main (574ad3fb9) vs test/pipeline-scan-nnx (574ad3fb9)| Metric | main 574ad3fb9 | test/pipeline-scan-nnx 574ad3fb9 | Diff (test/pipeline-scan-nnx − main) |
|---|---|---|---|
| Parameters | 1.104 billion | 1.104 billion | — |
| Final loss | 8.1790 | 8.1790 | 0 |
| TFLOP/s | 91.796 | 92.225 | +0.429 |
| Tok/s | 13836.4 | 13901.1 | +64.709 |
| Avg s/step | 3.081 | 2.441 | -0.64 |
| Memory % | 1.38 | 1.38 | 0 |
| JAX | 0.9.2 | 0.8.1 | — |
Diff = branch value − main value. Green = branch improved. Red = branch regressed.
XPK Start: Sat Apr 18 18:05:37 UTC 2026 PyTorch was not found. Models won't be available and only tokenizers, configuration and file/data utilities can be used. 2026-04-18 18:06:01.657639: 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) I0418 18:06:02.191496 139213532985152 max_utils.py:273] Attempting to initialize the jax distributed system... I0418 18:06:11.232153 139213532985152 distributed.py:149] Starting JAX distributed service on [::]:8482 I0418 18:06:11.234575 139213532985152 distributed.py:172] Connecting to JAX distributed service on mt-02-synthetic-1t9y5-slice-job-0-0.mt-02-synthetic-1t9y5:8482 I0418 18:06:12.509524 139213532985152 max_utils.py:284] Jax distributed system initialized! I0418 18:06:18.514478 139213532985152 max_utils.py:800] System Information: Jax Version: 0.9.2 I0418 18:06:18.514583 139213532985152 max_utils.py:801] System Information: Jaxlib Version: 0.9.2 I0418 18:06:18.514624 139213532985152 max_utils.py:802] System Information: Jax Backend: PJRT C API TFRT TPU v6 lite Built on Mar 4 2026 11:32:08 (1772652728) cl/878335365 I0418 18:06:18.514659 139213532985152 train_utils.py:348] WARNING: Sequence packing is essentially ignored for synthetic data. Please use a real dataset to use sequence packing. I0418 18:06:19.215611 139213532985152 maxtext_utils.py:1551] Num_devices: 32, shape (1, 1, 1, 32, 1, 1, 1, 1, 1, 1, 1, 1, 1) I0418 18:06:19.215882 139213532985152 checkpointing.py:677] Setting up checkpoint logger... I0418 18:06:19.215936 139213532985152 checkpointing.py:233] Creating checkpoint manager with ocdbt=True and zarr3=True I0418 18:06:19.215980 139213532985152 pytree_checkpoint_handler.py:592] save_device_host_concurrent_bytes=None I0418 18:06:19.216330 139213532985152 base_pytree_checkpoint_handler.py:441] 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 0x7e9c75b3c170>, enable_pinned_host_transfer=False, save_concurrent_bytes: 96000000000 (89.4 GiB), restore_concurrent_bytes: 96000000000 (89.4 GiB) I0418 18:06:23.367615 139213532985152 checkpointing.py:265] Enabling policy for fixed interval checkpointing. I0418 18:06:23.367856 139213532985152 checkpoint_manager.py:708] [process=7][thread=MainThread] CheckpointManager init: checkpointers=None, item_names=('items',), item_handlers={'items': <orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeCheckpointHandler object at 0x7e9b69fd6630>}, handler_registry=None I0418 18:06:23.368108 139213532985152 composite_checkpoint_handler.py:237] Deferred registration for item: "items". Adding handler `<orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeCheckpointHandler object at 0x7e9b69fd6630>` 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`. I0418 18:06:23.368158 139213532985152 composite_checkpoint_handler.py:237] Deferred registration for item: "metrics". Adding handler `<orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonCheckpointHandler object at 0x7e881448d580>` 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`. I0418 18:06:23.368190 139213532985152 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 0x7e9b69fd6630>, ('items', <class 'orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeRestoreArgs'>): <orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeCheckpointHandler object at 0x7e9b69fd6630>, ('metrics', <class 'orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonSaveArgs'>): <orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonCheckpointHandler object at 0x7e881448d580>, ('metrics', <class 'orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonRestoreArgs'>): <orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonCheckpointHandler object at 0x7e881448d580>}). I0418 18:06:23.368507 139213532985152 abstract_checkpointer.py:35] orbax-checkpoint version: 0.11.34 I0418 18:06:23.368576 139213532985152 async_checkpointer.py:192] [process=7][thread=MainThread] Using barrier_sync_fn: <function get_barrier_sync_fn.<locals>._fn at 0x7e87307cb880> timeout: 1200 secs and primary_host=0 for async checkpoint writes I0418 18:06:25.040515 139213532985152 checkpoint_manager.py:1812] Found 0 checkpoint steps in gs://lance-maxtext/linen_ckpt_xpk_main_20260418_180002/linen_xpk_main_20260418_180002_02_synthetic/checkpoints I0418 18:06:25.042809 139213532985152 checkpoint_manager.py:929] [process=7][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_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, lightweight_initialize=False), root_directory=gs://lance-maxtext/linen_ckpt_xpk_main_20260418_180002/linen_xpk_main_20260418_180002_02_synthetic/checkpoints: <orbax.checkpoint.checkpoint_manager.CheckpointManager object at 0x7e881448ab40> I0418 18:06:25.042919 139213532985152 checkpointing.py:301] Checkpoint manager created! I0418 18:06:25.968616 139213532985152 nnx_wrappers.py:437] Unknown Logical: bfloat16[32,2048,2048]...................................... ('activation_batch', 'activation_norm_length', 'activation_embed'). I0418 18:06:25.968731 139213532985152 nnx_wrappers.py:437] Unknown Physical: bfloat16[32,2048,2048]...................................... ('fsdp', None, None). I0418 18:06:26.349723 139213532985152 attentions.py:1088] attentions/inputs_q Logical: bfloat16[32,2048,2048]...................................... ('activation_batch', 'activation_attn_length', 'activation_attn_embed'). I0418 18:06:26.349818 139213532985152 attentions.py:1088] attentions/inputs_q Physical: bfloat16[32,2048,2048]...................................... ('fsdp', None, None). I0418 18:06:26.366260 139213532985152 attentions.py:1089] attentions/inputs_kv Logical: bfloat16[32,2048,2048]...................................... ('activation_batch', 'activation_attn_length', 'activation_attn_embed'). I0418 18:06:26.366317 139213532985152 attentions.py:1089] attentions/inputs_kv Physical: bfloat16[32,2048,2048]...................................... ('fsdp', None, None). I0418 18:06:26.390088 139213532985152 attentions.py:1154] attentions/query Logical: bfloat16[32,2048,16,128].................................... ('activation_kv_batch', 'activation_attn_length', 'activation_kv_heads', 'activation_kv_head_dim'). I0418 18:06:26.390157 139213532985152 attentions.py:1154] attentions/query Physical: bfloat16[32,2048,16,128].................................... ('fsdp', None, None, None). I0418 18:06:26.406702 139213532985152 attentions.py:1155] attentions/key Logical: bfloat16[32,2048,16,128].................................... ('activation_kv_batch', 'activation_attn_length', 'activation_kv_heads', 'activation_kv_head_dim'). I0418 18:06:26.406764 139213532985152 attentions.py:1155] attentions/key Physical: bfloat16[32,2048,16,128].................................... ('fsdp', None, None, None). I0418 18:06:26.423184 139213532985152 attentions.py:1156] attentions/value Logical: bfloat16[32,2048,16,128].................................... ('activation_kv_batch', 'activation_attn_length', 'activation_kv_heads', 'activation_kv_head_dim'). I0418 18:06:26.423245 139213532985152 attentions.py:1156] attentions/value Physical: bfloat16[32,2048,16,128].................................... ('fsdp', None, None, None). I0418 18:06:26.448630 139213532985152 attentions.py:1197] attentions/out Logical: bfloat16[32,2048,16,128].................................... ('activation_batch', 'activation_attn_length', 'activation_heads', 'activation_kv'). I0418 18:06:26.448699 139213532985152 attentions.py:1197] attentions/out Physical: bfloat16[32,2048,16,128].................................... ('fsdp', None, None, None). I0418 18:06:26.469554 139213532985152 linears.py:525] linears/x Logical: bfloat16[32,2048,7168]...................................... ('activation_batch', 'activation_length', 'activation_mlp'). I0418 18:06:26.469618 139213532985152 linears.py:525] linears/x Physical: bfloat16[32,2048,7168]...................................... ('fsdp', None, None). I0418 18:06:26.684895 139213532985152 checkpointing.py:577] checkpoint manager exists so trying to load this run's existing checkpoint I0418 18:06:26.685003 139213532985152 checkpointing.py:665] No existing checkpoints found, not restoring checkpoint. fsdp: 32 I0418 18:06:28.112239 139213532985152 maxtext_utils.py:1654] params/params/decoder/decoder_norm/scale Shape: float32[2048] Logical: P('norm',) Physical: (None,) I0418 18:06:28.112364 139213532985152 maxtext_utils.py:1654] params/params/decoder/layers/mlp/wi_0/kernel Shape: float32[2048,16,7168] Logical: P('embed', 'layers', 'mlp') Physical: ('fsdp', None, None) I0418 18:06:28.112416 139213532985152 maxtext_utils.py:1654] params/params/decoder/layers/mlp/wi_1/kernel Shape: float32[2048,16,7168] Logical: P('embed', 'layers', 'mlp') Physical: ('fsdp', None, None) I0418 18:06:28.112474 139213532985152 maxtext_utils.py:1654] params/params/decoder/layers/mlp/wo/kernel Shape: float32[7168,16,2048] Logical: P('mlp', 'layers', 'embed') Physical: (None, None, 'fsdp') I0418 18:06:28.112533 139213532985152 maxtext_utils.py:1654] params/params/decoder/layers/post_self_attention_layer_norm/scale Shape: float32[2048,16] Logical: P('norm', 'layers') Physical: (None, None) I0418 18:06:28.112574 139213532985152 maxtext_utils.py:1654] params/params/decoder/layers/pre_self_attention_layer_norm/scale Shape: float32[2048,16] Logical: P('norm', 'layers') Physical: (None, None) I0418 18:06:28.112627 139213532985152 maxtext_utils.py:1654] 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) I0418 18:06:28.112680 139213532985152 maxtext_utils.py:1654] 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') I0418 18:06:28.112719 139213532985152 maxtext_utils.py:1654] 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) I0418 18:06:28.112757 139213532985152 maxtext_utils.py:1654] 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) I0418 18:06:28.112804 139213532985152 maxtext_utils.py:1654] params/params/decoder/logits_dense/kernel Shape: float32[2048,32000] Logical: P('embed_vocab', 'vocab') Physical: ('fsdp', None) I0418 18:06:28.112853 139213532985152 maxtext_utils.py:1654] params/params/token_embedder/embedding Shape: float32[32000,2048] Logical: P('vocab', 'embed_vocab') Physical: (None, 'fsdp') I0418 18:06:28.605794 139213532985152 train.py:155] train/xent Logical: float32[32,2048]............................................ ('activation_embed_and_logits_batch', 'activation_length'). I0418 18:06:28.605895 139213532985152 train.py:155] train/xent Physical: float32[32,2048]............................................ ('fsdp', None). I0418 18:06:28.621365 139213532985152 train.py:162] train/z_loss Logical: float32[32,2048]............................................ ('activation_embed_and_logits_batch', 'activation_length'). I0418 18:06:28.621431 139213532985152 train.py:162] train/z_loss Physical: float32[32,2048]............................................ ('fsdp', None). I0418 18:06:39.413702 139213532985152 max_utils.py:791] Total memory size: 1.5 GB, Output size: 0.4 GB, Temp size: 1.1 GB, Argument size: 0.4 GB, Host temp size: 0.0 GB. I0418 18:06:39.414582 139213532985152 metric_logger.py:301] number parameters: 1.104 billion I0418 18:06:50.894916 139213532985152 checkpointing.py:772] Waiting for step 0 to finish before checkpoint... I0418 18:06:51.047773 139213532985152 checkpointing.py:776] Waited 0.15283632278442383 seconds for step 0 to finish before starting checkpointing. I0418 18:06:51.050359 139213532985152 checkpoint_manager.py:2009] [process=7][thread=MainThread][wait_until_finished] No Save Finalize thread to wait for. Returning. I0418 18:06:51.052017 139213532985152 checkpoint_manager.py:1512] [process=7] Saving checkpoint at step 0 I0418 18:06:51.053503 139213532985152 event_tracking.py:70] [process=7] [async] Started save checkpoint @ gs://lance-maxtext/linen_ckpt_xpk_main_20260418_180002/linen_xpk_main_20260418_180002_02_synthetic/checkpoints/0. I0418 18:06:51.815512 139213532985152 signaling_client.py:364] Using JaxDistributedSignalingClient I0418 18:06:51.816776 139213532985152 jax_array_handlers.py:360] Scheduling D2H of 39 prioritized jax.Array. I0418 18:06:51.816832 139213532985152 replica_slices.py:424] 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 I0418 18:06:52.093092 139213532985152 base_pytree_checkpoint_handler.py:154] [process=7][thread=MainThread] Initiated "orbax.checkpoint._src.serialization.jax_array_handlers.ArrayHandler".serialize. Time taken: 0.277638s I0418 18:06:52.093270 139213532985152 base_pytree_checkpoint_handler.py:130] [process=7] /jax/orbax/write/blocking_gbytes_per_sec: 5.450 GiB/s (total gbytes: 1.5 GiB) (time elapsed: 0.2830619812011719 s) (per-host) I0418 18:06:52.093325 139213532985152 base_pytree_checkpoint_handler.py:768] [process=7][thread=MainThread] Initiated Pytree async_save. Time taken: 0.283127s (batch_requests_ready=0.002185s, total_serialization_initiated=0.280868s, others=0.000074s) I0418 18:06:52.093415 139213532985152 composite_checkpoint_handler.py:715] [process=7][thread=MainThread] Initiated CompositeCheckpointHandler.async_save. Time taken: 0.287293s (all_items=0.000018s, per_item={'items': '0.00001764'}, temp_paths=0.287276) I0418 18:06:52.094237 139213532985152 event_tracking.py:125] [process=7] [async] Finished blocking save in 1.04 seconds. Continuing save @ gs://lance-maxtext/linen_ckpt_xpk_main_20260418_180002/linen_xpk_main_20260418_180002_02_synthetic/checkpoints/0. I0418 18:06:52.094519 139083852723968 async_checkpointer.py:76] [process=7][thread=async_save] Background save thread started. Deadline for this save operation is 2026-04-18 18:26:52.094495 I0418 18:06:52.100398 139213532985152 checkpoint_manager.py:1560] [process=7][thread=MainThread][step=0] Starting CheckpointManager Save Finalize thread=save_finalize I0418 18:06:52.100628 139082776700672 async_checkpointer.py:280] [process=7][thread=save_finalize] Waiting for background save thread=async_save. I0418 18:06:52.100745 139213532985152 standard_logger.py:34] {'step': 0, 'event_type': 'save', 'directory': 'gs://lance-maxtext/linen_ckpt_xpk_main_20260418_180002/linen_xpk_main_20260418_180002_02_synthetic/checkpoints', 'reached_preemption': False, 'preemption_received_at': None, 'synchronous': False, 'wait_for_prev_start_time': 1776535611.05034, 'wait_for_prev_duration_secs': 6.198883056640625e-05, 'time_between_consecutive_saves_sec': None, 'checkpointer_blocking_start_time': 1776535611.0520616, 'checkpointer_blocking_duration_secs': 1.042572259902954, 'get_old_steps_start_time': 1776535612.0946612, 'get_old_steps_duration_secs': 2.9802322387695312e-05, 'checkpoint_manager_blocking_start_time': 1776535611.048341, 'checkpoint_manager_blocking_duration_secs': 1.0523667335510254} I0418 18:06:52.100853 139213532985152 checkpointing.py:408] Started an asynchronous checkpoint save for step 0 I0418 18:06:52.100902 139213532985152 max_utils.py:750] Memstats: After params initialized: I0418 18:06:52.100948 139213532985152 max_utils.py:756] Using (GB) 0.43 / 31.25 (1.376000%) on TPU_26(process=7,(2,6,0,0)) I0418 18:06:52.100981 139213532985152 max_utils.py:756] Using (GB) 0.43 / 31.25 (1.376000%) on TPU_27(process=7,(3,6,0,0)) I0418 18:06:52.101009 139213532985152 max_utils.py:756] Using (GB) 0.43 / 31.25 (1.376000%) on TPU_30(process=7,(2,7,0,0)) I0418 18:06:52.101032 139213532985152 max_utils.py:756] Using (GB) 0.43 / 31.25 (1.376000%) on TPU_31(process=7,(3,7,0,0)) I0418 18:06:52.413053 139213532985152 metric_logger.py:196] completed step: 0, seconds: 11.480, TFLOP/s/device: 1.184, Tokens/s/device: 178.394, total_weights: 65536, loss: 10.874, lm_loss: 10.874, perplexity: 52779.875 I0418 18:06:52.599794 139213532985152 metric_logger.py:196] completed step: 1, seconds: 1.517, TFLOP/s/device: 8.959, Tokens/s/device: 1350.353, total_weights: 65536, loss: 10.874, lm_loss: 10.874, perplexity: 52779.875 I0418 18:06:53.016669 139213532985152 metric_logger.py:196] completed step: 2, seconds: 0.040, TFLOP/s/device: 341.445, Tokens/s/device: 51466.338, total_weights: 65536, loss: 10.262, lm_loss: 10.262, perplexity: 28637.129 I0418 18:06:53.164270 139213532985152 metric_logger.py:196] completed step: 3, seconds: 0.417, TFLOP/s/device: 32.594, Tokens/s/device: 4912.944, total_weights: 65536, loss: 9.734, lm_loss: 9.734, perplexity: 16889.203 I0418 18:06:53.459233 139213532985152 metric_logger.py:196] completed step: 4, seconds: 0.153, TFLOP/s/device: 88.918, Tokens/s/device: 13402.615, total_weights: 65536, loss: 9.277, lm_loss: 9.277, perplexity: 10694.614 I0418 18:06:53.465097 139213532985152 metric_logger.py:196] completed step: 5, seconds: 0.148, TFLOP/s/device: 92.064, Tokens/s/device: 13876.843, total_weights: 65536, loss: 8.892, lm_loss: 8.892, perplexity: 7270.717 I0418 18:06:57.601235 2858 google_auth_provider.cc:181] Running on GCE, using service account 562977990677-compute@developer.gserviceaccount.com I0418 18:07:01.081295 139082785093376 array_metadata_store.py:203] [process=7][thread=array_type_handler] Wrote 39 array_metadata.ArrayMetadata to gs://lance-maxtext/linen_ckpt_xpk_main_20260418_180002/linen_xpk_main_20260418_180002_02_synthetic/checkpoints/0/items/array_metadatas/process_7 I0418 18:07:18.472151 139213532985152 metric_logger.py:196] completed step: 6, seconds: 0.296, TFLOP/s/device: 45.973, Tokens/s/device: 6929.547, total_weights: 65536, loss: 8.592, lm_loss: 8.592, perplexity: 5390.759 I0418 18:07:18.619755 139213532985152 metric_logger.py:196] completed step: 7, seconds: 24.860, TFLOP/s/device: 0.547, Tokens/s/device: 82.382, total_weights: 65536, loss: 8.384, lm_loss: 8.384, perplexity: 4376.905 I0418 18:07:18.767367 139213532985152 metric_logger.py:196] completed step: 8, seconds: 0.153, TFLOP/s/device: 88.936, Tokens/s/device: 13405.335, total_weights: 65536, loss: 8.255, lm_loss: 8.255, perplexity: 3846.406 I0418 18:07:18.914693 139213532985152 checkpointing.py:772] Waiting for step 9 to finish before checkpoint... I0418 18:07:18.915394 139213532985152 checkpointing.py:776] Waited 0.0007214546203613281 seconds for step 9 to finish before starting checkpointing. I0418 18:07:18.917498 139213532985152 checkpoint_manager.py:2020] [process=7][thread=MainThread][step=0][wait_until_finished] Waiting for Save Finalize thread (save_finalize) to complete. I0418 18:07:31.226597 139083852723968 base_pytree_checkpoint_handler.py:130] [process=7] /jax/orbax/write/gbytes_per_sec: 40.075 MiB/s (total gbytes: 1.5 GiB) (time elapsed: 39.4163293838501 s) (per-host) I0418 18:07:31.226727 139083852723968 async_checkpointer.py:90] [process=7][thread=async_save] 3 Handler Commit operations completed. Time taken: 39.132128s. I0418 18:07:40.136698 139083852723968 async_checkpointer.py:160] [process=7][thread=async_save] Background save thread done. Time taken: 48.042082s. I0418 18:07:40.136959 139082776700672 async_checkpointer.py:288] [process=7][thread=save_finalize] Done with waiting for background save thread=async_save. I0418 18:07:40.137097 139082776700672 async_checkpointer.py:298] [process=7][thread=save_finalize] No errors found in background save thread=async_save. I0418 18:07:40.137150 139082776700672 checkpoint_manager.py:2137] [process=7][thread=save_finalize][step=0] CheckpointManager Save Finalize is syncing with other hosts... I0418 18:07:40.138911 139082776700672 checkpoint_manager.py:2146] [process=7][thread=save_finalize][step=0] CheckpointManager Save Finalize is done on all hosts. I0418 18:07:40.139150 139213532985152 checkpoint_manager.py:2032] [process=7][thread=MainThread][step=0][wait_until_finished] Done waiting for Save Finalize thread (save_finalize) running at step=0. W0418 18:07:40.139292 139213532985152 checkpoint_manager.py:1452] Waiting for previous save to complete took 21.221793 seconds. If this number is high, consider checkpointing less frequently. I0418 18:07:40.140958 139213532985152 checkpoint_manager.py:1512] [process=7] Saving checkpoint at step 9 I0418 18:07:40.142998 139213532985152 event_tracking.py:70] [process=7] [async] Started save checkpoint @ gs://lance-maxtext/linen_ckpt_xpk_main_20260418_180002/linen_xpk_main_20260418_180002_02_synthetic/checkpoints/9. I0418 18:07:40.861851 139213532985152 jax_array_handlers.py:360] Scheduling D2H of 39 prioritized jax.Array. I0418 18:07:40.861947 139213532985152 replica_slices.py:424] 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 I0418 18:07:40.902238 139213532985152 base_pytree_checkpoint_handler.py:154] [process=7][thread=MainThread] Initiated "orbax.checkpoint._src.serialization.jax_array_handlers.ArrayHandler".serialize. Time taken: 0.041449s I0418 18:07:40.902404 139213532985152 base_pytree_checkpoint_handler.py:130] [process=7] /jax/orbax/write/blocking_gbytes_per_sec: 34.410 GiB/s (total gbytes: 1.5 GiB) (time elapsed: 0.0448298454284668 s) (per-host) I0418 18:07:40.902456 139213532985152 base_pytree_checkpoint_handler.py:768] [process=7][thread=MainThread] Initiated Pytree async_save. Time taken: 0.044891s (batch_requests_ready=0.001755s, total_serialization_initiated=0.043068s, others=0.000067s) I0418 18:07:40.902544 139213532985152 composite_checkpoint_handler.py:715] [process=7][thread=MainThread] Initiated CompositeCheckpointHandler.async_save. Time taken: 0.049036s (all_items=0.000016s, per_item={'items': '0.00001574'}, temp_paths=0.049020) I0418 18:07:40.903293 139213532985152 event_tracking.py:125] [process=7] [async] Finished blocking save in 0.76 seconds. Continuing save @ gs://lance-maxtext/linen_ckpt_xpk_main_20260418_180002/linen_xpk_main_20260418_180002_02_synthetic/checkpoints/9. I0418 18:07:40.903562 139082776700672 async_checkpointer.py:76] [process=7][thread=async_save] Background save thread started. Deadline for this save operation is 2026-04-18 18:27:40.903538 I0418 18:07:40.905514 139213532985152 checkpoint_manager.py:1560] [process=7][thread=MainThread][step=9] Starting CheckpointManager Save Finalize thread=save_finalize I0418 18:07:40.905796 139062425929472 async_checkpointer.py:280] [process=7][thread=save_finalize] Waiting for background save thread=async_save. I0418 18:07:40.905934 139213532985152 standard_logger.py:34] {'step': 9, 'event_type': 'save', 'directory': 'gs://lance-maxtext/linen_ckpt_xpk_main_20260418_180002/linen_xpk_main_20260418_180002_02_synthetic/checkpoints', 'reached_preemption': False, 'preemption_received_at': None, 'synchronous': False, 'wait_for_prev_start_time': 1776535638.9174688, 'wait_for_prev_duration_secs': 21.221792936325073, 'time_between_consecutive_saves_sec': None, 'checkpointer_blocking_start_time': 1776535660.1409976, 'checkpointer_blocking_duration_secs': 0.7626664638519287, 'get_old_steps_start_time': 1776535660.9036894, 'get_old_steps_duration_secs': 3.123283386230469e-05, 'checkpoint_manager_blocking_start_time': 1776535638.9156184, 'checkpoint_manager_blocking_duration_secs': 21.99028182029724} I0418 18:07:40.906121 139213532985152 checkpointing.py:408] Started an asynchronous checkpoint save for step 9 I0418 18:07:40.906168 139213532985152 checkpoint_manager.py:2020] [process=7][thread=MainThread][step=9][wait_until_finished] Waiting for Save Finalize thread (save_finalize) to complete. I0418 18:07:49.572019 139082785093376 array_metadata_store.py:203] [process=7][thread=array_type_handler] Wrote 39 array_metadata.ArrayMetadata to gs://lance-maxtext/linen_ckpt_xpk_main_20260418_180002/linen_xpk_main_20260418_180002_02_synthetic/checkpoints/9/items/array_metadatas/process_7 I0418 18:08:23.994221 139082776700672 base_pytree_checkpoint_handler.py:130] [process=7] /jax/orbax/write/gbytes_per_sec: 36.618 MiB/s (total gbytes: 1.5 GiB) (time elapsed: 43.13660931587219 s) (per-host) I0418 18:08:23.994349 139082776700672 async_checkpointer.py:90] [process=7][thread=async_save] 3 Handler Commit operations completed. Time taken: 43.090707s. I0418 18:08:34.151354 139082776700672 async_checkpointer.py:160] [process=7][thread=async_save] Background save thread done. Time taken: 53.247697s. I0418 18:08:34.151639 139062425929472 async_checkpointer.py:288] [process=7][thread=save_finalize] Done with waiting for background save thread=async_save. I0418 18:08:34.151765 139062425929472 async_checkpointer.py:298] [process=7][thread=save_finalize] No errors found in background save thread=async_save. I0418 18:08:34.151825 139062425929472 checkpoint_manager.py:2137] [process=7][thread=save_finalize][step=9] CheckpointManager Save Finalize is syncing with other hosts... I0418 18:08:34.153744 139062425929472 checkpoint_manager.py:2146] [process=7][thread=save_finalize][step=9] CheckpointManager Save Finalize is done on all hosts. I0418 18:08:34.153940 139213532985152 checkpoint_manager.py:2032] [process=7][thread=MainThread][step=9][wait_until_finished] Done waiting for Save Finalize thread (save_finalize) running at step=9. I0418 18:08:34.154109 139213532985152 checkpoint_manager.py:2009] [process=7][thread=MainThread][wait_until_finished] No Save Finalize thread to wait for. Returning. I0418 18:08:34.155037 139213532985152 metric_logger.py:196] completed step: 9, seconds: 0.148, TFLOP/s/device: 91.796, Tokens/s/device: 13836.435, total_weights: 65536, loss: 8.179, lm_loss: 8.179, perplexity: 3564.635 Per train step: Total TFLOPs: 13.59 split as 93.93% learnable weight flops and 6.07% attention flops XPK End: Sat Apr 18 18:08:43 UTC 2026 EXIT_CODE=0
XPK Start: Sun Apr 19 08:59:32 UTC 2026 2026-04-19 08:59:36.425472: E external/local_xla/xla/stream_executor/cuda/cuda_fft.cc:467] Unable to register cuFFT factory: Attempting to register factory for plugin cuFFT when one has already been registered WARNING: All log messages before absl::InitializeLog() is called are written to STDERR E0000 00:00:1776589176.437988 10 cuda_dnn.cc:8579] Unable to register cuDNN factory: Attempting to register factory for plugin cuDNN when one has already been registered E0000 00:00:1776589176.441712 10 cuda_blas.cc:1407] Unable to register cuBLAS factory: Attempting to register factory for plugin cuBLAS when one has already been registered W0000 00:00:1776589176.453070 10 computation_placer.cc:177] computation placer already registered. Please check linkage and avoid linking the same target more than once. W0000 00:00:1776589176.453089 10 computation_placer.cc:177] computation placer already registered. Please check linkage and avoid linking the same target more than once. W0000 00:00:1776589176.453091 10 computation_placer.cc:177] computation placer already registered. Please check linkage and avoid linking the same target more than once. W0000 00:00:1776589176.453093 10 computation_placer.cc:177] computation placer already registered. Please check linkage and avoid linking the same target more than once. 2026-04-19 09:00:10.160932: 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) I0419 09:00:10.668677 133951773706048 max_utils.py:273] Attempting to initialize the jax distributed system... INFO:2026-04-19 09:00:19,708:jax._src.distributed:140: Starting JAX distributed service on [::]:8482 I0419 09:00:19.708698 133951773706048 distributed.py:140] Starting JAX distributed service on [::]:8482 INFO:2026-04-19 09:00:19,711:jax._src.distributed:157: Connecting to JAX distributed service on mt-02-synthetic-l2rig-slice-job-0-0.mt-02-synthetic-l2rig:8482 I0419 09:00:19.711211 133951773706048 distributed.py:157] Connecting to JAX distributed service on mt-02-synthetic-l2rig-slice-job-0-0.mt-02-synthetic-l2rig:8482 I0419 09:00:28.322265 133951773706048 max_utils.py:284] Jax distributed system initialized! I0419 09:00:35.348552 133951773706048 max_utils.py:800] System Information: Jax Version: 0.8.1 I0419 09:00:35.348657 133951773706048 max_utils.py:801] System Information: Jaxlib Version: 0.8.1 I0419 09:00:35.348697 133951773706048 max_utils.py:802] System Information: Jax Backend: PJRT C API TFRT TPU v6 lite Built on Nov 12 2025 14:16:36 (1762985796) cl/831091709 I0419 09:00:35.348736 133951773706048 train_utils.py:347] WARNING: Sequence packing is essentially ignored for synthetic data. Please use a real dataset to use sequence packing. I0419 09:00:35.972409 133951773706048 maxtext_utils.py:1398] Num_devices: 32, shape (1, 1, 1, 32, 1, 1, 1, 1, 1, 1, 1, 1, 1) I0419 09:00:35.972695 133951773706048 checkpointing.py:677] Setting up checkpoint logger... I0419 09:00:35.972757 133951773706048 checkpointing.py:233] Creating checkpoint manager with ocdbt=True and zarr3=True I0419 09:00:35.972800 133951773706048 pytree_checkpoint_handler.py:589] save_device_host_concurrent_bytes=None I0419 09:00:35.973171 133951773706048 base_pytree_checkpoint_handler.py:415] 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 0x79d3a1974e30>, enable_pinned_host_transfer=False, save_concurrent_bytes: 96000000000 (89.4 GiB), restore_concurrent_bytes: 96000000000 (89.4 GiB) I0419 09:00:39.218423 133951773706048 checkpointing.py:265] Enabling policy for fixed interval checkpointing. I0419 09:00:39.218660 133951773706048 checkpoint_manager.py:709] [process=4][thread=MainThread] CheckpointManager init: checkpointers=None, item_names=('items',), item_handlers={'items': <orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeCheckpointHandler object at 0x79c09cf95ee0>}, handler_registry=None I0419 09:00:39.218894 133951773706048 composite_checkpoint_handler.py:237] Deferred registration for item: "items". Adding handler `<orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeCheckpointHandler object at 0x79c09cf95ee0>` 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`. I0419 09:00:39.218954 133951773706048 composite_checkpoint_handler.py:237] Deferred registration for item: "metrics". Adding handler `<orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonCheckpointHandler object at 0x79c09cf98170>` 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`. I0419 09:00:39.218992 133951773706048 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 0x79c09cf95ee0>, ('items', <class 'orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeRestoreArgs'>): <orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeCheckpointHandler object at 0x79c09cf95ee0>, ('metrics', <class 'orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonSaveArgs'>): <orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonCheckpointHandler object at 0x79c09cf98170>, ('metrics', <class 'orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonRestoreArgs'>): <orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonCheckpointHandler object at 0x79c09cf98170>}). I0419 09:00:39.219479 133951773706048 abstract_checkpointer.py:35] orbax-checkpoint version: 0.11.33 I0419 09:00:39.219553 133951773706048 async_checkpointer.py:177] [process=4][thread=MainThread] Using barrier_sync_fn: <function get_barrier_sync_fn.<locals>._fn at 0x79bef81f8040> timeout: 600 secs and primary_host=0 for async checkpoint writes I0419 09:00:40.020436 133951773706048 checkpoint_manager.py:1818] Found 0 checkpoint steps in gs://lance-maxtext/linen_ckpt_xpk_test_pipeline_scan_nnx_20260419_085454/linen_xpk_test_pipeline_scan_nnx_20260419_085454_02_synthetic/checkpoints I0419 09:00:40.484757 133951773706048 checkpoint_manager.py:929] [process=4][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_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, lightweight_initialize=False), root_directory=gs://lance-maxtext/linen_ckpt_xpk_test_pipeline_scan_nnx_20260419_085454/linen_xpk_test_pipeline_scan_nnx_20260419_085454_02_synthetic/checkpoints: <orbax.checkpoint.checkpoint_manager.CheckpointManager object at 0x79c09cf02ae0> I0419 09:00:40.484929 133951773706048 checkpointing.py:301] Checkpoint manager created! I0419 09:00:41.386625 133951773706048 nnx_wrappers.py:453] Unknown Logical: bfloat16[32,2048,2048]...................................... ('activation_batch', 'activation_norm_length', 'activation_embed'). I0419 09:00:41.386736 133951773706048 nnx_wrappers.py:453] Unknown Physical: bfloat16[32,2048,2048]...................................... ('fsdp', None, None). I0419 09:00:41.759856 133951773706048 attentions.py:1084] attentions/inputs_q Logical: bfloat16[32,2048,2048]...................................... ('activation_batch', 'activation_attn_length', 'activation_attn_embed'). I0419 09:00:41.759962 133951773706048 attentions.py:1084] attentions/inputs_q Physical: bfloat16[32,2048,2048]...................................... ('fsdp', None, None). I0419 09:00:41.775545 133951773706048 attentions.py:1085] attentions/inputs_kv Logical: bfloat16[32,2048,2048]...................................... ('activation_batch', 'activation_attn_length', 'activation_attn_embed'). I0419 09:00:41.775604 133951773706048 attentions.py:1085] attentions/inputs_kv Physical: bfloat16[32,2048,2048]...................................... ('fsdp', None, None). I0419 09:00:41.798264 133951773706048 attentions.py:1150] attentions/query Logical: bfloat16[32,2048,16,128].................................... ('activation_kv_batch', 'activation_attn_length', 'activation_kv_heads', 'activation_kv_head_dim'). I0419 09:00:41.798333 133951773706048 attentions.py:1150] attentions/query Physical: bfloat16[32,2048,16,128].................................... ('fsdp', None, None, None). I0419 09:00:41.814094 133951773706048 attentions.py:1151] attentions/key Logical: bfloat16[32,2048,16,128].................................... ('activation_kv_batch', 'activation_attn_length', 'activation_kv_heads', 'activation_kv_head_dim'). I0419 09:00:41.814153 133951773706048 attentions.py:1151] attentions/key Physical: bfloat16[32,2048,16,128].................................... ('fsdp', None, None, None). I0419 09:00:41.829934 133951773706048 attentions.py:1152] attentions/value Logical: bfloat16[32,2048,16,128].................................... ('activation_kv_batch', 'activation_attn_length', 'activation_kv_heads', 'activation_kv_head_dim'). I0419 09:00:41.830007 133951773706048 attentions.py:1152] attentions/value Physical: bfloat16[32,2048,16,128].................................... ('fsdp', None, None, None). I0419 09:00:41.854592 133951773706048 attentions.py:1193] attentions/out Logical: bfloat16[32,2048,16,128].................................... ('activation_batch', 'activation_attn_length', 'activation_heads', 'activation_kv'). I0419 09:00:41.854668 133951773706048 attentions.py:1193] attentions/out Physical: bfloat16[32,2048,16,128].................................... ('fsdp', None, None, None). I0419 09:00:41.874052 133951773706048 linears.py:525] linears/x Logical: bfloat16[32,2048,7168]...................................... ('activation_batch', 'activation_length', 'activation_mlp'). I0419 09:00:41.874114 133951773706048 linears.py:525] linears/x Physical: bfloat16[32,2048,7168]...................................... ('fsdp', None, None). I0419 09:00:42.066029 133951773706048 checkpointing.py:577] checkpoint manager exists so trying to load this run's existing checkpoint I0419 09:00:42.066143 133951773706048 checkpointing.py:665] No existing checkpoints found, not restoring checkpoint. fsdp: 32 I0419 09:00:43.479969 133951773706048 maxtext_utils.py:1501] params/params/decoder/decoder_norm/scale Shape: float32[2048] Logical: PartitionSpec('norm',) Physical: (None,) I0419 09:00:43.480115 133951773706048 maxtext_utils.py:1501] params/params/decoder/layers/mlp/wi_0/kernel Shape: float32[2048,16,7168] Logical: PartitionSpec('embed', 'layers', 'mlp') Physical: ('fsdp', None, None) I0419 09:00:43.480170 133951773706048 maxtext_utils.py:1501] params/params/decoder/layers/mlp/wi_1/kernel Shape: float32[2048,16,7168] Logical: PartitionSpec('embed', 'layers', 'mlp') Physical: ('fsdp', None, None) I0419 09:00:43.480228 133951773706048 maxtext_utils.py:1501] params/params/decoder/layers/mlp/wo/kernel Shape: float32[7168,16,2048] Logical: PartitionSpec('mlp', 'layers', 'embed') Physical: (None, None, 'fsdp') I0419 09:00:43.480280 133951773706048 maxtext_utils.py:1501] params/params/decoder/layers/post_self_attention_layer_norm/scale Shape: float32[2048,16] Logical: PartitionSpec('norm', 'layers') Physical: (None, None) I0419 09:00:43.480318 133951773706048 maxtext_utils.py:1501] params/params/decoder/layers/pre_self_attention_layer_norm/scale Shape: float32[2048,16] Logical: PartitionSpec('norm', 'layers') Physical: (None, None) I0419 09:00:43.480370 133951773706048 maxtext_utils.py:1501] params/params/decoder/layers/self_attention/key/kernel Shape: float32[2048,16,16,128] Logical: PartitionSpec('embed', 'layers', 'kv_heads', 'kv_head_dim') Physical: ('fsdp', None, None, None) I0419 09:00:43.480421 133951773706048 maxtext_utils.py:1501] params/params/decoder/layers/self_attention/out/kernel Shape: float32[16,16,128,2048] Logical: PartitionSpec('heads', 'layers', 'kv', 'embed') Physical: (None, None, None, 'fsdp') I0419 09:00:43.480461 133951773706048 maxtext_utils.py:1501] params/params/decoder/layers/self_attention/query/kernel Shape: float32[2048,16,16,128] Logical: PartitionSpec('embed', 'layers', 'q_heads', 'kv') Physical: ('fsdp', None, None, None) I0419 09:00:43.480497 133951773706048 maxtext_utils.py:1501] params/params/decoder/layers/self_attention/value/kernel Shape: float32[2048,16,16,128] Logical: PartitionSpec('embed', 'layers', 'kv_heads', 'kv_head_dim') Physical: ('fsdp', None, None, None) I0419 09:00:43.480562 133951773706048 maxtext_utils.py:1501] params/params/decoder/logits_dense/kernel Shape: float32[2048,32000] Logical: PartitionSpec('embed', 'vocab') Physical: ('fsdp', None) I0419 09:00:43.480615 133951773706048 maxtext_utils.py:1501] params/params/token_embedder/embedding Shape: float32[32000,2048] Logical: PartitionSpec('vocab', 'embed') Physical: (None, 'fsdp') I0419 09:00:43.988734 133951773706048 train.py:156] train/xent Logical: float32[32,2048]............................................ ('activation_embed_and_logits_batch', 'activation_length'). I0419 09:00:43.988837 133951773706048 train.py:156] train/xent Physical: float32[32,2048]............................................ ('fsdp', None). I0419 09:00:44.003599 133951773706048 train.py:163] train/z_loss Logical: float32[32,2048]............................................ ('activation_embed_and_logits_batch', 'activation_length'). I0419 09:00:44.003656 133951773706048 train.py:163] train/z_loss Physical: float32[32,2048]............................................ ('fsdp', None). I0419 09:00:54.770881 133951773706048 max_utils.py:791] Total memory size: 1.5 GB, Output size: 0.4 GB, Temp size: 1.1 GB, Argument size: 0.4 GB, Host temp size: 0.0 GB. I0419 09:00:54.771725 133951773706048 metric_logger.py:289] number parameters: 1.104 billion I0419 09:01:05.825982 133951773706048 checkpointing.py:772] Waiting for step 0 to finish before checkpoint... I0419 09:01:05.978853 133951773706048 checkpointing.py:776] Waited 0.15285515785217285 seconds for step 0 to finish before starting checkpointing. I0419 09:01:05.981499 133951773706048 checkpoint_manager.py:2013] [process=4][thread=MainThread][wait_until_finished] No Save Finalize thread to wait for. Returning. I0419 09:01:05.983375 133951773706048 checkpoint_manager.py:1518] [process=4] Saving checkpoint at step 0 I0419 09:01:05.986232 133951773706048 async_checkpointer.py:452] [process=4] Started async saving checkpoint to gs://lance-maxtext/linen_ckpt_xpk_test_pipeline_scan_nnx_20260419_085454/linen_xpk_test_pipeline_scan_nnx_20260419_085454_02_synthetic/checkpoints/0. I0419 09:01:06.617465 133951773706048 signaling_client.py:364] Using JaxDistributedSignalingClient I0419 09:01:06.618573 133951773706048 jax_array_handlers.py:358] Scheduling D2H of 39 prioritized jax.Array. I0419 09:01:06.618631 133951773706048 replica_slices.py:424] 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 I0419 09:01:06.863907 133951773706048 base_pytree_checkpoint_handler.py:153] [process=4][thread=MainThread] Initiated "orbax.checkpoint._src.serialization.jax_array_handlers.ArrayHandler".serialize. Time taken: 0.246521s I0419 09:01:06.864093 133951773706048 base_pytree_checkpoint_handler.py:129] [process=4] /jax/checkpoint/write/blocking_gbytes_per_sec: 2.818 GiB/s (total gbytes: 1.5 GiB) (time elapsed: 0.5474205017089844 s) (per-host) I0419 09:01:06.864144 133951773706048 base_pytree_checkpoint_handler.py:737] [process=4][thread=MainThread] Initiated Pytree async_save. Time taken: 0.547481s (batch_requests_ready=0.297347s, total_serialization_initiated=0.250065s, others=0.000069s) I0419 09:01:06.864230 133951773706048 composite_checkpoint_handler.py:715] [process=4][thread=MainThread] Initiated CompositeCheckpointHandler.async_save. Time taken: 0.551567s (all_items=0.000020s, per_item={'items': '0.00002003'}, temp_paths=0.551547) I0419 09:01:06.865215 133825389426432 async_checkpointer.py:79] [process=4][thread=async_save] Background save thread started. I0419 09:01:06.865388 133951773706048 async_checkpointer.py:561] Finished blocking save. Time taken: 0.881942s. Continuing background save to gs://lance-maxtext/linen_ckpt_xpk_test_pipeline_scan_nnx_20260419_085454/linen_xpk_test_pipeline_scan_nnx_20260419_085454_02_synthetic/checkpoints/0. I0419 09:01:07.260625 133951773706048 checkpoint_manager.py:1566] [process=4][thread=MainThread][step=0] Starting CheckpointManager Save Finalize thread=save_finalize I0419 09:01:07.261044 133824151262976 async_checkpointer.py:265] [process=4][thread=save_finalize] Waiting for background save thread=async_save. I0419 09:01:07.261213 133951773706048 standard_logger.py:34] {'step': 0, 'event_type': 'save', 'directory': 'gs://lance-maxtext/linen_ckpt_xpk_test_pipeline_scan_nnx_20260419_085454/linen_xpk_test_pipeline_scan_nnx_20260419_085454_02_synthetic/checkpoints', 'reached_preemption': False, 'preemption_received_at': None, 'synchronous': False, 'wait_for_prev_start_time': 1776589265.9814813, 'wait_for_prev_duration_secs': 6.341934204101562e-05, 'time_between_consecutive_saves_sec': None, 'checkpointer_blocking_start_time': 1776589265.983414, 'checkpointer_blocking_duration_secs': 0.8820798397064209, 'get_old_steps_start_time': 1776589266.865515, 'get_old_steps_duration_secs': 2.5272369384765625e-05, 'checkpoint_manager_blocking_start_time': 1776589265.9794302, 'checkpoint_manager_blocking_duration_secs': 1.2817418575286865} I0419 09:01:07.261321 133951773706048 checkpointing.py:408] Started an asynchronous checkpoint save for step 0 I0419 09:01:07.261373 133951773706048 max_utils.py:750] Memstats: After params initialized: I0419 09:01:07.261430 133951773706048 max_utils.py:756] Using (GB) 0.43 / 31.25 (1.376000%) on TPU_16(process=4,(0,4,0,0)) I0419 09:01:07.261464 133951773706048 max_utils.py:756] Using (GB) 0.43 / 31.25 (1.376000%) on TPU_17(process=4,(1,4,0,0)) I0419 09:01:07.261492 133951773706048 max_utils.py:756] Using (GB) 0.43 / 31.25 (1.376000%) on TPU_20(process=4,(0,5,0,0)) I0419 09:01:07.261515 133951773706048 max_utils.py:756] Using (GB) 0.43 / 31.25 (1.376000%) on TPU_21(process=4,(1,5,0,0)) I0419 09:01:07.529726 133951773706048 metric_logger.py:185] completed step: 0, seconds: 11.054, TFLOP/s/device: 1.229, Tokens/s/device: 185.270, total_weights: 65536, loss: 10.874 I0419 09:01:07.706140 133951773706048 metric_logger.py:185] completed step: 1, seconds: 1.702, TFLOP/s/device: 7.981, Tokens/s/device: 1202.981, total_weights: 65536, loss: 10.874 I0419 09:01:08.549864 133951773706048 metric_logger.py:185] completed step: 2, seconds: 0.029, TFLOP/s/device: 467.056, Tokens/s/device: 70399.780, total_weights: 65536, loss: 10.263 I0419 09:01:08.697317 133951773706048 metric_logger.py:185] completed step: 3, seconds: 0.845, TFLOP/s/device: 16.088, Tokens/s/device: 2424.934, total_weights: 65536, loss: 9.734 I0419 09:01:08.992651 133951773706048 metric_logger.py:185] completed step: 4, seconds: 0.153, TFLOP/s/device: 89.049, Tokens/s/device: 13422.467, total_weights: 65536, loss: 9.278 I0419 09:01:08.998412 133951773706048 metric_logger.py:185] completed step: 5, seconds: 0.147, TFLOP/s/device: 92.249, Tokens/s/device: 13904.731, total_weights: 65536, loss: 8.892 I0419 09:01:27.468448 2650 google_auth_provider.cc:181] Running on GCE, using service account 562977990677-compute@developer.gserviceaccount.com I0419 09:01:27.638670 133951773706048 metric_logger.py:185] completed step: 6, seconds: 0.296, TFLOP/s/device: 45.948, Tokens/s/device: 6925.728, total_weights: 65536, loss: 8.592 I0419 09:01:27.786065 133951773706048 metric_logger.py:185] completed step: 7, seconds: 18.488, TFLOP/s/device: 0.735, Tokens/s/device: 110.777, total_weights: 65536, loss: 8.384 I0419 09:01:27.933640 133951773706048 metric_logger.py:185] completed step: 8, seconds: 0.158, TFLOP/s/device: 85.783, Tokens/s/device: 12930.109, total_weights: 65536, loss: 8.255 I0419 09:01:28.080685 133951773706048 checkpointing.py:772] Waiting for step 9 to finish before checkpoint... I0419 09:01:28.081372 133951773706048 checkpointing.py:776] Waited 0.0007061958312988281 seconds for step 9 to finish before starting checkpointing. I0419 09:01:28.083517 133951773706048 checkpoint_manager.py:2024] [process=4][thread=MainThread][step=0][wait_until_finished] Waiting for Save Finalize thread (save_finalize) to complete. I0419 09:01:30.354830 133825397819136 array_metadata_store.py:203] [process=4][thread=array_type_handler] Wrote 39 array_metadata.ArrayMetadata to gs://lance-maxtext/linen_ckpt_xpk_test_pipeline_scan_nnx_20260419_085454/linen_xpk_test_pipeline_scan_nnx_20260419_085454_02_synthetic/checkpoints/0/items/array_metadatas/process_4 I0419 09:02:00.336714 133825389426432 base_pytree_checkpoint_handler.py:129] [process=4] /jax/checkpoint/write/gbytes_per_sec: 29.241 MiB/s (total gbytes: 1.5 GiB) (time elapsed: 54.02000045776367 s) (per-host) I0419 09:02:00.336835 133825389426432 async_checkpointer.py:90] [process=4][thread=async_save] 3 Handler Commit operations completed. Time taken: 53.471507s. I0419 09:02:10.512462 133825389426432 async_checkpointer.py:144] [process=4][thread=async_save] Background save thread done. Time taken: 63.647119s. I0419 09:02:10.512752 133824151262976 async_checkpointer.py:273] [process=4][thread=save_finalize] Done with waiting for background save thread=async_save. I0419 09:02:10.512880 133824151262976 async_checkpointer.py:283] [process=4][thread=save_finalize] No errors found in background save thread=async_save. I0419 09:02:10.512930 133824151262976 checkpoint_manager.py:2133] [process=4][thread=save_finalize][step=0] CheckpointManager Save Finalize is syncing with other hosts... I0419 09:02:10.514393 133824151262976 checkpoint_manager.py:2142] [process=4][thread=save_finalize][step=0] CheckpointManager Save Finalize is done on all hosts. I0419 09:02:10.514583 133951773706048 checkpoint_manager.py:2036] [process=4][thread=MainThread][step=0][wait_until_finished] Done waiting for Save Finalize thread (save_finalize) running at step=0. W0419 09:02:10.514725 133951773706048 checkpoint_manager.py:1458] Waiting for previous save to complete took 42.431223 seconds. If this number is high, consider checkpointing less frequently. I0419 09:02:10.517056 133951773706048 checkpoint_manager.py:1518] [process=4] Saving checkpoint at step 9 I0419 09:02:10.520518 133951773706048 async_checkpointer.py:452] [process=4] Started async saving checkpoint to gs://lance-maxtext/linen_ckpt_xpk_test_pipeline_scan_nnx_20260419_085454/linen_xpk_test_pipeline_scan_nnx_20260419_085454_02_synthetic/checkpoints/9. I0419 09:02:11.509487 133951773706048 jax_array_handlers.py:358] Scheduling D2H of 39 prioritized jax.Array. I0419 09:02:11.509585 133951773706048 replica_slices.py:424] 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 I0419 09:02:11.555293 133951773706048 base_pytree_checkpoint_handler.py:153] [process=4][thread=MainThread] Initiated "orbax.checkpoint._src.serialization.jax_array_handlers.ArrayHandler".serialize. Time taken: 0.047038s I0419 09:02:11.555468 133951773706048 base_pytree_checkpoint_handler.py:129] [process=4] /jax/checkpoint/write/blocking_gbytes_per_sec: 5.112 GiB/s (total gbytes: 1.5 GiB) (time elapsed: 0.3017435073852539 s) (per-host) I0419 09:02:11.555522 133951773706048 base_pytree_checkpoint_handler.py:737] [process=4][thread=MainThread] Initiated Pytree async_save. Time taken: 0.301808s (batch_requests_ready=0.252851s, total_serialization_initiated=0.048883s, others=0.000073s) I0419 09:02:11.555612 133951773706048 composite_checkpoint_handler.py:715] [process=4][thread=MainThread] Initiated CompositeCheckpointHandler.async_save. Time taken: 0.305738s (all_items=0.000016s, per_item={'items': '0.00001621'}, temp_paths=0.305722) I0419 09:02:11.556551 133824764512000 async_checkpointer.py:79] [process=4][thread=async_save] Background save thread started. I0419 09:02:11.556714 133951773706048 async_checkpointer.py:561] Finished blocking save. Time taken: 1.039585s. Continuing background save to gs://lance-maxtext/linen_ckpt_xpk_test_pipeline_scan_nnx_20260419_085454/linen_xpk_test_pipeline_scan_nnx_20260419_085454_02_synthetic/checkpoints/9. I0419 09:02:11.869408 133951773706048 checkpoint_manager.py:1566] [process=4][thread=MainThread][step=9] Starting CheckpointManager Save Finalize thread=save_finalize I0419 09:02:11.869797 133824151262976 async_checkpointer.py:265] [process=4][thread=save_finalize] Waiting for background save thread=async_save. I0419 09:02:11.869987 133951773706048 standard_logger.py:34] {'step': 9, 'event_type': 'save', 'directory': 'gs://lance-maxtext/linen_ckpt_xpk_test_pipeline_scan_nnx_20260419_085454/linen_xpk_test_pipeline_scan_nnx_20260419_085454_02_synthetic/checkpoints', 'reached_preemption': False, 'preemption_received_at': None, 'synchronous': False, 'wait_for_prev_start_time': 1776589288.083473, 'wait_for_prev_duration_secs': 42.431222915649414, 'time_between_consecutive_saves_sec': None, 'checkpointer_blocking_start_time': 1776589330.5170949, 'checkpointer_blocking_duration_secs': 1.0397250652313232, 'get_old_steps_start_time': 1776589331.5568392, 'get_old_steps_duration_secs': 2.5987625122070312e-05, 'checkpoint_manager_blocking_start_time': 1776589288.0815995, 'checkpoint_manager_blocking_duration_secs': 43.78833341598511} I0419 09:02:11.870101 133951773706048 checkpointing.py:408] Started an asynchronous checkpoint save for step 9 I0419 09:02:11.870148 133951773706048 checkpoint_manager.py:2024] [process=4][thread=MainThread][step=9][wait_until_finished] Waiting for Save Finalize thread (save_finalize) to complete. I0419 09:02:16.809836 133825397819136 array_metadata_store.py:203] [process=4][thread=array_type_handler] Wrote 39 array_metadata.ArrayMetadata to gs://lance-maxtext/linen_ckpt_xpk_test_pipeline_scan_nnx_20260419_085454/linen_xpk_test_pipeline_scan_nnx_20260419_085454_02_synthetic/checkpoints/9/items/array_metadatas/process_4 I0419 09:02:53.421021 133824764512000 base_pytree_checkpoint_handler.py:129] [process=4] /jax/checkpoint/write/gbytes_per_sec: 37.460 MiB/s (total gbytes: 1.5 GiB) (time elapsed: 42.167232513427734 s) (per-host) I0419 09:02:53.421136 133824764512000 async_checkpointer.py:90] [process=4][thread=async_save] 3 Handler Commit operations completed. Time taken: 41.864471s. I0419 09:03:01.871672 133824764512000 async_checkpointer.py:144] [process=4][thread=async_save] Background save thread done. Time taken: 50.314993s. I0419 09:03:01.872130 133824151262976 async_checkpointer.py:273] [process=4][thread=save_finalize] Done with waiting for background save thread=async_save. I0419 09:03:01.872261 133824151262976 async_checkpointer.py:283] [process=4][thread=save_finalize] No errors found in background save thread=async_save. I0419 09:03:01.872309 133824151262976 checkpoint_manager.py:2133] [process=4][thread=save_finalize][step=9] CheckpointManager Save Finalize is syncing with other hosts... I0419 09:03:01.873599 133824151262976 checkpoint_manager.py:2142] [process=4][thread=save_finalize][step=9] CheckpointManager Save Finalize is done on all hosts. I0419 09:03:01.873799 133951773706048 checkpoint_manager.py:2036] [process=4][thread=MainThread][step=9][wait_until_finished] Done waiting for Save Finalize thread (save_finalize) running at step=9. I0419 09:03:01.873960 133951773706048 checkpoint_manager.py:2013] [process=4][thread=MainThread][wait_until_finished] No Save Finalize thread to wait for. Returning. I0419 09:03:01.874579 133951773706048 metric_logger.py:185] completed step: 9, seconds: 0.147, TFLOP/s/device: 92.225, Tokens/s/device: 13901.144, total_weights: 65536, loss: 8.179 Per train step: Total TFLOPs: 13.59 split as 93.93% learnable weight flops and 6.07% attention flops XPK End: Sun Apr 19 09:03:12 UTC 2026 EXIT_CODE=0