2026-04-16 20:16:06.070523: 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 20:16:06.185807 134367141866624 max_utils.py:238] Skipping jax distributed system due to skip_jax_distributed_system=True flag. I0416 20:17:03.580356 134367141866624 max_utils.py:800] System Information: Jax Version: 0.9.2 I0416 20:17:03.580461 134367141866624 max_utils.py:801] System Information: Jaxlib Version: 0.9.2 I0416 20:17:03.580495 134367141866624 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 20:17:03.580519 134367141866624 train_utils.py:364] WARNING: 'dataset_path' might be pointing your local file system I0416 20:17:03.580540 134367141866624 train_utils.py:377] WARNING: Sequence packing is essentially ignored for synthetic data. Please use a real dataset to use sequence packing. I0416 20:17:03.580627 134367141866624 train.py:811] [DECOUPLED NO-OP] skipping cloud diagnostics wrapper. W0416 20:17:03.673029 4091524 pjrt_executable.cc:642] Assume version compatibility. PjRt-IFRT does not track XLA executable versions. I0416 20:17:04.070676 134367141866624 maxtext_utils.py:1687] Num_devices: 8, shape (1, 1, 1, 8, 1, 1, 1, 1, 1, 1, 1, 1, 1) I0416 20:17:04.169694 134367141866624 checkpointing.py:688] Setting up checkpoint logger... I0416 20:17:04.169795 134367141866624 checkpointing.py:234] Creating checkpoint manager with ocdbt=True and zarr3=True I0416 20:17:04.169837 134367141866624 pytree_checkpoint_handler.py:577] save_device_host_concurrent_bytes=None I0416 20:17:04.170035 134367141866624 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 0x7a34337fcf20>, enable_pinned_host_transfer=False, save_concurrent_bytes: 96000000000 (89.4 GiB), restore_concurrent_bytes: 96000000000 (89.4 GiB) I0416 20:17:06.594425 134367141866624 checkpointing.py:266] Enabling policy for fixed interval checkpointing. I0416 20:17:06.594598 134367141866624 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 0x7a2db3b5f6e0>}, handler_registry=None I0416 20:17:06.594855 134367141866624 composite_checkpoint_handler.py:237] Deferred registration for item: "items". Adding handler `<orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeCheckpointHandler object at 0x7a2db3b5f6e0>` 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 20:17:06.594900 134367141866624 composite_checkpoint_handler.py:237] Deferred registration for item: "metrics". Adding handler `<orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonCheckpointHandler object at 0x7a2db3b947a0>` 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 20:17:06.594928 134367141866624 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 0x7a2db3b5f6e0>, ('items', <class 'orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeRestoreArgs'>): <orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeCheckpointHandler object at 0x7a2db3b5f6e0>, ('metrics', <class 'orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonSaveArgs'>): <orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonCheckpointHandler object at 0x7a2db3b947a0>, ('metrics', <class 'orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonRestoreArgs'>): <orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonCheckpointHandler object at 0x7a2db3b947a0>}). I0416 20:17:06.595243 134367141866624 abstract_checkpointer.py:35] orbax-checkpoint version: 0.11.28 I0416 20:17:06.595293 134367141866624 async_checkpointer.py:177] [process=0][thread=MainThread] Using barrier_sync_fn: <function get_barrier_sync_fn.<locals>.<lambda> at 0x7a2db4b79760> timeout: 600 secs and primary_host=0 for async checkpoint writes I0416 20:17:06.729452 134367141866624 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_11_optimizer_offload_false/checkpoints I0416 20:17:06.729671 134367141866624 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_11_optimizer_offload_false/checkpoints: <orbax.checkpoint.checkpoint_manager.CheckpointManager object at 0x7a2db3b97320> I0416 20:17:06.729748 134367141866624 checkpointing.py:302] Checkpoint manager created! W0416 20:17:06.745461 4091524 pjrt_executable.cc:642] Assume version compatibility. PjRt-IFRT does not track XLA executable versions. I0416 20:17:06.955265 134367141866624 checkpointing.py:578] checkpoint manager exists so trying to load this run's existing checkpoint I0416 20:17:06.955372 134367141866624 checkpointing.py:676] No existing checkpoints found, not restoring checkpoint. W0416 20:17:07.063988 4091524 pjrt_executable.cc:642] Assume version compatibility. PjRt-IFRT does not track XLA executable versions. [DECOUPLED NO-OP] gcs_storage: using stubs. [DECOUPLED NO-OP] mldiagnostics: using stub. [DECOUPLED NO-OP] mldiagnostics: using stub. [DECOUPLED NO-OP] mldiagnostics: using stub. [DECOUPLED NO-OP] workload_monitor: using stub. [DECOUPLED NO-OP] vertex_tensorboard: using stub. fsdp: 8 I0416 20:17:07.150821 134367141866624 maxtext_utils.py:1805] decoder/decoder_norm/scale/value Shape: float32[2048] Physical: (None,) I0416 20:17:07.150927 134367141866624 maxtext_utils.py:1805] decoder/layers/mlp/wi_0/kernel/value Shape: float32[2048,16,7168] Physical: ('fsdp', None, None) I0416 20:17:07.150968 134367141866624 maxtext_utils.py:1805] decoder/layers/mlp/wi_1/kernel/value Shape: float32[2048,16,7168] Physical: ('fsdp', None, None) I0416 20:17:07.151000 134367141866624 maxtext_utils.py:1805] decoder/layers/mlp/wo/kernel/value Shape: float32[7168,16,2048] Physical: (None, None, 'fsdp') I0416 20:17:07.151028 134367141866624 maxtext_utils.py:1805] decoder/layers/post_self_attention_layer_norm/scale/value Shape: float32[2048,16] Physical: (None, None) I0416 20:17:07.151070 134367141866624 maxtext_utils.py:1805] decoder/layers/pre_self_attention_layer_norm/scale/value Shape: float32[2048,16] Physical: (None, None) I0416 20:17:07.151099 134367141866624 maxtext_utils.py:1805] decoder/layers/self_attention/key/kernel/value Shape: float32[2048,16,16,128] Physical: ('fsdp', None, None, None) I0416 20:17:07.151124 134367141866624 maxtext_utils.py:1805] decoder/layers/self_attention/out/kernel/value Shape: float32[16,16,128,2048] Physical: (None, None, None, 'fsdp') I0416 20:17:07.151149 134367141866624 maxtext_utils.py:1805] decoder/layers/self_attention/query/kernel/value Shape: float32[2048,16,16,128] Physical: ('fsdp', None, None, None) I0416 20:17:07.151173 134367141866624 maxtext_utils.py:1805] decoder/layers/self_attention/value/kernel/value Shape: float32[2048,16,16,128] Physical: ('fsdp', None, None, None) I0416 20:17:07.151196 134367141866624 maxtext_utils.py:1805] decoder/logits_dense/kernel/value Shape: float32[2048,32000] Physical: ('fsdp', None) I0416 20:17:07.151218 134367141866624 maxtext_utils.py:1805] token_embedder/embedding/value Shape: float32[32000,2048] Physical: (None, 'fsdp') I0416 20:17:07.257875 134367141866624 nnx_decoders.py:465] nnx_decoders/carry Logical: bfloat16[8,2048,2048]....................................... ('activation_batch', 'activation_norm_length', 'activation_embed'). I0416 20:17:07.257956 134367141866624 nnx_decoders.py:465] nnx_decoders/carry Physical: bfloat16[8,2048,2048]....................................... ('fsdp', None, None). I0416 20:17:07.263108 134367141866624 nnx_decoders.py:465] Unknown Logical: bfloat16[8,2048,2048]....................................... ('activation_batch', 'activation_norm_length', 'activation_embed'). I0416 20:17:07.263153 134367141866624 nnx_decoders.py:465] Unknown Physical: bfloat16[8,2048,2048]....................................... ('fsdp', None, None). I0416 20:17:07.282244 134367141866624 attentions.py:1088] attentions/inputs_q Logical: bfloat16[8,2048,2048]....................................... ('activation_batch', 'activation_attn_length', 'activation_attn_embed'). I0416 20:17:07.282360 134367141866624 attentions.py:1088] attentions/inputs_q Physical: bfloat16[8,2048,2048]....................................... ('fsdp', None, None). I0416 20:17:07.296168 134367141866624 attentions.py:1089] attentions/inputs_kv Logical: bfloat16[8,2048,2048]....................................... ('activation_batch', 'activation_attn_length', 'activation_attn_embed'). I0416 20:17:07.296214 134367141866624 attentions.py:1089] attentions/inputs_kv Physical: bfloat16[8,2048,2048]....................................... ('fsdp', None, None). I0416 20:17:07.318289 134367141866624 attentions.py:1154] attentions/query Logical: bfloat16[8,2048,16,128]..................................... ('activation_kv_batch', 'activation_attn_length', 'activation_kv_heads', 'activation_kv_head_dim'). I0416 20:17:07.318352 134367141866624 attentions.py:1154] attentions/query Physical: bfloat16[8,2048,16,128]..................................... ('fsdp', None, None, None). I0416 20:17:07.332161 134367141866624 attentions.py:1155] attentions/key Logical: bfloat16[8,2048,16,128]..................................... ('activation_kv_batch', 'activation_attn_length', 'activation_kv_heads', 'activation_kv_head_dim'). I0416 20:17:07.332211 134367141866624 attentions.py:1155] attentions/key Physical: bfloat16[8,2048,16,128]..................................... ('fsdp', None, None, None). I0416 20:17:07.345894 134367141866624 attentions.py:1156] attentions/value Logical: bfloat16[8,2048,16,128]..................................... ('activation_kv_batch', 'activation_attn_length', 'activation_kv_heads', 'activation_kv_head_dim'). I0416 20:17:07.345941 134367141866624 attentions.py:1156] attentions/value Physical: bfloat16[8,2048,16,128]..................................... ('fsdp', None, None, None). I0416 20:17:07.373196 134367141866624 attentions.py:1197] attentions/out Logical: bfloat16[8,2048,16,128]..................................... ('activation_batch', 'activation_attn_length', 'activation_heads', 'activation_kv'). I0416 20:17:07.373253 134367141866624 attentions.py:1197] attentions/out Physical: bfloat16[8,2048,16,128]..................................... ('fsdp', None, None, None). I0416 20:17:07.391454 134367141866624 linears.py:525] linears/x Logical: bfloat16[8,2048,7168]....................................... ('activation_batch', 'activation_length', 'activation_mlp'). I0416 20:17:07.391508 134367141866624 linears.py:525] linears/x Physical: bfloat16[8,2048,7168]....................................... ('fsdp', None, None). W0416 20:17:08.122443 4091524 pjrt_executable.cc:642] Assume version compatibility. PjRt-IFRT does not track XLA executable versions. I0416 20:17:08.240656 134367141866624 max_utils.py:791] Total memory size: 3.6 GB, Output size: 1.5 GB, Temp size: 2.0 GB, Argument size: 1.5 GB, Host temp size: 0.0 GB. I0416 20:17:08.241209 134367141866624 max_utils.py:194] tensorboardX not available; using no-op SummaryWriter. I0416 20:17:08.242519 134367141866624 metric_logger.py:289] number parameters: 1.104 billion W0416 20:17:08.862993 4091524 pjrt_executable.cc:642] Assume version compatibility. PjRt-IFRT does not track XLA executable versions. I0416 20:17:08.964234 134367141866624 checkpointing.py:794] Waiting for step 0 to finish before checkpoint... I0416 20:17:09.080125 134367141866624 checkpointing.py:798] Waited 0.11588001251220703 seconds for step 0 to finish before starting checkpointing. I0416 20:17:09.080545 134367141866624 checkpoint_manager.py:1983] [process=0][thread=MainThread][wait_until_finished] No Save Finalize thread to wait for. Returning. I0416 20:17:09.080700 134367141866624 checkpoint_manager.py:1501] [process=0] Saving checkpoint at step 0 I0416 20:17:09.081074 134367141866624 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_11_optimizer_offload_false/checkpoints/0. I0416 20:17:09.172737 134367141866624 signaling_client.py:373] Using ThreadSafeKeyValueSignalingClient I0416 20:17:09.261315 134259700401728 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_11_optimizer_offload_false/checkpoints/0 I0416 20:17:09.284654 134367141866624 jax_array_handlers.py:347] Scheduling D2H of 69 prioritized jax.Array. I0416 20:17:09.284738 134367141866624 replica_slices.py:410] Transferring arrays to host memory with options: use_replica_parallel=True, min_slice_bytes_for_replica_parallel=None, max_replicas_for_replica_parallel=None, enable_pinned_host_transfer=False W0416 20:17:09.924380 4091524 pjrt_executable.cc:642] Assume version compatibility. PjRt-IFRT does not track XLA executable versions. I0416 20:17:09.929550 134259689915968 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_11_optimizer_offload_false/checkpoints/0/items W0416 20:17:09.935512 4091524 pjrt_executable.cc:642] Assume version compatibility. PjRt-IFRT does not track XLA executable versions. W0416 20:17:09.942016 4091524 pjrt_executable.cc:642] Assume version compatibility. PjRt-IFRT does not track XLA executable versions. W0416 20:17:09.952477 4091524 pjrt_executable.cc:642] Assume version compatibility. PjRt-IFRT does not track XLA executable versions. W0416 20:17:09.959784 4091524 pjrt_executable.cc:642] Assume version compatibility. PjRt-IFRT does not track XLA executable versions. W0416 20:17:09.965253 4091524 pjrt_executable.cc:642] Assume version compatibility. PjRt-IFRT does not track XLA executable versions. W0416 20:17:09.973604 4091524 pjrt_executable.cc:642] Assume version compatibility. PjRt-IFRT does not track XLA executable versions. W0416 20:17:09.979175 4091524 pjrt_executable.cc:642] Assume version compatibility. PjRt-IFRT does not track XLA executable versions. I0416 20:17:10.110923 134259650070080 checkpoint.py:188] Wrote Metadata={'item_handlers': None, 'metrics': {}, 'performance_metrics': {}, 'init_timestamp_nsecs': 1776370629841147923, 'commit_timestamp_nsecs': None, 'custom_metadata': {}}, json={"item_handlers": null, "metrics": {}, "performance_metrics": {}, "init_timestamp_nsecs": 1776370629841147923, "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_11_optimizer_offload_false/checkpoints/0/_CHECKPOINT_METADATA I0416 20:17:10.511088 4094036 google_auth_provider.cc:149] Using credentials at ~/.config/gcloud/application_default_credentials.json I0416 20:17:10.511143 4094036 google_auth_provider.cc:156] Using OAuth2 AuthProvider I0416 20:17:10.923299 134367141866624 base_pytree_checkpoint_handler.py:153] [process=0][thread=MainThread] Initiated "orbax.checkpoint._src.serialization.jax_array_handlers.ArrayHandler".serialize. Time taken: 1.639625s I0416 20:17:10.930005 134367141866624 base_pytree_checkpoint_handler.py:128] [process=0] /jax/checkpoint/write/blocking_gbytes_per_sec: 7.031 GiB/s (total gbytes: 12.3 GiB) (time elapsed: a second) (per-host) I0416 20:17:10.930155 134367141866624 base_pytree_checkpoint_handler.py:732] [process=0][thread=MainThread] Initiated Pytree async_save. Time taken: 1.755528s (batch_requests_ready=0.101566s, total_serialization_initiated=1.647359s, others=0.006603s) I0416 20:17:10.930280 134367141866624 composite_checkpoint_handler.py:715] [process=0][thread=MainThread] Initiated CompositeCheckpointHandler.async_save. Time taken: 1.756553s (all_items=0.000019s, per_item={'items': '0.00001931'}, temp_paths=1.756534) I0416 20:17:10.931974 134259576669760 async_checkpointer.py:79] [process=0][thread=async_save] Background save thread started. I0416 20:17:10.932078 134367141866624 async_checkpointer.py:561] Finished blocking save. Time taken: 1.851338s. Continuing background save to gs://wanglance-maxtext/nnx_ckpt_feat_nnx_post_train_fixes_20260416_181521/nnx_feat_nnx_post_train_fixes_20260416_181521_11_optimizer_offload_false/checkpoints/0. I0416 20:17:10.932264 134367141866624 checkpoint_manager.py:1549] [process=0][thread=MainThread][step=0] Starting CheckpointManager Save Finalize thread=save_finalize I0416 20:17:10.932435 134259710887488 async_checkpointer.py:265] [process=0][thread=save_finalize] Waiting for background save thread=async_save. I0416 20:17:10.932529 134367141866624 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_11_optimizer_offload_false/checkpoints', 'reached_preemption': False, 'preemption_received_at': None, 'synchronous': False, 'wait_for_prev_start_time': 1776370629.080531, 'wait_for_prev_duration_secs': 4.00543212890625e-05, 'time_between_consecutive_saves_sec': None, 'checkpointer_blocking_start_time': 1776370629.0807211, 'checkpointer_blocking_duration_secs': 1.851464033126831, 'get_old_steps_start_time': 1776370630.9322052, 'get_old_steps_duration_secs': 2.2172927856445312e-05, 'checkpoint_manager_blocking_start_time': 1776370629.0804524, 'checkpoint_manager_blocking_duration_secs': 1.8520479202270508} I0416 20:17:10.932652 134367141866624 checkpointing.py:409] Started an asynchronous checkpoint save for step 0 I0416 20:17:10.932707 134367141866624 max_utils.py:750] Memstats: After params initialized: I0416 20:17:10.932757 134367141866624 max_utils.py:756] Using (GB) 1.59 / 31.25 (5.088000%) on TPU_0(process=0,(0,0,0,0)) I0416 20:17:10.932784 134367141866624 max_utils.py:756] Using (GB) 1.59 / 31.25 (5.088000%) on TPU_1(process=0,(1,0,0,0)) I0416 20:17:10.932807 134367141866624 max_utils.py:756] Using (GB) 1.59 / 31.25 (5.088000%) on TPU_2(process=0,(0,1,0,0)) I0416 20:17:10.932826 134367141866624 max_utils.py:756] Using (GB) 1.59 / 31.25 (5.088000%) on TPU_3(process=0,(1,1,0,0)) I0416 20:17:10.932846 134367141866624 max_utils.py:756] Using (GB) 1.59 / 31.25 (5.088000%) on TPU_4(process=0,(0,2,0,0)) I0416 20:17:10.932863 134367141866624 max_utils.py:756] Using (GB) 1.59 / 31.25 (5.088000%) on TPU_5(process=0,(1,2,0,0)) I0416 20:17:10.932881 134367141866624 max_utils.py:756] Using (GB) 1.59 / 31.25 (5.088000%) on TPU_6(process=0,(0,3,0,0)) I0416 20:17:10.932899 134367141866624 max_utils.py:756] Using (GB) 1.59 / 31.25 (5.088000%) on TPU_7(process=0,(1,3,0,0)) W0416 20:17:10.940019 4091524 pjrt_executable.cc:642] Assume version compatibility. PjRt-IFRT does not track XLA executable versions. W0416 20:17:10.945652 4091524 pjrt_executable.cc:642] Assume version compatibility. PjRt-IFRT does not track XLA executable versions. W0416 20:17:10.950367 4091524 pjrt_executable.cc:642] Assume version compatibility. PjRt-IFRT does not track XLA executable versions. I0416 20:17:11.284143 134367141866624 metric_logger.py:185] completed step: 0, seconds: 0.720, TFLOP/s/device: 18.859, Tokens/s/device: 2842.605, total_weights: 16384, loss: 10.881 I0416 20:17:11.285126 134367141866624 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_11_optimizer_offload_false/tensorboard/' I0416 20:17:11.396578 134367141866624 metric_logger.py:185] completed step: 1, seconds: 2.316, TFLOP/s/device: 5.867, Tokens/s/device: 884.367, total_weights: 16384, loss: 10.881 I0416 20:17:11.515059 134367141866624 metric_logger.py:185] completed step: 2, seconds: 0.018, TFLOP/s/device: 745.726, Tokens/s/device: 112403.952, total_weights: 16384, loss: 9.779 I0416 20:17:11.633784 134367141866624 metric_logger.py:185] completed step: 3, seconds: 0.105, TFLOP/s/device: 129.292, Tokens/s/device: 19488.243, total_weights: 16384, loss: 8.837 I0416 20:17:12.119934 134259668944448 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_11_optimizer_offload_false/checkpoints/0/items/array_metadatas/process_0 I0416 20:17:31.329529 134367141866624 metric_logger.py:185] completed step: 4, seconds: 0.119, TFLOP/s/device: 114.125, Tokens/s/device: 17202.133, total_weights: 16384, loss: 7.961 I0416 20:17:31.338555 134367141866624 metric_logger.py:185] completed step: 5, seconds: 0.118, TFLOP/s/device: 114.864, Tokens/s/device: 17313.529, total_weights: 16384, loss: 7.226 I0416 20:17:31.453507 134367141866624 metric_logger.py:185] completed step: 6, seconds: 19.696, TFLOP/s/device: 0.690, Tokens/s/device: 103.979, total_weights: 16384, loss: 6.669 I0416 20:17:31.571890 134367141866624 metric_logger.py:185] completed step: 7, seconds: 0.006, TFLOP/s/device: 2128.310, Tokens/s/device: 320802.005, total_weights: 16384, loss: 6.286 I0416 20:17:31.690169 134367141866624 metric_logger.py:185] completed step: 8, seconds: 0.116, TFLOP/s/device: 116.665, Tokens/s/device: 17584.984, total_weights: 16384, loss: 6.048 I0416 20:17:31.809171 134367141866624 checkpointing.py:794] Waiting for step 9 to finish before checkpoint... I0416 20:17:31.810416 134367141866624 checkpointing.py:798] Waited 0.0012607574462890625 seconds for step 9 to finish before starting checkpointing. I0416 20:17:31.810694 134367141866624 checkpoint_manager.py:1994] [process=0][thread=MainThread][step=0][wait_until_finished] Waiting for Save Finalize thread (save_finalize) to complete. I0416 20:17:41.305819 134259597641280 base_pytree_checkpoint_handler.py:1217] [process=0][thread=write_metadata_after_commits] Commit + Array metadata written. Time taken: 30.374542s (commit=29.936940s, array_metadata_write=0.437602s) I0416 20:17:41.306882 134259576669760 base_pytree_checkpoint_handler.py:128] [process=0] /jax/checkpoint/write/gbytes_per_sec: 393.275 MiB/s (total gbytes: 12.3 GiB) (time elapsed: 32 seconds) (per-host) I0416 20:17:41.306941 134259576669760 async_checkpointer.py:90] [process=0][thread=async_save] 3 Handler Commit operations completed. Time taken: 30.374815s. I0416 20:17:41.564794 134259576669760 checkpoint.py:228] Read Metadata={'item_handlers': None, 'metrics': {}, 'performance_metrics': {}, 'init_timestamp_nsecs': 1776370629841147923, '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_11_optimizer_offload_false/checkpoints/0/_CHECKPOINT_METADATA I0416 20:17:41.762875 134259576669760 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 20:17:41.974621 134259650070080 checkpoint.py:247] Updated Metadata={'item_handlers': {'items': 'orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeCheckpointHandler'}, 'metrics': {}, 'performance_metrics': {}, 'init_timestamp_nsecs': 1776370629841147923, '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_11_optimizer_offload_false/checkpoints/0/_CHECKPOINT_METADATA I0416 20:17:42.215165 134259576669760 ocdbt_utils.py:56] Param validation support for Zarr3 will be added later (b/362328389). I0416 20:17:42.215809 134259576669760 base_pytree_checkpoint_handler.py:1342] [process=0][thread=async_save] Pytree save finalize (merge_ocdbt + ArrayMetadata validation) completed. Time taken: 0.601936s. 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_11_optimizer_offload_false/checkpoints/0/items I0416 20:17:42.216467 134259576669760 atomicity.py:608] Finalizing gs://wanglance-maxtext/nnx_ckpt_feat_nnx_post_train_fixes_20260416_181521/nnx_feat_nnx_post_train_fixes_20260416_181521_11_optimizer_offload_false/checkpoints/0/items I0416 20:17:42.479528 134259576669760 atomicity.py:608] Finalizing gs://wanglance-maxtext/nnx_ckpt_feat_nnx_post_train_fixes_20260416_181521/nnx_feat_nnx_post_train_fixes_20260416_181521_11_optimizer_offload_false/checkpoints/0 I0416 20:17:43.167414 134259576669760 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_11_optimizer_offload_false/checkpoints/0`. I0416 20:17:43.168070 134259576669760 async_checkpointer.py:420] Finished async_save (blocking + background). Time taken: 34.087335s. directory=gs://wanglance-maxtext/nnx_ckpt_feat_nnx_post_train_fixes_20260416_181521/nnx_feat_nnx_post_train_fixes_20260416_181521_11_optimizer_offload_false/checkpoints/0 I0416 20:17:43.168144 134259576669760 async_checkpointer.py:144] [process=0][thread=async_save] Background save thread done. Time taken: 32.236023s. I0416 20:17:43.168332 134259710887488 async_checkpointer.py:273] [process=0][thread=save_finalize] Done with waiting for background save thread=async_save. I0416 20:17:43.168441 134259710887488 async_checkpointer.py:283] [process=0][thread=save_finalize] No errors found in background save thread=async_save. I0416 20:17:43.168500 134259710887488 checkpoint_manager.py:2103] [process=0][thread=save_finalize][step=0] CheckpointManager Save Finalize is syncing with other hosts... I0416 20:17:43.168547 134259710887488 checkpoint_manager.py:2112] [process=0][thread=save_finalize][step=0] CheckpointManager Save Finalize is done on all hosts. I0416 20:17:43.168662 134367141866624 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 20:17:43.168735 134367141866624 checkpoint_manager.py:1441] Waiting for previous save to complete took 11.358070 seconds. If this number is high, consider checkpointing less frequently. I0416 20:17:43.169453 134367141866624 checkpoint_manager.py:1501] [process=0] Saving checkpoint at step 9 I0416 20:17:43.169705 134367141866624 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_11_optimizer_offload_false/checkpoints/9. I0416 20:17:43.355085 134259710887488 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_11_optimizer_offload_false/checkpoints/9 I0416 20:17:43.365309 134367141866624 jax_array_handlers.py:347] Scheduling D2H of 69 prioritized jax.Array. I0416 20:17:43.365404 134367141866624 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 20:17:44.050766 134259566184000 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_11_optimizer_offload_false/checkpoints/9/items I0416 20:17:52.649073 134367141866624 base_pytree_checkpoint_handler.py:153] [process=0][thread=MainThread] Initiated "orbax.checkpoint._src.serialization.jax_array_handlers.ArrayHandler".serialize. Time taken: 9.284773s I0416 20:17:52.655013 134367141866624 base_pytree_checkpoint_handler.py:128] [process=0] /jax/checkpoint/write/blocking_gbytes_per_sec: 1.314 GiB/s (total gbytes: 12.3 GiB) (time elapsed: 9 seconds) (per-host) I0416 20:17:52.655083 134367141866624 base_pytree_checkpoint_handler.py:732] [process=0][thread=MainThread] Initiated Pytree async_save. Time taken: 9.390954s (batch_requests_ready=0.092891s, total_serialization_initiated=9.292224s, others=0.005838s) I0416 20:17:52.655159 134367141866624 composite_checkpoint_handler.py:715] [process=0][thread=MainThread] Initiated CompositeCheckpointHandler.async_save. Time taken: 9.391478s (all_items=0.000033s, per_item={'items': '0.00003266'}, temp_paths=9.391445) I0416 20:17:52.656606 134259710887488 async_checkpointer.py:79] [process=0][thread=async_save] Background save thread started. I0416 20:17:52.656719 134367141866624 async_checkpointer.py:561] Finished blocking save. Time taken: 9.487216s. Continuing background save to gs://wanglance-maxtext/nnx_ckpt_feat_nnx_post_train_fixes_20260416_181521/nnx_feat_nnx_post_train_fixes_20260416_181521_11_optimizer_offload_false/checkpoints/9. I0416 20:17:52.656935 134367141866624 checkpoint_manager.py:1549] [process=0][thread=MainThread][step=9] Starting CheckpointManager Save Finalize thread=save_finalize I0416 20:17:52.657136 134259576669760 async_checkpointer.py:265] [process=0][thread=save_finalize] Waiting for background save thread=async_save. I0416 20:17:52.657250 134367141866624 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_11_optimizer_offload_false/checkpoints', 'reached_preemption': False, 'preemption_received_at': None, 'synchronous': False, 'wait_for_prev_start_time': 1776370651.8106446, 'wait_for_prev_duration_secs': 11.358069896697998, 'time_between_consecutive_saves_sec': None, 'checkpointer_blocking_start_time': 1776370663.1694849, 'checkpointer_blocking_duration_secs': 9.48734974861145, 'get_old_steps_start_time': 1776370672.6568582, 'get_old_steps_duration_secs': 2.9087066650390625e-05, 'checkpoint_manager_blocking_start_time': 1776370651.8106005, 'checkpoint_manager_blocking_duration_secs': 20.846617698669434} I0416 20:17:52.657381 134367141866624 checkpointing.py:409] Started an asynchronous checkpoint save for step 9 I0416 20:17:52.657415 134367141866624 checkpoint_manager.py:1994] [process=0][thread=MainThread][step=9][wait_until_finished] Waiting for Save Finalize thread (save_finalize) to complete. I0416 20:17:53.405619 134259668944448 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_11_optimizer_offload_false/checkpoints/9/items/array_metadatas/process_0 I0416 20:18:27.889122 134259566184000 base_pytree_checkpoint_handler.py:1217] [process=0][thread=write_metadata_after_commits] Commit + Array metadata written. Time taken: 35.233377s (commit=34.804972s, array_metadata_write=0.428406s) I0416 20:18:27.890163 134259710887488 base_pytree_checkpoint_handler.py:128] [process=0] /jax/checkpoint/write/gbytes_per_sec: 283.171 MiB/s (total gbytes: 12.3 GiB) (time elapsed: 44 seconds) (per-host) I0416 20:18:27.890275 134259710887488 async_checkpointer.py:90] [process=0][thread=async_save] 3 Handler Commit operations completed. Time taken: 35.233511s. I0416 20:18:28.286863 134259710887488 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 20:18:28.712914 134259710887488 ocdbt_utils.py:56] Param validation support for Zarr3 will be added later (b/362328389). I0416 20:18:28.713643 134259710887488 base_pytree_checkpoint_handler.py:1342] [process=0][thread=async_save] Pytree save finalize (merge_ocdbt + ArrayMetadata validation) completed. Time taken: 0.572179s. 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_11_optimizer_offload_false/checkpoints/9/items I0416 20:18:28.714308 134259710887488 atomicity.py:608] Finalizing gs://wanglance-maxtext/nnx_ckpt_feat_nnx_post_train_fixes_20260416_181521/nnx_feat_nnx_post_train_fixes_20260416_181521_11_optimizer_offload_false/checkpoints/9/items I0416 20:18:28.957294 134259710887488 atomicity.py:608] Finalizing gs://wanglance-maxtext/nnx_ckpt_feat_nnx_post_train_fixes_20260416_181521/nnx_feat_nnx_post_train_fixes_20260416_181521_11_optimizer_offload_false/checkpoints/9 I0416 20:18:29.674228 134259710887488 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_11_optimizer_offload_false/checkpoints/9`. I0416 20:18:29.674895 134259710887488 async_checkpointer.py:420] Finished async_save (blocking + background). Time taken: 46.505397s. directory=gs://wanglance-maxtext/nnx_ckpt_feat_nnx_post_train_fixes_20260416_181521/nnx_feat_nnx_post_train_fixes_20260416_181521_11_optimizer_offload_false/checkpoints/9 I0416 20:18:29.674967 134259710887488 async_checkpointer.py:144] [process=0][thread=async_save] Background save thread done. Time taken: 37.018206s. I0416 20:18:29.675121 134259576669760 async_checkpointer.py:273] [process=0][thread=save_finalize] Done with waiting for background save thread=async_save. I0416 20:18:29.675220 134259576669760 async_checkpointer.py:283] [process=0][thread=save_finalize] No errors found in background save thread=async_save. I0416 20:18:29.675276 134259576669760 checkpoint_manager.py:2103] [process=0][thread=save_finalize][step=9] CheckpointManager Save Finalize is syncing with other hosts... I0416 20:18:29.675322 134259576669760 checkpoint_manager.py:2112] [process=0][thread=save_finalize][step=9] CheckpointManager Save Finalize is done on all hosts. I0416 20:18:29.675579 134367141866624 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 20:18:29.675720 134367141866624 checkpoint_manager.py:1983] [process=0][thread=MainThread][wait_until_finished] No Save Finalize thread to wait for. Returning. I0416 20:18:29.676290 134367141866624 metric_logger.py:185] completed step: 9, seconds: 0.118, TFLOP/s/device: 115.305, Tokens/s/device: 17380.088, total_weights: 16384, loss: 5.908 Per train step: Total TFLOPs: 13.59 split as 93.93% learnable weight flops and 6.07% attention flops