main (574ad3fb9) vs feat/nnx-post-train-fixes (574ad3fb9)| Metric | main 574ad3fb9 | feat/nnx-post-train-fixes 574ad3fb9 | Diff (feat/nnx-post-train-fixes − main) |
|---|---|---|---|
| Parameters | 1.104 billion | 1.104 billion | — |
| Final loss | 5.3910 | 5.3910 | 0 |
| TFLOP/s | 92.177 | 92.184 | +0.007 |
| Tok/s | 13893.9 | 13895.0 | +1.131 |
| Avg s/step | 8.525 | 8.563 | +0.038 |
| 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: Mon Apr 20 07:31:07 UTC 2026 PyTorch was not found. Models won't be available and only tokenizers, configuration and file/data utilities can be used. 2026-04-20 07:31:31.755599: E external/local_xla/xla/stream_executor/cuda/cuda_platform.cc:51] failed call to cuInit: INTERNAL: CUDA error: Failed call to cuInit: UNKNOWN ERROR (303) I0420 07:31:32.291649 135088473790272 max_utils.py:273] Attempting to initialize the jax distributed system... I0420 07:31:41.333622 135088473790272 distributed.py:149] Starting JAX distributed service on [::]:8482 I0420 07:31:41.335987 135088473790272 distributed.py:172] Connecting to JAX distributed service on mt-16-zarr3-false-rs-cszoz-slice-job-0-0.mt-16-zarr3-false-rs-cszoz:8482 I0420 07:31:42.668657 135088473790272 max_utils.py:284] Jax distributed system initialized! I0420 07:31:47.999677 135088473790272 max_utils.py:800] System Information: Jax Version: 0.9.2 I0420 07:31:47.999779 135088473790272 max_utils.py:801] System Information: Jaxlib Version: 0.9.2 I0420 07:31:47.999823 135088473790272 max_utils.py:802] System Information: Jax Backend: PJRT C API TFRT TPU v6 lite Built on Mar 4 2026 11:32:08 (1772652728) cl/878335365 I0420 07:31:47.999857 135088473790272 train_utils.py:348] WARNING: Sequence packing is essentially ignored for synthetic data. Please use a real dataset to use sequence packing. I0420 07:31:48.707550 135088473790272 maxtext_utils.py:1551] Num_devices: 32, shape (1, 1, 1, 32, 1, 1, 1, 1, 1, 1, 1, 1, 1) I0420 07:31:48.707822 135088473790272 checkpointing.py:677] Setting up checkpoint logger... I0420 07:31:48.707875 135088473790272 checkpointing.py:233] Creating checkpoint manager with ocdbt=True and zarr3=False I0420 07:31:48.707924 135088473790272 pytree_checkpoint_handler.py:592] save_device_host_concurrent_bytes=None I0420 07:31:48.708254 135088473790272 base_pytree_checkpoint_handler.py:441] Created BasePyTreeCheckpointHandler: use_ocdbt=True, use_zarr3=False, pytree_metadata_options=PyTreeMetadataOptions(support_rich_types=False), array_metadata_store=<orbax.checkpoint._src.metadata.array_metadata_store.Store object at 0x7adc04e4f1d0>, enable_pinned_host_transfer=False, save_concurrent_bytes: 96000000000 (89.4 GiB), restore_concurrent_bytes: 96000000000 (89.4 GiB) I0420 07:31:49.930782 135088473790272 checkpointing.py:265] Enabling policy for fixed interval checkpointing. I0420 07:31:49.931012 135088473790272 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 0x7ac8784fe930>}, handler_registry=None I0420 07:31:49.931242 135088473790272 composite_checkpoint_handler.py:237] Deferred registration for item: "items". Adding handler `<orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeCheckpointHandler object at 0x7ac8784fe930>` for item "items" and save args `<class 'orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeSaveArgs'>` and restore args `<class 'orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeRestoreArgs'>` to `_handler_registry`. I0420 07:31:49.931290 135088473790272 composite_checkpoint_handler.py:237] Deferred registration for item: "metrics". Adding handler `<orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonCheckpointHandler object at 0x7ac87839f200>` for item "metrics" and save args `<class 'orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonSaveArgs'>` and restore args `<class 'orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonRestoreArgs'>` to `_handler_registry`. I0420 07:31:49.931324 135088473790272 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 0x7ac8784fe930>, ('items', <class 'orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeRestoreArgs'>): <orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeCheckpointHandler object at 0x7ac8784fe930>, ('metrics', <class 'orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonSaveArgs'>): <orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonCheckpointHandler object at 0x7ac87839f200>, ('metrics', <class 'orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonRestoreArgs'>): <orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonCheckpointHandler object at 0x7ac87839f200>}). I0420 07:31:49.931657 135088473790272 abstract_checkpointer.py:35] orbax-checkpoint version: 0.11.34 I0420 07:31:49.931727 135088473790272 async_checkpointer.py:192] [process=4][thread=MainThread] Using barrier_sync_fn: <function get_barrier_sync_fn.<locals>._fn at 0x7ac8601c3880> timeout: 1200 secs and primary_host=0 for async checkpoint writes I0420 07:31:51.477575 134969716356864 checkpoint.py:228] Read Metadata={'item_handlers': {'items': 'orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeCheckpointHandler'}, 'metrics': {}, 'performance_metrics': {}, 'init_timestamp_nsecs': 1776670094799312577, 'commit_timestamp_nsecs': 1776670142680280075, 'custom_metadata': {}} from gs://lance-maxtext/linen_ckpt_xpk_main_20260420_060028/linen_xpk_main_20260420_060028_16_zarr3_false/checkpoints/5/_CHECKPOINT_METADATA I0420 07:31:51.726522 134969716356864 checkpoint_manager.py:1871] Missing metrics for step 5 I0420 07:31:52.153673 135088473790272 checkpoint_manager.py:1812] Found 3 checkpoint steps in gs://lance-maxtext/linen_ckpt_xpk_main_20260420_060028/linen_xpk_main_20260420_060028_16_zarr3_false/checkpoints I0420 07:31:52.600018 135088473790272 checkpoint_manager.py:929] [process=4][thread=MainThread] CheckpointManager created, primary_host=0, CheckpointManagerOptions=CheckpointManagerOptions(save_interval_steps=1, max_to_keep=None, keep_time_interval=None, keep_period=None, should_keep_fn=None, best_fn=None, best_mode='max', keep_checkpoints_without_metrics=True, step_prefix=None, step_format_fixed_length=None, step_name_format=None, create=True, cleanup_tmp_directories=False, save_on_steps=frozenset(), single_host_load_and_broadcast=False, todelete_subdir=None, todelete_full_path=None, enable_background_delete=False, read_only=False, enable_async_checkpointing=True, async_options=None, multiprocessing_options=MultiprocessingOptions(primary_host=0, active_processes=None, barrier_sync_key_prefix=None), should_save_fn=None, file_options=FileOptions(path_permission_mode=None), save_root_metadata=True, temporary_path_class=None, save_decision_policy=FixedIntervalPolicy(interval=5), preservation_policy=LatestN(n=None), prevent_write_metrics=False, enable_should_save_is_saving_in_progress_check=True, enable_per_process_directory_creation=False, lightweight_initialize=False), root_directory=gs://lance-maxtext/linen_ckpt_xpk_main_20260420_060028/linen_xpk_main_20260420_060028_16_zarr3_false/checkpoints: <orbax.checkpoint.checkpoint_manager.CheckpointManager object at 0x7ac87871db50> I0420 07:31:52.600193 135088473790272 checkpointing.py:301] Checkpoint manager created! I0420 07:31:53.539827 135088473790272 nnx_wrappers.py:437] Unknown Logical: bfloat16[32,2048,2048]...................................... ('activation_batch', 'activation_norm_length', 'activation_embed'). I0420 07:31:53.539937 135088473790272 nnx_wrappers.py:437] Unknown Physical: bfloat16[32,2048,2048]...................................... ('fsdp', None, None). I0420 07:31:53.921438 135088473790272 attentions.py:1088] attentions/inputs_q Logical: bfloat16[32,2048,2048]...................................... ('activation_batch', 'activation_attn_length', 'activation_attn_embed'). I0420 07:31:53.921530 135088473790272 attentions.py:1088] attentions/inputs_q Physical: bfloat16[32,2048,2048]...................................... ('fsdp', None, None). I0420 07:31:53.937797 135088473790272 attentions.py:1089] attentions/inputs_kv Logical: bfloat16[32,2048,2048]...................................... ('activation_batch', 'activation_attn_length', 'activation_attn_embed'). I0420 07:31:53.937856 135088473790272 attentions.py:1089] attentions/inputs_kv Physical: bfloat16[32,2048,2048]...................................... ('fsdp', None, None). I0420 07:31:53.962330 135088473790272 attentions.py:1154] attentions/query Logical: bfloat16[32,2048,16,128].................................... ('activation_kv_batch', 'activation_attn_length', 'activation_kv_heads', 'activation_kv_head_dim'). I0420 07:31:53.962407 135088473790272 attentions.py:1154] attentions/query Physical: bfloat16[32,2048,16,128].................................... ('fsdp', None, None, None). I0420 07:31:53.978829 135088473790272 attentions.py:1155] attentions/key Logical: bfloat16[32,2048,16,128].................................... ('activation_kv_batch', 'activation_attn_length', 'activation_kv_heads', 'activation_kv_head_dim'). I0420 07:31:53.978888 135088473790272 attentions.py:1155] attentions/key Physical: bfloat16[32,2048,16,128].................................... ('fsdp', None, None, None). I0420 07:31:53.995278 135088473790272 attentions.py:1156] attentions/value Logical: bfloat16[32,2048,16,128].................................... ('activation_kv_batch', 'activation_attn_length', 'activation_kv_heads', 'activation_kv_head_dim'). I0420 07:31:53.995350 135088473790272 attentions.py:1156] attentions/value Physical: bfloat16[32,2048,16,128].................................... ('fsdp', None, None, None). I0420 07:31:54.020385 135088473790272 attentions.py:1197] attentions/out Logical: bfloat16[32,2048,16,128].................................... ('activation_batch', 'activation_attn_length', 'activation_heads', 'activation_kv'). I0420 07:31:54.020460 135088473790272 attentions.py:1197] attentions/out Physical: bfloat16[32,2048,16,128].................................... ('fsdp', None, None, None). I0420 07:31:54.041770 135088473790272 linears.py:525] linears/x Logical: bfloat16[32,2048,7168]...................................... ('activation_batch', 'activation_length', 'activation_mlp'). I0420 07:31:54.041832 135088473790272 linears.py:525] linears/x Physical: bfloat16[32,2048,7168]...................................... ('fsdp', None, None). I0420 07:31:54.260373 135088473790272 checkpointing.py:577] checkpoint manager exists so trying to load this run's existing checkpoint I0420 07:31:54.260480 135088473790272 checkpointing.py:581] restoring from this run's directory step 9 I0420 07:31:54.518152 135088473790272 event_tracking.py:70] [process=4] [sync] Started load checkpoint @ gs://lance-maxtext/linen_ckpt_xpk_main_20260420_060028/linen_xpk_main_20260420_060028_16_zarr3_false/checkpoints/9. I0420 07:31:54.765818 135088473790272 checkpointer.py:307] Restoring checkpoint from gs://lance-maxtext/linen_ckpt_xpk_main_20260420_060028/linen_xpk_main_20260420_060028_16_zarr3_false/checkpoints/9. I0420 07:31:54.766241 135088473790272 event_tracking.py:125] [process=4] [sync] Finished blocking load in 0.25 seconds. Continuing load @ gs://lance-maxtext/linen_ckpt_xpk_main_20260420_060028/linen_xpk_main_20260420_060028_16_zarr3_false/checkpoints/9. I0420 07:31:56.157637 1414 google_auth_provider.cc:181] Running on GCE, using service account 562977990677-compute@developer.gserviceaccount.com I0420 07:32:03.464491 135088473790272 jax_array_handlers.py:843] [process=4] /jax/orbax/read/worker/io/requested throughput: 216.100 MiB/s (total gbytes: 1.5 GiB) (time elapsed: 7.312694072723389 s) (per-host) I0420 07:32:03.469974 135088473790272 base_pytree_checkpoint_handler.py:130] [process=4] /jax/checkpoint/read/gbytes_per_sec: 205.033 MiB/s (total gbytes: 1.5 GiB) (time elapsed: 7.718711853027344 s) (per-host) I0420 07:32:04.311449 135088473790272 event_tracking.py:138] [process=4] [sync] Finished load in 9.79 seconds @ gs://lance-maxtext/linen_ckpt_xpk_main_20260420_060028/linen_xpk_main_20260420_060028_16_zarr3_false/checkpoints/9 I0420 07:32:04.311599 135088473790272 standard_logger.py:34] {'step': 9, 'event_type': 'restore', 'directory': 'gs://lance-maxtext/linen_ckpt_xpk_main_20260420_060028/linen_xpk_main_20260420_060028_16_zarr3_false/checkpoints', 'checkpointer_start_time': 1776670314.5180545, 'checkpointer_duration_secs': 9.793514013290405, 'checkpoint_manager_start_time': 1776670314.2607217, 'checkpoint_manager_duration_secs': 10.050849199295044} fsdp: 32 I0420 07:32:04.725955 135088473790272 maxtext_utils.py:1654] params/params/decoder/decoder_norm/scale Shape: float32[2048] Logical: P('norm',) Physical: (None,) I0420 07:32:04.726076 135088473790272 maxtext_utils.py:1654] params/params/decoder/layers/mlp/wi_0/kernel Shape: float32[2048,16,7168] Logical: P('embed', 'layers', 'mlp') Physical: ('fsdp', None, None) I0420 07:32:04.726130 135088473790272 maxtext_utils.py:1654] params/params/decoder/layers/mlp/wi_1/kernel Shape: float32[2048,16,7168] Logical: P('embed', 'layers', 'mlp') Physical: ('fsdp', None, None) I0420 07:32:04.726187 135088473790272 maxtext_utils.py:1654] params/params/decoder/layers/mlp/wo/kernel Shape: float32[7168,16,2048] Logical: P('mlp', 'layers', 'embed') Physical: (None, None, 'fsdp') I0420 07:32:04.726238 135088473790272 maxtext_utils.py:1654] params/params/decoder/layers/post_self_attention_layer_norm/scale Shape: float32[2048,16] Logical: P('norm', 'layers') Physical: (None, None) I0420 07:32:04.726276 135088473790272 maxtext_utils.py:1654] params/params/decoder/layers/pre_self_attention_layer_norm/scale Shape: float32[2048,16] Logical: P('norm', 'layers') Physical: (None, None) I0420 07:32:04.726325 135088473790272 maxtext_utils.py:1654] params/params/decoder/layers/self_attention/key/kernel Shape: float32[2048,16,16,128] Logical: P('embed', 'layers', 'kv_heads', 'kv_head_dim') Physical: ('fsdp', None, None, None) I0420 07:32:04.726491 135088473790272 maxtext_utils.py:1654] params/params/decoder/layers/self_attention/out/kernel Shape: float32[16,16,128,2048] Logical: P('heads', 'layers', 'kv', 'embed') Physical: (None, None, None, 'fsdp') I0420 07:32:04.726534 135088473790272 maxtext_utils.py:1654] params/params/decoder/layers/self_attention/query/kernel Shape: float32[2048,16,16,128] Logical: P('embed', 'layers', 'q_heads', 'kv') Physical: ('fsdp', None, None, None) I0420 07:32:04.726571 135088473790272 maxtext_utils.py:1654] params/params/decoder/layers/self_attention/value/kernel Shape: float32[2048,16,16,128] Logical: P('embed', 'layers', 'kv_heads', 'kv_head_dim') Physical: ('fsdp', None, None, None) I0420 07:32:04.726618 135088473790272 maxtext_utils.py:1654] params/params/decoder/logits_dense/kernel Shape: float32[2048,32000] Logical: P('embed_vocab', 'vocab') Physical: ('fsdp', None) I0420 07:32:04.726664 135088473790272 maxtext_utils.py:1654] params/params/token_embedder/embedding Shape: float32[32000,2048] Logical: P('vocab', 'embed_vocab') Physical: (None, 'fsdp') I0420 07:32:05.206961 135088473790272 train.py:155] train/xent Logical: float32[32,2048]............................................ ('activation_embed_and_logits_batch', 'activation_length'). I0420 07:32:05.207057 135088473790272 train.py:155] train/xent Physical: float32[32,2048]............................................ ('fsdp', None). I0420 07:32:05.222791 135088473790272 train.py:162] train/z_loss Logical: float32[32,2048]............................................ ('activation_embed_and_logits_batch', 'activation_length'). I0420 07:32:05.222849 135088473790272 train.py:162] train/z_loss Physical: float32[32,2048]............................................ ('fsdp', None). I0420 07:32:16.392876 135088473790272 max_utils.py:791] Total memory size: 1.5 GB, Output size: 0.4 GB, Temp size: 1.1 GB, Argument size: 0.4 GB, Host temp size: 0.0 GB. I0420 07:32:16.393684 135088473790272 metric_logger.py:301] number parameters: 1.104 billion I0420 07:32:27.775606 135088473790272 checkpointing.py:772] Waiting for step 10 to finish before checkpoint... I0420 07:32:27.927739 135088473790272 checkpointing.py:776] Waited 0.15210986137390137 seconds for step 10 to finish before starting checkpointing. I0420 07:32:27.930422 135088473790272 checkpoint_manager.py:2009] [process=4][thread=MainThread][wait_until_finished] No Save Finalize thread to wait for. Returning. I0420 07:32:27.932369 135088473790272 checkpoint_manager.py:1512] [process=4] Saving checkpoint at step 10 I0420 07:32:27.934048 135088473790272 event_tracking.py:70] [process=4] [async] Started save checkpoint @ gs://lance-maxtext/linen_ckpt_xpk_main_20260420_060028/linen_xpk_main_20260420_060028_16_zarr3_false/checkpoints/10. I0420 07:32:28.710674 135088473790272 signaling_client.py:364] Using JaxDistributedSignalingClient I0420 07:32:28.711577 135088473790272 jax_array_handlers.py:360] Scheduling D2H of 39 prioritized jax.Array. I0420 07:32:28.711634 135088473790272 replica_slices.py:424] Transferring arrays to host memory with options: use_replica_parallel=True, min_slice_bytes_for_replica_parallel=None, max_replicas_for_replica_parallel=None, enable_pinned_host_transfer=False I0420 07:32:28.989174 135088473790272 base_pytree_checkpoint_handler.py:154] [process=4][thread=MainThread] Initiated "orbax.checkpoint._src.serialization.jax_array_handlers.ArrayHandler".serialize. Time taken: 0.278582s I0420 07:32:28.989355 135088473790272 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: 0.28261613845825195 s) (per-host) I0420 07:32:28.989412 135088473790272 base_pytree_checkpoint_handler.py:768] [process=4][thread=MainThread] Initiated Pytree async_save. Time taken: 0.282696s (batch_requests_ready=0.002156s, total_serialization_initiated=0.280452s, others=0.000088s) I0420 07:32:28.989500 135088473790272 composite_checkpoint_handler.py:715] [process=4][thread=MainThread] Initiated CompositeCheckpointHandler.async_save. Time taken: 0.286880s (all_items=0.000017s, per_item={'items': '0.00001693'}, temp_paths=0.286864) I0420 07:32:28.990259 135088473790272 event_tracking.py:125] [process=4] [async] Finished blocking save in 1.06 seconds. Continuing save @ gs://lance-maxtext/linen_ckpt_xpk_main_20260420_060028/linen_xpk_main_20260420_060028_16_zarr3_false/checkpoints/10. I0420 07:32:28.990622 134967497578240 async_checkpointer.py:76] [process=4][thread=async_save] Background save thread started. Deadline for this save operation is 2026-04-20 07:52:28.990583 I0420 07:32:29.000119 135088473790272 checkpoint_manager.py:1560] [process=4][thread=MainThread][step=10] Starting CheckpointManager Save Finalize thread=save_finalize I0420 07:32:29.000432 134969103984384 async_checkpointer.py:280] [process=4][thread=save_finalize] Waiting for background save thread=async_save. I0420 07:32:29.000610 135088473790272 standard_logger.py:34] {'step': 10, 'event_type': 'save', 'directory': 'gs://lance-maxtext/linen_ckpt_xpk_main_20260420_060028/linen_xpk_main_20260420_060028_16_zarr3_false/checkpoints', 'reached_preemption': False, 'preemption_received_at': None, 'synchronous': False, 'wait_for_prev_start_time': 1776670347.9304047, 'wait_for_prev_duration_secs': 6.031990051269531e-05, 'time_between_consecutive_saves_sec': None, 'checkpointer_blocking_start_time': 1776670347.9324074, 'checkpointer_blocking_duration_secs': 1.0583722591400146, 'get_old_steps_start_time': 1776670348.9908028, 'get_old_steps_duration_secs': 3.123283386230469e-05, 'checkpoint_manager_blocking_start_time': 1776670347.928261, 'checkpoint_manager_blocking_duration_secs': 1.0723142623901367} I0420 07:32:29.000714 135088473790272 checkpointing.py:408] Started an asynchronous checkpoint save for step 10 I0420 07:32:29.000766 135088473790272 max_utils.py:750] Memstats: After params initialized: I0420 07:32:29.000817 135088473790272 max_utils.py:756] Using (GB) 0.43 / 31.25 (1.376000%) on TPU_16(process=4,(0,4,0,0)) I0420 07:32:29.000851 135088473790272 max_utils.py:756] Using (GB) 0.43 / 31.25 (1.376000%) on TPU_17(process=4,(1,4,0,0)) I0420 07:32:29.000885 135088473790272 max_utils.py:756] Using (GB) 0.43 / 31.25 (1.376000%) on TPU_20(process=4,(0,5,0,0)) I0420 07:32:29.000910 135088473790272 max_utils.py:756] Using (GB) 0.43 / 31.25 (1.376000%) on TPU_21(process=4,(1,5,0,0)) I0420 07:32:29.312306 135088473790272 metric_logger.py:196] completed step: 10, seconds: 11.382, TFLOP/s/device: 1.194, Tokens/s/device: 179.936, total_weights: 65536, loss: 6.833, lm_loss: 6.833, perplexity: 927.869 I0420 07:32:29.481695 135088473790272 metric_logger.py:196] completed step: 11, seconds: 1.535, TFLOP/s/device: 8.851, Tokens/s/device: 1334.053, total_weights: 65536, loss: 6.527, lm_loss: 6.527, perplexity: 683.659 I0420 07:32:29.629592 135088473790272 metric_logger.py:196] completed step: 12, seconds: 0.022, TFLOP/s/device: 611.317, Tokens/s/device: 92144.335, total_weights: 65536, loss: 6.265, lm_loss: 6.265, perplexity: 526.068 I0420 07:32:29.777244 135088473790272 metric_logger.py:196] completed step: 13, seconds: 0.153, TFLOP/s/device: 88.643, Tokens/s/device: 13361.169, total_weights: 65536, loss: 6.045, lm_loss: 6.045, perplexity: 422.001 I0420 07:32:29.782140 135088473790272 checkpointing.py:772] Waiting for step 15 to finish before checkpoint... I0420 07:32:30.071595 135088473790272 checkpointing.py:776] Waited 0.2894294261932373 seconds for step 15 to finish before starting checkpointing. I0420 07:32:30.074511 135088473790272 checkpoint_manager.py:2020] [process=4][thread=MainThread][step=10][wait_until_finished] Waiting for Save Finalize thread (save_finalize) to complete. I0420 07:32:34.717621 134968017663744 array_metadata_store.py:203] [process=4][thread=array_type_handler] Wrote 39 array_metadata.ArrayMetadata to gs://lance-maxtext/linen_ckpt_xpk_main_20260420_060028/linen_xpk_main_20260420_060028_16_zarr3_false/checkpoints/10/items/array_metadatas/process_4 I0420 07:33:11.767464 134967497578240 base_pytree_checkpoint_handler.py:130] [process=4] /jax/orbax/write/gbytes_per_sec: 36.683 MiB/s (total gbytes: 1.5 GiB) (time elapsed: 43.06070280075073 s) (per-host) I0420 07:33:11.767604 134967497578240 async_checkpointer.py:90] [process=4][thread=async_save] 3 Handler Commit operations completed. Time taken: 42.776868s. I0420 07:33:20.058192 134967497578240 async_checkpointer.py:160] [process=4][thread=async_save] Background save thread done. Time taken: 51.067439s. I0420 07:33:20.058499 134969103984384 async_checkpointer.py:288] [process=4][thread=save_finalize] Done with waiting for background save thread=async_save. I0420 07:33:20.058614 134969103984384 async_checkpointer.py:298] [process=4][thread=save_finalize] No errors found in background save thread=async_save. I0420 07:33:20.058661 134969103984384 checkpoint_manager.py:2137] [process=4][thread=save_finalize][step=10] CheckpointManager Save Finalize is syncing with other hosts... I0420 07:33:20.060370 134969103984384 checkpoint_manager.py:2146] [process=4][thread=save_finalize][step=10] CheckpointManager Save Finalize is done on all hosts. I0420 07:33:20.060560 135088473790272 checkpoint_manager.py:2032] [process=4][thread=MainThread][step=10][wait_until_finished] Done waiting for Save Finalize thread (save_finalize) running at step=10. W0420 07:33:20.060694 135088473790272 checkpoint_manager.py:1452] Waiting for previous save to complete took 49.986185 seconds. If this number is high, consider checkpointing less frequently. I0420 07:33:20.062741 135088473790272 checkpoint_manager.py:1512] [process=4] Saving checkpoint at step 15 I0420 07:33:20.065031 135088473790272 event_tracking.py:70] [process=4] [async] Started save checkpoint @ gs://lance-maxtext/linen_ckpt_xpk_main_20260420_060028/linen_xpk_main_20260420_060028_16_zarr3_false/checkpoints/15. I0420 07:33:20.372605 135088473790272 jax_array_handlers.py:360] Scheduling D2H of 39 prioritized jax.Array. I0420 07:33:20.372697 135088473790272 replica_slices.py:424] Transferring arrays to host memory with options: use_replica_parallel=True, min_slice_bytes_for_replica_parallel=None, max_replicas_for_replica_parallel=None, enable_pinned_host_transfer=False I0420 07:33:20.398423 135088473790272 base_pytree_checkpoint_handler.py:154] [process=4][thread=MainThread] Initiated "orbax.checkpoint._src.serialization.jax_array_handlers.ArrayHandler".serialize. Time taken: 0.026819s I0420 07:33:20.398606 135088473790272 base_pytree_checkpoint_handler.py:130] [process=4] /jax/orbax/write/blocking_gbytes_per_sec: 49.806 GiB/s (total gbytes: 1.5 GiB) (time elapsed: 0.030971527099609375 s) (per-host) I0420 07:33:20.398667 135088473790272 base_pytree_checkpoint_handler.py:768] [process=4][thread=MainThread] Initiated Pytree async_save. Time taken: 0.031047s (batch_requests_ready=0.002489s, total_serialization_initiated=0.028475s, others=0.000084s) I0420 07:33:20.398765 135088473790272 composite_checkpoint_handler.py:715] [process=4][thread=MainThread] Initiated CompositeCheckpointHandler.async_save. Time taken: 0.036235s (all_items=0.000012s, per_item={'items': '0.00001216'}, temp_paths=0.036223) I0420 07:33:20.399487 135088473790272 event_tracking.py:125] [process=4] [async] Finished blocking save in 0.34 seconds. Continuing save @ gs://lance-maxtext/linen_ckpt_xpk_main_20260420_060028/linen_xpk_main_20260420_060028_16_zarr3_false/checkpoints/15. I0420 07:33:20.399811 134969103984384 async_checkpointer.py:76] [process=4][thread=async_save] Background save thread started. Deadline for this save operation is 2026-04-20 07:53:20.399777 I0420 07:33:20.411523 135088473790272 checkpoint_manager.py:1560] [process=4][thread=MainThread][step=15] Starting CheckpointManager Save Finalize thread=save_finalize I0420 07:33:20.411789 134968017663744 async_checkpointer.py:280] [process=4][thread=save_finalize] Waiting for background save thread=async_save. I0420 07:33:20.411924 135088473790272 standard_logger.py:34] {'step': 15, 'event_type': 'save', 'directory': 'gs://lance-maxtext/linen_ckpt_xpk_main_20260420_060028/linen_xpk_main_20260420_060028_16_zarr3_false/checkpoints', 'reached_preemption': False, 'preemption_received_at': None, 'synchronous': False, 'wait_for_prev_start_time': 1776670350.0744796, 'wait_for_prev_duration_secs': 49.98618531227112, 'time_between_consecutive_saves_sec': None, 'checkpointer_blocking_start_time': 1776670400.0627797, 'checkpointer_blocking_duration_secs': 0.3371775150299072, 'get_old_steps_start_time': 1776670400.3999803, 'get_old_steps_duration_secs': 3.2901763916015625e-05, 'checkpoint_manager_blocking_start_time': 1776670350.0721602, 'checkpoint_manager_blocking_duration_secs': 50.339730978012085} I0420 07:33:20.412042 135088473790272 checkpointing.py:408] Started an asynchronous checkpoint save for step 15 I0420 07:33:20.412878 135088473790272 metric_logger.py:196] completed step: 14, seconds: 0.148, TFLOP/s/device: 91.959, Tokens/s/device: 13861.064, total_weights: 65536, loss: 5.864, lm_loss: 5.864, perplexity: 352.193 I0420 07:33:20.419152 135088473790272 metric_logger.py:196] completed step: 15, seconds: 0.148, TFLOP/s/device: 91.819, Tokens/s/device: 13839.989, total_weights: 65536, loss: 5.719, lm_loss: 5.719, perplexity: 304.661 I0420 07:33:26.412936 134964828083968 array_metadata_store.py:203] [process=4][thread=array_type_handler] Wrote 39 array_metadata.ArrayMetadata to gs://lance-maxtext/linen_ckpt_xpk_main_20260420_060028/linen_xpk_main_20260420_060028_16_zarr3_false/checkpoints/15/items/array_metadatas/process_4 I0420 07:33:41.491839 135088473790272 metric_logger.py:196] completed step: 16, seconds: 50.635, TFLOP/s/device: 0.268, Tokens/s/device: 40.446, total_weights: 65536, loss: 5.605, lm_loss: 5.605, perplexity: 271.872 I0420 07:33:41.639386 135088473790272 metric_logger.py:196] completed step: 17, seconds: 20.927, TFLOP/s/device: 0.649, Tokens/s/device: 97.864, total_weights: 65536, loss: 5.517, lm_loss: 5.517, perplexity: 248.975 I0420 07:33:41.786816 135088473790272 metric_logger.py:196] completed step: 18, seconds: 0.152, TFLOP/s/device: 89.316, Tokens/s/device: 13462.701, total_weights: 65536, loss: 5.448, lm_loss: 5.448, perplexity: 232.333 I0420 07:33:41.933727 135088473790272 checkpointing.py:772] Waiting for step 19 to finish before checkpoint... I0420 07:33:41.934455 135088473790272 checkpointing.py:776] Waited 0.0007483959197998047 seconds for step 19 to finish before starting checkpointing. I0420 07:33:41.936549 135088473790272 checkpoint_manager.py:2020] [process=4][thread=MainThread][step=15][wait_until_finished] Waiting for Save Finalize thread (save_finalize) to complete. I0420 07:34:02.617207 134969103984384 base_pytree_checkpoint_handler.py:130] [process=4] /jax/orbax/write/gbytes_per_sec: 37.387 MiB/s (total gbytes: 1.5 GiB) (time elapsed: 42.24954271316528 s) (per-host) I0420 07:34:02.617322 134969103984384 async_checkpointer.py:90] [process=4][thread=async_save] 3 Handler Commit operations completed. Time taken: 42.217398s. I0420 07:34:11.275758 134969103984384 async_checkpointer.py:160] [process=4][thread=async_save] Background save thread done. Time taken: 50.875818s. I0420 07:34:11.276052 134968017663744 async_checkpointer.py:288] [process=4][thread=save_finalize] Done with waiting for background save thread=async_save. I0420 07:34:11.276168 134968017663744 async_checkpointer.py:298] [process=4][thread=save_finalize] No errors found in background save thread=async_save. I0420 07:34:11.276215 134968017663744 checkpoint_manager.py:2137] [process=4][thread=save_finalize][step=15] CheckpointManager Save Finalize is syncing with other hosts... I0420 07:34:11.279026 134968017663744 checkpoint_manager.py:2146] [process=4][thread=save_finalize][step=15] CheckpointManager Save Finalize is done on all hosts. I0420 07:34:11.279207 135088473790272 checkpoint_manager.py:2032] [process=4][thread=MainThread][step=15][wait_until_finished] Done waiting for Save Finalize thread (save_finalize) running at step=15. W0420 07:34:11.279356 135088473790272 checkpoint_manager.py:1452] Waiting for previous save to complete took 29.342798 seconds. If this number is high, consider checkpointing less frequently. I0420 07:34:11.280902 135088473790272 checkpoint_manager.py:1512] [process=4] Saving checkpoint at step 19 I0420 07:34:11.283029 135088473790272 event_tracking.py:70] [process=4] [async] Started save checkpoint @ gs://lance-maxtext/linen_ckpt_xpk_main_20260420_060028/linen_xpk_main_20260420_060028_16_zarr3_false/checkpoints/19. I0420 07:34:11.593026 135088473790272 jax_array_handlers.py:360] Scheduling D2H of 39 prioritized jax.Array. I0420 07:34:11.593116 135088473790272 replica_slices.py:424] Transferring arrays to host memory with options: use_replica_parallel=True, min_slice_bytes_for_replica_parallel=None, max_replicas_for_replica_parallel=None, enable_pinned_host_transfer=False I0420 07:34:11.619109 135088473790272 base_pytree_checkpoint_handler.py:154] [process=4][thread=MainThread] Initiated "orbax.checkpoint._src.serialization.jax_array_handlers.ArrayHandler".serialize. Time taken: 0.027024s I0420 07:34:11.619251 135088473790272 base_pytree_checkpoint_handler.py:130] [process=4] /jax/orbax/write/blocking_gbytes_per_sec: 50.638 GiB/s (total gbytes: 1.5 GiB) (time elapsed: 0.03046250343322754 s) (per-host) I0420 07:34:11.619298 135088473790272 base_pytree_checkpoint_handler.py:768] [process=4][thread=MainThread] Initiated Pytree async_save. Time taken: 0.030519s (batch_requests_ready=0.001645s, total_serialization_initiated=0.028813s, others=0.000061s) I0420 07:34:11.619388 135088473790272 composite_checkpoint_handler.py:715] [process=4][thread=MainThread] Initiated CompositeCheckpointHandler.async_save. Time taken: 0.034687s (all_items=0.000011s, per_item={'items': '0.00001073'}, temp_paths=0.034676) I0420 07:34:11.620068 135088473790272 event_tracking.py:125] [process=4] [async] Finished blocking save in 0.34 seconds. Continuing save @ gs://lance-maxtext/linen_ckpt_xpk_main_20260420_060028/linen_xpk_main_20260420_060028_16_zarr3_false/checkpoints/19. I0420 07:34:11.620414 134968017663744 async_checkpointer.py:76] [process=4][thread=async_save] Background save thread started. Deadline for this save operation is 2026-04-20 07:54:11.620378 I0420 07:34:11.625538 135088473790272 checkpoint_manager.py:1560] [process=4][thread=MainThread][step=19] Starting CheckpointManager Save Finalize thread=save_finalize I0420 07:34:11.625861 134964828083968 async_checkpointer.py:280] [process=4][thread=save_finalize] Waiting for background save thread=async_save. I0420 07:34:11.626024 135088473790272 standard_logger.py:34] {'step': 19, 'event_type': 'save', 'directory': 'gs://lance-maxtext/linen_ckpt_xpk_main_20260420_060028/linen_xpk_main_20260420_060028_16_zarr3_false/checkpoints', 'reached_preemption': False, 'preemption_received_at': None, 'synchronous': False, 'wait_for_prev_start_time': 1776670421.936519, 'wait_for_prev_duration_secs': 29.34279775619507, 'time_between_consecutive_saves_sec': 21.876110076904297, 'checkpointer_blocking_start_time': 1776670451.28094, 'checkpointer_blocking_duration_secs': 0.3396131992340088, 'get_old_steps_start_time': 1776670451.6205702, 'get_old_steps_duration_secs': 2.5987625122070312e-05, 'checkpoint_manager_blocking_start_time': 1776670421.934702, 'checkpoint_manager_blocking_duration_secs': 29.69128966331482} I0420 07:34:11.626140 135088473790272 checkpointing.py:408] Started an asynchronous checkpoint save for step 19 I0420 07:34:11.626181 135088473790272 checkpoint_manager.py:2020] [process=4][thread=MainThread][step=19][wait_until_finished] Waiting for Save Finalize thread (save_finalize) to complete. I0420 07:34:17.577699 134962133055232 array_metadata_store.py:203] [process=4][thread=array_type_handler] Wrote 39 array_metadata.ArrayMetadata to gs://lance-maxtext/linen_ckpt_xpk_main_20260420_060028/linen_xpk_main_20260420_060028_16_zarr3_false/checkpoints/19/items/array_metadatas/process_4 I0420 07:34:54.119187 134968017663744 base_pytree_checkpoint_handler.py:130] [process=4] /jax/orbax/write/gbytes_per_sec: 37.140 MiB/s (total gbytes: 1.5 GiB) (time elapsed: 42.5303795337677 s) (per-host) I0420 07:34:54.119289 134968017663744 async_checkpointer.py:90] [process=4][thread=async_save] 3 Handler Commit operations completed. Time taken: 42.498767s. I0420 07:35:01.566816 134968017663744 async_checkpointer.py:160] [process=4][thread=async_save] Background save thread done. Time taken: 49.946277s. I0420 07:35:01.567121 134964828083968 async_checkpointer.py:288] [process=4][thread=save_finalize] Done with waiting for background save thread=async_save. I0420 07:35:01.567246 134964828083968 async_checkpointer.py:298] [process=4][thread=save_finalize] No errors found in background save thread=async_save. I0420 07:35:01.567288 134964828083968 checkpoint_manager.py:2137] [process=4][thread=save_finalize][step=19] CheckpointManager Save Finalize is syncing with other hosts... I0420 07:35:01.569294 134964828083968 checkpoint_manager.py:2146] [process=4][thread=save_finalize][step=19] CheckpointManager Save Finalize is done on all hosts. I0420 07:35:01.569466 135088473790272 checkpoint_manager.py:2032] [process=4][thread=MainThread][step=19][wait_until_finished] Done waiting for Save Finalize thread (save_finalize) running at step=19. I0420 07:35:01.569614 135088473790272 checkpoint_manager.py:2009] [process=4][thread=MainThread][wait_until_finished] No Save Finalize thread to wait for. Returning. I0420 07:35:01.570604 135088473790272 metric_logger.py:196] completed step: 19, seconds: 0.147, TFLOP/s/device: 92.177, Tokens/s/device: 13893.883, total_weights: 65536, loss: 5.391, lm_loss: 5.391, perplexity: 219.349 Per train step: Total TFLOPs: 13.59 split as 93.93% learnable weight flops and 6.07% attention flops XPK End: Mon Apr 20 07:35:13 UTC 2026 EXIT_CODE=0
XPK Start: Mon Apr 20 17:29:56 UTC 2026 PyTorch was not found. Models won't be available and only tokenizers, configuration and file/data utilities can be used. 2026-04-20 17:30:20.692925: E external/local_xla/xla/stream_executor/cuda/cuda_platform.cc:51] failed call to cuInit: INTERNAL: CUDA error: Failed call to cuInit: UNKNOWN ERROR (303) I0420 17:30:20.871632 140492519659328 max_utils.py:273] Attempting to initialize the jax distributed system... I0420 17:30:29.912387 140492519659328 distributed.py:149] Starting JAX distributed service on [::]:8482 I0420 17:30:29.914764 140492519659328 distributed.py:172] Connecting to JAX distributed service on mt-16-zarr3-false-rs-x93fq-slice-job-0-0.mt-16-zarr3-false-rs-x93fq:8482 I0420 17:30:31.168908 140492519659328 max_utils.py:284] Jax distributed system initialized! I0420 17:30:37.357240 140492519659328 max_utils.py:800] System Information: Jax Version: 0.9.2 I0420 17:30:37.357349 140492519659328 max_utils.py:801] System Information: Jaxlib Version: 0.9.2 I0420 17:30:37.357389 140492519659328 max_utils.py:802] System Information: Jax Backend: PJRT C API TFRT TPU v6 lite Built on Mar 4 2026 11:32:08 (1772652728) cl/878335365 I0420 17:30:37.357424 140492519659328 train_utils.py:378] WARNING: Sequence packing is essentially ignored for synthetic data. Please use a real dataset to use sequence packing. I0420 17:30:38.405976 140492519659328 maxtext_utils.py:1718] Num_devices: 32, shape (1, 1, 1, 32, 1, 1, 1, 1, 1, 1, 1, 1, 1) I0420 17:30:38.406701 140492519659328 maxtext_utils.py:1718] Num_devices: 32, shape (1, 1, 1, 32, 1, 1, 1, 1, 1, 1, 1, 1, 1) I0420 17:30:38.406911 140492519659328 checkpointing.py:688] Setting up checkpoint logger... I0420 17:30:38.406968 140492519659328 checkpointing.py:234] Creating checkpoint manager with ocdbt=True and zarr3=False I0420 17:30:38.407012 140492519659328 pytree_checkpoint_handler.py:592] save_device_host_concurrent_bytes=None I0420 17:30:38.407372 140492519659328 base_pytree_checkpoint_handler.py:441] Created BasePyTreeCheckpointHandler: use_ocdbt=True, use_zarr3=False, pytree_metadata_options=PyTreeMetadataOptions(support_rich_types=False), array_metadata_store=<orbax.checkpoint._src.metadata.array_metadata_store.Store object at 0x7fc65ad1c170>, enable_pinned_host_transfer=False, save_concurrent_bytes: 96000000000 (89.4 GiB), restore_concurrent_bytes: 96000000000 (89.4 GiB) I0420 17:30:39.741818 140492519659328 checkpointing.py:266] Enabling policy for fixed interval checkpointing. I0420 17:30:39.742054 140492519659328 checkpoint_manager.py:708] [process=0][thread=MainThread] CheckpointManager init: checkpointers=None, item_names=('items',), item_handlers={'items': <orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeCheckpointHandler object at 0x7fc5333ab800>}, handler_registry=None I0420 17:30:39.742302 140492519659328 composite_checkpoint_handler.py:237] Deferred registration for item: "items". Adding handler `<orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeCheckpointHandler object at 0x7fc5333ab800>` for item "items" and save args `<class 'orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeSaveArgs'>` and restore args `<class 'orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeRestoreArgs'>` to `_handler_registry`. I0420 17:30:39.742351 140492519659328 composite_checkpoint_handler.py:237] Deferred registration for item: "metrics". Adding handler `<orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonCheckpointHandler object at 0x7fc53338a090>` for item "metrics" and save args `<class 'orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonSaveArgs'>` and restore args `<class 'orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonRestoreArgs'>` to `_handler_registry`. I0420 17:30:39.742387 140492519659328 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 0x7fc5333ab800>, ('items', <class 'orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeRestoreArgs'>): <orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeCheckpointHandler object at 0x7fc5333ab800>, ('metrics', <class 'orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonSaveArgs'>): <orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonCheckpointHandler object at 0x7fc53338a090>, ('metrics', <class 'orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonRestoreArgs'>): <orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonCheckpointHandler object at 0x7fc53338a090>}). I0420 17:30:39.742695 140492519659328 abstract_checkpointer.py:35] orbax-checkpoint version: 0.11.34 I0420 17:30:39.742763 140492519659328 async_checkpointer.py:192] [process=0][thread=MainThread] Using barrier_sync_fn: <function get_barrier_sync_fn.<locals>._fn at 0x7fb21057ec00> timeout: 1200 secs and primary_host=0 for async checkpoint writes I0420 17:30:41.454436 140371426252544 checkpoint.py:228] Read Metadata={'item_handlers': {'items': 'orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeCheckpointHandler'}, 'metrics': {}, 'performance_metrics': {}, 'init_timestamp_nsecs': 1776706061168527914, 'commit_timestamp_nsecs': 1776706121296651082, 'custom_metadata': {}} from gs://lance-maxtext/linen_ckpt_xpk_feat_nnx_post_train_fixes_20260420_153543/linen_xpk_feat_nnx_post_train_fixes_20260420_153543_16_zarr3_false/checkpoints/9/_CHECKPOINT_METADATA I0420 17:30:41.701423 140371426252544 checkpoint_manager.py:1871] Missing metrics for step 9 I0420 17:30:42.013917 140492519659328 checkpoint_manager.py:1812] Found 3 checkpoint steps in gs://lance-maxtext/linen_ckpt_xpk_feat_nnx_post_train_fixes_20260420_153543/linen_xpk_feat_nnx_post_train_fixes_20260420_153543_16_zarr3_false/checkpoints I0420 17:30:42.963572 140492519659328 checkpoint_manager.py:929] [process=0][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_post_train_fixes_20260420_153543/linen_xpk_feat_nnx_post_train_fixes_20260420_153543_16_zarr3_false/checkpoints: <orbax.checkpoint.checkpoint_manager.CheckpointManager object at 0x7fc5331aecc0> I0420 17:30:42.963732 140492519659328 checkpointing.py:302] Checkpoint manager created! I0420 17:30:43.875760 140492519659328 nnx_wrappers.py:437] Unknown Logical: bfloat16[32,2048,2048]...................................... ('activation_batch', 'activation_norm_length', 'activation_embed'). I0420 17:30:43.875861 140492519659328 nnx_wrappers.py:437] Unknown Physical: bfloat16[32,2048,2048]...................................... ('fsdp', None, None). I0420 17:30:44.254218 140492519659328 attentions.py:1088] attentions/inputs_q Logical: bfloat16[32,2048,2048]...................................... ('activation_batch', 'activation_attn_length', 'activation_attn_embed'). I0420 17:30:44.254307 140492519659328 attentions.py:1088] attentions/inputs_q Physical: bfloat16[32,2048,2048]...................................... ('fsdp', None, None). I0420 17:30:44.275579 140492519659328 attentions.py:1089] attentions/inputs_kv Logical: bfloat16[32,2048,2048]...................................... ('activation_batch', 'activation_attn_length', 'activation_attn_embed'). I0420 17:30:44.275681 140492519659328 attentions.py:1089] attentions/inputs_kv Physical: bfloat16[32,2048,2048]...................................... ('fsdp', None, None). I0420 17:30:44.299357 140492519659328 attentions.py:1154] attentions/query Logical: bfloat16[32,2048,16,128].................................... ('activation_kv_batch', 'activation_attn_length', 'activation_kv_heads', 'activation_kv_head_dim'). I0420 17:30:44.299426 140492519659328 attentions.py:1154] attentions/query Physical: bfloat16[32,2048,16,128].................................... ('fsdp', None, None, None). I0420 17:30:44.315780 140492519659328 attentions.py:1155] attentions/key Logical: bfloat16[32,2048,16,128].................................... ('activation_kv_batch', 'activation_attn_length', 'activation_kv_heads', 'activation_kv_head_dim'). I0420 17:30:44.315843 140492519659328 attentions.py:1155] attentions/key Physical: bfloat16[32,2048,16,128].................................... ('fsdp', None, None, None). I0420 17:30:44.331972 140492519659328 attentions.py:1156] attentions/value Logical: bfloat16[32,2048,16,128].................................... ('activation_kv_batch', 'activation_attn_length', 'activation_kv_heads', 'activation_kv_head_dim'). I0420 17:30:44.332030 140492519659328 attentions.py:1156] attentions/value Physical: bfloat16[32,2048,16,128].................................... ('fsdp', None, None, None). I0420 17:30:44.357212 140492519659328 attentions.py:1197] attentions/out Logical: bfloat16[32,2048,16,128].................................... ('activation_batch', 'activation_attn_length', 'activation_heads', 'activation_kv'). I0420 17:30:44.357281 140492519659328 attentions.py:1197] attentions/out Physical: bfloat16[32,2048,16,128].................................... ('fsdp', None, None, None). I0420 17:30:44.377947 140492519659328 linears.py:525] linears/x Logical: bfloat16[32,2048,7168]...................................... ('activation_batch', 'activation_length', 'activation_mlp'). I0420 17:30:44.378013 140492519659328 linears.py:525] linears/x Physical: bfloat16[32,2048,7168]...................................... ('fsdp', None, None). I0420 17:30:44.593368 140492519659328 checkpointing.py:578] checkpoint manager exists so trying to load this run's existing checkpoint I0420 17:30:44.593474 140492519659328 checkpointing.py:582] restoring from this run's directory step 9 I0420 17:30:44.842710 140492519659328 event_tracking.py:70] [process=0] [sync] Started load checkpoint @ gs://lance-maxtext/linen_ckpt_xpk_feat_nnx_post_train_fixes_20260420_153543/linen_xpk_feat_nnx_post_train_fixes_20260420_153543_16_zarr3_false/checkpoints/9. I0420 17:30:45.094684 140492519659328 checkpointer.py:307] Restoring checkpoint from gs://lance-maxtext/linen_ckpt_xpk_feat_nnx_post_train_fixes_20260420_153543/linen_xpk_feat_nnx_post_train_fixes_20260420_153543_16_zarr3_false/checkpoints/9. I0420 17:30:45.095110 140492519659328 event_tracking.py:125] [process=0] [sync] Finished blocking load in 0.25 seconds. Continuing load @ gs://lance-maxtext/linen_ckpt_xpk_feat_nnx_post_train_fixes_20260420_153543/linen_xpk_feat_nnx_post_train_fixes_20260420_153543_16_zarr3_false/checkpoints/9. I0420 17:30:46.957357 1452 google_auth_provider.cc:181] Running on GCE, using service account 562977990677-compute@developer.gserviceaccount.com I0420 17:30:53.202164 140492519659328 jax_array_handlers.py:843] [process=0] /jax/orbax/read/worker/io/requested throughput: 252.823 MiB/s (total gbytes: 1.5 GiB) (time elapsed: 6.250504970550537 s) (per-host) I0420 17:30:53.207264 140492519659328 base_pytree_checkpoint_handler.py:130] [process=0] /jax/checkpoint/read/gbytes_per_sec: 237.949 MiB/s (total gbytes: 1.5 GiB) (time elapsed: 6.650965213775635 s) (per-host) I0420 17:30:54.551159 140492519659328 event_tracking.py:138] [process=0] [sync] Finished load in 9.71 seconds @ gs://lance-maxtext/linen_ckpt_xpk_feat_nnx_post_train_fixes_20260420_153543/linen_xpk_feat_nnx_post_train_fixes_20260420_153543_16_zarr3_false/checkpoints/9 I0420 17:30:54.551315 140492519659328 standard_logger.py:34] {'step': 9, 'event_type': 'restore', 'directory': 'gs://lance-maxtext/linen_ckpt_xpk_feat_nnx_post_train_fixes_20260420_153543/linen_xpk_feat_nnx_post_train_fixes_20260420_153543_16_zarr3_false/checkpoints', 'checkpointer_start_time': 1776706244.8426044, 'checkpointer_duration_secs': 9.708678245544434, 'checkpoint_manager_start_time': 1776706244.5937223, 'checkpoint_manager_duration_secs': 9.957562923431396} fsdp: 32 I0420 17:30:54.960178 140492519659328 maxtext_utils.py:1827] params/params/decoder/decoder_norm/scale Shape: float32[2048] Logical: P('norm',) Physical: (None,) I0420 17:30:54.960301 140492519659328 maxtext_utils.py:1827] params/params/decoder/layers/mlp/wi_0/kernel Shape: float32[2048,16,7168] Logical: P('embed', 'layers', 'mlp') Physical: ('fsdp', None, None) I0420 17:30:54.960354 140492519659328 maxtext_utils.py:1827] params/params/decoder/layers/mlp/wi_1/kernel Shape: float32[2048,16,7168] Logical: P('embed', 'layers', 'mlp') Physical: ('fsdp', None, None) I0420 17:30:54.960411 140492519659328 maxtext_utils.py:1827] params/params/decoder/layers/mlp/wo/kernel Shape: float32[7168,16,2048] Logical: P('mlp', 'layers', 'embed') Physical: (None, None, 'fsdp') I0420 17:30:54.960464 140492519659328 maxtext_utils.py:1827] params/params/decoder/layers/post_self_attention_layer_norm/scale Shape: float32[2048,16] Logical: P('norm', 'layers') Physical: (None, None) I0420 17:30:54.960504 140492519659328 maxtext_utils.py:1827] params/params/decoder/layers/pre_self_attention_layer_norm/scale Shape: float32[2048,16] Logical: P('norm', 'layers') Physical: (None, None) I0420 17:30:54.960560 140492519659328 maxtext_utils.py:1827] params/params/decoder/layers/self_attention/key/kernel Shape: float32[2048,16,16,128] Logical: P('embed', 'layers', 'kv_heads', 'kv_head_dim') Physical: ('fsdp', None, None, None) I0420 17:30:54.960612 140492519659328 maxtext_utils.py:1827] params/params/decoder/layers/self_attention/out/kernel Shape: float32[16,16,128,2048] Logical: P('heads', 'layers', 'kv', 'embed') Physical: (None, None, None, 'fsdp') I0420 17:30:54.960651 140492519659328 maxtext_utils.py:1827] params/params/decoder/layers/self_attention/query/kernel Shape: float32[2048,16,16,128] Logical: P('embed', 'layers', 'q_heads', 'kv') Physical: ('fsdp', None, None, None) I0420 17:30:54.960687 140492519659328 maxtext_utils.py:1827] params/params/decoder/layers/self_attention/value/kernel Shape: float32[2048,16,16,128] Logical: P('embed', 'layers', 'kv_heads', 'kv_head_dim') Physical: ('fsdp', None, None, None) I0420 17:30:54.960734 140492519659328 maxtext_utils.py:1827] params/params/decoder/logits_dense/kernel Shape: float32[2048,32000] Logical: P('embed_vocab', 'vocab') Physical: ('fsdp', None) I0420 17:30:54.960779 140492519659328 maxtext_utils.py:1827] params/params/token_embedder/embedding Shape: float32[32000,2048] Logical: P('vocab', 'embed_vocab') Physical: (None, 'fsdp') I0420 17:30:55.441667 140492519659328 train.py:157] train/xent Logical: float32[32,2048]............................................ ('activation_embed_and_logits_batch', 'activation_length'). I0420 17:30:55.441761 140492519659328 train.py:157] train/xent Physical: float32[32,2048]............................................ ('fsdp', None). I0420 17:30:55.457108 140492519659328 train.py:164] train/z_loss Logical: float32[32,2048]............................................ ('activation_embed_and_logits_batch', 'activation_length'). I0420 17:30:55.457170 140492519659328 train.py:164] train/z_loss Physical: float32[32,2048]............................................ ('fsdp', None). I0420 17:31:06.285702 140492519659328 max_utils.py:791] Total memory size: 1.5 GB, Output size: 0.4 GB, Temp size: 1.1 GB, Argument size: 0.4 GB, Host temp size: 0.0 GB. I0420 17:31:06.293525 140492519659328 metric_logger.py:301] number parameters: 1.104 billion I0420 17:31:17.984047 140492519659328 checkpointing.py:794] Waiting for step 10 to finish before checkpoint... I0420 17:31:18.136844 140492519659328 checkpointing.py:798] Waited 0.15277671813964844 seconds for step 10 to finish before starting checkpointing. I0420 17:31:18.139327 140492519659328 checkpoint_manager.py:2009] [process=0][thread=MainThread][wait_until_finished] No Save Finalize thread to wait for. Returning. I0420 17:31:18.141127 140492519659328 checkpoint_manager.py:1512] [process=0] Saving checkpoint at step 10 I0420 17:31:18.142514 140492519659328 event_tracking.py:70] [process=0] [async] Started save checkpoint @ gs://lance-maxtext/linen_ckpt_xpk_feat_nnx_post_train_fixes_20260420_153543/linen_xpk_feat_nnx_post_train_fixes_20260420_153543_16_zarr3_false/checkpoints/10. I0420 17:31:18.894958 140492519659328 signaling_client.py:364] Using JaxDistributedSignalingClient I0420 17:31:18.907832 140492519659328 jax_array_handlers.py:360] Scheduling D2H of 39 prioritized jax.Array. I0420 17:31:18.907898 140492519659328 replica_slices.py:424] Transferring arrays to host memory with options: use_replica_parallel=True, min_slice_bytes_for_replica_parallel=None, max_replicas_for_replica_parallel=None, enable_pinned_host_transfer=False I0420 17:31:19.179783 140492519659328 base_pytree_checkpoint_handler.py:154] [process=0][thread=MainThread] Initiated "orbax.checkpoint._src.serialization.jax_array_handlers.ArrayHandler".serialize. Time taken: 0.272991s I0420 17:31:19.180645 140492519659328 base_pytree_checkpoint_handler.py:130] [process=0] /jax/orbax/write/blocking_gbytes_per_sec: 5.557 GiB/s (total gbytes: 1.5 GiB) (time elapsed: 0.27759313583374023 s) (per-host) I0420 17:31:19.180709 140492519659328 base_pytree_checkpoint_handler.py:768] [process=0][thread=MainThread] Initiated Pytree async_save. Time taken: 0.277675s (batch_requests_ready=0.001992s, total_serialization_initiated=0.274906s, others=0.000777s) I0420 17:31:19.180784 140492519659328 composite_checkpoint_handler.py:715] [process=0][thread=MainThread] Initiated CompositeCheckpointHandler.async_save. Time taken: 0.281776s (all_items=0.000017s, per_item={'items': '0.00001717'}, temp_paths=0.281759) I0420 17:31:19.181494 140492519659328 event_tracking.py:125] [process=0] [async] Finished blocking save in 1.04 seconds. Continuing save @ gs://lance-maxtext/linen_ckpt_xpk_feat_nnx_post_train_fixes_20260420_153543/linen_xpk_feat_nnx_post_train_fixes_20260420_153543_16_zarr3_false/checkpoints/10. I0420 17:31:19.181837 140370847450880 async_checkpointer.py:76] [process=0][thread=async_save] Background save thread started. Deadline for this save operation is 2026-04-20 17:51:19.181799 I0420 17:31:19.182895 140370839058176 atomicity.py:140] Creating tmp directory gs://lance-maxtext/linen_ckpt_xpk_feat_nnx_post_train_fixes_20260420_153543/linen_xpk_feat_nnx_post_train_fixes_20260420_153543_16_zarr3_false/checkpoints/10 I0420 17:31:19.189716 140492519659328 checkpoint_manager.py:1560] [process=0][thread=MainThread][step=10] Starting CheckpointManager Save Finalize thread=save_finalize I0420 17:31:19.190000 140366569268992 async_checkpointer.py:280] [process=0][thread=save_finalize] Waiting for background save thread=async_save. I0420 17:31:19.190135 140492519659328 standard_logger.py:34] {'step': 10, 'event_type': 'save', 'directory': 'gs://lance-maxtext/linen_ckpt_xpk_feat_nnx_post_train_fixes_20260420_153543/linen_xpk_feat_nnx_post_train_fixes_20260420_153543_16_zarr3_false/checkpoints', 'reached_preemption': False, 'preemption_received_at': None, 'synchronous': False, 'wait_for_prev_start_time': 1776706278.1393094, 'wait_for_prev_duration_secs': 5.841255187988281e-05, 'time_between_consecutive_saves_sec': None, 'checkpointer_blocking_start_time': 1776706278.1411655, 'checkpointer_blocking_duration_secs': 1.040830373764038, 'get_old_steps_start_time': 1776706279.1820176, 'get_old_steps_duration_secs': 3.075599670410156e-05, 'checkpoint_manager_blocking_start_time': 1776706278.1373577, 'checkpoint_manager_blocking_duration_secs': 1.0527427196502686} I0420 17:31:19.190241 140492519659328 checkpointing.py:409] Started an asynchronous checkpoint save for step 10 I0420 17:31:19.190293 140492519659328 max_utils.py:750] Memstats: After params initialized: I0420 17:31:19.190342 140492519659328 max_utils.py:756] Using (GB) 0.43 / 31.25 (1.376000%) on TPU_0(process=0,(0,0,0,0)) I0420 17:31:19.190375 140492519659328 max_utils.py:756] Using (GB) 0.43 / 31.25 (1.376000%) on TPU_1(process=0,(1,0,0,0)) I0420 17:31:19.190402 140492519659328 max_utils.py:756] Using (GB) 0.43 / 31.25 (1.376000%) on TPU_4(process=0,(0,1,0,0)) I0420 17:31:19.190428 140492519659328 max_utils.py:756] Using (GB) 0.43 / 31.25 (1.376000%) on TPU_5(process=0,(1,1,0,0)) I0420 17:31:19.530543 140492519659328 metric_logger.py:196] completed step: 10, seconds: 11.396, TFLOP/s/device: 1.192, Tokens/s/device: 179.705, total_weights: 65536, loss: 6.833, lm_loss: 6.833, perplexity: 927.869 I0420 17:31:19.675554 140492519659328 metric_logger.py:196] completed step: 11, seconds: 1.545, TFLOP/s/device: 8.797, Tokens/s/device: 1325.919, total_weights: 65536, loss: 6.527, lm_loss: 6.527, perplexity: 683.659 I0420 17:31:19.823162 140492519659328 metric_logger.py:196] completed step: 12, seconds: 0.011, TFLOP/s/device: 1267.220, Tokens/s/device: 191009.140, total_weights: 65536, loss: 6.265, lm_loss: 6.265, perplexity: 526.068 I0420 17:31:19.970550 140492519659328 metric_logger.py:196] completed step: 13, seconds: 0.145, TFLOP/s/device: 93.536, Tokens/s/device: 14098.760, total_weights: 65536, loss: 6.045, lm_loss: 6.045, perplexity: 422.001 I0420 17:31:19.979577 140492519659328 checkpointing.py:794] Waiting for step 15 to finish before checkpoint... I0420 17:31:20.264634 140492519659328 checkpointing.py:798] Waited 0.28490185737609863 seconds for step 15 to finish before starting checkpointing. I0420 17:31:20.267736 140492519659328 checkpoint_manager.py:2020] [process=0][thread=MainThread][step=10][wait_until_finished] Waiting for Save Finalize thread (save_finalize) to complete. I0420 17:31:21.656682 140367601047296 atomicity.py:140] Creating tmp directory gs://lance-maxtext/linen_ckpt_xpk_feat_nnx_post_train_fixes_20260420_153543/linen_xpk_feat_nnx_post_train_fixes_20260420_153543_16_zarr3_false/checkpoints/10/items I0420 17:31:21.825758 140366523119360 checkpoint.py:188] Wrote Metadata={'item_handlers': None, 'metrics': {}, 'performance_metrics': {}, 'init_timestamp_nsecs': 1776706281377679796, 'commit_timestamp_nsecs': None, 'custom_metadata': {}}, json={"item_handlers": null, "metrics": {}, "performance_metrics": {}, "init_timestamp_nsecs": 1776706281377679796, "commit_timestamp_nsecs": null, "custom_metadata": {}} to gs://lance-maxtext/linen_ckpt_xpk_feat_nnx_post_train_fixes_20260420_153543/linen_xpk_feat_nnx_post_train_fixes_20260420_153543_16_zarr3_false/checkpoints/10/_CHECKPOINT_METADATA I0420 17:31:25.317631 140367106139904 array_metadata_store.py:203] [process=0][thread=array_type_handler] Wrote 39 array_metadata.ArrayMetadata to gs://lance-maxtext/linen_ckpt_xpk_feat_nnx_post_train_fixes_20260420_153543/linen_xpk_feat_nnx_post_train_fixes_20260420_153543_16_zarr3_false/checkpoints/10/items/array_metadatas/process_0 I0420 17:32:05.661186 140367074809600 base_pytree_checkpoint_handler.py:1282] [process=0][thread=write_metadata_after_commits] Commit + Array metadata written. Time taken: 42.284533s (commit=40.950991s, array_metadata_write=1.333542s) I0420 17:32:05.662623 140370847450880 base_pytree_checkpoint_handler.py:130] [process=0] /jax/orbax/write/gbytes_per_sec: 33.781 MiB/s (total gbytes: 1.5 GiB) (time elapsed: 46.75952982902527 s) (per-host) I0420 17:32:05.662789 140370847450880 async_checkpointer.py:90] [process=0][thread=async_save] 3 Handler Commit operations completed. Time taken: 46.480828s. I0420 17:32:07.350568 140370847450880 array_metadata_store.py:411] [process=0][thread=async_save] Validated ArrayMetadata from all 8 hosts. Time taken: 0.000391s. I0420 17:32:07.790454 140366523119360 checkpoint.py:247] Updated Metadata={'item_handlers': {'items': 'orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeCheckpointHandler'}, 'metrics': {}, 'performance_metrics': {}, 'init_timestamp_nsecs': 1776706281377679796, 'commit_timestamp_nsecs': None, 'custom_metadata': {}} to gs://lance-maxtext/linen_ckpt_xpk_feat_nnx_post_train_fixes_20260420_153543/linen_xpk_feat_nnx_post_train_fixes_20260420_153543_16_zarr3_false/checkpoints/10/_CHECKPOINT_METADATA I0420 17:32:08.617788 140370847450880 base_pytree_checkpoint_handler.py:1406] [process=0][thread=async_save] Pytree save finalize (merge_ocdbt + ArrayMetadata validation) completed. Time taken: 1.949631s. use_zarr3=False, enable_post_merge_validation=True, directory=gs://lance-maxtext/linen_ckpt_xpk_feat_nnx_post_train_fixes_20260420_153543/linen_xpk_feat_nnx_post_train_fixes_20260420_153543_16_zarr3_false/checkpoints/10/items I0420 17:32:08.620214 140370847450880 atomicity.py:666] Finalizing gs://lance-maxtext/linen_ckpt_xpk_feat_nnx_post_train_fixes_20260420_153543/linen_xpk_feat_nnx_post_train_fixes_20260420_153543_16_zarr3_false/checkpoints/10/items I0420 17:32:09.214557 140370847450880 atomicity.py:666] Finalizing gs://lance-maxtext/linen_ckpt_xpk_feat_nnx_post_train_fixes_20260420_153543/linen_xpk_feat_nnx_post_train_fixes_20260420_153543_16_zarr3_false/checkpoints/10 I0420 17:32:10.744695 140370847450880 atomicity.py:847] [process=0][thread=async_save] Finished saving checkpoint (finalized tmp dir) to `gs://lance-maxtext/linen_ckpt_xpk_feat_nnx_post_train_fixes_20260420_153543/linen_xpk_feat_nnx_post_train_fixes_20260420_153543_16_zarr3_false/checkpoints/10`. I0420 17:32:10.745871 140370847450880 event_tracking.py:138] [process=0] [async] Finished save (blocking + background) in 52.60 seconds @ gs://lance-maxtext/linen_ckpt_xpk_feat_nnx_post_train_fixes_20260420_153543/linen_xpk_feat_nnx_post_train_fixes_20260420_153543_16_zarr3_false/checkpoints/10 I0420 17:32:10.747557 140370847450880 async_checkpointer.py:160] [process=0][thread=async_save] Background save thread done. Time taken: 51.565592s. I0420 17:32:10.747858 140366569268992 async_checkpointer.py:288] [process=0][thread=save_finalize] Done with waiting for background save thread=async_save. I0420 17:32:10.748044 140366569268992 async_checkpointer.py:298] [process=0][thread=save_finalize] No errors found in background save thread=async_save. I0420 17:32:10.748193 140366569268992 checkpoint_manager.py:2137] [process=0][thread=save_finalize][step=10] CheckpointManager Save Finalize is syncing with other hosts... I0420 17:32:10.750722 140366569268992 checkpoint_manager.py:2146] [process=0][thread=save_finalize][step=10] CheckpointManager Save Finalize is done on all hosts. I0420 17:32:10.751031 140492519659328 checkpoint_manager.py:2032] [process=0][thread=MainThread][step=10][wait_until_finished] Done waiting for Save Finalize thread (save_finalize) running at step=10. W0420 17:32:10.751236 140492519659328 checkpoint_manager.py:1452] Waiting for previous save to complete took 50.483500 seconds. If this number is high, consider checkpointing less frequently. I0420 17:32:10.753130 140492519659328 checkpoint_manager.py:1512] [process=0] Saving checkpoint at step 15 I0420 17:32:10.755612 140492519659328 event_tracking.py:70] [process=0] [async] Started save checkpoint @ gs://lance-maxtext/linen_ckpt_xpk_feat_nnx_post_train_fixes_20260420_153543/linen_xpk_feat_nnx_post_train_fixes_20260420_153543_16_zarr3_false/checkpoints/15. I0420 17:32:11.083108 140492519659328 jax_array_handlers.py:360] Scheduling D2H of 39 prioritized jax.Array. I0420 17:32:11.083237 140492519659328 replica_slices.py:424] Transferring arrays to host memory with options: use_replica_parallel=True, min_slice_bytes_for_replica_parallel=None, max_replicas_for_replica_parallel=None, enable_pinned_host_transfer=False I0420 17:32:11.111428 140492519659328 base_pytree_checkpoint_handler.py:154] [process=0][thread=MainThread] Initiated "orbax.checkpoint._src.serialization.jax_array_handlers.ArrayHandler".serialize. Time taken: 0.029446s I0420 17:32:11.112433 140492519659328 base_pytree_checkpoint_handler.py:130] [process=0] /jax/orbax/write/blocking_gbytes_per_sec: 40.691 GiB/s (total gbytes: 1.5 GiB) (time elapsed: 0.03790903091430664 s) (per-host) I0420 17:32:11.112526 140492519659328 base_pytree_checkpoint_handler.py:768] [process=0][thread=MainThread] Initiated Pytree async_save. Time taken: 0.038025s (batch_requests_ready=0.003925s, total_serialization_initiated=0.033209s, others=0.000891s) I0420 17:32:11.112647 140492519659328 composite_checkpoint_handler.py:715] [process=0][thread=MainThread] Initiated CompositeCheckpointHandler.async_save. Time taken: 0.042873s (all_items=0.000021s, per_item={'items': '0.00002098'}, temp_paths=0.042852) I0420 17:32:11.113623 140492519659328 event_tracking.py:125] [process=0] [async] Finished blocking save in 0.36 seconds. Continuing save @ gs://lance-maxtext/linen_ckpt_xpk_feat_nnx_post_train_fixes_20260420_153543/linen_xpk_feat_nnx_post_train_fixes_20260420_153543_16_zarr3_false/checkpoints/15. I0420 17:32:11.114107 140366569268992 async_checkpointer.py:76] [process=0][thread=async_save] Background save thread started. Deadline for this save operation is 2026-04-20 17:52:11.114047 I0420 17:32:11.116684 140492519659328 checkpoint_manager.py:1560] [process=0][thread=MainThread][step=15] Starting CheckpointManager Save Finalize thread=save_finalize I0420 17:32:11.117121 140367106139904 async_checkpointer.py:280] [process=0][thread=save_finalize] Waiting for background save thread=async_save. I0420 17:32:11.117346 140492519659328 standard_logger.py:34] {'step': 15, 'event_type': 'save', 'directory': 'gs://lance-maxtext/linen_ckpt_xpk_feat_nnx_post_train_fixes_20260420_153543/linen_xpk_feat_nnx_post_train_fixes_20260420_153543_16_zarr3_false/checkpoints', 'reached_preemption': False, 'preemption_received_at': None, 'synchronous': False, 'wait_for_prev_start_time': 1776706280.2676919, 'wait_for_prev_duration_secs': 50.483500242233276, 'time_between_consecutive_saves_sec': None, 'checkpointer_blocking_start_time': 1776706330.7531981, 'checkpointer_blocking_duration_secs': 0.3611268997192383, 'get_old_steps_start_time': 1776706331.1143515, 'get_old_steps_duration_secs': 4.673004150390625e-05, 'checkpoint_manager_blocking_start_time': 1776706280.2653286, 'checkpoint_manager_blocking_duration_secs': 50.85196542739868} I0420 17:32:11.117496 140492519659328 checkpointing.py:409] Started an asynchronous checkpoint save for step 15 I0420 17:32:11.492060 140370847450880 atomicity.py:140] Creating tmp directory gs://lance-maxtext/linen_ckpt_xpk_feat_nnx_post_train_fixes_20260420_153543/linen_xpk_feat_nnx_post_train_fixes_20260420_153543_16_zarr3_false/checkpoints/15 I0420 17:32:13.357116 140370839058176 atomicity.py:140] Creating tmp directory gs://lance-maxtext/linen_ckpt_xpk_feat_nnx_post_train_fixes_20260420_153543/linen_xpk_feat_nnx_post_train_fixes_20260420_153543_16_zarr3_false/checkpoints/15/items I0420 17:32:16.948526 140367601047296 array_metadata_store.py:203] [process=0][thread=array_type_handler] Wrote 39 array_metadata.ArrayMetadata to gs://lance-maxtext/linen_ckpt_xpk_feat_nnx_post_train_fixes_20260420_153543/linen_xpk_feat_nnx_post_train_fixes_20260420_153543_16_zarr3_false/checkpoints/15/items/array_metadatas/process_0 I0420 17:32:31.911252 140492519659328 metric_logger.py:196] completed step: 14, seconds: 0.148, TFLOP/s/device: 92.025, Tokens/s/device: 13871.016, total_weights: 65536, loss: 5.864, lm_loss: 5.864, perplexity: 352.193 I0420 17:32:31.920500 140492519659328 metric_logger.py:196] completed step: 15, seconds: 0.147, TFLOP/s/device: 92.316, Tokens/s/device: 13914.934, total_weights: 65536, loss: 5.719, lm_loss: 5.719, perplexity: 304.661 I0420 17:32:32.066143 140492519659328 metric_logger.py:196] completed step: 16, seconds: 71.939, TFLOP/s/device: 0.189, Tokens/s/device: 28.468, total_weights: 65536, loss: 5.605, lm_loss: 5.605, perplexity: 271.872 I0420 17:32:32.213502 140492519659328 metric_logger.py:196] completed step: 17, seconds: 0.007, TFLOP/s/device: 1846.077, Tokens/s/device: 278260.870, total_weights: 65536, loss: 5.517, lm_loss: 5.517, perplexity: 248.975 I0420 17:32:32.361016 140492519659328 metric_logger.py:196] completed step: 18, seconds: 0.146, TFLOP/s/device: 92.788, Tokens/s/device: 13986.014, total_weights: 65536, loss: 5.448, lm_loss: 5.448, perplexity: 232.333 I0420 17:32:32.507582 140492519659328 checkpointing.py:794] Waiting for step 19 to finish before checkpoint... I0420 17:32:32.508300 140492519659328 checkpointing.py:798] Waited 0.0007379055023193359 seconds for step 19 to finish before starting checkpointing. I0420 17:32:32.510270 140492519659328 checkpoint_manager.py:2020] [process=0][thread=MainThread][step=15][wait_until_finished] Waiting for Save Finalize thread (save_finalize) to complete. I0420 17:32:57.134390 140367114532608 base_pytree_checkpoint_handler.py:1282] [process=0][thread=write_metadata_after_commits] Commit + Array metadata written. Time taken: 42.573396s (commit=41.323044s, array_metadata_write=1.250352s) I0420 17:32:57.135432 140366569268992 base_pytree_checkpoint_handler.py:130] [process=0] /jax/orbax/write/gbytes_per_sec: 34.294 MiB/s (total gbytes: 1.5 GiB) (time elapsed: 46.06090831756592 s) (per-host) I0420 17:32:57.135513 140366569268992 async_checkpointer.py:90] [process=0][thread=async_save] 3 Handler Commit operations completed. Time taken: 46.021228s. I0420 17:32:59.068747 140366569268992 array_metadata_store.py:411] [process=0][thread=async_save] Validated ArrayMetadata from all 8 hosts. Time taken: 0.000243s. I0420 17:33:00.280410 140366569268992 base_pytree_checkpoint_handler.py:1406] [process=0][thread=async_save] Pytree save finalize (merge_ocdbt + ArrayMetadata validation) completed. Time taken: 2.242553s. use_zarr3=False, enable_post_merge_validation=True, directory=gs://lance-maxtext/linen_ckpt_xpk_feat_nnx_post_train_fixes_20260420_153543/linen_xpk_feat_nnx_post_train_fixes_20260420_153543_16_zarr3_false/checkpoints/15/items I0420 17:33:00.282357 140366569268992 atomicity.py:666] Finalizing gs://lance-maxtext/linen_ckpt_xpk_feat_nnx_post_train_fixes_20260420_153543/linen_xpk_feat_nnx_post_train_fixes_20260420_153543_16_zarr3_false/checkpoints/15/items I0420 17:33:00.824010 140366569268992 atomicity.py:666] Finalizing gs://lance-maxtext/linen_ckpt_xpk_feat_nnx_post_train_fixes_20260420_153543/linen_xpk_feat_nnx_post_train_fixes_20260420_153543_16_zarr3_false/checkpoints/15 I0420 17:33:02.275808 140366569268992 atomicity.py:847] [process=0][thread=async_save] Finished saving checkpoint (finalized tmp dir) to `gs://lance-maxtext/linen_ckpt_xpk_feat_nnx_post_train_fixes_20260420_153543/linen_xpk_feat_nnx_post_train_fixes_20260420_153543_16_zarr3_false/checkpoints/15`. I0420 17:33:02.276659 140366569268992 event_tracking.py:138] [process=0] [async] Finished save (blocking + background) in 51.52 seconds @ gs://lance-maxtext/linen_ckpt_xpk_feat_nnx_post_train_fixes_20260420_153543/linen_xpk_feat_nnx_post_train_fixes_20260420_153543_16_zarr3_false/checkpoints/15 I0420 17:33:02.278354 140366569268992 async_checkpointer.py:160] [process=0][thread=async_save] Background save thread done. Time taken: 51.164067s. I0420 17:33:02.278563 140367106139904 async_checkpointer.py:288] [process=0][thread=save_finalize] Done with waiting for background save thread=async_save. I0420 17:33:02.278683 140367106139904 async_checkpointer.py:298] [process=0][thread=save_finalize] No errors found in background save thread=async_save. I0420 17:33:02.278752 140367106139904 checkpoint_manager.py:2137] [process=0][thread=save_finalize][step=15] CheckpointManager Save Finalize is syncing with other hosts... I0420 17:33:02.280303 140367106139904 checkpoint_manager.py:2146] [process=0][thread=save_finalize][step=15] CheckpointManager Save Finalize is done on all hosts. I0420 17:33:02.280431 140492519659328 checkpoint_manager.py:2032] [process=0][thread=MainThread][step=15][wait_until_finished] Done waiting for Save Finalize thread (save_finalize) running at step=15. W0420 17:33:02.280506 140492519659328 checkpoint_manager.py:1452] Waiting for previous save to complete took 29.770247 seconds. If this number is high, consider checkpointing less frequently. I0420 17:33:02.282097 140492519659328 checkpoint_manager.py:1512] [process=0] Saving checkpoint at step 19 I0420 17:33:02.284168 140492519659328 event_tracking.py:70] [process=0] [async] Started save checkpoint @ gs://lance-maxtext/linen_ckpt_xpk_feat_nnx_post_train_fixes_20260420_153543/linen_xpk_feat_nnx_post_train_fixes_20260420_153543_16_zarr3_false/checkpoints/19. I0420 17:33:02.594034 140492519659328 jax_array_handlers.py:360] Scheduling D2H of 39 prioritized jax.Array. I0420 17:33:02.594141 140492519659328 replica_slices.py:424] Transferring arrays to host memory with options: use_replica_parallel=True, min_slice_bytes_for_replica_parallel=None, max_replicas_for_replica_parallel=None, enable_pinned_host_transfer=False I0420 17:33:02.621201 140492519659328 base_pytree_checkpoint_handler.py:154] [process=0][thread=MainThread] Initiated "orbax.checkpoint._src.serialization.jax_array_handlers.ArrayHandler".serialize. Time taken: 0.028144s I0420 17:33:02.621979 140492519659328 base_pytree_checkpoint_handler.py:130] [process=0] /jax/orbax/write/blocking_gbytes_per_sec: 47.533 GiB/s (total gbytes: 1.5 GiB) (time elapsed: 0.03245258331298828 s) (per-host) I0420 17:33:02.622035 140492519659328 base_pytree_checkpoint_handler.py:768] [process=0][thread=MainThread] Initiated Pytree async_save. Time taken: 0.032526s (batch_requests_ready=0.001759s, total_serialization_initiated=0.030079s, others=0.000688s) I0420 17:33:02.622118 140492519659328 composite_checkpoint_handler.py:715] [process=0][thread=MainThread] Initiated CompositeCheckpointHandler.async_save. Time taken: 0.036450s (all_items=0.000015s, per_item={'items': '0.00001526'}, temp_paths=0.036434) I0420 17:33:02.622845 140492519659328 event_tracking.py:125] [process=0] [async] Finished blocking save in 0.34 seconds. Continuing save @ gs://lance-maxtext/linen_ckpt_xpk_feat_nnx_post_train_fixes_20260420_153543/linen_xpk_feat_nnx_post_train_fixes_20260420_153543_16_zarr3_false/checkpoints/19. I0420 17:33:02.623151 140367106139904 async_checkpointer.py:76] [process=0][thread=async_save] Background save thread started. Deadline for this save operation is 2026-04-20 17:53:02.623118 I0420 17:33:02.625198 140492519659328 checkpoint_manager.py:1560] [process=0][thread=MainThread][step=19] Starting CheckpointManager Save Finalize thread=save_finalize I0420 17:33:02.625423 140369773709056 async_checkpointer.py:280] [process=0][thread=save_finalize] Waiting for background save thread=async_save. I0420 17:33:02.625540 140492519659328 standard_logger.py:34] {'step': 19, 'event_type': 'save', 'directory': 'gs://lance-maxtext/linen_ckpt_xpk_feat_nnx_post_train_fixes_20260420_153543/linen_xpk_feat_nnx_post_train_fixes_20260420_153543_16_zarr3_false/checkpoints', 'reached_preemption': False, 'preemption_received_at': None, 'synchronous': False, 'wait_for_prev_start_time': 1776706352.5102367, 'wait_for_prev_duration_secs': 29.770246744155884, 'time_between_consecutive_saves_sec': 21.7594473361969, 'checkpointer_blocking_start_time': 1776706382.2821357, 'checkpointer_blocking_duration_secs': 0.34114909172058105, 'get_old_steps_start_time': 1776706382.6233044, 'get_old_steps_duration_secs': 3.0517578125e-05, 'checkpoint_manager_blocking_start_time': 1776706352.508532, 'checkpoint_manager_blocking_duration_secs': 30.116976976394653} I0420 17:33:02.625648 140492519659328 checkpointing.py:409] Started an asynchronous checkpoint save for step 19 I0420 17:33:02.625689 140492519659328 checkpoint_manager.py:2020] [process=0][thread=MainThread][step=19][wait_until_finished] Waiting for Save Finalize thread (save_finalize) to complete. I0420 17:33:02.836376 140366569268992 atomicity.py:140] Creating tmp directory gs://lance-maxtext/linen_ckpt_xpk_feat_nnx_post_train_fixes_20260420_153543/linen_xpk_feat_nnx_post_train_fixes_20260420_153543_16_zarr3_false/checkpoints/19 I0420 17:33:05.026999 140367074809600 atomicity.py:140] Creating tmp directory gs://lance-maxtext/linen_ckpt_xpk_feat_nnx_post_train_fixes_20260420_153543/linen_xpk_feat_nnx_post_train_fixes_20260420_153543_16_zarr3_false/checkpoints/19/items I0420 17:33:08.839321 140370839058176 array_metadata_store.py:203] [process=0][thread=array_type_handler] Wrote 39 array_metadata.ArrayMetadata to gs://lance-maxtext/linen_ckpt_xpk_feat_nnx_post_train_fixes_20260420_153543/linen_xpk_feat_nnx_post_train_fixes_20260420_153543_16_zarr3_false/checkpoints/19/items/array_metadatas/process_0 I0420 17:33:47.689737 140369798887168 base_pytree_checkpoint_handler.py:1282] [process=0][thread=write_metadata_after_commits] Commit + Array metadata written. Time taken: 40.649563s (commit=39.384392s, array_metadata_write=1.265171s) I0420 17:33:47.690847 140367106139904 base_pytree_checkpoint_handler.py:130] [process=0] /jax/orbax/write/gbytes_per_sec: 35.023 MiB/s (total gbytes: 1.5 GiB) (time elapsed: 45.101266622543335 s) (per-host) I0420 17:33:47.690964 140367106139904 async_checkpointer.py:90] [process=0][thread=async_save] 3 Handler Commit operations completed. Time taken: 45.067703s. I0420 17:33:49.628230 140367106139904 array_metadata_store.py:411] [process=0][thread=async_save] Validated ArrayMetadata from all 8 hosts. Time taken: 0.000277s. I0420 17:33:51.285506 140367106139904 base_pytree_checkpoint_handler.py:1406] [process=0][thread=async_save] Pytree save finalize (merge_ocdbt + ArrayMetadata validation) completed. Time taken: 2.246704s. use_zarr3=False, enable_post_merge_validation=True, directory=gs://lance-maxtext/linen_ckpt_xpk_feat_nnx_post_train_fixes_20260420_153543/linen_xpk_feat_nnx_post_train_fixes_20260420_153543_16_zarr3_false/checkpoints/19/items I0420 17:33:51.287284 140367106139904 atomicity.py:666] Finalizing gs://lance-maxtext/linen_ckpt_xpk_feat_nnx_post_train_fixes_20260420_153543/linen_xpk_feat_nnx_post_train_fixes_20260420_153543_16_zarr3_false/checkpoints/19/items I0420 17:33:51.829332 140367106139904 atomicity.py:666] Finalizing gs://lance-maxtext/linen_ckpt_xpk_feat_nnx_post_train_fixes_20260420_153543/linen_xpk_feat_nnx_post_train_fixes_20260420_153543_16_zarr3_false/checkpoints/19 I0420 17:33:53.557745 140367106139904 atomicity.py:847] [process=0][thread=async_save] Finished saving checkpoint (finalized tmp dir) to `gs://lance-maxtext/linen_ckpt_xpk_feat_nnx_post_train_fixes_20260420_153543/linen_xpk_feat_nnx_post_train_fixes_20260420_153543_16_zarr3_false/checkpoints/19`. I0420 17:33:53.558501 140367106139904 event_tracking.py:138] [process=0] [async] Finished save (blocking + background) in 51.28 seconds @ gs://lance-maxtext/linen_ckpt_xpk_feat_nnx_post_train_fixes_20260420_153543/linen_xpk_feat_nnx_post_train_fixes_20260420_153543_16_zarr3_false/checkpoints/19 I0420 17:33:53.559728 140367106139904 async_checkpointer.py:160] [process=0][thread=async_save] Background save thread done. Time taken: 50.936464s. I0420 17:33:53.559943 140369773709056 async_checkpointer.py:288] [process=0][thread=save_finalize] Done with waiting for background save thread=async_save. I0420 17:33:53.560062 140369773709056 async_checkpointer.py:298] [process=0][thread=save_finalize] No errors found in background save thread=async_save. I0420 17:33:53.560130 140369773709056 checkpoint_manager.py:2137] [process=0][thread=save_finalize][step=19] CheckpointManager Save Finalize is syncing with other hosts... I0420 17:33:53.562810 140369773709056 checkpoint_manager.py:2146] [process=0][thread=save_finalize][step=19] CheckpointManager Save Finalize is done on all hosts. I0420 17:33:53.562931 140492519659328 checkpoint_manager.py:2032] [process=0][thread=MainThread][step=19][wait_until_finished] Done waiting for Save Finalize thread (save_finalize) running at step=19. I0420 17:33:53.563019 140492519659328 checkpoint_manager.py:2009] [process=0][thread=MainThread][wait_until_finished] No Save Finalize thread to wait for. Returning. I0420 17:33:53.563887 140492519659328 metric_logger.py:196] completed step: 19, seconds: 0.147, TFLOP/s/device: 92.184, Tokens/s/device: 13895.014, total_weights: 65536, loss: 5.391, lm_loss: 5.391, perplexity: 219.349 Per train step: Total TFLOPs: 13.59 split as 93.93% learnable weight flops and 6.07% attention flops XPK End: Mon Apr 20 17:34:06 UTC 2026 EXIT_CODE=0