main (8a17c3d19) vs test/pipeline-scan-nnx (5b21b66fc)| Metric | main 8a17c3d19 | test/pipeline-scan-nnx 5b21b66fc | Diff (test/pipeline-scan-nnx − main) |
|---|---|---|---|
| Parameters | 1.104 billion | 1.104 billion | — |
| Final loss | 5.3910 | 5.3910 | 0 |
| TFLOP/s | 92.223 | 92.295 | +0.072 |
| Tok/s | 13900.9 | 13911.7 | +10.859 |
| Avg s/step | 8.490 | 8.544 | +0.054 |
| 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: Wed Apr 22 08:35:38 UTC 2026 PyTorch was not found. Models won't be available and only tokenizers, configuration and file/data utilities can be used. `rope_parameters`'s factor field must be a float >= 1, got 40 `rope_parameters`'s beta_fast field must be a float, got 32 `rope_parameters`'s beta_slow field must be a float, got 1 DeepseekV32Config got `key=rope_scaling` in kwargs but hasn't set it as attribute. For RoPE standardization you need to set `self.rope_parameters` in model's config. 2026-04-22 08:36:02.852799: 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) I0422 08:36:03.067228 140454833280832 max_utils.py:273] Attempting to initialize the jax distributed system... I0422 08:36:12.109440 140454833280832 distributed.py:149] Starting JAX distributed service on [::]:8482 I0422 08:36:12.114198 140454833280832 distributed.py:172] Connecting to JAX distributed service on mt-16-zarr3-true-rs-nfneu-slice-job-0-0.mt-16-zarr3-true-rs-nfneu:8482 I0422 08:36:13.364203 140454833280832 max_utils.py:284] Jax distributed system initialized! I0422 08:36:19.400674 140454833280832 max_utils.py:800] System Information: Jax Version: 0.9.2 I0422 08:36:19.400786 140454833280832 max_utils.py:801] System Information: Jaxlib Version: 0.9.2 I0422 08:36:19.400827 140454833280832 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 I0422 08:36:19.400863 140454833280832 train_utils.py:361] WARNING: Sequence packing is essentially ignored for synthetic data. Please use a real dataset to use sequence packing. I0422 08:36:20.096029 140454833280832 maxtext_utils.py:1565] Num_devices: 32, shape (1, 1, 1, 32, 1, 1, 1, 1, 1, 1, 1, 1, 1) I0422 08:36:20.096328 140454833280832 checkpointing.py:677] Setting up checkpoint logger... I0422 08:36:20.096384 140454833280832 checkpointing.py:233] Creating checkpoint manager with ocdbt=True and zarr3=True I0422 08:36:20.096428 140454833280832 pytree_checkpoint_handler.py:592] save_device_host_concurrent_bytes=None I0422 08:36:20.096790 140454833280832 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 0x7fbd8d91cfe0>, enable_pinned_host_transfer=False, save_concurrent_bytes: 96000000000 (89.4 GiB), restore_concurrent_bytes: 96000000000 (89.4 GiB) I0422 08:36:21.415061 140454833280832 checkpointing.py:265] Enabling policy for fixed interval checkpointing. I0422 08:36:21.415321 140454833280832 checkpoint_manager.py:708] [process=6][thread=MainThread] CheckpointManager init: checkpointers=None, item_names=('items',), item_handlers={'items': <orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeCheckpointHandler object at 0x7faa08304980>}, handler_registry=None I0422 08:36:21.415632 140454833280832 composite_checkpoint_handler.py:237] Deferred registration for item: "items". Adding handler `<orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeCheckpointHandler object at 0x7faa08304980>` 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`. I0422 08:36:21.415691 140454833280832 composite_checkpoint_handler.py:237] Deferred registration for item: "metrics". Adding handler `<orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonCheckpointHandler object at 0x7fbc6c1a7ec0>` 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`. I0422 08:36:21.415729 140454833280832 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 0x7faa08304980>, ('items', <class 'orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeRestoreArgs'>): <orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeCheckpointHandler object at 0x7faa08304980>, ('metrics', <class 'orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonSaveArgs'>): <orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonCheckpointHandler object at 0x7fbc6c1a7ec0>, ('metrics', <class 'orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonRestoreArgs'>): <orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonCheckpointHandler object at 0x7fbc6c1a7ec0>}). I0422 08:36:21.416053 140454833280832 abstract_checkpointer.py:35] orbax-checkpoint version: 0.11.34 I0422 08:36:21.416141 140454833280832 async_checkpointer.py:192] [process=6][thread=MainThread] Using barrier_sync_fn: <function get_barrier_sync_fn.<locals>._fn at 0x7fa978609580> timeout: 1200 secs and primary_host=0 for async checkpoint writes I0422 08:36:23.344017 140337225910016 checkpoint.py:228] Read Metadata={'item_handlers': {'items': 'orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeCheckpointHandler'}, 'metrics': {}, 'performance_metrics': {}, 'init_timestamp_nsecs': 1776846817358508401, 'commit_timestamp_nsecs': 1776846865133181729, 'custom_metadata': {}} from gs://lance-maxtext/linen_ckpt_xpk_main_20260422_071422/linen_xpk_main_20260422_071422_16_zarr3_true/checkpoints/9/_CHECKPOINT_METADATA I0422 08:36:23.586657 140337225910016 checkpoint_manager.py:1871] Missing metrics for step 9 I0422 08:36:24.411033 140454833280832 checkpoint_manager.py:1812] Found 3 checkpoint steps in gs://lance-maxtext/linen_ckpt_xpk_main_20260422_071422/linen_xpk_main_20260422_071422_16_zarr3_true/checkpoints I0422 08:36:24.459780 140454833280832 checkpoint_manager.py:929] [process=6][thread=MainThread] CheckpointManager created, primary_host=0, CheckpointManagerOptions=CheckpointManagerOptions(save_interval_steps=1, max_to_keep=None, keep_time_interval=None, keep_period=None, should_keep_fn=None, best_fn=None, best_mode='max', keep_checkpoints_without_metrics=True, step_prefix=None, step_format_fixed_length=None, step_name_format=None, create=True, cleanup_tmp_directories=False, save_on_steps=frozenset(), single_host_load_and_broadcast=False, todelete_subdir=None, todelete_full_path=None, enable_background_delete=False, read_only=False, enable_async_checkpointing=True, async_options=None, multiprocessing_options=MultiprocessingOptions(primary_host=0, active_processes=None, barrier_sync_key_prefix=None), should_save_fn=None, file_options=FileOptions(path_permission_mode=None), save_root_metadata=True, temporary_path_class=None, save_decision_policy=FixedIntervalPolicy(interval=5), preservation_policy=LatestN(n=None), prevent_write_metrics=False, enable_should_save_is_saving_in_progress_check=True, enable_per_process_directory_creation=False, lightweight_initialize=False), root_directory=gs://lance-maxtext/linen_ckpt_xpk_main_20260422_071422/linen_xpk_main_20260422_071422_16_zarr3_true/checkpoints: <orbax.checkpoint.checkpoint_manager.CheckpointManager object at 0x7fbc6c12d430> I0422 08:36:24.459931 140454833280832 checkpointing.py:301] Checkpoint manager created! I0422 08:36:25.421374 140454833280832 nnx_wrappers.py:437] Unknown Logical: bfloat16[32,2048,2048]...................................... ('activation_batch', 'activation_norm_length', 'activation_embed'). I0422 08:36:25.421497 140454833280832 nnx_wrappers.py:437] Unknown Physical: bfloat16[32,2048,2048]...................................... ('fsdp', None, None). I0422 08:36:25.800858 140454833280832 attentions.py:1088] attentions/inputs_q Logical: bfloat16[32,2048,2048]...................................... ('activation_batch', 'activation_attn_length', 'activation_attn_embed'). I0422 08:36:25.800953 140454833280832 attentions.py:1088] attentions/inputs_q Physical: bfloat16[32,2048,2048]...................................... ('fsdp', None, None). I0422 08:36:25.817328 140454833280832 attentions.py:1089] attentions/inputs_kv Logical: bfloat16[32,2048,2048]...................................... ('activation_batch', 'activation_attn_length', 'activation_attn_embed'). I0422 08:36:25.817386 140454833280832 attentions.py:1089] attentions/inputs_kv Physical: bfloat16[32,2048,2048]...................................... ('fsdp', None, None). I0422 08:36:25.841190 140454833280832 attentions.py:1154] attentions/query Logical: bfloat16[32,2048,16,128].................................... ('activation_kv_batch', 'activation_attn_length', 'activation_kv_heads', 'activation_kv_head_dim'). I0422 08:36:25.841262 140454833280832 attentions.py:1154] attentions/query Physical: bfloat16[32,2048,16,128].................................... ('fsdp', None, None, None). I0422 08:36:25.858016 140454833280832 attentions.py:1155] attentions/key Logical: bfloat16[32,2048,16,128].................................... ('activation_kv_batch', 'activation_attn_length', 'activation_kv_heads', 'activation_kv_head_dim'). I0422 08:36:25.858080 140454833280832 attentions.py:1155] attentions/key Physical: bfloat16[32,2048,16,128].................................... ('fsdp', None, None, None). I0422 08:36:25.874496 140454833280832 attentions.py:1156] attentions/value Logical: bfloat16[32,2048,16,128].................................... ('activation_kv_batch', 'activation_attn_length', 'activation_kv_heads', 'activation_kv_head_dim'). I0422 08:36:25.874558 140454833280832 attentions.py:1156] attentions/value Physical: bfloat16[32,2048,16,128].................................... ('fsdp', None, None, None). I0422 08:36:25.899599 140454833280832 attentions.py:1198] attentions/out Logical: bfloat16[32,2048,16,128].................................... ('activation_batch', 'activation_attn_length', 'activation_heads', 'activation_kv'). I0422 08:36:25.899669 140454833280832 attentions.py:1198] attentions/out Physical: bfloat16[32,2048,16,128].................................... ('fsdp', None, None, None). I0422 08:36:25.920881 140454833280832 linears.py:525] linears/x Logical: bfloat16[32,2048,7168]...................................... ('activation_batch', 'activation_length', 'activation_mlp'). I0422 08:36:25.920956 140454833280832 linears.py:525] linears/x Physical: bfloat16[32,2048,7168]...................................... ('fsdp', None, None). I0422 08:36:26.133347 140454833280832 checkpointing.py:577] checkpoint manager exists so trying to load this run's existing checkpoint I0422 08:36:26.133456 140454833280832 checkpointing.py:581] restoring from this run's directory step 9 I0422 08:36:26.385611 140454833280832 event_tracking.py:70] [process=6] [sync] Started load checkpoint @ gs://lance-maxtext/linen_ckpt_xpk_main_20260422_071422/linen_xpk_main_20260422_071422_16_zarr3_true/checkpoints/9. I0422 08:36:26.629319 140454833280832 checkpointer.py:307] Restoring checkpoint from gs://lance-maxtext/linen_ckpt_xpk_main_20260422_071422/linen_xpk_main_20260422_071422_16_zarr3_true/checkpoints/9. I0422 08:36:26.629781 140454833280832 event_tracking.py:125] [process=6] [sync] Finished blocking load in 0.24 seconds. Continuing load @ gs://lance-maxtext/linen_ckpt_xpk_main_20260422_071422/linen_xpk_main_20260422_071422_16_zarr3_true/checkpoints/9. I0422 08:36:28.031943 1410 google_auth_provider.cc:181] Running on GCE, using service account 562977990677-compute@developer.gserviceaccount.com I0422 08:36:35.144937 140454833280832 jax_array_handlers.py:843] [process=6] /jax/orbax/read/worker/io/requested throughput: 221.976 MiB/s (total gbytes: 1.5 GiB) (time elapsed: 7.119133472442627 s) (per-host) I0422 08:36:35.150531 140454833280832 base_pytree_checkpoint_handler.py:130] [process=6] /jax/checkpoint/read/gbytes_per_sec: 210.340 MiB/s (total gbytes: 1.5 GiB) (time elapsed: 7.523974418640137 s) (per-host) I0422 08:36:35.769075 140454833280832 event_tracking.py:138] [process=6] [sync] Finished load in 9.38 seconds @ gs://lance-maxtext/linen_ckpt_xpk_main_20260422_071422/linen_xpk_main_20260422_071422_16_zarr3_true/checkpoints/9 I0422 08:36:35.769268 140454833280832 standard_logger.py:34] {'step': 9, 'event_type': 'restore', 'directory': 'gs://lance-maxtext/linen_ckpt_xpk_main_20260422_071422/linen_xpk_main_20260422_071422_16_zarr3_true/checkpoints', 'checkpointer_start_time': 1776846986.385504, 'checkpointer_duration_secs': 9.383731365203857, 'checkpoint_manager_start_time': 1776846986.1336992, 'checkpoint_manager_duration_secs': 9.635538578033447} fsdp: 32 I0422 08:36:36.190516 140454833280832 maxtext_utils.py:1668] params/params/decoder/decoder_norm/scale Shape: float32[2048] Logical: P('norm',) Physical: (None,) I0422 08:36:36.190659 140454833280832 maxtext_utils.py:1668] params/params/decoder/layers/mlp/wi_0/kernel Shape: float32[2048,16,7168] Logical: P('embed', 'layers', 'mlp') Physical: ('fsdp', None, None) I0422 08:36:36.190714 140454833280832 maxtext_utils.py:1668] params/params/decoder/layers/mlp/wi_1/kernel Shape: float32[2048,16,7168] Logical: P('embed', 'layers', 'mlp') Physical: ('fsdp', None, None) I0422 08:36:36.190773 140454833280832 maxtext_utils.py:1668] params/params/decoder/layers/mlp/wo/kernel Shape: float32[7168,16,2048] Logical: P('mlp', 'layers', 'embed') Physical: (None, None, 'fsdp') I0422 08:36:36.190827 140454833280832 maxtext_utils.py:1668] params/params/decoder/layers/post_self_attention_layer_norm/scale Shape: float32[2048,16] Logical: P('norm', 'layers') Physical: (None, None) I0422 08:36:36.190864 140454833280832 maxtext_utils.py:1668] params/params/decoder/layers/pre_self_attention_layer_norm/scale Shape: float32[2048,16] Logical: P('norm', 'layers') Physical: (None, None) I0422 08:36:36.190917 140454833280832 maxtext_utils.py:1668] 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) I0422 08:36:36.190969 140454833280832 maxtext_utils.py:1668] 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') I0422 08:36:36.191011 140454833280832 maxtext_utils.py:1668] 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) I0422 08:36:36.191047 140454833280832 maxtext_utils.py:1668] 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) I0422 08:36:36.191105 140454833280832 maxtext_utils.py:1668] params/params/decoder/logits_dense/kernel Shape: float32[2048,32000] Logical: P('embed_vocab', 'vocab') Physical: ('fsdp', None) I0422 08:36:36.191154 140454833280832 maxtext_utils.py:1668] params/params/token_embedder/embedding Shape: float32[32000,2048] Logical: P('vocab', 'embed_vocab') Physical: (None, 'fsdp') I0422 08:36:36.686445 140454833280832 train.py:155] train/xent Logical: float32[32,2048]............................................ ('activation_embed_and_logits_batch', 'activation_length'). I0422 08:36:36.686549 140454833280832 train.py:155] train/xent Physical: float32[32,2048]............................................ ('fsdp', None). I0422 08:36:36.702235 140454833280832 train.py:162] train/z_loss Logical: float32[32,2048]............................................ ('activation_embed_and_logits_batch', 'activation_length'). I0422 08:36:36.702294 140454833280832 train.py:162] train/z_loss Physical: float32[32,2048]............................................ ('fsdp', None). I0422 08:36:47.643078 140454833280832 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. I0422 08:36:47.643911 140454833280832 metric_logger.py:301] number parameters: 1.104 billion I0422 08:36:58.768354 140454833280832 checkpointing.py:772] Waiting for step 10 to finish before checkpoint... I0422 08:36:59.212244 140454833280832 checkpointing.py:776] Waited 0.44386887550354004 seconds for step 10 to finish before starting checkpointing. I0422 08:36:59.214808 140454833280832 checkpoint_manager.py:2009] [process=6][thread=MainThread][wait_until_finished] No Save Finalize thread to wait for. Returning. I0422 08:36:59.217052 140454833280832 checkpoint_manager.py:1512] [process=6] Saving checkpoint at step 10 I0422 08:36:59.218656 140454833280832 event_tracking.py:70] [process=6] [async] Started save checkpoint @ gs://lance-maxtext/linen_ckpt_xpk_main_20260422_071422/linen_xpk_main_20260422_071422_16_zarr3_true/checkpoints/10. I0422 08:36:59.561897 140454833280832 signaling_client.py:364] Using JaxDistributedSignalingClient I0422 08:36:59.562943 140454833280832 jax_array_handlers.py:360] Scheduling D2H of 39 prioritized jax.Array. I0422 08:36:59.563001 140454833280832 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 I0422 08:36:59.840005 140454833280832 base_pytree_checkpoint_handler.py:154] [process=6][thread=MainThread] Initiated "orbax.checkpoint._src.serialization.jax_array_handlers.ArrayHandler".serialize. Time taken: 0.278183s I0422 08:36:59.840205 140454833280832 base_pytree_checkpoint_handler.py:130] [process=6] /jax/orbax/write/blocking_gbytes_per_sec: 5.466 GiB/s (total gbytes: 1.5 GiB) (time elapsed: 0.28223276138305664 s) (per-host) I0422 08:36:59.840265 140454833280832 base_pytree_checkpoint_handler.py:768] [process=6][thread=MainThread] Initiated Pytree async_save. Time taken: 0.282303s (batch_requests_ready=0.002331s, total_serialization_initiated=0.279893s, others=0.000080s) I0422 08:36:59.840356 140454833280832 composite_checkpoint_handler.py:715] [process=6][thread=MainThread] Initiated CompositeCheckpointHandler.async_save. Time taken: 0.286354s (all_items=0.000016s, per_item={'items': '0.00001621'}, temp_paths=0.286338) I0422 08:36:59.841057 140454833280832 event_tracking.py:125] [process=6] [async] Finished blocking save in 0.62 seconds. Continuing save @ gs://lance-maxtext/linen_ckpt_xpk_main_20260422_071422/linen_xpk_main_20260422_071422_16_zarr3_true/checkpoints/10. I0422 08:36:59.841357 140332880619264 async_checkpointer.py:76] [process=6][thread=async_save] Background save thread started. Deadline for this save operation is 2026-04-22 08:56:59.841322 I0422 08:36:59.851826 140454833280832 checkpoint_manager.py:1560] [process=6][thread=MainThread][step=10] Starting CheckpointManager Save Finalize thread=save_finalize I0422 08:36:59.852068 140336630322944 async_checkpointer.py:280] [process=6][thread=save_finalize] Waiting for background save thread=async_save. I0422 08:36:59.852208 140454833280832 standard_logger.py:34] {'step': 10, 'event_type': 'save', 'directory': 'gs://lance-maxtext/linen_ckpt_xpk_main_20260422_071422/linen_xpk_main_20260422_071422_16_zarr3_true/checkpoints', 'reached_preemption': False, 'preemption_received_at': None, 'synchronous': False, 'wait_for_prev_start_time': 1776847019.2147894, 'wait_for_prev_duration_secs': 6.151199340820312e-05, 'time_between_consecutive_saves_sec': None, 'checkpointer_blocking_start_time': 1776847019.217107, 'checkpointer_blocking_duration_secs': 0.6244065761566162, 'get_old_steps_start_time': 1776847019.8415334, 'get_old_steps_duration_secs': 3.314018249511719e-05, 'checkpoint_manager_blocking_start_time': 1776847019.2127185, 'checkpoint_manager_blocking_duration_secs': 0.6394531726837158} I0422 08:36:59.852323 140454833280832 checkpointing.py:408] Started an asynchronous checkpoint save for step 10 I0422 08:36:59.852374 140454833280832 max_utils.py:750] Memstats: After params initialized: I0422 08:36:59.852423 140454833280832 max_utils.py:756] Using (GB) 0.43 / 31.25 (1.376000%) on TPU_24(process=6,(0,6,0,0)) I0422 08:36:59.852455 140454833280832 max_utils.py:756] Using (GB) 0.43 / 31.25 (1.376000%) on TPU_25(process=6,(1,6,0,0)) I0422 08:36:59.852481 140454833280832 max_utils.py:756] Using (GB) 0.43 / 31.25 (1.376000%) on TPU_28(process=6,(0,7,0,0)) I0422 08:36:59.852506 140454833280832 max_utils.py:756] Using (GB) 0.43 / 31.25 (1.376000%) on TPU_29(process=6,(1,7,0,0)) I0422 08:37:00.164722 140454833280832 metric_logger.py:196] completed step: 10, seconds: 11.124, TFLOP/s/device: 1.221, Tokens/s/device: 184.101, total_weights: 65536, loss: 6.833, lm_loss: 6.833, perplexity: 927.869 I0422 08:37:00.341045 140454833280832 metric_logger.py:196] completed step: 11, seconds: 1.395, TFLOP/s/device: 9.741, Tokens/s/device: 1468.256, total_weights: 65536, loss: 6.527, lm_loss: 6.527, perplexity: 683.659 I0422 08:37:00.488447 140454833280832 metric_logger.py:196] completed step: 12, seconds: 0.029, TFLOP/s/device: 466.287, Tokens/s/device: 70283.812, total_weights: 65536, loss: 6.265, lm_loss: 6.265, perplexity: 526.068 I0422 08:37:00.635902 140454833280832 metric_logger.py:196] completed step: 13, seconds: 0.153, TFLOP/s/device: 88.708, Tokens/s/device: 13371.026, total_weights: 65536, loss: 6.045, lm_loss: 6.045, perplexity: 422.001 I0422 08:37:00.640432 140454833280832 checkpointing.py:772] Waiting for step 15 to finish before checkpoint... I0422 08:37:00.929850 140454833280832 checkpointing.py:776] Waited 0.2893836498260498 seconds for step 15 to finish before starting checkpointing. I0422 08:37:00.932918 140454833280832 checkpoint_manager.py:2020] [process=6][thread=MainThread][step=10][wait_until_finished] Waiting for Save Finalize thread (save_finalize) to complete. I0422 08:37:05.240857 140332889011968 array_metadata_store.py:203] [process=6][thread=array_type_handler] Wrote 39 array_metadata.ArrayMetadata to gs://lance-maxtext/linen_ckpt_xpk_main_20260422_071422/linen_xpk_main_20260422_071422_16_zarr3_true/checkpoints/10/items/array_metadatas/process_6 I0422 08:37:41.126994 140332880619264 base_pytree_checkpoint_handler.py:130] [process=6] /jax/orbax/write/gbytes_per_sec: 37.999 MiB/s (total gbytes: 1.5 GiB) (time elapsed: 41.568986654281616 s) (per-host) I0422 08:37:41.127138 140332880619264 async_checkpointer.py:90] [process=6][thread=async_save] 3 Handler Commit operations completed. Time taken: 41.285665s. I0422 08:37:50.808643 140332880619264 async_checkpointer.py:160] [process=6][thread=async_save] Background save thread done. Time taken: 50.967156s. I0422 08:37:50.808858 140336630322944 async_checkpointer.py:288] [process=6][thread=save_finalize] Done with waiting for background save thread=async_save. I0422 08:37:50.808923 140336630322944 async_checkpointer.py:298] [process=6][thread=save_finalize] No errors found in background save thread=async_save. I0422 08:37:50.808969 140336630322944 checkpoint_manager.py:2137] [process=6][thread=save_finalize][step=10] CheckpointManager Save Finalize is syncing with other hosts... I0422 08:37:50.810787 140336630322944 checkpoint_manager.py:2146] [process=6][thread=save_finalize][step=10] CheckpointManager Save Finalize is done on all hosts. I0422 08:37:50.810948 140454833280832 checkpoint_manager.py:2032] [process=6][thread=MainThread][step=10][wait_until_finished] Done waiting for Save Finalize thread (save_finalize) running at step=10. W0422 08:37:50.811082 140454833280832 checkpoint_manager.py:1452] Waiting for previous save to complete took 49.878165 seconds. If this number is high, consider checkpointing less frequently. I0422 08:37:50.813256 140454833280832 checkpoint_manager.py:1512] [process=6] Saving checkpoint at step 15 I0422 08:37:50.815771 140454833280832 event_tracking.py:70] [process=6] [async] Started save checkpoint @ gs://lance-maxtext/linen_ckpt_xpk_main_20260422_071422/linen_xpk_main_20260422_071422_16_zarr3_true/checkpoints/15. I0422 08:37:51.514388 140454833280832 jax_array_handlers.py:360] Scheduling D2H of 39 prioritized jax.Array. I0422 08:37:51.514483 140454833280832 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 I0422 08:37:51.540305 140454833280832 base_pytree_checkpoint_handler.py:154] [process=6][thread=MainThread] Initiated "orbax.checkpoint._src.serialization.jax_array_handlers.ArrayHandler".serialize. Time taken: 0.026986s I0422 08:37:51.540481 140454833280832 base_pytree_checkpoint_handler.py:130] [process=6] /jax/orbax/write/blocking_gbytes_per_sec: 50.050 GiB/s (total gbytes: 1.5 GiB) (time elapsed: 0.030820846557617188 s) (per-host) I0422 08:37:51.540541 140454833280832 base_pytree_checkpoint_handler.py:768] [process=6][thread=MainThread] Initiated Pytree async_save. Time taken: 0.030891s (batch_requests_ready=0.002068s, total_serialization_initiated=0.028747s, others=0.000077s) I0422 08:37:51.540637 140454833280832 composite_checkpoint_handler.py:715] [process=6][thread=MainThread] Initiated CompositeCheckpointHandler.async_save. Time taken: 0.035933s (all_items=0.000016s, per_item={'items': '0.00001597'}, temp_paths=0.035918) I0422 08:37:51.541324 140454833280832 event_tracking.py:125] [process=6] [async] Finished blocking save in 0.73 seconds. Continuing save @ gs://lance-maxtext/linen_ckpt_xpk_main_20260422_071422/linen_xpk_main_20260422_071422_16_zarr3_true/checkpoints/15. I0422 08:37:51.541647 140336630322944 async_checkpointer.py:76] [process=6][thread=async_save] Background save thread started. Deadline for this save operation is 2026-04-22 08:57:51.541610 I0422 08:37:51.553001 140454833280832 checkpoint_manager.py:1560] [process=6][thread=MainThread][step=15] Starting CheckpointManager Save Finalize thread=save_finalize I0422 08:37:51.553321 140330724742912 async_checkpointer.py:280] [process=6][thread=save_finalize] Waiting for background save thread=async_save. I0422 08:37:51.553504 140454833280832 standard_logger.py:34] {'step': 15, 'event_type': 'save', 'directory': 'gs://lance-maxtext/linen_ckpt_xpk_main_20260422_071422/linen_xpk_main_20260422_071422_16_zarr3_true/checkpoints', 'reached_preemption': False, 'preemption_received_at': None, 'synchronous': False, 'wait_for_prev_start_time': 1776847020.932889, 'wait_for_prev_duration_secs': 49.87816548347473, 'time_between_consecutive_saves_sec': None, 'checkpointer_blocking_start_time': 1776847070.8132958, 'checkpointer_blocking_duration_secs': 0.7284886837005615, 'get_old_steps_start_time': 1776847071.5418062, 'get_old_steps_duration_secs': 3.3855438232421875e-05, 'checkpoint_manager_blocking_start_time': 1776847020.9305365, 'checkpoint_manager_blocking_duration_secs': 50.6229190826416} I0422 08:37:51.553626 140454833280832 checkpointing.py:408] Started an asynchronous checkpoint save for step 15 I0422 08:37:51.554488 140454833280832 metric_logger.py:196] completed step: 14, seconds: 0.147, TFLOP/s/device: 92.135, Tokens/s/device: 13887.570, total_weights: 65536, loss: 5.864, lm_loss: 5.864, perplexity: 352.193 I0422 08:37:51.560772 140454833280832 metric_logger.py:196] completed step: 15, seconds: 0.147, TFLOP/s/device: 92.219, Tokens/s/device: 13900.201, total_weights: 65536, loss: 5.719, lm_loss: 5.719, perplexity: 304.661 I0422 08:37:56.677297 140332889011968 array_metadata_store.py:203] [process=6][thread=array_type_handler] Wrote 39 array_metadata.ArrayMetadata to gs://lance-maxtext/linen_ckpt_xpk_main_20260422_071422/linen_xpk_main_20260422_071422_16_zarr3_true/checkpoints/15/items/array_metadatas/process_6 I0422 08:38:12.391538 140454833280832 metric_logger.py:196] completed step: 16, seconds: 50.919, TFLOP/s/device: 0.267, Tokens/s/device: 40.221, total_weights: 65536, loss: 5.605, lm_loss: 5.605, perplexity: 271.872 I0422 08:38:12.538863 140454833280832 metric_logger.py:196] completed step: 17, seconds: 20.685, TFLOP/s/device: 0.657, Tokens/s/device: 99.009, total_weights: 65536, loss: 5.517, lm_loss: 5.517, perplexity: 248.975 I0422 08:38:12.686183 140454833280832 metric_logger.py:196] completed step: 18, seconds: 0.152, TFLOP/s/device: 89.275, Tokens/s/device: 13456.509, total_weights: 65536, loss: 5.448, lm_loss: 5.448, perplexity: 232.333 I0422 08:38:12.832802 140454833280832 checkpointing.py:772] Waiting for step 19 to finish before checkpoint... I0422 08:38:12.833549 140454833280832 checkpointing.py:776] Waited 0.0007677078247070312 seconds for step 19 to finish before starting checkpointing. I0422 08:38:12.835549 140454833280832 checkpoint_manager.py:2020] [process=6][thread=MainThread][step=15][wait_until_finished] Waiting for Save Finalize thread (save_finalize) to complete. I0422 08:38:33.813506 140336630322944 base_pytree_checkpoint_handler.py:130] [process=6] /jax/orbax/write/gbytes_per_sec: 37.339 MiB/s (total gbytes: 1.5 GiB) (time elapsed: 42.3038067817688 s) (per-host) I0422 08:38:33.813629 140336630322944 async_checkpointer.py:90] [process=6][thread=async_save] 3 Handler Commit operations completed. Time taken: 42.271872s. I0422 08:38:42.667127 140336630322944 async_checkpointer.py:160] [process=6][thread=async_save] Background save thread done. Time taken: 51.125336s. I0422 08:38:42.667426 140330724742912 async_checkpointer.py:288] [process=6][thread=save_finalize] Done with waiting for background save thread=async_save. I0422 08:38:42.667544 140330724742912 async_checkpointer.py:298] [process=6][thread=save_finalize] No errors found in background save thread=async_save. I0422 08:38:42.667593 140330724742912 checkpoint_manager.py:2137] [process=6][thread=save_finalize][step=15] CheckpointManager Save Finalize is syncing with other hosts... I0422 08:38:42.669001 140330724742912 checkpoint_manager.py:2146] [process=6][thread=save_finalize][step=15] CheckpointManager Save Finalize is done on all hosts. I0422 08:38:42.669200 140454833280832 checkpoint_manager.py:2032] [process=6][thread=MainThread][step=15][wait_until_finished] Done waiting for Save Finalize thread (save_finalize) running at step=15. W0422 08:38:42.669327 140454833280832 checkpoint_manager.py:1452] Waiting for previous save to complete took 29.833782 seconds. If this number is high, consider checkpointing less frequently. I0422 08:38:42.670748 140454833280832 checkpoint_manager.py:1512] [process=6] Saving checkpoint at step 19 I0422 08:38:42.672848 140454833280832 event_tracking.py:70] [process=6] [async] Started save checkpoint @ gs://lance-maxtext/linen_ckpt_xpk_main_20260422_071422/linen_xpk_main_20260422_071422_16_zarr3_true/checkpoints/19. I0422 08:38:42.975920 140454833280832 jax_array_handlers.py:360] Scheduling D2H of 39 prioritized jax.Array. I0422 08:38:42.976016 140454833280832 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 I0422 08:38:43.001733 140454833280832 base_pytree_checkpoint_handler.py:154] [process=6][thread=MainThread] Initiated "orbax.checkpoint._src.serialization.jax_array_handlers.ArrayHandler".serialize. Time taken: 0.026632s I0422 08:38:43.001910 140454833280832 base_pytree_checkpoint_handler.py:130] [process=6] /jax/orbax/write/blocking_gbytes_per_sec: 51.063 GiB/s (total gbytes: 1.5 GiB) (time elapsed: 0.03020954132080078 s) (per-host) I0422 08:38:43.001967 140454833280832 base_pytree_checkpoint_handler.py:768] [process=6][thread=MainThread] Initiated Pytree async_save. Time taken: 0.030277s (batch_requests_ready=0.001767s, total_serialization_initiated=0.028437s, others=0.000073s) I0422 08:38:43.002088 140454833280832 composite_checkpoint_handler.py:715] [process=6][thread=MainThread] Initiated CompositeCheckpointHandler.async_save. Time taken: 0.034329s (all_items=0.000012s, per_item={'items': '0.00001240'}, temp_paths=0.034317) I0422 08:38:43.002850 140454833280832 event_tracking.py:125] [process=6] [async] Finished blocking save in 0.33 seconds. Continuing save @ gs://lance-maxtext/linen_ckpt_xpk_main_20260422_071422/linen_xpk_main_20260422_071422_16_zarr3_true/checkpoints/19. I0422 08:38:43.003189 140330724742912 async_checkpointer.py:76] [process=6][thread=async_save] Background save thread started. Deadline for this save operation is 2026-04-22 08:58:43.003151 I0422 08:38:43.006282 140454833280832 checkpoint_manager.py:1560] [process=6][thread=MainThread][step=19] Starting CheckpointManager Save Finalize thread=save_finalize I0422 08:38:43.006574 140329634215680 async_checkpointer.py:280] [process=6][thread=save_finalize] Waiting for background save thread=async_save. I0422 08:38:43.006738 140454833280832 standard_logger.py:34] {'step': 19, 'event_type': 'save', 'directory': 'gs://lance-maxtext/linen_ckpt_xpk_main_20260422_071422/linen_xpk_main_20260422_071422_16_zarr3_true/checkpoints', 'reached_preemption': False, 'preemption_received_at': None, 'synchronous': False, 'wait_for_prev_start_time': 1776847092.835522, 'wait_for_prev_duration_secs': 29.833781719207764, 'time_between_consecutive_saves_sec': 22.024693727493286, 'checkpointer_blocking_start_time': 1776847122.670788, 'checkpointer_blocking_duration_secs': 0.3325495719909668, 'get_old_steps_start_time': 1776847123.003356, 'get_old_steps_duration_secs': 2.6702880859375e-05, 'checkpoint_manager_blocking_start_time': 1776847092.8337812, 'checkpoint_manager_blocking_duration_secs': 30.17292308807373} I0422 08:38:43.006857 140454833280832 checkpointing.py:408] Started an asynchronous checkpoint save for step 19 I0422 08:38:43.006902 140454833280832 checkpoint_manager.py:2020] [process=6][thread=MainThread][step=19][wait_until_finished] Waiting for Save Finalize thread (save_finalize) to complete. I0422 08:38:49.501410 140332889011968 array_metadata_store.py:203] [process=6][thread=array_type_handler] Wrote 39 array_metadata.ArrayMetadata to gs://lance-maxtext/linen_ckpt_xpk_main_20260422_071422/linen_xpk_main_20260422_071422_16_zarr3_true/checkpoints/19/items/array_metadatas/process_6 I0422 08:39:25.316810 140330724742912 base_pytree_checkpoint_handler.py:130] [process=6] /jax/orbax/write/gbytes_per_sec: 37.303 MiB/s (total gbytes: 1.5 GiB) (time elapsed: 42.34507131576538 s) (per-host) I0422 08:39:25.316938 140330724742912 async_checkpointer.py:90] [process=6][thread=async_save] 3 Handler Commit operations completed. Time taken: 42.313632s. I0422 08:39:34.292558 140330724742912 async_checkpointer.py:160] [process=6][thread=async_save] Background save thread done. Time taken: 51.289226s. I0422 08:39:34.292823 140329634215680 async_checkpointer.py:288] [process=6][thread=save_finalize] Done with waiting for background save thread=async_save. I0422 08:39:34.292943 140329634215680 async_checkpointer.py:298] [process=6][thread=save_finalize] No errors found in background save thread=async_save. I0422 08:39:34.292985 140329634215680 checkpoint_manager.py:2137] [process=6][thread=save_finalize][step=19] CheckpointManager Save Finalize is syncing with other hosts... I0422 08:39:34.294737 140329634215680 checkpoint_manager.py:2146] [process=6][thread=save_finalize][step=19] CheckpointManager Save Finalize is done on all hosts. I0422 08:39:34.294900 140454833280832 checkpoint_manager.py:2032] [process=6][thread=MainThread][step=19][wait_until_finished] Done waiting for Save Finalize thread (save_finalize) running at step=19. I0422 08:39:34.295042 140454833280832 checkpoint_manager.py:2009] [process=6][thread=MainThread][wait_until_finished] No Save Finalize thread to wait for. Returning. I0422 08:39:34.296018 140454833280832 metric_logger.py:196] completed step: 19, seconds: 0.147, TFLOP/s/device: 92.223, Tokens/s/device: 13900.861, total_weights: 65536, loss: 5.391, lm_loss: 5.391, perplexity: 219.349 Per train step: Total TFLOPs: 13.59 split as 93.93% learnable weight flops and 6.07% attention flops XPK End: Wed Apr 22 08:39:43 UTC 2026 EXIT_CODE=0
XPK Start: Wed Apr 22 22:47:56 UTC 2026 PyTorch was not found. Models won't be available and only tokenizers, configuration and file/data utilities can be used. `rope_parameters`'s factor field must be a float >= 1, got 40 `rope_parameters`'s beta_fast field must be a float, got 32 `rope_parameters`'s beta_slow field must be a float, got 1 DeepseekV32Config got `key=rope_scaling` in kwargs but hasn't set it as attribute. For RoPE standardization you need to set `self.rope_parameters` in model's config. 2026-04-22 22:48:21.866896: 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) I0422 22:48:22.079049 136183349827392 max_utils.py:273] Attempting to initialize the jax distributed system... I0422 22:48:31.120312 136183349827392 distributed.py:149] Starting JAX distributed service on [::]:8482 I0422 22:48:31.122664 136183349827392 distributed.py:172] Connecting to JAX distributed service on mt-16-zarr3-true-rs-wwfxh-slice-job-0-0.mt-16-zarr3-true-rs-wwfxh:8482 I0422 22:48:32.756399 136183349827392 max_utils.py:284] Jax distributed system initialized! I0422 22:48:38.745217 136183349827392 max_utils.py:800] System Information: Jax Version: 0.9.2 I0422 22:48:38.745325 136183349827392 max_utils.py:801] System Information: Jaxlib Version: 0.9.2 I0422 22:48:38.745367 136183349827392 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 I0422 22:48:38.745401 136183349827392 train_utils.py:361] WARNING: Sequence packing is essentially ignored for synthetic data. Please use a real dataset to use sequence packing. I0422 22:48:39.431568 136183349827392 maxtext_utils.py:1565] Num_devices: 32, shape (1, 1, 1, 32, 1, 1, 1, 1, 1, 1, 1, 1, 1) I0422 22:48:39.431857 136183349827392 checkpointing.py:677] Setting up checkpoint logger... I0422 22:48:39.431912 136183349827392 checkpointing.py:233] Creating checkpoint manager with ocdbt=True and zarr3=True I0422 22:48:39.431972 136183349827392 pytree_checkpoint_handler.py:592] save_device_host_concurrent_bytes=None I0422 22:48:39.432306 136183349827392 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 0x7bdb11b1c1a0>, enable_pinned_host_transfer=False, save_concurrent_bytes: 96000000000 (89.4 GiB), restore_concurrent_bytes: 96000000000 (89.4 GiB) I0422 22:48:41.109003 136183349827392 checkpointing.py:265] Enabling policy for fixed interval checkpointing. I0422 22:48:41.109239 136183349827392 checkpoint_manager.py:708] [process=3][thread=MainThread] CheckpointManager init: checkpointers=None, item_names=('items',), item_handlers={'items': <orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeCheckpointHandler object at 0x7bc61c757920>}, handler_registry=None I0422 22:48:41.109472 136183349827392 composite_checkpoint_handler.py:237] Deferred registration for item: "items". Adding handler `<orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeCheckpointHandler object at 0x7bc61c757920>` 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`. I0422 22:48:41.109523 136183349827392 composite_checkpoint_handler.py:237] Deferred registration for item: "metrics". Adding handler `<orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonCheckpointHandler object at 0x7bc61c75dd30>` 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`. I0422 22:48:41.109559 136183349827392 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 0x7bc61c757920>, ('items', <class 'orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeRestoreArgs'>): <orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeCheckpointHandler object at 0x7bc61c757920>, ('metrics', <class 'orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonSaveArgs'>): <orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonCheckpointHandler object at 0x7bc61c75dd30>, ('metrics', <class 'orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonRestoreArgs'>): <orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonCheckpointHandler object at 0x7bc61c75dd30>}). I0422 22:48:41.109879 136183349827392 abstract_checkpointer.py:35] orbax-checkpoint version: 0.11.34 I0422 22:48:41.109972 136183349827392 async_checkpointer.py:192] [process=3][thread=MainThread] Using barrier_sync_fn: <function get_barrier_sync_fn.<locals>._fn at 0x7bc5fc3432e0> timeout: 1200 secs and primary_host=0 for async checkpoint writes I0422 22:48:43.054471 136062584207104 checkpoint.py:228] Read Metadata={'item_handlers': {'items': 'orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeCheckpointHandler'}, 'metrics': {}, 'performance_metrics': {}, 'init_timestamp_nsecs': 1776897857966287684, 'commit_timestamp_nsecs': 1776897901287762356, 'custom_metadata': {}} from gs://lance-maxtext/linen_ckpt_xpk_test_pipeline_scan_nnx_20260422_212603/linen_xpk_test_pipeline_scan_nnx_20260422_212603_16_zarr3_true/checkpoints/0/_CHECKPOINT_METADATA I0422 22:48:43.298730 136062584207104 checkpoint_manager.py:1871] Missing metrics for step 0 I0422 22:48:43.598793 136183349827392 checkpoint_manager.py:1812] Found 3 checkpoint steps in gs://lance-maxtext/linen_ckpt_xpk_test_pipeline_scan_nnx_20260422_212603/linen_xpk_test_pipeline_scan_nnx_20260422_212603_16_zarr3_true/checkpoints I0422 22:48:44.990050 136183349827392 checkpoint_manager.py:929] [process=3][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_test_pipeline_scan_nnx_20260422_212603/linen_xpk_test_pipeline_scan_nnx_20260422_212603_16_zarr3_true/checkpoints: <orbax.checkpoint.checkpoint_manager.CheckpointManager object at 0x7bc6e431e780> I0422 22:48:44.990225 136183349827392 checkpointing.py:301] Checkpoint manager created! I0422 22:48:45.932720 136183349827392 nnx_wrappers.py:453] Unknown Logical: bfloat16[32,2048,2048]...................................... ('activation_batch', 'activation_norm_length', 'activation_embed'). I0422 22:48:45.932837 136183349827392 nnx_wrappers.py:453] Unknown Physical: bfloat16[32,2048,2048]...................................... ('fsdp', None, None). I0422 22:48:46.314892 136183349827392 attentions.py:1088] attentions/inputs_q Logical: bfloat16[32,2048,2048]...................................... ('activation_batch', 'activation_attn_length', 'activation_attn_embed'). I0422 22:48:46.314999 136183349827392 attentions.py:1088] attentions/inputs_q Physical: bfloat16[32,2048,2048]...................................... ('fsdp', None, None). I0422 22:48:46.331468 136183349827392 attentions.py:1089] attentions/inputs_kv Logical: bfloat16[32,2048,2048]...................................... ('activation_batch', 'activation_attn_length', 'activation_attn_embed'). I0422 22:48:46.331537 136183349827392 attentions.py:1089] attentions/inputs_kv Physical: bfloat16[32,2048,2048]...................................... ('fsdp', None, None). I0422 22:48:46.355066 136183349827392 attentions.py:1154] attentions/query Logical: bfloat16[32,2048,16,128].................................... ('activation_kv_batch', 'activation_attn_length', 'activation_kv_heads', 'activation_kv_head_dim'). I0422 22:48:46.355134 136183349827392 attentions.py:1154] attentions/query Physical: bfloat16[32,2048,16,128].................................... ('fsdp', None, None, None). I0422 22:48:46.371620 136183349827392 attentions.py:1155] attentions/key Logical: bfloat16[32,2048,16,128].................................... ('activation_kv_batch', 'activation_attn_length', 'activation_kv_heads', 'activation_kv_head_dim'). I0422 22:48:46.371685 136183349827392 attentions.py:1155] attentions/key Physical: bfloat16[32,2048,16,128].................................... ('fsdp', None, None, None). I0422 22:48:46.388116 136183349827392 attentions.py:1156] attentions/value Logical: bfloat16[32,2048,16,128].................................... ('activation_kv_batch', 'activation_attn_length', 'activation_kv_heads', 'activation_kv_head_dim'). I0422 22:48:46.388176 136183349827392 attentions.py:1156] attentions/value Physical: bfloat16[32,2048,16,128].................................... ('fsdp', None, None, None). I0422 22:48:46.413042 136183349827392 attentions.py:1198] attentions/out Logical: bfloat16[32,2048,16,128].................................... ('activation_batch', 'activation_attn_length', 'activation_heads', 'activation_kv'). I0422 22:48:46.413112 136183349827392 attentions.py:1198] attentions/out Physical: bfloat16[32,2048,16,128].................................... ('fsdp', None, None, None). I0422 22:48:46.434147 136183349827392 linears.py:525] linears/x Logical: bfloat16[32,2048,7168]...................................... ('activation_batch', 'activation_length', 'activation_mlp'). I0422 22:48:46.434219 136183349827392 linears.py:525] linears/x Physical: bfloat16[32,2048,7168]...................................... ('fsdp', None, None). I0422 22:48:46.646597 136183349827392 checkpointing.py:577] checkpoint manager exists so trying to load this run's existing checkpoint I0422 22:48:46.646726 136183349827392 checkpointing.py:581] restoring from this run's directory step 9 I0422 22:48:47.327228 136183349827392 event_tracking.py:70] [process=3] [sync] Started load checkpoint @ gs://lance-maxtext/linen_ckpt_xpk_test_pipeline_scan_nnx_20260422_212603/linen_xpk_test_pipeline_scan_nnx_20260422_212603_16_zarr3_true/checkpoints/9. I0422 22:48:47.581183 136183349827392 checkpointer.py:307] Restoring checkpoint from gs://lance-maxtext/linen_ckpt_xpk_test_pipeline_scan_nnx_20260422_212603/linen_xpk_test_pipeline_scan_nnx_20260422_212603_16_zarr3_true/checkpoints/9. I0422 22:48:47.581609 136183349827392 event_tracking.py:125] [process=3] [sync] Finished blocking load in 0.25 seconds. Continuing load @ gs://lance-maxtext/linen_ckpt_xpk_test_pipeline_scan_nnx_20260422_212603/linen_xpk_test_pipeline_scan_nnx_20260422_212603_16_zarr3_true/checkpoints/9. I0422 22:48:49.428964 1442 google_auth_provider.cc:181] Running on GCE, using service account 562977990677-compute@developer.gserviceaccount.com I0422 22:48:55.952477 136183349827392 jax_array_handlers.py:843] [process=3] /jax/orbax/read/worker/io/requested throughput: 242.021 MiB/s (total gbytes: 1.5 GiB) (time elapsed: 6.5294976234436035 s) (per-host) I0422 22:48:55.958208 136183349827392 base_pytree_checkpoint_handler.py:130] [process=3] /jax/checkpoint/read/gbytes_per_sec: 228.356 MiB/s (total gbytes: 1.5 GiB) (time elapsed: 6.930376768112183 s) (per-host) I0422 22:48:56.334690 136183349827392 event_tracking.py:138] [process=3] [sync] Finished load in 9.01 seconds @ gs://lance-maxtext/linen_ckpt_xpk_test_pipeline_scan_nnx_20260422_212603/linen_xpk_test_pipeline_scan_nnx_20260422_212603_16_zarr3_true/checkpoints/9 I0422 22:48:56.334848 136183349827392 standard_logger.py:34] {'step': 9, 'event_type': 'restore', 'directory': 'gs://lance-maxtext/linen_ckpt_xpk_test_pipeline_scan_nnx_20260422_212603/linen_xpk_test_pipeline_scan_nnx_20260422_212603_16_zarr3_true/checkpoints', 'checkpointer_start_time': 1776898127.327129, 'checkpointer_duration_secs': 9.007685899734497, 'checkpoint_manager_start_time': 1776898126.6470096, 'checkpoint_manager_duration_secs': 9.687807083129883} fsdp: 32 I0422 22:48:56.760187 136183349827392 maxtext_utils.py:1668] params/params/decoder/decoder_norm/scale Shape: float32[2048] Logical: P('norm',) Physical: (None,) I0422 22:48:56.760349 136183349827392 maxtext_utils.py:1668] params/params/decoder/layers/mlp/wi_0/kernel Shape: float32[2048,16,7168] Logical: P('embed', 'layers', 'mlp') Physical: ('fsdp', None, None) I0422 22:48:56.760427 136183349827392 maxtext_utils.py:1668] params/params/decoder/layers/mlp/wi_1/kernel Shape: float32[2048,16,7168] Logical: P('embed', 'layers', 'mlp') Physical: ('fsdp', None, None) I0422 22:48:56.760517 136183349827392 maxtext_utils.py:1668] params/params/decoder/layers/mlp/wo/kernel Shape: float32[7168,16,2048] Logical: P('mlp', 'layers', 'embed') Physical: (None, None, 'fsdp') I0422 22:48:56.760605 136183349827392 maxtext_utils.py:1668] params/params/decoder/layers/post_self_attention_layer_norm/scale Shape: float32[2048,16] Logical: P('norm', 'layers') Physical: (None, None) I0422 22:48:56.760667 136183349827392 maxtext_utils.py:1668] params/params/decoder/layers/pre_self_attention_layer_norm/scale Shape: float32[2048,16] Logical: P('norm', 'layers') Physical: (None, None) I0422 22:48:56.760762 136183349827392 maxtext_utils.py:1668] 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) I0422 22:48:56.760845 136183349827392 maxtext_utils.py:1668] 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') I0422 22:48:56.760905 136183349827392 maxtext_utils.py:1668] 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) I0422 22:48:56.760981 136183349827392 maxtext_utils.py:1668] 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) I0422 22:48:56.761055 136183349827392 maxtext_utils.py:1668] params/params/decoder/logits_dense/kernel Shape: float32[2048,32000] Logical: P('embed_vocab', 'vocab') Physical: ('fsdp', None) I0422 22:48:56.761131 136183349827392 maxtext_utils.py:1668] params/params/token_embedder/embedding Shape: float32[32000,2048] Logical: P('vocab', 'embed_vocab') Physical: (None, 'fsdp') I0422 22:48:57.245231 136183349827392 train.py:155] train/xent Logical: float32[32,2048]............................................ ('activation_embed_and_logits_batch', 'activation_length'). I0422 22:48:57.245328 136183349827392 train.py:155] train/xent Physical: float32[32,2048]............................................ ('fsdp', None). I0422 22:48:57.260860 136183349827392 train.py:162] train/z_loss Logical: float32[32,2048]............................................ ('activation_embed_and_logits_batch', 'activation_length'). I0422 22:48:57.260921 136183349827392 train.py:162] train/z_loss Physical: float32[32,2048]............................................ ('fsdp', None). I0422 22:49:08.328727 136183349827392 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. I0422 22:49:08.329574 136183349827392 metric_logger.py:301] number parameters: 1.104 billion I0422 22:49:19.699634 136183349827392 checkpointing.py:772] Waiting for step 10 to finish before checkpoint... I0422 22:49:19.864790 136183349827392 checkpointing.py:776] Waited 0.16512703895568848 seconds for step 10 to finish before starting checkpointing. I0422 22:49:19.867234 136183349827392 checkpoint_manager.py:2009] [process=3][thread=MainThread][wait_until_finished] No Save Finalize thread to wait for. Returning. I0422 22:49:19.869168 136183349827392 checkpoint_manager.py:1512] [process=3] Saving checkpoint at step 10 I0422 22:49:19.870696 136183349827392 event_tracking.py:70] [process=3] [async] Started save checkpoint @ gs://lance-maxtext/linen_ckpt_xpk_test_pipeline_scan_nnx_20260422_212603/linen_xpk_test_pipeline_scan_nnx_20260422_212603_16_zarr3_true/checkpoints/10. I0422 22:49:20.614897 136183349827392 signaling_client.py:364] Using JaxDistributedSignalingClient I0422 22:49:20.615842 136183349827392 jax_array_handlers.py:360] Scheduling D2H of 39 prioritized jax.Array. I0422 22:49:20.615900 136183349827392 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 I0422 22:49:20.896457 136183349827392 base_pytree_checkpoint_handler.py:154] [process=3][thread=MainThread] Initiated "orbax.checkpoint._src.serialization.jax_array_handlers.ArrayHandler".serialize. Time taken: 0.281636s I0422 22:49:20.896637 136183349827392 base_pytree_checkpoint_handler.py:130] [process=3] /jax/orbax/write/blocking_gbytes_per_sec: 5.403 GiB/s (total gbytes: 1.5 GiB) (time elapsed: 0.2854917049407959 s) (per-host) I0422 22:49:20.896694 136183349827392 base_pytree_checkpoint_handler.py:768] [process=3][thread=MainThread] Initiated Pytree async_save. Time taken: 0.285558s (batch_requests_ready=0.002142s, total_serialization_initiated=0.283341s, others=0.000075s) I0422 22:49:20.896784 136183349827392 composite_checkpoint_handler.py:715] [process=3][thread=MainThread] Initiated CompositeCheckpointHandler.async_save. Time taken: 0.289669s (all_items=0.000018s, per_item={'items': '0.00001764'}, temp_paths=0.289651) I0422 22:49:20.897526 136183349827392 event_tracking.py:125] [process=3] [async] Finished blocking save in 1.03 seconds. Continuing save @ gs://lance-maxtext/linen_ckpt_xpk_test_pipeline_scan_nnx_20260422_212603/linen_xpk_test_pipeline_scan_nnx_20260422_212603_16_zarr3_true/checkpoints/10. I0422 22:49:20.897892 136061997012736 async_checkpointer.py:76] [process=3][thread=async_save] Background save thread started. Deadline for this save operation is 2026-04-22 23:09:20.897853 I0422 22:49:20.908955 136183349827392 checkpoint_manager.py:1560] [process=3][thread=MainThread][step=10] Starting CheckpointManager Save Finalize thread=save_finalize I0422 22:49:20.909213 136058744456960 async_checkpointer.py:280] [process=3][thread=save_finalize] Waiting for background save thread=async_save. I0422 22:49:20.909367 136183349827392 standard_logger.py:34] {'step': 10, 'event_type': 'save', 'directory': 'gs://lance-maxtext/linen_ckpt_xpk_test_pipeline_scan_nnx_20260422_212603/linen_xpk_test_pipeline_scan_nnx_20260422_212603_16_zarr3_true/checkpoints', 'reached_preemption': False, 'preemption_received_at': None, 'synchronous': False, 'wait_for_prev_start_time': 1776898159.8672156, 'wait_for_prev_duration_secs': 6.031990051269531e-05, 'time_between_consecutive_saves_sec': None, 'checkpointer_blocking_start_time': 1776898159.8692048, 'checkpointer_blocking_duration_secs': 1.0288581848144531, 'get_old_steps_start_time': 1776898160.8980875, 'get_old_steps_duration_secs': 3.266334533691406e-05, 'checkpoint_manager_blocking_start_time': 1776898159.865318, 'checkpoint_manager_blocking_duration_secs': 1.0440123081207275} I0422 22:49:20.909486 136183349827392 checkpointing.py:408] Started an asynchronous checkpoint save for step 10 I0422 22:49:20.909535 136183349827392 max_utils.py:750] Memstats: After params initialized: I0422 22:49:20.909583 136183349827392 max_utils.py:756] Using (GB) 0.43 / 31.25 (1.376000%) on TPU_10(process=3,(2,2,0,0)) I0422 22:49:20.909615 136183349827392 max_utils.py:756] Using (GB) 0.43 / 31.25 (1.376000%) on TPU_11(process=3,(3,2,0,0)) I0422 22:49:20.909642 136183349827392 max_utils.py:756] Using (GB) 0.43 / 31.25 (1.376000%) on TPU_14(process=3,(2,3,0,0)) I0422 22:49:20.909666 136183349827392 max_utils.py:756] Using (GB) 0.43 / 31.25 (1.376000%) on TPU_15(process=3,(3,3,0,0)) I0422 22:49:21.230093 136183349827392 metric_logger.py:196] completed step: 10, seconds: 11.370, TFLOP/s/device: 1.195, Tokens/s/device: 180.124, total_weights: 65536, loss: 6.833, lm_loss: 6.833, perplexity: 927.869 I0422 22:49:21.801509 136183349827392 metric_logger.py:196] completed step: 11, seconds: 1.529, TFLOP/s/device: 8.887, Tokens/s/device: 1339.523, total_weights: 65536, loss: 6.527, lm_loss: 6.527, perplexity: 683.659 I0422 22:49:21.949143 136183349827392 metric_logger.py:196] completed step: 12, seconds: 0.426, TFLOP/s/device: 31.923, Tokens/s/device: 4811.827, total_weights: 65536, loss: 6.265, lm_loss: 6.265, perplexity: 526.068 I0422 22:49:22.096422 136183349827392 metric_logger.py:196] completed step: 13, seconds: 0.152, TFLOP/s/device: 89.491, Tokens/s/device: 13489.037, total_weights: 65536, loss: 6.045, lm_loss: 6.045, perplexity: 422.001 I0422 22:49:22.100705 136183349827392 checkpointing.py:772] Waiting for step 15 to finish before checkpoint... I0422 22:49:22.390467 136183349827392 checkpointing.py:776] Waited 0.2897157669067383 seconds for step 15 to finish before starting checkpointing. I0422 22:49:22.393562 136183349827392 checkpoint_manager.py:2020] [process=3][thread=MainThread][step=10][wait_until_finished] Waiting for Save Finalize thread (save_finalize) to complete. I0422 22:49:26.618825 136058752849664 array_metadata_store.py:203] [process=3][thread=array_type_handler] Wrote 39 array_metadata.ArrayMetadata to gs://lance-maxtext/linen_ckpt_xpk_test_pipeline_scan_nnx_20260422_212603/linen_xpk_test_pipeline_scan_nnx_20260422_212603_16_zarr3_true/checkpoints/10/items/array_metadatas/process_3 I0422 22:50:03.044291 136061997012736 base_pytree_checkpoint_handler.py:130] [process=3] /jax/orbax/write/gbytes_per_sec: 37.226 MiB/s (total gbytes: 1.5 GiB) (time elapsed: 42.43310761451721 s) (per-host) I0422 22:50:03.044420 136061997012736 async_checkpointer.py:90] [process=3][thread=async_save] 3 Handler Commit operations completed. Time taken: 42.146397s. I0422 22:50:11.830980 136061997012736 async_checkpointer.py:160] [process=3][thread=async_save] Background save thread done. Time taken: 50.932938s. I0422 22:50:11.831279 136058744456960 async_checkpointer.py:288] [process=3][thread=save_finalize] Done with waiting for background save thread=async_save. I0422 22:50:11.831398 136058744456960 async_checkpointer.py:298] [process=3][thread=save_finalize] No errors found in background save thread=async_save. I0422 22:50:11.831449 136058744456960 checkpoint_manager.py:2137] [process=3][thread=save_finalize][step=10] CheckpointManager Save Finalize is syncing with other hosts... I0422 22:50:11.833089 136058744456960 checkpoint_manager.py:2146] [process=3][thread=save_finalize][step=10] CheckpointManager Save Finalize is done on all hosts. I0422 22:50:11.833262 136183349827392 checkpoint_manager.py:2032] [process=3][thread=MainThread][step=10][wait_until_finished] Done waiting for Save Finalize thread (save_finalize) running at step=10. W0422 22:50:11.833398 136183349827392 checkpoint_manager.py:1452] Waiting for previous save to complete took 49.439837 seconds. If this number is high, consider checkpointing less frequently. I0422 22:50:11.835355 136183349827392 checkpoint_manager.py:1512] [process=3] Saving checkpoint at step 15 I0422 22:50:11.837904 136183349827392 event_tracking.py:70] [process=3] [async] Started save checkpoint @ gs://lance-maxtext/linen_ckpt_xpk_test_pipeline_scan_nnx_20260422_212603/linen_xpk_test_pipeline_scan_nnx_20260422_212603_16_zarr3_true/checkpoints/15. I0422 22:50:12.224983 136183349827392 jax_array_handlers.py:360] Scheduling D2H of 39 prioritized jax.Array. I0422 22:50:12.225074 136183349827392 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 I0422 22:50:12.265600 136183349827392 base_pytree_checkpoint_handler.py:154] [process=3][thread=MainThread] Initiated "orbax.checkpoint._src.serialization.jax_array_handlers.ArrayHandler".serialize. Time taken: 0.041811s I0422 22:50:12.265756 136183349827392 base_pytree_checkpoint_handler.py:130] [process=3] /jax/orbax/write/blocking_gbytes_per_sec: 34.117 GiB/s (total gbytes: 1.5 GiB) (time elapsed: 0.04521369934082031 s) (per-host) I0422 22:50:12.265807 136183349827392 base_pytree_checkpoint_handler.py:768] [process=3][thread=MainThread] Initiated Pytree async_save. Time taken: 0.045274s (batch_requests_ready=0.001774s, total_serialization_initiated=0.043434s, others=0.000065s) I0422 22:50:12.265893 136183349827392 composite_checkpoint_handler.py:715] [process=3][thread=MainThread] Initiated CompositeCheckpointHandler.async_save. Time taken: 0.050371s (all_items=0.000014s, per_item={'items': '0.00001359'}, temp_paths=0.050357) I0422 22:50:12.266615 136183349827392 event_tracking.py:125] [process=3] [async] Finished blocking save in 0.43 seconds. Continuing save @ gs://lance-maxtext/linen_ckpt_xpk_test_pipeline_scan_nnx_20260422_212603/linen_xpk_test_pipeline_scan_nnx_20260422_212603_16_zarr3_true/checkpoints/15. I0422 22:50:12.266911 136058744456960 async_checkpointer.py:76] [process=3][thread=async_save] Background save thread started. Deadline for this save operation is 2026-04-22 23:10:12.266886 I0422 22:50:12.268889 136183349827392 checkpoint_manager.py:1560] [process=3][thread=MainThread][step=15] Starting CheckpointManager Save Finalize thread=save_finalize I0422 22:50:12.269191 136058752849664 async_checkpointer.py:280] [process=3][thread=save_finalize] Waiting for background save thread=async_save. I0422 22:50:12.269351 136183349827392 standard_logger.py:34] {'step': 15, 'event_type': 'save', 'directory': 'gs://lance-maxtext/linen_ckpt_xpk_test_pipeline_scan_nnx_20260422_212603/linen_xpk_test_pipeline_scan_nnx_20260422_212603_16_zarr3_true/checkpoints', 'reached_preemption': False, 'preemption_received_at': None, 'synchronous': False, 'wait_for_prev_start_time': 1776898162.3935328, 'wait_for_prev_duration_secs': 49.43983697891235, 'time_between_consecutive_saves_sec': None, 'checkpointer_blocking_start_time': 1776898211.8353934, 'checkpointer_blocking_duration_secs': 0.4316372871398926, 'get_old_steps_start_time': 1776898212.2670543, 'get_old_steps_duration_secs': 2.9802322387695312e-05, 'checkpoint_manager_blocking_start_time': 1776898162.3910944, 'checkpoint_manager_blocking_duration_secs': 49.878222703933716} I0422 22:50:12.269461 136183349827392 checkpointing.py:408] Started an asynchronous checkpoint save for step 15 I0422 22:50:12.270240 136183349827392 metric_logger.py:196] completed step: 14, seconds: 0.147, TFLOP/s/device: 92.135, Tokens/s/device: 13887.570, total_weights: 65536, loss: 5.864, lm_loss: 5.864, perplexity: 352.193 I0422 22:50:12.276252 136183349827392 metric_logger.py:196] completed step: 15, seconds: 0.147, TFLOP/s/device: 92.333, Tokens/s/device: 13917.487, total_weights: 65536, loss: 5.719, lm_loss: 5.719, perplexity: 304.661 I0422 22:50:17.353567 136056603125504 array_metadata_store.py:203] [process=3][thread=array_type_handler] Wrote 39 array_metadata.ArrayMetadata to gs://lance-maxtext/linen_ckpt_xpk_test_pipeline_scan_nnx_20260422_212603/linen_xpk_test_pipeline_scan_nnx_20260422_212603_16_zarr3_true/checkpoints/15/items/array_metadatas/process_3 I0422 22:50:33.615983 136183349827392 metric_logger.py:196] completed step: 16, seconds: 50.174, TFLOP/s/device: 0.271, Tokens/s/device: 40.818, total_weights: 65536, loss: 5.605, lm_loss: 5.605, perplexity: 271.872 I0422 22:50:33.763258 136183349827392 metric_logger.py:196] completed step: 17, seconds: 21.194, TFLOP/s/device: 0.641, Tokens/s/device: 96.633, total_weights: 65536, loss: 5.517, lm_loss: 5.517, perplexity: 248.975 I0422 22:50:33.910873 136183349827392 metric_logger.py:196] completed step: 18, seconds: 0.152, TFLOP/s/device: 89.230, Tokens/s/device: 13449.705, total_weights: 65536, loss: 5.448, lm_loss: 5.448, perplexity: 232.333 I0422 22:50:34.057387 136183349827392 checkpointing.py:772] Waiting for step 19 to finish before checkpoint... I0422 22:50:34.058099 136183349827392 checkpointing.py:776] Waited 0.0007350444793701172 seconds for step 19 to finish before starting checkpointing. I0422 22:50:34.060219 136183349827392 checkpoint_manager.py:2020] [process=3][thread=MainThread][step=15][wait_until_finished] Waiting for Save Finalize thread (save_finalize) to complete. I0422 22:50:54.388884 136058744456960 base_pytree_checkpoint_handler.py:130] [process=3] /jax/orbax/write/gbytes_per_sec: 37.459 MiB/s (total gbytes: 1.5 GiB) (time elapsed: 42.16830396652222 s) (per-host) I0422 22:50:54.389030 136058744456960 async_checkpointer.py:90] [process=3][thread=async_save] 3 Handler Commit operations completed. Time taken: 42.122023s. I0422 22:51:01.277794 136058744456960 async_checkpointer.py:160] [process=3][thread=async_save] Background save thread done. Time taken: 49.010771s. I0422 22:51:01.278162 136058752849664 async_checkpointer.py:288] [process=3][thread=save_finalize] Done with waiting for background save thread=async_save. I0422 22:51:01.278298 136058752849664 async_checkpointer.py:298] [process=3][thread=save_finalize] No errors found in background save thread=async_save. I0422 22:51:01.278350 136058752849664 checkpoint_manager.py:2137] [process=3][thread=save_finalize][step=15] CheckpointManager Save Finalize is syncing with other hosts... I0422 22:51:01.282353 136058752849664 checkpoint_manager.py:2146] [process=3][thread=save_finalize][step=15] CheckpointManager Save Finalize is done on all hosts. I0422 22:51:01.282539 136183349827392 checkpoint_manager.py:2032] [process=3][thread=MainThread][step=15][wait_until_finished] Done waiting for Save Finalize thread (save_finalize) running at step=15. W0422 22:51:01.282666 136183349827392 checkpoint_manager.py:1452] Waiting for previous save to complete took 27.222455 seconds. If this number is high, consider checkpointing less frequently. I0422 22:51:01.284891 136183349827392 checkpoint_manager.py:1512] [process=3] Saving checkpoint at step 19 I0422 22:51:01.286908 136183349827392 event_tracking.py:70] [process=3] [async] Started save checkpoint @ gs://lance-maxtext/linen_ckpt_xpk_test_pipeline_scan_nnx_20260422_212603/linen_xpk_test_pipeline_scan_nnx_20260422_212603_16_zarr3_true/checkpoints/19. I0422 22:51:01.986338 136183349827392 jax_array_handlers.py:360] Scheduling D2H of 39 prioritized jax.Array. I0422 22:51:01.986455 136183349827392 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 I0422 22:51:02.026690 136183349827392 base_pytree_checkpoint_handler.py:154] [process=3][thread=MainThread] Initiated "orbax.checkpoint._src.serialization.jax_array_handlers.ArrayHandler".serialize. Time taken: 0.041353s I0422 22:51:02.026853 136183349827392 base_pytree_checkpoint_handler.py:130] [process=3] /jax/orbax/write/blocking_gbytes_per_sec: 34.430 GiB/s (total gbytes: 1.5 GiB) (time elapsed: 0.04480385780334473 s) (per-host) I0422 22:51:02.026900 136183349827392 base_pytree_checkpoint_handler.py:768] [process=3][thread=MainThread] Initiated Pytree async_save. Time taken: 0.044862s (batch_requests_ready=0.001629s, total_serialization_initiated=0.043170s, others=0.000063s) I0422 22:51:02.027000 136183349827392 composite_checkpoint_handler.py:715] [process=3][thread=MainThread] Initiated CompositeCheckpointHandler.async_save. Time taken: 0.049218s (all_items=0.000011s, per_item={'items': '0.00001073'}, temp_paths=0.049207) I0422 22:51:02.027739 136183349827392 event_tracking.py:125] [process=3] [async] Finished blocking save in 0.74 seconds. Continuing save @ gs://lance-maxtext/linen_ckpt_xpk_test_pipeline_scan_nnx_20260422_212603/linen_xpk_test_pipeline_scan_nnx_20260422_212603_16_zarr3_true/checkpoints/19. I0422 22:51:02.028064 136058752849664 async_checkpointer.py:76] [process=3][thread=async_save] Background save thread started. Deadline for this save operation is 2026-04-22 23:11:02.028026 I0422 22:51:02.030028 136183349827392 checkpoint_manager.py:1560] [process=3][thread=MainThread][step=19] Starting CheckpointManager Save Finalize thread=save_finalize I0422 22:51:02.030289 136061997012736 async_checkpointer.py:280] [process=3][thread=save_finalize] Waiting for background save thread=async_save. I0422 22:51:02.030443 136183349827392 standard_logger.py:34] {'step': 19, 'event_type': 'save', 'directory': 'gs://lance-maxtext/linen_ckpt_xpk_test_pipeline_scan_nnx_20260422_212603/linen_xpk_test_pipeline_scan_nnx_20260422_212603_16_zarr3_true/checkpoints', 'reached_preemption': False, 'preemption_received_at': None, 'synchronous': False, 'wait_for_prev_start_time': 1776898234.0601876, 'wait_for_prev_duration_secs': 27.22245478630066, 'time_between_consecutive_saves_sec': 22.227060317993164, 'checkpointer_blocking_start_time': 1776898261.2849824, 'checkpointer_blocking_duration_secs': 0.7432248592376709, 'get_old_steps_start_time': 1776898262.0282254, 'get_old_steps_duration_secs': 2.5033950805664062e-05, 'checkpoint_manager_blocking_start_time': 1776898234.0583284, 'checkpoint_manager_blocking_duration_secs': 27.97208309173584} I0422 22:51:02.030556 136183349827392 checkpointing.py:408] Started an asynchronous checkpoint save for step 19 I0422 22:51:02.030599 136183349827392 checkpoint_manager.py:2020] [process=3][thread=MainThread][step=19][wait_until_finished] Waiting for Save Finalize thread (save_finalize) to complete. I0422 22:51:07.578637 136055017690880 array_metadata_store.py:203] [process=3][thread=array_type_handler] Wrote 39 array_metadata.ArrayMetadata to gs://lance-maxtext/linen_ckpt_xpk_test_pipeline_scan_nnx_20260422_212603/linen_xpk_test_pipeline_scan_nnx_20260422_212603_16_zarr3_true/checkpoints/19/items/array_metadatas/process_3 I0422 22:51:44.872801 136058752849664 base_pytree_checkpoint_handler.py:130] [process=3] /jax/orbax/write/gbytes_per_sec: 36.828 MiB/s (total gbytes: 1.5 GiB) (time elapsed: 42.89071202278137 s) (per-host) I0422 22:51:44.872942 136058752849664 async_checkpointer.py:90] [process=3][thread=async_save] 3 Handler Commit operations completed. Time taken: 42.844751s. I0422 22:52:03.944739 136058752849664 async_checkpointer.py:160] [process=3][thread=async_save] Background save thread done. Time taken: 61.916548s. I0422 22:52:03.945012 136061997012736 async_checkpointer.py:288] [process=3][thread=save_finalize] Done with waiting for background save thread=async_save. I0422 22:52:03.945131 136061997012736 async_checkpointer.py:298] [process=3][thread=save_finalize] No errors found in background save thread=async_save. I0422 22:52:03.945174 136061997012736 checkpoint_manager.py:2137] [process=3][thread=save_finalize][step=19] CheckpointManager Save Finalize is syncing with other hosts... I0422 22:52:03.946818 136061997012736 checkpoint_manager.py:2146] [process=3][thread=save_finalize][step=19] CheckpointManager Save Finalize is done on all hosts. I0422 22:52:03.947043 136183349827392 checkpoint_manager.py:2032] [process=3][thread=MainThread][step=19][wait_until_finished] Done waiting for Save Finalize thread (save_finalize) running at step=19. I0422 22:52:03.947217 136183349827392 checkpoint_manager.py:2009] [process=3][thread=MainThread][wait_until_finished] No Save Finalize thread to wait for. Returning. I0422 22:52:03.948301 136183349827392 metric_logger.py:196] completed step: 19, seconds: 0.147, TFLOP/s/device: 92.295, Tokens/s/device: 13911.720, total_weights: 65536, loss: 5.391, lm_loss: 5.391, perplexity: 219.349 Per train step: Total TFLOPs: 13.59 split as 93.93% learnable weight flops and 6.07% attention flops XPK End: Wed Apr 22 22:52:17 UTC 2026 EXIT_CODE=0