2026-04-15 22:23:03.816448: 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) I0415 22:23:03.935229 124104330243200 max_utils.py:238] Skipping jax distributed system due to skip_jax_distributed_system=True flag. I0415 22:23:59.126346 124104330243200 max_utils.py:800] System Information: Jax Version: 0.9.2 I0415 22:23:59.126451 124104330243200 max_utils.py:801] System Information: Jaxlib Version: 0.9.2 I0415 22:23:59.126485 124104330243200 max_utils.py:802] System Information: Jax Backend: PJRT C API TFRT TPU v6 lite Built on Apr 6 2026 20:48:10 (1775533690) cl/895581894 I0415 22:23:59.126508 124104330243200 train_utils.py:334] WARNING: 'dataset_path' might be pointing your local file system I0415 22:23:59.126530 124104330243200 train_utils.py:347] WARNING: Sequence packing is essentially ignored for synthetic data. Please use a real dataset to use sequence packing. I0415 22:23:59.126606 124104330243200 train.py:683] [DECOUPLED NO-OP] skipping cloud diagnostics wrapper. W0415 22:23:59.216620 3298240 pjrt_executable.cc:642] Assume version compatibility. PjRt-IFRT does not track XLA executable versions. I0415 22:23:59.621203 124104330243200 maxtext_utils.py:1517] Num_devices: 8, shape (1, 1, 1, 8, 1, 1, 1, 1, 1, 1, 1, 1, 1) I0415 22:23:59.621463 124104330243200 checkpointing.py:677] Setting up checkpoint logger... I0415 22:23:59.621506 124104330243200 checkpointing.py:233] Creating checkpoint manager with ocdbt=True and zarr3=False I0415 22:23:59.621546 124104330243200 pytree_checkpoint_handler.py:577] save_device_host_concurrent_bytes=None I0415 22:23:59.621988 124104330243200 base_pytree_checkpoint_handler.py:411] Created BasePyTreeCheckpointHandler: use_ocdbt=True, use_zarr3=False, pytree_metadata_options=PyTreeMetadataOptions(support_rich_types=False), array_metadata_store=<orbax.checkpoint._src.metadata.array_metadata_store.Store object at 0x70deb39f8950>, enable_pinned_host_transfer=False, save_concurrent_bytes: 96000000000 (89.4 GiB), restore_concurrent_bytes: 96000000000 (89.4 GiB) I0415 22:24:01.324526 124104330243200 checkpointing.py:265] Enabling policy for fixed interval checkpointing. I0415 22:24:01.324700 124104330243200 checkpoint_manager.py:702] [process=0][thread=MainThread] CheckpointManager init: checkpointers=None, item_names=('items',), item_handlers={'items': <orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeCheckpointHandler object at 0x70d837735df0>}, handler_registry=None I0415 22:24:01.324953 124104330243200 composite_checkpoint_handler.py:237] Deferred registration for item: "items". Adding handler `<orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeCheckpointHandler object at 0x70d837735df0>` 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`. I0415 22:24:01.324994 124104330243200 composite_checkpoint_handler.py:237] Deferred registration for item: "metrics". Adding handler `<orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonCheckpointHandler object at 0x70d837736030>` 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`. I0415 22:24:01.325026 124104330243200 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 0x70d837735df0>, ('items', <class 'orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeRestoreArgs'>): <orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeCheckpointHandler object at 0x70d837735df0>, ('metrics', <class 'orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonSaveArgs'>): <orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonCheckpointHandler object at 0x70d837736030>, ('metrics', <class 'orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonRestoreArgs'>): <orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonCheckpointHandler object at 0x70d837736030>}). I0415 22:24:01.325347 124104330243200 abstract_checkpointer.py:35] orbax-checkpoint version: 0.11.28 I0415 22:24:01.325398 124104330243200 async_checkpointer.py:177] [process=0][thread=MainThread] Using barrier_sync_fn: <function get_barrier_sync_fn.<locals>.<lambda> at 0x70d8377313a0> timeout: 600 secs and primary_host=0 for async checkpoint writes I0415 22:24:01.777802 123998216128064 checkpoint.py:228] Read Metadata={'item_handlers': {'items': 'orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeCheckpointHandler'}, 'metrics': {}, 'performance_metrics': {}, 'init_timestamp_nsecs': 1776291663008947775, 'commit_timestamp_nsecs': 1776291714754726828, 'custom_metadata': {}} from gs://wanglance-maxtext/linen_ckpt_main_20260415_211126/linen_main_20260415_211126_16_zarr3_false/checkpoints/5/_CHECKPOINT_METADATA I0415 22:24:01.862711 123998216128064 checkpoint_manager.py:1847] Missing metrics for step 5 I0415 22:24:01.912735 124104330243200 checkpoint_manager.py:1788] Found 3 checkpoint steps in gs://wanglance-maxtext/linen_ckpt_main_20260415_211126/linen_main_20260415_211126_16_zarr3_false/checkpoints I0415 22:24:01.913214 124104330243200 checkpoint_manager.py:921] [process=0][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_hns=False, enable_background_delete=False, read_only=False, enable_async_checkpointing=True, async_options=None, multiprocessing_options=MultiprocessingOptions(primary_host=0, active_processes=None, barrier_sync_key_prefix=None), should_save_fn=None, file_options=FileOptions(path_permission_mode=None), save_root_metadata=True, temporary_path_class=None, save_decision_policy=FixedIntervalPolicy(interval=5), preservation_policy=LatestN(n=None), prevent_write_metrics=False, enable_should_save_is_saving_in_progress_check=True, enable_per_process_directory_creation=False), root_directory=gs://wanglance-maxtext/linen_ckpt_main_20260415_211126/linen_main_20260415_211126_16_zarr3_false/checkpoints: <orbax.checkpoint.checkpoint_manager.CheckpointManager object at 0x70d837735460> I0415 22:24:01.913301 124104330243200 checkpointing.py:301] Checkpoint manager created! W0415 22:24:01.930038 3298240 pjrt_executable.cc:642] Assume version compatibility. PjRt-IFRT does not track XLA executable versions. I0415 22:24:02.110311 124104330243200 nnx_wrappers.py:437] Unknown Logical: bfloat16[8,2048,2048]....................................... ('activation_batch', 'activation_norm_length', 'activation_embed'). I0415 22:24:02.110408 124104330243200 nnx_wrappers.py:437] Unknown Physical: bfloat16[8,2048,2048]....................................... ('fsdp', None, None). I0415 22:24:02.202414 124104330243200 attentions.py:1088] attentions/inputs_q Logical: bfloat16[8,2048,2048]....................................... ('activation_batch', 'activation_attn_length', 'activation_attn_embed'). I0415 22:24:02.202489 124104330243200 attentions.py:1088] attentions/inputs_q Physical: bfloat16[8,2048,2048]....................................... ('fsdp', None, None). I0415 22:24:02.216764 124104330243200 attentions.py:1089] attentions/inputs_kv Logical: bfloat16[8,2048,2048]....................................... ('activation_batch', 'activation_attn_length', 'activation_attn_embed'). I0415 22:24:02.216811 124104330243200 attentions.py:1089] attentions/inputs_kv Physical: bfloat16[8,2048,2048]....................................... ('fsdp', None, None). I0415 22:24:02.237385 124104330243200 attentions.py:1154] attentions/query Logical: bfloat16[8,2048,16,128]..................................... ('activation_kv_batch', 'activation_attn_length', 'activation_kv_heads', 'activation_kv_head_dim'). I0415 22:24:02.237438 124104330243200 attentions.py:1154] attentions/query Physical: bfloat16[8,2048,16,128]..................................... ('fsdp', None, None, None). I0415 22:24:02.251672 124104330243200 attentions.py:1155] attentions/key Logical: bfloat16[8,2048,16,128]..................................... ('activation_kv_batch', 'activation_attn_length', 'activation_kv_heads', 'activation_kv_head_dim'). I0415 22:24:02.251721 124104330243200 attentions.py:1155] attentions/key Physical: bfloat16[8,2048,16,128]..................................... ('fsdp', None, None, None). I0415 22:24:02.265914 124104330243200 attentions.py:1156] attentions/value Logical: bfloat16[8,2048,16,128]..................................... ('activation_kv_batch', 'activation_attn_length', 'activation_kv_heads', 'activation_kv_head_dim'). I0415 22:24:02.265963 124104330243200 attentions.py:1156] attentions/value Physical: bfloat16[8,2048,16,128]..................................... ('fsdp', None, None, None). I0415 22:24:02.287648 124104330243200 attentions.py:1197] attentions/out Logical: bfloat16[8,2048,16,128]..................................... ('activation_batch', 'activation_attn_length', 'activation_heads', 'activation_kv'). I0415 22:24:02.287703 124104330243200 attentions.py:1197] attentions/out Physical: bfloat16[8,2048,16,128]..................................... ('fsdp', None, None, None). I0415 22:24:02.305857 124104330243200 linears.py:525] linears/x Logical: bfloat16[8,2048,7168]....................................... ('activation_batch', 'activation_length', 'activation_mlp'). I0415 22:24:02.305909 124104330243200 linears.py:525] linears/x Physical: bfloat16[8,2048,7168]....................................... ('fsdp', None, None). I0415 22:24:02.491933 124104330243200 checkpointing.py:577] checkpoint manager exists so trying to load this run's existing checkpoint I0415 22:24:02.492034 124104330243200 checkpointing.py:581] restoring from this run's directory step 9 I0415 22:24:02.639090 124104330243200 checkpointer.py:304] Restoring checkpoint from gs://wanglance-maxtext/linen_ckpt_main_20260415_211126/linen_main_20260415_211126_16_zarr3_false/checkpoints/9. I0415 22:24:03.162112 3299813 google_auth_provider.cc:149] Using credentials at ~/.config/gcloud/application_default_credentials.json I0415 22:24:03.162177 3299813 google_auth_provider.cc:156] Using OAuth2 AuthProvider I0415 22:24:09.016471 124104330243200 base_pytree_checkpoint_handler.py:128] [process=0] /jax/checkpoint/read/gbytes_per_sec: 2.045 GiB/s (total gbytes: 12.3 GiB) (time elapsed: 6 seconds) (per-host) I0415 22:24:09.016855 124104330243200 checkpointer.py:318] Finished restoring checkpoint in 6.45 seconds from gs://wanglance-maxtext/linen_ckpt_main_20260415_211126/linen_main_20260415_211126_16_zarr3_false/checkpoints/9. I0415 22:24:09.016974 124104330243200 standard_logger.py:34] {'step': 9, 'event_type': 'restore', 'directory': 'gs://wanglance-maxtext/linen_ckpt_main_20260415_211126/linen_main_20260415_211126_16_zarr3_false/checkpoints', 'checkpointer_start_time': 1776291842.5697165, 'checkpointer_duration_secs': 6.447200775146484, 'checkpoint_manager_start_time': 1776291842.4922874, 'checkpoint_manager_duration_secs': 6.524632692337036} [DECOUPLED NO-OP] gcs_storage: using stubs. [DECOUPLED NO-OP] mldiagnostics: using stub. [DECOUPLED NO-OP] mldiagnostics: using stub. [DECOUPLED NO-OP] mldiagnostics: using stub. [DECOUPLED NO-OP] workload_monitor: using stub. [DECOUPLED NO-OP] vertex_tensorboard: using stub. fsdp: 8 I0415 22:24:09.697311 124104330243200 maxtext_utils.py:1620] params/params/decoder/decoder_norm/scale Shape: float32[2048] Logical: P('norm',) Physical: (None,) I0415 22:24:09.697443 124104330243200 maxtext_utils.py:1620] params/params/decoder/layers/mlp/wi_0/kernel Shape: float32[2048,16,7168] Logical: P('embed', 'layers', 'mlp') Physical: ('fsdp', None, None) I0415 22:24:09.697484 124104330243200 maxtext_utils.py:1620] params/params/decoder/layers/mlp/wi_1/kernel Shape: float32[2048,16,7168] Logical: P('embed', 'layers', 'mlp') Physical: ('fsdp', None, None) I0415 22:24:09.697529 124104330243200 maxtext_utils.py:1620] params/params/decoder/layers/mlp/wo/kernel Shape: float32[7168,16,2048] Logical: P('mlp', 'layers', 'embed') Physical: (None, None, 'fsdp') I0415 22:24:09.697567 124104330243200 maxtext_utils.py:1620] params/params/decoder/layers/post_self_attention_layer_norm/scale Shape: float32[2048,16] Logical: P('norm', 'layers') Physical: (None, None) I0415 22:24:09.697611 124104330243200 maxtext_utils.py:1620] params/params/decoder/layers/pre_self_attention_layer_norm/scale Shape: float32[2048,16] Logical: P('norm', 'layers') Physical: (None, None) I0415 22:24:09.697652 124104330243200 maxtext_utils.py:1620] 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) I0415 22:24:09.697690 124104330243200 maxtext_utils.py:1620] 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') I0415 22:24:09.697719 124104330243200 maxtext_utils.py:1620] 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) I0415 22:24:09.697745 124104330243200 maxtext_utils.py:1620] 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) I0415 22:24:09.697779 124104330243200 maxtext_utils.py:1620] params/params/decoder/logits_dense/kernel Shape: float32[2048,32000] Logical: P('embed_vocab', 'vocab') Physical: ('fsdp', None) I0415 22:24:09.697813 124104330243200 maxtext_utils.py:1620] params/params/token_embedder/embedding Shape: float32[32000,2048] Logical: P('vocab', 'embed_vocab') Physical: (None, 'fsdp') I0415 22:24:10.170707 124104330243200 train.py:155] train/xent Logical: float32[8,2048]............................................. ('activation_embed_and_logits_batch', 'activation_length'). I0415 22:24:10.170787 124104330243200 train.py:155] train/xent Physical: float32[8,2048]............................................. ('fsdp', None). I0415 22:24:10.184168 124104330243200 train.py:162] train/z_loss Logical: float32[8,2048]............................................. ('activation_embed_and_logits_batch', 'activation_length'). I0415 22:24:10.184216 124104330243200 train.py:162] train/z_loss Physical: float32[8,2048]............................................. ('fsdp', None). W0415 22:24:10.816079 3298240 pjrt_executable.cc:642] Assume version compatibility. PjRt-IFRT does not track XLA executable versions. I0415 22:24:10.879342 124104330243200 max_utils.py:791] Total memory size: 3.6 GB, Output size: 1.5 GB, Temp size: 2.0 GB, Argument size: 1.5 GB, Host temp size: 0.0 GB. I0415 22:24:10.879889 124104330243200 max_utils.py:194] tensorboardX not available; using no-op SummaryWriter. I0415 22:24:10.880023 124104330243200 metric_logger.py:289] number parameters: 1.104 billion W0415 22:24:11.690403 3298240 pjrt_executable.cc:642] Assume version compatibility. PjRt-IFRT does not track XLA executable versions. I0415 22:24:11.760310 124104330243200 checkpointing.py:772] Waiting for step 10 to finish before checkpoint... I0415 22:24:11.876713 124104330243200 checkpointing.py:776] Waited 0.11639094352722168 seconds for step 10 to finish before starting checkpointing. I0415 22:24:11.877139 124104330243200 checkpoint_manager.py:1983] [process=0][thread=MainThread][wait_until_finished] No Save Finalize thread to wait for. Returning. I0415 22:24:11.877307 124104330243200 checkpoint_manager.py:1501] [process=0] Saving checkpoint at step 10 I0415 22:24:11.877603 124104330243200 async_checkpointer.py:452] [process=0] Started async saving checkpoint to gs://wanglance-maxtext/linen_ckpt_main_20260415_211126/linen_main_20260415_211126_16_zarr3_false/checkpoints/10. I0415 22:24:11.950901 124104330243200 signaling_client.py:373] Using ThreadSafeKeyValueSignalingClient I0415 22:24:12.034864 124104330243200 jax_array_handlers.py:347] Scheduling D2H of 39 prioritized jax.Array. I0415 22:24:12.035000 123996662138432 atomicity.py:137] Creating tmp directory gs://wanglance-maxtext/linen_ckpt_main_20260415_211126/linen_main_20260415_211126_16_zarr3_false/checkpoints/10 I0415 22:24:12.035088 124104330243200 replica_slices.py:410] 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 W0415 22:24:12.256057 3298240 pjrt_executable.cc:642] Assume version compatibility. PjRt-IFRT does not track XLA executable versions. W0415 22:24:12.271706 3298240 pjrt_executable.cc:642] Assume version compatibility. PjRt-IFRT does not track XLA executable versions. W0415 22:24:12.284249 3298240 pjrt_executable.cc:642] Assume version compatibility. PjRt-IFRT does not track XLA executable versions. W0415 22:24:12.288899 3298240 pjrt_executable.cc:642] Assume version compatibility. PjRt-IFRT does not track XLA executable versions. W0415 22:24:12.293061 3298240 pjrt_executable.cc:642] Assume version compatibility. PjRt-IFRT does not track XLA executable versions. W0415 22:24:12.296984 3298240 pjrt_executable.cc:642] Assume version compatibility. PjRt-IFRT does not track XLA executable versions. W0415 22:24:12.300894 3298240 pjrt_executable.cc:642] Assume version compatibility. PjRt-IFRT does not track XLA executable versions. W0415 22:24:12.305338 3298240 pjrt_executable.cc:642] Assume version compatibility. PjRt-IFRT does not track XLA executable versions. I0415 22:24:12.748708 123996651652672 atomicity.py:137] Creating tmp directory gs://wanglance-maxtext/linen_ckpt_main_20260415_211126/linen_main_20260415_211126_16_zarr3_false/checkpoints/10/items I0415 22:24:12.900100 123996616001088 checkpoint.py:188] Wrote Metadata={'item_handlers': None, 'metrics': {}, 'performance_metrics': {}, 'init_timestamp_nsecs': 1776291852657368296, 'commit_timestamp_nsecs': None, 'custom_metadata': {}}, json={"item_handlers": null, "metrics": {}, "performance_metrics": {}, "init_timestamp_nsecs": 1776291852657368296, "commit_timestamp_nsecs": null, "custom_metadata": {}} to gs://wanglance-maxtext/linen_ckpt_main_20260415_211126/linen_main_20260415_211126_16_zarr3_false/checkpoints/10/_CHECKPOINT_METADATA I0415 22:24:15.869690 124104330243200 base_pytree_checkpoint_handler.py:153] [process=0][thread=MainThread] Initiated "orbax.checkpoint._src.serialization.jax_array_handlers.ArrayHandler".serialize. Time taken: 3.835540s I0415 22:24:15.873509 124104330243200 base_pytree_checkpoint_handler.py:128] [process=0] /jax/checkpoint/write/blocking_gbytes_per_sec: 3.147 GiB/s (total gbytes: 12.3 GiB) (time elapsed: 3 seconds) (per-host) I0415 22:24:15.873578 124104330243200 base_pytree_checkpoint_handler.py:732] [process=0][thread=MainThread] Initiated Pytree async_save. Time taken: 3.921529s (batch_requests_ready=0.079521s, total_serialization_initiated=3.838284s, others=0.003724s) I0415 22:24:15.873647 124104330243200 composite_checkpoint_handler.py:715] [process=0][thread=MainThread] Initiated CompositeCheckpointHandler.async_save. Time taken: 3.922316s (all_items=0.000018s, per_item={'items': '0.00001788'}, temp_paths=3.922298) I0415 22:24:15.875024 123996662138432 async_checkpointer.py:79] [process=0][thread=async_save] Background save thread started. I0415 22:24:15.875182 124104330243200 async_checkpointer.py:561] Finished blocking save. Time taken: 3.997835s. Continuing background save to gs://wanglance-maxtext/linen_ckpt_main_20260415_211126/linen_main_20260415_211126_16_zarr3_false/checkpoints/10. I0415 22:24:15.875377 124104330243200 checkpoint_manager.py:1549] [process=0][thread=MainThread][step=10] Starting CheckpointManager Save Finalize thread=save_finalize I0415 22:24:15.875538 123996672624192 async_checkpointer.py:265] [process=0][thread=save_finalize] Waiting for background save thread=async_save. I0415 22:24:15.875609 124104330243200 standard_logger.py:34] {'step': 10, 'event_type': 'save', 'directory': 'gs://wanglance-maxtext/linen_ckpt_main_20260415_211126/linen_main_20260415_211126_16_zarr3_false/checkpoints', 'reached_preemption': False, 'preemption_received_at': None, 'synchronous': False, 'wait_for_prev_start_time': 1776291851.8771262, 'wait_for_prev_duration_secs': 4.00543212890625e-05, 'time_between_consecutive_saves_sec': None, 'checkpointer_blocking_start_time': 1776291851.8773282, 'checkpointer_blocking_duration_secs': 3.997973918914795, 'get_old_steps_start_time': 1776291855.875318, 'get_old_steps_duration_secs': 2.2649765014648438e-05, 'checkpoint_manager_blocking_start_time': 1776291851.8770154, 'checkpoint_manager_blocking_duration_secs': 3.9985692501068115} I0415 22:24:15.875714 124104330243200 checkpointing.py:408] Started an asynchronous checkpoint save for step 10 I0415 22:24:15.875751 124104330243200 max_utils.py:750] Memstats: After params initialized: I0415 22:24:15.875793 124104330243200 max_utils.py:756] Using (GB) 1.59 / 31.25 (5.088000%) on TPU_0(process=0,(0,0,0,0)) I0415 22:24:15.875818 124104330243200 max_utils.py:756] Using (GB) 1.59 / 31.25 (5.088000%) on TPU_1(process=0,(1,0,0,0)) I0415 22:24:15.875837 124104330243200 max_utils.py:756] Using (GB) 1.59 / 31.25 (5.088000%) on TPU_2(process=0,(0,1,0,0)) I0415 22:24:15.875861 124104330243200 max_utils.py:756] Using (GB) 1.59 / 31.25 (5.088000%) on TPU_3(process=0,(1,1,0,0)) I0415 22:24:15.875879 124104330243200 max_utils.py:756] Using (GB) 1.59 / 31.25 (5.088000%) on TPU_4(process=0,(0,2,0,0)) I0415 22:24:15.875896 124104330243200 max_utils.py:756] Using (GB) 1.59 / 31.25 (5.088000%) on TPU_5(process=0,(1,2,0,0)) I0415 22:24:15.875913 124104330243200 max_utils.py:756] Using (GB) 1.59 / 31.25 (5.088000%) on TPU_6(process=0,(0,3,0,0)) I0415 22:24:15.875929 124104330243200 max_utils.py:756] Using (GB) 1.59 / 31.25 (5.088000%) on TPU_7(process=0,(1,3,0,0)) W0415 22:24:15.881646 3298240 pjrt_executable.cc:642] Assume version compatibility. PjRt-IFRT does not track XLA executable versions. W0415 22:24:15.887465 3298240 pjrt_executable.cc:642] Assume version compatibility. PjRt-IFRT does not track XLA executable versions. W0415 22:24:15.891499 3298240 pjrt_executable.cc:642] Assume version compatibility. PjRt-IFRT does not track XLA executable versions. I0415 22:24:16.076185 124104330243200 metric_logger.py:185] completed step: 10, seconds: 0.880, TFLOP/s/device: 15.443, Tokens/s/device: 2327.757, total_weights: 16384, loss: 3.602 I0415 22:24:16.192646 124104330243200 metric_logger.py:185] completed step: 11, seconds: 4.314, TFLOP/s/device: 3.149, Tokens/s/device: 474.707, total_weights: 16384, loss: 3.101 I0415 22:24:16.311282 124104330243200 metric_logger.py:185] completed step: 12, seconds: 0.005, TFLOP/s/device: 2570.399, Tokens/s/device: 387438.517, total_weights: 16384, loss: 2.684 I0415 22:24:16.429541 124104330243200 metric_logger.py:185] completed step: 13, seconds: 0.117, TFLOP/s/device: 115.797, Tokens/s/device: 17454.149, total_weights: 16384, loss: 2.348 I0415 22:24:16.434067 124104330243200 checkpointing.py:772] Waiting for step 15 to finish before checkpoint... I0415 22:24:16.588356 123996630681152 array_metadata_store.py:203] [process=0][thread=array_type_handler] Wrote 39 array_metadata.ArrayMetadata to gs://wanglance-maxtext/linen_ckpt_main_20260415_211126/linen_main_20260415_211126_16_zarr3_false/checkpoints/10/items/array_metadatas/process_0 I0415 22:24:16.666091 124104330243200 checkpointing.py:776] Waited 0.23196911811828613 seconds for step 15 to finish before starting checkpointing. I0415 22:24:16.666464 124104330243200 checkpoint_manager.py:1994] [process=0][thread=MainThread][step=10][wait_until_finished] Waiting for Save Finalize thread (save_finalize) to complete. I0415 22:24:43.074708 123996651652672 base_pytree_checkpoint_handler.py:1217] [process=0][thread=write_metadata_after_commits] Commit + Array metadata written. Time taken: 27.200443s (commit=26.779344s, array_metadata_write=0.421099s) I0415 22:24:43.075837 123996662138432 base_pytree_checkpoint_handler.py:128] [process=0] /jax/checkpoint/write/gbytes_per_sec: 406.018 MiB/s (total gbytes: 12.3 GiB) (time elapsed: 31 seconds) (per-host) I0415 22:24:43.075942 123996662138432 async_checkpointer.py:90] [process=0][thread=async_save] 3 Handler Commit operations completed. Time taken: 27.200718s. I0415 22:24:43.491761 123996662138432 array_metadata_store.py:367] [process=0][thread=async_save] Skipped cross-host ArrayMetadata validation because only one process is found: process_index=0. I0415 22:24:43.679091 123996616001088 checkpoint.py:247] Updated Metadata={'item_handlers': {'items': 'orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeCheckpointHandler'}, 'metrics': {}, 'performance_metrics': {}, 'init_timestamp_nsecs': 1776291852657368296, 'commit_timestamp_nsecs': None, 'custom_metadata': {}} to gs://wanglance-maxtext/linen_ckpt_main_20260415_211126/linen_main_20260415_211126_16_zarr3_false/checkpoints/10/_CHECKPOINT_METADATA I0415 22:24:43.959743 123996662138432 base_pytree_checkpoint_handler.py:1342] [process=0][thread=async_save] Pytree save finalize (merge_ocdbt + ArrayMetadata validation) completed. Time taken: 0.609763s. use_zarr3=False, enable_post_merge_validation=True, directory=gs://wanglance-maxtext/linen_ckpt_main_20260415_211126/linen_main_20260415_211126_16_zarr3_false/checkpoints/10/items I0415 22:24:43.960473 123996662138432 atomicity.py:608] Finalizing gs://wanglance-maxtext/linen_ckpt_main_20260415_211126/linen_main_20260415_211126_16_zarr3_false/checkpoints/10/items I0415 22:24:44.199980 123996662138432 atomicity.py:608] Finalizing gs://wanglance-maxtext/linen_ckpt_main_20260415_211126/linen_main_20260415_211126_16_zarr3_false/checkpoints/10 I0415 22:24:44.855137 123996662138432 atomicity.py:794] [process=0][thread=async_save] Finished saving checkpoint (finalized tmp dir) to `gs://wanglance-maxtext/linen_ckpt_main_20260415_211126/linen_main_20260415_211126_16_zarr3_false/checkpoints/10`. I0415 22:24:44.855770 123996662138432 async_checkpointer.py:420] Finished async_save (blocking + background). Time taken: 32.978430s. directory=gs://wanglance-maxtext/linen_ckpt_main_20260415_211126/linen_main_20260415_211126_16_zarr3_false/checkpoints/10 I0415 22:24:44.855847 123996662138432 async_checkpointer.py:144] [process=0][thread=async_save] Background save thread done. Time taken: 28.980625s. I0415 22:24:44.855959 123996672624192 async_checkpointer.py:273] [process=0][thread=save_finalize] Done with waiting for background save thread=async_save. I0415 22:24:44.856008 123996672624192 async_checkpointer.py:283] [process=0][thread=save_finalize] No errors found in background save thread=async_save. I0415 22:24:44.856069 123996672624192 checkpoint_manager.py:2103] [process=0][thread=save_finalize][step=10] CheckpointManager Save Finalize is syncing with other hosts... I0415 22:24:44.856111 123996672624192 checkpoint_manager.py:2112] [process=0][thread=save_finalize][step=10] CheckpointManager Save Finalize is done on all hosts. I0415 22:24:44.856394 124104330243200 checkpoint_manager.py:2006] [process=0][thread=MainThread][step=10][wait_until_finished] Done waiting for Save Finalize thread (save_finalize) running at step=10. W0415 22:24:44.857159 124104330243200 checkpoint_manager.py:1441] Waiting for previous save to complete took 28.190695 seconds. If this number is high, consider checkpointing less frequently. I0415 22:24:44.858059 124104330243200 checkpoint_manager.py:1501] [process=0] Saving checkpoint at step 15 I0415 22:24:44.858343 124104330243200 async_checkpointer.py:452] [process=0] Started async saving checkpoint to gs://wanglance-maxtext/linen_ckpt_main_20260415_211126/linen_main_20260415_211126_16_zarr3_false/checkpoints/15. I0415 22:24:45.030620 123996672624192 atomicity.py:137] Creating tmp directory gs://wanglance-maxtext/linen_ckpt_main_20260415_211126/linen_main_20260415_211126_16_zarr3_false/checkpoints/15 I0415 22:24:45.041335 124104330243200 jax_array_handlers.py:347] Scheduling D2H of 39 prioritized jax.Array. I0415 22:24:45.041427 124104330243200 replica_slices.py:410] 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 I0415 22:24:45.713993 123996605515328 atomicity.py:137] Creating tmp directory gs://wanglance-maxtext/linen_ckpt_main_20260415_211126/linen_main_20260415_211126_16_zarr3_false/checkpoints/15/items I0415 22:24:49.813190 124104330243200 base_pytree_checkpoint_handler.py:153] [process=0][thread=MainThread] Initiated "orbax.checkpoint._src.serialization.jax_array_handlers.ArrayHandler".serialize. Time taken: 4.772266s I0415 22:24:49.816886 124104330243200 base_pytree_checkpoint_handler.py:128] [process=0] /jax/checkpoint/write/blocking_gbytes_per_sec: 2.533 GiB/s (total gbytes: 12.3 GiB) (time elapsed: 4 seconds) (per-host) I0415 22:24:49.816945 124104330243200 base_pytree_checkpoint_handler.py:732] [process=0][thread=MainThread] Initiated Pytree async_save. Time taken: 4.871258s (batch_requests_ready=0.092137s, total_serialization_initiated=4.775534s, others=0.003587s) I0415 22:24:49.817014 124104330243200 composite_checkpoint_handler.py:715] [process=0][thread=MainThread] Initiated CompositeCheckpointHandler.async_save. Time taken: 4.871753s (all_items=0.000014s, per_item={'items': '0.00001407'}, temp_paths=4.871739) I0415 22:24:49.818469 123996672624192 async_checkpointer.py:79] [process=0][thread=async_save] Background save thread started. I0415 22:24:49.818625 124104330243200 async_checkpointer.py:561] Finished blocking save. Time taken: 4.960522s. Continuing background save to gs://wanglance-maxtext/linen_ckpt_main_20260415_211126/linen_main_20260415_211126_16_zarr3_false/checkpoints/15. I0415 22:24:49.818826 124104330243200 checkpoint_manager.py:1549] [process=0][thread=MainThread][step=15] Starting CheckpointManager Save Finalize thread=save_finalize I0415 22:24:49.818965 123996662138432 async_checkpointer.py:265] [process=0][thread=save_finalize] Waiting for background save thread=async_save. I0415 22:24:49.819032 124104330243200 standard_logger.py:34] {'step': 15, 'event_type': 'save', 'directory': 'gs://wanglance-maxtext/linen_ckpt_main_20260415_211126/linen_main_20260415_211126_16_zarr3_false/checkpoints', 'reached_preemption': False, 'preemption_received_at': None, 'synchronous': False, 'wait_for_prev_start_time': 1776291856.666439, 'wait_for_prev_duration_secs': 28.19069504737854, 'time_between_consecutive_saves_sec': None, 'checkpointer_blocking_start_time': 1776291884.8580859, 'checkpointer_blocking_duration_secs': 4.960655927658081, 'get_old_steps_start_time': 1776291889.818757, 'get_old_steps_duration_secs': 2.5510787963867188e-05, 'checkpoint_manager_blocking_start_time': 1776291856.6663759, 'checkpoint_manager_blocking_duration_secs': 33.15263319015503} I0415 22:24:49.819155 124104330243200 checkpointing.py:408] Started an asynchronous checkpoint save for step 15 I0415 22:24:49.819724 124104330243200 metric_logger.py:185] completed step: 14, seconds: 0.118, TFLOP/s/device: 114.733, Tokens/s/device: 17293.792, total_weights: 16384, loss: 2.083 I0415 22:24:49.827001 124104330243200 metric_logger.py:185] completed step: 15, seconds: 0.118, TFLOP/s/device: 114.833, Tokens/s/device: 17308.846, total_weights: 16384, loss: 1.880 I0415 22:24:49.943169 124104330243200 metric_logger.py:185] completed step: 16, seconds: 33.391, TFLOP/s/device: 0.407, Tokens/s/device: 61.334, total_weights: 16384, loss: 1.726 I0415 22:24:50.061439 124104330243200 metric_logger.py:185] completed step: 17, seconds: 0.005, TFLOP/s/device: 2488.485, Tokens/s/device: 375091.575, total_weights: 16384, loss: 1.610 I0415 22:24:50.179636 124104330243200 metric_logger.py:185] completed step: 18, seconds: 0.117, TFLOP/s/device: 115.957, Tokens/s/device: 17478.280, total_weights: 16384, loss: 1.521 I0415 22:24:50.297533 124104330243200 checkpointing.py:772] Waiting for step 19 to finish before checkpoint... I0415 22:24:50.298430 124104330243200 checkpointing.py:776] Waited 0.0009188652038574219 seconds for step 19 to finish before starting checkpointing. I0415 22:24:50.298676 124104330243200 checkpoint_manager.py:1994] [process=0][thread=MainThread][step=15][wait_until_finished] Waiting for Save Finalize thread (save_finalize) to complete. I0415 22:24:50.561392 123998174185024 array_metadata_store.py:203] [process=0][thread=array_type_handler] Wrote 39 array_metadata.ArrayMetadata to gs://wanglance-maxtext/linen_ckpt_main_20260415_211126/linen_main_20260415_211126_16_zarr3_false/checkpoints/15/items/array_metadatas/process_0 I0415 22:25:17.898990 123996605515328 base_pytree_checkpoint_handler.py:1217] [process=0][thread=write_metadata_after_commits] Commit + Array metadata written. Time taken: 28.081399s (commit=27.643569s, array_metadata_write=0.437830s) I0415 22:25:17.899989 123996672624192 base_pytree_checkpoint_handler.py:128] [process=0] /jax/checkpoint/write/gbytes_per_sec: 383.465 MiB/s (total gbytes: 12.3 GiB) (time elapsed: 32 seconds) (per-host) I0415 22:25:17.900105 123996672624192 async_checkpointer.py:90] [process=0][thread=async_save] 3 Handler Commit operations completed. Time taken: 28.081438s. I0415 22:25:18.309888 123996672624192 array_metadata_store.py:367] [process=0][thread=async_save] Skipped cross-host ArrayMetadata validation because only one process is found: process_index=0. I0415 22:25:18.811650 123996672624192 base_pytree_checkpoint_handler.py:1342] [process=0][thread=async_save] Pytree save finalize (merge_ocdbt + ArrayMetadata validation) completed. Time taken: 0.653653s. use_zarr3=False, enable_post_merge_validation=True, directory=gs://wanglance-maxtext/linen_ckpt_main_20260415_211126/linen_main_20260415_211126_16_zarr3_false/checkpoints/15/items I0415 22:25:18.812354 123996672624192 atomicity.py:608] Finalizing gs://wanglance-maxtext/linen_ckpt_main_20260415_211126/linen_main_20260415_211126_16_zarr3_false/checkpoints/15/items I0415 22:25:19.045930 123996672624192 atomicity.py:608] Finalizing gs://wanglance-maxtext/linen_ckpt_main_20260415_211126/linen_main_20260415_211126_16_zarr3_false/checkpoints/15 I0415 22:25:19.721165 123996672624192 atomicity.py:794] [process=0][thread=async_save] Finished saving checkpoint (finalized tmp dir) to `gs://wanglance-maxtext/linen_ckpt_main_20260415_211126/linen_main_20260415_211126_16_zarr3_false/checkpoints/15`. I0415 22:25:19.721785 123996672624192 async_checkpointer.py:420] Finished async_save (blocking + background). Time taken: 34.863687s. directory=gs://wanglance-maxtext/linen_ckpt_main_20260415_211126/linen_main_20260415_211126_16_zarr3_false/checkpoints/15 I0415 22:25:19.721871 123996672624192 async_checkpointer.py:144] [process=0][thread=async_save] Background save thread done. Time taken: 29.903206s. I0415 22:25:19.721972 123996662138432 async_checkpointer.py:273] [process=0][thread=save_finalize] Done with waiting for background save thread=async_save. I0415 22:25:19.722020 123996662138432 async_checkpointer.py:283] [process=0][thread=save_finalize] No errors found in background save thread=async_save. I0415 22:25:19.722087 123996662138432 checkpoint_manager.py:2103] [process=0][thread=save_finalize][step=15] CheckpointManager Save Finalize is syncing with other hosts... I0415 22:25:19.722129 123996662138432 checkpoint_manager.py:2112] [process=0][thread=save_finalize][step=15] CheckpointManager Save Finalize is done on all hosts. I0415 22:25:19.722347 124104330243200 checkpoint_manager.py:2006] [process=0][thread=MainThread][step=15][wait_until_finished] Done waiting for Save Finalize thread (save_finalize) running at step=15. W0415 22:25:19.722460 124104330243200 checkpoint_manager.py:1441] Waiting for previous save to complete took 29.423793 seconds. If this number is high, consider checkpointing less frequently. I0415 22:25:19.722654 124104330243200 checkpoint_manager.py:1501] [process=0] Saving checkpoint at step 19 I0415 22:25:19.722948 124104330243200 async_checkpointer.py:452] [process=0] Started async saving checkpoint to gs://wanglance-maxtext/linen_ckpt_main_20260415_211126/linen_main_20260415_211126_16_zarr3_false/checkpoints/19. I0415 22:25:19.890699 123996672624192 atomicity.py:137] Creating tmp directory gs://wanglance-maxtext/linen_ckpt_main_20260415_211126/linen_main_20260415_211126_16_zarr3_false/checkpoints/19 I0415 22:25:19.908031 124104330243200 jax_array_handlers.py:347] Scheduling D2H of 39 prioritized jax.Array. I0415 22:25:19.908131 124104330243200 replica_slices.py:410] 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 I0415 22:25:20.598939 123998163699264 atomicity.py:137] Creating tmp directory gs://wanglance-maxtext/linen_ckpt_main_20260415_211126/linen_main_20260415_211126_16_zarr3_false/checkpoints/19/items I0415 22:25:27.309555 124104330243200 base_pytree_checkpoint_handler.py:153] [process=0][thread=MainThread] Initiated "orbax.checkpoint._src.serialization.jax_array_handlers.ArrayHandler".serialize. Time taken: 7.401866s I0415 22:25:27.310090 124104330243200 base_pytree_checkpoint_handler.py:128] [process=0] /jax/checkpoint/write/blocking_gbytes_per_sec: 1.646 GiB/s (total gbytes: 12.3 GiB) (time elapsed: 7 seconds) (per-host) I0415 22:25:27.310142 124104330243200 base_pytree_checkpoint_handler.py:732] [process=0][thread=MainThread] Initiated Pytree async_save. Time taken: 7.499692s (batch_requests_ready=0.094179s, total_serialization_initiated=7.405090s, others=0.000423s) I0415 22:25:27.310203 124104330243200 composite_checkpoint_handler.py:715] [process=0][thread=MainThread] Initiated CompositeCheckpointHandler.async_save. Time taken: 7.500164s (all_items=0.000011s, per_item={'items': '0.00001121'}, temp_paths=7.500153) I0415 22:25:27.314568 123996605515328 async_checkpointer.py:79] [process=0][thread=async_save] Background save thread started. I0415 22:25:27.314729 124104330243200 async_checkpointer.py:561] Finished blocking save. Time taken: 7.592037s. Continuing background save to gs://wanglance-maxtext/linen_ckpt_main_20260415_211126/linen_main_20260415_211126_16_zarr3_false/checkpoints/19. I0415 22:25:27.314917 124104330243200 checkpoint_manager.py:1549] [process=0][thread=MainThread][step=19] Starting CheckpointManager Save Finalize thread=save_finalize I0415 22:25:27.315102 123996662138432 async_checkpointer.py:265] [process=0][thread=save_finalize] Waiting for background save thread=async_save. I0415 22:25:27.315211 124104330243200 standard_logger.py:34] {'step': 19, 'event_type': 'save', 'directory': 'gs://wanglance-maxtext/linen_ckpt_main_20260415_211126/linen_main_20260415_211126_16_zarr3_false/checkpoints', 'reached_preemption': False, 'preemption_received_at': None, 'synchronous': False, 'wait_for_prev_start_time': 1776291890.2986515, 'wait_for_prev_duration_secs': 29.423793077468872, 'time_between_consecutive_saves_sec': 5.442525148391724, 'checkpointer_blocking_start_time': 1776291919.7226758, 'checkpointer_blocking_duration_secs': 7.592172145843506, 'get_old_steps_start_time': 1776291927.3148582, 'get_old_steps_duration_secs': 2.4557113647460938e-05, 'checkpoint_manager_blocking_start_time': 1776291890.2986195, 'checkpoint_manager_blocking_duration_secs': 37.01656770706177} I0415 22:25:27.315408 124104330243200 checkpointing.py:408] Started an asynchronous checkpoint save for step 19 I0415 22:25:27.315441 124104330243200 checkpoint_manager.py:1994] [process=0][thread=MainThread][step=19][wait_until_finished] Waiting for Save Finalize thread (save_finalize) to complete. I0415 22:25:28.079859 123998174185024 array_metadata_store.py:203] [process=0][thread=array_type_handler] Wrote 39 array_metadata.ArrayMetadata to gs://wanglance-maxtext/linen_ckpt_main_20260415_211126/linen_main_20260415_211126_16_zarr3_false/checkpoints/19/items/array_metadatas/process_0 I0415 22:25:53.188524 123998163699264 base_pytree_checkpoint_handler.py:1217] [process=0][thread=write_metadata_after_commits] Commit + Array metadata written. Time taken: 25.874599s (commit=25.447566s, array_metadata_write=0.427033s) I0415 22:25:53.189658 123996605515328 base_pytree_checkpoint_handler.py:128] [process=0] /jax/checkpoint/write/gbytes_per_sec: 378.583 MiB/s (total gbytes: 12.3 GiB) (time elapsed: 33 seconds) (per-host) I0415 22:25:53.189773 123996605515328 async_checkpointer.py:90] [process=0][thread=async_save] 3 Handler Commit operations completed. Time taken: 25.875001s. I0415 22:25:53.642064 123996605515328 array_metadata_store.py:367] [process=0][thread=async_save] Skipped cross-host ArrayMetadata validation because only one process is found: process_index=0. I0415 22:25:54.139538 123996605515328 base_pytree_checkpoint_handler.py:1342] [process=0][thread=async_save] Pytree save finalize (merge_ocdbt + ArrayMetadata validation) completed. Time taken: 0.656966s. use_zarr3=False, enable_post_merge_validation=True, directory=gs://wanglance-maxtext/linen_ckpt_main_20260415_211126/linen_main_20260415_211126_16_zarr3_false/checkpoints/19/items I0415 22:25:54.140270 123996605515328 atomicity.py:608] Finalizing gs://wanglance-maxtext/linen_ckpt_main_20260415_211126/linen_main_20260415_211126_16_zarr3_false/checkpoints/19/items I0415 22:25:54.362908 123996605515328 atomicity.py:608] Finalizing gs://wanglance-maxtext/linen_ckpt_main_20260415_211126/linen_main_20260415_211126_16_zarr3_false/checkpoints/19 I0415 22:25:55.032589 123996605515328 atomicity.py:794] [process=0][thread=async_save] Finished saving checkpoint (finalized tmp dir) to `gs://wanglance-maxtext/linen_ckpt_main_20260415_211126/linen_main_20260415_211126_16_zarr3_false/checkpoints/19`. I0415 22:25:55.033231 123996605515328 async_checkpointer.py:420] Finished async_save (blocking + background). Time taken: 35.310544s. directory=gs://wanglance-maxtext/linen_ckpt_main_20260415_211126/linen_main_20260415_211126_16_zarr3_false/checkpoints/19 I0415 22:25:55.033296 123996605515328 async_checkpointer.py:144] [process=0][thread=async_save] Background save thread done. Time taken: 27.718525s. I0415 22:25:55.033452 123996662138432 async_checkpointer.py:273] [process=0][thread=save_finalize] Done with waiting for background save thread=async_save. I0415 22:25:55.033558 123996662138432 async_checkpointer.py:283] [process=0][thread=save_finalize] No errors found in background save thread=async_save. I0415 22:25:55.033613 123996662138432 checkpoint_manager.py:2103] [process=0][thread=save_finalize][step=19] CheckpointManager Save Finalize is syncing with other hosts... I0415 22:25:55.033654 123996662138432 checkpoint_manager.py:2112] [process=0][thread=save_finalize][step=19] CheckpointManager Save Finalize is done on all hosts. I0415 22:25:55.033784 124104330243200 checkpoint_manager.py:2006] [process=0][thread=MainThread][step=19][wait_until_finished] Done waiting for Save Finalize thread (save_finalize) running at step=19. I0415 22:25:55.033914 124104330243200 checkpoint_manager.py:1983] [process=0][thread=MainThread][wait_until_finished] No Save Finalize thread to wait for. Returning. I0415 22:25:55.034778 124104330243200 metric_logger.py:185] completed step: 19, seconds: 0.118, TFLOP/s/device: 114.772, Tokens/s/device: 17299.635, total_weights: 16384, loss: 1.449 Per train step: Total TFLOPs: 13.59 split as 93.93% learnable weight flops and 6.07% attention flops