main (59e0f1759) vs feat/Migrate-Decoder-And-Tests-to-NNX (4f4617640)| Metric | main 59e0f1759 | feat/Migrate-Decoder-And-Tests-to-NNX 4f4617640 | Diff (feat/Migrate-Decoder-And-Tests-to-NNX − main) |
|---|---|---|---|
| Parameters | 1.104 billion | 1.104 billion | — |
| Final loss | 8.1750 | 8.1750 | 0 |
| TFLOP/s | 59.249 | 59.125 | -0.124 |
| Tok/s | 8930.7 | 8911.9 | -18.77 |
| Avg s/step | 3.613 | 2.898 | -0.715 |
| 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 07:21:16 UTC 2026 PyTorch was not found. Models won't be available and only tokenizers, configuration and file/data utilities can be used. `rope_parameters`'s factor field must be a float >= 1, got 40 `rope_parameters`'s beta_fast field must be a float, got 32 `rope_parameters`'s beta_slow field must be a float, got 1 DeepseekV32Config got `key=rope_scaling` in kwargs but hasn't set it as attribute. For RoPE standardization you need to set `self.rope_parameters` in model's config. 2026-04-25 07:21:41.869574: 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 07:21:42.079255 135965884618560 max_utils.py:273] Attempting to initialize the jax distributed system... I0425 07:21:51.119862 135965884618560 distributed.py:149] Starting JAX distributed service on [::]:8482 I0425 07:21:51.122118 135965884618560 distributed.py:172] Connecting to JAX distributed service on mt-04-int8-tfv0h-slice-job-0-0.mt-04-int8-tfv0h:8482 I0425 07:21:51.944138 135965884618560 max_utils.py:284] Jax distributed system initialized! I0425 07:21:57.985152 135965884618560 max_utils.py:800] System Information: Jax Version: 0.9.2 I0425 07:21:57.985256 135965884618560 max_utils.py:801] System Information: Jaxlib Version: 0.9.2 I0425 07:21:57.985296 135965884618560 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 07:21:57.985332 135965884618560 train_utils.py:361] WARNING: Sequence packing is essentially ignored for synthetic data. Please use a real dataset to use sequence packing. I0425 07:21:58.676079 135965884618560 maxtext_utils.py:1604] Num_devices: 32, shape (1, 1, 1, 32, 1, 1, 1, 1, 1, 1, 1, 1, 1) I0425 07:21:58.676470 135965884618560 checkpointing.py:677] Setting up checkpoint logger... I0425 07:21:58.676526 135965884618560 checkpointing.py:233] Creating checkpoint manager with ocdbt=True and zarr3=True I0425 07:21:58.676569 135965884618560 pytree_checkpoint_handler.py:592] save_device_host_concurrent_bytes=None I0425 07:21:58.676913 135965884618560 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 0x7ba871f1ede0>, enable_pinned_host_transfer=False, save_concurrent_bytes: 96000000000 (89.4 GiB), restore_concurrent_bytes: 96000000000 (89.4 GiB) I0425 07:22:01.486215 135965884618560 checkpointing.py:265] Enabling policy for fixed interval checkpointing. I0425 07:22:01.486443 135965884618560 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 0x7ba74308a300>}, handler_registry=None I0425 07:22:01.486701 135965884618560 composite_checkpoint_handler.py:237] Deferred registration for item: "items". Adding handler `<orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeCheckpointHandler object at 0x7ba74308a300>` 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 07:22:01.486752 135965884618560 composite_checkpoint_handler.py:237] Deferred registration for item: "metrics". Adding handler `<orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonCheckpointHandler object at 0x7b946413c560>` 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 07:22:01.486789 135965884618560 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 0x7ba74308a300>, ('items', <class 'orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeRestoreArgs'>): <orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeCheckpointHandler object at 0x7ba74308a300>, ('metrics', <class 'orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonSaveArgs'>): <orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonCheckpointHandler object at 0x7b946413c560>, ('metrics', <class 'orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonRestoreArgs'>): <orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonCheckpointHandler object at 0x7b946413c560>}). I0425 07:22:01.487118 135965884618560 abstract_checkpointer.py:35] orbax-checkpoint version: 0.11.34 I0425 07:22:01.487190 135965884618560 async_checkpointer.py:192] [process=5][thread=MainThread] Using barrier_sync_fn: <function get_barrier_sync_fn.<locals>._fn at 0x7b94445d1760> timeout: 1200 secs and primary_host=0 for async checkpoint writes I0425 07:22:02.648800 135965884618560 checkpoint_manager.py:1812] Found 0 checkpoint steps in gs://lance-maxtext/linen_ckpt_xpk_main_20260425_071456/linen_xpk_main_20260425_071456_04_int8/checkpoints I0425 07:22:02.688567 135965884618560 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_20260425_071456/linen_xpk_main_20260425_071456_04_int8/checkpoints: <orbax.checkpoint.checkpoint_manager.CheckpointManager object at 0x7b9464135a60> I0425 07:22:02.688736 135965884618560 checkpointing.py:301] Checkpoint manager created! I0425 07:22:03.830742 135965884618560 nnx_wrappers.py:437] Unknown Logical: bfloat16[32,2048,2048]...................................... ('activation_batch', 'activation_norm_length', 'activation_embed'). I0425 07:22:03.830857 135965884618560 nnx_wrappers.py:437] Unknown Physical: bfloat16[32,2048,2048]...................................... ('fsdp', None, None). I0425 07:22:04.218037 135965884618560 attentions.py:1088] attentions/inputs_q Logical: bfloat16[32,2048,2048]...................................... ('activation_batch', 'activation_attn_length', 'activation_attn_embed'). I0425 07:22:04.218132 135965884618560 attentions.py:1088] attentions/inputs_q Physical: bfloat16[32,2048,2048]...................................... ('fsdp', None, None). I0425 07:22:04.234637 135965884618560 attentions.py:1089] attentions/inputs_kv Logical: bfloat16[32,2048,2048]...................................... ('activation_batch', 'activation_attn_length', 'activation_attn_embed'). I0425 07:22:04.234702 135965884618560 attentions.py:1089] attentions/inputs_kv Physical: bfloat16[32,2048,2048]...................................... ('fsdp', None, None). I0425 07:22:04.300613 135965884618560 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 07:22:04.300716 135965884618560 attentions.py:1154] attentions/query Physical: bfloat16[32,2048,16,128].................................... ('fsdp', None, None, None). I0425 07:22:04.317263 135965884618560 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 07:22:04.317322 135965884618560 attentions.py:1155] attentions/key Physical: bfloat16[32,2048,16,128].................................... ('fsdp', None, None, None). I0425 07:22:04.333912 135965884618560 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 07:22:04.333976 135965884618560 attentions.py:1156] attentions/value Physical: bfloat16[32,2048,16,128].................................... ('fsdp', None, None, None). I0425 07:22:04.358685 135965884618560 attentions.py:1198] attentions/out Logical: bfloat16[32,2048,16,128].................................... ('activation_batch', 'activation_attn_length', 'activation_heads', 'activation_kv'). I0425 07:22:04.358756 135965884618560 attentions.py:1198] attentions/out Physical: bfloat16[32,2048,16,128].................................... ('fsdp', None, None, None). I0425 07:22:04.424959 135965884618560 linears.py:525] linears/x Logical: bfloat16[32,2048,7168]...................................... ('activation_batch', 'activation_length', 'activation_mlp'). I0425 07:22:04.425045 135965884618560 linears.py:525] linears/x Physical: bfloat16[32,2048,7168]...................................... ('fsdp', None, None). I0425 07:22:04.932487 135965884618560 checkpointing.py:577] checkpoint manager exists so trying to load this run's existing checkpoint I0425 07:22:04.932595 135965884618560 checkpointing.py:665] No existing checkpoints found, not restoring checkpoint. fsdp: 32 I0425 07:22:07.426491 135965884618560 maxtext_utils.py:1707] params/params/decoder/decoder_norm/scale Shape: float32[2048] Logical: P('norm',) Physical: (None,) I0425 07:22:07.426615 135965884618560 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 07:22:07.426681 135965884618560 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 07:22:07.426740 135965884618560 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 07:22:07.426794 135965884618560 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 07:22:07.426833 135965884618560 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 07:22:07.426887 135965884618560 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 07:22:07.426943 135965884618560 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 07:22:07.426985 135965884618560 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 07:22:07.427020 135965884618560 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 07:22:07.427066 135965884618560 maxtext_utils.py:1707] params/params/decoder/logits_dense/kernel Shape: float32[2048,32000] Logical: P('embed_vocab', 'vocab') Physical: ('fsdp', None) I0425 07:22:07.427113 135965884618560 maxtext_utils.py:1707] params/params/token_embedder/embedding Shape: float32[32000,2048] Logical: P('vocab', 'embed_vocab') Physical: (None, 'fsdp') I0425 07:22:08.817587 135965884618560 train.py:155] train/xent Logical: float32[32,2048]............................................ ('activation_embed_and_logits_batch', 'activation_length'). I0425 07:22:08.817697 135965884618560 train.py:155] train/xent Physical: float32[32,2048]............................................ ('fsdp', None). I0425 07:22:08.834160 135965884618560 train.py:162] train/z_loss Logical: float32[32,2048]............................................ ('activation_embed_and_logits_batch', 'activation_length'). I0425 07:22:08.834223 135965884618560 train.py:162] train/z_loss Physical: float32[32,2048]............................................ ('fsdp', None). I0425 07:22:22.341513 135965884618560 max_utils.py:791] Total memory size: 1.8 GB, Output size: 0.4 GB, Temp size: 1.4 GB, Argument size: 0.4 GB, Host temp size: 0.0 GB. I0425 07:22:22.342331 135965884618560 metric_logger.py:301] number parameters: 1.104 billion I0425 07:22:37.212856 135965884618560 checkpointing.py:772] Waiting for step 0 to finish before checkpoint... I0425 07:22:37.453130 135965884618560 checkpointing.py:776] Waited 0.2402477264404297 seconds for step 0 to finish before starting checkpointing. I0425 07:22:37.455894 135965884618560 checkpoint_manager.py:2009] [process=5][thread=MainThread][wait_until_finished] No Save Finalize thread to wait for. Returning. I0425 07:22:37.459030 135965884618560 checkpoint_manager.py:1512] [process=5] Saving checkpoint at step 0 I0425 07:22:37.460455 135965884618560 event_tracking.py:70] [process=5] [async] Started save checkpoint @ gs://lance-maxtext/linen_ckpt_xpk_main_20260425_071456/linen_xpk_main_20260425_071456_04_int8/checkpoints/0. I0425 07:22:37.786480 135965884618560 signaling_client.py:364] Using JaxDistributedSignalingClient I0425 07:22:37.787490 135965884618560 jax_array_handlers.py:360] Scheduling D2H of 39 prioritized jax.Array. I0425 07:22:37.787547 135965884618560 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 07:22:38.066318 135965884618560 base_pytree_checkpoint_handler.py:154] [process=5][thread=MainThread] Initiated "orbax.checkpoint._src.serialization.jax_array_handlers.ArrayHandler".serialize. Time taken: 0.279913s I0425 07:22:38.066492 135965884618560 base_pytree_checkpoint_handler.py:130] [process=5] /jax/orbax/write/blocking_gbytes_per_sec: 5.406 GiB/s (total gbytes: 1.5 GiB) (time elapsed: 0.28536128997802734 s) (per-host) I0425 07:22:38.066548 135965884618560 base_pytree_checkpoint_handler.py:768] [process=5][thread=MainThread] Initiated Pytree async_save. Time taken: 0.285427s (batch_requests_ready=0.002229s, total_serialization_initiated=0.283124s, others=0.000073s) I0425 07:22:38.066646 135965884618560 composite_checkpoint_handler.py:715] [process=5][thread=MainThread] Initiated CompositeCheckpointHandler.async_save. Time taken: 0.289515s (all_items=0.000019s, per_item={'items': '0.00001860'}, temp_paths=0.289496) I0425 07:22:38.067538 135965884618560 event_tracking.py:125] [process=5] [async] Finished blocking save in 0.61 seconds. Continuing save @ gs://lance-maxtext/linen_ckpt_xpk_main_20260425_071456/linen_xpk_main_20260425_071456_04_int8/checkpoints/0. I0425 07:22:38.067868 135833875052288 async_checkpointer.py:76] [process=5][thread=async_save] Background save thread started. Deadline for this save operation is 2026-04-25 07:42:38.067830 I0425 07:22:38.078439 135965884618560 checkpoint_manager.py:1560] [process=5][thread=MainThread][step=0] Starting CheckpointManager Save Finalize thread=save_finalize I0425 07:22:38.078732 135831321896704 async_checkpointer.py:280] [process=5][thread=save_finalize] Waiting for background save thread=async_save. I0425 07:22:38.078892 135965884618560 standard_logger.py:34] {'step': 0, 'event_type': 'save', 'directory': 'gs://lance-maxtext/linen_ckpt_xpk_main_20260425_071456/linen_xpk_main_20260425_071456_04_int8/checkpoints', 'reached_preemption': False, 'preemption_received_at': None, 'synchronous': False, 'wait_for_prev_start_time': 1777101757.455877, 'wait_for_prev_duration_secs': 6.031990051269531e-05, 'time_between_consecutive_saves_sec': None, 'checkpointer_blocking_start_time': 1777101757.4590714, 'checkpointer_blocking_duration_secs': 0.6089563369750977, 'get_old_steps_start_time': 1777101758.0680528, 'get_old_steps_duration_secs': 3.2901763916015625e-05, 'checkpoint_manager_blocking_start_time': 1777101757.453697, 'checkpoint_manager_blocking_duration_secs': 0.6251547336578369} I0425 07:22:38.079006 135965884618560 checkpointing.py:408] Started an asynchronous checkpoint save for step 0 I0425 07:22:38.079055 135965884618560 max_utils.py:750] Memstats: After params initialized: I0425 07:22:38.079104 135965884618560 max_utils.py:756] Using (GB) 0.44 / 31.25 (1.408000%) on TPU_18(process=5,(2,4,0,0)) I0425 07:22:38.079138 135965884618560 max_utils.py:756] Using (GB) 0.43 / 31.25 (1.376000%) on TPU_19(process=5,(3,4,0,0)) I0425 07:22:38.079165 135965884618560 max_utils.py:756] Using (GB) 0.43 / 31.25 (1.376000%) on TPU_22(process=5,(2,5,0,0)) I0425 07:22:38.079191 135965884618560 max_utils.py:756] Using (GB) 0.43 / 31.25 (1.376000%) on TPU_23(process=5,(3,5,0,0)) I0425 07:22:38.393681 135965884618560 metric_logger.py:196] completed step: 0, seconds: 14.870, TFLOP/s/device: 0.914, Tokens/s/device: 137.723, total_weights: 65536, loss: 10.874, lm_loss: 10.874, perplexity: 52796.277 I0425 07:22:38.664057 135965884618560 metric_logger.py:196] completed step: 1, seconds: 1.179, TFLOP/s/device: 11.522, Tokens/s/device: 1736.650, total_weights: 65536, loss: 10.874, lm_loss: 10.874, perplexity: 52796.277 I0425 07:22:39.081716 135965884618560 metric_logger.py:196] completed step: 2, seconds: 0.042, TFLOP/s/device: 326.786, Tokens/s/device: 49256.819, total_weights: 65536, loss: 10.262, lm_loss: 10.262, perplexity: 28634.820 I0425 07:22:39.311253 135965884618560 metric_logger.py:196] completed step: 3, seconds: 0.418, TFLOP/s/device: 32.511, Tokens/s/device: 4900.459, total_weights: 65536, loss: 9.731, lm_loss: 9.731, perplexity: 16827.199 I0425 07:22:39.775644 135965884618560 metric_logger.py:196] completed step: 4, seconds: 0.235, TFLOP/s/device: 57.867, Tokens/s/device: 8722.317, total_weights: 65536, loss: 9.272, lm_loss: 9.272, perplexity: 10638.531 I0425 07:22:39.781880 135965884618560 metric_logger.py:196] completed step: 5, seconds: 0.230, TFLOP/s/device: 59.183, Tokens/s/device: 8920.677, total_weights: 65536, loss: 8.887, lm_loss: 8.887, perplexity: 7233.941 I0425 07:22:41.303051 2925 google_auth_provider.cc:181] Running on GCE, using service account 562977990677-compute@developer.gserviceaccount.com I0425 07:22:43.365815 135832799807232 array_metadata_store.py:203] [process=5][thread=array_type_handler] Wrote 39 array_metadata.ArrayMetadata to gs://lance-maxtext/linen_ckpt_xpk_main_20260425_071456/linen_xpk_main_20260425_071456_04_int8/checkpoints/0/items/array_metadatas/process_5 I0425 07:23:09.498006 135965884618560 metric_logger.py:196] completed step: 6, seconds: 0.465, TFLOP/s/device: 29.228, Tokens/s/device: 4405.628, total_weights: 65536, loss: 8.588, lm_loss: 8.588, perplexity: 5365.468 I0425 07:23:09.727385 135965884618560 metric_logger.py:196] completed step: 7, seconds: 29.488, TFLOP/s/device: 0.461, Tokens/s/device: 69.452, total_weights: 65536, loss: 8.380, lm_loss: 8.380, perplexity: 4358.779 I0425 07:23:09.956782 135965884618560 metric_logger.py:196] completed step: 8, seconds: 0.234, TFLOP/s/device: 58.054, Tokens/s/device: 8750.566, total_weights: 65536, loss: 8.251, lm_loss: 8.251, perplexity: 3831.453 I0425 07:23:10.185567 135965884618560 checkpointing.py:772] Waiting for step 9 to finish before checkpoint... I0425 07:23:10.186244 135965884618560 checkpointing.py:776] Waited 0.0006976127624511719 seconds for step 9 to finish before starting checkpointing. I0425 07:23:10.188390 135965884618560 checkpoint_manager.py:2020] [process=5][thread=MainThread][step=0][wait_until_finished] Waiting for Save Finalize thread (save_finalize) to complete. I0425 07:23:11.670372 135833875052288 base_pytree_checkpoint_handler.py:130] [process=5] /jax/orbax/write/gbytes_per_sec: 46.611 MiB/s (total gbytes: 1.5 GiB) (time elapsed: 33.88920211791992 s) (per-host) I0425 07:23:11.670495 135833875052288 async_checkpointer.py:90] [process=5][thread=async_save] 3 Handler Commit operations completed. Time taken: 33.602513s. I0425 07:23:19.863918 135833875052288 async_checkpointer.py:160] [process=5][thread=async_save] Background save thread done. Time taken: 41.795918s. I0425 07:23:19.864192 135831321896704 async_checkpointer.py:288] [process=5][thread=save_finalize] Done with waiting for background save thread=async_save. I0425 07:23:19.864320 135831321896704 async_checkpointer.py:298] [process=5][thread=save_finalize] No errors found in background save thread=async_save. I0425 07:23:19.864371 135831321896704 checkpoint_manager.py:2137] [process=5][thread=save_finalize][step=0] CheckpointManager Save Finalize is syncing with other hosts... I0425 07:23:19.866041 135831321896704 checkpoint_manager.py:2146] [process=5][thread=save_finalize][step=0] CheckpointManager Save Finalize is done on all hosts. I0425 07:23:19.866220 135965884618560 checkpoint_manager.py:2032] [process=5][thread=MainThread][step=0][wait_until_finished] Done waiting for Save Finalize thread (save_finalize) running at step=0. W0425 07:23:19.866360 135965884618560 checkpoint_manager.py:1452] Waiting for previous save to complete took 9.677972 seconds. If this number is high, consider checkpointing less frequently. I0425 07:23:19.868197 135965884618560 checkpoint_manager.py:1512] [process=5] Saving checkpoint at step 9 I0425 07:23:19.870316 135965884618560 event_tracking.py:70] [process=5] [async] Started save checkpoint @ gs://lance-maxtext/linen_ckpt_xpk_main_20260425_071456/linen_xpk_main_20260425_071456_04_int8/checkpoints/9. I0425 07:23:20.180980 135965884618560 jax_array_handlers.py:360] Scheduling D2H of 39 prioritized jax.Array. I0425 07:23:20.181073 135965884618560 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 07:23:20.224014 135965884618560 base_pytree_checkpoint_handler.py:154] [process=5][thread=MainThread] Initiated "orbax.checkpoint._src.serialization.jax_array_handlers.ArrayHandler".serialize. Time taken: 0.044906s I0425 07:23:20.224178 135965884618560 base_pytree_checkpoint_handler.py:130] [process=5] /jax/orbax/write/blocking_gbytes_per_sec: 31.903 GiB/s (total gbytes: 1.5 GiB) (time elapsed: 0.048351287841796875 s) (per-host) I0425 07:23:20.224235 135965884618560 base_pytree_checkpoint_handler.py:768] [process=5][thread=MainThread] Initiated Pytree async_save. Time taken: 0.048417s (batch_requests_ready=0.001795s, total_serialization_initiated=0.046549s, others=0.000073s) I0425 07:23:20.224331 135965884618560 composite_checkpoint_handler.py:715] [process=5][thread=MainThread] Initiated CompositeCheckpointHandler.async_save. Time taken: 0.053346s (all_items=0.000016s, per_item={'items': '0.00001597'}, temp_paths=0.053330) I0425 07:23:20.225048 135965884618560 event_tracking.py:125] [process=5] [async] Finished blocking save in 0.36 seconds. Continuing save @ gs://lance-maxtext/linen_ckpt_xpk_main_20260425_071456/linen_xpk_main_20260425_071456_04_int8/checkpoints/9. I0425 07:23:20.225403 135831321896704 async_checkpointer.py:76] [process=5][thread=async_save] Background save thread started. Deadline for this save operation is 2026-04-25 07:43:20.225366 I0425 07:23:20.227372 135965884618560 checkpoint_manager.py:1560] [process=5][thread=MainThread][step=9] Starting CheckpointManager Save Finalize thread=save_finalize I0425 07:23:20.227627 135833340327680 async_checkpointer.py:280] [process=5][thread=save_finalize] Waiting for background save thread=async_save. I0425 07:23:20.227800 135965884618560 standard_logger.py:34] {'step': 9, 'event_type': 'save', 'directory': 'gs://lance-maxtext/linen_ckpt_xpk_main_20260425_071456/linen_xpk_main_20260425_071456_04_int8/checkpoints', 'reached_preemption': False, 'preemption_received_at': None, 'synchronous': False, 'wait_for_prev_start_time': 1777101790.188358, 'wait_for_prev_duration_secs': 9.677971601486206, 'time_between_consecutive_saves_sec': None, 'checkpointer_blocking_start_time': 1777101799.8682365, 'checkpointer_blocking_duration_secs': 0.35731029510498047, 'get_old_steps_start_time': 1777101800.2255714, 'get_old_steps_duration_secs': 2.9802322387695312e-05, 'checkpoint_manager_blocking_start_time': 1777101790.186482, 'checkpoint_manager_blocking_duration_secs': 10.041281700134277} I0425 07:23:20.227904 135965884618560 checkpointing.py:408] Started an asynchronous checkpoint save for step 9 I0425 07:23:20.227949 135965884618560 checkpoint_manager.py:2020] [process=5][thread=MainThread][step=9][wait_until_finished] Waiting for Save Finalize thread (save_finalize) to complete. I0425 07:23:25.333943 135833859925760 array_metadata_store.py:203] [process=5][thread=array_type_handler] Wrote 39 array_metadata.ArrayMetadata to gs://lance-maxtext/linen_ckpt_xpk_main_20260425_071456/linen_xpk_main_20260425_071456_04_int8/checkpoints/9/items/array_metadatas/process_5 I0425 07:24:02.656580 135831321896704 base_pytree_checkpoint_handler.py:130] [process=5] /jax/orbax/write/gbytes_per_sec: 37.184 MiB/s (total gbytes: 1.5 GiB) (time elapsed: 42.48072957992554 s) (per-host) I0425 07:24:02.656706 135831321896704 async_checkpointer.py:90] [process=5][thread=async_save] 3 Handler Commit operations completed. Time taken: 42.431186s. I0425 07:24:10.118889 135831321896704 async_checkpointer.py:160] [process=5][thread=async_save] Background save thread done. Time taken: 49.893354s. I0425 07:24:10.119287 135833340327680 async_checkpointer.py:288] [process=5][thread=save_finalize] Done with waiting for background save thread=async_save. I0425 07:24:10.119402 135833340327680 async_checkpointer.py:298] [process=5][thread=save_finalize] No errors found in background save thread=async_save. I0425 07:24:10.119449 135833340327680 checkpoint_manager.py:2137] [process=5][thread=save_finalize][step=9] CheckpointManager Save Finalize is syncing with other hosts... I0425 07:24:10.120985 135833340327680 checkpoint_manager.py:2146] [process=5][thread=save_finalize][step=9] CheckpointManager Save Finalize is done on all hosts. I0425 07:24:10.121161 135965884618560 checkpoint_manager.py:2032] [process=5][thread=MainThread][step=9][wait_until_finished] Done waiting for Save Finalize thread (save_finalize) running at step=9. I0425 07:24:10.121315 135965884618560 checkpoint_manager.py:2009] [process=5][thread=MainThread][wait_until_finished] No Save Finalize thread to wait for. Returning. I0425 07:24:10.122251 135965884618560 metric_logger.py:196] completed step: 9, seconds: 0.229, TFLOP/s/device: 59.249, Tokens/s/device: 8930.674, total_weights: 65536, loss: 8.175, lm_loss: 8.175, perplexity: 3552.333 Per train step: Total TFLOPs: 13.59 split as 93.93% learnable weight flops and 6.07% attention flops XPK End: Sat Apr 25 07:24:26 UTC 2026 EXIT_CODE=0
4f4617640
· feat_Migrate_Decoder_And_Tests_to_NNX_20260425_180433
· full log
XPK Start: Sat Apr 25 18:11:04 UTC 2026 2026-04-25 18:11:08.824176: 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 00:00:1777140668.837084 10 cuda_dnn.cc:8579] Unable to register cuDNN factory: Attempting to register factory for plugin cuDNN when one has already been registered E0000 00:00:1777140668.840838 10 cuda_blas.cc:1407] Unable to register cuBLAS factory: Attempting to register factory for plugin cuBLAS when one has already been registered W0000 00:00:1777140668.852565 10 computation_placer.cc:177] computation placer already registered. Please check linkage and avoid linking the same target more than once. W0000 00:00:1777140668.852583 10 computation_placer.cc:177] computation placer already registered. Please check linkage and avoid linking the same target more than once. W0000 00:00:1777140668.852586 10 computation_placer.cc:177] computation placer already registered. Please check linkage and avoid linking the same target more than once. W0000 00:00:1777140668.852588 10 computation_placer.cc:177] computation placer already registered. Please check linkage and avoid linking the same target more than once. 2026-04-25 18:11:28.006522: 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 18:11:28.521991 136328432793408 max_utils.py:273] Attempting to initialize the jax distributed system... INFO:2026-04-25 18:11:37,564:jax._src.distributed:140: Starting JAX distributed service on [::]:8482 I0425 18:11:37.564607 136328432793408 distributed.py:140] Starting JAX distributed service on [::]:8482 INFO:2026-04-25 18:11:37,566:jax._src.distributed:157: Connecting to JAX distributed service on mt-04-int8-x4xd8-slice-job-0-0.mt-04-int8-x4xd8:8482 I0425 18:11:37.566834 136328432793408 distributed.py:157] Connecting to JAX distributed service on mt-04-int8-x4xd8-slice-job-0-0.mt-04-int8-x4xd8:8482 I0425 18:11:38.621506 136328432793408 max_utils.py:284] Jax distributed system initialized! I0425 18:11:45.646522 136328432793408 max_utils.py:800] System Information: Jax Version: 0.8.1 I0425 18:11:45.646631 136328432793408 max_utils.py:801] System Information: Jaxlib Version: 0.8.1 I0425 18:11:45.646671 136328432793408 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 18:11:45.646705 136328432793408 train_utils.py:347] WARNING: Sequence packing is essentially ignored for synthetic data. Please use a real dataset to use sequence packing. I0425 18:11:46.264520 136328432793408 maxtext_utils.py:1517] Num_devices: 32, shape (1, 1, 1, 32, 1, 1, 1, 1, 1, 1, 1, 1, 1) I0425 18:11:46.264940 136328432793408 checkpointing.py:677] Setting up checkpoint logger... I0425 18:11:46.264998 136328432793408 checkpointing.py:233] Creating checkpoint manager with ocdbt=True and zarr3=True I0425 18:11:46.265044 136328432793408 pytree_checkpoint_handler.py:589] save_device_host_concurrent_bytes=None I0425 18:11:46.265410 136328432793408 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 0x7bfce6774da0>, enable_pinned_host_transfer=False, save_concurrent_bytes: 96000000000 (89.4 GiB), restore_concurrent_bytes: 96000000000 (89.4 GiB) I0425 18:11:49.128985 136328432793408 checkpointing.py:265] Enabling policy for fixed interval checkpointing. I0425 18:11:49.129363 136328432793408 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 0x7bfbaa96db20>}, handler_registry=None I0425 18:11:49.129605 136328432793408 composite_checkpoint_handler.py:237] Deferred registration for item: "items". Adding handler `<orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeCheckpointHandler object at 0x7bfbaa96db20>` 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 18:11:49.129654 136328432793408 composite_checkpoint_handler.py:237] Deferred registration for item: "metrics". Adding handler `<orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonCheckpointHandler object at 0x7be8541a73b0>` 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 18:11:49.129690 136328432793408 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 0x7bfbaa96db20>, ('items', <class 'orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeRestoreArgs'>): <orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeCheckpointHandler object at 0x7bfbaa96db20>, ('metrics', <class 'orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonSaveArgs'>): <orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonCheckpointHandler object at 0x7be8541a73b0>, ('metrics', <class 'orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonRestoreArgs'>): <orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonCheckpointHandler object at 0x7be8541a73b0>}). I0425 18:11:49.130225 136328432793408 abstract_checkpointer.py:35] orbax-checkpoint version: 0.11.33 I0425 18:11:49.130303 136328432793408 async_checkpointer.py:177] [process=6][thread=MainThread] Using barrier_sync_fn: <function get_barrier_sync_fn.<locals>._fn at 0x7be85410aac0> timeout: 600 secs and primary_host=0 for async checkpoint writes I0425 18:11:51.963948 136328432793408 checkpoint_manager.py:1818] Found 0 checkpoint steps in gs://lance-maxtext/linen_ckpt_xpk_feat_Migrate_Decoder_And_Tests_to_NNX_20260425_180433/linen_xpk_feat_Migrate_Decoder_And_Tests_to_NNX_20260425_180433_04_int8/checkpoints I0425 18:11:51.966908 136328432793408 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_Migrate_Decoder_And_Tests_to_NNX_20260425_180433/linen_xpk_feat_Migrate_Decoder_And_Tests_to_NNX_20260425_180433_04_int8/checkpoints: <orbax.checkpoint.checkpoint_manager.CheckpointManager object at 0x7be8541a1880> I0425 18:11:51.967021 136328432793408 checkpointing.py:301] Checkpoint manager created! I0425 18:11:53.151298 136328432793408 nnx_wrappers.py:442] Unknown Logical: bfloat16[32,2048,2048]...................................... ('activation_batch', 'activation_norm_length', 'activation_embed'). I0425 18:11:53.151409 136328432793408 nnx_wrappers.py:442] Unknown Physical: bfloat16[32,2048,2048]...................................... ('fsdp', None, None). I0425 18:11:53.524302 136328432793408 attentions.py:1088] attentions/inputs_q Logical: bfloat16[32,2048,2048]...................................... ('activation_batch', 'activation_attn_length', 'activation_attn_embed'). I0425 18:11:53.524400 136328432793408 attentions.py:1088] attentions/inputs_q Physical: bfloat16[32,2048,2048]...................................... ('fsdp', None, None). I0425 18:11:53.540305 136328432793408 attentions.py:1089] attentions/inputs_kv Logical: bfloat16[32,2048,2048]...................................... ('activation_batch', 'activation_attn_length', 'activation_attn_embed'). I0425 18:11:53.540364 136328432793408 attentions.py:1089] attentions/inputs_kv Physical: bfloat16[32,2048,2048]...................................... ('fsdp', None, None). I0425 18:11:53.618634 136328432793408 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 18:11:53.618737 136328432793408 attentions.py:1154] attentions/query Physical: bfloat16[32,2048,16,128].................................... ('fsdp', None, None, None). I0425 18:11:53.634839 136328432793408 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 18:11:53.634903 136328432793408 attentions.py:1155] attentions/key Physical: bfloat16[32,2048,16,128].................................... ('fsdp', None, None, None). I0425 18:11:53.650830 136328432793408 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 18:11:53.650893 136328432793408 attentions.py:1156] attentions/value Physical: bfloat16[32,2048,16,128].................................... ('fsdp', None, None, None). I0425 18:11:53.675386 136328432793408 attentions.py:1197] attentions/out Logical: bfloat16[32,2048,16,128].................................... ('activation_batch', 'activation_attn_length', 'activation_heads', 'activation_kv'). I0425 18:11:53.675458 136328432793408 attentions.py:1197] attentions/out Physical: bfloat16[32,2048,16,128].................................... ('fsdp', None, None, None). I0425 18:11:53.756484 136328432793408 linears.py:525] linears/x Logical: bfloat16[32,2048,7168]...................................... ('activation_batch', 'activation_length', 'activation_mlp'). I0425 18:11:53.756576 136328432793408 linears.py:525] linears/x Physical: bfloat16[32,2048,7168]...................................... ('fsdp', None, None). I0425 18:11:54.355860 136328432793408 checkpointing.py:577] checkpoint manager exists so trying to load this run's existing checkpoint I0425 18:11:54.355977 136328432793408 checkpointing.py:665] No existing checkpoints found, not restoring checkpoint. fsdp: 32 I0425 18:11:57.028488 136328432793408 maxtext_utils.py:1620] params/params/decoder/decoder_norm/scale Shape: float32[2048] Logical: PartitionSpec('norm',) Physical: (None,) I0425 18:11:57.028624 136328432793408 maxtext_utils.py:1620] params/params/decoder/layers/mlp/wi_0/kernel Shape: float32[2048,16,7168] Logical: PartitionSpec('embed', 'layers', 'mlp') Physical: ('fsdp', None, None) I0425 18:11:57.028678 136328432793408 maxtext_utils.py:1620] params/params/decoder/layers/mlp/wi_1/kernel Shape: float32[2048,16,7168] Logical: PartitionSpec('embed', 'layers', 'mlp') Physical: ('fsdp', None, None) I0425 18:11:57.028736 136328432793408 maxtext_utils.py:1620] params/params/decoder/layers/mlp/wo/kernel Shape: float32[7168,16,2048] Logical: PartitionSpec('mlp', 'layers', 'embed') Physical: (None, None, 'fsdp') I0425 18:11:57.028789 136328432793408 maxtext_utils.py:1620] params/params/decoder/layers/post_self_attention_layer_norm/scale Shape: float32[2048,16] Logical: PartitionSpec('norm', 'layers') Physical: (None, None) I0425 18:11:57.028828 136328432793408 maxtext_utils.py:1620] params/params/decoder/layers/pre_self_attention_layer_norm/scale Shape: float32[2048,16] Logical: PartitionSpec('norm', 'layers') Physical: (None, None) I0425 18:11:57.028881 136328432793408 maxtext_utils.py:1620] 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 18:11:57.028933 136328432793408 maxtext_utils.py:1620] 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 18:11:57.028973 136328432793408 maxtext_utils.py:1620] 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 18:11:57.029009 136328432793408 maxtext_utils.py:1620] 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 18:11:57.029056 136328432793408 maxtext_utils.py:1620] params/params/decoder/logits_dense/kernel Shape: float32[2048,32000] Logical: PartitionSpec('embed', 'vocab') Physical: ('fsdp', None) I0425 18:11:57.029127 136328432793408 maxtext_utils.py:1620] params/params/token_embedder/embedding Shape: float32[32000,2048] Logical: PartitionSpec('vocab', 'embed') Physical: (None, 'fsdp') I0425 18:11:58.741808 136328432793408 train.py:155] train/xent Logical: float32[32,2048]............................................ ('activation_embed_and_logits_batch', 'activation_length'). I0425 18:11:58.741912 136328432793408 train.py:155] train/xent Physical: float32[32,2048]............................................ ('fsdp', None). I0425 18:11:58.756796 136328432793408 train.py:162] train/z_loss Logical: float32[32,2048]............................................ ('activation_embed_and_logits_batch', 'activation_length'). I0425 18:11:58.756856 136328432793408 train.py:162] train/z_loss Physical: float32[32,2048]............................................ ('fsdp', None). I0425 18:12:12.816176 136328432793408 max_utils.py:791] Total memory size: 1.8 GB, Output size: 0.4 GB, Temp size: 1.4 GB, Argument size: 0.4 GB, Host temp size: 0.0 GB. I0425 18:12:12.817058 136328432793408 metric_logger.py:289] number parameters: 1.104 billion I0425 18:12:28.438791 136328432793408 checkpointing.py:772] Waiting for step 0 to finish before checkpoint... I0425 18:12:28.680114 136328432793408 checkpointing.py:776] Waited 0.24128222465515137 seconds for step 0 to finish before starting checkpointing. I0425 18:12:28.682749 136328432793408 checkpoint_manager.py:2013] [process=6][thread=MainThread][wait_until_finished] No Save Finalize thread to wait for. Returning. I0425 18:12:28.684352 136328432793408 checkpoint_manager.py:1518] [process=6] Saving checkpoint at step 0 I0425 18:12:28.687063 136328432793408 async_checkpointer.py:452] [process=6] Started async saving checkpoint to gs://lance-maxtext/linen_ckpt_xpk_feat_Migrate_Decoder_And_Tests_to_NNX_20260425_180433/linen_xpk_feat_Migrate_Decoder_And_Tests_to_NNX_20260425_180433_04_int8/checkpoints/0. I0425 18:12:29.716378 136328432793408 signaling_client.py:364] Using JaxDistributedSignalingClient I0425 18:12:29.717335 136328432793408 jax_array_handlers.py:358] Scheduling D2H of 39 prioritized jax.Array. I0425 18:12:29.717393 136328432793408 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 18:12:29.964738 136328432793408 base_pytree_checkpoint_handler.py:153] [process=6][thread=MainThread] Initiated "orbax.checkpoint._src.serialization.jax_array_handlers.ArrayHandler".serialize. Time taken: 0.248436s I0425 18:12:29.964915 136328432793408 base_pytree_checkpoint_handler.py:129] [process=6] /jax/checkpoint/write/blocking_gbytes_per_sec: 2.996 GiB/s (total gbytes: 1.5 GiB) (time elapsed: 0.514939546585083 s) (per-host) I0425 18:12:29.964969 136328432793408 base_pytree_checkpoint_handler.py:737] [process=6][thread=MainThread] Initiated Pytree async_save. Time taken: 0.515004s (batch_requests_ready=0.262597s, total_serialization_initiated=0.252335s, others=0.000073s) I0425 18:12:29.965061 136328432793408 composite_checkpoint_handler.py:715] [process=6][thread=MainThread] Initiated CompositeCheckpointHandler.async_save. Time taken: 0.519222s (all_items=0.000021s, per_item={'items': '0.00002050'}, temp_paths=0.519201) I0425 18:12:29.965992 136205131826944 async_checkpointer.py:79] [process=6][thread=async_save] Background save thread started. I0425 18:12:29.966179 136328432793408 async_checkpointer.py:561] Finished blocking save. Time taken: 1.281756s. Continuing background save to gs://lance-maxtext/linen_ckpt_xpk_feat_Migrate_Decoder_And_Tests_to_NNX_20260425_180433/linen_xpk_feat_Migrate_Decoder_And_Tests_to_NNX_20260425_180433_04_int8/checkpoints/0. I0425 18:12:30.401154 136328432793408 checkpoint_manager.py:1566] [process=6][thread=MainThread][step=0] Starting CheckpointManager Save Finalize thread=save_finalize I0425 18:12:30.401596 136194419586816 async_checkpointer.py:265] [process=6][thread=save_finalize] Waiting for background save thread=async_save. I0425 18:12:30.401769 136328432793408 standard_logger.py:34] {'step': 0, 'event_type': 'save', 'directory': 'gs://lance-maxtext/linen_ckpt_xpk_feat_Migrate_Decoder_And_Tests_to_NNX_20260425_180433/linen_xpk_feat_Migrate_Decoder_And_Tests_to_NNX_20260425_180433_04_int8/checkpoints', 'reached_preemption': False, 'preemption_received_at': None, 'synchronous': False, 'wait_for_prev_start_time': 1777140748.68273, 'wait_for_prev_duration_secs': 6.556510925292969e-05, 'time_between_consecutive_saves_sec': None, 'checkpointer_blocking_start_time': 1777140748.6843903, 'checkpointer_blocking_duration_secs': 1.2818999290466309, 'get_old_steps_start_time': 1777140749.9663112, 'get_old_steps_duration_secs': 3.24249267578125e-05, 'checkpoint_manager_blocking_start_time': 1777140748.6807666, 'checkpoint_manager_blocking_duration_secs': 1.7209603786468506} I0425 18:12:30.401880 136328432793408 checkpointing.py:408] Started an asynchronous checkpoint save for step 0 I0425 18:12:30.401932 136328432793408 max_utils.py:750] Memstats: After params initialized: I0425 18:12:30.401990 136328432793408 max_utils.py:756] Using (GB) 0.43 / 31.25 (1.376000%) on TPU_24(process=6,(0,6,0,0)) I0425 18:12:30.402022 136328432793408 max_utils.py:756] Using (GB) 0.43 / 31.25 (1.376000%) on TPU_25(process=6,(1,6,0,0)) I0425 18:12:30.402050 136328432793408 max_utils.py:756] Using (GB) 0.43 / 31.25 (1.376000%) on TPU_28(process=6,(0,7,0,0)) I0425 18:12:30.402076 136328432793408 max_utils.py:756] Using (GB) 0.43 / 31.25 (1.376000%) on TPU_29(process=6,(1,7,0,0)) I0425 18:12:30.671845 136328432793408 metric_logger.py:185] completed step: 0, seconds: 15.622, TFLOP/s/device: 0.870, Tokens/s/device: 131.101, total_weights: 65536, loss: 10.874 I0425 18:12:30.929070 136328432793408 metric_logger.py:185] completed step: 1, seconds: 2.232, TFLOP/s/device: 6.088, Tokens/s/device: 917.721, total_weights: 65536, loss: 10.874 I0425 18:12:31.347723 136328432793408 metric_logger.py:185] completed step: 2, seconds: 0.028, TFLOP/s/device: 478.252, Tokens/s/device: 72087.293, total_weights: 65536, loss: 10.262 I0425 18:12:31.578049 136328432793408 metric_logger.py:185] completed step: 3, seconds: 0.419, TFLOP/s/device: 32.396, Tokens/s/device: 4883.120, total_weights: 65536, loss: 9.731 I0425 18:12:32.037731 136328432793408 metric_logger.py:185] completed step: 4, seconds: 0.235, TFLOP/s/device: 57.936, Tokens/s/device: 8732.768, total_weights: 65536, loss: 9.272 I0425 18:12:32.043243 136328432793408 metric_logger.py:185] completed step: 5, seconds: 0.230, TFLOP/s/device: 59.196, Tokens/s/device: 8922.736, total_weights: 65536, loss: 8.887 I0425 18:12:54.284942 136328432793408 metric_logger.py:185] completed step: 6, seconds: 0.460, TFLOP/s/device: 29.522, Tokens/s/device: 4449.930, total_weights: 65536, loss: 8.588 I0425 18:12:54.514584 136328432793408 metric_logger.py:185] completed step: 7, seconds: 22.014, TFLOP/s/device: 0.617, Tokens/s/device: 93.033, total_weights: 65536, loss: 8.380 I0425 18:12:54.745220 136328432793408 metric_logger.py:185] completed step: 8, seconds: 0.233, TFLOP/s/device: 58.331, Tokens/s/device: 8792.303, total_weights: 65536, loss: 8.251 I0425 18:12:54.975543 136328432793408 checkpointing.py:772] Waiting for step 9 to finish before checkpoint... I0425 18:12:54.976155 136328432793408 checkpointing.py:776] Waited 0.0006401538848876953 seconds for step 9 to finish before starting checkpointing. I0425 18:12:54.978230 136328432793408 checkpoint_manager.py:2024] [process=6][thread=MainThread][step=0][wait_until_finished] Waiting for Save Finalize thread (save_finalize) to complete. I0425 18:12:55.945448 2869 google_auth_provider.cc:181] Running on GCE, using service account 562977990677-compute@developer.gserviceaccount.com I0425 18:12:57.880513 136205140219648 array_metadata_store.py:203] [process=6][thread=array_type_handler] Wrote 39 array_metadata.ArrayMetadata to gs://lance-maxtext/linen_ckpt_xpk_feat_Migrate_Decoder_And_Tests_to_NNX_20260425_180433/linen_xpk_feat_Migrate_Decoder_And_Tests_to_NNX_20260425_180433_04_int8/checkpoints/0/items/array_metadatas/process_6 I0425 18:13:26.292318 136205131826944 base_pytree_checkpoint_handler.py:129] [process=6] /jax/checkpoint/write/gbytes_per_sec: 27.789 MiB/s (total gbytes: 1.5 GiB) (time elapsed: 56.8422966003418 s) (per-host) I0425 18:13:26.292442 136205131826944 async_checkpointer.py:90] [process=6][thread=async_save] 3 Handler Commit operations completed. Time taken: 56.326323s. I0425 18:13:33.826427 136205131826944 async_checkpointer.py:144] [process=6][thread=async_save] Background save thread done. Time taken: 63.860292s. I0425 18:13:33.826659 136194419586816 async_checkpointer.py:273] [process=6][thread=save_finalize] Done with waiting for background save thread=async_save. I0425 18:13:33.826726 136194419586816 async_checkpointer.py:283] [process=6][thread=save_finalize] No errors found in background save thread=async_save. I0425 18:13:33.826774 136194419586816 checkpoint_manager.py:2133] [process=6][thread=save_finalize][step=0] CheckpointManager Save Finalize is syncing with other hosts... I0425 18:13:33.828452 136194419586816 checkpoint_manager.py:2142] [process=6][thread=save_finalize][step=0] CheckpointManager Save Finalize is done on all hosts. I0425 18:13:33.828650 136328432793408 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 18:13:33.828783 136328432793408 checkpoint_manager.py:1458] Waiting for previous save to complete took 38.850559 seconds. If this number is high, consider checkpointing less frequently. I0425 18:13:33.830660 136328432793408 checkpoint_manager.py:1518] [process=6] Saving checkpoint at step 9 I0425 18:13:33.834174 136328432793408 async_checkpointer.py:452] [process=6] Started async saving checkpoint to gs://lance-maxtext/linen_ckpt_xpk_feat_Migrate_Decoder_And_Tests_to_NNX_20260425_180433/linen_xpk_feat_Migrate_Decoder_And_Tests_to_NNX_20260425_180433_04_int8/checkpoints/9. I0425 18:13:34.379449 136328432793408 jax_array_handlers.py:358] Scheduling D2H of 39 prioritized jax.Array. I0425 18:13:34.379553 136328432793408 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 18:13:34.422691 136328432793408 base_pytree_checkpoint_handler.py:153] [process=6][thread=MainThread] Initiated "orbax.checkpoint._src.serialization.jax_array_handlers.ArrayHandler".serialize. Time taken: 0.044497s I0425 18:13:34.422861 136328432793408 base_pytree_checkpoint_handler.py:129] [process=6] /jax/checkpoint/write/blocking_gbytes_per_sec: 5.096 GiB/s (total gbytes: 1.5 GiB) (time elapsed: 0.30272793769836426 s) (per-host) I0425 18:13:34.422913 136328432793408 base_pytree_checkpoint_handler.py:737] [process=6][thread=MainThread] Initiated Pytree async_save. Time taken: 0.302788s (batch_requests_ready=0.256293s, total_serialization_initiated=0.046425s, others=0.000070s) I0425 18:13:34.423006 136328432793408 composite_checkpoint_handler.py:715] [process=6][thread=MainThread] Initiated CompositeCheckpointHandler.async_save. Time taken: 0.306929s (all_items=0.000017s, per_item={'items': '0.00001717'}, temp_paths=0.306912) I0425 18:13:34.423910 136198169290496 async_checkpointer.py:79] [process=6][thread=async_save] Background save thread started. I0425 18:13:34.424075 136328432793408 async_checkpointer.py:561] Finished blocking save. Time taken: 0.593342s. Continuing background save to gs://lance-maxtext/linen_ckpt_xpk_feat_Migrate_Decoder_And_Tests_to_NNX_20260425_180433/linen_xpk_feat_Migrate_Decoder_And_Tests_to_NNX_20260425_180433_04_int8/checkpoints/9. I0425 18:13:34.440438 136328432793408 checkpoint_manager.py:1566] [process=6][thread=MainThread][step=9] Starting CheckpointManager Save Finalize thread=save_finalize I0425 18:13:34.440681 136194419586816 async_checkpointer.py:265] [process=6][thread=save_finalize] Waiting for background save thread=async_save. I0425 18:13:34.440808 136328432793408 standard_logger.py:34] {'step': 9, 'event_type': 'save', 'directory': 'gs://lance-maxtext/linen_ckpt_xpk_feat_Migrate_Decoder_And_Tests_to_NNX_20260425_180433/linen_xpk_feat_Migrate_Decoder_And_Tests_to_NNX_20260425_180433_04_int8/checkpoints', 'reached_preemption': False, 'preemption_received_at': None, 'synchronous': False, 'wait_for_prev_start_time': 1777140774.9781995, 'wait_for_prev_duration_secs': 38.85055947303772, 'time_between_consecutive_saves_sec': None, 'checkpointer_blocking_start_time': 1777140813.8307009, 'checkpointer_blocking_duration_secs': 0.5934960842132568, 'get_old_steps_start_time': 1777140814.4242163, 'get_old_steps_duration_secs': 2.6464462280273438e-05, 'checkpoint_manager_blocking_start_time': 1777140774.9763877, 'checkpoint_manager_blocking_duration_secs': 39.46438670158386} I0425 18:13:34.440922 136328432793408 checkpointing.py:408] Started an asynchronous checkpoint save for step 9 I0425 18:13:34.440968 136328432793408 checkpoint_manager.py:2024] [process=6][thread=MainThread][step=9][wait_until_finished] Waiting for Save Finalize thread (save_finalize) to complete. I0425 18:13:39.551858 136205140219648 array_metadata_store.py:203] [process=6][thread=array_type_handler] Wrote 39 array_metadata.ArrayMetadata to gs://lance-maxtext/linen_ckpt_xpk_feat_Migrate_Decoder_And_Tests_to_NNX_20260425_180433/linen_xpk_feat_Migrate_Decoder_And_Tests_to_NNX_20260425_180433_04_int8/checkpoints/9/items/array_metadatas/process_6 I0425 18:14:15.562183 136198169290496 base_pytree_checkpoint_handler.py:129] [process=6] /jax/checkpoint/write/gbytes_per_sec: 38.116 MiB/s (total gbytes: 1.5 GiB) (time elapsed: 41.442007064819336 s) (per-host) I0425 18:14:15.562318 136198169290496 async_checkpointer.py:90] [process=6][thread=async_save] 3 Handler Commit operations completed. Time taken: 41.138290s. I0425 18:14:24.119863 136198169290496 async_checkpointer.py:144] [process=6][thread=async_save] Background save thread done. Time taken: 49.695821s. I0425 18:14:24.120090 136194419586816 async_checkpointer.py:273] [process=6][thread=save_finalize] Done with waiting for background save thread=async_save. I0425 18:14:24.120170 136194419586816 async_checkpointer.py:283] [process=6][thread=save_finalize] No errors found in background save thread=async_save. I0425 18:14:24.120216 136194419586816 checkpoint_manager.py:2133] [process=6][thread=save_finalize][step=9] CheckpointManager Save Finalize is syncing with other hosts... I0425 18:14:24.122000 136194419586816 checkpoint_manager.py:2142] [process=6][thread=save_finalize][step=9] CheckpointManager Save Finalize is done on all hosts. I0425 18:14:24.122177 136328432793408 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 18:14:24.122322 136328432793408 checkpoint_manager.py:2013] [process=6][thread=MainThread][wait_until_finished] No Save Finalize thread to wait for. Returning. I0425 18:14:24.123068 136328432793408 metric_logger.py:185] completed step: 9, seconds: 0.230, TFLOP/s/device: 59.125, Tokens/s/device: 8911.904, total_weights: 65536, loss: 8.175 Per train step: Total TFLOPs: 13.59 split as 93.93% learnable weight flops and 6.07% attention flops XPK End: Sat Apr 25 18:14:35 UTC 2026 EXIT_CODE=0