2026-04-16 03:49:39.725648: 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 03:49:39.843275 136083091008640 max_utils.py:238] Skipping jax distributed system due to skip_jax_distributed_system=True flag. I0416 03:50:32.589668 136083091008640 max_utils.py:800] System Information: Jax Version: 0.9.2 I0416 03:50:32.589772 136083091008640 max_utils.py:801] System Information: Jaxlib Version: 0.9.2 I0416 03:50:32.589807 136083091008640 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 03:50:32.589831 136083091008640 train_utils.py:364] WARNING: 'dataset_path' might be pointing your local file system I0416 03:50:32.589859 136083091008640 train_utils.py:377] WARNING: Sequence packing is essentially ignored for synthetic data. Please use a real dataset to use sequence packing. I0416 03:50:32.589935 136083091008640 train.py:818] [DECOUPLED NO-OP] skipping cloud diagnostics wrapper. W0416 03:50:32.681823 3640194 pjrt_executable.cc:642] Assume version compatibility. PjRt-IFRT does not track XLA executable versions. I0416 03:50:33.078442 136083091008640 maxtext_utils.py:1687] Num_devices: 8, shape (1, 1, 1, 8, 1, 1, 1, 1, 1, 1, 1, 1, 1) I0416 03:50:33.078612 136083091008640 maxtext_utils.py:1687] Num_devices: 8, shape (1, 1, 1, 8, 1, 1, 1, 1, 1, 1, 1, 1, 1) I0416 03:50:33.078779 136083091008640 checkpointing.py:688] Setting up checkpoint logger... I0416 03:50:33.078819 136083091008640 checkpointing.py:234] Creating checkpoint manager with ocdbt=True and zarr3=True I0416 03:50:33.078861 136083091008640 pytree_checkpoint_handler.py:577] save_device_host_concurrent_bytes=None I0416 03:50:33.079324 136083091008640 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 0x7bc3b99f4f50>, enable_pinned_host_transfer=False, save_concurrent_bytes: 96000000000 (89.4 GiB), restore_concurrent_bytes: 96000000000 (89.4 GiB) I0416 03:50:35.509505 136083091008640 checkpointing.py:266] Enabling policy for fixed interval checkpointing. I0416 03:50:35.509685 136083091008640 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 0x7bbd3efd5820>}, handler_registry=None I0416 03:50:35.509932 136083091008640 composite_checkpoint_handler.py:237] Deferred registration for item: "items". Adding handler `<orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeCheckpointHandler object at 0x7bbd3efd5820>` 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 03:50:35.509980 136083091008640 composite_checkpoint_handler.py:237] Deferred registration for item: "metrics". Adding handler `<orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonCheckpointHandler object at 0x7bbd3d731010>` 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 03:50:35.510009 136083091008640 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 0x7bbd3efd5820>, ('items', <class 'orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeRestoreArgs'>): <orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeCheckpointHandler object at 0x7bbd3efd5820>, ('metrics', <class 'orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonSaveArgs'>): <orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonCheckpointHandler object at 0x7bbd3d731010>, ('metrics', <class 'orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonRestoreArgs'>): <orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonCheckpointHandler object at 0x7bbd3d731010>}). I0416 03:50:35.510319 136083091008640 abstract_checkpointer.py:35] orbax-checkpoint version: 0.11.28 I0416 03:50:35.510373 136083091008640 async_checkpointer.py:177] [process=0][thread=MainThread] Using barrier_sync_fn: <function get_barrier_sync_fn.<locals>.<lambda> at 0x7bbd3d73e0c0> timeout: 600 secs and primary_host=0 for async checkpoint writes I0416 03:50:35.632673 136083091008640 checkpoint_manager.py:1788] Found 0 checkpoint steps in gs://wanglance-maxtext/linen_ckpt_feat_nnx_trainstate_and_training_loop_20260416_033532/linen_feat_nnx_trainstate_and_training_loop_20260416_033532_06_grad_accum/checkpoints I0416 03:50:35.632893 136083091008640 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_trainstate_and_training_loop_20260416_033532/linen_feat_nnx_trainstate_and_training_loop_20260416_033532_06_grad_accum/checkpoints: <orbax.checkpoint.checkpoint_manager.CheckpointManager object at 0x7bbeb4ca9a00> I0416 03:50:35.632973 136083091008640 checkpointing.py:302] Checkpoint manager created! I0416 03:50:36.495006 136083091008640 nnx_wrappers.py:437] Unknown Logical: bfloat16[8,2048,2048]....................................... ('activation_batch', 'activation_norm_length', 'activation_embed'). I0416 03:50:36.495110 136083091008640 nnx_wrappers.py:437] Unknown Physical: bfloat16[8,2048,2048]....................................... ('fsdp', None, None). I0416 03:50:36.583889 136083091008640 attentions.py:1088] attentions/inputs_q Logical: bfloat16[8,2048,2048]....................................... ('activation_batch', 'activation_attn_length', 'activation_attn_embed'). I0416 03:50:36.583971 136083091008640 attentions.py:1088] attentions/inputs_q Physical: bfloat16[8,2048,2048]....................................... ('fsdp', None, None). I0416 03:50:36.601027 136083091008640 attentions.py:1089] attentions/inputs_kv Logical: bfloat16[8,2048,2048]....................................... ('activation_batch', 'activation_attn_length', 'activation_attn_embed'). I0416 03:50:36.601091 136083091008640 attentions.py:1089] attentions/inputs_kv Physical: bfloat16[8,2048,2048]....................................... ('fsdp', None, None). I0416 03:50:36.622358 136083091008640 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 03:50:36.622414 136083091008640 attentions.py:1154] attentions/query Physical: bfloat16[8,2048,16,128]..................................... ('fsdp', None, None, None). I0416 03:50:36.636605 136083091008640 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 03:50:36.636655 136083091008640 attentions.py:1155] attentions/key Physical: bfloat16[8,2048,16,128]..................................... ('fsdp', None, None, None). I0416 03:50:36.650837 136083091008640 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 03:50:36.650885 136083091008640 attentions.py:1156] attentions/value Physical: bfloat16[8,2048,16,128]..................................... ('fsdp', None, None, None). I0416 03:50:36.672422 136083091008640 attentions.py:1197] attentions/out Logical: bfloat16[8,2048,16,128]..................................... ('activation_batch', 'activation_attn_length', 'activation_heads', 'activation_kv'). I0416 03:50:36.672477 136083091008640 attentions.py:1197] attentions/out Physical: bfloat16[8,2048,16,128]..................................... ('fsdp', None, None, None). I0416 03:50:36.690450 136083091008640 linears.py:525] linears/x Logical: bfloat16[8,2048,7168]....................................... ('activation_batch', 'activation_length', 'activation_mlp'). I0416 03:50:36.690500 136083091008640 linears.py:525] linears/x Physical: bfloat16[8,2048,7168]....................................... ('fsdp', None, None). I0416 03:50:36.886079 136083091008640 checkpointing.py:578] checkpoint manager exists so trying to load this run's existing checkpoint I0416 03:50:36.886301 136083091008640 checkpointing.py:676] No existing checkpoints found, not restoring checkpoint. W0416 03:50:36.939535 3640194 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 03:50:37.330021 136083091008640 maxtext_utils.py:1790] params/params/decoder/decoder_norm/scale Shape: float32[2048] Logical: P('norm',) Physical: (None,) I0416 03:50:37.330147 136083091008640 maxtext_utils.py:1790] params/params/decoder/layers/mlp/wi_0/kernel Shape: float32[2048,16,7168] Logical: P('embed', 'layers', 'mlp') Physical: ('fsdp', None, None) I0416 03:50:37.330188 136083091008640 maxtext_utils.py:1790] params/params/decoder/layers/mlp/wi_1/kernel Shape: float32[2048,16,7168] Logical: P('embed', 'layers', 'mlp') Physical: ('fsdp', None, None) I0416 03:50:37.330236 136083091008640 maxtext_utils.py:1790] params/params/decoder/layers/mlp/wo/kernel Shape: float32[7168,16,2048] Logical: P('mlp', 'layers', 'embed') Physical: (None, None, 'fsdp') I0416 03:50:37.330276 136083091008640 maxtext_utils.py:1790] params/params/decoder/layers/post_self_attention_layer_norm/scale Shape: float32[2048,16] Logical: P('norm', 'layers') Physical: (None, None) I0416 03:50:37.330303 136083091008640 maxtext_utils.py:1790] params/params/decoder/layers/pre_self_attention_layer_norm/scale Shape: float32[2048,16] Logical: P('norm', 'layers') Physical: (None, None) I0416 03:50:37.330341 136083091008640 maxtext_utils.py:1790] 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 03:50:37.330379 136083091008640 maxtext_utils.py:1790] 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 03:50:37.330406 136083091008640 maxtext_utils.py:1790] 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 03:50:37.330432 136083091008640 maxtext_utils.py:1790] 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 03:50:37.330465 136083091008640 maxtext_utils.py:1790] params/params/decoder/logits_dense/kernel Shape: float32[2048,32000] Logical: P('embed_vocab', 'vocab') Physical: ('fsdp', None) I0416 03:50:37.330498 136083091008640 maxtext_utils.py:1790] params/params/token_embedder/embedding Shape: float32[32000,2048] Logical: P('vocab', 'embed_vocab') Physical: (None, 'fsdp') I0416 03:50:37.349035 136083091008640 gradient_accumulation.py:70] gradient_accumulation/inputs Logical: float32[2048]............................................... Unknown. I0416 03:50:37.349099 136083091008640 gradient_accumulation.py:70] gradient_accumulation/inputs Physical: float32[2048]............................................... (None,). I0416 03:50:37.362184 136083091008640 gradient_accumulation.py:70] gradient_accumulation/inputs Logical: float32[2048,16,7168]....................................... Unknown. I0416 03:50:37.362230 136083091008640 gradient_accumulation.py:70] gradient_accumulation/inputs Physical: float32[2048,16,7168]....................................... ('fsdp', None, None). I0416 03:50:37.388225 136083091008640 gradient_accumulation.py:70] gradient_accumulation/inputs Logical: float32[7168,16,2048]....................................... Unknown. I0416 03:50:37.388281 136083091008640 gradient_accumulation.py:70] gradient_accumulation/inputs Physical: float32[7168,16,2048]....................................... (None, None, 'fsdp'). I0416 03:50:37.401257 136083091008640 gradient_accumulation.py:70] gradient_accumulation/inputs Logical: float32[2048,16]............................................ Unknown. I0416 03:50:37.401305 136083091008640 gradient_accumulation.py:70] gradient_accumulation/inputs Physical: float32[2048,16]............................................ (None, None). I0416 03:50:37.427281 136083091008640 gradient_accumulation.py:70] gradient_accumulation/inputs Logical: float32[2048,16,16,128]..................................... Unknown. I0416 03:50:37.427332 136083091008640 gradient_accumulation.py:70] gradient_accumulation/inputs Physical: float32[2048,16,16,128]..................................... ('fsdp', None, None, None). I0416 03:50:37.440298 136083091008640 gradient_accumulation.py:70] gradient_accumulation/inputs Logical: float32[16,16,128,2048]..................................... Unknown. I0416 03:50:37.440344 136083091008640 gradient_accumulation.py:70] gradient_accumulation/inputs Physical: float32[16,16,128,2048]..................................... (None, None, None, 'fsdp'). I0416 03:50:37.479174 136083091008640 gradient_accumulation.py:70] gradient_accumulation/inputs Logical: float32[2048,32000]......................................... Unknown. I0416 03:50:37.479235 136083091008640 gradient_accumulation.py:70] gradient_accumulation/inputs Physical: float32[2048,32000]......................................... ('fsdp', None). I0416 03:50:37.492172 136083091008640 gradient_accumulation.py:70] gradient_accumulation/inputs Logical: float32[32000,2048]......................................... Unknown. I0416 03:50:37.492217 136083091008640 gradient_accumulation.py:70] gradient_accumulation/inputs Physical: float32[32000,2048]......................................... (None, 'fsdp'). I0416 03:50:38.074023 136083091008640 train.py:157] train/xent Logical: float32[8,2048]............................................. ('activation_embed_and_logits_batch', 'activation_length'). I0416 03:50:38.074124 136083091008640 train.py:157] train/xent Physical: float32[8,2048]............................................. ('fsdp', None). I0416 03:50:38.087776 136083091008640 train.py:164] train/z_loss Logical: float32[8,2048]............................................. ('activation_embed_and_logits_batch', 'activation_length'). I0416 03:50:38.087822 136083091008640 train.py:164] train/z_loss Physical: float32[8,2048]............................................. ('fsdp', None). W0416 03:50:38.946346 3640194 pjrt_executable.cc:642] Assume version compatibility. PjRt-IFRT does not track XLA executable versions. I0416 03:50:39.082399 136083091008640 max_utils.py:791] Total memory size: 4.5 GB, Output size: 1.5 GB, Temp size: 2.9 GB, Argument size: 1.5 GB, Host temp size: 0.0 GB. I0416 03:50:39.082855 136083091008640 max_utils.py:194] tensorboardX not available; using no-op SummaryWriter. I0416 03:50:39.083001 136083091008640 metric_logger.py:289] number parameters: 1.104 billion W0416 03:50:40.406458 3640194 pjrt_executable.cc:642] Assume version compatibility. PjRt-IFRT does not track XLA executable versions. I0416 03:50:40.535934 136083091008640 checkpointing.py:794] Waiting for step 0 to finish before checkpoint... I0416 03:50:40.958111 136083091008640 checkpointing.py:798] Waited 0.42215871810913086 seconds for step 0 to finish before starting checkpointing. I0416 03:50:40.958540 136083091008640 checkpoint_manager.py:1983] [process=0][thread=MainThread][wait_until_finished] No Save Finalize thread to wait for. Returning. I0416 03:50:40.958692 136083091008640 checkpoint_manager.py:1501] [process=0] Saving checkpoint at step 0 I0416 03:50:40.959063 136083091008640 async_checkpointer.py:452] [process=0] Started async saving checkpoint to gs://wanglance-maxtext/linen_ckpt_feat_nnx_trainstate_and_training_loop_20260416_033532/linen_feat_nnx_trainstate_and_training_loop_20260416_033532_06_grad_accum/checkpoints/0. I0416 03:50:41.049351 136083091008640 signaling_client.py:373] Using ThreadSafeKeyValueSignalingClient I0416 03:50:41.129987 135975527253568 atomicity.py:137] Creating tmp directory gs://wanglance-maxtext/linen_ckpt_feat_nnx_trainstate_and_training_loop_20260416_033532/linen_feat_nnx_trainstate_and_training_loop_20260416_033532_06_grad_accum/checkpoints/0 I0416 03:50:41.149995 136083091008640 jax_array_handlers.py:347] Scheduling D2H of 39 prioritized jax.Array. I0416 03:50:41.150094 136083091008640 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 03:50:41.377918 3640194 pjrt_executable.cc:642] Assume version compatibility. PjRt-IFRT does not track XLA executable versions. W0416 03:50:41.389068 3640194 pjrt_executable.cc:642] Assume version compatibility. PjRt-IFRT does not track XLA executable versions. W0416 03:50:41.405346 3640194 pjrt_executable.cc:642] Assume version compatibility. PjRt-IFRT does not track XLA executable versions. W0416 03:50:41.424502 3640194 pjrt_executable.cc:642] Assume version compatibility. PjRt-IFRT does not track XLA executable versions. W0416 03:50:41.430528 3640194 pjrt_executable.cc:642] Assume version compatibility. PjRt-IFRT does not track XLA executable versions. W0416 03:50:41.435375 3640194 pjrt_executable.cc:642] Assume version compatibility. PjRt-IFRT does not track XLA executable versions. W0416 03:50:41.440857 3640194 pjrt_executable.cc:642] Assume version compatibility. PjRt-IFRT does not track XLA executable versions. W0416 03:50:41.445314 3640194 pjrt_executable.cc:642] Assume version compatibility. PjRt-IFRT does not track XLA executable versions. I0416 03:50:41.867389 135975516767808 atomicity.py:137] Creating tmp directory gs://wanglance-maxtext/linen_ckpt_feat_nnx_trainstate_and_training_loop_20260416_033532/linen_feat_nnx_trainstate_and_training_loop_20260416_033532_06_grad_accum/checkpoints/0/items I0416 03:50:41.922632 136083091008640 base_pytree_checkpoint_handler.py:153] [process=0][thread=MainThread] Initiated "orbax.checkpoint._src.serialization.jax_array_handlers.ArrayHandler".serialize. Time taken: 0.773036s I0416 03:50:41.923020 136083091008640 base_pytree_checkpoint_handler.py:128] [process=0] /jax/checkpoint/write/blocking_gbytes_per_sec: 14.148 GiB/s (total gbytes: 12.3 GiB) (time elapsed: 872 milliseconds) (per-host) I0416 03:50:41.923084 136083091008640 base_pytree_checkpoint_handler.py:732] [process=0][thread=MainThread] Initiated Pytree async_save. Time taken: 0.872384s (batch_requests_ready=0.095288s, total_serialization_initiated=0.776794s, others=0.000302s) I0416 03:50:41.923140 136083091008640 composite_checkpoint_handler.py:715] [process=0][thread=MainThread] Initiated CompositeCheckpointHandler.async_save. Time taken: 0.873331s (all_items=0.000019s, per_item={'items': '0.00001860'}, temp_paths=0.873312) I0416 03:50:41.923887 135975464339008 async_checkpointer.py:79] [process=0][thread=async_save] Background save thread started. I0416 03:50:41.923983 136083091008640 async_checkpointer.py:561] Finished blocking save. Time taken: 0.965252s. Continuing background save to gs://wanglance-maxtext/linen_ckpt_feat_nnx_trainstate_and_training_loop_20260416_033532/linen_feat_nnx_trainstate_and_training_loop_20260416_033532_06_grad_accum/checkpoints/0. I0416 03:50:41.924187 136083091008640 checkpoint_manager.py:1549] [process=0][thread=MainThread][step=0] Starting CheckpointManager Save Finalize thread=save_finalize I0416 03:50:41.924405 136083091008640 standard_logger.py:34] {'step': 0, 'event_type': 'save', 'directory': 'gs://wanglance-maxtext/linen_ckpt_feat_nnx_trainstate_and_training_loop_20260416_033532/linen_feat_nnx_trainstate_and_training_loop_20260416_033532_06_grad_accum/checkpoints', 'reached_preemption': False, 'preemption_received_at': None, 'synchronous': False, 'wait_for_prev_start_time': 1776311440.9585278, 'wait_for_prev_duration_secs': 4.029273986816406e-05, 'time_between_consecutive_saves_sec': None, 'checkpointer_blocking_start_time': 1776311440.958714, 'checkpointer_blocking_duration_secs': 0.9654064178466797, 'get_old_steps_start_time': 1776311441.9241362, 'get_old_steps_duration_secs': 1.9788742065429688e-05, 'checkpoint_manager_blocking_start_time': 1776311440.9584467, 'checkpoint_manager_blocking_duration_secs': 0.9659357070922852} I0416 03:50:41.924469 136083091008640 checkpointing.py:409] Started an asynchronous checkpoint save for step 0 I0416 03:50:41.924505 136083091008640 max_utils.py:750] Memstats: After params initialized: I0416 03:50:41.924584 136083091008640 max_utils.py:756] Using (GB) 1.6 / 31.25 (5.120000%) on TPU_0(process=0,(0,0,0,0)) I0416 03:50:41.924611 136083091008640 max_utils.py:756] Using (GB) 1.59 / 31.25 (5.088000%) on TPU_1(process=0,(1,0,0,0)) I0416 03:50:41.924630 136083091008640 max_utils.py:756] Using (GB) 1.59 / 31.25 (5.088000%) on TPU_2(process=0,(0,1,0,0)) I0416 03:50:41.924647 136083091008640 max_utils.py:756] Using (GB) 1.59 / 31.25 (5.088000%) on TPU_3(process=0,(1,1,0,0)) I0416 03:50:41.924663 136083091008640 max_utils.py:756] Using (GB) 1.59 / 31.25 (5.088000%) on TPU_4(process=0,(0,2,0,0)) I0416 03:50:41.924685 136083091008640 max_utils.py:756] Using (GB) 1.59 / 31.25 (5.088000%) on TPU_5(process=0,(1,2,0,0)) I0416 03:50:41.924701 136083091008640 max_utils.py:756] Using (GB) 1.59 / 31.25 (5.088000%) on TPU_6(process=0,(0,3,0,0)) I0416 03:50:41.924715 136083091008640 max_utils.py:756] Using (GB) 1.59 / 31.25 (5.088000%) on TPU_7(process=0,(1,3,0,0)) W0416 03:50:41.930379 3640194 pjrt_executable.cc:642] Assume version compatibility. PjRt-IFRT does not track XLA executable versions. W0416 03:50:41.935886 3640194 pjrt_executable.cc:642] Assume version compatibility. PjRt-IFRT does not track XLA executable versions. W0416 03:50:41.940686 3640194 pjrt_executable.cc:642] Assume version compatibility. PjRt-IFRT does not track XLA executable versions. I0416 03:50:42.015553 135975485310528 checkpoint.py:188] Wrote Metadata={'item_handlers': None, 'metrics': {}, 'performance_metrics': {}, 'init_timestamp_nsecs': 1776311441773682482, 'commit_timestamp_nsecs': None, 'custom_metadata': {}}, json={"item_handlers": null, "metrics": {}, "performance_metrics": {}, "init_timestamp_nsecs": 1776311441773682482, "commit_timestamp_nsecs": null, "custom_metadata": {}} to gs://wanglance-maxtext/linen_ckpt_feat_nnx_trainstate_and_training_loop_20260416_033532/linen_feat_nnx_trainstate_and_training_loop_20260416_033532_06_grad_accum/checkpoints/0/_CHECKPOINT_METADATA I0416 03:50:42.015797 135975537739328 async_checkpointer.py:265] [process=0][thread=save_finalize] Waiting for background save thread=async_save. I0416 03:50:42.273596 136083091008640 metric_logger.py:185] completed step: 0, seconds: 1.452, TFLOP/s/device: 37.418, Tokens/s/device: 5640.063, total_weights: 65536, loss: 10.871 I0416 03:50:42.274589 136083091008640 metric_logger.py:269] To see full metrics 'tensorboard --logdir=gs://wanglance-maxtext/linen_ckpt_feat_nnx_trainstate_and_training_loop_20260416_033532/linen_feat_nnx_trainstate_and_training_loop_20260416_033532_06_grad_accum/tensorboard/' I0416 03:50:42.586179 3642533 google_auth_provider.cc:149] Using credentials at ~/.config/gcloud/application_default_credentials.json I0416 03:50:42.586237 3642533 google_auth_provider.cc:156] Using OAuth2 AuthProvider I0416 03:50:42.699800 136083091008640 metric_logger.py:185] completed step: 1, seconds: 1.736, TFLOP/s/device: 31.310, Tokens/s/device: 4719.359, total_weights: 65536, loss: 10.871 I0416 03:50:43.122140 136083091008640 metric_logger.py:185] completed step: 2, seconds: 0.007, TFLOP/s/device: 8191.186, Tokens/s/device: 1234664.657, total_weights: 65536, loss: 10.261 I0416 03:50:43.382599 135975506282048 array_metadata_store.py:203] [process=0][thread=array_type_handler] Wrote 39 array_metadata.ArrayMetadata to gs://wanglance-maxtext/linen_ckpt_feat_nnx_trainstate_and_training_loop_20260416_033532/linen_feat_nnx_trainstate_and_training_loop_20260416_033532_06_grad_accum/checkpoints/0/items/array_metadatas/process_0 I0416 03:50:43.540160 136083091008640 metric_logger.py:185] completed step: 3, seconds: 0.427, TFLOP/s/device: 127.277, Tokens/s/device: 19184.562, total_weights: 65536, loss: 9.732 I0416 03:51:08.923069 135975527253568 base_pytree_checkpoint_handler.py:1217] [process=0][thread=write_metadata_after_commits] Commit + Array metadata written. Time taken: 26.341310s (commit=25.917728s, array_metadata_write=0.423581s) I0416 03:51:08.924324 135975464339008 base_pytree_checkpoint_handler.py:128] [process=0] /jax/checkpoint/write/gbytes_per_sec: 453.361 MiB/s (total gbytes: 12.3 GiB) (time elapsed: 27 seconds) (per-host) I0416 03:51:08.924445 135975464339008 async_checkpointer.py:90] [process=0][thread=async_save] 3 Handler Commit operations completed. Time taken: 27.000415s. I0416 03:51:09.165860 135975464339008 checkpoint.py:228] Read Metadata={'item_handlers': None, 'metrics': {}, 'performance_metrics': {}, 'init_timestamp_nsecs': 1776311441773682482, 'commit_timestamp_nsecs': None, 'custom_metadata': {}} from gs://wanglance-maxtext/linen_ckpt_feat_nnx_trainstate_and_training_loop_20260416_033532/linen_feat_nnx_trainstate_and_training_loop_20260416_033532_06_grad_accum/checkpoints/0/_CHECKPOINT_METADATA I0416 03:51:09.340763 135975464339008 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 03:51:09.532931 135975485310528 checkpoint.py:247] Updated Metadata={'item_handlers': {'items': 'orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeCheckpointHandler'}, 'metrics': {}, 'performance_metrics': {}, 'init_timestamp_nsecs': 1776311441773682482, 'commit_timestamp_nsecs': None, 'custom_metadata': {}} to gs://wanglance-maxtext/linen_ckpt_feat_nnx_trainstate_and_training_loop_20260416_033532/linen_feat_nnx_trainstate_and_training_loop_20260416_033532_06_grad_accum/checkpoints/0/_CHECKPOINT_METADATA I0416 03:51:09.819245 135975464339008 ocdbt_utils.py:56] Param validation support for Zarr3 will be added later (b/362328389). I0416 03:51:09.819860 135975464339008 base_pytree_checkpoint_handler.py:1342] [process=0][thread=async_save] Pytree save finalize (merge_ocdbt + ArrayMetadata validation) completed. Time taken: 0.616447s. use_zarr3=True, enable_post_merge_validation=True, directory=gs://wanglance-maxtext/linen_ckpt_feat_nnx_trainstate_and_training_loop_20260416_033532/linen_feat_nnx_trainstate_and_training_loop_20260416_033532_06_grad_accum/checkpoints/0/items I0416 03:51:09.820539 135975464339008 atomicity.py:608] Finalizing gs://wanglance-maxtext/linen_ckpt_feat_nnx_trainstate_and_training_loop_20260416_033532/linen_feat_nnx_trainstate_and_training_loop_20260416_033532_06_grad_accum/checkpoints/0/items I0416 03:51:10.067466 135975464339008 atomicity.py:608] Finalizing gs://wanglance-maxtext/linen_ckpt_feat_nnx_trainstate_and_training_loop_20260416_033532/linen_feat_nnx_trainstate_and_training_loop_20260416_033532_06_grad_accum/checkpoints/0 I0416 03:51:10.761778 135975464339008 atomicity.py:794] [process=0][thread=async_save] Finished saving checkpoint (finalized tmp dir) to `gs://wanglance-maxtext/linen_ckpt_feat_nnx_trainstate_and_training_loop_20260416_033532/linen_feat_nnx_trainstate_and_training_loop_20260416_033532_06_grad_accum/checkpoints/0`. I0416 03:51:10.762414 135975464339008 async_checkpointer.py:420] Finished async_save (blocking + background). Time taken: 29.803688s. directory=gs://wanglance-maxtext/linen_ckpt_feat_nnx_trainstate_and_training_loop_20260416_033532/linen_feat_nnx_trainstate_and_training_loop_20260416_033532_06_grad_accum/checkpoints/0 I0416 03:51:10.762492 135975464339008 async_checkpointer.py:144] [process=0][thread=async_save] Background save thread done. Time taken: 28.838464s. I0416 03:51:10.762706 135975537739328 async_checkpointer.py:273] [process=0][thread=save_finalize] Done with waiting for background save thread=async_save. I0416 03:51:10.762817 135975537739328 async_checkpointer.py:283] [process=0][thread=save_finalize] No errors found in background save thread=async_save. I0416 03:51:10.762896 135975537739328 checkpoint_manager.py:2103] [process=0][thread=save_finalize][step=0] CheckpointManager Save Finalize is syncing with other hosts... I0416 03:51:10.762945 135975537739328 checkpoint_manager.py:2112] [process=0][thread=save_finalize][step=0] CheckpointManager Save Finalize is done on all hosts. I0416 03:51:20.195588 136083091008640 metric_logger.py:185] completed step: 4, seconds: 0.422, TFLOP/s/device: 128.733, Tokens/s/device: 19404.000, total_weights: 65536, loss: 9.274 I0416 03:51:20.203114 136083091008640 metric_logger.py:185] completed step: 5, seconds: 0.418, TFLOP/s/device: 129.928, Tokens/s/device: 19584.124, total_weights: 65536, loss: 8.888 I0416 03:51:20.630518 136083091008640 metric_logger.py:185] completed step: 6, seconds: 36.656, TFLOP/s/device: 1.483, Tokens/s/device: 223.482, total_weights: 65536, loss: 8.588 I0416 03:51:21.059561 136083091008640 metric_logger.py:185] completed step: 7, seconds: 0.005, TFLOP/s/device: 11585.699, Tokens/s/device: 1746322.746, total_weights: 65536, loss: 8.380 I0416 03:51:21.482882 136083091008640 metric_logger.py:185] completed step: 8, seconds: 0.429, TFLOP/s/device: 126.575, Tokens/s/device: 19078.805, total_weights: 65536, loss: 8.251 I0416 03:51:21.910820 136083091008640 checkpointing.py:794] Waiting for step 9 to finish before checkpoint... I0416 03:51:21.912035 136083091008640 checkpointing.py:798] Waited 0.0012629032135009766 seconds for step 9 to finish before starting checkpointing. I0416 03:51:21.912404 136083091008640 checkpoint_manager.py:1983] [process=0][thread=MainThread][wait_until_finished] No Save Finalize thread to wait for. Returning. I0416 03:51:21.912616 136083091008640 checkpoint_manager.py:1501] [process=0] Saving checkpoint at step 9 I0416 03:51:21.913004 136083091008640 async_checkpointer.py:452] [process=0] Started async saving checkpoint to gs://wanglance-maxtext/linen_ckpt_feat_nnx_trainstate_and_training_loop_20260416_033532/linen_feat_nnx_trainstate_and_training_loop_20260416_033532_06_grad_accum/checkpoints/9. I0416 03:51:22.085552 135975537739328 atomicity.py:137] Creating tmp directory gs://wanglance-maxtext/linen_ckpt_feat_nnx_trainstate_and_training_loop_20260416_033532/linen_feat_nnx_trainstate_and_training_loop_20260416_033532_06_grad_accum/checkpoints/9 I0416 03:51:22.093832 136083091008640 jax_array_handlers.py:347] Scheduling D2H of 39 prioritized jax.Array. I0416 03:51:22.093967 136083091008640 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 03:51:22.807702 135975527253568 atomicity.py:137] Creating tmp directory gs://wanglance-maxtext/linen_ckpt_feat_nnx_trainstate_and_training_loop_20260416_033532/linen_feat_nnx_trainstate_and_training_loop_20260416_033532_06_grad_accum/checkpoints/9/items I0416 03:51:33.178692 136083091008640 base_pytree_checkpoint_handler.py:153] [process=0][thread=MainThread] Initiated "orbax.checkpoint._src.serialization.jax_array_handlers.ArrayHandler".serialize. Time taken: 11.085459s I0416 03:51:33.184774 136083091008640 base_pytree_checkpoint_handler.py:128] [process=0] /jax/checkpoint/write/blocking_gbytes_per_sec: 1.103 GiB/s (total gbytes: 12.3 GiB) (time elapsed: 11 seconds) (per-host) I0416 03:51:33.184874 136083091008640 base_pytree_checkpoint_handler.py:732] [process=0][thread=MainThread] Initiated Pytree async_save. Time taken: 11.186155s (batch_requests_ready=0.090604s, total_serialization_initiated=11.089637s, others=0.005914s) I0416 03:51:33.185004 136083091008640 composite_checkpoint_handler.py:715] [process=0][thread=MainThread] Initiated CompositeCheckpointHandler.async_save. Time taken: 11.186924s (all_items=0.000027s, per_item={'items': '0.00002742'}, temp_paths=11.186896) I0416 03:51:33.187347 135975537739328 async_checkpointer.py:79] [process=0][thread=async_save] Background save thread started. I0416 03:51:33.187469 136083091008640 async_checkpointer.py:561] Finished blocking save. Time taken: 11.274803s. Continuing background save to gs://wanglance-maxtext/linen_ckpt_feat_nnx_trainstate_and_training_loop_20260416_033532/linen_feat_nnx_trainstate_and_training_loop_20260416_033532_06_grad_accum/checkpoints/9. I0416 03:51:33.187767 136083091008640 checkpoint_manager.py:1549] [process=0][thread=MainThread][step=9] Starting CheckpointManager Save Finalize thread=save_finalize I0416 03:51:33.188039 135975424493120 async_checkpointer.py:265] [process=0][thread=save_finalize] Waiting for background save thread=async_save. I0416 03:51:33.188236 136083091008640 standard_logger.py:34] {'step': 9, 'event_type': 'save', 'directory': 'gs://wanglance-maxtext/linen_ckpt_feat_nnx_trainstate_and_training_loop_20260416_033532/linen_feat_nnx_trainstate_and_training_loop_20260416_033532_06_grad_accum/checkpoints', 'reached_preemption': False, 'preemption_received_at': None, 'synchronous': False, 'wait_for_prev_start_time': 1776311481.9123507, 'wait_for_prev_duration_secs': 8.0108642578125e-05, 'time_between_consecutive_saves_sec': 11.149390459060669, 'checkpointer_blocking_start_time': 1776311481.912637, 'checkpointer_blocking_duration_secs': 11.275003671646118, 'get_old_steps_start_time': 1776311493.187667, 'get_old_steps_duration_secs': 4.1484832763671875e-05, 'checkpoint_manager_blocking_start_time': 1776311481.9122996, 'checkpoint_manager_blocking_duration_secs': 11.275896787643433} I0416 03:51:33.188433 136083091008640 checkpointing.py:409] Started an asynchronous checkpoint save for step 9 I0416 03:51:33.188479 136083091008640 checkpoint_manager.py:1994] [process=0][thread=MainThread][step=9][wait_until_finished] Waiting for Save Finalize thread (save_finalize) to complete. I0416 03:51:33.968769 135974849873472 array_metadata_store.py:203] [process=0][thread=array_type_handler] Wrote 39 array_metadata.ArrayMetadata to gs://wanglance-maxtext/linen_ckpt_feat_nnx_trainstate_and_training_loop_20260416_033532/linen_feat_nnx_trainstate_and_training_loop_20260416_033532_06_grad_accum/checkpoints/9/items/array_metadatas/process_0 I0416 03:52:14.145411 135974178784832 base_pytree_checkpoint_handler.py:1217] [process=0][thread=write_metadata_after_commits] Commit + Array metadata written. Time taken: 40.959348s (commit=40.451932s, array_metadata_write=0.507416s) I0416 03:52:14.146731 135975537739328 base_pytree_checkpoint_handler.py:128] [process=0] /jax/checkpoint/write/gbytes_per_sec: 242.325 MiB/s (total gbytes: 12.3 GiB) (time elapsed: 52 seconds) (per-host) I0416 03:52:14.146794 135975537739328 async_checkpointer.py:90] [process=0][thread=async_save] 3 Handler Commit operations completed. Time taken: 40.959263s. I0416 03:52:14.604355 135975537739328 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 03:52:15.066229 135975537739328 ocdbt_utils.py:56] Param validation support for Zarr3 will be added later (b/362328389). I0416 03:52:15.066484 135975537739328 base_pytree_checkpoint_handler.py:1342] [process=0][thread=async_save] Pytree save finalize (merge_ocdbt + ArrayMetadata validation) completed. Time taken: 0.625773s. use_zarr3=True, enable_post_merge_validation=True, directory=gs://wanglance-maxtext/linen_ckpt_feat_nnx_trainstate_and_training_loop_20260416_033532/linen_feat_nnx_trainstate_and_training_loop_20260416_033532_06_grad_accum/checkpoints/9/items I0416 03:52:15.067254 135975537739328 atomicity.py:608] Finalizing gs://wanglance-maxtext/linen_ckpt_feat_nnx_trainstate_and_training_loop_20260416_033532/linen_feat_nnx_trainstate_and_training_loop_20260416_033532_06_grad_accum/checkpoints/9/items I0416 03:52:15.346910 135975537739328 atomicity.py:608] Finalizing gs://wanglance-maxtext/linen_ckpt_feat_nnx_trainstate_and_training_loop_20260416_033532/linen_feat_nnx_trainstate_and_training_loop_20260416_033532_06_grad_accum/checkpoints/9 I0416 03:52:16.014580 135975537739328 atomicity.py:794] [process=0][thread=async_save] Finished saving checkpoint (finalized tmp dir) to `gs://wanglance-maxtext/linen_ckpt_feat_nnx_trainstate_and_training_loop_20260416_033532/linen_feat_nnx_trainstate_and_training_loop_20260416_033532_06_grad_accum/checkpoints/9`. I0416 03:52:16.015484 135975537739328 async_checkpointer.py:420] Finished async_save (blocking + background). Time taken: 54.102832s. directory=gs://wanglance-maxtext/linen_ckpt_feat_nnx_trainstate_and_training_loop_20260416_033532/linen_feat_nnx_trainstate_and_training_loop_20260416_033532_06_grad_accum/checkpoints/9 I0416 03:52:16.015598 135975537739328 async_checkpointer.py:144] [process=0][thread=async_save] Background save thread done. Time taken: 42.828073s. I0416 03:52:16.015792 135975424493120 async_checkpointer.py:273] [process=0][thread=save_finalize] Done with waiting for background save thread=async_save. I0416 03:52:16.015942 135975424493120 async_checkpointer.py:283] [process=0][thread=save_finalize] No errors found in background save thread=async_save. I0416 03:52:16.016010 135975424493120 checkpoint_manager.py:2103] [process=0][thread=save_finalize][step=9] CheckpointManager Save Finalize is syncing with other hosts... I0416 03:52:16.016097 135975424493120 checkpoint_manager.py:2112] [process=0][thread=save_finalize][step=9] CheckpointManager Save Finalize is done on all hosts. I0416 03:52:16.016428 136083091008640 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 03:52:16.016613 136083091008640 checkpoint_manager.py:1983] [process=0][thread=MainThread][wait_until_finished] No Save Finalize thread to wait for. Returning. I0416 03:52:16.017335 136083091008640 metric_logger.py:185] completed step: 9, seconds: 0.429, TFLOP/s/device: 126.658, Tokens/s/device: 19091.299, total_weights: 65536, loss: 8.175 Per train step: Total TFLOPs: 54.35 split as 93.93% learnable weight flops and 6.07% attention flops