2026-04-16 18:59:18.441164: 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 18:59:18.558107 125605802794112 max_utils.py:238] Skipping jax distributed system due to skip_jax_distributed_system=True flag. I0416 19:00:13.233282 125605802794112 max_utils.py:800] System Information: Jax Version: 0.9.2 I0416 19:00:13.233388 125605802794112 max_utils.py:801] System Information: Jaxlib Version: 0.9.2 I0416 19:00:13.233425 125605802794112 max_utils.py:802] System Information: Jax Backend: PJRT C API TFRT TPU v6 lite Built on Apr 6 2026 20:48:10 (1775533690) cl/895581894 I0416 19:00:13.233450 125605802794112 train_utils.py:364] WARNING: 'dataset_path' might be pointing your local file system I0416 19:00:13.233473 125605802794112 train_utils.py:377] WARNING: Sequence packing is essentially ignored for synthetic data. Please use a real dataset to use sequence packing. I0416 19:00:13.233545 125605802794112 train.py:811] [DECOUPLED NO-OP] skipping cloud diagnostics wrapper. W0416 19:00:13.324559 4014336 pjrt_executable.cc:642] Assume version compatibility. PjRt-IFRT does not track XLA executable versions. I0416 19:00:13.726971 125605802794112 maxtext_utils.py:1687] Num_devices: 8, shape (1, 1, 1, 8, 1, 1, 1, 1, 1, 1, 1, 1, 1) I0416 19:00:13.727147 125605802794112 maxtext_utils.py:1687] Num_devices: 8, shape (1, 1, 1, 8, 1, 1, 1, 1, 1, 1, 1, 1, 1) I0416 19:00:13.727314 125605802794112 checkpointing.py:688] Setting up checkpoint logger... I0416 19:00:13.727352 125605802794112 checkpointing.py:234] Creating checkpoint manager with ocdbt=True and zarr3=True I0416 19:00:13.727393 125605802794112 pytree_checkpoint_handler.py:577] save_device_host_concurrent_bytes=None I0416 19:00:13.727860 125605802794112 base_pytree_checkpoint_handler.py:411] 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 0x723c4afdba70>, enable_pinned_host_transfer=False, save_concurrent_bytes: 96000000000 (89.4 GiB), restore_concurrent_bytes: 96000000000 (89.4 GiB) I0416 19:00:16.088903 125605802794112 checkpointing.py:266] Enabling policy for fixed interval checkpointing. I0416 19:00:16.089123 125605802794112 checkpoint_manager.py:702] [process=0][thread=MainThread] CheckpointManager init: checkpointers=None, item_names=('items',), item_handlers={'items': <orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeCheckpointHandler object at 0x7235f25ae840>}, handler_registry=None I0416 19:00:16.089382 125605802794112 composite_checkpoint_handler.py:237] Deferred registration for item: "items". Adding handler `<orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeCheckpointHandler object at 0x7235f25ae840>` 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 19:00:16.089420 125605802794112 composite_checkpoint_handler.py:237] Deferred registration for item: "metrics". Adding handler `<orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonCheckpointHandler object at 0x7235ce740d10>` 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 19:00:16.089448 125605802794112 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 0x7235f25ae840>, ('items', <class 'orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeRestoreArgs'>): <orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeCheckpointHandler object at 0x7235f25ae840>, ('metrics', <class 'orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonSaveArgs'>): <orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonCheckpointHandler object at 0x7235ce740d10>, ('metrics', <class 'orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonRestoreArgs'>): <orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonCheckpointHandler object at 0x7235ce740d10>}). I0416 19:00:16.089766 125605802794112 abstract_checkpointer.py:35] orbax-checkpoint version: 0.11.28 I0416 19:00:16.089816 125605802794112 async_checkpointer.py:177] [process=0][thread=MainThread] Using barrier_sync_fn: <function get_barrier_sync_fn.<locals>.<lambda> at 0x7235ce731ee0> timeout: 600 secs and primary_host=0 for async checkpoint writes I0416 19:00:16.224928 125605802794112 checkpoint_manager.py:1788] Found 0 checkpoint steps in gs://wanglance-maxtext/linen_ckpt_feat_nnx_post_train_fixes_20260416_181521/linen_feat_nnx_post_train_fixes_20260416_181521_13_scan_layers_true/checkpoints I0416 19:00:16.225163 125605802794112 checkpoint_manager.py:921] [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_hns=False, 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), root_directory=gs://wanglance-maxtext/linen_ckpt_feat_nnx_post_train_fixes_20260416_181521/linen_feat_nnx_post_train_fixes_20260416_181521_13_scan_layers_true/checkpoints: <orbax.checkpoint.checkpoint_manager.CheckpointManager object at 0x7235ce740f50> I0416 19:00:16.225244 125605802794112 checkpointing.py:302] Checkpoint manager created! W0416 19:00:16.241388 4014336 pjrt_executable.cc:642] Assume version compatibility. PjRt-IFRT does not track XLA executable versions. I0416 19:00:16.423686 125605802794112 nnx_wrappers.py:437] Unknown Logical: bfloat16[8,2048,2048]....................................... ('activation_batch', 'activation_norm_length', 'activation_embed'). I0416 19:00:16.423774 125605802794112 nnx_wrappers.py:437] Unknown Physical: bfloat16[8,2048,2048]....................................... ('fsdp', None, None). I0416 19:00:16.516444 125605802794112 attentions.py:1088] attentions/inputs_q Logical: bfloat16[8,2048,2048]....................................... ('activation_batch', 'activation_attn_length', 'activation_attn_embed'). I0416 19:00:16.516522 125605802794112 attentions.py:1088] attentions/inputs_q Physical: bfloat16[8,2048,2048]....................................... ('fsdp', None, None). I0416 19:00:16.530713 125605802794112 attentions.py:1089] attentions/inputs_kv Logical: bfloat16[8,2048,2048]....................................... ('activation_batch', 'activation_attn_length', 'activation_attn_embed'). I0416 19:00:16.530761 125605802794112 attentions.py:1089] attentions/inputs_kv Physical: bfloat16[8,2048,2048]....................................... ('fsdp', None, None). I0416 19:00:16.552000 125605802794112 attentions.py:1154] attentions/query Logical: bfloat16[8,2048,16,128]..................................... ('activation_kv_batch', 'activation_attn_length', 'activation_kv_heads', 'activation_kv_head_dim'). I0416 19:00:16.552065 125605802794112 attentions.py:1154] attentions/query Physical: bfloat16[8,2048,16,128]..................................... ('fsdp', None, None, None). I0416 19:00:16.566358 125605802794112 attentions.py:1155] attentions/key Logical: bfloat16[8,2048,16,128]..................................... ('activation_kv_batch', 'activation_attn_length', 'activation_kv_heads', 'activation_kv_head_dim'). I0416 19:00:16.566405 125605802794112 attentions.py:1155] attentions/key Physical: bfloat16[8,2048,16,128]..................................... ('fsdp', None, None, None). I0416 19:00:16.580713 125605802794112 attentions.py:1156] attentions/value Logical: bfloat16[8,2048,16,128]..................................... ('activation_kv_batch', 'activation_attn_length', 'activation_kv_heads', 'activation_kv_head_dim'). I0416 19:00:16.580763 125605802794112 attentions.py:1156] attentions/value Physical: bfloat16[8,2048,16,128]..................................... ('fsdp', None, None, None). I0416 19:00:16.602454 125605802794112 attentions.py:1197] attentions/out Logical: bfloat16[8,2048,16,128]..................................... ('activation_batch', 'activation_attn_length', 'activation_heads', 'activation_kv'). I0416 19:00:16.602512 125605802794112 attentions.py:1197] attentions/out Physical: bfloat16[8,2048,16,128]..................................... ('fsdp', None, None, None). I0416 19:00:16.620575 125605802794112 linears.py:525] linears/x Logical: bfloat16[8,2048,7168]....................................... ('activation_batch', 'activation_length', 'activation_mlp'). I0416 19:00:16.620624 125605802794112 linears.py:525] linears/x Physical: bfloat16[8,2048,7168]....................................... ('fsdp', None, None). I0416 19:00:16.815304 125605802794112 checkpointing.py:578] checkpoint manager exists so trying to load this run's existing checkpoint I0416 19:00:16.815400 125605802794112 checkpointing.py:676] No existing checkpoints found, not restoring checkpoint. W0416 19:00:16.864994 4014336 pjrt_executable.cc:642] Assume version compatibility. PjRt-IFRT does not track XLA executable versions. [DECOUPLED NO-OP] gcs_storage: using stubs. [DECOUPLED NO-OP] mldiagnostics: using stub. [DECOUPLED NO-OP] mldiagnostics: using stub. [DECOUPLED NO-OP] mldiagnostics: using stub. [DECOUPLED NO-OP] workload_monitor: using stub. [DECOUPLED NO-OP] vertex_tensorboard: using stub. fsdp: 8 I0416 19:00:17.248901 125605802794112 maxtext_utils.py:1796] params/params/decoder/decoder_norm/scale Shape: float32[2048] Logical: P('norm',) Physical: (None,) I0416 19:00:17.249015 125605802794112 maxtext_utils.py:1796] params/params/decoder/layers/mlp/wi_0/kernel Shape: float32[2048,16,7168] Logical: P('embed', 'layers', 'mlp') Physical: ('fsdp', None, None) I0416 19:00:17.249091 125605802794112 maxtext_utils.py:1796] params/params/decoder/layers/mlp/wi_1/kernel Shape: float32[2048,16,7168] Logical: P('embed', 'layers', 'mlp') Physical: ('fsdp', None, None) I0416 19:00:17.249138 125605802794112 maxtext_utils.py:1796] params/params/decoder/layers/mlp/wo/kernel Shape: float32[7168,16,2048] Logical: P('mlp', 'layers', 'embed') Physical: (None, None, 'fsdp') I0416 19:00:17.249176 125605802794112 maxtext_utils.py:1796] params/params/decoder/layers/post_self_attention_layer_norm/scale Shape: float32[2048,16] Logical: P('norm', 'layers') Physical: (None, None) I0416 19:00:17.249203 125605802794112 maxtext_utils.py:1796] params/params/decoder/layers/pre_self_attention_layer_norm/scale Shape: float32[2048,16] Logical: P('norm', 'layers') Physical: (None, None) I0416 19:00:17.249241 125605802794112 maxtext_utils.py:1796] params/params/decoder/layers/self_attention/key/kernel Shape: float32[2048,16,16,128] Logical: P('embed', 'layers', 'kv_heads', 'kv_head_dim') Physical: ('fsdp', None, None, None) I0416 19:00:17.249278 125605802794112 maxtext_utils.py:1796] params/params/decoder/layers/self_attention/out/kernel Shape: float32[16,16,128,2048] Logical: P('heads', 'layers', 'kv', 'embed') Physical: (None, None, None, 'fsdp') I0416 19:00:17.249305 125605802794112 maxtext_utils.py:1796] params/params/decoder/layers/self_attention/query/kernel Shape: float32[2048,16,16,128] Logical: P('embed', 'layers', 'q_heads', 'kv') Physical: ('fsdp', None, None, None) I0416 19:00:17.249330 125605802794112 maxtext_utils.py:1796] params/params/decoder/layers/self_attention/value/kernel Shape: float32[2048,16,16,128] Logical: P('embed', 'layers', 'kv_heads', 'kv_head_dim') Physical: ('fsdp', None, None, None) I0416 19:00:17.249362 125605802794112 maxtext_utils.py:1796] params/params/decoder/logits_dense/kernel Shape: float32[2048,32000] Logical: P('embed_vocab', 'vocab') Physical: ('fsdp', None) I0416 19:00:17.249394 125605802794112 maxtext_utils.py:1796] params/params/token_embedder/embedding Shape: float32[32000,2048] Logical: P('vocab', 'embed_vocab') Physical: (None, 'fsdp') I0416 19:00:17.673206 125605802794112 train.py:157] train/xent Logical: float32[8,2048]............................................. ('activation_embed_and_logits_batch', 'activation_length'). I0416 19:00:17.673352 125605802794112 train.py:157] train/xent Physical: float32[8,2048]............................................. ('fsdp', None). I0416 19:00:17.687246 125605802794112 train.py:164] train/z_loss Logical: float32[8,2048]............................................. ('activation_embed_and_logits_batch', 'activation_length'). I0416 19:00:17.687294 125605802794112 train.py:164] train/z_loss Physical: float32[8,2048]............................................. ('fsdp', None). W0416 19:00:18.356201 4014336 pjrt_executable.cc:642] Assume version compatibility. PjRt-IFRT does not track XLA executable versions. I0416 19:00:18.479341 125605802794112 max_utils.py:791] Total memory size: 3.6 GB, Output size: 1.5 GB, Temp size: 2.0 GB, Argument size: 1.5 GB, Host temp size: 0.0 GB. I0416 19:00:18.479812 125605802794112 max_utils.py:194] tensorboardX not available; using no-op SummaryWriter. I0416 19:00:18.479959 125605802794112 metric_logger.py:289] number parameters: 1.104 billion W0416 19:00:19.345013 4014336 pjrt_executable.cc:642] Assume version compatibility. PjRt-IFRT does not track XLA executable versions. I0416 19:00:19.443317 125605802794112 checkpointing.py:794] Waiting for step 0 to finish before checkpoint... I0416 19:00:19.560907 125605802794112 checkpointing.py:798] Waited 0.11757993698120117 seconds for step 0 to finish before starting checkpointing. I0416 19:00:19.561349 125605802794112 checkpoint_manager.py:1983] [process=0][thread=MainThread][wait_until_finished] No Save Finalize thread to wait for. Returning. I0416 19:00:19.561499 125605802794112 checkpoint_manager.py:1501] [process=0] Saving checkpoint at step 0 I0416 19:00:19.561879 125605802794112 async_checkpointer.py:452] [process=0] Started async saving checkpoint to gs://wanglance-maxtext/linen_ckpt_feat_nnx_post_train_fixes_20260416_181521/linen_feat_nnx_post_train_fixes_20260416_181521_13_scan_layers_true/checkpoints/0. I0416 19:00:19.650399 125605802794112 signaling_client.py:373] Using ThreadSafeKeyValueSignalingClient I0416 19:00:19.729880 125498258622016 atomicity.py:137] Creating tmp directory gs://wanglance-maxtext/linen_ckpt_feat_nnx_post_train_fixes_20260416_181521/linen_feat_nnx_post_train_fixes_20260416_181521_13_scan_layers_true/checkpoints/0 I0416 19:00:19.745546 125605802794112 jax_array_handlers.py:347] Scheduling D2H of 39 prioritized jax.Array. I0416 19:00:19.745630 125605802794112 replica_slices.py:410] 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 W0416 19:00:19.957332 4014336 pjrt_executable.cc:642] Assume version compatibility. PjRt-IFRT does not track XLA executable versions. W0416 19:00:19.968550 4014336 pjrt_executable.cc:642] Assume version compatibility. PjRt-IFRT does not track XLA executable versions. W0416 19:00:19.974529 4014336 pjrt_executable.cc:642] Assume version compatibility. PjRt-IFRT does not track XLA executable versions. W0416 19:00:19.984202 4014336 pjrt_executable.cc:642] Assume version compatibility. PjRt-IFRT does not track XLA executable versions. W0416 19:00:20.003082 4014336 pjrt_executable.cc:642] Assume version compatibility. PjRt-IFRT does not track XLA executable versions. W0416 19:00:20.013630 4014336 pjrt_executable.cc:642] Assume version compatibility. PjRt-IFRT does not track XLA executable versions. W0416 19:00:20.019828 4014336 pjrt_executable.cc:642] Assume version compatibility. PjRt-IFRT does not track XLA executable versions. W0416 19:00:20.024829 4014336 pjrt_executable.cc:642] Assume version compatibility. PjRt-IFRT does not track XLA executable versions. I0416 19:00:20.404907 125605802794112 base_pytree_checkpoint_handler.py:153] [process=0][thread=MainThread] Initiated "orbax.checkpoint._src.serialization.jax_array_handlers.ArrayHandler".serialize. Time taken: 0.659732s I0416 19:00:20.405335 125605802794112 base_pytree_checkpoint_handler.py:128] [process=0] /jax/checkpoint/write/blocking_gbytes_per_sec: 16.377 GiB/s (total gbytes: 12.3 GiB) (time elapsed: 753 milliseconds) (per-host) I0416 19:00:20.405385 125605802794112 base_pytree_checkpoint_handler.py:732] [process=0][thread=MainThread] Initiated Pytree async_save. Time taken: 0.753640s (batch_requests_ready=0.089803s, total_serialization_initiated=0.663500s, others=0.000338s) I0416 19:00:20.405442 125605802794112 composite_checkpoint_handler.py:715] [process=0][thread=MainThread] Initiated CompositeCheckpointHandler.async_save. Time taken: 0.754569s (all_items=0.000021s, per_item={'items': '0.00002098'}, temp_paths=0.754548) I0416 19:00:20.406182 125498195707456 async_checkpointer.py:79] [process=0][thread=async_save] Background save thread started. I0416 19:00:20.406274 125605802794112 async_checkpointer.py:561] Finished blocking save. Time taken: 0.844737s. Continuing background save to gs://wanglance-maxtext/linen_ckpt_feat_nnx_post_train_fixes_20260416_181521/linen_feat_nnx_post_train_fixes_20260416_181521_13_scan_layers_true/checkpoints/0. I0416 19:00:20.406448 125605802794112 checkpoint_manager.py:1549] [process=0][thread=MainThread][step=0] Starting CheckpointManager Save Finalize thread=save_finalize I0416 19:00:20.406686 125605802794112 standard_logger.py:34] {'step': 0, 'event_type': 'save', 'directory': 'gs://wanglance-maxtext/linen_ckpt_feat_nnx_post_train_fixes_20260416_181521/linen_feat_nnx_post_train_fixes_20260416_181521_13_scan_layers_true/checkpoints', 'reached_preemption': False, 'preemption_received_at': None, 'synchronous': False, 'wait_for_prev_start_time': 1776366019.5613368, 'wait_for_prev_duration_secs': 4.1484832763671875e-05, 'time_between_consecutive_saves_sec': None, 'checkpointer_blocking_start_time': 1776366019.5615194, 'checkpointer_blocking_duration_secs': 0.8448622226715088, 'get_old_steps_start_time': 1776366020.4063964, 'get_old_steps_duration_secs': 1.9788742065429688e-05, 'checkpoint_manager_blocking_start_time': 1776366019.5612528, 'checkpoint_manager_blocking_duration_secs': 0.8454113006591797} I0416 19:00:20.406752 125605802794112 checkpointing.py:409] Started an asynchronous checkpoint save for step 0 I0416 19:00:20.406789 125605802794112 max_utils.py:750] Memstats: After params initialized: I0416 19:00:20.406869 125605802794112 max_utils.py:756] Using (GB) 1.6 / 31.25 (5.120000%) on TPU_0(process=0,(0,0,0,0)) I0416 19:00:20.406896 125605802794112 max_utils.py:756] Using (GB) 1.59 / 31.25 (5.088000%) on TPU_1(process=0,(1,0,0,0)) I0416 19:00:20.406918 125605802794112 max_utils.py:756] Using (GB) 1.59 / 31.25 (5.088000%) on TPU_2(process=0,(0,1,0,0)) I0416 19:00:20.406936 125605802794112 max_utils.py:756] Using (GB) 1.59 / 31.25 (5.088000%) on TPU_3(process=0,(1,1,0,0)) I0416 19:00:20.406955 125605802794112 max_utils.py:756] Using (GB) 1.59 / 31.25 (5.088000%) on TPU_4(process=0,(0,2,0,0)) I0416 19:00:20.406975 125605802794112 max_utils.py:756] Using (GB) 1.59 / 31.25 (5.088000%) on TPU_5(process=0,(1,2,0,0)) I0416 19:00:20.406993 125605802794112 max_utils.py:756] Using (GB) 1.59 / 31.25 (5.088000%) on TPU_6(process=0,(0,3,0,0)) I0416 19:00:20.407010 125605802794112 max_utils.py:756] Using (GB) 1.59 / 31.25 (5.088000%) on TPU_7(process=0,(1,3,0,0)) I0416 19:00:20.412509 125498248136256 atomicity.py:137] Creating tmp directory gs://wanglance-maxtext/linen_ckpt_feat_nnx_post_train_fixes_20260416_181521/linen_feat_nnx_post_train_fixes_20260416_181521_13_scan_layers_true/checkpoints/0/items W0416 19:00:20.412884 4014336 pjrt_executable.cc:642] Assume version compatibility. PjRt-IFRT does not track XLA executable versions. W0416 19:00:20.418346 4014336 pjrt_executable.cc:642] Assume version compatibility. PjRt-IFRT does not track XLA executable versions. W0416 19:00:20.423132 4014336 pjrt_executable.cc:642] Assume version compatibility. PjRt-IFRT does not track XLA executable versions. I0416 19:00:20.545468 125498216678976 checkpoint.py:188] Wrote Metadata={'item_handlers': None, 'metrics': {}, 'performance_metrics': {}, 'init_timestamp_nsecs': 1776366020314011475, 'commit_timestamp_nsecs': None, 'custom_metadata': {}}, json={"item_handlers": null, "metrics": {}, "performance_metrics": {}, "init_timestamp_nsecs": 1776366020314011475, "commit_timestamp_nsecs": null, "custom_metadata": {}} to gs://wanglance-maxtext/linen_ckpt_feat_nnx_post_train_fixes_20260416_181521/linen_feat_nnx_post_train_fixes_20260416_181521_13_scan_layers_true/checkpoints/0/_CHECKPOINT_METADATA I0416 19:00:20.545681 125498269107776 async_checkpointer.py:265] [process=0][thread=save_finalize] Waiting for background save thread=async_save. I0416 19:00:20.744922 125605802794112 metric_logger.py:185] completed step: 0, seconds: 0.963, TFLOP/s/device: 14.111, Tokens/s/device: 2126.994, total_weights: 16384, loss: 10.874 I0416 19:00:20.745922 125605802794112 metric_logger.py:269] To see full metrics 'tensorboard --logdir=gs://wanglance-maxtext/linen_ckpt_feat_nnx_post_train_fixes_20260416_181521/linen_feat_nnx_post_train_fixes_20260416_181521_13_scan_layers_true/tensorboard/' I0416 19:00:20.860807 125605802794112 metric_logger.py:185] completed step: 1, seconds: 1.300, TFLOP/s/device: 10.454, Tokens/s/device: 1575.723, total_weights: 16384, loss: 10.874 I0416 19:00:20.942173 4016525 google_auth_provider.cc:149] Using credentials at ~/.config/gcloud/application_default_credentials.json I0416 19:00:20.942229 4016525 google_auth_provider.cc:156] Using OAuth2 AuthProvider I0416 19:00:20.979795 125605802794112 metric_logger.py:185] completed step: 2, seconds: 0.007, TFLOP/s/device: 2055.852, Tokens/s/device: 309880.466, total_weights: 16384, loss: 9.775 I0416 19:00:21.098019 125605802794112 metric_logger.py:185] completed step: 3, seconds: 0.117, TFLOP/s/device: 116.514, Tokens/s/device: 17562.214, total_weights: 16384, loss: 8.838 I0416 19:00:21.680459 125498237650496 array_metadata_store.py:203] [process=0][thread=array_type_handler] Wrote 39 array_metadata.ArrayMetadata to gs://wanglance-maxtext/linen_ckpt_feat_nnx_post_train_fixes_20260416_181521/linen_feat_nnx_post_train_fixes_20260416_181521_13_scan_layers_true/checkpoints/0/items/array_metadatas/process_0 I0416 19:00:40.992903 125605802794112 metric_logger.py:185] completed step: 4, seconds: 0.119, TFLOP/s/device: 114.174, Tokens/s/device: 17209.506, total_weights: 16384, loss: 7.965 I0416 19:00:41.000693 125605802794112 metric_logger.py:185] completed step: 5, seconds: 0.118, TFLOP/s/device: 115.017, Tokens/s/device: 17336.686, total_weights: 16384, loss: 7.233 I0416 19:00:41.117525 125605802794112 metric_logger.py:185] completed step: 6, seconds: 19.896, TFLOP/s/device: 0.683, Tokens/s/device: 102.934, total_weights: 16384, loss: 6.676 I0416 19:00:41.236461 125605802794112 metric_logger.py:185] completed step: 7, seconds: 0.005, TFLOP/s/device: 2834.195, Tokens/s/device: 427200.668, total_weights: 16384, loss: 6.293 I0416 19:00:41.354820 125605802794112 metric_logger.py:185] completed step: 8, seconds: 0.118, TFLOP/s/device: 115.618, Tokens/s/device: 17427.266, total_weights: 16384, loss: 6.055 I0416 19:00:41.472758 125605802794112 checkpointing.py:794] Waiting for step 9 to finish before checkpoint... I0416 19:00:41.473605 125605802794112 checkpointing.py:798] Waited 0.0008776187896728516 seconds for step 9 to finish before starting checkpointing. I0416 19:00:41.473870 125605802794112 checkpoint_manager.py:1994] [process=0][thread=MainThread][step=0][wait_until_finished] Waiting for Save Finalize thread (save_finalize) to complete. I0416 19:00:58.078485 125498206193216 base_pytree_checkpoint_handler.py:1217] [process=0][thread=write_metadata_after_commits] Commit + Array metadata written. Time taken: 37.133143s (commit=36.711715s, array_metadata_write=0.421427s) I0416 19:00:58.079634 125498195707456 base_pytree_checkpoint_handler.py:128] [process=0] /jax/checkpoint/write/gbytes_per_sec: 328.845 MiB/s (total gbytes: 12.3 GiB) (time elapsed: 38 seconds) (per-host) I0416 19:00:58.079737 125498195707456 async_checkpointer.py:90] [process=0][thread=async_save] 3 Handler Commit operations completed. Time taken: 37.673418s. I0416 19:00:58.288162 125498195707456 checkpoint.py:228] Read Metadata={'item_handlers': None, 'metrics': {}, 'performance_metrics': {}, 'init_timestamp_nsecs': 1776366020314011475, 'commit_timestamp_nsecs': None, 'custom_metadata': {}} from gs://wanglance-maxtext/linen_ckpt_feat_nnx_post_train_fixes_20260416_181521/linen_feat_nnx_post_train_fixes_20260416_181521_13_scan_layers_true/checkpoints/0/_CHECKPOINT_METADATA I0416 19:00:58.457880 125498195707456 array_metadata_store.py:367] [process=0][thread=async_save] Skipped cross-host ArrayMetadata validation because only one process is found: process_index=0. I0416 19:00:58.655524 125498216678976 checkpoint.py:247] Updated Metadata={'item_handlers': {'items': 'orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeCheckpointHandler'}, 'metrics': {}, 'performance_metrics': {}, 'init_timestamp_nsecs': 1776366020314011475, 'commit_timestamp_nsecs': None, 'custom_metadata': {}} to gs://wanglance-maxtext/linen_ckpt_feat_nnx_post_train_fixes_20260416_181521/linen_feat_nnx_post_train_fixes_20260416_181521_13_scan_layers_true/checkpoints/0/_CHECKPOINT_METADATA I0416 19:00:58.891116 125498195707456 ocdbt_utils.py:56] Param validation support for Zarr3 will be added later (b/362328389). I0416 19:00:58.891296 125498195707456 base_pytree_checkpoint_handler.py:1342] [process=0][thread=async_save] Pytree save finalize (merge_ocdbt + ArrayMetadata validation) completed. Time taken: 0.562755s. use_zarr3=True, enable_post_merge_validation=True, directory=gs://wanglance-maxtext/linen_ckpt_feat_nnx_post_train_fixes_20260416_181521/linen_feat_nnx_post_train_fixes_20260416_181521_13_scan_layers_true/checkpoints/0/items I0416 19:00:58.892661 125498195707456 atomicity.py:608] Finalizing gs://wanglance-maxtext/linen_ckpt_feat_nnx_post_train_fixes_20260416_181521/linen_feat_nnx_post_train_fixes_20260416_181521_13_scan_layers_true/checkpoints/0/items I0416 19:00:59.140153 125498195707456 atomicity.py:608] Finalizing gs://wanglance-maxtext/linen_ckpt_feat_nnx_post_train_fixes_20260416_181521/linen_feat_nnx_post_train_fixes_20260416_181521_13_scan_layers_true/checkpoints/0 I0416 19:00:59.854319 125498195707456 atomicity.py:794] [process=0][thread=async_save] Finished saving checkpoint (finalized tmp dir) to `gs://wanglance-maxtext/linen_ckpt_feat_nnx_post_train_fixes_20260416_181521/linen_feat_nnx_post_train_fixes_20260416_181521_13_scan_layers_true/checkpoints/0`. I0416 19:00:59.854953 125498195707456 async_checkpointer.py:420] Finished async_save (blocking + background). Time taken: 40.293422s. directory=gs://wanglance-maxtext/linen_ckpt_feat_nnx_post_train_fixes_20260416_181521/linen_feat_nnx_post_train_fixes_20260416_181521_13_scan_layers_true/checkpoints/0 I0416 19:00:59.855030 125498195707456 async_checkpointer.py:144] [process=0][thread=async_save] Background save thread done. Time taken: 39.448713s. I0416 19:00:59.855198 125498269107776 async_checkpointer.py:273] [process=0][thread=save_finalize] Done with waiting for background save thread=async_save. I0416 19:00:59.855300 125498269107776 async_checkpointer.py:283] [process=0][thread=save_finalize] No errors found in background save thread=async_save. I0416 19:00:59.855356 125498269107776 checkpoint_manager.py:2103] [process=0][thread=save_finalize][step=0] CheckpointManager Save Finalize is syncing with other hosts... I0416 19:00:59.855401 125498269107776 checkpoint_manager.py:2112] [process=0][thread=save_finalize][step=0] CheckpointManager Save Finalize is done on all hosts. I0416 19:00:59.855610 125605802794112 checkpoint_manager.py:2006] [process=0][thread=MainThread][step=0][wait_until_finished] Done waiting for Save Finalize thread (save_finalize) running at step=0. W0416 19:00:59.855680 125605802794112 checkpoint_manager.py:1441] Waiting for previous save to complete took 18.381829 seconds. If this number is high, consider checkpointing less frequently. I0416 19:00:59.856466 125605802794112 checkpoint_manager.py:1501] [process=0] Saving checkpoint at step 9 I0416 19:00:59.856697 125605802794112 async_checkpointer.py:452] [process=0] Started async saving checkpoint to gs://wanglance-maxtext/linen_ckpt_feat_nnx_post_train_fixes_20260416_181521/linen_feat_nnx_post_train_fixes_20260416_181521_13_scan_layers_true/checkpoints/9. I0416 19:01:00.017423 125605802794112 jax_array_handlers.py:347] Scheduling D2H of 39 prioritized jax.Array. I0416 19:01:00.017513 125605802794112 replica_slices.py:410] 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 19:01:00.033524 125498269107776 atomicity.py:137] Creating tmp directory gs://wanglance-maxtext/linen_ckpt_feat_nnx_post_train_fixes_20260416_181521/linen_feat_nnx_post_train_fixes_20260416_181521_13_scan_layers_true/checkpoints/9 I0416 19:01:00.772343 125498206193216 atomicity.py:137] Creating tmp directory gs://wanglance-maxtext/linen_ckpt_feat_nnx_post_train_fixes_20260416_181521/linen_feat_nnx_post_train_fixes_20260416_181521_13_scan_layers_true/checkpoints/9/items I0416 19:01:09.393828 125605802794112 base_pytree_checkpoint_handler.py:153] [process=0][thread=MainThread] Initiated "orbax.checkpoint._src.serialization.jax_array_handlers.ArrayHandler".serialize. Time taken: 9.376789s I0416 19:01:09.398484 125605802794112 base_pytree_checkpoint_handler.py:128] [process=0] /jax/checkpoint/write/blocking_gbytes_per_sec: 1.304 GiB/s (total gbytes: 12.3 GiB) (time elapsed: 9 seconds) (per-host) I0416 19:01:09.398585 125605802794112 base_pytree_checkpoint_handler.py:732] [process=0][thread=MainThread] Initiated Pytree async_save. Time taken: 9.467038s (batch_requests_ready=0.082284s, total_serialization_initiated=9.380330s, others=0.004424s) I0416 19:01:09.398664 125605802794112 composite_checkpoint_handler.py:715] [process=0][thread=MainThread] Initiated CompositeCheckpointHandler.async_save. Time taken: 9.467799s (all_items=0.000018s, per_item={'items': '0.00001812'}, temp_paths=9.467781) I0416 19:01:09.400292 125498166347328 async_checkpointer.py:79] [process=0][thread=async_save] Background save thread started. I0416 19:01:09.400376 125605802794112 async_checkpointer.py:561] Finished blocking save. Time taken: 9.543863s. Continuing background save to gs://wanglance-maxtext/linen_ckpt_feat_nnx_post_train_fixes_20260416_181521/linen_feat_nnx_post_train_fixes_20260416_181521_13_scan_layers_true/checkpoints/9. I0416 19:01:09.400573 125605802794112 checkpoint_manager.py:1549] [process=0][thread=MainThread][step=9] Starting CheckpointManager Save Finalize thread=save_finalize I0416 19:01:09.400750 125498195707456 async_checkpointer.py:265] [process=0][thread=save_finalize] Waiting for background save thread=async_save. I0416 19:01:09.400850 125605802794112 standard_logger.py:34] {'step': 9, 'event_type': 'save', 'directory': 'gs://wanglance-maxtext/linen_ckpt_feat_nnx_post_train_fixes_20260416_181521/linen_feat_nnx_post_train_fixes_20260416_181521_13_scan_layers_true/checkpoints', 'reached_preemption': False, 'preemption_received_at': None, 'synchronous': False, 'wait_for_prev_start_time': 1776366041.4738302, 'wait_for_prev_duration_secs': 18.381829023361206, 'time_between_consecutive_saves_sec': None, 'checkpointer_blocking_start_time': 1776366059.8564916, 'checkpointer_blocking_duration_secs': 9.543994665145874, 'get_old_steps_start_time': 1776366069.4005034, 'get_old_steps_duration_secs': 2.86102294921875e-05, 'checkpoint_manager_blocking_start_time': 1776366041.4737856, 'checkpoint_manager_blocking_duration_secs': 27.92703628540039} I0416 19:01:09.400971 125605802794112 checkpointing.py:409] Started an asynchronous checkpoint save for step 9 I0416 19:01:09.401005 125605802794112 checkpoint_manager.py:1994] [process=0][thread=MainThread][step=9][wait_until_finished] Waiting for Save Finalize thread (save_finalize) to complete. I0416 19:01:10.140113 125497004525120 array_metadata_store.py:203] [process=0][thread=array_type_handler] Wrote 39 array_metadata.ArrayMetadata to gs://wanglance-maxtext/linen_ckpt_feat_nnx_post_train_fixes_20260416_181521/linen_feat_nnx_post_train_fixes_20260416_181521_13_scan_layers_true/checkpoints/9/items/array_metadatas/process_0 I0416 19:01:46.159156 125498237650496 base_pytree_checkpoint_handler.py:1217] [process=0][thread=write_metadata_after_commits] Commit + Array metadata written. Time taken: 36.759816s (commit=36.348705s, array_metadata_write=0.411111s) I0416 19:01:46.160182 125498166347328 base_pytree_checkpoint_handler.py:128] [process=0] /jax/checkpoint/write/gbytes_per_sec: 273.354 MiB/s (total gbytes: 12.3 GiB) (time elapsed: 46 seconds) (per-host) I0416 19:01:46.160243 125498166347328 async_checkpointer.py:90] [process=0][thread=async_save] 3 Handler Commit operations completed. Time taken: 36.759819s. I0416 19:01:46.591258 125498166347328 array_metadata_store.py:367] [process=0][thread=async_save] Skipped cross-host ArrayMetadata validation because only one process is found: process_index=0. I0416 19:01:47.006523 125498166347328 ocdbt_utils.py:56] Param validation support for Zarr3 will be added later (b/362328389). I0416 19:01:47.007183 125498166347328 base_pytree_checkpoint_handler.py:1342] [process=0][thread=async_save] Pytree save finalize (merge_ocdbt + ArrayMetadata validation) completed. Time taken: 0.549705s. use_zarr3=True, enable_post_merge_validation=True, directory=gs://wanglance-maxtext/linen_ckpt_feat_nnx_post_train_fixes_20260416_181521/linen_feat_nnx_post_train_fixes_20260416_181521_13_scan_layers_true/checkpoints/9/items I0416 19:01:47.007810 125498166347328 atomicity.py:608] Finalizing gs://wanglance-maxtext/linen_ckpt_feat_nnx_post_train_fixes_20260416_181521/linen_feat_nnx_post_train_fixes_20260416_181521_13_scan_layers_true/checkpoints/9/items I0416 19:01:47.254674 125498166347328 atomicity.py:608] Finalizing gs://wanglance-maxtext/linen_ckpt_feat_nnx_post_train_fixes_20260416_181521/linen_feat_nnx_post_train_fixes_20260416_181521_13_scan_layers_true/checkpoints/9 I0416 19:01:47.928239 125498166347328 atomicity.py:794] [process=0][thread=async_save] Finished saving checkpoint (finalized tmp dir) to `gs://wanglance-maxtext/linen_ckpt_feat_nnx_post_train_fixes_20260416_181521/linen_feat_nnx_post_train_fixes_20260416_181521_13_scan_layers_true/checkpoints/9`. I0416 19:01:47.928872 125498166347328 async_checkpointer.py:420] Finished async_save (blocking + background). Time taken: 48.072364s. directory=gs://wanglance-maxtext/linen_ckpt_feat_nnx_post_train_fixes_20260416_181521/linen_feat_nnx_post_train_fixes_20260416_181521_13_scan_layers_true/checkpoints/9 I0416 19:01:47.928956 125498166347328 async_checkpointer.py:144] [process=0][thread=async_save] Background save thread done. Time taken: 38.528535s. I0416 19:01:47.929144 125498195707456 async_checkpointer.py:273] [process=0][thread=save_finalize] Done with waiting for background save thread=async_save. I0416 19:01:47.929255 125498195707456 async_checkpointer.py:283] [process=0][thread=save_finalize] No errors found in background save thread=async_save. I0416 19:01:47.929341 125498195707456 checkpoint_manager.py:2103] [process=0][thread=save_finalize][step=9] CheckpointManager Save Finalize is syncing with other hosts... I0416 19:01:47.929387 125498195707456 checkpoint_manager.py:2112] [process=0][thread=save_finalize][step=9] CheckpointManager Save Finalize is done on all hosts. I0416 19:01:47.929692 125605802794112 checkpoint_manager.py:2006] [process=0][thread=MainThread][step=9][wait_until_finished] Done waiting for Save Finalize thread (save_finalize) running at step=9. I0416 19:01:47.929822 125605802794112 checkpoint_manager.py:1983] [process=0][thread=MainThread][wait_until_finished] No Save Finalize thread to wait for. Returning. I0416 19:01:47.930314 125605802794112 metric_logger.py:185] completed step: 9, seconds: 0.119, TFLOP/s/device: 114.068, Tokens/s/device: 17193.613, total_weights: 16384, loss: 5.915 Per train step: Total TFLOPs: 13.59 split as 93.93% learnable weight flops and 6.07% attention flops