main (59e0f1759) vs feat/nnx-set-defaults-true (73213e044)| Metric | main 59e0f1759 | 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 | 92.178 | 92.024 | -0.154 |
| Tok/s | 13894.1 | 13870.8 | -23.243 |
| Avg s/step | 2.719 | 2.357 | -0.362 |
| 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: Sat Apr 25 07:29:57 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. 2026-04-25 07:30:21.840068: 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) I0425 07:30:22.043651 140086283003712 max_utils.py:273] Attempting to initialize the jax distributed system... I0425 07:30:31.085296 140086283003712 distributed.py:149] Starting JAX distributed service on [::]:8482 I0425 07:30:31.087654 140086283003712 distributed.py:172] Connecting to JAX distributed service on mt-07-eval-6hjmi-slice-job-0-0.mt-07-eval-6hjmi:8482 I0425 07:30:32.479485 140086283003712 max_utils.py:284] Jax distributed system initialized! I0425 07:30:38.608567 140086283003712 max_utils.py:800] System Information: Jax Version: 0.9.2 I0425 07:30:38.608674 140086283003712 max_utils.py:801] System Information: Jaxlib Version: 0.9.2 I0425 07:30:38.608713 140086283003712 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 I0425 07:30:38.608748 140086283003712 train_utils.py:361] WARNING: Sequence packing is essentially ignored for synthetic data. Please use a real dataset to use sequence packing. I0425 07:30:39.313864 140086283003712 maxtext_utils.py:1604] Num_devices: 32, shape (1, 1, 1, 32, 1, 1, 1, 1, 1, 1, 1, 1) I0425 07:30:39.314153 140086283003712 checkpointing.py:677] Setting up checkpoint logger... I0425 07:30:39.314209 140086283003712 checkpointing.py:233] Creating checkpoint manager with ocdbt=True and zarr3=True I0425 07:30:39.314256 140086283003712 pytree_checkpoint_handler.py:592] save_device_host_concurrent_bytes=None I0425 07:30:39.314591 140086283003712 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 0x7f67c5b14dd0>, enable_pinned_host_transfer=False, save_concurrent_bytes: 96000000000 (89.4 GiB), restore_concurrent_bytes: 96000000000 (89.4 GiB) I0425 07:30:42.176279 140086283003712 checkpointing.py:265] Enabling policy for fixed interval checkpointing. I0425 07:30:42.176521 140086283003712 checkpoint_manager.py:708] [process=6][thread=MainThread] CheckpointManager init: checkpointers=None, item_names=('items',), item_handlers={'items': <orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeCheckpointHandler object at 0x7f669ccbbe30>}, handler_registry=None I0425 07:30:42.176757 140086283003712 composite_checkpoint_handler.py:237] Deferred registration for item: "items". Adding handler `<orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeCheckpointHandler object at 0x7f669ccbbe30>` 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`. I0425 07:30:42.176805 140086283003712 composite_checkpoint_handler.py:237] Deferred registration for item: "metrics". Adding handler `<orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonCheckpointHandler object at 0x7f53803290a0>` 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`. I0425 07:30:42.176841 140086283003712 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 0x7f669ccbbe30>, ('items', <class 'orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeRestoreArgs'>): <orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeCheckpointHandler object at 0x7f669ccbbe30>, ('metrics', <class 'orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonSaveArgs'>): <orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonCheckpointHandler object at 0x7f53803290a0>, ('metrics', <class 'orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonRestoreArgs'>): <orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonCheckpointHandler object at 0x7f53803290a0>}). I0425 07:30:42.177196 140086283003712 abstract_checkpointer.py:35] orbax-checkpoint version: 0.11.34 I0425 07:30:42.177281 140086283003712 async_checkpointer.py:192] [process=6][thread=MainThread] Using barrier_sync_fn: <function get_barrier_sync_fn.<locals>._fn at 0x7f52a06e5760> timeout: 1200 secs and primary_host=0 for async checkpoint writes I0425 07:30:43.020732 140086283003712 checkpoint_manager.py:1812] Found 0 checkpoint steps in gs://lance-maxtext/linen_ckpt_xpk_main_20260425_071456/linen_xpk_main_20260425_071456_07_eval/checkpoints I0425 07:30:43.052524 140086283003712 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_main_20260425_071456/linen_xpk_main_20260425_071456_07_eval/checkpoints: <orbax.checkpoint.checkpoint_manager.CheckpointManager object at 0x7f538032a300> I0425 07:30:43.052660 140086283003712 checkpointing.py:301] Checkpoint manager created! I0425 07:30:43.982335 140086283003712 nnx_wrappers.py:437] Unknown Logical: bfloat16[32,2048,2048]...................................... ('activation_batch', 'activation_norm_length', 'activation_embed'). I0425 07:30:43.982451 140086283003712 nnx_wrappers.py:437] Unknown Physical: bfloat16[32,2048,2048]...................................... ('fsdp', None, None). I0425 07:30:44.365956 140086283003712 attentions.py:1088] attentions/inputs_q Logical: bfloat16[32,2048,2048]...................................... ('activation_batch_attn', 'activation_length_attn', 'activation_embed_attn'). I0425 07:30:44.366050 140086283003712 attentions.py:1088] attentions/inputs_q Physical: bfloat16[32,2048,2048]...................................... ('fsdp', None, None). I0425 07:30:44.382715 140086283003712 attentions.py:1089] attentions/inputs_kv Logical: bfloat16[32,2048,2048]...................................... ('activation_batch_attn', 'activation_length_attn', 'activation_embed_attn'). I0425 07:30:44.382778 140086283003712 attentions.py:1089] attentions/inputs_kv Physical: bfloat16[32,2048,2048]...................................... ('fsdp', None, None). I0425 07:30:44.406744 140086283003712 attentions.py:1154] attentions/query Logical: bfloat16[32,2048,16,128].................................... ('activation_kv_batch', 'activation_length_attn', 'activation_kv_heads', 'activation_kv_head_dim'). I0425 07:30:44.406813 140086283003712 attentions.py:1154] attentions/query Physical: bfloat16[32,2048,16,128].................................... ('fsdp', None, None, None). I0425 07:30:44.423426 140086283003712 attentions.py:1155] attentions/key Logical: bfloat16[32,2048,16,128].................................... ('activation_kv_batch', 'activation_length_attn', 'activation_kv_heads', 'activation_kv_head_dim'). I0425 07:30:44.423493 140086283003712 attentions.py:1155] attentions/key Physical: bfloat16[32,2048,16,128].................................... ('fsdp', None, None, None). I0425 07:30:44.440217 140086283003712 attentions.py:1156] attentions/value Logical: bfloat16[32,2048,16,128].................................... ('activation_kv_batch', 'activation_length_attn', 'activation_kv_heads', 'activation_kv_head_dim'). I0425 07:30:44.440278 140086283003712 attentions.py:1156] attentions/value Physical: bfloat16[32,2048,16,128].................................... ('fsdp', None, None, None). I0425 07:30:44.465201 140086283003712 attentions.py:1198] attentions/out Logical: bfloat16[32,2048,16,128].................................... ('activation_batch_attn', 'activation_length_attn', 'activation_heads', 'activation_kv'). I0425 07:30:44.465271 140086283003712 attentions.py:1198] attentions/out Physical: bfloat16[32,2048,16,128].................................... ('fsdp', None, None, None). I0425 07:30:44.486516 140086283003712 linears.py:525] linears/x Logical: bfloat16[32,2048,7168]...................................... ('activation_batch', 'activation_length', 'activation_mlp'). I0425 07:30:44.486581 140086283003712 linears.py:525] linears/x Physical: bfloat16[32,2048,7168]...................................... ('fsdp', None, None). I0425 07:30:44.696981 140086283003712 checkpointing.py:577] checkpoint manager exists so trying to load this run's existing checkpoint I0425 07:30:44.697087 140086283003712 checkpointing.py:665] No existing checkpoints found, not restoring checkpoint. fsdp: 32 I0425 07:30:46.132336 140086283003712 maxtext_utils.py:1707] params/params/decoder/decoder_norm/scale Shape: float32[2048] Logical: P('norm',) Physical: (None,) I0425 07:30:46.132464 140086283003712 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) I0425 07:30:46.132516 140086283003712 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) I0425 07:30:46.132574 140086283003712 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') I0425 07:30:46.132627 140086283003712 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) I0425 07:30:46.132666 140086283003712 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) I0425 07:30:46.132723 140086283003712 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) I0425 07:30:46.132776 140086283003712 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') I0425 07:30:46.132816 140086283003712 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) I0425 07:30:46.132851 140086283003712 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) I0425 07:30:46.132899 140086283003712 maxtext_utils.py:1707] params/params/decoder/logits_dense/kernel Shape: float32[2048,32000] Logical: P('embed_vocab', 'vocab') Physical: ('fsdp', None) I0425 07:30:46.132946 140086283003712 maxtext_utils.py:1707] params/params/token_embedder/embedding Shape: float32[32000,2048] Logical: P('vocab', 'embed_vocab') Physical: (None, 'fsdp') I0425 07:30:46.627427 140086283003712 train.py:155] train/xent Logical: float32[32,2048]............................................ ('activation_embed_and_logits_batch', 'activation_length'). I0425 07:30:46.627525 140086283003712 train.py:155] train/xent Physical: float32[32,2048]............................................ ('fsdp', None). I0425 07:30:46.643202 140086283003712 train.py:162] train/z_loss Logical: float32[32,2048]............................................ ('activation_embed_and_logits_batch', 'activation_length'). I0425 07:30:46.643266 140086283003712 train.py:162] train/z_loss Physical: float32[32,2048]............................................ ('fsdp', None). I0425 07:30:57.703543 140086283003712 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. I0425 07:30:57.704318 140086283003712 metric_logger.py:301] number parameters: 1.104 billion I0425 07:31:09.397899 140086283003712 checkpointing.py:772] Waiting for step 0 to finish before checkpoint... I0425 07:31:09.551964 140086283003712 checkpointing.py:776] Waited 0.15404701232910156 seconds for step 0 to finish before starting checkpointing. I0425 07:31:09.554126 140086283003712 checkpoint_manager.py:2009] [process=6][thread=MainThread][wait_until_finished] No Save Finalize thread to wait for. Returning. I0425 07:31:09.555714 140086283003712 checkpoint_manager.py:1512] [process=6] Saving checkpoint at step 0 I0425 07:31:09.557067 140086283003712 event_tracking.py:70] [process=6] [async] Started save checkpoint @ gs://lance-maxtext/linen_ckpt_xpk_main_20260425_071456/linen_xpk_main_20260425_071456_07_eval/checkpoints/0. I0425 07:31:10.319557 140086283003712 signaling_client.py:364] Using JaxDistributedSignalingClient I0425 07:31:10.320452 140086283003712 jax_array_handlers.py:360] Scheduling D2H of 39 prioritized jax.Array. I0425 07:31:10.320521 140086283003712 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 I0425 07:31:10.596589 140086283003712 base_pytree_checkpoint_handler.py:154] [process=6][thread=MainThread] Initiated "orbax.checkpoint._src.serialization.jax_array_handlers.ArrayHandler".serialize. Time taken: 0.277117s I0425 07:31:10.596763 140086283003712 base_pytree_checkpoint_handler.py:130] [process=6] /jax/orbax/write/blocking_gbytes_per_sec: 5.454 GiB/s (total gbytes: 1.5 GiB) (time elapsed: 0.2828233242034912 s) (per-host) I0425 07:31:10.596818 140086283003712 base_pytree_checkpoint_handler.py:768] [process=6][thread=MainThread] Initiated Pytree async_save. Time taken: 0.282888s (batch_requests_ready=0.002419s, total_serialization_initiated=0.280397s, others=0.000072s) I0425 07:31:10.596910 140086283003712 composite_checkpoint_handler.py:715] [process=6][thread=MainThread] Initiated CompositeCheckpointHandler.async_save. Time taken: 0.287235s (all_items=0.000030s, per_item={'items': '0.00003004'}, temp_paths=0.287205) I0425 07:31:10.597758 140086283003712 event_tracking.py:125] [process=6] [async] Finished blocking save in 1.04 seconds. Continuing save @ gs://lance-maxtext/linen_ckpt_xpk_main_20260425_071456/linen_xpk_main_20260425_071456_07_eval/checkpoints/0. I0425 07:31:10.598088 139958035650304 async_checkpointer.py:76] [process=6][thread=async_save] Background save thread started. Deadline for this save operation is 2026-04-25 07:51:10.598049 I0425 07:31:10.603577 140086283003712 checkpoint_manager.py:1560] [process=6][thread=MainThread][step=0] Starting CheckpointManager Save Finalize thread=save_finalize I0425 07:31:10.603824 139955892373248 async_checkpointer.py:280] [process=6][thread=save_finalize] Waiting for background save thread=async_save. I0425 07:31:10.603972 140086283003712 standard_logger.py:34] {'step': 0, 'event_type': 'save', 'directory': 'gs://lance-maxtext/linen_ckpt_xpk_main_20260425_071456/linen_xpk_main_20260425_071456_07_eval/checkpoints', 'reached_preemption': False, 'preemption_received_at': None, 'synchronous': False, 'wait_for_prev_start_time': 1777102269.5541072, 'wait_for_prev_duration_secs': 6.365776062011719e-05, 'time_between_consecutive_saves_sec': None, 'checkpointer_blocking_start_time': 1777102269.555753, 'checkpointer_blocking_duration_secs': 1.0424506664276123, 'get_old_steps_start_time': 1777102270.5982287, 'get_old_steps_duration_secs': 3.1948089599609375e-05, 'checkpoint_manager_blocking_start_time': 1777102269.5524402, 'checkpoint_manager_blocking_duration_secs': 1.0514945983886719} I0425 07:31:10.604077 140086283003712 checkpointing.py:408] Started an asynchronous checkpoint save for step 0 I0425 07:31:10.604134 140086283003712 max_utils.py:750] Memstats: After params initialized: I0425 07:31:10.604180 140086283003712 max_utils.py:756] Using (GB) 0.43 / 31.25 (1.376000%) on TPU_24(process=6,(0,6,0,0)) I0425 07:31:10.604209 140086283003712 max_utils.py:756] Using (GB) 0.43 / 31.25 (1.376000%) on TPU_25(process=6,(1,6,0,0)) I0425 07:31:10.604236 140086283003712 max_utils.py:756] Using (GB) 0.43 / 31.25 (1.376000%) on TPU_28(process=6,(0,7,0,0)) I0425 07:31:10.604261 140086283003712 max_utils.py:756] Using (GB) 0.43 / 31.25 (1.376000%) on TPU_29(process=6,(1,7,0,0)) I0425 07:31:10.913857 140086283003712 metric_logger.py:196] completed step: 0, seconds: 11.693, TFLOP/s/device: 1.162, Tokens/s/device: 175.140, total_weights: 65536, loss: 10.874, lm_loss: 10.874, perplexity: 52779.875 I0425 07:31:11.098443 140086283003712 metric_logger.py:196] completed step: 1, seconds: 1.514, TFLOP/s/device: 8.972, Tokens/s/device: 1352.333, total_weights: 65536, loss: 10.874, lm_loss: 10.874, perplexity: 52779.875 I0425 07:31:11.526601 140086283003712 metric_logger.py:196] completed step: 2, seconds: 0.038, TFLOP/s/device: 362.101, Tokens/s/device: 54579.858, total_weights: 65536, loss: 10.262, lm_loss: 10.262, perplexity: 28637.129 I0425 07:31:11.673859 140086283003712 metric_logger.py:196] completed step: 3, seconds: 0.428, TFLOP/s/device: 31.722, Tokens/s/device: 4781.472, total_weights: 65536, loss: 9.734, lm_loss: 9.734, perplexity: 16889.203 I0425 07:31:11.969033 140086283003712 metric_logger.py:196] completed step: 4, seconds: 0.153, TFLOP/s/device: 88.843, Tokens/s/device: 13391.310, total_weights: 65536, loss: 9.277, lm_loss: 9.277, perplexity: 10694.614 I0425 07:31:11.974947 140086283003712 metric_logger.py:196] completed step: 5, seconds: 0.147, TFLOP/s/device: 92.298, Tokens/s/device: 13912.193, total_weights: 65536, loss: 8.892, lm_loss: 8.892, perplexity: 7270.717 I0425 07:31:14.071102 2797 google_auth_provider.cc:181] Running on GCE, using service account 562977990677-compute@developer.gserviceaccount.com I0425 07:31:16.094603 139955900765952 array_metadata_store.py:203] [process=6][thread=array_type_handler] Wrote 39 array_metadata.ArrayMetadata to gs://lance-maxtext/linen_ckpt_xpk_main_20260425_071456/linen_xpk_main_20260425_071456_07_eval/checkpoints/0/items/array_metadatas/process_6 I0425 07:31:33.718538 140086283003712 metric_logger.py:196] completed step: 6, seconds: 0.296, TFLOP/s/device: 45.964, Tokens/s/device: 6928.141, total_weights: 65536, loss: 8.592, lm_loss: 8.592, perplexity: 5390.759 I0425 07:31:33.865911 140086283003712 metric_logger.py:196] completed step: 7, seconds: 21.597, TFLOP/s/device: 0.629, Tokens/s/device: 94.828, total_weights: 65536, loss: 8.384, lm_loss: 8.384, perplexity: 4376.905 I0425 07:31:34.012928 140086283003712 metric_logger.py:196] completed step: 8, seconds: 0.152, TFLOP/s/device: 89.351, Tokens/s/device: 13467.925, total_weights: 65536, loss: 8.255, lm_loss: 8.255, perplexity: 3846.406 I0425 07:31:34.160906 140086283003712 checkpointing.py:772] Waiting for step 9 to finish before checkpoint... I0425 07:31:34.161579 140086283003712 checkpointing.py:776] Waited 0.0006940364837646484 seconds for step 9 to finish before starting checkpointing. I0425 07:31:34.163712 140086283003712 checkpoint_manager.py:2020] [process=6][thread=MainThread][step=0][wait_until_finished] Waiting for Save Finalize thread (save_finalize) to complete. I0425 07:31:48.592410 139958035650304 base_pytree_checkpoint_handler.py:130] [process=6] /jax/orbax/write/gbytes_per_sec: 41.266 MiB/s (total gbytes: 1.5 GiB) (time elapsed: 38.27843165397644 s) (per-host) I0425 07:31:48.592535 139958035650304 async_checkpointer.py:90] [process=6][thread=async_save] 3 Handler Commit operations completed. Time taken: 37.993795s. I0425 07:31:56.068517 139958035650304 async_checkpointer.py:160] [process=6][thread=async_save] Background save thread done. Time taken: 45.469760s. I0425 07:31:56.068814 139955892373248 async_checkpointer.py:288] [process=6][thread=save_finalize] Done with waiting for background save thread=async_save. I0425 07:31:56.068938 139955892373248 async_checkpointer.py:298] [process=6][thread=save_finalize] No errors found in background save thread=async_save. I0425 07:31:56.068987 139955892373248 checkpoint_manager.py:2137] [process=6][thread=save_finalize][step=0] CheckpointManager Save Finalize is syncing with other hosts... I0425 07:31:56.071333 139955892373248 checkpoint_manager.py:2146] [process=6][thread=save_finalize][step=0] CheckpointManager Save Finalize is done on all hosts. I0425 07:31:56.071525 140086283003712 checkpoint_manager.py:2032] [process=6][thread=MainThread][step=0][wait_until_finished] Done waiting for Save Finalize thread (save_finalize) running at step=0. W0425 07:31:56.071654 140086283003712 checkpoint_manager.py:1452] Waiting for previous save to complete took 21.907941 seconds. If this number is high, consider checkpointing less frequently. I0425 07:31:56.073552 140086283003712 checkpoint_manager.py:1512] [process=6] Saving checkpoint at step 9 I0425 07:31:56.075533 140086283003712 event_tracking.py:70] [process=6] [async] Started save checkpoint @ gs://lance-maxtext/linen_ckpt_xpk_main_20260425_071456/linen_xpk_main_20260425_071456_07_eval/checkpoints/9. I0425 07:31:56.348145 140086283003712 jax_array_handlers.py:360] Scheduling D2H of 39 prioritized jax.Array. I0425 07:31:56.348239 140086283003712 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 I0425 07:31:56.383889 140086283003712 base_pytree_checkpoint_handler.py:154] [process=6][thread=MainThread] Initiated "orbax.checkpoint._src.serialization.jax_array_handlers.ArrayHandler".serialize. Time taken: 0.036847s I0425 07:31:56.384050 140086283003712 base_pytree_checkpoint_handler.py:130] [process=6] /jax/orbax/write/blocking_gbytes_per_sec: 37.652 GiB/s (total gbytes: 1.5 GiB) (time elapsed: 0.0409693717956543 s) (per-host) I0425 07:31:56.384114 140086283003712 base_pytree_checkpoint_handler.py:768] [process=6][thread=MainThread] Initiated Pytree async_save. Time taken: 0.041030s (batch_requests_ready=0.002468s, total_serialization_initiated=0.038493s, others=0.000069s) I0425 07:31:56.384204 140086283003712 composite_checkpoint_handler.py:715] [process=6][thread=MainThread] Initiated CompositeCheckpointHandler.async_save. Time taken: 0.045083s (all_items=0.000016s, per_item={'items': '0.00001597'}, temp_paths=0.045067) I0425 07:31:56.384886 140086283003712 event_tracking.py:125] [process=6] [async] Finished blocking save in 0.31 seconds. Continuing save @ gs://lance-maxtext/linen_ckpt_xpk_main_20260425_071456/linen_xpk_main_20260425_071456_07_eval/checkpoints/9. I0425 07:31:56.385267 139955892373248 async_checkpointer.py:76] [process=6][thread=async_save] Background save thread started. Deadline for this save operation is 2026-04-25 07:51:56.385221 I0425 07:31:56.387278 140086283003712 checkpoint_manager.py:1560] [process=6][thread=MainThread][step=9] Starting CheckpointManager Save Finalize thread=save_finalize I0425 07:31:56.387579 139957531064064 async_checkpointer.py:280] [process=6][thread=save_finalize] Waiting for background save thread=async_save. I0425 07:31:56.387719 140086283003712 standard_logger.py:34] {'step': 9, 'event_type': 'save', 'directory': 'gs://lance-maxtext/linen_ckpt_xpk_main_20260425_071456/linen_xpk_main_20260425_071456_07_eval/checkpoints', 'reached_preemption': False, 'preemption_received_at': None, 'synchronous': False, 'wait_for_prev_start_time': 1777102294.163681, 'wait_for_prev_duration_secs': 21.907941341400146, 'time_between_consecutive_saves_sec': None, 'checkpointer_blocking_start_time': 1777102316.0735912, 'checkpointer_blocking_duration_secs': 0.31184887886047363, 'get_old_steps_start_time': 1777102316.3854637, 'get_old_steps_duration_secs': 3.719329833984375e-05, 'checkpoint_manager_blocking_start_time': 1777102294.1618028, 'checkpoint_manager_blocking_duration_secs': 22.22588062286377} I0425 07:31:56.387885 140086283003712 checkpointing.py:408] Started an asynchronous checkpoint save for step 9 I0425 07:31:56.387930 140086283003712 checkpoint_manager.py:2020] [process=6][thread=MainThread][step=9][wait_until_finished] Waiting for Save Finalize thread (save_finalize) to complete. I0425 07:32:01.765737 139955900765952 array_metadata_store.py:203] [process=6][thread=array_type_handler] Wrote 39 array_metadata.ArrayMetadata to gs://lance-maxtext/linen_ckpt_xpk_main_20260425_071456/linen_xpk_main_20260425_071456_07_eval/checkpoints/9/items/array_metadatas/process_6 I0425 07:32:37.767421 139955892373248 base_pytree_checkpoint_handler.py:130] [process=6] /jax/orbax/write/gbytes_per_sec: 38.132 MiB/s (total gbytes: 1.5 GiB) (time elapsed: 41.42431640625 s) (per-host) I0425 07:32:37.767536 139955892373248 async_checkpointer.py:90] [process=6][thread=async_save] 3 Handler Commit operations completed. Time taken: 41.382132s. I0425 07:32:45.928233 139955892373248 async_checkpointer.py:160] [process=6][thread=async_save] Background save thread done. Time taken: 49.542810s. I0425 07:32:45.928537 139957531064064 async_checkpointer.py:288] [process=6][thread=save_finalize] Done with waiting for background save thread=async_save. I0425 07:32:45.928684 139957531064064 async_checkpointer.py:298] [process=6][thread=save_finalize] No errors found in background save thread=async_save. I0425 07:32:45.928732 139957531064064 checkpoint_manager.py:2137] [process=6][thread=save_finalize][step=9] CheckpointManager Save Finalize is syncing with other hosts... I0425 07:32:45.930227 139957531064064 checkpoint_manager.py:2146] [process=6][thread=save_finalize][step=9] CheckpointManager Save Finalize is done on all hosts. I0425 07:32:45.930412 140086283003712 checkpoint_manager.py:2032] [process=6][thread=MainThread][step=9][wait_until_finished] Done waiting for Save Finalize thread (save_finalize) running at step=9. I0425 07:32:45.930558 140086283003712 checkpoint_manager.py:2009] [process=6][thread=MainThread][wait_until_finished] No Save Finalize thread to wait for. Returning. I0425 07:32:45.931623 140086283003712 metric_logger.py:196] completed step: 9, seconds: 0.147, TFLOP/s/device: 92.178, Tokens/s/device: 13894.071, 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: Sat Apr 25 07:32:57 UTC 2026 EXIT_CODE=0
XPK Start: Sat Apr 25 15:19:29 UTC 2026 2026-04-25 15:19:33.927019: 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 00:00:1777130373.939456 10 cuda_dnn.cc:8579] Unable to register cuDNN factory: Attempting to register factory for plugin cuDNN when one has already been registered E0000 00:00:1777130373.943267 10 cuda_blas.cc:1407] Unable to register cuBLAS factory: Attempting to register factory for plugin cuBLAS when one has already been registered W0000 00:00:1777130373.954453 10 computation_placer.cc:177] computation placer already registered. Please check linkage and avoid linking the same target more than once. W0000 00:00:1777130373.954471 10 computation_placer.cc:177] computation placer already registered. Please check linkage and avoid linking the same target more than once. W0000 00:00:1777130373.954473 10 computation_placer.cc:177] computation placer already registered. Please check linkage and avoid linking the same target more than once. W0000 00:00:1777130373.954475 10 computation_placer.cc:177] computation placer already registered. Please check linkage and avoid linking the same target more than once. 2026-04-25 15:19:53.158321: 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) I0425 15:19:53.681800 133019535910720 max_utils.py:273] Attempting to initialize the jax distributed system... INFO:2026-04-25 15:20:02,721:jax._src.distributed:140: Starting JAX distributed service on [::]:8482 I0425 15:20:02.721856 133019535910720 distributed.py:140] Starting JAX distributed service on [::]:8482 INFO:2026-04-25 15:20:02,724:jax._src.distributed:157: Connecting to JAX distributed service on mt-07-eval-c9gni-slice-job-0-0.mt-07-eval-c9gni:8482 I0425 15:20:02.724214 133019535910720 distributed.py:157] Connecting to JAX distributed service on mt-07-eval-c9gni-slice-job-0-0.mt-07-eval-c9gni:8482 I0425 15:20:04.323141 133019535910720 max_utils.py:284] Jax distributed system initialized! I0425 15:20:11.373933 133019535910720 max_utils.py:800] System Information: Jax Version: 0.8.1 I0425 15:20:11.374040 133019535910720 max_utils.py:801] System Information: Jaxlib Version: 0.8.1 I0425 15:20:11.374081 133019535910720 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 I0425 15:20:11.374116 133019535910720 train_utils.py:377] WARNING: Sequence packing is essentially ignored for synthetic data. Please use a real dataset to use sequence packing. I0425 15:20:11.997408 133019535910720 maxtext_utils.py:1631] Num_devices: 32, shape (1, 1, 1, 32, 1, 1, 1, 1, 1, 1, 1, 1, 1) I0425 15:20:11.998193 133019535910720 maxtext_utils.py:1631] Num_devices: 32, shape (1, 1, 1, 32, 1, 1, 1, 1, 1, 1, 1, 1, 1) I0425 15:20:11.998382 133019535910720 checkpointing.py:688] Setting up checkpoint logger... I0425 15:20:11.998435 133019535910720 checkpointing.py:234] Creating checkpoint manager with ocdbt=True and zarr3=True I0425 15:20:11.998479 133019535910720 pytree_checkpoint_handler.py:589] save_device_host_concurrent_bytes=None I0425 15:20:11.998860 133019535910720 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 0x78fa93d6d2b0>, enable_pinned_host_transfer=False, save_concurrent_bytes: 96000000000 (89.4 GiB), restore_concurrent_bytes: 96000000000 (89.4 GiB) I0425 15:20:14.905981 133019535910720 checkpointing.py:266] Enabling policy for fixed interval checkpointing. I0425 15:20:14.906331 133019535910720 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 0x78f93c562960>}, handler_registry=None I0425 15:20:14.906565 133019535910720 composite_checkpoint_handler.py:237] Deferred registration for item: "items". Adding handler `<orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeCheckpointHandler object at 0x78f93c562960>` 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`. I0425 15:20:14.906614 133019535910720 composite_checkpoint_handler.py:237] Deferred registration for item: "metrics". Adding handler `<orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonCheckpointHandler object at 0x78e6340ebda0>` 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`. I0425 15:20:14.906661 133019535910720 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 0x78f93c562960>, ('items', <class 'orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeRestoreArgs'>): <orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeCheckpointHandler object at 0x78f93c562960>, ('metrics', <class 'orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonSaveArgs'>): <orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonCheckpointHandler object at 0x78e6340ebda0>, ('metrics', <class 'orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonRestoreArgs'>): <orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonCheckpointHandler object at 0x78e6340ebda0>}). I0425 15:20:14.907260 133019535910720 abstract_checkpointer.py:35] orbax-checkpoint version: 0.11.33 I0425 15:20:14.907337 133019535910720 async_checkpointer.py:177] [process=5][thread=MainThread] Using barrier_sync_fn: <function get_barrier_sync_fn.<locals>._fn at 0x78e5dc684180> timeout: 600 secs and primary_host=0 for async checkpoint writes I0425 15:20:15.648892 133019535910720 checkpoint_manager.py:1818] Found 0 checkpoint steps in gs://lance-maxtext/linen_ckpt_xpk_feat_nnx_set_defaults_true_20260425_150412/linen_xpk_feat_nnx_set_defaults_true_20260425_150412_07_eval/checkpoints I0425 15:20:15.669369 133019535910720 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_20260425_150412/linen_xpk_feat_nnx_set_defaults_true_20260425_150412_07_eval/checkpoints: <orbax.checkpoint.checkpoint_manager.CheckpointManager object at 0x78e61c25fef0> I0425 15:20:15.669496 133019535910720 checkpointing.py:302] Checkpoint manager created! I0425 15:20:16.564646 133019535910720 nnx_wrappers.py:455] Unknown Logical: bfloat16[32,2048,2048]...................................... ('activation_batch', 'activation_norm_length', 'activation_embed'). I0425 15:20:16.564764 133019535910720 nnx_wrappers.py:455] Unknown Physical: bfloat16[32,2048,2048]...................................... ('fsdp', None, None). I0425 15:20:16.936518 133019535910720 attentions.py:1084] attentions/inputs_q Logical: bfloat16[32,2048,2048]...................................... ('activation_batch', 'activation_attn_length', 'activation_attn_embed'). I0425 15:20:16.936622 133019535910720 attentions.py:1084] attentions/inputs_q Physical: bfloat16[32,2048,2048]...................................... ('fsdp', None, None). I0425 15:20:16.953063 133019535910720 attentions.py:1085] attentions/inputs_kv Logical: bfloat16[32,2048,2048]...................................... ('activation_batch', 'activation_attn_length', 'activation_attn_embed'). I0425 15:20:16.953129 133019535910720 attentions.py:1085] attentions/inputs_kv Physical: bfloat16[32,2048,2048]...................................... ('fsdp', None, None). I0425 15:20:16.976362 133019535910720 attentions.py:1150] attentions/query Logical: bfloat16[32,2048,16,128].................................... ('activation_kv_batch', 'activation_attn_length', 'activation_kv_heads', 'activation_kv_head_dim'). I0425 15:20:16.976433 133019535910720 attentions.py:1150] attentions/query Physical: bfloat16[32,2048,16,128].................................... ('fsdp', None, None, None). I0425 15:20:16.992709 133019535910720 attentions.py:1151] attentions/key Logical: bfloat16[32,2048,16,128].................................... ('activation_kv_batch', 'activation_attn_length', 'activation_kv_heads', 'activation_kv_head_dim'). I0425 15:20:16.992770 133019535910720 attentions.py:1151] attentions/key Physical: bfloat16[32,2048,16,128].................................... ('fsdp', None, None, None). I0425 15:20:17.008884 133019535910720 attentions.py:1152] attentions/value Logical: bfloat16[32,2048,16,128].................................... ('activation_kv_batch', 'activation_attn_length', 'activation_kv_heads', 'activation_kv_head_dim'). I0425 15:20:17.008944 133019535910720 attentions.py:1152] attentions/value Physical: bfloat16[32,2048,16,128].................................... ('fsdp', None, None, None). I0425 15:20:17.034145 133019535910720 attentions.py:1193] attentions/out Logical: bfloat16[32,2048,16,128].................................... ('activation_batch', 'activation_attn_length', 'activation_heads', 'activation_kv'). I0425 15:20:17.034220 133019535910720 attentions.py:1193] attentions/out Physical: bfloat16[32,2048,16,128].................................... ('fsdp', None, None, None). I0425 15:20:17.053968 133019535910720 linears.py:541] linears/x Logical: bfloat16[32,2048,7168]...................................... ('activation_batch', 'activation_length', 'activation_mlp'). I0425 15:20:17.054039 133019535910720 linears.py:541] linears/x Physical: bfloat16[32,2048,7168]...................................... ('fsdp', None, None). I0425 15:20:17.244617 133019535910720 checkpointing.py:578] checkpoint manager exists so trying to load this run's existing checkpoint I0425 15:20:17.244744 133019535910720 checkpointing.py:676] No existing checkpoints found, not restoring checkpoint. fsdp: 32 I0425 15:20:18.690286 133019535910720 maxtext_utils.py:1740] params/params/decoder/decoder_norm/scale Shape: float32[2048] Logical: PartitionSpec('norm',) Physical: (None,) I0425 15:20:18.690427 133019535910720 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) I0425 15:20:18.690484 133019535910720 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) I0425 15:20:18.690545 133019535910720 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') I0425 15:20:18.690599 133019535910720 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) I0425 15:20:18.690639 133019535910720 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) I0425 15:20:18.690709 133019535910720 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) I0425 15:20:18.690766 133019535910720 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') I0425 15:20:18.690810 133019535910720 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) I0425 15:20:18.690851 133019535910720 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) I0425 15:20:18.690904 133019535910720 maxtext_utils.py:1740] params/params/decoder/logits_dense/kernel Shape: float32[2048,32000] Logical: PartitionSpec('embed', 'vocab') Physical: ('fsdp', None) I0425 15:20:18.690959 133019535910720 maxtext_utils.py:1740] params/params/token_embedder/embedding Shape: float32[32000,2048] Logical: PartitionSpec('vocab', 'embed') Physical: (None, 'fsdp') I0425 15:20:19.195165 133019535910720 train.py:158] train/xent Logical: float32[32,2048]............................................ ('activation_embed_and_logits_batch', 'activation_length'). I0425 15:20:19.195259 133019535910720 train.py:158] train/xent Physical: float32[32,2048]............................................ ('fsdp', None). I0425 15:20:19.210290 133019535910720 train.py:165] train/z_loss Logical: float32[32,2048]............................................ ('activation_embed_and_logits_batch', 'activation_length'). I0425 15:20:19.210350 133019535910720 train.py:165] train/z_loss Physical: float32[32,2048]............................................ ('fsdp', None). I0425 15:20:29.906025 133019535910720 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. I0425 15:20:29.906892 133019535910720 metric_logger.py:289] number parameters: 1.104 billion I0425 15:20:41.076799 133019535910720 checkpointing.py:794] Waiting for step 0 to finish before checkpoint... I0425 15:20:41.276606 133019535910720 checkpointing.py:798] Waited 0.19979071617126465 seconds for step 0 to finish before starting checkpointing. I0425 15:20:41.279219 133019535910720 checkpoint_manager.py:2013] [process=5][thread=MainThread][wait_until_finished] No Save Finalize thread to wait for. Returning. I0425 15:20:41.280909 133019535910720 checkpoint_manager.py:1518] [process=5] Saving checkpoint at step 0 I0425 15:20:41.283624 133019535910720 async_checkpointer.py:452] [process=5] Started async saving checkpoint to gs://lance-maxtext/linen_ckpt_xpk_feat_nnx_set_defaults_true_20260425_150412/linen_xpk_feat_nnx_set_defaults_true_20260425_150412_07_eval/checkpoints/0. I0425 15:20:41.916529 133019535910720 signaling_client.py:364] Using JaxDistributedSignalingClient I0425 15:20:41.917589 133019535910720 jax_array_handlers.py:358] Scheduling D2H of 39 prioritized jax.Array. I0425 15:20:41.917664 133019535910720 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 I0425 15:20:42.165854 133019535910720 base_pytree_checkpoint_handler.py:153] [process=5][thread=MainThread] Initiated "orbax.checkpoint._src.serialization.jax_array_handlers.ArrayHandler".serialize. Time taken: 0.249403s I0425 15:20:42.166024 133019535910720 base_pytree_checkpoint_handler.py:129] [process=5] /jax/checkpoint/write/blocking_gbytes_per_sec: 2.783 GiB/s (total gbytes: 1.5 GiB) (time elapsed: 0.5543646812438965 s) (per-host) I0425 15:20:42.166077 133019535910720 base_pytree_checkpoint_handler.py:737] [process=5][thread=MainThread] Initiated Pytree async_save. Time taken: 0.554427s (batch_requests_ready=0.301291s, total_serialization_initiated=0.253065s, others=0.000072s) I0425 15:20:42.166168 133019535910720 composite_checkpoint_handler.py:715] [process=5][thread=MainThread] Initiated CompositeCheckpointHandler.async_save. Time taken: 0.558290s (all_items=0.000019s, per_item={'items': '0.00001907'}, temp_paths=0.558270) I0425 15:20:42.167094 132893502322432 async_checkpointer.py:79] [process=5][thread=async_save] Background save thread started. I0425 15:20:42.167266 133019535910720 async_checkpointer.py:561] Finished blocking save. Time taken: 0.886283s. Continuing background save to gs://lance-maxtext/linen_ckpt_xpk_feat_nnx_set_defaults_true_20260425_150412/linen_xpk_feat_nnx_set_defaults_true_20260425_150412_07_eval/checkpoints/0. I0425 15:20:42.169192 133019535910720 checkpoint_manager.py:1566] [process=5][thread=MainThread][step=0] Starting CheckpointManager Save Finalize thread=save_finalize I0425 15:20:42.169486 132893527500544 async_checkpointer.py:265] [process=5][thread=save_finalize] Waiting for background save thread=async_save. I0425 15:20:42.169639 133019535910720 standard_logger.py:34] {'step': 0, 'event_type': 'save', 'directory': 'gs://lance-maxtext/linen_ckpt_xpk_feat_nnx_set_defaults_true_20260425_150412/linen_xpk_feat_nnx_set_defaults_true_20260425_150412_07_eval/checkpoints', 'reached_preemption': False, 'preemption_received_at': None, 'synchronous': False, 'wait_for_prev_start_time': 1777130441.2792, 'wait_for_prev_duration_secs': 6.270408630371094e-05, 'time_between_consecutive_saves_sec': None, 'checkpointer_blocking_start_time': 1777130441.2809505, 'checkpointer_blocking_duration_secs': 0.886420726776123, 'get_old_steps_start_time': 1777130442.1673915, 'get_old_steps_duration_secs': 2.5033950805664062e-05, 'checkpoint_manager_blocking_start_time': 1777130441.2772331, 'checkpoint_manager_blocking_duration_secs': 0.8923680782318115} I0425 15:20:42.169756 133019535910720 checkpointing.py:409] Started an asynchronous checkpoint save for step 0 I0425 15:20:42.169808 133019535910720 max_utils.py:750] Memstats: After params initialized: I0425 15:20:42.169857 133019535910720 max_utils.py:756] Using (GB) 0.43 / 31.25 (1.376000%) on TPU_18(process=5,(2,4,0,0)) I0425 15:20:42.169892 133019535910720 max_utils.py:756] Using (GB) 0.43 / 31.25 (1.376000%) on TPU_19(process=5,(3,4,0,0)) I0425 15:20:42.169918 133019535910720 max_utils.py:756] Using (GB) 0.43 / 31.25 (1.376000%) on TPU_22(process=5,(2,5,0,0)) I0425 15:20:42.169948 133019535910720 max_utils.py:756] Using (GB) 0.43 / 31.25 (1.376000%) on TPU_23(process=5,(3,5,0,0)) I0425 15:20:42.438111 133019535910720 metric_logger.py:185] completed step: 0, seconds: 11.170, TFLOP/s/device: 1.216, Tokens/s/device: 183.352, total_weights: 65536, loss: 10.874 I0425 15:20:42.609366 133019535910720 metric_logger.py:185] completed step: 1, seconds: 1.360, TFLOP/s/device: 9.991, Tokens/s/device: 1505.878, total_weights: 65536, loss: 10.874 I0425 15:20:43.048459 133019535910720 metric_logger.py:185] completed step: 2, seconds: 0.024, TFLOP/s/device: 564.625, Tokens/s/device: 85106.383, total_weights: 65536, loss: 10.263 I0425 15:20:43.195763 133019535910720 metric_logger.py:185] completed step: 3, seconds: 0.439, TFLOP/s/device: 30.918, Tokens/s/device: 4660.307, total_weights: 65536, loss: 9.734 I0425 15:20:43.491479 133019535910720 metric_logger.py:185] completed step: 4, seconds: 0.153, TFLOP/s/device: 89.033, Tokens/s/device: 13420.004, total_weights: 65536, loss: 9.278 I0425 15:20:43.497065 133019535910720 metric_logger.py:185] completed step: 5, seconds: 0.147, TFLOP/s/device: 92.229, Tokens/s/device: 13901.805, total_weights: 65536, loss: 8.892 I0425 15:21:02.137674 133019535910720 metric_logger.py:185] completed step: 6, seconds: 0.296, TFLOP/s/device: 45.895, Tokens/s/device: 6917.750, total_weights: 65536, loss: 8.592 I0425 15:21:02.285342 133019535910720 metric_logger.py:185] completed step: 7, seconds: 18.494, TFLOP/s/device: 0.735, Tokens/s/device: 110.739, total_weights: 65536, loss: 8.384 I0425 15:21:02.432643 133019535910720 metric_logger.py:185] completed step: 8, seconds: 0.152, TFLOP/s/device: 89.459, Tokens/s/device: 13484.241, total_weights: 65536, loss: 8.255 I0425 15:21:02.579818 133019535910720 checkpointing.py:794] Waiting for step 9 to finish before checkpoint... I0425 15:21:02.580455 133019535910720 checkpointing.py:798] Waited 0.0006573200225830078 seconds for step 9 to finish before starting checkpointing. I0425 15:21:02.582585 133019535910720 checkpoint_manager.py:2024] [process=5][thread=MainThread][step=0][wait_until_finished] Waiting for Save Finalize thread (save_finalize) to complete. I0425 15:21:03.870841 2623 google_auth_provider.cc:181] Running on GCE, using service account 562977990677-compute@developer.gserviceaccount.com I0425 15:21:05.792229 132893510715136 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_20260425_150412/linen_xpk_feat_nnx_set_defaults_true_20260425_150412_07_eval/checkpoints/0/items/array_metadatas/process_5 I0425 15:21:37.392025 132893502322432 base_pytree_checkpoint_handler.py:129] [process=5] /jax/checkpoint/write/gbytes_per_sec: 28.318 MiB/s (total gbytes: 1.5 GiB) (time elapsed: 55.780322313308716 s) (per-host) I0425 15:21:37.392144 132893502322432 async_checkpointer.py:90] [process=5][thread=async_save] 3 Handler Commit operations completed. Time taken: 55.224940s. I0425 15:21:45.595544 132893502322432 async_checkpointer.py:144] [process=5][thread=async_save] Background save thread done. Time taken: 63.428325s. I0425 15:21:45.595879 132893527500544 async_checkpointer.py:273] [process=5][thread=save_finalize] Done with waiting for background save thread=async_save. I0425 15:21:45.596006 132893527500544 async_checkpointer.py:283] [process=5][thread=save_finalize] No errors found in background save thread=async_save. I0425 15:21:45.596068 132893527500544 checkpoint_manager.py:2133] [process=5][thread=save_finalize][step=0] CheckpointManager Save Finalize is syncing with other hosts... I0425 15:21:45.597396 132893527500544 checkpoint_manager.py:2142] [process=5][thread=save_finalize][step=0] CheckpointManager Save Finalize is done on all hosts. I0425 15:21:45.597536 133019535910720 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. W0425 15:21:45.597615 133019535910720 checkpoint_manager.py:1458] Waiting for previous save to complete took 43.015037 seconds. If this number is high, consider checkpointing less frequently. I0425 15:21:45.599569 133019535910720 checkpoint_manager.py:1518] [process=5] Saving checkpoint at step 9 I0425 15:21:45.602999 133019535910720 async_checkpointer.py:452] [process=5] Started async saving checkpoint to gs://lance-maxtext/linen_ckpt_xpk_feat_nnx_set_defaults_true_20260425_150412/linen_xpk_feat_nnx_set_defaults_true_20260425_150412_07_eval/checkpoints/9. I0425 15:21:46.144018 133019535910720 jax_array_handlers.py:358] Scheduling D2H of 39 prioritized jax.Array. I0425 15:21:46.144118 133019535910720 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 I0425 15:21:46.189670 133019535910720 base_pytree_checkpoint_handler.py:153] [process=5][thread=MainThread] Initiated "orbax.checkpoint._src.serialization.jax_array_handlers.ArrayHandler".serialize. Time taken: 0.046971s I0425 15:21:46.189827 133019535910720 base_pytree_checkpoint_handler.py:129] [process=5] /jax/checkpoint/write/blocking_gbytes_per_sec: 5.236 GiB/s (total gbytes: 1.5 GiB) (time elapsed: 0.2946343421936035 s) (per-host) I0425 15:21:46.189877 133019535910720 base_pytree_checkpoint_handler.py:737] [process=5][thread=MainThread] Initiated Pytree async_save. Time taken: 0.294693s (batch_requests_ready=0.245807s, total_serialization_initiated=0.048820s, others=0.000066s) I0425 15:21:46.189966 133019535910720 composite_checkpoint_handler.py:715] [process=5][thread=MainThread] Initiated CompositeCheckpointHandler.async_save. Time taken: 0.298676s (all_items=0.000015s, per_item={'items': '0.00001550'}, temp_paths=0.298661) I0425 15:21:46.190874 132893510715136 async_checkpointer.py:79] [process=5][thread=async_save] Background save thread started. I0425 15:21:46.191056 133019535910720 async_checkpointer.py:561] Finished blocking save. Time taken: 0.591413s. Continuing background save to gs://lance-maxtext/linen_ckpt_xpk_feat_nnx_set_defaults_true_20260425_150412/linen_xpk_feat_nnx_set_defaults_true_20260425_150412_07_eval/checkpoints/9. I0425 15:21:46.227588 133019535910720 checkpoint_manager.py:1566] [process=5][thread=MainThread][step=9] Starting CheckpointManager Save Finalize thread=save_finalize I0425 15:21:46.227911 132893527500544 async_checkpointer.py:265] [process=5][thread=save_finalize] Waiting for background save thread=async_save. I0425 15:21:46.228059 133019535910720 standard_logger.py:34] {'step': 9, 'event_type': 'save', 'directory': 'gs://lance-maxtext/linen_ckpt_xpk_feat_nnx_set_defaults_true_20260425_150412/linen_xpk_feat_nnx_set_defaults_true_20260425_150412_07_eval/checkpoints', 'reached_preemption': False, 'preemption_received_at': None, 'synchronous': False, 'wait_for_prev_start_time': 1777130462.5825572, 'wait_for_prev_duration_secs': 43.015036821365356, 'time_between_consecutive_saves_sec': None, 'checkpointer_blocking_start_time': 1777130505.5996065, 'checkpointer_blocking_duration_secs': 0.5915775299072266, 'get_old_steps_start_time': 1777130506.1912093, 'get_old_steps_duration_secs': 2.956390380859375e-05, 'checkpoint_manager_blocking_start_time': 1777130462.5806952, 'checkpoint_manager_blocking_duration_secs': 43.64732813835144} I0425 15:21:46.228171 133019535910720 checkpointing.py:409] Started an asynchronous checkpoint save for step 9 I0425 15:21:46.228219 133019535910720 checkpoint_manager.py:2024] [process=5][thread=MainThread][step=9][wait_until_finished] Waiting for Save Finalize thread (save_finalize) to complete. I0425 15:21:51.160040 132893519107840 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_20260425_150412/linen_xpk_feat_nnx_set_defaults_true_20260425_150412_07_eval/checkpoints/9/items/array_metadatas/process_5 I0425 15:22:27.821790 132893510715136 base_pytree_checkpoint_handler.py:129] [process=5] /jax/checkpoint/write/gbytes_per_sec: 37.675 MiB/s (total gbytes: 1.5 GiB) (time elapsed: 41.92655396461487 s) (per-host) I0425 15:22:27.821908 132893510715136 async_checkpointer.py:90] [process=5][thread=async_save] 3 Handler Commit operations completed. Time taken: 41.630910s. I0425 15:22:35.284672 132893510715136 async_checkpointer.py:144] [process=5][thread=async_save] Background save thread done. Time taken: 49.093642s. I0425 15:22:35.284942 132893527500544 async_checkpointer.py:273] [process=5][thread=save_finalize] Done with waiting for background save thread=async_save. I0425 15:22:35.285056 132893527500544 async_checkpointer.py:283] [process=5][thread=save_finalize] No errors found in background save thread=async_save. I0425 15:22:35.285104 132893527500544 checkpoint_manager.py:2133] [process=5][thread=save_finalize][step=9] CheckpointManager Save Finalize is syncing with other hosts... I0425 15:22:35.286442 132893527500544 checkpoint_manager.py:2142] [process=5][thread=save_finalize][step=9] CheckpointManager Save Finalize is done on all hosts. I0425 15:22:35.286642 133019535910720 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. I0425 15:22:35.286795 133019535910720 checkpoint_manager.py:2013] [process=5][thread=MainThread][wait_until_finished] No Save Finalize thread to wait for. Returning. I0425 15:22:35.287522 133019535910720 metric_logger.py:185] completed step: 9, seconds: 0.148, TFLOP/s/device: 92.024, Tokens/s/device: 13870.828, 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: Sat Apr 25 15:22:45 UTC 2026 EXIT_CODE=0