XPK Start: Thu Apr 16 22:22:41 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 22:23:05.010815: 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 22:23:05.541899 136603194873664 max_utils.py:273] Attempting to initialize the jax distributed system... I0416 22:23:14.580832 136603194873664 distributed.py:149] Starting JAX distributed service on [::]:8482 I0416 22:23:14.583520 136603194873664 distributed.py:172] Connecting to JAX distributed service on mt-04-int8-7y1qt-slice-job-0-0.mt-04-int8-7y1qt:8482 I0416 22:23:15.588872 136603194873664 max_utils.py:284] Jax distributed system initialized! I0416 22:23:20.591897 136603194873664 max_utils.py:800] System Information: Jax Version: 0.9.2 I0416 22:23:20.591997 136603194873664 max_utils.py:801] System Information: Jaxlib Version: 0.9.2 I0416 22:23:20.592038 136603194873664 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 22:23:20.592072 136603194873664 train_utils.py:377] WARNING: Sequence packing is essentially ignored for synthetic data. Please use a real dataset to use sequence packing. I0416 22:23:21.257645 136603194873664 maxtext_utils.py:1687] Num_devices: 32, shape (1, 1, 1, 32, 1, 1, 1, 1, 1, 1, 1, 1, 1) I0416 22:23:21.817808 136603194873664 checkpointing.py:688] Setting up checkpoint logger... I0416 22:23:21.817933 136603194873664 checkpointing.py:234] Creating checkpoint manager with ocdbt=True and zarr3=True I0416 22:23:21.817978 136603194873664 pytree_checkpoint_handler.py:592] save_device_host_concurrent_bytes=None I0416 22:23:21.818200 136603194873664 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 0x7c3cd6f1c110>, enable_pinned_host_transfer=False, save_concurrent_bytes: 96000000000 (89.4 GiB), restore_concurrent_bytes: 96000000000 (89.4 GiB) I0416 22:23:24.708289 136603194873664 checkpointing.py:266] Enabling policy for fixed interval checkpointing. I0416 22:23:24.708518 136603194873664 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 0x7c291c3f38f0>}, handler_registry=None I0416 22:23:24.708763 136603194873664 composite_checkpoint_handler.py:237] Deferred registration for item: "items". Adding handler `<orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeCheckpointHandler object at 0x7c291c3f38f0>` 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 22:23:24.708809 136603194873664 composite_checkpoint_handler.py:237] Deferred registration for item: "metrics". Adding handler `<orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonCheckpointHandler object at 0x7c291c3d35c0>` 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 22:23:24.708857 136603194873664 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 0x7c291c3f38f0>, ('items', <class 'orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeRestoreArgs'>): <orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeCheckpointHandler object at 0x7c291c3f38f0>, ('metrics', <class 'orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonSaveArgs'>): <orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonCheckpointHandler object at 0x7c291c3d35c0>, ('metrics', <class 'orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonRestoreArgs'>): <orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonCheckpointHandler object at 0x7c291c3d35c0>}). I0416 22:23:24.709182 136603194873664 abstract_checkpointer.py:35] orbax-checkpoint version: 0.11.34 I0416 22:23:24.709252 136603194873664 async_checkpointer.py:192] [process=0][thread=MainThread] Using barrier_sync_fn: <function get_barrier_sync_fn.<locals>._fn at 0x7c291c48ed40> timeout: 1200 secs and primary_host=0 for async checkpoint writes I0416 22:23:25.851209 136603194873664 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_04_int8/checkpoints I0416 22:23:26.278252 136603194873664 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_04_int8/checkpoints: <orbax.checkpoint.checkpoint_manager.CheckpointManager object at 0x7c291c3b9100> I0416 22:23:26.278418 136603194873664 checkpointing.py:302] Checkpoint manager created! I0416 22:23:27.634141 136603194873664 checkpointing.py:578] checkpoint manager exists so trying to load this run's existing checkpoint I0416 22:23:27.634260 136603194873664 checkpointing.py:676] No existing checkpoints found, not restoring checkpoint. fsdp: 32 I0416 22:23:29.969690 136603194873664 maxtext_utils.py:1805] decoder/decoder_norm/scale/value Shape: float32[2048] Physical: (None,) I0416 22:23:29.969805 136603194873664 maxtext_utils.py:1805] decoder/layers/mlp/wi_0/kernel/value Shape: float32[2048,16,7168] Physical: ('fsdp', None, None) I0416 22:23:29.969884 136603194873664 maxtext_utils.py:1805] decoder/layers/mlp/wi_1/kernel/value Shape: float32[2048,16,7168] Physical: ('fsdp', None, None) I0416 22:23:29.969930 136603194873664 maxtext_utils.py:1805] decoder/layers/mlp/wo/kernel/value Shape: float32[7168,16,2048] Physical: (None, None, 'fsdp') I0416 22:23:29.969967 136603194873664 maxtext_utils.py:1805] decoder/layers/post_self_attention_layer_norm/scale/value Shape: float32[2048,16] Physical: (None, None) I0416 22:23:29.970001 136603194873664 maxtext_utils.py:1805] decoder/layers/pre_self_attention_layer_norm/scale/value Shape: float32[2048,16] Physical: (None, None) I0416 22:23:29.970036 136603194873664 maxtext_utils.py:1805] decoder/layers/self_attention/key/kernel/value Shape: float32[2048,16,16,128] Physical: ('fsdp', None, None, None) I0416 22:23:29.970070 136603194873664 maxtext_utils.py:1805] decoder/layers/self_attention/out/kernel/value Shape: float32[16,16,128,2048] Physical: (None, None, None, 'fsdp') I0416 22:23:29.970103 136603194873664 maxtext_utils.py:1805] decoder/layers/self_attention/query/kernel/value Shape: float32[2048,16,16,128] Physical: ('fsdp', None, None, None) I0416 22:23:29.970134 136603194873664 maxtext_utils.py:1805] decoder/layers/self_attention/value/kernel/value Shape: float32[2048,16,16,128] Physical: ('fsdp', None, None, None) I0416 22:23:29.970164 136603194873664 maxtext_utils.py:1805] decoder/logits_dense/kernel/value Shape: float32[2048,32000] Physical: ('fsdp', None) I0416 22:23:29.970194 136603194873664 maxtext_utils.py:1805] token_embedder/embedding/value Shape: float32[32000,2048] Physical: (None, 'fsdp') I0416 22:23:30.384698 136603194873664 nnx_decoders.py:465] nnx_decoders/carry Logical: bfloat16[32,2048,2048]...................................... ('activation_batch', 'activation_norm_length', 'activation_embed'). I0416 22:23:30.384794 136603194873664 nnx_decoders.py:465] nnx_decoders/carry Physical: bfloat16[32,2048,2048]...................................... ('fsdp', None, None). I0416 22:23:30.390042 136603194873664 nnx_decoders.py:465] Unknown Logical: bfloat16[32,2048,2048]...................................... ('activation_batch', 'activation_norm_length', 'activation_embed'). I0416 22:23:30.390102 136603194873664 nnx_decoders.py:465] Unknown Physical: bfloat16[32,2048,2048]...................................... ('fsdp', None, None). I0416 22:23:30.406123 136603194873664 attentions.py:1088] attentions/inputs_q Logical: bfloat16[32,2048,2048]...................................... ('activation_batch', 'activation_attn_length', 'activation_attn_embed'). I0416 22:23:30.406188 136603194873664 attentions.py:1088] attentions/inputs_q Physical: bfloat16[32,2048,2048]...................................... ('fsdp', None, None). I0416 22:23:30.421581 136603194873664 attentions.py:1089] attentions/inputs_kv Logical: bfloat16[32,2048,2048]...................................... ('activation_batch', 'activation_attn_length', 'activation_attn_embed'). I0416 22:23:30.421638 136603194873664 attentions.py:1089] attentions/inputs_kv Physical: bfloat16[32,2048,2048]...................................... ('fsdp', None, None). I0416 22:23:30.502790 136603194873664 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 22:23:30.502899 136603194873664 attentions.py:1154] attentions/query Physical: bfloat16[32,2048,16,128].................................... ('fsdp', None, None, None). I0416 22:23:30.518500 136603194873664 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 22:23:30.518558 136603194873664 attentions.py:1155] attentions/key Physical: bfloat16[32,2048,16,128].................................... ('fsdp', None, None, None). I0416 22:23:30.534024 136603194873664 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 22:23:30.534082 136603194873664 attentions.py:1156] attentions/value Physical: bfloat16[32,2048,16,128].................................... ('fsdp', None, None, None). I0416 22:23:30.564420 136603194873664 attentions.py:1197] attentions/out Logical: bfloat16[32,2048,16,128].................................... ('activation_batch', 'activation_attn_length', 'activation_heads', 'activation_kv'). I0416 22:23:30.564497 136603194873664 attentions.py:1197] attentions/out Physical: bfloat16[32,2048,16,128].................................... ('fsdp', None, None, None). I0416 22:23:30.644040 136603194873664 linears.py:525] linears/x Logical: bfloat16[32,2048,7168]...................................... ('activation_batch', 'activation_length', 'activation_mlp'). I0416 22:23:30.644133 136603194873664 linears.py:525] linears/x Physical: bfloat16[32,2048,7168]...................................... ('fsdp', None, None). I0416 22:23:44.526428 136603194873664 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. I0416 22:23:44.536724 136603194873664 metric_logger.py:289] number parameters: 1.104 billion I0416 22:23:58.925448 136603194873664 checkpointing.py:794] Waiting for step 0 to finish before checkpoint... I0416 22:23:59.164625 136603194873664 checkpointing.py:798] Waited 0.23916220664978027 seconds for step 0 to finish before starting checkpointing. I0416 22:23:59.166982 136603194873664 checkpoint_manager.py:2009] [process=0][thread=MainThread][wait_until_finished] No Save Finalize thread to wait for. Returning. I0416 22:23:59.169121 136603194873664 checkpoint_manager.py:1512] [process=0] Saving checkpoint at step 0 I0416 22:23:59.171453 136603194873664 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_04_int8/checkpoints/0. I0416 22:23:59.903814 136603194873664 signaling_client.py:364] Using JaxDistributedSignalingClient I0416 22:23:59.932306 136603194873664 jax_array_handlers.py:360] Scheduling D2H of 111 prioritized jax.Array. I0416 22:23:59.932386 136603194873664 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 22:24:00.200065 136471536416512 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_04_int8/checkpoints/0 I0416 22:24:00.215873 136603194873664 base_pytree_checkpoint_handler.py:154] [process=0][thread=MainThread] Initiated "orbax.checkpoint._src.serialization.jax_array_handlers.ArrayHandler".serialize. Time taken: 0.286045s I0416 22:24:00.216750 136603194873664 base_pytree_checkpoint_handler.py:130] [process=0] /jax/orbax/write/blocking_gbytes_per_sec: 5.072 GiB/s (total gbytes: 1.5 GiB) (time elapsed: 0.3041224479675293 s) (per-host) I0416 22:24:00.216830 136603194873664 base_pytree_checkpoint_handler.py:768] [process=0][thread=MainThread] Initiated Pytree async_save. Time taken: 0.304222s (batch_requests_ready=0.005382s, total_serialization_initiated=0.298046s, others=0.000795s) I0416 22:24:00.216941 136603194873664 composite_checkpoint_handler.py:715] [process=0][thread=MainThread] Initiated CompositeCheckpointHandler.async_save. Time taken: 0.308818s (all_items=0.000018s, per_item={'items': '0.00001764'}, temp_paths=0.308800) I0416 22:24:00.217731 136603194873664 event_tracking.py:125] [process=0] [async] Finished blocking save in 1.05 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_04_int8/checkpoints/0. I0416 22:24:00.218027 136471579195136 async_checkpointer.py:76] [process=0][thread=async_save] Background save thread started. Deadline for this save operation is 2026-04-16 22:44:00.217990 I0416 22:24:00.220513 136603194873664 checkpoint_manager.py:1560] [process=0][thread=MainThread][step=0] Starting CheckpointManager Save Finalize thread=save_finalize I0416 22:24:00.220749 136470987781888 async_checkpointer.py:280] [process=0][thread=save_finalize] Waiting for background save thread=async_save. I0416 22:24:00.220910 136603194873664 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_04_int8/checkpoints', 'reached_preemption': False, 'preemption_received_at': None, 'synchronous': False, 'wait_for_prev_start_time': 1776378239.1669629, 'wait_for_prev_duration_secs': 6.008148193359375e-05, 'time_between_consecutive_saves_sec': None, 'checkpointer_blocking_start_time': 1776378239.1692324, 'checkpointer_blocking_duration_secs': 1.0489411354064941, 'get_old_steps_start_time': 1776378240.2182, 'get_old_steps_duration_secs': 2.8848648071289062e-05, 'checkpoint_manager_blocking_start_time': 1776378239.1650686, 'checkpoint_manager_blocking_duration_secs': 1.0558030605316162} I0416 22:24:00.221024 136603194873664 checkpointing.py:409] Started an asynchronous checkpoint save for step 0 I0416 22:24:00.221102 136603194873664 max_utils.py:750] Memstats: After params initialized: I0416 22:24:00.221149 136603194873664 max_utils.py:756] Using (GB) 0.44 / 31.25 (1.408000%) on TPU_0(process=0,(0,0,0,0)) I0416 22:24:00.221183 136603194873664 max_utils.py:756] Using (GB) 0.43 / 31.25 (1.376000%) on TPU_1(process=0,(1,0,0,0)) I0416 22:24:00.221210 136603194873664 max_utils.py:756] Using (GB) 0.43 / 31.25 (1.376000%) on TPU_4(process=0,(0,1,0,0)) I0416 22:24:00.221234 136603194873664 max_utils.py:756] Using (GB) 0.43 / 31.25 (1.376000%) on TPU_5(process=0,(1,1,0,0)) I0416 22:24:00.568598 136603194873664 metric_logger.py:185] completed step: 0, seconds: 14.000, TFLOP/s/device: 0.970, Tokens/s/device: 146.282, total_weights: 65536, loss: 10.872 I0416 22:24:00.571873 136603194873664 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_04_int8/tensorboard/' I0416 22:24:01.008130 136603194873664 metric_logger.py:185] completed step: 1, seconds: 1.638, TFLOP/s/device: 8.295, Tokens/s/device: 1250.333, total_weights: 65536, loss: 10.872 I0416 22:24:01.227745 136603194873664 metric_logger.py:185] completed step: 2, seconds: 0.441, TFLOP/s/device: 30.828, Tokens/s/device: 4646.709, total_weights: 65536, loss: 10.260 I0416 22:24:01.457221 136603194873664 metric_logger.py:185] completed step: 3, seconds: 0.015, TFLOP/s/device: 877.495, Tokens/s/device: 132265.564, total_weights: 65536, loss: 9.728 I0416 22:24:02.088795 136468856104704 checkpoint.py:188] Wrote Metadata={'item_handlers': None, 'metrics': {}, 'performance_metrics': {}, 'init_timestamp_nsecs': 1776378241700493626, 'commit_timestamp_nsecs': None, 'custom_metadata': {}}, json={"item_handlers": null, "metrics": {}, "performance_metrics": {}, "init_timestamp_nsecs": 1776378241700493626, "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_04_int8/checkpoints/0/_CHECKPOINT_METADATA I0416 22:24:02.148990 136471042324224 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_04_int8/checkpoints/0/items I0416 22:24:03.323500 2916 google_auth_provider.cc:181] Running on GCE, using service account 562977990677-compute@developer.gserviceaccount.com I0416 22:24:06.236189 136479092332288 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_20260416_221011/nnx_xpk_feat_nnx_post_train_fixes_20260416_221011_04_int8/checkpoints/0/items/array_metadatas/process_0 I0416 22:24:29.573267 136603194873664 metric_logger.py:185] completed step: 4, seconds: 0.220, TFLOP/s/device: 61.703, Tokens/s/device: 9300.594, total_weights: 65536, loss: 9.271 I0416 22:24:29.621549 136603194873664 metric_logger.py:185] completed step: 5, seconds: 0.230, TFLOP/s/device: 59.021, Tokens/s/device: 8896.341, total_weights: 65536, loss: 8.885 I0416 22:24:29.810951 136603194873664 metric_logger.py:185] completed step: 6, seconds: 28.115, TFLOP/s/device: 0.483, Tokens/s/device: 72.845, total_weights: 65536, loss: 8.586 I0416 22:24:30.040315 136603194873664 metric_logger.py:185] completed step: 7, seconds: 0.045, TFLOP/s/device: 299.831, Tokens/s/device: 45193.751, total_weights: 65536, loss: 8.378 I0416 22:24:30.269881 136603194873664 metric_logger.py:185] completed step: 8, seconds: 0.200, TFLOP/s/device: 68.073, Tokens/s/device: 10260.675, total_weights: 65536, loss: 8.249 I0416 22:24:30.500994 136603194873664 checkpointing.py:794] Waiting for step 9 to finish before checkpoint... I0416 22:24:30.502016 136603194873664 checkpointing.py:798] Waited 0.0010328292846679688 seconds for step 9 to finish before starting checkpointing. I0416 22:24:30.504057 136603194873664 checkpoint_manager.py:2020] [process=0][thread=MainThread][step=0][wait_until_finished] Waiting for Save Finalize thread (save_finalize) to complete. I0416 22:24:36.698950 136470996174592 base_pytree_checkpoint_handler.py:1282] [process=0][thread=write_metadata_after_commits] Commit + Array metadata written. Time taken: 33.384690s (commit=32.017175s, array_metadata_write=1.367515s) I0416 22:24:36.700208 136471579195136 base_pytree_checkpoint_handler.py:130] [process=0] /jax/orbax/write/gbytes_per_sec: 42.939 MiB/s (total gbytes: 1.5 GiB) (time elapsed: 36.78755283355713 s) (per-host) I0416 22:24:36.700331 136471579195136 async_checkpointer.py:90] [process=0][thread=async_save] 3 Handler Commit operations completed. Time taken: 36.482198s. I0416 22:24:37.395326 136471579195136 checkpoint.py:228] Read Metadata={'item_handlers': None, 'metrics': {}, 'performance_metrics': {}, 'init_timestamp_nsecs': 1776378241700493626, '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_04_int8/checkpoints/0/_CHECKPOINT_METADATA I0416 22:24:38.132523 136468856104704 checkpoint.py:247] Updated Metadata={'item_handlers': {'items': 'orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeCheckpointHandler'}, 'metrics': {}, 'performance_metrics': {}, 'init_timestamp_nsecs': 1776378241700493626, '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_04_int8/checkpoints/0/_CHECKPOINT_METADATA I0416 22:24:38.138052 136471579195136 array_metadata_store.py:411] [process=0][thread=async_save] Validated ArrayMetadata from all 8 hosts. Time taken: 0.000865s. I0416 22:24:38.806420 136471579195136 ocdbt_utils.py:49] Param validation support for Zarr3 will be added later (b/362328389). I0416 22:24:39.738150 136471579195136 base_pytree_checkpoint_handler.py:1406] [process=0][thread=async_save] Pytree save finalize (merge_ocdbt + ArrayMetadata validation) completed. Time taken: 2.216993s. 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_04_int8/checkpoints/0/items I0416 22:24:39.739918 136471579195136 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_04_int8/checkpoints/0/items I0416 22:24:40.299105 136471579195136 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_04_int8/checkpoints/0 I0416 22:24:42.087286 136471579195136 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_04_int8/checkpoints/0`. I0416 22:24:42.088125 136471579195136 event_tracking.py:138] [process=0] [async] Finished save (blocking + background) in 42.92 seconds @ gs://lance-maxtext/nnx_ckpt_xpk_feat_nnx_post_train_fixes_20260416_221011/nnx_xpk_feat_nnx_post_train_fixes_20260416_221011_04_int8/checkpoints/0 I0416 22:24:42.089480 136471579195136 async_checkpointer.py:160] [process=0][thread=async_save] Background save thread done. Time taken: 41.871345s. I0416 22:24:42.089690 136470987781888 async_checkpointer.py:288] [process=0][thread=save_finalize] Done with waiting for background save thread=async_save. I0416 22:24:42.089816 136470987781888 async_checkpointer.py:298] [process=0][thread=save_finalize] No errors found in background save thread=async_save. I0416 22:24:42.089918 136470987781888 checkpoint_manager.py:2137] [process=0][thread=save_finalize][step=0] CheckpointManager Save Finalize is syncing with other hosts... I0416 22:24:42.092497 136470987781888 checkpoint_manager.py:2146] [process=0][thread=save_finalize][step=0] CheckpointManager Save Finalize is done on all hosts. I0416 22:24:42.092674 136603194873664 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 22:24:42.092805 136603194873664 checkpoint_manager.py:1452] Waiting for previous save to complete took 11.588747 seconds. If this number is high, consider checkpointing less frequently. I0416 22:24:42.094654 136603194873664 checkpoint_manager.py:1512] [process=0] Saving checkpoint at step 9 I0416 22:24:42.096698 136603194873664 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_04_int8/checkpoints/9. I0416 22:24:42.627946 136603194873664 jax_array_handlers.py:360] Scheduling D2H of 111 prioritized jax.Array. I0416 22:24:42.628044 136603194873664 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 22:24:42.657071 136603194873664 base_pytree_checkpoint_handler.py:154] [process=0][thread=MainThread] Initiated "orbax.checkpoint._src.serialization.jax_array_handlers.ArrayHandler".serialize. Time taken: 0.031505s I0416 22:24:42.657874 136603194873664 base_pytree_checkpoint_handler.py:130] [process=0] /jax/orbax/write/blocking_gbytes_per_sec: 33.372 GiB/s (total gbytes: 1.5 GiB) (time elapsed: 0.046224355697631836 s) (per-host) I0416 22:24:42.657952 136603194873664 base_pytree_checkpoint_handler.py:768] [process=0][thread=MainThread] Initiated Pytree async_save. Time taken: 0.046336s (batch_requests_ready=0.004444s, total_serialization_initiated=0.041165s, others=0.000727s) I0416 22:24:42.658040 136603194873664 composite_checkpoint_handler.py:715] [process=0][thread=MainThread] Initiated CompositeCheckpointHandler.async_save. Time taken: 0.050199s (all_items=0.000016s, per_item={'items': '0.00001621'}, temp_paths=0.050183) I0416 22:24:42.658790 136603194873664 event_tracking.py:125] [process=0] [async] Finished blocking save in 0.56 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_04_int8/checkpoints/9. I0416 22:24:42.659078 136470987781888 async_checkpointer.py:76] [process=0][thread=async_save] Background save thread started. Deadline for this save operation is 2026-04-16 22:44:42.659050 I0416 22:24:42.663177 136603194873664 checkpoint_manager.py:1560] [process=0][thread=MainThread][step=9] Starting CheckpointManager Save Finalize thread=save_finalize I0416 22:24:42.663441 136470497060608 async_checkpointer.py:280] [process=0][thread=save_finalize] Waiting for background save thread=async_save. I0416 22:24:42.663607 136603194873664 standard_logger.py:34] {'step': 9, '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_04_int8/checkpoints', 'reached_preemption': False, 'preemption_received_at': None, 'synchronous': False, 'wait_for_prev_start_time': 1776378270.504025, 'wait_for_prev_duration_secs': 11.588746547698975, 'time_between_consecutive_saves_sec': None, 'checkpointer_blocking_start_time': 1776378282.0946956, 'checkpointer_blocking_duration_secs': 0.5645034313201904, 'get_old_steps_start_time': 1776378282.6592238, 'get_old_steps_duration_secs': 2.9802322387695312e-05, 'checkpoint_manager_blocking_start_time': 1776378270.5024152, 'checkpoint_manager_blocking_duration_secs': 12.161155939102173} I0416 22:24:42.663723 136603194873664 checkpointing.py:409] Started an asynchronous checkpoint save for step 9 I0416 22:24:42.663770 136603194873664 checkpoint_manager.py:2020] [process=0][thread=MainThread][step=9][wait_until_finished] Waiting for Save Finalize thread (save_finalize) to complete. I0416 22:24:42.856624 136471579195136 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_04_int8/checkpoints/9 I0416 22:24:44.575683 136471025538816 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_04_int8/checkpoints/9/items I0416 22:24:47.463743 136471008167680 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_20260416_221011/nnx_xpk_feat_nnx_post_train_fixes_20260416_221011_04_int8/checkpoints/9/items/array_metadatas/process_0 I0416 22:25:26.515551 136470999774976 base_pytree_checkpoint_handler.py:1282] [process=0][thread=write_metadata_after_commits] Commit + Array metadata written. Time taken: 40.811417s (commit=39.584331s, array_metadata_write=1.227087s) I0416 22:25:26.516692 136470987781888 base_pytree_checkpoint_handler.py:130] [process=0] /jax/orbax/write/gbytes_per_sec: 35.978 MiB/s (total gbytes: 1.5 GiB) (time elapsed: 43.905022382736206 s) (per-host) I0416 22:25:26.516822 136470987781888 async_checkpointer.py:90] [process=0][thread=async_save] 3 Handler Commit operations completed. Time taken: 43.857645s. I0416 22:25:28.363406 136470987781888 array_metadata_store.py:411] [process=0][thread=async_save] Validated ArrayMetadata from all 8 hosts. Time taken: 0.000455s. I0416 22:25:28.614458 136470987781888 ocdbt_utils.py:49] Param validation support for Zarr3 will be added later (b/362328389). I0416 22:25:29.561712 136470987781888 base_pytree_checkpoint_handler.py:1406] [process=0][thread=async_save] Pytree save finalize (merge_ocdbt + ArrayMetadata validation) completed. Time taken: 2.190718s. 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_04_int8/checkpoints/9/items I0416 22:25:29.563427 136470987781888 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_04_int8/checkpoints/9/items I0416 22:25:30.092958 136470987781888 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_04_int8/checkpoints/9 I0416 22:25:31.466743 136470987781888 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_04_int8/checkpoints/9`. I0416 22:25:31.467632 136470987781888 event_tracking.py:138] [process=0] [async] Finished save (blocking + background) in 49.37 seconds @ gs://lance-maxtext/nnx_ckpt_xpk_feat_nnx_post_train_fixes_20260416_221011/nnx_xpk_feat_nnx_post_train_fixes_20260416_221011_04_int8/checkpoints/9 I0416 22:25:31.469129 136470987781888 async_checkpointer.py:160] [process=0][thread=async_save] Background save thread done. Time taken: 48.809949s. I0416 22:25:31.469291 136470497060608 async_checkpointer.py:288] [process=0][thread=save_finalize] Done with waiting for background save thread=async_save. I0416 22:25:31.469371 136470497060608 async_checkpointer.py:298] [process=0][thread=save_finalize] No errors found in background save thread=async_save. I0416 22:25:31.469432 136470497060608 checkpoint_manager.py:2137] [process=0][thread=save_finalize][step=9] CheckpointManager Save Finalize is syncing with other hosts... I0416 22:25:31.471060 136470497060608 checkpoint_manager.py:2146] [process=0][thread=save_finalize][step=9] CheckpointManager Save Finalize is done on all hosts. I0416 22:25:31.471262 136603194873664 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. I0416 22:25:31.471452 136603194873664 checkpoint_manager.py:2009] [process=0][thread=MainThread][wait_until_finished] No Save Finalize thread to wait for. Returning. I0416 22:25:31.472184 136603194873664 metric_logger.py:185] completed step: 9, seconds: 0.220, TFLOP/s/device: 61.750, Tokens/s/device: 9307.652, 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: Thu Apr 16 22:25:41 UTC 2026 EXIT_CODE=0