main (8a17c3d19) vs test/pipeline-scan-nnx (5b21b66fc)| Metric | main 8a17c3d19 | test/pipeline-scan-nnx 5b21b66fc | Diff (test/pipeline-scan-nnx − main) |
|---|---|---|---|
| Parameters | 1.104 billion | 1.104 billion | — |
| Final loss | 7.1810 | 7.1810 | 0 |
| TFLOP/s | 92.248 | 92.355 | +0.107 |
| Tok/s | 13904.6 | 13920.7 | +16.068 |
| Avg s/step | 7.834 | 7.838 | +0.004 |
| Memory % | 1.38 | 1.38 | 0 |
| JAX | 0.9.2 | 0.9.2 | — |
Diff = branch value − main value. Green = branch improved. Red = branch regressed.
XPK Start: Wed Apr 22 07:35:44 UTC 2026 PyTorch was not found. Models won't be available and only tokenizers, configuration and file/data utilities can be used. `rope_parameters`'s factor field must be a float >= 1, got 40 `rope_parameters`'s beta_fast field must be a float, got 32 `rope_parameters`'s beta_slow field must be a float, got 1 DeepseekV32Config got `key=rope_scaling` in kwargs but hasn't set it as attribute. For RoPE standardization you need to set `self.rope_parameters` in model's config. 2026-04-22 07:36:09.623787: E external/local_xla/xla/stream_executor/cuda/cuda_platform.cc:51] failed call to cuInit: INTERNAL: CUDA error: Failed call to cuInit: UNKNOWN ERROR (303) I0422 07:36:09.837476 140560251893568 max_utils.py:273] Attempting to initialize the jax distributed system... I0422 07:36:18.877778 140560251893568 distributed.py:149] Starting JAX distributed service on [::]:8482 I0422 07:36:18.880144 140560251893568 distributed.py:172] Connecting to JAX distributed service on mt-08-checkpoint-async--iuv4c-slice-job-0-0.mt-08-checkpoint-async--iuv4c:8482 I0422 07:36:19.965348 140560251893568 max_utils.py:284] Jax distributed system initialized! I0422 07:36:25.081165 140560251893568 max_utils.py:800] System Information: Jax Version: 0.9.2 I0422 07:36:25.081271 140560251893568 max_utils.py:801] System Information: Jaxlib Version: 0.9.2 I0422 07:36:25.081334 140560251893568 max_utils.py:802] System Information: Jax Backend: PJRT C API TFRT TPU v6 lite Built on Mar 4 2026 11:32:08 (1772652728) cl/878335365 I0422 07:36:25.081373 140560251893568 train_utils.py:361] WARNING: Sequence packing is essentially ignored for synthetic data. Please use a real dataset to use sequence packing. I0422 07:36:25.777177 140560251893568 maxtext_utils.py:1604] Num_devices: 32, shape (1, 1, 1, 32, 1, 1, 1, 1, 1, 1, 1, 1, 1) I0422 07:36:25.777505 140560251893568 checkpointing.py:677] Setting up checkpoint logger... I0422 07:36:25.777575 140560251893568 checkpointing.py:233] Creating checkpoint manager with ocdbt=True and zarr3=True I0422 07:36:25.777629 140560251893568 pytree_checkpoint_handler.py:592] save_device_host_concurrent_bytes=None I0422 07:36:25.777997 140560251893568 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 0x7fd60424c740>, enable_pinned_host_transfer=False, save_concurrent_bytes: 96000000000 (89.4 GiB), restore_concurrent_bytes: 96000000000 (89.4 GiB) I0422 07:36:28.635142 140560251893568 checkpointing.py:265] Enabling policy for fixed interval checkpointing. I0422 07:36:28.635404 140560251893568 checkpoint_manager.py:708] [process=6][thread=MainThread] CheckpointManager init: checkpointers=None, item_names=('items',), item_handlers={'items': <orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeCheckpointHandler object at 0x7fc26c7adf10>}, handler_registry=None I0422 07:36:28.635649 140560251893568 composite_checkpoint_handler.py:237] Deferred registration for item: "items". Adding handler `<orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeCheckpointHandler object at 0x7fc26c7adf10>` for item "items" and save args `<class 'orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeSaveArgs'>` and restore args `<class 'orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeRestoreArgs'>` to `_handler_registry`. I0422 07:36:28.635699 140560251893568 composite_checkpoint_handler.py:237] Deferred registration for item: "metrics". Adding handler `<orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonCheckpointHandler object at 0x7fc26c7b15b0>` for item "metrics" and save args `<class 'orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonSaveArgs'>` and restore args `<class 'orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonRestoreArgs'>` to `_handler_registry`. I0422 07:36:28.635735 140560251893568 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 0x7fc26c7adf10>, ('items', <class 'orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeRestoreArgs'>): <orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeCheckpointHandler object at 0x7fc26c7adf10>, ('metrics', <class 'orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonSaveArgs'>): <orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonCheckpointHandler object at 0x7fc26c7b15b0>, ('metrics', <class 'orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonRestoreArgs'>): <orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonCheckpointHandler object at 0x7fc26c7b15b0>}). I0422 07:36:28.636065 140560251893568 abstract_checkpointer.py:35] orbax-checkpoint version: 0.11.34 I0422 07:36:28.636136 140560251893568 async_checkpointer.py:192] [process=6][thread=MainThread] Using barrier_sync_fn: <function get_barrier_sync_fn.<locals>._fn at 0x7fc26c3d5800> timeout: 1200 secs and primary_host=0 for async checkpoint writes I0422 07:36:30.241306 140560251893568 checkpoint_manager.py:1812] Found 0 checkpoint steps in gs://lance-maxtext/linen_ckpt_xpk_main_20260422_071422/linen_xpk_main_20260422_071422_08_checkpoint_async_true/checkpoints I0422 07:36:30.706984 140560251893568 checkpoint_manager.py:929] [process=6][thread=MainThread] CheckpointManager created, primary_host=0, CheckpointManagerOptions=CheckpointManagerOptions(save_interval_steps=1, max_to_keep=None, keep_time_interval=None, keep_period=None, should_keep_fn=None, best_fn=None, best_mode='max', keep_checkpoints_without_metrics=True, step_prefix=None, step_format_fixed_length=None, step_name_format=None, create=True, cleanup_tmp_directories=False, save_on_steps=frozenset(), single_host_load_and_broadcast=False, todelete_subdir=None, todelete_full_path=None, enable_background_delete=False, read_only=False, enable_async_checkpointing=True, async_options=None, multiprocessing_options=MultiprocessingOptions(primary_host=0, active_processes=None, barrier_sync_key_prefix=None), should_save_fn=None, file_options=FileOptions(path_permission_mode=None), save_root_metadata=True, temporary_path_class=None, save_decision_policy=FixedIntervalPolicy(interval=5), preservation_policy=LatestN(n=None), prevent_write_metrics=False, enable_should_save_is_saving_in_progress_check=True, enable_per_process_directory_creation=False, lightweight_initialize=False), root_directory=gs://lance-maxtext/linen_ckpt_xpk_main_20260422_071422/linen_xpk_main_20260422_071422_08_checkpoint_async_true/checkpoints: <orbax.checkpoint.checkpoint_manager.CheckpointManager object at 0x7fc26c7ad970> I0422 07:36:30.707173 140560251893568 checkpointing.py:301] Checkpoint manager created! I0422 07:36:31.634084 140560251893568 nnx_wrappers.py:437] Unknown Logical: bfloat16[32,2048,2048]...................................... ('activation_batch', 'activation_norm_length', 'activation_embed'). I0422 07:36:31.634203 140560251893568 nnx_wrappers.py:437] Unknown Physical: bfloat16[32,2048,2048]...................................... ('fsdp', None, None). I0422 07:36:32.032988 140560251893568 attentions.py:1088] attentions/inputs_q Logical: bfloat16[32,2048,2048]...................................... ('activation_batch', 'activation_attn_length', 'activation_attn_embed'). I0422 07:36:32.033098 140560251893568 attentions.py:1088] attentions/inputs_q Physical: bfloat16[32,2048,2048]...................................... ('fsdp', None, None). I0422 07:36:32.049908 140560251893568 attentions.py:1089] attentions/inputs_kv Logical: bfloat16[32,2048,2048]...................................... ('activation_batch', 'activation_attn_length', 'activation_attn_embed'). I0422 07:36:32.049968 140560251893568 attentions.py:1089] attentions/inputs_kv Physical: bfloat16[32,2048,2048]...................................... ('fsdp', None, None). I0422 07:36:32.074053 140560251893568 attentions.py:1154] attentions/query Logical: bfloat16[32,2048,16,128].................................... ('activation_kv_batch', 'activation_attn_length', 'activation_kv_heads', 'activation_kv_head_dim'). I0422 07:36:32.074123 140560251893568 attentions.py:1154] attentions/query Physical: bfloat16[32,2048,16,128].................................... ('fsdp', None, None, None). I0422 07:36:32.091127 140560251893568 attentions.py:1155] attentions/key Logical: bfloat16[32,2048,16,128].................................... ('activation_kv_batch', 'activation_attn_length', 'activation_kv_heads', 'activation_kv_head_dim'). I0422 07:36:32.091187 140560251893568 attentions.py:1155] attentions/key Physical: bfloat16[32,2048,16,128].................................... ('fsdp', None, None, None). I0422 07:36:32.107995 140560251893568 attentions.py:1156] attentions/value Logical: bfloat16[32,2048,16,128].................................... ('activation_kv_batch', 'activation_attn_length', 'activation_kv_heads', 'activation_kv_head_dim'). I0422 07:36:32.108060 140560251893568 attentions.py:1156] attentions/value Physical: bfloat16[32,2048,16,128].................................... ('fsdp', None, None, None). I0422 07:36:32.133135 140560251893568 attentions.py:1198] attentions/out Logical: bfloat16[32,2048,16,128].................................... ('activation_batch', 'activation_attn_length', 'activation_heads', 'activation_kv'). I0422 07:36:32.133205 140560251893568 attentions.py:1198] attentions/out Physical: bfloat16[32,2048,16,128].................................... ('fsdp', None, None, None). I0422 07:36:32.154633 140560251893568 linears.py:525] linears/x Logical: bfloat16[32,2048,7168]...................................... ('activation_batch', 'activation_length', 'activation_mlp'). I0422 07:36:32.154702 140560251893568 linears.py:525] linears/x Physical: bfloat16[32,2048,7168]...................................... ('fsdp', None, None). I0422 07:36:32.373439 140560251893568 checkpointing.py:577] checkpoint manager exists so trying to load this run's existing checkpoint I0422 07:36:32.373548 140560251893568 checkpointing.py:665] No existing checkpoints found, not restoring checkpoint. fsdp: 32 I0422 07:36:33.802842 140560251893568 maxtext_utils.py:1707] params/params/decoder/decoder_norm/scale Shape: float32[2048] Logical: P('norm',) Physical: (None,) I0422 07:36:33.802972 140560251893568 maxtext_utils.py:1707] params/params/decoder/layers/mlp/wi_0/kernel Shape: float32[2048,16,7168] Logical: P('embed', 'layers', 'mlp') Physical: ('fsdp', None, None) I0422 07:36:33.803024 140560251893568 maxtext_utils.py:1707] params/params/decoder/layers/mlp/wi_1/kernel Shape: float32[2048,16,7168] Logical: P('embed', 'layers', 'mlp') Physical: ('fsdp', None, None) I0422 07:36:33.803081 140560251893568 maxtext_utils.py:1707] params/params/decoder/layers/mlp/wo/kernel Shape: float32[7168,16,2048] Logical: P('mlp', 'layers', 'embed') Physical: (None, None, 'fsdp') I0422 07:36:33.803133 140560251893568 maxtext_utils.py:1707] params/params/decoder/layers/post_self_attention_layer_norm/scale Shape: float32[2048,16] Logical: P('norm', 'layers') Physical: (None, None) I0422 07:36:33.803172 140560251893568 maxtext_utils.py:1707] params/params/decoder/layers/pre_self_attention_layer_norm/scale Shape: float32[2048,16] Logical: P('norm', 'layers') Physical: (None, None) I0422 07:36:33.803228 140560251893568 maxtext_utils.py:1707] params/params/decoder/layers/self_attention/key/kernel Shape: float32[2048,16,16,128] Logical: P('embed', 'layers', 'kv_heads', 'kv_head_dim') Physical: ('fsdp', None, None, None) I0422 07:36:33.803296 140560251893568 maxtext_utils.py:1707] params/params/decoder/layers/self_attention/out/kernel Shape: float32[16,16,128,2048] Logical: P('heads', 'layers', 'kv', 'embed') Physical: (None, None, None, 'fsdp') I0422 07:36:33.803347 140560251893568 maxtext_utils.py:1707] params/params/decoder/layers/self_attention/query/kernel Shape: float32[2048,16,16,128] Logical: P('embed', 'layers', 'q_heads', 'kv') Physical: ('fsdp', None, None, None) I0422 07:36:33.803386 140560251893568 maxtext_utils.py:1707] params/params/decoder/layers/self_attention/value/kernel Shape: float32[2048,16,16,128] Logical: P('embed', 'layers', 'kv_heads', 'kv_head_dim') Physical: ('fsdp', None, None, None) I0422 07:36:33.803436 140560251893568 maxtext_utils.py:1707] params/params/decoder/logits_dense/kernel Shape: float32[2048,32000] Logical: P('embed_vocab', 'vocab') Physical: ('fsdp', None) I0422 07:36:33.803488 140560251893568 maxtext_utils.py:1707] params/params/token_embedder/embedding Shape: float32[32000,2048] Logical: P('vocab', 'embed_vocab') Physical: (None, 'fsdp') I0422 07:36:34.303491 140560251893568 train.py:155] train/xent Logical: float32[32,2048]............................................ ('activation_embed_and_logits_batch', 'activation_length'). I0422 07:36:34.303587 140560251893568 train.py:155] train/xent Physical: float32[32,2048]............................................ ('fsdp', None). I0422 07:36:34.319278 140560251893568 train.py:162] train/z_loss Logical: float32[32,2048]............................................ ('activation_embed_and_logits_batch', 'activation_length'). I0422 07:36:34.319353 140560251893568 train.py:162] train/z_loss Physical: float32[32,2048]............................................ ('fsdp', None). I0422 07:36:44.730931 140560251893568 max_utils.py:791] Total memory size: 1.5 GB, Output size: 0.4 GB, Temp size: 1.1 GB, Argument size: 0.4 GB, Host temp size: 0.0 GB. I0422 07:36:44.731780 140560251893568 metric_logger.py:301] number parameters: 1.104 billion I0422 07:36:56.571122 140560251893568 checkpointing.py:772] Waiting for step 0 to finish before checkpoint... I0422 07:36:56.723988 140560251893568 checkpointing.py:776] Waited 0.15284442901611328 seconds for step 0 to finish before starting checkpointing. I0422 07:36:56.726442 140560251893568 checkpoint_manager.py:2009] [process=6][thread=MainThread][wait_until_finished] No Save Finalize thread to wait for. Returning. I0422 07:36:56.728053 140560251893568 checkpoint_manager.py:1512] [process=6] Saving checkpoint at step 0 I0422 07:36:56.729454 140560251893568 event_tracking.py:70] [process=6] [async] Started save checkpoint @ gs://lance-maxtext/linen_ckpt_xpk_main_20260422_071422/linen_xpk_main_20260422_071422_08_checkpoint_async_true/checkpoints/0. I0422 07:36:57.063702 140560251893568 signaling_client.py:364] Using JaxDistributedSignalingClient I0422 07:36:57.064702 140560251893568 jax_array_handlers.py:360] Scheduling D2H of 39 prioritized jax.Array. I0422 07:36:57.064758 140560251893568 replica_slices.py:424] Transferring arrays to host memory with options: use_replica_parallel=True, min_slice_bytes_for_replica_parallel=None, max_replicas_for_replica_parallel=None, enable_pinned_host_transfer=False I0422 07:36:57.341797 140560251893568 base_pytree_checkpoint_handler.py:154] [process=6][thread=MainThread] Initiated "orbax.checkpoint._src.serialization.jax_array_handlers.ArrayHandler".serialize. Time taken: 0.278168s I0422 07:36:57.341982 140560251893568 base_pytree_checkpoint_handler.py:130] [process=6] /jax/orbax/write/blocking_gbytes_per_sec: 5.431 GiB/s (total gbytes: 1.5 GiB) (time elapsed: 0.2840549945831299 s) (per-host) I0422 07:36:57.342034 140560251893568 base_pytree_checkpoint_handler.py:768] [process=6][thread=MainThread] Initiated Pytree async_save. Time taken: 0.284118s (batch_requests_ready=0.002489s, total_serialization_initiated=0.281557s, others=0.000072s) I0422 07:36:57.342133 140560251893568 composite_checkpoint_handler.py:715] [process=6][thread=MainThread] Initiated CompositeCheckpointHandler.async_save. Time taken: 0.288485s (all_items=0.000018s, per_item={'items': '0.00001836'}, temp_paths=0.288466) I0422 07:36:57.343019 140560251893568 event_tracking.py:125] [process=6] [async] Finished blocking save in 0.61 seconds. Continuing save @ gs://lance-maxtext/linen_ckpt_xpk_main_20260422_071422/linen_xpk_main_20260422_071422_08_checkpoint_async_true/checkpoints/0. I0422 07:36:57.343326 140434086123264 async_checkpointer.py:76] [process=6][thread=async_save] Background save thread started. Deadline for this save operation is 2026-04-22 07:56:57.343277 I0422 07:36:57.356343 140560251893568 checkpoint_manager.py:1560] [process=6][thread=MainThread][step=0] Starting CheckpointManager Save Finalize thread=save_finalize I0422 07:36:57.356590 140430826366720 async_checkpointer.py:280] [process=6][thread=save_finalize] Waiting for background save thread=async_save. I0422 07:36:57.356718 140560251893568 standard_logger.py:34] {'step': 0, 'event_type': 'save', 'directory': 'gs://lance-maxtext/linen_ckpt_xpk_main_20260422_071422/linen_xpk_main_20260422_071422_08_checkpoint_async_true/checkpoints', 'reached_preemption': False, 'preemption_received_at': None, 'synchronous': False, 'wait_for_prev_start_time': 1776843416.7264235, 'wait_for_prev_duration_secs': 6.198883056640625e-05, 'time_between_consecutive_saves_sec': None, 'checkpointer_blocking_start_time': 1776843416.7280934, 'checkpointer_blocking_duration_secs': 0.6153802871704102, 'get_old_steps_start_time': 1776843417.343499, 'get_old_steps_duration_secs': 3.0040740966796875e-05, 'checkpoint_manager_blocking_start_time': 1776843416.7244124, 'checkpoint_manager_blocking_duration_secs': 0.6322684288024902} I0422 07:36:57.356830 140560251893568 checkpointing.py:408] Started an asynchronous checkpoint save for step 0 I0422 07:36:57.356877 140560251893568 max_utils.py:750] Memstats: After params initialized: I0422 07:36:57.356926 140560251893568 max_utils.py:756] Using (GB) 0.43 / 31.25 (1.376000%) on TPU_24(process=6,(0,6,0,0)) I0422 07:36:57.356957 140560251893568 max_utils.py:756] Using (GB) 0.43 / 31.25 (1.376000%) on TPU_25(process=6,(1,6,0,0)) I0422 07:36:57.356984 140560251893568 max_utils.py:756] Using (GB) 0.43 / 31.25 (1.376000%) on TPU_28(process=6,(0,7,0,0)) I0422 07:36:57.357010 140560251893568 max_utils.py:756] Using (GB) 0.43 / 31.25 (1.376000%) on TPU_29(process=6,(1,7,0,0)) I0422 07:36:57.670893 140560251893568 metric_logger.py:196] completed step: 0, seconds: 11.839, TFLOP/s/device: 1.148, Tokens/s/device: 172.984, total_weights: 65536, loss: 10.874, lm_loss: 10.874, perplexity: 52779.875 I0422 07:36:57.848477 140560251893568 metric_logger.py:196] completed step: 1, seconds: 1.098, TFLOP/s/device: 12.372, Tokens/s/device: 1864.912, total_weights: 65536, loss: 10.874, lm_loss: 10.874, perplexity: 52779.875 I0422 07:36:58.267165 140560251893568 metric_logger.py:196] completed step: 2, seconds: 0.030, TFLOP/s/device: 451.700, Tokens/s/device: 68085.106, total_weights: 65536, loss: 10.560, lm_loss: 10.560, perplexity: 38547.719 I0422 07:36:58.414682 140560251893568 metric_logger.py:196] completed step: 3, seconds: 0.419, TFLOP/s/device: 32.405, Tokens/s/device: 4884.471, total_weights: 65536, loss: 9.980, lm_loss: 9.980, perplexity: 21590.506 I0422 07:36:58.419114 140560251893568 checkpointing.py:772] Waiting for step 5 to finish before checkpoint... I0422 07:36:58.708707 140560251893568 checkpointing.py:776] Waited 0.2895541191101074 seconds for step 5 to finish before starting checkpointing. I0422 07:36:58.711662 140560251893568 checkpoint_manager.py:2020] [process=6][thread=MainThread][step=0][wait_until_finished] Waiting for Save Finalize thread (save_finalize) to complete. I0422 07:37:00.245566 2762 google_auth_provider.cc:181] Running on GCE, using service account 562977990677-compute@developer.gserviceaccount.com I0422 07:37:02.313789 140431369848576 array_metadata_store.py:203] [process=6][thread=array_type_handler] Wrote 39 array_metadata.ArrayMetadata to gs://lance-maxtext/linen_ckpt_xpk_main_20260422_071422/linen_xpk_main_20260422_071422_08_checkpoint_async_true/checkpoints/0/items/array_metadatas/process_6 I0422 07:37:34.530602 140434086123264 base_pytree_checkpoint_handler.py:130] [process=6] /jax/orbax/write/gbytes_per_sec: 42.153 MiB/s (total gbytes: 1.5 GiB) (time elapsed: 37.47265386581421 s) (per-host) I0422 07:37:34.530709 140434086123264 async_checkpointer.py:90] [process=6][thread=async_save] 3 Handler Commit operations completed. Time taken: 37.187274s. I0422 07:37:43.147625 140434086123264 async_checkpointer.py:160] [process=6][thread=async_save] Background save thread done. Time taken: 45.804173s. I0422 07:37:43.147898 140430826366720 async_checkpointer.py:288] [process=6][thread=save_finalize] Done with waiting for background save thread=async_save. I0422 07:37:43.148017 140430826366720 async_checkpointer.py:298] [process=6][thread=save_finalize] No errors found in background save thread=async_save. I0422 07:37:43.148067 140430826366720 checkpoint_manager.py:2137] [process=6][thread=save_finalize][step=0] CheckpointManager Save Finalize is syncing with other hosts... I0422 07:37:43.152615 140430826366720 checkpoint_manager.py:2146] [process=6][thread=save_finalize][step=0] CheckpointManager Save Finalize is done on all hosts. I0422 07:37:43.152783 140560251893568 checkpoint_manager.py:2032] [process=6][thread=MainThread][step=0][wait_until_finished] Done waiting for Save Finalize thread (save_finalize) running at step=0. W0422 07:37:43.152914 140560251893568 checkpoint_manager.py:1452] Waiting for previous save to complete took 44.441254 seconds. If this number is high, consider checkpointing less frequently. I0422 07:37:43.154811 140560251893568 checkpoint_manager.py:1512] [process=6] Saving checkpoint at step 5 I0422 07:37:43.157379 140560251893568 event_tracking.py:70] [process=6] [async] Started save checkpoint @ gs://lance-maxtext/linen_ckpt_xpk_main_20260422_071422/linen_xpk_main_20260422_071422_08_checkpoint_async_true/checkpoints/5. I0422 07:37:43.859611 140560251893568 jax_array_handlers.py:360] Scheduling D2H of 39 prioritized jax.Array. I0422 07:37:43.859718 140560251893568 replica_slices.py:424] Transferring arrays to host memory with options: use_replica_parallel=True, min_slice_bytes_for_replica_parallel=None, max_replicas_for_replica_parallel=None, enable_pinned_host_transfer=False I0422 07:37:43.901375 140560251893568 base_pytree_checkpoint_handler.py:154] [process=6][thread=MainThread] Initiated "orbax.checkpoint._src.serialization.jax_array_handlers.ArrayHandler".serialize. Time taken: 0.042939s I0422 07:37:43.901535 140560251893568 base_pytree_checkpoint_handler.py:130] [process=6] /jax/orbax/write/blocking_gbytes_per_sec: 33.103 GiB/s (total gbytes: 1.5 GiB) (time elapsed: 0.046599626541137695 s) (per-host) I0422 07:37:43.901583 140560251893568 base_pytree_checkpoint_handler.py:768] [process=6][thread=MainThread] Initiated Pytree async_save. Time taken: 0.046657s (batch_requests_ready=0.001988s, total_serialization_initiated=0.044606s, others=0.000062s) I0422 07:37:43.901671 140560251893568 composite_checkpoint_handler.py:715] [process=6][thread=MainThread] Initiated CompositeCheckpointHandler.async_save. Time taken: 0.051837s (all_items=0.000015s, per_item={'items': '0.00001526'}, temp_paths=0.051822) I0422 07:37:43.902395 140560251893568 event_tracking.py:125] [process=6] [async] Finished blocking save in 0.75 seconds. Continuing save @ gs://lance-maxtext/linen_ckpt_xpk_main_20260422_071422/linen_xpk_main_20260422_071422_08_checkpoint_async_true/checkpoints/5. I0422 07:37:43.902750 140406725515008 async_checkpointer.py:76] [process=6][thread=async_save] Background save thread started. Deadline for this save operation is 2026-04-22 07:57:43.902712 I0422 07:37:43.904758 140560251893568 checkpoint_manager.py:1560] [process=6][thread=MainThread][step=5] Starting CheckpointManager Save Finalize thread=save_finalize I0422 07:37:43.905018 140430826366720 async_checkpointer.py:280] [process=6][thread=save_finalize] Waiting for background save thread=async_save. I0422 07:37:43.905182 140560251893568 standard_logger.py:34] {'step': 5, 'event_type': 'save', 'directory': 'gs://lance-maxtext/linen_ckpt_xpk_main_20260422_071422/linen_xpk_main_20260422_071422_08_checkpoint_async_true/checkpoints', 'reached_preemption': False, 'preemption_received_at': None, 'synchronous': False, 'wait_for_prev_start_time': 1776843418.7116315, 'wait_for_prev_duration_secs': 44.441253900527954, 'time_between_consecutive_saves_sec': None, 'checkpointer_blocking_start_time': 1776843463.1548498, 'checkpointer_blocking_duration_secs': 0.7480525970458984, 'get_old_steps_start_time': 1776843463.9029245, 'get_old_steps_duration_secs': 3.0994415283203125e-05, 'checkpoint_manager_blocking_start_time': 1776843418.709344, 'checkpoint_manager_blocking_duration_secs': 45.19580388069153} I0422 07:37:43.905317 140560251893568 checkpointing.py:408] Started an asynchronous checkpoint save for step 5 I0422 07:37:43.906095 140560251893568 metric_logger.py:196] completed step: 4, seconds: 0.153, TFLOP/s/device: 88.819, Tokens/s/device: 13387.808, total_weights: 65536, loss: 9.460, lm_loss: 9.460, perplexity: 12840.355 I0422 07:37:43.912480 140560251893568 metric_logger.py:196] completed step: 5, seconds: 0.147, TFLOP/s/device: 92.199, Tokens/s/device: 13897.183, total_weights: 65536, loss: 8.959, lm_loss: 8.959, perplexity: 7776.467 I0422 07:37:50.091002 140431369848576 array_metadata_store.py:203] [process=6][thread=array_type_handler] Wrote 39 array_metadata.ArrayMetadata to gs://lance-maxtext/linen_ckpt_xpk_main_20260422_071422/linen_xpk_main_20260422_071422_08_checkpoint_async_true/checkpoints/5/items/array_metadatas/process_6 I0422 07:38:06.926161 140560251893568 metric_logger.py:196] completed step: 6, seconds: 45.492, TFLOP/s/device: 0.299, Tokens/s/device: 45.019, total_weights: 65536, loss: 8.469, lm_loss: 8.469, perplexity: 4762.740 I0422 07:38:07.073631 140560251893568 metric_logger.py:196] completed step: 7, seconds: 22.868, TFLOP/s/device: 0.594, Tokens/s/device: 89.559, total_weights: 65536, loss: 8.003, lm_loss: 8.003, perplexity: 2989.151 I0422 07:38:07.220907 140560251893568 metric_logger.py:196] completed step: 8, seconds: 0.152, TFLOP/s/device: 89.219, Tokens/s/device: 13448.115, total_weights: 65536, loss: 7.572, lm_loss: 7.572, perplexity: 1943.098 I0422 07:38:07.367826 140560251893568 checkpointing.py:772] Waiting for step 9 to finish before checkpoint... I0422 07:38:07.368509 140560251893568 checkpointing.py:776] Waited 0.0007007122039794922 seconds for step 9 to finish before starting checkpointing. I0422 07:38:07.370620 140560251893568 checkpoint_manager.py:2020] [process=6][thread=MainThread][step=5][wait_until_finished] Waiting for Save Finalize thread (save_finalize) to complete. I0422 07:38:26.397056 140406725515008 base_pytree_checkpoint_handler.py:130] [process=6] /jax/orbax/write/gbytes_per_sec: 37.130 MiB/s (total gbytes: 1.5 GiB) (time elapsed: 42.542078733444214 s) (per-host) I0422 07:38:26.397180 140406725515008 async_checkpointer.py:90] [process=6][thread=async_save] 3 Handler Commit operations completed. Time taken: 42.494312s. I0422 07:38:34.162400 140406725515008 async_checkpointer.py:160] [process=6][thread=async_save] Background save thread done. Time taken: 50.259517s. I0422 07:38:34.162705 140430826366720 async_checkpointer.py:288] [process=6][thread=save_finalize] Done with waiting for background save thread=async_save. I0422 07:38:34.162824 140430826366720 async_checkpointer.py:298] [process=6][thread=save_finalize] No errors found in background save thread=async_save. I0422 07:38:34.162871 140430826366720 checkpoint_manager.py:2137] [process=6][thread=save_finalize][step=5] CheckpointManager Save Finalize is syncing with other hosts... I0422 07:38:34.164162 140430826366720 checkpoint_manager.py:2146] [process=6][thread=save_finalize][step=5] CheckpointManager Save Finalize is done on all hosts. I0422 07:38:34.164362 140560251893568 checkpoint_manager.py:2032] [process=6][thread=MainThread][step=5][wait_until_finished] Done waiting for Save Finalize thread (save_finalize) running at step=5. W0422 07:38:34.164487 140560251893568 checkpoint_manager.py:1452] Waiting for previous save to complete took 26.793874 seconds. If this number is high, consider checkpointing less frequently. I0422 07:38:34.169481 140560251893568 checkpoint_manager.py:1512] [process=6] Saving checkpoint at step 9 I0422 07:38:34.171680 140560251893568 event_tracking.py:70] [process=6] [async] Started save checkpoint @ gs://lance-maxtext/linen_ckpt_xpk_main_20260422_071422/linen_xpk_main_20260422_071422_08_checkpoint_async_true/checkpoints/9. I0422 07:38:34.887322 140560251893568 jax_array_handlers.py:360] Scheduling D2H of 39 prioritized jax.Array. I0422 07:38:34.887422 140560251893568 replica_slices.py:424] Transferring arrays to host memory with options: use_replica_parallel=True, min_slice_bytes_for_replica_parallel=None, max_replicas_for_replica_parallel=None, enable_pinned_host_transfer=False I0422 07:38:34.914475 140560251893568 base_pytree_checkpoint_handler.py:154] [process=6][thread=MainThread] Initiated "orbax.checkpoint._src.serialization.jax_array_handlers.ArrayHandler".serialize. Time taken: 0.028069s I0422 07:38:34.914644 140560251893568 base_pytree_checkpoint_handler.py:130] [process=6] /jax/orbax/write/blocking_gbytes_per_sec: 48.430 GiB/s (total gbytes: 1.5 GiB) (time elapsed: 0.03185153007507324 s) (per-host) I0422 07:38:34.914695 140560251893568 base_pytree_checkpoint_handler.py:768] [process=6][thread=MainThread] Initiated Pytree async_save. Time taken: 0.031912s (batch_requests_ready=0.001907s, total_serialization_initiated=0.029939s, others=0.000066s) I0422 07:38:34.914790 140560251893568 composite_checkpoint_handler.py:715] [process=6][thread=MainThread] Initiated CompositeCheckpointHandler.async_save. Time taken: 0.036143s (all_items=0.000010s, per_item={'items': '0.00001025'}, temp_paths=0.036133) I0422 07:38:34.915525 140560251893568 event_tracking.py:125] [process=6] [async] Finished blocking save in 0.75 seconds. Continuing save @ gs://lance-maxtext/linen_ckpt_xpk_main_20260422_071422/linen_xpk_main_20260422_071422_08_checkpoint_async_true/checkpoints/9. I0422 07:38:34.915811 140430826366720 async_checkpointer.py:76] [process=6][thread=async_save] Background save thread started. Deadline for this save operation is 2026-04-22 07:58:34.915778 I0422 07:38:34.923186 140560251893568 checkpoint_manager.py:1560] [process=6][thread=MainThread][step=9] Starting CheckpointManager Save Finalize thread=save_finalize I0422 07:38:34.923417 140431369848576 async_checkpointer.py:280] [process=6][thread=save_finalize] Waiting for background save thread=async_save. I0422 07:38:34.923542 140560251893568 standard_logger.py:34] {'step': 9, 'event_type': 'save', 'directory': 'gs://lance-maxtext/linen_ckpt_xpk_main_20260422_071422/linen_xpk_main_20260422_071422_08_checkpoint_async_true/checkpoints', 'reached_preemption': False, 'preemption_received_at': None, 'synchronous': False, 'wait_for_prev_start_time': 1776843487.3705914, 'wait_for_prev_duration_secs': 26.79387402534485, 'time_between_consecutive_saves_sec': 24.21793246269226, 'checkpointer_blocking_start_time': 1776843514.1695218, 'checkpointer_blocking_duration_secs': 0.7463865280151367, 'get_old_steps_start_time': 1776843514.9159234, 'get_old_steps_duration_secs': 2.7418136596679688e-05, 'checkpoint_manager_blocking_start_time': 1776843487.3687456, 'checkpoint_manager_blocking_duration_secs': 27.55476403236389} I0422 07:38:34.923650 140560251893568 checkpointing.py:408] Started an asynchronous checkpoint save for step 9 I0422 07:38:34.923691 140560251893568 checkpoint_manager.py:2020] [process=6][thread=MainThread][step=9][wait_until_finished] Waiting for Save Finalize thread (save_finalize) to complete. I0422 07:38:41.013799 140433552307968 array_metadata_store.py:203] [process=6][thread=array_type_handler] Wrote 39 array_metadata.ArrayMetadata to gs://lance-maxtext/linen_ckpt_xpk_main_20260422_071422/linen_xpk_main_20260422_071422_08_checkpoint_async_true/checkpoints/9/items/array_metadatas/process_6 I0422 07:39:16.985243 140430826366720 base_pytree_checkpoint_handler.py:130] [process=6] /jax/orbax/write/gbytes_per_sec: 37.518 MiB/s (total gbytes: 1.5 GiB) (time elapsed: 42.1024112701416 s) (per-host) I0422 07:39:16.985400 140430826366720 async_checkpointer.py:90] [process=6][thread=async_save] 3 Handler Commit operations completed. Time taken: 42.068955s. I0422 07:39:26.379457 140430826366720 async_checkpointer.py:160] [process=6][thread=async_save] Background save thread done. Time taken: 51.462996s. I0422 07:39:26.379732 140431369848576 async_checkpointer.py:288] [process=6][thread=save_finalize] Done with waiting for background save thread=async_save. I0422 07:39:26.379858 140431369848576 async_checkpointer.py:298] [process=6][thread=save_finalize] No errors found in background save thread=async_save. I0422 07:39:26.379900 140431369848576 checkpoint_manager.py:2137] [process=6][thread=save_finalize][step=9] CheckpointManager Save Finalize is syncing with other hosts... I0422 07:39:26.381727 140431369848576 checkpoint_manager.py:2146] [process=6][thread=save_finalize][step=9] CheckpointManager Save Finalize is done on all hosts. I0422 07:39:26.381849 140560251893568 checkpoint_manager.py:2032] [process=6][thread=MainThread][step=9][wait_until_finished] Done waiting for Save Finalize thread (save_finalize) running at step=9. I0422 07:39:26.381940 140560251893568 checkpoint_manager.py:2009] [process=6][thread=MainThread][wait_until_finished] No Save Finalize thread to wait for. Returning. I0422 07:39:26.382831 140560251893568 metric_logger.py:196] completed step: 9, seconds: 0.147, TFLOP/s/device: 92.248, Tokens/s/device: 13904.636, 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: Wed Apr 22 07:39:40 UTC 2026 EXIT_CODE=0
XPK Start: Wed Apr 22 21:47:23 UTC 2026 PyTorch was not found. Models won't be available and only tokenizers, configuration and file/data utilities can be used. `rope_parameters`'s factor field must be a float >= 1, got 40 `rope_parameters`'s beta_fast field must be a float, got 32 `rope_parameters`'s beta_slow field must be a float, got 1 DeepseekV32Config got `key=rope_scaling` in kwargs but hasn't set it as attribute. For RoPE standardization you need to set `self.rope_parameters` in model's config. 2026-04-22 21:47:48.362883: E external/local_xla/xla/stream_executor/cuda/cuda_platform.cc:51] failed call to cuInit: INTERNAL: CUDA error: Failed call to cuInit: UNKNOWN ERROR (303) I0422 21:47:48.573587 133252648376128 max_utils.py:273] Attempting to initialize the jax distributed system... I0422 21:47:57.613454 133252648376128 distributed.py:149] Starting JAX distributed service on [::]:8482 I0422 21:47:57.615841 133252648376128 distributed.py:172] Connecting to JAX distributed service on mt-08-checkpoint-async--zjo0x-slice-job-0-0.mt-08-checkpoint-async--zjo0x:8482 I0422 21:47:59.447662 133252648376128 max_utils.py:284] Jax distributed system initialized! I0422 21:48:04.920175 133252648376128 max_utils.py:800] System Information: Jax Version: 0.9.2 I0422 21:48:04.920278 133252648376128 max_utils.py:801] System Information: Jaxlib Version: 0.9.2 I0422 21:48:04.920318 133252648376128 max_utils.py:802] System Information: Jax Backend: PJRT C API TFRT TPU v6 lite Built on Mar 4 2026 11:32:08 (1772652728) cl/878335365 I0422 21:48:04.920353 133252648376128 train_utils.py:361] WARNING: Sequence packing is essentially ignored for synthetic data. Please use a real dataset to use sequence packing. I0422 21:48:05.610484 133252648376128 maxtext_utils.py:1565] Num_devices: 32, shape (1, 1, 1, 32, 1, 1, 1, 1, 1, 1, 1, 1, 1) I0422 21:48:05.610797 133252648376128 checkpointing.py:677] Setting up checkpoint logger... I0422 21:48:05.610854 133252648376128 checkpointing.py:233] Creating checkpoint manager with ocdbt=True and zarr3=True I0422 21:48:05.610900 133252648376128 pytree_checkpoint_handler.py:592] save_device_host_concurrent_bytes=None I0422 21:48:05.611296 133252648376128 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 0x7930ac71c9b0>, enable_pinned_host_transfer=False, save_concurrent_bytes: 96000000000 (89.4 GiB), restore_concurrent_bytes: 96000000000 (89.4 GiB) I0422 21:48:08.571077 133252648376128 checkpointing.py:265] Enabling policy for fixed interval checkpointing. I0422 21:48:08.571315 133252648376128 checkpoint_manager.py:708] [process=3][thread=MainThread] CheckpointManager init: checkpointers=None, item_names=('items',), item_handlers={'items': <orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeCheckpointHandler object at 0x791d343d5580>}, handler_registry=None I0422 21:48:08.571555 133252648376128 composite_checkpoint_handler.py:237] Deferred registration for item: "items". Adding handler `<orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeCheckpointHandler object at 0x791d343d5580>` for item "items" and save args `<class 'orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeSaveArgs'>` and restore args `<class 'orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeRestoreArgs'>` to `_handler_registry`. I0422 21:48:08.571603 133252648376128 composite_checkpoint_handler.py:237] Deferred registration for item: "metrics". Adding handler `<orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonCheckpointHandler object at 0x791d343db350>` for item "metrics" and save args `<class 'orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonSaveArgs'>` and restore args `<class 'orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonRestoreArgs'>` to `_handler_registry`. I0422 21:48:08.571640 133252648376128 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 0x791d343d5580>, ('items', <class 'orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeRestoreArgs'>): <orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeCheckpointHandler object at 0x791d343d5580>, ('metrics', <class 'orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonSaveArgs'>): <orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonCheckpointHandler object at 0x791d343db350>, ('metrics', <class 'orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonRestoreArgs'>): <orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonCheckpointHandler object at 0x791d343db350>}). I0422 21:48:08.572044 133252648376128 abstract_checkpointer.py:35] orbax-checkpoint version: 0.11.34 I0422 21:48:08.572116 133252648376128 async_checkpointer.py:192] [process=3][thread=MainThread] Using barrier_sync_fn: <function get_barrier_sync_fn.<locals>._fn at 0x791d342032e0> timeout: 1200 secs and primary_host=0 for async checkpoint writes I0422 21:48:09.869632 133252648376128 checkpoint_manager.py:1812] Found 0 checkpoint steps in gs://lance-maxtext/linen_ckpt_xpk_test_pipeline_scan_nnx_20260422_212603/linen_xpk_test_pipeline_scan_nnx_20260422_212603_08_checkpoint_async_true/checkpoints I0422 21:48:10.288543 133252648376128 checkpoint_manager.py:929] [process=3][thread=MainThread] CheckpointManager created, primary_host=0, CheckpointManagerOptions=CheckpointManagerOptions(save_interval_steps=1, max_to_keep=None, keep_time_interval=None, keep_period=None, should_keep_fn=None, best_fn=None, best_mode='max', keep_checkpoints_without_metrics=True, step_prefix=None, step_format_fixed_length=None, step_name_format=None, create=True, cleanup_tmp_directories=False, save_on_steps=frozenset(), single_host_load_and_broadcast=False, todelete_subdir=None, todelete_full_path=None, enable_background_delete=False, read_only=False, enable_async_checkpointing=True, async_options=None, multiprocessing_options=MultiprocessingOptions(primary_host=0, active_processes=None, barrier_sync_key_prefix=None), should_save_fn=None, file_options=FileOptions(path_permission_mode=None), save_root_metadata=True, temporary_path_class=None, save_decision_policy=FixedIntervalPolicy(interval=5), preservation_policy=LatestN(n=None), prevent_write_metrics=False, enable_should_save_is_saving_in_progress_check=True, enable_per_process_directory_creation=False, lightweight_initialize=False), root_directory=gs://lance-maxtext/linen_ckpt_xpk_test_pipeline_scan_nnx_20260422_212603/linen_xpk_test_pipeline_scan_nnx_20260422_212603_08_checkpoint_async_true/checkpoints: <orbax.checkpoint.checkpoint_manager.CheckpointManager object at 0x791d343d9b20> I0422 21:48:10.288723 133252648376128 checkpointing.py:301] Checkpoint manager created! I0422 21:48:11.229435 133252648376128 nnx_wrappers.py:453] Unknown Logical: bfloat16[32,2048,2048]...................................... ('activation_batch', 'activation_norm_length', 'activation_embed'). I0422 21:48:11.229560 133252648376128 nnx_wrappers.py:453] Unknown Physical: bfloat16[32,2048,2048]...................................... ('fsdp', None, None). I0422 21:48:11.614224 133252648376128 attentions.py:1088] attentions/inputs_q Logical: bfloat16[32,2048,2048]...................................... ('activation_batch', 'activation_attn_length', 'activation_attn_embed'). I0422 21:48:11.614318 133252648376128 attentions.py:1088] attentions/inputs_q Physical: bfloat16[32,2048,2048]...................................... ('fsdp', None, None). I0422 21:48:11.630804 133252648376128 attentions.py:1089] attentions/inputs_kv Logical: bfloat16[32,2048,2048]...................................... ('activation_batch', 'activation_attn_length', 'activation_attn_embed'). I0422 21:48:11.630874 133252648376128 attentions.py:1089] attentions/inputs_kv Physical: bfloat16[32,2048,2048]...................................... ('fsdp', None, None). I0422 21:48:11.654479 133252648376128 attentions.py:1154] attentions/query Logical: bfloat16[32,2048,16,128].................................... ('activation_kv_batch', 'activation_attn_length', 'activation_kv_heads', 'activation_kv_head_dim'). I0422 21:48:11.654549 133252648376128 attentions.py:1154] attentions/query Physical: bfloat16[32,2048,16,128].................................... ('fsdp', None, None, None). I0422 21:48:11.671043 133252648376128 attentions.py:1155] attentions/key Logical: bfloat16[32,2048,16,128].................................... ('activation_kv_batch', 'activation_attn_length', 'activation_kv_heads', 'activation_kv_head_dim'). I0422 21:48:11.671104 133252648376128 attentions.py:1155] attentions/key Physical: bfloat16[32,2048,16,128].................................... ('fsdp', None, None, None). I0422 21:48:11.687577 133252648376128 attentions.py:1156] attentions/value Logical: bfloat16[32,2048,16,128].................................... ('activation_kv_batch', 'activation_attn_length', 'activation_kv_heads', 'activation_kv_head_dim'). I0422 21:48:11.687638 133252648376128 attentions.py:1156] attentions/value Physical: bfloat16[32,2048,16,128].................................... ('fsdp', None, None, None). I0422 21:48:11.712433 133252648376128 attentions.py:1198] attentions/out Logical: bfloat16[32,2048,16,128].................................... ('activation_batch', 'activation_attn_length', 'activation_heads', 'activation_kv'). I0422 21:48:11.712532 133252648376128 attentions.py:1198] attentions/out Physical: bfloat16[32,2048,16,128].................................... ('fsdp', None, None, None). I0422 21:48:11.733667 133252648376128 linears.py:525] linears/x Logical: bfloat16[32,2048,7168]...................................... ('activation_batch', 'activation_length', 'activation_mlp'). I0422 21:48:11.733740 133252648376128 linears.py:525] linears/x Physical: bfloat16[32,2048,7168]...................................... ('fsdp', None, None). I0422 21:48:11.950081 133252648376128 checkpointing.py:577] checkpoint manager exists so trying to load this run's existing checkpoint I0422 21:48:11.950192 133252648376128 checkpointing.py:665] No existing checkpoints found, not restoring checkpoint. fsdp: 32 I0422 21:48:13.377679 133252648376128 maxtext_utils.py:1668] params/params/decoder/decoder_norm/scale Shape: float32[2048] Logical: P('norm',) Physical: (None,) I0422 21:48:13.377811 133252648376128 maxtext_utils.py:1668] params/params/decoder/layers/mlp/wi_0/kernel Shape: float32[2048,16,7168] Logical: P('embed', 'layers', 'mlp') Physical: ('fsdp', None, None) I0422 21:48:13.377868 133252648376128 maxtext_utils.py:1668] params/params/decoder/layers/mlp/wi_1/kernel Shape: float32[2048,16,7168] Logical: P('embed', 'layers', 'mlp') Physical: ('fsdp', None, None) I0422 21:48:13.377936 133252648376128 maxtext_utils.py:1668] params/params/decoder/layers/mlp/wo/kernel Shape: float32[7168,16,2048] Logical: P('mlp', 'layers', 'embed') Physical: (None, None, 'fsdp') I0422 21:48:13.377995 133252648376128 maxtext_utils.py:1668] params/params/decoder/layers/post_self_attention_layer_norm/scale Shape: float32[2048,16] Logical: P('norm', 'layers') Physical: (None, None) I0422 21:48:13.378036 133252648376128 maxtext_utils.py:1668] params/params/decoder/layers/pre_self_attention_layer_norm/scale Shape: float32[2048,16] Logical: P('norm', 'layers') Physical: (None, None) I0422 21:48:13.378089 133252648376128 maxtext_utils.py:1668] params/params/decoder/layers/self_attention/key/kernel Shape: float32[2048,16,16,128] Logical: P('embed', 'layers', 'kv_heads', 'kv_head_dim') Physical: ('fsdp', None, None, None) I0422 21:48:13.378142 133252648376128 maxtext_utils.py:1668] params/params/decoder/layers/self_attention/out/kernel Shape: float32[16,16,128,2048] Logical: P('heads', 'layers', 'kv', 'embed') Physical: (None, None, None, 'fsdp') I0422 21:48:13.378183 133252648376128 maxtext_utils.py:1668] params/params/decoder/layers/self_attention/query/kernel Shape: float32[2048,16,16,128] Logical: P('embed', 'layers', 'q_heads', 'kv') Physical: ('fsdp', None, None, None) I0422 21:48:13.378219 133252648376128 maxtext_utils.py:1668] params/params/decoder/layers/self_attention/value/kernel Shape: float32[2048,16,16,128] Logical: P('embed', 'layers', 'kv_heads', 'kv_head_dim') Physical: ('fsdp', None, None, None) I0422 21:48:13.378266 133252648376128 maxtext_utils.py:1668] params/params/decoder/logits_dense/kernel Shape: float32[2048,32000] Logical: P('embed_vocab', 'vocab') Physical: ('fsdp', None) I0422 21:48:13.378315 133252648376128 maxtext_utils.py:1668] params/params/token_embedder/embedding Shape: float32[32000,2048] Logical: P('vocab', 'embed_vocab') Physical: (None, 'fsdp') I0422 21:48:13.870315 133252648376128 train.py:155] train/xent Logical: float32[32,2048]............................................ ('activation_embed_and_logits_batch', 'activation_length'). I0422 21:48:13.870412 133252648376128 train.py:155] train/xent Physical: float32[32,2048]............................................ ('fsdp', None). I0422 21:48:13.885989 133252648376128 train.py:162] train/z_loss Logical: float32[32,2048]............................................ ('activation_embed_and_logits_batch', 'activation_length'). I0422 21:48:13.886052 133252648376128 train.py:162] train/z_loss Physical: float32[32,2048]............................................ ('fsdp', None). I0422 21:48:24.976769 133252648376128 max_utils.py:791] Total memory size: 1.5 GB, Output size: 0.4 GB, Temp size: 1.1 GB, Argument size: 0.4 GB, Host temp size: 0.0 GB. I0422 21:48:24.977616 133252648376128 metric_logger.py:301] number parameters: 1.104 billion I0422 21:48:36.264398 133252648376128 checkpointing.py:772] Waiting for step 0 to finish before checkpoint... I0422 21:48:36.731564 133252648376128 checkpointing.py:776] Waited 0.4671440124511719 seconds for step 0 to finish before starting checkpointing. I0422 21:48:36.733812 133252648376128 checkpoint_manager.py:2009] [process=3][thread=MainThread][wait_until_finished] No Save Finalize thread to wait for. Returning. I0422 21:48:36.735565 133252648376128 checkpoint_manager.py:1512] [process=3] Saving checkpoint at step 0 I0422 21:48:36.737012 133252648376128 event_tracking.py:70] [process=3] [async] Started save checkpoint @ gs://lance-maxtext/linen_ckpt_xpk_test_pipeline_scan_nnx_20260422_212603/linen_xpk_test_pipeline_scan_nnx_20260422_212603_08_checkpoint_async_true/checkpoints/0. I0422 21:48:37.459332 133252648376128 signaling_client.py:364] Using JaxDistributedSignalingClient I0422 21:48:37.460354 133252648376128 jax_array_handlers.py:360] Scheduling D2H of 39 prioritized jax.Array. I0422 21:48:37.460412 133252648376128 replica_slices.py:424] Transferring arrays to host memory with options: use_replica_parallel=True, min_slice_bytes_for_replica_parallel=None, max_replicas_for_replica_parallel=None, enable_pinned_host_transfer=False I0422 21:48:37.738136 133252648376128 base_pytree_checkpoint_handler.py:154] [process=3][thread=MainThread] Initiated "orbax.checkpoint._src.serialization.jax_array_handlers.ArrayHandler".serialize. Time taken: 0.278877s I0422 21:48:37.738313 133252648376128 base_pytree_checkpoint_handler.py:130] [process=3] /jax/orbax/write/blocking_gbytes_per_sec: 5.422 GiB/s (total gbytes: 1.5 GiB) (time elapsed: 0.28447699546813965 s) (per-host) I0422 21:48:37.738370 133252648376128 base_pytree_checkpoint_handler.py:768] [process=3][thread=MainThread] Initiated Pytree async_save. Time taken: 0.284544s (batch_requests_ready=0.002455s, total_serialization_initiated=0.282013s, others=0.000076s) I0422 21:48:37.738475 133252648376128 composite_checkpoint_handler.py:715] [process=3][thread=MainThread] Initiated CompositeCheckpointHandler.async_save. Time taken: 0.288743s (all_items=0.000017s, per_item={'items': '0.00001717'}, temp_paths=0.288725) I0422 21:48:37.739269 133252648376128 event_tracking.py:125] [process=3] [async] Finished blocking save in 1.00 seconds. Continuing save @ gs://lance-maxtext/linen_ckpt_xpk_test_pipeline_scan_nnx_20260422_212603/linen_xpk_test_pipeline_scan_nnx_20260422_212603_08_checkpoint_async_true/checkpoints/0. I0422 21:48:37.739608 133126537869056 async_checkpointer.py:76] [process=3][thread=async_save] Background save thread started. Deadline for this save operation is 2026-04-22 22:08:37.739571 I0422 21:48:37.742655 133252648376128 checkpoint_manager.py:1560] [process=3][thread=MainThread][step=0] Starting CheckpointManager Save Finalize thread=save_finalize I0422 21:48:37.742967 133126483326720 async_checkpointer.py:280] [process=3][thread=save_finalize] Waiting for background save thread=async_save. I0422 21:48:37.743124 133252648376128 standard_logger.py:34] {'step': 0, 'event_type': 'save', 'directory': 'gs://lance-maxtext/linen_ckpt_xpk_test_pipeline_scan_nnx_20260422_212603/linen_xpk_test_pipeline_scan_nnx_20260422_212603_08_checkpoint_async_true/checkpoints', 'reached_preemption': False, 'preemption_received_at': None, 'synchronous': False, 'wait_for_prev_start_time': 1776894516.7337935, 'wait_for_prev_duration_secs': 6.246566772460938e-05, 'time_between_consecutive_saves_sec': None, 'checkpointer_blocking_start_time': 1776894516.7356038, 'checkpointer_blocking_duration_secs': 1.004164695739746, 'get_old_steps_start_time': 1776894517.7397926, 'get_old_steps_duration_secs': 2.9087066650390625e-05, 'checkpoint_manager_blocking_start_time': 1776894516.7320135, 'checkpoint_manager_blocking_duration_secs': 1.0110726356506348} I0422 21:48:37.743229 133252648376128 checkpointing.py:408] Started an asynchronous checkpoint save for step 0 I0422 21:48:37.743276 133252648376128 max_utils.py:750] Memstats: After params initialized: I0422 21:48:37.743322 133252648376128 max_utils.py:756] Using (GB) 0.43 / 31.25 (1.376000%) on TPU_10(process=3,(2,2,0,0)) I0422 21:48:37.743352 133252648376128 max_utils.py:756] Using (GB) 0.43 / 31.25 (1.376000%) on TPU_11(process=3,(3,2,0,0)) I0422 21:48:37.743378 133252648376128 max_utils.py:756] Using (GB) 0.43 / 31.25 (1.376000%) on TPU_14(process=3,(2,3,0,0)) I0422 21:48:37.743402 133252648376128 max_utils.py:756] Using (GB) 0.43 / 31.25 (1.376000%) on TPU_15(process=3,(3,3,0,0)) I0422 21:48:38.057791 133252648376128 metric_logger.py:196] completed step: 0, seconds: 11.287, TFLOP/s/device: 1.204, Tokens/s/device: 181.453, total_weights: 65536, loss: 10.874, lm_loss: 10.874, perplexity: 52779.875 I0422 21:48:38.242092 133252648376128 metric_logger.py:196] completed step: 1, seconds: 1.792, TFLOP/s/device: 7.583, Tokens/s/device: 1142.924, total_weights: 65536, loss: 10.874, lm_loss: 10.874, perplexity: 52779.875 I0422 21:48:38.642230 133252648376128 metric_logger.py:196] completed step: 2, seconds: 0.037, TFLOP/s/device: 367.389, Tokens/s/device: 55376.795, total_weights: 65536, loss: 10.560, lm_loss: 10.560, perplexity: 38547.719 I0422 21:48:38.789768 133252648376128 metric_logger.py:196] completed step: 3, seconds: 0.400, TFLOP/s/device: 33.931, Tokens/s/device: 5114.438, total_weights: 65536, loss: 9.980, lm_loss: 9.980, perplexity: 21590.506 I0422 21:48:38.794204 133252648376128 checkpointing.py:772] Waiting for step 5 to finish before checkpoint... I0422 21:48:39.084123 133252648376128 checkpointing.py:776] Waited 0.2898852825164795 seconds for step 5 to finish before starting checkpointing. I0422 21:48:39.087204 133252648376128 checkpoint_manager.py:2020] [process=3][thread=MainThread][step=0][wait_until_finished] Waiting for Save Finalize thread (save_finalize) to complete. I0422 21:48:40.595588 2754 google_auth_provider.cc:181] Running on GCE, using service account 562977990677-compute@developer.gserviceaccount.com I0422 21:48:42.623198 133126491719424 array_metadata_store.py:203] [process=3][thread=array_type_handler] Wrote 39 array_metadata.ArrayMetadata to gs://lance-maxtext/linen_ckpt_xpk_test_pipeline_scan_nnx_20260422_212603/linen_xpk_test_pipeline_scan_nnx_20260422_212603_08_checkpoint_async_true/checkpoints/0/items/array_metadatas/process_3 I0422 21:49:14.502101 133126537869056 base_pytree_checkpoint_handler.py:130] [process=3] /jax/orbax/write/gbytes_per_sec: 42.636 MiB/s (total gbytes: 1.5 GiB) (time elapsed: 37.04823875427246 s) (per-host) I0422 21:49:14.502205 133126537869056 async_checkpointer.py:90] [process=3][thread=async_save] 3 Handler Commit operations completed. Time taken: 36.762480s. I0422 21:49:23.725062 133126537869056 async_checkpointer.py:160] [process=3][thread=async_save] Background save thread done. Time taken: 45.985320s. I0422 21:49:23.725432 133126483326720 async_checkpointer.py:288] [process=3][thread=save_finalize] Done with waiting for background save thread=async_save. I0422 21:49:23.725556 133126483326720 async_checkpointer.py:298] [process=3][thread=save_finalize] No errors found in background save thread=async_save. I0422 21:49:23.725608 133126483326720 checkpoint_manager.py:2137] [process=3][thread=save_finalize][step=0] CheckpointManager Save Finalize is syncing with other hosts... I0422 21:49:23.728230 133126483326720 checkpoint_manager.py:2146] [process=3][thread=save_finalize][step=0] CheckpointManager Save Finalize is done on all hosts. I0422 21:49:23.728414 133252648376128 checkpoint_manager.py:2032] [process=3][thread=MainThread][step=0][wait_until_finished] Done waiting for Save Finalize thread (save_finalize) running at step=0. W0422 21:49:23.728560 133252648376128 checkpoint_manager.py:1452] Waiting for previous save to complete took 44.641352 seconds. If this number is high, consider checkpointing less frequently. I0422 21:49:23.730626 133252648376128 checkpoint_manager.py:1512] [process=3] Saving checkpoint at step 5 I0422 21:49:23.733165 133252648376128 event_tracking.py:70] [process=3] [async] Started save checkpoint @ gs://lance-maxtext/linen_ckpt_xpk_test_pipeline_scan_nnx_20260422_212603/linen_xpk_test_pipeline_scan_nnx_20260422_212603_08_checkpoint_async_true/checkpoints/5. I0422 21:49:24.024988 133252648376128 jax_array_handlers.py:360] Scheduling D2H of 39 prioritized jax.Array. I0422 21:49:24.025081 133252648376128 replica_slices.py:424] Transferring arrays to host memory with options: use_replica_parallel=True, min_slice_bytes_for_replica_parallel=None, max_replicas_for_replica_parallel=None, enable_pinned_host_transfer=False I0422 21:49:24.061361 133252648376128 base_pytree_checkpoint_handler.py:154] [process=3][thread=MainThread] Initiated "orbax.checkpoint._src.serialization.jax_array_handlers.ArrayHandler".serialize. Time taken: 0.037491s I0422 21:49:24.061530 133252648376128 base_pytree_checkpoint_handler.py:130] [process=3] /jax/orbax/write/blocking_gbytes_per_sec: 37.733 GiB/s (total gbytes: 1.5 GiB) (time elapsed: 0.04088163375854492 s) (per-host) I0422 21:49:24.061583 133252648376128 base_pytree_checkpoint_handler.py:768] [process=3][thread=MainThread] Initiated Pytree async_save. Time taken: 0.040944s (batch_requests_ready=0.001760s, total_serialization_initiated=0.039115s, others=0.000069s) I0422 21:49:24.061672 133252648376128 composite_checkpoint_handler.py:715] [process=3][thread=MainThread] Initiated CompositeCheckpointHandler.async_save. Time taken: 0.046035s (all_items=0.000015s, per_item={'items': '0.00001526'}, temp_paths=0.046020) I0422 21:49:24.062348 133252648376128 event_tracking.py:125] [process=3] [async] Finished blocking save in 0.33 seconds. Continuing save @ gs://lance-maxtext/linen_ckpt_xpk_test_pipeline_scan_nnx_20260422_212603/linen_xpk_test_pipeline_scan_nnx_20260422_212603_08_checkpoint_async_true/checkpoints/5. I0422 21:49:24.062625 133126483326720 async_checkpointer.py:76] [process=3][thread=async_save] Background save thread started. Deadline for this save operation is 2026-04-22 22:09:24.062600 I0422 21:49:24.067959 133252648376128 checkpoint_manager.py:1560] [process=3][thread=MainThread][step=5] Starting CheckpointManager Save Finalize thread=save_finalize I0422 21:49:24.068156 133126491719424 async_checkpointer.py:280] [process=3][thread=save_finalize] Waiting for background save thread=async_save. I0422 21:49:24.068271 133252648376128 standard_logger.py:34] {'step': 5, 'event_type': 'save', 'directory': 'gs://lance-maxtext/linen_ckpt_xpk_test_pipeline_scan_nnx_20260422_212603/linen_xpk_test_pipeline_scan_nnx_20260422_212603_08_checkpoint_async_true/checkpoints', 'reached_preemption': False, 'preemption_received_at': None, 'synchronous': False, 'wait_for_prev_start_time': 1776894519.0871747, 'wait_for_prev_duration_secs': 44.64135241508484, 'time_between_consecutive_saves_sec': None, 'checkpointer_blocking_start_time': 1776894563.7306657, 'checkpointer_blocking_duration_secs': 0.3320634365081787, 'get_old_steps_start_time': 1776894564.0627525, 'get_old_steps_duration_secs': 2.9802322387695312e-05, 'checkpoint_manager_blocking_start_time': 1776894519.084728, 'checkpoint_manager_blocking_duration_secs': 44.98350954055786} I0422 21:49:24.068381 133252648376128 checkpointing.py:408] Started an asynchronous checkpoint save for step 5 I0422 21:49:24.069235 133252648376128 metric_logger.py:196] completed step: 4, seconds: 0.153, TFLOP/s/device: 88.828, Tokens/s/device: 13389.121, total_weights: 65536, loss: 9.460, lm_loss: 9.460, perplexity: 12840.355 I0422 21:49:24.075403 133252648376128 metric_logger.py:196] completed step: 5, seconds: 0.147, TFLOP/s/device: 92.160, Tokens/s/device: 13891.338, total_weights: 65536, loss: 8.959, lm_loss: 8.959, perplexity: 7776.467 I0422 21:49:29.441309 133120190899968 array_metadata_store.py:203] [process=3][thread=array_type_handler] Wrote 39 array_metadata.ArrayMetadata to gs://lance-maxtext/linen_ckpt_xpk_test_pipeline_scan_nnx_20260422_212603/linen_xpk_test_pipeline_scan_nnx_20260422_212603_08_checkpoint_async_true/checkpoints/5/items/array_metadatas/process_3 I0422 21:49:46.651422 133252648376128 metric_logger.py:196] completed step: 6, seconds: 45.280, TFLOP/s/device: 0.300, Tokens/s/device: 45.230, total_weights: 65536, loss: 8.469, lm_loss: 8.469, perplexity: 4762.740 I0422 21:49:46.798786 133252648376128 metric_logger.py:196] completed step: 7, seconds: 22.430, TFLOP/s/device: 0.606, Tokens/s/device: 91.307, total_weights: 65536, loss: 8.003, lm_loss: 8.003, perplexity: 2989.151 I0422 21:49:46.946196 133252648376128 metric_logger.py:196] completed step: 8, seconds: 0.153, TFLOP/s/device: 89.068, Tokens/s/device: 13425.371, total_weights: 65536, loss: 7.572, lm_loss: 7.572, perplexity: 1943.098 I0422 21:49:47.092659 133252648376128 checkpointing.py:772] Waiting for step 9 to finish before checkpoint... I0422 21:49:47.093333 133252648376128 checkpointing.py:776] Waited 0.0006952285766601562 seconds for step 9 to finish before starting checkpointing. I0422 21:49:47.095515 133252648376128 checkpoint_manager.py:2020] [process=3][thread=MainThread][step=5][wait_until_finished] Waiting for Save Finalize thread (save_finalize) to complete. I0422 21:50:05.089614 133126483326720 base_pytree_checkpoint_handler.py:130] [process=3] /jax/orbax/write/gbytes_per_sec: 38.462 MiB/s (total gbytes: 1.5 GiB) (time elapsed: 41.06892681121826 s) (per-host) I0422 21:50:05.089744 133126483326720 async_checkpointer.py:90] [process=3][thread=async_save] 3 Handler Commit operations completed. Time taken: 41.027039s. I0422 21:50:13.697832 133126483326720 async_checkpointer.py:160] [process=3][thread=async_save] Background save thread done. Time taken: 49.635111s. I0422 21:50:13.698094 133126491719424 async_checkpointer.py:288] [process=3][thread=save_finalize] Done with waiting for background save thread=async_save. I0422 21:50:13.698149 133126491719424 async_checkpointer.py:298] [process=3][thread=save_finalize] No errors found in background save thread=async_save. I0422 21:50:13.698194 133126491719424 checkpoint_manager.py:2137] [process=3][thread=save_finalize][step=5] CheckpointManager Save Finalize is syncing with other hosts... I0422 21:50:13.700170 133126491719424 checkpoint_manager.py:2146] [process=3][thread=save_finalize][step=5] CheckpointManager Save Finalize is done on all hosts. I0422 21:50:13.700350 133252648376128 checkpoint_manager.py:2032] [process=3][thread=MainThread][step=5][wait_until_finished] Done waiting for Save Finalize thread (save_finalize) running at step=5. W0422 21:50:13.700479 133252648376128 checkpoint_manager.py:1452] Waiting for previous save to complete took 26.604973 seconds. If this number is high, consider checkpointing less frequently. I0422 21:50:13.702010 133252648376128 checkpoint_manager.py:1512] [process=3] Saving checkpoint at step 9 I0422 21:50:13.704193 133252648376128 event_tracking.py:70] [process=3] [async] Started save checkpoint @ gs://lance-maxtext/linen_ckpt_xpk_test_pipeline_scan_nnx_20260422_212603/linen_xpk_test_pipeline_scan_nnx_20260422_212603_08_checkpoint_async_true/checkpoints/9. I0422 21:50:14.405811 133252648376128 jax_array_handlers.py:360] Scheduling D2H of 39 prioritized jax.Array. I0422 21:50:14.405905 133252648376128 replica_slices.py:424] Transferring arrays to host memory with options: use_replica_parallel=True, min_slice_bytes_for_replica_parallel=None, max_replicas_for_replica_parallel=None, enable_pinned_host_transfer=False I0422 21:50:14.443167 133252648376128 base_pytree_checkpoint_handler.py:154] [process=3][thread=MainThread] Initiated "orbax.checkpoint._src.serialization.jax_array_handlers.ArrayHandler".serialize. Time taken: 0.038296s I0422 21:50:14.443297 133252648376128 base_pytree_checkpoint_handler.py:130] [process=3] /jax/orbax/write/blocking_gbytes_per_sec: 37.016 GiB/s (total gbytes: 1.5 GiB) (time elapsed: 0.04167366027832031 s) (per-host) I0422 21:50:14.443345 133252648376128 base_pytree_checkpoint_handler.py:768] [process=3][thread=MainThread] Initiated Pytree async_save. Time taken: 0.041731s (batch_requests_ready=0.001614s, total_serialization_initiated=0.040055s, others=0.000062s) I0422 21:50:14.443418 133252648376128 composite_checkpoint_handler.py:715] [process=3][thread=MainThread] Initiated CompositeCheckpointHandler.async_save. Time taken: 0.046095s (all_items=0.000017s, per_item={'items': '0.00001717'}, temp_paths=0.046078) I0422 21:50:14.444095 133252648376128 event_tracking.py:125] [process=3] [async] Finished blocking save in 0.74 seconds. Continuing save @ gs://lance-maxtext/linen_ckpt_xpk_test_pipeline_scan_nnx_20260422_212603/linen_xpk_test_pipeline_scan_nnx_20260422_212603_08_checkpoint_async_true/checkpoints/9. I0422 21:50:14.444427 133125440456448 async_checkpointer.py:76] [process=3][thread=async_save] Background save thread started. Deadline for this save operation is 2026-04-22 22:10:14.444390 I0422 21:50:14.447124 133252648376128 checkpoint_manager.py:1560] [process=3][thread=MainThread][step=9] Starting CheckpointManager Save Finalize thread=save_finalize I0422 21:50:14.447391 133126491719424 async_checkpointer.py:280] [process=3][thread=save_finalize] Waiting for background save thread=async_save. I0422 21:50:14.447552 133252648376128 standard_logger.py:34] {'step': 9, 'event_type': 'save', 'directory': 'gs://lance-maxtext/linen_ckpt_xpk_test_pipeline_scan_nnx_20260422_212603/linen_xpk_test_pipeline_scan_nnx_20260422_212603_08_checkpoint_async_true/checkpoints', 'reached_preemption': False, 'preemption_received_at': None, 'synchronous': False, 'wait_for_prev_start_time': 1776894587.095483, 'wait_for_prev_duration_secs': 26.60497260093689, 'time_between_consecutive_saves_sec': 23.367213249206543, 'checkpointer_blocking_start_time': 1776894613.7020488, 'checkpointer_blocking_duration_secs': 0.7425336837768555, 'get_old_steps_start_time': 1776894614.444601, 'get_old_steps_duration_secs': 2.5987625122070312e-05, 'checkpoint_manager_blocking_start_time': 1776894587.0935545, 'checkpoint_manager_blocking_duration_secs': 27.35396695137024} I0422 21:50:14.447663 133252648376128 checkpointing.py:408] Started an asynchronous checkpoint save for step 9 I0422 21:50:14.447707 133252648376128 checkpoint_manager.py:2020] [process=3][thread=MainThread][step=9][wait_until_finished] Waiting for Save Finalize thread (save_finalize) to complete. I0422 21:50:19.628158 133124893685504 array_metadata_store.py:203] [process=3][thread=array_type_handler] Wrote 39 array_metadata.ArrayMetadata to gs://lance-maxtext/linen_ckpt_xpk_test_pipeline_scan_nnx_20260422_212603/linen_xpk_test_pipeline_scan_nnx_20260422_212603_08_checkpoint_async_true/checkpoints/9/items/array_metadatas/process_3 I0422 21:50:55.725586 133125440456448 base_pytree_checkpoint_handler.py:130] [process=3] /jax/orbax/write/gbytes_per_sec: 38.225 MiB/s (total gbytes: 1.5 GiB) (time elapsed: 41.32392334938049 s) (per-host) I0422 21:50:55.725707 133125440456448 async_checkpointer.py:90] [process=3][thread=async_save] 3 Handler Commit operations completed. Time taken: 41.281161s. I0422 21:51:04.511404 133125440456448 async_checkpointer.py:160] [process=3][thread=async_save] Background save thread done. Time taken: 50.066842s. I0422 21:51:04.511694 133126491719424 async_checkpointer.py:288] [process=3][thread=save_finalize] Done with waiting for background save thread=async_save. I0422 21:51:04.511825 133126491719424 async_checkpointer.py:298] [process=3][thread=save_finalize] No errors found in background save thread=async_save. I0422 21:51:04.511867 133126491719424 checkpoint_manager.py:2137] [process=3][thread=save_finalize][step=9] CheckpointManager Save Finalize is syncing with other hosts... I0422 21:51:04.514121 133126491719424 checkpoint_manager.py:2146] [process=3][thread=save_finalize][step=9] CheckpointManager Save Finalize is done on all hosts. I0422 21:51:04.514302 133252648376128 checkpoint_manager.py:2032] [process=3][thread=MainThread][step=9][wait_until_finished] Done waiting for Save Finalize thread (save_finalize) running at step=9. I0422 21:51:04.514448 133252648376128 checkpoint_manager.py:2009] [process=3][thread=MainThread][wait_until_finished] No Save Finalize thread to wait for. Returning. I0422 21:51:04.515408 133252648376128 metric_logger.py:196] completed step: 9, seconds: 0.147, TFLOP/s/device: 92.355, Tokens/s/device: 13920.704, 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: Wed Apr 22 21:51:13 UTC 2026 EXIT_CODE=0