main (574ad3fb9) vs test/pipeline-scan-nnx (574ad3fb9)| Metric | main 574ad3fb9 | test/pipeline-scan-nnx 574ad3fb9 | Diff (test/pipeline-scan-nnx − main) |
|---|---|---|---|
| Parameters | 1.104 billion | 1.104 billion | — |
| Final loss | 9.4570 | 9.4570 | 0 |
| TFLOP/s | 93.923 | 93.861 | -0.062 |
| Tok/s | 14157.1 | 14147.7 | -9.388 |
| Avg s/step | 6.899 | 6.046 | -0.853 |
| 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 18 18:20:34 UTC 2026 PyTorch was not found. Models won't be available and only tokenizers, configuration and file/data utilities can be used. 2026-04-18 18:20:58.747697: 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) I0418 18:20:59.279374 138120810325824 max_utils.py:273] Attempting to initialize the jax distributed system... I0418 18:21:08.318775 138120810325824 distributed.py:149] Starting JAX distributed service on [::]:8482 I0418 18:21:08.321148 138120810325824 distributed.py:172] Connecting to JAX distributed service on mt-06-grad-accum-0jggr-slice-job-0-0.mt-06-grad-accum-0jggr:8482 I0418 18:21:09.061537 138120810325824 max_utils.py:284] Jax distributed system initialized! I0418 18:21:15.305518 138120810325824 max_utils.py:800] System Information: Jax Version: 0.9.2 I0418 18:21:15.305618 138120810325824 max_utils.py:801] System Information: Jaxlib Version: 0.9.2 I0418 18:21:15.305668 138120810325824 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 I0418 18:21:15.305708 138120810325824 train_utils.py:348] WARNING: Sequence packing is essentially ignored for synthetic data. Please use a real dataset to use sequence packing. I0418 18:21:16.021094 138120810325824 maxtext_utils.py:1551] Num_devices: 32, shape (1, 1, 1, 32, 1, 1, 1, 1, 1, 1, 1, 1, 1) I0418 18:21:16.021367 138120810325824 checkpointing.py:677] Setting up checkpoint logger... I0418 18:21:16.021421 138120810325824 checkpointing.py:233] Creating checkpoint manager with ocdbt=True and zarr3=True I0418 18:21:16.021479 138120810325824 pytree_checkpoint_handler.py:592] save_device_host_concurrent_bytes=None I0418 18:21:16.021821 138120810325824 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 0x7d9e0a0481a0>, enable_pinned_host_transfer=False, save_concurrent_bytes: 96000000000 (89.4 GiB), restore_concurrent_bytes: 96000000000 (89.4 GiB) I0418 18:21:20.187852 138120810325824 checkpointing.py:265] Enabling policy for fixed interval checkpointing. I0418 18:21:20.188086 138120810325824 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 0x7d8a044ef200>}, handler_registry=None I0418 18:21:20.188320 138120810325824 composite_checkpoint_handler.py:237] Deferred registration for item: "items". Adding handler `<orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeCheckpointHandler object at 0x7d8a044ef200>` 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`. I0418 18:21:20.188367 138120810325824 composite_checkpoint_handler.py:237] Deferred registration for item: "metrics". Adding handler `<orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonCheckpointHandler object at 0x7d8a04304740>` 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`. I0418 18:21:20.188402 138120810325824 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 0x7d8a044ef200>, ('items', <class 'orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeRestoreArgs'>): <orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeCheckpointHandler object at 0x7d8a044ef200>, ('metrics', <class 'orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonSaveArgs'>): <orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonCheckpointHandler object at 0x7d8a04304740>, ('metrics', <class 'orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonRestoreArgs'>): <orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonCheckpointHandler object at 0x7d8a04304740>}). I0418 18:21:20.188730 138120810325824 abstract_checkpointer.py:35] orbax-checkpoint version: 0.11.34 I0418 18:21:20.188801 138120810325824 async_checkpointer.py:192] [process=5][thread=MainThread] Using barrier_sync_fn: <function get_barrier_sync_fn.<locals>._fn at 0x7d892c5cf880> timeout: 1200 secs and primary_host=0 for async checkpoint writes I0418 18:21:21.333163 138120810325824 checkpoint_manager.py:1812] Found 0 checkpoint steps in gs://lance-maxtext/linen_ckpt_xpk_main_20260418_180002/linen_xpk_main_20260418_180002_06_grad_accum/checkpoints I0418 18:21:22.190575 138120810325824 checkpoint_manager.py:929] [process=5][thread=MainThread] CheckpointManager created, primary_host=0, CheckpointManagerOptions=CheckpointManagerOptions(save_interval_steps=1, max_to_keep=None, keep_time_interval=None, keep_period=None, should_keep_fn=None, best_fn=None, best_mode='max', keep_checkpoints_without_metrics=True, step_prefix=None, step_format_fixed_length=None, step_name_format=None, create=True, cleanup_tmp_directories=False, save_on_steps=frozenset(), single_host_load_and_broadcast=False, todelete_subdir=None, todelete_full_path=None, enable_background_delete=False, read_only=False, enable_async_checkpointing=True, async_options=None, multiprocessing_options=MultiprocessingOptions(primary_host=0, active_processes=None, barrier_sync_key_prefix=None), should_save_fn=None, file_options=FileOptions(path_permission_mode=None), save_root_metadata=True, temporary_path_class=None, save_decision_policy=FixedIntervalPolicy(interval=10), preservation_policy=LatestN(n=None), prevent_write_metrics=False, enable_should_save_is_saving_in_progress_check=True, enable_per_process_directory_creation=False, lightweight_initialize=False), root_directory=gs://lance-maxtext/linen_ckpt_xpk_main_20260418_180002/linen_xpk_main_20260418_180002_06_grad_accum/checkpoints: <orbax.checkpoint.checkpoint_manager.CheckpointManager object at 0x7d8a04300fb0> I0418 18:21:22.190736 138120810325824 checkpointing.py:301] Checkpoint manager created! I0418 18:21:23.900480 138120810325824 nnx_wrappers.py:437] Unknown Logical: bfloat16[32,2048,2048]...................................... ('activation_batch', 'activation_norm_length', 'activation_embed'). I0418 18:21:23.900593 138120810325824 nnx_wrappers.py:437] Unknown Physical: bfloat16[32,2048,2048]...................................... ('fsdp', None, None). I0418 18:21:24.285784 138120810325824 attentions.py:1088] attentions/inputs_q Logical: bfloat16[32,2048,2048]...................................... ('activation_batch', 'activation_attn_length', 'activation_attn_embed'). I0418 18:21:24.285870 138120810325824 attentions.py:1088] attentions/inputs_q Physical: bfloat16[32,2048,2048]...................................... ('fsdp', None, None). I0418 18:21:24.302419 138120810325824 attentions.py:1089] attentions/inputs_kv Logical: bfloat16[32,2048,2048]...................................... ('activation_batch', 'activation_attn_length', 'activation_attn_embed'). I0418 18:21:24.302487 138120810325824 attentions.py:1089] attentions/inputs_kv Physical: bfloat16[32,2048,2048]...................................... ('fsdp', None, None). I0418 18:21:24.326354 138120810325824 attentions.py:1154] attentions/query Logical: bfloat16[32,2048,16,128].................................... ('activation_kv_batch', 'activation_attn_length', 'activation_kv_heads', 'activation_kv_head_dim'). I0418 18:21:24.326422 138120810325824 attentions.py:1154] attentions/query Physical: bfloat16[32,2048,16,128].................................... ('fsdp', None, None, None). I0418 18:21:24.343260 138120810325824 attentions.py:1155] attentions/key Logical: bfloat16[32,2048,16,128].................................... ('activation_kv_batch', 'activation_attn_length', 'activation_kv_heads', 'activation_kv_head_dim'). I0418 18:21:24.343318 138120810325824 attentions.py:1155] attentions/key Physical: bfloat16[32,2048,16,128].................................... ('fsdp', None, None, None). I0418 18:21:24.360033 138120810325824 attentions.py:1156] attentions/value Logical: bfloat16[32,2048,16,128].................................... ('activation_kv_batch', 'activation_attn_length', 'activation_kv_heads', 'activation_kv_head_dim'). I0418 18:21:24.360091 138120810325824 attentions.py:1156] attentions/value Physical: bfloat16[32,2048,16,128].................................... ('fsdp', None, None, None). I0418 18:21:24.385735 138120810325824 attentions.py:1197] attentions/out Logical: bfloat16[32,2048,16,128].................................... ('activation_batch', 'activation_attn_length', 'activation_heads', 'activation_kv'). I0418 18:21:24.385807 138120810325824 attentions.py:1197] attentions/out Physical: bfloat16[32,2048,16,128].................................... ('fsdp', None, None, None). I0418 18:21:24.406907 138120810325824 linears.py:525] linears/x Logical: bfloat16[32,2048,7168]...................................... ('activation_batch', 'activation_length', 'activation_mlp'). I0418 18:21:24.406972 138120810325824 linears.py:525] linears/x Physical: bfloat16[32,2048,7168]...................................... ('fsdp', None, None). I0418 18:21:24.623523 138120810325824 checkpointing.py:577] checkpoint manager exists so trying to load this run's existing checkpoint I0418 18:21:24.623625 138120810325824 checkpointing.py:665] No existing checkpoints found, not restoring checkpoint. fsdp: 32 I0418 18:21:26.055944 138120810325824 maxtext_utils.py:1654] params/params/decoder/decoder_norm/scale Shape: float32[2048] Logical: P('norm',) Physical: (None,) I0418 18:21:26.056072 138120810325824 maxtext_utils.py:1654] params/params/decoder/layers/mlp/wi_0/kernel Shape: float32[2048,16,7168] Logical: P('embed', 'layers', 'mlp') Physical: ('fsdp', None, None) I0418 18:21:26.056124 138120810325824 maxtext_utils.py:1654] params/params/decoder/layers/mlp/wi_1/kernel Shape: float32[2048,16,7168] Logical: P('embed', 'layers', 'mlp') Physical: ('fsdp', None, None) I0418 18:21:26.056180 138120810325824 maxtext_utils.py:1654] params/params/decoder/layers/mlp/wo/kernel Shape: float32[7168,16,2048] Logical: P('mlp', 'layers', 'embed') Physical: (None, None, 'fsdp') I0418 18:21:26.056235 138120810325824 maxtext_utils.py:1654] params/params/decoder/layers/post_self_attention_layer_norm/scale Shape: float32[2048,16] Logical: P('norm', 'layers') Physical: (None, None) I0418 18:21:26.056272 138120810325824 maxtext_utils.py:1654] params/params/decoder/layers/pre_self_attention_layer_norm/scale Shape: float32[2048,16] Logical: P('norm', 'layers') Physical: (None, None) I0418 18:21:26.056321 138120810325824 maxtext_utils.py:1654] 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) I0418 18:21:26.056371 138120810325824 maxtext_utils.py:1654] 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') I0418 18:21:26.056409 138120810325824 maxtext_utils.py:1654] 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) I0418 18:21:26.056444 138120810325824 maxtext_utils.py:1654] 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) I0418 18:21:26.056567 138120810325824 maxtext_utils.py:1654] params/params/decoder/logits_dense/kernel Shape: float32[2048,32000] Logical: P('embed_vocab', 'vocab') Physical: ('fsdp', None) I0418 18:21:26.056620 138120810325824 maxtext_utils.py:1654] params/params/token_embedder/embedding Shape: float32[32000,2048] Logical: P('vocab', 'embed_vocab') Physical: (None, 'fsdp') I0418 18:21:26.079357 138120810325824 gradient_accumulation.py:68] gradient_accumulation/inputs Logical: float32[2048]............................................... Unknown. I0418 18:21:26.079424 138120810325824 gradient_accumulation.py:68] gradient_accumulation/inputs Physical: float32[2048]............................................... (None,). I0418 18:21:26.094476 138120810325824 gradient_accumulation.py:68] gradient_accumulation/inputs Logical: float32[2048,16,7168]....................................... Unknown. I0418 18:21:26.094532 138120810325824 gradient_accumulation.py:68] gradient_accumulation/inputs Physical: float32[2048,16,7168]....................................... ('fsdp', None, None). I0418 18:21:26.124304 138120810325824 gradient_accumulation.py:68] gradient_accumulation/inputs Logical: float32[7168,16,2048]....................................... Unknown. I0418 18:21:26.124368 138120810325824 gradient_accumulation.py:68] gradient_accumulation/inputs Physical: float32[7168,16,2048]....................................... (None, None, 'fsdp'). I0418 18:21:26.139235 138120810325824 gradient_accumulation.py:68] gradient_accumulation/inputs Logical: float32[2048,16]............................................ Unknown. I0418 18:21:26.139289 138120810325824 gradient_accumulation.py:68] gradient_accumulation/inputs Physical: float32[2048,16]............................................ (None, None). I0418 18:21:26.169518 138120810325824 gradient_accumulation.py:68] gradient_accumulation/inputs Logical: float32[2048,16,16,128]..................................... Unknown. I0418 18:21:26.169583 138120810325824 gradient_accumulation.py:68] gradient_accumulation/inputs Physical: float32[2048,16,16,128]..................................... ('fsdp', None, None, None). I0418 18:21:26.184443 138120810325824 gradient_accumulation.py:68] gradient_accumulation/inputs Logical: float32[16,16,128,2048]..................................... Unknown. I0418 18:21:26.184508 138120810325824 gradient_accumulation.py:68] gradient_accumulation/inputs Physical: float32[16,16,128,2048]..................................... (None, None, None, 'fsdp'). I0418 18:21:26.230402 138120810325824 gradient_accumulation.py:68] gradient_accumulation/inputs Logical: float32[2048,32000]......................................... Unknown. I0418 18:21:26.230484 138120810325824 gradient_accumulation.py:68] gradient_accumulation/inputs Physical: float32[2048,32000]......................................... ('fsdp', None). I0418 18:21:26.245301 138120810325824 gradient_accumulation.py:68] gradient_accumulation/inputs Logical: float32[32000,2048]......................................... Unknown. I0418 18:21:26.245358 138120810325824 gradient_accumulation.py:68] gradient_accumulation/inputs Physical: float32[32000,2048]......................................... (None, 'fsdp'). I0418 18:21:26.911927 138120810325824 train.py:155] train/xent Logical: float32[32,2048]............................................ ('activation_embed_and_logits_batch', 'activation_length'). I0418 18:21:26.912023 138120810325824 train.py:155] train/xent Physical: float32[32,2048]............................................ ('fsdp', None). I0418 18:21:26.927740 138120810325824 train.py:162] train/z_loss Logical: float32[32,2048]............................................ ('activation_embed_and_logits_batch', 'activation_length'). I0418 18:21:26.927803 138120810325824 train.py:162] train/z_loss Physical: float32[32,2048]............................................ ('fsdp', None). I0418 18:21:38.331479 138120810325824 max_utils.py:791] Total memory size: 1.7 GB, Output size: 0.4 GB, Temp size: 1.3 GB, Argument size: 0.4 GB, Host temp size: 0.0 GB. I0418 18:21:38.332359 138120810325824 metric_logger.py:301] number parameters: 1.104 billion I0418 18:21:40.239638 138120810325824 checkpointing.py:772] Waiting for step 0 to finish before checkpoint... I0418 18:21:50.993809 138120810325824 checkpointing.py:776] Waited 10.754153490066528 seconds for step 0 to finish before starting checkpointing. I0418 18:21:50.996673 138120810325824 checkpoint_manager.py:2009] [process=5][thread=MainThread][wait_until_finished] No Save Finalize thread to wait for. Returning. I0418 18:21:50.998681 138120810325824 checkpoint_manager.py:1512] [process=5] Saving checkpoint at step 0 I0418 18:21:51.000091 138120810325824 event_tracking.py:70] [process=5] [async] Started save checkpoint @ gs://lance-maxtext/linen_ckpt_xpk_main_20260418_180002/linen_xpk_main_20260418_180002_06_grad_accum/checkpoints/0. I0418 18:21:51.756288 138120810325824 signaling_client.py:364] Using JaxDistributedSignalingClient I0418 18:21:51.757306 138120810325824 jax_array_handlers.py:360] Scheduling D2H of 39 prioritized jax.Array. I0418 18:21:51.757362 138120810325824 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 I0418 18:21:52.031582 138120810325824 base_pytree_checkpoint_handler.py:154] [process=5][thread=MainThread] Initiated "orbax.checkpoint._src.serialization.jax_array_handlers.ArrayHandler".serialize. Time taken: 0.275364s I0418 18:21:52.031753 138120810325824 base_pytree_checkpoint_handler.py:130] [process=5] /jax/orbax/write/blocking_gbytes_per_sec: 5.496 GiB/s (total gbytes: 1.5 GiB) (time elapsed: 0.2806868553161621 s) (per-host) I0418 18:21:52.031807 138120810325824 base_pytree_checkpoint_handler.py:768] [process=5][thread=MainThread] Initiated Pytree async_save. Time taken: 0.280751s (batch_requests_ready=0.002174s, total_serialization_initiated=0.278505s, others=0.000071s) I0418 18:21:52.031900 138120810325824 composite_checkpoint_handler.py:715] [process=5][thread=MainThread] Initiated CompositeCheckpointHandler.async_save. Time taken: 0.284828s (all_items=0.000018s, per_item={'items': '0.00001788'}, temp_paths=0.284810) I0418 18:21:52.032742 138120810325824 event_tracking.py:125] [process=5] [async] Finished blocking save in 1.03 seconds. Continuing save @ gs://lance-maxtext/linen_ckpt_xpk_main_20260418_180002/linen_xpk_main_20260418_180002_06_grad_accum/checkpoints/0. I0418 18:21:52.033103 137993369503488 async_checkpointer.py:76] [process=5][thread=async_save] Background save thread started. Deadline for this save operation is 2026-04-18 18:41:52.033065 I0418 18:21:52.043927 138120810325824 checkpoint_manager.py:1560] [process=5][thread=MainThread][step=0] Starting CheckpointManager Save Finalize thread=save_finalize I0418 18:21:52.044234 137993352718080 async_checkpointer.py:280] [process=5][thread=save_finalize] Waiting for background save thread=async_save. I0418 18:21:52.044387 138120810325824 standard_logger.py:34] {'step': 0, 'event_type': 'save', 'directory': 'gs://lance-maxtext/linen_ckpt_xpk_main_20260418_180002/linen_xpk_main_20260418_180002_06_grad_accum/checkpoints', 'reached_preemption': False, 'preemption_received_at': None, 'synchronous': False, 'wait_for_prev_start_time': 1776536510.996654, 'wait_for_prev_duration_secs': 6.246566772460938e-05, 'time_between_consecutive_saves_sec': None, 'checkpointer_blocking_start_time': 1776536510.9987192, 'checkpointer_blocking_duration_secs': 1.034545660018921, 'get_old_steps_start_time': 1776536512.0332894, 'get_old_steps_duration_secs': 2.9802322387695312e-05, 'checkpoint_manager_blocking_start_time': 1776536510.9943285, 'checkpoint_manager_blocking_duration_secs': 1.0500211715698242} I0418 18:21:52.044503 138120810325824 checkpointing.py:408] Started an asynchronous checkpoint save for step 0 I0418 18:21:52.044552 138120810325824 max_utils.py:750] Memstats: After params initialized: I0418 18:21:52.044598 138120810325824 max_utils.py:756] Using (GB) 0.43 / 31.25 (1.376000%) on TPU_18(process=5,(2,4,0,0)) I0418 18:21:52.044629 138120810325824 max_utils.py:756] Using (GB) 0.43 / 31.25 (1.376000%) on TPU_19(process=5,(3,4,0,0)) I0418 18:21:52.044655 138120810325824 max_utils.py:756] Using (GB) 0.43 / 31.25 (1.376000%) on TPU_22(process=5,(2,5,0,0)) I0418 18:21:52.044678 138120810325824 max_utils.py:756] Using (GB) 0.43 / 31.25 (1.376000%) on TPU_23(process=5,(3,5,0,0)) I0418 18:21:52.351405 138120810325824 metric_logger.py:196] completed step: 0, seconds: 1.907, TFLOP/s/device: 28.497, Tokens/s/device: 4295.370, total_weights: 262144, loss: 10.877, lm_loss: 10.877, perplexity: 52959.059 I0418 18:21:52.971392 138120810325824 metric_logger.py:196] completed step: 1, seconds: 12.110, TFLOP/s/device: 4.488, Tokens/s/device: 676.449, total_weights: 262144, loss: 10.877, lm_loss: 10.877, perplexity: 52959.059 I0418 18:21:53.549937 138120810325824 metric_logger.py:196] completed step: 2, seconds: 0.041, TFLOP/s/device: 1314.608, Tokens/s/device: 198152.000, total_weights: 262144, loss: 10.563, lm_loss: 10.563, perplexity: 38662.707 I0418 18:21:54.128334 138120810325824 metric_logger.py:196] completed step: 3, seconds: 0.584, TFLOP/s/device: 93.036, Tokens/s/device: 14023.435, total_weights: 262144, loss: 10.272, lm_loss: 10.272, perplexity: 28909.668 I0418 18:21:55.285424 138120810325824 metric_logger.py:196] completed step: 4, seconds: 0.578, TFLOP/s/device: 93.955, Tokens/s/device: 14161.960, total_weights: 262144, loss: 10.022, lm_loss: 10.022, perplexity: 22524.992 I0418 18:21:55.290740 138120810325824 metric_logger.py:196] completed step: 5, seconds: 0.578, TFLOP/s/device: 93.965, Tokens/s/device: 14163.503, total_weights: 262144, loss: 9.820, lm_loss: 9.820, perplexity: 18401.865 I0418 18:21:55.797282 2571 google_auth_provider.cc:181] Running on GCE, using service account 562977990677-compute@developer.gserviceaccount.com I0418 18:21:58.958063 137996609795840 array_metadata_store.py:203] [process=5][thread=array_type_handler] Wrote 39 array_metadata.ArrayMetadata to gs://lance-maxtext/linen_ckpt_xpk_main_20260418_180002/linen_xpk_main_20260418_180002_06_grad_accum/checkpoints/0/items/array_metadatas/process_5 I0418 18:22:29.362634 137993369503488 base_pytree_checkpoint_handler.py:130] [process=5] /jax/orbax/write/gbytes_per_sec: 41.998 MiB/s (total gbytes: 1.5 GiB) (time elapsed: 37.611533403396606 s) (per-host) I0418 18:22:29.362752 137993369503488 async_checkpointer.py:90] [process=5][thread=async_save] 3 Handler Commit operations completed. Time taken: 37.329531s. I0418 18:22:39.554933 137993369503488 async_checkpointer.py:160] [process=5][thread=async_save] Background save thread done. Time taken: 47.521695s. I0418 18:22:39.555216 137993352718080 async_checkpointer.py:288] [process=5][thread=save_finalize] Done with waiting for background save thread=async_save. I0418 18:22:39.555334 137993352718080 async_checkpointer.py:298] [process=5][thread=save_finalize] No errors found in background save thread=async_save. I0418 18:22:39.555385 137993352718080 checkpoint_manager.py:2137] [process=5][thread=save_finalize][step=0] CheckpointManager Save Finalize is syncing with other hosts... I0418 18:22:40.449532 137993352718080 checkpoint_manager.py:2146] [process=5][thread=save_finalize][step=0] CheckpointManager Save Finalize is done on all hosts. I0418 18:22:41.746440 138120810325824 metric_logger.py:196] completed step: 6, seconds: 1.158, TFLOP/s/device: 46.952, Tokens/s/device: 7077.163, total_weights: 262144, loss: 9.667, lm_loss: 9.667, perplexity: 15787.604 I0418 18:22:42.325114 138120810325824 metric_logger.py:196] completed step: 7, seconds: 45.878, TFLOP/s/device: 1.185, Tokens/s/device: 178.562, total_weights: 262144, loss: 9.561, lm_loss: 9.561, perplexity: 14203.827 I0418 18:22:42.903517 138120810325824 metric_logger.py:196] completed step: 8, seconds: 0.583, TFLOP/s/device: 93.208, Tokens/s/device: 14049.289, total_weights: 262144, loss: 9.496, lm_loss: 9.496, perplexity: 13302.920 I0418 18:22:43.481658 138120810325824 checkpointing.py:772] Waiting for step 9 to finish before checkpoint... I0418 18:22:43.482256 138120810325824 checkpointing.py:776] Waited 0.0006222724914550781 seconds for step 9 to finish before starting checkpointing. I0418 18:22:43.484622 138120810325824 checkpoint_manager.py:2009] [process=5][thread=MainThread][wait_until_finished] No Save Finalize thread to wait for. Returning. I0418 18:22:43.486362 138120810325824 checkpoint_manager.py:1512] [process=5] Saving checkpoint at step 9 I0418 18:22:43.487731 138120810325824 event_tracking.py:70] [process=5] [async] Started save checkpoint @ gs://lance-maxtext/linen_ckpt_xpk_main_20260418_180002/linen_xpk_main_20260418_180002_06_grad_accum/checkpoints/9. I0418 18:22:44.630259 138120810325824 jax_array_handlers.py:360] Scheduling D2H of 39 prioritized jax.Array. I0418 18:22:44.630355 138120810325824 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 I0418 18:22:44.666225 138120810325824 base_pytree_checkpoint_handler.py:154] [process=5][thread=MainThread] Initiated "orbax.checkpoint._src.serialization.jax_array_handlers.ArrayHandler".serialize. Time taken: 0.036979s I0418 18:22:44.666381 138120810325824 base_pytree_checkpoint_handler.py:130] [process=5] /jax/orbax/write/blocking_gbytes_per_sec: 38.234 GiB/s (total gbytes: 1.5 GiB) (time elapsed: 0.04034543037414551 s) (per-host) I0418 18:22:44.666432 138120810325824 base_pytree_checkpoint_handler.py:768] [process=5][thread=MainThread] Initiated Pytree async_save. Time taken: 0.040405s (batch_requests_ready=0.001625s, total_serialization_initiated=0.038715s, others=0.000066s) I0418 18:22:44.666536 138120810325824 composite_checkpoint_handler.py:715] [process=5][thread=MainThread] Initiated CompositeCheckpointHandler.async_save. Time taken: 0.044580s (all_items=0.000015s, per_item={'items': '0.00001502'}, temp_paths=0.044565) I0418 18:22:44.667249 138120810325824 event_tracking.py:125] [process=5] [async] Finished blocking save in 1.18 seconds. Continuing save @ gs://lance-maxtext/linen_ckpt_xpk_main_20260418_180002/linen_xpk_main_20260418_180002_06_grad_accum/checkpoints/9. I0418 18:22:44.667558 137993352718080 async_checkpointer.py:76] [process=5][thread=async_save] Background save thread started. Deadline for this save operation is 2026-04-18 18:42:44.667525 I0418 18:22:44.672501 138120810325824 checkpoint_manager.py:1560] [process=5][thread=MainThread][step=9] Starting CheckpointManager Save Finalize thread=save_finalize I0418 18:22:44.672764 137996609795840 async_checkpointer.py:280] [process=5][thread=save_finalize] Waiting for background save thread=async_save. I0418 18:22:44.672898 138120810325824 standard_logger.py:34] {'step': 9, 'event_type': 'save', 'directory': 'gs://lance-maxtext/linen_ckpt_xpk_main_20260418_180002/linen_xpk_main_20260418_180002_06_grad_accum/checkpoints', 'reached_preemption': False, 'preemption_received_at': None, 'synchronous': False, 'wait_for_prev_start_time': 1776536563.4845798, 'wait_for_prev_duration_secs': 0.00012946128845214844, 'time_between_consecutive_saves_sec': 3.0349531173706055, 'checkpointer_blocking_start_time': 1776536563.4864004, 'checkpointer_blocking_duration_secs': 1.1813042163848877, 'get_old_steps_start_time': 1776536564.66773, 'get_old_steps_duration_secs': 3.170967102050781e-05, 'checkpoint_manager_blocking_start_time': 1776536563.4824953, 'checkpoint_manager_blocking_duration_secs': 1.1903660297393799} I0418 18:22:44.673012 138120810325824 checkpointing.py:408] Started an asynchronous checkpoint save for step 9 I0418 18:22:44.673057 138120810325824 checkpoint_manager.py:2020] [process=5][thread=MainThread][step=9][wait_until_finished] Waiting for Save Finalize thread (save_finalize) to complete. I0418 18:22:51.177536 137987450967808 array_metadata_store.py:203] [process=5][thread=array_type_handler] Wrote 39 array_metadata.ArrayMetadata to gs://lance-maxtext/linen_ckpt_xpk_main_20260418_180002/linen_xpk_main_20260418_180002_06_grad_accum/checkpoints/9/items/array_metadatas/process_5 I0418 18:23:26.417444 137993352718080 base_pytree_checkpoint_handler.py:130] [process=5] /jax/orbax/write/gbytes_per_sec: 37.797 MiB/s (total gbytes: 1.5 GiB) (time elapsed: 41.79137110710144 s) (per-host) I0418 18:23:26.417586 137993352718080 async_checkpointer.py:90] [process=5][thread=async_save] 3 Handler Commit operations completed. Time taken: 41.749918s. I0418 18:23:34.962885 137993352718080 async_checkpointer.py:160] [process=5][thread=async_save] Background save thread done. Time taken: 50.295202s. I0418 18:23:34.963245 137996609795840 async_checkpointer.py:288] [process=5][thread=save_finalize] Done with waiting for background save thread=async_save. I0418 18:23:34.963366 137996609795840 async_checkpointer.py:298] [process=5][thread=save_finalize] No errors found in background save thread=async_save. I0418 18:23:34.963411 137996609795840 checkpoint_manager.py:2137] [process=5][thread=save_finalize][step=9] CheckpointManager Save Finalize is syncing with other hosts... I0418 18:23:34.964733 137996609795840 checkpoint_manager.py:2146] [process=5][thread=save_finalize][step=9] CheckpointManager Save Finalize is done on all hosts. I0418 18:23:34.964928 138120810325824 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. I0418 18:23:34.965081 138120810325824 checkpoint_manager.py:2009] [process=5][thread=MainThread][wait_until_finished] No Save Finalize thread to wait for. Returning. I0418 18:23:34.965996 138120810325824 metric_logger.py:196] completed step: 9, seconds: 0.579, TFLOP/s/device: 93.923, Tokens/s/device: 14157.065, total_weights: 262144, loss: 9.457, lm_loss: 9.457, perplexity: 12802.546 Per train step: Total TFLOPs: 54.35 split as 93.93% learnable weight flops and 6.07% attention flops XPK End: Sat Apr 18 18:23:45 UTC 2026 EXIT_CODE=0
XPK Start: Sun Apr 19 09:10:03 UTC 2026 2026-04-19 09:10:07.919370: 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 00:00:1776589807.931927 10 cuda_dnn.cc:8579] Unable to register cuDNN factory: Attempting to register factory for plugin cuDNN when one has already been registered E0000 00:00:1776589807.935588 10 cuda_blas.cc:1407] Unable to register cuBLAS factory: Attempting to register factory for plugin cuBLAS when one has already been registered W0000 00:00:1776589807.946721 10 computation_placer.cc:177] computation placer already registered. Please check linkage and avoid linking the same target more than once. W0000 00:00:1776589807.946741 10 computation_placer.cc:177] computation placer already registered. Please check linkage and avoid linking the same target more than once. W0000 00:00:1776589807.946743 10 computation_placer.cc:177] computation placer already registered. Please check linkage and avoid linking the same target more than once. W0000 00:00:1776589807.946745 10 computation_placer.cc:177] computation placer already registered. Please check linkage and avoid linking the same target more than once. 2026-04-19 09:10:26.931917: 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) I0419 09:10:27.444483 133050411046720 max_utils.py:273] Attempting to initialize the jax distributed system... INFO:2026-04-19 09:10:36,484:jax._src.distributed:140: Starting JAX distributed service on [::]:8482 I0419 09:10:36.484633 133050411046720 distributed.py:140] Starting JAX distributed service on [::]:8482 INFO:2026-04-19 09:10:36,486:jax._src.distributed:157: Connecting to JAX distributed service on mt-06-grad-accum-gnmny-slice-job-0-0.mt-06-grad-accum-gnmny:8482 I0419 09:10:36.486987 133050411046720 distributed.py:157] Connecting to JAX distributed service on mt-06-grad-accum-gnmny-slice-job-0-0.mt-06-grad-accum-gnmny:8482 I0419 09:10:38.493108 133050411046720 max_utils.py:284] Jax distributed system initialized! I0419 09:10:45.367599 133050411046720 max_utils.py:800] System Information: Jax Version: 0.8.1 I0419 09:10:45.367706 133050411046720 max_utils.py:801] System Information: Jaxlib Version: 0.8.1 I0419 09:10:45.367747 133050411046720 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 I0419 09:10:45.367780 133050411046720 train_utils.py:347] WARNING: Sequence packing is essentially ignored for synthetic data. Please use a real dataset to use sequence packing. I0419 09:10:45.985780 133050411046720 maxtext_utils.py:1398] Num_devices: 32, shape (1, 1, 1, 32, 1, 1, 1, 1, 1, 1, 1, 1, 1) I0419 09:10:45.986069 133050411046720 checkpointing.py:677] Setting up checkpoint logger... I0419 09:10:45.986128 133050411046720 checkpointing.py:233] Creating checkpoint manager with ocdbt=True and zarr3=True I0419 09:10:45.986173 133050411046720 pytree_checkpoint_handler.py:589] save_device_host_concurrent_bytes=None I0419 09:10:45.986529 133050411046720 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 0x7901bc178fe0>, enable_pinned_host_transfer=False, save_concurrent_bytes: 96000000000 (89.4 GiB), restore_concurrent_bytes: 96000000000 (89.4 GiB) I0419 09:10:48.779722 133050411046720 checkpointing.py:265] Enabling policy for fixed interval checkpointing. I0419 09:10:48.779962 133050411046720 checkpoint_manager.py:709] [process=3][thread=MainThread] CheckpointManager init: checkpointers=None, item_names=('items',), item_handlers={'items': <orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeCheckpointHandler object at 0x78ef4cd39af0>}, handler_registry=None I0419 09:10:48.780205 133050411046720 composite_checkpoint_handler.py:237] Deferred registration for item: "items". Adding handler `<orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeCheckpointHandler object at 0x78ef4cd39af0>` 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`. I0419 09:10:48.780267 133050411046720 composite_checkpoint_handler.py:237] Deferred registration for item: "metrics". Adding handler `<orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonCheckpointHandler object at 0x78ef4cddbc20>` 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`. I0419 09:10:48.780305 133050411046720 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 0x78ef4cd39af0>, ('items', <class 'orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeRestoreArgs'>): <orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeCheckpointHandler object at 0x78ef4cd39af0>, ('metrics', <class 'orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonSaveArgs'>): <orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonCheckpointHandler object at 0x78ef4cddbc20>, ('metrics', <class 'orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonRestoreArgs'>): <orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonCheckpointHandler object at 0x78ef4cddbc20>}). I0419 09:10:48.783009 133050411046720 abstract_checkpointer.py:35] orbax-checkpoint version: 0.11.33 I0419 09:10:48.783099 133050411046720 async_checkpointer.py:177] [process=3][thread=MainThread] Using barrier_sync_fn: <function get_barrier_sync_fn.<locals>._fn at 0x7901a55cb2e0> timeout: 600 secs and primary_host=0 for async checkpoint writes I0419 09:10:50.830840 133050411046720 checkpoint_manager.py:1818] Found 0 checkpoint steps in gs://lance-maxtext/linen_ckpt_xpk_test_pipeline_scan_nnx_20260419_085454/linen_xpk_test_pipeline_scan_nnx_20260419_085454_06_grad_accum/checkpoints I0419 09:10:50.833059 133050411046720 checkpoint_manager.py:929] [process=3][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_test_pipeline_scan_nnx_20260419_085454/linen_xpk_test_pipeline_scan_nnx_20260419_085454_06_grad_accum/checkpoints: <orbax.checkpoint.checkpoint_manager.CheckpointManager object at 0x78ef4d6e3470> I0419 09:10:50.833180 133050411046720 checkpointing.py:301] Checkpoint manager created! I0419 09:10:52.503949 133050411046720 nnx_wrappers.py:453] Unknown Logical: bfloat16[32,2048,2048]...................................... ('activation_batch', 'activation_norm_length', 'activation_embed'). I0419 09:10:52.504060 133050411046720 nnx_wrappers.py:453] Unknown Physical: bfloat16[32,2048,2048]...................................... ('fsdp', None, None). I0419 09:10:52.875081 133050411046720 attentions.py:1084] attentions/inputs_q Logical: bfloat16[32,2048,2048]...................................... ('activation_batch', 'activation_attn_length', 'activation_attn_embed'). I0419 09:10:52.875179 133050411046720 attentions.py:1084] attentions/inputs_q Physical: bfloat16[32,2048,2048]...................................... ('fsdp', None, None). I0419 09:10:52.890921 133050411046720 attentions.py:1085] attentions/inputs_kv Logical: bfloat16[32,2048,2048]...................................... ('activation_batch', 'activation_attn_length', 'activation_attn_embed'). I0419 09:10:52.890984 133050411046720 attentions.py:1085] attentions/inputs_kv Physical: bfloat16[32,2048,2048]...................................... ('fsdp', None, None). I0419 09:10:52.914126 133050411046720 attentions.py:1150] attentions/query Logical: bfloat16[32,2048,16,128].................................... ('activation_kv_batch', 'activation_attn_length', 'activation_kv_heads', 'activation_kv_head_dim'). I0419 09:10:52.914207 133050411046720 attentions.py:1150] attentions/query Physical: bfloat16[32,2048,16,128].................................... ('fsdp', None, None, None). I0419 09:10:52.929939 133050411046720 attentions.py:1151] attentions/key Logical: bfloat16[32,2048,16,128].................................... ('activation_kv_batch', 'activation_attn_length', 'activation_kv_heads', 'activation_kv_head_dim'). I0419 09:10:52.930001 133050411046720 attentions.py:1151] attentions/key Physical: bfloat16[32,2048,16,128].................................... ('fsdp', None, None, None). I0419 09:10:52.945745 133050411046720 attentions.py:1152] attentions/value Logical: bfloat16[32,2048,16,128].................................... ('activation_kv_batch', 'activation_attn_length', 'activation_kv_heads', 'activation_kv_head_dim'). I0419 09:10:52.945807 133050411046720 attentions.py:1152] attentions/value Physical: bfloat16[32,2048,16,128].................................... ('fsdp', None, None, None). I0419 09:10:52.969717 133050411046720 attentions.py:1193] attentions/out Logical: bfloat16[32,2048,16,128].................................... ('activation_batch', 'activation_attn_length', 'activation_heads', 'activation_kv'). I0419 09:10:52.969787 133050411046720 attentions.py:1193] attentions/out Physical: bfloat16[32,2048,16,128].................................... ('fsdp', None, None, None). I0419 09:10:52.989588 133050411046720 linears.py:525] linears/x Logical: bfloat16[32,2048,7168]...................................... ('activation_batch', 'activation_length', 'activation_mlp'). I0419 09:10:52.989659 133050411046720 linears.py:525] linears/x Physical: bfloat16[32,2048,7168]...................................... ('fsdp', None, None). I0419 09:10:53.185221 133050411046720 checkpointing.py:577] checkpoint manager exists so trying to load this run's existing checkpoint I0419 09:10:53.185346 133050411046720 checkpointing.py:665] No existing checkpoints found, not restoring checkpoint. fsdp: 32 I0419 09:10:54.601497 133050411046720 maxtext_utils.py:1501] params/params/decoder/decoder_norm/scale Shape: float32[2048] Logical: PartitionSpec('norm',) Physical: (None,) I0419 09:10:54.601629 133050411046720 maxtext_utils.py:1501] params/params/decoder/layers/mlp/wi_0/kernel Shape: float32[2048,16,7168] Logical: PartitionSpec('embed', 'layers', 'mlp') Physical: ('fsdp', None, None) I0419 09:10:54.601683 133050411046720 maxtext_utils.py:1501] params/params/decoder/layers/mlp/wi_1/kernel Shape: float32[2048,16,7168] Logical: PartitionSpec('embed', 'layers', 'mlp') Physical: ('fsdp', None, None) I0419 09:10:54.601742 133050411046720 maxtext_utils.py:1501] params/params/decoder/layers/mlp/wo/kernel Shape: float32[7168,16,2048] Logical: PartitionSpec('mlp', 'layers', 'embed') Physical: (None, None, 'fsdp') I0419 09:10:54.601796 133050411046720 maxtext_utils.py:1501] params/params/decoder/layers/post_self_attention_layer_norm/scale Shape: float32[2048,16] Logical: PartitionSpec('norm', 'layers') Physical: (None, None) I0419 09:10:54.601838 133050411046720 maxtext_utils.py:1501] params/params/decoder/layers/pre_self_attention_layer_norm/scale Shape: float32[2048,16] Logical: PartitionSpec('norm', 'layers') Physical: (None, None) I0419 09:10:54.601892 133050411046720 maxtext_utils.py:1501] 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) I0419 09:10:54.601943 133050411046720 maxtext_utils.py:1501] 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') I0419 09:10:54.601982 133050411046720 maxtext_utils.py:1501] 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) I0419 09:10:54.602018 133050411046720 maxtext_utils.py:1501] 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) I0419 09:10:54.602069 133050411046720 maxtext_utils.py:1501] params/params/decoder/logits_dense/kernel Shape: float32[2048,32000] Logical: PartitionSpec('embed', 'vocab') Physical: ('fsdp', None) I0419 09:10:54.602119 133050411046720 maxtext_utils.py:1501] params/params/token_embedder/embedding Shape: float32[32000,2048] Logical: PartitionSpec('vocab', 'embed') Physical: (None, 'fsdp') I0419 09:10:54.626219 133050411046720 gradient_accumulation.py:68] gradient_accumulation/inputs Logical: float32[2048]............................................... Unknown. I0419 09:10:54.626300 133050411046720 gradient_accumulation.py:68] gradient_accumulation/inputs Physical: float32[2048]............................................... (None,). I0419 09:10:54.641289 133050411046720 gradient_accumulation.py:68] gradient_accumulation/inputs Logical: float32[2048,16,7168]....................................... Unknown. I0419 09:10:54.641349 133050411046720 gradient_accumulation.py:68] gradient_accumulation/inputs Physical: float32[2048,16,7168]....................................... ('fsdp', None, None). I0419 09:10:54.670871 133050411046720 gradient_accumulation.py:68] gradient_accumulation/inputs Logical: float32[7168,16,2048]....................................... Unknown. I0419 09:10:54.670947 133050411046720 gradient_accumulation.py:68] gradient_accumulation/inputs Physical: float32[7168,16,2048]....................................... (None, None, 'fsdp'). I0419 09:10:54.685804 133050411046720 gradient_accumulation.py:68] gradient_accumulation/inputs Logical: float32[2048,16]............................................ Unknown. I0419 09:10:54.685858 133050411046720 gradient_accumulation.py:68] gradient_accumulation/inputs Physical: float32[2048,16]............................................ (None, None). I0419 09:10:54.715555 133050411046720 gradient_accumulation.py:68] gradient_accumulation/inputs Logical: float32[2048,16,16,128]..................................... Unknown. I0419 09:10:54.715619 133050411046720 gradient_accumulation.py:68] gradient_accumulation/inputs Physical: float32[2048,16,16,128]..................................... ('fsdp', None, None, None). I0419 09:10:54.730535 133050411046720 gradient_accumulation.py:68] gradient_accumulation/inputs Logical: float32[16,16,128,2048]..................................... Unknown. I0419 09:10:54.730591 133050411046720 gradient_accumulation.py:68] gradient_accumulation/inputs Physical: float32[16,16,128,2048]..................................... (None, None, None, 'fsdp'). I0419 09:10:54.775082 133050411046720 gradient_accumulation.py:68] gradient_accumulation/inputs Logical: float32[2048,32000]......................................... Unknown. I0419 09:10:54.775156 133050411046720 gradient_accumulation.py:68] gradient_accumulation/inputs Physical: float32[2048,32000]......................................... ('fsdp', None). I0419 09:10:54.789942 133050411046720 gradient_accumulation.py:68] gradient_accumulation/inputs Logical: float32[32000,2048]......................................... Unknown. I0419 09:10:54.789994 133050411046720 gradient_accumulation.py:68] gradient_accumulation/inputs Physical: float32[32000,2048]......................................... (None, 'fsdp'). I0419 09:10:55.460222 133050411046720 train.py:156] train/xent Logical: float32[32,2048]............................................ ('activation_embed_and_logits_batch', 'activation_length'). I0419 09:10:55.460336 133050411046720 train.py:156] train/xent Physical: float32[32,2048]............................................ ('fsdp', None). I0419 09:10:55.475984 133050411046720 train.py:163] train/z_loss Logical: float32[32,2048]............................................ ('activation_embed_and_logits_batch', 'activation_length'). I0419 09:10:55.476046 133050411046720 train.py:163] train/z_loss Physical: float32[32,2048]............................................ ('fsdp', None). I0419 09:11:06.609174 133050411046720 max_utils.py:791] Total memory size: 1.7 GB, Output size: 0.4 GB, Temp size: 1.3 GB, Argument size: 0.4 GB, Host temp size: 0.0 GB. I0419 09:11:06.610066 133050411046720 metric_logger.py:289] number parameters: 1.104 billion I0419 09:11:08.730699 133050411046720 checkpointing.py:772] Waiting for step 0 to finish before checkpoint... I0419 09:11:19.148830 133050411046720 checkpointing.py:776] Waited 10.418111085891724 seconds for step 0 to finish before starting checkpointing. I0419 09:11:19.151407 133050411046720 checkpoint_manager.py:2013] [process=3][thread=MainThread][wait_until_finished] No Save Finalize thread to wait for. Returning. I0419 09:11:19.153312 133050411046720 checkpoint_manager.py:1518] [process=3] Saving checkpoint at step 0 I0419 09:11:19.156128 133050411046720 async_checkpointer.py:452] [process=3] Started async saving checkpoint to gs://lance-maxtext/linen_ckpt_xpk_test_pipeline_scan_nnx_20260419_085454/linen_xpk_test_pipeline_scan_nnx_20260419_085454_06_grad_accum/checkpoints/0. I0419 09:11:20.143093 133050411046720 signaling_client.py:364] Using JaxDistributedSignalingClient I0419 09:11:20.144207 133050411046720 jax_array_handlers.py:358] Scheduling D2H of 39 prioritized jax.Array. I0419 09:11:20.144277 133050411046720 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 I0419 09:11:20.387936 133050411046720 base_pytree_checkpoint_handler.py:153] [process=3][thread=MainThread] Initiated "orbax.checkpoint._src.serialization.jax_array_handlers.ArrayHandler".serialize. Time taken: 0.244928s I0419 09:11:20.388120 133050411046720 base_pytree_checkpoint_handler.py:129] [process=3] /jax/checkpoint/write/blocking_gbytes_per_sec: 3.042 GiB/s (total gbytes: 1.5 GiB) (time elapsed: 0.5071663856506348 s) (per-host) I0419 09:11:20.388174 133050411046720 base_pytree_checkpoint_handler.py:737] [process=3][thread=MainThread] Initiated Pytree async_save. Time taken: 0.507231s (batch_requests_ready=0.258724s, total_serialization_initiated=0.248434s, others=0.000072s) I0419 09:11:20.388282 133050411046720 composite_checkpoint_handler.py:715] [process=3][thread=MainThread] Initiated CompositeCheckpointHandler.async_save. Time taken: 0.511383s (all_items=0.000021s, per_item={'items': '0.00002050'}, temp_paths=0.511363) I0419 09:11:20.389229 132928113710848 async_checkpointer.py:79] [process=3][thread=async_save] Background save thread started. I0419 09:11:20.389405 133050411046720 async_checkpointer.py:561] Finished blocking save. Time taken: 1.236021s. Continuing background save to gs://lance-maxtext/linen_ckpt_xpk_test_pipeline_scan_nnx_20260419_085454/linen_xpk_test_pipeline_scan_nnx_20260419_085454_06_grad_accum/checkpoints/0. I0419 09:11:20.839283 133050411046720 checkpoint_manager.py:1566] [process=3][thread=MainThread][step=0] Starting CheckpointManager Save Finalize thread=save_finalize I0419 09:11:20.839658 132924926056192 async_checkpointer.py:265] [process=3][thread=save_finalize] Waiting for background save thread=async_save. I0419 09:11:20.839833 133050411046720 standard_logger.py:34] {'step': 0, 'event_type': 'save', 'directory': 'gs://lance-maxtext/linen_ckpt_xpk_test_pipeline_scan_nnx_20260419_085454/linen_xpk_test_pipeline_scan_nnx_20260419_085454_06_grad_accum/checkpoints', 'reached_preemption': False, 'preemption_received_at': None, 'synchronous': False, 'wait_for_prev_start_time': 1776589879.1513882, 'wait_for_prev_duration_secs': 6.556510925292969e-05, 'time_between_consecutive_saves_sec': None, 'checkpointer_blocking_start_time': 1776589879.153351, 'checkpointer_blocking_duration_secs': 1.2361598014831543, 'get_old_steps_start_time': 1776589880.389532, 'get_old_steps_duration_secs': 2.574920654296875e-05, 'checkpoint_manager_blocking_start_time': 1776589879.149371, 'checkpoint_manager_blocking_duration_secs': 1.6904137134552002} I0419 09:11:20.839940 133050411046720 checkpointing.py:408] Started an asynchronous checkpoint save for step 0 I0419 09:11:20.839992 133050411046720 max_utils.py:750] Memstats: After params initialized: I0419 09:11:20.840048 133050411046720 max_utils.py:756] Using (GB) 0.43 / 31.25 (1.376000%) on TPU_10(process=3,(2,2,0,0)) I0419 09:11:20.840084 133050411046720 max_utils.py:756] Using (GB) 0.43 / 31.25 (1.376000%) on TPU_11(process=3,(3,2,0,0)) I0419 09:11:20.840113 133050411046720 max_utils.py:756] Using (GB) 0.43 / 31.25 (1.376000%) on TPU_14(process=3,(2,3,0,0)) I0419 09:11:20.840139 133050411046720 max_utils.py:756] Using (GB) 0.43 / 31.25 (1.376000%) on TPU_15(process=3,(3,3,0,0)) I0419 09:11:21.105098 133050411046720 metric_logger.py:185] completed step: 0, seconds: 2.121, TFLOP/s/device: 25.630, Tokens/s/device: 3863.207, total_weights: 262144, loss: 10.877 I0419 09:11:21.714163 133050411046720 metric_logger.py:185] completed step: 1, seconds: 12.373, TFLOP/s/device: 4.392, Tokens/s/device: 662.082, total_weights: 262144, loss: 10.877 I0419 09:11:22.293525 133050411046720 metric_logger.py:185] completed step: 2, seconds: 0.030, TFLOP/s/device: 1803.023, Tokens/s/device: 271771.224, total_weights: 262144, loss: 10.563 I0419 09:11:22.872450 133050411046720 metric_logger.py:185] completed step: 3, seconds: 0.585, TFLOP/s/device: 92.939, Tokens/s/device: 14008.783, total_weights: 262144, loss: 10.272 I0419 09:11:23.056341 2420 google_auth_provider.cc:181] Running on GCE, using service account 562977990677-compute@developer.gserviceaccount.com I0419 09:11:24.030492 133050411046720 metric_logger.py:185] completed step: 4, seconds: 0.579, TFLOP/s/device: 93.812, Tokens/s/device: 14140.326, total_weights: 262144, loss: 10.022 I0419 09:11:24.036227 133050411046720 metric_logger.py:185] completed step: 5, seconds: 0.579, TFLOP/s/device: 93.888, Tokens/s/device: 14151.783, total_weights: 262144, loss: 9.820 I0419 09:11:24.954680 132928122103552 array_metadata_store.py:203] [process=3][thread=array_type_handler] Wrote 39 array_metadata.ArrayMetadata to gs://lance-maxtext/linen_ckpt_xpk_test_pipeline_scan_nnx_20260419_085454/linen_xpk_test_pipeline_scan_nnx_20260419_085454_06_grad_accum/checkpoints/0/items/array_metadatas/process_3 I0419 09:11:58.083917 132928113710848 base_pytree_checkpoint_handler.py:129] [process=3] /jax/checkpoint/write/gbytes_per_sec: 41.348 MiB/s (total gbytes: 1.5 GiB) (time elapsed: 38.202919244766235 s) (per-host) I0419 09:11:58.084025 132928113710848 async_checkpointer.py:90] [process=3][thread=async_save] 3 Handler Commit operations completed. Time taken: 37.694674s. I0419 09:12:02.558277 133050411046720 metric_logger.py:185] completed step: 6, seconds: 1.159, TFLOP/s/device: 46.911, Tokens/s/device: 7070.926, total_weights: 262144, loss: 9.667 I0419 09:12:03.137341 133050411046720 metric_logger.py:185] completed step: 7, seconds: 37.943, TFLOP/s/device: 1.432, Tokens/s/device: 215.900, total_weights: 262144, loss: 9.561 I0419 09:12:03.716247 133050411046720 metric_logger.py:185] completed step: 8, seconds: 0.584, TFLOP/s/device: 93.092, Tokens/s/device: 14031.866, total_weights: 262144, loss: 9.496 I0419 09:12:04.295205 133050411046720 checkpointing.py:772] Waiting for step 9 to finish before checkpoint... I0419 09:12:04.295878 133050411046720 checkpointing.py:776] Waited 0.0006918907165527344 seconds for step 9 to finish before starting checkpointing. I0419 09:12:04.297916 133050411046720 checkpoint_manager.py:2024] [process=3][thread=MainThread][step=0][wait_until_finished] Waiting for Save Finalize thread (save_finalize) to complete. I0419 09:12:41.528071 132928113710848 async_checkpointer.py:144] [process=3][thread=async_save] Background save thread done. Time taken: 81.138704s. I0419 09:12:41.528390 132924926056192 async_checkpointer.py:273] [process=3][thread=save_finalize] Done with waiting for background save thread=async_save. I0419 09:12:41.528517 132924926056192 async_checkpointer.py:283] [process=3][thread=save_finalize] No errors found in background save thread=async_save. I0419 09:12:41.528587 132924926056192 checkpoint_manager.py:2133] [process=3][thread=save_finalize][step=0] CheckpointManager Save Finalize is syncing with other hosts... I0419 09:12:41.530060 132924926056192 checkpoint_manager.py:2142] [process=3][thread=save_finalize][step=0] CheckpointManager Save Finalize is done on all hosts. I0419 09:12:41.530221 133050411046720 checkpoint_manager.py:2036] [process=3][thread=MainThread][step=0][wait_until_finished] Done waiting for Save Finalize thread (save_finalize) running at step=0. W0419 09:12:41.530373 133050411046720 checkpoint_manager.py:1458] Waiting for previous save to complete took 37.232463 seconds. If this number is high, consider checkpointing less frequently. I0419 09:12:41.533805 133050411046720 checkpoint_manager.py:1518] [process=3] Saving checkpoint at step 9 I0419 09:12:41.537366 133050411046720 async_checkpointer.py:452] [process=3] Started async saving checkpoint to gs://lance-maxtext/linen_ckpt_xpk_test_pipeline_scan_nnx_20260419_085454/linen_xpk_test_pipeline_scan_nnx_20260419_085454_06_grad_accum/checkpoints/9. I0419 09:12:42.950083 133050411046720 jax_array_handlers.py:358] Scheduling D2H of 39 prioritized jax.Array. I0419 09:12:42.950193 133050411046720 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 I0419 09:12:42.995312 133050411046720 base_pytree_checkpoint_handler.py:153] [process=3][thread=MainThread] Initiated "orbax.checkpoint._src.serialization.jax_array_handlers.ArrayHandler".serialize. Time taken: 0.046357s I0419 09:12:42.995483 133050411046720 base_pytree_checkpoint_handler.py:129] [process=3] /jax/checkpoint/write/blocking_gbytes_per_sec: 2.118 GiB/s (total gbytes: 1.5 GiB) (time elapsed: 0.7284677028656006 s) (per-host) I0419 09:12:42.995534 133050411046720 base_pytree_checkpoint_handler.py:737] [process=3][thread=MainThread] Initiated Pytree async_save. Time taken: 0.728528s (batch_requests_ready=0.680230s, total_serialization_initiated=0.048229s, others=0.000069s) I0419 09:12:42.995623 133050411046720 composite_checkpoint_handler.py:715] [process=3][thread=MainThread] Initiated CompositeCheckpointHandler.async_save. Time taken: 0.732668s (all_items=0.000018s, per_item={'items': '0.00001788'}, temp_paths=0.732650) I0419 09:12:42.996500 132924405970688 async_checkpointer.py:79] [process=3][thread=async_save] Background save thread started. I0419 09:12:42.996656 133050411046720 async_checkpointer.py:561] Finished blocking save. Time taken: 1.462781s. Continuing background save to gs://lance-maxtext/linen_ckpt_xpk_test_pipeline_scan_nnx_20260419_085454/linen_xpk_test_pipeline_scan_nnx_20260419_085454_06_grad_accum/checkpoints/9. I0419 09:12:43.003718 133050411046720 checkpoint_manager.py:1566] [process=3][thread=MainThread][step=9] Starting CheckpointManager Save Finalize thread=save_finalize I0419 09:12:43.004012 132924926056192 async_checkpointer.py:265] [process=3][thread=save_finalize] Waiting for background save thread=async_save. I0419 09:12:43.004178 133050411046720 standard_logger.py:34] {'step': 9, 'event_type': 'save', 'directory': 'gs://lance-maxtext/linen_ckpt_xpk_test_pipeline_scan_nnx_20260419_085454/linen_xpk_test_pipeline_scan_nnx_20260419_085454_06_grad_accum/checkpoints', 'reached_preemption': False, 'preemption_received_at': None, 'synchronous': False, 'wait_for_prev_start_time': 1776589924.2978861, 'wait_for_prev_duration_secs': 37.232463359832764, 'time_between_consecutive_saves_sec': None, 'checkpointer_blocking_start_time': 1776589961.533847, 'checkpointer_blocking_duration_secs': 1.4629147052764893, 'get_old_steps_start_time': 1776589962.9967794, 'get_old_steps_duration_secs': 2.47955322265625e-05, 'checkpoint_manager_blocking_start_time': 1776589924.296101, 'checkpoint_manager_blocking_duration_secs': 38.70804452896118} I0419 09:12:43.004297 133050411046720 checkpointing.py:408] Started an asynchronous checkpoint save for step 9 I0419 09:12:43.004343 133050411046720 checkpoint_manager.py:2024] [process=3][thread=MainThread][step=9][wait_until_finished] Waiting for Save Finalize thread (save_finalize) to complete. I0419 09:12:47.869315 132928122103552 array_metadata_store.py:203] [process=3][thread=array_type_handler] Wrote 39 array_metadata.ArrayMetadata to gs://lance-maxtext/linen_ckpt_xpk_test_pipeline_scan_nnx_20260419_085454/linen_xpk_test_pipeline_scan_nnx_20260419_085454_06_grad_accum/checkpoints/9/items/array_metadatas/process_3 I0419 09:13:24.366936 132924405970688 base_pytree_checkpoint_handler.py:129] [process=3] /jax/checkpoint/write/gbytes_per_sec: 37.520 MiB/s (total gbytes: 1.5 GiB) (time elapsed: 42.09987950325012 s) (per-host) I0419 09:13:24.367058 132924405970688 async_checkpointer.py:90] [process=3][thread=async_save] 3 Handler Commit operations completed. Time taken: 41.370446s. I0419 09:13:30.196184 132924405970688 async_checkpointer.py:144] [process=3][thread=async_save] Background save thread done. Time taken: 47.199557s. I0419 09:13:30.196527 132924926056192 async_checkpointer.py:273] [process=3][thread=save_finalize] Done with waiting for background save thread=async_save. I0419 09:13:30.196650 132924926056192 async_checkpointer.py:283] [process=3][thread=save_finalize] No errors found in background save thread=async_save. I0419 09:13:30.196698 132924926056192 checkpoint_manager.py:2133] [process=3][thread=save_finalize][step=9] CheckpointManager Save Finalize is syncing with other hosts... I0419 09:13:30.198293 132924926056192 checkpoint_manager.py:2142] [process=3][thread=save_finalize][step=9] CheckpointManager Save Finalize is done on all hosts. I0419 09:13:30.198479 133050411046720 checkpoint_manager.py:2036] [process=3][thread=MainThread][step=9][wait_until_finished] Done waiting for Save Finalize thread (save_finalize) running at step=9. I0419 09:13:30.198621 133050411046720 checkpoint_manager.py:2013] [process=3][thread=MainThread][wait_until_finished] No Save Finalize thread to wait for. Returning. I0419 09:13:30.199327 133050411046720 metric_logger.py:185] completed step: 9, seconds: 0.579, TFLOP/s/device: 93.861, Tokens/s/device: 14147.677, total_weights: 262144, loss: 9.457 Per train step: Total TFLOPs: 54.35 split as 93.93% learnable weight flops and 6.07% attention flops XPK End: Sun Apr 19 09:13:39 UTC 2026 EXIT_CODE=0