MaxView

Timing Test Reference Updates
‹ 03_dropoutCase: 04_int805_fp8 ›

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 loss8.17508.17500
TFLOP/s59.22759.170-0.057
Tok/s8927.38918.7-8.592
Avg s/step3.5872.795-0.792
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 07:11:34 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 07:11:59.390322: 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 07:11:59.610493 139157606033216 max_utils.py:273] Attempting to initialize the jax distributed system...
I0424 07:12:08.650213 139157606033216 distributed.py:149] Starting JAX distributed service on [::]:8482
I0424 07:12:08.652699 139157606033216 distributed.py:172] Connecting to JAX distributed service on mt-04-int8-uwur0-slice-job-0-0.mt-04-int8-uwur0:8482
I0424 07:12:09.437019 139157606033216 max_utils.py:284] Jax distributed system initialized!
I0424 07:12:15.522265 139157606033216 max_utils.py:800] System Information: Jax Version: 0.9.2
I0424 07:12:15.522370 139157606033216 max_utils.py:801] System Information: Jaxlib Version: 0.9.2
I0424 07:12:15.522411 139157606033216 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 07:12:15.522446 139157606033216 train_utils.py:361] WARNING: Sequence packing is essentially ignored for synthetic data. Please use a real dataset to use sequence packing.
I0424 07:12:16.222576 139157606033216 maxtext_utils.py:1604] Num_devices: 32, shape (1, 1, 1, 32, 1, 1, 1, 1, 1, 1, 1, 1, 1)
I0424 07:12:16.222993 139157606033216 checkpointing.py:677] Setting up checkpoint logger...
I0424 07:12:16.223052 139157606033216 checkpointing.py:233] Creating checkpoint manager with ocdbt=True and zarr3=True
I0424 07:12:16.223096 139157606033216 pytree_checkpoint_handler.py:592] save_device_host_concurrent_bytes=None
I0424 07:12:16.223426 139157606033216 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 0x7e8fa0314350>, enable_pinned_host_transfer=False, save_concurrent_bytes: 96000000000 (89.4 GiB), restore_concurrent_bytes: 96000000000 (89.4 GiB)
I0424 07:12:19.157739 139157606033216 checkpointing.py:265] Enabling policy for fixed interval checkpointing.
I0424 07:12:19.157978 139157606033216 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 0x7e7b4436be90>}, handler_registry=None
I0424 07:12:19.158254 139157606033216 composite_checkpoint_handler.py:237] Deferred registration for item: "items". Adding handler `<orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeCheckpointHandler object at 0x7e7b4436be90>` 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 07:12:19.158307 139157606033216 composite_checkpoint_handler.py:237] Deferred registration for item: "metrics". Adding handler `<orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonCheckpointHandler object at 0x7e7b44375d60>` 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 07:12:19.158344 139157606033216 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 0x7e7b4436be90>, ('items', <class 'orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeRestoreArgs'>): <orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeCheckpointHandler object at 0x7e7b4436be90>, ('metrics', <class 'orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonSaveArgs'>): <orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonCheckpointHandler object at 0x7e7b44375d60>, ('metrics', <class 'orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonRestoreArgs'>): <orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonCheckpointHandler object at 0x7e7b44375d60>}).
I0424 07:12:19.158664 139157606033216 abstract_checkpointer.py:35] orbax-checkpoint version: 0.11.34
I0424 07:12:19.158735 139157606033216 async_checkpointer.py:192] [process=5][thread=MainThread] Using barrier_sync_fn: <function get_barrier_sync_fn.<locals>._fn at 0x7e7a506818a0> timeout: 1200 secs and primary_host=0 for async checkpoint writes
I0424 07:12:20.109352 139157606033216 checkpoint_manager.py:1812] Found 0 checkpoint steps in gs://lance-maxtext/linen_ckpt_xpk_main_20260424_070227/linen_xpk_main_20260424_070227_04_int8/checkpoints
I0424 07:12:20.114812 139157606033216 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_04_int8/checkpoints: <orbax.checkpoint.checkpoint_manager.CheckpointManager object at 0x7e7b44373a70>
I0424 07:12:20.114935 139157606033216 checkpointing.py:301] Checkpoint manager created!
I0424 07:12:21.263098 139157606033216 nnx_wrappers.py:437] Unknown Logical: bfloat16[32,2048,2048]...................................... ('activation_batch', 'activation_norm_length', 'activation_embed').
I0424 07:12:21.263217 139157606033216 nnx_wrappers.py:437] Unknown Physical: bfloat16[32,2048,2048]...................................... ('fsdp', None, None).
I0424 07:12:21.645049 139157606033216 attentions.py:1088] attentions/inputs_q Logical: bfloat16[32,2048,2048]...................................... ('activation_batch', 'activation_attn_length', 'activation_attn_embed').
I0424 07:12:21.645145 139157606033216 attentions.py:1088] attentions/inputs_q Physical: bfloat16[32,2048,2048]...................................... ('fsdp', None, None).
I0424 07:12:21.661672 139157606033216 attentions.py:1089] attentions/inputs_kv Logical: bfloat16[32,2048,2048]...................................... ('activation_batch', 'activation_attn_length', 'activation_attn_embed').
I0424 07:12:21.661735 139157606033216 attentions.py:1089] attentions/inputs_kv Physical: bfloat16[32,2048,2048]...................................... ('fsdp', None, None).
I0424 07:12:21.728067 139157606033216 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 07:12:21.728154 139157606033216 attentions.py:1154] attentions/query Physical: bfloat16[32,2048,16,128].................................... ('fsdp', None, None, None).
I0424 07:12:21.744702 139157606033216 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 07:12:21.744765 139157606033216 attentions.py:1155] attentions/key Physical: bfloat16[32,2048,16,128].................................... ('fsdp', None, None, None).
I0424 07:12:21.761294 139157606033216 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 07:12:21.761355 139157606033216 attentions.py:1156] attentions/value Physical: bfloat16[32,2048,16,128].................................... ('fsdp', None, None, None).
I0424 07:12:21.786704 139157606033216 attentions.py:1198] attentions/out Logical: bfloat16[32,2048,16,128].................................... ('activation_batch', 'activation_attn_length', 'activation_heads', 'activation_kv').
I0424 07:12:21.786770 139157606033216 attentions.py:1198] attentions/out Physical: bfloat16[32,2048,16,128].................................... ('fsdp', None, None, None).
I0424 07:12:21.852427 139157606033216 linears.py:525] linears/x Logical: bfloat16[32,2048,7168]...................................... ('activation_batch', 'activation_length', 'activation_mlp').
I0424 07:12:21.852518 139157606033216 linears.py:525] linears/x Physical: bfloat16[32,2048,7168]...................................... ('fsdp', None, None).
I0424 07:12:22.360509 139157606033216 checkpointing.py:577] checkpoint manager exists so trying to load this run's existing checkpoint
I0424 07:12:22.360623 139157606033216 checkpointing.py:665] No existing checkpoints found, not restoring checkpoint.
fsdp: 32
I0424 07:12:24.830140 139157606033216 maxtext_utils.py:1707]  params/params/decoder/decoder_norm/scale
    Shape:     float32[2048]
    Logical:   P('norm',)
    Physical:  (None,)
I0424 07:12:24.830272 139157606033216 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 07:12:24.830325 139157606033216 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 07:12:24.830383 139157606033216 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 07:12:24.830435 139157606033216 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 07:12:24.830474 139157606033216 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 07:12:24.830531 139157606033216 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 07:12:24.830586 139157606033216 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 07:12:24.830628 139157606033216 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 07:12:24.830687 139157606033216 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 07:12:24.830741 139157606033216 maxtext_utils.py:1707]  params/params/decoder/logits_dense/kernel
    Shape:     float32[2048,32000]
    Logical:   P('embed_vocab', 'vocab')
    Physical:  ('fsdp', None)

I0424 07:12:24.830789 139157606033216 maxtext_utils.py:1707]  params/params/token_embedder/embedding
    Shape:     float32[32000,2048]
    Logical:   P('vocab', 'embed_vocab')
    Physical:  (None, 'fsdp')
I0424 07:12:26.214032 139157606033216 train.py:155] train/xent Logical: float32[32,2048]............................................ ('activation_embed_and_logits_batch', 'activation_length').
I0424 07:12:26.214147 139157606033216 train.py:155] train/xent Physical: float32[32,2048]............................................ ('fsdp', None).
I0424 07:12:26.229685 139157606033216 train.py:162] train/z_loss Logical: float32[32,2048]............................................ ('activation_embed_and_logits_batch', 'activation_length').
I0424 07:12:26.229744 139157606033216 train.py:162] train/z_loss Physical: float32[32,2048]............................................ ('fsdp', None).
I0424 07:12:40.123329 139157606033216 max_utils.py:791] Total memory size: 1.8 GB, Output size: 0.4 GB, Temp size: 1.4 GB, Argument size: 0.4 GB, Host temp size: 0.0 GB.
I0424 07:12:40.124143 139157606033216 metric_logger.py:301] number parameters: 1.104 billion
I0424 07:12:54.821469 139157606033216 checkpointing.py:772] Waiting for step 0 to finish before checkpoint...
I0424 07:12:55.060811 139157606033216 checkpointing.py:776] Waited 0.23932433128356934 seconds for step 0 to finish before starting checkpointing.
I0424 07:12:55.063256 139157606033216 checkpoint_manager.py:2009] [process=5][thread=MainThread][wait_until_finished] No Save Finalize thread to wait for. Returning.
I0424 07:12:55.065171 139157606033216 checkpoint_manager.py:1512] [process=5] Saving checkpoint at step 0
I0424 07:12:55.066665 139157606033216 event_tracking.py:70] [process=5] [async] Started save checkpoint @ gs://lance-maxtext/linen_ckpt_xpk_main_20260424_070227/linen_xpk_main_20260424_070227_04_int8/checkpoints/0.
I0424 07:12:55.732296 139157606033216 signaling_client.py:364] Using JaxDistributedSignalingClient
I0424 07:12:55.733244 139157606033216 jax_array_handlers.py:360] Scheduling D2H of 39 prioritized jax.Array.
I0424 07:12:55.733301 139157606033216 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 07:12:56.020544 139157606033216 base_pytree_checkpoint_handler.py:154] [process=5][thread=MainThread] Initiated "orbax.checkpoint._src.serialization.jax_array_handlers.ArrayHandler".serialize. Time taken: 0.288320s
I0424 07:12:56.020728 139157606033216 base_pytree_checkpoint_handler.py:130] [process=5] /jax/orbax/write/blocking_gbytes_per_sec: 5.251 GiB/s (total gbytes: 1.5 GiB) (time elapsed: 0.2937946319580078 s) (per-host)
I0424 07:12:56.020783 139157606033216 base_pytree_checkpoint_handler.py:768] [process=5][thread=MainThread] Initiated Pytree async_save. Time taken: 0.293861s (batch_requests_ready=0.002258s, total_serialization_initiated=0.291528s, others=0.000074s)
I0424 07:12:56.020880 139157606033216 composite_checkpoint_handler.py:715] [process=5][thread=MainThread] Initiated CompositeCheckpointHandler.async_save. Time taken: 0.298027s (all_items=0.000018s, per_item={'items': '0.00001764'}, temp_paths=0.298010)
I0424 07:12:56.021755 139157606033216 event_tracking.py:125] [process=5] [async] Finished blocking save in 0.96 seconds. Continuing save @ gs://lance-maxtext/linen_ckpt_xpk_main_20260424_070227/linen_xpk_main_20260424_070227_04_int8/checkpoints/0.
I0424 07:12:56.022096 139024241719040 async_checkpointer.py:76] [process=5][thread=async_save] Background save thread started. Deadline for this save operation is 2026-04-24 07:32:56.022057
I0424 07:12:56.024055 139157606033216 checkpoint_manager.py:1560] [process=5][thread=MainThread][step=0] Starting CheckpointManager Save Finalize thread=save_finalize
I0424 07:12:56.024309 139022876858112 async_checkpointer.py:280] [process=5][thread=save_finalize] Waiting for background save thread=async_save.
I0424 07:12:56.024439 139157606033216 standard_logger.py:34] {'step': 0, 'event_type': 'save', 'directory': 'gs://lance-maxtext/linen_ckpt_xpk_main_20260424_070227/linen_xpk_main_20260424_070227_04_int8/checkpoints', 'reached_preemption': False, 'preemption_received_at': None, 'synchronous': False, 'wait_for_prev_start_time': 1777014775.0632372, 'wait_for_prev_duration_secs': 6.127357482910156e-05, 'time_between_consecutive_saves_sec': None, 'checkpointer_blocking_start_time': 1777014775.0652099, 'checkpointer_blocking_duration_secs': 0.957036018371582, 'get_old_steps_start_time': 1777014776.0222704, 'get_old_steps_duration_secs': 3.218650817871094e-05, 'checkpoint_manager_blocking_start_time': 1777014775.0613558, 'checkpoint_manager_blocking_duration_secs': 0.9630439281463623}
I0424 07:12:56.024601 139157606033216 checkpointing.py:408] Started an asynchronous checkpoint save for step 0
I0424 07:12:56.024667 139157606033216 max_utils.py:750] 
Memstats: After params initialized:
I0424 07:12:56.024724 139157606033216 max_utils.py:756] 	Using (GB) 0.44 / 31.25 (1.408000%) on TPU_18(process=5,(2,4,0,0))
I0424 07:12:56.024758 139157606033216 max_utils.py:756] 	Using (GB) 0.43 / 31.25 (1.376000%) on TPU_19(process=5,(3,4,0,0))
I0424 07:12:56.024786 139157606033216 max_utils.py:756] 	Using (GB) 0.43 / 31.25 (1.376000%) on TPU_22(process=5,(2,5,0,0))
I0424 07:12:56.024812 139157606033216 max_utils.py:756] 	Using (GB) 0.43 / 31.25 (1.376000%) on TPU_23(process=5,(3,5,0,0))
I0424 07:12:56.354707 139157606033216 metric_logger.py:196] completed step: 0, seconds: 14.697, TFLOP/s/device: 0.924, Tokens/s/device: 139.346, total_weights: 65536, loss: 10.874, lm_loss: 10.874, perplexity: 52796.277
I0424 07:12:56.595214 139157606033216 metric_logger.py:196] completed step: 1, seconds: 1.532, TFLOP/s/device: 8.871, Tokens/s/device: 1337.077, total_weights: 65536, loss: 10.874, lm_loss: 10.874, perplexity: 52796.277
I0424 07:12:57.004308 139157606033216 metric_logger.py:196] completed step: 2, seconds: 0.012, TFLOP/s/device: 1173.935, Tokens/s/device: 176948.332, total_weights: 65536, loss: 10.262, lm_loss: 10.262, perplexity: 28634.820
I0424 07:12:57.233771 139157606033216 metric_logger.py:196] completed step: 3, seconds: 0.409, TFLOP/s/device: 33.182, Tokens/s/device: 5001.514, total_weights: 65536, loss: 9.731, lm_loss: 9.731, perplexity: 16827.199
I0424 07:12:57.692815 139157606033216 metric_logger.py:196] completed step: 4, seconds: 0.235, TFLOP/s/device: 57.883, Tokens/s/device: 8724.806, total_weights: 65536, loss: 9.272, lm_loss: 9.272, perplexity: 10638.531
I0424 07:12:57.698796 139157606033216 metric_logger.py:196] completed step: 5, seconds: 0.229, TFLOP/s/device: 59.250, Tokens/s/device: 8930.791, total_weights: 65536, loss: 8.887, lm_loss: 8.887, perplexity: 7233.941
I0424 07:12:58.959826    2924 google_auth_provider.cc:181] Running on GCE, using service account 562977990677-compute@developer.gserviceaccount.com
I0424 07:13:01.475111 139022885250816 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_04_int8/checkpoints/0/items/array_metadatas/process_5
I0424 07:13:26.869181 139157606033216 metric_logger.py:196] completed step: 6, seconds: 0.460, TFLOP/s/device: 29.564, Tokens/s/device: 4456.136, total_weights: 65536, loss: 8.588, lm_loss: 8.588, perplexity: 5365.468
I0424 07:13:27.098499 139157606033216 metric_logger.py:196] completed step: 7, seconds: 28.942, TFLOP/s/device: 0.469, Tokens/s/device: 70.761, total_weights: 65536, loss: 8.380, lm_loss: 8.380, perplexity: 4358.779
I0424 07:13:27.328148 139157606033216 metric_logger.py:196] completed step: 8, seconds: 0.234, TFLOP/s/device: 58.172, Tokens/s/device: 8768.362, total_weights: 65536, loss: 8.251, lm_loss: 8.251, perplexity: 3831.453
I0424 07:13:27.556835 139157606033216 checkpointing.py:772] Waiting for step 9 to finish before checkpoint...
I0424 07:13:27.557534 139157606033216 checkpointing.py:776] Waited 0.0007200241088867188 seconds for step 9 to finish before starting checkpointing.
I0424 07:13:27.559574 139157606033216 checkpoint_manager.py:2020] [process=5][thread=MainThread][step=0][wait_until_finished] Waiting for Save Finalize thread (save_finalize) to complete.
I0424 07:13:28.872802 139024241719040 base_pytree_checkpoint_handler.py:130] [process=5] /jax/orbax/write/gbytes_per_sec: 47.656 MiB/s (total gbytes: 1.5 GiB) (time elapsed: 33.14583134651184 s) (per-host)
I0424 07:13:28.872924 139024241719040 async_checkpointer.py:90] [process=5][thread=async_save] 3 Handler Commit operations completed. Time taken: 32.850717s.
I0424 07:13:38.377211 139024241719040 async_checkpointer.py:160] [process=5][thread=async_save] Background save thread done. Time taken: 42.354988s.
I0424 07:13:38.377501 139022876858112 async_checkpointer.py:288] [process=5][thread=save_finalize] Done with waiting for background save thread=async_save.
I0424 07:13:38.377619 139022876858112 async_checkpointer.py:298] [process=5][thread=save_finalize] No errors found in background save thread=async_save.
I0424 07:13:38.377684 139022876858112 checkpoint_manager.py:2137] [process=5][thread=save_finalize][step=0] CheckpointManager Save Finalize is syncing with other hosts...
I0424 07:13:38.379979 139022876858112 checkpoint_manager.py:2146] [process=5][thread=save_finalize][step=0] CheckpointManager Save Finalize is done on all hosts.
I0424 07:13:38.380143 139157606033216 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 07:13:38.380276 139157606033216 checkpoint_manager.py:1452] Waiting for previous save to complete took 10.820704 seconds. If this number is high, consider checkpointing less frequently.
I0424 07:13:38.382047 139157606033216 checkpoint_manager.py:1512] [process=5] Saving checkpoint at step 9
I0424 07:13:38.384174 139157606033216 event_tracking.py:70] [process=5] [async] Started save checkpoint @ gs://lance-maxtext/linen_ckpt_xpk_main_20260424_070227/linen_xpk_main_20260424_070227_04_int8/checkpoints/9.
I0424 07:13:38.660115 139157606033216 jax_array_handlers.py:360] Scheduling D2H of 39 prioritized jax.Array.
I0424 07:13:38.660210 139157606033216 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 07:13:38.694504 139157606033216 base_pytree_checkpoint_handler.py:154] [process=5][thread=MainThread] Initiated "orbax.checkpoint._src.serialization.jax_array_handlers.ArrayHandler".serialize. Time taken: 0.036338s
I0424 07:13:38.694683 139157606033216 base_pytree_checkpoint_handler.py:130] [process=5] /jax/orbax/write/blocking_gbytes_per_sec: 38.732 GiB/s (total gbytes: 1.5 GiB) (time elapsed: 0.03982710838317871 s) (per-host)
I0424 07:13:38.694736 139157606033216 base_pytree_checkpoint_handler.py:768] [process=5][thread=MainThread] Initiated Pytree async_save. Time taken: 0.039891s (batch_requests_ready=0.001833s, total_serialization_initiated=0.037976s, others=0.000082s)
I0424 07:13:38.694833 139157606033216 composite_checkpoint_handler.py:715] [process=5][thread=MainThread] Initiated CompositeCheckpointHandler.async_save. Time taken: 0.044264s (all_items=0.000015s, per_item={'items': '0.00001550'}, temp_paths=0.044249)
I0424 07:13:38.695542 139157606033216 event_tracking.py:125] [process=5] [async] Finished blocking save in 0.31 seconds. Continuing save @ gs://lance-maxtext/linen_ckpt_xpk_main_20260424_070227/linen_xpk_main_20260424_070227_04_int8/checkpoints/9.
I0424 07:13:38.695857 139022876858112 async_checkpointer.py:76] [process=5][thread=async_save] Background save thread started. Deadline for this save operation is 2026-04-24 07:33:38.695825
I0424 07:13:38.697777 139157606033216 checkpoint_manager.py:1560] [process=5][thread=MainThread][step=9] Starting CheckpointManager Save Finalize thread=save_finalize
I0424 07:13:38.698029 139001709135616 async_checkpointer.py:280] [process=5][thread=save_finalize] Waiting for background save thread=async_save.
I0424 07:13:38.698190 139157606033216 standard_logger.py:34] {'step': 9, 'event_type': 'save', 'directory': 'gs://lance-maxtext/linen_ckpt_xpk_main_20260424_070227/linen_xpk_main_20260424_070227_04_int8/checkpoints', 'reached_preemption': False, 'preemption_received_at': None, 'synchronous': False, 'wait_for_prev_start_time': 1777014807.5595417, 'wait_for_prev_duration_secs': 10.820703506469727, 'time_between_consecutive_saves_sec': None, 'checkpointer_blocking_start_time': 1777014818.3820853, 'checkpointer_blocking_duration_secs': 0.3139059543609619, 'get_old_steps_start_time': 1777014818.6960135, 'get_old_steps_duration_secs': 3.218650817871094e-05, 'checkpoint_manager_blocking_start_time': 1777014807.5577981, 'checkpoint_manager_blocking_duration_secs': 11.140357255935669}
I0424 07:13:38.698305 139157606033216 checkpointing.py:408] Started an asynchronous checkpoint save for step 9
I0424 07:13:38.698352 139157606033216 checkpoint_manager.py:2020] [process=5][thread=MainThread][step=9][wait_until_finished] Waiting for Save Finalize thread (save_finalize) to complete.
I0424 07:13:44.237746 139001633601280 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_04_int8/checkpoints/9/items/array_metadatas/process_5
I0424 07:14:20.305140 139022876858112 base_pytree_checkpoint_handler.py:130] [process=5] /jax/orbax/write/gbytes_per_sec: 37.925 MiB/s (total gbytes: 1.5 GiB) (time elapsed: 41.65026116371155 s) (per-host)
I0424 07:14:20.305252 139022876858112 async_checkpointer.py:90] [process=5][thread=async_save] 3 Handler Commit operations completed. Time taken: 41.609289s.
I0424 07:14:29.628173 139022876858112 async_checkpointer.py:160] [process=5][thread=async_save] Background save thread done. Time taken: 50.932192s.
I0424 07:14:29.628504 139001709135616 async_checkpointer.py:288] [process=5][thread=save_finalize] Done with waiting for background save thread=async_save.
I0424 07:14:29.628620 139001709135616 async_checkpointer.py:298] [process=5][thread=save_finalize] No errors found in background save thread=async_save.
I0424 07:14:29.628693 139001709135616 checkpoint_manager.py:2137] [process=5][thread=save_finalize][step=9] CheckpointManager Save Finalize is syncing with other hosts...
I0424 07:14:29.629914 139001709135616 checkpoint_manager.py:2146] [process=5][thread=save_finalize][step=9] CheckpointManager Save Finalize is done on all hosts.
I0424 07:14:29.630142 139157606033216 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 07:14:29.630306 139157606033216 checkpoint_manager.py:2009] [process=5][thread=MainThread][wait_until_finished] No Save Finalize thread to wait for. Returning.
I0424 07:14:29.631201 139157606033216 metric_logger.py:196] completed step: 9, seconds: 0.229, TFLOP/s/device: 59.227, Tokens/s/device: 8927.326, total_weights: 65536, loss: 8.175, lm_loss: 8.175, perplexity: 3552.333
Per train step:
 Total TFLOPs: 13.59 
 split as 93.93% learnable weight flops and 6.07% attention flops
XPK End: Fri Apr 24 07:14:39 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 15:07:13 UTC 2026
2026-04-24 15:07:17.583913: 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 00:00:1777043237.596396      10 cuda_dnn.cc:8579] Unable to register cuDNN factory: Attempting to register factory for plugin cuDNN when one has already been registered
E0000 00:00:1777043237.600123      10 cuda_blas.cc:1407] Unable to register cuBLAS factory: Attempting to register factory for plugin cuBLAS when one has already been registered
W0000 00:00:1777043237.611348      10 computation_placer.cc:177] computation placer already registered. Please check linkage and avoid linking the same target more than once.
W0000 00:00:1777043237.611365      10 computation_placer.cc:177] computation placer already registered. Please check linkage and avoid linking the same target more than once.
W0000 00:00:1777043237.611367      10 computation_placer.cc:177] computation placer already registered. Please check linkage and avoid linking the same target more than once.
W0000 00:00:1777043237.611370      10 computation_placer.cc:177] computation placer already registered. Please check linkage and avoid linking the same target more than once.
2026-04-24 15:07:36.731900: 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 15:07:37.259340 138441071458112 max_utils.py:273] Attempting to initialize the jax distributed system...
INFO:2026-04-24 15:07:46,301:jax._src.distributed:140: Starting JAX distributed service on [::]:8482
I0424 15:07:46.301152 138441071458112 distributed.py:140] Starting JAX distributed service on [::]:8482
INFO:2026-04-24 15:07:46,303:jax._src.distributed:157: Connecting to JAX distributed service on mt-04-int8-za2tg-slice-job-0-0.mt-04-int8-za2tg:8482
I0424 15:07:46.303409 138441071458112 distributed.py:157] Connecting to JAX distributed service on mt-04-int8-za2tg-slice-job-0-0.mt-04-int8-za2tg:8482
I0424 15:07:47.684708 138441071458112 max_utils.py:284] Jax distributed system initialized!
I0424 15:07:54.671586 138441071458112 max_utils.py:800] System Information: Jax Version: 0.8.1
I0424 15:07:54.671711 138441071458112 max_utils.py:801] System Information: Jaxlib Version: 0.8.1
I0424 15:07:54.671753 138441071458112 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 15:07:54.671789 138441071458112 train_utils.py:377] WARNING: Sequence packing is essentially ignored for synthetic data. Please use a real dataset to use sequence packing.
I0424 15:07:55.294133 138441071458112 maxtext_utils.py:1631] Num_devices: 32, shape (1, 1, 1, 32, 1, 1, 1, 1, 1, 1, 1, 1, 1)
I0424 15:07:55.294749 138441071458112 maxtext_utils.py:1631] Num_devices: 32, shape (1, 1, 1, 32, 1, 1, 1, 1, 1, 1, 1, 1, 1)
I0424 15:07:55.295052 138441071458112 checkpointing.py:688] Setting up checkpoint logger...
I0424 15:07:55.295107 138441071458112 checkpointing.py:234] Creating checkpoint manager with ocdbt=True and zarr3=True
I0424 15:07:55.295149 138441071458112 pytree_checkpoint_handler.py:589] save_device_host_concurrent_bytes=None
I0424 15:07:55.295488 138441071458112 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 0x7de8d3975070>, enable_pinned_host_transfer=False, save_concurrent_bytes: 96000000000 (89.4 GiB), restore_concurrent_bytes: 96000000000 (89.4 GiB)
I0424 15:07:58.225603 138441071458112 checkpointing.py:266] Enabling policy for fixed interval checkpointing.
I0424 15:07:58.225978 138441071458112 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 0x7dd5fcb09d00>}, handler_registry=None
I0424 15:07:58.226216 138441071458112 composite_checkpoint_handler.py:237] Deferred registration for item: "items". Adding handler `<orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeCheckpointHandler object at 0x7dd5fcb09d00>` 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 15:07:58.226265 138441071458112 composite_checkpoint_handler.py:237] Deferred registration for item: "metrics". Adding handler `<orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonCheckpointHandler object at 0x7dd5fcb0eb10>` 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 15:07:58.226300 138441071458112 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 0x7dd5fcb09d00>, ('items', <class 'orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeRestoreArgs'>): <orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeCheckpointHandler object at 0x7dd5fcb09d00>, ('metrics', <class 'orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonSaveArgs'>): <orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonCheckpointHandler object at 0x7dd5fcb0eb10>, ('metrics', <class 'orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonRestoreArgs'>): <orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonCheckpointHandler object at 0x7dd5fcb0eb10>}).
I0424 15:07:58.226826 138441071458112 abstract_checkpointer.py:35] orbax-checkpoint version: 0.11.33
I0424 15:07:58.226905 138441071458112 async_checkpointer.py:177] [process=5][thread=MainThread] Using barrier_sync_fn: <function get_barrier_sync_fn.<locals>._fn at 0x7dd5fc0d8180> timeout: 600 secs and primary_host=0 for async checkpoint writes
I0424 15:07:59.200645 138441071458112 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_04_int8/checkpoints
I0424 15:07:59.478252 138441071458112 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_20260424_145744/linen_xpk_feat_nnx_set_defaults_true_20260424_145744_04_int8/checkpoints: <orbax.checkpoint.checkpoint_manager.CheckpointManager object at 0x7dd5fcb0a240>
I0424 15:07:59.478425 138441071458112 checkpointing.py:302] Checkpoint manager created!
I0424 15:08:00.662442 138441071458112 nnx_wrappers.py:455] Unknown Logical: bfloat16[32,2048,2048]...................................... ('activation_batch', 'activation_norm_length', 'activation_embed').
I0424 15:08:00.662550 138441071458112 nnx_wrappers.py:455] Unknown Physical: bfloat16[32,2048,2048]...................................... ('fsdp', None, None).
I0424 15:08:01.037354 138441071458112 attentions.py:1084] attentions/inputs_q Logical: bfloat16[32,2048,2048]...................................... ('activation_batch', 'activation_attn_length', 'activation_attn_embed').
I0424 15:08:01.037444 138441071458112 attentions.py:1084] attentions/inputs_q Physical: bfloat16[32,2048,2048]...................................... ('fsdp', None, None).
I0424 15:08:01.053305 138441071458112 attentions.py:1085] attentions/inputs_kv Logical: bfloat16[32,2048,2048]...................................... ('activation_batch', 'activation_attn_length', 'activation_attn_embed').
I0424 15:08:01.053363 138441071458112 attentions.py:1085] attentions/inputs_kv Physical: bfloat16[32,2048,2048]...................................... ('fsdp', None, None).
I0424 15:08:01.131710 138441071458112 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 15:08:01.131801 138441071458112 attentions.py:1150] attentions/query Physical: bfloat16[32,2048,16,128].................................... ('fsdp', None, None, None).
I0424 15:08:01.147800 138441071458112 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 15:08:01.147863 138441071458112 attentions.py:1151] attentions/key Physical: bfloat16[32,2048,16,128].................................... ('fsdp', None, None, None).
I0424 15:08:01.163642 138441071458112 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 15:08:01.163713 138441071458112 attentions.py:1152] attentions/value Physical: bfloat16[32,2048,16,128].................................... ('fsdp', None, None, None).
I0424 15:08:01.188093 138441071458112 attentions.py:1193] attentions/out Logical: bfloat16[32,2048,16,128].................................... ('activation_batch', 'activation_attn_length', 'activation_heads', 'activation_kv').
I0424 15:08:01.188163 138441071458112 attentions.py:1193] attentions/out Physical: bfloat16[32,2048,16,128].................................... ('fsdp', None, None, None).
I0424 15:08:01.269166 138441071458112 linears.py:541] linears/x Logical: bfloat16[32,2048,7168]...................................... ('activation_batch', 'activation_length', 'activation_mlp').
I0424 15:08:01.269258 138441071458112 linears.py:541] linears/x Physical: bfloat16[32,2048,7168]...................................... ('fsdp', None, None).
I0424 15:08:01.879811 138441071458112 checkpointing.py:578] checkpoint manager exists so trying to load this run's existing checkpoint
I0424 15:08:01.879922 138441071458112 checkpointing.py:676] No existing checkpoints found, not restoring checkpoint.
fsdp: 32
I0424 15:08:04.552958 138441071458112 maxtext_utils.py:1740]  params/params/decoder/decoder_norm/scale
    Shape:     float32[2048]
    Logical:   PartitionSpec('norm',)
    Physical:  (None,)
I0424 15:08:04.553099 138441071458112 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 15:08:04.553156 138441071458112 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 15:08:04.553215 138441071458112 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 15:08:04.553271 138441071458112 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 15:08:04.553312 138441071458112 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 15:08:04.553366 138441071458112 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 15:08:04.553420 138441071458112 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 15:08:04.553460 138441071458112 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 15:08:04.553497 138441071458112 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 15:08:04.553546 138441071458112 maxtext_utils.py:1740]  params/params/decoder/logits_dense/kernel
    Shape:     float32[2048,32000]
    Logical:   PartitionSpec('embed', 'vocab')
    Physical:  ('fsdp', None)
I0424 15:08:04.553594 138441071458112 maxtext_utils.py:1740]  params/params/token_embedder/embedding
    Shape:     float32[32000,2048]
    Logical:   PartitionSpec('vocab', 'embed')
    Physical:  (None, 'fsdp')

I0424 15:08:06.269335 138441071458112 train.py:158] train/xent Logical: float32[32,2048]............................................ ('activation_embed_and_logits_batch', 'activation_length').
I0424 15:08:06.269433 138441071458112 train.py:158] train/xent Physical: float32[32,2048]............................................ ('fsdp', None).
I0424 15:08:06.284370 138441071458112 train.py:165] train/z_loss Logical: float32[32,2048]............................................ ('activation_embed_and_logits_batch', 'activation_length').
I0424 15:08:06.284432 138441071458112 train.py:165] train/z_loss Physical: float32[32,2048]............................................ ('fsdp', None).
I0424 15:08:20.324487 138441071458112 max_utils.py:791] Total memory size: 1.8 GB, Output size: 0.4 GB, Temp size: 1.4 GB, Argument size: 0.4 GB, Host temp size: 0.0 GB.
I0424 15:08:20.325640 138441071458112 metric_logger.py:289] number parameters: 1.104 billion
I0424 15:08:35.277450 138441071458112 checkpointing.py:794] Waiting for step 0 to finish before checkpoint...
I0424 15:08:35.518733 138441071458112 checkpointing.py:798] Waited 0.24126195907592773 seconds for step 0 to finish before starting checkpointing.
I0424 15:08:35.521353 138441071458112 checkpoint_manager.py:2013] [process=5][thread=MainThread][wait_until_finished] No Save Finalize thread to wait for. Returning.
I0424 15:08:35.523319 138441071458112 checkpoint_manager.py:1518] [process=5] Saving checkpoint at step 0
I0424 15:08:35.526092 138441071458112 async_checkpointer.py:452] [process=5] 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_04_int8/checkpoints/0.
I0424 15:08:36.120328 138441071458112 signaling_client.py:364] Using JaxDistributedSignalingClient
I0424 15:08:36.121314 138441071458112 jax_array_handlers.py:358] Scheduling D2H of 39 prioritized jax.Array.
I0424 15:08:36.121373 138441071458112 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 15:08:36.372306 138441071458112 base_pytree_checkpoint_handler.py:153] [process=5][thread=MainThread] Initiated "orbax.checkpoint._src.serialization.jax_array_handlers.ArrayHandler".serialize. Time taken: 0.252059s
I0424 15:08:36.372472 138441071458112 base_pytree_checkpoint_handler.py:129] [process=5] /jax/checkpoint/write/blocking_gbytes_per_sec: 3.008 GiB/s (total gbytes: 1.5 GiB) (time elapsed: 0.5127687454223633 s) (per-host)
I0424 15:08:36.372526 138441071458112 base_pytree_checkpoint_handler.py:737] [process=5][thread=MainThread] Initiated Pytree async_save. Time taken: 0.512833s (batch_requests_ready=0.257065s, total_serialization_initiated=0.255697s, others=0.000071s)
I0424 15:08:36.372616 138441071458112 composite_checkpoint_handler.py:715] [process=5][thread=MainThread] Initiated CompositeCheckpointHandler.async_save. Time taken: 0.517352s (all_items=0.000020s, per_item={'items': '0.00002003'}, temp_paths=0.517332)
I0424 15:08:36.373535 138309087000320 async_checkpointer.py:79] [process=5][thread=async_save] Background save thread started.
I0424 15:08:36.373710 138441071458112 async_checkpointer.py:561] Finished blocking save. Time taken: 0.850318s. 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_04_int8/checkpoints/0.
I0424 15:08:36.408573 138441071458112 checkpoint_manager.py:1566] [process=5][thread=MainThread][step=0] Starting CheckpointManager Save Finalize thread=save_finalize
I0424 15:08:36.408899 138308499773184 async_checkpointer.py:265] [process=5][thread=save_finalize] Waiting for background save thread=async_save.
I0424 15:08:36.409047 138441071458112 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_04_int8/checkpoints', 'reached_preemption': False, 'preemption_received_at': None, 'synchronous': False, 'wait_for_prev_start_time': 1777043315.5213363, 'wait_for_prev_duration_secs': 6.389617919921875e-05, 'time_between_consecutive_saves_sec': None, 'checkpointer_blocking_start_time': 1777043315.523358, 'checkpointer_blocking_duration_secs': 0.8504576683044434, 'get_old_steps_start_time': 1777043316.3738356, 'get_old_steps_duration_secs': 2.5510787963867188e-05, 'checkpoint_manager_blocking_start_time': 1777043315.5193837, 'checkpoint_manager_blocking_duration_secs': 0.8896214962005615}
I0424 15:08:36.409155 138441071458112 checkpointing.py:409] Started an asynchronous checkpoint save for step 0
I0424 15:08:36.409207 138441071458112 max_utils.py:750] 
Memstats: After params initialized:
I0424 15:08:36.409268 138441071458112 max_utils.py:756] 	Using (GB) 0.43 / 31.25 (1.376000%) on TPU_18(process=5,(2,4,0,0))
I0424 15:08:36.409304 138441071458112 max_utils.py:756] 	Using (GB) 0.43 / 31.25 (1.376000%) on TPU_19(process=5,(3,4,0,0))
I0424 15:08:36.409332 138441071458112 max_utils.py:756] 	Using (GB) 0.43 / 31.25 (1.376000%) on TPU_22(process=5,(2,5,0,0))
I0424 15:08:36.409357 138441071458112 max_utils.py:756] 	Using (GB) 0.43 / 31.25 (1.376000%) on TPU_23(process=5,(3,5,0,0))
I0424 15:08:36.684440 138441071458112 metric_logger.py:185] completed step: 0, seconds: 14.952, TFLOP/s/device: 0.909, Tokens/s/device: 136.975, total_weights: 65536, loss: 10.874
I0424 15:08:36.940637 138441071458112 metric_logger.py:185] completed step: 1, seconds: 1.406, TFLOP/s/device: 9.666, Tokens/s/device: 1456.975, total_weights: 65536, loss: 10.874
I0424 15:08:37.367620 138441071458112 metric_logger.py:185] completed step: 2, seconds: 0.027, TFLOP/s/device: 499.766, Tokens/s/device: 75330.121, total_weights: 65536, loss: 10.262
I0424 15:08:37.597437 138441071458112 metric_logger.py:185] completed step: 3, seconds: 0.428, TFLOP/s/device: 31.767, Tokens/s/device: 4788.336, total_weights: 65536, loss: 9.731
I0424 15:08:38.057865 138441071458112 metric_logger.py:185] completed step: 4, seconds: 0.234, TFLOP/s/device: 57.998, Tokens/s/device: 8742.162, total_weights: 65536, loss: 9.272
I0424 15:08:38.063567 138441071458112 metric_logger.py:185] completed step: 5, seconds: 0.230, TFLOP/s/device: 59.132, Tokens/s/device: 8913.067, total_weights: 65536, loss: 8.887
I0424 15:09:00.199893 138441071458112 metric_logger.py:185] completed step: 6, seconds: 0.461, TFLOP/s/device: 29.488, Tokens/s/device: 4444.811, total_weights: 65536, loss: 8.588
I0424 15:09:00.429645 138441071458112 metric_logger.py:185] completed step: 7, seconds: 21.908, TFLOP/s/device: 0.620, Tokens/s/device: 93.481, total_weights: 65536, loss: 8.380
I0424 15:09:00.660099 138441071458112 metric_logger.py:185] completed step: 8, seconds: 0.234, TFLOP/s/device: 58.156, Tokens/s/device: 8765.922, total_weights: 65536, loss: 8.251
I0424 15:09:00.890448 138441071458112 checkpointing.py:794] Waiting for step 9 to finish before checkpoint...
I0424 15:09:00.891226 138441071458112 checkpointing.py:798] Waited 0.0007948875427246094 seconds for step 9 to finish before starting checkpointing.
I0424 15:09:00.893185 138441071458112 checkpoint_manager.py:2024] [process=5][thread=MainThread][step=0][wait_until_finished] Waiting for Save Finalize thread (save_finalize) to complete.
I0424 15:09:00.974881    2789 google_auth_provider.cc:181] Running on GCE, using service account 562977990677-compute@developer.gserviceaccount.com
I0424 15:09:02.904927 138310641940224 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_20260424_145744/linen_xpk_feat_nnx_set_defaults_true_20260424_145744_04_int8/checkpoints/0/items/array_metadatas/process_5
I0424 15:09:31.754297 138309087000320 base_pytree_checkpoint_handler.py:129] [process=5] /jax/checkpoint/write/gbytes_per_sec: 28.260 MiB/s (total gbytes: 1.5 GiB) (time elapsed: 55.894556522369385 s) (per-host)
I0424 15:09:31.754400 138309087000320 async_checkpointer.py:90] [process=5][thread=async_save] 3 Handler Commit operations completed. Time taken: 55.380758s.
I0424 15:09:39.900014 138309087000320 async_checkpointer.py:144] [process=5][thread=async_save] Background save thread done. Time taken: 63.526357s.
I0424 15:09:39.900264 138308499773184 async_checkpointer.py:273] [process=5][thread=save_finalize] Done with waiting for background save thread=async_save.
I0424 15:09:39.900330 138308499773184 async_checkpointer.py:283] [process=5][thread=save_finalize] No errors found in background save thread=async_save.
I0424 15:09:39.900377 138308499773184 checkpoint_manager.py:2133] [process=5][thread=save_finalize][step=0] CheckpointManager Save Finalize is syncing with other hosts...
I0424 15:09:39.902172 138308499773184 checkpoint_manager.py:2142] [process=5][thread=save_finalize][step=0] CheckpointManager Save Finalize is done on all hosts.
I0424 15:09:39.902390 138441071458112 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.
W0424 15:09:39.902538 138441071458112 checkpoint_manager.py:1458] Waiting for previous save to complete took 39.009356 seconds. If this number is high, consider checkpointing less frequently.
I0424 15:09:39.904324 138441071458112 checkpoint_manager.py:1518] [process=5] Saving checkpoint at step 9
I0424 15:09:39.907697 138441071458112 async_checkpointer.py:452] [process=5] 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_04_int8/checkpoints/9.
I0424 15:09:40.827411 138441071458112 jax_array_handlers.py:358] Scheduling D2H of 39 prioritized jax.Array.
I0424 15:09:40.827513 138441071458112 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 15:09:40.873517 138441071458112 base_pytree_checkpoint_handler.py:153] [process=5][thread=MainThread] Initiated "orbax.checkpoint._src.serialization.jax_array_handlers.ArrayHandler".serialize. Time taken: 0.047208s
I0424 15:09:40.873693 138441071458112 base_pytree_checkpoint_handler.py:129] [process=5] /jax/checkpoint/write/blocking_gbytes_per_sec: 4.928 GiB/s (total gbytes: 1.5 GiB) (time elapsed: 0.31302690505981445 s) (per-host)
I0424 15:09:40.873743 138441071458112 base_pytree_checkpoint_handler.py:737] [process=5][thread=MainThread] Initiated Pytree async_save. Time taken: 0.313086s (batch_requests_ready=0.263946s, total_serialization_initiated=0.049075s, others=0.000065s)
I0424 15:09:40.873831 138441071458112 composite_checkpoint_handler.py:715] [process=5][thread=MainThread] Initiated CompositeCheckpointHandler.async_save. Time taken: 0.317068s (all_items=0.000014s, per_item={'items': '0.00001431'}, temp_paths=0.317054)
I0424 15:09:40.874644 138311754569472 async_checkpointer.py:79] [process=5][thread=async_save] Background save thread started.
I0424 15:09:40.874822 138441071458112 async_checkpointer.py:561] Finished blocking save. Time taken: 0.970422s. 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_04_int8/checkpoints/9.
I0424 15:09:40.876674 138441071458112 checkpoint_manager.py:1566] [process=5][thread=MainThread][step=9] Starting CheckpointManager Save Finalize thread=save_finalize
I0424 15:09:40.876961 138308499773184 async_checkpointer.py:265] [process=5][thread=save_finalize] Waiting for background save thread=async_save.
I0424 15:09:40.877104 138441071458112 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_04_int8/checkpoints', 'reached_preemption': False, 'preemption_received_at': None, 'synchronous': False, 'wait_for_prev_start_time': 1777043340.893156, 'wait_for_prev_duration_secs': 39.0093560218811, 'time_between_consecutive_saves_sec': None, 'checkpointer_blocking_start_time': 1777043379.9043646, 'checkpointer_blocking_duration_secs': 0.9705696105957031, 'get_old_steps_start_time': 1777043380.8749518, 'get_old_steps_duration_secs': 3.147125244140625e-05, 'checkpoint_manager_blocking_start_time': 1777043340.8914957, 'checkpoint_manager_blocking_duration_secs': 39.98557114601135}
I0424 15:09:40.877275 138441071458112 checkpointing.py:409] Started an asynchronous checkpoint save for step 9
I0424 15:09:40.877330 138441071458112 checkpoint_manager.py:2024] [process=5][thread=MainThread][step=9][wait_until_finished] Waiting for Save Finalize thread (save_finalize) to complete.
I0424 15:09:45.959035 138311762962176 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_20260424_145744/linen_xpk_feat_nnx_set_defaults_true_20260424_145744_04_int8/checkpoints/9/items/array_metadatas/process_5
I0424 15:10:22.741797 138311754569472 base_pytree_checkpoint_handler.py:129] [process=5] /jax/checkpoint/write/gbytes_per_sec: 37.448 MiB/s (total gbytes: 1.5 GiB) (time elapsed: 42.18108677864075 s) (per-host)
I0424 15:10:22.741913 138311754569472 async_checkpointer.py:90] [process=5][thread=async_save] 3 Handler Commit operations completed. Time taken: 41.867144s.
I0424 15:10:30.683640 138311754569472 async_checkpointer.py:144] [process=5][thread=async_save] Background save thread done. Time taken: 49.808854s.
I0424 15:10:30.683939 138308499773184 async_checkpointer.py:273] [process=5][thread=save_finalize] Done with waiting for background save thread=async_save.
I0424 15:10:30.684066 138308499773184 async_checkpointer.py:283] [process=5][thread=save_finalize] No errors found in background save thread=async_save.
I0424 15:10:30.684114 138308499773184 checkpoint_manager.py:2133] [process=5][thread=save_finalize][step=9] CheckpointManager Save Finalize is syncing with other hosts...
I0424 15:10:30.685683 138308499773184 checkpoint_manager.py:2142] [process=5][thread=save_finalize][step=9] CheckpointManager Save Finalize is done on all hosts.
I0424 15:10:30.685877 138441071458112 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.
I0424 15:10:30.686019 138441071458112 checkpoint_manager.py:2013] [process=5][thread=MainThread][wait_until_finished] No Save Finalize thread to wait for. Returning.
I0424 15:10:30.686745 138441071458112 metric_logger.py:185] completed step: 9, seconds: 0.230, TFLOP/s/device: 59.170, Tokens/s/device: 8918.734, total_weights: 65536, loss: 8.175
Per train step:
 Total TFLOPs: 13.59 
 split as 93.93% learnable weight flops and 6.07% attention flops
XPK End: Fri Apr 24 15:10:44 UTC 2026
EXIT_CODE=0