main (574ad3fb9) vs feat/nnx-trainstate-and-training-loop (574ad3fb9)| Metric | main 574ad3fb9 | feat/nnx-trainstate-and-training-loop 574ad3fb9 | Diff (feat/nnx-trainstate-and-training-loop − main) |
|---|---|---|---|
| Parameters | 1.104 billion | 1.104 billion | — |
| Final loss | 8.1790 | 8.1790 | 0 |
| TFLOP/s | 91.735 | 92.222 | +0.487 |
| Tok/s | 13827.3 | 13900.8 | +73.486 |
| Avg s/step | 2.697 | 2.751 | +0.054 |
| Memory % | 1.38 | 1.38 | 0 |
| JAX | 0.9.2 | 0.9.2 | — |
Diff = branch value − main value. Green = branch improved. Red = branch regressed.
XPK Start: Mon Apr 20 06:04:23 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 06:05:03.169743: 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 06:05:03.705629 134537364629312 max_utils.py:273] Attempting to initialize the jax distributed system... I0420 06:05:12.747409 134537364629312 distributed.py:149] Starting JAX distributed service on [::]:8482 I0420 06:05:12.749641 134537364629312 distributed.py:172] Connecting to JAX distributed service on mt-02-synthetic-2y2f8-slice-job-0-0.mt-02-synthetic-2y2f8:8482 I0420 06:05:19.953547 134537364629312 max_utils.py:284] Jax distributed system initialized! I0420 06:05:26.425812 134537364629312 max_utils.py:800] System Information: Jax Version: 0.9.2 I0420 06:05:26.425916 134537364629312 max_utils.py:801] System Information: Jaxlib Version: 0.9.2 I0420 06:05:26.425956 134537364629312 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 06:05:26.425998 134537364629312 train_utils.py:348] WARNING: Sequence packing is essentially ignored for synthetic data. Please use a real dataset to use sequence packing. I0420 06:05:27.123493 134537364629312 maxtext_utils.py:1551] Num_devices: 32, shape (1, 1, 1, 32, 1, 1, 1, 1, 1, 1, 1, 1, 1) I0420 06:05:27.123769 134537364629312 checkpointing.py:677] Setting up checkpoint logger... I0420 06:05:27.123822 134537364629312 checkpointing.py:233] Creating checkpoint manager with ocdbt=True and zarr3=True I0420 06:05:27.123873 134537364629312 pytree_checkpoint_handler.py:592] save_device_host_concurrent_bytes=None I0420 06:05:27.124198 134537364629312 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 0x7a5bcd514050>, enable_pinned_host_transfer=False, save_concurrent_bytes: 96000000000 (89.4 GiB), restore_concurrent_bytes: 96000000000 (89.4 GiB) I0420 06:05:30.189278 134537364629312 checkpointing.py:265] Enabling policy for fixed interval checkpointing. I0420 06:05:30.189555 134537364629312 checkpoint_manager.py:708] [process=3][thread=MainThread] CheckpointManager init: checkpointers=None, item_names=('items',), item_handlers={'items': <orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeCheckpointHandler object at 0x7a4774447890>}, handler_registry=None I0420 06:05:30.189790 134537364629312 composite_checkpoint_handler.py:237] Deferred registration for item: "items". Adding handler `<orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeCheckpointHandler object at 0x7a4774447890>` 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 06:05:30.189842 134537364629312 composite_checkpoint_handler.py:237] Deferred registration for item: "metrics". Adding handler `<orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonCheckpointHandler object at 0x7a4774449520>` 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 06:05:30.189878 134537364629312 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 0x7a4774447890>, ('items', <class 'orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeRestoreArgs'>): <orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeCheckpointHandler object at 0x7a4774447890>, ('metrics', <class 'orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonSaveArgs'>): <orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonCheckpointHandler object at 0x7a4774449520>, ('metrics', <class 'orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonRestoreArgs'>): <orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonCheckpointHandler object at 0x7a4774449520>}). I0420 06:05:30.190189 134537364629312 abstract_checkpointer.py:35] orbax-checkpoint version: 0.11.34 I0420 06:05:30.190259 134537364629312 async_checkpointer.py:192] [process=3][thread=MainThread] Using barrier_sync_fn: <function get_barrier_sync_fn.<locals>._fn at 0x7a474c7d3880> timeout: 1200 secs and primary_host=0 for async checkpoint writes I0420 06:05:31.867599 134537364629312 checkpoint_manager.py:1812] Found 0 checkpoint steps in gs://lance-maxtext/linen_ckpt_xpk_main_20260420_060028/linen_xpk_main_20260420_060028_02_synthetic/checkpoints I0420 06:05:32.721962 134537364629312 checkpoint_manager.py:929] [process=3][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_02_synthetic/checkpoints: <orbax.checkpoint.checkpoint_manager.CheckpointManager object at 0x7a4774445160> I0420 06:05:32.722133 134537364629312 checkpointing.py:301] Checkpoint manager created! I0420 06:05:33.647060 134537364629312 nnx_wrappers.py:437] Unknown Logical: bfloat16[32,2048,2048]...................................... ('activation_batch', 'activation_norm_length', 'activation_embed'). I0420 06:05:33.647170 134537364629312 nnx_wrappers.py:437] Unknown Physical: bfloat16[32,2048,2048]...................................... ('fsdp', None, None). I0420 06:05:34.025807 134537364629312 attentions.py:1088] attentions/inputs_q Logical: bfloat16[32,2048,2048]...................................... ('activation_batch', 'activation_attn_length', 'activation_attn_embed'). I0420 06:05:34.025899 134537364629312 attentions.py:1088] attentions/inputs_q Physical: bfloat16[32,2048,2048]...................................... ('fsdp', None, None). I0420 06:05:34.042320 134537364629312 attentions.py:1089] attentions/inputs_kv Logical: bfloat16[32,2048,2048]...................................... ('activation_batch', 'activation_attn_length', 'activation_attn_embed'). I0420 06:05:34.042382 134537364629312 attentions.py:1089] attentions/inputs_kv Physical: bfloat16[32,2048,2048]...................................... ('fsdp', None, None). I0420 06:05:34.065995 134537364629312 attentions.py:1154] attentions/query Logical: bfloat16[32,2048,16,128].................................... ('activation_kv_batch', 'activation_attn_length', 'activation_kv_heads', 'activation_kv_head_dim'). I0420 06:05:34.066061 134537364629312 attentions.py:1154] attentions/query Physical: bfloat16[32,2048,16,128].................................... ('fsdp', None, None, None). I0420 06:05:34.082418 134537364629312 attentions.py:1155] attentions/key Logical: bfloat16[32,2048,16,128].................................... ('activation_kv_batch', 'activation_attn_length', 'activation_kv_heads', 'activation_kv_head_dim'). I0420 06:05:34.082478 134537364629312 attentions.py:1155] attentions/key Physical: bfloat16[32,2048,16,128].................................... ('fsdp', None, None, None). I0420 06:05:34.098690 134537364629312 attentions.py:1156] attentions/value Logical: bfloat16[32,2048,16,128].................................... ('activation_kv_batch', 'activation_attn_length', 'activation_kv_heads', 'activation_kv_head_dim'). I0420 06:05:34.098749 134537364629312 attentions.py:1156] attentions/value Physical: bfloat16[32,2048,16,128].................................... ('fsdp', None, None, None). I0420 06:05:34.124208 134537364629312 attentions.py:1197] attentions/out Logical: bfloat16[32,2048,16,128].................................... ('activation_batch', 'activation_attn_length', 'activation_heads', 'activation_kv'). I0420 06:05:34.124280 134537364629312 attentions.py:1197] attentions/out Physical: bfloat16[32,2048,16,128].................................... ('fsdp', None, None, None). I0420 06:05:34.145139 134537364629312 linears.py:525] linears/x Logical: bfloat16[32,2048,7168]...................................... ('activation_batch', 'activation_length', 'activation_mlp'). I0420 06:05:34.145202 134537364629312 linears.py:525] linears/x Physical: bfloat16[32,2048,7168]...................................... ('fsdp', None, None). I0420 06:05:34.358899 134537364629312 checkpointing.py:577] checkpoint manager exists so trying to load this run's existing checkpoint I0420 06:05:34.359006 134537364629312 checkpointing.py:665] No existing checkpoints found, not restoring checkpoint. fsdp: 32 I0420 06:05:35.770669 134537364629312 maxtext_utils.py:1654] params/params/decoder/decoder_norm/scale Shape: float32[2048] Logical: P('norm',) Physical: (None,) I0420 06:05:35.770787 134537364629312 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 06:05:35.770846 134537364629312 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 06:05:35.770904 134537364629312 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 06:05:35.770958 134537364629312 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 06:05:35.770997 134537364629312 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 06:05:35.771049 134537364629312 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 06:05:35.771100 134537364629312 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 06:05:35.771140 134537364629312 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 06:05:35.771177 134537364629312 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 06:05:35.771223 134537364629312 maxtext_utils.py:1654] params/params/decoder/logits_dense/kernel Shape: float32[2048,32000] Logical: P('embed_vocab', 'vocab') Physical: ('fsdp', None) I0420 06:05:35.771270 134537364629312 maxtext_utils.py:1654] params/params/token_embedder/embedding Shape: float32[32000,2048] Logical: P('vocab', 'embed_vocab') Physical: (None, 'fsdp') I0420 06:05:36.256777 134537364629312 train.py:155] train/xent Logical: float32[32,2048]............................................ ('activation_embed_and_logits_batch', 'activation_length'). I0420 06:05:36.256868 134537364629312 train.py:155] train/xent Physical: float32[32,2048]............................................ ('fsdp', None). I0420 06:05:36.272108 134537364629312 train.py:162] train/z_loss Logical: float32[32,2048]............................................ ('activation_embed_and_logits_batch', 'activation_length'). I0420 06:05:36.272166 134537364629312 train.py:162] train/z_loss Physical: float32[32,2048]............................................ ('fsdp', None). I0420 06:05:47.315834 134537364629312 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. I0420 06:05:47.316628 134537364629312 metric_logger.py:301] number parameters: 1.104 billion I0420 06:05:58.597790 134537364629312 checkpointing.py:772] Waiting for step 0 to finish before checkpoint... I0420 06:05:58.750764 134537364629312 checkpointing.py:776] Waited 0.15296649932861328 seconds for step 0 to finish before starting checkpointing. I0420 06:05:58.753222 134537364629312 checkpoint_manager.py:2009] [process=3][thread=MainThread][wait_until_finished] No Save Finalize thread to wait for. Returning. I0420 06:05:58.754627 134537364629312 checkpoint_manager.py:1512] [process=3] Saving checkpoint at step 0 I0420 06:05:58.756154 134537364629312 event_tracking.py:70] [process=3] [async] Started save checkpoint @ gs://lance-maxtext/linen_ckpt_xpk_main_20260420_060028/linen_xpk_main_20260420_060028_02_synthetic/checkpoints/0. I0420 06:05:59.495821 134537364629312 signaling_client.py:364] Using JaxDistributedSignalingClient I0420 06:05:59.496773 134537364629312 jax_array_handlers.py:360] Scheduling D2H of 39 prioritized jax.Array. I0420 06:05:59.496833 134537364629312 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 06:05:59.778315 134537364629312 base_pytree_checkpoint_handler.py:154] [process=3][thread=MainThread] Initiated "orbax.checkpoint._src.serialization.jax_array_handlers.ArrayHandler".serialize. Time taken: 0.282556s I0420 06:05:59.778496 134537364629312 base_pytree_checkpoint_handler.py:130] [process=3] /jax/orbax/write/blocking_gbytes_per_sec: 5.357 GiB/s (total gbytes: 1.5 GiB) (time elapsed: 0.2879605293273926 s) (per-host) I0420 06:05:59.778549 134537364629312 base_pytree_checkpoint_handler.py:768] [process=3][thread=MainThread] Initiated Pytree async_save. Time taken: 0.288026s (batch_requests_ready=0.002177s, total_serialization_initiated=0.285775s, others=0.000074s) I0420 06:05:59.778637 134537364629312 composite_checkpoint_handler.py:715] [process=3][thread=MainThread] Initiated CompositeCheckpointHandler.async_save. Time taken: 0.292061s (all_items=0.000018s, per_item={'items': '0.00001764'}, temp_paths=0.292043) I0420 06:05:59.779476 134537364629312 event_tracking.py:125] [process=3] [async] Finished blocking save in 1.02 seconds. Continuing save @ gs://lance-maxtext/linen_ckpt_xpk_main_20260420_060028/linen_xpk_main_20260420_060028_02_synthetic/checkpoints/0. I0420 06:05:59.779807 134408015165184 async_checkpointer.py:76] [process=3][thread=async_save] Background save thread started. Deadline for this save operation is 2026-04-20 06:25:59.779768 I0420 06:05:59.782449 134537364629312 checkpoint_manager.py:1560] [process=3][thread=MainThread][step=0] Starting CheckpointManager Save Finalize thread=save_finalize I0420 06:05:59.782698 134401551742720 async_checkpointer.py:280] [process=3][thread=save_finalize] Waiting for background save thread=async_save. I0420 06:05:59.782817 134537364629312 standard_logger.py:34] {'step': 0, 'event_type': 'save', 'directory': 'gs://lance-maxtext/linen_ckpt_xpk_main_20260420_060028/linen_xpk_main_20260420_060028_02_synthetic/checkpoints', 'reached_preemption': False, 'preemption_received_at': None, 'synchronous': False, 'wait_for_prev_start_time': 1776665158.7532036, 'wait_for_prev_duration_secs': 6.127357482910156e-05, 'time_between_consecutive_saves_sec': None, 'checkpointer_blocking_start_time': 1776665158.7546682, 'checkpointer_blocking_duration_secs': 1.025301456451416, 'get_old_steps_start_time': 1776665159.7799947, 'get_old_steps_duration_secs': 3.170967102050781e-05, 'checkpoint_manager_blocking_start_time': 1776665158.751357, 'checkpoint_manager_blocking_duration_secs': 1.0314197540283203} I0420 06:05:59.782932 134537364629312 checkpointing.py:408] Started an asynchronous checkpoint save for step 0 I0420 06:05:59.782985 134537364629312 max_utils.py:750] Memstats: After params initialized: I0420 06:05:59.783034 134537364629312 max_utils.py:756] Using (GB) 0.43 / 31.25 (1.376000%) on TPU_10(process=3,(2,2,0,0)) I0420 06:05:59.783066 134537364629312 max_utils.py:756] Using (GB) 0.43 / 31.25 (1.376000%) on TPU_11(process=3,(3,2,0,0)) I0420 06:05:59.783091 134537364629312 max_utils.py:756] Using (GB) 0.43 / 31.25 (1.376000%) on TPU_14(process=3,(2,3,0,0)) I0420 06:05:59.783114 134537364629312 max_utils.py:756] Using (GB) 0.43 / 31.25 (1.376000%) on TPU_15(process=3,(3,3,0,0)) I0420 06:06:00.093813 134537364629312 metric_logger.py:196] completed step: 0, seconds: 11.281, TFLOP/s/device: 1.204, Tokens/s/device: 181.543, total_weights: 65536, loss: 10.874, lm_loss: 10.874, perplexity: 52779.875 I0420 06:06:00.272749 134537364629312 metric_logger.py:196] completed step: 1, seconds: 1.495, TFLOP/s/device: 9.091, Tokens/s/device: 1370.334, total_weights: 65536, loss: 10.874, lm_loss: 10.874, perplexity: 52779.875 I0420 06:06:00.690477 134537364629312 metric_logger.py:196] completed step: 2, seconds: 0.032, TFLOP/s/device: 427.255, Tokens/s/device: 64400.491, total_weights: 65536, loss: 10.262, lm_loss: 10.262, perplexity: 28637.129 I0420 06:06:00.837826 134537364629312 metric_logger.py:196] completed step: 3, seconds: 0.418, TFLOP/s/device: 32.490, Tokens/s/device: 4897.295, total_weights: 65536, loss: 9.734, lm_loss: 9.734, perplexity: 16889.203 I0420 06:06:01.132997 134537364629312 metric_logger.py:196] completed step: 4, seconds: 0.152, TFLOP/s/device: 89.114, Tokens/s/device: 13432.239, total_weights: 65536, loss: 9.277, lm_loss: 9.277, perplexity: 10694.614 I0420 06:06:01.138906 134537364629312 metric_logger.py:196] completed step: 5, seconds: 0.147, TFLOP/s/device: 92.154, Tokens/s/device: 13890.490, total_weights: 65536, loss: 8.892, lm_loss: 8.892, perplexity: 7270.717 I0420 06:06:03.490737 2869 google_auth_provider.cc:181] Running on GCE, using service account 562977990677-compute@developer.gserviceaccount.com I0420 06:06:05.563695 134405335017216 array_metadata_store.py:203] [process=3][thread=array_type_handler] Wrote 39 array_metadata.ArrayMetadata to gs://lance-maxtext/linen_ckpt_xpk_main_20260420_060028/linen_xpk_main_20260420_060028_02_synthetic/checkpoints/0/items/array_metadatas/process_3 I0420 06:06:22.718634 134537364629312 metric_logger.py:196] completed step: 6, seconds: 0.296, TFLOP/s/device: 45.952, Tokens/s/device: 6926.313, total_weights: 65536, loss: 8.592, lm_loss: 8.592, perplexity: 5390.759 I0420 06:06:22.865889 134537364629312 metric_logger.py:196] completed step: 7, seconds: 21.433, TFLOP/s/device: 0.634, Tokens/s/device: 95.553, total_weights: 65536, loss: 8.384, lm_loss: 8.384, perplexity: 4376.905 I0420 06:06:23.013392 134537364629312 metric_logger.py:196] completed step: 8, seconds: 0.152, TFLOP/s/device: 89.364, Tokens/s/device: 13469.962, total_weights: 65536, loss: 8.255, lm_loss: 8.255, perplexity: 3846.406 I0420 06:06:23.159990 134537364629312 checkpointing.py:772] Waiting for step 9 to finish before checkpoint... I0420 06:06:23.160644 134537364629312 checkpointing.py:776] Waited 0.0006763935089111328 seconds for step 9 to finish before starting checkpointing. I0420 06:06:23.162776 134537364629312 checkpoint_manager.py:2020] [process=3][thread=MainThread][step=0][wait_until_finished] Waiting for Save Finalize thread (save_finalize) to complete. I0420 06:06:37.396360 134408015165184 base_pytree_checkpoint_handler.py:130] [process=3] /jax/orbax/write/gbytes_per_sec: 41.672 MiB/s (total gbytes: 1.5 GiB) (time elapsed: 37.90577483177185 s) (per-host) I0420 06:06:37.396485 134408015165184 async_checkpointer.py:90] [process=3][thread=async_save] 3 Handler Commit operations completed. Time taken: 37.616556s. I0420 06:06:47.372949 134408015165184 async_checkpointer.py:160] [process=3][thread=async_save] Background save thread done. Time taken: 47.593003s. I0420 06:06:47.373171 134401551742720 async_checkpointer.py:288] [process=3][thread=save_finalize] Done with waiting for background save thread=async_save. I0420 06:06:47.373242 134401551742720 async_checkpointer.py:298] [process=3][thread=save_finalize] No errors found in background save thread=async_save. I0420 06:06:47.373292 134401551742720 checkpoint_manager.py:2137] [process=3][thread=save_finalize][step=0] CheckpointManager Save Finalize is syncing with other hosts... I0420 06:06:47.375020 134401551742720 checkpoint_manager.py:2146] [process=3][thread=save_finalize][step=0] CheckpointManager Save Finalize is done on all hosts. I0420 06:06:47.375234 134537364629312 checkpoint_manager.py:2032] [process=3][thread=MainThread][step=0][wait_until_finished] Done waiting for Save Finalize thread (save_finalize) running at step=0. W0420 06:06:47.375392 134537364629312 checkpoint_manager.py:1452] Waiting for previous save to complete took 24.212615 seconds. If this number is high, consider checkpointing less frequently. I0420 06:06:47.377159 134537364629312 checkpoint_manager.py:1512] [process=3] Saving checkpoint at step 9 I0420 06:06:47.379285 134537364629312 event_tracking.py:70] [process=3] [async] Started save checkpoint @ gs://lance-maxtext/linen_ckpt_xpk_main_20260420_060028/linen_xpk_main_20260420_060028_02_synthetic/checkpoints/9. I0420 06:06:48.073672 134537364629312 jax_array_handlers.py:360] Scheduling D2H of 39 prioritized jax.Array. I0420 06:06:48.073765 134537364629312 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 06:06:48.109095 134537364629312 base_pytree_checkpoint_handler.py:154] [process=3][thread=MainThread] Initiated "orbax.checkpoint._src.serialization.jax_array_handlers.ArrayHandler".serialize. Time taken: 0.036632s I0420 06:06:48.109268 134537364629312 base_pytree_checkpoint_handler.py:130] [process=3] /jax/orbax/write/blocking_gbytes_per_sec: 38.559 GiB/s (total gbytes: 1.5 GiB) (time elapsed: 0.04000568389892578 s) (per-host) I0420 06:06:48.109331 134537364629312 base_pytree_checkpoint_handler.py:768] [process=3][thread=MainThread] Initiated Pytree async_save. Time taken: 0.040078s (batch_requests_ready=0.001720s, total_serialization_initiated=0.038276s, others=0.000082s) I0420 06:06:48.109419 134537364629312 composite_checkpoint_handler.py:715] [process=3][thread=MainThread] Initiated CompositeCheckpointHandler.async_save. Time taken: 0.044291s (all_items=0.000015s, per_item={'items': '0.00001526'}, temp_paths=0.044276) I0420 06:06:48.110059 134537364629312 event_tracking.py:125] [process=3] [async] Finished blocking save in 0.73 seconds. Continuing save @ gs://lance-maxtext/linen_ckpt_xpk_main_20260420_060028/linen_xpk_main_20260420_060028_02_synthetic/checkpoints/9. I0420 06:06:48.110407 134405335017216 async_checkpointer.py:76] [process=3][thread=async_save] Background save thread started. Deadline for this save operation is 2026-04-20 06:26:48.110371 I0420 06:06:48.115522 134537364629312 checkpoint_manager.py:1560] [process=3][thread=MainThread][step=9] Starting CheckpointManager Save Finalize thread=save_finalize I0420 06:06:48.115802 134401551742720 async_checkpointer.py:280] [process=3][thread=save_finalize] Waiting for background save thread=async_save. I0420 06:06:48.115960 134537364629312 standard_logger.py:34] {'step': 9, 'event_type': 'save', 'directory': 'gs://lance-maxtext/linen_ckpt_xpk_main_20260420_060028/linen_xpk_main_20260420_060028_02_synthetic/checkpoints', 'reached_preemption': False, 'preemption_received_at': None, 'synchronous': False, 'wait_for_prev_start_time': 1776665183.1627462, 'wait_for_prev_duration_secs': 24.21261477470398, 'time_between_consecutive_saves_sec': None, 'checkpointer_blocking_start_time': 1776665207.3771987, 'checkpointer_blocking_duration_secs': 0.7333498001098633, 'get_old_steps_start_time': 1776665208.110569, 'get_old_steps_duration_secs': 3.147125244140625e-05, 'checkpoint_manager_blocking_start_time': 1776665183.1608818, 'checkpoint_manager_blocking_duration_secs': 24.955042362213135} I0420 06:06:48.116071 134537364629312 checkpointing.py:408] Started an asynchronous checkpoint save for step 9 I0420 06:06:48.116117 134537364629312 checkpoint_manager.py:2020] [process=3][thread=MainThread][step=9][wait_until_finished] Waiting for Save Finalize thread (save_finalize) to complete. I0420 06:06:54.069175 134408015165184 array_metadata_store.py:203] [process=3][thread=array_type_handler] Wrote 39 array_metadata.ArrayMetadata to gs://lance-maxtext/linen_ckpt_xpk_main_20260420_060028/linen_xpk_main_20260420_060028_02_synthetic/checkpoints/9/items/array_metadatas/process_3 I0420 06:07:29.778391 134405335017216 base_pytree_checkpoint_handler.py:130] [process=3] /jax/orbax/write/gbytes_per_sec: 37.872 MiB/s (total gbytes: 1.5 GiB) (time elapsed: 41.7090950012207 s) (per-host) I0420 06:07:29.778497 134405335017216 async_checkpointer.py:90] [process=3][thread=async_save] 3 Handler Commit operations completed. Time taken: 41.667977s. I0420 06:07:40.600686 134405335017216 async_checkpointer.py:160] [process=3][thread=async_save] Background save thread done. Time taken: 52.490149s. I0420 06:07:40.600949 134401551742720 async_checkpointer.py:288] [process=3][thread=save_finalize] Done with waiting for background save thread=async_save. I0420 06:07:40.601093 134401551742720 async_checkpointer.py:298] [process=3][thread=save_finalize] No errors found in background save thread=async_save. I0420 06:07:40.601161 134401551742720 checkpoint_manager.py:2137] [process=3][thread=save_finalize][step=9] CheckpointManager Save Finalize is syncing with other hosts... I0420 06:07:40.603692 134401551742720 checkpoint_manager.py:2146] [process=3][thread=save_finalize][step=9] CheckpointManager Save Finalize is done on all hosts. I0420 06:07:40.603893 134537364629312 checkpoint_manager.py:2032] [process=3][thread=MainThread][step=9][wait_until_finished] Done waiting for Save Finalize thread (save_finalize) running at step=9. I0420 06:07:40.604041 134537364629312 checkpoint_manager.py:2009] [process=3][thread=MainThread][wait_until_finished] No Save Finalize thread to wait for. Returning. I0420 06:07:40.605084 134537364629312 metric_logger.py:196] completed step: 9, seconds: 0.148, TFLOP/s/device: 91.735, Tokens/s/device: 13827.281, 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: Mon Apr 20 06:07:48 UTC 2026 EXIT_CODE=0
574ad3fb9
· feat_nnx_trainstate_and_training_loop_20260420_081533
· full log
XPK Start: Mon Apr 20 08:17:58 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 08:18:23.834375: 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 08:18:24.013493 133563095811904 max_utils.py:273] Attempting to initialize the jax distributed system... I0420 08:18:33.053613 133563095811904 distributed.py:149] Starting JAX distributed service on [::]:8482 I0420 08:18:33.055877 133563095811904 distributed.py:172] Connecting to JAX distributed service on mt-02-synthetic-dkb2s-slice-job-0-0.mt-02-synthetic-dkb2s:8482 I0420 08:18:34.997037 133563095811904 max_utils.py:284] Jax distributed system initialized! I0420 08:18:41.154794 133563095811904 max_utils.py:800] System Information: Jax Version: 0.9.2 I0420 08:18:41.154896 133563095811904 max_utils.py:801] System Information: Jaxlib Version: 0.9.2 I0420 08:18:41.154934 133563095811904 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 08:18:41.154968 133563095811904 train_utils.py:378] WARNING: Sequence packing is essentially ignored for synthetic data. Please use a real dataset to use sequence packing. I0420 08:18:42.197620 133563095811904 maxtext_utils.py:1718] Num_devices: 32, shape (1, 1, 1, 32, 1, 1, 1, 1, 1, 1, 1, 1, 1) I0420 08:18:42.198217 133563095811904 maxtext_utils.py:1718] Num_devices: 32, shape (1, 1, 1, 32, 1, 1, 1, 1, 1, 1, 1, 1, 1) I0420 08:18:42.198392 133563095811904 checkpointing.py:688] Setting up checkpoint logger... I0420 08:18:42.198440 133563095811904 checkpointing.py:234] Creating checkpoint manager with ocdbt=True and zarr3=True I0420 08:18:42.198482 133563095811904 pytree_checkpoint_handler.py:592] save_device_host_concurrent_bytes=None I0420 08:18:42.198826 133563095811904 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 0x7978dd24c350>, enable_pinned_host_transfer=False, save_concurrent_bytes: 96000000000 (89.4 GiB), restore_concurrent_bytes: 96000000000 (89.4 GiB) I0420 08:18:44.918165 133563095811904 checkpointing.py:266] Enabling policy for fixed interval checkpointing. I0420 08:18:44.918355 133563095811904 checkpoint_manager.py:708] [process=1][thread=MainThread] CheckpointManager init: checkpointers=None, item_names=('items',), item_handlers={'items': <orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeCheckpointHandler object at 0x7977d1461e20>}, handler_registry=None I0420 08:18:44.918581 133563095811904 composite_checkpoint_handler.py:237] Deferred registration for item: "items". Adding handler `<orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeCheckpointHandler object at 0x7977d1461e20>` 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 08:18:44.918628 133563095811904 composite_checkpoint_handler.py:237] Deferred registration for item: "metrics". Adding handler `<orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonCheckpointHandler object at 0x7977d1533560>` 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 08:18:44.918663 133563095811904 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 0x7977d1461e20>, ('items', <class 'orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeRestoreArgs'>): <orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeCheckpointHandler object at 0x7977d1461e20>, ('metrics', <class 'orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonSaveArgs'>): <orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonCheckpointHandler object at 0x7977d1533560>, ('metrics', <class 'orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonRestoreArgs'>): <orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonCheckpointHandler object at 0x7977d1533560>}). I0420 08:18:44.918994 133563095811904 abstract_checkpointer.py:35] orbax-checkpoint version: 0.11.34 I0420 08:18:44.919066 133563095811904 async_checkpointer.py:192] [process=1][thread=MainThread] Using barrier_sync_fn: <function get_barrier_sync_fn.<locals>._fn at 0x796490687740> timeout: 1200 secs and primary_host=0 for async checkpoint writes I0420 08:18:46.015626 133563095811904 checkpoint_manager.py:1812] Found 0 checkpoint steps in gs://lance-maxtext/linen_ckpt_xpk_feat_nnx_trainstate_and_training_loop_20260420_081533/linen_xpk_feat_nnx_trainstate_and_training_loop_20260420_081533_02_synthetic/checkpoints I0420 08:18:46.885110 133563095811904 checkpoint_manager.py:929] [process=1][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_trainstate_and_training_loop_20260420_081533/linen_xpk_feat_nnx_trainstate_and_training_loop_20260420_081533_02_synthetic/checkpoints: <orbax.checkpoint.checkpoint_manager.CheckpointManager object at 0x7964c029aff0> I0420 08:18:46.885275 133563095811904 checkpointing.py:302] Checkpoint manager created! I0420 08:18:47.811537 133563095811904 nnx_wrappers.py:437] Unknown Logical: bfloat16[32,2048,2048]...................................... ('activation_batch', 'activation_norm_length', 'activation_embed'). I0420 08:18:47.811638 133563095811904 nnx_wrappers.py:437] Unknown Physical: bfloat16[32,2048,2048]...................................... ('fsdp', None, None). I0420 08:18:48.196182 133563095811904 attentions.py:1088] attentions/inputs_q Logical: bfloat16[32,2048,2048]...................................... ('activation_batch', 'activation_attn_length', 'activation_attn_embed'). I0420 08:18:48.196270 133563095811904 attentions.py:1088] attentions/inputs_q Physical: bfloat16[32,2048,2048]...................................... ('fsdp', None, None). I0420 08:18:48.212373 133563095811904 attentions.py:1089] attentions/inputs_kv Logical: bfloat16[32,2048,2048]...................................... ('activation_batch', 'activation_attn_length', 'activation_attn_embed'). I0420 08:18:48.212435 133563095811904 attentions.py:1089] attentions/inputs_kv Physical: bfloat16[32,2048,2048]...................................... ('fsdp', None, None). I0420 08:18:48.235906 133563095811904 attentions.py:1154] attentions/query Logical: bfloat16[32,2048,16,128].................................... ('activation_kv_batch', 'activation_attn_length', 'activation_kv_heads', 'activation_kv_head_dim'). I0420 08:18:48.235971 133563095811904 attentions.py:1154] attentions/query Physical: bfloat16[32,2048,16,128].................................... ('fsdp', None, None, None). I0420 08:18:48.252157 133563095811904 attentions.py:1155] attentions/key Logical: bfloat16[32,2048,16,128].................................... ('activation_kv_batch', 'activation_attn_length', 'activation_kv_heads', 'activation_kv_head_dim'). I0420 08:18:48.252223 133563095811904 attentions.py:1155] attentions/key Physical: bfloat16[32,2048,16,128].................................... ('fsdp', None, None, None). I0420 08:18:48.268370 133563095811904 attentions.py:1156] attentions/value Logical: bfloat16[32,2048,16,128].................................... ('activation_kv_batch', 'activation_attn_length', 'activation_kv_heads', 'activation_kv_head_dim'). I0420 08:18:48.268428 133563095811904 attentions.py:1156] attentions/value Physical: bfloat16[32,2048,16,128].................................... ('fsdp', None, None, None). I0420 08:18:48.292756 133563095811904 attentions.py:1197] attentions/out Logical: bfloat16[32,2048,16,128].................................... ('activation_batch', 'activation_attn_length', 'activation_heads', 'activation_kv'). I0420 08:18:48.292823 133563095811904 attentions.py:1197] attentions/out Physical: bfloat16[32,2048,16,128].................................... ('fsdp', None, None, None). I0420 08:18:48.314337 133563095811904 linears.py:525] linears/x Logical: bfloat16[32,2048,7168]...................................... ('activation_batch', 'activation_length', 'activation_mlp'). I0420 08:18:48.314407 133563095811904 linears.py:525] linears/x Physical: bfloat16[32,2048,7168]...................................... ('fsdp', None, None). I0420 08:18:48.529643 133563095811904 checkpointing.py:578] checkpoint manager exists so trying to load this run's existing checkpoint I0420 08:18:48.529767 133563095811904 checkpointing.py:676] No existing checkpoints found, not restoring checkpoint. fsdp: 32 I0420 08:18:49.962475 133563095811904 maxtext_utils.py:1821] params/params/decoder/decoder_norm/scale Shape: float32[2048] Logical: P('norm',) Physical: (None,) I0420 08:18:49.962599 133563095811904 maxtext_utils.py:1821] params/params/decoder/layers/mlp/wi_0/kernel Shape: float32[2048,16,7168] Logical: P('embed', 'layers', 'mlp') Physical: ('fsdp', None, None) I0420 08:18:49.962651 133563095811904 maxtext_utils.py:1821] params/params/decoder/layers/mlp/wi_1/kernel Shape: float32[2048,16,7168] Logical: P('embed', 'layers', 'mlp') Physical: ('fsdp', None, None) I0420 08:18:49.962737 133563095811904 maxtext_utils.py:1821] params/params/decoder/layers/mlp/wo/kernel Shape: float32[7168,16,2048] Logical: P('mlp', 'layers', 'embed') Physical: (None, None, 'fsdp') I0420 08:18:49.962799 133563095811904 maxtext_utils.py:1821] params/params/decoder/layers/post_self_attention_layer_norm/scale Shape: float32[2048,16] Logical: P('norm', 'layers') Physical: (None, None) I0420 08:18:49.962844 133563095811904 maxtext_utils.py:1821] params/params/decoder/layers/pre_self_attention_layer_norm/scale Shape: float32[2048,16] Logical: P('norm', 'layers') Physical: (None, None) I0420 08:18:49.962896 133563095811904 maxtext_utils.py:1821] 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 08:18:49.962948 133563095811904 maxtext_utils.py:1821] 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 08:18:49.962989 133563095811904 maxtext_utils.py:1821] 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 08:18:49.963026 133563095811904 maxtext_utils.py:1821] 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 08:18:49.963073 133563095811904 maxtext_utils.py:1821] params/params/decoder/logits_dense/kernel Shape: float32[2048,32000] Logical: P('embed_vocab', 'vocab') Physical: ('fsdp', None) I0420 08:18:49.963120 133563095811904 maxtext_utils.py:1821] params/params/token_embedder/embedding Shape: float32[32000,2048] Logical: P('vocab', 'embed_vocab') Physical: (None, 'fsdp') I0420 08:18:50.450308 133563095811904 train.py:157] train/xent Logical: float32[32,2048]............................................ ('activation_embed_and_logits_batch', 'activation_length'). I0420 08:18:50.450400 133563095811904 train.py:157] train/xent Physical: float32[32,2048]............................................ ('fsdp', None). I0420 08:18:50.465603 133563095811904 train.py:164] train/z_loss Logical: float32[32,2048]............................................ ('activation_embed_and_logits_batch', 'activation_length'). I0420 08:18:50.465665 133563095811904 train.py:164] train/z_loss Physical: float32[32,2048]............................................ ('fsdp', None). I0420 08:19:01.163746 133563095811904 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. I0420 08:19:01.164540 133563095811904 metric_logger.py:301] number parameters: 1.104 billion I0420 08:19:12.470362 133563095811904 checkpointing.py:794] Waiting for step 0 to finish before checkpoint... I0420 08:19:12.664999 133563095811904 checkpointing.py:798] Waited 0.1946125030517578 seconds for step 0 to finish before starting checkpointing. I0420 08:19:12.667622 133563095811904 checkpoint_manager.py:2009] [process=1][thread=MainThread][wait_until_finished] No Save Finalize thread to wait for. Returning. I0420 08:19:12.669363 133563095811904 checkpoint_manager.py:1512] [process=1] Saving checkpoint at step 0 I0420 08:19:12.671683 133563095811904 event_tracking.py:70] [process=1] [async] Started save checkpoint @ gs://lance-maxtext/linen_ckpt_xpk_feat_nnx_trainstate_and_training_loop_20260420_081533/linen_xpk_feat_nnx_trainstate_and_training_loop_20260420_081533_02_synthetic/checkpoints/0. I0420 08:19:13.439321 133563095811904 signaling_client.py:364] Using JaxDistributedSignalingClient I0420 08:19:13.440197 133563095811904 jax_array_handlers.py:360] Scheduling D2H of 39 prioritized jax.Array. I0420 08:19:13.440253 133563095811904 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 08:19:13.716171 133563095811904 base_pytree_checkpoint_handler.py:154] [process=1][thread=MainThread] Initiated "orbax.checkpoint._src.serialization.jax_array_handlers.ArrayHandler".serialize. Time taken: 0.276919s I0420 08:19:13.716346 133563095811904 base_pytree_checkpoint_handler.py:130] [process=1] /jax/orbax/write/blocking_gbytes_per_sec: 5.465 GiB/s (total gbytes: 1.5 GiB) (time elapsed: 0.2822580337524414 s) (per-host) I0420 08:19:13.716398 133563095811904 base_pytree_checkpoint_handler.py:768] [process=1][thread=MainThread] Initiated Pytree async_save. Time taken: 0.282320s (batch_requests_ready=0.002145s, total_serialization_initiated=0.280106s, others=0.000070s) I0420 08:19:13.716491 133563095811904 composite_checkpoint_handler.py:715] [process=1][thread=MainThread] Initiated CompositeCheckpointHandler.async_save. Time taken: 0.286641s (all_items=0.000018s, per_item={'items': '0.00001764'}, temp_paths=0.286624) I0420 08:19:13.717288 133563095811904 event_tracking.py:125] [process=1] [async] Finished blocking save in 1.05 seconds. Continuing save @ gs://lance-maxtext/linen_ckpt_xpk_feat_nnx_trainstate_and_training_loop_20260420_081533/linen_xpk_feat_nnx_trainstate_and_training_loop_20260420_081533_02_synthetic/checkpoints/0. I0420 08:19:13.717619 133434421888768 async_checkpointer.py:76] [process=1][thread=async_save] Background save thread started. Deadline for this save operation is 2026-04-20 08:39:13.717581 I0420 08:19:13.726696 133563095811904 checkpoint_manager.py:1560] [process=1][thread=MainThread][step=0] Starting CheckpointManager Save Finalize thread=save_finalize I0420 08:19:13.727011 133432822724352 async_checkpointer.py:280] [process=1][thread=save_finalize] Waiting for background save thread=async_save. I0420 08:19:13.727151 133563095811904 standard_logger.py:34] {'step': 0, 'event_type': 'save', 'directory': 'gs://lance-maxtext/linen_ckpt_xpk_feat_nnx_trainstate_and_training_loop_20260420_081533/linen_xpk_feat_nnx_trainstate_and_training_loop_20260420_081533_02_synthetic/checkpoints', 'reached_preemption': False, 'preemption_received_at': None, 'synchronous': False, 'wait_for_prev_start_time': 1776673152.667604, 'wait_for_prev_duration_secs': 6.079673767089844e-05, 'time_between_consecutive_saves_sec': None, 'checkpointer_blocking_start_time': 1776673152.669401, 'checkpointer_blocking_duration_secs': 1.048386573791504, 'get_old_steps_start_time': 1776673153.7178137, 'get_old_steps_duration_secs': 3.314018249511719e-05, 'checkpoint_manager_blocking_start_time': 1776673152.665523, 'checkpoint_manager_blocking_duration_secs': 1.061586856842041} I0420 08:19:13.727257 133563095811904 checkpointing.py:409] Started an asynchronous checkpoint save for step 0 I0420 08:19:13.727306 133563095811904 max_utils.py:750] Memstats: After params initialized: I0420 08:19:13.727355 133563095811904 max_utils.py:756] Using (GB) 0.43 / 31.25 (1.376000%) on TPU_2(process=1,(2,0,0,0)) I0420 08:19:13.727387 133563095811904 max_utils.py:756] Using (GB) 0.43 / 31.25 (1.376000%) on TPU_3(process=1,(3,0,0,0)) I0420 08:19:13.727413 133563095811904 max_utils.py:756] Using (GB) 0.43 / 31.25 (1.376000%) on TPU_6(process=1,(2,1,0,0)) I0420 08:19:13.727436 133563095811904 max_utils.py:756] Using (GB) 0.43 / 31.25 (1.376000%) on TPU_7(process=1,(3,1,0,0)) I0420 08:19:14.036653 133563095811904 metric_logger.py:196] completed step: 0, seconds: 11.306, TFLOP/s/device: 1.202, Tokens/s/device: 181.148, total_weights: 65536, loss: 10.874, lm_loss: 10.874, perplexity: 52779.875 I0420 08:19:14.220574 133563095811904 metric_logger.py:196] completed step: 1, seconds: 1.565, TFLOP/s/device: 8.684, Tokens/s/device: 1308.874, total_weights: 65536, loss: 10.874, lm_loss: 10.874, perplexity: 52779.875 I0420 08:19:14.648771 133563095811904 metric_logger.py:196] completed step: 2, seconds: 0.037, TFLOP/s/device: 369.366, Tokens/s/device: 55674.867, total_weights: 65536, loss: 10.262, lm_loss: 10.262, perplexity: 28637.129 I0420 08:19:14.795969 133563095811904 metric_logger.py:196] completed step: 3, seconds: 0.428, TFLOP/s/device: 31.711, Tokens/s/device: 4779.865, total_weights: 65536, loss: 9.734, lm_loss: 9.734, perplexity: 16889.203 I0420 08:19:15.091006 133563095811904 metric_logger.py:196] completed step: 4, seconds: 0.153, TFLOP/s/device: 89.006, Tokens/s/device: 13415.960, total_weights: 65536, loss: 9.277, lm_loss: 9.277, perplexity: 10694.614 I0420 08:19:15.096551 133563095811904 metric_logger.py:196] completed step: 5, seconds: 0.147, TFLOP/s/device: 92.291, Tokens/s/device: 13911.153, total_weights: 65536, loss: 8.892, lm_loss: 8.892, perplexity: 7270.717 I0420 08:19:17.172547 2856 google_auth_provider.cc:181] Running on GCE, using service account 562977990677-compute@developer.gserviceaccount.com I0420 08:19:19.167034 133433305765632 array_metadata_store.py:203] [process=1][thread=array_type_handler] Wrote 39 array_metadata.ArrayMetadata to gs://lance-maxtext/linen_ckpt_xpk_feat_nnx_trainstate_and_training_loop_20260420_081533/linen_xpk_feat_nnx_trainstate_and_training_loop_20260420_081533_02_synthetic/checkpoints/0/items/array_metadatas/process_1 I0420 08:19:37.075579 133563095811904 metric_logger.py:196] completed step: 6, seconds: 0.295, TFLOP/s/device: 45.996, Tokens/s/device: 6933.066, total_weights: 65536, loss: 8.592, lm_loss: 8.592, perplexity: 5390.759 I0420 08:19:37.222809 133563095811904 metric_logger.py:196] completed step: 7, seconds: 21.832, TFLOP/s/device: 0.622, Tokens/s/device: 93.806, total_weights: 65536, loss: 8.384, lm_loss: 8.384, perplexity: 4376.905 I0420 08:19:37.370023 133563095811904 metric_logger.py:196] completed step: 8, seconds: 0.152, TFLOP/s/device: 89.387, Tokens/s/device: 13473.330, total_weights: 65536, loss: 8.255, lm_loss: 8.255, perplexity: 3846.406 I0420 08:19:37.517115 133563095811904 checkpointing.py:794] Waiting for step 9 to finish before checkpoint... I0420 08:19:37.517684 133563095811904 checkpointing.py:798] Waited 0.0005896091461181641 seconds for step 9 to finish before starting checkpointing. I0420 08:19:37.519836 133563095811904 checkpoint_manager.py:2020] [process=1][thread=MainThread][step=0][wait_until_finished] Waiting for Save Finalize thread (save_finalize) to complete. I0420 08:19:51.426030 133434421888768 base_pytree_checkpoint_handler.py:130] [process=1] /jax/orbax/write/gbytes_per_sec: 41.577 MiB/s (total gbytes: 1.5 GiB) (time elapsed: 37.99188804626465 s) (per-host) I0420 08:19:51.426161 133434421888768 async_checkpointer.py:90] [process=1][thread=async_save] 3 Handler Commit operations completed. Time taken: 37.708415s. I0420 08:19:59.868571 133434421888768 async_checkpointer.py:160] [process=1][thread=async_save] Background save thread done. Time taken: 46.150809s. I0420 08:19:59.868884 133432822724352 async_checkpointer.py:288] [process=1][thread=save_finalize] Done with waiting for background save thread=async_save. I0420 08:19:59.869001 133432822724352 async_checkpointer.py:298] [process=1][thread=save_finalize] No errors found in background save thread=async_save. I0420 08:19:59.869052 133432822724352 checkpoint_manager.py:2137] [process=1][thread=save_finalize][step=0] CheckpointManager Save Finalize is syncing with other hosts... I0420 08:19:59.870635 133432822724352 checkpoint_manager.py:2146] [process=1][thread=save_finalize][step=0] CheckpointManager Save Finalize is done on all hosts. I0420 08:19:59.870782 133563095811904 checkpoint_manager.py:2032] [process=1][thread=MainThread][step=0][wait_until_finished] Done waiting for Save Finalize thread (save_finalize) running at step=0. W0420 08:19:59.870864 133563095811904 checkpoint_manager.py:1452] Waiting for previous save to complete took 22.351032 seconds. If this number is high, consider checkpointing less frequently. I0420 08:19:59.872984 133563095811904 checkpoint_manager.py:1512] [process=1] Saving checkpoint at step 9 I0420 08:19:59.875945 133563095811904 event_tracking.py:70] [process=1] [async] Started save checkpoint @ gs://lance-maxtext/linen_ckpt_xpk_feat_nnx_trainstate_and_training_loop_20260420_081533/linen_xpk_feat_nnx_trainstate_and_training_loop_20260420_081533_02_synthetic/checkpoints/9. I0420 08:20:00.584295 133563095811904 jax_array_handlers.py:360] Scheduling D2H of 39 prioritized jax.Array. I0420 08:20:00.584386 133563095811904 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 08:20:00.618774 133563095811904 base_pytree_checkpoint_handler.py:154] [process=1][thread=MainThread] Initiated "orbax.checkpoint._src.serialization.jax_array_handlers.ArrayHandler".serialize. Time taken: 0.035740s I0420 08:20:00.618913 133563095811904 base_pytree_checkpoint_handler.py:130] [process=1] /jax/orbax/write/blocking_gbytes_per_sec: 39.480 GiB/s (total gbytes: 1.5 GiB) (time elapsed: 0.03907203674316406 s) (per-host) I0420 08:20:00.618961 133563095811904 base_pytree_checkpoint_handler.py:768] [process=1][thread=MainThread] Initiated Pytree async_save. Time taken: 0.039128s (batch_requests_ready=0.001729s, total_serialization_initiated=0.037337s, others=0.000062s) I0420 08:20:00.619040 133563095811904 composite_checkpoint_handler.py:715] [process=1][thread=MainThread] Initiated CompositeCheckpointHandler.async_save. Time taken: 0.043271s (all_items=0.000019s, per_item={'items': '0.00001860'}, temp_paths=0.043253) I0420 08:20:00.619752 133563095811904 event_tracking.py:125] [process=1] [async] Finished blocking save in 0.75 seconds. Continuing save @ gs://lance-maxtext/linen_ckpt_xpk_feat_nnx_trainstate_and_training_loop_20260420_081533/linen_xpk_feat_nnx_trainstate_and_training_loop_20260420_081533_02_synthetic/checkpoints/9. I0420 08:20:00.620112 133432822724352 async_checkpointer.py:76] [process=1][thread=async_save] Background save thread started. Deadline for this save operation is 2026-04-20 08:40:00.620074 I0420 08:20:00.631208 133563095811904 checkpoint_manager.py:1560] [process=1][thread=MainThread][step=9] Starting CheckpointManager Save Finalize thread=save_finalize I0420 08:20:00.631468 133434401539840 async_checkpointer.py:280] [process=1][thread=save_finalize] Waiting for background save thread=async_save. I0420 08:20:00.631627 133563095811904 standard_logger.py:34] {'step': 9, 'event_type': 'save', 'directory': 'gs://lance-maxtext/linen_ckpt_xpk_feat_nnx_trainstate_and_training_loop_20260420_081533/linen_xpk_feat_nnx_trainstate_and_training_loop_20260420_081533_02_synthetic/checkpoints', 'reached_preemption': False, 'preemption_received_at': None, 'synchronous': False, 'wait_for_prev_start_time': 1776673177.5198057, 'wait_for_prev_duration_secs': 22.3510320186615, 'time_between_consecutive_saves_sec': None, 'checkpointer_blocking_start_time': 1776673199.873023, 'checkpointer_blocking_duration_secs': 0.7472336292266846, 'get_old_steps_start_time': 1776673200.620281, 'get_old_steps_duration_secs': 3.0040740966796875e-05, 'checkpoint_manager_blocking_start_time': 1776673177.5179265, 'checkpoint_manager_blocking_duration_secs': 23.113667488098145} I0420 08:20:00.631746 133563095811904 checkpointing.py:409] Started an asynchronous checkpoint save for step 9 I0420 08:20:00.631793 133563095811904 checkpoint_manager.py:2020] [process=1][thread=MainThread][step=9][wait_until_finished] Waiting for Save Finalize thread (save_finalize) to complete. I0420 08:20:05.822526 133434409932544 array_metadata_store.py:203] [process=1][thread=array_type_handler] Wrote 39 array_metadata.ArrayMetadata to gs://lance-maxtext/linen_ckpt_xpk_feat_nnx_trainstate_and_training_loop_20260420_081533/linen_xpk_feat_nnx_trainstate_and_training_loop_20260420_081533_02_synthetic/checkpoints/9/items/array_metadatas/process_1 I0420 08:20:42.615544 133432822724352 base_pytree_checkpoint_handler.py:130] [process=1] /jax/orbax/write/gbytes_per_sec: 37.578 MiB/s (total gbytes: 1.5 GiB) (time elapsed: 42.03563928604126 s) (per-host) I0420 08:20:42.615669 133432822724352 async_checkpointer.py:90] [process=1][thread=async_save] 3 Handler Commit operations completed. Time taken: 41.995442s. I0420 08:20:50.439882 133432822724352 async_checkpointer.py:160] [process=1][thread=async_save] Background save thread done. Time taken: 49.819639s. I0420 08:20:50.440193 133434401539840 async_checkpointer.py:288] [process=1][thread=save_finalize] Done with waiting for background save thread=async_save. I0420 08:20:50.440308 133434401539840 async_checkpointer.py:298] [process=1][thread=save_finalize] No errors found in background save thread=async_save. I0420 08:20:50.440354 133434401539840 checkpoint_manager.py:2137] [process=1][thread=save_finalize][step=9] CheckpointManager Save Finalize is syncing with other hosts... I0420 08:20:50.441770 133434401539840 checkpoint_manager.py:2146] [process=1][thread=save_finalize][step=9] CheckpointManager Save Finalize is done on all hosts. I0420 08:20:50.441949 133563095811904 checkpoint_manager.py:2032] [process=1][thread=MainThread][step=9][wait_until_finished] Done waiting for Save Finalize thread (save_finalize) running at step=9. I0420 08:20:50.442104 133563095811904 checkpoint_manager.py:2009] [process=1][thread=MainThread][wait_until_finished] No Save Finalize thread to wait for. Returning. I0420 08:20:50.443107 133563095811904 metric_logger.py:196] completed step: 9, seconds: 0.147, TFLOP/s/device: 92.222, Tokens/s/device: 13900.767, 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: Mon Apr 20 08:20:59 UTC 2026 EXIT_CODE=0