MaxView

‹ 14_async_ckpt_false_resumeCase: 14_async_ckpt_false_save15_ocdbt_false_resume ›

Metrics: Linen vs NNX  ·  main

MetricLinen  b117f50cfNNX  b117f50cfDiff (NNX − Linen)
Parameters1.104 billion
Final loss7.1810
TFLOP/s92.266
Tok/s13907.3
Avg s/step13.571
Memory %1.38
JAX0.9.20.9.2

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

Linen  ·  b117f50cf  ·  main_20260424_070227  ·  full log
XPK Start: Fri Apr 24 07:48:35 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 07:49:00.166757: 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 07:49:00.372381 136089302112064 max_utils.py:273] Attempting to initialize the jax distributed system...
I0424 07:49:09.413625 136089302112064 distributed.py:149] Starting JAX distributed service on [::]:8482
I0424 07:49:09.416107 136089302112064 distributed.py:172] Connecting to JAX distributed service on mt-14-async-ckpt-false--vfd7f-slice-job-0-0.mt-14-async-ckpt-false--vfd7f:8482
I0424 07:49:10.412172 136089302112064 max_utils.py:284] Jax distributed system initialized!
I0424 07:49:16.499609 136089302112064 max_utils.py:800] System Information: Jax Version: 0.9.2
I0424 07:49:16.499716 136089302112064 max_utils.py:801] System Information: Jaxlib Version: 0.9.2
I0424 07:49:16.499757 136089302112064 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 07:49:16.499791 136089302112064 train_utils.py:361] WARNING: Sequence packing is essentially ignored for synthetic data. Please use a real dataset to use sequence packing.
I0424 07:49:17.197314 136089302112064 maxtext_utils.py:1604] Num_devices: 32, shape (1, 1, 1, 32, 1, 1, 1, 1, 1, 1, 1, 1)
I0424 07:49:17.197597 136089302112064 checkpointing.py:677] Setting up checkpoint logger...
I0424 07:49:17.197653 136089302112064 checkpointing.py:233] Creating checkpoint manager with ocdbt=True and zarr3=True
I0424 07:49:17.197698 136089302112064 pytree_checkpoint_handler.py:592] save_device_host_concurrent_bytes=None
I0424 07:49:17.198041 136089302112064 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 0x7bc52691cce0>, enable_pinned_host_transfer=False, save_concurrent_bytes: 96000000000 (89.4 GiB), restore_concurrent_bytes: 96000000000 (89.4 GiB)
I0424 07:49:20.500078 136089302112064 checkpointing.py:265] Enabling policy for fixed interval checkpointing.
I0424 07:49:20.500343 136089302112064 checkpoint_manager.py:708] [process=6][thread=MainThread] CheckpointManager init: checkpointers=None, item_names=('items',), item_handlers={'items': <orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeCheckpointHandler object at 0x7bb0dc0902c0>}, handler_registry=None
I0424 07:49:20.500586 136089302112064 composite_checkpoint_handler.py:237] Deferred registration for item: "items". Adding handler `<orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeCheckpointHandler object at 0x7bb0dc0902c0>` 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 07:49:20.500636 136089302112064 composite_checkpoint_handler.py:237] Deferred registration for item: "metrics". Adding handler `<orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonCheckpointHandler object at 0x7bb0dc092480>` 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 07:49:20.500673 136089302112064 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 0x7bb0dc0902c0>, ('items', <class 'orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeRestoreArgs'>): <orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeCheckpointHandler object at 0x7bb0dc0902c0>, ('metrics', <class 'orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonSaveArgs'>): <orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonCheckpointHandler object at 0x7bb0dc092480>, ('metrics', <class 'orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonRestoreArgs'>): <orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonCheckpointHandler object at 0x7bb0dc092480>}).
I0424 07:49:20.500990 136089302112064 abstract_checkpointer.py:35] orbax-checkpoint version: 0.11.34
I0424 07:49:21.222249 136089302112064 checkpoint_manager.py:1812] Found 0 checkpoint steps in gs://lance-maxtext/linen_ckpt_xpk_main_20260424_070227/linen_xpk_main_20260424_070227_14_async_ckpt_false/checkpoints
I0424 07:49:21.240879 136089302112064 checkpoint_manager.py:929] [process=6][thread=MainThread] CheckpointManager created,  primary_host=0, CheckpointManagerOptions=CheckpointManagerOptions(save_interval_steps=1, max_to_keep=None, keep_time_interval=None, keep_period=None, should_keep_fn=None, best_fn=None, best_mode='max', keep_checkpoints_without_metrics=True, step_prefix=None, step_format_fixed_length=None, step_name_format=None, create=True, cleanup_tmp_directories=False, save_on_steps=frozenset(), single_host_load_and_broadcast=False, todelete_subdir=None, todelete_full_path=None, enable_background_delete=False, read_only=False, enable_async_checkpointing=False, 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=5), 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_14_async_ckpt_false/checkpoints: <orbax.checkpoint.checkpoint_manager.CheckpointManager object at 0x7bb0dc091520>
I0424 07:49:21.241004 136089302112064 checkpointing.py:301] Checkpoint manager created!
I0424 07:49:22.186602 136089302112064 nnx_wrappers.py:437] Unknown Logical: bfloat16[32,2048,2048]...................................... ('activation_batch', 'activation_norm_length', 'activation_embed').
I0424 07:49:22.186721 136089302112064 nnx_wrappers.py:437] Unknown Physical: bfloat16[32,2048,2048]...................................... ('fsdp', None, None).
I0424 07:49:22.567209 136089302112064 attentions.py:1088] attentions/inputs_q Logical: bfloat16[32,2048,2048]...................................... ('activation_batch_attn', 'activation_length_attn', 'activation_embed_attn').
I0424 07:49:22.567301 136089302112064 attentions.py:1088] attentions/inputs_q Physical: bfloat16[32,2048,2048]...................................... ('fsdp', None, None).
I0424 07:49:22.583890 136089302112064 attentions.py:1089] attentions/inputs_kv Logical: bfloat16[32,2048,2048]...................................... ('activation_batch_attn', 'activation_length_attn', 'activation_embed_attn').
I0424 07:49:22.583952 136089302112064 attentions.py:1089] attentions/inputs_kv Physical: bfloat16[32,2048,2048]...................................... ('fsdp', None, None).
I0424 07:49:22.607946 136089302112064 attentions.py:1154] attentions/query Logical: bfloat16[32,2048,16,128].................................... ('activation_kv_batch', 'activation_length_attn', 'activation_kv_heads', 'activation_kv_head_dim').
I0424 07:49:22.608020 136089302112064 attentions.py:1154] attentions/query Physical: bfloat16[32,2048,16,128].................................... ('fsdp', None, None, None).
I0424 07:49:22.624823 136089302112064 attentions.py:1155] attentions/key Logical: bfloat16[32,2048,16,128].................................... ('activation_kv_batch', 'activation_length_attn', 'activation_kv_heads', 'activation_kv_head_dim').
I0424 07:49:22.624890 136089302112064 attentions.py:1155] attentions/key Physical: bfloat16[32,2048,16,128].................................... ('fsdp', None, None, None).
I0424 07:49:22.641524 136089302112064 attentions.py:1156] attentions/value Logical: bfloat16[32,2048,16,128].................................... ('activation_kv_batch', 'activation_length_attn', 'activation_kv_heads', 'activation_kv_head_dim').
I0424 07:49:22.641585 136089302112064 attentions.py:1156] attentions/value Physical: bfloat16[32,2048,16,128].................................... ('fsdp', None, None, None).
I0424 07:49:22.666559 136089302112064 attentions.py:1198] attentions/out Logical: bfloat16[32,2048,16,128].................................... ('activation_batch_attn', 'activation_length_attn', 'activation_heads', 'activation_kv').
I0424 07:49:22.666628 136089302112064 attentions.py:1198] attentions/out Physical: bfloat16[32,2048,16,128].................................... ('fsdp', None, None, None).
I0424 07:49:22.688072 136089302112064 linears.py:525] linears/x Logical: bfloat16[32,2048,7168]...................................... ('activation_batch', 'activation_length', 'activation_mlp').
I0424 07:49:22.688153 136089302112064 linears.py:525] linears/x Physical: bfloat16[32,2048,7168]...................................... ('fsdp', None, None).
I0424 07:49:22.898514 136089302112064 checkpointing.py:577] checkpoint manager exists so trying to load this run's existing checkpoint
I0424 07:49:22.898625 136089302112064 checkpointing.py:665] No existing checkpoints found, not restoring checkpoint.
fsdp: 32
I0424 07:49:24.323312 136089302112064 maxtext_utils.py:1707]  params/params/decoder/decoder_norm/scale
    Shape:     float32[2048]
    Logical:   P('norm',)
    Physical:  (None,)
I0424 07:49:24.323444 136089302112064 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 07:49:24.323499 136089302112064 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 07:49:24.323559 136089302112064 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 07:49:24.323614 136089302112064 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 07:49:24.323656 136089302112064 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 07:49:24.323711 136089302112064 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 07:49:24.323765 136089302112064 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 07:49:24.323807 136089302112064 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 07:49:24.323844 136089302112064 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 07:49:24.323894 136089302112064 maxtext_utils.py:1707]  params/params/decoder/logits_dense/kernel
    Shape:     float32[2048,32000]
    Logical:   P('embed_vocab', 'vocab')
    Physical:  ('fsdp', None)

I0424 07:49:24.323945 136089302112064 maxtext_utils.py:1707]  params/params/token_embedder/embedding
    Shape:     float32[32000,2048]
    Logical:   P('vocab', 'embed_vocab')
    Physical:  (None, 'fsdp')
I0424 07:49:24.815869 136089302112064 train.py:155] train/xent Logical: float32[32,2048]............................................ ('activation_embed_and_logits_batch', 'activation_length').
I0424 07:49:24.815973 136089302112064 train.py:155] train/xent Physical: float32[32,2048]............................................ ('fsdp', None).
I0424 07:49:24.831629 136089302112064 train.py:162] train/z_loss Logical: float32[32,2048]............................................ ('activation_embed_and_logits_batch', 'activation_length').
I0424 07:49:24.831691 136089302112064 train.py:162] train/z_loss Physical: float32[32,2048]............................................ ('fsdp', None).
I0424 07:49:36.023793 136089302112064 max_utils.py:791] Total memory size: 1.5 GB, Output size: 0.4 GB, Temp size: 1.1 GB, Argument size: 0.4 GB, Host temp size: 0.0 GB.
I0424 07:49:36.024638 136089302112064 metric_logger.py:301] number parameters: 1.104 billion
I0424 07:49:47.849042 136089302112064 checkpointing.py:772] Waiting for step 0 to finish before checkpoint...
I0424 07:49:48.003216 136089302112064 checkpointing.py:776] Waited 0.15415549278259277 seconds for step 0 to finish before starting checkpointing.
I0424 07:49:48.005456 136089302112064 checkpoint_manager.py:2009] [process=6][thread=MainThread][wait_until_finished] No Save Finalize thread to wait for. Returning.
I0424 07:49:48.007155 136089302112064 checkpoint_manager.py:1512] [process=6] Saving checkpoint at step 0
I0424 07:49:48.008431 136089302112064 event_tracking.py:70] [process=6] [sync] Started save checkpoint @ gs://lance-maxtext/linen_ckpt_xpk_main_20260424_070227/linen_xpk_main_20260424_070227_14_async_ckpt_false/checkpoints/0.
I0424 07:49:50.597757 136089302112064 signaling_client.py:364] Using JaxDistributedSignalingClient
I0424 07:49:50.599139 136089302112064 future.py:372] [process=6][thread=MainThread][operation_id=1] _SignalingThread.join() waiting for signals ([]) blocking the main thread will slow down blocking save times. This is likely due to main thread calling result() on a CommitFuture.
I0424 07:49:51.968199 136089302112064 jax_array_handlers.py:360] Scheduling D2H of 39 prioritized jax.Array.
I0424 07:49:51.968308 136089302112064 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 07:49:52.245352 136089302112064 base_pytree_checkpoint_handler.py:154] [process=6][thread=MainThread] Initiated "orbax.checkpoint._src.serialization.jax_array_handlers.ArrayHandler".serialize. Time taken: 0.278189s
I0424 07:49:52.245536 136089302112064 base_pytree_checkpoint_handler.py:130] [process=6] /jax/orbax/write/blocking_gbytes_per_sec: 5.420 GiB/s (total gbytes: 1.5 GiB) (time elapsed: 0.2846336364746094 s) (per-host)
I0424 07:49:52.245591 136089302112064 base_pytree_checkpoint_handler.py:768] [process=6][thread=MainThread] Initiated Pytree async_save. Time taken: 0.284698s (batch_requests_ready=0.002880s, total_serialization_initiated=0.281745s, others=0.000073s)
I0424 07:49:52.245690 136089302112064 composite_checkpoint_handler.py:715] [process=6][thread=MainThread] Initiated CompositeCheckpointHandler.async_save. Time taken: 1.648019s (all_items=0.000033s, per_item={'items': '0.00003266'}, temp_paths=1.647986)
I0424 07:49:52.245735 136089302112064 future.py:372] [process=6][thread=MainThread][operation_id=1] _SignalingThread.join() waiting for signals ([]) blocking the main thread will slow down blocking save times. This is likely due to main thread calling result() on a CommitFuture.
I0424 07:49:52.245776 136089302112064 future.py:372] [process=6][thread=MainThread][operation_id=1] _SignalingThread.join() waiting for signals ([]) blocking the main thread will slow down blocking save times. This is likely due to main thread calling result() on a CommitFuture.
I0424 07:49:52.259022    2771 google_auth_provider.cc:181] Running on GCE, using service account 562977990677-compute@developer.gserviceaccount.com
I0424 07:49:54.263951 135960136431360 array_metadata_store.py:203] [process=6][thread=array_type_handler] Wrote 39 array_metadata.ArrayMetadata to gs://lance-maxtext/linen_ckpt_xpk_main_20260424_070227/linen_xpk_main_20260424_070227_14_async_ckpt_false/checkpoints/0/items/array_metadatas/process_6
I0424 07:50:25.891084 136089302112064 base_pytree_checkpoint_handler.py:130] [process=6] /jax/orbax/write/gbytes_per_sec: 46.554 MiB/s (total gbytes: 1.5 GiB) (time elapsed: 33.930142402648926 s) (per-host)
I0424 07:50:30.474292 136089302112064 event_tracking.py:125] [process=6] [sync] Finished blocking save in 42.47 seconds. Continuing save @ gs://lance-maxtext/linen_ckpt_xpk_main_20260424_070227/linen_xpk_main_20260424_070227_14_async_ckpt_false/checkpoints/0.
I0424 07:50:34.678927 136089302112064 event_tracking.py:138] [process=6] [sync] Finished save in 46.67 seconds @ gs://lance-maxtext/linen_ckpt_xpk_main_20260424_070227/linen_xpk_main_20260424_070227_14_async_ckpt_false/checkpoints/0
I0424 07:50:34.680944 136089302112064 checkpoint_manager.py:2137] [process=6][thread=MainThread][step=0] CheckpointManager Save Finalize is syncing with other hosts...
I0424 07:50:34.682678 136089302112064 checkpoint_manager.py:2146] [process=6][thread=MainThread][step=0] CheckpointManager Save Finalize is done on all hosts.
I0424 07:50:34.682728 136089302112064 checkpoint_manager.py:1581] [process=6][thread=MainThread][step=0] Finished synchronous save.
I0424 07:50:34.682783 136089302112064 standard_logger.py:34] {'step': 0, 'event_type': 'save', 'directory': 'gs://lance-maxtext/linen_ckpt_xpk_main_20260424_070227/linen_xpk_main_20260424_070227_14_async_ckpt_false/checkpoints', 'reached_preemption': False, 'preemption_received_at': None, 'synchronous': True, 'wait_for_prev_start_time': 1777016988.0054379, 'wait_for_prev_duration_secs': 6.198883056640625e-05, 'time_between_consecutive_saves_sec': None, 'checkpointer_blocking_start_time': 1777016988.0071938, 'checkpointer_blocking_duration_secs': 46.671849966049194, 'get_old_steps_start_time': 1777017034.6790674, 'get_old_steps_duration_secs': 4.57763671875e-05, 'checkpoint_manager_blocking_start_time': 1777016988.0037277, 'checkpoint_manager_blocking_duration_secs': 46.67902660369873}
I0424 07:50:34.682876 136089302112064 checkpointing.py:410] Saved a checkpoint at step 0.
I0424 07:50:34.682918 136089302112064 max_utils.py:750] 
Memstats: After params initialized:
I0424 07:50:34.682964 136089302112064 max_utils.py:756] 	Using (GB) 0.43 / 31.25 (1.376000%) on TPU_24(process=6,(0,6,0,0))
I0424 07:50:34.682995 136089302112064 max_utils.py:756] 	Using (GB) 0.43 / 31.25 (1.376000%) on TPU_25(process=6,(1,6,0,0))
I0424 07:50:34.683020 136089302112064 max_utils.py:756] 	Using (GB) 0.43 / 31.25 (1.376000%) on TPU_28(process=6,(0,7,0,0))
I0424 07:50:34.683044 136089302112064 max_utils.py:756] 	Using (GB) 0.43 / 31.25 (1.376000%) on TPU_29(process=6,(1,7,0,0))
I0424 07:50:35.014268 136089302112064 metric_logger.py:196] completed step: 0, seconds: 11.824, TFLOP/s/device: 1.149, Tokens/s/device: 173.203, total_weights: 65536, loss: 10.874, lm_loss: 10.874, perplexity: 52779.875
I0424 07:50:35.167812 136089302112064 metric_logger.py:196] completed step: 1, seconds: 47.147, TFLOP/s/device: 0.288, Tokens/s/device: 43.439, total_weights: 65536, loss: 10.874, lm_loss: 10.874, perplexity: 52779.875
I0424 07:50:35.589613 136089302112064 metric_logger.py:196] completed step: 2, seconds: 0.023, TFLOP/s/device: 582.189, Tokens/s/device: 87753.878, total_weights: 65536, loss: 10.560, lm_loss: 10.560, perplexity: 38547.719
I0424 07:50:35.737079 136089302112064 metric_logger.py:196] completed step: 3, seconds: 0.422, TFLOP/s/device: 32.194, Tokens/s/device: 4852.678, total_weights: 65536, loss: 9.980, lm_loss: 9.980, perplexity: 21590.506
I0424 07:50:35.741951 136089302112064 checkpointing.py:772] Waiting for step 5 to finish before checkpoint...
I0424 07:50:36.031052 136089302112064 checkpointing.py:776] Waited 0.2890758514404297 seconds for step 5 to finish before starting checkpointing.
I0424 07:50:36.033933 136089302112064 checkpoint_manager.py:2009] [process=6][thread=MainThread][wait_until_finished] No Save Finalize thread to wait for. Returning.
I0424 07:50:36.035824 136089302112064 checkpoint_manager.py:1512] [process=6] Saving checkpoint at step 5
I0424 07:50:36.037644 136089302112064 event_tracking.py:70] [process=6] [sync] Started save checkpoint @ gs://lance-maxtext/linen_ckpt_xpk_main_20260424_070227/linen_xpk_main_20260424_070227_14_async_ckpt_false/checkpoints/5.
I0424 07:50:38.810818 136089302112064 future.py:372] [process=6][thread=MainThread][operation_id=2] _SignalingThread.join() waiting for signals ([]) blocking the main thread will slow down blocking save times. This is likely due to main thread calling result() on a CommitFuture.
I0424 07:50:40.227766 136089302112064 jax_array_handlers.py:360] Scheduling D2H of 39 prioritized jax.Array.
I0424 07:50:40.227865 136089302112064 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 07:50:40.271354 136089302112064 base_pytree_checkpoint_handler.py:154] [process=6][thread=MainThread] Initiated "orbax.checkpoint._src.serialization.jax_array_handlers.ArrayHandler".serialize. Time taken: 0.045062s
I0424 07:50:40.271544 136089302112064 base_pytree_checkpoint_handler.py:130] [process=6] /jax/orbax/write/blocking_gbytes_per_sec: 31.563 GiB/s (total gbytes: 1.5 GiB) (time elapsed: 0.048872947692871094 s) (per-host)
I0424 07:50:40.271603 136089302112064 base_pytree_checkpoint_handler.py:768] [process=6][thread=MainThread] Initiated Pytree async_save. Time taken: 0.048943s (batch_requests_ready=0.002031s, total_serialization_initiated=0.046834s, others=0.000079s)
I0424 07:50:40.271708 136089302112064 composite_checkpoint_handler.py:715] [process=6][thread=MainThread] Initiated CompositeCheckpointHandler.async_save. Time taken: 1.462156s (all_items=0.000027s, per_item={'items': '0.00002718'}, temp_paths=1.462129)
I0424 07:50:40.271750 136089302112064 future.py:372] [process=6][thread=MainThread][operation_id=2] _SignalingThread.join() waiting for signals ([]) blocking the main thread will slow down blocking save times. This is likely due to main thread calling result() on a CommitFuture.
I0424 07:50:40.271788 136089302112064 future.py:372] [process=6][thread=MainThread][operation_id=2] _SignalingThread.join() waiting for signals ([]) blocking the main thread will slow down blocking save times. This is likely due to main thread calling result() on a CommitFuture.
I0424 07:50:42.319654 135960136431360 array_metadata_store.py:203] [process=6][thread=array_type_handler] Wrote 39 array_metadata.ArrayMetadata to gs://lance-maxtext/linen_ckpt_xpk_main_20260424_070227/linen_xpk_main_20260424_070227_14_async_ckpt_false/checkpoints/5/items/array_metadatas/process_6
I0424 07:51:19.245943 136089302112064 base_pytree_checkpoint_handler.py:130] [process=6] /jax/orbax/write/gbytes_per_sec: 40.478 MiB/s (total gbytes: 1.5 GiB) (time elapsed: 39.02323579788208 s) (per-host)
I0424 07:51:23.377777 136089302112064 event_tracking.py:125] [process=6] [sync] Finished blocking save in 47.34 seconds. Continuing save @ gs://lance-maxtext/linen_ckpt_xpk_main_20260424_070227/linen_xpk_main_20260424_070227_14_async_ckpt_false/checkpoints/5.
I0424 07:51:27.257226 136089302112064 event_tracking.py:138] [process=6] [sync] Finished save in 51.22 seconds @ gs://lance-maxtext/linen_ckpt_xpk_main_20260424_070227/linen_xpk_main_20260424_070227_14_async_ckpt_false/checkpoints/5
I0424 07:51:27.259112 136089302112064 checkpoint_manager.py:2137] [process=6][thread=MainThread][step=5] CheckpointManager Save Finalize is syncing with other hosts...
I0424 07:51:27.260741 136089302112064 checkpoint_manager.py:2146] [process=6][thread=MainThread][step=5] CheckpointManager Save Finalize is done on all hosts.
I0424 07:51:27.260792 136089302112064 checkpoint_manager.py:1581] [process=6][thread=MainThread][step=5] Finished synchronous save.
I0424 07:51:27.260843 136089302112064 standard_logger.py:34] {'step': 5, 'event_type': 'save', 'directory': 'gs://lance-maxtext/linen_ckpt_xpk_main_20260424_070227/linen_xpk_main_20260424_070227_14_async_ckpt_false/checkpoints', 'reached_preemption': False, 'preemption_received_at': None, 'synchronous': True, 'wait_for_prev_start_time': 1777017036.0339131, 'wait_for_prev_duration_secs': 6.031990051269531e-05, 'time_between_consecutive_saves_sec': 1.3511958122253418, 'checkpointer_blocking_start_time': 1777017036.0358617, 'checkpointer_blocking_duration_secs': 51.2214879989624, 'get_old_steps_start_time': 1777017087.257372, 'get_old_steps_duration_secs': 2.574920654296875e-05, 'checkpoint_manager_blocking_start_time': 1777017036.0317092, 'checkpoint_manager_blocking_duration_secs': 51.22910785675049}
I0424 07:51:27.260933 136089302112064 checkpointing.py:410] Saved a checkpoint at step 5.
I0424 07:51:27.261754 136089302112064 metric_logger.py:196] completed step: 4, seconds: 0.153, TFLOP/s/device: 88.905, Tokens/s/device: 13400.773, total_weights: 65536, loss: 9.460, lm_loss: 9.460, perplexity: 12840.355
I0424 07:51:27.285948 136089302112064 metric_logger.py:196] completed step: 5, seconds: 0.148, TFLOP/s/device: 91.917, Tokens/s/device: 13854.688, total_weights: 65536, loss: 8.959, lm_loss: 8.959, perplexity: 7776.467
I0424 07:51:49.839295 136089302112064 metric_logger.py:196] completed step: 6, seconds: 51.524, TFLOP/s/device: 0.264, Tokens/s/device: 39.748, total_weights: 65536, loss: 8.469, lm_loss: 8.469, perplexity: 4762.740
I0424 07:51:49.986777 136089302112064 metric_logger.py:196] completed step: 7, seconds: 22.425, TFLOP/s/device: 0.606, Tokens/s/device: 91.325, total_weights: 65536, loss: 8.003, lm_loss: 8.003, perplexity: 2989.151
I0424 07:51:50.134215 136089302112064 metric_logger.py:196] completed step: 8, seconds: 0.152, TFLOP/s/device: 89.205, Tokens/s/device: 13445.996, total_weights: 65536, loss: 7.572, lm_loss: 7.572, perplexity: 1943.098
I0424 07:51:50.281141 136089302112064 checkpointing.py:772] Waiting for step 9 to finish before checkpoint...
I0424 07:51:50.281839 136089302112064 checkpointing.py:776] Waited 0.0007488727569580078 seconds for step 9 to finish before starting checkpointing.
I0424 07:51:50.283854 136089302112064 checkpoint_manager.py:2009] [process=6][thread=MainThread][wait_until_finished] No Save Finalize thread to wait for. Returning.
I0424 07:51:50.285324 136089302112064 checkpoint_manager.py:1512] [process=6] Saving checkpoint at step 9
I0424 07:51:50.286761 136089302112064 event_tracking.py:70] [process=6] [sync] Started save checkpoint @ gs://lance-maxtext/linen_ckpt_xpk_main_20260424_070227/linen_xpk_main_20260424_070227_14_async_ckpt_false/checkpoints/9.
I0424 07:51:53.228693 136089302112064 future.py:372] [process=6][thread=MainThread][operation_id=3] _SignalingThread.join() waiting for signals ([]) blocking the main thread will slow down blocking save times. This is likely due to main thread calling result() on a CommitFuture.
I0424 07:51:54.707653 136089302112064 jax_array_handlers.py:360] Scheduling D2H of 39 prioritized jax.Array.
I0424 07:51:54.707752 136089302112064 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 07:51:54.749561 136089302112064 base_pytree_checkpoint_handler.py:154] [process=6][thread=MainThread] Initiated "orbax.checkpoint._src.serialization.jax_array_handlers.ArrayHandler".serialize. Time taken: 0.042964s
I0424 07:51:54.749737 136089302112064 base_pytree_checkpoint_handler.py:130] [process=6] /jax/orbax/write/blocking_gbytes_per_sec: 32.904 GiB/s (total gbytes: 1.5 GiB) (time elapsed: 0.04688072204589844 s) (per-host)
I0424 07:51:54.749790 136089302112064 base_pytree_checkpoint_handler.py:768] [process=6][thread=MainThread] Initiated Pytree async_save. Time taken: 0.046943s (batch_requests_ready=0.001995s, total_serialization_initiated=0.044881s, others=0.000067s)
I0424 07:51:54.749889 136089302112064 composite_checkpoint_handler.py:715] [process=6][thread=MainThread] Initiated CompositeCheckpointHandler.async_save. Time taken: 1.522378s (all_items=0.000021s, per_item={'items': '0.00002074'}, temp_paths=1.522357)
I0424 07:51:54.749930 136089302112064 future.py:372] [process=6][thread=MainThread][operation_id=3] _SignalingThread.join() waiting for signals ([]) blocking the main thread will slow down blocking save times. This is likely due to main thread calling result() on a CommitFuture.
I0424 07:51:54.749967 136089302112064 future.py:372] [process=6][thread=MainThread][operation_id=3] _SignalingThread.join() waiting for signals ([]) blocking the main thread will slow down blocking save times. This is likely due to main thread calling result() on a CommitFuture.
I0424 07:51:57.201070 135964437509888 array_metadata_store.py:203] [process=6][thread=array_type_handler] Wrote 39 array_metadata.ArrayMetadata to gs://lance-maxtext/linen_ckpt_xpk_main_20260424_070227/linen_xpk_main_20260424_070227_14_async_ckpt_false/checkpoints/9/items/array_metadatas/process_6
I0424 07:52:33.077620 136089302112064 base_pytree_checkpoint_handler.py:130] [process=6] /jax/orbax/write/gbytes_per_sec: 41.162 MiB/s (total gbytes: 1.5 GiB) (time elapsed: 38.37471151351929 s) (per-host)
I0424 07:52:37.687731 136089302112064 event_tracking.py:125] [process=6] [sync] Finished blocking save in 47.40 seconds. Continuing save @ gs://lance-maxtext/linen_ckpt_xpk_main_20260424_070227/linen_xpk_main_20260424_070227_14_async_ckpt_false/checkpoints/9.
I0424 07:52:42.222501 136089302112064 event_tracking.py:138] [process=6] [sync] Finished save in 51.94 seconds @ gs://lance-maxtext/linen_ckpt_xpk_main_20260424_070227/linen_xpk_main_20260424_070227_14_async_ckpt_false/checkpoints/9
I0424 07:52:42.224480 136089302112064 checkpoint_manager.py:2137] [process=6][thread=MainThread][step=9] CheckpointManager Save Finalize is syncing with other hosts...
I0424 07:52:42.226106 136089302112064 checkpoint_manager.py:2146] [process=6][thread=MainThread][step=9] CheckpointManager Save Finalize is done on all hosts.
I0424 07:52:42.226163 136089302112064 checkpoint_manager.py:1581] [process=6][thread=MainThread][step=9] Finished synchronous save.
I0424 07:52:42.226220 136089302112064 standard_logger.py:34] {'step': 9, 'event_type': 'save', 'directory': 'gs://lance-maxtext/linen_ckpt_xpk_main_20260424_070227/linen_xpk_main_20260424_070227_14_async_ckpt_false/checkpoints', 'reached_preemption': False, 'preemption_received_at': None, 'synchronous': True, 'wait_for_prev_start_time': 1777017110.283837, 'wait_for_prev_duration_secs': 5.745887756347656e-05, 'time_between_consecutive_saves_sec': 23.023056745529175, 'checkpointer_blocking_start_time': 1777017110.2853622, 'checkpointer_blocking_duration_secs': 51.93725848197937, 'get_old_steps_start_time': 1777017162.2226353, 'get_old_steps_duration_secs': 1.8835067749023438e-05, 'checkpoint_manager_blocking_start_time': 1777017110.282062, 'checkpoint_manager_blocking_duration_secs': 51.94412803649902}
I0424 07:52:42.226333 136089302112064 checkpointing.py:410] Saved a checkpoint at step 9.
I0424 07:52:42.226372 136089302112064 checkpoint_manager.py:2009] [process=6][thread=MainThread][wait_until_finished] No Save Finalize thread to wait for. Returning.
I0424 07:52:42.226408 136089302112064 checkpoint_manager.py:2009] [process=6][thread=MainThread][wait_until_finished] No Save Finalize thread to wait for. Returning.
I0424 07:52:42.227198 136089302112064 metric_logger.py:196] completed step: 9, seconds: 0.147, TFLOP/s/device: 92.266, Tokens/s/device: 13907.280, total_weights: 65536, loss: 7.181, lm_loss: 7.181, perplexity: 1314.495
Per train step:
 Total TFLOPs: 13.59 
 split as 93.93% learnable weight flops and 6.07% attention flops
XPK End: Fri Apr 24 07:52:55 UTC 2026
EXIT_CODE=0
NNX  ·  b117f50cf  ·  main_20260424_070227  ·  full log
XPK Start: Fri Apr 24 08:53:40 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 08:54:05.529467: 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 08:54:05.731936 140670237226816 max_utils.py:273] Attempting to initialize the jax distributed system...
I0424 08:54:14.773833 140670237226816 distributed.py:149] Starting JAX distributed service on [::]:8482
I0424 08:54:14.776200 140670237226816 distributed.py:172] Connecting to JAX distributed service on mt-14-async-ckpt-false--ihzyu-slice-job-0-0.mt-14-async-ckpt-false--ihzyu:8482
I0424 08:54:15.867282 140670237226816 max_utils.py:284] Jax distributed system initialized!
I0424 08:54:20.919392 140670237226816 max_utils.py:800] System Information: Jax Version: 0.9.2
I0424 08:54:20.919500 140670237226816 max_utils.py:801] System Information: Jaxlib Version: 0.9.2
I0424 08:54:20.919540 140670237226816 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 08:54:20.919574 140670237226816 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: Fri Apr 24 08:54:30 UTC 2026
EXIT_CODE=1