main (8a17c3d19) vs feat/nnx-set-defaults-true (73213e044)| Metric | main 8a17c3d19 | feat/nnx-set-defaults-true 73213e044 | Diff (feat/nnx-set-defaults-true − main) |
|---|---|---|---|
| Parameters | 1.104 billion | 1.104 billion | — |
| Final loss | 6.9810 | 6.9810 | 0 |
| TFLOP/s | 11.325 | 11.278 | -0.047 |
| Tok/s | 1810.0 | 1802.5 | -7.443 |
| Avg s/step | 1.379 | 1.371 | -0.008 |
| Memory % | 1.31 | 1.31 | 0 |
| JAX | 0.9.2 | 0.8.1 | — |
Diff = branch value − main value. Green = branch improved. Red = branch regressed.
XPK Start: Wed Apr 22 07:21:50 UTC 2026 PyTorch was not found. Models won't be available and only tokenizers, configuration and file/data utilities can be used. `rope_parameters`'s factor field must be a float >= 1, got 40 `rope_parameters`'s beta_fast field must be a float, got 32 `rope_parameters`'s beta_slow field must be a float, got 1 DeepseekV32Config got `key=rope_scaling` in kwargs but hasn't set it as attribute. For RoPE standardization you need to set `self.rope_parameters` in model's config. 2026-04-22 07:22:30.380535: 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) I0422 07:22:30.591233 133050128602944 max_utils.py:273] Attempting to initialize the jax distributed system... I0422 07:22:39.630258 133050128602944 distributed.py:149] Starting JAX distributed service on [::]:8482 I0422 07:22:39.632703 133050128602944 distributed.py:172] Connecting to JAX distributed service on mt-03-dropout-4sut3-slice-job-0-0.mt-03-dropout-4sut3:8482 I0422 07:22:43.262246 133050128602944 max_utils.py:284] Jax distributed system initialized! I0422 07:22:49.715569 133050128602944 max_utils.py:800] System Information: Jax Version: 0.9.2 I0422 07:22:49.715675 133050128602944 max_utils.py:801] System Information: Jaxlib Version: 0.9.2 I0422 07:22:49.715714 133050128602944 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 I0422 07:22:49.715750 133050128602944 train_utils.py:361] WARNING: Sequence packing is essentially ignored for synthetic data. Please use a real dataset to use sequence packing. I0422 07:22:50.406398 133050128602944 maxtext_utils.py:1604] Num_devices: 32, shape (1, 1, 1, 32, 1, 1, 1, 1, 1, 1, 1, 1, 1) I0422 07:22:50.406683 133050128602944 checkpointing.py:677] Setting up checkpoint logger... I0422 07:22:50.406738 133050128602944 checkpointing.py:233] Creating checkpoint manager with ocdbt=True and zarr3=True I0422 07:22:50.406782 133050128602944 pytree_checkpoint_handler.py:592] save_device_host_concurrent_bytes=None I0422 07:22:50.407139 133050128602944 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 0x79018a31c770>, enable_pinned_host_transfer=False, save_concurrent_bytes: 96000000000 (89.4 GiB), restore_concurrent_bytes: 96000000000 (89.4 GiB) I0422 07:22:53.470102 133050128602944 checkpointing.py:265] Enabling policy for fixed interval checkpointing. I0422 07:22:53.470373 133050128602944 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 0x78ec285823f0>}, handler_registry=None I0422 07:22:53.470616 133050128602944 composite_checkpoint_handler.py:237] Deferred registration for item: "items". Adding handler `<orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeCheckpointHandler object at 0x78ec285823f0>` 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`. I0422 07:22:53.470668 133050128602944 composite_checkpoint_handler.py:237] Deferred registration for item: "metrics". Adding handler `<orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonCheckpointHandler object at 0x78ed104ce0f0>` 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`. I0422 07:22:53.470702 133050128602944 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 0x78ec285823f0>, ('items', <class 'orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeRestoreArgs'>): <orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeCheckpointHandler object at 0x78ec285823f0>, ('metrics', <class 'orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonSaveArgs'>): <orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonCheckpointHandler object at 0x78ed104ce0f0>, ('metrics', <class 'orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonRestoreArgs'>): <orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonCheckpointHandler object at 0x78ed104ce0f0>}). I0422 07:22:53.471035 133050128602944 abstract_checkpointer.py:35] orbax-checkpoint version: 0.11.34 I0422 07:22:53.471108 133050128602944 async_checkpointer.py:192] [process=6][thread=MainThread] Using barrier_sync_fn: <function get_barrier_sync_fn.<locals>._fn at 0x78ec28741b20> timeout: 1200 secs and primary_host=0 for async checkpoint writes I0422 07:22:54.798958 133050128602944 checkpoint_manager.py:1812] Found 0 checkpoint steps in gs://lance-maxtext/linen_ckpt_xpk_main_20260422_071422/linen_xpk_main_20260422_071422_03_dropout/checkpoints I0422 07:22:55.217518 133050128602944 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_20260422_071422/linen_xpk_main_20260422_071422_03_dropout/checkpoints: <orbax.checkpoint.checkpoint_manager.CheckpointManager object at 0x78ed104caae0> I0422 07:22:55.217693 133050128602944 checkpointing.py:301] Checkpoint manager created! I0422 07:22:55.565344 133050128602944 nnx_wrappers.py:437] Unknown Logical: bfloat16[32,128,2048]....................................... ('activation_batch', 'activation_norm_length', 'activation_embed'). I0422 07:22:55.565466 133050128602944 nnx_wrappers.py:437] Unknown Physical: bfloat16[32,128,2048]....................................... ('fsdp', None, None). I0422 07:22:55.948063 133050128602944 attentions.py:1088] attentions/inputs_q Logical: bfloat16[32,128,2048]....................................... ('activation_batch', 'activation_attn_length', 'activation_attn_embed'). I0422 07:22:55.948156 133050128602944 attentions.py:1088] attentions/inputs_q Physical: bfloat16[32,128,2048]....................................... ('fsdp', None, None). I0422 07:22:55.970173 133050128602944 attentions.py:1089] attentions/inputs_kv Logical: bfloat16[32,128,2048]....................................... ('activation_batch', 'activation_attn_length', 'activation_attn_embed'). I0422 07:22:55.970273 133050128602944 attentions.py:1089] attentions/inputs_kv Physical: bfloat16[32,128,2048]....................................... ('fsdp', None, None). I0422 07:22:55.994108 133050128602944 attentions.py:1154] attentions/query Logical: bfloat16[32,128,16,128]..................................... ('activation_kv_batch', 'activation_attn_length', 'activation_kv_heads', 'activation_kv_head_dim'). I0422 07:22:55.994177 133050128602944 attentions.py:1154] attentions/query Physical: bfloat16[32,128,16,128]..................................... ('fsdp', None, None, None). I0422 07:22:56.010580 133050128602944 attentions.py:1155] attentions/key Logical: bfloat16[32,128,16,128]..................................... ('activation_kv_batch', 'activation_attn_length', 'activation_kv_heads', 'activation_kv_head_dim'). I0422 07:22:56.010638 133050128602944 attentions.py:1155] attentions/key Physical: bfloat16[32,128,16,128]..................................... ('fsdp', None, None, None). I0422 07:22:56.026985 133050128602944 attentions.py:1156] attentions/value Logical: bfloat16[32,128,16,128]..................................... ('activation_kv_batch', 'activation_attn_length', 'activation_kv_heads', 'activation_kv_head_dim'). I0422 07:22:56.027046 133050128602944 attentions.py:1156] attentions/value Physical: bfloat16[32,128,16,128]..................................... ('fsdp', None, None, None). I0422 07:22:56.051961 133050128602944 attentions.py:1198] attentions/out Logical: bfloat16[32,128,16,128]..................................... ('activation_batch', 'activation_attn_length', 'activation_heads', 'activation_kv'). I0422 07:22:56.052033 133050128602944 attentions.py:1198] attentions/out Physical: bfloat16[32,128,16,128]..................................... ('fsdp', None, None, None). I0422 07:22:56.079391 133050128602944 linears.py:525] linears/x Logical: bfloat16[32,128,7168]....................................... ('activation_batch', 'activation_length', 'activation_mlp'). I0422 07:22:56.079457 133050128602944 linears.py:525] linears/x Physical: bfloat16[32,128,7168]....................................... ('fsdp', None, None). I0422 07:22:56.297951 133050128602944 checkpointing.py:577] checkpoint manager exists so trying to load this run's existing checkpoint I0422 07:22:56.298065 133050128602944 checkpointing.py:665] No existing checkpoints found, not restoring checkpoint. fsdp: 32 I0422 07:22:57.738766 133050128602944 maxtext_utils.py:1707] params/params/decoder/decoder_norm/scale Shape: float32[2048] Logical: P('norm',) Physical: (None,) I0422 07:22:57.738892 133050128602944 maxtext_utils.py:1707] params/params/decoder/layers/mlp/wi_0/kernel Shape: float32[2048,16,7168] Logical: P('embed', 'layers', 'mlp') Physical: ('fsdp', None, None) I0422 07:22:57.738946 133050128602944 maxtext_utils.py:1707] params/params/decoder/layers/mlp/wi_1/kernel Shape: float32[2048,16,7168] Logical: P('embed', 'layers', 'mlp') Physical: ('fsdp', None, None) I0422 07:22:57.739005 133050128602944 maxtext_utils.py:1707] params/params/decoder/layers/mlp/wo/kernel Shape: float32[7168,16,2048] Logical: P('mlp', 'layers', 'embed') Physical: (None, None, 'fsdp') I0422 07:22:57.739058 133050128602944 maxtext_utils.py:1707] params/params/decoder/layers/post_self_attention_layer_norm/scale Shape: float32[2048,16] Logical: P('norm', 'layers') Physical: (None, None) I0422 07:22:57.739096 133050128602944 maxtext_utils.py:1707] params/params/decoder/layers/pre_self_attention_layer_norm/scale Shape: float32[2048,16] Logical: P('norm', 'layers') Physical: (None, None) I0422 07:22:57.739148 133050128602944 maxtext_utils.py:1707] 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) I0422 07:22:57.739200 133050128602944 maxtext_utils.py:1707] 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') I0422 07:22:57.739240 133050128602944 maxtext_utils.py:1707] 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) I0422 07:22:57.739275 133050128602944 maxtext_utils.py:1707] 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) I0422 07:22:57.739347 133050128602944 maxtext_utils.py:1707] params/params/decoder/logits_dense/kernel Shape: float32[2048,32000] Logical: P('embed_vocab', 'vocab') Physical: ('fsdp', None) I0422 07:22:57.739700 133050128602944 maxtext_utils.py:1707] params/params/token_embedder/embedding Shape: float32[32000,2048] Logical: P('vocab', 'embed_vocab') Physical: (None, 'fsdp') I0422 07:22:58.264539 133050128602944 train.py:155] train/xent Logical: float32[32,128]............................................. ('activation_embed_and_logits_batch', 'activation_length'). I0422 07:22:58.264635 133050128602944 train.py:155] train/xent Physical: float32[32,128]............................................. ('fsdp', None). I0422 07:22:58.280124 133050128602944 train.py:162] train/z_loss Logical: float32[32,128]............................................. ('activation_embed_and_logits_batch', 'activation_length'). I0422 07:22:58.280184 133050128602944 train.py:162] train/z_loss Physical: float32[32,128]............................................. ('fsdp', None). I0422 07:23:01.937496 133050128602944 max_utils.py:791] Total memory size: 0.8 GB, Output size: 0.4 GB, Temp size: 0.4 GB, Argument size: 0.4 GB, Host temp size: 0.0 GB. I0422 07:23:01.938299 133050128602944 metric_logger.py:301] number parameters: 1.104 billion I0422 07:23:06.505146 133050128602944 checkpointing.py:772] Waiting for step 0 to finish before checkpoint... I0422 07:23:06.581360 133050128602944 checkpointing.py:776] Waited 0.07619762420654297 seconds for step 0 to finish before starting checkpointing. I0422 07:23:06.583818 133050128602944 checkpoint_manager.py:2009] [process=6][thread=MainThread][wait_until_finished] No Save Finalize thread to wait for. Returning. I0422 07:23:06.585791 133050128602944 checkpoint_manager.py:1512] [process=6] Saving checkpoint at step 0 I0422 07:23:06.587258 133050128602944 event_tracking.py:70] [process=6] [async] Started save checkpoint @ gs://lance-maxtext/linen_ckpt_xpk_main_20260422_071422/linen_xpk_main_20260422_071422_03_dropout/checkpoints/0. I0422 07:23:07.340304 133050128602944 signaling_client.py:364] Using JaxDistributedSignalingClient I0422 07:23:07.341246 133050128602944 jax_array_handlers.py:360] Scheduling D2H of 39 prioritized jax.Array. I0422 07:23:07.341323 133050128602944 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 I0422 07:23:07.625578 133050128602944 base_pytree_checkpoint_handler.py:154] [process=6][thread=MainThread] Initiated "orbax.checkpoint._src.serialization.jax_array_handlers.ArrayHandler".serialize. Time taken: 0.285358s I0422 07:23:07.625751 133050128602944 base_pytree_checkpoint_handler.py:130] [process=6] /jax/orbax/write/blocking_gbytes_per_sec: 5.305 GiB/s (total gbytes: 1.5 GiB) (time elapsed: 0.2907707691192627 s) (per-host) I0422 07:23:07.625805 133050128602944 base_pytree_checkpoint_handler.py:768] [process=6][thread=MainThread] Initiated Pytree async_save. Time taken: 0.290834s (batch_requests_ready=0.002182s, total_serialization_initiated=0.288581s, others=0.000071s) I0422 07:23:07.625895 133050128602944 composite_checkpoint_handler.py:715] [process=6][thread=MainThread] Initiated CompositeCheckpointHandler.async_save. Time taken: 0.294980s (all_items=0.000018s, per_item={'items': '0.00001836'}, temp_paths=0.294961) I0422 07:23:07.626746 133050128602944 event_tracking.py:125] [process=6] [async] Finished blocking save in 1.04 seconds. Continuing save @ gs://lance-maxtext/linen_ckpt_xpk_main_20260422_071422/linen_xpk_main_20260422_071422_03_dropout/checkpoints/0. I0422 07:23:07.627102 132921511896832 async_checkpointer.py:76] [process=6][thread=async_save] Background save thread started. Deadline for this save operation is 2026-04-22 07:43:07.627064 I0422 07:23:07.629451 133050128602944 checkpoint_manager.py:1560] [process=6][thread=MainThread][step=0] Starting CheckpointManager Save Finalize thread=save_finalize I0422 07:23:07.629718 132920446547712 async_checkpointer.py:280] [process=6][thread=save_finalize] Waiting for background save thread=async_save. I0422 07:23:07.629883 133050128602944 standard_logger.py:34] {'step': 0, 'event_type': 'save', 'directory': 'gs://lance-maxtext/linen_ckpt_xpk_main_20260422_071422/linen_xpk_main_20260422_071422_03_dropout/checkpoints', 'reached_preemption': False, 'preemption_received_at': None, 'synchronous': False, 'wait_for_prev_start_time': 1776842586.5838006, 'wait_for_prev_duration_secs': 6.103515625e-05, 'time_between_consecutive_saves_sec': None, 'checkpointer_blocking_start_time': 1776842586.585831, 'checkpointer_blocking_duration_secs': 1.0414295196533203, 'get_old_steps_start_time': 1776842587.6273084, 'get_old_steps_duration_secs': 2.9087066650390625e-05, 'checkpoint_manager_blocking_start_time': 1776842586.5818605, 'checkpoint_manager_blocking_duration_secs': 1.0479793548583984} I0422 07:23:07.629992 133050128602944 checkpointing.py:408] Started an asynchronous checkpoint save for step 0 I0422 07:23:07.630043 133050128602944 max_utils.py:750] Memstats: After params initialized: I0422 07:23:07.630095 133050128602944 max_utils.py:756] Using (GB) 0.41 / 31.25 (1.312000%) on TPU_24(process=6,(0,6,0,0)) I0422 07:23:07.630128 133050128602944 max_utils.py:756] Using (GB) 0.41 / 31.25 (1.312000%) on TPU_25(process=6,(1,6,0,0)) I0422 07:23:07.630156 133050128602944 max_utils.py:756] Using (GB) 0.41 / 31.25 (1.312000%) on TPU_28(process=6,(0,7,0,0)) I0422 07:23:07.630182 133050128602944 max_utils.py:756] Using (GB) 0.41 / 31.25 (1.312000%) on TPU_29(process=6,(1,7,0,0)) I0422 07:23:07.942533 133050128602944 metric_logger.py:196] completed step: 0, seconds: 4.567, TFLOP/s/device: 0.175, Tokens/s/device: 28.029, total_weights: 4096, loss: 10.889, lm_loss: 10.889, perplexity: 53560.699 I0422 07:23:08.035938 133050128602944 metric_logger.py:196] completed step: 1, seconds: 1.436, TFLOP/s/device: 0.558, Tokens/s/device: 89.149, total_weights: 4096, loss: 10.902, lm_loss: 10.902, perplexity: 54298.523 I0422 07:23:08.466557 133050128602944 metric_logger.py:196] completed step: 2, seconds: 0.023, TFLOP/s/device: 34.280, Tokens/s/device: 5478.748, total_weights: 4096, loss: 9.956, lm_loss: 9.956, perplexity: 21082.051 I0422 07:23:08.537240 133050128602944 metric_logger.py:196] completed step: 3, seconds: 0.431, TFLOP/s/device: 1.858, Tokens/s/device: 296.950, total_weights: 4096, loss: 9.139, lm_loss: 9.139, perplexity: 9314.691 I0422 07:23:08.678953 133050128602944 metric_logger.py:196] completed step: 4, seconds: 0.076, TFLOP/s/device: 10.528, Tokens/s/device: 1682.572, total_weights: 4096, loss: 8.459, lm_loss: 8.459, perplexity: 4715.493 I0422 07:23:08.684755 133050128602944 metric_logger.py:196] completed step: 5, seconds: 0.071, TFLOP/s/device: 11.348, Tokens/s/device: 1813.725, total_weights: 4096, loss: 7.924, lm_loss: 7.924, perplexity: 2762.373 I0422 07:23:10.517609 2773 google_auth_provider.cc:181] Running on GCE, using service account 562977990677-compute@developer.gserviceaccount.com I0422 07:23:13.483686 132921465747200 array_metadata_store.py:203] [process=6][thread=array_type_handler] Wrote 39 array_metadata.ArrayMetadata to gs://lance-maxtext/linen_ckpt_xpk_main_20260422_071422/linen_xpk_main_20260422_071422_03_dropout/checkpoints/0/items/array_metadatas/process_6 I0422 07:23:18.839264 133050128602944 metric_logger.py:196] completed step: 6, seconds: 0.142, TFLOP/s/device: 5.646, Tokens/s/device: 902.374, total_weights: 4096, loss: 7.553, lm_loss: 7.553, perplexity: 1906.617 I0422 07:23:18.909924 133050128602944 metric_logger.py:196] completed step: 7, seconds: 10.085, TFLOP/s/device: 0.079, Tokens/s/device: 12.692, total_weights: 4096, loss: 7.267, lm_loss: 7.267, perplexity: 1431.653 I0422 07:23:18.982053 133050128602944 metric_logger.py:196] completed step: 8, seconds: 0.075, TFLOP/s/device: 10.655, Tokens/s/device: 1702.920, total_weights: 4096, loss: 7.111, lm_loss: 7.111, perplexity: 1225.983 I0422 07:23:19.053263 133050128602944 checkpointing.py:772] Waiting for step 9 to finish before checkpoint... I0422 07:23:19.053849 133050128602944 checkpointing.py:776] Waited 0.0006043910980224609 seconds for step 9 to finish before starting checkpointing. I0422 07:23:19.055932 133050128602944 checkpoint_manager.py:2020] [process=6][thread=MainThread][step=0][wait_until_finished] Waiting for Save Finalize thread (save_finalize) to complete. I0422 07:23:40.697907 132921511896832 base_pytree_checkpoint_handler.py:130] [process=6] /jax/orbax/write/gbytes_per_sec: 47.346 MiB/s (total gbytes: 1.5 GiB) (time elapsed: 33.36286902427673 s) (per-host) I0422 07:23:40.698032 132921511896832 async_checkpointer.py:90] [process=6][thread=async_save] 3 Handler Commit operations completed. Time taken: 33.070819s. I0422 07:23:49.725019 132921511896832 async_checkpointer.py:160] [process=6][thread=async_save] Background save thread done. Time taken: 42.097790s. I0422 07:23:49.725394 132920446547712 async_checkpointer.py:288] [process=6][thread=save_finalize] Done with waiting for background save thread=async_save. I0422 07:23:49.725518 132920446547712 async_checkpointer.py:298] [process=6][thread=save_finalize] No errors found in background save thread=async_save. I0422 07:23:49.725570 132920446547712 checkpoint_manager.py:2137] [process=6][thread=save_finalize][step=0] CheckpointManager Save Finalize is syncing with other hosts... I0422 07:23:49.728118 132920446547712 checkpoint_manager.py:2146] [process=6][thread=save_finalize][step=0] CheckpointManager Save Finalize is done on all hosts. I0422 07:23:49.728255 133050128602944 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. W0422 07:23:49.728362 133050128602944 checkpoint_manager.py:1452] Waiting for previous save to complete took 30.672431 seconds. If this number is high, consider checkpointing less frequently. I0422 07:23:49.730231 133050128602944 checkpoint_manager.py:1512] [process=6] Saving checkpoint at step 9 I0422 07:23:49.732261 133050128602944 event_tracking.py:70] [process=6] [async] Started save checkpoint @ gs://lance-maxtext/linen_ckpt_xpk_main_20260422_071422/linen_xpk_main_20260422_071422_03_dropout/checkpoints/9. I0422 07:23:50.457462 133050128602944 jax_array_handlers.py:360] Scheduling D2H of 39 prioritized jax.Array. I0422 07:23:50.457559 133050128602944 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 I0422 07:23:50.495778 133050128602944 base_pytree_checkpoint_handler.py:154] [process=6][thread=MainThread] Initiated "orbax.checkpoint._src.serialization.jax_array_handlers.ArrayHandler".serialize. Time taken: 0.039526s I0422 07:23:50.495941 133050128602944 base_pytree_checkpoint_handler.py:130] [process=6] /jax/orbax/write/blocking_gbytes_per_sec: 35.912 GiB/s (total gbytes: 1.5 GiB) (time elapsed: 0.042954206466674805 s) (per-host) I0422 07:23:50.495990 133050128602944 base_pytree_checkpoint_handler.py:768] [process=6][thread=MainThread] Initiated Pytree async_save. Time taken: 0.043012s (batch_requests_ready=0.001786s, total_serialization_initiated=0.041161s, others=0.000066s) I0422 07:23:50.496095 133050128602944 composite_checkpoint_handler.py:715] [process=6][thread=MainThread] Initiated CompositeCheckpointHandler.async_save. Time taken: 0.047304s (all_items=0.000016s, per_item={'items': '0.00001645'}, temp_paths=0.047287) I0422 07:23:50.496805 133050128602944 event_tracking.py:125] [process=6] [async] Finished blocking save in 0.77 seconds. Continuing save @ gs://lance-maxtext/linen_ckpt_xpk_main_20260422_071422/linen_xpk_main_20260422_071422_03_dropout/checkpoints/9. I0422 07:23:50.497134 132920446547712 async_checkpointer.py:76] [process=6][thread=async_save] Background save thread started. Deadline for this save operation is 2026-04-22 07:43:50.497097 I0422 07:23:50.499111 133050128602944 checkpoint_manager.py:1560] [process=6][thread=MainThread][step=9] Starting CheckpointManager Save Finalize thread=save_finalize I0422 07:23:50.499372 132921486132992 async_checkpointer.py:280] [process=6][thread=save_finalize] Waiting for background save thread=async_save. I0422 07:23:50.499540 133050128602944 standard_logger.py:34] {'step': 9, 'event_type': 'save', 'directory': 'gs://lance-maxtext/linen_ckpt_xpk_main_20260422_071422/linen_xpk_main_20260422_071422_03_dropout/checkpoints', 'reached_preemption': False, 'preemption_received_at': None, 'synchronous': False, 'wait_for_prev_start_time': 1776842599.0559027, 'wait_for_prev_duration_secs': 30.672431230545044, 'time_between_consecutive_saves_sec': None, 'checkpointer_blocking_start_time': 1776842629.7302697, 'checkpointer_blocking_duration_secs': 0.7670114040374756, 'get_old_steps_start_time': 1776842630.4973161, 'get_old_steps_duration_secs': 3.457069396972656e-05, 'checkpoint_manager_blocking_start_time': 1776842599.0540824, 'checkpoint_manager_blocking_duration_secs': 31.445420742034912} I0422 07:23:50.499658 133050128602944 checkpointing.py:408] Started an asynchronous checkpoint save for step 9 I0422 07:23:50.499703 133050128602944 checkpoint_manager.py:2020] [process=6][thread=MainThread][step=9][wait_until_finished] Waiting for Save Finalize thread (save_finalize) to complete. I0422 07:23:56.464667 132921494525696 array_metadata_store.py:203] [process=6][thread=array_type_handler] Wrote 39 array_metadata.ArrayMetadata to gs://lance-maxtext/linen_ckpt_xpk_main_20260422_071422/linen_xpk_main_20260422_071422_03_dropout/checkpoints/9/items/array_metadatas/process_6 I0422 07:24:31.740633 132920446547712 base_pytree_checkpoint_handler.py:130] [process=6] /jax/orbax/write/gbytes_per_sec: 38.258 MiB/s (total gbytes: 1.5 GiB) (time elapsed: 41.287593841552734 s) (per-host) I0422 07:24:31.740760 132920446547712 async_checkpointer.py:90] [process=6][thread=async_save] 3 Handler Commit operations completed. Time taken: 41.243513s. I0422 07:24:40.398932 132920446547712 async_checkpointer.py:160] [process=6][thread=async_save] Background save thread done. Time taken: 49.901670s. I0422 07:24:40.399232 132921486132992 async_checkpointer.py:288] [process=6][thread=save_finalize] Done with waiting for background save thread=async_save. I0422 07:24:40.399379 132921486132992 async_checkpointer.py:298] [process=6][thread=save_finalize] No errors found in background save thread=async_save. I0422 07:24:40.399427 132921486132992 checkpoint_manager.py:2137] [process=6][thread=save_finalize][step=9] CheckpointManager Save Finalize is syncing with other hosts... I0422 07:24:40.401111 132921486132992 checkpoint_manager.py:2146] [process=6][thread=save_finalize][step=9] CheckpointManager Save Finalize is done on all hosts. I0422 07:24:40.401310 133050128602944 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. I0422 07:24:40.401468 133050128602944 checkpoint_manager.py:2009] [process=6][thread=MainThread][wait_until_finished] No Save Finalize thread to wait for. Returning. I0422 07:24:40.402428 133050128602944 metric_logger.py:196] completed step: 9, seconds: 0.071, TFLOP/s/device: 11.325, Tokens/s/device: 1809.955, total_weights: 4096, loss: 6.981, lm_loss: 6.981, perplexity: 1075.867 Per train step: Total TFLOPs: 0.80 split as 99.60% learnable weight flops and 0.40% attention flops XPK End: Wed Apr 22 07:24:50 UTC 2026 EXIT_CODE=0
XPK Start: Wed Apr 22 15:54:37 UTC 2026 2026-04-22 15:54:41.898493: E external/local_xla/xla/stream_executor/cuda/cuda_fft.cc:467] Unable to register cuFFT factory: Attempting to register factory for plugin cuFFT when one has already been registered WARNING: All log messages before absl::InitializeLog() is called are written to STDERR E0000 00:00:1776873281.911337 10 cuda_dnn.cc:8579] Unable to register cuDNN factory: Attempting to register factory for plugin cuDNN when one has already been registered E0000 00:00:1776873281.915030 10 cuda_blas.cc:1407] Unable to register cuBLAS factory: Attempting to register factory for plugin cuBLAS when one has already been registered W0000 00:00:1776873281.926241 10 computation_placer.cc:177] computation placer already registered. Please check linkage and avoid linking the same target more than once. W0000 00:00:1776873281.926260 10 computation_placer.cc:177] computation placer already registered. Please check linkage and avoid linking the same target more than once. W0000 00:00:1776873281.926262 10 computation_placer.cc:177] computation placer already registered. Please check linkage and avoid linking the same target more than once. W0000 00:00:1776873281.926264 10 computation_placer.cc:177] computation placer already registered. Please check linkage and avoid linking the same target more than once. 2026-04-22 15:55:16.400022: 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) I0422 15:55:16.923345 135166776719168 max_utils.py:273] Attempting to initialize the jax distributed system... INFO:2026-04-22 15:55:25,965:jax._src.distributed:140: Starting JAX distributed service on [::]:8482 I0422 15:55:25.965564 135166776719168 distributed.py:140] Starting JAX distributed service on [::]:8482 INFO:2026-04-22 15:55:25,968:jax._src.distributed:157: Connecting to JAX distributed service on mt-03-dropout-35co5-slice-job-0-0.mt-03-dropout-35co5:8482 I0422 15:55:25.968151 135166776719168 distributed.py:157] Connecting to JAX distributed service on mt-03-dropout-35co5-slice-job-0-0.mt-03-dropout-35co5:8482 I0422 15:55:27.844243 135166776719168 max_utils.py:284] Jax distributed system initialized! I0422 15:55:34.908928 135166776719168 max_utils.py:800] System Information: Jax Version: 0.8.1 I0422 15:55:34.909034 135166776719168 max_utils.py:801] System Information: Jaxlib Version: 0.8.1 I0422 15:55:34.909075 135166776719168 max_utils.py:802] System Information: Jax Backend: PJRT C API TFRT TPU v6 lite Built on Nov 12 2025 14:16:36 (1762985796) cl/831091709 I0422 15:55:34.909109 135166776719168 train_utils.py:377] WARNING: Sequence packing is essentially ignored for synthetic data. Please use a real dataset to use sequence packing. I0422 15:55:35.522435 135166776719168 maxtext_utils.py:1631] Num_devices: 32, shape (1, 1, 1, 32, 1, 1, 1, 1, 1, 1, 1, 1, 1) I0422 15:55:35.523038 135166776719168 maxtext_utils.py:1631] Num_devices: 32, shape (1, 1, 1, 32, 1, 1, 1, 1, 1, 1, 1, 1, 1) I0422 15:55:35.523222 135166776719168 checkpointing.py:688] Setting up checkpoint logger... I0422 15:55:35.523272 135166776719168 checkpointing.py:234] Creating checkpoint manager with ocdbt=True and zarr3=True I0422 15:55:35.523323 135166776719168 pytree_checkpoint_handler.py:589] save_device_host_concurrent_bytes=None I0422 15:55:35.523685 135166776719168 base_pytree_checkpoint_handler.py:415] Created BasePyTreeCheckpointHandler: use_ocdbt=True, use_zarr3=True, pytree_metadata_options=PyTreeMetadataOptions(support_rich_types=False), array_metadata_store=<orbax.checkpoint._src.metadata.array_metadata_store.Store object at 0x7aee64553830>, enable_pinned_host_transfer=False, save_concurrent_bytes: 96000000000 (89.4 GiB), restore_concurrent_bytes: 96000000000 (89.4 GiB) I0422 15:55:38.590826 135166776719168 checkpointing.py:266] Enabling policy for fixed interval checkpointing. I0422 15:55:38.591167 135166776719168 checkpoint_manager.py:709] [process=5][thread=MainThread] CheckpointManager init: checkpointers=None, item_names=('items',), item_handlers={'items': <orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeCheckpointHandler object at 0x7ad9b40b5e50>}, handler_registry=None I0422 15:55:38.591408 135166776719168 composite_checkpoint_handler.py:237] Deferred registration for item: "items". Adding handler `<orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeCheckpointHandler object at 0x7ad9b40b5e50>` 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`. I0422 15:55:38.591457 135166776719168 composite_checkpoint_handler.py:237] Deferred registration for item: "metrics". Adding handler `<orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonCheckpointHandler object at 0x7aed327cc500>` 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`. I0422 15:55:38.591492 135166776719168 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 0x7ad9b40b5e50>, ('items', <class 'orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeRestoreArgs'>): <orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeCheckpointHandler object at 0x7ad9b40b5e50>, ('metrics', <class 'orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonSaveArgs'>): <orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonCheckpointHandler object at 0x7aed327cc500>, ('metrics', <class 'orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonRestoreArgs'>): <orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonCheckpointHandler object at 0x7aed327cc500>}). I0422 15:55:38.592008 135166776719168 abstract_checkpointer.py:35] orbax-checkpoint version: 0.11.33 I0422 15:55:38.592086 135166776719168 async_checkpointer.py:177] [process=5][thread=MainThread] Using barrier_sync_fn: <function get_barrier_sync_fn.<locals>._fn at 0x7ad93c760180> timeout: 600 secs and primary_host=0 for async checkpoint writes I0422 15:55:40.172848 135166776719168 checkpoint_manager.py:1818] Found 0 checkpoint steps in gs://lance-maxtext/linen_ckpt_xpk_feat_nnx_set_defaults_true_20260422_154647/linen_xpk_feat_nnx_set_defaults_true_20260422_154647_03_dropout/checkpoints I0422 15:55:40.205024 135166776719168 checkpoint_manager.py:929] [process=5][thread=MainThread] CheckpointManager created, primary_host=0, CheckpointManagerOptions=CheckpointManagerOptions(save_interval_steps=1, max_to_keep=None, keep_time_interval=None, keep_period=None, should_keep_fn=None, best_fn=None, best_mode='max', keep_checkpoints_without_metrics=True, step_prefix=None, step_format_fixed_length=None, step_name_format=None, create=True, cleanup_tmp_directories=False, save_on_steps=frozenset(), single_host_load_and_broadcast=False, todelete_subdir=None, todelete_full_path=None, enable_background_delete=False, read_only=False, enable_async_checkpointing=True, async_options=None, multiprocessing_options=MultiprocessingOptions(primary_host=0, active_processes=None, barrier_sync_key_prefix=None), should_save_fn=None, file_options=FileOptions(path_permission_mode=None), save_root_metadata=True, temporary_path_class=None, save_decision_policy=FixedIntervalPolicy(interval=10), preservation_policy=LatestN(n=None), prevent_write_metrics=False, enable_should_save_is_saving_in_progress_check=True, enable_per_process_directory_creation=False, lightweight_initialize=False), root_directory=gs://lance-maxtext/linen_ckpt_xpk_feat_nnx_set_defaults_true_20260422_154647/linen_xpk_feat_nnx_set_defaults_true_20260422_154647_03_dropout/checkpoints: <orbax.checkpoint.checkpoint_manager.CheckpointManager object at 0x7ad9b44a38f0> I0422 15:55:40.205156 135166776719168 checkpointing.py:302] Checkpoint manager created! I0422 15:55:40.519065 135166776719168 nnx_wrappers.py:455] Unknown Logical: bfloat16[32,128,2048]....................................... ('activation_batch', 'activation_norm_length', 'activation_embed'). I0422 15:55:40.519170 135166776719168 nnx_wrappers.py:455] Unknown Physical: bfloat16[32,128,2048]....................................... ('fsdp', None, None). I0422 15:55:40.890902 135166776719168 attentions.py:1084] attentions/inputs_q Logical: bfloat16[32,128,2048]....................................... ('activation_batch', 'activation_attn_length', 'activation_attn_embed'). I0422 15:55:40.890994 135166776719168 attentions.py:1084] attentions/inputs_q Physical: bfloat16[32,128,2048]....................................... ('fsdp', None, None). I0422 15:55:40.906942 135166776719168 attentions.py:1085] attentions/inputs_kv Logical: bfloat16[32,128,2048]....................................... ('activation_batch', 'activation_attn_length', 'activation_attn_embed'). I0422 15:55:40.907003 135166776719168 attentions.py:1085] attentions/inputs_kv Physical: bfloat16[32,128,2048]....................................... ('fsdp', None, None). I0422 15:55:40.929751 135166776719168 attentions.py:1150] attentions/query Logical: bfloat16[32,128,16,128]..................................... ('activation_kv_batch', 'activation_attn_length', 'activation_kv_heads', 'activation_kv_head_dim'). I0422 15:55:40.929826 135166776719168 attentions.py:1150] attentions/query Physical: bfloat16[32,128,16,128]..................................... ('fsdp', None, None, None). I0422 15:55:40.945747 135166776719168 attentions.py:1151] attentions/key Logical: bfloat16[32,128,16,128]..................................... ('activation_kv_batch', 'activation_attn_length', 'activation_kv_heads', 'activation_kv_head_dim'). I0422 15:55:40.945807 135166776719168 attentions.py:1151] attentions/key Physical: bfloat16[32,128,16,128]..................................... ('fsdp', None, None, None). I0422 15:55:40.961671 135166776719168 attentions.py:1152] attentions/value Logical: bfloat16[32,128,16,128]..................................... ('activation_kv_batch', 'activation_attn_length', 'activation_kv_heads', 'activation_kv_head_dim'). I0422 15:55:40.961727 135166776719168 attentions.py:1152] attentions/value Physical: bfloat16[32,128,16,128]..................................... ('fsdp', None, None, None). I0422 15:55:40.986331 135166776719168 attentions.py:1193] attentions/out Logical: bfloat16[32,128,16,128]..................................... ('activation_batch', 'activation_attn_length', 'activation_heads', 'activation_kv'). I0422 15:55:40.986397 135166776719168 attentions.py:1193] attentions/out Physical: bfloat16[32,128,16,128]..................................... ('fsdp', None, None, None). I0422 15:55:41.012077 135166776719168 linears.py:541] linears/x Logical: bfloat16[32,128,7168]....................................... ('activation_batch', 'activation_length', 'activation_mlp'). I0422 15:55:41.012153 135166776719168 linears.py:541] linears/x Physical: bfloat16[32,128,7168]....................................... ('fsdp', None, None). I0422 15:55:41.218458 135166776719168 checkpointing.py:578] checkpoint manager exists so trying to load this run's existing checkpoint I0422 15:55:41.218567 135166776719168 checkpointing.py:676] No existing checkpoints found, not restoring checkpoint. fsdp: 32 I0422 15:55:42.665132 135166776719168 maxtext_utils.py:1740] params/params/decoder/decoder_norm/scale Shape: float32[2048] Logical: PartitionSpec('norm',) Physical: (None,) I0422 15:55:42.665261 135166776719168 maxtext_utils.py:1740] params/params/decoder/layers/mlp/wi_0/kernel Shape: float32[2048,16,7168] Logical: PartitionSpec('embed', 'layers', 'mlp') Physical: ('fsdp', None, None) I0422 15:55:42.665317 135166776719168 maxtext_utils.py:1740] params/params/decoder/layers/mlp/wi_1/kernel Shape: float32[2048,16,7168] Logical: PartitionSpec('embed', 'layers', 'mlp') Physical: ('fsdp', None, None) I0422 15:55:42.665375 135166776719168 maxtext_utils.py:1740] params/params/decoder/layers/mlp/wo/kernel Shape: float32[7168,16,2048] Logical: PartitionSpec('mlp', 'layers', 'embed') Physical: (None, None, 'fsdp') I0422 15:55:42.665427 135166776719168 maxtext_utils.py:1740] params/params/decoder/layers/post_self_attention_layer_norm/scale Shape: float32[2048,16] Logical: PartitionSpec('norm', 'layers') Physical: (None, None) I0422 15:55:42.665466 135166776719168 maxtext_utils.py:1740] params/params/decoder/layers/pre_self_attention_layer_norm/scale Shape: float32[2048,16] Logical: PartitionSpec('norm', 'layers') Physical: (None, None) I0422 15:55:42.665519 135166776719168 maxtext_utils.py:1740] params/params/decoder/layers/self_attention/key/kernel Shape: float32[2048,16,16,128] Logical: PartitionSpec('embed', 'layers', 'kv_heads', 'kv_head_dim') Physical: ('fsdp', None, None, None) I0422 15:55:42.665571 135166776719168 maxtext_utils.py:1740] params/params/decoder/layers/self_attention/out/kernel Shape: float32[16,16,128,2048] Logical: PartitionSpec('heads', 'layers', 'kv', 'embed') Physical: (None, None, None, 'fsdp') I0422 15:55:42.665611 135166776719168 maxtext_utils.py:1740] params/params/decoder/layers/self_attention/query/kernel Shape: float32[2048,16,16,128] Logical: PartitionSpec('embed', 'layers', 'q_heads', 'kv') Physical: ('fsdp', None, None, None) I0422 15:55:42.665661 135166776719168 maxtext_utils.py:1740] params/params/decoder/layers/self_attention/value/kernel Shape: float32[2048,16,16,128] Logical: PartitionSpec('embed', 'layers', 'kv_heads', 'kv_head_dim') Physical: ('fsdp', None, None, None) I0422 15:55:42.665729 135166776719168 maxtext_utils.py:1740] params/params/decoder/logits_dense/kernel Shape: float32[2048,32000] Logical: PartitionSpec('embed', 'vocab') Physical: ('fsdp', None) I0422 15:55:42.665782 135166776719168 maxtext_utils.py:1740] params/params/token_embedder/embedding Shape: float32[32000,2048] Logical: PartitionSpec('vocab', 'embed') Physical: (None, 'fsdp') I0422 15:55:43.192766 135166776719168 train.py:158] train/xent Logical: float32[32,128]............................................. ('activation_embed_and_logits_batch', 'activation_length'). I0422 15:55:43.192858 135166776719168 train.py:158] train/xent Physical: float32[32,128]............................................. ('fsdp', None). I0422 15:55:43.207619 135166776719168 train.py:165] train/z_loss Logical: float32[32,128]............................................. ('activation_embed_and_logits_batch', 'activation_length'). I0422 15:55:43.207689 135166776719168 train.py:165] train/z_loss Physical: float32[32,128]............................................. ('fsdp', None). I0422 15:55:46.628247 135166776719168 max_utils.py:791] Total memory size: 0.8 GB, Output size: 0.4 GB, Temp size: 0.4 GB, Argument size: 0.4 GB, Host temp size: 0.0 GB. I0422 15:55:46.629097 135166776719168 metric_logger.py:289] number parameters: 1.104 billion I0422 15:55:50.484800 135166776719168 checkpointing.py:794] Waiting for step 0 to finish before checkpoint... I0422 15:55:50.801251 135166776719168 checkpointing.py:798] Waited 0.31643009185791016 seconds for step 0 to finish before starting checkpointing. I0422 15:55:50.803962 135166776719168 checkpoint_manager.py:2013] [process=5][thread=MainThread][wait_until_finished] No Save Finalize thread to wait for. Returning. I0422 15:55:50.805785 135166776719168 checkpoint_manager.py:1518] [process=5] Saving checkpoint at step 0 I0422 15:55:50.808532 135166776719168 async_checkpointer.py:452] [process=5] Started async saving checkpoint to gs://lance-maxtext/linen_ckpt_xpk_feat_nnx_set_defaults_true_20260422_154647/linen_xpk_feat_nnx_set_defaults_true_20260422_154647_03_dropout/checkpoints/0. I0422 15:55:51.853190 135166776719168 signaling_client.py:364] Using JaxDistributedSignalingClient I0422 15:55:51.854409 135166776719168 jax_array_handlers.py:358] Scheduling D2H of 39 prioritized jax.Array. I0422 15:55:51.854468 135166776719168 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 I0422 15:55:52.096156 135166776719168 base_pytree_checkpoint_handler.py:153] [process=5][thread=MainThread] Initiated "orbax.checkpoint._src.serialization.jax_array_handlers.ArrayHandler".serialize. Time taken: 0.243042s I0422 15:55:52.096335 135166776719168 base_pytree_checkpoint_handler.py:129] [process=5] /jax/checkpoint/write/blocking_gbytes_per_sec: 2.951 GiB/s (total gbytes: 1.5 GiB) (time elapsed: 0.5227904319763184 s) (per-host) I0422 15:55:52.096389 135166776719168 base_pytree_checkpoint_handler.py:737] [process=5][thread=MainThread] Initiated Pytree async_save. Time taken: 0.522854s (batch_requests_ready=0.275881s, total_serialization_initiated=0.246901s, others=0.000072s) I0422 15:55:52.096480 135166776719168 composite_checkpoint_handler.py:715] [process=5][thread=MainThread] Initiated CompositeCheckpointHandler.async_save. Time taken: 0.527047s (all_items=0.000021s, per_item={'items': '0.00002122'}, temp_paths=0.527025) I0422 15:55:52.097398 135043485783808 async_checkpointer.py:79] [process=5][thread=async_save] Background save thread started. I0422 15:55:52.097569 135166776719168 async_checkpointer.py:561] Finished blocking save. Time taken: 1.291713s. Continuing background save to gs://lance-maxtext/linen_ckpt_xpk_feat_nnx_set_defaults_true_20260422_154647/linen_xpk_feat_nnx_set_defaults_true_20260422_154647_03_dropout/checkpoints/0. I0422 15:55:52.536471 135166776719168 checkpoint_manager.py:1566] [process=5][thread=MainThread][step=0] Starting CheckpointManager Save Finalize thread=save_finalize I0422 15:55:52.536883 135038637160192 async_checkpointer.py:265] [process=5][thread=save_finalize] Waiting for background save thread=async_save. I0422 15:55:52.537024 135166776719168 standard_logger.py:34] {'step': 0, 'event_type': 'save', 'directory': 'gs://lance-maxtext/linen_ckpt_xpk_feat_nnx_set_defaults_true_20260422_154647/linen_xpk_feat_nnx_set_defaults_true_20260422_154647_03_dropout/checkpoints', 'reached_preemption': False, 'preemption_received_at': None, 'synchronous': False, 'wait_for_prev_start_time': 1776873350.803945, 'wait_for_prev_duration_secs': 6.437301635742188e-05, 'time_between_consecutive_saves_sec': None, 'checkpointer_blocking_start_time': 1776873350.805823, 'checkpointer_blocking_duration_secs': 1.2918658256530762, 'get_old_steps_start_time': 1776873352.097712, 'get_old_steps_duration_secs': 2.5033950805664062e-05, 'checkpoint_manager_blocking_start_time': 1776873350.8018594, 'checkpoint_manager_blocking_duration_secs': 1.735124111175537} I0422 15:55:52.537176 135166776719168 checkpointing.py:409] Started an asynchronous checkpoint save for step 0 I0422 15:55:52.537230 135166776719168 max_utils.py:750] Memstats: After params initialized: I0422 15:55:52.537292 135166776719168 max_utils.py:756] Using (GB) 0.41 / 31.25 (1.312000%) on TPU_18(process=5,(2,4,0,0)) I0422 15:55:52.537328 135166776719168 max_utils.py:756] Using (GB) 0.41 / 31.25 (1.312000%) on TPU_19(process=5,(3,4,0,0)) I0422 15:55:52.537357 135166776719168 max_utils.py:756] Using (GB) 0.41 / 31.25 (1.312000%) on TPU_22(process=5,(2,5,0,0)) I0422 15:55:52.537387 135166776719168 max_utils.py:756] Using (GB) 0.41 / 31.25 (1.312000%) on TPU_23(process=5,(3,5,0,0)) I0422 15:55:52.805330 135166776719168 metric_logger.py:185] completed step: 0, seconds: 3.856, TFLOP/s/device: 0.208, Tokens/s/device: 33.199, total_weights: 4096, loss: 10.889 I0422 15:55:52.892678 135166776719168 metric_logger.py:185] completed step: 1, seconds: 2.319, TFLOP/s/device: 0.345, Tokens/s/device: 55.190, total_weights: 4096, loss: 10.902 I0422 15:55:53.313973 135166776719168 metric_logger.py:185] completed step: 2, seconds: 0.017, TFLOP/s/device: 46.769, Tokens/s/device: 7474.889, total_weights: 4096, loss: 9.956 I0422 15:55:53.385993 135166776719168 metric_logger.py:185] completed step: 3, seconds: 0.422, TFLOP/s/device: 1.897, Tokens/s/device: 303.165, total_weights: 4096, loss: 9.139 I0422 15:55:53.527847 135166776719168 metric_logger.py:185] completed step: 4, seconds: 0.075, TFLOP/s/device: 10.628, Tokens/s/device: 1698.581, total_weights: 4096, loss: 8.458 I0422 15:55:53.533556 135166776719168 metric_logger.py:185] completed step: 5, seconds: 0.072, TFLOP/s/device: 11.151, Tokens/s/device: 1782.258, total_weights: 4096, loss: 7.924 I0422 15:56:02.751638 135166776719168 metric_logger.py:185] completed step: 6, seconds: 0.142, TFLOP/s/device: 5.630, Tokens/s/device: 899.793, total_weights: 4096, loss: 7.553 I0422 15:56:02.822618 135166776719168 metric_logger.py:185] completed step: 7, seconds: 9.149, TFLOP/s/device: 0.088, Tokens/s/device: 13.991, total_weights: 4096, loss: 7.267 I0422 15:56:02.893340 135166776719168 metric_logger.py:185] completed step: 8, seconds: 0.075, TFLOP/s/device: 10.738, Tokens/s/device: 1716.209, total_weights: 4096, loss: 7.111 I0422 15:56:02.964833 135166776719168 checkpointing.py:794] Waiting for step 9 to finish before checkpoint... I0422 15:56:02.965427 135166776719168 checkpointing.py:798] Waited 0.0006079673767089844 seconds for step 9 to finish before starting checkpointing. I0422 15:56:02.967410 135166776719168 checkpoint_manager.py:2024] [process=5][thread=MainThread][step=0][wait_until_finished] Waiting for Save Finalize thread (save_finalize) to complete. I0422 15:56:04.418928 2645 google_auth_provider.cc:181] Running on GCE, using service account 562977990677-compute@developer.gserviceaccount.com I0422 15:56:06.579352 135043494176512 array_metadata_store.py:203] [process=5][thread=array_type_handler] Wrote 39 array_metadata.ArrayMetadata to gs://lance-maxtext/linen_ckpt_xpk_feat_nnx_set_defaults_true_20260422_154647/linen_xpk_feat_nnx_set_defaults_true_20260422_154647_03_dropout/checkpoints/0/items/array_metadatas/process_5 I0422 15:56:33.549414 135043485783808 base_pytree_checkpoint_handler.py:129] [process=5] /jax/checkpoint/write/gbytes_per_sec: 37.631 MiB/s (total gbytes: 1.5 GiB) (time elapsed: 41.97582721710205 s) (per-host) I0422 15:56:33.549535 135043485783808 async_checkpointer.py:90] [process=5][thread=async_save] 3 Handler Commit operations completed. Time taken: 41.452024s. I0422 15:56:41.945847 135043485783808 async_checkpointer.py:144] [process=5][thread=async_save] Background save thread done. Time taken: 49.848322s. I0422 15:56:41.946129 135038637160192 async_checkpointer.py:273] [process=5][thread=save_finalize] Done with waiting for background save thread=async_save. I0422 15:56:41.946252 135038637160192 async_checkpointer.py:283] [process=5][thread=save_finalize] No errors found in background save thread=async_save. I0422 15:56:41.946305 135038637160192 checkpoint_manager.py:2133] [process=5][thread=save_finalize][step=0] CheckpointManager Save Finalize is syncing with other hosts... I0422 15:56:41.948007 135038637160192 checkpoint_manager.py:2142] [process=5][thread=save_finalize][step=0] CheckpointManager Save Finalize is done on all hosts. I0422 15:56:41.948183 135166776719168 checkpoint_manager.py:2036] [process=5][thread=MainThread][step=0][wait_until_finished] Done waiting for Save Finalize thread (save_finalize) running at step=0. W0422 15:56:41.948324 135166776719168 checkpoint_manager.py:1458] Waiting for previous save to complete took 38.980919 seconds. If this number is high, consider checkpointing less frequently. I0422 15:56:41.950217 135166776719168 checkpoint_manager.py:1518] [process=5] Saving checkpoint at step 9 I0422 15:56:41.953620 135166776719168 async_checkpointer.py:452] [process=5] Started async saving checkpoint to gs://lance-maxtext/linen_ckpt_xpk_feat_nnx_set_defaults_true_20260422_154647/linen_xpk_feat_nnx_set_defaults_true_20260422_154647_03_dropout/checkpoints/9. I0422 15:56:42.954837 135166776719168 jax_array_handlers.py:358] Scheduling D2H of 39 prioritized jax.Array. I0422 15:56:42.954943 135166776719168 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 I0422 15:56:42.997473 135166776719168 base_pytree_checkpoint_handler.py:153] [process=5][thread=MainThread] Initiated "orbax.checkpoint._src.serialization.jax_array_handlers.ArrayHandler".serialize. Time taken: 0.043606s I0422 15:56:42.997632 135166776719168 base_pytree_checkpoint_handler.py:129] [process=5] /jax/checkpoint/write/blocking_gbytes_per_sec: 4.723 GiB/s (total gbytes: 1.5 GiB) (time elapsed: 0.32657861709594727 s) (per-host) I0422 15:56:42.997696 135166776719168 base_pytree_checkpoint_handler.py:737] [process=5][thread=MainThread] Initiated Pytree async_save. Time taken: 0.326652s (batch_requests_ready=0.281109s, total_serialization_initiated=0.045462s, others=0.000081s) I0422 15:56:42.997787 135166776719168 composite_checkpoint_handler.py:715] [process=5][thread=MainThread] Initiated CompositeCheckpointHandler.async_save. Time taken: 0.330836s (all_items=0.000017s, per_item={'items': '0.00001693'}, temp_paths=0.330819) I0422 15:56:42.998698 135043494176512 async_checkpointer.py:79] [process=5][thread=async_save] Background save thread started. I0422 15:56:42.998858 135166776719168 async_checkpointer.py:561] Finished blocking save. Time taken: 1.048562s. Continuing background save to gs://lance-maxtext/linen_ckpt_xpk_feat_nnx_set_defaults_true_20260422_154647/linen_xpk_feat_nnx_set_defaults_true_20260422_154647_03_dropout/checkpoints/9. I0422 15:56:43.414243 135166776719168 checkpoint_manager.py:1566] [process=5][thread=MainThread][step=9] Starting CheckpointManager Save Finalize thread=save_finalize I0422 15:56:43.414621 135038637160192 async_checkpointer.py:265] [process=5][thread=save_finalize] Waiting for background save thread=async_save. I0422 15:56:43.414817 135166776719168 standard_logger.py:34] {'step': 9, 'event_type': 'save', 'directory': 'gs://lance-maxtext/linen_ckpt_xpk_feat_nnx_set_defaults_true_20260422_154647/linen_xpk_feat_nnx_set_defaults_true_20260422_154647_03_dropout/checkpoints', 'reached_preemption': False, 'preemption_received_at': None, 'synchronous': False, 'wait_for_prev_start_time': 1776873362.9673803, 'wait_for_prev_duration_secs': 38.980918645858765, 'time_between_consecutive_saves_sec': None, 'checkpointer_blocking_start_time': 1776873401.9502628, 'checkpointer_blocking_duration_secs': 1.0487041473388672, 'get_old_steps_start_time': 1776873402.9989877, 'get_old_steps_duration_secs': 2.5510787963867188e-05, 'checkpoint_manager_blocking_start_time': 1776873362.9656668, 'checkpoint_manager_blocking_duration_secs': 40.449110984802246} I0422 15:56:43.414925 135166776719168 checkpointing.py:409] Started an asynchronous checkpoint save for step 9 I0422 15:56:43.414980 135166776719168 checkpoint_manager.py:2024] [process=5][thread=MainThread][step=9][wait_until_finished] Waiting for Save Finalize thread (save_finalize) to complete. I0422 15:56:49.607990 135014456928000 array_metadata_store.py:203] [process=5][thread=array_type_handler] Wrote 39 array_metadata.ArrayMetadata to gs://lance-maxtext/linen_ckpt_xpk_feat_nnx_set_defaults_true_20260422_154647/linen_xpk_feat_nnx_set_defaults_true_20260422_154647_03_dropout/checkpoints/9/items/array_metadatas/process_5 I0422 15:57:24.440784 135043494176512 base_pytree_checkpoint_handler.py:129] [process=5] /jax/checkpoint/write/gbytes_per_sec: 37.817 MiB/s (total gbytes: 1.5 GiB) (time elapsed: 41.76968741416931 s) (per-host) I0422 15:57:24.440908 135043494176512 async_checkpointer.py:90] [process=5][thread=async_save] 3 Handler Commit operations completed. Time taken: 41.442098s. I0422 15:57:33.445502 135043494176512 async_checkpointer.py:144] [process=5][thread=async_save] Background save thread done. Time taken: 50.446676s. I0422 15:57:33.445833 135038637160192 async_checkpointer.py:273] [process=5][thread=save_finalize] Done with waiting for background save thread=async_save. I0422 15:57:33.445957 135038637160192 async_checkpointer.py:283] [process=5][thread=save_finalize] No errors found in background save thread=async_save. I0422 15:57:33.446007 135038637160192 checkpoint_manager.py:2133] [process=5][thread=save_finalize][step=9] CheckpointManager Save Finalize is syncing with other hosts... I0422 15:57:33.447718 135038637160192 checkpoint_manager.py:2142] [process=5][thread=save_finalize][step=9] CheckpointManager Save Finalize is done on all hosts. I0422 15:57:33.447893 135166776719168 checkpoint_manager.py:2036] [process=5][thread=MainThread][step=9][wait_until_finished] Done waiting for Save Finalize thread (save_finalize) running at step=9. I0422 15:57:33.448039 135166776719168 checkpoint_manager.py:2013] [process=5][thread=MainThread][wait_until_finished] No Save Finalize thread to wait for. Returning. I0422 15:57:33.448660 135166776719168 metric_logger.py:185] completed step: 9, seconds: 0.071, TFLOP/s/device: 11.278, Tokens/s/device: 1802.512, total_weights: 4096, loss: 6.981 Per train step: Total TFLOPs: 0.80 split as 99.60% learnable weight flops and 0.40% attention flops XPK End: Wed Apr 22 15:57:43 UTC 2026 EXIT_CODE=0