2026-04-16 19:42:20.145381: 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 19:42:20.260614 135003411475584 max_utils.py:238] Skipping jax distributed system due to skip_jax_distributed_system=True flag. I0416 19:43:18.226454 135003411475584 max_utils.py:800] System Information: Jax Version: 0.9.2 I0416 19:43:18.226564 135003411475584 max_utils.py:801] System Information: Jaxlib Version: 0.9.2 I0416 19:43:18.226602 135003411475584 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 19:43:18.226627 135003411475584 train_utils.py:364] WARNING: 'dataset_path' might be pointing your local file system I0416 19:43:18.226649 135003411475584 train_utils.py:377] WARNING: Sequence packing is essentially ignored for synthetic data. Please use a real dataset to use sequence packing. I0416 19:43:18.226720 135003411475584 train.py:811] [DECOUPLED NO-OP] skipping cloud diagnostics wrapper. W0416 19:43:18.319468 4049865 pjrt_executable.cc:642] Assume version compatibility. PjRt-IFRT does not track XLA executable versions. I0416 19:43:18.717715 135003411475584 maxtext_utils.py:1687] Num_devices: 8, shape (1, 1, 1, 8, 1, 1, 1, 1, 1, 1, 1, 1, 1) I0416 19:43:18.814275 135003411475584 checkpointing.py:688] Setting up checkpoint logger... I0416 19:43:18.814374 135003411475584 checkpointing.py:234] Creating checkpoint manager with ocdbt=True and zarr3=True I0416 19:43:18.814410 135003411475584 pytree_checkpoint_handler.py:577] save_device_host_concurrent_bytes=None I0416 19:43:18.814579 135003411475584 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 0x7ac8577f5010>, enable_pinned_host_transfer=False, save_concurrent_bytes: 96000000000 (89.4 GiB), restore_concurrent_bytes: 96000000000 (89.4 GiB) I0416 19:43:21.184893 135003411475584 checkpointing.py:266] Enabling policy for fixed interval checkpointing. I0416 19:43:21.185083 135003411475584 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 0x7ac1da910e00>}, handler_registry=None I0416 19:43:21.185331 135003411475584 composite_checkpoint_handler.py:237] Deferred registration for item: "items". Adding handler `<orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeCheckpointHandler object at 0x7ac1da910e00>` 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 19:43:21.185371 135003411475584 composite_checkpoint_handler.py:237] Deferred registration for item: "metrics". Adding handler `<orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonCheckpointHandler object at 0x7ac1da93ddf0>` 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 19:43:21.185400 135003411475584 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 0x7ac1da910e00>, ('items', <class 'orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeRestoreArgs'>): <orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeCheckpointHandler object at 0x7ac1da910e00>, ('metrics', <class 'orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonSaveArgs'>): <orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonCheckpointHandler object at 0x7ac1da93ddf0>, ('metrics', <class 'orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonRestoreArgs'>): <orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonCheckpointHandler object at 0x7ac1da93ddf0>}). I0416 19:43:21.185858 135003411475584 abstract_checkpointer.py:35] orbax-checkpoint version: 0.11.28 I0416 19:43:21.185908 135003411475584 async_checkpointer.py:177] [process=0][thread=MainThread] Using barrier_sync_fn: <function get_barrier_sync_fn.<locals>.<lambda> at 0x7ac1db575760> timeout: 600 secs and primary_host=0 for async checkpoint writes I0416 19:43:21.302819 135003411475584 checkpoint_manager.py:1788] Found 0 checkpoint steps in gs://wanglance-maxtext/nnx_ckpt_feat_nnx_post_train_fixes_20260416_181521/nnx_feat_nnx_post_train_fixes_20260416_181521_02_synthetic/checkpoints I0416 19:43:21.303037 135003411475584 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/nnx_ckpt_feat_nnx_post_train_fixes_20260416_181521/nnx_feat_nnx_post_train_fixes_20260416_181521_02_synthetic/checkpoints: <orbax.checkpoint.checkpoint_manager.CheckpointManager object at 0x7ac1da93ecf0> I0416 19:43:21.307074 135003411475584 checkpointing.py:302] Checkpoint manager created! W0416 19:43:21.326849 4049865 pjrt_executable.cc:642] Assume version compatibility. PjRt-IFRT does not track XLA executable versions. I0416 19:43:21.537079 135003411475584 checkpointing.py:578] checkpoint manager exists so trying to load this run's existing checkpoint I0416 19:43:21.537220 135003411475584 checkpointing.py:676] No existing checkpoints found, not restoring checkpoint. W0416 19:43:21.646402 4049865 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 19:43:21.732716 135003411475584 maxtext_utils.py:1805] decoder/decoder_norm/scale/value Shape: float32[2048] Physical: (None,) I0416 19:43:21.732813 135003411475584 maxtext_utils.py:1805] decoder/layers/mlp/wi_0/kernel/value Shape: float32[2048,16,7168] Physical: ('fsdp', None, None) I0416 19:43:21.732853 135003411475584 maxtext_utils.py:1805] decoder/layers/mlp/wi_1/kernel/value Shape: float32[2048,16,7168] Physical: ('fsdp', None, None) I0416 19:43:21.732890 135003411475584 maxtext_utils.py:1805] decoder/layers/mlp/wo/kernel/value Shape: float32[7168,16,2048] Physical: (None, None, 'fsdp') I0416 19:43:21.732918 135003411475584 maxtext_utils.py:1805] decoder/layers/post_self_attention_layer_norm/scale/value Shape: float32[2048,16] Physical: (None, None) I0416 19:43:21.732943 135003411475584 maxtext_utils.py:1805] decoder/layers/pre_self_attention_layer_norm/scale/value Shape: float32[2048,16] Physical: (None, None) I0416 19:43:21.732969 135003411475584 maxtext_utils.py:1805] decoder/layers/self_attention/key/kernel/value Shape: float32[2048,16,16,128] Physical: ('fsdp', None, None, None) I0416 19:43:21.732994 135003411475584 maxtext_utils.py:1805] decoder/layers/self_attention/out/kernel/value Shape: float32[16,16,128,2048] Physical: (None, None, None, 'fsdp') I0416 19:43:21.733017 135003411475584 maxtext_utils.py:1805] decoder/layers/self_attention/query/kernel/value Shape: float32[2048,16,16,128] Physical: ('fsdp', None, None, None) I0416 19:43:21.733041 135003411475584 maxtext_utils.py:1805] decoder/layers/self_attention/value/kernel/value Shape: float32[2048,16,16,128] Physical: ('fsdp', None, None, None) I0416 19:43:21.733080 135003411475584 maxtext_utils.py:1805] decoder/logits_dense/kernel/value Shape: float32[2048,32000] Physical: ('fsdp', None) I0416 19:43:21.733103 135003411475584 maxtext_utils.py:1805] token_embedder/embedding/value Shape: float32[32000,2048] Physical: (None, 'fsdp') I0416 19:43:21.838748 135003411475584 nnx_decoders.py:465] nnx_decoders/carry Logical: bfloat16[8,2048,2048]....................................... ('activation_batch', 'activation_norm_length', 'activation_embed'). I0416 19:43:21.838825 135003411475584 nnx_decoders.py:465] nnx_decoders/carry Physical: bfloat16[8,2048,2048]....................................... ('fsdp', None, None). I0416 19:43:21.844097 135003411475584 nnx_decoders.py:465] Unknown Logical: bfloat16[8,2048,2048]....................................... ('activation_batch', 'activation_norm_length', 'activation_embed'). I0416 19:43:21.844144 135003411475584 nnx_decoders.py:465] Unknown Physical: bfloat16[8,2048,2048]....................................... ('fsdp', None, None). I0416 19:43:21.858712 135003411475584 attentions.py:1088] attentions/inputs_q Logical: bfloat16[8,2048,2048]....................................... ('activation_batch', 'activation_attn_length', 'activation_attn_embed'). I0416 19:43:21.858758 135003411475584 attentions.py:1088] attentions/inputs_q Physical: bfloat16[8,2048,2048]....................................... ('fsdp', None, None). I0416 19:43:21.872377 135003411475584 attentions.py:1089] attentions/inputs_kv Logical: bfloat16[8,2048,2048]....................................... ('activation_batch', 'activation_attn_length', 'activation_attn_embed'). I0416 19:43:21.872424 135003411475584 attentions.py:1089] attentions/inputs_kv Physical: bfloat16[8,2048,2048]....................................... ('fsdp', None, None). I0416 19:43:21.893495 135003411475584 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 19:43:21.893550 135003411475584 attentions.py:1154] attentions/query Physical: bfloat16[8,2048,16,128]..................................... ('fsdp', None, None, None). I0416 19:43:21.907344 135003411475584 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 19:43:21.907393 135003411475584 attentions.py:1155] attentions/key Physical: bfloat16[8,2048,16,128]..................................... ('fsdp', None, None, None). I0416 19:43:21.921104 135003411475584 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 19:43:21.921152 135003411475584 attentions.py:1156] attentions/value Physical: bfloat16[8,2048,16,128]..................................... ('fsdp', None, None, None). I0416 19:43:21.947688 135003411475584 attentions.py:1197] attentions/out Logical: bfloat16[8,2048,16,128]..................................... ('activation_batch', 'activation_attn_length', 'activation_heads', 'activation_kv'). I0416 19:43:21.947745 135003411475584 attentions.py:1197] attentions/out Physical: bfloat16[8,2048,16,128]..................................... ('fsdp', None, None, None). I0416 19:43:21.966034 135003411475584 linears.py:525] linears/x Logical: bfloat16[8,2048,7168]....................................... ('activation_batch', 'activation_length', 'activation_mlp'). I0416 19:43:21.966098 135003411475584 linears.py:525] linears/x Physical: bfloat16[8,2048,7168]....................................... ('fsdp', None, None). W0416 19:43:22.714980 4049865 pjrt_executable.cc:642] Assume version compatibility. PjRt-IFRT does not track XLA executable versions. I0416 19:43:22.832742 135003411475584 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 19:43:22.833298 135003411475584 max_utils.py:194] tensorboardX not available; using no-op SummaryWriter. I0416 19:43:22.834624 135003411475584 metric_logger.py:289] number parameters: 1.104 billion W0416 19:43:23.455939 4049865 pjrt_executable.cc:642] Assume version compatibility. PjRt-IFRT does not track XLA executable versions. I0416 19:43:23.557288 135003411475584 checkpointing.py:794] Waiting for step 0 to finish before checkpoint... I0416 19:43:23.673627 135003411475584 checkpointing.py:798] Waited 0.11632871627807617 seconds for step 0 to finish before starting checkpointing. I0416 19:43:23.674062 135003411475584 checkpoint_manager.py:1983] [process=0][thread=MainThread][wait_until_finished] No Save Finalize thread to wait for. Returning. I0416 19:43:23.674207 135003411475584 checkpoint_manager.py:1501] [process=0] Saving checkpoint at step 0 I0416 19:43:23.674562 135003411475584 async_checkpointer.py:452] [process=0] Started async saving checkpoint to gs://wanglance-maxtext/nnx_ckpt_feat_nnx_post_train_fixes_20260416_181521/nnx_feat_nnx_post_train_fixes_20260416_181521_02_synthetic/checkpoints/0. I0416 19:43:23.761617 135003411475584 signaling_client.py:373] Using ThreadSafeKeyValueSignalingClient I0416 19:43:23.847728 134895984707136 atomicity.py:137] Creating tmp directory gs://wanglance-maxtext/nnx_ckpt_feat_nnx_post_train_fixes_20260416_181521/nnx_feat_nnx_post_train_fixes_20260416_181521_02_synthetic/checkpoints/0 I0416 19:43:23.868002 135003411475584 jax_array_handlers.py:347] Scheduling D2H of 69 prioritized jax.Array. I0416 19:43:23.868105 135003411475584 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 19:43:24.508064 134895974221376 atomicity.py:137] Creating tmp directory gs://wanglance-maxtext/nnx_ckpt_feat_nnx_post_train_fixes_20260416_181521/nnx_feat_nnx_post_train_fixes_20260416_181521_02_synthetic/checkpoints/0/items W0416 19:43:24.535350 4049865 pjrt_executable.cc:642] Assume version compatibility. PjRt-IFRT does not track XLA executable versions. W0416 19:43:24.544827 4049865 pjrt_executable.cc:642] Assume version compatibility. PjRt-IFRT does not track XLA executable versions. W0416 19:43:24.550134 4049865 pjrt_executable.cc:642] Assume version compatibility. PjRt-IFRT does not track XLA executable versions. W0416 19:43:24.559207 4049865 pjrt_executable.cc:642] Assume version compatibility. PjRt-IFRT does not track XLA executable versions. W0416 19:43:24.567169 4049865 pjrt_executable.cc:642] Assume version compatibility. PjRt-IFRT does not track XLA executable versions. W0416 19:43:24.580169 4049865 pjrt_executable.cc:642] Assume version compatibility. PjRt-IFRT does not track XLA executable versions. W0416 19:43:24.587126 4049865 pjrt_executable.cc:642] Assume version compatibility. PjRt-IFRT does not track XLA executable versions. W0416 19:43:24.592766 4049865 pjrt_executable.cc:642] Assume version compatibility. PjRt-IFRT does not track XLA executable versions. I0416 19:43:24.683767 134895934375488 checkpoint.py:188] Wrote Metadata={'item_handlers': None, 'metrics': {}, 'performance_metrics': {}, 'init_timestamp_nsecs': 1776368604426308649, 'commit_timestamp_nsecs': None, 'custom_metadata': {}}, json={"item_handlers": null, "metrics": {}, "performance_metrics": {}, "init_timestamp_nsecs": 1776368604426308649, "commit_timestamp_nsecs": null, "custom_metadata": {}} to gs://wanglance-maxtext/nnx_ckpt_feat_nnx_post_train_fixes_20260416_181521/nnx_feat_nnx_post_train_fixes_20260416_181521_02_synthetic/checkpoints/0/_CHECKPOINT_METADATA I0416 19:43:25.123006 4055683 google_auth_provider.cc:149] Using credentials at ~/.config/gcloud/application_default_credentials.json I0416 19:43:25.123077 4055683 google_auth_provider.cc:156] Using OAuth2 AuthProvider I0416 19:43:25.308892 135003411475584 base_pytree_checkpoint_handler.py:153] [process=0][thread=MainThread] Initiated "orbax.checkpoint._src.serialization.jax_array_handlers.ArrayHandler".serialize. Time taken: 1.441870s I0416 19:43:25.315651 135003411475584 base_pytree_checkpoint_handler.py:128] [process=0] /jax/checkpoint/write/blocking_gbytes_per_sec: 7.951 GiB/s (total gbytes: 12.3 GiB) (time elapsed: a second) (per-host) I0416 19:43:25.315815 135003411475584 base_pytree_checkpoint_handler.py:732] [process=0][thread=MainThread] Initiated Pytree async_save. Time taken: 1.552331s (batch_requests_ready=0.096040s, total_serialization_initiated=1.449625s, others=0.006666s) I0416 19:43:25.315918 135003411475584 composite_checkpoint_handler.py:715] [process=0][thread=MainThread] Initiated CompositeCheckpointHandler.async_save. Time taken: 1.553333s (all_items=0.000021s, per_item={'items': '0.00002050'}, temp_paths=1.553313) I0416 19:43:25.317023 134895995192896 async_checkpointer.py:79] [process=0][thread=async_save] Background save thread started. I0416 19:43:25.317118 135003411475584 async_checkpointer.py:561] Finished blocking save. Time taken: 1.642872s. Continuing background save to gs://wanglance-maxtext/nnx_ckpt_feat_nnx_post_train_fixes_20260416_181521/nnx_feat_nnx_post_train_fixes_20260416_181521_02_synthetic/checkpoints/0. I0416 19:43:25.317301 135003411475584 checkpoint_manager.py:1549] [process=0][thread=MainThread][step=0] Starting CheckpointManager Save Finalize thread=save_finalize I0416 19:43:25.317496 134895850489408 async_checkpointer.py:265] [process=0][thread=save_finalize] Waiting for background save thread=async_save. I0416 19:43:25.317623 135003411475584 standard_logger.py:34] {'step': 0, 'event_type': 'save', 'directory': 'gs://wanglance-maxtext/nnx_ckpt_feat_nnx_post_train_fixes_20260416_181521/nnx_feat_nnx_post_train_fixes_20260416_181521_02_synthetic/checkpoints', 'reached_preemption': False, 'preemption_received_at': None, 'synchronous': False, 'wait_for_prev_start_time': 1776368603.674036, 'wait_for_prev_duration_secs': 5.435943603515625e-05, 'time_between_consecutive_saves_sec': None, 'checkpointer_blocking_start_time': 1776368603.6742272, 'checkpointer_blocking_duration_secs': 1.6429972648620605, 'get_old_steps_start_time': 1776368605.317246, 'get_old_steps_duration_secs': 2.0503997802734375e-05, 'checkpoint_manager_blocking_start_time': 1776368603.6739576, 'checkpoint_manager_blocking_duration_secs': 1.6436357498168945} I0416 19:43:25.317741 135003411475584 checkpointing.py:409] Started an asynchronous checkpoint save for step 0 I0416 19:43:25.317792 135003411475584 max_utils.py:750] Memstats: After params initialized: I0416 19:43:25.317838 135003411475584 max_utils.py:756] Using (GB) 1.59 / 31.25 (5.088000%) on TPU_0(process=0,(0,0,0,0)) I0416 19:43:25.317869 135003411475584 max_utils.py:756] Using (GB) 1.59 / 31.25 (5.088000%) on TPU_1(process=0,(1,0,0,0)) I0416 19:43:25.317891 135003411475584 max_utils.py:756] Using (GB) 1.59 / 31.25 (5.088000%) on TPU_2(process=0,(0,1,0,0)) I0416 19:43:25.317909 135003411475584 max_utils.py:756] Using (GB) 1.59 / 31.25 (5.088000%) on TPU_3(process=0,(1,1,0,0)) I0416 19:43:25.317928 135003411475584 max_utils.py:756] Using (GB) 1.59 / 31.25 (5.088000%) on TPU_4(process=0,(0,2,0,0)) I0416 19:43:25.317945 135003411475584 max_utils.py:756] Using (GB) 1.59 / 31.25 (5.088000%) on TPU_5(process=0,(1,2,0,0)) I0416 19:43:25.317962 135003411475584 max_utils.py:756] Using (GB) 1.59 / 31.25 (5.088000%) on TPU_6(process=0,(0,3,0,0)) I0416 19:43:25.317981 135003411475584 max_utils.py:756] Using (GB) 1.59 / 31.25 (5.088000%) on TPU_7(process=0,(1,3,0,0)) W0416 19:43:25.324704 4049865 pjrt_executable.cc:642] Assume version compatibility. PjRt-IFRT does not track XLA executable versions. W0416 19:43:25.330127 4049865 pjrt_executable.cc:642] Assume version compatibility. PjRt-IFRT does not track XLA executable versions. W0416 19:43:25.334685 4049865 pjrt_executable.cc:642] Assume version compatibility. PjRt-IFRT does not track XLA executable versions. I0416 19:43:25.666577 135003411475584 metric_logger.py:185] completed step: 0, seconds: 0.721, TFLOP/s/device: 18.833, Tokens/s/device: 2838.767, total_weights: 16384, loss: 10.881 I0416 19:43:25.667571 135003411475584 metric_logger.py:269] To see full metrics 'tensorboard --logdir=gs://wanglance-maxtext/nnx_ckpt_feat_nnx_post_train_fixes_20260416_181521/nnx_feat_nnx_post_train_fixes_20260416_181521_02_synthetic/tensorboard/' I0416 19:43:25.778915 135003411475584 metric_logger.py:185] completed step: 1, seconds: 2.106, TFLOP/s/device: 6.453, Tokens/s/device: 972.687, total_weights: 16384, loss: 10.881 I0416 19:43:25.898058 135003411475584 metric_logger.py:185] completed step: 2, seconds: 0.018, TFLOP/s/device: 765.515, Tokens/s/device: 115386.782, total_weights: 16384, loss: 9.779 I0416 19:43:26.016771 135003411475584 metric_logger.py:185] completed step: 3, seconds: 0.105, TFLOP/s/device: 129.336, Tokens/s/device: 19494.922, total_weights: 16384, loss: 8.837 I0416 19:43:26.265394 134895881946688 array_metadata_store.py:203] [process=0][thread=array_type_handler] Wrote 69 array_metadata.ArrayMetadata to gs://wanglance-maxtext/nnx_ckpt_feat_nnx_post_train_fixes_20260416_181521/nnx_feat_nnx_post_train_fixes_20260416_181521_02_synthetic/checkpoints/0/items/array_metadatas/process_0 I0416 19:43:45.411634 135003411475584 metric_logger.py:185] completed step: 4, seconds: 0.119, TFLOP/s/device: 114.029, Tokens/s/device: 17187.697, total_weights: 16384, loss: 7.961 I0416 19:43:45.422943 135003411475584 metric_logger.py:185] completed step: 5, seconds: 0.120, TFLOP/s/device: 113.682, Tokens/s/device: 17135.351, total_weights: 16384, loss: 7.226 I0416 19:43:45.537029 135003411475584 metric_logger.py:185] completed step: 6, seconds: 19.396, TFLOP/s/device: 0.700, Tokens/s/device: 105.586, total_weights: 16384, loss: 6.669 I0416 19:43:45.655616 135003411475584 metric_logger.py:185] completed step: 7, seconds: 0.007, TFLOP/s/device: 1961.191, Tokens/s/device: 295612.009, total_weights: 16384, loss: 6.286 I0416 19:43:45.773946 135003411475584 metric_logger.py:185] completed step: 8, seconds: 0.116, TFLOP/s/device: 117.533, Tokens/s/device: 17715.803, total_weights: 16384, loss: 6.048 I0416 19:43:45.893027 135003411475584 checkpointing.py:794] Waiting for step 9 to finish before checkpoint... I0416 19:43:45.894218 135003411475584 checkpointing.py:798] Waited 0.0012049674987792969 seconds for step 9 to finish before starting checkpointing. I0416 19:43:45.894515 135003411475584 checkpoint_manager.py:1994] [process=0][thread=MainThread][step=0][wait_until_finished] Waiting for Save Finalize thread (save_finalize) to complete. I0416 19:43:49.117421 134895871460928 base_pytree_checkpoint_handler.py:1217] [process=0][thread=write_metadata_after_commits] Commit + Array metadata written. Time taken: 23.800961s (commit=23.361355s, array_metadata_write=0.439606s) I0416 19:43:49.118578 134895995192896 base_pytree_checkpoint_handler.py:128] [process=0] /jax/checkpoint/write/gbytes_per_sec: 498.393 MiB/s (total gbytes: 12.3 GiB) (time elapsed: 25 seconds) (per-host) I0416 19:43:49.118641 134895995192896 async_checkpointer.py:90] [process=0][thread=async_save] 3 Handler Commit operations completed. Time taken: 23.801476s. I0416 19:43:49.354700 134895995192896 checkpoint.py:228] Read Metadata={'item_handlers': None, 'metrics': {}, 'performance_metrics': {}, 'init_timestamp_nsecs': 1776368604426308649, 'commit_timestamp_nsecs': None, 'custom_metadata': {}} from gs://wanglance-maxtext/nnx_ckpt_feat_nnx_post_train_fixes_20260416_181521/nnx_feat_nnx_post_train_fixes_20260416_181521_02_synthetic/checkpoints/0/_CHECKPOINT_METADATA I0416 19:43:49.523258 134895995192896 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 19:43:49.731490 134895934375488 checkpoint.py:247] Updated Metadata={'item_handlers': {'items': 'orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeCheckpointHandler'}, 'metrics': {}, 'performance_metrics': {}, 'init_timestamp_nsecs': 1776368604426308649, 'commit_timestamp_nsecs': None, 'custom_metadata': {}} to gs://wanglance-maxtext/nnx_ckpt_feat_nnx_post_train_fixes_20260416_181521/nnx_feat_nnx_post_train_fixes_20260416_181521_02_synthetic/checkpoints/0/_CHECKPOINT_METADATA I0416 19:43:49.961912 134895995192896 ocdbt_utils.py:56] Param validation support for Zarr3 will be added later (b/362328389). I0416 19:43:49.962108 134895995192896 base_pytree_checkpoint_handler.py:1342] [process=0][thread=async_save] Pytree save finalize (merge_ocdbt + ArrayMetadata validation) completed. Time taken: 0.569266s. use_zarr3=True, enable_post_merge_validation=True, directory=gs://wanglance-maxtext/nnx_ckpt_feat_nnx_post_train_fixes_20260416_181521/nnx_feat_nnx_post_train_fixes_20260416_181521_02_synthetic/checkpoints/0/items I0416 19:43:49.962727 134895995192896 atomicity.py:608] Finalizing gs://wanglance-maxtext/nnx_ckpt_feat_nnx_post_train_fixes_20260416_181521/nnx_feat_nnx_post_train_fixes_20260416_181521_02_synthetic/checkpoints/0/items I0416 19:43:50.203868 134895995192896 atomicity.py:608] Finalizing gs://wanglance-maxtext/nnx_ckpt_feat_nnx_post_train_fixes_20260416_181521/nnx_feat_nnx_post_train_fixes_20260416_181521_02_synthetic/checkpoints/0 I0416 19:43:50.913341 134895995192896 atomicity.py:794] [process=0][thread=async_save] Finished saving checkpoint (finalized tmp dir) to `gs://wanglance-maxtext/nnx_ckpt_feat_nnx_post_train_fixes_20260416_181521/nnx_feat_nnx_post_train_fixes_20260416_181521_02_synthetic/checkpoints/0`. I0416 19:43:50.913984 134895995192896 async_checkpointer.py:420] Finished async_save (blocking + background). Time taken: 27.239741s. directory=gs://wanglance-maxtext/nnx_ckpt_feat_nnx_post_train_fixes_20260416_181521/nnx_feat_nnx_post_train_fixes_20260416_181521_02_synthetic/checkpoints/0 I0416 19:43:50.914086 134895995192896 async_checkpointer.py:144] [process=0][thread=async_save] Background save thread done. Time taken: 25.596924s. I0416 19:43:50.914289 134895850489408 async_checkpointer.py:273] [process=0][thread=save_finalize] Done with waiting for background save thread=async_save. I0416 19:43:50.914402 134895850489408 async_checkpointer.py:283] [process=0][thread=save_finalize] No errors found in background save thread=async_save. I0416 19:43:50.914467 134895850489408 checkpoint_manager.py:2103] [process=0][thread=save_finalize][step=0] CheckpointManager Save Finalize is syncing with other hosts... I0416 19:43:50.914517 134895850489408 checkpoint_manager.py:2112] [process=0][thread=save_finalize][step=0] CheckpointManager Save Finalize is done on all hosts. I0416 19:43:50.914630 135003411475584 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 19:43:50.914711 135003411475584 checkpoint_manager.py:1441] Waiting for previous save to complete took 5.020227 seconds. If this number is high, consider checkpointing less frequently. I0416 19:43:50.915857 135003411475584 checkpoint_manager.py:1501] [process=0] Saving checkpoint at step 9 I0416 19:43:50.916118 135003411475584 async_checkpointer.py:452] [process=0] Started async saving checkpoint to gs://wanglance-maxtext/nnx_ckpt_feat_nnx_post_train_fixes_20260416_181521/nnx_feat_nnx_post_train_fixes_20260416_181521_02_synthetic/checkpoints/9. I0416 19:43:51.088411 134895850489408 atomicity.py:137] Creating tmp directory gs://wanglance-maxtext/nnx_ckpt_feat_nnx_post_train_fixes_20260416_181521/nnx_feat_nnx_post_train_fixes_20260416_181521_02_synthetic/checkpoints/9 I0416 19:43:51.111142 135003411475584 jax_array_handlers.py:347] Scheduling D2H of 69 prioritized jax.Array. I0416 19:43:51.111240 135003411475584 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 19:43:51.792806 134895840003648 atomicity.py:137] Creating tmp directory gs://wanglance-maxtext/nnx_ckpt_feat_nnx_post_train_fixes_20260416_181521/nnx_feat_nnx_post_train_fixes_20260416_181521_02_synthetic/checkpoints/9/items I0416 19:44:01.574455 135003411475584 base_pytree_checkpoint_handler.py:153] [process=0][thread=MainThread] Initiated "orbax.checkpoint._src.serialization.jax_array_handlers.ArrayHandler".serialize. Time taken: 10.464318s I0416 19:44:01.581306 135003411475584 base_pytree_checkpoint_handler.py:128] [process=0] /jax/checkpoint/write/blocking_gbytes_per_sec: 1.167 GiB/s (total gbytes: 12.3 GiB) (time elapsed: 10 seconds) (per-host) I0416 19:44:01.581368 135003411475584 base_pytree_checkpoint_handler.py:732] [process=0][thread=MainThread] Initiated Pytree async_save. Time taken: 10.572932s (batch_requests_ready=0.094303s, total_serialization_initiated=10.471879s, others=0.006750s) I0416 19:44:01.581448 135003411475584 composite_checkpoint_handler.py:715] [process=0][thread=MainThread] Initiated CompositeCheckpointHandler.async_save. Time taken: 10.573485s (all_items=0.000023s, per_item={'items': '0.00002289'}, temp_paths=10.573462) I0416 19:44:01.582847 134895850489408 async_checkpointer.py:79] [process=0][thread=async_save] Background save thread started. I0416 19:44:01.582930 135003411475584 async_checkpointer.py:561] Finished blocking save. Time taken: 10.667026s. Continuing background save to gs://wanglance-maxtext/nnx_ckpt_feat_nnx_post_train_fixes_20260416_181521/nnx_feat_nnx_post_train_fixes_20260416_181521_02_synthetic/checkpoints/9. I0416 19:44:01.583149 135003411475584 checkpoint_manager.py:1549] [process=0][thread=MainThread][step=9] Starting CheckpointManager Save Finalize thread=save_finalize I0416 19:44:01.583333 134895995192896 async_checkpointer.py:265] [process=0][thread=save_finalize] Waiting for background save thread=async_save. I0416 19:44:01.583445 135003411475584 standard_logger.py:34] {'step': 9, 'event_type': 'save', 'directory': 'gs://wanglance-maxtext/nnx_ckpt_feat_nnx_post_train_fixes_20260416_181521/nnx_feat_nnx_post_train_fixes_20260416_181521_02_synthetic/checkpoints', 'reached_preemption': False, 'preemption_received_at': None, 'synchronous': False, 'wait_for_prev_start_time': 1776368625.8944612, 'wait_for_prev_duration_secs': 5.0202271938323975, 'time_between_consecutive_saves_sec': None, 'checkpointer_blocking_start_time': 1776368630.9158823, 'checkpointer_blocking_duration_secs': 10.66715383529663, 'get_old_steps_start_time': 1776368641.5830665, 'get_old_steps_duration_secs': 4.2438507080078125e-05, 'checkpoint_manager_blocking_start_time': 1776368625.8944077, 'checkpoint_manager_blocking_duration_secs': 15.689009428024292} I0416 19:44:01.583563 135003411475584 checkpointing.py:409] Started an asynchronous checkpoint save for step 9 I0416 19:44:01.583595 135003411475584 checkpoint_manager.py:1994] [process=0][thread=MainThread][step=9][wait_until_finished] Waiting for Save Finalize thread (save_finalize) to complete. I0416 19:44:02.326195 134895913403968 array_metadata_store.py:203] [process=0][thread=array_type_handler] Wrote 69 array_metadata.ArrayMetadata to gs://wanglance-maxtext/nnx_ckpt_feat_nnx_post_train_fixes_20260416_181521/nnx_feat_nnx_post_train_fixes_20260416_181521_02_synthetic/checkpoints/9/items/array_metadatas/process_0 I0416 19:44:39.777882 134895840003648 base_pytree_checkpoint_handler.py:1217] [process=0][thread=write_metadata_after_commits] Commit + Array metadata written. Time taken: 38.195683s (commit=37.752956s, array_metadata_write=0.442727s) I0416 19:44:39.779148 134895850489408 base_pytree_checkpoint_handler.py:128] [process=0] /jax/checkpoint/write/gbytes_per_sec: 259.106 MiB/s (total gbytes: 12.3 GiB) (time elapsed: 48 seconds) (per-host) I0416 19:44:39.779217 134895850489408 async_checkpointer.py:90] [process=0][thread=async_save] 3 Handler Commit operations completed. Time taken: 38.196237s. I0416 19:44:40.210184 134895850489408 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 19:44:40.634900 134895850489408 ocdbt_utils.py:56] Param validation support for Zarr3 will be added later (b/362328389). I0416 19:44:40.635539 134895850489408 base_pytree_checkpoint_handler.py:1342] [process=0][thread=async_save] Pytree save finalize (merge_ocdbt + ArrayMetadata validation) completed. Time taken: 0.576478s. use_zarr3=True, enable_post_merge_validation=True, directory=gs://wanglance-maxtext/nnx_ckpt_feat_nnx_post_train_fixes_20260416_181521/nnx_feat_nnx_post_train_fixes_20260416_181521_02_synthetic/checkpoints/9/items I0416 19:44:40.636191 134895850489408 atomicity.py:608] Finalizing gs://wanglance-maxtext/nnx_ckpt_feat_nnx_post_train_fixes_20260416_181521/nnx_feat_nnx_post_train_fixes_20260416_181521_02_synthetic/checkpoints/9/items I0416 19:44:40.882656 134895850489408 atomicity.py:608] Finalizing gs://wanglance-maxtext/nnx_ckpt_feat_nnx_post_train_fixes_20260416_181521/nnx_feat_nnx_post_train_fixes_20260416_181521_02_synthetic/checkpoints/9 I0416 19:44:41.593982 134895850489408 atomicity.py:794] [process=0][thread=async_save] Finished saving checkpoint (finalized tmp dir) to `gs://wanglance-maxtext/nnx_ckpt_feat_nnx_post_train_fixes_20260416_181521/nnx_feat_nnx_post_train_fixes_20260416_181521_02_synthetic/checkpoints/9`. I0416 19:44:41.594632 134895850489408 async_checkpointer.py:420] Finished async_save (blocking + background). Time taken: 50.678734s. directory=gs://wanglance-maxtext/nnx_ckpt_feat_nnx_post_train_fixes_20260416_181521/nnx_feat_nnx_post_train_fixes_20260416_181521_02_synthetic/checkpoints/9 I0416 19:44:41.594713 134895850489408 async_checkpointer.py:144] [process=0][thread=async_save] Background save thread done. Time taken: 40.011738s. I0416 19:44:41.594886 134895995192896 async_checkpointer.py:273] [process=0][thread=save_finalize] Done with waiting for background save thread=async_save. I0416 19:44:41.594993 134895995192896 async_checkpointer.py:283] [process=0][thread=save_finalize] No errors found in background save thread=async_save. I0416 19:44:41.595097 134895995192896 checkpoint_manager.py:2103] [process=0][thread=save_finalize][step=9] CheckpointManager Save Finalize is syncing with other hosts... I0416 19:44:41.595150 134895995192896 checkpoint_manager.py:2112] [process=0][thread=save_finalize][step=9] CheckpointManager Save Finalize is done on all hosts. I0416 19:44:41.595443 135003411475584 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 19:44:41.595624 135003411475584 checkpoint_manager.py:1983] [process=0][thread=MainThread][wait_until_finished] No Save Finalize thread to wait for. Returning. I0416 19:44:41.597031 135003411475584 metric_logger.py:185] completed step: 9, seconds: 0.118, TFLOP/s/device: 115.402, Tokens/s/device: 17394.702, total_weights: 16384, loss: 5.908 Per train step: Total TFLOPs: 13.59 split as 93.93% learnable weight flops and 6.07% attention flops