MaxView

XPK Start: Mon Apr 20 08:33:59 UTC 2026
2026-04-20 08:34:15.902388: 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)
I0420 08:34:19.455910 139954941523776 max_utils.py:273] Attempting to initialize the jax distributed system...
INFO:2026-04-20 08:34:28,495:jax._src.distributed:149: Starting JAX distributed service on [::]:8482
I0420 08:34:28.495689 139954941523776 distributed.py:149] Starting JAX distributed service on [::]:8482
INFO:2026-04-20 08:34:28,497:jax._src.distributed:166: Connecting to JAX distributed service on mt-07-distill-smoke-8rm6x-slice-job-0-0.mt-07-distill-smoke-8rm6x:8482
I0420 08:34:28.497924 139954941523776 distributed.py:166] Connecting to JAX distributed service on mt-07-distill-smoke-8rm6x-slice-job-0-0.mt-07-distill-smoke-8rm6x:8482
I0420 08:34:29.794170 139954941523776 max_utils.py:284] Jax distributed system initialized!
I0420 08:34:36.052828 139954941523776 max_utils.py:244] Jax distributed system is already initialized.
I0420 08:34:36.520008 139954941523776 max_utils.py:244] Jax distributed system is already initialized.
I0420 08:34:36.521322 139954941523776 pyconfig.py:432] Config param abort_on_inf_loss: True
I0420 08:34:36.521389 139954941523776 pyconfig.py:432] Config param abort_on_nan_loss: True
I0420 08:34:36.521417 139954941523776 pyconfig.py:432] Config param act_quantization_calibration_method: absmax
I0420 08:34:36.521439 139954941523776 pyconfig.py:432] Config param activation_dropout_for_audio: 0.0
I0420 08:34:36.521459 139954941523776 pyconfig.py:432] Config param activation_function_for_audio: gelu
I0420 08:34:36.521477 139954941523776 pyconfig.py:432] Config param activations_in_float32: False
I0420 08:34:36.521495 139954941523776 pyconfig.py:432] Config param adam_b1: 0.9
I0420 08:34:36.521513 139954941523776 pyconfig.py:432] Config param adam_b2: 0.95
I0420 08:34:36.521530 139954941523776 pyconfig.py:432] Config param adam_eps: 1e-08
I0420 08:34:36.521557 139954941523776 pyconfig.py:432] Config param adam_eps_root: 0.0
I0420 08:34:36.521573 139954941523776 pyconfig.py:432] Config param adam_weight_decay: 0.1
I0420 08:34:36.521589 139954941523776 pyconfig.py:432] Config param adamw_mask: []
I0420 08:34:36.521604 139954941523776 pyconfig.py:432] Config param add_bos: True
I0420 08:34:36.521620 139954941523776 pyconfig.py:432] Config param add_eos: True
I0420 08:34:36.521635 139954941523776 pyconfig.py:432] Config param allow_split_physical_axes: False
I0420 08:34:36.521651 139954941523776 pyconfig.py:432] Config param ar_cache_axis_order: 1,2,0,3
I0420 08:34:36.521666 139954941523776 pyconfig.py:432] Config param async_checkpointing: True
I0420 08:34:36.521682 139954941523776 pyconfig.py:432] Config param async_scheduling: False
I0420 08:34:36.521697 139954941523776 pyconfig.py:432] Config param attention: dot_product
I0420 08:34:36.521714 139954941523776 pyconfig.py:432] Config param attention_bias: False
I0420 08:34:36.521729 139954941523776 pyconfig.py:432] Config param attention_dropout_for_audio: 0.0
I0420 08:34:36.521747 139954941523776 pyconfig.py:432] Config param attention_out: RematLocation.REMAT
I0420 08:34:36.521767 139954941523776 pyconfig.py:432] Config param attention_output_dim: -1
I0420 08:34:36.521784 139954941523776 pyconfig.py:432] Config param attention_sink: False
I0420 08:34:36.521801 139954941523776 pyconfig.py:432] Config param attention_type: global
I0420 08:34:36.521817 139954941523776 pyconfig.py:432] Config param attn_logits_soft_cap: None
I0420 08:34:36.521832 139954941523776 pyconfig.py:432] Config param audio_path: 
I0420 08:34:36.521848 139954941523776 pyconfig.py:432] Config param audio_placeholder: <|audio|>
I0420 08:34:36.521863 139954941523776 pyconfig.py:432] Config param autoregressive_decode_assert: 
I0420 08:34:36.521878 139954941523776 pyconfig.py:432] Config param base_config: base.yml
I0420 08:34:36.521893 139954941523776 pyconfig.py:432] Config param base_emb_dim: 16
I0420 08:34:36.521909 139954941523776 pyconfig.py:432] Config param base_mlp_dim: 64
I0420 08:34:36.521924 139954941523776 pyconfig.py:432] Config param base_moe_mlp_dim: -1
I0420 08:34:36.521939 139954941523776 pyconfig.py:432] Config param base_num_decoder_layers: 1
I0420 08:34:36.521956 139954941523776 pyconfig.py:432] Config param base_num_kv_heads: 2
I0420 08:34:36.521972 139954941523776 pyconfig.py:432] Config param base_num_query_heads: 2
I0420 08:34:36.521987 139954941523776 pyconfig.py:432] Config param base_output_directory: 
I0420 08:34:36.522002 139954941523776 pyconfig.py:432] Config param batch_size: 1
I0420 08:34:36.522017 139954941523776 pyconfig.py:432] Config param batch_split_factor: 1
I0420 08:34:36.522033 139954941523776 pyconfig.py:432] Config param beta_fast: 32
I0420 08:34:36.522047 139954941523776 pyconfig.py:432] Config param beta_slow: 1
I0420 08:34:36.522063 139954941523776 pyconfig.py:432] Config param bwd_quantization_calibration_method: absmax
I0420 08:34:36.522078 139954941523776 pyconfig.py:432] Config param capacity_factor: -1.0
I0420 08:34:36.522095 139954941523776 pyconfig.py:432] Config param cast_logits_to_fp32: True
I0420 08:34:36.522111 139954941523776 pyconfig.py:432] Config param chat_template: 
I0420 08:34:36.522126 139954941523776 pyconfig.py:432] Config param chat_template_path: 
I0420 08:34:36.522142 139954941523776 pyconfig.py:432] Config param checkpoint_conversion_fn: None
I0420 08:34:36.522159 139954941523776 pyconfig.py:432] Config param checkpoint_dir: None
I0420 08:34:36.522177 139954941523776 pyconfig.py:432] Config param checkpoint_is_quantized: False
I0420 08:34:36.522195 139954941523776 pyconfig.py:432] Config param checkpoint_period: 2000
I0420 08:34:36.522212 139954941523776 pyconfig.py:432] Config param checkpoint_storage_concurrent_gb: 96
I0420 08:34:36.522227 139954941523776 pyconfig.py:432] Config param checkpoint_storage_target_data_file_size_bytes: 2147483648
I0420 08:34:36.522244 139954941523776 pyconfig.py:432] Config param checkpoint_storage_use_ocdbt: True
I0420 08:34:36.522260 139954941523776 pyconfig.py:432] Config param checkpoint_storage_use_zarr3: True
I0420 08:34:36.522276 139954941523776 pyconfig.py:432] Config param checkpoint_todelete_full_path: None
I0420 08:34:36.522290 139954941523776 pyconfig.py:432] Config param checkpoint_todelete_subdir: None
I0420 08:34:36.522305 139954941523776 pyconfig.py:432] Config param chips_per_vm: 4
I0420 08:34:36.522320 139954941523776 pyconfig.py:432] Config param chunk_attn_window_size: 0
I0420 08:34:36.522347 139954941523776 pyconfig.py:432] Config param collect_stack_trace: False
I0420 08:34:36.522371 139954941523776 pyconfig.py:432] Config param colocated_python_checkpointing: False
I0420 08:34:36.522387 139954941523776 pyconfig.py:432] Config param colocated_python_data_input: False
I0420 08:34:36.522402 139954941523776 pyconfig.py:432] Config param compile_topology: 
I0420 08:34:36.522417 139954941523776 pyconfig.py:432] Config param compile_topology_num_slices: -1
I0420 08:34:36.522431 139954941523776 pyconfig.py:432] Config param compile_xla_flags: 
I0420 08:34:36.522447 139954941523776 pyconfig.py:432] Config param compiled_trainstep_file: 
I0420 08:34:36.522461 139954941523776 pyconfig.py:432] Config param compute_axis_order: 0,1,2,3
I0420 08:34:36.522477 139954941523776 pyconfig.py:432] Config param constant_bound_config: []
I0420 08:34:36.522491 139954941523776 pyconfig.py:432] Config param context: RematLocation.REMAT
I0420 08:34:36.522508 139954941523776 pyconfig.py:432] Config param context_parallel_load_balance: True
I0420 08:34:36.522522 139954941523776 pyconfig.py:432] Config param context_parallel_size: 1
I0420 08:34:36.522537 139954941523776 pyconfig.py:432] Config param context_parallel_strategy: all_gather
I0420 08:34:36.522557 139954941523776 pyconfig.py:432] Config param context_sharding: context
I0420 08:34:36.522572 139954941523776 pyconfig.py:432] Config param conv_chunksize_for_audio: 500
I0420 08:34:36.522586 139954941523776 pyconfig.py:432] Config param conv_stride_for_vit: 14
I0420 08:34:36.522601 139954941523776 pyconfig.py:432] Config param cost_estimate_flops_bwd: -1
I0420 08:34:36.522616 139954941523776 pyconfig.py:432] Config param cost_estimate_flops_fwd: -1
I0420 08:34:36.522633 139954941523776 pyconfig.py:432] Config param custom_mesh: 
I0420 08:34:36.522648 139954941523776 pyconfig.py:432] Config param custom_mesh_and_rule: 
I0420 08:34:36.522663 139954941523776 pyconfig.py:432] Config param d_model_for_audio: 256
I0420 08:34:36.522678 139954941523776 pyconfig.py:432] Config param data_sharding: (('data', 'stage', 'fsdp', 'fsdp_transpose', 'sequence', 'context', 'context_autoregressive', 'tensor', 'tensor_transpose', 'tensor_sequence', 'expert', 'autoregressive'),)
I0420 08:34:36.522697 139954941523776 pyconfig.py:432] Config param data_shuffle_seed: 0
I0420 08:34:36.522713 139954941523776 pyconfig.py:432] Config param dataset_name: c4/en:3.0.1
I0420 08:34:36.522729 139954941523776 pyconfig.py:432] Config param dataset_path: 
I0420 08:34:36.522743 139954941523776 pyconfig.py:432] Config param dataset_type: DatasetType.HF
I0420 08:34:36.522760 139954941523776 pyconfig.py:432] Config param dcn_autoregressive_parallelism: 1
I0420 08:34:36.522774 139954941523776 pyconfig.py:432] Config param dcn_context_autoregressive_parallelism: 1
I0420 08:34:36.522788 139954941523776 pyconfig.py:432] Config param dcn_context_parallelism: 1
I0420 08:34:36.522802 139954941523776 pyconfig.py:432] Config param dcn_data_parallelism: -1
I0420 08:34:36.522818 139954941523776 pyconfig.py:432] Config param dcn_diloco_parallelism: 1
I0420 08:34:36.522833 139954941523776 pyconfig.py:432] Config param dcn_expert_parallelism: 1
I0420 08:34:36.522847 139954941523776 pyconfig.py:432] Config param dcn_fsdp_parallelism: 1
I0420 08:34:36.522870 139954941523776 pyconfig.py:432] Config param dcn_fsdp_transpose_parallelism: 1
I0420 08:34:36.522884 139954941523776 pyconfig.py:432] Config param dcn_parallelism: [1, -1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1]
I0420 08:34:36.522900 139954941523776 pyconfig.py:432] Config param dcn_pipeline_parallelism: 1
I0420 08:34:36.522916 139954941523776 pyconfig.py:432] Config param dcn_sequence_parallelism: 1
I0420 08:34:36.522930 139954941523776 pyconfig.py:432] Config param dcn_tensor_parallelism: 1
I0420 08:34:36.522945 139954941523776 pyconfig.py:432] Config param dcn_tensor_sequence_parallelism: 1
I0420 08:34:36.522959 139954941523776 pyconfig.py:432] Config param dcn_tensor_transpose_parallelism: 1
I0420 08:34:36.522974 139954941523776 pyconfig.py:432] Config param debug: {'rl': False}
I0420 08:34:36.522991 139954941523776 pyconfig.py:432] Config param debug_sharding: False
I0420 08:34:36.523007 139954941523776 pyconfig.py:432] Config param decode_sampling_nucleus_p: -1
I0420 08:34:36.523021 139954941523776 pyconfig.py:432] Config param decode_sampling_strategy: SamplingStrategy.GREEDY
I0420 08:34:36.523038 139954941523776 pyconfig.py:432] Config param decode_sampling_temperature: 1.0
I0420 08:34:36.523053 139954941523776 pyconfig.py:432] Config param decode_sampling_top_k: 0
I0420 08:34:36.523068 139954941523776 pyconfig.py:432] Config param decoder_block: DecoderBlockType.GPT3
I0420 08:34:36.523084 139954941523776 pyconfig.py:432] Config param decoder_layer_input: RematLocation.DEVICE
I0420 08:34:36.523100 139954941523776 pyconfig.py:432] Config param deepstack_visual_indexes_for_vit: []
I0420 08:34:36.523116 139954941523776 pyconfig.py:432] Config param degenerate_group_masking: True
I0420 08:34:36.523132 139954941523776 pyconfig.py:432] Config param dense_init_scale: 1.0
I0420 08:34:36.523147 139954941523776 pyconfig.py:432] Config param diloco_outer_lr: 0.3
I0420 08:34:36.523163 139954941523776 pyconfig.py:432] Config param diloco_outer_momentum: 0.9
I0420 08:34:36.523179 139954941523776 pyconfig.py:432] Config param diloco_sync_period: 36
I0420 08:34:36.523193 139954941523776 pyconfig.py:432] Config param distill_alpha: 0.5
I0420 08:34:36.523208 139954941523776 pyconfig.py:432] Config param distill_beta: 0.0
I0420 08:34:36.523223 139954941523776 pyconfig.py:432] Config param distill_feature_loss_type: cosine
I0420 08:34:36.523238 139954941523776 pyconfig.py:432] Config param distill_layer_indices: None
I0420 08:34:36.523254 139954941523776 pyconfig.py:432] Config param distill_temperature: 1.0
I0420 08:34:36.523270 139954941523776 pyconfig.py:432] Config param downsample_hidden_size_for_audio: 256
I0420 08:34:36.523284 139954941523776 pyconfig.py:432] Config param dpo_beta: 0.1
I0420 08:34:36.523301 139954941523776 pyconfig.py:432] Config param dpo_label_smoothing: 0.0
I0420 08:34:36.523316 139954941523776 pyconfig.py:432] Config param dq_reduction_steps: 0
I0420 08:34:36.523330 139954941523776 pyconfig.py:432] Config param dropout_rate: 0.0
I0420 08:34:36.523387 139954941523776 pyconfig.py:432] Config param dtype: bfloat16
I0420 08:34:36.523418 139954941523776 pyconfig.py:432] Config param dtype_mm: float32
I0420 08:34:36.523435 139954941523776 pyconfig.py:432] Config param dump_hlo: False
I0420 08:34:36.523450 139954941523776 pyconfig.py:432] Config param dump_hlo_delete_local_after: True
I0420 08:34:36.523466 139954941523776 pyconfig.py:432] Config param dump_hlo_gcs_dir: gpt3-52k_2026-04-20-08-34/xla_dump
I0420 08:34:36.523481 139954941523776 pyconfig.py:432] Config param dump_hlo_local_dir: /tmp/xla_dump/
I0420 08:34:36.523496 139954941523776 pyconfig.py:432] Config param dump_hlo_local_module_name: jit_train_step
I0420 08:34:36.523512 139954941523776 pyconfig.py:432] Config param dump_hlo_module_name: jit_train_step
I0420 08:34:36.523526 139954941523776 pyconfig.py:432] Config param dump_hlo_upload_all: False
I0420 08:34:36.523546 139954941523776 pyconfig.py:432] Config param dump_hlo_xla_flags: 
I0420 08:34:36.523561 139954941523776 pyconfig.py:432] Config param dump_jaxpr: False
I0420 08:34:36.523576 139954941523776 pyconfig.py:432] Config param dump_jaxpr_delete_local_after: True
I0420 08:34:36.523591 139954941523776 pyconfig.py:432] Config param dump_jaxpr_gcs_dir: gpt3-52k_2026-04-20-08-34/jaxpr_dump
I0420 08:34:36.523606 139954941523776 pyconfig.py:432] Config param dump_jaxpr_local_dir: /tmp/jaxpr_dump/
I0420 08:34:36.523621 139954941523776 pyconfig.py:432] Config param dump_step: -1
I0420 08:34:36.523635 139954941523776 pyconfig.py:432] Config param elastic_enabled: False
I0420 08:34:36.523651 139954941523776 pyconfig.py:432] Config param elastic_max_retries: 10
I0420 08:34:36.523667 139954941523776 pyconfig.py:432] Config param elastic_timeout_seconds: 300
I0420 08:34:36.523682 139954941523776 pyconfig.py:432] Config param emb_dim: 16
I0420 08:34:36.523697 139954941523776 pyconfig.py:432] Config param enable_autocheckpoint: False
I0420 08:34:36.523711 139954941523776 pyconfig.py:432] Config param enable_checkpoint_cloud_logger: False
I0420 08:34:36.523727 139954941523776 pyconfig.py:432] Config param enable_checkpointing: True
I0420 08:34:36.523741 139954941523776 pyconfig.py:432] Config param enable_continuous_checkpointing: False
I0420 08:34:36.523756 139954941523776 pyconfig.py:432] Config param enable_data_shuffling: True
I0420 08:34:36.523772 139954941523776 pyconfig.py:432] Config param enable_diloco: False
I0420 08:34:36.523785 139954941523776 pyconfig.py:432] Config param enable_dp_attention: False
I0420 08:34:36.523801 139954941523776 pyconfig.py:432] Config param enable_dropout: False
I0420 08:34:36.523814 139954941523776 pyconfig.py:432] Config param enable_emergency_checkpoint: False
I0420 08:34:36.523830 139954941523776 pyconfig.py:432] Config param enable_gcp_goodput_metrics: True
I0420 08:34:36.523845 139954941523776 pyconfig.py:432] Config param enable_gcp_step_deviation_metrics: True
I0420 08:34:36.523860 139954941523776 pyconfig.py:432] Config param enable_goodput_recording: False
I0420 08:34:36.523876 139954941523776 pyconfig.py:432] Config param enable_jax_profiler: False
I0420 08:34:36.523891 139954941523776 pyconfig.py:432] Config param enable_llm_inference_pool: False
I0420 08:34:36.523907 139954941523776 pyconfig.py:432] Config param enable_model_warmup: False
I0420 08:34:36.523921 139954941523776 pyconfig.py:432] Config param enable_multi_tier_checkpointing: False
I0420 08:34:36.523936 139954941523776 pyconfig.py:432] Config param enable_nnx: False
I0420 08:34:36.523952 139954941523776 pyconfig.py:432] Config param enable_orbax_v1: False
I0420 08:34:36.523968 139954941523776 pyconfig.py:432] Config param enable_padding_causal_mask: True
I0420 08:34:36.523982 139954941523776 pyconfig.py:432] Config param enable_pathways_goodput: False
I0420 08:34:36.523996 139954941523776 pyconfig.py:432] Config param enable_prefix_caching: False
I0420 08:34:36.524011 139954941523776 pyconfig.py:432] Config param enable_rampup_batch_size: False
I0420 08:34:36.524026 139954941523776 pyconfig.py:432] Config param enable_single_controller: False
I0420 08:34:36.524041 139954941523776 pyconfig.py:432] Config param enable_single_replica_ckpt_restoring: False
I0420 08:34:36.524056 139954941523776 pyconfig.py:432] Config param enable_tensorboard: True
I0420 08:34:36.524071 139954941523776 pyconfig.py:432] Config param enable_tunix_perf_metrics: False
I0420 08:34:36.524085 139954941523776 pyconfig.py:432] Config param encoder_attention_heads_for_audio: 4
I0420 08:34:36.524101 139954941523776 pyconfig.py:432] Config param encoder_ffn_dim_for_audio: 512
I0420 08:34:36.524116 139954941523776 pyconfig.py:432] Config param encoder_layers_for_audio: 2
I0420 08:34:36.524130 139954941523776 pyconfig.py:432] Config param engram: RematLocation.REMAT
I0420 08:34:36.524147 139954941523776 pyconfig.py:432] Config param engram_head_dim: 1280
I0420 08:34:36.524161 139954941523776 pyconfig.py:432] Config param engram_kernel_size: 4
I0420 08:34:36.524176 139954941523776 pyconfig.py:432] Config param engram_layers: []
I0420 08:34:36.524191 139954941523776 pyconfig.py:432] Config param engram_max_ngram_size: 3
I0420 08:34:36.524206 139954941523776 pyconfig.py:432] Config param engram_num_heads: 8
I0420 08:34:36.524222 139954941523776 pyconfig.py:432] Config param engram_seed: 0
I0420 08:34:36.524237 139954941523776 pyconfig.py:432] Config param engram_vocab_bases: []
I0420 08:34:36.524253 139954941523776 pyconfig.py:432] Config param epsilon_high: None
I0420 08:34:36.524267 139954941523776 pyconfig.py:432] Config param eval_corr_lst: False
I0420 08:34:36.524283 139954941523776 pyconfig.py:432] Config param eval_data_columns: ['text']
I0420 08:34:36.524298 139954941523776 pyconfig.py:432] Config param eval_dataset_name: c4/en:3.0.1
I0420 08:34:36.524312 139954941523776 pyconfig.py:432] Config param eval_image_column: image
I0420 08:34:36.524328 139954941523776 pyconfig.py:432] Config param eval_interval: -1
I0420 08:34:36.524357 139954941523776 pyconfig.py:432] Config param eval_make_lst: False
I0420 08:34:36.524380 139954941523776 pyconfig.py:432] Config param eval_per_device_batch_size: 2
I0420 08:34:36.524402 139954941523776 pyconfig.py:432] Config param eval_sampling_strategy: greedy
I0420 08:34:36.524424 139954941523776 pyconfig.py:432] Config param eval_split: validation
I0420 08:34:36.524446 139954941523776 pyconfig.py:432] Config param eval_steps: -1
I0420 08:34:36.524467 139954941523776 pyconfig.py:432] Config param expansion_factor_real_data: -1.0
I0420 08:34:36.524489 139954941523776 pyconfig.py:432] Config param final_logits_soft_cap: None
I0420 08:34:36.524511 139954941523776 pyconfig.py:432] Config param first_num_dense_layers: 0
I0420 08:34:36.524533 139954941523776 pyconfig.py:432] Config param float32_gate_logits: False
I0420 08:34:36.524558 139954941523776 pyconfig.py:432] Config param float32_logits: False
I0420 08:34:36.524579 139954941523776 pyconfig.py:432] Config param float32_qk_product: False
I0420 08:34:36.524601 139954941523776 pyconfig.py:432] Config param float32_weight_sum: True
I0420 08:34:36.524622 139954941523776 pyconfig.py:432] Config param force_q_layout: False
I0420 08:34:36.524643 139954941523776 pyconfig.py:432] Config param force_unroll: False
I0420 08:34:36.524658 139954941523776 pyconfig.py:432] Config param freeze_audio_encoder_params: True
I0420 08:34:36.524672 139954941523776 pyconfig.py:432] Config param freeze_vision_encoder_params: True
I0420 08:34:36.524687 139954941523776 pyconfig.py:432] Config param fused_mlp: False
I0420 08:34:36.524702 139954941523776 pyconfig.py:432] Config param fused_qkv: True
I0420 08:34:36.524717 139954941523776 pyconfig.py:432] Config param gcs_metrics: False
I0420 08:34:36.524731 139954941523776 pyconfig.py:432] Config param gdn_chunk_size: 64
I0420 08:34:36.524746 139954941523776 pyconfig.py:432] Config param gdn_conv_kernel_dim: 4
I0420 08:34:36.524761 139954941523776 pyconfig.py:432] Config param gdn_key_head_dim: 128
I0420 08:34:36.524776 139954941523776 pyconfig.py:432] Config param gdn_num_key_heads: 16
I0420 08:34:36.524793 139954941523776 pyconfig.py:432] Config param gdn_num_value_heads: 32
I0420 08:34:36.524807 139954941523776 pyconfig.py:432] Config param gdn_value_head_dim: 128
I0420 08:34:36.524822 139954941523776 pyconfig.py:432] Config param generate_padding_batch_eval: False
I0420 08:34:36.524837 139954941523776 pyconfig.py:432] Config param generate_padding_batch_train: False
I0420 08:34:36.524852 139954941523776 pyconfig.py:432] Config param generate_slice: v5e-16
I0420 08:34:36.524867 139954941523776 pyconfig.py:432] Config param generation_configs: {}
I0420 08:34:36.524882 139954941523776 pyconfig.py:432] Config param global_batch_size_to_eval_on: 64
I0420 08:34:36.524898 139954941523776 pyconfig.py:432] Config param global_batch_size_to_load: 512
I0420 08:34:36.524912 139954941523776 pyconfig.py:432] Config param global_batch_size_to_load_eval: 64
I0420 08:34:36.524927 139954941523776 pyconfig.py:432] Config param global_batch_size_to_load_increment: None
I0420 08:34:36.524941 139954941523776 pyconfig.py:432] Config param global_batch_size_to_load_start: None
I0420 08:34:36.524957 139954941523776 pyconfig.py:432] Config param global_batch_size_to_train_on: 512
I0420 08:34:36.524972 139954941523776 pyconfig.py:432] Config param global_head_dim: 0
I0420 08:34:36.524986 139954941523776 pyconfig.py:432] Config param global_num_kv_heads: 0
I0420 08:34:36.525000 139954941523776 pyconfig.py:432] Config param global_parameter_scale: 1
I0420 08:34:36.525015 139954941523776 pyconfig.py:432] Config param global_rampup_samples: 500
I0420 08:34:36.525031 139954941523776 pyconfig.py:432] Config param global_rope_max_timescale: -1
I0420 08:34:36.525046 139954941523776 pyconfig.py:432] Config param global_rope_proportion: 0.25
I0420 08:34:36.525061 139954941523776 pyconfig.py:432] Config param goodput_upload_interval_seconds: 30
I0420 08:34:36.525075 139954941523776 pyconfig.py:432] Config param grad_dtype: float32
I0420 08:34:36.525110 139954941523776 pyconfig.py:432] Config param gradient_accumulation_steps: 8
I0420 08:34:36.525126 139954941523776 pyconfig.py:432] Config param gradient_clipping_threshold: 1.0
I0420 08:34:36.525141 139954941523776 pyconfig.py:432] Config param grain_data_source_max_workers: 16
I0420 08:34:36.525156 139954941523776 pyconfig.py:432] Config param grain_eval_files: 
I0420 08:34:36.525171 139954941523776 pyconfig.py:432] Config param grain_file_type: arrayrecord
I0420 08:34:36.525186 139954941523776 pyconfig.py:432] Config param grain_num_threads: 16
I0420 08:34:36.525200 139954941523776 pyconfig.py:432] Config param grain_num_threads_eval: 16
I0420 08:34:36.525216 139954941523776 pyconfig.py:432] Config param grain_packing_type: first_fit
I0420 08:34:36.525230 139954941523776 pyconfig.py:432] Config param grain_per_worker_buffer_size: 1
I0420 08:34:36.525245 139954941523776 pyconfig.py:432] Config param grain_per_worker_buffer_size_eval: 1
I0420 08:34:36.525261 139954941523776 pyconfig.py:432] Config param grain_prefetch_buffer_size: 500
I0420 08:34:36.525276 139954941523776 pyconfig.py:432] Config param grain_prefetch_buffer_size_eval: 500
I0420 08:34:36.525291 139954941523776 pyconfig.py:432] Config param grain_ram_budget_mb: 1024
I0420 08:34:36.525307 139954941523776 pyconfig.py:432] Config param grain_shuffle_buffer_size: 100
I0420 08:34:36.525322 139954941523776 pyconfig.py:432] Config param grain_train_files: 
I0420 08:34:36.525363 139954941523776 pyconfig.py:432] Config param grain_train_mixture_config_path: 
I0420 08:34:36.525388 139954941523776 pyconfig.py:432] Config param grain_worker_count: 1
I0420 08:34:36.525411 139954941523776 pyconfig.py:432] Config param grain_worker_count_eval: 1
I0420 08:34:36.525433 139954941523776 pyconfig.py:432] Config param grpo_beta: 0.08
I0420 08:34:36.525454 139954941523776 pyconfig.py:432] Config param grpo_epsilon: 0.2
I0420 08:34:36.525470 139954941523776 pyconfig.py:432] Config param hardware: tpu
I0420 08:34:36.525485 139954941523776 pyconfig.py:432] Config param hbm_utilization_vllm: 0.72
I0420 08:34:36.525500 139954941523776 pyconfig.py:432] Config param head_dim: 8
I0420 08:34:36.525516 139954941523776 pyconfig.py:432] Config param heartbeat_reporting_interval_in_seconds: 5
I0420 08:34:36.525530 139954941523776 pyconfig.py:432] Config param hf_data_dir: None
I0420 08:34:36.525548 139954941523776 pyconfig.py:432] Config param hf_eval_files: None
I0420 08:34:36.525564 139954941523776 pyconfig.py:432] Config param hf_eval_split: None
I0420 08:34:36.525578 139954941523776 pyconfig.py:432] Config param hf_name: None
I0420 08:34:36.525593 139954941523776 pyconfig.py:432] Config param hf_path: OptimalScale/ClimbMix
I0420 08:34:36.525608 139954941523776 pyconfig.py:432] Config param hf_train_files: None
I0420 08:34:36.525624 139954941523776 pyconfig.py:432] Config param hidden_size_for_vit: 1408
I0420 08:34:36.525638 139954941523776 pyconfig.py:432] Config param hide_profiler_step_metric: False
I0420 08:34:36.525653 139954941523776 pyconfig.py:432] Config param ici_autoregressive_parallelism: 1
I0420 08:34:36.525667 139954941523776 pyconfig.py:432] Config param ici_context_autoregressive_parallelism: 1
I0420 08:34:36.525683 139954941523776 pyconfig.py:432] Config param ici_context_parallelism: 1
I0420 08:34:36.525697 139954941523776 pyconfig.py:432] Config param ici_data_parallelism: 1
I0420 08:34:36.525712 139954941523776 pyconfig.py:432] Config param ici_diloco_parallelism: 1
I0420 08:34:36.525727 139954941523776 pyconfig.py:432] Config param ici_expert_parallelism: 1
I0420 08:34:36.525742 139954941523776 pyconfig.py:432] Config param ici_fsdp_parallelism: -1
I0420 08:34:36.525757 139954941523776 pyconfig.py:432] Config param ici_fsdp_transpose_parallelism: 1
I0420 08:34:36.525771 139954941523776 pyconfig.py:432] Config param ici_parallelism: [1, 1, 1, -1, 1, 1, 1, 1, 1, 1, 1, 1, 1]
I0420 08:34:36.525788 139954941523776 pyconfig.py:432] Config param ici_pipeline_parallelism: 1
I0420 08:34:36.525801 139954941523776 pyconfig.py:432] Config param ici_sequence_parallelism: 1
I0420 08:34:36.525817 139954941523776 pyconfig.py:432] Config param ici_tensor_parallelism: 1
I0420 08:34:36.525831 139954941523776 pyconfig.py:432] Config param ici_tensor_sequence_parallelism: 1
I0420 08:34:36.525846 139954941523776 pyconfig.py:432] Config param ici_tensor_transpose_parallelism: 1
I0420 08:34:36.525862 139954941523776 pyconfig.py:432] Config param image_path: 
I0420 08:34:36.525876 139954941523776 pyconfig.py:432] Config param image_placeholder: <|image|>
I0420 08:34:36.525891 139954941523776 pyconfig.py:432] Config param image_size_for_vit: 896
I0420 08:34:36.525905 139954941523776 pyconfig.py:432] Config param indexer_head_dim: 128
I0420 08:34:36.525921 139954941523776 pyconfig.py:432] Config param indexer_loss_scaling_factor: 0.0
I0420 08:34:36.525935 139954941523776 pyconfig.py:432] Config param indexer_n_heads: 64
I0420 08:34:36.525950 139954941523776 pyconfig.py:432] Config param indexer_sparse_training: False
I0420 08:34:36.525964 139954941523776 pyconfig.py:432] Config param indexer_topk: 2048
I0420 08:34:36.525979 139954941523776 pyconfig.py:432] Config param inference_benchmark_test: False
I0420 08:34:36.525994 139954941523776 pyconfig.py:432] Config param inference_metadata_file: 
I0420 08:34:36.526008 139954941523776 pyconfig.py:432] Config param inference_microbenchmark_log_file_path: 
I0420 08:34:36.526024 139954941523776 pyconfig.py:432] Config param inference_microbenchmark_loop_iters: 10
I0420 08:34:36.526039 139954941523776 pyconfig.py:432] Config param inference_microbenchmark_num_samples: [1, 2, 3, 4, 5]
I0420 08:34:36.526054 139954941523776 pyconfig.py:432] Config param inference_microbenchmark_prefill_lengths: 64,128,256,512,1024
I0420 08:34:36.526069 139954941523776 pyconfig.py:432] Config param inference_microbenchmark_stages: prefill,generate
I0420 08:34:36.526085 139954941523776 pyconfig.py:432] Config param inference_server: MaxtextInterleavedServer
I0420 08:34:36.526098 139954941523776 pyconfig.py:432] Config param inhomogeneous_layer_cycle_interval: 1
I0420 08:34:36.526114 139954941523776 pyconfig.py:432] Config param init_weights_seed: 0
I0420 08:34:36.526129 139954941523776 pyconfig.py:432] Config param input_data_sharding_logical_axes: ['activation_embed_and_logits_batch', 'activation_norm_length']
I0420 08:34:36.526144 139954941523776 pyconfig.py:432] Config param interleave_moe_layer_step: 1
I0420 08:34:36.526159 139954941523776 pyconfig.py:432] Config param intermediate_size_for_vit: 5632
I0420 08:34:36.526174 139954941523776 pyconfig.py:432] Config param internal_compile: False
I0420 08:34:36.526188 139954941523776 pyconfig.py:432] Config param internal_compile_num_devices: -1
I0420 08:34:36.526203 139954941523776 pyconfig.py:432] Config param jax_cache_dir: ~/jax_cache
I0420 08:34:36.526219 139954941523776 pyconfig.py:432] Config param jax_debug_log_modules: 
I0420 08:34:36.526234 139954941523776 pyconfig.py:432] Config param jax_distributed_initialization_timeout: 300
I0420 08:34:36.526249 139954941523776 pyconfig.py:432] Config param jax_profiler_port: 9999
I0420 08:34:36.526263 139954941523776 pyconfig.py:432] Config param key_proj: RematLocation.REMAT
I0420 08:34:36.526280 139954941523776 pyconfig.py:432] Config param kv_cache_buffer: 256
I0420 08:34:36.526294 139954941523776 pyconfig.py:432] Config param kv_lora_rank: 512
I0420 08:34:36.526310 139954941523776 pyconfig.py:432] Config param kv_quant_axis: KvQuantAxis.HEADS_AND_DKV
I0420 08:34:36.526327 139954941523776 pyconfig.py:432] Config param kv_quant_dtype: int8
I0420 08:34:36.526357 139954941523776 pyconfig.py:432] Config param kv_wa_proj: RematLocation.REMAT
I0420 08:34:36.526381 139954941523776 pyconfig.py:432] Config param learning_rate: 0.0002
I0420 08:34:36.526404 139954941523776 pyconfig.py:432] Config param learning_rate_final_fraction: 0.1
I0420 08:34:36.526427 139954941523776 pyconfig.py:432] Config param learning_rate_schedule_steps: 200000
I0420 08:34:36.526449 139954941523776 pyconfig.py:432] Config param load_balance_loss_weight: 0.0
I0420 08:34:36.526469 139954941523776 pyconfig.py:432] Config param load_checkpoint_only_once: False
I0420 08:34:36.526484 139954941523776 pyconfig.py:432] Config param load_from_prefill_dir: False
I0420 08:34:36.526499 139954941523776 pyconfig.py:432] Config param load_full_state_path: 
I0420 08:34:36.526513 139954941523776 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
I0420 08:34:36.526529 139954941523776 pyconfig.py:432] Config param local_checkpoint_directory: 
I0420 08:34:36.526547 139954941523776 pyconfig.py:432] Config param local_checkpoint_period: 0
I0420 08:34:36.526562 139954941523776 pyconfig.py:432] Config param local_rope_max_timescale: -1
I0420 08:34:36.526576 139954941523776 pyconfig.py:432] Config param local_rope_proportion: 1.0
I0420 08:34:36.526592 139954941523776 pyconfig.py:432] Config param log_config: True
I0420 08:34:36.526606 139954941523776 pyconfig.py:432] Config param log_period: 10
I0420 08:34:36.526621 139954941523776 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'))
I0420 08:34:36.526699 139954941523776 pyconfig.py:432] Config param logits_dot_in_fp32: False
I0420 08:34:36.526716 139954941523776 pyconfig.py:432] Config param logits_via_embedding: True
I0420 08:34:36.526732 139954941523776 pyconfig.py:432] Config param lora_input_adapters_path: 
I0420 08:34:36.526747 139954941523776 pyconfig.py:432] Config param loss_algo: grpo
I0420 08:34:36.526763 139954941523776 pyconfig.py:432] Config param lr_schedule_type: LearningRateScheduleType.COSINE
I0420 08:34:36.526781 139954941523776 pyconfig.py:432] Config param managed_mldiagnostics: False
I0420 08:34:36.526795 139954941523776 pyconfig.py:432] Config param managed_mldiagnostics_dir: None
I0420 08:34:36.526809 139954941523776 pyconfig.py:432] Config param managed_mldiagnostics_run_group: 
I0420 08:34:36.526824 139954941523776 pyconfig.py:432] Config param matmul_precision: MatmulPrecision.DEFAULT
I0420 08:34:36.526842 139954941523776 pyconfig.py:432] Config param max_checkify: False
I0420 08:34:36.526856 139954941523776 pyconfig.py:432] Config param max_concurrency: 256
I0420 08:34:36.526871 139954941523776 pyconfig.py:432] Config param max_corpus_chars: 10000000
I0420 08:34:36.526886 139954941523776 pyconfig.py:432] Config param max_num_batched_tokens: None
I0420 08:34:36.526900 139954941523776 pyconfig.py:432] Config param max_num_checkpoints_to_keep: None
I0420 08:34:36.526914 139954941523776 pyconfig.py:432] Config param max_num_images_per_example: -1
I0420 08:34:36.526930 139954941523776 pyconfig.py:432] Config param max_num_seqs: None
I0420 08:34:36.526945 139954941523776 pyconfig.py:432] Config param max_position_embeddings: 163840
I0420 08:34:36.526960 139954941523776 pyconfig.py:432] Config param max_prefill_predict_length: 64
I0420 08:34:36.526974 139954941523776 pyconfig.py:432] Config param max_sample_len_for_audio: 10000
I0420 08:34:36.526990 139954941523776 pyconfig.py:432] Config param max_segments_per_seq: -1
I0420 08:34:36.527004 139954941523776 pyconfig.py:432] Config param max_source_positions_for_audio: 1500
I0420 08:34:36.527019 139954941523776 pyconfig.py:432] Config param max_target_length: 2048
I0420 08:34:36.527035 139954941523776 pyconfig.py:432] Config param max_timescale_for_audio: 10000.0
I0420 08:34:36.527051 139954941523776 pyconfig.py:432] Config param megablox: True
I0420 08:34:36.527066 139954941523776 pyconfig.py:432] Config param merge_gating_gmm: False
I0420 08:34:36.527080 139954941523776 pyconfig.py:432] Config param mesh_axes: ['diloco', 'data', 'stage', 'fsdp', 'fsdp_transpose', 'sequence', 'context', 'context_autoregressive', 'tensor', 'tensor_transpose', 'tensor_sequence', 'expert', 'autoregressive']
I0420 08:34:36.527098 139954941523776 pyconfig.py:432] Config param metrics_dir: None
I0420 08:34:36.527113 139954941523776 pyconfig.py:432] Config param metrics_file: 
I0420 08:34:36.527129 139954941523776 pyconfig.py:432] Config param mhc_expansion_rate: 1
I0420 08:34:36.527143 139954941523776 pyconfig.py:432] Config param micro_batch_size_to_eval_on: 64
I0420 08:34:36.527158 139954941523776 pyconfig.py:432] Config param micro_batch_size_to_train_on: 64
I0420 08:34:36.527171 139954941523776 pyconfig.py:432] Config param mla_kv: RematLocation.REMAT
I0420 08:34:36.527187 139954941523776 pyconfig.py:432] Config param mla_naive_kvcache: True
I0420 08:34:36.527201 139954941523776 pyconfig.py:432] Config param mla_q: RematLocation.REMAT
I0420 08:34:36.527217 139954941523776 pyconfig.py:432] Config param mlp_activations: ['gelu']
I0420 08:34:36.527231 139954941523776 pyconfig.py:432] Config param mlp_activations_limit: -1.0
I0420 08:34:36.527245 139954941523776 pyconfig.py:432] Config param mlp_bias: False
I0420 08:34:36.527261 139954941523776 pyconfig.py:432] Config param mlp_dim: 64
I0420 08:34:36.527275 139954941523776 pyconfig.py:432] Config param mlpwi: RematLocation.REMAT
I0420 08:34:36.527291 139954941523776 pyconfig.py:432] Config param mlpwi_0: RematLocation.REMAT
I0420 08:34:36.527305 139954941523776 pyconfig.py:432] Config param mlpwi_1: RematLocation.REMAT
I0420 08:34:36.527321 139954941523776 pyconfig.py:432] Config param mlpwo: RematLocation.REMAT
I0420 08:34:36.527346 139954941523776 pyconfig.py:432] Config param moba: False
I0420 08:34:36.527369 139954941523776 pyconfig.py:432] Config param moba_chunk_size: 1024
I0420 08:34:36.527385 139954941523776 pyconfig.py:432] Config param moba_topk: 8
I0420 08:34:36.527400 139954941523776 pyconfig.py:432] Config param model_call_mode: 
I0420 08:34:36.527415 139954941523776 pyconfig.py:432] Config param model_name: gpt3-52k
I0420 08:34:36.527430 139954941523776 pyconfig.py:432] Config param moe_expert_input_dim: -1
I0420 08:34:36.527443 139954941523776 pyconfig.py:432] Config param moe_fsdp_use_two_stage_all_gather: False
I0420 08:34:36.527459 139954941523776 pyconfig.py:432] Config param moe_mlp_dim: -1
I0420 08:34:36.527473 139954941523776 pyconfig.py:432] Config param moe_mlpwi_0: RematLocation.REMAT
I0420 08:34:36.527488 139954941523776 pyconfig.py:432] Config param moe_mlpwi_1: RematLocation.REMAT
I0420 08:34:36.527503 139954941523776 pyconfig.py:432] Config param moe_mlpwo: RematLocation.REMAT
I0420 08:34:36.527518 139954941523776 pyconfig.py:432] Config param monitor_goodput: False
I0420 08:34:36.527532 139954941523776 pyconfig.py:432] Config param monitor_step_time_deviation: True
I0420 08:34:36.527551 139954941523776 pyconfig.py:432] Config param mrope_section: [24, 20, 20]
I0420 08:34:36.527567 139954941523776 pyconfig.py:432] Config param mscale: 1.0
I0420 08:34:36.527581 139954941523776 pyconfig.py:432] Config param mtc_data_parallelism: 0
I0420 08:34:36.527596 139954941523776 pyconfig.py:432] Config param mtp_eval_target_module: 0
I0420 08:34:36.527611 139954941523776 pyconfig.py:432] Config param mtp_loss_scaling_factor: 0.1
I0420 08:34:36.527626 139954941523776 pyconfig.py:432] Config param mtp_num_layers: 0
I0420 08:34:36.527640 139954941523776 pyconfig.py:432] Config param mu_dtype: float32
I0420 08:34:36.527665 139954941523776 pyconfig.py:432] Config param multi_sampling: False
I0420 08:34:36.527679 139954941523776 pyconfig.py:432] Config param multi_tier_checkpointing_backup_interval_minutes: 0
I0420 08:34:36.527695 139954941523776 pyconfig.py:432] Config param muon_beta: 0.95
I0420 08:34:36.527710 139954941523776 pyconfig.py:432] Config param muon_consistent_rms: None
I0420 08:34:36.527724 139954941523776 pyconfig.py:432] Config param muon_weight_decay: 0.0
I0420 08:34:36.527740 139954941523776 pyconfig.py:432] Config param n_routing_groups: -1
I0420 08:34:36.527755 139954941523776 pyconfig.py:432] Config param n_window_for_audio: 50
I0420 08:34:36.527769 139954941523776 pyconfig.py:432] Config param n_window_infer_for_audio: 800
I0420 08:34:36.527785 139954941523776 pyconfig.py:432] Config param nope_layer_interval: -1
I0420 08:34:36.527800 139954941523776 pyconfig.py:432] Config param norm_topk_prob: False
I0420 08:34:36.527814 139954941523776 pyconfig.py:432] Config param normalization_layer_epsilon: 1e-05
I0420 08:34:36.527832 139954941523776 pyconfig.py:432] Config param normalize_embedding_logits: False
I0420 08:34:36.527846 139954941523776 pyconfig.py:432] Config param num_attention_heads_for_vit: 16
I0420 08:34:36.527862 139954941523776 pyconfig.py:432] Config param num_batches: 4
I0420 08:34:36.527877 139954941523776 pyconfig.py:432] Config param num_channels_for_vit: 3
I0420 08:34:36.527891 139954941523776 pyconfig.py:432] Config param num_conv_layers_for_audio: 3
I0420 08:34:36.527906 139954941523776 pyconfig.py:432] Config param num_decoder_layers: 1
I0420 08:34:36.527920 139954941523776 pyconfig.py:432] Config param num_diloco_replicas: 1
I0420 08:34:36.527935 139954941523776 pyconfig.py:432] Config param num_epoch: 1
I0420 08:34:36.527950 139954941523776 pyconfig.py:432] Config param num_eval_passes: 1
I0420 08:34:36.527964 139954941523776 pyconfig.py:432] Config param num_experts: 1
I0420 08:34:36.527980 139954941523776 pyconfig.py:432] Config param num_experts_per_tok: 1
I0420 08:34:36.527995 139954941523776 pyconfig.py:432] Config param num_generations: 2
I0420 08:34:36.528010 139954941523776 pyconfig.py:432] Config param num_hidden_layers_for_vit: 34
I0420 08:34:36.528024 139954941523776 pyconfig.py:432] Config param num_iterations: 1
I0420 08:34:36.528039 139954941523776 pyconfig.py:432] Config param num_kv_heads: 2
I0420 08:34:36.528054 139954941523776 pyconfig.py:432] Config param num_layers_per_pipeline_stage: 1
I0420 08:34:36.528069 139954941523776 pyconfig.py:432] Config param num_mel_bins_for_audio: 128
I0420 08:34:36.528083 139954941523776 pyconfig.py:432] Config param num_pipeline_microbatches: -1
I0420 08:34:36.528097 139954941523776 pyconfig.py:432] Config param num_pipeline_repeats: -1
I0420 08:34:36.528112 139954941523776 pyconfig.py:432] Config param num_position_embeddings_for_vit: 1024
I0420 08:34:36.528127 139954941523776 pyconfig.py:432] Config param num_query_heads: 2
I0420 08:34:36.528142 139954941523776 pyconfig.py:432] Config param num_samplers_slices: -1
I0420 08:34:36.528157 139954941523776 pyconfig.py:432] Config param num_slices: 1
I0420 08:34:36.528171 139954941523776 pyconfig.py:432] Config param num_target_devices: 32
I0420 08:34:36.528186 139954941523776 pyconfig.py:432] Config param num_test_batches: 5
I0420 08:34:36.528202 139954941523776 pyconfig.py:432] Config param num_trainer_slices: -1
I0420 08:34:36.528216 139954941523776 pyconfig.py:432] Config param num_vocab_tiling: 1
I0420 08:34:36.528232 139954941523776 pyconfig.py:432] Config param off_policy_steps: 0
I0420 08:34:36.528246 139954941523776 pyconfig.py:432] Config param offline_data_dir: None
I0420 08:34:36.528261 139954941523776 pyconfig.py:432] Config param opt_type: OptimizerType.ADAM_PAX
I0420 08:34:36.528278 139954941523776 pyconfig.py:432] Config param optimize_mesh_for_tpu_v6e: False
I0420 08:34:36.528292 139954941523776 pyconfig.py:432] Config param optimizer_memory_host_offload: False
I0420 08:34:36.528308 139954941523776 pyconfig.py:432] Config param original_max_position_embeddings: 4096
I0420 08:34:36.528323 139954941523776 pyconfig.py:432] Config param out_hidden_size_for_vit: 512
I0420 08:34:36.528348 139954941523776 pyconfig.py:432] Config param out_proj: RematLocation.REMAT
I0420 08:34:36.528373 139954941523776 pyconfig.py:432] Config param output_dim_for_audio: 512
I0420 08:34:36.528394 139954941523776 pyconfig.py:432] Config param override_logical_axis_rules: False
I0420 08:34:36.528416 139954941523776 pyconfig.py:432] Config param override_model_config: True
I0420 08:34:36.528438 139954941523776 pyconfig.py:432] Config param packing: True
I0420 08:34:36.528461 139954941523776 pyconfig.py:432] Config param pagedattn_head_dim_alignment: 128
I0420 08:34:36.528485 139954941523776 pyconfig.py:432] Config param pagedattn_max_pages_per_group: -1
I0420 08:34:36.528507 139954941523776 pyconfig.py:432] Config param pagedattn_num_pages: 64
I0420 08:34:36.528530 139954941523776 pyconfig.py:432] Config param pagedattn_pages_per_compute_block: 4
I0420 08:34:36.528556 139954941523776 pyconfig.py:432] Config param pagedattn_tokens_per_page: 32
I0420 08:34:36.528577 139954941523776 pyconfig.py:432] Config param param_scan_axis: 1
I0420 08:34:36.528593 139954941523776 pyconfig.py:432] Config param parameter_memory_host_offload: False
I0420 08:34:36.528609 139954941523776 pyconfig.py:432] Config param partial_rotary_factor: 1.0
I0420 08:34:36.528625 139954941523776 pyconfig.py:432] Config param patch_size_for_vit: 14
I0420 08:34:36.528640 139954941523776 pyconfig.py:432] Config param penalty_incorrect_answer: -1.0
I0420 08:34:36.528655 139954941523776 pyconfig.py:432] Config param penalty_incorrect_format: -0.5
I0420 08:34:36.528670 139954941523776 pyconfig.py:432] Config param per_device_batch_size: 2
I0420 08:34:36.528686 139954941523776 pyconfig.py:432] Config param per_device_batch_size_increment: 2.0
I0420 08:34:36.528700 139954941523776 pyconfig.py:432] Config param per_device_batch_size_start: 4.0
I0420 08:34:36.528716 139954941523776 pyconfig.py:432] Config param pipeline_delay_activation_forwarding: False
I0420 08:34:36.528730 139954941523776 pyconfig.py:432] Config param pipeline_fsdp_ag_once: False
I0420 08:34:36.528745 139954941523776 pyconfig.py:432] Config param pipeline_fsdp_ag_per_repeat: False
I0420 08:34:36.528759 139954941523776 pyconfig.py:432] Config param pipeline_parallel_layers: 1
I0420 08:34:36.528774 139954941523776 pyconfig.py:432] Config param pixel_shuffle_ratio_for_vit: 0.5
I0420 08:34:36.528789 139954941523776 pyconfig.py:432] Config param posemb_type_for_vit: learn
I0420 08:34:36.528804 139954941523776 pyconfig.py:432] Config param position_id_per_seconds: 25
I0420 08:34:36.528818 139954941523776 pyconfig.py:432] Config param prefill_cache_axis_order: 1,2,0,3
I0420 08:34:36.528833 139954941523776 pyconfig.py:432] Config param prefill_cache_dir: 
I0420 08:34:36.528847 139954941523776 pyconfig.py:432] Config param prefill_chunk_size: 256
I0420 08:34:36.528862 139954941523776 pyconfig.py:432] Config param prefill_slice: v5e-16
I0420 08:34:36.528876 139954941523776 pyconfig.py:432] Config param prefix_caching_dram_byte: 100000000000
I0420 08:34:36.528891 139954941523776 pyconfig.py:432] Config param prefix_caching_hbm_byte: 10000000000
I0420 08:34:36.528905 139954941523776 pyconfig.py:432] Config param profile_cleanly: True
I0420 08:34:36.528921 139954941523776 pyconfig.py:432] Config param profile_periodically_period: -1
I0420 08:34:36.528935 139954941523776 pyconfig.py:432] Config param profile_power_events: False
I0420 08:34:36.528951 139954941523776 pyconfig.py:432] Config param profiler: ProfilerType.NONE
I0420 08:34:36.528968 139954941523776 pyconfig.py:432] Config param profiler_steps: 5
I0420 08:34:36.528983 139954941523776 pyconfig.py:432] Config param projector_dropout_for_vit: 0.0
I0420 08:34:36.528998 139954941523776 pyconfig.py:432] Config param projector_input_dim_for_vit: 4096
I0420 08:34:36.529012 139954941523776 pyconfig.py:432] Config param projector_output_dim_for_vit: 4096
I0420 08:34:36.529028 139954941523776 pyconfig.py:432] Config param prometheus_port: 0
I0420 08:34:36.529043 139954941523776 pyconfig.py:432] Config param prompt: I love to
I0420 08:34:36.529058 139954941523776 pyconfig.py:432] Config param pure_nnx: False
I0420 08:34:36.529072 139954941523776 pyconfig.py:432] Config param pure_nnx_decoder: False
I0420 08:34:36.529088 139954941523776 pyconfig.py:432] Config param q_lora_rank: 0
I0420 08:34:36.529102 139954941523776 pyconfig.py:432] Config param qk_clip_threshold: 100.0
I0420 08:34:36.529117 139954941523776 pyconfig.py:432] Config param qk_nope_head_dim: 128
I0420 08:34:36.529131 139954941523776 pyconfig.py:432] Config param qk_norm_with_scale: True
I0420 08:34:36.529146 139954941523776 pyconfig.py:432] Config param qk_rope_head_dim: 64
I0420 08:34:36.529163 139954941523776 pyconfig.py:432] Config param qkv_proj: RematLocation.REMAT
I0420 08:34:36.529177 139954941523776 pyconfig.py:432] Config param quant_cfg_path: 
I0420 08:34:36.529192 139954941523776 pyconfig.py:432] Config param quantization: QuantizationType.NONE
I0420 08:34:36.529209 139954941523776 pyconfig.py:432] Config param quantization_local_shard_count: 4
I0420 08:34:36.529225 139954941523776 pyconfig.py:432] Config param quantize_kvcache: False
I0420 08:34:36.529240 139954941523776 pyconfig.py:432] Config param query_proj: RematLocation.REMAT
I0420 08:34:36.529255 139954941523776 pyconfig.py:432] Config param query_wa_proj: RematLocation.REMAT
I0420 08:34:36.529269 139954941523776 pyconfig.py:432] Config param ragged_block_size: 256
I0420 08:34:36.529284 139954941523776 pyconfig.py:432] Config param ragged_buffer_factor: -1.0
I0420 08:34:36.529300 139954941523776 pyconfig.py:432] Config param rampup_end_step: 0
I0420 08:34:36.529316 139954941523776 pyconfig.py:432] Config param rampup_samples_per_increment_to_load: None
I0420 08:34:36.529331 139954941523776 pyconfig.py:432] Config param reasoning_end_token: </reasoning>
I0420 08:34:36.529371 139954941523776 pyconfig.py:432] Config param reasoning_start_token: <reasoning>
I0420 08:34:36.529393 139954941523776 pyconfig.py:432] Config param record_internal_nn_metrics: 0
I0420 08:34:36.529415 139954941523776 pyconfig.py:432] Config param remat_policy: full
I0420 08:34:36.529437 139954941523776 pyconfig.py:432] Config param remat_policy_for_vit: minimal
I0420 08:34:36.529458 139954941523776 pyconfig.py:432] Config param remove_size_one_mesh_axis_from_type: True
I0420 08:34:36.529478 139954941523776 pyconfig.py:432] Config param replicate_quant_scale: False
I0420 08:34:36.529493 139954941523776 pyconfig.py:432] Config param replicator_backup_interval_minutes: 0
I0420 08:34:36.529509 139954941523776 pyconfig.py:432] Config param report_heartbeat_metric_for_gcp_monitoring: False
I0420 08:34:36.529524 139954941523776 pyconfig.py:432] Config param report_performance_metric_for_gcp_monitoring: False
I0420 08:34:36.529543 139954941523776 pyconfig.py:432] Config param reshape_q: False
I0420 08:34:36.529559 139954941523776 pyconfig.py:432] Config param return_log_prob: False
I0420 08:34:36.529573 139954941523776 pyconfig.py:432] Config param reuse_example_batch: 0
I0420 08:34:36.529588 139954941523776 pyconfig.py:432] Config param reward_exact_answer: 5.0
I0420 08:34:36.529603 139954941523776 pyconfig.py:432] Config param reward_exact_format_match: 3.0
I0420 08:34:36.529618 139954941523776 pyconfig.py:432] Config param reward_partial_format_match: 0.5
I0420 08:34:36.529633 139954941523776 pyconfig.py:432] Config param reward_ratio_guess_to_answer_high: 0.5
I0420 08:34:36.529649 139954941523776 pyconfig.py:432] Config param reward_ratio_guess_to_answer_low: 0.25
I0420 08:34:36.529663 139954941523776 pyconfig.py:432] Config param reward_white_space_format_match: 1.5
I0420 08:34:36.529679 139954941523776 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}
I0420 08:34:36.529700 139954941523776 pyconfig.py:432] Config param rollout_data_parallelism: -1
I0420 08:34:36.529715 139954941523776 pyconfig.py:432] Config param rollout_expert_parallelism: 1
I0420 08:34:36.529730 139954941523776 pyconfig.py:432] Config param rollout_micro_batch_size: -1
I0420 08:34:36.529745 139954941523776 pyconfig.py:432] Config param rollout_tensor_parallelism: -1
I0420 08:34:36.529758 139954941523776 pyconfig.py:432] Config param rope_attention_scaling: False
I0420 08:34:36.529774 139954941523776 pyconfig.py:432] Config param rope_factor: 40
I0420 08:34:36.529788 139954941523776 pyconfig.py:432] Config param rope_interleave: True
I0420 08:34:36.529804 139954941523776 pyconfig.py:432] Config param rope_linear_scaling_factor: 1.0
I0420 08:34:36.529818 139954941523776 pyconfig.py:432] Config param rope_max_timescale: 10000
I0420 08:34:36.529833 139954941523776 pyconfig.py:432] Config param rope_min_timescale: 1
I0420 08:34:36.529848 139954941523776 pyconfig.py:432] Config param rope_theta_for_vit: 10000
I0420 08:34:36.529862 139954941523776 pyconfig.py:432] Config param rope_truncate: True
I0420 08:34:36.529877 139954941523776 pyconfig.py:432] Config param rope_type: RopeType.DEFAULT
I0420 08:34:36.529894 139954941523776 pyconfig.py:432] Config param rope_use_scale: True
I0420 08:34:36.529910 139954941523776 pyconfig.py:432] Config param routed_bias: False
I0420 08:34:36.529925 139954941523776 pyconfig.py:432] Config param routed_bias_update_rate: 0.0
I0420 08:34:36.529939 139954941523776 pyconfig.py:432] Config param routed_scaling_factor: 1.0
I0420 08:34:36.529954 139954941523776 pyconfig.py:432] Config param routed_score_func: 
I0420 08:34:36.529970 139954941523776 pyconfig.py:432] Config param run_name: gpt3-52k_2026-04-20-08-34
I0420 08:34:36.529985 139954941523776 pyconfig.py:432] Config param sa_block_kv: 512
I0420 08:34:36.530000 139954941523776 pyconfig.py:432] Config param sa_block_kv_compute: 512
I0420 08:34:36.530015 139954941523776 pyconfig.py:432] Config param sa_block_kv_dkv: 512
I0420 08:34:36.530030 139954941523776 pyconfig.py:432] Config param sa_block_kv_dkv_compute: 512
I0420 08:34:36.530045 139954941523776 pyconfig.py:432] Config param sa_block_kv_dq: 512
I0420 08:34:36.530060 139954941523776 pyconfig.py:432] Config param sa_block_q: 512
I0420 08:34:36.530075 139954941523776 pyconfig.py:432] Config param sa_block_q_dkv: 512
I0420 08:34:36.530090 139954941523776 pyconfig.py:432] Config param sa_block_q_dq: 512
I0420 08:34:36.530104 139954941523776 pyconfig.py:432] Config param sa_k_layout: HEAD_DIM_MINOR
I0420 08:34:36.530119 139954941523776 pyconfig.py:432] Config param sa_q_layout: HEAD_DIM_MINOR
I0420 08:34:36.530134 139954941523776 pyconfig.py:432] Config param sa_use_fused_bwd_kernel: False
I0420 08:34:36.530148 139954941523776 pyconfig.py:432] Config param sa_v_layout: HEAD_DIM_MINOR
I0420 08:34:36.530164 139954941523776 pyconfig.py:432] Config param sampler_devices_fraction: 0.5
I0420 08:34:36.530179 139954941523776 pyconfig.py:432] Config param save_checkpoint_on_completion: True
I0420 08:34:36.530194 139954941523776 pyconfig.py:432] Config param save_config_to_gcs: False
I0420 08:34:36.530209 139954941523776 pyconfig.py:432] Config param save_quantized_params_path: 
I0420 08:34:36.530224 139954941523776 pyconfig.py:432] Config param scale_embedding_for_audio: True
I0420 08:34:36.530238 139954941523776 pyconfig.py:432] Config param scan_layers: True
I0420 08:34:36.530253 139954941523776 pyconfig.py:432] Config param scan_layers_per_stage: False
I0420 08:34:36.530267 139954941523776 pyconfig.py:432] Config param scan_pipeline_iterations: True
I0420 08:34:36.530282 139954941523776 pyconfig.py:432] Config param scan_pipeline_repeats: False
I0420 08:34:36.530296 139954941523776 pyconfig.py:432] Config param set_remat_policy_on_layers_per_stage: False
I0420 08:34:36.530311 139954941523776 pyconfig.py:432] Config param set_remat_policy_on_pipeline_iterations: True
I0420 08:34:36.530327 139954941523776 pyconfig.py:432] Config param sft_train_on_completion_only: False
I0420 08:34:36.530359 139954941523776 pyconfig.py:432] Config param shard_exp_on_fsdp: False
I0420 08:34:36.530382 139954941523776 pyconfig.py:432] Config param shard_mode: ShardMode.AUTO
I0420 08:34:36.530405 139954941523776 pyconfig.py:432] Config param shard_optimizer_over_data: False
I0420 08:34:36.530427 139954941523776 pyconfig.py:432] Config param sharding_strategy: None
I0420 08:34:36.530449 139954941523776 pyconfig.py:432] Config param sharding_tolerance: 0.02
I0420 08:34:36.530469 139954941523776 pyconfig.py:432] Config param shardy: True
I0420 08:34:36.530484 139954941523776 pyconfig.py:432] Config param share_kv_projections: False
I0420 08:34:36.530499 139954941523776 pyconfig.py:432] Config param shared_experts: 0
I0420 08:34:36.530514 139954941523776 pyconfig.py:432] Config param sinkhorn_iterations: 20
I0420 08:34:36.530528 139954941523776 pyconfig.py:432] Config param skip_first_n_steps_for_profiler: 1
I0420 08:34:36.530547 139954941523776 pyconfig.py:432] Config param skip_jax_distributed_system: False
I0420 08:34:36.530562 139954941523776 pyconfig.py:432] Config param skip_step_interval: 128
I0420 08:34:36.530576 139954941523776 pyconfig.py:432] Config param skip_step_on_spikes: False
I0420 08:34:36.530591 139954941523776 pyconfig.py:432] Config param skip_step_scaling_factor: 6.0
I0420 08:34:36.530607 139954941523776 pyconfig.py:432] Config param sliding_window_size: 0
I0420 08:34:36.530621 139954941523776 pyconfig.py:432] Config param solution_end_token: </answer>
I0420 08:34:36.530637 139954941523776 pyconfig.py:432] Config param solution_start_token: <answer>
I0420 08:34:36.530651 139954941523776 pyconfig.py:432] Config param source_checkpoint_layout: orbax
I0420 08:34:36.530666 139954941523776 pyconfig.py:432] Config param sparse_matmul: True
I0420 08:34:36.530682 139954941523776 pyconfig.py:432] Config param spatial_merge_size_for_vit: 2
I0420 08:34:36.530697 139954941523776 pyconfig.py:432] Config param stack_prefill_result_cache: False
I0420 08:34:36.530713 139954941523776 pyconfig.py:432] Config param stack_trace_interval_seconds: 600
I0420 08:34:36.530727 139954941523776 pyconfig.py:432] Config param stack_trace_to_cloud: False
I0420 08:34:36.530742 139954941523776 pyconfig.py:432] Config param step_deviation_interval_seconds: 30
I0420 08:34:36.530756 139954941523776 pyconfig.py:432] Config param steps: 200000
I0420 08:34:36.530771 139954941523776 pyconfig.py:432] Config param stop_strings: None
I0420 08:34:36.530786 139954941523776 pyconfig.py:432] Config param student_overrides: {'model_name': 'llama3.1-8b'}
I0420 08:34:36.530803 139954941523776 pyconfig.py:432] Config param student_params_to_update: None
I0420 08:34:36.530819 139954941523776 pyconfig.py:432] Config param subslice_shape: 
I0420 08:34:36.530833 139954941523776 pyconfig.py:432] Config param swap_space_vllm_gb: 2
I0420 08:34:36.530849 139954941523776 pyconfig.py:432] Config param system_prompt: 
I0420 08:34:36.530863 139954941523776 pyconfig.py:432] Config param target_eval_loss: 0.0
I0420 08:34:36.530878 139954941523776 pyconfig.py:432] Config param teacher_overrides: {'model_name': 'llama3.1-8b'}
I0420 08:34:36.530893 139954941523776 pyconfig.py:432] Config param temperature_tuning: False
I0420 08:34:36.530908 139954941523776 pyconfig.py:432] Config param temporal_patch_size_for_vit: 2
I0420 08:34:36.530922 139954941523776 pyconfig.py:432] Config param tensorboard_dir: None
I0420 08:34:36.530937 139954941523776 pyconfig.py:432] Config param tensors_on_device: None
I0420 08:34:36.530951 139954941523776 pyconfig.py:432] Config param tensors_to_offload: None
I0420 08:34:36.530966 139954941523776 pyconfig.py:432] Config param test_batch_start_index: 0
I0420 08:34:36.530979 139954941523776 pyconfig.py:432] Config param tile_size_for_vit: 336
I0420 08:34:36.530995 139954941523776 pyconfig.py:432] Config param tokenize_eval_data: True
I0420 08:34:36.531008 139954941523776 pyconfig.py:432] Config param tokenize_train_data: True
I0420 08:34:36.531024 139954941523776 pyconfig.py:432] Config param tokenizer_path: meta-llama/Llama-3.1-8B
I0420 08:34:36.531038 139954941523776 pyconfig.py:432] Config param tokenizer_type: TokenizerType.HUGGINGFACE
I0420 08:34:36.531055 139954941523776 pyconfig.py:432] Config param topk_routing_group: -1
I0420 08:34:36.531069 139954941523776 pyconfig.py:432] Config param train_data_columns: ['text']
I0420 08:34:36.531084 139954941523776 pyconfig.py:432] Config param train_fraction: 1.0
I0420 08:34:36.531100 139954941523776 pyconfig.py:432] Config param train_image_column: image
I0420 08:34:36.531114 139954941523776 pyconfig.py:432] Config param train_micro_batch_size: -1
I0420 08:34:36.531129 139954941523776 pyconfig.py:432] Config param train_split: train
I0420 08:34:36.531143 139954941523776 pyconfig.py:432] Config param trainable_parameters_mask: []
I0420 08:34:36.531158 139954941523776 pyconfig.py:432] Config param trainable_position_size: 2048
I0420 08:34:36.531172 139954941523776 pyconfig.py:432] Config param trainer_devices_fraction: 0.5
I0420 08:34:36.531188 139954941523776 pyconfig.py:432] Config param upload_all_profiler_results: False
I0420 08:34:36.531202 139954941523776 pyconfig.py:432] Config param use_2d_fsdp_sharding: False
I0420 08:34:36.531217 139954941523776 pyconfig.py:432] Config param use_agentic_rollout: False
I0420 08:34:36.531231 139954941523776 pyconfig.py:432] Config param use_audio: False
I0420 08:34:36.531246 139954941523776 pyconfig.py:432] Config param use_audio_in_video: False
I0420 08:34:36.531260 139954941523776 pyconfig.py:432] Config param use_batch_split_schedule: False
I0420 08:34:36.531275 139954941523776 pyconfig.py:432] Config param use_chat_template: False
I0420 08:34:36.531289 139954941523776 pyconfig.py:432] Config param use_chunked_prefill: False
I0420 08:34:36.531305 139954941523776 pyconfig.py:432] Config param use_custom_sort_vjp: True
I0420 08:34:36.531318 139954941523776 pyconfig.py:432] Config param use_dpo: False
I0420 08:34:36.531346 139954941523776 pyconfig.py:432] Config param use_gather_mosaic_kernel: False
I0420 08:34:36.531368 139954941523776 pyconfig.py:432] Config param use_grpo: True
I0420 08:34:36.531383 139954941523776 pyconfig.py:432] Config param use_indexer: False
I0420 08:34:36.531397 139954941523776 pyconfig.py:432] Config param use_iota_embed: True
I0420 08:34:36.531411 139954941523776 pyconfig.py:432] Config param use_jax_splash: False
I0420 08:34:36.531425 139954941523776 pyconfig.py:432] Config param use_max_logit_estimate: -1
I0420 08:34:36.531439 139954941523776 pyconfig.py:432] Config param use_mrope: False
I0420 08:34:36.531453 139954941523776 pyconfig.py:432] Config param use_multimodal: False
I0420 08:34:36.531467 139954941523776 pyconfig.py:432] Config param use_pathways: True
I0420 08:34:36.531482 139954941523776 pyconfig.py:432] Config param use_post_attn_norm: False
I0420 08:34:36.531496 139954941523776 pyconfig.py:432] Config param use_post_ffw_norm: False
I0420 08:34:36.531511 139954941523776 pyconfig.py:432] Config param use_qk_clip: False
I0420 08:34:36.531525 139954941523776 pyconfig.py:432] Config param use_qk_norm: False
I0420 08:34:36.531543 139954941523776 pyconfig.py:432] Config param use_qk_norm_in_gdn: True
I0420 08:34:36.531557 139954941523776 pyconfig.py:432] Config param use_qwix_quantization: False
I0420 08:34:36.531573 139954941523776 pyconfig.py:432] Config param use_ragged_attention: False
I0420 08:34:36.531587 139954941523776 pyconfig.py:432] Config param use_random_routing: False
I0420 08:34:36.531602 139954941523776 pyconfig.py:432] Config param use_replicator_service: False
I0420 08:34:36.531615 139954941523776 pyconfig.py:432] Config param use_ring_of_experts: False
I0420 08:34:36.531630 139954941523776 pyconfig.py:432] Config param use_sft: False
I0420 08:34:36.531644 139954941523776 pyconfig.py:432] Config param use_splash_scheduler: False
I0420 08:34:36.531660 139954941523776 pyconfig.py:432] Config param use_tokamax_gmm: False
I0420 08:34:36.531674 139954941523776 pyconfig.py:432] Config param use_tokamax_splash: False
I0420 08:34:36.531688 139954941523776 pyconfig.py:432] Config param use_truncation: True
I0420 08:34:36.531702 139954941523776 pyconfig.py:432] Config param use_tunix_gradient_accumulation: False
I0420 08:34:36.531717 139954941523776 pyconfig.py:432] Config param use_untrainable_positional_embedding: False
I0420 08:34:36.531731 139954941523776 pyconfig.py:432] Config param use_vertex_tensorboard: False
I0420 08:34:36.531746 139954941523776 pyconfig.py:432] Config param using_pipeline_parallelism: False
I0420 08:34:36.531760 139954941523776 pyconfig.py:432] Config param v_head_dim: 128
I0420 08:34:36.531775 139954941523776 pyconfig.py:432] Config param v_norm_with_scale: True
I0420 08:34:36.531789 139954941523776 pyconfig.py:432] Config param value_proj: RematLocation.REMAT
I0420 08:34:36.531807 139954941523776 pyconfig.py:432] Config param vertex_tensorboard_project: 
I0420 08:34:36.531821 139954941523776 pyconfig.py:432] Config param vertex_tensorboard_region: 
I0420 08:34:36.531836 139954941523776 pyconfig.py:432] Config param video_path: 
I0420 08:34:36.531850 139954941523776 pyconfig.py:432] Config param video_placeholder: <|video|>
I0420 08:34:36.531865 139954941523776 pyconfig.py:432] Config param vision_output_dim_for_vit: 4096
I0420 08:34:36.531879 139954941523776 pyconfig.py:432] Config param vision_output_length: -1
I0420 08:34:36.531895 139954941523776 pyconfig.py:432] Config param vllm_additional_config: {}
I0420 08:34:36.531909 139954941523776 pyconfig.py:432] Config param vllm_hf_config_path: 
I0420 08:34:36.531923 139954941523776 pyconfig.py:432] Config param vllm_hf_overrides: {}
I0420 08:34:36.531939 139954941523776 pyconfig.py:432] Config param vocab_size: 32000
I0420 08:34:36.531953 139954941523776 pyconfig.py:432] Config param warmup_steps_fraction: 0.1
I0420 08:34:36.531969 139954941523776 pyconfig.py:432] Config param weight_dtype: float32
I0420 08:34:36.531993 139954941523776 pyconfig.py:432] Config param weight_quantization_calibration_method: absmax
I0420 08:34:36.532007 139954941523776 pyconfig.py:432] Config param wi_tile_dlhs_batch_seq: 512
I0420 08:34:36.532022 139954941523776 pyconfig.py:432] Config param wi_tile_dlhs_embed_dim: 1024
I0420 08:34:36.532036 139954941523776 pyconfig.py:432] Config param wi_tile_dlhs_mlp_dim: 1024
I0420 08:34:36.532051 139954941523776 pyconfig.py:432] Config param wi_tile_drhs_batch_seq: 512
I0420 08:34:36.532065 139954941523776 pyconfig.py:432] Config param wi_tile_drhs_embed_dim: 1024
I0420 08:34:36.532080 139954941523776 pyconfig.py:432] Config param wi_tile_drhs_mlp_dim: 1024
I0420 08:34:36.532094 139954941523776 pyconfig.py:432] Config param wi_tile_fwd_batch_seq: 512
I0420 08:34:36.532109 139954941523776 pyconfig.py:432] Config param wi_tile_fwd_embed_dim: 1024
I0420 08:34:36.532123 139954941523776 pyconfig.py:432] Config param wi_tile_fwd_mlp_dim: 1024
I0420 08:34:36.532138 139954941523776 pyconfig.py:432] Config param wo_tile_dlhs_batch_seq: 512
I0420 08:34:36.532152 139954941523776 pyconfig.py:432] Config param wo_tile_dlhs_embed_dim: 1024
I0420 08:34:36.532167 139954941523776 pyconfig.py:432] Config param wo_tile_dlhs_mlp_dim: 1024
I0420 08:34:36.532183 139954941523776 pyconfig.py:432] Config param wo_tile_drhs_batch_seq: 512
I0420 08:34:36.532198 139954941523776 pyconfig.py:432] Config param wo_tile_drhs_embed_dim: 1024
I0420 08:34:36.532212 139954941523776 pyconfig.py:432] Config param wo_tile_drhs_mlp_dim: 1024
I0420 08:34:36.532227 139954941523776 pyconfig.py:432] Config param wo_tile_fwd_batch_seq: 512
I0420 08:34:36.532241 139954941523776 pyconfig.py:432] Config param wo_tile_fwd_embed_dim: 1024
I0420 08:34:36.532256 139954941523776 pyconfig.py:432] Config param wo_tile_fwd_mlp_dim: 1024
I0420 08:34:36.532270 139954941523776 pyconfig.py:432] Config param wsd_decay_steps_fraction: 0.1
I0420 08:34:36.532286 139954941523776 pyconfig.py:432] Config param wsd_decay_style: WsdDecayStyle.LINEAR
I0420 08:34:36.532301 139954941523776 pyconfig.py:432] Config param xprof_e2e_enable_fw_power_level_event: False
I0420 08:34:36.532318 139954941523776 pyconfig.py:432] Config param xprof_e2e_enable_fw_thermal_event: False
I0420 08:34:36.532354 139954941523776 pyconfig.py:432] Config param xprof_e2e_enable_fw_throttle_event: False
I0420 08:34:36.532380 139954941523776 pyconfig.py:432] Config param xprof_tpu_power_trace_level: 0
I0420 08:34:36.532397 139954941523776 pyconfig.py:432] Config param z_loss_multiplier: 0.0
I0420 08:34:36.532734 139954941523776 tokenizer.py:245] Tokenizer path: meta-llama/Llama-2-7b-chat-hf
I0420 08:34:36.532773 139954941523776 tokenizer.py:224] Loading HF tokenizer: meta-llama/Llama-2-7b-chat-hf
I0420 08:34:40.476153 139954941523776 _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`.
I0420 08:34:40.479247 139954941523776 maxtext_utils.py:1718] Num_devices: 32, shape (1, 4, 1, 8, 1, 1, 1, 1, 1, 1, 1, 1, 1)
I0420 08:34:40.479382 139954941523776 train_distill.py:584] Applying logical axis rules for model initialization and training...
I0420 08:34:40.479457 139954941523776 train_distill.py:588] Loading Student from ...
I0420 08:34:40.479487 139954941523776 train_distill.py:169] --- Student Configuration ---
I0420 08:34:40.479510 139954941523776 train_distill.py:170]   Model Name:      gpt3-52k
I0420 08:34:40.479530 139954941523776 train_distill.py:171]   Dimensions:      1 Layers, 16 Emb Dim, 8 Head Dim
I0420 08:34:40.479547 139954941523776 train_distill.py:174]   Attention Heads: 2 Query, 2 KV
I0420 08:34:40.479566 139954941523776 train_distill.py:175]   Vocab Size:      32000
I0420 08:34:40.479583 139954941523776 train_distill.py:176]   Checkpoint:      
I0420 08:34:40.479601 139954941523776 train_distill.py:460] Initializing model: gpt3-52k...
I0420 08:34:41.756906 139954941523776 train_distill.py:602] Loading Teacher from gs://lance-maxtext/pt_seed_ckpts/pt_seed_ckpts/pt_seed_ckpt_gpt352k_v32k_linen/checkpoints/4/items...
I0420 08:34:41.757016 139954941523776 train_distill.py:169] --- Teacher Configuration ---
I0420 08:34:41.757044 139954941523776 train_distill.py:170]   Model Name:      gpt3-52k
I0420 08:34:41.757067 139954941523776 train_distill.py:171]   Dimensions:      1 Layers, 16 Emb Dim, 8 Head Dim
I0420 08:34:41.757088 139954941523776 train_distill.py:174]   Attention Heads: 2 Query, 2 KV
I0420 08:34:41.757108 139954941523776 train_distill.py:175]   Vocab Size:      32000
I0420 08:34:41.757125 139954941523776 train_distill.py:176]   Checkpoint:      gs://lance-maxtext/pt_seed_ckpts/pt_seed_ckpts/pt_seed_ckpt_gpt352k_v32k_linen/checkpoints/4/items
I0420 08:34:41.757144 139954941523776 train_distill.py:460] Initializing model: gpt3-52k...
I0420 08:34:42.904348 139954941523776 pytree_checkpoint_handler.py:577] save_device_host_concurrent_bytes=None
I0420 08:34:42.904791 139954941523776 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 0x7f4914f2c4a0>, enable_pinned_host_transfer=False, save_concurrent_bytes: 96000000000 (89.4 GiB), restore_concurrent_bytes: 96000000000 (89.4 GiB)
I0420 08:34:42.904851 139954941523776 abstract_checkpointer.py:35] orbax-checkpoint version: 0.11.28
W0420 08:34:43.452655 139954941523776 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
I0420 08:34:44.421083    2141 google_auth_provider.cc:181] Running on GCE, using service account 562977990677-compute@developer.gserviceaccount.com
I0420 08:34:45.556076 139954941523776 checkpointer.py:304] Restoring checkpoint from gs://lance-maxtext/pt_seed_ckpts/pt_seed_ckpts/pt_seed_ckpt_gpt352k_v32k_linen/checkpoints/4/items.
W0420 08:34:48.037318 139954941523776 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.
I0420 08:34:48.037725 139954941523776 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
I0420 08:34:48.347036 139954941523776 checkpointer.py:318] Finished restoring checkpoint in 3.17 seconds from gs://lance-maxtext/pt_seed_ckpts/pt_seed_ckpts/pt_seed_ckpt_gpt352k_v32k_linen/checkpoints/4/items.
I0420 08:34:49.039439 139954941523776 train_distill.py:628] Initializing Data Iterators via MaxText pipeline...
I0420 08:34:49.102778 139954941523776 config.py:112] TensorFlow version 2.20.0 available.
I0420 08:34:49.103261 139954941523776 config.py:125] JAX version 0.8.3 available.
E0420 08:34:51.159690 139954941523776 packing.py:209] PackAndBatchOperation is deprecated. Please use lazy_dataset.FirstFitPackIterDataset instead.
I0420 08:34:51.159911 139954941523776 data_loader.py:408] Adding CopyNumPyArrayToSharedMemory MapTransform.
I0420 08:34:51.162970 139954941523776 train_distill.py:405] Input Pipeline Checkpointing: DISABLED
I0420 08:34:51.163036 139954941523776 train_distill.py:409] Reason: Iterator 'MultiHostDataLoadIterator' is not recognized as Grain (dataset_type='DatasetType.HF', has_save=False)
I0420 08:34:51.163099 139954941523776 pytree_checkpoint_handler.py:577] save_device_host_concurrent_bytes=None
I0420 08:34:51.163176 139954941523776 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 0x7f4914f2c4a0>, enable_pinned_host_transfer=False, save_concurrent_bytes: 96000000000 (89.4 GiB), restore_concurrent_bytes: 96000000000 (89.4 GiB)
I0420 08:34:51.163218 139954941523776 pytree_checkpoint_handler.py:577] save_device_host_concurrent_bytes=None
I0420 08:34:51.163249 139954941523776 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 0x7f4914f2c4a0>, enable_pinned_host_transfer=False, save_concurrent_bytes: 96000000000 (89.4 GiB), restore_concurrent_bytes: 96000000000 (89.4 GiB)
I0420 08:34:51.163292 139954941523776 checkpoint_manager.py:702] [process=4][thread=MainThread] CheckpointManager init: checkpointers=None, item_names=None, item_handlers={'model_params': <orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeCheckpointHandler object at 0x7f31229a1190>, 'optimizer_state': <orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeCheckpointHandler object at 0x7f31228cf830>, 'custom_metadata': <orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonCheckpointHandler object at 0x7f31228cf770>}, handler_registry=None
I0420 08:34:51.163514 139954941523776 composite_checkpoint_handler.py:237] Deferred registration for item: "model_params". Adding handler `<orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeCheckpointHandler object at 0x7f31229a1190>` 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`.
I0420 08:34:51.163558 139954941523776 composite_checkpoint_handler.py:237] Deferred registration for item: "optimizer_state". Adding handler `<orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeCheckpointHandler object at 0x7f31228cf830>` 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`.
I0420 08:34:51.163584 139954941523776 composite_checkpoint_handler.py:237] Deferred registration for item: "custom_metadata". Adding handler `<orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonCheckpointHandler object at 0x7f31228cf770>` 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`.
I0420 08:34:51.163607 139954941523776 composite_checkpoint_handler.py:237] Deferred registration for item: "metrics". Adding handler `<orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonCheckpointHandler object at 0x7f3123e308c0>` 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`.
I0420 08:34:51.163638 139954941523776 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 0x7f31229a1190>, ('model_params', <class 'orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeRestoreArgs'>): <orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeCheckpointHandler object at 0x7f31229a1190>, ('optimizer_state', <class 'orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeSaveArgs'>): <orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeCheckpointHandler object at 0x7f31228cf830>, ('optimizer_state', <class 'orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeRestoreArgs'>): <orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeCheckpointHandler object at 0x7f31228cf830>, ('custom_metadata', <class 'orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonSaveArgs'>): <orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonCheckpointHandler object at 0x7f31228cf770>, ('custom_metadata', <class 'orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonRestoreArgs'>): <orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonCheckpointHandler object at 0x7f31228cf770>, ('metrics', <class 'orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonSaveArgs'>): <orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonCheckpointHandler object at 0x7f3123e308c0>, ('metrics', <class 'orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonRestoreArgs'>): <orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonCheckpointHandler object at 0x7f3123e308c0>}).
I0420 08:34:51.164081 139954941523776 async_checkpointer.py:177] [process=4][thread=MainThread] Using barrier_sync_fn: <function get_barrier_sync_fn.<locals>._fn at 0x7f31227807c0> timeout: 600 secs and primary_host=0 for async checkpoint writes
I0420 08:34:54.394271 139954941523776 checkpoint_manager.py:1788] Found 0 checkpoint steps in gs://lance-maxtext/pt_ckpt_xpk_feat_nnx_trainstate_and_training_loop_20260420_081542/pt_distill_nnx_xpk_feat_nnx_trainstate_and_training_loop_20260420_081542_07_distill_smoke/checkpoints
I0420 08:34:54.396296 139954941523776 checkpoint_manager.py:921] [process=4][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_20260420_081542/pt_distill_nnx_xpk_feat_nnx_trainstate_and_training_loop_20260420_081542_07_distill_smoke/checkpoints: <orbax.checkpoint.checkpoint_manager.CheckpointManager object at 0x7f31228cf740>
I0420 08:34:54.396427 139954941523776 pytree_checkpoint_handler.py:577] save_device_host_concurrent_bytes=None
I0420 08:34:54.396492 139954941523776 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 0x7f4914f2c4a0>, enable_pinned_host_transfer=False, save_concurrent_bytes: 96000000000 (89.4 GiB), restore_concurrent_bytes: 96000000000 (89.4 GiB)
I0420 08:34:54.396528 139954941523776 pytree_checkpoint_handler.py:577] save_device_host_concurrent_bytes=None
I0420 08:34:54.396559 139954941523776 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 0x7f4914f2c4a0>, enable_pinned_host_transfer=False, save_concurrent_bytes: 96000000000 (89.4 GiB), restore_concurrent_bytes: 96000000000 (89.4 GiB)
I0420 08:34:54.396593 139954941523776 checkpoint_manager.py:1983] [process=4][thread=MainThread][wait_until_finished] No Save Finalize thread to wait for. Returning.
I0420 08:34:54.396647 139954941523776 checkpoint.py:459] Closing _NonBlockingMetadataStore(enable_write=True, _write_lock=<locked _thread.RLock object owner=139954941523776 count=1 at 0x7f3274d9df80>, _store_impl=<orbax.checkpoint._src.metadata.checkpoint._MetadataStoreImpl object at 0x7f31228cf530>, _single_thread_executor=<concurrent.futures.thread.ThreadPoolExecutor object at 0x7f31228cf500>, _write_futures=[])
I0420 08:34:54.396981 139954941523776 checkpoint.py:459] Closing _NonBlockingMetadataStore(enable_write=True, _write_lock=<locked _thread.RLock object owner=139954941523776 count=1 at 0x7f3274d9df80>, _store_impl=<orbax.checkpoint._src.metadata.checkpoint._MetadataStoreImpl object at 0x7f31228cf530>, _single_thread_executor=<concurrent.futures.thread.ThreadPoolExecutor object at 0x7f31228cf500>, _write_futures=[])
I0420 08:34:54.397007 139954941523776 checkpoint.py:459] Closing _NonBlockingMetadataStore(enable_write=True, _write_lock=<locked _thread.RLock object owner=139954941523776 count=1 at 0x7f3274d9df80>, _store_impl=<orbax.checkpoint._src.metadata.checkpoint._MetadataStoreImpl object at 0x7f31228cf530>, _single_thread_executor=<concurrent.futures.thread.ThreadPoolExecutor object at 0x7f31228cf500>, _write_futures=[])
I0420 08:34:54.397037 139954941523776 checkpoint_manager.py:702] [process=4][thread=MainThread] CheckpointManager init: checkpointers=None, item_names=None, item_handlers={'model_params': <orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeCheckpointHandler object at 0x7f31228cf710>, 'optimizer_state': <orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeCheckpointHandler object at 0x7f31228ced50>, 'custom_metadata': <orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonCheckpointHandler object at 0x7f31228cd760>, 'iter': <maxtext.common.checkpointing.GrainCheckpointHandler object at 0x7f31228cd460>}, handler_registry=None
I0420 08:34:54.397131 139954941523776 composite_checkpoint_handler.py:237] Deferred registration for item: "model_params". Adding handler `<orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeCheckpointHandler object at 0x7f31228cf710>` 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`.
I0420 08:34:54.397163 139954941523776 composite_checkpoint_handler.py:237] Deferred registration for item: "optimizer_state". Adding handler `<orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeCheckpointHandler object at 0x7f31228ced50>` 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`.
I0420 08:34:54.397186 139954941523776 composite_checkpoint_handler.py:237] Deferred registration for item: "custom_metadata". Adding handler `<orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonCheckpointHandler object at 0x7f31228cd760>` 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`.
I0420 08:34:54.397212 139954941523776 composite_checkpoint_handler.py:237] Deferred registration for item: "iter". Adding handler `<maxtext.common.checkpointing.GrainCheckpointHandler object at 0x7f31228cd460>` for item "iter" and save args `<class 'maxtext.common.checkpointing.GrainCheckpointSave'>` and restore args `<class 'maxtext.common.checkpointing.GrainCheckpointRestore'>` to `_handler_registry`.
I0420 08:34:54.397235 139954941523776 composite_checkpoint_handler.py:237] Deferred registration for item: "metrics". Adding handler `<orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonCheckpointHandler object at 0x7f31228cd5e0>` 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`.
I0420 08:34:54.397260 139954941523776 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 0x7f31228cf710>, ('model_params', <class 'orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeRestoreArgs'>): <orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeCheckpointHandler object at 0x7f31228cf710>, ('optimizer_state', <class 'orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeSaveArgs'>): <orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeCheckpointHandler object at 0x7f31228ced50>, ('optimizer_state', <class 'orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeRestoreArgs'>): <orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeCheckpointHandler object at 0x7f31228ced50>, ('custom_metadata', <class 'orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonSaveArgs'>): <orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonCheckpointHandler object at 0x7f31228cd760>, ('custom_metadata', <class 'orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonRestoreArgs'>): <orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonCheckpointHandler object at 0x7f31228cd760>, ('iter', <class 'maxtext.common.checkpointing.GrainCheckpointSave'>): <maxtext.common.checkpointing.GrainCheckpointHandler object at 0x7f31228cd460>, ('iter', <class 'maxtext.common.checkpointing.GrainCheckpointRestore'>): <maxtext.common.checkpointing.GrainCheckpointHandler object at 0x7f31228cd460>, ('metrics', <class 'orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonSaveArgs'>): <orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonCheckpointHandler object at 0x7f31228cd5e0>, ('metrics', <class 'orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonRestoreArgs'>): <orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonCheckpointHandler object at 0x7f31228cd5e0>}).
I0420 08:34:54.397328 139954941523776 async_checkpointer.py:177] [process=4][thread=MainThread] Using barrier_sync_fn: <function get_barrier_sync_fn.<locals>._fn at 0x7f3122780900> timeout: 600 secs and primary_host=0 for async checkpoint writes
I0420 08:34:54.791211 139954941523776 checkpoint_manager.py:1788] Found 0 checkpoint steps in gs://lance-maxtext/pt_ckpt_xpk_feat_nnx_trainstate_and_training_loop_20260420_081542/pt_distill_nnx_xpk_feat_nnx_trainstate_and_training_loop_20260420_081542_07_distill_smoke/checkpoints
I0420 08:34:55.208937 139954941523776 checkpoint_manager.py:921] [process=4][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_20260420_081542/pt_distill_nnx_xpk_feat_nnx_trainstate_and_training_loop_20260420_081542_07_distill_smoke/checkpoints: <orbax.checkpoint.checkpoint_manager.CheckpointManager object at 0x7f32749685f0>
I0420 08:34:55.209170 139954941523776 train_distill.py:675] Starting Distillation Training...
I0420 08:34:55.209268 139954941523776 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))
I0420 08:34:55.331913 139954941523776 peft_trainer.py:600] Compiled train_step cache size: 0

Training:   0%|          | 0/5 [00:00<?, ?step/s]I0420 08:34:55.333629 139811075630848 grain_pool.py:367] Grain pool will use 1 processes.
I0420 08:34:55.360108 139811075630848 grain_pool.py:440] Grain pool will start child processes.
I0420 08:34:55.365237 139811075630848 grain_pool.py:448] Grain pool started all child processes.
2026-04-20 08:35:01.379597: 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)
I0420 08:35:04.564817 139954941523776 utils.py:86] Train loop finished in: 9.2324 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 749, 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 745, 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 677, 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=24, process_index=6, coords=(0,6,0), core_on_chip=0), TpuDevice(id=18, process_index=5, coords=(2,4,0), core_on_chip=0), TpuDevice(id=26, process_index=7, coords=(2,6,0), core_on_chip=0), TpuDevice(id=0, process_index=0, coords=(0,0,0), core_on_chip=0), TpuDevice(id=1, process_index=0, coords=(1,0,0), core_on_chip=0), TpuDevice(id=3, process_index=1, coords=(3,0,0), core_on_chip=0), TpuDevice(id=17, process_index=4, coords=(1,4,0), core_on_chip=0), TpuDevice(id=5, process_index=0, coords=(1,1,0), core_on_chip=0), TpuDevice(id=14, process_index=3, coords=(2,3,0), core_on_chip=0), TpuDevice(id=6, process_index=1, coords=(2,1,0), core_on_chip=0), TpuDevice(id=25, process_index=6, coords=(1,6,0), core_on_chip=0), TpuDevice(id=15, process_index=3, coords=(3,3,0), core_on_chip=0), TpuDevice(id=19, process_index=5, coords=(3,4,0), core_on_chip=0), TpuDevice(id=27, process_index=7, coords=(3,6,0), core_on_chip=0), TpuDevice(id=4, process_index=0, coords=(0,1,0), core_on_chip=0), TpuDevice(id=13, process_index=2, coords=(1,3,0), core_on_chip=0), TpuDevice(id=2, process_index=1, coords=(2,0,0), core_on_chip=0), TpuDevice(id=8, process_index=2, coords=(0,2,0), core_on_chip=0), TpuDevice(id=20, process_index=4, coords=(0,5,0), core_on_chip=0), TpuDevice(id=28, process_index=6, coords=(0,7,0), core_on_chip=0), TpuDevice(id=22, process_index=5, coords=(2,5,0), core_on_chip=0), TpuDevice(id=30, process_index=7, coords=(2,7,0), core_on_chip=0), TpuDevice(id=7, process_index=1, coords=(3,1,0), core_on_chip=0), TpuDevice(id=10, process_index=3, coords=(2,2,0), core_on_chip=0), TpuDevice(id=21, process_index=4, coords=(1,5,0), core_on_chip=0), TpuDevice(id=29, process_index=6, coords=(1,7,0), core_on_chip=0), TpuDevice(id=23, process_index=5, coords=(3,5,0), core_on_chip=0), TpuDevice(id=31, process_index=7, coords=(3,7,0), core_on_chip=0), TpuDevice(id=16, process_index=4, coords=(0,4,0), core_on_chip=0), TpuDevice(id=11, process_index=3, coords=(3,2,0), core_on_chip=0), TpuDevice(id=12, process_index=2, coords=(0,3,0), core_on_chip=0), TpuDevice(id=9, process_index=2, coords=(1,2,0), core_on_chip=0)}
I0420 08:35:04.915948 139811075630848 grain_pool.py:542] Grain pool is exiting.
I0420 08:35:04.916046 139811075630848 grain_pool.py:547] Shutting down multiprocessing system.
I0420 08:35:06.343433 139811075630848 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: Mon Apr 20 08:35:14 UTC 2026
EXIT_CODE=1