2026-04-17 13:42:36.086997: 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) I0417 13:42:36.203429 126746257050752 max_utils.py:238] Skipping jax distributed system due to skip_jax_distributed_system=True flag. I0417 13:43:24.760071 126746257050752 max_utils.py:800] System Information: Jax Version: 0.9.2 I0417 13:43:24.760186 126746257050752 max_utils.py:801] System Information: Jaxlib Version: 0.9.2 I0417 13:43:24.760223 126746257050752 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 I0417 13:43:24.760255 126746257050752 train_utils.py:335] WARNING: 'dataset_path' might be pointing your local file system I0417 13:43:24.760279 126746257050752 train_utils.py:348] WARNING: Sequence packing is essentially ignored for synthetic data. Please use a real dataset to use sequence packing. I0417 13:43:24.760390 126746257050752 train.py:671] [DECOUPLED NO-OP] skipping cloud diagnostics wrapper. W0417 13:43:24.853406 485973 pjrt_executable.cc:642] Assume version compatibility. PjRt-IFRT does not track XLA executable versions. I0417 13:43:25.255290 126746257050752 maxtext_utils.py:1548] Num_devices: 8, shape (1, 1, 1, 8, 1, 1, 1, 1, 1, 1, 1, 1, 1) I0417 13:43:25.255647 126746257050752 checkpointing.py:677] Setting up checkpoint logger... I0417 13:43:25.255695 126746257050752 checkpointing.py:233] Creating checkpoint manager with ocdbt=True and zarr3=True I0417 13:43:25.255757 126746257050752 pytree_checkpoint_handler.py:577] save_device_host_concurrent_bytes=None I0417 13:43:25.256432 126746257050752 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 0x7345d2bf8b00>, enable_pinned_host_transfer=False, save_concurrent_bytes: 96000000000 (89.4 GiB), restore_concurrent_bytes: 96000000000 (89.4 GiB) I0417 13:43:26.942419 126746257050752 checkpointing.py:265] Enabling policy for fixed interval checkpointing. I0417 13:43:26.942613 126746257050752 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 0x733f56932300>}, handler_registry=None I0417 13:43:26.942937 126746257050752 composite_checkpoint_handler.py:237] Deferred registration for item: "items". Adding handler `<orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeCheckpointHandler object at 0x733f56932300>` 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`. I0417 13:43:26.942980 126746257050752 composite_checkpoint_handler.py:237] Deferred registration for item: "metrics". Adding handler `<orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonCheckpointHandler object at 0x733f569325d0>` 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`. I0417 13:43:26.943008 126746257050752 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 0x733f56932300>, ('items', <class 'orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeRestoreArgs'>): <orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeCheckpointHandler object at 0x733f56932300>, ('metrics', <class 'orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonSaveArgs'>): <orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonCheckpointHandler object at 0x733f569325d0>, ('metrics', <class 'orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonRestoreArgs'>): <orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonCheckpointHandler object at 0x733f569325d0>}). I0417 13:43:26.943320 126746257050752 abstract_checkpointer.py:35] orbax-checkpoint version: 0.11.28 I0417 13:43:27.350917 126640132720192 checkpoint.py:228] Read Metadata={'item_handlers': {'items': 'orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeCheckpointHandler'}, 'metrics': {}, 'performance_metrics': {}, 'init_timestamp_nsecs': 1776433277347791063, 'commit_timestamp_nsecs': 1776433328582315048, 'custom_metadata': {}} from gs://wanglance-maxtext/linen_ckpt_main_20260417_125240/linen_main_20260417_125240_14_async_ckpt_false/checkpoints/9/_CHECKPOINT_METADATA I0417 13:43:27.435071 126640132720192 checkpoint_manager.py:1847] Missing metrics for step 9 I0417 13:43:27.479986 126746257050752 checkpoint_manager.py:1788] Found 3 checkpoint steps in gs://wanglance-maxtext/linen_ckpt_main_20260417_125240/linen_main_20260417_125240_14_async_ckpt_false/checkpoints I0417 13:43:27.480495 126746257050752 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=False, 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=5), preservation_policy=LatestN(n=None), prevent_write_metrics=False, enable_should_save_is_saving_in_progress_check=True, enable_per_process_directory_creation=False), root_directory=gs://wanglance-maxtext/linen_ckpt_main_20260417_125240/linen_main_20260417_125240_14_async_ckpt_false/checkpoints: <orbax.checkpoint.checkpoint_manager.CheckpointManager object at 0x733f56932570> I0417 13:43:27.480588 126746257050752 checkpointing.py:301] Checkpoint manager created! W0417 13:43:27.497486 485973 pjrt_executable.cc:642] Assume version compatibility. PjRt-IFRT does not track XLA executable versions. I0417 13:43:28.038085 126746257050752 nnx_wrappers.py:437] Unknown Logical: bfloat16[8,2048,2048]....................................... ('activation_batch', 'activation_norm_length', 'activation_embed'). I0417 13:43:28.038183 126746257050752 nnx_wrappers.py:437] Unknown Physical: bfloat16[8,2048,2048]....................................... ('fsdp', None, None). I0417 13:43:28.121722 126746257050752 attentions.py:1088] attentions/inputs_q Logical: bfloat16[8,2048,2048]....................................... ('activation_batch', 'activation_attn_length', 'activation_attn_embed'). I0417 13:43:28.121800 126746257050752 attentions.py:1088] attentions/inputs_q Physical: bfloat16[8,2048,2048]....................................... ('fsdp', None, None). I0417 13:43:28.135885 126746257050752 attentions.py:1089] attentions/inputs_kv Logical: bfloat16[8,2048,2048]....................................... ('activation_batch', 'activation_attn_length', 'activation_attn_embed'). I0417 13:43:28.135935 126746257050752 attentions.py:1089] attentions/inputs_kv Physical: bfloat16[8,2048,2048]....................................... ('fsdp', None, None). I0417 13:43:28.157003 126746257050752 attentions.py:1154] attentions/query Logical: bfloat16[8,2048,16,128]..................................... ('activation_kv_batch', 'activation_attn_length', 'activation_kv_heads', 'activation_kv_head_dim'). I0417 13:43:28.157077 126746257050752 attentions.py:1154] attentions/query Physical: bfloat16[8,2048,16,128]..................................... ('fsdp', None, None, None). I0417 13:43:28.171203 126746257050752 attentions.py:1155] attentions/key Logical: bfloat16[8,2048,16,128]..................................... ('activation_kv_batch', 'activation_attn_length', 'activation_kv_heads', 'activation_kv_head_dim'). I0417 13:43:28.171251 126746257050752 attentions.py:1155] attentions/key Physical: bfloat16[8,2048,16,128]..................................... ('fsdp', None, None, None). I0417 13:43:28.185361 126746257050752 attentions.py:1156] attentions/value Logical: bfloat16[8,2048,16,128]..................................... ('activation_kv_batch', 'activation_attn_length', 'activation_kv_heads', 'activation_kv_head_dim'). I0417 13:43:28.185409 126746257050752 attentions.py:1156] attentions/value Physical: bfloat16[8,2048,16,128]..................................... ('fsdp', None, None, None). I0417 13:43:28.206843 126746257050752 attentions.py:1197] attentions/out Logical: bfloat16[8,2048,16,128]..................................... ('activation_batch', 'activation_attn_length', 'activation_heads', 'activation_kv'). I0417 13:43:28.206902 126746257050752 attentions.py:1197] attentions/out Physical: bfloat16[8,2048,16,128]..................................... ('fsdp', None, None, None). I0417 13:43:28.224983 126746257050752 linears.py:525] linears/x Logical: bfloat16[8,2048,7168]....................................... ('activation_batch', 'activation_length', 'activation_mlp'). I0417 13:43:28.225035 126746257050752 linears.py:525] linears/x Physical: bfloat16[8,2048,7168]....................................... ('fsdp', None, None). I0417 13:43:28.419026 126746257050752 checkpointing.py:577] checkpoint manager exists so trying to load this run's existing checkpoint I0417 13:43:28.419133 126746257050752 checkpointing.py:581] restoring from this run's directory step 9 I0417 13:43:28.578481 126746257050752 checkpointer.py:304] Restoring checkpoint from gs://wanglance-maxtext/linen_ckpt_main_20260417_125240/linen_main_20260417_125240_14_async_ckpt_false/checkpoints/9. I0417 13:43:29.103065 487608 google_auth_provider.cc:149] Using credentials at ~/.config/gcloud/application_default_credentials.json I0417 13:43:29.103121 487608 google_auth_provider.cc:156] Using OAuth2 AuthProvider I0417 13:43:35.274184 126746257050752 base_pytree_checkpoint_handler.py:128] [process=0] /jax/checkpoint/read/gbytes_per_sec: 1.944 GiB/s (total gbytes: 12.3 GiB) (time elapsed: 6 seconds) (per-host) I0417 13:43:35.274437 126746257050752 checkpointer.py:318] Finished restoring checkpoint in 6.77 seconds from gs://wanglance-maxtext/linen_ckpt_main_20260417_125240/linen_main_20260417_125240_14_async_ckpt_false/checkpoints/9. I0417 13:43:35.274506 126746257050752 standard_logger.py:34] {'step': 9, 'event_type': 'restore', 'directory': 'gs://wanglance-maxtext/linen_ckpt_main_20260417_125240/linen_main_20260417_125240_14_async_ckpt_false/checkpoints', 'checkpointer_start_time': 1776433408.502181, 'checkpointer_duration_secs': 6.772300958633423, 'checkpoint_manager_start_time': 1776433408.4193385, 'checkpoint_manager_duration_secs': 6.85514497756958} [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 I0417 13:43:35.798294 126746257050752 maxtext_utils.py:1651] params/params/decoder/decoder_norm/scale Shape: float32[2048] Logical: P('norm',) Physical: (None,) I0417 13:43:35.798403 126746257050752 maxtext_utils.py:1651] params/params/decoder/layers/mlp/wi_0/kernel Shape: float32[2048,16,7168] Logical: P('embed', 'layers', 'mlp') Physical: ('fsdp', None, None) I0417 13:43:35.798444 126746257050752 maxtext_utils.py:1651] params/params/decoder/layers/mlp/wi_1/kernel Shape: float32[2048,16,7168] Logical: P('embed', 'layers', 'mlp') Physical: ('fsdp', None, None) I0417 13:43:35.798487 126746257050752 maxtext_utils.py:1651] params/params/decoder/layers/mlp/wo/kernel Shape: float32[7168,16,2048] Logical: P('mlp', 'layers', 'embed') Physical: (None, None, 'fsdp') I0417 13:43:35.798527 126746257050752 maxtext_utils.py:1651] params/params/decoder/layers/post_self_attention_layer_norm/scale Shape: float32[2048,16] Logical: P('norm', 'layers') Physical: (None, None) I0417 13:43:35.798558 126746257050752 maxtext_utils.py:1651] params/params/decoder/layers/pre_self_attention_layer_norm/scale Shape: float32[2048,16] Logical: P('norm', 'layers') Physical: (None, None) I0417 13:43:35.798599 126746257050752 maxtext_utils.py:1651] 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) I0417 13:43:35.798637 126746257050752 maxtext_utils.py:1651] 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') I0417 13:43:35.798666 126746257050752 maxtext_utils.py:1651] 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) I0417 13:43:35.798700 126746257050752 maxtext_utils.py:1651] 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) I0417 13:43:35.798738 126746257050752 maxtext_utils.py:1651] params/params/decoder/logits_dense/kernel Shape: float32[2048,32000] Logical: P('embed_vocab', 'vocab') Physical: ('fsdp', None) I0417 13:43:35.798773 126746257050752 maxtext_utils.py:1651] params/params/token_embedder/embedding Shape: float32[32000,2048] Logical: P('vocab', 'embed_vocab') Physical: (None, 'fsdp') I0417 13:43:36.360238 126746257050752 train.py:155] train/xent Logical: float32[8,2048]............................................. ('activation_embed_and_logits_batch', 'activation_length'). I0417 13:43:36.360347 126746257050752 train.py:155] train/xent Physical: float32[8,2048]............................................. ('fsdp', None). I0417 13:43:36.374852 126746257050752 train.py:162] train/z_loss Logical: float32[8,2048]............................................. ('activation_embed_and_logits_batch', 'activation_length'). I0417 13:43:36.374902 126746257050752 train.py:162] train/z_loss Physical: float32[8,2048]............................................. ('fsdp', None). W0417 13:43:36.713682 485973 pjrt_executable.cc:642] Assume version compatibility. PjRt-IFRT does not track XLA executable versions. I0417 13:43:36.779981 126746257050752 max_utils.py:791] Total memory size: 3.6 GB, Output size: 1.5 GB, Temp size: 2.0 GB, Argument size: 1.5 GB, Host temp size: 0.0 GB. I0417 13:43:36.815873 126746257050752 max_utils.py:194] tensorboardX not available; using no-op SummaryWriter. I0417 13:43:36.816082 126746257050752 metric_logger.py:301] number parameters: 1.104 billion W0417 13:43:37.755218 485973 pjrt_executable.cc:642] Assume version compatibility. PjRt-IFRT does not track XLA executable versions. I0417 13:43:37.824975 126746257050752 checkpointing.py:772] Waiting for step 10 to finish before checkpoint... I0417 13:43:37.941327 126746257050752 checkpointing.py:776] Waited 0.11634302139282227 seconds for step 10 to finish before starting checkpointing. I0417 13:43:37.941767 126746257050752 checkpoint_manager.py:1983] [process=0][thread=MainThread][wait_until_finished] No Save Finalize thread to wait for. Returning. I0417 13:43:37.941925 126746257050752 checkpoint_manager.py:1501] [process=0] Saving checkpoint at step 10 I0417 13:43:37.941999 126746257050752 checkpointer.py:241] [process=0] Started saving checkpoint to gs://wanglance-maxtext/linen_ckpt_main_20260417_125240/linen_main_20260417_125240_14_async_ckpt_false/checkpoints/10. I0417 13:43:38.109423 126746257050752 atomicity.py:137] Creating tmp directory gs://wanglance-maxtext/linen_ckpt_main_20260417_125240/linen_main_20260417_125240_14_async_ckpt_false/checkpoints/10 I0417 13:43:38.937727 126746257050752 checkpoint.py:188] Wrote Metadata={'item_handlers': None, 'metrics': {}, 'performance_metrics': {}, 'init_timestamp_nsecs': 1776433418699673789, 'commit_timestamp_nsecs': None, 'custom_metadata': {}}, json={"item_handlers": null, "metrics": {}, "performance_metrics": {}, "init_timestamp_nsecs": 1776433418699673789, "commit_timestamp_nsecs": null, "custom_metadata": {}} to gs://wanglance-maxtext/linen_ckpt_main_20260417_125240/linen_main_20260417_125240_14_async_ckpt_false/checkpoints/10/_CHECKPOINT_METADATA I0417 13:43:38.938713 126746257050752 signaling_client.py:373] Using ThreadSafeKeyValueSignalingClient W0417 13:43:38.939041 126746257050752 future.py:328] [process=0][thread=MainThread][operation_id=1] _SignalingThread.join() waiting for signals ([]) blocking the main thread will slow down blocking save times. This is likely due to main thread calling result() on a CommitFuture. I0417 13:43:39.046356 126638614382144 atomicity.py:137] Creating tmp directory gs://wanglance-maxtext/linen_ckpt_main_20260417_125240/linen_main_20260417_125240_14_async_ckpt_false/checkpoints/10/items I0417 13:43:39.596635 126746257050752 jax_array_handlers.py:347] Scheduling D2H of 39 prioritized jax.Array. I0417 13:43:39.596746 126746257050752 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 W0417 13:43:39.847510 485973 pjrt_executable.cc:642] Assume version compatibility. PjRt-IFRT does not track XLA executable versions. W0417 13:43:39.882816 485973 pjrt_executable.cc:642] Assume version compatibility. PjRt-IFRT does not track XLA executable versions. W0417 13:43:39.887996 485973 pjrt_executable.cc:642] Assume version compatibility. PjRt-IFRT does not track XLA executable versions. W0417 13:43:39.892365 485973 pjrt_executable.cc:642] Assume version compatibility. PjRt-IFRT does not track XLA executable versions. W0417 13:43:39.896538 485973 pjrt_executable.cc:642] Assume version compatibility. PjRt-IFRT does not track XLA executable versions. W0417 13:43:39.900532 485973 pjrt_executable.cc:642] Assume version compatibility. PjRt-IFRT does not track XLA executable versions. W0417 13:43:39.904517 485973 pjrt_executable.cc:642] Assume version compatibility. PjRt-IFRT does not track XLA executable versions. W0417 13:43:39.908479 485973 pjrt_executable.cc:642] Assume version compatibility. PjRt-IFRT does not track XLA executable versions. I0417 13:43:43.622260 126746257050752 base_pytree_checkpoint_handler.py:153] [process=0][thread=MainThread] Initiated "orbax.checkpoint._src.serialization.jax_array_handlers.ArrayHandler".serialize. Time taken: 4.025988s I0417 13:43:43.626616 126746257050752 base_pytree_checkpoint_handler.py:128] [process=0] /jax/checkpoint/write/blocking_gbytes_per_sec: 2.991 GiB/s (total gbytes: 12.3 GiB) (time elapsed: 4 seconds) (per-host) I0417 13:43:43.626678 126746257050752 base_pytree_checkpoint_handler.py:732] [process=0][thread=MainThread] Initiated Pytree async_save. Time taken: 4.125725s (batch_requests_ready=0.092675s, total_serialization_initiated=4.028802s, others=0.004247s) I0417 13:43:43.626750 126746257050752 composite_checkpoint_handler.py:715] [process=0][thread=MainThread] Initiated CompositeCheckpointHandler.async_save. Time taken: 4.688116s (all_items=0.000046s, per_item={'items': '0.00004601'}, temp_paths=4.688070) W0417 13:43:43.626787 126746257050752 future.py:328] [process=0][thread=MainThread][operation_id=1] _SignalingThread.join() waiting for signals ([]) blocking the main thread will slow down blocking save times. This is likely due to main thread calling result() on a CommitFuture. I0417 13:43:44.403938 126638624867904 array_metadata_store.py:203] [process=0][thread=array_type_handler] Wrote 39 array_metadata.ArrayMetadata to gs://wanglance-maxtext/linen_ckpt_main_20260417_125240/linen_main_20260417_125240_14_async_ckpt_false/checkpoints/10/items/array_metadatas/process_0 I0417 13:44:26.513381 126638991869504 base_pytree_checkpoint_handler.py:1217] [process=0][thread=write_metadata_after_commits] Commit + Array metadata written. Time taken: 42.886379s (commit=42.410340s, array_metadata_write=0.476038s) I0417 13:44:26.514418 126746257050752 base_pytree_checkpoint_handler.py:128] [process=0] /jax/checkpoint/write/gbytes_per_sec: 268.791 MiB/s (total gbytes: 12.3 GiB) (time elapsed: 47 seconds) (per-host) I0417 13:44:27.158434 126746257050752 checkpoint.py:247] Updated Metadata={'item_handlers': {'items': 'orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeCheckpointHandler'}, 'metrics': {}, 'performance_metrics': {}, 'init_timestamp_nsecs': 1776433418699673789, 'commit_timestamp_nsecs': None, 'custom_metadata': {}} to gs://wanglance-maxtext/linen_ckpt_main_20260417_125240/linen_main_20260417_125240_14_async_ckpt_false/checkpoints/10/_CHECKPOINT_METADATA I0417 13:44:27.302257 126746257050752 array_metadata_store.py:367] [process=0][thread=MainThread] Skipped cross-host ArrayMetadata validation because only one process is found: process_index=0. I0417 13:44:27.716438 126746257050752 ocdbt_utils.py:56] Param validation support for Zarr3 will be added later (b/362328389). I0417 13:44:27.716597 126746257050752 base_pytree_checkpoint_handler.py:1342] [process=0][thread=MainThread] Pytree save finalize (merge_ocdbt + ArrayMetadata validation) completed. Time taken: 0.557787s. use_zarr3=True, enable_post_merge_validation=True, directory=gs://wanglance-maxtext/linen_ckpt_main_20260417_125240/linen_main_20260417_125240_14_async_ckpt_false/checkpoints/10/items I0417 13:44:27.717240 126746257050752 atomicity.py:608] Finalizing gs://wanglance-maxtext/linen_ckpt_main_20260417_125240/linen_main_20260417_125240_14_async_ckpt_false/checkpoints/10/items I0417 13:44:27.984555 126746257050752 atomicity.py:608] Finalizing gs://wanglance-maxtext/linen_ckpt_main_20260417_125240/linen_main_20260417_125240_14_async_ckpt_false/checkpoints/10 I0417 13:44:28.642407 126746257050752 atomicity.py:794] [process=0][thread=MainThread] Finished saving checkpoint (finalized tmp dir) to `gs://wanglance-maxtext/linen_ckpt_main_20260417_125240/linen_main_20260417_125240_14_async_ckpt_false/checkpoints/10`. I0417 13:44:28.643071 126746257050752 checkpointer.py:290] Finished synchronous save in 50.70 seconds to gs://wanglance-maxtext/linen_ckpt_main_20260417_125240/linen_main_20260417_125240_14_async_ckpt_false/checkpoints/10 I0417 13:44:28.643248 126746257050752 checkpoint_manager.py:2103] [process=0][thread=MainThread][step=10] CheckpointManager Save Finalize is syncing with other hosts... I0417 13:44:28.643298 126746257050752 checkpoint_manager.py:2112] [process=0][thread=MainThread][step=10] CheckpointManager Save Finalize is done on all hosts. I0417 13:44:28.643320 126746257050752 checkpoint_manager.py:1570] [process=0][thread=MainThread][step=10] Finished synchronous save. I0417 13:44:28.643356 126746257050752 standard_logger.py:34] {'step': 10, 'event_type': 'save', 'directory': 'gs://wanglance-maxtext/linen_ckpt_main_20260417_125240/linen_main_20260417_125240_14_async_ckpt_false/checkpoints', 'reached_preemption': False, 'preemption_received_at': None, 'synchronous': True, 'wait_for_prev_start_time': 1776433417.9417534, 'wait_for_prev_duration_secs': 3.933906555175781e-05, 'time_between_consecutive_saves_sec': None, 'checkpointer_blocking_start_time': 1776433417.9419525, 'checkpointer_blocking_duration_secs': 50.70115923881531, 'get_old_steps_start_time': 1776433468.643132, 'get_old_steps_duration_secs': 3.337860107421875e-05, 'checkpoint_manager_blocking_start_time': 1776433417.9416497, 'checkpoint_manager_blocking_duration_secs': 50.70168495178223} I0417 13:44:28.643453 126746257050752 checkpointing.py:410] Saved a checkpoint at step 10. I0417 13:44:28.643493 126746257050752 max_utils.py:750] Memstats: After params initialized: I0417 13:44:28.643543 126746257050752 max_utils.py:756] Using (GB) 1.59 / 31.25 (5.088000%) on TPU_0(process=0,(0,0,0,0)) I0417 13:44:28.643568 126746257050752 max_utils.py:756] Using (GB) 1.59 / 31.25 (5.088000%) on TPU_1(process=0,(1,0,0,0)) I0417 13:44:28.643589 126746257050752 max_utils.py:756] Using (GB) 1.59 / 31.25 (5.088000%) on TPU_2(process=0,(0,1,0,0)) I0417 13:44:28.643608 126746257050752 max_utils.py:756] Using (GB) 1.59 / 31.25 (5.088000%) on TPU_3(process=0,(1,1,0,0)) I0417 13:44:28.643633 126746257050752 max_utils.py:756] Using (GB) 1.59 / 31.25 (5.088000%) on TPU_4(process=0,(0,2,0,0)) I0417 13:44:28.643652 126746257050752 max_utils.py:756] Using (GB) 1.59 / 31.25 (5.088000%) on TPU_5(process=0,(1,2,0,0)) I0417 13:44:28.643671 126746257050752 max_utils.py:756] Using (GB) 1.59 / 31.25 (5.088000%) on TPU_6(process=0,(0,3,0,0)) I0417 13:44:28.643689 126746257050752 max_utils.py:756] Using (GB) 1.59 / 31.25 (5.088000%) on TPU_7(process=0,(1,3,0,0)) W0417 13:44:28.649981 485973 pjrt_executable.cc:642] Assume version compatibility. PjRt-IFRT does not track XLA executable versions. W0417 13:44:28.655610 485973 pjrt_executable.cc:642] Assume version compatibility. PjRt-IFRT does not track XLA executable versions. W0417 13:44:28.660189 485973 pjrt_executable.cc:642] Assume version compatibility. PjRt-IFRT does not track XLA executable versions. I0417 13:44:28.849907 126746257050752 metric_logger.py:196] completed step: 10, seconds: 1.008, TFLOP/s/device: 13.474, Tokens/s/device: 2030.970, total_weights: 16384, loss: 3.602, lm_loss: 3.602, perplexity: 36.686 I0417 13:44:28.964826 126746257050752 metric_logger.py:196] completed step: 11, seconds: 51.022, TFLOP/s/device: 0.266, Tokens/s/device: 40.140, total_weights: 16384, loss: 3.101, lm_loss: 3.101, perplexity: 22.211 I0417 13:44:29.083247 126746257050752 metric_logger.py:196] completed step: 12, seconds: 0.007, TFLOP/s/device: 2075.956, Tokens/s/device: 312910.619, total_weights: 16384, loss: 2.684, lm_loss: 2.684, perplexity: 14.644 I0417 13:44:29.201081 126746257050752 metric_logger.py:196] completed step: 13, seconds: 0.116, TFLOP/s/device: 117.347, Tokens/s/device: 17687.803, total_weights: 16384, loss: 2.348, lm_loss: 2.348, perplexity: 10.461 I0417 13:44:29.205447 126746257050752 checkpointing.py:772] Waiting for step 15 to finish before checkpoint... I0417 13:44:29.436928 126746257050752 checkpointing.py:776] Waited 0.23145127296447754 seconds for step 15 to finish before starting checkpointing. I0417 13:44:29.437325 126746257050752 checkpoint_manager.py:1983] [process=0][thread=MainThread][wait_until_finished] No Save Finalize thread to wait for. Returning. I0417 13:44:29.437470 126746257050752 checkpoint_manager.py:1501] [process=0] Saving checkpoint at step 15 I0417 13:44:29.437532 126746257050752 checkpointer.py:241] [process=0] Started saving checkpoint to gs://wanglance-maxtext/linen_ckpt_main_20260417_125240/linen_main_20260417_125240_14_async_ckpt_false/checkpoints/15. I0417 13:44:29.605153 126746257050752 atomicity.py:137] Creating tmp directory gs://wanglance-maxtext/linen_ckpt_main_20260417_125240/linen_main_20260417_125240_14_async_ckpt_false/checkpoints/15 W0417 13:44:30.422093 126746257050752 future.py:328] [process=0][thread=MainThread][operation_id=2] _SignalingThread.join() waiting for signals ([]) blocking the main thread will slow down blocking save times. This is likely due to main thread calling result() on a CommitFuture. I0417 13:44:30.508669 126638991869504 atomicity.py:137] Creating tmp directory gs://wanglance-maxtext/linen_ckpt_main_20260417_125240/linen_main_20260417_125240_14_async_ckpt_false/checkpoints/15/items I0417 13:44:31.053908 126746257050752 jax_array_handlers.py:347] Scheduling D2H of 39 prioritized jax.Array. I0417 13:44:31.054012 126746257050752 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 I0417 13:44:36.001160 126746257050752 base_pytree_checkpoint_handler.py:153] [process=0][thread=MainThread] Initiated "orbax.checkpoint._src.serialization.jax_array_handlers.ArrayHandler".serialize. Time taken: 4.947612s I0417 13:44:36.005277 126746257050752 base_pytree_checkpoint_handler.py:128] [process=0] /jax/checkpoint/write/blocking_gbytes_per_sec: 2.449 GiB/s (total gbytes: 12.3 GiB) (time elapsed: 5 seconds) (per-host) I0417 13:44:36.005346 126746257050752 base_pytree_checkpoint_handler.py:732] [process=0][thread=MainThread] Initiated Pytree async_save. Time taken: 5.039399s (batch_requests_ready=0.084632s, total_serialization_initiated=4.950756s, others=0.004011s) I0417 13:44:36.005419 126746257050752 composite_checkpoint_handler.py:715] [process=0][thread=MainThread] Initiated CompositeCheckpointHandler.async_save. Time taken: 5.583670s (all_items=0.000022s, per_item={'items': '0.00002170'}, temp_paths=5.583649) W0417 13:44:36.005452 126746257050752 future.py:328] [process=0][thread=MainThread][operation_id=2] _SignalingThread.join() waiting for signals ([]) blocking the main thread will slow down blocking save times. This is likely due to main thread calling result() on a CommitFuture. I0417 13:44:36.728371 126638614382144 array_metadata_store.py:203] [process=0][thread=array_type_handler] Wrote 39 array_metadata.ArrayMetadata to gs://wanglance-maxtext/linen_ckpt_main_20260417_125240/linen_main_20260417_125240_14_async_ckpt_false/checkpoints/15/items/array_metadatas/process_0 I0417 13:45:15.501596 126638624867904 base_pytree_checkpoint_handler.py:1217] [process=0][thread=write_metadata_after_commits] Commit + Array metadata written. Time taken: 39.495927s (commit=39.051689s, array_metadata_write=0.444238s) I0417 13:45:15.503114 126746257050752 base_pytree_checkpoint_handler.py:128] [process=0] /jax/checkpoint/write/gbytes_per_sec: 283.736 MiB/s (total gbytes: 12.3 GiB) (time elapsed: 44 seconds) (per-host) I0417 13:45:16.240256 126746257050752 array_metadata_store.py:367] [process=0][thread=MainThread] Skipped cross-host ArrayMetadata validation because only one process is found: process_index=0. I0417 13:45:16.650616 126746257050752 ocdbt_utils.py:56] Param validation support for Zarr3 will be added later (b/362328389). I0417 13:45:16.650790 126746257050752 base_pytree_checkpoint_handler.py:1342] [process=0][thread=MainThread] Pytree save finalize (merge_ocdbt + ArrayMetadata validation) completed. Time taken: 0.545187s. use_zarr3=True, enable_post_merge_validation=True, directory=gs://wanglance-maxtext/linen_ckpt_main_20260417_125240/linen_main_20260417_125240_14_async_ckpt_false/checkpoints/15/items I0417 13:45:16.651444 126746257050752 atomicity.py:608] Finalizing gs://wanglance-maxtext/linen_ckpt_main_20260417_125240/linen_main_20260417_125240_14_async_ckpt_false/checkpoints/15/items I0417 13:45:16.887829 126746257050752 atomicity.py:608] Finalizing gs://wanglance-maxtext/linen_ckpt_main_20260417_125240/linen_main_20260417_125240_14_async_ckpt_false/checkpoints/15 I0417 13:45:17.578370 126746257050752 atomicity.py:794] [process=0][thread=MainThread] Finished saving checkpoint (finalized tmp dir) to `gs://wanglance-maxtext/linen_ckpt_main_20260417_125240/linen_main_20260417_125240_14_async_ckpt_false/checkpoints/15`. I0417 13:45:17.579017 126746257050752 checkpointer.py:290] Finished synchronous save in 48.14 seconds to gs://wanglance-maxtext/linen_ckpt_main_20260417_125240/linen_main_20260417_125240_14_async_ckpt_false/checkpoints/15 I0417 13:45:17.579212 126746257050752 checkpoint_manager.py:2103] [process=0][thread=MainThread][step=15] CheckpointManager Save Finalize is syncing with other hosts... I0417 13:45:17.579256 126746257050752 checkpoint_manager.py:2112] [process=0][thread=MainThread][step=15] CheckpointManager Save Finalize is done on all hosts. I0417 13:45:17.579280 126746257050752 checkpoint_manager.py:1570] [process=0][thread=MainThread][step=15] Finished synchronous save. I0417 13:45:17.579320 126746257050752 standard_logger.py:34] {'step': 15, 'event_type': 'save', 'directory': 'gs://wanglance-maxtext/linen_ckpt_main_20260417_125240/linen_main_20260417_125240_14_async_ckpt_false/checkpoints', 'reached_preemption': False, 'preemption_received_at': None, 'synchronous': True, 'wait_for_prev_start_time': 1776433469.4372993, 'wait_for_prev_duration_secs': 5.316734313964844e-05, 'time_between_consecutive_saves_sec': 0.7939858436584473, 'checkpointer_blocking_start_time': 1776433469.437491, 'checkpointer_blocking_duration_secs': 48.1415913105011, 'get_old_steps_start_time': 1776433517.5791144, 'get_old_steps_duration_secs': 3.4809112548828125e-05, 'checkpoint_manager_blocking_start_time': 1776433469.4372325, 'checkpoint_manager_blocking_duration_secs': 48.142062187194824} I0417 13:45:17.579415 126746257050752 checkpointing.py:410] Saved a checkpoint at step 15. I0417 13:45:17.580593 126746257050752 metric_logger.py:196] completed step: 14, seconds: 0.118, TFLOP/s/device: 114.675, Tokens/s/device: 17285.034, total_weights: 16384, loss: 2.083, lm_loss: 2.083, perplexity: 8.030 I0417 13:45:17.591884 126746257050752 metric_logger.py:196] completed step: 15, seconds: 0.118, TFLOP/s/device: 115.526, Tokens/s/device: 17413.337, total_weights: 16384, loss: 1.880, lm_loss: 1.880, perplexity: 6.551 I0417 13:45:17.705620 126746257050752 metric_logger.py:196] completed step: 16, seconds: 48.382, TFLOP/s/device: 0.281, Tokens/s/device: 42.330, total_weights: 16384, loss: 1.726, lm_loss: 1.726, perplexity: 5.616 I0417 13:45:17.823450 126746257050752 metric_logger.py:196] completed step: 17, seconds: 0.008, TFLOP/s/device: 1643.339, Tokens/s/device: 247701.984, total_weights: 16384, loss: 1.610, lm_loss: 1.610, perplexity: 5.002 I0417 13:45:17.941895 126746257050752 metric_logger.py:196] completed step: 18, seconds: 0.114, TFLOP/s/device: 118.684, Tokens/s/device: 17889.275, total_weights: 16384, loss: 1.521, lm_loss: 1.521, perplexity: 4.577 I0417 13:45:18.059822 126746257050752 checkpointing.py:772] Waiting for step 19 to finish before checkpoint... I0417 13:45:18.060642 126746257050752 checkpointing.py:776] Waited 0.0008444786071777344 seconds for step 19 to finish before starting checkpointing. I0417 13:45:18.060878 126746257050752 checkpoint_manager.py:1983] [process=0][thread=MainThread][wait_until_finished] No Save Finalize thread to wait for. Returning. I0417 13:45:18.061027 126746257050752 checkpoint_manager.py:1501] [process=0] Saving checkpoint at step 19 I0417 13:45:18.061100 126746257050752 checkpointer.py:241] [process=0] Started saving checkpoint to gs://wanglance-maxtext/linen_ckpt_main_20260417_125240/linen_main_20260417_125240_14_async_ckpt_false/checkpoints/19. I0417 13:45:18.233203 126746257050752 atomicity.py:137] Creating tmp directory gs://wanglance-maxtext/linen_ckpt_main_20260417_125240/linen_main_20260417_125240_14_async_ckpt_false/checkpoints/19 W0417 13:45:19.069488 126746257050752 future.py:328] [process=0][thread=MainThread][operation_id=3] _SignalingThread.join() waiting for signals ([]) blocking the main thread will slow down blocking save times. This is likely due to main thread calling result() on a CommitFuture. I0417 13:45:19.156005 126638991869504 atomicity.py:137] Creating tmp directory gs://wanglance-maxtext/linen_ckpt_main_20260417_125240/linen_main_20260417_125240_14_async_ckpt_false/checkpoints/19/items I0417 13:45:19.727112 126746257050752 jax_array_handlers.py:347] Scheduling D2H of 39 prioritized jax.Array. I0417 13:45:19.727217 126746257050752 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 I0417 13:45:21.080019 126746257050752 base_pytree_checkpoint_handler.py:153] [process=0][thread=MainThread] Initiated "orbax.checkpoint._src.serialization.jax_array_handlers.ArrayHandler".serialize. Time taken: 1.353265s I0417 13:45:21.080525 126746257050752 base_pytree_checkpoint_handler.py:128] [process=0] /jax/checkpoint/write/blocking_gbytes_per_sec: 8.537 GiB/s (total gbytes: 12.3 GiB) (time elapsed: a second) (per-host) I0417 13:45:21.080581 126746257050752 base_pytree_checkpoint_handler.py:732] [process=0][thread=MainThread] Initiated Pytree async_save. Time taken: 1.445637s (batch_requests_ready=0.088925s, total_serialization_initiated=1.356325s, others=0.000387s) I0417 13:45:21.080647 126746257050752 composite_checkpoint_handler.py:715] [process=0][thread=MainThread] Initiated CompositeCheckpointHandler.async_save. Time taken: 2.011476s (all_items=0.000029s, per_item={'items': '0.00002885'}, temp_paths=2.011447) W0417 13:45:21.080678 126746257050752 future.py:328] [process=0][thread=MainThread][operation_id=3] _SignalingThread.join() waiting for signals ([]) blocking the main thread will slow down blocking save times. This is likely due to main thread calling result() on a CommitFuture. I0417 13:45:21.809393 126638624867904 array_metadata_store.py:203] [process=0][thread=array_type_handler] Wrote 39 array_metadata.ArrayMetadata to gs://wanglance-maxtext/linen_ckpt_main_20260417_125240/linen_main_20260417_125240_14_async_ckpt_false/checkpoints/19/items/array_metadatas/process_0 I0417 13:45:54.031386 126638614382144 base_pytree_checkpoint_handler.py:1217] [process=0][thread=write_metadata_after_commits] Commit + Array metadata written. Time taken: 32.946947s (commit=32.437612s, array_metadata_write=0.509335s) I0417 13:45:54.032893 126746257050752 base_pytree_checkpoint_handler.py:128] [process=0] /jax/checkpoint/write/gbytes_per_sec: 367.371 MiB/s (total gbytes: 12.3 GiB) (time elapsed: 34 seconds) (per-host) I0417 13:45:54.861603 126746257050752 array_metadata_store.py:367] [process=0][thread=MainThread] Skipped cross-host ArrayMetadata validation because only one process is found: process_index=0. I0417 13:45:55.289161 126746257050752 ocdbt_utils.py:56] Param validation support for Zarr3 will be added later (b/362328389). I0417 13:45:55.289314 126746257050752 base_pytree_checkpoint_handler.py:1342] [process=0][thread=MainThread] Pytree save finalize (merge_ocdbt + ArrayMetadata validation) completed. Time taken: 0.572805s. use_zarr3=True, enable_post_merge_validation=True, directory=gs://wanglance-maxtext/linen_ckpt_main_20260417_125240/linen_main_20260417_125240_14_async_ckpt_false/checkpoints/19/items I0417 13:45:55.289953 126746257050752 atomicity.py:608] Finalizing gs://wanglance-maxtext/linen_ckpt_main_20260417_125240/linen_main_20260417_125240_14_async_ckpt_false/checkpoints/19/items I0417 13:45:55.557657 126746257050752 atomicity.py:608] Finalizing gs://wanglance-maxtext/linen_ckpt_main_20260417_125240/linen_main_20260417_125240_14_async_ckpt_false/checkpoints/19 I0417 13:45:56.260782 126746257050752 atomicity.py:794] [process=0][thread=MainThread] Finished saving checkpoint (finalized tmp dir) to `gs://wanglance-maxtext/linen_ckpt_main_20260417_125240/linen_main_20260417_125240_14_async_ckpt_false/checkpoints/19`. I0417 13:45:56.261454 126746257050752 checkpointer.py:290] Finished synchronous save in 38.20 seconds to gs://wanglance-maxtext/linen_ckpt_main_20260417_125240/linen_main_20260417_125240_14_async_ckpt_false/checkpoints/19 I0417 13:45:56.261641 126746257050752 checkpoint_manager.py:2103] [process=0][thread=MainThread][step=19] CheckpointManager Save Finalize is syncing with other hosts... I0417 13:45:56.261687 126746257050752 checkpoint_manager.py:2112] [process=0][thread=MainThread][step=19] CheckpointManager Save Finalize is done on all hosts. I0417 13:45:56.261711 126746257050752 checkpoint_manager.py:1570] [process=0][thread=MainThread][step=19] Finished synchronous save. I0417 13:45:56.261749 126746257050752 standard_logger.py:34] {'step': 19, 'event_type': 'save', 'directory': 'gs://wanglance-maxtext/linen_ckpt_main_20260417_125240/linen_main_20260417_125240_14_async_ckpt_false/checkpoints', 'reached_preemption': False, 'preemption_received_at': None, 'synchronous': True, 'wait_for_prev_start_time': 1776433518.0608597, 'wait_for_prev_duration_secs': 4.1961669921875e-05, 'time_between_consecutive_saves_sec': 0.48158860206604004, 'checkpointer_blocking_start_time': 1776433518.0610611, 'checkpointer_blocking_duration_secs': 38.200445890426636, 'get_old_steps_start_time': 1776433556.2615328, 'get_old_steps_duration_secs': 3.0517578125e-05, 'checkpoint_manager_blocking_start_time': 1776433518.0608282, 'checkpoint_manager_blocking_duration_secs': 38.200896978378296} I0417 13:45:56.261875 126746257050752 checkpointing.py:410] Saved a checkpoint at step 19. I0417 13:45:56.261904 126746257050752 checkpoint_manager.py:1983] [process=0][thread=MainThread][wait_until_finished] No Save Finalize thread to wait for. Returning. I0417 13:45:56.261928 126746257050752 checkpoint_manager.py:1983] [process=0][thread=MainThread][wait_until_finished] No Save Finalize thread to wait for. Returning. I0417 13:45:56.263690 126746257050752 metric_logger.py:196] completed step: 19, seconds: 0.118, TFLOP/s/device: 115.169, Tokens/s/device: 17359.463, total_weights: 16384, loss: 1.449, lm_loss: 1.449, perplexity: 4.260 Per train step: Total TFLOPs: 13.59 split as 93.93% learnable weight flops and 6.07% attention flops