XPK Start: Fri Apr 17 00:47:12 UTC 2026 PyTorch was not found. Models won't be available and only tokenizers, configuration and file/data utilities can be used. 2026-04-17 00:47:36.058702: 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) I0417 00:47:36.588848 136175587092288 max_utils.py:273] Attempting to initialize the jax distributed system... I0417 00:47:45.627669 136175587092288 distributed.py:149] Starting JAX distributed service on [::]:8482 I0417 00:47:45.629976 136175587092288 distributed.py:172] Connecting to JAX distributed service on mt-04-int8-0tvi2-slice-job-0-0.mt-04-int8-0tvi2:8482 I0417 00:47:46.662917 136175587092288 max_utils.py:284] Jax distributed system initialized! I0417 00:47:52.693361 136175587092288 max_utils.py:800] System Information: Jax Version: 0.9.2 I0417 00:47:52.693462 136175587092288 max_utils.py:801] System Information: Jaxlib Version: 0.9.2 I0417 00:47:52.693502 136175587092288 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 I0417 00:47:52.693537 136175587092288 train_utils.py:377] WARNING: Sequence packing is essentially ignored for synthetic data. Please use a real dataset to use sequence packing. I0417 00:47:53.370528 136175587092288 maxtext_utils.py:1687] Num_devices: 32, shape (1, 1, 1, 32, 1, 1, 1, 1, 1, 1, 1, 1, 1) I0417 00:47:53.924564 136175587092288 checkpointing.py:688] Setting up checkpoint logger... I0417 00:47:53.924683 136175587092288 checkpointing.py:234] Creating checkpoint manager with ocdbt=True and zarr3=True I0417 00:47:53.924728 136175587092288 pytree_checkpoint_handler.py:592] save_device_host_concurrent_bytes=None I0417 00:47:53.924955 136175587092288 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 0x7bd921b48440>, enable_pinned_host_transfer=False, save_concurrent_bytes: 96000000000 (89.4 GiB), restore_concurrent_bytes: 96000000000 (89.4 GiB) I0417 00:47:57.191517 136175587092288 checkpointing.py:266] Enabling policy for fixed interval checkpointing. I0417 00:47:57.191748 136175587092288 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 0x7bc50c4a89b0>}, handler_registry=None I0417 00:47:57.192094 136175587092288 composite_checkpoint_handler.py:237] Deferred registration for item: "items". Adding handler `<orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeCheckpointHandler object at 0x7bc50c4a89b0>` 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`. I0417 00:47:57.192275 136175587092288 composite_checkpoint_handler.py:237] Deferred registration for item: "metrics". Adding handler `<orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonCheckpointHandler object at 0x7bc50c4f1640>` 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`. I0417 00:47:57.192325 136175587092288 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 0x7bc50c4a89b0>, ('items', <class 'orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeRestoreArgs'>): <orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeCheckpointHandler object at 0x7bc50c4a89b0>, ('metrics', <class 'orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonSaveArgs'>): <orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonCheckpointHandler object at 0x7bc50c4f1640>, ('metrics', <class 'orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonRestoreArgs'>): <orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonCheckpointHandler object at 0x7bc50c4f1640>}). I0417 00:47:57.192738 136175587092288 abstract_checkpointer.py:35] orbax-checkpoint version: 0.11.34 I0417 00:47:57.192829 136175587092288 async_checkpointer.py:192] [process=0][thread=MainThread] Using barrier_sync_fn: <function get_barrier_sync_fn.<locals>._fn at 0x7bc50c592ca0> timeout: 1200 secs and primary_host=0 for async checkpoint writes I0417 00:47:58.173886 136175587092288 checkpoint_manager.py:1812] Found 0 checkpoint steps in gs://lance-maxtext/nnx_ckpt_xpk_feat_nnx_post_train_fixes_20260417_003418/nnx_xpk_feat_nnx_post_train_fixes_20260417_003418_04_int8/checkpoints I0417 00:47:58.622951 136175587092288 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_20260417_003418/nnx_xpk_feat_nnx_post_train_fixes_20260417_003418_04_int8/checkpoints: <orbax.checkpoint.checkpoint_manager.CheckpointManager object at 0x7bc50c4c2210> I0417 00:47:58.623115 136175587092288 checkpointing.py:302] Checkpoint manager created! I0417 00:47:59.982296 136175587092288 checkpointing.py:578] checkpoint manager exists so trying to load this run's existing checkpoint I0417 00:47:59.982413 136175587092288 checkpointing.py:676] No existing checkpoints found, not restoring checkpoint. fsdp: 32 I0417 00:48:02.346416 136175587092288 maxtext_utils.py:1805] decoder/decoder_norm/scale/value Shape: float32[2048] Physical: (None,) I0417 00:48:02.346533 136175587092288 maxtext_utils.py:1805] decoder/layers/mlp/wi_0/kernel/value Shape: float32[2048,16,7168] Physical: ('fsdp', None, None) I0417 00:48:02.346586 136175587092288 maxtext_utils.py:1805] decoder/layers/mlp/wi_1/kernel/value Shape: float32[2048,16,7168] Physical: ('fsdp', None, None) I0417 00:48:02.346631 136175587092288 maxtext_utils.py:1805] decoder/layers/mlp/wo/kernel/value Shape: float32[7168,16,2048] Physical: (None, None, 'fsdp') I0417 00:48:02.346701 136175587092288 maxtext_utils.py:1805] decoder/layers/post_self_attention_layer_norm/scale/value Shape: float32[2048,16] Physical: (None, None) I0417 00:48:02.346740 136175587092288 maxtext_utils.py:1805] decoder/layers/pre_self_attention_layer_norm/scale/value Shape: float32[2048,16] Physical: (None, None) I0417 00:48:02.346778 136175587092288 maxtext_utils.py:1805] decoder/layers/self_attention/key/kernel/value Shape: float32[2048,16,16,128] Physical: ('fsdp', None, None, None) I0417 00:48:02.346813 136175587092288 maxtext_utils.py:1805] decoder/layers/self_attention/out/kernel/value Shape: float32[16,16,128,2048] Physical: (None, None, None, 'fsdp') I0417 00:48:02.346859 136175587092288 maxtext_utils.py:1805] decoder/layers/self_attention/query/kernel/value Shape: float32[2048,16,16,128] Physical: ('fsdp', None, None, None) I0417 00:48:02.346895 136175587092288 maxtext_utils.py:1805] decoder/layers/self_attention/value/kernel/value Shape: float32[2048,16,16,128] Physical: ('fsdp', None, None, None) I0417 00:48:02.346926 136175587092288 maxtext_utils.py:1805] decoder/logits_dense/kernel/value Shape: float32[2048,32000] Physical: ('fsdp', None) I0417 00:48:02.346958 136175587092288 maxtext_utils.py:1805] token_embedder/embedding/value Shape: float32[32000,2048] Physical: (None, 'fsdp') I0417 00:48:02.758376 136175587092288 nnx_decoders.py:465] nnx_decoders/carry Logical: bfloat16[32,2048,2048]...................................... ('activation_batch', 'activation_norm_length', 'activation_embed'). I0417 00:48:02.758468 136175587092288 nnx_decoders.py:465] nnx_decoders/carry Physical: bfloat16[32,2048,2048]...................................... ('fsdp', None, None). I0417 00:48:02.763710 136175587092288 nnx_decoders.py:465] Unknown Logical: bfloat16[32,2048,2048]...................................... ('activation_batch', 'activation_norm_length', 'activation_embed'). I0417 00:48:02.763765 136175587092288 nnx_decoders.py:465] Unknown Physical: bfloat16[32,2048,2048]...................................... ('fsdp', None, None). I0417 00:48:02.779995 136175587092288 attentions.py:1088] attentions/inputs_q Logical: bfloat16[32,2048,2048]...................................... ('activation_batch', 'activation_attn_length', 'activation_attn_embed'). I0417 00:48:02.780054 136175587092288 attentions.py:1088] attentions/inputs_q Physical: bfloat16[32,2048,2048]...................................... ('fsdp', None, None). I0417 00:48:02.795751 136175587092288 attentions.py:1089] attentions/inputs_kv Logical: bfloat16[32,2048,2048]...................................... ('activation_batch', 'activation_attn_length', 'activation_attn_embed'). I0417 00:48:02.795818 136175587092288 attentions.py:1089] attentions/inputs_kv Physical: bfloat16[32,2048,2048]...................................... ('fsdp', None, None). I0417 00:48:02.874273 136175587092288 attentions.py:1154] attentions/query Logical: bfloat16[32,2048,16,128].................................... ('activation_kv_batch', 'activation_attn_length', 'activation_kv_heads', 'activation_kv_head_dim'). I0417 00:48:02.874362 136175587092288 attentions.py:1154] attentions/query Physical: bfloat16[32,2048,16,128].................................... ('fsdp', None, None, None). I0417 00:48:02.890102 136175587092288 attentions.py:1155] attentions/key Logical: bfloat16[32,2048,16,128].................................... ('activation_kv_batch', 'activation_attn_length', 'activation_kv_heads', 'activation_kv_head_dim'). I0417 00:48:02.890163 136175587092288 attentions.py:1155] attentions/key Physical: bfloat16[32,2048,16,128].................................... ('fsdp', None, None, None). I0417 00:48:02.905685 136175587092288 attentions.py:1156] attentions/value Logical: bfloat16[32,2048,16,128].................................... ('activation_kv_batch', 'activation_attn_length', 'activation_kv_heads', 'activation_kv_head_dim'). I0417 00:48:02.905743 136175587092288 attentions.py:1156] attentions/value Physical: bfloat16[32,2048,16,128].................................... ('fsdp', None, None, None). I0417 00:48:02.936342 136175587092288 attentions.py:1197] attentions/out Logical: bfloat16[32,2048,16,128].................................... ('activation_batch', 'activation_attn_length', 'activation_heads', 'activation_kv'). I0417 00:48:02.936409 136175587092288 attentions.py:1197] attentions/out Physical: bfloat16[32,2048,16,128].................................... ('fsdp', None, None, None). I0417 00:48:03.017210 136175587092288 linears.py:525] linears/x Logical: bfloat16[32,2048,7168]...................................... ('activation_batch', 'activation_length', 'activation_mlp'). I0417 00:48:03.017299 136175587092288 linears.py:525] linears/x Physical: bfloat16[32,2048,7168]...................................... ('fsdp', None, None). I0417 00:48:16.931524 136175587092288 max_utils.py:791] Total memory size: 1.8 GB, Output size: 0.4 GB, Temp size: 1.4 GB, Argument size: 0.4 GB, Host temp size: 0.0 GB. I0417 00:48:16.941805 136175587092288 metric_logger.py:289] number parameters: 1.104 billion I0417 00:48:31.505489 136175587092288 checkpointing.py:794] Waiting for step 0 to finish before checkpoint... I0417 00:48:31.745568 136175587092288 checkpointing.py:798] Waited 0.24006056785583496 seconds for step 0 to finish before starting checkpointing. I0417 00:48:31.748207 136175587092288 checkpoint_manager.py:2009] [process=0][thread=MainThread][wait_until_finished] No Save Finalize thread to wait for. Returning. I0417 00:48:31.750036 136175587092288 checkpoint_manager.py:1512] [process=0] Saving checkpoint at step 0 I0417 00:48:31.751859 136175587092288 event_tracking.py:70] [process=0] [async] Started save checkpoint @ gs://lance-maxtext/nnx_ckpt_xpk_feat_nnx_post_train_fixes_20260417_003418/nnx_xpk_feat_nnx_post_train_fixes_20260417_003418_04_int8/checkpoints/0. I0417 00:48:32.473448 136175587092288 signaling_client.py:364] Using JaxDistributedSignalingClient I0417 00:48:32.500914 136175587092288 jax_array_handlers.py:360] Scheduling D2H of 111 prioritized jax.Array. I0417 00:48:32.500998 136175587092288 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 I0417 00:48:32.756967 136044892645120 atomicity.py:140] Creating tmp directory gs://lance-maxtext/nnx_ckpt_xpk_feat_nnx_post_train_fixes_20260417_003418/nnx_xpk_feat_nnx_post_train_fixes_20260417_003418_04_int8/checkpoints/0 I0417 00:48:32.782607 136175587092288 base_pytree_checkpoint_handler.py:154] [process=0][thread=MainThread] Initiated "orbax.checkpoint._src.serialization.jax_array_handlers.ArrayHandler".serialize. Time taken: 0.284069s I0417 00:48:32.783512 136175587092288 base_pytree_checkpoint_handler.py:130] [process=0] /jax/orbax/write/blocking_gbytes_per_sec: 5.114 GiB/s (total gbytes: 1.5 GiB) (time elapsed: 0.30163145065307617 s) (per-host) I0417 00:48:32.783587 136175587092288 base_pytree_checkpoint_handler.py:768] [process=0][thread=MainThread] Initiated Pytree async_save. Time taken: 0.301728s (batch_requests_ready=0.005084s, total_serialization_initiated=0.295817s, others=0.000827s) I0417 00:48:32.783682 136175587092288 composite_checkpoint_handler.py:715] [process=0][thread=MainThread] Initiated CompositeCheckpointHandler.async_save. Time taken: 0.305719s (all_items=0.000017s, per_item={'items': '0.00001740'}, temp_paths=0.305701) I0417 00:48:32.784477 136175587092288 event_tracking.py:125] [process=0] [async] Finished blocking save in 1.03 seconds. Continuing save @ gs://lance-maxtext/nnx_ckpt_xpk_feat_nnx_post_train_fixes_20260417_003418/nnx_xpk_feat_nnx_post_train_fixes_20260417_003418_04_int8/checkpoints/0. I0417 00:48:32.784769 136044901037824 async_checkpointer.py:76] [process=0][thread=async_save] Background save thread started. Deadline for this save operation is 2026-04-17 01:08:32.784733 I0417 00:48:32.787245 136175587092288 checkpoint_manager.py:1560] [process=0][thread=MainThread][step=0] Starting CheckpointManager Save Finalize thread=save_finalize I0417 00:48:32.787493 136044306102016 async_checkpointer.py:280] [process=0][thread=save_finalize] Waiting for background save thread=async_save. I0417 00:48:32.787617 136175587092288 standard_logger.py:34] {'step': 0, 'event_type': 'save', 'directory': 'gs://lance-maxtext/nnx_ckpt_xpk_feat_nnx_post_train_fixes_20260417_003418/nnx_xpk_feat_nnx_post_train_fixes_20260417_003418_04_int8/checkpoints', 'reached_preemption': False, 'preemption_received_at': None, 'synchronous': False, 'wait_for_prev_start_time': 1776386911.7481868, 'wait_for_prev_duration_secs': 6.246566772460938e-05, 'time_between_consecutive_saves_sec': None, 'checkpointer_blocking_start_time': 1776386911.7500758, 'checkpointer_blocking_duration_secs': 1.0347979068756104, 'get_old_steps_start_time': 1776386912.784898, 'get_old_steps_duration_secs': 2.8848648071289062e-05, 'checkpoint_manager_blocking_start_time': 1776386911.7460449, 'checkpoint_manager_blocking_duration_secs': 1.0415329933166504} I0417 00:48:32.787733 136175587092288 checkpointing.py:409] Started an asynchronous checkpoint save for step 0 I0417 00:48:32.787811 136175587092288 max_utils.py:750] Memstats: After params initialized: I0417 00:48:32.787889 136175587092288 max_utils.py:756] Using (GB) 0.44 / 31.25 (1.408000%) on TPU_0(process=0,(0,0,0,0)) I0417 00:48:32.787935 136175587092288 max_utils.py:756] Using (GB) 0.43 / 31.25 (1.376000%) on TPU_1(process=0,(1,0,0,0)) I0417 00:48:32.787963 136175587092288 max_utils.py:756] Using (GB) 0.43 / 31.25 (1.376000%) on TPU_4(process=0,(0,1,0,0)) I0417 00:48:32.787988 136175587092288 max_utils.py:756] Using (GB) 0.43 / 31.25 (1.376000%) on TPU_5(process=0,(1,1,0,0)) I0417 00:48:33.132833 136175587092288 metric_logger.py:185] completed step: 0, seconds: 14.164, TFLOP/s/device: 0.959, Tokens/s/device: 144.592, total_weights: 65536, loss: 10.872 I0417 00:48:33.136071 136175587092288 metric_logger.py:269] To see full metrics 'tensorboard --logdir=gs://lance-maxtext/nnx_ckpt_xpk_feat_nnx_post_train_fixes_20260417_003418/nnx_xpk_feat_nnx_post_train_fixes_20260417_003418_04_int8/tensorboard/' I0417 00:48:33.588940 136175587092288 metric_logger.py:185] completed step: 1, seconds: 1.622, TFLOP/s/device: 8.376, Tokens/s/device: 1262.502, total_weights: 65536, loss: 10.872 I0417 00:48:33.808565 136175587092288 metric_logger.py:185] completed step: 2, seconds: 0.457, TFLOP/s/device: 29.721, Tokens/s/device: 4479.920, total_weights: 65536, loss: 10.260 I0417 00:48:34.038085 136175587092288 metric_logger.py:185] completed step: 3, seconds: 0.015, TFLOP/s/device: 879.540, Tokens/s/device: 132573.796, total_weights: 65536, loss: 9.728 I0417 00:48:34.564279 136044875859712 atomicity.py:140] Creating tmp directory gs://lance-maxtext/nnx_ckpt_xpk_feat_nnx_post_train_fixes_20260417_003418/nnx_xpk_feat_nnx_post_train_fixes_20260417_003418_04_int8/checkpoints/0/items I0417 00:48:34.715004 136043778406144 checkpoint.py:188] Wrote Metadata={'item_handlers': None, 'metrics': {}, 'performance_metrics': {}, 'init_timestamp_nsecs': 1776386914291582026, 'commit_timestamp_nsecs': None, 'custom_metadata': {}}, json={"item_handlers": null, "metrics": {}, "performance_metrics": {}, "init_timestamp_nsecs": 1776386914291582026, "commit_timestamp_nsecs": null, "custom_metadata": {}} to gs://lance-maxtext/nnx_ckpt_xpk_feat_nnx_post_train_fixes_20260417_003418/nnx_xpk_feat_nnx_post_train_fixes_20260417_003418_04_int8/checkpoints/0/_CHECKPOINT_METADATA I0417 00:48:35.688144 2955 google_auth_provider.cc:181] Running on GCE, using service account 562977990677-compute@developer.gserviceaccount.com I0417 00:48:38.161561 136044850681600 array_metadata_store.py:203] [process=0][thread=array_type_handler] Wrote 111 array_metadata.ArrayMetadata to gs://lance-maxtext/nnx_ckpt_xpk_feat_nnx_post_train_fixes_20260417_003418/nnx_xpk_feat_nnx_post_train_fixes_20260417_003418_04_int8/checkpoints/0/items/array_metadatas/process_0 I0417 00:49:03.615221 136175587092288 metric_logger.py:185] completed step: 4, seconds: 0.220, TFLOP/s/device: 61.638, Tokens/s/device: 9290.805, total_weights: 65536, loss: 9.271 I0417 00:49:03.671869 136175587092288 metric_logger.py:185] completed step: 5, seconds: 0.229, TFLOP/s/device: 59.247, Tokens/s/device: 8930.401, total_weights: 65536, loss: 8.885 I0417 00:49:03.852667 136175587092288 metric_logger.py:185] completed step: 6, seconds: 29.576, TFLOP/s/device: 0.459, Tokens/s/device: 69.245, total_weights: 65536, loss: 8.586 I0417 00:49:04.082383 136175587092288 metric_logger.py:185] completed step: 7, seconds: 0.055, TFLOP/s/device: 246.940, Tokens/s/device: 37221.475, total_weights: 65536, loss: 8.378 I0417 00:49:04.312475 136175587092288 metric_logger.py:185] completed step: 8, seconds: 0.182, TFLOP/s/device: 74.786, Tokens/s/device: 11272.505, total_weights: 65536, loss: 8.249 I0417 00:49:04.543550 136175587092288 checkpointing.py:794] Waiting for step 9 to finish before checkpoint... I0417 00:49:04.544640 136175587092288 checkpointing.py:798] Waited 0.0011012554168701172 seconds for step 9 to finish before starting checkpointing. I0417 00:49:04.546756 136175587092288 checkpoint_manager.py:2020] [process=0][thread=MainThread][step=0][wait_until_finished] Waiting for Save Finalize thread (save_finalize) to complete. I0417 00:49:11.555282 136044314494720 base_pytree_checkpoint_handler.py:1282] [process=0][thread=write_metadata_after_commits] Commit + Array metadata written. Time taken: 35.875149s (commit=32.007585s, array_metadata_write=3.867565s) I0417 00:49:11.556565 136044901037824 base_pytree_checkpoint_handler.py:130] [process=0] /jax/orbax/write/gbytes_per_sec: 40.425 MiB/s (total gbytes: 1.5 GiB) (time elapsed: 39.074657917022705 s) (per-host) I0417 00:49:11.556688 136044901037824 async_checkpointer.py:90] [process=0][thread=async_save] 3 Handler Commit operations completed. Time taken: 38.771284s. I0417 00:49:12.275983 136044901037824 checkpoint.py:228] Read Metadata={'item_handlers': None, 'metrics': {}, 'performance_metrics': {}, 'init_timestamp_nsecs': 1776386914291582026, 'commit_timestamp_nsecs': None, 'custom_metadata': {}} from gs://lance-maxtext/nnx_ckpt_xpk_feat_nnx_post_train_fixes_20260417_003418/nnx_xpk_feat_nnx_post_train_fixes_20260417_003418_04_int8/checkpoints/0/_CHECKPOINT_METADATA I0417 00:49:13.866728 136043778406144 checkpoint.py:247] Updated Metadata={'item_handlers': {'items': 'orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeCheckpointHandler'}, 'metrics': {}, 'performance_metrics': {}, 'init_timestamp_nsecs': 1776386914291582026, 'commit_timestamp_nsecs': None, 'custom_metadata': {}} to gs://lance-maxtext/nnx_ckpt_xpk_feat_nnx_post_train_fixes_20260417_003418/nnx_xpk_feat_nnx_post_train_fixes_20260417_003418_04_int8/checkpoints/0/_CHECKPOINT_METADATA I0417 00:49:14.352434 136044901037824 array_metadata_store.py:411] [process=0][thread=async_save] Validated ArrayMetadata from all 8 hosts. Time taken: 0.000606s. I0417 00:49:14.750081 136044901037824 ocdbt_utils.py:49] Param validation support for Zarr3 will be added later (b/362328389). I0417 00:49:15.540135 136044901037824 base_pytree_checkpoint_handler.py:1406] [process=0][thread=async_save] Pytree save finalize (merge_ocdbt + ArrayMetadata validation) completed. Time taken: 2.687920s. use_zarr3=True, enable_post_merge_validation=True, directory=gs://lance-maxtext/nnx_ckpt_xpk_feat_nnx_post_train_fixes_20260417_003418/nnx_xpk_feat_nnx_post_train_fixes_20260417_003418_04_int8/checkpoints/0/items I0417 00:49:15.541998 136044901037824 atomicity.py:666] Finalizing gs://lance-maxtext/nnx_ckpt_xpk_feat_nnx_post_train_fixes_20260417_003418/nnx_xpk_feat_nnx_post_train_fixes_20260417_003418_04_int8/checkpoints/0/items I0417 00:49:16.920507 136044901037824 atomicity.py:666] Finalizing gs://lance-maxtext/nnx_ckpt_xpk_feat_nnx_post_train_fixes_20260417_003418/nnx_xpk_feat_nnx_post_train_fixes_20260417_003418_04_int8/checkpoints/0 I0417 00:49:19.359144 136044901037824 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_20260417_003418/nnx_xpk_feat_nnx_post_train_fixes_20260417_003418_04_int8/checkpoints/0`. I0417 00:49:19.360056 136044901037824 event_tracking.py:138] [process=0] [async] Finished save (blocking + background) in 47.61 seconds @ gs://lance-maxtext/nnx_ckpt_xpk_feat_nnx_post_train_fixes_20260417_003418/nnx_xpk_feat_nnx_post_train_fixes_20260417_003418_04_int8/checkpoints/0 I0417 00:49:19.361652 136044901037824 async_checkpointer.py:160] [process=0][thread=async_save] Background save thread done. Time taken: 46.576246s. I0417 00:49:19.361882 136044306102016 async_checkpointer.py:288] [process=0][thread=save_finalize] Done with waiting for background save thread=async_save. I0417 00:49:19.362018 136044306102016 async_checkpointer.py:298] [process=0][thread=save_finalize] No errors found in background save thread=async_save. I0417 00:49:19.362083 136044306102016 checkpoint_manager.py:2137] [process=0][thread=save_finalize][step=0] CheckpointManager Save Finalize is syncing with other hosts... I0417 00:49:19.364871 136044306102016 checkpoint_manager.py:2146] [process=0][thread=save_finalize][step=0] CheckpointManager Save Finalize is done on all hosts. I0417 00:49:19.365061 136175587092288 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. W0417 00:49:19.365201 136175587092288 checkpoint_manager.py:1452] Waiting for previous save to complete took 14.818441 seconds. If this number is high, consider checkpointing less frequently. I0417 00:49:19.367301 136175587092288 checkpoint_manager.py:1512] [process=0] Saving checkpoint at step 9 I0417 00:49:19.369433 136175587092288 event_tracking.py:70] [process=0] [async] Started save checkpoint @ gs://lance-maxtext/nnx_ckpt_xpk_feat_nnx_post_train_fixes_20260417_003418/nnx_xpk_feat_nnx_post_train_fixes_20260417_003418_04_int8/checkpoints/9. I0417 00:49:20.087505 136175587092288 jax_array_handlers.py:360] Scheduling D2H of 111 prioritized jax.Array. I0417 00:49:20.087600 136175587092288 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 I0417 00:49:20.116568 136175587092288 base_pytree_checkpoint_handler.py:154] [process=0][thread=MainThread] Initiated "orbax.checkpoint._src.serialization.jax_array_handlers.ArrayHandler".serialize. Time taken: 0.031383s I0417 00:49:20.118704 136175587092288 base_pytree_checkpoint_handler.py:130] [process=0] /jax/orbax/write/blocking_gbytes_per_sec: 32.765 GiB/s (total gbytes: 1.5 GiB) (time elapsed: 0.04708051681518555 s) (per-host) I0417 00:49:20.118775 136175587092288 base_pytree_checkpoint_handler.py:768] [process=0][thread=MainThread] Initiated Pytree async_save. Time taken: 0.047167s (batch_requests_ready=0.004143s, total_serialization_initiated=0.040933s, others=0.002092s) I0417 00:49:20.118864 136175587092288 composite_checkpoint_handler.py:715] [process=0][thread=MainThread] Initiated CompositeCheckpointHandler.async_save. Time taken: 0.051306s (all_items=0.000017s, per_item={'items': '0.00001669'}, temp_paths=0.051289) I0417 00:49:20.119627 136175587092288 event_tracking.py:125] [process=0] [async] Finished blocking save in 0.75 seconds. Continuing save @ gs://lance-maxtext/nnx_ckpt_xpk_feat_nnx_post_train_fixes_20260417_003418/nnx_xpk_feat_nnx_post_train_fixes_20260417_003418_04_int8/checkpoints/9. I0417 00:49:20.119939 136044306102016 async_checkpointer.py:76] [process=0][thread=async_save] Background save thread started. Deadline for this save operation is 2026-04-17 01:09:20.119904 I0417 00:49:20.123243 136175587092288 checkpoint_manager.py:1560] [process=0][thread=MainThread][step=9] Starting CheckpointManager Save Finalize thread=save_finalize I0417 00:49:20.123489 136044314494720 async_checkpointer.py:280] [process=0][thread=save_finalize] Waiting for background save thread=async_save. I0417 00:49:20.123621 136175587092288 standard_logger.py:34] {'step': 9, 'event_type': 'save', 'directory': 'gs://lance-maxtext/nnx_ckpt_xpk_feat_nnx_post_train_fixes_20260417_003418/nnx_xpk_feat_nnx_post_train_fixes_20260417_003418_04_int8/checkpoints', 'reached_preemption': False, 'preemption_received_at': None, 'synchronous': False, 'wait_for_prev_start_time': 1776386944.5467236, 'wait_for_prev_duration_secs': 14.818440914154053, 'time_between_consecutive_saves_sec': None, 'checkpointer_blocking_start_time': 1776386959.3673491, 'checkpointer_blocking_duration_secs': 0.7527554035186768, 'get_old_steps_start_time': 1776386960.120131, 'get_old_steps_duration_secs': 3.933906555175781e-05, 'checkpoint_manager_blocking_start_time': 1776386944.5450852, 'checkpoint_manager_blocking_duration_secs': 15.578500032424927} I0417 00:49:20.123738 136175587092288 checkpointing.py:409] Started an asynchronous checkpoint save for step 9 I0417 00:49:20.123787 136175587092288 checkpoint_manager.py:2020] [process=0][thread=MainThread][step=9][wait_until_finished] Waiting for Save Finalize thread (save_finalize) to complete. I0417 00:49:20.317891 136044901037824 atomicity.py:140] Creating tmp directory gs://lance-maxtext/nnx_ckpt_xpk_feat_nnx_post_train_fixes_20260417_003418/nnx_xpk_feat_nnx_post_train_fixes_20260417_003418_04_int8/checkpoints/9 I0417 00:49:22.068113 136043818903296 atomicity.py:140] Creating tmp directory gs://lance-maxtext/nnx_ckpt_xpk_feat_nnx_post_train_fixes_20260417_003418/nnx_xpk_feat_nnx_post_train_fixes_20260417_003418_04_int8/checkpoints/9/items I0417 00:49:27.596284 136044859074304 array_metadata_store.py:203] [process=0][thread=array_type_handler] Wrote 111 array_metadata.ArrayMetadata to gs://lance-maxtext/nnx_ckpt_xpk_feat_nnx_post_train_fixes_20260417_003418/nnx_xpk_feat_nnx_post_train_fixes_20260417_003418_04_int8/checkpoints/9/items/array_metadatas/process_0 I0417 00:50:04.778697 136044850681600 base_pytree_checkpoint_handler.py:1282] [process=0][thread=write_metadata_after_commits] Commit + Array metadata written. Time taken: 41.556249s (commit=39.825851s, array_metadata_write=1.730398s) I0417 00:50:04.779982 136044306102016 base_pytree_checkpoint_handler.py:130] [process=0] /jax/orbax/write/gbytes_per_sec: 35.331 MiB/s (total gbytes: 1.5 GiB) (time elapsed: 44.70832586288452 s) (per-host) I0417 00:50:04.780118 136044306102016 async_checkpointer.py:90] [process=0][thread=async_save] 3 Handler Commit operations completed. Time taken: 44.660052s. I0417 00:50:07.043551 136044306102016 array_metadata_store.py:411] [process=0][thread=async_save] Validated ArrayMetadata from all 8 hosts. Time taken: 0.000480s. I0417 00:50:07.145704 136044306102016 ocdbt_utils.py:49] Param validation support for Zarr3 will be added later (b/362328389). I0417 00:50:08.114541 136044306102016 base_pytree_checkpoint_handler.py:1406] [process=0][thread=async_save] Pytree save finalize (merge_ocdbt + ArrayMetadata validation) completed. Time taken: 1.962893s. use_zarr3=True, enable_post_merge_validation=True, directory=gs://lance-maxtext/nnx_ckpt_xpk_feat_nnx_post_train_fixes_20260417_003418/nnx_xpk_feat_nnx_post_train_fixes_20260417_003418_04_int8/checkpoints/9/items I0417 00:50:08.116251 136044306102016 atomicity.py:666] Finalizing gs://lance-maxtext/nnx_ckpt_xpk_feat_nnx_post_train_fixes_20260417_003418/nnx_xpk_feat_nnx_post_train_fixes_20260417_003418_04_int8/checkpoints/9/items I0417 00:50:08.646528 136044306102016 atomicity.py:666] Finalizing gs://lance-maxtext/nnx_ckpt_xpk_feat_nnx_post_train_fixes_20260417_003418/nnx_xpk_feat_nnx_post_train_fixes_20260417_003418_04_int8/checkpoints/9 I0417 00:50:10.081202 136044306102016 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_20260417_003418/nnx_xpk_feat_nnx_post_train_fixes_20260417_003418_04_int8/checkpoints/9`. I0417 00:50:10.082060 136044306102016 event_tracking.py:138] [process=0] [async] Finished save (blocking + background) in 50.71 seconds @ gs://lance-maxtext/nnx_ckpt_xpk_feat_nnx_post_train_fixes_20260417_003418/nnx_xpk_feat_nnx_post_train_fixes_20260417_003418_04_int8/checkpoints/9 I0417 00:50:10.083299 136044306102016 async_checkpointer.py:160] [process=0][thread=async_save] Background save thread done. Time taken: 49.963230s. I0417 00:50:10.083495 136044314494720 async_checkpointer.py:288] [process=0][thread=save_finalize] Done with waiting for background save thread=async_save. I0417 00:50:10.083618 136044314494720 async_checkpointer.py:298] [process=0][thread=save_finalize] No errors found in background save thread=async_save. I0417 00:50:10.083682 136044314494720 checkpoint_manager.py:2137] [process=0][thread=save_finalize][step=9] CheckpointManager Save Finalize is syncing with other hosts... I0417 00:50:10.085719 136044314494720 checkpoint_manager.py:2146] [process=0][thread=save_finalize][step=9] CheckpointManager Save Finalize is done on all hosts. I0417 00:50:10.085876 136175587092288 checkpoint_manager.py:2032] [process=0][thread=MainThread][step=9][wait_until_finished] Done waiting for Save Finalize thread (save_finalize) running at step=9. I0417 00:50:10.086009 136175587092288 checkpoint_manager.py:2009] [process=0][thread=MainThread][wait_until_finished] No Save Finalize thread to wait for. Returning. I0417 00:50:10.086610 136175587092288 metric_logger.py:185] completed step: 9, seconds: 0.229, TFLOP/s/device: 59.280, Tokens/s/device: 8935.389, total_weights: 65536, loss: 8.174 Per train step: Total TFLOPs: 13.59 split as 93.93% learnable weight flops and 6.07% attention flops XPK End: Fri Apr 17 00:50:18 UTC 2026 EXIT_CODE=0