MaxView

‹ 08_checkpoint_async_true_saveCase: 09_pdb_lt_110_shardy_false ›

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

MetricLinen  5b21b66fcNNX  5b21b66fcDiff (NNX − Linen)
Parameters1.104 billion
Final loss5.8920
TFLOP/s58.230
Tok/s8777.0
Avg s/step3.348
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 21:47:46 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 21:48:10.941722: 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 21:48:11.151238 136048720987968 max_utils.py:273] Attempting to initialize the jax distributed system...
I0423 21:48:20.191963 136048720987968 distributed.py:149] Starting JAX distributed service on [::]:8482
I0423 21:48:20.194463 136048720987968 distributed.py:172] Connecting to JAX distributed service on mt-09-pdb-lt-1-f6yfr-slice-job-0-0.mt-09-pdb-lt-1-f6yfr:8482
I0423 21:48:21.283526 136048720987968 max_utils.py:284] Jax distributed system initialized!
I0423 21:48:27.432862 136048720987968 max_utils.py:800] System Information: Jax Version: 0.9.2
I0423 21:48:27.432972 136048720987968 max_utils.py:801] System Information: Jaxlib Version: 0.9.2
I0423 21:48:27.433027 136048720987968 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 21:48:27.433074 136048720987968 train_utils.py:361] WARNING: Sequence packing is essentially ignored for synthetic data. Please use a real dataset to use sequence packing.
I0423 21:48:28.133789 136048720987968 maxtext_utils.py:1565] Num_devices: 32, shape (1, 1, 1, 8, 1, 1, 1, 1, 4, 1, 1, 1, 1)
I0423 21:48:28.134069 136048720987968 checkpointing.py:677] Setting up checkpoint logger...
I0423 21:48:28.134124 136048720987968 checkpointing.py:233] Creating checkpoint manager with ocdbt=True and zarr3=True
I0423 21:48:28.134169 136048720987968 pytree_checkpoint_handler.py:592] save_device_host_concurrent_bytes=None
I0423 21:48:28.134523 136048720987968 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 0x7bbbb6b18d40>, enable_pinned_host_transfer=False, save_concurrent_bytes: 96000000000 (89.4 GiB), restore_concurrent_bytes: 96000000000 (89.4 GiB)
I0423 21:48:31.045355 136048720987968 checkpointing.py:265] Enabling policy for fixed interval checkpointing.
I0423 21:48:31.045593 136048720987968 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 0x7ba76c05cc50>}, handler_registry=None
I0423 21:48:31.045845 136048720987968 composite_checkpoint_handler.py:237] Deferred registration for item: "items". Adding handler `<orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeCheckpointHandler object at 0x7ba76c05cc50>` 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 21:48:31.045896 136048720987968 composite_checkpoint_handler.py:237] Deferred registration for item: "metrics". Adding handler `<orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonCheckpointHandler object at 0x7ba73c64da00>` 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 21:48:31.045932 136048720987968 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 0x7ba76c05cc50>, ('items', <class 'orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeRestoreArgs'>): <orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeCheckpointHandler object at 0x7ba76c05cc50>, ('metrics', <class 'orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonSaveArgs'>): <orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonCheckpointHandler object at 0x7ba73c64da00>, ('metrics', <class 'orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonRestoreArgs'>): <orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonCheckpointHandler object at 0x7ba73c64da00>}).
I0423 21:48:31.046314 136048720987968 abstract_checkpointer.py:35] orbax-checkpoint version: 0.11.34
I0423 21:48:31.046394 136048720987968 async_checkpointer.py:192] [process=5][thread=MainThread] Using barrier_sync_fn: <function get_barrier_sync_fn.<locals>._fn at 0x7ba72c227560> timeout: 1200 secs and primary_host=0 for async checkpoint writes
I0423 21:48:32.564320 136048720987968 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_09_pdb_lt_1/checkpoints
I0423 21:48:32.575150 136048720987968 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_09_pdb_lt_1/checkpoints: <orbax.checkpoint.checkpoint_manager.CheckpointManager object at 0x7ba76c0910a0>
I0423 21:48:32.575276 136048720987968 checkpointing.py:301] Checkpoint manager created!
I0423 21:48:33.504026 136048720987968 nnx_wrappers.py:453] Unknown Logical: bfloat16[8,2048,2048]....................................... ('activation_batch', 'activation_norm_length', 'activation_embed').
I0423 21:48:33.504142 136048720987968 nnx_wrappers.py:453] Unknown Physical: bfloat16[8,2048,2048]....................................... ('fsdp', None, 'tensor').
I0423 21:48:33.883614 136048720987968 attentions.py:1088] attentions/inputs_q Logical: bfloat16[8,2048,2048]....................................... ('activation_batch', 'activation_attn_length', 'activation_attn_embed').
I0423 21:48:33.883719 136048720987968 attentions.py:1088] attentions/inputs_q Physical: bfloat16[8,2048,2048]....................................... ('fsdp', None, 'tensor').
I0423 21:48:33.900256 136048720987968 attentions.py:1089] attentions/inputs_kv Logical: bfloat16[8,2048,2048]....................................... ('activation_batch', 'activation_attn_length', 'activation_attn_embed').
I0423 21:48:33.900313 136048720987968 attentions.py:1089] attentions/inputs_kv Physical: bfloat16[8,2048,2048]....................................... ('fsdp', None, 'tensor').
I0423 21:48:33.923917 136048720987968 attentions.py:1154] attentions/query Logical: bfloat16[8,2048,16,128]..................................... ('activation_kv_batch', 'activation_attn_length', 'activation_kv_heads', 'activation_kv_head_dim').
I0423 21:48:33.923986 136048720987968 attentions.py:1154] attentions/query Physical: bfloat16[8,2048,16,128]..................................... ('fsdp', None, 'tensor', None).
I0423 21:48:33.940489 136048720987968 attentions.py:1155] attentions/key Logical: bfloat16[8,2048,16,128]..................................... ('activation_kv_batch', 'activation_attn_length', 'activation_kv_heads', 'activation_kv_head_dim').
I0423 21:48:33.940557 136048720987968 attentions.py:1155] attentions/key Physical: bfloat16[8,2048,16,128]..................................... ('fsdp', None, 'tensor', None).
I0423 21:48:33.957019 136048720987968 attentions.py:1156] attentions/value Logical: bfloat16[8,2048,16,128]..................................... ('activation_kv_batch', 'activation_attn_length', 'activation_kv_heads', 'activation_kv_head_dim').
I0423 21:48:33.957081 136048720987968 attentions.py:1156] attentions/value Physical: bfloat16[8,2048,16,128]..................................... ('fsdp', None, 'tensor', None).
I0423 21:48:33.981984 136048720987968 attentions.py:1198] attentions/out Logical: bfloat16[8,2048,16,128]..................................... ('activation_batch', 'activation_attn_length', 'activation_heads', 'activation_kv').
I0423 21:48:33.982056 136048720987968 attentions.py:1198] attentions/out Physical: bfloat16[8,2048,16,128]..................................... ('fsdp', None, 'tensor', None).
I0423 21:48:34.003091 136048720987968 linears.py:525] linears/x Logical: bfloat16[8,2048,7168]....................................... ('activation_batch', 'activation_length', 'activation_mlp').
I0423 21:48:34.003154 136048720987968 linears.py:525] linears/x Physical: bfloat16[8,2048,7168]....................................... ('fsdp', None, 'tensor').
I0423 21:48:34.213324 136048720987968 checkpointing.py:577] checkpoint manager exists so trying to load this run's existing checkpoint
I0423 21:48:34.213433 136048720987968 checkpointing.py:665] No existing checkpoints found, not restoring checkpoint.
fsdp: 8
tensor: 4
I0423 21:48:35.644825 136048720987968 maxtext_utils.py:1668]  params/params/decoder/decoder_norm/scale
    Shape:     float32[2048]
    Logical:   P('norm',)
    Physical:  ('tensor',)
I0423 21:48:35.644952 136048720987968 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, 'tensor')
I0423 21:48:35.645006 136048720987968 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, 'tensor')
I0423 21:48:35.645069 136048720987968 maxtext_utils.py:1668]  params/params/decoder/layers/mlp/wo/kernel
    Shape:     float32[7168,16,2048]
    Logical:   P('mlp', 'layers', 'embed')
    Physical:  ('tensor', None, 'fsdp')
I0423 21:48:35.645125 136048720987968 maxtext_utils.py:1668]  params/params/decoder/layers/post_self_attention_layer_norm/scale
    Shape:     float32[2048,16]
    Logical:   P('norm', 'layers')
    Physical:  ('tensor', None)
I0423 21:48:35.645166 136048720987968 maxtext_utils.py:1668]  params/params/decoder/layers/pre_self_attention_layer_norm/scale
    Shape:     float32[2048,16]
    Logical:   P('norm', 'layers')
    Physical:  ('tensor', None)
I0423 21:48:35.645220 136048720987968 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, 'tensor', None)
I0423 21:48:35.645277 136048720987968 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:  ('tensor', None, None, 'fsdp')
I0423 21:48:35.645319 136048720987968 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, 'tensor', None)
I0423 21:48:35.645357 136048720987968 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, 'tensor', None)
I0423 21:48:35.645404 136048720987968 maxtext_utils.py:1668]  params/params/decoder/logits_dense/kernel
    Shape:     float32[2048,32000]
    Logical:   P('embed_vocab', 'vocab')
    Physical:  ('fsdp', 'tensor')

I0423 21:48:35.645452 136048720987968 maxtext_utils.py:1668]  params/params/token_embedder/embedding
    Shape:     float32[32000,2048]
    Logical:   P('vocab', 'embed_vocab')
    Physical:  ('tensor', 'fsdp')
I0423 21:48:36.134724 136048720987968 train.py:155] train/xent Logical: float32[8,2048]............................................. ('activation_embed_and_logits_batch', 'activation_length').
I0423 21:48:36.134818 136048720987968 train.py:155] train/xent Physical: float32[8,2048]............................................. ('fsdp', None).
I0423 21:48:36.150428 136048720987968 train.py:162] train/z_loss Logical: float32[8,2048]............................................. ('activation_embed_and_logits_batch', 'activation_length').
I0423 21:48:36.150488 136048720987968 train.py:162] train/z_loss Physical: float32[8,2048]............................................. ('fsdp', None).
I0423 21:48:51.244885 136048720987968 max_utils.py:791] Total memory size: 0.9 GB, Output size: 0.4 GB, Temp size: 0.5 GB, Argument size: 0.4 GB, Host temp size: 0.0 GB.
I0423 21:48:51.245783 136048720987968 metric_logger.py:301] number parameters: 1.104 billion
I0423 21:48:52.860803 136048720987968 checkpointing.py:772] Waiting for step 0 to finish before checkpoint...
I0423 21:49:07.353025 136048720987968 checkpointing.py:776] Waited 14.492198705673218 seconds for step 0 to finish before starting checkpointing.
I0423 21:49:07.355554 136048720987968 checkpoint_manager.py:2009] [process=5][thread=MainThread][wait_until_finished] No Save Finalize thread to wait for. Returning.
I0423 21:49:07.357566 136048720987968 checkpoint_manager.py:1512] [process=5] Saving checkpoint at step 0
I0423 21:49:07.358924 136048720987968 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_09_pdb_lt_1/checkpoints/0.
I0423 21:49:07.690006 136048720987968 signaling_client.py:364] Using JaxDistributedSignalingClient
I0423 21:49:07.690933 136048720987968 jax_array_handlers.py:360] Scheduling D2H of 39 prioritized jax.Array.
I0423 21:49:07.690991 136048720987968 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 21:49:07.957119 136048720987968 base_pytree_checkpoint_handler.py:154] [process=5][thread=MainThread] Initiated "orbax.checkpoint._src.serialization.jax_array_handlers.ArrayHandler".serialize. Time taken: 0.267188s
I0423 21:49:07.957302 136048720987968 base_pytree_checkpoint_handler.py:130] [process=5] /jax/orbax/write/blocking_gbytes_per_sec: 5.656 GiB/s (total gbytes: 1.5 GiB) (time elapsed: 0.27274203300476074 s) (per-host)
I0423 21:49:07.957358 136048720987968 base_pytree_checkpoint_handler.py:768] [process=5][thread=MainThread] Initiated Pytree async_save. Time taken: 0.272809s (batch_requests_ready=0.002244s, total_serialization_initiated=0.270490s, others=0.000075s)
I0423 21:49:07.957452 136048720987968 composite_checkpoint_handler.py:715] [process=5][thread=MainThread] Initiated CompositeCheckpointHandler.async_save. Time taken: 0.277485s (all_items=0.000018s, per_item={'items': '0.00001812'}, temp_paths=0.277467)
I0423 21:49:07.958256 136048720987968 event_tracking.py:125] [process=5] [async] Finished blocking save in 0.60 seconds. Continuing save @ gs://lance-maxtext/linen_ckpt_xpk_test_pipeline_scan_nnx_20260423_211617/linen_xpk_test_pipeline_scan_nnx_20260423_211617_09_pdb_lt_1/checkpoints/0.
I0423 21:49:07.958545 135920407983872 async_checkpointer.py:76] [process=5][thread=async_save] Background save thread started. Deadline for this save operation is 2026-04-23 22:09:07.958513
I0423 21:49:07.964939 136048720987968 checkpoint_manager.py:1560] [process=5][thread=MainThread][step=0] Starting CheckpointManager Save Finalize thread=save_finalize
I0423 21:49:07.965255 135920382805760 async_checkpointer.py:280] [process=5][thread=save_finalize] Waiting for background save thread=async_save.
I0423 21:49:07.965414 136048720987968 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_09_pdb_lt_1/checkpoints', 'reached_preemption': False, 'preemption_received_at': None, 'synchronous': False, 'wait_for_prev_start_time': 1776980947.3555348, 'wait_for_prev_duration_secs': 6.318092346191406e-05, 'time_between_consecutive_saves_sec': None, 'checkpointer_blocking_start_time': 1776980947.3576045, 'checkpointer_blocking_duration_secs': 0.6011149883270264, 'get_old_steps_start_time': 1776980947.9587464, 'get_old_steps_duration_secs': 3.075599670410156e-05, 'checkpoint_manager_blocking_start_time': 1776980947.3535862, 'checkpoint_manager_blocking_duration_secs': 0.6117861270904541}
I0423 21:49:07.965529 136048720987968 checkpointing.py:408] Started an asynchronous checkpoint save for step 0
I0423 21:49:07.965581 136048720987968 max_utils.py:750] 
Memstats: After params initialized:
I0423 21:49:07.965632 136048720987968 max_utils.py:756] 	Using (GB) 0.43 / 31.25 (1.376000%) on TPU_18(process=5,(2,4,0,0))
I0423 21:49:07.965682 136048720987968 max_utils.py:756] 	Using (GB) 0.43 / 31.25 (1.376000%) on TPU_19(process=5,(3,4,0,0))
I0423 21:49:07.965711 136048720987968 max_utils.py:756] 	Using (GB) 0.43 / 31.25 (1.376000%) on TPU_22(process=5,(2,5,0,0))
I0423 21:49:07.965738 136048720987968 max_utils.py:756] 	Using (GB) 0.43 / 31.25 (1.376000%) on TPU_23(process=5,(3,5,0,0))
I0423 21:49:08.278835 136048720987968 metric_logger.py:196] completed step: 0, seconds: 1.615, TFLOP/s/device: 2.103, Tokens/s/device: 317.048, total_weights: 16384, loss: 10.862, lm_loss: 10.862, perplexity: 52153.723
I0423 21:49:08.382639 136048720987968 metric_logger.py:196] completed step: 1, seconds: 15.416, TFLOP/s/device: 0.220, Tokens/s/device: 33.211, total_weights: 16384, loss: 10.862, lm_loss: 10.862, perplexity: 52153.723
I0423 21:49:08.802454 136048720987968 metric_logger.py:196] completed step: 2, seconds: 0.046, TFLOP/s/device: 74.528, Tokens/s/device: 11233.736, total_weights: 16384, loss: 9.763, lm_loss: 9.763, perplexity: 17373.129
I0423 21:49:08.860742 136048720987968 metric_logger.py:196] completed step: 3, seconds: 0.420, TFLOP/s/device: 8.082, Tokens/s/device: 1218.204, total_weights: 16384, loss: 8.823, lm_loss: 8.823, perplexity: 6791.817
I0423 21:49:08.977920 136048720987968 metric_logger.py:196] completed step: 4, seconds: 0.064, TFLOP/s/device: 53.125, Tokens/s/device: 8007.507, total_weights: 16384, loss: 7.947, lm_loss: 7.947, perplexity: 2828.417
I0423 21:49:08.984023 136048720987968 metric_logger.py:196] completed step: 5, seconds: 0.058, TFLOP/s/device: 58.551, Tokens/s/device: 8825.456, total_weights: 16384, loss: 7.212, lm_loss: 7.212, perplexity: 1355.339
I0423 21:49:10.901674    2537 google_auth_provider.cc:181] Running on GCE, using service account 562977990677-compute@developer.gserviceaccount.com
I0423 21:49:12.921710 135920391198464 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_09_pdb_lt_1/checkpoints/0/items/array_metadatas/process_5
I0423 21:49:22.927630 136048720987968 metric_logger.py:196] completed step: 6, seconds: 0.118, TFLOP/s/device: 28.811, Tokens/s/device: 4342.700, total_weights: 16384, loss: 6.652, lm_loss: 6.652, perplexity: 774.659
I0423 21:49:22.986291 136048720987968 metric_logger.py:196] completed step: 7, seconds: 13.886, TFLOP/s/device: 0.245, Tokens/s/device: 36.872, total_weights: 16384, loss: 6.269, lm_loss: 6.269, perplexity: 527.696
I0423 21:49:23.044458 136048720987968 metric_logger.py:196] completed step: 8, seconds: 0.063, TFLOP/s/device: 53.615, Tokens/s/device: 8081.446, total_weights: 16384, loss: 6.031, lm_loss: 6.031, perplexity: 416.165
I0423 21:49:23.102215 136048720987968 checkpointing.py:772] Waiting for step 9 to finish before checkpoint...
I0423 21:49:23.102866 136048720987968 checkpointing.py:776] Waited 0.0006685256958007812 seconds for step 9 to finish before starting checkpointing.
I0423 21:49:23.105968 136048720987968 checkpoint_manager.py:2020] [process=5][thread=MainThread][step=0][wait_until_finished] Waiting for Save Finalize thread (save_finalize) to complete.
I0423 21:49:43.878856 135920407983872 base_pytree_checkpoint_handler.py:130] [process=5] /jax/orbax/write/gbytes_per_sec: 43.642 MiB/s (total gbytes: 1.5 GiB) (time elapsed: 36.19425630569458 s) (per-host)
I0423 21:49:43.878983 135920407983872 async_checkpointer.py:90] [process=5][thread=async_save] 3 Handler Commit operations completed. Time taken: 35.920311s.
I0423 21:49:53.262937 135920407983872 async_checkpointer.py:160] [process=5][thread=async_save] Background save thread done. Time taken: 45.304249s.
I0423 21:49:53.263246 135920382805760 async_checkpointer.py:288] [process=5][thread=save_finalize] Done with waiting for background save thread=async_save.
I0423 21:49:53.263361 135920382805760 async_checkpointer.py:298] [process=5][thread=save_finalize] No errors found in background save thread=async_save.
I0423 21:49:53.263413 135920382805760 checkpoint_manager.py:2137] [process=5][thread=save_finalize][step=0] CheckpointManager Save Finalize is syncing with other hosts...
I0423 21:49:53.265325 135920382805760 checkpoint_manager.py:2146] [process=5][thread=save_finalize][step=0] CheckpointManager Save Finalize is done on all hosts.
I0423 21:49:53.265501 136048720987968 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 21:49:53.265660 136048720987968 checkpoint_manager.py:1452] Waiting for previous save to complete took 30.159679 seconds. If this number is high, consider checkpointing less frequently.
I0423 21:49:53.267626 136048720987968 checkpoint_manager.py:1512] [process=5] Saving checkpoint at step 9
I0423 21:49:53.269632 136048720987968 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_09_pdb_lt_1/checkpoints/9.
I0423 21:49:53.553108 136048720987968 jax_array_handlers.py:360] Scheduling D2H of 39 prioritized jax.Array.
I0423 21:49:53.553202 136048720987968 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 21:49:53.578809 136048720987968 base_pytree_checkpoint_handler.py:154] [process=5][thread=MainThread] Initiated "orbax.checkpoint._src.serialization.jax_array_handlers.ArrayHandler".serialize. Time taken: 0.026888s
I0423 21:49:53.578946 136048720987968 base_pytree_checkpoint_handler.py:130] [process=5] /jax/orbax/write/blocking_gbytes_per_sec: 50.875 GiB/s (total gbytes: 1.5 GiB) (time elapsed: 0.03032088279724121 s) (per-host)
I0423 21:49:53.578996 136048720987968 base_pytree_checkpoint_handler.py:768] [process=5][thread=MainThread] Initiated Pytree async_save. Time taken: 0.030379s (batch_requests_ready=0.001791s, total_serialization_initiated=0.028518s, others=0.000069s)
I0423 21:49:53.579073 136048720987968 composite_checkpoint_handler.py:715] [process=5][thread=MainThread] Initiated CompositeCheckpointHandler.async_save. Time taken: 0.034483s (all_items=0.000015s, per_item={'items': '0.00001550'}, temp_paths=0.034467)
I0423 21:49:53.579789 136048720987968 event_tracking.py:125] [process=5] [async] Finished blocking save in 0.31 seconds. Continuing save @ gs://lance-maxtext/linen_ckpt_xpk_test_pipeline_scan_nnx_20260423_211617/linen_xpk_test_pipeline_scan_nnx_20260423_211617_09_pdb_lt_1/checkpoints/9.
I0423 21:49:53.580128 135920382805760 async_checkpointer.py:76] [process=5][thread=async_save] Background save thread started. Deadline for this save operation is 2026-04-23 22:09:53.580090
I0423 21:49:53.583176 136048720987968 checkpoint_manager.py:1560] [process=5][thread=MainThread][step=9] Starting CheckpointManager Save Finalize thread=save_finalize
I0423 21:49:53.583411 135919827404544 async_checkpointer.py:280] [process=5][thread=save_finalize] Waiting for background save thread=async_save.
I0423 21:49:53.583539 136048720987968 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_09_pdb_lt_1/checkpoints', 'reached_preemption': False, 'preemption_received_at': None, 'synchronous': False, 'wait_for_prev_start_time': 1776980963.105939, 'wait_for_prev_duration_secs': 30.159679174423218, 'time_between_consecutive_saves_sec': None, 'checkpointer_blocking_start_time': 1776980993.2676775, 'checkpointer_blocking_duration_secs': 0.3126089572906494, 'get_old_steps_start_time': 1776980993.5803106, 'get_old_steps_duration_secs': 2.956390380859375e-05, 'checkpoint_manager_blocking_start_time': 1776980963.1030934, 'checkpoint_manager_blocking_duration_secs': 30.480411291122437}
I0423 21:49:53.583661 136048720987968 checkpointing.py:408] Started an asynchronous checkpoint save for step 9
I0423 21:49:53.583707 136048720987968 checkpoint_manager.py:2020] [process=5][thread=MainThread][step=9][wait_until_finished] Waiting for Save Finalize thread (save_finalize) to complete.
I0423 21:49:59.220260 135913434773248 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_09_pdb_lt_1/checkpoints/9/items/array_metadatas/process_5
I0423 21:50:35.534951 135920382805760 base_pytree_checkpoint_handler.py:130] [process=5] /jax/orbax/write/gbytes_per_sec: 37.622 MiB/s (total gbytes: 1.5 GiB) (time elapsed: 41.98628115653992 s) (per-host)
I0423 21:50:35.535080 135920382805760 async_checkpointer.py:90] [process=5][thread=async_save] 3 Handler Commit operations completed. Time taken: 41.954829s.
I0423 21:50:42.580158 135920382805760 async_checkpointer.py:160] [process=5][thread=async_save] Background save thread done. Time taken: 48.999892s.
I0423 21:50:42.580398 135919827404544 async_checkpointer.py:288] [process=5][thread=save_finalize] Done with waiting for background save thread=async_save.
I0423 21:50:42.580470 135919827404544 async_checkpointer.py:298] [process=5][thread=save_finalize] No errors found in background save thread=async_save.
I0423 21:50:42.580530 135919827404544 checkpoint_manager.py:2137] [process=5][thread=save_finalize][step=9] CheckpointManager Save Finalize is syncing with other hosts...
I0423 21:50:42.583033 135919827404544 checkpoint_manager.py:2146] [process=5][thread=save_finalize][step=9] CheckpointManager Save Finalize is done on all hosts.
I0423 21:50:42.583209 136048720987968 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 21:50:42.583360 136048720987968 checkpoint_manager.py:2009] [process=5][thread=MainThread][wait_until_finished] No Save Finalize thread to wait for. Returning.
I0423 21:50:42.584220 136048720987968 metric_logger.py:196] completed step: 9, seconds: 0.058, TFLOP/s/device: 58.230, Tokens/s/device: 8777.043, total_weights: 16384, loss: 5.892, lm_loss: 5.892, perplexity: 361.955
Per train step:
 Total TFLOPs: 3.40 
 split as 93.93% learnable weight flops and 6.07% attention flops
XPK End: Thu Apr 23 21:50:51 UTC 2026
EXIT_CODE=0
NNX  ·  5b21b66fc  ·  test_pipeline_scan_nnx_20260423_211617  ·  full log
XPK Start: Thu Apr 23 23:00:50 UTC 2026
PyTorch was not found. Models won't be available and only tokenizers, configuration and file/data utilities can be used.
`rope_parameters`'s factor field must be a float >= 1, got 40
`rope_parameters`'s beta_fast field must be a float, got 32
`rope_parameters`'s beta_slow field must be a float, got 1
DeepseekV32Config got `key=rope_scaling` in kwargs but hasn't set it as attribute. For RoPE standardization you need to set `self.rope_parameters` in model's config. 
2026-04-23 23:01:15.506371: 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 23:01:15.715388 138874543998784 max_utils.py:273] Attempting to initialize the jax distributed system...
I0423 23:01:24.756636 138874543998784 distributed.py:149] Starting JAX distributed service on [::]:8482
I0423 23:01:24.758975 138874543998784 distributed.py:172] Connecting to JAX distributed service on mt-09-pdb-lt-1-72qvl-slice-job-0-0.mt-09-pdb-lt-1-72qvl:8482
I0423 23:01:26.715044 138874543998784 max_utils.py:284] Jax distributed system initialized!
I0423 23:01:32.903406 138874543998784 max_utils.py:800] System Information: Jax Version: 0.9.2
I0423 23:01:32.903510 138874543998784 max_utils.py:801] System Information: Jaxlib Version: 0.9.2
I0423 23:01:32.903548 138874543998784 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 23:01:32.903586 138874543998784 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 23:01:40 UTC 2026
EXIT_CODE=1