main (574ad3fb9) vs test/pipeline-scan-nnx (e27fc1e97)| Metric | main 574ad3fb9 | test/pipeline-scan-nnx e27fc1e97 | Diff (test/pipeline-scan-nnx − main) |
|---|---|---|---|
| Parameters | 1.104 billion | 1.104 billion | — |
| Final loss | 7.1810 | 7.1810 | 0 |
| TFLOP/s | 92.082 | 92.111 | +0.029 |
| Tok/s | 13879.6 | 13884.0 | +4.423 |
| Avg s/step | 7.745 | 8.375 | +0.63 |
| Memory % | 1.38 | 1.38 | 0 |
| JAX | 0.9.2 | 0.8.1 | — |
Diff = branch value − main value. Green = branch improved. Red = branch regressed.
XPK Start: Mon Apr 20 06:59:18 UTC 2026 PyTorch was not found. Models won't be available and only tokenizers, configuration and file/data utilities can be used. 2026-04-20 06:59:42.953772: 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) I0420 06:59:43.134013 137378039891776 max_utils.py:273] Attempting to initialize the jax distributed system... I0420 06:59:52.174803 137378039891776 distributed.py:149] Starting JAX distributed service on [::]:8482 I0420 06:59:52.177190 137378039891776 distributed.py:172] Connecting to JAX distributed service on mt-15-ocdbt-true-sv-hte1v-slice-job-0-0.mt-15-ocdbt-true-sv-hte1v:8482 I0420 06:59:53.410859 137378039891776 max_utils.py:284] Jax distributed system initialized! I0420 06:59:59.590697 137378039891776 max_utils.py:800] System Information: Jax Version: 0.9.2 I0420 06:59:59.590798 137378039891776 max_utils.py:801] System Information: Jaxlib Version: 0.9.2 I0420 06:59:59.590838 137378039891776 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 I0420 06:59:59.590871 137378039891776 train_utils.py:348] WARNING: Sequence packing is essentially ignored for synthetic data. Please use a real dataset to use sequence packing. I0420 07:00:00.285116 137378039891776 maxtext_utils.py:1551] Num_devices: 32, shape (1, 1, 1, 32, 1, 1, 1, 1, 1, 1, 1, 1, 1) I0420 07:00:00.285410 137378039891776 checkpointing.py:677] Setting up checkpoint logger... I0420 07:00:00.285464 137378039891776 checkpointing.py:233] Creating checkpoint manager with ocdbt=True and zarr3=True I0420 07:00:00.285510 137378039891776 pytree_checkpoint_handler.py:592] save_device_host_concurrent_bytes=None I0420 07:00:00.285850 137378039891776 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 0x7cf142900860>, enable_pinned_host_transfer=False, save_concurrent_bytes: 96000000000 (89.4 GiB), restore_concurrent_bytes: 96000000000 (89.4 GiB) I0420 07:00:03.187960 137378039891776 checkpointing.py:265] Enabling policy for fixed interval checkpointing. I0420 07:00:03.188193 137378039891776 checkpoint_manager.py:708] [process=4][thread=MainThread] CheckpointManager init: checkpointers=None, item_names=('items',), item_handlers={'items': <orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeCheckpointHandler object at 0x7cdcf068a9f0>}, handler_registry=None I0420 07:00:03.188441 137378039891776 composite_checkpoint_handler.py:237] Deferred registration for item: "items". Adding handler `<orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeCheckpointHandler object at 0x7cdcf068a9f0>` 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`. I0420 07:00:03.188489 137378039891776 composite_checkpoint_handler.py:237] Deferred registration for item: "metrics". Adding handler `<orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonCheckpointHandler object at 0x7cdcf068c9b0>` 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`. I0420 07:00:03.188524 137378039891776 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 0x7cdcf068a9f0>, ('items', <class 'orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeRestoreArgs'>): <orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeCheckpointHandler object at 0x7cdcf068a9f0>, ('metrics', <class 'orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonSaveArgs'>): <orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonCheckpointHandler object at 0x7cdcf068c9b0>, ('metrics', <class 'orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonRestoreArgs'>): <orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonCheckpointHandler object at 0x7cdcf068c9b0>}). I0420 07:00:03.188860 137378039891776 abstract_checkpointer.py:35] orbax-checkpoint version: 0.11.34 I0420 07:00:03.188932 137378039891776 async_checkpointer.py:192] [process=4][thread=MainThread] Using barrier_sync_fn: <function get_barrier_sync_fn.<locals>._fn at 0x7cdcf02afc40> timeout: 1200 secs and primary_host=0 for async checkpoint writes I0420 07:00:04.311014 137378039891776 checkpoint_manager.py:1812] Found 0 checkpoint steps in gs://lance-maxtext/linen_ckpt_xpk_main_20260420_060028/linen_xpk_main_20260420_060028_15_ocdbt_true/checkpoints I0420 07:00:04.743789 137378039891776 checkpoint_manager.py:929] [process=4][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_20260420_060028/linen_xpk_main_20260420_060028_15_ocdbt_true/checkpoints: <orbax.checkpoint.checkpoint_manager.CheckpointManager object at 0x7cdcf0688e30> I0420 07:00:04.743957 137378039891776 checkpointing.py:301] Checkpoint manager created! I0420 07:00:05.665940 137378039891776 nnx_wrappers.py:437] Unknown Logical: bfloat16[32,2048,2048]...................................... ('activation_batch', 'activation_norm_length', 'activation_embed'). I0420 07:00:05.666054 137378039891776 nnx_wrappers.py:437] Unknown Physical: bfloat16[32,2048,2048]...................................... ('fsdp', None, None). I0420 07:00:06.045390 137378039891776 attentions.py:1088] attentions/inputs_q Logical: bfloat16[32,2048,2048]...................................... ('activation_batch', 'activation_attn_length', 'activation_attn_embed'). I0420 07:00:06.045476 137378039891776 attentions.py:1088] attentions/inputs_q Physical: bfloat16[32,2048,2048]...................................... ('fsdp', None, None). I0420 07:00:06.061789 137378039891776 attentions.py:1089] attentions/inputs_kv Logical: bfloat16[32,2048,2048]...................................... ('activation_batch', 'activation_attn_length', 'activation_attn_embed'). I0420 07:00:06.061841 137378039891776 attentions.py:1089] attentions/inputs_kv Physical: bfloat16[32,2048,2048]...................................... ('fsdp', None, None). I0420 07:00:06.085455 137378039891776 attentions.py:1154] attentions/query Logical: bfloat16[32,2048,16,128].................................... ('activation_kv_batch', 'activation_attn_length', 'activation_kv_heads', 'activation_kv_head_dim'). I0420 07:00:06.085522 137378039891776 attentions.py:1154] attentions/query Physical: bfloat16[32,2048,16,128].................................... ('fsdp', None, None, None). I0420 07:00:06.102010 137378039891776 attentions.py:1155] attentions/key Logical: bfloat16[32,2048,16,128].................................... ('activation_kv_batch', 'activation_attn_length', 'activation_kv_heads', 'activation_kv_head_dim'). I0420 07:00:06.102076 137378039891776 attentions.py:1155] attentions/key Physical: bfloat16[32,2048,16,128].................................... ('fsdp', None, None, None). I0420 07:00:06.118563 137378039891776 attentions.py:1156] attentions/value Logical: bfloat16[32,2048,16,128].................................... ('activation_kv_batch', 'activation_attn_length', 'activation_kv_heads', 'activation_kv_head_dim'). I0420 07:00:06.118620 137378039891776 attentions.py:1156] attentions/value Physical: bfloat16[32,2048,16,128].................................... ('fsdp', None, None, None). I0420 07:00:06.143270 137378039891776 attentions.py:1197] attentions/out Logical: bfloat16[32,2048,16,128].................................... ('activation_batch', 'activation_attn_length', 'activation_heads', 'activation_kv'). I0420 07:00:06.143356 137378039891776 attentions.py:1197] attentions/out Physical: bfloat16[32,2048,16,128].................................... ('fsdp', None, None, None). I0420 07:00:06.164289 137378039891776 linears.py:525] linears/x Logical: bfloat16[32,2048,7168]...................................... ('activation_batch', 'activation_length', 'activation_mlp'). I0420 07:00:06.164358 137378039891776 linears.py:525] linears/x Physical: bfloat16[32,2048,7168]...................................... ('fsdp', None, None). I0420 07:00:06.379727 137378039891776 checkpointing.py:577] checkpoint manager exists so trying to load this run's existing checkpoint I0420 07:00:06.379830 137378039891776 checkpointing.py:665] No existing checkpoints found, not restoring checkpoint. fsdp: 32 I0420 07:00:07.803615 137378039891776 maxtext_utils.py:1654] params/params/decoder/decoder_norm/scale Shape: float32[2048] Logical: P('norm',) Physical: (None,) I0420 07:00:07.803733 137378039891776 maxtext_utils.py:1654] params/params/decoder/layers/mlp/wi_0/kernel Shape: float32[2048,16,7168] Logical: P('embed', 'layers', 'mlp') Physical: ('fsdp', None, None) I0420 07:00:07.803785 137378039891776 maxtext_utils.py:1654] params/params/decoder/layers/mlp/wi_1/kernel Shape: float32[2048,16,7168] Logical: P('embed', 'layers', 'mlp') Physical: ('fsdp', None, None) I0420 07:00:07.803846 137378039891776 maxtext_utils.py:1654] params/params/decoder/layers/mlp/wo/kernel Shape: float32[7168,16,2048] Logical: P('mlp', 'layers', 'embed') Physical: (None, None, 'fsdp') I0420 07:00:07.803899 137378039891776 maxtext_utils.py:1654] params/params/decoder/layers/post_self_attention_layer_norm/scale Shape: float32[2048,16] Logical: P('norm', 'layers') Physical: (None, None) I0420 07:00:07.803936 137378039891776 maxtext_utils.py:1654] params/params/decoder/layers/pre_self_attention_layer_norm/scale Shape: float32[2048,16] Logical: P('norm', 'layers') Physical: (None, None) I0420 07:00:07.803987 137378039891776 maxtext_utils.py:1654] params/params/decoder/layers/self_attention/key/kernel Shape: float32[2048,16,16,128] Logical: P('embed', 'layers', 'kv_heads', 'kv_head_dim') Physical: ('fsdp', None, None, None) I0420 07:00:07.804039 137378039891776 maxtext_utils.py:1654] params/params/decoder/layers/self_attention/out/kernel Shape: float32[16,16,128,2048] Logical: P('heads', 'layers', 'kv', 'embed') Physical: (None, None, None, 'fsdp') I0420 07:00:07.804077 137378039891776 maxtext_utils.py:1654] params/params/decoder/layers/self_attention/query/kernel Shape: float32[2048,16,16,128] Logical: P('embed', 'layers', 'q_heads', 'kv') Physical: ('fsdp', None, None, None) I0420 07:00:07.804113 137378039891776 maxtext_utils.py:1654] params/params/decoder/layers/self_attention/value/kernel Shape: float32[2048,16,16,128] Logical: P('embed', 'layers', 'kv_heads', 'kv_head_dim') Physical: ('fsdp', None, None, None) I0420 07:00:07.804159 137378039891776 maxtext_utils.py:1654] params/params/decoder/logits_dense/kernel Shape: float32[2048,32000] Logical: P('embed_vocab', 'vocab') Physical: ('fsdp', None) I0420 07:00:07.804205 137378039891776 maxtext_utils.py:1654] params/params/token_embedder/embedding Shape: float32[32000,2048] Logical: P('vocab', 'embed_vocab') Physical: (None, 'fsdp') I0420 07:00:08.297189 137378039891776 train.py:155] train/xent Logical: float32[32,2048]............................................ ('activation_embed_and_logits_batch', 'activation_length'). I0420 07:00:08.297282 137378039891776 train.py:155] train/xent Physical: float32[32,2048]............................................ ('fsdp', None). I0420 07:00:08.312966 137378039891776 train.py:162] train/z_loss Logical: float32[32,2048]............................................ ('activation_embed_and_logits_batch', 'activation_length'). I0420 07:00:08.313025 137378039891776 train.py:162] train/z_loss Physical: float32[32,2048]............................................ ('fsdp', None). I0420 07:00:19.265494 137378039891776 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. I0420 07:00:19.266297 137378039891776 metric_logger.py:301] number parameters: 1.104 billion I0420 07:00:30.514466 137378039891776 checkpointing.py:772] Waiting for step 0 to finish before checkpoint... I0420 07:00:30.670985 137378039891776 checkpointing.py:776] Waited 0.15649843215942383 seconds for step 0 to finish before starting checkpointing. I0420 07:00:30.673423 137378039891776 checkpoint_manager.py:2009] [process=4][thread=MainThread][wait_until_finished] No Save Finalize thread to wait for. Returning. I0420 07:00:30.675140 137378039891776 checkpoint_manager.py:1512] [process=4] Saving checkpoint at step 0 I0420 07:00:30.676567 137378039891776 event_tracking.py:70] [process=4] [async] Started save checkpoint @ gs://lance-maxtext/linen_ckpt_xpk_main_20260420_060028/linen_xpk_main_20260420_060028_15_ocdbt_true/checkpoints/0. I0420 07:00:31.010890 137378039891776 signaling_client.py:364] Using JaxDistributedSignalingClient I0420 07:00:31.011779 137378039891776 jax_array_handlers.py:360] Scheduling D2H of 39 prioritized jax.Array. I0420 07:00:31.011841 137378039891776 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 I0420 07:00:31.286088 137378039891776 base_pytree_checkpoint_handler.py:154] [process=4][thread=MainThread] Initiated "orbax.checkpoint._src.serialization.jax_array_handlers.ArrayHandler".serialize. Time taken: 0.275324s I0420 07:00:31.286265 137378039891776 base_pytree_checkpoint_handler.py:130] [process=4] /jax/orbax/write/blocking_gbytes_per_sec: 5.488 GiB/s (total gbytes: 1.5 GiB) (time elapsed: 0.28107357025146484 s) (per-host) I0420 07:00:31.286316 137378039891776 base_pytree_checkpoint_handler.py:768] [process=4][thread=MainThread] Initiated Pytree async_save. Time taken: 0.281135s (batch_requests_ready=0.002596s, total_serialization_initiated=0.278469s, others=0.000070s) I0420 07:00:31.286426 137378039891776 composite_checkpoint_handler.py:715] [process=4][thread=MainThread] Initiated CompositeCheckpointHandler.async_save. Time taken: 0.285267s (all_items=0.000018s, per_item={'items': '0.00001788'}, temp_paths=0.285249) I0420 07:00:31.287352 137378039891776 event_tracking.py:125] [process=4] [async] Finished blocking save in 0.61 seconds. Continuing save @ gs://lance-maxtext/linen_ckpt_xpk_main_20260420_060028/linen_xpk_main_20260420_060028_15_ocdbt_true/checkpoints/0. I0420 07:00:31.287653 137249151268608 async_checkpointer.py:76] [process=4][thread=async_save] Background save thread started. Deadline for this save operation is 2026-04-20 07:20:31.287621 I0420 07:00:31.300463 137378039891776 checkpoint_manager.py:1560] [process=4][thread=MainThread][step=0] Starting CheckpointManager Save Finalize thread=save_finalize I0420 07:00:31.300711 137246485247744 async_checkpointer.py:280] [process=4][thread=save_finalize] Waiting for background save thread=async_save. I0420 07:00:31.300830 137378039891776 standard_logger.py:34] {'step': 0, 'event_type': 'save', 'directory': 'gs://lance-maxtext/linen_ckpt_xpk_main_20260420_060028/linen_xpk_main_20260420_060028_15_ocdbt_true/checkpoints', 'reached_preemption': False, 'preemption_received_at': None, 'synchronous': False, 'wait_for_prev_start_time': 1776668430.673404, 'wait_for_prev_duration_secs': 6.103515625e-05, 'time_between_consecutive_saves_sec': None, 'checkpointer_blocking_start_time': 1776668430.6751788, 'checkpointer_blocking_duration_secs': 0.6126282215118408, 'get_old_steps_start_time': 1776668431.2878346, 'get_old_steps_duration_secs': 3.147125244140625e-05, 'checkpoint_manager_blocking_start_time': 1776668430.6715488, 'checkpoint_manager_blocking_duration_secs': 0.6292433738708496} I0420 07:00:31.300942 137378039891776 checkpointing.py:408] Started an asynchronous checkpoint save for step 0 I0420 07:00:31.300992 137378039891776 max_utils.py:750] Memstats: After params initialized: I0420 07:00:31.301041 137378039891776 max_utils.py:756] Using (GB) 0.43 / 31.25 (1.376000%) on TPU_16(process=4,(0,4,0,0)) I0420 07:00:31.301072 137378039891776 max_utils.py:756] Using (GB) 0.43 / 31.25 (1.376000%) on TPU_17(process=4,(1,4,0,0)) I0420 07:00:31.301097 137378039891776 max_utils.py:756] Using (GB) 0.43 / 31.25 (1.376000%) on TPU_20(process=4,(0,5,0,0)) I0420 07:00:31.301120 137378039891776 max_utils.py:756] Using (GB) 0.43 / 31.25 (1.376000%) on TPU_21(process=4,(1,5,0,0)) I0420 07:00:31.618642 137378039891776 metric_logger.py:196] completed step: 0, seconds: 11.248, TFLOP/s/device: 1.208, Tokens/s/device: 182.076, total_weights: 65536, loss: 10.874, lm_loss: 10.874, perplexity: 52779.875 I0420 07:00:31.785450 137378039891776 metric_logger.py:196] completed step: 1, seconds: 1.103, TFLOP/s/device: 12.322, Tokens/s/device: 1857.357, total_weights: 65536, loss: 10.874, lm_loss: 10.874, perplexity: 52779.875 I0420 07:00:32.205521 137378039891776 metric_logger.py:196] completed step: 2, seconds: 0.019, TFLOP/s/device: 699.574, Tokens/s/device: 105447.431, total_weights: 65536, loss: 10.560, lm_loss: 10.560, perplexity: 38547.719 I0420 07:00:32.352802 137378039891776 metric_logger.py:196] completed step: 3, seconds: 0.421, TFLOP/s/device: 32.284, Tokens/s/device: 4866.134, total_weights: 65536, loss: 9.980, lm_loss: 9.980, perplexity: 21590.506 I0420 07:00:32.357072 137378039891776 checkpointing.py:772] Waiting for step 5 to finish before checkpoint... I0420 07:00:32.647099 137378039891776 checkpointing.py:776] Waited 0.2899913787841797 seconds for step 5 to finish before starting checkpointing. I0420 07:00:32.650109 137378039891776 checkpoint_manager.py:2020] [process=4][thread=MainThread][step=0][wait_until_finished] Waiting for Save Finalize thread (save_finalize) to complete. I0420 07:00:34.287438 2859 google_auth_provider.cc:181] Running on GCE, using service account 562977990677-compute@developer.gserviceaccount.com I0420 07:00:36.338839 137248039737088 array_metadata_store.py:203] [process=4][thread=array_type_handler] Wrote 39 array_metadata.ArrayMetadata to gs://lance-maxtext/linen_ckpt_xpk_main_20260420_060028/linen_xpk_main_20260420_060028_15_ocdbt_true/checkpoints/0/items/array_metadatas/process_4 I0420 07:01:07.847415 137249151268608 base_pytree_checkpoint_handler.py:130] [process=4] /jax/orbax/write/gbytes_per_sec: 42.875 MiB/s (total gbytes: 1.5 GiB) (time elapsed: 36.84215235710144 s) (per-host) I0420 07:01:07.847540 137249151268608 async_checkpointer.py:90] [process=4][thread=async_save] 3 Handler Commit operations completed. Time taken: 36.559777s. I0420 07:01:16.715395 137249151268608 async_checkpointer.py:160] [process=4][thread=async_save] Background save thread done. Time taken: 45.427616s. I0420 07:01:16.715754 137246485247744 async_checkpointer.py:288] [process=4][thread=save_finalize] Done with waiting for background save thread=async_save. I0420 07:01:16.715873 137246485247744 async_checkpointer.py:298] [process=4][thread=save_finalize] No errors found in background save thread=async_save. I0420 07:01:16.715925 137246485247744 checkpoint_manager.py:2137] [process=4][thread=save_finalize][step=0] CheckpointManager Save Finalize is syncing with other hosts... I0420 07:01:16.717241 137246485247744 checkpoint_manager.py:2146] [process=4][thread=save_finalize][step=0] CheckpointManager Save Finalize is done on all hosts. I0420 07:01:16.717375 137378039891776 checkpoint_manager.py:2032] [process=4][thread=MainThread][step=0][wait_until_finished] Done waiting for Save Finalize thread (save_finalize) running at step=0. W0420 07:01:16.717462 137378039891776 checkpoint_manager.py:1452] Waiting for previous save to complete took 44.067359 seconds. If this number is high, consider checkpointing less frequently. I0420 07:01:16.719802 137378039891776 checkpoint_manager.py:1512] [process=4] Saving checkpoint at step 5 I0420 07:01:16.722302 137378039891776 event_tracking.py:70] [process=4] [async] Started save checkpoint @ gs://lance-maxtext/linen_ckpt_xpk_main_20260420_060028/linen_xpk_main_20260420_060028_15_ocdbt_true/checkpoints/5. I0420 07:01:17.017805 137378039891776 jax_array_handlers.py:360] Scheduling D2H of 39 prioritized jax.Array. I0420 07:01:17.017898 137378039891776 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 I0420 07:01:17.053414 137378039891776 base_pytree_checkpoint_handler.py:154] [process=4][thread=MainThread] Initiated "orbax.checkpoint._src.serialization.jax_array_handlers.ArrayHandler".serialize. Time taken: 0.036678s I0420 07:01:17.053582 137378039891776 base_pytree_checkpoint_handler.py:130] [process=4] /jax/orbax/write/blocking_gbytes_per_sec: 38.387 GiB/s (total gbytes: 1.5 GiB) (time elapsed: 0.04018521308898926 s) (per-host) I0420 07:01:17.053633 137378039891776 base_pytree_checkpoint_handler.py:768] [process=4][thread=MainThread] Initiated Pytree async_save. Time taken: 0.040246s (batch_requests_ready=0.001862s, total_serialization_initiated=0.038315s, others=0.000069s) I0420 07:01:17.053728 137378039891776 composite_checkpoint_handler.py:715] [process=4][thread=MainThread] Initiated CompositeCheckpointHandler.async_save. Time taken: 0.045100s (all_items=0.000015s, per_item={'items': '0.00001526'}, temp_paths=0.045085) I0420 07:01:17.054458 137378039891776 event_tracking.py:125] [process=4] [async] Finished blocking save in 0.33 seconds. Continuing save @ gs://lance-maxtext/linen_ckpt_xpk_main_20260420_060028/linen_xpk_main_20260420_060028_15_ocdbt_true/checkpoints/5. I0420 07:01:17.054798 137246485247744 async_checkpointer.py:76] [process=4][thread=async_save] Background save thread started. Deadline for this save operation is 2026-04-20 07:21:17.054763 I0420 07:01:17.064903 137378039891776 checkpoint_manager.py:1560] [process=4][thread=MainThread][step=5] Starting CheckpointManager Save Finalize thread=save_finalize I0420 07:01:17.065183 137243082675968 async_checkpointer.py:280] [process=4][thread=save_finalize] Waiting for background save thread=async_save. I0420 07:01:17.065351 137378039891776 standard_logger.py:34] {'step': 5, 'event_type': 'save', 'directory': 'gs://lance-maxtext/linen_ckpt_xpk_main_20260420_060028/linen_xpk_main_20260420_060028_15_ocdbt_true/checkpoints', 'reached_preemption': False, 'preemption_received_at': None, 'synchronous': False, 'wait_for_prev_start_time': 1776668432.6500785, 'wait_for_prev_duration_secs': 44.06735873222351, 'time_between_consecutive_saves_sec': None, 'checkpointer_blocking_start_time': 1776668476.7198462, 'checkpointer_blocking_duration_secs': 0.3350963592529297, 'get_old_steps_start_time': 1776668477.0549676, 'get_old_steps_duration_secs': 2.7894973754882812e-05, 'checkpoint_manager_blocking_start_time': 1776668432.6477435, 'checkpoint_manager_blocking_duration_secs': 44.417564868927} I0420 07:01:17.065459 137378039891776 checkpointing.py:408] Started an asynchronous checkpoint save for step 5 I0420 07:01:17.066268 137378039891776 metric_logger.py:196] completed step: 4, seconds: 0.153, TFLOP/s/device: 89.002, Tokens/s/device: 13415.345, total_weights: 65536, loss: 9.460, lm_loss: 9.460, perplexity: 12840.355 I0420 07:01:17.072293 137378039891776 metric_logger.py:196] completed step: 5, seconds: 0.147, TFLOP/s/device: 92.485, Tokens/s/device: 13940.413, total_weights: 65536, loss: 8.959, lm_loss: 8.959, perplexity: 7776.467 I0420 07:01:23.147635 137248039737088 array_metadata_store.py:203] [process=4][thread=array_type_handler] Wrote 39 array_metadata.ArrayMetadata to gs://lance-maxtext/linen_ckpt_xpk_main_20260420_060028/linen_xpk_main_20260420_060028_15_ocdbt_true/checkpoints/5/items/array_metadatas/process_4 I0420 07:01:40.064192 137378039891776 metric_logger.py:196] completed step: 6, seconds: 44.714, TFLOP/s/device: 0.304, Tokens/s/device: 45.802, total_weights: 65536, loss: 8.469, lm_loss: 8.469, perplexity: 4762.740 I0420 07:01:40.211625 137378039891776 metric_logger.py:196] completed step: 7, seconds: 22.846, TFLOP/s/device: 0.595, Tokens/s/device: 89.644, total_weights: 65536, loss: 8.003, lm_loss: 8.003, perplexity: 2989.151 I0420 07:01:40.359124 137378039891776 metric_logger.py:196] completed step: 8, seconds: 0.152, TFLOP/s/device: 89.329, Tokens/s/device: 13464.649, total_weights: 65536, loss: 7.572, lm_loss: 7.572, perplexity: 1943.098 I0420 07:01:40.505934 137378039891776 checkpointing.py:772] Waiting for step 9 to finish before checkpoint... I0420 07:01:40.506632 137378039891776 checkpointing.py:776] Waited 0.0007174015045166016 seconds for step 9 to finish before starting checkpointing. I0420 07:01:40.508862 137378039891776 checkpoint_manager.py:2020] [process=4][thread=MainThread][step=5][wait_until_finished] Waiting for Save Finalize thread (save_finalize) to complete. I0420 07:02:00.886527 137246485247744 base_pytree_checkpoint_handler.py:130] [process=4] /jax/orbax/write/gbytes_per_sec: 36.004 MiB/s (total gbytes: 1.5 GiB) (time elapsed: 43.87310338020325 s) (per-host) I0420 07:02:00.886634 137246485247744 async_checkpointer.py:90] [process=4][thread=async_save] 3 Handler Commit operations completed. Time taken: 43.831720s. I0420 07:02:07.692399 137246485247744 async_checkpointer.py:160] [process=4][thread=async_save] Background save thread done. Time taken: 50.637470s. I0420 07:02:07.692746 137243082675968 async_checkpointer.py:288] [process=4][thread=save_finalize] Done with waiting for background save thread=async_save. I0420 07:02:07.692865 137243082675968 async_checkpointer.py:298] [process=4][thread=save_finalize] No errors found in background save thread=async_save. I0420 07:02:07.692916 137243082675968 checkpoint_manager.py:2137] [process=4][thread=save_finalize][step=5] CheckpointManager Save Finalize is syncing with other hosts... I0420 07:02:07.694302 137243082675968 checkpoint_manager.py:2146] [process=4][thread=save_finalize][step=5] CheckpointManager Save Finalize is done on all hosts. I0420 07:02:07.694508 137378039891776 checkpoint_manager.py:2032] [process=4][thread=MainThread][step=5][wait_until_finished] Done waiting for Save Finalize thread (save_finalize) running at step=5. W0420 07:02:07.694649 137378039891776 checkpoint_manager.py:1452] Waiting for previous save to complete took 27.185793 seconds. If this number is high, consider checkpointing less frequently. I0420 07:02:07.695944 137378039891776 checkpoint_manager.py:1512] [process=4] Saving checkpoint at step 9 I0420 07:02:07.698049 137378039891776 event_tracking.py:70] [process=4] [async] Started save checkpoint @ gs://lance-maxtext/linen_ckpt_xpk_main_20260420_060028/linen_xpk_main_20260420_060028_15_ocdbt_true/checkpoints/9. I0420 07:02:08.147034 137378039891776 jax_array_handlers.py:360] Scheduling D2H of 39 prioritized jax.Array. I0420 07:02:08.147125 137378039891776 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 I0420 07:02:08.179914 137378039891776 base_pytree_checkpoint_handler.py:154] [process=4][thread=MainThread] Initiated "orbax.checkpoint._src.serialization.jax_array_handlers.ArrayHandler".serialize. Time taken: 0.033764s I0420 07:02:08.180046 137378039891776 base_pytree_checkpoint_handler.py:130] [process=4] /jax/orbax/write/blocking_gbytes_per_sec: 41.393 GiB/s (total gbytes: 1.5 GiB) (time elapsed: 0.03726649284362793 s) (per-host) I0420 07:02:08.180090 137378039891776 base_pytree_checkpoint_handler.py:768] [process=4][thread=MainThread] Initiated Pytree async_save. Time taken: 0.037320s (batch_requests_ready=0.001747s, total_serialization_initiated=0.035515s, others=0.000057s) I0420 07:02:08.180165 137378039891776 composite_checkpoint_handler.py:715] [process=4][thread=MainThread] Initiated CompositeCheckpointHandler.async_save. Time taken: 0.041862s (all_items=0.000010s, per_item={'items': '0.00001025'}, temp_paths=0.041852) I0420 07:02:08.180864 137378039891776 event_tracking.py:125] [process=4] [async] Finished blocking save in 0.48 seconds. Continuing save @ gs://lance-maxtext/linen_ckpt_xpk_main_20260420_060028/linen_xpk_main_20260420_060028_15_ocdbt_true/checkpoints/9. I0420 07:02:08.181157 137243082675968 async_checkpointer.py:76] [process=4][thread=async_save] Background save thread started. Deadline for this save operation is 2026-04-20 07:22:08.181123 I0420 07:02:08.190757 137378039891776 checkpoint_manager.py:1560] [process=4][thread=MainThread][step=9] Starting CheckpointManager Save Finalize thread=save_finalize I0420 07:02:08.190998 137227388573440 async_checkpointer.py:280] [process=4][thread=save_finalize] Waiting for background save thread=async_save. I0420 07:02:08.191123 137378039891776 standard_logger.py:34] {'step': 9, 'event_type': 'save', 'directory': 'gs://lance-maxtext/linen_ckpt_xpk_main_20260420_060028/linen_xpk_main_20260420_060028_15_ocdbt_true/checkpoints', 'reached_preemption': False, 'preemption_received_at': None, 'synchronous': False, 'wait_for_prev_start_time': 1776668500.508832, 'wait_for_prev_duration_secs': 27.185792684555054, 'time_between_consecutive_saves_sec': 23.791550636291504, 'checkpointer_blocking_start_time': 1776668527.695986, 'checkpointer_blocking_duration_secs': 0.4853076934814453, 'get_old_steps_start_time': 1776668528.1813123, 'get_old_steps_duration_secs': 3.4332275390625e-05, 'checkpoint_manager_blocking_start_time': 1776668500.50686, 'checkpoint_manager_blocking_duration_secs': 27.68423056602478} I0420 07:02:08.191238 137378039891776 checkpointing.py:408] Started an asynchronous checkpoint save for step 9 I0420 07:02:08.191280 137378039891776 checkpoint_manager.py:2020] [process=4][thread=MainThread][step=9][wait_until_finished] Waiting for Save Finalize thread (save_finalize) to complete. I0420 07:02:14.240790 137248039737088 array_metadata_store.py:203] [process=4][thread=array_type_handler] Wrote 39 array_metadata.ArrayMetadata to gs://lance-maxtext/linen_ckpt_xpk_main_20260420_060028/linen_xpk_main_20260420_060028_15_ocdbt_true/checkpoints/9/items/array_metadatas/process_4 I0420 07:02:49.786872 137243082675968 base_pytree_checkpoint_handler.py:130] [process=4] /jax/orbax/write/gbytes_per_sec: 37.931 MiB/s (total gbytes: 1.5 GiB) (time elapsed: 41.644062757492065 s) (per-host) I0420 07:02:49.786979 137243082675968 async_checkpointer.py:90] [process=4][thread=async_save] 3 Handler Commit operations completed. Time taken: 41.605714s. I0420 07:02:59.087204 137243082675968 async_checkpointer.py:160] [process=4][thread=async_save] Background save thread done. Time taken: 50.905924s. I0420 07:02:59.087443 137227388573440 async_checkpointer.py:288] [process=4][thread=save_finalize] Done with waiting for background save thread=async_save. I0420 07:02:59.087503 137227388573440 async_checkpointer.py:298] [process=4][thread=save_finalize] No errors found in background save thread=async_save. I0420 07:02:59.087543 137227388573440 checkpoint_manager.py:2137] [process=4][thread=save_finalize][step=9] CheckpointManager Save Finalize is syncing with other hosts... I0420 07:02:59.089019 137227388573440 checkpoint_manager.py:2146] [process=4][thread=save_finalize][step=9] CheckpointManager Save Finalize is done on all hosts. I0420 07:02:59.089169 137378039891776 checkpoint_manager.py:2032] [process=4][thread=MainThread][step=9][wait_until_finished] Done waiting for Save Finalize thread (save_finalize) running at step=9. I0420 07:02:59.089316 137378039891776 checkpoint_manager.py:2009] [process=4][thread=MainThread][wait_until_finished] No Save Finalize thread to wait for. Returning. I0420 07:02:59.090348 137378039891776 metric_logger.py:196] completed step: 9, seconds: 0.148, TFLOP/s/device: 92.082, Tokens/s/device: 13879.570, 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: Mon Apr 20 07:03:09 UTC 2026 EXIT_CODE=0
XPK Start: Tue Apr 21 01:53:26 UTC 2026 2026-04-21 01:53:30.704453: E external/local_xla/xla/stream_executor/cuda/cuda_fft.cc:467] Unable to register cuFFT factory: Attempting to register factory for plugin cuFFT when one has already been registered WARNING: All log messages before absl::InitializeLog() is called are written to STDERR E0000 00:00:1776736410.717161 10 cuda_dnn.cc:8579] Unable to register cuDNN factory: Attempting to register factory for plugin cuDNN when one has already been registered E0000 00:00:1776736410.720873 10 cuda_blas.cc:1407] Unable to register cuBLAS factory: Attempting to register factory for plugin cuBLAS when one has already been registered W0000 00:00:1776736410.732161 10 computation_placer.cc:177] computation placer already registered. Please check linkage and avoid linking the same target more than once. W0000 00:00:1776736410.732178 10 computation_placer.cc:177] computation placer already registered. Please check linkage and avoid linking the same target more than once. W0000 00:00:1776736410.732180 10 computation_placer.cc:177] computation placer already registered. Please check linkage and avoid linking the same target more than once. W0000 00:00:1776736410.732182 10 computation_placer.cc:177] computation placer already registered. Please check linkage and avoid linking the same target more than once. 2026-04-21 01:53:49.795250: 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) I0421 01:53:50.315515 138293463746368 max_utils.py:273] Attempting to initialize the jax distributed system... INFO:2026-04-21 01:53:59,357:jax._src.distributed:140: Starting JAX distributed service on [::]:8482 I0421 01:53:59.357281 138293463746368 distributed.py:140] Starting JAX distributed service on [::]:8482 INFO:2026-04-21 01:53:59,359:jax._src.distributed:157: Connecting to JAX distributed service on mt-15-ocdbt-true-sv-ew2rm-slice-job-0-0.mt-15-ocdbt-true-sv-ew2rm:8482 I0421 01:53:59.359772 138293463746368 distributed.py:157] Connecting to JAX distributed service on mt-15-ocdbt-true-sv-ew2rm-slice-job-0-0.mt-15-ocdbt-true-sv-ew2rm:8482 I0421 01:54:00.790592 138293463746368 max_utils.py:284] Jax distributed system initialized! I0421 01:54:06.589093 138293463746368 max_utils.py:800] System Information: Jax Version: 0.8.1 I0421 01:54:06.589204 138293463746368 max_utils.py:801] System Information: Jaxlib Version: 0.8.1 I0421 01:54:06.589246 138293463746368 max_utils.py:802] System Information: Jax Backend: PJRT C API TFRT TPU v6 lite Built on Nov 12 2025 14:16:36 (1762985796) cl/831091709 I0421 01:54:06.589279 138293463746368 train_utils.py:347] WARNING: Sequence packing is essentially ignored for synthetic data. Please use a real dataset to use sequence packing. I0421 01:54:07.209679 138293463746368 maxtext_utils.py:1398] Num_devices: 32, shape (1, 1, 1, 32, 1, 1, 1, 1, 1, 1, 1, 1, 1) I0421 01:54:07.209978 138293463746368 checkpointing.py:677] Setting up checkpoint logger... I0421 01:54:07.210036 138293463746368 checkpointing.py:233] Creating checkpoint manager with ocdbt=True and zarr3=True I0421 01:54:07.210081 138293463746368 pytree_checkpoint_handler.py:589] save_device_host_concurrent_bytes=None I0421 01:54:07.210462 138293463746368 base_pytree_checkpoint_handler.py:415] Created BasePyTreeCheckpointHandler: use_ocdbt=True, use_zarr3=True, pytree_metadata_options=PyTreeMetadataOptions(support_rich_types=False), array_metadata_store=<orbax.checkpoint._src.metadata.array_metadata_store.Store object at 0x7dc66df74fe0>, enable_pinned_host_transfer=False, save_concurrent_bytes: 96000000000 (89.4 GiB), restore_concurrent_bytes: 96000000000 (89.4 GiB) I0421 01:54:10.869566 138293463746368 checkpointing.py:265] Enabling policy for fixed interval checkpointing. I0421 01:54:10.869811 138293463746368 checkpoint_manager.py:709] [process=1][thread=MainThread] CheckpointManager init: checkpointers=None, item_names=('items',), item_handlers={'items': <orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeCheckpointHandler object at 0x7db46ee97d40>}, handler_registry=None I0421 01:54:10.870045 138293463746368 composite_checkpoint_handler.py:237] Deferred registration for item: "items". Adding handler `<orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeCheckpointHandler object at 0x7db46ee97d40>` 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`. I0421 01:54:10.870098 138293463746368 composite_checkpoint_handler.py:237] Deferred registration for item: "metrics". Adding handler `<orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonCheckpointHandler object at 0x7db46ee9bce0>` 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`. I0421 01:54:10.870134 138293463746368 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 0x7db46ee97d40>, ('items', <class 'orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeRestoreArgs'>): <orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeCheckpointHandler object at 0x7db46ee97d40>, ('metrics', <class 'orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonSaveArgs'>): <orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonCheckpointHandler object at 0x7db46ee9bce0>, ('metrics', <class 'orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonRestoreArgs'>): <orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonCheckpointHandler object at 0x7db46ee9bce0>}). I0421 01:54:10.872806 138293463746368 abstract_checkpointer.py:35] orbax-checkpoint version: 0.11.33 I0421 01:54:10.872889 138293463746368 async_checkpointer.py:177] [process=1][thread=MainThread] Using barrier_sync_fn: <function get_barrier_sync_fn.<locals>._fn at 0x7dc6543fb2e0> timeout: 600 secs and primary_host=0 for async checkpoint writes I0421 01:54:11.987227 138293463746368 checkpoint_manager.py:1818] Found 0 checkpoint steps in gs://lance-maxtext/linen_ckpt_xpk_test_pipeline_scan_nnx_20260421_005327/linen_xpk_test_pipeline_scan_nnx_20260421_005327_15_ocdbt_true/checkpoints I0421 01:54:12.472015 138293463746368 checkpoint_manager.py:929] [process=1][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_20260421_005327/linen_xpk_test_pipeline_scan_nnx_20260421_005327_15_ocdbt_true/checkpoints: <orbax.checkpoint.checkpoint_manager.CheckpointManager object at 0x7db46ee9bdd0> I0421 01:54:12.472203 138293463746368 checkpointing.py:301] Checkpoint manager created! I0421 01:54:13.356527 138293463746368 nnx_wrappers.py:453] Unknown Logical: bfloat16[32,2048,2048]...................................... ('activation_batch', 'activation_norm_length', 'activation_embed'). I0421 01:54:13.356632 138293463746368 nnx_wrappers.py:453] Unknown Physical: bfloat16[32,2048,2048]...................................... ('fsdp', None, None). I0421 01:54:13.725012 138293463746368 attentions.py:1084] attentions/inputs_q Logical: bfloat16[32,2048,2048]...................................... ('activation_batch', 'activation_attn_length', 'activation_attn_embed'). I0421 01:54:13.725111 138293463746368 attentions.py:1084] attentions/inputs_q Physical: bfloat16[32,2048,2048]...................................... ('fsdp', None, None). I0421 01:54:13.740769 138293463746368 attentions.py:1085] attentions/inputs_kv Logical: bfloat16[32,2048,2048]...................................... ('activation_batch', 'activation_attn_length', 'activation_attn_embed'). I0421 01:54:13.740829 138293463746368 attentions.py:1085] attentions/inputs_kv Physical: bfloat16[32,2048,2048]...................................... ('fsdp', None, None). I0421 01:54:13.764076 138293463746368 attentions.py:1150] attentions/query Logical: bfloat16[32,2048,16,128].................................... ('activation_kv_batch', 'activation_attn_length', 'activation_kv_heads', 'activation_kv_head_dim'). I0421 01:54:13.764150 138293463746368 attentions.py:1150] attentions/query Physical: bfloat16[32,2048,16,128].................................... ('fsdp', None, None, None). I0421 01:54:13.779956 138293463746368 attentions.py:1151] attentions/key Logical: bfloat16[32,2048,16,128].................................... ('activation_kv_batch', 'activation_attn_length', 'activation_kv_heads', 'activation_kv_head_dim'). I0421 01:54:13.780018 138293463746368 attentions.py:1151] attentions/key Physical: bfloat16[32,2048,16,128].................................... ('fsdp', None, None, None). I0421 01:54:13.795814 138293463746368 attentions.py:1152] attentions/value Logical: bfloat16[32,2048,16,128].................................... ('activation_kv_batch', 'activation_attn_length', 'activation_kv_heads', 'activation_kv_head_dim'). I0421 01:54:13.795874 138293463746368 attentions.py:1152] attentions/value Physical: bfloat16[32,2048,16,128].................................... ('fsdp', None, None, None). I0421 01:54:13.819816 138293463746368 attentions.py:1193] attentions/out Logical: bfloat16[32,2048,16,128].................................... ('activation_batch', 'activation_attn_length', 'activation_heads', 'activation_kv'). I0421 01:54:13.819901 138293463746368 attentions.py:1193] attentions/out Physical: bfloat16[32,2048,16,128].................................... ('fsdp', None, None, None). I0421 01:54:13.839377 138293463746368 linears.py:525] linears/x Logical: bfloat16[32,2048,7168]...................................... ('activation_batch', 'activation_length', 'activation_mlp'). I0421 01:54:13.839452 138293463746368 linears.py:525] linears/x Physical: bfloat16[32,2048,7168]...................................... ('fsdp', None, None). I0421 01:54:14.028310 138293463746368 checkpointing.py:577] checkpoint manager exists so trying to load this run's existing checkpoint I0421 01:54:14.028430 138293463746368 checkpointing.py:665] No existing checkpoints found, not restoring checkpoint. fsdp: 32 I0421 01:54:15.444854 138293463746368 maxtext_utils.py:1501] params/params/decoder/decoder_norm/scale Shape: float32[2048] Logical: PartitionSpec('norm',) Physical: (None,) I0421 01:54:15.444986 138293463746368 maxtext_utils.py:1501] params/params/decoder/layers/mlp/wi_0/kernel Shape: float32[2048,16,7168] Logical: PartitionSpec('embed', 'layers', 'mlp') Physical: ('fsdp', None, None) I0421 01:54:15.445041 138293463746368 maxtext_utils.py:1501] params/params/decoder/layers/mlp/wi_1/kernel Shape: float32[2048,16,7168] Logical: PartitionSpec('embed', 'layers', 'mlp') Physical: ('fsdp', None, None) I0421 01:54:15.445101 138293463746368 maxtext_utils.py:1501] params/params/decoder/layers/mlp/wo/kernel Shape: float32[7168,16,2048] Logical: PartitionSpec('mlp', 'layers', 'embed') Physical: (None, None, 'fsdp') I0421 01:54:15.445153 138293463746368 maxtext_utils.py:1501] params/params/decoder/layers/post_self_attention_layer_norm/scale Shape: float32[2048,16] Logical: PartitionSpec('norm', 'layers') Physical: (None, None) I0421 01:54:15.445194 138293463746368 maxtext_utils.py:1501] params/params/decoder/layers/pre_self_attention_layer_norm/scale Shape: float32[2048,16] Logical: PartitionSpec('norm', 'layers') Physical: (None, None) I0421 01:54:15.445246 138293463746368 maxtext_utils.py:1501] params/params/decoder/layers/self_attention/key/kernel Shape: float32[2048,16,16,128] Logical: PartitionSpec('embed', 'layers', 'kv_heads', 'kv_head_dim') Physical: ('fsdp', None, None, None) I0421 01:54:15.445297 138293463746368 maxtext_utils.py:1501] params/params/decoder/layers/self_attention/out/kernel Shape: float32[16,16,128,2048] Logical: PartitionSpec('heads', 'layers', 'kv', 'embed') Physical: (None, None, None, 'fsdp') I0421 01:54:15.445336 138293463746368 maxtext_utils.py:1501] params/params/decoder/layers/self_attention/query/kernel Shape: float32[2048,16,16,128] Logical: PartitionSpec('embed', 'layers', 'q_heads', 'kv') Physical: ('fsdp', None, None, None) I0421 01:54:15.445371 138293463746368 maxtext_utils.py:1501] params/params/decoder/layers/self_attention/value/kernel Shape: float32[2048,16,16,128] Logical: PartitionSpec('embed', 'layers', 'kv_heads', 'kv_head_dim') Physical: ('fsdp', None, None, None) I0421 01:54:15.445432 138293463746368 maxtext_utils.py:1501] params/params/decoder/logits_dense/kernel Shape: float32[2048,32000] Logical: PartitionSpec('embed', 'vocab') Physical: ('fsdp', None) I0421 01:54:15.445482 138293463746368 maxtext_utils.py:1501] params/params/token_embedder/embedding Shape: float32[32000,2048] Logical: PartitionSpec('vocab', 'embed') Physical: (None, 'fsdp') I0421 01:54:15.947529 138293463746368 train.py:156] train/xent Logical: float32[32,2048]............................................ ('activation_embed_and_logits_batch', 'activation_length'). I0421 01:54:15.947626 138293463746368 train.py:156] train/xent Physical: float32[32,2048]............................................ ('fsdp', None). I0421 01:54:15.962388 138293463746368 train.py:163] train/z_loss Logical: float32[32,2048]............................................ ('activation_embed_and_logits_batch', 'activation_length'). I0421 01:54:15.962463 138293463746368 train.py:163] train/z_loss Physical: float32[32,2048]............................................ ('fsdp', None). I0421 01:54:26.682219 138293463746368 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. I0421 01:54:26.683108 138293463746368 metric_logger.py:289] number parameters: 1.104 billion I0421 01:54:37.566020 138293463746368 checkpointing.py:772] Waiting for step 0 to finish before checkpoint... I0421 01:54:38.040543 138293463746368 checkpointing.py:776] Waited 0.474506139755249 seconds for step 0 to finish before starting checkpointing. I0421 01:54:38.042996 138293463746368 checkpoint_manager.py:2013] [process=1][thread=MainThread][wait_until_finished] No Save Finalize thread to wait for. Returning. I0421 01:54:38.044659 138293463746368 checkpoint_manager.py:1518] [process=1] Saving checkpoint at step 0 I0421 01:54:38.047514 138293463746368 async_checkpointer.py:452] [process=1] Started async saving checkpoint to gs://lance-maxtext/linen_ckpt_xpk_test_pipeline_scan_nnx_20260421_005327/linen_xpk_test_pipeline_scan_nnx_20260421_005327_15_ocdbt_true/checkpoints/0. I0421 01:54:39.055514 138293463746368 signaling_client.py:364] Using JaxDistributedSignalingClient I0421 01:54:39.056456 138293463746368 jax_array_handlers.py:358] Scheduling D2H of 39 prioritized jax.Array. I0421 01:54:39.056514 138293463746368 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 I0421 01:54:39.299499 138293463746368 base_pytree_checkpoint_handler.py:153] [process=1][thread=MainThread] Initiated "orbax.checkpoint._src.serialization.jax_array_handlers.ArrayHandler".serialize. Time taken: 0.244066s I0421 01:54:39.299664 138293463746368 base_pytree_checkpoint_handler.py:129] [process=1] /jax/checkpoint/write/blocking_gbytes_per_sec: 3.059 GiB/s (total gbytes: 1.5 GiB) (time elapsed: 0.5042135715484619 s) (per-host) I0421 01:54:39.299713 138293463746368 base_pytree_checkpoint_handler.py:737] [process=1][thread=MainThread] Initiated Pytree async_save. Time taken: 0.504272s (batch_requests_ready=0.256583s, total_serialization_initiated=0.247622s, others=0.000066s) I0421 01:54:39.299801 138293463746368 composite_checkpoint_handler.py:715] [process=1][thread=MainThread] Initiated CompositeCheckpointHandler.async_save. Time taken: 0.508554s (all_items=0.000020s, per_item={'items': '0.00001955'}, temp_paths=0.508534) I0421 01:54:39.300707 138172830828288 async_checkpointer.py:79] [process=1][thread=async_save] Background save thread started. I0421 01:54:39.300875 138293463746368 async_checkpointer.py:561] Finished blocking save. Time taken: 1.256148s. Continuing background save to gs://lance-maxtext/linen_ckpt_xpk_test_pipeline_scan_nnx_20260421_005327/linen_xpk_test_pipeline_scan_nnx_20260421_005327_15_ocdbt_true/checkpoints/0. I0421 01:54:39.749988 138293463746368 checkpoint_manager.py:1566] [process=1][thread=MainThread][step=0] Starting CheckpointManager Save Finalize thread=save_finalize I0421 01:54:39.750376 138169525708544 async_checkpointer.py:265] [process=1][thread=save_finalize] Waiting for background save thread=async_save. I0421 01:54:39.750554 138293463746368 standard_logger.py:34] {'step': 0, 'event_type': 'save', 'directory': 'gs://lance-maxtext/linen_ckpt_xpk_test_pipeline_scan_nnx_20260421_005327/linen_xpk_test_pipeline_scan_nnx_20260421_005327_15_ocdbt_true/checkpoints', 'reached_preemption': False, 'preemption_received_at': None, 'synchronous': False, 'wait_for_prev_start_time': 1776736478.0429783, 'wait_for_prev_duration_secs': 6.341934204101562e-05, 'time_between_consecutive_saves_sec': None, 'checkpointer_blocking_start_time': 1776736478.0446975, 'checkpointer_blocking_duration_secs': 1.2562782764434814, 'get_old_steps_start_time': 1776736479.3009963, 'get_old_steps_duration_secs': 2.5033950805664062e-05, 'checkpoint_manager_blocking_start_time': 1776736478.04112, 'checkpoint_manager_blocking_duration_secs': 1.7093915939331055} I0421 01:54:39.750662 138293463746368 checkpointing.py:408] Started an asynchronous checkpoint save for step 0 I0421 01:54:39.750715 138293463746368 max_utils.py:750] Memstats: After params initialized: I0421 01:54:39.750771 138293463746368 max_utils.py:756] Using (GB) 0.43 / 31.25 (1.376000%) on TPU_2(process=1,(2,0,0,0)) I0421 01:54:39.750806 138293463746368 max_utils.py:756] Using (GB) 0.43 / 31.25 (1.376000%) on TPU_3(process=1,(3,0,0,0)) I0421 01:54:39.750834 138293463746368 max_utils.py:756] Using (GB) 0.43 / 31.25 (1.376000%) on TPU_6(process=1,(2,1,0,0)) I0421 01:54:39.750859 138293463746368 max_utils.py:756] Using (GB) 0.43 / 31.25 (1.376000%) on TPU_7(process=1,(3,1,0,0)) I0421 01:54:40.015197 138293463746368 metric_logger.py:185] completed step: 0, seconds: 10.883, TFLOP/s/device: 1.248, Tokens/s/device: 188.187, total_weights: 65536, loss: 10.874 I0421 01:54:40.189779 138293463746368 metric_logger.py:185] completed step: 1, seconds: 2.448, TFLOP/s/device: 5.551, Tokens/s/device: 836.643, total_weights: 65536, loss: 10.874 I0421 01:54:40.564893 138293463746368 metric_logger.py:185] completed step: 2, seconds: 0.027, TFLOP/s/device: 498.482, Tokens/s/device: 75136.662, total_weights: 65536, loss: 10.560 I0421 01:54:40.712358 138293463746368 metric_logger.py:185] completed step: 3, seconds: 0.376, TFLOP/s/device: 36.168, Tokens/s/device: 5451.593, total_weights: 65536, loss: 9.980 I0421 01:54:40.716828 138293463746368 checkpointing.py:772] Waiting for step 5 to finish before checkpoint... I0421 01:54:41.007159 138293463746368 checkpointing.py:776] Waited 0.2902998924255371 seconds for step 5 to finish before starting checkpointing. I0421 01:54:41.010005 138293463746368 checkpoint_manager.py:2024] [process=1][thread=MainThread][step=0][wait_until_finished] Waiting for Save Finalize thread (save_finalize) to complete. I0421 01:54:45.986218 2549 google_auth_provider.cc:181] Running on GCE, using service account 562977990677-compute@developer.gserviceaccount.com I0421 01:54:49.154915 138172839220992 array_metadata_store.py:203] [process=1][thread=array_type_handler] Wrote 39 array_metadata.ArrayMetadata to gs://lance-maxtext/linen_ckpt_xpk_test_pipeline_scan_nnx_20260421_005327/linen_xpk_test_pipeline_scan_nnx_20260421_005327_15_ocdbt_true/checkpoints/0/items/array_metadatas/process_1 I0421 01:55:19.525155 138172830828288 base_pytree_checkpoint_handler.py:129] [process=1] /jax/checkpoint/write/gbytes_per_sec: 38.782 MiB/s (total gbytes: 1.5 GiB) (time elapsed: 40.729655742645264 s) (per-host) I0421 01:55:19.525270 138172830828288 async_checkpointer.py:90] [process=1][thread=async_save] 3 Handler Commit operations completed. Time taken: 40.224452s. I0421 01:55:29.582442 138172830828288 async_checkpointer.py:144] [process=1][thread=async_save] Background save thread done. Time taken: 50.281607s. I0421 01:55:29.582727 138169525708544 async_checkpointer.py:273] [process=1][thread=save_finalize] Done with waiting for background save thread=async_save. I0421 01:55:29.582845 138169525708544 async_checkpointer.py:283] [process=1][thread=save_finalize] No errors found in background save thread=async_save. I0421 01:55:29.582894 138169525708544 checkpoint_manager.py:2133] [process=1][thread=save_finalize][step=0] CheckpointManager Save Finalize is syncing with other hosts... I0421 01:55:29.585823 138169525708544 checkpoint_manager.py:2142] [process=1][thread=save_finalize][step=0] CheckpointManager Save Finalize is done on all hosts. I0421 01:55:29.586009 138293463746368 checkpoint_manager.py:2036] [process=1][thread=MainThread][step=0][wait_until_finished] Done waiting for Save Finalize thread (save_finalize) running at step=0. W0421 01:55:29.586134 138293463746368 checkpoint_manager.py:1458] Waiting for previous save to complete took 48.576131 seconds. If this number is high, consider checkpointing less frequently. I0421 01:55:29.588237 138293463746368 checkpoint_manager.py:1518] [process=1] Saving checkpoint at step 5 I0421 01:55:29.592363 138293463746368 async_checkpointer.py:452] [process=1] Started async saving checkpoint to gs://lance-maxtext/linen_ckpt_xpk_test_pipeline_scan_nnx_20260421_005327/linen_xpk_test_pipeline_scan_nnx_20260421_005327_15_ocdbt_true/checkpoints/5. I0421 01:55:30.965107 138293463746368 jax_array_handlers.py:358] Scheduling D2H of 39 prioritized jax.Array. I0421 01:55:30.965210 138293463746368 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 I0421 01:55:31.005047 138293463746368 base_pytree_checkpoint_handler.py:153] [process=1][thread=MainThread] Initiated "orbax.checkpoint._src.serialization.jax_array_handlers.ArrayHandler".serialize. Time taken: 0.041104s I0421 01:55:31.005210 138293463746368 base_pytree_checkpoint_handler.py:129] [process=1] /jax/checkpoint/write/blocking_gbytes_per_sec: 5.317 GiB/s (total gbytes: 1.5 GiB) (time elapsed: 0.29010701179504395 s) (per-host) I0421 01:55:31.005261 138293463746368 base_pytree_checkpoint_handler.py:737] [process=1][thread=MainThread] Initiated Pytree async_save. Time taken: 0.290167s (batch_requests_ready=0.247176s, total_serialization_initiated=0.042924s, others=0.000067s) I0421 01:55:31.005346 138293463746368 composite_checkpoint_handler.py:715] [process=1][thread=MainThread] Initiated CompositeCheckpointHandler.async_save. Time taken: 0.295236s (all_items=0.000016s, per_item={'items': '0.00001574'}, temp_paths=0.295221) I0421 01:55:31.006189 138166836750080 async_checkpointer.py:79] [process=1][thread=async_save] Background save thread started. I0421 01:55:31.006312 138293463746368 async_checkpointer.py:561] Finished blocking save. Time taken: 1.418008s. Continuing background save to gs://lance-maxtext/linen_ckpt_xpk_test_pipeline_scan_nnx_20260421_005327/linen_xpk_test_pipeline_scan_nnx_20260421_005327_15_ocdbt_true/checkpoints/5. I0421 01:55:31.451194 138293463746368 checkpoint_manager.py:1566] [process=1][thread=MainThread][step=5] Starting CheckpointManager Save Finalize thread=save_finalize I0421 01:55:31.451579 138169525708544 async_checkpointer.py:265] [process=1][thread=save_finalize] Waiting for background save thread=async_save. I0421 01:55:31.451728 138293463746368 standard_logger.py:34] {'step': 5, 'event_type': 'save', 'directory': 'gs://lance-maxtext/linen_ckpt_xpk_test_pipeline_scan_nnx_20260421_005327/linen_xpk_test_pipeline_scan_nnx_20260421_005327_15_ocdbt_true/checkpoints', 'reached_preemption': False, 'preemption_received_at': None, 'synchronous': False, 'wait_for_prev_start_time': 1776736481.0099788, 'wait_for_prev_duration_secs': 48.576130628585815, 'time_between_consecutive_saves_sec': None, 'checkpointer_blocking_start_time': 1776736529.5882761, 'checkpointer_blocking_duration_secs': 1.4182090759277344, 'get_old_steps_start_time': 1776736531.0065038, 'get_old_steps_duration_secs': 2.765655517578125e-05, 'checkpoint_manager_blocking_start_time': 1776736481.0077825, 'checkpoint_manager_blocking_duration_secs': 50.44390940666199} I0421 01:55:31.451907 138293463746368 checkpointing.py:408] Started an asynchronous checkpoint save for step 5 I0421 01:55:31.452540 138293463746368 metric_logger.py:185] completed step: 4, seconds: 0.153, TFLOP/s/device: 89.094, Tokens/s/device: 13429.156, total_weights: 65536, loss: 9.460 I0421 01:55:31.458174 138293463746368 metric_logger.py:185] completed step: 5, seconds: 0.147, TFLOP/s/device: 92.183, Tokens/s/device: 13894.825, total_weights: 65536, loss: 8.959 I0421 01:55:52.784470 138293463746368 metric_logger.py:185] completed step: 6, seconds: 50.740, TFLOP/s/device: 0.268, Tokens/s/device: 40.362, total_weights: 65536, loss: 8.469 I0421 01:55:52.931924 138293463746368 metric_logger.py:185] completed step: 7, seconds: 21.180, TFLOP/s/device: 0.642, Tokens/s/device: 96.695, total_weights: 65536, loss: 8.003 I0421 01:55:53.079352 138293463746368 metric_logger.py:185] completed step: 8, seconds: 0.152, TFLOP/s/device: 89.291, Tokens/s/device: 13458.986, total_weights: 65536, loss: 7.572 I0421 01:55:53.226673 138293463746368 checkpointing.py:772] Waiting for step 9 to finish before checkpoint... I0421 01:55:53.227284 138293463746368 checkpointing.py:776] Waited 0.0006320476531982422 seconds for step 9 to finish before starting checkpointing. I0421 01:55:53.229377 138293463746368 checkpoint_manager.py:2024] [process=1][thread=MainThread][step=5][wait_until_finished] Waiting for Save Finalize thread (save_finalize) to complete. I0421 01:55:59.174887 138172839220992 array_metadata_store.py:203] [process=1][thread=array_type_handler] Wrote 39 array_metadata.ArrayMetadata to gs://lance-maxtext/linen_ckpt_xpk_test_pipeline_scan_nnx_20260421_005327/linen_xpk_test_pipeline_scan_nnx_20260421_005327_15_ocdbt_true/checkpoints/5/items/array_metadatas/process_1 I0421 01:56:34.314941 138166836750080 base_pytree_checkpoint_handler.py:129] [process=1] /jax/checkpoint/write/gbytes_per_sec: 24.837 MiB/s (total gbytes: 1.5 GiB) (time elapsed: 63.59978795051575 s) (per-host) I0421 01:56:34.315060 138166836750080 async_checkpointer.py:90] [process=1][thread=async_save] 3 Handler Commit operations completed. Time taken: 63.308673s. I0421 01:56:44.460165 138166836750080 async_checkpointer.py:144] [process=1][thread=async_save] Background save thread done. Time taken: 73.453762s. I0421 01:56:44.460503 138169525708544 async_checkpointer.py:273] [process=1][thread=save_finalize] Done with waiting for background save thread=async_save. I0421 01:56:44.460629 138169525708544 async_checkpointer.py:283] [process=1][thread=save_finalize] No errors found in background save thread=async_save. I0421 01:56:44.460679 138169525708544 checkpoint_manager.py:2133] [process=1][thread=save_finalize][step=5] CheckpointManager Save Finalize is syncing with other hosts... I0421 01:56:44.462246 138169525708544 checkpoint_manager.py:2142] [process=1][thread=save_finalize][step=5] CheckpointManager Save Finalize is done on all hosts. I0421 01:56:44.462460 138293463746368 checkpoint_manager.py:2036] [process=1][thread=MainThread][step=5][wait_until_finished] Done waiting for Save Finalize thread (save_finalize) running at step=5. W0421 01:56:44.462597 138293463746368 checkpoint_manager.py:1458] Waiting for previous save to complete took 51.233230 seconds. If this number is high, consider checkpointing less frequently. I0421 01:56:44.464245 138293463746368 checkpoint_manager.py:1518] [process=1] Saving checkpoint at step 9 I0421 01:56:44.467742 138293463746368 async_checkpointer.py:452] [process=1] Started async saving checkpoint to gs://lance-maxtext/linen_ckpt_xpk_test_pipeline_scan_nnx_20260421_005327/linen_xpk_test_pipeline_scan_nnx_20260421_005327_15_ocdbt_true/checkpoints/9. I0421 01:56:45.402169 138293463746368 jax_array_handlers.py:358] Scheduling D2H of 39 prioritized jax.Array. I0421 01:56:45.402268 138293463746368 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 I0421 01:56:45.443747 138293463746368 base_pytree_checkpoint_handler.py:153] [process=1][thread=MainThread] Initiated "orbax.checkpoint._src.serialization.jax_array_handlers.ArrayHandler".serialize. Time taken: 0.042464s I0421 01:56:45.443917 138293463746368 base_pytree_checkpoint_handler.py:129] [process=1] /jax/checkpoint/write/blocking_gbytes_per_sec: 5.351 GiB/s (total gbytes: 1.5 GiB) (time elapsed: 0.2882661819458008 s) (per-host) I0421 01:56:45.443967 138293463746368 base_pytree_checkpoint_handler.py:737] [process=1][thread=MainThread] Initiated Pytree async_save. Time taken: 0.288328s (batch_requests_ready=0.243822s, total_serialization_initiated=0.044439s, others=0.000067s) I0421 01:56:45.444055 138293463746368 composite_checkpoint_handler.py:715] [process=1][thread=MainThread] Initiated CompositeCheckpointHandler.async_save. Time taken: 0.292452s (all_items=0.000012s, per_item={'items': '0.00001168'}, temp_paths=0.292440) I0421 01:56:45.444893 138170337203968 async_checkpointer.py:79] [process=1][thread=async_save] Background save thread started. I0421 01:56:45.445053 138293463746368 async_checkpointer.py:561] Finished blocking save. Time taken: 0.980737s. Continuing background save to gs://lance-maxtext/linen_ckpt_xpk_test_pipeline_scan_nnx_20260421_005327/linen_xpk_test_pipeline_scan_nnx_20260421_005327_15_ocdbt_true/checkpoints/9. I0421 01:56:45.874259 138293463746368 checkpoint_manager.py:1566] [process=1][thread=MainThread][step=9] Starting CheckpointManager Save Finalize thread=save_finalize I0421 01:56:45.874665 138169525708544 async_checkpointer.py:265] [process=1][thread=save_finalize] Waiting for background save thread=async_save. I0421 01:56:45.874814 138293463746368 standard_logger.py:34] {'step': 9, 'event_type': 'save', 'directory': 'gs://lance-maxtext/linen_ckpt_xpk_test_pipeline_scan_nnx_20260421_005327/linen_xpk_test_pipeline_scan_nnx_20260421_005327_15_ocdbt_true/checkpoints', 'reached_preemption': False, 'preemption_received_at': None, 'synchronous': False, 'wait_for_prev_start_time': 1776736553.2293494, 'wait_for_prev_duration_secs': 51.23323035240173, 'time_between_consecutive_saves_sec': 23.64347767829895, 'checkpointer_blocking_start_time': 1776736604.4642844, 'checkpointer_blocking_duration_secs': 0.98087477684021, 'get_old_steps_start_time': 1776736605.4451723, 'get_old_steps_duration_secs': 2.4318695068359375e-05, 'checkpoint_manager_blocking_start_time': 1776736553.2275312, 'checkpoint_manager_blocking_duration_secs': 52.6472487449646} I0421 01:56:45.874979 138293463746368 checkpointing.py:408] Started an asynchronous checkpoint save for step 9 I0421 01:56:45.875022 138293463746368 checkpoint_manager.py:2024] [process=1][thread=MainThread][step=9][wait_until_finished] Waiting for Save Finalize thread (save_finalize) to complete. I0421 01:56:51.521051 138172839220992 array_metadata_store.py:203] [process=1][thread=array_type_handler] Wrote 39 array_metadata.ArrayMetadata to gs://lance-maxtext/linen_ckpt_xpk_test_pipeline_scan_nnx_20260421_005327/linen_xpk_test_pipeline_scan_nnx_20260421_005327_15_ocdbt_true/checkpoints/9/items/array_metadatas/process_1 I0421 01:57:27.421836 138170337203968 base_pytree_checkpoint_handler.py:129] [process=1] /jax/checkpoint/write/gbytes_per_sec: 37.373 MiB/s (total gbytes: 1.5 GiB) (time elapsed: 42.26614713668823 s) (per-host) I0421 01:57:27.421958 138170337203968 async_checkpointer.py:90] [process=1][thread=async_save] 3 Handler Commit operations completed. Time taken: 41.976954s. I0421 01:57:35.517319 138170337203968 async_checkpointer.py:144] [process=1][thread=async_save] Background save thread done. Time taken: 50.072300s. I0421 01:57:35.517630 138169525708544 async_checkpointer.py:273] [process=1][thread=save_finalize] Done with waiting for background save thread=async_save. I0421 01:57:35.517751 138169525708544 async_checkpointer.py:283] [process=1][thread=save_finalize] No errors found in background save thread=async_save. I0421 01:57:35.517791 138169525708544 checkpoint_manager.py:2133] [process=1][thread=save_finalize][step=9] CheckpointManager Save Finalize is syncing with other hosts... I0421 01:57:35.519261 138169525708544 checkpoint_manager.py:2142] [process=1][thread=save_finalize][step=9] CheckpointManager Save Finalize is done on all hosts. I0421 01:57:35.519455 138293463746368 checkpoint_manager.py:2036] [process=1][thread=MainThread][step=9][wait_until_finished] Done waiting for Save Finalize thread (save_finalize) running at step=9. I0421 01:57:35.519605 138293463746368 checkpoint_manager.py:2013] [process=1][thread=MainThread][wait_until_finished] No Save Finalize thread to wait for. Returning. I0421 01:57:35.520219 138293463746368 metric_logger.py:185] completed step: 9, seconds: 0.148, TFLOP/s/device: 92.111, Tokens/s/device: 13883.993, 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: Tue Apr 21 01:57:47 UTC 2026 EXIT_CODE=0