MaxView

Timing Test Reference Updates
‹ 09_pdb_lt_1Case: 10_shardy_false11_optimizer_offload_true ›

Metrics: main (b117f50cf) vs feat/nnx-set-defaults-true (73213e044)

Metricmain  b117f50cffeat/nnx-set-defaults-true  73213e044Diff (feat/nnx-set-defaults-true − main)
Parameters1.104 billion1.104 billion—
Final loss8.17908.17900
TFLOP/s92.39792.130-0.267
Tok/s13927.013886.8-40.229
Avg s/step2.7372.481-0.256
Memory %1.381.380
JAX0.9.20.8.1—

Diff = branch value − main value. Green = branch improved. Red = branch regressed.

main  ·  b117f50cf  ·  main_20260424_070227  ·  full log 
XPK Start: Fri Apr 24 07:38:47 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. 
2026-04-24 07:39:12.538972: 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)
I0424 07:39:12.750144 138411120113472 max_utils.py:273] Attempting to initialize the jax distributed system...
I0424 07:39:21.790947 138411120113472 distributed.py:149] Starting JAX distributed service on [::]:8482
I0424 07:39:21.793248 138411120113472 distributed.py:172] Connecting to JAX distributed service on mt-10-shardy-false-1ubfd-slice-job-0-0.mt-10-shardy-false-1ubfd:8482
I0424 07:39:23.169116 138411120113472 max_utils.py:284] Jax distributed system initialized!
I0424 07:39:29.398111 138411120113472 max_utils.py:800] System Information: Jax Version: 0.9.2
I0424 07:39:29.398216 138411120113472 max_utils.py:801] System Information: Jaxlib Version: 0.9.2
I0424 07:39:29.398257 138411120113472 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
I0424 07:39:29.398293 138411120113472 train_utils.py:361] WARNING: Sequence packing is essentially ignored for synthetic data. Please use a real dataset to use sequence packing.
I0424 07:39:30.094225 138411120113472 maxtext_utils.py:1604] Num_devices: 32, shape (1, 1, 1, 32, 1, 1, 1, 1, 1, 1, 1, 1, 1)
I0424 07:39:30.094512 138411120113472 checkpointing.py:677] Setting up checkpoint logger...
I0424 07:39:30.094568 138411120113472 checkpointing.py:233] Creating checkpoint manager with ocdbt=True and zarr3=True
I0424 07:39:30.094610 138411120113472 pytree_checkpoint_handler.py:592] save_device_host_concurrent_bytes=None
I0424 07:39:30.094970 138411120113472 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 0x7de1c2b1c080>, enable_pinned_host_transfer=False, save_concurrent_bytes: 96000000000 (89.4 GiB), restore_concurrent_bytes: 96000000000 (89.4 GiB)
I0424 07:39:33.022744 138411120113472 checkpointing.py:265] Enabling policy for fixed interval checkpointing.
I0424 07:39:33.022990 138411120113472 checkpoint_manager.py:708] [process=5][thread=MainThread] CheckpointManager init: checkpointers=None, item_names=('items',), item_handlers={'items': <orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeCheckpointHandler object at 0x7dcd7c1f9fd0>}, handler_registry=None
I0424 07:39:33.023224 138411120113472 composite_checkpoint_handler.py:237] Deferred registration for item: "items". Adding handler `<orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeCheckpointHandler object at 0x7dcd7c1f9fd0>` 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`.
I0424 07:39:33.023274 138411120113472 composite_checkpoint_handler.py:237] Deferred registration for item: "metrics". Adding handler `<orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonCheckpointHandler object at 0x7dcd7c3f5730>` 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`.
I0424 07:39:33.023310 138411120113472 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 0x7dcd7c1f9fd0>, ('items', <class 'orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeRestoreArgs'>): <orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeCheckpointHandler object at 0x7dcd7c1f9fd0>, ('metrics', <class 'orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonSaveArgs'>): <orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonCheckpointHandler object at 0x7dcd7c3f5730>, ('metrics', <class 'orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonRestoreArgs'>): <orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonCheckpointHandler object at 0x7dcd7c3f5730>}).
I0424 07:39:33.023627 138411120113472 abstract_checkpointer.py:35] orbax-checkpoint version: 0.11.34
I0424 07:39:33.023710 138411120113472 async_checkpointer.py:192] [process=5][thread=MainThread] Using barrier_sync_fn: <function get_barrier_sync_fn.<locals>._fn at 0x7dcd643a1a80> timeout: 1200 secs and primary_host=0 for async checkpoint writes
I0424 07:39:34.078434 138411120113472 checkpoint_manager.py:1812] Found 0 checkpoint steps in gs://lance-maxtext/linen_ckpt_xpk_main_20260424_070227/linen_xpk_main_20260424_070227_10_shardy_false/checkpoints
I0424 07:39:34.114173 138411120113472 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_main_20260424_070227/linen_xpk_main_20260424_070227_10_shardy_false/checkpoints: <orbax.checkpoint.checkpoint_manager.CheckpointManager object at 0x7dcd7c1fbda0>
I0424 07:39:34.114344 138411120113472 checkpointing.py:301] Checkpoint manager created!
I0424 07:39:35.053183 138411120113472 nnx_wrappers.py:437] Unknown Logical: bfloat16[32,2048,2048]...................................... ('activation_batch', 'activation_norm_length', 'activation_embed').
I0424 07:39:35.053304 138411120113472 nnx_wrappers.py:437] Unknown Physical: bfloat16[32,2048,2048]...................................... ('fsdp', None, None).
I0424 07:39:35.435963 138411120113472 attentions.py:1088] attentions/inputs_q Logical: bfloat16[32,2048,2048]...................................... ('activation_batch', 'activation_attn_length', 'activation_attn_embed').
I0424 07:39:35.436056 138411120113472 attentions.py:1088] attentions/inputs_q Physical: bfloat16[32,2048,2048]...................................... ('fsdp', None, None).
I0424 07:39:35.452497 138411120113472 attentions.py:1089] attentions/inputs_kv Logical: bfloat16[32,2048,2048]...................................... ('activation_batch', 'activation_attn_length', 'activation_attn_embed').
I0424 07:39:35.452556 138411120113472 attentions.py:1089] attentions/inputs_kv Physical: bfloat16[32,2048,2048]...................................... ('fsdp', None, None).
I0424 07:39:35.476729 138411120113472 attentions.py:1154] attentions/query Logical: bfloat16[32,2048,16,128].................................... ('activation_kv_batch', 'activation_attn_length', 'activation_kv_heads', 'activation_kv_head_dim').
I0424 07:39:35.476795 138411120113472 attentions.py:1154] attentions/query Physical: bfloat16[32,2048,16,128].................................... ('fsdp', None, None, None).
I0424 07:39:35.493172 138411120113472 attentions.py:1155] attentions/key Logical: bfloat16[32,2048,16,128].................................... ('activation_kv_batch', 'activation_attn_length', 'activation_kv_heads', 'activation_kv_head_dim').
I0424 07:39:35.493236 138411120113472 attentions.py:1155] attentions/key Physical: bfloat16[32,2048,16,128].................................... ('fsdp', None, None, None).
I0424 07:39:35.509510 138411120113472 attentions.py:1156] attentions/value Logical: bfloat16[32,2048,16,128].................................... ('activation_kv_batch', 'activation_attn_length', 'activation_kv_heads', 'activation_kv_head_dim').
I0424 07:39:35.509571 138411120113472 attentions.py:1156] attentions/value Physical: bfloat16[32,2048,16,128].................................... ('fsdp', None, None, None).
I0424 07:39:35.534181 138411120113472 attentions.py:1198] attentions/out Logical: bfloat16[32,2048,16,128].................................... ('activation_batch', 'activation_attn_length', 'activation_heads', 'activation_kv').
I0424 07:39:35.534256 138411120113472 attentions.py:1198] attentions/out Physical: bfloat16[32,2048,16,128].................................... ('fsdp', None, None, None).
I0424 07:39:35.555304 138411120113472 linears.py:525] linears/x Logical: bfloat16[32,2048,7168]...................................... ('activation_batch', 'activation_length', 'activation_mlp').
I0424 07:39:35.555366 138411120113472 linears.py:525] linears/x Physical: bfloat16[32,2048,7168]...................................... ('fsdp', None, None).
I0424 07:39:35.770153 138411120113472 checkpointing.py:577] checkpoint manager exists so trying to load this run's existing checkpoint
I0424 07:39:35.770264 138411120113472 checkpointing.py:665] No existing checkpoints found, not restoring checkpoint.
fsdp: 32
I0424 07:39:37.181834 138411120113472 maxtext_utils.py:1707]  params/params/decoder/decoder_norm/scale
    Shape:     float32[2048]
    Logical:   P('norm',)
    Physical:  (None,)
I0424 07:39:37.181959 138411120113472 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)
I0424 07:39:37.182016 138411120113472 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)
I0424 07:39:37.182075 138411120113472 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')
I0424 07:39:37.182127 138411120113472 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)
I0424 07:39:37.182165 138411120113472 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)
I0424 07:39:37.182215 138411120113472 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)
I0424 07:39:37.182265 138411120113472 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')
I0424 07:39:37.182303 138411120113472 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)
I0424 07:39:37.182338 138411120113472 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)
I0424 07:39:37.182383 138411120113472 maxtext_utils.py:1707]  params/params/decoder/logits_dense/kernel
    Shape:     float32[2048,32000]
    Logical:   P('embed_vocab', 'vocab')
    Physical:  ('fsdp', None)

I0424 07:39:37.182430 138411120113472 maxtext_utils.py:1707]  params/params/token_embedder/embedding
    Shape:     float32[32000,2048]
    Logical:   P('vocab', 'embed_vocab')
    Physical:  (None, 'fsdp')
I0424 07:39:37.670079 138411120113472 train.py:155] train/xent Logical: float32[32,2048]............................................ ('activation_embed_and_logits_batch', 'activation_length').
I0424 07:39:37.670168 138411120113472 train.py:155] train/xent Physical: float32[32,2048]............................................ ('fsdp', None).
I0424 07:39:37.686027 138411120113472 train.py:162] train/z_loss Logical: float32[32,2048]............................................ ('activation_embed_and_logits_batch', 'activation_length').
I0424 07:39:37.686088 138411120113472 train.py:162] train/z_loss Physical: float32[32,2048]............................................ ('fsdp', None).
I0424 07:39:48.608884 138411120113472 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.
I0424 07:39:48.609716 138411120113472 metric_logger.py:301] number parameters: 1.104 billion
I0424 07:40:00.097942 138411120113472 checkpointing.py:772] Waiting for step 0 to finish before checkpoint...
I0424 07:40:00.277922 138411120113472 checkpointing.py:776] Waited 0.17995238304138184 seconds for step 0 to finish before starting checkpointing.
I0424 07:40:00.280550 138411120113472 checkpoint_manager.py:2009] [process=5][thread=MainThread][wait_until_finished] No Save Finalize thread to wait for. Returning.
I0424 07:40:00.282358 138411120113472 checkpoint_manager.py:1512] [process=5] Saving checkpoint at step 0
I0424 07:40:00.283863 138411120113472 event_tracking.py:70] [process=5] [async] Started save checkpoint @ gs://lance-maxtext/linen_ckpt_xpk_main_20260424_070227/linen_xpk_main_20260424_070227_10_shardy_false/checkpoints/0.
I0424 07:40:00.598110 138411120113472 signaling_client.py:364] Using JaxDistributedSignalingClient
I0424 07:40:00.599143 138411120113472 jax_array_handlers.py:360] Scheduling D2H of 39 prioritized jax.Array.
I0424 07:40:00.599204 138411120113472 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
I0424 07:40:00.870079 138411120113472 base_pytree_checkpoint_handler.py:154] [process=5][thread=MainThread] Initiated "orbax.checkpoint._src.serialization.jax_array_handlers.ArrayHandler".serialize. Time taken: 0.272042s
I0424 07:40:00.870256 138411120113472 base_pytree_checkpoint_handler.py:130] [process=5] /jax/orbax/write/blocking_gbytes_per_sec: 5.556 GiB/s (total gbytes: 1.5 GiB) (time elapsed: 0.27762436866760254 s) (per-host)
I0424 07:40:00.870311 138411120113472 base_pytree_checkpoint_handler.py:768] [process=5][thread=MainThread] Initiated Pytree async_save. Time taken: 0.277694s (batch_requests_ready=0.002331s, total_serialization_initiated=0.275285s, others=0.000079s)
I0424 07:40:00.870405 138411120113472 composite_checkpoint_handler.py:715] [process=5][thread=MainThread] Initiated CompositeCheckpointHandler.async_save. Time taken: 0.282004s (all_items=0.000018s, per_item={'items': '0.00001764'}, temp_paths=0.281986)
I0424 07:40:00.871283 138411120113472 event_tracking.py:125] [process=5] [async] Finished blocking save in 0.59 seconds. Continuing save @ gs://lance-maxtext/linen_ckpt_xpk_main_20260424_070227/linen_xpk_main_20260424_070227_10_shardy_false/checkpoints/0.
I0424 07:40:00.871614 138282444781312 async_checkpointer.py:76] [process=5][thread=async_save] Background save thread started. Deadline for this save operation is 2026-04-24 08:00:00.871577
I0424 07:40:01.317759 138411120113472 checkpoint_manager.py:1560] [process=5][thread=MainThread][step=0] Starting CheckpointManager Save Finalize thread=save_finalize
I0424 07:40:01.318146 138279717648128 async_checkpointer.py:280] [process=5][thread=save_finalize] Waiting for background save thread=async_save.
I0424 07:40:01.318313 138411120113472 standard_logger.py:34] {'step': 0, 'event_type': 'save', 'directory': 'gs://lance-maxtext/linen_ckpt_xpk_main_20260424_070227/linen_xpk_main_20260424_070227_10_shardy_false/checkpoints', 'reached_preemption': False, 'preemption_received_at': None, 'synchronous': False, 'wait_for_prev_start_time': 1777016400.2805312, 'wait_for_prev_duration_secs': 6.175041198730469e-05, 'time_between_consecutive_saves_sec': None, 'checkpointer_blocking_start_time': 1777016400.2823956, 'checkpointer_blocking_duration_secs': 0.5893909931182861, 'get_old_steps_start_time': 1777016400.871813, 'get_old_steps_duration_secs': 2.956390380859375e-05, 'checkpoint_manager_blocking_start_time': 1777016400.2784798, 'checkpoint_manager_blocking_duration_secs': 1.0397922992706299}
I0424 07:40:01.318427 138411120113472 checkpointing.py:408] Started an asynchronous checkpoint save for step 0
I0424 07:40:01.318478 138411120113472 max_utils.py:750] 
Memstats: After params initialized:
I0424 07:40:01.318532 138411120113472 max_utils.py:756] 	Using (GB) 0.43 / 31.25 (1.376000%) on TPU_18(process=5,(2,4,0,0))
I0424 07:40:01.318564 138411120113472 max_utils.py:756] 	Using (GB) 0.43 / 31.25 (1.376000%) on TPU_19(process=5,(3,4,0,0))
I0424 07:40:01.318591 138411120113472 max_utils.py:756] 	Using (GB) 0.43 / 31.25 (1.376000%) on TPU_22(process=5,(2,5,0,0))
I0424 07:40:01.318616 138411120113472 max_utils.py:756] 	Using (GB) 0.43 / 31.25 (1.376000%) on TPU_23(process=5,(3,5,0,0))
I0424 07:40:01.633461 138411120113472 metric_logger.py:196] completed step: 0, seconds: 11.488, TFLOP/s/device: 1.183, Tokens/s/device: 178.271, total_weights: 65536, loss: 10.874, lm_loss: 10.874, perplexity: 52779.875
I0424 07:40:01.818990 138411120113472 metric_logger.py:196] completed step: 1, seconds: 1.534, TFLOP/s/device: 8.859, Tokens/s/device: 1335.395, total_weights: 65536, loss: 10.874, lm_loss: 10.874, perplexity: 52779.875
I0424 07:40:02.275708 138411120113472 metric_logger.py:196] completed step: 2, seconds: 0.039, TFLOP/s/device: 352.080, Tokens/s/device: 53069.369, total_weights: 65536, loss: 10.262, lm_loss: 10.262, perplexity: 28637.129
I0424 07:40:02.422853 138411120113472 metric_logger.py:196] completed step: 3, seconds: 0.457, TFLOP/s/device: 29.740, Tokens/s/device: 4482.685, total_weights: 65536, loss: 9.734, lm_loss: 9.734, perplexity: 16889.203
I0424 07:40:02.717979 138411120113472 metric_logger.py:196] completed step: 4, seconds: 0.153, TFLOP/s/device: 88.550, Tokens/s/device: 13347.237, total_weights: 65536, loss: 9.277, lm_loss: 9.277, perplexity: 10694.614
I0424 07:40:02.723958 138411120113472 metric_logger.py:196] completed step: 5, seconds: 0.147, TFLOP/s/device: 92.525, Tokens/s/device: 13946.394, total_weights: 65536, loss: 8.892, lm_loss: 8.892, perplexity: 7270.717
I0424 07:40:03.923209    2828 google_auth_provider.cc:181] Running on GCE, using service account 562977990677-compute@developer.gserviceaccount.com
I0424 07:40:06.385799 138279760426752 array_metadata_store.py:203] [process=5][thread=array_type_handler] Wrote 39 array_metadata.ArrayMetadata to gs://lance-maxtext/linen_ckpt_xpk_main_20260424_070227/linen_xpk_main_20260424_070227_10_shardy_false/checkpoints/0/items/array_metadatas/process_5
I0424 07:40:24.574975 138411120113472 metric_logger.py:196] completed step: 6, seconds: 0.296, TFLOP/s/device: 45.964, Tokens/s/device: 6928.235, total_weights: 65536, loss: 8.592, lm_loss: 8.592, perplexity: 5390.759
I0424 07:40:24.722224 138411120113472 metric_logger.py:196] completed step: 7, seconds: 21.704, TFLOP/s/device: 0.626, Tokens/s/device: 94.359, total_weights: 65536, loss: 8.384, lm_loss: 8.384, perplexity: 4376.905
I0424 07:40:24.869506 138411120113472 metric_logger.py:196] completed step: 8, seconds: 0.152, TFLOP/s/device: 89.189, Tokens/s/device: 13443.525, total_weights: 65536, loss: 8.255, lm_loss: 8.255, perplexity: 3846.406
I0424 07:40:25.016248 138411120113472 checkpointing.py:772] Waiting for step 9 to finish before checkpoint...
I0424 07:40:25.016910 138411120113472 checkpointing.py:776] Waited 0.0006840229034423828 seconds for step 9 to finish before starting checkpointing.
I0424 07:40:25.019034 138411120113472 checkpoint_manager.py:2020] [process=5][thread=MainThread][step=0][wait_until_finished] Waiting for Save Finalize thread (save_finalize) to complete.
I0424 07:40:37.508049 138282444781312 base_pytree_checkpoint_handler.py:130] [process=5] /jax/orbax/write/gbytes_per_sec: 42.790 MiB/s (total gbytes: 1.5 GiB) (time elapsed: 36.91538429260254 s) (per-host)
I0424 07:40:37.508175 138282444781312 async_checkpointer.py:90] [process=5][thread=async_save] 3 Handler Commit operations completed. Time taken: 36.636420s.
I0424 07:40:47.441344 138282444781312 async_checkpointer.py:160] [process=5][thread=async_save] Background save thread done. Time taken: 46.569580s.
I0424 07:40:47.441637 138279717648128 async_checkpointer.py:288] [process=5][thread=save_finalize] Done with waiting for background save thread=async_save.
I0424 07:40:47.441761 138279717648128 async_checkpointer.py:298] [process=5][thread=save_finalize] No errors found in background save thread=async_save.
I0424 07:40:47.441811 138279717648128 checkpoint_manager.py:2137] [process=5][thread=save_finalize][step=0] CheckpointManager Save Finalize is syncing with other hosts...
I0424 07:40:47.444147 138279717648128 checkpoint_manager.py:2146] [process=5][thread=save_finalize][step=0] CheckpointManager Save Finalize is done on all hosts.
I0424 07:40:47.444291 138411120113472 checkpoint_manager.py:2032] [process=5][thread=MainThread][step=0][wait_until_finished] Done waiting for Save Finalize thread (save_finalize) running at step=0.
W0424 07:40:47.444378 138411120113472 checkpoint_manager.py:1452] Waiting for previous save to complete took 22.425347 seconds. If this number is high, consider checkpointing less frequently.
I0424 07:40:47.446234 138411120113472 checkpoint_manager.py:1512] [process=5] Saving checkpoint at step 9
I0424 07:40:47.448367 138411120113472 event_tracking.py:70] [process=5] [async] Started save checkpoint @ gs://lance-maxtext/linen_ckpt_xpk_main_20260424_070227/linen_xpk_main_20260424_070227_10_shardy_false/checkpoints/9.
I0424 07:40:47.758399 138411120113472 jax_array_handlers.py:360] Scheduling D2H of 39 prioritized jax.Array.
I0424 07:40:47.758489 138411120113472 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
I0424 07:40:47.792888 138411120113472 base_pytree_checkpoint_handler.py:154] [process=5][thread=MainThread] Initiated "orbax.checkpoint._src.serialization.jax_array_handlers.ArrayHandler".serialize. Time taken: 0.035667s
I0424 07:40:47.793013 138411120113472 base_pytree_checkpoint_handler.py:130] [process=5] /jax/orbax/write/blocking_gbytes_per_sec: 39.444 GiB/s (total gbytes: 1.5 GiB) (time elapsed: 0.0391082763671875 s) (per-host)
I0424 07:40:47.793061 138411120113472 base_pytree_checkpoint_handler.py:768] [process=5][thread=MainThread] Initiated Pytree async_save. Time taken: 0.039164s (batch_requests_ready=0.001811s, total_serialization_initiated=0.037291s, others=0.000062s)
I0424 07:40:47.793139 138411120113472 composite_checkpoint_handler.py:715] [process=5][thread=MainThread] Initiated CompositeCheckpointHandler.async_save. Time taken: 0.043432s (all_items=0.000017s, per_item={'items': '0.00001693'}, temp_paths=0.043416)
I0424 07:40:47.793831 138411120113472 event_tracking.py:125] [process=5] [async] Finished blocking save in 0.35 seconds. Continuing save @ gs://lance-maxtext/linen_ckpt_xpk_main_20260424_070227/linen_xpk_main_20260424_070227_10_shardy_false/checkpoints/9.
I0424 07:40:47.794155 138279717648128 async_checkpointer.py:76] [process=5][thread=async_save] Background save thread started. Deadline for this save operation is 2026-04-24 08:00:47.794118
I0424 07:40:47.796082 138411120113472 checkpoint_manager.py:1560] [process=5][thread=MainThread][step=9] Starting CheckpointManager Save Finalize thread=save_finalize
I0424 07:40:47.796315 138262421161728 async_checkpointer.py:280] [process=5][thread=save_finalize] Waiting for background save thread=async_save.
I0424 07:40:47.796447 138411120113472 standard_logger.py:34] {'step': 9, 'event_type': 'save', 'directory': 'gs://lance-maxtext/linen_ckpt_xpk_main_20260424_070227/linen_xpk_main_20260424_070227_10_shardy_false/checkpoints', 'reached_preemption': False, 'preemption_received_at': None, 'synchronous': False, 'wait_for_prev_start_time': 1777016425.0190036, 'wait_for_prev_duration_secs': 22.425346612930298, 'time_between_consecutive_saves_sec': None, 'checkpointer_blocking_start_time': 1777016447.446272, 'checkpointer_blocking_duration_secs': 0.34803295135498047, 'get_old_steps_start_time': 1777016447.7943318, 'get_old_steps_duration_secs': 2.9802322387695312e-05, 'checkpoint_manager_blocking_start_time': 1777016425.0171356, 'checkpoint_manager_blocking_duration_secs': 22.77927803993225}
I0424 07:40:47.796617 138411120113472 checkpointing.py:408] Started an asynchronous checkpoint save for step 9
I0424 07:40:47.796673 138411120113472 checkpoint_manager.py:2020] [process=5][thread=MainThread][step=9][wait_until_finished] Waiting for Save Finalize thread (save_finalize) to complete.
I0424 07:40:53.026312 138262446339840 array_metadata_store.py:203] [process=5][thread=array_type_handler] Wrote 39 array_metadata.ArrayMetadata to gs://lance-maxtext/linen_ckpt_xpk_main_20260424_070227/linen_xpk_main_20260424_070227_10_shardy_false/checkpoints/9/items/array_metadatas/process_5
I0424 07:41:29.265553 138279717648128 base_pytree_checkpoint_handler.py:130] [process=5] /jax/orbax/write/gbytes_per_sec: 38.052 MiB/s (total gbytes: 1.5 GiB) (time elapsed: 41.51160669326782 s) (per-host)
I0424 07:41:29.265703 138279717648128 async_checkpointer.py:90] [process=5][thread=async_save] 3 Handler Commit operations completed. Time taken: 41.471432s.
I0424 07:41:38.897273 138279717648128 async_checkpointer.py:160] [process=5][thread=async_save] Background save thread done. Time taken: 51.102986s.
I0424 07:41:38.897547 138262421161728 async_checkpointer.py:288] [process=5][thread=save_finalize] Done with waiting for background save thread=async_save.
I0424 07:41:38.897699 138262421161728 async_checkpointer.py:298] [process=5][thread=save_finalize] No errors found in background save thread=async_save.
I0424 07:41:38.897759 138262421161728 checkpoint_manager.py:2137] [process=5][thread=save_finalize][step=9] CheckpointManager Save Finalize is syncing with other hosts...
I0424 07:41:38.900094 138262421161728 checkpoint_manager.py:2146] [process=5][thread=save_finalize][step=9] CheckpointManager Save Finalize is done on all hosts.
I0424 07:41:38.900279 138411120113472 checkpoint_manager.py:2032] [process=5][thread=MainThread][step=9][wait_until_finished] Done waiting for Save Finalize thread (save_finalize) running at step=9.
I0424 07:41:38.900425 138411120113472 checkpoint_manager.py:2009] [process=5][thread=MainThread][wait_until_finished] No Save Finalize thread to wait for. Returning.
I0424 07:41:38.901467 138411120113472 metric_logger.py:196] completed step: 9, seconds: 0.147, TFLOP/s/device: 92.397, Tokens/s/device: 13927.046, 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: Fri Apr 24 07:41:47 UTC 2026
EXIT_CODE=0
feat/nnx-set-defaults-true  ·  73213e044  ·  feat_nnx_set_defaults_true_20260424_145744  ·  full log 
XPK Start: Fri Apr 24 15:31:23 UTC 2026
2026-04-24 15:31:27.448662: 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 00:00:1777044687.461780      10 cuda_dnn.cc:8579] Unable to register cuDNN factory: Attempting to register factory for plugin cuDNN when one has already been registered
E0000 00:00:1777044687.465566      10 cuda_blas.cc:1407] Unable to register cuBLAS factory: Attempting to register factory for plugin cuBLAS when one has already been registered
W0000 00:00:1777044687.477495      10 computation_placer.cc:177] computation placer already registered. Please check linkage and avoid linking the same target more than once.
W0000 00:00:1777044687.477515      10 computation_placer.cc:177] computation placer already registered. Please check linkage and avoid linking the same target more than once.
W0000 00:00:1777044687.477517      10 computation_placer.cc:177] computation placer already registered. Please check linkage and avoid linking the same target more than once.
W0000 00:00:1777044687.477519      10 computation_placer.cc:177] computation placer already registered. Please check linkage and avoid linking the same target more than once.
2026-04-24 15:31:46.582963: 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)
I0424 15:31:47.108931 139424412055360 max_utils.py:273] Attempting to initialize the jax distributed system...
INFO:2026-04-24 15:31:56,150:jax._src.distributed:140: Starting JAX distributed service on [::]:8482
I0424 15:31:56.150407 139424412055360 distributed.py:140] Starting JAX distributed service on [::]:8482
INFO:2026-04-24 15:31:56,152:jax._src.distributed:157: Connecting to JAX distributed service on mt-10-shardy-false-adznj-slice-job-0-0.mt-10-shardy-false-adznj:8482
I0424 15:31:56.152771 139424412055360 distributed.py:157] Connecting to JAX distributed service on mt-10-shardy-false-adznj-slice-job-0-0.mt-10-shardy-false-adznj:8482
I0424 15:31:57.416534 139424412055360 max_utils.py:284] Jax distributed system initialized!
I0424 15:32:04.852493 139424412055360 max_utils.py:800] System Information: Jax Version: 0.8.1
I0424 15:32:04.852604 139424412055360 max_utils.py:801] System Information: Jaxlib Version: 0.8.1
I0424 15:32:04.852648 139424412055360 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
I0424 15:32:04.852682 139424412055360 train_utils.py:377] WARNING: Sequence packing is essentially ignored for synthetic data. Please use a real dataset to use sequence packing.
I0424 15:32:05.479961 139424412055360 maxtext_utils.py:1631] Num_devices: 32, shape (1, 1, 1, 32, 1, 1, 1, 1, 1, 1, 1, 1, 1)
I0424 15:32:05.480568 139424412055360 maxtext_utils.py:1631] Num_devices: 32, shape (1, 1, 1, 32, 1, 1, 1, 1, 1, 1, 1, 1, 1)
I0424 15:32:05.480755 139424412055360 checkpointing.py:688] Setting up checkpoint logger...
I0424 15:32:05.480806 139424412055360 checkpointing.py:234] Creating checkpoint manager with ocdbt=True and zarr3=True
I0424 15:32:05.480849 139424412055360 pytree_checkpoint_handler.py:589] save_device_host_concurrent_bytes=None
I0424 15:32:05.481217 139424412055360 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 0x7ecdc7379190>, enable_pinned_host_transfer=False, save_concurrent_bytes: 96000000000 (89.4 GiB), restore_concurrent_bytes: 96000000000 (89.4 GiB)
I0424 15:32:08.398834 139424412055360 checkpointing.py:266] Enabling policy for fixed interval checkpointing.
I0424 15:32:08.399200 139424412055360 checkpoint_manager.py:709] [process=6][thread=MainThread] CheckpointManager init: checkpointers=None, item_names=('items',), item_handlers={'items': <orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeCheckpointHandler object at 0x7eb924731880>}, handler_registry=None
I0424 15:32:08.399436 139424412055360 composite_checkpoint_handler.py:237] Deferred registration for item: "items". Adding handler `<orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeCheckpointHandler object at 0x7eb924731880>` 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`.
I0424 15:32:08.399486 139424412055360 composite_checkpoint_handler.py:237] Deferred registration for item: "metrics". Adding handler `<orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonCheckpointHandler object at 0x7eb924736750>` 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`.
I0424 15:32:08.399521 139424412055360 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 0x7eb924731880>, ('items', <class 'orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeRestoreArgs'>): <orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeCheckpointHandler object at 0x7eb924731880>, ('metrics', <class 'orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonSaveArgs'>): <orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonCheckpointHandler object at 0x7eb924736750>, ('metrics', <class 'orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonRestoreArgs'>): <orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonCheckpointHandler object at 0x7eb924736750>}).
I0424 15:32:08.400022 139424412055360 abstract_checkpointer.py:35] orbax-checkpoint version: 0.11.33
I0424 15:32:08.400113 139424412055360 async_checkpointer.py:177] [process=6][thread=MainThread] Using barrier_sync_fn: <function get_barrier_sync_fn.<locals>._fn at 0x7eb92410c180> timeout: 600 secs and primary_host=0 for async checkpoint writes
I0424 15:32:09.416318 139424412055360 checkpoint_manager.py:1818] Found 0 checkpoint steps in gs://lance-maxtext/linen_ckpt_xpk_feat_nnx_set_defaults_true_20260424_145744/linen_xpk_feat_nnx_set_defaults_true_20260424_145744_10_shardy_false/checkpoints
I0424 15:32:09.428076 139424412055360 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_feat_nnx_set_defaults_true_20260424_145744/linen_xpk_feat_nnx_set_defaults_true_20260424_145744_10_shardy_false/checkpoints: <orbax.checkpoint.checkpoint_manager.CheckpointManager object at 0x7eb94c124e30>
I0424 15:32:09.428217 139424412055360 checkpointing.py:302] Checkpoint manager created!
I0424 15:32:10.324448 139424412055360 nnx_wrappers.py:455] Unknown Logical: bfloat16[32,2048,2048]...................................... ('activation_batch', 'activation_norm_length', 'activation_embed').
I0424 15:32:10.324557 139424412055360 nnx_wrappers.py:455] Unknown Physical: bfloat16[32,2048,2048]...................................... ('fsdp', None, None).
I0424 15:32:10.695389 139424412055360 attentions.py:1084] attentions/inputs_q Logical: bfloat16[32,2048,2048]...................................... ('activation_batch', 'activation_attn_length', 'activation_attn_embed').
I0424 15:32:10.695485 139424412055360 attentions.py:1084] attentions/inputs_q Physical: bfloat16[32,2048,2048]...................................... ('fsdp', None, None).
I0424 15:32:10.711471 139424412055360 attentions.py:1085] attentions/inputs_kv Logical: bfloat16[32,2048,2048]...................................... ('activation_batch', 'activation_attn_length', 'activation_attn_embed').
I0424 15:32:10.711532 139424412055360 attentions.py:1085] attentions/inputs_kv Physical: bfloat16[32,2048,2048]...................................... ('fsdp', None, None).
I0424 15:32:10.734474 139424412055360 attentions.py:1150] attentions/query Logical: bfloat16[32,2048,16,128].................................... ('activation_kv_batch', 'activation_attn_length', 'activation_kv_heads', 'activation_kv_head_dim').
I0424 15:32:10.734542 139424412055360 attentions.py:1150] attentions/query Physical: bfloat16[32,2048,16,128].................................... ('fsdp', None, None, None).
I0424 15:32:10.750782 139424412055360 attentions.py:1151] attentions/key Logical: bfloat16[32,2048,16,128].................................... ('activation_kv_batch', 'activation_attn_length', 'activation_kv_heads', 'activation_kv_head_dim').
I0424 15:32:10.750848 139424412055360 attentions.py:1151] attentions/key Physical: bfloat16[32,2048,16,128].................................... ('fsdp', None, None, None).
I0424 15:32:10.766819 139424412055360 attentions.py:1152] attentions/value Logical: bfloat16[32,2048,16,128].................................... ('activation_kv_batch', 'activation_attn_length', 'activation_kv_heads', 'activation_kv_head_dim').
I0424 15:32:10.766879 139424412055360 attentions.py:1152] attentions/value Physical: bfloat16[32,2048,16,128].................................... ('fsdp', None, None, None).
I0424 15:32:10.791764 139424412055360 attentions.py:1193] attentions/out Logical: bfloat16[32,2048,16,128].................................... ('activation_batch', 'activation_attn_length', 'activation_heads', 'activation_kv').
I0424 15:32:10.791836 139424412055360 attentions.py:1193] attentions/out Physical: bfloat16[32,2048,16,128].................................... ('fsdp', None, None, None).
I0424 15:32:10.811613 139424412055360 linears.py:541] linears/x Logical: bfloat16[32,2048,7168]...................................... ('activation_batch', 'activation_length', 'activation_mlp').
I0424 15:32:10.811677 139424412055360 linears.py:541] linears/x Physical: bfloat16[32,2048,7168]...................................... ('fsdp', None, None).
I0424 15:32:11.004408 139424412055360 checkpointing.py:578] checkpoint manager exists so trying to load this run's existing checkpoint
I0424 15:32:11.004524 139424412055360 checkpointing.py:676] No existing checkpoints found, not restoring checkpoint.
fsdp: 32
I0424 15:32:12.400986 139424412055360 maxtext_utils.py:1740]  params/params/decoder/decoder_norm/scale
    Shape:     float32[2048]
    Logical:   PartitionSpec('norm',)
    Physical:  (None,)
I0424 15:32:12.401146 139424412055360 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)
I0424 15:32:12.401210 139424412055360 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)
I0424 15:32:12.401274 139424412055360 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')
I0424 15:32:12.401328 139424412055360 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)
I0424 15:32:12.401367 139424412055360 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)
I0424 15:32:12.401420 139424412055360 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)
I0424 15:32:12.401473 139424412055360 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')
I0424 15:32:12.401522 139424412055360 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)
I0424 15:32:12.401560 139424412055360 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)
I0424 15:32:12.401610 139424412055360 maxtext_utils.py:1740]  params/params/decoder/logits_dense/kernel
    Shape:     float32[2048,32000]
    Logical:   PartitionSpec('embed', 'vocab')
    Physical:  ('fsdp', None)

I0424 15:32:12.401659 139424412055360 maxtext_utils.py:1740]  params/params/token_embedder/embedding
    Shape:     float32[32000,2048]
    Logical:   PartitionSpec('vocab', 'embed')
    Physical:  (None, 'fsdp')
I0424 15:32:12.904163 139424412055360 train.py:158] train/xent Logical: float32[32,2048]............................................ ('activation_embed_and_logits_batch', 'activation_length').
I0424 15:32:12.904267 139424412055360 train.py:158] train/xent Physical: float32[32,2048]............................................ ('fsdp', None).
I0424 15:32:12.922959 139424412055360 train.py:165] train/z_loss Logical: float32[32,2048]............................................ ('activation_embed_and_logits_batch', 'activation_length').
I0424 15:32:12.923071 139424412055360 train.py:165] train/z_loss Physical: float32[32,2048]............................................ ('fsdp', None).
I0424 15:32:23.931848 139424412055360 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.
I0424 15:32:23.932671 139424412055360 metric_logger.py:289] number parameters: 1.104 billion
I0424 15:32:35.003152 139424412055360 checkpointing.py:794] Waiting for step 0 to finish before checkpoint...
I0424 15:32:35.155889 139424412055360 checkpointing.py:798] Waited 0.15272021293640137 seconds for step 0 to finish before starting checkpointing.
I0424 15:32:35.158466 139424412055360 checkpoint_manager.py:2013] [process=6][thread=MainThread][wait_until_finished] No Save Finalize thread to wait for. Returning.
I0424 15:32:35.160227 139424412055360 checkpoint_manager.py:1518] [process=6] Saving checkpoint at step 0
I0424 15:32:35.162904 139424412055360 async_checkpointer.py:452] [process=6] Started async saving checkpoint to gs://lance-maxtext/linen_ckpt_xpk_feat_nnx_set_defaults_true_20260424_145744/linen_xpk_feat_nnx_set_defaults_true_20260424_145744_10_shardy_false/checkpoints/0.
I0424 15:32:35.962990 139424412055360 signaling_client.py:364] Using JaxDistributedSignalingClient
I0424 15:32:35.964263 139424412055360 jax_array_handlers.py:358] Scheduling D2H of 39 prioritized jax.Array.
I0424 15:32:35.964323 139424412055360 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
I0424 15:32:36.209970 139424412055360 base_pytree_checkpoint_handler.py:153] [process=6][thread=MainThread] Initiated "orbax.checkpoint._src.serialization.jax_array_handlers.ArrayHandler".serialize. Time taken: 0.247065s
I0424 15:32:36.210158 139424412055360 base_pytree_checkpoint_handler.py:129] [process=6] /jax/checkpoint/write/blocking_gbytes_per_sec: 3.071 GiB/s (total gbytes: 1.5 GiB) (time elapsed: 0.502249002456665 s) (per-host)
I0424 15:32:36.210216 139424412055360 base_pytree_checkpoint_handler.py:737] [process=6][thread=MainThread] Initiated Pytree async_save. Time taken: 0.502315s (batch_requests_ready=0.251580s, total_serialization_initiated=0.250661s, others=0.000074s)
I0424 15:32:36.210305 139424412055360 composite_checkpoint_handler.py:715] [process=6][thread=MainThread] Initiated CompositeCheckpointHandler.async_save. Time taken: 0.506451s (all_items=0.000018s, per_item={'items': '0.00001812'}, temp_paths=0.506433)
I0424 15:32:36.211248 139299051783936 async_checkpointer.py:79] [process=6][thread=async_save] Background save thread started.
I0424 15:32:36.211412 139424412055360 async_checkpointer.py:561] Finished blocking save. Time taken: 1.051116s. Continuing background save to gs://lance-maxtext/linen_ckpt_xpk_feat_nnx_set_defaults_true_20260424_145744/linen_xpk_feat_nnx_set_defaults_true_20260424_145744_10_shardy_false/checkpoints/0.
I0424 15:32:36.257682 139424412055360 checkpoint_manager.py:1566] [process=6][thread=MainThread][step=0] Starting CheckpointManager Save Finalize thread=save_finalize
I0424 15:32:36.258004 139297332131584 async_checkpointer.py:265] [process=6][thread=save_finalize] Waiting for background save thread=async_save.
I0424 15:32:36.258150 139424412055360 standard_logger.py:34] {'step': 0, 'event_type': 'save', 'directory': 'gs://lance-maxtext/linen_ckpt_xpk_feat_nnx_set_defaults_true_20260424_145744/linen_xpk_feat_nnx_set_defaults_true_20260424_145744_10_shardy_false/checkpoints', 'reached_preemption': False, 'preemption_received_at': None, 'synchronous': False, 'wait_for_prev_start_time': 1777044755.1584473, 'wait_for_prev_duration_secs': 6.771087646484375e-05, 'time_between_consecutive_saves_sec': None, 'checkpointer_blocking_start_time': 1777044755.1602662, 'checkpointer_blocking_duration_secs': 1.0512518882751465, 'get_old_steps_start_time': 1777044756.2115376, 'get_old_steps_duration_secs': 2.4080276489257812e-05, 'checkpoint_manager_blocking_start_time': 1777044755.156365, 'checkpoint_manager_blocking_duration_secs': 1.1017467975616455}
I0424 15:32:36.258313 139424412055360 checkpointing.py:409] Started an asynchronous checkpoint save for step 0
I0424 15:32:36.258368 139424412055360 max_utils.py:750] 
Memstats: After params initialized:
I0424 15:32:36.258423 139424412055360 max_utils.py:756] 	Using (GB) 0.43 / 31.25 (1.376000%) on TPU_24(process=6,(0,6,0,0))
I0424 15:32:36.258458 139424412055360 max_utils.py:756] 	Using (GB) 0.43 / 31.25 (1.376000%) on TPU_25(process=6,(1,6,0,0))
I0424 15:32:36.258487 139424412055360 max_utils.py:756] 	Using (GB) 0.43 / 31.25 (1.376000%) on TPU_28(process=6,(0,7,0,0))
I0424 15:32:36.258513 139424412055360 max_utils.py:756] 	Using (GB) 0.43 / 31.25 (1.376000%) on TPU_29(process=6,(1,7,0,0))
I0424 15:32:36.519747 139424412055360 metric_logger.py:185] completed step: 0, seconds: 11.070, TFLOP/s/device: 1.227, Tokens/s/device: 184.999, total_weights: 65536, loss: 10.874
I0424 15:32:36.688010 139424412055360 metric_logger.py:185] completed step: 1, seconds: 1.515, TFLOP/s/device: 8.967, Tokens/s/device: 1351.557, total_weights: 65536, loss: 10.874
I0424 15:32:37.120591 139424412055360 metric_logger.py:185] completed step: 2, seconds: 0.021, TFLOP/s/device: 644.245, Tokens/s/device: 97107.634, total_weights: 65536, loss: 10.263
I0424 15:32:37.268026 139424412055360 metric_logger.py:185] completed step: 3, seconds: 0.433, TFLOP/s/device: 31.381, Tokens/s/device: 4730.054, total_weights: 65536, loss: 9.734
I0424 15:32:37.563298 139424412055360 metric_logger.py:185] completed step: 4, seconds: 0.153, TFLOP/s/device: 88.936, Tokens/s/device: 13405.335, total_weights: 65536, loss: 9.278
I0424 15:32:37.568989 139424412055360 metric_logger.py:185] completed step: 5, seconds: 0.147, TFLOP/s/device: 92.196, Tokens/s/device: 13896.805, total_weights: 65536, loss: 8.892
I0424 15:32:57.180695 139424412055360 metric_logger.py:185] completed step: 6, seconds: 0.296, TFLOP/s/device: 45.971, Tokens/s/device: 6929.313, total_weights: 65536, loss: 8.592
I0424 15:32:57.328127 139424412055360 metric_logger.py:185] completed step: 7, seconds: 19.465, TFLOP/s/device: 0.698, Tokens/s/device: 105.214, total_weights: 65536, loss: 8.384
I0424 15:32:57.475874 139424412055360 metric_logger.py:185] completed step: 8, seconds: 0.152, TFLOP/s/device: 89.318, Tokens/s/device: 13462.967, total_weights: 65536, loss: 8.255
I0424 15:32:57.622948 139424412055360 checkpointing.py:794] Waiting for step 9 to finish before checkpoint...
I0424 15:32:57.623644 139424412055360 checkpointing.py:798] Waited 0.0007154941558837891 seconds for step 9 to finish before starting checkpointing.
I0424 15:32:57.625613 139424412055360 checkpoint_manager.py:2024] [process=6][thread=MainThread][step=0][wait_until_finished] Waiting for Save Finalize thread (save_finalize) to complete.
I0424 15:32:58.271552    2634 google_auth_provider.cc:181] Running on GCE, using service account 562977990677-compute@developer.gserviceaccount.com
I0424 15:33:00.200798 139299060176640 array_metadata_store.py:203] [process=6][thread=array_type_handler] Wrote 39 array_metadata.ArrayMetadata to gs://lance-maxtext/linen_ckpt_xpk_feat_nnx_set_defaults_true_20260424_145744/linen_xpk_feat_nnx_set_defaults_true_20260424_145744_10_shardy_false/checkpoints/0/items/array_metadatas/process_6
I0424 15:33:31.366553 139299051783936 base_pytree_checkpoint_handler.py:129] [process=6] /jax/checkpoint/write/gbytes_per_sec: 28.380 MiB/s (total gbytes: 1.5 GiB) (time elapsed: 55.658604860305786 s) (per-host)
I0424 15:33:31.366674 139299051783936 async_checkpointer.py:90] [process=6][thread=async_save] 3 Handler Commit operations completed. Time taken: 55.155315s.
I0424 15:33:41.146024 139299051783936 async_checkpointer.py:144] [process=6][thread=async_save] Background save thread done. Time taken: 64.934650s.
I0424 15:33:41.146364 139297332131584 async_checkpointer.py:273] [process=6][thread=save_finalize] Done with waiting for background save thread=async_save.
I0424 15:33:41.146497 139297332131584 async_checkpointer.py:283] [process=6][thread=save_finalize] No errors found in background save thread=async_save.
I0424 15:33:41.146567 139297332131584 checkpoint_manager.py:2133] [process=6][thread=save_finalize][step=0] CheckpointManager Save Finalize is syncing with other hosts...
I0424 15:33:41.148612 139297332131584 checkpoint_manager.py:2142] [process=6][thread=save_finalize][step=0] CheckpointManager Save Finalize is done on all hosts.
I0424 15:33:41.148806 139424412055360 checkpoint_manager.py:2036] [process=6][thread=MainThread][step=0][wait_until_finished] Done waiting for Save Finalize thread (save_finalize) running at step=0.
W0424 15:33:41.148953 139424412055360 checkpoint_manager.py:1458] Waiting for previous save to complete took 43.523343 seconds. If this number is high, consider checkpointing less frequently.
I0424 15:33:41.150827 139424412055360 checkpoint_manager.py:1518] [process=6] Saving checkpoint at step 9
I0424 15:33:41.154325 139424412055360 async_checkpointer.py:452] [process=6] Started async saving checkpoint to gs://lance-maxtext/linen_ckpt_xpk_feat_nnx_set_defaults_true_20260424_145744/linen_xpk_feat_nnx_set_defaults_true_20260424_145744_10_shardy_false/checkpoints/9.
I0424 15:33:41.703510 139424412055360 jax_array_handlers.py:358] Scheduling D2H of 39 prioritized jax.Array.
I0424 15:33:41.703614 139424412055360 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
I0424 15:33:41.748139 139424412055360 base_pytree_checkpoint_handler.py:153] [process=6][thread=MainThread] Initiated "orbax.checkpoint._src.serialization.jax_array_handlers.ArrayHandler".serialize. Time taken: 0.045882s
I0424 15:33:41.748319 139424412055360 base_pytree_checkpoint_handler.py:129] [process=6] /jax/checkpoint/write/blocking_gbytes_per_sec: 5.196 GiB/s (total gbytes: 1.5 GiB) (time elapsed: 0.2968904972076416 s) (per-host)
I0424 15:33:41.748373 139424412055360 base_pytree_checkpoint_handler.py:737] [process=6][thread=MainThread] Initiated Pytree async_save. Time taken: 0.296955s (batch_requests_ready=0.249161s, total_serialization_initiated=0.047721s, others=0.000072s)
I0424 15:33:41.748466 139424412055360 composite_checkpoint_handler.py:715] [process=6][thread=MainThread] Initiated CompositeCheckpointHandler.async_save. Time taken: 0.300896s (all_items=0.000014s, per_item={'items': '0.00001383'}, temp_paths=0.300882)
I0424 15:33:41.749721 139299043391232 async_checkpointer.py:79] [process=6][thread=async_save] Background save thread started.
I0424 15:33:41.749877 139424412055360 async_checkpointer.py:561] Finished blocking save. Time taken: 0.598979s. Continuing background save to gs://lance-maxtext/linen_ckpt_xpk_feat_nnx_set_defaults_true_20260424_145744/linen_xpk_feat_nnx_set_defaults_true_20260424_145744_10_shardy_false/checkpoints/9.
I0424 15:33:41.929755 139424412055360 checkpoint_manager.py:1566] [process=6][thread=MainThread][step=9] Starting CheckpointManager Save Finalize thread=save_finalize
I0424 15:33:41.930145 139297332131584 async_checkpointer.py:265] [process=6][thread=save_finalize] Waiting for background save thread=async_save.
I0424 15:33:41.930319 139424412055360 standard_logger.py:34] {'step': 9, 'event_type': 'save', 'directory': 'gs://lance-maxtext/linen_ckpt_xpk_feat_nnx_set_defaults_true_20260424_145744/linen_xpk_feat_nnx_set_defaults_true_20260424_145744_10_shardy_false/checkpoints', 'reached_preemption': False, 'preemption_received_at': None, 'synchronous': False, 'wait_for_prev_start_time': 1777044777.6255836, 'wait_for_prev_duration_secs': 43.5233428478241, 'time_between_consecutive_saves_sec': None, 'checkpointer_blocking_start_time': 1777044821.150866, 'checkpointer_blocking_duration_secs': 0.5991182327270508, 'get_old_steps_start_time': 1777044821.7500029, 'get_old_steps_duration_secs': 2.5987625122070312e-05, 'checkpoint_manager_blocking_start_time': 1777044777.6238716, 'checkpoint_manager_blocking_duration_secs': 44.30641031265259}
I0424 15:33:41.930439 139424412055360 checkpointing.py:409] Started an asynchronous checkpoint save for step 9
I0424 15:33:41.930485 139424412055360 checkpoint_manager.py:2024] [process=6][thread=MainThread][step=9][wait_until_finished] Waiting for Save Finalize thread (save_finalize) to complete.
I0424 15:33:46.559253 139299060176640 array_metadata_store.py:203] [process=6][thread=array_type_handler] Wrote 39 array_metadata.ArrayMetadata to gs://lance-maxtext/linen_ckpt_xpk_feat_nnx_set_defaults_true_20260424_145744/linen_xpk_feat_nnx_set_defaults_true_20260424_145744_10_shardy_false/checkpoints/9/items/array_metadatas/process_6
I0424 15:34:22.937610 139299043391232 base_pytree_checkpoint_handler.py:129] [process=6] /jax/checkpoint/write/gbytes_per_sec: 38.075 MiB/s (total gbytes: 1.5 GiB) (time elapsed: 41.48614573478699 s) (per-host)
I0424 15:34:22.937719 139299043391232 async_checkpointer.py:90] [process=6][thread=async_save] 3 Handler Commit operations completed. Time taken: 41.187887s.
I0424 15:34:31.429510 139299043391232 async_checkpointer.py:144] [process=6][thread=async_save] Background save thread done. Time taken: 49.679661s.
I0424 15:34:31.429776 139297332131584 async_checkpointer.py:273] [process=6][thread=save_finalize] Done with waiting for background save thread=async_save.
I0424 15:34:31.429898 139297332131584 async_checkpointer.py:283] [process=6][thread=save_finalize] No errors found in background save thread=async_save.
I0424 15:34:31.429946 139297332131584 checkpoint_manager.py:2133] [process=6][thread=save_finalize][step=9] CheckpointManager Save Finalize is syncing with other hosts...
I0424 15:34:31.431438 139297332131584 checkpoint_manager.py:2142] [process=6][thread=save_finalize][step=9] CheckpointManager Save Finalize is done on all hosts.
I0424 15:34:31.431633 139424412055360 checkpoint_manager.py:2036] [process=6][thread=MainThread][step=9][wait_until_finished] Done waiting for Save Finalize thread (save_finalize) running at step=9.
I0424 15:34:31.431782 139424412055360 checkpoint_manager.py:2013] [process=6][thread=MainThread][wait_until_finished] No Save Finalize thread to wait for. Returning.
I0424 15:34:31.432518 139424412055360 metric_logger.py:185] completed step: 9, seconds: 0.147, TFLOP/s/device: 92.130, Tokens/s/device: 13886.817, total_weights: 65536, loss: 8.179
Per train step:
 Total TFLOPs: 13.59 
 split as 93.93% learnable weight flops and 6.07% attention flops
XPK End: Fri Apr 24 15:34:43 UTC 2026
EXIT_CODE=0