MaxView

‹ 03_dropoutCase: 04_int805_fp8 ›

Metrics: Linen vs NNX  ·  test/pipeline-scan-nnx

MetricLinen  5b21b66fcNNX  5b21b66fcDiff (NNX − Linen)
Parameters1.104 billion
Final loss8.1750
TFLOP/s59.195
Tok/s8922.6
Avg s/step3.595
Memory %1.38
JAX0.9.20.9.2

Diff = NNX value − Linen value. Green = NNX improved. Red = NNX regressed.

Linen  ·  5b21b66fc  ·  test_pipeline_scan_nnx_20260423_211617  ·  full log
XPK Start: Thu Apr 23 21:25:23 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. 
2026-04-23 21:25:47.962851: 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)
I0423 21:25:48.172762 138075086112576 max_utils.py:273] Attempting to initialize the jax distributed system...
I0423 21:25:57.213625 138075086112576 distributed.py:149] Starting JAX distributed service on [::]:8482
I0423 21:25:57.215944 138075086112576 distributed.py:172] Connecting to JAX distributed service on mt-04-int8-4elo3-slice-job-0-0.mt-04-int8-4elo3:8482
I0423 21:25:58.757200 138075086112576 max_utils.py:284] Jax distributed system initialized!
I0423 21:26:04.139388 138075086112576 max_utils.py:800] System Information: Jax Version: 0.9.2
I0423 21:26:04.139496 138075086112576 max_utils.py:801] System Information: Jaxlib Version: 0.9.2
I0423 21:26:04.139536 138075086112576 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
I0423 21:26:04.139573 138075086112576 train_utils.py:361] WARNING: Sequence packing is essentially ignored for synthetic data. Please use a real dataset to use sequence packing.
I0423 21:26:04.824767 138075086112576 maxtext_utils.py:1565] Num_devices: 32, shape (1, 1, 1, 32, 1, 1, 1, 1, 1, 1, 1, 1, 1)
I0423 21:26:04.825167 138075086112576 checkpointing.py:677] Setting up checkpoint logger...
I0423 21:26:04.825226 138075086112576 checkpointing.py:233] Creating checkpoint manager with ocdbt=True and zarr3=True
I0423 21:26:04.825271 138075086112576 pytree_checkpoint_handler.py:592] save_device_host_concurrent_bytes=None
I0423 21:26:04.825603 138075086112576 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 0x7d9379714440>, enable_pinned_host_transfer=False, save_concurrent_bytes: 96000000000 (89.4 GiB), restore_concurrent_bytes: 96000000000 (89.4 GiB)
I0423 21:26:07.654077 138075086112576 checkpointing.py:265] Enabling policy for fixed interval checkpointing.
I0423 21:26:07.654319 138075086112576 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 0x7d7fc453bb00>}, handler_registry=None
I0423 21:26:07.654561 138075086112576 composite_checkpoint_handler.py:237] Deferred registration for item: "items". Adding handler `<orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeCheckpointHandler object at 0x7d7fc453bb00>` 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`.
I0423 21:26:07.654610 138075086112576 composite_checkpoint_handler.py:237] Deferred registration for item: "metrics". Adding handler `<orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonCheckpointHandler object at 0x7d7fc453fad0>` 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`.
I0423 21:26:07.654645 138075086112576 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 0x7d7fc453bb00>, ('items', <class 'orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeRestoreArgs'>): <orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeCheckpointHandler object at 0x7d7fc453bb00>, ('metrics', <class 'orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonSaveArgs'>): <orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonCheckpointHandler object at 0x7d7fc453fad0>, ('metrics', <class 'orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonRestoreArgs'>): <orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonCheckpointHandler object at 0x7d7fc453fad0>}).
I0423 21:26:07.654988 138075086112576 abstract_checkpointer.py:35] orbax-checkpoint version: 0.11.34
I0423 21:26:07.655058 138075086112576 async_checkpointer.py:192] [process=5][thread=MainThread] Using barrier_sync_fn: <function get_barrier_sync_fn.<locals>._fn at 0x7d7fc4387560> timeout: 1200 secs and primary_host=0 for async checkpoint writes
I0423 21:26:08.522958 138075086112576 checkpoint_manager.py:1812] Found 0 checkpoint steps in gs://lance-maxtext/linen_ckpt_xpk_test_pipeline_scan_nnx_20260423_211617/linen_xpk_test_pipeline_scan_nnx_20260423_211617_04_int8/checkpoints
I0423 21:26:08.545735 138075086112576 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_test_pipeline_scan_nnx_20260423_211617/linen_xpk_test_pipeline_scan_nnx_20260423_211617_04_int8/checkpoints: <orbax.checkpoint.checkpoint_manager.CheckpointManager object at 0x7d7fc46a10d0>
I0423 21:26:08.545862 138075086112576 checkpointing.py:301] Checkpoint manager created!
I0423 21:26:09.685781 138075086112576 nnx_wrappers.py:453] Unknown Logical: bfloat16[32,2048,2048]...................................... ('activation_batch', 'activation_norm_length', 'activation_embed').
I0423 21:26:09.685901 138075086112576 nnx_wrappers.py:453] Unknown Physical: bfloat16[32,2048,2048]...................................... ('fsdp', None, None).
I0423 21:26:10.071062 138075086112576 attentions.py:1088] attentions/inputs_q Logical: bfloat16[32,2048,2048]...................................... ('activation_batch', 'activation_attn_length', 'activation_attn_embed').
I0423 21:26:10.071156 138075086112576 attentions.py:1088] attentions/inputs_q Physical: bfloat16[32,2048,2048]...................................... ('fsdp', None, None).
I0423 21:26:10.087564 138075086112576 attentions.py:1089] attentions/inputs_kv Logical: bfloat16[32,2048,2048]...................................... ('activation_batch', 'activation_attn_length', 'activation_attn_embed').
I0423 21:26:10.087625 138075086112576 attentions.py:1089] attentions/inputs_kv Physical: bfloat16[32,2048,2048]...................................... ('fsdp', None, None).
I0423 21:26:10.154425 138075086112576 attentions.py:1154] attentions/query Logical: bfloat16[32,2048,16,128].................................... ('activation_kv_batch', 'activation_attn_length', 'activation_kv_heads', 'activation_kv_head_dim').
I0423 21:26:10.154515 138075086112576 attentions.py:1154] attentions/query Physical: bfloat16[32,2048,16,128].................................... ('fsdp', None, None, None).
I0423 21:26:10.170977 138075086112576 attentions.py:1155] attentions/key Logical: bfloat16[32,2048,16,128].................................... ('activation_kv_batch', 'activation_attn_length', 'activation_kv_heads', 'activation_kv_head_dim').
I0423 21:26:10.171047 138075086112576 attentions.py:1155] attentions/key Physical: bfloat16[32,2048,16,128].................................... ('fsdp', None, None, None).
I0423 21:26:10.187597 138075086112576 attentions.py:1156] attentions/value Logical: bfloat16[32,2048,16,128].................................... ('activation_kv_batch', 'activation_attn_length', 'activation_kv_heads', 'activation_kv_head_dim').
I0423 21:26:10.187669 138075086112576 attentions.py:1156] attentions/value Physical: bfloat16[32,2048,16,128].................................... ('fsdp', None, None, None).
I0423 21:26:10.213052 138075086112576 attentions.py:1198] attentions/out Logical: bfloat16[32,2048,16,128].................................... ('activation_batch', 'activation_attn_length', 'activation_heads', 'activation_kv').
I0423 21:26:10.213123 138075086112576 attentions.py:1198] attentions/out Physical: bfloat16[32,2048,16,128].................................... ('fsdp', None, None, None).
I0423 21:26:10.279516 138075086112576 linears.py:525] linears/x Logical: bfloat16[32,2048,7168]...................................... ('activation_batch', 'activation_length', 'activation_mlp').
I0423 21:26:10.279603 138075086112576 linears.py:525] linears/x Physical: bfloat16[32,2048,7168]...................................... ('fsdp', None, None).
I0423 21:26:10.789018 138075086112576 checkpointing.py:577] checkpoint manager exists so trying to load this run's existing checkpoint
I0423 21:26:10.789130 138075086112576 checkpointing.py:665] No existing checkpoints found, not restoring checkpoint.
fsdp: 32
I0423 21:26:13.270006 138075086112576 maxtext_utils.py:1668]  params/params/decoder/decoder_norm/scale
    Shape:     float32[2048]
    Logical:   P('norm',)
    Physical:  (None,)
I0423 21:26:13.270135 138075086112576 maxtext_utils.py:1668]  params/params/decoder/layers/mlp/wi_0/kernel
    Shape:     float32[2048,16,7168]
    Logical:   P('embed', 'layers', 'mlp')
    Physical:  ('fsdp', None, None)
I0423 21:26:13.270194 138075086112576 maxtext_utils.py:1668]  params/params/decoder/layers/mlp/wi_1/kernel
    Shape:     float32[2048,16,7168]
    Logical:   P('embed', 'layers', 'mlp')
    Physical:  ('fsdp', None, None)
I0423 21:26:13.270252 138075086112576 maxtext_utils.py:1668]  params/params/decoder/layers/mlp/wo/kernel
    Shape:     float32[7168,16,2048]
    Logical:   P('mlp', 'layers', 'embed')
    Physical:  (None, None, 'fsdp')
I0423 21:26:13.270304 138075086112576 maxtext_utils.py:1668]  params/params/decoder/layers/post_self_attention_layer_norm/scale
    Shape:     float32[2048,16]
    Logical:   P('norm', 'layers')
    Physical:  (None, None)
I0423 21:26:13.270343 138075086112576 maxtext_utils.py:1668]  params/params/decoder/layers/pre_self_attention_layer_norm/scale
    Shape:     float32[2048,16]
    Logical:   P('norm', 'layers')
    Physical:  (None, None)
I0423 21:26:13.270396 138075086112576 maxtext_utils.py:1668]  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)
I0423 21:26:13.270449 138075086112576 maxtext_utils.py:1668]  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')
I0423 21:26:13.270490 138075086112576 maxtext_utils.py:1668]  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)
I0423 21:26:13.270525 138075086112576 maxtext_utils.py:1668]  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)
I0423 21:26:13.270573 138075086112576 maxtext_utils.py:1668]  params/params/decoder/logits_dense/kernel
    Shape:     float32[2048,32000]
    Logical:   P('embed_vocab', 'vocab')
    Physical:  ('fsdp', None)
I0423 21:26:13.270619 138075086112576 maxtext_utils.py:1668]  params/params/token_embedder/embedding

    Shape:     float32[32000,2048]
    Logical:   P('vocab', 'embed_vocab')
    Physical:  (None, 'fsdp')
I0423 21:26:14.662225 138075086112576 train.py:155] train/xent Logical: float32[32,2048]............................................ ('activation_embed_and_logits_batch', 'activation_length').
I0423 21:26:14.662321 138075086112576 train.py:155] train/xent Physical: float32[32,2048]............................................ ('fsdp', None).
I0423 21:26:14.677966 138075086112576 train.py:162] train/z_loss Logical: float32[32,2048]............................................ ('activation_embed_and_logits_batch', 'activation_length').
I0423 21:26:14.678028 138075086112576 train.py:162] train/z_loss Physical: float32[32,2048]............................................ ('fsdp', None).
I0423 21:26:28.612636 138075086112576 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.
I0423 21:26:28.613417 138075086112576 metric_logger.py:301] number parameters: 1.104 billion
I0423 21:26:43.270348 138075086112576 checkpointing.py:772] Waiting for step 0 to finish before checkpoint...
I0423 21:26:43.510987 138075086112576 checkpointing.py:776] Waited 0.24061059951782227 seconds for step 0 to finish before starting checkpointing.
I0423 21:26:43.513442 138075086112576 checkpoint_manager.py:2009] [process=5][thread=MainThread][wait_until_finished] No Save Finalize thread to wait for. Returning.
I0423 21:26:43.515607 138075086112576 checkpoint_manager.py:1512] [process=5] Saving checkpoint at step 0
I0423 21:26:43.517143 138075086112576 event_tracking.py:70] [process=5] [async] Started save checkpoint @ gs://lance-maxtext/linen_ckpt_xpk_test_pipeline_scan_nnx_20260423_211617/linen_xpk_test_pipeline_scan_nnx_20260423_211617_04_int8/checkpoints/0.
I0423 21:26:43.846317 138075086112576 signaling_client.py:364] Using JaxDistributedSignalingClient
I0423 21:26:43.847161 138075086112576 jax_array_handlers.py:360] Scheduling D2H of 39 prioritized jax.Array.
I0423 21:26:43.847218 138075086112576 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
I0423 21:26:44.124800 138075086112576 base_pytree_checkpoint_handler.py:154] [process=5][thread=MainThread] Initiated "orbax.checkpoint._src.serialization.jax_array_handlers.ArrayHandler".serialize. Time taken: 0.278556s
I0423 21:26:44.124979 138075086112576 base_pytree_checkpoint_handler.py:130] [process=5] /jax/orbax/write/blocking_gbytes_per_sec: 5.427 GiB/s (total gbytes: 1.5 GiB) (time elapsed: 0.2842409610748291 s) (per-host)
I0423 21:26:44.125038 138075086112576 base_pytree_checkpoint_handler.py:768] [process=5][thread=MainThread] Initiated Pytree async_save. Time taken: 0.284310s (batch_requests_ready=0.002248s, total_serialization_initiated=0.281986s, others=0.000077s)
I0423 21:26:44.125138 138075086112576 composite_checkpoint_handler.py:715] [process=5][thread=MainThread] Initiated CompositeCheckpointHandler.async_save. Time taken: 0.288577s (all_items=0.000018s, per_item={'items': '0.00001812'}, temp_paths=0.288559)
I0423 21:26:44.125934 138075086112576 event_tracking.py:125] [process=5] [async] Finished blocking save in 0.61 seconds. Continuing save @ gs://lance-maxtext/linen_ckpt_xpk_test_pipeline_scan_nnx_20260423_211617/linen_xpk_test_pipeline_scan_nnx_20260423_211617_04_int8/checkpoints/0.
I0423 21:26:44.126277 137944939026176 async_checkpointer.py:76] [process=5][thread=async_save] Background save thread started. Deadline for this save operation is 2026-04-23 21:46:44.126234
I0423 21:26:44.142538 138075086112576 checkpoint_manager.py:1560] [process=5][thread=MainThread][step=0] Starting CheckpointManager Save Finalize thread=save_finalize
I0423 21:26:44.142899 137943305934592 async_checkpointer.py:280] [process=5][thread=save_finalize] Waiting for background save thread=async_save.
I0423 21:26:44.143067 138075086112576 standard_logger.py:34] {'step': 0, 'event_type': 'save', 'directory': 'gs://lance-maxtext/linen_ckpt_xpk_test_pipeline_scan_nnx_20260423_211617/linen_xpk_test_pipeline_scan_nnx_20260423_211617_04_int8/checkpoints', 'reached_preemption': False, 'preemption_received_at': None, 'synchronous': False, 'wait_for_prev_start_time': 1776979603.513424, 'wait_for_prev_duration_secs': 6.222724914550781e-05, 'time_between_consecutive_saves_sec': None, 'checkpointer_blocking_start_time': 1776979603.515647, 'checkpointer_blocking_duration_secs': 0.6107809543609619, 'get_old_steps_start_time': 1776979604.126452, 'get_old_steps_duration_secs': 2.9087066650390625e-05, 'checkpoint_manager_blocking_start_time': 1776979603.511483, 'checkpoint_manager_blocking_duration_secs': 0.6315464973449707}
I0423 21:26:44.143177 138075086112576 checkpointing.py:408] Started an asynchronous checkpoint save for step 0
I0423 21:26:44.143229 138075086112576 max_utils.py:750] 
Memstats: After params initialized:
I0423 21:26:44.143280 138075086112576 max_utils.py:756] 	Using (GB) 0.44 / 31.25 (1.408000%) on TPU_18(process=5,(2,4,0,0))
I0423 21:26:44.143314 138075086112576 max_utils.py:756] 	Using (GB) 0.43 / 31.25 (1.376000%) on TPU_19(process=5,(3,4,0,0))
I0423 21:26:44.143341 138075086112576 max_utils.py:756] 	Using (GB) 0.43 / 31.25 (1.376000%) on TPU_22(process=5,(2,5,0,0))
I0423 21:26:44.143368 138075086112576 max_utils.py:756] 	Using (GB) 0.43 / 31.25 (1.376000%) on TPU_23(process=5,(3,5,0,0))
I0423 21:26:44.466888 138075086112576 metric_logger.py:196] completed step: 0, seconds: 14.657, TFLOP/s/device: 0.927, Tokens/s/device: 139.730, total_weights: 65536, loss: 10.874, lm_loss: 10.874, perplexity: 52796.277
I0423 21:26:44.715673 138075086112576 metric_logger.py:196] completed step: 1, seconds: 1.195, TFLOP/s/device: 11.370, Tokens/s/device: 1713.818, total_weights: 65536, loss: 10.874, lm_loss: 10.874, perplexity: 52796.277
I0423 21:26:45.146034 138075086112576 metric_logger.py:196] completed step: 2, seconds: 0.020, TFLOP/s/device: 680.002, Tokens/s/device: 102497.373, total_weights: 65536, loss: 10.262, lm_loss: 10.262, perplexity: 28634.820
I0423 21:26:45.375512 138075086112576 metric_logger.py:196] completed step: 3, seconds: 0.431, TFLOP/s/device: 31.555, Tokens/s/device: 4756.353, total_weights: 65536, loss: 9.731, lm_loss: 9.731, perplexity: 16827.199
I0423 21:26:45.835005 138075086112576 metric_logger.py:196] completed step: 4, seconds: 0.235, TFLOP/s/device: 57.920, Tokens/s/device: 8730.348, total_weights: 65536, loss: 9.272, lm_loss: 9.272, perplexity: 10638.531
I0423 21:26:45.841083 138075086112576 metric_logger.py:196] completed step: 5, seconds: 0.230, TFLOP/s/device: 59.192, Tokens/s/device: 8922.076, total_weights: 65536, loss: 8.887, lm_loss: 8.887, perplexity: 7233.941
I0423 21:26:47.210594    2885 google_auth_provider.cc:181] Running on GCE, using service account 562977990677-compute@developer.gserviceaccount.com
I0423 21:26:49.750532 137943314327296 array_metadata_store.py:203] [process=5][thread=array_type_handler] Wrote 39 array_metadata.ArrayMetadata to gs://lance-maxtext/linen_ckpt_xpk_test_pipeline_scan_nnx_20260423_211617/linen_xpk_test_pipeline_scan_nnx_20260423_211617_04_int8/checkpoints/0/items/array_metadatas/process_5
I0423 21:27:15.390110 138075086112576 metric_logger.py:196] completed step: 6, seconds: 0.460, TFLOP/s/device: 29.532, Tokens/s/device: 4451.448, total_weights: 65536, loss: 8.588, lm_loss: 8.588, perplexity: 5365.468
I0423 21:27:15.619760 138075086112576 metric_logger.py:196] completed step: 7, seconds: 29.321, TFLOP/s/device: 0.463, Tokens/s/device: 69.848, total_weights: 65536, loss: 8.380, lm_loss: 8.380, perplexity: 4358.779
I0423 21:27:15.851735 138075086112576 metric_logger.py:196] completed step: 8, seconds: 0.234, TFLOP/s/device: 58.062, Tokens/s/device: 8751.725, total_weights: 65536, loss: 8.251, lm_loss: 8.251, perplexity: 3831.453
I0423 21:27:16.081530 138075086112576 checkpointing.py:772] Waiting for step 9 to finish before checkpoint...
I0423 21:27:16.082215 138075086112576 checkpointing.py:776] Waited 0.0007083415985107422 seconds for step 9 to finish before starting checkpointing.
I0423 21:27:16.084279 138075086112576 checkpoint_manager.py:2020] [process=5][thread=MainThread][step=0][wait_until_finished] Waiting for Save Finalize thread (save_finalize) to complete.
I0423 21:27:17.668262 137944939026176 base_pytree_checkpoint_handler.py:130] [process=5] /jax/orbax/write/gbytes_per_sec: 46.696 MiB/s (total gbytes: 1.5 GiB) (time elapsed: 33.82748532295227 s) (per-host)
I0423 21:27:17.668387 137944939026176 async_checkpointer.py:90] [process=5][thread=async_save] 3 Handler Commit operations completed. Time taken: 33.541996s.
I0423 21:27:25.828265 137944939026176 async_checkpointer.py:160] [process=5][thread=async_save] Background save thread done. Time taken: 41.701859s.
I0423 21:27:25.828564 137943305934592 async_checkpointer.py:288] [process=5][thread=save_finalize] Done with waiting for background save thread=async_save.
I0423 21:27:25.828734 137943305934592 async_checkpointer.py:298] [process=5][thread=save_finalize] No errors found in background save thread=async_save.
I0423 21:27:25.829017 137943305934592 checkpoint_manager.py:2137] [process=5][thread=save_finalize][step=0] CheckpointManager Save Finalize is syncing with other hosts...
I0423 21:27:25.831714 137943305934592 checkpoint_manager.py:2146] [process=5][thread=save_finalize][step=0] CheckpointManager Save Finalize is done on all hosts.
I0423 21:27:25.831888 138075086112576 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.
W0423 21:27:25.832027 138075086112576 checkpoint_manager.py:1452] Waiting for previous save to complete took 9.747744 seconds. If this number is high, consider checkpointing less frequently.
I0423 21:27:25.833879 138075086112576 checkpoint_manager.py:1512] [process=5] Saving checkpoint at step 9
I0423 21:27:25.835880 138075086112576 event_tracking.py:70] [process=5] [async] Started save checkpoint @ gs://lance-maxtext/linen_ckpt_xpk_test_pipeline_scan_nnx_20260423_211617/linen_xpk_test_pipeline_scan_nnx_20260423_211617_04_int8/checkpoints/9.
I0423 21:27:26.121724 138075086112576 jax_array_handlers.py:360] Scheduling D2H of 39 prioritized jax.Array.
I0423 21:27:26.121818 138075086112576 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
I0423 21:27:26.160848 138075086112576 base_pytree_checkpoint_handler.py:154] [process=5][thread=MainThread] Initiated "orbax.checkpoint._src.serialization.jax_array_handlers.ArrayHandler".serialize. Time taken: 0.040151s
I0423 21:27:26.161018 138075086112576 base_pytree_checkpoint_handler.py:130] [process=5] /jax/orbax/write/blocking_gbytes_per_sec: 35.361 GiB/s (total gbytes: 1.5 GiB) (time elapsed: 0.043623924255371094 s) (per-host)
I0423 21:27:26.161066 138075086112576 base_pytree_checkpoint_handler.py:768] [process=5][thread=MainThread] Initiated Pytree async_save. Time taken: 0.043681s (batch_requests_ready=0.001812s, total_serialization_initiated=0.041805s, others=0.000065s)
I0423 21:27:26.161168 138075086112576 composite_checkpoint_handler.py:715] [process=5][thread=MainThread] Initiated CompositeCheckpointHandler.async_save. Time taken: 0.047816s (all_items=0.000016s, per_item={'items': '0.00001621'}, temp_paths=0.047799)
I0423 21:27:26.161924 138075086112576 event_tracking.py:125] [process=5] [async] Finished blocking save in 0.33 seconds. Continuing save @ gs://lance-maxtext/linen_ckpt_xpk_test_pipeline_scan_nnx_20260423_211617/linen_xpk_test_pipeline_scan_nnx_20260423_211617_04_int8/checkpoints/9.
I0423 21:27:26.162292 137943305934592 async_checkpointer.py:76] [process=5][thread=async_save] Background save thread started. Deadline for this save operation is 2026-04-23 21:47:26.162250
I0423 21:27:26.164249 138075086112576 checkpoint_manager.py:1560] [process=5][thread=MainThread][step=9] Starting CheckpointManager Save Finalize thread=save_finalize
I0423 21:27:26.164490 137943314327296 async_checkpointer.py:280] [process=5][thread=save_finalize] Waiting for background save thread=async_save.
I0423 21:27:26.164624 138075086112576 standard_logger.py:34] {'step': 9, 'event_type': 'save', 'directory': 'gs://lance-maxtext/linen_ckpt_xpk_test_pipeline_scan_nnx_20260423_211617/linen_xpk_test_pipeline_scan_nnx_20260423_211617_04_int8/checkpoints', 'reached_preemption': False, 'preemption_received_at': None, 'synchronous': False, 'wait_for_prev_start_time': 1776979636.0842493, 'wait_for_prev_duration_secs': 9.74774432182312, 'time_between_consecutive_saves_sec': None, 'checkpointer_blocking_start_time': 1776979645.8339193, 'checkpointer_blocking_duration_secs': 0.3285229206085205, 'get_old_steps_start_time': 1776979646.162465, 'get_old_steps_duration_secs': 3.1948089599609375e-05, 'checkpoint_manager_blocking_start_time': 1776979636.0824423, 'checkpoint_manager_blocking_duration_secs': 10.082147359848022}
I0423 21:27:26.164799 138075086112576 checkpointing.py:408] Started an asynchronous checkpoint save for step 9
I0423 21:27:26.164847 138075086112576 checkpoint_manager.py:2020] [process=5][thread=MainThread][step=9][wait_until_finished] Waiting for Save Finalize thread (save_finalize) to complete.
I0423 21:27:31.488003 137926509192960 array_metadata_store.py:203] [process=5][thread=array_type_handler] Wrote 39 array_metadata.ArrayMetadata to gs://lance-maxtext/linen_ckpt_xpk_test_pipeline_scan_nnx_20260423_211617/linen_xpk_test_pipeline_scan_nnx_20260423_211617_04_int8/checkpoints/9/items/array_metadatas/process_5
I0423 21:28:08.169352 137943305934592 base_pytree_checkpoint_handler.py:130] [process=5] /jax/orbax/write/gbytes_per_sec: 37.563 MiB/s (total gbytes: 1.5 GiB) (time elapsed: 42.05191659927368 s) (per-host)
I0423 21:28:08.169474 137943305934592 async_checkpointer.py:90] [process=5][thread=async_save] 3 Handler Commit operations completed. Time taken: 42.007066s.
I0423 21:28:15.870336 137943305934592 async_checkpointer.py:160] [process=5][thread=async_save] Background save thread done. Time taken: 49.707913s.
I0423 21:28:15.870624 137943314327296 async_checkpointer.py:288] [process=5][thread=save_finalize] Done with waiting for background save thread=async_save.
I0423 21:28:15.870765 137943314327296 async_checkpointer.py:298] [process=5][thread=save_finalize] No errors found in background save thread=async_save.
I0423 21:28:15.870814 137943314327296 checkpoint_manager.py:2137] [process=5][thread=save_finalize][step=9] CheckpointManager Save Finalize is syncing with other hosts...
I0423 21:28:15.873324 137943314327296 checkpoint_manager.py:2146] [process=5][thread=save_finalize][step=9] CheckpointManager Save Finalize is done on all hosts.
I0423 21:28:15.873492 138075086112576 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.
I0423 21:28:15.873643 138075086112576 checkpoint_manager.py:2009] [process=5][thread=MainThread][wait_until_finished] No Save Finalize thread to wait for. Returning.
I0423 21:28:15.874551 138075086112576 metric_logger.py:196] completed step: 9, seconds: 0.230, TFLOP/s/device: 59.195, Tokens/s/device: 8922.581, 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: Thu Apr 23 21:28:28 UTC 2026
EXIT_CODE=0
NNX  ·  5b21b66fc  ·  test_pipeline_scan_nnx_20260423_211617  ·  full log
XPK Start: Thu Apr 23 22:51:39 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. 
2026-04-23 22:52:04.671313: 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)
I0423 22:52:04.881352 138425582925632 max_utils.py:273] Attempting to initialize the jax distributed system...
I0423 22:52:13.923433 138425582925632 distributed.py:149] Starting JAX distributed service on [::]:8482
I0423 22:52:13.925844 138425582925632 distributed.py:172] Connecting to JAX distributed service on mt-04-int8-jafjj-slice-job-0-0.mt-04-int8-jafjj:8482
I0423 22:52:15.101362 138425582925632 max_utils.py:284] Jax distributed system initialized!
I0423 22:52:21.045574 138425582925632 max_utils.py:800] System Information: Jax Version: 0.9.2
I0423 22:52:21.045682 138425582925632 max_utils.py:801] System Information: Jaxlib Version: 0.9.2
I0423 22:52:21.045728 138425582925632 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
I0423 22:52:21.045766 138425582925632 train_utils.py:361] WARNING: Sequence packing is essentially ignored for synthetic data. Please use a real dataset to use sequence packing.
Traceback (most recent call last):
  File "<frozen runpy>", line 198, in _run_module_as_main
  File "<frozen runpy>", line 88, in _run_code
  File "/deps/src/maxtext/trainers/pre_train/train.py", line 744, in <module>
    app.run(main)
  File "/usr/local/lib/python3.12/site-packages/absl/app.py", line 367, in run
    _run_main(main, args)
  File "/usr/local/lib/python3.12/site-packages/absl/app.py", line 312, in _run_main
    sys.exit(main(argv))
             ^^^^^^^^^^
  File "/deps/src/maxtext/trainers/pre_train/train.py", line 740, in main
    train_func()
  File "/deps/src/maxtext/trainers/pre_train/train.py", line 730, in train_func
    run(config, recorder, diagnostic_config)
  File "/deps/src/maxtext/trainers/pre_train/train.py", line 709, in run
    train_loop(config, recorder)
  File "/deps/src/maxtext/trainers/pre_train/train.py", line 536, in train_loop
    ) = train_utils.setup_train_loop(config, recorder)
        ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
  File "/deps/src/maxtext/utils/train_utils.py", line 218, in setup_train_loop
    raise NotImplementedError("Pure NNX support has not been implemented yet.")
NotImplementedError: Pure NNX support has not been implemented yet.
XPK End: Thu Apr 23 22:52:30 UTC 2026
EXIT_CODE=1