main (8a17c3d19) vs feat/nnx-set-defaults-true (73213e044)| Metric | main 8a17c3d19 | feat/nnx-set-defaults-true 73213e044 | Diff (feat/nnx-set-defaults-true − main) |
|---|---|---|---|
| Parameters | 1.104 billion | 1.104 billion | — |
| Final loss | 7.1810 | 7.1810 | 0 |
| TFLOP/s | 92.149 | 92.185 | +0.036 |
| Tok/s | 13889.7 | 13895.1 | +5.371 |
| Avg s/step | 4.661 | 4.217 | -0.444 |
| Memory % | 1.38 | 1.38 | 0 |
| JAX | 0.9.2 | 0.8.1 | — |
Diff = branch value − main value. Green = branch improved. Red = branch regressed.
XPK Start: Wed Apr 22 08:23:48 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 08:24:13.109354: 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 08:24:13.321731 136953331668800 max_utils.py:273] Attempting to initialize the jax distributed system... I0422 08:24:22.360110 136953331668800 distributed.py:149] Starting JAX distributed service on [::]:8482 I0422 08:24:22.362589 136953331668800 distributed.py:172] Connecting to JAX distributed service on mt-15-ocdbt-false-sv-ezjru-slice-job-0-0.mt-15-ocdbt-false-sv-ezjru:8482 I0422 08:24:23.555696 136953331668800 max_utils.py:284] Jax distributed system initialized! I0422 08:24:29.860690 136953331668800 max_utils.py:800] System Information: Jax Version: 0.9.2 I0422 08:24:29.860796 136953331668800 max_utils.py:801] System Information: Jaxlib Version: 0.9.2 I0422 08:24:29.860836 136953331668800 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 08:24:29.860871 136953331668800 train_utils.py:361] WARNING: Sequence packing is essentially ignored for synthetic data. Please use a real dataset to use sequence packing. I0422 08:24:30.551675 136953331668800 maxtext_utils.py:1565] Num_devices: 32, shape (1, 1, 1, 32, 1, 1, 1, 1, 1, 1, 1, 1, 1) I0422 08:24:30.551982 136953331668800 checkpointing.py:677] Setting up checkpoint logger... I0422 08:24:30.552038 136953331668800 checkpointing.py:233] Creating checkpoint manager with ocdbt=False and zarr3=True I0422 08:24:30.552083 136953331668800 pytree_checkpoint_handler.py:592] save_device_host_concurrent_bytes=None I0422 08:24:30.552436 136953331668800 base_pytree_checkpoint_handler.py:441] Created BasePyTreeCheckpointHandler: use_ocdbt=False, use_zarr3=True, pytree_metadata_options=PyTreeMetadataOptions(support_rich_types=False), array_metadata_store=<orbax.checkpoint._src.metadata.array_metadata_store.Store object at 0x7c8e52b1cbc0>, enable_pinned_host_transfer=False, save_concurrent_bytes: 96000000000 (89.4 GiB), restore_concurrent_bytes: 96000000000 (89.4 GiB) I0422 08:24:33.474181 136953331668800 checkpointing.py:265] Enabling policy for fixed interval checkpointing. I0422 08:24:33.474418 136953331668800 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 0x7c79f471de20>}, handler_registry=None I0422 08:24:33.474663 136953331668800 composite_checkpoint_handler.py:237] Deferred registration for item: "items". Adding handler `<orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeCheckpointHandler object at 0x7c79f471de20>` 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 08:24:33.474713 136953331668800 composite_checkpoint_handler.py:237] Deferred registration for item: "metrics". Adding handler `<orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonCheckpointHandler object at 0x7c79f47205f0>` 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 08:24:33.474750 136953331668800 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 0x7c79f471de20>, ('items', <class 'orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeRestoreArgs'>): <orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeCheckpointHandler object at 0x7c79f471de20>, ('metrics', <class 'orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonSaveArgs'>): <orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonCheckpointHandler object at 0x7c79f47205f0>, ('metrics', <class 'orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonRestoreArgs'>): <orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonCheckpointHandler object at 0x7c79f47205f0>}). I0422 08:24:33.475085 136953331668800 abstract_checkpointer.py:35] orbax-checkpoint version: 0.11.34 I0422 08:24:33.475155 136953331668800 async_checkpointer.py:192] [process=3][thread=MainThread] Using barrier_sync_fn: <function get_barrier_sync_fn.<locals>._fn at 0x7c78fc4f1800> timeout: 1200 secs and primary_host=0 for async checkpoint writes I0422 08:24:34.602389 136953331668800 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_false/checkpoints I0422 08:24:35.048096 136953331668800 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_main_20260422_071422/linen_xpk_main_20260422_071422_15_ocdbt_false/checkpoints: <orbax.checkpoint.checkpoint_manager.CheckpointManager object at 0x7c79f471ed80> I0422 08:24:35.048280 136953331668800 checkpointing.py:301] Checkpoint manager created! I0422 08:24:35.990427 136953331668800 nnx_wrappers.py:437] Unknown Logical: bfloat16[32,2048,2048]...................................... ('activation_batch', 'activation_norm_length', 'activation_embed'). I0422 08:24:35.990549 136953331668800 nnx_wrappers.py:437] Unknown Physical: bfloat16[32,2048,2048]...................................... ('fsdp', None, None). I0422 08:24:36.370740 136953331668800 attentions.py:1088] attentions/inputs_q Logical: bfloat16[32,2048,2048]...................................... ('activation_batch', 'activation_attn_length', 'activation_attn_embed'). I0422 08:24:36.370836 136953331668800 attentions.py:1088] attentions/inputs_q Physical: bfloat16[32,2048,2048]...................................... ('fsdp', None, None). I0422 08:24:36.387109 136953331668800 attentions.py:1089] attentions/inputs_kv Logical: bfloat16[32,2048,2048]...................................... ('activation_batch', 'activation_attn_length', 'activation_attn_embed'). I0422 08:24:36.387168 136953331668800 attentions.py:1089] attentions/inputs_kv Physical: bfloat16[32,2048,2048]...................................... ('fsdp', None, None). I0422 08:24:36.416241 136953331668800 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 08:24:36.416345 136953331668800 attentions.py:1154] attentions/query Physical: bfloat16[32,2048,16,128].................................... ('fsdp', None, None, None). I0422 08:24:36.432718 136953331668800 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 08:24:36.432778 136953331668800 attentions.py:1155] attentions/key Physical: bfloat16[32,2048,16,128].................................... ('fsdp', None, None, None). I0422 08:24:36.449033 136953331668800 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 08:24:36.449093 136953331668800 attentions.py:1156] attentions/value Physical: bfloat16[32,2048,16,128].................................... ('fsdp', None, None, None). I0422 08:24:36.473771 136953331668800 attentions.py:1198] attentions/out Logical: bfloat16[32,2048,16,128].................................... ('activation_batch', 'activation_attn_length', 'activation_heads', 'activation_kv'). I0422 08:24:36.473841 136953331668800 attentions.py:1198] attentions/out Physical: bfloat16[32,2048,16,128].................................... ('fsdp', None, None, None). I0422 08:24:36.494915 136953331668800 linears.py:525] linears/x Logical: bfloat16[32,2048,7168]...................................... ('activation_batch', 'activation_length', 'activation_mlp'). I0422 08:24:36.494994 136953331668800 linears.py:525] linears/x Physical: bfloat16[32,2048,7168]...................................... ('fsdp', None, None). I0422 08:24:36.704665 136953331668800 checkpointing.py:577] checkpoint manager exists so trying to load this run's existing checkpoint I0422 08:24:36.704781 136953331668800 checkpointing.py:665] No existing checkpoints found, not restoring checkpoint. fsdp: 32 I0422 08:24:38.157965 136953331668800 maxtext_utils.py:1668] params/params/decoder/decoder_norm/scale Shape: float32[2048] Logical: P('norm',) Physical: (None,) I0422 08:24:38.158097 136953331668800 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 08:24:38.158152 136953331668800 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 08:24:38.158212 136953331668800 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 08:24:38.158267 136953331668800 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 08:24:38.158307 136953331668800 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 08:24:38.158362 136953331668800 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 08:24:38.158416 136953331668800 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 08:24:38.158457 136953331668800 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 08:24:38.158495 136953331668800 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 08:24:38.158544 136953331668800 maxtext_utils.py:1668] params/params/decoder/logits_dense/kernel Shape: float32[2048,32000] Logical: P('embed_vocab', 'vocab') Physical: ('fsdp', None) I0422 08:24:38.158593 136953331668800 maxtext_utils.py:1668] params/params/token_embedder/embedding Shape: float32[32000,2048] Logical: P('vocab', 'embed_vocab') Physical: (None, 'fsdp') I0422 08:24:38.657458 136953331668800 train.py:155] train/xent Logical: float32[32,2048]............................................ ('activation_embed_and_logits_batch', 'activation_length'). I0422 08:24:38.657555 136953331668800 train.py:155] train/xent Physical: float32[32,2048]............................................ ('fsdp', None). I0422 08:24:38.672920 136953331668800 train.py:162] train/z_loss Logical: float32[32,2048]............................................ ('activation_embed_and_logits_batch', 'activation_length'). I0422 08:24:38.672991 136953331668800 train.py:162] train/z_loss Physical: float32[32,2048]............................................ ('fsdp', None). I0422 08:24:49.854768 136953331668800 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 08:24:49.855643 136953331668800 metric_logger.py:301] number parameters: 1.104 billion I0422 08:25:01.475031 136953331668800 checkpointing.py:772] Waiting for step 0 to finish before checkpoint... I0422 08:25:01.629542 136953331668800 checkpointing.py:776] Waited 0.15448999404907227 seconds for step 0 to finish before starting checkpointing. I0422 08:25:01.632953 136953331668800 checkpoint_manager.py:2009] [process=3][thread=MainThread][wait_until_finished] No Save Finalize thread to wait for. Returning. I0422 08:25:01.634500 136953331668800 checkpoint_manager.py:1512] [process=3] Saving checkpoint at step 0 I0422 08:25:01.635972 136953331668800 event_tracking.py:70] [process=3] [async] Started save checkpoint @ gs://lance-maxtext/linen_ckpt_xpk_main_20260422_071422/linen_xpk_main_20260422_071422_15_ocdbt_false/checkpoints/0. I0422 08:25:02.370804 136953331668800 signaling_client.py:364] Using JaxDistributedSignalingClient I0422 08:25:02.371762 136953331668800 jax_array_handlers.py:360] Scheduling D2H of 39 prioritized jax.Array. I0422 08:25:02.371822 136953331668800 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 08:25:02.648649 136953331668800 base_pytree_checkpoint_handler.py:154] [process=3][thread=MainThread] Initiated "orbax.checkpoint._src.serialization.jax_array_handlers.ArrayHandler".serialize. Time taken: 0.277923s I0422 08:25:02.648838 136953331668800 base_pytree_checkpoint_handler.py:130] [process=3] /jax/orbax/write/blocking_gbytes_per_sec: 5.442 GiB/s (total gbytes: 1.5 GiB) (time elapsed: 0.28343820571899414 s) (per-host) I0422 08:25:02.648895 136953331668800 base_pytree_checkpoint_handler.py:768] [process=3][thread=MainThread] Initiated Pytree async_save. Time taken: 0.283506s (batch_requests_ready=0.002216s, total_serialization_initiated=0.281214s, others=0.000077s) I0422 08:25:02.649021 136953331668800 composite_checkpoint_handler.py:715] [process=3][thread=MainThread] Initiated CompositeCheckpointHandler.async_save. Time taken: 0.287686s (all_items=0.000017s, per_item={'items': '0.00001669'}, temp_paths=0.287669) I0422 08:25:02.649803 136953331668800 event_tracking.py:125] [process=3] [async] Finished blocking save in 1.02 seconds. Continuing save @ gs://lance-maxtext/linen_ckpt_xpk_main_20260422_071422/linen_xpk_main_20260422_071422_15_ocdbt_false/checkpoints/0. I0422 08:25:02.650104 136822327863040 async_checkpointer.py:76] [process=3][thread=async_save] Background save thread started. Deadline for this save operation is 2026-04-22 08:45:02.650069 I0422 08:25:03.063983 136953331668800 checkpoint_manager.py:1560] [process=3][thread=MainThread][step=0] Starting CheckpointManager Save Finalize thread=save_finalize I0422 08:25:03.064371 136823933495040 async_checkpointer.py:280] [process=3][thread=save_finalize] Waiting for background save thread=async_save. I0422 08:25:03.064545 136953331668800 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_false/checkpoints', 'reached_preemption': False, 'preemption_received_at': None, 'synchronous': False, 'wait_for_prev_start_time': 1776846301.6329226, 'wait_for_prev_duration_secs': 6.198883056640625e-05, 'time_between_consecutive_saves_sec': None, 'checkpointer_blocking_start_time': 1776846301.6345396, 'checkpointer_blocking_duration_secs': 1.0156922340393066, 'get_old_steps_start_time': 1776846302.650257, 'get_old_steps_duration_secs': 3.0517578125e-05, 'checkpoint_manager_blocking_start_time': 1776846301.6300714, 'checkpoint_manager_blocking_duration_secs': 1.4344301223754883} I0422 08:25:03.064654 136953331668800 checkpointing.py:408] Started an asynchronous checkpoint save for step 0 I0422 08:25:03.064705 136953331668800 max_utils.py:750] Memstats: After params initialized: I0422 08:25:03.064769 136953331668800 max_utils.py:756] Using (GB) 0.43 / 31.25 (1.376000%) on TPU_10(process=3,(2,2,0,0)) I0422 08:25:03.064804 136953331668800 max_utils.py:756] Using (GB) 0.43 / 31.25 (1.376000%) on TPU_11(process=3,(3,2,0,0)) I0422 08:25:03.064832 136953331668800 max_utils.py:756] Using (GB) 0.43 / 31.25 (1.376000%) on TPU_14(process=3,(2,3,0,0)) I0422 08:25:03.064858 136953331668800 max_utils.py:756] Using (GB) 0.43 / 31.25 (1.376000%) on TPU_15(process=3,(3,3,0,0)) I0422 08:25:03.382185 136953331668800 metric_logger.py:196] completed step: 0, seconds: 11.619, TFLOP/s/device: 1.169, Tokens/s/device: 176.259, total_weights: 65536, loss: 10.874, lm_loss: 10.874, perplexity: 52779.875 I0422 08:25:03.561919 136953331668800 metric_logger.py:196] completed step: 1, seconds: 1.906, TFLOP/s/device: 7.130, Tokens/s/device: 1074.723, total_weights: 65536, loss: 10.874, lm_loss: 10.874, perplexity: 52779.875 I0422 08:25:04.002132 136953331668800 metric_logger.py:196] completed step: 2, seconds: 0.033, TFLOP/s/device: 417.975, Tokens/s/device: 63001.815, total_weights: 65536, loss: 10.560, lm_loss: 10.560, perplexity: 38547.719 I0422 08:25:04.149432 136953331668800 metric_logger.py:196] completed step: 3, seconds: 0.440, TFLOP/s/device: 30.855, Tokens/s/device: 4650.793, total_weights: 65536, loss: 9.980, lm_loss: 9.980, perplexity: 21590.506 I0422 08:25:04.153783 136953331668800 checkpointing.py:772] Waiting for step 5 to finish before checkpoint... I0422 08:25:04.443536 136953331668800 checkpointing.py:776] Waited 0.2897176742553711 seconds for step 5 to finish before starting checkpointing. I0422 08:25:04.446559 136953331668800 checkpoint_manager.py:2020] [process=3][thread=MainThread][step=0][wait_until_finished] Waiting for Save Finalize thread (save_finalize) to complete. I0422 08:25:06.203547 2809 google_auth_provider.cc:181] Running on GCE, using service account 562977990677-compute@developer.gserviceaccount.com I0422 08:25:08.660766 136822840407808 array_metadata_store.py:203] [process=3][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_false/checkpoints/0/items/array_metadatas/process_3 I0422 08:25:13.325314 136822327863040 base_pytree_checkpoint_handler.py:130] [process=3] /jax/orbax/write/gbytes_per_sec: 144.125 MiB/s (total gbytes: 1.5 GiB) (time elapsed: 10.95987844467163 s) (per-host) I0422 08:25:13.325434 136822327863040 async_checkpointer.py:90] [process=3][thread=async_save] 3 Handler Commit operations completed. Time taken: 10.675238s. I0422 08:25:20.103124 136822327863040 async_checkpointer.py:160] [process=3][thread=async_save] Background save thread done. Time taken: 17.452912s. I0422 08:25:20.103380 136823933495040 async_checkpointer.py:288] [process=3][thread=save_finalize] Done with waiting for background save thread=async_save. I0422 08:25:20.103514 136823933495040 async_checkpointer.py:298] [process=3][thread=save_finalize] No errors found in background save thread=async_save. I0422 08:25:20.103567 136823933495040 checkpoint_manager.py:2137] [process=3][thread=save_finalize][step=0] CheckpointManager Save Finalize is syncing with other hosts... I0422 08:25:20.105040 136823933495040 checkpoint_manager.py:2146] [process=3][thread=save_finalize][step=0] CheckpointManager Save Finalize is done on all hosts. I0422 08:25:20.105147 136953331668800 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 08:25:20.105223 136953331668800 checkpoint_manager.py:1452] Waiting for previous save to complete took 15.658666 seconds. If this number is high, consider checkpointing less frequently. I0422 08:25:20.108247 136953331668800 checkpoint_manager.py:1512] [process=3] Saving checkpoint at step 5 I0422 08:25:20.110815 136953331668800 event_tracking.py:70] [process=3] [async] Started save checkpoint @ gs://lance-maxtext/linen_ckpt_xpk_main_20260422_071422/linen_xpk_main_20260422_071422_15_ocdbt_false/checkpoints/5. I0422 08:25:20.398978 136953331668800 jax_array_handlers.py:360] Scheduling D2H of 39 prioritized jax.Array. I0422 08:25:20.399067 136953331668800 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 08:25:20.439103 136953331668800 base_pytree_checkpoint_handler.py:154] [process=3][thread=MainThread] Initiated "orbax.checkpoint._src.serialization.jax_array_handlers.ArrayHandler".serialize. Time taken: 0.041225s I0422 08:25:20.439267 136953331668800 base_pytree_checkpoint_handler.py:130] [process=3] /jax/orbax/write/blocking_gbytes_per_sec: 34.528 GiB/s (total gbytes: 1.5 GiB) (time elapsed: 0.04467606544494629 s) (per-host) I0422 08:25:20.439316 136953331668800 base_pytree_checkpoint_handler.py:768] [process=3][thread=MainThread] Initiated Pytree async_save. Time taken: 0.044734s (batch_requests_ready=0.001781s, total_serialization_initiated=0.042889s, others=0.000064s) I0422 08:25:20.439404 136953331668800 composite_checkpoint_handler.py:715] [process=3][thread=MainThread] Initiated CompositeCheckpointHandler.async_save. Time taken: 0.049572s (all_items=0.000017s, per_item={'items': '0.00001669'}, temp_paths=0.049555) I0422 08:25:20.440160 136953331668800 event_tracking.py:125] [process=3] [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_false/checkpoints/5. I0422 08:25:20.440459 136823933495040 async_checkpointer.py:76] [process=3][thread=async_save] Background save thread started. Deadline for this save operation is 2026-04-22 08:45:20.440426 I0422 08:25:20.448321 136953331668800 checkpoint_manager.py:1560] [process=3][thread=MainThread][step=5] Starting CheckpointManager Save Finalize thread=save_finalize I0422 08:25:20.448633 136823430969088 async_checkpointer.py:280] [process=3][thread=save_finalize] Waiting for background save thread=async_save. I0422 08:25:20.448800 136953331668800 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_false/checkpoints', 'reached_preemption': False, 'preemption_received_at': None, 'synchronous': False, 'wait_for_prev_start_time': 1776846304.4465313, 'wait_for_prev_duration_secs': 15.658666133880615, 'time_between_consecutive_saves_sec': None, 'checkpointer_blocking_start_time': 1776846320.1082869, 'checkpointer_blocking_duration_secs': 0.3323397636413574, 'get_old_steps_start_time': 1776846320.44065, 'get_old_steps_duration_secs': 3.719329833984375e-05, 'checkpoint_manager_blocking_start_time': 1776846304.444206, 'checkpoint_manager_blocking_duration_secs': 16.004557371139526} I0422 08:25:20.448917 136953331668800 checkpointing.py:408] Started an asynchronous checkpoint save for step 5 I0422 08:25:20.449756 136953331668800 metric_logger.py:196] completed step: 4, seconds: 0.153, TFLOP/s/device: 88.871, Tokens/s/device: 13395.602, total_weights: 65536, loss: 9.460, lm_loss: 9.460, perplexity: 12840.355 I0422 08:25:20.455999 136953331668800 metric_logger.py:196] completed step: 5, seconds: 0.147, TFLOP/s/device: 92.251, Tokens/s/device: 13905.014, total_weights: 65536, loss: 8.959, lm_loss: 8.959, perplexity: 7776.467 I0422 08:25:26.914072 136822840407808 array_metadata_store.py:203] [process=3][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_false/checkpoints/5/items/array_metadatas/process_3 I0422 08:25:31.926570 136823933495040 base_pytree_checkpoint_handler.py:130] [process=3] /jax/orbax/write/gbytes_per_sec: 136.976 MiB/s (total gbytes: 1.5 GiB) (time elapsed: 11.531921148300171 s) (per-host) I0422 08:25:31.926697 136823933495040 async_checkpointer.py:90] [process=3][thread=async_save] 3 Handler Commit operations completed. Time taken: 11.486111s. I0422 08:25:37.899536 136823933495040 async_checkpointer.py:160] [process=3][thread=async_save] Background save thread done. Time taken: 17.458934s. I0422 08:25:37.899837 136823430969088 async_checkpointer.py:288] [process=3][thread=save_finalize] Done with waiting for background save thread=async_save. I0422 08:25:37.899970 136823430969088 async_checkpointer.py:298] [process=3][thread=save_finalize] No errors found in background save thread=async_save. I0422 08:25:37.900019 136823430969088 checkpoint_manager.py:2137] [process=3][thread=save_finalize][step=5] CheckpointManager Save Finalize is syncing with other hosts... I0422 08:25:37.901397 136823430969088 checkpoint_manager.py:2146] [process=3][thread=save_finalize][step=5] CheckpointManager Save Finalize is done on all hosts. I0422 08:25:43.275292 136953331668800 metric_logger.py:196] completed step: 6, seconds: 16.301, TFLOP/s/device: 0.834, Tokens/s/device: 125.639, total_weights: 65536, loss: 8.469, lm_loss: 8.469, perplexity: 4762.740 I0422 08:25:43.422845 136953331668800 metric_logger.py:196] completed step: 7, seconds: 22.673, TFLOP/s/device: 0.599, Tokens/s/device: 90.326, total_weights: 65536, loss: 8.003, lm_loss: 8.003, perplexity: 2989.151 I0422 08:25:43.570162 136953331668800 metric_logger.py:196] completed step: 8, seconds: 0.152, TFLOP/s/device: 89.390, Tokens/s/device: 13473.861, total_weights: 65536, loss: 7.572, lm_loss: 7.572, perplexity: 1943.098 I0422 08:25:43.717012 136953331668800 checkpointing.py:772] Waiting for step 9 to finish before checkpoint... I0422 08:25:43.717698 136953331668800 checkpointing.py:776] Waited 0.0007061958312988281 seconds for step 9 to finish before starting checkpointing. I0422 08:25:43.719728 136953331668800 checkpoint_manager.py:2009] [process=3][thread=MainThread][wait_until_finished] No Save Finalize thread to wait for. Returning. I0422 08:25:43.721373 136953331668800 checkpoint_manager.py:1512] [process=3] Saving checkpoint at step 9 I0422 08:25:43.722731 136953331668800 event_tracking.py:70] [process=3] [async] Started save checkpoint @ gs://lance-maxtext/linen_ckpt_xpk_main_20260422_071422/linen_xpk_main_20260422_071422_15_ocdbt_false/checkpoints/9. I0422 08:25:44.036419 136953331668800 jax_array_handlers.py:360] Scheduling D2H of 39 prioritized jax.Array. I0422 08:25:44.036510 136953331668800 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 08:25:44.077788 136953331668800 base_pytree_checkpoint_handler.py:154] [process=3][thread=MainThread] Initiated "orbax.checkpoint._src.serialization.jax_array_handlers.ArrayHandler".serialize. Time taken: 0.042233s I0422 08:25:44.077971 136953331668800 base_pytree_checkpoint_handler.py:130] [process=3] /jax/orbax/write/blocking_gbytes_per_sec: 33.748 GiB/s (total gbytes: 1.5 GiB) (time elapsed: 0.04570889472961426 s) (per-host) I0422 08:25:44.078023 136953331668800 base_pytree_checkpoint_handler.py:768] [process=3][thread=MainThread] Initiated Pytree async_save. Time taken: 0.045773s (batch_requests_ready=0.001651s, total_serialization_initiated=0.044053s, others=0.000069s) I0422 08:25:44.078109 136953331668800 composite_checkpoint_handler.py:715] [process=3][thread=MainThread] Initiated CompositeCheckpointHandler.async_save. Time taken: 0.050030s (all_items=0.000011s, per_item={'items': '0.00001097'}, temp_paths=0.050019) I0422 08:25:44.078766 136953331668800 event_tracking.py:125] [process=3] [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_false/checkpoints/9. I0422 08:25:44.079099 136822840407808 async_checkpointer.py:76] [process=3][thread=async_save] Background save thread started. Deadline for this save operation is 2026-04-22 08:45:44.079060 I0422 08:25:44.081129 136953331668800 checkpoint_manager.py:1560] [process=3][thread=MainThread][step=9] Starting CheckpointManager Save Finalize thread=save_finalize I0422 08:25:44.081398 136823430969088 async_checkpointer.py:280] [process=3][thread=save_finalize] Waiting for background save thread=async_save. I0422 08:25:44.081532 136953331668800 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_false/checkpoints', 'reached_preemption': False, 'preemption_received_at': None, 'synchronous': False, 'wait_for_prev_start_time': 1776846343.7196977, 'wait_for_prev_duration_secs': 6.914138793945312e-05, 'time_between_consecutive_saves_sec': 5.818260431289673, 'checkpointer_blocking_start_time': 1776846343.7214103, 'checkpointer_blocking_duration_secs': 0.3578364849090576, 'get_old_steps_start_time': 1776846344.0792649, 'get_old_steps_duration_secs': 2.47955322265625e-05, 'checkpoint_manager_blocking_start_time': 1776846343.717942, 'checkpoint_manager_blocking_duration_secs': 0.3635599613189697} I0422 08:25:44.081691 136953331668800 checkpointing.py:408] Started an asynchronous checkpoint save for step 9 I0422 08:25:44.081738 136953331668800 checkpoint_manager.py:2020] [process=3][thread=MainThread][step=9][wait_until_finished] Waiting for Save Finalize thread (save_finalize) to complete. I0422 08:25:49.663063 136823422576384 array_metadata_store.py:203] [process=3][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_false/checkpoints/9/items/array_metadatas/process_3 I0422 08:25:55.004242 136822840407808 base_pytree_checkpoint_handler.py:130] [process=3] /jax/orbax/write/gbytes_per_sec: 143.967 MiB/s (total gbytes: 1.5 GiB) (time elapsed: 10.97196364402771 s) (per-host) I0422 08:25:55.004353 136822840407808 async_checkpointer.py:90] [process=3][thread=async_save] 3 Handler Commit operations completed. Time taken: 10.925141s. I0422 08:26:01.127055 136822840407808 async_checkpointer.py:160] [process=3][thread=async_save] Background save thread done. Time taken: 17.047829s. I0422 08:26:01.127331 136823430969088 async_checkpointer.py:288] [process=3][thread=save_finalize] Done with waiting for background save thread=async_save. I0422 08:26:01.127454 136823430969088 async_checkpointer.py:298] [process=3][thread=save_finalize] No errors found in background save thread=async_save. I0422 08:26:01.127496 136823430969088 checkpoint_manager.py:2137] [process=3][thread=save_finalize][step=9] CheckpointManager Save Finalize is syncing with other hosts... I0422 08:26:01.129061 136823430969088 checkpoint_manager.py:2146] [process=3][thread=save_finalize][step=9] CheckpointManager Save Finalize is done on all hosts. I0422 08:26:01.129234 136953331668800 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 08:26:01.129384 136953331668800 checkpoint_manager.py:2009] [process=3][thread=MainThread][wait_until_finished] No Save Finalize thread to wait for. Returning. I0422 08:26:01.130364 136953331668800 metric_logger.py:196] completed step: 9, seconds: 0.147, TFLOP/s/device: 92.149, Tokens/s/device: 13889.737, 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 08:26:13 UTC 2026 EXIT_CODE=0
XPK Start: Wed Apr 22 17:04:01 UTC 2026 2026-04-22 17:04:05.825972: E external/local_xla/xla/stream_executor/cuda/cuda_fft.cc:467] Unable to register cuFFT factory: Attempting to register factory for plugin cuFFT when one has already been registered WARNING: All log messages before absl::InitializeLog() is called are written to STDERR E0000 00:00:1776877445.838723 10 cuda_dnn.cc:8579] Unable to register cuDNN factory: Attempting to register factory for plugin cuDNN when one has already been registered E0000 00:00:1776877445.842455 10 cuda_blas.cc:1407] Unable to register cuBLAS factory: Attempting to register factory for plugin cuBLAS when one has already been registered W0000 00:00:1776877445.853988 10 computation_placer.cc:177] computation placer already registered. Please check linkage and avoid linking the same target more than once. W0000 00:00:1776877445.854005 10 computation_placer.cc:177] computation placer already registered. Please check linkage and avoid linking the same target more than once. W0000 00:00:1776877445.854008 10 computation_placer.cc:177] computation placer already registered. Please check linkage and avoid linking the same target more than once. W0000 00:00:1776877445.854010 10 computation_placer.cc:177] computation placer already registered. Please check linkage and avoid linking the same target more than once. 2026-04-22 17:04:25.070047: 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 17:04:25.594604 134043235866432 max_utils.py:273] Attempting to initialize the jax distributed system... INFO:2026-04-22 17:04:34,636:jax._src.distributed:140: Starting JAX distributed service on [::]:8482 I0422 17:04:34.636157 134043235866432 distributed.py:140] Starting JAX distributed service on [::]:8482 INFO:2026-04-22 17:04:34,638:jax._src.distributed:157: Connecting to JAX distributed service on mt-15-ocdbt-false-sv-zp839-slice-job-0-0.mt-15-ocdbt-false-sv-zp839:8482 I0422 17:04:34.638734 134043235866432 distributed.py:157] Connecting to JAX distributed service on mt-15-ocdbt-false-sv-zp839-slice-job-0-0.mt-15-ocdbt-false-sv-zp839:8482 I0422 17:04:35.740076 134043235866432 max_utils.py:284] Jax distributed system initialized! I0422 17:04:42.596638 134043235866432 max_utils.py:800] System Information: Jax Version: 0.8.1 I0422 17:04:42.596743 134043235866432 max_utils.py:801] System Information: Jaxlib Version: 0.8.1 I0422 17:04:42.596783 134043235866432 max_utils.py:802] System Information: Jax Backend: PJRT C API TFRT TPU v6 lite Built on Nov 12 2025 14:16:36 (1762985796) cl/831091709 I0422 17:04:42.596816 134043235866432 train_utils.py:377] WARNING: Sequence packing is essentially ignored for synthetic data. Please use a real dataset to use sequence packing. I0422 17:04:43.220369 134043235866432 maxtext_utils.py:1631] Num_devices: 32, shape (1, 1, 1, 32, 1, 1, 1, 1, 1, 1, 1, 1, 1) I0422 17:04:43.220978 134043235866432 maxtext_utils.py:1631] Num_devices: 32, shape (1, 1, 1, 32, 1, 1, 1, 1, 1, 1, 1, 1, 1) I0422 17:04:43.221160 134043235866432 checkpointing.py:688] Setting up checkpoint logger... I0422 17:04:43.221210 134043235866432 checkpointing.py:234] Creating checkpoint manager with ocdbt=False and zarr3=True I0422 17:04:43.221252 134043235866432 pytree_checkpoint_handler.py:589] save_device_host_concurrent_bytes=None I0422 17:04:43.221606 134043235866432 base_pytree_checkpoint_handler.py:415] Created BasePyTreeCheckpointHandler: use_ocdbt=False, use_zarr3=True, pytree_metadata_options=PyTreeMetadataOptions(support_rich_types=False), array_metadata_store=<orbax.checkpoint._src.metadata.array_metadata_store.Store object at 0x79e8d4178e90>, enable_pinned_host_transfer=False, save_concurrent_bytes: 96000000000 (89.4 GiB), restore_concurrent_bytes: 96000000000 (89.4 GiB) I0422 17:04:46.108207 134043235866432 checkpointing.py:266] Enabling policy for fixed interval checkpointing. I0422 17:04:46.108441 134043235866432 checkpoint_manager.py:709] [process=3][thread=MainThread] CheckpointManager init: checkpointers=None, item_names=('items',), item_handlers={'items': <orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeCheckpointHandler object at 0x79d670f60c50>}, handler_registry=None I0422 17:04:46.108673 134043235866432 composite_checkpoint_handler.py:237] Deferred registration for item: "items". Adding handler `<orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeCheckpointHandler object at 0x79d670f60c50>` 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 17:04:46.108729 134043235866432 composite_checkpoint_handler.py:237] Deferred registration for item: "metrics". Adding handler `<orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonCheckpointHandler object at 0x79d671b7b080>` 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 17:04:46.108767 134043235866432 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 0x79d670f60c50>, ('items', <class 'orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeRestoreArgs'>): <orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeCheckpointHandler object at 0x79d670f60c50>, ('metrics', <class 'orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonSaveArgs'>): <orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonCheckpointHandler object at 0x79d671b7b080>, ('metrics', <class 'orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonRestoreArgs'>): <orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonCheckpointHandler object at 0x79d671b7b080>}). I0422 17:04:46.109273 134043235866432 abstract_checkpointer.py:35] orbax-checkpoint version: 0.11.33 I0422 17:04:46.109350 134043235866432 async_checkpointer.py:177] [process=3][thread=MainThread] Using barrier_sync_fn: <function get_barrier_sync_fn.<locals>._fn at 0x79d67050c180> timeout: 600 secs and primary_host=0 for async checkpoint writes I0422 17:04:47.278337 134043235866432 checkpoint_manager.py:1818] Found 0 checkpoint steps in gs://lance-maxtext/linen_ckpt_xpk_feat_nnx_set_defaults_true_20260422_154647/linen_xpk_feat_nnx_set_defaults_true_20260422_154647_15_ocdbt_false/checkpoints I0422 17:04:47.333082 134043235866432 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_feat_nnx_set_defaults_true_20260422_154647/linen_xpk_feat_nnx_set_defaults_true_20260422_154647_15_ocdbt_false/checkpoints: <orbax.checkpoint.checkpoint_manager.CheckpointManager object at 0x79d670f63080> I0422 17:04:47.333233 134043235866432 checkpointing.py:302] Checkpoint manager created! I0422 17:04:48.234910 134043235866432 nnx_wrappers.py:455] Unknown Logical: bfloat16[32,2048,2048]...................................... ('activation_batch', 'activation_norm_length', 'activation_embed'). I0422 17:04:48.235041 134043235866432 nnx_wrappers.py:455] Unknown Physical: bfloat16[32,2048,2048]...................................... ('fsdp', None, None). I0422 17:04:48.607046 134043235866432 attentions.py:1084] attentions/inputs_q Logical: bfloat16[32,2048,2048]...................................... ('activation_batch', 'activation_attn_length', 'activation_attn_embed'). I0422 17:04:48.607139 134043235866432 attentions.py:1084] attentions/inputs_q Physical: bfloat16[32,2048,2048]...................................... ('fsdp', None, None). I0422 17:04:48.622654 134043235866432 attentions.py:1085] attentions/inputs_kv Logical: bfloat16[32,2048,2048]...................................... ('activation_batch', 'activation_attn_length', 'activation_attn_embed'). I0422 17:04:48.622720 134043235866432 attentions.py:1085] attentions/inputs_kv Physical: bfloat16[32,2048,2048]...................................... ('fsdp', None, None). I0422 17:04:48.645226 134043235866432 attentions.py:1150] attentions/query Logical: bfloat16[32,2048,16,128].................................... ('activation_kv_batch', 'activation_attn_length', 'activation_kv_heads', 'activation_kv_head_dim'). I0422 17:04:48.645296 134043235866432 attentions.py:1150] attentions/query Physical: bfloat16[32,2048,16,128].................................... ('fsdp', None, None, None). I0422 17:04:48.661062 134043235866432 attentions.py:1151] attentions/key Logical: bfloat16[32,2048,16,128].................................... ('activation_kv_batch', 'activation_attn_length', 'activation_kv_heads', 'activation_kv_head_dim'). I0422 17:04:48.661123 134043235866432 attentions.py:1151] attentions/key Physical: bfloat16[32,2048,16,128].................................... ('fsdp', None, None, None). I0422 17:04:48.676640 134043235866432 attentions.py:1152] attentions/value Logical: bfloat16[32,2048,16,128].................................... ('activation_kv_batch', 'activation_attn_length', 'activation_kv_heads', 'activation_kv_head_dim'). I0422 17:04:48.676706 134043235866432 attentions.py:1152] attentions/value Physical: bfloat16[32,2048,16,128].................................... ('fsdp', None, None, None). I0422 17:04:48.701174 134043235866432 attentions.py:1193] attentions/out Logical: bfloat16[32,2048,16,128].................................... ('activation_batch', 'activation_attn_length', 'activation_heads', 'activation_kv'). I0422 17:04:48.701244 134043235866432 attentions.py:1193] attentions/out Physical: bfloat16[32,2048,16,128].................................... ('fsdp', None, None, None). I0422 17:04:48.720452 134043235866432 linears.py:541] linears/x Logical: bfloat16[32,2048,7168]...................................... ('activation_batch', 'activation_length', 'activation_mlp'). I0422 17:04:48.720518 134043235866432 linears.py:541] linears/x Physical: bfloat16[32,2048,7168]...................................... ('fsdp', None, None). I0422 17:04:48.908987 134043235866432 checkpointing.py:578] checkpoint manager exists so trying to load this run's existing checkpoint I0422 17:04:48.909095 134043235866432 checkpointing.py:676] No existing checkpoints found, not restoring checkpoint. fsdp: 32 I0422 17:04:50.343677 134043235866432 maxtext_utils.py:1740] params/params/decoder/decoder_norm/scale Shape: float32[2048] Logical: PartitionSpec('norm',) Physical: (None,) I0422 17:04:50.343813 134043235866432 maxtext_utils.py:1740] params/params/decoder/layers/mlp/wi_0/kernel Shape: float32[2048,16,7168] Logical: PartitionSpec('embed', 'layers', 'mlp') Physical: ('fsdp', None, None) I0422 17:04:50.343869 134043235866432 maxtext_utils.py:1740] params/params/decoder/layers/mlp/wi_1/kernel Shape: float32[2048,16,7168] Logical: PartitionSpec('embed', 'layers', 'mlp') Physical: ('fsdp', None, None) I0422 17:04:50.343944 134043235866432 maxtext_utils.py:1740] params/params/decoder/layers/mlp/wo/kernel Shape: float32[7168,16,2048] Logical: PartitionSpec('mlp', 'layers', 'embed') Physical: (None, None, 'fsdp') I0422 17:04:50.344012 134043235866432 maxtext_utils.py:1740] params/params/decoder/layers/post_self_attention_layer_norm/scale Shape: float32[2048,16] Logical: PartitionSpec('norm', 'layers') Physical: (None, None) I0422 17:04:50.344054 134043235866432 maxtext_utils.py:1740] params/params/decoder/layers/pre_self_attention_layer_norm/scale Shape: float32[2048,16] Logical: PartitionSpec('norm', 'layers') Physical: (None, None) I0422 17:04:50.344109 134043235866432 maxtext_utils.py:1740] params/params/decoder/layers/self_attention/key/kernel Shape: float32[2048,16,16,128] Logical: PartitionSpec('embed', 'layers', 'kv_heads', 'kv_head_dim') Physical: ('fsdp', None, None, None) I0422 17:04:50.344162 134043235866432 maxtext_utils.py:1740] params/params/decoder/layers/self_attention/out/kernel Shape: float32[16,16,128,2048] Logical: PartitionSpec('heads', 'layers', 'kv', 'embed') Physical: (None, None, None, 'fsdp') I0422 17:04:50.344203 134043235866432 maxtext_utils.py:1740] params/params/decoder/layers/self_attention/query/kernel Shape: float32[2048,16,16,128] Logical: PartitionSpec('embed', 'layers', 'q_heads', 'kv') Physical: ('fsdp', None, None, None) I0422 17:04:50.344240 134043235866432 maxtext_utils.py:1740] params/params/decoder/layers/self_attention/value/kernel Shape: float32[2048,16,16,128] Logical: PartitionSpec('embed', 'layers', 'kv_heads', 'kv_head_dim') Physical: ('fsdp', None, None, None) I0422 17:04:50.344290 134043235866432 maxtext_utils.py:1740] params/params/decoder/logits_dense/kernel Shape: float32[2048,32000] Logical: PartitionSpec('embed', 'vocab') Physical: ('fsdp', None) I0422 17:04:50.344340 134043235866432 maxtext_utils.py:1740] params/params/token_embedder/embedding Shape: float32[32000,2048] Logical: PartitionSpec('vocab', 'embed') Physical: (None, 'fsdp') I0422 17:04:50.847592 134043235866432 train.py:158] train/xent Logical: float32[32,2048]............................................ ('activation_embed_and_logits_batch', 'activation_length'). I0422 17:04:50.847689 134043235866432 train.py:158] train/xent Physical: float32[32,2048]............................................ ('fsdp', None). I0422 17:04:50.862367 134043235866432 train.py:165] train/z_loss Logical: float32[32,2048]............................................ ('activation_embed_and_logits_batch', 'activation_length'). I0422 17:04:50.862428 134043235866432 train.py:165] train/z_loss Physical: float32[32,2048]............................................ ('fsdp', None). I0422 17:05:01.529616 134043235866432 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 17:05:01.530487 134043235866432 metric_logger.py:289] number parameters: 1.104 billion I0422 17:05:12.530102 134043235866432 checkpointing.py:794] Waiting for step 0 to finish before checkpoint... I0422 17:05:12.973706 134043235866432 checkpointing.py:798] Waited 0.44357991218566895 seconds for step 0 to finish before starting checkpointing. I0422 17:05:12.976148 134043235866432 checkpoint_manager.py:2013] [process=3][thread=MainThread][wait_until_finished] No Save Finalize thread to wait for. Returning. I0422 17:05:12.978102 134043235866432 checkpoint_manager.py:1518] [process=3] Saving checkpoint at step 0 I0422 17:05:12.981159 134043235866432 async_checkpointer.py:452] [process=3] Started async saving checkpoint to gs://lance-maxtext/linen_ckpt_xpk_feat_nnx_set_defaults_true_20260422_154647/linen_xpk_feat_nnx_set_defaults_true_20260422_154647_15_ocdbt_false/checkpoints/0. I0422 17:05:13.542217 134043235866432 signaling_client.py:364] Using JaxDistributedSignalingClient I0422 17:05:13.543222 134043235866432 jax_array_handlers.py:358] Scheduling D2H of 39 prioritized jax.Array. I0422 17:05:13.543280 134043235866432 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 17:05:13.791831 134043235866432 base_pytree_checkpoint_handler.py:153] [process=3][thread=MainThread] Initiated "orbax.checkpoint._src.serialization.jax_array_handlers.ArrayHandler".serialize. Time taken: 0.249692s I0422 17:05:13.792019 134043235866432 base_pytree_checkpoint_handler.py:129] [process=3] /jax/checkpoint/write/blocking_gbytes_per_sec: 2.949 GiB/s (total gbytes: 1.5 GiB) (time elapsed: 0.5230188369750977 s) (per-host) I0422 17:05:13.792077 134043235866432 base_pytree_checkpoint_handler.py:737] [process=3][thread=MainThread] Initiated Pytree async_save. Time taken: 0.523087s (batch_requests_ready=0.269768s, total_serialization_initiated=0.253242s, others=0.000076s) I0422 17:05:13.792171 134043235866432 composite_checkpoint_handler.py:715] [process=3][thread=MainThread] Initiated CompositeCheckpointHandler.async_save. Time taken: 0.527202s (all_items=0.000018s, per_item={'items': '0.00001836'}, temp_paths=0.527184) I0422 17:05:13.793102 133920192440064 async_checkpointer.py:79] [process=3][thread=async_save] Background save thread started. I0422 17:05:13.793265 134043235866432 async_checkpointer.py:561] Finished blocking save. Time taken: 0.815048s. Continuing background save to gs://lance-maxtext/linen_ckpt_xpk_feat_nnx_set_defaults_true_20260422_154647/linen_xpk_feat_nnx_set_defaults_true_20260422_154647_15_ocdbt_false/checkpoints/0. I0422 17:05:14.236833 134043235866432 checkpoint_manager.py:1566] [process=3][thread=MainThread][step=0] Starting CheckpointManager Save Finalize thread=save_finalize I0422 17:05:14.237236 133916442736384 async_checkpointer.py:265] [process=3][thread=save_finalize] Waiting for background save thread=async_save. I0422 17:05:14.237401 134043235866432 standard_logger.py:34] {'step': 0, 'event_type': 'save', 'directory': 'gs://lance-maxtext/linen_ckpt_xpk_feat_nnx_set_defaults_true_20260422_154647/linen_xpk_feat_nnx_set_defaults_true_20260422_154647_15_ocdbt_false/checkpoints', 'reached_preemption': False, 'preemption_received_at': None, 'synchronous': False, 'wait_for_prev_start_time': 1776877512.9761307, 'wait_for_prev_duration_secs': 6.127357482910156e-05, 'time_between_consecutive_saves_sec': None, 'checkpointer_blocking_start_time': 1776877512.9781854, 'checkpointer_blocking_duration_secs': 0.8152000904083252, 'get_old_steps_start_time': 1776877513.793404, 'get_old_steps_duration_secs': 2.3365020751953125e-05, 'checkpoint_manager_blocking_start_time': 1776877512.9743397, 'checkpoint_manager_blocking_duration_secs': 1.2630195617675781} I0422 17:05:14.237511 134043235866432 checkpointing.py:409] Started an asynchronous checkpoint save for step 0 I0422 17:05:14.237565 134043235866432 max_utils.py:750] Memstats: After params initialized: I0422 17:05:14.237623 134043235866432 max_utils.py:756] Using (GB) 0.43 / 31.25 (1.376000%) on TPU_10(process=3,(2,2,0,0)) I0422 17:05:14.237658 134043235866432 max_utils.py:756] Using (GB) 0.43 / 31.25 (1.376000%) on TPU_11(process=3,(3,2,0,0)) I0422 17:05:14.237687 134043235866432 max_utils.py:756] Using (GB) 0.43 / 31.25 (1.376000%) on TPU_14(process=3,(2,3,0,0)) I0422 17:05:14.237715 134043235866432 max_utils.py:756] Using (GB) 0.43 / 31.25 (1.376000%) on TPU_15(process=3,(3,3,0,0)) I0422 17:05:14.509128 134043235866432 metric_logger.py:185] completed step: 0, seconds: 11.000, TFLOP/s/device: 1.235, Tokens/s/device: 186.190, total_weights: 65536, loss: 10.874 I0422 17:05:14.667778 134043235866432 metric_logger.py:185] completed step: 1, seconds: 1.978, TFLOP/s/device: 6.870, Tokens/s/device: 1035.581, total_weights: 65536, loss: 10.874 I0422 17:05:15.071725 134043235866432 metric_logger.py:185] completed step: 2, seconds: 0.011, TFLOP/s/device: 1190.809, Tokens/s/device: 179491.674, total_weights: 65536, loss: 10.560 I0422 17:05:15.219073 134043235866432 metric_logger.py:185] completed step: 3, seconds: 0.405, TFLOP/s/device: 33.573, Tokens/s/device: 5060.551, total_weights: 65536, loss: 9.980 I0422 17:05:15.223767 134043235866432 checkpointing.py:794] Waiting for step 5 to finish before checkpoint... I0422 17:05:15.513723 134043235866432 checkpointing.py:798] Waited 0.28992700576782227 seconds for step 5 to finish before starting checkpointing. I0422 17:05:15.516556 134043235866432 checkpoint_manager.py:2024] [process=3][thread=MainThread][step=0][wait_until_finished] Waiting for Save Finalize thread (save_finalize) to complete. I0422 17:05:16.871053 2598 google_auth_provider.cc:181] Running on GCE, using service account 562977990677-compute@developer.gserviceaccount.com I0422 17:05:18.771387 133920200832768 array_metadata_store.py:203] [process=3][thread=array_type_handler] Wrote 39 array_metadata.ArrayMetadata to gs://lance-maxtext/linen_ckpt_xpk_feat_nnx_set_defaults_true_20260422_154647/linen_xpk_feat_nnx_set_defaults_true_20260422_154647_15_ocdbt_false/checkpoints/0/items/array_metadatas/process_3 I0422 17:05:24.052941 133920192440064 base_pytree_checkpoint_handler.py:129] [process=3] /jax/checkpoint/write/gbytes_per_sec: 146.478 MiB/s (total gbytes: 1.5 GiB) (time elapsed: 10.783883810043335 s) (per-host) I0422 17:05:24.053064 133920192440064 async_checkpointer.py:90] [process=3][thread=async_save] 3 Handler Commit operations completed. Time taken: 10.259855s. I0422 17:05:30.258686 133920192440064 async_checkpointer.py:144] [process=3][thread=async_save] Background save thread done. Time taken: 16.465451s. I0422 17:05:30.258998 133916442736384 async_checkpointer.py:273] [process=3][thread=save_finalize] Done with waiting for background save thread=async_save. I0422 17:05:30.259118 133916442736384 async_checkpointer.py:283] [process=3][thread=save_finalize] No errors found in background save thread=async_save. I0422 17:05:30.259169 133916442736384 checkpoint_manager.py:2133] [process=3][thread=save_finalize][step=0] CheckpointManager Save Finalize is syncing with other hosts... I0422 17:05:30.260783 133916442736384 checkpoint_manager.py:2142] [process=3][thread=save_finalize][step=0] CheckpointManager Save Finalize is done on all hosts. I0422 17:05:30.261008 134043235866432 checkpoint_manager.py:2036] [process=3][thread=MainThread][step=0][wait_until_finished] Done waiting for Save Finalize thread (save_finalize) running at step=0. W0422 17:05:30.261152 134043235866432 checkpoint_manager.py:1458] Waiting for previous save to complete took 14.744594 seconds. If this number is high, consider checkpointing less frequently. I0422 17:05:30.263053 134043235866432 checkpoint_manager.py:1518] [process=3] Saving checkpoint at step 5 I0422 17:05:30.267196 134043235866432 async_checkpointer.py:452] [process=3] Started async saving checkpoint to gs://lance-maxtext/linen_ckpt_xpk_feat_nnx_set_defaults_true_20260422_154647/linen_xpk_feat_nnx_set_defaults_true_20260422_154647_15_ocdbt_false/checkpoints/5. I0422 17:05:30.801844 134043235866432 jax_array_handlers.py:358] Scheduling D2H of 39 prioritized jax.Array. I0422 17:05:30.801956 134043235866432 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 17:05:30.843257 134043235866432 base_pytree_checkpoint_handler.py:153] [process=3][thread=MainThread] Initiated "orbax.checkpoint._src.serialization.jax_array_handlers.ArrayHandler".serialize. Time taken: 0.042651s I0422 17:05:30.843421 134043235866432 base_pytree_checkpoint_handler.py:129] [process=3] /jax/checkpoint/write/blocking_gbytes_per_sec: 5.186 GiB/s (total gbytes: 1.5 GiB) (time elapsed: 0.29744815826416016 s) (per-host) I0422 17:05:30.843474 134043235866432 base_pytree_checkpoint_handler.py:737] [process=3][thread=MainThread] Initiated Pytree async_save. Time taken: 0.297508s (batch_requests_ready=0.252997s, total_serialization_initiated=0.044444s, others=0.000068s) I0422 17:05:30.843569 134043235866432 composite_checkpoint_handler.py:715] [process=3][thread=MainThread] Initiated CompositeCheckpointHandler.async_save. Time taken: 0.302233s (all_items=0.000015s, per_item={'items': '0.00001502'}, temp_paths=0.302218) I0422 17:05:30.844384 133911070971648 async_checkpointer.py:79] [process=3][thread=async_save] Background save thread started. I0422 17:05:30.844546 134043235866432 async_checkpointer.py:561] Finished blocking save. Time taken: 0.581428s. Continuing background save to gs://lance-maxtext/linen_ckpt_xpk_feat_nnx_set_defaults_true_20260422_154647/linen_xpk_feat_nnx_set_defaults_true_20260422_154647_15_ocdbt_false/checkpoints/5. I0422 17:05:30.859549 134043235866432 checkpoint_manager.py:1566] [process=3][thread=MainThread][step=5] Starting CheckpointManager Save Finalize thread=save_finalize I0422 17:05:30.859856 133916442736384 async_checkpointer.py:265] [process=3][thread=save_finalize] Waiting for background save thread=async_save. I0422 17:05:30.860033 134043235866432 standard_logger.py:34] {'step': 5, 'event_type': 'save', 'directory': 'gs://lance-maxtext/linen_ckpt_xpk_feat_nnx_set_defaults_true_20260422_154647/linen_xpk_feat_nnx_set_defaults_true_20260422_154647_15_ocdbt_false/checkpoints', 'reached_preemption': False, 'preemption_received_at': None, 'synchronous': False, 'wait_for_prev_start_time': 1776877515.516528, 'wait_for_prev_duration_secs': 14.744594097137451, 'time_between_consecutive_saves_sec': None, 'checkpointer_blocking_start_time': 1776877530.263091, 'checkpointer_blocking_duration_secs': 0.5815553665161133, 'get_old_steps_start_time': 1776877530.8446648, 'get_old_steps_duration_secs': 2.3603439331054688e-05, 'checkpoint_manager_blocking_start_time': 1776877515.5143538, 'checkpoint_manager_blocking_duration_secs': 15.345645427703857} I0422 17:05:30.860143 134043235866432 checkpointing.py:409] Started an asynchronous checkpoint save for step 5 I0422 17:05:30.860746 134043235866432 metric_logger.py:185] completed step: 4, seconds: 0.153, TFLOP/s/device: 89.080, Tokens/s/device: 13427.131, total_weights: 65536, loss: 9.460 I0422 17:05:30.866199 134043235866432 metric_logger.py:185] completed step: 5, seconds: 0.147, TFLOP/s/device: 92.252, Tokens/s/device: 13905.297, total_weights: 65536, loss: 8.959 I0422 17:05:50.327725 134043235866432 metric_logger.py:185] completed step: 6, seconds: 15.641, TFLOP/s/device: 0.869, Tokens/s/device: 130.936, total_weights: 65536, loss: 8.469 I0422 17:05:50.474983 134043235866432 metric_logger.py:185] completed step: 7, seconds: 19.315, TFLOP/s/device: 0.703, Tokens/s/device: 106.032, total_weights: 65536, loss: 8.003 I0422 17:05:50.622244 134043235866432 metric_logger.py:185] completed step: 8, seconds: 0.152, TFLOP/s/device: 89.101, Tokens/s/device: 13430.213, total_weights: 65536, loss: 7.572 I0422 17:05:50.769333 134043235866432 checkpointing.py:794] Waiting for step 9 to finish before checkpoint... I0422 17:05:50.770010 134043235866432 checkpointing.py:798] Waited 0.0006968975067138672 seconds for step 9 to finish before starting checkpointing. I0422 17:05:50.771977 134043235866432 checkpoint_manager.py:2024] [process=3][thread=MainThread][step=5][wait_until_finished] Waiting for Save Finalize thread (save_finalize) to complete. I0422 17:05:56.371113 133920200832768 array_metadata_store.py:203] [process=3][thread=array_type_handler] Wrote 39 array_metadata.ArrayMetadata to gs://lance-maxtext/linen_ckpt_xpk_feat_nnx_set_defaults_true_20260422_154647/linen_xpk_feat_nnx_set_defaults_true_20260422_154647_15_ocdbt_false/checkpoints/5/items/array_metadatas/process_3 I0422 17:06:01.587770 133911070971648 base_pytree_checkpoint_handler.py:129] [process=3] /jax/checkpoint/write/gbytes_per_sec: 50.886 MiB/s (total gbytes: 1.5 GiB) (time elapsed: 31.041758060455322 s) (per-host) I0422 17:06:01.587888 133911070971648 async_checkpointer.py:90] [process=3][thread=async_save] 3 Handler Commit operations completed. Time taken: 30.743385s. I0422 17:06:07.169784 133911070971648 async_checkpointer.py:144] [process=3][thread=async_save] Background save thread done. Time taken: 36.325266s. I0422 17:06:07.170099 133916442736384 async_checkpointer.py:273] [process=3][thread=save_finalize] Done with waiting for background save thread=async_save. I0422 17:06:07.170213 133916442736384 async_checkpointer.py:283] [process=3][thread=save_finalize] No errors found in background save thread=async_save. I0422 17:06:07.170259 133916442736384 checkpoint_manager.py:2133] [process=3][thread=save_finalize][step=5] CheckpointManager Save Finalize is syncing with other hosts... I0422 17:06:07.172672 133916442736384 checkpoint_manager.py:2142] [process=3][thread=save_finalize][step=5] CheckpointManager Save Finalize is done on all hosts. I0422 17:06:07.172856 134043235866432 checkpoint_manager.py:2036] [process=3][thread=MainThread][step=5][wait_until_finished] Done waiting for Save Finalize thread (save_finalize) running at step=5. W0422 17:06:07.172986 134043235866432 checkpoint_manager.py:1458] Waiting for previous save to complete took 16.401022 seconds. If this number is high, consider checkpointing less frequently. I0422 17:06:07.174380 134043235866432 checkpoint_manager.py:1518] [process=3] Saving checkpoint at step 9 I0422 17:06:07.177699 134043235866432 async_checkpointer.py:452] [process=3] Started async saving checkpoint to gs://lance-maxtext/linen_ckpt_xpk_feat_nnx_set_defaults_true_20260422_154647/linen_xpk_feat_nnx_set_defaults_true_20260422_154647_15_ocdbt_false/checkpoints/9. I0422 17:06:07.721284 134043235866432 jax_array_handlers.py:358] Scheduling D2H of 39 prioritized jax.Array. I0422 17:06:07.721378 134043235866432 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 17:06:07.762555 134043235866432 base_pytree_checkpoint_handler.py:153] [process=3][thread=MainThread] Initiated "orbax.checkpoint._src.serialization.jax_array_handlers.ArrayHandler".serialize. Time taken: 0.042168s I0422 17:06:07.762719 134043235866432 base_pytree_checkpoint_handler.py:129] [process=3] /jax/checkpoint/write/blocking_gbytes_per_sec: 5.100 GiB/s (total gbytes: 1.5 GiB) (time elapsed: 0.3024570941925049 s) (per-host) I0422 17:06:07.762770 134043235866432 base_pytree_checkpoint_handler.py:737] [process=3][thread=MainThread] Initiated Pytree async_save. Time taken: 0.302524s (batch_requests_ready=0.258292s, total_serialization_initiated=0.044160s, others=0.000072s) I0422 17:06:07.762855 134043235866432 composite_checkpoint_handler.py:715] [process=3][thread=MainThread] Initiated CompositeCheckpointHandler.async_save. Time taken: 0.306640s (all_items=0.000009s, per_item={'items': '0.00000930'}, temp_paths=0.306630) I0422 17:06:07.763690 133912172947200 async_checkpointer.py:79] [process=3][thread=async_save] Background save thread started. I0422 17:06:07.763853 134043235866432 async_checkpointer.py:561] Finished blocking save. Time taken: 0.589405s. Continuing background save to gs://lance-maxtext/linen_ckpt_xpk_feat_nnx_set_defaults_true_20260422_154647/linen_xpk_feat_nnx_set_defaults_true_20260422_154647_15_ocdbt_false/checkpoints/9. I0422 17:06:08.195481 134043235866432 checkpoint_manager.py:1566] [process=3][thread=MainThread][step=9] Starting CheckpointManager Save Finalize thread=save_finalize I0422 17:06:08.195879 133916442736384 async_checkpointer.py:265] [process=3][thread=save_finalize] Waiting for background save thread=async_save. I0422 17:06:08.196076 134043235866432 standard_logger.py:34] {'step': 9, 'event_type': 'save', 'directory': 'gs://lance-maxtext/linen_ckpt_xpk_feat_nnx_set_defaults_true_20260422_154647/linen_xpk_feat_nnx_set_defaults_true_20260422_154647_15_ocdbt_false/checkpoints', 'reached_preemption': False, 'preemption_received_at': None, 'synchronous': False, 'wait_for_prev_start_time': 1776877550.7719476, 'wait_for_prev_duration_secs': 16.40102195739746, 'time_between_consecutive_saves_sec': 20.51112413406372, 'checkpointer_blocking_start_time': 1776877567.1744182, 'checkpointer_blocking_duration_secs': 0.5895512104034424, 'get_old_steps_start_time': 1776877567.7639825, 'get_old_steps_duration_secs': 1.9550323486328125e-05, 'checkpoint_manager_blocking_start_time': 1776877550.7702312, 'checkpoint_manager_blocking_duration_secs': 17.425811290740967} I0422 17:06:08.196185 134043235866432 checkpointing.py:409] Started an asynchronous checkpoint save for step 9 I0422 17:06:08.196228 134043235866432 checkpoint_manager.py:2024] [process=3][thread=MainThread][step=9][wait_until_finished] Waiting for Save Finalize thread (save_finalize) to complete. I0422 17:06:12.919365 133920200832768 array_metadata_store.py:203] [process=3][thread=array_type_handler] Wrote 39 array_metadata.ArrayMetadata to gs://lance-maxtext/linen_ckpt_xpk_feat_nnx_set_defaults_true_20260422_154647/linen_xpk_feat_nnx_set_defaults_true_20260422_154647_15_ocdbt_false/checkpoints/9/items/array_metadatas/process_3 I0422 17:06:18.359199 133912172947200 base_pytree_checkpoint_handler.py:129] [process=3] /jax/checkpoint/write/gbytes_per_sec: 144.932 MiB/s (total gbytes: 1.5 GiB) (time elapsed: 10.898897647857666 s) (per-host) I0422 17:06:18.359319 133912172947200 async_checkpointer.py:90] [process=3][thread=async_save] 3 Handler Commit operations completed. Time taken: 10.595510s. I0422 17:06:24.404878 133912172947200 async_checkpointer.py:144] [process=3][thread=async_save] Background save thread done. Time taken: 16.641056s. I0422 17:06:24.405133 133916442736384 async_checkpointer.py:273] [process=3][thread=save_finalize] Done with waiting for background save thread=async_save. I0422 17:06:24.405203 133916442736384 async_checkpointer.py:283] [process=3][thread=save_finalize] No errors found in background save thread=async_save. I0422 17:06:24.405242 133916442736384 checkpoint_manager.py:2133] [process=3][thread=save_finalize][step=9] CheckpointManager Save Finalize is syncing with other hosts... I0422 17:06:24.407677 133916442736384 checkpoint_manager.py:2142] [process=3][thread=save_finalize][step=9] CheckpointManager Save Finalize is done on all hosts. I0422 17:06:24.407824 134043235866432 checkpoint_manager.py:2036] [process=3][thread=MainThread][step=9][wait_until_finished] Done waiting for Save Finalize thread (save_finalize) running at step=9. I0422 17:06:24.408014 134043235866432 checkpoint_manager.py:2013] [process=3][thread=MainThread][wait_until_finished] No Save Finalize thread to wait for. Returning. I0422 17:06:24.408737 134043235866432 metric_logger.py:185] completed step: 9, seconds: 0.147, TFLOP/s/device: 92.185, Tokens/s/device: 13895.108, total_weights: 65536, loss: 7.181 Per train step: Total TFLOPs: 13.59 split as 93.93% learnable weight flops and 6.07% attention flops XPK End: Wed Apr 22 17:06:37 UTC 2026 EXIT_CODE=0