XPK Start: Wed Apr 22 14:43:07 UTC 2026 PyTorch was not found. Models won't be available and only tokenizers, configuration and file/data utilities can be used. 2026-04-22 14:43:31.312685: E external/local_xla/xla/stream_executor/cuda/cuda_platform.cc:51] failed call to cuInit: INTERNAL: CUDA error: Failed call to cuInit: UNKNOWN ERROR (303) I0422 14:43:31.844339 139648287532864 max_utils.py:273] Attempting to initialize the jax distributed system... I0422 14:43:40.886927 139648287532864 distributed.py:149] Starting JAX distributed service on [::]:8482 I0422 14:43:40.889194 139648287532864 distributed.py:172] Connecting to JAX distributed service on mt-09-pdb-lt-1-1xmej-slice-job-0-0.mt-09-pdb-lt-1-1xmej:8482 I0422 14:43:42.408779 139648287532864 max_utils.py:284] Jax distributed system initialized! I0422 14:43:48.539824 139648287532864 max_utils.py:800] System Information: Jax Version: 0.9.2 I0422 14:43:48.539921 139648287532864 max_utils.py:801] System Information: Jaxlib Version: 0.9.2 I0422 14:43:48.539960 139648287532864 max_utils.py:802] System Information: Jax Backend: PJRT C API TFRT TPU v6 lite Built on Mar 4 2026 11:32:08 (1772652728) cl/878335365 I0422 14:43:48.539992 139648287532864 train_utils.py:378] WARNING: Sequence packing is essentially ignored for synthetic data. Please use a real dataset to use sequence packing. I0422 14:43:49.243661 139648287532864 maxtext_utils.py:1718] Num_devices: 32, shape (1, 1, 1, 8, 1, 1, 1, 1, 4, 1, 1, 1, 1) I0422 14:43:49.358034 139648287532864 checkpointing.py:688] Setting up checkpoint logger... I0422 14:43:49.358160 139648287532864 checkpointing.py:234] Creating checkpoint manager with ocdbt=True and zarr3=True I0422 14:43:49.358206 139648287532864 pytree_checkpoint_handler.py:592] save_device_host_concurrent_bytes=None I0422 14:43:49.358402 139648287532864 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 0x7f01cdbde6c0>, enable_pinned_host_transfer=False, save_concurrent_bytes: 96000000000 (89.4 GiB), restore_concurrent_bytes: 96000000000 (89.4 GiB) I0422 14:43:52.290083 139648287532864 checkpointing.py:266] Enabling policy for fixed interval checkpointing. I0422 14:43:52.290338 139648287532864 checkpoint_manager.py:708] [process=6][thread=MainThread] CheckpointManager init: checkpointers=None, item_names=('items',), item_handlers={'items': <orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeCheckpointHandler object at 0x7eedb8230b30>}, handler_registry=None I0422 14:43:52.290583 139648287532864 composite_checkpoint_handler.py:237] Deferred registration for item: "items". Adding handler `<orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeCheckpointHandler object at 0x7eedb8230b30>` for item "items" and save args `<class 'orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeSaveArgs'>` and restore args `<class 'orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeRestoreArgs'>` to `_handler_registry`. I0422 14:43:52.290633 139648287532864 composite_checkpoint_handler.py:237] Deferred registration for item: "metrics". Adding handler `<orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonCheckpointHandler object at 0x7eecb4510a70>` for item "metrics" and save args `<class 'orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonSaveArgs'>` and restore args `<class 'orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonRestoreArgs'>` to `_handler_registry`. I0422 14:43:52.290671 139648287532864 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 0x7eedb8230b30>, ('items', <class 'orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeRestoreArgs'>): <orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeCheckpointHandler object at 0x7eedb8230b30>, ('metrics', <class 'orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonSaveArgs'>): <orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonCheckpointHandler object at 0x7eecb4510a70>, ('metrics', <class 'orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonRestoreArgs'>): <orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonCheckpointHandler object at 0x7eecb4510a70>}). I0422 14:43:52.290983 139648287532864 abstract_checkpointer.py:35] orbax-checkpoint version: 0.11.34 I0422 14:43:52.291053 139648287532864 async_checkpointer.py:192] [process=6][thread=MainThread] Using barrier_sync_fn: <function get_barrier_sync_fn.<locals>._fn at 0x7eed884399e0> timeout: 1200 secs and primary_host=0 for async checkpoint writes I0422 14:43:53.149742 139648287532864 checkpoint_manager.py:1812] Found 0 checkpoint steps in gs://lance-maxtext/nnx_ckpt_xpk_feat_nnx_post_train_fixes_20260422_123906/nnx_xpk_feat_nnx_post_train_fixes_20260422_123906_09_pdb_lt_1/checkpoints I0422 14:43:53.452828 139648287532864 checkpoint_manager.py:929] [process=6][thread=MainThread] CheckpointManager created, primary_host=0, CheckpointManagerOptions=CheckpointManagerOptions(save_interval_steps=1, max_to_keep=None, keep_time_interval=None, keep_period=None, should_keep_fn=None, best_fn=None, best_mode='max', keep_checkpoints_without_metrics=True, step_prefix=None, step_format_fixed_length=None, step_name_format=None, create=True, cleanup_tmp_directories=False, save_on_steps=frozenset(), single_host_load_and_broadcast=False, todelete_subdir=None, todelete_full_path=None, enable_background_delete=False, read_only=False, enable_async_checkpointing=True, async_options=None, multiprocessing_options=MultiprocessingOptions(primary_host=0, active_processes=None, barrier_sync_key_prefix=None), should_save_fn=None, file_options=FileOptions(path_permission_mode=None), save_root_metadata=True, temporary_path_class=None, save_decision_policy=FixedIntervalPolicy(interval=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_20260422_123906/nnx_xpk_feat_nnx_post_train_fixes_20260422_123906_09_pdb_lt_1/checkpoints: <orbax.checkpoint.checkpoint_manager.CheckpointManager object at 0x7eecb46e4770> I0422 14:43:53.453002 139648287532864 checkpointing.py:302] Checkpoint manager created! I0422 14:43:54.422537 139648287532864 checkpointing.py:578] checkpoint manager exists so trying to load this run's existing checkpoint I0422 14:43:54.422658 139648287532864 checkpointing.py:676] No existing checkpoints found, not restoring checkpoint. fsdp: 8 tensor: 4 I0422 14:43:55.830393 139648287532864 maxtext_utils.py:1836] decoder/decoder_norm/scale/value Shape: float32[2048] Physical: ('tensor',) I0422 14:43:55.830508 139648287532864 maxtext_utils.py:1836] decoder/layers/mlp/wi_0/kernel/value Shape: float32[2048,16,7168] Physical: ('fsdp', None, 'tensor') I0422 14:43:55.830565 139648287532864 maxtext_utils.py:1836] decoder/layers/mlp/wi_1/kernel/value Shape: float32[2048,16,7168] Physical: ('fsdp', None, 'tensor') I0422 14:43:55.830609 139648287532864 maxtext_utils.py:1836] decoder/layers/mlp/wo/kernel/value Shape: float32[7168,16,2048] Physical: ('tensor', None, 'fsdp') I0422 14:43:55.830644 139648287532864 maxtext_utils.py:1836] decoder/layers/post_self_attention_layer_norm/scale/value Shape: float32[2048,16] Physical: ('tensor', None) I0422 14:43:55.830678 139648287532864 maxtext_utils.py:1836] decoder/layers/pre_self_attention_layer_norm/scale/value Shape: float32[2048,16] Physical: ('tensor', None) I0422 14:43:55.830713 139648287532864 maxtext_utils.py:1836] decoder/layers/self_attention/key/kernel/value Shape: float32[2048,16,16,128] Physical: ('fsdp', None, 'tensor', None) I0422 14:43:55.830746 139648287532864 maxtext_utils.py:1836] decoder/layers/self_attention/out/kernel/value Shape: float32[16,16,128,2048] Physical: ('tensor', None, None, 'fsdp') I0422 14:43:55.830778 139648287532864 maxtext_utils.py:1836] decoder/layers/self_attention/query/kernel/value Shape: float32[2048,16,16,128] Physical: ('fsdp', None, 'tensor', None) I0422 14:43:55.830810 139648287532864 maxtext_utils.py:1836] decoder/layers/self_attention/value/kernel/value Shape: float32[2048,16,16,128] Physical: ('fsdp', None, 'tensor', None) I0422 14:43:55.830840 139648287532864 maxtext_utils.py:1836] decoder/logits_dense/kernel/value Shape: float32[2048,32000] Physical: ('fsdp', 'tensor') I0422 14:43:55.830871 139648287532864 maxtext_utils.py:1836] token_embedder/embedding/value Shape: float32[32000,2048] Physical: ('tensor', 'fsdp') I0422 14:43:56.235253 139648287532864 nnx_decoders.py:465] nnx_decoders/carry Logical: bfloat16[8,2048,2048]....................................... ('activation_batch', 'activation_norm_length', 'activation_embed'). I0422 14:43:56.235346 139648287532864 nnx_decoders.py:465] nnx_decoders/carry Physical: bfloat16[8,2048,2048]....................................... ('fsdp', None, 'tensor'). I0422 14:43:56.240653 139648287532864 nnx_decoders.py:465] Unknown Logical: bfloat16[8,2048,2048]....................................... ('activation_batch', 'activation_norm_length', 'activation_embed'). I0422 14:43:56.240709 139648287532864 nnx_decoders.py:465] Unknown Physical: bfloat16[8,2048,2048]....................................... ('fsdp', None, 'tensor'). I0422 14:43:56.257118 139648287532864 attentions.py:1088] attentions/inputs_q Logical: bfloat16[8,2048,2048]....................................... ('activation_batch', 'activation_attn_length', 'activation_attn_embed'). I0422 14:43:56.257177 139648287532864 attentions.py:1088] attentions/inputs_q Physical: bfloat16[8,2048,2048]....................................... ('fsdp', None, 'tensor'). I0422 14:43:56.272850 139648287532864 attentions.py:1089] attentions/inputs_kv Logical: bfloat16[8,2048,2048]....................................... ('activation_batch', 'activation_attn_length', 'activation_attn_embed'). I0422 14:43:56.272916 139648287532864 attentions.py:1089] attentions/inputs_kv Physical: bfloat16[8,2048,2048]....................................... ('fsdp', None, 'tensor'). I0422 14:43:56.296971 139648287532864 attentions.py:1154] attentions/query Logical: bfloat16[8,2048,16,128]..................................... ('activation_kv_batch', 'activation_attn_length', 'activation_kv_heads', 'activation_kv_head_dim'). I0422 14:43:56.297040 139648287532864 attentions.py:1154] attentions/query Physical: bfloat16[8,2048,16,128]..................................... ('fsdp', None, 'tensor', None). I0422 14:43:56.312908 139648287532864 attentions.py:1155] attentions/key Logical: bfloat16[8,2048,16,128]..................................... ('activation_kv_batch', 'activation_attn_length', 'activation_kv_heads', 'activation_kv_head_dim'). I0422 14:43:56.312970 139648287532864 attentions.py:1155] attentions/key Physical: bfloat16[8,2048,16,128]..................................... ('fsdp', None, 'tensor', None). I0422 14:43:56.328632 139648287532864 attentions.py:1156] attentions/value Logical: bfloat16[8,2048,16,128]..................................... ('activation_kv_batch', 'activation_attn_length', 'activation_kv_heads', 'activation_kv_head_dim'). I0422 14:43:56.328693 139648287532864 attentions.py:1156] attentions/value Physical: bfloat16[8,2048,16,128]..................................... ('fsdp', None, 'tensor', None). I0422 14:43:56.358870 139648287532864 attentions.py:1197] attentions/out Logical: bfloat16[8,2048,16,128]..................................... ('activation_batch', 'activation_attn_length', 'activation_heads', 'activation_kv'). I0422 14:43:56.358943 139648287532864 attentions.py:1197] attentions/out Physical: bfloat16[8,2048,16,128]..................................... ('fsdp', None, 'tensor', None). I0422 14:43:56.378815 139648287532864 linears.py:525] linears/x Logical: bfloat16[8,2048,7168]....................................... ('activation_batch', 'activation_length', 'activation_mlp'). I0422 14:43:56.378880 139648287532864 linears.py:525] linears/x Physical: bfloat16[8,2048,7168]....................................... ('fsdp', None, 'tensor'). I0422 14:44:11.684468 139648287532864 max_utils.py:791] Total memory size: 0.9 GB, Output size: 0.4 GB, Temp size: 0.5 GB, Argument size: 0.4 GB, Host temp size: 0.0 GB. I0422 14:44:11.686917 139648287532864 metric_logger.py:301] number parameters: 1.104 billion I0422 14:44:12.640878 139648287532864 checkpointing.py:794] Waiting for step 0 to finish before checkpoint... I0422 14:44:27.330285 139648287532864 checkpointing.py:798] Waited 14.689387559890747 seconds for step 0 to finish before starting checkpointing. I0422 14:44:27.332506 139648287532864 checkpoint_manager.py:2009] [process=6][thread=MainThread][wait_until_finished] No Save Finalize thread to wait for. Returning. I0422 14:44:27.334479 139648287532864 checkpoint_manager.py:1512] [process=6] Saving checkpoint at step 0 I0422 14:44:27.335993 139648287532864 event_tracking.py:70] [process=6] [async] Started save checkpoint @ gs://lance-maxtext/nnx_ckpt_xpk_feat_nnx_post_train_fixes_20260422_123906/nnx_xpk_feat_nnx_post_train_fixes_20260422_123906_09_pdb_lt_1/checkpoints/0. I0422 14:44:27.653887 139648287532864 signaling_client.py:364] Using JaxDistributedSignalingClient I0422 14:44:27.654928 139648287532864 jax_array_handlers.py:360] Scheduling D2H of 69 prioritized jax.Array. I0422 14:44:27.654984 139648287532864 replica_slices.py:424] Transferring arrays to host memory with options: use_replica_parallel=True, min_slice_bytes_for_replica_parallel=None, max_replicas_for_replica_parallel=None, enable_pinned_host_transfer=False I0422 14:44:27.925565 139648287532864 base_pytree_checkpoint_handler.py:154] [process=6][thread=MainThread] Initiated "orbax.checkpoint._src.serialization.jax_array_handlers.ArrayHandler".serialize. Time taken: 0.272475s I0422 14:44:27.925735 139648287532864 base_pytree_checkpoint_handler.py:130] [process=6] /jax/orbax/write/blocking_gbytes_per_sec: 5.480 GiB/s (total gbytes: 1.5 GiB) (time elapsed: 0.2814815044403076 s) (per-host) I0422 14:44:27.925787 139648287532864 base_pytree_checkpoint_handler.py:768] [process=6][thread=MainThread] Initiated Pytree async_save. Time taken: 0.281543s (batch_requests_ready=0.002942s, total_serialization_initiated=0.278531s, others=0.000070s) I0422 14:44:27.925885 139648287532864 composite_checkpoint_handler.py:715] [process=6][thread=MainThread] Initiated CompositeCheckpointHandler.async_save. Time taken: 0.285672s (all_items=0.000018s, per_item={'items': '0.00001788'}, temp_paths=0.285655) I0422 14:44:27.926696 139648287532864 event_tracking.py:125] [process=6] [async] Finished blocking save in 0.59 seconds. Continuing save @ gs://lance-maxtext/nnx_ckpt_xpk_feat_nnx_post_train_fixes_20260422_123906/nnx_xpk_feat_nnx_post_train_fixes_20260422_123906_09_pdb_lt_1/checkpoints/0. I0422 14:44:27.927034 139520883906304 async_checkpointer.py:76] [process=6][thread=async_save] Background save thread started. Deadline for this save operation is 2026-04-22 15:04:27.926997 I0422 14:44:27.944038 139648287532864 checkpoint_manager.py:1560] [process=6][thread=MainThread][step=0] Starting CheckpointManager Save Finalize thread=save_finalize I0422 14:44:27.944380 139520393602816 async_checkpointer.py:280] [process=6][thread=save_finalize] Waiting for background save thread=async_save. I0422 14:44:27.944537 139648287532864 standard_logger.py:34] {'step': 0, 'event_type': 'save', 'directory': 'gs://lance-maxtext/nnx_ckpt_xpk_feat_nnx_post_train_fixes_20260422_123906/nnx_xpk_feat_nnx_post_train_fixes_20260422_123906_09_pdb_lt_1/checkpoints', 'reached_preemption': False, 'preemption_received_at': None, 'synchronous': False, 'wait_for_prev_start_time': 1776869067.3324876, 'wait_for_prev_duration_secs': 6.246566772460938e-05, 'time_between_consecutive_saves_sec': None, 'checkpointer_blocking_start_time': 1776869067.3345182, 'checkpointer_blocking_duration_secs': 0.5926282405853271, 'get_old_steps_start_time': 1776869067.9271715, 'get_old_steps_duration_secs': 2.8133392333984375e-05, 'checkpoint_manager_blocking_start_time': 1776869067.3306773, 'checkpoint_manager_blocking_duration_secs': 0.6138217449188232} I0422 14:44:27.944641 139648287532864 checkpointing.py:409] Started an asynchronous checkpoint save for step 0 I0422 14:44:27.944705 139648287532864 max_utils.py:750] Memstats: After params initialized: I0422 14:44:27.944755 139648287532864 max_utils.py:756] Using (GB) 0.43 / 31.25 (1.376000%) on TPU_24(process=6,(0,6,0,0)) I0422 14:44:27.944787 139648287532864 max_utils.py:756] Using (GB) 0.43 / 31.25 (1.376000%) on TPU_25(process=6,(1,6,0,0)) I0422 14:44:27.944814 139648287532864 max_utils.py:756] Using (GB) 0.43 / 31.25 (1.376000%) on TPU_28(process=6,(0,7,0,0)) I0422 14:44:27.944839 139648287532864 max_utils.py:756] Using (GB) 0.43 / 31.25 (1.376000%) on TPU_29(process=6,(1,7,0,0)) I0422 14:44:28.265394 139648287532864 metric_logger.py:196] completed step: 0, seconds: 0.953, TFLOP/s/device: 3.565, Tokens/s/device: 537.325, total_weights: 16384, loss: 10.864, lm_loss: 10.864, perplexity: 52258.910 I0422 14:44:28.338068 139648287532864 metric_logger.py:196] completed step: 1, seconds: 15.623, TFLOP/s/device: 0.217, Tokens/s/device: 32.773, total_weights: 16384, loss: 10.864, lm_loss: 10.864, perplexity: 52258.910 I0422 14:44:28.772041 139648287532864 metric_logger.py:196] completed step: 2, seconds: 0.015, TFLOP/s/device: 220.599, Tokens/s/device: 33251.072, total_weights: 16384, loss: 9.760, lm_loss: 9.760, perplexity: 17319.064 I0422 14:44:28.830446 139648287532864 metric_logger.py:196] completed step: 3, seconds: 0.424, TFLOP/s/device: 8.014, Tokens/s/device: 1207.915, total_weights: 16384, loss: 8.820, lm_loss: 8.820, perplexity: 6771.344 I0422 14:44:28.947965 139648287532864 metric_logger.py:196] completed step: 4, seconds: 0.076, TFLOP/s/device: 44.975, Tokens/s/device: 6779.122, total_weights: 16384, loss: 7.947, lm_loss: 7.947, perplexity: 2825.893 I0422 14:44:28.955591 139648287532864 metric_logger.py:196] completed step: 5, seconds: 0.058, TFLOP/s/device: 58.745, Tokens/s/device: 8854.761, total_weights: 16384, loss: 7.209, lm_loss: 7.209, perplexity: 1352.040 I0422 14:44:31.238479 2538 google_auth_provider.cc:181] Running on GCE, using service account 562977990677-compute@developer.gserviceaccount.com I0422 14:44:33.695008 139523598776064 array_metadata_store.py:203] [process=6][thread=array_type_handler] Wrote 69 array_metadata.ArrayMetadata to gs://lance-maxtext/nnx_ckpt_xpk_feat_nnx_post_train_fixes_20260422_123906/nnx_xpk_feat_nnx_post_train_fixes_20260422_123906_09_pdb_lt_1/checkpoints/0/items/array_metadatas/process_6 I0422 14:44:44.309026 139648287532864 metric_logger.py:196] completed step: 6, seconds: 0.118, TFLOP/s/device: 28.727, Tokens/s/device: 4330.029, total_weights: 16384, loss: 6.651, lm_loss: 6.651, perplexity: 773.347 I0422 14:44:44.367501 139648287532864 metric_logger.py:196] completed step: 7, seconds: 15.296, TFLOP/s/device: 0.222, Tokens/s/device: 33.473, total_weights: 16384, loss: 6.267, lm_loss: 6.267, perplexity: 526.951 I0422 14:44:44.426347 139648287532864 metric_logger.py:196] completed step: 8, seconds: 0.064, TFLOP/s/device: 52.703, Tokens/s/device: 7944.019, total_weights: 16384, loss: 6.030, lm_loss: 6.030, perplexity: 415.600 I0422 14:44:44.485643 139648287532864 checkpointing.py:794] Waiting for step 9 to finish before checkpoint... I0422 14:44:44.486262 139648287532864 checkpointing.py:798] Waited 0.0006275177001953125 seconds for step 9 to finish before starting checkpointing. I0422 14:44:44.488327 139648287532864 checkpoint_manager.py:2020] [process=6][thread=MainThread][step=0][wait_until_finished] Waiting for Save Finalize thread (save_finalize) to complete. I0422 14:45:04.067326 139520883906304 base_pytree_checkpoint_handler.py:130] [process=6] /jax/orbax/write/gbytes_per_sec: 43.368 MiB/s (total gbytes: 1.5 GiB) (time elapsed: 36.42303252220154 s) (per-host) I0422 14:45:04.067448 139520883906304 async_checkpointer.py:90] [process=6][thread=async_save] 3 Handler Commit operations completed. Time taken: 36.139780s. I0422 14:45:12.496601 139520883906304 async_checkpointer.py:160] [process=6][thread=async_save] Background save thread done. Time taken: 44.568916s. I0422 14:45:12.496913 139520393602816 async_checkpointer.py:288] [process=6][thread=save_finalize] Done with waiting for background save thread=async_save. I0422 14:45:12.497041 139520393602816 async_checkpointer.py:298] [process=6][thread=save_finalize] No errors found in background save thread=async_save. I0422 14:45:12.497108 139520393602816 checkpoint_manager.py:2137] [process=6][thread=save_finalize][step=0] CheckpointManager Save Finalize is syncing with other hosts... I0422 14:45:12.498382 139520393602816 checkpoint_manager.py:2146] [process=6][thread=save_finalize][step=0] CheckpointManager Save Finalize is done on all hosts. I0422 14:45:12.498576 139648287532864 checkpoint_manager.py:2032] [process=6][thread=MainThread][step=0][wait_until_finished] Done waiting for Save Finalize thread (save_finalize) running at step=0. W0422 14:45:12.498712 139648287532864 checkpoint_manager.py:1452] Waiting for previous save to complete took 28.010382 seconds. If this number is high, consider checkpointing less frequently. I0422 14:45:12.500846 139648287532864 checkpoint_manager.py:1512] [process=6] Saving checkpoint at step 9 I0422 14:45:12.502810 139648287532864 event_tracking.py:70] [process=6] [async] Started save checkpoint @ gs://lance-maxtext/nnx_ckpt_xpk_feat_nnx_post_train_fixes_20260422_123906/nnx_xpk_feat_nnx_post_train_fixes_20260422_123906_09_pdb_lt_1/checkpoints/9. I0422 14:45:13.081954 139648287532864 jax_array_handlers.py:360] Scheduling D2H of 69 prioritized jax.Array. I0422 14:45:13.082052 139648287532864 replica_slices.py:424] Transferring arrays to host memory with options: use_replica_parallel=True, min_slice_bytes_for_replica_parallel=None, max_replicas_for_replica_parallel=None, enable_pinned_host_transfer=False I0422 14:45:13.108331 139648287532864 base_pytree_checkpoint_handler.py:154] [process=6][thread=MainThread] Initiated "orbax.checkpoint._src.serialization.jax_array_handlers.ArrayHandler".serialize. Time taken: 0.028127s I0422 14:45:13.108460 139648287532864 base_pytree_checkpoint_handler.py:130] [process=6] /jax/orbax/write/blocking_gbytes_per_sec: 43.941 GiB/s (total gbytes: 1.5 GiB) (time elapsed: 0.03510546684265137 s) (per-host) I0422 14:45:13.108510 139648287532864 base_pytree_checkpoint_handler.py:768] [process=6][thread=MainThread] Initiated Pytree async_save. Time taken: 0.035164s (batch_requests_ready=0.002671s, total_serialization_initiated=0.032428s, others=0.000066s) I0422 14:45:13.108592 139648287532864 composite_checkpoint_handler.py:715] [process=6][thread=MainThread] Initiated CompositeCheckpointHandler.async_save. Time taken: 0.039288s (all_items=0.000015s, per_item={'items': '0.00001550'}, temp_paths=0.039272) I0422 14:45:13.109308 139648287532864 event_tracking.py:125] [process=6] [async] Finished blocking save in 0.61 seconds. Continuing save @ gs://lance-maxtext/nnx_ckpt_xpk_feat_nnx_post_train_fixes_20260422_123906/nnx_xpk_feat_nnx_post_train_fixes_20260422_123906_09_pdb_lt_1/checkpoints/9. I0422 14:45:13.109598 139520393602816 async_checkpointer.py:76] [process=6][thread=async_save] Background save thread started. Deadline for this save operation is 2026-04-22 15:05:13.109565 I0422 14:45:13.114412 139648287532864 checkpoint_manager.py:1560] [process=6][thread=MainThread][step=9] Starting CheckpointManager Save Finalize thread=save_finalize I0422 14:45:13.114647 139524654999296 async_checkpointer.py:280] [process=6][thread=save_finalize] Waiting for background save thread=async_save. I0422 14:45:13.114777 139648287532864 standard_logger.py:34] {'step': 9, 'event_type': 'save', 'directory': 'gs://lance-maxtext/nnx_ckpt_xpk_feat_nnx_post_train_fixes_20260422_123906/nnx_xpk_feat_nnx_post_train_fixes_20260422_123906_09_pdb_lt_1/checkpoints', 'reached_preemption': False, 'preemption_received_at': None, 'synchronous': False, 'wait_for_prev_start_time': 1776869084.4882972, 'wait_for_prev_duration_secs': 28.010381937026978, 'time_between_consecutive_saves_sec': None, 'checkpointer_blocking_start_time': 1776869112.5008857, 'checkpointer_blocking_duration_secs': 0.6088566780090332, 'get_old_steps_start_time': 1776869113.109766, 'get_old_steps_duration_secs': 3.0517578125e-05, 'checkpoint_manager_blocking_start_time': 1776869084.4864998, 'checkpoint_manager_blocking_duration_secs': 28.628241300582886} I0422 14:45:13.114895 139648287532864 checkpointing.py:409] Started an asynchronous checkpoint save for step 9 I0422 14:45:13.114939 139648287532864 checkpoint_manager.py:2020] [process=6][thread=MainThread][step=9][wait_until_finished] Waiting for Save Finalize thread (save_finalize) to complete. I0422 14:45:18.272360 139513959544576 array_metadata_store.py:203] [process=6][thread=array_type_handler] Wrote 69 array_metadata.ArrayMetadata to gs://lance-maxtext/nnx_ckpt_xpk_feat_nnx_post_train_fixes_20260422_123906/nnx_xpk_feat_nnx_post_train_fixes_20260422_123906_09_pdb_lt_1/checkpoints/9/items/array_metadatas/process_6 I0422 14:45:54.136421 139520393602816 base_pytree_checkpoint_handler.py:130] [process=6] /jax/orbax/write/gbytes_per_sec: 38.468 MiB/s (total gbytes: 1.5 GiB) (time elapsed: 41.063021183013916 s) (per-host) I0422 14:45:54.136549 139520393602816 async_checkpointer.py:90] [process=6][thread=async_save] 3 Handler Commit operations completed. Time taken: 41.026835s. I0422 14:46:02.058748 139520393602816 async_checkpointer.py:160] [process=6][thread=async_save] Background save thread done. Time taken: 48.949018s. I0422 14:46:02.059036 139524654999296 async_checkpointer.py:288] [process=6][thread=save_finalize] Done with waiting for background save thread=async_save. I0422 14:46:02.059194 139524654999296 async_checkpointer.py:298] [process=6][thread=save_finalize] No errors found in background save thread=async_save. I0422 14:46:02.059245 139524654999296 checkpoint_manager.py:2137] [process=6][thread=save_finalize][step=9] CheckpointManager Save Finalize is syncing with other hosts... I0422 14:46:02.060706 139524654999296 checkpoint_manager.py:2146] [process=6][thread=save_finalize][step=9] CheckpointManager Save Finalize is done on all hosts. I0422 14:46:02.060889 139648287532864 checkpoint_manager.py:2032] [process=6][thread=MainThread][step=9][wait_until_finished] Done waiting for Save Finalize thread (save_finalize) running at step=9. I0422 14:46:02.061058 139648287532864 checkpoint_manager.py:2009] [process=6][thread=MainThread][wait_until_finished] No Save Finalize thread to wait for. Returning. I0422 14:46:02.062091 139648287532864 metric_logger.py:196] completed step: 9, seconds: 0.059, TFLOP/s/device: 57.814, Tokens/s/device: 8714.300, total_weights: 16384, loss: 5.890, lm_loss: 5.890, perplexity: 361.390 Per train step: Total TFLOPs: 3.40 split as 93.93% learnable weight flops and 6.07% attention flops XPK End: Wed Apr 22 14:46:13 UTC 2026 EXIT_CODE=0