main (574ad3fb9) vs feat/nnx-post-train-fixes (574ad3fb9)| Metric | main 574ad3fb9 | feat/nnx-post-train-fixes 574ad3fb9 | Diff (feat/nnx-post-train-fixes − main) |
|---|---|---|---|
| Parameters | 1.104 billion | 1.104 billion | — |
| Final loss | 5.3910 | 5.3910 | 0 |
| TFLOP/s | 92.211 | 92.214 | +0.003 |
| Tok/s | 13899.1 | 13899.5 | +0.472 |
| Avg s/step | 8.318 | 8.613 | +0.295 |
| 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 07:21:07 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 07:21:31.198255: 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 07:21:31.723657 138625162618688 max_utils.py:273] Attempting to initialize the jax distributed system... I0420 07:21:40.763985 138625162618688 distributed.py:149] Starting JAX distributed service on [::]:8482 I0420 07:21:40.766244 138625162618688 distributed.py:172] Connecting to JAX distributed service on mt-16-zarr3-true-rs-s58kf-slice-job-0-0.mt-16-zarr3-true-rs-s58kf:8482 I0420 07:21:42.049892 138625162618688 max_utils.py:284] Jax distributed system initialized! I0420 07:21:48.246217 138625162618688 max_utils.py:800] System Information: Jax Version: 0.9.2 I0420 07:21:48.246319 138625162618688 max_utils.py:801] System Information: Jaxlib Version: 0.9.2 I0420 07:21:48.246378 138625162618688 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 07:21:48.246416 138625162618688 train_utils.py:348] WARNING: Sequence packing is essentially ignored for synthetic data. Please use a real dataset to use sequence packing. I0420 07:21:48.938263 138625162618688 maxtext_utils.py:1551] Num_devices: 32, shape (1, 1, 1, 32, 1, 1, 1, 1, 1, 1, 1, 1, 1) I0420 07:21:48.938565 138625162618688 checkpointing.py:677] Setting up checkpoint logger... I0420 07:21:48.938620 138625162618688 checkpointing.py:233] Creating checkpoint manager with ocdbt=True and zarr3=True I0420 07:21:48.938663 138625162618688 pytree_checkpoint_handler.py:592] save_device_host_concurrent_bytes=None I0420 07:21:48.939043 138625162618688 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 0x7e1377d48110>, enable_pinned_host_transfer=False, save_concurrent_bytes: 96000000000 (89.4 GiB), restore_concurrent_bytes: 96000000000 (89.4 GiB) I0420 07:21:50.205137 138625162618688 checkpointing.py:265] Enabling policy for fixed interval checkpointing. I0420 07:21:50.205477 138625162618688 checkpoint_manager.py:708] [process=4][thread=MainThread] CheckpointManager init: checkpointers=None, item_names=('items',), item_handlers={'items': <orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeCheckpointHandler object at 0x7dff380709b0>}, handler_registry=None I0420 07:21:50.205835 138625162618688 composite_checkpoint_handler.py:237] Deferred registration for item: "items". Adding handler `<orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeCheckpointHandler object at 0x7dff380709b0>` 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 07:21:50.205889 138625162618688 composite_checkpoint_handler.py:237] Deferred registration for item: "metrics". Adding handler `<orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonCheckpointHandler object at 0x7dff187a4110>` 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 07:21:50.205926 138625162618688 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 0x7dff380709b0>, ('items', <class 'orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeRestoreArgs'>): <orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeCheckpointHandler object at 0x7dff380709b0>, ('metrics', <class 'orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonSaveArgs'>): <orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonCheckpointHandler object at 0x7dff187a4110>, ('metrics', <class 'orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonRestoreArgs'>): <orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonCheckpointHandler object at 0x7dff187a4110>}). I0420 07:21:50.206245 138625162618688 abstract_checkpointer.py:35] orbax-checkpoint version: 0.11.34 I0420 07:21:50.206327 138625162618688 async_checkpointer.py:192] [process=4][thread=MainThread] Using barrier_sync_fn: <function get_barrier_sync_fn.<locals>._fn at 0x7dff183df880> timeout: 1200 secs and primary_host=0 for async checkpoint writes I0420 07:21:52.133888 138503123871488 checkpoint.py:228] Read Metadata={'item_handlers': {'items': 'orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeCheckpointHandler'}, 'metrics': {}, 'performance_metrics': {}, 'init_timestamp_nsecs': 1776669448364875749, 'commit_timestamp_nsecs': 1776669490729619239, 'custom_metadata': {}} from gs://lance-maxtext/linen_ckpt_xpk_main_20260420_060028/linen_xpk_main_20260420_060028_16_zarr3_true/checkpoints/0/_CHECKPOINT_METADATA I0420 07:21:52.382230 138503123871488 checkpoint_manager.py:1871] Missing metrics for step 0 I0420 07:21:52.546398 138625162618688 checkpoint_manager.py:1812] Found 3 checkpoint steps in gs://lance-maxtext/linen_ckpt_xpk_main_20260420_060028/linen_xpk_main_20260420_060028_16_zarr3_true/checkpoints I0420 07:21:53.240026 138625162618688 checkpoint_manager.py:929] [process=4][thread=MainThread] CheckpointManager created, primary_host=0, CheckpointManagerOptions=CheckpointManagerOptions(save_interval_steps=1, max_to_keep=None, keep_time_interval=None, keep_period=None, should_keep_fn=None, best_fn=None, best_mode='max', keep_checkpoints_without_metrics=True, step_prefix=None, step_format_fixed_length=None, step_name_format=None, create=True, cleanup_tmp_directories=False, save_on_steps=frozenset(), single_host_load_and_broadcast=False, todelete_subdir=None, todelete_full_path=None, enable_background_delete=False, read_only=False, enable_async_checkpointing=True, async_options=None, multiprocessing_options=MultiprocessingOptions(primary_host=0, active_processes=None, barrier_sync_key_prefix=None), should_save_fn=None, file_options=FileOptions(path_permission_mode=None), save_root_metadata=True, temporary_path_class=None, save_decision_policy=FixedIntervalPolicy(interval=5), preservation_policy=LatestN(n=None), prevent_write_metrics=False, enable_should_save_is_saving_in_progress_check=True, enable_per_process_directory_creation=False, lightweight_initialize=False), root_directory=gs://lance-maxtext/linen_ckpt_xpk_main_20260420_060028/linen_xpk_main_20260420_060028_16_zarr3_true/checkpoints: <orbax.checkpoint.checkpoint_manager.CheckpointManager object at 0x7dff187a2540> I0420 07:21:53.240196 138625162618688 checkpointing.py:301] Checkpoint manager created! I0420 07:21:54.168537 138625162618688 nnx_wrappers.py:437] Unknown Logical: bfloat16[32,2048,2048]...................................... ('activation_batch', 'activation_norm_length', 'activation_embed'). I0420 07:21:54.168654 138625162618688 nnx_wrappers.py:437] Unknown Physical: bfloat16[32,2048,2048]...................................... ('fsdp', None, None). I0420 07:21:54.546042 138625162618688 attentions.py:1088] attentions/inputs_q Logical: bfloat16[32,2048,2048]...................................... ('activation_batch', 'activation_attn_length', 'activation_attn_embed'). I0420 07:21:54.546131 138625162618688 attentions.py:1088] attentions/inputs_q Physical: bfloat16[32,2048,2048]...................................... ('fsdp', None, None). I0420 07:21:54.562330 138625162618688 attentions.py:1089] attentions/inputs_kv Logical: bfloat16[32,2048,2048]...................................... ('activation_batch', 'activation_attn_length', 'activation_attn_embed'). I0420 07:21:54.562403 138625162618688 attentions.py:1089] attentions/inputs_kv Physical: bfloat16[32,2048,2048]...................................... ('fsdp', None, None). I0420 07:21:54.586753 138625162618688 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 07:21:54.586820 138625162618688 attentions.py:1154] attentions/query Physical: bfloat16[32,2048,16,128].................................... ('fsdp', None, None, None). I0420 07:21:54.603081 138625162618688 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 07:21:54.603135 138625162618688 attentions.py:1155] attentions/key Physical: bfloat16[32,2048,16,128].................................... ('fsdp', None, None, None). I0420 07:21:54.619269 138625162618688 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 07:21:54.619328 138625162618688 attentions.py:1156] attentions/value Physical: bfloat16[32,2048,16,128].................................... ('fsdp', None, None, None). I0420 07:21:54.643829 138625162618688 attentions.py:1197] attentions/out Logical: bfloat16[32,2048,16,128].................................... ('activation_batch', 'activation_attn_length', 'activation_heads', 'activation_kv'). I0420 07:21:54.643903 138625162618688 attentions.py:1197] attentions/out Physical: bfloat16[32,2048,16,128].................................... ('fsdp', None, None, None). I0420 07:21:54.664796 138625162618688 linears.py:525] linears/x Logical: bfloat16[32,2048,7168]...................................... ('activation_batch', 'activation_length', 'activation_mlp'). I0420 07:21:54.664857 138625162618688 linears.py:525] linears/x Physical: bfloat16[32,2048,7168]...................................... ('fsdp', None, None). I0420 07:21:54.879772 138625162618688 checkpointing.py:577] checkpoint manager exists so trying to load this run's existing checkpoint I0420 07:21:54.879882 138625162618688 checkpointing.py:581] restoring from this run's directory step 9 I0420 07:21:55.132125 138625162618688 event_tracking.py:70] [process=4] [sync] Started load checkpoint @ gs://lance-maxtext/linen_ckpt_xpk_main_20260420_060028/linen_xpk_main_20260420_060028_16_zarr3_true/checkpoints/9. I0420 07:21:55.814858 138625162618688 checkpointer.py:307] Restoring checkpoint from gs://lance-maxtext/linen_ckpt_xpk_main_20260420_060028/linen_xpk_main_20260420_060028_16_zarr3_true/checkpoints/9. I0420 07:21:55.815278 138625162618688 event_tracking.py:125] [process=4] [sync] Finished blocking load in 0.68 seconds. Continuing load @ gs://lance-maxtext/linen_ckpt_xpk_main_20260420_060028/linen_xpk_main_20260420_060028_16_zarr3_true/checkpoints/9. I0420 07:21:57.198985 1463 google_auth_provider.cc:181] Running on GCE, using service account 562977990677-compute@developer.gserviceaccount.com I0420 07:22:03.278620 138625162618688 jax_array_handlers.py:843] [process=4] /jax/orbax/read/worker/io/requested throughput: 259.685 MiB/s (total gbytes: 1.5 GiB) (time elapsed: 6.085346937179565 s) (per-host) I0420 07:22:03.283064 138625162618688 base_pytree_checkpoint_handler.py:130] [process=4] /jax/checkpoint/read/gbytes_per_sec: 244.125 MiB/s (total gbytes: 1.5 GiB) (time elapsed: 6.482720851898193 s) (per-host) I0420 07:22:04.268560 138625162618688 event_tracking.py:138] [process=4] [sync] Finished load in 9.14 seconds @ gs://lance-maxtext/linen_ckpt_xpk_main_20260420_060028/linen_xpk_main_20260420_060028_16_zarr3_true/checkpoints/9 I0420 07:22:04.268713 138625162618688 standard_logger.py:34] {'step': 9, 'event_type': 'restore', 'directory': 'gs://lance-maxtext/linen_ckpt_xpk_main_20260420_060028/linen_xpk_main_20260420_060028_16_zarr3_true/checkpoints', 'checkpointer_start_time': 1776669715.1320245, 'checkpointer_duration_secs': 9.136656999588013, 'checkpoint_manager_start_time': 1776669714.8801172, 'checkpoint_manager_duration_secs': 9.388566970825195} fsdp: 32 I0420 07:22:04.678521 138625162618688 maxtext_utils.py:1654] params/params/decoder/decoder_norm/scale Shape: float32[2048] Logical: P('norm',) Physical: (None,) I0420 07:22:04.678649 138625162618688 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 07:22:04.678703 138625162618688 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 07:22:04.678760 138625162618688 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 07:22:04.678812 138625162618688 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 07:22:04.678854 138625162618688 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 07:22:04.678907 138625162618688 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 07:22:04.678961 138625162618688 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 07:22:04.679002 138625162618688 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 07:22:04.679039 138625162618688 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 07:22:04.679087 138625162618688 maxtext_utils.py:1654] params/params/decoder/logits_dense/kernel Shape: float32[2048,32000] Logical: P('embed_vocab', 'vocab') Physical: ('fsdp', None) I0420 07:22:04.679134 138625162618688 maxtext_utils.py:1654] params/params/token_embedder/embedding Shape: float32[32000,2048] Logical: P('vocab', 'embed_vocab') Physical: (None, 'fsdp') I0420 07:22:05.158703 138625162618688 train.py:155] train/xent Logical: float32[32,2048]............................................ ('activation_embed_and_logits_batch', 'activation_length'). I0420 07:22:05.158798 138625162618688 train.py:155] train/xent Physical: float32[32,2048]............................................ ('fsdp', None). I0420 07:22:05.174182 138625162618688 train.py:162] train/z_loss Logical: float32[32,2048]............................................ ('activation_embed_and_logits_batch', 'activation_length'). I0420 07:22:05.174239 138625162618688 train.py:162] train/z_loss Physical: float32[32,2048]............................................ ('fsdp', None). I0420 07:22:16.249839 138625162618688 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 07:22:16.250630 138625162618688 metric_logger.py:301] number parameters: 1.104 billion I0420 07:22:27.511813 138625162618688 checkpointing.py:772] Waiting for step 10 to finish before checkpoint... I0420 07:22:27.666837 138625162618688 checkpointing.py:776] Waited 0.15500450134277344 seconds for step 10 to finish before starting checkpointing. I0420 07:22:27.669228 138625162618688 checkpoint_manager.py:2009] [process=4][thread=MainThread][wait_until_finished] No Save Finalize thread to wait for. Returning. I0420 07:22:27.671289 138625162618688 checkpoint_manager.py:1512] [process=4] Saving checkpoint at step 10 I0420 07:22:27.672688 138625162618688 event_tracking.py:70] [process=4] [async] Started save checkpoint @ gs://lance-maxtext/linen_ckpt_xpk_main_20260420_060028/linen_xpk_main_20260420_060028_16_zarr3_true/checkpoints/10. I0420 07:22:28.003225 138625162618688 signaling_client.py:364] Using JaxDistributedSignalingClient I0420 07:22:28.004155 138625162618688 jax_array_handlers.py:360] Scheduling D2H of 39 prioritized jax.Array. I0420 07:22:28.004213 138625162618688 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 07:22:28.279022 138625162618688 base_pytree_checkpoint_handler.py:154] [process=4][thread=MainThread] Initiated "orbax.checkpoint._src.serialization.jax_array_handlers.ArrayHandler".serialize. Time taken: 0.275870s I0420 07:22:28.279190 138625162618688 base_pytree_checkpoint_handler.py:130] [process=4] /jax/orbax/write/blocking_gbytes_per_sec: 5.511 GiB/s (total gbytes: 1.5 GiB) (time elapsed: 0.27989912033081055 s) (per-host) I0420 07:22:28.279244 138625162618688 base_pytree_checkpoint_handler.py:768] [process=4][thread=MainThread] Initiated Pytree async_save. Time taken: 0.279963s (batch_requests_ready=0.002140s, total_serialization_initiated=0.277750s, others=0.000072s) I0420 07:22:28.279345 138625162618688 composite_checkpoint_handler.py:715] [process=4][thread=MainThread] Initiated CompositeCheckpointHandler.async_save. Time taken: 0.284276s (all_items=0.000019s, per_item={'items': '0.00001860'}, temp_paths=0.284257) I0420 07:22:28.280032 138625162618688 event_tracking.py:125] [process=4] [async] Finished blocking save in 0.61 seconds. Continuing save @ gs://lance-maxtext/linen_ckpt_xpk_main_20260420_060028/linen_xpk_main_20260420_060028_16_zarr3_true/checkpoints/10. I0420 07:22:28.280388 138498266887936 async_checkpointer.py:76] [process=4][thread=async_save] Background save thread started. Deadline for this save operation is 2026-04-20 07:42:28.280347 I0420 07:22:28.287830 138625162618688 checkpoint_manager.py:1560] [process=4][thread=MainThread][step=10] Starting CheckpointManager Save Finalize thread=save_finalize I0420 07:22:28.288129 138502536677120 async_checkpointer.py:280] [process=4][thread=save_finalize] Waiting for background save thread=async_save. I0420 07:22:28.288278 138625162618688 standard_logger.py:34] {'step': 10, 'event_type': 'save', 'directory': 'gs://lance-maxtext/linen_ckpt_xpk_main_20260420_060028/linen_xpk_main_20260420_060028_16_zarr3_true/checkpoints', 'reached_preemption': False, 'preemption_received_at': None, 'synchronous': False, 'wait_for_prev_start_time': 1776669747.6692092, 'wait_for_prev_duration_secs': 6.222724914550781e-05, 'time_between_consecutive_saves_sec': None, 'checkpointer_blocking_start_time': 1776669747.6713288, 'checkpointer_blocking_duration_secs': 0.609220027923584, 'get_old_steps_start_time': 1776669748.280573, 'get_old_steps_duration_secs': 3.218650817871094e-05, 'checkpoint_manager_blocking_start_time': 1776669747.6673496, 'checkpoint_manager_blocking_duration_secs': 0.6208961009979248} I0420 07:22:28.288398 138625162618688 checkpointing.py:408] Started an asynchronous checkpoint save for step 10 I0420 07:22:28.288450 138625162618688 max_utils.py:750] Memstats: After params initialized: I0420 07:22:28.288499 138625162618688 max_utils.py:756] Using (GB) 0.43 / 31.25 (1.376000%) on TPU_16(process=4,(0,4,0,0)) I0420 07:22:28.288532 138625162618688 max_utils.py:756] Using (GB) 0.43 / 31.25 (1.376000%) on TPU_17(process=4,(1,4,0,0)) I0420 07:22:28.288557 138625162618688 max_utils.py:756] Using (GB) 0.43 / 31.25 (1.376000%) on TPU_20(process=4,(0,5,0,0)) I0420 07:22:28.288581 138625162618688 max_utils.py:756] Using (GB) 0.43 / 31.25 (1.376000%) on TPU_21(process=4,(1,5,0,0)) I0420 07:22:28.597245 138625162618688 metric_logger.py:196] completed step: 10, seconds: 11.261, TFLOP/s/device: 1.207, Tokens/s/device: 181.866, total_weights: 65536, loss: 6.833, lm_loss: 6.833, perplexity: 927.869 I0420 07:22:28.767947 138625162618688 metric_logger.py:196] completed step: 11, seconds: 1.084, TFLOP/s/device: 12.536, Tokens/s/device: 1889.541, total_weights: 65536, loss: 6.527, lm_loss: 6.527, perplexity: 683.659 I0420 07:22:28.915512 138625162618688 metric_logger.py:196] completed step: 12, seconds: 0.023, TFLOP/s/device: 578.323, Tokens/s/device: 87171.193, total_weights: 65536, loss: 6.265, lm_loss: 6.265, perplexity: 526.068 I0420 07:22:29.062962 138625162618688 metric_logger.py:196] completed step: 13, seconds: 0.153, TFLOP/s/device: 88.619, Tokens/s/device: 13357.683, total_weights: 65536, loss: 6.045, lm_loss: 6.045, perplexity: 422.001 I0420 07:22:29.067406 138625162618688 checkpointing.py:772] Waiting for step 15 to finish before checkpoint... I0420 07:22:29.357245 138625162618688 checkpointing.py:776] Waited 0.28980112075805664 seconds for step 15 to finish before starting checkpointing. I0420 07:22:29.360301 138625162618688 checkpoint_manager.py:2020] [process=4][thread=MainThread][step=10][wait_until_finished] Waiting for Save Finalize thread (save_finalize) to complete. I0420 07:22:33.600286 138500951242496 array_metadata_store.py:203] [process=4][thread=array_type_handler] Wrote 39 array_metadata.ArrayMetadata to gs://lance-maxtext/linen_ckpt_xpk_main_20260420_060028/linen_xpk_main_20260420_060028_16_zarr3_true/checkpoints/10/items/array_metadatas/process_4 I0420 07:23:09.616220 138498266887936 base_pytree_checkpoint_handler.py:130] [process=4] /jax/orbax/write/gbytes_per_sec: 37.956 MiB/s (total gbytes: 1.5 GiB) (time elapsed: 41.616905212402344 s) (per-host) I0420 07:23:09.616323 138498266887936 async_checkpointer.py:90] [process=4][thread=async_save] 3 Handler Commit operations completed. Time taken: 41.335817s. I0420 07:23:17.810713 138498266887936 async_checkpointer.py:160] [process=4][thread=async_save] Background save thread done. Time taken: 49.530189s. I0420 07:23:17.810942 138502536677120 async_checkpointer.py:288] [process=4][thread=save_finalize] Done with waiting for background save thread=async_save. I0420 07:23:17.810998 138502536677120 async_checkpointer.py:298] [process=4][thread=save_finalize] No errors found in background save thread=async_save. I0420 07:23:17.811043 138502536677120 checkpoint_manager.py:2137] [process=4][thread=save_finalize][step=10] CheckpointManager Save Finalize is syncing with other hosts... I0420 07:23:17.812750 138502536677120 checkpoint_manager.py:2146] [process=4][thread=save_finalize][step=10] CheckpointManager Save Finalize is done on all hosts. I0420 07:23:17.812909 138625162618688 checkpoint_manager.py:2032] [process=4][thread=MainThread][step=10][wait_until_finished] Done waiting for Save Finalize thread (save_finalize) running at step=10. W0420 07:23:17.813046 138625162618688 checkpoint_manager.py:1452] Waiting for previous save to complete took 48.452744 seconds. If this number is high, consider checkpointing less frequently. I0420 07:23:17.815104 138625162618688 checkpoint_manager.py:1512] [process=4] Saving checkpoint at step 15 I0420 07:23:17.817465 138625162618688 event_tracking.py:70] [process=4] [async] Started save checkpoint @ gs://lance-maxtext/linen_ckpt_xpk_main_20260420_060028/linen_xpk_main_20260420_060028_16_zarr3_true/checkpoints/15. I0420 07:23:18.103400 138625162618688 jax_array_handlers.py:360] Scheduling D2H of 39 prioritized jax.Array. I0420 07:23:18.103492 138625162618688 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 07:23:18.146310 138625162618688 base_pytree_checkpoint_handler.py:154] [process=4][thread=MainThread] Initiated "orbax.checkpoint._src.serialization.jax_array_handlers.ArrayHandler".serialize. Time taken: 0.044072s I0420 07:23:18.146489 138625162618688 base_pytree_checkpoint_handler.py:130] [process=4] /jax/orbax/write/blocking_gbytes_per_sec: 32.467 GiB/s (total gbytes: 1.5 GiB) (time elapsed: 0.04751229286193848 s) (per-host) I0420 07:23:18.146539 138625162618688 base_pytree_checkpoint_handler.py:768] [process=4][thread=MainThread] Initiated Pytree async_save. Time taken: 0.047571s (batch_requests_ready=0.001765s, total_serialization_initiated=0.045741s, others=0.000065s) I0420 07:23:18.146628 138625162618688 composite_checkpoint_handler.py:715] [process=4][thread=MainThread] Initiated CompositeCheckpointHandler.async_save. Time taken: 0.052589s (all_items=0.000013s, per_item={'items': '0.00001311'}, temp_paths=0.052576) I0420 07:23:18.147271 138625162618688 event_tracking.py:125] [process=4] [async] Finished blocking save in 0.33 seconds. Continuing save @ gs://lance-maxtext/linen_ckpt_xpk_main_20260420_060028/linen_xpk_main_20260420_060028_16_zarr3_true/checkpoints/15. I0420 07:23:18.147664 138500951242496 async_checkpointer.py:76] [process=4][thread=async_save] Background save thread started. Deadline for this save operation is 2026-04-20 07:43:18.147626 I0420 07:23:18.149592 138625162618688 checkpoint_manager.py:1560] [process=4][thread=MainThread][step=15] Starting CheckpointManager Save Finalize thread=save_finalize I0420 07:23:18.149903 138502536677120 async_checkpointer.py:280] [process=4][thread=save_finalize] Waiting for background save thread=async_save. I0420 07:23:18.150034 138625162618688 standard_logger.py:34] {'step': 15, 'event_type': 'save', 'directory': 'gs://lance-maxtext/linen_ckpt_xpk_main_20260420_060028/linen_xpk_main_20260420_060028_16_zarr3_true/checkpoints', 'reached_preemption': False, 'preemption_received_at': None, 'synchronous': False, 'wait_for_prev_start_time': 1776669749.3602722, 'wait_for_prev_duration_secs': 48.45274353027344, 'time_between_consecutive_saves_sec': None, 'checkpointer_blocking_start_time': 1776669797.8151424, 'checkpointer_blocking_duration_secs': 0.3326723575592041, 'get_old_steps_start_time': 1776669798.147838, 'get_old_steps_duration_secs': 3.0040740966796875e-05, 'checkpoint_manager_blocking_start_time': 1776669749.3578653, 'checkpoint_manager_blocking_duration_secs': 48.792139291763306} I0420 07:23:18.150211 138625162618688 checkpointing.py:408] Started an asynchronous checkpoint save for step 15 I0420 07:23:18.150996 138625162618688 metric_logger.py:196] completed step: 14, seconds: 0.147, TFLOP/s/device: 92.157, Tokens/s/device: 13890.867, total_weights: 65536, loss: 5.864, lm_loss: 5.864, perplexity: 352.193 I0420 07:23:18.156947 138625162618688 metric_logger.py:196] completed step: 15, seconds: 0.147, TFLOP/s/device: 92.206, Tokens/s/device: 13898.220, total_weights: 65536, loss: 5.719, lm_loss: 5.719, perplexity: 304.661 I0420 07:23:23.243679 138495020484352 array_metadata_store.py:203] [process=4][thread=array_type_handler] Wrote 39 array_metadata.ArrayMetadata to gs://lance-maxtext/linen_ckpt_xpk_main_20260420_060028/linen_xpk_main_20260420_060028_16_zarr3_true/checkpoints/15/items/array_metadatas/process_4 I0420 07:23:39.282000 138625162618688 metric_logger.py:196] completed step: 16, seconds: 49.088, TFLOP/s/device: 0.277, Tokens/s/device: 41.721, total_weights: 65536, loss: 5.605, lm_loss: 5.605, perplexity: 271.872 I0420 07:23:39.429329 138625162618688 metric_logger.py:196] completed step: 17, seconds: 20.979, TFLOP/s/device: 0.648, Tokens/s/device: 97.623, total_weights: 65536, loss: 5.517, lm_loss: 5.517, perplexity: 248.975 I0420 07:23:39.576690 138625162618688 metric_logger.py:196] completed step: 18, seconds: 0.152, TFLOP/s/device: 89.154, Tokens/s/device: 13438.232, total_weights: 65536, loss: 5.448, lm_loss: 5.448, perplexity: 232.333 I0420 07:23:39.723434 138625162618688 checkpointing.py:772] Waiting for step 19 to finish before checkpoint... I0420 07:23:39.724094 138625162618688 checkpointing.py:776] Waited 0.0006814002990722656 seconds for step 19 to finish before starting checkpointing. I0420 07:23:39.726281 138625162618688 checkpoint_manager.py:2020] [process=4][thread=MainThread][step=15][wait_until_finished] Waiting for Save Finalize thread (save_finalize) to complete. I0420 07:23:59.974196 138500951242496 base_pytree_checkpoint_handler.py:130] [process=4] /jax/orbax/write/gbytes_per_sec: 37.722 MiB/s (total gbytes: 1.5 GiB) (time elapsed: 41.87518787384033 s) (per-host) I0420 07:23:59.974314 138500951242496 async_checkpointer.py:90] [process=4][thread=async_save] 3 Handler Commit operations completed. Time taken: 41.826536s. I0420 07:24:08.559727 138500951242496 async_checkpointer.py:160] [process=4][thread=async_save] Background save thread done. Time taken: 50.411932s. I0420 07:24:08.560040 138502536677120 async_checkpointer.py:288] [process=4][thread=save_finalize] Done with waiting for background save thread=async_save. I0420 07:24:08.560161 138502536677120 async_checkpointer.py:298] [process=4][thread=save_finalize] No errors found in background save thread=async_save. I0420 07:24:08.560210 138502536677120 checkpoint_manager.py:2137] [process=4][thread=save_finalize][step=15] CheckpointManager Save Finalize is syncing with other hosts... I0420 07:24:08.561813 138502536677120 checkpoint_manager.py:2146] [process=4][thread=save_finalize][step=15] CheckpointManager Save Finalize is done on all hosts. I0420 07:24:08.562013 138625162618688 checkpoint_manager.py:2032] [process=4][thread=MainThread][step=15][wait_until_finished] Done waiting for Save Finalize thread (save_finalize) running at step=15. W0420 07:24:08.562139 138625162618688 checkpoint_manager.py:1452] Waiting for previous save to complete took 28.835865 seconds. If this number is high, consider checkpointing less frequently. I0420 07:24:08.563518 138625162618688 checkpoint_manager.py:1512] [process=4] Saving checkpoint at step 19 I0420 07:24:08.565608 138625162618688 event_tracking.py:70] [process=4] [async] Started save checkpoint @ gs://lance-maxtext/linen_ckpt_xpk_main_20260420_060028/linen_xpk_main_20260420_060028_16_zarr3_true/checkpoints/19. I0420 07:24:09.277420 138625162618688 jax_array_handlers.py:360] Scheduling D2H of 39 prioritized jax.Array. I0420 07:24:09.277516 138625162618688 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 07:24:09.303776 138625162618688 base_pytree_checkpoint_handler.py:154] [process=4][thread=MainThread] Initiated "orbax.checkpoint._src.serialization.jax_array_handlers.ArrayHandler".serialize. Time taken: 0.027415s I0420 07:24:09.303943 138625162618688 base_pytree_checkpoint_handler.py:130] [process=4] /jax/orbax/write/blocking_gbytes_per_sec: 49.890 GiB/s (total gbytes: 1.5 GiB) (time elapsed: 0.030919313430786133 s) (per-host) I0420 07:24:09.303994 138625162618688 base_pytree_checkpoint_handler.py:768] [process=4][thread=MainThread] Initiated Pytree async_save. Time taken: 0.030980s (batch_requests_ready=0.001663s, total_serialization_initiated=0.029251s, others=0.000066s) I0420 07:24:09.304081 138625162618688 composite_checkpoint_handler.py:715] [process=4][thread=MainThread] Initiated CompositeCheckpointHandler.async_save. Time taken: 0.035183s (all_items=0.000011s, per_item={'items': '0.00001121'}, temp_paths=0.035172) I0420 07:24:09.304819 138625162618688 event_tracking.py:125] [process=4] [async] Finished blocking save in 0.74 seconds. Continuing save @ gs://lance-maxtext/linen_ckpt_xpk_main_20260420_060028/linen_xpk_main_20260420_060028_16_zarr3_true/checkpoints/19. I0420 07:24:09.305135 138502536677120 async_checkpointer.py:76] [process=4][thread=async_save] Background save thread started. Deadline for this save operation is 2026-04-20 07:44:09.305098 I0420 07:24:09.307751 138625162618688 checkpoint_manager.py:1560] [process=4][thread=MainThread][step=19] Starting CheckpointManager Save Finalize thread=save_finalize I0420 07:24:09.308026 138494995306240 async_checkpointer.py:280] [process=4][thread=save_finalize] Waiting for background save thread=async_save. I0420 07:24:09.308179 138625162618688 standard_logger.py:34] {'step': 19, 'event_type': 'save', 'directory': 'gs://lance-maxtext/linen_ckpt_xpk_main_20260420_060028/linen_xpk_main_20260420_060028_16_zarr3_true/checkpoints', 'reached_preemption': False, 'preemption_received_at': None, 'synchronous': False, 'wait_for_prev_start_time': 1776669819.7262526, 'wait_for_prev_duration_secs': 28.835864782333374, 'time_between_consecutive_saves_sec': 21.91346263885498, 'checkpointer_blocking_start_time': 1776669848.563556, 'checkpointer_blocking_duration_secs': 0.7417206764221191, 'get_old_steps_start_time': 1776669849.3052943, 'get_old_steps_duration_secs': 2.9087066650390625e-05, 'checkpoint_manager_blocking_start_time': 1776669819.7243135, 'checkpoint_manager_blocking_duration_secs': 29.583833694458008} I0420 07:24:09.308282 138625162618688 checkpointing.py:408] Started an asynchronous checkpoint save for step 19 I0420 07:24:09.308326 138625162618688 checkpoint_manager.py:2020] [process=4][thread=MainThread][step=19][wait_until_finished] Waiting for Save Finalize thread (save_finalize) to complete. I0420 07:24:14.425625 138495020484352 array_metadata_store.py:203] [process=4][thread=array_type_handler] Wrote 39 array_metadata.ArrayMetadata to gs://lance-maxtext/linen_ckpt_xpk_main_20260420_060028/linen_xpk_main_20260420_060028_16_zarr3_true/checkpoints/19/items/array_metadatas/process_4 I0420 07:24:53.068471 138502536677120 base_pytree_checkpoint_handler.py:130] [process=4] /jax/orbax/write/gbytes_per_sec: 36.068 MiB/s (total gbytes: 1.5 GiB) (time elapsed: 43.79540729522705 s) (per-host) I0420 07:24:53.068600 138502536677120 async_checkpointer.py:90] [process=4][thread=async_save] 3 Handler Commit operations completed. Time taken: 43.763354s. I0420 07:24:58.424109 138502536677120 async_checkpointer.py:160] [process=4][thread=async_save] Background save thread done. Time taken: 49.118848s. I0420 07:24:58.424425 138494995306240 async_checkpointer.py:288] [process=4][thread=save_finalize] Done with waiting for background save thread=async_save. I0420 07:24:58.424553 138494995306240 async_checkpointer.py:298] [process=4][thread=save_finalize] No errors found in background save thread=async_save. I0420 07:24:58.424601 138494995306240 checkpoint_manager.py:2137] [process=4][thread=save_finalize][step=19] CheckpointManager Save Finalize is syncing with other hosts... I0420 07:24:58.426188 138494995306240 checkpoint_manager.py:2146] [process=4][thread=save_finalize][step=19] CheckpointManager Save Finalize is done on all hosts. I0420 07:24:58.426359 138625162618688 checkpoint_manager.py:2032] [process=4][thread=MainThread][step=19][wait_until_finished] Done waiting for Save Finalize thread (save_finalize) running at step=19. I0420 07:24:58.426504 138625162618688 checkpoint_manager.py:2009] [process=4][thread=MainThread][wait_until_finished] No Save Finalize thread to wait for. Returning. I0420 07:24:58.427465 138625162618688 metric_logger.py:196] completed step: 19, seconds: 0.147, TFLOP/s/device: 92.211, Tokens/s/device: 13899.069, total_weights: 65536, loss: 5.391, lm_loss: 5.391, perplexity: 219.349 Per train step: Total TFLOPs: 13.59 split as 93.93% learnable weight flops and 6.07% attention flops XPK End: Mon Apr 20 07:25:07 UTC 2026 EXIT_CODE=0
XPK Start: Mon Apr 20 17:19:31 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 17:19:55.424280: 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 17:19:55.603208 135780113520448 max_utils.py:273] Attempting to initialize the jax distributed system... I0420 17:20:04.643183 135780113520448 distributed.py:149] Starting JAX distributed service on [::]:8482 I0420 17:20:04.645477 135780113520448 distributed.py:172] Connecting to JAX distributed service on mt-16-zarr3-true-rs-sxnik-slice-job-0-0.mt-16-zarr3-true-rs-sxnik:8482 I0420 17:20:05.674493 135780113520448 max_utils.py:284] Jax distributed system initialized! I0420 17:20:11.894048 135780113520448 max_utils.py:800] System Information: Jax Version: 0.9.2 I0420 17:20:11.894245 135780113520448 max_utils.py:801] System Information: Jaxlib Version: 0.9.2 I0420 17:20:11.894293 135780113520448 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 17:20:11.894339 135780113520448 train_utils.py:378] WARNING: Sequence packing is essentially ignored for synthetic data. Please use a real dataset to use sequence packing. I0420 17:20:12.578742 135780113520448 maxtext_utils.py:1718] Num_devices: 32, shape (1, 1, 1, 32, 1, 1, 1, 1, 1, 1, 1, 1, 1) I0420 17:20:12.579345 135780113520448 maxtext_utils.py:1718] Num_devices: 32, shape (1, 1, 1, 32, 1, 1, 1, 1, 1, 1, 1, 1, 1) I0420 17:20:12.579519 135780113520448 checkpointing.py:688] Setting up checkpoint logger... I0420 17:20:12.579582 135780113520448 checkpointing.py:234] Creating checkpoint manager with ocdbt=True and zarr3=True I0420 17:20:12.579627 135780113520448 pytree_checkpoint_handler.py:592] save_device_host_concurrent_bytes=None I0420 17:20:12.579949 135780113520448 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 0x7b7d0db50fb0>, enable_pinned_host_transfer=False, save_concurrent_bytes: 96000000000 (89.4 GiB), restore_concurrent_bytes: 96000000000 (89.4 GiB) I0420 17:20:13.916945 135780113520448 checkpointing.py:266] Enabling policy for fixed interval checkpointing. I0420 17:20:13.917191 135780113520448 checkpoint_manager.py:708] [process=0][thread=MainThread] CheckpointManager init: checkpointers=None, item_names=('items',), item_handlers={'items': <orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeCheckpointHandler object at 0x7b68a8643dd0>}, handler_registry=None I0420 17:20:13.917429 135780113520448 composite_checkpoint_handler.py:237] Deferred registration for item: "items". Adding handler `<orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeCheckpointHandler object at 0x7b68a8643dd0>` 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 17:20:13.917489 135780113520448 composite_checkpoint_handler.py:237] Deferred registration for item: "metrics". Adding handler `<orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonCheckpointHandler object at 0x7b7c0177d640>` 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 17:20:13.917533 135780113520448 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 0x7b68a8643dd0>, ('items', <class 'orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeRestoreArgs'>): <orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeCheckpointHandler object at 0x7b68a8643dd0>, ('metrics', <class 'orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonSaveArgs'>): <orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonCheckpointHandler object at 0x7b7c0177d640>, ('metrics', <class 'orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonRestoreArgs'>): <orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonCheckpointHandler object at 0x7b7c0177d640>}). I0420 17:20:13.917856 135780113520448 abstract_checkpointer.py:35] orbax-checkpoint version: 0.11.34 I0420 17:20:13.917925 135780113520448 async_checkpointer.py:192] [process=0][thread=MainThread] Using barrier_sync_fn: <function get_barrier_sync_fn.<locals>._fn at 0x7b68a83e7740> timeout: 1200 secs and primary_host=0 for async checkpoint writes I0420 17:20:16.532021 135657984866048 checkpoint.py:228] Read Metadata={'item_handlers': {'items': 'orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeCheckpointHandler'}, 'metrics': {}, 'performance_metrics': {}, 'init_timestamp_nsecs': 1776705447245166102, 'commit_timestamp_nsecs': 1776705496241505819, 'custom_metadata': {}} from gs://lance-maxtext/linen_ckpt_xpk_feat_nnx_post_train_fixes_20260420_153543/linen_xpk_feat_nnx_post_train_fixes_20260420_153543_16_zarr3_true/checkpoints/9/_CHECKPOINT_METADATA I0420 17:20:16.793997 135657984866048 checkpoint_manager.py:1871] Missing metrics for step 9 I0420 17:20:16.807336 135780113520448 checkpoint_manager.py:1812] Found 3 checkpoint steps in gs://lance-maxtext/linen_ckpt_xpk_feat_nnx_post_train_fixes_20260420_153543/linen_xpk_feat_nnx_post_train_fixes_20260420_153543_16_zarr3_true/checkpoints I0420 17:20:17.280507 135780113520448 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=5), preservation_policy=LatestN(n=None), prevent_write_metrics=False, enable_should_save_is_saving_in_progress_check=True, enable_per_process_directory_creation=False, lightweight_initialize=False), root_directory=gs://lance-maxtext/linen_ckpt_xpk_feat_nnx_post_train_fixes_20260420_153543/linen_xpk_feat_nnx_post_train_fixes_20260420_153543_16_zarr3_true/checkpoints: <orbax.checkpoint.checkpoint_manager.CheckpointManager object at 0x7b68a8646c90> I0420 17:20:17.280678 135780113520448 checkpointing.py:302] Checkpoint manager created! I0420 17:20:18.196312 135780113520448 nnx_wrappers.py:437] Unknown Logical: bfloat16[32,2048,2048]...................................... ('activation_batch', 'activation_norm_length', 'activation_embed'). I0420 17:20:18.196414 135780113520448 nnx_wrappers.py:437] Unknown Physical: bfloat16[32,2048,2048]...................................... ('fsdp', None, None). I0420 17:20:18.573093 135780113520448 attentions.py:1088] attentions/inputs_q Logical: bfloat16[32,2048,2048]...................................... ('activation_batch', 'activation_attn_length', 'activation_attn_embed'). I0420 17:20:18.573179 135780113520448 attentions.py:1088] attentions/inputs_q Physical: bfloat16[32,2048,2048]...................................... ('fsdp', None, None). I0420 17:20:18.589859 135780113520448 attentions.py:1089] attentions/inputs_kv Logical: bfloat16[32,2048,2048]...................................... ('activation_batch', 'activation_attn_length', 'activation_attn_embed'). I0420 17:20:18.589921 135780113520448 attentions.py:1089] attentions/inputs_kv Physical: bfloat16[32,2048,2048]...................................... ('fsdp', None, None). I0420 17:20:18.613998 135780113520448 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 17:20:18.614066 135780113520448 attentions.py:1154] attentions/query Physical: bfloat16[32,2048,16,128].................................... ('fsdp', None, None, None). I0420 17:20:18.630710 135780113520448 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 17:20:18.630770 135780113520448 attentions.py:1155] attentions/key Physical: bfloat16[32,2048,16,128].................................... ('fsdp', None, None, None). I0420 17:20:18.647381 135780113520448 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 17:20:18.647441 135780113520448 attentions.py:1156] attentions/value Physical: bfloat16[32,2048,16,128].................................... ('fsdp', None, None, None). I0420 17:20:18.672559 135780113520448 attentions.py:1197] attentions/out Logical: bfloat16[32,2048,16,128].................................... ('activation_batch', 'activation_attn_length', 'activation_heads', 'activation_kv'). I0420 17:20:18.672633 135780113520448 attentions.py:1197] attentions/out Physical: bfloat16[32,2048,16,128].................................... ('fsdp', None, None, None). I0420 17:20:18.693564 135780113520448 linears.py:525] linears/x Logical: bfloat16[32,2048,7168]...................................... ('activation_batch', 'activation_length', 'activation_mlp'). I0420 17:20:18.693636 135780113520448 linears.py:525] linears/x Physical: bfloat16[32,2048,7168]...................................... ('fsdp', None, None). I0420 17:20:18.909207 135780113520448 checkpointing.py:578] checkpoint manager exists so trying to load this run's existing checkpoint I0420 17:20:18.909326 135780113520448 checkpointing.py:582] restoring from this run's directory step 9 I0420 17:20:19.159428 135780113520448 event_tracking.py:70] [process=0] [sync] Started load checkpoint @ gs://lance-maxtext/linen_ckpt_xpk_feat_nnx_post_train_fixes_20260420_153543/linen_xpk_feat_nnx_post_train_fixes_20260420_153543_16_zarr3_true/checkpoints/9. I0420 17:20:19.448510 135780113520448 checkpointer.py:307] Restoring checkpoint from gs://lance-maxtext/linen_ckpt_xpk_feat_nnx_post_train_fixes_20260420_153543/linen_xpk_feat_nnx_post_train_fixes_20260420_153543_16_zarr3_true/checkpoints/9. I0420 17:20:19.448920 135780113520448 event_tracking.py:125] [process=0] [sync] Finished blocking load in 0.29 seconds. Continuing load @ gs://lance-maxtext/linen_ckpt_xpk_feat_nnx_post_train_fixes_20260420_153543/linen_xpk_feat_nnx_post_train_fixes_20260420_153543_16_zarr3_true/checkpoints/9. I0420 17:20:20.899608 1462 google_auth_provider.cc:181] Running on GCE, using service account 562977990677-compute@developer.gserviceaccount.com I0420 17:20:27.380373 135780113520448 jax_array_handlers.py:843] [process=0] /jax/orbax/read/worker/io/requested throughput: 243.610 MiB/s (total gbytes: 1.5 GiB) (time elapsed: 6.486887216567993 s) (per-host) I0420 17:20:27.385311 135780113520448 base_pytree_checkpoint_handler.py:130] [process=0] /jax/checkpoint/read/gbytes_per_sec: 229.631 MiB/s (total gbytes: 1.5 GiB) (time elapsed: 6.891902685165405 s) (per-host) I0420 17:20:28.665537 135780113520448 event_tracking.py:138] [process=0] [sync] Finished load in 9.51 seconds @ gs://lance-maxtext/linen_ckpt_xpk_feat_nnx_post_train_fixes_20260420_153543/linen_xpk_feat_nnx_post_train_fixes_20260420_153543_16_zarr3_true/checkpoints/9 I0420 17:20:28.665701 135780113520448 standard_logger.py:34] {'step': 9, 'event_type': 'restore', 'directory': 'gs://lance-maxtext/linen_ckpt_xpk_feat_nnx_post_train_fixes_20260420_153543/linen_xpk_feat_nnx_post_train_fixes_20260420_153543_16_zarr3_true/checkpoints', 'checkpointer_start_time': 1776705619.1593308, 'checkpointer_duration_secs': 9.506339073181152, 'checkpoint_manager_start_time': 1776705618.9095976, 'checkpoint_manager_duration_secs': 9.756074905395508} fsdp: 32 I0420 17:20:29.077651 135780113520448 maxtext_utils.py:1827] params/params/decoder/decoder_norm/scale Shape: float32[2048] Logical: P('norm',) Physical: (None,) I0420 17:20:29.077775 135780113520448 maxtext_utils.py:1827] params/params/decoder/layers/mlp/wi_0/kernel Shape: float32[2048,16,7168] Logical: P('embed', 'layers', 'mlp') Physical: ('fsdp', None, None) I0420 17:20:29.077835 135780113520448 maxtext_utils.py:1827] params/params/decoder/layers/mlp/wi_1/kernel Shape: float32[2048,16,7168] Logical: P('embed', 'layers', 'mlp') Physical: ('fsdp', None, None) I0420 17:20:29.077896 135780113520448 maxtext_utils.py:1827] params/params/decoder/layers/mlp/wo/kernel Shape: float32[7168,16,2048] Logical: P('mlp', 'layers', 'embed') Physical: (None, None, 'fsdp') I0420 17:20:29.077950 135780113520448 maxtext_utils.py:1827] params/params/decoder/layers/post_self_attention_layer_norm/scale Shape: float32[2048,16] Logical: P('norm', 'layers') Physical: (None, None) I0420 17:20:29.077991 135780113520448 maxtext_utils.py:1827] params/params/decoder/layers/pre_self_attention_layer_norm/scale Shape: float32[2048,16] Logical: P('norm', 'layers') Physical: (None, None) I0420 17:20:29.078046 135780113520448 maxtext_utils.py:1827] 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 17:20:29.078141 135780113520448 maxtext_utils.py:1827] 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 17:20:29.078190 135780113520448 maxtext_utils.py:1827] 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 17:20:29.078230 135780113520448 maxtext_utils.py:1827] 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 17:20:29.078280 135780113520448 maxtext_utils.py:1827] params/params/decoder/logits_dense/kernel Shape: float32[2048,32000] Logical: P('embed_vocab', 'vocab') Physical: ('fsdp', None) I0420 17:20:29.078329 135780113520448 maxtext_utils.py:1827] params/params/token_embedder/embedding Shape: float32[32000,2048] Logical: P('vocab', 'embed_vocab') Physical: (None, 'fsdp') I0420 17:20:29.562225 135780113520448 train.py:157] train/xent Logical: float32[32,2048]............................................ ('activation_embed_and_logits_batch', 'activation_length'). I0420 17:20:29.562316 135780113520448 train.py:157] train/xent Physical: float32[32,2048]............................................ ('fsdp', None). I0420 17:20:29.577645 135780113520448 train.py:164] train/z_loss Logical: float32[32,2048]............................................ ('activation_embed_and_logits_batch', 'activation_length'). I0420 17:20:29.577707 135780113520448 train.py:164] train/z_loss Physical: float32[32,2048]............................................ ('fsdp', None). I0420 17:20:40.519809 135780113520448 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 17:20:40.528168 135780113520448 metric_logger.py:301] number parameters: 1.104 billion I0420 17:20:52.103987 135780113520448 checkpointing.py:794] Waiting for step 10 to finish before checkpoint... I0420 17:20:52.257978 135780113520448 checkpointing.py:798] Waited 0.1539771556854248 seconds for step 10 to finish before starting checkpointing. I0420 17:20:52.261204 135780113520448 checkpoint_manager.py:2009] [process=0][thread=MainThread][wait_until_finished] No Save Finalize thread to wait for. Returning. I0420 17:20:52.263478 135780113520448 checkpoint_manager.py:1512] [process=0] Saving checkpoint at step 10 I0420 17:20:52.265268 135780113520448 event_tracking.py:70] [process=0] [async] Started save checkpoint @ gs://lance-maxtext/linen_ckpt_xpk_feat_nnx_post_train_fixes_20260420_153543/linen_xpk_feat_nnx_post_train_fixes_20260420_153543_16_zarr3_true/checkpoints/10. I0420 17:20:52.580795 135780113520448 signaling_client.py:364] Using JaxDistributedSignalingClient I0420 17:20:52.593535 135780113520448 jax_array_handlers.py:360] Scheduling D2H of 39 prioritized jax.Array. I0420 17:20:52.593601 135780113520448 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 17:20:52.843225 135656835622656 atomicity.py:140] Creating tmp directory gs://lance-maxtext/linen_ckpt_xpk_feat_nnx_post_train_fixes_20260420_153543/linen_xpk_feat_nnx_post_train_fixes_20260420_153543_16_zarr3_true/checkpoints/10 I0420 17:20:52.871054 135780113520448 base_pytree_checkpoint_handler.py:154] [process=0][thread=MainThread] Initiated "orbax.checkpoint._src.serialization.jax_array_handlers.ArrayHandler".serialize. Time taken: 0.278474s I0420 17:20:52.871800 135780113520448 base_pytree_checkpoint_handler.py:130] [process=0] /jax/orbax/write/blocking_gbytes_per_sec: 5.450 GiB/s (total gbytes: 1.5 GiB) (time elapsed: 0.28304457664489746 s) (per-host) I0420 17:20:52.871864 135780113520448 base_pytree_checkpoint_handler.py:768] [process=0][thread=MainThread] Initiated Pytree async_save. Time taken: 0.283125s (batch_requests_ready=0.002098s, total_serialization_initiated=0.280383s, others=0.000644s) I0420 17:20:52.871937 135780113520448 composite_checkpoint_handler.py:715] [process=0][thread=MainThread] Initiated CompositeCheckpointHandler.async_save. Time taken: 0.287317s (all_items=0.000016s, per_item={'items': '0.00001645'}, temp_paths=0.287301) I0420 17:20:52.872715 135780113520448 event_tracking.py:125] [process=0] [async] Finished blocking save in 0.61 seconds. Continuing save @ gs://lance-maxtext/linen_ckpt_xpk_feat_nnx_post_train_fixes_20260420_153543/linen_xpk_feat_nnx_post_train_fixes_20260420_153543_16_zarr3_true/checkpoints/10. I0420 17:20:52.873009 135657406064384 async_checkpointer.py:76] [process=0][thread=async_save] Background save thread started. Deadline for this save operation is 2026-04-20 17:40:52.872977 I0420 17:20:52.877124 135780113520448 checkpoint_manager.py:1560] [process=0][thread=MainThread][step=10] Starting CheckpointManager Save Finalize thread=save_finalize I0420 17:20:52.877362 135653622015744 async_checkpointer.py:280] [process=0][thread=save_finalize] Waiting for background save thread=async_save. I0420 17:20:52.877517 135780113520448 standard_logger.py:34] {'step': 10, 'event_type': 'save', 'directory': 'gs://lance-maxtext/linen_ckpt_xpk_feat_nnx_post_train_fixes_20260420_153543/linen_xpk_feat_nnx_post_train_fixes_20260420_153543_16_zarr3_true/checkpoints', 'reached_preemption': False, 'preemption_received_at': None, 'synchronous': False, 'wait_for_prev_start_time': 1776705652.261185, 'wait_for_prev_duration_secs': 6.127357482910156e-05, 'time_between_consecutive_saves_sec': None, 'checkpointer_blocking_start_time': 1776705652.2635624, 'checkpointer_blocking_duration_secs': 0.6095561981201172, 'get_old_steps_start_time': 1776705652.8731441, 'get_old_steps_duration_secs': 2.8848648071289062e-05, 'checkpoint_manager_blocking_start_time': 1776705652.2585068, 'checkpoint_manager_blocking_duration_secs': 0.6189768314361572} I0420 17:20:52.877622 135780113520448 checkpointing.py:409] Started an asynchronous checkpoint save for step 10 I0420 17:20:52.877672 135780113520448 max_utils.py:750] Memstats: After params initialized: I0420 17:20:52.877721 135780113520448 max_utils.py:756] Using (GB) 0.43 / 31.25 (1.376000%) on TPU_0(process=0,(0,0,0,0)) I0420 17:20:52.877758 135780113520448 max_utils.py:756] Using (GB) 0.43 / 31.25 (1.376000%) on TPU_1(process=0,(1,0,0,0)) I0420 17:20:52.877786 135780113520448 max_utils.py:756] Using (GB) 0.43 / 31.25 (1.376000%) on TPU_4(process=0,(0,1,0,0)) I0420 17:20:52.877812 135780113520448 max_utils.py:756] Using (GB) 0.43 / 31.25 (1.376000%) on TPU_5(process=0,(1,1,0,0)) I0420 17:20:53.315049 135780113520448 metric_logger.py:196] completed step: 10, seconds: 11.261, TFLOP/s/device: 1.207, Tokens/s/device: 181.867, total_weights: 65536, loss: 6.833, lm_loss: 6.833, perplexity: 927.869 I0420 17:20:53.460360 135780113520448 metric_logger.py:196] completed step: 11, seconds: 1.209, TFLOP/s/device: 11.237, Tokens/s/device: 1693.689, total_weights: 65536, loss: 6.527, lm_loss: 6.527, perplexity: 683.659 I0420 17:20:53.608000 135780113520448 metric_logger.py:196] completed step: 12, seconds: 0.012, TFLOP/s/device: 1143.698, Tokens/s/device: 172390.572, total_weights: 65536, loss: 6.265, lm_loss: 6.265, perplexity: 526.068 I0420 17:20:53.755474 135780113520448 metric_logger.py:196] completed step: 13, seconds: 0.145, TFLOP/s/device: 93.457, Tokens/s/device: 14086.929, total_weights: 65536, loss: 6.045, lm_loss: 6.045, perplexity: 422.001 I0420 17:20:53.765227 135780113520448 checkpointing.py:794] Waiting for step 15 to finish before checkpoint... I0420 17:20:54.049305 135780113520448 checkpointing.py:798] Waited 0.28391361236572266 seconds for step 15 to finish before starting checkpointing. I0420 17:20:54.052380 135780113520448 checkpoint_manager.py:2020] [process=0][thread=MainThread][step=10][wait_until_finished] Waiting for Save Finalize thread (save_finalize) to complete. I0420 17:20:55.073277 135653613623040 checkpoint.py:188] Wrote Metadata={'item_handlers': None, 'metrics': {}, 'performance_metrics': {}, 'init_timestamp_nsecs': 1776705654643534395, 'commit_timestamp_nsecs': None, 'custom_metadata': {}}, json={"item_handlers": null, "metrics": {}, "performance_metrics": {}, "init_timestamp_nsecs": 1776705654643534395, "commit_timestamp_nsecs": null, "custom_metadata": {}} to gs://lance-maxtext/linen_ckpt_xpk_feat_nnx_post_train_fixes_20260420_153543/linen_xpk_feat_nnx_post_train_fixes_20260420_153543_16_zarr3_true/checkpoints/10/_CHECKPOINT_METADATA I0420 17:20:55.340220 135654168053504 atomicity.py:140] Creating tmp directory gs://lance-maxtext/linen_ckpt_xpk_feat_nnx_post_train_fixes_20260420_153543/linen_xpk_feat_nnx_post_train_fixes_20260420_153543_16_zarr3_true/checkpoints/10/items I0420 17:20:58.336402 135653656360704 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_post_train_fixes_20260420_153543/linen_xpk_feat_nnx_post_train_fixes_20260420_153543_16_zarr3_true/checkpoints/10/items/array_metadatas/process_0 I0420 17:21:39.064876 135653647968000 base_pytree_checkpoint_handler.py:1282] [process=0][thread=write_metadata_after_commits] Commit + Array metadata written. Time taken: 42.584871s (commit=41.314606s, array_metadata_write=1.270264s) I0420 17:21:39.066213 135657406064384 base_pytree_checkpoint_handler.py:130] [process=0] /jax/orbax/write/gbytes_per_sec: 33.986 MiB/s (total gbytes: 1.5 GiB) (time elapsed: 46.477412939071655 s) (per-host) I0420 17:21:39.066385 135657406064384 async_checkpointer.py:90] [process=0][thread=async_save] 3 Handler Commit operations completed. Time taken: 46.192749s. I0420 17:21:41.013028 135657406064384 array_metadata_store.py:411] [process=0][thread=async_save] Validated ArrayMetadata from all 8 hosts. Time taken: 0.000407s. I0420 17:21:41.062487 135653613623040 checkpoint.py:247] Updated Metadata={'item_handlers': {'items': 'orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeCheckpointHandler'}, 'metrics': {}, 'performance_metrics': {}, 'init_timestamp_nsecs': 1776705654643534395, 'commit_timestamp_nsecs': None, 'custom_metadata': {}} to gs://lance-maxtext/linen_ckpt_xpk_feat_nnx_post_train_fixes_20260420_153543/linen_xpk_feat_nnx_post_train_fixes_20260420_153543_16_zarr3_true/checkpoints/10/_CHECKPOINT_METADATA I0420 17:21:41.659946 135657406064384 ocdbt_utils.py:49] Param validation support for Zarr3 will be added later (b/362328389). I0420 17:21:42.612444 135657406064384 base_pytree_checkpoint_handler.py:1406] [process=0][thread=async_save] Pytree save finalize (merge_ocdbt + ArrayMetadata validation) completed. Time taken: 2.205292s. use_zarr3=True, enable_post_merge_validation=True, directory=gs://lance-maxtext/linen_ckpt_xpk_feat_nnx_post_train_fixes_20260420_153543/linen_xpk_feat_nnx_post_train_fixes_20260420_153543_16_zarr3_true/checkpoints/10/items I0420 17:21:42.614812 135657406064384 atomicity.py:666] Finalizing gs://lance-maxtext/linen_ckpt_xpk_feat_nnx_post_train_fixes_20260420_153543/linen_xpk_feat_nnx_post_train_fixes_20260420_153543_16_zarr3_true/checkpoints/10/items I0420 17:21:43.194359 135657406064384 atomicity.py:666] Finalizing gs://lance-maxtext/linen_ckpt_xpk_feat_nnx_post_train_fixes_20260420_153543/linen_xpk_feat_nnx_post_train_fixes_20260420_153543_16_zarr3_true/checkpoints/10 I0420 17:21:44.619107 135657406064384 atomicity.py:847] [process=0][thread=async_save] Finished saving checkpoint (finalized tmp dir) to `gs://lance-maxtext/linen_ckpt_xpk_feat_nnx_post_train_fixes_20260420_153543/linen_xpk_feat_nnx_post_train_fixes_20260420_153543_16_zarr3_true/checkpoints/10`. I0420 17:21:44.620300 135657406064384 event_tracking.py:138] [process=0] [async] Finished save (blocking + background) in 52.36 seconds @ gs://lance-maxtext/linen_ckpt_xpk_feat_nnx_post_train_fixes_20260420_153543/linen_xpk_feat_nnx_post_train_fixes_20260420_153543_16_zarr3_true/checkpoints/10 I0420 17:21:44.621801 135657406064384 async_checkpointer.py:160] [process=0][thread=async_save] Background save thread done. Time taken: 51.748142s. I0420 17:21:44.622278 135653622015744 async_checkpointer.py:288] [process=0][thread=save_finalize] Done with waiting for background save thread=async_save. I0420 17:21:44.622448 135653622015744 async_checkpointer.py:298] [process=0][thread=save_finalize] No errors found in background save thread=async_save. I0420 17:21:44.622558 135653622015744 checkpoint_manager.py:2137] [process=0][thread=save_finalize][step=10] CheckpointManager Save Finalize is syncing with other hosts... I0420 17:21:44.624416 135653622015744 checkpoint_manager.py:2146] [process=0][thread=save_finalize][step=10] CheckpointManager Save Finalize is done on all hosts. I0420 17:21:44.624695 135780113520448 checkpoint_manager.py:2032] [process=0][thread=MainThread][step=10][wait_until_finished] Done waiting for Save Finalize thread (save_finalize) running at step=10. W0420 17:21:44.624822 135780113520448 checkpoint_manager.py:1452] Waiting for previous save to complete took 50.572446 seconds. If this number is high, consider checkpointing less frequently. I0420 17:21:44.626715 135780113520448 checkpoint_manager.py:1512] [process=0] Saving checkpoint at step 15 I0420 17:21:44.629285 135780113520448 event_tracking.py:70] [process=0] [async] Started save checkpoint @ gs://lance-maxtext/linen_ckpt_xpk_feat_nnx_post_train_fixes_20260420_153543/linen_xpk_feat_nnx_post_train_fixes_20260420_153543_16_zarr3_true/checkpoints/15. I0420 17:21:44.922735 135780113520448 jax_array_handlers.py:360] Scheduling D2H of 39 prioritized jax.Array. I0420 17:21:44.922875 135780113520448 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 17:21:44.952758 135780113520448 base_pytree_checkpoint_handler.py:154] [process=0][thread=MainThread] Initiated "orbax.checkpoint._src.serialization.jax_array_handlers.ArrayHandler".serialize. Time taken: 0.031063s I0420 17:21:44.953789 135780113520448 base_pytree_checkpoint_handler.py:130] [process=0] /jax/orbax/write/blocking_gbytes_per_sec: 38.225 GiB/s (total gbytes: 1.5 GiB) (time elapsed: 0.04035472869873047 s) (per-host) I0420 17:21:44.953903 135780113520448 base_pytree_checkpoint_handler.py:768] [process=0][thread=MainThread] Initiated Pytree async_save. Time taken: 0.040493s (batch_requests_ready=0.004262s, total_serialization_initiated=0.035300s, others=0.000931s) I0420 17:21:44.954033 135780113520448 composite_checkpoint_handler.py:715] [process=0][thread=MainThread] Initiated CompositeCheckpointHandler.async_save. Time taken: 0.045459s (all_items=0.000022s, per_item={'items': '0.00002193'}, temp_paths=0.045437) I0420 17:21:44.955114 135780113520448 event_tracking.py:125] [process=0] [async] Finished blocking save in 0.33 seconds. Continuing save @ gs://lance-maxtext/linen_ckpt_xpk_feat_nnx_post_train_fixes_20260420_153543/linen_xpk_feat_nnx_post_train_fixes_20260420_153543_16_zarr3_true/checkpoints/15. I0420 17:21:44.955491 135653622015744 async_checkpointer.py:76] [process=0][thread=async_save] Background save thread started. Deadline for this save operation is 2026-04-20 17:41:44.955454 I0420 17:21:44.958006 135780113520448 checkpoint_manager.py:1560] [process=0][thread=MainThread][step=15] Starting CheckpointManager Save Finalize thread=save_finalize I0420 17:21:44.958348 135653656360704 async_checkpointer.py:280] [process=0][thread=save_finalize] Waiting for background save thread=async_save. I0420 17:21:44.958534 135780113520448 standard_logger.py:34] {'step': 15, 'event_type': 'save', 'directory': 'gs://lance-maxtext/linen_ckpt_xpk_feat_nnx_post_train_fixes_20260420_153543/linen_xpk_feat_nnx_post_train_fixes_20260420_153543_16_zarr3_true/checkpoints', 'reached_preemption': False, 'preemption_received_at': None, 'synchronous': False, 'wait_for_prev_start_time': 1776705654.052336, 'wait_for_prev_duration_secs': 50.57244634628296, 'time_between_consecutive_saves_sec': None, 'checkpointer_blocking_start_time': 1776705704.6267805, 'checkpointer_blocking_duration_secs': 0.32889318466186523, 'get_old_steps_start_time': 1776705704.9557016, 'get_old_steps_duration_secs': 4.673004150390625e-05, 'checkpoint_manager_blocking_start_time': 1776705654.05, 'checkpoint_manager_blocking_duration_secs': 50.90847969055176} I0420 17:21:44.958688 135780113520448 checkpointing.py:409] Started an asynchronous checkpoint save for step 15 I0420 17:21:45.165723 135657406064384 atomicity.py:140] Creating tmp directory gs://lance-maxtext/linen_ckpt_xpk_feat_nnx_post_train_fixes_20260420_153543/linen_xpk_feat_nnx_post_train_fixes_20260420_153543_16_zarr3_true/checkpoints/15 I0420 17:21:47.368461 135656835622656 atomicity.py:140] Creating tmp directory gs://lance-maxtext/linen_ckpt_xpk_feat_nnx_post_train_fixes_20260420_153543/linen_xpk_feat_nnx_post_train_fixes_20260420_153543_16_zarr3_true/checkpoints/15/items I0420 17:21:50.700879 135654168053504 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_post_train_fixes_20260420_153543/linen_xpk_feat_nnx_post_train_fixes_20260420_153543_16_zarr3_true/checkpoints/15/items/array_metadatas/process_0 I0420 17:22:06.665413 135780113520448 metric_logger.py:196] completed step: 14, seconds: 0.148, TFLOP/s/device: 92.011, Tokens/s/device: 13868.949, total_weights: 65536, loss: 5.864, lm_loss: 5.864, perplexity: 352.193 I0420 17:22:06.674914 135780113520448 metric_logger.py:196] completed step: 15, seconds: 0.147, TFLOP/s/device: 92.441, Tokens/s/device: 13933.774, total_weights: 65536, loss: 5.719, lm_loss: 5.719, perplexity: 304.661 I0420 17:22:06.820376 135780113520448 metric_logger.py:196] completed step: 16, seconds: 72.908, TFLOP/s/device: 0.186, Tokens/s/device: 28.090, total_weights: 65536, loss: 5.605, lm_loss: 5.605, perplexity: 271.872 I0420 17:22:06.967818 135780113520448 metric_logger.py:196] completed step: 17, seconds: 0.007, TFLOP/s/device: 1869.960, Tokens/s/device: 281860.721, total_weights: 65536, loss: 5.517, lm_loss: 5.517, perplexity: 248.975 I0420 17:22:07.115092 135780113520448 metric_logger.py:196] completed step: 18, seconds: 0.146, TFLOP/s/device: 92.821, Tokens/s/device: 13990.982, total_weights: 65536, loss: 5.448, lm_loss: 5.448, perplexity: 232.333 I0420 17:22:07.261721 135780113520448 checkpointing.py:794] Waiting for step 19 to finish before checkpoint... I0420 17:22:07.262426 135780113520448 checkpointing.py:798] Waited 0.0007283687591552734 seconds for step 19 to finish before starting checkpointing. I0420 17:22:07.264420 135780113520448 checkpoint_manager.py:2020] [process=0][thread=MainThread][step=15][wait_until_finished] Waiting for Save Finalize thread (save_finalize) to complete. I0420 17:22:30.725323 135654144341760 base_pytree_checkpoint_handler.py:1282] [process=0][thread=write_metadata_after_commits] Commit + Array metadata written. Time taken: 41.822470s (commit=40.609502s, array_metadata_write=1.212969s) I0420 17:22:30.726417 135653622015744 base_pytree_checkpoint_handler.py:130] [process=0] /jax/orbax/write/gbytes_per_sec: 34.479 MiB/s (total gbytes: 1.5 GiB) (time elapsed: 45.81296491622925 s) (per-host) I0420 17:22:30.726546 135653622015744 async_checkpointer.py:90] [process=0][thread=async_save] 3 Handler Commit operations completed. Time taken: 45.770908s. I0420 17:22:32.570119 135653622015744 array_metadata_store.py:411] [process=0][thread=async_save] Validated ArrayMetadata from all 8 hosts. Time taken: 0.000224s. I0420 17:22:32.915787 135653622015744 ocdbt_utils.py:49] Param validation support for Zarr3 will be added later (b/362328389). I0420 17:22:33.874934 135653622015744 base_pytree_checkpoint_handler.py:1406] [process=0][thread=async_save] Pytree save finalize (merge_ocdbt + ArrayMetadata validation) completed. Time taken: 2.214068s. use_zarr3=True, enable_post_merge_validation=True, directory=gs://lance-maxtext/linen_ckpt_xpk_feat_nnx_post_train_fixes_20260420_153543/linen_xpk_feat_nnx_post_train_fixes_20260420_153543_16_zarr3_true/checkpoints/15/items I0420 17:22:33.876426 135653622015744 atomicity.py:666] Finalizing gs://lance-maxtext/linen_ckpt_xpk_feat_nnx_post_train_fixes_20260420_153543/linen_xpk_feat_nnx_post_train_fixes_20260420_153543_16_zarr3_true/checkpoints/15/items I0420 17:22:34.422848 135653622015744 atomicity.py:666] Finalizing gs://lance-maxtext/linen_ckpt_xpk_feat_nnx_post_train_fixes_20260420_153543/linen_xpk_feat_nnx_post_train_fixes_20260420_153543_16_zarr3_true/checkpoints/15 I0420 17:22:35.843896 135653622015744 atomicity.py:847] [process=0][thread=async_save] Finished saving checkpoint (finalized tmp dir) to `gs://lance-maxtext/linen_ckpt_xpk_feat_nnx_post_train_fixes_20260420_153543/linen_xpk_feat_nnx_post_train_fixes_20260420_153543_16_zarr3_true/checkpoints/15`. I0420 17:22:35.844731 135653622015744 event_tracking.py:138] [process=0] [async] Finished save (blocking + background) in 51.22 seconds @ gs://lance-maxtext/linen_ckpt_xpk_feat_nnx_post_train_fixes_20260420_153543/linen_xpk_feat_nnx_post_train_fixes_20260420_153543_16_zarr3_true/checkpoints/15 I0420 17:22:35.846063 135653622015744 async_checkpointer.py:160] [process=0][thread=async_save] Background save thread done. Time taken: 50.890422s. I0420 17:22:35.846315 135653656360704 async_checkpointer.py:288] [process=0][thread=save_finalize] Done with waiting for background save thread=async_save. I0420 17:22:35.846436 135653656360704 async_checkpointer.py:298] [process=0][thread=save_finalize] No errors found in background save thread=async_save. I0420 17:22:35.846503 135653656360704 checkpoint_manager.py:2137] [process=0][thread=save_finalize][step=15] CheckpointManager Save Finalize is syncing with other hosts... I0420 17:22:35.848838 135653656360704 checkpoint_manager.py:2146] [process=0][thread=save_finalize][step=15] CheckpointManager Save Finalize is done on all hosts. I0420 17:22:35.849013 135780113520448 checkpoint_manager.py:2032] [process=0][thread=MainThread][step=15][wait_until_finished] Done waiting for Save Finalize thread (save_finalize) running at step=15. W0420 17:22:35.849156 135780113520448 checkpoint_manager.py:1452] Waiting for previous save to complete took 28.584742 seconds. If this number is high, consider checkpointing less frequently. I0420 17:22:35.850666 135780113520448 checkpoint_manager.py:1512] [process=0] Saving checkpoint at step 19 I0420 17:22:35.852661 135780113520448 event_tracking.py:70] [process=0] [async] Started save checkpoint @ gs://lance-maxtext/linen_ckpt_xpk_feat_nnx_post_train_fixes_20260420_153543/linen_xpk_feat_nnx_post_train_fixes_20260420_153543_16_zarr3_true/checkpoints/19. I0420 17:22:36.143004 135780113520448 jax_array_handlers.py:360] Scheduling D2H of 39 prioritized jax.Array. I0420 17:22:36.143107 135780113520448 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 17:22:36.168546 135780113520448 base_pytree_checkpoint_handler.py:154] [process=0][thread=MainThread] Initiated "orbax.checkpoint._src.serialization.jax_array_handlers.ArrayHandler".serialize. Time taken: 0.026222s I0420 17:22:36.169332 135780113520448 base_pytree_checkpoint_handler.py:130] [process=0] /jax/orbax/write/blocking_gbytes_per_sec: 49.968 GiB/s (total gbytes: 1.5 GiB) (time elapsed: 0.03087139129638672 s) (per-host) I0420 17:22:36.169394 135780113520448 base_pytree_checkpoint_handler.py:768] [process=0][thread=MainThread] Initiated Pytree async_save. Time taken: 0.030950s (batch_requests_ready=0.001904s, total_serialization_initiated=0.028360s, others=0.000687s) I0420 17:22:36.169467 135780113520448 composite_checkpoint_handler.py:715] [process=0][thread=MainThread] Initiated CompositeCheckpointHandler.async_save. Time taken: 0.034910s (all_items=0.000016s, per_item={'items': '0.00001574'}, temp_paths=0.034894) I0420 17:22:36.170171 135780113520448 event_tracking.py:125] [process=0] [async] Finished blocking save in 0.32 seconds. Continuing save @ gs://lance-maxtext/linen_ckpt_xpk_feat_nnx_post_train_fixes_20260420_153543/linen_xpk_feat_nnx_post_train_fixes_20260420_153543_16_zarr3_true/checkpoints/19. I0420 17:22:36.170481 135653656360704 async_checkpointer.py:76] [process=0][thread=async_save] Background save thread started. Deadline for this save operation is 2026-04-20 17:42:36.170443 I0420 17:22:36.173362 135780113520448 checkpoint_manager.py:1560] [process=0][thread=MainThread][step=19] Starting CheckpointManager Save Finalize thread=save_finalize I0420 17:22:36.173579 135654176446208 async_checkpointer.py:280] [process=0][thread=save_finalize] Waiting for background save thread=async_save. I0420 17:22:36.173696 135780113520448 standard_logger.py:34] {'step': 19, 'event_type': 'save', 'directory': 'gs://lance-maxtext/linen_ckpt_xpk_feat_nnx_post_train_fixes_20260420_153543/linen_xpk_feat_nnx_post_train_fixes_20260420_153543_16_zarr3_true/checkpoints', 'reached_preemption': False, 'preemption_received_at': None, 'synchronous': False, 'wait_for_prev_start_time': 1776705727.2643845, 'wait_for_prev_duration_secs': 28.584742069244385, 'time_between_consecutive_saves_sec': 22.63990092277527, 'checkpointer_blocking_start_time': 1776705755.850705, 'checkpointer_blocking_duration_secs': 0.3199188709259033, 'get_old_steps_start_time': 1776705756.170643, 'get_old_steps_duration_secs': 2.765655517578125e-05, 'checkpoint_manager_blocking_start_time': 1776705727.2626595, 'checkpoint_manager_blocking_duration_secs': 28.911006927490234} I0420 17:22:36.173806 135780113520448 checkpointing.py:409] Started an asynchronous checkpoint save for step 19 I0420 17:22:36.173847 135780113520448 checkpoint_manager.py:2020] [process=0][thread=MainThread][step=19][wait_until_finished] Waiting for Save Finalize thread (save_finalize) to complete. I0420 17:22:36.389816 135653622015744 atomicity.py:140] Creating tmp directory gs://lance-maxtext/linen_ckpt_xpk_feat_nnx_post_train_fixes_20260420_153543/linen_xpk_feat_nnx_post_train_fixes_20260420_153543_16_zarr3_true/checkpoints/19 I0420 17:22:38.611614 135656835622656 atomicity.py:140] Creating tmp directory gs://lance-maxtext/linen_ckpt_xpk_feat_nnx_post_train_fixes_20260420_153543/linen_xpk_feat_nnx_post_train_fixes_20260420_153543_16_zarr3_true/checkpoints/19/items I0420 17:22:42.134814 135657397671680 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_post_train_fixes_20260420_153543/linen_xpk_feat_nnx_post_train_fixes_20260420_153543_16_zarr3_true/checkpoints/19/items/array_metadatas/process_0 I0420 17:23:22.662718 135656827229952 base_pytree_checkpoint_handler.py:1282] [process=0][thread=write_metadata_after_commits] Commit + Array metadata written. Time taken: 42.870893s (commit=41.628989s, array_metadata_write=1.241904s) I0420 17:23:22.663864 135653656360704 base_pytree_checkpoint_handler.py:130] [process=0] /jax/orbax/write/gbytes_per_sec: 33.951 MiB/s (total gbytes: 1.5 GiB) (time elapsed: 46.525373697280884 s) (per-host) I0420 17:23:22.664001 135653656360704 async_checkpointer.py:90] [process=0][thread=async_save] 3 Handler Commit operations completed. Time taken: 46.493409s. I0420 17:23:25.273093 135653656360704 array_metadata_store.py:411] [process=0][thread=async_save] Validated ArrayMetadata from all 8 hosts. Time taken: 0.000224s. I0420 17:23:25.339633 135653656360704 ocdbt_utils.py:49] Param validation support for Zarr3 will be added later (b/362328389). I0420 17:23:26.312453 135653656360704 base_pytree_checkpoint_handler.py:1406] [process=0][thread=async_save] Pytree save finalize (merge_ocdbt + ArrayMetadata validation) completed. Time taken: 2.277998s. use_zarr3=True, enable_post_merge_validation=True, directory=gs://lance-maxtext/linen_ckpt_xpk_feat_nnx_post_train_fixes_20260420_153543/linen_xpk_feat_nnx_post_train_fixes_20260420_153543_16_zarr3_true/checkpoints/19/items I0420 17:23:26.317415 135653656360704 atomicity.py:666] Finalizing gs://lance-maxtext/linen_ckpt_xpk_feat_nnx_post_train_fixes_20260420_153543/linen_xpk_feat_nnx_post_train_fixes_20260420_153543_16_zarr3_true/checkpoints/19/items I0420 17:23:26.921761 135653656360704 atomicity.py:666] Finalizing gs://lance-maxtext/linen_ckpt_xpk_feat_nnx_post_train_fixes_20260420_153543/linen_xpk_feat_nnx_post_train_fixes_20260420_153543_16_zarr3_true/checkpoints/19 I0420 17:23:28.447060 135653656360704 atomicity.py:847] [process=0][thread=async_save] Finished saving checkpoint (finalized tmp dir) to `gs://lance-maxtext/linen_ckpt_xpk_feat_nnx_post_train_fixes_20260420_153543/linen_xpk_feat_nnx_post_train_fixes_20260420_153543_16_zarr3_true/checkpoints/19`. I0420 17:23:28.447926 135653656360704 event_tracking.py:138] [process=0] [async] Finished save (blocking + background) in 52.60 seconds @ gs://lance-maxtext/linen_ckpt_xpk_feat_nnx_post_train_fixes_20260420_153543/linen_xpk_feat_nnx_post_train_fixes_20260420_153543_16_zarr3_true/checkpoints/19 I0420 17:23:28.449214 135653656360704 async_checkpointer.py:160] [process=0][thread=async_save] Background save thread done. Time taken: 52.278620s. I0420 17:23:28.449358 135654176446208 async_checkpointer.py:288] [process=0][thread=save_finalize] Done with waiting for background save thread=async_save. I0420 17:23:28.449419 135654176446208 async_checkpointer.py:298] [process=0][thread=save_finalize] No errors found in background save thread=async_save. I0420 17:23:28.449475 135654176446208 checkpoint_manager.py:2137] [process=0][thread=save_finalize][step=19] CheckpointManager Save Finalize is syncing with other hosts... I0420 17:23:28.451797 135654176446208 checkpoint_manager.py:2146] [process=0][thread=save_finalize][step=19] CheckpointManager Save Finalize is done on all hosts. I0420 17:23:28.451928 135780113520448 checkpoint_manager.py:2032] [process=0][thread=MainThread][step=19][wait_until_finished] Done waiting for Save Finalize thread (save_finalize) running at step=19. I0420 17:23:28.452023 135780113520448 checkpoint_manager.py:2009] [process=0][thread=MainThread][wait_until_finished] No Save Finalize thread to wait for. Returning. I0420 17:23:28.453073 135780113520448 metric_logger.py:196] completed step: 19, seconds: 0.147, TFLOP/s/device: 92.214, Tokens/s/device: 13899.541, total_weights: 65536, loss: 5.391, lm_loss: 5.391, perplexity: 219.349 Per train step: Total TFLOPs: 13.59 split as 93.93% learnable weight flops and 6.07% attention flops XPK End: Mon Apr 20 17:23:41 UTC 2026 EXIT_CODE=0