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 | 92.350 | 92.114 | -0.236 |
| Tok/s | 13919.9 | 13884.5 | -35.484 |
| Avg s/step | 2.831 | 3.184 | +0.353 |
| 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: Sun Apr 19 18:29:38 UTC 2026 PyTorch was not found. Models won't be available and only tokenizers, configuration and file/data utilities can be used. 2026-04-19 18:30:03.585125: E external/local_xla/xla/stream_executor/cuda/cuda_platform.cc:51] failed call to cuInit: INTERNAL: CUDA error: Failed call to cuInit: UNKNOWN ERROR (303) I0419 18:30:04.122726 131973296785216 max_utils.py:273] Attempting to initialize the jax distributed system... I0419 18:30:13.164701 131973296785216 distributed.py:149] Starting JAX distributed service on [::]:8482 I0419 18:30:13.167041 131973296785216 distributed.py:172] Connecting to JAX distributed service on mt-07-eval-4sae1-slice-job-0-0.mt-07-eval-4sae1:8482 I0419 18:30:14.824723 131973296785216 max_utils.py:284] Jax distributed system initialized! I0419 18:30:20.956590 131973296785216 max_utils.py:800] System Information: Jax Version: 0.9.2 I0419 18:30:20.956693 131973296785216 max_utils.py:801] System Information: Jaxlib Version: 0.9.2 I0419 18:30:20.956733 131973296785216 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 I0419 18:30:20.956768 131973296785216 train_utils.py:348] WARNING: Sequence packing is essentially ignored for synthetic data. Please use a real dataset to use sequence packing. I0419 18:30:21.653605 131973296785216 maxtext_utils.py:1551] Num_devices: 32, shape (1, 1, 1, 32, 1, 1, 1, 1, 1, 1, 1, 1, 1) I0419 18:30:21.653874 131973296785216 checkpointing.py:677] Setting up checkpoint logger... I0419 18:30:21.653925 131973296785216 checkpointing.py:233] Creating checkpoint manager with ocdbt=True and zarr3=True I0419 18:30:21.653969 131973296785216 pytree_checkpoint_handler.py:592] save_device_host_concurrent_bytes=None I0419 18:30:21.654289 131973296785216 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 0x7806b5f38050>, enable_pinned_host_transfer=False, save_concurrent_bytes: 96000000000 (89.4 GiB), restore_concurrent_bytes: 96000000000 (89.4 GiB) I0419 18:30:24.528636 131973296785216 checkpointing.py:265] Enabling policy for fixed interval checkpointing. I0419 18:30:24.528867 131973296785216 checkpoint_manager.py:708] [process=6][thread=MainThread] CheckpointManager init: checkpointers=None, item_names=('items',), item_handlers={'items': <orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeCheckpointHandler object at 0x77f2d8430f50>}, handler_registry=None I0419 18:30:24.529101 131973296785216 composite_checkpoint_handler.py:237] Deferred registration for item: "items". Adding handler `<orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeCheckpointHandler object at 0x77f2d8430f50>` for item "items" and save args `<class 'orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeSaveArgs'>` and restore args `<class 'orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeRestoreArgs'>` to `_handler_registry`. I0419 18:30:24.529156 131973296785216 composite_checkpoint_handler.py:237] Deferred registration for item: "metrics". Adding handler `<orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonCheckpointHandler object at 0x77f2d8436ed0>` for item "metrics" and save args `<class 'orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonSaveArgs'>` and restore args `<class 'orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonRestoreArgs'>` to `_handler_registry`. I0419 18:30:24.529192 131973296785216 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 0x77f2d8430f50>, ('items', <class 'orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeRestoreArgs'>): <orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeCheckpointHandler object at 0x77f2d8430f50>, ('metrics', <class 'orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonSaveArgs'>): <orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonCheckpointHandler object at 0x77f2d8436ed0>, ('metrics', <class 'orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonRestoreArgs'>): <orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonCheckpointHandler object at 0x77f2d8436ed0>}). I0419 18:30:24.529528 131973296785216 abstract_checkpointer.py:35] orbax-checkpoint version: 0.11.34 I0419 18:30:24.529599 131973296785216 async_checkpointer.py:192] [process=6][thread=MainThread] Using barrier_sync_fn: <function get_barrier_sync_fn.<locals>._fn at 0x77f2987837e0> timeout: 1200 secs and primary_host=0 for async checkpoint writes I0419 18:30:26.461011 131973296785216 checkpoint_manager.py:1812] Found 0 checkpoint steps in gs://lance-maxtext/linen_ckpt_xpk_main_20260419_180002/linen_xpk_main_20260419_180002_07_eval/checkpoints I0419 18:30:26.922790 131973296785216 checkpoint_manager.py:929] [process=6][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_20260419_180002/linen_xpk_main_20260419_180002_07_eval/checkpoints: <orbax.checkpoint.checkpoint_manager.CheckpointManager object at 0x77f2e022b440> I0419 18:30:26.922953 131973296785216 checkpointing.py:301] Checkpoint manager created! I0419 18:30:27.845157 131973296785216 nnx_wrappers.py:437] Unknown Logical: bfloat16[32,2048,2048]...................................... ('activation_batch', 'activation_norm_length', 'activation_embed'). I0419 18:30:27.845269 131973296785216 nnx_wrappers.py:437] Unknown Physical: bfloat16[32,2048,2048]...................................... ('fsdp', None, None). I0419 18:30:28.227533 131973296785216 attentions.py:1088] attentions/inputs_q Logical: bfloat16[32,2048,2048]...................................... ('activation_batch', 'activation_attn_length', 'activation_attn_embed'). I0419 18:30:28.227621 131973296785216 attentions.py:1088] attentions/inputs_q Physical: bfloat16[32,2048,2048]...................................... ('fsdp', None, None). I0419 18:30:28.243757 131973296785216 attentions.py:1089] attentions/inputs_kv Logical: bfloat16[32,2048,2048]...................................... ('activation_batch', 'activation_attn_length', 'activation_attn_embed'). I0419 18:30:28.243823 131973296785216 attentions.py:1089] attentions/inputs_kv Physical: bfloat16[32,2048,2048]...................................... ('fsdp', None, None). I0419 18:30:28.267262 131973296785216 attentions.py:1154] attentions/query Logical: bfloat16[32,2048,16,128].................................... ('activation_kv_batch', 'activation_attn_length', 'activation_kv_heads', 'activation_kv_head_dim'). I0419 18:30:28.267329 131973296785216 attentions.py:1154] attentions/query Physical: bfloat16[32,2048,16,128].................................... ('fsdp', None, None, None). I0419 18:30:28.289037 131973296785216 attentions.py:1155] attentions/key Logical: bfloat16[32,2048,16,128].................................... ('activation_kv_batch', 'activation_attn_length', 'activation_kv_heads', 'activation_kv_head_dim'). I0419 18:30:28.289136 131973296785216 attentions.py:1155] attentions/key Physical: bfloat16[32,2048,16,128].................................... ('fsdp', None, None, None). I0419 18:30:28.305759 131973296785216 attentions.py:1156] attentions/value Logical: bfloat16[32,2048,16,128].................................... ('activation_kv_batch', 'activation_attn_length', 'activation_kv_heads', 'activation_kv_head_dim'). I0419 18:30:28.305831 131973296785216 attentions.py:1156] attentions/value Physical: bfloat16[32,2048,16,128].................................... ('fsdp', None, None, None). I0419 18:30:28.331428 131973296785216 attentions.py:1197] attentions/out Logical: bfloat16[32,2048,16,128].................................... ('activation_batch', 'activation_attn_length', 'activation_heads', 'activation_kv'). I0419 18:30:28.331499 131973296785216 attentions.py:1197] attentions/out Physical: bfloat16[32,2048,16,128].................................... ('fsdp', None, None, None). I0419 18:30:28.352385 131973296785216 linears.py:525] linears/x Logical: bfloat16[32,2048,7168]...................................... ('activation_batch', 'activation_length', 'activation_mlp'). I0419 18:30:28.352475 131973296785216 linears.py:525] linears/x Physical: bfloat16[32,2048,7168]...................................... ('fsdp', None, None). I0419 18:30:28.576989 131973296785216 checkpointing.py:577] checkpoint manager exists so trying to load this run's existing checkpoint I0419 18:30:28.577100 131973296785216 checkpointing.py:665] No existing checkpoints found, not restoring checkpoint. fsdp: 32 I0419 18:30:30.006798 131973296785216 maxtext_utils.py:1654] params/params/decoder/decoder_norm/scale Shape: float32[2048] Logical: P('norm',) Physical: (None,) I0419 18:30:30.006922 131973296785216 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) I0419 18:30:30.006975 131973296785216 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) I0419 18:30:30.007032 131973296785216 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') I0419 18:30:30.007090 131973296785216 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) I0419 18:30:30.007130 131973296785216 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) I0419 18:30:30.007184 131973296785216 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) I0419 18:30:30.007236 131973296785216 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') I0419 18:30:30.007277 131973296785216 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) I0419 18:30:30.007315 131973296785216 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) I0419 18:30:30.007363 131973296785216 maxtext_utils.py:1654] params/params/decoder/logits_dense/kernel Shape: float32[2048,32000] Logical: P('embed_vocab', 'vocab') Physical: ('fsdp', None) I0419 18:30:30.007436 131973296785216 maxtext_utils.py:1654] params/params/token_embedder/embedding Shape: float32[32000,2048] Logical: P('vocab', 'embed_vocab') Physical: (None, 'fsdp') I0419 18:30:30.493458 131973296785216 train.py:155] train/xent Logical: float32[32,2048]............................................ ('activation_embed_and_logits_batch', 'activation_length'). I0419 18:30:30.493551 131973296785216 train.py:155] train/xent Physical: float32[32,2048]............................................ ('fsdp', None). I0419 18:30:30.508754 131973296785216 train.py:162] train/z_loss Logical: float32[32,2048]............................................ ('activation_embed_and_logits_batch', 'activation_length'). I0419 18:30:30.508815 131973296785216 train.py:162] train/z_loss Physical: float32[32,2048]............................................ ('fsdp', None). I0419 18:30:41.481258 131973296785216 max_utils.py:791] Total memory size: 1.5 GB, Output size: 0.4 GB, Temp size: 1.1 GB, Argument size: 0.4 GB, Host temp size: 0.0 GB. I0419 18:30:41.482091 131973296785216 metric_logger.py:301] number parameters: 1.104 billion I0419 18:30:52.902930 131973296785216 checkpointing.py:772] Waiting for step 0 to finish before checkpoint... I0419 18:30:53.055904 131973296785216 checkpointing.py:776] Waited 0.15295767784118652 seconds for step 0 to finish before starting checkpointing. I0419 18:30:53.058470 131973296785216 checkpoint_manager.py:2009] [process=6][thread=MainThread][wait_until_finished] No Save Finalize thread to wait for. Returning. I0419 18:30:53.060133 131973296785216 checkpoint_manager.py:1512] [process=6] Saving checkpoint at step 0 I0419 18:30:53.061850 131973296785216 event_tracking.py:70] [process=6] [async] Started save checkpoint @ gs://lance-maxtext/linen_ckpt_xpk_main_20260419_180002/linen_xpk_main_20260419_180002_07_eval/checkpoints/0. I0419 18:30:53.825946 131973296785216 signaling_client.py:364] Using JaxDistributedSignalingClient I0419 18:30:53.827016 131973296785216 jax_array_handlers.py:360] Scheduling D2H of 39 prioritized jax.Array. I0419 18:30:53.827072 131973296785216 replica_slices.py:424] Transferring arrays to host memory with options: use_replica_parallel=True, min_slice_bytes_for_replica_parallel=None, max_replicas_for_replica_parallel=None, enable_pinned_host_transfer=False I0419 18:30:54.104160 131973296785216 base_pytree_checkpoint_handler.py:154] [process=6][thread=MainThread] Initiated "orbax.checkpoint._src.serialization.jax_array_handlers.ArrayHandler".serialize. Time taken: 0.278282s I0419 18:30:54.104331 131973296785216 base_pytree_checkpoint_handler.py:130] [process=6] /jax/orbax/write/blocking_gbytes_per_sec: 5.437 GiB/s (total gbytes: 1.5 GiB) (time elapsed: 0.2837355136871338 s) (per-host) I0419 18:30:54.104387 131973296785216 base_pytree_checkpoint_handler.py:768] [process=6][thread=MainThread] Initiated Pytree async_save. Time taken: 0.283801s (batch_requests_ready=0.002205s, total_serialization_initiated=0.281522s, others=0.000074s) I0419 18:30:54.104492 131973296785216 composite_checkpoint_handler.py:715] [process=6][thread=MainThread] Initiated CompositeCheckpointHandler.async_save. Time taken: 0.287966s (all_items=0.000017s, per_item={'items': '0.00001669'}, temp_paths=0.287950) I0419 18:30:54.105276 131973296785216 event_tracking.py:125] [process=6] [async] Finished blocking save in 1.05 seconds. Continuing save @ gs://lance-maxtext/linen_ckpt_xpk_main_20260419_180002/linen_xpk_main_20260419_180002_07_eval/checkpoints/0. I0419 18:30:54.105624 131844817282816 async_checkpointer.py:76] [process=6][thread=async_save] Background save thread started. Deadline for this save operation is 2026-04-19 18:50:54.105587 I0419 18:30:54.110279 131973296785216 checkpoint_manager.py:1560] [process=6][thread=MainThread][step=0] Starting CheckpointManager Save Finalize thread=save_finalize I0419 18:30:54.110561 131843168913152 async_checkpointer.py:280] [process=6][thread=save_finalize] Waiting for background save thread=async_save. I0419 18:30:54.110688 131973296785216 standard_logger.py:34] {'step': 0, 'event_type': 'save', 'directory': 'gs://lance-maxtext/linen_ckpt_xpk_main_20260419_180002/linen_xpk_main_20260419_180002_07_eval/checkpoints', 'reached_preemption': False, 'preemption_received_at': None, 'synchronous': False, 'wait_for_prev_start_time': 1776623453.0584524, 'wait_for_prev_duration_secs': 6.008148193359375e-05, 'time_between_consecutive_saves_sec': None, 'checkpointer_blocking_start_time': 1776623453.060219, 'checkpointer_blocking_duration_secs': 1.0455586910247803, 'get_old_steps_start_time': 1776623454.105804, 'get_old_steps_duration_secs': 3.170967102050781e-05, 'checkpoint_manager_blocking_start_time': 1776623453.056454, 'checkpoint_manager_blocking_duration_secs': 1.054192304611206} I0419 18:30:54.110799 131973296785216 checkpointing.py:408] Started an asynchronous checkpoint save for step 0 I0419 18:30:54.110847 131973296785216 max_utils.py:750] Memstats: After params initialized: I0419 18:30:54.110895 131973296785216 max_utils.py:756] Using (GB) 0.43 / 31.25 (1.376000%) on TPU_24(process=6,(0,6,0,0)) I0419 18:30:54.110927 131973296785216 max_utils.py:756] Using (GB) 0.43 / 31.25 (1.376000%) on TPU_25(process=6,(1,6,0,0)) I0419 18:30:54.110953 131973296785216 max_utils.py:756] Using (GB) 0.43 / 31.25 (1.376000%) on TPU_28(process=6,(0,7,0,0)) I0419 18:30:54.110978 131973296785216 max_utils.py:756] Using (GB) 0.43 / 31.25 (1.376000%) on TPU_29(process=6,(1,7,0,0)) I0419 18:30:54.423501 131973296785216 metric_logger.py:196] completed step: 0, seconds: 11.421, TFLOP/s/device: 1.190, Tokens/s/device: 179.323, total_weights: 65536, loss: 10.874, lm_loss: 10.874, perplexity: 52779.875 I0419 18:30:54.609318 131973296785216 metric_logger.py:196] completed step: 1, seconds: 1.519, TFLOP/s/device: 8.945, Tokens/s/device: 1348.300, total_weights: 65536, loss: 10.874, lm_loss: 10.874, perplexity: 52779.875 I0419 18:30:55.024252 131973296785216 metric_logger.py:196] completed step: 2, seconds: 0.039, TFLOP/s/device: 351.034, Tokens/s/device: 52911.693, total_weights: 65536, loss: 10.262, lm_loss: 10.262, perplexity: 28637.129 I0419 18:30:55.171479 131973296785216 metric_logger.py:196] completed step: 3, seconds: 0.415, TFLOP/s/device: 32.737, Tokens/s/device: 4934.417, total_weights: 65536, loss: 9.734, lm_loss: 9.734, perplexity: 16889.203 I0419 18:30:55.466487 131973296785216 metric_logger.py:196] completed step: 4, seconds: 0.153, TFLOP/s/device: 88.703, Tokens/s/device: 13370.241, total_weights: 65536, loss: 9.277, lm_loss: 9.277, perplexity: 10694.614 I0419 18:30:55.472471 131973296785216 metric_logger.py:196] completed step: 5, seconds: 0.147, TFLOP/s/device: 92.335, Tokens/s/device: 13917.676, total_weights: 65536, loss: 8.892, lm_loss: 8.892, perplexity: 7270.717 I0419 18:30:57.890576 2783 google_auth_provider.cc:181] Running on GCE, using service account 562977990677-compute@developer.gserviceaccount.com I0419 18:31:00.422579 131843177305856 array_metadata_store.py:203] [process=6][thread=array_type_handler] Wrote 39 array_metadata.ArrayMetadata to gs://lance-maxtext/linen_ckpt_xpk_main_20260419_180002/linen_xpk_main_20260419_180002_07_eval/checkpoints/0/items/array_metadatas/process_6 I0419 18:31:18.233723 131973296785216 metric_logger.py:196] completed step: 6, seconds: 0.295, TFLOP/s/device: 45.985, Tokens/s/device: 6931.400, total_weights: 65536, loss: 8.592, lm_loss: 8.592, perplexity: 5390.759 I0419 18:31:18.381006 131973296785216 metric_logger.py:196] completed step: 7, seconds: 22.615, TFLOP/s/device: 0.601, Tokens/s/device: 90.561, total_weights: 65536, loss: 8.384, lm_loss: 8.384, perplexity: 4376.905 I0419 18:31:18.528430 131973296785216 metric_logger.py:196] completed step: 8, seconds: 0.152, TFLOP/s/device: 89.137, Tokens/s/device: 13435.764, total_weights: 65536, loss: 8.255, lm_loss: 8.255, perplexity: 3846.406 I0419 18:31:18.675166 131973296785216 checkpointing.py:772] Waiting for step 9 to finish before checkpoint... I0419 18:31:18.675834 131973296785216 checkpointing.py:776] Waited 0.0006954669952392578 seconds for step 9 to finish before starting checkpointing. I0419 18:31:18.677968 131973296785216 checkpoint_manager.py:2020] [process=6][thread=MainThread][step=0][wait_until_finished] Waiting for Save Finalize thread (save_finalize) to complete. I0419 18:31:31.978878 131844817282816 base_pytree_checkpoint_handler.py:130] [process=6] /jax/orbax/write/gbytes_per_sec: 41.396 MiB/s (total gbytes: 1.5 GiB) (time elapsed: 38.15824365615845 s) (per-host) I0419 18:31:31.979001 131844817282816 async_checkpointer.py:90] [process=6][thread=async_save] 3 Handler Commit operations completed. Time taken: 37.873265s. I0419 18:31:40.220470 131844817282816 async_checkpointer.py:160] [process=6][thread=async_save] Background save thread done. Time taken: 46.114717s. I0419 18:31:40.220738 131843168913152 async_checkpointer.py:288] [process=6][thread=save_finalize] Done with waiting for background save thread=async_save. I0419 18:31:40.220858 131843168913152 async_checkpointer.py:298] [process=6][thread=save_finalize] No errors found in background save thread=async_save. I0419 18:31:40.220910 131843168913152 checkpoint_manager.py:2137] [process=6][thread=save_finalize][step=0] CheckpointManager Save Finalize is syncing with other hosts... I0419 18:31:40.223660 131843168913152 checkpoint_manager.py:2146] [process=6][thread=save_finalize][step=0] CheckpointManager Save Finalize is done on all hosts. I0419 18:31:40.223840 131973296785216 checkpoint_manager.py:2032] [process=6][thread=MainThread][step=0][wait_until_finished] Done waiting for Save Finalize thread (save_finalize) running at step=0. W0419 18:31:40.223972 131973296785216 checkpoint_manager.py:1452] Waiting for previous save to complete took 21.546000 seconds. If this number is high, consider checkpointing less frequently. I0419 18:31:40.225724 131973296785216 checkpoint_manager.py:1512] [process=6] Saving checkpoint at step 9 I0419 18:31:40.227663 131973296785216 event_tracking.py:70] [process=6] [async] Started save checkpoint @ gs://lance-maxtext/linen_ckpt_xpk_main_20260419_180002/linen_xpk_main_20260419_180002_07_eval/checkpoints/9. I0419 18:31:40.776001 131973296785216 jax_array_handlers.py:360] Scheduling D2H of 39 prioritized jax.Array. I0419 18:31:40.776098 131973296785216 replica_slices.py:424] Transferring arrays to host memory with options: use_replica_parallel=True, min_slice_bytes_for_replica_parallel=None, max_replicas_for_replica_parallel=None, enable_pinned_host_transfer=False I0419 18:31:40.808231 131973296785216 base_pytree_checkpoint_handler.py:154] [process=6][thread=MainThread] Initiated "orbax.checkpoint._src.serialization.jax_array_handlers.ArrayHandler".serialize. Time taken: 0.033342s I0419 18:31:40.808384 131973296785216 base_pytree_checkpoint_handler.py:130] [process=6] /jax/orbax/write/blocking_gbytes_per_sec: 41.934 GiB/s (total gbytes: 1.5 GiB) (time elapsed: 0.03678607940673828 s) (per-host) I0419 18:31:40.808447 131973296785216 base_pytree_checkpoint_handler.py:768] [process=6][thread=MainThread] Initiated Pytree async_save. Time taken: 0.036859s (batch_requests_ready=0.001805s, total_serialization_initiated=0.034973s, others=0.000081s) I0419 18:31:40.808534 131973296785216 composite_checkpoint_handler.py:715] [process=6][thread=MainThread] Initiated CompositeCheckpointHandler.async_save. Time taken: 0.040922s (all_items=0.000016s, per_item={'items': '0.00001621'}, temp_paths=0.040906) I0419 18:31:40.809230 131973296785216 event_tracking.py:125] [process=6] [async] Finished blocking save in 0.58 seconds. Continuing save @ gs://lance-maxtext/linen_ckpt_xpk_main_20260419_180002/linen_xpk_main_20260419_180002_07_eval/checkpoints/9. I0419 18:31:40.809590 131843168913152 async_checkpointer.py:76] [process=6][thread=async_save] Background save thread started. Deadline for this save operation is 2026-04-19 18:51:40.809559 I0419 18:31:40.820924 131973296785216 checkpoint_manager.py:1560] [process=6][thread=MainThread][step=9] Starting CheckpointManager Save Finalize thread=save_finalize I0419 18:31:40.821175 131844259190528 async_checkpointer.py:280] [process=6][thread=save_finalize] Waiting for background save thread=async_save. I0419 18:31:40.821319 131973296785216 standard_logger.py:34] {'step': 9, 'event_type': 'save', 'directory': 'gs://lance-maxtext/linen_ckpt_xpk_main_20260419_180002/linen_xpk_main_20260419_180002_07_eval/checkpoints', 'reached_preemption': False, 'preemption_received_at': None, 'synchronous': False, 'wait_for_prev_start_time': 1776623478.67794, 'wait_for_prev_duration_secs': 21.54599952697754, 'time_between_consecutive_saves_sec': None, 'checkpointer_blocking_start_time': 1776623500.2257645, 'checkpointer_blocking_duration_secs': 0.5839657783508301, 'get_old_steps_start_time': 1776623500.8097553, 'get_old_steps_duration_secs': 3.147125244140625e-05, 'checkpoint_manager_blocking_start_time': 1776623478.6760862, 'checkpoint_manager_blocking_duration_secs': 22.14519739151001} I0419 18:31:40.821445 131973296785216 checkpointing.py:408] Started an asynchronous checkpoint save for step 9 I0419 18:31:40.821495 131973296785216 checkpoint_manager.py:2020] [process=6][thread=MainThread][step=9][wait_until_finished] Waiting for Save Finalize thread (save_finalize) to complete. I0419 18:31:46.073509 131843177305856 array_metadata_store.py:203] [process=6][thread=array_type_handler] Wrote 39 array_metadata.ArrayMetadata to gs://lance-maxtext/linen_ckpt_xpk_main_20260419_180002/linen_xpk_main_20260419_180002_07_eval/checkpoints/9/items/array_metadatas/process_6 I0419 18:32:23.707731 131843168913152 base_pytree_checkpoint_handler.py:130] [process=6] /jax/orbax/write/gbytes_per_sec: 36.789 MiB/s (total gbytes: 1.5 GiB) (time elapsed: 42.93610143661499 s) (per-host) I0419 18:32:23.707841 131843168913152 async_checkpointer.py:90] [process=6][thread=async_save] 3 Handler Commit operations completed. Time taken: 42.898137s. I0419 18:32:30.151853 131843168913152 async_checkpointer.py:160] [process=6][thread=async_save] Background save thread done. Time taken: 49.342134s. I0419 18:32:30.152080 131844259190528 async_checkpointer.py:288] [process=6][thread=save_finalize] Done with waiting for background save thread=async_save. I0419 18:32:30.152146 131844259190528 async_checkpointer.py:298] [process=6][thread=save_finalize] No errors found in background save thread=async_save. I0419 18:32:30.152193 131844259190528 checkpoint_manager.py:2137] [process=6][thread=save_finalize][step=9] CheckpointManager Save Finalize is syncing with other hosts... I0419 18:32:30.153878 131844259190528 checkpoint_manager.py:2146] [process=6][thread=save_finalize][step=9] CheckpointManager Save Finalize is done on all hosts. I0419 18:32:30.154047 131973296785216 checkpoint_manager.py:2032] [process=6][thread=MainThread][step=9][wait_until_finished] Done waiting for Save Finalize thread (save_finalize) running at step=9. I0419 18:32:30.154206 131973296785216 checkpoint_manager.py:2009] [process=6][thread=MainThread][wait_until_finished] No Save Finalize thread to wait for. Returning. I0419 18:32:30.155069 131973296785216 metric_logger.py:196] completed step: 9, seconds: 0.147, TFLOP/s/device: 92.350, Tokens/s/device: 13919.947, 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: Sun Apr 19 18:32:38 UTC 2026 EXIT_CODE=0
574ad3fb9
· feat_nnx_trainstate_and_training_loop_20260419_214525
· full log
XPK Start: Sun Apr 19 22:14:30 UTC 2026 PyTorch was not found. Models won't be available and only tokenizers, configuration and file/data utilities can be used. 2026-04-19 22:14:54.813504: E external/local_xla/xla/stream_executor/cuda/cuda_platform.cc:51] failed call to cuInit: INTERNAL: CUDA error: Failed call to cuInit: UNKNOWN ERROR (303) I0419 22:14:54.992481 136025315080000 max_utils.py:273] Attempting to initialize the jax distributed system... I0419 22:15:04.031090 136025315080000 distributed.py:149] Starting JAX distributed service on [::]:8482 I0419 22:15:04.033461 136025315080000 distributed.py:172] Connecting to JAX distributed service on mt-07-eval-c0uh8-slice-job-0-0.mt-07-eval-c0uh8:8482 I0419 22:15:05.902734 136025315080000 max_utils.py:284] Jax distributed system initialized! I0419 22:15:12.045748 136025315080000 max_utils.py:800] System Information: Jax Version: 0.9.2 I0419 22:15:12.045850 136025315080000 max_utils.py:801] System Information: Jaxlib Version: 0.9.2 I0419 22:15:12.045889 136025315080000 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 I0419 22:15:12.045922 136025315080000 train_utils.py:378] WARNING: Sequence packing is essentially ignored for synthetic data. Please use a real dataset to use sequence packing. I0419 22:15:13.096575 136025315080000 maxtext_utils.py:1718] Num_devices: 32, shape (1, 1, 1, 32, 1, 1, 1, 1, 1, 1, 1, 1, 1) I0419 22:15:13.097170 136025315080000 maxtext_utils.py:1718] Num_devices: 32, shape (1, 1, 1, 32, 1, 1, 1, 1, 1, 1, 1, 1, 1) I0419 22:15:13.097373 136025315080000 checkpointing.py:688] Setting up checkpoint logger... I0419 22:15:13.097423 136025315080000 checkpointing.py:234] Creating checkpoint manager with ocdbt=True and zarr3=True I0419 22:15:13.097467 136025315080000 pytree_checkpoint_handler.py:592] save_device_host_concurrent_bytes=None I0419 22:15:13.097799 136025315080000 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 0x7bb624f3c590>, enable_pinned_host_transfer=False, save_concurrent_bytes: 96000000000 (89.4 GiB), restore_concurrent_bytes: 96000000000 (89.4 GiB) I0419 22:15:17.454259 136025315080000 checkpointing.py:266] Enabling policy for fixed interval checkpointing. I0419 22:15:17.454526 136025315080000 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 0x7bb518fa3a70>}, handler_registry=None I0419 22:15:17.454911 136025315080000 composite_checkpoint_handler.py:237] Deferred registration for item: "items". Adding handler `<orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeCheckpointHandler object at 0x7bb518fa3a70>` for item "items" and save args `<class 'orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeSaveArgs'>` and restore args `<class 'orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeRestoreArgs'>` to `_handler_registry`. I0419 22:15:17.454976 136025315080000 composite_checkpoint_handler.py:237] Deferred registration for item: "metrics". Adding handler `<orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonCheckpointHandler object at 0x7bb518fc5850>` for item "metrics" and save args `<class 'orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonSaveArgs'>` and restore args `<class 'orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonRestoreArgs'>` to `_handler_registry`. I0419 22:15:17.455030 136025315080000 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 0x7bb518fa3a70>, ('items', <class 'orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeRestoreArgs'>): <orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeCheckpointHandler object at 0x7bb518fa3a70>, ('metrics', <class 'orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonSaveArgs'>): <orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonCheckpointHandler object at 0x7bb518fc5850>, ('metrics', <class 'orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonRestoreArgs'>): <orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonCheckpointHandler object at 0x7bb518fc5850>}). I0419 22:15:17.455385 136025315080000 abstract_checkpointer.py:35] orbax-checkpoint version: 0.11.34 I0419 22:15:17.455481 136025315080000 async_checkpointer.py:192] [process=4][thread=MainThread] Using barrier_sync_fn: <function get_barrier_sync_fn.<locals>._fn at 0x7ba1c0593740> timeout: 1200 secs and primary_host=0 for async checkpoint writes I0419 22:15:19.200947 136025315080000 checkpoint_manager.py:1812] Found 0 checkpoint steps in gs://lance-maxtext/linen_ckpt_xpk_feat_nnx_trainstate_and_training_loop_20260419_214525/linen_xpk_feat_nnx_trainstate_and_training_loop_20260419_214525_07_eval/checkpoints I0419 22:15:19.203212 136025315080000 checkpoint_manager.py:929] [process=4][thread=MainThread] CheckpointManager created, primary_host=0, CheckpointManagerOptions=CheckpointManagerOptions(save_interval_steps=1, max_to_keep=None, keep_time_interval=None, keep_period=None, should_keep_fn=None, best_fn=None, best_mode='max', keep_checkpoints_without_metrics=True, step_prefix=None, step_format_fixed_length=None, step_name_format=None, create=True, cleanup_tmp_directories=False, save_on_steps=frozenset(), single_host_load_and_broadcast=False, todelete_subdir=None, todelete_full_path=None, enable_background_delete=False, read_only=False, enable_async_checkpointing=True, async_options=None, multiprocessing_options=MultiprocessingOptions(primary_host=0, active_processes=None, barrier_sync_key_prefix=None), should_save_fn=None, file_options=FileOptions(path_permission_mode=None), save_root_metadata=True, temporary_path_class=None, save_decision_policy=FixedIntervalPolicy(interval=10), preservation_policy=LatestN(n=None), prevent_write_metrics=False, enable_should_save_is_saving_in_progress_check=True, enable_per_process_directory_creation=False, lightweight_initialize=False), root_directory=gs://lance-maxtext/linen_ckpt_xpk_feat_nnx_trainstate_and_training_loop_20260419_214525/linen_xpk_feat_nnx_trainstate_and_training_loop_20260419_214525_07_eval/checkpoints: <orbax.checkpoint.checkpoint_manager.CheckpointManager object at 0x7bb518fc53d0> I0419 22:15:19.203325 136025315080000 checkpointing.py:302] Checkpoint manager created! I0419 22:15:20.132354 136025315080000 nnx_wrappers.py:437] Unknown Logical: bfloat16[32,2048,2048]...................................... ('activation_batch', 'activation_norm_length', 'activation_embed'). I0419 22:15:20.132455 136025315080000 nnx_wrappers.py:437] Unknown Physical: bfloat16[32,2048,2048]...................................... ('fsdp', None, None). I0419 22:15:20.512914 136025315080000 attentions.py:1088] attentions/inputs_q Logical: bfloat16[32,2048,2048]...................................... ('activation_batch', 'activation_attn_length', 'activation_attn_embed'). I0419 22:15:20.513004 136025315080000 attentions.py:1088] attentions/inputs_q Physical: bfloat16[32,2048,2048]...................................... ('fsdp', None, None). I0419 22:15:20.529264 136025315080000 attentions.py:1089] attentions/inputs_kv Logical: bfloat16[32,2048,2048]...................................... ('activation_batch', 'activation_attn_length', 'activation_attn_embed'). I0419 22:15:20.529320 136025315080000 attentions.py:1089] attentions/inputs_kv Physical: bfloat16[32,2048,2048]...................................... ('fsdp', None, None). I0419 22:15:20.552745 136025315080000 attentions.py:1154] attentions/query Logical: bfloat16[32,2048,16,128].................................... ('activation_kv_batch', 'activation_attn_length', 'activation_kv_heads', 'activation_kv_head_dim'). I0419 22:15:20.552819 136025315080000 attentions.py:1154] attentions/query Physical: bfloat16[32,2048,16,128].................................... ('fsdp', None, None, None). I0419 22:15:20.569011 136025315080000 attentions.py:1155] attentions/key Logical: bfloat16[32,2048,16,128].................................... ('activation_kv_batch', 'activation_attn_length', 'activation_kv_heads', 'activation_kv_head_dim'). I0419 22:15:20.569076 136025315080000 attentions.py:1155] attentions/key Physical: bfloat16[32,2048,16,128].................................... ('fsdp', None, None, None). I0419 22:15:20.585185 136025315080000 attentions.py:1156] attentions/value Logical: bfloat16[32,2048,16,128].................................... ('activation_kv_batch', 'activation_attn_length', 'activation_kv_heads', 'activation_kv_head_dim'). I0419 22:15:20.585256 136025315080000 attentions.py:1156] attentions/value Physical: bfloat16[32,2048,16,128].................................... ('fsdp', None, None, None). I0419 22:15:20.609833 136025315080000 attentions.py:1197] attentions/out Logical: bfloat16[32,2048,16,128].................................... ('activation_batch', 'activation_attn_length', 'activation_heads', 'activation_kv'). I0419 22:15:20.609897 136025315080000 attentions.py:1197] attentions/out Physical: bfloat16[32,2048,16,128].................................... ('fsdp', None, None, None). I0419 22:15:20.631458 136025315080000 linears.py:525] linears/x Logical: bfloat16[32,2048,7168]...................................... ('activation_batch', 'activation_length', 'activation_mlp'). I0419 22:15:20.631519 136025315080000 linears.py:525] linears/x Physical: bfloat16[32,2048,7168]...................................... ('fsdp', None, None). I0419 22:15:20.846487 136025315080000 checkpointing.py:578] checkpoint manager exists so trying to load this run's existing checkpoint I0419 22:15:20.846593 136025315080000 checkpointing.py:676] No existing checkpoints found, not restoring checkpoint. fsdp: 32 I0419 22:15:22.263571 136025315080000 maxtext_utils.py:1821] params/params/decoder/decoder_norm/scale Shape: float32[2048] Logical: P('norm',) Physical: (None,) I0419 22:15:22.263695 136025315080000 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) I0419 22:15:22.263748 136025315080000 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) I0419 22:15:22.263806 136025315080000 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') I0419 22:15:22.263882 136025315080000 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) I0419 22:15:22.263945 136025315080000 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) I0419 22:15:22.264021 136025315080000 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) I0419 22:15:22.264078 136025315080000 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') I0419 22:15:22.264119 136025315080000 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) I0419 22:15:22.264155 136025315080000 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) I0419 22:15:22.264203 136025315080000 maxtext_utils.py:1821] params/params/decoder/logits_dense/kernel Shape: float32[2048,32000] Logical: P('embed_vocab', 'vocab') Physical: ('fsdp', None) I0419 22:15:22.264250 136025315080000 maxtext_utils.py:1821] params/params/token_embedder/embedding Shape: float32[32000,2048] Logical: P('vocab', 'embed_vocab') Physical: (None, 'fsdp') I0419 22:15:22.756108 136025315080000 train.py:157] train/xent Logical: float32[32,2048]............................................ ('activation_embed_and_logits_batch', 'activation_length'). I0419 22:15:22.756198 136025315080000 train.py:157] train/xent Physical: float32[32,2048]............................................ ('fsdp', None). I0419 22:15:22.771449 136025315080000 train.py:164] train/z_loss Logical: float32[32,2048]............................................ ('activation_embed_and_logits_batch', 'activation_length'). I0419 22:15:22.771509 136025315080000 train.py:164] train/z_loss Physical: float32[32,2048]............................................ ('fsdp', None). I0419 22:15:33.615466 136025315080000 max_utils.py:791] Total memory size: 1.5 GB, Output size: 0.4 GB, Temp size: 1.1 GB, Argument size: 0.4 GB, Host temp size: 0.0 GB. I0419 22:15:33.616202 136025315080000 metric_logger.py:301] number parameters: 1.104 billion I0419 22:15:44.801060 136025315080000 checkpointing.py:794] Waiting for step 0 to finish before checkpoint... I0419 22:15:44.954884 136025315080000 checkpointing.py:798] Waited 0.15381121635437012 seconds for step 0 to finish before starting checkpointing. I0419 22:15:44.957278 136025315080000 checkpoint_manager.py:2009] [process=4][thread=MainThread][wait_until_finished] No Save Finalize thread to wait for. Returning. I0419 22:15:44.958908 136025315080000 checkpoint_manager.py:1512] [process=4] Saving checkpoint at step 0 I0419 22:15:44.960324 136025315080000 event_tracking.py:70] [process=4] [async] Started save checkpoint @ gs://lance-maxtext/linen_ckpt_xpk_feat_nnx_trainstate_and_training_loop_20260419_214525/linen_xpk_feat_nnx_trainstate_and_training_loop_20260419_214525_07_eval/checkpoints/0. I0419 22:15:45.693359 136025315080000 signaling_client.py:364] Using JaxDistributedSignalingClient I0419 22:15:45.694309 136025315080000 jax_array_handlers.py:360] Scheduling D2H of 39 prioritized jax.Array. I0419 22:15:45.694383 136025315080000 replica_slices.py:424] Transferring arrays to host memory with options: use_replica_parallel=True, min_slice_bytes_for_replica_parallel=None, max_replicas_for_replica_parallel=None, enable_pinned_host_transfer=False I0419 22:15:45.965262 136025315080000 base_pytree_checkpoint_handler.py:154] [process=4][thread=MainThread] Initiated "orbax.checkpoint._src.serialization.jax_array_handlers.ArrayHandler".serialize. Time taken: 0.271992s I0419 22:15:45.965459 136025315080000 base_pytree_checkpoint_handler.py:130] [process=4] /jax/orbax/write/blocking_gbytes_per_sec: 5.561 GiB/s (total gbytes: 1.5 GiB) (time elapsed: 0.2774035930633545 s) (per-host) I0419 22:15:45.965585 136025315080000 base_pytree_checkpoint_handler.py:768] [process=4][thread=MainThread] Initiated Pytree async_save. Time taken: 0.277539s (batch_requests_ready=0.002190s, total_serialization_initiated=0.275205s, others=0.000144s) I0419 22:15:45.965692 136025315080000 composite_checkpoint_handler.py:715] [process=4][thread=MainThread] Initiated CompositeCheckpointHandler.async_save. Time taken: 0.281612s (all_items=0.000018s, per_item={'items': '0.00001788'}, temp_paths=0.281594) I0419 22:15:45.966485 136025315080000 event_tracking.py:125] [process=4] [async] Finished blocking save in 1.01 seconds. Continuing save @ gs://lance-maxtext/linen_ckpt_xpk_feat_nnx_trainstate_and_training_loop_20260419_214525/linen_xpk_feat_nnx_trainstate_and_training_loop_20260419_214525_07_eval/checkpoints/0. I0419 22:15:45.966853 135895718033152 async_checkpointer.py:76] [process=4][thread=async_save] Background save thread started. Deadline for this save operation is 2026-04-19 22:35:45.966817 I0419 22:15:45.980087 136025315080000 checkpoint_manager.py:1560] [process=4][thread=MainThread][step=0] Starting CheckpointManager Save Finalize thread=save_finalize I0419 22:15:45.980424 135892456806144 async_checkpointer.py:280] [process=4][thread=save_finalize] Waiting for background save thread=async_save. I0419 22:15:45.980583 136025315080000 standard_logger.py:34] {'step': 0, 'event_type': 'save', 'directory': 'gs://lance-maxtext/linen_ckpt_xpk_feat_nnx_trainstate_and_training_loop_20260419_214525/linen_xpk_feat_nnx_trainstate_and_training_loop_20260419_214525_07_eval/checkpoints', 'reached_preemption': False, 'preemption_received_at': None, 'synchronous': False, 'wait_for_prev_start_time': 1776636944.95726, 'wait_for_prev_duration_secs': 6.127357482910156e-05, 'time_between_consecutive_saves_sec': None, 'checkpointer_blocking_start_time': 1776636944.9589467, 'checkpointer_blocking_duration_secs': 1.0080616474151611, 'get_old_steps_start_time': 1776636945.967033, 'get_old_steps_duration_secs': 3.075599670410156e-05, 'checkpoint_manager_blocking_start_time': 1776636944.9553697, 'checkpoint_manager_blocking_duration_secs': 1.0251736640930176} I0419 22:15:45.980688 136025315080000 checkpointing.py:409] Started an asynchronous checkpoint save for step 0 I0419 22:15:45.980737 136025315080000 max_utils.py:750] Memstats: After params initialized: I0419 22:15:45.980785 136025315080000 max_utils.py:756] Using (GB) 0.43 / 31.25 (1.376000%) on TPU_16(process=4,(0,4,0,0)) I0419 22:15:45.980818 136025315080000 max_utils.py:756] Using (GB) 0.43 / 31.25 (1.376000%) on TPU_17(process=4,(1,4,0,0)) I0419 22:15:45.980851 136025315080000 max_utils.py:756] Using (GB) 0.43 / 31.25 (1.376000%) on TPU_20(process=4,(0,5,0,0)) I0419 22:15:45.980876 136025315080000 max_utils.py:756] Using (GB) 0.43 / 31.25 (1.376000%) on TPU_21(process=4,(1,5,0,0)) I0419 22:15:46.293849 136025315080000 metric_logger.py:196] completed step: 0, seconds: 11.185, TFLOP/s/device: 1.215, Tokens/s/device: 183.106, total_weights: 65536, loss: 10.874, lm_loss: 10.874, perplexity: 52779.875 I0419 22:15:46.485038 136025315080000 metric_logger.py:196] completed step: 1, seconds: 1.491, TFLOP/s/device: 9.111, Tokens/s/device: 1373.340, total_weights: 65536, loss: 10.874, lm_loss: 10.874, perplexity: 52779.875 I0419 22:15:47.337770 136025315080000 metric_logger.py:196] completed step: 2, seconds: 0.044, TFLOP/s/device: 307.972, Tokens/s/device: 46420.962, total_weights: 65536, loss: 10.262, lm_loss: 10.262, perplexity: 28637.129 I0419 22:15:47.485215 136025315080000 metric_logger.py:196] completed step: 3, seconds: 0.853, TFLOP/s/device: 15.932, Tokens/s/device: 2401.433, total_weights: 65536, loss: 9.734, lm_loss: 9.734, perplexity: 16889.203 I0419 22:15:47.780484 136025315080000 metric_logger.py:196] completed step: 4, seconds: 0.153, TFLOP/s/device: 88.876, Tokens/s/device: 13396.391, total_weights: 65536, loss: 9.277, lm_loss: 9.277, perplexity: 10694.614 I0419 22:15:47.786531 136025315080000 metric_logger.py:196] completed step: 5, seconds: 0.147, TFLOP/s/device: 92.199, Tokens/s/device: 13897.277, total_weights: 65536, loss: 8.892, lm_loss: 8.892, perplexity: 7270.717 I0419 22:15:49.749369 2829 google_auth_provider.cc:181] Running on GCE, using service account 562977990677-compute@developer.gserviceaccount.com I0419 22:15:52.234467 135892465198848 array_metadata_store.py:203] [process=4][thread=array_type_handler] Wrote 39 array_metadata.ArrayMetadata to gs://lance-maxtext/linen_ckpt_xpk_feat_nnx_trainstate_and_training_loop_20260419_214525/linen_xpk_feat_nnx_trainstate_and_training_loop_20260419_214525_07_eval/checkpoints/0/items/array_metadatas/process_4 I0419 22:16:13.308075 136025315080000 metric_logger.py:196] completed step: 6, seconds: 0.296, TFLOP/s/device: 45.887, Tokens/s/device: 6916.582, total_weights: 65536, loss: 8.592, lm_loss: 8.592, perplexity: 5390.759 I0419 22:16:13.455653 136025315080000 metric_logger.py:196] completed step: 7, seconds: 25.375, TFLOP/s/device: 0.535, Tokens/s/device: 80.710, total_weights: 65536, loss: 8.384, lm_loss: 8.384, perplexity: 4376.905 I0419 22:16:13.603051 136025315080000 metric_logger.py:196] completed step: 8, seconds: 0.152, TFLOP/s/device: 89.296, Tokens/s/device: 13459.605, total_weights: 65536, loss: 8.255, lm_loss: 8.255, perplexity: 3846.406 I0419 22:16:13.749742 136025315080000 checkpointing.py:794] Waiting for step 9 to finish before checkpoint... I0419 22:16:13.750303 136025315080000 checkpointing.py:798] Waited 0.0005829334259033203 seconds for step 9 to finish before starting checkpointing. I0419 22:16:13.752709 136025315080000 checkpoint_manager.py:2020] [process=4][thread=MainThread][step=0][wait_until_finished] Waiting for Save Finalize thread (save_finalize) to complete. I0419 22:16:24.024197 135895718033152 base_pytree_checkpoint_handler.py:130] [process=4] /jax/orbax/write/gbytes_per_sec: 41.204 MiB/s (total gbytes: 1.5 GiB) (time elapsed: 38.3361074924469 s) (per-host) I0419 22:16:24.024316 135895718033152 async_checkpointer.py:90] [process=4][thread=async_save] 3 Handler Commit operations completed. Time taken: 38.057352s. I0419 22:16:34.568915 135895718033152 async_checkpointer.py:160] [process=4][thread=async_save] Background save thread done. Time taken: 48.601933s. I0419 22:16:34.569207 135892456806144 async_checkpointer.py:288] [process=4][thread=save_finalize] Done with waiting for background save thread=async_save. I0419 22:16:34.569320 135892456806144 async_checkpointer.py:298] [process=4][thread=save_finalize] No errors found in background save thread=async_save. I0419 22:16:34.569405 135892456806144 checkpoint_manager.py:2137] [process=4][thread=save_finalize][step=0] CheckpointManager Save Finalize is syncing with other hosts... I0419 22:16:34.570894 135892456806144 checkpoint_manager.py:2146] [process=4][thread=save_finalize][step=0] CheckpointManager Save Finalize is done on all hosts. I0419 22:16:34.571061 136025315080000 checkpoint_manager.py:2032] [process=4][thread=MainThread][step=0][wait_until_finished] Done waiting for Save Finalize thread (save_finalize) running at step=0. W0419 22:16:34.571200 136025315080000 checkpoint_manager.py:1452] Waiting for previous save to complete took 20.818491 seconds. If this number is high, consider checkpointing less frequently. I0419 22:16:34.573032 136025315080000 checkpoint_manager.py:1512] [process=4] Saving checkpoint at step 9 I0419 22:16:34.575098 136025315080000 event_tracking.py:70] [process=4] [async] Started save checkpoint @ gs://lance-maxtext/linen_ckpt_xpk_feat_nnx_trainstate_and_training_loop_20260419_214525/linen_xpk_feat_nnx_trainstate_and_training_loop_20260419_214525_07_eval/checkpoints/9. I0419 22:16:35.715369 136025315080000 jax_array_handlers.py:360] Scheduling D2H of 39 prioritized jax.Array. I0419 22:16:35.715463 136025315080000 replica_slices.py:424] Transferring arrays to host memory with options: use_replica_parallel=True, min_slice_bytes_for_replica_parallel=None, max_replicas_for_replica_parallel=None, enable_pinned_host_transfer=False I0419 22:16:35.751483 136025315080000 base_pytree_checkpoint_handler.py:154] [process=4][thread=MainThread] Initiated "orbax.checkpoint._src.serialization.jax_array_handlers.ArrayHandler".serialize. Time taken: 0.037351s I0419 22:16:35.751653 136025315080000 base_pytree_checkpoint_handler.py:130] [process=4] /jax/orbax/write/blocking_gbytes_per_sec: 37.794 GiB/s (total gbytes: 1.5 GiB) (time elapsed: 0.040815114974975586 s) (per-host) I0419 22:16:35.751706 136025315080000 base_pytree_checkpoint_handler.py:768] [process=4][thread=MainThread] Initiated Pytree async_save. Time taken: 0.040878s (batch_requests_ready=0.001821s, total_serialization_initiated=0.038986s, others=0.000071s) I0419 22:16:35.751795 136025315080000 composite_checkpoint_handler.py:715] [process=4][thread=MainThread] Initiated CompositeCheckpointHandler.async_save. Time taken: 0.044803s (all_items=0.000015s, per_item={'items': '0.00001502'}, temp_paths=0.044788) I0419 22:16:35.752492 136025315080000 event_tracking.py:125] [process=4] [async] Finished blocking save in 1.18 seconds. Continuing save @ gs://lance-maxtext/linen_ckpt_xpk_feat_nnx_trainstate_and_training_loop_20260419_214525/linen_xpk_feat_nnx_trainstate_and_training_loop_20260419_214525_07_eval/checkpoints/9. I0419 22:16:35.752777 135892456806144 async_checkpointer.py:76] [process=4][thread=async_save] Background save thread started. Deadline for this save operation is 2026-04-19 22:36:35.752753 I0419 22:16:35.763093 136025315080000 checkpoint_manager.py:1560] [process=4][thread=MainThread][step=9] Starting CheckpointManager Save Finalize thread=save_finalize I0419 22:16:35.763350 135895181162240 async_checkpointer.py:280] [process=4][thread=save_finalize] Waiting for background save thread=async_save. I0419 22:16:35.763480 136025315080000 standard_logger.py:34] {'step': 9, 'event_type': 'save', 'directory': 'gs://lance-maxtext/linen_ckpt_xpk_feat_nnx_trainstate_and_training_loop_20260419_214525/linen_xpk_feat_nnx_trainstate_and_training_loop_20260419_214525_07_eval/checkpoints', 'reached_preemption': False, 'preemption_received_at': None, 'synchronous': False, 'wait_for_prev_start_time': 1776636973.752678, 'wait_for_prev_duration_secs': 20.818490505218506, 'time_between_consecutive_saves_sec': None, 'checkpointer_blocking_start_time': 1776636994.5730708, 'checkpointer_blocking_duration_secs': 1.1798136234283447, 'get_old_steps_start_time': 1776636995.7529063, 'get_old_steps_duration_secs': 2.956390380859375e-05, 'checkpoint_manager_blocking_start_time': 1776636973.7505462, 'checkpoint_manager_blocking_duration_secs': 22.012897729873657} I0419 22:16:35.763585 136025315080000 checkpointing.py:409] Started an asynchronous checkpoint save for step 9 I0419 22:16:35.763631 136025315080000 checkpoint_manager.py:2020] [process=4][thread=MainThread][step=9][wait_until_finished] Waiting for Save Finalize thread (save_finalize) to complete. I0419 22:16:46.620830 135892465198848 array_metadata_store.py:203] [process=4][thread=array_type_handler] Wrote 39 array_metadata.ArrayMetadata to gs://lance-maxtext/linen_ckpt_xpk_feat_nnx_trainstate_and_training_loop_20260419_214525/linen_xpk_feat_nnx_trainstate_and_training_loop_20260419_214525_07_eval/checkpoints/9/items/array_metadatas/process_4 I0419 22:17:21.109883 135892456806144 base_pytree_checkpoint_handler.py:130] [process=4] /jax/orbax/write/gbytes_per_sec: 34.794 MiB/s (total gbytes: 1.5 GiB) (time elapsed: 45.39899802207947 s) (per-host) I0419 22:17:21.109993 135892456806144 async_checkpointer.py:90] [process=4][thread=async_save] 3 Handler Commit operations completed. Time taken: 45.357133s. I0419 22:17:32.212434 135892456806144 async_checkpointer.py:160] [process=4][thread=async_save] Background save thread done. Time taken: 56.459558s. I0419 22:17:32.212659 135895181162240 async_checkpointer.py:288] [process=4][thread=save_finalize] Done with waiting for background save thread=async_save. I0419 22:17:32.212724 135895181162240 async_checkpointer.py:298] [process=4][thread=save_finalize] No errors found in background save thread=async_save. I0419 22:17:32.212772 135895181162240 checkpoint_manager.py:2137] [process=4][thread=save_finalize][step=9] CheckpointManager Save Finalize is syncing with other hosts... I0419 22:17:32.214631 135895181162240 checkpoint_manager.py:2146] [process=4][thread=save_finalize][step=9] CheckpointManager Save Finalize is done on all hosts. I0419 22:17:32.214801 136025315080000 checkpoint_manager.py:2032] [process=4][thread=MainThread][step=9][wait_until_finished] Done waiting for Save Finalize thread (save_finalize) running at step=9. I0419 22:17:32.214949 136025315080000 checkpoint_manager.py:2009] [process=4][thread=MainThread][wait_until_finished] No Save Finalize thread to wait for. Returning. I0419 22:17:32.215904 136025315080000 metric_logger.py:196] completed step: 9, seconds: 0.148, TFLOP/s/device: 92.114, Tokens/s/device: 13884.463, 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: Sun Apr 19 22:17:40 UTC 2026 EXIT_CODE=0