_xpk_feat_nnx_post_train_fixes_20260417_023710/nnx_xpk_feat_nnx_post_train_fixes_20260417_023710_15_ocdbt_true/checkpoints/9 I0417 06:04:10.712296 138959480932096 atomicity.py:847] [process=0][thread=async_save] Finished saving checkpoint (finalized tmp dir) to `gs://lance-maxtext/nnx_ckpt_xpk_feat_nnx_post_train_fixes_20260417_023710/nnx_xpk_feat_nnx_post_train_fixes_20260417_023710_15_ocdbt_true/checkpoints/9`. I0417 06:04:10.713150 138959480932096 event_tracking.py:138] [process=0] [async] Finished save (blocking + background) in 50.58 seconds @ gs://lance-maxtext/nnx_ckpt_xpk_feat_nnx_post_train_fixes_20260417_023710/nnx_xpk_feat_nnx_post_train_fixes_20260417_023710_15_ocdbt_true/checkpoints/9 I0417 06:04:10.714501 138959480932096 async_checkpointer.py:160] [process=0][thread=async_save] Background save thread done. Time taken: 49.822842s. I0417 06:04:10.714705 138960697284352 async_checkpointer.py:288] [process=0][thread=save_finalize] Done with waiting for background save thread=async_save. I0417 06:04:10.714828 138960697284352 async_checkpointer.py:298] [process=0][thread=save_finalize] No errors found in background save thread=async_save. I0417 06:04:10.714893 138960697284352 checkpoint_manager.py:2137] [process=0][thread=save_finalize][step=9] CheckpointManager Save Finalize is syncing with other hosts... I0417 06:04:10.717246 138960697284352 checkpoint_manager.py:2146] [process=0][thread=save_finalize][step=9] CheckpointManager Save Finalize is done on all hosts. I0417 06:04:10.717365 139090603628352 checkpoint_manager.py:2032] [process=0][thread=MainThread][step=9][wait_until_finished] Done waiting for Save Finalize thread (save_finalize) running at step=9. I0417 06:04:10.717477 139090603628352 checkpoint_manager.py:2009] [process=0][thread=MainThread][wait_until_finished] No Save Finalize thread to wait for. Returning. I0417 06:04:10.718184 139090603628352 metric_logger.py:185] completed step: 9, seconds: 0.147, TFLOP/s/device: 92.351, Tokens/s/device: 13920.136, total_weights: 65536, loss: 7.176 Per train step: Total TFLOPs: 13.59 split as 93.93% learnable weight flops and 6.07% attention flops XPK End: Fri Apr 17 06:04:20 UTC 2026 EXIT_CODE=0