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 | 91.983 | 92.367 | +0.384 |
| Tok/s | 13864.7 | 13922.5 | +57.778 |
| Avg s/step | 7.822 | 8.030 | +0.208 |
| Memory % | 1.38 | 1.38 | 0 |
| JAX | 0.9.2 | 0.9.2 | — |
Diff = branch value − main value. Green = branch improved. Red = branch regressed.
XPK Start: Wed Apr 22 08:13:50 UTC 2026 PyTorch was not found. Models won't be available and only tokenizers, configuration and file/data utilities can be used. `rope_parameters`'s factor field must be a float >= 1, got 40 `rope_parameters`'s beta_fast field must be a float, got 32 `rope_parameters`'s beta_slow field must be a float, got 1 DeepseekV32Config got `key=rope_scaling` in kwargs but hasn't set it as attribute. For RoPE standardization you need to set `self.rope_parameters` in model's config. 2026-04-22 08:14:15.082044: E external/local_xla/xla/stream_executor/cuda/cuda_platform.cc:51] failed call to cuInit: INTERNAL: CUDA error: Failed call to cuInit: UNKNOWN ERROR (303) I0422 08:14:15.294312 132569523226432 max_utils.py:273] Attempting to initialize the jax distributed system... I0422 08:14:24.336060 132569523226432 distributed.py:149] Starting JAX distributed service on [::]:8482 I0422 08:14:24.338380 132569523226432 distributed.py:172] Connecting to JAX distributed service on mt-15-ocdbt-true-sv-cklh4-slice-job-0-0.mt-15-ocdbt-true-sv-cklh4:8482 I0422 08:14:25.613103 132569523226432 max_utils.py:284] Jax distributed system initialized! I0422 08:14:31.705891 132569523226432 max_utils.py:800] System Information: Jax Version: 0.9.2 I0422 08:14:31.705998 132569523226432 max_utils.py:801] System Information: Jaxlib Version: 0.9.2 I0422 08:14:31.706037 132569523226432 max_utils.py:802] System Information: Jax Backend: PJRT C API TFRT TPU v6 lite Built on Mar 4 2026 11:32:08 (1772652728) cl/878335365 I0422 08:14:31.706072 132569523226432 train_utils.py:361] WARNING: Sequence packing is essentially ignored for synthetic data. Please use a real dataset to use sequence packing. I0422 08:14:32.391656 132569523226432 maxtext_utils.py:1565] Num_devices: 32, shape (1, 1, 1, 32, 1, 1, 1, 1, 1, 1, 1, 1, 1) I0422 08:14:32.391950 132569523226432 checkpointing.py:677] Setting up checkpoint logger... I0422 08:14:32.392006 132569523226432 checkpointing.py:233] Creating checkpoint manager with ocdbt=True and zarr3=True I0422 08:14:32.392050 132569523226432 pytree_checkpoint_handler.py:592] save_device_host_concurrent_bytes=None I0422 08:14:32.392400 132569523226432 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 0x7891ab5181d0>, enable_pinned_host_transfer=False, save_concurrent_bytes: 96000000000 (89.4 GiB), restore_concurrent_bytes: 96000000000 (89.4 GiB) I0422 08:14:35.345453 132569523226432 checkpointing.py:265] Enabling policy for fixed interval checkpointing. I0422 08:14:35.345715 132569523226432 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 0x787c78681970>}, handler_registry=None I0422 08:14:35.345954 132569523226432 composite_checkpoint_handler.py:237] Deferred registration for item: "items". Adding handler `<orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeCheckpointHandler object at 0x787c78681970>` for item "items" and save args `<class 'orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeSaveArgs'>` and restore args `<class 'orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeRestoreArgs'>` to `_handler_registry`. I0422 08:14:35.346003 132569523226432 composite_checkpoint_handler.py:237] Deferred registration for item: "metrics". Adding handler `<orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonCheckpointHandler object at 0x787c786851c0>` for item "metrics" and save args `<class 'orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonSaveArgs'>` and restore args `<class 'orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonRestoreArgs'>` to `_handler_registry`. I0422 08:14:35.346039 132569523226432 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 0x787c78681970>, ('items', <class 'orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeRestoreArgs'>): <orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeCheckpointHandler object at 0x787c78681970>, ('metrics', <class 'orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonSaveArgs'>): <orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonCheckpointHandler object at 0x787c786851c0>, ('metrics', <class 'orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonRestoreArgs'>): <orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonCheckpointHandler object at 0x787c786851c0>}). I0422 08:14:35.346362 132569523226432 abstract_checkpointer.py:35] orbax-checkpoint version: 0.11.34 I0422 08:14:35.346434 132569523226432 async_checkpointer.py:192] [process=4][thread=MainThread] Using barrier_sync_fn: <function get_barrier_sync_fn.<locals>._fn at 0x787c4c205580> timeout: 1200 secs and primary_host=0 for async checkpoint writes I0422 08:14:37.764852 132569523226432 checkpoint_manager.py:1812] Found 0 checkpoint steps in gs://lance-maxtext/linen_ckpt_xpk_main_20260422_071422/linen_xpk_main_20260422_071422_15_ocdbt_true/checkpoints I0422 08:14:38.202899 132569523226432 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_20260422_071422/linen_xpk_main_20260422_071422_15_ocdbt_true/checkpoints: <orbax.checkpoint.checkpoint_manager.CheckpointManager object at 0x787c78682480> I0422 08:14:38.203075 132569523226432 checkpointing.py:301] Checkpoint manager created! I0422 08:14:39.126626 132569523226432 nnx_wrappers.py:437] Unknown Logical: bfloat16[32,2048,2048]...................................... ('activation_batch', 'activation_norm_length', 'activation_embed'). I0422 08:14:39.126760 132569523226432 nnx_wrappers.py:437] Unknown Physical: bfloat16[32,2048,2048]...................................... ('fsdp', None, None). I0422 08:14:39.505791 132569523226432 attentions.py:1088] attentions/inputs_q Logical: bfloat16[32,2048,2048]...................................... ('activation_batch', 'activation_attn_length', 'activation_attn_embed'). I0422 08:14:39.505883 132569523226432 attentions.py:1088] attentions/inputs_q Physical: bfloat16[32,2048,2048]...................................... ('fsdp', None, None). I0422 08:14:39.522366 132569523226432 attentions.py:1089] attentions/inputs_kv Logical: bfloat16[32,2048,2048]...................................... ('activation_batch', 'activation_attn_length', 'activation_attn_embed'). I0422 08:14:39.522422 132569523226432 attentions.py:1089] attentions/inputs_kv Physical: bfloat16[32,2048,2048]...................................... ('fsdp', None, None). I0422 08:14:39.546438 132569523226432 attentions.py:1154] attentions/query Logical: bfloat16[32,2048,16,128].................................... ('activation_kv_batch', 'activation_attn_length', 'activation_kv_heads', 'activation_kv_head_dim'). I0422 08:14:39.546510 132569523226432 attentions.py:1154] attentions/query Physical: bfloat16[32,2048,16,128].................................... ('fsdp', None, None, None). I0422 08:14:39.563019 132569523226432 attentions.py:1155] attentions/key Logical: bfloat16[32,2048,16,128].................................... ('activation_kv_batch', 'activation_attn_length', 'activation_kv_heads', 'activation_kv_head_dim'). I0422 08:14:39.563081 132569523226432 attentions.py:1155] attentions/key Physical: bfloat16[32,2048,16,128].................................... ('fsdp', None, None, None). I0422 08:14:39.579586 132569523226432 attentions.py:1156] attentions/value Logical: bfloat16[32,2048,16,128].................................... ('activation_kv_batch', 'activation_attn_length', 'activation_kv_heads', 'activation_kv_head_dim'). I0422 08:14:39.579644 132569523226432 attentions.py:1156] attentions/value Physical: bfloat16[32,2048,16,128].................................... ('fsdp', None, None, None). I0422 08:14:39.604299 132569523226432 attentions.py:1198] attentions/out Logical: bfloat16[32,2048,16,128].................................... ('activation_batch', 'activation_attn_length', 'activation_heads', 'activation_kv'). I0422 08:14:39.604377 132569523226432 attentions.py:1198] attentions/out Physical: bfloat16[32,2048,16,128].................................... ('fsdp', None, None, None). I0422 08:14:39.625468 132569523226432 linears.py:525] linears/x Logical: bfloat16[32,2048,7168]...................................... ('activation_batch', 'activation_length', 'activation_mlp'). I0422 08:14:39.625537 132569523226432 linears.py:525] linears/x Physical: bfloat16[32,2048,7168]...................................... ('fsdp', None, None). I0422 08:14:39.834147 132569523226432 checkpointing.py:577] checkpoint manager exists so trying to load this run's existing checkpoint I0422 08:14:39.834255 132569523226432 checkpointing.py:665] No existing checkpoints found, not restoring checkpoint. fsdp: 32 I0422 08:14:41.256609 132569523226432 maxtext_utils.py:1668] params/params/decoder/decoder_norm/scale Shape: float32[2048] Logical: P('norm',) Physical: (None,) I0422 08:14:41.256749 132569523226432 maxtext_utils.py:1668] params/params/decoder/layers/mlp/wi_0/kernel Shape: float32[2048,16,7168] Logical: P('embed', 'layers', 'mlp') Physical: ('fsdp', None, None) I0422 08:14:41.256808 132569523226432 maxtext_utils.py:1668] params/params/decoder/layers/mlp/wi_1/kernel Shape: float32[2048,16,7168] Logical: P('embed', 'layers', 'mlp') Physical: ('fsdp', None, None) I0422 08:14:41.256868 132569523226432 maxtext_utils.py:1668] params/params/decoder/layers/mlp/wo/kernel Shape: float32[7168,16,2048] Logical: P('mlp', 'layers', 'embed') Physical: (None, None, 'fsdp') I0422 08:14:41.256919 132569523226432 maxtext_utils.py:1668] params/params/decoder/layers/post_self_attention_layer_norm/scale Shape: float32[2048,16] Logical: P('norm', 'layers') Physical: (None, None) I0422 08:14:41.256956 132569523226432 maxtext_utils.py:1668] params/params/decoder/layers/pre_self_attention_layer_norm/scale Shape: float32[2048,16] Logical: P('norm', 'layers') Physical: (None, None) I0422 08:14:41.257016 132569523226432 maxtext_utils.py:1668] params/params/decoder/layers/self_attention/key/kernel Shape: float32[2048,16,16,128] Logical: P('embed', 'layers', 'kv_heads', 'kv_head_dim') Physical: ('fsdp', None, None, None) I0422 08:14:41.257095 132569523226432 maxtext_utils.py:1668] params/params/decoder/layers/self_attention/out/kernel Shape: float32[16,16,128,2048] Logical: P('heads', 'layers', 'kv', 'embed') Physical: (None, None, None, 'fsdp') I0422 08:14:41.257158 132569523226432 maxtext_utils.py:1668] params/params/decoder/layers/self_attention/query/kernel Shape: float32[2048,16,16,128] Logical: P('embed', 'layers', 'q_heads', 'kv') Physical: ('fsdp', None, None, None) I0422 08:14:41.257220 132569523226432 maxtext_utils.py:1668] params/params/decoder/layers/self_attention/value/kernel Shape: float32[2048,16,16,128] Logical: P('embed', 'layers', 'kv_heads', 'kv_head_dim') Physical: ('fsdp', None, None, None) I0422 08:14:41.257304 132569523226432 maxtext_utils.py:1668] params/params/decoder/logits_dense/kernel Shape: float32[2048,32000] Logical: P('embed_vocab', 'vocab') Physical: ('fsdp', None) I0422 08:14:41.257380 132569523226432 maxtext_utils.py:1668] params/params/token_embedder/embedding Shape: float32[32000,2048] Logical: P('vocab', 'embed_vocab') Physical: (None, 'fsdp') I0422 08:14:41.756035 132569523226432 train.py:155] train/xent Logical: float32[32,2048]............................................ ('activation_embed_and_logits_batch', 'activation_length'). I0422 08:14:41.756132 132569523226432 train.py:155] train/xent Physical: float32[32,2048]............................................ ('fsdp', None). I0422 08:14:41.771609 132569523226432 train.py:162] train/z_loss Logical: float32[32,2048]............................................ ('activation_embed_and_logits_batch', 'activation_length'). I0422 08:14:41.771671 132569523226432 train.py:162] train/z_loss Physical: float32[32,2048]............................................ ('fsdp', None). I0422 08:14:52.758003 132569523226432 max_utils.py:791] Total memory size: 1.5 GB, Output size: 0.4 GB, Temp size: 1.1 GB, Argument size: 0.4 GB, Host temp size: 0.0 GB. I0422 08:14:52.758894 132569523226432 metric_logger.py:301] number parameters: 1.104 billion I0422 08:15:04.020994 132569523226432 checkpointing.py:772] Waiting for step 0 to finish before checkpoint... I0422 08:15:04.251848 132569523226432 checkpointing.py:776] Waited 0.23083877563476562 seconds for step 0 to finish before starting checkpointing. I0422 08:15:04.254736 132569523226432 checkpoint_manager.py:2009] [process=4][thread=MainThread][wait_until_finished] No Save Finalize thread to wait for. Returning. I0422 08:15:04.256237 132569523226432 checkpoint_manager.py:1512] [process=4] Saving checkpoint at step 0 I0422 08:15:04.257890 132569523226432 event_tracking.py:70] [process=4] [async] Started save checkpoint @ gs://lance-maxtext/linen_ckpt_xpk_main_20260422_071422/linen_xpk_main_20260422_071422_15_ocdbt_true/checkpoints/0. I0422 08:15:04.592621 132569523226432 signaling_client.py:364] Using JaxDistributedSignalingClient I0422 08:15:04.593497 132569523226432 jax_array_handlers.py:360] Scheduling D2H of 39 prioritized jax.Array. I0422 08:15:04.593555 132569523226432 replica_slices.py:424] Transferring arrays to host memory with options: use_replica_parallel=True, min_slice_bytes_for_replica_parallel=None, max_replicas_for_replica_parallel=None, enable_pinned_host_transfer=False I0422 08:15:04.869503 132569523226432 base_pytree_checkpoint_handler.py:154] [process=4][thread=MainThread] Initiated "orbax.checkpoint._src.serialization.jax_array_handlers.ArrayHandler".serialize. Time taken: 0.276955s I0422 08:15:04.869673 132569523226432 base_pytree_checkpoint_handler.py:130] [process=4] /jax/orbax/write/blocking_gbytes_per_sec: 5.458 GiB/s (total gbytes: 1.5 GiB) (time elapsed: 0.2826213836669922 s) (per-host) I0422 08:15:04.869735 132569523226432 base_pytree_checkpoint_handler.py:768] [process=4][thread=MainThread] Initiated Pytree async_save. Time taken: 0.282694s (batch_requests_ready=0.002494s, total_serialization_initiated=0.280119s, others=0.000081s) I0422 08:15:04.869829 132569523226432 composite_checkpoint_handler.py:715] [process=4][thread=MainThread] Initiated CompositeCheckpointHandler.async_save. Time taken: 0.286681s (all_items=0.000019s, per_item={'items': '0.00001860'}, temp_paths=0.286662) I0422 08:15:04.870643 132569523226432 event_tracking.py:125] [process=4] [async] Finished blocking save in 0.61 seconds. Continuing save @ gs://lance-maxtext/linen_ckpt_xpk_main_20260422_071422/linen_xpk_main_20260422_071422_15_ocdbt_true/checkpoints/0. I0422 08:15:04.870929 132441188579072 async_checkpointer.py:76] [process=4][thread=async_save] Background save thread started. Deadline for this save operation is 2026-04-22 08:35:04.870897 I0422 08:15:04.880644 132569523226432 checkpoint_manager.py:1560] [process=4][thread=MainThread][step=0] Starting CheckpointManager Save Finalize thread=save_finalize I0422 08:15:04.880964 132440632252160 async_checkpointer.py:280] [process=4][thread=save_finalize] Waiting for background save thread=async_save. I0422 08:15:04.881130 132569523226432 standard_logger.py:34] {'step': 0, 'event_type': 'save', 'directory': 'gs://lance-maxtext/linen_ckpt_xpk_main_20260422_071422/linen_xpk_main_20260422_071422_15_ocdbt_true/checkpoints', 'reached_preemption': False, 'preemption_received_at': None, 'synchronous': False, 'wait_for_prev_start_time': 1776845704.2547169, 'wait_for_prev_duration_secs': 7.176399230957031e-05, 'time_between_consecutive_saves_sec': None, 'checkpointer_blocking_start_time': 1776845704.256277, 'checkpointer_blocking_duration_secs': 0.6148061752319336, 'get_old_steps_start_time': 1776845704.8711076, 'get_old_steps_duration_secs': 3.695487976074219e-05, 'checkpoint_manager_blocking_start_time': 1776845704.2524104, 'checkpoint_manager_blocking_duration_secs': 0.6286730766296387} I0422 08:15:04.881235 132569523226432 checkpointing.py:408] Started an asynchronous checkpoint save for step 0 I0422 08:15:04.881286 132569523226432 max_utils.py:750] Memstats: After params initialized: I0422 08:15:04.881336 132569523226432 max_utils.py:756] Using (GB) 0.43 / 31.25 (1.376000%) on TPU_16(process=4,(0,4,0,0)) I0422 08:15:04.881369 132569523226432 max_utils.py:756] Using (GB) 0.43 / 31.25 (1.376000%) on TPU_17(process=4,(1,4,0,0)) I0422 08:15:04.881396 132569523226432 max_utils.py:756] Using (GB) 0.43 / 31.25 (1.376000%) on TPU_20(process=4,(0,5,0,0)) I0422 08:15:04.881422 132569523226432 max_utils.py:756] Using (GB) 0.43 / 31.25 (1.376000%) on TPU_21(process=4,(1,5,0,0)) I0422 08:15:05.193351 132569523226432 metric_logger.py:196] completed step: 0, seconds: 11.262, TFLOP/s/device: 1.206, Tokens/s/device: 181.851, total_weights: 65536, loss: 10.874, lm_loss: 10.874, perplexity: 52779.875 I0422 08:15:05.384293 132569523226432 metric_logger.py:196] completed step: 1, seconds: 1.171, TFLOP/s/device: 11.605, Tokens/s/device: 1749.239, total_weights: 65536, loss: 10.874, lm_loss: 10.874, perplexity: 52779.875 I0422 08:15:05.793337 132569523226432 metric_logger.py:196] completed step: 2, seconds: 0.044, TFLOP/s/device: 310.712, Tokens/s/device: 46833.909, total_weights: 65536, loss: 10.560, lm_loss: 10.560, perplexity: 38547.719 I0422 08:15:05.940659 132569523226432 metric_logger.py:196] completed step: 3, seconds: 0.409, TFLOP/s/device: 33.196, Tokens/s/device: 5003.689, total_weights: 65536, loss: 9.980, lm_loss: 9.980, perplexity: 21590.506 I0422 08:15:05.945118 132569523226432 checkpointing.py:772] Waiting for step 5 to finish before checkpoint... I0422 08:15:06.234772 132569523226432 checkpointing.py:776] Waited 0.28963327407836914 seconds for step 5 to finish before starting checkpointing. I0422 08:15:06.237812 132569523226432 checkpoint_manager.py:2020] [process=4][thread=MainThread][step=0][wait_until_finished] Waiting for Save Finalize thread (save_finalize) to complete. I0422 08:15:08.257106 2798 google_auth_provider.cc:181] Running on GCE, using service account 562977990677-compute@developer.gserviceaccount.com I0422 08:15:10.668636 132440640644864 array_metadata_store.py:203] [process=4][thread=array_type_handler] Wrote 39 array_metadata.ArrayMetadata to gs://lance-maxtext/linen_ckpt_xpk_main_20260422_071422/linen_xpk_main_20260422_071422_15_ocdbt_true/checkpoints/0/items/array_metadatas/process_4 I0422 08:15:42.184009 132441188579072 base_pytree_checkpoint_handler.py:130] [process=4] /jax/orbax/write/gbytes_per_sec: 42.014 MiB/s (total gbytes: 1.5 GiB) (time elapsed: 37.596923828125 s) (per-host) I0422 08:15:42.184117 132441188579072 async_checkpointer.py:90] [process=4][thread=async_save] 3 Handler Commit operations completed. Time taken: 37.313077s. I0422 08:15:50.907137 132441188579072 async_checkpointer.py:160] [process=4][thread=async_save] Background save thread done. Time taken: 46.036081s. I0422 08:15:50.907412 132440632252160 async_checkpointer.py:288] [process=4][thread=save_finalize] Done with waiting for background save thread=async_save. I0422 08:15:50.907532 132440632252160 async_checkpointer.py:298] [process=4][thread=save_finalize] No errors found in background save thread=async_save. I0422 08:15:50.907582 132440632252160 checkpoint_manager.py:2137] [process=4][thread=save_finalize][step=0] CheckpointManager Save Finalize is syncing with other hosts... I0422 08:15:50.909871 132440632252160 checkpoint_manager.py:2146] [process=4][thread=save_finalize][step=0] CheckpointManager Save Finalize is done on all hosts. I0422 08:15:50.910034 132569523226432 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. W0422 08:15:50.910168 132569523226432 checkpoint_manager.py:1452] Waiting for previous save to complete took 44.672356 seconds. If this number is high, consider checkpointing less frequently. I0422 08:15:50.912082 132569523226432 checkpoint_manager.py:1512] [process=4] Saving checkpoint at step 5 I0422 08:15:50.914484 132569523226432 event_tracking.py:70] [process=4] [async] Started save checkpoint @ gs://lance-maxtext/linen_ckpt_xpk_main_20260422_071422/linen_xpk_main_20260422_071422_15_ocdbt_true/checkpoints/5. I0422 08:15:51.238445 132569523226432 jax_array_handlers.py:360] Scheduling D2H of 39 prioritized jax.Array. I0422 08:15:51.238542 132569523226432 replica_slices.py:424] Transferring arrays to host memory with options: use_replica_parallel=True, min_slice_bytes_for_replica_parallel=None, max_replicas_for_replica_parallel=None, enable_pinned_host_transfer=False I0422 08:15:51.272052 132569523226432 base_pytree_checkpoint_handler.py:154] [process=4][thread=MainThread] Initiated "orbax.checkpoint._src.serialization.jax_array_handlers.ArrayHandler".serialize. Time taken: 0.034768s I0422 08:15:51.272205 132569523226432 base_pytree_checkpoint_handler.py:130] [process=4] /jax/orbax/write/blocking_gbytes_per_sec: 40.420 GiB/s (total gbytes: 1.5 GiB) (time elapsed: 0.03816366195678711 s) (per-host) I0422 08:15:51.272257 132569523226432 base_pytree_checkpoint_handler.py:768] [process=4][thread=MainThread] Initiated Pytree async_save. Time taken: 0.038225s (batch_requests_ready=0.001779s, total_serialization_initiated=0.036377s, others=0.000069s) I0422 08:15:51.272340 132569523226432 composite_checkpoint_handler.py:715] [process=4][thread=MainThread] Initiated CompositeCheckpointHandler.async_save. Time taken: 0.043179s (all_items=0.000017s, per_item={'items': '0.00001669'}, temp_paths=0.043162) I0422 08:15:51.272994 132569523226432 event_tracking.py:125] [process=4] [async] Finished blocking save in 0.36 seconds. Continuing save @ gs://lance-maxtext/linen_ckpt_xpk_main_20260422_071422/linen_xpk_main_20260422_071422_15_ocdbt_true/checkpoints/5. I0422 08:15:51.273319 132440632252160 async_checkpointer.py:76] [process=4][thread=async_save] Background save thread started. Deadline for this save operation is 2026-04-22 08:35:51.273282 I0422 08:15:51.281899 132569523226432 checkpoint_manager.py:1560] [process=4][thread=MainThread][step=5] Starting CheckpointManager Save Finalize thread=save_finalize I0422 08:15:51.282144 132440137770752 async_checkpointer.py:280] [process=4][thread=save_finalize] Waiting for background save thread=async_save. I0422 08:15:51.282276 132569523226432 standard_logger.py:34] {'step': 5, 'event_type': 'save', 'directory': 'gs://lance-maxtext/linen_ckpt_xpk_main_20260422_071422/linen_xpk_main_20260422_071422_15_ocdbt_true/checkpoints', 'reached_preemption': False, 'preemption_received_at': None, 'synchronous': False, 'wait_for_prev_start_time': 1776845706.2377841, 'wait_for_prev_duration_secs': 44.67235589027405, 'time_between_consecutive_saves_sec': None, 'checkpointer_blocking_start_time': 1776845750.9121265, 'checkpointer_blocking_duration_secs': 0.36133623123168945, 'get_old_steps_start_time': 1776845751.2734847, 'get_old_steps_duration_secs': 3.123283386230469e-05, 'checkpoint_manager_blocking_start_time': 1776845706.2353773, 'checkpoint_manager_blocking_duration_secs': 45.0468635559082} I0422 08:15:51.282384 132569523226432 checkpointing.py:408] Started an asynchronous checkpoint save for step 5 I0422 08:15:51.283216 132569523226432 metric_logger.py:196] completed step: 4, seconds: 0.153, TFLOP/s/device: 88.838, Tokens/s/device: 13390.610, total_weights: 65536, loss: 9.460, lm_loss: 9.460, perplexity: 12840.355 I0422 08:15:51.289101 132569523226432 metric_logger.py:196] completed step: 5, seconds: 0.147, TFLOP/s/device: 92.274, Tokens/s/device: 13908.508, total_weights: 65536, loss: 8.959, lm_loss: 8.959, perplexity: 7776.467 I0422 08:15:57.420032 132440146163456 array_metadata_store.py:203] [process=4][thread=array_type_handler] Wrote 39 array_metadata.ArrayMetadata to gs://lance-maxtext/linen_ckpt_xpk_main_20260422_071422/linen_xpk_main_20260422_071422_15_ocdbt_true/checkpoints/5/items/array_metadatas/process_4 I0422 08:16:14.265588 132569523226432 metric_logger.py:196] completed step: 6, seconds: 45.343, TFLOP/s/device: 0.300, Tokens/s/device: 45.167, total_weights: 65536, loss: 8.469, lm_loss: 8.469, perplexity: 4762.740 I0422 08:16:14.413235 132569523226432 metric_logger.py:196] completed step: 7, seconds: 22.831, TFLOP/s/device: 0.595, Tokens/s/device: 89.704, total_weights: 65536, loss: 8.003, lm_loss: 8.003, perplexity: 2989.151 I0422 08:16:14.560460 132569523226432 metric_logger.py:196] completed step: 8, seconds: 0.152, TFLOP/s/device: 89.386, Tokens/s/device: 13473.241, total_weights: 65536, loss: 7.572, lm_loss: 7.572, perplexity: 1943.098 I0422 08:16:14.707252 132569523226432 checkpointing.py:772] Waiting for step 9 to finish before checkpoint... I0422 08:16:14.707952 132569523226432 checkpointing.py:776] Waited 0.000720977783203125 seconds for step 9 to finish before starting checkpointing. I0422 08:16:14.710095 132569523226432 checkpoint_manager.py:2020] [process=4][thread=MainThread][step=5][wait_until_finished] Waiting for Save Finalize thread (save_finalize) to complete. I0422 08:16:32.357051 132440632252160 base_pytree_checkpoint_handler.py:130] [process=4] /jax/orbax/write/gbytes_per_sec: 38.412 MiB/s (total gbytes: 1.5 GiB) (time elapsed: 41.12297320365906 s) (per-host) I0422 08:16:32.357159 132440632252160 async_checkpointer.py:90] [process=4][thread=async_save] 3 Handler Commit operations completed. Time taken: 41.083728s. I0422 08:16:39.642744 132440632252160 async_checkpointer.py:160] [process=4][thread=async_save] Background save thread done. Time taken: 48.369298s. I0422 08:16:39.643067 132440137770752 async_checkpointer.py:288] [process=4][thread=save_finalize] Done with waiting for background save thread=async_save. I0422 08:16:39.643182 132440137770752 async_checkpointer.py:298] [process=4][thread=save_finalize] No errors found in background save thread=async_save. I0422 08:16:39.643232 132440137770752 checkpoint_manager.py:2137] [process=4][thread=save_finalize][step=5] CheckpointManager Save Finalize is syncing with other hosts... I0422 08:16:39.644611 132440137770752 checkpoint_manager.py:2146] [process=4][thread=save_finalize][step=5] CheckpointManager Save Finalize is done on all hosts. I0422 08:16:39.644779 132569523226432 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. W0422 08:16:39.644912 132569523226432 checkpoint_manager.py:1452] Waiting for previous save to complete took 24.934825 seconds. If this number is high, consider checkpointing less frequently. I0422 08:16:39.646500 132569523226432 checkpoint_manager.py:1512] [process=4] Saving checkpoint at step 9 I0422 08:16:39.648623 132569523226432 event_tracking.py:70] [process=4] [async] Started save checkpoint @ gs://lance-maxtext/linen_ckpt_xpk_main_20260422_071422/linen_xpk_main_20260422_071422_15_ocdbt_true/checkpoints/9. I0422 08:16:39.937834 132569523226432 jax_array_handlers.py:360] Scheduling D2H of 39 prioritized jax.Array. I0422 08:16:39.937922 132569523226432 replica_slices.py:424] Transferring arrays to host memory with options: use_replica_parallel=True, min_slice_bytes_for_replica_parallel=None, max_replicas_for_replica_parallel=None, enable_pinned_host_transfer=False I0422 08:16:39.974632 132569523226432 base_pytree_checkpoint_handler.py:154] [process=4][thread=MainThread] Initiated "orbax.checkpoint._src.serialization.jax_array_handlers.ArrayHandler".serialize. Time taken: 0.037757s I0422 08:16:39.974785 132569523226432 base_pytree_checkpoint_handler.py:130] [process=4] /jax/orbax/write/blocking_gbytes_per_sec: 37.489 GiB/s (total gbytes: 1.5 GiB) (time elapsed: 0.04114723205566406 s) (per-host) I0422 08:16:39.974835 132569523226432 base_pytree_checkpoint_handler.py:768] [process=4][thread=MainThread] Initiated Pytree async_save. Time taken: 0.041207s (batch_requests_ready=0.001640s, total_serialization_initiated=0.039503s, others=0.000064s) I0422 08:16:39.974910 132569523226432 composite_checkpoint_handler.py:715] [process=4][thread=MainThread] Initiated CompositeCheckpointHandler.async_save. Time taken: 0.045083s (all_items=0.000010s, per_item={'items': '0.00001049'}, temp_paths=0.045073) I0422 08:16:39.975556 132569523226432 event_tracking.py:125] [process=4] [async] Finished blocking save in 0.33 seconds. Continuing save @ gs://lance-maxtext/linen_ckpt_xpk_main_20260422_071422/linen_xpk_main_20260422_071422_15_ocdbt_true/checkpoints/9. I0422 08:16:39.975906 132440137770752 async_checkpointer.py:76] [process=4][thread=async_save] Background save thread started. Deadline for this save operation is 2026-04-22 08:36:39.975867 I0422 08:16:39.980126 132569523226432 checkpoint_manager.py:1560] [process=4][thread=MainThread][step=9] Starting CheckpointManager Save Finalize thread=save_finalize I0422 08:16:39.980437 132445491939072 async_checkpointer.py:280] [process=4][thread=save_finalize] Waiting for background save thread=async_save. I0422 08:16:39.980596 132569523226432 standard_logger.py:34] {'step': 9, 'event_type': 'save', 'directory': 'gs://lance-maxtext/linen_ckpt_xpk_main_20260422_071422/linen_xpk_main_20260422_071422_15_ocdbt_true/checkpoints', 'reached_preemption': False, 'preemption_received_at': None, 'synchronous': False, 'wait_for_prev_start_time': 1776845774.7100651, 'wait_for_prev_duration_secs': 24.93482494354248, 'time_between_consecutive_saves_sec': 23.80015254020691, 'checkpointer_blocking_start_time': 1776845799.646539, 'checkpointer_blocking_duration_secs': 0.32952451705932617, 'get_old_steps_start_time': 1776845799.9760811, 'get_old_steps_duration_secs': 2.47955322265625e-05, 'checkpoint_manager_blocking_start_time': 1776845774.7081726, 'checkpoint_manager_blocking_duration_secs': 25.272390127182007} I0422 08:16:39.980750 132569523226432 checkpointing.py:408] Started an asynchronous checkpoint save for step 9 I0422 08:16:39.980799 132569523226432 checkpoint_manager.py:2020] [process=4][thread=MainThread][step=9][wait_until_finished] Waiting for Save Finalize thread (save_finalize) to complete. I0422 08:16:45.104814 132440146163456 array_metadata_store.py:203] [process=4][thread=array_type_handler] Wrote 39 array_metadata.ArrayMetadata to gs://lance-maxtext/linen_ckpt_xpk_main_20260422_071422/linen_xpk_main_20260422_071422_15_ocdbt_true/checkpoints/9/items/array_metadatas/process_4 I0422 08:17:21.066217 132440137770752 base_pytree_checkpoint_handler.py:130] [process=4] /jax/orbax/write/gbytes_per_sec: 38.403 MiB/s (total gbytes: 1.5 GiB) (time elapsed: 41.13253855705261 s) (per-host) I0422 08:17:21.066348 132440137770752 async_checkpointer.py:90] [process=4][thread=async_save] 3 Handler Commit operations completed. Time taken: 41.090324s. I0422 08:17:28.225474 132440137770752 async_checkpointer.py:160] [process=4][thread=async_save] Background save thread done. Time taken: 48.249436s. I0422 08:17:28.225791 132445491939072 async_checkpointer.py:288] [process=4][thread=save_finalize] Done with waiting for background save thread=async_save. I0422 08:17:28.225915 132445491939072 async_checkpointer.py:298] [process=4][thread=save_finalize] No errors found in background save thread=async_save. I0422 08:17:28.225958 132445491939072 checkpoint_manager.py:2137] [process=4][thread=save_finalize][step=9] CheckpointManager Save Finalize is syncing with other hosts... I0422 08:17:28.228725 132445491939072 checkpoint_manager.py:2146] [process=4][thread=save_finalize][step=9] CheckpointManager Save Finalize is done on all hosts. I0422 08:17:28.228896 132569523226432 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. I0422 08:17:28.229043 132569523226432 checkpoint_manager.py:2009] [process=4][thread=MainThread][wait_until_finished] No Save Finalize thread to wait for. Returning. I0422 08:17:28.230075 132569523226432 metric_logger.py:196] completed step: 9, seconds: 0.148, TFLOP/s/device: 91.983, Tokens/s/device: 13864.724, 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 08:17:40 UTC 2026 EXIT_CODE=0
XPK Start: Wed Apr 22 22:25:51 UTC 2026 PyTorch was not found. Models won't be available and only tokenizers, configuration and file/data utilities can be used. `rope_parameters`'s factor field must be a float >= 1, got 40 `rope_parameters`'s beta_fast field must be a float, got 32 `rope_parameters`'s beta_slow field must be a float, got 1 DeepseekV32Config got `key=rope_scaling` in kwargs but hasn't set it as attribute. For RoPE standardization you need to set `self.rope_parameters` in model's config. 2026-04-22 22:26:15.905174: E external/local_xla/xla/stream_executor/cuda/cuda_platform.cc:51] failed call to cuInit: INTERNAL: CUDA error: Failed call to cuInit: UNKNOWN ERROR (303) I0422 22:26:16.117696 139838915245888 max_utils.py:273] Attempting to initialize the jax distributed system... I0422 22:26:25.160089 139838915245888 distributed.py:149] Starting JAX distributed service on [::]:8482 I0422 22:26:25.162471 139838915245888 distributed.py:172] Connecting to JAX distributed service on mt-15-ocdbt-true-sv-nukyj-slice-job-0-0.mt-15-ocdbt-true-sv-nukyj:8482 I0422 22:26:26.168396 139838915245888 max_utils.py:284] Jax distributed system initialized! I0422 22:26:32.194521 139838915245888 max_utils.py:800] System Information: Jax Version: 0.9.2 I0422 22:26:32.194627 139838915245888 max_utils.py:801] System Information: Jaxlib Version: 0.9.2 I0422 22:26:32.194667 139838915245888 max_utils.py:802] System Information: Jax Backend: PJRT C API TFRT TPU v6 lite Built on Mar 4 2026 11:32:08 (1772652728) cl/878335365 I0422 22:26:32.194701 139838915245888 train_utils.py:361] WARNING: Sequence packing is essentially ignored for synthetic data. Please use a real dataset to use sequence packing. I0422 22:26:32.889224 139838915245888 maxtext_utils.py:1565] Num_devices: 32, shape (1, 1, 1, 32, 1, 1, 1, 1, 1, 1, 1, 1, 1) I0422 22:26:32.889507 139838915245888 checkpointing.py:677] Setting up checkpoint logger... I0422 22:26:32.889563 139838915245888 checkpointing.py:233] Creating checkpoint manager with ocdbt=True and zarr3=True I0422 22:26:32.889608 139838915245888 pytree_checkpoint_handler.py:592] save_device_host_concurrent_bytes=None I0422 22:26:32.889951 139838915245888 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 0x7f2e3251c3b0>, enable_pinned_host_transfer=False, save_concurrent_bytes: 96000000000 (89.4 GiB), restore_concurrent_bytes: 96000000000 (89.4 GiB) I0422 22:26:36.064233 139838915245888 checkpointing.py:265] Enabling policy for fixed interval checkpointing. I0422 22:26:36.064476 139838915245888 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 0x7f191c2b3860>}, handler_registry=None I0422 22:26:36.064713 139838915245888 composite_checkpoint_handler.py:237] Deferred registration for item: "items". Adding handler `<orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeCheckpointHandler object at 0x7f191c2b3860>` for item "items" and save args `<class 'orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeSaveArgs'>` and restore args `<class 'orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeRestoreArgs'>` to `_handler_registry`. I0422 22:26:36.064761 139838915245888 composite_checkpoint_handler.py:237] Deferred registration for item: "metrics". Adding handler `<orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonCheckpointHandler object at 0x7f191c2b4b90>` for item "metrics" and save args `<class 'orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonSaveArgs'>` and restore args `<class 'orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonRestoreArgs'>` to `_handler_registry`. I0422 22:26:36.064796 139838915245888 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 0x7f191c2b3860>, ('items', <class 'orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeRestoreArgs'>): <orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeCheckpointHandler object at 0x7f191c2b3860>, ('metrics', <class 'orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonSaveArgs'>): <orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonCheckpointHandler object at 0x7f191c2b4b90>, ('metrics', <class 'orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonRestoreArgs'>): <orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonCheckpointHandler object at 0x7f191c2b4b90>}). I0422 22:26:36.065126 139838915245888 abstract_checkpointer.py:35] orbax-checkpoint version: 0.11.34 I0422 22:26:36.065196 139838915245888 async_checkpointer.py:192] [process=6][thread=MainThread] Using barrier_sync_fn: <function get_barrier_sync_fn.<locals>._fn at 0x7f18f05432e0> timeout: 1200 secs and primary_host=0 for async checkpoint writes I0422 22:26:37.431017 139838915245888 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_15_ocdbt_true/checkpoints I0422 22:26:38.282362 139838915245888 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_test_pipeline_scan_nnx_20260422_212603/linen_xpk_test_pipeline_scan_nnx_20260422_212603_15_ocdbt_true/checkpoints: <orbax.checkpoint.checkpoint_manager.CheckpointManager object at 0x7f191c2b37d0> I0422 22:26:38.282538 139838915245888 checkpointing.py:301] Checkpoint manager created! I0422 22:26:39.240654 139838915245888 nnx_wrappers.py:453] Unknown Logical: bfloat16[32,2048,2048]...................................... ('activation_batch', 'activation_norm_length', 'activation_embed'). I0422 22:26:39.240774 139838915245888 nnx_wrappers.py:453] Unknown Physical: bfloat16[32,2048,2048]...................................... ('fsdp', None, None). I0422 22:26:39.619777 139838915245888 attentions.py:1088] attentions/inputs_q Logical: bfloat16[32,2048,2048]...................................... ('activation_batch', 'activation_attn_length', 'activation_attn_embed'). I0422 22:26:39.619874 139838915245888 attentions.py:1088] attentions/inputs_q Physical: bfloat16[32,2048,2048]...................................... ('fsdp', None, None). I0422 22:26:39.636634 139838915245888 attentions.py:1089] attentions/inputs_kv Logical: bfloat16[32,2048,2048]...................................... ('activation_batch', 'activation_attn_length', 'activation_attn_embed'). I0422 22:26:39.636693 139838915245888 attentions.py:1089] attentions/inputs_kv Physical: bfloat16[32,2048,2048]...................................... ('fsdp', None, None). I0422 22:26:39.660734 139838915245888 attentions.py:1154] attentions/query Logical: bfloat16[32,2048,16,128].................................... ('activation_kv_batch', 'activation_attn_length', 'activation_kv_heads', 'activation_kv_head_dim'). I0422 22:26:39.660806 139838915245888 attentions.py:1154] attentions/query Physical: bfloat16[32,2048,16,128].................................... ('fsdp', None, None, None). I0422 22:26:39.678310 139838915245888 attentions.py:1155] attentions/key Logical: bfloat16[32,2048,16,128].................................... ('activation_kv_batch', 'activation_attn_length', 'activation_kv_heads', 'activation_kv_head_dim'). I0422 22:26:39.678373 139838915245888 attentions.py:1155] attentions/key Physical: bfloat16[32,2048,16,128].................................... ('fsdp', None, None, None). I0422 22:26:39.695143 139838915245888 attentions.py:1156] attentions/value Logical: bfloat16[32,2048,16,128].................................... ('activation_kv_batch', 'activation_attn_length', 'activation_kv_heads', 'activation_kv_head_dim'). I0422 22:26:39.695202 139838915245888 attentions.py:1156] attentions/value Physical: bfloat16[32,2048,16,128].................................... ('fsdp', None, None, None). I0422 22:26:39.720234 139838915245888 attentions.py:1198] attentions/out Logical: bfloat16[32,2048,16,128].................................... ('activation_batch', 'activation_attn_length', 'activation_heads', 'activation_kv'). I0422 22:26:39.720302 139838915245888 attentions.py:1198] attentions/out Physical: bfloat16[32,2048,16,128].................................... ('fsdp', None, None, None). I0422 22:26:39.741711 139838915245888 linears.py:525] linears/x Logical: bfloat16[32,2048,7168]...................................... ('activation_batch', 'activation_length', 'activation_mlp'). I0422 22:26:39.741784 139838915245888 linears.py:525] linears/x Physical: bfloat16[32,2048,7168]...................................... ('fsdp', None, None). I0422 22:26:39.958667 139838915245888 checkpointing.py:577] checkpoint manager exists so trying to load this run's existing checkpoint I0422 22:26:39.958779 139838915245888 checkpointing.py:665] No existing checkpoints found, not restoring checkpoint. fsdp: 32 I0422 22:26:41.394750 139838915245888 maxtext_utils.py:1668] params/params/decoder/decoder_norm/scale Shape: float32[2048] Logical: P('norm',) Physical: (None,) I0422 22:26:41.394876 139838915245888 maxtext_utils.py:1668] params/params/decoder/layers/mlp/wi_0/kernel Shape: float32[2048,16,7168] Logical: P('embed', 'layers', 'mlp') Physical: ('fsdp', None, None) I0422 22:26:41.394929 139838915245888 maxtext_utils.py:1668] params/params/decoder/layers/mlp/wi_1/kernel Shape: float32[2048,16,7168] Logical: P('embed', 'layers', 'mlp') Physical: ('fsdp', None, None) I0422 22:26:41.394990 139838915245888 maxtext_utils.py:1668] params/params/decoder/layers/mlp/wo/kernel Shape: float32[7168,16,2048] Logical: P('mlp', 'layers', 'embed') Physical: (None, None, 'fsdp') I0422 22:26:41.395043 139838915245888 maxtext_utils.py:1668] params/params/decoder/layers/post_self_attention_layer_norm/scale Shape: float32[2048,16] Logical: P('norm', 'layers') Physical: (None, None) I0422 22:26:41.395082 139838915245888 maxtext_utils.py:1668] params/params/decoder/layers/pre_self_attention_layer_norm/scale Shape: float32[2048,16] Logical: P('norm', 'layers') Physical: (None, None) I0422 22:26:41.395150 139838915245888 maxtext_utils.py:1668] params/params/decoder/layers/self_attention/key/kernel Shape: float32[2048,16,16,128] Logical: P('embed', 'layers', 'kv_heads', 'kv_head_dim') Physical: ('fsdp', None, None, None) I0422 22:26:41.395205 139838915245888 maxtext_utils.py:1668] params/params/decoder/layers/self_attention/out/kernel Shape: float32[16,16,128,2048] Logical: P('heads', 'layers', 'kv', 'embed') Physical: (None, None, None, 'fsdp') I0422 22:26:41.395245 139838915245888 maxtext_utils.py:1668] params/params/decoder/layers/self_attention/query/kernel Shape: float32[2048,16,16,128] Logical: P('embed', 'layers', 'q_heads', 'kv') Physical: ('fsdp', None, None, None) I0422 22:26:41.395280 139838915245888 maxtext_utils.py:1668] params/params/decoder/layers/self_attention/value/kernel Shape: float32[2048,16,16,128] Logical: P('embed', 'layers', 'kv_heads', 'kv_head_dim') Physical: ('fsdp', None, None, None) I0422 22:26:41.395326 139838915245888 maxtext_utils.py:1668] params/params/decoder/logits_dense/kernel Shape: float32[2048,32000] Logical: P('embed_vocab', 'vocab') Physical: ('fsdp', None) I0422 22:26:41.395380 139838915245888 maxtext_utils.py:1668] params/params/token_embedder/embedding Shape: float32[32000,2048] Logical: P('vocab', 'embed_vocab') Physical: (None, 'fsdp') I0422 22:26:41.897918 139838915245888 train.py:155] train/xent Logical: float32[32,2048]............................................ ('activation_embed_and_logits_batch', 'activation_length'). I0422 22:26:41.898013 139838915245888 train.py:155] train/xent Physical: float32[32,2048]............................................ ('fsdp', None). I0422 22:26:41.913866 139838915245888 train.py:162] train/z_loss Logical: float32[32,2048]............................................ ('activation_embed_and_logits_batch', 'activation_length'). I0422 22:26:41.913927 139838915245888 train.py:162] train/z_loss Physical: float32[32,2048]............................................ ('fsdp', None). I0422 22:26:52.604458 139838915245888 max_utils.py:791] Total memory size: 1.5 GB, Output size: 0.4 GB, Temp size: 1.1 GB, Argument size: 0.4 GB, Host temp size: 0.0 GB. I0422 22:26:52.605337 139838915245888 metric_logger.py:301] number parameters: 1.104 billion I0422 22:27:03.908230 139838915245888 checkpointing.py:772] Waiting for step 0 to finish before checkpoint... I0422 22:27:04.201745 139838915245888 checkpointing.py:776] Waited 0.29349422454833984 seconds for step 0 to finish before starting checkpointing. I0422 22:27:04.206050 139838915245888 checkpoint_manager.py:2009] [process=6][thread=MainThread][wait_until_finished] No Save Finalize thread to wait for. Returning. I0422 22:27:04.207565 139838915245888 checkpoint_manager.py:1512] [process=6] Saving checkpoint at step 0 I0422 22:27:04.208952 139838915245888 event_tracking.py:70] [process=6] [async] Started save checkpoint @ gs://lance-maxtext/linen_ckpt_xpk_test_pipeline_scan_nnx_20260422_212603/linen_xpk_test_pipeline_scan_nnx_20260422_212603_15_ocdbt_true/checkpoints/0. I0422 22:27:04.965674 139838915245888 signaling_client.py:364] Using JaxDistributedSignalingClient I0422 22:27:04.966690 139838915245888 jax_array_handlers.py:360] Scheduling D2H of 39 prioritized jax.Array. I0422 22:27:04.966747 139838915245888 replica_slices.py:424] Transferring arrays to host memory with options: use_replica_parallel=True, min_slice_bytes_for_replica_parallel=None, max_replicas_for_replica_parallel=None, enable_pinned_host_transfer=False I0422 22:27:05.245845 139838915245888 base_pytree_checkpoint_handler.py:154] [process=6][thread=MainThread] Initiated "orbax.checkpoint._src.serialization.jax_array_handlers.ArrayHandler".serialize. Time taken: 0.280247s I0422 22:27:05.246032 139838915245888 base_pytree_checkpoint_handler.py:130] [process=6] /jax/orbax/write/blocking_gbytes_per_sec: 5.395 GiB/s (total gbytes: 1.5 GiB) (time elapsed: 0.285952091217041 s) (per-host) I0422 22:27:05.246112 139838915245888 base_pytree_checkpoint_handler.py:768] [process=6][thread=MainThread] Initiated Pytree async_save. Time taken: 0.286030s (batch_requests_ready=0.002497s, total_serialization_initiated=0.283446s, others=0.000087s) I0422 22:27:05.246243 139838915245888 composite_checkpoint_handler.py:715] [process=6][thread=MainThread] Initiated CompositeCheckpointHandler.async_save. Time taken: 0.290197s (all_items=0.000017s, per_item={'items': '0.00001740'}, temp_paths=0.290179) I0422 22:27:05.247111 139838915245888 event_tracking.py:125] [process=6] [async] Finished blocking save in 1.04 seconds. Continuing save @ gs://lance-maxtext/linen_ckpt_xpk_test_pipeline_scan_nnx_20260422_212603/linen_xpk_test_pipeline_scan_nnx_20260422_212603_15_ocdbt_true/checkpoints/0. I0422 22:27:05.247451 139711030777600 async_checkpointer.py:76] [process=6][thread=async_save] Background save thread started. Deadline for this save operation is 2026-04-22 22:47:05.247413 I0422 22:27:05.253489 139838915245888 checkpoint_manager.py:1560] [process=6][thread=MainThread][step=0] Starting CheckpointManager Save Finalize thread=save_finalize I0422 22:27:05.253741 139709439272704 async_checkpointer.py:280] [process=6][thread=save_finalize] Waiting for background save thread=async_save. I0422 22:27:05.253875 139838915245888 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_15_ocdbt_true/checkpoints', 'reached_preemption': False, 'preemption_received_at': None, 'synchronous': False, 'wait_for_prev_start_time': 1776896824.2060313, 'wait_for_prev_duration_secs': 7.748603820800781e-05, 'time_between_consecutive_saves_sec': None, 'checkpointer_blocking_start_time': 1776896824.2076037, 'checkpointer_blocking_duration_secs': 1.0400214195251465, 'get_old_steps_start_time': 1776896825.2476535, 'get_old_steps_duration_secs': 3.218650817871094e-05, 'checkpoint_manager_blocking_start_time': 1776896824.202285, 'checkpoint_manager_blocking_duration_secs': 1.0515506267547607} I0422 22:27:05.253985 139838915245888 checkpointing.py:408] Started an asynchronous checkpoint save for step 0 I0422 22:27:05.254031 139838915245888 max_utils.py:750] Memstats: After params initialized: I0422 22:27:05.254078 139838915245888 max_utils.py:756] Using (GB) 0.43 / 31.25 (1.376000%) on TPU_24(process=6,(0,6,0,0)) I0422 22:27:05.254134 139838915245888 max_utils.py:756] Using (GB) 0.43 / 31.25 (1.376000%) on TPU_25(process=6,(1,6,0,0)) I0422 22:27:05.254165 139838915245888 max_utils.py:756] Using (GB) 0.43 / 31.25 (1.376000%) on TPU_28(process=6,(0,7,0,0)) I0422 22:27:05.254191 139838915245888 max_utils.py:756] Using (GB) 0.43 / 31.25 (1.376000%) on TPU_29(process=6,(1,7,0,0)) I0422 22:27:05.563866 139838915245888 metric_logger.py:196] completed step: 0, seconds: 11.303, TFLOP/s/device: 1.202, Tokens/s/device: 181.194, total_weights: 65536, loss: 10.874, lm_loss: 10.874, perplexity: 52779.875 I0422 22:27:05.737990 139838915245888 metric_logger.py:196] completed step: 1, seconds: 1.654, TFLOP/s/device: 8.214, Tokens/s/device: 1238.176, total_weights: 65536, loss: 10.874, lm_loss: 10.874, perplexity: 52779.875 I0422 22:27:06.146553 139838915245888 metric_logger.py:196] completed step: 2, seconds: 0.027, TFLOP/s/device: 499.325, Tokens/s/device: 75263.680, total_weights: 65536, loss: 10.560, lm_loss: 10.560, perplexity: 38547.719 I0422 22:27:06.293941 139838915245888 metric_logger.py:196] completed step: 3, seconds: 0.409, TFLOP/s/device: 33.243, Tokens/s/device: 5010.802, total_weights: 65536, loss: 9.980, lm_loss: 9.980, perplexity: 21590.506 I0422 22:27:06.298533 139838915245888 checkpointing.py:772] Waiting for step 5 to finish before checkpoint... I0422 22:27:06.588194 139838915245888 checkpointing.py:776] Waited 0.28962278366088867 seconds for step 5 to finish before starting checkpointing. I0422 22:27:06.591358 139838915245888 checkpoint_manager.py:2020] [process=6][thread=MainThread][step=0][wait_until_finished] Waiting for Save Finalize thread (save_finalize) to complete. I0422 22:27:08.922713 2800 google_auth_provider.cc:181] Running on GCE, using service account 562977990677-compute@developer.gserviceaccount.com I0422 22:27:12.212787 139709447665408 array_metadata_store.py:203] [process=6][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_15_ocdbt_true/checkpoints/0/items/array_metadatas/process_6 I0422 22:27:42.967146 139711030777600 base_pytree_checkpoint_handler.py:130] [process=6] /jax/orbax/write/gbytes_per_sec: 41.561 MiB/s (total gbytes: 1.5 GiB) (time elapsed: 38.00702476501465 s) (per-host) I0422 22:27:42.967254 139711030777600 async_checkpointer.py:90] [process=6][thread=async_save] 3 Handler Commit operations completed. Time taken: 37.719691s. I0422 22:27:51.946952 139711030777600 async_checkpointer.py:160] [process=6][thread=async_save] Background save thread done. Time taken: 46.699373s. I0422 22:27:51.947255 139709439272704 async_checkpointer.py:288] [process=6][thread=save_finalize] Done with waiting for background save thread=async_save. I0422 22:27:51.947373 139709439272704 async_checkpointer.py:298] [process=6][thread=save_finalize] No errors found in background save thread=async_save. I0422 22:27:51.947425 139709439272704 checkpoint_manager.py:2137] [process=6][thread=save_finalize][step=0] CheckpointManager Save Finalize is syncing with other hosts... I0422 22:27:51.949161 139709439272704 checkpoint_manager.py:2146] [process=6][thread=save_finalize][step=0] CheckpointManager Save Finalize is done on all hosts. I0422 22:27:51.949353 139838915245888 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 22:27:51.949482 139838915245888 checkpoint_manager.py:1452] Waiting for previous save to complete took 45.358126 seconds. If this number is high, consider checkpointing less frequently. I0422 22:27:51.951621 139838915245888 checkpoint_manager.py:1512] [process=6] Saving checkpoint at step 5 I0422 22:27:51.954184 139838915245888 event_tracking.py:70] [process=6] [async] Started save checkpoint @ gs://lance-maxtext/linen_ckpt_xpk_test_pipeline_scan_nnx_20260422_212603/linen_xpk_test_pipeline_scan_nnx_20260422_212603_15_ocdbt_true/checkpoints/5. I0422 22:27:52.666214 139838915245888 jax_array_handlers.py:360] Scheduling D2H of 39 prioritized jax.Array. I0422 22:27:52.666306 139838915245888 replica_slices.py:424] Transferring arrays to host memory with options: use_replica_parallel=True, min_slice_bytes_for_replica_parallel=None, max_replicas_for_replica_parallel=None, enable_pinned_host_transfer=False I0422 22:27:52.713318 139838915245888 base_pytree_checkpoint_handler.py:154] [process=6][thread=MainThread] Initiated "orbax.checkpoint._src.serialization.jax_array_handlers.ArrayHandler".serialize. Time taken: 0.048342s I0422 22:27:52.713494 139838915245888 base_pytree_checkpoint_handler.py:130] [process=6] /jax/orbax/write/blocking_gbytes_per_sec: 29.800 GiB/s (total gbytes: 1.5 GiB) (time elapsed: 0.05176424980163574 s) (per-host) I0422 22:27:52.713544 139838915245888 base_pytree_checkpoint_handler.py:768] [process=6][thread=MainThread] Initiated Pytree async_save. Time taken: 0.051823s (batch_requests_ready=0.001777s, total_serialization_initiated=0.049980s, others=0.000066s) I0422 22:27:52.713632 139838915245888 composite_checkpoint_handler.py:715] [process=6][thread=MainThread] Initiated CompositeCheckpointHandler.async_save. Time taken: 0.056846s (all_items=0.000015s, per_item={'items': '0.00001550'}, temp_paths=0.056830) I0422 22:27:52.714302 139838915245888 event_tracking.py:125] [process=6] [async] Finished blocking save in 0.76 seconds. Continuing save @ gs://lance-maxtext/linen_ckpt_xpk_test_pipeline_scan_nnx_20260422_212603/linen_xpk_test_pipeline_scan_nnx_20260422_212603_15_ocdbt_true/checkpoints/5. I0422 22:27:52.714641 139688333518592 async_checkpointer.py:76] [process=6][thread=async_save] Background save thread started. Deadline for this save operation is 2026-04-22 22:47:52.714606 I0422 22:27:52.716550 139838915245888 checkpoint_manager.py:1560] [process=6][thread=MainThread][step=5] Starting CheckpointManager Save Finalize thread=save_finalize I0422 22:27:52.716768 139709439272704 async_checkpointer.py:280] [process=6][thread=save_finalize] Waiting for background save thread=async_save. I0422 22:27:52.716894 139838915245888 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_15_ocdbt_true/checkpoints', 'reached_preemption': False, 'preemption_received_at': None, 'synchronous': False, 'wait_for_prev_start_time': 1776896826.5913286, 'wait_for_prev_duration_secs': 45.358126401901245, 'time_between_consecutive_saves_sec': None, 'checkpointer_blocking_start_time': 1776896871.9517035, 'checkpointer_blocking_duration_secs': 0.7630858421325684, 'get_old_steps_start_time': 1776896872.7148142, 'get_old_steps_duration_secs': 2.86102294921875e-05, 'checkpoint_manager_blocking_start_time': 1776896826.5888355, 'checkpoint_manager_blocking_duration_secs': 46.1280243396759} I0422 22:27:52.717004 139838915245888 checkpointing.py:408] Started an asynchronous checkpoint save for step 5 I0422 22:27:52.717863 139838915245888 metric_logger.py:196] completed step: 4, seconds: 0.153, TFLOP/s/device: 88.899, Tokens/s/device: 13399.809, total_weights: 65536, loss: 9.460, lm_loss: 9.460, perplexity: 12840.355 I0422 22:27:52.724379 139838915245888 metric_logger.py:196] completed step: 5, seconds: 0.147, TFLOP/s/device: 92.188, Tokens/s/device: 13895.580, total_weights: 65536, loss: 8.959, lm_loss: 8.959, perplexity: 7776.467 I0422 22:27:58.226069 139709447665408 array_metadata_store.py:203] [process=6][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_15_ocdbt_true/checkpoints/5/items/array_metadatas/process_6 I0422 22:28:16.023180 139838915245888 metric_logger.py:196] completed step: 6, seconds: 46.424, TFLOP/s/device: 0.293, Tokens/s/device: 44.115, total_weights: 65536, loss: 8.469, lm_loss: 8.469, perplexity: 4762.740 I0422 22:28:16.170671 139838915245888 metric_logger.py:196] completed step: 7, seconds: 23.153, TFLOP/s/device: 0.587, Tokens/s/device: 88.455, total_weights: 65536, loss: 8.003, lm_loss: 8.003, perplexity: 2989.151 I0422 22:28:16.317819 139838915245888 metric_logger.py:196] completed step: 8, seconds: 0.153, TFLOP/s/device: 89.035, Tokens/s/device: 13420.268, total_weights: 65536, loss: 7.572, lm_loss: 7.572, perplexity: 1943.098 I0422 22:28:16.464580 139838915245888 checkpointing.py:772] Waiting for step 9 to finish before checkpoint... I0422 22:28:16.465278 139838915245888 checkpointing.py:776] Waited 0.000720977783203125 seconds for step 9 to finish before starting checkpointing. I0422 22:28:16.467352 139838915245888 checkpoint_manager.py:2020] [process=6][thread=MainThread][step=5][wait_until_finished] Waiting for Save Finalize thread (save_finalize) to complete. I0422 22:28:35.460932 139688333518592 base_pytree_checkpoint_handler.py:130] [process=6] /jax/orbax/write/gbytes_per_sec: 36.907 MiB/s (total gbytes: 1.5 GiB) (time elapsed: 42.79916286468506 s) (per-host) I0422 22:28:35.461060 139688333518592 async_checkpointer.py:90] [process=6][thread=async_save] 3 Handler Commit operations completed. Time taken: 42.746302s. I0422 22:28:43.410493 139688333518592 async_checkpointer.py:160] [process=6][thread=async_save] Background save thread done. Time taken: 50.695718s. I0422 22:28:43.410746 139709439272704 async_checkpointer.py:288] [process=6][thread=save_finalize] Done with waiting for background save thread=async_save. I0422 22:28:43.410804 139709439272704 async_checkpointer.py:298] [process=6][thread=save_finalize] No errors found in background save thread=async_save. I0422 22:28:43.410857 139709439272704 checkpoint_manager.py:2137] [process=6][thread=save_finalize][step=5] CheckpointManager Save Finalize is syncing with other hosts... I0422 22:28:43.412412 139709439272704 checkpoint_manager.py:2146] [process=6][thread=save_finalize][step=5] CheckpointManager Save Finalize is done on all hosts. I0422 22:28:43.412609 139838915245888 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 22:28:43.412734 139838915245888 checkpoint_manager.py:1452] Waiting for previous save to complete took 26.945388 seconds. If this number is high, consider checkpointing less frequently. I0422 22:28:43.414334 139838915245888 checkpoint_manager.py:1512] [process=6] Saving checkpoint at step 9 I0422 22:28:43.416321 139838915245888 event_tracking.py:70] [process=6] [async] Started save checkpoint @ gs://lance-maxtext/linen_ckpt_xpk_test_pipeline_scan_nnx_20260422_212603/linen_xpk_test_pipeline_scan_nnx_20260422_212603_15_ocdbt_true/checkpoints/9. I0422 22:28:44.127625 139838915245888 jax_array_handlers.py:360] Scheduling D2H of 39 prioritized jax.Array. I0422 22:28:44.127739 139838915245888 replica_slices.py:424] Transferring arrays to host memory with options: use_replica_parallel=True, min_slice_bytes_for_replica_parallel=None, max_replicas_for_replica_parallel=None, enable_pinned_host_transfer=False I0422 22:28:44.173074 139838915245888 base_pytree_checkpoint_handler.py:154] [process=6][thread=MainThread] Initiated "orbax.checkpoint._src.serialization.jax_array_handlers.ArrayHandler".serialize. Time taken: 0.046474s I0422 22:28:44.173240 139838915245888 base_pytree_checkpoint_handler.py:130] [process=6] /jax/orbax/write/blocking_gbytes_per_sec: 30.909 GiB/s (total gbytes: 1.5 GiB) (time elapsed: 0.04990720748901367 s) (per-host) I0422 22:28:44.173289 139838915245888 base_pytree_checkpoint_handler.py:768] [process=6][thread=MainThread] Initiated Pytree async_save. Time taken: 0.049964s (batch_requests_ready=0.001647s, total_serialization_initiated=0.048255s, others=0.000062s) I0422 22:28:44.173376 139838915245888 composite_checkpoint_handler.py:715] [process=6][thread=MainThread] Initiated CompositeCheckpointHandler.async_save. Time taken: 0.053918s (all_items=0.000012s, per_item={'items': '0.00001192'}, temp_paths=0.053906) I0422 22:28:44.174141 139838915245888 event_tracking.py:125] [process=6] [async] Finished blocking save in 0.76 seconds. Continuing save @ gs://lance-maxtext/linen_ckpt_xpk_test_pipeline_scan_nnx_20260422_212603/linen_xpk_test_pipeline_scan_nnx_20260422_212603_15_ocdbt_true/checkpoints/9. I0422 22:28:44.174465 139709439272704 async_checkpointer.py:76] [process=6][thread=async_save] Background save thread started. Deadline for this save operation is 2026-04-22 22:48:44.174427 I0422 22:28:44.176434 139838915245888 checkpoint_manager.py:1560] [process=6][thread=MainThread][step=9] Starting CheckpointManager Save Finalize thread=save_finalize I0422 22:28:44.176692 139709447665408 async_checkpointer.py:280] [process=6][thread=save_finalize] Waiting for background save thread=async_save. I0422 22:28:44.176849 139838915245888 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_15_ocdbt_true/checkpoints', 'reached_preemption': False, 'preemption_received_at': None, 'synchronous': False, 'wait_for_prev_start_time': 1776896896.467324, 'wait_for_prev_duration_secs': 26.945387840270996, 'time_between_consecutive_saves_sec': 24.5181245803833, 'checkpointer_blocking_start_time': 1776896923.4143734, 'checkpointer_blocking_duration_secs': 0.7601957321166992, 'get_old_steps_start_time': 1776896924.1745868, 'get_old_steps_duration_secs': 2.5510787963867188e-05, 'checkpoint_manager_blocking_start_time': 1776896896.465514, 'checkpoint_manager_blocking_duration_secs': 27.711303234100342} I0422 22:28:44.176962 139838915245888 checkpointing.py:408] Started an asynchronous checkpoint save for step 9 I0422 22:28:44.177005 139838915245888 checkpoint_manager.py:2020] [process=6][thread=MainThread][step=9][wait_until_finished] Waiting for Save Finalize thread (save_finalize) to complete. I0422 22:28:49.253381 139708907738880 array_metadata_store.py:203] [process=6][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_15_ocdbt_true/checkpoints/9/items/array_metadatas/process_6 I0422 22:29:25.531209 139709439272704 base_pytree_checkpoint_handler.py:130] [process=6] /jax/orbax/write/gbytes_per_sec: 38.147 MiB/s (total gbytes: 1.5 GiB) (time elapsed: 41.407835960388184 s) (per-host) I0422 22:29:25.531332 139709439272704 async_checkpointer.py:90] [process=6][thread=async_save] 3 Handler Commit operations completed. Time taken: 41.356268s. I0422 22:29:35.252749 139709439272704 async_checkpointer.py:160] [process=6][thread=async_save] Background save thread done. Time taken: 51.077667s. I0422 22:29:35.253036 139709447665408 async_checkpointer.py:288] [process=6][thread=save_finalize] Done with waiting for background save thread=async_save. I0422 22:29:35.253169 139709447665408 async_checkpointer.py:298] [process=6][thread=save_finalize] No errors found in background save thread=async_save. I0422 22:29:35.253211 139709447665408 checkpoint_manager.py:2137] [process=6][thread=save_finalize][step=9] CheckpointManager Save Finalize is syncing with other hosts... I0422 22:29:35.255116 139709447665408 checkpoint_manager.py:2146] [process=6][thread=save_finalize][step=9] CheckpointManager Save Finalize is done on all hosts. I0422 22:29:35.255299 139838915245888 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 22:29:35.255453 139838915245888 checkpoint_manager.py:2009] [process=6][thread=MainThread][wait_until_finished] No Save Finalize thread to wait for. Returning. I0422 22:29:35.256335 139838915245888 metric_logger.py:196] completed step: 9, seconds: 0.147, TFLOP/s/device: 92.367, Tokens/s/device: 13922.502, 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 22:29:46 UTC 2026 EXIT_CODE=0