XPK Start: Fri Apr 17 05:03:41 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 05:04:05.755636: 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 05:04:05.932393 137741326899008 max_utils.py:273] Attempting to initialize the jax distributed system... I0417 05:04:14.971423 137741326899008 distributed.py:149] Starting JAX distributed service on [::]:8482 I0417 05:04:14.973809 137741326899008 distributed.py:172] Connecting to JAX distributed service on mt-07-eval-2hlai-slice-job-0-0.mt-07-eval-2hlai:8482 I0417 05:04:16.533199 137741326899008 max_utils.py:284] Jax distributed system initialized! I0417 05:04:22.702253 137741326899008 max_utils.py:800] System Information: Jax Version: 0.9.2 I0417 05:04:22.702355 137741326899008 max_utils.py:801] System Information: Jaxlib Version: 0.9.2 I0417 05:04:22.702394 137741326899008 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 05:04:22.702429 137741326899008 train_utils.py:377] WARNING: Sequence packing is essentially ignored for synthetic data. Please use a real dataset to use sequence packing. I0417 05:04:23.371312 137741326899008 maxtext_utils.py:1687] Num_devices: 32, shape (1, 1, 1, 32, 1, 1, 1, 1, 1, 1, 1, 1, 1) I0417 05:04:23.484453 137741326899008 checkpointing.py:688] Setting up checkpoint logger... I0417 05:04:23.484566 137741326899008 checkpointing.py:234] Creating checkpoint manager with ocdbt=True and zarr3=True I0417 05:04:23.484611 137741326899008 pytree_checkpoint_handler.py:592] save_device_host_concurrent_bytes=None I0417 05:04:23.484797 137741326899008 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 0x7d45d81b1520>, enable_pinned_host_transfer=False, save_concurrent_bytes: 96000000000 (89.4 GiB), restore_concurrent_bytes: 96000000000 (89.4 GiB) I0417 05:04:26.373177 137741326899008 checkpointing.py:266] Enabling policy for fixed interval checkpointing. I0417 05:04:26.373409 137741326899008 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 0x7d30746e4170>}, handler_registry=None I0417 05:04:26.373652 137741326899008 composite_checkpoint_handler.py:237] Deferred registration for item: "items". Adding handler `<orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeCheckpointHandler object at 0x7d30746e4170>` 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 05:04:26.373701 137741326899008 composite_checkpoint_handler.py:237] Deferred registration for item: "metrics". Adding handler `<orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonCheckpointHandler object at 0x7d3094400290>` 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 05:04:26.373736 137741326899008 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 0x7d30746e4170>, ('items', <class 'orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeRestoreArgs'>): <orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeCheckpointHandler object at 0x7d30746e4170>, ('metrics', <class 'orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonSaveArgs'>): <orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonCheckpointHandler object at 0x7d3094400290>, ('metrics', <class 'orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonRestoreArgs'>): <orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonCheckpointHandler object at 0x7d3094400290>}). I0417 05:04:26.374142 137741326899008 abstract_checkpointer.py:35] orbax-checkpoint version: 0.11.34 I0417 05:04:26.374215 137741326899008 async_checkpointer.py:192] [process=0][thread=MainThread] Using barrier_sync_fn: <function get_barrier_sync_fn.<locals>._fn at 0x7d3094482020> timeout: 1200 secs and primary_host=0 for async checkpoint writes I0417 05:04:28.362889 137741326899008 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_07_eval/checkpoints I0417 05:04:28.579339 137741326899008 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=10), preservation_policy=LatestN(n=None), prevent_write_metrics=False, enable_should_save_is_saving_in_progress_check=True, enable_per_process_directory_creation=False, lightweight_initialize=False), root_directory=gs://lance-maxtext/nnx_ckpt_xpk_feat_nnx_post_train_fixes_20260417_023710/nnx_xpk_feat_nnx_post_train_fixes_20260417_023710_07_eval/checkpoints: <orbax.checkpoint.checkpoint_manager.CheckpointManager object at 0x7d30740fc590> I0417 05:04:28.579511 137741326899008 checkpointing.py:302] Checkpoint manager created! I0417 05:04:29.527008 137741326899008 checkpointing.py:578] checkpoint manager exists so trying to load this run's existing checkpoint I0417 05:04:29.527124 137741326899008 checkpointing.py:676] No existing checkpoints found, not restoring checkpoint. fsdp: 32 I0417 05:04:30.873378 137741326899008 maxtext_utils.py:1805] decoder/decoder_norm/scale/value Shape: float32[2048] Physical: (None,) I0417 05:04:30.873492 137741326899008 maxtext_utils.py:1805] decoder/layers/mlp/wi_0/kernel/value Shape: float32[2048,16,7168] Physical: ('fsdp', None, None) I0417 05:04:30.873541 137741326899008 maxtext_utils.py:1805] decoder/layers/mlp/wi_1/kernel/value Shape: float32[2048,16,7168] Physical: ('fsdp', None, None) I0417 05:04:30.873583 137741326899008 maxtext_utils.py:1805] decoder/layers/mlp/wo/kernel/value Shape: float32[7168,16,2048] Physical: (None, None, 'fsdp') I0417 05:04:30.873619 137741326899008 maxtext_utils.py:1805] decoder/layers/post_self_attention_layer_norm/scale/value Shape: float32[2048,16] Physical: (None, None) I0417 05:04:30.873653 137741326899008 maxtext_utils.py:1805] decoder/layers/pre_self_attention_layer_norm/scale/value Shape: float32[2048,16] Physical: (None, None) I0417 05:04:30.873688 137741326899008 maxtext_utils.py:1805] decoder/layers/self_attention/key/kernel/value Shape: float32[2048,16,16,128] Physical: ('fsdp', None, None, None) I0417 05:04:30.873722 137741326899008 maxtext_utils.py:1805] decoder/layers/self_attention/out/kernel/value Shape: float32[16,16,128,2048] Physical: (None, None, None, 'fsdp') I0417 05:04:30.873755 137741326899008 maxtext_utils.py:1805] decoder/layers/self_attention/query/kernel/value Shape: float32[2048,16,16,128] Physical: ('fsdp', None, None, None) I0417 05:04:30.873788 137741326899008 maxtext_utils.py:1805] decoder/layers/self_attention/value/kernel/value Shape: float32[2048,16,16,128] Physical: ('fsdp', None, None, None) I0417 05:04:30.873821 137741326899008 maxtext_utils.py:1805] decoder/logits_dense/kernel/value Shape: float32[2048,32000] Physical: ('fsdp', None) I0417 05:04:30.873867 137741326899008 maxtext_utils.py:1805] token_embedder/embedding/value Shape: float32[32000,2048] Physical: (None, 'fsdp') I0417 05:04:31.268913 137741326899008 nnx_decoders.py:465] nnx_decoders/carry Logical: bfloat16[32,2048,2048]...................................... ('activation_batch', 'activation_norm_length', 'activation_embed'). I0417 05:04:31.269010 137741326899008 nnx_decoders.py:465] nnx_decoders/carry Physical: bfloat16[32,2048,2048]...................................... ('fsdp', None, None). I0417 05:04:31.274239 137741326899008 nnx_decoders.py:465] Unknown Logical: bfloat16[32,2048,2048]...................................... ('activation_batch', 'activation_norm_length', 'activation_embed'). I0417 05:04:31.274295 137741326899008 nnx_decoders.py:465] Unknown Physical: bfloat16[32,2048,2048]...................................... ('fsdp', None, None). I0417 05:04:31.290395 137741326899008 attentions.py:1088] attentions/inputs_q Logical: bfloat16[32,2048,2048]...................................... ('activation_batch', 'activation_attn_length', 'activation_attn_embed'). I0417 05:04:31.290457 137741326899008 attentions.py:1088] attentions/inputs_q Physical: bfloat16[32,2048,2048]...................................... ('fsdp', None, None). I0417 05:04:31.305913 137741326899008 attentions.py:1089] attentions/inputs_kv Logical: bfloat16[32,2048,2048]...................................... ('activation_batch', 'activation_attn_length', 'activation_attn_embed'). I0417 05:04:31.305968 137741326899008 attentions.py:1089] attentions/inputs_kv Physical: bfloat16[32,2048,2048]...................................... ('fsdp', None, None). I0417 05:04:31.334754 137741326899008 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 05:04:31.334874 137741326899008 attentions.py:1154] attentions/query Physical: bfloat16[32,2048,16,128].................................... ('fsdp', None, None, None). I0417 05:04:31.350509 137741326899008 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 05:04:31.350573 137741326899008 attentions.py:1155] attentions/key Physical: bfloat16[32,2048,16,128].................................... ('fsdp', None, None, None). I0417 05:04:31.366051 137741326899008 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 05:04:31.366119 137741326899008 attentions.py:1156] attentions/value Physical: bfloat16[32,2048,16,128].................................... ('fsdp', None, None, None). I0417 05:04:31.395956 137741326899008 attentions.py:1197] attentions/out Logical: bfloat16[32,2048,16,128].................................... ('activation_batch', 'activation_attn_length', 'activation_heads', 'activation_kv'). I0417 05:04:31.396023 137741326899008 attentions.py:1197] attentions/out Physical: bfloat16[32,2048,16,128].................................... ('fsdp', None, None, None). I0417 05:04:31.415546 137741326899008 linears.py:525] linears/x Logical: bfloat16[32,2048,7168]...................................... ('activation_batch', 'activation_length', 'activation_mlp'). I0417 05:04:31.415614 137741326899008 linears.py:525] linears/x Physical: bfloat16[32,2048,7168]...................................... ('fsdp', None, None). I0417 05:04:42.715883 137741326899008 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 05:04:42.725036 137741326899008 metric_logger.py:289] number parameters: 1.104 billion I0417 05:04:54.332659 137741326899008 checkpointing.py:794] Waiting for step 0 to finish before checkpoint... I0417 05:04:54.484415 137741326899008 checkpointing.py:798] Waited 0.15174078941345215 seconds for step 0 to finish before starting checkpointing. I0417 05:04:54.487353 137741326899008 checkpoint_manager.py:2009] [process=0][thread=MainThread][wait_until_finished] No Save Finalize thread to wait for. Returning. I0417 05:04:54.489160 137741326899008 checkpoint_manager.py:1512] [process=0] Saving checkpoint at step 0 I0417 05:04:54.490529 137741326899008 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_07_eval/checkpoints/0. I0417 05:04:55.653505 137741326899008 signaling_client.py:364] Using JaxDistributedSignalingClient I0417 05:04:55.674212 137741326899008 jax_array_handlers.py:360] Scheduling D2H of 69 prioritized jax.Array. I0417 05:04:55.674286 137741326899008 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 05:04:55.911071 137612860696320 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_07_eval/checkpoints/0 I0417 05:04:55.943944 137741326899008 base_pytree_checkpoint_handler.py:154] [process=0][thread=MainThread] Initiated "orbax.checkpoint._src.serialization.jax_array_handlers.ArrayHandler".serialize. Time taken: 0.271132s I0417 05:04:55.944931 137741326899008 base_pytree_checkpoint_handler.py:130] [process=0] /jax/orbax/write/blocking_gbytes_per_sec: 5.467 GiB/s (total gbytes: 1.5 GiB) (time elapsed: 0.28217458724975586 s) (per-host) I0417 05:04:55.945011 137741326899008 base_pytree_checkpoint_handler.py:768] [process=0][thread=MainThread] Initiated Pytree async_save. Time taken: 0.282276s (batch_requests_ready=0.003060s, total_serialization_initiated=0.278308s, others=0.000908s) I0417 05:04:55.945098 137741326899008 composite_checkpoint_handler.py:715] [process=0][thread=MainThread] Initiated CompositeCheckpointHandler.async_save. Time taken: 0.286365s (all_items=0.000016s, per_item={'items': '0.00001645'}, temp_paths=0.286348) I0417 05:04:55.945859 137741326899008 event_tracking.py:125] [process=0] [async] Finished blocking save in 1.46 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_07_eval/checkpoints/0. I0417 05:04:55.946146 137610437576448 async_checkpointer.py:76] [process=0][thread=async_save] Background save thread started. Deadline for this save operation is 2026-04-17 05:24:55.946108 I0417 05:04:55.959793 137741326899008 checkpoint_manager.py:1560] [process=0][thread=MainThread][step=0] Starting CheckpointManager Save Finalize thread=save_finalize I0417 05:04:55.960128 137612869089024 async_checkpointer.py:280] [process=0][thread=save_finalize] Waiting for background save thread=async_save. I0417 05:04:55.960285 137741326899008 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_07_eval/checkpoints', 'reached_preemption': False, 'preemption_received_at': None, 'synchronous': False, 'wait_for_prev_start_time': 1776402294.4873226, 'wait_for_prev_duration_secs': 0.00011801719665527344, 'time_between_consecutive_saves_sec': None, 'checkpointer_blocking_start_time': 1776402294.4891984, 'checkpointer_blocking_duration_secs': 1.4571044445037842, 'get_old_steps_start_time': 1776402295.9463284, 'get_old_steps_duration_secs': 2.9802322387695312e-05, 'checkpoint_manager_blocking_start_time': 1776402294.4848042, 'checkpoint_manager_blocking_duration_secs': 1.4754419326782227} I0417 05:04:55.960391 137741326899008 checkpointing.py:409] Started an asynchronous checkpoint save for step 0 I0417 05:04:55.960458 137741326899008 max_utils.py:750] Memstats: After params initialized: I0417 05:04:55.960508 137741326899008 max_utils.py:756] Using (GB) 0.43 / 31.25 (1.376000%) on TPU_0(process=0,(0,0,0,0)) I0417 05:04:55.960541 137741326899008 max_utils.py:756] Using (GB) 0.43 / 31.25 (1.376000%) on TPU_1(process=0,(1,0,0,0)) I0417 05:04:55.960569 137741326899008 max_utils.py:756] Using (GB) 0.43 / 31.25 (1.376000%) on TPU_4(process=0,(0,1,0,0)) I0417 05:04:55.960594 137741326899008 max_utils.py:756] Using (GB) 0.43 / 31.25 (1.376000%) on TPU_5(process=0,(1,1,0,0)) I0417 05:04:56.301707 137741326899008 metric_logger.py:185] completed step: 0, seconds: 11.283, TFLOP/s/device: 1.204, Tokens/s/device: 181.509, total_weights: 65536, loss: 10.874 I0417 05:04:56.304593 137741326899008 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_07_eval/tensorboard/' I0417 05:04:56.740755 137741326899008 metric_logger.py:185] completed step: 1, seconds: 1.966, TFLOP/s/device: 6.912, Tokens/s/device: 1041.881, total_weights: 65536, loss: 10.874 I0417 05:04:56.882592 137741326899008 metric_logger.py:185] completed step: 2, seconds: 0.440, TFLOP/s/device: 30.877, Tokens/s/device: 4654.059, total_weights: 65536, loss: 10.262 I0417 05:04:57.030001 137741326899008 metric_logger.py:185] completed step: 3, seconds: 0.012, TFLOP/s/device: 1108.791, Tokens/s/device: 167129.101, total_weights: 65536, loss: 9.732 I0417 05:04:58.559414 137612843910912 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_07_eval/checkpoints/0/items I0417 05:04:58.753200 137609688663808 checkpoint.py:188] Wrote Metadata={'item_handlers': None, 'metrics': {}, 'performance_metrics': {}, 'init_timestamp_nsecs': 1776402298314739275, 'commit_timestamp_nsecs': None, 'custom_metadata': {}}, json={"item_handlers": null, "metrics": {}, "performance_metrics": {}, "init_timestamp_nsecs": 1776402298314739275, "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_07_eval/checkpoints/0/_CHECKPOINT_METADATA I0417 05:05:00.113558 2804 google_auth_provider.cc:181] Running on GCE, using service account 562977990677-compute@developer.gserviceaccount.com I0417 05:05:01.885341 137610454361856 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_07_eval/checkpoints/0/items/array_metadatas/process_0 I0417 05:05:19.365542 137741326899008 metric_logger.py:185] completed step: 4, seconds: 0.143, TFLOP/s/device: 94.694, Tokens/s/device: 14273.269, total_weights: 65536, loss: 9.274 I0417 05:05:19.375633 137741326899008 metric_logger.py:185] completed step: 5, seconds: 0.146, TFLOP/s/device: 92.933, Tokens/s/device: 14007.920, total_weights: 65536, loss: 8.888 I0417 05:05:19.520164 137741326899008 metric_logger.py:185] completed step: 6, seconds: 22.334, TFLOP/s/device: 0.608, Tokens/s/device: 91.698, total_weights: 65536, loss: 8.589 I0417 05:05:19.667644 137741326899008 metric_logger.py:185] completed step: 7, seconds: 0.008, TFLOP/s/device: 1729.963, Tokens/s/device: 260758.849, total_weights: 65536, loss: 8.380 I0417 05:05:19.814953 137741326899008 metric_logger.py:185] completed step: 8, seconds: 0.146, TFLOP/s/device: 93.066, Tokens/s/device: 14027.974, total_weights: 65536, loss: 8.251 I0417 05:05:19.963398 137741326899008 checkpointing.py:794] Waiting for step 9 to finish before checkpoint... I0417 05:05:19.964076 137741326899008 checkpointing.py:798] Waited 0.0006890296936035156 seconds for step 9 to finish before starting checkpointing. I0417 05:05:19.966004 137741326899008 checkpoint_manager.py:2020] [process=0][thread=MainThread][step=0][wait_until_finished] Waiting for Save Finalize thread (save_finalize) to complete. I0417 05:05:36.570445 137610445969152 base_pytree_checkpoint_handler.py:1282] [process=0][thread=write_metadata_after_commits] Commit + Array metadata written. Time taken: 36.464273s (commit=35.258311s, array_metadata_write=1.205962s) I0417 05:05:36.571561 137610437576448 base_pytree_checkpoint_handler.py:130] [process=0] /jax/orbax/write/gbytes_per_sec: 38.613 MiB/s (total gbytes: 1.5 GiB) (time elapsed: 40.90880846977234 s) (per-host) I0417 05:05:36.571629 137610437576448 async_checkpointer.py:90] [process=0][thread=async_save] 3 Handler Commit operations completed. Time taken: 40.625365s. I0417 05:05:37.363819 137610437576448 checkpoint.py:228] Read Metadata={'item_handlers': None, 'metrics': {}, 'performance_metrics': {}, 'init_timestamp_nsecs': 1776402298314739275, '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_07_eval/checkpoints/0/_CHECKPOINT_METADATA I0417 05:05:38.106490 137609688663808 checkpoint.py:247] Updated Metadata={'item_handlers': {'items': 'orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeCheckpointHandler'}, 'metrics': {}, 'performance_metrics': {}, 'init_timestamp_nsecs': 1776402298314739275, '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_07_eval/checkpoints/0/_CHECKPOINT_METADATA I0417 05:05:38.467026 137610437576448 array_metadata_store.py:411] [process=0][thread=async_save] Validated ArrayMetadata from all 8 hosts. Time taken: 0.000447s. I0417 05:05:38.726449 137610437576448 ocdbt_utils.py:49] Param validation support for Zarr3 will be added later (b/362328389). I0417 05:05:39.657257 137610437576448 base_pytree_checkpoint_handler.py:1406] [process=0][thread=async_save] Pytree save finalize (merge_ocdbt + ArrayMetadata validation) completed. Time taken: 2.172359s. 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_07_eval/checkpoints/0/items I0417 05:05:39.659056 137610437576448 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_07_eval/checkpoints/0/items I0417 05:05:40.593065 137610437576448 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_07_eval/checkpoints/0 I0417 05:05:42.401708 137610437576448 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_07_eval/checkpoints/0`. I0417 05:05:42.402566 137610437576448 event_tracking.py:138] [process=0] [async] Finished save (blocking + background) in 47.91 seconds @ gs://lance-maxtext/nnx_ckpt_xpk_feat_nnx_post_train_fixes_20260417_023710/nnx_xpk_feat_nnx_post_train_fixes_20260417_023710_07_eval/checkpoints/0 I0417 05:05:42.404000 137610437576448 async_checkpointer.py:160] [process=0][thread=async_save] Background save thread done. Time taken: 46.457732s. I0417 05:05:42.404199 137612869089024 async_checkpointer.py:288] [process=0][thread=save_finalize] Done with waiting for background save thread=async_save. I0417 05:05:42.404326 137612869089024 async_checkpointer.py:298] [process=0][thread=save_finalize] No errors found in background save thread=async_save. I0417 05:05:42.404392 137612869089024 checkpoint_manager.py:2137] [process=0][thread=save_finalize][step=0] CheckpointManager Save Finalize is syncing with other hosts... I0417 05:05:42.407191 137612869089024 checkpoint_manager.py:2146] [process=0][thread=save_finalize][step=0] CheckpointManager Save Finalize is done on all hosts. I0417 05:05:42.407341 137741326899008 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 05:05:42.407440 137741326899008 checkpoint_manager.py:1452] Waiting for previous save to complete took 22.441438 seconds. If this number is high, consider checkpointing less frequently. I0417 05:05:42.409194 137741326899008 checkpoint_manager.py:1512] [process=0] Saving checkpoint at step 9 I0417 05:05:42.411054 137741326899008 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_07_eval/checkpoints/9. I0417 05:05:43.126523 137741326899008 jax_array_handlers.py:360] Scheduling D2H of 69 prioritized jax.Array. I0417 05:05:43.126618 137741326899008 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 05:05:43.153395 137741326899008 base_pytree_checkpoint_handler.py:154] [process=0][thread=MainThread] Initiated "orbax.checkpoint._src.serialization.jax_array_handlers.ArrayHandler".serialize. Time taken: 0.028331s I0417 05:05:43.154200 137741326899008 base_pytree_checkpoint_handler.py:130] [process=0] /jax/orbax/write/blocking_gbytes_per_sec: 41.619 GiB/s (total gbytes: 1.5 GiB) (time elapsed: 0.0370638370513916 s) (per-host) I0417 05:05:43.154274 137741326899008 base_pytree_checkpoint_handler.py:768] [process=0][thread=MainThread] Initiated Pytree async_save. Time taken: 0.037157s (batch_requests_ready=0.002657s, total_serialization_initiated=0.033759s, others=0.000741s) I0417 05:05:43.154356 137741326899008 composite_checkpoint_handler.py:715] [process=0][thread=MainThread] Initiated CompositeCheckpointHandler.async_save. Time taken: 0.041778s (all_items=0.000020s, per_item={'items': '0.00001979'}, temp_paths=0.041758) I0417 05:05:43.155118 137741326899008 event_tracking.py:125] [process=0] [async] Finished blocking save in 0.75 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_07_eval/checkpoints/9. I0417 05:05:43.155492 137612869089024 async_checkpointer.py:76] [process=0][thread=async_save] Background save thread started. Deadline for this save operation is 2026-04-17 05:25:43.155456 I0417 05:05:43.162690 137741326899008 checkpoint_manager.py:1560] [process=0][thread=MainThread][step=9] Starting CheckpointManager Save Finalize thread=save_finalize I0417 05:05:43.162982 137611817510656 async_checkpointer.py:280] [process=0][thread=save_finalize] Waiting for background save thread=async_save. I0417 05:05:43.163155 137741326899008 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_07_eval/checkpoints', 'reached_preemption': False, 'preemption_received_at': None, 'synchronous': False, 'wait_for_prev_start_time': 1776402319.9659736, 'wait_for_prev_duration_secs': 22.44143772125244, 'time_between_consecutive_saves_sec': None, 'checkpointer_blocking_start_time': 1776402342.409232, 'checkpointer_blocking_duration_secs': 0.7463960647583008, 'get_old_steps_start_time': 1776402343.1556525, 'get_old_steps_duration_secs': 3.147125244140625e-05, 'checkpoint_manager_blocking_start_time': 1776402319.9642954, 'checkpoint_manager_blocking_duration_secs': 23.198823928833008} I0417 05:05:43.163264 137741326899008 checkpointing.py:409] Started an asynchronous checkpoint save for step 9 I0417 05:05:43.163309 137741326899008 checkpoint_manager.py:2020] [process=0][thread=MainThread][step=9][wait_until_finished] Waiting for Save Finalize thread (save_finalize) to complete. I0417 05:05:43.374467 137610437576448 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_07_eval/checkpoints/9 I0417 05:05:45.101212 137612312418048 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_07_eval/checkpoints/9/items I0417 05:05:48.000183 137612329203456 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_07_eval/checkpoints/9/items/array_metadatas/process_0 I0417 05:06:28.242758 137611825903360 base_pytree_checkpoint_handler.py:1282] [process=0][thread=write_metadata_after_commits] Commit + Array metadata written. Time taken: 42.066581s (commit=40.396002s, array_metadata_write=1.670580s) I0417 05:06:28.243946 137612869089024 base_pytree_checkpoint_handler.py:130] [process=0] /jax/orbax/write/gbytes_per_sec: 35.004 MiB/s (total gbytes: 1.5 GiB) (time elapsed: 45.126781702041626 s) (per-host) I0417 05:06:28.244077 137612869089024 async_checkpointer.py:90] [process=0][thread=async_save] 3 Handler Commit operations completed. Time taken: 45.088466s. I0417 05:06:30.170057 137612869089024 array_metadata_store.py:411] [process=0][thread=async_save] Validated ArrayMetadata from all 8 hosts. Time taken: 0.000420s. I0417 05:06:30.561506 137612869089024 ocdbt_utils.py:49] Param validation support for Zarr3 will be added later (b/362328389). I0417 05:06:31.478857 137612869089024 base_pytree_checkpoint_handler.py:1406] [process=0][thread=async_save] Pytree save finalize (merge_ocdbt + ArrayMetadata validation) completed. Time taken: 2.320083s. 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_07_eval/checkpoints/9/items I0417 05:06:31.484052 137612869089024 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_07_eval/checkpoints/9/items I0417 05:06:32.022286 137612869089024 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_07_eval/checkpoints/9 I0417 05:06:33.869234 137612869089024 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_07_eval/checkpoints/9`. I0417 05:06:33.870079 137612869089024 event_tracking.py:138] [process=0] [async] Finished save (blocking + background) in 51.46 seconds @ gs://lance-maxtext/nnx_ckpt_xpk_feat_nnx_post_train_fixes_20260417_023710/nnx_xpk_feat_nnx_post_train_fixes_20260417_023710_07_eval/checkpoints/9 I0417 05:06:33.871729 137612869089024 async_checkpointer.py:160] [process=0][thread=async_save] Background save thread done. Time taken: 50.716122s. I0417 05:06:33.871874 137611817510656 async_checkpointer.py:288] [process=0][thread=save_finalize] Done with waiting for background save thread=async_save. I0417 05:06:33.871937 137611817510656 async_checkpointer.py:298] [process=0][thread=save_finalize] No errors found in background save thread=async_save. I0417 05:06:33.871994 137611817510656 checkpoint_manager.py:2137] [process=0][thread=save_finalize][step=9] CheckpointManager Save Finalize is syncing with other hosts... I0417 05:06:33.874969 137611817510656 checkpoint_manager.py:2146] [process=0][thread=save_finalize][step=9] CheckpointManager Save Finalize is done on all hosts. I0417 05:06:33.875149 137741326899008 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 05:06:33.875338 137741326899008 checkpoint_manager.py:2009] [process=0][thread=MainThread][wait_until_finished] No Save Finalize thread to wait for. Returning. I0417 05:06:33.875998 137741326899008 metric_logger.py:185] completed step: 9, seconds: 0.147, TFLOP/s/device: 92.338, Tokens/s/device: 13918.149, total_weights: 65536, loss: 8.175 Per train step: Total TFLOPs: 13.59 split as 93.93% learnable weight flops and 6.07% attention flops XPK End: Fri Apr 17 05:06:42 UTC 2026 EXIT_CODE=0