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 | 102.484 | 92.283 | -10.201 |
| Tok/s | 15447.5 | 13909.9 | -1537.585 |
| Avg s/step | 7.562 | 7.729 | +0.167 |
| 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 06:22:32 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 06:22:56.205397: 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 06:22:56.733500 139490510141248 max_utils.py:273] Attempting to initialize the jax distributed system... I0420 06:23:05.776466 139490510141248 distributed.py:149] Starting JAX distributed service on [::]:8482 I0420 06:23:05.778829 139490510141248 distributed.py:172] Connecting to JAX distributed service on mt-08-checkpoint-async--2jpgu-slice-job-0-0.mt-08-checkpoint-async--2jpgu:8482 I0420 06:23:06.969423 139490510141248 max_utils.py:284] Jax distributed system initialized! I0420 06:23:12.954278 139490510141248 max_utils.py:800] System Information: Jax Version: 0.9.2 I0420 06:23:12.954394 139490510141248 max_utils.py:801] System Information: Jaxlib Version: 0.9.2 I0420 06:23:12.954436 139490510141248 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 06:23:12.954469 139490510141248 train_utils.py:348] WARNING: Sequence packing is essentially ignored for synthetic data. Please use a real dataset to use sequence packing. I0420 06:23:13.644230 139490510141248 maxtext_utils.py:1551] Num_devices: 32, shape (1, 1, 1, 32, 1, 1, 1, 1, 1, 1, 1, 1, 1) I0420 06:23:13.644564 139490510141248 checkpointing.py:677] Setting up checkpoint logger... I0420 06:23:13.644624 139490510141248 checkpointing.py:233] Creating checkpoint manager with ocdbt=True and zarr3=True I0420 06:23:13.644671 139490510141248 pytree_checkpoint_handler.py:592] save_device_host_concurrent_bytes=None I0420 06:23:13.645008 139490510141248 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 0x7edcf29482c0>, enable_pinned_host_transfer=False, save_concurrent_bytes: 96000000000 (89.4 GiB), restore_concurrent_bytes: 96000000000 (89.4 GiB) I0420 06:23:16.476360 139490510141248 checkpointing.py:265] Enabling policy for fixed interval checkpointing. I0420 06:23:16.476594 139490510141248 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 0x7ec8e4343860>}, handler_registry=None I0420 06:23:16.476826 139490510141248 composite_checkpoint_handler.py:237] Deferred registration for item: "items". Adding handler `<orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeCheckpointHandler object at 0x7ec8e4343860>` 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 06:23:16.476873 139490510141248 composite_checkpoint_handler.py:237] Deferred registration for item: "metrics". Adding handler `<orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonCheckpointHandler object at 0x7ec8e43455e0>` 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 06:23:16.476908 139490510141248 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 0x7ec8e4343860>, ('items', <class 'orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeRestoreArgs'>): <orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeCheckpointHandler object at 0x7ec8e4343860>, ('metrics', <class 'orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonSaveArgs'>): <orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonCheckpointHandler object at 0x7ec8e43455e0>, ('metrics', <class 'orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonRestoreArgs'>): <orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonCheckpointHandler object at 0x7ec8e43455e0>}). I0420 06:23:16.477223 139490510141248 abstract_checkpointer.py:35] orbax-checkpoint version: 0.11.34 I0420 06:23:16.477291 139490510141248 async_checkpointer.py:192] [process=4][thread=MainThread] Using barrier_sync_fn: <function get_barrier_sync_fn.<locals>._fn at 0x7ec8cc6c7600> timeout: 1200 secs and primary_host=0 for async checkpoint writes I0420 06:23:17.720064 139490510141248 checkpoint_manager.py:1812] Found 0 checkpoint steps in gs://lance-maxtext/linen_ckpt_xpk_main_20260420_060028/linen_xpk_main_20260420_060028_08_checkpoint_async_true/checkpoints I0420 06:23:17.741835 139490510141248 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_08_checkpoint_async_true/checkpoints: <orbax.checkpoint.checkpoint_manager.CheckpointManager object at 0x7ec8e4346e10> I0420 06:23:17.741957 139490510141248 checkpointing.py:301] Checkpoint manager created! I0420 06:23:18.663979 139490510141248 nnx_wrappers.py:437] Unknown Logical: bfloat16[32,2048,2048]...................................... ('activation_batch', 'activation_norm_length', 'activation_embed'). I0420 06:23:18.664089 139490510141248 nnx_wrappers.py:437] Unknown Physical: bfloat16[32,2048,2048]...................................... ('fsdp', None, None). I0420 06:23:19.041111 139490510141248 attentions.py:1088] attentions/inputs_q Logical: bfloat16[32,2048,2048]...................................... ('activation_batch', 'activation_attn_length', 'activation_attn_embed'). I0420 06:23:19.041203 139490510141248 attentions.py:1088] attentions/inputs_q Physical: bfloat16[32,2048,2048]...................................... ('fsdp', None, None). I0420 06:23:19.057490 139490510141248 attentions.py:1089] attentions/inputs_kv Logical: bfloat16[32,2048,2048]...................................... ('activation_batch', 'activation_attn_length', 'activation_attn_embed'). I0420 06:23:19.057558 139490510141248 attentions.py:1089] attentions/inputs_kv Physical: bfloat16[32,2048,2048]...................................... ('fsdp', None, None). I0420 06:23:19.081776 139490510141248 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 06:23:19.081845 139490510141248 attentions.py:1154] attentions/query Physical: bfloat16[32,2048,16,128].................................... ('fsdp', None, None, None). I0420 06:23:19.098102 139490510141248 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 06:23:19.098162 139490510141248 attentions.py:1155] attentions/key Physical: bfloat16[32,2048,16,128].................................... ('fsdp', None, None, None). I0420 06:23:19.114362 139490510141248 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 06:23:19.114422 139490510141248 attentions.py:1156] attentions/value Physical: bfloat16[32,2048,16,128].................................... ('fsdp', None, None, None). I0420 06:23:19.138797 139490510141248 attentions.py:1197] attentions/out Logical: bfloat16[32,2048,16,128].................................... ('activation_batch', 'activation_attn_length', 'activation_heads', 'activation_kv'). I0420 06:23:19.138876 139490510141248 attentions.py:1197] attentions/out Physical: bfloat16[32,2048,16,128].................................... ('fsdp', None, None, None). I0420 06:23:19.159745 139490510141248 linears.py:525] linears/x Logical: bfloat16[32,2048,7168]...................................... ('activation_batch', 'activation_length', 'activation_mlp'). I0420 06:23:19.159807 139490510141248 linears.py:525] linears/x Physical: bfloat16[32,2048,7168]...................................... ('fsdp', None, None). I0420 06:23:19.374869 139490510141248 checkpointing.py:577] checkpoint manager exists so trying to load this run's existing checkpoint I0420 06:23:19.374974 139490510141248 checkpointing.py:665] No existing checkpoints found, not restoring checkpoint. fsdp: 32 I0420 06:23:20.783277 139490510141248 maxtext_utils.py:1654] params/params/decoder/decoder_norm/scale Shape: float32[2048] Logical: P('norm',) Physical: (None,) I0420 06:23:20.783472 139490510141248 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 06:23:20.783528 139490510141248 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 06:23:20.783590 139490510141248 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 06:23:20.783645 139490510141248 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 06:23:20.783683 139490510141248 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 06:23:20.783735 139490510141248 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 06:23:20.783785 139490510141248 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 06:23:20.783824 139490510141248 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 06:23:20.783860 139490510141248 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 06:23:20.783906 139490510141248 maxtext_utils.py:1654] params/params/decoder/logits_dense/kernel Shape: float32[2048,32000] Logical: P('embed_vocab', 'vocab') Physical: ('fsdp', None) I0420 06:23:20.783952 139490510141248 maxtext_utils.py:1654] params/params/token_embedder/embedding Shape: float32[32000,2048] Logical: P('vocab', 'embed_vocab') Physical: (None, 'fsdp') I0420 06:23:21.271374 139490510141248 train.py:155] train/xent Logical: float32[32,2048]............................................ ('activation_embed_and_logits_batch', 'activation_length'). I0420 06:23:21.271467 139490510141248 train.py:155] train/xent Physical: float32[32,2048]............................................ ('fsdp', None). I0420 06:23:21.286656 139490510141248 train.py:162] train/z_loss Logical: float32[32,2048]............................................ ('activation_embed_and_logits_batch', 'activation_length'). I0420 06:23:21.286715 139490510141248 train.py:162] train/z_loss Physical: float32[32,2048]............................................ ('fsdp', None). I0420 06:23:32.474270 139490510141248 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 06:23:32.475141 139490510141248 metric_logger.py:301] number parameters: 1.104 billion I0420 06:23:43.652383 139490510141248 checkpointing.py:772] Waiting for step 0 to finish before checkpoint... I0420 06:23:43.806169 139490510141248 checkpointing.py:776] Waited 0.15376734733581543 seconds for step 0 to finish before starting checkpointing. I0420 06:23:43.808836 139490510141248 checkpoint_manager.py:2009] [process=4][thread=MainThread][wait_until_finished] No Save Finalize thread to wait for. Returning. I0420 06:23:43.810487 139490510141248 checkpoint_manager.py:1512] [process=4] Saving checkpoint at step 0 I0420 06:23:43.811894 139490510141248 event_tracking.py:70] [process=4] [async] Started save checkpoint @ gs://lance-maxtext/linen_ckpt_xpk_main_20260420_060028/linen_xpk_main_20260420_060028_08_checkpoint_async_true/checkpoints/0. I0420 06:23:44.124069 139490510141248 signaling_client.py:364] Using JaxDistributedSignalingClient I0420 06:23:44.125064 139490510141248 jax_array_handlers.py:360] Scheduling D2H of 39 prioritized jax.Array. I0420 06:23:44.125124 139490510141248 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 06:23:44.402734 139490510141248 base_pytree_checkpoint_handler.py:154] [process=4][thread=MainThread] Initiated "orbax.checkpoint._src.serialization.jax_array_handlers.ArrayHandler".serialize. Time taken: 0.278737s I0420 06:23:44.402899 139490510141248 base_pytree_checkpoint_handler.py:130] [process=4] /jax/orbax/write/blocking_gbytes_per_sec: 5.428 GiB/s (total gbytes: 1.5 GiB) (time elapsed: 0.28418660163879395 s) (per-host) I0420 06:23:44.402955 139490510141248 base_pytree_checkpoint_handler.py:768] [process=4][thread=MainThread] Initiated Pytree async_save. Time taken: 0.284252s (batch_requests_ready=0.002224s, total_serialization_initiated=0.281954s, others=0.000075s) I0420 06:23:44.403046 139490510141248 composite_checkpoint_handler.py:715] [process=4][thread=MainThread] Initiated CompositeCheckpointHandler.async_save. Time taken: 0.288371s (all_items=0.000018s, per_item={'items': '0.00001764'}, temp_paths=0.288353) I0420 06:23:44.403835 139490510141248 event_tracking.py:125] [process=4] [async] Finished blocking save in 0.59 seconds. Continuing save @ gs://lance-maxtext/linen_ckpt_xpk_main_20260420_060028/linen_xpk_main_20260420_060028_08_checkpoint_async_true/checkpoints/0. I0420 06:23:44.404166 139362775422720 async_checkpointer.py:76] [process=4][thread=async_save] Background save thread started. Deadline for this save operation is 2026-04-20 06:43:44.404128 I0420 06:23:44.416144 139490510141248 checkpoint_manager.py:1560] [process=4][thread=MainThread][step=0] Starting CheckpointManager Save Finalize thread=save_finalize I0420 06:23:44.416473 139359579793152 async_checkpointer.py:280] [process=4][thread=save_finalize] Waiting for background save thread=async_save. I0420 06:23:44.416635 139490510141248 standard_logger.py:34] {'step': 0, 'event_type': 'save', 'directory': 'gs://lance-maxtext/linen_ckpt_xpk_main_20260420_060028/linen_xpk_main_20260420_060028_08_checkpoint_async_true/checkpoints', 'reached_preemption': False, 'preemption_received_at': None, 'synchronous': False, 'wait_for_prev_start_time': 1776666223.8088179, 'wait_for_prev_duration_secs': 6.103515625e-05, 'time_between_consecutive_saves_sec': None, 'checkpointer_blocking_start_time': 1776666223.810525, 'checkpointer_blocking_duration_secs': 0.593794584274292, 'get_old_steps_start_time': 1776666224.4043584, 'get_old_steps_duration_secs': 3.0994415283203125e-05, 'checkpoint_manager_blocking_start_time': 1776666223.8067305, 'checkpoint_manager_blocking_duration_secs': 0.6098651885986328} I0420 06:23:44.416741 139490510141248 checkpointing.py:408] Started an asynchronous checkpoint save for step 0 I0420 06:23:44.416792 139490510141248 max_utils.py:750] Memstats: After params initialized: I0420 06:23:44.416840 139490510141248 max_utils.py:756] Using (GB) 0.43 / 31.25 (1.376000%) on TPU_16(process=4,(0,4,0,0)) I0420 06:23:44.416875 139490510141248 max_utils.py:756] Using (GB) 0.43 / 31.25 (1.376000%) on TPU_17(process=4,(1,4,0,0)) I0420 06:23:44.416902 139490510141248 max_utils.py:756] Using (GB) 0.43 / 31.25 (1.376000%) on TPU_20(process=4,(0,5,0,0)) I0420 06:23:44.416925 139490510141248 max_utils.py:756] Using (GB) 0.43 / 31.25 (1.376000%) on TPU_21(process=4,(1,5,0,0)) I0420 06:23:44.728371 139490510141248 metric_logger.py:196] completed step: 0, seconds: 11.177, TFLOP/s/device: 1.216, Tokens/s/device: 183.232, total_weights: 65536, loss: 10.874, lm_loss: 10.874, perplexity: 52779.875 I0420 06:23:44.902188 139490510141248 metric_logger.py:196] completed step: 1, seconds: 1.074, TFLOP/s/device: 12.646, Tokens/s/device: 1906.109, total_weights: 65536, loss: 10.874, lm_loss: 10.874, perplexity: 52779.875 I0420 06:23:45.312674 139490510141248 metric_logger.py:196] completed step: 2, seconds: 0.026, TFLOP/s/device: 514.372, Tokens/s/device: 77531.705, total_weights: 65536, loss: 10.560, lm_loss: 10.560, perplexity: 38547.719 I0420 06:23:45.459937 139490510141248 metric_logger.py:196] completed step: 3, seconds: 0.411, TFLOP/s/device: 33.092, Tokens/s/device: 4988.054, total_weights: 65536, loss: 9.980, lm_loss: 9.980, perplexity: 21590.506 I0420 06:23:45.464440 139490510141248 checkpointing.py:772] Waiting for step 5 to finish before checkpoint... I0420 06:23:45.754045 139490510141248 checkpointing.py:776] Waited 0.2895686626434326 seconds for step 5 to finish before starting checkpointing. I0420 06:23:45.757053 139490510141248 checkpoint_manager.py:2020] [process=4][thread=MainThread][step=0][wait_until_finished] Waiting for Save Finalize thread (save_finalize) to complete. I0420 06:23:47.452594 2818 google_auth_provider.cc:181] Running on GCE, using service account 562977990677-compute@developer.gserviceaccount.com I0420 06:23:49.471199 139360636393216 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_08_checkpoint_async_true/checkpoints/0/items/array_metadatas/process_4 I0420 06:24:21.259870 139362775422720 base_pytree_checkpoint_handler.py:130] [process=4] /jax/orbax/write/gbytes_per_sec: 42.530 MiB/s (total gbytes: 1.5 GiB) (time elapsed: 37.141117811203 s) (per-host) I0420 06:24:21.259994 139362775422720 async_checkpointer.py:90] [process=4][thread=async_save] 3 Handler Commit operations completed. Time taken: 36.855718s. I0420 06:24:27.853293 139362775422720 async_checkpointer.py:160] [process=4][thread=async_save] Background save thread done. Time taken: 43.449000s. I0420 06:24:27.853524 139359579793152 async_checkpointer.py:288] [process=4][thread=save_finalize] Done with waiting for background save thread=async_save. I0420 06:24:27.853598 139359579793152 async_checkpointer.py:298] [process=4][thread=save_finalize] No errors found in background save thread=async_save. I0420 06:24:27.853658 139359579793152 checkpoint_manager.py:2137] [process=4][thread=save_finalize][step=0] CheckpointManager Save Finalize is syncing with other hosts... I0420 06:24:27.855292 139359579793152 checkpoint_manager.py:2146] [process=4][thread=save_finalize][step=0] CheckpointManager Save Finalize is done on all hosts. I0420 06:24:27.855489 139490510141248 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 06:24:27.855645 139490510141248 checkpoint_manager.py:1452] Waiting for previous save to complete took 42.098590 seconds. If this number is high, consider checkpointing less frequently. I0420 06:24:27.857653 139490510141248 checkpoint_manager.py:1512] [process=4] Saving checkpoint at step 5 I0420 06:24:27.860151 139490510141248 event_tracking.py:70] [process=4] [async] Started save checkpoint @ gs://lance-maxtext/linen_ckpt_xpk_main_20260420_060028/linen_xpk_main_20260420_060028_08_checkpoint_async_true/checkpoints/5. I0420 06:24:28.344617 139490510141248 jax_array_handlers.py:360] Scheduling D2H of 39 prioritized jax.Array. I0420 06:24:28.344712 139490510141248 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 06:24:28.377941 139490510141248 base_pytree_checkpoint_handler.py:154] [process=4][thread=MainThread] Initiated "orbax.checkpoint._src.serialization.jax_array_handlers.ArrayHandler".serialize. Time taken: 0.034405s I0420 06:24:28.378074 139490510141248 base_pytree_checkpoint_handler.py:130] [process=4] /jax/orbax/write/blocking_gbytes_per_sec: 40.818 GiB/s (total gbytes: 1.5 GiB) (time elapsed: 0.03779196739196777 s) (per-host) I0420 06:24:28.378123 139490510141248 base_pytree_checkpoint_handler.py:768] [process=4][thread=MainThread] Initiated Pytree async_save. Time taken: 0.037849s (batch_requests_ready=0.001779s, total_serialization_initiated=0.036008s, others=0.000062s) I0420 06:24:28.378200 139490510141248 composite_checkpoint_handler.py:715] [process=4][thread=MainThread] Initiated CompositeCheckpointHandler.async_save. Time taken: 0.042792s (all_items=0.000015s, per_item={'items': '0.00001550'}, temp_paths=0.042776) I0420 06:24:28.378859 139490510141248 event_tracking.py:125] [process=4] [async] Finished blocking save in 0.52 seconds. Continuing save @ gs://lance-maxtext/linen_ckpt_xpk_main_20260420_060028/linen_xpk_main_20260420_060028_08_checkpoint_async_true/checkpoints/5. I0420 06:24:28.379130 139359579793152 async_checkpointer.py:76] [process=4][thread=async_save] Background save thread started. Deadline for this save operation is 2026-04-20 06:44:28.379106 I0420 06:24:28.390983 139490510141248 checkpoint_manager.py:1560] [process=4][thread=MainThread][step=5] Starting CheckpointManager Save Finalize thread=save_finalize I0420 06:24:28.391246 139339937867520 async_checkpointer.py:280] [process=4][thread=save_finalize] Waiting for background save thread=async_save. I0420 06:24:28.391419 139490510141248 standard_logger.py:34] {'step': 5, 'event_type': 'save', 'directory': 'gs://lance-maxtext/linen_ckpt_xpk_main_20260420_060028/linen_xpk_main_20260420_060028_08_checkpoint_async_true/checkpoints', 'reached_preemption': False, 'preemption_received_at': None, 'synchronous': False, 'wait_for_prev_start_time': 1776666225.7570252, 'wait_for_prev_duration_secs': 42.09858965873718, 'time_between_consecutive_saves_sec': None, 'checkpointer_blocking_start_time': 1776666267.8576922, 'checkpointer_blocking_duration_secs': 0.521538257598877, 'get_old_steps_start_time': 1776666268.3792534, 'get_old_steps_duration_secs': 3.0517578125e-05, 'checkpoint_manager_blocking_start_time': 1776666225.7547266, 'checkpoint_manager_blocking_duration_secs': 42.63665843009949} I0420 06:24:28.391533 139490510141248 checkpointing.py:408] Started an asynchronous checkpoint save for step 5 I0420 06:24:28.392375 139490510141248 metric_logger.py:196] completed step: 4, seconds: 0.153, TFLOP/s/device: 88.582, Tokens/s/device: 13352.110, total_weights: 65536, loss: 9.460, lm_loss: 9.460, perplexity: 12840.355 I0420 06:24:28.398319 139490510141248 metric_logger.py:196] completed step: 5, seconds: 0.147, TFLOP/s/device: 92.347, Tokens/s/device: 13919.474, total_weights: 65536, loss: 8.959, lm_loss: 8.959, perplexity: 7776.467 I0420 06:24:33.540182 139360636393216 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_08_checkpoint_async_true/checkpoints/5/items/array_metadatas/process_4 I0420 06:24:51.563786 139490510141248 metric_logger.py:196] completed step: 6, seconds: 42.932, TFLOP/s/device: 0.316, Tokens/s/device: 47.703, total_weights: 65536, loss: 8.469, lm_loss: 8.469, perplexity: 4762.740 I0420 06:24:51.703749 139490510141248 metric_logger.py:196] completed step: 7, seconds: 23.012, TFLOP/s/device: 0.590, Tokens/s/device: 88.998, total_weights: 65536, loss: 8.003, lm_loss: 8.003, perplexity: 2989.151 I0420 06:24:51.851097 139490510141248 metric_logger.py:196] completed step: 8, seconds: 0.167, TFLOP/s/device: 81.384, Tokens/s/device: 12267.072, total_weights: 65536, loss: 7.572, lm_loss: 7.572, perplexity: 1943.098 I0420 06:24:51.997742 139490510141248 checkpointing.py:772] Waiting for step 9 to finish before checkpoint... I0420 06:24:51.998397 139490510141248 checkpointing.py:776] Waited 0.0006740093231201172 seconds for step 9 to finish before starting checkpointing. I0420 06:24:52.000579 139490510141248 checkpoint_manager.py:2020] [process=4][thread=MainThread][step=5][wait_until_finished] Waiting for Save Finalize thread (save_finalize) to complete. I0420 06:25:09.882387 139359579793152 base_pytree_checkpoint_handler.py:130] [process=4] /jax/orbax/write/gbytes_per_sec: 38.024 MiB/s (total gbytes: 1.5 GiB) (time elapsed: 41.54208040237427 s) (per-host) I0420 06:25:09.882498 139359579793152 async_checkpointer.py:90] [process=4][thread=async_save] 3 Handler Commit operations completed. Time taken: 41.503291s. I0420 06:25:15.941138 139359579793152 async_checkpointer.py:160] [process=4][thread=async_save] Background save thread done. Time taken: 47.561914s. I0420 06:25:15.941452 139339937867520 async_checkpointer.py:288] [process=4][thread=save_finalize] Done with waiting for background save thread=async_save. I0420 06:25:15.941597 139339937867520 async_checkpointer.py:298] [process=4][thread=save_finalize] No errors found in background save thread=async_save. I0420 06:25:15.941647 139339937867520 checkpoint_manager.py:2137] [process=4][thread=save_finalize][step=5] CheckpointManager Save Finalize is syncing with other hosts... I0420 06:25:15.942907 139339937867520 checkpoint_manager.py:2146] [process=4][thread=save_finalize][step=5] CheckpointManager Save Finalize is done on all hosts. I0420 06:25:15.943086 139490510141248 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 06:25:15.943209 139490510141248 checkpoint_manager.py:1452] Waiting for previous save to complete took 23.942636 seconds. If this number is high, consider checkpointing less frequently. I0420 06:25:15.944791 139490510141248 checkpoint_manager.py:1512] [process=4] Saving checkpoint at step 9 I0420 06:25:15.946874 139490510141248 event_tracking.py:70] [process=4] [async] Started save checkpoint @ gs://lance-maxtext/linen_ckpt_xpk_main_20260420_060028/linen_xpk_main_20260420_060028_08_checkpoint_async_true/checkpoints/9. I0420 06:25:16.274900 139490510141248 jax_array_handlers.py:360] Scheduling D2H of 39 prioritized jax.Array. I0420 06:25:16.274990 139490510141248 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 06:25:16.307147 139490510141248 base_pytree_checkpoint_handler.py:154] [process=4][thread=MainThread] Initiated "orbax.checkpoint._src.serialization.jax_array_handlers.ArrayHandler".serialize. Time taken: 0.033115s I0420 06:25:16.307312 139490510141248 base_pytree_checkpoint_handler.py:130] [process=4] /jax/orbax/write/blocking_gbytes_per_sec: 42.270 GiB/s (total gbytes: 1.5 GiB) (time elapsed: 0.036493778228759766 s) (per-host) I0420 06:25:16.307381 139490510141248 base_pytree_checkpoint_handler.py:768] [process=4][thread=MainThread] Initiated Pytree async_save. Time taken: 0.036572s (batch_requests_ready=0.001619s, total_serialization_initiated=0.034870s, others=0.000083s) I0420 06:25:16.307475 139490510141248 composite_checkpoint_handler.py:715] [process=4][thread=MainThread] Initiated CompositeCheckpointHandler.async_save. Time taken: 0.041006s (all_items=0.000010s, per_item={'items': '0.00001025'}, temp_paths=0.040996) I0420 06:25:16.308196 139490510141248 event_tracking.py:125] [process=4] [async] Finished blocking save in 0.36 seconds. Continuing save @ gs://lance-maxtext/linen_ckpt_xpk_main_20260420_060028/linen_xpk_main_20260420_060028_08_checkpoint_async_true/checkpoints/9. I0420 06:25:16.308513 139339937867520 async_checkpointer.py:76] [process=4][thread=async_save] Background save thread started. Deadline for this save operation is 2026-04-20 06:45:16.308476 I0420 06:25:16.311855 139490510141248 checkpoint_manager.py:1560] [process=4][thread=MainThread][step=9] Starting CheckpointManager Save Finalize thread=save_finalize I0420 06:25:16.312140 139360636393216 async_checkpointer.py:280] [process=4][thread=save_finalize] Waiting for background save thread=async_save. I0420 06:25:16.312301 139490510141248 standard_logger.py:34] {'step': 9, 'event_type': 'save', 'directory': 'gs://lance-maxtext/linen_ckpt_xpk_main_20260420_060028/linen_xpk_main_20260420_060028_08_checkpoint_async_true/checkpoints', 'reached_preemption': False, 'preemption_received_at': None, 'synchronous': False, 'wait_for_prev_start_time': 1776666292.0005488, 'wait_for_prev_duration_secs': 23.942635536193848, 'time_between_consecutive_saves_sec': 24.14519453048706, 'checkpointer_blocking_start_time': 1776666315.9448287, 'checkpointer_blocking_duration_secs': 0.3638331890106201, 'get_old_steps_start_time': 1776666316.3086812, 'get_old_steps_duration_secs': 2.5987625122070312e-05, 'checkpoint_manager_blocking_start_time': 1776666291.9986403, 'checkpoint_manager_blocking_duration_secs': 24.313626289367676} I0420 06:25:16.312425 139490510141248 checkpointing.py:408] Started an asynchronous checkpoint save for step 9 I0420 06:25:16.312469 139490510141248 checkpoint_manager.py:2020] [process=4][thread=MainThread][step=9][wait_until_finished] Waiting for Save Finalize thread (save_finalize) to complete. I0420 06:25:21.938204 139358505633536 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_08_checkpoint_async_true/checkpoints/9/items/array_metadatas/process_4 I0420 06:25:58.995863 139339937867520 base_pytree_checkpoint_handler.py:130] [process=4] /jax/orbax/write/gbytes_per_sec: 36.971 MiB/s (total gbytes: 1.5 GiB) (time elapsed: 42.72500538825989 s) (per-host) I0420 06:25:58.995987 139339937867520 async_checkpointer.py:90] [process=4][thread=async_save] 3 Handler Commit operations completed. Time taken: 42.687361s. I0420 06:26:05.326957 139339937867520 async_checkpointer.py:160] [process=4][thread=async_save] Background save thread done. Time taken: 49.018314s. I0420 06:26:05.327231 139360636393216 async_checkpointer.py:288] [process=4][thread=save_finalize] Done with waiting for background save thread=async_save. I0420 06:26:05.327362 139360636393216 async_checkpointer.py:298] [process=4][thread=save_finalize] No errors found in background save thread=async_save. I0420 06:26:05.327415 139360636393216 checkpoint_manager.py:2137] [process=4][thread=save_finalize][step=9] CheckpointManager Save Finalize is syncing with other hosts... I0420 06:26:05.328874 139360636393216 checkpoint_manager.py:2146] [process=4][thread=save_finalize][step=9] CheckpointManager Save Finalize is done on all hosts. I0420 06:26:05.329060 139490510141248 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 06:26:05.329223 139490510141248 checkpoint_manager.py:2009] [process=4][thread=MainThread][wait_until_finished] No Save Finalize thread to wait for. Returning. I0420 06:26:05.330109 139490510141248 metric_logger.py:196] completed step: 9, seconds: 0.133, TFLOP/s/device: 102.484, Tokens/s/device: 15447.510, 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 06:26:17 UTC 2026 EXIT_CODE=0
XPK Start: Tue Apr 21 01:13:58 UTC 2026 2026-04-21 01:14:02.599630: 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 00:00:1776734042.612470 10 cuda_dnn.cc:8579] Unable to register cuDNN factory: Attempting to register factory for plugin cuDNN when one has already been registered E0000 00:00:1776734042.616300 10 cuda_blas.cc:1407] Unable to register cuBLAS factory: Attempting to register factory for plugin cuBLAS when one has already been registered W0000 00:00:1776734042.627493 10 computation_placer.cc:177] computation placer already registered. Please check linkage and avoid linking the same target more than once. W0000 00:00:1776734042.627510 10 computation_placer.cc:177] computation placer already registered. Please check linkage and avoid linking the same target more than once. W0000 00:00:1776734042.627513 10 computation_placer.cc:177] computation placer already registered. Please check linkage and avoid linking the same target more than once. W0000 00:00:1776734042.627515 10 computation_placer.cc:177] computation placer already registered. Please check linkage and avoid linking the same target more than once. 2026-04-21 01:14:21.799246: 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 01:14:22.318120 134279449679680 max_utils.py:273] Attempting to initialize the jax distributed system... INFO:2026-04-21 01:14:31,360:jax._src.distributed:140: Starting JAX distributed service on [::]:8482 I0421 01:14:31.360152 134279449679680 distributed.py:140] Starting JAX distributed service on [::]:8482 INFO:2026-04-21 01:14:31,362:jax._src.distributed:157: Connecting to JAX distributed service on mt-08-checkpoint-async--bcr5r-slice-job-0-0.mt-08-checkpoint-async--bcr5r:8482 I0421 01:14:31.362556 134279449679680 distributed.py:157] Connecting to JAX distributed service on mt-08-checkpoint-async--bcr5r-slice-job-0-0.mt-08-checkpoint-async--bcr5r:8482 I0421 01:14:32.543552 134279449679680 max_utils.py:284] Jax distributed system initialized! I0421 01:14:39.466960 134279449679680 max_utils.py:800] System Information: Jax Version: 0.8.1 I0421 01:14:39.467069 134279449679680 max_utils.py:801] System Information: Jaxlib Version: 0.8.1 I0421 01:14:39.467109 134279449679680 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 01:14:39.467145 134279449679680 train_utils.py:347] WARNING: Sequence packing is essentially ignored for synthetic data. Please use a real dataset to use sequence packing. I0421 01:14:40.087154 134279449679680 maxtext_utils.py:1398] Num_devices: 32, shape (1, 1, 1, 32, 1, 1, 1, 1, 1, 1, 1, 1, 1) I0421 01:14:40.087454 134279449679680 checkpointing.py:677] Setting up checkpoint logger... I0421 01:14:40.087510 134279449679680 checkpointing.py:233] Creating checkpoint manager with ocdbt=True and zarr3=True I0421 01:14:40.087554 134279449679680 pytree_checkpoint_handler.py:589] save_device_host_concurrent_bytes=None I0421 01:14:40.087933 134279449679680 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 0x7a1fe5179010>, enable_pinned_host_transfer=False, save_concurrent_bytes: 96000000000 (89.4 GiB), restore_concurrent_bytes: 96000000000 (89.4 GiB) I0421 01:14:43.094487 134279449679680 checkpointing.py:265] Enabling policy for fixed interval checkpointing. I0421 01:14:43.094735 134279449679680 checkpoint_manager.py:709] [process=4][thread=MainThread] CheckpointManager init: checkpointers=None, item_names=('items',), item_handlers={'items': <orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeCheckpointHandler object at 0x7a0c3feb8710>}, handler_registry=None I0421 01:14:43.094969 134279449679680 composite_checkpoint_handler.py:237] Deferred registration for item: "items". Adding handler `<orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeCheckpointHandler object at 0x7a0c3feb8710>` 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 01:14:43.095018 134279449679680 composite_checkpoint_handler.py:237] Deferred registration for item: "metrics". Adding handler `<orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonCheckpointHandler object at 0x7a0c3febbbf0>` 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 01:14:43.095054 134279449679680 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 0x7a0c3feb8710>, ('items', <class 'orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeRestoreArgs'>): <orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeCheckpointHandler object at 0x7a0c3feb8710>, ('metrics', <class 'orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonSaveArgs'>): <orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonCheckpointHandler object at 0x7a0c3febbbf0>, ('metrics', <class 'orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonRestoreArgs'>): <orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonCheckpointHandler object at 0x7a0c3febbbf0>}). I0421 01:14:43.097670 134279449679680 abstract_checkpointer.py:35] orbax-checkpoint version: 0.11.33 I0421 01:14:43.097769 134279449679680 async_checkpointer.py:177] [process=4][thread=MainThread] Using barrier_sync_fn: <function get_barrier_sync_fn.<locals>._fn at 0x7a1fcfde32e0> timeout: 600 secs and primary_host=0 for async checkpoint writes I0421 01:14:45.034687 134279449679680 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_08_checkpoint_async_true/checkpoints I0421 01:14:45.890595 134279449679680 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_test_pipeline_scan_nnx_20260421_005327/linen_xpk_test_pipeline_scan_nnx_20260421_005327_08_checkpoint_async_true/checkpoints: <orbax.checkpoint.checkpoint_manager.CheckpointManager object at 0x7a0c3feb9040> I0421 01:14:45.890804 134279449679680 checkpointing.py:301] Checkpoint manager created! I0421 01:14:46.792055 134279449679680 nnx_wrappers.py:453] Unknown Logical: bfloat16[32,2048,2048]...................................... ('activation_batch', 'activation_norm_length', 'activation_embed'). I0421 01:14:46.792161 134279449679680 nnx_wrappers.py:453] Unknown Physical: bfloat16[32,2048,2048]...................................... ('fsdp', None, None). I0421 01:14:47.165028 134279449679680 attentions.py:1084] attentions/inputs_q Logical: bfloat16[32,2048,2048]...................................... ('activation_batch', 'activation_attn_length', 'activation_attn_embed'). I0421 01:14:47.165120 134279449679680 attentions.py:1084] attentions/inputs_q Physical: bfloat16[32,2048,2048]...................................... ('fsdp', None, None). I0421 01:14:47.180723 134279449679680 attentions.py:1085] attentions/inputs_kv Logical: bfloat16[32,2048,2048]...................................... ('activation_batch', 'activation_attn_length', 'activation_attn_embed'). I0421 01:14:47.180778 134279449679680 attentions.py:1085] attentions/inputs_kv Physical: bfloat16[32,2048,2048]...................................... ('fsdp', None, None). I0421 01:14:47.203954 134279449679680 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 01:14:47.204024 134279449679680 attentions.py:1150] attentions/query Physical: bfloat16[32,2048,16,128].................................... ('fsdp', None, None, None). I0421 01:14:47.219742 134279449679680 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 01:14:47.219804 134279449679680 attentions.py:1151] attentions/key Physical: bfloat16[32,2048,16,128].................................... ('fsdp', None, None, None). I0421 01:14:47.235291 134279449679680 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 01:14:47.235352 134279449679680 attentions.py:1152] attentions/value Physical: bfloat16[32,2048,16,128].................................... ('fsdp', None, None, None). I0421 01:14:47.258945 134279449679680 attentions.py:1193] attentions/out Logical: bfloat16[32,2048,16,128].................................... ('activation_batch', 'activation_attn_length', 'activation_heads', 'activation_kv'). I0421 01:14:47.259015 134279449679680 attentions.py:1193] attentions/out Physical: bfloat16[32,2048,16,128].................................... ('fsdp', None, None, None). I0421 01:14:47.278267 134279449679680 linears.py:525] linears/x Logical: bfloat16[32,2048,7168]...................................... ('activation_batch', 'activation_length', 'activation_mlp'). I0421 01:14:47.278331 134279449679680 linears.py:525] linears/x Physical: bfloat16[32,2048,7168]...................................... ('fsdp', None, None). I0421 01:14:47.466265 134279449679680 checkpointing.py:577] checkpoint manager exists so trying to load this run's existing checkpoint I0421 01:14:47.466374 134279449679680 checkpointing.py:665] No existing checkpoints found, not restoring checkpoint. fsdp: 32 I0421 01:14:48.876045 134279449679680 maxtext_utils.py:1501] params/params/decoder/decoder_norm/scale Shape: float32[2048] Logical: PartitionSpec('norm',) Physical: (None,) I0421 01:14:48.876179 134279449679680 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 01:14:48.876237 134279449679680 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 01:14:48.876296 134279449679680 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 01:14:48.876349 134279449679680 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 01:14:48.876389 134279449679680 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 01:14:48.876449 134279449679680 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 01:14:48.876505 134279449679680 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 01:14:48.876548 134279449679680 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 01:14:48.876587 134279449679680 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 01:14:48.876637 134279449679680 maxtext_utils.py:1501] params/params/decoder/logits_dense/kernel Shape: float32[2048,32000] Logical: PartitionSpec('embed', 'vocab') Physical: ('fsdp', None) I0421 01:14:48.876685 134279449679680 maxtext_utils.py:1501] params/params/token_embedder/embedding Shape: float32[32000,2048] Logical: PartitionSpec('vocab', 'embed') Physical: (None, 'fsdp') I0421 01:14:49.382770 134279449679680 train.py:156] train/xent Logical: float32[32,2048]............................................ ('activation_embed_and_logits_batch', 'activation_length'). I0421 01:14:49.382870 134279449679680 train.py:156] train/xent Physical: float32[32,2048]............................................ ('fsdp', None). I0421 01:14:49.397837 134279449679680 train.py:163] train/z_loss Logical: float32[32,2048]............................................ ('activation_embed_and_logits_batch', 'activation_length'). I0421 01:14:49.397896 134279449679680 train.py:163] train/z_loss Physical: float32[32,2048]............................................ ('fsdp', None). I0421 01:15:00.124896 134279449679680 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 01:15:00.125789 134279449679680 metric_logger.py:289] number parameters: 1.104 billion I0421 01:15:11.098251 134279449679680 checkpointing.py:772] Waiting for step 0 to finish before checkpoint... I0421 01:15:11.254278 134279449679680 checkpointing.py:776] Waited 0.15600872039794922 seconds for step 0 to finish before starting checkpointing. I0421 01:15:11.256777 134279449679680 checkpoint_manager.py:2013] [process=4][thread=MainThread][wait_until_finished] No Save Finalize thread to wait for. Returning. I0421 01:15:11.258347 134279449679680 checkpoint_manager.py:1518] [process=4] Saving checkpoint at step 0 I0421 01:15:11.261117 134279449679680 async_checkpointer.py:452] [process=4] 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_08_checkpoint_async_true/checkpoints/0. I0421 01:15:13.106612 134279449679680 signaling_client.py:364] Using JaxDistributedSignalingClient I0421 01:15:13.107608 134279449679680 jax_array_handlers.py:358] Scheduling D2H of 39 prioritized jax.Array. I0421 01:15:13.107668 134279449679680 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 01:15:13.359639 134279449679680 base_pytree_checkpoint_handler.py:153] [process=4][thread=MainThread] Initiated "orbax.checkpoint._src.serialization.jax_array_handlers.ArrayHandler".serialize. Time taken: 0.253110s I0421 01:15:13.359826 134279449679680 base_pytree_checkpoint_handler.py:129] [process=4] /jax/checkpoint/write/blocking_gbytes_per_sec: 1.647 GiB/s (total gbytes: 1.5 GiB) (time elapsed: 0.9368727207183838 s) (per-host) I0421 01:15:13.359877 134279449679680 base_pytree_checkpoint_handler.py:737] [process=4][thread=MainThread] Initiated Pytree async_save. Time taken: 0.936934s (batch_requests_ready=0.680177s, total_serialization_initiated=0.256688s, others=0.000069s) I0421 01:15:13.359966 134279449679680 composite_checkpoint_handler.py:715] [process=4][thread=MainThread] Initiated CompositeCheckpointHandler.async_save. Time taken: 0.940760s (all_items=0.000025s, per_item={'items': '0.00002527'}, temp_paths=0.940735) I0421 01:15:13.360882 134156977690368 async_checkpointer.py:79] [process=4][thread=async_save] Background save thread started. I0421 01:15:13.361060 134279449679680 async_checkpointer.py:561] Finished blocking save. Time taken: 2.102640s. Continuing background save to gs://lance-maxtext/linen_ckpt_xpk_test_pipeline_scan_nnx_20260421_005327/linen_xpk_test_pipeline_scan_nnx_20260421_005327_08_checkpoint_async_true/checkpoints/0. I0421 01:15:13.390614 134279449679680 checkpoint_manager.py:1566] [process=4][thread=MainThread][step=0] Starting CheckpointManager Save Finalize thread=save_finalize I0421 01:15:13.390943 134152657544960 async_checkpointer.py:265] [process=4][thread=save_finalize] Waiting for background save thread=async_save. I0421 01:15:13.391105 134279449679680 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_08_checkpoint_async_true/checkpoints', 'reached_preemption': False, 'preemption_received_at': None, 'synchronous': False, 'wait_for_prev_start_time': 1776734111.2567594, 'wait_for_prev_duration_secs': 6.246566772460938e-05, 'time_between_consecutive_saves_sec': None, 'checkpointer_blocking_start_time': 1776734111.2583854, 'checkpointer_blocking_duration_secs': 2.1027979850769043, 'get_old_steps_start_time': 1776734113.3612053, 'get_old_steps_duration_secs': 2.5987625122070312e-05, 'checkpoint_manager_blocking_start_time': 1776734111.254771, 'checkpoint_manager_blocking_duration_secs': 2.136293411254883} I0421 01:15:13.391240 134279449679680 checkpointing.py:408] Started an asynchronous checkpoint save for step 0 I0421 01:15:13.391298 134279449679680 max_utils.py:750] Memstats: After params initialized: I0421 01:15:13.391364 134279449679680 max_utils.py:756] Using (GB) 0.43 / 31.25 (1.376000%) on TPU_16(process=4,(0,4,0,0)) I0421 01:15:13.391423 134279449679680 max_utils.py:756] Using (GB) 0.43 / 31.25 (1.376000%) on TPU_17(process=4,(1,4,0,0)) I0421 01:15:13.391470 134279449679680 max_utils.py:756] Using (GB) 0.43 / 31.25 (1.376000%) on TPU_20(process=4,(0,5,0,0)) I0421 01:15:13.391524 134279449679680 max_utils.py:756] Using (GB) 0.43 / 31.25 (1.376000%) on TPU_21(process=4,(1,5,0,0)) I0421 01:15:13.658926 134279449679680 metric_logger.py:185] completed step: 0, seconds: 10.972, TFLOP/s/device: 1.238, Tokens/s/device: 186.651, total_weights: 65536, loss: 10.874 I0421 01:15:13.826464 134279449679680 metric_logger.py:185] completed step: 1, seconds: 2.559, TFLOP/s/device: 5.309, Tokens/s/device: 800.202, total_weights: 65536, loss: 10.874 I0421 01:15:14.243731 134279449679680 metric_logger.py:185] completed step: 2, seconds: 0.021, TFLOP/s/device: 662.012, Tokens/s/device: 99785.617, total_weights: 65536, loss: 10.560 I0421 01:15:14.391079 134279449679680 metric_logger.py:185] completed step: 3, seconds: 0.418, TFLOP/s/device: 32.532, Tokens/s/device: 4903.651, total_weights: 65536, loss: 9.980 I0421 01:15:14.395552 134279449679680 checkpointing.py:772] Waiting for step 5 to finish before checkpoint... I0421 01:15:14.685633 134279449679680 checkpointing.py:776] Waited 0.29004812240600586 seconds for step 5 to finish before starting checkpointing. I0421 01:15:14.688410 134279449679680 checkpoint_manager.py:2024] [process=4][thread=MainThread][step=0][wait_until_finished] Waiting for Save Finalize thread (save_finalize) to complete. I0421 01:15:18.095530 2623 google_auth_provider.cc:181] Running on GCE, using service account 562977990677-compute@developer.gserviceaccount.com I0421 01:15:20.883622 134156986083072 array_metadata_store.py:203] [process=4][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_08_checkpoint_async_true/checkpoints/0/items/array_metadatas/process_4 I0421 01:15:51.719785 134156977690368 base_pytree_checkpoint_handler.py:129] [process=4] /jax/checkpoint/write/gbytes_per_sec: 40.197 MiB/s (total gbytes: 1.5 GiB) (time elapsed: 39.296791076660156 s) (per-host) I0421 01:15:51.719911 134156977690368 async_checkpointer.py:90] [process=4][thread=async_save] 3 Handler Commit operations completed. Time taken: 38.358916s. I0421 01:15:58.765643 134156977690368 async_checkpointer.py:144] [process=4][thread=async_save] Background save thread done. Time taken: 45.404631s. I0421 01:15:58.765970 134152657544960 async_checkpointer.py:273] [process=4][thread=save_finalize] Done with waiting for background save thread=async_save. I0421 01:15:58.766089 134152657544960 async_checkpointer.py:283] [process=4][thread=save_finalize] No errors found in background save thread=async_save. I0421 01:15:58.766140 134152657544960 checkpoint_manager.py:2133] [process=4][thread=save_finalize][step=0] CheckpointManager Save Finalize is syncing with other hosts... I0421 01:15:58.767590 134152657544960 checkpoint_manager.py:2142] [process=4][thread=save_finalize][step=0] CheckpointManager Save Finalize is done on all hosts. I0421 01:15:58.767776 134279449679680 checkpoint_manager.py:2036] [process=4][thread=MainThread][step=0][wait_until_finished] Done waiting for Save Finalize thread (save_finalize) running at step=0. W0421 01:15:58.767935 134279449679680 checkpoint_manager.py:1458] Waiting for previous save to complete took 44.079526 seconds. If this number is high, consider checkpointing less frequently. I0421 01:15:58.769998 134279449679680 checkpoint_manager.py:1518] [process=4] Saving checkpoint at step 5 I0421 01:15:58.774100 134279449679680 async_checkpointer.py:452] [process=4] 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_08_checkpoint_async_true/checkpoints/5. I0421 01:15:59.740348 134279449679680 jax_array_handlers.py:358] Scheduling D2H of 39 prioritized jax.Array. I0421 01:15:59.740449 134279449679680 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 01:15:59.781711 134279449679680 base_pytree_checkpoint_handler.py:153] [process=4][thread=MainThread] Initiated "orbax.checkpoint._src.serialization.jax_array_handlers.ArrayHandler".serialize. Time taken: 0.042486s I0421 01:15:59.781875 134279449679680 base_pytree_checkpoint_handler.py:129] [process=4] /jax/checkpoint/write/blocking_gbytes_per_sec: 5.232 GiB/s (total gbytes: 1.5 GiB) (time elapsed: 0.29482269287109375 s) (per-host) I0421 01:15:59.781929 134279449679680 base_pytree_checkpoint_handler.py:737] [process=4][thread=MainThread] Initiated Pytree async_save. Time taken: 0.294886s (batch_requests_ready=0.250513s, total_serialization_initiated=0.044301s, others=0.000071s) I0421 01:15:59.782019 134279449679680 composite_checkpoint_handler.py:715] [process=4][thread=MainThread] Initiated CompositeCheckpointHandler.async_save. Time taken: 0.299509s (all_items=0.000016s, per_item={'items': '0.00001645'}, temp_paths=0.299492) I0421 01:15:59.782919 134156986083072 async_checkpointer.py:79] [process=4][thread=async_save] Background save thread started. I0421 01:15:59.783079 134279449679680 async_checkpointer.py:561] Finished blocking save. Time taken: 1.013009s. Continuing background save to gs://lance-maxtext/linen_ckpt_xpk_test_pipeline_scan_nnx_20260421_005327/linen_xpk_test_pipeline_scan_nnx_20260421_005327_08_checkpoint_async_true/checkpoints/5. I0421 01:16:00.222281 134279449679680 checkpoint_manager.py:1566] [process=4][thread=MainThread][step=5] Starting CheckpointManager Save Finalize thread=save_finalize I0421 01:16:00.222653 134152657544960 async_checkpointer.py:265] [process=4][thread=save_finalize] Waiting for background save thread=async_save. I0421 01:16:00.222850 134279449679680 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_08_checkpoint_async_true/checkpoints', 'reached_preemption': False, 'preemption_received_at': None, 'synchronous': False, 'wait_for_prev_start_time': 1776734114.6883805, 'wait_for_prev_duration_secs': 44.07952570915222, 'time_between_consecutive_saves_sec': None, 'checkpointer_blocking_start_time': 1776734158.7700381, 'checkpointer_blocking_duration_secs': 1.0131738185882568, 'get_old_steps_start_time': 1776734159.7832346, 'get_old_steps_duration_secs': 2.86102294921875e-05, 'checkpoint_manager_blocking_start_time': 1776734114.6863256, 'checkpoint_manager_blocking_duration_secs': 45.536486864089966} I0421 01:16:00.222963 134279449679680 checkpointing.py:408] Started an asynchronous checkpoint save for step 5 I0421 01:16:00.223571 134279449679680 metric_logger.py:185] completed step: 4, seconds: 0.152, TFLOP/s/device: 89.192, Tokens/s/device: 13444.054, total_weights: 65536, loss: 9.460 I0421 01:16:00.229546 134279449679680 metric_logger.py:185] completed step: 5, seconds: 0.147, TFLOP/s/device: 92.218, Tokens/s/device: 13900.107, total_weights: 65536, loss: 8.959 I0421 01:16:20.509407 134279449679680 metric_logger.py:185] completed step: 6, seconds: 45.833, TFLOP/s/device: 0.296, Tokens/s/device: 44.684, total_weights: 65536, loss: 8.469 I0421 01:16:20.656905 134279449679680 metric_logger.py:185] completed step: 7, seconds: 20.134, TFLOP/s/device: 0.675, Tokens/s/device: 101.719, total_weights: 65536, loss: 8.003 I0421 01:16:20.804111 134279449679680 metric_logger.py:185] completed step: 8, seconds: 0.152, TFLOP/s/device: 89.351, Tokens/s/device: 13468.013, total_weights: 65536, loss: 7.572 I0421 01:16:20.951137 134279449679680 checkpointing.py:772] Waiting for step 9 to finish before checkpoint... I0421 01:16:20.951865 134279449679680 checkpointing.py:776] Waited 0.0007486343383789062 seconds for step 9 to finish before starting checkpointing. I0421 01:16:20.953824 134279449679680 checkpoint_manager.py:2024] [process=4][thread=MainThread][step=5][wait_until_finished] Waiting for Save Finalize thread (save_finalize) to complete. I0421 01:16:26.652609 134149453838080 array_metadata_store.py:203] [process=4][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_08_checkpoint_async_true/checkpoints/5/items/array_metadatas/process_4 I0421 01:17:00.503639 134156986083072 base_pytree_checkpoint_handler.py:129] [process=4] /jax/checkpoint/write/gbytes_per_sec: 25.888 MiB/s (total gbytes: 1.5 GiB) (time elapsed: 61.01654410362244 s) (per-host) I0421 01:17:00.503775 134156986083072 async_checkpointer.py:90] [process=4][thread=async_save] 3 Handler Commit operations completed. Time taken: 60.720745s. I0421 01:17:09.893029 134156986083072 async_checkpointer.py:144] [process=4][thread=async_save] Background save thread done. Time taken: 70.109982s. I0421 01:17:09.893360 134152657544960 async_checkpointer.py:273] [process=4][thread=save_finalize] Done with waiting for background save thread=async_save. I0421 01:17:09.893481 134152657544960 async_checkpointer.py:283] [process=4][thread=save_finalize] No errors found in background save thread=async_save. I0421 01:17:09.893531 134152657544960 checkpoint_manager.py:2133] [process=4][thread=save_finalize][step=5] CheckpointManager Save Finalize is syncing with other hosts... I0421 01:17:09.894820 134152657544960 checkpoint_manager.py:2142] [process=4][thread=save_finalize][step=5] CheckpointManager Save Finalize is done on all hosts. I0421 01:17:09.895004 134279449679680 checkpoint_manager.py:2036] [process=4][thread=MainThread][step=5][wait_until_finished] Done waiting for Save Finalize thread (save_finalize) running at step=5. W0421 01:17:09.895186 134279449679680 checkpoint_manager.py:1458] Waiting for previous save to complete took 48.941369 seconds. If this number is high, consider checkpointing less frequently. I0421 01:17:09.897426 134279449679680 checkpoint_manager.py:1518] [process=4] Saving checkpoint at step 9 I0421 01:17:09.900810 134279449679680 async_checkpointer.py:452] [process=4] 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_08_checkpoint_async_true/checkpoints/9. I0421 01:17:10.430578 134279449679680 jax_array_handlers.py:358] Scheduling D2H of 39 prioritized jax.Array. I0421 01:17:10.430676 134279449679680 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 01:17:10.477666 134279449679680 base_pytree_checkpoint_handler.py:153] [process=4][thread=MainThread] Initiated "orbax.checkpoint._src.serialization.jax_array_handlers.ArrayHandler".serialize. Time taken: 0.048021s I0421 01:17:10.477843 134279449679680 base_pytree_checkpoint_handler.py:129] [process=4] /jax/checkpoint/write/blocking_gbytes_per_sec: 5.121 GiB/s (total gbytes: 1.5 GiB) (time elapsed: 0.3012106418609619 s) (per-host) I0421 01:17:10.477893 134279449679680 base_pytree_checkpoint_handler.py:737] [process=4][thread=MainThread] Initiated Pytree async_save. Time taken: 0.301270s (batch_requests_ready=0.251190s, total_serialization_initiated=0.050015s, others=0.000065s) I0421 01:17:10.477976 134279449679680 composite_checkpoint_handler.py:715] [process=4][thread=MainThread] Initiated CompositeCheckpointHandler.async_save. Time taken: 0.305304s (all_items=0.000012s, per_item={'items': '0.00001192'}, temp_paths=0.305292) I0421 01:17:10.478848 134152673302272 async_checkpointer.py:79] [process=4][thread=async_save] Background save thread started. I0421 01:17:10.479021 134279449679680 async_checkpointer.py:561] Finished blocking save. Time taken: 0.581524s. Continuing background save to gs://lance-maxtext/linen_ckpt_xpk_test_pipeline_scan_nnx_20260421_005327/linen_xpk_test_pipeline_scan_nnx_20260421_005327_08_checkpoint_async_true/checkpoints/9. I0421 01:17:11.355501 134279449679680 checkpoint_manager.py:1566] [process=4][thread=MainThread][step=9] Starting CheckpointManager Save Finalize thread=save_finalize I0421 01:17:11.355858 134152657544960 async_checkpointer.py:265] [process=4][thread=save_finalize] Waiting for background save thread=async_save. I0421 01:17:11.356024 134279449679680 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_08_checkpoint_async_true/checkpoints', 'reached_preemption': False, 'preemption_received_at': None, 'synchronous': False, 'wait_for_prev_start_time': 1776734180.9537961, 'wait_for_prev_duration_secs': 48.94136929512024, 'time_between_consecutive_saves_sec': 22.186167001724243, 'checkpointer_blocking_start_time': 1776734229.8974671, 'checkpointer_blocking_duration_secs': 0.5816619396209717, 'get_old_steps_start_time': 1776734230.479143, 'get_old_steps_duration_secs': 2.0265579223632812e-05, 'checkpoint_manager_blocking_start_time': 1776734180.952113, 'checkpoint_manager_blocking_duration_secs': 50.403876304626465} I0421 01:17:11.356133 134279449679680 checkpointing.py:408] Started an asynchronous checkpoint save for step 9 I0421 01:17:11.356176 134279449679680 checkpoint_manager.py:2024] [process=4][thread=MainThread][step=9][wait_until_finished] Waiting for Save Finalize thread (save_finalize) to complete. I0421 01:17:17.294776 134156977690368 array_metadata_store.py:203] [process=4][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_08_checkpoint_async_true/checkpoints/9/items/array_metadatas/process_4 I0421 01:17:53.128642 134152673302272 base_pytree_checkpoint_handler.py:129] [process=4] /jax/checkpoint/write/gbytes_per_sec: 36.776 MiB/s (total gbytes: 1.5 GiB) (time elapsed: 42.95196747779846 s) (per-host) I0421 01:17:53.128774 134152673302272 async_checkpointer.py:90] [process=4][thread=async_save] 3 Handler Commit operations completed. Time taken: 42.649810s. I0421 01:18:01.333477 134152673302272 async_checkpointer.py:144] [process=4][thread=async_save] Background save thread done. Time taken: 50.854489s. I0421 01:18:01.333808 134152657544960 async_checkpointer.py:273] [process=4][thread=save_finalize] Done with waiting for background save thread=async_save. I0421 01:18:01.333955 134152657544960 async_checkpointer.py:283] [process=4][thread=save_finalize] No errors found in background save thread=async_save. I0421 01:18:01.334010 134152657544960 checkpoint_manager.py:2133] [process=4][thread=save_finalize][step=9] CheckpointManager Save Finalize is syncing with other hosts... I0421 01:18:01.335350 134152657544960 checkpoint_manager.py:2142] [process=4][thread=save_finalize][step=9] CheckpointManager Save Finalize is done on all hosts. I0421 01:18:01.335519 134279449679680 checkpoint_manager.py:2036] [process=4][thread=MainThread][step=9][wait_until_finished] Done waiting for Save Finalize thread (save_finalize) running at step=9. I0421 01:18:01.335662 134279449679680 checkpoint_manager.py:2013] [process=4][thread=MainThread][wait_until_finished] No Save Finalize thread to wait for. Returning. I0421 01:18:01.336377 134279449679680 metric_logger.py:185] completed step: 9, seconds: 0.147, TFLOP/s/device: 92.283, Tokens/s/device: 13909.925, 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 01:18:13 UTC 2026 EXIT_CODE=0