2026-04-16 13:22:57.330251: 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 13:22:57.446214 127153807072384 max_utils.py:238] Skipping jax distributed system due to skip_jax_distributed_system=True flag. I0416 13:23:56.788343 127153807072384 max_utils.py:800] System Information: Jax Version: 0.9.2 I0416 13:23:56.788467 127153807072384 max_utils.py:801] System Information: Jaxlib Version: 0.9.2 I0416 13:23:56.788509 127153807072384 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 13:23:56.788533 127153807072384 train_utils.py:334] WARNING: 'dataset_path' might be pointing your local file system I0416 13:23:56.788554 127153807072384 train_utils.py:347] WARNING: Sequence packing is essentially ignored for synthetic data. Please use a real dataset to use sequence packing. I0416 13:23:56.788631 127153807072384 train.py:683] [DECOUPLED NO-OP] skipping cloud diagnostics wrapper. W0416 13:23:56.880533 3868421 pjrt_executable.cc:642] Assume version compatibility. PjRt-IFRT does not track XLA executable versions. I0416 13:23:57.290068 127153807072384 maxtext_utils.py:1517] Num_devices: 8, shape (1, 1, 1, 8, 1, 1, 1, 1, 1, 1, 1, 1, 1) I0416 13:23:57.290388 127153807072384 checkpointing.py:677] Setting up checkpoint logger... I0416 13:23:57.290431 127153807072384 checkpointing.py:233] Creating checkpoint manager with ocdbt=True and zarr3=True I0416 13:23:57.290509 127153807072384 pytree_checkpoint_handler.py:577] save_device_host_concurrent_bytes=None I0416 13:23:57.291247 127153807072384 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 0x73a4b69f8a70>, enable_pinned_host_transfer=False, save_concurrent_bytes: 96000000000 (89.4 GiB), restore_concurrent_bytes: 96000000000 (89.4 GiB) I0416 13:23:59.689010 127153807072384 checkpointing.py:265] Enabling policy for fixed interval checkpointing. I0416 13:23:59.689254 127153807072384 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 0x739e38fa24e0>}, handler_registry=None I0416 13:23:59.689600 127153807072384 composite_checkpoint_handler.py:237] Deferred registration for item: "items". Adding handler `<orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeCheckpointHandler object at 0x739e38fa24e0>` 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 13:23:59.689644 127153807072384 composite_checkpoint_handler.py:237] Deferred registration for item: "metrics". Adding handler `<orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonCheckpointHandler object at 0x739e37f3a1e0>` 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 13:23:59.689672 127153807072384 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 0x739e38fa24e0>, ('items', <class 'orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeRestoreArgs'>): <orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeCheckpointHandler object at 0x739e38fa24e0>, ('metrics', <class 'orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonSaveArgs'>): <orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonCheckpointHandler object at 0x739e37f3a1e0>, ('metrics', <class 'orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonRestoreArgs'>): <orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonCheckpointHandler object at 0x739e37f3a1e0>}). I0416 13:23:59.689977 127153807072384 abstract_checkpointer.py:35] orbax-checkpoint version: 0.11.28 I0416 13:23:59.803717 127153807072384 checkpoint_manager.py:1788] Found 0 checkpoint steps in gs://wanglance-maxtext/linen_ckpt_main_20260416_123638/linen_main_20260416_123638_14_async_ckpt_false/checkpoints I0416 13:23:59.803949 127153807072384 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_20260416_123638/linen_main_20260416_123638_14_async_ckpt_false/checkpoints: <orbax.checkpoint.checkpoint_manager.CheckpointManager object at 0x739e37f3a150> I0416 13:23:59.804039 127153807072384 checkpointing.py:301] Checkpoint manager created! W0416 13:23:59.820387 3868421 pjrt_executable.cc:642] Assume version compatibility. PjRt-IFRT does not track XLA executable versions. I0416 13:23:59.998858 127153807072384 nnx_wrappers.py:437] Unknown Logical: bfloat16[8,2048,2048]....................................... ('activation_batch', 'activation_norm_length', 'activation_embed'). I0416 13:23:59.998954 127153807072384 nnx_wrappers.py:437] Unknown Physical: bfloat16[8,2048,2048]....................................... ('fsdp', None, None). I0416 13:24:00.086566 127153807072384 attentions.py:1088] attentions/inputs_q Logical: bfloat16[8,2048,2048]....................................... ('activation_batch', 'activation_attn_length', 'activation_attn_embed'). I0416 13:24:00.086644 127153807072384 attentions.py:1088] attentions/inputs_q Physical: bfloat16[8,2048,2048]....................................... ('fsdp', None, None). I0416 13:24:00.100950 127153807072384 attentions.py:1089] attentions/inputs_kv Logical: bfloat16[8,2048,2048]....................................... ('activation_batch', 'activation_attn_length', 'activation_attn_embed'). I0416 13:24:00.100998 127153807072384 attentions.py:1089] attentions/inputs_kv Physical: bfloat16[8,2048,2048]....................................... ('fsdp', None, None). I0416 13:24:00.121564 127153807072384 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 13:24:00.121624 127153807072384 attentions.py:1154] attentions/query Physical: bfloat16[8,2048,16,128]..................................... ('fsdp', None, None, None). I0416 13:24:00.135920 127153807072384 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 13:24:00.135967 127153807072384 attentions.py:1155] attentions/key Physical: bfloat16[8,2048,16,128]..................................... ('fsdp', None, None, None). I0416 13:24:00.150154 127153807072384 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 13:24:00.150203 127153807072384 attentions.py:1156] attentions/value Physical: bfloat16[8,2048,16,128]..................................... ('fsdp', None, None, None). I0416 13:24:00.171812 127153807072384 attentions.py:1197] attentions/out Logical: bfloat16[8,2048,16,128]..................................... ('activation_batch', 'activation_attn_length', 'activation_heads', 'activation_kv'). I0416 13:24:00.171866 127153807072384 attentions.py:1197] attentions/out Physical: bfloat16[8,2048,16,128]..................................... ('fsdp', None, None, None). I0416 13:24:00.190513 127153807072384 linears.py:525] linears/x Logical: bfloat16[8,2048,7168]....................................... ('activation_batch', 'activation_length', 'activation_mlp'). I0416 13:24:00.190563 127153807072384 linears.py:525] linears/x Physical: bfloat16[8,2048,7168]....................................... ('fsdp', None, None). I0416 13:24:00.372725 127153807072384 checkpointing.py:577] checkpoint manager exists so trying to load this run's existing checkpoint I0416 13:24:00.372819 127153807072384 checkpointing.py:665] No existing checkpoints found, not restoring checkpoint. W0416 13:24:00.422916 3868421 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 13:24:00.803020 127153807072384 maxtext_utils.py:1620] params/params/decoder/decoder_norm/scale Shape: float32[2048] Logical: P('norm',) Physical: (None,) I0416 13:24:00.803183 127153807072384 maxtext_utils.py:1620] params/params/decoder/layers/mlp/wi_0/kernel Shape: float32[2048,16,7168] Logical: P('embed', 'layers', 'mlp') Physical: ('fsdp', None, None) I0416 13:24:00.803225 127153807072384 maxtext_utils.py:1620] params/params/decoder/layers/mlp/wi_1/kernel Shape: float32[2048,16,7168] Logical: P('embed', 'layers', 'mlp') Physical: ('fsdp', None, None) I0416 13:24:00.803268 127153807072384 maxtext_utils.py:1620] params/params/decoder/layers/mlp/wo/kernel Shape: float32[7168,16,2048] Logical: P('mlp', 'layers', 'embed') Physical: (None, None, 'fsdp') I0416 13:24:00.803307 127153807072384 maxtext_utils.py:1620] params/params/decoder/layers/post_self_attention_layer_norm/scale Shape: float32[2048,16] Logical: P('norm', 'layers') Physical: (None, None) I0416 13:24:00.803341 127153807072384 maxtext_utils.py:1620] params/params/decoder/layers/pre_self_attention_layer_norm/scale Shape: float32[2048,16] Logical: P('norm', 'layers') Physical: (None, None) I0416 13:24:00.803381 127153807072384 maxtext_utils.py:1620] 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 13:24:00.803419 127153807072384 maxtext_utils.py:1620] 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 13:24:00.803447 127153807072384 maxtext_utils.py:1620] 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 13:24:00.803472 127153807072384 maxtext_utils.py:1620] 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 13:24:00.803505 127153807072384 maxtext_utils.py:1620] params/params/decoder/logits_dense/kernel Shape: float32[2048,32000] Logical: P('embed_vocab', 'vocab') Physical: ('fsdp', None) I0416 13:24:00.803539 127153807072384 maxtext_utils.py:1620] params/params/token_embedder/embedding Shape: float32[32000,2048] Logical: P('vocab', 'embed_vocab') Physical: (None, 'fsdp') I0416 13:24:01.224180 127153807072384 train.py:155] train/xent Logical: float32[8,2048]............................................. ('activation_embed_and_logits_batch', 'activation_length'). I0416 13:24:01.224260 127153807072384 train.py:155] train/xent Physical: float32[8,2048]............................................. ('fsdp', None). I0416 13:24:01.237786 127153807072384 train.py:162] train/z_loss Logical: float32[8,2048]............................................. ('activation_embed_and_logits_batch', 'activation_length'). I0416 13:24:01.237837 127153807072384 train.py:162] train/z_loss Physical: float32[8,2048]............................................. ('fsdp', None). W0416 13:24:01.861128 3868421 pjrt_executable.cc:642] Assume version compatibility. PjRt-IFRT does not track XLA executable versions. I0416 13:24:01.985972 127153807072384 max_utils.py:791] Total memory size: 3.6 GB, Output size: 1.5 GB, Temp size: 2.0 GB, Argument size: 1.5 GB, Host temp size: 0.0 GB. I0416 13:24:01.986428 127153807072384 max_utils.py:194] tensorboardX not available; using no-op SummaryWriter. I0416 13:24:01.986565 127153807072384 metric_logger.py:289] number parameters: 1.104 billion W0416 13:24:02.815372 3868421 pjrt_executable.cc:642] Assume version compatibility. PjRt-IFRT does not track XLA executable versions. I0416 13:24:02.910904 127153807072384 checkpointing.py:772] Waiting for step 0 to finish before checkpoint... I0416 13:24:03.028208 127153807072384 checkpointing.py:776] Waited 0.11728739738464355 seconds for step 0 to finish before starting checkpointing. I0416 13:24:03.028630 127153807072384 checkpoint_manager.py:1983] [process=0][thread=MainThread][wait_until_finished] No Save Finalize thread to wait for. Returning. I0416 13:24:03.028775 127153807072384 checkpoint_manager.py:1501] [process=0] Saving checkpoint at step 0 I0416 13:24:03.028831 127153807072384 checkpointer.py:241] [process=0] Started saving checkpoint to gs://wanglance-maxtext/linen_ckpt_main_20260416_123638/linen_main_20260416_123638_14_async_ckpt_false/checkpoints/0. I0416 13:24:03.178260 127153807072384 atomicity.py:137] Creating tmp directory gs://wanglance-maxtext/linen_ckpt_main_20260416_123638/linen_main_20260416_123638_14_async_ckpt_false/checkpoints/0 I0416 13:24:03.959722 127153807072384 checkpoint.py:188] Wrote Metadata={'item_handlers': None, 'metrics': {}, 'performance_metrics': {}, 'init_timestamp_nsecs': 1776345843723741027, 'commit_timestamp_nsecs': None, 'custom_metadata': {}}, json={"item_handlers": null, "metrics": {}, "performance_metrics": {}, "init_timestamp_nsecs": 1776345843723741027, "commit_timestamp_nsecs": null, "custom_metadata": {}} to gs://wanglance-maxtext/linen_ckpt_main_20260416_123638/linen_main_20260416_123638_14_async_ckpt_false/checkpoints/0/_CHECKPOINT_METADATA I0416 13:24:03.960727 127153807072384 signaling_client.py:373] Using ThreadSafeKeyValueSignalingClient W0416 13:24:03.961035 127153807072384 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. I0416 13:24:04.051595 127046298633792 atomicity.py:137] Creating tmp directory gs://wanglance-maxtext/linen_ckpt_main_20260416_123638/linen_main_20260416_123638_14_async_ckpt_false/checkpoints/0/items I0416 13:24:04.598878 127153807072384 jax_array_handlers.py:347] Scheduling D2H of 39 prioritized jax.Array. I0416 13:24:04.598982 127153807072384 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 13:24:04.616463 3870320 google_auth_provider.cc:149] Using credentials at ~/.config/gcloud/application_default_credentials.json I0416 13:24:04.616509 3870320 google_auth_provider.cc:156] Using OAuth2 AuthProvider W0416 13:24:04.826831 3868421 pjrt_executable.cc:642] Assume version compatibility. PjRt-IFRT does not track XLA executable versions. W0416 13:24:04.839036 3868421 pjrt_executable.cc:642] Assume version compatibility. PjRt-IFRT does not track XLA executable versions. W0416 13:24:04.849443 3868421 pjrt_executable.cc:642] Assume version compatibility. PjRt-IFRT does not track XLA executable versions. W0416 13:24:04.868169 3868421 pjrt_executable.cc:642] Assume version compatibility. PjRt-IFRT does not track XLA executable versions. W0416 13:24:04.878629 3868421 pjrt_executable.cc:642] Assume version compatibility. PjRt-IFRT does not track XLA executable versions. W0416 13:24:04.884590 3868421 pjrt_executable.cc:642] Assume version compatibility. PjRt-IFRT does not track XLA executable versions. W0416 13:24:04.888697 3868421 pjrt_executable.cc:642] Assume version compatibility. PjRt-IFRT does not track XLA executable versions. W0416 13:24:04.892732 3868421 pjrt_executable.cc:642] Assume version compatibility. PjRt-IFRT does not track XLA executable versions. I0416 13:24:05.761522 127153807072384 base_pytree_checkpoint_handler.py:153] [process=0][thread=MainThread] Initiated "orbax.checkpoint._src.serialization.jax_array_handlers.ArrayHandler".serialize. Time taken: 1.163000s I0416 13:24:05.762028 127153807072384 base_pytree_checkpoint_handler.py:128] [process=0] /jax/checkpoint/write/blocking_gbytes_per_sec: 9.839 GiB/s (total gbytes: 12.3 GiB) (time elapsed: a second) (per-host) I0416 13:24:05.762094 127153807072384 base_pytree_checkpoint_handler.py:732] [process=0][thread=MainThread] Initiated Pytree async_save. Time taken: 1.254307s (batch_requests_ready=0.087096s, total_serialization_initiated=1.166801s, others=0.000410s) I0416 13:24:05.762157 127153807072384 composite_checkpoint_handler.py:715] [process=0][thread=MainThread] Initiated CompositeCheckpointHandler.async_save. Time taken: 1.801492s (all_items=0.000031s, per_item={'items': '0.00003147'}, temp_paths=1.801461) W0416 13:24:05.762193 127153807072384 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. I0416 13:24:06.515121 127046309119552 array_metadata_store.py:203] [process=0][thread=array_type_handler] Wrote 39 array_metadata.ArrayMetadata to gs://wanglance-maxtext/linen_ckpt_main_20260416_123638/linen_main_20260416_123638_14_async_ckpt_false/checkpoints/0/items/array_metadatas/process_0 I0416 13:24:33.422189 127056360769088 base_pytree_checkpoint_handler.py:1217] [process=0][thread=write_metadata_after_commits] Commit + Array metadata written. Time taken: 27.655866s (commit=27.236061s, array_metadata_write=0.419805s) I0416 13:24:33.423200 127153807072384 base_pytree_checkpoint_handler.py:128] [process=0] /jax/checkpoint/write/gbytes_per_sec: 437.027 MiB/s (total gbytes: 12.3 GiB) (time elapsed: 28 seconds) (per-host) I0416 13:24:33.651660 127153807072384 checkpoint.py:228] Read Metadata={'item_handlers': None, 'metrics': {}, 'performance_metrics': {}, 'init_timestamp_nsecs': 1776345843723741027, 'commit_timestamp_nsecs': None, 'custom_metadata': {}} from gs://wanglance-maxtext/linen_ckpt_main_20260416_123638/linen_main_20260416_123638_14_async_ckpt_false/checkpoints/0/_CHECKPOINT_METADATA I0416 13:24:34.006887 127153807072384 checkpoint.py:247] Updated Metadata={'item_handlers': {'items': 'orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeCheckpointHandler'}, 'metrics': {}, 'performance_metrics': {}, 'init_timestamp_nsecs': 1776345843723741027, 'commit_timestamp_nsecs': None, 'custom_metadata': {}} to gs://wanglance-maxtext/linen_ckpt_main_20260416_123638/linen_main_20260416_123638_14_async_ckpt_false/checkpoints/0/_CHECKPOINT_METADATA I0416 13:24:34.148471 127153807072384 array_metadata_store.py:367] [process=0][thread=MainThread] Skipped cross-host ArrayMetadata validation because only one process is found: process_index=0. I0416 13:24:34.576340 127153807072384 ocdbt_utils.py:56] Param validation support for Zarr3 will be added later (b/362328389). I0416 13:24:34.576503 127153807072384 base_pytree_checkpoint_handler.py:1342] [process=0][thread=MainThread] Pytree save finalize (merge_ocdbt + ArrayMetadata validation) completed. Time taken: 0.569223s. use_zarr3=True, enable_post_merge_validation=True, directory=gs://wanglance-maxtext/linen_ckpt_main_20260416_123638/linen_main_20260416_123638_14_async_ckpt_false/checkpoints/0/items I0416 13:24:34.577172 127153807072384 atomicity.py:608] Finalizing gs://wanglance-maxtext/linen_ckpt_main_20260416_123638/linen_main_20260416_123638_14_async_ckpt_false/checkpoints/0/items I0416 13:24:34.832013 127153807072384 atomicity.py:608] Finalizing gs://wanglance-maxtext/linen_ckpt_main_20260416_123638/linen_main_20260416_123638_14_async_ckpt_false/checkpoints/0 I0416 13:24:35.510035 127153807072384 atomicity.py:794] [process=0][thread=MainThread] Finished saving checkpoint (finalized tmp dir) to `gs://wanglance-maxtext/linen_ckpt_main_20260416_123638/linen_main_20260416_123638_14_async_ckpt_false/checkpoints/0`. I0416 13:24:35.510700 127153807072384 checkpointer.py:290] Finished synchronous save in 32.48 seconds to gs://wanglance-maxtext/linen_ckpt_main_20260416_123638/linen_main_20260416_123638_14_async_ckpt_false/checkpoints/0 I0416 13:24:35.510879 127153807072384 checkpoint_manager.py:2103] [process=0][thread=MainThread][step=0] CheckpointManager Save Finalize is syncing with other hosts... I0416 13:24:35.510924 127153807072384 checkpoint_manager.py:2112] [process=0][thread=MainThread][step=0] CheckpointManager Save Finalize is done on all hosts. I0416 13:24:35.510947 127153807072384 checkpoint_manager.py:1570] [process=0][thread=MainThread][step=0] Finished synchronous save. I0416 13:24:35.510984 127153807072384 standard_logger.py:34] {'step': 0, 'event_type': 'save', 'directory': 'gs://wanglance-maxtext/linen_ckpt_main_20260416_123638/linen_main_20260416_123638_14_async_ckpt_false/checkpoints', 'reached_preemption': False, 'preemption_received_at': None, 'synchronous': True, 'wait_for_prev_start_time': 1776345843.0286167, 'wait_for_prev_duration_secs': 4.00543212890625e-05, 'time_between_consecutive_saves_sec': None, 'checkpointer_blocking_start_time': 1776345843.0287957, 'checkpointer_blocking_duration_secs': 32.481948137283325, 'get_old_steps_start_time': 1776345875.5107806, 'get_old_steps_duration_secs': 2.765655517578125e-05, 'checkpoint_manager_blocking_start_time': 1776345843.028532, 'checkpoint_manager_blocking_duration_secs': 32.482428789138794} I0416 13:24:35.511101 127153807072384 checkpointing.py:410] Saved a checkpoint at step 0. I0416 13:24:35.511152 127153807072384 max_utils.py:750] Memstats: After params initialized: I0416 13:24:35.511204 127153807072384 max_utils.py:756] Using (GB) 1.59 / 31.25 (5.088000%) on TPU_0(process=0,(0,0,0,0)) I0416 13:24:35.511230 127153807072384 max_utils.py:756] Using (GB) 1.59 / 31.25 (5.088000%) on TPU_1(process=0,(1,0,0,0)) I0416 13:24:35.511251 127153807072384 max_utils.py:756] Using (GB) 1.59 / 31.25 (5.088000%) on TPU_2(process=0,(0,1,0,0)) I0416 13:24:35.511271 127153807072384 max_utils.py:756] Using (GB) 1.59 / 31.25 (5.088000%) on TPU_3(process=0,(1,1,0,0)) I0416 13:24:35.511288 127153807072384 max_utils.py:756] Using (GB) 1.59 / 31.25 (5.088000%) on TPU_4(process=0,(0,2,0,0)) I0416 13:24:35.511307 127153807072384 max_utils.py:756] Using (GB) 1.59 / 31.25 (5.088000%) on TPU_5(process=0,(1,2,0,0)) I0416 13:24:35.511324 127153807072384 max_utils.py:756] Using (GB) 1.59 / 31.25 (5.088000%) on TPU_6(process=0,(0,3,0,0)) I0416 13:24:35.511340 127153807072384 max_utils.py:756] Using (GB) 1.59 / 31.25 (5.088000%) on TPU_7(process=0,(1,3,0,0)) W0416 13:24:35.517501 3868421 pjrt_executable.cc:642] Assume version compatibility. PjRt-IFRT does not track XLA executable versions. W0416 13:24:35.523295 3868421 pjrt_executable.cc:642] Assume version compatibility. PjRt-IFRT does not track XLA executable versions. W0416 13:24:35.527778 3868421 pjrt_executable.cc:642] Assume version compatibility. PjRt-IFRT does not track XLA executable versions. I0416 13:24:35.722320 127153807072384 metric_logger.py:185] completed step: 0, seconds: 0.924, TFLOP/s/device: 14.707, Tokens/s/device: 2216.743, total_weights: 16384, loss: 10.874 I0416 13:24:35.723079 127153807072384 metric_logger.py:269] To see full metrics 'tensorboard --logdir=gs://wanglance-maxtext/linen_ckpt_main_20260416_123638/linen_main_20260416_123638_14_async_ckpt_false/tensorboard/' I0416 13:24:35.836940 127153807072384 metric_logger.py:185] completed step: 1, seconds: 32.809, TFLOP/s/device: 0.414, Tokens/s/device: 62.423, total_weights: 16384, loss: 10.874 I0416 13:24:35.956288 127153807072384 metric_logger.py:185] completed step: 2, seconds: 0.006, TFLOP/s/device: 2123.321, Tokens/s/device: 320050.008, total_weights: 16384, loss: 10.276 I0416 13:24:36.074983 127153807072384 metric_logger.py:185] completed step: 3, seconds: 0.115, TFLOP/s/device: 117.681, Tokens/s/device: 17738.206, total_weights: 16384, loss: 9.282 I0416 13:24:36.079472 127153807072384 checkpointing.py:772] Waiting for step 5 to finish before checkpoint... I0416 13:24:36.310849 127153807072384 checkpointing.py:776] Waited 0.23135685920715332 seconds for step 5 to finish before starting checkpointing. I0416 13:24:36.311253 127153807072384 checkpoint_manager.py:1983] [process=0][thread=MainThread][wait_until_finished] No Save Finalize thread to wait for. Returning. I0416 13:24:36.311403 127153807072384 checkpoint_manager.py:1501] [process=0] Saving checkpoint at step 5 I0416 13:24:36.311460 127153807072384 checkpointer.py:241] [process=0] Started saving checkpoint to gs://wanglance-maxtext/linen_ckpt_main_20260416_123638/linen_main_20260416_123638_14_async_ckpt_false/checkpoints/5. I0416 13:24:36.496950 127153807072384 atomicity.py:137] Creating tmp directory gs://wanglance-maxtext/linen_ckpt_main_20260416_123638/linen_main_20260416_123638_14_async_ckpt_false/checkpoints/5 W0416 13:24:37.330103 127153807072384 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. I0416 13:24:37.425186 127056360769088 atomicity.py:137] Creating tmp directory gs://wanglance-maxtext/linen_ckpt_main_20260416_123638/linen_main_20260416_123638_14_async_ckpt_false/checkpoints/5/items I0416 13:24:37.975989 127153807072384 jax_array_handlers.py:347] Scheduling D2H of 39 prioritized jax.Array. I0416 13:24:37.976110 127153807072384 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 13:24:38.235134 127153807072384 base_pytree_checkpoint_handler.py:153] [process=0][thread=MainThread] Initiated "orbax.checkpoint._src.serialization.jax_array_handlers.ArrayHandler".serialize. Time taken: 0.259512s I0416 13:24:38.239541 127153807072384 base_pytree_checkpoint_handler.py:128] [process=0] /jax/checkpoint/write/blocking_gbytes_per_sec: 35.070 GiB/s (total gbytes: 12.3 GiB) (time elapsed: 351 milliseconds) (per-host) I0416 13:24:38.239599 127153807072384 base_pytree_checkpoint_handler.py:732] [process=0][thread=MainThread] Initiated Pytree async_save. Time taken: 0.351989s (batch_requests_ready=0.082293s, total_serialization_initiated=0.265398s, others=0.004298s) I0416 13:24:38.239667 127153807072384 composite_checkpoint_handler.py:715] [process=0][thread=MainThread] Initiated CompositeCheckpointHandler.async_save. Time taken: 0.909906s (all_items=0.000033s, per_item={'items': '0.00003338'}, temp_paths=0.909873) W0416 13:24:38.239699 127153807072384 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. I0416 13:24:38.971905 127046430754368 array_metadata_store.py:203] [process=0][thread=array_type_handler] Wrote 39 array_metadata.ArrayMetadata to gs://wanglance-maxtext/linen_ckpt_main_20260416_123638/linen_main_20260416_123638_14_async_ckpt_false/checkpoints/5/items/array_metadatas/process_0 I0416 13:25:20.216299 127046309119552 base_pytree_checkpoint_handler.py:1217] [process=0][thread=write_metadata_after_commits] Commit + Array metadata written. Time taken: 41.976381s (commit=41.486342s, array_metadata_write=0.490039s) I0416 13:25:20.217420 127153807072384 base_pytree_checkpoint_handler.py:128] [process=0] /jax/checkpoint/write/gbytes_per_sec: 298.532 MiB/s (total gbytes: 12.3 GiB) (time elapsed: 42 seconds) (per-host) I0416 13:25:20.951744 127153807072384 array_metadata_store.py:367] [process=0][thread=MainThread] Skipped cross-host ArrayMetadata validation because only one process is found: process_index=0. I0416 13:25:21.385235 127153807072384 ocdbt_utils.py:56] Param validation support for Zarr3 will be added later (b/362328389). I0416 13:25:21.385384 127153807072384 base_pytree_checkpoint_handler.py:1342] [process=0][thread=MainThread] Pytree save finalize (merge_ocdbt + ArrayMetadata validation) completed. Time taken: 0.566958s. use_zarr3=True, enable_post_merge_validation=True, directory=gs://wanglance-maxtext/linen_ckpt_main_20260416_123638/linen_main_20260416_123638_14_async_ckpt_false/checkpoints/5/items I0416 13:25:21.386027 127153807072384 atomicity.py:608] Finalizing gs://wanglance-maxtext/linen_ckpt_main_20260416_123638/linen_main_20260416_123638_14_async_ckpt_false/checkpoints/5/items I0416 13:25:21.616001 127153807072384 atomicity.py:608] Finalizing gs://wanglance-maxtext/linen_ckpt_main_20260416_123638/linen_main_20260416_123638_14_async_ckpt_false/checkpoints/5 I0416 13:25:22.275466 127153807072384 atomicity.py:794] [process=0][thread=MainThread] Finished saving checkpoint (finalized tmp dir) to `gs://wanglance-maxtext/linen_ckpt_main_20260416_123638/linen_main_20260416_123638_14_async_ckpt_false/checkpoints/5`. I0416 13:25:22.276148 127153807072384 checkpointer.py:290] Finished synchronous save in 45.96 seconds to gs://wanglance-maxtext/linen_ckpt_main_20260416_123638/linen_main_20260416_123638_14_async_ckpt_false/checkpoints/5 I0416 13:25:22.276346 127153807072384 checkpoint_manager.py:2103] [process=0][thread=MainThread][step=5] CheckpointManager Save Finalize is syncing with other hosts... I0416 13:25:22.276389 127153807072384 checkpoint_manager.py:2112] [process=0][thread=MainThread][step=5] CheckpointManager Save Finalize is done on all hosts. I0416 13:25:22.276413 127153807072384 checkpoint_manager.py:1570] [process=0][thread=MainThread][step=5] Finished synchronous save. I0416 13:25:22.276451 127153807072384 standard_logger.py:34] {'step': 5, 'event_type': 'save', 'directory': 'gs://wanglance-maxtext/linen_ckpt_main_20260416_123638/linen_main_20260416_123638_14_async_ckpt_false/checkpoints', 'reached_preemption': False, 'preemption_received_at': None, 'synchronous': True, 'wait_for_prev_start_time': 1776345876.3112319, 'wait_for_prev_duration_secs': 4.76837158203125e-05, 'time_between_consecutive_saves_sec': 0.8002934455871582, 'checkpointer_blocking_start_time': 1776345876.3114238, 'checkpointer_blocking_duration_secs': 45.964778900146484, 'get_old_steps_start_time': 1776345922.276233, 'get_old_steps_duration_secs': 3.6716461181640625e-05, 'checkpoint_manager_blocking_start_time': 1776345876.311165, 'checkpoint_manager_blocking_duration_secs': 45.96526122093201} I0416 13:25:22.276582 127153807072384 checkpointing.py:410] Saved a checkpoint at step 5. I0416 13:25:22.277275 127153807072384 metric_logger.py:185] completed step: 4, seconds: 0.119, TFLOP/s/device: 113.768, Tokens/s/device: 17148.407, total_weights: 16384, loss: 8.327 I0416 13:25:22.288121 127153807072384 metric_logger.py:185] completed step: 5, seconds: 0.119, TFLOP/s/device: 114.586, Tokens/s/device: 17271.623, total_weights: 16384, loss: 7.367 I0416 13:25:22.400947 127153807072384 metric_logger.py:185] completed step: 6, seconds: 46.204, TFLOP/s/device: 0.294, Tokens/s/device: 44.325, total_weights: 16384, loss: 6.459 I0416 13:25:22.519117 127153807072384 metric_logger.py:185] completed step: 7, seconds: 0.008, TFLOP/s/device: 1611.377, Tokens/s/device: 242884.250, total_weights: 16384, loss: 5.622 I0416 13:25:22.637061 127153807072384 metric_logger.py:185] completed step: 8, seconds: 0.114, TFLOP/s/device: 119.422, Tokens/s/device: 18000.598, total_weights: 16384, loss: 4.864 I0416 13:25:22.755016 127153807072384 checkpointing.py:772] Waiting for step 9 to finish before checkpoint... I0416 13:25:22.755850 127153807072384 checkpointing.py:776] Waited 0.0008540153503417969 seconds for step 9 to finish before starting checkpointing. I0416 13:25:22.756099 127153807072384 checkpoint_manager.py:1983] [process=0][thread=MainThread][wait_until_finished] No Save Finalize thread to wait for. Returning. I0416 13:25:22.756242 127153807072384 checkpoint_manager.py:1501] [process=0] Saving checkpoint at step 9 I0416 13:25:22.756294 127153807072384 checkpointer.py:241] [process=0] Started saving checkpoint to gs://wanglance-maxtext/linen_ckpt_main_20260416_123638/linen_main_20260416_123638_14_async_ckpt_false/checkpoints/9. I0416 13:25:22.931284 127153807072384 atomicity.py:137] Creating tmp directory gs://wanglance-maxtext/linen_ckpt_main_20260416_123638/linen_main_20260416_123638_14_async_ckpt_false/checkpoints/9 W0416 13:25:23.759528 127153807072384 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. I0416 13:25:23.843432 127056360769088 atomicity.py:137] Creating tmp directory gs://wanglance-maxtext/linen_ckpt_main_20260416_123638/linen_main_20260416_123638_14_async_ckpt_false/checkpoints/9/items I0416 13:25:24.401885 127153807072384 jax_array_handlers.py:347] Scheduling D2H of 39 prioritized jax.Array. I0416 13:25:24.401991 127153807072384 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 13:25:26.543694 127153807072384 base_pytree_checkpoint_handler.py:153] [process=0][thread=MainThread] Initiated "orbax.checkpoint._src.serialization.jax_array_handlers.ArrayHandler".serialize. Time taken: 2.142156s I0416 13:25:26.547969 127153807072384 base_pytree_checkpoint_handler.py:128] [process=0] /jax/checkpoint/write/blocking_gbytes_per_sec: 5.513 GiB/s (total gbytes: 12.3 GiB) (time elapsed: 2 seconds) (per-host) I0416 13:25:26.548031 127153807072384 base_pytree_checkpoint_handler.py:732] [process=0][thread=MainThread] Initiated Pytree async_save. Time taken: 2.238614s (batch_requests_ready=0.089232s, total_serialization_initiated=2.145221s, others=0.004161s) I0416 13:25:26.548112 127153807072384 composite_checkpoint_handler.py:715] [process=0][thread=MainThread] Initiated CompositeCheckpointHandler.async_save. Time taken: 2.788903s (all_items=0.000026s, per_item={'items': '0.00002599'}, temp_paths=2.788877) W0416 13:25:26.548142 127153807072384 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. I0416 13:25:27.279084 127046309119552 array_metadata_store.py:203] [process=0][thread=array_type_handler] Wrote 39 array_metadata.ArrayMetadata to gs://wanglance-maxtext/linen_ckpt_main_20260416_123638/linen_main_20260416_123638_14_async_ckpt_false/checkpoints/9/items/array_metadatas/process_0 I0416 13:25:59.586640 127046430754368 base_pytree_checkpoint_handler.py:1217] [process=0][thread=write_metadata_after_commits] Commit + Array metadata written. Time taken: 33.038130s (commit=32.615673s, array_metadata_write=0.422457s) I0416 13:25:59.587763 127153807072384 base_pytree_checkpoint_handler.py:128] [process=0] /jax/checkpoint/write/gbytes_per_sec: 358.203 MiB/s (total gbytes: 12.3 GiB) (time elapsed: 35 seconds) (per-host) I0416 13:26:00.317322 127153807072384 array_metadata_store.py:367] [process=0][thread=MainThread] Skipped cross-host ArrayMetadata validation because only one process is found: process_index=0. I0416 13:26:00.745197 127153807072384 ocdbt_utils.py:56] Param validation support for Zarr3 will be added later (b/362328389). I0416 13:26:00.745360 127153807072384 base_pytree_checkpoint_handler.py:1342] [process=0][thread=MainThread] Pytree save finalize (merge_ocdbt + ArrayMetadata validation) completed. Time taken: 0.571174s. use_zarr3=True, enable_post_merge_validation=True, directory=gs://wanglance-maxtext/linen_ckpt_main_20260416_123638/linen_main_20260416_123638_14_async_ckpt_false/checkpoints/9/items I0416 13:26:00.746019 127153807072384 atomicity.py:608] Finalizing gs://wanglance-maxtext/linen_ckpt_main_20260416_123638/linen_main_20260416_123638_14_async_ckpt_false/checkpoints/9/items I0416 13:26:00.986398 127153807072384 atomicity.py:608] Finalizing gs://wanglance-maxtext/linen_ckpt_main_20260416_123638/linen_main_20260416_123638_14_async_ckpt_false/checkpoints/9 I0416 13:26:01.642564 127153807072384 atomicity.py:794] [process=0][thread=MainThread] Finished saving checkpoint (finalized tmp dir) to `gs://wanglance-maxtext/linen_ckpt_main_20260416_123638/linen_main_20260416_123638_14_async_ckpt_false/checkpoints/9`. I0416 13:26:01.643221 127153807072384 checkpointer.py:290] Finished synchronous save in 38.89 seconds to gs://wanglance-maxtext/linen_ckpt_main_20260416_123638/linen_main_20260416_123638_14_async_ckpt_false/checkpoints/9 I0416 13:26:01.643416 127153807072384 checkpoint_manager.py:2103] [process=0][thread=MainThread][step=9] CheckpointManager Save Finalize is syncing with other hosts... I0416 13:26:01.643465 127153807072384 checkpoint_manager.py:2112] [process=0][thread=MainThread][step=9] CheckpointManager Save Finalize is done on all hosts. I0416 13:26:01.643489 127153807072384 checkpoint_manager.py:1570] [process=0][thread=MainThread][step=9] Finished synchronous save. I0416 13:26:01.643521 127153807072384 standard_logger.py:34] {'step': 9, 'event_type': 'save', 'directory': 'gs://wanglance-maxtext/linen_ckpt_main_20260416_123638/linen_main_20260416_123638_14_async_ckpt_false/checkpoints', 'reached_preemption': False, 'preemption_received_at': None, 'synchronous': True, 'wait_for_prev_start_time': 1776345922.7560778, 'wait_for_prev_duration_secs': 4.553794860839844e-05, 'time_between_consecutive_saves_sec': 0.4796748161315918, 'checkpointer_blocking_start_time': 1776345922.7562616, 'checkpointer_blocking_duration_secs': 38.887004137039185, 'get_old_steps_start_time': 1776345961.643298, 'get_old_steps_duration_secs': 5.054473876953125e-05, 'checkpoint_manager_blocking_start_time': 1776345922.756034, 'checkpoint_manager_blocking_duration_secs': 38.887468099594116} I0416 13:26:01.643600 127153807072384 checkpointing.py:410] Saved a checkpoint at step 9. I0416 13:26:01.643632 127153807072384 checkpoint_manager.py:1983] [process=0][thread=MainThread][wait_until_finished] No Save Finalize thread to wait for. Returning. I0416 13:26:01.643650 127153807072384 checkpoint_manager.py:1983] [process=0][thread=MainThread][wait_until_finished] No Save Finalize thread to wait for. Returning. I0416 13:26:01.644372 127153807072384 metric_logger.py:185] completed step: 9, seconds: 0.118, TFLOP/s/device: 114.898, Tokens/s/device: 17318.653, total_weights: 16384, loss: 4.189 Per train step: Total TFLOPs: 13.59 split as 93.93% learnable weight flops and 6.07% attention flops