main (7f7822822) vs feat/nnx-post-train-fixes (2fbc8c2f1)| Metric | main 7f7822822 | feat/nnx-post-train-fixes 2fbc8c2f1 | Diff (feat/nnx-post-train-fixes − main) |
|---|---|---|---|
| Parameters | 1.104 billion | 1.104 billion | — |
| Final loss | 7.1810 | 7.1810 | 0 |
| TFLOP/s | 92.195 | 92.220 | +0.025 |
| Tok/s | 13896.6 | 13900.4 | +3.773 |
| Avg s/step | 7.649 | 7.730 | +0.081 |
| Memory % | 1.38 | 1.38 | 0 |
| JAX | 0.9.2 | 0.9.2 | — |
Diff = branch value − main value. Green = branch improved. Red = branch regressed.
XPK Start: Fri Apr 17 14:47:55 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 14:48:19.332080: 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 14:48:19.511034 135646693517120 max_utils.py:273] Attempting to initialize the jax distributed system... I0417 14:48:28.551275 135646693517120 distributed.py:149] Starting JAX distributed service on [::]:8482 I0417 14:48:28.553786 135646693517120 distributed.py:172] Connecting to JAX distributed service on mt-16-zarr3-true-sv-e7xye-slice-job-0-0.mt-16-zarr3-true-sv-e7xye:8482 I0417 14:48:29.596595 135646693517120 max_utils.py:284] Jax distributed system initialized! I0417 14:48:35.660345 135646693517120 max_utils.py:800] System Information: Jax Version: 0.9.2 I0417 14:48:35.660448 135646693517120 max_utils.py:801] System Information: Jaxlib Version: 0.9.2 I0417 14:48:35.660492 135646693517120 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 14:48:35.660535 135646693517120 train_utils.py:348] WARNING: Sequence packing is essentially ignored for synthetic data. Please use a real dataset to use sequence packing. I0417 14:48:36.332295 135646693517120 maxtext_utils.py:1548] Num_devices: 32, shape (1, 1, 1, 32, 1, 1, 1, 1, 1, 1, 1, 1, 1) I0417 14:48:36.332581 135646693517120 checkpointing.py:677] Setting up checkpoint logger... I0417 14:48:36.332638 135646693517120 checkpointing.py:233] Creating checkpoint manager with ocdbt=True and zarr3=True I0417 14:48:36.332683 135646693517120 pytree_checkpoint_handler.py:592] save_device_host_concurrent_bytes=None I0417 14:48:36.333052 135646693517120 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 0x7b5e14919d30>, enable_pinned_host_transfer=False, save_concurrent_bytes: 96000000000 (89.4 GiB), restore_concurrent_bytes: 96000000000 (89.4 GiB) I0417 14:48:39.300322 135646693517120 checkpointing.py:265] Enabling policy for fixed interval checkpointing. I0417 14:48:39.300652 135646693517120 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 0x7b49ac7cc920>}, handler_registry=None I0417 14:48:39.300909 135646693517120 composite_checkpoint_handler.py:237] Deferred registration for item: "items". Adding handler `<orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeCheckpointHandler object at 0x7b49ac7cc920>` 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 14:48:39.300958 135646693517120 composite_checkpoint_handler.py:237] Deferred registration for item: "metrics". Adding handler `<orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonCheckpointHandler object at 0x7b49ac7d06e0>` 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 14:48:39.300995 135646693517120 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 0x7b49ac7cc920>, ('items', <class 'orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeRestoreArgs'>): <orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeCheckpointHandler object at 0x7b49ac7cc920>, ('metrics', <class 'orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonSaveArgs'>): <orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonCheckpointHandler object at 0x7b49ac7d06e0>, ('metrics', <class 'orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonRestoreArgs'>): <orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonCheckpointHandler object at 0x7b49ac7d06e0>}). I0417 14:48:39.301304 135646693517120 abstract_checkpointer.py:35] orbax-checkpoint version: 0.11.34 I0417 14:48:39.301374 135646693517120 async_checkpointer.py:192] [process=0][thread=MainThread] Using barrier_sync_fn: <function get_barrier_sync_fn.<locals>._fn at 0x7b49ac1b0a40> timeout: 1200 secs and primary_host=0 for async checkpoint writes I0417 14:48:40.372404 135646693517120 checkpoint_manager.py:1812] Found 0 checkpoint steps in gs://lance-maxtext/linen_ckpt_xpk_main_20260417_125726/linen_xpk_main_20260417_125726_16_zarr3_true/checkpoints I0417 14:48:40.994585 135646693517120 checkpoint_manager.py:929] [process=0][thread=MainThread] CheckpointManager created, primary_host=0, CheckpointManagerOptions=CheckpointManagerOptions(save_interval_steps=1, max_to_keep=None, keep_time_interval=None, keep_period=None, should_keep_fn=None, best_fn=None, best_mode='max', keep_checkpoints_without_metrics=True, step_prefix=None, step_format_fixed_length=None, step_name_format=None, create=True, cleanup_tmp_directories=False, save_on_steps=frozenset(), single_host_load_and_broadcast=False, todelete_subdir=None, todelete_full_path=None, enable_background_delete=False, read_only=False, enable_async_checkpointing=True, async_options=None, multiprocessing_options=MultiprocessingOptions(primary_host=0, active_processes=None, barrier_sync_key_prefix=None), should_save_fn=None, file_options=FileOptions(path_permission_mode=None), save_root_metadata=True, temporary_path_class=None, save_decision_policy=FixedIntervalPolicy(interval=5), preservation_policy=LatestN(n=None), prevent_write_metrics=False, enable_should_save_is_saving_in_progress_check=True, enable_per_process_directory_creation=False, lightweight_initialize=False), root_directory=gs://lance-maxtext/linen_ckpt_xpk_main_20260417_125726/linen_xpk_main_20260417_125726_16_zarr3_true/checkpoints: <orbax.checkpoint.checkpoint_manager.CheckpointManager object at 0x7b49ac7ce690> I0417 14:48:40.994755 135646693517120 checkpointing.py:301] Checkpoint manager created! I0417 14:48:41.904882 135646693517120 nnx_wrappers.py:437] Unknown Logical: bfloat16[32,2048,2048]...................................... ('activation_batch', 'activation_norm_length', 'activation_embed'). I0417 14:48:41.904994 135646693517120 nnx_wrappers.py:437] Unknown Physical: bfloat16[32,2048,2048]...................................... ('fsdp', None, None). I0417 14:48:42.281774 135646693517120 attentions.py:1088] attentions/inputs_q Logical: bfloat16[32,2048,2048]...................................... ('activation_batch', 'activation_attn_length', 'activation_attn_embed'). I0417 14:48:42.281876 135646693517120 attentions.py:1088] attentions/inputs_q Physical: bfloat16[32,2048,2048]...................................... ('fsdp', None, None). I0417 14:48:42.298039 135646693517120 attentions.py:1089] attentions/inputs_kv Logical: bfloat16[32,2048,2048]...................................... ('activation_batch', 'activation_attn_length', 'activation_attn_embed'). I0417 14:48:42.298109 135646693517120 attentions.py:1089] attentions/inputs_kv Physical: bfloat16[32,2048,2048]...................................... ('fsdp', None, None). I0417 14:48:42.321949 135646693517120 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 14:48:42.322014 135646693517120 attentions.py:1154] attentions/query Physical: bfloat16[32,2048,16,128].................................... ('fsdp', None, None, None). I0417 14:48:42.338394 135646693517120 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 14:48:42.338457 135646693517120 attentions.py:1155] attentions/key Physical: bfloat16[32,2048,16,128].................................... ('fsdp', None, None, None). I0417 14:48:42.354653 135646693517120 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 14:48:42.354715 135646693517120 attentions.py:1156] attentions/value Physical: bfloat16[32,2048,16,128].................................... ('fsdp', None, None, None). I0417 14:48:42.379236 135646693517120 attentions.py:1197] attentions/out Logical: bfloat16[32,2048,16,128].................................... ('activation_batch', 'activation_attn_length', 'activation_heads', 'activation_kv'). I0417 14:48:42.379304 135646693517120 attentions.py:1197] attentions/out Physical: bfloat16[32,2048,16,128].................................... ('fsdp', None, None, None). I0417 14:48:42.401163 135646693517120 linears.py:525] linears/x Logical: bfloat16[32,2048,7168]...................................... ('activation_batch', 'activation_length', 'activation_mlp'). I0417 14:48:42.401226 135646693517120 linears.py:525] linears/x Physical: bfloat16[32,2048,7168]...................................... ('fsdp', None, None). I0417 14:48:42.616519 135646693517120 checkpointing.py:577] checkpoint manager exists so trying to load this run's existing checkpoint I0417 14:48:42.616625 135646693517120 checkpointing.py:665] No existing checkpoints found, not restoring checkpoint. fsdp: 32 I0417 14:48:44.053216 135646693517120 maxtext_utils.py:1651] params/params/decoder/decoder_norm/scale Shape: float32[2048] Logical: P('norm',) Physical: (None,) I0417 14:48:44.053341 135646693517120 maxtext_utils.py:1651] params/params/decoder/layers/mlp/wi_0/kernel Shape: float32[2048,16,7168] Logical: P('embed', 'layers', 'mlp') Physical: ('fsdp', None, None) I0417 14:48:44.053395 135646693517120 maxtext_utils.py:1651] params/params/decoder/layers/mlp/wi_1/kernel Shape: float32[2048,16,7168] Logical: P('embed', 'layers', 'mlp') Physical: ('fsdp', None, None) I0417 14:48:44.053453 135646693517120 maxtext_utils.py:1651] params/params/decoder/layers/mlp/wo/kernel Shape: float32[7168,16,2048] Logical: P('mlp', 'layers', 'embed') Physical: (None, None, 'fsdp') I0417 14:48:44.053507 135646693517120 maxtext_utils.py:1651] params/params/decoder/layers/post_self_attention_layer_norm/scale Shape: float32[2048,16] Logical: P('norm', 'layers') Physical: (None, None) I0417 14:48:44.053549 135646693517120 maxtext_utils.py:1651] params/params/decoder/layers/pre_self_attention_layer_norm/scale Shape: float32[2048,16] Logical: P('norm', 'layers') Physical: (None, None) I0417 14:48:44.053642 135646693517120 maxtext_utils.py:1651] 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) I0417 14:48:44.053744 135646693517120 maxtext_utils.py:1651] 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') I0417 14:48:44.053821 135646693517120 maxtext_utils.py:1651] 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) I0417 14:48:44.053912 135646693517120 maxtext_utils.py:1651] 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) I0417 14:48:44.054003 135646693517120 maxtext_utils.py:1651] params/params/decoder/logits_dense/kernel Shape: float32[2048,32000] Logical: P('embed_vocab', 'vocab') Physical: ('fsdp', None) I0417 14:48:44.054091 135646693517120 maxtext_utils.py:1651] params/params/token_embedder/embedding Shape: float32[32000,2048] Logical: P('vocab', 'embed_vocab') Physical: (None, 'fsdp') I0417 14:48:44.546349 135646693517120 train.py:155] train/xent Logical: float32[32,2048]............................................ ('activation_embed_and_logits_batch', 'activation_length'). I0417 14:48:44.546444 135646693517120 train.py:155] train/xent Physical: float32[32,2048]............................................ ('fsdp', None). I0417 14:48:44.561783 135646693517120 train.py:162] train/z_loss Logical: float32[32,2048]............................................ ('activation_embed_and_logits_batch', 'activation_length'). I0417 14:48:44.561854 135646693517120 train.py:162] train/z_loss Physical: float32[32,2048]............................................ ('fsdp', None). I0417 14:48:55.293828 135646693517120 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 14:48:55.301575 135646693517120 metric_logger.py:301] number parameters: 1.104 billion I0417 14:49:06.580766 135646693517120 checkpointing.py:772] Waiting for step 0 to finish before checkpoint... I0417 14:49:06.735260 135646693517120 checkpointing.py:776] Waited 0.1544787883758545 seconds for step 0 to finish before starting checkpointing. I0417 14:49:06.737648 135646693517120 checkpoint_manager.py:2009] [process=0][thread=MainThread][wait_until_finished] No Save Finalize thread to wait for. Returning. I0417 14:49:06.739452 135646693517120 checkpoint_manager.py:1512] [process=0] Saving checkpoint at step 0 I0417 14:49:06.741120 135646693517120 event_tracking.py:70] [process=0] [async] Started save checkpoint @ gs://lance-maxtext/linen_ckpt_xpk_main_20260417_125726/linen_xpk_main_20260417_125726_16_zarr3_true/checkpoints/0. I0417 14:49:07.063064 135646693517120 signaling_client.py:364] Using JaxDistributedSignalingClient I0417 14:49:07.077010 135646693517120 jax_array_handlers.py:360] Scheduling D2H of 39 prioritized jax.Array. I0417 14:49:07.077088 135646693517120 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 14:49:07.343805 135646693517120 base_pytree_checkpoint_handler.py:154] [process=0][thread=MainThread] Initiated "orbax.checkpoint._src.serialization.jax_array_handlers.ArrayHandler".serialize. Time taken: 0.267480s I0417 14:49:07.344593 135646693517120 base_pytree_checkpoint_handler.py:130] [process=0] /jax/orbax/write/blocking_gbytes_per_sec: 5.633 GiB/s (total gbytes: 1.5 GiB) (time elapsed: 0.273867130279541 s) (per-host) I0417 14:49:07.344666 135646693517120 base_pytree_checkpoint_handler.py:768] [process=0][thread=MainThread] Initiated Pytree async_save. Time taken: 0.273961s (batch_requests_ready=0.002226s, total_serialization_initiated=0.271049s, others=0.000686s) I0417 14:49:07.344748 135646693517120 composite_checkpoint_handler.py:715] [process=0][thread=MainThread] Initiated CompositeCheckpointHandler.async_save. Time taken: 0.277933s (all_items=0.000017s, per_item={'items': '0.00001717'}, temp_paths=0.277916) I0417 14:49:07.345576 135646693517120 event_tracking.py:125] [process=0] [async] Finished blocking save in 0.61 seconds. Continuing save @ gs://lance-maxtext/linen_ckpt_xpk_main_20260417_125726/linen_xpk_main_20260417_125726_16_zarr3_true/checkpoints/0. I0417 14:49:07.345863 135517953070848 async_checkpointer.py:76] [process=0][thread=async_save] Background save thread started. Deadline for this save operation is 2026-04-17 15:09:07.345829 I0417 14:49:07.356563 135517944678144 atomicity.py:140] Creating tmp directory gs://lance-maxtext/linen_ckpt_xpk_main_20260417_125726/linen_xpk_main_20260417_125726_16_zarr3_true/checkpoints/0 I0417 14:49:07.360948 135646693517120 checkpoint_manager.py:1560] [process=0][thread=MainThread][step=0] Starting CheckpointManager Save Finalize thread=save_finalize I0417 14:49:07.361208 135515277883136 async_checkpointer.py:280] [process=0][thread=save_finalize] Waiting for background save thread=async_save. I0417 14:49:07.361334 135646693517120 standard_logger.py:34] {'step': 0, 'event_type': 'save', 'directory': 'gs://lance-maxtext/linen_ckpt_xpk_main_20260417_125726/linen_xpk_main_20260417_125726_16_zarr3_true/checkpoints', 'reached_preemption': False, 'preemption_received_at': None, 'synchronous': False, 'wait_for_prev_start_time': 1776437346.7376277, 'wait_for_prev_duration_secs': 6.556510925292969e-05, 'time_between_consecutive_saves_sec': None, 'checkpointer_blocking_start_time': 1776437346.7394905, 'checkpointer_blocking_duration_secs': 0.6064724922180176, 'get_old_steps_start_time': 1776437347.3459888, 'get_old_steps_duration_secs': 3.170967102050781e-05, 'checkpoint_manager_blocking_start_time': 1776437346.735666, 'checkpoint_manager_blocking_duration_secs': 0.6256275177001953} I0417 14:49:07.361438 135646693517120 checkpointing.py:408] Started an asynchronous checkpoint save for step 0 I0417 14:49:07.361489 135646693517120 max_utils.py:750] Memstats: After params initialized: I0417 14:49:07.361540 135646693517120 max_utils.py:756] Using (GB) 0.43 / 31.25 (1.376000%) on TPU_0(process=0,(0,0,0,0)) I0417 14:49:07.361574 135646693517120 max_utils.py:756] Using (GB) 0.43 / 31.25 (1.376000%) on TPU_1(process=0,(1,0,0,0)) I0417 14:49:07.361601 135646693517120 max_utils.py:756] Using (GB) 0.43 / 31.25 (1.376000%) on TPU_4(process=0,(0,1,0,0)) I0417 14:49:07.361626 135646693517120 max_utils.py:756] Using (GB) 0.43 / 31.25 (1.376000%) on TPU_5(process=0,(1,1,0,0)) I0417 14:49:07.694880 135646693517120 metric_logger.py:196] completed step: 0, seconds: 10.988, TFLOP/s/device: 1.237, Tokens/s/device: 186.381, total_weights: 65536, loss: 10.874, lm_loss: 10.874, perplexity: 52779.875 I0417 14:49:07.697564 135646693517120 metric_logger.py:281] To see full metrics 'tensorboard --logdir=gs://lance-maxtext/linen_ckpt_xpk_main_20260417_125726/linen_xpk_main_20260417_125726_16_zarr3_true/tensorboard/' I0417 14:49:08.106380 135646693517120 metric_logger.py:196] completed step: 1, seconds: 1.112, TFLOP/s/device: 12.216, Tokens/s/device: 1841.328, total_weights: 65536, loss: 10.874, lm_loss: 10.874, perplexity: 52779.875 I0417 14:49:08.252539 135646693517120 metric_logger.py:196] completed step: 2, seconds: 0.412, TFLOP/s/device: 32.962, Tokens/s/device: 4968.353, total_weights: 65536, loss: 10.560, lm_loss: 10.560, perplexity: 38547.719 I0417 14:49:08.399905 135646693517120 metric_logger.py:196] completed step: 3, seconds: 0.009, TFLOP/s/device: 1444.210, Tokens/s/device: 217687.075, total_weights: 65536, loss: 9.980, lm_loss: 9.980, perplexity: 21590.506 I0417 14:49:08.409327 135646693517120 checkpointing.py:772] Waiting for step 5 to finish before checkpoint... I0417 14:49:08.693936 135646693517120 checkpointing.py:776] Waited 0.28444385528564453 seconds for step 5 to finish before starting checkpointing. I0417 14:49:08.696903 135646693517120 checkpoint_manager.py:2020] [process=0][thread=MainThread][step=0][wait_until_finished] Waiting for Save Finalize thread (save_finalize) to complete. I0417 14:49:09.231541 135517927892736 atomicity.py:140] Creating tmp directory gs://lance-maxtext/linen_ckpt_xpk_main_20260417_125726/linen_xpk_main_20260417_125726_16_zarr3_true/checkpoints/0/items I0417 14:49:09.352056 135515228993280 checkpoint.py:188] Wrote Metadata={'item_handlers': None, 'metrics': {}, 'performance_metrics': {}, 'init_timestamp_nsecs': 1776437348952059370, 'commit_timestamp_nsecs': None, 'custom_metadata': {}}, json={"item_handlers": null, "metrics": {}, "performance_metrics": {}, "init_timestamp_nsecs": 1776437348952059370, "commit_timestamp_nsecs": null, "custom_metadata": {}} to gs://lance-maxtext/linen_ckpt_xpk_main_20260417_125726/linen_xpk_main_20260417_125726_16_zarr3_true/checkpoints/0/_CHECKPOINT_METADATA I0417 14:49:10.400444 2785 google_auth_provider.cc:181] Running on GCE, using service account 562977990677-compute@developer.gserviceaccount.com I0417 14:49:12.149561 135517418632960 array_metadata_store.py:203] [process=0][thread=array_type_handler] Wrote 39 array_metadata.ArrayMetadata to gs://lance-maxtext/linen_ckpt_xpk_main_20260417_125726/linen_xpk_main_20260417_125726_16_zarr3_true/checkpoints/0/items/array_metadatas/process_0 I0417 14:49:47.206114 135517370824448 base_pytree_checkpoint_handler.py:1282] [process=0][thread=write_metadata_after_commits] Commit + Array metadata written. Time taken: 36.804484s (commit=35.160823s, array_metadata_write=1.643661s) I0417 14:49:47.207556 135517953070848 base_pytree_checkpoint_handler.py:130] [process=0] /jax/orbax/write/gbytes_per_sec: 39.355 MiB/s (total gbytes: 1.5 GiB) (time elapsed: 40.13682174682617 s) (per-host) I0417 14:49:47.207660 135517953070848 async_checkpointer.py:90] [process=0][thread=async_save] 3 Handler Commit operations completed. Time taken: 39.861721s. I0417 14:49:47.981476 135517953070848 checkpoint.py:228] Read Metadata={'item_handlers': None, 'metrics': {}, 'performance_metrics': {}, 'init_timestamp_nsecs': 1776437348952059370, 'commit_timestamp_nsecs': None, 'custom_metadata': {}} from gs://lance-maxtext/linen_ckpt_xpk_main_20260417_125726/linen_xpk_main_20260417_125726_16_zarr3_true/checkpoints/0/_CHECKPOINT_METADATA I0417 14:49:49.175606 135515228993280 checkpoint.py:247] Updated Metadata={'item_handlers': {'items': 'orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeCheckpointHandler'}, 'metrics': {}, 'performance_metrics': {}, 'init_timestamp_nsecs': 1776437348952059370, 'commit_timestamp_nsecs': None, 'custom_metadata': {}} to gs://lance-maxtext/linen_ckpt_xpk_main_20260417_125726/linen_xpk_main_20260417_125726_16_zarr3_true/checkpoints/0/_CHECKPOINT_METADATA I0417 14:49:49.382081 135517953070848 ocdbt_utils.py:49] Param validation support for Zarr3 will be added later (b/362328389). I0417 14:49:49.527642 135517953070848 array_metadata_store.py:411] [process=0][thread=async_save] Validated ArrayMetadata from all 8 hosts. Time taken: 0.000393s. I0417 14:49:50.000500 135517953070848 base_pytree_checkpoint_handler.py:1406] [process=0][thread=async_save] Pytree save finalize (merge_ocdbt + ArrayMetadata validation) completed. Time taken: 1.465238s. use_zarr3=True, enable_post_merge_validation=True, directory=gs://lance-maxtext/linen_ckpt_xpk_main_20260417_125726/linen_xpk_main_20260417_125726_16_zarr3_true/checkpoints/0/items I0417 14:49:50.002679 135517953070848 atomicity.py:666] Finalizing gs://lance-maxtext/linen_ckpt_xpk_main_20260417_125726/linen_xpk_main_20260417_125726_16_zarr3_true/checkpoints/0/items I0417 14:49:50.535915 135517953070848 atomicity.py:666] Finalizing gs://lance-maxtext/linen_ckpt_xpk_main_20260417_125726/linen_xpk_main_20260417_125726_16_zarr3_true/checkpoints/0 I0417 14:49:51.956215 135517953070848 atomicity.py:847] [process=0][thread=async_save] Finished saving checkpoint (finalized tmp dir) to `gs://lance-maxtext/linen_ckpt_xpk_main_20260417_125726/linen_xpk_main_20260417_125726_16_zarr3_true/checkpoints/0`. I0417 14:49:51.957355 135517953070848 event_tracking.py:138] [process=0] [async] Finished save (blocking + background) in 45.22 seconds @ gs://lance-maxtext/linen_ckpt_xpk_main_20260417_125726/linen_xpk_main_20260417_125726_16_zarr3_true/checkpoints/0 I0417 14:49:51.958833 135517953070848 async_checkpointer.py:160] [process=0][thread=async_save] Background save thread done. Time taken: 44.612891s. I0417 14:49:51.959020 135515277883136 async_checkpointer.py:288] [process=0][thread=save_finalize] Done with waiting for background save thread=async_save. I0417 14:49:51.959124 135515277883136 async_checkpointer.py:298] [process=0][thread=save_finalize] No errors found in background save thread=async_save. I0417 14:49:51.959223 135515277883136 checkpoint_manager.py:2137] [process=0][thread=save_finalize][step=0] CheckpointManager Save Finalize is syncing with other hosts... I0417 14:49:51.960952 135515277883136 checkpoint_manager.py:2146] [process=0][thread=save_finalize][step=0] CheckpointManager Save Finalize is done on all hosts. I0417 14:49:51.961299 135646693517120 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 14:49:51.961414 135646693517120 checkpoint_manager.py:1452] Waiting for previous save to complete took 43.264522 seconds. If this number is high, consider checkpointing less frequently. I0417 14:49:51.963334 135646693517120 checkpoint_manager.py:1512] [process=0] Saving checkpoint at step 5 I0417 14:49:51.965753 135646693517120 event_tracking.py:70] [process=0] [async] Started save checkpoint @ gs://lance-maxtext/linen_ckpt_xpk_main_20260417_125726/linen_xpk_main_20260417_125726_16_zarr3_true/checkpoints/5. I0417 14:49:52.682570 135646693517120 jax_array_handlers.py:360] Scheduling D2H of 39 prioritized jax.Array. I0417 14:49:52.682705 135646693517120 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 14:49:52.720913 135646693517120 base_pytree_checkpoint_handler.py:154] [process=0][thread=MainThread] Initiated "orbax.checkpoint._src.serialization.jax_array_handlers.ArrayHandler".serialize. Time taken: 0.039339s I0417 14:49:52.721812 135646693517120 base_pytree_checkpoint_handler.py:130] [process=0] /jax/orbax/write/blocking_gbytes_per_sec: 32.311 GiB/s (total gbytes: 1.5 GiB) (time elapsed: 0.04774117469787598 s) (per-host) I0417 14:49:52.721937 135646693517120 base_pytree_checkpoint_handler.py:768] [process=0][thread=MainThread] Initiated Pytree async_save. Time taken: 0.047891s (batch_requests_ready=0.003960s, total_serialization_initiated=0.043135s, others=0.000796s) I0417 14:49:52.722071 135646693517120 composite_checkpoint_handler.py:715] [process=0][thread=MainThread] Initiated CompositeCheckpointHandler.async_save. Time taken: 0.052779s (all_items=0.000025s, per_item={'items': '0.00002527'}, temp_paths=0.052754) I0417 14:49:52.723035 135646693517120 event_tracking.py:125] [process=0] [async] Finished blocking save in 0.76 seconds. Continuing save @ gs://lance-maxtext/linen_ckpt_xpk_main_20260417_125726/linen_xpk_main_20260417_125726_16_zarr3_true/checkpoints/5. I0417 14:49:52.723451 135515277883136 async_checkpointer.py:76] [process=0][thread=async_save] Background save thread started. Deadline for this save operation is 2026-04-17 15:09:52.723410 I0417 14:49:52.725785 135646693517120 checkpoint_manager.py:1560] [process=0][thread=MainThread][step=5] Starting CheckpointManager Save Finalize thread=save_finalize I0417 14:49:52.726180 135516333717248 async_checkpointer.py:280] [process=0][thread=save_finalize] Waiting for background save thread=async_save. I0417 14:49:52.726429 135646693517120 standard_logger.py:34] {'step': 5, 'event_type': 'save', 'directory': 'gs://lance-maxtext/linen_ckpt_xpk_main_20260417_125726/linen_xpk_main_20260417_125726_16_zarr3_true/checkpoints', 'reached_preemption': False, 'preemption_received_at': None, 'synchronous': False, 'wait_for_prev_start_time': 1776437348.6968565, 'wait_for_prev_duration_secs': 43.2645218372345, 'time_between_consecutive_saves_sec': None, 'checkpointer_blocking_start_time': 1776437391.9634504, 'checkpointer_blocking_duration_secs': 0.7601926326751709, 'get_old_steps_start_time': 1776437392.7236698, 'get_old_steps_duration_secs': 4.029273986816406e-05, 'checkpoint_manager_blocking_start_time': 1776437348.6945968, 'checkpoint_manager_blocking_duration_secs': 44.031779527664185} I0417 14:49:52.726574 135646693517120 checkpointing.py:408] Started an asynchronous checkpoint save for step 5 I0417 14:49:52.916018 135517953070848 atomicity.py:140] Creating tmp directory gs://lance-maxtext/linen_ckpt_xpk_main_20260417_125726/linen_xpk_main_20260417_125726_16_zarr3_true/checkpoints/5 I0417 14:49:54.643575 135517391795968 atomicity.py:140] Creating tmp directory gs://lance-maxtext/linen_ckpt_xpk_main_20260417_125726/linen_xpk_main_20260417_125726_16_zarr3_true/checkpoints/5/items I0417 14:49:57.927236 135516350502656 array_metadata_store.py:203] [process=0][thread=array_type_handler] Wrote 39 array_metadata.ArrayMetadata to gs://lance-maxtext/linen_ckpt_xpk_main_20260417_125726/linen_xpk_main_20260417_125726_16_zarr3_true/checkpoints/5/items/array_metadatas/process_0 I0417 14:50:15.111409 135646693517120 metric_logger.py:196] completed step: 4, seconds: 0.148, TFLOP/s/device: 92.077, Tokens/s/device: 13878.818, total_weights: 65536, loss: 9.460, lm_loss: 9.460, perplexity: 12840.355 I0417 14:50:15.120565 135646693517120 metric_logger.py:196] completed step: 5, seconds: 0.147, TFLOP/s/device: 92.423, Tokens/s/device: 13931.025, total_weights: 65536, loss: 8.959, lm_loss: 8.959, perplexity: 7776.467 I0417 14:50:15.266169 135646693517120 metric_logger.py:196] completed step: 6, seconds: 66.710, TFLOP/s/device: 0.204, Tokens/s/device: 30.700, total_weights: 65536, loss: 8.469, lm_loss: 8.469, perplexity: 4762.740 I0417 14:50:15.413482 135646693517120 metric_logger.py:196] completed step: 7, seconds: 0.007, TFLOP/s/device: 1905.094, Tokens/s/device: 287156.478, total_weights: 65536, loss: 8.003, lm_loss: 8.003, perplexity: 2989.151 I0417 14:50:15.560795 135646693517120 metric_logger.py:196] completed step: 8, seconds: 0.147, TFLOP/s/device: 92.696, Tokens/s/device: 13972.083, total_weights: 65536, loss: 7.572, lm_loss: 7.572, perplexity: 1943.098 I0417 14:50:15.707353 135646693517120 checkpointing.py:772] Waiting for step 9 to finish before checkpoint... I0417 14:50:15.708119 135646693517120 checkpointing.py:776] Waited 0.0007855892181396484 seconds for step 9 to finish before starting checkpointing. I0417 14:50:15.710171 135646693517120 checkpoint_manager.py:2020] [process=0][thread=MainThread][step=5][wait_until_finished] Waiting for Save Finalize thread (save_finalize) to complete. I0417 14:50:36.527317 135516342109952 base_pytree_checkpoint_handler.py:1282] [process=0][thread=write_metadata_after_commits] Commit + Array metadata written. Time taken: 40.765060s (commit=39.590306s, array_metadata_write=1.174754s) I0417 14:50:36.528363 135515277883136 base_pytree_checkpoint_handler.py:130] [process=0] /jax/orbax/write/gbytes_per_sec: 36.019 MiB/s (total gbytes: 1.5 GiB) (time elapsed: 43.85429859161377 s) (per-host) I0417 14:50:36.528448 135515277883136 async_checkpointer.py:90] [process=0][thread=async_save] 3 Handler Commit operations completed. Time taken: 43.804848s. I0417 14:50:38.138416 135515277883136 ocdbt_utils.py:49] Param validation support for Zarr3 will be added later (b/362328389). I0417 14:50:38.241651 135515277883136 array_metadata_store.py:411] [process=0][thread=async_save] Validated ArrayMetadata from all 8 hosts. Time taken: 0.000232s. I0417 14:50:38.742458 135515277883136 base_pytree_checkpoint_handler.py:1406] [process=0][thread=async_save] Pytree save finalize (merge_ocdbt + ArrayMetadata validation) completed. Time taken: 1.417035s. use_zarr3=True, enable_post_merge_validation=True, directory=gs://lance-maxtext/linen_ckpt_xpk_main_20260417_125726/linen_xpk_main_20260417_125726_16_zarr3_true/checkpoints/5/items I0417 14:50:38.744201 135515277883136 atomicity.py:666] Finalizing gs://lance-maxtext/linen_ckpt_xpk_main_20260417_125726/linen_xpk_main_20260417_125726_16_zarr3_true/checkpoints/5/items I0417 14:50:39.300571 135515277883136 atomicity.py:666] Finalizing gs://lance-maxtext/linen_ckpt_xpk_main_20260417_125726/linen_xpk_main_20260417_125726_16_zarr3_true/checkpoints/5 I0417 14:50:40.666207 135515277883136 atomicity.py:847] [process=0][thread=async_save] Finished saving checkpoint (finalized tmp dir) to `gs://lance-maxtext/linen_ckpt_xpk_main_20260417_125726/linen_xpk_main_20260417_125726_16_zarr3_true/checkpoints/5`. I0417 14:50:40.667047 135515277883136 event_tracking.py:138] [process=0] [async] Finished save (blocking + background) in 48.70 seconds @ gs://lance-maxtext/linen_ckpt_xpk_main_20260417_125726/linen_xpk_main_20260417_125726_16_zarr3_true/checkpoints/5 I0417 14:50:40.668587 135515277883136 async_checkpointer.py:160] [process=0][thread=async_save] Background save thread done. Time taken: 47.944986s. I0417 14:50:40.668792 135516333717248 async_checkpointer.py:288] [process=0][thread=save_finalize] Done with waiting for background save thread=async_save. I0417 14:50:40.668944 135516333717248 async_checkpointer.py:298] [process=0][thread=save_finalize] No errors found in background save thread=async_save. I0417 14:50:40.669013 135516333717248 checkpoint_manager.py:2137] [process=0][thread=save_finalize][step=5] CheckpointManager Save Finalize is syncing with other hosts... I0417 14:50:40.673213 135516333717248 checkpoint_manager.py:2146] [process=0][thread=save_finalize][step=5] CheckpointManager Save Finalize is done on all hosts. I0417 14:50:40.673372 135646693517120 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 14:50:40.673508 135646693517120 checkpoint_manager.py:1452] Waiting for previous save to complete took 24.963342 seconds. If this number is high, consider checkpointing less frequently. I0417 14:50:40.674848 135646693517120 checkpoint_manager.py:1512] [process=0] Saving checkpoint at step 9 I0417 14:50:40.677053 135646693517120 event_tracking.py:70] [process=0] [async] Started save checkpoint @ gs://lance-maxtext/linen_ckpt_xpk_main_20260417_125726/linen_xpk_main_20260417_125726_16_zarr3_true/checkpoints/9. I0417 14:50:40.979181 135646693517120 jax_array_handlers.py:360] Scheduling D2H of 39 prioritized jax.Array. I0417 14:50:40.979274 135646693517120 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 14:50:41.005177 135646693517120 base_pytree_checkpoint_handler.py:154] [process=0][thread=MainThread] Initiated "orbax.checkpoint._src.serialization.jax_array_handlers.ArrayHandler".serialize. Time taken: 0.026934s I0417 14:50:41.006007 135646693517120 base_pytree_checkpoint_handler.py:130] [process=0] /jax/orbax/write/blocking_gbytes_per_sec: 49.311 GiB/s (total gbytes: 1.5 GiB) (time elapsed: 0.031282663345336914 s) (per-host) I0417 14:50:41.006076 135646693517120 base_pytree_checkpoint_handler.py:768] [process=0][thread=MainThread] Initiated Pytree async_save. Time taken: 0.031368s (batch_requests_ready=0.001781s, total_serialization_initiated=0.028802s, others=0.000786s) I0417 14:50:41.006152 135646693517120 composite_checkpoint_handler.py:715] [process=0][thread=MainThread] Initiated CompositeCheckpointHandler.async_save. Time taken: 0.035304s (all_items=0.000015s, per_item={'items': '0.00001502'}, temp_paths=0.035289) I0417 14:50:41.006797 135646693517120 event_tracking.py:125] [process=0] [async] Finished blocking save in 0.33 seconds. Continuing save @ gs://lance-maxtext/linen_ckpt_xpk_main_20260417_125726/linen_xpk_main_20260417_125726_16_zarr3_true/checkpoints/9. I0417 14:50:41.007126 135517408593664 async_checkpointer.py:76] [process=0][thread=async_save] Background save thread started. Deadline for this save operation is 2026-04-17 15:10:41.007082 I0417 14:50:41.016411 135646693517120 checkpoint_manager.py:1560] [process=0][thread=MainThread][step=9] Starting CheckpointManager Save Finalize thread=save_finalize I0417 14:50:41.016770 135516333717248 async_checkpointer.py:280] [process=0][thread=save_finalize] Waiting for background save thread=async_save. I0417 14:50:41.016927 135646693517120 standard_logger.py:34] {'step': 9, 'event_type': 'save', 'directory': 'gs://lance-maxtext/linen_ckpt_xpk_main_20260417_125726/linen_xpk_main_20260417_125726_16_zarr3_true/checkpoints', 'reached_preemption': False, 'preemption_received_at': None, 'synchronous': False, 'wait_for_prev_start_time': 1776437415.7101367, 'wait_for_prev_duration_secs': 24.96334171295166, 'time_between_consecutive_saves_sec': 23.749117612838745, 'checkpointer_blocking_start_time': 1776437440.6748874, 'checkpointer_blocking_duration_secs': 0.33237719535827637, 'get_old_steps_start_time': 1776437441.0072842, 'get_old_steps_duration_secs': 2.6941299438476562e-05, 'checkpoint_manager_blocking_start_time': 1776437415.7083468, 'checkpoint_manager_blocking_duration_secs': 25.308546543121338} I0417 14:50:41.017086 135646693517120 checkpointing.py:408] Started an asynchronous checkpoint save for step 9 I0417 14:50:41.017130 135646693517120 checkpoint_manager.py:2020] [process=0][thread=MainThread][step=9][wait_until_finished] Waiting for Save Finalize thread (save_finalize) to complete. I0417 14:50:41.218430 135515277883136 atomicity.py:140] Creating tmp directory gs://lance-maxtext/linen_ckpt_xpk_main_20260417_125726/linen_xpk_main_20260417_125726_16_zarr3_true/checkpoints/9 I0417 14:50:43.425488 135516350502656 atomicity.py:140] Creating tmp directory gs://lance-maxtext/linen_ckpt_xpk_main_20260417_125726/linen_xpk_main_20260417_125726_16_zarr3_true/checkpoints/9/items I0417 14:50:46.731572 135517953070848 array_metadata_store.py:203] [process=0][thread=array_type_handler] Wrote 39 array_metadata.ArrayMetadata to gs://lance-maxtext/linen_ckpt_xpk_main_20260417_125726/linen_xpk_main_20260417_125726_16_zarr3_true/checkpoints/9/items/array_metadatas/process_0 I0417 14:51:24.356894 135517418632960 base_pytree_checkpoint_handler.py:1282] [process=0][thread=write_metadata_after_commits] Commit + Array metadata written. Time taken: 39.805597s (commit=38.616496s, array_metadata_write=1.189101s) I0417 14:51:24.363371 135517408593664 base_pytree_checkpoint_handler.py:130] [process=0] /jax/orbax/write/gbytes_per_sec: 36.406 MiB/s (total gbytes: 1.5 GiB) (time elapsed: 43.38864207267761 s) (per-host) I0417 14:51:24.363436 135517408593664 async_checkpointer.py:90] [process=0][thread=async_save] 3 Handler Commit operations completed. Time taken: 43.356198s. I0417 14:51:25.832599 135517408593664 array_metadata_store.py:411] [process=0][thread=async_save] Validated ArrayMetadata from all 8 hosts. Time taken: 0.000247s. I0417 14:51:26.013870 135517408593664 ocdbt_utils.py:49] Param validation support for Zarr3 will be added later (b/362328389). I0417 14:51:26.630009 135517408593664 base_pytree_checkpoint_handler.py:1406] [process=0][thread=async_save] Pytree save finalize (merge_ocdbt + ArrayMetadata validation) completed. Time taken: 1.419015s. use_zarr3=True, enable_post_merge_validation=True, directory=gs://lance-maxtext/linen_ckpt_xpk_main_20260417_125726/linen_xpk_main_20260417_125726_16_zarr3_true/checkpoints/9/items I0417 14:51:26.631744 135517408593664 atomicity.py:666] Finalizing gs://lance-maxtext/linen_ckpt_xpk_main_20260417_125726/linen_xpk_main_20260417_125726_16_zarr3_true/checkpoints/9/items I0417 14:51:27.180728 135517408593664 atomicity.py:666] Finalizing gs://lance-maxtext/linen_ckpt_xpk_main_20260417_125726/linen_xpk_main_20260417_125726_16_zarr3_true/checkpoints/9 I0417 14:51:28.596229 135517408593664 atomicity.py:847] [process=0][thread=async_save] Finished saving checkpoint (finalized tmp dir) to `gs://lance-maxtext/linen_ckpt_xpk_main_20260417_125726/linen_xpk_main_20260417_125726_16_zarr3_true/checkpoints/9`. I0417 14:51:28.597040 135517408593664 event_tracking.py:138] [process=0] [async] Finished save (blocking + background) in 47.92 seconds @ gs://lance-maxtext/linen_ckpt_xpk_main_20260417_125726/linen_xpk_main_20260417_125726_16_zarr3_true/checkpoints/9 I0417 14:51:28.598172 135517408593664 async_checkpointer.py:160] [process=0][thread=async_save] Background save thread done. Time taken: 47.590930s. I0417 14:51:28.598377 135516333717248 async_checkpointer.py:288] [process=0][thread=save_finalize] Done with waiting for background save thread=async_save. I0417 14:51:28.598540 135516333717248 async_checkpointer.py:298] [process=0][thread=save_finalize] No errors found in background save thread=async_save. I0417 14:51:28.598619 135516333717248 checkpoint_manager.py:2137] [process=0][thread=save_finalize][step=9] CheckpointManager Save Finalize is syncing with other hosts... I0417 14:51:28.600346 135516333717248 checkpoint_manager.py:2146] [process=0][thread=save_finalize][step=9] CheckpointManager Save Finalize is done on all hosts. I0417 14:51:28.600559 135646693517120 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 14:51:28.600713 135646693517120 checkpoint_manager.py:2009] [process=0][thread=MainThread][wait_until_finished] No Save Finalize thread to wait for. Returning. I0417 14:51:28.601703 135646693517120 metric_logger.py:196] completed step: 9, seconds: 0.147, TFLOP/s/device: 92.195, Tokens/s/device: 13896.617, 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: Fri Apr 17 14:51:41 UTC 2026 EXIT_CODE=0
XPK Start: Fri Apr 17 04:17:57 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 04:18:21.380954: 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 04:18:21.560209 134953940940608 max_utils.py:273] Attempting to initialize the jax distributed system... I0417 04:18:30.600489 134953940940608 distributed.py:149] Starting JAX distributed service on [::]:8482 I0417 04:18:30.603259 134953940940608 distributed.py:172] Connecting to JAX distributed service on mt-16-zarr3-true-sv-s1ua0-slice-job-0-0.mt-16-zarr3-true-sv-s1ua0:8482 I0417 04:18:31.709997 134953940940608 max_utils.py:284] Jax distributed system initialized! I0417 04:18:37.981010 134953940940608 max_utils.py:800] System Information: Jax Version: 0.9.2 I0417 04:18:37.981111 134953940940608 max_utils.py:801] System Information: Jaxlib Version: 0.9.2 I0417 04:18:37.981150 134953940940608 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 04:18:37.981183 134953940940608 train_utils.py:377] WARNING: Sequence packing is essentially ignored for synthetic data. Please use a real dataset to use sequence packing. I0417 04:18:38.648082 134953940940608 maxtext_utils.py:1687] Num_devices: 32, shape (1, 1, 1, 32, 1, 1, 1, 1, 1, 1, 1, 1, 1) I0417 04:18:38.648652 134953940940608 maxtext_utils.py:1687] Num_devices: 32, shape (1, 1, 1, 32, 1, 1, 1, 1, 1, 1, 1, 1, 1) I0417 04:18:38.648823 134953940940608 checkpointing.py:688] Setting up checkpoint logger... I0417 04:18:38.648892 134953940940608 checkpointing.py:234] Creating checkpoint manager with ocdbt=True and zarr3=True I0417 04:18:38.648935 134953940940608 pytree_checkpoint_handler.py:592] save_device_host_concurrent_bytes=None I0417 04:18:38.649253 134953940940608 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 0x7abcb1e50140>, enable_pinned_host_transfer=False, save_concurrent_bytes: 96000000000 (89.4 GiB), restore_concurrent_bytes: 96000000000 (89.4 GiB) I0417 04:18:41.541877 134953940940608 checkpointing.py:266] Enabling policy for fixed interval checkpointing. I0417 04:18:41.542219 134953940940608 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 0x7aa7b8538560>}, handler_registry=None I0417 04:18:41.542458 134953940940608 composite_checkpoint_handler.py:237] Deferred registration for item: "items". Adding handler `<orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeCheckpointHandler object at 0x7aa7b8538560>` 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 04:18:41.542507 134953940940608 composite_checkpoint_handler.py:237] Deferred registration for item: "metrics". Adding handler `<orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonCheckpointHandler object at 0x7aa7b85397c0>` 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 04:18:41.542541 134953940940608 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 0x7aa7b8538560>, ('items', <class 'orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeRestoreArgs'>): <orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeCheckpointHandler object at 0x7aa7b8538560>, ('metrics', <class 'orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonSaveArgs'>): <orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonCheckpointHandler object at 0x7aa7b85397c0>, ('metrics', <class 'orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonRestoreArgs'>): <orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonCheckpointHandler object at 0x7aa7b85397c0>}). I0417 04:18:41.542864 134953940940608 abstract_checkpointer.py:35] orbax-checkpoint version: 0.11.34 I0417 04:18:41.542935 134953940940608 async_checkpointer.py:192] [process=0][thread=MainThread] Using barrier_sync_fn: <function get_barrier_sync_fn.<locals>._fn at 0x7aa7b81b07c0> timeout: 1200 secs and primary_host=0 for async checkpoint writes I0417 04:18:43.140955 134953940940608 checkpoint_manager.py:1812] Found 0 checkpoint steps in gs://lance-maxtext/linen_ckpt_xpk_feat_nnx_post_train_fixes_20260417_023710/linen_xpk_feat_nnx_post_train_fixes_20260417_023710_16_zarr3_true/checkpoints I0417 04:18:43.588275 134953940940608 checkpoint_manager.py:929] [process=0][thread=MainThread] CheckpointManager created, primary_host=0, CheckpointManagerOptions=CheckpointManagerOptions(save_interval_steps=1, max_to_keep=None, keep_time_interval=None, keep_period=None, should_keep_fn=None, best_fn=None, best_mode='max', keep_checkpoints_without_metrics=True, step_prefix=None, step_format_fixed_length=None, step_name_format=None, create=True, cleanup_tmp_directories=False, save_on_steps=frozenset(), single_host_load_and_broadcast=False, todelete_subdir=None, todelete_full_path=None, enable_background_delete=False, read_only=False, enable_async_checkpointing=True, async_options=None, multiprocessing_options=MultiprocessingOptions(primary_host=0, active_processes=None, barrier_sync_key_prefix=None), should_save_fn=None, file_options=FileOptions(path_permission_mode=None), save_root_metadata=True, temporary_path_class=None, save_decision_policy=FixedIntervalPolicy(interval=5), preservation_policy=LatestN(n=None), prevent_write_metrics=False, enable_should_save_is_saving_in_progress_check=True, enable_per_process_directory_creation=False, lightweight_initialize=False), root_directory=gs://lance-maxtext/linen_ckpt_xpk_feat_nnx_post_train_fixes_20260417_023710/linen_xpk_feat_nnx_post_train_fixes_20260417_023710_16_zarr3_true/checkpoints: <orbax.checkpoint.checkpoint_manager.CheckpointManager object at 0x7aa7b853bec0> I0417 04:18:43.588446 134953940940608 checkpointing.py:302] Checkpoint manager created! I0417 04:18:44.507232 134953940940608 nnx_wrappers.py:437] Unknown Logical: bfloat16[32,2048,2048]...................................... ('activation_batch', 'activation_norm_length', 'activation_embed'). I0417 04:18:44.507345 134953940940608 nnx_wrappers.py:437] Unknown Physical: bfloat16[32,2048,2048]...................................... ('fsdp', None, None). I0417 04:18:44.889958 134953940940608 attentions.py:1088] attentions/inputs_q Logical: bfloat16[32,2048,2048]...................................... ('activation_batch', 'activation_attn_length', 'activation_attn_embed'). I0417 04:18:44.890044 134953940940608 attentions.py:1088] attentions/inputs_q Physical: bfloat16[32,2048,2048]...................................... ('fsdp', None, None). I0417 04:18:44.906148 134953940940608 attentions.py:1089] attentions/inputs_kv Logical: bfloat16[32,2048,2048]...................................... ('activation_batch', 'activation_attn_length', 'activation_attn_embed'). I0417 04:18:44.906208 134953940940608 attentions.py:1089] attentions/inputs_kv Physical: bfloat16[32,2048,2048]...................................... ('fsdp', None, None). I0417 04:18:44.929735 134953940940608 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 04:18:44.929800 134953940940608 attentions.py:1154] attentions/query Physical: bfloat16[32,2048,16,128].................................... ('fsdp', None, None, None). I0417 04:18:44.946014 134953940940608 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 04:18:44.946072 134953940940608 attentions.py:1155] attentions/key Physical: bfloat16[32,2048,16,128].................................... ('fsdp', None, None, None). I0417 04:18:44.962193 134953940940608 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 04:18:44.962251 134953940940608 attentions.py:1156] attentions/value Physical: bfloat16[32,2048,16,128].................................... ('fsdp', None, None, None). I0417 04:18:44.986942 134953940940608 attentions.py:1197] attentions/out Logical: bfloat16[32,2048,16,128].................................... ('activation_batch', 'activation_attn_length', 'activation_heads', 'activation_kv'). I0417 04:18:44.987011 134953940940608 attentions.py:1197] attentions/out Physical: bfloat16[32,2048,16,128].................................... ('fsdp', None, None, None). I0417 04:18:45.008297 134953940940608 linears.py:525] linears/x Logical: bfloat16[32,2048,7168]...................................... ('activation_batch', 'activation_length', 'activation_mlp'). I0417 04:18:45.008360 134953940940608 linears.py:525] linears/x Physical: bfloat16[32,2048,7168]...................................... ('fsdp', None, None). I0417 04:18:45.224237 134953940940608 checkpointing.py:578] checkpoint manager exists so trying to load this run's existing checkpoint I0417 04:18:45.224343 134953940940608 checkpointing.py:676] No existing checkpoints found, not restoring checkpoint. fsdp: 32 I0417 04:18:46.646650 134953940940608 maxtext_utils.py:1796] params/params/decoder/decoder_norm/scale Shape: float32[2048] Logical: P('norm',) Physical: (None,) I0417 04:18:46.646771 134953940940608 maxtext_utils.py:1796] params/params/decoder/layers/mlp/wi_0/kernel Shape: float32[2048,16,7168] Logical: P('embed', 'layers', 'mlp') Physical: ('fsdp', None, None) I0417 04:18:46.646825 134953940940608 maxtext_utils.py:1796] params/params/decoder/layers/mlp/wi_1/kernel Shape: float32[2048,16,7168] Logical: P('embed', 'layers', 'mlp') Physical: ('fsdp', None, None) I0417 04:18:46.646928 134953940940608 maxtext_utils.py:1796] params/params/decoder/layers/mlp/wo/kernel Shape: float32[7168,16,2048] Logical: P('mlp', 'layers', 'embed') Physical: (None, None, 'fsdp') I0417 04:18:46.646983 134953940940608 maxtext_utils.py:1796] params/params/decoder/layers/post_self_attention_layer_norm/scale Shape: float32[2048,16] Logical: P('norm', 'layers') Physical: (None, None) I0417 04:18:46.647023 134953940940608 maxtext_utils.py:1796] params/params/decoder/layers/pre_self_attention_layer_norm/scale Shape: float32[2048,16] Logical: P('norm', 'layers') Physical: (None, None) I0417 04:18:46.647081 134953940940608 maxtext_utils.py:1796] 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) I0417 04:18:46.647134 134953940940608 maxtext_utils.py:1796] 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') I0417 04:18:46.647175 134953940940608 maxtext_utils.py:1796] 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) I0417 04:18:46.647212 134953940940608 maxtext_utils.py:1796] 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) I0417 04:18:46.647259 134953940940608 maxtext_utils.py:1796] params/params/decoder/logits_dense/kernel Shape: float32[2048,32000] Logical: P('embed_vocab', 'vocab') Physical: ('fsdp', None) I0417 04:18:46.647306 134953940940608 maxtext_utils.py:1796] params/params/token_embedder/embedding Shape: float32[32000,2048] Logical: P('vocab', 'embed_vocab') Physical: (None, 'fsdp') I0417 04:18:47.147100 134953940940608 train.py:157] train/xent Logical: float32[32,2048]............................................ ('activation_embed_and_logits_batch', 'activation_length'). I0417 04:18:47.147193 134953940940608 train.py:157] train/xent Physical: float32[32,2048]............................................ ('fsdp', None). I0417 04:18:47.162975 134953940940608 train.py:164] train/z_loss Logical: float32[32,2048]............................................ ('activation_embed_and_logits_batch', 'activation_length'). I0417 04:18:47.163042 134953940940608 train.py:164] train/z_loss Physical: float32[32,2048]............................................ ('fsdp', None). I0417 04:18:58.010583 134953940940608 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 04:18:58.018178 134953940940608 metric_logger.py:289] number parameters: 1.104 billion I0417 04:19:09.551908 134953940940608 checkpointing.py:794] Waiting for step 0 to finish before checkpoint... I0417 04:19:09.704763 134953940940608 checkpointing.py:798] Waited 0.15283203125 seconds for step 0 to finish before starting checkpointing. I0417 04:19:09.707373 134953940940608 checkpoint_manager.py:2009] [process=0][thread=MainThread][wait_until_finished] No Save Finalize thread to wait for. Returning. I0417 04:19:09.709109 134953940940608 checkpoint_manager.py:1512] [process=0] Saving checkpoint at step 0 I0417 04:19:09.710544 134953940940608 event_tracking.py:70] [process=0] [async] Started save checkpoint @ gs://lance-maxtext/linen_ckpt_xpk_feat_nnx_post_train_fixes_20260417_023710/linen_xpk_feat_nnx_post_train_fixes_20260417_023710_16_zarr3_true/checkpoints/0. I0417 04:19:10.443251 134953940940608 signaling_client.py:364] Using JaxDistributedSignalingClient I0417 04:19:10.457731 134953940940608 jax_array_handlers.py:360] Scheduling D2H of 39 prioritized jax.Array. I0417 04:19:10.457795 134953940940608 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 04:19:10.710257 134825013561088 atomicity.py:140] Creating tmp directory gs://lance-maxtext/linen_ckpt_xpk_feat_nnx_post_train_fixes_20260417_023710/linen_xpk_feat_nnx_post_train_fixes_20260417_023710_16_zarr3_true/checkpoints/0 I0417 04:19:10.725322 134953940940608 base_pytree_checkpoint_handler.py:154] [process=0][thread=MainThread] Initiated "orbax.checkpoint._src.serialization.jax_array_handlers.ArrayHandler".serialize. Time taken: 0.268339s I0417 04:19:10.726172 134953940940608 base_pytree_checkpoint_handler.py:130] [process=0] /jax/orbax/write/blocking_gbytes_per_sec: 5.617 GiB/s (total gbytes: 1.5 GiB) (time elapsed: 0.2746288776397705 s) (per-host) I0417 04:19:10.726244 134953940940608 base_pytree_checkpoint_handler.py:768] [process=0][thread=MainThread] Initiated Pytree async_save. Time taken: 0.274721s (batch_requests_ready=0.002061s, total_serialization_initiated=0.271895s, others=0.000766s) I0417 04:19:10.726324 134953940940608 composite_checkpoint_handler.py:715] [process=0][thread=MainThread] Initiated CompositeCheckpointHandler.async_save. Time taken: 0.278800s (all_items=0.000019s, per_item={'items': '0.00001884'}, temp_paths=0.278782) I0417 04:19:10.727152 134953940940608 event_tracking.py:125] [process=0] [async] Finished blocking save in 1.02 seconds. Continuing save @ gs://lance-maxtext/linen_ckpt_xpk_feat_nnx_post_train_fixes_20260417_023710/linen_xpk_feat_nnx_post_train_fixes_20260417_023710_16_zarr3_true/checkpoints/0. I0417 04:19:10.727437 134825021953792 async_checkpointer.py:76] [process=0][thread=async_save] Background save thread started. Deadline for this save operation is 2026-04-17 04:39:10.727406 I0417 04:19:10.738995 134953940940608 checkpoint_manager.py:1560] [process=0][thread=MainThread][step=0] Starting CheckpointManager Save Finalize thread=save_finalize I0417 04:19:10.739303 134820768249600 async_checkpointer.py:280] [process=0][thread=save_finalize] Waiting for background save thread=async_save. I0417 04:19:10.739457 134953940940608 standard_logger.py:34] {'step': 0, 'event_type': 'save', 'directory': 'gs://lance-maxtext/linen_ckpt_xpk_feat_nnx_post_train_fixes_20260417_023710/linen_xpk_feat_nnx_post_train_fixes_20260417_023710_16_zarr3_true/checkpoints', 'reached_preemption': False, 'preemption_received_at': None, 'synchronous': False, 'wait_for_prev_start_time': 1776399549.707348, 'wait_for_prev_duration_secs': 8.630752563476562e-05, 'time_between_consecutive_saves_sec': None, 'checkpointer_blocking_start_time': 1776399549.7091608, 'checkpointer_blocking_duration_secs': 1.0184144973754883, 'get_old_steps_start_time': 1776399550.7276003, 'get_old_steps_duration_secs': 2.8133392333984375e-05, 'checkpoint_manager_blocking_start_time': 1776399549.7053108, 'checkpoint_manager_blocking_duration_secs': 1.034109115600586} I0417 04:19:10.739558 134953940940608 checkpointing.py:409] Started an asynchronous checkpoint save for step 0 I0417 04:19:10.739609 134953940940608 max_utils.py:750] Memstats: After params initialized: I0417 04:19:10.739661 134953940940608 max_utils.py:756] Using (GB) 0.43 / 31.25 (1.376000%) on TPU_0(process=0,(0,0,0,0)) I0417 04:19:10.739694 134953940940608 max_utils.py:756] Using (GB) 0.43 / 31.25 (1.376000%) on TPU_1(process=0,(1,0,0,0)) I0417 04:19:10.739722 134953940940608 max_utils.py:756] Using (GB) 0.43 / 31.25 (1.376000%) on TPU_4(process=0,(0,1,0,0)) I0417 04:19:10.739747 134953940940608 max_utils.py:756] Using (GB) 0.43 / 31.25 (1.376000%) on TPU_5(process=0,(1,1,0,0)) I0417 04:19:11.074721 134953940940608 metric_logger.py:185] completed step: 0, seconds: 11.236, TFLOP/s/device: 1.209, Tokens/s/device: 182.264, total_weights: 65536, loss: 10.874 I0417 04:19:11.077628 134953940940608 metric_logger.py:269] To see full metrics 'tensorboard --logdir=gs://lance-maxtext/linen_ckpt_xpk_feat_nnx_post_train_fixes_20260417_023710/linen_xpk_feat_nnx_post_train_fixes_20260417_023710_16_zarr3_true/tensorboard/' I0417 04:19:11.478586 134953940940608 metric_logger.py:185] completed step: 1, seconds: 1.521, TFLOP/s/device: 8.931, Tokens/s/device: 1346.241, total_weights: 65536, loss: 10.874 I0417 04:19:11.624608 134953940940608 metric_logger.py:185] completed step: 2, seconds: 0.404, TFLOP/s/device: 33.590, Tokens/s/device: 5063.053, total_weights: 65536, loss: 10.560 I0417 04:19:11.772049 134953940940608 metric_logger.py:185] completed step: 3, seconds: 0.008, TFLOP/s/device: 1620.409, Tokens/s/device: 244245.677, total_weights: 65536, loss: 9.980 I0417 04:19:11.780607 134953940940608 checkpointing.py:794] Waiting for step 5 to finish before checkpoint... I0417 04:19:12.066245 134953940940608 checkpointing.py:798] Waited 0.28550195693969727 seconds for step 5 to finish before starting checkpointing. I0417 04:19:12.069122 134953940940608 checkpoint_manager.py:2020] [process=0][thread=MainThread][step=0][wait_until_finished] Waiting for Save Finalize thread (save_finalize) to complete. I0417 04:19:12.483056 134824492734208 atomicity.py:140] Creating tmp directory gs://lance-maxtext/linen_ckpt_xpk_feat_nnx_post_train_fixes_20260417_023710/linen_xpk_feat_nnx_post_train_fixes_20260417_023710_16_zarr3_true/checkpoints/0/items I0417 04:19:12.639800 134820722415360 checkpoint.py:188] Wrote Metadata={'item_handlers': None, 'metrics': {}, 'performance_metrics': {}, 'init_timestamp_nsecs': 1776399552196413019, 'commit_timestamp_nsecs': None, 'custom_metadata': {}}, json={"item_handlers": null, "metrics": {}, "performance_metrics": {}, "init_timestamp_nsecs": 1776399552196413019, "commit_timestamp_nsecs": null, "custom_metadata": {}} to gs://lance-maxtext/linen_ckpt_xpk_feat_nnx_post_train_fixes_20260417_023710/linen_xpk_feat_nnx_post_train_fixes_20260417_023710_16_zarr3_true/checkpoints/0/_CHECKPOINT_METADATA I0417 04:19:14.053525 2858 google_auth_provider.cc:181] Running on GCE, using service account 562977990677-compute@developer.gserviceaccount.com I0417 04:19:15.793899 134823401379584 array_metadata_store.py:203] [process=0][thread=array_type_handler] Wrote 39 array_metadata.ArrayMetadata to gs://lance-maxtext/linen_ckpt_xpk_feat_nnx_post_train_fixes_20260417_023710/linen_xpk_feat_nnx_post_train_fixes_20260417_023710_16_zarr3_true/checkpoints/0/items/array_metadatas/process_0 I0417 04:19:50.722587 134822857676544 base_pytree_checkpoint_handler.py:1282] [process=0][thread=write_metadata_after_commits] Commit + Array metadata written. Time taken: 36.670080s (commit=35.451739s, array_metadata_write=1.218341s) I0417 04:19:50.724091 134825021953792 base_pytree_checkpoint_handler.py:130] [process=0] /jax/orbax/write/gbytes_per_sec: 39.223 MiB/s (total gbytes: 1.5 GiB) (time elapsed: 40.27249622344971 s) (per-host) I0417 04:19:50.724266 134825021953792 async_checkpointer.py:90] [process=0][thread=async_save] 3 Handler Commit operations completed. Time taken: 39.996717s. I0417 04:19:51.443038 134825021953792 checkpoint.py:228] Read Metadata={'item_handlers': None, 'metrics': {}, 'performance_metrics': {}, 'init_timestamp_nsecs': 1776399552196413019, 'commit_timestamp_nsecs': None, 'custom_metadata': {}} from gs://lance-maxtext/linen_ckpt_xpk_feat_nnx_post_train_fixes_20260417_023710/linen_xpk_feat_nnx_post_train_fixes_20260417_023710_16_zarr3_true/checkpoints/0/_CHECKPOINT_METADATA I0417 04:19:52.179680 134825021953792 array_metadata_store.py:411] [process=0][thread=async_save] Validated ArrayMetadata from all 8 hosts. Time taken: 0.000380s. I0417 04:19:52.219493 134820722415360 checkpoint.py:247] Updated Metadata={'item_handlers': {'items': 'orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeCheckpointHandler'}, 'metrics': {}, 'performance_metrics': {}, 'init_timestamp_nsecs': 1776399552196413019, 'commit_timestamp_nsecs': None, 'custom_metadata': {}} to gs://lance-maxtext/linen_ckpt_xpk_feat_nnx_post_train_fixes_20260417_023710/linen_xpk_feat_nnx_post_train_fixes_20260417_023710_16_zarr3_true/checkpoints/0/_CHECKPOINT_METADATA I0417 04:19:52.853630 134825021953792 ocdbt_utils.py:49] Param validation support for Zarr3 will be added later (b/362328389). I0417 04:19:53.774957 134825021953792 base_pytree_checkpoint_handler.py:1406] [process=0][thread=async_save] Pytree save finalize (merge_ocdbt + ArrayMetadata validation) completed. Time taken: 2.207155s. use_zarr3=True, enable_post_merge_validation=True, directory=gs://lance-maxtext/linen_ckpt_xpk_feat_nnx_post_train_fixes_20260417_023710/linen_xpk_feat_nnx_post_train_fixes_20260417_023710_16_zarr3_true/checkpoints/0/items I0417 04:19:53.777251 134825021953792 atomicity.py:666] Finalizing gs://lance-maxtext/linen_ckpt_xpk_feat_nnx_post_train_fixes_20260417_023710/linen_xpk_feat_nnx_post_train_fixes_20260417_023710_16_zarr3_true/checkpoints/0/items I0417 04:19:54.330404 134825021953792 atomicity.py:666] Finalizing gs://lance-maxtext/linen_ckpt_xpk_feat_nnx_post_train_fixes_20260417_023710/linen_xpk_feat_nnx_post_train_fixes_20260417_023710_16_zarr3_true/checkpoints/0 I0417 04:19:55.770750 134825021953792 atomicity.py:847] [process=0][thread=async_save] Finished saving checkpoint (finalized tmp dir) to `gs://lance-maxtext/linen_ckpt_xpk_feat_nnx_post_train_fixes_20260417_023710/linen_xpk_feat_nnx_post_train_fixes_20260417_023710_16_zarr3_true/checkpoints/0`. I0417 04:19:55.771921 134825021953792 event_tracking.py:138] [process=0] [async] Finished save (blocking + background) in 46.06 seconds @ gs://lance-maxtext/linen_ckpt_xpk_feat_nnx_post_train_fixes_20260417_023710/linen_xpk_feat_nnx_post_train_fixes_20260417_023710_16_zarr3_true/checkpoints/0 I0417 04:19:55.773633 134825021953792 async_checkpointer.py:160] [process=0][thread=async_save] Background save thread done. Time taken: 45.046081s. I0417 04:19:55.773945 134820768249600 async_checkpointer.py:288] [process=0][thread=save_finalize] Done with waiting for background save thread=async_save. I0417 04:19:55.774103 134820768249600 async_checkpointer.py:298] [process=0][thread=save_finalize] No errors found in background save thread=async_save. I0417 04:19:55.774211 134820768249600 checkpoint_manager.py:2137] [process=0][thread=save_finalize][step=0] CheckpointManager Save Finalize is syncing with other hosts... I0417 04:19:55.776289 134820768249600 checkpoint_manager.py:2146] [process=0][thread=save_finalize][step=0] CheckpointManager Save Finalize is done on all hosts. I0417 04:19:55.776534 134953940940608 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 04:19:55.776659 134953940940608 checkpoint_manager.py:1452] Waiting for previous save to complete took 43.707546 seconds. If this number is high, consider checkpointing less frequently. I0417 04:19:55.778879 134953940940608 checkpoint_manager.py:1512] [process=0] Saving checkpoint at step 5 I0417 04:19:55.781331 134953940940608 event_tracking.py:70] [process=0] [async] Started save checkpoint @ gs://lance-maxtext/linen_ckpt_xpk_feat_nnx_post_train_fixes_20260417_023710/linen_xpk_feat_nnx_post_train_fixes_20260417_023710_16_zarr3_true/checkpoints/5. I0417 04:19:56.480813 134953940940608 jax_array_handlers.py:360] Scheduling D2H of 39 prioritized jax.Array. I0417 04:19:56.480963 134953940940608 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 04:19:56.525895 134953940940608 base_pytree_checkpoint_handler.py:154] [process=0][thread=MainThread] Initiated "orbax.checkpoint._src.serialization.jax_array_handlers.ArrayHandler".serialize. Time taken: 0.046040s I0417 04:19:56.526806 134953940940608 base_pytree_checkpoint_handler.py:130] [process=0] /jax/orbax/write/blocking_gbytes_per_sec: 28.236 GiB/s (total gbytes: 1.5 GiB) (time elapsed: 0.054631710052490234 s) (per-host) I0417 04:19:56.526946 134953940940608 base_pytree_checkpoint_handler.py:768] [process=0][thread=MainThread] Initiated Pytree async_save. Time taken: 0.054795s (batch_requests_ready=0.004168s, total_serialization_initiated=0.049787s, others=0.000840s) I0417 04:19:56.527079 134953940940608 composite_checkpoint_handler.py:715] [process=0][thread=MainThread] Initiated CompositeCheckpointHandler.async_save. Time taken: 0.059769s (all_items=0.000024s, per_item={'items': '0.00002384'}, temp_paths=0.059745) I0417 04:19:56.528736 134953940940608 event_tracking.py:125] [process=0] [async] Finished blocking save in 0.75 seconds. Continuing save @ gs://lance-maxtext/linen_ckpt_xpk_feat_nnx_post_train_fixes_20260417_023710/linen_xpk_feat_nnx_post_train_fixes_20260417_023710_16_zarr3_true/checkpoints/5. I0417 04:19:56.529325 134820768249600 async_checkpointer.py:76] [process=0][thread=async_save] Background save thread started. Deadline for this save operation is 2026-04-17 04:39:56.529274 I0417 04:19:56.531922 134953940940608 checkpoint_manager.py:1560] [process=0][thread=MainThread][step=5] Starting CheckpointManager Save Finalize thread=save_finalize I0417 04:19:56.532330 134823947511552 async_checkpointer.py:280] [process=0][thread=save_finalize] Waiting for background save thread=async_save. I0417 04:19:56.532571 134953940940608 standard_logger.py:34] {'step': 5, 'event_type': 'save', 'directory': 'gs://lance-maxtext/linen_ckpt_xpk_feat_nnx_post_train_fixes_20260417_023710/linen_xpk_feat_nnx_post_train_fixes_20260417_023710_16_zarr3_true/checkpoints', 'reached_preemption': False, 'preemption_received_at': None, 'synchronous': False, 'wait_for_prev_start_time': 1776399552.0690775, 'wait_for_prev_duration_secs': 43.7075457572937, 'time_between_consecutive_saves_sec': None, 'checkpointer_blocking_start_time': 1776399595.7789493, 'checkpointer_blocking_duration_secs': 0.7506067752838135, 'get_old_steps_start_time': 1776399596.529585, 'get_old_steps_duration_secs': 4.410743713378906e-05, 'checkpoint_manager_blocking_start_time': 1776399552.0669155, 'checkpoint_manager_blocking_duration_secs': 44.465598821640015} I0417 04:19:56.532720 134953940940608 checkpointing.py:409] Started an asynchronous checkpoint save for step 5 I0417 04:19:56.716248 134825021953792 atomicity.py:140] Creating tmp directory gs://lance-maxtext/linen_ckpt_xpk_feat_nnx_post_train_fixes_20260417_023710/linen_xpk_feat_nnx_post_train_fixes_20260417_023710_16_zarr3_true/checkpoints/5 I0417 04:19:58.447684 134823964296960 atomicity.py:140] Creating tmp directory gs://lance-maxtext/linen_ckpt_xpk_feat_nnx_post_train_fixes_20260417_023710/linen_xpk_feat_nnx_post_train_fixes_20260417_023710_16_zarr3_true/checkpoints/5/items I0417 04:20:01.733717 134823981082368 array_metadata_store.py:203] [process=0][thread=array_type_handler] Wrote 39 array_metadata.ArrayMetadata to gs://lance-maxtext/linen_ckpt_xpk_feat_nnx_post_train_fixes_20260417_023710/linen_xpk_feat_nnx_post_train_fixes_20260417_023710_16_zarr3_true/checkpoints/5/items/array_metadatas/process_0 I0417 04:20:18.814260 134953940940608 metric_logger.py:185] completed step: 4, seconds: 0.147, TFLOP/s/device: 92.404, Tokens/s/device: 13928.183, total_weights: 65536, loss: 9.460 I0417 04:20:18.822926 134953940940608 metric_logger.py:185] completed step: 5, seconds: 0.147, TFLOP/s/device: 92.197, Tokens/s/device: 13896.994, total_weights: 65536, loss: 8.959 I0417 04:20:18.968376 134953940940608 metric_logger.py:185] completed step: 6, seconds: 67.041, TFLOP/s/device: 0.203, Tokens/s/device: 30.548, total_weights: 65536, loss: 8.469 I0417 04:20:19.115825 134953940940608 metric_logger.py:185] completed step: 7, seconds: 0.007, TFLOP/s/device: 2061.154, Tokens/s/device: 310679.612, total_weights: 65536, loss: 8.003 I0417 04:20:19.263142 134953940940608 metric_logger.py:185] completed step: 8, seconds: 0.147, TFLOP/s/device: 92.571, Tokens/s/device: 13953.235, total_weights: 65536, loss: 7.572 I0417 04:20:19.410280 134953940940608 checkpointing.py:794] Waiting for step 9 to finish before checkpoint... I0417 04:20:19.410995 134953940940608 checkpointing.py:798] Waited 0.0007345676422119141 seconds for step 9 to finish before starting checkpointing. I0417 04:20:19.412981 134953940940608 checkpoint_manager.py:2020] [process=0][thread=MainThread][step=5][wait_until_finished] Waiting for Save Finalize thread (save_finalize) to complete. I0417 04:20:40.937920 134823972689664 base_pytree_checkpoint_handler.py:1282] [process=0][thread=write_metadata_after_commits] Commit + Array metadata written. Time taken: 41.385438s (commit=40.172310s, array_metadata_write=1.213128s) I0417 04:20:40.939100 134820768249600 base_pytree_checkpoint_handler.py:130] [process=0] /jax/orbax/write/gbytes_per_sec: 35.523 MiB/s (total gbytes: 1.5 GiB) (time elapsed: 44.46690487861633 s) (per-host) I0417 04:20:40.939226 134820768249600 async_checkpointer.py:90] [process=0][thread=async_save] 3 Handler Commit operations completed. Time taken: 44.409720s. I0417 04:20:42.628788 134820768249600 array_metadata_store.py:411] [process=0][thread=async_save] Validated ArrayMetadata from all 8 hosts. Time taken: 0.000232s. I0417 04:20:42.990187 134820768249600 ocdbt_utils.py:49] Param validation support for Zarr3 will be added later (b/362328389). I0417 04:20:43.920508 134820768249600 base_pytree_checkpoint_handler.py:1406] [process=0][thread=async_save] Pytree save finalize (merge_ocdbt + ArrayMetadata validation) completed. Time taken: 2.184048s. use_zarr3=True, enable_post_merge_validation=True, directory=gs://lance-maxtext/linen_ckpt_xpk_feat_nnx_post_train_fixes_20260417_023710/linen_xpk_feat_nnx_post_train_fixes_20260417_023710_16_zarr3_true/checkpoints/5/items I0417 04:20:43.921991 134820768249600 atomicity.py:666] Finalizing gs://lance-maxtext/linen_ckpt_xpk_feat_nnx_post_train_fixes_20260417_023710/linen_xpk_feat_nnx_post_train_fixes_20260417_023710_16_zarr3_true/checkpoints/5/items I0417 04:20:44.513382 134820768249600 atomicity.py:666] Finalizing gs://lance-maxtext/linen_ckpt_xpk_feat_nnx_post_train_fixes_20260417_023710/linen_xpk_feat_nnx_post_train_fixes_20260417_023710_16_zarr3_true/checkpoints/5 I0417 04:20:45.933741 134820768249600 atomicity.py:847] [process=0][thread=async_save] Finished saving checkpoint (finalized tmp dir) to `gs://lance-maxtext/linen_ckpt_xpk_feat_nnx_post_train_fixes_20260417_023710/linen_xpk_feat_nnx_post_train_fixes_20260417_023710_16_zarr3_true/checkpoints/5`. I0417 04:20:45.934468 134820768249600 event_tracking.py:138] [process=0] [async] Finished save (blocking + background) in 50.16 seconds @ gs://lance-maxtext/linen_ckpt_xpk_feat_nnx_post_train_fixes_20260417_023710/linen_xpk_feat_nnx_post_train_fixes_20260417_023710_16_zarr3_true/checkpoints/5 I0417 04:20:45.935851 134820768249600 async_checkpointer.py:160] [process=0][thread=async_save] Background save thread done. Time taken: 49.406332s. I0417 04:20:45.936033 134823947511552 async_checkpointer.py:288] [process=0][thread=save_finalize] Done with waiting for background save thread=async_save. I0417 04:20:45.936150 134823947511552 async_checkpointer.py:298] [process=0][thread=save_finalize] No errors found in background save thread=async_save. I0417 04:20:45.936227 134823947511552 checkpoint_manager.py:2137] [process=0][thread=save_finalize][step=5] CheckpointManager Save Finalize is syncing with other hosts... I0417 04:20:45.938865 134823947511552 checkpoint_manager.py:2146] [process=0][thread=save_finalize][step=5] CheckpointManager Save Finalize is done on all hosts. I0417 04:20:45.939033 134953940940608 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 04:20:45.939163 134953940940608 checkpoint_manager.py:1452] Waiting for previous save to complete took 26.526187 seconds. If this number is high, consider checkpointing less frequently. I0417 04:20:45.940777 134953940940608 checkpoint_manager.py:1512] [process=0] Saving checkpoint at step 9 I0417 04:20:45.942885 134953940940608 event_tracking.py:70] [process=0] [async] Started save checkpoint @ gs://lance-maxtext/linen_ckpt_xpk_feat_nnx_post_train_fixes_20260417_023710/linen_xpk_feat_nnx_post_train_fixes_20260417_023710_16_zarr3_true/checkpoints/9. I0417 04:20:46.646104 134953940940608 jax_array_handlers.py:360] Scheduling D2H of 39 prioritized jax.Array. I0417 04:20:46.646199 134953940940608 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 04:20:46.671694 134953940940608 base_pytree_checkpoint_handler.py:154] [process=0][thread=MainThread] Initiated "orbax.checkpoint._src.serialization.jax_array_handlers.ArrayHandler".serialize. Time taken: 0.026481s I0417 04:20:46.672459 134953940940608 base_pytree_checkpoint_handler.py:130] [process=0] /jax/orbax/write/blocking_gbytes_per_sec: 50.077 GiB/s (total gbytes: 1.5 GiB) (time elapsed: 0.030803918838500977 s) (per-host) I0417 04:20:46.672523 134953940940608 base_pytree_checkpoint_handler.py:768] [process=0][thread=MainThread] Initiated Pytree async_save. Time taken: 0.030887s (batch_requests_ready=0.001819s, total_serialization_initiated=0.028349s, others=0.000719s) I0417 04:20:46.672595 134953940940608 composite_checkpoint_handler.py:715] [process=0][thread=MainThread] Initiated CompositeCheckpointHandler.async_save. Time taken: 0.034947s (all_items=0.000017s, per_item={'items': '0.00001669'}, temp_paths=0.034930) I0417 04:20:46.673260 134953940940608 event_tracking.py:125] [process=0] [async] Finished blocking save in 0.73 seconds. Continuing save @ gs://lance-maxtext/linen_ckpt_xpk_feat_nnx_post_train_fixes_20260417_023710/linen_xpk_feat_nnx_post_train_fixes_20260417_023710_16_zarr3_true/checkpoints/9. I0417 04:20:46.673553 134823444211456 async_checkpointer.py:76] [process=0][thread=async_save] Background save thread started. Deadline for this save operation is 2026-04-17 04:40:46.673513 I0417 04:20:46.687464 134953940940608 checkpoint_manager.py:1560] [process=0][thread=MainThread][step=9] Starting CheckpointManager Save Finalize thread=save_finalize I0417 04:20:46.687720 134823947511552 async_checkpointer.py:280] [process=0][thread=save_finalize] Waiting for background save thread=async_save. I0417 04:20:46.687870 134953940940608 standard_logger.py:34] {'step': 9, 'event_type': 'save', 'directory': 'gs://lance-maxtext/linen_ckpt_xpk_feat_nnx_post_train_fixes_20260417_023710/linen_xpk_feat_nnx_post_train_fixes_20260417_023710_16_zarr3_true/checkpoints', 'reached_preemption': False, 'preemption_received_at': None, 'synchronous': False, 'wait_for_prev_start_time': 1776399619.4129488, 'wait_for_prev_duration_secs': 26.526187419891357, 'time_between_consecutive_saves_sec': 23.636590719223022, 'checkpointer_blocking_start_time': 1776399645.9408164, 'checkpointer_blocking_duration_secs': 0.7328839302062988, 'get_old_steps_start_time': 1776399646.6737196, 'get_old_steps_duration_secs': 2.6464462280273438e-05, 'checkpoint_manager_blocking_start_time': 1776399619.4112222, 'checkpoint_manager_blocking_duration_secs': 27.276604175567627} I0417 04:20:46.688002 134953940940608 checkpointing.py:409] Started an asynchronous checkpoint save for step 9 I0417 04:20:46.688054 134953940940608 checkpoint_manager.py:2020] [process=0][thread=MainThread][step=9][wait_until_finished] Waiting for Save Finalize thread (save_finalize) to complete. I0417 04:20:46.880758 134820768249600 atomicity.py:140] Creating tmp directory gs://lance-maxtext/linen_ckpt_xpk_feat_nnx_post_train_fixes_20260417_023710/linen_xpk_feat_nnx_post_train_fixes_20260417_023710_16_zarr3_true/checkpoints/9 I0417 04:20:48.600916 134825021953792 atomicity.py:140] Creating tmp directory gs://lance-maxtext/linen_ckpt_xpk_feat_nnx_post_train_fixes_20260417_023710/linen_xpk_feat_nnx_post_train_fixes_20260417_023710_16_zarr3_true/checkpoints/9/items I0417 04:20:52.285887 134823978252032 array_metadata_store.py:203] [process=0][thread=array_type_handler] Wrote 39 array_metadata.ArrayMetadata to gs://lance-maxtext/linen_ckpt_xpk_feat_nnx_post_train_fixes_20260417_023710/linen_xpk_feat_nnx_post_train_fixes_20260417_023710_16_zarr3_true/checkpoints/9/items/array_metadatas/process_0 I0417 04:21:31.491495 134823969859328 base_pytree_checkpoint_handler.py:1282] [process=0][thread=write_metadata_after_commits] Commit + Array metadata written. Time taken: 41.359541s (commit=40.176889s, array_metadata_write=1.182652s) I0417 04:21:31.493806 134823444211456 base_pytree_checkpoint_handler.py:130] [process=0] /jax/orbax/write/gbytes_per_sec: 35.218 MiB/s (total gbytes: 1.5 GiB) (time elapsed: 44.852121353149414 s) (per-host) I0417 04:21:31.493946 134823444211456 async_checkpointer.py:90] [process=0][thread=async_save] 3 Handler Commit operations completed. Time taken: 44.820271s. I0417 04:21:33.269132 134823444211456 array_metadata_store.py:411] [process=0][thread=async_save] Validated ArrayMetadata from all 8 hosts. Time taken: 0.000242s. I0417 04:21:33.638570 134823444211456 ocdbt_utils.py:49] Param validation support for Zarr3 will be added later (b/362328389). I0417 04:21:34.568174 134823444211456 base_pytree_checkpoint_handler.py:1406] [process=0][thread=async_save] Pytree save finalize (merge_ocdbt + ArrayMetadata validation) completed. Time taken: 2.196440s. use_zarr3=True, enable_post_merge_validation=True, directory=gs://lance-maxtext/linen_ckpt_xpk_feat_nnx_post_train_fixes_20260417_023710/linen_xpk_feat_nnx_post_train_fixes_20260417_023710_16_zarr3_true/checkpoints/9/items I0417 04:21:34.569801 134823444211456 atomicity.py:666] Finalizing gs://lance-maxtext/linen_ckpt_xpk_feat_nnx_post_train_fixes_20260417_023710/linen_xpk_feat_nnx_post_train_fixes_20260417_023710_16_zarr3_true/checkpoints/9/items I0417 04:21:35.096067 134823444211456 atomicity.py:666] Finalizing gs://lance-maxtext/linen_ckpt_xpk_feat_nnx_post_train_fixes_20260417_023710/linen_xpk_feat_nnx_post_train_fixes_20260417_023710_16_zarr3_true/checkpoints/9 I0417 04:21:36.521773 134823444211456 atomicity.py:847] [process=0][thread=async_save] Finished saving checkpoint (finalized tmp dir) to `gs://lance-maxtext/linen_ckpt_xpk_feat_nnx_post_train_fixes_20260417_023710/linen_xpk_feat_nnx_post_train_fixes_20260417_023710_16_zarr3_true/checkpoints/9`. I0417 04:21:36.522610 134823444211456 event_tracking.py:138] [process=0] [async] Finished save (blocking + background) in 50.58 seconds @ gs://lance-maxtext/linen_ckpt_xpk_feat_nnx_post_train_fixes_20260417_023710/linen_xpk_feat_nnx_post_train_fixes_20260417_023710_16_zarr3_true/checkpoints/9 I0417 04:21:36.524428 134823444211456 async_checkpointer.py:160] [process=0][thread=async_save] Background save thread done. Time taken: 49.850750s. I0417 04:21:36.524555 134823947511552 async_checkpointer.py:288] [process=0][thread=save_finalize] Done with waiting for background save thread=async_save. I0417 04:21:36.524621 134823947511552 async_checkpointer.py:298] [process=0][thread=save_finalize] No errors found in background save thread=async_save. I0417 04:21:36.524684 134823947511552 checkpoint_manager.py:2137] [process=0][thread=save_finalize][step=9] CheckpointManager Save Finalize is syncing with other hosts... I0417 04:21:36.528148 134823947511552 checkpoint_manager.py:2146] [process=0][thread=save_finalize][step=9] CheckpointManager Save Finalize is done on all hosts. I0417 04:21:36.528328 134953940940608 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 04:21:36.528473 134953940940608 checkpoint_manager.py:2009] [process=0][thread=MainThread][wait_until_finished] No Save Finalize thread to wait for. Returning. I0417 04:21:36.529135 134953940940608 metric_logger.py:185] completed step: 9, seconds: 0.147, TFLOP/s/device: 92.220, Tokens/s/device: 13900.390, 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: Fri Apr 17 04:21:48 UTC 2026 EXIT_CODE=0