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 | 92.241 | 92.245 | +0.004 |
| Tok/s | 13903.6 | 13904.2 | +0.566 |
| Avg s/step | 2.728 | 2.385 | -0.343 |
| 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 07:46:56 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 07:47:21.267773: 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 07:47:21.477220 135379629762368 max_utils.py:273] Attempting to initialize the jax distributed system... I0425 07:47:30.517835 135379629762368 distributed.py:149] Starting JAX distributed service on [::]:8482 I0425 07:47:30.520167 135379629762368 distributed.py:172] Connecting to JAX distributed service on mt-10-shardy-false-2gfn1-slice-job-0-0.mt-10-shardy-false-2gfn1:8482 I0425 07:47:31.800594 135379629762368 max_utils.py:284] Jax distributed system initialized! I0425 07:47:36.916332 135379629762368 max_utils.py:800] System Information: Jax Version: 0.9.2 I0425 07:47:36.916436 135379629762368 max_utils.py:801] System Information: Jaxlib Version: 0.9.2 I0425 07:47:36.916477 135379629762368 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 07:47:36.916514 135379629762368 train_utils.py:361] WARNING: Sequence packing is essentially ignored for synthetic data. Please use a real dataset to use sequence packing. I0425 07:47:37.614609 135379629762368 maxtext_utils.py:1604] Num_devices: 32, shape (1, 1, 1, 32, 1, 1, 1, 1, 1, 1, 1, 1, 1) I0425 07:47:37.614897 135379629762368 checkpointing.py:677] Setting up checkpoint logger... I0425 07:47:37.614953 135379629762368 checkpointing.py:233] Creating checkpoint manager with ocdbt=True and zarr3=True I0425 07:47:37.614998 135379629762368 pytree_checkpoint_handler.py:592] save_device_host_concurrent_bytes=None I0425 07:47:37.615333 135379629762368 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 0x7b1fe3b14140>, enable_pinned_host_transfer=False, save_concurrent_bytes: 96000000000 (89.4 GiB), restore_concurrent_bytes: 96000000000 (89.4 GiB) I0425 07:47:40.538362 135379629762368 checkpointing.py:265] Enabling policy for fixed interval checkpointing. I0425 07:47:40.538655 135379629762368 checkpoint_manager.py:708] [process=4][thread=MainThread] CheckpointManager init: checkpointers=None, item_names=('items',), item_handlers={'items': <orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeCheckpointHandler object at 0x7b0c383d98b0>}, handler_registry=None I0425 07:47:40.538909 135379629762368 composite_checkpoint_handler.py:237] Deferred registration for item: "items". Adding handler `<orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeCheckpointHandler object at 0x7b0c383d98b0>` 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 07:47:40.538960 135379629762368 composite_checkpoint_handler.py:237] Deferred registration for item: "metrics". Adding handler `<orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonCheckpointHandler object at 0x7b0c383dae70>` 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 07:47:40.538996 135379629762368 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 0x7b0c383d98b0>, ('items', <class 'orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeRestoreArgs'>): <orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeCheckpointHandler object at 0x7b0c383d98b0>, ('metrics', <class 'orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonSaveArgs'>): <orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonCheckpointHandler object at 0x7b0c383dae70>, ('metrics', <class 'orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonRestoreArgs'>): <orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonCheckpointHandler object at 0x7b0c383dae70>}). I0425 07:47:40.539324 135379629762368 abstract_checkpointer.py:35] orbax-checkpoint version: 0.11.34 I0425 07:47:40.539394 135379629762368 async_checkpointer.py:192] [process=4][thread=MainThread] Using barrier_sync_fn: <function get_barrier_sync_fn.<locals>._fn at 0x7b0b2874d800> timeout: 1200 secs and primary_host=0 for async checkpoint writes I0425 07:47:41.226478 135379629762368 checkpoint_manager.py:1812] Found 0 checkpoint steps in gs://lance-maxtext/linen_ckpt_xpk_main_20260425_071456/linen_xpk_main_20260425_071456_10_shardy_false/checkpoints I0425 07:47:41.370240 135379629762368 checkpoint_manager.py:929] [process=4][thread=MainThread] CheckpointManager created, primary_host=0, CheckpointManagerOptions=CheckpointManagerOptions(save_interval_steps=1, max_to_keep=None, keep_time_interval=None, keep_period=None, should_keep_fn=None, best_fn=None, best_mode='max', keep_checkpoints_without_metrics=True, step_prefix=None, step_format_fixed_length=None, step_name_format=None, create=True, cleanup_tmp_directories=False, save_on_steps=frozenset(), single_host_load_and_broadcast=False, todelete_subdir=None, todelete_full_path=None, enable_background_delete=False, read_only=False, enable_async_checkpointing=True, async_options=None, multiprocessing_options=MultiprocessingOptions(primary_host=0, active_processes=None, barrier_sync_key_prefix=None), should_save_fn=None, file_options=FileOptions(path_permission_mode=None), save_root_metadata=True, temporary_path_class=None, save_decision_policy=FixedIntervalPolicy(interval=10), preservation_policy=LatestN(n=None), prevent_write_metrics=False, enable_should_save_is_saving_in_progress_check=True, enable_per_process_directory_creation=False, lightweight_initialize=False), root_directory=gs://lance-maxtext/linen_ckpt_xpk_main_20260425_071456/linen_xpk_main_20260425_071456_10_shardy_false/checkpoints: <orbax.checkpoint.checkpoint_manager.CheckpointManager object at 0x7b0c385fe4b0> I0425 07:47:41.370409 135379629762368 checkpointing.py:301] Checkpoint manager created! I0425 07:47:42.338371 135379629762368 nnx_wrappers.py:437] Unknown Logical: bfloat16[32,2048,2048]...................................... ('activation_batch', 'activation_norm_length', 'activation_embed'). I0425 07:47:42.338489 135379629762368 nnx_wrappers.py:437] Unknown Physical: bfloat16[32,2048,2048]...................................... ('fsdp', None, None). I0425 07:47:42.717695 135379629762368 attentions.py:1088] attentions/inputs_q Logical: bfloat16[32,2048,2048]...................................... ('activation_batch', 'activation_attn_length', 'activation_attn_embed'). I0425 07:47:42.717784 135379629762368 attentions.py:1088] attentions/inputs_q Physical: bfloat16[32,2048,2048]...................................... ('fsdp', None, None). I0425 07:47:42.734222 135379629762368 attentions.py:1089] attentions/inputs_kv Logical: bfloat16[32,2048,2048]...................................... ('activation_batch', 'activation_attn_length', 'activation_attn_embed'). I0425 07:47:42.734280 135379629762368 attentions.py:1089] attentions/inputs_kv Physical: bfloat16[32,2048,2048]...................................... ('fsdp', None, None). I0425 07:47:42.758165 135379629762368 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 07:47:42.758234 135379629762368 attentions.py:1154] attentions/query Physical: bfloat16[32,2048,16,128].................................... ('fsdp', None, None, None). I0425 07:47:42.774929 135379629762368 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 07:47:42.774989 135379629762368 attentions.py:1155] attentions/key Physical: bfloat16[32,2048,16,128].................................... ('fsdp', None, None, None). I0425 07:47:42.791528 135379629762368 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 07:47:42.791587 135379629762368 attentions.py:1156] attentions/value Physical: bfloat16[32,2048,16,128].................................... ('fsdp', None, None, None). I0425 07:47:42.816250 135379629762368 attentions.py:1198] attentions/out Logical: bfloat16[32,2048,16,128].................................... ('activation_batch', 'activation_attn_length', 'activation_heads', 'activation_kv'). I0425 07:47:42.816318 135379629762368 attentions.py:1198] attentions/out Physical: bfloat16[32,2048,16,128].................................... ('fsdp', None, None, None). I0425 07:47:42.837651 135379629762368 linears.py:525] linears/x Logical: bfloat16[32,2048,7168]...................................... ('activation_batch', 'activation_length', 'activation_mlp'). I0425 07:47:42.837715 135379629762368 linears.py:525] linears/x Physical: bfloat16[32,2048,7168]...................................... ('fsdp', None, None). I0425 07:47:43.047374 135379629762368 checkpointing.py:577] checkpoint manager exists so trying to load this run's existing checkpoint I0425 07:47:43.047480 135379629762368 checkpointing.py:665] No existing checkpoints found, not restoring checkpoint. fsdp: 32 I0425 07:47:44.432560 135379629762368 maxtext_utils.py:1707] params/params/decoder/decoder_norm/scale Shape: float32[2048] Logical: P('norm',) Physical: (None,) I0425 07:47:44.432682 135379629762368 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 07:47:44.432734 135379629762368 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 07:47:44.432790 135379629762368 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 07:47:44.432841 135379629762368 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 07:47:44.432900 135379629762368 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 07:47:44.432958 135379629762368 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 07:47:44.433010 135379629762368 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 07:47:44.433049 135379629762368 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 07:47:44.433084 135379629762368 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 07:47:44.433130 135379629762368 maxtext_utils.py:1707] params/params/decoder/logits_dense/kernel Shape: float32[2048,32000] Logical: P('embed_vocab', 'vocab') Physical: ('fsdp', None) I0425 07:47:44.433177 135379629762368 maxtext_utils.py:1707] params/params/token_embedder/embedding Shape: float32[32000,2048] Logical: P('vocab', 'embed_vocab') Physical: (None, 'fsdp') I0425 07:47:44.921638 135379629762368 train.py:155] train/xent Logical: float32[32,2048]............................................ ('activation_embed_and_logits_batch', 'activation_length'). I0425 07:47:44.921730 135379629762368 train.py:155] train/xent Physical: float32[32,2048]............................................ ('fsdp', None). I0425 07:47:44.937569 135379629762368 train.py:162] train/z_loss Logical: float32[32,2048]............................................ ('activation_embed_and_logits_batch', 'activation_length'). I0425 07:47:44.937635 135379629762368 train.py:162] train/z_loss Physical: float32[32,2048]............................................ ('fsdp', None). I0425 07:47:55.659672 135379629762368 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 07:47:55.660467 135379629762368 metric_logger.py:301] number parameters: 1.104 billion I0425 07:48:06.617445 135379629762368 checkpointing.py:772] Waiting for step 0 to finish before checkpoint... I0425 07:48:07.480184 135379629762368 checkpointing.py:776] Waited 0.862717866897583 seconds for step 0 to finish before starting checkpointing. I0425 07:48:07.484290 135379629762368 checkpoint_manager.py:2009] [process=4][thread=MainThread][wait_until_finished] No Save Finalize thread to wait for. Returning. I0425 07:48:07.485629 135379629762368 checkpoint_manager.py:1512] [process=4] Saving checkpoint at step 0 I0425 07:48:07.487043 135379629762368 event_tracking.py:70] [process=4] [async] Started save checkpoint @ gs://lance-maxtext/linen_ckpt_xpk_main_20260425_071456/linen_xpk_main_20260425_071456_10_shardy_false/checkpoints/0. I0425 07:48:07.768981 135379629762368 signaling_client.py:364] Using JaxDistributedSignalingClient I0425 07:48:07.769939 135379629762368 jax_array_handlers.py:360] Scheduling D2H of 39 prioritized jax.Array. I0425 07:48:07.769996 135379629762368 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 07:48:08.056205 135379629762368 base_pytree_checkpoint_handler.py:154] [process=4][thread=MainThread] Initiated "orbax.checkpoint._src.serialization.jax_array_handlers.ArrayHandler".serialize. Time taken: 0.287296s I0425 07:48:08.056377 135379629762368 base_pytree_checkpoint_handler.py:130] [process=4] /jax/orbax/write/blocking_gbytes_per_sec: 5.268 GiB/s (total gbytes: 1.5 GiB) (time elapsed: 0.2927978038787842 s) (per-host) I0425 07:48:08.056432 135379629762368 base_pytree_checkpoint_handler.py:768] [process=4][thread=MainThread] Initiated Pytree async_save. Time taken: 0.292864s (batch_requests_ready=0.002303s, total_serialization_initiated=0.290486s, others=0.000074s) I0425 07:48:08.056539 135379629762368 composite_checkpoint_handler.py:715] [process=4][thread=MainThread] Initiated CompositeCheckpointHandler.async_save. Time taken: 0.297461s (all_items=0.000018s, per_item={'items': '0.00001836'}, temp_paths=0.297442) I0425 07:48:08.057380 135379629762368 event_tracking.py:125] [process=4] [async] Finished blocking save in 0.57 seconds. Continuing save @ gs://lance-maxtext/linen_ckpt_xpk_main_20260425_071456/linen_xpk_main_20260425_071456_10_shardy_false/checkpoints/0. I0425 07:48:08.057715 135254193846016 async_checkpointer.py:76] [process=4][thread=async_save] Background save thread started. Deadline for this save operation is 2026-04-25 08:08:08.057676 I0425 07:48:08.061488 135379629762368 checkpoint_manager.py:1560] [process=4][thread=MainThread][step=0] Starting CheckpointManager Save Finalize thread=save_finalize I0425 07:48:08.061729 135252560336640 async_checkpointer.py:280] [process=4][thread=save_finalize] Waiting for background save thread=async_save. I0425 07:48:08.061854 135379629762368 standard_logger.py:34] {'step': 0, 'event_type': 'save', 'directory': 'gs://lance-maxtext/linen_ckpt_xpk_main_20260425_071456/linen_xpk_main_20260425_071456_10_shardy_false/checkpoints', 'reached_preemption': False, 'preemption_received_at': None, 'synchronous': False, 'wait_for_prev_start_time': 1777103287.484271, 'wait_for_prev_duration_secs': 6.222724914550781e-05, 'time_between_consecutive_saves_sec': None, 'checkpointer_blocking_start_time': 1777103287.485666, 'checkpointer_blocking_duration_secs': 0.5722157955169678, 'get_old_steps_start_time': 1777103288.0579066, 'get_old_steps_duration_secs': 2.9802322387695312e-05, 'checkpoint_manager_blocking_start_time': 1777103287.4807022, 'checkpoint_manager_blocking_duration_secs': 0.5811111927032471} I0425 07:48:08.061981 135379629762368 checkpointing.py:408] Started an asynchronous checkpoint save for step 0 I0425 07:48:08.062031 135379629762368 max_utils.py:750] Memstats: After params initialized: I0425 07:48:08.062081 135379629762368 max_utils.py:756] Using (GB) 0.43 / 31.25 (1.376000%) on TPU_16(process=4,(0,4,0,0)) I0425 07:48:08.062113 135379629762368 max_utils.py:756] Using (GB) 0.43 / 31.25 (1.376000%) on TPU_17(process=4,(1,4,0,0)) I0425 07:48:08.062141 135379629762368 max_utils.py:756] Using (GB) 0.43 / 31.25 (1.376000%) on TPU_20(process=4,(0,5,0,0)) I0425 07:48:08.062166 135379629762368 max_utils.py:756] Using (GB) 0.43 / 31.25 (1.376000%) on TPU_21(process=4,(1,5,0,0)) I0425 07:48:08.378576 135379629762368 metric_logger.py:196] completed step: 0, seconds: 10.957, TFLOP/s/device: 1.240, Tokens/s/device: 186.915, total_weights: 65536, loss: 10.874, lm_loss: 10.874, perplexity: 52779.875 I0425 07:48:08.549315 135379629762368 metric_logger.py:196] completed step: 1, seconds: 1.760, TFLOP/s/device: 7.721, Tokens/s/device: 1163.866, total_weights: 65536, loss: 10.874, lm_loss: 10.874, perplexity: 52779.875 I0425 07:48:08.946773 135379629762368 metric_logger.py:196] completed step: 2, seconds: 0.023, TFLOP/s/device: 581.442, Tokens/s/device: 87641.219, total_weights: 65536, loss: 10.262, lm_loss: 10.262, perplexity: 28637.129 I0425 07:48:09.094144 135379629762368 metric_logger.py:196] completed step: 3, seconds: 0.398, TFLOP/s/device: 34.149, Tokens/s/device: 5147.281, total_weights: 65536, loss: 9.734, lm_loss: 9.734, perplexity: 16889.203 I0425 07:48:09.389454 135379629762368 metric_logger.py:196] completed step: 4, seconds: 0.153, TFLOP/s/device: 88.717, Tokens/s/device: 13372.423, total_weights: 65536, loss: 9.277, lm_loss: 9.277, perplexity: 10694.614 I0425 07:48:09.395136 135379629762368 metric_logger.py:196] completed step: 5, seconds: 0.147, TFLOP/s/device: 92.474, Tokens/s/device: 13938.705, total_weights: 65536, loss: 8.892, lm_loss: 8.892, perplexity: 7270.717 I0425 07:48:16.485436 2747 google_auth_provider.cc:181] Running on GCE, using service account 562977990677-compute@developer.gserviceaccount.com I0425 07:48:18.389842 135252568729344 array_metadata_store.py:203] [process=4][thread=array_type_handler] Wrote 39 array_metadata.ArrayMetadata to gs://lance-maxtext/linen_ckpt_xpk_main_20260425_071456/linen_xpk_main_20260425_071456_10_shardy_false/checkpoints/0/items/array_metadatas/process_4 I0425 07:48:31.017835 135379629762368 metric_logger.py:196] completed step: 6, seconds: 0.296, TFLOP/s/device: 45.946, Tokens/s/device: 6925.517, total_weights: 65536, loss: 8.592, lm_loss: 8.592, perplexity: 5390.759 I0425 07:48:31.165260 135379629762368 metric_logger.py:196] completed step: 7, seconds: 21.476, TFLOP/s/device: 0.633, Tokens/s/device: 95.364, total_weights: 65536, loss: 8.384, lm_loss: 8.384, perplexity: 4376.905 I0425 07:48:31.312661 135379629762368 metric_logger.py:196] completed step: 8, seconds: 0.152, TFLOP/s/device: 89.118, Tokens/s/device: 13432.855, total_weights: 65536, loss: 8.255, lm_loss: 8.255, perplexity: 3846.406 I0425 07:48:31.459189 135379629762368 checkpointing.py:772] Waiting for step 9 to finish before checkpoint... I0425 07:48:31.459842 135379629762368 checkpointing.py:776] Waited 0.0006737709045410156 seconds for step 9 to finish before starting checkpointing. I0425 07:48:31.461975 135379629762368 checkpoint_manager.py:2020] [process=4][thread=MainThread][step=0][wait_until_finished] Waiting for Save Finalize thread (save_finalize) to complete. I0425 07:48:49.580753 135254193846016 base_pytree_checkpoint_handler.py:130] [process=4] /jax/orbax/write/gbytes_per_sec: 37.774 MiB/s (total gbytes: 1.5 GiB) (time elapsed: 41.81713604927063 s) (per-host) I0425 07:48:49.580896 135254193846016 async_checkpointer.py:90] [process=4][thread=async_save] 3 Handler Commit operations completed. Time taken: 41.523067s. I0425 07:48:58.644487 135254193846016 async_checkpointer.py:160] [process=4][thread=async_save] Background save thread done. Time taken: 50.586642s. I0425 07:48:58.644797 135252560336640 async_checkpointer.py:288] [process=4][thread=save_finalize] Done with waiting for background save thread=async_save. I0425 07:48:58.644946 135252560336640 async_checkpointer.py:298] [process=4][thread=save_finalize] No errors found in background save thread=async_save. I0425 07:48:58.645012 135252560336640 checkpoint_manager.py:2137] [process=4][thread=save_finalize][step=0] CheckpointManager Save Finalize is syncing with other hosts... I0425 07:48:58.646472 135252560336640 checkpoint_manager.py:2146] [process=4][thread=save_finalize][step=0] CheckpointManager Save Finalize is done on all hosts. I0425 07:48:58.646631 135379629762368 checkpoint_manager.py:2032] [process=4][thread=MainThread][step=0][wait_until_finished] Done waiting for Save Finalize thread (save_finalize) running at step=0. W0425 07:48:58.646757 135379629762368 checkpoint_manager.py:1452] Waiting for previous save to complete took 27.184784 seconds. If this number is high, consider checkpointing less frequently. I0425 07:48:58.648523 135379629762368 checkpoint_manager.py:1512] [process=4] Saving checkpoint at step 9 I0425 07:48:58.650680 135379629762368 event_tracking.py:70] [process=4] [async] Started save checkpoint @ gs://lance-maxtext/linen_ckpt_xpk_main_20260425_071456/linen_xpk_main_20260425_071456_10_shardy_false/checkpoints/9. I0425 07:48:58.934998 135379629762368 jax_array_handlers.py:360] Scheduling D2H of 39 prioritized jax.Array. I0425 07:48:58.935089 135379629762368 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 07:48:58.970600 135379629762368 base_pytree_checkpoint_handler.py:154] [process=4][thread=MainThread] Initiated "orbax.checkpoint._src.serialization.jax_array_handlers.ArrayHandler".serialize. Time taken: 0.036761s I0425 07:48:58.970769 135379629762368 base_pytree_checkpoint_handler.py:130] [process=4] /jax/orbax/write/blocking_gbytes_per_sec: 38.317 GiB/s (total gbytes: 1.5 GiB) (time elapsed: 0.040258169174194336 s) (per-host) I0425 07:48:58.970821 135379629762368 base_pytree_checkpoint_handler.py:768] [process=4][thread=MainThread] Initiated Pytree async_save. Time taken: 0.040318s (batch_requests_ready=0.001873s, total_serialization_initiated=0.038378s, others=0.000068s) I0425 07:48:58.970927 135379629762368 composite_checkpoint_handler.py:715] [process=4][thread=MainThread] Initiated CompositeCheckpointHandler.async_save. Time taken: 0.044387s (all_items=0.000017s, per_item={'items': '0.00001740'}, temp_paths=0.044370) I0425 07:48:58.971591 135379629762368 event_tracking.py:125] [process=4] [async] Finished blocking save in 0.32 seconds. Continuing save @ gs://lance-maxtext/linen_ckpt_xpk_main_20260425_071456/linen_xpk_main_20260425_071456_10_shardy_false/checkpoints/9. I0425 07:48:58.971922 135231575144192 async_checkpointer.py:76] [process=4][thread=async_save] Background save thread started. Deadline for this save operation is 2026-04-25 08:08:58.971886 I0425 07:48:58.973765 135379629762368 checkpoint_manager.py:1560] [process=4][thread=MainThread][step=9] Starting CheckpointManager Save Finalize thread=save_finalize I0425 07:48:58.974073 135252560336640 async_checkpointer.py:280] [process=4][thread=save_finalize] Waiting for background save thread=async_save. I0425 07:48:58.974230 135379629762368 standard_logger.py:34] {'step': 9, 'event_type': 'save', 'directory': 'gs://lance-maxtext/linen_ckpt_xpk_main_20260425_071456/linen_xpk_main_20260425_071456_10_shardy_false/checkpoints', 'reached_preemption': False, 'preemption_received_at': None, 'synchronous': False, 'wait_for_prev_start_time': 1777103311.461944, 'wait_for_prev_duration_secs': 27.184783935546875, 'time_between_consecutive_saves_sec': None, 'checkpointer_blocking_start_time': 1777103338.6485605, 'checkpointer_blocking_duration_secs': 0.32350802421569824, 'get_old_steps_start_time': 1777103338.9720912, 'get_old_steps_duration_secs': 2.956390380859375e-05, 'checkpoint_manager_blocking_start_time': 1777103311.4600816, 'checkpoint_manager_blocking_duration_secs': 27.51411533355713} I0425 07:48:58.974337 135379629762368 checkpointing.py:408] Started an asynchronous checkpoint save for step 9 I0425 07:48:58.974381 135379629762368 checkpoint_manager.py:2020] [process=4][thread=MainThread][step=9][wait_until_finished] Waiting for Save Finalize thread (save_finalize) to complete. I0425 07:49:04.453830 135247063734016 array_metadata_store.py:203] [process=4][thread=array_type_handler] Wrote 39 array_metadata.ArrayMetadata to gs://lance-maxtext/linen_ckpt_xpk_main_20260425_071456/linen_xpk_main_20260425_071456_10_shardy_false/checkpoints/9/items/array_metadatas/process_4 I0425 07:49:40.374290 135231575144192 base_pytree_checkpoint_handler.py:130] [process=4] /jax/orbax/write/gbytes_per_sec: 38.114 MiB/s (total gbytes: 1.5 GiB) (time elapsed: 41.44373369216919 s) (per-host) I0425 07:49:40.374405 135231575144192 async_checkpointer.py:90] [process=4][thread=async_save] 3 Handler Commit operations completed. Time taken: 41.402366s. I0425 07:49:47.829248 135231575144192 async_checkpointer.py:160] [process=4][thread=async_save] Background save thread done. Time taken: 48.857194s. I0425 07:49:47.829539 135252560336640 async_checkpointer.py:288] [process=4][thread=save_finalize] Done with waiting for background save thread=async_save. I0425 07:49:47.829661 135252560336640 async_checkpointer.py:298] [process=4][thread=save_finalize] No errors found in background save thread=async_save. I0425 07:49:47.829712 135252560336640 checkpoint_manager.py:2137] [process=4][thread=save_finalize][step=9] CheckpointManager Save Finalize is syncing with other hosts... I0425 07:49:47.831162 135252560336640 checkpoint_manager.py:2146] [process=4][thread=save_finalize][step=9] CheckpointManager Save Finalize is done on all hosts. I0425 07:49:47.831372 135379629762368 checkpoint_manager.py:2032] [process=4][thread=MainThread][step=9][wait_until_finished] Done waiting for Save Finalize thread (save_finalize) running at step=9. I0425 07:49:47.831524 135379629762368 checkpoint_manager.py:2009] [process=4][thread=MainThread][wait_until_finished] No Save Finalize thread to wait for. Returning. I0425 07:49:47.832726 135379629762368 metric_logger.py:196] completed step: 9, seconds: 0.147, TFLOP/s/device: 92.241, Tokens/s/device: 13903.598, 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 07:49:56 UTC 2026 EXIT_CODE=0
XPK Start: Sat Apr 25 15:34:48 UTC 2026 2026-04-25 15:34:52.826752: 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 00:00:1777131292.839714 10 cuda_dnn.cc:8579] Unable to register cuDNN factory: Attempting to register factory for plugin cuDNN when one has already been registered E0000 00:00:1777131292.843679 10 cuda_blas.cc:1407] Unable to register cuBLAS factory: Attempting to register factory for plugin cuBLAS when one has already been registered W0000 00:00:1777131292.855151 10 computation_placer.cc:177] computation placer already registered. Please check linkage and avoid linking the same target more than once. W0000 00:00:1777131292.855170 10 computation_placer.cc:177] computation placer already registered. Please check linkage and avoid linking the same target more than once. W0000 00:00:1777131292.855173 10 computation_placer.cc:177] computation placer already registered. Please check linkage and avoid linking the same target more than once. W0000 00:00:1777131292.855174 10 computation_placer.cc:177] computation placer already registered. Please check linkage and avoid linking the same target more than once. 2026-04-25 15:35:11.997696: 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 15:35:12.530410 138886580102976 max_utils.py:273] Attempting to initialize the jax distributed system... INFO:2026-04-25 15:35:21,570:jax._src.distributed:140: Starting JAX distributed service on [::]:8482 I0425 15:35:21.570865 138886580102976 distributed.py:140] Starting JAX distributed service on [::]:8482 INFO:2026-04-25 15:35:21,573:jax._src.distributed:157: Connecting to JAX distributed service on mt-10-shardy-false-w1hfk-slice-job-0-0.mt-10-shardy-false-w1hfk:8482 I0425 15:35:21.573333 138886580102976 distributed.py:157] Connecting to JAX distributed service on mt-10-shardy-false-w1hfk-slice-job-0-0.mt-10-shardy-false-w1hfk:8482 I0425 15:35:22.741552 138886580102976 max_utils.py:284] Jax distributed system initialized! I0425 15:35:29.835371 138886580102976 max_utils.py:800] System Information: Jax Version: 0.8.1 I0425 15:35:29.835476 138886580102976 max_utils.py:801] System Information: Jaxlib Version: 0.8.1 I0425 15:35:29.835515 138886580102976 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 15:35:29.835548 138886580102976 train_utils.py:377] WARNING: Sequence packing is essentially ignored for synthetic data. Please use a real dataset to use sequence packing. I0425 15:35:30.461810 138886580102976 maxtext_utils.py:1631] Num_devices: 32, shape (1, 1, 1, 32, 1, 1, 1, 1, 1, 1, 1, 1, 1) I0425 15:35:30.462419 138886580102976 maxtext_utils.py:1631] Num_devices: 32, shape (1, 1, 1, 32, 1, 1, 1, 1, 1, 1, 1, 1, 1) I0425 15:35:30.462600 138886580102976 checkpointing.py:688] Setting up checkpoint logger... I0425 15:35:30.462651 138886580102976 checkpointing.py:234] Creating checkpoint manager with ocdbt=True and zarr3=True I0425 15:35:30.462694 138886580102976 pytree_checkpoint_handler.py:589] save_device_host_concurrent_bytes=None I0425 15:35:30.463039 138886580102976 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 0x7e5095f71310>, enable_pinned_host_transfer=False, save_concurrent_bytes: 96000000000 (89.4 GiB), restore_concurrent_bytes: 96000000000 (89.4 GiB) I0425 15:35:33.326255 138886580102976 checkpointing.py:266] Enabling policy for fixed interval checkpointing. I0425 15:35:33.326610 138886580102976 checkpoint_manager.py:709] [process=6][thread=MainThread] CheckpointManager init: checkpointers=None, item_names=('items',), item_handlers={'items': <orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeCheckpointHandler object at 0x7e3c004a8440>}, handler_registry=None I0425 15:35:33.326844 138886580102976 composite_checkpoint_handler.py:237] Deferred registration for item: "items". Adding handler `<orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeCheckpointHandler object at 0x7e3c004a8440>` 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 15:35:33.326893 138886580102976 composite_checkpoint_handler.py:237] Deferred registration for item: "metrics". Adding handler `<orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonCheckpointHandler object at 0x7e3c002b58b0>` 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 15:35:33.326930 138886580102976 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 0x7e3c004a8440>, ('items', <class 'orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeRestoreArgs'>): <orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeCheckpointHandler object at 0x7e3c004a8440>, ('metrics', <class 'orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonSaveArgs'>): <orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonCheckpointHandler object at 0x7e3c002b58b0>, ('metrics', <class 'orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonRestoreArgs'>): <orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonCheckpointHandler object at 0x7e3c002b58b0>}). I0425 15:35:33.327469 138886580102976 abstract_checkpointer.py:35] orbax-checkpoint version: 0.11.33 I0425 15:35:33.327550 138886580102976 async_checkpointer.py:177] [process=6][thread=MainThread] Using barrier_sync_fn: <function get_barrier_sync_fn.<locals>._fn at 0x7e3c00218180> timeout: 600 secs and primary_host=0 for async checkpoint writes I0425 15:35:35.774960 138886580102976 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_10_shardy_false/checkpoints I0425 15:35:36.202451 138886580102976 checkpoint_manager.py:929] [process=6][thread=MainThread] CheckpointManager created, primary_host=0, CheckpointManagerOptions=CheckpointManagerOptions(save_interval_steps=1, max_to_keep=None, keep_time_interval=None, keep_period=None, should_keep_fn=None, best_fn=None, best_mode='max', keep_checkpoints_without_metrics=True, step_prefix=None, step_format_fixed_length=None, step_name_format=None, create=True, cleanup_tmp_directories=False, save_on_steps=frozenset(), single_host_load_and_broadcast=False, todelete_subdir=None, todelete_full_path=None, enable_background_delete=False, read_only=False, enable_async_checkpointing=True, async_options=None, multiprocessing_options=MultiprocessingOptions(primary_host=0, active_processes=None, barrier_sync_key_prefix=None), should_save_fn=None, file_options=FileOptions(path_permission_mode=None), save_root_metadata=True, temporary_path_class=None, save_decision_policy=FixedIntervalPolicy(interval=10), preservation_policy=LatestN(n=None), prevent_write_metrics=False, enable_should_save_is_saving_in_progress_check=True, enable_per_process_directory_creation=False, lightweight_initialize=False), root_directory=gs://lance-maxtext/linen_ckpt_xpk_feat_nnx_set_defaults_true_20260425_150412/linen_xpk_feat_nnx_set_defaults_true_20260425_150412_10_shardy_false/checkpoints: <orbax.checkpoint.checkpoint_manager.CheckpointManager object at 0x7e3c00542900> I0425 15:35:36.202628 138886580102976 checkpointing.py:302] Checkpoint manager created! I0425 15:35:37.104871 138886580102976 nnx_wrappers.py:455] Unknown Logical: bfloat16[32,2048,2048]...................................... ('activation_batch', 'activation_norm_length', 'activation_embed'). I0425 15:35:37.104979 138886580102976 nnx_wrappers.py:455] Unknown Physical: bfloat16[32,2048,2048]...................................... ('fsdp', None, None). I0425 15:35:37.474511 138886580102976 attentions.py:1084] attentions/inputs_q Logical: bfloat16[32,2048,2048]...................................... ('activation_batch', 'activation_attn_length', 'activation_attn_embed'). I0425 15:35:37.474605 138886580102976 attentions.py:1084] attentions/inputs_q Physical: bfloat16[32,2048,2048]...................................... ('fsdp', None, None). I0425 15:35:37.490487 138886580102976 attentions.py:1085] attentions/inputs_kv Logical: bfloat16[32,2048,2048]...................................... ('activation_batch', 'activation_attn_length', 'activation_attn_embed'). I0425 15:35:37.490551 138886580102976 attentions.py:1085] attentions/inputs_kv Physical: bfloat16[32,2048,2048]...................................... ('fsdp', None, None). I0425 15:35:37.513448 138886580102976 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 15:35:37.513522 138886580102976 attentions.py:1150] attentions/query Physical: bfloat16[32,2048,16,128].................................... ('fsdp', None, None, None). I0425 15:35:37.529491 138886580102976 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 15:35:37.529559 138886580102976 attentions.py:1151] attentions/key Physical: bfloat16[32,2048,16,128].................................... ('fsdp', None, None, None). I0425 15:35:37.545424 138886580102976 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 15:35:37.545488 138886580102976 attentions.py:1152] attentions/value Physical: bfloat16[32,2048,16,128].................................... ('fsdp', None, None, None). I0425 15:35:37.570312 138886580102976 attentions.py:1193] attentions/out Logical: bfloat16[32,2048,16,128].................................... ('activation_batch', 'activation_attn_length', 'activation_heads', 'activation_kv'). I0425 15:35:37.570384 138886580102976 attentions.py:1193] attentions/out Physical: bfloat16[32,2048,16,128].................................... ('fsdp', None, None, None). I0425 15:35:37.590017 138886580102976 linears.py:541] linears/x Logical: bfloat16[32,2048,7168]...................................... ('activation_batch', 'activation_length', 'activation_mlp'). I0425 15:35:37.590090 138886580102976 linears.py:541] linears/x Physical: bfloat16[32,2048,7168]...................................... ('fsdp', None, None). I0425 15:35:37.780024 138886580102976 checkpointing.py:578] checkpoint manager exists so trying to load this run's existing checkpoint I0425 15:35:37.780152 138886580102976 checkpointing.py:676] No existing checkpoints found, not restoring checkpoint. fsdp: 32 I0425 15:35:39.188023 138886580102976 maxtext_utils.py:1740] params/params/decoder/decoder_norm/scale Shape: float32[2048] Logical: PartitionSpec('norm',) Physical: (None,) I0425 15:35:39.188167 138886580102976 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 15:35:39.188223 138886580102976 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 15:35:39.188286 138886580102976 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 15:35:39.188341 138886580102976 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 15:35:39.188378 138886580102976 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 15:35:39.188429 138886580102976 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 15:35:39.188480 138886580102976 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 15:35:39.188520 138886580102976 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 15:35:39.188558 138886580102976 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 15:35:39.188606 138886580102976 maxtext_utils.py:1740] params/params/decoder/logits_dense/kernel Shape: float32[2048,32000] Logical: PartitionSpec('embed', 'vocab') Physical: ('fsdp', None) I0425 15:35:39.188652 138886580102976 maxtext_utils.py:1740] params/params/token_embedder/embedding Shape: float32[32000,2048] Logical: PartitionSpec('vocab', 'embed') Physical: (None, 'fsdp') I0425 15:35:39.689629 138886580102976 train.py:158] train/xent Logical: float32[32,2048]............................................ ('activation_embed_and_logits_batch', 'activation_length'). I0425 15:35:39.689722 138886580102976 train.py:158] train/xent Physical: float32[32,2048]............................................ ('fsdp', None). I0425 15:35:39.704816 138886580102976 train.py:165] train/z_loss Logical: float32[32,2048]............................................ ('activation_embed_and_logits_batch', 'activation_length'). I0425 15:35:39.704875 138886580102976 train.py:165] train/z_loss Physical: float32[32,2048]............................................ ('fsdp', None). I0425 15:35:50.726926 138886580102976 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 15:35:50.727874 138886580102976 metric_logger.py:289] number parameters: 1.104 billion I0425 15:36:01.868975 138886580102976 checkpointing.py:794] Waiting for step 0 to finish before checkpoint... I0425 15:36:02.023029 138886580102976 checkpointing.py:798] Waited 0.1540384292602539 seconds for step 0 to finish before starting checkpointing. I0425 15:36:02.026064 138886580102976 checkpoint_manager.py:2013] [process=6][thread=MainThread][wait_until_finished] No Save Finalize thread to wait for. Returning. I0425 15:36:02.027849 138886580102976 checkpoint_manager.py:1518] [process=6] Saving checkpoint at step 0 I0425 15:36:02.030665 138886580102976 async_checkpointer.py:452] [process=6] 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_10_shardy_false/checkpoints/0. I0425 15:36:02.611107 138886580102976 signaling_client.py:364] Using JaxDistributedSignalingClient I0425 15:36:02.612352 138886580102976 jax_array_handlers.py:358] Scheduling D2H of 39 prioritized jax.Array. I0425 15:36:02.612411 138886580102976 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 15:36:02.861610 138886580102976 base_pytree_checkpoint_handler.py:153] [process=6][thread=MainThread] Initiated "orbax.checkpoint._src.serialization.jax_array_handlers.ArrayHandler".serialize. Time taken: 0.250597s I0425 15:36:02.861783 138886580102976 base_pytree_checkpoint_handler.py:129] [process=6] /jax/checkpoint/write/blocking_gbytes_per_sec: 3.073 GiB/s (total gbytes: 1.5 GiB) (time elapsed: 0.5019736289978027 s) (per-host) I0425 15:36:02.861832 138886580102976 base_pytree_checkpoint_handler.py:737] [process=6][thread=MainThread] Initiated Pytree async_save. Time taken: 0.502032s (batch_requests_ready=0.247516s, total_serialization_initiated=0.254450s, others=0.000066s) I0425 15:36:02.861920 138886580102976 composite_checkpoint_handler.py:715] [process=6][thread=MainThread] Initiated CompositeCheckpointHandler.async_save. Time taken: 0.506516s (all_items=0.000019s, per_item={'items': '0.00001860'}, temp_paths=0.506498) I0425 15:36:02.862877 138761040492288 async_checkpointer.py:79] [process=6][thread=async_save] Background save thread started. I0425 15:36:02.863043 138886580102976 async_checkpointer.py:561] Finished blocking save. Time taken: 0.835124s. 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_10_shardy_false/checkpoints/0. I0425 15:36:02.905719 138886580102976 checkpoint_manager.py:1566] [process=6][thread=MainThread][step=0] Starting CheckpointManager Save Finalize thread=save_finalize I0425 15:36:02.906042 138759538448128 async_checkpointer.py:265] [process=6][thread=save_finalize] Waiting for background save thread=async_save. I0425 15:36:02.906208 138886580102976 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_10_shardy_false/checkpoints', 'reached_preemption': False, 'preemption_received_at': None, 'synchronous': False, 'wait_for_prev_start_time': 1777131362.0260453, 'wait_for_prev_duration_secs': 7.724761962890625e-05, 'time_between_consecutive_saves_sec': None, 'checkpointer_blocking_start_time': 1777131362.0278883, 'checkpointer_blocking_duration_secs': 0.8352851867675781, 'get_old_steps_start_time': 1777131362.8631933, 'get_old_steps_duration_secs': 2.6702880859375e-05, 'checkpoint_manager_blocking_start_time': 1777131362.023659, 'checkpoint_manager_blocking_duration_secs': 0.8825082778930664} I0425 15:36:02.906315 138886580102976 checkpointing.py:409] Started an asynchronous checkpoint save for step 0 I0425 15:36:02.906367 138886580102976 max_utils.py:750] Memstats: After params initialized: I0425 15:36:02.906420 138886580102976 max_utils.py:756] Using (GB) 0.43 / 31.25 (1.376000%) on TPU_24(process=6,(0,6,0,0)) I0425 15:36:02.906454 138886580102976 max_utils.py:756] Using (GB) 0.43 / 31.25 (1.376000%) on TPU_25(process=6,(1,6,0,0)) I0425 15:36:02.906481 138886580102976 max_utils.py:756] Using (GB) 0.43 / 31.25 (1.376000%) on TPU_28(process=6,(0,7,0,0)) I0425 15:36:02.906506 138886580102976 max_utils.py:756] Using (GB) 0.43 / 31.25 (1.376000%) on TPU_29(process=6,(1,7,0,0)) I0425 15:36:03.167169 138886580102976 metric_logger.py:185] completed step: 0, seconds: 11.141, TFLOP/s/device: 1.220, Tokens/s/device: 183.826, total_weights: 65536, loss: 10.874 I0425 15:36:03.354977 138886580102976 metric_logger.py:185] completed step: 1, seconds: 1.297, TFLOP/s/device: 10.477, Tokens/s/device: 1579.180, total_weights: 65536, loss: 10.874 I0425 15:36:03.764145 138886580102976 metric_logger.py:185] completed step: 2, seconds: 0.041, TFLOP/s/device: 334.197, Tokens/s/device: 50373.869, total_weights: 65536, loss: 10.263 I0425 15:36:03.911366 138886580102976 metric_logger.py:185] completed step: 3, seconds: 0.409, TFLOP/s/device: 33.217, Tokens/s/device: 5006.772, total_weights: 65536, loss: 9.734 I0425 15:36:04.207252 138886580102976 metric_logger.py:185] completed step: 4, seconds: 0.153, TFLOP/s/device: 88.716, Tokens/s/device: 13372.249, total_weights: 65536, loss: 9.278 I0425 15:36:04.213106 138886580102976 metric_logger.py:185] completed step: 5, seconds: 0.147, TFLOP/s/device: 92.219, Tokens/s/device: 13900.201, total_weights: 65536, loss: 8.892 I0425 15:36:23.181653 138886580102976 metric_logger.py:185] completed step: 6, seconds: 0.296, TFLOP/s/device: 45.856, Tokens/s/device: 6911.960, total_weights: 65536, loss: 8.592 I0425 15:36:23.329195 138886580102976 metric_logger.py:185] completed step: 7, seconds: 18.822, TFLOP/s/device: 0.722, Tokens/s/device: 108.809, total_weights: 65536, loss: 8.384 I0425 15:36:23.476637 138886580102976 metric_logger.py:185] completed step: 8, seconds: 0.152, TFLOP/s/device: 89.237, Tokens/s/device: 13450.765, total_weights: 65536, loss: 8.255 I0425 15:36:23.623718 138886580102976 checkpointing.py:794] Waiting for step 9 to finish before checkpoint... I0425 15:36:23.624413 138886580102976 checkpointing.py:798] Waited 0.0007150173187255859 seconds for step 9 to finish before starting checkpointing. I0425 15:36:23.626439 138886580102976 checkpoint_manager.py:2024] [process=6][thread=MainThread][step=0][wait_until_finished] Waiting for Save Finalize thread (save_finalize) to complete. I0425 15:36:24.588225 2751 google_auth_provider.cc:181] Running on GCE, using service account 562977990677-compute@developer.gserviceaccount.com I0425 15:36:26.485158 138761048884992 array_metadata_store.py:203] [process=6][thread=array_type_handler] Wrote 39 array_metadata.ArrayMetadata to gs://lance-maxtext/linen_ckpt_xpk_feat_nnx_set_defaults_true_20260425_150412/linen_xpk_feat_nnx_set_defaults_true_20260425_150412_10_shardy_false/checkpoints/0/items/array_metadatas/process_6 I0425 15:36:57.888005 138761040492288 base_pytree_checkpoint_handler.py:129] [process=6] /jax/checkpoint/write/gbytes_per_sec: 28.447 MiB/s (total gbytes: 1.5 GiB) (time elapsed: 55.52816462516785 s) (per-host) I0425 15:36:57.888139 138761040492288 async_checkpointer.py:90] [process=6][thread=async_save] 3 Handler Commit operations completed. Time taken: 55.025151s. I0425 15:37:07.422568 138761040492288 async_checkpointer.py:144] [process=6][thread=async_save] Background save thread done. Time taken: 64.559563s. I0425 15:37:07.422857 138759538448128 async_checkpointer.py:273] [process=6][thread=save_finalize] Done with waiting for background save thread=async_save. I0425 15:37:07.422985 138759538448128 async_checkpointer.py:283] [process=6][thread=save_finalize] No errors found in background save thread=async_save. I0425 15:37:07.423034 138759538448128 checkpoint_manager.py:2133] [process=6][thread=save_finalize][step=0] CheckpointManager Save Finalize is syncing with other hosts... I0425 15:37:07.424651 138759538448128 checkpoint_manager.py:2142] [process=6][thread=save_finalize][step=0] CheckpointManager Save Finalize is done on all hosts. I0425 15:37:07.424828 138886580102976 checkpoint_manager.py:2036] [process=6][thread=MainThread][step=0][wait_until_finished] Done waiting for Save Finalize thread (save_finalize) running at step=0. W0425 15:37:07.424963 138886580102976 checkpoint_manager.py:1458] Waiting for previous save to complete took 43.798530 seconds. If this number is high, consider checkpointing less frequently. I0425 15:37:07.426747 138886580102976 checkpoint_manager.py:1518] [process=6] Saving checkpoint at step 9 I0425 15:37:07.430041 138886580102976 async_checkpointer.py:452] [process=6] 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_10_shardy_false/checkpoints/9. I0425 15:37:07.954673 138886580102976 jax_array_handlers.py:358] Scheduling D2H of 39 prioritized jax.Array. I0425 15:37:07.954773 138886580102976 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 15:37:08.003636 138886580102976 base_pytree_checkpoint_handler.py:153] [process=6][thread=MainThread] Initiated "orbax.checkpoint._src.serialization.jax_array_handlers.ArrayHandler".serialize. Time taken: 0.050134s I0425 15:37:08.003803 138886580102976 base_pytree_checkpoint_handler.py:129] [process=6] /jax/checkpoint/write/blocking_gbytes_per_sec: 5.158 GiB/s (total gbytes: 1.5 GiB) (time elapsed: 0.2990453243255615 s) (per-host) I0425 15:37:08.003853 138886580102976 base_pytree_checkpoint_handler.py:737] [process=6][thread=MainThread] Initiated Pytree async_save. Time taken: 0.299104s (batch_requests_ready=0.247049s, total_serialization_initiated=0.051990s, others=0.000066s) I0425 15:37:08.003942 138886580102976 composite_checkpoint_handler.py:715] [process=6][thread=MainThread] Initiated CompositeCheckpointHandler.async_save. Time taken: 0.303339s (all_items=0.000015s, per_item={'items': '0.00001454'}, temp_paths=0.303325) I0425 15:37:08.004868 138759552804608 async_checkpointer.py:79] [process=6][thread=async_save] Background save thread started. I0425 15:37:08.005026 138886580102976 async_checkpointer.py:561] Finished blocking save. Time taken: 0.578207s. 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_10_shardy_false/checkpoints/9. I0425 15:37:08.213632 138886580102976 checkpoint_manager.py:1566] [process=6][thread=MainThread][step=9] Starting CheckpointManager Save Finalize thread=save_finalize I0425 15:37:08.214003 138759538448128 async_checkpointer.py:265] [process=6][thread=save_finalize] Waiting for background save thread=async_save. I0425 15:37:08.214154 138886580102976 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_10_shardy_false/checkpoints', 'reached_preemption': False, 'preemption_received_at': None, 'synchronous': False, 'wait_for_prev_start_time': 1777131383.6264095, 'wait_for_prev_duration_secs': 43.798529624938965, 'time_between_consecutive_saves_sec': None, 'checkpointer_blocking_start_time': 1777131427.4267857, 'checkpointer_blocking_duration_secs': 0.5783681869506836, 'get_old_steps_start_time': 1777131428.0051732, 'get_old_steps_duration_secs': 2.47955322265625e-05, 'checkpoint_manager_blocking_start_time': 1777131383.6246507, 'checkpoint_manager_blocking_duration_secs': 44.589465618133545} I0425 15:37:08.214322 138886580102976 checkpointing.py:409] Started an asynchronous checkpoint save for step 9 I0425 15:37:08.214372 138886580102976 checkpoint_manager.py:2024] [process=6][thread=MainThread][step=9][wait_until_finished] Waiting for Save Finalize thread (save_finalize) to complete. I0425 15:37:12.744585 138761048884992 array_metadata_store.py:203] [process=6][thread=array_type_handler] Wrote 39 array_metadata.ArrayMetadata to gs://lance-maxtext/linen_ckpt_xpk_feat_nnx_set_defaults_true_20260425_150412/linen_xpk_feat_nnx_set_defaults_true_20260425_150412_10_shardy_false/checkpoints/9/items/array_metadatas/process_6 I0425 15:37:48.890293 138759552804608 base_pytree_checkpoint_handler.py:129] [process=6] /jax/checkpoint/write/gbytes_per_sec: 38.353 MiB/s (total gbytes: 1.5 GiB) (time elapsed: 41.18549418449402 s) (per-host) I0425 15:37:48.890413 138759552804608 async_checkpointer.py:90] [process=6][thread=async_save] 3 Handler Commit operations completed. Time taken: 40.885436s. I0425 15:37:57.568114 138759552804608 async_checkpointer.py:144] [process=6][thread=async_save] Background save thread done. Time taken: 49.563106s. I0425 15:37:57.568404 138759538448128 async_checkpointer.py:273] [process=6][thread=save_finalize] Done with waiting for background save thread=async_save. I0425 15:37:57.568522 138759538448128 async_checkpointer.py:283] [process=6][thread=save_finalize] No errors found in background save thread=async_save. I0425 15:37:57.568568 138759538448128 checkpoint_manager.py:2133] [process=6][thread=save_finalize][step=9] CheckpointManager Save Finalize is syncing with other hosts... I0425 15:37:57.569983 138759538448128 checkpoint_manager.py:2142] [process=6][thread=save_finalize][step=9] CheckpointManager Save Finalize is done on all hosts. I0425 15:37:57.570162 138886580102976 checkpoint_manager.py:2036] [process=6][thread=MainThread][step=9][wait_until_finished] Done waiting for Save Finalize thread (save_finalize) running at step=9. I0425 15:37:57.570311 138886580102976 checkpoint_manager.py:2013] [process=6][thread=MainThread][wait_until_finished] No Save Finalize thread to wait for. Returning. I0425 15:37:57.570935 138886580102976 metric_logger.py:185] completed step: 9, seconds: 0.147, TFLOP/s/device: 92.245, Tokens/s/device: 13904.164, 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 15:38:09 UTC 2026 EXIT_CODE=0