MaxView

← Back to run

Log Summary

XPK Start: Thu Apr 16 22:26:53 UTC 2026
PyTorch was not found. Models won't be available and only tokenizers, configuration and file/data utilities can be used.
2026-04-16 22:27:17.584279: 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)
I0416 22:27:18.117335 133834931758912 max_utils.py:273] Attempting to initialize the jax distributed system...
I0416 22:27:27.157818 133834931758912 distributed.py:149] Starting JAX distributed service on [::]:8482
I0416 22:27:27.160350 133834931758912 distributed.py:172] Connecting to JAX distributed service on mt-05-fp8-fdrcw-slice-job-0-0.mt-05-fp8-fdrcw:8482
I0416 22:27:28.212692 133834931758912 max_utils.py:284] Jax distributed system initialized!
I0416 22:27:34.238610 133834931758912 max_utils.py:800] System Information: Jax Version: 0.9.2
I0416 22:27:34.238714 133834931758912 max_utils.py:801] System Information: Jaxlib Version: 0.9.2
I0416 22:27:34.238754 133834931758912 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
I0416 22:27:34.238787 133834931758912 train_utils.py:377] WARNING: Sequence packing is essentially ignored for synthetic data. Please use a real dataset to use sequence packing.
I0416 22:27:34.905041 133834931758912 maxtext_utils.py:1687] Num_devices: 32, shape (1, 1, 1, 32, 1, 1, 1, 1, 1, 1, 1, 1, 1)
I0416 22:27:35.330931 133834931758912 checkpointing.py:688] Setting up checkpoint logger...
I0416 22:27:35.331049 133834931758912 checkpointing.py:234] Creating checkpoint manager with ocdbt=True and zarr3=True
I0416 22:27:35.331094 133834931758912 pytree_checkpoint_handler.py:592] save_device_host_concurrent_bytes=None
I0416 22:27:35.331304 133834931758912 base_pytree_checkpoint_handler.py:441] Created BasePyTreeCheckpointHandler: use_ocdbt=True, use_zarr3=True, pytree_metadata_options=PyTreeMetadataOptions(support_rich_types=False), array_metadata_store=<orbax.checkpoint._src.metadata.array_metadata_store.Store object at 0x79b8419141a0>, enable_pinned_host_transfer=False, save_concurrent_bytes: 96000000000 (89.4 GiB), restore_concurrent_bytes: 96000000000 (89.4 GiB)
I0416 22:27:38.197897 133834931758912 checkpointing.py:266] Enabling policy for fixed interval checkpointing.
I0416 22:27:38.198127 133834931758912 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 0x79a3f07ea090>}, handler_registry=None
I0416 22:27:38.198371 133834931758912 composite_checkpoint_handler.py:237] Deferred registration for item: "items". Adding handler `<orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeCheckpointHandler object at 0x79a3f07ea090>` 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`.
I0416 22:27:38.198421 133834931758912 composite_checkpoint_handler.py:237] Deferred registration for item: "metrics". Adding handler `<orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonCheckpointHandler object at 0x79a3f07970b0>` 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`.
I0416 22:27:38.198457 133834931758912 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 0x79a3f07ea090>, ('items', <class 'orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeRestoreArgs'>): <orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeCheckpointHandler object at 0x79a3f07ea090>, ('metrics', <class 'orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonSaveArgs'>): <orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonCheckpointHandler object at 0x79a3f07970b0>, ('metrics', <class 'orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonRestoreArgs'>): <orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonCheckpointHandler object at 0x79a3f07970b0>}).
I0416 22:27:38.198787 133834931758912 abstract_checkpointer.py:35] orbax-checkpoint version: 0.11.34
I0416 22:27:38.198876 133834931758912 async_checkpointer.py:192] [process=0][thread=MainThread] Using barrier_sync_fn: <function get_barrier_sync_fn.<locals>._fn at 0x79a3f02211c0> timeout: 1200 secs and primary_host=0 for async checkpoint writes
I0416 22:27:39.787477 133834931758912 checkpoint_manager.py:1812] Found 0 checkpoint steps in gs://lance-maxtext/nnx_ckpt_xpk_feat_nnx_post_train_fixes_20260416_221011/nnx_xpk_feat_nnx_post_train_fixes_20260416_221011_05_fp8/checkpoints
I0416 22:27:39.789776 133834931758912 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=10), 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/nnx_ckpt_xpk_feat_nnx_post_train_fixes_20260416_221011/nnx_xpk_feat_nnx_post_train_fixes_20260416_221011_05_fp8/checkpoints: <orbax.checkpoint.checkpoint_manager.CheckpointManager object at 0x79a3d84fa4b0>
I0416 22:27:39.789905 133834931758912 checkpointing.py:302] Checkpoint manager created!
I0416 22:27:41.029473 133834931758912 checkpointing.py:578] checkpoint manager exists so trying to load this run's existing checkpoint
I0416 22:27:41.029591 133834931758912 checkpointing.py:676] No existing checkpoints found, not restoring checkpoint.
fsdp: 32
I0416 22:27:43.093716 133834931758912 maxtext_utils.py:1805]  decoder/decoder_norm/scale/value
    Shape:     float32[2048]
    Physical:  (None,)
I0416 22:27:43.093832 133834931758912 maxtext_utils.py:1805]  decoder/layers/mlp/wi_0/kernel/value
    Shape:     float32[2048,16,7168]
    Physical:  ('fsdp', None, None)
I0416 22:27:43.093899 133834931758912 maxtext_utils.py:1805]  decoder/layers/mlp/wi_1/kernel/value
    Shape:     float32[2048,16,7168]
    Physical:  ('fsdp', None, None)
I0416 22:27:43.093944 133834931758912 maxtext_utils.py:1805]  decoder/layers/mlp/wo/kernel/value
    Shape:     float32[7168,16,2048]
    Physical:  (None, None, 'fsdp')
I0416 22:27:43.093981 133834931758912 maxtext_utils.py:1805]  decoder/layers/post_self_attention_layer_norm/scale/value
    Shape:     float32[2048,16]
    Physical:  (None, None)
I0416 22:27:43.094016 133834931758912 maxtext_utils.py:1805]  decoder/layers/pre_self_attention_layer_norm/scale/value
    Shape:     float32[2048,16]
    Physical:  (None, None)
I0416 22:27:43.094052 133834931758912 maxtext_utils.py:1805]  decoder/layers/self_attention/key/kernel/value
    Shape:     float32[2048,16,16,128]
    Physical:  ('fsdp', None, None, None)
I0416 22:27:43.094092 133834931758912 maxtext_utils.py:1805]  decoder/layers/self_attention/out/kernel/value
    Shape:     float32[16,16,128,2048]
    Physical:  (None, None, None, 'fsdp')
I0416 22:27:43.094127 133834931758912 maxtext_utils.py:1805]  decoder/layers/self_attention/query/kernel/value
    Shape:     float32[2048,16,16,128]
    Physical:  ('fsdp', None, None, None)
I0416 22:27:43.094170 133834931758912 maxtext_utils.py:1805]  decoder/layers/self_attention/value/kernel/value
    Shape:     float32[2048,16,16,128]
    Physical:  ('fsdp', None, None, None)
I0416 22:27:43.094203 133834931758912 maxtext_utils.py:1805]  decoder/logits_dense/kernel/value
    Shape:     float32[2048,32000]
    Physical:  ('fsdp', None)
I0416 22:27:43.094234 133834931758912 maxtext_utils.py:1805]  token_embedder/embedding/value
    Shape:     float32[32000,2048]
    Physical:  (None, 'fsdp')

I0416 22:27:43.512230 133834931758912 nnx_decoders.py:465] nnx_decoders/carry Logical: bfloat16[32,2048,2048]...................................... ('activation_batch', 'activation_norm_length', 'activation_embed').
I0416 22:27:43.512320 133834931758912 nnx_decoders.py:465] nnx_decoders/carry Physical: bfloat16[32,2048,2048]...................................... ('fsdp', None, None).
I0416 22:27:43.517649 133834931758912 nnx_decoders.py:465] Unknown Logical: bfloat16[32,2048,2048]...................................... ('activation_batch', 'activation_norm_length', 'activation_embed').
I0416 22:27:43.517706 133834931758912 nnx_decoders.py:465] Unknown Physical: bfloat16[32,2048,2048]...................................... ('fsdp', None, None).
I0416 22:27:43.533733 133834931758912 attentions.py:1088] attentions/inputs_q Logical: bfloat16[32,2048,2048]...................................... ('activation_batch', 'activation_attn_length', 'activation_attn_embed').
I0416 22:27:43.533791 133834931758912 attentions.py:1088] attentions/inputs_q Physical: bfloat16[32,2048,2048]...................................... ('fsdp', None, None).
I0416 22:27:43.549223 133834931758912 attentions.py:1089] attentions/inputs_kv Logical: bfloat16[32,2048,2048]...................................... ('activation_batch', 'activation_attn_length', 'activation_attn_embed').
I0416 22:27:43.549281 133834931758912 attentions.py:1089] attentions/inputs_kv Physical: bfloat16[32,2048,2048]...................................... ('fsdp', None, None).
I0416 22:27:43.609292 133834931758912 attentions.py:1154] attentions/query Logical: bfloat16[32,2048,16,128].................................... ('activation_kv_batch', 'activation_attn_length', 'activation_kv_heads', 'activation_kv_head_dim').
I0416 22:27:43.609376 133834931758912 attentions.py:1154] attentions/query Physical: bfloat16[32,2048,16,128].................................... ('fsdp', None, None, None).
I0416 22:27:43.624894 133834931758912 attentions.py:1155] attentions/key Logical: bfloat16[32,2048,16,128].................................... ('activation_kv_batch', 'activation_attn_length', 'activation_kv_heads', 'activation_kv_head_dim').
I0416 22:27:43.624955 133834931758912 attentions.py:1155] attentions/key Physical: bfloat16[32,2048,16,128].................................... ('fsdp', None, None, None).
I0416 22:27:43.640497 133834931758912 attentions.py:1156] attentions/value Logical: bfloat16[32,2048,16,128].................................... ('activation_kv_batch', 'activation_attn_length', 'activation_kv_heads', 'activation_kv_head_dim').
I0416 22:27:43.640559 133834931758912 attentions.py:1156] attentions/value Physical: bfloat16[32,2048,16,128].................................... ('fsdp', None, None, None).
I0416 22:27:43.669761 133834931758912 attentions.py:1197] attentions/out Logical: bfloat16[32,2048,16,128].................................... ('activation_batch', 'activation_attn_length', 'activation_heads', 'activation_kv').
I0416 22:27:43.669852 133834931758912 attentions.py:1197] attentions/out Physical: bfloat16[32,2048,16,128].................................... ('fsdp', None, None, None).
I0416 22:27:43.733855 133834931758912 linears.py:525] linears/x Logical: bfloat16[32,2048,7168]...................................... ('activation_batch', 'activation_length', 'activation_mlp').
I0416 22:27:43.733946 133834931758912 linears.py:525] linears/x Physical: bfloat16[32,2048,7168]...................................... ('fsdp', None, None).
I0416 22:27:57.687587 133834931758912 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.
I0416 22:27:57.698173 133834931758912 metric_logger.py:289] number parameters: 1.104 billion
I0416 22:28:11.646687 133834931758912 checkpointing.py:794] Waiting for step 0 to finish before checkpoint...
I0416 22:28:11.814460 133834931758912 checkpointing.py:798] Waited 0.1677563190460205 seconds for step 0 to finish before starting checkpointing.
I0416 22:28:11.816856 133834931758912 checkpoint_manager.py:2009] [process=0][thread=MainThread][wait_until_finished] No Save Finalize thread to wait for. Returning.
I0416 22:28:11.818814 133834931758912 checkpoint_manager.py:1512] [process=0] Saving checkpoint at step 0
I0416 22:28:11.820236 133834931758912 event_tracking.py:70] [process=0] [async] Started save checkpoint @ gs://lance-maxtext/nnx_ckpt_xpk_feat_nnx_post_train_fixes_20260416_221011/nnx_xpk_feat_nnx_post_train_fixes_20260416_221011_05_fp8/checkpoints/0.
I0416 22:28:12.547529 133834931758912 signaling_client.py:364] Using JaxDistributedSignalingClient
I0416 22:28:12.578120 133834931758912 jax_array_handlers.py:360] Scheduling D2H of 153 prioritized jax.Array.
I0416 22:28:12.578202 133834931758912 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
I0416 22:28:12.836602 133705435682560 atomicity.py:140] Creating tmp directory gs://lance-maxtext/nnx_ckpt_xpk_feat_nnx_post_train_fixes_20260416_221011/nnx_xpk_feat_nnx_post_train_fixes_20260416_221011_05_fp8/checkpoints/0
I0416 22:28:12.978899 133834931758912 base_pytree_checkpoint_handler.py:154] [process=0][thread=MainThread] Initiated "orbax.checkpoint._src.serialization.jax_array_handlers.ArrayHandler".serialize. Time taken: 0.403213s
I0416 22:28:12.979769 133834931758912 base_pytree_checkpoint_handler.py:130] [process=0] /jax/orbax/write/blocking_gbytes_per_sec: 3.635 GiB/s (total gbytes: 1.5 GiB) (time elapsed: 0.4244096279144287 s) (per-host)
I0416 22:28:12.979848 133834931758912 base_pytree_checkpoint_handler.py:768] [process=0][thread=MainThread] Initiated Pytree async_save. Time taken: 0.424497s (batch_requests_ready=0.006224s, total_serialization_initiated=0.417487s, others=0.000785s)
I0416 22:28:12.979939 133834931758912 composite_checkpoint_handler.py:715] [process=0][thread=MainThread] Initiated CompositeCheckpointHandler.async_save. Time taken: 0.428468s (all_items=0.000018s, per_item={'items': '0.00001764'}, temp_paths=0.428450)
I0416 22:28:12.980764 133834931758912 event_tracking.py:125] [process=0] [async] Finished blocking save in 1.16 seconds. Continuing save @ gs://lance-maxtext/nnx_ckpt_xpk_feat_nnx_post_train_fixes_20260416_221011/nnx_xpk_feat_nnx_post_train_fixes_20260416_221011_05_fp8/checkpoints/0.
I0416 22:28:12.981134 133705444075264 async_checkpointer.py:76] [process=0][thread=async_save] Background save thread started. Deadline for this save operation is 2026-04-16 22:48:12.981095
I0416 22:28:12.994947 133834931758912 checkpoint_manager.py:1560] [process=0][thread=MainThread][step=0] Starting CheckpointManager Save Finalize thread=save_finalize
I0416 22:28:12.995268 133703732815616 async_checkpointer.py:280] [process=0][thread=save_finalize] Waiting for background save thread=async_save.
I0416 22:28:12.995423 133834931758912 standard_logger.py:34] {'step': 0, 'event_type': 'save', 'directory': 'gs://lance-maxtext/nnx_ckpt_xpk_feat_nnx_post_train_fixes_20260416_221011/nnx_xpk_feat_nnx_post_train_fixes_20260416_221011_05_fp8/checkpoints', 'reached_preemption': False, 'preemption_received_at': None, 'synchronous': False, 'wait_for_prev_start_time': 1776378491.8168206, 'wait_for_prev_duration_secs': 7.557868957519531e-05, 'time_between_consecutive_saves_sec': None, 'checkpointer_blocking_start_time': 1776378491.8188643, 'checkpointer_blocking_duration_secs': 1.1624228954315186, 'get_old_steps_start_time': 1776378492.9813128, 'get_old_steps_duration_secs': 2.8371810913085938e-05, 'checkpoint_manager_blocking_start_time': 1776378491.8149562, 'checkpoint_manager_blocking_duration_secs': 1.1804273128509521}
I0416 22:28:12.995552 133834931758912 checkpointing.py:409] Started an asynchronous checkpoint save for step 0
I0416 22:28:12.995627 133834931758912 max_utils.py:750] 
Memstats: After params initialized:
I0416 22:28:12.995677 133834931758912 max_utils.py:756] 	Using (GB) 0.45 / 31.25 (1.440000%) on TPU_0(process=0,(0,0,0,0))
I0416 22:28:12.995712 133834931758912 max_utils.py:756] 	Using (GB) 0.45 / 31.25 (1.440000%) on TPU_1(process=0,(1,0,0,0))
I0416 22:28:12.995743 133834931758912 max_utils.py:756] 	Using (GB) 0.45 / 31.25 (1.440000%) on TPU_4(process=0,(0,1,0,0))
I0416 22:28:12.995770 133834931758912 max_utils.py:756] 	Using (GB) 0.45 / 31.25 (1.440000%) on TPU_5(process=0,(1,1,0,0))
I0416 22:28:13.352055 133834931758912 metric_logger.py:185] completed step: 0, seconds: 13.528, TFLOP/s/device: 1.004, Tokens/s/device: 151.384, total_weights: 65536, loss: 10.872
I0416 22:28:13.355266 133834931758912 metric_logger.py:269] To see full metrics 'tensorboard --logdir=gs://lance-maxtext/nnx_ckpt_xpk_feat_nnx_post_train_fixes_20260416_221011/nnx_xpk_feat_nnx_post_train_fixes_20260416_221011_05_fp8/tensorboard/'
I0416 22:28:13.778582 133834931758912 metric_logger.py:185] completed step: 1, seconds: 1.699, TFLOP/s/device: 7.997, Tokens/s/device: 1205.372, total_weights: 65536, loss: 10.872
I0416 22:28:13.924416 133834931758912 metric_logger.py:185] completed step: 2, seconds: 0.428, TFLOP/s/device: 31.738, Tokens/s/device: 4783.895, total_weights: 65536, loss: 10.856
I0416 22:28:14.081967 133834931758912 metric_logger.py:185] completed step: 3, seconds: 0.017, TFLOP/s/device: 794.244, Tokens/s/device: 119717.075, total_weights: 65536, loss: 10.824
I0416 22:28:15.073064 133704403904256 atomicity.py:140] Creating tmp directory gs://lance-maxtext/nnx_ckpt_xpk_feat_nnx_post_train_fixes_20260416_221011/nnx_xpk_feat_nnx_post_train_fixes_20260416_221011_05_fp8/checkpoints/0/items
I0416 22:28:15.213278 133703706089216 checkpoint.py:188] Wrote Metadata={'item_handlers': None, 'metrics': {}, 'performance_metrics': {}, 'init_timestamp_nsecs': 1776378494813364592, 'commit_timestamp_nsecs': None, 'custom_metadata': {}}, json={"item_handlers": null, "metrics": {}, "performance_metrics": {}, "init_timestamp_nsecs": 1776378494813364592, "commit_timestamp_nsecs": null, "custom_metadata": {}} to gs://lance-maxtext/nnx_ckpt_xpk_feat_nnx_post_train_fixes_20260416_221011/nnx_xpk_feat_nnx_post_train_fixes_20260416_221011_05_fp8/checkpoints/0/_CHECKPOINT_METADATA
I0416 22:28:16.189726    2813 google_auth_provider.cc:181] Running on GCE, using service account 562977990677-compute@developer.gserviceaccount.com
I0416 22:28:18.224572 133704387118848 array_metadata_store.py:203] [process=0][thread=array_type_handler] Wrote 153 array_metadata.ArrayMetadata to gs://lance-maxtext/nnx_ckpt_xpk_feat_nnx_post_train_fixes_20260416_221011/nnx_xpk_feat_nnx_post_train_fixes_20260416_221011_05_fp8/checkpoints/0/items/array_metadatas/process_0
I0416 22:28:35.368196 133703741208320 base_pytree_checkpoint_handler.py:1282] [process=0][thread=write_metadata_after_commits] Commit + Array metadata written. Time taken: 19.191023s (commit=18.000859s, array_metadata_write=1.190165s)
I0416 22:28:35.369503 133705444075264 base_pytree_checkpoint_handler.py:130] [process=0] /jax/orbax/write/gbytes_per_sec: 69.245 MiB/s (total gbytes: 1.5 GiB) (time elapsed: 22.814112663269043 s) (per-host)
I0416 22:28:35.369626 133705444075264 async_checkpointer.py:90] [process=0][thread=async_save] 3 Handler Commit operations completed. Time taken: 22.388376s.
I0416 22:28:36.015127 133705444075264 checkpoint.py:228] Read Metadata={'item_handlers': None, 'metrics': {}, 'performance_metrics': {}, 'init_timestamp_nsecs': 1776378494813364592, 'commit_timestamp_nsecs': None, 'custom_metadata': {}} from gs://lance-maxtext/nnx_ckpt_xpk_feat_nnx_post_train_fixes_20260416_221011/nnx_xpk_feat_nnx_post_train_fixes_20260416_221011_05_fp8/checkpoints/0/_CHECKPOINT_METADATA
I0416 22:28:36.730241 133705444075264 array_metadata_store.py:411] [process=0][thread=async_save] Validated ArrayMetadata from all 8 hosts. Time taken: 0.001018s.
I0416 22:28:36.748619 133703706089216 checkpoint.py:247] Updated Metadata={'item_handlers': {'items': 'orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeCheckpointHandler'}, 'metrics': {}, 'performance_metrics': {}, 'init_timestamp_nsecs': 1776378494813364592, 'commit_timestamp_nsecs': None, 'custom_metadata': {}} to gs://lance-maxtext/nnx_ckpt_xpk_feat_nnx_post_train_fixes_20260416_221011/nnx_xpk_feat_nnx_post_train_fixes_20260416_221011_05_fp8/checkpoints/0/_CHECKPOINT_METADATA
I0416 22:28:37.524624 133705444075264 ocdbt_utils.py:49] Param validation support for Zarr3 will be added later (b/362328389).
I0416 22:28:38.473472 133705444075264 base_pytree_checkpoint_handler.py:1406] [process=0][thread=async_save] Pytree save finalize (merge_ocdbt + ArrayMetadata validation) completed. Time taken: 2.333722s. use_zarr3=True, enable_post_merge_validation=True, directory=gs://lance-maxtext/nnx_ckpt_xpk_feat_nnx_post_train_fixes_20260416_221011/nnx_xpk_feat_nnx_post_train_fixes_20260416_221011_05_fp8/checkpoints/0/items
I0416 22:28:38.475116 133705444075264 atomicity.py:666] Finalizing gs://lance-maxtext/nnx_ckpt_xpk_feat_nnx_post_train_fixes_20260416_221011/nnx_xpk_feat_nnx_post_train_fixes_20260416_221011_05_fp8/checkpoints/0/items
I0416 22:28:38.490858 133834931758912 metric_logger.py:185] completed step: 4, seconds: 0.146, TFLOP/s/device: 92.774, Tokens/s/device: 13983.913, total_weights: 65536, loss: 10.793
I0416 22:28:38.503054 133834931758912 metric_logger.py:185] completed step: 5, seconds: 0.158, TFLOP/s/device: 86.077, Tokens/s/device: 12974.507, total_weights: 65536, loss: 10.763
I0416 22:28:38.654716 133834931758912 metric_logger.py:185] completed step: 6, seconds: 24.406, TFLOP/s/device: 0.557, Tokens/s/device: 83.913, total_weights: 65536, loss: 10.733
I0416 22:28:38.811822 133834931758912 metric_logger.py:185] completed step: 7, seconds: 0.010, TFLOP/s/device: 1363.896, Tokens/s/device: 205581.209, total_weights: 65536, loss: 10.706
I0416 22:28:38.968728 133834931758912 metric_logger.py:185] completed step: 8, seconds: 0.153, TFLOP/s/device: 88.734, Tokens/s/device: 13374.956, total_weights: 65536, loss: 10.680
I0416 22:28:38.978794 133834931758912 checkpointing.py:794] Waiting for step 10 to finish before checkpoint...
I0416 22:28:38.990779 133705444075264 atomicity.py:666] Finalizing gs://lance-maxtext/nnx_ckpt_xpk_feat_nnx_post_train_fixes_20260416_221011/nnx_xpk_feat_nnx_post_train_fixes_20260416_221011_05_fp8/checkpoints/0
I0416 22:28:39.283460 133834931758912 checkpointing.py:798] Waited 0.3046376705169678 seconds for step 10 to finish before starting checkpointing.
I0416 22:28:39.285661 133834931758912 checkpoint_manager.py:2020] [process=0][thread=MainThread][step=0][wait_until_finished] Waiting for Save Finalize thread (save_finalize) to complete.
I0416 22:28:40.362563 133705444075264 atomicity.py:847] [process=0][thread=async_save] Finished saving checkpoint (finalized tmp dir) to `gs://lance-maxtext/nnx_ckpt_xpk_feat_nnx_post_train_fixes_20260416_221011/nnx_xpk_feat_nnx_post_train_fixes_20260416_221011_05_fp8/checkpoints/0`.
I0416 22:28:40.363447 133705444075264 event_tracking.py:138] [process=0] [async] Finished save (blocking + background) in 28.54 seconds @ gs://lance-maxtext/nnx_ckpt_xpk_feat_nnx_post_train_fixes_20260416_221011/nnx_xpk_feat_nnx_post_train_fixes_20260416_221011_05_fp8/checkpoints/0
I0416 22:28:40.365012 133705444075264 async_checkpointer.py:160] [process=0][thread=async_save] Background save thread done. Time taken: 27.383760s.
I0416 22:28:40.365236 133703732815616 async_checkpointer.py:288] [process=0][thread=save_finalize] Done with waiting for background save thread=async_save.
I0416 22:28:40.365358 133703732815616 async_checkpointer.py:298] [process=0][thread=save_finalize] No errors found in background save thread=async_save.
I0416 22:28:40.365421 133703732815616 checkpoint_manager.py:2137] [process=0][thread=save_finalize][step=0] CheckpointManager Save Finalize is syncing with other hosts...
I0416 22:28:40.367907 133703732815616 checkpoint_manager.py:2146] [process=0][thread=save_finalize][step=0] CheckpointManager Save Finalize is done on all hosts.
I0416 22:28:40.368044 133834931758912 checkpoint_manager.py:2032] [process=0][thread=MainThread][step=0][wait_until_finished] Done waiting for Save Finalize thread (save_finalize) running at step=0.
W0416 22:28:40.368126 133834931758912 checkpoint_manager.py:1452] Waiting for previous save to complete took 1.082469 seconds. If this number is high, consider checkpointing less frequently.
I0416 22:28:40.369861 133834931758912 checkpoint_manager.py:1512] [process=0] Saving checkpoint at step 10
I0416 22:28:40.371754 133834931758912 event_tracking.py:70] [process=0] [async] Started save checkpoint @ gs://lance-maxtext/nnx_ckpt_xpk_feat_nnx_post_train_fixes_20260416_221011/nnx_xpk_feat_nnx_post_train_fixes_20260416_221011_05_fp8/checkpoints/10.
I0416 22:28:41.107132 133834931758912 jax_array_handlers.py:360] Scheduling D2H of 153 prioritized jax.Array.
I0416 22:28:41.107235 133834931758912 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
I0416 22:28:41.151368 133834931758912 base_pytree_checkpoint_handler.py:154] [process=0][thread=MainThread] Initiated "orbax.checkpoint._src.serialization.jax_array_handlers.ArrayHandler".serialize. Time taken: 0.046566s
I0416 22:28:41.152188 133834931758912 base_pytree_checkpoint_handler.py:130] [process=0] /jax/orbax/write/blocking_gbytes_per_sec: 24.071 GiB/s (total gbytes: 1.5 GiB) (time elapsed: 0.06409025192260742 s) (per-host)
I0416 22:28:41.152267 133834931758912 base_pytree_checkpoint_handler.py:768] [process=0][thread=MainThread] Initiated Pytree async_save. Time taken: 0.064186s (batch_requests_ready=0.005453s, total_serialization_initiated=0.057980s, others=0.000753s)
I0416 22:28:41.152360 133834931758912 composite_checkpoint_handler.py:715] [process=0][thread=MainThread] Initiated CompositeCheckpointHandler.async_save. Time taken: 0.068589s (all_items=0.000016s, per_item={'items': '0.00001597'}, temp_paths=0.068573)
I0416 22:28:41.153043 133834931758912 event_tracking.py:125] [process=0] [async] Finished blocking save in 0.78 seconds. Continuing save @ gs://lance-maxtext/nnx_ckpt_xpk_feat_nnx_post_train_fixes_20260416_221011/nnx_xpk_feat_nnx_post_train_fixes_20260416_221011_05_fp8/checkpoints/10.
I0416 22:28:41.153381 133703732815616 async_checkpointer.py:76] [process=0][thread=async_save] Background save thread started. Deadline for this save operation is 2026-04-16 22:48:41.153342
I0416 22:28:41.155377 133834931758912 checkpoint_manager.py:1560] [process=0][thread=MainThread][step=10] Starting CheckpointManager Save Finalize thread=save_finalize
I0416 22:28:41.155634 133701719549696 async_checkpointer.py:280] [process=0][thread=save_finalize] Waiting for background save thread=async_save.
I0416 22:28:41.155799 133834931758912 standard_logger.py:34] {'step': 10, 'event_type': 'save', 'directory': 'gs://lance-maxtext/nnx_ckpt_xpk_feat_nnx_post_train_fixes_20260416_221011/nnx_xpk_feat_nnx_post_train_fixes_20260416_221011_05_fp8/checkpoints', 'reached_preemption': False, 'preemption_received_at': None, 'synchronous': False, 'wait_for_prev_start_time': 1776378519.285628, 'wait_for_prev_duration_secs': 1.0824694633483887, 'time_between_consecutive_saves_sec': None, 'checkpointer_blocking_start_time': 1776378520.3699, 'checkpointer_blocking_duration_secs': 0.7836349010467529, 'get_old_steps_start_time': 1776378521.1535616, 'get_old_steps_duration_secs': 3.457069396972656e-05, 'checkpoint_manager_blocking_start_time': 1776378519.2839305, 'checkpoint_manager_blocking_duration_secs': 1.871833324432373}
I0416 22:28:41.155930 133834931758912 checkpointing.py:409] Started an asynchronous checkpoint save for step 10
I0416 22:28:41.156347 133834931758912 metric_logger.py:185] completed step: 9, seconds: 0.157, TFLOP/s/device: 86.354, Tokens/s/device: 13016.232, total_weights: 65536, loss: 10.656
I0416 22:28:41.170915 133834931758912 metric_logger.py:185] completed step: 10, seconds: 0.156, TFLOP/s/device: 86.849, Tokens/s/device: 13090.863, total_weights: 65536, loss: 10.636
I0416 22:28:41.319802 133834931758912 metric_logger.py:185] completed step: 11, seconds: 2.189, TFLOP/s/device: 6.206, Tokens/s/device: 935.403, total_weights: 65536, loss: 10.618
I0416 22:28:41.332770 133705444075264 atomicity.py:140] Creating tmp directory gs://lance-maxtext/nnx_ckpt_xpk_feat_nnx_post_train_fixes_20260416_221011/nnx_xpk_feat_nnx_post_train_fixes_20260416_221011_05_fp8/checkpoints/10
I0416 22:28:41.477323 133834931758912 metric_logger.py:185] completed step: 12, seconds: 0.012, TFLOP/s/device: 1125.321, Tokens/s/device: 169620.673, total_weights: 65536, loss: 10.602
I0416 22:28:41.633632 133834931758912 metric_logger.py:185] completed step: 13, seconds: 0.150, TFLOP/s/device: 90.627, Tokens/s/device: 13660.255, total_weights: 65536, loss: 10.588
I0416 22:28:41.790455 133834931758912 metric_logger.py:185] completed step: 14, seconds: 0.157, TFLOP/s/device: 86.381, Tokens/s/device: 13020.287, total_weights: 65536, loss: 10.577
I0416 22:28:41.947448 133834931758912 metric_logger.py:185] completed step: 15, seconds: 0.156, TFLOP/s/device: 86.838, Tokens/s/device: 13089.189, total_weights: 65536, loss: 10.568
I0416 22:28:42.447746 133834931758912 metric_logger.py:185] completed step: 16, seconds: 0.157, TFLOP/s/device: 86.665, Tokens/s/device: 13063.141, total_weights: 65536, loss: 10.562
I0416 22:28:42.600104 133834931758912 metric_logger.py:185] completed step: 17, seconds: 0.646, TFLOP/s/device: 21.026, Tokens/s/device: 3169.209, total_weights: 65536, loss: 10.556
I0416 22:28:42.756672 133834931758912 metric_logger.py:185] completed step: 18, seconds: 0.010, TFLOP/s/device: 1415.179, Tokens/s/device: 213311.113, total_weights: 65536, loss: 10.552
I0416 22:28:42.915709 133834931758912 checkpointing.py:794] Waiting for step 19 to finish before checkpoint...
I0416 22:28:42.917007 133834931758912 checkpointing.py:798] Waited 0.001312255859375 seconds for step 19 to finish before starting checkpointing.
I0416 22:28:42.919192 133834931758912 checkpoint_manager.py:2020] [process=0][thread=MainThread][step=10][wait_until_finished] Waiting for Save Finalize thread (save_finalize) to complete.
I0416 22:28:43.565774 133704387118848 atomicity.py:140] Creating tmp directory gs://lance-maxtext/nnx_ckpt_xpk_feat_nnx_post_train_fixes_20260416_221011/nnx_xpk_feat_nnx_post_train_fixes_20260416_221011_05_fp8/checkpoints/10/items
I0416 22:28:47.290729 133702751328000 array_metadata_store.py:203] [process=0][thread=array_type_handler] Wrote 153 array_metadata.ArrayMetadata to gs://lance-maxtext/nnx_ckpt_xpk_feat_nnx_post_train_fixes_20260416_221011/nnx_xpk_feat_nnx_post_train_fixes_20260416_221011_05_fp8/checkpoints/10/items/array_metadatas/process_0
I0416 22:29:05.568667 133702207530752 base_pytree_checkpoint_handler.py:1282] [process=0][thread=write_metadata_after_commits] Commit + Array metadata written. Time taken: 20.834008s (commit=19.211728s, array_metadata_write=1.622280s)
I0416 22:29:05.569965 133703732815616 base_pytree_checkpoint_handler.py:130] [process=0] /jax/orbax/write/gbytes_per_sec: 64.528 MiB/s (total gbytes: 1.5 GiB) (time elapsed: 24.481837034225464 s) (per-host)
I0416 22:29:05.570098 133703732815616 async_checkpointer.py:90] [process=0][thread=async_save] 3 Handler Commit operations completed. Time taken: 24.416600s.
I0416 22:29:07.601123 133703732815616 ocdbt_utils.py:49] Param validation support for Zarr3 will be added later (b/362328389).
I0416 22:29:07.727313 133703732815616 array_metadata_store.py:411] [process=0][thread=async_save] Validated ArrayMetadata from all 8 hosts. Time taken: 0.000883s.
I0416 22:29:08.697062 133703732815616 base_pytree_checkpoint_handler.py:1406] [process=0][thread=async_save] Pytree save finalize (merge_ocdbt + ArrayMetadata validation) completed. Time taken: 1.912831s. use_zarr3=True, enable_post_merge_validation=True, directory=gs://lance-maxtext/nnx_ckpt_xpk_feat_nnx_post_train_fixes_20260416_221011/nnx_xpk_feat_nnx_post_train_fixes_20260416_221011_05_fp8/checkpoints/10/items
I0416 22:29:08.698866 133703732815616 atomicity.py:666] Finalizing gs://lance-maxtext/nnx_ckpt_xpk_feat_nnx_post_train_fixes_20260416_221011/nnx_xpk_feat_nnx_post_train_fixes_20260416_221011_05_fp8/checkpoints/10/items
I0416 22:29:09.227012 133703732815616 atomicity.py:666] Finalizing gs://lance-maxtext/nnx_ckpt_xpk_feat_nnx_post_train_fixes_20260416_221011/nnx_xpk_feat_nnx_post_train_fixes_20260416_221011_05_fp8/checkpoints/10
I0416 22:29:11.761966 133703732815616 atomicity.py:847] [process=0][thread=async_save] Finished saving checkpoint (finalized tmp dir) to `gs://lance-maxtext/nnx_ckpt_xpk_feat_nnx_post_train_fixes_20260416_221011/nnx_xpk_feat_nnx_post_train_fixes_20260416_221011_05_fp8/checkpoints/10`.
I0416 22:29:11.762903 133703732815616 event_tracking.py:138] [process=0] [async] Finished save (blocking + background) in 31.39 seconds @ gs://lance-maxtext/nnx_ckpt_xpk_feat_nnx_post_train_fixes_20260416_221011/nnx_xpk_feat_nnx_post_train_fixes_20260416_221011_05_fp8/checkpoints/10
I0416 22:29:11.764376 133703732815616 async_checkpointer.py:160] [process=0][thread=async_save] Background save thread done. Time taken: 30.610876s.
I0416 22:29:11.764566 133701719549696 async_checkpointer.py:288] [process=0][thread=save_finalize] Done with waiting for background save thread=async_save.
I0416 22:29:11.764686 133701719549696 async_checkpointer.py:298] [process=0][thread=save_finalize] No errors found in background save thread=async_save.
I0416 22:29:11.764750 133701719549696 checkpoint_manager.py:2137] [process=0][thread=save_finalize][step=10] CheckpointManager Save Finalize is syncing with other hosts...
I0416 22:29:11.767618 133701719549696 checkpoint_manager.py:2146] [process=0][thread=save_finalize][step=10] CheckpointManager Save Finalize is done on all hosts.
I0416 22:29:11.767769 133834931758912 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.
W0416 22:29:11.767863 133834931758912 checkpoint_manager.py:1452] Waiting for previous save to complete took 28.848666 seconds. If this number is high, consider checkpointing less frequently.
I0416 22:29:11.769972 133834931758912 checkpoint_manager.py:1512] [process=0] Saving checkpoint at step 19
I0416 22:29:11.772211 133834931758912 event_tracking.py:70] [process=0] [async] Started save checkpoint @ gs://lance-maxtext/nnx_ckpt_xpk_feat_nnx_post_train_fixes_20260416_221011/nnx_xpk_feat_nnx_post_train_fixes_20260416_221011_05_fp8/checkpoints/19.
I0416 22:29:12.488119 133834931758912 jax_array_handlers.py:360] Scheduling D2H of 153 prioritized jax.Array.
I0416 22:29:12.488282 133834931758912 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
I0416 22:29:12.528977 133834931758912 base_pytree_checkpoint_handler.py:154] [process=0][thread=MainThread] Initiated "orbax.checkpoint._src.serialization.jax_array_handlers.ArrayHandler".serialize. Time taken: 0.043227s
I0416 22:29:12.529848 133834931758912 base_pytree_checkpoint_handler.py:130] [process=0] /jax/orbax/write/blocking_gbytes_per_sec: 25.284 GiB/s (total gbytes: 1.5 GiB) (time elapsed: 0.06101584434509277 s) (per-host)
I0416 22:29:12.529920 133834931758912 base_pytree_checkpoint_handler.py:768] [process=0][thread=MainThread] Initiated Pytree async_save. Time taken: 0.061117s (batch_requests_ready=0.005843s, total_serialization_initiated=0.054488s, others=0.000786s)
I0416 22:29:12.530015 133834931758912 composite_checkpoint_handler.py:715] [process=0][thread=MainThread] Initiated CompositeCheckpointHandler.async_save. Time taken: 0.065112s (all_items=0.000010s, per_item={'items': '0.00001049'}, temp_paths=0.065101)
I0416 22:29:12.530720 133834931758912 event_tracking.py:125] [process=0] [async] Finished blocking save in 0.76 seconds. Continuing save @ gs://lance-maxtext/nnx_ckpt_xpk_feat_nnx_post_train_fixes_20260416_221011/nnx_xpk_feat_nnx_post_train_fixes_20260416_221011_05_fp8/checkpoints/19.
I0416 22:29:12.531060 133701719549696 async_checkpointer.py:76] [process=0][thread=async_save] Background save thread started. Deadline for this save operation is 2026-04-16 22:49:12.531023
I0416 22:29:12.533040 133834931758912 checkpoint_manager.py:1560] [process=0][thread=MainThread][step=19] Starting CheckpointManager Save Finalize thread=save_finalize
I0416 22:29:12.533291 133702207530752 async_checkpointer.py:280] [process=0][thread=save_finalize] Waiting for background save thread=async_save.
I0416 22:29:12.533439 133834931758912 standard_logger.py:34] {'step': 19, 'event_type': 'save', 'directory': 'gs://lance-maxtext/nnx_ckpt_xpk_feat_nnx_post_train_fixes_20260416_221011/nnx_xpk_feat_nnx_post_train_fixes_20260416_221011_05_fp8/checkpoints', 'reached_preemption': False, 'preemption_received_at': None, 'synchronous': False, 'wait_for_prev_start_time': 1776378522.9191637, 'wait_for_prev_duration_secs': 28.848666191101074, 'time_between_consecutive_saves_sec': 2.551217555999756, 'checkpointer_blocking_start_time': 1776378551.770011, 'checkpointer_blocking_duration_secs': 0.7611918449401855, 'get_old_steps_start_time': 1776378552.5312212, 'get_old_steps_duration_secs': 2.4557113647460938e-05, 'checkpoint_manager_blocking_start_time': 1776378522.917312, 'checkpoint_manager_blocking_duration_secs': 29.616095066070557}
I0416 22:29:12.533647 133834931758912 checkpointing.py:409] Started an asynchronous checkpoint save for step 19
I0416 22:29:12.533695 133834931758912 checkpoint_manager.py:2020] [process=0][thread=MainThread][step=19][wait_until_finished] Waiting for Save Finalize thread (save_finalize) to complete.
I0416 22:29:12.714957 133702801684224 atomicity.py:140] Creating tmp directory gs://lance-maxtext/nnx_ckpt_xpk_feat_nnx_post_train_fixes_20260416_221011/nnx_xpk_feat_nnx_post_train_fixes_20260416_221011_05_fp8/checkpoints/19
I0416 22:29:14.539184 133700092151552 atomicity.py:140] Creating tmp directory gs://lance-maxtext/nnx_ckpt_xpk_feat_nnx_post_train_fixes_20260416_221011/nnx_xpk_feat_nnx_post_train_fixes_20260416_221011_05_fp8/checkpoints/19/items
I0416 22:29:17.440584 133702751328000 array_metadata_store.py:203] [process=0][thread=array_type_handler] Wrote 153 array_metadata.ArrayMetadata to gs://lance-maxtext/nnx_ckpt_xpk_feat_nnx_post_train_fixes_20260416_221011/nnx_xpk_feat_nnx_post_train_fixes_20260416_221011_05_fp8/checkpoints/19/items/array_metadatas/process_0
I0416 22:29:37.488055 133702264813312 base_pytree_checkpoint_handler.py:1282] [process=0][thread=write_metadata_after_commits] Commit + Array metadata written. Time taken: 21.820948s (commit=19.748705s, array_metadata_write=2.072243s)
I0416 22:29:37.489359 133701719549696 base_pytree_checkpoint_handler.py:130] [process=0] /jax/orbax/write/gbytes_per_sec: 63.139 MiB/s (total gbytes: 1.5 GiB) (time elapsed: 25.020507097244263 s) (per-host)
I0416 22:29:37.489483 133701719549696 async_checkpointer.py:90] [process=0][thread=async_save] 3 Handler Commit operations completed. Time taken: 24.958310s.
I0416 22:29:39.691758 133701719549696 array_metadata_store.py:411] [process=0][thread=async_save] Validated ArrayMetadata from all 8 hosts. Time taken: 0.000819s.
I0416 22:29:39.854488 133701719549696 ocdbt_utils.py:49] Param validation support for Zarr3 will be added later (b/362328389).
I0416 22:29:40.475225 133701719549696 base_pytree_checkpoint_handler.py:1406] [process=0][thread=async_save] Pytree save finalize (merge_ocdbt + ArrayMetadata validation) completed. Time taken: 2.072071s. use_zarr3=True, enable_post_merge_validation=True, directory=gs://lance-maxtext/nnx_ckpt_xpk_feat_nnx_post_train_fixes_20260416_221011/nnx_xpk_feat_nnx_post_train_fixes_20260416_221011_05_fp8/checkpoints/19/items
I0416 22:29:40.476969 133701719549696 atomicity.py:666] Finalizing gs://lance-maxtext/nnx_ckpt_xpk_feat_nnx_post_train_fixes_20260416_221011/nnx_xpk_feat_nnx_post_train_fixes_20260416_221011_05_fp8/checkpoints/19/items
I0416 22:29:40.989729 133701719549696 atomicity.py:666] Finalizing gs://lance-maxtext/nnx_ckpt_xpk_feat_nnx_post_train_fixes_20260416_221011/nnx_xpk_feat_nnx_post_train_fixes_20260416_221011_05_fp8/checkpoints/19
I0416 22:29:42.834736 133701719549696 atomicity.py:847] [process=0][thread=async_save] Finished saving checkpoint (finalized tmp dir) to `gs://lance-maxtext/nnx_ckpt_xpk_feat_nnx_post_train_fixes_20260416_221011/nnx_xpk_feat_nnx_post_train_fixes_20260416_221011_05_fp8/checkpoints/19`.
I0416 22:29:42.835593 133701719549696 event_tracking.py:138] [process=0] [async] Finished save (blocking + background) in 31.07 seconds @ gs://lance-maxtext/nnx_ckpt_xpk_feat_nnx_post_train_fixes_20260416_221011/nnx_xpk_feat_nnx_post_train_fixes_20260416_221011_05_fp8/checkpoints/19
I0416 22:29:42.837216 133701719549696 async_checkpointer.py:160] [process=0][thread=async_save] Background save thread done. Time taken: 30.306041s.
I0416 22:29:42.837406 133702207530752 async_checkpointer.py:288] [process=0][thread=save_finalize] Done with waiting for background save thread=async_save.
I0416 22:29:42.837521 133702207530752 async_checkpointer.py:298] [process=0][thread=save_finalize] No errors found in background save thread=async_save.
I0416 22:29:42.837574 133702207530752 checkpoint_manager.py:2137] [process=0][thread=save_finalize][step=19] CheckpointManager Save Finalize is syncing with other hosts...
I0416 22:29:42.839933 133702207530752 checkpoint_manager.py:2146] [process=0][thread=save_finalize][step=19] CheckpointManager Save Finalize is done on all hosts.
I0416 22:29:42.840102 133834931758912 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.
I0416 22:29:42.840221 133834931758912 checkpoint_manager.py:2009] [process=0][thread=MainThread][wait_until_finished] No Save Finalize thread to wait for. Returning.
I0416 22:29:42.840859 133834931758912 metric_logger.py:185] completed step: 19, seconds: 0.154, TFLOP/s/device: 88.350, Tokens/s/device: 13317.034, total_weights: 65536, loss: 10.549
Per train step:
 Total TFLOPs: 13.59 
 split as 93.93% learnable weight flops and 6.07% attention flops
XPK End: Thu Apr 16 22:29:55 UTC 2026
EXIT_CODE=0