main (51c7f2b8e) vs feat/migrate-nnx-utils (dc4d58891)| Metric | main 51c7f2b8e | feat/migrate-nnx-utils dc4d58891 | Diff (feat/migrate-nnx-utils − main) |
|---|---|---|---|
| Parameters | 1.104 billion | 1.104 billion | — |
| Final loss | 5.9150 | 5.9150 | 0 |
| TFLOP/s | 114.570 | 114.329 | -0.241 |
| Tok/s | 17269.3 | 17233.0 | -36.328 |
| Avg s/step | 2.472 | 2.434 | -0.038 |
| Memory % | 5.09 | 5.09 | 0 |
| JAX | 0.9.2 | 0.9.2 | — |
Diff = branch value − main value. Green = branch improved. Red = branch regressed.
2026-04-15 21:14:04.886548: 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) I0415 21:14:05.002152 125104306547840 max_utils.py:238] Skipping jax distributed system due to skip_jax_distributed_system=True flag. I0415 21:14:48.515319 125104306547840 max_utils.py:800] System Information: Jax Version: 0.9.2 I0415 21:14:48.515448 125104306547840 max_utils.py:801] System Information: Jaxlib Version: 0.9.2 I0415 21:14:48.515486 125104306547840 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 I0415 21:14:48.515510 125104306547840 train_utils.py:334] WARNING: 'dataset_path' might be pointing your local file system I0415 21:14:48.515532 125104306547840 train_utils.py:347] WARNING: Sequence packing is essentially ignored for synthetic data. Please use a real dataset to use sequence packing. I0415 21:14:48.515631 125104306547840 train.py:683] [DECOUPLED NO-OP] skipping cloud diagnostics wrapper. W0415 21:14:48.608867 3226711 pjrt_executable.cc:642] Assume version compatibility. PjRt-IFRT does not track XLA executable versions. I0415 21:14:49.029210 125104306547840 maxtext_utils.py:1517] Num_devices: 8, shape (1, 1, 1, 8, 1, 1, 1, 1, 1, 1, 1, 1, 1) I0415 21:14:49.029573 125104306547840 checkpointing.py:677] Setting up checkpoint logger... I0415 21:14:49.029622 125104306547840 checkpointing.py:233] Creating checkpoint manager with ocdbt=True and zarr3=True I0415 21:14:49.029670 125104306547840 pytree_checkpoint_handler.py:577] save_device_host_concurrent_bytes=None I0415 21:14:49.030230 125104306547840 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 0x71c7877d7860>, enable_pinned_host_transfer=False, save_concurrent_bytes: 96000000000 (89.4 GiB), restore_concurrent_bytes: 96000000000 (89.4 GiB) I0415 21:14:51.415340 125104306547840 checkpointing.py:265] Enabling policy for fixed interval checkpointing. I0415 21:14:51.415554 125104306547840 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 0x71c10af3ce90>}, handler_registry=None I0415 21:14:51.415873 125104306547840 composite_checkpoint_handler.py:237] Deferred registration for item: "items". Adding handler `<orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeCheckpointHandler object at 0x71c10af3ce90>` 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`. I0415 21:14:51.415917 125104306547840 composite_checkpoint_handler.py:237] Deferred registration for item: "metrics". Adding handler `<orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonCheckpointHandler object at 0x71c10af3e1e0>` 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`. I0415 21:14:51.415951 125104306547840 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 0x71c10af3ce90>, ('items', <class 'orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeRestoreArgs'>): <orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeCheckpointHandler object at 0x71c10af3ce90>, ('metrics', <class 'orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonSaveArgs'>): <orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonCheckpointHandler object at 0x71c10af3e1e0>, ('metrics', <class 'orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonRestoreArgs'>): <orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonCheckpointHandler object at 0x71c10af3e1e0>}). I0415 21:14:51.416304 125104306547840 abstract_checkpointer.py:35] orbax-checkpoint version: 0.11.28 I0415 21:14:51.416358 125104306547840 async_checkpointer.py:177] [process=0][thread=MainThread] Using barrier_sync_fn: <function get_barrier_sync_fn.<locals>.<lambda> at 0x71c10af35260> timeout: 600 secs and primary_host=0 for async checkpoint writes I0415 21:14:51.540738 125104306547840 checkpoint_manager.py:1788] Found 0 checkpoint steps in gs://wanglance-maxtext/linen_ckpt_main_20260415_211126/linen_main_20260415_211126_02_synthetic/checkpoints I0415 21:14:51.541020 125104306547840 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_main_20260415_211126/linen_main_20260415_211126_02_synthetic/checkpoints: <orbax.checkpoint.checkpoint_manager.CheckpointManager object at 0x71c2cc181580> I0415 21:14:51.541127 125104306547840 checkpointing.py:301] Checkpoint manager created! W0415 21:14:51.559109 3226711 pjrt_executable.cc:642] Assume version compatibility. PjRt-IFRT does not track XLA executable versions. I0415 21:14:51.745264 125104306547840 nnx_wrappers.py:437] Unknown Logical: bfloat16[8,2048,2048]....................................... ('activation_batch', 'activation_norm_length', 'activation_embed'). I0415 21:14:51.745384 125104306547840 nnx_wrappers.py:437] Unknown Physical: bfloat16[8,2048,2048]....................................... ('fsdp', None, None). I0415 21:14:51.837209 125104306547840 attentions.py:1088] attentions/inputs_q Logical: bfloat16[8,2048,2048]....................................... ('activation_batch', 'activation_attn_length', 'activation_attn_embed'). I0415 21:14:51.837294 125104306547840 attentions.py:1088] attentions/inputs_q Physical: bfloat16[8,2048,2048]....................................... ('fsdp', None, None). I0415 21:14:51.851651 125104306547840 attentions.py:1089] attentions/inputs_kv Logical: bfloat16[8,2048,2048]....................................... ('activation_batch', 'activation_attn_length', 'activation_attn_embed'). I0415 21:14:51.851703 125104306547840 attentions.py:1089] attentions/inputs_kv Physical: bfloat16[8,2048,2048]....................................... ('fsdp', None, None). I0415 21:14:51.872569 125104306547840 attentions.py:1154] attentions/query Logical: bfloat16[8,2048,16,128]..................................... ('activation_kv_batch', 'activation_attn_length', 'activation_kv_heads', 'activation_kv_head_dim'). I0415 21:14:51.872637 125104306547840 attentions.py:1154] attentions/query Physical: bfloat16[8,2048,16,128]..................................... ('fsdp', None, None, None). I0415 21:14:51.886972 125104306547840 attentions.py:1155] attentions/key Logical: bfloat16[8,2048,16,128]..................................... ('activation_kv_batch', 'activation_attn_length', 'activation_kv_heads', 'activation_kv_head_dim'). I0415 21:14:51.887023 125104306547840 attentions.py:1155] attentions/key Physical: bfloat16[8,2048,16,128]..................................... ('fsdp', None, None, None). I0415 21:14:51.901399 125104306547840 attentions.py:1156] attentions/value Logical: bfloat16[8,2048,16,128]..................................... ('activation_kv_batch', 'activation_attn_length', 'activation_kv_heads', 'activation_kv_head_dim'). I0415 21:14:51.901448 125104306547840 attentions.py:1156] attentions/value Physical: bfloat16[8,2048,16,128]..................................... ('fsdp', None, None, None). I0415 21:14:51.923287 125104306547840 attentions.py:1197] attentions/out Logical: bfloat16[8,2048,16,128]..................................... ('activation_batch', 'activation_attn_length', 'activation_heads', 'activation_kv'). I0415 21:14:51.923346 125104306547840 attentions.py:1197] attentions/out Physical: bfloat16[8,2048,16,128]..................................... ('fsdp', None, None, None). I0415 21:14:51.942869 125104306547840 linears.py:525] linears/x Logical: bfloat16[8,2048,7168]....................................... ('activation_batch', 'activation_length', 'activation_mlp'). I0415 21:14:51.942930 125104306547840 linears.py:525] linears/x Physical: bfloat16[8,2048,7168]....................................... ('fsdp', None, None). I0415 21:14:52.127954 125104306547840 checkpointing.py:577] checkpoint manager exists so trying to load this run's existing checkpoint I0415 21:14:52.128078 125104306547840 checkpointing.py:665] No existing checkpoints found, not restoring checkpoint. W0415 21:14:52.179933 3226711 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 I0415 21:14:52.564379 125104306547840 maxtext_utils.py:1620] params/params/decoder/decoder_norm/scale Shape: float32[2048] Logical: P('norm',) Physical: (None,) I0415 21:14:52.564490 125104306547840 maxtext_utils.py:1620] params/params/decoder/layers/mlp/wi_0/kernel Shape: float32[2048,16,7168] Logical: P('embed', 'layers', 'mlp') Physical: ('fsdp', None, None) I0415 21:14:52.564530 125104306547840 maxtext_utils.py:1620] params/params/decoder/layers/mlp/wi_1/kernel Shape: float32[2048,16,7168] Logical: P('embed', 'layers', 'mlp') Physical: ('fsdp', None, None) I0415 21:14:52.564573 125104306547840 maxtext_utils.py:1620] params/params/decoder/layers/mlp/wo/kernel Shape: float32[7168,16,2048] Logical: P('mlp', 'layers', 'embed') Physical: (None, None, 'fsdp') I0415 21:14:52.564611 125104306547840 maxtext_utils.py:1620] params/params/decoder/layers/post_self_attention_layer_norm/scale Shape: float32[2048,16] Logical: P('norm', 'layers') Physical: (None, None) I0415 21:14:52.564639 125104306547840 maxtext_utils.py:1620] params/params/decoder/layers/pre_self_attention_layer_norm/scale Shape: float32[2048,16] Logical: P('norm', 'layers') Physical: (None, None) I0415 21:14:52.564677 125104306547840 maxtext_utils.py:1620] 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) I0415 21:14:52.564715 125104306547840 maxtext_utils.py:1620] 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') I0415 21:14:52.564744 125104306547840 maxtext_utils.py:1620] 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) I0415 21:14:52.564770 125104306547840 maxtext_utils.py:1620] 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) I0415 21:14:52.564805 125104306547840 maxtext_utils.py:1620] params/params/decoder/logits_dense/kernel Shape: float32[2048,32000] Logical: P('embed_vocab', 'vocab') Physical: ('fsdp', None) I0415 21:14:52.564839 125104306547840 maxtext_utils.py:1620] params/params/token_embedder/embedding Shape: float32[32000,2048] Logical: P('vocab', 'embed_vocab') Physical: (None, 'fsdp') I0415 21:14:52.991935 125104306547840 train.py:155] train/xent Logical: float32[8,2048]............................................. ('activation_embed_and_logits_batch', 'activation_length'). I0415 21:14:52.992027 125104306547840 train.py:155] train/xent Physical: float32[8,2048]............................................. ('fsdp', None). I0415 21:14:53.005634 125104306547840 train.py:162] train/z_loss Logical: float32[8,2048]............................................. ('activation_embed_and_logits_batch', 'activation_length'). I0415 21:14:53.005685 125104306547840 train.py:162] train/z_loss Physical: float32[8,2048]............................................. ('fsdp', None). W0415 21:14:53.651129 3226711 pjrt_executable.cc:642] Assume version compatibility. PjRt-IFRT does not track XLA executable versions. I0415 21:14:53.778494 125104306547840 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. I0415 21:14:53.779028 125104306547840 max_utils.py:194] tensorboardX not available; using no-op SummaryWriter. I0415 21:14:53.779183 125104306547840 metric_logger.py:289] number parameters: 1.104 billion W0415 21:14:54.625956 3226711 pjrt_executable.cc:642] Assume version compatibility. PjRt-IFRT does not track XLA executable versions. I0415 21:14:54.721431 125104306547840 checkpointing.py:772] Waiting for step 0 to finish before checkpoint... I0415 21:14:54.838824 125104306547840 checkpointing.py:776] Waited 0.11737251281738281 seconds for step 0 to finish before starting checkpointing. I0415 21:14:54.839345 125104306547840 checkpoint_manager.py:1983] [process=0][thread=MainThread][wait_until_finished] No Save Finalize thread to wait for. Returning. I0415 21:14:54.839502 125104306547840 checkpoint_manager.py:1501] [process=0] Saving checkpoint at step 0 I0415 21:14:54.839912 125104306547840 async_checkpointer.py:452] [process=0] Started async saving checkpoint to gs://wanglance-maxtext/linen_ckpt_main_20260415_211126/linen_main_20260415_211126_02_synthetic/checkpoints/0. I0415 21:14:54.926788 125104306547840 signaling_client.py:373] Using ThreadSafeKeyValueSignalingClient I0415 21:14:55.014494 124996884104768 atomicity.py:137] Creating tmp directory gs://wanglance-maxtext/linen_ckpt_main_20260415_211126/linen_main_20260415_211126_02_synthetic/checkpoints/0 I0415 21:14:55.019958 125104306547840 jax_array_handlers.py:347] Scheduling D2H of 39 prioritized jax.Array. I0415 21:14:55.020081 125104306547840 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 W0415 21:14:55.245614 3226711 pjrt_executable.cc:642] Assume version compatibility. PjRt-IFRT does not track XLA executable versions. W0415 21:14:55.259341 3226711 pjrt_executable.cc:642] Assume version compatibility. PjRt-IFRT does not track XLA executable versions. W0415 21:14:55.267005 3226711 pjrt_executable.cc:642] Assume version compatibility. PjRt-IFRT does not track XLA executable versions. W0415 21:14:55.274721 3226711 pjrt_executable.cc:642] Assume version compatibility. PjRt-IFRT does not track XLA executable versions. W0415 21:14:55.291748 3226711 pjrt_executable.cc:642] Assume version compatibility. PjRt-IFRT does not track XLA executable versions. W0415 21:14:55.299028 3226711 pjrt_executable.cc:642] Assume version compatibility. PjRt-IFRT does not track XLA executable versions. W0415 21:14:55.303386 3226711 pjrt_executable.cc:642] Assume version compatibility. PjRt-IFRT does not track XLA executable versions. W0415 21:14:55.308009 3226711 pjrt_executable.cc:642] Assume version compatibility. PjRt-IFRT does not track XLA executable versions. I0415 21:14:55.725949 124996873619008 atomicity.py:137] Creating tmp directory gs://wanglance-maxtext/linen_ckpt_main_20260415_211126/linen_main_20260415_211126_02_synthetic/checkpoints/0/items I0415 21:14:55.901255 124996837967424 checkpoint.py:188] Wrote Metadata={'item_handlers': None, 'metrics': {}, 'performance_metrics': {}, 'init_timestamp_nsecs': 1776287695630194654, 'commit_timestamp_nsecs': None, 'custom_metadata': {}}, json={"item_handlers": null, "metrics": {}, "performance_metrics": {}, "init_timestamp_nsecs": 1776287695630194654, "commit_timestamp_nsecs": null, "custom_metadata": {}} to gs://wanglance-maxtext/linen_ckpt_main_20260415_211126/linen_main_20260415_211126_02_synthetic/checkpoints/0/_CHECKPOINT_METADATA I0415 21:14:56.084562 125104306547840 base_pytree_checkpoint_handler.py:153] [process=0][thread=MainThread] Initiated "orbax.checkpoint._src.serialization.jax_array_handlers.ArrayHandler".serialize. Time taken: 1.065083s I0415 21:14:56.085063 125104306547840 base_pytree_checkpoint_handler.py:128] [process=0] /jax/checkpoint/write/blocking_gbytes_per_sec: 10.668 GiB/s (total gbytes: 12.3 GiB) (time elapsed: a second) (per-host) I0415 21:14:56.085116 125104306547840 base_pytree_checkpoint_handler.py:732] [process=0][thread=MainThread] Initiated Pytree async_save. Time taken: 1.156856s (batch_requests_ready=0.087410s, total_serialization_initiated=1.069063s, others=0.000382s) I0415 21:14:56.085184 125104306547840 composite_checkpoint_handler.py:715] [process=0][thread=MainThread] Initiated CompositeCheckpointHandler.async_save. Time taken: 1.157820s (all_items=0.000019s, per_item={'items': '0.00001931'}, temp_paths=1.157801) I0415 21:14:56.086078 124996816995904 async_checkpointer.py:79] [process=0][thread=async_save] Background save thread started. I0415 21:14:56.086190 125104306547840 async_checkpointer.py:561] Finished blocking save. Time taken: 1.246649s. Continuing background save to gs://wanglance-maxtext/linen_ckpt_main_20260415_211126/linen_main_20260415_211126_02_synthetic/checkpoints/0. I0415 21:14:56.086385 125104306547840 checkpoint_manager.py:1549] [process=0][thread=MainThread][step=0] Starting CheckpointManager Save Finalize thread=save_finalize I0415 21:14:56.086682 124996894590528 async_checkpointer.py:265] [process=0][thread=save_finalize] Waiting for background save thread=async_save. I0415 21:14:56.086801 125104306547840 standard_logger.py:34] {'step': 0, 'event_type': 'save', 'directory': 'gs://wanglance-maxtext/linen_ckpt_main_20260415_211126/linen_main_20260415_211126_02_synthetic/checkpoints', 'reached_preemption': False, 'preemption_received_at': None, 'synchronous': False, 'wait_for_prev_start_time': 1776287694.839332, 'wait_for_prev_duration_secs': 4.2438507080078125e-05, 'time_between_consecutive_saves_sec': None, 'checkpointer_blocking_start_time': 1776287694.8395228, 'checkpointer_blocking_duration_secs': 1.2467832565307617, 'get_old_steps_start_time': 1776287696.0863295, 'get_old_steps_duration_secs': 2.2172927856445312e-05, 'checkpoint_manager_blocking_start_time': 1776287694.8392425, 'checkpoint_manager_blocking_duration_secs': 1.2475333213806152} I0415 21:14:56.086927 125104306547840 checkpointing.py:408] Started an asynchronous checkpoint save for step 0 I0415 21:14:56.086983 125104306547840 max_utils.py:750] Memstats: After params initialized: I0415 21:14:56.087071 125104306547840 max_utils.py:756] Using (GB) 1.59 / 31.25 (5.088000%) on TPU_0(process=0,(0,0,0,0)) I0415 21:14:56.087098 125104306547840 max_utils.py:756] Using (GB) 1.59 / 31.25 (5.088000%) on TPU_1(process=0,(1,0,0,0)) I0415 21:14:56.087117 125104306547840 max_utils.py:756] Using (GB) 1.59 / 31.25 (5.088000%) on TPU_2(process=0,(0,1,0,0)) I0415 21:14:56.087134 125104306547840 max_utils.py:756] Using (GB) 1.59 / 31.25 (5.088000%) on TPU_3(process=0,(1,1,0,0)) I0415 21:14:56.087150 125104306547840 max_utils.py:756] Using (GB) 1.59 / 31.25 (5.088000%) on TPU_4(process=0,(0,2,0,0)) I0415 21:14:56.087167 125104306547840 max_utils.py:756] Using (GB) 1.59 / 31.25 (5.088000%) on TPU_5(process=0,(1,2,0,0)) I0415 21:14:56.087182 125104306547840 max_utils.py:756] Using (GB) 1.59 / 31.25 (5.088000%) on TPU_6(process=0,(0,3,0,0)) I0415 21:14:56.087198 125104306547840 max_utils.py:756] Using (GB) 1.59 / 31.25 (5.088000%) on TPU_7(process=0,(1,3,0,0)) W0415 21:14:56.093444 3226711 pjrt_executable.cc:642] Assume version compatibility. PjRt-IFRT does not track XLA executable versions. W0415 21:14:56.099357 3226711 pjrt_executable.cc:642] Assume version compatibility. PjRt-IFRT does not track XLA executable versions. W0415 21:14:56.104400 3226711 pjrt_executable.cc:642] Assume version compatibility. PjRt-IFRT does not track XLA executable versions. I0415 21:14:56.204528 3228888 google_auth_provider.cc:149] Using credentials at ~/.config/gcloud/application_default_credentials.json I0415 21:14:56.204618 3228888 google_auth_provider.cc:156] Using OAuth2 AuthProvider I0415 21:14:56.475031 125104306547840 metric_logger.py:185] completed step: 0, seconds: 0.942, TFLOP/s/device: 14.427, Tokens/s/device: 2174.622, total_weights: 16384, loss: 10.874 I0415 21:14:56.476090 125104306547840 metric_logger.py:269] To see full metrics 'tensorboard --logdir=gs://wanglance-maxtext/linen_ckpt_main_20260415_211126/linen_main_20260415_211126_02_synthetic/tensorboard/' I0415 21:14:56.591552 125104306547840 metric_logger.py:185] completed step: 1, seconds: 1.752, TFLOP/s/device: 7.757, Tokens/s/device: 1169.211, total_weights: 16384, loss: 10.874 I0415 21:14:56.710727 125104306547840 metric_logger.py:185] completed step: 2, seconds: 0.007, TFLOP/s/device: 1965.446, Tokens/s/device: 296253.436, total_weights: 16384, loss: 9.775 I0415 21:14:56.829605 125104306547840 metric_logger.py:185] completed step: 3, seconds: 0.118, TFLOP/s/device: 115.629, Tokens/s/device: 17428.897, total_weights: 16384, loss: 8.838 I0415 21:14:57.190747 124996884104768 array_metadata_store.py:203] [process=0][thread=array_type_handler] Wrote 39 array_metadata.ArrayMetadata to gs://wanglance-maxtext/linen_ckpt_main_20260415_211126/linen_main_20260415_211126_02_synthetic/checkpoints/0/items/array_metadatas/process_0 I0415 21:15:16.716282 125104306547840 metric_logger.py:185] completed step: 4, seconds: 0.119, TFLOP/s/device: 114.019, Tokens/s/device: 17186.254, total_weights: 16384, loss: 7.965 I0415 21:15:16.723850 125104306547840 metric_logger.py:185] completed step: 5, seconds: 0.119, TFLOP/s/device: 114.246, Tokens/s/device: 17220.358, total_weights: 16384, loss: 7.233 I0415 21:15:16.841177 125104306547840 metric_logger.py:185] completed step: 6, seconds: 19.887, TFLOP/s/device: 0.683, Tokens/s/device: 102.980, total_weights: 16384, loss: 6.676 I0415 21:15:16.959825 125104306547840 metric_logger.py:185] completed step: 7, seconds: 0.005, TFLOP/s/device: 2867.693, Tokens/s/device: 432249.894, total_weights: 16384, loss: 6.293 I0415 21:15:17.078507 125104306547840 metric_logger.py:185] completed step: 8, seconds: 0.118, TFLOP/s/device: 114.826, Tokens/s/device: 17307.822, total_weights: 16384, loss: 6.055 I0415 21:15:17.196362 125104306547840 checkpointing.py:772] Waiting for step 9 to finish before checkpoint... I0415 21:15:17.197253 125104306547840 checkpointing.py:776] Waited 0.0009205341339111328 seconds for step 9 to finish before starting checkpointing. I0415 21:15:17.197555 125104306547840 checkpoint_manager.py:1994] [process=0][thread=MainThread][step=0][wait_until_finished] Waiting for Save Finalize thread (save_finalize) to complete. I0415 21:15:19.740896 124996827481664 base_pytree_checkpoint_handler.py:1217] [process=0][thread=write_metadata_after_commits] Commit + Array metadata written. Time taken: 23.533303s (commit=23.071575s, array_metadata_write=0.461728s) I0415 21:15:19.742200 124996816995904 base_pytree_checkpoint_handler.py:128] [process=0] /jax/checkpoint/write/gbytes_per_sec: 509.263 MiB/s (total gbytes: 12.3 GiB) (time elapsed: 24 seconds) (per-host) I0415 21:15:19.742316 124996816995904 async_checkpointer.py:90] [process=0][thread=async_save] 3 Handler Commit operations completed. Time taken: 23.656079s. I0415 21:15:19.964223 124996816995904 checkpoint.py:228] Read Metadata={'item_handlers': None, 'metrics': {}, 'performance_metrics': {}, 'init_timestamp_nsecs': 1776287695630194654, 'commit_timestamp_nsecs': None, 'custom_metadata': {}} from gs://wanglance-maxtext/linen_ckpt_main_20260415_211126/linen_main_20260415_211126_02_synthetic/checkpoints/0/_CHECKPOINT_METADATA I0415 21:15:20.145598 124996816995904 array_metadata_store.py:367] [process=0][thread=async_save] Skipped cross-host ArrayMetadata validation because only one process is found: process_index=0. I0415 21:15:20.364684 124996837967424 checkpoint.py:247] Updated Metadata={'item_handlers': {'items': 'orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeCheckpointHandler'}, 'metrics': {}, 'performance_metrics': {}, 'init_timestamp_nsecs': 1776287695630194654, 'commit_timestamp_nsecs': None, 'custom_metadata': {}} to gs://wanglance-maxtext/linen_ckpt_main_20260415_211126/linen_main_20260415_211126_02_synthetic/checkpoints/0/_CHECKPOINT_METADATA I0415 21:15:20.603388 124996816995904 ocdbt_utils.py:56] Param validation support for Zarr3 will be added later (b/362328389). I0415 21:15:20.604031 124996816995904 base_pytree_checkpoint_handler.py:1342] [process=0][thread=async_save] Pytree save finalize (merge_ocdbt + ArrayMetadata validation) completed. Time taken: 0.609719s. use_zarr3=True, enable_post_merge_validation=True, directory=gs://wanglance-maxtext/linen_ckpt_main_20260415_211126/linen_main_20260415_211126_02_synthetic/checkpoints/0/items I0415 21:15:20.604857 124996816995904 atomicity.py:608] Finalizing gs://wanglance-maxtext/linen_ckpt_main_20260415_211126/linen_main_20260415_211126_02_synthetic/checkpoints/0/items I0415 21:15:20.865710 124996816995904 atomicity.py:608] Finalizing gs://wanglance-maxtext/linen_ckpt_main_20260415_211126/linen_main_20260415_211126_02_synthetic/checkpoints/0 I0415 21:15:21.564556 124996816995904 atomicity.py:794] [process=0][thread=async_save] Finished saving checkpoint (finalized tmp dir) to `gs://wanglance-maxtext/linen_ckpt_main_20260415_211126/linen_main_20260415_211126_02_synthetic/checkpoints/0`. I0415 21:15:21.565346 124996816995904 async_checkpointer.py:420] Finished async_save (blocking + background). Time taken: 26.725811s. directory=gs://wanglance-maxtext/linen_ckpt_main_20260415_211126/linen_main_20260415_211126_02_synthetic/checkpoints/0 I0415 21:15:21.565428 124996816995904 async_checkpointer.py:144] [process=0][thread=async_save] Background save thread done. Time taken: 25.479194s. I0415 21:15:21.565598 124996894590528 async_checkpointer.py:273] [process=0][thread=save_finalize] Done with waiting for background save thread=async_save. I0415 21:15:21.565721 124996894590528 async_checkpointer.py:283] [process=0][thread=save_finalize] No errors found in background save thread=async_save. I0415 21:15:21.565779 124996894590528 checkpoint_manager.py:2103] [process=0][thread=save_finalize][step=0] CheckpointManager Save Finalize is syncing with other hosts... I0415 21:15:21.565827 124996894590528 checkpoint_manager.py:2112] [process=0][thread=save_finalize][step=0] CheckpointManager Save Finalize is done on all hosts. I0415 21:15:21.565984 125104306547840 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. W0415 21:15:21.566152 125104306547840 checkpoint_manager.py:1441] Waiting for previous save to complete took 4.368603 seconds. If this number is high, consider checkpointing less frequently. I0415 21:15:21.567028 125104306547840 checkpoint_manager.py:1501] [process=0] Saving checkpoint at step 9 I0415 21:15:21.567368 125104306547840 async_checkpointer.py:452] [process=0] Started async saving checkpoint to gs://wanglance-maxtext/linen_ckpt_main_20260415_211126/linen_main_20260415_211126_02_synthetic/checkpoints/9. I0415 21:15:21.752362 125104306547840 jax_array_handlers.py:347] Scheduling D2H of 39 prioritized jax.Array. I0415 21:15:21.752466 125104306547840 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 I0415 21:15:21.754418 124996894590528 atomicity.py:137] Creating tmp directory gs://wanglance-maxtext/linen_ckpt_main_20260415_211126/linen_main_20260415_211126_02_synthetic/checkpoints/9 I0415 21:15:22.471398 124996745692736 atomicity.py:137] Creating tmp directory gs://wanglance-maxtext/linen_ckpt_main_20260415_211126/linen_main_20260415_211126_02_synthetic/checkpoints/9/items I0415 21:15:30.332244 125104306547840 base_pytree_checkpoint_handler.py:153] [process=0][thread=MainThread] Initiated "orbax.checkpoint._src.serialization.jax_array_handlers.ArrayHandler".serialize. Time taken: 8.580375s I0415 21:15:30.337155 125104306547840 base_pytree_checkpoint_handler.py:128] [process=0] /jax/checkpoint/write/blocking_gbytes_per_sec: 1.422 GiB/s (total gbytes: 12.3 GiB) (time elapsed: 8 seconds) (per-host) I0415 21:15:30.337238 125104306547840 base_pytree_checkpoint_handler.py:732] [process=0][thread=MainThread] Initiated Pytree async_save. Time taken: 8.679564s (batch_requests_ready=0.091189s, total_serialization_initiated=8.583614s, others=0.004761s) I0415 21:15:30.337346 125104306547840 composite_checkpoint_handler.py:715] [process=0][thread=MainThread] Initiated CompositeCheckpointHandler.async_save. Time taken: 8.680225s (all_items=0.000034s, per_item={'items': '0.00003386'}, temp_paths=8.680192) I0415 21:15:30.339514 124996827481664 async_checkpointer.py:79] [process=0][thread=async_save] Background save thread started. I0415 21:15:30.339649 125104306547840 async_checkpointer.py:561] Finished blocking save. Time taken: 8.772559s. Continuing background save to gs://wanglance-maxtext/linen_ckpt_main_20260415_211126/linen_main_20260415_211126_02_synthetic/checkpoints/9. I0415 21:15:30.339931 125104306547840 checkpoint_manager.py:1549] [process=0][thread=MainThread][step=9] Starting CheckpointManager Save Finalize thread=save_finalize I0415 21:15:30.340149 124996816995904 async_checkpointer.py:265] [process=0][thread=save_finalize] Waiting for background save thread=async_save. I0415 21:15:30.340284 125104306547840 standard_logger.py:34] {'step': 9, 'event_type': 'save', 'directory': 'gs://wanglance-maxtext/linen_ckpt_main_20260415_211126/linen_main_20260415_211126_02_synthetic/checkpoints', 'reached_preemption': False, 'preemption_received_at': None, 'synchronous': False, 'wait_for_prev_start_time': 1776287717.1975126, 'wait_for_prev_duration_secs': 4.368603467941284, 'time_between_consecutive_saves_sec': None, 'checkpointer_blocking_start_time': 1776287721.5670636, 'checkpointer_blocking_duration_secs': 8.77273178100586, 'get_old_steps_start_time': 1776287730.3398209, 'get_old_steps_duration_secs': 5.054473876953125e-05, 'checkpoint_manager_blocking_start_time': 1776287717.197467, 'checkpoint_manager_blocking_duration_secs': 13.142783880233765} I0415 21:15:30.340435 125104306547840 checkpointing.py:408] Started an asynchronous checkpoint save for step 9 I0415 21:15:30.340475 125104306547840 checkpoint_manager.py:1994] [process=0][thread=MainThread][step=9][wait_until_finished] Waiting for Save Finalize thread (save_finalize) to complete. I0415 21:15:31.117439 124996884104768 array_metadata_store.py:203] [process=0][thread=array_type_handler] Wrote 39 array_metadata.ArrayMetadata to gs://wanglance-maxtext/linen_ckpt_main_20260415_211126/linen_main_20260415_211126_02_synthetic/checkpoints/9/items/array_metadatas/process_0 I0415 21:16:07.519540 124996745692736 base_pytree_checkpoint_handler.py:1217] [process=0][thread=write_metadata_after_commits] Commit + Array metadata written. Time taken: 37.181555s (commit=36.722147s, array_metadata_write=0.459408s) I0415 21:16:07.521075 124996827481664 base_pytree_checkpoint_handler.py:128] [process=0] /jax/checkpoint/write/gbytes_per_sec: 275.531 MiB/s (total gbytes: 12.3 GiB) (time elapsed: 45 seconds) (per-host) I0415 21:16:07.521227 124996827481664 async_checkpointer.py:90] [process=0][thread=async_save] 3 Handler Commit operations completed. Time taken: 37.181517s. I0415 21:16:07.955710 124996827481664 array_metadata_store.py:367] [process=0][thread=async_save] Skipped cross-host ArrayMetadata validation because only one process is found: process_index=0. I0415 21:16:08.388362 124996827481664 ocdbt_utils.py:56] Param validation support for Zarr3 will be added later (b/362328389). I0415 21:16:08.389276 124996827481664 base_pytree_checkpoint_handler.py:1342] [process=0][thread=async_save] Pytree save finalize (merge_ocdbt + ArrayMetadata validation) completed. Time taken: 0.592542s. use_zarr3=True, enable_post_merge_validation=True, directory=gs://wanglance-maxtext/linen_ckpt_main_20260415_211126/linen_main_20260415_211126_02_synthetic/checkpoints/9/items I0415 21:16:08.390101 124996827481664 atomicity.py:608] Finalizing gs://wanglance-maxtext/linen_ckpt_main_20260415_211126/linen_main_20260415_211126_02_synthetic/checkpoints/9/items I0415 21:16:08.641376 124996827481664 atomicity.py:608] Finalizing gs://wanglance-maxtext/linen_ckpt_main_20260415_211126/linen_main_20260415_211126_02_synthetic/checkpoints/9 I0415 21:16:09.367434 124996827481664 atomicity.py:794] [process=0][thread=async_save] Finished saving checkpoint (finalized tmp dir) to `gs://wanglance-maxtext/linen_ckpt_main_20260415_211126/linen_main_20260415_211126_02_synthetic/checkpoints/9`. I0415 21:16:09.368214 124996827481664 async_checkpointer.py:420] Finished async_save (blocking + background). Time taken: 47.801133s. directory=gs://wanglance-maxtext/linen_ckpt_main_20260415_211126/linen_main_20260415_211126_02_synthetic/checkpoints/9 I0415 21:16:09.368305 124996827481664 async_checkpointer.py:144] [process=0][thread=async_save] Background save thread done. Time taken: 39.028602s. I0415 21:16:09.368594 124996816995904 async_checkpointer.py:273] [process=0][thread=save_finalize] Done with waiting for background save thread=async_save. I0415 21:16:09.368745 124996816995904 async_checkpointer.py:283] [process=0][thread=save_finalize] No errors found in background save thread=async_save. I0415 21:16:09.368826 124996816995904 checkpoint_manager.py:2103] [process=0][thread=save_finalize][step=9] CheckpointManager Save Finalize is syncing with other hosts... I0415 21:16:09.368875 124996816995904 checkpoint_manager.py:2112] [process=0][thread=save_finalize][step=9] CheckpointManager Save Finalize is done on all hosts. I0415 21:16:09.369447 125104306547840 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. I0415 21:16:09.369626 125104306547840 checkpoint_manager.py:1983] [process=0][thread=MainThread][wait_until_finished] No Save Finalize thread to wait for. Returning. I0415 21:16:09.370739 125104306547840 metric_logger.py:185] completed step: 9, seconds: 0.119, TFLOP/s/device: 114.570, Tokens/s/device: 17269.293, 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
2026-04-15 23:10:41.109894: 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) I0415 23:10:41.228703 124327861365888 max_utils.py:238] Skipping jax distributed system due to skip_jax_distributed_system=True flag. I0415 23:11:28.136871 124327861365888 max_utils.py:800] System Information: Jax Version: 0.9.2 I0415 23:11:28.136990 124327861365888 max_utils.py:801] System Information: Jaxlib Version: 0.9.2 I0415 23:11:28.137024 124327861365888 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 I0415 23:11:28.137064 124327861365888 train_utils.py:334] WARNING: 'dataset_path' might be pointing your local file system I0415 23:11:28.137088 124327861365888 train_utils.py:347] WARNING: Sequence packing is essentially ignored for synthetic data. Please use a real dataset to use sequence packing. I0415 23:11:28.137169 124327861365888 train.py:715] [DECOUPLED NO-OP] skipping cloud diagnostics wrapper. W0415 23:11:28.229713 3351818 pjrt_executable.cc:642] Assume version compatibility. PjRt-IFRT does not track XLA executable versions. I0415 23:11:28.637109 124327861365888 maxtext_utils.py:1520] Num_devices: 8, shape (1, 1, 1, 8, 1, 1, 1, 1, 1, 1, 1, 1, 1) I0415 23:11:28.637447 124327861365888 checkpointing.py:677] Setting up checkpoint logger... I0415 23:11:28.637492 124327861365888 checkpointing.py:233] Creating checkpoint manager with ocdbt=True and zarr3=True I0415 23:11:28.637532 124327861365888 pytree_checkpoint_handler.py:577] save_device_host_concurrent_bytes=None I0415 23:11:28.638104 124327861365888 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 0x7112bf100a70>, enable_pinned_host_transfer=False, save_concurrent_bytes: 96000000000 (89.4 GiB), restore_concurrent_bytes: 96000000000 (89.4 GiB) I0415 23:11:31.115347 124327861365888 checkpointing.py:265] Enabling policy for fixed interval checkpointing. I0415 23:11:31.115517 124327861365888 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 0x710c41388530>}, handler_registry=None I0415 23:11:31.115772 124327861365888 composite_checkpoint_handler.py:237] Deferred registration for item: "items". Adding handler `<orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeCheckpointHandler object at 0x710c41388530>` 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`. I0415 23:11:31.115811 124327861365888 composite_checkpoint_handler.py:237] Deferred registration for item: "metrics". Adding handler `<orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonCheckpointHandler object at 0x710c40f4fc20>` 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`. I0415 23:11:31.115840 124327861365888 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 0x710c41388530>, ('items', <class 'orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeRestoreArgs'>): <orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeCheckpointHandler object at 0x710c41388530>, ('metrics', <class 'orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonSaveArgs'>): <orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonCheckpointHandler object at 0x710c40f4fc20>, ('metrics', <class 'orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonRestoreArgs'>): <orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonCheckpointHandler object at 0x710c40f4fc20>}). I0415 23:11:31.116131 124327861365888 abstract_checkpointer.py:35] orbax-checkpoint version: 0.11.28 I0415 23:11:31.116179 124327861365888 async_checkpointer.py:177] [process=0][thread=MainThread] Using barrier_sync_fn: <function get_barrier_sync_fn.<locals>.<lambda> at 0x710c40b394e0> timeout: 600 secs and primary_host=0 for async checkpoint writes I0415 23:11:31.249433 124327861365888 checkpoint_manager.py:1788] Found 0 checkpoint steps in gs://wanglance-maxtext/linen_ckpt_feat_migrate_nnx_utils_20260415_230805/linen_feat_migrate_nnx_utils_20260415_230805_02_synthetic/checkpoints I0415 23:11:31.249677 124327861365888 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_migrate_nnx_utils_20260415_230805/linen_feat_migrate_nnx_utils_20260415_230805_02_synthetic/checkpoints: <orbax.checkpoint.checkpoint_manager.CheckpointManager object at 0x710c41718680> I0415 23:11:31.249756 124327861365888 checkpointing.py:301] Checkpoint manager created! W0415 23:11:31.265838 3351818 pjrt_executable.cc:642] Assume version compatibility. PjRt-IFRT does not track XLA executable versions. I0415 23:11:31.448316 124327861365888 nnx_wrappers.py:437] Unknown Logical: bfloat16[8,2048,2048]....................................... ('activation_batch', 'activation_norm_length', 'activation_embed'). I0415 23:11:31.448415 124327861365888 nnx_wrappers.py:437] Unknown Physical: bfloat16[8,2048,2048]....................................... ('fsdp', None, None). I0415 23:11:31.542965 124327861365888 attentions.py:1088] attentions/inputs_q Logical: bfloat16[8,2048,2048]....................................... ('activation_batch', 'activation_attn_length', 'activation_attn_embed'). I0415 23:11:31.543037 124327861365888 attentions.py:1088] attentions/inputs_q Physical: bfloat16[8,2048,2048]....................................... ('fsdp', None, None). I0415 23:11:31.557181 124327861365888 attentions.py:1089] attentions/inputs_kv Logical: bfloat16[8,2048,2048]....................................... ('activation_batch', 'activation_attn_length', 'activation_attn_embed'). I0415 23:11:31.557226 124327861365888 attentions.py:1089] attentions/inputs_kv Physical: bfloat16[8,2048,2048]....................................... ('fsdp', None, None). I0415 23:11:31.577630 124327861365888 attentions.py:1154] attentions/query Logical: bfloat16[8,2048,16,128]..................................... ('activation_kv_batch', 'activation_attn_length', 'activation_kv_heads', 'activation_kv_head_dim'). I0415 23:11:31.577686 124327861365888 attentions.py:1154] attentions/query Physical: bfloat16[8,2048,16,128]..................................... ('fsdp', None, None, None). I0415 23:11:31.591834 124327861365888 attentions.py:1155] attentions/key Logical: bfloat16[8,2048,16,128]..................................... ('activation_kv_batch', 'activation_attn_length', 'activation_kv_heads', 'activation_kv_head_dim'). I0415 23:11:31.591882 124327861365888 attentions.py:1155] attentions/key Physical: bfloat16[8,2048,16,128]..................................... ('fsdp', None, None, None). I0415 23:11:31.605996 124327861365888 attentions.py:1156] attentions/value Logical: bfloat16[8,2048,16,128]..................................... ('activation_kv_batch', 'activation_attn_length', 'activation_kv_heads', 'activation_kv_head_dim'). I0415 23:11:31.606056 124327861365888 attentions.py:1156] attentions/value Physical: bfloat16[8,2048,16,128]..................................... ('fsdp', None, None, None). I0415 23:11:31.627892 124327861365888 attentions.py:1197] attentions/out Logical: bfloat16[8,2048,16,128]..................................... ('activation_batch', 'activation_attn_length', 'activation_heads', 'activation_kv'). I0415 23:11:31.627946 124327861365888 attentions.py:1197] attentions/out Physical: bfloat16[8,2048,16,128]..................................... ('fsdp', None, None, None). I0415 23:11:31.646228 124327861365888 linears.py:525] linears/x Logical: bfloat16[8,2048,7168]....................................... ('activation_batch', 'activation_length', 'activation_mlp'). I0415 23:11:31.646277 124327861365888 linears.py:525] linears/x Physical: bfloat16[8,2048,7168]....................................... ('fsdp', None, None). I0415 23:11:31.838849 124327861365888 checkpointing.py:577] checkpoint manager exists so trying to load this run's existing checkpoint I0415 23:11:31.838964 124327861365888 checkpointing.py:665] No existing checkpoints found, not restoring checkpoint. W0415 23:11:31.889768 3351818 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 I0415 23:11:32.277753 124327861365888 maxtext_utils.py:1623] params/params/decoder/decoder_norm/scale Shape: float32[2048] Logical: P('norm',) Physical: (None,) I0415 23:11:32.277922 124327861365888 maxtext_utils.py:1623] params/params/decoder/layers/mlp/wi_0/kernel Shape: float32[2048,16,7168] Logical: P('embed', 'layers', 'mlp') Physical: ('fsdp', None, None) I0415 23:11:32.277963 124327861365888 maxtext_utils.py:1623] params/params/decoder/layers/mlp/wi_1/kernel Shape: float32[2048,16,7168] Logical: P('embed', 'layers', 'mlp') Physical: ('fsdp', None, None) I0415 23:11:32.278006 124327861365888 maxtext_utils.py:1623] params/params/decoder/layers/mlp/wo/kernel Shape: float32[7168,16,2048] Logical: P('mlp', 'layers', 'embed') Physical: (None, None, 'fsdp') I0415 23:11:32.278059 124327861365888 maxtext_utils.py:1623] params/params/decoder/layers/post_self_attention_layer_norm/scale Shape: float32[2048,16] Logical: P('norm', 'layers') Physical: (None, None) I0415 23:11:32.278090 124327861365888 maxtext_utils.py:1623] params/params/decoder/layers/pre_self_attention_layer_norm/scale Shape: float32[2048,16] Logical: P('norm', 'layers') Physical: (None, None) I0415 23:11:32.278134 124327861365888 maxtext_utils.py:1623] 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) I0415 23:11:32.278173 124327861365888 maxtext_utils.py:1623] 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') I0415 23:11:32.278201 124327861365888 maxtext_utils.py:1623] 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) I0415 23:11:32.278226 124327861365888 maxtext_utils.py:1623] 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) I0415 23:11:32.278260 124327861365888 maxtext_utils.py:1623] params/params/decoder/logits_dense/kernel Shape: float32[2048,32000] Logical: P('embed', 'vocab') Physical: ('fsdp', None) I0415 23:11:32.278293 124327861365888 maxtext_utils.py:1623] params/params/token_embedder/embedding Shape: float32[32000,2048] Logical: P('vocab', 'embed') Physical: (None, 'fsdp') I0415 23:11:32.700338 124327861365888 train.py:155] train/xent Logical: float32[8,2048]............................................. ('activation_embed_and_logits_batch', 'activation_length'). I0415 23:11:32.700417 124327861365888 train.py:155] train/xent Physical: float32[8,2048]............................................. ('fsdp', None). I0415 23:11:32.713933 124327861365888 train.py:162] train/z_loss Logical: float32[8,2048]............................................. ('activation_embed_and_logits_batch', 'activation_length'). I0415 23:11:32.713983 124327861365888 train.py:162] train/z_loss Physical: float32[8,2048]............................................. ('fsdp', None). W0415 23:11:33.398396 3351818 pjrt_executable.cc:642] Assume version compatibility. PjRt-IFRT does not track XLA executable versions. I0415 23:11:33.521283 124327861365888 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. I0415 23:11:33.521727 124327861365888 max_utils.py:194] tensorboardX not available; using no-op SummaryWriter. I0415 23:11:33.521864 124327861365888 metric_logger.py:289] number parameters: 1.104 billion W0415 23:11:34.373845 3351818 pjrt_executable.cc:642] Assume version compatibility. PjRt-IFRT does not track XLA executable versions. I0415 23:11:34.472615 124327861365888 checkpointing.py:772] Waiting for step 0 to finish before checkpoint... I0415 23:11:34.589920 124327861365888 checkpointing.py:776] Waited 0.1172943115234375 seconds for step 0 to finish before starting checkpointing. I0415 23:11:34.590344 124327861365888 checkpoint_manager.py:1983] [process=0][thread=MainThread][wait_until_finished] No Save Finalize thread to wait for. Returning. I0415 23:11:34.590491 124327861365888 checkpoint_manager.py:1501] [process=0] Saving checkpoint at step 0 I0415 23:11:34.590864 124327861365888 async_checkpointer.py:452] [process=0] Started async saving checkpoint to gs://wanglance-maxtext/linen_ckpt_feat_migrate_nnx_utils_20260415_230805/linen_feat_migrate_nnx_utils_20260415_230805_02_synthetic/checkpoints/0. I0415 23:11:34.681916 124327861365888 signaling_client.py:373] Using ThreadSafeKeyValueSignalingClient I0415 23:11:34.770944 124327861365888 jax_array_handlers.py:347] Scheduling D2H of 39 prioritized jax.Array. I0415 23:11:34.771028 124327861365888 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 I0415 23:11:34.776736 124220308719168 atomicity.py:137] Creating tmp directory gs://wanglance-maxtext/linen_ckpt_feat_migrate_nnx_utils_20260415_230805/linen_feat_migrate_nnx_utils_20260415_230805_02_synthetic/checkpoints/0 W0415 23:11:34.984201 3351818 pjrt_executable.cc:642] Assume version compatibility. PjRt-IFRT does not track XLA executable versions. W0415 23:11:34.997227 3351818 pjrt_executable.cc:642] Assume version compatibility. PjRt-IFRT does not track XLA executable versions. W0415 23:11:35.006776 3351818 pjrt_executable.cc:642] Assume version compatibility. PjRt-IFRT does not track XLA executable versions. W0415 23:11:35.024559 3351818 pjrt_executable.cc:642] Assume version compatibility. PjRt-IFRT does not track XLA executable versions. W0415 23:11:35.038185 3351818 pjrt_executable.cc:642] Assume version compatibility. PjRt-IFRT does not track XLA executable versions. W0415 23:11:35.044579 3351818 pjrt_executable.cc:642] Assume version compatibility. PjRt-IFRT does not track XLA executable versions. W0415 23:11:35.050013 3351818 pjrt_executable.cc:642] Assume version compatibility. PjRt-IFRT does not track XLA executable versions. W0415 23:11:35.056145 3351818 pjrt_executable.cc:642] Assume version compatibility. PjRt-IFRT does not track XLA executable versions. I0415 23:11:35.463701 124327861365888 base_pytree_checkpoint_handler.py:153] [process=0][thread=MainThread] Initiated "orbax.checkpoint._src.serialization.jax_array_handlers.ArrayHandler".serialize. Time taken: 0.693182s I0415 23:11:35.464267 124327861365888 base_pytree_checkpoint_handler.py:128] [process=0] /jax/checkpoint/write/blocking_gbytes_per_sec: 15.803 GiB/s (total gbytes: 12.3 GiB) (time elapsed: 780 milliseconds) (per-host) I0415 23:11:35.464423 124327861365888 base_pytree_checkpoint_handler.py:732] [process=0][thread=MainThread] Initiated Pytree async_save. Time taken: 0.781248s (batch_requests_ready=0.083655s, total_serialization_initiated=0.697100s, others=0.000492s) I0415 23:11:35.464541 124327861365888 composite_checkpoint_handler.py:715] [process=0][thread=MainThread] Initiated CompositeCheckpointHandler.async_save. Time taken: 0.782152s (all_items=0.000019s, per_item={'items': '0.00001884'}, temp_paths=0.782134) I0415 23:11:35.465403 124220245804608 async_checkpointer.py:79] [process=0][thread=async_save] Background save thread started. I0415 23:11:35.465498 124327861365888 async_checkpointer.py:561] Finished blocking save. Time taken: 0.874969s. Continuing background save to gs://wanglance-maxtext/linen_ckpt_feat_migrate_nnx_utils_20260415_230805/linen_feat_migrate_nnx_utils_20260415_230805_02_synthetic/checkpoints/0. I0415 23:11:35.465701 124327861365888 checkpoint_manager.py:1549] [process=0][thread=MainThread][step=0] Starting CheckpointManager Save Finalize thread=save_finalize I0415 23:11:35.465932 124327861365888 standard_logger.py:34] {'step': 0, 'event_type': 'save', 'directory': 'gs://wanglance-maxtext/linen_ckpt_feat_migrate_nnx_utils_20260415_230805/linen_feat_migrate_nnx_utils_20260415_230805_02_synthetic/checkpoints', 'reached_preemption': False, 'preemption_received_at': None, 'synchronous': False, 'wait_for_prev_start_time': 1776294694.5903308, 'wait_for_prev_duration_secs': 4.1961669921875e-05, 'time_between_consecutive_saves_sec': None, 'checkpointer_blocking_start_time': 1776294694.5905106, 'checkpointer_blocking_duration_secs': 0.8751130104064941, 'get_old_steps_start_time': 1776294695.4656425, 'get_old_steps_duration_secs': 2.288818359375e-05, 'checkpoint_manager_blocking_start_time': 1776294694.590249, 'checkpoint_manager_blocking_duration_secs': 0.8756585121154785} I0415 23:11:35.466003 124327861365888 checkpointing.py:408] Started an asynchronous checkpoint save for step 0 I0415 23:11:35.466057 124327861365888 max_utils.py:750] Memstats: After params initialized: I0415 23:11:35.466151 124327861365888 max_utils.py:756] Using (GB) 1.6 / 31.25 (5.120000%) on TPU_0(process=0,(0,0,0,0)) I0415 23:11:35.466178 124327861365888 max_utils.py:756] Using (GB) 1.59 / 31.25 (5.088000%) on TPU_1(process=0,(1,0,0,0)) I0415 23:11:35.466200 124327861365888 max_utils.py:756] Using (GB) 1.59 / 31.25 (5.088000%) on TPU_2(process=0,(0,1,0,0)) I0415 23:11:35.466221 124327861365888 max_utils.py:756] Using (GB) 1.59 / 31.25 (5.088000%) on TPU_3(process=0,(1,1,0,0)) I0415 23:11:35.466239 124327861365888 max_utils.py:756] Using (GB) 1.59 / 31.25 (5.088000%) on TPU_4(process=0,(0,2,0,0)) I0415 23:11:35.466257 124327861365888 max_utils.py:756] Using (GB) 1.59 / 31.25 (5.088000%) on TPU_5(process=0,(1,2,0,0)) I0415 23:11:35.466275 124327861365888 max_utils.py:756] Using (GB) 1.59 / 31.25 (5.088000%) on TPU_6(process=0,(0,3,0,0)) I0415 23:11:35.466293 124327861365888 max_utils.py:756] Using (GB) 1.59 / 31.25 (5.088000%) on TPU_7(process=0,(1,3,0,0)) I0415 23:11:35.472507 124220298233408 atomicity.py:137] Creating tmp directory gs://wanglance-maxtext/linen_ckpt_feat_migrate_nnx_utils_20260415_230805/linen_feat_migrate_nnx_utils_20260415_230805_02_synthetic/checkpoints/0/items W0415 23:11:35.472897 3351818 pjrt_executable.cc:642] Assume version compatibility. PjRt-IFRT does not track XLA executable versions. W0415 23:11:35.478534 3351818 pjrt_executable.cc:642] Assume version compatibility. PjRt-IFRT does not track XLA executable versions. W0415 23:11:35.483385 3351818 pjrt_executable.cc:642] Assume version compatibility. PjRt-IFRT does not track XLA executable versions. I0415 23:11:35.631895 124220266776128 checkpoint.py:188] Wrote Metadata={'item_handlers': None, 'metrics': {}, 'performance_metrics': {}, 'init_timestamp_nsecs': 1776294695379068601, 'commit_timestamp_nsecs': None, 'custom_metadata': {}}, json={"item_handlers": null, "metrics": {}, "performance_metrics": {}, "init_timestamp_nsecs": 1776294695379068601, "commit_timestamp_nsecs": null, "custom_metadata": {}} to gs://wanglance-maxtext/linen_ckpt_feat_migrate_nnx_utils_20260415_230805/linen_feat_migrate_nnx_utils_20260415_230805_02_synthetic/checkpoints/0/_CHECKPOINT_METADATA I0415 23:11:35.632171 124220319204928 async_checkpointer.py:265] [process=0][thread=save_finalize] Waiting for background save thread=async_save. I0415 23:11:35.796834 124327861365888 metric_logger.py:185] completed step: 0, seconds: 0.950, TFLOP/s/device: 14.298, Tokens/s/device: 2155.177, total_weights: 16384, loss: 10.874 I0415 23:11:35.797878 124327861365888 metric_logger.py:269] To see full metrics 'tensorboard --logdir=gs://wanglance-maxtext/linen_ckpt_feat_migrate_nnx_utils_20260415_230805/linen_feat_migrate_nnx_utils_20260415_230805_02_synthetic/tensorboard/' I0415 23:11:35.912815 124327861365888 metric_logger.py:185] completed step: 1, seconds: 1.322, TFLOP/s/device: 10.276, Tokens/s/device: 1548.914, total_weights: 16384, loss: 10.874 I0415 23:11:36.031547 124327861365888 metric_logger.py:185] completed step: 2, seconds: 0.007, TFLOP/s/device: 1963.174, Tokens/s/device: 295910.996, total_weights: 16384, loss: 9.775 I0415 23:11:36.150220 124327861365888 metric_logger.py:185] completed step: 3, seconds: 0.117, TFLOP/s/device: 116.493, Tokens/s/device: 17559.052, total_weights: 16384, loss: 8.838 I0415 23:11:36.169033 3354114 google_auth_provider.cc:149] Using credentials at ~/.config/gcloud/application_default_credentials.json I0415 23:11:36.169104 3354114 google_auth_provider.cc:156] Using OAuth2 AuthProvider I0415 23:11:37.176191 124220287747648 array_metadata_store.py:203] [process=0][thread=array_type_handler] Wrote 39 array_metadata.ArrayMetadata to gs://wanglance-maxtext/linen_ckpt_feat_migrate_nnx_utils_20260415_230805/linen_feat_migrate_nnx_utils_20260415_230805_02_synthetic/checkpoints/0/items/array_metadatas/process_0 I0415 23:11:56.128585 124327861365888 metric_logger.py:185] completed step: 4, seconds: 0.119, TFLOP/s/device: 114.413, Tokens/s/device: 17245.590, total_weights: 16384, loss: 7.965 I0415 23:11:56.136313 124327861365888 metric_logger.py:185] completed step: 5, seconds: 0.119, TFLOP/s/device: 114.504, Tokens/s/device: 17259.251, total_weights: 16384, loss: 7.233 I0415 23:11:56.253193 124327861365888 metric_logger.py:185] completed step: 6, seconds: 19.980, TFLOP/s/device: 0.680, Tokens/s/device: 102.505, total_weights: 16384, loss: 6.676 I0415 23:11:56.371896 124327861365888 metric_logger.py:185] completed step: 7, seconds: 0.005, TFLOP/s/device: 2944.123, Tokens/s/device: 443770.314, total_weights: 16384, loss: 6.293 I0415 23:11:56.490186 124327861365888 metric_logger.py:185] completed step: 8, seconds: 0.118, TFLOP/s/device: 115.406, Tokens/s/device: 17395.293, total_weights: 16384, loss: 6.055 I0415 23:11:56.608296 124327861365888 checkpointing.py:772] Waiting for step 9 to finish before checkpoint... I0415 23:11:56.609144 124327861365888 checkpointing.py:776] Waited 0.000865936279296875 seconds for step 9 to finish before starting checkpointing. I0415 23:11:56.609407 124327861365888 checkpoint_manager.py:1994] [process=0][thread=MainThread][step=0][wait_until_finished] Waiting for Save Finalize thread (save_finalize) to complete. I0415 23:12:01.075901 124220256290368 base_pytree_checkpoint_handler.py:1217] [process=0][thread=write_metadata_after_commits] Commit + Array metadata written. Time taken: 24.913215s (commit=24.486937s, array_metadata_write=0.426279s) I0415 23:12:01.077102 124220245804608 base_pytree_checkpoint_handler.py:128] [process=0] /jax/checkpoint/write/gbytes_per_sec: 478.777 MiB/s (total gbytes: 12.3 GiB) (time elapsed: 26 seconds) (per-host) I0415 23:12:01.077211 124220245804608 async_checkpointer.py:90] [process=0][thread=async_save] 3 Handler Commit operations completed. Time taken: 25.611668s. I0415 23:12:01.305080 124220245804608 checkpoint.py:228] Read Metadata={'item_handlers': None, 'metrics': {}, 'performance_metrics': {}, 'init_timestamp_nsecs': 1776294695379068601, 'commit_timestamp_nsecs': None, 'custom_metadata': {}} from gs://wanglance-maxtext/linen_ckpt_feat_migrate_nnx_utils_20260415_230805/linen_feat_migrate_nnx_utils_20260415_230805_02_synthetic/checkpoints/0/_CHECKPOINT_METADATA I0415 23:12:01.513718 124220245804608 array_metadata_store.py:367] [process=0][thread=async_save] Skipped cross-host ArrayMetadata validation because only one process is found: process_index=0. I0415 23:12:01.705731 124220266776128 checkpoint.py:247] Updated Metadata={'item_handlers': {'items': 'orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeCheckpointHandler'}, 'metrics': {}, 'performance_metrics': {}, 'init_timestamp_nsecs': 1776294695379068601, 'commit_timestamp_nsecs': None, 'custom_metadata': {}} to gs://wanglance-maxtext/linen_ckpt_feat_migrate_nnx_utils_20260415_230805/linen_feat_migrate_nnx_utils_20260415_230805_02_synthetic/checkpoints/0/_CHECKPOINT_METADATA I0415 23:12:01.928602 124220245804608 ocdbt_utils.py:56] Param validation support for Zarr3 will be added later (b/362328389). I0415 23:12:01.929249 124220245804608 base_pytree_checkpoint_handler.py:1342] [process=0][thread=async_save] Pytree save finalize (merge_ocdbt + ArrayMetadata validation) completed. Time taken: 0.583574s. use_zarr3=True, enable_post_merge_validation=True, directory=gs://wanglance-maxtext/linen_ckpt_feat_migrate_nnx_utils_20260415_230805/linen_feat_migrate_nnx_utils_20260415_230805_02_synthetic/checkpoints/0/items I0415 23:12:01.929905 124220245804608 atomicity.py:608] Finalizing gs://wanglance-maxtext/linen_ckpt_feat_migrate_nnx_utils_20260415_230805/linen_feat_migrate_nnx_utils_20260415_230805_02_synthetic/checkpoints/0/items I0415 23:12:02.179535 124220245804608 atomicity.py:608] Finalizing gs://wanglance-maxtext/linen_ckpt_feat_migrate_nnx_utils_20260415_230805/linen_feat_migrate_nnx_utils_20260415_230805_02_synthetic/checkpoints/0 I0415 23:12:02.860709 124220245804608 atomicity.py:794] [process=0][thread=async_save] Finished saving checkpoint (finalized tmp dir) to `gs://wanglance-maxtext/linen_ckpt_feat_migrate_nnx_utils_20260415_230805/linen_feat_migrate_nnx_utils_20260415_230805_02_synthetic/checkpoints/0`. I0415 23:12:02.861350 124220245804608 async_checkpointer.py:420] Finished async_save (blocking + background). Time taken: 28.270827s. directory=gs://wanglance-maxtext/linen_ckpt_feat_migrate_nnx_utils_20260415_230805/linen_feat_migrate_nnx_utils_20260415_230805_02_synthetic/checkpoints/0 I0415 23:12:02.861428 124220245804608 async_checkpointer.py:144] [process=0][thread=async_save] Background save thread done. Time taken: 27.395888s. I0415 23:12:02.861603 124220319204928 async_checkpointer.py:273] [process=0][thread=save_finalize] Done with waiting for background save thread=async_save. I0415 23:12:02.861700 124220319204928 async_checkpointer.py:283] [process=0][thread=save_finalize] No errors found in background save thread=async_save. I0415 23:12:02.861759 124220319204928 checkpoint_manager.py:2103] [process=0][thread=save_finalize][step=0] CheckpointManager Save Finalize is syncing with other hosts... I0415 23:12:02.861804 124220319204928 checkpoint_manager.py:2112] [process=0][thread=save_finalize][step=0] CheckpointManager Save Finalize is done on all hosts. I0415 23:12:02.861936 124327861365888 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. W0415 23:12:02.862075 124327861365888 checkpoint_manager.py:1441] Waiting for previous save to complete took 6.252660 seconds. If this number is high, consider checkpointing less frequently. I0415 23:12:02.862882 124327861365888 checkpoint_manager.py:1501] [process=0] Saving checkpoint at step 9 I0415 23:12:02.863194 124327861365888 async_checkpointer.py:452] [process=0] Started async saving checkpoint to gs://wanglance-maxtext/linen_ckpt_feat_migrate_nnx_utils_20260415_230805/linen_feat_migrate_nnx_utils_20260415_230805_02_synthetic/checkpoints/9. I0415 23:12:03.047594 124220319204928 atomicity.py:137] Creating tmp directory gs://wanglance-maxtext/linen_ckpt_feat_migrate_nnx_utils_20260415_230805/linen_feat_migrate_nnx_utils_20260415_230805_02_synthetic/checkpoints/9 I0415 23:12:03.064708 124327861365888 jax_array_handlers.py:347] Scheduling D2H of 39 prioritized jax.Array. I0415 23:12:03.064800 124327861365888 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 I0415 23:12:03.778093 124220256290368 atomicity.py:137] Creating tmp directory gs://wanglance-maxtext/linen_ckpt_feat_migrate_nnx_utils_20260415_230805/linen_feat_migrate_nnx_utils_20260415_230805_02_synthetic/checkpoints/9/items I0415 23:12:08.526167 124327861365888 base_pytree_checkpoint_handler.py:153] [process=0][thread=MainThread] Initiated "orbax.checkpoint._src.serialization.jax_array_handlers.ArrayHandler".serialize. Time taken: 5.461867s I0415 23:12:08.530422 124327861365888 base_pytree_checkpoint_handler.py:128] [process=0] /jax/checkpoint/write/blocking_gbytes_per_sec: 2.215 GiB/s (total gbytes: 12.3 GiB) (time elapsed: 5 seconds) (per-host) I0415 23:12:08.530478 124327861365888 base_pytree_checkpoint_handler.py:732] [process=0][thread=MainThread] Initiated Pytree async_save. Time taken: 5.572355s (batch_requests_ready=0.103174s, total_serialization_initiated=5.465039s, others=0.004143s) I0415 23:12:08.530548 124327861365888 composite_checkpoint_handler.py:715] [process=0][thread=MainThread] Initiated CompositeCheckpointHandler.async_save. Time taken: 5.573157s (all_items=0.000018s, per_item={'items': '0.00001764'}, temp_paths=5.573139) I0415 23:12:08.531978 124220245804608 async_checkpointer.py:79] [process=0][thread=async_save] Background save thread started. I0415 23:12:08.532072 124327861365888 async_checkpointer.py:561] Finished blocking save. Time taken: 5.669143s. Continuing background save to gs://wanglance-maxtext/linen_ckpt_feat_migrate_nnx_utils_20260415_230805/linen_feat_migrate_nnx_utils_20260415_230805_02_synthetic/checkpoints/9. I0415 23:12:08.532258 124327861365888 checkpoint_manager.py:1549] [process=0][thread=MainThread][step=9] Starting CheckpointManager Save Finalize thread=save_finalize I0415 23:12:08.532435 124220298233408 async_checkpointer.py:265] [process=0][thread=save_finalize] Waiting for background save thread=async_save. I0415 23:12:08.532544 124327861365888 standard_logger.py:34] {'step': 9, 'event_type': 'save', 'directory': 'gs://wanglance-maxtext/linen_ckpt_feat_migrate_nnx_utils_20260415_230805/linen_feat_migrate_nnx_utils_20260415_230805_02_synthetic/checkpoints', 'reached_preemption': False, 'preemption_received_at': None, 'synchronous': False, 'wait_for_prev_start_time': 1776294716.6093755, 'wait_for_prev_duration_secs': 6.252659797668457, 'time_between_consecutive_saves_sec': None, 'checkpointer_blocking_start_time': 1776294722.8629084, 'checkpointer_blocking_duration_secs': 5.669263124465942, 'get_old_steps_start_time': 1776294728.5321877, 'get_old_steps_duration_secs': 3.170967102050781e-05, 'checkpoint_manager_blocking_start_time': 1776294716.6093338, 'checkpoint_manager_blocking_duration_secs': 11.923186540603638} I0415 23:12:08.532666 124327861365888 checkpointing.py:408] Started an asynchronous checkpoint save for step 9 I0415 23:12:08.532699 124327861365888 checkpoint_manager.py:1994] [process=0][thread=MainThread][step=9][wait_until_finished] Waiting for Save Finalize thread (save_finalize) to complete. I0415 23:12:09.295620 124220287747648 array_metadata_store.py:203] [process=0][thread=array_type_handler] Wrote 39 array_metadata.ArrayMetadata to gs://wanglance-maxtext/linen_ckpt_feat_migrate_nnx_utils_20260415_230805/linen_feat_migrate_nnx_utils_20260415_230805_02_synthetic/checkpoints/9/items/array_metadatas/process_0 I0415 23:12:45.893706 124220256290368 base_pytree_checkpoint_handler.py:1217] [process=0][thread=write_metadata_after_commits] Commit + Array metadata written. Time taken: 37.362581s (commit=36.895165s, array_metadata_write=0.467416s) I0415 23:12:45.894732 124220245804608 base_pytree_checkpoint_handler.py:128] [process=0] /jax/checkpoint/write/gbytes_per_sec: 294.313 MiB/s (total gbytes: 12.3 GiB) (time elapsed: 42 seconds) (per-host) I0415 23:12:45.894786 124220245804608 async_checkpointer.py:90] [process=0][thread=async_save] 3 Handler Commit operations completed. Time taken: 37.362672s. I0415 23:12:46.383485 124220245804608 array_metadata_store.py:367] [process=0][thread=async_save] Skipped cross-host ArrayMetadata validation because only one process is found: process_index=0. I0415 23:12:46.828654 124220245804608 ocdbt_utils.py:56] Param validation support for Zarr3 will be added later (b/362328389). I0415 23:12:46.829271 124220245804608 base_pytree_checkpoint_handler.py:1342] [process=0][thread=async_save] Pytree save finalize (merge_ocdbt + ArrayMetadata validation) completed. Time taken: 0.601574s. use_zarr3=True, enable_post_merge_validation=True, directory=gs://wanglance-maxtext/linen_ckpt_feat_migrate_nnx_utils_20260415_230805/linen_feat_migrate_nnx_utils_20260415_230805_02_synthetic/checkpoints/9/items I0415 23:12:46.829925 124220245804608 atomicity.py:608] Finalizing gs://wanglance-maxtext/linen_ckpt_feat_migrate_nnx_utils_20260415_230805/linen_feat_migrate_nnx_utils_20260415_230805_02_synthetic/checkpoints/9/items I0415 23:12:47.076901 124220245804608 atomicity.py:608] Finalizing gs://wanglance-maxtext/linen_ckpt_feat_migrate_nnx_utils_20260415_230805/linen_feat_migrate_nnx_utils_20260415_230805_02_synthetic/checkpoints/9 I0415 23:12:47.767763 124220245804608 atomicity.py:794] [process=0][thread=async_save] Finished saving checkpoint (finalized tmp dir) to `gs://wanglance-maxtext/linen_ckpt_feat_migrate_nnx_utils_20260415_230805/linen_feat_migrate_nnx_utils_20260415_230805_02_synthetic/checkpoints/9`. I0415 23:12:47.768421 124220245804608 async_checkpointer.py:420] Finished async_save (blocking + background). Time taken: 44.905497s. directory=gs://wanglance-maxtext/linen_ckpt_feat_migrate_nnx_utils_20260415_230805/linen_feat_migrate_nnx_utils_20260415_230805_02_synthetic/checkpoints/9 I0415 23:12:47.768499 124220245804608 async_checkpointer.py:144] [process=0][thread=async_save] Background save thread done. Time taken: 39.236386s. I0415 23:12:47.768703 124220298233408 async_checkpointer.py:273] [process=0][thread=save_finalize] Done with waiting for background save thread=async_save. I0415 23:12:47.768809 124220298233408 async_checkpointer.py:283] [process=0][thread=save_finalize] No errors found in background save thread=async_save. I0415 23:12:47.768869 124220298233408 checkpoint_manager.py:2103] [process=0][thread=save_finalize][step=9] CheckpointManager Save Finalize is syncing with other hosts... I0415 23:12:47.768923 124220298233408 checkpoint_manager.py:2112] [process=0][thread=save_finalize][step=9] CheckpointManager Save Finalize is done on all hosts. I0415 23:12:47.769151 124327861365888 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. I0415 23:12:47.769270 124327861365888 checkpoint_manager.py:1983] [process=0][thread=MainThread][wait_until_finished] No Save Finalize thread to wait for. Returning. I0415 23:12:47.769807 124327861365888 metric_logger.py:185] completed step: 9, seconds: 0.119, TFLOP/s/device: 114.329, Tokens/s/device: 17232.965, 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