2026-04-16 18:53:54.240966: E external/local_xla/xla/stream_executor/cuda/cuda_platform.cc:51] failed call to cuInit: INTERNAL: CUDA error: Failed call to cuInit: UNKNOWN ERROR (303) I0416 18:53:54.356101 133106201832576 max_utils.py:238] Skipping jax distributed system due to skip_jax_distributed_system=True flag. I0416 18:54:42.981728 133106201832576 max_utils.py:800] System Information: Jax Version: 0.9.2 I0416 18:54:42.981839 133106201832576 max_utils.py:801] System Information: Jaxlib Version: 0.9.2 I0416 18:54:42.981879 133106201832576 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 18:54:42.981905 133106201832576 train_utils.py:364] WARNING: 'dataset_path' might be pointing your local file system I0416 18:54:42.981927 133106201832576 train_utils.py:377] WARNING: Sequence packing is essentially ignored for synthetic data. Please use a real dataset to use sequence packing. I0416 18:54:42.982006 133106201832576 train.py:811] [DECOUPLED NO-OP] skipping cloud diagnostics wrapper. W0416 18:54:43.074127 4009027 pjrt_executable.cc:642] Assume version compatibility. PjRt-IFRT does not track XLA executable versions. I0416 18:54:43.470718 133106201832576 maxtext_utils.py:1687] Num_devices: 8, shape (1, 1, 1, 8, 1, 1, 1, 1, 1, 1, 1, 1, 1) I0416 18:54:43.470882 133106201832576 maxtext_utils.py:1687] Num_devices: 8, shape (1, 1, 1, 8, 1, 1, 1, 1, 1, 1, 1, 1, 1) I0416 18:54:43.471101 133106201832576 checkpointing.py:688] Setting up checkpoint logger... I0416 18:54:43.471142 133106201832576 checkpointing.py:234] Creating checkpoint manager with ocdbt=True and zarr3=True I0416 18:54:43.471178 133106201832576 pytree_checkpoint_handler.py:577] save_device_host_concurrent_bytes=None I0416 18:54:43.471690 133106201832576 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 0x790e9cff4d40>, enable_pinned_host_transfer=False, save_concurrent_bytes: 96000000000 (89.4 GiB), restore_concurrent_bytes: 96000000000 (89.4 GiB) I0416 18:54:45.856182 133106201832576 checkpointing.py:266] Enabling policy for fixed interval checkpointing. I0416 18:54:45.856360 133106201832576 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 0x790820d362a0>}, handler_registry=None I0416 18:54:45.856608 133106201832576 composite_checkpoint_handler.py:237] Deferred registration for item: "items". Adding handler `<orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeCheckpointHandler object at 0x790820d362a0>` 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 18:54:45.856648 133106201832576 composite_checkpoint_handler.py:237] Deferred registration for item: "metrics". Adding handler `<orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonCheckpointHandler object at 0x790820d36690>` 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 18:54:45.856677 133106201832576 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 0x790820d362a0>, ('items', <class 'orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeRestoreArgs'>): <orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeCheckpointHandler object at 0x790820d362a0>, ('metrics', <class 'orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonSaveArgs'>): <orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonCheckpointHandler object at 0x790820d36690>, ('metrics', <class 'orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonRestoreArgs'>): <orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonCheckpointHandler object at 0x790820d36690>}). I0416 18:54:45.857168 133106201832576 abstract_checkpointer.py:35] orbax-checkpoint version: 0.11.28 I0416 18:54:45.857219 133106201832576 async_checkpointer.py:177] [process=0][thread=MainThread] Using barrier_sync_fn: <function get_barrier_sync_fn.<locals>.<lambda> at 0x790820d2dee0> timeout: 600 secs and primary_host=0 for async checkpoint writes I0416 18:54:45.974256 133106201832576 checkpoint_manager.py:1788] Found 0 checkpoint steps in gs://wanglance-maxtext/linen_ckpt_feat_nnx_post_train_fixes_20260416_181521/linen_feat_nnx_post_train_fixes_20260416_181521_11_optimizer_offload_true/checkpoints I0416 18:54:45.974469 133106201832576 checkpoint_manager.py:921] [process=0][thread=MainThread] CheckpointManager created, primary_host=0, CheckpointManagerOptions=CheckpointManagerOptions(save_interval_steps=1, max_to_keep=None, keep_time_interval=None, keep_period=None, should_keep_fn=None, best_fn=None, best_mode='max', keep_checkpoints_without_metrics=True, step_prefix=None, step_format_fixed_length=None, step_name_format=None, create=True, cleanup_tmp_directories=False, save_on_steps=frozenset(), single_host_load_and_broadcast=False, todelete_subdir=None, todelete_full_path=None, enable_hns=False, enable_background_delete=False, read_only=False, enable_async_checkpointing=True, async_options=None, multiprocessing_options=MultiprocessingOptions(primary_host=0, active_processes=None, barrier_sync_key_prefix=None), should_save_fn=None, file_options=FileOptions(path_permission_mode=None), save_root_metadata=True, temporary_path_class=None, save_decision_policy=FixedIntervalPolicy(interval=10), preservation_policy=LatestN(n=None), prevent_write_metrics=False, enable_should_save_is_saving_in_progress_check=True, enable_per_process_directory_creation=False), root_directory=gs://wanglance-maxtext/linen_ckpt_feat_nnx_post_train_fixes_20260416_181521/linen_feat_nnx_post_train_fixes_20260416_181521_11_optimizer_offload_true/checkpoints: <orbax.checkpoint.checkpoint_manager.CheckpointManager object at 0x790820d35d00> I0416 18:54:45.974547 133106201832576 checkpointing.py:302] Checkpoint manager created! W0416 18:54:45.990813 4009027 pjrt_executable.cc:642] Assume version compatibility. PjRt-IFRT does not track XLA executable versions. I0416 18:54:46.176808 133106201832576 nnx_wrappers.py:437] Unknown Logical: bfloat16[8,2048,2048]....................................... ('activation_batch', 'activation_norm_length', 'activation_embed'). I0416 18:54:46.176965 133106201832576 nnx_wrappers.py:437] Unknown Physical: bfloat16[8,2048,2048]....................................... ('fsdp', None, None). I0416 18:54:46.269822 133106201832576 attentions.py:1088] attentions/inputs_q Logical: bfloat16[8,2048,2048]....................................... ('activation_batch', 'activation_attn_length', 'activation_attn_embed'). I0416 18:54:46.269898 133106201832576 attentions.py:1088] attentions/inputs_q Physical: bfloat16[8,2048,2048]....................................... ('fsdp', None, None). I0416 18:54:46.284189 133106201832576 attentions.py:1089] attentions/inputs_kv Logical: bfloat16[8,2048,2048]....................................... ('activation_batch', 'activation_attn_length', 'activation_attn_embed'). I0416 18:54:46.284237 133106201832576 attentions.py:1089] attentions/inputs_kv Physical: bfloat16[8,2048,2048]....................................... ('fsdp', None, None). I0416 18:54:46.305492 133106201832576 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 18:54:46.305546 133106201832576 attentions.py:1154] attentions/query Physical: bfloat16[8,2048,16,128]..................................... ('fsdp', None, None, None). I0416 18:54:46.319803 133106201832576 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 18:54:46.319855 133106201832576 attentions.py:1155] attentions/key Physical: bfloat16[8,2048,16,128]..................................... ('fsdp', None, None, None). I0416 18:54:46.334137 133106201832576 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 18:54:46.334184 133106201832576 attentions.py:1156] attentions/value Physical: bfloat16[8,2048,16,128]..................................... ('fsdp', None, None, None). I0416 18:54:46.355912 133106201832576 attentions.py:1197] attentions/out Logical: bfloat16[8,2048,16,128]..................................... ('activation_batch', 'activation_attn_length', 'activation_heads', 'activation_kv'). I0416 18:54:46.355970 133106201832576 attentions.py:1197] attentions/out Physical: bfloat16[8,2048,16,128]..................................... ('fsdp', None, None, None). I0416 18:54:46.374105 133106201832576 linears.py:525] linears/x Logical: bfloat16[8,2048,7168]....................................... ('activation_batch', 'activation_length', 'activation_mlp'). I0416 18:54:46.374156 133106201832576 linears.py:525] linears/x Physical: bfloat16[8,2048,7168]....................................... ('fsdp', None, None). I0416 18:54:46.562571 133106201832576 checkpointing.py:578] checkpoint manager exists so trying to load this run's existing checkpoint I0416 18:54:46.562664 133106201832576 checkpointing.py:676] No existing checkpoints found, not restoring checkpoint. W0416 18:54:46.613690 4009027 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 18:54:47.645298 133106201832576 maxtext_utils.py:1796] params/params/decoder/decoder_norm/scale Shape: float32[2048] Logical: P('norm',) Physical: (None,) I0416 18:54:47.645412 133106201832576 maxtext_utils.py:1796] params/params/decoder/layers/mlp/wi_0/kernel Shape: float32[2048,16,7168] Logical: P('embed', 'layers', 'mlp') Physical: ('fsdp', None, None) I0416 18:54:47.645452 133106201832576 maxtext_utils.py:1796] params/params/decoder/layers/mlp/wi_1/kernel Shape: float32[2048,16,7168] Logical: P('embed', 'layers', 'mlp') Physical: ('fsdp', None, None) I0416 18:54:47.645497 133106201832576 maxtext_utils.py:1796] params/params/decoder/layers/mlp/wo/kernel Shape: float32[7168,16,2048] Logical: P('mlp', 'layers', 'embed') Physical: (None, None, 'fsdp') I0416 18:54:47.645539 133106201832576 maxtext_utils.py:1796] params/params/decoder/layers/post_self_attention_layer_norm/scale Shape: float32[2048,16] Logical: P('norm', 'layers') Physical: (None, None) I0416 18:54:47.645571 133106201832576 maxtext_utils.py:1796] params/params/decoder/layers/pre_self_attention_layer_norm/scale Shape: float32[2048,16] Logical: P('norm', 'layers') Physical: (None, None) I0416 18:54:47.645612 133106201832576 maxtext_utils.py:1796] params/params/decoder/layers/self_attention/key/kernel Shape: float32[2048,16,16,128] Logical: P('embed', 'layers', 'kv_heads', 'kv_head_dim') Physical: ('fsdp', None, None, None) I0416 18:54:47.645649 133106201832576 maxtext_utils.py:1796] params/params/decoder/layers/self_attention/out/kernel Shape: float32[16,16,128,2048] Logical: P('heads', 'layers', 'kv', 'embed') Physical: (None, None, None, 'fsdp') I0416 18:54:47.645677 133106201832576 maxtext_utils.py:1796] params/params/decoder/layers/self_attention/query/kernel Shape: float32[2048,16,16,128] Logical: P('embed', 'layers', 'q_heads', 'kv') Physical: ('fsdp', None, None, None) I0416 18:54:47.645702 133106201832576 maxtext_utils.py:1796] params/params/decoder/layers/self_attention/value/kernel Shape: float32[2048,16,16,128] Logical: P('embed', 'layers', 'kv_heads', 'kv_head_dim') Physical: ('fsdp', None, None, None) I0416 18:54:47.645737 133106201832576 maxtext_utils.py:1796] params/params/decoder/logits_dense/kernel Shape: float32[2048,32000] Logical: P('embed_vocab', 'vocab') Physical: ('fsdp', None) I0416 18:54:47.645775 133106201832576 maxtext_utils.py:1796] params/params/token_embedder/embedding Shape: float32[32000,2048] Logical: P('vocab', 'embed_vocab') Physical: (None, 'fsdp') I0416 18:54:48.472826 133106201832576 train.py:157] train/xent Logical: float32[8,2048]............................................. ('activation_embed_and_logits_batch', 'activation_length'). I0416 18:54:48.472917 133106201832576 train.py:157] train/xent Physical: float32[8,2048]............................................. ('fsdp', None). I0416 18:54:48.487017 133106201832576 train.py:164] train/z_loss Logical: float32[8,2048]............................................. ('activation_embed_and_logits_batch', 'activation_length'). I0416 18:54:48.487082 133106201832576 train.py:164] train/z_loss Physical: float32[8,2048]............................................. ('fsdp', None). W0416 18:54:49.217168 4009027 pjrt_executable.cc:642] Assume version compatibility. PjRt-IFRT does not track XLA executable versions. I0416 18:54:49.358650 133106201832576 max_utils.py:791] Total memory size: 2.4 GB, Output size: 0.5 GB, Temp size: 1.9 GB, Argument size: 0.5 GB, Host temp size: 0.0 GB. I0416 18:54:50.629959 133106201832576 max_utils.py:194] tensorboardX not available; using no-op SummaryWriter. I0416 18:54:50.630217 133106201832576 metric_logger.py:289] number parameters: 1.104 billion W0416 18:54:51.470943 4009027 pjrt_executable.cc:642] Assume version compatibility. PjRt-IFRT does not track XLA executable versions. I0416 18:54:51.591926 133106201832576 checkpointing.py:794] Waiting for step 0 to finish before checkpoint... I0416 18:54:51.781543 133106201832576 checkpointing.py:798] Waited 0.18960809707641602 seconds for step 0 to finish before starting checkpointing. I0416 18:54:51.781977 133106201832576 checkpoint_manager.py:1983] [process=0][thread=MainThread][wait_until_finished] No Save Finalize thread to wait for. Returning. I0416 18:54:51.782158 133106201832576 checkpoint_manager.py:1501] [process=0] Saving checkpoint at step 0 I0416 18:54:51.782515 133106201832576 async_checkpointer.py:452] [process=0] Started async saving checkpoint to gs://wanglance-maxtext/linen_ckpt_feat_nnx_post_train_fixes_20260416_181521/linen_feat_nnx_post_train_fixes_20260416_181521_11_optimizer_offload_true/checkpoints/0. I0416 18:54:51.868978 133106201832576 signaling_client.py:373] Using ThreadSafeKeyValueSignalingClient I0416 18:54:51.960985 133106201832576 jax_array_handlers.py:347] Scheduling D2H of 39 prioritized jax.Array. I0416 18:54:51.961082 133106201832576 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 18:54:51.975282 132998794053184 atomicity.py:137] Creating tmp directory gs://wanglance-maxtext/linen_ckpt_feat_nnx_post_train_fixes_20260416_181521/linen_feat_nnx_post_train_fixes_20260416_181521_11_optimizer_offload_true/checkpoints/0 W0416 18:54:52.177074 4009027 pjrt_executable.cc:642] Assume version compatibility. PjRt-IFRT does not track XLA executable versions. W0416 18:54:52.189620 4009027 pjrt_executable.cc:642] Assume version compatibility. PjRt-IFRT does not track XLA executable versions. W0416 18:54:52.201283 4009027 pjrt_executable.cc:642] Assume version compatibility. PjRt-IFRT does not track XLA executable versions. W0416 18:54:52.208522 4009027 pjrt_executable.cc:642] Assume version compatibility. PjRt-IFRT does not track XLA executable versions. W0416 18:54:52.215950 4009027 pjrt_executable.cc:642] Assume version compatibility. PjRt-IFRT does not track XLA executable versions. W0416 18:54:52.221850 4009027 pjrt_executable.cc:642] Assume version compatibility. PjRt-IFRT does not track XLA executable versions. W0416 18:54:52.226751 4009027 pjrt_executable.cc:642] Assume version compatibility. PjRt-IFRT does not track XLA executable versions. W0416 18:54:52.231314 4009027 pjrt_executable.cc:642] Assume version compatibility. PjRt-IFRT does not track XLA executable versions. I0416 18:54:52.402294 133106201832576 base_pytree_checkpoint_handler.py:153] [process=0][thread=MainThread] Initiated "orbax.checkpoint._src.serialization.jax_array_handlers.ArrayHandler".serialize. Time taken: 0.441697s I0416 18:54:52.402824 133106201832576 base_pytree_checkpoint_handler.py:128] [process=0] /jax/checkpoint/write/blocking_gbytes_per_sec: 23.161 GiB/s (total gbytes: 12.3 GiB) (time elapsed: 532 milliseconds) (per-host) I0416 18:54:52.402895 133106201832576 base_pytree_checkpoint_handler.py:732] [process=0][thread=MainThread] Initiated Pytree async_save. Time taken: 0.532992s (batch_requests_ready=0.087103s, total_serialization_initiated=0.445517s, others=0.000372s) I0416 18:54:52.402976 133106201832576 composite_checkpoint_handler.py:715] [process=0][thread=MainThread] Initiated CompositeCheckpointHandler.async_save. Time taken: 0.533537s (all_items=0.000024s, per_item={'items': '0.00002408'}, temp_paths=0.533513) I0416 18:54:52.403872 132998731138624 async_checkpointer.py:79] [process=0][thread=async_save] Background save thread started. I0416 18:54:52.403992 133106201832576 async_checkpointer.py:561] Finished blocking save. Time taken: 0.621785s. Continuing background save to gs://wanglance-maxtext/linen_ckpt_feat_nnx_post_train_fixes_20260416_181521/linen_feat_nnx_post_train_fixes_20260416_181521_11_optimizer_offload_true/checkpoints/0. I0416 18:54:52.404240 133106201832576 checkpoint_manager.py:1549] [process=0][thread=MainThread][step=0] Starting CheckpointManager Save Finalize thread=save_finalize I0416 18:54:52.404482 132998804538944 async_checkpointer.py:265] [process=0][thread=save_finalize] Waiting for background save thread=async_save. I0416 18:54:52.404618 133106201832576 standard_logger.py:34] {'step': 0, 'event_type': 'save', 'directory': 'gs://wanglance-maxtext/linen_ckpt_feat_nnx_post_train_fixes_20260416_181521/linen_feat_nnx_post_train_fixes_20260416_181521_11_optimizer_offload_true/checkpoints', 'reached_preemption': False, 'preemption_received_at': None, 'synchronous': False, 'wait_for_prev_start_time': 1776365691.781964, 'wait_for_prev_duration_secs': 4.076957702636719e-05, 'time_between_consecutive_saves_sec': None, 'checkpointer_blocking_start_time': 1776365691.782183, 'checkpointer_blocking_duration_secs': 0.6219563484191895, 'get_old_steps_start_time': 1776365692.4041648, 'get_old_steps_duration_secs': 3.0279159545898438e-05, 'checkpoint_manager_blocking_start_time': 1776365691.7818813, 'checkpoint_manager_blocking_duration_secs': 0.6226918697357178} I0416 18:54:52.404777 133106201832576 checkpointing.py:409] Started an asynchronous checkpoint save for step 0 I0416 18:54:52.404828 133106201832576 max_utils.py:750] Memstats: After params initialized: I0416 18:54:52.404931 133106201832576 max_utils.py:756] Using (GB) 0.57 / 31.25 (1.824000%) on TPU_0(process=0,(0,0,0,0)) I0416 18:54:52.404962 133106201832576 max_utils.py:756] Using (GB) 0.56 / 31.25 (1.792000%) on TPU_1(process=0,(1,0,0,0)) I0416 18:54:52.404983 133106201832576 max_utils.py:756] Using (GB) 0.56 / 31.25 (1.792000%) on TPU_2(process=0,(0,1,0,0)) I0416 18:54:52.405001 133106201832576 max_utils.py:756] Using (GB) 0.56 / 31.25 (1.792000%) on TPU_3(process=0,(1,1,0,0)) I0416 18:54:52.405019 133106201832576 max_utils.py:756] Using (GB) 0.56 / 31.25 (1.792000%) on TPU_4(process=0,(0,2,0,0)) I0416 18:54:52.405037 133106201832576 max_utils.py:756] Using (GB) 0.56 / 31.25 (1.792000%) on TPU_5(process=0,(1,2,0,0)) I0416 18:54:52.405070 133106201832576 max_utils.py:756] Using (GB) 0.56 / 31.25 (1.792000%) on TPU_6(process=0,(0,3,0,0)) I0416 18:54:52.405088 133106201832576 max_utils.py:756] Using (GB) 0.56 / 31.25 (1.792000%) on TPU_7(process=0,(1,3,0,0)) W0416 18:54:52.411260 4009027 pjrt_executable.cc:642] Assume version compatibility. PjRt-IFRT does not track XLA executable versions. W0416 18:54:52.417022 4009027 pjrt_executable.cc:642] Assume version compatibility. PjRt-IFRT does not track XLA executable versions. W0416 18:54:52.421789 4009027 pjrt_executable.cc:642] Assume version compatibility. PjRt-IFRT does not track XLA executable versions. I0416 18:54:52.739447 133106201832576 metric_logger.py:185] completed step: 0, seconds: 0.961, TFLOP/s/device: 14.135, Tokens/s/device: 2130.617, total_weights: 16384, loss: 10.874 I0416 18:54:52.740166 133106201832576 metric_logger.py:269] To see full metrics 'tensorboard --logdir=gs://wanglance-maxtext/linen_ckpt_feat_nnx_post_train_fixes_20260416_181521/linen_feat_nnx_post_train_fixes_20260416_181521_11_optimizer_offload_true/tensorboard/' I0416 18:54:52.923775 133106201832576 metric_logger.py:185] completed step: 1, seconds: 1.145, TFLOP/s/device: 11.862, Tokens/s/device: 1788.022, total_weights: 16384, loss: 10.874 I0416 18:54:52.944504 132998783567424 atomicity.py:137] Creating tmp directory gs://wanglance-maxtext/linen_ckpt_feat_nnx_post_train_fixes_20260416_181521/linen_feat_nnx_post_train_fixes_20260416_181521_11_optimizer_offload_true/checkpoints/0/items I0416 18:54:53.092635 132998640961088 checkpoint.py:188] Wrote Metadata={'item_handlers': None, 'metrics': {}, 'performance_metrics': {}, 'init_timestamp_nsecs': 1776365692855124901, 'commit_timestamp_nsecs': None, 'custom_metadata': {}}, json={"item_handlers": null, "metrics": {}, "performance_metrics": {}, "init_timestamp_nsecs": 1776365692855124901, "commit_timestamp_nsecs": null, "custom_metadata": {}} to gs://wanglance-maxtext/linen_ckpt_feat_nnx_post_train_fixes_20260416_181521/linen_feat_nnx_post_train_fixes_20260416_181521_11_optimizer_offload_true/checkpoints/0/_CHECKPOINT_METADATA I0416 18:54:53.112881 133106201832576 metric_logger.py:185] completed step: 2, seconds: 0.007, TFLOP/s/device: 2080.724, Tokens/s/device: 313629.403, total_weights: 16384, loss: 9.775 I0416 18:54:53.303335 133106201832576 metric_logger.py:185] completed step: 3, seconds: 0.186, TFLOP/s/device: 73.141, Tokens/s/device: 11024.563, total_weights: 16384, loss: 8.838 I0416 18:54:53.408238 4011180 google_auth_provider.cc:149] Using credentials at ~/.config/gcloud/application_default_credentials.json I0416 18:54:53.408299 4011180 google_auth_provider.cc:156] Using OAuth2 AuthProvider I0416 18:54:54.158078 132998752110144 array_metadata_store.py:203] [process=0][thread=array_type_handler] Wrote 39 array_metadata.ArrayMetadata to gs://wanglance-maxtext/linen_ckpt_feat_nnx_post_train_fixes_20260416_181521/linen_feat_nnx_post_train_fixes_20260416_181521_11_optimizer_offload_true/checkpoints/0/items/array_metadatas/process_0 I0416 18:55:13.361019 133106201832576 metric_logger.py:185] completed step: 4, seconds: 0.189, TFLOP/s/device: 71.836, Tokens/s/device: 10827.958, total_weights: 16384, loss: 7.965 I0416 18:55:13.368899 133106201832576 metric_logger.py:185] completed step: 5, seconds: 0.191, TFLOP/s/device: 71.273, Tokens/s/device: 10743.043, total_weights: 16384, loss: 7.233 I0416 18:55:13.553997 133106201832576 metric_logger.py:185] completed step: 6, seconds: 20.058, TFLOP/s/device: 0.677, Tokens/s/device: 102.102, total_weights: 16384, loss: 6.677 I0416 18:55:13.743310 133106201832576 metric_logger.py:185] completed step: 7, seconds: 0.005, TFLOP/s/device: 3002.680, Tokens/s/device: 452596.685, total_weights: 16384, loss: 6.293 I0416 18:55:13.932801 133106201832576 metric_logger.py:185] completed step: 8, seconds: 0.186, TFLOP/s/device: 73.061, Tokens/s/device: 11012.588, total_weights: 16384, loss: 6.055 I0416 18:55:14.121809 133106201832576 checkpointing.py:794] Waiting for step 9 to finish before checkpoint... I0416 18:55:14.126137 133106201832576 checkpointing.py:798] Waited 0.004359245300292969 seconds for step 9 to finish before starting checkpointing. I0416 18:55:14.126416 133106201832576 checkpoint_manager.py:1994] [process=0][thread=MainThread][step=0][wait_until_finished] Waiting for Save Finalize thread (save_finalize) to complete. I0416 18:55:19.817744 132998741624384 base_pytree_checkpoint_handler.py:1217] [process=0][thread=write_metadata_after_commits] Commit + Array metadata written. Time taken: 26.414056s (commit=25.978675s, array_metadata_write=0.435381s) I0416 18:55:19.818834 132998731138624 base_pytree_checkpoint_handler.py:128] [process=0] /jax/checkpoint/write/gbytes_per_sec: 452.140 MiB/s (total gbytes: 12.3 GiB) (time elapsed: 27 seconds) (per-host) I0416 18:55:19.818946 132998731138624 async_checkpointer.py:90] [process=0][thread=async_save] 3 Handler Commit operations completed. Time taken: 27.414885s. I0416 18:55:20.042402 132998731138624 checkpoint.py:228] Read Metadata={'item_handlers': None, 'metrics': {}, 'performance_metrics': {}, 'init_timestamp_nsecs': 1776365692855124901, 'commit_timestamp_nsecs': None, 'custom_metadata': {}} from gs://wanglance-maxtext/linen_ckpt_feat_nnx_post_train_fixes_20260416_181521/linen_feat_nnx_post_train_fixes_20260416_181521_11_optimizer_offload_true/checkpoints/0/_CHECKPOINT_METADATA I0416 18:55:20.224732 132998731138624 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 18:55:20.435221 132998640961088 checkpoint.py:247] Updated Metadata={'item_handlers': {'items': 'orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeCheckpointHandler'}, 'metrics': {}, 'performance_metrics': {}, 'init_timestamp_nsecs': 1776365692855124901, 'commit_timestamp_nsecs': None, 'custom_metadata': {}} to gs://wanglance-maxtext/linen_ckpt_feat_nnx_post_train_fixes_20260416_181521/linen_feat_nnx_post_train_fixes_20260416_181521_11_optimizer_offload_true/checkpoints/0/_CHECKPOINT_METADATA I0416 18:55:20.628150 132998731138624 ocdbt_utils.py:56] Param validation support for Zarr3 will be added later (b/362328389). I0416 18:55:20.628882 132998731138624 base_pytree_checkpoint_handler.py:1342] [process=0][thread=async_save] Pytree save finalize (merge_ocdbt + ArrayMetadata validation) completed. Time taken: 0.546689s. use_zarr3=True, enable_post_merge_validation=True, directory=gs://wanglance-maxtext/linen_ckpt_feat_nnx_post_train_fixes_20260416_181521/linen_feat_nnx_post_train_fixes_20260416_181521_11_optimizer_offload_true/checkpoints/0/items I0416 18:55:20.629531 132998731138624 atomicity.py:608] Finalizing gs://wanglance-maxtext/linen_ckpt_feat_nnx_post_train_fixes_20260416_181521/linen_feat_nnx_post_train_fixes_20260416_181521_11_optimizer_offload_true/checkpoints/0/items I0416 18:55:20.883090 132998731138624 atomicity.py:608] Finalizing gs://wanglance-maxtext/linen_ckpt_feat_nnx_post_train_fixes_20260416_181521/linen_feat_nnx_post_train_fixes_20260416_181521_11_optimizer_offload_true/checkpoints/0 I0416 18:55:21.521881 132998731138624 atomicity.py:794] [process=0][thread=async_save] Finished saving checkpoint (finalized tmp dir) to `gs://wanglance-maxtext/linen_ckpt_feat_nnx_post_train_fixes_20260416_181521/linen_feat_nnx_post_train_fixes_20260416_181521_11_optimizer_offload_true/checkpoints/0`. I0416 18:55:21.522507 132998731138624 async_checkpointer.py:420] Finished async_save (blocking + background). Time taken: 29.740312s. directory=gs://wanglance-maxtext/linen_ckpt_feat_nnx_post_train_fixes_20260416_181521/linen_feat_nnx_post_train_fixes_20260416_181521_11_optimizer_offload_true/checkpoints/0 I0416 18:55:21.522582 132998731138624 async_checkpointer.py:144] [process=0][thread=async_save] Background save thread done. Time taken: 29.118523s. I0416 18:55:21.522782 132998804538944 async_checkpointer.py:273] [process=0][thread=save_finalize] Done with waiting for background save thread=async_save. I0416 18:55:21.522890 132998804538944 async_checkpointer.py:283] [process=0][thread=save_finalize] No errors found in background save thread=async_save. I0416 18:55:21.522948 132998804538944 checkpoint_manager.py:2103] [process=0][thread=save_finalize][step=0] CheckpointManager Save Finalize is syncing with other hosts... I0416 18:55:21.522993 132998804538944 checkpoint_manager.py:2112] [process=0][thread=save_finalize][step=0] CheckpointManager Save Finalize is done on all hosts. I0416 18:55:21.523277 133106201832576 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 18:55:21.523405 133106201832576 checkpoint_manager.py:1441] Waiting for previous save to complete took 7.396998 seconds. If this number is high, consider checkpointing less frequently. I0416 18:55:21.524180 133106201832576 checkpoint_manager.py:1501] [process=0] Saving checkpoint at step 9 I0416 18:55:21.524464 133106201832576 async_checkpointer.py:452] [process=0] Started async saving checkpoint to gs://wanglance-maxtext/linen_ckpt_feat_nnx_post_train_fixes_20260416_181521/linen_feat_nnx_post_train_fixes_20260416_181521_11_optimizer_offload_true/checkpoints/9. I0416 18:55:21.690180 132998804538944 atomicity.py:137] Creating tmp directory gs://wanglance-maxtext/linen_ckpt_feat_nnx_post_train_fixes_20260416_181521/linen_feat_nnx_post_train_fixes_20260416_181521_11_optimizer_offload_true/checkpoints/9 I0416 18:55:21.695583 133106201832576 jax_array_handlers.py:347] Scheduling D2H of 39 prioritized jax.Array. I0416 18:55:21.695675 133106201832576 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 18:55:21.983898 133106201832576 base_pytree_checkpoint_handler.py:153] [process=0][thread=MainThread] Initiated "orbax.checkpoint._src.serialization.jax_array_handlers.ArrayHandler".serialize. Time taken: 0.288691s I0416 18:55:21.984357 133106201832576 base_pytree_checkpoint_handler.py:128] [process=0] /jax/checkpoint/write/blocking_gbytes_per_sec: 32.517 GiB/s (total gbytes: 12.3 GiB) (time elapsed: 379 milliseconds) (per-host) I0416 18:55:21.984416 133106201832576 base_pytree_checkpoint_handler.py:732] [process=0][thread=MainThread] Initiated Pytree async_save. Time taken: 0.379627s (batch_requests_ready=0.087595s, total_serialization_initiated=0.291744s, others=0.000288s) I0416 18:55:21.984486 133106201832576 composite_checkpoint_handler.py:715] [process=0][thread=MainThread] Initiated CompositeCheckpointHandler.async_save. Time taken: 0.380128s (all_items=0.000016s, per_item={'items': '0.00001645'}, temp_paths=0.380111) I0416 18:55:21.985540 132998691292736 async_checkpointer.py:79] [process=0][thread=async_save] Background save thread started. I0416 18:55:21.985666 133106201832576 async_checkpointer.py:561] Finished blocking save. Time taken: 0.461437s. Continuing background save to gs://wanglance-maxtext/linen_ckpt_feat_nnx_post_train_fixes_20260416_181521/linen_feat_nnx_post_train_fixes_20260416_181521_11_optimizer_offload_true/checkpoints/9. I0416 18:55:21.986034 133106201832576 checkpoint_manager.py:1549] [process=0][thread=MainThread][step=9] Starting CheckpointManager Save Finalize thread=save_finalize I0416 18:55:21.986352 132998680806976 async_checkpointer.py:265] [process=0][thread=save_finalize] Waiting for background save thread=async_save. I0416 18:55:21.986454 133106201832576 standard_logger.py:34] {'step': 9, 'event_type': 'save', 'directory': 'gs://wanglance-maxtext/linen_ckpt_feat_nnx_post_train_fixes_20260416_181521/linen_feat_nnx_post_train_fixes_20260416_181521_11_optimizer_offload_true/checkpoints', 'reached_preemption': False, 'preemption_received_at': None, 'synchronous': False, 'wait_for_prev_start_time': 1776365714.1263754, 'wait_for_prev_duration_secs': 7.396998405456543, 'time_between_consecutive_saves_sec': None, 'checkpointer_blocking_start_time': 1776365721.5242047, 'checkpointer_blocking_duration_secs': 0.4615929126739502, 'get_old_steps_start_time': 1776365721.9858947, 'get_old_steps_duration_secs': 5.888938903808594e-05, 'checkpoint_manager_blocking_start_time': 1776365714.1263344, 'checkpoint_manager_blocking_duration_secs': 7.860077857971191} I0416 18:55:21.986595 133106201832576 checkpointing.py:409] Started an asynchronous checkpoint save for step 9 I0416 18:55:21.986633 133106201832576 checkpoint_manager.py:1994] [process=0][thread=MainThread][step=9][wait_until_finished] Waiting for Save Finalize thread (save_finalize) to complete. I0416 18:55:22.408983 132998619989568 atomicity.py:137] Creating tmp directory gs://wanglance-maxtext/linen_ckpt_feat_nnx_post_train_fixes_20260416_181521/linen_feat_nnx_post_train_fixes_20260416_181521_11_optimizer_offload_true/checkpoints/9/items I0416 18:55:23.626155 132998712264256 array_metadata_store.py:203] [process=0][thread=array_type_handler] Wrote 39 array_metadata.ArrayMetadata to gs://wanglance-maxtext/linen_ckpt_feat_nnx_post_train_fixes_20260416_181521/linen_feat_nnx_post_train_fixes_20260416_181521_11_optimizer_offload_true/checkpoints/9/items/array_metadatas/process_0 I0416 18:55:52.512449 132998701778496 base_pytree_checkpoint_handler.py:1217] [process=0][thread=write_metadata_after_commits] Commit + Array metadata written. Time taken: 29.628120s (commit=29.149262s, array_metadata_write=0.478858s) I0416 18:55:52.513501 132998691292736 base_pytree_checkpoint_handler.py:128] [process=0] /jax/checkpoint/write/gbytes_per_sec: 408.843 MiB/s (total gbytes: 12.3 GiB) (time elapsed: 30 seconds) (per-host) I0416 18:55:52.513607 132998691292736 async_checkpointer.py:90] [process=0][thread=async_save] 3 Handler Commit operations completed. Time taken: 30.527879s. I0416 18:55:52.925256 132998691292736 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 18:55:53.362299 132998691292736 ocdbt_utils.py:56] Param validation support for Zarr3 will be added later (b/362328389). I0416 18:55:53.362470 132998691292736 base_pytree_checkpoint_handler.py:1342] [process=0][thread=async_save] Pytree save finalize (merge_ocdbt + ArrayMetadata validation) completed. Time taken: 0.582904s. use_zarr3=True, enable_post_merge_validation=True, directory=gs://wanglance-maxtext/linen_ckpt_feat_nnx_post_train_fixes_20260416_181521/linen_feat_nnx_post_train_fixes_20260416_181521_11_optimizer_offload_true/checkpoints/9/items I0416 18:55:53.363119 132998691292736 atomicity.py:608] Finalizing gs://wanglance-maxtext/linen_ckpt_feat_nnx_post_train_fixes_20260416_181521/linen_feat_nnx_post_train_fixes_20260416_181521_11_optimizer_offload_true/checkpoints/9/items I0416 18:55:53.618549 132998691292736 atomicity.py:608] Finalizing gs://wanglance-maxtext/linen_ckpt_feat_nnx_post_train_fixes_20260416_181521/linen_feat_nnx_post_train_fixes_20260416_181521_11_optimizer_offload_true/checkpoints/9 I0416 18:55:54.318362 132998691292736 atomicity.py:794] [process=0][thread=async_save] Finished saving checkpoint (finalized tmp dir) to `gs://wanglance-maxtext/linen_ckpt_feat_nnx_post_train_fixes_20260416_181521/linen_feat_nnx_post_train_fixes_20260416_181521_11_optimizer_offload_true/checkpoints/9`. I0416 18:55:54.318998 132998691292736 async_checkpointer.py:420] Finished async_save (blocking + background). Time taken: 32.794781s. directory=gs://wanglance-maxtext/linen_ckpt_feat_nnx_post_train_fixes_20260416_181521/linen_feat_nnx_post_train_fixes_20260416_181521_11_optimizer_offload_true/checkpoints/9 I0416 18:55:54.319099 132998691292736 async_checkpointer.py:144] [process=0][thread=async_save] Background save thread done. Time taken: 32.333375s. I0416 18:55:54.319224 132998680806976 async_checkpointer.py:273] [process=0][thread=save_finalize] Done with waiting for background save thread=async_save. I0416 18:55:54.319273 132998680806976 async_checkpointer.py:283] [process=0][thread=save_finalize] No errors found in background save thread=async_save. I0416 18:55:54.319331 132998680806976 checkpoint_manager.py:2103] [process=0][thread=save_finalize][step=9] CheckpointManager Save Finalize is syncing with other hosts... I0416 18:55:54.319371 132998680806976 checkpoint_manager.py:2112] [process=0][thread=save_finalize][step=9] CheckpointManager Save Finalize is done on all hosts. I0416 18:55:54.319648 133106201832576 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 18:55:54.319786 133106201832576 checkpoint_manager.py:1983] [process=0][thread=MainThread][wait_until_finished] No Save Finalize thread to wait for. Returning. I0416 18:55:54.320476 133106201832576 metric_logger.py:185] completed step: 9, seconds: 0.190, TFLOP/s/device: 71.673, Tokens/s/device: 10803.340, 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