2026-04-16 03:43:41.042307: 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 03:43:41.158282 123881254395008 max_utils.py:238] Skipping jax distributed system due to skip_jax_distributed_system=True flag. I0416 03:44:32.832621 123881254395008 max_utils.py:800] System Information: Jax Version: 0.9.2 I0416 03:44:32.832728 123881254395008 max_utils.py:801] System Information: Jaxlib Version: 0.9.2 I0416 03:44:32.832764 123881254395008 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 03:44:32.832792 123881254395008 train_utils.py:364] WARNING: 'dataset_path' might be pointing your local file system I0416 03:44:32.832884 123881254395008 train.py:818] [DECOUPLED NO-OP] skipping cloud diagnostics wrapper. W0416 03:44:32.926716 3633034 pjrt_executable.cc:642] Assume version compatibility. PjRt-IFRT does not track XLA executable versions. I0416 03:44:33.330150 123881254395008 maxtext_utils.py:1687] Num_devices: 8, shape (1, 1, 1, 8, 1, 1, 1, 1, 1, 1, 1, 1, 1) I0416 03:44:33.330317 123881254395008 maxtext_utils.py:1687] Num_devices: 8, shape (1, 1, 1, 8, 1, 1, 1, 1, 1, 1, 1, 1, 1) I0416 03:44:33.330773 123881254395008 checkpointing.py:688] Setting up checkpoint logger... I0416 03:44:33.330816 123881254395008 checkpointing.py:234] Creating checkpoint manager with ocdbt=True and zarr3=True I0416 03:44:33.330866 123881254395008 pytree_checkpoint_handler.py:577] save_device_host_concurrent_bytes=None I0416 03:44:33.331410 123881254395008 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 0x70aac37f5040>, enable_pinned_host_transfer=False, save_concurrent_bytes: 96000000000 (89.4 GiB), restore_concurrent_bytes: 96000000000 (89.4 GiB) I0416 03:44:35.640526 123881254395008 checkpointing.py:266] Enabling policy for fixed interval checkpointing. I0416 03:44:35.640706 123881254395008 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 0x70a447536f60>}, handler_registry=None I0416 03:44:35.640976 123881254395008 composite_checkpoint_handler.py:237] Deferred registration for item: "items". Adding handler `<orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeCheckpointHandler object at 0x70a447536f60>` 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 03:44:35.641017 123881254395008 composite_checkpoint_handler.py:237] Deferred registration for item: "metrics". Adding handler `<orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonCheckpointHandler object at 0x70a447535cd0>` 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 03:44:35.641062 123881254395008 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 0x70a447536f60>, ('items', <class 'orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeRestoreArgs'>): <orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeCheckpointHandler object at 0x70a447536f60>, ('metrics', <class 'orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonSaveArgs'>): <orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonCheckpointHandler object at 0x70a447535cd0>, ('metrics', <class 'orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonRestoreArgs'>): <orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonCheckpointHandler object at 0x70a447535cd0>}). I0416 03:44:35.641529 123881254395008 abstract_checkpointer.py:35] orbax-checkpoint version: 0.11.28 I0416 03:44:35.641582 123881254395008 async_checkpointer.py:177] [process=0][thread=MainThread] Using barrier_sync_fn: <function get_barrier_sync_fn.<locals>.<lambda> at 0x70a4475420c0> timeout: 600 secs and primary_host=0 for async checkpoint writes I0416 03:44:35.758921 123881254395008 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_04_int8/checkpoints I0416 03:44:35.759183 123881254395008 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_04_int8/checkpoints: <orbax.checkpoint.checkpoint_manager.CheckpointManager object at 0x70a4487a3d40> I0416 03:44:35.759261 123881254395008 checkpointing.py:302] Checkpoint manager created! I0416 03:44:35.840794 123881254395008 dataset_info.py:707] Load dataset info from tests/assets/local_datasets/c4_en_dataset_minimal/c4/en/3.1.0 I0416 03:44:35.843821 123881254395008 reader.py:262] Creating a tf.data.Dataset reading 8 files located in folders: tests/assets/local_datasets/c4_en_dataset_minimal/c4/en/3.1.0. I0416 03:44:35.898266 123881254395008 logging_logger.py:49] Constructing tf.data.Dataset __local_c4_builder for split train, from tests/assets/local_datasets/c4_en_dataset_minimal/c4/en/3.1.0 I0416 03:44:35.929807 123881254395008 tokenizer.py:245] Tokenizer path: src/maxtext/assets/tokenizers/tokenizer.llama2 I0416 03:44:35.929873 123881254395008 tokenizer.py:187] Loading sentencepiece tokenizer: src/maxtext/assets/tokenizers/tokenizer.llama2 I0416 03:44:37.107654 123881254395008 nnx_wrappers.py:437] Unknown Logical: bfloat16[8,2048,2048]....................................... ('activation_batch', 'activation_norm_length', 'activation_embed'). I0416 03:44:37.107754 123881254395008 nnx_wrappers.py:437] Unknown Physical: bfloat16[8,2048,2048]....................................... ('fsdp', None, None). I0416 03:44:37.211150 123881254395008 attentions.py:1088] attentions/inputs_q Logical: bfloat16[8,2048,2048]....................................... ('activation_batch', 'activation_attn_length', 'activation_attn_embed'). I0416 03:44:37.211228 123881254395008 attentions.py:1088] attentions/inputs_q Physical: bfloat16[8,2048,2048]....................................... ('fsdp', None, None). I0416 03:44:37.226289 123881254395008 attentions.py:1089] attentions/inputs_kv Logical: bfloat16[8,2048,2048]....................................... ('activation_batch', 'activation_attn_length', 'activation_attn_embed'). I0416 03:44:37.226339 123881254395008 attentions.py:1089] attentions/inputs_kv Physical: bfloat16[8,2048,2048]....................................... ('fsdp', None, None). I0416 03:44:37.283931 123881254395008 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 03:44:37.284007 123881254395008 attentions.py:1154] attentions/query Physical: bfloat16[8,2048,16,128]..................................... ('fsdp', None, None, None). I0416 03:44:37.299399 123881254395008 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 03:44:37.299451 123881254395008 attentions.py:1155] attentions/key Physical: bfloat16[8,2048,16,128]..................................... ('fsdp', None, None, None). I0416 03:44:37.314590 123881254395008 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 03:44:37.314641 123881254395008 attentions.py:1156] attentions/value Physical: bfloat16[8,2048,16,128]..................................... ('fsdp', None, None, None). I0416 03:44:37.336978 123881254395008 attentions.py:1197] attentions/out Logical: bfloat16[8,2048,16,128]..................................... ('activation_batch', 'activation_attn_length', 'activation_heads', 'activation_kv'). I0416 03:44:37.337037 123881254395008 attentions.py:1197] attentions/out Physical: bfloat16[8,2048,16,128]..................................... ('fsdp', None, None, None). I0416 03:44:37.394199 123881254395008 linears.py:525] linears/x Logical: bfloat16[8,2048,7168]....................................... ('activation_batch', 'activation_length', 'activation_mlp'). I0416 03:44:37.394267 123881254395008 linears.py:525] linears/x Physical: bfloat16[8,2048,7168]....................................... ('fsdp', None, None). I0416 03:44:37.839660 123881254395008 checkpointing.py:578] checkpoint manager exists so trying to load this run's existing checkpoint I0416 03:44:37.839757 123881254395008 checkpointing.py:676] No existing checkpoints found, not restoring checkpoint. [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 03:44:39.753911 123881254395008 maxtext_utils.py:1790] params/params/decoder/decoder_norm/scale Shape: float32[2048] Logical: P('norm',) Physical: (None,) I0416 03:44:39.754024 123881254395008 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 03:44:39.754082 123881254395008 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 03:44:39.754130 123881254395008 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 03:44:39.754173 123881254395008 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 03:44:39.754203 123881254395008 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 03:44:39.754252 123881254395008 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 03:44:39.754297 123881254395008 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 03:44:39.754328 123881254395008 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 03:44:39.754357 123881254395008 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 03:44:39.754392 123881254395008 maxtext_utils.py:1790] params/params/decoder/logits_dense/kernel Shape: float32[2048,32000] Logical: P('embed_vocab', 'vocab') Physical: ('fsdp', None) I0416 03:44:39.754426 123881254395008 maxtext_utils.py:1790] params/params/token_embedder/embedding Shape: float32[32000,2048] Logical: P('vocab', 'embed_vocab') Physical: (None, 'fsdp') I0416 03:44:41.411495 123881254395008 train.py:157] train/xent Logical: float32[8,2048]............................................. ('activation_embed_and_logits_batch', 'activation_length'). I0416 03:44:41.411579 123881254395008 train.py:157] train/xent Physical: float32[8,2048]............................................. ('fsdp', None). I0416 03:44:41.426531 123881254395008 train.py:164] train/z_loss Logical: float32[8,2048]............................................. ('activation_embed_and_logits_batch', 'activation_length'). I0416 03:44:41.426582 123881254395008 train.py:164] train/z_loss Physical: float32[8,2048]............................................. ('fsdp', None). W0416 03:44:42.002091 3633034 pjrt_executable.cc:642] Assume version compatibility. PjRt-IFRT does not track XLA executable versions. I0416 03:44:42.192733 123881254395008 max_utils.py:791] Total memory size: 3.7 GB, Output size: 1.5 GB, Temp size: 2.2 GB, Argument size: 1.5 GB, Host temp size: 0.0 GB. I0416 03:44:42.193218 123881254395008 max_utils.py:194] tensorboardX not available; using no-op SummaryWriter. I0416 03:44:42.193366 123881254395008 metric_logger.py:289] number parameters: 1.104 billion W0416 03:44:46.727729 3633034 pjrt_executable.cc:642] Assume version compatibility. PjRt-IFRT does not track XLA executable versions. I0416 03:44:46.954326 123881254395008 checkpointing.py:794] Waiting for step 0 to finish before checkpoint... I0416 03:44:47.079466 123881254395008 checkpointing.py:798] Waited 0.12512993812561035 seconds for step 0 to finish before starting checkpointing. I0416 03:44:47.079900 123881254395008 checkpoint_manager.py:1983] [process=0][thread=MainThread][wait_until_finished] No Save Finalize thread to wait for. Returning. I0416 03:44:47.080070 123881254395008 checkpoint_manager.py:1501] [process=0] Saving checkpoint at step 0 I0416 03:44:47.080431 123881254395008 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_04_int8/checkpoints/0. I0416 03:44:47.165803 123881254395008 signaling_client.py:373] Using ThreadSafeKeyValueSignalingClient I0416 03:44:47.257118 123881254395008 jax_array_handlers.py:347] Scheduling D2H of 39 prioritized jax.Array. I0416 03:44:47.257211 123881254395008 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 03:44:47.263217 123765837006400 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_04_int8/checkpoints/0 W0416 03:44:47.490738 3633034 pjrt_executable.cc:642] Assume version compatibility. PjRt-IFRT does not track XLA executable versions. W0416 03:44:47.499324 3633034 pjrt_executable.cc:642] Assume version compatibility. PjRt-IFRT does not track XLA executable versions. W0416 03:44:47.513116 3633034 pjrt_executable.cc:642] Assume version compatibility. PjRt-IFRT does not track XLA executable versions. W0416 03:44:47.527209 3633034 pjrt_executable.cc:642] Assume version compatibility. PjRt-IFRT does not track XLA executable versions. W0416 03:44:47.540615 3633034 pjrt_executable.cc:642] Assume version compatibility. PjRt-IFRT does not track XLA executable versions. W0416 03:44:47.547070 3633034 pjrt_executable.cc:642] Assume version compatibility. PjRt-IFRT does not track XLA executable versions. W0416 03:44:47.552350 3633034 pjrt_executable.cc:642] Assume version compatibility. PjRt-IFRT does not track XLA executable versions. W0416 03:44:47.557421 3633034 pjrt_executable.cc:642] Assume version compatibility. PjRt-IFRT does not track XLA executable versions. I0416 03:44:48.009571 123765826520640 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_04_int8/checkpoints/0/items I0416 03:44:48.121842 123881254395008 base_pytree_checkpoint_handler.py:153] [process=0][thread=MainThread] Initiated "orbax.checkpoint._src.serialization.jax_array_handlers.ArrayHandler".serialize. Time taken: 0.865148s I0416 03:44:48.122265 123881254395008 base_pytree_checkpoint_handler.py:128] [process=0] /jax/checkpoint/write/blocking_gbytes_per_sec: 12.920 GiB/s (total gbytes: 12.3 GiB) (time elapsed: 955 milliseconds) (per-host) I0416 03:44:48.122316 123881254395008 base_pytree_checkpoint_handler.py:732] [process=0][thread=MainThread] Initiated Pytree async_save. Time taken: 0.955290s (batch_requests_ready=0.086220s, total_serialization_initiated=0.868738s, others=0.000332s) I0416 03:44:48.122375 123881254395008 composite_checkpoint_handler.py:715] [process=0][thread=MainThread] Initiated CompositeCheckpointHandler.async_save. Time taken: 0.956082s (all_items=0.000018s, per_item={'items': '0.00001836'}, temp_paths=0.956064) I0416 03:44:48.123113 123765765703232 async_checkpointer.py:79] [process=0][thread=async_save] Background save thread started. I0416 03:44:48.123209 123881254395008 async_checkpointer.py:561] Finished blocking save. Time taken: 1.043098s. 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_04_int8/checkpoints/0. I0416 03:44:48.123391 123881254395008 checkpoint_manager.py:1549] [process=0][thread=MainThread][step=0] Starting CheckpointManager Save Finalize thread=save_finalize I0416 03:44:48.123658 123881254395008 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_04_int8/checkpoints', 'reached_preemption': False, 'preemption_received_at': None, 'synchronous': False, 'wait_for_prev_start_time': 1776311087.0798867, 'wait_for_prev_duration_secs': 4.124641418457031e-05, 'time_between_consecutive_saves_sec': None, 'checkpointer_blocking_start_time': 1776311087.0800927, 'checkpointer_blocking_duration_secs': 1.043227195739746, 'get_old_steps_start_time': 1776311088.1233358, 'get_old_steps_duration_secs': 2.193450927734375e-05, 'checkpoint_manager_blocking_start_time': 1776311087.0798025, 'checkpoint_manager_blocking_duration_secs': 1.0438315868377686} I0416 03:44:48.123728 123881254395008 checkpointing.py:409] Started an asynchronous checkpoint save for step 0 I0416 03:44:48.123766 123881254395008 max_utils.py:750] Memstats: After params initialized: I0416 03:44:48.123851 123881254395008 max_utils.py:756] Using (GB) 1.6 / 31.25 (5.120000%) on TPU_0(process=0,(0,0,0,0)) I0416 03:44:48.123877 123881254395008 max_utils.py:756] Using (GB) 1.6 / 31.25 (5.120000%) on TPU_1(process=0,(1,0,0,0)) I0416 03:44:48.123897 123881254395008 max_utils.py:756] Using (GB) 1.6 / 31.25 (5.120000%) on TPU_2(process=0,(0,1,0,0)) I0416 03:44:48.123915 123881254395008 max_utils.py:756] Using (GB) 1.6 / 31.25 (5.120000%) on TPU_3(process=0,(1,1,0,0)) I0416 03:44:48.123932 123881254395008 max_utils.py:756] Using (GB) 1.6 / 31.25 (5.120000%) on TPU_4(process=0,(0,2,0,0)) I0416 03:44:48.123948 123881254395008 max_utils.py:756] Using (GB) 1.6 / 31.25 (5.120000%) on TPU_5(process=0,(1,2,0,0)) I0416 03:44:48.123965 123881254395008 max_utils.py:756] Using (GB) 1.6 / 31.25 (5.120000%) on TPU_6(process=0,(0,3,0,0)) I0416 03:44:48.123980 123881254395008 max_utils.py:756] Using (GB) 1.6 / 31.25 (5.120000%) on TPU_7(process=0,(1,3,0,0)) W0416 03:44:48.130029 3633034 pjrt_executable.cc:642] Assume version compatibility. PjRt-IFRT does not track XLA executable versions. W0416 03:44:48.135893 3633034 pjrt_executable.cc:642] Assume version compatibility. PjRt-IFRT does not track XLA executable versions. W0416 03:44:48.141446 3633034 pjrt_executable.cc:642] Assume version compatibility. PjRt-IFRT does not track XLA executable versions. I0416 03:44:48.164647 123765786674752 checkpoint.py:188] Wrote Metadata={'item_handlers': None, 'metrics': {}, 'performance_metrics': {}, 'init_timestamp_nsecs': 1776311087921738807, 'commit_timestamp_nsecs': None, 'custom_metadata': {}}, json={"item_handlers": null, "metrics": {}, "performance_metrics": {}, "init_timestamp_nsecs": 1776311087921738807, "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_04_int8/checkpoints/0/_CHECKPOINT_METADATA I0416 03:44:48.164857 123765755217472 async_checkpointer.py:265] [process=0][thread=save_finalize] Waiting for background save thread=async_save. I0416 03:44:48.475206 123881254395008 metric_logger.py:185] completed step: 0, seconds: 4.760, TFLOP/s/device: 2.854, Tokens/s/device: 430.208, total_weights: 13328, loss: 10.840 I0416 03:44:48.477124 123881254395008 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_04_int8/tensorboard/' I0416 03:44:48.482845 3636226 google_auth_provider.cc:149] Using credentials at ~/.config/gcloud/application_default_credentials.json I0416 03:44:48.482901 3636226 google_auth_provider.cc:156] Using OAuth2 AuthProvider I0416 03:44:48.594887 123881254395008 metric_logger.py:185] completed step: 1, seconds: 1.519, TFLOP/s/device: 8.945, Tokens/s/device: 1348.295, total_weights: 12332, loss: 10.869 I0416 03:44:48.717077 123881254395008 metric_logger.py:185] completed step: 2, seconds: 0.032, TFLOP/s/device: 428.346, Tokens/s/device: 64564.943, total_weights: 15161, loss: 9.874 I0416 03:44:48.838888 123881254395008 metric_logger.py:185] completed step: 3, seconds: 0.099, TFLOP/s/device: 137.309, Tokens/s/device: 20696.694, total_weights: 13327, loss: 9.406 I0416 03:44:49.215681 123765837006400 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_04_int8/checkpoints/0/items/array_metadatas/process_0 I0416 03:45:13.574162 123765816034880 base_pytree_checkpoint_handler.py:1217] [process=0][thread=write_metadata_after_commits] Commit + Array metadata written. Time taken: 25.088369s (commit=24.648435s, array_metadata_write=0.439934s) I0416 03:45:13.575396 123765765703232 base_pytree_checkpoint_handler.py:128] [process=0] /jax/checkpoint/write/gbytes_per_sec: 478.516 MiB/s (total gbytes: 12.3 GiB) (time elapsed: 26 seconds) (per-host) I0416 03:45:13.575512 123765765703232 async_checkpointer.py:90] [process=0][thread=async_save] 3 Handler Commit operations completed. Time taken: 25.452257s. I0416 03:45:13.785886 123765765703232 checkpoint.py:228] Read Metadata={'item_handlers': None, 'metrics': {}, 'performance_metrics': {}, 'init_timestamp_nsecs': 1776311087921738807, '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_04_int8/checkpoints/0/_CHECKPOINT_METADATA I0416 03:45:13.962899 123765765703232 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 03:45:14.162631 123765786674752 checkpoint.py:247] Updated Metadata={'item_handlers': {'items': 'orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeCheckpointHandler'}, 'metrics': {}, 'performance_metrics': {}, 'init_timestamp_nsecs': 1776311087921738807, '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_04_int8/checkpoints/0/_CHECKPOINT_METADATA I0416 03:45:14.421414 123765765703232 ocdbt_utils.py:56] Param validation support for Zarr3 will be added later (b/362328389). I0416 03:45:14.422087 123765765703232 base_pytree_checkpoint_handler.py:1342] [process=0][thread=async_save] Pytree save finalize (merge_ocdbt + ArrayMetadata validation) completed. Time taken: 0.598282s. 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_04_int8/checkpoints/0/items I0416 03:45:14.422737 123765765703232 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_04_int8/checkpoints/0/items I0416 03:45:14.643705 123765765703232 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_04_int8/checkpoints/0 I0416 03:45:15.299770 123765765703232 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_04_int8/checkpoints/0`. I0416 03:45:15.300427 123765765703232 async_checkpointer.py:420] Finished async_save (blocking + background). Time taken: 28.220323s. directory=gs://wanglance-maxtext/linen_ckpt_feat_nnx_trainstate_and_training_loop_20260416_033532/linen_feat_nnx_trainstate_and_training_loop_20260416_033532_04_int8/checkpoints/0 I0416 03:45:15.300503 123765765703232 async_checkpointer.py:144] [process=0][thread=async_save] Background save thread done. Time taken: 27.177249s. I0416 03:45:15.300674 123765755217472 async_checkpointer.py:273] [process=0][thread=save_finalize] Done with waiting for background save thread=async_save. I0416 03:45:15.300781 123765755217472 async_checkpointer.py:283] [process=0][thread=save_finalize] No errors found in background save thread=async_save. I0416 03:45:15.300858 123765755217472 checkpoint_manager.py:2103] [process=0][thread=save_finalize][step=0] CheckpointManager Save Finalize is syncing with other hosts... I0416 03:45:15.300915 123765755217472 checkpoint_manager.py:2112] [process=0][thread=save_finalize][step=0] CheckpointManager Save Finalize is done on all hosts. I0416 03:45:17.572753 123881254395008 metric_logger.py:185] completed step: 4, seconds: 0.122, TFLOP/s/device: 111.275, Tokens/s/device: 16772.587, total_weights: 11939, loss: 9.149 I0416 03:45:17.582299 123881254395008 metric_logger.py:185] completed step: 5, seconds: 0.122, TFLOP/s/device: 111.481, Tokens/s/device: 16803.689, total_weights: 15502, loss: 9.058 I0416 03:45:17.702988 123881254395008 metric_logger.py:185] completed step: 6, seconds: 28.733, TFLOP/s/device: 0.473, Tokens/s/device: 71.276, total_weights: 13864, loss: 8.892 I0416 03:45:17.825243 123881254395008 metric_logger.py:185] completed step: 7, seconds: 0.006, TFLOP/s/device: 2133.657, Tokens/s/device: 321608.040, total_weights: 12988, loss: 8.751 I0416 03:45:17.947813 123881254395008 metric_logger.py:185] completed step: 8, seconds: 0.122, TFLOP/s/device: 111.755, Tokens/s/device: 16844.876, total_weights: 13820, loss: 8.791 I0416 03:45:18.069291 123881254395008 checkpointing.py:794] Waiting for step 9 to finish before checkpoint... I0416 03:45:18.070083 123881254395008 checkpointing.py:798] Waited 0.0008192062377929688 seconds for step 9 to finish before starting checkpointing. I0416 03:45:18.070336 123881254395008 checkpoint_manager.py:1983] [process=0][thread=MainThread][wait_until_finished] No Save Finalize thread to wait for. Returning. I0416 03:45:18.070496 123881254395008 checkpoint_manager.py:1501] [process=0] Saving checkpoint at step 9 I0416 03:45:18.070771 123881254395008 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_04_int8/checkpoints/9. I0416 03:45:18.229595 123765755217472 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_04_int8/checkpoints/9 I0416 03:45:18.241662 123881254395008 jax_array_handlers.py:347] Scheduling D2H of 39 prioritized jax.Array. I0416 03:45:18.241755 123881254395008 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 03:45:18.921509 123765816034880 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_04_int8/checkpoints/9/items I0416 03:45:25.995700 123881254395008 base_pytree_checkpoint_handler.py:153] [process=0][thread=MainThread] Initiated "orbax.checkpoint._src.serialization.jax_array_handlers.ArrayHandler".serialize. Time taken: 7.754498s I0416 03:45:26.000981 123881254395008 base_pytree_checkpoint_handler.py:128] [process=0] /jax/checkpoint/write/blocking_gbytes_per_sec: 1.571 GiB/s (total gbytes: 12.3 GiB) (time elapsed: 7 seconds) (per-host) I0416 03:45:26.001074 123881254395008 base_pytree_checkpoint_handler.py:732] [process=0][thread=MainThread] Initiated Pytree async_save. Time taken: 7.853403s (batch_requests_ready=0.090572s, total_serialization_initiated=7.757713s, others=0.005117s) I0416 03:45:26.001165 123881254395008 composite_checkpoint_handler.py:715] [process=0][thread=MainThread] Initiated CompositeCheckpointHandler.async_save. Time taken: 7.854199s (all_items=0.000018s, per_item={'items': '0.00001764'}, temp_paths=7.854181) I0416 03:45:26.003727 123765620999744 async_checkpointer.py:79] [process=0][thread=async_save] Background save thread started. I0416 03:45:26.003878 123881254395008 async_checkpointer.py:561] Finished blocking save. Time taken: 7.933335s. 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_04_int8/checkpoints/9. I0416 03:45:26.004187 123881254395008 checkpoint_manager.py:1549] [process=0][thread=MainThread][step=9] Starting CheckpointManager Save Finalize thread=save_finalize I0416 03:45:26.004440 123765765703232 async_checkpointer.py:265] [process=0][thread=save_finalize] Waiting for background save thread=async_save. I0416 03:45:26.004619 123881254395008 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_04_int8/checkpoints', 'reached_preemption': False, 'preemption_received_at': None, 'synchronous': False, 'wait_for_prev_start_time': 1776311118.0703006, 'wait_for_prev_duration_secs': 5.984306335449219e-05, 'time_between_consecutive_saves_sec': 2.7693700790405273, 'checkpointer_blocking_start_time': 1776311118.0705173, 'checkpointer_blocking_duration_secs': 7.93351411819458, 'get_old_steps_start_time': 1776311126.0040715, 'get_old_steps_duration_secs': 6.127357482910156e-05, 'checkpoint_manager_blocking_start_time': 1776311118.0702677, 'checkpoint_manager_blocking_duration_secs': 7.934315204620361} I0416 03:45:26.004780 123881254395008 checkpointing.py:409] Started an asynchronous checkpoint save for step 9 I0416 03:45:26.004820 123881254395008 checkpoint_manager.py:1994] [process=0][thread=MainThread][step=9][wait_until_finished] Waiting for Save Finalize thread (save_finalize) to complete. I0416 03:45:26.744187 123765837006400 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_04_int8/checkpoints/9/items/array_metadatas/process_0 I0416 03:45:57.639235 123765816034880 base_pytree_checkpoint_handler.py:1217] [process=0][thread=write_metadata_after_commits] Commit + Array metadata written. Time taken: 31.637350s (commit=31.163842s, array_metadata_write=0.473508s) I0416 03:45:57.641583 123765620999744 base_pytree_checkpoint_handler.py:128] [process=0] /jax/checkpoint/write/gbytes_per_sec: 319.969 MiB/s (total gbytes: 12.3 GiB) (time elapsed: 39 seconds) (per-host) I0416 03:45:57.641772 123765620999744 async_checkpointer.py:90] [process=0][thread=async_save] 3 Handler Commit operations completed. Time taken: 31.637835s. I0416 03:45:58.095602 123765620999744 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 03:45:58.452950 123765620999744 ocdbt_utils.py:56] Param validation support for Zarr3 will be added later (b/362328389). I0416 03:45:58.453844 123765620999744 base_pytree_checkpoint_handler.py:1342] [process=0][thread=async_save] Pytree save finalize (merge_ocdbt + ArrayMetadata validation) completed. Time taken: 0.552897s. 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_04_int8/checkpoints/9/items I0416 03:45:58.454772 123765620999744 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_04_int8/checkpoints/9/items I0416 03:45:58.696241 123765620999744 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_04_int8/checkpoints/9 I0416 03:45:59.381438 123765620999744 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_04_int8/checkpoints/9`. I0416 03:45:59.382379 123765620999744 async_checkpointer.py:420] Finished async_save (blocking + background). Time taken: 41.311848s. directory=gs://wanglance-maxtext/linen_ckpt_feat_nnx_trainstate_and_training_loop_20260416_033532/linen_feat_nnx_trainstate_and_training_loop_20260416_033532_04_int8/checkpoints/9 I0416 03:45:59.382505 123765620999744 async_checkpointer.py:144] [process=0][thread=async_save] Background save thread done. Time taken: 33.378571s. I0416 03:45:59.382874 123765765703232 async_checkpointer.py:273] [process=0][thread=save_finalize] Done with waiting for background save thread=async_save. I0416 03:45:59.383040 123765765703232 async_checkpointer.py:283] [process=0][thread=save_finalize] No errors found in background save thread=async_save. I0416 03:45:59.383186 123765765703232 checkpoint_manager.py:2103] [process=0][thread=save_finalize][step=9] CheckpointManager Save Finalize is syncing with other hosts... I0416 03:45:59.383249 123765765703232 checkpoint_manager.py:2112] [process=0][thread=save_finalize][step=9] CheckpointManager Save Finalize is done on all hosts. I0416 03:45:59.383626 123881254395008 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 03:45:59.384639 123881254395008 checkpoint_manager.py:1983] [process=0][thread=MainThread][wait_until_finished] No Save Finalize thread to wait for. Returning. I0416 03:45:59.386377 123881254395008 metric_logger.py:185] completed step: 9, seconds: 0.122, TFLOP/s/device: 111.381, Tokens/s/device: 16788.537, total_weights: 12300, loss: 8.774 Per train step: Total TFLOPs: 13.59 split as 93.93% learnable weight flops and 6.07% attention flops