MaxView

XPK Start: Tue Apr 21 08:50:21 UTC 2026
2026-04-21 08:50:38.753747: 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)
I0421 08:50:42.324108 138379323594560 max_utils.py:273] Attempting to initialize the jax distributed system...
INFO:2026-04-21 08:50:51,364:jax._src.distributed:149: Starting JAX distributed service on [::]:8482
I0421 08:50:51.364160 138379323594560 distributed.py:149] Starting JAX distributed service on [::]:8482
INFO:2026-04-21 08:50:51,366:jax._src.distributed:166: Connecting to JAX distributed service on mt-07-distill-smoke-c4pjp-slice-job-0-0.mt-07-distill-smoke-c4pjp:8482
I0421 08:50:51.366410 138379323594560 distributed.py:166] Connecting to JAX distributed service on mt-07-distill-smoke-c4pjp-slice-job-0-0.mt-07-distill-smoke-c4pjp:8482
I0421 08:50:52.776022 138379323594560 max_utils.py:284] Jax distributed system initialized!
I0421 08:50:58.977718 138379323594560 max_utils.py:244] Jax distributed system is already initialized.
I0421 08:50:59.450324 138379323594560 max_utils.py:244] Jax distributed system is already initialized.
I0421 08:50:59.451509 138379323594560 pyconfig.py:432] Config param abort_on_inf_loss: True
I0421 08:50:59.451561 138379323594560 pyconfig.py:432] Config param abort_on_nan_loss: True
I0421 08:50:59.451587 138379323594560 pyconfig.py:432] Config param act_quantization_calibration_method: absmax
I0421 08:50:59.451608 138379323594560 pyconfig.py:432] Config param activation_dropout_for_audio: 0.0
I0421 08:50:59.451629 138379323594560 pyconfig.py:432] Config param activation_function_for_audio: gelu
I0421 08:50:59.451645 138379323594560 pyconfig.py:432] Config param activations_in_float32: False
I0421 08:50:59.451664 138379323594560 pyconfig.py:432] Config param adam_b1: 0.9
I0421 08:50:59.451685 138379323594560 pyconfig.py:432] Config param adam_b2: 0.95
I0421 08:50:59.451703 138379323594560 pyconfig.py:432] Config param adam_eps: 1e-08
I0421 08:50:59.451727 138379323594560 pyconfig.py:432] Config param adam_eps_root: 0.0
I0421 08:50:59.451756 138379323594560 pyconfig.py:432] Config param adam_weight_decay: 0.1
I0421 08:50:59.451772 138379323594560 pyconfig.py:432] Config param adamw_mask: []
I0421 08:50:59.451789 138379323594560 pyconfig.py:432] Config param add_bos: True
I0421 08:50:59.451806 138379323594560 pyconfig.py:432] Config param add_eos: True
I0421 08:50:59.451820 138379323594560 pyconfig.py:432] Config param allow_split_physical_axes: False
I0421 08:50:59.451837 138379323594560 pyconfig.py:432] Config param ar_cache_axis_order: 1,2,0,3
I0421 08:50:59.451852 138379323594560 pyconfig.py:432] Config param async_checkpointing: True
I0421 08:50:59.451870 138379323594560 pyconfig.py:432] Config param async_scheduling: False
I0421 08:50:59.451886 138379323594560 pyconfig.py:432] Config param attention: dot_product
I0421 08:50:59.451902 138379323594560 pyconfig.py:432] Config param attention_bias: False
I0421 08:50:59.451918 138379323594560 pyconfig.py:432] Config param attention_dropout_for_audio: 0.0
I0421 08:50:59.451936 138379323594560 pyconfig.py:432] Config param attention_out: RematLocation.REMAT
I0421 08:50:59.451957 138379323594560 pyconfig.py:432] Config param attention_output_dim: -1
I0421 08:50:59.451972 138379323594560 pyconfig.py:432] Config param attention_sink: False
I0421 08:50:59.451989 138379323594560 pyconfig.py:432] Config param attention_type: global
I0421 08:50:59.452006 138379323594560 pyconfig.py:432] Config param attn_logits_soft_cap: None
I0421 08:50:59.452021 138379323594560 pyconfig.py:432] Config param audio_path: 
I0421 08:50:59.452037 138379323594560 pyconfig.py:432] Config param audio_placeholder: <|audio|>
I0421 08:50:59.452052 138379323594560 pyconfig.py:432] Config param autoregressive_decode_assert: 
I0421 08:50:59.452068 138379323594560 pyconfig.py:432] Config param base_config: base.yml
I0421 08:50:59.452088 138379323594560 pyconfig.py:432] Config param base_emb_dim: 16
I0421 08:50:59.452104 138379323594560 pyconfig.py:432] Config param base_mlp_dim: 64
I0421 08:50:59.452120 138379323594560 pyconfig.py:432] Config param base_moe_mlp_dim: -1
I0421 08:50:59.452137 138379323594560 pyconfig.py:432] Config param base_num_decoder_layers: 1
I0421 08:50:59.452153 138379323594560 pyconfig.py:432] Config param base_num_kv_heads: 2
I0421 08:50:59.452169 138379323594560 pyconfig.py:432] Config param base_num_query_heads: 2
I0421 08:50:59.452185 138379323594560 pyconfig.py:432] Config param base_output_directory: 
I0421 08:50:59.452201 138379323594560 pyconfig.py:432] Config param batch_size: 1
I0421 08:50:59.452217 138379323594560 pyconfig.py:432] Config param batch_split_factor: 1
I0421 08:50:59.452232 138379323594560 pyconfig.py:432] Config param beta_fast: 32
I0421 08:50:59.452249 138379323594560 pyconfig.py:432] Config param beta_slow: 1
I0421 08:50:59.452265 138379323594560 pyconfig.py:432] Config param bwd_quantization_calibration_method: absmax
I0421 08:50:59.452281 138379323594560 pyconfig.py:432] Config param capacity_factor: -1.0
I0421 08:50:59.452297 138379323594560 pyconfig.py:432] Config param cast_logits_to_fp32: True
I0421 08:50:59.452313 138379323594560 pyconfig.py:432] Config param chat_template: 
I0421 08:50:59.452327 138379323594560 pyconfig.py:432] Config param chat_template_path: 
I0421 08:50:59.452345 138379323594560 pyconfig.py:432] Config param checkpoint_conversion_fn: None
I0421 08:50:59.452361 138379323594560 pyconfig.py:432] Config param checkpoint_dir: None
I0421 08:50:59.452379 138379323594560 pyconfig.py:432] Config param checkpoint_is_quantized: False
I0421 08:50:59.452396 138379323594560 pyconfig.py:432] Config param checkpoint_period: 2000
I0421 08:50:59.452413 138379323594560 pyconfig.py:432] Config param checkpoint_storage_concurrent_gb: 96
I0421 08:50:59.452428 138379323594560 pyconfig.py:432] Config param checkpoint_storage_target_data_file_size_bytes: 2147483648
I0421 08:50:59.452452 138379323594560 pyconfig.py:432] Config param checkpoint_storage_use_ocdbt: True
I0421 08:50:59.452476 138379323594560 pyconfig.py:432] Config param checkpoint_storage_use_zarr3: True
I0421 08:50:59.452499 138379323594560 pyconfig.py:432] Config param checkpoint_todelete_full_path: None
I0421 08:50:59.452533 138379323594560 pyconfig.py:432] Config param checkpoint_todelete_subdir: None
I0421 08:50:59.452550 138379323594560 pyconfig.py:432] Config param chips_per_vm: 4
I0421 08:50:59.452566 138379323594560 pyconfig.py:432] Config param chunk_attn_window_size: 0
I0421 08:50:59.452581 138379323594560 pyconfig.py:432] Config param collect_stack_trace: False
I0421 08:50:59.452595 138379323594560 pyconfig.py:432] Config param colocated_python_checkpointing: False
I0421 08:50:59.452611 138379323594560 pyconfig.py:432] Config param colocated_python_data_input: False
I0421 08:50:59.452627 138379323594560 pyconfig.py:432] Config param compile_topology: 
I0421 08:50:59.452642 138379323594560 pyconfig.py:432] Config param compile_topology_num_slices: -1
I0421 08:50:59.452658 138379323594560 pyconfig.py:432] Config param compile_xla_flags: 
I0421 08:50:59.452674 138379323594560 pyconfig.py:432] Config param compiled_trainstep_file: 
I0421 08:50:59.452689 138379323594560 pyconfig.py:432] Config param compute_axis_order: 0,1,2,3
I0421 08:50:59.452706 138379323594560 pyconfig.py:432] Config param constant_bound_config: []
I0421 08:50:59.452720 138379323594560 pyconfig.py:432] Config param context: RematLocation.REMAT
I0421 08:50:59.452759 138379323594560 pyconfig.py:432] Config param context_parallel_load_balance: True
I0421 08:50:59.452776 138379323594560 pyconfig.py:432] Config param context_parallel_size: 1
I0421 08:50:59.452793 138379323594560 pyconfig.py:432] Config param context_parallel_strategy: all_gather
I0421 08:50:59.452807 138379323594560 pyconfig.py:432] Config param context_sharding: context
I0421 08:50:59.452823 138379323594560 pyconfig.py:432] Config param conv_chunksize_for_audio: 500
I0421 08:50:59.452843 138379323594560 pyconfig.py:432] Config param conv_stride_for_vit: 14
I0421 08:50:59.452858 138379323594560 pyconfig.py:432] Config param cost_estimate_flops_bwd: -1
I0421 08:50:59.452874 138379323594560 pyconfig.py:432] Config param cost_estimate_flops_fwd: -1
I0421 08:50:59.452889 138379323594560 pyconfig.py:432] Config param custom_mesh: 
I0421 08:50:59.452903 138379323594560 pyconfig.py:432] Config param custom_mesh_and_rule: 
I0421 08:50:59.452919 138379323594560 pyconfig.py:432] Config param d_model_for_audio: 256
I0421 08:50:59.452934 138379323594560 pyconfig.py:432] Config param data_sharding: (('data', 'stage', 'fsdp', 'fsdp_transpose', 'sequence', 'context', 'context_autoregressive', 'tensor', 'tensor_transpose', 'tensor_sequence', 'expert', 'autoregressive'),)
I0421 08:50:59.452954 138379323594560 pyconfig.py:432] Config param data_shuffle_seed: 0
I0421 08:50:59.452969 138379323594560 pyconfig.py:432] Config param dataset_name: c4/en:3.0.1
I0421 08:50:59.452984 138379323594560 pyconfig.py:432] Config param dataset_path: 
I0421 08:50:59.453000 138379323594560 pyconfig.py:432] Config param dataset_type: DatasetType.HF
I0421 08:50:59.453018 138379323594560 pyconfig.py:432] Config param dcn_autoregressive_parallelism: 1
I0421 08:50:59.453033 138379323594560 pyconfig.py:432] Config param dcn_context_autoregressive_parallelism: 1
I0421 08:50:59.453047 138379323594560 pyconfig.py:432] Config param dcn_context_parallelism: 1
I0421 08:50:59.453063 138379323594560 pyconfig.py:432] Config param dcn_data_parallelism: -1
I0421 08:50:59.453079 138379323594560 pyconfig.py:432] Config param dcn_diloco_parallelism: 1
I0421 08:50:59.453098 138379323594560 pyconfig.py:432] Config param dcn_expert_parallelism: 1
I0421 08:50:59.453113 138379323594560 pyconfig.py:432] Config param dcn_fsdp_parallelism: 1
I0421 08:50:59.453129 138379323594560 pyconfig.py:432] Config param dcn_fsdp_transpose_parallelism: 1
I0421 08:50:59.453143 138379323594560 pyconfig.py:432] Config param dcn_parallelism: [1, -1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1]
I0421 08:50:59.453160 138379323594560 pyconfig.py:432] Config param dcn_pipeline_parallelism: 1
I0421 08:50:59.453175 138379323594560 pyconfig.py:432] Config param dcn_sequence_parallelism: 1
I0421 08:50:59.453190 138379323594560 pyconfig.py:432] Config param dcn_tensor_parallelism: 1
I0421 08:50:59.453206 138379323594560 pyconfig.py:432] Config param dcn_tensor_sequence_parallelism: 1
I0421 08:50:59.453220 138379323594560 pyconfig.py:432] Config param dcn_tensor_transpose_parallelism: 1
I0421 08:50:59.453237 138379323594560 pyconfig.py:432] Config param debug: {'rl': False}
I0421 08:50:59.453252 138379323594560 pyconfig.py:432] Config param debug_sharding: False
I0421 08:50:59.453268 138379323594560 pyconfig.py:432] Config param decode_sampling_nucleus_p: -1
I0421 08:50:59.453283 138379323594560 pyconfig.py:432] Config param decode_sampling_strategy: SamplingStrategy.GREEDY
I0421 08:50:59.453300 138379323594560 pyconfig.py:432] Config param decode_sampling_temperature: 1.0
I0421 08:50:59.453316 138379323594560 pyconfig.py:432] Config param decode_sampling_top_k: 0
I0421 08:50:59.453332 138379323594560 pyconfig.py:432] Config param decoder_block: DecoderBlockType.GPT3
I0421 08:50:59.453350 138379323594560 pyconfig.py:432] Config param decoder_layer_input: RematLocation.DEVICE
I0421 08:50:59.453366 138379323594560 pyconfig.py:432] Config param deepstack_visual_indexes_for_vit: []
I0421 08:50:59.453382 138379323594560 pyconfig.py:432] Config param degenerate_group_masking: True
I0421 08:50:59.453396 138379323594560 pyconfig.py:432] Config param dense_init_scale: 1.0
I0421 08:50:59.453412 138379323594560 pyconfig.py:432] Config param diloco_outer_lr: 0.3
I0421 08:50:59.453428 138379323594560 pyconfig.py:432] Config param diloco_outer_momentum: 0.9
I0421 08:50:59.453444 138379323594560 pyconfig.py:432] Config param diloco_sync_period: 36
I0421 08:50:59.453459 138379323594560 pyconfig.py:432] Config param distill_alpha: 0.5
I0421 08:50:59.453474 138379323594560 pyconfig.py:432] Config param distill_alpha_end: None
I0421 08:50:59.453490 138379323594560 pyconfig.py:432] Config param distill_alpha_schedule: constant
I0421 08:50:59.453505 138379323594560 pyconfig.py:432] Config param distill_beta: 0.0
I0421 08:50:59.453521 138379323594560 pyconfig.py:432] Config param distill_beta_end: None
I0421 08:50:59.453535 138379323594560 pyconfig.py:432] Config param distill_beta_schedule: constant
I0421 08:50:59.453551 138379323594560 pyconfig.py:432] Config param distill_feature_loss_type: cosine
I0421 08:50:59.453567 138379323594560 pyconfig.py:432] Config param distill_layer_indices: None
I0421 08:50:59.453581 138379323594560 pyconfig.py:432] Config param distill_temperature: 1.0
I0421 08:50:59.453597 138379323594560 pyconfig.py:432] Config param distill_temperature_end: None
I0421 08:50:59.453612 138379323594560 pyconfig.py:432] Config param distill_temperature_schedule: constant
I0421 08:50:59.453628 138379323594560 pyconfig.py:432] Config param downsample_hidden_size_for_audio: 256
I0421 08:50:59.453642 138379323594560 pyconfig.py:432] Config param dpo_beta: 0.1
I0421 08:50:59.453657 138379323594560 pyconfig.py:432] Config param dpo_label_smoothing: 0.0
I0421 08:50:59.453673 138379323594560 pyconfig.py:432] Config param dq_reduction_steps: 0
I0421 08:50:59.453688 138379323594560 pyconfig.py:432] Config param dropout_rate: 0.0
I0421 08:50:59.453704 138379323594560 pyconfig.py:432] Config param dtype: bfloat16
I0421 08:50:59.453746 138379323594560 pyconfig.py:432] Config param dtype_mm: float32
I0421 08:50:59.453763 138379323594560 pyconfig.py:432] Config param dump_hlo: False
I0421 08:50:59.453778 138379323594560 pyconfig.py:432] Config param dump_hlo_delete_local_after: True
I0421 08:50:59.453793 138379323594560 pyconfig.py:432] Config param dump_hlo_gcs_dir: gpt3-52k_2026-04-21-08-50/xla_dump
I0421 08:50:59.453809 138379323594560 pyconfig.py:432] Config param dump_hlo_local_dir: /tmp/xla_dump/
I0421 08:50:59.453823 138379323594560 pyconfig.py:432] Config param dump_hlo_local_module_name: jit_train_step
I0421 08:50:59.453839 138379323594560 pyconfig.py:432] Config param dump_hlo_module_name: jit_train_step
I0421 08:50:59.453855 138379323594560 pyconfig.py:432] Config param dump_hlo_upload_all: False
I0421 08:50:59.453869 138379323594560 pyconfig.py:432] Config param dump_hlo_xla_flags: 
I0421 08:50:59.453885 138379323594560 pyconfig.py:432] Config param dump_jaxpr: False
I0421 08:50:59.453900 138379323594560 pyconfig.py:432] Config param dump_jaxpr_delete_local_after: True
I0421 08:50:59.453915 138379323594560 pyconfig.py:432] Config param dump_jaxpr_gcs_dir: gpt3-52k_2026-04-21-08-50/jaxpr_dump
I0421 08:50:59.453931 138379323594560 pyconfig.py:432] Config param dump_jaxpr_local_dir: /tmp/jaxpr_dump/
I0421 08:50:59.453945 138379323594560 pyconfig.py:432] Config param dump_step: -1
I0421 08:50:59.453961 138379323594560 pyconfig.py:432] Config param elastic_enabled: False
I0421 08:50:59.453975 138379323594560 pyconfig.py:432] Config param elastic_max_retries: 10
I0421 08:50:59.453991 138379323594560 pyconfig.py:432] Config param elastic_timeout_seconds: 300
I0421 08:50:59.454005 138379323594560 pyconfig.py:432] Config param emb_dim: 16
I0421 08:50:59.454020 138379323594560 pyconfig.py:432] Config param enable_autocheckpoint: False
I0421 08:50:59.454036 138379323594560 pyconfig.py:432] Config param enable_checkpoint_cloud_logger: False
I0421 08:50:59.454050 138379323594560 pyconfig.py:432] Config param enable_checkpointing: True
I0421 08:50:59.454066 138379323594560 pyconfig.py:432] Config param enable_continuous_checkpointing: False
I0421 08:50:59.454084 138379323594560 pyconfig.py:432] Config param enable_data_shuffling: True
I0421 08:50:59.454101 138379323594560 pyconfig.py:432] Config param enable_diloco: False
I0421 08:50:59.454115 138379323594560 pyconfig.py:432] Config param enable_dp_attention: False
I0421 08:50:59.454131 138379323594560 pyconfig.py:432] Config param enable_dropout: False
I0421 08:50:59.454145 138379323594560 pyconfig.py:432] Config param enable_emergency_checkpoint: False
I0421 08:50:59.454160 138379323594560 pyconfig.py:432] Config param enable_expert_parallel: False
I0421 08:50:59.454175 138379323594560 pyconfig.py:432] Config param enable_gcp_goodput_metrics: True
I0421 08:50:59.454189 138379323594560 pyconfig.py:432] Config param enable_gcp_step_deviation_metrics: True
I0421 08:50:59.454205 138379323594560 pyconfig.py:432] Config param enable_goodput_recording: False
I0421 08:50:59.454221 138379323594560 pyconfig.py:432] Config param enable_jax_profiler: False
I0421 08:50:59.454235 138379323594560 pyconfig.py:432] Config param enable_llm_inference_pool: False
I0421 08:50:59.454251 138379323594560 pyconfig.py:432] Config param enable_model_warmup: False
I0421 08:50:59.454265 138379323594560 pyconfig.py:432] Config param enable_multi_tier_checkpointing: False
I0421 08:50:59.454280 138379323594560 pyconfig.py:432] Config param enable_nnx: False
I0421 08:50:59.454296 138379323594560 pyconfig.py:432] Config param enable_orbax_v1: False
I0421 08:50:59.454310 138379323594560 pyconfig.py:432] Config param enable_padding_causal_mask: True
I0421 08:50:59.454326 138379323594560 pyconfig.py:432] Config param enable_pathways_goodput: False
I0421 08:50:59.454340 138379323594560 pyconfig.py:432] Config param enable_prefix_caching: False
I0421 08:50:59.454356 138379323594560 pyconfig.py:432] Config param enable_rampup_batch_size: False
I0421 08:50:59.454370 138379323594560 pyconfig.py:432] Config param enable_single_controller: False
I0421 08:50:59.454386 138379323594560 pyconfig.py:432] Config param enable_single_replica_ckpt_restoring: False
I0421 08:50:59.454400 138379323594560 pyconfig.py:432] Config param enable_tensorboard: True
I0421 08:50:59.454415 138379323594560 pyconfig.py:432] Config param enable_tunix_perf_metrics: False
I0421 08:50:59.454430 138379323594560 pyconfig.py:432] Config param encoder_attention_heads_for_audio: 4
I0421 08:50:59.454444 138379323594560 pyconfig.py:432] Config param encoder_ffn_dim_for_audio: 512
I0421 08:50:59.454460 138379323594560 pyconfig.py:432] Config param encoder_layers_for_audio: 2
I0421 08:50:59.454476 138379323594560 pyconfig.py:432] Config param engram: RematLocation.REMAT
I0421 08:50:59.454492 138379323594560 pyconfig.py:432] Config param engram_head_dim: 1280
I0421 08:50:59.454507 138379323594560 pyconfig.py:432] Config param engram_kernel_size: 4
I0421 08:50:59.454522 138379323594560 pyconfig.py:432] Config param engram_layers: []
I0421 08:50:59.454536 138379323594560 pyconfig.py:432] Config param engram_max_ngram_size: 3
I0421 08:50:59.454552 138379323594560 pyconfig.py:432] Config param engram_num_heads: 8
I0421 08:50:59.454568 138379323594560 pyconfig.py:432] Config param engram_seed: 0
I0421 08:50:59.454584 138379323594560 pyconfig.py:432] Config param engram_vocab_bases: []
I0421 08:50:59.454600 138379323594560 pyconfig.py:432] Config param epsilon_high: None
I0421 08:50:59.454614 138379323594560 pyconfig.py:432] Config param eval_corr_lst: False
I0421 08:50:59.454630 138379323594560 pyconfig.py:432] Config param eval_data_columns: ['text']
I0421 08:50:59.454645 138379323594560 pyconfig.py:432] Config param eval_dataset_name: c4/en:3.0.1
I0421 08:50:59.454661 138379323594560 pyconfig.py:432] Config param eval_image_column: image
I0421 08:50:59.454676 138379323594560 pyconfig.py:432] Config param eval_interval: -1
I0421 08:50:59.454692 138379323594560 pyconfig.py:432] Config param eval_make_lst: False
I0421 08:50:59.454706 138379323594560 pyconfig.py:432] Config param eval_per_device_batch_size: 2
I0421 08:50:59.454722 138379323594560 pyconfig.py:432] Config param eval_sampling_strategy: greedy
I0421 08:50:59.454745 138379323594560 pyconfig.py:432] Config param eval_split: validation
I0421 08:50:59.454761 138379323594560 pyconfig.py:432] Config param eval_steps: -1
I0421 08:50:59.454777 138379323594560 pyconfig.py:432] Config param expansion_factor_real_data: -1.0
I0421 08:50:59.454792 138379323594560 pyconfig.py:432] Config param final_logits_soft_cap: None
I0421 08:50:59.454807 138379323594560 pyconfig.py:432] Config param first_num_dense_layers: 0
I0421 08:50:59.454823 138379323594560 pyconfig.py:432] Config param float32_gate_logits: False
I0421 08:50:59.454839 138379323594560 pyconfig.py:432] Config param float32_logits: False
I0421 08:50:59.454854 138379323594560 pyconfig.py:432] Config param float32_qk_product: False
I0421 08:50:59.454869 138379323594560 pyconfig.py:432] Config param float32_weight_sum: True
I0421 08:50:59.454884 138379323594560 pyconfig.py:432] Config param force_q_layout: False
I0421 08:50:59.454899 138379323594560 pyconfig.py:432] Config param force_unroll: False
I0421 08:50:59.454915 138379323594560 pyconfig.py:432] Config param freeze_audio_encoder_params: True
I0421 08:50:59.454929 138379323594560 pyconfig.py:432] Config param freeze_vision_encoder_params: True
I0421 08:50:59.454945 138379323594560 pyconfig.py:432] Config param fused_mlp: False
I0421 08:50:59.454960 138379323594560 pyconfig.py:432] Config param fused_qkv: True
I0421 08:50:59.454975 138379323594560 pyconfig.py:432] Config param gcs_metrics: False
I0421 08:50:59.454990 138379323594560 pyconfig.py:432] Config param gdn_chunk_size: 64
I0421 08:50:59.455006 138379323594560 pyconfig.py:432] Config param gdn_conv_kernel_dim: 4
I0421 08:50:59.455022 138379323594560 pyconfig.py:432] Config param gdn_key_head_dim: 128
I0421 08:50:59.455038 138379323594560 pyconfig.py:432] Config param gdn_num_key_heads: 16
I0421 08:50:59.455052 138379323594560 pyconfig.py:432] Config param gdn_num_value_heads: 32
I0421 08:50:59.455067 138379323594560 pyconfig.py:432] Config param gdn_value_head_dim: 128
I0421 08:50:59.455086 138379323594560 pyconfig.py:432] Config param generate_padding_batch_eval: False
I0421 08:50:59.455101 138379323594560 pyconfig.py:432] Config param generate_padding_batch_train: False
I0421 08:50:59.455116 138379323594560 pyconfig.py:432] Config param generate_slice: v5e-16
I0421 08:50:59.455132 138379323594560 pyconfig.py:432] Config param generation_configs: {}
I0421 08:50:59.455148 138379323594560 pyconfig.py:432] Config param global_batch_size_to_eval_on: 64
I0421 08:50:59.455164 138379323594560 pyconfig.py:432] Config param global_batch_size_to_load: 512
I0421 08:50:59.455178 138379323594560 pyconfig.py:432] Config param global_batch_size_to_load_eval: 64
I0421 08:50:59.455194 138379323594560 pyconfig.py:432] Config param global_batch_size_to_load_increment: None
I0421 08:50:59.455209 138379323594560 pyconfig.py:432] Config param global_batch_size_to_load_start: None
I0421 08:50:59.455224 138379323594560 pyconfig.py:432] Config param global_batch_size_to_train_on: 512
I0421 08:50:59.455239 138379323594560 pyconfig.py:432] Config param global_head_dim: 0
I0421 08:50:59.455255 138379323594560 pyconfig.py:432] Config param global_num_kv_heads: 0
I0421 08:50:59.455270 138379323594560 pyconfig.py:432] Config param global_parameter_scale: 1
I0421 08:50:59.455284 138379323594560 pyconfig.py:432] Config param global_rampup_samples: 500
I0421 08:50:59.455300 138379323594560 pyconfig.py:432] Config param global_rope_max_timescale: -1
I0421 08:50:59.455314 138379323594560 pyconfig.py:432] Config param global_rope_proportion: 0.25
I0421 08:50:59.455330 138379323594560 pyconfig.py:432] Config param goodput_upload_interval_seconds: 30
I0421 08:50:59.455346 138379323594560 pyconfig.py:432] Config param grad_dtype: float32
I0421 08:50:59.455379 138379323594560 pyconfig.py:432] Config param gradient_accumulation_steps: 8
I0421 08:50:59.455395 138379323594560 pyconfig.py:432] Config param gradient_clipping_threshold: 1.0
I0421 08:50:59.455412 138379323594560 pyconfig.py:432] Config param grain_data_source_max_workers: 16
I0421 08:50:59.455428 138379323594560 pyconfig.py:432] Config param grain_eval_files: 
I0421 08:50:59.455442 138379323594560 pyconfig.py:432] Config param grain_file_type: arrayrecord
I0421 08:50:59.455458 138379323594560 pyconfig.py:432] Config param grain_num_threads: 16
I0421 08:50:59.455474 138379323594560 pyconfig.py:432] Config param grain_num_threads_eval: 16
I0421 08:50:59.455490 138379323594560 pyconfig.py:432] Config param grain_packing_type: first_fit
I0421 08:50:59.455505 138379323594560 pyconfig.py:432] Config param grain_per_worker_buffer_size: 1
I0421 08:50:59.455521 138379323594560 pyconfig.py:432] Config param grain_per_worker_buffer_size_eval: 1
I0421 08:50:59.455537 138379323594560 pyconfig.py:432] Config param grain_prefetch_buffer_size: 500
I0421 08:50:59.455552 138379323594560 pyconfig.py:432] Config param grain_prefetch_buffer_size_eval: 500
I0421 08:50:59.455568 138379323594560 pyconfig.py:432] Config param grain_ram_budget_mb: 1024
I0421 08:50:59.455582 138379323594560 pyconfig.py:432] Config param grain_shuffle_buffer_size: 100
I0421 08:50:59.455598 138379323594560 pyconfig.py:432] Config param grain_train_files: 
I0421 08:50:59.455614 138379323594560 pyconfig.py:432] Config param grain_train_mixture_config_path: 
I0421 08:50:59.455630 138379323594560 pyconfig.py:432] Config param grain_worker_count: 1
I0421 08:50:59.455644 138379323594560 pyconfig.py:432] Config param grain_worker_count_eval: 1
I0421 08:50:59.455660 138379323594560 pyconfig.py:432] Config param grpo_beta: 0.08
I0421 08:50:59.455675 138379323594560 pyconfig.py:432] Config param grpo_epsilon: 0.2
I0421 08:50:59.455691 138379323594560 pyconfig.py:432] Config param hardware: tpu
I0421 08:50:59.455706 138379323594560 pyconfig.py:432] Config param hbm_utilization_vllm: 0.72
I0421 08:50:59.455723 138379323594560 pyconfig.py:432] Config param head_dim: 8
I0421 08:50:59.455752 138379323594560 pyconfig.py:432] Config param heartbeat_reporting_interval_in_seconds: 5
I0421 08:50:59.455767 138379323594560 pyconfig.py:432] Config param hf_data_dir: None
I0421 08:50:59.455783 138379323594560 pyconfig.py:432] Config param hf_eval_files: None
I0421 08:50:59.455800 138379323594560 pyconfig.py:432] Config param hf_eval_split: None
I0421 08:50:59.455816 138379323594560 pyconfig.py:432] Config param hf_name: None
I0421 08:50:59.455830 138379323594560 pyconfig.py:432] Config param hf_path: OptimalScale/ClimbMix
I0421 08:50:59.455846 138379323594560 pyconfig.py:432] Config param hf_train_files: None
I0421 08:50:59.455862 138379323594560 pyconfig.py:432] Config param hidden_size_for_vit: 1408
I0421 08:50:59.455877 138379323594560 pyconfig.py:432] Config param hide_profiler_step_metric: False
I0421 08:50:59.455893 138379323594560 pyconfig.py:432] Config param ici_autoregressive_parallelism: 1
I0421 08:50:59.455907 138379323594560 pyconfig.py:432] Config param ici_context_autoregressive_parallelism: 1
I0421 08:50:59.455923 138379323594560 pyconfig.py:432] Config param ici_context_parallelism: 1
I0421 08:50:59.455937 138379323594560 pyconfig.py:432] Config param ici_data_parallelism: 1
I0421 08:50:59.455953 138379323594560 pyconfig.py:432] Config param ici_diloco_parallelism: 1
I0421 08:50:59.455967 138379323594560 pyconfig.py:432] Config param ici_expert_parallelism: 1
I0421 08:50:59.455983 138379323594560 pyconfig.py:432] Config param ici_fsdp_parallelism: -1
I0421 08:50:59.455998 138379323594560 pyconfig.py:432] Config param ici_fsdp_transpose_parallelism: 1
I0421 08:50:59.456013 138379323594560 pyconfig.py:432] Config param ici_parallelism: [1, 1, 1, -1, 1, 1, 1, 1, 1, 1, 1, 1, 1]
I0421 08:50:59.456030 138379323594560 pyconfig.py:432] Config param ici_pipeline_parallelism: 1
I0421 08:50:59.456044 138379323594560 pyconfig.py:432] Config param ici_sequence_parallelism: 1
I0421 08:50:59.456060 138379323594560 pyconfig.py:432] Config param ici_tensor_parallelism: 1
I0421 08:50:59.456074 138379323594560 pyconfig.py:432] Config param ici_tensor_sequence_parallelism: 1
I0421 08:50:59.456093 138379323594560 pyconfig.py:432] Config param ici_tensor_transpose_parallelism: 1
I0421 08:50:59.456108 138379323594560 pyconfig.py:432] Config param image_path: 
I0421 08:50:59.456123 138379323594560 pyconfig.py:432] Config param image_placeholder: <|image|>
I0421 08:50:59.456139 138379323594560 pyconfig.py:432] Config param image_size_for_vit: 896
I0421 08:50:59.456155 138379323594560 pyconfig.py:432] Config param indexer_head_dim: 128
I0421 08:50:59.456171 138379323594560 pyconfig.py:432] Config param indexer_loss_scaling_factor: 0.0
I0421 08:50:59.456188 138379323594560 pyconfig.py:432] Config param indexer_n_heads: 64
I0421 08:50:59.456205 138379323594560 pyconfig.py:432] Config param indexer_sparse_training: False
I0421 08:50:59.456220 138379323594560 pyconfig.py:432] Config param indexer_topk: 2048
I0421 08:50:59.456234 138379323594560 pyconfig.py:432] Config param inference_benchmark_test: False
I0421 08:50:59.456249 138379323594560 pyconfig.py:432] Config param inference_metadata_file: 
I0421 08:50:59.456265 138379323594560 pyconfig.py:432] Config param inference_microbenchmark_log_file_path: 
I0421 08:50:59.456279 138379323594560 pyconfig.py:432] Config param inference_microbenchmark_loop_iters: 10
I0421 08:50:59.456295 138379323594560 pyconfig.py:432] Config param inference_microbenchmark_num_samples: [1, 2, 3, 4, 5]
I0421 08:50:59.456311 138379323594560 pyconfig.py:432] Config param inference_microbenchmark_prefill_lengths: 64,128,256,512,1024
I0421 08:50:59.456326 138379323594560 pyconfig.py:432] Config param inference_microbenchmark_stages: prefill,generate
I0421 08:50:59.456342 138379323594560 pyconfig.py:432] Config param inference_server: MaxtextInterleavedServer
I0421 08:50:59.456358 138379323594560 pyconfig.py:432] Config param inhomogeneous_layer_cycle_interval: 1
I0421 08:50:59.456372 138379323594560 pyconfig.py:432] Config param init_weights_seed: 0
I0421 08:50:59.456388 138379323594560 pyconfig.py:432] Config param input_data_sharding_logical_axes: ['activation_embed_and_logits_batch', 'activation_norm_length']
I0421 08:50:59.456404 138379323594560 pyconfig.py:432] Config param interleave_moe_layer_step: 1
I0421 08:50:59.456419 138379323594560 pyconfig.py:432] Config param intermediate_size_for_vit: 5632
I0421 08:50:59.456434 138379323594560 pyconfig.py:432] Config param internal_compile: False
I0421 08:50:59.456450 138379323594560 pyconfig.py:432] Config param internal_compile_num_devices: -1
I0421 08:50:59.456465 138379323594560 pyconfig.py:432] Config param jax_cache_dir: ~/jax_cache
I0421 08:50:59.456480 138379323594560 pyconfig.py:432] Config param jax_debug_log_modules: 
I0421 08:50:59.456496 138379323594560 pyconfig.py:432] Config param jax_distributed_initialization_timeout: 300
I0421 08:50:59.456510 138379323594560 pyconfig.py:432] Config param jax_profiler_port: 9999
I0421 08:50:59.456526 138379323594560 pyconfig.py:432] Config param key_proj: RematLocation.REMAT
I0421 08:50:59.456541 138379323594560 pyconfig.py:432] Config param kv_cache_buffer: 256
I0421 08:50:59.456556 138379323594560 pyconfig.py:432] Config param kv_lora_rank: 512
I0421 08:50:59.456572 138379323594560 pyconfig.py:432] Config param kv_quant_axis: KvQuantAxis.HEADS_AND_DKV
I0421 08:50:59.456590 138379323594560 pyconfig.py:432] Config param kv_quant_dtype: int8
I0421 08:50:59.456606 138379323594560 pyconfig.py:432] Config param kv_wa_proj: RematLocation.REMAT
I0421 08:50:59.456622 138379323594560 pyconfig.py:432] Config param learning_rate: 0.0002
I0421 08:50:59.456637 138379323594560 pyconfig.py:432] Config param learning_rate_final_fraction: 0.1
I0421 08:50:59.456653 138379323594560 pyconfig.py:432] Config param learning_rate_schedule_steps: 200000
I0421 08:50:59.456669 138379323594560 pyconfig.py:432] Config param load_balance_loss_weight: 0.0
I0421 08:50:59.456683 138379323594560 pyconfig.py:432] Config param load_checkpoint_only_once: False
I0421 08:50:59.456700 138379323594560 pyconfig.py:432] Config param load_from_prefill_dir: False
I0421 08:50:59.456715 138379323594560 pyconfig.py:432] Config param load_full_state_path: 
I0421 08:50:59.456730 138379323594560 pyconfig.py:432] Config param load_parameters_path: gs://lance-maxtext/pt_seed_ckpts/pt_seed_ckpts/pt_seed_ckpt_gpt352k_v32k_linen/checkpoints/4/items
I0421 08:50:59.456755 138379323594560 pyconfig.py:432] Config param local_checkpoint_directory: 
I0421 08:50:59.456769 138379323594560 pyconfig.py:432] Config param local_checkpoint_period: 0
I0421 08:50:59.456785 138379323594560 pyconfig.py:432] Config param local_rope_max_timescale: -1
I0421 08:50:59.456799 138379323594560 pyconfig.py:432] Config param local_rope_proportion: 1.0
I0421 08:50:59.456815 138379323594560 pyconfig.py:432] Config param log_config: True
I0421 08:50:59.456830 138379323594560 pyconfig.py:432] Config param log_period: 10
I0421 08:50:59.456845 138379323594560 pyconfig.py:432] Config param logical_axis_rules: (('activation_embed_and_logits_batch', ('data', 'stage', 'fsdp', 'fsdp_transpose', 'expert')), ('activation_embed_and_logits_batch_sequence', ('data', 'stage', 'fsdp', 'fsdp_transpose', 'sequence', 'context', 'expert')), ('activation_vocab', ('tensor', 'tensor_transpose', 'tensor_sequence')), ('activation_vocab', ('tensor', 'tensor_transpose')), ('activation_vocab', 'tensor_sequence'), ('activation_vocab', ('sequence', 'context')), ('vocab', ('tensor', 'tensor_transpose', 'tensor_sequence', 'autoregressive')), ('embed_vocab', ('fsdp', 'fsdp_transpose', 'sequence', 'context', 'expert')), ('activation_heads', ('tensor', 'tensor_transpose', 'sequence', 'tensor_sequence', 'autoregressive')), ('activation_kv_heads', ('tensor', 'tensor_transpose', 'sequence', 'tensor_sequence')), ('activation_attn_length', ('sequence', 'context')), ('activation_attn_length', ('context',)), ('activation_q_length', ('context',)), ('activation_kv_length', ()), ('activation_attn_embed', ('tensor', 'tensor_transpose')), ('activation_kv', ('tensor', 'tensor_transpose', 'tensor_sequence')), ('activation_kv_batch', ('data', 'fsdp', 'fsdp_transpose', 'expert')), ('activation_kv_head_dim', ('tensor', 'tensor_transpose', 'tensor_sequence')), ('heads', ('tensor', 'tensor_transpose', 'tensor_sequence', 'autoregressive')), ('q_heads', ('tensor', 'tensor_transpose', 'tensor_sequence', 'autoregressive')), ('kv_heads', ('tensor', 'tensor_transpose', 'tensor_sequence', 'autoregressive')), ('qkv', ()), ('kv', ()), ('kv_head_dim', ()), ('q_lora', ('fsdp', 'fsdp_transpose', 'sequence', 'context', 'tensor_transpose', 'expert')), ('q_lora', ('fsdp', 'sequence', 'context', 'tensor_transpose', 'expert')), ('q_lora', ('fsdp', 'fsdp_transpose', 'sequence', 'context', 'expert')), ('q_lora', ('fsdp', 'sequence', 'context', 'expert')), ('q_lora_up_proj', ()), ('kv_lora', ('fsdp', 'fsdp_transpose', 'sequence', 'context', 'tensor_transpose', 'expert')), ('kv_lora', ('fsdp', 'sequence', 'context', 'tensor_transpose', 'expert')), ('kv_lora', ('fsdp', 'fsdp_transpose', 'sequence', 'context', 'expert')), ('kv_lora', ('fsdp', 'sequence', 'context', 'expert')), ('kv_lora_up_proj', ()), ('activation_batch_moe', ('data', 'fsdp', 'fsdp_transpose')), ('activation_length_moe', ('sequence', 'context')), ('activation_length_moe', ('context',)), ('activation_norm_length_moe', ('tensor_sequence', 'context', 'sequence')), ('activation_embed_moe', ('tensor', 'tensor_transpose')), ('activation_mlp_moe', ('tensor', 'tensor_transpose', 'tensor_sequence')), ('activation_exp', ('expert',)), ('exp', 'expert'), ('mlp_moe', ('fsdp_transpose', 'tensor', 'tensor_sequence', 'autoregressive')), ('embed_moe', ('fsdp', 'fsdp_transpose', 'sequence', 'tensor_transpose', 'context')), ('embed_moe', ('fsdp', 'sequence', 'tensor_transpose', 'context')), ('embed_moe', ('fsdp', 'fsdp_transpose', 'sequence', 'context')), ('embed_moe', ('fsdp', 'sequence', 'context')), ('activation_mlp', ('tensor', 'tensor_transpose', 'tensor_sequence')), ('activation_batch', ('data', 'fsdp', 'fsdp_transpose', 'expert')), ('activation_length', ('sequence', 'context')), ('activation_length', ('context',)), ('activation_norm_length', ('tensor_sequence', 'context', 'sequence')), ('activation_embed', ('tensor', 'tensor_transpose')), ('activation_stage', 'stage'), ('mlp', ('fsdp_transpose', 'tensor', 'tensor_sequence', 'autoregressive')), ('embed', ('fsdp', 'fsdp_transpose', 'sequence', 'tensor_transpose', 'context', 'expert')), ('embed', ('fsdp', 'sequence', 'tensor_transpose', 'context', 'expert')), ('embed', ('fsdp', 'fsdp_transpose', 'sequence', 'context', 'expert')), ('embed', ('fsdp', 'sequence', 'context', 'expert')), ('norm', ('tensor', 'tensor_transpose')), ('layers', 'stage'), ('diloco', 'diloco'), ('engram_dim', ('tensor',)), ('dense_layers', ()), ('moe_layers', ()), ('mhc', ()), ('prefill_activation_length', ('sequence', 'context')), ('prefill_activation_norm_length', ('tensor_sequence', 'context', 'sequence')), ('activation_prefill_kv_batch', ('data', 'fsdp', 'fsdp_transpose', 'expert')), ('decode_batch', ('data', 'fsdp', 'fsdp_transpose', 'expert')), ('decode_length', ('sequence',)), ('cache_heads', ('autoregressive', 'tensor', 'tensor_transpose', 'tensor_sequence')), ('cache_heads', ('autoregressive', 'tensor', 'tensor_sequence')), ('paged_kv_heads', ('tensor',)), ('cache_batch_prefill', ()), ('cache_batch', ()), ('cache_heads_none', ()), ('cache_kv', ()), ('cache_sequence', ()), ('num_pages', ()), ('tokens_per_page', ()), ('paged_kv_head_dim_size', ()), ('mlp_no_fsdp', ('tensor', 'tensor_sequence', 'autoregressive')), ('embed_tensor_transpose', ('tensor_transpose',)), ('exp_with_fsdp', 'fsdp'))
I0421 08:50:59.456925 138379323594560 pyconfig.py:432] Config param logits_dot_in_fp32: False
I0421 08:50:59.456942 138379323594560 pyconfig.py:432] Config param logits_via_embedding: True
I0421 08:50:59.456959 138379323594560 pyconfig.py:432] Config param lora_input_adapters_path: 
I0421 08:50:59.456975 138379323594560 pyconfig.py:432] Config param loss_algo: grpo
I0421 08:50:59.456990 138379323594560 pyconfig.py:432] Config param lr_schedule_type: LearningRateScheduleType.COSINE
I0421 08:50:59.457008 138379323594560 pyconfig.py:432] Config param managed_mldiagnostics: False
I0421 08:50:59.457022 138379323594560 pyconfig.py:432] Config param managed_mldiagnostics_dir: None
I0421 08:50:59.457039 138379323594560 pyconfig.py:432] Config param managed_mldiagnostics_run_group: 
I0421 08:50:59.457053 138379323594560 pyconfig.py:432] Config param matmul_precision: MatmulPrecision.DEFAULT
I0421 08:50:59.457072 138379323594560 pyconfig.py:432] Config param max_checkify: False
I0421 08:50:59.457090 138379323594560 pyconfig.py:432] Config param max_concurrency: 256
I0421 08:50:59.457106 138379323594560 pyconfig.py:432] Config param max_corpus_chars: 10000000
I0421 08:50:59.457120 138379323594560 pyconfig.py:432] Config param max_num_batched_tokens: None
I0421 08:50:59.457136 138379323594560 pyconfig.py:432] Config param max_num_checkpoints_to_keep: None
I0421 08:50:59.457151 138379323594560 pyconfig.py:432] Config param max_num_images_per_example: -1
I0421 08:50:59.457167 138379323594560 pyconfig.py:432] Config param max_num_seqs: None
I0421 08:50:59.457181 138379323594560 pyconfig.py:432] Config param max_position_embeddings: 163840
I0421 08:50:59.457197 138379323594560 pyconfig.py:432] Config param max_prefill_predict_length: 64
I0421 08:50:59.457211 138379323594560 pyconfig.py:432] Config param max_sample_len_for_audio: 10000
I0421 08:50:59.457227 138379323594560 pyconfig.py:432] Config param max_segments_per_seq: -1
I0421 08:50:59.457241 138379323594560 pyconfig.py:432] Config param max_source_positions_for_audio: 1500
I0421 08:50:59.457257 138379323594560 pyconfig.py:432] Config param max_target_length: 2048
I0421 08:50:59.457272 138379323594560 pyconfig.py:432] Config param max_timescale_for_audio: 10000.0
I0421 08:50:59.457288 138379323594560 pyconfig.py:432] Config param megablox: True
I0421 08:50:59.457302 138379323594560 pyconfig.py:432] Config param merge_gating_gmm: False
I0421 08:50:59.457317 138379323594560 pyconfig.py:432] Config param mesh_axes: ['diloco', 'data', 'stage', 'fsdp', 'fsdp_transpose', 'sequence', 'context', 'context_autoregressive', 'tensor', 'tensor_transpose', 'tensor_sequence', 'expert', 'autoregressive']
I0421 08:50:59.457335 138379323594560 pyconfig.py:432] Config param metrics_dir: None
I0421 08:50:59.457350 138379323594560 pyconfig.py:432] Config param metrics_file: 
I0421 08:50:59.457365 138379323594560 pyconfig.py:432] Config param mhc_expansion_rate: 1
I0421 08:50:59.457380 138379323594560 pyconfig.py:432] Config param micro_batch_size_to_eval_on: 64
I0421 08:50:59.457396 138379323594560 pyconfig.py:432] Config param micro_batch_size_to_train_on: 64
I0421 08:50:59.457410 138379323594560 pyconfig.py:432] Config param mla_kv: RematLocation.REMAT
I0421 08:50:59.457426 138379323594560 pyconfig.py:432] Config param mla_naive_kvcache: True
I0421 08:50:59.457441 138379323594560 pyconfig.py:432] Config param mla_q: RematLocation.REMAT
I0421 08:50:59.457457 138379323594560 pyconfig.py:432] Config param mlp_activations: ['gelu']
I0421 08:50:59.457471 138379323594560 pyconfig.py:432] Config param mlp_activations_limit: -1.0
I0421 08:50:59.457488 138379323594560 pyconfig.py:432] Config param mlp_bias: False
I0421 08:50:59.457502 138379323594560 pyconfig.py:432] Config param mlp_dim: 64
I0421 08:50:59.457518 138379323594560 pyconfig.py:432] Config param mlpwi: RematLocation.REMAT
I0421 08:50:59.457532 138379323594560 pyconfig.py:432] Config param mlpwi_0: RematLocation.REMAT
I0421 08:50:59.457548 138379323594560 pyconfig.py:432] Config param mlpwi_1: RematLocation.REMAT
I0421 08:50:59.457563 138379323594560 pyconfig.py:432] Config param mlpwo: RematLocation.REMAT
I0421 08:50:59.457578 138379323594560 pyconfig.py:432] Config param moba: False
I0421 08:50:59.457593 138379323594560 pyconfig.py:432] Config param moba_chunk_size: 1024
I0421 08:50:59.457608 138379323594560 pyconfig.py:432] Config param moba_topk: 8
I0421 08:50:59.457624 138379323594560 pyconfig.py:432] Config param model_call_mode: 
I0421 08:50:59.457638 138379323594560 pyconfig.py:432] Config param model_name: gpt3-52k
I0421 08:50:59.457654 138379323594560 pyconfig.py:432] Config param moe_expert_input_dim: -1
I0421 08:50:59.457668 138379323594560 pyconfig.py:432] Config param moe_fsdp_use_two_stage_all_gather: False
I0421 08:50:59.457684 138379323594560 pyconfig.py:432] Config param moe_mlp_dim: -1
I0421 08:50:59.457698 138379323594560 pyconfig.py:432] Config param moe_mlpwi_0: RematLocation.REMAT
I0421 08:50:59.457714 138379323594560 pyconfig.py:432] Config param moe_mlpwi_1: RematLocation.REMAT
I0421 08:50:59.457739 138379323594560 pyconfig.py:432] Config param moe_mlpwo: RematLocation.REMAT
I0421 08:50:59.457755 138379323594560 pyconfig.py:432] Config param monitor_goodput: False
I0421 08:50:59.457770 138379323594560 pyconfig.py:432] Config param monitor_step_time_deviation: True
I0421 08:50:59.457784 138379323594560 pyconfig.py:432] Config param mrope_section: [24, 20, 20]
I0421 08:50:59.457800 138379323594560 pyconfig.py:432] Config param mscale: 1.0
I0421 08:50:59.457814 138379323594560 pyconfig.py:432] Config param mtc_data_parallelism: 0
I0421 08:50:59.457830 138379323594560 pyconfig.py:432] Config param mtp_eval_target_module: 0
I0421 08:50:59.457844 138379323594560 pyconfig.py:432] Config param mtp_loss_scaling_factor: 0.1
I0421 08:50:59.457861 138379323594560 pyconfig.py:432] Config param mtp_num_layers: 0
I0421 08:50:59.457875 138379323594560 pyconfig.py:432] Config param mu_dtype: float32
I0421 08:50:59.457899 138379323594560 pyconfig.py:432] Config param multi_sampling: False
I0421 08:50:59.457914 138379323594560 pyconfig.py:432] Config param multi_tier_checkpointing_backup_interval_minutes: 0
I0421 08:50:59.457930 138379323594560 pyconfig.py:432] Config param muon_beta: 0.95
I0421 08:50:59.457945 138379323594560 pyconfig.py:432] Config param muon_consistent_rms: None
I0421 08:50:59.457961 138379323594560 pyconfig.py:432] Config param muon_weight_decay: 0.0
I0421 08:50:59.457976 138379323594560 pyconfig.py:432] Config param n_routing_groups: -1
I0421 08:50:59.457991 138379323594560 pyconfig.py:432] Config param n_window_for_audio: 50
I0421 08:50:59.458006 138379323594560 pyconfig.py:432] Config param n_window_infer_for_audio: 800
I0421 08:50:59.458021 138379323594560 pyconfig.py:432] Config param nope_layer_interval: -1
I0421 08:50:59.458036 138379323594560 pyconfig.py:432] Config param norm_topk_prob: False
I0421 08:50:59.458051 138379323594560 pyconfig.py:432] Config param normalization_layer_epsilon: 1e-05
I0421 08:50:59.458067 138379323594560 pyconfig.py:432] Config param normalize_embedding_logits: False
I0421 08:50:59.458086 138379323594560 pyconfig.py:432] Config param num_attention_heads_for_vit: 16
I0421 08:50:59.458101 138379323594560 pyconfig.py:432] Config param num_batches: 4
I0421 08:50:59.458116 138379323594560 pyconfig.py:432] Config param num_channels_for_vit: 3
I0421 08:50:59.458132 138379323594560 pyconfig.py:432] Config param num_conv_layers_for_audio: 3
I0421 08:50:59.458148 138379323594560 pyconfig.py:432] Config param num_decoder_layers: 1
I0421 08:50:59.458162 138379323594560 pyconfig.py:432] Config param num_diloco_replicas: 1
I0421 08:50:59.458178 138379323594560 pyconfig.py:432] Config param num_epoch: 1
I0421 08:50:59.458192 138379323594560 pyconfig.py:432] Config param num_eval_passes: 1
I0421 08:50:59.458208 138379323594560 pyconfig.py:432] Config param num_experts: 1
I0421 08:50:59.458222 138379323594560 pyconfig.py:432] Config param num_experts_per_tok: 1
I0421 08:50:59.458237 138379323594560 pyconfig.py:432] Config param num_generations: 2
I0421 08:50:59.458251 138379323594560 pyconfig.py:432] Config param num_hidden_layers_for_vit: 34
I0421 08:50:59.458267 138379323594560 pyconfig.py:432] Config param num_iterations: 1
I0421 08:50:59.458281 138379323594560 pyconfig.py:432] Config param num_kv_heads: 2
I0421 08:50:59.458296 138379323594560 pyconfig.py:432] Config param num_layers_per_pipeline_stage: 1
I0421 08:50:59.458310 138379323594560 pyconfig.py:432] Config param num_mel_bins_for_audio: 128
I0421 08:50:59.458326 138379323594560 pyconfig.py:432] Config param num_pipeline_microbatches: -1
I0421 08:50:59.458340 138379323594560 pyconfig.py:432] Config param num_pipeline_repeats: -1
I0421 08:50:59.458355 138379323594560 pyconfig.py:432] Config param num_position_embeddings_for_vit: 1024
I0421 08:50:59.458369 138379323594560 pyconfig.py:432] Config param num_query_heads: 2
I0421 08:50:59.458384 138379323594560 pyconfig.py:432] Config param num_samplers_slices: -1
I0421 08:50:59.458398 138379323594560 pyconfig.py:432] Config param num_slices: 1
I0421 08:50:59.458414 138379323594560 pyconfig.py:432] Config param num_target_devices: 32
I0421 08:50:59.458427 138379323594560 pyconfig.py:432] Config param num_test_batches: 5
I0421 08:50:59.458443 138379323594560 pyconfig.py:432] Config param num_trainer_slices: -1
I0421 08:50:59.458457 138379323594560 pyconfig.py:432] Config param num_vocab_tiling: 1
I0421 08:50:59.458473 138379323594560 pyconfig.py:432] Config param off_policy_steps: 0
I0421 08:50:59.458487 138379323594560 pyconfig.py:432] Config param offline_data_dir: None
I0421 08:50:59.458503 138379323594560 pyconfig.py:432] Config param opt_type: OptimizerType.ADAM_PAX
I0421 08:50:59.458518 138379323594560 pyconfig.py:432] Config param optimize_mesh_for_tpu_v6e: False
I0421 08:50:59.458534 138379323594560 pyconfig.py:432] Config param optimizer_memory_host_offload: False
I0421 08:50:59.458549 138379323594560 pyconfig.py:432] Config param original_max_position_embeddings: 4096
I0421 08:50:59.458565 138379323594560 pyconfig.py:432] Config param out_hidden_size_for_vit: 512
I0421 08:50:59.458579 138379323594560 pyconfig.py:432] Config param out_proj: RematLocation.REMAT
I0421 08:50:59.458595 138379323594560 pyconfig.py:432] Config param output_dim_for_audio: 512
I0421 08:50:59.458609 138379323594560 pyconfig.py:432] Config param override_logical_axis_rules: False
I0421 08:50:59.458625 138379323594560 pyconfig.py:432] Config param override_model_config: True
I0421 08:50:59.458639 138379323594560 pyconfig.py:432] Config param packing: True
I0421 08:50:59.458655 138379323594560 pyconfig.py:432] Config param pagedattn_head_dim_alignment: 128
I0421 08:50:59.458669 138379323594560 pyconfig.py:432] Config param pagedattn_max_pages_per_group: -1
I0421 08:50:59.458685 138379323594560 pyconfig.py:432] Config param pagedattn_num_pages: 64
I0421 08:50:59.458699 138379323594560 pyconfig.py:432] Config param pagedattn_pages_per_compute_block: 4
I0421 08:50:59.458715 138379323594560 pyconfig.py:432] Config param pagedattn_tokens_per_page: 32
I0421 08:50:59.458729 138379323594560 pyconfig.py:432] Config param param_scan_axis: 1
I0421 08:50:59.458754 138379323594560 pyconfig.py:432] Config param parameter_memory_host_offload: False
I0421 08:50:59.458768 138379323594560 pyconfig.py:432] Config param partial_rotary_factor: 1.0
I0421 08:50:59.458784 138379323594560 pyconfig.py:432] Config param patch_size_for_vit: 14
I0421 08:50:59.458798 138379323594560 pyconfig.py:432] Config param penalty_incorrect_answer: -1.0
I0421 08:50:59.458813 138379323594560 pyconfig.py:432] Config param penalty_incorrect_format: -0.5
I0421 08:50:59.458829 138379323594560 pyconfig.py:432] Config param per_device_batch_size: 2
I0421 08:50:59.458843 138379323594560 pyconfig.py:432] Config param per_device_batch_size_increment: 2.0
I0421 08:50:59.458859 138379323594560 pyconfig.py:432] Config param per_device_batch_size_start: 4.0
I0421 08:50:59.458874 138379323594560 pyconfig.py:432] Config param pipeline_delay_activation_forwarding: False
I0421 08:50:59.458889 138379323594560 pyconfig.py:432] Config param pipeline_fsdp_ag_once: False
I0421 08:50:59.458903 138379323594560 pyconfig.py:432] Config param pipeline_fsdp_ag_per_repeat: False
I0421 08:50:59.458919 138379323594560 pyconfig.py:432] Config param pipeline_parallel_layers: 1
I0421 08:50:59.458933 138379323594560 pyconfig.py:432] Config param pixel_shuffle_ratio_for_vit: 0.5
I0421 08:50:59.458949 138379323594560 pyconfig.py:432] Config param posemb_type_for_vit: learn
I0421 08:50:59.458963 138379323594560 pyconfig.py:432] Config param position_id_per_seconds: 25
I0421 08:50:59.458979 138379323594560 pyconfig.py:432] Config param prefill_cache_axis_order: 1,2,0,3
I0421 08:50:59.458993 138379323594560 pyconfig.py:432] Config param prefill_cache_dir: 
I0421 08:50:59.459008 138379323594560 pyconfig.py:432] Config param prefill_chunk_size: 256
I0421 08:50:59.459023 138379323594560 pyconfig.py:432] Config param prefill_slice: v5e-16
I0421 08:50:59.459039 138379323594560 pyconfig.py:432] Config param prefix_caching_dram_byte: 100000000000
I0421 08:50:59.459054 138379323594560 pyconfig.py:432] Config param prefix_caching_hbm_byte: 10000000000
I0421 08:50:59.459069 138379323594560 pyconfig.py:432] Config param profile_cleanly: True
I0421 08:50:59.459088 138379323594560 pyconfig.py:432] Config param profile_periodically_period: -1
I0421 08:50:59.459104 138379323594560 pyconfig.py:432] Config param profile_power_events: False
I0421 08:50:59.459118 138379323594560 pyconfig.py:432] Config param profiler: ProfilerType.NONE
I0421 08:50:59.459136 138379323594560 pyconfig.py:432] Config param profiler_steps: 5
I0421 08:50:59.459150 138379323594560 pyconfig.py:432] Config param projector_dropout_for_vit: 0.0
I0421 08:50:59.459166 138379323594560 pyconfig.py:432] Config param projector_input_dim_for_vit: 4096
I0421 08:50:59.459180 138379323594560 pyconfig.py:432] Config param projector_output_dim_for_vit: 4096
I0421 08:50:59.459195 138379323594560 pyconfig.py:432] Config param prometheus_port: 0
I0421 08:50:59.459211 138379323594560 pyconfig.py:432] Config param prompt: I love to
I0421 08:50:59.459225 138379323594560 pyconfig.py:432] Config param pure_nnx: False
I0421 08:50:59.459241 138379323594560 pyconfig.py:432] Config param pure_nnx_decoder: False
I0421 08:50:59.459256 138379323594560 pyconfig.py:432] Config param q_lora_rank: 0
I0421 08:50:59.459270 138379323594560 pyconfig.py:432] Config param qk_clip_threshold: 100.0
I0421 08:50:59.459286 138379323594560 pyconfig.py:432] Config param qk_nope_head_dim: 128
I0421 08:50:59.459301 138379323594560 pyconfig.py:432] Config param qk_norm_with_scale: True
I0421 08:50:59.459316 138379323594560 pyconfig.py:432] Config param qk_rope_head_dim: 64
I0421 08:50:59.459332 138379323594560 pyconfig.py:432] Config param qkv_proj: RematLocation.REMAT
I0421 08:50:59.459347 138379323594560 pyconfig.py:432] Config param quant_cfg_path: 
I0421 08:50:59.459362 138379323594560 pyconfig.py:432] Config param quantization: QuantizationType.NONE
I0421 08:50:59.459378 138379323594560 pyconfig.py:432] Config param quantization_local_shard_count: 4
I0421 08:50:59.459394 138379323594560 pyconfig.py:432] Config param quantize_kvcache: False
I0421 08:50:59.459409 138379323594560 pyconfig.py:432] Config param query_proj: RematLocation.REMAT
I0421 08:50:59.459424 138379323594560 pyconfig.py:432] Config param query_wa_proj: RematLocation.REMAT
I0421 08:50:59.459440 138379323594560 pyconfig.py:432] Config param ragged_block_size: 256
I0421 08:50:59.459456 138379323594560 pyconfig.py:432] Config param ragged_buffer_factor: -1.0
I0421 08:50:59.459470 138379323594560 pyconfig.py:432] Config param rampup_end_step: 0
I0421 08:50:59.459486 138379323594560 pyconfig.py:432] Config param rampup_samples_per_increment_to_load: None
I0421 08:50:59.459502 138379323594560 pyconfig.py:432] Config param reasoning_end_token: </reasoning>
I0421 08:50:59.459516 138379323594560 pyconfig.py:432] Config param reasoning_start_token: <reasoning>
I0421 08:50:59.459533 138379323594560 pyconfig.py:432] Config param record_internal_nn_metrics: 0
I0421 08:50:59.459548 138379323594560 pyconfig.py:432] Config param remat_policy: full
I0421 08:50:59.459563 138379323594560 pyconfig.py:432] Config param remat_policy_for_vit: minimal
I0421 08:50:59.459578 138379323594560 pyconfig.py:432] Config param remove_size_one_mesh_axis_from_type: True
I0421 08:50:59.459594 138379323594560 pyconfig.py:432] Config param replicate_quant_scale: False
I0421 08:50:59.459610 138379323594560 pyconfig.py:432] Config param replicator_backup_interval_minutes: 0
I0421 08:50:59.459625 138379323594560 pyconfig.py:432] Config param report_heartbeat_metric_for_gcp_monitoring: False
I0421 08:50:59.459640 138379323594560 pyconfig.py:432] Config param report_performance_metric_for_gcp_monitoring: False
I0421 08:50:59.459656 138379323594560 pyconfig.py:432] Config param reshape_q: False
I0421 08:50:59.459670 138379323594560 pyconfig.py:432] Config param return_log_prob: False
I0421 08:50:59.459686 138379323594560 pyconfig.py:432] Config param reuse_example_batch: 0
I0421 08:50:59.459700 138379323594560 pyconfig.py:432] Config param reward_exact_answer: 5.0
I0421 08:50:59.459716 138379323594560 pyconfig.py:432] Config param reward_exact_format_match: 3.0
I0421 08:50:59.459740 138379323594560 pyconfig.py:432] Config param reward_partial_format_match: 0.5
I0421 08:50:59.459757 138379323594560 pyconfig.py:432] Config param reward_ratio_guess_to_answer_high: 0.5
I0421 08:50:59.459772 138379323594560 pyconfig.py:432] Config param reward_ratio_guess_to_answer_low: 0.25
I0421 08:50:59.459788 138379323594560 pyconfig.py:432] Config param reward_white_space_format_match: 1.5
I0421 08:50:59.459803 138379323594560 pyconfig.py:432] Config param rl: {'num_generations': 2, 'num_iterations': 1, 'grpo_beta': 0.08, 'grpo_epsilon': 0.2, 'loss_algo': 'grpo', 'use_agentic_rollout': False, 'max_concurrency': 256, 'off_policy_steps': 0, 'system_prompt': '', 'degenerate_group_masking': True, 'epsilon_high': None}
I0421 08:50:59.459824 138379323594560 pyconfig.py:432] Config param rollout_data_parallelism: -1
I0421 08:50:59.459839 138379323594560 pyconfig.py:432] Config param rollout_expert_parallelism: 1
I0421 08:50:59.459855 138379323594560 pyconfig.py:432] Config param rollout_micro_batch_size: -1
I0421 08:50:59.459870 138379323594560 pyconfig.py:432] Config param rollout_tensor_parallelism: -1
I0421 08:50:59.459886 138379323594560 pyconfig.py:432] Config param rope_attention_scaling: False
I0421 08:50:59.459900 138379323594560 pyconfig.py:432] Config param rope_factor: 40
I0421 08:50:59.459915 138379323594560 pyconfig.py:432] Config param rope_interleave: True
I0421 08:50:59.459929 138379323594560 pyconfig.py:432] Config param rope_linear_scaling_factor: 1.0
I0421 08:50:59.459945 138379323594560 pyconfig.py:432] Config param rope_max_timescale: 10000
I0421 08:50:59.459960 138379323594560 pyconfig.py:432] Config param rope_min_timescale: 1
I0421 08:50:59.459975 138379323594560 pyconfig.py:432] Config param rope_theta_for_vit: 10000
I0421 08:50:59.459990 138379323594560 pyconfig.py:432] Config param rope_truncate: True
I0421 08:50:59.460005 138379323594560 pyconfig.py:432] Config param rope_type: RopeType.DEFAULT
I0421 08:50:59.460024 138379323594560 pyconfig.py:432] Config param rope_use_scale: True
I0421 08:50:59.460039 138379323594560 pyconfig.py:432] Config param routed_bias: False
I0421 08:50:59.460054 138379323594560 pyconfig.py:432] Config param routed_bias_update_rate: 0.0
I0421 08:50:59.460069 138379323594560 pyconfig.py:432] Config param routed_scaling_factor: 1.0
I0421 08:50:59.460087 138379323594560 pyconfig.py:432] Config param routed_score_func: 
I0421 08:50:59.460103 138379323594560 pyconfig.py:432] Config param run_name: gpt3-52k_2026-04-21-08-50
I0421 08:50:59.460117 138379323594560 pyconfig.py:432] Config param sa_block_kv: 512
I0421 08:50:59.460132 138379323594560 pyconfig.py:432] Config param sa_block_kv_compute: 512
I0421 08:50:59.460147 138379323594560 pyconfig.py:432] Config param sa_block_kv_dkv: 512
I0421 08:50:59.460162 138379323594560 pyconfig.py:432] Config param sa_block_kv_dkv_compute: 512
I0421 08:50:59.460177 138379323594560 pyconfig.py:432] Config param sa_block_kv_dq: 512
I0421 08:50:59.460192 138379323594560 pyconfig.py:432] Config param sa_block_q: 512
I0421 08:50:59.460208 138379323594560 pyconfig.py:432] Config param sa_block_q_dkv: 512
I0421 08:50:59.460222 138379323594560 pyconfig.py:432] Config param sa_block_q_dq: 512
I0421 08:50:59.460238 138379323594560 pyconfig.py:432] Config param sa_k_layout: HEAD_DIM_MINOR
I0421 08:50:59.460253 138379323594560 pyconfig.py:432] Config param sa_q_layout: HEAD_DIM_MINOR
I0421 08:50:59.460267 138379323594560 pyconfig.py:432] Config param sa_use_fused_bwd_kernel: False
I0421 08:50:59.460282 138379323594560 pyconfig.py:432] Config param sa_v_layout: HEAD_DIM_MINOR
I0421 08:50:59.460297 138379323594560 pyconfig.py:432] Config param sampler_devices_fraction: 0.5
I0421 08:50:59.460313 138379323594560 pyconfig.py:432] Config param save_checkpoint_on_completion: True
I0421 08:50:59.460328 138379323594560 pyconfig.py:432] Config param save_config_to_gcs: False
I0421 08:50:59.460343 138379323594560 pyconfig.py:432] Config param save_quantized_params_path: 
I0421 08:50:59.460358 138379323594560 pyconfig.py:432] Config param scale_embedding_for_audio: True
I0421 08:50:59.460373 138379323594560 pyconfig.py:432] Config param scan_layers: True
I0421 08:50:59.460388 138379323594560 pyconfig.py:432] Config param scan_layers_per_stage: False
I0421 08:50:59.460403 138379323594560 pyconfig.py:432] Config param scan_pipeline_iterations: True
I0421 08:50:59.460418 138379323594560 pyconfig.py:432] Config param scan_pipeline_repeats: False
I0421 08:50:59.460433 138379323594560 pyconfig.py:432] Config param set_remat_policy_on_layers_per_stage: False
I0421 08:50:59.460448 138379323594560 pyconfig.py:432] Config param set_remat_policy_on_pipeline_iterations: True
I0421 08:50:59.460464 138379323594560 pyconfig.py:432] Config param sft_train_on_completion_only: False
I0421 08:50:59.460480 138379323594560 pyconfig.py:432] Config param shard_exp_on_fsdp: False
I0421 08:50:59.460494 138379323594560 pyconfig.py:432] Config param shard_mode: ShardMode.AUTO
I0421 08:50:59.460511 138379323594560 pyconfig.py:432] Config param shard_optimizer_over_data: False
I0421 08:50:59.460527 138379323594560 pyconfig.py:432] Config param sharding_strategy: None
I0421 08:50:59.460541 138379323594560 pyconfig.py:432] Config param sharding_tolerance: 0.02
I0421 08:50:59.460557 138379323594560 pyconfig.py:432] Config param shardy: True
I0421 08:50:59.460572 138379323594560 pyconfig.py:432] Config param share_kv_projections: False
I0421 08:50:59.460587 138379323594560 pyconfig.py:432] Config param shared_experts: 0
I0421 08:50:59.460603 138379323594560 pyconfig.py:432] Config param sinkhorn_iterations: 20
I0421 08:50:59.460617 138379323594560 pyconfig.py:432] Config param skip_first_n_steps_for_profiler: 1
I0421 08:50:59.460633 138379323594560 pyconfig.py:432] Config param skip_jax_distributed_system: False
I0421 08:50:59.460647 138379323594560 pyconfig.py:432] Config param skip_step_interval: 128
I0421 08:50:59.460663 138379323594560 pyconfig.py:432] Config param skip_step_on_spikes: False
I0421 08:50:59.460678 138379323594560 pyconfig.py:432] Config param skip_step_scaling_factor: 6.0
I0421 08:50:59.460693 138379323594560 pyconfig.py:432] Config param sliding_window_size: 0
I0421 08:50:59.460709 138379323594560 pyconfig.py:432] Config param solution_end_token: </answer>
I0421 08:50:59.460725 138379323594560 pyconfig.py:432] Config param solution_start_token: <answer>
I0421 08:50:59.460747 138379323594560 pyconfig.py:432] Config param source_checkpoint_layout: orbax
I0421 08:50:59.460763 138379323594560 pyconfig.py:432] Config param sparse_matmul: True
I0421 08:50:59.460779 138379323594560 pyconfig.py:432] Config param spatial_merge_size_for_vit: 2
I0421 08:50:59.460793 138379323594560 pyconfig.py:432] Config param stack_prefill_result_cache: False
I0421 08:50:59.460809 138379323594560 pyconfig.py:432] Config param stack_trace_interval_seconds: 600
I0421 08:50:59.460823 138379323594560 pyconfig.py:432] Config param stack_trace_to_cloud: False
I0421 08:50:59.460839 138379323594560 pyconfig.py:432] Config param step_deviation_interval_seconds: 30
I0421 08:50:59.460853 138379323594560 pyconfig.py:432] Config param steps: 200000
I0421 08:50:59.460869 138379323594560 pyconfig.py:432] Config param stop_strings: None
I0421 08:50:59.460883 138379323594560 pyconfig.py:432] Config param student_overrides: {'model_name': 'llama3.1-8b'}
I0421 08:50:59.460900 138379323594560 pyconfig.py:432] Config param student_params_to_update: None
I0421 08:50:59.460914 138379323594560 pyconfig.py:432] Config param subslice_shape: 
I0421 08:50:59.460930 138379323594560 pyconfig.py:432] Config param swap_space_vllm_gb: 2
I0421 08:50:59.460944 138379323594560 pyconfig.py:432] Config param system_prompt: 
I0421 08:50:59.460960 138379323594560 pyconfig.py:432] Config param target_eval_loss: 0.0
I0421 08:50:59.460974 138379323594560 pyconfig.py:432] Config param teacher_overrides: {'model_name': 'llama3.1-8b'}
I0421 08:50:59.460990 138379323594560 pyconfig.py:432] Config param temperature_tuning: False
I0421 08:50:59.461004 138379323594560 pyconfig.py:432] Config param temporal_patch_size_for_vit: 2
I0421 08:50:59.461020 138379323594560 pyconfig.py:432] Config param tensorboard_dir: None
I0421 08:50:59.461034 138379323594560 pyconfig.py:432] Config param tensors_on_device: None
I0421 08:50:59.461050 138379323594560 pyconfig.py:432] Config param tensors_to_offload: None
I0421 08:50:59.461064 138379323594560 pyconfig.py:432] Config param test_batch_start_index: 0
I0421 08:50:59.461083 138379323594560 pyconfig.py:432] Config param tile_size_for_vit: 336
I0421 08:50:59.461098 138379323594560 pyconfig.py:432] Config param tokenize_eval_data: True
I0421 08:50:59.461114 138379323594560 pyconfig.py:432] Config param tokenize_train_data: True
I0421 08:50:59.461128 138379323594560 pyconfig.py:432] Config param tokenizer_path: meta-llama/Llama-3.1-8B
I0421 08:50:59.461144 138379323594560 pyconfig.py:432] Config param tokenizer_type: TokenizerType.HUGGINGFACE
I0421 08:50:59.461160 138379323594560 pyconfig.py:432] Config param topk_routing_group: -1
I0421 08:50:59.461175 138379323594560 pyconfig.py:432] Config param train_data_columns: ['text']
I0421 08:50:59.461191 138379323594560 pyconfig.py:432] Config param train_fraction: 1.0
I0421 08:50:59.461205 138379323594560 pyconfig.py:432] Config param train_image_column: image
I0421 08:50:59.461221 138379323594560 pyconfig.py:432] Config param train_micro_batch_size: -1
I0421 08:50:59.461236 138379323594560 pyconfig.py:432] Config param train_split: train
I0421 08:50:59.461251 138379323594560 pyconfig.py:432] Config param trainable_parameters_mask: []
I0421 08:50:59.461265 138379323594560 pyconfig.py:432] Config param trainable_position_size: 2048
I0421 08:50:59.461281 138379323594560 pyconfig.py:432] Config param trainer_devices_fraction: 0.5
I0421 08:50:59.461295 138379323594560 pyconfig.py:432] Config param upload_all_profiler_results: False
I0421 08:50:59.461310 138379323594560 pyconfig.py:432] Config param use_2d_fsdp_sharding: False
I0421 08:50:59.461325 138379323594560 pyconfig.py:432] Config param use_agentic_rollout: False
I0421 08:50:59.461339 138379323594560 pyconfig.py:432] Config param use_audio: False
I0421 08:50:59.461354 138379323594560 pyconfig.py:432] Config param use_audio_in_video: False
I0421 08:50:59.461368 138379323594560 pyconfig.py:432] Config param use_batch_split_schedule: False
I0421 08:50:59.461384 138379323594560 pyconfig.py:432] Config param use_chat_template: False
I0421 08:50:59.461400 138379323594560 pyconfig.py:432] Config param use_chunked_prefill: False
I0421 08:50:59.461413 138379323594560 pyconfig.py:432] Config param use_custom_sort_vjp: True
I0421 08:50:59.461429 138379323594560 pyconfig.py:432] Config param use_dpo: False
I0421 08:50:59.461443 138379323594560 pyconfig.py:432] Config param use_gather_mosaic_kernel: False
I0421 08:50:59.461459 138379323594560 pyconfig.py:432] Config param use_grpo: True
I0421 08:50:59.461473 138379323594560 pyconfig.py:432] Config param use_indexer: False
I0421 08:50:59.461489 138379323594560 pyconfig.py:432] Config param use_iota_embed: True
I0421 08:50:59.461503 138379323594560 pyconfig.py:432] Config param use_jax_splash: False
I0421 08:50:59.461518 138379323594560 pyconfig.py:432] Config param use_max_logit_estimate: -1
I0421 08:50:59.461532 138379323594560 pyconfig.py:432] Config param use_mrope: False
I0421 08:50:59.461548 138379323594560 pyconfig.py:432] Config param use_multimodal: False
I0421 08:50:59.461562 138379323594560 pyconfig.py:432] Config param use_pathways: True
I0421 08:50:59.461578 138379323594560 pyconfig.py:432] Config param use_post_attn_norm: False
I0421 08:50:59.461592 138379323594560 pyconfig.py:432] Config param use_post_ffw_norm: False
I0421 08:50:59.461607 138379323594560 pyconfig.py:432] Config param use_qk_clip: False
I0421 08:50:59.461621 138379323594560 pyconfig.py:432] Config param use_qk_norm: False
I0421 08:50:59.461637 138379323594560 pyconfig.py:432] Config param use_qk_norm_in_gdn: True
I0421 08:50:59.461652 138379323594560 pyconfig.py:432] Config param use_qwix_quantization: False
I0421 08:50:59.461667 138379323594560 pyconfig.py:432] Config param use_ragged_attention: False
I0421 08:50:59.461681 138379323594560 pyconfig.py:432] Config param use_random_routing: False
I0421 08:50:59.461697 138379323594560 pyconfig.py:432] Config param use_replicator_service: False
I0421 08:50:59.461711 138379323594560 pyconfig.py:432] Config param use_ring_of_experts: False
I0421 08:50:59.461726 138379323594560 pyconfig.py:432] Config param use_sft: False
I0421 08:50:59.461748 138379323594560 pyconfig.py:432] Config param use_splash_scheduler: False
I0421 08:50:59.461764 138379323594560 pyconfig.py:432] Config param use_tokamax_gmm: False
I0421 08:50:59.461778 138379323594560 pyconfig.py:432] Config param use_tokamax_splash: False
I0421 08:50:59.461793 138379323594560 pyconfig.py:432] Config param use_truncation: True
I0421 08:50:59.461808 138379323594560 pyconfig.py:432] Config param use_tunix_gradient_accumulation: False
I0421 08:50:59.461824 138379323594560 pyconfig.py:432] Config param use_untrainable_positional_embedding: False
I0421 08:50:59.461838 138379323594560 pyconfig.py:432] Config param use_vertex_tensorboard: False
I0421 08:50:59.461854 138379323594560 pyconfig.py:432] Config param using_pipeline_parallelism: False
I0421 08:50:59.461870 138379323594560 pyconfig.py:432] Config param v_head_dim: 128
I0421 08:50:59.461886 138379323594560 pyconfig.py:432] Config param v_norm_with_scale: True
I0421 08:50:59.461900 138379323594560 pyconfig.py:432] Config param value_proj: RematLocation.REMAT
I0421 08:50:59.461918 138379323594560 pyconfig.py:432] Config param vertex_tensorboard_project: 
I0421 08:50:59.461932 138379323594560 pyconfig.py:432] Config param vertex_tensorboard_region: 
I0421 08:50:59.461947 138379323594560 pyconfig.py:432] Config param video_path: 
I0421 08:50:59.461962 138379323594560 pyconfig.py:432] Config param video_placeholder: <|video|>
I0421 08:50:59.461979 138379323594560 pyconfig.py:432] Config param vision_output_dim_for_vit: 4096
I0421 08:50:59.461994 138379323594560 pyconfig.py:432] Config param vision_output_length: -1
I0421 08:50:59.462009 138379323594560 pyconfig.py:432] Config param vllm_additional_config: {}
I0421 08:50:59.462024 138379323594560 pyconfig.py:432] Config param vllm_hf_config_path: 
I0421 08:50:59.462039 138379323594560 pyconfig.py:432] Config param vllm_hf_overrides: {}
I0421 08:50:59.462054 138379323594560 pyconfig.py:432] Config param vocab_size: 32000
I0421 08:50:59.462069 138379323594560 pyconfig.py:432] Config param warmup_steps_fraction: 0.1
I0421 08:50:59.462088 138379323594560 pyconfig.py:432] Config param weight_dtype: float32
I0421 08:50:59.462112 138379323594560 pyconfig.py:432] Config param weight_quantization_calibration_method: absmax
I0421 08:50:59.462127 138379323594560 pyconfig.py:432] Config param wi_tile_dlhs_batch_seq: 512
I0421 08:50:59.462142 138379323594560 pyconfig.py:432] Config param wi_tile_dlhs_embed_dim: 1024
I0421 08:50:59.462157 138379323594560 pyconfig.py:432] Config param wi_tile_dlhs_mlp_dim: 1024
I0421 08:50:59.462171 138379323594560 pyconfig.py:432] Config param wi_tile_drhs_batch_seq: 512
I0421 08:50:59.462186 138379323594560 pyconfig.py:432] Config param wi_tile_drhs_embed_dim: 1024
I0421 08:50:59.462201 138379323594560 pyconfig.py:432] Config param wi_tile_drhs_mlp_dim: 1024
I0421 08:50:59.462217 138379323594560 pyconfig.py:432] Config param wi_tile_fwd_batch_seq: 512
I0421 08:50:59.462233 138379323594560 pyconfig.py:432] Config param wi_tile_fwd_embed_dim: 1024
I0421 08:50:59.462247 138379323594560 pyconfig.py:432] Config param wi_tile_fwd_mlp_dim: 1024
I0421 08:50:59.462261 138379323594560 pyconfig.py:432] Config param wo_tile_dlhs_batch_seq: 512
I0421 08:50:59.462277 138379323594560 pyconfig.py:432] Config param wo_tile_dlhs_embed_dim: 1024
I0421 08:50:59.462292 138379323594560 pyconfig.py:432] Config param wo_tile_dlhs_mlp_dim: 1024
I0421 08:50:59.462308 138379323594560 pyconfig.py:432] Config param wo_tile_drhs_batch_seq: 512
I0421 08:50:59.462322 138379323594560 pyconfig.py:432] Config param wo_tile_drhs_embed_dim: 1024
I0421 08:50:59.462337 138379323594560 pyconfig.py:432] Config param wo_tile_drhs_mlp_dim: 1024
I0421 08:50:59.462352 138379323594560 pyconfig.py:432] Config param wo_tile_fwd_batch_seq: 512
I0421 08:50:59.462368 138379323594560 pyconfig.py:432] Config param wo_tile_fwd_embed_dim: 1024
I0421 08:50:59.462382 138379323594560 pyconfig.py:432] Config param wo_tile_fwd_mlp_dim: 1024
I0421 08:50:59.462398 138379323594560 pyconfig.py:432] Config param wsd_decay_steps_fraction: 0.1
I0421 08:50:59.462412 138379323594560 pyconfig.py:432] Config param wsd_decay_style: WsdDecayStyle.LINEAR
I0421 08:50:59.462430 138379323594560 pyconfig.py:432] Config param xprof_e2e_enable_fw_power_level_event: False
I0421 08:50:59.462445 138379323594560 pyconfig.py:432] Config param xprof_e2e_enable_fw_thermal_event: False
I0421 08:50:59.462460 138379323594560 pyconfig.py:432] Config param xprof_e2e_enable_fw_throttle_event: False
I0421 08:50:59.462475 138379323594560 pyconfig.py:432] Config param xprof_tpu_power_trace_level: 0
I0421 08:50:59.462492 138379323594560 pyconfig.py:432] Config param z_loss_multiplier: 0.0
I0421 08:50:59.462835 138379323594560 tokenizer.py:245] Tokenizer path: meta-llama/Llama-2-7b-chat-hf
I0421 08:50:59.462873 138379323594560 tokenizer.py:224] Loading HF tokenizer: meta-llama/Llama-2-7b-chat-hf
I0421 08:51:03.580931 138379323594560 _schedule.py:129] A polynomial schedule was set with a non-positive `transition_steps` value; this results in a constant schedule with value `init_value`.
I0421 08:51:03.583968 138379323594560 maxtext_utils.py:1718] Num_devices: 32, shape (1, 4, 1, 8, 1, 1, 1, 1, 1, 1, 1, 1, 1)
I0421 08:51:03.584085 138379323594560 train_distill.py:596] Applying logical axis rules for model initialization and training...
I0421 08:51:03.584156 138379323594560 train_distill.py:600] Loading Student from ...
I0421 08:51:03.584184 138379323594560 train_distill.py:169] --- Student Configuration ---
I0421 08:51:03.584205 138379323594560 train_distill.py:170]   Model Name:      gpt3-52k
I0421 08:51:03.584227 138379323594560 train_distill.py:171]   Dimensions:      1 Layers, 16 Emb Dim, 8 Head Dim
I0421 08:51:03.584246 138379323594560 train_distill.py:174]   Attention Heads: 2 Query, 2 KV
I0421 08:51:03.584264 138379323594560 train_distill.py:175]   Vocab Size:      32000
I0421 08:51:03.584282 138379323594560 train_distill.py:176]   Checkpoint:      
I0421 08:51:03.584298 138379323594560 train_distill.py:465] Initializing model: gpt3-52k...
I0421 08:51:04.989072 138379323594560 train_distill.py:614] Loading Teacher from gs://lance-maxtext/pt_seed_ckpts/pt_seed_ckpts/pt_seed_ckpt_gpt352k_v32k_linen/checkpoints/4/items...
I0421 08:51:04.989182 138379323594560 train_distill.py:169] --- Teacher Configuration ---
I0421 08:51:04.989210 138379323594560 train_distill.py:170]   Model Name:      gpt3-52k
I0421 08:51:04.989236 138379323594560 train_distill.py:171]   Dimensions:      1 Layers, 16 Emb Dim, 8 Head Dim
I0421 08:51:04.989258 138379323594560 train_distill.py:174]   Attention Heads: 2 Query, 2 KV
I0421 08:51:04.989277 138379323594560 train_distill.py:175]   Vocab Size:      32000
I0421 08:51:04.989296 138379323594560 train_distill.py:176]   Checkpoint:      gs://lance-maxtext/pt_seed_ckpts/pt_seed_ckpts/pt_seed_ckpt_gpt352k_v32k_linen/checkpoints/4/items
I0421 08:51:04.989317 138379323594560 train_distill.py:465] Initializing model: gpt3-52k...
I0421 08:51:06.046034 138379323594560 pytree_checkpoint_handler.py:577] save_device_host_concurrent_bytes=None
I0421 08:51:06.046478 138379323594560 base_pytree_checkpoint_handler.py:411] 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 0x7dda3aea96a0>, enable_pinned_host_transfer=False, save_concurrent_bytes: 96000000000 (89.4 GiB), restore_concurrent_bytes: 96000000000 (89.4 GiB)
I0421 08:51:06.046539 138379323594560 abstract_checkpointer.py:35] orbax-checkpoint version: 0.11.28
W0421 08:51:07.000986 138379323594560 checkpoint.py:202] Metadata file does not exist: gs://lance-maxtext/pt_seed_ckpts/pt_seed_ckpts/pt_seed_ckpt_gpt352k_v32k_linen/checkpoints/4/items/_CHECKPOINT_METADATA
I0421 08:51:07.523865    2134 google_auth_provider.cc:181] Running on GCE, using service account 562977990677-compute@developer.gserviceaccount.com
I0421 08:51:08.337331 138379323594560 checkpointer.py:304] Restoring checkpoint from gs://lance-maxtext/pt_seed_ckpts/pt_seed_ckpts/pt_seed_ckpt_gpt352k_v32k_linen/checkpoints/4/items.
W0421 08:51:10.595832 138379323594560 transform_utils.py:230] The transformations API will eventually be replaced by an upgraded design. The current API will not be removed until this point, but it will no longer be actively worked on.
I0421 08:51:10.596199 138379323594560 transform_utils.py:288] The following keys are not loaded from the original tree after applying specified transforms: params/params/decoder/to_nnx__rngs/aqt/count, params/params/decoder/to_nnx__rngs/aqt/key, params/params/decoder/to_nnx__rngs/dropout/count, params/params/decoder/to_nnx__rngs/dropout/key, params/params/decoder/to_nnx__rngs/params/count, params/params/decoder/to_nnx__rngs/params/key
I0421 08:51:11.934052 138379323594560 checkpointer.py:318] Finished restoring checkpoint in 3.96 seconds from gs://lance-maxtext/pt_seed_ckpts/pt_seed_ckpts/pt_seed_ckpt_gpt352k_v32k_linen/checkpoints/4/items.
I0421 08:51:12.619513 138379323594560 train_distill.py:640] Initializing Data Iterators via MaxText pipeline...
I0421 08:51:12.683311 138379323594560 config.py:112] TensorFlow version 2.20.0 available.
I0421 08:51:12.683811 138379323594560 config.py:125] JAX version 0.8.3 available.
E0421 08:51:14.782948 138379323594560 packing.py:209] PackAndBatchOperation is deprecated. Please use lazy_dataset.FirstFitPackIterDataset instead.
I0421 08:51:14.783169 138379323594560 data_loader.py:408] Adding CopyNumPyArrayToSharedMemory MapTransform.
I0421 08:51:14.786171 138379323594560 train_distill.py:410] Input Pipeline Checkpointing: DISABLED
I0421 08:51:14.786234 138379323594560 train_distill.py:414] Reason: Iterator 'MultiHostDataLoadIterator' is not recognized as Grain (dataset_type='DatasetType.HF', has_save=False)
I0421 08:51:14.786296 138379323594560 pytree_checkpoint_handler.py:577] save_device_host_concurrent_bytes=None
I0421 08:51:14.786376 138379323594560 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 0x7dda3aea96a0>, enable_pinned_host_transfer=False, save_concurrent_bytes: 96000000000 (89.4 GiB), restore_concurrent_bytes: 96000000000 (89.4 GiB)
I0421 08:51:14.786417 138379323594560 pytree_checkpoint_handler.py:577] save_device_host_concurrent_bytes=None
I0421 08:51:14.786448 138379323594560 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 0x7dda3aea96a0>, enable_pinned_host_transfer=False, save_concurrent_bytes: 96000000000 (89.4 GiB), restore_concurrent_bytes: 96000000000 (89.4 GiB)
I0421 08:51:14.786490 138379323594560 checkpoint_manager.py:702] [process=7][thread=MainThread] CheckpointManager init: checkpointers=None, item_names=None, item_handlers={'model_params': <orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeCheckpointHandler object at 0x7dceff64d100>, 'optimizer_state': <orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeCheckpointHandler object at 0x7dc2c0e0b3b0>, 'custom_metadata': <orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonCheckpointHandler object at 0x7dc2c0e0b2f0>}, handler_registry=None
I0421 08:51:14.786682 138379323594560 composite_checkpoint_handler.py:237] Deferred registration for item: "model_params". Adding handler `<orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeCheckpointHandler object at 0x7dceff64d100>` for item "model_params" 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`.
I0421 08:51:14.786723 138379323594560 composite_checkpoint_handler.py:237] Deferred registration for item: "optimizer_state". Adding handler `<orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeCheckpointHandler object at 0x7dc2c0e0b3b0>` for item "optimizer_state" 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`.
I0421 08:51:14.786762 138379323594560 composite_checkpoint_handler.py:237] Deferred registration for item: "custom_metadata". Adding handler `<orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonCheckpointHandler object at 0x7dc2c0e0b2f0>` for item "custom_metadata" 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`.
I0421 08:51:14.786793 138379323594560 composite_checkpoint_handler.py:237] Deferred registration for item: "metrics". Adding handler `<orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonCheckpointHandler object at 0x7dc2c0e0af30>` 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`.
I0421 08:51:14.786820 138379323594560 composite_checkpoint_handler.py:505] Initialized registry DefaultCheckpointHandlerRegistry({('model_params', <class 'orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeSaveArgs'>): <orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeCheckpointHandler object at 0x7dceff64d100>, ('model_params', <class 'orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeRestoreArgs'>): <orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeCheckpointHandler object at 0x7dceff64d100>, ('optimizer_state', <class 'orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeSaveArgs'>): <orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeCheckpointHandler object at 0x7dc2c0e0b3b0>, ('optimizer_state', <class 'orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeRestoreArgs'>): <orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeCheckpointHandler object at 0x7dc2c0e0b3b0>, ('custom_metadata', <class 'orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonSaveArgs'>): <orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonCheckpointHandler object at 0x7dc2c0e0b2f0>, ('custom_metadata', <class 'orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonRestoreArgs'>): <orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonCheckpointHandler object at 0x7dc2c0e0b2f0>, ('metrics', <class 'orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonSaveArgs'>): <orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonCheckpointHandler object at 0x7dc2c0e0af30>, ('metrics', <class 'orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonRestoreArgs'>): <orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonCheckpointHandler object at 0x7dc2c0e0af30>}).
I0421 08:51:14.787224 138379323594560 async_checkpointer.py:177] [process=7][thread=MainThread] Using barrier_sync_fn: <function get_barrier_sync_fn.<locals>._fn at 0x7dc2c0b34040> timeout: 600 secs and primary_host=0 for async checkpoint writes
I0421 08:51:17.295549 138379323594560 checkpoint_manager.py:1788] Found 0 checkpoint steps in gs://lance-maxtext/pt_ckpt_xpk_feat_nnx_trainstate_and_training_loop_20260421_083247/pt_distill_linen_xpk_feat_nnx_trainstate_and_training_loop_20260421_083247_07_distill_smoke/checkpoints
I0421 08:51:17.320507 138379323594560 checkpoint_manager.py:921] [process=7][thread=MainThread] CheckpointManager created,  primary_host=0, CheckpointManagerOptions=CheckpointManagerOptions(save_interval_steps=2000, 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=None, preservation_policy=None, prevent_write_metrics=False, enable_should_save_is_saving_in_progress_check=True, enable_per_process_directory_creation=False), root_directory=gs://lance-maxtext/pt_ckpt_xpk_feat_nnx_trainstate_and_training_loop_20260421_083247/pt_distill_linen_xpk_feat_nnx_trainstate_and_training_loop_20260421_083247_07_distill_smoke/checkpoints: <orbax.checkpoint.checkpoint_manager.CheckpointManager object at 0x7dc2c0e0b2c0>
I0421 08:51:17.320640 138379323594560 pytree_checkpoint_handler.py:577] save_device_host_concurrent_bytes=None
I0421 08:51:17.320706 138379323594560 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 0x7dda3aea96a0>, enable_pinned_host_transfer=False, save_concurrent_bytes: 96000000000 (89.4 GiB), restore_concurrent_bytes: 96000000000 (89.4 GiB)
I0421 08:51:17.320755 138379323594560 pytree_checkpoint_handler.py:577] save_device_host_concurrent_bytes=None
I0421 08:51:17.320797 138379323594560 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 0x7dda3aea96a0>, enable_pinned_host_transfer=False, save_concurrent_bytes: 96000000000 (89.4 GiB), restore_concurrent_bytes: 96000000000 (89.4 GiB)
I0421 08:51:17.320835 138379323594560 checkpoint_manager.py:1983] [process=7][thread=MainThread][wait_until_finished] No Save Finalize thread to wait for. Returning.
I0421 08:51:17.320887 138379323594560 checkpoint.py:459] Closing _NonBlockingMetadataStore(enable_write=True, _write_lock=<locked _thread.RLock object owner=138379323594560 count=1 at 0x7dc10045a140>, _store_impl=<orbax.checkpoint._src.metadata.checkpoint._MetadataStoreImpl object at 0x7dc2c0e0b0b0>, _single_thread_executor=<concurrent.futures.thread.ThreadPoolExecutor object at 0x7dc2c0e0b080>, _write_futures=[])
I0421 08:51:17.321240 138379323594560 checkpoint.py:459] Closing _NonBlockingMetadataStore(enable_write=True, _write_lock=<locked _thread.RLock object owner=138379323594560 count=1 at 0x7dc10045a140>, _store_impl=<orbax.checkpoint._src.metadata.checkpoint._MetadataStoreImpl object at 0x7dc2c0e0b0b0>, _single_thread_executor=<concurrent.futures.thread.ThreadPoolExecutor object at 0x7dc2c0e0b080>, _write_futures=[])
I0421 08:51:17.321268 138379323594560 checkpoint.py:459] Closing _NonBlockingMetadataStore(enable_write=True, _write_lock=<locked _thread.RLock object owner=138379323594560 count=1 at 0x7dc10045a140>, _store_impl=<orbax.checkpoint._src.metadata.checkpoint._MetadataStoreImpl object at 0x7dc2c0e0b0b0>, _single_thread_executor=<concurrent.futures.thread.ThreadPoolExecutor object at 0x7dc2c0e0b080>, _write_futures=[])
I0421 08:51:17.321300 138379323594560 checkpoint_manager.py:702] [process=7][thread=MainThread] CheckpointManager init: checkpointers=None, item_names=None, item_handlers={'model_params': <orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeCheckpointHandler object at 0x7dc2c0e0b290>, 'optimizer_state': <orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeCheckpointHandler object at 0x7dc2c0a3a000>, 'custom_metadata': <orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonCheckpointHandler object at 0x7dc2c0a396d0>, 'iter': <maxtext.common.checkpointing.GrainCheckpointHandler object at 0x7dc2c0a39ee0>}, handler_registry=None
I0421 08:51:17.321401 138379323594560 composite_checkpoint_handler.py:237] Deferred registration for item: "model_params". Adding handler `<orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeCheckpointHandler object at 0x7dc2c0e0b290>` for item "model_params" 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`.
I0421 08:51:17.321433 138379323594560 composite_checkpoint_handler.py:237] Deferred registration for item: "optimizer_state". Adding handler `<orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeCheckpointHandler object at 0x7dc2c0a3a000>` for item "optimizer_state" 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`.
I0421 08:51:17.321456 138379323594560 composite_checkpoint_handler.py:237] Deferred registration for item: "custom_metadata". Adding handler `<orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonCheckpointHandler object at 0x7dc2c0a396d0>` for item "custom_metadata" 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`.
I0421 08:51:17.321485 138379323594560 composite_checkpoint_handler.py:237] Deferred registration for item: "iter". Adding handler `<maxtext.common.checkpointing.GrainCheckpointHandler object at 0x7dc2c0a39ee0>` for item "iter" and save args `<class 'maxtext.common.checkpointing.GrainCheckpointSave'>` and restore args `<class 'maxtext.common.checkpointing.GrainCheckpointRestore'>` to `_handler_registry`.
I0421 08:51:17.321508 138379323594560 composite_checkpoint_handler.py:237] Deferred registration for item: "metrics". Adding handler `<orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonCheckpointHandler object at 0x7dc2c0a3a270>` 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`.
I0421 08:51:17.321532 138379323594560 composite_checkpoint_handler.py:505] Initialized registry DefaultCheckpointHandlerRegistry({('model_params', <class 'orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeSaveArgs'>): <orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeCheckpointHandler object at 0x7dc2c0e0b290>, ('model_params', <class 'orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeRestoreArgs'>): <orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeCheckpointHandler object at 0x7dc2c0e0b290>, ('optimizer_state', <class 'orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeSaveArgs'>): <orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeCheckpointHandler object at 0x7dc2c0a3a000>, ('optimizer_state', <class 'orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeRestoreArgs'>): <orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeCheckpointHandler object at 0x7dc2c0a3a000>, ('custom_metadata', <class 'orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonSaveArgs'>): <orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonCheckpointHandler object at 0x7dc2c0a396d0>, ('custom_metadata', <class 'orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonRestoreArgs'>): <orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonCheckpointHandler object at 0x7dc2c0a396d0>, ('iter', <class 'maxtext.common.checkpointing.GrainCheckpointSave'>): <maxtext.common.checkpointing.GrainCheckpointHandler object at 0x7dc2c0a39ee0>, ('iter', <class 'maxtext.common.checkpointing.GrainCheckpointRestore'>): <maxtext.common.checkpointing.GrainCheckpointHandler object at 0x7dc2c0a39ee0>, ('metrics', <class 'orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonSaveArgs'>): <orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonCheckpointHandler object at 0x7dc2c0a3a270>, ('metrics', <class 'orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonRestoreArgs'>): <orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonCheckpointHandler object at 0x7dc2c0a3a270>}).
I0421 08:51:17.321604 138379323594560 async_checkpointer.py:177] [process=7][thread=MainThread] Using barrier_sync_fn: <function get_barrier_sync_fn.<locals>._fn at 0x7dc2c0b34220> timeout: 600 secs and primary_host=0 for async checkpoint writes
I0421 08:51:17.698167 138379323594560 checkpoint_manager.py:1788] Found 0 checkpoint steps in gs://lance-maxtext/pt_ckpt_xpk_feat_nnx_trainstate_and_training_loop_20260421_083247/pt_distill_linen_xpk_feat_nnx_trainstate_and_training_loop_20260421_083247_07_distill_smoke/checkpoints
I0421 08:51:18.129145 138379323594560 checkpoint_manager.py:921] [process=7][thread=MainThread] CheckpointManager created,  primary_host=0, CheckpointManagerOptions=CheckpointManagerOptions(save_interval_steps=2000, 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=None, preservation_policy=None, prevent_write_metrics=False, enable_should_save_is_saving_in_progress_check=True, enable_per_process_directory_creation=False), root_directory=gs://lance-maxtext/pt_ckpt_xpk_feat_nnx_trainstate_and_training_loop_20260421_083247/pt_distill_linen_xpk_feat_nnx_trainstate_and_training_loop_20260421_083247_07_distill_smoke/checkpoints: <orbax.checkpoint.checkpoint_manager.CheckpointManager object at 0x7dc2c0e0a7e0>
I0421 08:51:18.129755 138379323594560 train_distill.py:691] Starting Distillation Training...
I0421 08:51:18.129870 138379323594560 peft_trainer.py:590] Training with mesh: Mesh('diloco': 1, 'data': 4, 'stage': 1, 'fsdp': 8, 'fsdp_transpose': 1, 'sequence': 1, 'context': 1, 'context_autoregressive': 1, 'tensor': 1, 'tensor_transpose': 1, 'tensor_sequence': 1, 'expert': 1, 'autoregressive': 1, axis_types=(Auto, Auto, Auto, Auto, Auto, Auto, Auto, Auto, Auto, Auto, Auto, Auto, Auto))
I0421 08:51:18.478284 138379323594560 peft_trainer.py:600] Compiled train_step cache size: 0

Training:   0%|          | 0/5 [00:00<?, ?step/s]I0421 08:51:18.480178 138235443398400 grain_pool.py:367] Grain pool will use 1 processes.
I0421 08:51:18.506704 138235443398400 grain_pool.py:440] Grain pool will start child processes.
I0421 08:51:18.512191 138235443398400 grain_pool.py:448] Grain pool started all child processes.
2026-04-21 08:51:24.570743: 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)
I0421 08:51:27.458694 138379323594560 utils.py:86] Train loop finished in: 8.9798 seconds
Traceback (most recent call last):
  File "<frozen runpy>", line 198, in _run_module_as_main
  File "<frozen runpy>", line 88, in _run_code
  File "/deps/src/maxtext/trainers/post_train/distillation/train_distill.py", line 765, in <module>
    app.run(main)
  File "/usr/local/lib/python3.12/site-packages/absl/app.py", line 316, in run
    _run_main(main, args)
  File "/usr/local/lib/python3.12/site-packages/absl/app.py", line 261, in _run_main
    sys.exit(main(argv))
             ^^^^^^^^^^
  File "/deps/src/maxtext/trainers/post_train/distillation/train_distill.py", line 761, in main
    train_distill(student_config, teacher_config, is_offline, global_config.offline_data_dir)
  File "/deps/src/maxtext/trainers/post_train/distillation/train_distill.py", line 693, in train_distill
    trainer.train(train_iter, eval_iter)
  File "/usr/local/lib/python3.12/site-packages/tunix/sft/peft_trainer.py", line 659, in train
    train_example = sharding_utils.shard_input(
                    ^^^^^^^^^^^^^^^^^^^^^^^^^^^
  File "/usr/local/lib/python3.12/site-packages/tunix/sft/sharding_utils.py", line 58, in shard_input
    return jax.tree.map(
           ^^^^^^^^^^^^^
  File "/usr/local/lib/python3.12/site-packages/jax/_src/tree.py", line 155, in map
    return tree_util.tree_map(f, tree, *rest, is_leaf=is_leaf)
           ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
  File "/usr/local/lib/python3.12/site-packages/jax/_src/tree_util.py", line 362, in tree_map
    return treedef.unflatten(f(*xs) for xs in zip(*all_leaves))
           ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
  File "/usr/local/lib/python3.12/site-packages/jax/_src/tree_util.py", line 362, in <genexpr>
    return treedef.unflatten(f(*xs) for xs in zip(*all_leaves))
                             ^^^^^^
  File "/usr/local/lib/python3.12/site-packages/tunix/sft/sharding_utils.py", line 59, in <lambda>
    lambda x: jax.make_array_from_process_local_data(
              ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
  File "/usr/local/lib/python3.12/site-packages/jax/_src/array.py", line 986, in make_array_from_process_local_data
    out = [_array_from_process_local_data(data, s, shape)
           ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
  File "/usr/local/lib/python3.12/site-packages/jax/_src/array.py", line 1048, in _array_from_process_local_data
    return make_array_from_callback(global_shape, sharding, cb)
           ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
  File "/usr/local/lib/python3.12/site-packages/jax/_src/array.py", line 845, in make_array_from_callback
    per_device_values = api.device_put(per_device_values, devices)
                        ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
  File "/usr/local/lib/python3.12/site-packages/jax/_src/api.py", line 2729, in device_put
    out_flat = dispatch._batched_device_put_impl(
               ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
  File "/usr/local/lib/python3.12/site-packages/jax/_src/dispatch.py", line 558, in _batched_device_put_impl
    y = _device_put_impl(x, device=device, src=src, copy=cp, aval=aval)
        ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
  File "/usr/local/lib/python3.12/site-packages/jax/_src/dispatch.py", line 545, in _device_put_impl
    return _device_put_sharding_impl(x, aval, device, copy)
           ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
  File "/usr/local/lib/python3.12/site-packages/jax/_src/dispatch.py", line 487, in _device_put_sharding_impl
    raise ValueError(
ValueError: device_put's first argument must be a fully addressable array, but got value with devices {TpuDevice(id=25, process_index=6, coords=(1,6,0), core_on_chip=0), TpuDevice(id=0, process_index=0, coords=(0,0,0), core_on_chip=0), TpuDevice(id=22, process_index=5, coords=(2,5,0), core_on_chip=0), TpuDevice(id=31, process_index=7, coords=(3,7,0), core_on_chip=0), TpuDevice(id=21, process_index=4, coords=(1,5,0), core_on_chip=0), TpuDevice(id=13, process_index=2, coords=(1,3,0), core_on_chip=0), TpuDevice(id=7, process_index=1, coords=(3,1,0), core_on_chip=0), TpuDevice(id=15, process_index=3, coords=(3,3,0), core_on_chip=0), TpuDevice(id=2, process_index=1, coords=(2,0,0), core_on_chip=0), TpuDevice(id=1, process_index=0, coords=(1,0,0), core_on_chip=0), TpuDevice(id=28, process_index=6, coords=(0,7,0), core_on_chip=0), TpuDevice(id=27, process_index=7, coords=(3,6,0), core_on_chip=0), TpuDevice(id=18, process_index=5, coords=(2,4,0), core_on_chip=0), TpuDevice(id=23, process_index=5, coords=(3,5,0), core_on_chip=0), TpuDevice(id=10, process_index=3, coords=(2,2,0), core_on_chip=0), TpuDevice(id=16, process_index=4, coords=(0,4,0), core_on_chip=0), TpuDevice(id=8, process_index=2, coords=(0,2,0), core_on_chip=0), TpuDevice(id=24, process_index=6, coords=(0,6,0), core_on_chip=0), TpuDevice(id=30, process_index=7, coords=(2,7,0), core_on_chip=0), TpuDevice(id=19, process_index=5, coords=(3,4,0), core_on_chip=0), TpuDevice(id=4, process_index=0, coords=(0,1,0), core_on_chip=0), TpuDevice(id=3, process_index=1, coords=(3,0,0), core_on_chip=0), TpuDevice(id=29, process_index=6, coords=(1,7,0), core_on_chip=0), TpuDevice(id=5, process_index=0, coords=(1,1,0), core_on_chip=0), TpuDevice(id=9, process_index=2, coords=(1,2,0), core_on_chip=0), TpuDevice(id=17, process_index=4, coords=(1,4,0), core_on_chip=0), TpuDevice(id=11, process_index=3, coords=(3,2,0), core_on_chip=0), TpuDevice(id=26, process_index=7, coords=(2,6,0), core_on_chip=0), TpuDevice(id=6, process_index=1, coords=(2,1,0), core_on_chip=0), TpuDevice(id=12, process_index=2, coords=(0,3,0), core_on_chip=0), TpuDevice(id=20, process_index=4, coords=(0,5,0), core_on_chip=0), TpuDevice(id=14, process_index=3, coords=(2,3,0), core_on_chip=0)}
I0421 08:51:27.800003 138235443398400 grain_pool.py:542] Grain pool is exiting.
I0421 08:51:27.800119 138235443398400 grain_pool.py:547] Shutting down multiprocessing system.
I0421 08:51:29.255345 138235443398400 grain_pool.py:547] Shutting down multiprocessing system.

Training:   0%|          | 0/5 [00:13<?, ?step/s]
/usr/local/lib/python3.12/multiprocessing/resource_tracker.py:279: UserWarning: resource_tracker: There appear to be 15 leaked shared_memory objects to clean up at shutdown
  warnings.warn('resource_tracker: There appear to be %d '
XPK End: Tue Apr 21 08:51:36 UTC 2026
EXIT_CODE=1