main (b117f50cf) vs main (59e0f1759)| Metric | main b117f50cf | main 59e0f1759 | Diff (main − main) |
|---|---|---|---|
| Parameters | 1.104 billion | 1.104 billion | — |
| Final loss | 7.1810 | 7.1810 | 0 |
| TFLOP/s | 92.169 | 92.252 | +0.083 |
| Tok/s | 13892.7 | 13905.3 | +12.639 |
| Avg s/step | 7.514 | 7.668 | +0.154 |
| Memory % | 1.38 | 1.38 | 0 |
| JAX | 0.9.2 | 0.9.2 | — |
Diff = branch value − main value. Green = branch improved. Red = branch regressed.
XPK Start: Fri Apr 24 07:25:49 UTC 2026 PyTorch was not found. Models won't be available and only tokenizers, configuration and file/data utilities can be used. `rope_parameters`'s factor field must be a float >= 1, got 40 `rope_parameters`'s beta_fast field must be a float, got 32 `rope_parameters`'s beta_slow field must be a float, got 1 DeepseekV32Config got `key=rope_scaling` in kwargs but hasn't set it as attribute. For RoPE standardization you need to set `self.rope_parameters` in model's config. 2026-04-24 07:26:14.714987: E external/local_xla/xla/stream_executor/cuda/cuda_platform.cc:51] failed call to cuInit: INTERNAL: CUDA error: Failed call to cuInit: UNKNOWN ERROR (303) I0424 07:26:14.926568 132182297855808 max_utils.py:273] Attempting to initialize the jax distributed system... I0424 07:26:23.966484 132182297855808 distributed.py:149] Starting JAX distributed service on [::]:8482 I0424 07:26:23.968842 132182297855808 distributed.py:172] Connecting to JAX distributed service on mt-08-checkpoint-async--qpjns-slice-job-0-0.mt-08-checkpoint-async--qpjns:8482 I0424 07:26:24.820416 132182297855808 max_utils.py:284] Jax distributed system initialized! I0424 07:26:30.858678 132182297855808 max_utils.py:800] System Information: Jax Version: 0.9.2 I0424 07:26:30.858782 132182297855808 max_utils.py:801] System Information: Jaxlib Version: 0.9.2 I0424 07:26:30.858822 132182297855808 max_utils.py:802] System Information: Jax Backend: PJRT C API TFRT TPU v6 lite Built on Mar 4 2026 11:32:08 (1772652728) cl/878335365 I0424 07:26:30.858859 132182297855808 train_utils.py:361] WARNING: Sequence packing is essentially ignored for synthetic data. Please use a real dataset to use sequence packing. I0424 07:26:31.551337 132182297855808 maxtext_utils.py:1604] Num_devices: 32, shape (1, 1, 1, 32, 1, 1, 1, 1, 1, 1, 1, 1, 1) I0424 07:26:31.551615 132182297855808 checkpointing.py:677] Setting up checkpoint logger... I0424 07:26:31.551687 132182297855808 checkpointing.py:233] Creating checkpoint manager with ocdbt=True and zarr3=True I0424 07:26:31.551733 132182297855808 pytree_checkpoint_handler.py:592] save_device_host_concurrent_bytes=None I0424 07:26:31.552076 132182297855808 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 0x78377b914f80>, enable_pinned_host_transfer=False, save_concurrent_bytes: 96000000000 (89.4 GiB), restore_concurrent_bytes: 96000000000 (89.4 GiB) I0424 07:26:34.470185 132182297855808 checkpointing.py:265] Enabling policy for fixed interval checkpointing. I0424 07:26:34.470418 132182297855808 checkpoint_manager.py:708] [process=5][thread=MainThread] CheckpointManager init: checkpointers=None, item_names=('items',), item_handlers={'items': <orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeCheckpointHandler object at 0x782424048380>}, handler_registry=None I0424 07:26:34.470666 132182297855808 composite_checkpoint_handler.py:237] Deferred registration for item: "items". Adding handler `<orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeCheckpointHandler object at 0x782424048380>` for item "items" and save args `<class 'orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeSaveArgs'>` and restore args `<class 'orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeRestoreArgs'>` to `_handler_registry`. I0424 07:26:34.470715 132182297855808 composite_checkpoint_handler.py:237] Deferred registration for item: "metrics". Adding handler `<orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonCheckpointHandler object at 0x78235c4655b0>` for item "metrics" and save args `<class 'orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonSaveArgs'>` and restore args `<class 'orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonRestoreArgs'>` to `_handler_registry`. I0424 07:26:34.470749 132182297855808 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 0x782424048380>, ('items', <class 'orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeRestoreArgs'>): <orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeCheckpointHandler object at 0x782424048380>, ('metrics', <class 'orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonSaveArgs'>): <orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonCheckpointHandler object at 0x78235c4655b0>, ('metrics', <class 'orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonRestoreArgs'>): <orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonCheckpointHandler object at 0x78235c4655b0>}). I0424 07:26:34.471084 132182297855808 abstract_checkpointer.py:35] orbax-checkpoint version: 0.11.34 I0424 07:26:34.471151 132182297855808 async_checkpointer.py:192] [process=5][thread=MainThread] Using barrier_sync_fn: <function get_barrier_sync_fn.<locals>._fn at 0x78231c7e56c0> timeout: 1200 secs and primary_host=0 for async checkpoint writes I0424 07:26:35.188666 132182297855808 checkpoint_manager.py:1812] Found 0 checkpoint steps in gs://lance-maxtext/linen_ckpt_xpk_main_20260424_070227/linen_xpk_main_20260424_070227_08_checkpoint_async_true/checkpoints I0424 07:26:35.417847 132182297855808 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=5), preservation_policy=LatestN(n=None), prevent_write_metrics=False, enable_should_save_is_saving_in_progress_check=True, enable_per_process_directory_creation=False, lightweight_initialize=False), root_directory=gs://lance-maxtext/linen_ckpt_xpk_main_20260424_070227/linen_xpk_main_20260424_070227_08_checkpoint_async_true/checkpoints: <orbax.checkpoint.checkpoint_manager.CheckpointManager object at 0x78235c3023f0> I0424 07:26:35.418023 132182297855808 checkpointing.py:301] Checkpoint manager created! I0424 07:26:36.350642 132182297855808 nnx_wrappers.py:437] Unknown Logical: bfloat16[32,2048,2048]...................................... ('activation_batch', 'activation_norm_length', 'activation_embed'). I0424 07:26:36.350768 132182297855808 nnx_wrappers.py:437] Unknown Physical: bfloat16[32,2048,2048]...................................... ('fsdp', None, None). I0424 07:26:36.727665 132182297855808 attentions.py:1088] attentions/inputs_q Logical: bfloat16[32,2048,2048]...................................... ('activation_batch', 'activation_attn_length', 'activation_attn_embed'). I0424 07:26:36.727756 132182297855808 attentions.py:1088] attentions/inputs_q Physical: bfloat16[32,2048,2048]...................................... ('fsdp', None, None). I0424 07:26:36.744346 132182297855808 attentions.py:1089] attentions/inputs_kv Logical: bfloat16[32,2048,2048]...................................... ('activation_batch', 'activation_attn_length', 'activation_attn_embed'). I0424 07:26:36.744404 132182297855808 attentions.py:1089] attentions/inputs_kv Physical: bfloat16[32,2048,2048]...................................... ('fsdp', None, None). I0424 07:26:36.768352 132182297855808 attentions.py:1154] attentions/query Logical: bfloat16[32,2048,16,128].................................... ('activation_kv_batch', 'activation_attn_length', 'activation_kv_heads', 'activation_kv_head_dim'). I0424 07:26:36.768418 132182297855808 attentions.py:1154] attentions/query Physical: bfloat16[32,2048,16,128].................................... ('fsdp', None, None, None). I0424 07:26:36.784975 132182297855808 attentions.py:1155] attentions/key Logical: bfloat16[32,2048,16,128].................................... ('activation_kv_batch', 'activation_attn_length', 'activation_kv_heads', 'activation_kv_head_dim'). I0424 07:26:36.785037 132182297855808 attentions.py:1155] attentions/key Physical: bfloat16[32,2048,16,128].................................... ('fsdp', None, None, None). I0424 07:26:36.801455 132182297855808 attentions.py:1156] attentions/value Logical: bfloat16[32,2048,16,128].................................... ('activation_kv_batch', 'activation_attn_length', 'activation_kv_heads', 'activation_kv_head_dim'). I0424 07:26:36.801525 132182297855808 attentions.py:1156] attentions/value Physical: bfloat16[32,2048,16,128].................................... ('fsdp', None, None, None). I0424 07:26:36.826418 132182297855808 attentions.py:1198] attentions/out Logical: bfloat16[32,2048,16,128].................................... ('activation_batch', 'activation_attn_length', 'activation_heads', 'activation_kv'). I0424 07:26:36.826486 132182297855808 attentions.py:1198] attentions/out Physical: bfloat16[32,2048,16,128].................................... ('fsdp', None, None, None). I0424 07:26:36.847548 132182297855808 linears.py:525] linears/x Logical: bfloat16[32,2048,7168]...................................... ('activation_batch', 'activation_length', 'activation_mlp'). I0424 07:26:36.847616 132182297855808 linears.py:525] linears/x Physical: bfloat16[32,2048,7168]...................................... ('fsdp', None, None). I0424 07:26:37.056057 132182297855808 checkpointing.py:577] checkpoint manager exists so trying to load this run's existing checkpoint I0424 07:26:37.056163 132182297855808 checkpointing.py:665] No existing checkpoints found, not restoring checkpoint. fsdp: 32 I0424 07:26:38.499033 132182297855808 maxtext_utils.py:1707] params/params/decoder/decoder_norm/scale Shape: float32[2048] Logical: P('norm',) Physical: (None,) I0424 07:26:38.499160 132182297855808 maxtext_utils.py:1707] params/params/decoder/layers/mlp/wi_0/kernel Shape: float32[2048,16,7168] Logical: P('embed', 'layers', 'mlp') Physical: ('fsdp', None, None) I0424 07:26:38.499213 132182297855808 maxtext_utils.py:1707] params/params/decoder/layers/mlp/wi_1/kernel Shape: float32[2048,16,7168] Logical: P('embed', 'layers', 'mlp') Physical: ('fsdp', None, None) I0424 07:26:38.499270 132182297855808 maxtext_utils.py:1707] params/params/decoder/layers/mlp/wo/kernel Shape: float32[7168,16,2048] Logical: P('mlp', 'layers', 'embed') Physical: (None, None, 'fsdp') I0424 07:26:38.499320 132182297855808 maxtext_utils.py:1707] params/params/decoder/layers/post_self_attention_layer_norm/scale Shape: float32[2048,16] Logical: P('norm', 'layers') Physical: (None, None) I0424 07:26:38.499358 132182297855808 maxtext_utils.py:1707] params/params/decoder/layers/pre_self_attention_layer_norm/scale Shape: float32[2048,16] Logical: P('norm', 'layers') Physical: (None, None) I0424 07:26:38.499413 132182297855808 maxtext_utils.py:1707] params/params/decoder/layers/self_attention/key/kernel Shape: float32[2048,16,16,128] Logical: P('embed', 'layers', 'kv_heads', 'kv_head_dim') Physical: ('fsdp', None, None, None) I0424 07:26:38.499464 132182297855808 maxtext_utils.py:1707] params/params/decoder/layers/self_attention/out/kernel Shape: float32[16,16,128,2048] Logical: P('heads', 'layers', 'kv', 'embed') Physical: (None, None, None, 'fsdp') I0424 07:26:38.499504 132182297855808 maxtext_utils.py:1707] params/params/decoder/layers/self_attention/query/kernel Shape: float32[2048,16,16,128] Logical: P('embed', 'layers', 'q_heads', 'kv') Physical: ('fsdp', None, None, None) I0424 07:26:38.499538 132182297855808 maxtext_utils.py:1707] params/params/decoder/layers/self_attention/value/kernel Shape: float32[2048,16,16,128] Logical: P('embed', 'layers', 'kv_heads', 'kv_head_dim') Physical: ('fsdp', None, None, None) I0424 07:26:38.499584 132182297855808 maxtext_utils.py:1707] params/params/decoder/logits_dense/kernel Shape: float32[2048,32000] Logical: P('embed_vocab', 'vocab') Physical: ('fsdp', None) I0424 07:26:38.499629 132182297855808 maxtext_utils.py:1707] params/params/token_embedder/embedding Shape: float32[32000,2048] Logical: P('vocab', 'embed_vocab') Physical: (None, 'fsdp') I0424 07:26:38.995270 132182297855808 train.py:155] train/xent Logical: float32[32,2048]............................................ ('activation_embed_and_logits_batch', 'activation_length'). I0424 07:26:38.995367 132182297855808 train.py:155] train/xent Physical: float32[32,2048]............................................ ('fsdp', None). I0424 07:26:39.012314 132182297855808 train.py:162] train/z_loss Logical: float32[32,2048]............................................ ('activation_embed_and_logits_batch', 'activation_length'). I0424 07:26:39.012374 132182297855808 train.py:162] train/z_loss Physical: float32[32,2048]............................................ ('fsdp', None). I0424 07:26:50.111450 132182297855808 max_utils.py:791] Total memory size: 1.5 GB, Output size: 0.4 GB, Temp size: 1.1 GB, Argument size: 0.4 GB, Host temp size: 0.0 GB. I0424 07:26:50.112218 132182297855808 metric_logger.py:301] number parameters: 1.104 billion I0424 07:27:01.364001 132182297855808 checkpointing.py:772] Waiting for step 0 to finish before checkpoint... I0424 07:27:01.655081 132182297855808 checkpointing.py:776] Waited 0.291057825088501 seconds for step 0 to finish before starting checkpointing. I0424 07:27:01.657471 132182297855808 checkpoint_manager.py:2009] [process=5][thread=MainThread][wait_until_finished] No Save Finalize thread to wait for. Returning. I0424 07:27:01.659106 132182297855808 checkpoint_manager.py:1512] [process=5] Saving checkpoint at step 0 I0424 07:27:01.660540 132182297855808 event_tracking.py:70] [process=5] [async] Started save checkpoint @ gs://lance-maxtext/linen_ckpt_xpk_main_20260424_070227/linen_xpk_main_20260424_070227_08_checkpoint_async_true/checkpoints/0. I0424 07:27:02.116543 132182297855808 signaling_client.py:364] Using JaxDistributedSignalingClient I0424 07:27:02.117486 132182297855808 jax_array_handlers.py:360] Scheduling D2H of 39 prioritized jax.Array. I0424 07:27:02.117541 132182297855808 replica_slices.py:424] Transferring arrays to host memory with options: use_replica_parallel=True, min_slice_bytes_for_replica_parallel=None, max_replicas_for_replica_parallel=None, enable_pinned_host_transfer=False I0424 07:27:02.390022 132182297855808 base_pytree_checkpoint_handler.py:154] [process=5][thread=MainThread] Initiated "orbax.checkpoint._src.serialization.jax_array_handlers.ArrayHandler".serialize. Time taken: 0.273550s I0424 07:27:02.390195 132182297855808 base_pytree_checkpoint_handler.py:130] [process=5] /jax/orbax/write/blocking_gbytes_per_sec: 5.526 GiB/s (total gbytes: 1.5 GiB) (time elapsed: 0.27915406227111816 s) (per-host) I0424 07:27:02.390246 132182297855808 base_pytree_checkpoint_handler.py:768] [process=5][thread=MainThread] Initiated Pytree async_save. Time taken: 0.279215s (batch_requests_ready=0.002474s, total_serialization_initiated=0.276672s, others=0.000069s) I0424 07:27:02.390341 132182297855808 composite_checkpoint_handler.py:715] [process=5][thread=MainThread] Initiated CompositeCheckpointHandler.async_save. Time taken: 0.283264s (all_items=0.000018s, per_item={'items': '0.00001788'}, temp_paths=0.283246) I0424 07:27:02.391202 132182297855808 event_tracking.py:125] [process=5] [async] Finished blocking save in 0.73 seconds. Continuing save @ gs://lance-maxtext/linen_ckpt_xpk_main_20260424_070227/linen_xpk_main_20260424_070227_08_checkpoint_async_true/checkpoints/0. I0424 07:27:02.391535 132052924151552 async_checkpointer.py:76] [process=5][thread=async_save] Background save thread started. Deadline for this save operation is 2026-04-24 07:47:02.391498 I0424 07:27:02.401011 132182297855808 checkpoint_manager.py:1560] [process=5][thread=MainThread][step=0] Starting CheckpointManager Save Finalize thread=save_finalize I0424 07:27:02.401299 132050237515520 async_checkpointer.py:280] [process=5][thread=save_finalize] Waiting for background save thread=async_save. I0424 07:27:02.401455 132182297855808 standard_logger.py:34] {'step': 0, 'event_type': 'save', 'directory': 'gs://lance-maxtext/linen_ckpt_xpk_main_20260424_070227/linen_xpk_main_20260424_070227_08_checkpoint_async_true/checkpoints', 'reached_preemption': False, 'preemption_received_at': None, 'synchronous': False, 'wait_for_prev_start_time': 1777015621.6574535, 'wait_for_prev_duration_secs': 5.817413330078125e-05, 'time_between_consecutive_saves_sec': None, 'checkpointer_blocking_start_time': 1777015621.6591444, 'checkpointer_blocking_duration_secs': 0.7325608730316162, 'get_old_steps_start_time': 1777015622.3917296, 'get_old_steps_duration_secs': 3.170967102050781e-05, 'checkpoint_manager_blocking_start_time': 1777015621.655488, 'checkpoint_manager_blocking_duration_secs': 0.7459287643432617} I0424 07:27:02.401562 132182297855808 checkpointing.py:408] Started an asynchronous checkpoint save for step 0 I0424 07:27:02.401611 132182297855808 max_utils.py:750] Memstats: After params initialized: I0424 07:27:02.401673 132182297855808 max_utils.py:756] Using (GB) 0.43 / 31.25 (1.376000%) on TPU_18(process=5,(2,4,0,0)) I0424 07:27:02.401708 132182297855808 max_utils.py:756] Using (GB) 0.43 / 31.25 (1.376000%) on TPU_19(process=5,(3,4,0,0)) I0424 07:27:02.401735 132182297855808 max_utils.py:756] Using (GB) 0.43 / 31.25 (1.376000%) on TPU_22(process=5,(2,5,0,0)) I0424 07:27:02.401761 132182297855808 max_utils.py:756] Using (GB) 0.43 / 31.25 (1.376000%) on TPU_23(process=5,(3,5,0,0)) I0424 07:27:02.712519 132182297855808 metric_logger.py:196] completed step: 0, seconds: 11.252, TFLOP/s/device: 1.208, Tokens/s/device: 182.017, total_weights: 65536, loss: 10.874, lm_loss: 10.874, perplexity: 52779.875 I0424 07:27:02.902503 132182297855808 metric_logger.py:196] completed step: 1, seconds: 1.347, TFLOP/s/device: 10.087, Tokens/s/device: 1520.399, total_weights: 65536, loss: 10.874, lm_loss: 10.874, perplexity: 52779.875 I0424 07:27:03.325387 132182297855808 metric_logger.py:196] completed step: 2, seconds: 0.043, TFLOP/s/device: 318.745, Tokens/s/device: 48044.667, total_weights: 65536, loss: 10.560, lm_loss: 10.560, perplexity: 38547.719 I0424 07:27:03.472686 132182297855808 metric_logger.py:196] completed step: 3, seconds: 0.424, TFLOP/s/device: 32.073, Tokens/s/device: 4834.350, total_weights: 65536, loss: 9.980, lm_loss: 9.980, perplexity: 21590.506 I0424 07:27:03.477046 132182297855808 checkpointing.py:772] Waiting for step 5 to finish before checkpoint... I0424 07:27:03.766779 132182297855808 checkpointing.py:776] Waited 0.28969550132751465 seconds for step 5 to finish before starting checkpointing. I0424 07:27:03.769819 132182297855808 checkpoint_manager.py:2020] [process=5][thread=MainThread][step=0][wait_until_finished] Waiting for Save Finalize thread (save_finalize) to complete. I0424 07:27:05.473943 2850 google_auth_provider.cc:181] Running on GCE, using service account 562977990677-compute@developer.gserviceaccount.com I0424 07:27:07.692886 132050245908224 array_metadata_store.py:203] [process=5][thread=array_type_handler] Wrote 39 array_metadata.ArrayMetadata to gs://lance-maxtext/linen_ckpt_xpk_main_20260424_070227/linen_xpk_main_20260424_070227_08_checkpoint_async_true/checkpoints/0/items/array_metadatas/process_5 I0424 07:27:39.457156 132052924151552 base_pytree_checkpoint_handler.py:130] [process=5] /jax/orbax/write/gbytes_per_sec: 42.296 MiB/s (total gbytes: 1.5 GiB) (time elapsed: 37.34607672691345 s) (per-host) I0424 07:27:39.457284 132052924151552 async_checkpointer.py:90] [process=5][thread=async_save] 3 Handler Commit operations completed. Time taken: 37.065639s. I0424 07:27:45.803025 132052924151552 async_checkpointer.py:160] [process=5][thread=async_save] Background save thread done. Time taken: 43.411358s. I0424 07:27:45.803399 132050237515520 async_checkpointer.py:288] [process=5][thread=save_finalize] Done with waiting for background save thread=async_save. I0424 07:27:45.803515 132050237515520 async_checkpointer.py:298] [process=5][thread=save_finalize] No errors found in background save thread=async_save. I0424 07:27:45.803565 132050237515520 checkpoint_manager.py:2137] [process=5][thread=save_finalize][step=0] CheckpointManager Save Finalize is syncing with other hosts... I0424 07:27:45.805198 132050237515520 checkpoint_manager.py:2146] [process=5][thread=save_finalize][step=0] CheckpointManager Save Finalize is done on all hosts. I0424 07:27:45.805411 132182297855808 checkpoint_manager.py:2032] [process=5][thread=MainThread][step=0][wait_until_finished] Done waiting for Save Finalize thread (save_finalize) running at step=0. W0424 07:27:45.805541 132182297855808 checkpoint_manager.py:1452] Waiting for previous save to complete took 42.035719 seconds. If this number is high, consider checkpointing less frequently. I0424 07:27:45.807333 132182297855808 checkpoint_manager.py:1512] [process=5] Saving checkpoint at step 5 I0424 07:27:45.809852 132182297855808 event_tracking.py:70] [process=5] [async] Started save checkpoint @ gs://lance-maxtext/linen_ckpt_xpk_main_20260424_070227/linen_xpk_main_20260424_070227_08_checkpoint_async_true/checkpoints/5. I0424 07:27:46.096232 132182297855808 jax_array_handlers.py:360] Scheduling D2H of 39 prioritized jax.Array. I0424 07:27:46.096329 132182297855808 replica_slices.py:424] Transferring arrays to host memory with options: use_replica_parallel=True, min_slice_bytes_for_replica_parallel=None, max_replicas_for_replica_parallel=None, enable_pinned_host_transfer=False I0424 07:27:46.137612 132182297855808 base_pytree_checkpoint_handler.py:154] [process=5][thread=MainThread] Initiated "orbax.checkpoint._src.serialization.jax_array_handlers.ArrayHandler".serialize. Time taken: 0.042288s I0424 07:27:46.137782 132182297855808 base_pytree_checkpoint_handler.py:130] [process=5] /jax/orbax/write/blocking_gbytes_per_sec: 33.705 GiB/s (total gbytes: 1.5 GiB) (time elapsed: 0.04576706886291504 s) (per-host) I0424 07:27:46.137833 132182297855808 base_pytree_checkpoint_handler.py:768] [process=5][thread=MainThread] Initiated Pytree async_save. Time taken: 0.045827s (batch_requests_ready=0.001760s, total_serialization_initiated=0.043998s, others=0.000069s) I0424 07:27:46.137919 132182297855808 composite_checkpoint_handler.py:715] [process=5][thread=MainThread] Initiated CompositeCheckpointHandler.async_save. Time taken: 0.050530s (all_items=0.000016s, per_item={'items': '0.00001597'}, temp_paths=0.050514) I0424 07:27:46.138668 132182297855808 event_tracking.py:125] [process=5] [async] Finished blocking save in 0.33 seconds. Continuing save @ gs://lance-maxtext/linen_ckpt_xpk_main_20260424_070227/linen_xpk_main_20260424_070227_08_checkpoint_async_true/checkpoints/5. I0424 07:27:46.138944 132050237515520 async_checkpointer.py:76] [process=5][thread=async_save] Background save thread started. Deadline for this save operation is 2026-04-24 07:47:46.138920 I0424 07:27:46.140966 132182297855808 checkpoint_manager.py:1560] [process=5][thread=MainThread][step=5] Starting CheckpointManager Save Finalize thread=save_finalize I0424 07:27:46.141226 132028351633152 async_checkpointer.py:280] [process=5][thread=save_finalize] Waiting for background save thread=async_save. I0424 07:27:46.141383 132182297855808 standard_logger.py:34] {'step': 5, 'event_type': 'save', 'directory': 'gs://lance-maxtext/linen_ckpt_xpk_main_20260424_070227/linen_xpk_main_20260424_070227_08_checkpoint_async_true/checkpoints', 'reached_preemption': False, 'preemption_received_at': None, 'synchronous': False, 'wait_for_prev_start_time': 1777015623.7697906, 'wait_for_prev_duration_secs': 42.03571891784668, 'time_between_consecutive_saves_sec': None, 'checkpointer_blocking_start_time': 1777015665.8073726, 'checkpointer_blocking_duration_secs': 0.3316676616668701, 'get_old_steps_start_time': 1777015666.1390646, 'get_old_steps_duration_secs': 3.0040740966796875e-05, 'checkpoint_manager_blocking_start_time': 1777015623.7674108, 'checkpoint_manager_blocking_duration_secs': 42.373936891555786} I0424 07:27:46.141495 132182297855808 checkpointing.py:408] Started an asynchronous checkpoint save for step 5 I0424 07:27:46.142325 132182297855808 metric_logger.py:196] completed step: 4, seconds: 0.153, TFLOP/s/device: 89.049, Tokens/s/device: 13422.379, total_weights: 65536, loss: 9.460, lm_loss: 9.460, perplexity: 12840.355 I0424 07:27:46.148316 132182297855808 metric_logger.py:196] completed step: 5, seconds: 0.147, TFLOP/s/device: 92.384, Tokens/s/device: 13925.152, total_weights: 65536, loss: 8.959, lm_loss: 8.959, perplexity: 7776.467 I0424 07:27:51.207192 132050245908224 array_metadata_store.py:203] [process=5][thread=array_type_handler] Wrote 39 array_metadata.ArrayMetadata to gs://lance-maxtext/linen_ckpt_xpk_main_20260424_070227/linen_xpk_main_20260424_070227_08_checkpoint_async_true/checkpoints/5/items/array_metadatas/process_5 I0424 07:28:08.839838 132182297855808 metric_logger.py:196] completed step: 6, seconds: 42.670, TFLOP/s/device: 0.318, Tokens/s/device: 47.997, total_weights: 65536, loss: 8.469, lm_loss: 8.469, perplexity: 4762.740 I0424 07:28:08.987189 132182297855808 metric_logger.py:196] completed step: 7, seconds: 22.546, TFLOP/s/device: 0.603, Tokens/s/device: 90.838, total_weights: 65536, loss: 8.003, lm_loss: 8.003, perplexity: 2989.151 I0424 07:28:09.134722 132182297855808 metric_logger.py:196] completed step: 8, seconds: 0.152, TFLOP/s/device: 89.386, Tokens/s/device: 13473.241, total_weights: 65536, loss: 7.572, lm_loss: 7.572, perplexity: 1943.098 I0424 07:28:09.281468 132182297855808 checkpointing.py:772] Waiting for step 9 to finish before checkpoint... I0424 07:28:09.282068 132182297855808 checkpointing.py:776] Waited 0.000621795654296875 seconds for step 9 to finish before starting checkpointing. I0424 07:28:09.284101 132182297855808 checkpoint_manager.py:2020] [process=5][thread=MainThread][step=5][wait_until_finished] Waiting for Save Finalize thread (save_finalize) to complete. I0424 07:28:28.131081 132050237515520 base_pytree_checkpoint_handler.py:130] [process=5] /jax/orbax/write/gbytes_per_sec: 37.575 MiB/s (total gbytes: 1.5 GiB) (time elapsed: 42.0390191078186 s) (per-host) I0424 07:28:28.131207 132050237515520 async_checkpointer.py:90] [process=5][thread=async_save] 3 Handler Commit operations completed. Time taken: 41.992188s. I0424 07:28:34.376891 132050237515520 async_checkpointer.py:160] [process=5][thread=async_save] Background save thread done. Time taken: 48.237857s. I0424 07:28:34.377184 132028351633152 async_checkpointer.py:288] [process=5][thread=save_finalize] Done with waiting for background save thread=async_save. I0424 07:28:34.377297 132028351633152 async_checkpointer.py:298] [process=5][thread=save_finalize] No errors found in background save thread=async_save. I0424 07:28:34.377346 132028351633152 checkpoint_manager.py:2137] [process=5][thread=save_finalize][step=5] CheckpointManager Save Finalize is syncing with other hosts... I0424 07:28:34.378902 132028351633152 checkpoint_manager.py:2146] [process=5][thread=save_finalize][step=5] CheckpointManager Save Finalize is done on all hosts. I0424 07:28:34.379077 132182297855808 checkpoint_manager.py:2032] [process=5][thread=MainThread][step=5][wait_until_finished] Done waiting for Save Finalize thread (save_finalize) running at step=5. W0424 07:28:34.379205 132182297855808 checkpoint_manager.py:1452] Waiting for previous save to complete took 25.095105 seconds. If this number is high, consider checkpointing less frequently. I0424 07:28:34.380578 132182297855808 checkpoint_manager.py:1512] [process=5] Saving checkpoint at step 9 I0424 07:28:34.382733 132182297855808 event_tracking.py:70] [process=5] [async] Started save checkpoint @ gs://lance-maxtext/linen_ckpt_xpk_main_20260424_070227/linen_xpk_main_20260424_070227_08_checkpoint_async_true/checkpoints/9. I0424 07:28:34.687895 132182297855808 jax_array_handlers.py:360] Scheduling D2H of 39 prioritized jax.Array. I0424 07:28:34.687988 132182297855808 replica_slices.py:424] Transferring arrays to host memory with options: use_replica_parallel=True, min_slice_bytes_for_replica_parallel=None, max_replicas_for_replica_parallel=None, enable_pinned_host_transfer=False I0424 07:28:34.727708 132182297855808 base_pytree_checkpoint_handler.py:154] [process=5][thread=MainThread] Initiated "orbax.checkpoint._src.serialization.jax_array_handlers.ArrayHandler".serialize. Time taken: 0.040644s I0424 07:28:34.727870 132182297855808 base_pytree_checkpoint_handler.py:130] [process=5] /jax/orbax/write/blocking_gbytes_per_sec: 34.972 GiB/s (total gbytes: 1.5 GiB) (time elapsed: 0.04410886764526367 s) (per-host) I0424 07:28:34.727922 132182297855808 base_pytree_checkpoint_handler.py:768] [process=5][thread=MainThread] Initiated Pytree async_save. Time taken: 0.044171s (batch_requests_ready=0.001640s, total_serialization_initiated=0.042464s, others=0.000067s) I0424 07:28:34.728013 132182297855808 composite_checkpoint_handler.py:715] [process=5][thread=MainThread] Initiated CompositeCheckpointHandler.async_save. Time taken: 0.048144s (all_items=0.000011s, per_item={'items': '0.00001121'}, temp_paths=0.048133) I0424 07:28:34.728754 132182297855808 event_tracking.py:125] [process=5] [async] Finished blocking save in 0.35 seconds. Continuing save @ gs://lance-maxtext/linen_ckpt_xpk_main_20260424_070227/linen_xpk_main_20260424_070227_08_checkpoint_async_true/checkpoints/9. I0424 07:28:34.729083 132028351633152 async_checkpointer.py:76] [process=5][thread=async_save] Background save thread started. Deadline for this save operation is 2026-04-24 07:48:34.729046 I0424 07:28:34.731059 132182297855808 checkpoint_manager.py:1560] [process=5][thread=MainThread][step=9] Starting CheckpointManager Save Finalize thread=save_finalize I0424 07:28:34.731337 132050245908224 async_checkpointer.py:280] [process=5][thread=save_finalize] Waiting for background save thread=async_save. I0424 07:28:34.731512 132182297855808 standard_logger.py:34] {'step': 9, 'event_type': 'save', 'directory': 'gs://lance-maxtext/linen_ckpt_xpk_main_20260424_070227/linen_xpk_main_20260424_070227_08_checkpoint_async_true/checkpoints', 'reached_preemption': False, 'preemption_received_at': None, 'synchronous': False, 'wait_for_prev_start_time': 1777015689.2840707, 'wait_for_prev_duration_secs': 25.095104932785034, 'time_between_consecutive_saves_sec': 23.478833436965942, 'checkpointer_blocking_start_time': 1777015714.3806174, 'checkpointer_blocking_duration_secs': 0.34860992431640625, 'get_old_steps_start_time': 1777015714.7292442, 'get_old_steps_duration_secs': 2.47955322265625e-05, 'checkpoint_manager_blocking_start_time': 1777015689.2823067, 'checkpoint_manager_blocking_duration_secs': 25.449174165725708} I0424 07:28:34.731620 132182297855808 checkpointing.py:408] Started an asynchronous checkpoint save for step 9 I0424 07:28:34.731680 132182297855808 checkpoint_manager.py:2020] [process=5][thread=MainThread][step=9][wait_until_finished] Waiting for Save Finalize thread (save_finalize) to complete. I0424 07:28:40.238512 132050757601024 array_metadata_store.py:203] [process=5][thread=array_type_handler] Wrote 39 array_metadata.ArrayMetadata to gs://lance-maxtext/linen_ckpt_xpk_main_20260424_070227/linen_xpk_main_20260424_070227_08_checkpoint_async_true/checkpoints/9/items/array_metadatas/process_5 I0424 07:29:17.489492 132028351633152 base_pytree_checkpoint_handler.py:130] [process=5] /jax/orbax/write/gbytes_per_sec: 36.902 MiB/s (total gbytes: 1.5 GiB) (time elapsed: 42.80569052696228 s) (per-host) I0424 07:29:17.489612 132028351633152 async_checkpointer.py:90] [process=5][thread=async_save] 3 Handler Commit operations completed. Time taken: 42.760418s. I0424 07:29:23.017033 132028351633152 async_checkpointer.py:160] [process=5][thread=async_save] Background save thread done. Time taken: 48.287822s. I0424 07:29:23.017348 132050245908224 async_checkpointer.py:288] [process=5][thread=save_finalize] Done with waiting for background save thread=async_save. I0424 07:29:23.017466 132050245908224 async_checkpointer.py:298] [process=5][thread=save_finalize] No errors found in background save thread=async_save. I0424 07:29:23.017507 132050245908224 checkpoint_manager.py:2137] [process=5][thread=save_finalize][step=9] CheckpointManager Save Finalize is syncing with other hosts... I0424 07:29:23.018854 132050245908224 checkpoint_manager.py:2146] [process=5][thread=save_finalize][step=9] CheckpointManager Save Finalize is done on all hosts. I0424 07:29:23.019030 132182297855808 checkpoint_manager.py:2032] [process=5][thread=MainThread][step=9][wait_until_finished] Done waiting for Save Finalize thread (save_finalize) running at step=9. I0424 07:29:23.019180 132182297855808 checkpoint_manager.py:2009] [process=5][thread=MainThread][wait_until_finished] No Save Finalize thread to wait for. Returning. I0424 07:29:23.020132 132182297855808 metric_logger.py:196] completed step: 9, seconds: 0.147, TFLOP/s/device: 92.169, Tokens/s/device: 13892.658, total_weights: 65536, loss: 7.181, lm_loss: 7.181, perplexity: 1314.495 Per train step: Total TFLOPs: 13.59 split as 93.93% learnable weight flops and 6.07% attention flops XPK End: Fri Apr 24 07:29:35 UTC 2026 EXIT_CODE=0
XPK Start: Sat Apr 25 07:33:56 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 07:34:21.450062: 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 07:34:21.661243 139204217857856 max_utils.py:273] Attempting to initialize the jax distributed system... I0425 07:34:30.701868 139204217857856 distributed.py:149] Starting JAX distributed service on [::]:8482 I0425 07:34:30.704129 139204217857856 distributed.py:172] Connecting to JAX distributed service on mt-08-checkpoint-async--1r37d-slice-job-0-0.mt-08-checkpoint-async--1r37d:8482 I0425 07:34:31.918335 139204217857856 max_utils.py:284] Jax distributed system initialized! I0425 07:34:38.025483 139204217857856 max_utils.py:800] System Information: Jax Version: 0.9.2 I0425 07:34:38.025585 139204217857856 max_utils.py:801] System Information: Jaxlib Version: 0.9.2 I0425 07:34:38.025633 139204217857856 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 07:34:38.025692 139204217857856 train_utils.py:361] WARNING: Sequence packing is essentially ignored for synthetic data. Please use a real dataset to use sequence packing. I0425 07:34:38.719029 139204217857856 maxtext_utils.py:1604] Num_devices: 32, shape (1, 1, 1, 32, 1, 1, 1, 1, 1, 1, 1, 1, 1) I0425 07:34:38.719330 139204217857856 checkpointing.py:677] Setting up checkpoint logger... I0425 07:34:38.719394 139204217857856 checkpointing.py:233] Creating checkpoint manager with ocdbt=True and zarr3=True I0425 07:34:38.719449 139204217857856 pytree_checkpoint_handler.py:592] save_device_host_concurrent_bytes=None I0425 07:34:38.719814 139204217857856 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 0x7e9a709e7680>, enable_pinned_host_transfer=False, save_concurrent_bytes: 96000000000 (89.4 GiB), restore_concurrent_bytes: 96000000000 (89.4 GiB) I0425 07:34:41.552226 139204217857856 checkpointing.py:265] Enabling policy for fixed interval checkpointing. I0425 07:34:41.552459 139204217857856 checkpoint_manager.py:708] [process=5][thread=MainThread] CheckpointManager init: checkpointers=None, item_names=('items',), item_handlers={'items': <orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeCheckpointHandler object at 0x7e861c40d8b0>}, handler_registry=None I0425 07:34:41.552722 139204217857856 composite_checkpoint_handler.py:237] Deferred registration for item: "items". Adding handler `<orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeCheckpointHandler object at 0x7e861c40d8b0>` 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 07:34:41.552786 139204217857856 composite_checkpoint_handler.py:237] Deferred registration for item: "metrics". Adding handler `<orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonCheckpointHandler object at 0x7e861c40df70>` 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 07:34:41.552838 139204217857856 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 0x7e861c40d8b0>, ('items', <class 'orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeRestoreArgs'>): <orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeCheckpointHandler object at 0x7e861c40d8b0>, ('metrics', <class 'orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonSaveArgs'>): <orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonCheckpointHandler object at 0x7e861c40df70>, ('metrics', <class 'orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonRestoreArgs'>): <orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonCheckpointHandler object at 0x7e861c40df70>}). I0425 07:34:41.553194 139204217857856 abstract_checkpointer.py:35] orbax-checkpoint version: 0.11.34 I0425 07:34:41.553286 139204217857856 async_checkpointer.py:192] [process=5][thread=MainThread] Using barrier_sync_fn: <function get_barrier_sync_fn.<locals>._fn at 0x7e861c1b9800> timeout: 1200 secs and primary_host=0 for async checkpoint writes I0425 07:34:42.995147 139204217857856 checkpoint_manager.py:1812] Found 0 checkpoint steps in gs://lance-maxtext/linen_ckpt_xpk_main_20260425_071456/linen_xpk_main_20260425_071456_08_checkpoint_async_true/checkpoints I0425 07:34:43.160899 139204217857856 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=5), preservation_policy=LatestN(n=None), prevent_write_metrics=False, enable_should_save_is_saving_in_progress_check=True, enable_per_process_directory_creation=False, lightweight_initialize=False), root_directory=gs://lance-maxtext/linen_ckpt_xpk_main_20260425_071456/linen_xpk_main_20260425_071456_08_checkpoint_async_true/checkpoints: <orbax.checkpoint.checkpoint_manager.CheckpointManager object at 0x7e861c40ddc0> I0425 07:34:43.161085 139204217857856 checkpointing.py:301] Checkpoint manager created! I0425 07:34:44.097968 139204217857856 nnx_wrappers.py:437] Unknown Logical: bfloat16[32,2048,2048]...................................... ('activation_batch', 'activation_norm_length', 'activation_embed'). I0425 07:34:44.098085 139204217857856 nnx_wrappers.py:437] Unknown Physical: bfloat16[32,2048,2048]...................................... ('fsdp', None, None). I0425 07:34:44.477067 139204217857856 attentions.py:1088] attentions/inputs_q Logical: bfloat16[32,2048,2048]...................................... ('activation_batch', 'activation_attn_length', 'activation_attn_embed'). I0425 07:34:44.477160 139204217857856 attentions.py:1088] attentions/inputs_q Physical: bfloat16[32,2048,2048]...................................... ('fsdp', None, None). I0425 07:34:44.493710 139204217857856 attentions.py:1089] attentions/inputs_kv Logical: bfloat16[32,2048,2048]...................................... ('activation_batch', 'activation_attn_length', 'activation_attn_embed'). I0425 07:34:44.493769 139204217857856 attentions.py:1089] attentions/inputs_kv Physical: bfloat16[32,2048,2048]...................................... ('fsdp', None, None). I0425 07:34:44.517847 139204217857856 attentions.py:1154] attentions/query Logical: bfloat16[32,2048,16,128].................................... ('activation_kv_batch', 'activation_attn_length', 'activation_kv_heads', 'activation_kv_head_dim'). I0425 07:34:44.517917 139204217857856 attentions.py:1154] attentions/query Physical: bfloat16[32,2048,16,128].................................... ('fsdp', None, None, None). I0425 07:34:44.534538 139204217857856 attentions.py:1155] attentions/key Logical: bfloat16[32,2048,16,128].................................... ('activation_kv_batch', 'activation_attn_length', 'activation_kv_heads', 'activation_kv_head_dim'). I0425 07:34:44.534600 139204217857856 attentions.py:1155] attentions/key Physical: bfloat16[32,2048,16,128].................................... ('fsdp', None, None, None). I0425 07:34:44.551131 139204217857856 attentions.py:1156] attentions/value Logical: bfloat16[32,2048,16,128].................................... ('activation_kv_batch', 'activation_attn_length', 'activation_kv_heads', 'activation_kv_head_dim'). I0425 07:34:44.551208 139204217857856 attentions.py:1156] attentions/value Physical: bfloat16[32,2048,16,128].................................... ('fsdp', None, None, None). I0425 07:34:44.576379 139204217857856 attentions.py:1198] attentions/out Logical: bfloat16[32,2048,16,128].................................... ('activation_batch', 'activation_attn_length', 'activation_heads', 'activation_kv'). I0425 07:34:44.576450 139204217857856 attentions.py:1198] attentions/out Physical: bfloat16[32,2048,16,128].................................... ('fsdp', None, None, None). I0425 07:34:44.598085 139204217857856 linears.py:525] linears/x Logical: bfloat16[32,2048,7168]...................................... ('activation_batch', 'activation_length', 'activation_mlp'). I0425 07:34:44.598152 139204217857856 linears.py:525] linears/x Physical: bfloat16[32,2048,7168]...................................... ('fsdp', None, None). I0425 07:34:44.811853 139204217857856 checkpointing.py:577] checkpoint manager exists so trying to load this run's existing checkpoint I0425 07:34:44.811963 139204217857856 checkpointing.py:665] No existing checkpoints found, not restoring checkpoint. fsdp: 32 I0425 07:34:46.245573 139204217857856 maxtext_utils.py:1707] params/params/decoder/decoder_norm/scale Shape: float32[2048] Logical: P('norm',) Physical: (None,) I0425 07:34:46.245720 139204217857856 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 07:34:46.245779 139204217857856 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 07:34:46.245839 139204217857856 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 07:34:46.245891 139204217857856 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 07:34:46.245930 139204217857856 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 07:34:46.245988 139204217857856 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 07:34:46.246041 139204217857856 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 07:34:46.246080 139204217857856 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 07:34:46.246117 139204217857856 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 07:34:46.246166 139204217857856 maxtext_utils.py:1707] params/params/decoder/logits_dense/kernel Shape: float32[2048,32000] Logical: P('embed_vocab', 'vocab') Physical: ('fsdp', None) I0425 07:34:46.246221 139204217857856 maxtext_utils.py:1707] params/params/token_embedder/embedding Shape: float32[32000,2048] Logical: P('vocab', 'embed_vocab') Physical: (None, 'fsdp') I0425 07:34:46.741486 139204217857856 train.py:155] train/xent Logical: float32[32,2048]............................................ ('activation_embed_and_logits_batch', 'activation_length'). I0425 07:34:46.741580 139204217857856 train.py:155] train/xent Physical: float32[32,2048]............................................ ('fsdp', None). I0425 07:34:46.757268 139204217857856 train.py:162] train/z_loss Logical: float32[32,2048]............................................ ('activation_embed_and_logits_batch', 'activation_length'). I0425 07:34:46.757326 139204217857856 train.py:162] train/z_loss Physical: float32[32,2048]............................................ ('fsdp', None). I0425 07:34:57.607289 139204217857856 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 07:34:57.608129 139204217857856 metric_logger.py:301] number parameters: 1.104 billion I0425 07:35:09.193685 139204217857856 checkpointing.py:772] Waiting for step 0 to finish before checkpoint... I0425 07:35:09.347340 139204217857856 checkpointing.py:776] Waited 0.15365052223205566 seconds for step 0 to finish before starting checkpointing. I0425 07:35:09.349579 139204217857856 checkpoint_manager.py:2009] [process=5][thread=MainThread][wait_until_finished] No Save Finalize thread to wait for. Returning. I0425 07:35:09.351580 139204217857856 checkpoint_manager.py:1512] [process=5] Saving checkpoint at step 0 I0425 07:35:09.353010 139204217857856 event_tracking.py:70] [process=5] [async] Started save checkpoint @ gs://lance-maxtext/linen_ckpt_xpk_main_20260425_071456/linen_xpk_main_20260425_071456_08_checkpoint_async_true/checkpoints/0. I0425 07:35:09.665458 139204217857856 signaling_client.py:364] Using JaxDistributedSignalingClient I0425 07:35:09.666443 139204217857856 jax_array_handlers.py:360] Scheduling D2H of 39 prioritized jax.Array. I0425 07:35:09.666502 139204217857856 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 07:35:09.940307 139204217857856 base_pytree_checkpoint_handler.py:154] [process=5][thread=MainThread] Initiated "orbax.checkpoint._src.serialization.jax_array_handlers.ArrayHandler".serialize. Time taken: 0.274924s I0425 07:35:09.940485 139204217857856 base_pytree_checkpoint_handler.py:130] [process=5] /jax/orbax/write/blocking_gbytes_per_sec: 5.500 GiB/s (total gbytes: 1.5 GiB) (time elapsed: 0.28044867515563965 s) (per-host) I0425 07:35:09.940538 139204217857856 base_pytree_checkpoint_handler.py:768] [process=5][thread=MainThread] Initiated Pytree async_save. Time taken: 0.280512s (batch_requests_ready=0.002240s, total_serialization_initiated=0.278200s, others=0.000071s) I0425 07:35:09.940629 139204217857856 composite_checkpoint_handler.py:715] [process=5][thread=MainThread] Initiated CompositeCheckpointHandler.async_save. Time taken: 0.284742s (all_items=0.000018s, per_item={'items': '0.00001788'}, temp_paths=0.284724) I0425 07:35:09.941461 139204217857856 event_tracking.py:125] [process=5] [async] Finished blocking save in 0.59 seconds. Continuing save @ gs://lance-maxtext/linen_ckpt_xpk_main_20260425_071456/linen_xpk_main_20260425_071456_08_checkpoint_async_true/checkpoints/0. I0425 07:35:09.941767 139075212355328 async_checkpointer.py:76] [process=5][thread=async_save] Background save thread started. Deadline for this save operation is 2026-04-25 07:55:09.941733 I0425 07:35:09.953262 139204217857856 checkpoint_manager.py:1560] [process=5][thread=MainThread][step=0] Starting CheckpointManager Save Finalize thread=save_finalize I0425 07:35:09.953574 139070386308864 async_checkpointer.py:280] [process=5][thread=save_finalize] Waiting for background save thread=async_save. I0425 07:35:09.953723 139204217857856 standard_logger.py:34] {'step': 0, 'event_type': 'save', 'directory': 'gs://lance-maxtext/linen_ckpt_xpk_main_20260425_071456/linen_xpk_main_20260425_071456_08_checkpoint_async_true/checkpoints', 'reached_preemption': False, 'preemption_received_at': None, 'synchronous': False, 'wait_for_prev_start_time': 1777102509.3495603, 'wait_for_prev_duration_secs': 6.151199340820312e-05, 'time_between_consecutive_saves_sec': None, 'checkpointer_blocking_start_time': 1777102509.3516188, 'checkpointer_blocking_duration_secs': 0.5902988910675049, 'get_old_steps_start_time': 1777102509.9419436, 'get_old_steps_duration_secs': 3.147125244140625e-05, 'checkpoint_manager_blocking_start_time': 1777102509.3478587, 'checkpoint_manager_blocking_duration_secs': 0.605825662612915} I0425 07:35:09.953883 139204217857856 checkpointing.py:408] Started an asynchronous checkpoint save for step 0 I0425 07:35:09.953937 139204217857856 max_utils.py:750] Memstats: After params initialized: I0425 07:35:09.953989 139204217857856 max_utils.py:756] Using (GB) 0.43 / 31.25 (1.376000%) on TPU_18(process=5,(2,4,0,0)) I0425 07:35:09.954023 139204217857856 max_utils.py:756] Using (GB) 0.43 / 31.25 (1.376000%) on TPU_19(process=5,(3,4,0,0)) I0425 07:35:09.954051 139204217857856 max_utils.py:756] Using (GB) 0.43 / 31.25 (1.376000%) on TPU_22(process=5,(2,5,0,0)) I0425 07:35:09.954078 139204217857856 max_utils.py:756] Using (GB) 0.43 / 31.25 (1.376000%) on TPU_23(process=5,(3,5,0,0)) I0425 07:35:10.274363 139204217857856 metric_logger.py:196] completed step: 0, seconds: 11.585, TFLOP/s/device: 1.173, Tokens/s/device: 176.774, total_weights: 65536, loss: 10.874, lm_loss: 10.874, perplexity: 52779.875 I0425 07:35:10.521266 139204217857856 metric_logger.py:196] completed step: 1, seconds: 1.079, TFLOP/s/device: 12.591, Tokens/s/device: 1897.848, total_weights: 65536, loss: 10.874, lm_loss: 10.874, perplexity: 52779.875 I0425 07:35:10.934001 139204217857856 metric_logger.py:196] completed step: 2, seconds: 0.100, TFLOP/s/device: 136.260, Tokens/s/device: 20538.535, total_weights: 65536, loss: 10.560, lm_loss: 10.560, perplexity: 38547.719 I0425 07:35:11.081317 139204217857856 metric_logger.py:196] completed step: 3, seconds: 0.413, TFLOP/s/device: 32.886, Tokens/s/device: 4956.893, total_weights: 65536, loss: 9.980, lm_loss: 9.980, perplexity: 21590.506 I0425 07:35:11.085866 139204217857856 checkpointing.py:772] Waiting for step 5 to finish before checkpoint... I0425 07:35:11.375416 139204217857856 checkpointing.py:776] Waited 0.2895162105560303 seconds for step 5 to finish before starting checkpointing. I0425 07:35:11.378314 139204217857856 checkpoint_manager.py:2020] [process=5][thread=MainThread][step=0][wait_until_finished] Waiting for Save Finalize thread (save_finalize) to complete. I0425 07:35:13.043615 2864 google_auth_provider.cc:181] Running on GCE, using service account 562977990677-compute@developer.gserviceaccount.com I0425 07:35:15.543012 139070929733376 array_metadata_store.py:203] [process=5][thread=array_type_handler] Wrote 39 array_metadata.ArrayMetadata to gs://lance-maxtext/linen_ckpt_xpk_main_20260425_071456/linen_xpk_main_20260425_071456_08_checkpoint_async_true/checkpoints/0/items/array_metadatas/process_5 I0425 07:35:46.620497 139075212355328 base_pytree_checkpoint_handler.py:130] [process=5] /jax/orbax/write/gbytes_per_sec: 42.738 MiB/s (total gbytes: 1.5 GiB) (time elapsed: 36.96042084693909 s) (per-host) I0425 07:35:46.620627 139075212355328 async_checkpointer.py:90] [process=5][thread=async_save] 3 Handler Commit operations completed. Time taken: 36.678748s. I0425 07:35:55.186698 139075212355328 async_checkpointer.py:160] [process=5][thread=async_save] Background save thread done. Time taken: 45.244771s. I0425 07:35:55.186964 139070386308864 async_checkpointer.py:288] [process=5][thread=save_finalize] Done with waiting for background save thread=async_save. I0425 07:35:55.187085 139070386308864 async_checkpointer.py:298] [process=5][thread=save_finalize] No errors found in background save thread=async_save. I0425 07:35:55.187138 139070386308864 checkpoint_manager.py:2137] [process=5][thread=save_finalize][step=0] CheckpointManager Save Finalize is syncing with other hosts... I0425 07:35:55.188789 139070386308864 checkpoint_manager.py:2146] [process=5][thread=save_finalize][step=0] CheckpointManager Save Finalize is done on all hosts. I0425 07:35:55.188978 139204217857856 checkpoint_manager.py:2032] [process=5][thread=MainThread][step=0][wait_until_finished] Done waiting for Save Finalize thread (save_finalize) running at step=0. W0425 07:35:55.189111 139204217857856 checkpoint_manager.py:1452] Waiting for previous save to complete took 43.810795 seconds. If this number is high, consider checkpointing less frequently. I0425 07:35:55.190945 139204217857856 checkpoint_manager.py:1512] [process=5] Saving checkpoint at step 5 I0425 07:35:55.193447 139204217857856 event_tracking.py:70] [process=5] [async] Started save checkpoint @ gs://lance-maxtext/linen_ckpt_xpk_main_20260425_071456/linen_xpk_main_20260425_071456_08_checkpoint_async_true/checkpoints/5. I0425 07:35:55.481607 139204217857856 jax_array_handlers.py:360] Scheduling D2H of 39 prioritized jax.Array. I0425 07:35:55.481713 139204217857856 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 07:35:55.516243 139204217857856 base_pytree_checkpoint_handler.py:154] [process=5][thread=MainThread] Initiated "orbax.checkpoint._src.serialization.jax_array_handlers.ArrayHandler".serialize. Time taken: 0.035793s I0425 07:35:55.516419 139204217857856 base_pytree_checkpoint_handler.py:130] [process=5] /jax/orbax/write/blocking_gbytes_per_sec: 39.282 GiB/s (total gbytes: 1.5 GiB) (time elapsed: 0.039269208908081055 s) (per-host) I0425 07:35:55.516472 139204217857856 base_pytree_checkpoint_handler.py:768] [process=5][thread=MainThread] Initiated Pytree async_save. Time taken: 0.039331s (batch_requests_ready=0.001776s, total_serialization_initiated=0.037483s, others=0.000072s) I0425 07:35:55.516564 139204217857856 composite_checkpoint_handler.py:715] [process=5][thread=MainThread] Initiated CompositeCheckpointHandler.async_save. Time taken: 0.044365s (all_items=0.000015s, per_item={'items': '0.00001526'}, temp_paths=0.044350) I0425 07:35:55.518621 139204217857856 event_tracking.py:125] [process=5] [async] Finished blocking save in 0.33 seconds. Continuing save @ gs://lance-maxtext/linen_ckpt_xpk_main_20260425_071456/linen_xpk_main_20260425_071456_08_checkpoint_async_true/checkpoints/5. I0425 07:35:55.518929 139070386308864 async_checkpointer.py:76] [process=5][thread=async_save] Background save thread started. Deadline for this save operation is 2026-04-25 07:55:55.518896 I0425 07:35:55.530670 139204217857856 checkpoint_manager.py:1560] [process=5][thread=MainThread][step=5] Starting CheckpointManager Save Finalize thread=save_finalize I0425 07:35:55.530980 139070929733376 async_checkpointer.py:280] [process=5][thread=save_finalize] Waiting for background save thread=async_save. I0425 07:35:55.531141 139204217857856 standard_logger.py:34] {'step': 5, 'event_type': 'save', 'directory': 'gs://lance-maxtext/linen_ckpt_xpk_main_20260425_071456/linen_xpk_main_20260425_071456_08_checkpoint_async_true/checkpoints', 'reached_preemption': False, 'preemption_received_at': None, 'synchronous': False, 'wait_for_prev_start_time': 1777102511.3782854, 'wait_for_prev_duration_secs': 43.81079459190369, 'time_between_consecutive_saves_sec': None, 'checkpointer_blocking_start_time': 1777102555.190984, 'checkpointer_blocking_duration_secs': 0.3281073570251465, 'get_old_steps_start_time': 1777102555.519115, 'get_old_steps_duration_secs': 3.4332275390625e-05, 'checkpoint_manager_blocking_start_time': 1777102511.3760743, 'checkpoint_manager_blocking_duration_secs': 44.15503144264221} I0425 07:35:55.531259 139204217857856 checkpointing.py:408] Started an asynchronous checkpoint save for step 5 I0425 07:35:55.532133 139204217857856 metric_logger.py:196] completed step: 4, seconds: 0.153, TFLOP/s/device: 88.881, Tokens/s/device: 13397.179, total_weights: 65536, loss: 9.460, lm_loss: 9.460, perplexity: 12840.355 I0425 07:35:55.538544 139204217857856 metric_logger.py:196] completed step: 5, seconds: 0.147, TFLOP/s/device: 92.285, Tokens/s/device: 13910.209, total_weights: 65536, loss: 8.959, lm_loss: 8.959, perplexity: 7776.467 I0425 07:36:00.757247 139053219436288 array_metadata_store.py:203] [process=5][thread=array_type_handler] Wrote 39 array_metadata.ArrayMetadata to gs://lance-maxtext/linen_ckpt_xpk_main_20260425_071456/linen_xpk_main_20260425_071456_08_checkpoint_async_true/checkpoints/5/items/array_metadatas/process_5 I0425 07:36:18.054320 139204217857856 metric_logger.py:196] completed step: 6, seconds: 44.451, TFLOP/s/device: 0.306, Tokens/s/device: 46.073, total_weights: 65536, loss: 8.469, lm_loss: 8.469, perplexity: 4762.740 I0425 07:36:18.201704 139204217857856 metric_logger.py:196] completed step: 7, seconds: 22.370, TFLOP/s/device: 0.607, Tokens/s/device: 91.553, total_weights: 65536, loss: 8.003, lm_loss: 8.003, perplexity: 2989.151 I0425 07:36:18.349203 139204217857856 metric_logger.py:196] completed step: 8, seconds: 0.152, TFLOP/s/device: 89.164, Tokens/s/device: 13439.819, total_weights: 65536, loss: 7.572, lm_loss: 7.572, perplexity: 1943.098 I0425 07:36:18.495793 139204217857856 checkpointing.py:772] Waiting for step 9 to finish before checkpoint... I0425 07:36:18.496416 139204217857856 checkpointing.py:776] Waited 0.0006437301635742188 seconds for step 9 to finish before starting checkpointing. I0425 07:36:18.498522 139204217857856 checkpoint_manager.py:2020] [process=5][thread=MainThread][step=5][wait_until_finished] Waiting for Save Finalize thread (save_finalize) to complete. I0425 07:36:36.639727 139070386308864 base_pytree_checkpoint_handler.py:130] [process=5] /jax/orbax/write/gbytes_per_sec: 38.375 MiB/s (total gbytes: 1.5 GiB) (time elapsed: 41.162537574768066 s) (per-host) I0425 07:36:36.639852 139070386308864 async_checkpointer.py:90] [process=5][thread=async_save] 3 Handler Commit operations completed. Time taken: 41.120805s. I0425 07:36:44.671406 139070386308864 async_checkpointer.py:160] [process=5][thread=async_save] Background save thread done. Time taken: 49.152340s. I0425 07:36:44.671729 139070929733376 async_checkpointer.py:288] [process=5][thread=save_finalize] Done with waiting for background save thread=async_save. I0425 07:36:44.671846 139070929733376 async_checkpointer.py:298] [process=5][thread=save_finalize] No errors found in background save thread=async_save. I0425 07:36:44.671896 139070929733376 checkpoint_manager.py:2137] [process=5][thread=save_finalize][step=5] CheckpointManager Save Finalize is syncing with other hosts... I0425 07:36:44.673525 139070929733376 checkpoint_manager.py:2146] [process=5][thread=save_finalize][step=5] CheckpointManager Save Finalize is done on all hosts. I0425 07:36:44.673718 139204217857856 checkpoint_manager.py:2032] [process=5][thread=MainThread][step=5][wait_until_finished] Done waiting for Save Finalize thread (save_finalize) running at step=5. W0425 07:36:44.673846 139204217857856 checkpoint_manager.py:1452] Waiting for previous save to complete took 26.175329 seconds. If this number is high, consider checkpointing less frequently. I0425 07:36:44.675390 139204217857856 checkpoint_manager.py:1512] [process=5] Saving checkpoint at step 9 I0425 07:36:44.677469 139204217857856 event_tracking.py:70] [process=5] [async] Started save checkpoint @ gs://lance-maxtext/linen_ckpt_xpk_main_20260425_071456/linen_xpk_main_20260425_071456_08_checkpoint_async_true/checkpoints/9. I0425 07:36:44.962568 139204217857856 jax_array_handlers.py:360] Scheduling D2H of 39 prioritized jax.Array. I0425 07:36:44.962671 139204217857856 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 07:36:44.999148 139204217857856 base_pytree_checkpoint_handler.py:154] [process=5][thread=MainThread] Initiated "orbax.checkpoint._src.serialization.jax_array_handlers.ArrayHandler".serialize. Time taken: 0.037389s I0425 07:36:44.999301 139204217857856 base_pytree_checkpoint_handler.py:130] [process=5] /jax/orbax/write/blocking_gbytes_per_sec: 37.767 GiB/s (total gbytes: 1.5 GiB) (time elapsed: 0.04084467887878418 s) (per-host) I0425 07:36:44.999366 139204217857856 base_pytree_checkpoint_handler.py:768] [process=5][thread=MainThread] Initiated Pytree async_save. Time taken: 0.040921s (batch_requests_ready=0.001649s, total_serialization_initiated=0.039190s, others=0.000082s) I0425 07:36:44.999471 139204217857856 composite_checkpoint_handler.py:715] [process=5][thread=MainThread] Initiated CompositeCheckpointHandler.async_save. Time taken: 0.045449s (all_items=0.000010s, per_item={'items': '0.00001025'}, temp_paths=0.045439) I0425 07:36:45.000184 139204217857856 event_tracking.py:125] [process=5] [async] Finished blocking save in 0.32 seconds. Continuing save @ gs://lance-maxtext/linen_ckpt_xpk_main_20260425_071456/linen_xpk_main_20260425_071456_08_checkpoint_async_true/checkpoints/9. I0425 07:36:45.000465 139070929733376 async_checkpointer.py:76] [process=5][thread=async_save] Background save thread started. Deadline for this save operation is 2026-04-25 07:56:45.000431 I0425 07:36:45.009576 139204217857856 checkpoint_manager.py:1560] [process=5][thread=MainThread][step=9] Starting CheckpointManager Save Finalize thread=save_finalize I0425 07:36:45.009883 139053219436288 async_checkpointer.py:280] [process=5][thread=save_finalize] Waiting for background save thread=async_save. I0425 07:36:45.010050 139204217857856 standard_logger.py:34] {'step': 9, 'event_type': 'save', 'directory': 'gs://lance-maxtext/linen_ckpt_xpk_main_20260425_071456/linen_xpk_main_20260425_071456_08_checkpoint_async_true/checkpoints', 'reached_preemption': False, 'preemption_received_at': None, 'synchronous': False, 'wait_for_prev_start_time': 1777102578.4984932, 'wait_for_prev_duration_secs': 26.175329208374023, 'time_between_consecutive_saves_sec': 23.30966353416443, 'checkpointer_blocking_start_time': 1777102604.6754274, 'checkpointer_blocking_duration_secs': 0.3251767158508301, 'get_old_steps_start_time': 1777102605.0006208, 'get_old_steps_duration_secs': 2.5272369384765625e-05, 'checkpoint_manager_blocking_start_time': 1777102578.4966462, 'checkpoint_manager_blocking_duration_secs': 26.513370752334595} I0425 07:36:45.010163 139204217857856 checkpointing.py:408] Started an asynchronous checkpoint save for step 9 I0425 07:36:45.010206 139204217857856 checkpoint_manager.py:2020] [process=5][thread=MainThread][step=9][wait_until_finished] Waiting for Save Finalize thread (save_finalize) to complete. I0425 07:36:50.061917 139053122975488 array_metadata_store.py:203] [process=5][thread=array_type_handler] Wrote 39 array_metadata.ArrayMetadata to gs://lance-maxtext/linen_ckpt_xpk_main_20260425_071456/linen_xpk_main_20260425_071456_08_checkpoint_async_true/checkpoints/9/items/array_metadatas/process_5 I0425 07:37:26.909519 139070929733376 base_pytree_checkpoint_handler.py:130] [process=5] /jax/orbax/write/gbytes_per_sec: 37.653 MiB/s (total gbytes: 1.5 GiB) (time elapsed: 41.95103311538696 s) (per-host) I0425 07:37:26.909623 139070929733376 async_checkpointer.py:90] [process=5][thread=async_save] 3 Handler Commit operations completed. Time taken: 41.909051s. I0425 07:37:32.388534 139070929733376 async_checkpointer.py:160] [process=5][thread=async_save] Background save thread done. Time taken: 47.387947s. I0425 07:37:32.388808 139053219436288 async_checkpointer.py:288] [process=5][thread=save_finalize] Done with waiting for background save thread=async_save. I0425 07:37:32.388923 139053219436288 async_checkpointer.py:298] [process=5][thread=save_finalize] No errors found in background save thread=async_save. I0425 07:37:32.388970 139053219436288 checkpoint_manager.py:2137] [process=5][thread=save_finalize][step=9] CheckpointManager Save Finalize is syncing with other hosts... I0425 07:37:32.390679 139053219436288 checkpoint_manager.py:2146] [process=5][thread=save_finalize][step=9] CheckpointManager Save Finalize is done on all hosts. I0425 07:37:32.390839 139204217857856 checkpoint_manager.py:2032] [process=5][thread=MainThread][step=9][wait_until_finished] Done waiting for Save Finalize thread (save_finalize) running at step=9. I0425 07:37:32.390989 139204217857856 checkpoint_manager.py:2009] [process=5][thread=MainThread][wait_until_finished] No Save Finalize thread to wait for. Returning. I0425 07:37:32.391906 139204217857856 metric_logger.py:196] completed step: 9, seconds: 0.147, TFLOP/s/device: 92.252, Tokens/s/device: 13905.297, total_weights: 65536, loss: 7.181, lm_loss: 7.181, perplexity: 1314.495 Per train step: Total TFLOPs: 13.59 split as 93.93% learnable weight flops and 6.07% attention flops XPK End: Sat Apr 25 07:37:41 UTC 2026 EXIT_CODE=0