main (8a17c3d19) vs feat/nnx-set-defaults-true (73213e044)| Metric | main 8a17c3d19 | feat/nnx-set-defaults-true 73213e044 | Diff (feat/nnx-set-defaults-true − main) |
|---|---|---|---|
| Parameters | 1.104 billion | 1.104 billion | — |
| Final loss | 8.1790 | 8.1790 | 0 |
| TFLOP/s | 92.293 | 92.009 | -0.284 |
| Tok/s | 13911.3 | 13868.6 | -42.768 |
| Avg s/step | 2.664 | 2.389 | -0.275 |
| 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: Wed Apr 22 07:18:16 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. 2026-04-22 07:18:41.004172: 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) I0422 07:18:41.214670 138450151987008 max_utils.py:273] Attempting to initialize the jax distributed system... I0422 07:18:50.256325 138450151987008 distributed.py:149] Starting JAX distributed service on [::]:8482 I0422 07:18:50.258615 138450151987008 distributed.py:172] Connecting to JAX distributed service on mt-02-synthetic-lt7wg-slice-job-0-0.mt-02-synthetic-lt7wg:8482 I0422 07:18:52.331493 138450151987008 max_utils.py:284] Jax distributed system initialized! I0422 07:18:58.472026 138450151987008 max_utils.py:800] System Information: Jax Version: 0.9.2 I0422 07:18:58.472136 138450151987008 max_utils.py:801] System Information: Jaxlib Version: 0.9.2 I0422 07:18:58.472176 138450151987008 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 I0422 07:18:58.472212 138450151987008 train_utils.py:361] WARNING: Sequence packing is essentially ignored for synthetic data. Please use a real dataset to use sequence packing. I0422 07:18:59.161847 138450151987008 maxtext_utils.py:1565] Num_devices: 32, shape (1, 1, 1, 32, 1, 1, 1, 1, 1, 1, 1, 1, 1) I0422 07:18:59.162125 138450151987008 checkpointing.py:677] Setting up checkpoint logger... I0422 07:18:59.162187 138450151987008 checkpointing.py:233] Creating checkpoint manager with ocdbt=True and zarr3=True I0422 07:18:59.162231 138450151987008 pytree_checkpoint_handler.py:592] save_device_host_concurrent_bytes=None I0422 07:18:59.162565 138450151987008 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 0x7deacffb17c0>, enable_pinned_host_transfer=False, save_concurrent_bytes: 96000000000 (89.4 GiB), restore_concurrent_bytes: 96000000000 (89.4 GiB) I0422 07:19:02.004953 138450151987008 checkpointing.py:265] Enabling policy for fixed interval checkpointing. I0422 07:19:02.005195 138450151987008 checkpoint_manager.py:708] [process=4][thread=MainThread] CheckpointManager init: checkpointers=None, item_names=('items',), item_handlers={'items': <orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeCheckpointHandler object at 0x7dd6a40809b0>}, handler_registry=None I0422 07:19:02.005432 138450151987008 composite_checkpoint_handler.py:237] Deferred registration for item: "items". Adding handler `<orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeCheckpointHandler object at 0x7dd6a40809b0>` 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`. I0422 07:19:02.005481 138450151987008 composite_checkpoint_handler.py:237] Deferred registration for item: "metrics". Adding handler `<orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonCheckpointHandler object at 0x7dd684700500>` 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`. I0422 07:19:02.005518 138450151987008 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 0x7dd6a40809b0>, ('items', <class 'orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeRestoreArgs'>): <orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeCheckpointHandler object at 0x7dd6a40809b0>, ('metrics', <class 'orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonSaveArgs'>): <orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonCheckpointHandler object at 0x7dd684700500>, ('metrics', <class 'orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonRestoreArgs'>): <orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonCheckpointHandler object at 0x7dd684700500>}). I0422 07:19:02.005851 138450151987008 abstract_checkpointer.py:35] orbax-checkpoint version: 0.11.34 I0422 07:19:02.005922 138450151987008 async_checkpointer.py:192] [process=4][thread=MainThread] Using barrier_sync_fn: <function get_barrier_sync_fn.<locals>._fn at 0x7dd5b047d580> timeout: 1200 secs and primary_host=0 for async checkpoint writes I0422 07:19:04.137309 138450151987008 checkpoint_manager.py:1812] Found 0 checkpoint steps in gs://lance-maxtext/linen_ckpt_xpk_main_20260422_071422/linen_xpk_main_20260422_071422_02_synthetic/checkpoints I0422 07:19:04.139620 138450151987008 checkpoint_manager.py:929] [process=4][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_20260422_071422/linen_xpk_main_20260422_071422_02_synthetic/checkpoints: <orbax.checkpoint.checkpoint_manager.CheckpointManager object at 0x7dd6846fecc0> I0422 07:19:04.139750 138450151987008 checkpointing.py:301] Checkpoint manager created! I0422 07:19:05.062641 138450151987008 nnx_wrappers.py:437] Unknown Logical: bfloat16[32,2048,2048]...................................... ('activation_batch', 'activation_norm_length', 'activation_embed'). I0422 07:19:05.062774 138450151987008 nnx_wrappers.py:437] Unknown Physical: bfloat16[32,2048,2048]...................................... ('fsdp', None, None). I0422 07:19:05.443550 138450151987008 attentions.py:1088] attentions/inputs_q Logical: bfloat16[32,2048,2048]...................................... ('activation_batch', 'activation_attn_length', 'activation_attn_embed'). I0422 07:19:05.443641 138450151987008 attentions.py:1088] attentions/inputs_q Physical: bfloat16[32,2048,2048]...................................... ('fsdp', None, None). I0422 07:19:05.465356 138450151987008 attentions.py:1089] attentions/inputs_kv Logical: bfloat16[32,2048,2048]...................................... ('activation_batch', 'activation_attn_length', 'activation_attn_embed'). I0422 07:19:05.465457 138450151987008 attentions.py:1089] attentions/inputs_kv Physical: bfloat16[32,2048,2048]...................................... ('fsdp', None, None). I0422 07:19:05.489353 138450151987008 attentions.py:1154] attentions/query Logical: bfloat16[32,2048,16,128].................................... ('activation_kv_batch', 'activation_attn_length', 'activation_kv_heads', 'activation_kv_head_dim'). I0422 07:19:05.489425 138450151987008 attentions.py:1154] attentions/query Physical: bfloat16[32,2048,16,128].................................... ('fsdp', None, None, None). I0422 07:19:05.505884 138450151987008 attentions.py:1155] attentions/key Logical: bfloat16[32,2048,16,128].................................... ('activation_kv_batch', 'activation_attn_length', 'activation_kv_heads', 'activation_kv_head_dim'). I0422 07:19:05.505957 138450151987008 attentions.py:1155] attentions/key Physical: bfloat16[32,2048,16,128].................................... ('fsdp', None, None, None). I0422 07:19:05.522363 138450151987008 attentions.py:1156] attentions/value Logical: bfloat16[32,2048,16,128].................................... ('activation_kv_batch', 'activation_attn_length', 'activation_kv_heads', 'activation_kv_head_dim'). I0422 07:19:05.522426 138450151987008 attentions.py:1156] attentions/value Physical: bfloat16[32,2048,16,128].................................... ('fsdp', None, None, None). I0422 07:19:05.547207 138450151987008 attentions.py:1198] attentions/out Logical: bfloat16[32,2048,16,128].................................... ('activation_batch', 'activation_attn_length', 'activation_heads', 'activation_kv'). I0422 07:19:05.547276 138450151987008 attentions.py:1198] attentions/out Physical: bfloat16[32,2048,16,128].................................... ('fsdp', None, None, None). I0422 07:19:05.568303 138450151987008 linears.py:525] linears/x Logical: bfloat16[32,2048,7168]...................................... ('activation_batch', 'activation_length', 'activation_mlp'). I0422 07:19:05.568372 138450151987008 linears.py:525] linears/x Physical: bfloat16[32,2048,7168]...................................... ('fsdp', None, None). I0422 07:19:05.780521 138450151987008 checkpointing.py:577] checkpoint manager exists so trying to load this run's existing checkpoint I0422 07:19:05.780652 138450151987008 checkpointing.py:665] No existing checkpoints found, not restoring checkpoint. fsdp: 32 I0422 07:19:07.201890 138450151987008 maxtext_utils.py:1668] params/params/decoder/decoder_norm/scale Shape: float32[2048] Logical: P('norm',) Physical: (None,) I0422 07:19:07.202021 138450151987008 maxtext_utils.py:1668] params/params/decoder/layers/mlp/wi_0/kernel Shape: float32[2048,16,7168] Logical: P('embed', 'layers', 'mlp') Physical: ('fsdp', None, None) I0422 07:19:07.202075 138450151987008 maxtext_utils.py:1668] params/params/decoder/layers/mlp/wi_1/kernel Shape: float32[2048,16,7168] Logical: P('embed', 'layers', 'mlp') Physical: ('fsdp', None, None) I0422 07:19:07.202135 138450151987008 maxtext_utils.py:1668] params/params/decoder/layers/mlp/wo/kernel Shape: float32[7168,16,2048] Logical: P('mlp', 'layers', 'embed') Physical: (None, None, 'fsdp') I0422 07:19:07.202189 138450151987008 maxtext_utils.py:1668] params/params/decoder/layers/post_self_attention_layer_norm/scale Shape: float32[2048,16] Logical: P('norm', 'layers') Physical: (None, None) I0422 07:19:07.202229 138450151987008 maxtext_utils.py:1668] params/params/decoder/layers/pre_self_attention_layer_norm/scale Shape: float32[2048,16] Logical: P('norm', 'layers') Physical: (None, None) I0422 07:19:07.202281 138450151987008 maxtext_utils.py:1668] params/params/decoder/layers/self_attention/key/kernel Shape: float32[2048,16,16,128] Logical: P('embed', 'layers', 'kv_heads', 'kv_head_dim') Physical: ('fsdp', None, None, None) I0422 07:19:07.202333 138450151987008 maxtext_utils.py:1668] params/params/decoder/layers/self_attention/out/kernel Shape: float32[16,16,128,2048] Logical: P('heads', 'layers', 'kv', 'embed') Physical: (None, None, None, 'fsdp') I0422 07:19:07.202374 138450151987008 maxtext_utils.py:1668] params/params/decoder/layers/self_attention/query/kernel Shape: float32[2048,16,16,128] Logical: P('embed', 'layers', 'q_heads', 'kv') Physical: ('fsdp', None, None, None) I0422 07:19:07.202414 138450151987008 maxtext_utils.py:1668] params/params/decoder/layers/self_attention/value/kernel Shape: float32[2048,16,16,128] Logical: P('embed', 'layers', 'kv_heads', 'kv_head_dim') Physical: ('fsdp', None, None, None) I0422 07:19:07.202464 138450151987008 maxtext_utils.py:1668] params/params/decoder/logits_dense/kernel Shape: float32[2048,32000] Logical: P('embed_vocab', 'vocab') Physical: ('fsdp', None) I0422 07:19:07.202514 138450151987008 maxtext_utils.py:1668] params/params/token_embedder/embedding Shape: float32[32000,2048] Logical: P('vocab', 'embed_vocab') Physical: (None, 'fsdp') I0422 07:19:07.694016 138450151987008 train.py:155] train/xent Logical: float32[32,2048]............................................ ('activation_embed_and_logits_batch', 'activation_length'). I0422 07:19:07.694110 138450151987008 train.py:155] train/xent Physical: float32[32,2048]............................................ ('fsdp', None). I0422 07:19:07.709539 138450151987008 train.py:162] train/z_loss Logical: float32[32,2048]............................................ ('activation_embed_and_logits_batch', 'activation_length'). I0422 07:19:07.709601 138450151987008 train.py:162] train/z_loss Physical: float32[32,2048]............................................ ('fsdp', None). I0422 07:19:18.805956 138450151987008 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. I0422 07:19:18.806715 138450151987008 metric_logger.py:301] number parameters: 1.104 billion I0422 07:19:30.121442 138450151987008 checkpointing.py:772] Waiting for step 0 to finish before checkpoint... I0422 07:19:30.402004 138450151987008 checkpointing.py:776] Waited 0.28054285049438477 seconds for step 0 to finish before starting checkpointing. I0422 07:19:30.406497 138450151987008 checkpoint_manager.py:2009] [process=4][thread=MainThread][wait_until_finished] No Save Finalize thread to wait for. Returning. I0422 07:19:30.407977 138450151987008 checkpoint_manager.py:1512] [process=4] Saving checkpoint at step 0 I0422 07:19:30.409474 138450151987008 event_tracking.py:70] [process=4] [async] Started save checkpoint @ gs://lance-maxtext/linen_ckpt_xpk_main_20260422_071422/linen_xpk_main_20260422_071422_02_synthetic/checkpoints/0. I0422 07:19:30.741931 138450151987008 signaling_client.py:364] Using JaxDistributedSignalingClient I0422 07:19:30.742975 138450151987008 jax_array_handlers.py:360] Scheduling D2H of 39 prioritized jax.Array. I0422 07:19:30.743031 138450151987008 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 I0422 07:19:31.019124 138450151987008 base_pytree_checkpoint_handler.py:154] [process=4][thread=MainThread] Initiated "orbax.checkpoint._src.serialization.jax_array_handlers.ArrayHandler".serialize. Time taken: 0.277266s I0422 07:19:31.019295 138450151987008 base_pytree_checkpoint_handler.py:130] [process=4] /jax/orbax/write/blocking_gbytes_per_sec: 5.455 GiB/s (total gbytes: 1.5 GiB) (time elapsed: 0.2827951908111572 s) (per-host) I0422 07:19:31.019347 138450151987008 base_pytree_checkpoint_handler.py:768] [process=4][thread=MainThread] Initiated Pytree async_save. Time taken: 0.282857s (batch_requests_ready=0.002343s, total_serialization_initiated=0.280445s, others=0.000070s) I0422 07:19:31.019444 138450151987008 composite_checkpoint_handler.py:715] [process=4][thread=MainThread] Initiated CompositeCheckpointHandler.async_save. Time taken: 0.287462s (all_items=0.000018s, per_item={'items': '0.00001812'}, temp_paths=0.287444) I0422 07:19:31.020288 138450151987008 event_tracking.py:125] [process=4] [async] Finished blocking save in 0.61 seconds. Continuing save @ gs://lance-maxtext/linen_ckpt_xpk_main_20260422_071422/linen_xpk_main_20260422_071422_02_synthetic/checkpoints/0. I0422 07:19:31.020635 138322441664256 async_checkpointer.py:76] [process=4][thread=async_save] Background save thread started. Deadline for this save operation is 2026-04-22 07:39:31.020598 I0422 07:19:31.030309 138450151987008 checkpoint_manager.py:1560] [process=4][thread=MainThread][step=0] Starting CheckpointManager Save Finalize thread=save_finalize I0422 07:19:31.030594 138317735651072 async_checkpointer.py:280] [process=4][thread=save_finalize] Waiting for background save thread=async_save. I0422 07:19:31.030761 138450151987008 standard_logger.py:34] {'step': 0, 'event_type': 'save', 'directory': 'gs://lance-maxtext/linen_ckpt_xpk_main_20260422_071422/linen_xpk_main_20260422_071422_02_synthetic/checkpoints', 'reached_preemption': False, 'preemption_received_at': None, 'synchronous': False, 'wait_for_prev_start_time': 1776842370.406478, 'wait_for_prev_duration_secs': 6.175041198730469e-05, 'time_between_consecutive_saves_sec': None, 'checkpointer_blocking_start_time': 1776842370.4080174, 'checkpointer_blocking_duration_secs': 0.6127405166625977, 'get_old_steps_start_time': 1776842371.020784, 'get_old_steps_duration_secs': 3.147125244140625e-05, 'checkpoint_manager_blocking_start_time': 1776842370.4025886, 'checkpoint_manager_blocking_duration_secs': 0.6281337738037109} I0422 07:19:31.030867 138450151987008 checkpointing.py:408] Started an asynchronous checkpoint save for step 0 I0422 07:19:31.030917 138450151987008 max_utils.py:750] Memstats: After params initialized: I0422 07:19:31.030965 138450151987008 max_utils.py:756] Using (GB) 0.43 / 31.25 (1.376000%) on TPU_16(process=4,(0,4,0,0)) I0422 07:19:31.030998 138450151987008 max_utils.py:756] Using (GB) 0.43 / 31.25 (1.376000%) on TPU_17(process=4,(1,4,0,0)) I0422 07:19:31.031026 138450151987008 max_utils.py:756] Using (GB) 0.43 / 31.25 (1.376000%) on TPU_20(process=4,(0,5,0,0)) I0422 07:19:31.031052 138450151987008 max_utils.py:756] Using (GB) 0.43 / 31.25 (1.376000%) on TPU_21(process=4,(1,5,0,0)) I0422 07:19:31.346553 138450151987008 metric_logger.py:196] completed step: 0, seconds: 11.315, TFLOP/s/device: 1.201, Tokens/s/device: 181.005, total_weights: 65536, loss: 10.874, lm_loss: 10.874, perplexity: 52779.875 I0422 07:19:31.534859 138450151987008 metric_logger.py:196] completed step: 1, seconds: 1.224, TFLOP/s/device: 11.105, Tokens/s/device: 1673.806, total_weights: 65536, loss: 10.874, lm_loss: 10.874, perplexity: 52779.875 I0422 07:19:31.923787 138450151987008 metric_logger.py:196] completed step: 2, seconds: 0.040, TFLOP/s/device: 343.943, Tokens/s/device: 51842.851, total_weights: 65536, loss: 10.262, lm_loss: 10.262, perplexity: 28637.129 I0422 07:19:32.071157 138450151987008 metric_logger.py:196] completed step: 3, seconds: 0.391, TFLOP/s/device: 34.793, Tokens/s/device: 5244.384, total_weights: 65536, loss: 9.734, lm_loss: 9.734, perplexity: 16889.203 I0422 07:19:32.365962 138450151987008 metric_logger.py:196] completed step: 4, seconds: 0.153, TFLOP/s/device: 88.816, Tokens/s/device: 13387.371, total_weights: 65536, loss: 9.277, lm_loss: 9.277, perplexity: 10694.614 I0422 07:19:32.371506 138450151987008 metric_logger.py:196] completed step: 5, seconds: 0.147, TFLOP/s/device: 92.205, Tokens/s/device: 13898.126, total_weights: 65536, loss: 8.892, lm_loss: 8.892, perplexity: 7270.717 I0422 07:19:33.983350 2792 google_auth_provider.cc:181] Running on GCE, using service account 562977990677-compute@developer.gserviceaccount.com I0422 07:19:36.444442 138317744043776 array_metadata_store.py:203] [process=4][thread=array_type_handler] Wrote 39 array_metadata.ArrayMetadata to gs://lance-maxtext/linen_ckpt_xpk_main_20260422_071422/linen_xpk_main_20260422_071422_02_synthetic/checkpoints/0/items/array_metadatas/process_4 I0422 07:19:53.941778 138450151987008 metric_logger.py:196] completed step: 6, seconds: 0.295, TFLOP/s/device: 46.034, Tokens/s/device: 6938.751, total_weights: 65536, loss: 8.592, lm_loss: 8.592, perplexity: 5390.759 I0422 07:19:54.088933 138450151987008 metric_logger.py:196] completed step: 7, seconds: 21.423, TFLOP/s/device: 0.634, Tokens/s/device: 95.596, total_weights: 65536, loss: 8.384, lm_loss: 8.384, perplexity: 4376.905 I0422 07:19:54.236409 138450151987008 metric_logger.py:196] completed step: 8, seconds: 0.152, TFLOP/s/device: 89.324, Tokens/s/device: 13463.852, total_weights: 65536, loss: 8.255, lm_loss: 8.255, perplexity: 3846.406 I0422 07:19:54.382993 138450151987008 checkpointing.py:772] Waiting for step 9 to finish before checkpoint... I0422 07:19:54.383660 138450151987008 checkpointing.py:776] Waited 0.0006885528564453125 seconds for step 9 to finish before starting checkpointing. I0422 07:19:54.385748 138450151987008 checkpoint_manager.py:2020] [process=4][thread=MainThread][step=0][wait_until_finished] Waiting for Save Finalize thread (save_finalize) to complete. I0422 07:20:07.981016 138322441664256 base_pytree_checkpoint_handler.py:130] [process=4] /jax/orbax/write/gbytes_per_sec: 42.412 MiB/s (total gbytes: 1.5 GiB) (time elapsed: 37.24447321891785 s) (per-host) I0422 07:20:07.981143 138322441664256 async_checkpointer.py:90] [process=4][thread=async_save] 3 Handler Commit operations completed. Time taken: 36.959847s. I0422 07:20:15.195404 138322441664256 async_checkpointer.py:160] [process=4][thread=async_save] Background save thread done. Time taken: 44.174091s. I0422 07:20:15.195731 138317735651072 async_checkpointer.py:288] [process=4][thread=save_finalize] Done with waiting for background save thread=async_save. I0422 07:20:15.195865 138317735651072 async_checkpointer.py:298] [process=4][thread=save_finalize] No errors found in background save thread=async_save. I0422 07:20:15.195915 138317735651072 checkpoint_manager.py:2137] [process=4][thread=save_finalize][step=0] CheckpointManager Save Finalize is syncing with other hosts... I0422 07:20:15.197429 138317735651072 checkpoint_manager.py:2146] [process=4][thread=save_finalize][step=0] CheckpointManager Save Finalize is done on all hosts. I0422 07:20:15.197607 138450151987008 checkpoint_manager.py:2032] [process=4][thread=MainThread][step=0][wait_until_finished] Done waiting for Save Finalize thread (save_finalize) running at step=0. W0422 07:20:15.197757 138450151987008 checkpoint_manager.py:1452] Waiting for previous save to complete took 20.812008 seconds. If this number is high, consider checkpointing less frequently. I0422 07:20:15.199619 138450151987008 checkpoint_manager.py:1512] [process=4] Saving checkpoint at step 9 I0422 07:20:15.201570 138450151987008 event_tracking.py:70] [process=4] [async] Started save checkpoint @ gs://lance-maxtext/linen_ckpt_xpk_main_20260422_071422/linen_xpk_main_20260422_071422_02_synthetic/checkpoints/9. I0422 07:20:15.924078 138450151987008 jax_array_handlers.py:360] Scheduling D2H of 39 prioritized jax.Array. I0422 07:20:15.924175 138450151987008 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 I0422 07:20:15.957827 138450151987008 base_pytree_checkpoint_handler.py:154] [process=4][thread=MainThread] Initiated "orbax.checkpoint._src.serialization.jax_array_handlers.ArrayHandler".serialize. Time taken: 0.034847s I0422 07:20:15.957997 138450151987008 base_pytree_checkpoint_handler.py:130] [process=4] /jax/orbax/write/blocking_gbytes_per_sec: 39.487 GiB/s (total gbytes: 1.5 GiB) (time elapsed: 0.03906583786010742 s) (per-host) I0422 07:20:15.958050 138450151987008 base_pytree_checkpoint_handler.py:768] [process=4][thread=MainThread] Initiated Pytree async_save. Time taken: 0.039128s (batch_requests_ready=0.002587s, total_serialization_initiated=0.036470s, others=0.000071s) I0422 07:20:15.958142 138450151987008 composite_checkpoint_handler.py:715] [process=4][thread=MainThread] Initiated CompositeCheckpointHandler.async_save. Time taken: 0.043256s (all_items=0.000016s, per_item={'items': '0.00001597'}, temp_paths=0.043240) I0422 07:20:15.958824 138450151987008 event_tracking.py:125] [process=4] [async] Finished blocking save in 0.76 seconds. Continuing save @ gs://lance-maxtext/linen_ckpt_xpk_main_20260422_071422/linen_xpk_main_20260422_071422_02_synthetic/checkpoints/9. I0422 07:20:15.959152 138317735651072 async_checkpointer.py:76] [process=4][thread=async_save] Background save thread started. Deadline for this save operation is 2026-04-22 07:40:15.959115 I0422 07:20:15.961089 138450151987008 checkpoint_manager.py:1560] [process=4][thread=MainThread][step=9] Starting CheckpointManager Save Finalize thread=save_finalize I0422 07:20:15.961394 138296361461504 async_checkpointer.py:280] [process=4][thread=save_finalize] Waiting for background save thread=async_save. I0422 07:20:15.961561 138450151987008 standard_logger.py:34] {'step': 9, 'event_type': 'save', 'directory': 'gs://lance-maxtext/linen_ckpt_xpk_main_20260422_071422/linen_xpk_main_20260422_071422_02_synthetic/checkpoints', 'reached_preemption': False, 'preemption_received_at': None, 'synchronous': False, 'wait_for_prev_start_time': 1776842394.3857193, 'wait_for_prev_duration_secs': 20.812007665634155, 'time_between_consecutive_saves_sec': None, 'checkpointer_blocking_start_time': 1776842415.199713, 'checkpointer_blocking_duration_secs': 0.7595860958099365, 'get_old_steps_start_time': 1776842415.9593232, 'get_old_steps_duration_secs': 3.170967102050781e-05, 'checkpoint_manager_blocking_start_time': 1776842394.3839035, 'checkpoint_manager_blocking_duration_secs': 21.577622652053833} I0422 07:20:15.961671 138450151987008 checkpointing.py:408] Started an asynchronous checkpoint save for step 9 I0422 07:20:15.961726 138450151987008 checkpoint_manager.py:2020] [process=4][thread=MainThread][step=9][wait_until_finished] Waiting for Save Finalize thread (save_finalize) to complete. I0422 07:20:22.008949 138317744043776 array_metadata_store.py:203] [process=4][thread=array_type_handler] Wrote 39 array_metadata.ArrayMetadata to gs://lance-maxtext/linen_ckpt_xpk_main_20260422_071422/linen_xpk_main_20260422_071422_02_synthetic/checkpoints/9/items/array_metadatas/process_4 I0422 07:20:58.924137 138317735651072 base_pytree_checkpoint_handler.py:130] [process=4] /jax/orbax/write/gbytes_per_sec: 36.730 MiB/s (total gbytes: 1.5 GiB) (time elapsed: 43.00516629219055 s) (per-host) I0422 07:20:58.924270 138317735651072 async_checkpointer.py:90] [process=4][thread=async_save] 3 Handler Commit operations completed. Time taken: 42.965002s. I0422 07:21:06.268014 138317735651072 async_checkpointer.py:160] [process=4][thread=async_save] Background save thread done. Time taken: 50.308731s. I0422 07:21:06.268308 138296361461504 async_checkpointer.py:288] [process=4][thread=save_finalize] Done with waiting for background save thread=async_save. I0422 07:21:06.268443 138296361461504 async_checkpointer.py:298] [process=4][thread=save_finalize] No errors found in background save thread=async_save. I0422 07:21:06.268492 138296361461504 checkpoint_manager.py:2137] [process=4][thread=save_finalize][step=9] CheckpointManager Save Finalize is syncing with other hosts... I0422 07:21:06.270091 138296361461504 checkpoint_manager.py:2146] [process=4][thread=save_finalize][step=9] CheckpointManager Save Finalize is done on all hosts. I0422 07:21:06.270284 138450151987008 checkpoint_manager.py:2032] [process=4][thread=MainThread][step=9][wait_until_finished] Done waiting for Save Finalize thread (save_finalize) running at step=9. I0422 07:21:06.270435 138450151987008 checkpoint_manager.py:2009] [process=4][thread=MainThread][wait_until_finished] No Save Finalize thread to wait for. Returning. I0422 07:21:06.271422 138450151987008 metric_logger.py:196] completed step: 9, seconds: 0.147, TFLOP/s/device: 92.293, Tokens/s/device: 13911.342, total_weights: 65536, loss: 8.179, lm_loss: 8.179, perplexity: 3564.635 Per train step: Total TFLOPs: 13.59 split as 93.93% learnable weight flops and 6.07% attention flops XPK End: Wed Apr 22 07:21:16 UTC 2026 EXIT_CODE=0
XPK Start: Wed Apr 22 15:52:10 UTC 2026 2026-04-22 15:52:14.568853: 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 00:00:1776873134.582602 10 cuda_dnn.cc:8579] Unable to register cuDNN factory: Attempting to register factory for plugin cuDNN when one has already been registered E0000 00:00:1776873134.586497 10 cuda_blas.cc:1407] Unable to register cuBLAS factory: Attempting to register factory for plugin cuBLAS when one has already been registered W0000 00:00:1776873134.598127 10 computation_placer.cc:177] computation placer already registered. Please check linkage and avoid linking the same target more than once. W0000 00:00:1776873134.598152 10 computation_placer.cc:177] computation placer already registered. Please check linkage and avoid linking the same target more than once. W0000 00:00:1776873134.598155 10 computation_placer.cc:177] computation placer already registered. Please check linkage and avoid linking the same target more than once. W0000 00:00:1776873134.598157 10 computation_placer.cc:177] computation placer already registered. Please check linkage and avoid linking the same target more than once. 2026-04-22 15:52:34.261553: 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) I0422 15:52:34.789295 140014694840128 max_utils.py:273] Attempting to initialize the jax distributed system... INFO:2026-04-22 15:52:43,830:jax._src.distributed:140: Starting JAX distributed service on [::]:8482 I0422 15:52:43.830500 140014694840128 distributed.py:140] Starting JAX distributed service on [::]:8482 INFO:2026-04-22 15:52:43,832:jax._src.distributed:157: Connecting to JAX distributed service on mt-02-synthetic-c7568-slice-job-0-0.mt-02-synthetic-c7568:8482 I0422 15:52:43.832853 140014694840128 distributed.py:157] Connecting to JAX distributed service on mt-02-synthetic-c7568-slice-job-0-0.mt-02-synthetic-c7568:8482 I0422 15:52:45.273488 140014694840128 max_utils.py:284] Jax distributed system initialized! I0422 15:52:52.611402 140014694840128 max_utils.py:800] System Information: Jax Version: 0.8.1 I0422 15:52:52.611509 140014694840128 max_utils.py:801] System Information: Jaxlib Version: 0.8.1 I0422 15:52:52.611550 140014694840128 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 I0422 15:52:52.611583 140014694840128 train_utils.py:377] WARNING: Sequence packing is essentially ignored for synthetic data. Please use a real dataset to use sequence packing. I0422 15:52:53.234894 140014694840128 maxtext_utils.py:1631] Num_devices: 32, shape (1, 1, 1, 32, 1, 1, 1, 1, 1, 1, 1, 1, 1) I0422 15:52:53.235494 140014694840128 maxtext_utils.py:1631] Num_devices: 32, shape (1, 1, 1, 32, 1, 1, 1, 1, 1, 1, 1, 1, 1) I0422 15:52:53.235681 140014694840128 checkpointing.py:688] Setting up checkpoint logger... I0422 15:52:53.235732 140014694840128 checkpointing.py:234] Creating checkpoint manager with ocdbt=True and zarr3=True I0422 15:52:53.235777 140014694840128 pytree_checkpoint_handler.py:589] save_device_host_concurrent_bytes=None I0422 15:52:53.236201 140014694840128 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 0x7f573e5790a0>, enable_pinned_host_transfer=False, save_concurrent_bytes: 96000000000 (89.4 GiB), restore_concurrent_bytes: 96000000000 (89.4 GiB) I0422 15:52:56.346147 140014694840128 checkpointing.py:266] Enabling policy for fixed interval checkpointing. I0422 15:52:56.346391 140014694840128 checkpoint_manager.py:709] [process=6][thread=MainThread] CheckpointManager init: checkpointers=None, item_names=('items',), item_handlers={'items': <orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeCheckpointHandler object at 0x7f42a83b1a00>}, handler_registry=None I0422 15:52:56.346633 140014694840128 composite_checkpoint_handler.py:237] Deferred registration for item: "items". Adding handler `<orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeCheckpointHandler object at 0x7f42a83b1a00>` 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`. I0422 15:52:56.346683 140014694840128 composite_checkpoint_handler.py:237] Deferred registration for item: "metrics". Adding handler `<orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonCheckpointHandler object at 0x7f42a83b61e0>` 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`. I0422 15:52:56.346719 140014694840128 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 0x7f42a83b1a00>, ('items', <class 'orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeRestoreArgs'>): <orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeCheckpointHandler object at 0x7f42a83b1a00>, ('metrics', <class 'orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonSaveArgs'>): <orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonCheckpointHandler object at 0x7f42a83b61e0>, ('metrics', <class 'orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonRestoreArgs'>): <orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonCheckpointHandler object at 0x7f42a83b61e0>}). I0422 15:52:56.347239 140014694840128 abstract_checkpointer.py:35] orbax-checkpoint version: 0.11.33 I0422 15:52:56.347317 140014694840128 async_checkpointer.py:177] [process=6][thread=MainThread] Using barrier_sync_fn: <function get_barrier_sync_fn.<locals>._fn at 0x7f42a81a8180> timeout: 600 secs and primary_host=0 for async checkpoint writes I0422 15:52:57.428112 140014694840128 checkpoint_manager.py:1818] Found 0 checkpoint steps in gs://lance-maxtext/linen_ckpt_xpk_feat_nnx_set_defaults_true_20260422_154647/linen_xpk_feat_nnx_set_defaults_true_20260422_154647_02_synthetic/checkpoints I0422 15:52:57.992918 140014694840128 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=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_20260422_154647/linen_xpk_feat_nnx_set_defaults_true_20260422_154647_02_synthetic/checkpoints: <orbax.checkpoint.checkpoint_manager.CheckpointManager object at 0x7f42a831f4a0> I0422 15:52:57.993118 140014694840128 checkpointing.py:302] Checkpoint manager created! I0422 15:52:58.894496 140014694840128 nnx_wrappers.py:455] Unknown Logical: bfloat16[32,2048,2048]...................................... ('activation_batch', 'activation_norm_length', 'activation_embed'). I0422 15:52:58.894605 140014694840128 nnx_wrappers.py:455] Unknown Physical: bfloat16[32,2048,2048]...................................... ('fsdp', None, None). I0422 15:52:59.263952 140014694840128 attentions.py:1084] attentions/inputs_q Logical: bfloat16[32,2048,2048]...................................... ('activation_batch', 'activation_attn_length', 'activation_attn_embed'). I0422 15:52:59.264047 140014694840128 attentions.py:1084] attentions/inputs_q Physical: bfloat16[32,2048,2048]...................................... ('fsdp', None, None). I0422 15:52:59.279889 140014694840128 attentions.py:1085] attentions/inputs_kv Logical: bfloat16[32,2048,2048]...................................... ('activation_batch', 'activation_attn_length', 'activation_attn_embed'). I0422 15:52:59.279950 140014694840128 attentions.py:1085] attentions/inputs_kv Physical: bfloat16[32,2048,2048]...................................... ('fsdp', None, None). I0422 15:52:59.302585 140014694840128 attentions.py:1150] attentions/query Logical: bfloat16[32,2048,16,128].................................... ('activation_kv_batch', 'activation_attn_length', 'activation_kv_heads', 'activation_kv_head_dim'). I0422 15:52:59.302651 140014694840128 attentions.py:1150] attentions/query Physical: bfloat16[32,2048,16,128].................................... ('fsdp', None, None, None). I0422 15:52:59.318537 140014694840128 attentions.py:1151] attentions/key Logical: bfloat16[32,2048,16,128].................................... ('activation_kv_batch', 'activation_attn_length', 'activation_kv_heads', 'activation_kv_head_dim'). I0422 15:52:59.318598 140014694840128 attentions.py:1151] attentions/key Physical: bfloat16[32,2048,16,128].................................... ('fsdp', None, None, None). I0422 15:52:59.334475 140014694840128 attentions.py:1152] attentions/value Logical: bfloat16[32,2048,16,128].................................... ('activation_kv_batch', 'activation_attn_length', 'activation_kv_heads', 'activation_kv_head_dim'). I0422 15:52:59.334542 140014694840128 attentions.py:1152] attentions/value Physical: bfloat16[32,2048,16,128].................................... ('fsdp', None, None, None). I0422 15:52:59.359564 140014694840128 attentions.py:1193] attentions/out Logical: bfloat16[32,2048,16,128].................................... ('activation_batch', 'activation_attn_length', 'activation_heads', 'activation_kv'). I0422 15:52:59.359634 140014694840128 attentions.py:1193] attentions/out Physical: bfloat16[32,2048,16,128].................................... ('fsdp', None, None, None). I0422 15:52:59.379170 140014694840128 linears.py:541] linears/x Logical: bfloat16[32,2048,7168]...................................... ('activation_batch', 'activation_length', 'activation_mlp'). I0422 15:52:59.379235 140014694840128 linears.py:541] linears/x Physical: bfloat16[32,2048,7168]...................................... ('fsdp', None, None). I0422 15:52:59.568077 140014694840128 checkpointing.py:578] checkpoint manager exists so trying to load this run's existing checkpoint I0422 15:52:59.568222 140014694840128 checkpointing.py:676] No existing checkpoints found, not restoring checkpoint. fsdp: 32 I0422 15:53:01.008252 140014694840128 maxtext_utils.py:1740] params/params/decoder/decoder_norm/scale Shape: float32[2048] Logical: PartitionSpec('norm',) Physical: (None,) I0422 15:53:01.008414 140014694840128 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) I0422 15:53:01.008496 140014694840128 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) I0422 15:53:01.008581 140014694840128 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') I0422 15:53:01.008660 140014694840128 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) I0422 15:53:01.008720 140014694840128 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) I0422 15:53:01.008795 140014694840128 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) I0422 15:53:01.008877 140014694840128 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') I0422 15:53:01.008935 140014694840128 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) I0422 15:53:01.009004 140014694840128 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) I0422 15:53:01.009079 140014694840128 maxtext_utils.py:1740] params/params/decoder/logits_dense/kernel Shape: float32[2048,32000] Logical: PartitionSpec('embed', 'vocab') Physical: ('fsdp', None) I0422 15:53:01.009193 140014694840128 maxtext_utils.py:1740] params/params/token_embedder/embedding Shape: float32[32000,2048] Logical: PartitionSpec('vocab', 'embed') Physical: (None, 'fsdp') I0422 15:53:01.507035 140014694840128 train.py:158] train/xent Logical: float32[32,2048]............................................ ('activation_embed_and_logits_batch', 'activation_length'). I0422 15:53:01.507167 140014694840128 train.py:158] train/xent Physical: float32[32,2048]............................................ ('fsdp', None). I0422 15:53:01.522040 140014694840128 train.py:165] train/z_loss Logical: float32[32,2048]............................................ ('activation_embed_and_logits_batch', 'activation_length'). I0422 15:53:01.522116 140014694840128 train.py:165] train/z_loss Physical: float32[32,2048]............................................ ('fsdp', None). I0422 15:53:12.248585 140014694840128 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. I0422 15:53:12.249542 140014694840128 metric_logger.py:289] number parameters: 1.104 billion I0422 15:53:23.305069 140014694840128 checkpointing.py:794] Waiting for step 0 to finish before checkpoint... I0422 15:53:23.458912 140014694840128 checkpointing.py:798] Waited 0.15382790565490723 seconds for step 0 to finish before starting checkpointing. I0422 15:53:23.461260 140014694840128 checkpoint_manager.py:2013] [process=6][thread=MainThread][wait_until_finished] No Save Finalize thread to wait for. Returning. I0422 15:53:23.462866 140014694840128 checkpoint_manager.py:1518] [process=6] Saving checkpoint at step 0 I0422 15:53:23.465631 140014694840128 async_checkpointer.py:452] [process=6] Started async saving checkpoint to gs://lance-maxtext/linen_ckpt_xpk_feat_nnx_set_defaults_true_20260422_154647/linen_xpk_feat_nnx_set_defaults_true_20260422_154647_02_synthetic/checkpoints/0. I0422 15:53:24.167091 140014694840128 signaling_client.py:364] Using JaxDistributedSignalingClient I0422 15:53:24.168293 140014694840128 jax_array_handlers.py:358] Scheduling D2H of 39 prioritized jax.Array. I0422 15:53:24.168351 140014694840128 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 I0422 15:53:24.417499 140014694840128 base_pytree_checkpoint_handler.py:153] [process=6][thread=MainThread] Initiated "orbax.checkpoint._src.serialization.jax_array_handlers.ArrayHandler".serialize. Time taken: 0.250484s I0422 15:53:24.417668 140014694840128 base_pytree_checkpoint_handler.py:129] [process=6] /jax/checkpoint/write/blocking_gbytes_per_sec: 2.802 GiB/s (total gbytes: 1.5 GiB) (time elapsed: 0.5504834651947021 s) (per-host) I0422 15:53:24.417720 140014694840128 base_pytree_checkpoint_handler.py:737] [process=6][thread=MainThread] Initiated Pytree async_save. Time taken: 0.550546s (batch_requests_ready=0.296417s, total_serialization_initiated=0.254058s, others=0.000071s) I0422 15:53:24.417806 140014694840128 composite_checkpoint_handler.py:715] [process=6][thread=MainThread] Initiated CompositeCheckpointHandler.async_save. Time taken: 0.554567s (all_items=0.000019s, per_item={'items': '0.00001931'}, temp_paths=0.554548) I0422 15:53:24.418732 139888041981696 async_checkpointer.py:79] [process=6][thread=async_save] Background save thread started. I0422 15:53:24.418858 140014694840128 async_checkpointer.py:561] Finished blocking save. Time taken: 0.955920s. Continuing background save to gs://lance-maxtext/linen_ckpt_xpk_feat_nnx_set_defaults_true_20260422_154647/linen_xpk_feat_nnx_set_defaults_true_20260422_154647_02_synthetic/checkpoints/0. I0422 15:53:24.431082 140014694840128 checkpoint_manager.py:1566] [process=6][thread=MainThread][step=0] Starting CheckpointManager Save Finalize thread=save_finalize I0422 15:53:24.431336 139888586368768 async_checkpointer.py:265] [process=6][thread=save_finalize] Waiting for background save thread=async_save. I0422 15:53:24.431463 140014694840128 standard_logger.py:34] {'step': 0, 'event_type': 'save', 'directory': 'gs://lance-maxtext/linen_ckpt_xpk_feat_nnx_set_defaults_true_20260422_154647/linen_xpk_feat_nnx_set_defaults_true_20260422_154647_02_synthetic/checkpoints', 'reached_preemption': False, 'preemption_received_at': None, 'synchronous': False, 'wait_for_prev_start_time': 1776873203.4612424, 'wait_for_prev_duration_secs': 6.461143493652344e-05, 'time_between_consecutive_saves_sec': None, 'checkpointer_blocking_start_time': 1776873203.4629068, 'checkpointer_blocking_duration_secs': 0.956141471862793, 'get_old_steps_start_time': 1776873204.419068, 'get_old_steps_duration_secs': 3.695487976074219e-05, 'checkpoint_manager_blocking_start_time': 1776873203.4595175, 'checkpoint_manager_blocking_duration_secs': 0.9719078540802002} I0422 15:53:24.431561 140014694840128 checkpointing.py:409] Started an asynchronous checkpoint save for step 0 I0422 15:53:24.431610 140014694840128 max_utils.py:750] Memstats: After params initialized: I0422 15:53:24.431656 140014694840128 max_utils.py:756] Using (GB) 0.43 / 31.25 (1.376000%) on TPU_24(process=6,(0,6,0,0)) I0422 15:53:24.431689 140014694840128 max_utils.py:756] Using (GB) 0.43 / 31.25 (1.376000%) on TPU_25(process=6,(1,6,0,0)) I0422 15:53:24.431715 140014694840128 max_utils.py:756] Using (GB) 0.43 / 31.25 (1.376000%) on TPU_28(process=6,(0,7,0,0)) I0422 15:53:24.431740 140014694840128 max_utils.py:756] Using (GB) 0.43 / 31.25 (1.376000%) on TPU_29(process=6,(1,7,0,0)) I0422 15:53:24.696937 140014694840128 metric_logger.py:185] completed step: 0, seconds: 11.055, TFLOP/s/device: 1.229, Tokens/s/device: 185.249, total_weights: 65536, loss: 10.874 I0422 15:53:24.879273 140014694840128 metric_logger.py:185] completed step: 1, seconds: 1.391, TFLOP/s/device: 9.771, Tokens/s/device: 1472.791, total_weights: 65536, loss: 10.874 I0422 15:53:25.301930 140014694840128 metric_logger.py:185] completed step: 2, seconds: 0.035, TFLOP/s/device: 387.296, Tokens/s/device: 58377.516, total_weights: 65536, loss: 10.263 I0422 15:53:25.449335 140014694840128 metric_logger.py:185] completed step: 3, seconds: 0.423, TFLOP/s/device: 32.113, Tokens/s/device: 4840.406, total_weights: 65536, loss: 9.734 I0422 15:53:25.744381 140014694840128 metric_logger.py:185] completed step: 4, seconds: 0.152, TFLOP/s/device: 89.119, Tokens/s/device: 13433.032, total_weights: 65536, loss: 9.278 I0422 15:53:25.750179 140014694840128 metric_logger.py:185] completed step: 5, seconds: 0.147, TFLOP/s/device: 92.142, Tokens/s/device: 13888.606, total_weights: 65536, loss: 8.892 I0422 15:53:44.650484 140014694840128 metric_logger.py:185] completed step: 6, seconds: 0.296, TFLOP/s/device: 45.950, Tokens/s/device: 6926.079, total_weights: 65536, loss: 8.592 I0422 15:53:44.797973 140014694840128 metric_logger.py:185] completed step: 7, seconds: 18.754, TFLOP/s/device: 0.725, Tokens/s/device: 109.205, total_weights: 65536, loss: 8.384 I0422 15:53:44.945371 140014694840128 metric_logger.py:185] completed step: 8, seconds: 0.152, TFLOP/s/device: 89.380, Tokens/s/device: 13472.266, total_weights: 65536, loss: 8.255 I0422 15:53:45.092478 140014694840128 checkpointing.py:794] Waiting for step 9 to finish before checkpoint... I0422 15:53:45.093126 140014694840128 checkpointing.py:798] Waited 0.0006437301635742188 seconds for step 9 to finish before starting checkpointing. I0422 15:53:45.095106 140014694840128 checkpoint_manager.py:2024] [process=6][thread=MainThread][step=0][wait_until_finished] Waiting for Save Finalize thread (save_finalize) to complete. I0422 15:53:46.988430 2641 google_auth_provider.cc:181] Running on GCE, using service account 562977990677-compute@developer.gserviceaccount.com I0422 15:53:50.362630 139888050374400 array_metadata_store.py:203] [process=6][thread=array_type_handler] Wrote 39 array_metadata.ArrayMetadata to gs://lance-maxtext/linen_ckpt_xpk_feat_nnx_set_defaults_true_20260422_154647/linen_xpk_feat_nnx_set_defaults_true_20260422_154647_02_synthetic/checkpoints/0/items/array_metadatas/process_6 I0422 15:54:21.095857 139888041981696 base_pytree_checkpoint_handler.py:129] [process=6] /jax/checkpoint/write/gbytes_per_sec: 27.602 MiB/s (total gbytes: 1.5 GiB) (time elapsed: 57.228639364242554 s) (per-host) I0422 15:54:21.095959 139888041981696 async_checkpointer.py:90] [process=6][thread=async_save] 3 Handler Commit operations completed. Time taken: 56.677026s. I0422 15:54:29.003360 139888041981696 async_checkpointer.py:144] [process=6][thread=async_save] Background save thread done. Time taken: 64.584410s. I0422 15:54:29.003697 139888586368768 async_checkpointer.py:273] [process=6][thread=save_finalize] Done with waiting for background save thread=async_save. I0422 15:54:29.003828 139888586368768 async_checkpointer.py:283] [process=6][thread=save_finalize] No errors found in background save thread=async_save. I0422 15:54:29.003876 139888586368768 checkpoint_manager.py:2133] [process=6][thread=save_finalize][step=0] CheckpointManager Save Finalize is syncing with other hosts... I0422 15:54:29.005234 139888586368768 checkpoint_manager.py:2142] [process=6][thread=save_finalize][step=0] CheckpointManager Save Finalize is done on all hosts. I0422 15:54:29.005428 140014694840128 checkpoint_manager.py:2036] [process=6][thread=MainThread][step=0][wait_until_finished] Done waiting for Save Finalize thread (save_finalize) running at step=0. W0422 15:54:29.005602 140014694840128 checkpoint_manager.py:1458] Waiting for previous save to complete took 43.910514 seconds. If this number is high, consider checkpointing less frequently. I0422 15:54:29.007377 140014694840128 checkpoint_manager.py:1518] [process=6] Saving checkpoint at step 9 I0422 15:54:29.010918 140014694840128 async_checkpointer.py:452] [process=6] Started async saving checkpoint to gs://lance-maxtext/linen_ckpt_xpk_feat_nnx_set_defaults_true_20260422_154647/linen_xpk_feat_nnx_set_defaults_true_20260422_154647_02_synthetic/checkpoints/9. I0422 15:54:29.612042 140014694840128 jax_array_handlers.py:358] Scheduling D2H of 39 prioritized jax.Array. I0422 15:54:29.612155 140014694840128 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 I0422 15:54:29.657617 140014694840128 base_pytree_checkpoint_handler.py:153] [process=6][thread=MainThread] Initiated "orbax.checkpoint._src.serialization.jax_array_handlers.ArrayHandler".serialize. Time taken: 0.046772s I0422 15:54:29.657788 140014694840128 base_pytree_checkpoint_handler.py:129] [process=6] /jax/checkpoint/write/blocking_gbytes_per_sec: 4.805 GiB/s (total gbytes: 1.5 GiB) (time elapsed: 0.321012020111084 s) (per-host) I0422 15:54:29.657844 140014694840128 base_pytree_checkpoint_handler.py:737] [process=6][thread=MainThread] Initiated Pytree async_save. Time taken: 0.321077s (batch_requests_ready=0.272415s, total_serialization_initiated=0.048588s, others=0.000074s) I0422 15:54:29.657932 140014694840128 composite_checkpoint_handler.py:715] [process=6][thread=MainThread] Initiated CompositeCheckpointHandler.async_save. Time taken: 0.325118s (all_items=0.000016s, per_item={'items': '0.00001574'}, temp_paths=0.325102) I0422 15:54:29.658814 139887529412352 async_checkpointer.py:79] [process=6][thread=async_save] Background save thread started. I0422 15:54:29.658971 140014694840128 async_checkpointer.py:561] Finished blocking save. Time taken: 0.651522s. Continuing background save to gs://lance-maxtext/linen_ckpt_xpk_feat_nnx_set_defaults_true_20260422_154647/linen_xpk_feat_nnx_set_defaults_true_20260422_154647_02_synthetic/checkpoints/9. I0422 15:54:30.015003 140014694840128 checkpoint_manager.py:1566] [process=6][thread=MainThread][step=9] Starting CheckpointManager Save Finalize thread=save_finalize I0422 15:54:30.015385 139888586368768 async_checkpointer.py:265] [process=6][thread=save_finalize] Waiting for background save thread=async_save. I0422 15:54:30.015564 140014694840128 standard_logger.py:34] {'step': 9, 'event_type': 'save', 'directory': 'gs://lance-maxtext/linen_ckpt_xpk_feat_nnx_set_defaults_true_20260422_154647/linen_xpk_feat_nnx_set_defaults_true_20260422_154647_02_synthetic/checkpoints', 'reached_preemption': False, 'preemption_received_at': None, 'synchronous': False, 'wait_for_prev_start_time': 1776873225.0950649, 'wait_for_prev_duration_secs': 43.91051363945007, 'time_between_consecutive_saves_sec': None, 'checkpointer_blocking_start_time': 1776873269.0074155, 'checkpointer_blocking_duration_secs': 0.6516621112823486, 'get_old_steps_start_time': 1776873269.6591065, 'get_old_steps_duration_secs': 2.5272369384765625e-05, 'checkpoint_manager_blocking_start_time': 1776873225.0933576, 'checkpoint_manager_blocking_duration_secs': 44.922168254852295} I0422 15:54:30.015685 140014694840128 checkpointing.py:409] Started an asynchronous checkpoint save for step 9 I0422 15:54:30.015732 140014694840128 checkpoint_manager.py:2024] [process=6][thread=MainThread][step=9][wait_until_finished] Waiting for Save Finalize thread (save_finalize) to complete. I0422 15:54:35.551032 139888050374400 array_metadata_store.py:203] [process=6][thread=array_type_handler] Wrote 39 array_metadata.ArrayMetadata to gs://lance-maxtext/linen_ckpt_xpk_feat_nnx_set_defaults_true_20260422_154647/linen_xpk_feat_nnx_set_defaults_true_20260422_154647_02_synthetic/checkpoints/9/items/array_metadatas/process_6 I0422 15:55:11.799145 139887529412352 base_pytree_checkpoint_handler.py:129] [process=6] /jax/checkpoint/write/gbytes_per_sec: 37.200 MiB/s (total gbytes: 1.5 GiB) (time elapsed: 42.46230888366699 s) (per-host) I0422 15:55:11.799273 139887529412352 async_checkpointer.py:90] [process=6][thread=async_save] 3 Handler Commit operations completed. Time taken: 42.140350s. I0422 15:55:20.405002 139887529412352 async_checkpointer.py:144] [process=6][thread=async_save] Background save thread done. Time taken: 50.746065s. I0422 15:55:20.405262 139888586368768 async_checkpointer.py:273] [process=6][thread=save_finalize] Done with waiting for background save thread=async_save. I0422 15:55:20.405318 139888586368768 async_checkpointer.py:283] [process=6][thread=save_finalize] No errors found in background save thread=async_save. I0422 15:55:20.405363 139888586368768 checkpoint_manager.py:2133] [process=6][thread=save_finalize][step=9] CheckpointManager Save Finalize is syncing with other hosts... I0422 15:55:20.406901 139888586368768 checkpoint_manager.py:2142] [process=6][thread=save_finalize][step=9] CheckpointManager Save Finalize is done on all hosts. I0422 15:55:20.407068 140014694840128 checkpoint_manager.py:2036] [process=6][thread=MainThread][step=9][wait_until_finished] Done waiting for Save Finalize thread (save_finalize) running at step=9. I0422 15:55:20.407225 140014694840128 checkpoint_manager.py:2013] [process=6][thread=MainThread][wait_until_finished] No Save Finalize thread to wait for. Returning. I0422 15:55:20.407847 140014694840128 metric_logger.py:185] completed step: 9, seconds: 0.148, TFLOP/s/device: 92.009, Tokens/s/device: 13868.574, total_weights: 65536, loss: 8.179 Per train step: Total TFLOPs: 13.59 split as 93.93% learnable weight flops and 6.07% attention flops XPK End: Wed Apr 22 15:55:31 UTC 2026 EXIT_CODE=0