main (574ad3fb9) vs feat/nnx-set-defaults-true (e27fc1e97)| Metric | main 574ad3fb9 | feat/nnx-set-defaults-true e27fc1e97 | Diff (feat/nnx-set-defaults-true − main) |
|---|---|---|---|
| Parameters | 1.104 billion | 1.104 billion | — |
| Final loss | 6.9810 | 6.9810 | 0 |
| TFLOP/s | 11.358 | 11.136 | -0.222 |
| Tok/s | 1815.2 | 1779.8 | -35.487 |
| Avg s/step | 1.393 | 1.253 | -0.14 |
| Memory % | 1.31 | 1.31 | 0 |
| JAX | 0.9.2 | 0.8.1 | — |
Diff = branch value − main value. Green = branch improved. Red = branch regressed.
XPK Start: Mon Apr 20 06:07:42 UTC 2026 PyTorch was not found. Models won't be available and only tokenizers, configuration and file/data utilities can be used. 2026-04-20 06:08:06.667903: 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) I0420 06:08:07.206369 132708396721984 max_utils.py:273] Attempting to initialize the jax distributed system... I0420 06:08:16.245934 132708396721984 distributed.py:149] Starting JAX distributed service on [::]:8482 I0420 06:08:16.248331 132708396721984 distributed.py:172] Connecting to JAX distributed service on mt-03-dropout-5xwjy-slice-job-0-0.mt-03-dropout-5xwjy:8482 I0420 06:08:17.784921 132708396721984 max_utils.py:284] Jax distributed system initialized! I0420 06:08:23.874588 132708396721984 max_utils.py:800] System Information: Jax Version: 0.9.2 I0420 06:08:23.874690 132708396721984 max_utils.py:801] System Information: Jaxlib Version: 0.9.2 I0420 06:08:23.874732 132708396721984 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 I0420 06:08:23.874766 132708396721984 train_utils.py:348] WARNING: Sequence packing is essentially ignored for synthetic data. Please use a real dataset to use sequence packing. I0420 06:08:24.574064 132708396721984 maxtext_utils.py:1551] Num_devices: 32, shape (1, 1, 1, 32, 1, 1, 1, 1, 1, 1, 1, 1, 1) I0420 06:08:24.574332 132708396721984 checkpointing.py:677] Setting up checkpoint logger... I0420 06:08:24.574384 132708396721984 checkpointing.py:233] Creating checkpoint manager with ocdbt=True and zarr3=True I0420 06:08:24.574432 132708396721984 pytree_checkpoint_handler.py:592] save_device_host_concurrent_bytes=None I0420 06:08:24.574772 132708396721984 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 0x78b200514e90>, enable_pinned_host_transfer=False, save_concurrent_bytes: 96000000000 (89.4 GiB), restore_concurrent_bytes: 96000000000 (89.4 GiB) I0420 06:08:27.859263 132708396721984 checkpointing.py:265] Enabling policy for fixed interval checkpointing. I0420 06:08:27.859499 132708396721984 checkpoint_manager.py:708] [process=2][thread=MainThread] CheckpointManager init: checkpointers=None, item_names=('items',), item_handlers={'items': <orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeCheckpointHandler object at 0x789d9c57f5f0>}, handler_registry=None I0420 06:08:27.859731 132708396721984 composite_checkpoint_handler.py:237] Deferred registration for item: "items". Adding handler `<orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeCheckpointHandler object at 0x789d9c57f5f0>` 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`. I0420 06:08:27.859780 132708396721984 composite_checkpoint_handler.py:237] Deferred registration for item: "metrics". Adding handler `<orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonCheckpointHandler object at 0x789d9c583cb0>` 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`. I0420 06:08:27.859813 132708396721984 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 0x789d9c57f5f0>, ('items', <class 'orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeRestoreArgs'>): <orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeCheckpointHandler object at 0x789d9c57f5f0>, ('metrics', <class 'orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonSaveArgs'>): <orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonCheckpointHandler object at 0x789d9c583cb0>, ('metrics', <class 'orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonRestoreArgs'>): <orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonCheckpointHandler object at 0x789d9c583cb0>}). I0420 06:08:27.860137 132708396721984 abstract_checkpointer.py:35] orbax-checkpoint version: 0.11.34 I0420 06:08:27.860207 132708396721984 async_checkpointer.py:192] [process=2][thread=MainThread] Using barrier_sync_fn: <function get_barrier_sync_fn.<locals>._fn at 0x789d9c1eb920> timeout: 1200 secs and primary_host=0 for async checkpoint writes I0420 06:08:29.099502 132708396721984 checkpoint_manager.py:1812] Found 0 checkpoint steps in gs://lance-maxtext/linen_ckpt_xpk_main_20260420_060028/linen_xpk_main_20260420_060028_03_dropout/checkpoints I0420 06:08:29.534219 132708396721984 checkpoint_manager.py:929] [process=2][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_20260420_060028/linen_xpk_main_20260420_060028_03_dropout/checkpoints: <orbax.checkpoint.checkpoint_manager.CheckpointManager object at 0x78b0d2de9700> I0420 06:08:29.534383 132708396721984 checkpointing.py:301] Checkpoint manager created! I0420 06:08:29.877303 132708396721984 nnx_wrappers.py:437] Unknown Logical: bfloat16[32,128,2048]....................................... ('activation_batch', 'activation_norm_length', 'activation_embed'). I0420 06:08:29.877417 132708396721984 nnx_wrappers.py:437] Unknown Physical: bfloat16[32,128,2048]....................................... ('fsdp', None, None). I0420 06:08:30.257851 132708396721984 attentions.py:1088] attentions/inputs_q Logical: bfloat16[32,128,2048]....................................... ('activation_batch', 'activation_attn_length', 'activation_attn_embed'). I0420 06:08:30.257941 132708396721984 attentions.py:1088] attentions/inputs_q Physical: bfloat16[32,128,2048]....................................... ('fsdp', None, None). I0420 06:08:30.274247 132708396721984 attentions.py:1089] attentions/inputs_kv Logical: bfloat16[32,128,2048]....................................... ('activation_batch', 'activation_attn_length', 'activation_attn_embed'). I0420 06:08:30.274307 132708396721984 attentions.py:1089] attentions/inputs_kv Physical: bfloat16[32,128,2048]....................................... ('fsdp', None, None). I0420 06:08:30.298028 132708396721984 attentions.py:1154] attentions/query Logical: bfloat16[32,128,16,128]..................................... ('activation_kv_batch', 'activation_attn_length', 'activation_kv_heads', 'activation_kv_head_dim'). I0420 06:08:30.298094 132708396721984 attentions.py:1154] attentions/query Physical: bfloat16[32,128,16,128]..................................... ('fsdp', None, None, None). I0420 06:08:30.314486 132708396721984 attentions.py:1155] attentions/key Logical: bfloat16[32,128,16,128]..................................... ('activation_kv_batch', 'activation_attn_length', 'activation_kv_heads', 'activation_kv_head_dim'). I0420 06:08:30.314544 132708396721984 attentions.py:1155] attentions/key Physical: bfloat16[32,128,16,128]..................................... ('fsdp', None, None, None). I0420 06:08:30.330896 132708396721984 attentions.py:1156] attentions/value Logical: bfloat16[32,128,16,128]..................................... ('activation_kv_batch', 'activation_attn_length', 'activation_kv_heads', 'activation_kv_head_dim'). I0420 06:08:30.330957 132708396721984 attentions.py:1156] attentions/value Physical: bfloat16[32,128,16,128]..................................... ('fsdp', None, None, None). I0420 06:08:30.356551 132708396721984 attentions.py:1197] attentions/out Logical: bfloat16[32,128,16,128]..................................... ('activation_batch', 'activation_attn_length', 'activation_heads', 'activation_kv'). I0420 06:08:30.356621 132708396721984 attentions.py:1197] attentions/out Physical: bfloat16[32,128,16,128]..................................... ('fsdp', None, None, None). I0420 06:08:30.382763 132708396721984 linears.py:525] linears/x Logical: bfloat16[32,128,7168]....................................... ('activation_batch', 'activation_length', 'activation_mlp'). I0420 06:08:30.382828 132708396721984 linears.py:525] linears/x Physical: bfloat16[32,128,7168]....................................... ('fsdp', None, None). I0420 06:08:30.611488 132708396721984 checkpointing.py:577] checkpoint manager exists so trying to load this run's existing checkpoint I0420 06:08:30.611595 132708396721984 checkpointing.py:665] No existing checkpoints found, not restoring checkpoint. fsdp: 32 I0420 06:08:32.048372 132708396721984 maxtext_utils.py:1654] params/params/decoder/decoder_norm/scale Shape: float32[2048] Logical: P('norm',) Physical: (None,) I0420 06:08:32.048504 132708396721984 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) I0420 06:08:32.048557 132708396721984 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) I0420 06:08:32.048615 132708396721984 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') I0420 06:08:32.048667 132708396721984 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) I0420 06:08:32.048712 132708396721984 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) I0420 06:08:32.048766 132708396721984 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) I0420 06:08:32.048818 132708396721984 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') I0420 06:08:32.048858 132708396721984 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) I0420 06:08:32.048896 132708396721984 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) I0420 06:08:32.048945 132708396721984 maxtext_utils.py:1654] params/params/decoder/logits_dense/kernel Shape: float32[2048,32000] Logical: P('embed_vocab', 'vocab') Physical: ('fsdp', None) I0420 06:08:32.048996 132708396721984 maxtext_utils.py:1654] params/params/token_embedder/embedding Shape: float32[32000,2048] Logical: P('vocab', 'embed_vocab') Physical: (None, 'fsdp') I0420 06:08:32.571006 132708396721984 train.py:155] train/xent Logical: float32[32,128]............................................. ('activation_embed_and_logits_batch', 'activation_length'). I0420 06:08:32.571112 132708396721984 train.py:155] train/xent Physical: float32[32,128]............................................. ('fsdp', None). I0420 06:08:32.586625 132708396721984 train.py:162] train/z_loss Logical: float32[32,128]............................................. ('activation_embed_and_logits_batch', 'activation_length'). I0420 06:08:32.586687 132708396721984 train.py:162] train/z_loss Physical: float32[32,128]............................................. ('fsdp', None). I0420 06:08:36.146428 132708396721984 max_utils.py:791] Total memory size: 0.8 GB, Output size: 0.4 GB, Temp size: 0.4 GB, Argument size: 0.4 GB, Host temp size: 0.0 GB. I0420 06:08:36.147165 132708396721984 metric_logger.py:301] number parameters: 1.104 billion I0420 06:08:40.228787 132708396721984 checkpointing.py:772] Waiting for step 0 to finish before checkpoint... I0420 06:08:40.836998 132708396721984 checkpointing.py:776] Waited 0.6081898212432861 seconds for step 0 to finish before starting checkpointing. I0420 06:08:40.840556 132708396721984 checkpoint_manager.py:2009] [process=2][thread=MainThread][wait_until_finished] No Save Finalize thread to wait for. Returning. I0420 06:08:40.842542 132708396721984 checkpoint_manager.py:1512] [process=2] Saving checkpoint at step 0 I0420 06:08:40.843913 132708396721984 event_tracking.py:70] [process=2] [async] Started save checkpoint @ gs://lance-maxtext/linen_ckpt_xpk_main_20260420_060028/linen_xpk_main_20260420_060028_03_dropout/checkpoints/0. I0420 06:08:41.562984 132708396721984 signaling_client.py:364] Using JaxDistributedSignalingClient I0420 06:08:41.563925 132708396721984 jax_array_handlers.py:360] Scheduling D2H of 39 prioritized jax.Array. I0420 06:08:41.563982 132708396721984 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 I0420 06:08:41.843361 132708396721984 base_pytree_checkpoint_handler.py:154] [process=2][thread=MainThread] Initiated "orbax.checkpoint._src.serialization.jax_array_handlers.ArrayHandler".serialize. Time taken: 0.280447s I0420 06:08:41.843551 132708396721984 base_pytree_checkpoint_handler.py:130] [process=2] /jax/orbax/write/blocking_gbytes_per_sec: 5.397 GiB/s (total gbytes: 1.5 GiB) (time elapsed: 0.28581738471984863 s) (per-host) I0420 06:08:41.843626 132708396721984 base_pytree_checkpoint_handler.py:768] [process=2][thread=MainThread] Initiated Pytree async_save. Time taken: 0.285903s (batch_requests_ready=0.002206s, total_serialization_initiated=0.283604s, others=0.000094s) I0420 06:08:41.843758 132708396721984 composite_checkpoint_handler.py:715] [process=2][thread=MainThread] Initiated CompositeCheckpointHandler.async_save. Time taken: 0.289957s (all_items=0.000018s, per_item={'items': '0.00001836'}, temp_paths=0.289938) I0420 06:08:41.844596 132708396721984 event_tracking.py:125] [process=2] [async] Finished blocking save in 1.00 seconds. Continuing save @ gs://lance-maxtext/linen_ckpt_xpk_main_20260420_060028/linen_xpk_main_20260420_060028_03_dropout/checkpoints/0. I0420 06:08:41.844948 132579785168640 async_checkpointer.py:76] [process=2][thread=async_save] Background save thread started. Deadline for this save operation is 2026-04-20 06:28:41.844910 I0420 06:08:41.856558 132708396721984 checkpoint_manager.py:1560] [process=2][thread=MainThread][step=0] Starting CheckpointManager Save Finalize thread=save_finalize I0420 06:08:41.856824 132578673669888 async_checkpointer.py:280] [process=2][thread=save_finalize] Waiting for background save thread=async_save. I0420 06:08:41.856980 132708396721984 standard_logger.py:34] {'step': 0, 'event_type': 'save', 'directory': 'gs://lance-maxtext/linen_ckpt_xpk_main_20260420_060028/linen_xpk_main_20260420_060028_03_dropout/checkpoints', 'reached_preemption': False, 'preemption_received_at': None, 'synchronous': False, 'wait_for_prev_start_time': 1776665320.8405373, 'wait_for_prev_duration_secs': 5.9604644775390625e-05, 'time_between_consecutive_saves_sec': None, 'checkpointer_blocking_start_time': 1776665320.8425808, 'checkpointer_blocking_duration_secs': 1.0025408267974854, 'get_old_steps_start_time': 1776665321.8451452, 'get_old_steps_duration_secs': 3.1948089599609375e-05, 'checkpoint_manager_blocking_start_time': 1776665320.8375306, 'checkpoint_manager_blocking_duration_secs': 1.0194108486175537} I0420 06:08:41.857108 132708396721984 checkpointing.py:408] Started an asynchronous checkpoint save for step 0 I0420 06:08:41.857162 132708396721984 max_utils.py:750] Memstats: After params initialized: I0420 06:08:41.857209 132708396721984 max_utils.py:756] Using (GB) 0.41 / 31.25 (1.312000%) on TPU_8(process=2,(0,2,0,0)) I0420 06:08:41.857241 132708396721984 max_utils.py:756] Using (GB) 0.41 / 31.25 (1.312000%) on TPU_9(process=2,(1,2,0,0)) I0420 06:08:41.857267 132708396721984 max_utils.py:756] Using (GB) 0.41 / 31.25 (1.312000%) on TPU_12(process=2,(0,3,0,0)) I0420 06:08:41.857291 132708396721984 max_utils.py:756] Using (GB) 0.41 / 31.25 (1.312000%) on TPU_13(process=2,(1,3,0,0)) I0420 06:08:42.173907 132708396721984 metric_logger.py:196] completed step: 0, seconds: 4.082, TFLOP/s/device: 0.196, Tokens/s/device: 31.361, total_weights: 4096, loss: 10.889, lm_loss: 10.889, perplexity: 53560.699 I0420 06:08:42.265519 132708396721984 metric_logger.py:196] completed step: 1, seconds: 1.944, TFLOP/s/device: 0.412, Tokens/s/device: 65.857, total_weights: 4096, loss: 10.902, lm_loss: 10.902, perplexity: 54298.523 I0420 06:08:42.671440 132708396721984 metric_logger.py:196] completed step: 2, seconds: 0.021, TFLOP/s/device: 38.132, Tokens/s/device: 6094.367, total_weights: 4096, loss: 9.956, lm_loss: 9.956, perplexity: 21082.051 I0420 06:08:42.743266 132708396721984 metric_logger.py:196] completed step: 3, seconds: 0.407, TFLOP/s/device: 1.970, Tokens/s/device: 314.783, total_weights: 4096, loss: 9.139, lm_loss: 9.139, perplexity: 9314.691 I0420 06:08:42.887635 132708396721984 metric_logger.py:196] completed step: 4, seconds: 0.076, TFLOP/s/device: 10.524, Tokens/s/device: 1682.042, total_weights: 4096, loss: 8.459, lm_loss: 8.459, perplexity: 4715.493 I0420 06:08:42.893641 132708396721984 metric_logger.py:196] completed step: 5, seconds: 0.072, TFLOP/s/device: 11.201, Tokens/s/device: 1790.135, total_weights: 4096, loss: 7.924, lm_loss: 7.924, perplexity: 2762.373 I0420 06:08:45.779687 2852 google_auth_provider.cc:181] Running on GCE, using service account 562977990677-compute@developer.gserviceaccount.com I0420 06:08:48.388337 132578721478400 array_metadata_store.py:203] [process=2][thread=array_type_handler] Wrote 39 array_metadata.ArrayMetadata to gs://lance-maxtext/linen_ckpt_xpk_main_20260420_060028/linen_xpk_main_20260420_060028_03_dropout/checkpoints/0/items/array_metadatas/process_2 I0420 06:08:52.687745 132708396721984 metric_logger.py:196] completed step: 6, seconds: 0.145, TFLOP/s/device: 5.524, Tokens/s/device: 882.862, total_weights: 4096, loss: 7.553, lm_loss: 7.553, perplexity: 1906.617 I0420 06:08:52.758408 132708396721984 metric_logger.py:196] completed step: 7, seconds: 9.725, TFLOP/s/device: 0.082, Tokens/s/device: 13.162, total_weights: 4096, loss: 7.267, lm_loss: 7.267, perplexity: 1431.653 I0420 06:08:52.829071 132708396721984 metric_logger.py:196] completed step: 8, seconds: 0.075, TFLOP/s/device: 10.685, Tokens/s/device: 1707.805, total_weights: 4096, loss: 7.111, lm_loss: 7.111, perplexity: 1225.983 I0420 06:08:52.899036 132708396721984 checkpointing.py:772] Waiting for step 9 to finish before checkpoint... I0420 06:08:52.899688 132708396721984 checkpointing.py:776] Waited 0.0006897449493408203 seconds for step 9 to finish before starting checkpointing. I0420 06:08:52.901739 132708396721984 checkpoint_manager.py:2020] [process=2][thread=MainThread][step=0][wait_until_finished] Waiting for Save Finalize thread (save_finalize) to complete. I0420 06:09:15.963989 132579785168640 base_pytree_checkpoint_handler.py:130] [process=2] /jax/orbax/write/gbytes_per_sec: 45.910 MiB/s (total gbytes: 1.5 GiB) (time elapsed: 34.406221866607666 s) (per-host) I0420 06:09:15.964116 132579785168640 async_checkpointer.py:90] [process=2][thread=async_save] 3 Handler Commit operations completed. Time taken: 34.119034s. I0420 06:09:23.718404 132579785168640 async_checkpointer.py:160] [process=2][thread=async_save] Background save thread done. Time taken: 41.873305s. I0420 06:09:23.718705 132578673669888 async_checkpointer.py:288] [process=2][thread=save_finalize] Done with waiting for background save thread=async_save. I0420 06:09:23.718827 132578673669888 async_checkpointer.py:298] [process=2][thread=save_finalize] No errors found in background save thread=async_save. I0420 06:09:23.718876 132578673669888 checkpoint_manager.py:2137] [process=2][thread=save_finalize][step=0] CheckpointManager Save Finalize is syncing with other hosts... I0420 06:09:23.720355 132578673669888 checkpoint_manager.py:2146] [process=2][thread=save_finalize][step=0] CheckpointManager Save Finalize is done on all hosts. I0420 06:09:23.720535 132708396721984 checkpoint_manager.py:2032] [process=2][thread=MainThread][step=0][wait_until_finished] Done waiting for Save Finalize thread (save_finalize) running at step=0. W0420 06:09:23.720674 132708396721984 checkpoint_manager.py:1452] Waiting for previous save to complete took 30.818937 seconds. If this number is high, consider checkpointing less frequently. I0420 06:09:23.722307 132708396721984 checkpoint_manager.py:1512] [process=2] Saving checkpoint at step 9 I0420 06:09:23.724368 132708396721984 event_tracking.py:70] [process=2] [async] Started save checkpoint @ gs://lance-maxtext/linen_ckpt_xpk_main_20260420_060028/linen_xpk_main_20260420_060028_03_dropout/checkpoints/9. I0420 06:09:24.428623 132708396721984 jax_array_handlers.py:360] Scheduling D2H of 39 prioritized jax.Array. I0420 06:09:24.428717 132708396721984 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 I0420 06:09:24.468863 132708396721984 base_pytree_checkpoint_handler.py:154] [process=2][thread=MainThread] Initiated "orbax.checkpoint._src.serialization.jax_array_handlers.ArrayHandler".serialize. Time taken: 0.041515s I0420 06:09:24.469002 132708396721984 base_pytree_checkpoint_handler.py:130] [process=2] /jax/orbax/write/blocking_gbytes_per_sec: 34.368 GiB/s (total gbytes: 1.5 GiB) (time elapsed: 0.04488372802734375 s) (per-host) I0420 06:09:24.469067 132708396721984 base_pytree_checkpoint_handler.py:768] [process=2][thread=MainThread] Initiated Pytree async_save. Time taken: 0.044957s (batch_requests_ready=0.001728s, total_serialization_initiated=0.043151s, others=0.000079s) I0420 06:09:24.469148 132708396721984 composite_checkpoint_handler.py:715] [process=2][thread=MainThread] Initiated CompositeCheckpointHandler.async_save. Time taken: 0.049599s (all_items=0.000033s, per_item={'items': '0.00003338'}, temp_paths=0.049566) I0420 06:09:24.469837 132708396721984 event_tracking.py:125] [process=2] [async] Finished blocking save in 0.75 seconds. Continuing save @ gs://lance-maxtext/linen_ckpt_xpk_main_20260420_060028/linen_xpk_main_20260420_060028_03_dropout/checkpoints/9. I0420 06:09:24.470126 132578673669888 async_checkpointer.py:76] [process=2][thread=async_save] Background save thread started. Deadline for this save operation is 2026-04-20 06:29:24.470094 I0420 06:09:24.473672 132708396721984 checkpoint_manager.py:1560] [process=2][thread=MainThread][step=9] Starting CheckpointManager Save Finalize thread=save_finalize I0420 06:09:24.473893 132578721478400 async_checkpointer.py:280] [process=2][thread=save_finalize] Waiting for background save thread=async_save. I0420 06:09:24.474042 132708396721984 standard_logger.py:34] {'step': 9, 'event_type': 'save', 'directory': 'gs://lance-maxtext/linen_ckpt_xpk_main_20260420_060028/linen_xpk_main_20260420_060028_03_dropout/checkpoints', 'reached_preemption': False, 'preemption_received_at': None, 'synchronous': False, 'wait_for_prev_start_time': 1776665332.9017081, 'wait_for_prev_duration_secs': 30.818936824798584, 'time_between_consecutive_saves_sec': None, 'checkpointer_blocking_start_time': 1776665363.7223446, 'checkpointer_blocking_duration_secs': 0.747920036315918, 'get_old_steps_start_time': 1776665364.4702892, 'get_old_steps_duration_secs': 2.956390380859375e-05, 'checkpoint_manager_blocking_start_time': 1776665332.8999114, 'checkpoint_manager_blocking_duration_secs': 31.574082612991333} I0420 06:09:24.474161 132708396721984 checkpointing.py:408] Started an asynchronous checkpoint save for step 9 I0420 06:09:24.474205 132708396721984 checkpoint_manager.py:2020] [process=2][thread=MainThread][step=9][wait_until_finished] Waiting for Save Finalize thread (save_finalize) to complete. I0420 06:09:30.352488 132574912833280 array_metadata_store.py:203] [process=2][thread=array_type_handler] Wrote 39 array_metadata.ArrayMetadata to gs://lance-maxtext/linen_ckpt_xpk_main_20260420_060028/linen_xpk_main_20260420_060028_03_dropout/checkpoints/9/items/array_metadatas/process_2 I0420 06:10:05.625214 132578673669888 base_pytree_checkpoint_handler.py:130] [process=2] /jax/orbax/write/gbytes_per_sec: 38.339 MiB/s (total gbytes: 1.5 GiB) (time elapsed: 41.20104622840881 s) (per-host) I0420 06:10:05.625337 132578673669888 async_checkpointer.py:90] [process=2][thread=async_save] 3 Handler Commit operations completed. Time taken: 41.155102s. I0420 06:10:14.011339 132578673669888 async_checkpointer.py:160] [process=2][thread=async_save] Background save thread done. Time taken: 49.541089s. I0420 06:10:14.011612 132578721478400 async_checkpointer.py:288] [process=2][thread=save_finalize] Done with waiting for background save thread=async_save. I0420 06:10:14.011727 132578721478400 async_checkpointer.py:298] [process=2][thread=save_finalize] No errors found in background save thread=async_save. I0420 06:10:14.011772 132578721478400 checkpoint_manager.py:2137] [process=2][thread=save_finalize][step=9] CheckpointManager Save Finalize is syncing with other hosts... I0420 06:10:14.014469 132578721478400 checkpoint_manager.py:2146] [process=2][thread=save_finalize][step=9] CheckpointManager Save Finalize is done on all hosts. I0420 06:10:14.014635 132708396721984 checkpoint_manager.py:2032] [process=2][thread=MainThread][step=9][wait_until_finished] Done waiting for Save Finalize thread (save_finalize) running at step=9. I0420 06:10:14.014801 132708396721984 checkpoint_manager.py:2009] [process=2][thread=MainThread][wait_until_finished] No Save Finalize thread to wait for. Returning. I0420 06:10:14.015815 132708396721984 metric_logger.py:196] completed step: 9, seconds: 0.071, TFLOP/s/device: 11.358, Tokens/s/device: 1815.242, total_weights: 4096, loss: 6.981, lm_loss: 6.981, perplexity: 1075.867 Per train step: Total TFLOPs: 0.80 split as 99.60% learnable weight flops and 0.40% attention flops XPK End: Mon Apr 20 06:10:22 UTC 2026 EXIT_CODE=0
XPK Start: Mon Apr 20 19:20:14 UTC 2026 2026-04-20 19:20:18.683755: 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 00:00:1776712818.696868 10 cuda_dnn.cc:8579] Unable to register cuDNN factory: Attempting to register factory for plugin cuDNN when one has already been registered E0000 00:00:1776712818.700561 10 cuda_blas.cc:1407] Unable to register cuBLAS factory: Attempting to register factory for plugin cuBLAS when one has already been registered W0000 00:00:1776712818.712279 10 computation_placer.cc:177] computation placer already registered. Please check linkage and avoid linking the same target more than once. W0000 00:00:1776712818.712297 10 computation_placer.cc:177] computation placer already registered. Please check linkage and avoid linking the same target more than once. W0000 00:00:1776712818.712299 10 computation_placer.cc:177] computation placer already registered. Please check linkage and avoid linking the same target more than once. W0000 00:00:1776712818.712301 10 computation_placer.cc:177] computation placer already registered. Please check linkage and avoid linking the same target more than once. 2026-04-20 19:20:37.765119: 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) I0420 19:20:38.291322 137996531795776 max_utils.py:273] Attempting to initialize the jax distributed system... INFO:2026-04-20 19:20:47,332:jax._src.distributed:140: Starting JAX distributed service on [::]:8482 I0420 19:20:47.332234 137996531795776 distributed.py:140] Starting JAX distributed service on [::]:8482 INFO:2026-04-20 19:20:47,334:jax._src.distributed:157: Connecting to JAX distributed service on mt-03-dropout-fmbma-slice-job-0-0.mt-03-dropout-fmbma:8482 I0420 19:20:47.334561 137996531795776 distributed.py:157] Connecting to JAX distributed service on mt-03-dropout-fmbma-slice-job-0-0.mt-03-dropout-fmbma:8482 I0420 19:20:48.989691 137996531795776 max_utils.py:284] Jax distributed system initialized! I0420 19:20:55.952999 137996531795776 max_utils.py:800] System Information: Jax Version: 0.8.1 I0420 19:20:55.953120 137996531795776 max_utils.py:801] System Information: Jaxlib Version: 0.8.1 I0420 19:20:55.953164 137996531795776 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 I0420 19:20:55.953200 137996531795776 train_utils.py:377] WARNING: Sequence packing is essentially ignored for synthetic data. Please use a real dataset to use sequence packing. I0420 19:20:56.563912 137996531795776 maxtext_utils.py:1631] Num_devices: 32, shape (1, 1, 1, 32, 1, 1, 1, 1, 1, 1, 1, 1, 1) I0420 19:20:56.564527 137996531795776 maxtext_utils.py:1631] Num_devices: 32, shape (1, 1, 1, 32, 1, 1, 1, 1, 1, 1, 1, 1, 1) I0420 19:20:56.564714 137996531795776 checkpointing.py:688] Setting up checkpoint logger... I0420 19:20:56.564765 137996531795776 checkpointing.py:234] Creating checkpoint manager with ocdbt=True and zarr3=True I0420 19:20:56.564813 137996531795776 pytree_checkpoint_handler.py:589] save_device_host_concurrent_bytes=None I0420 19:20:56.565191 137996531795776 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 0x7d814676d310>, enable_pinned_host_transfer=False, save_concurrent_bytes: 96000000000 (89.4 GiB), restore_concurrent_bytes: 96000000000 (89.4 GiB) I0420 19:20:59.556548 137996531795776 checkpointing.py:266] Enabling policy for fixed interval checkpointing. I0420 19:20:59.556902 137996531795776 checkpoint_manager.py:709] [process=0][thread=MainThread] CheckpointManager init: checkpointers=None, item_names=('items',), item_handlers={'items': <orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeCheckpointHandler object at 0x7d6c987dd9a0>}, handler_registry=None I0420 19:20:59.557153 137996531795776 composite_checkpoint_handler.py:237] Deferred registration for item: "items". Adding handler `<orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeCheckpointHandler object at 0x7d6c987dd9a0>` 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`. I0420 19:20:59.557207 137996531795776 composite_checkpoint_handler.py:237] Deferred registration for item: "metrics". Adding handler `<orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonCheckpointHandler object at 0x7d6c987e2d80>` 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`. I0420 19:20:59.557253 137996531795776 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 0x7d6c987dd9a0>, ('items', <class 'orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeRestoreArgs'>): <orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeCheckpointHandler object at 0x7d6c987dd9a0>, ('metrics', <class 'orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonSaveArgs'>): <orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonCheckpointHandler object at 0x7d6c987e2d80>, ('metrics', <class 'orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonRestoreArgs'>): <orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonCheckpointHandler object at 0x7d6c987e2d80>}). I0420 19:20:59.557830 137996531795776 abstract_checkpointer.py:35] orbax-checkpoint version: 0.11.33 I0420 19:20:59.557943 137996531795776 async_checkpointer.py:177] [process=0][thread=MainThread] Using barrier_sync_fn: <function get_barrier_sync_fn.<locals>._fn at 0x7d6c983b8180> timeout: 600 secs and primary_host=0 for async checkpoint writes I0420 19:21:00.889772 137996531795776 checkpoint_manager.py:1818] Found 0 checkpoint steps in gs://lance-maxtext/linen_ckpt_xpk_feat_nnx_set_defaults_true_20260420_190404/linen_xpk_feat_nnx_set_defaults_true_20260420_190404_03_dropout/checkpoints I0420 19:21:00.892038 137996531795776 checkpoint_manager.py:929] [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_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_feat_nnx_set_defaults_true_20260420_190404/linen_xpk_feat_nnx_set_defaults_true_20260420_190404_03_dropout/checkpoints: <orbax.checkpoint.checkpoint_manager.CheckpointManager object at 0x7d6c987dce90> I0420 19:21:00.892172 137996531795776 checkpointing.py:302] Checkpoint manager created! I0420 19:21:01.194626 137996531795776 nnx_wrappers.py:455] Unknown Logical: bfloat16[32,128,2048]....................................... ('activation_batch', 'activation_norm_length', 'activation_embed'). I0420 19:21:01.194731 137996531795776 nnx_wrappers.py:455] Unknown Physical: bfloat16[32,128,2048]....................................... ('fsdp', None, None). I0420 19:21:01.564372 137996531795776 attentions.py:1084] attentions/inputs_q Logical: bfloat16[32,128,2048]....................................... ('activation_batch', 'activation_attn_length', 'activation_attn_embed'). I0420 19:21:01.564466 137996531795776 attentions.py:1084] attentions/inputs_q Physical: bfloat16[32,128,2048]....................................... ('fsdp', None, None). I0420 19:21:01.580356 137996531795776 attentions.py:1085] attentions/inputs_kv Logical: bfloat16[32,128,2048]....................................... ('activation_batch', 'activation_attn_length', 'activation_attn_embed'). I0420 19:21:01.580419 137996531795776 attentions.py:1085] attentions/inputs_kv Physical: bfloat16[32,128,2048]....................................... ('fsdp', None, None). I0420 19:21:01.603214 137996531795776 attentions.py:1150] attentions/query Logical: bfloat16[32,128,16,128]..................................... ('activation_kv_batch', 'activation_attn_length', 'activation_kv_heads', 'activation_kv_head_dim'). I0420 19:21:01.603284 137996531795776 attentions.py:1150] attentions/query Physical: bfloat16[32,128,16,128]..................................... ('fsdp', None, None, None). I0420 19:21:01.619216 137996531795776 attentions.py:1151] attentions/key Logical: bfloat16[32,128,16,128]..................................... ('activation_kv_batch', 'activation_attn_length', 'activation_kv_heads', 'activation_kv_head_dim'). I0420 19:21:01.619282 137996531795776 attentions.py:1151] attentions/key Physical: bfloat16[32,128,16,128]..................................... ('fsdp', None, None, None). I0420 19:21:01.635229 137996531795776 attentions.py:1152] attentions/value Logical: bfloat16[32,128,16,128]..................................... ('activation_kv_batch', 'activation_attn_length', 'activation_kv_heads', 'activation_kv_head_dim'). I0420 19:21:01.635298 137996531795776 attentions.py:1152] attentions/value Physical: bfloat16[32,128,16,128]..................................... ('fsdp', None, None, None). I0420 19:21:01.659983 137996531795776 attentions.py:1193] attentions/out Logical: bfloat16[32,128,16,128]..................................... ('activation_batch', 'activation_attn_length', 'activation_heads', 'activation_kv'). I0420 19:21:01.660054 137996531795776 attentions.py:1193] attentions/out Physical: bfloat16[32,128,16,128]..................................... ('fsdp', None, None, None). I0420 19:21:01.685983 137996531795776 linears.py:541] linears/x Logical: bfloat16[32,128,7168]....................................... ('activation_batch', 'activation_length', 'activation_mlp'). I0420 19:21:01.686058 137996531795776 linears.py:541] linears/x Physical: bfloat16[32,128,7168]....................................... ('fsdp', None, None). I0420 19:21:01.885096 137996531795776 checkpointing.py:578] checkpoint manager exists so trying to load this run's existing checkpoint I0420 19:21:01.885205 137996531795776 checkpointing.py:676] No existing checkpoints found, not restoring checkpoint. fsdp: 32 I0420 19:21:03.345872 137996531795776 maxtext_utils.py:1740] params/params/decoder/decoder_norm/scale Shape: float32[2048] Logical: PartitionSpec('norm',) Physical: (None,) I0420 19:21:03.346001 137996531795776 maxtext_utils.py:1740] params/params/decoder/layers/mlp/wi_0/kernel Shape: float32[2048,16,7168] Logical: PartitionSpec('embed', 'layers', 'mlp') Physical: ('fsdp', None, None) I0420 19:21:03.346058 137996531795776 maxtext_utils.py:1740] params/params/decoder/layers/mlp/wi_1/kernel Shape: float32[2048,16,7168] Logical: PartitionSpec('embed', 'layers', 'mlp') Physical: ('fsdp', None, None) I0420 19:21:03.346130 137996531795776 maxtext_utils.py:1740] params/params/decoder/layers/mlp/wo/kernel Shape: float32[7168,16,2048] Logical: PartitionSpec('mlp', 'layers', 'embed') Physical: (None, None, 'fsdp') I0420 19:21:03.346184 137996531795776 maxtext_utils.py:1740] params/params/decoder/layers/post_self_attention_layer_norm/scale Shape: float32[2048,16] Logical: PartitionSpec('norm', 'layers') Physical: (None, None) I0420 19:21:03.346223 137996531795776 maxtext_utils.py:1740] params/params/decoder/layers/pre_self_attention_layer_norm/scale Shape: float32[2048,16] Logical: PartitionSpec('norm', 'layers') Physical: (None, None) I0420 19:21:03.346275 137996531795776 maxtext_utils.py:1740] 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) I0420 19:21:03.346328 137996531795776 maxtext_utils.py:1740] 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') I0420 19:21:03.346368 137996531795776 maxtext_utils.py:1740] 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) I0420 19:21:03.346407 137996531795776 maxtext_utils.py:1740] 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) I0420 19:21:03.346456 137996531795776 maxtext_utils.py:1740] params/params/decoder/logits_dense/kernel Shape: float32[2048,32000] Logical: PartitionSpec('embed', 'vocab') Physical: ('fsdp', None) I0420 19:21:03.346503 137996531795776 maxtext_utils.py:1740] params/params/token_embedder/embedding Shape: float32[32000,2048] Logical: PartitionSpec('vocab', 'embed') Physical: (None, 'fsdp') I0420 19:21:03.869573 137996531795776 train.py:158] train/xent Logical: float32[32,128]............................................. ('activation_embed_and_logits_batch', 'activation_length'). I0420 19:21:03.869678 137996531795776 train.py:158] train/xent Physical: float32[32,128]............................................. ('fsdp', None). I0420 19:21:03.884564 137996531795776 train.py:165] train/z_loss Logical: float32[32,128]............................................. ('activation_embed_and_logits_batch', 'activation_length'). I0420 19:21:03.884629 137996531795776 train.py:165] train/z_loss Physical: float32[32,128]............................................. ('fsdp', None). I0420 19:21:07.255343 137996531795776 max_utils.py:791] Total memory size: 0.8 GB, Output size: 0.4 GB, Temp size: 0.4 GB, Argument size: 0.4 GB, Host temp size: 0.0 GB. I0420 19:21:07.261524 137996531795776 metric_logger.py:289] number parameters: 1.104 billion I0420 19:21:11.420972 137996531795776 checkpointing.py:794] Waiting for step 0 to finish before checkpoint... I0420 19:21:11.495759 137996531795776 checkpointing.py:798] Waited 0.07477617263793945 seconds for step 0 to finish before starting checkpointing. I0420 19:21:11.498461 137996531795776 checkpoint_manager.py:2013] [process=0][thread=MainThread][wait_until_finished] No Save Finalize thread to wait for. Returning. I0420 19:21:11.500478 137996531795776 checkpoint_manager.py:1518] [process=0] Saving checkpoint at step 0 I0420 19:21:11.503259 137996531795776 async_checkpointer.py:452] [process=0] Started async saving checkpoint to gs://lance-maxtext/linen_ckpt_xpk_feat_nnx_set_defaults_true_20260420_190404/linen_xpk_feat_nnx_set_defaults_true_20260420_190404_03_dropout/checkpoints/0. I0420 19:21:12.037918 137996531795776 signaling_client.py:364] Using JaxDistributedSignalingClient I0420 19:21:12.292131 137868626347776 atomicity.py:137] Creating tmp directory gs://lance-maxtext/linen_ckpt_xpk_feat_nnx_set_defaults_true_20260420_190404/linen_xpk_feat_nnx_set_defaults_true_20260420_190404_03_dropout/checkpoints/0 I0420 19:21:12.313281 137996531795776 jax_array_handlers.py:358] Scheduling D2H of 39 prioritized jax.Array. I0420 19:21:12.313380 137996531795776 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 I0420 19:21:12.554641 137996531795776 base_pytree_checkpoint_handler.py:153] [process=0][thread=MainThread] Initiated "orbax.checkpoint._src.serialization.jax_array_handlers.ArrayHandler".serialize. Time taken: 0.242380s I0420 19:21:12.555428 137996531795776 base_pytree_checkpoint_handler.py:129] [process=0] /jax/checkpoint/write/blocking_gbytes_per_sec: 3.030 GiB/s (total gbytes: 1.5 GiB) (time elapsed: 0.509129524230957 s) (per-host) I0420 19:21:12.555504 137996531795776 base_pytree_checkpoint_handler.py:737] [process=0][thread=MainThread] Initiated Pytree async_save. Time taken: 0.509225s (batch_requests_ready=0.261855s, total_serialization_initiated=0.246668s, others=0.000702s) I0420 19:21:12.555586 137996531795776 composite_checkpoint_handler.py:715] [process=0][thread=MainThread] Initiated CompositeCheckpointHandler.async_save. Time taken: 0.513399s (all_items=0.000019s, per_item={'items': '0.00001884'}, temp_paths=0.513380) I0420 19:21:12.556527 137869695903488 async_checkpointer.py:79] [process=0][thread=async_save] Background save thread started. I0420 19:21:12.556688 137996531795776 async_checkpointer.py:561] Finished blocking save. Time taken: 1.056142s. Continuing background save to gs://lance-maxtext/linen_ckpt_xpk_feat_nnx_set_defaults_true_20260420_190404/linen_xpk_feat_nnx_set_defaults_true_20260420_190404_03_dropout/checkpoints/0. I0420 19:21:12.570755 137996531795776 checkpoint_manager.py:1566] [process=0][thread=MainThread][step=0] Starting CheckpointManager Save Finalize thread=save_finalize I0420 19:21:12.571068 137868617955072 async_checkpointer.py:265] [process=0][thread=save_finalize] Waiting for background save thread=async_save. I0420 19:21:12.571237 137996531795776 standard_logger.py:34] {'step': 0, 'event_type': 'save', 'directory': 'gs://lance-maxtext/linen_ckpt_xpk_feat_nnx_set_defaults_true_20260420_190404/linen_xpk_feat_nnx_set_defaults_true_20260420_190404_03_dropout/checkpoints', 'reached_preemption': False, 'preemption_received_at': None, 'synchronous': False, 'wait_for_prev_start_time': 1776712871.4984431, 'wait_for_prev_duration_secs': 6.699562072753906e-05, 'time_between_consecutive_saves_sec': None, 'checkpointer_blocking_start_time': 1776712871.5005178, 'checkpointer_blocking_duration_secs': 1.0562763214111328, 'get_old_steps_start_time': 1776712872.5568216, 'get_old_steps_duration_secs': 2.4080276489257812e-05, 'checkpoint_manager_blocking_start_time': 1776712871.4963229, 'checkpoint_manager_blocking_duration_secs': 1.0748770236968994} I0420 19:21:12.571337 137996531795776 checkpointing.py:409] Started an asynchronous checkpoint save for step 0 I0420 19:21:12.571390 137996531795776 max_utils.py:750] Memstats: After params initialized: I0420 19:21:12.571435 137996531795776 max_utils.py:756] Using (GB) 0.41 / 31.25 (1.312000%) on TPU_0(process=0,(0,0,0,0)) I0420 19:21:12.571467 137996531795776 max_utils.py:756] Using (GB) 0.41 / 31.25 (1.312000%) on TPU_1(process=0,(1,0,0,0)) I0420 19:21:12.571495 137996531795776 max_utils.py:756] Using (GB) 0.41 / 31.25 (1.312000%) on TPU_4(process=0,(0,1,0,0)) I0420 19:21:12.571520 137996531795776 max_utils.py:756] Using (GB) 0.41 / 31.25 (1.312000%) on TPU_5(process=0,(1,1,0,0)) I0420 19:21:12.882019 137996531795776 metric_logger.py:185] completed step: 0, seconds: 3.976, TFLOP/s/device: 0.201, Tokens/s/device: 32.194, total_weights: 4096, loss: 10.889 I0420 19:21:12.884920 137996531795776 metric_logger.py:269] To see full metrics 'tensorboard --logdir=gs://lance-maxtext/linen_ckpt_xpk_feat_nnx_set_defaults_true_20260420_190404/linen_xpk_feat_nnx_set_defaults_true_20260420_190404_03_dropout/tensorboard/' I0420 19:21:13.304287 137996531795776 metric_logger.py:185] completed step: 1, seconds: 1.459, TFLOP/s/device: 0.549, Tokens/s/device: 87.710, total_weights: 4096, loss: 10.902 I0420 19:21:13.373259 137996531795776 metric_logger.py:185] completed step: 2, seconds: 0.423, TFLOP/s/device: 1.893, Tokens/s/device: 302.524, total_weights: 4096, loss: 9.956 I0420 19:21:13.443890 137996531795776 metric_logger.py:185] completed step: 3, seconds: 0.009, TFLOP/s/device: 93.484, Tokens/s/device: 14941.053, total_weights: 4096, loss: 9.139 I0420 19:21:22.543231 137996531795776 metric_logger.py:185] completed step: 4, seconds: 0.070, TFLOP/s/device: 11.417, Tokens/s/device: 1824.739, total_weights: 4096, loss: 8.458 I0420 19:21:22.551711 137996531795776 metric_logger.py:185] completed step: 5, seconds: 0.071, TFLOP/s/device: 11.306, Tokens/s/device: 1806.965, total_weights: 4096, loss: 7.924 I0420 19:21:22.620636 137996531795776 metric_logger.py:185] completed step: 6, seconds: 9.098, TFLOP/s/device: 0.088, Tokens/s/device: 14.069, total_weights: 4096, loss: 7.553 I0420 19:21:22.692480 137996531795776 metric_logger.py:185] completed step: 7, seconds: 0.006, TFLOP/s/device: 125.392, Tokens/s/device: 20040.708, total_weights: 4096, loss: 7.267 I0420 19:21:22.764456 137996531795776 metric_logger.py:185] completed step: 8, seconds: 0.070, TFLOP/s/device: 11.452, Tokens/s/device: 1830.297, total_weights: 4096, loss: 7.111 I0420 19:21:22.834764 137996531795776 checkpointing.py:794] Waiting for step 9 to finish before checkpoint... I0420 19:21:22.835470 137996531795776 checkpointing.py:798] Waited 0.0007226467132568359 seconds for step 9 to finish before starting checkpointing. I0420 19:21:22.837408 137996531795776 checkpoint_manager.py:2024] [process=0][thread=MainThread][step=0][wait_until_finished] Waiting for Save Finalize thread (save_finalize) to complete. I0420 19:21:23.051041 137870236960512 atomicity.py:137] Creating tmp directory gs://lance-maxtext/linen_ckpt_xpk_feat_nnx_set_defaults_true_20260420_190404/linen_xpk_feat_nnx_set_defaults_true_20260420_190404_03_dropout/checkpoints/0/items I0420 19:21:23.214377 137868636083968 checkpoint.py:188] Wrote Metadata={'item_handlers': None, 'metrics': {}, 'performance_metrics': {}, 'init_timestamp_nsecs': 1776712882794637785, 'commit_timestamp_nsecs': None, 'custom_metadata': {}}, json={"item_handlers": null, "metrics": {}, "performance_metrics": {}, "init_timestamp_nsecs": 1776712882794637785, "commit_timestamp_nsecs": null, "custom_metadata": {}} to gs://lance-maxtext/linen_ckpt_xpk_feat_nnx_set_defaults_true_20260420_190404/linen_xpk_feat_nnx_set_defaults_true_20260420_190404_03_dropout/checkpoints/0/_CHECKPOINT_METADATA I0420 19:21:24.213218 2706 google_auth_provider.cc:181] Running on GCE, using service account 562977990677-compute@developer.gserviceaccount.com I0420 19:21:26.044207 137870220175104 array_metadata_store.py:203] [process=0][thread=array_type_handler] Wrote 39 array_metadata.ArrayMetadata to gs://lance-maxtext/linen_ckpt_xpk_feat_nnx_set_defaults_true_20260420_190404/linen_xpk_feat_nnx_set_defaults_true_20260420_190404_03_dropout/checkpoints/0/items/array_metadatas/process_0 I0420 19:21:55.475282 137870211782400 base_pytree_checkpoint_handler.py:1232] [process=0][thread=write_metadata_after_commits] Commit + Array metadata written. Time taken: 31.266033s (commit=30.123227s, array_metadata_write=1.142807s) I0420 19:21:55.476290 137869695903488 base_pytree_checkpoint_handler.py:129] [process=0] /jax/checkpoint/write/gbytes_per_sec: 36.371 MiB/s (total gbytes: 1.5 GiB) (time elapsed: 43.429962158203125 s) (per-host) I0420 19:21:55.476419 137869695903488 async_checkpointer.py:90] [process=0][thread=async_save] 3 Handler Commit operations completed. Time taken: 42.919777s. I0420 19:21:56.426314 137869695903488 checkpoint.py:228] Read Metadata={'item_handlers': None, 'metrics': {}, 'performance_metrics': {}, 'init_timestamp_nsecs': 1776712882794637785, 'commit_timestamp_nsecs': None, 'custom_metadata': {}} from gs://lance-maxtext/linen_ckpt_xpk_feat_nnx_set_defaults_true_20260420_190404/linen_xpk_feat_nnx_set_defaults_true_20260420_190404_03_dropout/checkpoints/0/_CHECKPOINT_METADATA I0420 19:21:57.193435 137868636083968 checkpoint.py:247] Updated Metadata={'item_handlers': {'items': 'orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeCheckpointHandler'}, 'metrics': {}, 'performance_metrics': {}, 'init_timestamp_nsecs': 1776712882794637785, 'commit_timestamp_nsecs': None, 'custom_metadata': {}} to gs://lance-maxtext/linen_ckpt_xpk_feat_nnx_set_defaults_true_20260420_190404/linen_xpk_feat_nnx_set_defaults_true_20260420_190404_03_dropout/checkpoints/0/_CHECKPOINT_METADATA I0420 19:21:57.520605 137869695903488 array_metadata_store.py:411] [process=0][thread=async_save] Validated ArrayMetadata from all 8 hosts. Time taken: 0.000245s. I0420 19:21:57.832484 137869695903488 ocdbt_utils.py:49] Param validation support for Zarr3 will be added later (b/362328389). I0420 19:21:58.465844 137869695903488 base_pytree_checkpoint_handler.py:1356] [process=0][thread=async_save] Pytree save finalize (merge_ocdbt + ArrayMetadata validation) completed. Time taken: 1.912069s. use_zarr3=True, enable_post_merge_validation=True, directory=gs://lance-maxtext/linen_ckpt_xpk_feat_nnx_set_defaults_true_20260420_190404/linen_xpk_feat_nnx_set_defaults_true_20260420_190404_03_dropout/checkpoints/0/items I0420 19:21:58.467636 137869695903488 atomicity.py:608] Finalizing gs://lance-maxtext/linen_ckpt_xpk_feat_nnx_set_defaults_true_20260420_190404/linen_xpk_feat_nnx_set_defaults_true_20260420_190404_03_dropout/checkpoints/0/items I0420 19:21:58.876771 137869695903488 atomicity.py:608] Finalizing gs://lance-maxtext/linen_ckpt_xpk_feat_nnx_set_defaults_true_20260420_190404/linen_xpk_feat_nnx_set_defaults_true_20260420_190404_03_dropout/checkpoints/0 I0420 19:22:00.238514 137869695903488 atomicity.py:790] [process=0][thread=async_save] Finished saving checkpoint (finalized tmp dir) to `gs://lance-maxtext/linen_ckpt_xpk_feat_nnx_set_defaults_true_20260420_190404/linen_xpk_feat_nnx_set_defaults_true_20260420_190404_03_dropout/checkpoints/0`. I0420 19:22:00.239333 137869695903488 async_checkpointer.py:420] Finished async_save (blocking + background). Time taken: 48.738791s. directory=gs://lance-maxtext/linen_ckpt_xpk_feat_nnx_set_defaults_true_20260420_190404/linen_xpk_feat_nnx_set_defaults_true_20260420_190404_03_dropout/checkpoints/0 I0420 19:22:00.241003 137869695903488 async_checkpointer.py:144] [process=0][thread=async_save] Background save thread done. Time taken: 47.684360s. I0420 19:22:00.241184 137868617955072 async_checkpointer.py:273] [process=0][thread=save_finalize] Done with waiting for background save thread=async_save. I0420 19:22:00.241263 137868617955072 async_checkpointer.py:283] [process=0][thread=save_finalize] No errors found in background save thread=async_save. I0420 19:22:00.241335 137868617955072 checkpoint_manager.py:2133] [process=0][thread=save_finalize][step=0] CheckpointManager Save Finalize is syncing with other hosts... I0420 19:22:00.242953 137868617955072 checkpoint_manager.py:2142] [process=0][thread=save_finalize][step=0] CheckpointManager Save Finalize is done on all hosts. I0420 19:22:00.243102 137996531795776 checkpoint_manager.py:2036] [process=0][thread=MainThread][step=0][wait_until_finished] Done waiting for Save Finalize thread (save_finalize) running at step=0. W0420 19:22:00.243174 137996531795776 checkpoint_manager.py:1458] Waiting for previous save to complete took 37.405777 seconds. If this number is high, consider checkpointing less frequently. I0420 19:22:00.245309 137996531795776 checkpoint_manager.py:1518] [process=0] Saving checkpoint at step 9 I0420 19:22:00.248719 137996531795776 async_checkpointer.py:452] [process=0] Started async saving checkpoint to gs://lance-maxtext/linen_ckpt_xpk_feat_nnx_set_defaults_true_20260420_190404/linen_xpk_feat_nnx_set_defaults_true_20260420_190404_03_dropout/checkpoints/9. I0420 19:22:01.205557 137869695903488 atomicity.py:137] Creating tmp directory gs://lance-maxtext/linen_ckpt_xpk_feat_nnx_set_defaults_true_20260420_190404/linen_xpk_feat_nnx_set_defaults_true_20260420_190404_03_dropout/checkpoints/9 I0420 19:22:01.210124 137996531795776 jax_array_handlers.py:358] Scheduling D2H of 39 prioritized jax.Array. I0420 19:22:01.210223 137996531795776 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 I0420 19:22:01.243740 137996531795776 base_pytree_checkpoint_handler.py:153] [process=0][thread=MainThread] Initiated "orbax.checkpoint._src.serialization.jax_array_handlers.ArrayHandler".serialize. Time taken: 0.034569s I0420 19:22:01.244606 137996531795776 base_pytree_checkpoint_handler.py:129] [process=0] /jax/checkpoint/write/blocking_gbytes_per_sec: 5.268 GiB/s (total gbytes: 1.5 GiB) (time elapsed: 0.29280567169189453 s) (per-host) I0420 19:22:01.244669 137996531795776 base_pytree_checkpoint_handler.py:737] [process=0][thread=MainThread] Initiated Pytree async_save. Time taken: 0.292889s (batch_requests_ready=0.255224s, total_serialization_initiated=0.036890s, others=0.000775s) I0420 19:22:01.244742 137996531795776 composite_checkpoint_handler.py:715] [process=0][thread=MainThread] Initiated CompositeCheckpointHandler.async_save. Time taken: 0.297178s (all_items=0.000017s, per_item={'items': '0.00001693'}, temp_paths=0.297161) I0420 19:22:01.245708 137869171611392 async_checkpointer.py:79] [process=0][thread=async_save] Background save thread started. I0420 19:22:01.245867 137996531795776 async_checkpointer.py:561] Finished blocking save. Time taken: 1.000484s. Continuing background save to gs://lance-maxtext/linen_ckpt_xpk_feat_nnx_set_defaults_true_20260420_190404/linen_xpk_feat_nnx_set_defaults_true_20260420_190404_03_dropout/checkpoints/9. I0420 19:22:01.263587 137996531795776 checkpoint_manager.py:1566] [process=0][thread=MainThread][step=9] Starting CheckpointManager Save Finalize thread=save_finalize I0420 19:22:01.263858 137868617955072 async_checkpointer.py:265] [process=0][thread=save_finalize] Waiting for background save thread=async_save. I0420 19:22:01.264029 137996531795776 standard_logger.py:34] {'step': 9, 'event_type': 'save', 'directory': 'gs://lance-maxtext/linen_ckpt_xpk_feat_nnx_set_defaults_true_20260420_190404/linen_xpk_feat_nnx_set_defaults_true_20260420_190404_03_dropout/checkpoints', 'reached_preemption': False, 'preemption_received_at': None, 'synchronous': False, 'wait_for_prev_start_time': 1776712882.8373768, 'wait_for_prev_duration_secs': 37.405776500701904, 'time_between_consecutive_saves_sec': None, 'checkpointer_blocking_start_time': 1776712920.2453496, 'checkpointer_blocking_duration_secs': 1.0006310939788818, 'get_old_steps_start_time': 1776712921.2460024, 'get_old_steps_duration_secs': 2.6702880859375e-05, 'checkpoint_manager_blocking_start_time': 1776712882.8356872, 'checkpoint_manager_blocking_duration_secs': 38.42830729484558} I0420 19:22:01.264145 137996531795776 checkpointing.py:409] Started an asynchronous checkpoint save for step 9 I0420 19:22:01.264189 137996531795776 checkpoint_manager.py:2024] [process=0][thread=MainThread][step=9][wait_until_finished] Waiting for Save Finalize thread (save_finalize) to complete. I0420 19:22:03.202197 137869708482304 atomicity.py:137] Creating tmp directory gs://lance-maxtext/linen_ckpt_xpk_feat_nnx_set_defaults_true_20260420_190404/linen_xpk_feat_nnx_set_defaults_true_20260420_190404_03_dropout/checkpoints/9/items I0420 19:22:06.102053 137869721900800 array_metadata_store.py:203] [process=0][thread=array_type_handler] Wrote 39 array_metadata.ArrayMetadata to gs://lance-maxtext/linen_ckpt_xpk_feat_nnx_set_defaults_true_20260420_190404/linen_xpk_feat_nnx_set_defaults_true_20260420_190404_03_dropout/checkpoints/9/items/array_metadatas/process_0 I0420 19:22:44.755834 137869188396800 base_pytree_checkpoint_handler.py:1232] [process=0][thread=write_metadata_after_commits] Commit + Array metadata written. Time taken: 40.374227s (commit=38.466247s, array_metadata_write=1.907980s) I0420 19:22:44.757999 137869171611392 base_pytree_checkpoint_handler.py:129] [process=0] /jax/checkpoint/write/gbytes_per_sec: 36.059 MiB/s (total gbytes: 1.5 GiB) (time elapsed: 43.80616116523743 s) (per-host) I0420 19:22:44.758140 137869171611392 async_checkpointer.py:90] [process=0][thread=async_save] 3 Handler Commit operations completed. Time taken: 43.512319s. I0420 19:22:46.266796 137869171611392 array_metadata_store.py:411] [process=0][thread=async_save] Validated ArrayMetadata from all 8 hosts. Time taken: 0.000285s. I0420 19:22:46.617341 137869171611392 ocdbt_utils.py:49] Param validation support for Zarr3 will be added later (b/362328389). I0420 19:22:47.238803 137869171611392 base_pytree_checkpoint_handler.py:1356] [process=0][thread=async_save] Pytree save finalize (merge_ocdbt + ArrayMetadata validation) completed. Time taken: 1.595841s. use_zarr3=True, enable_post_merge_validation=True, directory=gs://lance-maxtext/linen_ckpt_xpk_feat_nnx_set_defaults_true_20260420_190404/linen_xpk_feat_nnx_set_defaults_true_20260420_190404_03_dropout/checkpoints/9/items I0420 19:22:47.244112 137869171611392 atomicity.py:608] Finalizing gs://lance-maxtext/linen_ckpt_xpk_feat_nnx_set_defaults_true_20260420_190404/linen_xpk_feat_nnx_set_defaults_true_20260420_190404_03_dropout/checkpoints/9/items I0420 19:22:48.103626 137869171611392 atomicity.py:608] Finalizing gs://lance-maxtext/linen_ckpt_xpk_feat_nnx_set_defaults_true_20260420_190404/linen_xpk_feat_nnx_set_defaults_true_20260420_190404_03_dropout/checkpoints/9 I0420 19:22:49.406368 137869171611392 atomicity.py:790] [process=0][thread=async_save] Finished saving checkpoint (finalized tmp dir) to `gs://lance-maxtext/linen_ckpt_xpk_feat_nnx_set_defaults_true_20260420_190404/linen_xpk_feat_nnx_set_defaults_true_20260420_190404_03_dropout/checkpoints/9`. I0420 19:22:49.407038 137869171611392 async_checkpointer.py:420] Finished async_save (blocking + background). Time taken: 49.161676s. directory=gs://lance-maxtext/linen_ckpt_xpk_feat_nnx_set_defaults_true_20260420_190404/linen_xpk_feat_nnx_set_defaults_true_20260420_190404_03_dropout/checkpoints/9 I0420 19:22:49.408556 137869171611392 async_checkpointer.py:144] [process=0][thread=async_save] Background save thread done. Time taken: 48.162736s. I0420 19:22:49.408710 137868617955072 async_checkpointer.py:273] [process=0][thread=save_finalize] Done with waiting for background save thread=async_save. I0420 19:22:49.408771 137868617955072 async_checkpointer.py:283] [process=0][thread=save_finalize] No errors found in background save thread=async_save. I0420 19:22:49.408836 137868617955072 checkpoint_manager.py:2133] [process=0][thread=save_finalize][step=9] CheckpointManager Save Finalize is syncing with other hosts... I0420 19:22:49.410494 137868617955072 checkpoint_manager.py:2142] [process=0][thread=save_finalize][step=9] CheckpointManager Save Finalize is done on all hosts. I0420 19:22:49.410649 137996531795776 checkpoint_manager.py:2036] [process=0][thread=MainThread][step=9][wait_until_finished] Done waiting for Save Finalize thread (save_finalize) running at step=9. I0420 19:22:49.410788 137996531795776 checkpoint_manager.py:2013] [process=0][thread=MainThread][wait_until_finished] No Save Finalize thread to wait for. Returning. I0420 19:22:49.411541 137996531795776 metric_logger.py:185] completed step: 9, seconds: 0.072, TFLOP/s/device: 11.136, Tokens/s/device: 1779.755, total_weights: 4096, loss: 6.981 Per train step: Total TFLOPs: 0.80 split as 99.60% learnable weight flops and 0.40% attention flops XPK End: Mon Apr 20 19:23:00 UTC 2026 EXIT_CODE=0