XPK Start: Fri Apr 17 00:39:05 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 00:39:28.995815: 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 00:39:29.173062 140240525641536 max_utils.py:273] Attempting to initialize the jax distributed system... I0417 00:39:38.213505 140240525641536 distributed.py:149] Starting JAX distributed service on [::]:8482 I0417 00:39:38.215758 140240525641536 distributed.py:172] Connecting to JAX distributed service on mt-02-synthetic-ql1o8-slice-job-0-0.mt-02-synthetic-ql1o8:8482 I0417 00:39:39.251541 140240525641536 max_utils.py:284] Jax distributed system initialized! I0417 00:39:45.401647 140240525641536 max_utils.py:800] System Information: Jax Version: 0.9.2 I0417 00:39:45.401748 140240525641536 max_utils.py:801] System Information: Jaxlib Version: 0.9.2 I0417 00:39:45.401788 140240525641536 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 00:39:45.401823 140240525641536 train_utils.py:377] WARNING: Sequence packing is essentially ignored for synthetic data. Please use a real dataset to use sequence packing. I0417 00:39:46.077269 140240525641536 maxtext_utils.py:1687] Num_devices: 32, shape (1, 1, 1, 32, 1, 1, 1, 1, 1, 1, 1, 1, 1) I0417 00:39:46.189171 140240525641536 checkpointing.py:688] Setting up checkpoint logger... I0417 00:39:46.189290 140240525641536 checkpointing.py:234] Creating checkpoint manager with ocdbt=True and zarr3=True I0417 00:39:46.189335 140240525641536 pytree_checkpoint_handler.py:592] save_device_host_concurrent_bytes=None I0417 00:39:46.189538 140240525641536 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 0x7f8b93238230>, enable_pinned_host_transfer=False, save_concurrent_bytes: 96000000000 (89.4 GiB), restore_concurrent_bytes: 96000000000 (89.4 GiB) I0417 00:39:48.969581 140240525641536 checkpointing.py:266] Enabling policy for fixed interval checkpointing. I0417 00:39:48.969815 140240525641536 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 0x7f778824bf50>}, handler_registry=None I0417 00:39:48.970074 140240525641536 composite_checkpoint_handler.py:237] Deferred registration for item: "items". Adding handler `<orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeCheckpointHandler object at 0x7f778824bf50>` 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 00:39:48.970123 140240525641536 composite_checkpoint_handler.py:237] Deferred registration for item: "metrics". Adding handler `<orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonCheckpointHandler object at 0x7f7788248800>` 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 00:39:48.970157 140240525641536 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 0x7f778824bf50>, ('items', <class 'orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeRestoreArgs'>): <orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeCheckpointHandler object at 0x7f778824bf50>, ('metrics', <class 'orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonSaveArgs'>): <orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonCheckpointHandler object at 0x7f7788248800>, ('metrics', <class 'orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonRestoreArgs'>): <orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonCheckpointHandler object at 0x7f7788248800>}). I0417 00:39:48.970476 140240525641536 abstract_checkpointer.py:35] orbax-checkpoint version: 0.11.34 I0417 00:39:48.970546 140240525641536 async_checkpointer.py:192] [process=0][thread=MainThread] Using barrier_sync_fn: <function get_barrier_sync_fn.<locals>._fn at 0x7f77881462a0> timeout: 1200 secs and primary_host=0 for async checkpoint writes I0417 00:39:50.355272 140240525641536 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_02_synthetic/checkpoints I0417 00:39:50.758264 140240525641536 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_02_synthetic/checkpoints: <orbax.checkpoint.checkpoint_manager.CheckpointManager object at 0x7f77704a0e00> I0417 00:39:50.758426 140240525641536 checkpointing.py:302] Checkpoint manager created! I0417 00:39:51.711337 140240525641536 checkpointing.py:578] checkpoint manager exists so trying to load this run's existing checkpoint I0417 00:39:51.711455 140240525641536 checkpointing.py:676] No existing checkpoints found, not restoring checkpoint. fsdp: 32 I0417 00:39:53.078113 140240525641536 maxtext_utils.py:1805] decoder/decoder_norm/scale/value Shape: float32[2048] Physical: (None,) I0417 00:39:53.078228 140240525641536 maxtext_utils.py:1805] decoder/layers/mlp/wi_0/kernel/value Shape: float32[2048,16,7168] Physical: ('fsdp', None, None) I0417 00:39:53.078278 140240525641536 maxtext_utils.py:1805] decoder/layers/mlp/wi_1/kernel/value Shape: float32[2048,16,7168] Physical: ('fsdp', None, None) I0417 00:39:53.078319 140240525641536 maxtext_utils.py:1805] decoder/layers/mlp/wo/kernel/value Shape: float32[7168,16,2048] Physical: (None, None, 'fsdp') I0417 00:39:53.078356 140240525641536 maxtext_utils.py:1805] decoder/layers/post_self_attention_layer_norm/scale/value Shape: float32[2048,16] Physical: (None, None) I0417 00:39:53.078391 140240525641536 maxtext_utils.py:1805] decoder/layers/pre_self_attention_layer_norm/scale/value Shape: float32[2048,16] Physical: (None, None) I0417 00:39:53.078426 140240525641536 maxtext_utils.py:1805] decoder/layers/self_attention/key/kernel/value Shape: float32[2048,16,16,128] Physical: ('fsdp', None, None, None) I0417 00:39:53.078461 140240525641536 maxtext_utils.py:1805] decoder/layers/self_attention/out/kernel/value Shape: float32[16,16,128,2048] Physical: (None, None, None, 'fsdp') I0417 00:39:53.078494 140240525641536 maxtext_utils.py:1805] decoder/layers/self_attention/query/kernel/value Shape: float32[2048,16,16,128] Physical: ('fsdp', None, None, None) I0417 00:39:53.078525 140240525641536 maxtext_utils.py:1805] decoder/layers/self_attention/value/kernel/value Shape: float32[2048,16,16,128] Physical: ('fsdp', None, None, None) I0417 00:39:53.078556 140240525641536 maxtext_utils.py:1805] decoder/logits_dense/kernel/value Shape: float32[2048,32000] Physical: ('fsdp', None) I0417 00:39:53.078587 140240525641536 maxtext_utils.py:1805] token_embedder/embedding/value Shape: float32[32000,2048] Physical: (None, 'fsdp') I0417 00:39:53.474262 140240525641536 nnx_decoders.py:465] nnx_decoders/carry Logical: bfloat16[32,2048,2048]...................................... ('activation_batch', 'activation_norm_length', 'activation_embed'). I0417 00:39:53.474352 140240525641536 nnx_decoders.py:465] nnx_decoders/carry Physical: bfloat16[32,2048,2048]...................................... ('fsdp', None, None). I0417 00:39:53.479577 140240525641536 nnx_decoders.py:465] Unknown Logical: bfloat16[32,2048,2048]...................................... ('activation_batch', 'activation_norm_length', 'activation_embed'). I0417 00:39:53.479632 140240525641536 nnx_decoders.py:465] Unknown Physical: bfloat16[32,2048,2048]...................................... ('fsdp', None, None). I0417 00:39:53.496409 140240525641536 attentions.py:1088] attentions/inputs_q Logical: bfloat16[32,2048,2048]...................................... ('activation_batch', 'activation_attn_length', 'activation_attn_embed'). I0417 00:39:53.496470 140240525641536 attentions.py:1088] attentions/inputs_q Physical: bfloat16[32,2048,2048]...................................... ('fsdp', None, None). I0417 00:39:53.512511 140240525641536 attentions.py:1089] attentions/inputs_kv Logical: bfloat16[32,2048,2048]...................................... ('activation_batch', 'activation_attn_length', 'activation_attn_embed'). I0417 00:39:53.512567 140240525641536 attentions.py:1089] attentions/inputs_kv Physical: bfloat16[32,2048,2048]...................................... ('fsdp', None, None). I0417 00:39:53.536759 140240525641536 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 00:39:53.536826 140240525641536 attentions.py:1154] attentions/query Physical: bfloat16[32,2048,16,128].................................... ('fsdp', None, None, None). I0417 00:39:53.553043 140240525641536 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 00:39:53.553100 140240525641536 attentions.py:1155] attentions/key Physical: bfloat16[32,2048,16,128].................................... ('fsdp', None, None, None). I0417 00:39:53.570122 140240525641536 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 00:39:53.570187 140240525641536 attentions.py:1156] attentions/value Physical: bfloat16[32,2048,16,128].................................... ('fsdp', None, None, None). I0417 00:39:53.602429 140240525641536 attentions.py:1197] attentions/out Logical: bfloat16[32,2048,16,128].................................... ('activation_batch', 'activation_attn_length', 'activation_heads', 'activation_kv'). I0417 00:39:53.602499 140240525641536 attentions.py:1197] attentions/out Physical: bfloat16[32,2048,16,128].................................... ('fsdp', None, None, None). I0417 00:39:53.627768 140240525641536 linears.py:525] linears/x Logical: bfloat16[32,2048,7168]...................................... ('activation_batch', 'activation_length', 'activation_mlp'). I0417 00:39:53.627871 140240525641536 linears.py:525] linears/x Physical: bfloat16[32,2048,7168]...................................... ('fsdp', None, None). I0417 00:40:04.676988 140240525641536 max_utils.py:791] Total memory size: 1.5 GB, Output size: 0.4 GB, Temp size: 1.1 GB, Argument size: 0.4 GB, Host temp size: 0.0 GB. I0417 00:40:04.686091 140240525641536 metric_logger.py:289] number parameters: 1.104 billion I0417 00:40:16.068712 140240525641536 checkpointing.py:794] Waiting for step 0 to finish before checkpoint... I0417 00:40:16.220700 140240525641536 checkpointing.py:798] Waited 0.15197205543518066 seconds for step 0 to finish before starting checkpointing. I0417 00:40:16.223148 140240525641536 checkpoint_manager.py:2009] [process=0][thread=MainThread][wait_until_finished] No Save Finalize thread to wait for. Returning. I0417 00:40:16.224949 140240525641536 checkpoint_manager.py:1512] [process=0] Saving checkpoint at step 0 I0417 00:40:16.226613 140240525641536 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_02_synthetic/checkpoints/0. I0417 00:40:16.954713 140240525641536 signaling_client.py:364] Using JaxDistributedSignalingClient I0417 00:40:16.974791 140240525641536 jax_array_handlers.py:360] Scheduling D2H of 69 prioritized jax.Array. I0417 00:40:16.974885 140240525641536 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 00:40:17.241993 140111677581056 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_02_synthetic/checkpoints/0 I0417 00:40:17.244679 140240525641536 base_pytree_checkpoint_handler.py:154] [process=0][thread=MainThread] Initiated "orbax.checkpoint._src.serialization.jax_array_handlers.ArrayHandler".serialize. Time taken: 0.271363s I0417 00:40:17.245567 140240525641536 base_pytree_checkpoint_handler.py:130] [process=0] /jax/orbax/write/blocking_gbytes_per_sec: 5.460 GiB/s (total gbytes: 1.5 GiB) (time elapsed: 0.2825014591217041 s) (per-host) I0417 00:40:17.245649 140240525641536 base_pytree_checkpoint_handler.py:768] [process=0][thread=MainThread] Initiated Pytree async_save. Time taken: 0.282607s (batch_requests_ready=0.003379s, total_serialization_initiated=0.278424s, others=0.000804s) I0417 00:40:17.245738 140240525641536 composite_checkpoint_handler.py:715] [process=0][thread=MainThread] Initiated CompositeCheckpointHandler.async_save. Time taken: 0.287054s (all_items=0.000017s, per_item={'items': '0.00001740'}, temp_paths=0.287036) I0417 00:40:17.246431 140240525641536 event_tracking.py:125] [process=0] [async] Finished blocking save in 1.02 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_02_synthetic/checkpoints/0. I0417 00:40:17.246683 140108476970752 async_checkpointer.py:76] [process=0][thread=async_save] Background save thread started. Deadline for this save operation is 2026-04-17 01:00:17.246646 I0417 00:40:17.271814 140240525641536 checkpoint_manager.py:1560] [process=0][thread=MainThread][step=0] Starting CheckpointManager Save Finalize thread=save_finalize I0417 00:40:17.272141 140111685973760 async_checkpointer.py:280] [process=0][thread=save_finalize] Waiting for background save thread=async_save. I0417 00:40:17.272292 140240525641536 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_02_synthetic/checkpoints', 'reached_preemption': False, 'preemption_received_at': None, 'synchronous': False, 'wait_for_prev_start_time': 1776386416.2231295, 'wait_for_prev_duration_secs': 6.079673767089844e-05, 'time_between_consecutive_saves_sec': None, 'checkpointer_blocking_start_time': 1776386416.225032, 'checkpointer_blocking_duration_secs': 1.0217986106872559, 'get_old_steps_start_time': 1776386417.246864, 'get_old_steps_duration_secs': 2.7894973754882812e-05, 'checkpoint_manager_blocking_start_time': 1776386416.2211185, 'checkpoint_manager_blocking_duration_secs': 1.0511386394500732} I0417 00:40:17.272399 140240525641536 checkpointing.py:409] Started an asynchronous checkpoint save for step 0 I0417 00:40:17.272466 140240525641536 max_utils.py:750] Memstats: After params initialized: I0417 00:40:17.272516 140240525641536 max_utils.py:756] Using (GB) 0.43 / 31.25 (1.376000%) on TPU_0(process=0,(0,0,0,0)) I0417 00:40:17.272547 140240525641536 max_utils.py:756] Using (GB) 0.43 / 31.25 (1.376000%) on TPU_1(process=0,(1,0,0,0)) I0417 00:40:17.272572 140240525641536 max_utils.py:756] Using (GB) 0.43 / 31.25 (1.376000%) on TPU_4(process=0,(0,1,0,0)) I0417 00:40:17.272595 140240525641536 max_utils.py:756] Using (GB) 0.43 / 31.25 (1.376000%) on TPU_5(process=0,(1,1,0,0)) I0417 00:40:17.623342 140240525641536 metric_logger.py:185] completed step: 0, seconds: 11.080, TFLOP/s/device: 1.226, Tokens/s/device: 184.837, total_weights: 65536, loss: 10.874 I0417 00:40:17.626486 140240525641536 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_02_synthetic/tensorboard/' I0417 00:40:18.068002 140240525641536 metric_logger.py:185] completed step: 1, seconds: 1.551, TFLOP/s/device: 8.759, Tokens/s/device: 1320.219, total_weights: 65536, loss: 10.874 I0417 00:40:18.208745 140240525641536 metric_logger.py:185] completed step: 2, seconds: 0.446, TFLOP/s/device: 30.491, Tokens/s/device: 4595.896, total_weights: 65536, loss: 10.262 I0417 00:40:18.356201 140240525641536 metric_logger.py:185] completed step: 3, seconds: 0.012, TFLOP/s/device: 1150.477, Tokens/s/device: 173412.362, total_weights: 65536, loss: 9.732 I0417 00:40:24.111480 140111660795648 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_02_synthetic/checkpoints/0/items I0417 00:40:24.240865 140108466472704 checkpoint.py:188] Wrote Metadata={'item_handlers': None, 'metrics': {}, 'performance_metrics': {}, 'init_timestamp_nsecs': 1776386423801162943, 'commit_timestamp_nsecs': None, 'custom_metadata': {}}, json={"item_handlers": null, "metrics": {}, "performance_metrics": {}, "init_timestamp_nsecs": 1776386423801162943, "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_02_synthetic/checkpoints/0/_CHECKPOINT_METADATA I0417 00:40:25.307354 2876 google_auth_provider.cc:181] Running on GCE, using service account 562977990677-compute@developer.gserviceaccount.com I0417 00:40:27.526619 140108967388928 array_metadata_store.py:203] [process=0][thread=array_type_handler] Wrote 69 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_02_synthetic/checkpoints/0/items/array_metadatas/process_0 I0417 00:40:41.052546 140240525641536 metric_logger.py:185] completed step: 4, seconds: 0.142, TFLOP/s/device: 95.354, Tokens/s/device: 14372.737, total_weights: 65536, loss: 9.274 I0417 00:40:41.062602 140240525641536 metric_logger.py:185] completed step: 5, seconds: 0.147, TFLOP/s/device: 92.701, Tokens/s/device: 13972.846, total_weights: 65536, loss: 8.888 I0417 00:40:41.207101 140240525641536 metric_logger.py:185] completed step: 6, seconds: 22.695, TFLOP/s/device: 0.599, Tokens/s/device: 90.240, total_weights: 65536, loss: 8.589 I0417 00:40:41.354499 140240525641536 metric_logger.py:185] completed step: 7, seconds: 0.008, TFLOP/s/device: 1768.697, Tokens/s/device: 266597.240, total_weights: 65536, loss: 8.380 I0417 00:40:41.501942 140240525641536 metric_logger.py:185] completed step: 8, seconds: 0.146, TFLOP/s/device: 92.968, Tokens/s/device: 14013.096, total_weights: 65536, loss: 8.251 I0417 00:40:41.650458 140240525641536 checkpointing.py:794] Waiting for step 9 to finish before checkpoint... I0417 00:40:41.651152 140240525641536 checkpointing.py:798] Waited 0.0007066726684570312 seconds for step 9 to finish before starting checkpointing. I0417 00:40:41.653155 140240525641536 checkpoint_manager.py:2020] [process=0][thread=MainThread][step=0][wait_until_finished] Waiting for Save Finalize thread (save_finalize) to complete. I0417 00:41:01.150033 140108958996224 base_pytree_checkpoint_handler.py:1282] [process=0][thread=write_metadata_after_commits] Commit + Array metadata written. Time taken: 35.840649s (commit=34.655587s, array_metadata_write=1.185061s) I0417 00:41:01.151189 140108476970752 base_pytree_checkpoint_handler.py:130] [process=0] /jax/orbax/write/gbytes_per_sec: 35.747 MiB/s (total gbytes: 1.5 GiB) (time elapsed: 44.18809795379639 s) (per-host) I0417 00:41:01.151312 140108476970752 async_checkpointer.py:90] [process=0][thread=async_save] 3 Handler Commit operations completed. Time taken: 43.904517s. I0417 00:41:02.303113 140108476970752 checkpoint.py:228] Read Metadata={'item_handlers': None, 'metrics': {}, 'performance_metrics': {}, 'init_timestamp_nsecs': 1776386423801162943, '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_02_synthetic/checkpoints/0/_CHECKPOINT_METADATA I0417 00:41:03.093369 140108466472704 checkpoint.py:247] Updated Metadata={'item_handlers': {'items': 'orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeCheckpointHandler'}, 'metrics': {}, 'performance_metrics': {}, 'init_timestamp_nsecs': 1776386423801162943, '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_02_synthetic/checkpoints/0/_CHECKPOINT_METADATA I0417 00:41:03.515828 140108476970752 array_metadata_store.py:411] [process=0][thread=async_save] Validated ArrayMetadata from all 8 hosts. Time taken: 0.000378s. I0417 00:41:03.702110 140108476970752 ocdbt_utils.py:49] Param validation support for Zarr3 will be added later (b/362328389). I0417 00:41:04.644230 140108476970752 base_pytree_checkpoint_handler.py:1406] [process=0][thread=async_save] Pytree save finalize (merge_ocdbt + ArrayMetadata validation) completed. Time taken: 2.184145s. 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_02_synthetic/checkpoints/0/items I0417 00:41:04.645960 140108476970752 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_02_synthetic/checkpoints/0/items I0417 00:41:05.170896 140108476970752 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_02_synthetic/checkpoints/0 I0417 00:41:06.561780 140108476970752 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_02_synthetic/checkpoints/0`. I0417 00:41:06.562578 140108476970752 event_tracking.py:138] [process=0] [async] Finished save (blocking + background) in 50.34 seconds @ gs://lance-maxtext/nnx_ckpt_xpk_feat_nnx_post_train_fixes_20260417_003418/nnx_xpk_feat_nnx_post_train_fixes_20260417_003418_02_synthetic/checkpoints/0 I0417 00:41:06.563796 140108476970752 async_checkpointer.py:160] [process=0][thread=async_save] Background save thread done. Time taken: 49.316999s. I0417 00:41:06.563924 140111685973760 async_checkpointer.py:288] [process=0][thread=save_finalize] Done with waiting for background save thread=async_save. I0417 00:41:06.563984 140111685973760 async_checkpointer.py:298] [process=0][thread=save_finalize] No errors found in background save thread=async_save. I0417 00:41:06.564056 140111685973760 checkpoint_manager.py:2137] [process=0][thread=save_finalize][step=0] CheckpointManager Save Finalize is syncing with other hosts... I0417 00:41:06.566804 140111685973760 checkpoint_manager.py:2146] [process=0][thread=save_finalize][step=0] CheckpointManager Save Finalize is done on all hosts. I0417 00:41:06.567731 140240525641536 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 00:41:06.567802 140240525641536 checkpoint_manager.py:1452] Waiting for previous save to complete took 24.914654 seconds. If this number is high, consider checkpointing less frequently. I0417 00:41:06.569472 140240525641536 checkpoint_manager.py:1512] [process=0] Saving checkpoint at step 9 I0417 00:41:06.571651 140240525641536 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_02_synthetic/checkpoints/9. I0417 00:41:06.934450 140240525641536 jax_array_handlers.py:360] Scheduling D2H of 69 prioritized jax.Array. I0417 00:41:06.934547 140240525641536 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 00:41:06.961683 140240525641536 base_pytree_checkpoint_handler.py:154] [process=0][thread=MainThread] Initiated "orbax.checkpoint._src.serialization.jax_array_handlers.ArrayHandler".serialize. Time taken: 0.028622s I0417 00:41:06.962524 140240525641536 base_pytree_checkpoint_handler.py:130] [process=0] /jax/orbax/write/blocking_gbytes_per_sec: 41.343 GiB/s (total gbytes: 1.5 GiB) (time elapsed: 0.03731131553649902 s) (per-host) I0417 00:41:06.962599 140240525641536 base_pytree_checkpoint_handler.py:768] [process=0][thread=MainThread] Initiated Pytree async_save. Time taken: 0.037405s (batch_requests_ready=0.002597s, total_serialization_initiated=0.034051s, others=0.000757s) I0417 00:41:06.962691 140240525641536 composite_checkpoint_handler.py:715] [process=0][thread=MainThread] Initiated CompositeCheckpointHandler.async_save. Time taken: 0.041736s (all_items=0.000018s, per_item={'items': '0.00001788'}, temp_paths=0.041718) I0417 00:41:06.963406 140240525641536 event_tracking.py:125] [process=0] [async] Finished blocking save in 0.39 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_02_synthetic/checkpoints/9. I0417 00:41:06.963693 140111685973760 async_checkpointer.py:76] [process=0][thread=async_save] Background save thread started. Deadline for this save operation is 2026-04-17 01:01:06.963660 I0417 00:41:06.976701 140240525641536 checkpoint_manager.py:1560] [process=0][thread=MainThread][step=9] Starting CheckpointManager Save Finalize thread=save_finalize I0417 00:41:06.976998 140111136147200 async_checkpointer.py:280] [process=0][thread=save_finalize] Waiting for background save thread=async_save. I0417 00:41:06.977161 140240525641536 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_02_synthetic/checkpoints', 'reached_preemption': False, 'preemption_received_at': None, 'synchronous': False, 'wait_for_prev_start_time': 1776386441.6531236, 'wait_for_prev_duration_secs': 24.91465425491333, 'time_between_consecutive_saves_sec': None, 'checkpointer_blocking_start_time': 1776386466.569509, 'checkpointer_blocking_duration_secs': 0.3943178653717041, 'get_old_steps_start_time': 1776386466.9638612, 'get_old_steps_duration_secs': 3.0517578125e-05, 'checkpoint_manager_blocking_start_time': 1776386441.6513767, 'checkpoint_manager_blocking_duration_secs': 25.32575011253357} I0417 00:41:06.977267 140240525641536 checkpointing.py:409] Started an asynchronous checkpoint save for step 9 I0417 00:41:06.977310 140240525641536 checkpoint_manager.py:2020] [process=0][thread=MainThread][step=9][wait_until_finished] Waiting for Save Finalize thread (save_finalize) to complete. I0417 00:41:07.587486 140108476970752 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_02_synthetic/checkpoints/9 I0417 00:41:09.328530 140087270569728 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_02_synthetic/checkpoints/9/items I0417 00:41:13.144969 140111656306432 array_metadata_store.py:203] [process=0][thread=array_type_handler] Wrote 69 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_02_synthetic/checkpoints/9/items/array_metadatas/process_0 I0417 00:41:52.317825 140111152932608 base_pytree_checkpoint_handler.py:1282] [process=0][thread=write_metadata_after_commits] Commit + Array metadata written. Time taken: 41.347456s (commit=40.110089s, array_metadata_write=1.237367s) I0417 00:41:52.318875 140111685973760 base_pytree_checkpoint_handler.py:130] [process=0] /jax/orbax/write/gbytes_per_sec: 34.798 MiB/s (total gbytes: 1.5 GiB) (time elapsed: 45.39361906051636 s) (per-host) I0417 00:41:52.319005 140111685973760 async_checkpointer.py:90] [process=0][thread=async_save] 3 Handler Commit operations completed. Time taken: 45.355201s. I0417 00:41:54.471728 140111685973760 array_metadata_store.py:411] [process=0][thread=async_save] Validated ArrayMetadata from all 8 hosts. Time taken: 0.000330s. I0417 00:41:54.713676 140111685973760 ocdbt_utils.py:49] Param validation support for Zarr3 will be added later (b/362328389). I0417 00:41:55.776915 140111685973760 base_pytree_checkpoint_handler.py:1406] [process=0][thread=async_save] Pytree save finalize (merge_ocdbt + ArrayMetadata validation) completed. Time taken: 2.197592s. 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_02_synthetic/checkpoints/9/items I0417 00:41:55.778666 140111685973760 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_02_synthetic/checkpoints/9/items I0417 00:41:56.310501 140111685973760 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_02_synthetic/checkpoints/9 I0417 00:41:57.705881 140111685973760 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_02_synthetic/checkpoints/9`. I0417 00:41:57.706712 140111685973760 event_tracking.py:138] [process=0] [async] Finished save (blocking + background) in 51.14 seconds @ gs://lance-maxtext/nnx_ckpt_xpk_feat_nnx_post_train_fixes_20260417_003418/nnx_xpk_feat_nnx_post_train_fixes_20260417_003418_02_synthetic/checkpoints/9 I0417 00:41:57.708499 140111685973760 async_checkpointer.py:160] [process=0][thread=async_save] Background save thread done. Time taken: 50.744680s. I0417 00:41:57.708758 140111136147200 async_checkpointer.py:288] [process=0][thread=save_finalize] Done with waiting for background save thread=async_save. I0417 00:41:57.708825 140111136147200 async_checkpointer.py:298] [process=0][thread=save_finalize] No errors found in background save thread=async_save. I0417 00:41:57.708896 140111136147200 checkpoint_manager.py:2137] [process=0][thread=save_finalize][step=9] CheckpointManager Save Finalize is syncing with other hosts... I0417 00:41:57.713741 140111136147200 checkpoint_manager.py:2146] [process=0][thread=save_finalize][step=9] CheckpointManager Save Finalize is done on all hosts. I0417 00:41:57.713876 140240525641536 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 00:41:57.713988 140240525641536 checkpoint_manager.py:2009] [process=0][thread=MainThread][wait_until_finished] No Save Finalize thread to wait for. Returning. I0417 00:41:57.714580 140240525641536 metric_logger.py:185] completed step: 9, seconds: 0.147, TFLOP/s/device: 92.267, Tokens/s/device: 13907.469, total_weights: 65536, loss: 8.175 Per train step: Total TFLOPs: 13.59 split as 93.93% learnable weight flops and 6.07% attention flops XPK End: Fri Apr 17 00:42:10 UTC 2026 EXIT_CODE=0