main (586e69205) vs feat/nnx-set-defaults-true (73213e044)| Metric | main 586e69205 | feat/nnx-set-defaults-true 73213e044 | Diff (feat/nnx-set-defaults-true − main) |
|---|---|---|---|
| Parameters | 1.104 billion | 1.104 billion | — |
| Final loss | 8.1750 | 8.1750 | 0 |
| TFLOP/s | 59.240 | 58.797 | -0.443 |
| Tok/s | 8929.3 | 8862.5 | -66.81 |
| Avg s/step | 3.624 | 2.858 | -0.766 |
| Memory % | 1.38 | 1.38 | 0 |
| JAX | 0.9.2 | 0.8.1 | — |
Diff = branch value − main value. Green = branch improved. Red = branch regressed.
XPK Start: Thu Apr 23 07:29:14 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 07:29:38.816194: 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 07:29:39.026206 137191704463168 max_utils.py:273] Attempting to initialize the jax distributed system... I0423 07:29:48.065587 137191704463168 distributed.py:149] Starting JAX distributed service on [::]:8482 I0423 07:29:48.067880 137191704463168 distributed.py:172] Connecting to JAX distributed service on mt-04-int8-wfyyn-slice-job-0-0.mt-04-int8-wfyyn:8482 I0423 07:29:49.604523 137191704463168 max_utils.py:284] Jax distributed system initialized! I0423 07:29:55.625925 137191704463168 max_utils.py:800] System Information: Jax Version: 0.9.2 I0423 07:29:55.626030 137191704463168 max_utils.py:801] System Information: Jaxlib Version: 0.9.2 I0423 07:29:55.626071 137191704463168 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 07:29:55.626106 137191704463168 train_utils.py:361] WARNING: Sequence packing is essentially ignored for synthetic data. Please use a real dataset to use sequence packing. I0423 07:29:56.333821 137191704463168 maxtext_utils.py:1604] Num_devices: 32, shape (1, 1, 1, 32, 1, 1, 1, 1, 1, 1, 1, 1, 1) I0423 07:29:56.334224 137191704463168 checkpointing.py:677] Setting up checkpoint logger... I0423 07:29:56.334279 137191704463168 checkpointing.py:233] Creating checkpoint manager with ocdbt=True and zarr3=True I0423 07:29:56.334323 137191704463168 pytree_checkpoint_handler.py:592] save_device_host_concurrent_bytes=None I0423 07:29:56.334662 137191704463168 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 0x7cc5dad18110>, enable_pinned_host_transfer=False, save_concurrent_bytes: 96000000000 (89.4 GiB), restore_concurrent_bytes: 96000000000 (89.4 GiB) I0423 07:29:59.183049 137191704463168 checkpointing.py:265] Enabling policy for fixed interval checkpointing. I0423 07:29:59.183289 137191704463168 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 0x7cb1743943b0>}, handler_registry=None I0423 07:29:59.183523 137191704463168 composite_checkpoint_handler.py:237] Deferred registration for item: "items". Adding handler `<orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeCheckpointHandler object at 0x7cb1743943b0>` 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 07:29:59.183571 137191704463168 composite_checkpoint_handler.py:237] Deferred registration for item: "metrics". Adding handler `<orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonCheckpointHandler object at 0x7cb130516120>` 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 07:29:59.183608 137191704463168 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 0x7cb1743943b0>, ('items', <class 'orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeRestoreArgs'>): <orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeCheckpointHandler object at 0x7cb1743943b0>, ('metrics', <class 'orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonSaveArgs'>): <orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonCheckpointHandler object at 0x7cb130516120>, ('metrics', <class 'orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonRestoreArgs'>): <orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonCheckpointHandler object at 0x7cb130516120>}). I0423 07:29:59.183938 137191704463168 abstract_checkpointer.py:35] orbax-checkpoint version: 0.11.34 I0423 07:29:59.184010 137191704463168 async_checkpointer.py:192] [process=5][thread=MainThread] Using barrier_sync_fn: <function get_barrier_sync_fn.<locals>._fn at 0x7cb1307a94e0> timeout: 1200 secs and primary_host=0 for async checkpoint writes I0423 07:30:01.245287 137191704463168 checkpoint_manager.py:1812] Found 0 checkpoint steps in gs://lance-maxtext/linen_ckpt_xpk_main_20260423_071538/linen_xpk_main_20260423_071538_04_int8/checkpoints I0423 07:30:01.681879 137191704463168 checkpoint_manager.py:929] [process=5][thread=MainThread] CheckpointManager created, primary_host=0, CheckpointManagerOptions=CheckpointManagerOptions(save_interval_steps=1, max_to_keep=None, keep_time_interval=None, keep_period=None, should_keep_fn=None, best_fn=None, best_mode='max', keep_checkpoints_without_metrics=True, step_prefix=None, step_format_fixed_length=None, step_name_format=None, create=True, cleanup_tmp_directories=False, save_on_steps=frozenset(), single_host_load_and_broadcast=False, todelete_subdir=None, todelete_full_path=None, enable_background_delete=False, read_only=False, enable_async_checkpointing=True, async_options=None, multiprocessing_options=MultiprocessingOptions(primary_host=0, active_processes=None, barrier_sync_key_prefix=None), should_save_fn=None, file_options=FileOptions(path_permission_mode=None), save_root_metadata=True, temporary_path_class=None, save_decision_policy=FixedIntervalPolicy(interval=10), preservation_policy=LatestN(n=None), prevent_write_metrics=False, enable_should_save_is_saving_in_progress_check=True, enable_per_process_directory_creation=False, lightweight_initialize=False), root_directory=gs://lance-maxtext/linen_ckpt_xpk_main_20260423_071538/linen_xpk_main_20260423_071538_04_int8/checkpoints: <orbax.checkpoint.checkpoint_manager.CheckpointManager object at 0x7cc625fcf6b0> I0423 07:30:01.682051 137191704463168 checkpointing.py:301] Checkpoint manager created! I0423 07:30:02.817699 137191704463168 nnx_wrappers.py:437] Unknown Logical: bfloat16[32,2048,2048]...................................... ('activation_batch', 'activation_norm_length', 'activation_embed'). I0423 07:30:02.817821 137191704463168 nnx_wrappers.py:437] Unknown Physical: bfloat16[32,2048,2048]...................................... ('fsdp', None, None). I0423 07:30:03.201369 137191704463168 attentions.py:1088] attentions/inputs_q Logical: bfloat16[32,2048,2048]...................................... ('activation_batch', 'activation_attn_length', 'activation_attn_embed'). I0423 07:30:03.201461 137191704463168 attentions.py:1088] attentions/inputs_q Physical: bfloat16[32,2048,2048]...................................... ('fsdp', None, None). I0423 07:30:03.217774 137191704463168 attentions.py:1089] attentions/inputs_kv Logical: bfloat16[32,2048,2048]...................................... ('activation_batch', 'activation_attn_length', 'activation_attn_embed'). I0423 07:30:03.217830 137191704463168 attentions.py:1089] attentions/inputs_kv Physical: bfloat16[32,2048,2048]...................................... ('fsdp', None, None). I0423 07:30:03.283683 137191704463168 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 07:30:03.283767 137191704463168 attentions.py:1154] attentions/query Physical: bfloat16[32,2048,16,128].................................... ('fsdp', None, None, None). I0423 07:30:03.300258 137191704463168 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 07:30:03.300321 137191704463168 attentions.py:1155] attentions/key Physical: bfloat16[32,2048,16,128].................................... ('fsdp', None, None, None). I0423 07:30:03.316884 137191704463168 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 07:30:03.316946 137191704463168 attentions.py:1156] attentions/value Physical: bfloat16[32,2048,16,128].................................... ('fsdp', None, None, None). I0423 07:30:03.341585 137191704463168 attentions.py:1198] attentions/out Logical: bfloat16[32,2048,16,128].................................... ('activation_batch', 'activation_attn_length', 'activation_heads', 'activation_kv'). I0423 07:30:03.341662 137191704463168 attentions.py:1198] attentions/out Physical: bfloat16[32,2048,16,128].................................... ('fsdp', None, None, None). I0423 07:30:03.408001 137191704463168 linears.py:525] linears/x Logical: bfloat16[32,2048,7168]...................................... ('activation_batch', 'activation_length', 'activation_mlp'). I0423 07:30:03.408090 137191704463168 linears.py:525] linears/x Physical: bfloat16[32,2048,7168]...................................... ('fsdp', None, None). I0423 07:30:03.915595 137191704463168 checkpointing.py:577] checkpoint manager exists so trying to load this run's existing checkpoint I0423 07:30:03.915714 137191704463168 checkpointing.py:665] No existing checkpoints found, not restoring checkpoint. fsdp: 32 I0423 07:30:06.401934 137191704463168 maxtext_utils.py:1707] params/params/decoder/decoder_norm/scale Shape: float32[2048] Logical: P('norm',) Physical: (None,) I0423 07:30:06.402069 137191704463168 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) I0423 07:30:06.402123 137191704463168 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) I0423 07:30:06.402181 137191704463168 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') I0423 07:30:06.402235 137191704463168 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) I0423 07:30:06.402276 137191704463168 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) I0423 07:30:06.402328 137191704463168 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) I0423 07:30:06.402384 137191704463168 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') I0423 07:30:06.402424 137191704463168 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) I0423 07:30:06.402459 137191704463168 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) I0423 07:30:06.402505 137191704463168 maxtext_utils.py:1707] params/params/decoder/logits_dense/kernel Shape: float32[2048,32000] Logical: P('embed_vocab', 'vocab') Physical: ('fsdp', None) I0423 07:30:06.402550 137191704463168 maxtext_utils.py:1707] params/params/token_embedder/embedding Shape: float32[32000,2048] Logical: P('vocab', 'embed_vocab') Physical: (None, 'fsdp') I0423 07:30:07.790900 137191704463168 train.py:155] train/xent Logical: float32[32,2048]............................................ ('activation_embed_and_logits_batch', 'activation_length'). I0423 07:30:07.790996 137191704463168 train.py:155] train/xent Physical: float32[32,2048]............................................ ('fsdp', None). I0423 07:30:07.806458 137191704463168 train.py:162] train/z_loss Logical: float32[32,2048]............................................ ('activation_embed_and_logits_batch', 'activation_length'). I0423 07:30:07.806521 137191704463168 train.py:162] train/z_loss Physical: float32[32,2048]............................................ ('fsdp', None). I0423 07:30:21.227624 137191704463168 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 07:30:21.228364 137191704463168 metric_logger.py:301] number parameters: 1.104 billion I0423 07:30:36.259264 137191704463168 checkpointing.py:772] Waiting for step 0 to finish before checkpoint... I0423 07:30:36.502389 137191704463168 checkpointing.py:776] Waited 0.24310922622680664 seconds for step 0 to finish before starting checkpointing. I0423 07:30:36.504845 137191704463168 checkpoint_manager.py:2009] [process=5][thread=MainThread][wait_until_finished] No Save Finalize thread to wait for. Returning. I0423 07:30:36.506710 137191704463168 checkpoint_manager.py:1512] [process=5] Saving checkpoint at step 0 I0423 07:30:36.508125 137191704463168 event_tracking.py:70] [process=5] [async] Started save checkpoint @ gs://lance-maxtext/linen_ckpt_xpk_main_20260423_071538/linen_xpk_main_20260423_071538_04_int8/checkpoints/0. I0423 07:30:37.264388 137191704463168 signaling_client.py:364] Using JaxDistributedSignalingClient I0423 07:30:37.265328 137191704463168 jax_array_handlers.py:360] Scheduling D2H of 39 prioritized jax.Array. I0423 07:30:37.265385 137191704463168 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 07:30:37.550566 137191704463168 base_pytree_checkpoint_handler.py:154] [process=5][thread=MainThread] Initiated "orbax.checkpoint._src.serialization.jax_array_handlers.ArrayHandler".serialize. Time taken: 0.286254s I0423 07:30:37.550769 137191704463168 base_pytree_checkpoint_handler.py:130] [process=5] /jax/orbax/write/blocking_gbytes_per_sec: 5.279 GiB/s (total gbytes: 1.5 GiB) (time elapsed: 0.2921931743621826 s) (per-host) I0423 07:30:37.550827 137191704463168 base_pytree_checkpoint_handler.py:768] [process=5][thread=MainThread] Initiated Pytree async_save. Time taken: 0.292262s (batch_requests_ready=0.002385s, total_serialization_initiated=0.289800s, others=0.000077s) I0423 07:30:37.550920 137191704463168 composite_checkpoint_handler.py:715] [process=5][thread=MainThread] Initiated CompositeCheckpointHandler.async_save. Time taken: 0.296517s (all_items=0.000019s, per_item={'items': '0.00001931'}, temp_paths=0.296498) I0423 07:30:37.551775 137191704463168 event_tracking.py:125] [process=5] [async] Finished blocking save in 1.04 seconds. Continuing save @ gs://lance-maxtext/linen_ckpt_xpk_main_20260423_071538/linen_xpk_main_20260423_071538_04_int8/checkpoints/0. I0423 07:30:37.552077 137058507208448 async_checkpointer.py:76] [process=5][thread=async_save] Background save thread started. Deadline for this save operation is 2026-04-23 07:50:37.552044 I0423 07:30:37.553954 137191704463168 checkpoint_manager.py:1560] [process=5][thread=MainThread][step=0] Starting CheckpointManager Save Finalize thread=save_finalize I0423 07:30:37.554244 137056848164608 async_checkpointer.py:280] [process=5][thread=save_finalize] Waiting for background save thread=async_save. I0423 07:30:37.554371 137191704463168 standard_logger.py:34] {'step': 0, 'event_type': 'save', 'directory': 'gs://lance-maxtext/linen_ckpt_xpk_main_20260423_071538/linen_xpk_main_20260423_071538_04_int8/checkpoints', 'reached_preemption': False, 'preemption_received_at': None, 'synchronous': False, 'wait_for_prev_start_time': 1776929436.504826, 'wait_for_prev_duration_secs': 6.29425048828125e-05, 'time_between_consecutive_saves_sec': None, 'checkpointer_blocking_start_time': 1776929436.506748, 'checkpointer_blocking_duration_secs': 1.0454840660095215, 'get_old_steps_start_time': 1776929437.552257, 'get_old_steps_duration_secs': 3.075599670410156e-05, 'checkpoint_manager_blocking_start_time': 1776929436.5029309, 'checkpoint_manager_blocking_duration_secs': 1.051405429840088} I0423 07:30:37.554526 137191704463168 checkpointing.py:408] Started an asynchronous checkpoint save for step 0 I0423 07:30:37.554575 137191704463168 max_utils.py:750] Memstats: After params initialized: I0423 07:30:37.554620 137191704463168 max_utils.py:756] Using (GB) 0.44 / 31.25 (1.408000%) on TPU_18(process=5,(2,4,0,0)) I0423 07:30:37.554665 137191704463168 max_utils.py:756] Using (GB) 0.43 / 31.25 (1.376000%) on TPU_19(process=5,(3,4,0,0)) I0423 07:30:37.554696 137191704463168 max_utils.py:756] Using (GB) 0.43 / 31.25 (1.376000%) on TPU_22(process=5,(2,5,0,0)) I0423 07:30:37.554722 137191704463168 max_utils.py:756] Using (GB) 0.43 / 31.25 (1.376000%) on TPU_23(process=5,(3,5,0,0)) I0423 07:30:37.876575 137191704463168 metric_logger.py:196] completed step: 0, seconds: 15.031, TFLOP/s/device: 0.904, Tokens/s/device: 136.254, total_weights: 65536, loss: 10.874, lm_loss: 10.874, perplexity: 52796.277 I0423 07:30:38.122787 137191704463168 metric_logger.py:196] completed step: 1, seconds: 1.616, TFLOP/s/device: 8.409, Tokens/s/device: 1267.456, total_weights: 65536, loss: 10.874, lm_loss: 10.874, perplexity: 52796.277 I0423 07:30:38.534336 137191704463168 metric_logger.py:196] completed step: 2, seconds: 0.017, TFLOP/s/device: 784.386, Tokens/s/device: 118231.151, total_weights: 65536, loss: 10.262, lm_loss: 10.262, perplexity: 28634.820 I0423 07:30:38.763913 137191704463168 metric_logger.py:196] completed step: 3, seconds: 0.412, TFLOP/s/device: 32.993, Tokens/s/device: 4973.010, total_weights: 65536, loss: 9.731, lm_loss: 9.731, perplexity: 16827.199 I0423 07:30:39.222977 137191704463168 metric_logger.py:196] completed step: 4, seconds: 0.235, TFLOP/s/device: 57.830, Tokens/s/device: 8716.785, total_weights: 65536, loss: 9.272, lm_loss: 9.272, perplexity: 10638.531 I0423 07:30:39.229384 137191704463168 metric_logger.py:196] completed step: 5, seconds: 0.230, TFLOP/s/device: 59.177, Tokens/s/device: 8919.744, total_weights: 65536, loss: 8.887, lm_loss: 8.887, perplexity: 7233.941 I0423 07:30:40.970092 2934 google_auth_provider.cc:181] Running on GCE, using service account 562977990677-compute@developer.gserviceaccount.com I0423 07:30:44.019488 137057402357504 array_metadata_store.py:203] [process=5][thread=array_type_handler] Wrote 39 array_metadata.ArrayMetadata to gs://lance-maxtext/linen_ckpt_xpk_main_20260423_071538/linen_xpk_main_20260423_071538_04_int8/checkpoints/0/items/array_metadatas/process_5 I0423 07:31:08.637168 137191704463168 metric_logger.py:196] completed step: 6, seconds: 0.460, TFLOP/s/device: 29.564, Tokens/s/device: 4456.272, total_weights: 65536, loss: 8.588, lm_loss: 8.588, perplexity: 5365.468 I0423 07:31:08.866616 137191704463168 metric_logger.py:196] completed step: 7, seconds: 29.180, TFLOP/s/device: 0.466, Tokens/s/device: 70.186, total_weights: 65536, loss: 8.380, lm_loss: 8.380, perplexity: 4358.779 I0423 07:31:09.096235 137191704463168 metric_logger.py:196] completed step: 8, seconds: 0.234, TFLOP/s/device: 58.061, Tokens/s/device: 8751.613, total_weights: 65536, loss: 8.251, lm_loss: 8.251, perplexity: 3831.453 I0423 07:31:09.324913 137191704463168 checkpointing.py:772] Waiting for step 9 to finish before checkpoint... I0423 07:31:09.325575 137191704463168 checkpointing.py:776] Waited 0.0006837844848632812 seconds for step 9 to finish before starting checkpointing. I0423 07:31:09.327621 137191704463168 checkpoint_manager.py:2020] [process=5][thread=MainThread][step=0][wait_until_finished] Waiting for Save Finalize thread (save_finalize) to complete. I0423 07:31:11.340154 137058507208448 base_pytree_checkpoint_handler.py:130] [process=5] /jax/orbax/write/gbytes_per_sec: 46.348 MiB/s (total gbytes: 1.5 GiB) (time elapsed: 34.08154249191284 s) (per-host) I0423 07:31:11.340275 137058507208448 async_checkpointer.py:90] [process=5][thread=async_save] 3 Handler Commit operations completed. Time taken: 33.788090s. I0423 07:31:19.360929 137058507208448 async_checkpointer.py:160] [process=5][thread=async_save] Background save thread done. Time taken: 41.808727s. I0423 07:31:19.361246 137056848164608 async_checkpointer.py:288] [process=5][thread=save_finalize] Done with waiting for background save thread=async_save. I0423 07:31:19.361377 137056848164608 async_checkpointer.py:298] [process=5][thread=save_finalize] No errors found in background save thread=async_save. I0423 07:31:19.361428 137056848164608 checkpoint_manager.py:2137] [process=5][thread=save_finalize][step=0] CheckpointManager Save Finalize is syncing with other hosts... I0423 07:31:19.362913 137056848164608 checkpoint_manager.py:2146] [process=5][thread=save_finalize][step=0] CheckpointManager Save Finalize is done on all hosts. I0423 07:31:19.363123 137191704463168 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 07:31:19.363257 137191704463168 checkpoint_manager.py:1452] Waiting for previous save to complete took 10.035638 seconds. If this number is high, consider checkpointing less frequently. I0423 07:31:19.365170 137191704463168 checkpoint_manager.py:1512] [process=5] Saving checkpoint at step 9 I0423 07:31:19.367163 137191704463168 event_tracking.py:70] [process=5] [async] Started save checkpoint @ gs://lance-maxtext/linen_ckpt_xpk_main_20260423_071538/linen_xpk_main_20260423_071538_04_int8/checkpoints/9. I0423 07:31:20.112314 137191704463168 jax_array_handlers.py:360] Scheduling D2H of 39 prioritized jax.Array. I0423 07:31:20.112409 137191704463168 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 07:31:20.146781 137191704463168 base_pytree_checkpoint_handler.py:154] [process=5][thread=MainThread] Initiated "orbax.checkpoint._src.serialization.jax_array_handlers.ArrayHandler".serialize. Time taken: 0.035601s I0423 07:31:20.146950 137191704463168 base_pytree_checkpoint_handler.py:130] [process=5] /jax/orbax/write/blocking_gbytes_per_sec: 39.364 GiB/s (total gbytes: 1.5 GiB) (time elapsed: 0.03918766975402832 s) (per-host) I0423 07:31:20.147001 137191704463168 base_pytree_checkpoint_handler.py:768] [process=5][thread=MainThread] Initiated Pytree async_save. Time taken: 0.039248s (batch_requests_ready=0.001922s, total_serialization_initiated=0.037258s, others=0.000069s) I0423 07:31:20.147096 137191704463168 composite_checkpoint_handler.py:715] [process=5][thread=MainThread] Initiated CompositeCheckpointHandler.async_save. Time taken: 0.043460s (all_items=0.000033s, per_item={'items': '0.00003266'}, temp_paths=0.043427) I0423 07:31:20.147820 137191704463168 event_tracking.py:125] [process=5] [async] Finished blocking save in 0.78 seconds. Continuing save @ gs://lance-maxtext/linen_ckpt_xpk_main_20260423_071538/linen_xpk_main_20260423_071538_04_int8/checkpoints/9. I0423 07:31:20.148157 137056848164608 async_checkpointer.py:76] [process=5][thread=async_save] Background save thread started. Deadline for this save operation is 2026-04-23 07:51:20.148120 I0423 07:31:20.154807 137191704463168 checkpoint_manager.py:1560] [process=5][thread=MainThread][step=9] Starting CheckpointManager Save Finalize thread=save_finalize I0423 07:31:20.155067 137057402357504 async_checkpointer.py:280] [process=5][thread=save_finalize] Waiting for background save thread=async_save. I0423 07:31:20.155201 137191704463168 standard_logger.py:34] {'step': 9, 'event_type': 'save', 'directory': 'gs://lance-maxtext/linen_ckpt_xpk_main_20260423_071538/linen_xpk_main_20260423_071538_04_int8/checkpoints', 'reached_preemption': False, 'preemption_received_at': None, 'synchronous': False, 'wait_for_prev_start_time': 1776929469.3275907, 'wait_for_prev_duration_secs': 10.035637617111206, 'time_between_consecutive_saves_sec': None, 'checkpointer_blocking_start_time': 1776929479.3652086, 'checkpointer_blocking_duration_secs': 0.7830929756164551, 'get_old_steps_start_time': 1776929480.1483245, 'get_old_steps_duration_secs': 3.123283386230469e-05, 'checkpoint_manager_blocking_start_time': 1776929469.3258126, 'checkpoint_manager_blocking_duration_secs': 10.829352855682373} I0423 07:31:20.155321 137191704463168 checkpointing.py:408] Started an asynchronous checkpoint save for step 9 I0423 07:31:20.155365 137191704463168 checkpoint_manager.py:2020] [process=5][thread=MainThread][step=9][wait_until_finished] Waiting for Save Finalize thread (save_finalize) to complete. I0423 07:31:26.817645 137037017499392 array_metadata_store.py:203] [process=5][thread=array_type_handler] Wrote 39 array_metadata.ArrayMetadata to gs://lance-maxtext/linen_ckpt_xpk_main_20260423_071538/linen_xpk_main_20260423_071538_04_int8/checkpoints/9/items/array_metadatas/process_5 I0423 07:32:03.778078 137056848164608 base_pytree_checkpoint_handler.py:130] [process=5] /jax/orbax/write/gbytes_per_sec: 36.171 MiB/s (total gbytes: 1.5 GiB) (time elapsed: 43.67027568817139 s) (per-host) I0423 07:32:03.778208 137056848164608 async_checkpointer.py:90] [process=5][thread=async_save] 3 Handler Commit operations completed. Time taken: 43.629936s. I0423 07:32:12.030685 137056848164608 async_checkpointer.py:160] [process=5][thread=async_save] Background save thread done. Time taken: 51.882382s. I0423 07:32:12.030971 137057402357504 async_checkpointer.py:288] [process=5][thread=save_finalize] Done with waiting for background save thread=async_save. I0423 07:32:12.031091 137057402357504 async_checkpointer.py:298] [process=5][thread=save_finalize] No errors found in background save thread=async_save. I0423 07:32:12.031140 137057402357504 checkpoint_manager.py:2137] [process=5][thread=save_finalize][step=9] CheckpointManager Save Finalize is syncing with other hosts... I0423 07:32:12.032776 137057402357504 checkpoint_manager.py:2146] [process=5][thread=save_finalize][step=9] CheckpointManager Save Finalize is done on all hosts. I0423 07:32:12.032962 137191704463168 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 07:32:12.033113 137191704463168 checkpoint_manager.py:2009] [process=5][thread=MainThread][wait_until_finished] No Save Finalize thread to wait for. Returning. I0423 07:32:12.034093 137191704463168 metric_logger.py:196] completed step: 9, seconds: 0.229, TFLOP/s/device: 59.240, Tokens/s/device: 8929.311, 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 07:32:24 UTC 2026 EXIT_CODE=0
XPK Start: Thu Apr 23 16:02:14 UTC 2026 2026-04-23 16:02:18.903574: E external/local_xla/xla/stream_executor/cuda/cuda_fft.cc:467] Unable to register cuFFT factory: Attempting to register factory for plugin cuFFT when one has already been registered WARNING: All log messages before absl::InitializeLog() is called are written to STDERR E0000 00:00:1776960138.916138 10 cuda_dnn.cc:8579] Unable to register cuDNN factory: Attempting to register factory for plugin cuDNN when one has already been registered E0000 00:00:1776960138.919956 10 cuda_blas.cc:1407] Unable to register cuBLAS factory: Attempting to register factory for plugin cuBLAS when one has already been registered W0000 00:00:1776960138.931391 10 computation_placer.cc:177] computation placer already registered. Please check linkage and avoid linking the same target more than once. W0000 00:00:1776960138.931409 10 computation_placer.cc:177] computation placer already registered. Please check linkage and avoid linking the same target more than once. W0000 00:00:1776960138.931412 10 computation_placer.cc:177] computation placer already registered. Please check linkage and avoid linking the same target more than once. W0000 00:00:1776960138.931414 10 computation_placer.cc:177] computation placer already registered. Please check linkage and avoid linking the same target more than once. 2026-04-23 16:02:37.951869: 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 16:02:38.477598 137110758524736 max_utils.py:273] Attempting to initialize the jax distributed system... INFO:2026-04-23 16:02:47,517:jax._src.distributed:140: Starting JAX distributed service on [::]:8482 I0423 16:02:47.517716 137110758524736 distributed.py:140] Starting JAX distributed service on [::]:8482 INFO:2026-04-23 16:02:47,520:jax._src.distributed:157: Connecting to JAX distributed service on mt-04-int8-nzdye-slice-job-0-0.mt-04-int8-nzdye:8482 I0423 16:02:47.520016 137110758524736 distributed.py:157] Connecting to JAX distributed service on mt-04-int8-nzdye-slice-job-0-0.mt-04-int8-nzdye:8482 I0423 16:02:49.520671 137110758524736 max_utils.py:284] Jax distributed system initialized! I0423 16:02:56.379232 137110758524736 max_utils.py:800] System Information: Jax Version: 0.8.1 I0423 16:02:56.379346 137110758524736 max_utils.py:801] System Information: Jaxlib Version: 0.8.1 I0423 16:02:56.379387 137110758524736 max_utils.py:802] System Information: Jax Backend: PJRT C API TFRT TPU v6 lite Built on Nov 12 2025 14:16:36 (1762985796) cl/831091709 I0423 16:02:56.379423 137110758524736 train_utils.py:377] WARNING: Sequence packing is essentially ignored for synthetic data. Please use a real dataset to use sequence packing. I0423 16:02:56.996012 137110758524736 maxtext_utils.py:1631] Num_devices: 32, shape (1, 1, 1, 32, 1, 1, 1, 1, 1, 1, 1, 1, 1) I0423 16:02:56.996607 137110758524736 maxtext_utils.py:1631] Num_devices: 32, shape (1, 1, 1, 32, 1, 1, 1, 1, 1, 1, 1, 1, 1) I0423 16:02:56.996917 137110758524736 checkpointing.py:688] Setting up checkpoint logger... I0423 16:02:56.996973 137110758524736 checkpointing.py:234] Creating checkpoint manager with ocdbt=True and zarr3=True I0423 16:02:56.997017 137110758524736 pytree_checkpoint_handler.py:589] save_device_host_concurrent_bytes=None I0423 16:02:56.997414 137110758524736 base_pytree_checkpoint_handler.py:415] Created BasePyTreeCheckpointHandler: use_ocdbt=True, use_zarr3=True, pytree_metadata_options=PyTreeMetadataOptions(support_rich_types=False), array_metadata_store=<orbax.checkpoint._src.metadata.array_metadata_store.Store object at 0x7cb319b75700>, enable_pinned_host_transfer=False, save_concurrent_bytes: 96000000000 (89.4 GiB), restore_concurrent_bytes: 96000000000 (89.4 GiB) I0423 16:02:59.881870 137110758524736 checkpointing.py:266] Enabling policy for fixed interval checkpointing. I0423 16:02:59.882108 137110758524736 checkpoint_manager.py:709] [process=5][thread=MainThread] CheckpointManager init: checkpointers=None, item_names=('items',), item_handlers={'items': <orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeCheckpointHandler object at 0x7ca070e80680>}, handler_registry=None I0423 16:02:59.882353 137110758524736 composite_checkpoint_handler.py:237] Deferred registration for item: "items". Adding handler `<orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeCheckpointHandler object at 0x7ca070e80680>` 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 16:02:59.882405 137110758524736 composite_checkpoint_handler.py:237] Deferred registration for item: "metrics". Adding handler `<orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonCheckpointHandler object at 0x7c9e7877adb0>` 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 16:02:59.882441 137110758524736 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 0x7ca070e80680>, ('items', <class 'orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeRestoreArgs'>): <orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeCheckpointHandler object at 0x7ca070e80680>, ('metrics', <class 'orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonSaveArgs'>): <orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonCheckpointHandler object at 0x7c9e7877adb0>, ('metrics', <class 'orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonRestoreArgs'>): <orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonCheckpointHandler object at 0x7c9e7877adb0>}). I0423 16:02:59.882975 137110758524736 abstract_checkpointer.py:35] orbax-checkpoint version: 0.11.33 I0423 16:02:59.883057 137110758524736 async_checkpointer.py:177] [process=5][thread=MainThread] Using barrier_sync_fn: <function get_barrier_sync_fn.<locals>._fn at 0x7c9e78188180> timeout: 600 secs and primary_host=0 for async checkpoint writes I0423 16:03:00.617791 137110758524736 checkpoint_manager.py:1818] Found 0 checkpoint steps in gs://lance-maxtext/linen_ckpt_xpk_feat_nnx_set_defaults_true_20260423_155242/linen_xpk_feat_nnx_set_defaults_true_20260423_155242_04_int8/checkpoints I0423 16:03:00.644792 137110758524736 checkpoint_manager.py:929] [process=5][thread=MainThread] CheckpointManager created, primary_host=0, CheckpointManagerOptions=CheckpointManagerOptions(save_interval_steps=1, max_to_keep=None, keep_time_interval=None, keep_period=None, should_keep_fn=None, best_fn=None, best_mode='max', keep_checkpoints_without_metrics=True, step_prefix=None, step_format_fixed_length=None, step_name_format=None, create=True, cleanup_tmp_directories=False, save_on_steps=frozenset(), single_host_load_and_broadcast=False, todelete_subdir=None, todelete_full_path=None, enable_background_delete=False, read_only=False, enable_async_checkpointing=True, async_options=None, multiprocessing_options=MultiprocessingOptions(primary_host=0, active_processes=None, barrier_sync_key_prefix=None), should_save_fn=None, file_options=FileOptions(path_permission_mode=None), save_root_metadata=True, temporary_path_class=None, save_decision_policy=FixedIntervalPolicy(interval=10), preservation_policy=LatestN(n=None), prevent_write_metrics=False, enable_should_save_is_saving_in_progress_check=True, enable_per_process_directory_creation=False, lightweight_initialize=False), root_directory=gs://lance-maxtext/linen_ckpt_xpk_feat_nnx_set_defaults_true_20260423_155242/linen_xpk_feat_nnx_set_defaults_true_20260423_155242_04_int8/checkpoints: <orbax.checkpoint.checkpoint_manager.CheckpointManager object at 0x7cb1d0e5a180> I0423 16:03:00.644930 137110758524736 checkpointing.py:302] Checkpoint manager created! I0423 16:03:01.835285 137110758524736 nnx_wrappers.py:455] Unknown Logical: bfloat16[32,2048,2048]...................................... ('activation_batch', 'activation_norm_length', 'activation_embed'). I0423 16:03:01.835391 137110758524736 nnx_wrappers.py:455] Unknown Physical: bfloat16[32,2048,2048]...................................... ('fsdp', None, None). I0423 16:03:02.210670 137110758524736 attentions.py:1084] attentions/inputs_q Logical: bfloat16[32,2048,2048]...................................... ('activation_batch', 'activation_attn_length', 'activation_attn_embed'). I0423 16:03:02.210766 137110758524736 attentions.py:1084] attentions/inputs_q Physical: bfloat16[32,2048,2048]...................................... ('fsdp', None, None). I0423 16:03:02.226671 137110758524736 attentions.py:1085] attentions/inputs_kv Logical: bfloat16[32,2048,2048]...................................... ('activation_batch', 'activation_attn_length', 'activation_attn_embed'). I0423 16:03:02.226735 137110758524736 attentions.py:1085] attentions/inputs_kv Physical: bfloat16[32,2048,2048]...................................... ('fsdp', None, None). I0423 16:03:02.304568 137110758524736 attentions.py:1150] attentions/query Logical: bfloat16[32,2048,16,128].................................... ('activation_kv_batch', 'activation_attn_length', 'activation_kv_heads', 'activation_kv_head_dim'). I0423 16:03:02.304673 137110758524736 attentions.py:1150] attentions/query Physical: bfloat16[32,2048,16,128].................................... ('fsdp', None, None, None). I0423 16:03:02.320675 137110758524736 attentions.py:1151] attentions/key Logical: bfloat16[32,2048,16,128].................................... ('activation_kv_batch', 'activation_attn_length', 'activation_kv_heads', 'activation_kv_head_dim'). I0423 16:03:02.320736 137110758524736 attentions.py:1151] attentions/key Physical: bfloat16[32,2048,16,128].................................... ('fsdp', None, None, None). I0423 16:03:02.336688 137110758524736 attentions.py:1152] attentions/value Logical: bfloat16[32,2048,16,128].................................... ('activation_kv_batch', 'activation_attn_length', 'activation_kv_heads', 'activation_kv_head_dim'). I0423 16:03:02.336754 137110758524736 attentions.py:1152] attentions/value Physical: bfloat16[32,2048,16,128].................................... ('fsdp', None, None, None). I0423 16:03:02.360960 137110758524736 attentions.py:1193] attentions/out Logical: bfloat16[32,2048,16,128].................................... ('activation_batch', 'activation_attn_length', 'activation_heads', 'activation_kv'). I0423 16:03:02.361029 137110758524736 attentions.py:1193] attentions/out Physical: bfloat16[32,2048,16,128].................................... ('fsdp', None, None, None). I0423 16:03:02.441293 137110758524736 linears.py:541] linears/x Logical: bfloat16[32,2048,7168]...................................... ('activation_batch', 'activation_length', 'activation_mlp'). I0423 16:03:02.441382 137110758524736 linears.py:541] linears/x Physical: bfloat16[32,2048,7168]...................................... ('fsdp', None, None). I0423 16:03:03.039770 137110758524736 checkpointing.py:578] checkpoint manager exists so trying to load this run's existing checkpoint I0423 16:03:03.039881 137110758524736 checkpointing.py:676] No existing checkpoints found, not restoring checkpoint. fsdp: 32 I0423 16:03:05.723705 137110758524736 maxtext_utils.py:1740] params/params/decoder/decoder_norm/scale Shape: float32[2048] Logical: PartitionSpec('norm',) Physical: (None,) I0423 16:03:05.723844 137110758524736 maxtext_utils.py:1740] params/params/decoder/layers/mlp/wi_0/kernel Shape: float32[2048,16,7168] Logical: PartitionSpec('embed', 'layers', 'mlp') Physical: ('fsdp', None, None) I0423 16:03:05.723899 137110758524736 maxtext_utils.py:1740] params/params/decoder/layers/mlp/wi_1/kernel Shape: float32[2048,16,7168] Logical: PartitionSpec('embed', 'layers', 'mlp') Physical: ('fsdp', None, None) I0423 16:03:05.723958 137110758524736 maxtext_utils.py:1740] params/params/decoder/layers/mlp/wo/kernel Shape: float32[7168,16,2048] Logical: PartitionSpec('mlp', 'layers', 'embed') Physical: (None, None, 'fsdp') I0423 16:03:05.724016 137110758524736 maxtext_utils.py:1740] params/params/decoder/layers/post_self_attention_layer_norm/scale Shape: float32[2048,16] Logical: PartitionSpec('norm', 'layers') Physical: (None, None) I0423 16:03:05.724057 137110758524736 maxtext_utils.py:1740] params/params/decoder/layers/pre_self_attention_layer_norm/scale Shape: float32[2048,16] Logical: PartitionSpec('norm', 'layers') Physical: (None, None) I0423 16:03:05.724112 137110758524736 maxtext_utils.py:1740] params/params/decoder/layers/self_attention/key/kernel Shape: float32[2048,16,16,128] Logical: PartitionSpec('embed', 'layers', 'kv_heads', 'kv_head_dim') Physical: ('fsdp', None, None, None) I0423 16:03:05.724167 137110758524736 maxtext_utils.py:1740] params/params/decoder/layers/self_attention/out/kernel Shape: float32[16,16,128,2048] Logical: PartitionSpec('heads', 'layers', 'kv', 'embed') Physical: (None, None, None, 'fsdp') I0423 16:03:05.724214 137110758524736 maxtext_utils.py:1740] params/params/decoder/layers/self_attention/query/kernel Shape: float32[2048,16,16,128] Logical: PartitionSpec('embed', 'layers', 'q_heads', 'kv') Physical: ('fsdp', None, None, None) I0423 16:03:05.724255 137110758524736 maxtext_utils.py:1740] params/params/decoder/layers/self_attention/value/kernel Shape: float32[2048,16,16,128] Logical: PartitionSpec('embed', 'layers', 'kv_heads', 'kv_head_dim') Physical: ('fsdp', None, None, None) I0423 16:03:05.724305 137110758524736 maxtext_utils.py:1740] params/params/decoder/logits_dense/kernel Shape: float32[2048,32000] Logical: PartitionSpec('embed', 'vocab') Physical: ('fsdp', None) I0423 16:03:05.724353 137110758524736 maxtext_utils.py:1740] params/params/token_embedder/embedding Shape: float32[32000,2048] Logical: PartitionSpec('vocab', 'embed') Physical: (None, 'fsdp') I0423 16:03:07.421037 137110758524736 train.py:158] train/xent Logical: float32[32,2048]............................................ ('activation_embed_and_logits_batch', 'activation_length'). I0423 16:03:07.421131 137110758524736 train.py:158] train/xent Physical: float32[32,2048]............................................ ('fsdp', None). I0423 16:03:07.436030 137110758524736 train.py:165] train/z_loss Logical: float32[32,2048]............................................ ('activation_embed_and_logits_batch', 'activation_length'). I0423 16:03:07.436091 137110758524736 train.py:165] train/z_loss Physical: float32[32,2048]............................................ ('fsdp', None). I0423 16:03:21.374807 137110758524736 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 16:03:21.375750 137110758524736 metric_logger.py:289] number parameters: 1.104 billion I0423 16:03:36.393352 137110758524736 checkpointing.py:794] Waiting for step 0 to finish before checkpoint... I0423 16:03:36.634175 137110758524736 checkpointing.py:798] Waited 0.24080204963684082 seconds for step 0 to finish before starting checkpointing. I0423 16:03:36.636789 137110758524736 checkpoint_manager.py:2013] [process=5][thread=MainThread][wait_until_finished] No Save Finalize thread to wait for. Returning. I0423 16:03:36.638993 137110758524736 checkpoint_manager.py:1518] [process=5] Saving checkpoint at step 0 I0423 16:03:36.641809 137110758524736 async_checkpointer.py:452] [process=5] Started async saving checkpoint to gs://lance-maxtext/linen_ckpt_xpk_feat_nnx_set_defaults_true_20260423_155242/linen_xpk_feat_nnx_set_defaults_true_20260423_155242_04_int8/checkpoints/0. I0423 16:03:37.193236 137110758524736 signaling_client.py:364] Using JaxDistributedSignalingClient I0423 16:03:37.194364 137110758524736 jax_array_handlers.py:358] Scheduling D2H of 39 prioritized jax.Array. I0423 16:03:37.194423 137110758524736 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 16:03:37.443419 137110758524736 base_pytree_checkpoint_handler.py:153] [process=5][thread=MainThread] Initiated "orbax.checkpoint._src.serialization.jax_array_handlers.ArrayHandler".serialize. Time taken: 0.250262s I0423 16:03:37.443595 137110758524736 base_pytree_checkpoint_handler.py:129] [process=5] /jax/checkpoint/write/blocking_gbytes_per_sec: 3.116 GiB/s (total gbytes: 1.5 GiB) (time elapsed: 0.4950876235961914 s) (per-host) I0423 16:03:37.443661 137110758524736 base_pytree_checkpoint_handler.py:737] [process=5][thread=MainThread] Initiated Pytree async_save. Time taken: 0.495154s (batch_requests_ready=0.241279s, total_serialization_initiated=0.253801s, others=0.000074s) I0423 16:03:37.443764 137110758524736 composite_checkpoint_handler.py:715] [process=5][thread=MainThread] Initiated CompositeCheckpointHandler.async_save. Time taken: 0.499819s (all_items=0.000020s, per_item={'items': '0.00002003'}, temp_paths=0.499799) I0423 16:03:37.444688 136979544405760 async_checkpointer.py:79] [process=5][thread=async_save] Background save thread started. I0423 16:03:37.444857 137110758524736 async_checkpointer.py:561] Finished blocking save. Time taken: 0.805792s. Continuing background save to gs://lance-maxtext/linen_ckpt_xpk_feat_nnx_set_defaults_true_20260423_155242/linen_xpk_feat_nnx_set_defaults_true_20260423_155242_04_int8/checkpoints/0. I0423 16:03:37.471687 137110758524736 checkpoint_manager.py:1566] [process=5][thread=MainThread][step=0] Starting CheckpointManager Save Finalize thread=save_finalize I0423 16:03:37.471961 136974711101184 async_checkpointer.py:265] [process=5][thread=save_finalize] Waiting for background save thread=async_save. I0423 16:03:37.472089 137110758524736 standard_logger.py:34] {'step': 0, 'event_type': 'save', 'directory': 'gs://lance-maxtext/linen_ckpt_xpk_feat_nnx_set_defaults_true_20260423_155242/linen_xpk_feat_nnx_set_defaults_true_20260423_155242_04_int8/checkpoints', 'reached_preemption': False, 'preemption_received_at': None, 'synchronous': False, 'wait_for_prev_start_time': 1776960216.636771, 'wait_for_prev_duration_secs': 6.222724914550781e-05, 'time_between_consecutive_saves_sec': None, 'checkpointer_blocking_start_time': 1776960216.6390321, 'checkpointer_blocking_duration_secs': 0.8059325218200684, 'get_old_steps_start_time': 1776960217.4449852, 'get_old_steps_duration_secs': 2.7418136596679688e-05, 'checkpoint_manager_blocking_start_time': 1776960216.634802, 'checkpoint_manager_blocking_duration_secs': 0.8372478485107422} I0423 16:03:37.472202 137110758524736 checkpointing.py:409] Started an asynchronous checkpoint save for step 0 I0423 16:03:37.472256 137110758524736 max_utils.py:750] Memstats: After params initialized: I0423 16:03:37.472307 137110758524736 max_utils.py:756] Using (GB) 0.43 / 31.25 (1.376000%) on TPU_18(process=5,(2,4,0,0)) I0423 16:03:37.472343 137110758524736 max_utils.py:756] Using (GB) 0.43 / 31.25 (1.376000%) on TPU_19(process=5,(3,4,0,0)) I0423 16:03:37.472371 137110758524736 max_utils.py:756] Using (GB) 0.43 / 31.25 (1.376000%) on TPU_22(process=5,(2,5,0,0)) I0423 16:03:37.472395 137110758524736 max_utils.py:756] Using (GB) 0.43 / 31.25 (1.376000%) on TPU_23(process=5,(3,5,0,0)) I0423 16:03:37.745153 137110758524736 metric_logger.py:185] completed step: 0, seconds: 15.017, TFLOP/s/device: 0.905, Tokens/s/device: 136.374, total_weights: 65536, loss: 10.874 I0423 16:03:37.995212 137110758524736 metric_logger.py:185] completed step: 1, seconds: 1.350, TFLOP/s/device: 10.062, Tokens/s/device: 1516.614, total_weights: 65536, loss: 10.874 I0423 16:03:38.416160 137110758524736 metric_logger.py:185] completed step: 2, seconds: 0.022, TFLOP/s/device: 628.946, Tokens/s/device: 94801.648, total_weights: 65536, loss: 10.262 I0423 16:03:38.645746 137110758524736 metric_logger.py:185] completed step: 3, seconds: 0.421, TFLOP/s/device: 32.249, Tokens/s/device: 4860.959, total_weights: 65536, loss: 9.731 I0423 16:03:39.105033 137110758524736 metric_logger.py:185] completed step: 4, seconds: 0.234, TFLOP/s/device: 58.001, Tokens/s/device: 8742.610, total_weights: 65536, loss: 9.272 I0423 16:03:39.110827 137110758524736 metric_logger.py:185] completed step: 5, seconds: 0.229, TFLOP/s/device: 59.209, Tokens/s/device: 8924.681, total_weights: 65536, loss: 8.887 I0423 16:04:01.880767 137110758524736 metric_logger.py:185] completed step: 6, seconds: 0.460, TFLOP/s/device: 29.556, Tokens/s/device: 4455.031, total_weights: 65536, loss: 8.588 I0423 16:04:02.111871 137110758524736 metric_logger.py:185] completed step: 7, seconds: 22.542, TFLOP/s/device: 0.603, Tokens/s/device: 90.852, total_weights: 65536, loss: 8.380 I0423 16:04:02.341307 137110758524736 metric_logger.py:185] completed step: 8, seconds: 0.233, TFLOP/s/device: 58.231, Tokens/s/device: 8777.155, total_weights: 65536, loss: 8.251 I0423 16:04:02.571359 137110758524736 checkpointing.py:794] Waiting for step 9 to finish before checkpoint... I0423 16:04:02.572061 137110758524736 checkpointing.py:798] Waited 0.0007219314575195312 seconds for step 9 to finish before starting checkpointing. I0423 16:04:02.574041 137110758524736 checkpoint_manager.py:2024] [process=5][thread=MainThread][step=0][wait_until_finished] Waiting for Save Finalize thread (save_finalize) to complete. I0423 16:04:03.392278 2788 google_auth_provider.cc:181] Running on GCE, using service account 562977990677-compute@developer.gserviceaccount.com I0423 16:04:05.342007 136979593295616 array_metadata_store.py:203] [process=5][thread=array_type_handler] Wrote 39 array_metadata.ArrayMetadata to gs://lance-maxtext/linen_ckpt_xpk_feat_nnx_set_defaults_true_20260423_155242/linen_xpk_feat_nnx_set_defaults_true_20260423_155242_04_int8/checkpoints/0/items/array_metadatas/process_5 I0423 16:04:33.485913 136979544405760 base_pytree_checkpoint_handler.py:129] [process=5] /jax/checkpoint/write/gbytes_per_sec: 27.939 MiB/s (total gbytes: 1.5 GiB) (time elapsed: 56.53736853599548 s) (per-host) I0423 16:04:33.486018 136979544405760 async_checkpointer.py:90] [process=5][thread=async_save] 3 Handler Commit operations completed. Time taken: 56.041216s. I0423 16:04:41.643502 136979544405760 async_checkpointer.py:144] [process=5][thread=async_save] Background save thread done. Time taken: 64.198685s. I0423 16:04:41.643754 136974711101184 async_checkpointer.py:273] [process=5][thread=save_finalize] Done with waiting for background save thread=async_save. I0423 16:04:41.643822 136974711101184 async_checkpointer.py:283] [process=5][thread=save_finalize] No errors found in background save thread=async_save. I0423 16:04:41.643872 136974711101184 checkpoint_manager.py:2133] [process=5][thread=save_finalize][step=0] CheckpointManager Save Finalize is syncing with other hosts... I0423 16:04:41.646019 136974711101184 checkpoint_manager.py:2142] [process=5][thread=save_finalize][step=0] CheckpointManager Save Finalize is done on all hosts. I0423 16:04:41.646213 137110758524736 checkpoint_manager.py:2036] [process=5][thread=MainThread][step=0][wait_until_finished] Done waiting for Save Finalize thread (save_finalize) running at step=0. W0423 16:04:41.646359 137110758524736 checkpoint_manager.py:1458] Waiting for previous save to complete took 39.072322 seconds. If this number is high, consider checkpointing less frequently. I0423 16:04:41.648136 137110758524736 checkpoint_manager.py:1518] [process=5] Saving checkpoint at step 9 I0423 16:04:41.651686 137110758524736 async_checkpointer.py:452] [process=5] Started async saving checkpoint to gs://lance-maxtext/linen_ckpt_xpk_feat_nnx_set_defaults_true_20260423_155242/linen_xpk_feat_nnx_set_defaults_true_20260423_155242_04_int8/checkpoints/9. I0423 16:04:42.208838 137110758524736 jax_array_handlers.py:358] Scheduling D2H of 39 prioritized jax.Array. I0423 16:04:42.208935 137110758524736 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 16:04:42.251160 137110758524736 base_pytree_checkpoint_handler.py:153] [process=5][thread=MainThread] Initiated "orbax.checkpoint._src.serialization.jax_array_handlers.ArrayHandler".serialize. Time taken: 0.043629s I0423 16:04:42.251327 137110758524736 base_pytree_checkpoint_handler.py:129] [process=5] /jax/checkpoint/write/blocking_gbytes_per_sec: 5.205 GiB/s (total gbytes: 1.5 GiB) (time elapsed: 0.2963752746582031 s) (per-host) I0423 16:04:42.251379 137110758524736 base_pytree_checkpoint_handler.py:737] [process=5][thread=MainThread] Initiated Pytree async_save. Time taken: 0.296437s (batch_requests_ready=0.250911s, total_serialization_initiated=0.045459s, others=0.000067s) I0423 16:04:42.251468 137110758524736 composite_checkpoint_handler.py:715] [process=5][thread=MainThread] Initiated CompositeCheckpointHandler.async_save. Time taken: 0.300573s (all_items=0.000016s, per_item={'items': '0.00001621'}, temp_paths=0.300557) I0423 16:04:42.252366 136978519553792 async_checkpointer.py:79] [process=5][thread=async_save] Background save thread started. I0423 16:04:42.252492 137110758524736 async_checkpointer.py:561] Finished blocking save. Time taken: 0.604287s. Continuing background save to gs://lance-maxtext/linen_ckpt_xpk_feat_nnx_set_defaults_true_20260423_155242/linen_xpk_feat_nnx_set_defaults_true_20260423_155242_04_int8/checkpoints/9. I0423 16:04:42.271619 137110758524736 checkpoint_manager.py:1566] [process=5][thread=MainThread][step=9] Starting CheckpointManager Save Finalize thread=save_finalize I0423 16:04:42.271935 136974711101184 async_checkpointer.py:265] [process=5][thread=save_finalize] Waiting for background save thread=async_save. I0423 16:04:42.272103 137110758524736 standard_logger.py:34] {'step': 9, 'event_type': 'save', 'directory': 'gs://lance-maxtext/linen_ckpt_xpk_feat_nnx_set_defaults_true_20260423_155242/linen_xpk_feat_nnx_set_defaults_true_20260423_155242_04_int8/checkpoints', 'reached_preemption': False, 'preemption_received_at': None, 'synchronous': False, 'wait_for_prev_start_time': 1776960242.574013, 'wait_for_prev_duration_secs': 39.07232213020325, 'time_between_consecutive_saves_sec': None, 'checkpointer_blocking_start_time': 1776960281.648175, 'checkpointer_blocking_duration_secs': 0.6045033931732178, 'get_old_steps_start_time': 1776960282.2526991, 'get_old_steps_duration_secs': 2.7179718017578125e-05, 'checkpoint_manager_blocking_start_time': 1776960242.5722954, 'checkpoint_manager_blocking_duration_secs': 39.699772357940674} I0423 16:04:42.272212 137110758524736 checkpointing.py:409] Started an asynchronous checkpoint save for step 9 I0423 16:04:42.272257 137110758524736 checkpoint_manager.py:2024] [process=5][thread=MainThread][step=9][wait_until_finished] Waiting for Save Finalize thread (save_finalize) to complete. I0423 16:04:47.013745 136979593295616 array_metadata_store.py:203] [process=5][thread=array_type_handler] Wrote 39 array_metadata.ArrayMetadata to gs://lance-maxtext/linen_ckpt_xpk_feat_nnx_set_defaults_true_20260423_155242/linen_xpk_feat_nnx_set_defaults_true_20260423_155242_04_int8/checkpoints/9/items/array_metadatas/process_5 I0423 16:05:24.689878 136978519553792 base_pytree_checkpoint_handler.py:129] [process=5] /jax/checkpoint/write/gbytes_per_sec: 36.963 MiB/s (total gbytes: 1.5 GiB) (time elapsed: 42.734886169433594 s) (per-host) I0423 16:05:24.690002 136978519553792 async_checkpointer.py:90] [process=5][thread=async_save] 3 Handler Commit operations completed. Time taken: 42.437430s. I0423 16:05:32.990792 136978519553792 async_checkpointer.py:144] [process=5][thread=async_save] Background save thread done. Time taken: 50.738205s. I0423 16:05:32.991100 136974711101184 async_checkpointer.py:273] [process=5][thread=save_finalize] Done with waiting for background save thread=async_save. I0423 16:05:32.991220 136974711101184 async_checkpointer.py:283] [process=5][thread=save_finalize] No errors found in background save thread=async_save. I0423 16:05:32.991268 136974711101184 checkpoint_manager.py:2133] [process=5][thread=save_finalize][step=9] CheckpointManager Save Finalize is syncing with other hosts... I0423 16:05:32.992908 136974711101184 checkpoint_manager.py:2142] [process=5][thread=save_finalize][step=9] CheckpointManager Save Finalize is done on all hosts. I0423 16:05:32.993081 137110758524736 checkpoint_manager.py:2036] [process=5][thread=MainThread][step=9][wait_until_finished] Done waiting for Save Finalize thread (save_finalize) running at step=9. I0423 16:05:32.993233 137110758524736 checkpoint_manager.py:2013] [process=5][thread=MainThread][wait_until_finished] No Save Finalize thread to wait for. Returning. I0423 16:05:32.993892 137110758524736 metric_logger.py:185] completed step: 9, seconds: 0.231, TFLOP/s/device: 58.797, Tokens/s/device: 8862.501, total_weights: 65536, loss: 8.175 Per train step: Total TFLOPs: 13.59 split as 93.93% learnable weight flops and 6.07% attention flops XPK End: Thu Apr 23 16:05:45 UTC 2026 EXIT_CODE=0