2026-04-16 20:06:44.747333: 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 20:06:44.861850 131200677657728 max_utils.py:238] Skipping jax distributed system due to skip_jax_distributed_system=True flag. ~/maxtext_venv/lib/python3.12/site-packages/jax/_src/xla_bridge.py:219: UserWarning: TPU backend initialization is taking more than 60.0 seconds. Did you run your code on all TPU hosts? See https://docs.jax.dev/en/latest/multi_process.html for more information. warnings.warn( I0416 20:07:45.251153 131200677657728 max_utils.py:800] System Information: Jax Version: 0.9.2 I0416 20:07:45.251259 131200677657728 max_utils.py:801] System Information: Jaxlib Version: 0.9.2 I0416 20:07:45.251293 131200677657728 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 20:07:45.251318 131200677657728 train_utils.py:364] WARNING: 'dataset_path' might be pointing your local file system I0416 20:07:45.251396 131200677657728 train.py:811] [DECOUPLED NO-OP] skipping cloud diagnostics wrapper. W0416 20:07:45.343449 4079847 pjrt_executable.cc:642] Assume version compatibility. PjRt-IFRT does not track XLA executable versions. I0416 20:07:45.745706 131200677657728 maxtext_utils.py:1687] Num_devices: 8, shape (1, 1, 1, 2, 1, 1, 1, 1, 4, 1, 1, 1, 1) I0416 20:07:45.843896 131200677657728 checkpointing.py:688] Setting up checkpoint logger... I0416 20:07:45.843993 131200677657728 checkpointing.py:234] Creating checkpoint manager with ocdbt=True and zarr3=True I0416 20:07:45.844033 131200677657728 pytree_checkpoint_handler.py:577] save_device_host_concurrent_bytes=None I0416 20:07:45.844233 131200677657728 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 0x7752f31f4d40>, enable_pinned_host_transfer=False, save_concurrent_bytes: 96000000000 (89.4 GiB), restore_concurrent_bytes: 96000000000 (89.4 GiB) I0416 20:07:48.311154 131200677657728 checkpointing.py:266] Enabling policy for fixed interval checkpointing. I0416 20:07:48.311333 131200677657728 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 0x774c76fc5730>}, handler_registry=None I0416 20:07:48.311589 131200677657728 composite_checkpoint_handler.py:237] Deferred registration for item: "items". Adding handler `<orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeCheckpointHandler object at 0x774c76fc5730>` 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 20:07:48.311633 131200677657728 composite_checkpoint_handler.py:237] Deferred registration for item: "metrics". Adding handler `<orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonCheckpointHandler object at 0x774c76fc4b00>` 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 20:07:48.311661 131200677657728 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 0x774c76fc5730>, ('items', <class 'orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeRestoreArgs'>): <orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeCheckpointHandler object at 0x774c76fc5730>, ('metrics', <class 'orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonSaveArgs'>): <orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonCheckpointHandler object at 0x774c76fc4b00>, ('metrics', <class 'orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonRestoreArgs'>): <orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonCheckpointHandler object at 0x774c76fc4b00>}). I0416 20:07:48.312236 131200677657728 abstract_checkpointer.py:35] orbax-checkpoint version: 0.11.28 I0416 20:07:48.312286 131200677657728 async_checkpointer.py:177] [process=0][thread=MainThread] Using barrier_sync_fn: <function get_barrier_sync_fn.<locals>.<lambda> at 0x774c76f91800> timeout: 600 secs and primary_host=0 for async checkpoint writes I0416 20:07:48.433212 131200677657728 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_09_pdb_lt_1/checkpoints I0416 20:07:48.433432 131200677657728 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_09_pdb_lt_1/checkpoints: <orbax.checkpoint.checkpoint_manager.CheckpointManager object at 0x774c76fc7e00> I0416 20:07:48.433509 131200677657728 checkpointing.py:302] Checkpoint manager created! I0416 20:07:48.514392 131200677657728 dataset_info.py:707] Load dataset info from tests/assets/local_datasets/c4_en_dataset_minimal/c4/en/3.1.0 I0416 20:07:48.517427 131200677657728 reader.py:262] Creating a tf.data.Dataset reading 8 files located in folders: tests/assets/local_datasets/c4_en_dataset_minimal/c4/en/3.1.0. I0416 20:07:48.570148 131200677657728 logging_logger.py:49] Constructing tf.data.Dataset __local_c4_builder for split train, from tests/assets/local_datasets/c4_en_dataset_minimal/c4/en/3.1.0 I0416 20:07:48.601341 131200677657728 tokenizer.py:245] Tokenizer path: src/maxtext/assets/tokenizers/tokenizer.llama2 I0416 20:07:48.601403 131200677657728 tokenizer.py:187] Loading sentencepiece tokenizer: src/maxtext/assets/tokenizers/tokenizer.llama2 I0416 20:07:49.610821 131200677657728 checkpointing.py:578] checkpoint manager exists so trying to load this run's existing checkpoint I0416 20:07:49.610925 131200677657728 checkpointing.py:676] No existing checkpoints found, not restoring checkpoint. W0416 20:07:49.721678 4079847 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: 2 tensor: 4 I0416 20:07:49.808676 131200677657728 maxtext_utils.py:1805] decoder/decoder_norm/scale/value Shape: float32[2048] Physical: ('tensor',) I0416 20:07:49.808779 131200677657728 maxtext_utils.py:1805] decoder/layers/mlp/wi_0/kernel/value Shape: float32[2048,16,7168] Physical: ('fsdp', None, 'tensor') I0416 20:07:49.808821 131200677657728 maxtext_utils.py:1805] decoder/layers/mlp/wi_1/kernel/value Shape: float32[2048,16,7168] Physical: ('fsdp', None, 'tensor') I0416 20:07:49.808853 131200677657728 maxtext_utils.py:1805] decoder/layers/mlp/wo/kernel/value Shape: float32[7168,16,2048] Physical: ('tensor', None, 'fsdp') I0416 20:07:49.808881 131200677657728 maxtext_utils.py:1805] decoder/layers/post_self_attention_layer_norm/scale/value Shape: float32[2048,16] Physical: ('tensor', None) I0416 20:07:49.808906 131200677657728 maxtext_utils.py:1805] decoder/layers/pre_self_attention_layer_norm/scale/value Shape: float32[2048,16] Physical: ('tensor', None) I0416 20:07:49.808932 131200677657728 maxtext_utils.py:1805] decoder/layers/self_attention/key/kernel/value Shape: float32[2048,16,16,128] Physical: ('fsdp', None, 'tensor', None) I0416 20:07:49.808958 131200677657728 maxtext_utils.py:1805] decoder/layers/self_attention/out/kernel/value Shape: float32[16,16,128,2048] Physical: ('tensor', None, None, 'fsdp') I0416 20:07:49.808983 131200677657728 maxtext_utils.py:1805] decoder/layers/self_attention/query/kernel/value Shape: float32[2048,16,16,128] Physical: ('fsdp', None, 'tensor', None) I0416 20:07:49.809006 131200677657728 maxtext_utils.py:1805] decoder/layers/self_attention/value/kernel/value Shape: float32[2048,16,16,128] Physical: ('fsdp', None, 'tensor', None) I0416 20:07:49.809029 131200677657728 maxtext_utils.py:1805] decoder/logits_dense/kernel/value Shape: float32[2048,32000] Physical: ('fsdp', 'tensor') I0416 20:07:49.809071 131200677657728 maxtext_utils.py:1805] token_embedder/embedding/value Shape: float32[32000,2048] Physical: ('tensor', 'fsdp') I0416 20:07:49.925958 131200677657728 nnx_decoders.py:465] nnx_decoders/carry Logical: bfloat16[2,2048,2048]....................................... ('activation_batch', 'activation_norm_length', 'activation_embed'). I0416 20:07:49.926055 131200677657728 nnx_decoders.py:465] nnx_decoders/carry Physical: bfloat16[2,2048,2048]....................................... ('fsdp', None, 'tensor'). I0416 20:07:49.931182 131200677657728 nnx_decoders.py:465] Unknown Logical: bfloat16[2,2048,2048]....................................... ('activation_batch', 'activation_norm_length', 'activation_embed'). I0416 20:07:49.931230 131200677657728 nnx_decoders.py:465] Unknown Physical: bfloat16[2,2048,2048]....................................... ('fsdp', None, 'tensor'). I0416 20:07:49.946220 131200677657728 attentions.py:1088] attentions/inputs_q Logical: bfloat16[2,2048,2048]....................................... ('activation_batch', 'activation_attn_length', 'activation_attn_embed'). I0416 20:07:49.946271 131200677657728 attentions.py:1088] attentions/inputs_q Physical: bfloat16[2,2048,2048]....................................... ('fsdp', None, 'tensor'). I0416 20:07:49.960736 131200677657728 attentions.py:1089] attentions/inputs_kv Logical: bfloat16[2,2048,2048]....................................... ('activation_batch', 'activation_attn_length', 'activation_attn_embed'). I0416 20:07:49.960785 131200677657728 attentions.py:1089] attentions/inputs_kv Physical: bfloat16[2,2048,2048]....................................... ('fsdp', None, 'tensor'). I0416 20:07:49.982642 131200677657728 attentions.py:1154] attentions/query Logical: bfloat16[2,2048,16,128]..................................... ('activation_kv_batch', 'activation_attn_length', 'activation_kv_heads', 'activation_kv_head_dim'). I0416 20:07:49.982706 131200677657728 attentions.py:1154] attentions/query Physical: bfloat16[2,2048,16,128]..................................... ('fsdp', None, 'tensor', None). I0416 20:07:49.997300 131200677657728 attentions.py:1155] attentions/key Logical: bfloat16[2,2048,16,128]..................................... ('activation_kv_batch', 'activation_attn_length', 'activation_kv_heads', 'activation_kv_head_dim'). I0416 20:07:49.997349 131200677657728 attentions.py:1155] attentions/key Physical: bfloat16[2,2048,16,128]..................................... ('fsdp', None, 'tensor', None). I0416 20:07:50.011885 131200677657728 attentions.py:1156] attentions/value Logical: bfloat16[2,2048,16,128]..................................... ('activation_kv_batch', 'activation_attn_length', 'activation_kv_heads', 'activation_kv_head_dim'). I0416 20:07:50.011934 131200677657728 attentions.py:1156] attentions/value Physical: bfloat16[2,2048,16,128]..................................... ('fsdp', None, 'tensor', None). I0416 20:07:50.039634 131200677657728 attentions.py:1197] attentions/out Logical: bfloat16[2,2048,16,128]..................................... ('activation_batch', 'activation_attn_length', 'activation_heads', 'activation_kv'). I0416 20:07:50.039700 131200677657728 attentions.py:1197] attentions/out Physical: bfloat16[2,2048,16,128]..................................... ('fsdp', None, 'tensor', None). I0416 20:07:50.057700 131200677657728 linears.py:525] linears/x Logical: bfloat16[2,2048,7168]....................................... ('activation_batch', 'activation_length', 'activation_mlp'). I0416 20:07:50.057754 131200677657728 linears.py:525] linears/x Physical: bfloat16[2,2048,7168]....................................... ('fsdp', None, 'tensor'). W0416 20:07:50.411383 4079847 pjrt_executable.cc:642] Assume version compatibility. PjRt-IFRT does not track XLA executable versions. I0416 20:07:50.490625 131200677657728 max_utils.py:791] Total memory size: 2.5 GB, Output size: 1.5 GB, Temp size: 0.9 GB, Argument size: 1.5 GB, Host temp size: 0.0 GB. I0416 20:07:50.491160 131200677657728 max_utils.py:194] tensorboardX not available; using no-op SummaryWriter. I0416 20:07:50.492617 131200677657728 metric_logger.py:289] number parameters: 1.104 billion W0416 20:07:53.302545 4079847 pjrt_executable.cc:642] Assume version compatibility. PjRt-IFRT does not track XLA executable versions. I0416 20:07:53.400276 131200677657728 checkpointing.py:794] Waiting for step 0 to finish before checkpoint... I0416 20:07:53.454931 131200677657728 checkpointing.py:798] Waited 0.054648399353027344 seconds for step 0 to finish before starting checkpointing. I0416 20:07:53.455344 131200677657728 checkpoint_manager.py:1983] [process=0][thread=MainThread][wait_until_finished] No Save Finalize thread to wait for. Returning. I0416 20:07:53.455490 131200677657728 checkpoint_manager.py:1501] [process=0] Saving checkpoint at step 0 I0416 20:07:53.455851 131200677657728 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_09_pdb_lt_1/checkpoints/0. I0416 20:07:53.543982 131200677657728 signaling_client.py:373] Using ThreadSafeKeyValueSignalingClient I0416 20:07:53.631916 131087493760576 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_09_pdb_lt_1/checkpoints/0 I0416 20:07:53.650263 131200677657728 jax_array_handlers.py:347] Scheduling D2H of 69 prioritized jax.Array. I0416 20:07:53.650354 131200677657728 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 20:07:54.350208 131087483274816 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_09_pdb_lt_1/checkpoints/0/items I0416 20:07:54.487102 131087439234624 checkpoint.py:188] Wrote Metadata={'item_handlers': None, 'metrics': {}, 'performance_metrics': {}, 'init_timestamp_nsecs': 1776370074243022597, 'commit_timestamp_nsecs': None, 'custom_metadata': {}}, json={"item_handlers": null, "metrics": {}, "performance_metrics": {}, "init_timestamp_nsecs": 1776370074243022597, "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_09_pdb_lt_1/checkpoints/0/_CHECKPOINT_METADATA I0416 20:07:54.937973 4082618 google_auth_provider.cc:149] Using credentials at ~/.config/gcloud/application_default_credentials.json I0416 20:07:54.938032 4082618 google_auth_provider.cc:156] Using OAuth2 AuthProvider I0416 20:07:55.408720 131200677657728 base_pytree_checkpoint_handler.py:153] [process=0][thread=MainThread] Initiated "orbax.checkpoint._src.serialization.jax_array_handlers.ArrayHandler".serialize. Time taken: 1.759454s I0416 20:07:55.414827 131200677657728 base_pytree_checkpoint_handler.py:128] [process=0] /jax/checkpoint/write/blocking_gbytes_per_sec: 6.601 GiB/s (total gbytes: 12.3 GiB) (time elapsed: a second) (per-host) I0416 20:07:55.414895 131200677657728 base_pytree_checkpoint_handler.py:732] [process=0][thread=MainThread] Initiated Pytree async_save. Time taken: 1.869707s (batch_requests_ready=0.096349s, total_serialization_initiated=1.767345s, others=0.006013s) I0416 20:07:55.414971 131200677657728 composite_checkpoint_handler.py:715] [process=0][thread=MainThread] Initiated CompositeCheckpointHandler.async_save. Time taken: 1.870260s (all_items=0.000020s, per_item={'items': '0.00001955'}, temp_paths=1.870240) I0416 20:07:55.416188 131087365834304 async_checkpointer.py:79] [process=0][thread=async_save] Background save thread started. I0416 20:07:55.416268 131200677657728 async_checkpointer.py:561] Finished blocking save. Time taken: 1.960740s. Continuing background save to gs://wanglance-maxtext/nnx_ckpt_feat_nnx_post_train_fixes_20260416_181521/nnx_feat_nnx_post_train_fixes_20260416_181521_09_pdb_lt_1/checkpoints/0. I0416 20:07:55.416456 131200677657728 checkpoint_manager.py:1549] [process=0][thread=MainThread][step=0] Starting CheckpointManager Save Finalize thread=save_finalize I0416 20:07:55.416617 131087504246336 async_checkpointer.py:265] [process=0][thread=save_finalize] Waiting for background save thread=async_save. I0416 20:07:55.416706 131200677657728 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_09_pdb_lt_1/checkpoints', 'reached_preemption': False, 'preemption_received_at': None, 'synchronous': False, 'wait_for_prev_start_time': 1776370073.4553316, 'wait_for_prev_duration_secs': 4.076957702636719e-05, 'time_between_consecutive_saves_sec': None, 'checkpointer_blocking_start_time': 1776370073.45551, 'checkpointer_blocking_duration_secs': 1.9608747959136963, 'get_old_steps_start_time': 1776370075.4164023, 'get_old_steps_duration_secs': 2.002716064453125e-05, 'checkpoint_manager_blocking_start_time': 1776370073.4552476, 'checkpoint_manager_blocking_duration_secs': 1.9614336490631104} I0416 20:07:55.416817 131200677657728 checkpointing.py:409] Started an asynchronous checkpoint save for step 0 I0416 20:07:55.416867 131200677657728 max_utils.py:750] Memstats: After params initialized: I0416 20:07:55.416909 131200677657728 max_utils.py:756] Using (GB) 1.58 / 31.25 (5.056000%) on TPU_0(process=0,(0,0,0,0)) I0416 20:07:55.416944 131200677657728 max_utils.py:756] Using (GB) 1.58 / 31.25 (5.056000%) on TPU_1(process=0,(1,0,0,0)) I0416 20:07:55.416966 131200677657728 max_utils.py:756] Using (GB) 1.58 / 31.25 (5.056000%) on TPU_2(process=0,(0,1,0,0)) I0416 20:07:55.416984 131200677657728 max_utils.py:756] Using (GB) 1.58 / 31.25 (5.056000%) on TPU_3(process=0,(1,1,0,0)) I0416 20:07:55.417004 131200677657728 max_utils.py:756] Using (GB) 1.58 / 31.25 (5.056000%) on TPU_4(process=0,(0,2,0,0)) I0416 20:07:55.417019 131200677657728 max_utils.py:756] Using (GB) 1.58 / 31.25 (5.056000%) on TPU_5(process=0,(1,2,0,0)) I0416 20:07:55.417035 131200677657728 max_utils.py:756] Using (GB) 1.58 / 31.25 (5.056000%) on TPU_6(process=0,(0,3,0,0)) I0416 20:07:55.417064 131200677657728 max_utils.py:756] Using (GB) 1.58 / 31.25 (5.056000%) on TPU_7(process=0,(1,3,0,0)) W0416 20:07:55.422686 4079847 pjrt_executable.cc:642] Assume version compatibility. PjRt-IFRT does not track XLA executable versions. W0416 20:07:55.429613 4079847 pjrt_executable.cc:642] Assume version compatibility. PjRt-IFRT does not track XLA executable versions. W0416 20:07:55.434203 4079847 pjrt_executable.cc:642] Assume version compatibility. PjRt-IFRT does not track XLA executable versions. I0416 20:07:55.760823 131200677657728 metric_logger.py:185] completed step: 0, seconds: 2.906, TFLOP/s/device: 1.169, Tokens/s/device: 176.166, total_weights: 3650, loss: 10.858 I0416 20:07:55.761716 131200677657728 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_09_pdb_lt_1/tensorboard/' I0416 20:07:55.808590 131200677657728 metric_logger.py:185] completed step: 1, seconds: 2.357, TFLOP/s/device: 1.441, Tokens/s/device: 217.232, total_weights: 2676, loss: 10.887 I0416 20:07:55.860975 131200677657728 metric_logger.py:185] completed step: 2, seconds: 0.024, TFLOP/s/device: 142.638, Tokens/s/device: 21499.958, total_weights: 3783, loss: 10.228 I0416 20:07:55.913307 131200677657728 metric_logger.py:185] completed step: 3, seconds: 0.040, TFLOP/s/device: 84.204, Tokens/s/device: 12692.117, total_weights: 3680, loss: 9.523 I0416 20:07:56.412038 131087462303296 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_09_pdb_lt_1/checkpoints/0/items/array_metadatas/process_0 I0416 20:08:06.383519 131200677657728 metric_logger.py:185] completed step: 4, seconds: 0.053, TFLOP/s/device: 64.124, Tokens/s/device: 9665.484, total_weights: 3736, loss: 9.170 I0416 20:08:06.396542 131200677657728 metric_logger.py:185] completed step: 5, seconds: 0.052, TFLOP/s/device: 65.711, Tokens/s/device: 9904.629, total_weights: 4096, loss: 9.052 I0416 20:08:06.446343 131200677657728 metric_logger.py:185] completed step: 6, seconds: 10.469, TFLOP/s/device: 0.324, Tokens/s/device: 48.905, total_weights: 2931, loss: 8.916 I0416 20:08:06.498827 131200677657728 metric_logger.py:185] completed step: 7, seconds: 0.011, TFLOP/s/device: 323.195, Tokens/s/device: 48715.509, total_weights: 3526, loss: 8.819 I0416 20:08:06.551298 131200677657728 metric_logger.py:185] completed step: 8, seconds: 0.050, TFLOP/s/device: 68.203, Tokens/s/device: 10280.299, total_weights: 3012, loss: 8.753 I0416 20:08:06.604518 131200677657728 checkpointing.py:794] Waiting for step 9 to finish before checkpoint... I0416 20:08:06.605660 131200677657728 checkpointing.py:798] Waited 0.0011525154113769531 seconds for step 9 to finish before starting checkpointing. I0416 20:08:06.605930 131200677657728 checkpoint_manager.py:1994] [process=0][thread=MainThread][step=0][wait_until_finished] Waiting for Save Finalize thread (save_finalize) to complete. I0416 20:08:31.346290 131087386805824 base_pytree_checkpoint_handler.py:1217] [process=0][thread=write_metadata_after_commits] Commit + Array metadata written. Time taken: 35.930859s (commit=35.479630s, array_metadata_write=0.451229s) I0416 20:08:31.347288 131087365834304 base_pytree_checkpoint_handler.py:128] [process=0] /jax/checkpoint/write/gbytes_per_sec: 334.288 MiB/s (total gbytes: 12.3 GiB) (time elapsed: 37 seconds) (per-host) I0416 20:08:31.347344 131087365834304 async_checkpointer.py:90] [process=0][thread=async_save] 3 Handler Commit operations completed. Time taken: 35.931028s. I0416 20:08:31.602600 131087365834304 checkpoint.py:228] Read Metadata={'item_handlers': None, 'metrics': {}, 'performance_metrics': {}, 'init_timestamp_nsecs': 1776370074243022597, '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_09_pdb_lt_1/checkpoints/0/_CHECKPOINT_METADATA I0416 20:08:31.784377 131087365834304 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 20:08:31.999534 131087439234624 checkpoint.py:247] Updated Metadata={'item_handlers': {'items': 'orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeCheckpointHandler'}, 'metrics': {}, 'performance_metrics': {}, 'init_timestamp_nsecs': 1776370074243022597, '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_09_pdb_lt_1/checkpoints/0/_CHECKPOINT_METADATA I0416 20:08:32.259471 131087365834304 ocdbt_utils.py:56] Param validation support for Zarr3 will be added later (b/362328389). I0416 20:08:32.259670 131087365834304 base_pytree_checkpoint_handler.py:1342] [process=0][thread=async_save] Pytree save finalize (merge_ocdbt + ArrayMetadata validation) completed. Time taken: 0.615962s. 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_09_pdb_lt_1/checkpoints/0/items I0416 20:08:32.260323 131087365834304 atomicity.py:608] Finalizing gs://wanglance-maxtext/nnx_ckpt_feat_nnx_post_train_fixes_20260416_181521/nnx_feat_nnx_post_train_fixes_20260416_181521_09_pdb_lt_1/checkpoints/0/items I0416 20:08:32.525072 131087365834304 atomicity.py:608] Finalizing gs://wanglance-maxtext/nnx_ckpt_feat_nnx_post_train_fixes_20260416_181521/nnx_feat_nnx_post_train_fixes_20260416_181521_09_pdb_lt_1/checkpoints/0 I0416 20:08:33.250800 131087365834304 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_09_pdb_lt_1/checkpoints/0`. I0416 20:08:33.251450 131087365834304 async_checkpointer.py:420] Finished async_save (blocking + background). Time taken: 39.795928s. directory=gs://wanglance-maxtext/nnx_ckpt_feat_nnx_post_train_fixes_20260416_181521/nnx_feat_nnx_post_train_fixes_20260416_181521_09_pdb_lt_1/checkpoints/0 I0416 20:08:33.251524 131087365834304 async_checkpointer.py:144] [process=0][thread=async_save] Background save thread done. Time taken: 37.835208s. I0416 20:08:33.251681 131087504246336 async_checkpointer.py:273] [process=0][thread=save_finalize] Done with waiting for background save thread=async_save. I0416 20:08:33.251783 131087504246336 async_checkpointer.py:283] [process=0][thread=save_finalize] No errors found in background save thread=async_save. I0416 20:08:33.251840 131087504246336 checkpoint_manager.py:2103] [process=0][thread=save_finalize][step=0] CheckpointManager Save Finalize is syncing with other hosts... I0416 20:08:33.251884 131087504246336 checkpoint_manager.py:2112] [process=0][thread=save_finalize][step=0] CheckpointManager Save Finalize is done on all hosts. I0416 20:08:33.252011 131200677657728 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 20:08:33.252147 131200677657728 checkpoint_manager.py:1441] Waiting for previous save to complete took 26.646223 seconds. If this number is high, consider checkpointing less frequently. I0416 20:08:33.252888 131200677657728 checkpoint_manager.py:1501] [process=0] Saving checkpoint at step 9 I0416 20:08:33.253182 131200677657728 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_09_pdb_lt_1/checkpoints/9. I0416 20:08:33.438245 131087504246336 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_09_pdb_lt_1/checkpoints/9 I0416 20:08:33.438404 131200677657728 jax_array_handlers.py:347] Scheduling D2H of 69 prioritized jax.Array. I0416 20:08:33.438633 131200677657728 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 20:08:34.158728 131087386805824 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_09_pdb_lt_1/checkpoints/9/items I0416 20:08:44.492632 131200677657728 base_pytree_checkpoint_handler.py:153] [process=0][thread=MainThread] Initiated "orbax.checkpoint._src.serialization.jax_array_handlers.ArrayHandler".serialize. Time taken: 11.055442s I0416 20:08:44.498754 131200677657728 base_pytree_checkpoint_handler.py:128] [process=0] /jax/checkpoint/write/blocking_gbytes_per_sec: 1.106 GiB/s (total gbytes: 12.3 GiB) (time elapsed: 11 seconds) (per-host) I0416 20:08:44.498816 131200677657728 base_pytree_checkpoint_handler.py:732] [process=0][thread=MainThread] Initiated Pytree async_save. Time taken: 11.158870s (batch_requests_ready=0.089713s, total_serialization_initiated=11.063146s, others=0.006011s) I0416 20:08:44.498898 131200677657728 composite_checkpoint_handler.py:715] [process=0][thread=MainThread] Initiated CompositeCheckpointHandler.async_save. Time taken: 11.159369s (all_items=0.000017s, per_item={'items': '0.00001693'}, temp_paths=11.159352) I0416 20:08:44.500090 131087355348544 async_checkpointer.py:79] [process=0][thread=async_save] Background save thread started. I0416 20:08:44.500190 131200677657728 async_checkpointer.py:561] Finished blocking save. Time taken: 11.247249s. Continuing background save to gs://wanglance-maxtext/nnx_ckpt_feat_nnx_post_train_fixes_20260416_181521/nnx_feat_nnx_post_train_fixes_20260416_181521_09_pdb_lt_1/checkpoints/9. I0416 20:08:44.500473 131200677657728 checkpoint_manager.py:1549] [process=0][thread=MainThread][step=9] Starting CheckpointManager Save Finalize thread=save_finalize I0416 20:08:44.500679 131087365834304 async_checkpointer.py:265] [process=0][thread=save_finalize] Waiting for background save thread=async_save. I0416 20:08:44.500793 131200677657728 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_09_pdb_lt_1/checkpoints', 'reached_preemption': False, 'preemption_received_at': None, 'synchronous': False, 'wait_for_prev_start_time': 1776370086.6058939, 'wait_for_prev_duration_secs': 26.646223306655884, 'time_between_consecutive_saves_sec': None, 'checkpointer_blocking_start_time': 1776370113.2529137, 'checkpointer_blocking_duration_secs': 11.247461318969727, 'get_old_steps_start_time': 1776370124.5003953, 'get_old_steps_duration_secs': 3.170967102050781e-05, 'checkpoint_manager_blocking_start_time': 1776370086.6058576, 'checkpoint_manager_blocking_duration_secs': 37.894909381866455} I0416 20:08:44.500924 131200677657728 checkpointing.py:409] Started an asynchronous checkpoint save for step 9 I0416 20:08:44.500959 131200677657728 checkpoint_manager.py:1994] [process=0][thread=MainThread][step=9][wait_until_finished] Waiting for Save Finalize thread (save_finalize) to complete. I0416 20:08:45.348711 131087462303296 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_09_pdb_lt_1/checkpoints/9/items/array_metadatas/process_0 I0416 20:09:29.783710 131087386805824 base_pytree_checkpoint_handler.py:1217] [process=0][thread=write_metadata_after_commits] Commit + Array metadata written. Time taken: 45.284200s (commit=44.842688s, array_metadata_write=0.441512s) I0416 20:09:29.784809 131087355348544 base_pytree_checkpoint_handler.py:128] [process=0] /jax/checkpoint/write/gbytes_per_sec: 223.879 MiB/s (total gbytes: 12.3 GiB) (time elapsed: 56 seconds) (per-host) I0416 20:09:29.784923 131087355348544 async_checkpointer.py:90] [process=0][thread=async_save] 3 Handler Commit operations completed. Time taken: 45.284684s. I0416 20:09:30.243820 131087355348544 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 20:09:30.709062 131087355348544 ocdbt_utils.py:56] Param validation support for Zarr3 will be added later (b/362328389). I0416 20:09:30.709762 131087355348544 base_pytree_checkpoint_handler.py:1342] [process=0][thread=async_save] Pytree save finalize (merge_ocdbt + ArrayMetadata validation) completed. Time taken: 0.608676s. 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_09_pdb_lt_1/checkpoints/9/items I0416 20:09:30.710427 131087355348544 atomicity.py:608] Finalizing gs://wanglance-maxtext/nnx_ckpt_feat_nnx_post_train_fixes_20260416_181521/nnx_feat_nnx_post_train_fixes_20260416_181521_09_pdb_lt_1/checkpoints/9/items I0416 20:09:30.955549 131087355348544 atomicity.py:608] Finalizing gs://wanglance-maxtext/nnx_ckpt_feat_nnx_post_train_fixes_20260416_181521/nnx_feat_nnx_post_train_fixes_20260416_181521_09_pdb_lt_1/checkpoints/9 I0416 20:09:31.669915 131087355348544 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_09_pdb_lt_1/checkpoints/9`. I0416 20:09:31.670560 131087355348544 async_checkpointer.py:420] Finished async_save (blocking + background). Time taken: 58.417626s. directory=gs://wanglance-maxtext/nnx_ckpt_feat_nnx_post_train_fixes_20260416_181521/nnx_feat_nnx_post_train_fixes_20260416_181521_09_pdb_lt_1/checkpoints/9 I0416 20:09:31.670641 131087355348544 async_checkpointer.py:144] [process=0][thread=async_save] Background save thread done. Time taken: 47.170408s. I0416 20:09:31.670807 131087365834304 async_checkpointer.py:273] [process=0][thread=save_finalize] Done with waiting for background save thread=async_save. I0416 20:09:31.670912 131087365834304 async_checkpointer.py:283] [process=0][thread=save_finalize] No errors found in background save thread=async_save. I0416 20:09:31.670963 131087365834304 checkpoint_manager.py:2103] [process=0][thread=save_finalize][step=9] CheckpointManager Save Finalize is syncing with other hosts... I0416 20:09:31.671012 131087365834304 checkpoint_manager.py:2112] [process=0][thread=save_finalize][step=9] CheckpointManager Save Finalize is done on all hosts. I0416 20:09:31.671182 131200677657728 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 20:09:31.671339 131200677657728 checkpoint_manager.py:1983] [process=0][thread=MainThread][wait_until_finished] No Save Finalize thread to wait for. Returning. I0416 20:09:31.671894 131200677657728 metric_logger.py:185] completed step: 9, seconds: 0.053, TFLOP/s/device: 64.312, Tokens/s/device: 9693.849, total_weights: 3004, loss: 8.931 Per train step: Total TFLOPs: 3.40 split as 93.93% learnable weight flops and 6.07% attention flops